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 "InterCheckerAPI.h"
15#include "clang/Basic/Builtins.h"
16#include "clang/Basic/CharInfo.h"
17#include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.h"
18#include "clang/StaticAnalyzer/Core/BugReporter/BugType.h"
19#include "clang/StaticAnalyzer/Core/Checker.h"
20#include "clang/StaticAnalyzer/Core/CheckerManager.h"
21#include "clang/StaticAnalyzer/Core/PathSensitive/CallDescription.h"
22#include "clang/StaticAnalyzer/Core/PathSensitive/CallEvent.h"
23#include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h"
24#include "clang/StaticAnalyzer/Core/PathSensitive/DynamicExtent.h"
25#include "clang/StaticAnalyzer/Core/PathSensitive/ProgramStateTrait.h"
26#include "llvm/ADT/STLExtras.h"
27#include "llvm/ADT/SmallString.h"
28#include "llvm/ADT/StringExtras.h"
29#include "llvm/Support/raw_ostream.h"
30#include <functional>
31#include <optional>
32
33using namespace clang;
34using namespace ento;
35using namespace std::placeholders;
36
37namespace {
38struct AnyArgExpr {
39 const Expr *Expression;
40 unsigned ArgumentIndex;
41};
42struct SourceArgExpr : AnyArgExpr {};
43struct DestinationArgExpr : AnyArgExpr {};
44struct SizeArgExpr : AnyArgExpr {};
45
46using ErrorMessage = SmallString<128>;
47enum class AccessKind { write, read };
48
49static ErrorMessage createOutOfBoundErrorMsg(StringRef FunctionDescription,
50 AccessKind Access) {
51 ErrorMessage Message;
52 llvm::raw_svector_ostream Os(Message);
53
54 // Function classification like: Memory copy function
55 Os << toUppercase(c: FunctionDescription.front())
56 << &FunctionDescription.data()[1];
57
58 if (Access == AccessKind::write) {
59 Os << " overflows the destination buffer";
60 } else { // read access
61 Os << " accesses out-of-bound array element";
62 }
63
64 return Message;
65}
66
67enum class ConcatFnKind { none = 0, strcat = 1, strlcat = 2 };
68
69enum class CharKind { Regular = 0, Wide };
70constexpr CharKind CK_Regular = CharKind::Regular;
71constexpr CharKind CK_Wide = CharKind::Wide;
72
73static QualType getCharPtrType(ASTContext &Ctx, CharKind CK) {
74 return Ctx.getPointerType(CK == CharKind::Regular ? Ctx.CharTy
75 : Ctx.WideCharTy);
76}
77
78class CStringChecker : public Checker< eval::Call,
79 check::PreStmt<DeclStmt>,
80 check::LiveSymbols,
81 check::DeadSymbols,
82 check::RegionChanges
83 > {
84 mutable std::unique_ptr<BugType> BT_Null, BT_Bounds, BT_Overlap,
85 BT_NotCString, BT_AdditionOverflow, BT_UninitRead;
86
87 mutable const char *CurrentFunctionDescription = nullptr;
88
89public:
90 /// The filter is used to filter out the diagnostics which are not enabled by
91 /// the user.
92 struct CStringChecksFilter {
93 bool CheckCStringNullArg = false;
94 bool CheckCStringOutOfBounds = false;
95 bool CheckCStringBufferOverlap = false;
96 bool CheckCStringNotNullTerm = false;
97 bool CheckCStringUninitializedRead = false;
98
99 CheckerNameRef CheckNameCStringNullArg;
100 CheckerNameRef CheckNameCStringOutOfBounds;
101 CheckerNameRef CheckNameCStringBufferOverlap;
102 CheckerNameRef CheckNameCStringNotNullTerm;
103 CheckerNameRef CheckNameCStringUninitializedRead;
104 };
105
106 CStringChecksFilter Filter;
107
108 static void *getTag() { static int tag; return &tag; }
109
110 bool evalCall(const CallEvent &Call, CheckerContext &C) const;
111 void checkPreStmt(const DeclStmt *DS, CheckerContext &C) const;
112 void checkLiveSymbols(ProgramStateRef state, SymbolReaper &SR) const;
113 void checkDeadSymbols(SymbolReaper &SR, CheckerContext &C) const;
114
115 ProgramStateRef
116 checkRegionChanges(ProgramStateRef state,
117 const InvalidatedSymbols *,
118 ArrayRef<const MemRegion *> ExplicitRegions,
119 ArrayRef<const MemRegion *> Regions,
120 const LocationContext *LCtx,
121 const CallEvent *Call) const;
122
123 using FnCheck = std::function<void(const CStringChecker *, CheckerContext &,
124 const CallEvent &)>;
125
126 CallDescriptionMap<FnCheck> Callbacks = {
127 {{CDM::CLibraryMaybeHardened, {"memcpy"}, 3},
128 std::bind(f: &CStringChecker::evalMemcpy, args: _1, args: _2, args: _3, args: CK_Regular)},
129 {{CDM::CLibraryMaybeHardened, {"wmemcpy"}, 3},
130 std::bind(f: &CStringChecker::evalMemcpy, args: _1, args: _2, args: _3, args: CK_Wide)},
131 {{CDM::CLibraryMaybeHardened, {"mempcpy"}, 3},
132 std::bind(f: &CStringChecker::evalMempcpy, args: _1, args: _2, args: _3, args: CK_Regular)},
133 {{CDM::CLibraryMaybeHardened, {"wmempcpy"}, 3},
134 std::bind(f: &CStringChecker::evalMempcpy, args: _1, args: _2, args: _3, args: CK_Wide)},
135 {{CDM::CLibrary, {"memcmp"}, 3},
136 std::bind(f: &CStringChecker::evalMemcmp, args: _1, args: _2, args: _3, args: CK_Regular)},
137 {{CDM::CLibrary, {"wmemcmp"}, 3},
138 std::bind(f: &CStringChecker::evalMemcmp, args: _1, args: _2, args: _3, args: CK_Wide)},
139 {{CDM::CLibraryMaybeHardened, {"memmove"}, 3},
140 std::bind(f: &CStringChecker::evalMemmove, args: _1, args: _2, args: _3, args: CK_Regular)},
141 {{CDM::CLibraryMaybeHardened, {"wmemmove"}, 3},
142 std::bind(f: &CStringChecker::evalMemmove, args: _1, args: _2, args: _3, args: CK_Wide)},
143 {{CDM::CLibraryMaybeHardened, {"memset"}, 3},
144 &CStringChecker::evalMemset},
145 {{CDM::CLibrary, {"explicit_memset"}, 3}, &CStringChecker::evalMemset},
146 // FIXME: C23 introduces 'memset_explicit', maybe also model that
147 {{CDM::CLibraryMaybeHardened, {"strcpy"}, 2},
148 &CStringChecker::evalStrcpy},
149 {{CDM::CLibraryMaybeHardened, {"strncpy"}, 3},
150 &CStringChecker::evalStrncpy},
151 {{CDM::CLibraryMaybeHardened, {"stpcpy"}, 2},
152 &CStringChecker::evalStpcpy},
153 {{CDM::CLibraryMaybeHardened, {"strlcpy"}, 3},
154 &CStringChecker::evalStrlcpy},
155 {{CDM::CLibraryMaybeHardened, {"strcat"}, 2},
156 &CStringChecker::evalStrcat},
157 {{CDM::CLibraryMaybeHardened, {"strncat"}, 3},
158 &CStringChecker::evalStrncat},
159 {{CDM::CLibraryMaybeHardened, {"strlcat"}, 3},
160 &CStringChecker::evalStrlcat},
161 {{CDM::CLibraryMaybeHardened, {"strlen"}, 1},
162 &CStringChecker::evalstrLength},
163 {{CDM::CLibrary, {"wcslen"}, 1}, &CStringChecker::evalstrLength},
164 {{CDM::CLibraryMaybeHardened, {"strnlen"}, 2},
165 &CStringChecker::evalstrnLength},
166 {{CDM::CLibrary, {"wcsnlen"}, 2}, &CStringChecker::evalstrnLength},
167 {{CDM::CLibrary, {"strcmp"}, 2}, &CStringChecker::evalStrcmp},
168 {{CDM::CLibrary, {"strncmp"}, 3}, &CStringChecker::evalStrncmp},
169 {{CDM::CLibrary, {"strcasecmp"}, 2}, &CStringChecker::evalStrcasecmp},
170 {{CDM::CLibrary, {"strncasecmp"}, 3}, &CStringChecker::evalStrncasecmp},
171 {{CDM::CLibrary, {"strsep"}, 2}, &CStringChecker::evalStrsep},
172 {{CDM::CLibrary, {"bcopy"}, 3}, &CStringChecker::evalBcopy},
173 {{CDM::CLibrary, {"bcmp"}, 3},
174 std::bind(f: &CStringChecker::evalMemcmp, args: _1, args: _2, args: _3, args: CK_Regular)},
175 {{CDM::CLibrary, {"bzero"}, 2}, &CStringChecker::evalBzero},
176 {{CDM::CLibraryMaybeHardened, {"explicit_bzero"}, 2},
177 &CStringChecker::evalBzero},
178
179 // When recognizing calls to the following variadic functions, we accept
180 // any number of arguments in the call (std::nullopt = accept any
181 // number), but check that in the declaration there are 2 and 3
182 // parameters respectively. (Note that the parameter count does not
183 // include the "...". Calls where the number of arguments is too small
184 // will be discarded by the callback.)
185 {{CDM::CLibraryMaybeHardened, {"sprintf"}, std::nullopt, 2},
186 &CStringChecker::evalSprintf},
187 {{CDM::CLibraryMaybeHardened, {"snprintf"}, std::nullopt, 3},
188 &CStringChecker::evalSnprintf},
189 };
190
191 // These require a bit of special handling.
192 CallDescription StdCopy{{"std", "copy"}, 3},
193 StdCopyBackward{{"std", "copy_backward"}, 3};
194
195 FnCheck identifyCall(const CallEvent &Call, CheckerContext &C) const;
196 void evalMemcpy(CheckerContext &C, const CallEvent &Call, CharKind CK) const;
197 void evalMempcpy(CheckerContext &C, const CallEvent &Call, CharKind CK) const;
198 void evalMemmove(CheckerContext &C, const CallEvent &Call, CharKind CK) const;
199 void evalBcopy(CheckerContext &C, const CallEvent &Call) const;
200 void evalCopyCommon(CheckerContext &C, const CallEvent &Call,
201 ProgramStateRef state, SizeArgExpr Size,
202 DestinationArgExpr Dest, SourceArgExpr Source,
203 bool Restricted, bool IsMempcpy, CharKind CK) const;
204
205 void evalMemcmp(CheckerContext &C, const CallEvent &Call, CharKind CK) const;
206
207 void evalstrLength(CheckerContext &C, const CallEvent &Call) const;
208 void evalstrnLength(CheckerContext &C, const CallEvent &Call) const;
209 void evalstrLengthCommon(CheckerContext &C, const CallEvent &Call,
210 bool IsStrnlen = false) const;
211
212 void evalStrcpy(CheckerContext &C, const CallEvent &Call) const;
213 void evalStrncpy(CheckerContext &C, const CallEvent &Call) const;
214 void evalStpcpy(CheckerContext &C, const CallEvent &Call) const;
215 void evalStrlcpy(CheckerContext &C, const CallEvent &Call) const;
216 void evalStrcpyCommon(CheckerContext &C, const CallEvent &Call,
217 bool ReturnEnd, bool IsBounded, ConcatFnKind appendK,
218 bool returnPtr = true) const;
219
220 void evalStrcat(CheckerContext &C, const CallEvent &Call) const;
221 void evalStrncat(CheckerContext &C, const CallEvent &Call) const;
222 void evalStrlcat(CheckerContext &C, const CallEvent &Call) const;
223
224 void evalStrcmp(CheckerContext &C, const CallEvent &Call) const;
225 void evalStrncmp(CheckerContext &C, const CallEvent &Call) const;
226 void evalStrcasecmp(CheckerContext &C, const CallEvent &Call) const;
227 void evalStrncasecmp(CheckerContext &C, const CallEvent &Call) const;
228 void evalStrcmpCommon(CheckerContext &C, const CallEvent &Call,
229 bool IsBounded = false, bool IgnoreCase = false) const;
230
231 void evalStrsep(CheckerContext &C, const CallEvent &Call) const;
232
233 void evalStdCopy(CheckerContext &C, const CallEvent &Call) const;
234 void evalStdCopyBackward(CheckerContext &C, const CallEvent &Call) const;
235 void evalStdCopyCommon(CheckerContext &C, const CallEvent &Call) const;
236 void evalMemset(CheckerContext &C, const CallEvent &Call) const;
237 void evalBzero(CheckerContext &C, const CallEvent &Call) const;
238
239 void evalSprintf(CheckerContext &C, const CallEvent &Call) const;
240 void evalSnprintf(CheckerContext &C, const CallEvent &Call) const;
241 void evalSprintfCommon(CheckerContext &C, const CallEvent &Call,
242 bool IsBounded) const;
243
244 // Utility methods
245 std::pair<ProgramStateRef , ProgramStateRef >
246 static assumeZero(CheckerContext &C,
247 ProgramStateRef state, SVal V, QualType Ty);
248
249 static ProgramStateRef setCStringLength(ProgramStateRef state,
250 const MemRegion *MR,
251 SVal strLength);
252 static SVal getCStringLengthForRegion(CheckerContext &C,
253 ProgramStateRef &state,
254 const Expr *Ex,
255 const MemRegion *MR,
256 bool hypothetical);
257 SVal getCStringLength(CheckerContext &C,
258 ProgramStateRef &state,
259 const Expr *Ex,
260 SVal Buf,
261 bool hypothetical = false) const;
262
263 const StringLiteral *getCStringLiteral(CheckerContext &C,
264 ProgramStateRef &state,
265 const Expr *expr,
266 SVal val) const;
267
268 /// Invalidate the destination buffer determined by characters copied.
269 static ProgramStateRef
270 invalidateDestinationBufferBySize(CheckerContext &C, ProgramStateRef S,
271 const Expr *BufE, SVal BufV, SVal SizeV,
272 QualType SizeTy);
273
274 /// Operation never overflows, do not invalidate the super region.
275 static ProgramStateRef invalidateDestinationBufferNeverOverflows(
276 CheckerContext &C, ProgramStateRef S, const Expr *BufE, SVal BufV);
277
278 /// We do not know whether the operation can overflow (e.g. size is unknown),
279 /// invalidate the super region and escape related pointers.
280 static ProgramStateRef invalidateDestinationBufferAlwaysEscapeSuperRegion(
281 CheckerContext &C, ProgramStateRef S, const Expr *BufE, SVal BufV);
282
283 /// Invalidate the source buffer for escaping pointers.
284 static ProgramStateRef invalidateSourceBuffer(CheckerContext &C,
285 ProgramStateRef S,
286 const Expr *BufE, SVal BufV);
287
288 /// @param InvalidationTraitOperations Determine how to invlidate the
289 /// MemRegion by setting the invalidation traits. Return true to cause pointer
290 /// escape, or false otherwise.
