clang 19.0.0git
PointerArithChecker.cpp
Go to the documentation of this file.
1//=== PointerArithChecker.cpp - Pointer arithmetic checker -----*- 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 files defines PointerArithChecker, a builtin checker that checks for
10// pointer arithmetic on locations other than array elements.
11//
12//===----------------------------------------------------------------------===//
13
14#include "clang/AST/DeclCXX.h"
15#include "clang/AST/ExprCXX.h"
21#include "llvm/ADT/StringRef.h"
22
23using namespace clang;
24using namespace ento;
25
26namespace {
27enum class AllocKind {
28 SingleObject,
29 Array,
30 Unknown,
31 Reinterpreted // Single object interpreted as an array.
32};
33} // end namespace
34
35namespace llvm {
36template <> struct FoldingSetTrait<AllocKind> {
37 static inline void Profile(AllocKind X, FoldingSetNodeID &ID) {
38 ID.AddInteger(static_cast<int>(X));
39 }
40};
41} // end namespace llvm
42
43namespace {
44class PointerArithChecker
45 : public Checker<
46 check::PreStmt<BinaryOperator>, check::PreStmt<UnaryOperator>,
47 check::PreStmt<ArraySubscriptExpr>, check::PreStmt<CastExpr>,
48 check::PostStmt<CastExpr>, check::PostStmt<CXXNewExpr>,
49 check::PostStmt<CallExpr>, check::DeadSymbols> {
50 AllocKind getKindOfNewOp(const CXXNewExpr *NE, const FunctionDecl *FD) const;
51 const MemRegion *getArrayRegion(const MemRegion *Region, bool &Polymorphic,
52 AllocKind &AKind, CheckerContext &C) const;
53 const MemRegion *getPointedRegion(const MemRegion *Region,
54 CheckerContext &C) const;
55 void reportPointerArithMisuse(const Expr *E, CheckerContext &C,
56 bool PointedNeeded = false) const;
57 void initAllocIdentifiers(ASTContext &C) const;
58
59 const BugType BT_pointerArith{this, "Dangerous pointer arithmetic"};
60 const BugType BT_polyArray{this, "Dangerous pointer arithmetic"};
61 mutable llvm::SmallSet<IdentifierInfo *, 8> AllocFunctions;
62
63public:
64 void checkPreStmt(const UnaryOperator *UOp, CheckerContext &C) const;
65 void checkPreStmt(const BinaryOperator *BOp, CheckerContext &C) const;
66 void checkPreStmt(const ArraySubscriptExpr *SubExpr, CheckerContext &C) const;
67 void checkPreStmt(const CastExpr *CE, CheckerContext &C) const;
68 void checkPostStmt(const CastExpr *CE, CheckerContext &C) const;
69 void checkPostStmt(const CXXNewExpr *NE, CheckerContext &C) const;
70 void checkPostStmt(const CallExpr *CE, CheckerContext &C) const;
71 void checkDeadSymbols(SymbolReaper &SR, CheckerContext &C) const;
72};
73} // end namespace
74
75REGISTER_MAP_WITH_PROGRAMSTATE(RegionState, const MemRegion *, AllocKind)
76
77void PointerArithChecker::checkDeadSymbols(SymbolReaper &SR,
78 CheckerContext &C) const {
79 // TODO: intentional leak. Some information is garbage collected too early,
80 // see http://reviews.llvm.org/D14203 for further information.
81 /*ProgramStateRef State = C.getState();
82 RegionStateTy RegionStates = State->get<RegionState>();
83 for (const MemRegion *Reg: llvm::make_first_range(RegionStates)) {
84 if (!SR.isLiveRegion(Reg))
85 State = State->remove<RegionState>(Reg);
86 }
87 C.addTransition(State);*/
88}
89
90AllocKind PointerArithChecker::getKindOfNewOp(const CXXNewExpr *NE,
91 const FunctionDecl *FD) const {
92 // This checker try not to assume anything about placement and overloaded
93 // new to avoid false positives.
94 if (isa<CXXMethodDecl>(FD))
95 return AllocKind::Unknown;
96 if (FD->getNumParams() != 1 || FD->isVariadic())
97 return AllocKind::Unknown;
98 if (NE->isArray())
99 return AllocKind::Array;
100
101 return AllocKind::SingleObject;
102}
103
104const MemRegion *
105PointerArithChecker::getPointedRegion(const MemRegion *Region,
106 CheckerContext &C) const {
107 assert(Region);
108 ProgramStateRef State = C.getState();
109 SVal S = State->getSVal(Region);
110 return S.getAsRegion();
111}
112
113/// Checks whether a region is the part of an array.
