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 "MCFunction.h" 17 #include "llvm/Object/ObjectFile.h" 18 #include "llvm/ADT/OwningPtr.h" 19 #include "llvm/ADT/Triple.h" 20 #include "llvm/ADT/STLExtras.h" 21 #include "llvm/MC/MCAsmInfo.h" 22 #include "llvm/MC/MCDisassembler.h" 23 #include "llvm/MC/MCInst.h" 24 #include "llvm/MC/MCInstPrinter.h" 25 #include "llvm/MC/MCInstrDesc.h" 26 #include "llvm/MC/MCInstrInfo.h" 27 #include "llvm/Support/CommandLine.h" 28 #include "llvm/Support/Debug.h" 29 #include "llvm/Support/Format.h" 30 #include "llvm/Support/GraphWriter.h" 31 #include "llvm/Support/Host.h" 32 #include "llvm/Support/ManagedStatic.h" 33 #include "llvm/Support/MemoryBuffer.h" 34 #include "llvm/Support/MemoryObject.h" 35 #include "llvm/Support/PrettyStackTrace.h" 36 #include "llvm/Support/Signals.h" 37 #include "llvm/Support/SourceMgr.h" 38 #include "llvm/Support/raw_ostream.h" 39 #include "llvm/Support/system_error.h" 40 #include "llvm/Target/TargetRegistry.h" 41 #include "llvm/Target/TargetSelect.h" 42 #include <algorithm> 43 #include <cstring> 44 using namespace llvm; 45 using namespace object; 46 47 namespace { 48 cl::list<std::string> 49 InputFilenames(cl::Positional, cl::desc("<input object files>"), 50 cl::ZeroOrMore); 51 52 cl::opt<bool> 53 Disassemble("disassemble", 54 cl::desc("Display assembler mnemonics for the machine instructions")); 55 cl::alias 56 Disassembled("d", cl::desc("Alias for --disassemble"), 57 cl::aliasopt(Disassemble)); 58 59 cl::opt<bool> 60 CFG("cfg", cl::desc("Create a CFG for every symbol in the object file and" 61 "write it to a graphviz file")); 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 const MCInstrInfo *InstrInfo = TheTarget->createMCInstrInfo(); 168 169 outs() << '\n'; 170 outs() << Filename 171 << ":\tfile format " << Obj->getFileFormatName() << "\n\n"; 172 173 error_code ec; 174 for (ObjectFile::section_iterator i = Obj->begin_sections(), 175 e = Obj->end_sections(); 176 i != e; i.increment(ec)) { 177 if (error(ec)) break; 178 bool text; 179 if (error(i->isText(text))) break; 180 if (!text) continue; 181 182 // Make a list of all the symbols in this section. 183 std::vector<std::pair<uint64_t, StringRef> > Symbols; 184 for (ObjectFile::symbol_iterator si = Obj->begin_symbols(), 185 se = Obj->end_symbols(); 186 si != se; si.increment(ec)) { 187 bool contains; 188 if (!error(i->containsSymbol(*si, contains)) && contains) { 189 uint64_t Address; 190 if (error(si->getAddress(Address))) break; 191 StringRef Name; 192 if (error(si->getName(Name))) break; 193 Symbols.push_back(std::make_pair(Address, Name)); 194 } 195 } 196 197 // Sort the symbols by address, just in case they didn't come in that way. 198 array_pod_sort(Symbols.begin(), Symbols.end()); 199 200 StringRef name; 201 if (error(i->getName(name))) break; 202 outs() << "Disassembly of section " << name << ':'; 203 204 // If the section has no symbols just insert a dummy one and disassemble 205 // the whole section. 206 if (Symbols.empty()) 207 Symbols.push_back(std::make_pair(0, name)); 208 209 // Set up disassembler. 210 OwningPtr<const MCAsmInfo> AsmInfo(TheTarget->createMCAsmInfo(TripleName)); 211 212 if (!AsmInfo) { 213 errs() << "error: no assembly info for target " << TripleName << "\n"; 214 return; 215 } 216 217 OwningPtr<const MCDisassembler> DisAsm(TheTarget->createMCDisassembler()); 218 if (!DisAsm) { 219 errs() << "error: no disassembler for target " << TripleName << "\n"; 220 return; 221 } 222 223 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 224 OwningPtr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 225 AsmPrinterVariant, *AsmInfo)); 226 if (!IP) { 227 errs() << "error: no instruction printer for target " << TripleName << '\n'; 228 return; 229 } 230 231 StringRef Bytes; 232 if (error(i->getContents(Bytes))) break; 233 StringRefMemoryObject memoryObject(Bytes); 234 uint64_t Size; 235 uint64_t Index; 236 uint64_t SectSize; 237 if (error(i->getSize(SectSize))) break; 238 239 // Disassemble symbol by symbol. 