LTO: introduce object file-based on-disk module format.

This format is simply a regular object file with the bitcode stored in a
section named ".llvmbc", plus any number of other (non-allocated) sections.

One immediate use case for this is to accommodate compilation processes
which expect the object file to contain metadata in non-allocated sections,
such as the ".go_export" section used by some Go compilers [1], although I
imagine that in the future we could consider compiling parts of the module
(such as large non-inlinable functions) directly into the object file to
improve LTO efficiency.

[1] http://golang.org/doc/install/gccgo#Imports

Differential Revision: http://reviews.llvm.org/D4371

llvm-svn: 218078

llvm/docs/BitCodeFormat.rst

@@ -28,8 +28,9 @@ Unlike XML, the bitstream format is a binary encoding, and unlike XML it
 provides a mechanism for the file to self-describe "abbreviations", which are
 effectively size optimizations for the content.
 
-LLVM IR files may be optionally embedded into a `wrapper`_ structure that makes
-it easy to embed extra data along with LLVM IR files.
+LLVM IR files may be optionally embedded into a `wrapper`_ structure, or in a
+`native object file`_. Both of these mechanisms make it easy to embed extra
+data along with LLVM IR files.
 
 This document first describes the LLVM bitstream format, describes the wrapper
 format, then describes the record structure used by LLVM IR files.
@@ -460,6 +461,19 @@ to the start of the bitcode stream in the file, and the Size field is the size
 in bytes of the stream. CPUType is a target-specific value that can be used to
 encode the CPU of the target.
 
+.. _native object file:
+
+Native Object File Wrapper Format
+=================================
+
+Bitcode files for LLVM IR may also be wrapped in a native object file
+(i.e. ELF, COFF, Mach-O).  The bitcode must be stored in a section of the
+object file named ``.llvmbc``.  This wrapper format is useful for accommodating
+LTO in compilation pipelines where intermediate objects must be native object
+files which contain metadata in other sections.
+
+Not all tools support this format.
+
 .. _encoding of LLVM IR:
 
 LLVM IR Encoding

llvm/docs/ReleaseNotes.rst

@@ -43,6 +43,9 @@ Non-comprehensive list of changes in this release
 
 * Support for AuroraUX has been removed.
 
+* Added support for a `native object file-based bitcode wrapper format
+  <BitCodeFormat.html#native-object-file>`_.
+
 * ... next change ...
 
 .. NOTE

llvm/include/llvm/Object/Error.h

@@ -26,7 +26,8 @@ enum class object_error {
   arch_not_found,
   invalid_file_type,
   parse_failed,
-  unexpected_eof
+  unexpected_eof,
+  bitcode_section_not_found,
 };
 
 inline std::error_code make_error_code(object_error e) {

llvm/include/llvm/Object/IRObjectFile.h

@@ -22,6 +22,8 @@ class Module;
 class GlobalValue;
 
 namespace object {
+class ObjectFile;
+
 class IRObjectFile : public SymbolicFile {
   std::unique_ptr<Module> M;
   std::unique_ptr<Mangler> Mang;
@@ -49,6 +51,16 @@ public:
     return v->isIR();
   }
 
+  /// \brief Finds and returns bitcode embedded in the given object file, or an
+  /// error code if not found.
+  static ErrorOr<MemoryBufferRef> findBitcodeInObject(const ObjectFile &Obj);
+
+  /// \brief Finds and returns bitcode in the given memory buffer (which may
+  /// be either a bitcode file or a native object file with embedded bitcode),
+  /// or an error code if not found.
+  static ErrorOr<MemoryBufferRef>
+  findBitcodeInMemBuffer(MemoryBufferRef Object);
+
   static ErrorOr<std::unique_ptr<IRObjectFile>>
   createIRObjectFile(MemoryBufferRef Object, LLVMContext &Context);
 };

llvm/lib/LTO/LTOModule.cpp

@@ -29,6 +29,8 @@
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/MC/MCTargetAsmParser.h"
 #include "llvm/MC/SubtargetFeature.h"
+#include "llvm/Object/IRObjectFile.h"
+#include "llvm/Object/ObjectFile.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/Host.h"
@@ -44,6 +46,7 @@
 #include "llvm/Transforms/Utils/GlobalStatus.h"
 #include <system_error>
 using namespace llvm;
+using namespace llvm::object;
 
 LTOModule::LTOModule(std::unique_ptr<object::IRObjectFile> Obj,
                      llvm::TargetMachine *TM)
@@ -51,23 +54,31 @@ LTOModule::LTOModule(std::unique_ptr<object::IRObjectFile> Obj,
 
