1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===// 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 class implements the parser for assembly files. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "llvm/ADT/APFloat.h" 15 #include "llvm/ADT/SmallString.h" 16 #include "llvm/ADT/StringMap.h" 17 #include "llvm/ADT/Twine.h" 18 #include "llvm/MC/MCAsmInfo.h" 19 #include "llvm/MC/MCContext.h" 20 #include "llvm/MC/MCDwarf.h" 21 #include "llvm/MC/MCExpr.h" 22 #include "llvm/MC/MCParser/AsmCond.h" 23 #include "llvm/MC/MCParser/AsmLexer.h" 24 #include "llvm/MC/MCParser/MCAsmParser.h" 25 #include "llvm/MC/MCParser/MCParsedAsmOperand.h" 26 #include "llvm/MC/MCRegisterInfo.h" 27 #include "llvm/MC/MCSectionMachO.h" 28 #include "llvm/MC/MCStreamer.h" 29 #include "llvm/MC/MCSymbol.h" 30 #include "llvm/MC/MCTargetAsmParser.h" 31 #include "llvm/Support/CommandLine.h" 32 #include "llvm/Support/ErrorHandling.h" 33 #include "llvm/Support/MathExtras.h" 34 #include "llvm/Support/MemoryBuffer.h" 35 #include "llvm/Support/SourceMgr.h" 36 #include "llvm/Support/raw_ostream.h" 37 #include <cctype> 38 #include <vector> 39 using namespace llvm; 40 41 static cl::opt<bool> 42 FatalAssemblerWarnings("fatal-assembler-warnings", 43 cl::desc("Consider warnings as error")); 44 45 namespace { 46 47 /// \brief Helper class for tracking macro definitions. 48 struct Macro { 49 StringRef Name; 50 StringRef Body; 51 std::vector<StringRef> Parameters; 52 53 public: 54 Macro(StringRef N, StringRef B, const std::vector<StringRef> &P) : 55 Name(N), Body(B), Parameters(P) {} 56 }; 57 58 /// \brief Helper class for storing information about an active macro 59 /// instantiation. 60 struct MacroInstantiation { 61 /// The macro being instantiated. 62 const Macro *TheMacro; 63 64 /// The macro instantiation with substitutions. 65 MemoryBuffer *Instantiation; 66 67 /// The location of the instantiation. 68 SMLoc InstantiationLoc; 69 70 /// The location where parsing should resume upon instantiation completion. 71 SMLoc ExitLoc; 72 73 public: 74 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL, 75 MemoryBuffer *I); 76 }; 77 78 /// \brief The concrete assembly parser instance. 79 class AsmParser : public MCAsmParser { 80 friend class GenericAsmParser; 81 82 AsmParser(const AsmParser &); // DO NOT IMPLEMENT 83 void operator=(const AsmParser &); // DO NOT IMPLEMENT 84 private: 85 AsmLexer Lexer; 86 MCContext &Ctx; 87 MCStreamer &Out; 88 const MCAsmInfo &MAI; 89 SourceMgr &SrcMgr; 90 SourceMgr::DiagHandlerTy SavedDiagHandler; 91 void *SavedDiagContext; 92 MCAsmParserExtension *GenericParser; 93 MCAsmParserExtension *PlatformParser; 94 95 /// This is the current buffer index we're lexing from as managed by the 96 /// SourceMgr object. 97 int CurBuffer; 98 99 AsmCond TheCondState; 100 std::vector<AsmCond> TheCondStack; 101 102 /// DirectiveMap - This is a table handlers for directives. Each handler is 103 /// invoked after the directive identifier is read and is responsible for 104 /// parsing and validating the rest of the directive. The handler is passed 105 /// in the directive name and the location of the directive keyword. 106 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap; 107 108 /// MacroMap - Map of currently defined macros. 109 StringMap<Macro*> MacroMap; 110 111 /// ActiveMacros - Stack of active macro instantiations. 112 std::vector<MacroInstantiation*> ActiveMacros; 113 114 /// Boolean tracking whether macro substitution is enabled. 115 unsigned MacrosEnabled : 1; 116 117 /// Flag tracking whether any errors have been encountered. 118 unsigned HadError : 1; 119 120 /// The values from the last parsed cpp hash file line comment if any. 121 StringRef CppHashFilename; 122 int64_t CppHashLineNumber; 123 SMLoc CppHashLoc; 124 125 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 126 unsigned AssemblerDialect; 127 128 public: 129 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 130 const MCAsmInfo &MAI); 131 ~AsmParser(); 132 133 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false); 134 135 void AddDirectiveHandler(MCAsmParserExtension *Object, 136 StringRef Directive, 137 DirectiveHandler Handler) { 138 DirectiveMap[Directive] = std::make_pair(Object, Handler); 139 } 140 141 public: 142 /// @name MCAsmParser Interface 143 /// { 144 145 virtual SourceMgr &getSourceManager() { return SrcMgr; } 146 virtual MCAsmLexer &getLexer() { return Lexer; } 147 virtual MCContext &getContext() { return Ctx; } 148 virtual MCStreamer &getStreamer() { return Out; } 149 virtual unsigned getAssemblerDialect() { 150 if (AssemblerDialect == ~0U) 151 return MAI.getAssemblerDialect(); 152 else 153 return AssemblerDialect; 154 } 155 virtual void setAssemblerDialect(unsigned i) { 156 AssemblerDialect = i; 157 } 158 159 virtual bool Warning(SMLoc L, const Twine &Msg, 160 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()); 161 virtual bool Error(SMLoc L, const Twine &Msg, 162 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()); 163 164 const AsmToken &Lex(); 165 166 bool ParseExpression(const MCExpr *&Res); 167 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc); 168 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc); 169 virtual bool ParseAbsoluteExpression(int64_t &Res); 170 171 /// } 172 173 private: 174 void CheckForValidSection(); 175 176 bool ParseStatement(); 177 void EatToEndOfLine(); 178 bool ParseCppHashLineFilenameComment(const SMLoc &L); 179 180 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M); 181 bool expandMacro(SmallString<256> &Buf, StringRef Body, 182 const std::vector<StringRef> &Parameters, 183 const std::vector<std::vector<AsmToken> > &A, 184 const SMLoc &L); 185 void HandleMacroExit(); 186 187 void PrintMacroInstantiations(); 188 void PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 189 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()) const { 190 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 191 } 192 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 193 194 /// EnterIncludeFile - Enter the specified file. This returns true on failure. 195 bool EnterIncludeFile(const std::string &Filename); 196 /// ProcessIncbinFile - Process the specified file for the .incbin directive. 197 /// This returns true on failure. 198 bool ProcessIncbinFile(const std::string &Filename); 199 200 /// \brief Reset the current lexer position to that given by \arg Loc. The 201 /// current token is not set; clients should ensure Lex() is called 202 /// subsequently. 203 void JumpToLoc(SMLoc Loc); 204 205 void EatToEndOfStatement(); 206 207 /// \brief Parse up to the end of statement and a return the contents from the 208 /// current token until the end of the statement; the current token on exit 209 /// will be either the EndOfStatement or EOF. 210 StringRef ParseStringToEndOfStatement(); 211 212 bool ParseAssignment(StringRef Name, bool allow_redef); 213 214 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc); 215 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 216 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 217 bool ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 218 219 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier) 220 /// and set \arg Res to the identifier contents. 221 bool ParseIdentifier(StringRef &Res); 222 223 // Directive Parsing. 224 225 // ".ascii", ".asciiz", ".string" 226 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 227 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ... 228 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ... 229 bool ParseDirectiveFill(); // ".fill" 230 bool ParseDirectiveSpace(); // ".space" 231 bool ParseDirectiveZero(); // ".zero" 232 bool ParseDirectiveSet(StringRef IDVal, bool allow_redef); // ".set", ".equ", ".equiv" 233 bool ParseDirectiveOrg(); // ".org" 234 // ".align{,32}", ".p2align{,w,l}" 235 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize); 236 237 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which 238 /// accepts a single symbol (which should be a label or an external). 239 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr); 240 241 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 242 243 bool ParseDirectiveAbort(); // ".abort" 244 bool ParseDirectiveInclude(); // ".include" 245 bool ParseDirectiveIncbin(); // ".incbin" 246 247 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if" 248 // ".ifdef" or ".ifndef", depending on expect_defined 249 bool ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 250 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 251 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else" 252 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 253 254 /// ParseEscapedString - Parse the current token as a string which may include 255 /// escaped characters and return the string contents. 256 bool ParseEscapedString(std::string &Data); 257 258 const MCExpr *ApplyModifierToExpr(const MCExpr *E, 259 MCSymbolRefExpr::VariantKind Variant); 260 }; 261 262 /// \brief Generic implementations of directive handling, etc. which is shared 263 /// (or the default, at least) for all assembler parser. 264 class GenericAsmParser : public MCAsmParserExtension { 265 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)> 266 void AddDirectiveHandler(StringRef Directive) { 267 getParser().AddDirectiveHandler(this, Directive, 268 HandleDirective<GenericAsmParser, Handler>); 269 } 270 public: 271 GenericAsmParser() {} 272 273 AsmParser &getParser() { 274 return (AsmParser&) this->MCAsmParserExtension::getParser(); 275 } 276 277 virtual void Initialize(MCAsmParser &Parser) { 278 // Call the base implementation. 279 this->MCAsmParserExtension::Initialize(Parser); 280 281 // Debugging directives. 282 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file"); 283 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line"); 284 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc"); 285 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs"); 286 287 // CFI directives. 288 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFISections>( 289 ".cfi_sections"); 290 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>( 291 ".cfi_startproc"); 292 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>( 293 ".cfi_endproc"); 294 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>( 295 ".cfi_def_cfa"); 296 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaOffset>( 297 ".cfi_def_cfa_offset"); 298 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset>( 299 ".cfi_adjust_cfa_offset"); 300 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaRegister>( 301 ".cfi_def_cfa_register"); 302 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIOffset>( 303 ".cfi_offset"); 304 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIRelOffset>( 305 ".cfi_rel_offset"); 306 AddDirectiveHandler< 307 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality"); 308 AddDirectiveHandler< 309 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda"); 310 AddDirectiveHandler< 311 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state"); 312 AddDirectiveHandler< 313 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state"); 314 AddDirectiveHandler< 315 &GenericAsmParser::ParseDirectiveCFISameValue>(".cfi_same_value"); 316 AddDirectiveHandler< 317 &GenericAsmParser::ParseDirectiveCFIRestore>(".cfi_restore"); 318 AddDirectiveHandler< 319 &GenericAsmParser::ParseDirectiveCFIEscape>(".cfi_escape"); 320 AddDirectiveHandler< 321 &GenericAsmParser::ParseDirectiveCFISignalFrame>(".cfi_signal_frame"); 322 323 // Macro directives. 