291 static ProgramStateRef invalidateBufferAux(
292 CheckerContext &C, ProgramStateRef State, const Expr *Ex, SVal V,
293 llvm::function_ref<bool(RegionAndSymbolInvalidationTraits &,
294 const MemRegion *)>
295 InvalidationTraitOperations);
296
297 static bool SummarizeRegion(raw_ostream &os, ASTContext &Ctx,
298 const MemRegion *MR);
299
300 static bool memsetAux(const Expr *DstBuffer, SVal CharE,
301 const Expr *Size, CheckerContext &C,
302 ProgramStateRef &State);
303
304 // Re-usable checks
305 ProgramStateRef checkNonNull(CheckerContext &C, ProgramStateRef State,
306 AnyArgExpr Arg, SVal l) const;
307 ProgramStateRef CheckLocation(CheckerContext &C, ProgramStateRef state,
308 AnyArgExpr Buffer, SVal Element,
309 AccessKind Access,
310 CharKind CK = CharKind::Regular) const;
311 ProgramStateRef CheckBufferAccess(CheckerContext &C, ProgramStateRef State,
312 AnyArgExpr Buffer, SizeArgExpr Size,
313 AccessKind Access,
314 CharKind CK = CharKind::Regular) const;
315 ProgramStateRef CheckOverlap(CheckerContext &C, ProgramStateRef state,
316 SizeArgExpr Size, AnyArgExpr First,
317 AnyArgExpr Second,
318 CharKind CK = CharKind::Regular) const;
319 void emitOverlapBug(CheckerContext &C,
320 ProgramStateRef state,
321 const Stmt *First,
322 const Stmt *Second) const;
323
324 void emitNullArgBug(CheckerContext &C, ProgramStateRef State, const Stmt *S,
325 StringRef WarningMsg) const;
326 void emitOutOfBoundsBug(CheckerContext &C, ProgramStateRef State,
327 const Stmt *S, StringRef WarningMsg) const;
328 void emitNotCStringBug(CheckerContext &C, ProgramStateRef State,
329 const Stmt *S, StringRef WarningMsg) const;
330 void emitAdditionOverflowBug(CheckerContext &C, ProgramStateRef State) const;
331 void emitUninitializedReadBug(CheckerContext &C, ProgramStateRef State,
332 const Expr *E) const;
333 ProgramStateRef checkAdditionOverflow(CheckerContext &C,
334 ProgramStateRef state,
335 NonLoc left,
336 NonLoc right) const;
337
338 // Return true if the destination buffer of the copy function may be in bound.
339 // Expects SVal of Size to be positive and unsigned.
340 // Expects SVal of FirstBuf to be a FieldRegion.
341 static bool isFirstBufInBound(CheckerContext &C, ProgramStateRef State,
342 SVal BufVal, QualType BufTy, SVal LengthVal,
343 QualType LengthTy);
344};
345
346} //end anonymous namespace
347
348REGISTER_MAP_WITH_PROGRAMSTATE(CStringLength, const MemRegion *, SVal)
349
350//===----------------------------------------------------------------------===//
351// Individual checks and utility methods.
352//===----------------------------------------------------------------------===//
353
354std::pair<ProgramStateRef , ProgramStateRef >
355CStringChecker::assumeZero(CheckerContext &C, ProgramStateRef state, SVal V,
356 QualType Ty) {
357 std::optional<DefinedSVal> val = V.getAs<DefinedSVal>();
358 if (!val)
359 return std::pair<ProgramStateRef , ProgramStateRef >(state, state);
360
361 SValBuilder &svalBuilder = C.getSValBuilder();
362 DefinedOrUnknownSVal zero = svalBuilder.makeZeroVal(type: Ty);
363 return state->assume(Cond: svalBuilder.evalEQ(state, lhs: *val, rhs: zero));
364}
365
366ProgramStateRef CStringChecker::checkNonNull(CheckerContext &C,
367 ProgramStateRef State,
368 AnyArgExpr Arg, SVal l) const {
369 // If a previous check has failed, propagate the failure.
370 if (!State)
371 return nullptr;
372
373 ProgramStateRef stateNull, stateNonNull;
374 std::tie(args&: stateNull, args&: stateNonNull) =
375 assumeZero(C, state: State, V: l, Ty: Arg.Expression->getType());
376
377 if (stateNull && !stateNonNull) {
378 if (Filter.CheckCStringNullArg) {
379 SmallString<80> buf;
380 llvm::raw_svector_ostream OS(buf);
381 assert(CurrentFunctionDescription);
382 OS << "Null pointer passed as " << (Arg.ArgumentIndex + 1)
383 << llvm::getOrdinalSuffix(Val: Arg.ArgumentIndex + 1) << " argument to "
384 << CurrentFunctionDescription;
385
386 emitNullArgBug(C, stateNull, Arg.Expression, OS.str());
387 }
388 return nullptr;
389 }
390
391 // From here on, assume that the value is non-null.
392 assert(stateNonNull);
393 return stateNonNull;
394}
395
396// FIXME: This was originally copied from ArrayBoundChecker.cpp. Refactor?
397ProgramStateRef CStringChecker::CheckLocation(CheckerContext &C,
398 ProgramStateRef state,
399 AnyArgExpr Buffer, SVal Element,
400 AccessKind Access,
401 CharKind CK) const {
402
403 // If a previous check has failed, propagate the failure.
404 if (!state)
405 return nullptr;
406
407 // Check for out of bound array element access.
408 const MemRegion *R = Element.getAsRegion();
409 if (!R)
410 return state;
411
412 const auto *ER = dyn_cast<ElementRegion>(Val: R);
413 if (!ER)
414 return state;
415
416 SValBuilder &svalBuilder = C.getSValBuilder();
417 ASTContext &Ctx = svalBuilder.getContext();
418
419 // Get the index of the accessed element.
420 NonLoc Idx = ER->getIndex();
421
422 if (CK == CharKind::Regular) {
423 if (ER->getValueType() != Ctx.CharTy)
424 return state;
425 } else {
426 if (ER->getValueType() != Ctx.WideCharTy)
427 return state;
428
429 QualType SizeTy = Ctx.getSizeType();
430 NonLoc WideSize =
431 svalBuilder
432 .makeIntVal(Ctx.getTypeSizeInChars(Ctx.WideCharTy).getQuantity(),
433 SizeTy)
434 .castAs<NonLoc>();
435 SVal Offset = svalBuilder.evalBinOpNN(state, op: BO_Mul, lhs: Idx, rhs: WideSize, resultTy: SizeTy);
436 if (Offset.isUnknown())
437 return state;
438 Idx = Offset.castAs<NonLoc>();
439 }
440
441 // Get the size of the array.
442 const auto *superReg = cast<SubRegion>(ER->getSuperRegion());
443 DefinedOrUnknownSVal Size =
444 getDynamicExtent(state, superReg, C.getSValBuilder());
445
446 ProgramStateRef StInBound, StOutBound;
447 std::tie(args&: StInBound, args&: StOutBound) = state->assumeInBoundDual(idx: Idx, upperBound: Size);
448 if (StOutBound && !StInBound) {
449 // These checks are either enabled by the CString out-of-bounds checker
450 // explicitly or implicitly by the Malloc checker.
451 // In the latter case we only do modeling but do not emit warning.
452 if (!Filter.CheckCStringOutOfBounds)
453 return nullptr;
454
455 // Emit a bug report.
456 ErrorMessage Message =
457 createOutOfBoundErrorMsg(FunctionDescription: CurrentFunctionDescription, Access);
458 emitOutOfBoundsBug(C, StOutBound, Buffer.Expression, Message);
459 return nullptr;
460 }
461
462 // Ensure that we wouldn't read uninitialized value.
463 if (Access == AccessKind::read) {
464 if (Filter.CheckCStringUninitializedRead &&
465 StInBound->getSVal(ER).isUndef()) {
466 emitUninitializedReadBug(C, State: StInBound, E: Buffer.Expression);
467 return nullptr;
468 }
469 }
470
471 // Array bound check succeeded. From this point forward the array bound
472 // should always succeed.
473 return StInBound;
474}
475
476ProgramStateRef
477CStringChecker::CheckBufferAccess(CheckerContext &C, ProgramStateRef State,
478 AnyArgExpr Buffer, SizeArgExpr Size,
479 AccessKind Access, CharKind CK) const {
480 // If a previous check has failed, propagate the failure.
481 if (!State)
482 return nullptr;
483
484 SValBuilder &svalBuilder = C.getSValBuilder();
485 ASTContext &Ctx = svalBuilder.getContext();
486
487 QualType SizeTy = Size.Expression->getType();
488 QualType PtrTy = getCharPtrType(Ctx, CK);
489
490 // Check that the first buffer is non-null.
491 SVal BufVal = C.getSVal(Buffer.Expression);
492 State = checkNonNull(C, State, Arg: Buffer, l: BufVal);
493 if (!State)
494 return nullptr;
495
496 // If out-of-bounds checking is turned off, skip the rest.
497 if (!Filter.CheckCStringOutOfBounds)
498 return State;
499
500 SVal BufStart =
501 svalBuilder.evalCast(V: BufVal, CastTy: PtrTy, OriginalTy: Buffer.Expression->getType());
502
503 // Check if the first byte of the buffer is accessible.
504 State = CheckLocation(C, state: State, Buffer, Element: BufStart, Access, CK);
505 if (!State)
506 return nullptr;
507
508 // Get the access length and make sure it is known.
509 // FIXME: This assumes the caller has already checked that the access length
510 // is positive. And that it's unsigned.
511 SVal LengthVal = C.getSVal(Size.Expression);
512 std::optional<NonLoc> Length = LengthVal.getAs<NonLoc>();
513 if (!Length)
514 return State;
515
516 // Compute the offset of the last element to be accessed: size-1.
517 NonLoc One = svalBuilder.makeIntVal(integer: 1, type: SizeTy).castAs<NonLoc>();
518 SVal Offset = svalBuilder.evalBinOpNN(state: State, op: BO_Sub, lhs: *Length, rhs: One, resultTy: SizeTy);
519 if (Offset.isUnknown())
520 return nullptr;
521 NonLoc LastOffset = Offset.castAs<NonLoc>();
522
523 // Check that the first buffer is sufficiently long.
524 if (std::optional<Loc> BufLoc = BufStart.getAs<Loc>()) {
525
526 SVal BufEnd =
527 svalBuilder.evalBinOpLN(state: State, op: BO_Add, lhs: *BufLoc, rhs: LastOffset, resultTy: PtrTy);
528 State = CheckLocation(C, state: State, Buffer, Element: BufEnd, Access, CK);
529
530 // If the buffer isn't large enough, abort.
531 if (!State)
532 return nullptr;
533 }
534
535 // Large enough or not, return this state!
536 return State;
537}
538
539ProgramStateRef CStringChecker::CheckOverlap(CheckerContext &C,
540 ProgramStateRef state,
541 SizeArgExpr Size, AnyArgExpr First,
542 AnyArgExpr Second,
543 CharKind CK) const {
544 if (!Filter.CheckCStringBufferOverlap)
545 return state;
546
547 // Do a simple check for overlap: if the two arguments are from the same
548 // buffer, see if the end of the first is greater than the start of the second
549 // or vice versa.
550
551 // If a previous check has failed, propagate the failure.
552 if (!state)
553 return nullptr;
554
555 ProgramStateRef stateTrue, stateFalse;
556
557 // Assume different address spaces cannot overlap.
558 if (First.Expression->getType()->getPointeeType().getAddressSpace() !=
559 Second.Expression->getType()->getPointeeType().getAddressSpace())
560 return state;
561
562 // Get the buffer values and make sure they're known locations.
563 const LocationContext *LCtx = C.getLocationContext();
564 SVal firstVal = state->getSVal(First.Expression, LCtx);
565 SVal secondVal = state->getSVal(Second.Expression, LCtx);
566
567 std::optional<Loc> firstLoc = firstVal.getAs<Loc>();
568 if (!firstLoc)
569 return state;
570
571 std::optional<Loc> secondLoc = secondVal.getAs<Loc>();
572 if (!secondLoc)
573 return state;
574
575 // Are the two values the same?
576 SValBuilder &svalBuilder = C.getSValBuilder();
577 std::tie(args&: stateTrue, args&: stateFalse) =
578 state->assume(Cond: svalBuilder.evalEQ(state, lhs: *firstLoc, rhs: *secondLoc));
579
580 if (stateTrue && !stateFalse) {
581 // If the values are known to be equal, that's automatically an overlap.
582 emitOverlapBug(C, stateTrue, First.Expression, Second.Expression);
583 return nullptr;
584 }
585
586 // assume the two expressions are not equal.
587 assert(stateFalse);
588 state = stateFalse;
589
590 // Which value comes first?
591 QualType cmpTy = svalBuilder.getConditionType();
592 SVal reverse =
593 svalBuilder.evalBinOpLL(state, op: BO_GT, lhs: *firstLoc, rhs: *secondLoc, resultTy: cmpTy);
594 std::optional<DefinedOrUnknownSVal> reverseTest =
595 reverse.getAs<DefinedOrUnknownSVal>();
596 if (!reverseTest)
597 return state;
598
599 std::tie(args&: stateTrue, args&: stateFalse) = state->assume(Cond: *reverseTest);
600 if (stateTrue) {
601 if (stateFalse) {
602 // If we don't know which one comes first, we can't perform this test.
603 return state;
604 } else {
605 // Switch the values so that firstVal is before secondVal.
606 std::swap(lhs&: firstLoc, rhs&: secondLoc);
607
608 // Switch the Exprs as well, so that they still correspond.
609 std::swap(a&: First, b&: Second);
610 }
611 }
612
613 // Get the length, and make sure it too is known.
614 SVal LengthVal = state->getSVal(Size.Expression, LCtx);
615 std::optional<NonLoc> Length = LengthVal.getAs<NonLoc>();
616 if (!Length)
617 return state;
618
619 // Convert the first buffer's start address to char*.
620 // Bail out if the cast fails.
621 ASTContext &Ctx = svalBuilder.getContext();
622 QualType CharPtrTy = getCharPtrType(Ctx, CK);
623 SVal FirstStart =
624 svalBuilder.evalCast(V: *firstLoc, CastTy: CharPtrTy, OriginalTy: First.Expression->getType());
625 std::optional<Loc> FirstStartLoc = FirstStart.getAs<Loc>();
626 if (!FirstStartLoc)
627 return state;
628
629 // Compute the end of the first buffer. Bail out if THAT fails.
630 SVal FirstEnd = svalBuilder.evalBinOpLN(state, op: BO_Add, lhs: *FirstStartLoc,
631 rhs: *Length, resultTy: CharPtrTy);
632 std::optional<Loc> FirstEndLoc = FirstEnd.getAs<Loc>();
633 if (!FirstEndLoc)
634 return state;
635
636 // Is the end of the first buffer past the start of the second buffer?
637 SVal Overlap =
638 svalBuilder.evalBinOpLL(state, op: BO_GT, lhs: *FirstEndLoc, rhs: *secondLoc, resultTy: cmpTy);
639 std::optional<DefinedOrUnknownSVal> OverlapTest =
640 Overlap.getAs<DefinedOrUnknownSVal>();
641 if (!OverlapTest)
642 return state;
643
644 std::tie(args&: stateTrue, args&: stateFalse) = state->assume(Cond: *OverlapTest);
645
646 if (stateTrue && !stateFalse) {
647 // Overlap!
648 emitOverlapBug(C, stateTrue, First.Expression, Second.Expression);
649 return nullptr;
650 }
651
652 // assume the two expressions don't overlap.
653 assert(stateFalse);
654 return stateFalse;
655}
656
657void CStringChecker::emitOverlapBug(CheckerContext &C, ProgramStateRef state,
658 const Stmt *First, const Stmt *Second) const {
659 ExplodedNode *N = C.generateErrorNode(State: state);
660 if (!N)
661 return;
662
663 if (!BT_Overlap)
664 BT_Overlap.reset(p: new BugType(Filter.CheckNameCStringBufferOverlap,
665 categories::UnixAPI, "Improper arguments"));
666
667 // Generate a report for this bug.