 /// isBitcodeFile - Returns 'true' if the file (or memory contents) is LLVM
 /// bitcode.
-bool LTOModule::isBitcodeFile(const void *mem, size_t length) {
-  return sys::fs::identify_magic(StringRef((const char *)mem, length)) ==
-         sys::fs::file_magic::bitcode;
+bool LTOModule::isBitcodeFile(const void *Mem, size_t Length) {
+  ErrorOr<MemoryBufferRef> BCData = IRObjectFile::findBitcodeInMemBuffer(
+      MemoryBufferRef(StringRef((const char *)Mem, Length), "<mem>"));
+  return bool(BCData);
 }
 
-bool LTOModule::isBitcodeFile(const char *path) {
-  sys::fs::file_magic type;
-  if (sys::fs::identify_magic(path, type))
+bool LTOModule::isBitcodeFile(const char *Path) {
+  ErrorOr<std::unique_ptr<MemoryBuffer>> BufferOrErr =
+      MemoryBuffer::getFile(Path);
+  if (!BufferOrErr)
     return false;
-  return type == sys::fs::file_magic::bitcode;
+
+  ErrorOr<MemoryBufferRef> BCData = IRObjectFile::findBitcodeInMemBuffer(
+      BufferOrErr.get()->getMemBufferRef());
+  return bool(BCData);
 }
 
-bool LTOModule::isBitcodeForTarget(MemoryBuffer *buffer,
-                                   StringRef triplePrefix) {
-  std::string Triple =
-      getBitcodeTargetTriple(buffer->getMemBufferRef(), getGlobalContext());
-  return StringRef(Triple).startswith(triplePrefix);
+bool LTOModule::isBitcodeForTarget(MemoryBuffer *Buffer,
+                                   StringRef TriplePrefix) {
+  ErrorOr<MemoryBufferRef> BCOrErr =
+      IRObjectFile::findBitcodeInMemBuffer(Buffer->getMemBufferRef());
+  if (!BCOrErr)
+    return false;
+  std::string Triple = getBitcodeTargetTriple(*BCOrErr, getGlobalContext());
+  return StringRef(Triple).startswith(TriplePrefix);
 }
 
 LTOModule *LTOModule::createFromFile(const char *path, TargetOptions options,
@@ -113,7 +124,13 @@ LTOModule *LTOModule::createFromBuffer(const void *mem, size_t length,
 LTOModule *LTOModule::makeLTOModule(MemoryBufferRef Buffer,
                                     TargetOptions options,
                                     std::string &errMsg) {
-  ErrorOr<Module *> MOrErr = parseBitcodeFile(Buffer, getGlobalContext());
+  ErrorOr<MemoryBufferRef> MBOrErr =
+      IRObjectFile::findBitcodeInMemBuffer(Buffer);
+  if (std::error_code EC = MBOrErr.getError()) {
+    errMsg = EC.message();
+    return nullptr;
+  }
+  ErrorOr<Module *> MOrErr = parseBitcodeFile(*MBOrErr, getGlobalContext());
   if (std::error_code EC = MOrErr.getError()) {
     errMsg = EC.message();
     return nullptr;

llvm/lib/Object/Error.cpp

@@ -41,6 +41,8 @@ std::string _object_error_category::message(int EV) const {
     return "Invalid data was encountered while parsing the file";
   case object_error::unexpected_eof:
     return "The end of the file was unexpectedly encountered";
+  case object_error::bitcode_section_not_found:
+    return "Bitcode section not found in object file";
   }
   llvm_unreachable("An enumerator of object_error does not have a message "
                    "defined.");

llvm/lib/Object/IRObjectFile.cpp

@@ -25,6 +25,7 @@
 #include "llvm/MC/MCObjectFileInfo.h"
 #include "llvm/MC/MCTargetAsmParser.h"
 #include "llvm/MC/MCParser/MCAsmParser.h"
+#include "llvm/Object/ObjectFile.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/SourceMgr.h"
 #include "llvm/Support/TargetRegistry.h"
@@ -264,11 +265,50 @@ basic_symbol_iterator IRObjectFile::symbol_end_impl() const {
   return basic_symbol_iterator(BasicSymbolRef(Ret, this));
 }
 
