[analyzer] Remove some false negatives in StackAddrEscapeChecker

clang/lib/StaticAnalyzer/Checkers/StackAddrEscapeChecker.cpp

@@ -54,8 +54,8 @@ class StackAddrEscapeChecker
                                   CheckerContext &C) const;
   void checkAsyncExecutedBlockCaptures(const BlockDataRegion &B,
                                        CheckerContext &C) const;
-  void EmitStackError(CheckerContext &C, const MemRegion *R,
-                      const Expr *RetE) const;
+  void EmitReturnLeakError(CheckerContext &C, const MemRegion *LeakedRegion,
+                           const Expr *RetE) const;
   bool isSemaphoreCaptured(const BlockDecl &B) const;
   static SourceRange genName(raw_ostream &os, const MemRegion *R,
                              ASTContext &Ctx);
@@ -147,21 +147,38 @@ StackAddrEscapeChecker::getCapturedStackRegions(const BlockDataRegion &B,
   return Regions;
 }
 
-void StackAddrEscapeChecker::EmitStackError(CheckerContext &C,
-                                            const MemRegion *R,
-                                            const Expr *RetE) const {
+static void EmitReturnedAsPartOfError(llvm::raw_ostream &OS, SVal ReturnedVal,
+                                      const MemRegion *LeakedRegion) {
+  if (const MemRegion *ReturnedRegion = ReturnedVal.getAsRegion()) {
+    if (isa<BlockDataRegion>(ReturnedRegion)) {
+      OS << " is captured by a returned block";
+      return;
+    }
+  }
+
+  // Generic message
+  OS << " returned to caller";
+}
+
+void StackAddrEscapeChecker::EmitReturnLeakError(CheckerContext &C,
+                                                 const MemRegion *R,
+                                                 const Expr *RetE) const {
   ExplodedNode *N = C.generateNonFatalErrorNode();
   if (!N)
     return;
   if (!BT_returnstack)
     BT_returnstack = std::make_unique<BugType>(
         CheckNames[CK_StackAddrEscapeChecker],
         "Return of address to stack-allocated memory");
+
   // Generate a report for this bug.
   SmallString<128> buf;
   llvm::raw_svector_ostream os(buf);
+
+  // Error message formatting
   SourceRange range = genName(os, R, C.getASTContext());
-  os << " returned to caller";
+  EmitReturnedAsPartOfError(os, C.getSVal(RetE), R);
+
   auto report =
       std::make_unique<PathSensitiveBugReport>(*BT_returnstack, os.str(), N);
   report->addRange(RetE->getSourceRange());
@@ -209,30 +226,6 @@ void StackAddrEscapeChecker::checkAsyncExecutedBlockCaptures(
   }
 }
 
-void StackAddrEscapeChecker::checkReturnedBlockCaptures(
-    const BlockDataRegion &B, CheckerContext &C) const {
-  for (const MemRegion *Region : getCapturedStackRegions(B, C)) {
-    if (isNotInCurrentFrame(Region, C))
-      continue;
-    ExplodedNode *N = C.generateNonFatalErrorNode();
-    if (!N)
-      continue;
-    if (!BT_capturedstackret)
-      BT_capturedstackret = std::make_unique<BugType>(
-          CheckNames[CK_StackAddrEscapeChecker],
-          "Address of stack-allocated memory is captured");
-    SmallString<128> Buf;
-    llvm::raw_svector_ostream Out(Buf);
-    SourceRange Range = genName(Out, Region, C.getASTContext());
-    Out << " is captured by a returned block";
-    auto Report = std::make_unique<PathSensitiveBugReport>(*BT_capturedstackret,
-                                                           Out.str(), N);
-    if (Range.isValid())
-      Report->addRange(Range);
-    C.emitReport(std::move(Report));
-  }
-}
-
 void StackAddrEscapeChecker::checkPreCall(const CallEvent &Call,
                                           CheckerContext &C) const {
   if (!ChecksEnabled[CK_StackAddrAsyncEscapeChecker])
@@ -247,45 +240,134 @@ void StackAddrEscapeChecker::checkPreCall(const CallEvent &Call,
   }
 }
 
