Index: llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
===================================================================
--- llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
+++ llvm/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
@@ -951,8 +951,8 @@
     addAddress(CallSiteDIE, getDwarf5OrGNUAttr(dwarf::DW_AT_call_target),
                MachineLocation(CallReg));
   } else {
-    DIE *CalleeDIE = getOrCreateSubprogramDIE(CalleeSP);
-    assert(CalleeDIE && "Could not create DIE for call site entry origin");
+    DIE *CalleeDIE = getDIE(CalleeSP);
+    assert(CalleeDIE && "Could not find DIE for call site entry origin");
     addDIEEntry(CallSiteDIE, getDwarf5OrGNUAttr(dwarf::DW_AT_call_origin),
                 *CalleeDIE);
   }
Index: llvm/lib/CodeGen/AsmPrinter/DwarfDebug.h
===================================================================
--- llvm/lib/CodeGen/AsmPrinter/DwarfDebug.h
+++ llvm/lib/CodeGen/AsmPrinter/DwarfDebug.h
@@ -437,6 +437,9 @@
   /// Construct a DIE for this abstract scope.
   void constructAbstractSubprogramScopeDIE(DwarfCompileUnit &SrcCU, LexicalScope *Scope);
 
+  /// Construct a DIE for the subprogram definition \p SP and return it.
+  DIE &constructSubprogramDefinitionDIE(const DISubprogram *SP);
+
   /// Construct DIEs for call site entries describing the calls in \p MF.
   void constructCallSiteEntryDIEs(const DISubprogram &SP, DwarfCompileUnit &CU,
                                   DIE &ScopeDIE, const MachineFunction &MF);
Index: llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
===================================================================
--- llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
+++ llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
@@ -535,6 +535,12 @@
   }
 }
 
+DIE &DwarfDebug::constructSubprogramDefinitionDIE(const DISubprogram *SP) {
+  assert(SP->isDefinition());
+  auto &CU = getOrCreateDwarfCompileUnit(SP->getUnit());
+  return *CU.getOrCreateSubprogramDIE(SP);
+}
+
 /// Try to interpret values loaded into registers that forward parameters
 /// for \p CallMI. Store parameters with interpreted value into \p Params.
 static void collectCallSiteParameters(const MachineInstr *CallMI,
@@ -748,6 +754,15 @@
         if (!CalleeDecl || !CalleeDecl->getSubprogram())
           continue;
         CalleeSP = CalleeDecl->getSubprogram();
+
+        if (CalleeSP->isDefinition()) {
+          // Ensure that a subprogram definition for the callee and a containing
+          // CU are available.
+          constructSubprogramDefinitionDIE(CalleeSP);
+        } else {
+          assert(CU.getDIE(CalleeSP) &&
+                 "Expected declaration subprogram DIE for callee");
+        }
       }
 
       // TODO: Omit call site entries for runtime calls (objc_msgSend, etc).
Index: llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
===================================================================
--- llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
+++ llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
@@ -188,8 +188,9 @@
 
 /// Check whether the DIE for this MDNode can be shared across CUs.
 bool DwarfUnit::isShareableAcrossCUs(const DINode *D) const {
-  // When the MDNode can be part of the type system, the DIE can be shared
-  // across CUs.
+  // When the MDNode can be part of the type system (this includes subprogram
+  // declarations *and* subprogram definitions, even local definitions), the
+  // DIE must be shared across CUs.