+ErrorOr<MemoryBufferRef> IRObjectFile::findBitcodeInObject(const ObjectFile &Obj) {
+  for (const SectionRef &Sec : Obj.sections()) {
+    StringRef SecName;
+    if (std::error_code EC = Sec.getName(SecName))
+      return EC;
+    if (SecName == ".llvmbc") {
+      StringRef SecContents;
+      if (std::error_code EC = Sec.getContents(SecContents))
+        return EC;
+      return MemoryBufferRef(SecContents, Obj.getFileName());
+    }
+  }
+
+  return object_error::bitcode_section_not_found;
+}
+
+ErrorOr<MemoryBufferRef> IRObjectFile::findBitcodeInMemBuffer(MemoryBufferRef Object) {
+  sys::fs::file_magic Type = sys::fs::identify_magic(Object.getBuffer());
+  switch (Type) {
+  case sys::fs::file_magic::bitcode:
+    return Object;
+  case sys::fs::file_magic::elf_relocatable:
+  case sys::fs::file_magic::macho_object:
+  case sys::fs::file_magic::coff_object: {
+    ErrorOr<std::unique_ptr<ObjectFile>> ObjFile =
+        ObjectFile::createObjectFile(Object, Type);
+    if (!ObjFile)
+      return ObjFile.getError();
+    return findBitcodeInObject(*ObjFile->get());
+  }
+  default:
+    return object_error::invalid_file_type;
+  }
+}
+
 ErrorOr<std::unique_ptr<IRObjectFile>>
 llvm::object::IRObjectFile::createIRObjectFile(MemoryBufferRef Object,
                                                LLVMContext &Context) {
+  ErrorOr<MemoryBufferRef> BCOrErr = findBitcodeInMemBuffer(Object);
+  if (!BCOrErr)
+    return BCOrErr.getError();
 
-  std::unique_ptr<MemoryBuffer> Buff(MemoryBuffer::getMemBuffer(Object, false));
+  std::unique_ptr<MemoryBuffer> Buff(
+      MemoryBuffer::getMemBuffer(BCOrErr.get(), false));
 
   ErrorOr<Module *> MOrErr = getLazyBitcodeModule(std::move(Buff), Context);
   if (std::error_code EC = MOrErr.getError())

llvm/lib/Object/SymbolicFile.cpp

@@ -40,11 +40,9 @@ ErrorOr<std::unique_ptr<SymbolicFile>> SymbolicFile::createSymbolicFile(
   case sys::fs::file_magic::macho_universal_binary:
   case sys::fs::file_magic::windows_resource:
     return object_error::invalid_file_type;
-  case sys::fs::file_magic::elf_relocatable:
   case sys::fs::file_magic::elf_executable:
   case sys::fs::file_magic::elf_shared_object:
   case sys::fs::file_magic::elf_core:
-  case sys::fs::file_magic::macho_object:
   case sys::fs::file_magic::macho_executable:
   case sys::fs::file_magic::macho_fixed_virtual_memory_shared_lib:
   case sys::fs::file_magic::macho_core:
@@ -54,10 +52,26 @@ ErrorOr<std::unique_ptr<SymbolicFile>> SymbolicFile::createSymbolicFile(
   case sys::fs::file_magic::macho_bundle:
   case sys::fs::file_magic::macho_dynamically_linked_shared_lib_stub:
   case sys::fs::file_magic::macho_dsym_companion:
-  case sys::fs::file_magic::coff_object:
   case sys::fs::file_magic::coff_import_library:
   case sys::fs::file_magic::pecoff_executable:
     return ObjectFile::createObjectFile(Object, Type);
+  case sys::fs::file_magic::elf_relocatable:
+  case sys::fs::file_magic::macho_object:
+  case sys::fs::file_magic::coff_object: {
+    ErrorOr<std::unique_ptr<ObjectFile>> Obj =
+        ObjectFile::createObjectFile(Object, Type);
+    if (!Obj || !Context)
+      return std::move(Obj);
+
+    ErrorOr<MemoryBufferRef> BCData =
+        IRObjectFile::findBitcodeInObject(*Obj->get());
+    if (!BCData)
+      return std::move(Obj);
+
+    return IRObjectFile::createIRObjectFile(
+        MemoryBufferRef(BCData->getBuffer(), Object.getBufferIdentifier()),
+        *Context);
+  }
   }
   llvm_unreachable("Unexpected Binary File Type");
 }

llvm/test/LTO/Inputs/bcsection.macho.s

@@ -0,0 +1,2 @@
+.section .llvmbc,.llvmbc
+.incbin "bcsection.bc"

llvm/test/LTO/Inputs/bcsection.s

@@ -0,0 +1,2 @@
+.section .llvmbc
+.incbin "bcsection.bc"

