[llvm-tapi] initial commit, supports ELF text stubs

http://lists.llvm.org/pipermail/llvm-dev/2018-September/126472.html

TextAPI is a library and accompanying tool that allows conversion between binary shared object stubs and textual counterparts. The motivations and uses cases for this are explained thoroughly in the llvm-dev proposal [1]. This initial commit proposes a potential structure for the TAPI library, also including support for reading/writing text-based ELF stubs (.tbe) in addition to preliminary support for reading binary ELF files. The goal for this patch is to ensure the project architecture appropriately welcomes integration of Mach-O stubbing from Apple's TAPI [2].

Added:

 - TextAPI library
 - .tbe read support
 - .tbe write (to raw_ostream) support

[1] http://lists.llvm.org/pipermail/llvm-dev/2018-September/126472.html
[2] https://github.com/ributzka/tapi

Differential Revision: https://reviews.llvm.org/D53051

llvm-svn: 348170

llvm/include/llvm/TextAPI/ELF/ELFStub.h

@@ -0,0 +1,69 @@
+//===- ELFStub.h ------------------------------------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===-----------------------------------------------------------------------===/
+///
+/// \file
+/// This file defines an internal representation of an ELF stub.
+///
+//===-----------------------------------------------------------------------===/
+
+#ifndef LLVM_TEXTAPI_ELF_ELFSTUB_H
+#define LLVM_TEXTAPI_ELF_ELFSTUB_H
+
+#include "llvm/BinaryFormat/ELF.h"
+#include "llvm/Support/VersionTuple.h"
+#include <vector>
+#include <set>
+
+namespace llvm {
+namespace elfabi {
+
+typedef uint16_t ELFArch;
+
+enum class ELFSymbolType {
+  NoType = ELF::STT_NOTYPE,
+  Object = ELF::STT_OBJECT,
+  Func = ELF::STT_FUNC,
+  TLS = ELF::STT_TLS,
+
+  // Type information is 4 bits, so 16 is safely out of range.
+  Unknown = 16,
+};
+
+struct ELFSymbol {
+  ELFSymbol(std::string SymbolName) : Name(SymbolName) {}
+  std::string Name;
+  uint64_t Size;
+  ELFSymbolType Type;
+  bool Undefined;
+  bool Weak;
+  Optional<std::string> Warning;
+  bool operator<(const ELFSymbol &RHS) const {
+    return Name < RHS.Name;
+  }
+};
+
+// A cumulative representation of ELF stubs.
+// Both textual and binary stubs will read into and write from this object.
+class ELFStub {
+// TODO: Add support for symbol versioning.
+public:
+  VersionTuple TbeVersion;
+  std::string SoName;
+  ELFArch Arch;
+  std::vector<std::string> NeededLibs;
+  std::set<ELFSymbol> Symbols;
+
+  ELFStub() {}
+  ELFStub(const ELFStub &Stub);
+  ELFStub(ELFStub &&Stub);
+};
+} // end namespace elfabi
+} // end namespace llvm
+
+#endif // LLVM_TEXTAPI_ELF_ELFSTUB_H

llvm/include/llvm/TextAPI/ELF/TBEHandler.h

@@ -0,0 +1,46 @@
+//===- TBEHandler.h ---------------------------------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===-----------------------------------------------------------------------===/
+///
+/// \file
+/// This file declares an interface for reading and writing .tbe (text-based
+/// ELF) files.
+///
+//===-----------------------------------------------------------------------===/
+
+#ifndef LLVM_TEXTAPI_ELF_TBEHANDLER_H
+#define LLVM_TEXTAPI_ELF_TBEHANDLER_H
+
+#include "llvm/Support/VersionTuple.h"
+#include <memory>
+
+namespace llvm {
+
+class raw_ostream;
+class Error;
+class StringRef;
+class VersionTuple;
+
+namespace elfabi {
+
+class ELFStub;
+
+const VersionTuple TBEVersionCurrent(1, 0);
+
+class TBEHandler {
+public:
+  /// Attempts to read an ELF interface file from a StringRef buffer.
+  std::unique_ptr<ELFStub> readFile(StringRef Buf);
+
+  /// Attempts to write an ELF interface file to a raw_ostream.
+  Error writeFile(raw_ostream &OS, const ELFStub &Stub);
+};
+} // end namespace elfabi
+} // end namespace llvm
+
+#endif // LLVM_TEXTAPI_ELF_TBEHANDLER_H

llvm/lib/CMakeLists.txt

@@ -23,6 +23,7 @@ add_subdirectory(AsmParser)
 add_subdirectory(LineEditor)
 add_subdirectory(ProfileData)
 add_subdirectory(Passes)
+add_subdirectory(TextAPI)
 add_subdirectory(ToolDrivers)
 add_subdirectory(XRay)
 add_subdirectory(Testing)

llvm/lib/LLVMBuild.txt

@@ -40,6 +40,7 @@ subdirectories =
  ProfileData
  Support
  TableGen
+ TextAPI
  Target
  Testing
  ToolDrivers

