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 "Disassembler.h" 16 #include "llvm/ADT/OwningPtr.h" 17 #include "llvm/MC/MCAsmBackend.h" 18 #include "llvm/MC/MCAsmInfo.h" 19 #include "llvm/MC/MCContext.h" 20 #include "llvm/MC/MCInstPrinter.h" 21 #include "llvm/MC/MCInstrInfo.h" 22 #include "llvm/MC/MCObjectFileInfo.h" 23 #include "llvm/MC/MCParser/AsmLexer.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/MCTargetAsmParser.h" 29 #include "llvm/Support/CommandLine.h" 30 #include "llvm/Support/FileUtilities.h" 31 #include "llvm/Support/FormattedStream.h" 32 #include "llvm/Support/Host.h" 33 #include "llvm/Support/ManagedStatic.h" 34 #include "llvm/Support/MemoryBuffer.h" 35 #include "llvm/Support/PrettyStackTrace.h" 36 #include "llvm/Support/Signals.h" 37 #include "llvm/Support/SourceMgr.h" 38 #include "llvm/Support/TargetRegistry.h" 39 #include "llvm/Support/TargetSelect.h" 40 #include "llvm/Support/ToolOutputFile.h" 41 using namespace llvm; 42 43 static cl::opt<std::string> 44 InputFilename(cl::Positional, cl::desc("<input file>"), cl::init("-")); 45 46 static cl::opt<std::string> 47 OutputFilename("o", cl::desc("Output filename"), 48 cl::value_desc("filename")); 49 50 static cl::opt<bool> 51 ShowEncoding("show-encoding", cl::desc("Show instruction encodings")); 52 53 static cl::opt<bool> 54 ShowInst("show-inst", cl::desc("Show internal instruction representation")); 55 56 static cl::opt<bool> 57 ShowInstOperands("show-inst-operands", 58 cl::desc("Show instructions operands as parsed")); 59 60 static cl::opt<unsigned> 61 OutputAsmVariant("output-asm-variant", 62 cl::desc("Syntax variant to use for output printing")); 63 64 static cl::opt<bool> 65 RelaxAll("mc-relax-all", cl::desc("Relax all fixups")); 66 67 static cl::opt<bool> 68 DisableCFI("disable-cfi", cl::desc("Do not use .cfi_* directives")); 69 70 static cl::opt<bool> 71 NoExecStack("mc-no-exec-stack", cl::desc("File doesn't need an exec stack")); 72 73 enum OutputFileType { 74 OFT_Null, 75 OFT_AssemblyFile, 76 OFT_ObjectFile 77 }; 78 static cl::opt<OutputFileType> 79 FileType("filetype", cl::init(OFT_AssemblyFile), 80 cl::desc("Choose an output file type:"), 81 cl::values( 82 clEnumValN(OFT_AssemblyFile, "asm", 83 "Emit an assembly ('.s') file"), 84 clEnumValN(OFT_Null, "null", 85 "Don't emit anything (for timing purposes)"), 86 clEnumValN(OFT_ObjectFile, "obj", 87 "Emit a native object ('.o') file"), 88 clEnumValEnd)); 89 90 static cl::list<std::string> 91 IncludeDirs("I", cl::desc("Directory of include files"), 92 cl::value_desc("directory"), cl::Prefix); 93 94 static cl::opt<std::string> 95 ArchName("arch", cl::desc("Target arch to assemble for, " 96 "see -version for available targets")); 97 98 static cl::opt<std::string> 99 TripleName("triple", cl::desc("Target triple to assemble for, " 100 "see -version for available targets")); 101 102 static cl::opt<std::string> 103 MCPU("mcpu", 104 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 105 cl::value_desc("cpu-name"), 106 cl::init("")); 107 108 static cl::list<std::string> 109 MAttrs("mattr", 110 cl::CommaSeparated, 111 