llvm-mc.cpp revision e76a33b9567d78a5744dc52fcec3a6056d6fb576
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/MCContext.h" 18#include "llvm/MC/MCCodeEmitter.h" 19#include "llvm/MC/MCInstPrinter.h" 20#include "llvm/MC/MCInstrInfo.h" 21#include "llvm/MC/MCObjectFileInfo.h" 22#include "llvm/MC/MCRegisterInfo.h" 23#include "llvm/MC/MCSectionMachO.h" 24#include "llvm/MC/MCStreamer.h" 25#include "llvm/MC/MCSubtargetInfo.h" 26#include "llvm/MC/SubtargetFeature.h" 27#include "llvm/Target/TargetAsmBackend.h" 28#include "llvm/Target/TargetAsmParser.h" 29#include "llvm/Target/TargetData.h" 30#include "llvm/Target/TargetRegistry.h" 31#include "llvm/Target/TargetAsmInfo.h" // FIXME. 32#include "llvm/Target/TargetLowering.h" // FIXME. 33#include "llvm/Target/TargetLoweringObjectFile.h" // FIXME. 34#include "llvm/Target/TargetMachine.h" // FIXME. 35#include "llvm/Target/TargetSelect.h" 36#include "llvm/ADT/OwningPtr.h" 37#include "llvm/Support/CommandLine.h" 38#include "llvm/Support/FileUtilities.h" 39#include "llvm/Support/FormattedStream.h" 40#include "llvm/Support/ManagedStatic.h" 41#include "llvm/Support/MemoryBuffer.h" 42#include "llvm/Support/PrettyStackTrace.h" 43#include "llvm/Support/SourceMgr.h" 44#include "llvm/Support/ToolOutputFile.h" 45#include "llvm/Support/Host.h" 46#include "llvm/Support/Signals.h" 47#include "llvm/Support/system_error.h" 48#include "Disassembler.h" 49using namespace llvm; 50 51static cl::opt<std::string> 52InputFilename(cl::Positional, cl::desc("<input file>"), cl::init("-")); 53 54static cl::opt<std::string> 55OutputFilename("o", cl::desc("Output filename"), 56 cl::value_desc("filename")); 57 58static cl::opt<bool> 59ShowEncoding("show-encoding", cl::desc("Show instruction encodings")); 60 61static cl::opt<bool> 62ShowInst("show-inst", cl::desc("Show internal instruction representation")); 63 64static cl::opt<bool> 65ShowInstOperands("show-inst-operands", 66 cl::desc("Show instructions operands as parsed")); 67 68static cl::opt<unsigned> 69OutputAsmVariant("output-asm-variant", 70 cl::desc("Syntax variant to use for output printing")); 71 72static cl::opt<bool> 73RelaxAll("mc-relax-all", cl::desc("Relax all fixups")); 74 75static cl::opt<bool> 76NoExecStack("mc-no-exec-stack", cl::desc("File doesn't need an exec stack")); 77 78static cl::opt<bool> 79EnableLogging("enable-api-logging", cl::desc("Enable MC API logging")); 80 81enum OutputFileType { 82 OFT_Null, 83 OFT_AssemblyFile, 84 OFT_ObjectFile 85}; 86static cl::opt<OutputFileType> 87FileType("filetype", cl::init(OFT_AssemblyFile), 88 cl::desc("Choose an output file type:"), 89 cl::values( 90 clEnumValN(OFT_AssemblyFile, "asm", 91 "Emit an assembly ('.s') file"), 92 clEnumValN(OFT_Null, "null", 93 "Don't emit anything (for timing purposes)"), 94 clEnumValN(OFT_ObjectFile, "obj", 95 "Emit a native object ('.o') file"), 96 clEnumValEnd)); 97 98static cl::list<std::string> 99IncludeDirs("I", cl::desc("Directory of include files"), 100 cl::value_desc("directory"), cl::Prefix); 101 102static cl::opt<std::string> 103ArchName("arch", cl::desc("Target arch to assemble for, " 104 "see -version for available targets")); 105 106static cl::opt<std::string> 