324 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>( 325 ".macros_on"); 326 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>( 327 ".macros_off"); 328 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro"); 329 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm"); 330 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro"); 331 332 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128"); 333 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128"); 334 } 335 336 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 337 338 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc); 339 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc); 340 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc); 341 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc); 342 bool ParseDirectiveCFISections(StringRef, SMLoc DirectiveLoc); 343 bool ParseDirectiveCFIStartProc(StringRef, SMLoc DirectiveLoc); 344 bool ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc); 345 bool ParseDirectiveCFIDefCfa(StringRef, SMLoc DirectiveLoc); 346 bool ParseDirectiveCFIDefCfaOffset(StringRef, SMLoc DirectiveLoc); 347 bool ParseDirectiveCFIAdjustCfaOffset(StringRef, SMLoc DirectiveLoc); 348 bool ParseDirectiveCFIDefCfaRegister(StringRef, SMLoc DirectiveLoc); 349 bool ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc); 350 bool ParseDirectiveCFIRelOffset(StringRef, SMLoc DirectiveLoc); 351 bool ParseDirectiveCFIPersonalityOrLsda(StringRef, SMLoc DirectiveLoc); 352 bool ParseDirectiveCFIRememberState(StringRef, SMLoc DirectiveLoc); 353 bool ParseDirectiveCFIRestoreState(StringRef, SMLoc DirectiveLoc); 354 bool ParseDirectiveCFISameValue(StringRef, SMLoc DirectiveLoc); 355 bool ParseDirectiveCFIRestore(StringRef, SMLoc DirectiveLoc); 356 bool ParseDirectiveCFIEscape(StringRef, SMLoc DirectiveLoc); 357 bool ParseDirectiveCFISignalFrame(StringRef, SMLoc DirectiveLoc); 358 359 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc); 360 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc); 361 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc); 362 363 bool ParseDirectiveLEB128(StringRef, SMLoc); 364 }; 365 366 } 367 368 namespace llvm { 369 370 extern MCAsmParserExtension *createDarwinAsmParser(); 371 extern MCAsmParserExtension *createELFAsmParser(); 372 extern MCAsmParserExtension *createCOFFAsmParser(); 373 374 } 375 376 enum { DEFAULT_ADDRSPACE = 0 }; 377 378 AsmParser::AsmParser(SourceMgr &_SM, MCContext &_Ctx, 379 MCStreamer &_Out, const MCAsmInfo &_MAI) 380 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), MAI(_MAI), SrcMgr(_SM), 381 GenericParser(new GenericAsmParser), PlatformParser(0), 382 CurBuffer(0), MacrosEnabled(true), CppHashLineNumber(0), 383 AssemblerDialect(~0U) { 384 // Save the old handler. 385 SavedDiagHandler = SrcMgr.getDiagHandler(); 386 SavedDiagContext = SrcMgr.getDiagContext(); 387 // Set our own handler which calls the saved handler. 388 SrcMgr.setDiagHandler(DiagHandler, this); 389 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 390 391 // Initialize the generic parser. 392 GenericParser->Initialize(*this); 393 394 // Initialize the platform / file format parser. 395 // 396 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are 397 // created. 398 if (_MAI.hasMicrosoftFastStdCallMangling()) { 399 PlatformParser = createCOFFAsmParser(); 400 PlatformParser->Initialize(*this); 401 } else if (_MAI.hasSubsectionsViaSymbols()) { 402 PlatformParser = createDarwinAsmParser(); 403 PlatformParser->Initialize(*this); 404 } else { 405 PlatformParser = createELFAsmParser(); 406 PlatformParser->Initialize(*this); 407 } 408 } 409 410 AsmParser::~AsmParser() { 411 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!"); 412 413 // Destroy any macros. 414 for (StringMap<Macro*>::iterator it = MacroMap.begin(), 415 ie = MacroMap.end(); it != ie; ++it) 416 delete it->getValue(); 417 418 delete PlatformParser; 419 delete GenericParser; 420 } 421 422 void AsmParser::PrintMacroInstantiations() { 423 // Print the active macro instantiation stack. 424 for (std::vector<MacroInstantiation*>::const_reverse_iterator 425 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it) 426 PrintMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 427 "while in macro instantiation"); 428 } 429 430 bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) { 431 if (FatalAssemblerWarnings) 432 return Error(L, Msg, Ranges); 433 PrintMessage(L, SourceMgr::DK_Warning, Msg, Ranges); 434 PrintMacroInstantiations(); 435 return false; 436 } 437 438 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) { 439 HadError = true; 440 PrintMessage(L, SourceMgr::DK_Error, Msg, Ranges); 441 PrintMacroInstantiations(); 442 return true; 443 } 444 445 bool AsmParser::EnterIncludeFile(const std::string &Filename) { 446 std::string IncludedFile; 447 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 448 if (NewBuf == -1) 449 return true; 450 451 CurBuffer = NewBuf; 452 453 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 454 455 return false; 456 } 457 458 /// Process the specified .incbin file by seaching for it in the include paths 459 /// then just emiting the byte contents of the file to the streamer. This 460 /// returns true on failure. 461 bool AsmParser::ProcessIncbinFile(const std::string &Filename) { 462 std::string IncludedFile; 463 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 464 if (NewBuf == -1) 465 return true; 466 467 // Pick up the bytes from the file and emit them. 468 getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(), 469 DEFAULT_ADDRSPACE); 470 return false; 471 } 472 473 void AsmParser::JumpToLoc(SMLoc Loc) { 474 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc); 475 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer()); 476 } 477 478 const AsmToken &AsmParser::Lex() { 479 const AsmToken *tok = &Lexer.Lex(); 480 481 if (tok->is(AsmToken::Eof)) { 482 // If this is the end of an included file, pop the parent file off the 483 // include stack. 484 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 485 if (ParentIncludeLoc != SMLoc()) { 486 JumpToLoc(ParentIncludeLoc); 487 tok = &Lexer.Lex(); 488 } 489 } 490 491 if (tok->is(AsmToken::Error)) 492 Error(Lexer.getErrLoc(), Lexer.getErr()); 493 494 return *tok; 495 } 496 497 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 498 // Create the initial section, if requested. 499 if (!NoInitialTextSection) 500 Out.InitSections(); 501 502 // Prime the lexer. 503 Lex(); 504 505 HadError = false; 506 AsmCond StartingCondState = TheCondState; 507 508 // If we are generating dwarf for assembly source files save the initial text 509 // section and generate a .file directive. 510 if (getContext().getGenDwarfForAssembly()) { 511 getContext().setGenDwarfSection(getStreamer().getCurrentSection()); 512 MCSymbol *SectionStartSym = getContext().CreateTempSymbol(); 513 getStreamer().EmitLabel(SectionStartSym); 514 getContext().setGenDwarfSectionStartSym(SectionStartSym); 515 getStreamer().EmitDwarfFileDirective(getContext().nextGenDwarfFileNumber(), 516 StringRef(), SrcMgr.getMemoryBuffer(CurBuffer)->getBufferIdentifier()); 517 } 518 519 // While we have input, parse each statement. 520 while (Lexer.isNot(AsmToken::Eof)) { 521 if (!ParseStatement()) continue; 522 523 // We had an error, validate that one was emitted and recover by skipping to 524 // the next line. 525 assert(HadError && "Parse statement returned an error, but none emitted!"); 526 EatToEndOfStatement(); 527 } 528 529 if (TheCondState.TheCond != StartingCondState.TheCond || 530 TheCondState.Ignore != StartingCondState.Ignore) 531 return TokError("unmatched .ifs or .elses"); 532 533 // Check to see there are no empty DwarfFile slots. 534 const std::vector<MCDwarfFile *> &MCDwarfFiles = 535 getContext().getMCDwarfFiles(); 536 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) { 537 if (!MCDwarfFiles[i]) 538 TokError("unassigned file number: " + Twine(i) + " for .file directives"); 539 } 540 541 // Check to see that all assembler local symbols were actually defined. 542 // Targets that don't do subsections via symbols may not want this, though, 543 // so conservatively exclude them. Only do this if we're finalizing, though, 544 // as otherwise we won't necessarilly have seen everything yet. 545 if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) { 546 const MCContext::SymbolTable &Symbols = getContext().getSymbols(); 547 for (MCContext::SymbolTable::const_iterator i = Symbols.begin(), 548 e = Symbols.end(); 549 i != e; ++i) { 550 MCSymbol *Sym = i->getValue(); 551 // Variable symbols may not be marked as defined, so check those 552 // explicitly. If we know it's a variable, we have a definition for 553 // the purposes of this check. 554 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 555 // FIXME: We would really like to refer back to where the symbol was 556 // first referenced for a source location. We need to add something 557 // to track that. Currently, we just point to the end of the file. 558 PrintMessage(getLexer().getLoc(), SourceMgr::DK_Error, 559 "assembler local symbol '" + Sym->getName() + 560 "' not defined"); 561 } 562 } 563 564 565 // Finalize the output stream if there are no errors and if the client wants 566 // us to. 567 if (!HadError && !NoFinalize) 568 Out.Finish(); 569 570 return HadError; 571 } 572 573 void AsmParser::CheckForValidSection() { 574 if (!getStreamer().getCurrentSection()) { 575 TokError("expected section directive before assembly directive"); 576 Out.SwitchSection(Ctx.getMachOSection( 577 "__TEXT", "__text", 578 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS, 579 0, SectionKind::getText())); 580 } 581 } 582 583 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes. 584 void AsmParser::EatToEndOfStatement() { 585 while (Lexer.isNot(AsmToken::EndOfStatement) && 586 Lexer.isNot(AsmToken::Eof)) 587 Lex(); 588 589 // Eat EOL. 590 if (Lexer.is(AsmToken::EndOfStatement)) 591 Lex(); 592 } 593 594 StringRef AsmParser::ParseStringToEndOfStatement() { 595 const char *Start = getTok().getLoc().getPointer(); 596 597 while (Lexer.isNot(AsmToken::EndOfStatement) && 598 Lexer.isNot(AsmToken::Eof)) 599 Lex(); 600 601 const char *End = getTok().getLoc().getPointer(); 602 return StringRef(Start, End - Start); 603 } 604 605 /// ParseParenExpr - Parse a paren expression and return it. 606 /// NOTE: This assumes the leading '(' has already been consumed. 607 /// 608 /// parenexpr ::= expr) 609 /// 610 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 611 if (ParseExpression(Res)) return true; 612 if (Lexer.isNot(AsmToken::RParen)) 613 return TokError("expected ')' in parentheses expression"); 614 EndLoc = Lexer.getLoc(); 615 Lex(); 616 return false; 617 } 618 619 /// ParseBracketExpr - Parse a bracket expression and return it. 620 /// NOTE: This assumes the leading '[' has already been consumed. 621 /// 622 /// bracketexpr ::= expr] 623 /// 624 bool AsmParser::ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 625 if (ParseExpression(Res)) return true; 626 if (Lexer.isNot(AsmToken::RBrac)) 627 return TokError("expected ']' in brackets expression"); 628 EndLoc = Lexer.getLoc(); 629 Lex(); 630 return false; 631 } 632 633 /// ParsePrimaryExpr - Parse a primary expression and return it. 634 /// primaryexpr ::= (parenexpr 635 /// primaryexpr ::= symbol 636 /// primaryexpr ::= number 637 /// primaryexpr ::= '.' 638 /// primaryexpr ::= ~,+,- primaryexpr 639 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 640 switch (Lexer.getKind()) { 641 default: 642 return TokError("unknown token in expression"); 643 // If we have an error assume that we've already handled it. 644 case AsmToken::Error: 645 return true; 646 case AsmToken::Exclaim: 647 Lex(); // Eat the operator. 648 if (ParsePrimaryExpr(Res, EndLoc)) 649 return true; 650 Res = MCUnaryExpr::CreateLNot(Res, getContext()); 651 return false; 652 case AsmToken::Dollar: 653 case AsmToken::String: 654 case AsmToken::Identifier: { 655 EndLoc = Lexer.getLoc(); 656 657 StringRef Identifier; 658 if (ParseIdentifier(Identifier)) 659 return true; 660 661 // This is a symbol reference. 662 std::pair<StringRef, StringRef> Split = Identifier.split('@'); 663 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first); 664 665 // Lookup the symbol variant if used. 666 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 667 if (Split.first.size() != Identifier.size()) { 668 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 669 if (Variant == MCSymbolRefExpr::VK_Invalid) { 670 Variant = MCSymbolRefExpr::VK_None; 671 return TokError("invalid variant '" + Split.second + "'"); 672 } 673 } 674 675 // If this is an absolute variable reference, substitute it now to preserve 676 // semantics in the face of reassignment. 