1 //===----- UninitializedObject.h ---------------------------------*- C++ -*-==//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file defines helper classes for UninitializedObjectChecker and
11 // documentation about the logic of it.
12 //
13 // The checker reports uninitialized fields in objects created after a
14 // constructor call.
15 //
16 // This checker has several options:
17 // - "Pedantic" (boolean). If its not set or is set to false, the checker
18 // won't emit warnings for objects that don't have at least one initialized
19 // field. This may be set with
20 //
21 // `-analyzer-config alpha.cplusplus.UninitializedObject:Pedantic=true`.
22 //
23 // - "NotesAsWarnings" (boolean). If set to true, the checker will emit a
24 // warning for each uninitialized field, as opposed to emitting one warning
25 // per constructor call, and listing the uninitialized fields that belongs
26 // to it in notes. Defaults to false.
27 //
28 // `-analyzer-config \
29 // alpha.cplusplus.UninitializedObject:NotesAsWarnings=true`.
30 //
31 // - "CheckPointeeInitialization" (boolean). If set to false, the checker will
32 // not analyze the pointee of pointer/reference fields, and will only check
33 // whether the object itself is initialized. Defaults to false.
34 //
35 // `-analyzer-config \
36 // alpha.cplusplus.UninitializedObject:CheckPointeeInitialization=true`.
37 //
38 // - "IgnoreRecordsWithField" (string). If supplied, the checker will not
39 // analyze structures that have a field with a name or type name that
40 // matches the given pattern. Defaults to "".
41 //
42 // `-analyzer-config \
43 // alpha.cplusplus.UninitializedObject:IgnoreRecordsWithField="[Tt]ag|[Kk]ind"`.
44 //
45 // TODO: With some clever heuristics, some pointers should be dereferenced
46 // by default. For example, if the pointee is constructed within the
47 // constructor call, it's reasonable to say that no external object
48 // references it, and we wouldn't generate multiple report on the same
49 // pointee.
50 //
51 // Most of the following methods as well as the checker itself is defined in
52 // UninitializedObjectChecker.cpp.
53 //
54 // Some methods are implemented in UninitializedPointee.cpp, to reduce the
55 // complexity of the main checker file.
56 //
57 //===----------------------------------------------------------------------===//
58
59 #ifndef LLVM_CLANG_STATICANALYZER_UNINITIALIZEDOBJECT_H
60 #define LLVM_CLANG_STATICANALYZER_UNINITIALIZEDOBJECT_H
61
62 #include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h"
63
64 namespace clang {
65 namespace ento {
66
67 struct UninitObjCheckerOptions {
68 bool IsPedantic = false;
69 bool ShouldConvertNotesToWarnings = false;
70 bool CheckPointeeInitialization = false;
71 std::string IgnoredRecordsWithFieldPattern;
72 };
73
74 /// A lightweight polymorphic wrapper around FieldRegion *. We'll use this
75 /// interface to store addinitional information about fields. As described
76 /// later, a list of these objects (i.e. "fieldchain") will be constructed and
77 /// used for printing note messages should an uninitialized value be found.
78 class FieldNode {
79 protected:
80 const FieldRegion *FR;
81
82 /// FieldNodes are never meant to be created on the heap, see
83 /// FindUninitializedFields::addFieldToUninits().
84 /* non-virtual */ ~FieldNode() = default;
85
86 public:
FieldNode(const FieldRegion * FR)87 FieldNode(const FieldRegion *FR) : FR(FR) {}
88
89 // We'll delete all of these special member functions to force the users of
90 // this interface to only store references to FieldNode objects in containers.
91 FieldNode() = delete;
92 FieldNode(const FieldNode &) = delete;
93 FieldNode(FieldNode &&) = delete;
94 FieldNode &operator=(const FieldNode &) = delete;
95 FieldNode &operator=(const FieldNode &&) = delete;
96
Profile(llvm::FoldingSetNodeID & ID)97 void Profile(llvm::FoldingSetNodeID &ID) const { ID.AddPointer(this); }
98
99 /// Helper method for uniqueing.
isSameRegion(const FieldRegion * OtherFR)100 bool isSameRegion(const FieldRegion *OtherFR) const {
101 // Special FieldNode descendants may wrap nullpointers (for example if they
102 // describe a special relationship between two elements of the fieldchain)
103 // -- we wouldn't like to unique these objects.
104 if (FR == nullptr)
105 return false;
106
107 return FR == OtherFR;
108 }
109
getRegion()110 const FieldRegion *getRegion() const { return FR; }
getDecl()111 const FieldDecl *getDecl() const {
112 assert(FR);
113 return FR->getDecl();
114 }
115
116 // When a fieldchain is printed, it will have the following format (without
117 // newline, indices are in order of insertion, from 1 to n):
118 //
119 // <note_message_n>'<prefix_n><prefix_n-1>...<prefix_1>
120 // this-><node_1><separator_1><node_2><separator_2>...<node_n>'
121
122 /// If this is the last element of the fieldchain, this method will print the
123 /// note message associated with it.
