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/AsmLexer.h" 16#include "llvm/MC/MCParser/MCAsmLexer.h" 17#include "llvm/MC/MCAsmBackend.h" 18#include "llvm/MC/MCContext.h" 19#include "llvm/MC/MCCodeEmitter.h" 20#include "llvm/MC/MCInstPrinter.h" 21#include "llvm/MC/MCInstrInfo.h" 22#include "llvm/MC/MCObjectFileInfo.h" 23#include "llvm/MC/MCRegisterInfo.h" 24#include "llvm/MC/MCSectionMachO.h" 25#include "llvm/MC/MCStreamer.h" 26#include "llvm/MC/MCSubtargetInfo.h" 27#include "llvm/MC/MCTargetAsmParser.h" 28#include "llvm/MC/SubtargetFeature.h" 29#include "llvm/ADT/OwningPtr.h" 30#include "llvm/Support/CommandLine.h" 31#include "llvm/Support/FileUtilities.h" 32#include "llvm/Support/FormattedStream.h" 33#include "llvm/Support/ManagedStatic.h" 34#include "llvm/Support/MemoryBuffer.h" 35#include "llvm/Support/PrettyStackTrace.h" 36#include "llvm/Support/SourceMgr.h" 37#include "llvm/Support/ToolOutputFile.h" 38#include "llvm/Support/Host.h" 39#include "llvm/Support/Signals.h" 40#include "llvm/Support/TargetRegistry.h" 41#include "llvm/Support/TargetSelect.h" 42#include "llvm/Support/system_error.h" 43#include "Disassembler.h" 44using namespace llvm; 45 46static cl::opt<std::string> 47InputFilename(cl::Positional, cl::desc("<input file>"), cl::init("-")); 48 49static cl::opt<std::string> 50OutputFilename("o", cl::desc("Output filename"), 51 cl::value_desc("filename")); 52 53static cl::opt<bool> 54ShowEncoding("show-encoding", cl::desc("Show instruction encodings")); 55 56static cl::opt<bool> 57ShowInst("show-inst", cl::desc("Show internal instruction representation")); 58 59static cl::opt<bool> 60ShowInstOperands("show-inst-operands", 61 cl::desc("Show instructions operands as parsed")); 62 63static cl::opt<unsigned> 64OutputAsmVariant("output-asm-variant", 65 cl::desc("Syntax variant to use for output printing")); 66 67static cl::opt<bool> 68RelaxAll("mc-relax-all", cl::desc("Relax all fixups")); 69 70static cl::opt<bool> 71NoExecStack("mc-no-exec-stack", cl::desc("File doesn't need an exec stack")); 72 73static cl::opt<bool> 74EnableLogging("enable-api-logging", cl::desc("Enable MC API logging")); 75 76enum OutputFileType { 77 OFT_Null, 78 OFT_AssemblyFile, 79 OFT_ObjectFile 80}; 81static cl::opt<OutputFileType> 82FileType("filetype", cl::init(OFT_AssemblyFile), 83 cl::desc("Choose an output file type:"), 84 cl::values( 85 clEnumValN(OFT_AssemblyFile, "asm", 86 "Emit an assembly ('.s') file"), 87 clEnumValN(OFT_Null, "null", 88 "Don't emit anything (for timing purposes)"), 89 clEnumValN(OFT_ObjectFile, "obj", 90 "Emit a native object ('.o') file"), 91 clEnumValEnd)); 92 93static cl::list<std::string> 94IncludeDirs("I", cl::desc("Directory of include files"), 95 cl::value_desc("directory"), cl::Prefix); 96 97static cl::opt<std::string> 98ArchName("arch", cl::desc("Target arch to assemble for, " 99 "see -version for available targets")); 100 101static cl::opt<std::string> 102TripleName("triple", cl::desc("Target triple to assemble for, " 103 "see -version for available targets")); 104 105static cl::opt<std::string> 106MCPU("mcpu", 107 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 108 cl::value_desc("cpu-name"), 109 cl::init("")); 110 111static cl::list<std::string> 112MAttrs("mattr", 113 cl::CommaSeparated, 