1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This class implements the parser for assembly files. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "llvm/ADT/APFloat.h" 14 #include "llvm/ADT/APInt.h" 15 #include "llvm/ADT/ArrayRef.h" 16 #include "llvm/ADT/None.h" 17 #include "llvm/ADT/STLExtras.h" 18 #include "llvm/ADT/SmallString.h" 19 #include "llvm/ADT/SmallVector.h" 20 #include "llvm/ADT/StringExtras.h" 21 #include "llvm/ADT/StringMap.h" 22 #include "llvm/ADT/StringRef.h" 23 #include "llvm/ADT/Twine.h" 24 #include "llvm/BinaryFormat/Dwarf.h" 25 #include "llvm/DebugInfo/CodeView/SymbolRecord.h" 26 #include "llvm/MC/MCAsmInfo.h" 27 #include "llvm/MC/MCCodeView.h" 28 #include "llvm/MC/MCContext.h" 29 #include "llvm/MC/MCDirectives.h" 30 #include "llvm/MC/MCDwarf.h" 31 #include "llvm/MC/MCExpr.h" 32 #include "llvm/MC/MCInstPrinter.h" 33 #include "llvm/MC/MCInstrDesc.h" 34 #include "llvm/MC/MCInstrInfo.h" 35 #include "llvm/MC/MCObjectFileInfo.h" 36 #include "llvm/MC/MCParser/AsmCond.h" 37 #include "llvm/MC/MCParser/AsmLexer.h" 38 #include "llvm/MC/MCParser/MCAsmLexer.h" 39 #include "llvm/MC/MCParser/MCAsmParser.h" 40 #include "llvm/MC/MCParser/MCAsmParserExtension.h" 41 #include "llvm/MC/MCParser/MCAsmParserUtils.h" 42 #include "llvm/MC/MCParser/MCParsedAsmOperand.h" 43 #include "llvm/MC/MCParser/MCTargetAsmParser.h" 44 #include "llvm/MC/MCRegisterInfo.h" 45 #include "llvm/MC/MCSection.h" 46 #include "llvm/MC/MCStreamer.h" 47 #include "llvm/MC/MCSymbol.h" 48 #include "llvm/MC/MCTargetOptions.h" 49 #include "llvm/MC/MCValue.h" 50 #include "llvm/Support/Casting.h" 51 #include "llvm/Support/CommandLine.h" 52 #include "llvm/Support/ErrorHandling.h" 53 #include "llvm/Support/MD5.h" 54 #include "llvm/Support/MathExtras.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 #include "llvm/Support/SMLoc.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/raw_ostream.h" 59 #include <algorithm> 60 #include <cassert> 61 #include <cctype> 62 #include <climits> 63 #include <cstddef> 64 #include <cstdint> 65 #include <deque> 66 #include <memory> 67 #include <sstream> 68 #include <string> 69 #include <tuple> 70 #include <utility> 71 #include <vector> 72 73 using namespace llvm; 74 75 extern cl::opt<unsigned> AsmMacroMaxNestingDepth; 76 77 namespace { 78 79 /// Helper types for tracking macro definitions. 80 typedef std::vector<AsmToken> MCAsmMacroArgument; 81 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; 82 83 /// Helper class for storing information about an active macro instantiation. 84 struct MacroInstantiation { 85 /// The location of the instantiation. 86 SMLoc InstantiationLoc; 87 88 /// The buffer where parsing should resume upon instantiation completion. 89 unsigned ExitBuffer; 90 91 /// The location where parsing should resume upon instantiation completion. 92 SMLoc ExitLoc; 93 94 /// The depth of TheCondStack at the start of the instantiation. 95 size_t CondStackDepth; 96 }; 97 98 struct ParseStatementInfo { 99 /// The parsed operands from the last parsed statement. 100 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; 101 102 /// The opcode from the last parsed instruction. 103 unsigned Opcode = ~0U; 104 105 /// Was there an error parsing the inline assembly? 106 bool ParseError = false; 107 108 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr; 109 110 ParseStatementInfo() = delete; 111 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites) 112 : AsmRewrites(rewrites) {} 113 }; 114 115 enum FieldType { 116 FT_INTEGRAL, // Initializer: integer expression, stored as an MCExpr. 117 FT_REAL, // Initializer: real number, stored as an APInt. 118 FT_STRUCT // Initializer: struct initializer, stored recursively. 119 }; 120 121 struct FieldInfo; 122 struct StructInfo { 123 StringRef Name; 124 bool IsUnion = false; 125 size_t Alignment = 0; 126 size_t Size = 0; 127 std::vector<FieldInfo> Fields; 128 StringMap<size_t> FieldsByName; 129 130 FieldInfo &addField(StringRef FieldName, FieldType FT, size_t FieldSize); 131 132 StructInfo() = default; 133 134 StructInfo(StringRef StructName, bool Union, unsigned AlignmentValue) 135 : Name(StructName), IsUnion(Union), Alignment(AlignmentValue) {} 136 }; 137 138 // FIXME: This should probably use a class hierarchy, raw pointers between the 139 // objects, and dynamic type resolution instead of a union. On the other hand, 140 // ownership then becomes much more complicated; the obvious thing would be to 141 // use BumpPtrAllocator, but the lack of a destructor makes that messy. 142 143 struct StructInitializer; 144 struct IntFieldInfo { 145 SmallVector<const MCExpr *, 1> Values; 146 147 IntFieldInfo() = default; 148 IntFieldInfo(const SmallVector<const MCExpr *, 1> &V) { Values = V; } 149 IntFieldInfo(SmallVector<const MCExpr *, 1> &&V) { Values = V; } 150 }; 151 struct RealFieldInfo { 152 SmallVector<APInt, 1> AsIntValues; 153 154 RealFieldInfo() = default; 155 RealFieldInfo(const SmallVector<APInt, 1> &V) { AsIntValues = V; } 156 RealFieldInfo(SmallVector<APInt, 1> &&V) { AsIntValues = V; } 157 }; 158 struct StructFieldInfo { 159 std::vector<StructInitializer> Initializers; 160 StructInfo Structure; 161 162 StructFieldInfo() = default; 163 StructFieldInfo(const std::vector<StructInitializer> &V, StructInfo S) { 164 Initializers = V; 165 Structure = S; 166 } 167 StructFieldInfo(std::vector<StructInitializer> &&V, StructInfo S) { 168 Initializers = V; 169 Structure = S; 170 } 171 }; 172 173 class FieldInitializer { 174 public: 175 FieldType FT; 176 union { 177 IntFieldInfo IntInfo; 178 RealFieldInfo RealInfo; 179 StructFieldInfo StructInfo; 180 }; 181 182 ~FieldInitializer() { 183 switch (FT) { 184 case FT_INTEGRAL: 185 IntInfo.~IntFieldInfo(); 186 break; 187 case FT_REAL: 188 RealInfo.~RealFieldInfo(); 189 break; 190 case FT_STRUCT: 191 StructInfo.~StructFieldInfo(); 192 break; 193 } 194 } 195 196 FieldInitializer(FieldType FT) : FT(FT) { 197 switch (FT) { 198 case FT_INTEGRAL: 199 new (&IntInfo) IntFieldInfo(); 200 break; 201 case FT_REAL: 202 new (&RealInfo) RealFieldInfo(); 203 break; 204 case FT_STRUCT: 205 new (&StructInfo) StructFieldInfo(); 206 break; 207 } 208 } 209 210 FieldInitializer(SmallVector<const MCExpr *, 1> &&Values) : FT(FT_INTEGRAL) { 211 new (&IntInfo) IntFieldInfo(Values); 212 } 213 214 FieldInitializer(SmallVector<APInt, 1> &&AsIntValues) : FT(FT_REAL) { 215 new (&RealInfo) RealFieldInfo(AsIntValues); 216 } 217 218 FieldInitializer(std::vector<StructInitializer> &&Initializers, 219 struct StructInfo Structure) 220 : FT(FT_STRUCT) { 221 new (&StructInfo) StructFieldInfo(Initializers, Structure); 222 } 223 224 FieldInitializer(const FieldInitializer &Initializer) : FT(Initializer.FT) { 225 switch (FT) { 226 case FT_INTEGRAL: 227 new (&IntInfo) IntFieldInfo(Initializer.IntInfo); 228 break; 229 case FT_REAL: 230 new (&RealInfo) RealFieldInfo(Initializer.RealInfo); 231 break; 232 case FT_STRUCT: 233 new (&StructInfo) StructFieldInfo(Initializer.StructInfo); 234 break; 235 } 236 } 237 238 FieldInitializer(FieldInitializer &&Initializer) : FT(Initializer.FT) { 239 switch (FT) { 240 case FT_INTEGRAL: 241 new (&IntInfo) IntFieldInfo(Initializer.IntInfo); 242 break; 243 case FT_REAL: 244 new (&RealInfo) RealFieldInfo(Initializer.RealInfo); 245 break; 246 case FT_STRUCT: 247 new (&StructInfo) StructFieldInfo(Initializer.StructInfo); 248 break; 249 } 250 } 251 252 FieldInitializer &operator=(const FieldInitializer &Initializer) { 253 if (FT != Initializer.FT) { 254 switch (FT) { 255 case FT_INTEGRAL: 256 IntInfo.~IntFieldInfo(); 257 break; 258 case FT_REAL: 259 RealInfo.~RealFieldInfo(); 260 break; 261 case FT_STRUCT: 262 StructInfo.~StructFieldInfo(); 263 break; 264 } 265 } 266 FT = Initializer.FT; 267 switch (FT) { 268 case FT_INTEGRAL: 269 IntInfo = Initializer.IntInfo; 270 break; 271 case FT_REAL: 272 RealInfo = Initializer.RealInfo; 273 break; 274 case FT_STRUCT: 275 StructInfo = Initializer.StructInfo; 276 break; 277 } 278 return *this; 279 } 280 281 FieldInitializer &operator=(FieldInitializer &&Initializer) { 282 if (FT != Initializer.FT) { 283 switch (FT) { 284 case FT_INTEGRAL: 285 IntInfo.~IntFieldInfo(); 286 break; 287 case FT_REAL: 288 RealInfo.~RealFieldInfo(); 289 break; 290 case FT_STRUCT: 291 StructInfo.~StructFieldInfo(); 292 break; 293 } 294 } 295 FT = Initializer.FT; 296 switch (FT) { 297 case FT_INTEGRAL: 298 IntInfo = Initializer.IntInfo; 299 break; 300 case FT_REAL: 301 RealInfo = Initializer.RealInfo; 302 break; 303 case FT_STRUCT: 304 StructInfo = Initializer.StructInfo; 305 break; 306 } 307 return *this; 308 } 309 }; 310 311 struct StructInitializer { 312 std::vector<FieldInitializer> FieldInitializers; 313 }; 314 315 struct FieldInfo { 316 // Offset of the field within the containing STRUCT. 317 size_t Offset = 0; 318 319 // Total size of the field (= LengthOf * Type). 320 size_t SizeOf = 0; 321 322 // Number of elements in the field (1 if scalar, >1 if an array). 323 size_t LengthOf = 0; 324 325 // Size of a single entry in this field, in bytes ("type" in MASM standards). 326 size_t Type = 0; 327 328 FieldInitializer Contents; 329 330 FieldInfo(FieldType FT) : Contents(FT) {} 331 }; 332 333 FieldInfo &StructInfo::addField(StringRef FieldName, FieldType FT, 334 size_t FieldSize) { 335 if (!FieldName.empty()) 336 FieldsByName[FieldName] = Fields.size(); 337 Fields.emplace_back(FT); 338 FieldInfo &Field = Fields.back(); 339 if (IsUnion) { 340 Field.Offset = 0; 341 } else { 342 Size = llvm::alignTo(Size, std::min(Alignment, FieldSize)); 343 Field.Offset = Size; 344 } 345 return Field; 346 } 347 348 /// The concrete assembly parser instance. 349 // Note that this is a full MCAsmParser, not an MCAsmParserExtension! 350 // It's a peer of AsmParser, not of COFFAsmParser, WasmAsmParser, etc. 351 class MasmParser : public MCAsmParser { 352 private: 353 AsmLexer Lexer; 354 MCContext &Ctx; 355 MCStreamer &Out; 356 const MCAsmInfo &MAI; 357 SourceMgr &SrcMgr; 358 SourceMgr::DiagHandlerTy SavedDiagHandler; 359 void *SavedDiagContext; 360 std::unique_ptr<MCAsmParserExtension> PlatformParser; 361 362 /// This is the current buffer index we're lexing from as managed by the 363 /// SourceMgr object. 