668 auto report = std::make_unique<PathSensitiveBugReport>(
669 args&: *BT_Overlap, args: "Arguments must not be overlapping buffers", args&: N);
670 report->addRange(R: First->getSourceRange());
671 report->addRange(R: Second->getSourceRange());
672
673 C.emitReport(R: std::move(report));
674}
675
676void CStringChecker::emitNullArgBug(CheckerContext &C, ProgramStateRef State,
677 const Stmt *S, StringRef WarningMsg) const {
678 if (ExplodedNode *N = C.generateErrorNode(State)) {
679 if (!BT_Null) {
680 // FIXME: This call uses the string constant 'categories::UnixAPI' as the
681 // description of the bug; it should be replaced by a real description.
682 BT_Null.reset(
683 p: new BugType(Filter.CheckNameCStringNullArg, categories::UnixAPI));
684 }
685
686 auto Report =
687 std::make_unique<PathSensitiveBugReport>(args&: *BT_Null, args&: WarningMsg, args&: N);
688 Report->addRange(R: S->getSourceRange());
689 if (const auto *Ex = dyn_cast<Expr>(Val: S))
690 bugreporter::trackExpressionValue(N, E: Ex, R&: *Report);
691 C.emitReport(R: std::move(Report));
692 }
693}
694
695void CStringChecker::emitUninitializedReadBug(CheckerContext &C,
696 ProgramStateRef State,
697 const Expr *E) const {
698 if (ExplodedNode *N = C.generateErrorNode(State)) {
699 const char *Msg =
700 "Bytes string function accesses uninitialized/garbage values";
701 if (!BT_UninitRead)
702 BT_UninitRead.reset(p: new BugType(Filter.CheckNameCStringUninitializedRead,
703 "Accessing unitialized/garbage values"));
704
705 auto Report =
706 std::make_unique<PathSensitiveBugReport>(args&: *BT_UninitRead, args&: Msg, args&: N);
707 Report->addRange(R: E->getSourceRange());
708 bugreporter::trackExpressionValue(N, E, R&: *Report);
709 C.emitReport(R: std::move(Report));
710 }
711}
712
713void CStringChecker::emitOutOfBoundsBug(CheckerContext &C,
714 ProgramStateRef State, const Stmt *S,
715 StringRef WarningMsg) const {
716 if (ExplodedNode *N = C.generateErrorNode(State)) {
717 if (!BT_Bounds)
718 BT_Bounds.reset(p: new BugType(Filter.CheckCStringOutOfBounds
719 ? Filter.CheckNameCStringOutOfBounds
720 : Filter.CheckNameCStringNullArg,
721 "Out-of-bound array access"));
722
723 // FIXME: It would be nice to eventually make this diagnostic more clear,
724 // e.g., by referencing the original declaration or by saying *why* this
725 // reference is outside the range.
726 auto Report =
727 std::make_unique<PathSensitiveBugReport>(args&: *BT_Bounds, args&: WarningMsg, args&: N);
728 Report->addRange(R: S->getSourceRange());
729 C.emitReport(R: std::move(Report));
730 }
731}
732
733void CStringChecker::emitNotCStringBug(CheckerContext &C, ProgramStateRef State,
734 const Stmt *S,
735 StringRef WarningMsg) const {
736 if (ExplodedNode *N = C.generateNonFatalErrorNode(State)) {
737 if (!BT_NotCString) {
738 // FIXME: This call uses the string constant 'categories::UnixAPI' as the
739 // description of the bug; it should be replaced by a real description.
740 BT_NotCString.reset(
741 p: new BugType(Filter.CheckNameCStringNotNullTerm, categories::UnixAPI));
742 }
743
744 auto Report =
745 std::make_unique<PathSensitiveBugReport>(args&: *BT_NotCString, args&: WarningMsg, args&: N);
746
747 Report->addRange(R: S->getSourceRange());
748 C.emitReport(R: std::move(Report));
749 }
750}
751
752void CStringChecker::emitAdditionOverflowBug(CheckerContext &C,
753 ProgramStateRef State) const {
754 if (ExplodedNode *N = C.generateErrorNode(State)) {
755 if (!BT_AdditionOverflow) {
756 // FIXME: This call uses the word "API" as the description of the bug;
757 // it should be replaced by a better error message (if this unlikely
758 // situation continues to exist as a separate bug type).
759 BT_AdditionOverflow.reset(
760 p: new BugType(Filter.CheckNameCStringOutOfBounds, "API"));
761 }
762
763 // This isn't a great error message, but this should never occur in real
764 // code anyway -- you'd have to create a buffer longer than a size_t can
765 // represent, which is sort of a contradiction.
766 const char *WarningMsg =
767 "This expression will create a string whose length is too big to "
768 "be represented as a size_t";
769
770 auto Report = std::make_unique<PathSensitiveBugReport>(args&: *BT_AdditionOverflow,
771 args&: WarningMsg, args&: N);
772 C.emitReport(R: std::move(Report));
773 }
774}
775
776ProgramStateRef CStringChecker::checkAdditionOverflow(CheckerContext &C,
777 ProgramStateRef state,
778 NonLoc left,
779 NonLoc right) const {
780 // If out-of-bounds checking is turned off, skip the rest.
781 if (!Filter.CheckCStringOutOfBounds)
782 return state;
783
784 // If a previous check has failed, propagate the failure.
785 if (!state)
786 return nullptr;
787
788 SValBuilder &svalBuilder = C.getSValBuilder();
789 BasicValueFactory &BVF = svalBuilder.getBasicValueFactory();
790
791 QualType sizeTy = svalBuilder.getContext().getSizeType();
792 const llvm::APSInt &maxValInt = BVF.getMaxValue(T: sizeTy);
793 NonLoc maxVal = svalBuilder.makeIntVal(integer: maxValInt);
794
795 SVal maxMinusRight;
796 if (isa<nonloc::ConcreteInt>(Val: right)) {
797 maxMinusRight = svalBuilder.evalBinOpNN(state, op: BO_Sub, lhs: maxVal, rhs: right,
798 resultTy: sizeTy);
799 } else {
800 // Try switching the operands. (The order of these two assignments is
801 // important!)
802 maxMinusRight = svalBuilder.evalBinOpNN(state, op: BO_Sub, lhs: maxVal, rhs: left,
803 resultTy: sizeTy);
804 left = right;
805 }
806
807 if (std::optional<NonLoc> maxMinusRightNL = maxMinusRight.getAs<NonLoc>()) {
808 QualType cmpTy = svalBuilder.getConditionType();
809 // If left > max - right, we have an overflow.
810 SVal willOverflow = svalBuilder.evalBinOpNN(state, op: BO_GT, lhs: left,
811 rhs: *maxMinusRightNL, resultTy: cmpTy);
812
813 ProgramStateRef stateOverflow, stateOkay;
814 std::tie(args&: stateOverflow, args&: stateOkay) =
815 state->assume(Cond: willOverflow.castAs<DefinedOrUnknownSVal>());
816
817 if (stateOverflow && !stateOkay) {
818 // We have an overflow. Emit a bug report.
819 emitAdditionOverflowBug(C, State: stateOverflow);
820 return nullptr;
821 }
822
823 // From now on, assume an overflow didn't occur.
824 assert(stateOkay);
825 state = stateOkay;
826 }
827
828 return state;
829}
830
831ProgramStateRef CStringChecker::setCStringLength(ProgramStateRef state,
832 const MemRegion *MR,
833 SVal strLength) {
834 assert(!strLength.isUndef() && "Attempt to set an undefined string length");
835
836 MR = MR->StripCasts();
837
838 switch (MR->getKind()) {
839 case MemRegion::StringRegionKind:
840 // FIXME: This can happen if we strcpy() into a string region. This is
841 // undefined [C99 6.4.5p6], but we should still warn about it.
842 return state;
843
844 case MemRegion::SymbolicRegionKind:
845 case MemRegion::AllocaRegionKind:
846 case MemRegion::NonParamVarRegionKind:
847 case MemRegion::ParamVarRegionKind:
848 case MemRegion::FieldRegionKind:
849 case MemRegion::ObjCIvarRegionKind:
850 // These are the types we can currently track string lengths for.
851 break;
852
853 case MemRegion::ElementRegionKind:
854 // FIXME: Handle element regions by upper-bounding the parent region's
855 // string length.
856 return state;
857
858 default:
859 // Other regions (mostly non-data) can't have a reliable C string length.
860 // For now, just ignore the change.
861 // FIXME: These are rare but not impossible. We should output some kind of
862 // warning for things like strcpy((char[]){'a', 0}, "b");
863 return state;
864 }
865
866 if (strLength.isUnknown())
867 return state->remove<CStringLength>(K: MR);
868
869 return state->set<CStringLength>(K: MR, E: strLength);
870}
871
872SVal CStringChecker::getCStringLengthForRegion(CheckerContext &C,
873 ProgramStateRef &state,
874 const Expr *Ex,
875 const MemRegion *MR,
876 bool hypothetical) {
877 if (!hypothetical) {
878 // If there's a recorded length, go ahead and return it.
879 const SVal *Recorded = state->get<CStringLength>(key: MR);
880 if (Recorded)
881 return *Recorded;
882 }
883
884 // Otherwise, get a new symbol and update the state.
885 SValBuilder &svalBuilder = C.getSValBuilder();
886 QualType sizeTy = svalBuilder.getContext().getSizeType();
887 SVal strLength = svalBuilder.getMetadataSymbolVal(symbolTag: CStringChecker::getTag(),
888 region: MR, expr: Ex, type: sizeTy,
889 LCtx: C.getLocationContext(),
890 count: C.blockCount());
891
892 if (!hypothetical) {
893 if (std::optional<NonLoc> strLn = strLength.getAs<NonLoc>()) {
894 // In case of unbounded calls strlen etc bound the range to SIZE_MAX/4
895 BasicValueFactory &BVF = svalBuilder.getBasicValueFactory();
896 const llvm::APSInt &maxValInt = BVF.getMaxValue(T: sizeTy);
897 llvm::APSInt fourInt = APSIntType(maxValInt).getValue(RawValue: 4);
898 const llvm::APSInt *maxLengthInt = BVF.evalAPSInt(Op: BO_Div, V1: maxValInt,
899 V2: fourInt);
900 NonLoc maxLength = svalBuilder.makeIntVal(integer: *maxLengthInt);
901 SVal evalLength = svalBuilder.evalBinOpNN(state, op: BO_LE, lhs: *strLn, rhs: maxLength,
902 resultTy: svalBuilder.getConditionType());
903 state = state->assume(Cond: evalLength.castAs<DefinedOrUnknownSVal>(), Assumption: true);
904 }
905 state = state->set<CStringLength>(K: MR, E: strLength);
906 }
907
908 return strLength;
909}
910
911SVal CStringChecker::getCStringLength(CheckerContext &C, ProgramStateRef &state,
912 const Expr *Ex, SVal Buf,
913 bool hypothetical) const {
914 const MemRegion *MR = Buf.getAsRegion();
915 if (!MR) {
916 // If we can't get a region, see if it's something we /know/ isn't a
917 // C string. In the context of locations, the only time we can issue such
918 // a warning is for labels.
919 if (std::optional<loc::GotoLabel> Label = Buf.getAs<loc::GotoLabel>()) {
920 if (Filter.CheckCStringNotNullTerm) {
921 SmallString<120> buf;
922 llvm::raw_svector_ostream os(buf);
923 assert(CurrentFunctionDescription);
924 os << "Argument to " << CurrentFunctionDescription
925 << " is the address of the label '" << Label->getLabel()->getName()
926 << "', which is not a null-terminated string";
927
928 emitNotCStringBug(C, state, Ex, os.str());
929 }
930 return UndefinedVal();
931 }
932
933 // If it's not a region and not a label, give up.
934 return UnknownVal();
935 }
936
937 // If we have a region, strip casts from it and see if we can figure out
938 // its length. For anything we can't figure out, just return UnknownVal.
939 MR = MR->StripCasts();
940
941 switch (MR->getKind()) {
942 case MemRegion::StringRegionKind: {
943 // Modifying the contents of string regions is undefined [C99 6.4.5p6],
944 // so we can assume that the byte length is the correct C string length.
945 SValBuilder &svalBuilder = C.getSValBuilder();
946 QualType sizeTy = svalBuilder.getContext().getSizeType();
947 const StringLiteral *strLit = cast<StringRegion>(Val: MR)->getStringLiteral();
948 return svalBuilder.makeIntVal(integer: strLit->getLength(), type: sizeTy);
949 }
950 case MemRegion::NonParamVarRegionKind: {
951 // If we have a global constant with a string literal initializer,
952 // compute the initializer's length.
953 const VarDecl *Decl = cast<NonParamVarRegion>(Val: MR)->getDecl();
954 if (Decl->getType().isConstQualified() && Decl->hasGlobalStorage()) {
955 if (const Expr *Init = Decl->getInit()) {
956 if (auto *StrLit = dyn_cast<StringLiteral>(Val: Init)) {
957 SValBuilder &SvalBuilder = C.getSValBuilder();
958 QualType SizeTy = SvalBuilder.getContext().getSizeType();
959 return SvalBuilder.makeIntVal(integer: StrLit->getLength(), type: SizeTy);
960 }
961 }
962 }
963 [[fallthrough]];
964 }
965 case MemRegion::SymbolicRegionKind:
966 case MemRegion::AllocaRegionKind:
967 case MemRegion::ParamVarRegionKind:
968 case MemRegion::FieldRegionKind:
969 case MemRegion::ObjCIvarRegionKind:
970 return getCStringLengthForRegion(C, state, Ex, MR, hypothetical);
971 case MemRegion::CompoundLiteralRegionKind:
972 // FIXME: Can we track this? Is it necessary?
973 return UnknownVal();
974 case MemRegion::ElementRegionKind:
975 // FIXME: How can we handle this? It's not good enough to subtract the
976 // offset from the base string length; consider "123\x00567" and &a[5].
977 return UnknownVal();
978 default:
979 // Other regions (mostly non-data) can't have a reliable C string length.
980 // In this case, an error is emitted and UndefinedVal is returned.
981 // The caller should always be prepared to handle this case.
982 if (Filter.CheckCStringNotNullTerm) {
983 SmallString<120> buf;
984 llvm::raw_svector_ostream os(buf);
985
986 assert(CurrentFunctionDescription);
987 os << "Argument to " << CurrentFunctionDescription << " is ";
988
989 if (SummarizeRegion(os, Ctx&: C.getASTContext(), MR))
990 os << ", which is not a null-terminated string";
991 else
992 os << "not a null-terminated string";
993
994 emitNotCStringBug(C, state, Ex, os.str());
995 }
996 return UndefinedVal();
997 }
998}
999
1000const StringLiteral *CStringChecker::getCStringLiteral(CheckerContext &C,
1001 ProgramStateRef &state, const Expr *expr, SVal val) const {
1002
1003 // Get the memory region pointed to by the val.
1004 const MemRegion *bufRegion = val.getAsRegion();
1005 if (!bufRegion)
1006 return nullptr;
1007
1008 // Strip casts off the memory region.
1009 bufRegion = bufRegion->StripCasts();
1010
1011 // Cast the memory region to a string region.
1012 const StringRegion *strRegion= dyn_cast<StringRegion>(Val: bufRegion);
1013 if (!strRegion)
1014 return nullptr;
1015
1016 // Return the actual string in the string region.
1017 return strRegion->getStringLiteral();
1018}
1019
1020bool CStringChecker::isFirstBufInBound(CheckerContext &C, ProgramStateRef State,
1021 SVal BufVal, QualType BufTy,
1022 SVal LengthVal, QualType LengthTy) {
1023 // If we do not know that the buffer is long enough we return 'true'.
1024 // Otherwise the parent region of this field region would also get
1025 // invalidated, which would lead to warnings based on an unknown state.
1026
1027 if (LengthVal.isUnknown())
1028 return false;
1029
1030 // Originally copied from CheckBufferAccess and CheckLocation.
