lib/BinaryFormat/Magic.cpp

   1 //===- llvm/BinaryFormat/Magic.cpp - File magic identification --*- C++ -*-===//
   2 //
   3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
   4 // See https://llvm.org/LICENSE.txt for license information.
   5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
   6 //
   7 //===----------------------------------------------------------------------===//
   8
   9 #include "llvm/BinaryFormat/Magic.h"
  10
  11 #include "llvm/BinaryFormat/COFF.h"
  12 #include "llvm/BinaryFormat/ELF.h"
  13 #include "llvm/BinaryFormat/MachO.h"
  14 #include "llvm/Support/Endian.h"
  15 #include "llvm/Support/FileSystem.h"
  16 #include "llvm/Support/MemoryBuffer.h"
  17
  18 #if !defined(_MSC_VER) && !defined(__MINGW32__)
  19 #include <unistd.h>
  20 #else
  21 #include <io.h>
  22 #endif
  23
  24 using namespace llvm;
  25 using namespace llvm::support::endian;
  26 using namespace llvm::sys::fs;
  27
  28 template <size_t N>
  29 static bool startswith(StringRef Magic, const char (&S)[N]) {
  30   return Magic.startswith(StringRef(S, N - 1));
  31 }
  32
  33 /// Identify the magic in magic.
  34 file_magic llvm::identify_magic(StringRef Magic) {
  35   if (Magic.size() < 4)
  36     return file_magic::unknown;
  37   switch ((unsigned char)Magic[0]) {
  38   case 0x00: {
  39     // COFF bigobj, CL.exe's LTO object file, or short import library file
  40     if (startswith(Magic, "\0\0\xFF\xFF")) {
  41       size_t MinSize =
  42           offsetof(COFF::BigObjHeader, UUID) + sizeof(COFF::BigObjMagic);
  43       if (Magic.size() < MinSize)
  44         return file_magic::coff_import_library;
  45
  46       const char *Start = Magic.data() + offsetof(COFF::BigObjHeader, UUID);
  47       if (memcmp(Start, COFF::BigObjMagic, sizeof(COFF::BigObjMagic)) == 0)
  48         return file_magic::coff_object;
  49       if (memcmp(Start, COFF::ClGlObjMagic, sizeof(COFF::BigObjMagic)) == 0)
  50         return file_magic::coff_cl_gl_object;
  51       return file_magic::coff_import_library;
  52     }
  53     // Windows resource file
  54     if (Magic.size() >= sizeof(COFF::WinResMagic) &&
  55         memcmp(Magic.data(), COFF::WinResMagic, sizeof(COFF::WinResMagic)) == 0)
  56       return file_magic::windows_resource;
  57     // 0x0000 = COFF unknown machine type
  58     if (Magic[1] == 0)
  59       return file_magic::coff_object;
  60     if (startswith(Magic, "\0asm"))
  61       return file_magic::wasm_object;
  62     break;
  63   }
  64
  65   case 0x01:
  66     // XCOFF format
  67     if (startswith(Magic, "\x01\xDF"))
  68       return file_magic::xcoff_object_32;
  69     if (startswith(Magic, "\x01\xF7"))
  70       return file_magic::xcoff_object_64;
  71     break;
  72
  73   case 0xDE: // 0x0B17C0DE = BC wraper
  74     if (startswith(Magic, "\xDE\xC0\x17\x0B"))
  75       return file_magic::bitcode;
  76     break;
  77   case 'B':
  78     if (startswith(Magic, "BC\xC0\xDE"))
  79       return file_magic::bitcode;
  80     break;
  81   case '!':
  82     if (startswith(Magic, "!<arch>\n") || startswith(Magic, "!<thin>\n"))
  83       return file_magic::archive;
  84     break;
  85
  86   case '\177':
  87     if (startswith(Magic, "\177ELF") && Magic.size() >= 18) {
  88       bool Data2MSB = Magic[5] == 2;
  89       unsigned high = Data2MSB ? 16 : 17;
  90       unsigned low = Data2MSB ? 17 : 16;
  91       if (Magic[high] == 0) {
  92         switch (Magic[low]) {
  93         default:
  94           return file_magic::elf;
  95         case 1:
  96           return file_magic::elf_relocatable;
  97         case 2:
  98           return file_magic::elf_executable;
  99         case 3:
 100           return file_magic::elf_shared_object;
 101         case 4:
 102           return file_magic::elf_core;
 103         }
 104       }
 105       // It's still some type of ELF file.
 106       return file_magic::elf;
 107     }
 108     break;
 109
 110   case 0xCA:
 111     if (startswith(Magic, "\xCA\xFE\xBA\xBE") ||
 112         startswith(Magic, "\xCA\xFE\xBA\xBF")) {
 113       // This is complicated by an overlap with Java class files.
 114       // See the Mach-O section in /usr/share/file/magic for details.