114 cl::desc("Target specific attributes (-mattr=help for details)"), 115 cl::value_desc("a1,+a2,-a3,...")); 116 117static cl::opt<Reloc::Model> 118RelocModel("relocation-model", 119 cl::desc("Choose relocation model"), 120 cl::init(Reloc::Default), 121 cl::values( 122 clEnumValN(Reloc::Default, "default", 123 "Target default relocation model"), 124 clEnumValN(Reloc::Static, "static", 125 "Non-relocatable code"), 126 clEnumValN(Reloc::PIC_, "pic", 127 "Fully relocatable, position independent code"), 128 clEnumValN(Reloc::DynamicNoPIC, "dynamic-no-pic", 129 "Relocatable external references, non-relocatable code"), 130 clEnumValEnd)); 131 132static cl::opt<llvm::CodeModel::Model> 133CMModel("code-model", 134 cl::desc("Choose code model"), 135 cl::init(CodeModel::Default), 136 cl::values(clEnumValN(CodeModel::Default, "default", 137 "Target default code model"), 138 clEnumValN(CodeModel::Small, "small", 139 "Small code model"), 140 clEnumValN(CodeModel::Kernel, "kernel", 141 "Kernel code model"), 142 clEnumValN(CodeModel::Medium, "medium", 143 "Medium code model"), 144 clEnumValN(CodeModel::Large, "large", 145 "Large code model"), 146 clEnumValEnd)); 147 148static cl::opt<bool> 149NoInitialTextSection("n", cl::desc("Don't assume assembly file starts " 150 "in the text section")); 151 152static cl::opt<bool> 153SaveTempLabels("L", cl::desc("Don't discard temporary labels")); 154 155enum ActionType { 156 AC_AsLex, 157 AC_Assemble, 158 AC_Disassemble, 159 AC_EDisassemble 160}; 161 162static cl::opt<ActionType> 163Action(cl::desc("Action to perform:"), 164 cl::init(AC_Assemble), 165 cl::values(clEnumValN(AC_AsLex, "as-lex", 166 "Lex tokens from a .s file"), 167 clEnumValN(AC_Assemble, "assemble", 168 "Assemble a .s file (default)"), 169 clEnumValN(AC_Disassemble, "disassemble", 170 "Disassemble strings of hex bytes"), 171 clEnumValN(AC_EDisassemble, "edis", 172 "Enhanced disassembly of strings of hex bytes"), 173 clEnumValEnd)); 174 175static const Target *GetTarget(const char *ProgName) { 176 // Figure out the target triple. 177 if (TripleName.empty()) 178 TripleName = sys::getHostTriple(); 179 Triple TheTriple(Triple::normalize(TripleName)); 180 181 const Target *TheTarget = 0; 182 if (!ArchName.empty()) { 183 for (TargetRegistry::iterator it = TargetRegistry::begin(), 184 ie = TargetRegistry::end(); it != ie; ++it) { 185 if (ArchName == it->getName()) { 186 TheTarget = &*it; 187 break; 188 } 189 } 190 191 if (!TheTarget) { 192 errs() << ProgName << ": error: invalid target '" << ArchName << "'.\n"; 193 return 0; 194 } 195 196 // Adjust the triple to match (if known), otherwise stick with the 197 // module/host triple. 198 Triple::ArchType Type = Triple::getArchTypeForLLVMName(ArchName); 199 if (Type != Triple::UnknownArch) 200 TheTriple.setArch(Type); 201 } else { 202 // Get the target specific parser. 203 std::string Error; 204 TheTarget = TargetRegistry::lookupTarget(TheTriple.getTriple(), Error); 205 if (TheTarget == 0) { 206 errs() << ProgName << ": error: unable to get target for '" 207 << TheTriple.getTriple() 208 << "', see --version and --triple.