cl::desc("Target specific attributes (-mattr=help for details)"), 112 cl::value_desc("a1,+a2,-a3,...")); 113 114 static cl::opt<Reloc::Model> 115 RelocModel("relocation-model", 116 cl::desc("Choose relocation model"), 117 cl::init(Reloc::Default), 118 cl::values( 119 clEnumValN(Reloc::Default, "default", 120 "Target default relocation model"), 121 clEnumValN(Reloc::Static, "static", 122 "Non-relocatable code"), 123 clEnumValN(Reloc::PIC_, "pic", 124 "Fully relocatable, position independent code"), 125 clEnumValN(Reloc::DynamicNoPIC, "dynamic-no-pic", 126 "Relocatable external references, non-relocatable code"), 127 clEnumValEnd)); 128 129 static cl::opt<llvm::CodeModel::Model> 130 CMModel("code-model", 131 cl::desc("Choose code model"), 132 cl::init(CodeModel::Default), 133 cl::values(clEnumValN(CodeModel::Default, "default", 134 "Target default code model"), 135 clEnumValN(CodeModel::Small, "small", 136 "Small code model"), 137 clEnumValN(CodeModel::Kernel, "kernel", 138 "Kernel code model"), 139 clEnumValN(CodeModel::Medium, "medium", 140 "Medium code model"), 141 clEnumValN(CodeModel::Large, "large", 142 "Large code model"), 143 clEnumValEnd)); 144 145 static cl::opt<bool> 146 NoInitialTextSection("n", cl::desc("Don't assume assembly file starts " 147 "in the text section")); 148 149 static cl::opt<bool> 150 SaveTempLabels("L", cl::desc("Don't discard temporary labels")); 151 152 static cl::opt<bool> 153 GenDwarfForAssembly("g", cl::desc("Generate dwarf debugging info for assembly " 154 "source files")); 155 156 static cl::opt<std::string> 157 DebugCompilationDir("fdebug-compilation-dir", 158 cl::desc("Specifies the debug info's compilation dir")); 159 160 static cl::opt<std::string> 161 MainFileName("main-file-name", 162 cl::desc("Specifies the name we should consider the input file")); 163 164 enum ActionType { 165 AC_AsLex, 166 AC_Assemble, 167 AC_Disassemble, 168 AC_MDisassemble, 169 AC_HDisassemble 170 }; 171 172 static cl::opt<ActionType> 173 Action(cl::desc("Action to perform:"), 174 cl::init(AC_Assemble), 175 cl::values(clEnumValN(AC_AsLex, "as-lex", 176 "Lex tokens from a .s file"), 177 clEnumValN(AC_Assemble, "assemble", 178 "Assemble a .s file (default)"), 179 clEnumValN(AC_Disassemble, "disassemble", 180 "Disassemble strings of hex bytes"), 181 clEnumValN(AC_MDisassemble, "mdis", 182 "Marked up disassembly of strings of hex bytes"), 183 clEnumValN(AC_HDisassemble, "hdis", 184 "Disassemble strings of hex bytes printing " 185 "immediates as hex"), 186 clEnumValEnd)); 187 188 static const Target *GetTarget(const char *ProgName) { 189 // Figure out the target triple. 190 if (TripleName.empty()) 191 TripleName = sys::getDefaultTargetTriple(); 192 Triple TheTriple(Triple::normalize(TripleName)); 193 194 // Get the target specific parser. 195 std::string Error; 196 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 197 Error); 198 if (!TheTarget) { 199 errs() << ProgName << ": " << Error; 200 return 0; 201 } 202 203 // Update the triple name and return the found target. 