1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This program is a utility that works like binutils "objdump", that is, it 11 // dumps out a plethora of information about an object file depending on the 12 // flags. 13 // 14 //===----------------------------------------------------------------------===// 15 16 #include "llvm/Object/ObjectFile.h" 17 // This config must be included before llvm-config.h. 18 #include "llvm/Config/config.h" 19 #include "../../lib/MC/MCDisassembler/EDDisassembler.h" 20 #include "../../lib/MC/MCDisassembler/EDInst.h" 21 #include "../../lib/MC/MCDisassembler/EDOperand.h" 22 #include "../../lib/MC/MCDisassembler/EDToken.h" 23 #include "llvm/ADT/OwningPtr.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/MC/MCAsmInfo.h" 26 #include "llvm/MC/MCDisassembler.h" 27 #include "llvm/MC/MCInst.h" 28 #include "llvm/MC/MCInstPrinter.h" 29 #include "llvm/Support/CommandLine.h" 30 #include "llvm/Support/Debug.h" 31 #include "llvm/Support/Format.h" 32 #include "llvm/Support/Host.h" 33 #include "llvm/Support/ManagedStatic.h" 34 #include "llvm/Support/MemoryBuffer.h" 35 #include "llvm/Support/MemoryObject.h" 36 #include "llvm/Support/PrettyStackTrace.h" 37 #include "llvm/Support/Signals.h" 38 #include "llvm/Support/SourceMgr.h" 39 #include "llvm/Support/raw_ostream.h" 40 #include "llvm/Support/system_error.h" 41 #include "llvm/Target/TargetMachine.h" 42 #include "llvm/Target/TargetRegistry.h" 43 #include "llvm/Target/TargetSelect.h" 44 #include <algorithm> 45 #include <cctype> 46 #include <cerrno> 47 #include <cstring> 48 using namespace llvm; 49 using namespace object; 50 51 namespace { 52 cl::list<std::string> 53 InputFilenames(cl::Positional, cl::desc("<input object files>"), 54 cl::ZeroOrMore); 55 56 cl::opt<bool> 57 Disassemble("disassemble", 58 cl::desc("Display assembler mnemonics for the machine instructions")); 59 cl::alias 60 Disassembled("d", cl::desc("Alias for --disassemble"), 61 cl::aliasopt(Disassemble)); 62 63 cl::opt<std::string> 64 TripleName("triple", cl::desc("Target triple to disassemble for, " 65 "see -version for available targets")); 66 67 cl::opt<std::string> 68 ArchName("arch", cl::desc("Target arch to disassemble for, " 69 "see -version for available targets")); 70 71 StringRef ToolName; 72 73 bool error(error_code ec) { 74 if (!ec) return false; 75 76 outs() << ToolName << ": error reading file: " << ec.message() << ".\n"; 77 outs().flush(); 78 return true; 79 } 80 } 81 82 static const Target *GetTarget(const ObjectFile *Obj = NULL) { 83 // Figure out the target triple. 84 llvm::Triple TT("unknown-unknown-unknown"); 85 if (TripleName.empty()) { 86 if (Obj) 87 TT.setArch(Triple::ArchType(Obj->getArch())); 88 } else 89 TT.setTriple(Triple::normalize(TripleName)); 90 91 if (!ArchName.empty()) 92 TT.setArchName(ArchName); 93 94 TripleName = TT.str(); 95 96 // Get the target specific parser. 97 std::string Error; 98 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 99 if (TheTarget) 100 return TheTarget; 101 102 errs() << ToolName << ": error: unable to get target for '" << TripleName 103 << "', see --version and --triple.\n"; 104 return 0; 105 } 106 107 namespace { 108 class StringRefMemoryObject : public MemoryObject { 109 private: 110 StringRef Bytes; 111 public: 112 StringRefMemoryObject(StringRef bytes) : Bytes(bytes) {} 113 114 uint64_t getBase() const { return 0; } 115 uint64_t getExtent() const { return Bytes.size(); } 116 117 int readByte(uint64_t Addr, uint8_t *Byte) const { 118 if (Addr > getExtent()) 119 return -1; 120 *Byte = Bytes[Addr]; 121 return 0; 122 } 123 }; 124 } 125 126 static void DumpBytes(StringRef bytes) { 127 static char hex_rep[] = "0123456789abcdef"; 128 // FIXME: The real way to do this is to figure out the longest instruction 129 // and align to that size before printing. I'll fix this when I get 130 // around to outputting relocations. 131 // 15 is the longest x86 instruction 132 // 3 is for the hex rep of a byte + a space. 133 // 1 is for the null terminator. 