677 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) { 678 if (Variant) 679 return Error(EndLoc, "unexpected modifier on variable reference"); 680 681 Res = Sym->getVariableValue(); 682 return false; 683 } 684 685 // Otherwise create a symbol ref. 686 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext()); 687 return false; 688 } 689 case AsmToken::Integer: { 690 SMLoc Loc = getTok().getLoc(); 691 int64_t IntVal = getTok().getIntVal(); 692 Res = MCConstantExpr::Create(IntVal, getContext()); 693 EndLoc = Lexer.getLoc(); 694 Lex(); // Eat token. 695 // Look for 'b' or 'f' following an Integer as a directional label 696 if (Lexer.getKind() == AsmToken::Identifier) { 697 StringRef IDVal = getTok().getString(); 698 if (IDVal == "f" || IDVal == "b"){ 699 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal, 700 IDVal == "f" ? 1 : 0); 701 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, 702 getContext()); 703 if(IDVal == "b" && Sym->isUndefined()) 704 return Error(Loc, "invalid reference to undefined symbol"); 705 EndLoc = Lexer.getLoc(); 706 Lex(); // Eat identifier. 707 } 708 } 709 return false; 710 } 711 case AsmToken::Real: { 712 APFloat RealVal(APFloat::IEEEdouble, getTok().getString()); 713 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 714 Res = MCConstantExpr::Create(IntVal, getContext()); 715 Lex(); // Eat token. 716 return false; 717 } 718 case AsmToken::Dot: { 719 // This is a '.' reference, which references the current PC. Emit a 720 // temporary label to the streamer and refer to it. 721 MCSymbol *Sym = Ctx.CreateTempSymbol(); 722 Out.EmitLabel(Sym); 723 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext()); 724 EndLoc = Lexer.getLoc(); 725 Lex(); // Eat identifier. 726 return false; 727 } 728 case AsmToken::LParen: 729 Lex(); // Eat the '('. 730 return ParseParenExpr(Res, EndLoc); 731 case AsmToken::LBrac: 732 if (!PlatformParser->HasBracketExpressions()) 733 return TokError("brackets expression not supported on this target"); 734 Lex(); // Eat the '['. 735 return ParseBracketExpr(Res, EndLoc); 736 case AsmToken::Minus: 737 Lex(); // Eat the operator. 738 if (ParsePrimaryExpr(Res, EndLoc)) 739 return true; 740 Res = MCUnaryExpr::CreateMinus(Res, getContext()); 741 return false; 742 case AsmToken::Plus: 743 Lex(); // Eat the operator. 744 if (ParsePrimaryExpr(Res, EndLoc)) 745 return true; 746 Res = MCUnaryExpr::CreatePlus(Res, getContext()); 747 return false; 748 case AsmToken::Tilde: 749 Lex(); // Eat the operator. 750 if (ParsePrimaryExpr(Res, EndLoc)) 751 return true; 752 Res = MCUnaryExpr::CreateNot(Res, getContext()); 753 return false; 754 } 755 } 756 757 bool AsmParser::ParseExpression(const MCExpr *&Res) { 758 SMLoc EndLoc; 759 return ParseExpression(Res, EndLoc); 760 } 761 762 const MCExpr * 763 AsmParser::ApplyModifierToExpr(const MCExpr *E, 764 MCSymbolRefExpr::VariantKind Variant) { 765 // Recurse over the given expression, rebuilding it to apply the given variant 766 // if there is exactly one symbol. 767 switch (E->getKind()) { 768 case MCExpr::Target: 769 case MCExpr::Constant: 770 return 0; 771 772 case MCExpr::SymbolRef: { 773 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 774 775 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 776 TokError("invalid variant on expression '" + 777 getTok().getIdentifier() + "' (already modified)"); 778 return E; 779 } 780 781 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext()); 782 } 783 784 case MCExpr::Unary: { 785 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 786 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant); 787 if (!Sub) 788 return 0; 789 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext()); 790 } 791 792 case MCExpr::Binary: { 793 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 794 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant); 795 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant); 796 797 if (!LHS && !RHS) 798 return 0; 799 800 if (!LHS) LHS = BE->getLHS(); 801 if (!RHS) RHS = BE->getRHS(); 802 803 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext()); 804 } 805 } 806 807 llvm_unreachable("Invalid expression kind!"); 808 } 809 810 /// ParseExpression - Parse an expression and return it. 811 /// 812 /// expr ::= expr &&,|| expr -> lowest. 813 /// expr ::= expr |,^,&,! expr 814 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 815 /// expr ::= expr <<,>> expr 816 /// expr ::= expr +,- expr 817 /// expr ::= expr *,/,% expr -> highest. 818 /// expr ::= primaryexpr 819 /// 820 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 821 // Parse the expression. 822 Res = 0; 823 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc)) 824 return true; 825 826 // As a special case, we support 'a op b @ modifier' by rewriting the 827 // expression to include the modifier. This is inefficient, but in general we 828 // expect users to use 'a@modifier op b'. 829 if (Lexer.getKind() == AsmToken::At) { 830 Lex(); 831 832 if (Lexer.isNot(AsmToken::Identifier)) 833 return TokError("unexpected symbol modifier following '@'"); 834 835 MCSymbolRefExpr::VariantKind Variant = 836 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 837 if (Variant == MCSymbolRefExpr::VK_Invalid) 838 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 839 840 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant); 841 if (!ModifiedRes) { 842 return TokError("invalid modifier '" + getTok().getIdentifier() + 843 "' (no symbols present)"); 844 } 845 846 Res = ModifiedRes; 847 Lex(); 848 } 849 850 // Try to constant fold it up front, if possible. 851 int64_t Value; 852 if (Res->EvaluateAsAbsolute(Value)) 853 Res = MCConstantExpr::Create(Value, getContext()); 854 855 return false; 856 } 857 858 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 859 Res = 0; 860 return ParseParenExpr(Res, EndLoc) || 861 ParseBinOpRHS(1, Res, EndLoc); 862 } 863 864 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) { 865 const MCExpr *Expr; 866 867 SMLoc StartLoc = Lexer.getLoc(); 868 if (ParseExpression(Expr)) 869 return true; 870 871 if (!Expr->EvaluateAsAbsolute(Res)) 872 return Error(StartLoc, "expected absolute expression"); 873 874 return false; 875 } 876 877 static unsigned getBinOpPrecedence(AsmToken::TokenKind K, 878 MCBinaryExpr::Opcode &Kind) { 879 switch (K) { 880 default: 881 return 0; // not a binop. 882 883 // Lowest Precedence: &&, || 884 case AsmToken::AmpAmp: 885 Kind = MCBinaryExpr::LAnd; 886 return 1; 887 case AsmToken::PipePipe: 888 Kind = MCBinaryExpr::LOr; 889 return 1; 890 891 892 // Low Precedence: |, &, ^ 893 // 894 // FIXME: gas seems to support '!' as an infix operator? 895 case AsmToken::Pipe: 896 Kind = MCBinaryExpr::Or; 897 return 2; 898 case AsmToken::Caret: 899 Kind = MCBinaryExpr::Xor; 900 return 2; 901 case AsmToken::Amp: 902 Kind = MCBinaryExpr::And; 903 return 2; 904 905 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 906 case AsmToken::EqualEqual: 907 Kind = MCBinaryExpr::EQ; 908 return 3; 909 case AsmToken::ExclaimEqual: 910 case AsmToken::LessGreater: 911 Kind = MCBinaryExpr::NE; 912 return 3; 913 case AsmToken::Less: 914 Kind = MCBinaryExpr::LT; 915 return 3; 916 case AsmToken::LessEqual: 917 Kind = MCBinaryExpr::LTE; 918 return 3; 919 case AsmToken::Greater: 920 Kind = MCBinaryExpr::GT; 921 return 3; 922 case AsmToken::GreaterEqual: 923 Kind = MCBinaryExpr::GTE; 924 return 3; 925 926 // Intermediate Precedence: <<, >> 927 case AsmToken::LessLess: 928 Kind = MCBinaryExpr::Shl; 929 return 4; 930 case AsmToken::GreaterGreater: 931 Kind = MCBinaryExpr::Shr; 932 return 4; 933 934 // High Intermediate Precedence: +, - 935 case AsmToken::Plus: 936 Kind = MCBinaryExpr::Add; 937 return 5; 938 case AsmToken::Minus: 939 Kind = MCBinaryExpr::Sub; 940 return 5; 941 942 // Highest Precedence: *, /, % 943 case AsmToken::Star: 944 Kind = MCBinaryExpr::Mul; 945 return 6; 946 case AsmToken::Slash: 947 Kind = MCBinaryExpr::Div; 948 return 6; 949 case AsmToken::Percent: 950 Kind = MCBinaryExpr::Mod; 951 return 6; 952 } 953 } 954 955 956 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'. 957 /// Res contains the LHS of the expression on input. 958 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 959 SMLoc &EndLoc) { 960 while (1) { 961 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 962 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 963 964 // If the next token is lower precedence than we are allowed to eat, return 965 // successfully with what we ate already. 966 if (TokPrec < Precedence) 967 return false; 968 969 Lex(); 970 971 // Eat the next primary expression. 972 const MCExpr *RHS; 973 if (ParsePrimaryExpr(RHS, EndLoc)) return true; 974 975 // If BinOp binds less tightly with RHS than the operator after RHS, let 976 // the pending operator take RHS as its LHS. 977 MCBinaryExpr::Opcode Dummy; 978 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 979 if (TokPrec < NextTokPrec) { 980 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true; 981 } 982 983 // Merge LHS and RHS according to operator. 984 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext()); 985 } 986 } 987 988 989 990 991 /// ParseStatement: 992 /// ::= EndOfStatement 993 /// ::= Label* Directive ...Operands... EndOfStatement 994 /// ::= Label* Identifier OperandList* EndOfStatement 995 bool AsmParser::ParseStatement() { 996 if (Lexer.is(AsmToken::EndOfStatement)) { 997 Out.AddBlankLine(); 998 Lex(); 999 return false; 1000 } 1001 1002 // Statements always start with an identifier or are a full line comment. 1003 AsmToken ID = getTok(); 1004 SMLoc IDLoc = ID.getLoc(); 1005 StringRef IDVal; 1006 int64_t LocalLabelVal = -1; 1007 // A full line comment is a '#' as the first token. 1008 if (Lexer.is(AsmToken::Hash)) 1009 return ParseCppHashLineFilenameComment(IDLoc); 1010 1011 // Allow an integer followed by a ':' as a directional local label. 1012 if (Lexer.is(AsmToken::Integer)) { 1013 LocalLabelVal = getTok().getIntVal(); 1014 if (LocalLabelVal < 0) { 1015 if (!TheCondState.Ignore) 1016 return TokError("unexpected token at start of statement"); 1017 IDVal = ""; 1018 } 1019 else { 1020 IDVal = getTok().getString(); 1021 Lex(); // Consume the integer token to be used as an identifier token. 1022 if (Lexer.getKind() != AsmToken::Colon) { 1023 if (!TheCondState.Ignore) 1024 return TokError("unexpected token at start of statement"); 1025 } 1026 } 1027 1028 } else if (Lexer.is(AsmToken::Dot)) { 1029 // Treat '.' as a valid identifier in this context. 1030 Lex(); 1031 IDVal = "."; 1032 1033 } else if (ParseIdentifier(IDVal)) { 1034 if (!TheCondState.Ignore) 1035 return TokError("unexpected token at start of statement"); 1036 IDVal = ""; 1037 } 1038 1039 1040 // Handle conditional assembly here before checking for skipping. We 1041 // have to do this so that .endif isn't skipped in a ".if 0" block for 1042 // example. 1043 if (IDVal == ".if") 1044 return ParseDirectiveIf(IDLoc); 1045 if (IDVal == ".ifdef") 1046 return ParseDirectiveIfdef(IDLoc, true); 1047 if (IDVal == ".ifndef" || IDVal == ".ifnotdef") 1048 return ParseDirectiveIfdef(IDLoc, false); 1049 if (IDVal == ".elseif") 1050 return ParseDirectiveElseIf(IDLoc); 1051 if (IDVal == ".else") 1052 return ParseDirectiveElse(IDLoc); 1053 if (IDVal == ".endif") 1054 return ParseDirectiveEndIf(IDLoc); 1055 1056 // If we are in a ".if 0" block, ignore this statement. 1057 if (TheCondState.Ignore) { 1058 EatToEndOfStatement(); 1059 return false; 1060 } 1061 1062 // FIXME: Recurse on local labels? 1063 1064 // See what kind of statement we have. 1065 switch (Lexer.getKind()) { 1066 case AsmToken::Colon: { 1067 CheckForValidSection(); 1068 1069 // identifier ':' -> Label. 1070 Lex(); 1071 1072 // Diagnose attempt to use '.' as a label. 1073 if (IDVal == ".") 1074 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1075 1076 // Diagnose attempt to use a variable as a label. 1077 // 1078 // FIXME: Diagnostics. Note the location of the definition as a label. 1079 // FIXME: This doesn't diagnose assignment to a symbol which has been 1080 // implicitly marked as external. 