107TripleName("triple", cl::desc("Target triple to assemble for, " 108 "see -version for available targets")); 109 110static cl::opt<std::string> 111MCPU("mcpu", 112 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 113 cl::value_desc("cpu-name"), 114 cl::init("")); 115 116static cl::opt<Reloc::Model> 117RelocModel("relocation-model", 118 cl::desc("Choose relocation model"), 119 cl::init(Reloc::Default), 120 cl::values( 121 clEnumValN(Reloc::Default, "default", 122 "Target default relocation model"), 123 clEnumValN(Reloc::Static, "static", 124 "Non-relocatable code"), 125 clEnumValN(Reloc::PIC_, "pic", 126 "Fully relocatable, position independent code"), 127 clEnumValN(Reloc::DynamicNoPIC, "dynamic-no-pic", 128 "Relocatable external references, non-relocatable code"), 129 clEnumValEnd)); 130 131static cl::opt<bool> 132NoInitialTextSection("n", cl::desc("Don't assume assembly file starts " 133 "in the text section")); 134 135static cl::opt<bool> 136SaveTempLabels("L", cl::desc("Don't discard temporary labels")); 137 138enum ActionType { 139 AC_AsLex, 140 AC_Assemble, 141 AC_Disassemble, 142 AC_EDisassemble 143}; 144 145static cl::opt<ActionType> 146Action(cl::desc("Action to perform:"), 147 cl::init(AC_Assemble), 148 cl::values(clEnumValN(AC_AsLex, "as-lex", 149 "Lex tokens from a .s file"), 150 clEnumValN(AC_Assemble, "assemble", 151 "Assemble a .s file (default)"), 152 clEnumValN(AC_Disassemble, "disassemble", 153 "Disassemble strings of hex bytes"), 154 clEnumValN(AC_EDisassemble, "edis", 155 "Enhanced disassembly of strings of hex bytes"), 156 clEnumValEnd)); 157 158static const Target *GetTarget(const char *ProgName) { 159 // Figure out the target triple. 160 if (TripleName.empty()) 161 TripleName = sys::getHostTriple(); 162 if (!ArchName.empty()) { 163 llvm::Triple TT(TripleName); 164 TT.setArchName(ArchName); 165 TripleName = TT.str(); 166 } 167 168 // Get the target specific parser. 169 std::string Error; 170 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 171 if (TheTarget) 172 return TheTarget; 173 174 errs() << ProgName << ": error: unable to get target for '" << TripleName 175 << "', see --version and --triple.\n"; 176 return 0; 177} 178 179static tool_output_file *GetOutputStream() { 180 if (OutputFilename == "") 181 OutputFilename = "-"; 182 183 std::string Err; 184 tool_output_file *Out = new tool_output_file(OutputFilename.c_str(), Err, 185 raw_fd_ostream::F_Binary); 186 if (!Err.empty()) { 187 errs() << Err << '\n'; 188 delete Out; 189 return 0; 190 } 191 192 return Out; 193} 194 195static int AsLexInput(const char *ProgName) { 196 OwningPtr<MemoryBuffer> BufferPtr; 197 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) { 198 errs() << ProgName << ": " << ec.message() << '\n'; 199 return 1; 200 } 201 MemoryBuffer *Buffer = BufferPtr.take(); 202 203 SourceMgr SrcMgr; 204 205 // Tell SrcMgr about this buffer, which is what TGParser will pick up. 206 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc()); 207 208 // Record the location of the include directories so that the lexer can find 209 // it later. 