Index: include/llvm/Analysis/TypeMetadataUtils.h
===================================================================
--- include/llvm/Analysis/TypeMetadataUtils.h
+++ include/llvm/Analysis/TypeMetadataUtils.h
@@ -20,6 +20,13 @@
 
 namespace llvm {
 
+/// The type of CFI jumptable needed for a function.
+enum CfiFunctionLinkage {
+  CFL_Definition = 0,
+  CFL_Declaration = 1,
+  CFL_WeakDeclaration = 2
+};
+
 /// A call site that could be devirtualized.
 struct DevirtCallSite {
   /// The offset from the address point to the virtual function.
Index: include/llvm/Bitcode/BitcodeWriter.h
===================================================================
--- include/llvm/Bitcode/BitcodeWriter.h
+++ include/llvm/Bitcode/BitcodeWriter.h
@@ -67,6 +67,10 @@
     void writeModule(const Module *M, bool ShouldPreserveUseListOrder = false,
                      const ModuleSummaryIndex *Index = nullptr,
                      bool GenerateHash = false, ModuleHash *ModHash = nullptr);
+
+    void writeIndex(
+        const ModuleSummaryIndex *Index,
+        const std::map<std::string, GVSummaryMapTy> *ModuleToSummariesForIndex);
   };
 
   /// \brief Write the specified module to the specified raw output stream.
Index: include/llvm/Bitcode/LLVMBitCodes.h
===================================================================
--- include/llvm/Bitcode/LLVMBitCodes.h
+++ include/llvm/Bitcode/LLVMBitCodes.h
@@ -240,6 +240,14 @@
   // summaries, but it can also appear in per-module summaries for PGO data.
   // [valueid, guid]
   FS_VALUE_GUID = 16,
+  // The list of local functions with CFI jump tables. Function names are
+  // strings in strtab.
+  // [n * name]
+  FS_CFI_FUNCTION_DEFS = 17,
+  // The list of external functions with CFI jump tables. Function names are
+  // strings in strtab.
+  // [n * name]
+  FS_CFI_FUNCTION_DECLS = 18,
 };
 
 enum MetadataCodes {
Index: include/llvm/IR/ModuleSummaryIndex.h
===================================================================
--- include/llvm/IR/ModuleSummaryIndex.h
+++ include/llvm/IR/ModuleSummaryIndex.h
@@ -32,6 +32,7 @@
 #include <cstdint>
 #include <map>
 #include <memory>
+#include <set>
 #include <string>
 #include <utility>
 #include <vector>
@@ -542,6 +543,9 @@
   /// considered live.
   bool WithGlobalValueDeadStripping = false;
 
+  std::set<std::string> CfiFunctionDefs;
+  std::set<std::string> CfiFunctionDecls;
+
   // YAML I/O support.
   friend yaml::MappingTraits<ModuleSummaryIndex>;
 
@@ -592,6 +596,12 @@
     return I == OidGuidMap.end() ? 0 : I->second;
   }
 
+  std::set<std::string> &cfiFunctionDefs() { return CfiFunctionDefs; }
+  const std::set<std::string> &cfiFunctionDefs() const { return CfiFunctionDefs; }
+
+  std::set<std::string> &cfiFunctionDecls() { return CfiFunctionDecls; }
+  const std::set<std::string> &cfiFunctionDecls() const { return CfiFunctionDecls; }
+
   /// Add a global value summary for a value of the given name.
   void addGlobalValueSummary(StringRef ValueName,
                              std::unique_ptr<GlobalValueSummary> Summary) {
Index: include/llvm/IR/ModuleSummaryIndexYAML.h
===================================================================
--- include/llvm/IR/ModuleSummaryIndexYAML.h
+++ include/llvm/IR/ModuleSummaryIndexYAML.h
@@ -188,6 +188,7 @@
 
 LLVM_YAML_IS_STRING_MAP(TypeIdSummary)
 LLVM_YAML_IS_SEQUENCE_VECTOR(FunctionSummaryYaml)
+LLVM_YAML_IS_SEQUENCE_VECTOR(std::string)
 
 namespace llvm {
 namespace yaml {
@@ -240,6 +241,23 @@
     io.mapOptional("TypeIdMap", index.TypeIdMap);
     io.mapOptional("WithGlobalValueDeadStripping",
                    index.WithGlobalValueDeadStripping);
+
+    if (io.outputting()) {
+      std::vector<std::string> CfiFunctionDefs(index.CfiFunctionDefs.begin(),
+                                               index.CfiFunctionDefs.end());
+      io.mapOptional("CfiFunctionDefs", CfiFunctionDefs);
+      std::vector<std::string> CfiFunctionDecls(index.CfiFunctionDecls.begin(),
+                                                index.CfiFunctionDecls.end());
+      io.mapOptional("CfiFunctionDecls", CfiFunctionDecls);
+    } else {
+      std::vector<std::string> CfiFunctionDefs;
+      io.mapOptional("CfiFunctionDefs", CfiFunctionDefs);
+      index.CfiFunctionDefs = {CfiFunctionDefs.begin(), CfiFunctionDefs.end()};
+      std::vector<std::string> CfiFunctionDecls;
+      io.mapOptional("CfiFunctionDecls", CfiFunctionDecls);
+      index.CfiFunctionDecls = {CfiFunctionDecls.begin(),
+                                CfiFunctionDecls.end()};
+    }
   }
 };
 