\n"; 209 return 0; 210 } 211 } 212 213 TripleName = TheTriple.getTriple(); 214 return TheTarget; 215} 216 217static tool_output_file *GetOutputStream() { 218 if (OutputFilename == "") 219 OutputFilename = "-"; 220 221 std::string Err; 222 tool_output_file *Out = new tool_output_file(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 Out; 231} 232 233static int AsLexInput(const char *ProgName) { 234 OwningPtr<MemoryBuffer> BufferPtr; 235 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) { 236 errs() << ProgName << ": " << ec.message() << '\n'; 237 return 1; 238 } 239 MemoryBuffer *Buffer = BufferPtr.take(); 240 241 SourceMgr SrcMgr; 242 243 // Tell SrcMgr about this buffer, which is what TGParser will pick up. 244 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc()); 245 246 // Record the location of the include directories so that the lexer can find 247 // it later. 248 SrcMgr.setIncludeDirs(IncludeDirs); 249 250 const Target *TheTarget = GetTarget(ProgName); 251 if (!TheTarget) 252 return 1; 253 254 llvm::OwningPtr<MCAsmInfo> MAI(TheTarget->createMCAsmInfo(TripleName)); 255 assert(MAI && "Unable to create target asm info!"); 256 257 AsmLexer Lexer(*MAI); 258 Lexer.setBuffer(SrcMgr.getMemoryBuffer(0)); 259 260 OwningPtr<tool_output_file> Out(GetOutputStream()); 261 if (!Out) 262 return 1; 263 264 bool Error = false; 265 while (Lexer.Lex().isNot(AsmToken::Eof)) { 266 AsmToken Tok = Lexer.getTok(); 267 268 switch (Tok.getKind()) { 269 default: 270 SrcMgr.PrintMessage(Lexer.getLoc(), "unknown token", "warning"); 271 Error = true; 272 break; 273 case AsmToken::Error: 274 Error = true; // error already printed. 275 break; 276 case AsmToken::Identifier: 277 Out->os() << "identifier: " << Lexer.getTok().getString(); 278 break; 279 case AsmToken::Integer: 280 Out->os() << "int: " << Lexer.getTok().getString(); 281 break; 282 case AsmToken::Real: 283 Out->os() << "real: " << Lexer.getTok().getString(); 284 break; 285 case AsmToken::Register: 286 Out->os() << "register: " << Lexer.getTok().getRegVal(); 287 break; 288 case AsmToken::String: 289 Out->os() << "string: " << Lexer.getTok().getString(); 290 break; 291 292 case AsmToken::Amp: Out->os() << "Amp"; break; 293 case AsmToken::AmpAmp: Out->os() << "AmpAmp"; break; 294 case AsmToken::At: Out->os() << "At"; break; 295 case AsmToken::Caret: Out->os() << "Caret"; break; 296 case AsmToken::Colon: Out->os() << "Colon"; break; 297 case AsmToken::Comma: Out->os() << "Comma"; break; 298 case AsmToken::Dollar: Out->os() << "Dollar"; break; 299 case AsmToken::Dot: Out->os() << "Dot"; break; 300 case AsmToken::EndOfStatement: Out->os() << "EndOfStatement"; break; 301 case AsmToken::Eof: Out->os() << "Eof"; break; 302 case AsmToken::Equal: Out->os() << "Equal"; break; 303 case AsmToken::EqualEqual: Out->os() << "EqualEqual"; break; 304 case AsmToken::Exclaim: Out->os() << "Exclaim"; break; 305 case AsmToken::ExclaimEqual: Out->os() << "ExclaimEqual"; break; 306 case AsmToken::Greater: Out->os() << "Greater"; break; 307 case AsmToken::GreaterEqual: Out->os() << "GreaterEqual"; break; 308 case AsmToken::GreaterGreater: Out->os() << "GreaterGreater"; break; 309 case AsmToken::Hash: Out->os() << "Hash"; break; 310 case AsmToken::LBrac: Out->os() << "LBrac"; break; 311 case AsmToken::LCurly: Out->os() << "LCurly"; break; 312 case AsmToken::LParen: Out->os() << "LParen"; break; 313 case AsmToken::Less: Out->os() << "Less"; break; 314 case AsmToken::LessEqual: Out->os() << "LessEqual"; break; 315 case AsmToken::LessGreater: Out->os() << "LessGreater"; break; 316 case AsmToken::LessLess: Out->os() << "LessLess"; break; 317 case AsmToken::Minus: Out->os() << "Minus"; break; 318 case AsmToken::Percent: Out->os() << "Percent"; break; 319 case AsmToken::Pipe: Out->os() << "Pipe"; break; 320 case AsmToken::PipePipe: Out->os() << "PipePipe"; break; 321 case AsmToken::Plus: Out->os() << "Plus"; break; 322 case AsmToken::RBrac: Out->os() << "RBrac"; break; 323 case AsmToken::RCurly: Out->os() << "RCurly"; break; 324 case AsmToken::RParen: Out->os() << "RParen"; break; 325 case AsmToken::Slash: Out->os() << "Slash"; break; 326 case AsmToken::Star: Out->os() << "Star"; break; 327 case AsmToken::Tilde: Out->os() << "Tilde"; break; 328 } 329 330 // Print the token string. 331 Out->os() << " (\""; 332 Out->os().write_escaped(Tok.getString()); 333 Out->os() << "\")\n"; 334 } 335 336 // Keep output if no errors. 337 if (Error == 0) Out->keep(); 338 339 return Error; 340} 341 342static int AssembleInput(const char *ProgName) { 343 const Target *TheTarget = GetTarget(ProgName); 344 if (!TheTarget) 345 return 1; 346 347 OwningPtr<MemoryBuffer> BufferPtr; 348 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) { 349 errs() << ProgName << ": " << ec.message() << '\n'; 350 return 1; 351 } 352 MemoryBuffer *Buffer = BufferPtr.take(); 353 354 SourceMgr SrcMgr; 355 356 // Tell SrcMgr about this buffer, which is what the parser will pick up. 357 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc()); 358 359 // Record the location of the include directories so that the lexer can find 360 // it later. 361 SrcMgr.setIncludeDirs(IncludeDirs); 362 363 364 llvm::OwningPtr<MCAsmInfo> MAI(TheTarget->createMCAsmInfo(TripleName)); 365 assert(MAI && "Unable to create target asm info!"); 366 367 llvm::OwningPtr<MCRegisterInfo> MRI(TheTarget->createMCRegInfo(TripleName)); 368 assert(MRI && "Unable to create target register info!"); 369 370 // FIXME: This is not pretty. MCContext has a ptr to MCObjectFileInfo and 371 // MCObjectFileInfo needs a MCContext reference in order to initialize itself. 372 OwningPtr<MCObjectFileInfo> MOFI(new MCObjectFileInfo()); 373 MCContext Ctx(*MAI, *MRI, MOFI.get()); 374 MOFI->InitMCObjectFileInfo(TripleName, RelocModel, CMModel, Ctx); 375 376 if (SaveTempLabels) 377 Ctx.setAllowTemporaryLabels(false); 378 379 // Package up features to be passed to target/subtarget 380 std::string FeaturesStr; 381 if (MAttrs.size()) { 382 SubtargetFeatures Features; 383 for (unsigned i = 0; i != MAttrs.size(); ++i) 384 Features.AddFeature(MAttrs[i]); 385 FeaturesStr = Features.getString(); 386 } 387 388 OwningPtr<tool_output_file> Out(GetOutputStream()); 389 if (!Out) 390 return 1; 391 392 formatted_raw_ostream FOS(Out->os()); 393 OwningPtr<MCStreamer> Str; 394 395 OwningPtr<MCInstrInfo> MCII(TheTarget->createMCInstrInfo()); 396 OwningPtr<MCSubtargetInfo> 397 STI(TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 398 399 // FIXME: There is a bit of code duplication with addPassesToEmitFile. 