-void StackAddrEscapeChecker::checkPreStmt(const ReturnStmt *RS,
-                                          CheckerContext &C) const {
-  if (!ChecksEnabled[CK_StackAddrEscapeChecker])
-    return;
+/// A visitor made for use with a ScanReachableSymbols scanner, used
+/// for finding stack regions within an SVal that live on the current
+/// stack frame of the given checker context. This visitor excludes
+/// NonParamVarRegion that data is bound to in a BlockDataRegion's
+/// bindings, since these are likely uninteresting, e.g., in case a
+/// temporary is constructed on the stack, but it captures values
+/// that would leak.
+class FindStackRegionsSymbolVisitor final : public SymbolVisitor {
+  CheckerContext &Ctxt;
+  const StackFrameContext *StackFrameContext;
+  SmallVector<const MemRegion *> &EscapingStackRegions;
 
-  const Expr *RetE = RS->getRetValue();
-  if (!RetE)
-    return;
-  RetE = RetE->IgnoreParens();
+public:
+  explicit FindStackRegionsSymbolVisitor(
+      CheckerContext &Ctxt,
+      SmallVector<const MemRegion *> &StorageForStackRegions)
+      : Ctxt(Ctxt), StackFrameContext(Ctxt.getStackFrame()),
+        EscapingStackRegions(StorageForStackRegions) {}
 
-  SVal V = C.getSVal(RetE);
-  const MemRegion *R = V.getAsRegion();
-  if (!R)
-    return;
+  bool VisitSymbol(SymbolRef sym) override { return true; }
 
-  if (const BlockDataRegion *B = dyn_cast<BlockDataRegion>(R))
-    checkReturnedBlockCaptures(*B, C);
+  bool VisitMemRegion(const MemRegion *MR) override {
+    SaveIfEscapes(MR);
 
-  if (!isa<StackSpaceRegion>(R->getMemorySpace()) || isNotInCurrentFrame(R, C))
-    return;
+    if (const BlockDataRegion *BDR = MR->getAs<BlockDataRegion>())
+      return VisitBlockDataRegionCaptures(BDR);
+
+    return true;
+  }
+
+private:
+  void SaveIfEscapes(const MemRegion *MR) {
+    const StackSpaceRegion *SSR =
+        MR->getMemorySpace()->getAs<StackSpaceRegion>();
+    if (SSR && SSR->getStackFrame() == StackFrameContext)
+      EscapingStackRegions.push_back(MR);
+  }
+
+  bool VisitBlockDataRegionCaptures(const BlockDataRegion *BDR) {
+    for (auto Var : BDR->referenced_vars()) {
+      SVal Val = Ctxt.getState()->getSVal(Var.getCapturedRegion());
+      const MemRegion *Region = Val.getAsRegion();
+      if (Region) {
+        SaveIfEscapes(Region);
+        VisitMemRegion(Region);
+      }
+    }
+
+    return false;
+  }
+};
+
+/// Given some memory regions that are flagged by FindStackRegionsSymbolVisitor,
+/// this function filters out memory regions that are being returned that are
+/// likely not true leaks:
+/// 1. If returning a block data region that has stack memory space
+/// 2. If returning a constructed object that has stack memory space
+static SmallVector<const MemRegion *>
+FilterReturnExpressionLeaks(const SmallVector<const MemRegion *> &MaybeEscaped,
+                            CheckerContext &C, const Expr *RetE, SVal &RetVal) {
+
+  SmallVector<const MemRegion *> WillEscape;
+
+  const MemRegion *RetRegion = RetVal.getAsRegion();
 
