1 //===-- llvm-mc.cpp - Machine Code Hacking Driver -------------------------===// 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 utility is a simple driver that allows command line hacking on machine 11 // code. 12 // 13 //===----------------------------------------------------------------------===// 14 15 #include "llvm/MC/MCParser/MCAsmLexer.h" 16 #include "llvm/MC/MCContext.h" 17 #include "llvm/MC/MCCodeEmitter.h" 18 #include "llvm/MC/MCInstPrinter.h" 19 #include "llvm/MC/MCSectionMachO.h" 20 #include "llvm/MC/MCStreamer.h" 21 #include "llvm/ADT/OwningPtr.h" 22 #include "llvm/Support/CommandLine.h" 23 #include "llvm/Support/FormattedStream.h" 24 #include "llvm/Support/ManagedStatic.h" 25 #include "llvm/Support/MemoryBuffer.h" 26 #include "llvm/Support/PrettyStackTrace.h" 27 #include "llvm/Support/SourceMgr.h" 28 #include "llvm/Support/raw_ostream.h" 29 #include "llvm/System/Host.h" 30 #include "llvm/System/Signals.h" 31 #include "llvm/Target/TargetAsmBackend.h" 32 #include "llvm/Target/TargetAsmParser.h" 33 #include "llvm/Target/TargetData.h" 34 #include "llvm/Target/TargetRegistry.h" 35 #include "llvm/Target/TargetMachine.h" // FIXME. 36 #include "llvm/Target/TargetSelect.h" 37 #include "llvm/MC/MCParser/AsmParser.h" 38 #include "Disassembler.h" 39 using namespace llvm; 40 41 static cl::opt<std::string> 42 InputFilename(cl::Positional, cl::desc("<input file>"), cl::init("-")); 43 44 static cl::opt<std::string> 45 OutputFilename("o", cl::desc("Output filename"), 46 cl::value_desc("filename")); 47 48 static cl::opt<bool> 49 ShowEncoding("show-encoding", cl::desc("Show instruction encodings")); 50 51 static cl::opt<bool> 52 ShowInst("show-inst", cl::desc("Show internal instruction representation")); 53 54 static cl::opt<unsigned> 55 OutputAsmVariant("output-asm-variant", 56 cl::desc("Syntax variant to use for output printing")); 57 58 enum OutputFileType { 59 OFT_AssemblyFile, 60 OFT_ObjectFile 61 }; 62 static cl::opt<OutputFileType> 63 FileType("filetype", cl::init(OFT_AssemblyFile), 64 cl::desc("Choose an output file type:"), 65 cl::values( 66 clEnumValN(OFT_AssemblyFile, "asm", 67 "Emit an assembly ('.s') file"), 68 clEnumValN(OFT_ObjectFile, "obj", 69 "Emit a native object ('.o') file"), 70 clEnumValEnd)); 71 72 static cl::opt<bool> 73 Force("f", cl::desc("Enable binary output on terminals")); 74 75 static cl::list<std::string> 76 IncludeDirs("I", cl::desc("Directory of include files"), 77 cl::value_desc("directory"), cl::Prefix); 78 79 static cl::opt<std::string> 80 ArchName("arch", cl::desc("Target arch to assemble for, " 81 "see -version for available targets")); 82 83 static cl::opt<std::string> 84 TripleName("triple", cl::desc("Target triple to assemble for, " 85 "see -version for available targets")); 86 87 enum ActionType { 88 AC_AsLex, 89 AC_Assemble, 90 AC_Disassemble 91 }; 92 93 static cl::opt<ActionType> 94 Action(cl::desc("Action to perform:"), 95 cl::init(AC_Assemble), 96 cl::values(clEnumValN(AC_AsLex, "as-lex", 97 "Lex tokens from a .s file"), 98 clEnumValN(AC_Assemble, "assemble", 99 "Assemble a .s file (default)"), 100 clEnumValN(AC_Disassemble, "disassemble", 101 "Disassemble strings of hex bytes"), 102 clEnumValEnd)); 103 104 static const Target *GetTarget(const char *ProgName) { 105 // Figure out the target triple. 106 if (TripleName.empty()) 107 TripleName = sys::getHostTriple(); 108 if (!ArchName.empty()) { 109 llvm::Triple TT(TripleName); 110 TT.setArchName(ArchName); 111 TripleName = TT.str(); 112 } 113 114 // Get the target specific parser. 115 std::string Error; 116 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 117 if (TheTarget) 118 return TheTarget; 119 120 errs() << ProgName << ": error: unable to get target for '" << TripleName 121 << "', see --version and --triple.