Index: lib/Bitcode/Reader/BitcodeReader.cpp
===================================================================
--- lib/Bitcode/Reader/BitcodeReader.cpp
+++ lib/Bitcode/Reader/BitcodeReader.cpp
@@ -5251,6 +5251,18 @@
           {{Record[0], Record[1]}, {Record.begin() + 2, Record.end()}});
       break;
     }
+    case bitc::FS_CFI_FUNCTION_DEFS: {
+      std::set<std::string> &CfiFunctionDefs = TheIndex.cfiFunctionDefs();
+      for (unsigned I = 0; I != Record.size(); I += 2)
+        CfiFunctionDefs.insert({Strtab.data() + Record[I], Record[I+1]});
+      break;
+    }
+    case bitc::FS_CFI_FUNCTION_DECLS: {
+      std::set<std::string> &CfiFunctionDecls = TheIndex.cfiFunctionDecls();
+      for (unsigned I = 0; I != Record.size(); I += 2)
+        CfiFunctionDecls.insert({Strtab.data() + Record[I], Record[I+1]});
+      break;
+    }
     }
   }
   llvm_unreachable("Exit infinite loop");
Index: lib/Bitcode/Writer/BitcodeWriter.cpp
===================================================================
--- lib/Bitcode/Writer/BitcodeWriter.cpp
+++ lib/Bitcode/Writer/BitcodeWriter.cpp
@@ -77,10 +77,13 @@
   /// The stream created and owned by the client.
   BitstreamWriter &Stream;
 
+  StringTableBuilder &StrtabBuilder;
+
 public:
   /// Constructs a BitcodeWriterBase object that writes to the provided
   /// \p Stream.
-  BitcodeWriterBase(BitstreamWriter &Stream) : Stream(Stream) {}
+  BitcodeWriterBase(BitstreamWriter &Stream, StringTableBuilder &StrtabBuilder)
+      : Stream(Stream), StrtabBuilder(StrtabBuilder) {}
 
 protected:
   void writeBitcodeHeader();
@@ -97,8 +100,6 @@
   /// Pointer to the buffer allocated by caller for bitcode writing.
   const SmallVectorImpl<char> &Buffer;
 
-  StringTableBuilder &StrtabBuilder;
-
   /// The Module to write to bitcode.
   const Module &M;
 
@@ -142,8 +143,8 @@
                       BitstreamWriter &Stream, bool ShouldPreserveUseListOrder,
                       const ModuleSummaryIndex *Index, bool GenerateHash,
                       ModuleHash *ModHash = nullptr)
-      : BitcodeWriterBase(Stream), Buffer(Buffer), StrtabBuilder(StrtabBuilder),
-        M(*M), VE(*M, ShouldPreserveUseListOrder), Index(Index),
+      : BitcodeWriterBase(Stream, StrtabBuilder), Buffer(Buffer), M(*M),
+        VE(*M, ShouldPreserveUseListOrder), Index(Index),
         GenerateHash(GenerateHash), ModHash(ModHash),
         BitcodeStartBit(Stream.GetCurrentBitNo()) {
     // Assign ValueIds to any callee values in the index that came from
@@ -331,10 +332,11 @@
   /// Constructs a IndexBitcodeWriter object for the given combined index,
   /// writing to the provided \p Buffer. When writing a subset of the index
   /// for a distributed backend, provide a \p ModuleToSummariesForIndex map.
-  IndexBitcodeWriter(BitstreamWriter &Stream, const ModuleSummaryIndex &Index,
+  IndexBitcodeWriter(BitstreamWriter &Stream, StringTableBuilder &StrtabBuilder,
+                     const ModuleSummaryIndex &Index,
                      const std::map<std::string, GVSummaryMapTy>
                          *ModuleToSummariesForIndex = nullptr)
-      : BitcodeWriterBase(Stream), Index(Index),
+      : BitcodeWriterBase(Stream, StrtabBuilder), Index(Index),
         ModuleToSummariesForIndex(ModuleToSummariesForIndex) {
     // Assign unique value ids to all summaries to be written, for use
     // in writing out the call graph edges. Save the mapping from GUID
@@ -3595,6 +3597,24 @@
     MaybeEmitOriginalName(*AS);
   }
 