210 SrcMgr.setIncludeDirs(IncludeDirs); 211 212 const Target *TheTarget = GetTarget(ProgName); 213 if (!TheTarget) 214 return 1; 215 216 llvm::OwningPtr<MCAsmInfo> MAI(TheTarget->createMCAsmInfo(TripleName)); 217 assert(MAI && "Unable to create target asm info!"); 218 219 AsmLexer Lexer(*MAI); 220 Lexer.setBuffer(SrcMgr.getMemoryBuffer(0)); 221 222 OwningPtr<tool_output_file> Out(GetOutputStream()); 223 if (!Out) 224 return 1; 225 226 bool Error = false; 227 while (Lexer.Lex().isNot(AsmToken::Eof)) { 228 AsmToken Tok = Lexer.getTok(); 229 230 switch (Tok.getKind()) { 231 default: 232 SrcMgr.PrintMessage(Lexer.getLoc(), "unknown token", "warning"); 233 Error = true; 234 break; 235 case AsmToken::Error: 236 Error = true; // error already printed. 237 break; 238 case AsmToken::Identifier: 239 Out->os() << "identifier: " << Lexer.getTok().getString(); 240 break; 241 case AsmToken::Integer: 242 Out->os() << "int: " << Lexer.getTok().getString(); 243 break; 244 case AsmToken::Real: 245 Out->os() << "real: " << Lexer.getTok().getString(); 246 break; 247 case AsmToken::Register: 248 Out->os() << "register: " << Lexer.getTok().getRegVal(); 249 break; 250 case AsmToken::String: 251 Out->os() << "string: " << Lexer.getTok().getString(); 252 break; 253 254 case AsmToken::Amp: Out->os() << "Amp"; break; 255 case AsmToken::AmpAmp: Out->os() << "AmpAmp"; break; 256 case AsmToken::At: Out->os() << "At"; break; 257 case AsmToken::Caret: Out->os() << "Caret"; break; 258 case AsmToken::Colon: Out->os() << "Colon"; break; 259 case AsmToken::Comma: Out->os() << "Comma"; break; 260 case AsmToken::Dollar: Out->os() << "Dollar"; break; 261 case AsmToken::Dot: Out->os() << "Dot"; break; 262 case AsmToken::EndOfStatement: Out->os() << "EndOfStatement"; break; 263 case AsmToken::Eof: Out->os() << "Eof"; break; 264 case AsmToken::Equal: Out->os() << "Equal"; break; 265 case AsmToken::EqualEqual: Out->os() << "EqualEqual"; break; 266 case AsmToken::Exclaim: Out->os() << "Exclaim"; break; 267 case AsmToken::ExclaimEqual: Out->os() << "ExclaimEqual"; break; 268 case AsmToken::Greater: Out->os() << "Greater"; break; 269 case AsmToken::GreaterEqual: Out->os() << "GreaterEqual"; break; 270 case AsmToken::GreaterGreater: Out->os() << "GreaterGreater"; break; 271 case AsmToken::Hash: Out->os() << "Hash"; break; 272 case AsmToken::LBrac: Out->os() << "LBrac"; break; 273 case AsmToken::LCurly: Out->os() << "LCurly"; break; 274 case AsmToken::LParen: Out->os() << "LParen"; break; 275 case AsmToken::Less: Out->os() << "Less"; break; 276 case AsmToken::LessEqual: Out->os() << "LessEqual"; break; 277 case AsmToken::LessGreater: Out->os() << "LessGreater"; break; 278 case AsmToken::LessLess: Out->os() << "LessLess"; break; 279 case AsmToken::Minus: Out->os() << "Minus"; break; 280 case AsmToken::Percent: Out->os() << "Percent"; break; 281 case AsmToken::Pipe: Out->os() << "Pipe"; break; 282 case AsmToken::PipePipe: Out->os() << "PipePipe"; break; 283 case AsmToken::Plus: Out->os() << "Plus"; break; 284 case AsmToken::RBrac: Out->os() << "RBrac"; break; 285 case AsmToken::RCurly: Out->os() << "RCurly"; break; 286 case AsmToken::RParen: Out->os() << "RParen"; break; 287 case AsmToken::Slash: Out->os() << "Slash"; break; 288 case AsmToken::Star: Out->os() << "Star"; break; 289 case AsmToken::Tilde: Out->os() << "Tilde"; break; 290 } 291 292 // Print the token string. 