\n"; 122 return 0; 123 } 124 125 static int AsLexInput(const char *ProgName) { 126 std::string ErrorMessage; 127 MemoryBuffer *Buffer = MemoryBuffer::getFileOrSTDIN(InputFilename, 128 &ErrorMessage); 129 if (Buffer == 0) { 130 errs() << ProgName << ": "; 131 if (ErrorMessage.size()) 132 errs() << ErrorMessage << "\n"; 133 else 134 errs() << "input file didn't read correctly.\n"; 135 return 1; 136 } 137 138 SourceMgr SrcMgr; 139 140 // Tell SrcMgr about this buffer, which is what TGParser will pick up. 141 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc()); 142 143 // Record the location of the include directories so that the lexer can find 144 // it later. 145 SrcMgr.setIncludeDirs(IncludeDirs); 146 147 const Target *TheTarget = GetTarget(ProgName); 148 if (!TheTarget) 149 return 1; 150 151 const MCAsmInfo *MAI = TheTarget->createAsmInfo(TripleName); 152 assert(MAI && "Unable to create target asm info!"); 153 154 AsmLexer Lexer(*MAI); 155 156 bool Error = false; 157 158 while (Lexer.Lex().isNot(AsmToken::Eof)) { 159 switch (Lexer.getKind()) { 160 default: 161 SrcMgr.PrintMessage(Lexer.getLoc(), "unknown token", "warning"); 162 Error = true; 163 break; 164 case AsmToken::Error: 165 Error = true; // error already printed. 166 break; 167 case AsmToken::Identifier: 168 outs() << "identifier: " << Lexer.getTok().getString() << '\n'; 169 break; 170 case AsmToken::String: 171 outs() << "string: " << Lexer.getTok().getString() << '\n'; 172 break; 173 case AsmToken::Integer: 174 outs() << "int: " << Lexer.getTok().getString() << '\n'; 175 break; 176 177 case AsmToken::Amp: outs() << "Amp\n"; break; 178 case AsmToken::AmpAmp: outs() << "AmpAmp\n"; break; 179 case AsmToken::Caret: outs() << "Caret\n"; break; 180 case AsmToken::Colon: outs() << "Colon\n"; break; 181 case AsmToken::Comma: outs() << "Comma\n"; break; 182 case AsmToken::Dollar: outs() << "Dollar\n"; break; 183 case AsmToken::EndOfStatement: outs() << "EndOfStatement\n"; break; 184 case AsmToken::Eof: outs() << "Eof\n"; break; 185 case AsmToken::Equal: outs() << "Equal\n"; break; 186 case AsmToken::EqualEqual: outs() << "EqualEqual\n"; break; 187 case AsmToken::Exclaim: outs() << "Exclaim\n"; break; 188 case AsmToken::ExclaimEqual: outs() << "ExclaimEqual\n"; break; 189 case AsmToken::Greater: outs() << "Greater\n"; break; 190 case AsmToken::GreaterEqual: outs() << "GreaterEqual\n"; break; 191 case AsmToken::GreaterGreater: outs() << "GreaterGreater\n"; break; 192 case AsmToken::LParen: outs() << "LParen\n"; break; 193 case AsmToken::Less: outs() << "Less\n"; break; 194 case AsmToken::LessEqual: outs() << "LessEqual\n"; break; 195 case AsmToken::LessGreater: outs() << "LessGreater\n"; break; 196 case AsmToken::LessLess: outs() << "LessLess\n"; break; 197 case AsmToken::Minus: outs() << "Minus\n"; break; 198 case AsmToken::Percent: outs() << "Percent\n"; break; 199 case AsmToken::Pipe: outs() << "Pipe\n"; break; 200 case AsmToken::PipePipe: outs() << "PipePipe\n"; break; 201 case AsmToken::Plus: outs() << "Plus\n"; break; 202 case AsmToken::RParen: outs() << "RParen\n"; break; 203 case AsmToken::Slash: outs() << "Slash\n"; break; 204 case AsmToken::Star: outs() << "Star\n"; break; 205 case AsmToken::Tilde: outs() << "Tilde\n"; break; 206 } 207 } 208 209 return Error; 210 } 211 212 static formatted_raw_ostream *GetOutputStream() { 213 if (OutputFilename == "") 214 OutputFilename = "-"; 215 216 // Make sure that the Out file gets unlinked from the disk if we get a 217 // SIGINT. 218 if (OutputFilename != "-") 219 sys::RemoveFileOnSignal(sys::Path(OutputFilename)); 220 221 std::string Err; 222 raw_fd_ostream *Out = new raw_fd_ostream(OutputFilename.c_str(), Err, 223 raw_fd_ostream::F_Binary); 224 if (!Err.empty()) { 225 errs() << Err << '\n'; 226 delete Out; 227 return 0; 228 } 229 230 return new formatted_raw_ostream(*Out, formatted_raw_ostream::DELETE_STREAM); 231 } 232 233 static int AssembleInput(const char *ProgName) { 234 const Target *TheTarget = GetTarget(ProgName); 235 if (!TheTarget) 236 return 1; 237 238 std::string Error; 239 MemoryBuffer *Buffer = MemoryBuffer::getFileOrSTDIN(InputFilename, &Error); 240 if (Buffer == 0) { 241 errs() << ProgName << ": "; 242 if (Error.size()) 243 errs() << Error << "\n"; 244 else 245 errs() << "input file didn't read correctly.