+  if (!Index.cfiFunctionDefs().empty()) {
+    for (auto &S : Index.cfiFunctionDefs()) {
+      NameVals.push_back(StrtabBuilder.add(S));
+      NameVals.push_back(S.size());
+    }
+    Stream.EmitRecord(bitc::FS_CFI_FUNCTION_DEFS, NameVals);
+    NameVals.clear();
+  }
+
+  if (!Index.cfiFunctionDecls().empty()) {
+    for (auto &S : Index.cfiFunctionDecls()) {
+      NameVals.push_back(StrtabBuilder.add(S));
+      NameVals.push_back(S.size());
+    }
+    Stream.EmitRecord(bitc::FS_CFI_FUNCTION_DECLS, NameVals);
+    NameVals.clear();
+  }
+
   Stream.ExitBlock();
 }
 
@@ -3829,6 +3849,14 @@
   ModuleWriter.write();
 }
 
+void BitcodeWriter::writeIndex(
+    const ModuleSummaryIndex *Index,
+    const std::map<std::string, GVSummaryMapTy> *ModuleToSummariesForIndex) {
+  IndexBitcodeWriter IndexWriter(*Stream, StrtabBuilder, *Index,
+                                 ModuleToSummariesForIndex);
+  IndexWriter.write();
+}
+
 /// WriteBitcodeToFile - Write the specified module to the specified output
 /// stream.
 void llvm::WriteBitcodeToFile(const Module *M, raw_ostream &Out,
@@ -3880,11 +3908,9 @@
   SmallVector<char, 0> Buffer;
   Buffer.reserve(256 * 1024);
 
-  BitstreamWriter Stream(Buffer);
-  writeBitcodeHeader(Stream);
-
-  IndexBitcodeWriter IndexWriter(Stream, Index, ModuleToSummariesForIndex);
-  IndexWriter.write();
+  BitcodeWriter Writer(Buffer);
+  Writer.writeIndex(&Index, ModuleToSummariesForIndex);
+  Writer.writeStrtab();
 
   Out.write((char *)&Buffer.front(), Buffer.size());
 }
Index: lib/Transforms/IPO/CrossDSOCFI.cpp
===================================================================
--- lib/Transforms/IPO/CrossDSOCFI.cpp
+++ lib/Transforms/IPO/CrossDSOCFI.cpp
@@ -95,6 +95,17 @@
     }
   }
 
+  NamedMDNode *CfiFunctionsMD = M.getNamedMetadata("cfi.functions");
+  if (CfiFunctionsMD) {
+    for (auto Func : CfiFunctionsMD->operands()) {
+      assert(Func->getNumOperands() >= 2);
+      for (unsigned I = 2; I < Func->getNumOperands(); ++I)
+        if (ConstantInt *TypeId =
+                extractNumericTypeId(cast<MDNode>(Func->getOperand(I).get())))
+          TypeIds.insert(TypeId->getZExtValue());
+    }
+  }
+
   LLVMContext &Ctx = M.getContext();
   Constant *C = M.getOrInsertFunction(
       "__cfi_check", Type::getVoidTy(Ctx), Type::getInt64Ty(Ctx),
Index: lib/Transforms/IPO/LowerTypeTests.cpp
===================================================================
--- lib/Transforms/IPO/LowerTypeTests.cpp
+++ lib/Transforms/IPO/LowerTypeTests.cpp
@@ -17,6 +17,7 @@
 #include "llvm/ADT/SetVector.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/Triple.h"
+#include "llvm/Analysis/TypeMetadataUtils.h"
 #include "llvm/IR/Constant.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/Function.h"
@@ -206,17 +207,26 @@
 class GlobalTypeMember final : TrailingObjects<GlobalTypeMember, MDNode *> {
   GlobalObject *GO;
   size_t NTypes;
+  // For functions: true if this is a definition (either in the merged module or
+  // in one of the thinlto modules).
+  bool IsDefinition;
+  // For functions: true if this function is either defined or used in a thinlto
+  // module and its jumptable entry needs to be exported to thinlto backends.
+  bool IsExported;
 
   friend TrailingObjects;
   size_t numTrailingObjects(OverloadToken<MDNode *>) const { return NTypes; }
 
 public:
   static GlobalTypeMember *create(BumpPtrAllocator &Alloc, GlobalObject *GO,
+                                  bool IsDefinition, bool IsExported,
                                   ArrayRef<MDNode *> Types) {
     auto *GTM = static_cast<GlobalTypeMember *>(Alloc.Allocate(
         totalSizeToAlloc<MDNode *>(Types.size()), alignof(GlobalTypeMember)));
     GTM->GO = GO;
     GTM->NTypes = Types.size();
+    GTM->IsDefinition = IsDefinition;
+    GTM->IsExported = IsExported;
     std::uninitialized_copy(Types.begin(), Types.end(),
                             GTM->getTrailingObjects<MDNode *>());
     return GTM;
@@ -224,6 +234,12 @@
   GlobalObject *getGlobal() const {
     return GO;
   }
+  bool isDefinition() const {
+    return IsDefinition;
+  }
+  bool isExported() const {
+    return IsExported;
+  }
   ArrayRef<MDNode *> types() const {
     return makeArrayRef(getTrailingObjects<MDNode *>(), NTypes);
   }
@@ -294,6 +310,7 @@
   void exportTypeId(StringRef TypeId, const TypeIdLowering &TIL);
   TypeIdLowering importTypeId(StringRef TypeId);
   void importTypeTest(CallInst *CI);
+  void importFunction(Function *F, bool isDefinition);
 