 115       if (Magic.size() >= 8 && Magic[7] < 43)
 116         return file_magic::macho_universal_binary;
 117     }
 118     break;
 119
 120   // The two magic numbers for mach-o are:
 121   // 0xfeedface - 32-bit mach-o
 122   // 0xfeedfacf - 64-bit mach-o
 123   case 0xFE:
 124   case 0xCE:
 125   case 0xCF: {
 126     uint16_t type = 0;
 127     if (startswith(Magic, "\xFE\xED\xFA\xCE") ||
 128         startswith(Magic, "\xFE\xED\xFA\xCF")) {
 129       /* Native endian */
 130       size_t MinSize;
 131       if (Magic[3] == char(0xCE))
 132         MinSize = sizeof(MachO::mach_header);
 133       else
 134         MinSize = sizeof(MachO::mach_header_64);
 135       if (Magic.size() >= MinSize)
 136         type = Magic[12] << 24 | Magic[13] << 12 | Magic[14] << 8 | Magic[15];
 137     } else if (startswith(Magic, "\xCE\xFA\xED\xFE") ||
 138                startswith(Magic, "\xCF\xFA\xED\xFE")) {
 139       /* Reverse endian */
 140       size_t MinSize;
 141       if (Magic[0] == char(0xCE))
 142         MinSize = sizeof(MachO::mach_header);
 143       else
 144         MinSize = sizeof(MachO::mach_header_64);
 145       if (Magic.size() >= MinSize)
 146         type = Magic[15] << 24 | Magic[14] << 12 | Magic[13] << 8 | Magic[12];
 147     }
 148     switch (type) {
 149     default:
 150       break;
 151     case 1:
 152       return file_magic::macho_object;
 153     case 2:
 154       return file_magic::macho_executable;
 155     case 3:
 156       return file_magic::macho_fixed_virtual_memory_shared_lib;
 157     case 4:
 158       return file_magic::macho_core;
 159     case 5:
 160       return file_magic::macho_preload_executable;
 161     case 6:
 162       return file_magic::macho_dynamically_linked_shared_lib;
 163     case 7:
 164       return file_magic::macho_dynamic_linker;
 165     case 8:
 166       return file_magic::macho_bundle;
 167     case 9:
 168       return file_magic::macho_dynamically_linked_shared_lib_stub;
 169     case 10:
 170       return file_magic::macho_dsym_companion;
 171     case 11:
 172       return file_magic::macho_kext_bundle;
 173     }
 174     break;
 175   }
 176   case 0xF0: // PowerPC Windows
 177   case 0x83: // Alpha 32-bit
 178   case 0x84: // Alpha 64-bit
 179   case 0x66: // MPS R4000 Windows
 180   case 0x50: // mc68K
 181   case 0x4c: // 80386 Windows
 182   case 0xc4: // ARMNT Windows
 183     if (Magic[1] == 0x01)
 184       return file_magic::coff_object;
 185     LLVM_FALLTHROUGH;
 186
 187   case 0x90: // PA-RISC Windows
 188   case 0x68: // mc68K Windows
 189     if (Magic[1] == 0x02)
 190       return file_magic::coff_object;
 191     break;
 192
 193   case 'M': // Possible MS-DOS stub on Windows PE file, MSF/PDB file or a
 194             // Minidump file.
 195     if (startswith(Magic, "MZ") && Magic.size() >= 0x3c + 4) {
 196       uint32_t off = read32le(Magic.data() + 0x3c);
 197       // PE/COFF file, either EXE or DLL.
 198       if (Magic.substr(off).startswith(
 199               StringRef(COFF::PEMagic, sizeof(COFF::PEMagic))))
 200         return file_magic::pecoff_executable;
 201     }
 202     if (Magic.startswith("Microsoft C/C++ MSF 7.00\r\n"))
 203       return file_magic::pdb;
 204     if (startswith(Magic, "MDMP"))
 205       return file_magic::minidump;
 206     break;
 207
 208   case 0x64: // x86-64 or ARM64 Windows.
 209     if (Magic[1] == char(0x86) || Magic[1] == char(0xaa))
 210       return file_magic::coff_object;
 211     break;
 212
 213   case 0x2d: // YAML '-'
 214     if (startswith(Magic, "--- !tapi") || startswith(Magic, "---\narchs:"))
 215       return file_magic::tapi_file;
 216     break;
 217
 218   default:
 219     break;
 220   }
 221   return file_magic::unknown;
 222 }
 223
 224 std::error_code llvm::identify_magic(const Twine &Path, file_magic &Result) {
 225   auto FileOrError = MemoryBuffer::getFile(Path, -1LL, false);
 226   if (!FileOrError)
 227     return FileOrError.getError();
 228
 229   std::unique_ptr<MemoryBuffer> FileBuffer = std::move(*FileOrError);
 230   Result = identify_magic(FileBuffer->getBuffer());
 231
 232   return std::error_code();
 233 }