124 /// The note message should state something like "uninitialized field" or
125 /// "uninitialized pointee" etc.
126 virtual void printNoteMsg(llvm::raw_ostream &Out) const = 0;
127
128 /// Print any prefixes before the fieldchain. Could contain casts, etc.
129 virtual void printPrefix(llvm::raw_ostream &Out) const = 0;
130
131 /// Print the node. Should contain the name of the field stored in FR.
132 virtual void printNode(llvm::raw_ostream &Out) const = 0;
133
134 /// Print the separator. For example, fields may be separated with '.' or
135 /// "->".
136 virtual void printSeparator(llvm::raw_ostream &Out) const = 0;
137
isBase()138 virtual bool isBase() const { return false; }
139 };
140
141 /// Returns with Field's name. This is a helper function to get the correct name
142 /// even if Field is a captured lambda variable.
143 std::string getVariableName(const FieldDecl *Field);
144
145 /// Represents a field chain. A field chain is a list of fields where the first
146 /// element of the chain is the object under checking (not stored), and every
147 /// other element is a field, and the element that precedes it is the object
148 /// that contains it.
149 ///
150 /// Note that this class is immutable (essentially a wrapper around an
151 /// ImmutableList), new FieldChainInfo objects may be created by member
152 /// functions such as add() and replaceHead().
153 class FieldChainInfo {
154 public:
155 using FieldChain = llvm::ImmutableList<const FieldNode &>;
156
157 private:
158 FieldChain::Factory &ChainFactory;
159 FieldChain Chain;
160
FieldChainInfo(FieldChain::Factory & F,FieldChain NewChain)161 FieldChainInfo(FieldChain::Factory &F, FieldChain NewChain)
162 : FieldChainInfo(F) {
163 Chain = NewChain;
164 }
165
166 public:
167 FieldChainInfo() = delete;
FieldChainInfo(FieldChain::Factory & F)168 FieldChainInfo(FieldChain::Factory &F) : ChainFactory(F) {}
169 FieldChainInfo(const FieldChainInfo &Other) = default;
170
171 /// Constructs a new FieldChainInfo object with \p FN appended.
172 template <class FieldNodeT> FieldChainInfo add(const FieldNodeT &FN);
173
174 /// Constructs a new FieldChainInfo object with \p FN as the new head of the
175 /// list.
176 template <class FieldNodeT> FieldChainInfo replaceHead(const FieldNodeT &FN);
177
178 bool contains(const FieldRegion *FR) const;
isEmpty()179 bool isEmpty() const { return Chain.isEmpty(); }
180
getHead()181 const FieldNode &getHead() const { return Chain.getHead(); }
getUninitRegion()182 const FieldRegion *getUninitRegion() const { return getHead().getRegion(); }
183
184 void printNoteMsg(llvm::raw_ostream &Out) const;
185 };
186
187 using UninitFieldMap = std::map<const FieldRegion *, llvm::SmallString<50>>;
188
189 /// Searches for and stores uninitialized fields in a non-union object.
190 class FindUninitializedFields {
191 ProgramStateRef State;
192 const TypedValueRegion *const ObjectR;
193
194 const UninitObjCheckerOptions Opts;
195 bool IsAnyFieldInitialized = false;
196
197 FieldChainInfo::FieldChain::Factory ChainFactory;
198
199 /// A map for assigning uninitialized regions to note messages. For example,
200 ///
201 /// struct A {
202 /// int x;
203 /// };
204 ///
205 /// A a;
206 ///
207 /// After analyzing `a`, the map will contain a pair for `a.x`'s region and
208 /// the note message "uninitialized field 'this->x'.
209 UninitFieldMap UninitFields;
210
211 public:
212 /// Constructs the FindUninitializedField object, searches for and stores
213 /// uninitialized fields in R.
214 FindUninitializedFields(ProgramStateRef State,
215 const TypedValueRegion *const R,
216 const UninitObjCheckerOptions &Opts);
217
218 /// Returns with the modified state and a map of (uninitialized region,
219 /// note message) pairs.
getResults()220 std::pair<ProgramStateRef, const UninitFieldMap &> getResults() {
221 return {State, UninitFields};
222 }
223
224 /// Returns whether the analyzed region contains at least one initialized
225 /// field. Note that this includes subfields as well, not just direct ones,
226 /// and will return false if an uninitialized pointee is found with
227 /// CheckPointeeInitialization enabled.
isAnyFieldInitialized()228 bool isAnyFieldInitialized() { return IsAnyFieldInitialized; }
229
230 private:
231 // For the purposes of this checker, we'll regard the analyzed region as a
232 // directed tree, where
233 // * the root is the object under checking
234 // * every node is an object that is
235 // - a union
236 // - a non-union record
237 // - dereferenceable (see isDereferencableType())
238 // - an array
239 // - of a primitive type (see isPrimitiveType())
240 // * the parent of each node is the object that contains it
241 // * every leaf is an array, a primitive object, a nullptr or an undefined
242 // pointer.