   BitSetInfo
   buildBitSet(Metadata *TypeId,
@@ -820,6 +837,41 @@
   CI->eraseFromParent();
 }
 
+// ThinLTO backend: the function F has a jump table entry; update this module
+// accordingly. isDefinition describes the type of the jump table entry.
+void LowerTypeTestsModule::importFunction(Function *F, bool isDefinition) {
+  assert(F->getType()->getAddressSpace() == 0);
+
+  // Declaration of a local function - nothing to do.
+  if (F->isDeclarationForLinker() && isDefinition)
+    return;
+
+  GlobalValue::VisibilityTypes Visibility = F->getVisibility();
+  std::string Name = F->getName();
+  Function *FDecl;
+
+  if (F->isDeclarationForLinker() && !isDefinition) {
+    // Declaration of an external function.
+    FDecl = Function::Create(F->getFunctionType(), GlobalValue::ExternalLinkage,
+                             Name + ".cfi_jt", &M);
+    FDecl->setVisibility(GlobalValue::HiddenVisibility);
+  } else {
+    // Definition.
+    assert(isDefinition);
+    F->setName(Name + ".cfi");
+    F->setLinkage(GlobalValue::ExternalLinkage);
+    F->setVisibility(GlobalValue::HiddenVisibility);
+    FDecl = Function::Create(F->getFunctionType(), GlobalValue::ExternalLinkage,
+                             Name, &M);
+    FDecl->setVisibility(Visibility);
+  }
+
+  if (F->isWeakForLinker())
+    replaceWeakDeclarationWithJumpTablePtr(F, FDecl);
+  else
+    F->replaceAllUsesWith(FDecl);
+}
+
 void LowerTypeTestsModule::lowerTypeTestCalls(
     ArrayRef<Metadata *> TypeIds, Constant *CombinedGlobalAddr,
     const DenseMap<GlobalTypeMember *, uint64_t> &GlobalLayout) {
@@ -1143,7 +1195,6 @@
   // arithmetic that we normally use for globals.
 
   // FIXME: find a better way to represent the jumptable in the IR.
-
   assert(!Functions.empty());
 
   // Build a simple layout based on the regular layout of jump tables.
@@ -1167,6 +1218,7 @@
   // references to the original functions with references to the aliases.
   for (unsigned I = 0; I != Functions.size(); ++I) {
     Function *F = cast<Function>(Functions[I]->getGlobal());
+    bool IsDefinition = Functions[I]->isDefinition();
 
     Constant *CombinedGlobalElemPtr = ConstantExpr::getBitCast(
         ConstantExpr::getInBoundsGetElementPtr(
@@ -1174,7 +1226,18 @@
             ArrayRef<Constant *>{ConstantInt::get(IntPtrTy, 0),
                                  ConstantInt::get(IntPtrTy, I)}),
         F->getType());
-    if (F->isDeclarationForLinker()) {
+    if (Functions[I]->isExported()) {
+      if (IsDefinition) {
+        ExportSummary->cfiFunctionDefs().insert(F->getName());
+      } else {
+        GlobalAlias *JtAlias = GlobalAlias::create(
+            F->getValueType(), 0, GlobalValue::ExternalLinkage,
+            F->getName() + ".cfi_jt", CombinedGlobalElemPtr, &M);
+        JtAlias->setVisibility(GlobalValue::HiddenVisibility);
+        ExportSummary->cfiFunctionDecls().insert(F->getName());
+      }
+    }
+    if (!IsDefinition) {
       if (F->isWeakForLinker())
         replaceWeakDeclarationWithJumpTablePtr(F, CombinedGlobalElemPtr);
       else
@@ -1182,9 +1245,8 @@
     } else {
       assert(F->getType()->getAddressSpace() == 0);
 
-      GlobalAlias *FAlias = GlobalAlias::create(F->getValueType(), 0,
-                                                F->getLinkage(), "",
-                                                CombinedGlobalElemPtr, &M);
+      GlobalAlias *FAlias = GlobalAlias::create(
+          F->getValueType(), 0, F->getLinkage(), "", CombinedGlobalElemPtr, &M);
       FAlias->setVisibility(F->getVisibility());
       FAlias->takeName(F);
       if (FAlias->hasName())
@@ -1350,18 +1412,51 @@
   return Changed;
 }
 