364 unsigned CurBuffer; 365 366 AsmCond TheCondState; 367 std::vector<AsmCond> TheCondStack; 368 369 /// maps directive names to handler methods in parser 370 /// extensions. Extensions register themselves in this map by calling 371 /// addDirectiveHandler. 372 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap; 373 374 /// maps assembly-time variable names to variables. 375 struct Variable { 376 StringRef Name; 377 bool Redefinable = true; 378 bool IsText = false; 379 int64_t NumericValue = 0; 380 std::string TextValue; 381 }; 382 StringMap<Variable> Variables; 383 384 /// Stack of active struct definitions. 385 SmallVector<StructInfo, 1> StructInProgress; 386 387 /// Maps struct tags to struct definitions. 388 StringMap<StructInfo> Structs; 389 390 /// Maps data location names to user-defined types. 391 StringMap<const StructInfo *> KnownType; 392 393 /// Stack of active macro instantiations. 394 std::vector<MacroInstantiation*> ActiveMacros; 395 396 /// List of bodies of anonymous macros. 397 std::deque<MCAsmMacro> MacroLikeBodies; 398 399 /// Keeps track of how many .macro's have been instantiated. 400 unsigned NumOfMacroInstantiations; 401 402 /// The values from the last parsed cpp hash file line comment if any. 403 struct CppHashInfoTy { 404 StringRef Filename; 405 int64_t LineNumber; 406 SMLoc Loc; 407 unsigned Buf; 408 CppHashInfoTy() : Filename(), LineNumber(0), Loc(), Buf(0) {} 409 }; 410 CppHashInfoTy CppHashInfo; 411 412 /// The filename from the first cpp hash file line comment, if any. 413 StringRef FirstCppHashFilename; 414 415 /// List of forward directional labels for diagnosis at the end. 416 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; 417 418 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 419 /// Defaults to 1U, meaning Intel. 420 unsigned AssemblerDialect = 1U; 421 422 /// is Darwin compatibility enabled? 423 bool IsDarwin = false; 424 425 /// Are we parsing ms-style inline assembly? 426 bool ParsingMSInlineAsm = false; 427 428 /// Did we already inform the user about inconsistent MD5 usage? 429 bool ReportedInconsistentMD5 = false; 430 431 // Is alt macro mode enabled. 432 bool AltMacroMode = false; 433 434 // Current <...> expression depth. 435 unsigned AngleBracketDepth = 0U; 436 437 public: 438 MasmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 439 const MCAsmInfo &MAI, unsigned CB); 440 MasmParser(const MasmParser &) = delete; 441 MasmParser &operator=(const MasmParser &) = delete; 442 ~MasmParser() override; 443 444 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override; 445 446 void addDirectiveHandler(StringRef Directive, 447 ExtensionDirectiveHandler Handler) override { 448 ExtensionDirectiveMap[Directive] = Handler; 449 if (DirectiveKindMap.find(Directive) == DirectiveKindMap.end()) { 450 DirectiveKindMap[Directive] = DK_HANDLER_DIRECTIVE; 451 } 452 } 453 454 void addAliasForDirective(StringRef Directive, StringRef Alias) override { 455 DirectiveKindMap[Directive] = DirectiveKindMap[Alias]; 456 } 457 458 /// @name MCAsmParser Interface 459 /// { 460 461 SourceMgr &getSourceManager() override { return SrcMgr; } 462 MCAsmLexer &getLexer() override { return Lexer; } 463 MCContext &getContext() override { return Ctx; } 464 MCStreamer &getStreamer() override { return Out; } 465 466 CodeViewContext &getCVContext() { return Ctx.getCVContext(); } 467 468 unsigned getAssemblerDialect() override { 469 if (AssemblerDialect == ~0U) 470 return MAI.getAssemblerDialect(); 471 else 472 return AssemblerDialect; 473 } 474 void setAssemblerDialect(unsigned i) override { 475 AssemblerDialect = i; 476 } 477 478 void Note(SMLoc L, const Twine &Msg, SMRange Range = None) override; 479 bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) override; 480 bool printError(SMLoc L, const Twine &Msg, SMRange Range = None) override; 481 482 const AsmToken &Lex() override; 483 484 void setParsingMSInlineAsm(bool V) override { 485 ParsingMSInlineAsm = V; 486 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and 487 // hex integer literals. 488 Lexer.setLexMasmIntegers(V); 489 } 490 bool isParsingMSInlineAsm() override { return ParsingMSInlineAsm; } 491 492 bool isParsingMasm() const override { return true; } 493 494 bool lookUpField(StringRef Name, StringRef &Type, 495 unsigned &Offset) const override; 496 bool lookUpField(StringRef Base, StringRef Member, StringRef &Type, 497 unsigned &Offset) const override; 498 499 bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString, 500 unsigned &NumOutputs, unsigned &NumInputs, 501 SmallVectorImpl<std::pair<void *,bool>> &OpDecls, 502 SmallVectorImpl<std::string> &Constraints, 503 SmallVectorImpl<std::string> &Clobbers, 504 const MCInstrInfo *MII, const MCInstPrinter *IP, 505 MCAsmParserSemaCallback &SI) override; 506 507 bool parseExpression(const MCExpr *&Res); 508 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 509 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override; 510 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 511 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 512 SMLoc &EndLoc) override; 513 bool parseAbsoluteExpression(int64_t &Res) override; 514 515 /// Parse a floating point expression using the float \p Semantics 516 /// and set \p Res to the value. 517 bool parseRealValue(const fltSemantics &Semantics, APInt &Res); 518 519 /// Parse an identifier or string (as a quoted identifier) 520 /// and set \p Res to the identifier contents. 521 bool parseIdentifier(StringRef &Res) override; 522 void eatToEndOfStatement() override; 523 524 bool checkForValidSection() override; 525 526 /// } 527 528 private: 529 bool parseStatement(ParseStatementInfo &Info, 530 MCAsmParserSemaCallback *SI); 531 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); 532 bool parseCppHashLineFilenameComment(SMLoc L); 533 534 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body, 535 ArrayRef<MCAsmMacroParameter> Parameters); 536 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 537 ArrayRef<MCAsmMacroParameter> Parameters, 538 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable, 539 SMLoc L); 540 541 /// Are we inside a macro instantiation? 542 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();} 543 544 /// Handle entry to macro instantiation. 545 /// 546 /// \param M The macro. 547 /// \param NameLoc Instantiation location. 548 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc); 549 550 /// Handle exit from macro instantiation. 551 void handleMacroExit(); 552 553 /// Extract AsmTokens for a macro argument. 554 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg); 555 556 /// Parse all macro arguments for a given macro. 557 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A); 558 559 void printMacroInstantiations(); 560 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 561 SMRange Range = None) const { 562 ArrayRef<SMRange> Ranges(Range); 563 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 564 } 565 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 566 567 bool lookUpField(const StructInfo &Structure, StringRef Member, 568 StringRef &Type, unsigned &Offset) const; 569 570 /// Should we emit DWARF describing this assembler source? (Returns false if 571 /// the source has .file directives, which means we don't want to generate 572 /// info describing the assembler source itself.) 573 bool enabledGenDwarfForAssembly(); 574 575 /// Enter the specified file. This returns true on failure. 576 bool enterIncludeFile(const std::string &Filename); 577 578 /// Reset the current lexer position to that given by \p Loc. The 579 /// current token is not set; clients should ensure Lex() is called 580 /// subsequently. 581 /// 582 /// \param InBuffer If not 0, should be the known buffer id that contains the 583 /// location. 584 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0); 585 586 /// Parse up to the end of statement and a return the contents from the 587 /// current token until the end of the statement; the current token on exit 588 /// will be either the EndOfStatement or EOF. 589 StringRef parseStringToEndOfStatement() override; 590 591 bool parseTextItem(std::string &Data); 592 593 unsigned getBinOpPrecedence(AsmToken::TokenKind K, 594 MCBinaryExpr::Opcode &Kind); 595 596 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 597 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 598 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 599 600 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 601 602 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName); 603 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName); 604 605 // Generic (target and platform independent) directive parsing. 606 enum DirectiveKind { 607 DK_NO_DIRECTIVE, // Placeholder 608 DK_HANDLER_DIRECTIVE, 609 DK_ASSIGN, 610 DK_EQU, 611 DK_TEXTEQU, 612 DK_ASCII, 613 DK_ASCIZ, 614 DK_STRING, 615 DK_BYTE, 616 DK_SBYTE, 617 DK_WORD, 618 DK_SWORD, 619 DK_DWORD, 620 DK_SDWORD, 621 DK_FWORD, 622 DK_QWORD, 623 DK_SQWORD, 624 DK_DB, 625 DK_DD, 626 DK_DQ, 627 DK_DW, 628 DK_REAL4, 629 DK_REAL8, 630 DK_ALIGN, 631 DK_ORG, 632 DK_ENDR, 633 DK_EXTERN, 634 DK_PUBLIC, 635 DK_COMM, 636 DK_COMMENT, 637 DK_INCLUDE, 638 DK_REPT, 639 DK_IRP, 640 DK_IRPC, 641 DK_IF, 642 DK_IFE, 643 DK_IFB, 644 DK_IFNB, 645 DK_IFDEF, 646 DK_IFNDEF, 647 DK_IFDIF, 648 DK_IFDIFI, 649 DK_IFIDN, 650 DK_IFIDNI, 651 DK_ELSEIF, 652 DK_ELSEIFE, 653 DK_ELSEIFB, 654 DK_ELSEIFNB, 655 DK_ELSEIFDEF, 656 DK_ELSEIFNDEF, 657 DK_ELSEIFDIF, 658 DK_ELSEIFDIFI, 659 DK_ELSEIFIDN, 660 DK_ELSEIFIDNI, 661 DK_ELSE, 662 DK_ENDIF, 663 DK_FILE, 664 DK_LINE, 665 DK_LOC, 666 DK_STABS, 667 DK_CV_FILE, 668 DK_CV_FUNC_ID, 669 DK_CV_INLINE_SITE_ID, 670 DK_CV_LOC, 671 DK_CV_LINETABLE, 672 DK_CV_INLINE_LINETABLE, 673 DK_CV_DEF_RANGE, 674 DK_CV_STRINGTABLE, 675 DK_CV_STRING, 676 DK_CV_FILECHECKSUMS, 677 DK_CV_FILECHECKSUM_OFFSET, 678 DK_CV_FPO_DATA, 679 DK_CFI_SECTIONS, 680 DK_CFI_STARTPROC, 681 DK_CFI_ENDPROC, 682 DK_CFI_DEF_CFA, 683 DK_CFI_DEF_CFA_OFFSET, 684 DK_CFI_ADJUST_CFA_OFFSET, 685 DK_CFI_DEF_CFA_REGISTER, 686 DK_CFI_OFFSET, 687 DK_CFI_REL_OFFSET, 688 DK_CFI_PERSONALITY, 689 DK_CFI_LSDA, 690 DK_CFI_REMEMBER_STATE, 691 DK_CFI_RESTORE_STATE, 692 DK_CFI_SAME_VALUE, 693 DK_CFI_RESTORE, 694 DK_CFI_ESCAPE, 695 DK_CFI_RETURN_COLUMN, 696 DK_CFI_SIGNAL_FRAME, 697 DK_CFI_UNDEFINED, 698 DK_CFI_REGISTER, 699 DK_CFI_WINDOW_SAVE, 700 DK_CFI_B_KEY_FRAME, 701 DK_ALTMACRO, 702 DK_NOALTMACRO, 703 DK_MACRO, 704 DK_EXITM, 705 DK_ENDM, 706 DK_PURGEM, 707 DK_ERR, 708 DK_ERRB, 709 DK_ERRNB, 710 DK_ERRDEF, 711 DK_ERRNDEF, 712 DK_ERRDIF, 713 DK_ERRDIFI, 714 DK_ERRIDN, 715 DK_ERRIDNI, 716 DK_ERRE, 717 DK_ERRNZ, 718 DK_ECHO, 719 DK_STRUCT, 720 DK_UNION, 721 DK_ENDS, 722 DK_END 723 }; 724 725 /// Maps directive name --> DirectiveKind enum, for directives parsed by this 726 /// class. 