   // Returning a record by value is fine. (In this case, the returned
   // expression will be a copy-constructor, possibly wrapped in an
   // ExprWithCleanups node.)
   if (const ExprWithCleanups *Cleanup = dyn_cast<ExprWithCleanups>(RetE))
     RetE = Cleanup->getSubExpr();
-  if (isa<CXXConstructExpr>(RetE) && RetE->getType()->isRecordType())
-    return;
+  bool IsConstructExpr =
+      isa<CXXConstructExpr>(RetE) && RetE->getType()->isRecordType();
 
   // The CK_CopyAndAutoreleaseBlockObject cast causes the block to be copied
   // so the stack address is not escaping here.
+  bool IsCopyAndAutoreleaseBlockObj = false;
   if (const auto *ICE = dyn_cast<ImplicitCastExpr>(RetE)) {
-    if (isa<BlockDataRegion>(R) &&
-        ICE->getCastKind() == CK_CopyAndAutoreleaseBlockObject) {
-      return;
-    }
+    IsCopyAndAutoreleaseBlockObj =
+        isa_and_nonnull<BlockDataRegion>(RetRegion) &&
+        ICE->getCastKind() == CK_CopyAndAutoreleaseBlockObject;
+  }
+
+  for (const MemRegion *MR : MaybeEscaped) {
+    if (RetRegion == MR && (IsCopyAndAutoreleaseBlockObj || IsConstructExpr))
+      continue;
+
+    // If this is a construct expr of an unelided return value copy, then don't
+    // warn about returning a region that currently lives on the stack.
+    if (IsConstructExpr && RetVal.getAs<nonloc::LazyCompoundVal>() &&
+        isa<CXXTempObjectRegion>(MR))
+      continue;
+
+    WillEscape.push_back(MR);
   }
 
-  EmitStackError(C, R, RetE);
+  return WillEscape;
+}
+
+/// For use in finding regions that live on the checker context's current
+/// stack frame, deep in the SVal representing the return value.
+static SmallVector<const MemRegion *>
+FindEscapingStackRegions(CheckerContext &C, const Expr *RetE, SVal RetVal) {
+  SmallVector<const MemRegion *> FoundStackRegions;
+
+  FindStackRegionsSymbolVisitor Finder(C, FoundStackRegions);
+  ScanReachableSymbols Scanner(C.getState(), Finder);
+  Scanner.scan(RetVal);
+
+  return FilterReturnExpressionLeaks(FoundStackRegions, C, RetE, RetVal);
+}
+
+void StackAddrEscapeChecker::checkPreStmt(const ReturnStmt *RS,
+                                          CheckerContext &C) const {
+  if (!ChecksEnabled[CK_StackAddrEscapeChecker])
+    return;
+
+  const Expr *RetE = RS->getRetValue();
+  if (!RetE)
+    return;
+  RetE = RetE->IgnoreParens();
+
+  SVal V = C.getSVal(RetE);
+
+  SmallVector<const MemRegion *> EscapedStackRegions =
+      FindEscapingStackRegions(C, RetE, V);
+
+  for (const MemRegion *ER : EscapedStackRegions)
+    EmitReturnLeakError(C, ER, RetE);
 }
 
 static const MemSpaceRegion *getStackOrGlobalSpaceRegion(const MemRegion *R) {

clang/test/Analysis/stack-addr-ps.cpp

@@ -251,7 +251,7 @@ void* lambda_to_context_direct_pointer_uncalled() {
     int local = 42;
     p = &local; // no-warning: analyzed only as top-level, ignored explicitly by the checker
   };
-  return new MyFunction(&lambda);
+  return new MyFunction(&lambda); // expected-warning{{Address of stack memory associated with local variable 'lambda' returned to caller}}
 }
 
