[IRGen] Emit lifetime intrinsics around temporary aggregate argument allocas

These temporaries are only used in the callee, and their memory can be reused
after the call is complete.

rdar://58552124

Differential revision: https://reviews.llvm.org/D74094

clang/lib/CodeGen/CGCall.cpp

@@ -3689,7 +3689,22 @@ void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
     return;
   }
 
-  args.add(EmitAnyExprToTemp(E), type);
+  AggValueSlot ArgSlot = AggValueSlot::ignored();
+  if (hasAggregateEvaluationKind(E->getType())) {
+    ArgSlot = CreateAggTemp(E->getType(), "agg.tmp");
+
+    // Emit a lifetime start/end for this temporary. If the type has a
+    // destructor, then we need to keep it alive. FIXME: We should still be able
+    // to end the lifetime after the destructor returns.
+    if (!E->getType().isDestructedType()) {
+      uint64_t size =
+          CGM.getDataLayout().getTypeAllocSize(ConvertTypeForMem(E->getType()));
+      if (auto *lifetimeSize = EmitLifetimeStart(size, ArgSlot.getPointer()))
+        args.addLifetimeCleanup({ArgSlot.getPointer(), lifetimeSize});
+    }
+  }
+
+  args.add(EmitAnyExpr(E, ArgSlot), type);
 }
 
 QualType CodeGenFunction::getVarArgType(const Expr *Arg) {
@@ -4769,6 +4784,9 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   for (CallLifetimeEnd &LifetimeEnd : CallLifetimeEndAfterCall)
     LifetimeEnd.Emit(*this, /*Flags=*/{});
 
+  for (auto &LT : CallArgs.getLifetimeCleanups())
+    EmitLifetimeEnd(LT.Size, LT.Addr);
+
   return Ret;
 }
 

clang/lib/CodeGen/CGCall.h

@@ -283,6 +283,11 @@ public:
     llvm::Instruction *IsActiveIP;
   };
 
+  struct EndLifetimeInfo {
+    llvm::Value *Addr;
+    llvm::Value *Size;
+  };
+
   void add(RValue rvalue, QualType type) { push_back(CallArg(rvalue, type)); }
 
   void addUncopiedAggregate(LValue LV, QualType type) {
@@ -299,6 +304,9 @@ public:
     CleanupsToDeactivate.insert(CleanupsToDeactivate.end(),
                                 other.CleanupsToDeactivate.begin(),
                                 other.CleanupsToDeactivate.end());
+    LifetimeCleanups.insert(LifetimeCleanups.end(),
+                            other.LifetimeCleanups.begin(),
+                            other.LifetimeCleanups.end());
     assert(!(StackBase && other.StackBase) && "can't merge stackbases");
     if (!StackBase)
       StackBase = other.StackBase;
@@ -338,6 +346,14 @@ public:
   /// memory.
   bool isUsingInAlloca() const { return StackBase; }
 
+  void addLifetimeCleanup(EndLifetimeInfo Info) {
+    LifetimeCleanups.push_back(Info);
+  }
+
+  ArrayRef<EndLifetimeInfo> getLifetimeCleanups() const {
+    return LifetimeCleanups;
+  }
+
 private:
   SmallVector<Writeback, 1> Writebacks;
 
@@ -346,6 +362,10 @@ private:
   /// occurs.
   SmallVector<CallArgCleanup, 1> CleanupsToDeactivate;
 
+  /// Lifetime information needed to call llvm.lifetime.end for any temporary
+  /// argument allocas.
+  SmallVector<EndLifetimeInfo, 2> LifetimeCleanups;
+
   /// The stacksave call.  It dominates all of the argument evaluation.
   llvm::CallInst *StackBase;
 };