1081 MCSymbol *Sym; 1082 if (LocalLabelVal == -1) 1083 Sym = getContext().GetOrCreateSymbol(IDVal); 1084 else 1085 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal); 1086 if (!Sym->isUndefined() || Sym->isVariable()) 1087 return Error(IDLoc, "invalid symbol redefinition"); 1088 1089 // Emit the label. 1090 Out.EmitLabel(Sym); 1091 1092 // If we are generating dwarf for assembly source files then gather the 1093 // info to make a dwarf label entry for this label if needed. 1094 if (getContext().getGenDwarfForAssembly()) 1095 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 1096 IDLoc); 1097 1098 // Consume any end of statement token, if present, to avoid spurious 1099 // AddBlankLine calls(). 1100 if (Lexer.is(AsmToken::EndOfStatement)) { 1101 Lex(); 1102 if (Lexer.is(AsmToken::Eof)) 1103 return false; 1104 } 1105 1106 return ParseStatement(); 1107 } 1108 1109 case AsmToken::Equal: 1110 // identifier '=' ... -> assignment statement 1111 Lex(); 1112 1113 return ParseAssignment(IDVal, true); 1114 1115 default: // Normal instruction or directive. 1116 break; 1117 } 1118 1119 // If macros are enabled, check to see if this is a macro instantiation. 1120 if (MacrosEnabled) 1121 if (const Macro *M = MacroMap.lookup(IDVal)) 1122 return HandleMacroEntry(IDVal, IDLoc, M); 1123 1124 // Otherwise, we have a normal instruction or directive. 1125 if (IDVal[0] == '.' && IDVal != ".") { 1126 // Assembler features 1127 if (IDVal == ".set" || IDVal == ".equ") 1128 return ParseDirectiveSet(IDVal, true); 1129 if (IDVal == ".equiv") 1130 return ParseDirectiveSet(IDVal, false); 1131 1132 // Data directives 1133 1134 if (IDVal == ".ascii") 1135 return ParseDirectiveAscii(IDVal, false); 1136 if (IDVal == ".asciz" || IDVal == ".string") 1137 return ParseDirectiveAscii(IDVal, true); 1138 1139 if (IDVal == ".byte") 1140 return ParseDirectiveValue(1); 1141 if (IDVal == ".short") 1142 return ParseDirectiveValue(2); 1143 if (IDVal == ".value") 1144 return ParseDirectiveValue(2); 1145 if (IDVal == ".2byte") 1146 return ParseDirectiveValue(2); 1147 if (IDVal == ".long") 1148 return ParseDirectiveValue(4); 1149 if (IDVal == ".int") 1150 return ParseDirectiveValue(4); 1151 if (IDVal == ".4byte") 1152 return ParseDirectiveValue(4); 1153 if (IDVal == ".quad") 1154 return ParseDirectiveValue(8); 1155 if (IDVal == ".8byte") 1156 return ParseDirectiveValue(8); 1157 if (IDVal == ".single" || IDVal == ".float") 1158 return ParseDirectiveRealValue(APFloat::IEEEsingle); 1159 if (IDVal == ".double") 1160 return ParseDirectiveRealValue(APFloat::IEEEdouble); 1161 1162 if (IDVal == ".align") { 1163 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes(); 1164 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1); 1165 } 1166 if (IDVal == ".align32") { 1167 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes(); 1168 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4); 1169 } 1170 if (IDVal == ".balign") 1171 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 1172 if (IDVal == ".balignw") 1173 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 1174 if (IDVal == ".balignl") 1175 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 1176 if (IDVal == ".p2align") 1177 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 1178 if (IDVal == ".p2alignw") 1179 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 1180 if (IDVal == ".p2alignl") 1181 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 1182 1183 if (IDVal == ".org") 1184 return ParseDirectiveOrg(); 1185 1186 if (IDVal == ".fill") 1187 return ParseDirectiveFill(); 1188 if (IDVal == ".space" || IDVal == ".skip") 1189 return ParseDirectiveSpace(); 1190 if (IDVal == ".zero") 1191 return ParseDirectiveZero(); 1192 1193 // Symbol attribute directives 1194 1195 if (IDVal == ".globl" || IDVal == ".global") 1196 return ParseDirectiveSymbolAttribute(MCSA_Global); 1197 if (IDVal == ".indirect_symbol") 1198 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol); 1199 if (IDVal == ".lazy_reference") 1200 return ParseDirectiveSymbolAttribute(MCSA_LazyReference); 1201 if (IDVal == ".no_dead_strip") 1202 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 1203 if (IDVal == ".symbol_resolver") 1204 return ParseDirectiveSymbolAttribute(MCSA_SymbolResolver); 1205 if (IDVal == ".private_extern") 1206 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern); 1207 if (IDVal == ".reference") 1208 return ParseDirectiveSymbolAttribute(MCSA_Reference); 1209 if (IDVal == ".weak_definition") 1210 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition); 1211 if (IDVal == ".weak_reference") 1212 return ParseDirectiveSymbolAttribute(MCSA_WeakReference); 1213 if (IDVal == ".weak_def_can_be_hidden") 1214 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 1215 1216 if (IDVal == ".comm" || IDVal == ".common") 1217 return ParseDirectiveComm(/*IsLocal=*/false); 1218 if (IDVal == ".lcomm") 1219 return ParseDirectiveComm(/*IsLocal=*/true); 1220 1221 if (IDVal == ".abort") 1222 return ParseDirectiveAbort(); 1223 if (IDVal == ".include") 1224 return ParseDirectiveInclude(); 1225 if (IDVal == ".incbin") 1226 return ParseDirectiveIncbin(); 1227 1228 if (IDVal == ".code16") 1229 return TokError(Twine(IDVal) + " not supported yet"); 1230 1231 // Look up the handler in the handler table. 1232 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler = 1233 DirectiveMap.lookup(IDVal); 1234 if (Handler.first) 1235 return (*Handler.second)(Handler.first, IDVal, IDLoc); 1236 1237 // Target hook for parsing target specific directives. 1238 if (!getTargetParser().ParseDirective(ID)) 1239 return false; 1240 1241 bool retval = Warning(IDLoc, "ignoring directive for now"); 1242 EatToEndOfStatement(); 1243 return retval; 1244 } 1245 1246 CheckForValidSection(); 1247 1248 // Canonicalize the opcode to lower case. 1249 SmallString<128> Opcode; 1250 for (unsigned i = 0, e = IDVal.size(); i != e; ++i) 1251 Opcode.push_back(tolower(IDVal[i])); 1252 1253 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands; 1254 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc, 1255 ParsedOperands); 1256 1257 // Dump the parsed representation, if requested. 1258 if (getShowParsedOperands()) { 1259 SmallString<256> Str; 1260 raw_svector_ostream OS(Str); 1261 OS << "parsed instruction: ["; 1262 for (unsigned i = 0; i != ParsedOperands.size(); ++i) { 1263 if (i != 0) 1264 OS << ", "; 1265 ParsedOperands[i]->print(OS); 1266 } 1267 OS << "]"; 1268 1269 PrintMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 1270 } 1271 1272 // If we are generating dwarf for assembly source files and the current 1273 // section is the initial text section then generate a .loc directive for 1274 // the instruction. 1275 if (!HadError && getContext().getGenDwarfForAssembly() && 1276 getContext().getGenDwarfSection() == getStreamer().getCurrentSection() ) { 1277 getStreamer().EmitDwarfLocDirective(getContext().getGenDwarfFileNumber(), 1278 SrcMgr.FindLineNumber(IDLoc, CurBuffer), 1279 0, DWARF2_LINE_DEFAULT_IS_STMT ? 1280 DWARF2_FLAG_IS_STMT : 0, 0, 0, 1281 StringRef()); 1282 } 1283 1284 // If parsing succeeded, match the instruction. 1285 if (!HadError) 1286 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands, 1287 Out); 1288 1289 // Free any parsed operands. 1290 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i) 1291 delete ParsedOperands[i]; 1292 1293 // Don't skip the rest of the line, the instruction parser is responsible for 1294 // that. 1295 return false; 1296 } 1297 1298 /// EatToEndOfLine uses the Lexer to eat the characters to the end of the line 1299 /// since they may not be able to be tokenized to get to the end of line token. 1300 void AsmParser::EatToEndOfLine() { 1301 if (!Lexer.is(AsmToken::EndOfStatement)) 1302 Lexer.LexUntilEndOfLine(); 1303 // Eat EOL. 1304 Lex(); 1305 } 1306 1307 /// ParseCppHashLineFilenameComment as this: 1308 /// ::= # number "filename" 1309 /// or just as a full line comment if it doesn't have a number and a string. 1310 bool AsmParser::ParseCppHashLineFilenameComment(const SMLoc &L) { 1311 Lex(); // Eat the hash token. 1312 1313 if (getLexer().isNot(AsmToken::Integer)) { 1314 // Consume the line since in cases it is not a well-formed line directive, 1315 // as if were simply a full line comment. 1316 EatToEndOfLine(); 1317 return false; 1318 } 1319 1320 int64_t LineNumber = getTok().getIntVal(); 1321 Lex(); 1322 1323 if (getLexer().isNot(AsmToken::String)) { 1324 EatToEndOfLine(); 1325 return false; 1326 } 1327 1328 StringRef Filename = getTok().getString(); 1329 // Get rid of the enclosing quotes. 1330 Filename = Filename.substr(1, Filename.size()-2); 1331 1332 // Save the SMLoc, Filename and LineNumber for later use by diagnostics. 1333 CppHashLoc = L; 1334 CppHashFilename = Filename; 1335 CppHashLineNumber = LineNumber; 1336 1337 // Ignore any trailing characters, they're just comment. 1338 EatToEndOfLine(); 1339 return false; 1340 } 1341 1342 /// DiagHandler - will use the the last parsed cpp hash line filename comment 1343 /// for the Filename and LineNo if any in the diagnostic. 1344 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 1345 const AsmParser *Parser = static_cast<const AsmParser*>(Context); 1346 raw_ostream &OS = errs(); 1347 1348 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 1349 const SMLoc &DiagLoc = Diag.getLoc(); 1350 int DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 1351 int CppHashBuf = Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc); 1352 1353 // Like SourceMgr::PrintMessage() we need to print the include stack if any 1354 // before printing the message. 1355 int DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 1356 if (!Parser->SavedDiagHandler && DiagCurBuffer > 0) { 1357 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 1358 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 1359 } 1360 1361 // If we have not parsed a cpp hash line filename comment or the source 1362 // manager changed or buffer changed (like in a nested include) then just 1363 // print the normal diagnostic using its Filename and LineNo. 1364 if (!Parser->CppHashLineNumber || 1365 &DiagSrcMgr != &Parser->SrcMgr || 1366 DiagBuf != CppHashBuf) { 1367 if (Parser->SavedDiagHandler) 1368 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 1369 else 1370 Diag.print(0, OS); 1371 return; 1372 } 1373 1374 // Use the CppHashFilename and calculate a line number based on the 1375 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for 1376 // the diagnostic. 1377 const std::string Filename = Parser->CppHashFilename; 1378 1379 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 1380 int CppHashLocLineNo = 1381 Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf); 1382 int LineNo = Parser->CppHashLineNumber - 1 + 1383 (DiagLocLineNo - CppHashLocLineNo); 1384 1385 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), 1386 Filename, LineNo, Diag.getColumnNo(), 1387 Diag.getKind(), Diag.getMessage(), 1388 Diag.getLineContents(), Diag.getRanges()); 1389 1390 if (Parser->SavedDiagHandler) 1391 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); 1392 else 1393 NewDiag.print(0, OS); 1394 } 1395 1396 bool AsmParser::expandMacro(SmallString<256> &Buf, StringRef Body, 1397 const std::vector<StringRef> &Parameters, 1398 const std::vector<std::vector<AsmToken> > &A, 1399 const SMLoc &L) { 1400 raw_svector_ostream OS(Buf); 1401 unsigned NParameters = Parameters.size(); 1402 if (NParameters != 0 && NParameters != A.size()) 1403 return Error(L, "Wrong number of arguments"); 1404 1405 while (!Body.empty()) { 1406 // Scan for the next substitution. 1407 std::size_t End = Body.size(), Pos = 0; 1408 for (; Pos != End; ++Pos) { 1409 // Check for a substitution or escape. 1410 if (!NParameters) { 1411 // This macro has no parameters, look for $0, $1, etc. 1412 if (Body[Pos] != '$' || Pos + 1 == End) 1413 continue; 1414 1415 char Next = Body[Pos + 1]; 1416 if (Next == '$' || Next == 'n' || isdigit(Next)) 1417 break; 1418 } else { 1419 // This macro has parameters, look for \foo, \bar, etc. 1420 if (Body[Pos] == '\\' && Pos + 1 != End) 1421 break; 1422 } 1423 } 1424 1425 // Add the prefix. 