   // Combining type units and cross-CU DIE sharing is lower value (since
   // cross-CU DIE sharing is used in LTO and removes type redundancy at that
   // level already) but may be implementable for some value in projects
@@ -197,9 +198,7 @@
   // together.
   if (isDwoUnit() && !DD->shareAcrossDWOCUs())
     return false;
-  return (isa<DIType>(D) ||
-          (isa<DISubprogram>(D) && !cast<DISubprogram>(D)->isDefinition())) &&
-         !DD->generateTypeUnits();
+  return (isa<DIType>(D) || isa<DISubprogram>(D)) && !DD->generateTypeUnits();
 }
 
 DIE *DwarfUnit::getDIE(const DINode *D) const {
Index: llvm/test/DebugInfo/X86/lto-cross-cu-call-origin-ref.ll
===================================================================
--- /dev/null
+++ llvm/test/DebugInfo/X86/lto-cross-cu-call-origin-ref.ll
@@ -0,0 +1,200 @@
+; RUN: llc -mtriple=x86_64-apple-darwin -filetype=obj < %s | llvm-dwarfdump - \
+; RUN:   | FileCheck %s -implicit-check-not=DW_TAG_subprogram
+
+; Source:
+; // a.c
+; extern void helper_from_b(void);
+; __attribute__((optnone)) void bar() {}
+; __attribute__((optnone)) static void baz() {}
+; __attribute__((always_inline)) void foo() {
+;   bar();
+;   baz();
+; }
+; __attribute__((noinline)) void noinline_in_a() {
+;   helper_from_b();
+; }
+; // b.c
+; extern void foo(void);
+; extern void noinline_in_a(void);
+; __attribute__((optnone)) void baz() {}
+; __attribute__((always_inline)) void helper_from_b(void) { baz(); }
+; int main(int argc, char **argv) {
+;   foo();
+;   baz();
+;   noinline_in_a();
+;   return 0;
+; }
+
+; Command:
+; clang -O2 -Xclang -femit-debug-entry-values -g -flto -o a.o -c a.c
+; clang -O2 -Xclang -femit-debug-entry-values -g -flto -o b.o -c b.c
+; clang -O2 -Xclang -femit-debug-entry-values -g -flto -o main a.o b.o -Wl,-object_path_lto,lto.o,-save-temps
+
+; === CU for a.c ===
+
+; CHECK: DW_TAG_compile_unit
+; CHECK:   DW_AT_name ("a.c")
+
+; Match & ignore the declaration of "helper_from_b".
+; FIXME: There is probably no need for this declaration in an LTO setting.
+; CHECK: DW_TAG_subprogram
+; CHECK:   DW_AT_name ("helper_from_b")
+; CHECK:   DW_AT_declaration (true)
+
+; Check for the definitions expected in a.c. Save the DIE references for these
+; definitions so that, later, we can check that the cross-CU references in b.c
+; are correct.
+; CHECK: 0x{{0+}}[[BAR_IN_A_DIE:.*]]: DW_TAG_subprogram
+; CHECK:   DW_AT_call_all_calls (true)
+; CHECK:   DW_AT_name ("bar")
+
+; CHECK: 0x{{0+}}[[BAZ_IN_A_DIE:.*]]: DW_TAG_subprogram
+; CHECK:   DW_AT_call_all_calls (true)
+; CHECK:   DW_AT_name ("baz")
+
+; Check that `helper_from_b` is inlined into `noinline_in_a`, which in turn
+; produces a cross-CU call into b.c.