727 StringMap<DirectiveKind> DirectiveKindMap; 728 729 // Codeview def_range type parsing. 730 enum CVDefRangeType { 731 CVDR_DEFRANGE = 0, // Placeholder 732 CVDR_DEFRANGE_REGISTER, 733 CVDR_DEFRANGE_FRAMEPOINTER_REL, 734 CVDR_DEFRANGE_SUBFIELD_REGISTER, 735 CVDR_DEFRANGE_REGISTER_REL 736 }; 737 738 /// Maps Codeview def_range types --> CVDefRangeType enum, for Codeview 739 /// def_range types parsed by this class. 740 StringMap<CVDefRangeType> CVDefRangeTypeMap; 741 742 bool parseInitValue(unsigned Size); 743 744 // ".ascii", ".asciz", ".string" 745 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 746 747 // "byte", "word", ... 748 bool emitIntValue(const MCExpr *Value, unsigned Size); 749 bool parseScalarInitializer(unsigned Size, 750 SmallVectorImpl<const MCExpr *> &Values, 751 unsigned StringPadLength = 0); 752 bool parseScalarInstList( 753 unsigned Size, SmallVectorImpl<const MCExpr *> &Values, 754 const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); 755 bool emitIntegralValues(unsigned Size); 756 bool addIntegralField(StringRef Name, unsigned Size); 757 bool parseDirectiveValue(StringRef IDVal, unsigned Size); 758 bool parseDirectiveNamedValue(StringRef IDVal, unsigned Size, StringRef Name, 759 SMLoc NameLoc); 760 761 // "real4", "real8" 762 bool emitRealValues(const fltSemantics &Semantics); 763 bool addRealField(StringRef Name, const fltSemantics &Semantics, size_t Size); 764 bool parseDirectiveRealValue(StringRef IDVal, const fltSemantics &Semantics, 765 size_t Size); 766 bool parseRealInstList( 767 const fltSemantics &Semantics, SmallVectorImpl<APInt> &Values, 768 const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); 769 bool parseDirectiveNamedRealValue(StringRef IDVal, 770 const fltSemantics &Semantics, size_t Size, 771 StringRef Name, SMLoc NameLoc); 772 773 bool parseOptionalAngleBracketOpen(); 774 bool parseAngleBracketClose(const Twine &Msg = "expected '>'"); 775 776 bool parseFieldInitializer(const FieldInfo &Field, 777 FieldInitializer &Initializer); 778 bool parseFieldInitializer(const FieldInfo &Field, 779 const IntFieldInfo &Contents, 780 FieldInitializer &Initializer); 781 bool parseFieldInitializer(const FieldInfo &Field, 782 const RealFieldInfo &Contents, 783 FieldInitializer &Initializer); 784 bool parseFieldInitializer(const FieldInfo &Field, 785 const StructFieldInfo &Contents, 786 FieldInitializer &Initializer); 787 788 bool parseStructInitializer(const StructInfo &Structure, 789 StructInitializer &Initializer); 790 bool parseStructInstList( 791 const StructInfo &Structure, std::vector<StructInitializer> &Initializers, 792 const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); 793 794 bool emitFieldValue(const FieldInfo &Field); 795 bool emitFieldValue(const FieldInfo &Field, const IntFieldInfo &Contents); 796 bool emitFieldValue(const FieldInfo &Field, const RealFieldInfo &Contents); 797 bool emitFieldValue(const FieldInfo &Field, const StructFieldInfo &Contents); 798 799 bool emitFieldInitializer(const FieldInfo &Field, 800 const FieldInitializer &Initializer); 801 bool emitFieldInitializer(const FieldInfo &Field, 802 const IntFieldInfo &Contents, 803 const IntFieldInfo &Initializer); 804 bool emitFieldInitializer(const FieldInfo &Field, 805 const RealFieldInfo &Contents, 806 const RealFieldInfo &Initializer); 807 bool emitFieldInitializer(const FieldInfo &Field, 808 const StructFieldInfo &Contents, 809 const StructFieldInfo &Initializer); 810 811 bool emitStructInitializer(const StructInfo &Structure, 812 const StructInitializer &Initializer); 813 814 // User-defined types (structs, unions): 815 bool emitStructValues(const StructInfo &Structure); 816 bool addStructField(StringRef Name, const StructInfo &Structure); 817 bool parseDirectiveStructValue(const StructInfo &Structure, 818 StringRef Directive, SMLoc DirLoc); 819 bool parseDirectiveNamedStructValue(const StructInfo &Structure, 820 StringRef Directive, SMLoc DirLoc, 821 StringRef Name); 822 823 // "=", "equ", "textequ" 824 bool parseDirectiveEquate(StringRef IDVal, StringRef Name, 825 DirectiveKind DirKind); 826 827 bool parseDirectiveOrg(); // ".org" 828 bool parseDirectiveAlign(); // "align" 829 830 // ".file", ".line", ".loc", ".stabs" 831 bool parseDirectiveFile(SMLoc DirectiveLoc); 832 bool parseDirectiveLine(); 833 bool parseDirectiveLoc(); 834 bool parseDirectiveStabs(); 835 836 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable", 837 // ".cv_inline_linetable", ".cv_def_range", ".cv_string" 838 bool parseDirectiveCVFile(); 839 bool parseDirectiveCVFuncId(); 840 bool parseDirectiveCVInlineSiteId(); 841 bool parseDirectiveCVLoc(); 842 bool parseDirectiveCVLinetable(); 843 bool parseDirectiveCVInlineLinetable(); 844 bool parseDirectiveCVDefRange(); 845 bool parseDirectiveCVString(); 846 bool parseDirectiveCVStringTable(); 847 bool parseDirectiveCVFileChecksums(); 848 bool parseDirectiveCVFileChecksumOffset(); 849 bool parseDirectiveCVFPOData(); 850 851 // .cfi directives 852 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc); 853 bool parseDirectiveCFIWindowSave(); 854 bool parseDirectiveCFISections(); 855 bool parseDirectiveCFIStartProc(); 856 bool parseDirectiveCFIEndProc(); 857 bool parseDirectiveCFIDefCfaOffset(); 858 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc); 859 bool parseDirectiveCFIAdjustCfaOffset(); 860 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc); 861 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc); 862 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc); 863 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality); 864 bool parseDirectiveCFIRememberState(); 865 bool parseDirectiveCFIRestoreState(); 866 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc); 867 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc); 868 bool parseDirectiveCFIEscape(); 869 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc); 870 bool parseDirectiveCFISignalFrame(); 871 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc); 872 873 // macro directives 874 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc); 875 bool parseDirectiveExitMacro(StringRef Directive); 876 bool parseDirectiveEndMacro(StringRef Directive); 877 bool parseDirectiveMacro(SMLoc DirectiveLoc); 878 // alternate macro mode directives 879 bool parseDirectiveAltmacro(StringRef Directive); 880 881 bool parseDirectiveStruct(StringRef Directive, DirectiveKind DirKind, 882 StringRef Name, SMLoc NameLoc); 883 bool parseDirectiveNestedStruct(StringRef Directive, DirectiveKind DirKind); 884 bool parseDirectiveEnds(StringRef Name, SMLoc NameLoc); 885 bool parseDirectiveNestedEnds(); 886 887 /// Parse a directive like ".globl" which accepts a single symbol (which 888 /// should be a label or an external). 889 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr); 890 891 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 892 893 bool parseDirectiveComment(SMLoc DirectiveLoc); // "comment" 894 895 bool parseDirectiveInclude(); // "include" 896 897 // "if" or "ife" 898 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 899 // "ifb" or "ifnb", depending on ExpectBlank. 900 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 901 // "ifidn", "ifdif", "ifidni", or "ifdifi", depending on ExpectEqual and 902 // CaseInsensitive. 903 bool parseDirectiveIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 904 bool CaseInsensitive); 905 // "ifdef" or "ifndef", depending on expect_defined 906 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 907 // "elseif" or "elseife" 908 bool parseDirectiveElseIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 909 // "elseifb" or "elseifnb", depending on ExpectBlank. 910 bool parseDirectiveElseIfb(SMLoc DirectiveLoc, bool ExpectBlank); 911 // ".elseifdef" or ".elseifndef", depending on expect_defined 912 bool parseDirectiveElseIfdef(SMLoc DirectiveLoc, bool expect_defined); 913 // "elseifidn", "elseifdif", "elseifidni", or "elseifdifi", depending on 914 // ExpectEqual and CaseInsensitive. 915 bool parseDirectiveElseIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 916 bool CaseInsensitive); 917 bool parseDirectiveElse(SMLoc DirectiveLoc); // "else" 918 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // "endif" 919 bool parseEscapedString(std::string &Data) override; 920 bool parseAngleBracketString(std::string &Data) override; 921 922 // Macro-like directives 923 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc); 924 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 925 raw_svector_ostream &OS); 926 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive); 927 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 928 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc" 929 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 930 931 // "_emit" or "__emit" 932 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info, 933 size_t Len); 934 935 // "align" 936 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info); 937 938 // "end" 939 bool parseDirectiveEnd(SMLoc DirectiveLoc); 940 941 // ".err" 942 bool parseDirectiveError(SMLoc DirectiveLoc); 943 // ".errb" or ".errnb", depending on ExpectBlank. 