1426 OS << Body.slice(0, Pos); 1427 1428 // Check if we reached the end. 1429 if (Pos == End) 1430 break; 1431 1432 if (!NParameters) { 1433 switch (Body[Pos+1]) { 1434 // $$ => $ 1435 case '$': 1436 OS << '$'; 1437 break; 1438 1439 // $n => number of arguments 1440 case 'n': 1441 OS << A.size(); 1442 break; 1443 1444 // $[0-9] => argument 1445 default: { 1446 // Missing arguments are ignored. 1447 unsigned Index = Body[Pos+1] - '0'; 1448 if (Index >= A.size()) 1449 break; 1450 1451 // Otherwise substitute with the token values, with spaces eliminated. 1452 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(), 1453 ie = A[Index].end(); it != ie; ++it) 1454 OS << it->getString(); 1455 break; 1456 } 1457 } 1458 Pos += 2; 1459 } else { 1460 unsigned I = Pos + 1; 1461 while (isalnum(Body[I]) && I + 1 != End) 1462 ++I; 1463 1464 const char *Begin = Body.data() + Pos +1; 1465 StringRef Argument(Begin, I - (Pos +1)); 1466 unsigned Index = 0; 1467 for (; Index < NParameters; ++Index) 1468 if (Parameters[Index] == Argument) 1469 break; 1470 1471 // FIXME: We should error at the macro definition. 1472 if (Index == NParameters) 1473 return Error(L, "Parameter not found"); 1474 1475 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(), 1476 ie = A[Index].end(); it != ie; ++it) 1477 OS << it->getString(); 1478 1479 Pos += 1 + Argument.size(); 1480 } 1481 // Update the scan point. 1482 Body = Body.substr(Pos); 1483 } 1484 1485 // We include the .endmacro in the buffer as our queue to exit the macro 1486 // instantiation. 1487 OS << ".endmacro\n"; 1488 return false; 1489 } 1490 1491 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL, 1492 MemoryBuffer *I) 1493 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitLoc(EL) 1494 { 1495 } 1496 1497 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc, 1498 const Macro *M) { 1499 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate 1500 // this, although we should protect against infinite loops. 1501 if (ActiveMacros.size() == 20) 1502 return TokError("macros cannot be nested more than 20 levels deep"); 1503 1504 // Parse the macro instantiation arguments. 1505 std::vector<std::vector<AsmToken> > MacroArguments; 1506 MacroArguments.push_back(std::vector<AsmToken>()); 1507 unsigned ParenLevel = 0; 1508 for (;;) { 1509 if (Lexer.is(AsmToken::Eof)) 1510 return TokError("unexpected token in macro instantiation"); 1511 if (Lexer.is(AsmToken::EndOfStatement)) 1512 break; 1513 1514 // If we aren't inside parentheses and this is a comma, start a new token 1515 // list. 1516 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) { 1517 MacroArguments.push_back(std::vector<AsmToken>()); 1518 } else { 1519 // Adjust the current parentheses level. 1520 if (Lexer.is(AsmToken::LParen)) 1521 ++ParenLevel; 1522 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 1523 --ParenLevel; 1524 1525 // Append the token to the current argument list. 1526 MacroArguments.back().push_back(getTok()); 1527 } 1528 Lex(); 1529 } 1530 // If the last argument didn't end up with any tokens, it's not a real 1531 // argument and we should remove it from the list. This happens with either 1532 // a tailing comma or an empty argument list. 1533 if (MacroArguments.back().empty()) 1534 MacroArguments.pop_back(); 1535 1536 // Macro instantiation is lexical, unfortunately. We construct a new buffer 1537 // to hold the macro body with substitutions. 1538 SmallString<256> Buf; 1539 StringRef Body = M->Body; 1540 1541 if (expandMacro(Buf, Body, M->Parameters, MacroArguments, getTok().getLoc())) 1542 return true; 1543 1544 MemoryBuffer *Instantiation = 1545 MemoryBuffer::getMemBufferCopy(Buf.str(), "<instantiation>"); 1546 1547 // Create the macro instantiation object and add to the current macro 1548 // instantiation stack. 1549 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc, 1550 getTok().getLoc(), 1551 Instantiation); 1552 ActiveMacros.push_back(MI); 1553 1554 // Jump to the macro instantiation and prime the lexer. 1555 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc()); 1556 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 1557 Lex(); 1558 1559 return false; 1560 } 1561 1562 void AsmParser::HandleMacroExit() { 1563 // Jump to the EndOfStatement we should return to, and consume it. 1564 JumpToLoc(ActiveMacros.back()->ExitLoc); 1565 Lex(); 1566 1567 // Pop the instantiation entry. 1568 delete ActiveMacros.back(); 1569 ActiveMacros.pop_back(); 1570 } 1571 1572 static bool IsUsedIn(const MCSymbol *Sym, const MCExpr *Value) { 1573 switch (Value->getKind()) { 1574 case MCExpr::Binary: { 1575 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr*>(Value); 1576 return IsUsedIn(Sym, BE->getLHS()) || IsUsedIn(Sym, BE->getRHS()); 1577 break; 1578 } 1579 case MCExpr::Target: 1580 case MCExpr::Constant: 1581 return false; 1582 case MCExpr::SymbolRef: { 1583 const MCSymbol &S = static_cast<const MCSymbolRefExpr*>(Value)->getSymbol(); 1584 if (S.isVariable()) 1585 return IsUsedIn(Sym, S.getVariableValue()); 1586 return &S == Sym; 1587 } 1588 case MCExpr::Unary: 1589 return IsUsedIn(Sym, static_cast<const MCUnaryExpr*>(Value)->getSubExpr()); 1590 } 1591 1592 llvm_unreachable("Unknown expr kind!"); 1593 } 1594 1595 bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) { 1596 // FIXME: Use better location, we should use proper tokens. 1597 SMLoc EqualLoc = Lexer.getLoc(); 1598 1599 const MCExpr *Value; 1600 if (ParseExpression(Value)) 1601 return true; 1602 1603 // Note: we don't count b as used in "a = b". This is to allow 1604 // a = b 1605 // b = c 1606 1607 if (Lexer.isNot(AsmToken::EndOfStatement)) 1608 return TokError("unexpected token in assignment"); 1609 1610 // Error on assignment to '.'. 1611 if (Name == ".") { 1612 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported " 1613 "(use '.space' or '.org').)")); 1614 } 1615 1616 // Eat the end of statement marker. 1617 Lex(); 1618 1619 // Validate that the LHS is allowed to be a variable (either it has not been 1620 // used as a symbol, or it is an absolute symbol). 1621 MCSymbol *Sym = getContext().LookupSymbol(Name); 1622 if (Sym) { 1623 // Diagnose assignment to a label. 1624 // 1625 // FIXME: Diagnostics. Note the location of the definition as a label. 1626 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 1627 if (IsUsedIn(Sym, Value)) 1628 return Error(EqualLoc, "Recursive use of '" + Name + "'"); 1629 else if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable()) 1630 ; // Allow redefinitions of undefined symbols only used in directives. 1631 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef) 1632 ; // Allow redefinitions of variables that haven't yet been used. 1633 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 1634 return Error(EqualLoc, "redefinition of '" + Name + "'"); 1635 else if (!Sym->isVariable()) 1636 return Error(EqualLoc, "invalid assignment to '" + Name + "'"); 1637 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 1638 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" + 1639 Name + "'"); 1640 1641 // Don't count these checks as uses. 1642 Sym->setUsed(false); 1643 } else 1644 Sym = getContext().GetOrCreateSymbol(Name); 1645 1646 // FIXME: Handle '.'. 1647 1648 // Do the assignment. 1649 Out.EmitAssignment(Sym, Value); 1650 1651 return false; 1652 } 1653 1654 /// ParseIdentifier: 1655 /// ::= identifier 1656 /// ::= string 1657 bool AsmParser::ParseIdentifier(StringRef &Res) { 1658 // The assembler has relaxed rules for accepting identifiers, in particular we 1659 // allow things like '.globl $foo', which would normally be separate 1660 // tokens. At this level, we have already lexed so we cannot (currently) 1661 // handle this as a context dependent token, instead we detect adjacent tokens 1662 // and return the combined identifier. 1663 if (Lexer.is(AsmToken::Dollar)) { 1664 SMLoc DollarLoc = getLexer().getLoc(); 1665 1666 // Consume the dollar sign, and check for a following identifier. 1667 Lex(); 1668 if (Lexer.isNot(AsmToken::Identifier)) 1669 return true; 1670 1671 // We have a '$' followed by an identifier, make sure they are adjacent. 1672 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer()) 1673 return true; 1674 1675 // Construct the joined identifier and consume the token. 1676 Res = StringRef(DollarLoc.getPointer(), 1677 getTok().getIdentifier().size() + 1); 1678 Lex(); 1679 return false; 1680 } 1681 1682 if (Lexer.isNot(AsmToken::Identifier) && 1683 Lexer.isNot(AsmToken::String)) 1684 return true; 1685 1686 Res = getTok().getIdentifier(); 1687 1688 Lex(); // Consume the identifier token. 1689 1690 return false; 1691 } 1692 1693 /// ParseDirectiveSet: 1694 /// ::= .equ identifier ',' expression 1695 /// ::= .equiv identifier ',' expression 1696 /// ::= .set identifier ',' expression 1697 bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) { 1698 StringRef Name; 1699 1700 if (ParseIdentifier(Name)) 1701 return TokError("expected identifier after '" + Twine(IDVal) + "'"); 1702 1703 if (getLexer().isNot(AsmToken::Comma)) 1704 return TokError("unexpected token in '" + Twine(IDVal) + "'"); 1705 Lex(); 1706 1707 return ParseAssignment(Name, allow_redef); 1708 } 1709 1710 bool AsmParser::ParseEscapedString(std::string &Data) { 1711 assert(getLexer().is(AsmToken::String) && "Unexpected current token!"); 1712 1713 Data = ""; 1714 StringRef Str = getTok().getStringContents(); 1715 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 1716 if (Str[i] != '\\') { 1717 Data += Str[i]; 1718 continue; 1719 } 1720 1721 // Recognize escaped characters. Note that this escape semantics currently 1722 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes. 1723 ++i; 1724 if (i == e) 1725 return TokError("unexpected backslash at end of string"); 1726 1727 // Recognize octal sequences. 1728 if ((unsigned) (Str[i] - '0') <= 7) { 1729 // Consume up to three octal characters. 1730 unsigned Value = Str[i] - '0'; 1731 1732 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) { 1733 ++i; 1734 Value = Value * 8 + (Str[i] - '0'); 1735 1736 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) { 1737 ++i; 1738 Value = Value * 8 + (Str[i] - '0'); 1739 } 1740 } 1741 1742 if (Value > 255) 1743 return TokError("invalid octal escape sequence (out of range)"); 1744 1745 Data += (unsigned char) Value; 1746 continue; 1747 } 1748 1749 // Otherwise recognize individual escapes. 1750 switch (Str[i]) { 1751 default: 1752 // Just reject invalid escape sequences for now. 1753 return TokError("invalid escape sequence (unrecognized character)"); 1754 1755 case 'b': Data += '\b'; break; 1756 case 'f': Data += '\f'; break; 1757 case 'n': Data += '\n'; break; 1758 case 'r': Data += '\r'; break; 1759 case 't': Data += '\t'; break; 1760 case '"': Data += '"'; break; 1761 case '\\': Data += '\\'; break; 1762 } 1763 } 1764 1765 return false; 1766 } 1767 1768 /// ParseDirectiveAscii: 1769 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 1770 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 1771 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1772 CheckForValidSection(); 1773 1774 for (;;) { 1775 if (getLexer().isNot(AsmToken::String)) 1776 return TokError("expected string in '" + Twine(IDVal) + "' directive"); 1777 1778 std::string Data; 1779 if (ParseEscapedString(Data)) 1780 return true; 1781 1782 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE); 1783 if (ZeroTerminated) 1784 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE); 1785 1786 Lex(); 1787 1788 if (getLexer().is(AsmToken::EndOfStatement)) 1789 break; 1790 1791 if (getLexer().isNot(AsmToken::Comma)) 1792 return TokError("unexpected token in '" + Twine(IDVal) + "' directive"); 1793 Lex(); 1794 } 1795 } 1796 1797 Lex(); 1798 return false; 1799 } 1800 1801 /// ParseDirectiveValue 1802 /// ::= (.byte | .short | ... ) [ expression (, expression)* ] 1803 bool AsmParser::ParseDirectiveValue(unsigned Size) { 1804 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1805 CheckForValidSection(); 1806 1807 for (;;) { 1808 const MCExpr *Value; 1809 SMLoc ExprLoc = getLexer().getLoc(); 1810 if (ParseExpression(Value)) 1811 return true; 1812 1813 // Special case constant expressions to match code generator. 