240 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 241 uint64_t Start = Symbols[si].first; 242 uint64_t End = si == se-1 ? SectSize : Symbols[si + 1].first - 1; 243 outs() << '\n' << Symbols[si].second << ":\n"; 244 245 #ifndef NDEBUG 246 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 247 #else 248 raw_ostream &DebugOut = nulls(); 249 #endif 250 251 if (!CFG) { 252 for (Index = Start; Index < End; Index += Size) { 253 MCInst Inst; 254 if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, 255 DebugOut)) { 256 uint64_t addr; 257 if (error(i->getAddress(addr))) break; 258 outs() << format("%8x:\t", addr + Index); 259 DumpBytes(StringRef(Bytes.data() + Index, Size)); 260 IP->printInst(&Inst, outs()); 261 outs() << "\n"; 262 } else { 263 errs() << ToolName << ": warning: invalid instruction encoding\n"; 264 if (Size == 0) 265 Size = 1; // skip illegible bytes 266 } 267 } 268 269 } else { 270 // Create CFG and use it for disassembly. 271 MCFunction f = 272 MCFunction::createFunctionFromMC(Symbols[si].second, DisAsm.get(), 273 memoryObject, Start, End, InstrInfo, 274 DebugOut); 275 276 for (MCFunction::iterator fi = f.begin(), fe = f.end(); fi != fe; ++fi){ 277 bool hasPreds = false; 278 // Only print blocks that have predecessors. 279 // FIXME: Slow. 280 for (MCFunction::iterator pi = f.begin(), pe = f.end(); pi != pe; 281 ++pi) 282 if (pi->second.contains(&fi->second)) { 283 hasPreds = true; 284 break; 285 } 286 287 // Data block. 288 if (!hasPreds && fi != f.begin()) { 289 uint64_t End = llvm::next(fi) == fe ? SectSize : 290 llvm::next(fi)->first; 291 uint64_t addr; 292 if (error(i->getAddress(addr))) break; 293 outs() << "# " << End-fi->first << " bytes of data:\n"; 294 for (unsigned pos = fi->first; pos != End; ++pos) { 295 outs() << format("%8x:\t", addr + pos); 296 DumpBytes(StringRef(Bytes.data() + pos, 1)); 297 outs() << format("\t.byte 0x%02x\n", (uint8_t)Bytes[pos]); 298 } 299 continue; 300 } 301 302 if (fi->second.contains(&fi->second)) 303 outs() << "# Loop begin:\n"; 304 305 for (unsigned ii = 0, ie = fi->second.getInsts().size(); ii != ie; 306 ++ii) { 307 uint64_t addr; 308 if (error(i->getAddress(addr))) break; 309 const MCDecodedInst &Inst = fi->second.getInsts()[ii]; 310 outs() << format("%8x:\t", addr + Inst.Address); 311 DumpBytes(StringRef(Bytes.data() + Inst.Address, Inst.Size)); 312 // Simple loops. 313 if (fi->second.contains(&fi->second)) 314 outs() << '\t'; 315 IP->printInst(&Inst.Inst, outs()); 316 outs() << '\n'; 317 } 318 } 319 320 // Start a new dot file. 321 std::string Error; 322 raw_fd_ostream Out((f.getName().str() + ".dot").c_str(), Error); 323 if (!Error.empty()) { 324 errs() << ToolName << ": warning: " << Error << '\n'; 325 continue; 326 } 327 328 Out << "digraph " << f.getName() << " {\n"; 329 Out << "graph [ rankdir = \"LR\" ];\n"; 330 for (MCFunction::iterator i = f.begin(), e = f.end(); i != e; ++i) { 331 bool hasPreds = false; 332 // Only print blocks that have predecessors. 333 // FIXME: Slow. 334 for (MCFunction::iterator pi = f.begin(), pe = f.end(); pi != pe; 335 ++pi) 336 if (pi->second.contains(&i->second)) { 337 hasPreds = true; 338 break; 339 } 340 341 if (!hasPreds && i != f.begin()) 342 continue; 343 344 Out << '"' << (uintptr_t)&i->second << "\" [ label=\"<a>"; 345 // Print instructions. 346 for (unsigned ii = 0, ie = i->second.getInsts().size(); ii != ie; 347 ++ii) { 348 // Escape special chars and print the instruction in mnemonic form. 349 std::string Str; 350 raw_string_ostream OS(Str); 351 IP->printInst(&i->second.getInsts()[ii].Inst, OS); 352 Out << DOT::EscapeString(OS.str()) << '|'; 353 } 354 Out << "<o>\" shape=\"record\" ];\n"; 355 356 // Add edges. 357 for (MCBasicBlock::succ_iterator si = i->second.succ_begin(), 358 se = i->second.succ_end(); si != se; ++si) 359 Out << (uintptr_t)&i->second << ":o -> " << (uintptr_t)*si <<":a\n"; 360 } 361 Out << "}\n"; 362 } 363 } 364 } 365 } 366 367 int main(int argc, char **argv) { 368 // Print a stack trace if we signal out. 369 sys::PrintStackTraceOnErrorSignal(); 370 PrettyStackTraceProgram X(argc, argv); 371 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 372 373 // Initialize targets and assembly printers/parsers. 374 llvm::InitializeAllTargetInfos(); 375 llvm::InitializeAllTargetMCs(); 376 llvm::InitializeAllAsmParsers(); 377 llvm::InitializeAllDisassemblers(); 378 379 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 380 TripleName = Triple::normalize(TripleName); 381 382 ToolName = argv[0]; 383 384 // Defaults to a.out if no filenames specified. 385 if (InputFilenames.size() == 0) 386 InputFilenames.push_back("a.out"); 387 388 // -d is the only flag that is currently implemented, so just print help if 389 // it is not set. 390 if (!Disassemble) { 391 cl::PrintHelpMessage(); 392 return 2; 393 } 394 395 std::for_each(InputFilenames.begin(), InputFilenames.end(), 396 DisassembleInput); 397 398 return 0; 399 } 400