1031 SValBuilder &SB = C.getSValBuilder();
1032 ASTContext &Ctx = C.getASTContext();
1033
1034 QualType PtrTy = Ctx.getPointerType(Ctx.CharTy);
1035
1036 std::optional<NonLoc> Length = LengthVal.getAs<NonLoc>();
1037 if (!Length)
1038 return true; // cf top comment.
1039
1040 // Compute the offset of the last element to be accessed: size-1.
1041 NonLoc One = SB.makeIntVal(integer: 1, type: LengthTy).castAs<NonLoc>();
1042 SVal Offset = SB.evalBinOpNN(state: State, op: BO_Sub, lhs: *Length, rhs: One, resultTy: LengthTy);
1043 if (Offset.isUnknown())
1044 return true; // cf top comment
1045 NonLoc LastOffset = Offset.castAs<NonLoc>();
1046
1047 // Check that the first buffer is sufficiently long.
1048 SVal BufStart = SB.evalCast(V: BufVal, CastTy: PtrTy, OriginalTy: BufTy);
1049 std::optional<Loc> BufLoc = BufStart.getAs<Loc>();
1050 if (!BufLoc)
1051 return true; // cf top comment.
1052
1053 SVal BufEnd = SB.evalBinOpLN(state: State, op: BO_Add, lhs: *BufLoc, rhs: LastOffset, resultTy: PtrTy);
1054
1055 // Check for out of bound array element access.
1056 const MemRegion *R = BufEnd.getAsRegion();
1057 if (!R)
1058 return true; // cf top comment.
1059
1060 const ElementRegion *ER = dyn_cast<ElementRegion>(Val: R);
1061 if (!ER)
1062 return true; // cf top comment.
1063
1064 // FIXME: Does this crash when a non-standard definition
1065 // of a library function is encountered?
1066 assert(ER->getValueType() == C.getASTContext().CharTy &&
1067 "isFirstBufInBound should only be called with char* ElementRegions");
1068
1069 // Get the size of the array.
1070 const SubRegion *superReg = cast<SubRegion>(ER->getSuperRegion());
1071 DefinedOrUnknownSVal SizeDV = getDynamicExtent(State, MR: superReg, SVB&: SB);
1072
1073 // Get the index of the accessed element.
1074 DefinedOrUnknownSVal Idx = ER->getIndex().castAs<DefinedOrUnknownSVal>();
1075
1076 ProgramStateRef StInBound = State->assumeInBound(idx: Idx, upperBound: SizeDV, assumption: true);
1077
1078 return static_cast<bool>(StInBound);
1079}
1080
1081ProgramStateRef CStringChecker::invalidateDestinationBufferBySize(
1082 CheckerContext &C, ProgramStateRef S, const Expr *BufE, SVal BufV,
1083 SVal SizeV, QualType SizeTy) {
1084 auto InvalidationTraitOperations =
1085 [&C, S, BufTy = BufE->getType(), BufV, SizeV,
1086 SizeTy](RegionAndSymbolInvalidationTraits &ITraits, const MemRegion *R) {
1087 // If destination buffer is a field region and access is in bound, do
1088 // not invalidate its super region.
1089 if (MemRegion::FieldRegionKind == R->getKind() &&
1090 isFirstBufInBound(C, State: S, BufVal: BufV, BufTy, LengthVal: SizeV, LengthTy: SizeTy)) {
1091 ITraits.setTrait(
1092 MR: R,
1093 IK: RegionAndSymbolInvalidationTraits::TK_DoNotInvalidateSuperRegion);
1094 }
1095 return false;
1096 };
1097
1098 return invalidateBufferAux(C, State: S, Ex: BufE, V: BufV, InvalidationTraitOperations);
1099}
1100
1101ProgramStateRef
1102CStringChecker::invalidateDestinationBufferAlwaysEscapeSuperRegion(
1103 CheckerContext &C, ProgramStateRef S, const Expr *BufE, SVal BufV) {
1104 auto InvalidationTraitOperations = [](RegionAndSymbolInvalidationTraits &,
1105 const MemRegion *R) {
1106 return isa<FieldRegion>(Val: R);
1107 };
1108
1109 return invalidateBufferAux(C, State: S, Ex: BufE, V: BufV, InvalidationTraitOperations);
1110}
1111
1112ProgramStateRef CStringChecker::invalidateDestinationBufferNeverOverflows(
1113 CheckerContext &C, ProgramStateRef S, const Expr *BufE, SVal BufV) {
1114 auto InvalidationTraitOperations =
1115 [](RegionAndSymbolInvalidationTraits &ITraits, const MemRegion *R) {
1116 if (MemRegion::FieldRegionKind == R->getKind())
1117 ITraits.setTrait(
1118 MR: R,
1119 IK: RegionAndSymbolInvalidationTraits::TK_DoNotInvalidateSuperRegion);
1120 return false;
1121 };
1122
1123 return invalidateBufferAux(C, State: S, Ex: BufE, V: BufV, InvalidationTraitOperations);
1124}
1125
1126ProgramStateRef CStringChecker::invalidateSourceBuffer(CheckerContext &C,
1127 ProgramStateRef S,
1128 const Expr *BufE,
1129 SVal BufV) {
1130 auto InvalidationTraitOperations =
1131 [](RegionAndSymbolInvalidationTraits &ITraits, const MemRegion *R) {
1132 ITraits.setTrait(
1133 MR: R->getBaseRegion(),
1134 IK: RegionAndSymbolInvalidationTraits::TK_PreserveContents);
1135 ITraits.setTrait(MR: R,
1136 IK: RegionAndSymbolInvalidationTraits::TK_SuppressEscape);
1137 return true;
1138 };
1139
1140 return invalidateBufferAux(C, State: S, Ex: BufE, V: BufV, InvalidationTraitOperations);
1141}
1142
1143ProgramStateRef CStringChecker::invalidateBufferAux(
1144 CheckerContext &C, ProgramStateRef State, const Expr *E, SVal V,
1145 llvm::function_ref<bool(RegionAndSymbolInvalidationTraits &,
1146 const MemRegion *)>
1147 InvalidationTraitOperations) {
1148 std::optional<Loc> L = V.getAs<Loc>();
1149 if (!L)
1150 return State;
1151
1152 // FIXME: This is a simplified version of what's in CFRefCount.cpp -- it makes
1153 // some assumptions about the value that CFRefCount can't. Even so, it should
1154 // probably be refactored.
1155 if (std::optional<loc::MemRegionVal> MR = L->getAs<loc::MemRegionVal>()) {
1156 const MemRegion *R = MR->getRegion()->StripCasts();
1157
1158 // Are we dealing with an ElementRegion? If so, we should be invalidating
1159 // the super-region.
1160 if (const ElementRegion *ER = dyn_cast<ElementRegion>(Val: R)) {
1161 R = ER->getSuperRegion();
1162 // FIXME: What about layers of ElementRegions?
1163 }
1164
1165 // Invalidate this region.
1166 const LocationContext *LCtx = C.getPredecessor()->getLocationContext();
1167 RegionAndSymbolInvalidationTraits ITraits;
1168 bool CausesPointerEscape = InvalidationTraitOperations(ITraits, R);
1169
1170 return State->invalidateRegions(Regions: R, E, BlockCount: C.blockCount(), LCtx,
1171 CausesPointerEscape, IS: nullptr, Call: nullptr,
1172 ITraits: &ITraits);
1173 }
1174
1175 // If we have a non-region value by chance, just remove the binding.
1176 // FIXME: is this necessary or correct? This handles the non-Region
1177 // cases. Is it ever valid to store to these?
1178 return State->killBinding(LV: *L);
1179}
1180
1181bool CStringChecker::SummarizeRegion(raw_ostream &os, ASTContext &Ctx,
1182 const MemRegion *MR) {
1183 switch (MR->getKind()) {
1184 case MemRegion::FunctionCodeRegionKind: {
1185 if (const auto *FD = cast<FunctionCodeRegion>(Val: MR)->getDecl())
1186 os << "the address of the function '" << *FD << '\'';
1187 else
1188 os << "the address of a function";
1189 return true;
1190 }
1191 case MemRegion::BlockCodeRegionKind:
1192 os << "block text";
1193 return true;
1194 case MemRegion::BlockDataRegionKind:
1195 os << "a block";
1196 return true;
1197 case MemRegion::CXXThisRegionKind:
1198 case MemRegion::CXXTempObjectRegionKind:
1199 os << "a C++ temp object of type "
1200 << cast<TypedValueRegion>(Val: MR)->getValueType();
1201 return true;
1202 case MemRegion::NonParamVarRegionKind:
1203 os << "a variable of type" << cast<TypedValueRegion>(Val: MR)->getValueType();
1204 return true;
1205 case MemRegion::ParamVarRegionKind:
1206 os << "a parameter of type" << cast<TypedValueRegion>(Val: MR)->getValueType();
1207 return true;
1208 case MemRegion::FieldRegionKind:
1209 os << "a field of type " << cast<TypedValueRegion>(Val: MR)->getValueType();
1210 return true;
1211 case MemRegion::ObjCIvarRegionKind:
1212 os << "an instance variable of type "
1213 << cast<TypedValueRegion>(Val: MR)->getValueType();
1214 return true;
1215 default:
1216 return false;
1217 }
1218}
1219
1220bool CStringChecker::memsetAux(const Expr *DstBuffer, SVal CharVal,
1221 const Expr *Size, CheckerContext &C,
1222 ProgramStateRef &State) {
1223 SVal MemVal = C.getSVal(DstBuffer);
1224 SVal SizeVal = C.getSVal(Size);
1225 const MemRegion *MR = MemVal.getAsRegion();
1226 if (!MR)
1227 return false;
1228
1229 // We're about to model memset by producing a "default binding" in the Store.
1230 // Our current implementation - RegionStore - doesn't support default bindings
1231 // that don't cover the whole base region. So we should first get the offset
1232 // and the base region to figure out whether the offset of buffer is 0.
1233 RegionOffset Offset = MR->getAsOffset();
1234 const MemRegion *BR = Offset.getRegion();
1235
1236 std::optional<NonLoc> SizeNL = SizeVal.getAs<NonLoc>();
1237 if (!SizeNL)
1238 return false;
1239
1240 SValBuilder &svalBuilder = C.getSValBuilder();
1241 ASTContext &Ctx = C.getASTContext();
1242
1243 // void *memset(void *dest, int ch, size_t count);
1244 // For now we can only handle the case of offset is 0 and concrete char value.
1245 if (Offset.isValid() && !Offset.hasSymbolicOffset() &&
1246 Offset.getOffset() == 0) {
1247 // Get the base region's size.
1248 DefinedOrUnknownSVal SizeDV = getDynamicExtent(State, MR: BR, SVB&: svalBuilder);
1249
1250 ProgramStateRef StateWholeReg, StateNotWholeReg;
1251 std::tie(args&: StateWholeReg, args&: StateNotWholeReg) =
1252 State->assume(Cond: svalBuilder.evalEQ(state: State, lhs: SizeDV, rhs: *SizeNL));
1253
1254 // With the semantic of 'memset()', we should convert the CharVal to
1255 // unsigned char.
1256 CharVal = svalBuilder.evalCast(V: CharVal, CastTy: Ctx.UnsignedCharTy, OriginalTy: Ctx.IntTy);
1257
1258 ProgramStateRef StateNullChar, StateNonNullChar;
1259 std::tie(args&: StateNullChar, args&: StateNonNullChar) =
1260 assumeZero(C, state: State, V: CharVal, Ty: Ctx.UnsignedCharTy);
1261
1262 if (StateWholeReg && !StateNotWholeReg && StateNullChar &&
1263 !StateNonNullChar) {
1264 // If the 'memset()' acts on the whole region of destination buffer and
1265 // the value of the second argument of 'memset()' is zero, bind the second
1266 // argument's value to the destination buffer with 'default binding'.
1267 // FIXME: Since there is no perfect way to bind the non-zero character, we
1268 // can only deal with zero value here. In the future, we need to deal with
1269 // the binding of non-zero value in the case of whole region.
1270 State = State->bindDefaultZero(loc: svalBuilder.makeLoc(region: BR),
1271 LCtx: C.getLocationContext());
1272 } else {
1273 // If the destination buffer's extent is not equal to the value of
1274 // third argument, just invalidate buffer.
1275 State = invalidateDestinationBufferBySize(C, S: State, BufE: DstBuffer, BufV: MemVal,
1276 SizeV: SizeVal, SizeTy: Size->getType());
1277 }
1278
1279 if (StateNullChar && !StateNonNullChar) {
1280 // If the value of the second argument of 'memset()' is zero, set the
1281 // string length of destination buffer to 0 directly.
1282 State = setCStringLength(state: State, MR,
1283 strLength: svalBuilder.makeZeroVal(type: Ctx.getSizeType()));
1284 } else if (!StateNullChar && StateNonNullChar) {
1285 SVal NewStrLen = svalBuilder.getMetadataSymbolVal(
1286 symbolTag: CStringChecker::getTag(), region: MR, expr: DstBuffer, type: Ctx.getSizeType(),
1287 LCtx: C.getLocationContext(), count: C.blockCount());
1288
1289 // If the value of second argument is not zero, then the string length
1290 // is at least the size argument.
1291 SVal NewStrLenGESize = svalBuilder.evalBinOp(
1292 state: State, op: BO_GE, lhs: NewStrLen, rhs: SizeVal, type: svalBuilder.getConditionType());
1293
1294 State = setCStringLength(
1295 state: State->assume(Cond: NewStrLenGESize.castAs<DefinedOrUnknownSVal>(), Assumption: true),
1296 MR, strLength: NewStrLen);
1297 }
1298 } else {
1299 // If the offset is not zero and char value is not concrete, we can do
1300 // nothing but invalidate the buffer.
1301 State = invalidateDestinationBufferBySize(C, S: State, BufE: DstBuffer, BufV: MemVal,
1302 SizeV: SizeVal, SizeTy: Size->getType());
1303 }
1304 return true;
1305}
1306
1307//===----------------------------------------------------------------------===//
1308// evaluation of individual function calls.
1309//===----------------------------------------------------------------------===//
1310
1311void CStringChecker::evalCopyCommon(CheckerContext &C, const CallEvent &Call,
1312 ProgramStateRef state, SizeArgExpr Size,
1313 DestinationArgExpr Dest,
1314 SourceArgExpr Source, bool Restricted,
1315 bool IsMempcpy, CharKind CK) const {
1316 CurrentFunctionDescription = "memory copy function";
1317
1318 // See if the size argument is zero.
1319 const LocationContext *LCtx = C.getLocationContext();
1320 SVal sizeVal = state->getSVal(Size.Expression, LCtx);
1321 QualType sizeTy = Size.Expression->getType();
1322
1323 ProgramStateRef stateZeroSize, stateNonZeroSize;
1324 std::tie(args&: stateZeroSize, args&: stateNonZeroSize) =
1325 assumeZero(C, state, V: sizeVal, Ty: sizeTy);
1326
1327 // Get the value of the Dest.
1328 SVal destVal = state->getSVal(Dest.Expression, LCtx);
1329
1330 // If the size is zero, there won't be any actual memory access, so
1331 // just bind the return value to the destination buffer and return.
1332 if (stateZeroSize && !stateNonZeroSize) {
1333 stateZeroSize =
1334 stateZeroSize->BindExpr(Call.getOriginExpr(), LCtx, destVal);
1335 C.addTransition(State: stateZeroSize);
1336 return;
1337 }
1338
1339 // If the size can be nonzero, we have to check the other arguments.
1340 if (stateNonZeroSize) {
1341 state = stateNonZeroSize;
1342
1343 // Ensure the destination is not null. If it is NULL there will be a
1344 // NULL pointer dereference.
1345 state = checkNonNull(C, State: state, Arg: Dest, l: destVal);
1346 if (!state)
1347 return;
1348
1349 // Get the value of the Src.