1814 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 1815 assert(Size <= 8 && "Invalid size"); 1816 uint64_t IntValue = MCE->getValue(); 1817 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 1818 return Error(ExprLoc, "literal value out of range for directive"); 1819 getStreamer().EmitIntValue(IntValue, Size, DEFAULT_ADDRSPACE); 1820 } else 1821 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE); 1822 1823 if (getLexer().is(AsmToken::EndOfStatement)) 1824 break; 1825 1826 // FIXME: Improve diagnostic. 1827 if (getLexer().isNot(AsmToken::Comma)) 1828 return TokError("unexpected token in directive"); 1829 Lex(); 1830 } 1831 } 1832 1833 Lex(); 1834 return false; 1835 } 1836 1837 /// ParseDirectiveRealValue 1838 /// ::= (.single | .double) [ expression (, expression)* ] 1839 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) { 1840 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1841 CheckForValidSection(); 1842 1843 for (;;) { 1844 // We don't truly support arithmetic on floating point expressions, so we 1845 // have to manually parse unary prefixes. 1846 bool IsNeg = false; 1847 if (getLexer().is(AsmToken::Minus)) { 1848 Lex(); 1849 IsNeg = true; 1850 } else if (getLexer().is(AsmToken::Plus)) 1851 Lex(); 1852 1853 if (getLexer().isNot(AsmToken::Integer) && 1854 getLexer().isNot(AsmToken::Real) && 1855 getLexer().isNot(AsmToken::Identifier)) 1856 return TokError("unexpected token in directive"); 1857 1858 // Convert to an APFloat. 1859 APFloat Value(Semantics); 1860 StringRef IDVal = getTok().getString(); 1861 if (getLexer().is(AsmToken::Identifier)) { 1862 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf")) 1863 Value = APFloat::getInf(Semantics); 1864 else if (!IDVal.compare_lower("nan")) 1865 Value = APFloat::getNaN(Semantics, false, ~0); 1866 else 1867 return TokError("invalid floating point literal"); 1868 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) == 1869 APFloat::opInvalidOp) 1870 return TokError("invalid floating point literal"); 1871 if (IsNeg) 1872 Value.changeSign(); 1873 1874 // Consume the numeric token. 1875 Lex(); 1876 1877 // Emit the value as an integer. 1878 APInt AsInt = Value.bitcastToAPInt(); 1879 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 1880 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE); 1881 1882 if (getLexer().is(AsmToken::EndOfStatement)) 1883 break; 1884 1885 if (getLexer().isNot(AsmToken::Comma)) 1886 return TokError("unexpected token in directive"); 1887 Lex(); 1888 } 1889 } 1890 1891 Lex(); 1892 return false; 1893 } 1894 1895 /// ParseDirectiveSpace 1896 /// ::= .space expression [ , expression ] 1897 bool AsmParser::ParseDirectiveSpace() { 1898 CheckForValidSection(); 1899 1900 int64_t NumBytes; 1901 if (ParseAbsoluteExpression(NumBytes)) 1902 return true; 1903 1904 int64_t FillExpr = 0; 1905 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1906 if (getLexer().isNot(AsmToken::Comma)) 1907 return TokError("unexpected token in '.space' directive"); 1908 Lex(); 1909 1910 if (ParseAbsoluteExpression(FillExpr)) 1911 return true; 1912 1913 if (getLexer().isNot(AsmToken::EndOfStatement)) 1914 return TokError("unexpected token in '.space' directive"); 1915 } 1916 1917 Lex(); 1918 1919 if (NumBytes <= 0) 1920 return TokError("invalid number of bytes in '.space' directive"); 1921 1922 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 1923 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE); 1924 1925 return false; 1926 } 1927 1928 /// ParseDirectiveZero 1929 /// ::= .zero expression 1930 bool AsmParser::ParseDirectiveZero() { 1931 CheckForValidSection(); 1932 1933 int64_t NumBytes; 1934 if (ParseAbsoluteExpression(NumBytes)) 1935 return true; 1936 1937 int64_t Val = 0; 1938 if (getLexer().is(AsmToken::Comma)) { 1939 Lex(); 1940 if (ParseAbsoluteExpression(Val)) 1941 return true; 1942 } 1943 1944 if (getLexer().isNot(AsmToken::EndOfStatement)) 1945 return TokError("unexpected token in '.zero' directive"); 1946 1947 Lex(); 1948 1949 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE); 1950 1951 return false; 1952 } 1953 1954 /// ParseDirectiveFill 1955 /// ::= .fill expression , expression , expression 1956 bool AsmParser::ParseDirectiveFill() { 1957 CheckForValidSection(); 1958 1959 int64_t NumValues; 1960 if (ParseAbsoluteExpression(NumValues)) 1961 return true; 1962 1963 if (getLexer().isNot(AsmToken::Comma)) 1964 return TokError("unexpected token in '.fill' directive"); 1965 Lex(); 1966 1967 int64_t FillSize; 1968 if (ParseAbsoluteExpression(FillSize)) 1969 return true; 1970 1971 if (getLexer().isNot(AsmToken::Comma)) 1972 return TokError("unexpected token in '.fill' directive"); 1973 Lex(); 1974 1975 int64_t FillExpr; 1976 if (ParseAbsoluteExpression(FillExpr)) 1977 return true; 1978 1979 if (getLexer().isNot(AsmToken::EndOfStatement)) 1980 return TokError("unexpected token in '.fill' directive"); 1981 1982 Lex(); 1983 1984 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8) 1985 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8"); 1986 1987 for (uint64_t i = 0, e = NumValues; i != e; ++i) 1988 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE); 1989 1990 return false; 1991 } 1992 1993 /// ParseDirectiveOrg 1994 /// ::= .org expression [ , expression ] 1995 bool AsmParser::ParseDirectiveOrg() { 1996 CheckForValidSection(); 1997 1998 const MCExpr *Offset; 1999 SMLoc Loc = getTok().getLoc(); 2000 if (ParseExpression(Offset)) 2001 return true; 2002 2003 // Parse optional fill expression. 2004 int64_t FillExpr = 0; 2005 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2006 if (getLexer().isNot(AsmToken::Comma)) 2007 return TokError("unexpected token in '.org' directive"); 2008 Lex(); 2009 2010 if (ParseAbsoluteExpression(FillExpr)) 2011 return true; 2012 2013 if (getLexer().isNot(AsmToken::EndOfStatement)) 2014 return TokError("unexpected token in '.org' directive"); 2015 } 2016 2017 Lex(); 2018 2019 // Only limited forms of relocatable expressions are accepted here, it 2020 // has to be relative to the current section. The streamer will return 2021 // 'true' if the expression wasn't evaluatable. 2022 if (getStreamer().EmitValueToOffset(Offset, FillExpr)) 2023 return Error(Loc, "expected assembly-time absolute expression"); 2024 2025 return false; 2026 } 2027 2028 /// ParseDirectiveAlign 2029 /// ::= {.align, ...} expression [ , expression [ , expression ]] 2030 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 2031 CheckForValidSection(); 2032 2033 SMLoc AlignmentLoc = getLexer().getLoc(); 2034 int64_t Alignment; 2035 if (ParseAbsoluteExpression(Alignment)) 2036 return true; 2037 2038 SMLoc MaxBytesLoc; 2039 bool HasFillExpr = false; 2040 int64_t FillExpr = 0; 2041 int64_t MaxBytesToFill = 0; 2042 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2043 if (getLexer().isNot(AsmToken::Comma)) 2044 return TokError("unexpected token in directive"); 2045 Lex(); 2046 2047 // The fill expression can be omitted while specifying a maximum number of 2048 // alignment bytes, e.g: 2049 // .align 3,,4 2050 if (getLexer().isNot(AsmToken::Comma)) { 2051 HasFillExpr = true; 2052 if (ParseAbsoluteExpression(FillExpr)) 2053 return true; 2054 } 2055 2056 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2057 if (getLexer().isNot(AsmToken::Comma)) 2058 return TokError("unexpected token in directive"); 2059 Lex(); 2060 2061 MaxBytesLoc = getLexer().getLoc(); 2062 if (ParseAbsoluteExpression(MaxBytesToFill)) 2063 return true; 2064 2065 if (getLexer().isNot(AsmToken::EndOfStatement)) 2066 return TokError("unexpected token in directive"); 2067 } 2068 } 2069 2070 Lex(); 2071 2072 if (!HasFillExpr) 2073 FillExpr = 0; 2074 2075 // Compute alignment in bytes. 2076 if (IsPow2) { 2077 // FIXME: Diagnose overflow. 2078 if (Alignment >= 32) { 2079 Error(AlignmentLoc, "invalid alignment value"); 2080 Alignment = 31; 2081 } 2082 2083 Alignment = 1ULL << Alignment; 2084 } 2085 2086 // Diagnose non-sensical max bytes to align. 2087 if (MaxBytesLoc.isValid()) { 2088 if (MaxBytesToFill < 1) { 2089 Error(MaxBytesLoc, "alignment directive can never be satisfied in this " 2090 "many bytes, ignoring maximum bytes expression"); 2091 MaxBytesToFill = 0; 2092 } 2093 2094 if (MaxBytesToFill >= Alignment) { 2095 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 2096 "has no effect"); 2097 MaxBytesToFill = 0; 2098 } 2099 } 2100 2101 // Check whether we should use optimal code alignment for this .align 2102 // directive. 2103 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign(); 2104 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 2105 ValueSize == 1 && UseCodeAlign) { 2106 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill); 2107 } else { 2108 // FIXME: Target specific behavior about how the "extra" bytes are filled. 2109 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize, 2110 MaxBytesToFill); 2111 } 2112 2113 return false; 2114 } 2115 2116 /// ParseDirectiveSymbolAttribute 2117 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 2118 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 2119 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2120 for (;;) { 2121 StringRef Name; 2122 SMLoc Loc = getTok().getLoc(); 2123 2124 if (ParseIdentifier(Name)) 2125 return Error(Loc, "expected identifier in directive"); 2126 2127 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2128 2129 // Assembler local symbols don't make any sense here. Complain loudly. 2130 if (Sym->isTemporary()) 2131 return Error(Loc, "non-local symbol required in directive"); 2132 2133 getStreamer().EmitSymbolAttribute(Sym, Attr); 2134 2135 if (getLexer().is(AsmToken::EndOfStatement)) 2136 break; 2137 2138 if (getLexer().isNot(AsmToken::Comma)) 2139 return TokError("unexpected token in directive"); 2140 Lex(); 2141 } 2142 } 2143 2144 Lex(); 2145 return false; 2146 } 2147 2148 /// ParseDirectiveComm 2149 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 2150 bool AsmParser::ParseDirectiveComm(bool IsLocal) { 2151 CheckForValidSection(); 2152 2153 SMLoc IDLoc = getLexer().getLoc(); 2154 StringRef Name; 2155 if (ParseIdentifier(Name)) 2156 return TokError("expected identifier in directive"); 2157 2158 // Handle the identifier as the key symbol. 2159 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2160 2161 if (getLexer().isNot(AsmToken::Comma)) 2162 return TokError("unexpected token in directive"); 2163 Lex(); 2164 2165 int64_t Size; 2166 SMLoc SizeLoc = getLexer().getLoc(); 2167 if (ParseAbsoluteExpression(Size)) 2168 return true; 2169 2170 int64_t Pow2Alignment = 0; 2171 SMLoc Pow2AlignmentLoc; 2172 if (getLexer().is(AsmToken::Comma)) { 2173 Lex(); 2174 Pow2AlignmentLoc = getLexer().getLoc(); 2175 if (ParseAbsoluteExpression(Pow2Alignment)) 2176 return true; 2177 2178 // If this target takes alignments in bytes (not log) validate and convert. 2179 if (Lexer.getMAI().getAlignmentIsInBytes()) { 2180 if (!isPowerOf2_64(Pow2Alignment)) 2181 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 2182 Pow2Alignment = Log2_64(Pow2Alignment); 2183 } 2184 } 2185 2186 if (getLexer().isNot(AsmToken::EndOfStatement)) 2187 return TokError("unexpected token in '.comm' or '.lcomm' directive"); 2188 2189 Lex(); 2190 2191 // NOTE: a size of zero for a .comm should create a undefined symbol 2192 // but a size of .lcomm creates a bss symbol of size zero. 2193 if (Size < 0) 2194 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 2195 "be less than zero"); 2196 2197 // NOTE: The alignment in the directive is a power of 2 value, the assembler 2198 // may internally end up wanting an alignment in bytes. 2199 // FIXME: Diagnose overflow. 2200 if (Pow2Alignment < 0) 2201 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 2202 "alignment, can't be less than zero"); 2203 2204 if (!Sym->isUndefined()) 2205 return Error(IDLoc, "invalid symbol redefinition"); 2206 2207 // '.lcomm' is equivalent to '.zerofill'. 2208 // Create the Symbol as a common or local common with Size and Pow2Alignment 2209 if (IsLocal) { 2210 getStreamer().EmitZerofill(Ctx.getMachOSection( 2211 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL, 2212 0, SectionKind::getBSS()), 2213 Sym, Size, 1 << Pow2Alignment); 2214 return false; 2215 } 2216 2217 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 2218 return false; 2219 } 2220 2221 /// ParseDirectiveAbort 2222 /// ::= .abort [... message ...] 