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-objdump.h" 17 #include "MCFunction.h" 18 #include "llvm/Object/Archive.h" 19 #include "llvm/Object/ObjectFile.h" 20 #include "llvm/ADT/OwningPtr.h" 21 #include "llvm/ADT/Triple.h" 22 #include "llvm/ADT/STLExtras.h" 23 #include "llvm/MC/MCAsmInfo.h" 24 #include "llvm/MC/MCDisassembler.h" 25 #include "llvm/MC/MCInst.h" 26 #include "llvm/MC/MCInstPrinter.h" 27 #include "llvm/MC/MCInstrAnalysis.h" 28 #include "llvm/MC/MCInstrDesc.h" 29 #include "llvm/MC/MCInstrInfo.h" 30 #include "llvm/MC/MCSubtargetInfo.h" 31 #include "llvm/Support/Casting.h" 32 #include "llvm/Support/CommandLine.h" 33 #include "llvm/Support/Debug.h" 34 #include "llvm/Support/FileSystem.h" 35 #include "llvm/Support/Format.h" 36 #include "llvm/Support/GraphWriter.h" 37 #include "llvm/Support/Host.h" 38 #include "llvm/Support/ManagedStatic.h" 39 #include "llvm/Support/MemoryBuffer.h" 40 #include "llvm/Support/MemoryObject.h" 41 #include "llvm/Support/PrettyStackTrace.h" 42 #include "llvm/Support/Signals.h" 43 #include "llvm/Support/SourceMgr.h" 44 #include "llvm/Support/TargetRegistry.h" 45 #include "llvm/Support/TargetSelect.h" 46 #include "llvm/Support/raw_ostream.h" 47 #include "llvm/Support/system_error.h" 48 #include <algorithm> 49 #include <cstring> 50 using namespace llvm; 51 using namespace object; 52 53 static cl::list<std::string> 54 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 55 56 static cl::opt<bool> 57 Disassemble("disassemble", 58 cl::desc("Display assembler mnemonics for the machine instructions")); 59 static cl::alias 60 Disassembled("d", cl::desc("Alias for --disassemble"), 61 cl::aliasopt(Disassemble)); 62 63 static cl::opt<bool> 64 Relocations("r", cl::desc("Display the relocation entries in the file")); 65 66 static cl::opt<bool> 67 MachO("macho", cl::desc("Use MachO specific object file parser")); 68 static cl::alias 69 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachO)); 70 71 cl::opt<std::string> 72 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 73 "see -version for available targets")); 74 75 cl::opt<std::string> 76 llvm::ArchName("arch", cl::desc("Target arch to disassemble for, " 77 "see -version for available targets")); 78 79 static StringRef ToolName; 80 81 static bool error(error_code ec) { 82 if (!ec) return false; 83 84 outs() << ToolName << ": error reading file: " << ec.message() << ".\n"; 85 outs().flush(); 86 return true; 87 } 88 89 static const Target *GetTarget(const ObjectFile *Obj = NULL) { 90 // Figure out the target triple. 91 llvm::Triple TT("unknown-unknown-unknown"); 92 if (TripleName.empty()) { 93 if (Obj) 94 TT.setArch(Triple::ArchType(Obj->getArch())); 95 } else 96 TT.setTriple(Triple::normalize(TripleName)); 97 98 if (!ArchName.empty()) 99 TT.setArchName(ArchName); 100 101 TripleName = TT.str(); 102 103 // Get the target specific parser. 104 std::string Error; 105 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 106 if (TheTarget) 107 return TheTarget; 108 109 errs() << ToolName << ": error: unable to get target for '" << TripleName 110 << "', see --version and --triple.\n"; 111 return 0; 112 } 113 114 void llvm::DumpBytes(StringRef bytes) { 115 static const char hex_rep[] = "0123456789abcdef"; 116 // FIXME: The real way to do this is to figure out the longest instruction 117 // and align to that size before printing. I'll fix this when I get 118 // around to outputting relocations. 119 // 15 is the longest x86 instruction 120 // 3 is for the hex rep of a byte + a space. 121 // 1 is for the null terminator. 