293 Out->os() << " (\""; 294 Out->os().write_escaped(Tok.getString()); 295 Out->os() << "\")\n"; 296 } 297 298 // Keep output if no errors. 299 if (Error == 0) Out->keep(); 300 301 return Error; 302} 303 304static int AssembleInput(const char *ProgName) { 305 const Target *TheTarget = GetTarget(ProgName); 306 if (!TheTarget) 307 return 1; 308 309 OwningPtr<MemoryBuffer> BufferPtr; 310 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) { 311 errs() << ProgName << ": " << ec.message() << '\n'; 312 return 1; 313 } 314 MemoryBuffer *Buffer = BufferPtr.take(); 315 316 SourceMgr SrcMgr; 317 318 // Tell SrcMgr about this buffer, which is what the parser will pick up. 319 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc()); 320 321 // Record the location of the include directories so that the lexer can find 322 // it later. 323 SrcMgr.setIncludeDirs(IncludeDirs); 324 325 326 llvm::OwningPtr<MCAsmInfo> MAI(TheTarget->createMCAsmInfo(TripleName)); 327 assert(MAI && "Unable to create target asm info!"); 328 329 llvm::OwningPtr<MCRegisterInfo> MRI(TheTarget->createMCRegInfo(TripleName)); 330 assert(MRI && "Unable to create target register info!"); 331 332 // Package up features to be passed to target/subtarget 333 std::string FeaturesStr; 334 335 // FIXME: We shouldn't need to do this (and link in codegen). 336 // When we split this out, we should do it in a way that makes 337 // it straightforward to switch subtargets on the fly (.e.g, 338 // the .cpu and .code16 directives). 339 OwningPtr<TargetMachine> TM(TheTarget->createTargetMachine(TripleName, 340 MCPU, 341 FeaturesStr, 342 RelocModel)); 343 344 if (!TM) { 345 errs() << ProgName << ": error: could not create target for triple '" 346 << TripleName << "'.\n"; 347 return 1; 348 } 349 350 const TargetAsmInfo *tai = new TargetAsmInfo(*TM); 351 // FIXME: This is not pretty. MCContext has a ptr to MCObjectFileInfo and 352 // MCObjectFileInfo needs a MCContext reference in order to initialize itself. 353 OwningPtr<MCObjectFileInfo> MOFI(new MCObjectFileInfo()); 354 MCContext Ctx(*MAI, *MRI, MOFI.get(), tai); 355 MOFI->InitMCObjectFileInfo(TripleName, RelocModel, Ctx); 356 357 if (SaveTempLabels) 358 Ctx.setAllowTemporaryLabels(false); 359 360 OwningPtr<tool_output_file> Out(GetOutputStream()); 361 if (!Out) 362 return 1; 363 364 formatted_raw_ostream FOS(Out->os()); 365 OwningPtr<MCStreamer> Str; 366 367 const TargetLoweringObjectFile &TLOF = 368 TM->getTargetLowering()->getObjFileLowering(); 369 const_cast<TargetLoweringObjectFile&>(TLOF).Initialize(Ctx, *TM); 370 371 OwningPtr<MCInstrInfo> MCII(TheTarget->createMCInstrInfo()); 372 OwningPtr<MCSubtargetInfo> 373 STI(TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 374 375 // FIXME: There is a bit of code duplication with addPassesToEmitFile. 