2223 bool AsmParser::ParseDirectiveAbort() { 2224 // FIXME: Use loc from directive. 2225 SMLoc Loc = getLexer().getLoc(); 2226 2227 StringRef Str = ParseStringToEndOfStatement(); 2228 if (getLexer().isNot(AsmToken::EndOfStatement)) 2229 return TokError("unexpected token in '.abort' directive"); 2230 2231 Lex(); 2232 2233 if (Str.empty()) 2234 Error(Loc, ".abort detected. Assembly stopping."); 2235 else 2236 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 2237 // FIXME: Actually abort assembly here. 2238 2239 return false; 2240 } 2241 2242 /// ParseDirectiveInclude 2243 /// ::= .include "filename" 2244 bool AsmParser::ParseDirectiveInclude() { 2245 if (getLexer().isNot(AsmToken::String)) 2246 return TokError("expected string in '.include' directive"); 2247 2248 std::string Filename = getTok().getString(); 2249 SMLoc IncludeLoc = getLexer().getLoc(); 2250 Lex(); 2251 2252 if (getLexer().isNot(AsmToken::EndOfStatement)) 2253 return TokError("unexpected token in '.include' directive"); 2254 2255 // Strip the quotes. 2256 Filename = Filename.substr(1, Filename.size()-2); 2257 2258 // Attempt to switch the lexer to the included file before consuming the end 2259 // of statement to avoid losing it when we switch. 2260 if (EnterIncludeFile(Filename)) { 2261 Error(IncludeLoc, "Could not find include file '" + Filename + "'"); 2262 return true; 2263 } 2264 2265 return false; 2266 } 2267 2268 /// ParseDirectiveIncbin 2269 /// ::= .incbin "filename" 2270 bool AsmParser::ParseDirectiveIncbin() { 2271 if (getLexer().isNot(AsmToken::String)) 2272 return TokError("expected string in '.incbin' directive"); 2273 2274 std::string Filename = getTok().getString(); 2275 SMLoc IncbinLoc = getLexer().getLoc(); 2276 Lex(); 2277 2278 if (getLexer().isNot(AsmToken::EndOfStatement)) 2279 return TokError("unexpected token in '.incbin' directive"); 2280 2281 // Strip the quotes. 2282 Filename = Filename.substr(1, Filename.size()-2); 2283 2284 // Attempt to process the included file. 2285 if (ProcessIncbinFile(Filename)) { 2286 Error(IncbinLoc, "Could not find incbin file '" + Filename + "'"); 2287 return true; 2288 } 2289 2290 return false; 2291 } 2292 2293 /// ParseDirectiveIf 2294 /// ::= .if expression 2295 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) { 2296 TheCondStack.push_back(TheCondState); 2297 TheCondState.TheCond = AsmCond::IfCond; 2298 if(TheCondState.Ignore) { 2299 EatToEndOfStatement(); 2300 } 2301 else { 2302 int64_t ExprValue; 2303 if (ParseAbsoluteExpression(ExprValue)) 2304 return true; 2305 2306 if (getLexer().isNot(AsmToken::EndOfStatement)) 2307 return TokError("unexpected token in '.if' directive"); 2308 2309 Lex(); 2310 2311 TheCondState.CondMet = ExprValue; 2312 TheCondState.Ignore = !TheCondState.CondMet; 2313 } 2314 2315 return false; 2316 } 2317 2318 bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 2319 StringRef Name; 2320 TheCondStack.push_back(TheCondState); 2321 TheCondState.TheCond = AsmCond::IfCond; 2322 2323 if (TheCondState.Ignore) { 2324 EatToEndOfStatement(); 2325 } else { 2326 if (ParseIdentifier(Name)) 2327 return TokError("expected identifier after '.ifdef'"); 2328 2329 Lex(); 2330 2331 MCSymbol *Sym = getContext().LookupSymbol(Name); 2332 2333 if (expect_defined) 2334 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined()); 2335 else 2336 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined()); 2337 TheCondState.Ignore = !TheCondState.CondMet; 2338 } 2339 2340 return false; 2341 } 2342 2343 /// ParseDirectiveElseIf 2344 /// ::= .elseif expression 2345 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) { 2346 if (TheCondState.TheCond != AsmCond::IfCond && 2347 TheCondState.TheCond != AsmCond::ElseIfCond) 2348 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or " 2349 " an .elseif"); 2350 TheCondState.TheCond = AsmCond::ElseIfCond; 2351 2352 bool LastIgnoreState = false; 2353 if (!TheCondStack.empty()) 2354 LastIgnoreState = TheCondStack.back().Ignore; 2355 if (LastIgnoreState || TheCondState.CondMet) { 2356 TheCondState.Ignore = true; 2357 EatToEndOfStatement(); 2358 } 2359 else { 2360 int64_t ExprValue; 2361 if (ParseAbsoluteExpression(ExprValue)) 2362 return true; 2363 2364 if (getLexer().isNot(AsmToken::EndOfStatement)) 2365 return TokError("unexpected token in '.elseif' directive"); 2366 2367 Lex(); 2368 TheCondState.CondMet = ExprValue; 2369 TheCondState.Ignore = !TheCondState.CondMet; 2370 } 2371 2372 return false; 2373 } 2374 2375 /// ParseDirectiveElse 2376 /// ::= .else 2377 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) { 2378 if (getLexer().isNot(AsmToken::EndOfStatement)) 2379 return TokError("unexpected token in '.else' directive"); 2380 2381 Lex(); 2382 2383 if (TheCondState.TheCond != AsmCond::IfCond && 2384 TheCondState.TheCond != AsmCond::ElseIfCond) 2385 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an " 2386 ".elseif"); 2387 TheCondState.TheCond = AsmCond::ElseCond; 2388 bool LastIgnoreState = false; 2389 if (!TheCondStack.empty()) 2390 LastIgnoreState = TheCondStack.back().Ignore; 2391 if (LastIgnoreState || TheCondState.CondMet) 2392 TheCondState.Ignore = true; 2393 else 2394 TheCondState.Ignore = false; 2395 2396 return false; 2397 } 2398 2399 /// ParseDirectiveEndIf 2400 /// ::= .endif 2401 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) { 2402 if (getLexer().isNot(AsmToken::EndOfStatement)) 2403 return TokError("unexpected token in '.endif' directive"); 2404 2405 Lex(); 2406 2407 if ((TheCondState.TheCond == AsmCond::NoCond) || 2408 TheCondStack.empty()) 2409 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or " 2410 ".else"); 2411 if (!TheCondStack.empty()) { 2412 TheCondState = TheCondStack.back(); 2413 TheCondStack.pop_back(); 2414 } 2415 2416 return false; 2417 } 2418 2419 /// ParseDirectiveFile 2420 /// ::= .file [number] filename 2421 /// ::= .file number directory filename 2422 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) { 2423 // FIXME: I'm not sure what this is. 2424 int64_t FileNumber = -1; 2425 SMLoc FileNumberLoc = getLexer().getLoc(); 2426 if (getLexer().is(AsmToken::Integer)) { 2427 FileNumber = getTok().getIntVal(); 2428 Lex(); 2429 2430 if (FileNumber < 1) 2431 return TokError("file number less than one"); 2432 } 2433 2434 if (getLexer().isNot(AsmToken::String)) 2435 return TokError("unexpected token in '.file' directive"); 2436 2437 // Usually the directory and filename together, otherwise just the directory. 2438 StringRef Path = getTok().getString(); 2439 Path = Path.substr(1, Path.size()-2); 2440 Lex(); 2441 2442 StringRef Directory; 2443 StringRef Filename; 2444 if (getLexer().is(AsmToken::String)) { 2445 if (FileNumber == -1) 2446 return TokError("explicit path specified, but no file number"); 2447 Filename = getTok().getString(); 2448 Filename = Filename.substr(1, Filename.size()-2); 2449 Directory = Path; 2450 Lex(); 2451 } else { 2452 Filename = Path; 2453 } 2454 2455 if (getLexer().isNot(AsmToken::EndOfStatement)) 2456 return TokError("unexpected token in '.file' directive"); 2457 2458 if (FileNumber == -1) 2459 getStreamer().EmitFileDirective(Filename); 2460 else { 2461 if (getContext().getGenDwarfForAssembly() == true) 2462 Error(DirectiveLoc, "input can't have .file dwarf directives when -g is " 2463 "used to generate dwarf debug info for assembly code"); 2464 2465 if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename)) 2466 Error(FileNumberLoc, "file number already allocated"); 2467 } 2468 2469 return false; 2470 } 2471 2472 /// ParseDirectiveLine 2473 /// ::= .line [number] 2474 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) { 2475 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2476 if (getLexer().isNot(AsmToken::Integer)) 2477 return TokError("unexpected token in '.line' directive"); 2478 2479 int64_t LineNumber = getTok().getIntVal(); 2480 (void) LineNumber; 2481 Lex(); 2482 2483 // FIXME: Do something with the .line. 2484 } 2485 2486 if (getLexer().isNot(AsmToken::EndOfStatement)) 2487 return TokError("unexpected token in '.line' directive"); 2488 2489 return false; 2490 } 2491 2492 2493 /// ParseDirectiveLoc 2494 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 2495 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 2496 /// The first number is a file number, must have been previously assigned with 2497 /// a .file directive, the second number is the line number and optionally the 2498 /// third number is a column position (zero if not specified). The remaining 2499 /// optional items are .loc sub-directives. 2500 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) { 2501 2502 if (getLexer().isNot(AsmToken::Integer)) 2503 return TokError("unexpected token in '.loc' directive"); 2504 int64_t FileNumber = getTok().getIntVal(); 2505 if (FileNumber < 1) 2506 return TokError("file number less than one in '.loc' directive"); 2507 if (!getContext().isValidDwarfFileNumber(FileNumber)) 2508 return TokError("unassigned file number in '.loc' directive"); 2509 Lex(); 2510 2511 int64_t LineNumber = 0; 2512 if (getLexer().is(AsmToken::Integer)) { 2513 LineNumber = getTok().getIntVal(); 2514 if (LineNumber < 1) 2515 return TokError("line number less than one in '.loc' directive"); 2516 Lex(); 2517 } 2518 2519 int64_t ColumnPos = 0; 2520 if (getLexer().is(AsmToken::Integer)) { 2521 ColumnPos = getTok().getIntVal(); 2522 if (ColumnPos < 0) 2523 return TokError("column position less than zero in '.loc' directive"); 2524 Lex(); 2525 } 2526 2527 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0; 2528 unsigned Isa = 0; 2529 int64_t Discriminator = 0; 2530 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2531 for (;;) { 2532 if (getLexer().is(AsmToken::EndOfStatement)) 2533 break; 2534 2535 StringRef Name; 2536 SMLoc Loc = getTok().getLoc(); 2537 if (getParser().ParseIdentifier(Name)) 2538 return TokError("unexpected token in '.loc' directive"); 2539 2540 if (Name == "basic_block") 2541 Flags |= DWARF2_FLAG_BASIC_BLOCK; 2542 else if (Name == "prologue_end") 2543 Flags |= DWARF2_FLAG_PROLOGUE_END; 2544 else if (Name == "epilogue_begin") 2545 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 2546 else if (Name == "is_stmt") { 2547 SMLoc Loc = getTok().getLoc(); 2548 const MCExpr *Value; 2549 if (getParser().ParseExpression(Value)) 2550 return true; 2551 // The expression must be the constant 0 or 1. 2552 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 2553 int Value = MCE->getValue(); 2554 if (Value == 0) 2555 Flags &= ~DWARF2_FLAG_IS_STMT; 2556 else if (Value == 1) 2557 Flags |= DWARF2_FLAG_IS_STMT; 2558 else 2559 return Error(Loc, "is_stmt value not 0 or 1"); 2560 } 2561 else { 2562 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 2563 } 2564 } 2565 else if (Name == "isa") { 2566 SMLoc Loc = getTok().getLoc(); 2567 const MCExpr *Value; 2568 if (getParser().ParseExpression(Value)) 2569 return true; 2570 // The expression must be a constant greater or equal to 0. 2571 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 2572 int Value = MCE->getValue(); 2573 if (Value < 0) 2574 return Error(Loc, "isa number less than zero"); 2575 Isa = Value; 2576 } 2577 else { 2578 return Error(Loc, "isa number not a constant value"); 2579 } 2580 } 2581 else if (Name == "discriminator") { 2582 if (getParser().ParseAbsoluteExpression(Discriminator)) 2583 return true; 2584 } 2585 else { 2586 return Error(Loc, "unknown sub-directive in '.loc' directive"); 2587 } 2588 2589 if (getLexer().is(AsmToken::EndOfStatement)) 2590 break; 2591 } 2592 } 2593 2594 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 2595 Isa, Discriminator, StringRef()); 2596 2597 return false; 2598 } 2599 2600 /// ParseDirectiveStabs 2601 /// ::= .stabs string, number, number, number 2602 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive, 2603 SMLoc DirectiveLoc) { 2604 return TokError("unsupported directive '" + Directive + "'"); 2605 } 2606 2607 /// ParseDirectiveCFISections 2608 /// ::= .cfi_sections section [, section] 2609 bool GenericAsmParser::ParseDirectiveCFISections(StringRef, 2610 SMLoc DirectiveLoc) { 2611 StringRef Name; 2612 bool EH = false; 2613 bool Debug = false; 2614 2615 if (getParser().ParseIdentifier(Name)) 2616 return TokError("Expected an identifier"); 2617 2618 if (Name == ".eh_frame") 2619 EH = true; 2620 else if (Name == ".debug_frame") 2621 Debug = true; 2622 2623 if (getLexer().is(AsmToken::Comma)) { 2624 Lex(); 2625 2626 if (getParser().ParseIdentifier(Name)) 2627 return TokError("Expected an identifier"); 2628 2629 if (Name == ".eh_frame") 2630 EH = true; 2631 else if (Name == ".debug_frame") 2632 Debug = true; 2633 } 2634 2635 getStreamer().EmitCFISections(EH, Debug); 2636 2637 return false; 2638 } 2639 2640 /// ParseDirectiveCFIStartProc 2641 /// ::= .cfi_startproc 2642 bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef, 2643 SMLoc DirectiveLoc) { 2644 getStreamer().EmitCFIStartProc(); 2645 return false; 2646 } 2647 2648 /// ParseDirectiveCFIEndProc 2649 /// ::= .cfi_endproc 2650 bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) { 2651 getStreamer().EmitCFIEndProc(); 2652 return false; 2653 } 2654 2655 /// ParseRegisterOrRegisterNumber - parse register name or number. 