944 bool parseDirectiveErrorIfb(SMLoc DirectiveLoc, bool ExpectBlank); 945 // ".errdef" or ".errndef", depending on ExpectBlank. 946 bool parseDirectiveErrorIfdef(SMLoc DirectiveLoc, bool ExpectDefined); 947 // ".erridn", ".errdif", ".erridni", or ".errdifi", depending on ExpectEqual 948 // and CaseInsensitive. 949 bool parseDirectiveErrorIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 950 bool CaseInsensitive); 951 // ".erre" or ".errnz", depending on ExpectZero. 952 bool parseDirectiveErrorIfe(SMLoc DirectiveLoc, bool ExpectZero); 953 954 // "echo" 955 bool parseDirectiveEcho(); 956 957 void initializeDirectiveKindMap(); 958 void initializeCVDefRangeTypeMap(); 959 }; 960 961 } // end anonymous namespace 962 963 namespace llvm { 964 965 extern MCAsmParserExtension *createCOFFMasmParser(); 966 967 } // end namespace llvm 968 969 enum { DEFAULT_ADDRSPACE = 0 }; 970 971 MasmParser::MasmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 972 const MCAsmInfo &MAI, unsigned CB = 0) 973 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM), 974 CurBuffer(CB ? CB : SM.getMainFileID()) { 975 HadError = false; 976 // Save the old handler. 977 SavedDiagHandler = SrcMgr.getDiagHandler(); 978 SavedDiagContext = SrcMgr.getDiagContext(); 979 // Set our own handler which calls the saved handler. 980 SrcMgr.setDiagHandler(DiagHandler, this); 981 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 982 983 // Initialize the platform / file format parser. 984 switch (Ctx.getObjectFileInfo()->getObjectFileType()) { 985 case MCObjectFileInfo::IsCOFF: 986 PlatformParser.reset(createCOFFMasmParser()); 987 break; 988 default: 989 report_fatal_error("llvm-ml currently supports only COFF output."); 990 break; 991 } 992 993 initializeDirectiveKindMap(); 994 PlatformParser->Initialize(*this); 995 initializeCVDefRangeTypeMap(); 996 997 NumOfMacroInstantiations = 0; 998 } 999 1000 MasmParser::~MasmParser() { 1001 assert((HadError || ActiveMacros.empty()) && 1002 "Unexpected active macro instantiation!"); 1003 1004 // Restore the saved diagnostics handler and context for use during 1005 // finalization. 1006 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext); 1007 } 1008 1009 void MasmParser::printMacroInstantiations() { 1010 // Print the active macro instantiation stack. 1011 for (std::vector<MacroInstantiation *>::const_reverse_iterator 1012 it = ActiveMacros.rbegin(), 1013 ie = ActiveMacros.rend(); 1014 it != ie; ++it) 1015 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 1016 "while in macro instantiation"); 1017 } 1018 1019 void MasmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { 1020 printPendingErrors(); 1021 printMessage(L, SourceMgr::DK_Note, Msg, Range); 1022 printMacroInstantiations(); 1023 } 1024 1025 bool MasmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) { 1026 if (getTargetParser().getTargetOptions().MCNoWarn) 1027 return false; 1028 if (getTargetParser().getTargetOptions().MCFatalWarnings) 1029 return Error(L, Msg, Range); 1030 printMessage(L, SourceMgr::DK_Warning, Msg, Range); 1031 printMacroInstantiations(); 1032 return false; 1033 } 1034 1035 bool MasmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { 1036 HadError = true; 1037 printMessage(L, SourceMgr::DK_Error, Msg, Range); 1038 printMacroInstantiations(); 1039 return true; 1040 } 1041 1042 bool MasmParser::enterIncludeFile(const std::string &Filename) { 1043 std::string IncludedFile; 1044 unsigned NewBuf = 1045 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 1046 if (!NewBuf) 1047 return true; 1048 1049 CurBuffer = NewBuf; 1050 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 1051 return false; 1052 } 1053 1054 void MasmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) { 1055 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); 1056 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), 1057 Loc.getPointer()); 1058 } 1059 1060 const AsmToken &MasmParser::Lex() { 1061 if (Lexer.getTok().is(AsmToken::Error)) 1062 Error(Lexer.getErrLoc(), Lexer.getErr()); 1063 1064 // if it's a end of statement with a comment in it 1065 if (getTok().is(AsmToken::EndOfStatement)) { 1066 // if this is a line comment output it. 1067 if (!getTok().getString().empty() && getTok().getString().front() != '\n' && 1068 getTok().getString().front() != '\r' && MAI.preserveAsmComments()) 1069 Out.addExplicitComment(Twine(getTok().getString())); 1070 } 1071 1072 const AsmToken *tok = &Lexer.Lex(); 1073 1074 while (tok->is(AsmToken::Identifier)) { 1075 auto it = Variables.find(tok->getIdentifier()); 1076 if (it != Variables.end() && it->second.IsText) { 1077 std::unique_ptr<MemoryBuffer> Instantiation = 1078 MemoryBuffer::getMemBufferCopy(it->second.TextValue, 1079 "<instantiation>"); 1080 1081 // Jump to the macro instantiation and prime the lexer. 1082 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), 1083 getTok().getEndLoc()); 1084 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), nullptr, 1085 /*EndStatementAtEOF=*/false); 1086 tok = &Lexer.Lex(); 1087 } else { 1088 break; 1089 } 1090 } 1091 1092 // Parse comments here to be deferred until end of next statement. 1093 while (tok->is(AsmToken::Comment)) { 1094 if (MAI.preserveAsmComments()) 1095 Out.addExplicitComment(Twine(tok->getString())); 1096 tok = &Lexer.Lex(); 1097 } 1098 1099 // Recognize and bypass line continuations. 1100 while (tok->is(AsmToken::BackSlash) && 1101 Lexer.peekTok().is(AsmToken::EndOfStatement)) { 1102 // Eat both the backslash and the end of statement. 1103 Lexer.Lex(); 1104 tok = &Lexer.Lex(); 1105 } 1106 1107 if (tok->is(AsmToken::Eof)) { 1108 // If this is the end of an included file, pop the parent file off the 1109 // include stack. 1110 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 1111 if (ParentIncludeLoc != SMLoc()) { 1112 jumpToLoc(ParentIncludeLoc); 1113 return Lex(); 1114 } 1115 } 1116 1117 return *tok; 1118 } 1119 1120 bool MasmParser::enabledGenDwarfForAssembly() { 1121 // Check whether the user specified -g. 1122 if (!getContext().getGenDwarfForAssembly()) 1123 return false; 1124 // If we haven't encountered any .file directives (which would imply that 1125 // the assembler source was produced with debug info already) then emit one 1126 // describing the assembler source file itself. 1127 if (getContext().getGenDwarfFileNumber() == 0) { 1128 // Use the first #line directive for this, if any. It's preprocessed, so 1129 // there is no checksum, and of course no source directive. 1130 if (!FirstCppHashFilename.empty()) 1131 getContext().setMCLineTableRootFile(/*CUID=*/0, 1132 getContext().getCompilationDir(), 1133 FirstCppHashFilename, 1134 /*Cksum=*/None, /*Source=*/None); 1135 const MCDwarfFile &RootFile = 1136 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); 1137 getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective( 1138 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name, 1139 RootFile.Checksum, RootFile.Source)); 1140 } 1141 return true; 1142 } 1143 1144 bool MasmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 1145 // Create the initial section, if requested. 1146 if (!NoInitialTextSection) 1147 Out.InitSections(false); 1148 1149 // Prime the lexer. 1150 Lex(); 1151 1152 HadError = false; 1153 AsmCond StartingCondState = TheCondState; 1154 SmallVector<AsmRewrite, 4> AsmStrRewrites; 1155 1156 // If we are generating dwarf for assembly source files save the initial text 1157 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't 1158 // emitting any actual debug info yet and haven't had a chance to parse any 1159 // embedded .file directives.) 1160 if (getContext().getGenDwarfForAssembly()) { 1161 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 1162 if (!Sec->getBeginSymbol()) { 1163 MCSymbol *SectionStartSym = getContext().createTempSymbol(); 1164 getStreamer().emitLabel(SectionStartSym); 1165 Sec->setBeginSymbol(SectionStartSym); 1166 } 1167 bool InsertResult = getContext().addGenDwarfSection(Sec); 1168 assert(InsertResult && ".text section should not have debug info yet"); 1169 (void)InsertResult; 1170 } 1171 1172 // While we have input, parse each statement. 1173 while (Lexer.isNot(AsmToken::Eof)) { 1174 ParseStatementInfo Info(&AsmStrRewrites); 1175 bool Parsed = parseStatement(Info, nullptr); 1176 1177 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error 1178 // for printing ErrMsg via Lex() only if no (presumably better) parser error 1179 // exists. 1180 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) { 1181 Lex(); 1182 } 1183 1184 // parseStatement returned true so may need to emit an error. 1185 printPendingErrors(); 1186 1187 // Skipping to the next line if needed. 1188 if (Parsed && !getLexer().isAtStartOfStatement()) 1189 eatToEndOfStatement(); 1190 } 1191 1192 getTargetParser().onEndOfFile(); 1193 printPendingErrors(); 1194 1195 // All errors should have been emitted. 1196 assert(!hasPendingError() && "unexpected error from parseStatement"); 1197 1198 getTargetParser().flushPendingInstructions(getStreamer()); 1199 1200 if (TheCondState.TheCond != StartingCondState.TheCond || 1201 TheCondState.Ignore != StartingCondState.Ignore) 1202 printError(getTok().getLoc(), "unmatched .ifs or .elses"); 1203 // Check to see there are no empty DwarfFile slots. 1204 const auto &LineTables = getContext().getMCDwarfLineTables(); 1205 if (!LineTables.empty()) { 1206 unsigned Index = 0; 1207 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) { 1208 if (File.Name.empty() && Index != 0) 1209 printError(getTok().getLoc(), "unassigned file number: " + 1210 Twine(Index) + 1211 " for .file directives"); 1212 ++Index; 1213 } 1214 } 1215 1216 // Check to see that all assembler local symbols were actually defined. 1217 // Targets that don't do subsections via symbols may not want this, though, 1218 // so conservatively exclude them. Only do this if we're finalizing, though, 1219 // as otherwise we won't necessarilly have seen everything yet. 1220 if (!NoFinalize) { 1221 if (MAI.hasSubsectionsViaSymbols()) { 1222 for (const auto &TableEntry : getContext().getSymbols()) { 1223 MCSymbol *Sym = TableEntry.getValue(); 1224 // Variable symbols may not be marked as defined, so check those 1225 // explicitly. If we know it's a variable, we have a definition for 1226 // the purposes of this check. 