diff --git a/lib/Transforms/Scalar/SCCP.cpp b/lib/Transforms/Scalar/SCCP.cpp
index 8d06690f2a4..5a0f62083ef 100644
--- a/lib/Transforms/Scalar/SCCP.cpp
+++ b/lib/Transforms/Scalar/SCCP.cpp
@@ -1918,6 +1918,14 @@ bool IPSCCP::runOnModule(Module &M) {
   // all call uses with the inferred value.  This means we don't need to bother
   // actually returning anything from the function.  Replace all return
   // instructions with return undef.
+  //
+  // Do this in two stages: first identify the functions we should process, then
+  // actually zap their returns.  This is important because we can only do this
+  // the address of the function isn't taken.  In cases where a return is the
+  // last use of a function, the order of processing functions would affect
+  // whether we other functions are optimizable.
+  SmallVector<ReturnInst*, 8> ReturnsToZap;
+  
   // TODO: Process multiple value ret instructions also.
   const DenseMap<Function*, LatticeVal> &RV = Solver.getTrackedRetVals();
   for (DenseMap<Function*, LatticeVal>::const_iterator I = RV.begin(),
@@ -1933,7 +1941,13 @@ bool IPSCCP::runOnModule(Module &M) {
     for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
       if (ReturnInst *RI = dyn_cast<ReturnInst>(BB->getTerminator()))
         if (!isa<UndefValue>(RI->getOperand(0)))
-          RI->setOperand(0, UndefValue::get(F->getReturnType()));
+          ReturnsToZap.push_back(RI);
+  }
+
+  // Zap all returns which we've identified as zap to change.
+  for (unsigned i = 0, e = ReturnsToZap.size(); i != e; ++i) {
+    Function *F = ReturnsToZap[i]->getParent()->getParent();
+    ReturnsToZap[i]->setOperand(0, UndefValue::get(F->getReturnType()));
   }
     
   // If we infered constant or undef values for globals variables, we can delete
diff --git a/test/Transforms/SCCP/retvalue-undef.ll b/test/Transforms/SCCP/retvalue-undef.ll
new file mode 100644
index 00000000000..389561f8a11
--- /dev/null
+++ b/test/Transforms/SCCP/retvalue-undef.ll
@@ -0,0 +1,32 @@
+; RUN: opt -ipsccp -S %s | FileCheck %s
+; PR6414
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define internal i32 ()* @f() {
+  ret i32 ()* @g
+}
+
+define internal i32 @g() {
+  ret i32 8
+}
+
+; CHECK: internal i32 @g()
+; CHECK-NEXT: ret i32 8
+
+define internal void @outer_mod() {
+  %1 = call i32 ()* ()* @f()                      ; <i32 ()*> [#uses=1]
+  %2 = call i32 %1()                              ; <i32> [#uses=0]
+  ret void
+}
+
+define internal void @module_init() {
+  call void @register_outer_mod(void ()* @outer_mod)
+  ret void
+}
+
+declare void @register_outer_mod(void ()*)
+
+define i32 @main() {
+  ret i32 0
+}