llvm/lib/TextAPI/CMakeLists.txt

@@ -0,0 +1,7 @@
+add_llvm_library(LLVMTextAPI
+  ELF/ELFStub.cpp
+  ELF/TBEHandler.cpp
+
+  ADDITIONAL_HEADER_DIRS
+  "${LLVM_MAIN_INCLUDE_DIR}/llvm/TextAPI"
+)

llvm/lib/TextAPI/ELF/ELFStub.cpp

@@ -0,0 +1,29 @@
+//===- ELFStub.cpp --------------------------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===-----------------------------------------------------------------------===/
+
+#include "llvm/TextAPI/ELF/ELFStub.h"
+
+using namespace llvm;
+using namespace llvm::elfabi;
+
+ELFStub::ELFStub(ELFStub const &Stub) {
+  TbeVersion = Stub.TbeVersion;
+  Arch = Stub.Arch;
+  SoName = Stub.SoName;
+  NeededLibs = Stub.NeededLibs;
+  Symbols = Stub.Symbols;
+}
+
+ELFStub::ELFStub(ELFStub &&Stub) {
+  TbeVersion = std::move(Stub.TbeVersion);
+  Arch = std::move(Stub.Arch);
+  SoName = std::move(Stub.SoName);
+  NeededLibs = std::move(Stub.NeededLibs);
+  Symbols = std::move(Stub.Symbols);
+}