1227 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 1228 // FIXME: We would really like to refer back to where the symbol was 1229 // first referenced for a source location. We need to add something 1230 // to track that. Currently, we just point to the end of the file. 1231 printError(getTok().getLoc(), "assembler local symbol '" + 1232 Sym->getName() + "' not defined"); 1233 } 1234 } 1235 1236 // Temporary symbols like the ones for directional jumps don't go in the 1237 // symbol table. They also need to be diagnosed in all (final) cases. 1238 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { 1239 if (std::get<2>(LocSym)->isUndefined()) { 1240 // Reset the state of any "# line file" directives we've seen to the 1241 // context as it was at the diagnostic site. 1242 CppHashInfo = std::get<1>(LocSym); 1243 printError(std::get<0>(LocSym), "directional label undefined"); 1244 } 1245 } 1246 } 1247 1248 // Finalize the output stream if there are no errors and if the client wants 1249 // us to. 1250 if (!HadError && !NoFinalize) 1251 Out.Finish(); 1252 1253 return HadError || getContext().hadError(); 1254 } 1255 1256 bool MasmParser::checkForValidSection() { 1257 if (!ParsingMSInlineAsm && !getStreamer().getCurrentSectionOnly()) { 1258 Out.InitSections(false); 1259 return Error(getTok().getLoc(), 1260 "expected section directive before assembly directive"); 1261 } 1262 return false; 1263 } 1264 1265 /// Throw away the rest of the line for testing purposes. 1266 void MasmParser::eatToEndOfStatement() { 1267 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1268 Lexer.Lex(); 1269 1270 // Eat EOL. 1271 if (Lexer.is(AsmToken::EndOfStatement)) 1272 Lexer.Lex(); 1273 } 1274 1275 StringRef MasmParser::parseStringToEndOfStatement() { 1276 const char *Start = getTok().getLoc().getPointer(); 1277 1278 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1279 Lexer.Lex(); 1280 1281 const char *End = getTok().getLoc().getPointer(); 1282 return StringRef(Start, End - Start); 1283 } 1284 1285 /// Parse a paren expression and return it. 1286 /// NOTE: This assumes the leading '(' has already been consumed. 1287 /// 1288 /// parenexpr ::= expr) 1289 /// 1290 bool MasmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1291 if (parseExpression(Res)) 1292 return true; 1293 if (Lexer.isNot(AsmToken::RParen)) 1294 return TokError("expected ')' in parentheses expression"); 1295 EndLoc = Lexer.getTok().getEndLoc(); 1296 Lex(); 1297 return false; 1298 } 1299 1300 /// Parse a bracket expression and return it. 1301 /// NOTE: This assumes the leading '[' has already been consumed. 1302 /// 1303 /// bracketexpr ::= expr] 1304 /// 1305 bool MasmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1306 if (parseExpression(Res)) 1307 return true; 1308 EndLoc = getTok().getEndLoc(); 1309 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression")) 1310 return true; 1311 return false; 1312 } 1313 1314 /// Parse a primary expression and return it. 1315 /// primaryexpr ::= (parenexpr 1316 /// primaryexpr ::= symbol 1317 /// primaryexpr ::= number 1318 /// primaryexpr ::= '.' 1319 /// primaryexpr ::= ~,+,-,'not' primaryexpr 1320 bool MasmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1321 SMLoc FirstTokenLoc = getLexer().getLoc(); 1322 AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); 1323 switch (FirstTokenKind) { 1324 default: 1325 return TokError("unknown token in expression"); 1326 // If we have an error assume that we've already handled it. 1327 case AsmToken::Error: 1328 return true; 1329 case AsmToken::Exclaim: 1330 Lex(); // Eat the operator. 1331 if (parsePrimaryExpr(Res, EndLoc)) 1332 return true; 1333 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc); 1334 return false; 1335 case AsmToken::Dollar: 1336 case AsmToken::At: 1337 case AsmToken::String: 1338 case AsmToken::Identifier: { 1339 StringRef Identifier; 1340 if (parseIdentifier(Identifier)) { 1341 // We may have failed but $ may be a valid token. 1342 if (getTok().is(AsmToken::Dollar)) { 1343 if (Lexer.getMAI().getDollarIsPC()) { 1344 Lex(); 1345 // This is a '$' reference, which references the current PC. Emit a 1346 // temporary label to the streamer and refer to it. 1347 MCSymbol *Sym = Ctx.createTempSymbol(); 1348 Out.emitLabel(Sym); 1349 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, 1350 getContext()); 1351 EndLoc = FirstTokenLoc; 1352 return false; 1353 } 1354 return Error(FirstTokenLoc, "invalid token in expression"); 1355 } 1356 } 1357 // Parse named bitwise negation. 1358 if (Identifier.equals_lower("not")) { 1359 if (parsePrimaryExpr(Res, EndLoc)) 1360 return true; 1361 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1362 return false; 1363 } 1364 // Parse symbol variant. 1365 std::pair<StringRef, StringRef> Split; 1366 if (!MAI.useParensForSymbolVariant()) { 1367 if (FirstTokenKind == AsmToken::String) { 1368 if (Lexer.is(AsmToken::At)) { 1369 Lex(); // eat @ 1370 SMLoc AtLoc = getLexer().getLoc(); 1371 StringRef VName; 1372 if (parseIdentifier(VName)) 1373 return Error(AtLoc, "expected symbol variant after '@'"); 1374 1375 Split = std::make_pair(Identifier, VName); 1376 } 1377 } else { 1378 Split = Identifier.split('@'); 1379 } 1380 } else if (Lexer.is(AsmToken::LParen)) { 1381 Lex(); // eat '('. 1382 StringRef VName; 1383 parseIdentifier(VName); 1384 // eat ')'. 1385 if (parseToken(AsmToken::RParen, 1386 "unexpected token in variant, expected ')'")) 1387 return true; 1388 Split = std::make_pair(Identifier, VName); 1389 } 1390 1391 EndLoc = SMLoc::getFromPointer(Identifier.end()); 1392 1393 // This is a symbol reference. 1394 StringRef SymbolName = Identifier; 1395 if (SymbolName.empty()) 1396 return Error(getLexer().getLoc(), "expected a symbol reference"); 1397 1398 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1399 1400 // Look up the symbol variant if used. 1401 if (!Split.second.empty()) { 1402 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1403 if (Variant != MCSymbolRefExpr::VK_Invalid) { 1404 SymbolName = Split.first; 1405 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { 1406 Variant = MCSymbolRefExpr::VK_None; 1407 } else { 1408 return Error(SMLoc::getFromPointer(Split.second.begin()), 1409 "invalid variant '" + Split.second + "'"); 1410 } 1411 } 1412 1413 // Find the field offset if used. 1414 StringRef Type; 1415 unsigned Offset = 0; 1416 Split = SymbolName.split('.'); 1417 if (!Split.second.empty()) { 1418 SymbolName = Split.first; 1419 if (Structs.count(SymbolName.lower()) && 1420 !lookUpField(SymbolName, Split.second, Type, Offset)) { 1421 // This is actually a reference to a field offset. 1422 Res = MCConstantExpr::create(Offset, getContext()); 1423 return false; 1424 } 1425 1426 auto TypeIt = KnownType.find(SymbolName); 1427 if (TypeIt == KnownType.end() || 1428 lookUpField(*TypeIt->second, Split.second, Type, Offset)) { 1429 std::pair<StringRef, StringRef> BaseMember = Split.second.split('.'); 1430 StringRef Base = BaseMember.first, Member = BaseMember.second; 1431 lookUpField(Base, Member, Type, Offset); 1432 } 1433 } 1434 1435 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName); 1436 if (!Sym) 1437 Sym = getContext().getOrCreateSymbol(SymbolName); 1438 1439 // If this is an absolute variable reference, substitute it now to preserve 1440 // semantics in the face of reassignment. 1441 if (Sym->isVariable()) { 1442 auto V = Sym->getVariableValue(/*SetUsed*/ false); 1443 bool DoInline = isa<MCConstantExpr>(V) && !Variant; 1444 if (auto TV = dyn_cast<MCTargetExpr>(V)) 1445 DoInline = TV->inlineAssignedExpr(); 1446 if (DoInline) { 1447 if (Variant) 1448 return Error(EndLoc, "unexpected modifier on variable reference"); 1449 Res = Sym->getVariableValue(/*SetUsed*/ false); 1450 return false; 1451 } 1452 } 1453 1454 // Otherwise create a symbol ref. 1455 const MCExpr *SymRef = 1456 MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc); 1457 if (Offset) { 1458 Res = MCBinaryExpr::create(MCBinaryExpr::Add, SymRef, 1459 MCConstantExpr::create(Offset, getContext()), 1460 getContext()); 1461 } else { 1462 Res = SymRef; 1463 } 1464 return false; 1465 } 1466 case AsmToken::BigNum: 1467 return TokError("literal value out of range for directive"); 1468 case AsmToken::Integer: { 1469 SMLoc Loc = getTok().getLoc(); 1470 int64_t IntVal = getTok().getIntVal(); 1471 Res = MCConstantExpr::create(IntVal, getContext()); 1472 EndLoc = Lexer.getTok().getEndLoc(); 1473 Lex(); // Eat token. 1474 // Look for 'b' or 'f' following an Integer as a directional label. 1475 if (Lexer.getKind() == AsmToken::Identifier) { 1476 StringRef IDVal = getTok().getString(); 1477 // Look up the symbol variant if used. 1478 std::pair<StringRef, StringRef> Split = IDVal.split('@'); 1479 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1480 if (Split.first.size() != IDVal.size()) { 1481 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1482 if (Variant == MCSymbolRefExpr::VK_Invalid) 1483 return TokError("invalid variant '" + Split.second + "'"); 1484 IDVal = Split.first; 1485 } 1486 if (IDVal == "f" || IDVal == "b") { 1487 MCSymbol *Sym = 1488 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b"); 1489 Res = MCSymbolRefExpr::create(Sym, Variant, getContext()); 1490 if (IDVal == "b" && Sym->isUndefined()) 1491 return Error(Loc, "directional label undefined"); 1492 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym)); 1493 EndLoc = Lexer.getTok().getEndLoc(); 1494 Lex(); // Eat identifier. 1495 } 1496 } 1497 return false; 1498 } 1499 case AsmToken::Real: { 1500 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); 1501 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 1502 Res = MCConstantExpr::create(IntVal, getContext()); 1503 EndLoc = Lexer.getTok().getEndLoc(); 1504 Lex(); // Eat token. 