1350 SVal srcVal = state->getSVal(Source.Expression, LCtx);
1351
1352 // Ensure the source is not null. If it is NULL there will be a
1353 // NULL pointer dereference.
1354 state = checkNonNull(C, State: state, Arg: Source, l: srcVal);
1355 if (!state)
1356 return;
1357
1358 // Ensure the accesses are valid and that the buffers do not overlap.
1359 state = CheckBufferAccess(C, State: state, Buffer: Dest, Size, Access: AccessKind::write, CK);
1360 state = CheckBufferAccess(C, State: state, Buffer: Source, Size, Access: AccessKind::read, CK);
1361
1362 if (Restricted)
1363 state = CheckOverlap(C, state, Size, First: Dest, Second: Source, CK);
1364
1365 if (!state)
1366 return;
1367
1368 // If this is mempcpy, get the byte after the last byte copied and
1369 // bind the expr.
1370 if (IsMempcpy) {
1371 // Get the byte after the last byte copied.
1372 SValBuilder &SvalBuilder = C.getSValBuilder();
1373 ASTContext &Ctx = SvalBuilder.getContext();
1374 QualType CharPtrTy = getCharPtrType(Ctx, CK);
1375 SVal DestRegCharVal =
1376 SvalBuilder.evalCast(V: destVal, CastTy: CharPtrTy, OriginalTy: Dest.Expression->getType());
1377 SVal lastElement = C.getSValBuilder().evalBinOp(
1378 state, op: BO_Add, lhs: DestRegCharVal, rhs: sizeVal, type: Dest.Expression->getType());
1379 // If we don't know how much we copied, we can at least
1380 // conjure a return value for later.
1381 if (lastElement.isUnknown())
1382 lastElement = C.getSValBuilder().conjureSymbolVal(
1383 symbolTag: nullptr, expr: Call.getOriginExpr(), LCtx, count: C.blockCount());
1384
1385 // The byte after the last byte copied is the return value.
1386 state = state->BindExpr(Call.getOriginExpr(), LCtx, lastElement);
1387 } else {
1388 // All other copies return the destination buffer.
1389 // (Well, bcopy() has a void return type, but this won't hurt.)
1390 state = state->BindExpr(Call.getOriginExpr(), LCtx, destVal);
1391 }
1392
1393 // Invalidate the destination (regular invalidation without pointer-escaping
1394 // the address of the top-level region).
1395 // FIXME: Even if we can't perfectly model the copy, we should see if we
1396 // can use LazyCompoundVals to copy the source values into the destination.
1397 // This would probably remove any existing bindings past the end of the
1398 // copied region, but that's still an improvement over blank invalidation.
1399 state = invalidateDestinationBufferBySize(
1400 C, S: state, BufE: Dest.Expression, BufV: C.getSVal(Dest.Expression), SizeV: sizeVal,
1401 SizeTy: Size.Expression->getType());
1402
1403 // Invalidate the source (const-invalidation without const-pointer-escaping
1404 // the address of the top-level region).
1405 state = invalidateSourceBuffer(C, S: state, BufE: Source.Expression,
1406 BufV: C.getSVal(Source.Expression));
1407
1408 C.addTransition(State: state);
1409 }
1410}
1411
1412void CStringChecker::evalMemcpy(CheckerContext &C, const CallEvent &Call,
1413 CharKind CK) const {
1414 // void *memcpy(void *restrict dst, const void *restrict src, size_t n);
1415 // The return value is the address of the destination buffer.
1416 DestinationArgExpr Dest = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
1417 SourceArgExpr Src = {{.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1}};
1418 SizeArgExpr Size = {{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}};
1419
1420 ProgramStateRef State = C.getState();
1421
1422 constexpr bool IsRestricted = true;
1423 constexpr bool IsMempcpy = false;
1424 evalCopyCommon(C, Call, state: State, Size, Dest, Source: Src, Restricted: IsRestricted, IsMempcpy, CK);
1425}
1426
1427void CStringChecker::evalMempcpy(CheckerContext &C, const CallEvent &Call,
1428 CharKind CK) const {
1429 // void *mempcpy(void *restrict dst, const void *restrict src, size_t n);
1430 // The return value is a pointer to the byte following the last written byte.
1431 DestinationArgExpr Dest = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
1432 SourceArgExpr Src = {{.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1}};
1433 SizeArgExpr Size = {{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}};
1434
1435 constexpr bool IsRestricted = true;
1436 constexpr bool IsMempcpy = true;
1437 evalCopyCommon(C, Call, state: C.getState(), Size, Dest, Source: Src, Restricted: IsRestricted,
1438 IsMempcpy, CK);
1439}
1440
1441void CStringChecker::evalMemmove(CheckerContext &C, const CallEvent &Call,
1442 CharKind CK) const {
1443 // void *memmove(void *dst, const void *src, size_t n);
1444 // The return value is the address of the destination buffer.
1445 DestinationArgExpr Dest = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
1446 SourceArgExpr Src = {{.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1}};
1447 SizeArgExpr Size = {{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}};
1448
1449 constexpr bool IsRestricted = false;
1450 constexpr bool IsMempcpy = false;
1451 evalCopyCommon(C, Call, state: C.getState(), Size, Dest, Source: Src, Restricted: IsRestricted,
1452 IsMempcpy, CK);
1453}
1454
1455void CStringChecker::evalBcopy(CheckerContext &C, const CallEvent &Call) const {
1456 // void bcopy(const void *src, void *dst, size_t n);
1457 SourceArgExpr Src{{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
1458 DestinationArgExpr Dest = {{.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1}};
1459 SizeArgExpr Size = {{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}};
1460
1461 constexpr bool IsRestricted = false;
1462 constexpr bool IsMempcpy = false;
1463 evalCopyCommon(C, Call, state: C.getState(), Size, Dest, Source: Src, Restricted: IsRestricted,
1464 IsMempcpy, CK: CharKind::Regular);
1465}
1466
1467void CStringChecker::evalMemcmp(CheckerContext &C, const CallEvent &Call,
1468 CharKind CK) const {
1469 // int memcmp(const void *s1, const void *s2, size_t n);
1470 CurrentFunctionDescription = "memory comparison function";
1471
1472 AnyArgExpr Left = {.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0};
1473 AnyArgExpr Right = {.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1};
1474 SizeArgExpr Size = {{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}};
1475
1476 ProgramStateRef State = C.getState();
1477 SValBuilder &Builder = C.getSValBuilder();
1478 const LocationContext *LCtx = C.getLocationContext();
1479
1480 // See if the size argument is zero.
1481 SVal sizeVal = State->getSVal(Size.Expression, LCtx);
1482 QualType sizeTy = Size.Expression->getType();
1483
1484 ProgramStateRef stateZeroSize, stateNonZeroSize;
1485 std::tie(args&: stateZeroSize, args&: stateNonZeroSize) =
1486 assumeZero(C, state: State, V: sizeVal, Ty: sizeTy);
1487
1488 // If the size can be zero, the result will be 0 in that case, and we don't
1489 // have to check either of the buffers.
1490 if (stateZeroSize) {
1491 State = stateZeroSize;
1492 State = State->BindExpr(Call.getOriginExpr(), LCtx,
1493 Builder.makeZeroVal(type: Call.getResultType()));
1494 C.addTransition(State);
1495 }
1496
1497 // If the size can be nonzero, we have to check the other arguments.
1498 if (stateNonZeroSize) {
1499 State = stateNonZeroSize;
1500 // If we know the two buffers are the same, we know the result is 0.
1501 // First, get the two buffers' addresses. Another checker will have already
1502 // made sure they're not undefined.
1503 DefinedOrUnknownSVal LV =
1504 State->getSVal(Left.Expression, LCtx).castAs<DefinedOrUnknownSVal>();
1505 DefinedOrUnknownSVal RV =
1506 State->getSVal(Right.Expression, LCtx).castAs<DefinedOrUnknownSVal>();
1507
1508 // See if they are the same.
1509 ProgramStateRef SameBuffer, NotSameBuffer;
1510 std::tie(args&: SameBuffer, args&: NotSameBuffer) =
1511 State->assume(Cond: Builder.evalEQ(state: State, lhs: LV, rhs: RV));
1512
1513 // If the two arguments are the same buffer, we know the result is 0,
1514 // and we only need to check one size.
1515 if (SameBuffer && !NotSameBuffer) {
1516 State = SameBuffer;
1517 State = CheckBufferAccess(C, State, Buffer: Left, Size, Access: AccessKind::read);
1518 if (State) {
1519 State = SameBuffer->BindExpr(Call.getOriginExpr(), LCtx,
1520 Builder.makeZeroVal(type: Call.getResultType()));
1521 C.addTransition(State);
1522 }
1523 return;
1524 }
1525
1526 // If the two arguments might be different buffers, we have to check
1527 // the size of both of them.
1528 assert(NotSameBuffer);
1529 State = CheckBufferAccess(C, State, Buffer: Right, Size, Access: AccessKind::read, CK);
1530 State = CheckBufferAccess(C, State, Buffer: Left, Size, Access: AccessKind::read, CK);
1531 if (State) {
1532 // The return value is the comparison result, which we don't know.
1533 SVal CmpV = Builder.conjureSymbolVal(symbolTag: nullptr, expr: Call.getOriginExpr(), LCtx,
1534 count: C.blockCount());
1535 State = State->BindExpr(Call.getOriginExpr(), LCtx, CmpV);
1536 C.addTransition(State);
1537 }
1538 }
1539}
1540
1541void CStringChecker::evalstrLength(CheckerContext &C,
1542 const CallEvent &Call) const {
1543 // size_t strlen(const char *s);
1544 evalstrLengthCommon(C, Call, /* IsStrnlen = */ false);
1545}
1546
1547void CStringChecker::evalstrnLength(CheckerContext &C,
1548 const CallEvent &Call) const {
1549 // size_t strnlen(const char *s, size_t maxlen);
1550 evalstrLengthCommon(C, Call, /* IsStrnlen = */ true);
1551}
1552
1553void CStringChecker::evalstrLengthCommon(CheckerContext &C,
1554 const CallEvent &Call,
1555 bool IsStrnlen) const {
1556 CurrentFunctionDescription = "string length function";
1557 ProgramStateRef state = C.getState();
1558 const LocationContext *LCtx = C.getLocationContext();
1559
1560 if (IsStrnlen) {
1561 const Expr *maxlenExpr = Call.getArgExpr(Index: 1);
1562 SVal maxlenVal = state->getSVal(maxlenExpr, LCtx);
1563
1564 ProgramStateRef stateZeroSize, stateNonZeroSize;
1565 std::tie(args&: stateZeroSize, args&: stateNonZeroSize) =
1566 assumeZero(C, state, V: maxlenVal, Ty: maxlenExpr->getType());
1567
1568 // If the size can be zero, the result will be 0 in that case, and we don't
1569 // have to check the string itself.
1570 if (stateZeroSize) {
1571 SVal zero = C.getSValBuilder().makeZeroVal(type: Call.getResultType());
1572 stateZeroSize = stateZeroSize->BindExpr(Call.getOriginExpr(), LCtx, zero);
1573 C.addTransition(State: stateZeroSize);
1574 }
1575
1576 // If the size is GUARANTEED to be zero, we're done!
1577 if (!stateNonZeroSize)
1578 return;
1579
1580 // Otherwise, record the assumption that the size is nonzero.
1581 state = stateNonZeroSize;
1582 }
1583
1584 // Check that the string argument is non-null.
1585 AnyArgExpr Arg = {.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0};
1586 SVal ArgVal = state->getSVal(Arg.Expression, LCtx);
1587 state = checkNonNull(C, State: state, Arg, l: ArgVal);
1588
1589 if (!state)
1590 return;
1591
1592 SVal strLength = getCStringLength(C, state, Ex: Arg.Expression, Buf: ArgVal);
1593
1594 // If the argument isn't a valid C string, there's no valid state to
1595 // transition to.
1596 if (strLength.isUndef())
1597 return;
1598
1599 DefinedOrUnknownSVal result = UnknownVal();
1600
1601 // If the check is for strnlen() then bind the return value to no more than
1602 // the maxlen value.
1603 if (IsStrnlen) {
1604 QualType cmpTy = C.getSValBuilder().getConditionType();
1605
1606 // It's a little unfortunate to be getting this again,
1607 // but it's not that expensive...
1608 const Expr *maxlenExpr = Call.getArgExpr(Index: 1);
1609 SVal maxlenVal = state->getSVal(maxlenExpr, LCtx);
1610
1611 std::optional<NonLoc> strLengthNL = strLength.getAs<NonLoc>();
1612 std::optional<NonLoc> maxlenValNL = maxlenVal.getAs<NonLoc>();
1613
1614 if (strLengthNL && maxlenValNL) {
1615 ProgramStateRef stateStringTooLong, stateStringNotTooLong;
1616
1617 // Check if the strLength is greater than the maxlen.
1618 std::tie(args&: stateStringTooLong, args&: stateStringNotTooLong) = state->assume(
1619 Cond: C.getSValBuilder()
1620 .evalBinOpNN(state, op: BO_GT, lhs: *strLengthNL, rhs: *maxlenValNL, resultTy: cmpTy)
1621 .castAs<DefinedOrUnknownSVal>());
1622
1623 if (stateStringTooLong && !stateStringNotTooLong) {
1624 // If the string is longer than maxlen, return maxlen.
1625 result = *maxlenValNL;
1626 } else if (stateStringNotTooLong && !stateStringTooLong) {
1627 // If the string is shorter than maxlen, return its length.
1628 result = *strLengthNL;
1629 }
1630 }
1631
1632 if (result.isUnknown()) {
1633 // If we don't have enough information for a comparison, there's
1634 // no guarantee the full string length will actually be returned.
1635 // All we know is the return value is the min of the string length
1636 // and the limit. This is better than nothing.
1637 result = C.getSValBuilder().conjureSymbolVal(
1638 symbolTag: nullptr, expr: Call.getOriginExpr(), LCtx, count: C.blockCount());
1639 NonLoc resultNL = result.castAs<NonLoc>();
1640
1641 if (strLengthNL) {
1642 state = state->assume(Cond: C.getSValBuilder().evalBinOpNN(
1643 state, op: BO_LE, lhs: resultNL, rhs: *strLengthNL, resultTy: cmpTy)
1644 .castAs<DefinedOrUnknownSVal>(), Assumption: true);
1645 }
1646
1647 if (maxlenValNL) {
1648 state = state->assume(Cond: C.getSValBuilder().evalBinOpNN(
1649 state, op: BO_LE, lhs: resultNL, rhs: *maxlenValNL, resultTy: cmpTy)
1650 .castAs<DefinedOrUnknownSVal>(), Assumption: true);
1651 }
1652 }
1653
1654 } else {
1655 // This is a plain strlen(), not strnlen().
1656 result = strLength.castAs<DefinedOrUnknownSVal>();
1657
1658 // If we don't know the length of the string, conjure a return
1659 // value, so it can be used in constraints, at least.
1660 if (result.isUnknown()) {
1661 result = C.getSValBuilder().conjureSymbolVal(
1662 symbolTag: nullptr, expr: Call.getOriginExpr(), LCtx, count: C.blockCount());
1663 }
1664 }
1665
1666 // Bind the return value.