204 TripleName = TheTriple.getTriple(); 205 return TheTarget; 206 } 207 208 static tool_output_file *GetOutputStream() { 209 if (OutputFilename == "") 210 OutputFilename = "-"; 211 212 std::string Err; 213 tool_output_file *Out = 214 new tool_output_file(OutputFilename.c_str(), Err, sys::fs::F_Binary); 215 if (!Err.empty()) { 216 errs() << Err << '\n'; 217 delete Out; 218 return 0; 219 } 220 221 return Out; 222 } 223 224 static std::string DwarfDebugFlags; 225 static void setDwarfDebugFlags(int argc, char **argv) { 226 if (!getenv("RC_DEBUG_OPTIONS")) 227 return; 228 for (int i = 0; i < argc; i++) { 229 DwarfDebugFlags += argv[i]; 230 if (i + 1 < argc) 231 DwarfDebugFlags += " "; 232 } 233 } 234 235 static std::string DwarfDebugProducer; 236 static void setDwarfDebugProducer(void) { 237 if(!getenv("DEBUG_PRODUCER")) 238 return; 239 DwarfDebugProducer += getenv("DEBUG_PRODUCER"); 240 } 241 242 static int AsLexInput(SourceMgr &SrcMgr, MCAsmInfo &MAI, tool_output_file *Out) { 243 244 AsmLexer Lexer(MAI); 245 Lexer.setBuffer(SrcMgr.getMemoryBuffer(0)); 246 247 bool Error = false; 248 while (Lexer.Lex().isNot(AsmToken::Eof)) { 249 AsmToken Tok = Lexer.getTok(); 250 251 switch (Tok.getKind()) { 252 default: 253 SrcMgr.PrintMessage(Lexer.getLoc(), SourceMgr::DK_Warning, 254 "unknown token"); 255 Error = true; 256 break; 257 case AsmToken::Error: 258 Error = true; // error already printed. 259 break; 260 case AsmToken::Identifier: 261 Out->os() << "identifier: " << Lexer.getTok().getString(); 262 break; 263 case AsmToken::Integer: 264 Out->os() << "int: " << Lexer.getTok().getString(); 265 break; 266 case AsmToken::Real: 267 Out->os() << "real: " << Lexer.getTok().getString(); 268 break; 269 case AsmToken::String: 270 Out->os() << "string: " << Lexer.getTok().getString(); 271 break; 272 273 case AsmToken::Amp: Out->os() << "Amp"; break; 274 case AsmToken::AmpAmp: Out->os() << "AmpAmp"; break; 275 case AsmToken::At: Out->os() << "At"; break; 276 case AsmToken::Caret: Out->os() << "Caret"; break; 277 case AsmToken::Colon: Out->os() << "Colon"; break; 278 case AsmToken::Comma: Out->os() << "Comma"; break; 279 case AsmToken::Dollar: Out->os() << "Dollar"; break; 280 case AsmToken::Dot: Out->os() << "Dot"; break; 281 case AsmToken::EndOfStatement: Out->os() << "EndOfStatement"; break; 282 case AsmToken::Eof: Out->os() << "Eof"; break; 283 case AsmToken::Equal: Out->os() << "Equal"; break; 284 case AsmToken::EqualEqual: Out->os() << "EqualEqual"; break; 285 case AsmToken::Exclaim: Out->os() << "Exclaim"; break; 286 case AsmToken::ExclaimEqual: Out->os() << "ExclaimEqual"; break; 287 case AsmToken::Greater: Out->os() << "Greater"; break; 288 case AsmToken::GreaterEqual: Out->os() << "GreaterEqual"; break; 289 case AsmToken::GreaterGreater: Out->os() << "GreaterGreater"; break; 290 case AsmToken::Hash: Out->os() << "Hash"; break; 291 case AsmToken::LBrac: Out->os() << "LBrac"; break; 292 case AsmToken::LCurly: Out->os() << "LCurly"; break; 293 case AsmToken::LParen: Out->os() << "LParen"; break; 294 case AsmToken::Less: Out->os() << "Less"; break; 295 case AsmToken::LessEqual: Out->os() << "LessEqual"; break; 296 case