1505 return false; 1506 } 1507 case AsmToken::Dot: { 1508 // This is a '.' reference, which references the current PC. Emit a 1509 // temporary label to the streamer and refer to it. 1510 MCSymbol *Sym = Ctx.createTempSymbol(); 1511 Out.emitLabel(Sym); 1512 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext()); 1513 EndLoc = Lexer.getTok().getEndLoc(); 1514 Lex(); // Eat identifier. 1515 return false; 1516 } 1517 case AsmToken::LParen: 1518 Lex(); // Eat the '('. 1519 return parseParenExpr(Res, EndLoc); 1520 case AsmToken::LBrac: 1521 if (!PlatformParser->HasBracketExpressions()) 1522 return TokError("brackets expression not supported on this target"); 1523 Lex(); // Eat the '['. 1524 return parseBracketExpr(Res, EndLoc); 1525 case AsmToken::Minus: 1526 Lex(); // Eat the operator. 1527 if (parsePrimaryExpr(Res, EndLoc)) 1528 return true; 1529 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc); 1530 return false; 1531 case AsmToken::Plus: 1532 Lex(); // Eat the operator. 1533 if (parsePrimaryExpr(Res, EndLoc)) 1534 return true; 1535 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc); 1536 return false; 1537 case AsmToken::Tilde: 1538 Lex(); // Eat the operator. 1539 if (parsePrimaryExpr(Res, EndLoc)) 1540 return true; 1541 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1542 return false; 1543 // MIPS unary expression operators. The lexer won't generate these tokens if 1544 // MCAsmInfo::HasMipsExpressions is false for the target. 1545 case AsmToken::PercentCall16: 1546 case AsmToken::PercentCall_Hi: 1547 case AsmToken::PercentCall_Lo: 1548 case AsmToken::PercentDtprel_Hi: 1549 case AsmToken::PercentDtprel_Lo: 1550 case AsmToken::PercentGot: 1551 case AsmToken::PercentGot_Disp: 1552 case AsmToken::PercentGot_Hi: 1553 case AsmToken::PercentGot_Lo: 1554 case AsmToken::PercentGot_Ofst: 1555 case AsmToken::PercentGot_Page: 1556 case AsmToken::PercentGottprel: 1557 case AsmToken::PercentGp_Rel: 1558 case AsmToken::PercentHi: 1559 case AsmToken::PercentHigher: 1560 case AsmToken::PercentHighest: 1561 case AsmToken::PercentLo: 1562 case AsmToken::PercentNeg: 1563 case AsmToken::PercentPcrel_Hi: 1564 case AsmToken::PercentPcrel_Lo: 1565 case AsmToken::PercentTlsgd: 1566 case AsmToken::PercentTlsldm: 1567 case AsmToken::PercentTprel_Hi: 1568 case AsmToken::PercentTprel_Lo: 1569 Lex(); // Eat the operator. 1570 if (Lexer.isNot(AsmToken::LParen)) 1571 return TokError("expected '(' after operator"); 1572 Lex(); // Eat the operator. 1573 if (parseExpression(Res, EndLoc)) 1574 return true; 1575 if (Lexer.isNot(AsmToken::RParen)) 1576 return TokError("expected ')'"); 1577 Lex(); // Eat the operator. 1578 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx); 1579 return !Res; 1580 } 1581 } 1582 1583 bool MasmParser::parseExpression(const MCExpr *&Res) { 1584 SMLoc EndLoc; 1585 return parseExpression(Res, EndLoc); 1586 } 1587 1588 /// This function checks if the next token is <string> type or arithmetic. 1589 /// string that begin with character '<' must end with character '>'. 1590 /// otherwise it is arithmetics. 1591 /// If the function returns a 'true' value, 1592 /// the End argument will be filled with the last location pointed to the '>' 1593 /// character. 1594 1595 /// There is a gap between the AltMacro's documentation and the single quote 1596 /// implementation. GCC does not fully support this feature and so we will not 1597 /// support it. 1598 /// TODO: Adding single quote as a string. 1599 static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) { 1600 assert((StrLoc.getPointer() != nullptr) && 1601 "Argument to the function cannot be a NULL value"); 1602 const char *CharPtr = StrLoc.getPointer(); 1603 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && 1604 (*CharPtr != '\0')) { 1605 if (*CharPtr == '!') 1606 CharPtr++; 1607 CharPtr++; 1608 } 1609 if (*CharPtr == '>') { 1610 EndLoc = StrLoc.getFromPointer(CharPtr + 1); 1611 return true; 1612 } 1613 return false; 1614 } 1615 1616 /// creating a string without the escape characters '!'. 1617 static std::string angleBracketString(StringRef AltMacroStr) { 1618 std::string Res; 1619 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) { 1620 if (AltMacroStr[Pos] == '!') 1621 Pos++; 1622 Res += AltMacroStr[Pos]; 1623 } 1624 return Res; 1625 } 1626 1627 /// Parse an expression and return it. 1628 /// 1629 /// expr ::= expr &&,|| expr -> lowest. 1630 /// expr ::= expr |,^,&,! expr 1631 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 1632 /// expr ::= expr <<,>> expr 1633 /// expr ::= expr +,- expr 1634 /// expr ::= expr *,/,% expr -> highest. 1635 /// expr ::= primaryexpr 1636 /// 1637 bool MasmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1638 // Parse the expression. 1639 Res = nullptr; 1640 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || 1641 parseBinOpRHS(1, Res, EndLoc)) 1642 return true; 1643 1644 // Try to constant fold it up front, if possible. Do not exploit 1645 // assembler here. 1646 int64_t Value; 1647 if (Res->evaluateAsAbsolute(Value)) 1648 Res = MCConstantExpr::create(Value, getContext()); 1649 1650 return false; 1651 } 1652 1653 bool MasmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1654 Res = nullptr; 1655 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc); 1656 } 1657 1658 bool MasmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 1659 SMLoc &EndLoc) { 1660 if (parseParenExpr(Res, EndLoc)) 1661 return true; 1662 1663 for (; ParenDepth > 0; --ParenDepth) { 1664 if (parseBinOpRHS(1, Res, EndLoc)) 1665 return true; 1666 1667 // We don't Lex() the last RParen. 1668 // This is the same behavior as parseParenExpression(). 1669 if (ParenDepth - 1 > 0) { 1670 EndLoc = getTok().getEndLoc(); 1671 if (parseToken(AsmToken::RParen, 1672 "expected ')' in parentheses expression")) 1673 return true; 1674 } 1675 } 1676 return false; 1677 } 1678 1679 bool MasmParser::parseAbsoluteExpression(int64_t &Res) { 1680 const MCExpr *Expr; 1681 1682 SMLoc StartLoc = Lexer.getLoc(); 1683 if (parseExpression(Expr)) 1684 return true; 1685 1686 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 1687 return Error(StartLoc, "expected absolute expression"); 1688 1689 return false; 1690 } 1691 1692 static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K, 1693 MCBinaryExpr::Opcode &Kind, 1694 bool ShouldUseLogicalShr, 1695 bool EndExpressionAtGreater) { 1696 switch (K) { 1697 default: 1698 return 0; // not a binop. 1699 1700 // Lowest Precedence: &&, || 1701 case AsmToken::AmpAmp: 1702 Kind = MCBinaryExpr::LAnd; 1703 return 2; 1704 case AsmToken::PipePipe: 1705 Kind = MCBinaryExpr::LOr; 1706 return 1; 1707 1708 // Low Precedence: ==, !=, <>, <, <=, >, >= 1709 case AsmToken::EqualEqual: 1710 Kind = MCBinaryExpr::EQ; 1711 return 3; 1712 case AsmToken::ExclaimEqual: 1713 case AsmToken::LessGreater: 1714 Kind = MCBinaryExpr::NE; 1715 return 3; 1716 case AsmToken::Less: 1717 Kind = MCBinaryExpr::LT; 1718 return 3; 1719 case AsmToken::LessEqual: 1720 Kind = MCBinaryExpr::LTE; 1721 return 3; 1722 case AsmToken::Greater: 1723 if (EndExpressionAtGreater) 1724 return 0; 1725 Kind = MCBinaryExpr::GT; 1726 return 3; 1727 case AsmToken::GreaterEqual: 1728 Kind = MCBinaryExpr::GTE; 1729 return 3; 1730 1731 // Low Intermediate Precedence: +, - 1732 case AsmToken::Plus: 1733 Kind = MCBinaryExpr::Add; 1734 return 4; 1735 case AsmToken::Minus: 1736 Kind = MCBinaryExpr::Sub; 1737 return 4; 1738 1739 // High Intermediate Precedence: |, &, ^ 1740 case AsmToken::Pipe: 1741 Kind = MCBinaryExpr::Or; 1742 return 5; 1743 case AsmToken::Caret: 1744 Kind = MCBinaryExpr::Xor; 1745 return 5; 1746 case AsmToken::Amp: 1747 Kind = MCBinaryExpr::And; 1748 return 5; 1749 1750 // Highest Precedence: *, /, %, <<, >> 1751 case AsmToken::Star: 1752 Kind = MCBinaryExpr::Mul; 1753 return 6; 1754 case AsmToken::Slash: 1755 Kind = MCBinaryExpr::Div; 1756 return 6; 1757 case AsmToken::Percent: 1758 Kind = MCBinaryExpr::Mod; 1759 return 6; 1760 case AsmToken::LessLess: 1761 Kind = MCBinaryExpr::Shl; 1762 return 6; 1763 case AsmToken::GreaterGreater: 1764 if (EndExpressionAtGreater) 1765 return 0; 1766 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1767 return 6; 1768 } 1769 } 1770 1771 unsigned MasmParser::getBinOpPrecedence(AsmToken::TokenKind K, 1772 MCBinaryExpr::Opcode &Kind) { 1773 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); 1774 return getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr, 1775 AngleBracketDepth > 0); 1776 } 1777 1778 /// Parse all binary operators with precedence >= 'Precedence'. 1779 /// Res contains the LHS of the expression on input. 1780 bool MasmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 1781 SMLoc &EndLoc) { 1782 SMLoc StartLoc = Lexer.getLoc(); 1783 while (true) { 1784 AsmToken::TokenKind TokKind = Lexer.getKind(); 1785 if (Lexer.getKind() == AsmToken::Identifier) { 1786 StringRef Identifier = Lexer.getTok().getString(); 1787 if (Identifier.equals_lower("and")) 1788 TokKind = AsmToken::Amp; 1789 else if (Identifier.equals_lower("not")) 1790 TokKind = AsmToken::Exclaim; 1791 else if (Identifier.equals_lower("or")) 1792 TokKind = AsmToken::Pipe; 1793 } 1794 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 1795 unsigned TokPrec = getBinOpPrecedence(TokKind, Kind); 1796 1797 // If the next token is lower precedence than we are allowed to eat, return 1798 // successfully with what we ate already. 1799 if (TokPrec < Precedence) 1800 return false; 1801 1802 Lex(); 1803 1804 // Eat the next primary expression. 1805 const MCExpr *RHS; 1806 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc)) 1807 return true; 1808 1809 // If BinOp binds less tightly with RHS than the operator after RHS, let 1810 // the pending operator take RHS as its LHS. 1811 MCBinaryExpr::Opcode Dummy; 1812 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1813 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc)) 1814 return true; 1815 1816 // Merge LHS and RHS according to operator. 