+static bool IsFunctionLive(StringRef Name, const ModuleSummaryIndex *Summary) {
+  auto GUID = GlobalValue::getGUID(GlobalValue::dropLLVMManglingEscape(Name));
+  ValueInfo Info = Summary->getValueInfo(GUID);
+  if (!Info)
+    return true;
+  const auto &SummaryList = Info.getSummaryList();
+  if (SummaryList.empty())
+    return true;
+  return Summary->isGlobalValueLive(SummaryList[0].get());
+}
+
 bool LowerTypeTestsModule::lower() {
   Function *TypeTestFunc =
       M.getFunction(Intrinsic::getName(Intrinsic::type_test));
-  if ((!TypeTestFunc || TypeTestFunc->use_empty()) && !ExportSummary)
+  if ((!TypeTestFunc || TypeTestFunc->use_empty()) && !ExportSummary &&
+      !ImportSummary)
     return false;
 
   if (ImportSummary) {
-    for (auto UI = TypeTestFunc->use_begin(), UE = TypeTestFunc->use_end();
-         UI != UE;) {
-      auto *CI = cast<CallInst>((*UI++).getUser());
-      importTypeTest(CI);
+    if (TypeTestFunc) {
+      for (auto UI = TypeTestFunc->use_begin(), UE = TypeTestFunc->use_end();
+           UI != UE;) {
+        auto *CI = cast<CallInst>((*UI++).getUser());
+        importTypeTest(CI);
+      }
     }
+
+    SmallVector<Function *, 8> Defs;
+    SmallVector<Function *, 8> Decls;
+    for (auto &F : M) {
+      // CFI functions are either external, or promoted. A local function may
+      // have the same name, but it's not the one we are looking for.
+      if (F.hasLocalLinkage())
+        continue;
+      if (ImportSummary->cfiFunctionDefs().count(F.getName()))
+        Defs.push_back(&F);
+      else if (ImportSummary->cfiFunctionDecls().count(F.getName()))
+        Decls.push_back(&F);
+    }
+
+    for (auto F : Defs)
+      importFunction(F, /*isDefinition*/ true);
+    for (auto F : Decls)
+      importFunction(F, /*isDefinition*/ false);
+
     return true;
   }
 
@@ -1387,6 +1482,53 @@
   llvm::DenseMap<Metadata *, TIInfo> TypeIdInfo;
   unsigned I = 0;
   SmallVector<MDNode *, 2> Types;
+
+  struct ExportedFunctionInfo {
+    CfiFunctionLinkage Linkage;
+    MDNode *FuncMD; // {name, linkage, type[, type...]}
+  };
+  DenseMap<StringRef, ExportedFunctionInfo> ExportedFunctions;
+  if (ExportSummary) {
+    NamedMDNode *CfiFunctionsMD = M.getNamedMetadata("cfi.functions");
+    if (CfiFunctionsMD) {
+      for (auto FuncMD : CfiFunctionsMD->operands()) {
+        assert(FuncMD->getNumOperands() >= 2);
+        StringRef FunctionName =
+            cast<MDString>(FuncMD->getOperand(0))->getString();
+        if (!IsFunctionLive(FunctionName, ExportSummary)) {
+          continue;
+        }
+        CfiFunctionLinkage Linkage = static_cast<CfiFunctionLinkage>(
+            cast<ConstantAsMetadata>(FuncMD->getOperand(1))
+                ->getValue()
+                ->getUniqueInteger()
+                .getZExtValue());
+        auto P = ExportedFunctions.insert({FunctionName, {Linkage, FuncMD}});
+        if (!P.second && P.first->second.Linkage != CFL_Definition)
+          P.first->second = {Linkage, FuncMD};
+      }
+
+      for (const auto &P : ExportedFunctions) {
+        StringRef FunctionName = P.first;
+        CfiFunctionLinkage Linkage = P.second.Linkage;
+        MDNode *FuncMD = P.second.FuncMD;
+        Function *F = M.getFunction(FunctionName);
+        if (!F)
+          F = Function::Create(
+              FunctionType::get(Type::getVoidTy(M.getContext()), false),
+              GlobalVariable::ExternalLinkage, FunctionName, &M);
+
+        if (Linkage == CFL_WeakDeclaration)
+          F->setLinkage(GlobalValue::ExternalWeakLinkage);
+
+        SmallVector<MDNode *, 2> Types;
+        for (unsigned I = 2; I < FuncMD->getNumOperands(); ++I)
+          F->addMetadata(LLVMContext::MD_type,
+                         *cast<MDNode>(FuncMD->getOperand(I).get()));
+      }
+    }
+  }
+
   for (GlobalObject &GO : M.global_objects()) {
     if (isa<GlobalVariable>(GO) && GO.isDeclarationForLinker())
       continue;
@@ -1396,7 +1538,15 @@
     if (Types.empty())
       continue;
 