clang/test/CodeGen/lifetime-call-temp.c

@@ -0,0 +1,83 @@
+// RUN: %clang -cc1                  -triple x86_64-apple-macos -O1 -disable-llvm-passes %s -S -emit-llvm -o - | FileCheck %s --implicit-check-not=llvm.lifetime
+// RUN: %clang -cc1 -xc++ -std=c++17 -triple x86_64-apple-macos -O1 -disable-llvm-passes %s -S -emit-llvm -o - | FileCheck %s --implicit-check-not=llvm.lifetime --check-prefix=CHECK --check-prefix=CXX
+// RUN: %clang -cc1 -xobjective-c    -triple x86_64-apple-macos -O1 -disable-llvm-passes %s -S -emit-llvm -o - | FileCheck %s --implicit-check-not=llvm.lifetime --check-prefix=CHECK --check-prefix=OBJC
+
+typedef struct { int x[100]; } aggregate;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void takes_aggregate(aggregate);
+aggregate gives_aggregate();
+
+// CHECK-LABEL: define void @t1
+void t1() {
+  takes_aggregate(gives_aggregate());
+
+  // CHECK: [[AGGTMP:%.*]] = alloca %struct.aggregate, align 8
+  // CHECK: [[CAST:%.*]] = bitcast %struct.aggregate* [[AGGTMP]] to i8*
+  // CHECK: call void @llvm.lifetime.start.p0i8(i64 400, i8* [[CAST]])
+  // CHECK: call void{{.*}} @gives_aggregate(%struct.aggregate* sret [[AGGTMP]])
+  // CHECK: call void @takes_aggregate(%struct.aggregate* byval(%struct.aggregate) align 8 [[AGGTMP]])
+  // CHECK: [[CAST:%.*]] = bitcast %struct.aggregate* [[AGGTMP]] to i8*
+  // CHECK: call void @llvm.lifetime.end.p0i8(i64 400, i8* [[CAST]])
+}
+
+// CHECK: declare {{.*}}llvm.lifetime.start
+// CHECK: declare {{.*}}llvm.lifetime.end
+
+#ifdef __cplusplus
+// CXX: define void @t2
+void t2() {
+  struct S {
+    S(aggregate) {}
+  };
+  S{gives_aggregate()};
+
+  // CXX: [[AGG:%.*]] = alloca %struct.aggregate
+  // CXX: call void @llvm.lifetime.start.p0i8(i64 400, i8*
+  // CXX: call void @gives_aggregate(%struct.aggregate* sret [[AGG]])
+  // CXX: call void @_ZZ2t2EN1SC1E9aggregate(%struct.S* {{.*}}, %struct.aggregate* byval(%struct.aggregate) align 8 [[AGG]])
+  // CXX: call void @llvm.lifetime.end.p0i8(i64 400, i8*
+}
+
+struct Dtor {
+  ~Dtor();
+};
+
+void takes_dtor(Dtor);
+Dtor gives_dtor();
+
+// CXX: define void @t3
+void t3() {
+  takes_dtor(gives_dtor());
+
+  // CXX-NOT @llvm.lifetime
+  // CXX: ret void
+}
+
+#endif
+
+#ifdef __OBJC__
+
+@interface X
+-m:(aggregate)x;
+@end
+
+// OBJC: define void @t4
+void t4(X *x) {
+  [x m: gives_aggregate()];
+
+  // OBJC: [[AGG:%.*]] = alloca %struct.aggregate
+  // OBJC: call void @llvm.lifetime.start.p0i8(i64 400, i8*
+  // OBJC: call void{{.*}} @gives_aggregate(%struct.aggregate* sret [[AGGTMP]])
+  // OBJC: call {{.*}}@objc_msgSend
+  // OBJC: call void @llvm.lifetime.end.p0i8(i64 400, i8*
+}
+
+#endif
+
+#ifdef __cplusplus
+}
+#endif

clang/test/CodeGenCXX/stack-reuse-miscompile.cpp

@@ -26,6 +26,8 @@ const char * f(S s)
 // CHECK: [[T2:%.*]] = alloca %class.T, align 4
 // CHECK: [[T3:%.*]] = alloca %class.T, align 4
 //
+// CHECK: [[AGG:%.*]] = alloca %class.S, align 4
+//
 // FIXME: We could defer starting the lifetime of the return object of concat
 // until the call.
 // CHECK: [[T1i8:%.*]] = bitcast %class.T* [[T1]] to i8*
@@ -37,8 +39,15 @@ const char * f(S s)
 //
 // CHECK: [[T3i8:%.*]] = bitcast %class.T* [[T3]] to i8*
 // CHECK: call void @llvm.lifetime.start.p0i8(i64 16, i8* [[T3i8]])
+//
+// CHECK: [[AGGi8:%.*]] = bitcast %class.S* [[AGG]] to i8*
+// CHECK: call void @llvm.lifetime.start.p0i8(i64 8, i8* [[AGGi8]])
+//
 // CHECK: [[T5:%.*]] = call %class.T* @_ZN1TC1E1S(%class.T* [[T3]], [2 x i32] %{{.*}})
 //
+// CHECK: [[AGGi8:%.*]] = bitcast %class.S* {{.*}} to i8*
+// CHECK: call void @llvm.lifetime.end.p0i8(i64 8, i8* [[AGGi8]])
+//
 // CHECK: call void @_ZNK1T6concatERKS_(%class.T* sret [[T1]], %class.T* [[T2]], %class.T* dereferenceable(16) [[T3]])
 // CHECK: [[T6:%.*]] = call i8* @_ZNK1T3strEv(%class.T* [[T1]])
 //