114/// In case there is a derived to base cast above the array element, the
115/// Polymorphic output value is set to true. AKind output value is set to the
116/// allocation kind of the inspected region.
117const MemRegion *PointerArithChecker::getArrayRegion(const MemRegion *Region,
118 bool &Polymorphic,
119 AllocKind &AKind,
120 CheckerContext &C) const {
121 assert(Region);
122 while (const auto *BaseRegion = dyn_cast<CXXBaseObjectRegion>(Region)) {
123 Region = BaseRegion->getSuperRegion();
124 Polymorphic = true;
125 }
126 if (const auto *ElemRegion = dyn_cast<ElementRegion>(Region)) {
127 Region = ElemRegion->getSuperRegion();
128 }
129
130 ProgramStateRef State = C.getState();
131 if (const AllocKind *Kind = State->get<RegionState>(Region)) {
132 AKind = *Kind;
133 if (*Kind == AllocKind::Array)
134 return Region;
135 else
136 return nullptr;
137 }
138 // When the region is symbolic and we do not have any information about it,
139 // assume that this is an array to avoid false positives.
140 if (isa<SymbolicRegion>(Region))
141 return Region;
142
143 // No AllocKind stored and not symbolic, assume that it points to a single
144 // object.
145 return nullptr;
146}
147
148void PointerArithChecker::reportPointerArithMisuse(const Expr *E,
150 bool PointedNeeded) const {
152 if (SR.isInvalid())
153 return;
154
155 ProgramStateRef State = C.getState();
156 const MemRegion *Region = C.getSVal(E).getAsRegion();
157 if (!Region)
158 return;
159 if (PointedNeeded)
160 Region = getPointedRegion(Region, C);
161 if (!Region)
162 return;
163
164 bool IsPolymorphic = false;
165 AllocKind Kind = AllocKind::Unknown;
166 if (const MemRegion *ArrayRegion =
167 getArrayRegion(Region, IsPolymorphic, Kind, C)) {
168 if (!IsPolymorphic)
169 return;
170 if (ExplodedNode *N = C.generateNonFatalErrorNode()) {
171 constexpr llvm::StringLiteral Msg =
172 "Pointer arithmetic on a pointer to base class is dangerous "
173 "because derived and base class may have different size.";
174 auto R = std::make_unique<PathSensitiveBugReport>(BT_polyArray, Msg, N);
175 R->addRange(E->getSourceRange());
176 R->markInteresting(ArrayRegion);
177 C.emitReport(std::move(R));
178 }
179 return;
180 }
181
182 if (Kind == AllocKind::Reinterpreted)
183 return;
184
185 // We might not have enough information about symbolic regions.
186 if (Kind != AllocKind::SingleObject &&
187 Region->getKind() == MemRegion::Kind::SymbolicRegionKind)
188 return;
189
190 if (ExplodedNode *N = C.generateNonFatalErrorNode()) {
191 constexpr llvm::StringLiteral Msg =
192 "Pointer arithmetic on non-array variables relies on memory layout, "
193 "which is dangerous.";
194 auto R = std::make_unique<PathSensitiveBugReport>(BT_pointerArith, Msg, N);
195 R->addRange(SR);
196 R->markInteresting(Region);
197 C.emitReport(std::move(R));
198 }
199}
200
201void PointerArithChecker::initAllocIdentifiers(ASTContext &C) const {
202 if (!AllocFunctions.empty())
203 return;
204 AllocFunctions.insert(&C.Idents.get("alloca"));
205 AllocFunctions.insert(&C.Idents.get("malloc"));
206 AllocFunctions.insert(&C.Idents.get("realloc"));
207 AllocFunctions.insert(&C.Idents.get("calloc"));
208 AllocFunctions.insert(&C.Idents.get("valloc"));
209}
210
211void PointerArithChecker::checkPostStmt(const CallExpr *CE,
212 CheckerContext &C) const {
213 ProgramStateRef State = C.getState();
214 const FunctionDecl *FD = C.getCalleeDecl(CE);
215 if (!FD)
216 return;
217 IdentifierInfo *FunI = FD->getIdentifier();
218 initAllocIdentifiers(C.getASTContext());
219 if (AllocFunctions.count(FunI) == 0)
220 return;
221
222 SVal SV = C.getSVal(CE);
223 const MemRegion *Region = SV.getAsRegion();
224 if (!Region)
225 return;
226 // Assume that C allocation functions allocate arrays to avoid false
227 // positives.