1667 assert(!result.isUnknown() && "Should have conjured a value by now");
1668 state = state->BindExpr(Call.getOriginExpr(), LCtx, result);
1669 C.addTransition(State: state);
1670}
1671
1672void CStringChecker::evalStrcpy(CheckerContext &C,
1673 const CallEvent &Call) const {
1674 // char *strcpy(char *restrict dst, const char *restrict src);
1675 evalStrcpyCommon(C, Call,
1676 /* ReturnEnd = */ false,
1677 /* IsBounded = */ false,
1678 /* appendK = */ ConcatFnKind::none);
1679}
1680
1681void CStringChecker::evalStrncpy(CheckerContext &C,
1682 const CallEvent &Call) const {
1683 // char *strncpy(char *restrict dst, const char *restrict src, size_t n);
1684 evalStrcpyCommon(C, Call,
1685 /* ReturnEnd = */ false,
1686 /* IsBounded = */ true,
1687 /* appendK = */ ConcatFnKind::none);
1688}
1689
1690void CStringChecker::evalStpcpy(CheckerContext &C,
1691 const CallEvent &Call) const {
1692 // char *stpcpy(char *restrict dst, const char *restrict src);
1693 evalStrcpyCommon(C, Call,
1694 /* ReturnEnd = */ true,
1695 /* IsBounded = */ false,
1696 /* appendK = */ ConcatFnKind::none);
1697}
1698
1699void CStringChecker::evalStrlcpy(CheckerContext &C,
1700 const CallEvent &Call) const {
1701 // size_t strlcpy(char *dest, const char *src, size_t size);
1702 evalStrcpyCommon(C, Call,
1703 /* ReturnEnd = */ true,
1704 /* IsBounded = */ true,
1705 /* appendK = */ ConcatFnKind::none,
1706 /* returnPtr = */ false);
1707}
1708
1709void CStringChecker::evalStrcat(CheckerContext &C,
1710 const CallEvent &Call) const {
1711 // char *strcat(char *restrict s1, const char *restrict s2);
1712 evalStrcpyCommon(C, Call,
1713 /* ReturnEnd = */ false,
1714 /* IsBounded = */ false,
1715 /* appendK = */ ConcatFnKind::strcat);
1716}
1717
1718void CStringChecker::evalStrncat(CheckerContext &C,
1719 const CallEvent &Call) const {
1720 // char *strncat(char *restrict s1, const char *restrict s2, size_t n);
1721 evalStrcpyCommon(C, Call,
1722 /* ReturnEnd = */ false,
1723 /* IsBounded = */ true,
1724 /* appendK = */ ConcatFnKind::strcat);
1725}
1726
1727void CStringChecker::evalStrlcat(CheckerContext &C,
1728 const CallEvent &Call) const {
1729 // size_t strlcat(char *dst, const char *src, size_t size);
1730 // It will append at most size - strlen(dst) - 1 bytes,
1731 // NULL-terminating the result.
1732 evalStrcpyCommon(C, Call,
1733 /* ReturnEnd = */ false,
1734 /* IsBounded = */ true,
1735 /* appendK = */ ConcatFnKind::strlcat,
1736 /* returnPtr = */ false);
1737}
1738
1739void CStringChecker::evalStrcpyCommon(CheckerContext &C, const CallEvent &Call,
1740 bool ReturnEnd, bool IsBounded,
1741 ConcatFnKind appendK,
1742 bool returnPtr) const {
1743 if (appendK == ConcatFnKind::none)
1744 CurrentFunctionDescription = "string copy function";
1745 else
1746 CurrentFunctionDescription = "string concatenation function";
1747
1748 ProgramStateRef state = C.getState();
1749 const LocationContext *LCtx = C.getLocationContext();
1750
1751 // Check that the destination is non-null.
1752 DestinationArgExpr Dst = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
1753 SVal DstVal = state->getSVal(Dst.Expression, LCtx);
1754 state = checkNonNull(C, State: state, Arg: Dst, l: DstVal);
1755 if (!state)
1756 return;
1757
1758 // Check that the source is non-null.
1759 SourceArgExpr srcExpr = {{.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1}};
1760 SVal srcVal = state->getSVal(srcExpr.Expression, LCtx);
1761 state = checkNonNull(C, State: state, Arg: srcExpr, l: srcVal);
1762 if (!state)
1763 return;
1764
1765 // Get the string length of the source.
1766 SVal strLength = getCStringLength(C, state, Ex: srcExpr.Expression, Buf: srcVal);
1767 std::optional<NonLoc> strLengthNL = strLength.getAs<NonLoc>();
1768
1769 // Get the string length of the destination buffer.
1770 SVal dstStrLength = getCStringLength(C, state, Ex: Dst.Expression, Buf: DstVal);
1771 std::optional<NonLoc> dstStrLengthNL = dstStrLength.getAs<NonLoc>();
1772
1773 // If the source isn't a valid C string, give up.
1774 if (strLength.isUndef())
1775 return;
1776
1777 SValBuilder &svalBuilder = C.getSValBuilder();
1778 QualType cmpTy = svalBuilder.getConditionType();
1779 QualType sizeTy = svalBuilder.getContext().getSizeType();
1780
1781 // These two values allow checking two kinds of errors:
1782 // - actual overflows caused by a source that doesn't fit in the destination
1783 // - potential overflows caused by a bound that could exceed the destination
1784 SVal amountCopied = UnknownVal();
1785 SVal maxLastElementIndex = UnknownVal();
1786 const char *boundWarning = nullptr;
1787
1788 // FIXME: Why do we choose the srcExpr if the access has no size?
1789 // Note that the 3rd argument of the call would be the size parameter.
1790 SizeArgExpr SrcExprAsSizeDummy = {
1791 {.Expression: srcExpr.Expression, .ArgumentIndex: srcExpr.ArgumentIndex}};
1792 state = CheckOverlap(
1793 C, state,
1794 Size: (IsBounded ? SizeArgExpr{{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}} : SrcExprAsSizeDummy),
1795 First: Dst, Second: srcExpr);
1796
1797 if (!state)
1798 return;
1799
1800 // If the function is strncpy, strncat, etc... it is bounded.
1801 if (IsBounded) {
1802 // Get the max number of characters to copy.
1803 SizeArgExpr lenExpr = {{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}};
1804 SVal lenVal = state->getSVal(lenExpr.Expression, LCtx);
1805
1806 // Protect against misdeclared strncpy().
1807 lenVal =
1808 svalBuilder.evalCast(V: lenVal, CastTy: sizeTy, OriginalTy: lenExpr.Expression->getType());
1809
1810 std::optional<NonLoc> lenValNL = lenVal.getAs<NonLoc>();
1811
1812 // If we know both values, we might be able to figure out how much
1813 // we're copying.
1814 if (strLengthNL && lenValNL) {
1815 switch (appendK) {
1816 case ConcatFnKind::none:
1817 case ConcatFnKind::strcat: {
1818 ProgramStateRef stateSourceTooLong, stateSourceNotTooLong;
1819 // Check if the max number to copy is less than the length of the src.
1820 // If the bound is equal to the source length, strncpy won't null-
1821 // terminate the result!
1822 std::tie(args&: stateSourceTooLong, args&: stateSourceNotTooLong) = state->assume(
1823 Cond: svalBuilder
1824 .evalBinOpNN(state, op: BO_GE, lhs: *strLengthNL, rhs: *lenValNL, resultTy: cmpTy)
1825 .castAs<DefinedOrUnknownSVal>());
1826
1827 if (stateSourceTooLong && !stateSourceNotTooLong) {
1828 // Max number to copy is less than the length of the src, so the
1829 // actual strLength copied is the max number arg.
1830 state = stateSourceTooLong;
1831 amountCopied = lenVal;
1832
1833 } else if (!stateSourceTooLong && stateSourceNotTooLong) {
1834 // The source buffer entirely fits in the bound.
1835 state = stateSourceNotTooLong;
1836 amountCopied = strLength;
1837 }
1838 break;
1839 }
1840 case ConcatFnKind::strlcat:
1841 if (!dstStrLengthNL)
1842 return;
1843
1844 // amountCopied = min (size - dstLen - 1 , srcLen)
1845 SVal freeSpace = svalBuilder.evalBinOpNN(state, op: BO_Sub, lhs: *lenValNL,
1846 rhs: *dstStrLengthNL, resultTy: sizeTy);
1847 if (!isa<NonLoc>(Val: freeSpace))
1848 return;
1849 freeSpace =
1850 svalBuilder.evalBinOp(state, op: BO_Sub, lhs: freeSpace,
1851 rhs: svalBuilder.makeIntVal(integer: 1, type: sizeTy), type: sizeTy);
1852 std::optional<NonLoc> freeSpaceNL = freeSpace.getAs<NonLoc>();
1853
1854 // While unlikely, it is possible that the subtraction is
1855 // too complex to compute, let's check whether it succeeded.
1856 if (!freeSpaceNL)
1857 return;
1858 SVal hasEnoughSpace = svalBuilder.evalBinOpNN(
1859 state, op: BO_LE, lhs: *strLengthNL, rhs: *freeSpaceNL, resultTy: cmpTy);
1860
1861 ProgramStateRef TrueState, FalseState;
1862 std::tie(args&: TrueState, args&: FalseState) =
1863 state->assume(Cond: hasEnoughSpace.castAs<DefinedOrUnknownSVal>());
1864
1865 // srcStrLength <= size - dstStrLength -1
1866 if (TrueState && !FalseState) {
1867 amountCopied = strLength;
1868 }
1869
1870 // srcStrLength > size - dstStrLength -1
1871 if (!TrueState && FalseState) {
1872 amountCopied = freeSpace;
1873 }
1874
1875 if (TrueState && FalseState)
1876 amountCopied = UnknownVal();
1877 break;
1878 }
1879 }
1880 // We still want to know if the bound is known to be too large.
1881 if (lenValNL) {
1882 switch (appendK) {
1883 case ConcatFnKind::strcat:
1884 // For strncat, the check is strlen(dst) + lenVal < sizeof(dst)
1885
1886 // Get the string length of the destination. If the destination is
1887 // memory that can't have a string length, we shouldn't be copying
1888 // into it anyway.
1889 if (dstStrLength.isUndef())
1890 return;
1891
1892 if (dstStrLengthNL) {
1893 maxLastElementIndex = svalBuilder.evalBinOpNN(
1894 state, op: BO_Add, lhs: *lenValNL, rhs: *dstStrLengthNL, resultTy: sizeTy);
1895
1896 boundWarning = "Size argument is greater than the free space in the "
1897 "destination buffer";
1898 }
1899 break;
1900 case ConcatFnKind::none:
1901 case ConcatFnKind::strlcat:
1902 // For strncpy and strlcat, this is just checking
1903 // that lenVal <= sizeof(dst).
1904 // (Yes, strncpy and strncat differ in how they treat termination.
1905 // strncat ALWAYS terminates, but strncpy doesn't.)
1906
1907 // We need a special case for when the copy size is zero, in which
1908 // case strncpy will do no work at all. Our bounds check uses n-1
1909 // as the last element accessed, so n == 0 is problematic.
1910 ProgramStateRef StateZeroSize, StateNonZeroSize;
1911 std::tie(args&: StateZeroSize, args&: StateNonZeroSize) =
1912 assumeZero(C, state, V: *lenValNL, Ty: sizeTy);
1913
1914 // If the size is known to be zero, we're done.
1915 if (StateZeroSize && !StateNonZeroSize) {
1916 if (returnPtr) {
1917 StateZeroSize =
1918 StateZeroSize->BindExpr(Call.getOriginExpr(), LCtx, DstVal);
1919 } else {
1920 if (appendK == ConcatFnKind::none) {
1921 // strlcpy returns strlen(src)
1922 StateZeroSize = StateZeroSize->BindExpr(Call.getOriginExpr(),
1923 LCtx, strLength);
1924 } else {
1925 // strlcat returns strlen(src) + strlen(dst)
1926 SVal retSize = svalBuilder.evalBinOp(
1927 state, op: BO_Add, lhs: strLength, rhs: dstStrLength, type: sizeTy);
1928 StateZeroSize =
1929 StateZeroSize->BindExpr(Call.getOriginExpr(), LCtx, retSize);
1930 }
1931 }
1932 C.addTransition(State: StateZeroSize);
1933 return;
1934 }
1935
1936 // Otherwise, go ahead and figure out the last element we'll touch.
1937 // We don't record the non-zero assumption here because we can't
1938 // be sure. We won't warn on a possible zero.
1939 NonLoc one = svalBuilder.makeIntVal(integer: 1, type: sizeTy).castAs<NonLoc>();
1940 maxLastElementIndex =
1941 svalBuilder.evalBinOpNN(state, op: BO_Sub, lhs: *lenValNL, rhs: one, resultTy: sizeTy);
1942 boundWarning = "Size argument is greater than the length of the "
1943 "destination buffer";
1944 break;
1945 }
1946 }
1947 } else {
1948 // The function isn't bounded. The amount copied should match the length
1949 // of the source buffer.
1950 amountCopied = strLength;
1951 }
1952
1953 assert(state);
1954
1955 // This represents the number of characters copied into the destination
1956 // buffer. (It may not actually be the strlen if the destination buffer
1957 // is not terminated.)
1958 SVal finalStrLength = UnknownVal();
1959 SVal strlRetVal = UnknownVal();
1960
1961 if (appendK == ConcatFnKind::none && !returnPtr) {
1962 // strlcpy returns the sizeof(src)
1963 strlRetVal = strLength;
1964 }
1965
1966 // If this is an appending function (strcat, strncat...) then set the
1967 // string length to strlen(src) + strlen(dst) since the buffer will
1968 // ultimately contain both.
1969 if (appendK != ConcatFnKind::none) {
1970 // Get the string length of the destination. If the destination is memory
1971 // that can't have a string length, we shouldn't be copying into it anyway.
1972 if (dstStrLength.isUndef())
1973 return;
1974
1975 if (appendK == ConcatFnKind::strlcat && dstStrLengthNL && strLengthNL) {
1976 strlRetVal = svalBuilder.evalBinOpNN(state, op: BO_Add, lhs: *strLengthNL,
1977 rhs: *dstStrLengthNL, resultTy: sizeTy);
1978 }
1979
1980 std::optional<NonLoc> amountCopiedNL = amountCopied.getAs<NonLoc>();
1981
1982 // If we know both string lengths, we might know the final string length.
1983 if (amountCopiedNL && dstStrLengthNL) {
1984 // Make sure the two lengths together don't overflow a size_t.
1985 state = checkAdditionOverflow(C, state, left: *amountCopiedNL, right: *dstStrLengthNL);
1986 if (!state)
1987 return;
1988
1989 finalStrLength = svalBuilder.evalBinOpNN(state, op: BO_Add, lhs: *amountCopiedNL,
1990 rhs: *dstStrLengthNL, resultTy: sizeTy);
1991 }
1992
1993 // If we couldn't get a single value for the final string length,
1994 // we can at least bound it by the individual lengths.
1995 if (finalStrLength.isUnknown()) {
1996 // Try to get a "hypothetical" string length symbol, which we can later
1997 // set as a real value if that turns out to be the case.
1998 finalStrLength =
1999 getCStringLength(C, state, Ex: Call.getOriginExpr(), Buf: DstVal, hypothetical: true);
2000 assert(!finalStrLength.isUndef());
2001
2002 if (std::optional<NonLoc> finalStrLengthNL =
2003 finalStrLength.getAs<NonLoc>()) {
2004 if (amountCopiedNL && appendK == ConcatFnKind::none) {
2005 // we overwrite dst string with the src
2006 // finalStrLength >= srcStrLength
2007 SVal sourceInResult = svalBuilder.evalBinOpNN(
2008 state, op: BO_GE, lhs: *finalStrLengthNL, rhs: *amountCopiedNL, resultTy: cmpTy);
2009 state = state->assume(Cond: sourceInResult.castAs<DefinedOrUnknownSVal>(),
2010 Assumption: true);
2011 if (!state)
2012 return;
2013 }
2014
2015 if (dstStrLengthNL && appendK != ConcatFnKind::none) {
2016 // we extend the dst string with the src
2017 // finalStrLength >= dstStrLength
2018 SVal destInResult = svalBuilder.evalBinOpNN(state, op: BO_GE,
2019 lhs: *finalStrLengthNL,
2020 rhs: *dstStrLengthNL,
2021 resultTy: cmpTy);
2022 state =
2023 state->assume(Cond: destInResult.castAs<DefinedOrUnknownSVal>(), Assumption: true);
2024 if (!state)
2025 return;
2026 }
2027 }
2028 }
2029
2030 } else {
2031 // Otherwise, this is a copy-over function (strcpy, strncpy, ...), and
2032 // the final string length will match the input string length.