AsmToken::LessGreater: Out->os() << "LessGreater"; break; 297 case AsmToken::LessLess: Out->os() << "LessLess"; break; 298 case AsmToken::Minus: Out->os() << "Minus"; break; 299 case AsmToken::Percent: Out->os() << "Percent"; break; 300 case AsmToken::Pipe: Out->os() << "Pipe"; break; 301 case AsmToken::PipePipe: Out->os() << "PipePipe"; break; 302 case AsmToken::Plus: Out->os() << "Plus"; break; 303 case AsmToken::RBrac: Out->os() << "RBrac"; break; 304 case AsmToken::RCurly: Out->os() << "RCurly"; break; 305 case AsmToken::RParen: Out->os() << "RParen"; break; 306 case AsmToken::Slash: Out->os() << "Slash"; break; 307 case AsmToken::Star: Out->os() << "Star"; break; 308 case AsmToken::Tilde: Out->os() << "Tilde"; break; 309 } 310 311 // Print the token string. 312 Out->os() << " (\""; 313 Out->os().write_escaped(Tok.getString()); 314 Out->os() << "\")\n"; 315 } 316 317 return Error; 318 } 319 320 static int AssembleInput(const char *ProgName, const Target *TheTarget, 321 SourceMgr &SrcMgr, MCContext &Ctx, MCStreamer &Str, 322 MCAsmInfo &MAI, MCSubtargetInfo &STI) { 323 OwningPtr<MCAsmParser> Parser(createMCAsmParser(SrcMgr, Ctx, 324 Str, MAI)); 325 OwningPtr<MCTargetAsmParser> TAP(TheTarget->createMCAsmParser(STI, *Parser)); 326 if (!TAP) { 327 errs() << ProgName 328 << ": error: this target does not support assembly parsing.\n"; 329 return 1; 330 } 331 332 Parser->setShowParsedOperands(ShowInstOperands); 333 Parser->setTargetParser(*TAP.get()); 334 335 int Res = Parser->Run(NoInitialTextSection); 336 337 return Res; 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 llvm::InitializeAllTargetMCs(); 349 llvm::InitializeAllAsmParsers(); 350 llvm::InitializeAllDisassemblers(); 351 352 // Register the target printer for --version. 353 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 354 355 cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n"); 356 TripleName = Triple::normalize(TripleName); 357 setDwarfDebugFlags(argc, argv); 358 359 setDwarfDebugProducer(); 360 361 const char *ProgName = argv[0]; 362 const Target *TheTarget = GetTarget(ProgName); 363 if (!TheTarget) 364 return 1; 365 366 OwningPtr<MemoryBuffer> BufferPtr; 367 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) { 368 errs() << ProgName << ": " << ec.message() << '\n'; 369 return 1; 370 } 371 MemoryBuffer *Buffer = BufferPtr.take(); 372 373 SourceMgr SrcMgr; 374 375 // Tell SrcMgr about this buffer, which is what the parser will pick up. 376 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc()); 377 378 // Record the location of the include directories so that the lexer can find 379 // it later. 380 SrcMgr.setIncludeDirs(IncludeDirs); 381 382 llvm::OwningPtr<MCRegisterInfo> MRI(TheTarget->createMCRegInfo(TripleName)); 383 assert(MRI && "Unable to create target register info!"); 384 385 llvm::OwningPtr<MCAsmInfo> MAI(TheTarget->createMCAsmInfo(*MRI, TripleName)); 386 assert(MAI && "Unable to create target asm info!"); 387 388 // FIXME: This is not pretty. MCContext has a ptr to MCObjectFileInfo and 389 // MCObjectFileInfo needs a MCContext reference in order to initialize itself. 