228 // TODO: Add heuristics to distinguish alloc calls that allocates single
229 // objecs.
230 State = State->set<RegionState>(Region, AllocKind::Array);
231 C.addTransition(State);
232}
233
234void PointerArithChecker::checkPostStmt(const CXXNewExpr *NE,
235 CheckerContext &C) const {
236 const FunctionDecl *FD = NE->getOperatorNew();
237 if (!FD)
238 return;
239
240 AllocKind Kind = getKindOfNewOp(NE, FD);
241
242 ProgramStateRef State = C.getState();
243 SVal AllocedVal = C.getSVal(NE);
244 const MemRegion *Region = AllocedVal.getAsRegion();
245 if (!Region)
246 return;
247 State = State->set<RegionState>(Region, Kind);
248 C.addTransition(State);
249}
250
251void PointerArithChecker::checkPostStmt(const CastExpr *CE,
252 CheckerContext &C) const {
253 if (CE->getCastKind() != CastKind::CK_BitCast)
254 return;
255
256 const Expr *CastedExpr = CE->getSubExpr();
257 ProgramStateRef State = C.getState();
258 SVal CastedVal = C.getSVal(CastedExpr);
259
260 const MemRegion *Region = CastedVal.getAsRegion();
261 if (!Region)
262 return;
263
264 // Suppress reinterpret casted hits.
265 State = State->set<RegionState>(Region, AllocKind::Reinterpreted);
266 C.addTransition(State);
267}
268
269void PointerArithChecker::checkPreStmt(const CastExpr *CE,
270 CheckerContext &C) const {
271 if (CE->getCastKind() != CastKind::CK_ArrayToPointerDecay)
272 return;
273
274 const Expr *CastedExpr = CE->getSubExpr();
275 ProgramStateRef State = C.getState();
276 SVal CastedVal = C.getSVal(CastedExpr);
277
278 const MemRegion *Region = CastedVal.getAsRegion();
279 if (!Region)
280 return;
281
282 if (const AllocKind *Kind = State->get<RegionState>(Region)) {
283 if (*Kind == AllocKind::Array || *Kind == AllocKind::Reinterpreted)
284 return;
285 }
286 State = State->set<RegionState>(Region, AllocKind::Array);
287 C.addTransition(State);
288}
289
290void PointerArithChecker::checkPreStmt(const UnaryOperator *UOp,
291 CheckerContext &C) const {
292 if (!UOp->isIncrementDecrementOp() || !UOp->getType()->isPointerType())
293 return;
294 reportPointerArithMisuse(UOp->getSubExpr(), C, true);
295}
296
297void PointerArithChecker::checkPreStmt(const ArraySubscriptExpr *SubsExpr,
298 CheckerContext &C) const {
299 SVal Idx = C.getSVal(SubsExpr->getIdx());
300
301 // Indexing with 0 is OK.
302 if (Idx.isZeroConstant())
303 return;
304
305 // Indexing vector-type expressions is also OK.
306 if (SubsExpr->getBase()->getType()->isVectorType())
307 return;
308 reportPointerArithMisuse(SubsExpr->getBase(), C);
309}
310
311void PointerArithChecker::checkPreStmt(const BinaryOperator *BOp,
312 CheckerContext &C) const {
313 BinaryOperatorKind OpKind = BOp->getOpcode();
314 if (!BOp->isAdditiveOp() && OpKind != BO_AddAssign && OpKind != BO_SubAssign)
315 return;
316
317 const Expr *Lhs = BOp->getLHS();
318 const Expr *Rhs = BOp->getRHS();
319 ProgramStateRef State = C.getState();
320
321 if (Rhs->getType()->isIntegerType() && Lhs->getType()->isPointerType()) {
322 SVal RHSVal = C.getSVal(Rhs);
323 if (State->isNull(RHSVal).isConstrainedTrue())
324 return;
325 reportPointerArithMisuse(Lhs, C, !BOp->isAdditiveOp());
326 }
327 // The int += ptr; case is not valid C++.