1817 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc); 1818 } 1819 } 1820 1821 /// ParseStatement: 1822 /// ::= EndOfStatement 1823 /// ::= Label* Directive ...Operands... EndOfStatement 1824 /// ::= Label* Identifier OperandList* EndOfStatement 1825 bool MasmParser::parseStatement(ParseStatementInfo &Info, 1826 MCAsmParserSemaCallback *SI) { 1827 assert(!hasPendingError() && "parseStatement started with pending error"); 1828 // Eat initial spaces and comments. 1829 while (Lexer.is(AsmToken::Space)) 1830 Lex(); 1831 if (Lexer.is(AsmToken::EndOfStatement)) { 1832 // If this is a line comment we can drop it safely. 1833 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 1834 getTok().getString().front() == '\n') 1835 Out.AddBlankLine(); 1836 Lex(); 1837 return false; 1838 } 1839 // Statements always start with an identifier, unless we're dealing with a 1840 // processor directive (.386, .686, etc.) that lexes as a real. 1841 AsmToken ID = getTok(); 1842 SMLoc IDLoc = ID.getLoc(); 1843 StringRef IDVal; 1844 int64_t LocalLabelVal = -1; 1845 if (Lexer.is(AsmToken::HashDirective)) 1846 return parseCppHashLineFilenameComment(IDLoc); 1847 // Allow an integer followed by a ':' as a directional local label. 1848 if (Lexer.is(AsmToken::Integer)) { 1849 LocalLabelVal = getTok().getIntVal(); 1850 if (LocalLabelVal < 0) { 1851 if (!TheCondState.Ignore) { 1852 Lex(); // always eat a token 1853 return Error(IDLoc, "unexpected token at start of statement"); 1854 } 1855 IDVal = ""; 1856 } else { 1857 IDVal = getTok().getString(); 1858 Lex(); // Consume the integer token to be used as an identifier token. 1859 if (Lexer.getKind() != AsmToken::Colon) { 1860 if (!TheCondState.Ignore) { 1861 Lex(); // always eat a token 1862 return Error(IDLoc, "unexpected token at start of statement"); 1863 } 1864 } 1865 } 1866 } else if (Lexer.is(AsmToken::Dot)) { 1867 // Treat '.' as a valid identifier in this context. 1868 Lex(); 1869 IDVal = "."; 1870 } else if (Lexer.is(AsmToken::LCurly)) { 1871 // Treat '{' as a valid identifier in this context. 1872 Lex(); 1873 IDVal = "{"; 1874 1875 } else if (Lexer.is(AsmToken::RCurly)) { 1876 // Treat '}' as a valid identifier in this context. 1877 Lex(); 1878 IDVal = "}"; 1879 } else if (Lexer.is(AsmToken::Star) && 1880 getTargetParser().starIsStartOfStatement()) { 1881 // Accept '*' as a valid start of statement. 1882 Lex(); 1883 IDVal = "*"; 1884 } else if (Lexer.is(AsmToken::Real)) { 1885 // Treat ".<number>" as a valid identifier in this context. 1886 IDVal = getTok().getString(); 1887 Lex(); // always eat a token 1888 if (!IDVal.startswith(".")) 1889 return Error(IDLoc, "unexpected token at start of statement"); 1890 } else if (parseIdentifier(IDVal)) { 1891 if (!TheCondState.Ignore) { 1892 Lex(); // always eat a token 1893 return Error(IDLoc, "unexpected token at start of statement"); 1894 } 1895 IDVal = ""; 1896 } 1897 1898 // Handle conditional assembly here before checking for skipping. We 1899 // have to do this so that .endif isn't skipped in a ".if 0" block for 1900 // example. 1901 StringMap<DirectiveKind>::const_iterator DirKindIt = 1902 DirectiveKindMap.find(IDVal.lower()); 1903 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) 1904 ? DK_NO_DIRECTIVE 1905 : DirKindIt->getValue(); 1906 switch (DirKind) { 1907 default: 1908 break; 1909 case DK_IF: 1910 case DK_IFE: 1911 return parseDirectiveIf(IDLoc, DirKind); 1912 case DK_IFB: 1913 return parseDirectiveIfb(IDLoc, true); 1914 case DK_IFNB: 1915 return parseDirectiveIfb(IDLoc, false); 1916 case DK_IFDEF: 1917 return parseDirectiveIfdef(IDLoc, true); 1918 case DK_IFNDEF: 1919 return parseDirectiveIfdef(IDLoc, false); 1920 case DK_IFDIF: 1921 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/false, 1922 /*CaseInsensitive=*/false); 1923 case DK_IFDIFI: 1924 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/false, 1925 /*CaseInsensitive=*/true); 1926 case DK_IFIDN: 1927 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/true, 1928 /*CaseInsensitive=*/false); 1929 case DK_IFIDNI: 1930 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/true, 1931 /*CaseInsensitive=*/true); 1932 case DK_ELSEIF: 1933 case DK_ELSEIFE: 1934 return parseDirectiveElseIf(IDLoc, DirKind); 1935 case DK_ELSEIFB: 1936 return parseDirectiveElseIfb(IDLoc, true); 1937 case DK_ELSEIFNB: 1938 return parseDirectiveElseIfb(IDLoc, false); 1939 case DK_ELSEIFDEF: 1940 return parseDirectiveElseIfdef(IDLoc, true); 1941 case DK_ELSEIFNDEF: 1942 return parseDirectiveElseIfdef(IDLoc, false); 1943 case DK_ELSEIFDIF: 1944 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/false, 1945 /*CaseInsensitive=*/false); 1946 case DK_ELSEIFDIFI: 1947 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/false, 1948 /*CaseInsensitive=*/true); 1949 case DK_ELSEIFIDN: 1950 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/true, 1951 /*CaseInsensitive=*/false); 1952 case DK_ELSEIFIDNI: 1953 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/true, 1954 /*CaseInsensitive=*/true); 1955 case DK_ELSE: 1956 return parseDirectiveElse(IDLoc); 1957 case DK_ENDIF: 1958 return parseDirectiveEndIf(IDLoc); 1959 } 1960 1961 // Ignore the statement if in the middle of inactive conditional 1962 // (e.g. ".if 0"). 1963 if (TheCondState.Ignore) { 1964 eatToEndOfStatement(); 1965 return false; 1966 } 1967 1968 // FIXME: Recurse on local labels? 1969 1970 // See what kind of statement we have. 1971 switch (Lexer.getKind()) { 1972 case AsmToken::Colon: { 1973 if (!getTargetParser().isLabel(ID)) 1974 break; 1975 if (checkForValidSection()) 1976 return true; 1977 1978 // identifier ':' -> Label. 1979 Lex(); 1980 1981 // Diagnose attempt to use '.' as a label. 1982 if (IDVal == ".") 1983 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1984 1985 // Diagnose attempt to use a variable as a label. 1986 // 1987 // FIXME: Diagnostics. Note the location of the definition as a label. 1988 // FIXME: This doesn't diagnose assignment to a symbol which has been 1989 // implicitly marked as external. 1990 MCSymbol *Sym; 1991 if (LocalLabelVal == -1) { 1992 if (ParsingMSInlineAsm && SI) { 1993 StringRef RewrittenLabel = 1994 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true); 1995 assert(!RewrittenLabel.empty() && 1996 "We should have an internal name here."); 1997 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(), 1998 RewrittenLabel); 1999 IDVal = RewrittenLabel; 2000 } 2001 Sym = getContext().getOrCreateSymbol(IDVal); 2002 } else 2003 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal); 2004 // End of Labels should be treated as end of line for lexing 2005 // purposes but that information is not available to the Lexer who 2006 // does not understand Labels. This may cause us to see a Hash 2007 // here instead of a preprocessor line comment. 2008 if (getTok().is(AsmToken::Hash)) { 2009 StringRef CommentStr = parseStringToEndOfStatement(); 2010 Lexer.Lex(); 2011 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr)); 2012 } 2013 2014 // Consume any end of statement token, if present, to avoid spurious 2015 // AddBlankLine calls(). 2016 if (getTok().is(AsmToken::EndOfStatement)) { 2017 Lex(); 2018 } 2019 2020 getTargetParser().doBeforeLabelEmit(Sym); 2021 2022 // Emit the label. 2023 if (!getTargetParser().isParsingMSInlineAsm()) 2024 Out.emitLabel(Sym, IDLoc); 2025 2026 // If we are generating dwarf for assembly source files then gather the 2027 // info to make a dwarf label entry for this label if needed. 2028 if (enabledGenDwarfForAssembly()) 2029 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 2030 IDLoc); 2031 2032 getTargetParser().onLabelParsed(Sym); 2033 2034 return false; 2035 } 2036 2037 default: // Normal instruction or directive. 2038 break; 2039 } 2040 2041 // If macros are enabled, check to see if this is a macro instantiation. 2042 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) { 2043 return handleMacroEntry(M, IDLoc); 2044 } 2045 2046 // Otherwise, we have a normal instruction or directive. 2047 2048 if (DirKind != DK_NO_DIRECTIVE) { 2049 // There are several entities interested in parsing directives: 2050 // 2051 // 1. Asm parser extensions. For example, platform-specific parsers 2052 // (like the ELF parser) register themselves as extensions. 2053 // 2. The target-specific assembly parser. Some directives are target 2054 // specific or may potentially behave differently on certain targets. 2055 // 3. The generic directive parser implemented by this class. These are 2056 // all the directives that behave in a target and platform independent 2057 // manner, or at least have a default behavior that's shared between 2058 // all targets and platforms. 2059 2060 getTargetParser().flushPendingInstructions(getStreamer()); 2061 2062 // Special-case handling of structure-end directives at higher priority, 2063 // since ENDS is overloaded as a segment-end directive. 2064 if (IDVal.equals_lower("ends") && StructInProgress.size() > 1 && 2065 getTok().is(AsmToken::EndOfStatement)) { 2066 return parseDirectiveNestedEnds(); 2067 } 2068 2069 // First, check the extension directive map to see if any extension has 2070 // registered itself to parse this directive. 2071 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 2072 ExtensionDirectiveMap.lookup(IDVal.lower()); 2073 if (Handler.first) 2074 return (*Handler.second)(Handler.first, IDVal, IDLoc); 2075 2076 // Next, let the target-specific assembly parser try. 2077 SMLoc StartTokLoc = getTok().getLoc(); 2078 bool TPDirectiveReturn = 2079 ID.is(AsmToken::Identifier) && getTargetParser().ParseDirective(ID); 2080 2081 if (hasPendingError()) 2082 return true; 2083 // Currently the return value should be true if we are 2084 // uninterested but as this is at odds with the standard parsing 2085 // convention (return true = error) we have instances of a parsed 2086 // directive that fails returning true as an error. Catch these 2087 // cases as best as possible errors here. 2088 if (TPDirectiveReturn && StartTokLoc != getTok().getLoc()) 2089 return true; 2090 // Return if we did some parsing or believe we succeeded. 