+; CHECK: 0x{{0+}}[[NOINLINE_IN_A_DIE:.*]]: DW_TAG_subprogram
+; CHECK:   DW_AT_call_all_calls (true)
+; CHECK:   DW_AT_name ("noinline_in_a")
+; CHECK:   DW_TAG_inlined_subroutine
+; CHECK-NEXT: DW_AT_abstract_origin (0x{{0+}}[[HELPER_FROM_B_DIE:.*]] "helper_from_b")
+; CHECK:   DW_TAG_call_site
+; CHECK-NEXT: DW_AT_call_origin (0x{{0+}}[[BAZ_IN_B_DIE:.*]])
+
+; CHECK: 0x{{0+}}[[FOO_IN_A_DIE:.*]]: DW_TAG_subprogram
+; CHECK:   DW_AT_name ("foo")
+; CHECK:   DW_AT_inline (DW_INL_inlined)
+
+; === CU for b.c ===
+
+; CHECK: DW_TAG_compile_unit
+; CHECK:   DW_AT_name ("b.c")
+
+; CHECK: DW_TAG_subprogram
+; CHECK:   DW_AT_name ("foo")
+; CHECK:   DW_AT_declaration (true)
+
+; CHECK: DW_TAG_subprogram
+; CHECK:   DW_AT_name ("noinline_in_a")
+; CHECK:   DW_AT_declaration (true)
+
+; CHECK: 0x{{0+}}[[HELPER_FROM_B_DIE]]: DW_TAG_subprogram
+; CHECK:   DW_AT_name ("helper_from_b")
+; CHECK:   DW_AT_inline (DW_INL_inlined)
+
+; Check for an external definition subprogram for "baz", distinct from the
+; "baz" local to a.c.
+; CHECK: 0x{{0+}}[[BAZ_IN_B_DIE]]: DW_TAG_subprogram
+; CHECK:   DW_AT_call_all_calls (true)
+; CHECK:   DW_AT_name ("baz")
+; CHECK:   DW_AT_external (true)
+
+; Check that "main" references, in order:
+;
+; 1) The definition of "bar" in "a.c". (Previously we would not share
+; definition subprograms across CUs, so the call site info logic would create an
+; empty 'definition' subprogram for "bar" which never got filled in. This breaks
+; both entry value evaluation & artificial tail call frame synthesis in the LTO
+; setting.)
+;
+; 2) The non-external definition of "baz" in "a.c". This is inlined via "foo".
+;
+; 3) The external definition of "baz" in "b.c".
+; CHECK: DW_TAG_subprogram
+; CHECK:   DW_AT_name ("main")
+; CHECK:   DW_TAG_inlined_subroutine
+; CHECK-NEXT: DW_AT_abstract_origin (0x{{0+}}[[FOO_IN_A_DIE]] "foo")
+; CHECK:   DW_TAG_call_site
+; CHECK-NEXT: DW_AT_call_origin (0x{{0+}}[[BAR_IN_A_DIE]])
+; CHECK:   DW_TAG_call_site
+; CHECK-NEXT: DW_AT_call_origin (0x{{0+}}[[BAZ_IN_A_DIE]])
+; CHECK:   DW_TAG_call_site
+; CHECK-NEXT: DW_AT_call_origin (0x{{0+}}[[BAZ_IN_B_DIE]])
+
+target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.14.0"
+
+define internal fastcc void @bar() unnamed_addr #0 !dbg !18 {
+entry:
+  ret void, !dbg !19
+}
+
+define internal fastcc void @baz.2() unnamed_addr #0 !dbg !20 {
+entry:
+  ret void, !dbg !21
+}
+
+define internal fastcc void @noinline_in_a() unnamed_addr #0 !dbg !22 {
+entry:
+  tail call fastcc void @baz(), !dbg !23
+  ret void, !dbg !26
+}
+
+define internal fastcc void @baz() unnamed_addr #0 !dbg !27 {
+entry:
+  ret void, !dbg !28
+}
+
+define i32 @main(i32 %argc, i8** nocapture readnone %argv) local_unnamed_addr !dbg !29 {
+entry:
+  tail call fastcc void @bar(), !dbg !40
+  tail call fastcc void @baz.2(), !dbg !43
+  tail call fastcc void @baz(), !dbg !44
+  tail call fastcc void @noinline_in_a(), !dbg !45
+  ret i32 0, !dbg !46
+}
+
+attributes #0 = { noinline }
+
+!llvm.dbg.cu = !{!0, !7}
+!llvm.ident = !{!12, !12}
+!llvm.module.flags = !{!13, !14, !15, !16, !17}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 10.0.0 (git@github.com:llvm/llvm-project.git 38530f9070cb06bb3d5bf6471befb70c7930da30)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !3, nameTableKind: None)
+!1 = !DIFile(filename: "a.c", directory: "/Users/vsk/tmp/lto-entry-vals")
+!2 = !{}
+!3 = !{!4}