122 enum { OutputSize = (15 * 3) + 1 }; 123 char output[OutputSize]; 124 125 assert(bytes.size() <= 15 126 && "DumpBytes only supports instructions of up to 15 bytes"); 127 memset(output, ' ', sizeof(output)); 128 unsigned index = 0; 129 for (StringRef::iterator i = bytes.begin(), 130 e = bytes.end(); i != e; ++i) { 131 output[index] = hex_rep[(*i & 0xF0) >> 4]; 132 output[index + 1] = hex_rep[*i & 0xF]; 133 index += 3; 134 } 135 136 output[sizeof(output) - 1] = 0; 137 outs() << output; 138 } 139 140 static void DisassembleObject(const ObjectFile *Obj) { 141 const Target *TheTarget = GetTarget(Obj); 142 if (!TheTarget) { 143 // GetTarget prints out stuff. 144 return; 145 } 146 const MCInstrInfo *InstrInfo = TheTarget->createMCInstrInfo(); 147 OwningPtr<MCInstrAnalysis> 148 InstrAnalysis(TheTarget->createMCInstrAnalysis(InstrInfo)); 149 150 outs() << '\n'; 151 outs() << Obj->getFileName() 152 << ":\tfile format " << Obj->getFileFormatName() << "\n\n"; 153 154 error_code ec; 155 for (section_iterator i = Obj->begin_sections(), 156 e = Obj->end_sections(); 157 i != e; i.increment(ec)) { 158 if (error(ec)) break; 159 bool text; 160 if (error(i->isText(text))) break; 161 if (!text) continue; 162 163 // Make a list of all the symbols in this section. 164 std::vector<std::pair<uint64_t, StringRef> > Symbols; 165 for (symbol_iterator si = Obj->begin_symbols(), 166 se = Obj->end_symbols(); 167 si != se; si.increment(ec)) { 168 bool contains; 169 if (!error(i->containsSymbol(*si, contains)) && contains) { 170 uint64_t Address; 171 if (error(si->getOffset(Address))) break; 172 StringRef Name; 173 if (error(si->getName(Name))) break; 174 Symbols.push_back(std::make_pair(Address, Name)); 175 } 176 } 177 178 // Sort the symbols by address, just in case they didn't come in that way. 179 array_pod_sort(Symbols.begin(), Symbols.end()); 180 181 StringRef name; 182 if (error(i->getName(name))) break; 183 outs() << "Disassembly of section " << name << ':'; 184 185 // If the section has no symbols just insert a dummy one and disassemble 186 // the whole section. 187 if (Symbols.empty()) 188 Symbols.push_back(std::make_pair(0, name)); 189 190 // Set up disassembler. 191 OwningPtr<const MCAsmInfo> AsmInfo(TheTarget->createMCAsmInfo(TripleName)); 192 193 if (!AsmInfo) { 194 errs() << "error: no assembly info for target " << TripleName << "\n"; 195 return; 196 } 197 198 OwningPtr<const MCSubtargetInfo> STI( 199 TheTarget->createMCSubtargetInfo(TripleName, "", "")); 200 201 if (!STI) { 202 errs() << "error: no subtarget info for target " << TripleName << "\n"; 203 return; 204 } 205 206 OwningPtr<const MCDisassembler> DisAsm( 207 TheTarget->createMCDisassembler(*STI)); 208 if (!DisAsm) { 209 errs() << "error: no disassembler for target " << TripleName << "\n"; 210 return; 211 } 212 213 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 214 OwningPtr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 215 AsmPrinterVariant, *AsmInfo, *STI)); 216 if (!IP) { 217 errs() << "error: no instruction printer for target " << TripleName 218 << '\n'; 219 return; 220 } 221 222 StringRef Bytes; 223 if (error(i->getContents(Bytes))) break; 224 StringRefMemoryObject memoryObject(Bytes); 225 uint64_t Size; 226 uint64_t Index; 227 uint64_t SectSize; 228 if (error(i->getSize(SectSize))) break; 229 230 // Disassemble symbol by symbol. 231 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 232 uint64_t Start = Symbols[si].first; 233 uint64_t End = si == se-1 ? SectSize : Symbols[si + 1].first - 1; 234 outs() << '\n' << Symbols[si].second << ":\n"; 235 236 #ifndef NDEBUG 237 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 238 #else 239 raw_ostream &DebugOut = nulls(); 240 #endif 241 242 for (Index = Start; Index < End; Index += Size) { 243 MCInst Inst; 244 245 if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, 246 DebugOut, nulls())) { 247 uint64_t addr; 248 if (error(i->getAddress(addr))) break; 249 