2656 bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register, 2657 SMLoc DirectiveLoc) { 2658 unsigned RegNo; 2659 2660 if (getLexer().isNot(AsmToken::Integer)) { 2661 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc, 2662 DirectiveLoc)) 2663 return true; 2664 Register = getContext().getRegisterInfo().getDwarfRegNum(RegNo, true); 2665 } else 2666 return getParser().ParseAbsoluteExpression(Register); 2667 2668 return false; 2669 } 2670 2671 /// ParseDirectiveCFIDefCfa 2672 /// ::= .cfi_def_cfa register, offset 2673 bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef, 2674 SMLoc DirectiveLoc) { 2675 int64_t Register = 0; 2676 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2677 return true; 2678 2679 if (getLexer().isNot(AsmToken::Comma)) 2680 return TokError("unexpected token in directive"); 2681 Lex(); 2682 2683 int64_t Offset = 0; 2684 if (getParser().ParseAbsoluteExpression(Offset)) 2685 return true; 2686 2687 getStreamer().EmitCFIDefCfa(Register, Offset); 2688 return false; 2689 } 2690 2691 /// ParseDirectiveCFIDefCfaOffset 2692 /// ::= .cfi_def_cfa_offset offset 2693 bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef, 2694 SMLoc DirectiveLoc) { 2695 int64_t Offset = 0; 2696 if (getParser().ParseAbsoluteExpression(Offset)) 2697 return true; 2698 2699 getStreamer().EmitCFIDefCfaOffset(Offset); 2700 return false; 2701 } 2702 2703 /// ParseDirectiveCFIAdjustCfaOffset 2704 /// ::= .cfi_adjust_cfa_offset adjustment 2705 bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef, 2706 SMLoc DirectiveLoc) { 2707 int64_t Adjustment = 0; 2708 if (getParser().ParseAbsoluteExpression(Adjustment)) 2709 return true; 2710 2711 getStreamer().EmitCFIAdjustCfaOffset(Adjustment); 2712 return false; 2713 } 2714 2715 /// ParseDirectiveCFIDefCfaRegister 2716 /// ::= .cfi_def_cfa_register register 2717 bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef, 2718 SMLoc DirectiveLoc) { 2719 int64_t Register = 0; 2720 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2721 return true; 2722 2723 getStreamer().EmitCFIDefCfaRegister(Register); 2724 return false; 2725 } 2726 2727 /// ParseDirectiveCFIOffset 2728 /// ::= .cfi_offset register, offset 2729 bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) { 2730 int64_t Register = 0; 2731 int64_t Offset = 0; 2732 2733 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2734 return true; 2735 2736 if (getLexer().isNot(AsmToken::Comma)) 2737 return TokError("unexpected token in directive"); 2738 Lex(); 2739 2740 if (getParser().ParseAbsoluteExpression(Offset)) 2741 return true; 2742 2743 getStreamer().EmitCFIOffset(Register, Offset); 2744 return false; 2745 } 2746 2747 /// ParseDirectiveCFIRelOffset 2748 /// ::= .cfi_rel_offset register, offset 2749 bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef, 2750 SMLoc DirectiveLoc) { 2751 int64_t Register = 0; 2752 2753 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2754 return true; 2755 2756 if (getLexer().isNot(AsmToken::Comma)) 2757 return TokError("unexpected token in directive"); 2758 Lex(); 2759 2760 int64_t Offset = 0; 2761 if (getParser().ParseAbsoluteExpression(Offset)) 2762 return true; 2763 2764 getStreamer().EmitCFIRelOffset(Register, Offset); 2765 return false; 2766 } 2767 2768 static bool isValidEncoding(int64_t Encoding) { 2769 if (Encoding & ~0xff) 2770 return false; 2771 2772 if (Encoding == dwarf::DW_EH_PE_omit) 2773 return true; 2774 2775 const unsigned Format = Encoding & 0xf; 2776 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 2777 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 2778 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 2779 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 2780 return false; 2781 2782 const unsigned Application = Encoding & 0x70; 2783 if (Application != dwarf::DW_EH_PE_absptr && 2784 Application != dwarf::DW_EH_PE_pcrel) 2785 return false; 2786 2787 return true; 2788 } 2789 2790 /// ParseDirectiveCFIPersonalityOrLsda 2791 /// ::= .cfi_personality encoding, [symbol_name] 2792 /// ::= .cfi_lsda encoding, [symbol_name] 2793 bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal, 2794 SMLoc DirectiveLoc) { 2795 int64_t Encoding = 0; 2796 if (getParser().ParseAbsoluteExpression(Encoding)) 2797 return true; 2798 if (Encoding == dwarf::DW_EH_PE_omit) 2799 return false; 2800 2801 if (!isValidEncoding(Encoding)) 2802 return TokError("unsupported encoding."); 2803 2804 if (getLexer().isNot(AsmToken::Comma)) 2805 return TokError("unexpected token in directive"); 2806 Lex(); 2807 2808 StringRef Name; 2809 if (getParser().ParseIdentifier(Name)) 2810 return TokError("expected identifier in directive"); 2811 2812 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2813 2814 if (IDVal == ".cfi_personality") 2815 getStreamer().EmitCFIPersonality(Sym, Encoding); 2816 else { 2817 assert(IDVal == ".cfi_lsda"); 2818 getStreamer().EmitCFILsda(Sym, Encoding); 2819 } 2820 return false; 2821 } 2822 2823 /// ParseDirectiveCFIRememberState 2824 /// ::= .cfi_remember_state 2825 bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal, 2826 SMLoc DirectiveLoc) { 2827 getStreamer().EmitCFIRememberState(); 2828 return false; 2829 } 2830 2831 /// ParseDirectiveCFIRestoreState 2832 /// ::= .cfi_remember_state 2833 bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal, 2834 SMLoc DirectiveLoc) { 2835 getStreamer().EmitCFIRestoreState(); 2836 return false; 2837 } 2838 2839 /// ParseDirectiveCFISameValue 2840 /// ::= .cfi_same_value register 2841 bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal, 2842 SMLoc DirectiveLoc) { 2843 int64_t Register = 0; 2844 2845 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2846 return true; 2847 2848 getStreamer().EmitCFISameValue(Register); 2849 2850 return false; 2851 } 2852 2853 /// ParseDirectiveCFIRestore 2854 /// ::= .cfi_restore register 2855 bool GenericAsmParser::ParseDirectiveCFIRestore(StringRef IDVal, 2856 SMLoc DirectiveLoc) { 2857 int64_t Register = 0; 2858 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2859 return true; 2860 2861 getStreamer().EmitCFIRestore(Register); 2862 2863 return false; 2864 } 2865 2866 /// ParseDirectiveCFIEscape 2867 /// ::= .cfi_escape expression[,...] 2868 bool GenericAsmParser::ParseDirectiveCFIEscape(StringRef IDVal, 2869 SMLoc DirectiveLoc) { 2870 std::string Values; 2871 int64_t CurrValue; 2872 if (getParser().ParseAbsoluteExpression(CurrValue)) 2873 return true; 2874 2875 Values.push_back((uint8_t)CurrValue); 2876 2877 while (getLexer().is(AsmToken::Comma)) { 2878 Lex(); 2879 2880 if (getParser().ParseAbsoluteExpression(CurrValue)) 2881 return true; 2882 2883 Values.push_back((uint8_t)CurrValue); 2884 } 2885 2886 getStreamer().EmitCFIEscape(Values); 2887 return false; 2888 } 2889 2890 /// ParseDirectiveCFISignalFrame 2891 /// ::= .cfi_signal_frame 2892 bool GenericAsmParser::ParseDirectiveCFISignalFrame(StringRef Directive, 2893 SMLoc DirectiveLoc) { 2894 if (getLexer().isNot(AsmToken::EndOfStatement)) 2895 return Error(getLexer().getLoc(), 2896 "unexpected token in '" + Directive + "' directive"); 2897 2898 getStreamer().EmitCFISignalFrame(); 2899 2900 return false; 2901 } 2902 2903 /// ParseDirectiveMacrosOnOff 2904 /// ::= .macros_on 2905 /// ::= .macros_off 2906 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive, 2907 SMLoc DirectiveLoc) { 2908 if (getLexer().isNot(AsmToken::EndOfStatement)) 2909 return Error(getLexer().getLoc(), 2910 "unexpected token in '" + Directive + "' directive"); 2911 2912 getParser().MacrosEnabled = Directive == ".macros_on"; 2913 2914 return false; 2915 } 2916 2917 /// ParseDirectiveMacro 2918 /// ::= .macro name [parameters] 2919 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive, 2920 SMLoc DirectiveLoc) { 2921 StringRef Name; 2922 if (getParser().ParseIdentifier(Name)) 2923 return TokError("expected identifier in directive"); 2924 2925 std::vector<StringRef> Parameters; 2926 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2927 for(;;) { 2928 StringRef Parameter; 2929 if (getParser().ParseIdentifier(Parameter)) 2930 return TokError("expected identifier in directive"); 2931 Parameters.push_back(Parameter); 2932 2933 if (getLexer().isNot(AsmToken::Comma)) 2934 break; 2935 Lex(); 2936 } 2937 } 2938 2939 if (getLexer().isNot(AsmToken::EndOfStatement)) 2940 return TokError("unexpected token in '.macro' directive"); 2941 2942 // Eat the end of statement. 2943 Lex(); 2944 2945 AsmToken EndToken, StartToken = getTok(); 2946 2947 // Lex the macro definition. 2948 for (;;) { 2949 // Check whether we have reached the end of the file. 2950 if (getLexer().is(AsmToken::Eof)) 2951 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 2952 2953 // Otherwise, check whether we have reach the .endmacro. 2954 if (getLexer().is(AsmToken::Identifier) && 2955 (getTok().getIdentifier() == ".endm" || 2956 getTok().getIdentifier() == ".endmacro")) { 2957 EndToken = getTok(); 2958 Lex(); 2959 if (getLexer().isNot(AsmToken::EndOfStatement)) 2960 return TokError("unexpected token in '" + EndToken.getIdentifier() + 2961 "' directive"); 2962 break; 2963 } 2964 2965 // Otherwise, scan til the end of the statement. 2966 getParser().EatToEndOfStatement(); 2967 } 2968 2969 if (getParser().MacroMap.lookup(Name)) { 2970 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 2971 } 2972 2973 const char *BodyStart = StartToken.getLoc().getPointer(); 2974 const char *BodyEnd = EndToken.getLoc().getPointer(); 2975 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 2976 getParser().MacroMap[Name] = new Macro(Name, Body, Parameters); 2977 return false; 2978 } 2979 2980 /// ParseDirectiveEndMacro 2981 /// ::= .endm 2982 /// ::= .endmacro 2983 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive, 2984 SMLoc DirectiveLoc) { 2985 if (getLexer().isNot(AsmToken::EndOfStatement)) 2986 return TokError("unexpected token in '" + Directive + "' directive"); 2987 2988 // If we are inside a macro instantiation, terminate the current 2989 // instantiation. 2990 if (!getParser().ActiveMacros.empty()) { 2991 getParser().HandleMacroExit(); 2992 return false; 2993 } 2994 2995 // Otherwise, this .endmacro is a stray entry in the file; well formed 2996 // .endmacro directives are handled during the macro definition parsing. 2997 return TokError("unexpected '" + Directive + "' in file, " 2998 "no current macro definition"); 2999 } 3000 3001 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) { 3002 getParser().CheckForValidSection(); 3003 3004 const MCExpr *Value; 3005 3006 if (getParser().ParseExpression(Value)) 3007 return true; 3008 3009 if (getLexer().isNot(AsmToken::EndOfStatement)) 3010 return TokError("unexpected token in directive"); 3011 3012 if (DirName[1] == 's') 3013 getStreamer().EmitSLEB128Value(Value); 3014 else 3015 getStreamer().EmitULEB128Value(Value); 3016 3017 return false; 3018 } 3019 3020 3021 /// \brief Create an MCAsmParser instance. 3022 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, 3023 MCContext &C, MCStreamer &Out, 3024 const MCAsmInfo &MAI) { 3025 return new AsmParser(SM, C, Out, MAI); 3026 } 3027