134 enum { OutputSize = (15 * 3) + 1 }; 135 char output[OutputSize]; 136 137 assert(bytes.size() <= 15 138 && "DumpBytes only supports instructions of up to 15 bytes"); 139 memset(output, ' ', sizeof(output)); 140 unsigned index = 0; 141 for (StringRef::iterator i = bytes.begin(), 142 e = bytes.end(); i != e; ++i) { 143 output[index] = hex_rep[(*i & 0xF0) >> 4]; 144 output[index + 1] = hex_rep[*i & 0xF]; 145 index += 3; 146 } 147 148 output[sizeof(output) - 1] = 0; 149 outs() << output; 150 } 151 152 static void DisassembleInput(const StringRef &Filename) { 153 OwningPtr<MemoryBuffer> Buff; 154 155 if (error_code ec = MemoryBuffer::getFileOrSTDIN(Filename, Buff)) { 156 errs() << ToolName << ": " << Filename << ": " << ec.message() << "\n"; 157 return; 158 } 159 160 OwningPtr<ObjectFile> Obj(ObjectFile::createObjectFile(Buff.take())); 161 162 const Target *TheTarget = GetTarget(Obj.get()); 163 if (!TheTarget) { 164 // GetTarget prints out stuff. 165 return; 166 } 167 168 outs() << '\n'; 169 outs() << Filename 170 << ":\tfile format " << Obj->getFileFormatName() << "\n\n\n"; 171 172 error_code ec; 173 for (ObjectFile::section_iterator i = Obj->begin_sections(), 174 e = Obj->end_sections(); 175 i != e; i.increment(ec)) { 176 if (error(ec)) break; 177 bool text; 178 if (error(i->isText(text))) break; 179 if (!text) continue; 180 181 StringRef name; 182 if (error(i->getName(name))) break; 183 outs() << "Disassembly of section " << name << ":\n\n"; 184 185 // Set up disassembler. 186 OwningPtr<const MCAsmInfo> AsmInfo(TheTarget->createAsmInfo(TripleName)); 187 188 if (!AsmInfo) { 189 errs() << "error: no assembly info for target " << TripleName << "\n"; 190 return; 191 } 192 193 OwningPtr<const MCDisassembler> DisAsm(TheTarget->createMCDisassembler()); 194 if (!DisAsm) { 195 errs() << "error: no disassembler for target " << TripleName << "\n"; 196 return; 197 } 198 199 // FIXME: We shouldn't need to do this (and link in codegen). 200 // When we split this out, we should do it in a way that makes 201 // it straightforward to switch subtargets on the fly (.e.g, 202 // the .cpu and .code16 directives). 203 std::string FeaturesStr; 204 std::string CPU; 205 OwningPtr<TargetMachine> TM(TheTarget->createTargetMachine(TripleName, CPU, 206 FeaturesStr)); 207 if (!TM) { 208 errs() << "error: could not create target for triple " << TripleName << "\n"; 209 return; 210 } 211 212 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 213 OwningPtr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 214 *TM, AsmPrinterVariant, *AsmInfo)); 215 if (!IP) { 216 errs() << "error: no instruction printer for target " << TripleName << '\n'; 217 return; 218 } 219 220 StringRef Bytes; 221 if (error(i->getContents(Bytes))) break; 222 StringRefMemoryObject memoryObject(Bytes); 223 uint64_t Size; 224 uint64_t Index; 225 226 for (Index = 0; Index < Bytes.size(); Index += Size) { 227 MCInst Inst; 228 229 # ifndef NDEBUG 230 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 231 # else 232 raw_ostream &DebugOut = nulls(); 233 # endif 234 235 if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, DebugOut)) { 236 uint64_t addr; 237 if (error(i->getAddress(addr))) break; 238 outs() << format("%8x:\t", addr + Index); 239 DumpBytes(StringRef(Bytes.data() + Index, Size)); 240 IP->printInst(&Inst, outs()); 241 outs() << "\n"; 242 } else { 243 errs() << ToolName << ": warning: invalid instruction encoding\n"; 244 if (Size == 0) 245 Size = 1; // skip illegible bytes 246 } 247 } 248 } 249 } 250 251 int main(int argc, char **argv) { 252 // Print a stack trace if we signal out. 253 sys::PrintStackTraceOnErrorSignal(); 254 PrettyStackTraceProgram X(argc, argv); 255 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 256 257 // Initialize targets and assembly printers/parsers. 258 llvm::InitializeAllTargetInfos(); 259 // FIXME: We shouldn't need to initialize the Target(Machine)s. 260 llvm::InitializeAllTargets(); 261 llvm::InitializeAllAsmPrinters(); 262 llvm::InitializeAllAsmParsers(); 263 llvm::InitializeAllDisassemblers(); 264 265 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 266 TripleName = Triple::normalize(TripleName); 267 268 ToolName = argv[0]; 269 270 // Defaults to a.out if no filenames specified. 271 if (InputFilenames.size() == 0) 272 InputFilenames.push_back("a.out"); 273 274 // -d is the only flag that is currently implemented, so just print help if 275 // it is not set. 276 if (!Disassemble) { 277 cl::PrintHelpMessage(); 278 return 2; 279 } 280 281 std::for_each(InputFilenames.begin(), InputFilenames.end(), 282 DisassembleInput); 283 284 return 0; 285 } 286