outs() << format("%8x:\t", addr + Index); 250 DumpBytes(StringRef(Bytes.data() + Index, Size)); 251 IP->printInst(&Inst, outs(), ""); 252 outs() << "\n"; 253 } else { 254 errs() << ToolName << ": warning: invalid instruction encoding\n"; 255 if (Size == 0) 256 Size = 1; // skip illegible bytes 257 } 258 } 259 } 260 } 261 } 262 263 static void PrintRelocations(const ObjectFile *o) { 264 error_code ec; 265 for (section_iterator si = o->begin_sections(), se = o->end_sections(); 266 si != se; si.increment(ec)){ 267 if (error(ec)) return; 268 if (si->begin_relocations() == si->end_relocations()) 269 continue; 270 StringRef secname; 271 if (error(si->getName(secname))) continue; 272 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 273 for (relocation_iterator ri = si->begin_relocations(), 274 re = si->end_relocations(); 275 ri != re; ri.increment(ec)) { 276 if (error(ec)) return; 277 278 uint64_t address; 279 SmallString<32> relocname; 280 SmallString<32> valuestr; 281 if (error(ri->getTypeName(relocname))) continue; 282 if (error(ri->getAddress(address))) continue; 283 if (error(ri->getValueString(valuestr))) continue; 284 outs() << address << " " << relocname << " " << valuestr << "\n"; 285 } 286 outs() << "\n"; 287 } 288 } 289 290 static void DumpObject(const ObjectFile *o) { 291 if (Disassemble) 292 DisassembleObject(o); 293 if (Relocations) 294 PrintRelocations(o); 295 } 296 297 /// @brief Dump each object file in \a a; 298 static void DumpArchive(const Archive *a) { 299 for (Archive::child_iterator i = a->begin_children(), 300 e = a->end_children(); i != e; ++i) { 301 OwningPtr<Binary> child; 302 if (error_code ec = i->getAsBinary(child)) { 303 errs() << ToolName << ": '" << a->getFileName() << "': " << ec.message() 304 << ".\n"; 305 continue; 306 } 307 if (ObjectFile *o = dyn_cast<ObjectFile>(child.get())) 308 DumpObject(o); 309 else 310 errs() << ToolName << ": '" << a->getFileName() << "': " 311 << "Unrecognized file type.\n"; 312 } 313 } 314 315 /// @brief Open file and figure out how to dump it. 316 static void DumpInput(StringRef file) { 317 // If file isn't stdin, check that it exists. 318 if (file != "-" && !sys::fs::exists(file)) { 319 errs() << ToolName << ": '" << file << "': " << "No such file\n"; 320 return; 321 } 322 323 if (MachO && Disassemble) { 324 DisassembleInputMachO(file); 325 return; 326 } 327 328 // Attempt to open the binary. 329 OwningPtr<Binary> binary; 330 if (error_code ec = createBinary(file, binary)) { 331 errs() << ToolName << ": '" << file << "': " << ec.message() << ".\n"; 332 return; 333 } 334 335 if (Archive *a = dyn_cast<Archive>(binary.get())) { 336 DumpArchive(a); 337 } else if (ObjectFile *o = dyn_cast<ObjectFile>(binary.get())) { 338 DumpObject(o); 339 } else { 340 errs() << ToolName << ": '" << file << "': " << "Unrecognized file type.\n"; 341 } 342 } 343 344 int main(int argc, char **argv) { 345 // Print a stack trace if we signal out. 346 sys::PrintStackTraceOnErrorSignal(); 347 PrettyStackTraceProgram X(argc, argv); 348 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 349 350 // Initialize targets and assembly printers/parsers. 351 llvm::InitializeAllTargetInfos(); 352 llvm::InitializeAllTargetMCs(); 353 llvm::InitializeAllAsmParsers(); 354 llvm::InitializeAllDisassemblers(); 355 356 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 357 TripleName = Triple::normalize(TripleName); 358 359 ToolName = argv[0]; 360 361 // Defaults to a.out if no filenames specified. 362 if (InputFilenames.size() == 0) 363 InputFilenames.push_back("a.out"); 364 365 if (!Disassemble && !Relocations) { 366 cl::PrintHelpMessage(); 367 return 2; 368 } 369 370 std::for_each(InputFilenames.begin(), InputFilenames.end(), 371 DumpInput); 372 373 return 0; 374 } 375