243 //
244 // Example:
245 //
246 // struct A {
247 // struct B {
248 // int x, y = 0;
249 // };
250 // B b;
251 // int *iptr = new int;
252 // B* bptr;
253 //
254 // A() {}
255 // };
256 //
257 // The directed tree:
258 //
259 // ->x
260 // /
261 // ->b--->y
262 // /
263 // A-->iptr->(int value)
264 // \
265 // ->bptr
266 //
267 // From this we'll construct a vector of fieldchains, where each fieldchain
268 // represents an uninitialized field. An uninitialized field may be a
269 // primitive object, a pointer, a pointee or a union without a single
270 // initialized field.
271 // In the above example, for the default constructor call we'll end up with
272 // these fieldchains:
273 //
274 // this->b.x
275 // this->iptr (pointee uninit)
276 // this->bptr (pointer uninit)
277 //
278 // We'll traverse each node of the above graph with the appropriate one of
279 // these methods:
280
281 /// Checks the region of a union object, and returns true if no field is
282 /// initialized within the region.
283 bool isUnionUninit(const TypedValueRegion *R);
284
285 /// Checks a region of a non-union object, and returns true if an
286 /// uninitialized field is found within the region.
287 bool isNonUnionUninit(const TypedValueRegion *R, FieldChainInfo LocalChain);
288
289 /// Checks a region of a pointer or reference object, and returns true if the
290 /// ptr/ref object itself or any field within the pointee's region is
291 /// uninitialized.
292 bool isDereferencableUninit(const FieldRegion *FR, FieldChainInfo LocalChain);
293
294 /// Returns true if the value of a primitive object is uninitialized.
295 bool isPrimitiveUninit(const SVal &V);
296
297 // Note that we don't have a method for arrays -- the elements of an array are
298 // often left uninitialized intentionally even when it is of a C++ record
299 // type, so we'll assume that an array is always initialized.
300 // TODO: Add a support for nonloc::LocAsInteger.
301
302 /// Processes LocalChain and attempts to insert it into UninitFields. Returns
303 /// true on success. Also adds the head of the list and \p PointeeR (if
304 /// supplied) to the GDM as already analyzed objects.
305 ///
306 /// Since this class analyzes regions with recursion, we'll only store
307 /// references to temporary FieldNode objects created on the stack. This means
308 /// that after analyzing a leaf of the directed tree described above, the
309 /// elements LocalChain references will be destructed, so we can't store it
310 /// directly.
311 bool addFieldToUninits(FieldChainInfo LocalChain,
312 const MemRegion *PointeeR = nullptr);
313 };
314
315 /// Returns true if T is a primitive type. An object of a primitive type only
316 /// needs to be analyzed as much as checking whether their value is undefined.
isPrimitiveType(const QualType & T)317 inline bool isPrimitiveType(const QualType &T) {
318 return T->isBuiltinType() || T->isEnumeralType() ||
319 T->isMemberPointerType() || T->isBlockPointerType() ||
320 T->isFunctionType();
321 }
322
isDereferencableType(const QualType & T)323 inline bool isDereferencableType(const QualType &T) {
324 return T->isAnyPointerType() || T->isReferenceType();
325 }
326
327 // Template method definitions.
328
329 template <class FieldNodeT>
add(const FieldNodeT & FN)330 inline FieldChainInfo FieldChainInfo::add(const FieldNodeT &FN) {
331 assert(!contains(FN.getRegion()) &&
332 "Can't add a field that is already a part of the "
333 "fieldchain! Is this a cyclic reference?");
334
335 FieldChainInfo NewChain = *this;
336 NewChain.Chain = ChainFactory.add(FN, Chain);
337 return NewChain;
338 }
339
340 template <class FieldNodeT>
replaceHead(const FieldNodeT & FN)341 inline FieldChainInfo FieldChainInfo::replaceHead(const FieldNodeT &FN) {
342 FieldChainInfo NewChain(ChainFactory, Chain.getTail());
343 return NewChain.add(FN);
344 }
345
346 } // end of namespace ento
347 } // end of namespace clang
348
349 #endif // LLVM_CLANG_STATICANALYZER_UNINITIALIZEDOBJECT_H
350