 void lambda_to_context_direct_pointer_lifetime_extended() {
@@ -410,16 +410,16 @@ void** returned_arr_of_ptr_top() {
   int* p = &local;
   void** arr = new void*[2];
   arr[1] = p;
-  return arr;
-} // no-warning False Negative
+  return arr; // expected-warning{{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 void** returned_arr_of_ptr_callee() {
   int local = 42;
   int* p = &local;
   void** arr = new void*[2];
   arr[1] = p;
-  return arr;
-} // no-warning False Negative
+  return arr; // expected-warning{{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 void returned_arr_of_ptr_caller() {
   void** arr = returned_arr_of_ptr_callee();
@@ -466,16 +466,16 @@ void** returned_arr_of_ptr_top(int idx) {
   int* p = &local;
   void** arr = new void*[2];
   arr[idx] = p;
-  return arr;
-} // no-warning False Negative
+  return arr; // expected-warning{{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 void** returned_arr_of_ptr_callee(int idx) {
   int local = 42;
   int* p = &local;
   void** arr = new void*[2];
   arr[idx] = p;
-  return arr;
-} // no-warning False Negative
+  return arr; // expected-warning{{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 void returned_arr_of_ptr_caller(int idx) {
   void** arr = returned_arr_of_ptr_callee(idx);
@@ -525,14 +525,25 @@ S returned_struct_with_ptr_top() {
   int local = 42;
   S s;
   s.p = &local;
-  return s;
-} // no-warning False Negative, requires traversing returned LazyCompoundVals
+  return s; // expected-warning{{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 S returned_struct_with_ptr_callee() {
   int local = 42;
   S s;
   s.p = &local;
-  return s; // expected-warning{{'local' is still referred to by the caller variable 's'}}
+  return s; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}} expected-warning{{Address of stack memory associated with local variable 'local' is still referred to by the caller variable 's' upon returning to the caller.  This will be a dangling reference}}
+}
+
+S leak_through_field_of_returned_object() {
+  int local = 14;
+  S s{&local};
+  return s; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
+}
+
+S leak_through_compound_literal() {
+  int local = 0;
+  return (S) { &local }; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
 }
 
 void returned_struct_with_ptr_caller() {
@@ -555,6 +566,30 @@ void static_struct_with_ptr() {
 }
 } // namespace leaking_via_struct_with_ptr
 
+namespace leaking_via_nested_structs_with_ptr {
+struct Inner {
+  int *ptr;
+};
+
+struct Outer {
+  Inner I;
+};
+
+struct Deriving : public Outer {};
+
+Outer leaks_through_nested_objects() {
+  int local = 0;
+  Outer O{&local};
+  return O; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
+}
+
+Deriving leaks_through_base_objects() {
+  int local = 0;
+  Deriving D{&local};
+  return D; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
+}
+} // namespace leaking_via_nested_structs_with_ptr
+
 namespace leaking_via_ref_to_struct_with_ptr {
 struct S {
   int* p;
@@ -613,15 +648,15 @@ S* returned_ptr_to_struct_with_ptr_top() {
   int local = 42;
   S* s = new S;
   s->p = &local;
-  return s;
-} // no-warning False Negative
+  return s; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 S* returned_ptr_to_struct_with_ptr_callee() {
   int local = 42;
   S* s = new S;
   s->p = &local;
-  return s;
-} // no-warning False Negative
+  return s; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 void returned_ptr_to_struct_with_ptr_caller() {
   S* s = returned_ptr_to_struct_with_ptr_callee();
@@ -676,15 +711,15 @@ S* returned_ptr_to_struct_with_ptr_top() {
   int local = 42;
   S* s = new S[2];
   s[1].p = &local;
-  return s;
-} // no-warning False Negative
+  return s; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 S* returned_ptr_to_struct_with_ptr_callee() {
   int local = 42;
   S* s = new S[2];
   s[1].p = &local;
-  return s;
-} // no-warning  False Negative
+  return s; // expected-warning {{Address of stack memory associated with local variable 'local' returned to caller}}
+}
 
 void returned_ptr_to_struct_with_ptr_caller() {
   S* s = returned_ptr_to_struct_with_ptr_callee();