\n"; 246 return 1; 247 } 248 249 SourceMgr SrcMgr; 250 251 // Tell SrcMgr about this buffer, which is what the parser will pick up. 252 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc()); 253 254 // Record the location of the include directories so that the lexer can find 255 // it later. 256 SrcMgr.setIncludeDirs(IncludeDirs); 257 258 259 const MCAsmInfo *MAI = TheTarget->createAsmInfo(TripleName); 260 assert(MAI && "Unable to create target asm info!"); 261 262 MCContext Ctx(*MAI); 263 formatted_raw_ostream *Out = GetOutputStream(); 264 if (!Out) 265 return 1; 266 267 268 // FIXME: We shouldn't need to do this (and link in codegen). 269 OwningPtr<TargetMachine> TM(TheTarget->createTargetMachine(TripleName, "")); 270 271 if (!TM) { 272 errs() << ProgName << ": error: could not create target for triple '" 273 << TripleName << "'.\n"; 274 return 1; 275 } 276 277 OwningPtr<MCInstPrinter> IP; 278 OwningPtr<MCCodeEmitter> CE; 279 OwningPtr<MCStreamer> Str; 280 OwningPtr<TargetAsmBackend> TAB; 281 282 if (FileType == OFT_AssemblyFile) { 283 IP.reset(TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI, *Out)); 284 if (ShowEncoding) 285 CE.reset(TheTarget->createCodeEmitter(*TM, Ctx)); 286 Str.reset(createAsmStreamer(Ctx, *Out,TM->getTargetData()->isLittleEndian(), 287 /*asmverbose*/true, IP.get(), CE.get(), 288 ShowInst)); 289 } else { 290 assert(FileType == OFT_ObjectFile && "Invalid file type!"); 291 CE.reset(TheTarget->createCodeEmitter(*TM, Ctx)); 292 TAB.reset(TheTarget->createAsmBackend(TripleName)); 293 Str.reset(createMachOStreamer(Ctx, *TAB, *Out, CE.get())); 294 } 295 296 AsmParser Parser(SrcMgr, Ctx, *Str.get(), *MAI); 297 OwningPtr<TargetAsmParser> TAP(TheTarget->createAsmParser(Parser)); 298 if (!TAP) { 299 errs() << ProgName 300 << ": error: this target does not support assembly parsing.\n"; 301 return 1; 302 } 303 304 Parser.setTargetParser(*TAP.get()); 305 306 int Res = Parser.Run(); 307 if (Out != &fouts()) 308 delete Out; 309 310 return Res; 311 } 312 313 static int DisassembleInput(const char *ProgName) { 314 std::string Error; 315 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 316 if (TheTarget == 0) { 317 errs() << ProgName << ": error: unable to get target for '" << TripleName 318 << "', see --version and --triple.\n"; 319 return 0; 320 } 321 322 std::string ErrorMessage; 323 324 MemoryBuffer *Buffer = MemoryBuffer::getFileOrSTDIN(InputFilename, 325 &ErrorMessage); 326 327 if (Buffer == 0) { 328 errs() << ProgName << ": "; 329 if (ErrorMessage.size()) 330 errs() << ErrorMessage << "\n"; 331 else 332 errs() << "input file didn't read correctly.\n"; 333 return 1; 334 } 335 336 return Disassembler::disassemble(*TheTarget, TripleName, *Buffer); 337 } 338 339 340 int main(int argc, char **argv) { 341 // Print a stack trace if we signal out. 342 sys::PrintStackTraceOnErrorSignal(); 343 PrettyStackTraceProgram X(argc, argv); 344 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 345 346 // Initialize targets and assembly printers/parsers. 347 llvm::InitializeAllTargetInfos(); 348 // FIXME: We shouldn't need to initialize the Target(Machine)s. 349 llvm::InitializeAllTargets(); 350 llvm::InitializeAllAsmPrinters(); 351 llvm::InitializeAllAsmParsers(); 352 llvm::InitializeAllDisassemblers(); 353 354 cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n"); 355 356 switch (Action) { 357 default: 358 case AC_AsLex: 359 return AsLexInput(argv[0]); 360 case AC_Assemble: 361 return AssembleInput(argv[0]); 362 case AC_Disassemble: 363 return DisassembleInput(argv[0]); 364 } 365 366 return 0; 367 } 368 369