400 if (FileType == OFT_AssemblyFile) { 401 MCInstPrinter *IP = 402 TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI, *STI); 403 MCCodeEmitter *CE = 0; 404 MCAsmBackend *MAB = 0; 405 if (ShowEncoding) { 406 CE = TheTarget->createMCCodeEmitter(*MCII, *STI, Ctx); 407 MAB = TheTarget->createMCAsmBackend(TripleName); 408 } 409 Str.reset(TheTarget->createAsmStreamer(Ctx, FOS, /*asmverbose*/true, 410 /*useLoc*/ true, 411 /*useCFI*/ true, IP, CE, MAB, 412 ShowInst)); 413 } else if (FileType == OFT_Null) { 414 Str.reset(createNullStreamer(Ctx)); 415 } else { 416 assert(FileType == OFT_ObjectFile && "Invalid file type!"); 417 MCCodeEmitter *CE = TheTarget->createMCCodeEmitter(*MCII, *STI, Ctx); 418 MCAsmBackend *MAB = TheTarget->createMCAsmBackend(TripleName); 419 Str.reset(TheTarget->createMCObjectStreamer(TripleName, Ctx, *MAB, 420 FOS, CE, RelaxAll, 421 NoExecStack)); 422 } 423 424 if (EnableLogging) { 425 Str.reset(createLoggingStreamer(Str.take(), errs())); 426 } 427 428 OwningPtr<MCAsmParser> Parser(createMCAsmParser(SrcMgr, Ctx, 429 *Str.get(), *MAI)); 430 OwningPtr<MCTargetAsmParser> TAP(TheTarget->createMCAsmParser(*STI, *Parser)); 431 if (!TAP) { 432 errs() << ProgName 433 << ": error: this target does not support assembly parsing.\n"; 434 return 1; 435 } 436 437 Parser->setShowParsedOperands(ShowInstOperands); 438 Parser->setTargetParser(*TAP.get()); 439 440 int Res = Parser->Run(NoInitialTextSection); 441 442 // Keep output if no errors. 443 if (Res == 0) Out->keep(); 444 445 return Res; 446} 447 448static int DisassembleInput(const char *ProgName, bool Enhanced) { 449 const Target *TheTarget = GetTarget(ProgName); 450 if (!TheTarget) 451 return 0; 452 453 OwningPtr<MemoryBuffer> Buffer; 454 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, Buffer)) { 455 errs() << ProgName << ": " << ec.message() << '\n'; 456 return 1; 457 } 458 459 OwningPtr<tool_output_file> Out(GetOutputStream()); 460 if (!Out) 461 return 1; 462 463 int Res; 464 if (Enhanced) { 465 Res = 466 Disassembler::disassembleEnhanced(TripleName, *Buffer.take(), Out->os()); 467 } else { 468 // Package up features to be passed to target/subtarget 469 std::string FeaturesStr; 470 if (MAttrs.size()) { 471 SubtargetFeatures Features; 472 for (unsigned i = 0; i != MAttrs.size(); ++i) 473 Features.AddFeature(MAttrs[i]); 474 FeaturesStr = Features.getString(); 475 } 476 477 Res = Disassembler::disassemble(*TheTarget, TripleName, MCPU, FeaturesStr, 478 *Buffer.take(), Out->os()); 479 } 480 481 // Keep output if no errors. 482 if (Res == 0) Out->keep(); 483 484 return Res; 485} 486 487 488int main(int argc, char **argv) { 489 // Print a stack trace if we signal out. 490 sys::PrintStackTraceOnErrorSignal(); 491 PrettyStackTraceProgram X(argc, argv); 492 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 493 494 // Initialize targets and assembly printers/parsers. 495 llvm::InitializeAllTargetInfos(); 496 llvm::InitializeAllTargetMCs(); 497 llvm::InitializeAllAsmParsers(); 498 llvm::InitializeAllDisassemblers(); 499 500 cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n"); 501 TripleName = Triple::normalize(TripleName); 502 503 switch (Action) { 504 default: 505 case AC_AsLex: 506 return AsLexInput(argv[0]); 507 case AC_Assemble: 508 return AssembleInput(argv[0]); 509 case AC_Disassemble: 510 return DisassembleInput(argv[0], false); 511 case AC_EDisassemble: 512 return DisassembleInput(argv[0], true); 513 } 514 515 return 0; 516} 517 518