+!4 = !DISubprogram(name: "helper_from_b", scope: !1, file: !1, line: 1, type: !5, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+!5 = !DISubroutineType(types: !6)
+!6 = !{null}
+!7 = distinct !DICompileUnit(language: DW_LANG_C99, file: !8, producer: "clang version 10.0.0 (git@github.com:llvm/llvm-project.git 38530f9070cb06bb3d5bf6471befb70c7930da30)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !9, nameTableKind: None)
+!8 = !DIFile(filename: "b.c", directory: "/Users/vsk/tmp/lto-entry-vals")
+!9 = !{!10, !11}
+!10 = !DISubprogram(name: "foo", scope: !8, file: !8, line: 1, type: !5, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+!11 = !DISubprogram(name: "noinline_in_a", scope: !8, file: !8, line: 2, type: !5, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !2)
+!12 = !{!"clang version 10.0.0 (git@github.com:llvm/llvm-project.git 38530f9070cb06bb3d5bf6471befb70c7930da30)"}
+!13 = !{i32 7, !"Dwarf Version", i32 4}
+!14 = !{i32 2, !"Debug Info Version", i32 3}
+!15 = !{i32 1, !"wchar_size", i32 4}
+!16 = !{i32 7, !"PIC Level", i32 2}
+!17 = !{i32 1, !"LTOPostLink", i32 1}
+!18 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 2, type: !5, scopeLine: 2, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!19 = !DILocation(line: 2, column: 38, scope: !18)
+!20 = distinct !DISubprogram(name: "baz", scope: !1, file: !1, line: 3, type: !5, scopeLine: 3, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagLocalToUnit | DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!21 = !DILocation(line: 3, column: 45, scope: !20)
+!22 = distinct !DISubprogram(name: "noinline_in_a", scope: !1, file: !1, line: 8, type: !5, scopeLine: 8, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!23 = !DILocation(line: 4, column: 59, scope: !24, inlinedAt: !25)
+!24 = distinct !DISubprogram(name: "helper_from_b", scope: !8, file: !8, line: 4, type: !5, scopeLine: 4, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !2)
+!25 = distinct !DILocation(line: 9, column: 3, scope: !22)
+!26 = !DILocation(line: 10, column: 1, scope: !22)
+!27 = distinct !DISubprogram(name: "baz", scope: !8, file: !8, line: 3, type: !5, scopeLine: 3, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !2)
+!28 = !DILocation(line: 3, column: 38, scope: !27)
+!29 = distinct !DISubprogram(name: "main", scope: !8, file: !8, line: 5, type: !30, scopeLine: 5, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !36)
+!30 = !DISubroutineType(types: !31)
+!31 = !{!32, !32, !33}
+!32 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!33 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !34, size: 64)
+!34 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !35, size: 64)
+!35 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_signed_char)
+!36 = !{!37, !38}
+!37 = !DILocalVariable(name: "argc", arg: 1, scope: !29, file: !8, line: 5, type: !32)
+!38 = !DILocalVariable(name: "argv", arg: 2, scope: !29, file: !8, line: 5, type: !33)
+!39 = !DILocation(line: 0, scope: !29)
+!40 = !DILocation(line: 5, column: 3, scope: !41, inlinedAt: !42)
+!41 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 4, type: !5, scopeLine: 4, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+!42 = distinct !DILocation(line: 6, column: 3, scope: !29)
+!43 = !DILocation(line: 6, column: 3, scope: !41, inlinedAt: !42)
+!44 = !DILocation(line: 7, column: 3, scope: !29)
+!45 = !DILocation(line: 8, column: 3, scope: !29)
+!46 = !DILocation(line: 9, column: 3, scope: !29)