390 OwningPtr<MCObjectFileInfo> MOFI(new MCObjectFileInfo()); 391 MCContext Ctx(MAI.get(), MRI.get(), MOFI.get(), &SrcMgr); 392 MOFI->InitMCObjectFileInfo(TripleName, RelocModel, CMModel, Ctx); 393 394 if (SaveTempLabels) 395 Ctx.setAllowTemporaryLabels(false); 396 397 Ctx.setGenDwarfForAssembly(GenDwarfForAssembly); 398 if (!DwarfDebugFlags.empty()) 399 Ctx.setDwarfDebugFlags(StringRef(DwarfDebugFlags)); 400 if (!DwarfDebugProducer.empty()) 401 Ctx.setDwarfDebugProducer(StringRef(DwarfDebugProducer)); 402 if (!DebugCompilationDir.empty()) 403 Ctx.setCompilationDir(DebugCompilationDir); 404 if (!MainFileName.empty()) 405 Ctx.setMainFileName(MainFileName); 406 407 // Package up features to be passed to target/subtarget 408 std::string FeaturesStr; 409 if (MAttrs.size()) { 410 SubtargetFeatures Features; 411 for (unsigned i = 0; i != MAttrs.size(); ++i) 412 Features.AddFeature(MAttrs[i]); 413 FeaturesStr = Features.getString(); 414 } 415 416 OwningPtr<tool_output_file> Out(GetOutputStream()); 417 if (!Out) 418 return 1; 419 420 formatted_raw_ostream FOS(Out->os()); 421 OwningPtr<MCStreamer> Str; 422 423 OwningPtr<MCInstrInfo> MCII(TheTarget->createMCInstrInfo()); 424 OwningPtr<MCSubtargetInfo> 425 STI(TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 426 427 MCInstPrinter *IP = NULL; 428 if (FileType == OFT_AssemblyFile) { 429 IP = 430 TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI, *MCII, *MRI, *STI); 431 MCCodeEmitter *CE = 0; 432 MCAsmBackend *MAB = 0; 433 if (ShowEncoding) { 434 CE = TheTarget->createMCCodeEmitter(*MCII, *MRI, *STI, Ctx); 435 MAB = TheTarget->createMCAsmBackend(TripleName, MCPU); 436 } 437 bool UseCFI = !DisableCFI; 438 Str.reset(TheTarget->createAsmStreamer(Ctx, FOS, /*asmverbose*/true, 439 /*useLoc*/ true, 440 UseCFI, 441 /*useDwarfDirectory*/ true, 442 IP, CE, MAB, ShowInst)); 443 444 } else if (FileType == OFT_Null) { 445 Str.reset(createNullStreamer(Ctx)); 446 } else { 447 assert(FileType == OFT_ObjectFile && "Invalid file type!"); 448 MCCodeEmitter *CE = TheTarget->createMCCodeEmitter(*MCII, *MRI, *STI, Ctx); 449 MCAsmBackend *MAB = TheTarget->createMCAsmBackend(TripleName, MCPU); 450 Str.reset(TheTarget->createMCObjectStreamer(TripleName, Ctx, *MAB, 451 FOS, CE, RelaxAll, 452 NoExecStack)); 453 } 454 455 int Res = 1; 456 bool disassemble = false; 457 switch (Action) { 458 case AC_AsLex: 459 Res = AsLexInput(SrcMgr, *MAI, Out.get()); 460 break; 461 case AC_Assemble: 462 Res = AssembleInput(ProgName, TheTarget, SrcMgr, Ctx, *Str, *MAI, *STI); 463 break; 464 case AC_MDisassemble: 465 assert(IP && "Expected assembly output"); 466 IP->setUseMarkup(1); 467 disassemble = true; 468 break; 469 case AC_HDisassemble: 470 assert(IP && "Expected assembly output"); 471 IP->setPrintImmHex(1); 472 disassemble = true; 473 break; 474 case AC_Disassemble: 475 disassemble = true; 476 break; 477 } 478 if (disassemble) 479 Res = Disassembler::disassemble(*TheTarget, TripleName, *STI, *Str, 480 *Buffer, SrcMgr, Out->os()); 481 482 // Keep output if no errors. 483 if (Res == 0) Out->keep(); 484 return Res; 485 } 486