2033 finalStrLength = amountCopied;
2034 }
2035
2036 SVal Result;
2037
2038 if (returnPtr) {
2039 // The final result of the function will either be a pointer past the last
2040 // copied element, or a pointer to the start of the destination buffer.
2041 Result = (ReturnEnd ? UnknownVal() : DstVal);
2042 } else {
2043 if (appendK == ConcatFnKind::strlcat || appendK == ConcatFnKind::none)
2044 //strlcpy, strlcat
2045 Result = strlRetVal;
2046 else
2047 Result = finalStrLength;
2048 }
2049
2050 assert(state);
2051
2052 // If the destination is a MemRegion, try to check for a buffer overflow and
2053 // record the new string length.
2054 if (std::optional<loc::MemRegionVal> dstRegVal =
2055 DstVal.getAs<loc::MemRegionVal>()) {
2056 QualType ptrTy = Dst.Expression->getType();
2057
2058 // If we have an exact value on a bounded copy, use that to check for
2059 // overflows, rather than our estimate about how much is actually copied.
2060 if (std::optional<NonLoc> maxLastNL = maxLastElementIndex.getAs<NonLoc>()) {
2061 SVal maxLastElement =
2062 svalBuilder.evalBinOpLN(state, op: BO_Add, lhs: *dstRegVal, rhs: *maxLastNL, resultTy: ptrTy);
2063
2064 // Check if the first byte of the destination is writable.
2065 state = CheckLocation(C, state, Buffer: Dst, Element: DstVal, Access: AccessKind::write);
2066 if (!state)
2067 return;
2068 // Check if the last byte of the destination is writable.
2069 state = CheckLocation(C, state, Buffer: Dst, Element: maxLastElement, Access: AccessKind::write);
2070 if (!state)
2071 return;
2072 }
2073
2074 // Then, if the final length is known...
2075 if (std::optional<NonLoc> knownStrLength = finalStrLength.getAs<NonLoc>()) {
2076 SVal lastElement = svalBuilder.evalBinOpLN(state, op: BO_Add, lhs: *dstRegVal,
2077 rhs: *knownStrLength, resultTy: ptrTy);
2078
2079 // ...and we haven't checked the bound, we'll check the actual copy.
2080 if (!boundWarning) {
2081 // Check if the first byte of the destination is writable.
2082 state = CheckLocation(C, state, Buffer: Dst, Element: DstVal, Access: AccessKind::write);
2083 if (!state)
2084 return;
2085 // Check if the last byte of the destination is writable.
2086 state = CheckLocation(C, state, Buffer: Dst, Element: lastElement, Access: AccessKind::write);
2087 if (!state)
2088 return;
2089 }
2090
2091 // If this is a stpcpy-style copy, the last element is the return value.
2092 if (returnPtr && ReturnEnd)
2093 Result = lastElement;
2094 }
2095
2096 // Invalidate the destination (regular invalidation without pointer-escaping
2097 // the address of the top-level region). This must happen before we set the
2098 // C string length because invalidation will clear the length.
2099 // FIXME: Even if we can't perfectly model the copy, we should see if we
2100 // can use LazyCompoundVals to copy the source values into the destination.
2101 // This would probably remove any existing bindings past the end of the
2102 // string, but that's still an improvement over blank invalidation.
2103 state = invalidateDestinationBufferBySize(C, S: state, BufE: Dst.Expression,
2104 BufV: *dstRegVal, SizeV: amountCopied,
2105 SizeTy: C.getASTContext().getSizeType());
2106
2107 // Invalidate the source (const-invalidation without const-pointer-escaping
2108 // the address of the top-level region).
2109 state = invalidateSourceBuffer(C, S: state, BufE: srcExpr.Expression, BufV: srcVal);
2110
2111 // Set the C string length of the destination, if we know it.
2112 if (IsBounded && (appendK == ConcatFnKind::none)) {
2113 // strncpy is annoying in that it doesn't guarantee to null-terminate
2114 // the result string. If the original string didn't fit entirely inside
2115 // the bound (including the null-terminator), we don't know how long the
2116 // result is.
2117 if (amountCopied != strLength)
2118 finalStrLength = UnknownVal();
2119 }
2120 state = setCStringLength(state, MR: dstRegVal->getRegion(), strLength: finalStrLength);
2121 }
2122
2123 assert(state);
2124
2125 if (returnPtr) {
2126 // If this is a stpcpy-style copy, but we were unable to check for a buffer
2127 // overflow, we still need a result. Conjure a return value.
2128 if (ReturnEnd && Result.isUnknown()) {
2129 Result = svalBuilder.conjureSymbolVal(symbolTag: nullptr, expr: Call.getOriginExpr(), LCtx,
2130 count: C.blockCount());
2131 }
2132 }
2133 // Set the return value.
2134 state = state->BindExpr(Call.getOriginExpr(), LCtx, Result);
2135 C.addTransition(State: state);
2136}
2137
2138void CStringChecker::evalStrcmp(CheckerContext &C,
2139 const CallEvent &Call) const {
2140 //int strcmp(const char *s1, const char *s2);
2141 evalStrcmpCommon(C, Call, /* IsBounded = */ false, /* IgnoreCase = */ false);
2142}
2143
2144void CStringChecker::evalStrncmp(CheckerContext &C,
2145 const CallEvent &Call) const {
2146 //int strncmp(const char *s1, const char *s2, size_t n);
2147 evalStrcmpCommon(C, Call, /* IsBounded = */ true, /* IgnoreCase = */ false);
2148}
2149
2150void CStringChecker::evalStrcasecmp(CheckerContext &C,
2151 const CallEvent &Call) const {
2152 //int strcasecmp(const char *s1, const char *s2);
2153 evalStrcmpCommon(C, Call, /* IsBounded = */ false, /* IgnoreCase = */ true);
2154}
2155
2156void CStringChecker::evalStrncasecmp(CheckerContext &C,
2157 const CallEvent &Call) const {
2158 //int strncasecmp(const char *s1, const char *s2, size_t n);
2159 evalStrcmpCommon(C, Call, /* IsBounded = */ true, /* IgnoreCase = */ true);
2160}
2161
2162void CStringChecker::evalStrcmpCommon(CheckerContext &C, const CallEvent &Call,
2163 bool IsBounded, bool IgnoreCase) const {
2164 CurrentFunctionDescription = "string comparison function";
2165 ProgramStateRef state = C.getState();
2166 const LocationContext *LCtx = C.getLocationContext();
2167
2168 // Check that the first string is non-null
2169 AnyArgExpr Left = {.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0};
2170 SVal LeftVal = state->getSVal(Left.Expression, LCtx);
2171 state = checkNonNull(C, State: state, Arg: Left, l: LeftVal);
2172 if (!state)
2173 return;
2174
2175 // Check that the second string is non-null.
2176 AnyArgExpr Right = {.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1};
2177 SVal RightVal = state->getSVal(Right.Expression, LCtx);
2178 state = checkNonNull(C, State: state, Arg: Right, l: RightVal);
2179 if (!state)
2180 return;
2181
2182 // Get the string length of the first string or give up.
2183 SVal LeftLength = getCStringLength(C, state, Ex: Left.Expression, Buf: LeftVal);
2184 if (LeftLength.isUndef())
2185 return;
2186
2187 // Get the string length of the second string or give up.
2188 SVal RightLength = getCStringLength(C, state, Ex: Right.Expression, Buf: RightVal);
2189 if (RightLength.isUndef())
2190 return;
2191
2192 // If we know the two buffers are the same, we know the result is 0.
2193 // First, get the two buffers' addresses. Another checker will have already
2194 // made sure they're not undefined.
2195 DefinedOrUnknownSVal LV = LeftVal.castAs<DefinedOrUnknownSVal>();
2196 DefinedOrUnknownSVal RV = RightVal.castAs<DefinedOrUnknownSVal>();
2197
2198 // See if they are the same.
2199 SValBuilder &svalBuilder = C.getSValBuilder();
2200 DefinedOrUnknownSVal SameBuf = svalBuilder.evalEQ(state, lhs: LV, rhs: RV);
2201 ProgramStateRef StSameBuf, StNotSameBuf;
2202 std::tie(args&: StSameBuf, args&: StNotSameBuf) = state->assume(Cond: SameBuf);
2203
2204 // If the two arguments might be the same buffer, we know the result is 0,
2205 // and we only need to check one size.
2206 if (StSameBuf) {
2207 StSameBuf =
2208 StSameBuf->BindExpr(Call.getOriginExpr(), LCtx,
2209 svalBuilder.makeZeroVal(type: Call.getResultType()));
2210 C.addTransition(State: StSameBuf);
2211
2212 // If the two arguments are GUARANTEED to be the same, we're done!
2213 if (!StNotSameBuf)
2214 return;
2215 }
2216
2217 assert(StNotSameBuf);
2218 state = StNotSameBuf;
2219
2220 // At this point we can go about comparing the two buffers.
2221 // For now, we only do this if they're both known string literals.
2222
2223 // Attempt to extract string literals from both expressions.
2224 const StringLiteral *LeftStrLiteral =
2225 getCStringLiteral(C, state, expr: Left.Expression, val: LeftVal);
2226 const StringLiteral *RightStrLiteral =
2227 getCStringLiteral(C, state, expr: Right.Expression, val: RightVal);
2228 bool canComputeResult = false;
2229 SVal resultVal = svalBuilder.conjureSymbolVal(symbolTag: nullptr, expr: Call.getOriginExpr(),
2230 LCtx, count: C.blockCount());
2231
2232 if (LeftStrLiteral && RightStrLiteral) {
2233 StringRef LeftStrRef = LeftStrLiteral->getString();
2234 StringRef RightStrRef = RightStrLiteral->getString();
2235
2236 if (IsBounded) {
2237 // Get the max number of characters to compare.
2238 const Expr *lenExpr = Call.getArgExpr(Index: 2);
2239 SVal lenVal = state->getSVal(lenExpr, LCtx);
2240
2241 // If the length is known, we can get the right substrings.
2242 if (const llvm::APSInt *len = svalBuilder.getKnownValue(state, val: lenVal)) {
2243 // Create substrings of each to compare the prefix.
2244 LeftStrRef = LeftStrRef.substr(Start: 0, N: (size_t)len->getZExtValue());
2245 RightStrRef = RightStrRef.substr(Start: 0, N: (size_t)len->getZExtValue());
2246 canComputeResult = true;
2247 }
2248 } else {
2249 // This is a normal, unbounded strcmp.
2250 canComputeResult = true;
2251 }
2252
2253 if (canComputeResult) {
2254 // Real strcmp stops at null characters.
2255 size_t s1Term = LeftStrRef.find(C: '\0');
2256 if (s1Term != StringRef::npos)
2257 LeftStrRef = LeftStrRef.substr(Start: 0, N: s1Term);
2258
2259 size_t s2Term = RightStrRef.find(C: '\0');
2260 if (s2Term != StringRef::npos)
2261 RightStrRef = RightStrRef.substr(Start: 0, N: s2Term);
2262
2263 // Use StringRef's comparison methods to compute the actual result.
2264 int compareRes = IgnoreCase ? LeftStrRef.compare_insensitive(RHS: RightStrRef)
2265 : LeftStrRef.compare(RHS: RightStrRef);
2266
2267 // The strcmp function returns an integer greater than, equal to, or less
2268 // than zero, [c11, p7.24.4.2].
2269 if (compareRes == 0) {
2270 resultVal = svalBuilder.makeIntVal(integer: compareRes, type: Call.getResultType());
2271 }
2272 else {
2273 DefinedSVal zeroVal = svalBuilder.makeIntVal(integer: 0, type: Call.getResultType());
2274 // Constrain strcmp's result range based on the result of StringRef's
2275 // comparison methods.
2276 BinaryOperatorKind op = (compareRes > 0) ? BO_GT : BO_LT;
2277 SVal compareWithZero =
2278 svalBuilder.evalBinOp(state, op, lhs: resultVal, rhs: zeroVal,
2279 type: svalBuilder.getConditionType());
2280 DefinedSVal compareWithZeroVal = compareWithZero.castAs<DefinedSVal>();
2281 state = state->assume(Cond: compareWithZeroVal, Assumption: true);
2282 }
2283 }
2284 }
2285
2286 state = state->BindExpr(Call.getOriginExpr(), LCtx, resultVal);
2287
2288 // Record this as a possible path.
2289 C.addTransition(State: state);
2290}
2291
2292void CStringChecker::evalStrsep(CheckerContext &C,
2293 const CallEvent &Call) const {
2294 // char *strsep(char **stringp, const char *delim);
2295 // Verify whether the search string parameter matches the return type.
2296 SourceArgExpr SearchStrPtr = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
2297
2298 QualType CharPtrTy = SearchStrPtr.Expression->getType()->getPointeeType();
2299 if (CharPtrTy.isNull() || Call.getResultType().getUnqualifiedType() !=
2300 CharPtrTy.getUnqualifiedType())
2301 return;
2302
2303 CurrentFunctionDescription = "strsep()";
2304 ProgramStateRef State = C.getState();
2305 const LocationContext *LCtx = C.getLocationContext();
2306
2307 // Check that the search string pointer is non-null (though it may point to
2308 // a null string).
2309 SVal SearchStrVal = State->getSVal(SearchStrPtr.Expression, LCtx);
2310 State = checkNonNull(C, State, Arg: SearchStrPtr, l: SearchStrVal);
2311 if (!State)
2312 return;
2313
2314 // Check that the delimiter string is non-null.
2315 AnyArgExpr DelimStr = {.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1};
2316 SVal DelimStrVal = State->getSVal(DelimStr.Expression, LCtx);
2317 State = checkNonNull(C, State, Arg: DelimStr, l: DelimStrVal);
2318 if (!State)
2319 return;
2320
2321 SValBuilder &SVB = C.getSValBuilder();
2322 SVal Result;
2323 if (std::optional<Loc> SearchStrLoc = SearchStrVal.getAs<Loc>()) {
2324 // Get the current value of the search string pointer, as a char*.
2325 Result = State->getSVal(LV: *SearchStrLoc, T: CharPtrTy);
2326
2327 // Invalidate the search string, representing the change of one delimiter
2328 // character to NUL.
2329 // As the replacement never overflows, do not invalidate its super region.
2330 State = invalidateDestinationBufferNeverOverflows(
2331 C, S: State, BufE: SearchStrPtr.Expression, BufV: Result);
2332
2333 // Overwrite the search string pointer. The new value is either an address
2334 // further along in the same string, or NULL if there are no more tokens.
2335 State =
2336 State->bindLoc(location: *SearchStrLoc,
2337 V: SVB.conjureSymbolVal(symbolTag: getTag(), expr: Call.getOriginExpr(),
2338 LCtx, type: CharPtrTy, count: C.blockCount()),
2339 LCtx);
2340 } else {
2341 assert(SearchStrVal.isUnknown());
2342 // Conjure a symbolic value. It's the best we can do.
2343 Result = SVB.conjureSymbolVal(symbolTag: nullptr, expr: Call.getOriginExpr(), LCtx,
2344 count: C.blockCount());
2345 }
2346
2347 // Set the return value, and finish.
2348 State = State->BindExpr(Call.getOriginExpr(), LCtx, Result);
2349 C.addTransition(State);
2350}
2351
2352// These should probably be moved into a C++ standard library checker.
