Index: llvm/trunk/lib/Analysis/ModuleSummaryAnalysis.cpp
===================================================================
--- llvm/trunk/lib/Analysis/ModuleSummaryAnalysis.cpp
+++ llvm/trunk/lib/Analysis/ModuleSummaryAnalysis.cpp
@@ -291,6 +291,18 @@
         if (!CalledValue || isa<Constant>(CalledValue))
           continue;
 
+        // Check if the instruction has a callees metadata. If so, add callees
+        // to CallGraphEdges to reflect the references from the metadata, and
+        // to enable importing for subsequent indirect call promotion and
+        // inlining.
+        if (auto *MD = I.getMetadata(LLVMContext::MD_callees)) {
+          for (auto &Op : MD->operands()) {
+            Function *Callee = mdconst::extract_or_null<Function>(Op);
+            if (Callee)
+              CallGraphEdges[Index.getOrInsertValueInfo(Callee)];
+          }
+        }
+
         uint32_t NumVals, NumCandidates;
         uint64_t TotalCount;
         auto CandidateProfileData =
Index: llvm/trunk/test/ThinLTO/X86/Inputs/callees-metadata.ll
===================================================================
--- llvm/trunk/test/ThinLTO/X86/Inputs/callees-metadata.ll
+++ llvm/trunk/test/ThinLTO/X86/Inputs/callees-metadata.ll
@@ -0,0 +1,34 @@
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+@fptr = internal unnamed_addr global i32 (i32)* @f2, align 8
+
+define dso_local i32 @foo(i32 %x) local_unnamed_addr {
+entry:
+  %0 = load i32 (i32)*, i32 (i32)** @fptr, align 8
+  %call = tail call i32 %0(i32 %x), !callees !0
+  ret i32 %call
+}
+
+define internal i32 @f2(i32 %x) {
+entry:
+  %tobool = icmp eq i32 %x, 0
+  br i1 %tobool, label %if.end, label %if.then
+
+if.then:                                          ; preds = %entry
+  store i32 (i32)* @f1, i32 (i32)** @fptr, align 8
+  %sub.i = add nsw i32 %x, -1
+  br label %if.end
+
+if.end:                                           ; preds = %entry, %if.then
+  %phi.call = phi i32 [ %sub.i, %if.then ], [ -1, %entry ]
+  ret i32 %phi.call
+}
+
+define internal i32 @f1(i32 %x) {
+entry:
+  %sub = add nsw i32 %x, -1
+  ret i32 %sub
+}
+
+!0 = !{i32 (i32)* @f1, i32 (i32)* @f2}
Index: llvm/trunk/test/ThinLTO/X86/callees-metadata.ll
===================================================================
--- llvm/trunk/test/ThinLTO/X86/callees-metadata.ll
+++ llvm/trunk/test/ThinLTO/X86/callees-metadata.ll
@@ -0,0 +1,22 @@
+; Do setup work: generate bitcode and combined index
+; RUN: opt -module-summary %s -o %t1.bc
+; RUN: opt -module-summary %p/Inputs/callees-metadata.ll -o %t2.bc
+
+; RUN: llvm-lto2 run %t1.bc %t2.bc -o %t.o -save-temps \
+; RUN:     -r=%t1.bc,bar,plx \
+; RUN:     -r=%t1.bc,foo,l \
+; RUN:     -r=%t2.bc,foo,pl
+; RUN: llvm-dis %t.o.1.3.import.bc -o - | FileCheck %s
+; CHECK: define {{.*}} i32 @f1.llvm.0
+; CHECK: define {{.*}} i32 @f2.llvm.0
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define dso_local i32 @bar(i32 %x) {
+entry:
+  %call = call i32 @foo(i32 %x)
+  ret i32 %call
+}
+
+declare dso_local i32 @foo(i32)