376 if (FileType == OFT_AssemblyFile) { 377 MCInstPrinter *IP = 378 TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI); 379 MCCodeEmitter *CE = 0; 380 TargetAsmBackend *TAB = 0; 381 if (ShowEncoding) { 382 CE = TheTarget->createCodeEmitter(*MCII, *STI, Ctx); 383 TAB = TheTarget->createAsmBackend(TripleName); 384 } 385 Str.reset(TheTarget->createAsmStreamer(Ctx, FOS, /*asmverbose*/true, 386 /*useLoc*/ true, 387 /*useCFI*/ true, IP, CE, TAB, 388 ShowInst)); 389 } else if (FileType == OFT_Null) { 390 Str.reset(createNullStreamer(Ctx)); 391 } else { 392 assert(FileType == OFT_ObjectFile && "Invalid file type!"); 393 MCCodeEmitter *CE = TheTarget->createCodeEmitter(*MCII, *STI, Ctx); 394 TargetAsmBackend *TAB = TheTarget->createAsmBackend(TripleName); 395 Str.reset(TheTarget->createObjectStreamer(TripleName, Ctx, *TAB, 396 FOS, CE, RelaxAll, 397 NoExecStack)); 398 } 399 400 if (EnableLogging) { 401 Str.reset(createLoggingStreamer(Str.take(), errs())); 402 } 403 404 OwningPtr<MCAsmParser> Parser(createMCAsmParser(*TheTarget, SrcMgr, Ctx, 405 *Str.get(), *MAI)); 406 OwningPtr<TargetAsmParser> TAP(TheTarget->createAsmParser(*STI, *Parser)); 407 if (!TAP) { 408 errs() << ProgName 409 << ": error: this target does not support assembly parsing.\n"; 410 return 1; 411 } 412 413 Parser->setShowParsedOperands(ShowInstOperands); 414 Parser->setTargetParser(*TAP.get()); 415 416 int Res = Parser->Run(NoInitialTextSection); 417 418 // Keep output if no errors. 419 if (Res == 0) Out->keep(); 420 421 return Res; 422} 423 424static int DisassembleInput(const char *ProgName, bool Enhanced) { 425 const Target *TheTarget = GetTarget(ProgName); 426 if (!TheTarget) 427 return 0; 428 429 OwningPtr<MemoryBuffer> Buffer; 430 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, Buffer)) { 431 errs() << ProgName << ": " << ec.message() << '\n'; 432 return 1; 433 } 434 435 OwningPtr<tool_output_file> Out(GetOutputStream()); 436 if (!Out) 437 return 1; 438 439 int Res; 440 if (Enhanced) { 441 Res = 442 Disassembler::disassembleEnhanced(TripleName, *Buffer.take(), Out->os()); 443 } else { 444 Res = Disassembler::disassemble(*TheTarget, TripleName, 445 *Buffer.take(), Out->os()); 446 } 447 448 // Keep output if no errors. 449 if (Res == 0) Out->keep(); 450 451 return Res; 452} 453 454 455int main(int argc, char **argv) { 456 // Print a stack trace if we signal out. 457 sys::PrintStackTraceOnErrorSignal(); 458 PrettyStackTraceProgram X(argc, argv); 459 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 460 461 // Initialize targets and assembly printers/parsers. 462 llvm::InitializeAllTargetInfos(); 463 // FIXME: We shouldn't need to initialize the Target(Machine)s. 464 llvm::InitializeAllTargets(); 465 llvm::InitializeAllMCAsmInfos(); 466 llvm::InitializeAllMCCodeGenInfos(); 467 llvm::InitializeAllMCInstrInfos(); 468 llvm::InitializeAllMCRegisterInfos(); 469 llvm::InitializeAllMCSubtargetInfos(); 470 llvm::InitializeAllAsmPrinters(); 471 llvm::InitializeAllAsmParsers(); 472 llvm::InitializeAllDisassemblers(); 473 474 cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n"); 475 TripleName = Triple::normalize(TripleName); 476 477 switch (Action) { 478 default: 479 case AC_AsLex: 480 return AsLexInput(argv[0]); 481 case AC_Assemble: 482 return AssembleInput(argv[0]); 483 case AC_Disassemble: 484 return DisassembleInput(argv[0], false); 485 case AC_EDisassemble: 486 return DisassembleInput(argv[0], true); 487 } 488 489 return 0; 490} 491 492