-    auto *GTM = GlobalTypeMember::create(Alloc, &GO, Types);
+    bool IsDefinition = !GO.isDeclarationForLinker();
+    bool IsExported = false;
+    if (isa<Function>(GO) && ExportedFunctions.count(GO.getName())) {
+      IsDefinition |= ExportedFunctions[GO.getName()].Linkage == CFL_Definition;
+      IsExported = true;
+    }
+
+    auto *GTM =
+        GlobalTypeMember::create(Alloc, &GO, IsDefinition, IsExported, Types);
     for (MDNode *Type : Types) {
       verifyTypeMDNode(&GO, Type);
       auto &Info = TypeIdInfo[cast<MDNode>(Type)->getOperand(1)];
Index: lib/Transforms/IPO/ThinLTOBitcodeWriter.cpp
===================================================================
--- lib/Transforms/IPO/ThinLTOBitcodeWriter.cpp
+++ lib/Transforms/IPO/ThinLTOBitcodeWriter.cpp
@@ -32,7 +32,8 @@
 
 // Promote each local-linkage entity defined by ExportM and used by ImportM by
 // changing visibility and appending the given ModuleId.
-void promoteInternals(Module &ExportM, Module &ImportM, StringRef ModuleId) {
+void promoteInternals(Module &ExportM, Module &ImportM, StringRef ModuleId,
+                      SetVector<GlobalValue *> &PromoteExtra) {
   DenseMap<const Comdat *, Comdat *> RenamedComdats;
   for (auto &ExportGV : ExportM.global_values()) {
     if (!ExportGV.hasLocalLinkage())
@@ -40,7 +41,7 @@
 
     auto Name = ExportGV.getName();
     GlobalValue *ImportGV = ImportM.getNamedValue(Name);
-    if (!ImportGV || ImportGV->use_empty())
+    if ((!ImportGV || ImportGV->use_empty()) && !PromoteExtra.count(&ExportGV))
       continue;
 
     std::string NewName = (Name + ModuleId).str();
@@ -53,8 +54,10 @@
     ExportGV.setLinkage(GlobalValue::ExternalLinkage);
     ExportGV.setVisibility(GlobalValue::HiddenVisibility);
 
-    ImportGV->setName(NewName);
-    ImportGV->setVisibility(GlobalValue::HiddenVisibility);
+    if (ImportGV) {
+      ImportGV->setName(NewName);
+      ImportGV->setVisibility(GlobalValue::HiddenVisibility);
+    }
   }
 
   if (!RenamedComdats.empty())
@@ -296,6 +299,11 @@
       F.setComdat(nullptr);
     }
 
+  SetVector<GlobalValue *> CfiFunctions;
+  for (auto &F : M)
+    if ((!F.hasLocalLinkage() || F.hasAddressTaken()) && HasTypeMetadata(&F))
+      CfiFunctions.insert(&F);
+
   // Remove all globals with type metadata, globals with comdats that live in
   // MergedM, and aliases pointing to such globals from the thin LTO module.
   filterModule(&M, [&](const GlobalValue *GV) {
@@ -308,11 +316,39 @@
     return true;
   });
 
-  promoteInternals(*MergedM, M, ModuleId);
-  promoteInternals(M, *MergedM, ModuleId);
+  promoteInternals(*MergedM, M, ModuleId, CfiFunctions);
+  promoteInternals(M, *MergedM, ModuleId, CfiFunctions);
+
+  SmallVector<MDNode *, 8> CfiFunctionMDs;
+  for (auto V : CfiFunctions) {
+    Function &F = *cast<Function>(V);
+    SmallVector<MDNode *, 2> Types;
+    F.getMetadata(LLVMContext::MD_type, Types);
+
+    auto &Ctx = MergedM->getContext();
+    SmallVector<Metadata *, 4> Elts;
+    Elts.push_back(MDString::get(Ctx, F.getName()));
+    CfiFunctionLinkage Linkage;
+    if (!F.isDeclarationForLinker())
+      Linkage = CFL_Definition;
+    else if (F.isWeakForLinker())
+      Linkage = CFL_WeakDeclaration;
+    else
+      Linkage = CFL_Declaration;
+    Elts.push_back(ConstantAsMetadata::get(
+        llvm::ConstantInt::get(Type::getInt8Ty(Ctx), Linkage)));
+    for (auto Type : Types)
+      Elts.push_back(Type);
+    CfiFunctionMDs.push_back(MDTuple::get(Ctx, Elts));
+  }
 
-  simplifyExternals(*MergedM);
+  if(!CfiFunctionMDs.empty()) {
+    NamedMDNode *NMD = MergedM->getOrInsertNamedMetadata("cfi.functions");
+    for (auto MD : CfiFunctionMDs)
+      NMD->addOperand(MD);
+  }
 
+  simplifyExternals(*MergedM);
 
   // FIXME: Try to re-use BSI and PFI from the original module here.
   ProfileSummaryInfo PSI(M);
Index: test/ThinLTO/X86/cfi-icall.ll
===================================================================
--- /dev/null
+++ test/ThinLTO/X86/cfi-icall.ll
@@ -0,0 +1,29 @@
+; RUN: opt -thinlto-bc %s -o %t1.bc
+; RUN: llvm-lto2 run  -thinlto-distributed-indexes %t1.bc -o %t.out -save-temps \
+; RUN:   -r %t1.bc,foo,plx \
+; RUN:   -r %t1.bc,bar,x
+; RUN: llvm-bcanalyzer -dump %t.out.index.bc | FileCheck %s --check-prefix=COMBINED
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define i1 @foo(i8* %p) !type !0 {
+entry:
+  %x = call i1 @llvm.type.test(i8* %p, metadata !"typeid1")
+  ret i1 %x
+}
+
+declare !type !0 void @bar()
+
+declare i1 @llvm.type.test(i8* %ptr, metadata %type) nounwind readnone
+
+!0 = !{i64 0, !"typeid1"}
+
+; COMBINED:   <GLOBALVAL_SUMMARY_BLOCK
+; COMBINED:     <CFI_FUNCTION_DEFS op0=0 op1=3/>
+; COMBINED:     <CFI_FUNCTION_DECLS op0=3 op1=3/>
+; COMBINED:   </GLOBALVAL_SUMMARY_BLOCK>
+
+; COMBINED:      <STRTAB_BLOCK
+; COMBINED-NEXT:   <BLOB abbrevid=4/> blob data = 'foobar'
+; COMBINED-NEXT: </STRTAB_BLOCK>
Index: test/Transforms/CrossDSOCFI/cfi_functions.ll
===================================================================
--- /dev/null
+++ test/Transforms/CrossDSOCFI/cfi_functions.ll
@@ -0,0 +1,23 @@
+; Test that types referenced in ThinLTO-style !cfi.functions are known to __cfi_check.
+; RUN: opt -S -cross-dso-cfi < %s | FileCheck %s
+; RUN: opt -S -passes=cross-dso-cfi < %s | FileCheck %s
+
+; CHECK:      define void @__cfi_check(
+; CHECK:        switch i64
+; CHECK-NEXT:     i64 1234, label
+; CHECK-NEXT:     i64 5678, label
+; CHECK-NEXT:   ]
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+!cfi.functions = !{!0, !1}
+!llvm.module.flags = !{!6}
+
+!0 = !{!"f", i8 0, !2, !4}
+!1 = !{!"g", i8 1, !3, !5}
+!2 = !{i64 0, !"typeid1"}
+!3 = !{i64 0, !"typeid2"}
+!4 = !{i64 0, i64 1234}
+!5 = !{i64 0, i64 5678}
+!6 = !{i32 4, !"Cross-DSO CFI", i32 1}
Index: test/Transforms/LowerTypeTests/Inputs/import-icall.yaml
===================================================================
--- /dev/null
+++ test/Transforms/LowerTypeTests/Inputs/import-icall.yaml
@@ -0,0 +1,19 @@
+---
+TypeIdMap:
+  typeid1:
+    TTRes:
+      Kind:            AllOnes
+      SizeM1BitWidth:  7
+  typeid2:
+    TTRes:
+      Kind:            Single
+      SizeM1BitWidth:  0
+WithGlobalValueDeadStripping: false
+CfiFunctionDefs:
+  - local_a
+  - local_b
+  - does_not_exist
+CfiFunctionDecls:
+  - external
+  - external_weak
+...
Index: test/Transforms/LowerTypeTests/export-icall.ll
===================================================================
--- /dev/null
+++ test/Transforms/LowerTypeTests/export-icall.ll
@@ -0,0 +1,62 @@
+; RUN: opt -S -lowertypetests -lowertypetests-summary-action=export -lowertypetests-read-summary=%S/Inputs/use-typeid1-typeid2.yaml -lowertypetests-write-summary=%t < %s | FileCheck %s
+; RUN: FileCheck --check-prefix=SUMMARY %s < %t
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @h(i8 %x) !type !2 {
+  ret void
+}
+
+!cfi.functions = !{!0, !1, !3, !4, !5, !6}
+
+!0 = !{!"h", i8 0, !2}
+!1 = !{!"h", i8 1, !2}
+!2 = !{i64 0, !"typeid1"}
+!3 = !{!"f", i8 0, !2}
+!4 = !{!"external", i8 1, !2}
+!5 = !{!"external_weak", i8 2, !2}
+!6 = !{!"g", i8 0, !7}
+!7 = !{i64 0, !"typeid2"}
+
+
+; CHECK: @__typeid_typeid1_global_addr = hidden alias i8, bitcast (void ()* [[JT1:.*]] to i8*)
+; CHECK: @__typeid_typeid1_align = hidden alias i8, inttoptr (i8 3 to i8*)
+; CHECK: @__typeid_typeid1_size_m1 = hidden alias i8, inttoptr (i64 3 to i8*)
+
+; CHECK: @h                    = alias void (i8), bitcast (void ()* [[JT1]] to void (i8)*)
+; CHECK: @external.cfi_jt      = hidden alias void (), {{.*}}getelementptr {{.*}}void ()* [[JT1]]
+; CHECK: @f                    = alias void (), {{.*}}getelementptr {{.*}}void ()* [[JT1]]
+; CHECK: @external_weak.cfi_jt = hidden alias void (), {{.*}}getelementptr {{.*}}void ()* [[JT1]]
+
+; CHECK: @__typeid_typeid2_global_addr = hidden alias i8, bitcast (void ()* [[JT2:.*]] to i8*)
+
+; CHECK: @g                    = alias void (), void ()* [[JT2]]
+
+; CHECK: define internal void @h.cfi(i8 {{.*}}) !type !{{.*}}
+; CHECK: declare !type !{{.*}} void @external()
+; CHECK: declare !type !{{.*}} void @f.cfi()
+; CHECK: declare !type !{{.*}} void @external_weak()
+; CHECK: declare !type !{{.*}} void @g.cfi()
+
+
+; SUMMARY:      TypeIdMap:
+; SUMMARY-NEXT:   typeid1:
+; SUMMARY-NEXT:     TTRes:
+; SUMMARY-NEXT:       Kind:            AllOnes
+; SUMMARY-NEXT:       SizeM1BitWidth:  7
+; SUMMARY-NEXT:     WPDRes:
+; SUMMARY-NEXT:   typeid2:
+; SUMMARY-NEXT:     TTRes:
+; SUMMARY-NEXT:       Kind:            Single
+; SUMMARY-NEXT:       SizeM1BitWidth:  0
+; SUMMARY-NEXT:     WPDRes:
+
+; SUMMARY:      CfiFunctionDefs:
+; SUMMARY-NEXT:   - f
+; SUMMARY-NEXT:   - g
+; SUMMARY-NEXT:   - h
+; SUMMARY-NEXT: CfiFunctionDecls:
+; SUMMARY-NEXT:   - external
+; SUMMARY-NEXT:   - external_weak
+; SUMMARY-NEXT: ...
Index: test/Transforms/LowerTypeTests/import-icall.ll
===================================================================
--- /dev/null
+++ test/Transforms/LowerTypeTests/import-icall.ll
@@ -0,0 +1,40 @@
+; RUN: opt -S -lowertypetests -lowertypetests-summary-action=import -lowertypetests-read-summary=%S/Inputs/import-icall.yaml < %s | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define i8 @local_a() {
+  call void @external()
+  call void @external_weak()
+  ret i8 1
+}
+
+define internal i8 @local_b() {
+  %x = call i8 @local_a()
+  ret i8 %x
+}
+
+define i8 @use_b() {
+  %x = call i8 @local_b()
+  ret i8 %x
+}
+
+
+declare void @external()
+declare extern_weak void @external_weak()
+
+; CHECK:      define hidden i8 @local_a.cfi() {
+; CHECK-NEXT:   call void @external.cfi_jt()
+; CHECK-NEXT:   call void select (i1 icmp ne (void ()* @external_weak, void ()* null), void ()* @external_weak.cfi_jt, void ()* null)()
+; CHECK-NEXT:   ret i8 1
+; CHECK-NEXT: }
+
+; internal @local_b is not the same function as "local_b" in the summary.
+; CHECK:      define internal i8 @local_b() {
+; CHECK-NEXT:   call i8 @local_a()
+
+; CHECK: declare void @external()
+; CHECK: declare extern_weak void @external_weak()
+; CHECK: declare i8 @local_a()
+; CHECK: declare hidden void @external.cfi_jt()
+; CHECK: declare hidden void @external_weak.cfi_jt()
Index: tools/llvm-bcanalyzer/llvm-bcanalyzer.cpp
===================================================================
--- tools/llvm-bcanalyzer/llvm-bcanalyzer.cpp
+++ tools/llvm-bcanalyzer/llvm-bcanalyzer.cpp
@@ -320,6 +320,8 @@
       STRINGIFY_CODE(FS, TYPE_TEST_ASSUME_CONST_VCALL)
       STRINGIFY_CODE(FS, TYPE_CHECKED_LOAD_CONST_VCALL)
       STRINGIFY_CODE(FS, VALUE_GUID)
+      STRINGIFY_CODE(FS, CFI_FUNCTION_DEFS)
+      STRINGIFY_CODE(FS, CFI_FUNCTION_DECLS)
     }
   case bitc::METADATA_ATTACHMENT_ID:
     switch(CodeID) {