llvm/test/LTO/bcsection.ll

@@ -0,0 +1,21 @@
+; RUN: llvm-as -o %T/bcsection.bc %s
+
+; RUN: llvm-mc -I=%T -filetype=obj -triple=x86_64-pc-win32 -o %T/bcsection.coff.bco %p/Inputs/bcsection.s
+; RUN: llvm-nm %T/bcsection.coff.bco | FileCheck %s
+; RUN: llvm-lto -exported-symbol=main -o %T/bcsection.coff.o %T/bcsection.coff.bco
+; RUN: llvm-nm %T/bcsection.coff.o | FileCheck %s
+
+; RUN: llvm-mc -I=%T -filetype=obj -triple=x86_64-unknown-linux-gnu -o %T/bcsection.elf.bco %p/Inputs/bcsection.s
+; RUN: llvm-nm %T/bcsection.elf.bco | FileCheck %s
+; RUN: llvm-lto -exported-symbol=main -o %T/bcsection.elf.o %T/bcsection.elf.bco
+; RUN: llvm-nm %T/bcsection.elf.o | FileCheck %s
+
+; RUN: llvm-mc -I=%T -filetype=obj -triple=x86_64-apple-darwin11 -o %T/bcsection.macho.bco %p/Inputs/bcsection.macho.s
+; RUN: llvm-nm %T/bcsection.macho.bco | FileCheck %s
+; RUN: llvm-lto -exported-symbol=main -o %T/bcsection.macho.o %T/bcsection.macho.bco
+; RUN: llvm-nm %T/bcsection.macho.o | FileCheck %s
+
+; CHECK: main
+define i32 @main() {
+  ret i32 0
+}

llvm/test/tools/gold/Inputs/bcsection.s

@@ -0,0 +1,2 @@
+.section .llvmbc
+.incbin "bcsection.bc"

llvm/test/tools/gold/bcsection.ll

@@ -0,0 +1,11 @@
+; RUN: llvm-as -o %T/bcsection.bc %s
+
+; RUN: llvm-mc -I=%T -filetype=obj -o %T/bcsection.bco %p/Inputs/bcsection.s
+; RUN: llvm-nm -no-llvm-bc %T/bcsection.bco | count 0
+; RUN: ld -r -o %T/bcsection.o -plugin %llvmshlibdir/LLVMgold.so %T/bcsection.bco
+; RUN: llvm-nm -no-llvm-bc %T/bcsection.o | FileCheck %s
+
+; CHECK: main
+define i32 @main() {
+  ret i32 0
+}

llvm/tools/gold/gold-plugin.cpp

@@ -300,7 +300,9 @@ static ld_plugin_status claim_file_hook(const ld_plugin_input_file *file,
   ErrorOr<std::unique_ptr<object::IRObjectFile>> ObjOrErr =
       object::IRObjectFile::createIRObjectFile(BufferRef, Context);
   std::error_code EC = ObjOrErr.getError();
-  if (EC == BitcodeError::InvalidBitcodeSignature)
+  if (EC == BitcodeError::InvalidBitcodeSignature ||
+      EC == object::object_error::invalid_file_type ||
+      EC == object::object_error::bitcode_section_not_found)
     return LDPS_OK;
 
   *claimed = 1;
@@ -548,8 +550,15 @@ getModuleForFile(LLVMContext &Context, claimed_file &F, raw_fd_ostream *ApiFile,
   if (get_view(F.handle, &View) != LDPS_OK)
     message(LDPL_FATAL, "Failed to get a view of file");
 
-  std::unique_ptr<MemoryBuffer> Buffer = MemoryBuffer::getMemBuffer(
-      StringRef((char *)View, File.filesize), "", false);
+  llvm::ErrorOr<MemoryBufferRef> MBOrErr =
+      object::IRObjectFile::findBitcodeInMemBuffer(
+          MemoryBufferRef(StringRef((const char *)View, File.filesize), ""));
+  if (std::error_code EC = MBOrErr.getError())
+    message(LDPL_FATAL, "Could not read bitcode from file : %s",
+            EC.message().c_str());
+
+  std::unique_ptr<MemoryBuffer> Buffer =
+      MemoryBuffer::getMemBuffer(MBOrErr->getBuffer(), "", false);
 
   if (release_input_file(F.handle) != LDPS_OK)
     message(LDPL_FATAL, "Failed to release file information");

llvm/tools/llvm-nm/llvm-nm.cpp

@@ -149,6 +149,9 @@ cl::list<std::string> SegSect("s", cl::Positional, cl::ZeroOrMore,
 cl::opt<bool> FormatMachOasHex("x", cl::desc("Print symbol entry in hex, "
                                              "Mach-O only"));
 
+cl::opt<bool> NoLLVMBitcode("no-llvm-bc",
+                            cl::desc("Disable LLVM bitcode reader"));
+
 bool PrintAddress = true;
 
 bool MultipleFiles = false;
@@ -1009,8 +1012,8 @@ static void dumpSymbolNamesFromFile(std::string &Filename) {
     return;
 
   LLVMContext &Context = getGlobalContext();
-  ErrorOr<std::unique_ptr<Binary>> BinaryOrErr =
-      createBinary(BufferOrErr.get()->getMemBufferRef(), &Context);
+  ErrorOr<std::unique_ptr<Binary>> BinaryOrErr = createBinary(
+      BufferOrErr.get()->getMemBufferRef(), NoLLVMBitcode ? nullptr : &Context);
   if (error(BinaryOrErr.getError(), Filename))
     return;
   Binary &Bin = *BinaryOrErr.get();