2353void CStringChecker::evalStdCopy(CheckerContext &C,
2354 const CallEvent &Call) const {
2355 evalStdCopyCommon(C, Call);
2356}
2357
2358void CStringChecker::evalStdCopyBackward(CheckerContext &C,
2359 const CallEvent &Call) const {
2360 evalStdCopyCommon(C, Call);
2361}
2362
2363void CStringChecker::evalStdCopyCommon(CheckerContext &C,
2364 const CallEvent &Call) const {
2365 if (!Call.getArgExpr(Index: 2)->getType()->isPointerType())
2366 return;
2367
2368 ProgramStateRef State = C.getState();
2369
2370 const LocationContext *LCtx = C.getLocationContext();
2371
2372 // template <class _InputIterator, class _OutputIterator>
2373 // _OutputIterator
2374 // copy(_InputIterator __first, _InputIterator __last,
2375 // _OutputIterator __result)
2376
2377 // Invalidate the destination buffer
2378 const Expr *Dst = Call.getArgExpr(Index: 2);
2379 SVal DstVal = State->getSVal(Dst, LCtx);
2380 // FIXME: As we do not know how many items are copied, we also invalidate the
2381 // super region containing the target location.
2382 State =
2383 invalidateDestinationBufferAlwaysEscapeSuperRegion(C, S: State, BufE: Dst, BufV: DstVal);
2384
2385 SValBuilder &SVB = C.getSValBuilder();
2386
2387 SVal ResultVal =
2388 SVB.conjureSymbolVal(symbolTag: nullptr, expr: Call.getOriginExpr(), LCtx, count: C.blockCount());
2389 State = State->BindExpr(Call.getOriginExpr(), LCtx, ResultVal);
2390
2391 C.addTransition(State);
2392}
2393
2394void CStringChecker::evalMemset(CheckerContext &C,
2395 const CallEvent &Call) const {
2396 // void *memset(void *s, int c, size_t n);
2397 CurrentFunctionDescription = "memory set function";
2398
2399 DestinationArgExpr Buffer = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
2400 AnyArgExpr CharE = {.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1};
2401 SizeArgExpr Size = {{.Expression: Call.getArgExpr(Index: 2), .ArgumentIndex: 2}};
2402
2403 ProgramStateRef State = C.getState();
2404
2405 // See if the size argument is zero.
2406 const LocationContext *LCtx = C.getLocationContext();
2407 SVal SizeVal = C.getSVal(Size.Expression);
2408 QualType SizeTy = Size.Expression->getType();
2409
2410 ProgramStateRef ZeroSize, NonZeroSize;
2411 std::tie(args&: ZeroSize, args&: NonZeroSize) = assumeZero(C, state: State, V: SizeVal, Ty: SizeTy);
2412
2413 // Get the value of the memory area.
2414 SVal BufferPtrVal = C.getSVal(Buffer.Expression);
2415
2416 // If the size is zero, there won't be any actual memory access, so
2417 // just bind the return value to the buffer and return.
2418 if (ZeroSize && !NonZeroSize) {
2419 ZeroSize = ZeroSize->BindExpr(Call.getOriginExpr(), LCtx, BufferPtrVal);
2420 C.addTransition(State: ZeroSize);
2421 return;
2422 }
2423
2424 // Ensure the memory area is not null.
2425 // If it is NULL there will be a NULL pointer dereference.
2426 State = checkNonNull(C, State: NonZeroSize, Arg: Buffer, l: BufferPtrVal);
2427 if (!State)
2428 return;
2429
2430 State = CheckBufferAccess(C, State, Buffer, Size, Access: AccessKind::write);
2431 if (!State)
2432 return;
2433
2434 // According to the values of the arguments, bind the value of the second
2435 // argument to the destination buffer and set string length, or just
2436 // invalidate the destination buffer.
2437 if (!memsetAux(DstBuffer: Buffer.Expression, CharVal: C.getSVal(CharE.Expression),
2438 Size: Size.Expression, C, State))
2439 return;
2440
2441 State = State->BindExpr(Call.getOriginExpr(), LCtx, BufferPtrVal);
2442 C.addTransition(State);
2443}
2444
2445void CStringChecker::evalBzero(CheckerContext &C, const CallEvent &Call) const {
2446 CurrentFunctionDescription = "memory clearance function";
2447
2448 DestinationArgExpr Buffer = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
2449 SizeArgExpr Size = {{.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1}};
2450 SVal Zero = C.getSValBuilder().makeZeroVal(type: C.getASTContext().IntTy);
2451
2452 ProgramStateRef State = C.getState();
2453
2454 // See if the size argument is zero.
2455 SVal SizeVal = C.getSVal(Size.Expression);
2456 QualType SizeTy = Size.Expression->getType();
2457
2458 ProgramStateRef StateZeroSize, StateNonZeroSize;
2459 std::tie(args&: StateZeroSize, args&: StateNonZeroSize) =
2460 assumeZero(C, state: State, V: SizeVal, Ty: SizeTy);
2461
2462 // If the size is zero, there won't be any actual memory access,
2463 // In this case we just return.
2464 if (StateZeroSize && !StateNonZeroSize) {
2465 C.addTransition(State: StateZeroSize);
2466 return;
2467 }
2468
2469 // Get the value of the memory area.
2470 SVal MemVal = C.getSVal(Buffer.Expression);
2471
2472 // Ensure the memory area is not null.
2473 // If it is NULL there will be a NULL pointer dereference.
2474 State = checkNonNull(C, State: StateNonZeroSize, Arg: Buffer, l: MemVal);
2475 if (!State)
2476 return;
2477
2478 State = CheckBufferAccess(C, State, Buffer, Size, Access: AccessKind::write);
2479 if (!State)
2480 return;
2481
2482 if (!memsetAux(DstBuffer: Buffer.Expression, CharVal: Zero, Size: Size.Expression, C, State))
2483 return;
2484
2485 C.addTransition(State);
2486}
2487
2488void CStringChecker::evalSprintf(CheckerContext &C,
2489 const CallEvent &Call) const {
2490 CurrentFunctionDescription = "'sprintf'";
2491 evalSprintfCommon(C, Call, /* IsBounded = */ false);
2492}
2493
2494void CStringChecker::evalSnprintf(CheckerContext &C,
2495 const CallEvent &Call) const {
2496 CurrentFunctionDescription = "'snprintf'";
2497 evalSprintfCommon(C, Call, /* IsBounded = */ true);
2498}
2499
2500void CStringChecker::evalSprintfCommon(CheckerContext &C, const CallEvent &Call,
2501 bool IsBounded) const {
2502 ProgramStateRef State = C.getState();
2503 const auto *CE = cast<CallExpr>(Val: Call.getOriginExpr());
2504 DestinationArgExpr Dest = {{.Expression: Call.getArgExpr(Index: 0), .ArgumentIndex: 0}};
2505
2506 const auto NumParams = Call.parameters().size();
2507 if (CE->getNumArgs() < NumParams) {
2508 // This is an invalid call, let's just ignore it.
2509 return;
2510 }
2511
2512 const auto AllArguments =
2513 llvm::make_range(x: CE->getArgs(), y: CE->getArgs() + CE->getNumArgs());
2514 const auto VariadicArguments = drop_begin(RangeOrContainer: enumerate(First: AllArguments), N: NumParams);
2515
2516 for (const auto &[ArgIdx, ArgExpr] : VariadicArguments) {
2517 // We consider only string buffers
2518 if (const QualType type = ArgExpr->getType();
2519 !type->isAnyPointerType() ||
2520 !type->getPointeeType()->isAnyCharacterType())
2521 continue;
2522 SourceArgExpr Source = {{.Expression: ArgExpr, .ArgumentIndex: unsigned(ArgIdx)}};
2523
2524 // Ensure the buffers do not overlap.
2525 SizeArgExpr SrcExprAsSizeDummy = {
2526 {.Expression: Source.Expression, .ArgumentIndex: Source.ArgumentIndex}};
2527 State = CheckOverlap(
2528 C, state: State,
2529 Size: (IsBounded ? SizeArgExpr{{.Expression: Call.getArgExpr(Index: 1), .ArgumentIndex: 1}} : SrcExprAsSizeDummy),
2530 First: Dest, Second: Source);
2531 if (!State)
2532 return;
2533 }
2534
2535 C.addTransition(State);
2536}
2537
2538//===----------------------------------------------------------------------===//
2539// The driver method, and other Checker callbacks.
2540//===----------------------------------------------------------------------===//
2541
2542CStringChecker::FnCheck CStringChecker::identifyCall(const CallEvent &Call,
2543 CheckerContext &C) const {
2544 const auto *CE = dyn_cast_or_null<CallExpr>(Val: Call.getOriginExpr());
2545 if (!CE)
2546 return nullptr;
2547
2548 const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(Val: Call.getDecl());
2549 if (!FD)
2550 return nullptr;
2551
2552 if (StdCopy.matches(Call))
2553 return &CStringChecker::evalStdCopy;
2554 if (StdCopyBackward.matches(Call))
2555 return &CStringChecker::evalStdCopyBackward;
2556
2557 // Pro-actively check that argument types are safe to do arithmetic upon.
2558 // We do not want to crash if someone accidentally passes a structure
2559 // into, say, a C++ overload of any of these functions. We could not check
2560 // that for std::copy because they may have arguments of other types.
2561 for (auto I : CE->arguments()) {
2562 QualType T = I->getType();
2563 if (!T->isIntegralOrEnumerationType() && !T->isPointerType())
2564 return nullptr;
2565 }
2566
2567 const FnCheck *Callback = Callbacks.lookup(Call);
2568 if (Callback)
2569 return *Callback;
2570
2571 return nullptr;
2572}
2573
2574bool CStringChecker::evalCall(const CallEvent &Call, CheckerContext &C) const {
2575 FnCheck Callback = identifyCall(Call, C);
2576
2577 // If the callee isn't a string function, let another checker handle it.
2578 if (!Callback)
2579 return false;
2580
2581 // Check and evaluate the call.
2582 assert(isa<CallExpr>(Call.getOriginExpr()));
2583 Callback(this, C, Call);
2584
2585 // If the evaluate call resulted in no change, chain to the next eval call
2586 // handler.
2587 // Note, the custom CString evaluation calls assume that basic safety
2588 // properties are held. However, if the user chooses to turn off some of these
2589 // checks, we ignore the issues and leave the call evaluation to a generic
2590 // handler.
2591 return C.isDifferent();
2592}
2593
2594void CStringChecker::checkPreStmt(const DeclStmt *DS, CheckerContext &C) const {
2595 // Record string length for char a[] = "abc";
2596 ProgramStateRef state = C.getState();
2597
2598 for (const auto *I : DS->decls()) {
2599 const VarDecl *D = dyn_cast<VarDecl>(Val: I);
2600 if (!D)
2601 continue;
2602
2603 // FIXME: Handle array fields of structs.
2604 if (!D->getType()->isArrayType())
2605 continue;
2606
2607 const Expr *Init = D->getInit();
2608 if (!Init)
2609 continue;
2610 if (!isa<StringLiteral>(Val: Init))
2611 continue;
2612
2613 Loc VarLoc = state->getLValue(VD: D, LC: C.getLocationContext());
2614 const MemRegion *MR = VarLoc.getAsRegion();
2615 if (!MR)
2616 continue;
2617
2618 SVal StrVal = C.getSVal(Init);
2619 assert(StrVal.isValid() && "Initializer string is unknown or undefined");
2620 DefinedOrUnknownSVal strLength =
2621 getCStringLength(C, state, Ex: Init, Buf: StrVal).castAs<DefinedOrUnknownSVal>();
2622
2623 state = state->set<CStringLength>(K: MR, E: strLength);
2624 }
2625
2626 C.addTransition(State: state);
2627}
2628
2629ProgramStateRef
2630CStringChecker::checkRegionChanges(ProgramStateRef state,
2631 const InvalidatedSymbols *,
2632 ArrayRef<const MemRegion *> ExplicitRegions,
2633 ArrayRef<const MemRegion *> Regions,
2634 const LocationContext *LCtx,
2635 const CallEvent *Call) const {
2636 CStringLengthTy Entries = state->get<CStringLength>();
2637 if (Entries.isEmpty())
2638 return state;
2639
2640 llvm::SmallPtrSet<const MemRegion *, 8> Invalidated;
2641 llvm::SmallPtrSet<const MemRegion *, 32> SuperRegions;
2642
2643 // First build sets for the changed regions and their super-regions.
2644 for (const MemRegion *MR : Regions) {
2645 Invalidated.insert(Ptr: MR);
2646
2647 SuperRegions.insert(Ptr: MR);
2648 while (const SubRegion *SR = dyn_cast<SubRegion>(Val: MR)) {
2649 MR = SR->getSuperRegion();
2650 SuperRegions.insert(Ptr: MR);
2651 }
2652 }
2653
2654 CStringLengthTy::Factory &F = state->get_context<CStringLength>();
2655
2656 // Then loop over the entries in the current state.
2657 for (const MemRegion *MR : llvm::make_first_range(c&: Entries)) {
2658 // Is this entry for a super-region of a changed region?
2659 if (SuperRegions.count(Ptr: MR)) {
2660 Entries = F.remove(Old: Entries, K: MR);
2661 continue;
2662 }
2663
2664 // Is this entry for a sub-region of a changed region?
2665 const MemRegion *Super = MR;
2666 while (const SubRegion *SR = dyn_cast<SubRegion>(Val: Super)) {
2667 Super = SR->getSuperRegion();
2668 if (Invalidated.count(Ptr: Super)) {
2669 Entries = F.remove(Old: Entries, K: MR);
2670 break;
2671 }
2672 }
2673 }
2674
2675 return state->set<CStringLength>(Entries);
2676}
2677
2678void CStringChecker::checkLiveSymbols(ProgramStateRef state,
2679 SymbolReaper &SR) const {
2680 // Mark all symbols in our string length map as valid.
2681 CStringLengthTy Entries = state->get<CStringLength>();
2682
2683 for (SVal Len : llvm::make_second_range(c&: Entries)) {
2684 for (SymbolRef Sym : Len.symbols())
2685 SR.markInUse(sym: Sym);
2686 }
2687}
2688
2689void CStringChecker::checkDeadSymbols(SymbolReaper &SR,
2690 CheckerContext &C) const {
2691 ProgramStateRef state = C.getState();
2692 CStringLengthTy Entries = state->get<CStringLength>();
2693 if (Entries.isEmpty())
2694 return;
2695
2696 CStringLengthTy::Factory &F = state->get_context<CStringLength>();
2697 for (auto [Reg, Len] : Entries) {
2698 if (SymbolRef Sym = Len.getAsSymbol()) {
2699 if (SR.isDead(sym: Sym))
2700 Entries = F.remove(Old: Entries, K: Reg);
2701 }
2702 }
2703
2704 state = state->set<CStringLength>(Entries);
2705 C.addTransition(State: state);
2706}
2707
2708void ento::registerCStringModeling(CheckerManager &Mgr) {
2709 Mgr.registerChecker<CStringChecker>();
2710}
2711
2712bool ento::shouldRegisterCStringModeling(const CheckerManager &mgr) {
2713 return true;
2714}
2715
2716#define REGISTER_CHECKER(name) \
2717 void ento::register##name(CheckerManager &mgr) { \
2718 CStringChecker *checker = mgr.getChecker<CStringChecker>(); \
2719 checker->Filter.Check##name = true; \
2720 checker->Filter.CheckName##name = mgr.getCurrentCheckerName(); \
2721 } \
2722 \
2723 bool ento::shouldRegister##name(const CheckerManager &mgr) { return true; }
2724
2725REGISTER_CHECKER(CStringNullArg)
2726REGISTER_CHECKER(CStringOutOfBounds)
2727REGISTER_CHECKER(CStringBufferOverlap)
2728REGISTER_CHECKER(CStringNotNullTerm)
2729REGISTER_CHECKER(CStringUninitializedRead)
2730

source code of clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp