1 //===-- llvm-mc.cpp - Machine Code Hacking Driver ---------------*- C++ -*-===// 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 "Disassembler.h" 16 #include "llvm/MC/MCAsmBackend.h" 17 #include "llvm/MC/MCAsmInfo.h" 18 #include "llvm/MC/MCContext.h" 19 #include "llvm/MC/MCInstPrinter.h" 20 #include "llvm/MC/MCInstrInfo.h" 21 #include "llvm/MC/MCObjectFileInfo.h" 22 #include "llvm/MC/MCParser/AsmLexer.h" 23 #include "llvm/MC/MCParser/MCTargetAsmParser.h" 24 #include "llvm/MC/MCRegisterInfo.h" 25 #include "llvm/MC/MCSectionMachO.h" 26 #include "llvm/MC/MCStreamer.h" 27 #include "llvm/MC/MCSubtargetInfo.h" 28 #include "llvm/MC/MCTargetOptionsCommandFlags.h" 29 #include "llvm/Support/CommandLine.h" 30 #include "llvm/Support/Compression.h" 31 #include "llvm/Support/FileUtilities.h" 32 #include "llvm/Support/FormattedStream.h" 33 #include "llvm/Support/Host.h" 34 #include "llvm/Support/ManagedStatic.h" 35 #include "llvm/Support/MemoryBuffer.h" 36 #include "llvm/Support/PrettyStackTrace.h" 37 #include "llvm/Support/Signals.h" 38 #include "llvm/Support/SourceMgr.h" 39 #include "llvm/Support/TargetRegistry.h" 40 #include "llvm/Support/TargetSelect.h" 41 #include "llvm/Support/ToolOutputFile.h" 42 43 using namespace llvm; 44 45 static cl::opt<std::string> 46 InputFilename(cl::Positional, cl::desc("<input file>"), cl::init("-")); 47 48 static cl::opt<std::string> 49 OutputFilename("o", cl::desc("Output filename"), 50 cl::value_desc("filename")); 51 52 static cl::opt<bool> 53 ShowEncoding("show-encoding", cl::desc("Show instruction encodings")); 54 55 static cl::opt<bool> 56 CompressDebugSections("compress-debug-sections", 57 cl::desc("Compress DWARF debug sections")); 58 59 static cl::opt<bool> 60 ShowInst("show-inst", cl::desc("Show internal instruction representation")); 61 62 static cl::opt<bool> 63 ShowInstOperands("show-inst-operands", 64 cl::desc("Show instructions operands as parsed")); 65 66 static cl::opt<unsigned> 67 OutputAsmVariant("output-asm-variant", 68 cl::desc("Syntax variant to use for output printing")); 69 70 static cl::opt<bool> 71 PrintImmHex("print-imm-hex", cl::init(false), 72 cl::desc("Prefer hex format for immediate values")); 73 74 static cl::list<std::string> 75 DefineSymbol("defsym", cl::desc("Defines a symbol to be an integer constant")); 76 77 enum OutputFileType { 78 OFT_Null, 79 OFT_AssemblyFile, 80 OFT_ObjectFile 81 }; 82 static cl::opt<OutputFileType> 83 FileType("filetype", cl::init(OFT_AssemblyFile), 84 cl::desc("Choose an output file type:"), 85 cl::values( 86 clEnumValN(OFT_AssemblyFile, "asm", 87 "Emit an assembly ('.s') file"), 88 clEnumValN(OFT_Null, "null", 89 "Don't emit anything (for timing purposes)"), 90 clEnumValN(OFT_ObjectFile, "obj", 91 "Emit a native object ('.o') file"), 92 clEnumValEnd)); 93 94 static cl::list<std::string> 95 IncludeDirs("I", cl::desc("Directory of include files"), 96 cl::value_desc("directory"), cl::Prefix); 97 98 static cl::opt<std::string> 99 ArchName("arch", cl::desc("Target arch to assemble for, " 100 "see -version for available targets")); 101 102 static cl::opt<std::string> 103 TripleName("triple", cl::desc("Target triple to assemble for, " 104 "see -version for available targets")); 105 106 static cl::opt<std::string> 107 MCPU("mcpu", 108 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 109 cl::value_desc("cpu-name"), 110 cl::init("")); 111 112 static cl::list<std::string> 113 MAttrs("mattr", 114 cl::CommaSeparated, 115 cl::desc("Target specific attributes (-mattr=help for details)"), 116 cl::value_desc("a1,+a2,-a3,...")); 117 118 static cl::opt<bool> PIC("position-independent", 119 cl::desc("Position independent"), cl::init(false)); 120 121 static cl::opt<llvm::CodeModel::Model> 122 CMModel("code-model", 123 cl::desc("Choose code model"), 124 cl::init(CodeModel::Default), 125 cl::values(clEnumValN(CodeModel::Default, "default", 126 "Target default code model"), 127 clEnumValN(CodeModel::Small, "small", 128 "Small code model"), 129 clEnumValN(CodeModel::Kernel, "kernel", 130 "Kernel code model"), 131 clEnumValN(CodeModel::Medium, "medium", 132 "Medium code model"), 133 clEnumValN(CodeModel::Large, "large", 134 "Large code model"), 135 clEnumValEnd)); 136 137 static cl::opt<bool> 138 NoInitialTextSection("n", cl::desc("Don't assume assembly file starts " 139 "in the text section")); 140 141 static cl::opt<bool> 142 GenDwarfForAssembly("g", cl::desc("Generate dwarf debugging info for assembly " 143 "source files")); 144 145 static cl::opt<std::string> 146 DebugCompilationDir("fdebug-compilation-dir", 147 cl::desc("Specifies the debug info's compilation dir")); 148 149 static cl::opt<std::string> 150 MainFileName("main-file-name", 151 cl::desc("Specifies the name we should consider the input file")); 152 153 static cl::opt<bool> SaveTempLabels("save-temp-labels", 154 cl::desc("Don't discard temporary labels")); 155 156 static cl::opt<bool> NoExecStack("no-exec-stack", 157 cl::desc("File doesn't need an exec stack")); 158 159 enum ActionType { 160 AC_AsLex, 161 AC_Assemble, 162 AC_Disassemble, 163 AC_MDisassemble, 164 }; 165 166 static cl::opt<ActionType> 167 Action(cl::desc("Action to perform:"), 168 cl::init(AC_Assemble), 169 cl::values(clEnumValN(AC_AsLex, "as-lex", 170 "Lex tokens from a .s file"), 171 clEnumValN(AC_Assemble, "assemble", 172 "Assemble a .s file (default)"), 173 clEnumValN(AC_Disassemble, "disassemble", 174 "Disassemble strings of hex bytes"), 175 clEnumValN(AC_MDisassemble, "mdis", 176 "Marked up disassembly of strings of hex bytes"), 177 clEnumValEnd)); 178 179 static const Target *GetTarget(const char *ProgName) { 180 // Figure out the target triple. 181 if (TripleName.empty()) 182 TripleName = sys::getDefaultTargetTriple(); 183 Triple TheTriple(Triple::normalize(TripleName)); 184 185 // Get the target specific parser. 186 std::string Error; 187 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 188 Error); 189 if (!TheTarget) { 190 errs() << ProgName << ": " << Error; 191 return nullptr; 192 } 193 194 // Update the triple name and return the found target. 195 TripleName = TheTriple.getTriple(); 196 return TheTarget; 197 } 198 199 static std::unique_ptr<tool_output_file> GetOutputStream() { 200 if (OutputFilename == "") 201 OutputFilename = "-"; 202 203 std::error_code EC; 204 auto Out = llvm::make_unique<tool_output_file>(OutputFilename, EC, 205 sys::fs::F_None); 206 if (EC) { 207 errs() << EC.message() << '\n'; 208 return nullptr; 209 } 210 211 return Out; 212 } 213 214 static std::string DwarfDebugFlags; 215 static void setDwarfDebugFlags(int argc, char **argv) { 216 if (!getenv("RC_DEBUG_OPTIONS")) 217 return; 218 for (int i = 0; i < argc; i++) { 219 DwarfDebugFlags += argv[i]; 220 if (i + 1 < argc) 221 DwarfDebugFlags += " "; 222 } 223 } 224 225 static std::string DwarfDebugProducer; 226 static void setDwarfDebugProducer() { 227 if(!getenv("DEBUG_PRODUCER")) 228 return; 229 DwarfDebugProducer += getenv("DEBUG_PRODUCER"); 230 } 231 232 static int AsLexInput(SourceMgr &SrcMgr, MCAsmInfo &MAI, 233 raw_ostream &OS) { 234 235 AsmLexer Lexer(MAI); 236 Lexer.setBuffer(SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer()); 237 238 bool Error = false; 239 while (Lexer.Lex().isNot(AsmToken::Eof)) { 240 AsmToken Tok = Lexer.getTok(); 241 242 switch (Tok.getKind()) { 243 default: 244 SrcMgr.PrintMessage(Lexer.getLoc(), SourceMgr::DK_Warning, 245 "unknown token"); 246 Error = true; 247 break; 248 case AsmToken::Error: 249 Error = true; // error already printed. 250 break; 251 case AsmToken::Identifier: 252 OS << "identifier: " << Lexer.getTok().getString(); 253 break; 254 case AsmToken::Integer: 255 OS << "int: " << Lexer.getTok().getString(); 256 break; 257 case AsmToken::Real: 258 OS << "real: " << Lexer.getTok().getString(); 259 break; 260 case AsmToken::String: 261 OS << "string: " << Lexer.getTok().getString(); 262 break; 263 264 case AsmToken::Amp: OS << "Amp"; break; 265 case AsmToken::AmpAmp: OS << "AmpAmp"; break; 266 case AsmToken::At: OS << "At"; break; 267 case AsmToken::Caret: OS << "Caret"; break; 268 case AsmToken::Colon: OS << "Colon"; break; 269 case AsmToken::Comma: OS << "Comma"; break; 270 case AsmToken::Dollar: OS << "Dollar"; break; 271 case AsmToken::Dot: OS << "Dot"; break; 272 case AsmToken::EndOfStatement: OS << "EndOfStatement"; break; 273 case AsmToken::Eof: OS << "Eof"; break; 274 case AsmToken::Equal: OS << "Equal"; break; 275 case AsmToken::EqualEqual: OS << "EqualEqual"; break; 276 case AsmToken::Exclaim: OS << "Exclaim"; break; 277 case AsmToken::ExclaimEqual: OS << "ExclaimEqual"; break; 278 case AsmToken::Greater: OS << "Greater"; break; 279 case AsmToken::GreaterEqual: OS << "GreaterEqual"; break; 280 case AsmToken::GreaterGreater: OS << "GreaterGreater"; break; 281 case AsmToken::Hash: OS << "Hash"; break; 282 case AsmToken::LBrac: OS << "LBrac"; break; 283 case AsmToken::LCurly: OS << "LCurly"; break; 284 case AsmToken::LParen: OS << "LParen"; break; 285 case AsmToken::Less: OS << "Less"; break; 286 case AsmToken::LessEqual: OS << "LessEqual"; break; 287 case AsmToken::LessGreater: OS << "LessGreater"; break; 288 case AsmToken::LessLess: OS << "LessLess"; break; 289 case AsmToken::Minus: OS << "Minus"; break; 290 case AsmToken::Percent: OS << "Percent"; break; 291 case AsmToken::Pipe: OS << "Pipe"; break; 292 case AsmToken::PipePipe: OS << "PipePipe"; break; 293 case AsmToken::Plus: OS << "Plus"; break; 294 case AsmToken::RBrac: OS << "RBrac"; break; 295 case AsmToken::RCurly: OS << "RCurly"; break; 296 case AsmToken::RParen: OS << "RParen"; break; 297 case AsmToken::Slash: OS << "Slash"; break; 298 case AsmToken::Star: OS << "Star"; break; 299 case AsmToken::Tilde: OS << "Tilde"; break; 300 } 301 302 // Print the token string. 303 OS << " (\""; 304 OS.write_escaped(Tok.getString()); 305 OS << "\")\n"; 306 } 307 308 return Error; 309 } 310 311 static int fillCommandLineSymbols(MCAsmParser &Parser){ 312 for(auto &I: DefineSymbol){ 313 auto Pair = StringRef(I).split('='); 314 if(Pair.second.empty()){ 315 errs() << "error: defsym must be of the form: sym=value: " << I; 316 return 1; 317 } 318 int64_t Value; 319 if(Pair.second.getAsInteger(0, Value)){ 320 errs() << "error: Value is not an integer: " << Pair.second; 321 return 1; 322 } 323 auto &Context = Parser.getContext(); 324 auto Symbol = Context.getOrCreateSymbol(Pair.first); 325 Parser.getStreamer().EmitAssignment(Symbol, 326 MCConstantExpr::create(Value, Context)); 327 } 328 return 0; 329 } 330 331 static int AssembleInput(const char *ProgName, const Target *TheTarget, 332 SourceMgr &SrcMgr, MCContext &Ctx, MCStreamer &Str, 333 MCAsmInfo &MAI, MCSubtargetInfo &STI, 334 MCInstrInfo &MCII, MCTargetOptions &MCOptions) { 335 std::unique_ptr<MCAsmParser> Parser( 336 createMCAsmParser(SrcMgr, Ctx, Str, MAI)); 337 std::unique_ptr<MCTargetAsmParser> TAP( 338 TheTarget->createMCAsmParser(STI, *Parser, MCII, MCOptions)); 339 340 if (!TAP) { 341 errs() << ProgName 342 << ": error: this target does not support assembly parsing.\n"; 343 return 1; 344 } 345 346 int SymbolResult = fillCommandLineSymbols(*Parser); 347 if(SymbolResult) 348 return SymbolResult; 349 Parser->setShowParsedOperands(ShowInstOperands); 350 Parser->setTargetParser(*TAP); 351 352 int Res = Parser->Run(NoInitialTextSection); 353 354 return Res; 355 } 356 357 int main(int argc, char **argv) { 358 // Print a stack trace if we signal out. 359 sys::PrintStackTraceOnErrorSignal(); 360 PrettyStackTraceProgram X(argc, argv); 361 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 362 363 // Initialize targets and assembly printers/parsers. 364 llvm::InitializeAllTargetInfos(); 365 llvm::InitializeAllTargetMCs(); 366 llvm::InitializeAllAsmParsers(); 367 llvm::InitializeAllDisassemblers(); 368 369 // Register the target printer for --version. 370 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 371 372 cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n"); 373 MCTargetOptions MCOptions = InitMCTargetOptionsFromFlags(); 374 TripleName = Triple::normalize(TripleName); 375 setDwarfDebugFlags(argc, argv); 376 377 setDwarfDebugProducer(); 378 379 const char *ProgName = argv[0]; 380 const Target *TheTarget = GetTarget(ProgName); 381 if (!TheTarget) 382 return 1; 383 // Now that GetTarget() has (potentially) replaced TripleName, it's safe to 384 // construct the Triple object. 385 Triple TheTriple(TripleName); 386 387 ErrorOr<std::unique_ptr<MemoryBuffer>> BufferPtr = 388 MemoryBuffer::getFileOrSTDIN(InputFilename); 389 if (std::error_code EC = BufferPtr.getError()) { 390 errs() << InputFilename << ": " << EC.message() << '\n'; 391 return 1; 392 } 393 MemoryBuffer *Buffer = BufferPtr->get(); 394 395 SourceMgr SrcMgr; 396 397 // Tell SrcMgr about this buffer, which is what the parser will pick up. 398 SrcMgr.AddNewSourceBuffer(std::move(*BufferPtr), SMLoc()); 399 400 // Record the location of the include directories so that the lexer can find 401 // it later. 402 SrcMgr.setIncludeDirs(IncludeDirs); 403 404 std::unique_ptr<MCRegisterInfo> MRI(TheTarget->createMCRegInfo(TripleName)); 405 assert(MRI && "Unable to create target register info!"); 406 407 std::unique_ptr<MCAsmInfo> MAI(TheTarget->createMCAsmInfo(*MRI, TripleName)); 408 assert(MAI && "Unable to create target asm info!"); 409 410 if (CompressDebugSections) { 411 if (!zlib::isAvailable()) { 412 errs() << ProgName 413 << ": build tools with zlib to enable -compress-debug-sections"; 414 return 1; 415 } 416 MAI->setCompressDebugSections(true); 417 } 418 419 // FIXME: This is not pretty. MCContext has a ptr to MCObjectFileInfo and 420 // MCObjectFileInfo needs a MCContext reference in order to initialize itself. 421 MCObjectFileInfo MOFI; 422 MCContext Ctx(MAI.get(), MRI.get(), &MOFI, &SrcMgr); 423 MOFI.InitMCObjectFileInfo(TheTriple, PIC, CMModel, Ctx); 424 425 if (SaveTempLabels) 426 Ctx.setAllowTemporaryLabels(false); 427 428 Ctx.setGenDwarfForAssembly(GenDwarfForAssembly); 429 // Default to 4 for dwarf version. 430 unsigned DwarfVersion = MCOptions.DwarfVersion ? MCOptions.DwarfVersion : 4; 431 if (DwarfVersion < 2 || DwarfVersion > 4) { 432 errs() << ProgName << ": Dwarf version " << DwarfVersion 433 << " is not supported." << '\n'; 434 return 1; 435 } 436 Ctx.setDwarfVersion(DwarfVersion); 437 if (!DwarfDebugFlags.empty()) 438 Ctx.setDwarfDebugFlags(StringRef(DwarfDebugFlags)); 439 if (!DwarfDebugProducer.empty()) 440 Ctx.setDwarfDebugProducer(StringRef(DwarfDebugProducer)); 441 if (!DebugCompilationDir.empty()) 442 Ctx.setCompilationDir(DebugCompilationDir); 443 else { 444 // If no compilation dir is set, try to use the current directory. 445 SmallString<128> CWD; 446 if (!sys::fs::current_path(CWD)) 447 Ctx.setCompilationDir(CWD); 448 } 449 if (!MainFileName.empty()) 450 Ctx.setMainFileName(MainFileName); 451 452 // Package up features to be passed to target/subtarget 453 std::string FeaturesStr; 454 if (MAttrs.size()) { 455 SubtargetFeatures Features; 456 for (unsigned i = 0; i != MAttrs.size(); ++i) 457 Features.AddFeature(MAttrs[i]); 458 FeaturesStr = Features.getString(); 459 } 460 461 std::unique_ptr<tool_output_file> Out = GetOutputStream(); 462 if (!Out) 463 return 1; 464 465 std::unique_ptr<buffer_ostream> BOS; 466 raw_pwrite_stream *OS = &Out->os(); 467 std::unique_ptr<MCStreamer> Str; 468 469 std::unique_ptr<MCInstrInfo> MCII(TheTarget->createMCInstrInfo()); 470 std::unique_ptr<MCSubtargetInfo> STI( 471 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 472 473 MCInstPrinter *IP = nullptr; 474 if (FileType == OFT_AssemblyFile) { 475 IP = TheTarget->createMCInstPrinter(Triple(TripleName), OutputAsmVariant, 476 *MAI, *MCII, *MRI); 477 478 // Set the display preference for hex vs. decimal immediates. 479 IP->setPrintImmHex(PrintImmHex); 480 481 // Set up the AsmStreamer. 482 MCCodeEmitter *CE = nullptr; 483 MCAsmBackend *MAB = nullptr; 484 if (ShowEncoding) { 485 CE = TheTarget->createMCCodeEmitter(*MCII, *MRI, Ctx); 486 MAB = TheTarget->createMCAsmBackend(*MRI, TripleName, MCPU); 487 } 488 auto FOut = llvm::make_unique<formatted_raw_ostream>(*OS); 489 Str.reset(TheTarget->createAsmStreamer( 490 Ctx, std::move(FOut), /*asmverbose*/ true, 491 /*useDwarfDirectory*/ true, IP, CE, MAB, ShowInst)); 492 493 } else if (FileType == OFT_Null) { 494 Str.reset(TheTarget->createNullStreamer(Ctx)); 495 } else { 496 assert(FileType == OFT_ObjectFile && "Invalid file type!"); 497 498 // Don't waste memory on names of temp labels. 499 Ctx.setUseNamesOnTempLabels(false); 500 501 if (!Out->os().supportsSeeking()) { 502 BOS = make_unique<buffer_ostream>(Out->os()); 503 OS = BOS.get(); 504 } 505 506 MCCodeEmitter *CE = TheTarget->createMCCodeEmitter(*MCII, *MRI, Ctx); 507 MCAsmBackend *MAB = TheTarget->createMCAsmBackend(*MRI, TripleName, MCPU); 508 Str.reset(TheTarget->createMCObjectStreamer( 509 TheTriple, Ctx, *MAB, *OS, CE, *STI, MCOptions.MCRelaxAll, 510 MCOptions.MCIncrementalLinkerCompatible, 511 /*DWARFMustBeAtTheEnd*/ false)); 512 if (NoExecStack) 513 Str->InitSections(true); 514 } 515 516 int Res = 1; 517 bool disassemble = false; 518 switch (Action) { 519 case AC_AsLex: 520 Res = AsLexInput(SrcMgr, *MAI, Out->os()); 521 break; 522 case AC_Assemble: 523 Res = AssembleInput(ProgName, TheTarget, SrcMgr, Ctx, *Str, *MAI, *STI, 524 *MCII, MCOptions); 525 break; 526 case AC_MDisassemble: 527 assert(IP && "Expected assembly output"); 528 IP->setUseMarkup(1); 529 disassemble = true; 530 break; 531 case AC_Disassemble: 532 disassemble = true; 533 break; 534 } 535 if (disassemble) 536 Res = Disassembler::disassemble(*TheTarget, TripleName, *STI, *Str, 537 *Buffer, SrcMgr, Out->os()); 538 539 // Keep output if no errors. 540 if (Res == 0) Out->keep(); 541 return Res; 542 } 543