328 if (Lhs->getType()->isIntegerType() && Rhs->getType()->isPointerType()) {
329 SVal LHSVal = C.getSVal(Lhs);
330 if (State->isNull(LHSVal).isConstrainedTrue())
331 return;
332 reportPointerArithMisuse(Rhs, C);
333 }
334}
335
336void ento::registerPointerArithChecker(CheckerManager &mgr) {
337 mgr.registerChecker<PointerArithChecker>();
338}
339
340bool ento::shouldRegisterPointerArithChecker(const CheckerManager &mgr) {
341 return true;
342}
Expr * E
Defines the C++ Decl subclasses, other than those for templates (found in DeclTemplate....
Defines the clang::Expr interface and subclasses for C++ expressions.
#define X(type, name)
Definition: Value.h:143
#define REGISTER_MAP_WITH_PROGRAMSTATE(Name, Key, Value)
Declares an immutable map of type NameTy, suitable for placement into the ProgramState.
Holds long-lived AST nodes (such as types and decls) that can be referred to throughout the semantic ...
Definition: ASTContext.h:186
ArraySubscriptExpr - [C99 6.5.2.1] Array Subscripting.
Definition: Expr.h:2669
A builtin binary operation expression such as "x + y" or "x <= y".
Definition: Expr.h:3845
Expr * getLHS() const
Definition: Expr.h:3894
Expr * getRHS() const
Definition: Expr.h:3896
static bool isAdditiveOp(Opcode Opc)
Definition: Expr.h:3930
Opcode getOpcode() const
Definition: Expr.h:3889
Represents a new-expression for memory allocation and constructor calls, e.g: "new CXXNewExpr(foo)".
Definition: ExprCXX.h:2240
CallExpr - Represents a function call (C99 6.5.2.2, C++ [expr.call]).
Definition: Expr.h:2825
CastExpr - Base class for type casts, including both implicit casts (ImplicitCastExpr) and explicit c...
Definition: Expr.h:3488
CastKind getCastKind() const
Definition: Expr.h:3532
Expr * getSubExpr()
Definition: Expr.h:3538
This represents one expression.
Definition: Expr.h:110
QualType getType() const
Definition: Expr.h:142
Represents a function declaration or definition.
Definition: Decl.h:1971
bool isVariadic() const
Whether this function is variadic.
Definition: Decl.cpp:3077
unsigned getNumParams() const
Return the number of parameters this function must have based on its FunctionType.
Definition: Decl.cpp:3680
One of these records is kept for each identifier that is lexed.
IdentifierInfo * getIdentifier() const
Get the identifier that names this declaration, if there is one.
Definition: Decl.h:270
A trivial tuple used to represent a source range.
bool isInvalid() const
SourceRange getSourceRange() const LLVM_READONLY
SourceLocation tokens are not useful in isolation - they are low level value objects created/interpre...
Definition: Stmt.cpp:326
bool isPointerType() const
Definition: Type.h:7997
bool isIntegerType() const
isIntegerType() does not include complex integers (a GCC extension).
Definition: Type.h:8336
bool isVectorType() const
Definition: Type.h:8105
UnaryOperator - This represents the unary-expression's (except sizeof and alignof),...
Definition: Expr.h:2188
Expr * getSubExpr() const
Definition: Expr.h:2233
static bool isIncrementDecrementOp(Opcode Op)
Definition: Expr.h:2288
CHECKER * registerChecker(AT &&... Args)
Used to register checkers.
MemRegion - The root abstract class for all memory regions.
Definition: MemRegion.h:97
Kind getKind() const
Definition: MemRegion.h:175
SVal - This represents a symbolic expression, which can be either an L-value or an R-value.
Definition: SVals.h:55
bool isZeroConstant() const
Definition: SVals.cpp:258
const MemRegion * getAsRegion() const
Definition: SVals.cpp:120
A class responsible for cleaning up unused symbols.
bool NE(InterpState &S, CodePtr OpPC)
Definition: Interp.h:1042
The JSON file list parser is used to communicate input to InstallAPI.
BinaryOperatorKind
Diagnostic wrappers for TextAPI types for error reporting.
Definition: Dominators.h:30
static void Profile(AllocKind X, FoldingSetNodeID &ID)