llvm/lib/TextAPI/ELF/TBEHandler.cpp

@@ -0,0 +1,176 @@
+//===- TBEHandler.cpp -----------------------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===-----------------------------------------------------------------------===/
+
+#include "llvm/TextAPI/ELF/TBEHandler.h"
+#include "llvm/ADT/StringSwitch.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/Error.h"
+#include "llvm/Support/YAMLTraits.h"
+#include "llvm/TextAPI/ELF/ELFStub.h"
+
+using namespace llvm;
+using namespace llvm::elfabi;
+
+LLVM_YAML_STRONG_TYPEDEF(ELFArch, ELFArchMapper);
+
+namespace llvm {
+namespace yaml {
+
+/// YAML traits for ELFSymbolType.
+template <> struct ScalarEnumerationTraits<ELFSymbolType> {
+  static void enumeration(IO &IO, ELFSymbolType &SymbolType) {
+    IO.enumCase(SymbolType, "NoType", ELFSymbolType::NoType);
+    IO.enumCase(SymbolType, "Func", ELFSymbolType::Func);
+    IO.enumCase(SymbolType, "Object", ELFSymbolType::Object);
+    IO.enumCase(SymbolType, "TLS", ELFSymbolType::TLS);
+    IO.enumCase(SymbolType, "Unknown", ELFSymbolType::Unknown);
+    // Treat other symbol types as noise, and map to Unknown.
+    if (!IO.outputting() && IO.matchEnumFallback())
+      SymbolType = ELFSymbolType::Unknown;
+  }
+};
+
+/// YAML traits for ELFArch.
+template <> struct ScalarTraits<ELFArchMapper> {
+  static void output(const ELFArchMapper &Value, void *,
+                     llvm::raw_ostream &Out) {
+    // Map from integer to architecture string.
+    switch (Value) {
+    case (ELFArch)ELF::EM_X86_64:
+      Out << "x86_64";
+      break;
+    case (ELFArch)ELF::EM_AARCH64:
+      Out << "AArch64";
+      break;
+    case (ELFArch)ELF::EM_NONE:
+    default:
+      Out << "Unknown";
+    }
+  }
+
+  static StringRef input(StringRef Scalar, void *, ELFArchMapper &Value) {
+    // Map from architecture string to integer.
+    Value = StringSwitch<ELFArch>(Scalar)
+                .Case("x86_64", ELF::EM_X86_64)
+                .Case("AArch64", ELF::EM_AARCH64)
+                .Case("Unknown", ELF::EM_NONE)
+                .Default(ELF::EM_NONE);
+
+    // Returning empty StringRef indicates successful parse.
+    return StringRef();
+  }
+
+  // Don't place quotation marks around architecture value.
+  static QuotingType mustQuote(StringRef) { return QuotingType::None; }
+};
+
+/// YAML traits for TbeVersion.
+template <> struct ScalarTraits<VersionTuple> {
+  static void output(const VersionTuple &Value, void *,
+                     llvm::raw_ostream &Out) {
+    Out << Value.getAsString();
+  }
+
+  static StringRef input(StringRef Scalar, void *, VersionTuple &Value) {
+    if (Value.tryParse(Scalar))
+      return StringRef("Can't parse version: invalid version format.");
+
+    if (Value > TBEVersionCurrent)
+      return StringRef("Unsupported TBE version.");
+
+    // Returning empty StringRef indicates successful parse.
+    return StringRef();
+  }
+
+  // Don't place quotation marks around version value.
+  static QuotingType mustQuote(StringRef) { return QuotingType::None; }
+};
+
+/// YAML traits for ELFSymbol.
+template <> struct MappingTraits<ELFSymbol> {
+  static void mapping(IO &IO, ELFSymbol &Symbol) {
+    IO.mapRequired("Type", Symbol.Type);
+    // The need for symbol size depends on the symbol type.
+    if (Symbol.Type == ELFSymbolType::NoType) {
+      IO.mapOptional("Size", Symbol.Size, (uint64_t)0);
+    } else if (Symbol.Type == ELFSymbolType::Func) {
+      Symbol.Size = 0;
+    } else {
+      IO.mapRequired("Size", Symbol.Size);
+    }
+    IO.mapOptional("Undefined", Symbol.Undefined, false);
+    IO.mapOptional("Warning", Symbol.Warning);
+  }
+
+  // Compacts symbol information into a single line.
+  static const bool flow = true;
+};
+
+/// YAML traits for set of ELFSymbols.
+template <> struct CustomMappingTraits<std::set<ELFSymbol>> {
+  static void inputOne(IO &IO, StringRef Key, std::set<ELFSymbol> &Set) {
+    ELFSymbol Sym(Key.str());
+    IO.mapRequired(Key.str().c_str(), Sym);
+    Set.insert(Sym);
+  }
+
+  static void output(IO &IO, std::set<ELFSymbol> &Set) {
+    for (auto &Sym : Set)
+      IO.mapRequired(Sym.Name.c_str(), const_cast<ELFSymbol &>(Sym));
+  }
+};
+
+/// YAML traits for generic string vectors (i.e. list of needed libraries).
+template <> struct SequenceTraits<std::vector<std::string>> {
+  static size_t size(IO &IO, std::vector<std::string> &List) {
+    return List.size();
+  }
+
+  static std::string &element(IO &IO, std::vector<std::string> &List,
+                              size_t Index) {
+    if (Index >= List.size())
+      List.resize(Index + 1);
+    return List[Index];
+  }
+
+  // Compacts list of needed libraries into a single line.
+  static const bool flow = true;
+};
+
+/// YAML traits for ELFStub objects.
+template <> struct MappingTraits<ELFStub> {
+  static void mapping(IO &IO, ELFStub &Stub) {
+    if (!IO.mapTag("!tapi-tbe", true))
+      IO.setError("Not a .tbe YAML file.");
+    IO.mapRequired("TbeVersion", Stub.TbeVersion);
+    IO.mapRequired("SoName", Stub.SoName);
+    IO.mapRequired("Arch", (ELFArchMapper &)Stub.Arch);
+    IO.mapOptional("NeededLibs", Stub.NeededLibs);
+    IO.mapRequired("Symbols", Stub.Symbols);
+  }
+};
+
+} // end namespace yaml
+} // end namespace llvm
+
+std::unique_ptr<ELFStub> TBEHandler::readFile(StringRef Buf) {
+  yaml::Input YamlIn(Buf);
+  std::unique_ptr<ELFStub> Stub(new ELFStub());
+  YamlIn >> *Stub;
+  if (YamlIn.error())
+    return nullptr;
+  return Stub;
+}
+
+Error TBEHandler::writeFile(raw_ostream &OS, const ELFStub &Stub) {
+  yaml::Output YamlOut(OS, NULL, /*WrapColumn =*/0);
+
+  YamlOut << const_cast<ELFStub &>(Stub);
+  return Error::success();
+}

llvm/lib/TextAPI/LLVMBuild.txt

@@ -0,0 +1,22 @@
+;===- ./lib/TextAPI/LLVMBuild.txt ------------------------------*- Conf -*--===;
+;
+;                     The LLVM Compiler Infrastructure
+;
+; This file is distributed under the University of Illinois Open Source
+; License. See LICENSE.TXT for details.
+;
+;===------------------------------------------------------------------------===;
+;
+; This is an LLVMBuild description file for the components in this subdirectory.
+;
+; For more information on the LLVMBuild system, please see:
+;
+;   http://llvm.org/docs/LLVMBuild.html
+;
+;===------------------------------------------------------------------------===;
+
+[component_0]
+type = Library
+name = TextAPI
+parent = Libraries
+required_libraries = Support BinaryFormat

llvm/unittests/CMakeLists.txt

@@ -30,6 +30,7 @@ add_subdirectory(OptRemarks)
 add_subdirectory(Passes)
 add_subdirectory(ProfileData)
 add_subdirectory(Support)
+add_subdirectory(TextAPI)
 add_subdirectory(Target)
 add_subdirectory(Transforms)
 add_subdirectory(XRay)

llvm/unittests/TextAPI/CMakeLists.txt

@@ -0,0 +1,7 @@
+set(LLVM_LINK_COMPONENTS
+  TextAPI
+)
+
+add_llvm_unittest(TapiTests
+  ELFYAMLTest.cpp
+)