2091 if (!TPDirectiveReturn || StartTokLoc != getTok().getLoc()) 2092 return false; 2093 2094 // Finally, if no one else is interested in this directive, it must be 2095 // generic and familiar to this class. 2096 switch (DirKind) { 2097 default: 2098 break; 2099 case DK_ASCII: 2100 return parseDirectiveAscii(IDVal, false); 2101 case DK_ASCIZ: 2102 case DK_STRING: 2103 return parseDirectiveAscii(IDVal, true); 2104 case DK_BYTE: 2105 case DK_SBYTE: 2106 case DK_DB: 2107 return parseDirectiveValue(IDVal, 1); 2108 case DK_WORD: 2109 case DK_SWORD: 2110 case DK_DW: 2111 return parseDirectiveValue(IDVal, 2); 2112 case DK_DWORD: 2113 case DK_SDWORD: 2114 case DK_DD: 2115 return parseDirectiveValue(IDVal, 4); 2116 case DK_FWORD: 2117 return parseDirectiveValue(IDVal, 6); 2118 case DK_QWORD: 2119 case DK_SQWORD: 2120 case DK_DQ: 2121 return parseDirectiveValue(IDVal, 8); 2122 case DK_REAL4: 2123 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle(), 4); 2124 case DK_REAL8: 2125 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble(), 8); 2126 case DK_STRUCT: 2127 case DK_UNION: 2128 return parseDirectiveNestedStruct(IDVal, DirKind); 2129 case DK_ENDS: 2130 return parseDirectiveNestedEnds(); 2131 case DK_ALIGN: 2132 return parseDirectiveAlign(); 2133 case DK_ORG: 2134 return parseDirectiveOrg(); 2135 case DK_EXTERN: 2136 eatToEndOfStatement(); // .extern is the default, ignore it. 2137 return false; 2138 case DK_PUBLIC: 2139 return parseDirectiveSymbolAttribute(MCSA_Global); 2140 case DK_COMM: 2141 return parseDirectiveComm(/*IsLocal=*/false); 2142 case DK_COMMENT: 2143 return parseDirectiveComment(IDLoc); 2144 case DK_INCLUDE: 2145 return parseDirectiveInclude(); 2146 case DK_REPT: 2147 return parseDirectiveRept(IDLoc, IDVal); 2148 case DK_IRP: 2149 return parseDirectiveIrp(IDLoc); 2150 case DK_IRPC: 2151 return parseDirectiveIrpc(IDLoc); 2152 case DK_ENDR: 2153 return parseDirectiveEndr(IDLoc); 2154 case DK_FILE: 2155 return parseDirectiveFile(IDLoc); 2156 case DK_LINE: 2157 return parseDirectiveLine(); 2158 case DK_LOC: 2159 return parseDirectiveLoc(); 2160 case DK_STABS: 2161 return parseDirectiveStabs(); 2162 case DK_CV_FILE: 2163 return parseDirectiveCVFile(); 2164 case DK_CV_FUNC_ID: 2165 return parseDirectiveCVFuncId(); 2166 case DK_CV_INLINE_SITE_ID: 2167 return parseDirectiveCVInlineSiteId(); 2168 case DK_CV_LOC: 2169 return parseDirectiveCVLoc(); 2170 case DK_CV_LINETABLE: 2171 return parseDirectiveCVLinetable(); 2172 case DK_CV_INLINE_LINETABLE: 2173 return parseDirectiveCVInlineLinetable(); 2174 case DK_CV_DEF_RANGE: 2175 return parseDirectiveCVDefRange(); 2176 case DK_CV_STRING: 2177 return parseDirectiveCVString(); 2178 case DK_CV_STRINGTABLE: 2179 return parseDirectiveCVStringTable(); 2180 case DK_CV_FILECHECKSUMS: 2181 return parseDirectiveCVFileChecksums(); 2182 case DK_CV_FILECHECKSUM_OFFSET: 2183 return parseDirectiveCVFileChecksumOffset(); 2184 case DK_CV_FPO_DATA: 2185 return parseDirectiveCVFPOData(); 2186 case DK_CFI_SECTIONS: 2187 return parseDirectiveCFISections(); 2188 case DK_CFI_STARTPROC: 2189 return parseDirectiveCFIStartProc(); 2190 case DK_CFI_ENDPROC: 2191 return parseDirectiveCFIEndProc(); 2192 case DK_CFI_DEF_CFA: 2193 return parseDirectiveCFIDefCfa(IDLoc); 2194 case DK_CFI_DEF_CFA_OFFSET: 2195 return parseDirectiveCFIDefCfaOffset(); 2196 case DK_CFI_ADJUST_CFA_OFFSET: 2197 return parseDirectiveCFIAdjustCfaOffset(); 2198 case DK_CFI_DEF_CFA_REGISTER: 2199 return parseDirectiveCFIDefCfaRegister(IDLoc); 2200 case DK_CFI_OFFSET: 2201 return parseDirectiveCFIOffset(IDLoc); 2202 case DK_CFI_REL_OFFSET: 2203 return parseDirectiveCFIRelOffset(IDLoc); 2204 case DK_CFI_PERSONALITY: 2205 return parseDirectiveCFIPersonalityOrLsda(true); 2206 case DK_CFI_LSDA: 2207 return parseDirectiveCFIPersonalityOrLsda(false); 2208 case DK_CFI_REMEMBER_STATE: 2209 return parseDirectiveCFIRememberState(); 2210 case DK_CFI_RESTORE_STATE: 2211 return parseDirectiveCFIRestoreState(); 2212 case DK_CFI_SAME_VALUE: 2213 return parseDirectiveCFISameValue(IDLoc); 2214 case DK_CFI_RESTORE: 2215 return parseDirectiveCFIRestore(IDLoc); 2216 case DK_CFI_ESCAPE: 2217 return parseDirectiveCFIEscape(); 2218 case DK_CFI_RETURN_COLUMN: 2219 return parseDirectiveCFIReturnColumn(IDLoc); 2220 case DK_CFI_SIGNAL_FRAME: 2221 return parseDirectiveCFISignalFrame(); 2222 case DK_CFI_UNDEFINED: 2223 return parseDirectiveCFIUndefined(IDLoc); 2224 case DK_CFI_REGISTER: 2225 return parseDirectiveCFIRegister(IDLoc); 2226 case DK_CFI_WINDOW_SAVE: 2227 return parseDirectiveCFIWindowSave(); 2228 case DK_MACRO: 2229 return parseDirectiveMacro(IDLoc); 2230 case DK_ALTMACRO: 2231 case DK_NOALTMACRO: 2232 return parseDirectiveAltmacro(IDVal); 2233 case DK_EXITM: 2234 return parseDirectiveExitMacro(IDVal); 2235 case DK_ENDM: 2236 return parseDirectiveEndMacro(IDVal); 2237 case DK_PURGEM: 2238 return parseDirectivePurgeMacro(IDLoc); 2239 case DK_END: 2240 return parseDirectiveEnd(IDLoc); 2241 case DK_ERR: 2242 return parseDirectiveError(IDLoc); 2243 case DK_ERRB: 2244 return parseDirectiveErrorIfb(IDLoc, true); 2245 case DK_ERRNB: 2246 return parseDirectiveErrorIfb(IDLoc, false); 2247 case DK_ERRDEF: 2248 return parseDirectiveErrorIfdef(IDLoc, true); 2249 case DK_ERRNDEF: 2250 return parseDirectiveErrorIfdef(IDLoc, false); 2251 case DK_ERRDIF: 2252 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/false, 2253 /*CaseInsensitive=*/false); 2254 case DK_ERRDIFI: 2255 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/false, 2256 /*CaseInsensitive=*/true); 2257 case DK_ERRIDN: 2258 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/true, 2259 /*CaseInsensitive=*/false); 2260 case DK_ERRIDNI: 2261 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/true, 2262 /*CaseInsensitive=*/true); 2263 case DK_ERRE: 2264 return parseDirectiveErrorIfe(IDLoc, true); 2265 case DK_ERRNZ: 2266 return parseDirectiveErrorIfe(IDLoc, false); 2267 case DK_ECHO: 2268 return parseDirectiveEcho(); 2269 } 2270 2271 return Error(IDLoc, "unknown directive"); 2272 } 2273 2274 // We also check if this is allocating memory with user-defined type. 2275 auto IDIt = Structs.find(IDVal.lower()); 2276 if (IDIt != Structs.end()) 2277 return parseDirectiveStructValue(/*Structure=*/IDIt->getValue(), IDVal, 2278 IDLoc); 2279 2280 // Non-conditional Microsoft directives sometimes follow their first argument. 2281 const AsmToken nextTok = getTok(); 2282 const StringRef nextVal = nextTok.getString(); 2283 const SMLoc nextLoc = nextTok.getLoc(); 2284 2285 // There are several entities interested in parsing infix directives: 2286 // 2287 // 1. Asm parser extensions. For example, platform-specific parsers 2288 // (like the ELF parser) register themselves as extensions. 2289 // 2. The generic directive parser implemented by this class. These are 2290 // all the directives that behave in a target and platform independent 2291 // manner, or at least have a default behavior that's shared between 2292 // all targets and platforms. 2293 2294 getTargetParser().flushPendingInstructions(getStreamer()); 2295 2296 // Special-case handling of structure-end directives at higher priority, since 2297 // ENDS is overloaded as a segment-end directive. 2298 if (nextVal.equals_lower("ends") && StructInProgress.size() == 1) { 2299 Lex(); 2300 return parseDirectiveEnds(IDVal, IDLoc); 2301 } 2302 2303 // First, check the extension directive map to see if any extension has 2304 // registered itself to parse this directive. 2305 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 2306 ExtensionDirectiveMap.lookup(nextVal.lower()); 2307 if (Handler.first) { 2308 Lex(); 2309 Lexer.UnLex(ID); 2310 return (*Handler.second)(Handler.first, nextVal, nextLoc); 2311 } 2312 2313 // If no one else is interested in this directive, it must be 2314 // generic and familiar to this class. 2315 DirKindIt = DirectiveKindMap.find(nextVal.lower()); 2316 DirKind = (DirKindIt == DirectiveKindMap.end()) 2317 ? DK_NO_DIRECTIVE 2318 : DirKindIt->getValue(); 2319 switch (DirKind) { 2320 default: 2321 break; 2322 case DK_ASSIGN: 2323 case DK_EQU: 2324 case DK_TEXTEQU: 2325 Lex(); 2326 return parseDirectiveEquate(nextVal, IDVal, DirKind); 2327 case DK_BYTE: 2328 case DK_DB: 2329 Lex(); 2330 return parseDirectiveNamedValue(nextVal, 1, IDVal, IDLoc); 2331 case DK_WORD: 2332 case DK_DW: 2333 Lex(); 2334 return parseDirectiveNamedValue(nextVal, 2, IDVal, IDLoc); 2335 case DK_DWORD: 2336 case DK_DD: 2337 Lex(); 2338 return parseDirectiveNamedValue(nextVal, 4, IDVal, IDLoc); 2339 case DK_FWORD: 2340 Lex(); 2341 return parseDirectiveNamedValue(nextVal, 6, IDVal, IDLoc); 2342 case DK_QWORD: 2343 case DK_DQ: 2344 Lex(); 2345 return parseDirectiveNamedValue(nextVal, 8, IDVal, IDLoc); 2346 case DK_REAL4: 2347 Lex(); 2348 return parseDirectiveNamedRealValue(nextVal, APFloat::IEEEsingle(), 4, 2349 IDVal, IDLoc); 2350 case DK_REAL8: 2351 Lex(); 2352 return parseDirectiveNamedRealValue(nextVal, APFloat::IEEEdouble(), 8, 2353 IDVal, IDLoc); 2354 case DK_STRUCT: 2355 case DK_UNION: 2356 Lex(); 2357 return parseDirectiveStruct(nextVal, DirKind, IDVal, IDLoc); 2358 case DK_ENDS: 2359 Lex(); 2360 return parseDirectiveEnds(IDVal, IDLoc); 2361 } 2362 2363 // Finally, we check if this is allocating a variable with user-defined type. 2364 auto NextIt = Structs.find(nextVal.lower()); 2365 if (NextIt != Structs.end()) { 2366 Lex(); 2367 return parseDirectiveNamedStructValue(/*Structure=*/NextIt->getValue(), 2368 nextVal, nextLoc, IDVal); 2369 } 2370 2371 // __asm _emit or __asm __emit 2372 if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || 2373 IDVal == "_EMIT" || IDVal == "__EMIT")) 2374 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size()); 2375 2376 // __asm align 2377 if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN")) 2378 return parseDirectiveMSAlign(IDLoc, Info); 2379 2380 if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN")) 2381 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4); 2382 if (checkForValidSection()) 2383 return true; 2384 2385 // Canonicalize the opcode to lower case. 2386 std::string OpcodeStr = IDVal.lower(); 2387 ParseInstructionInfo IInfo(Info.AsmRewrites); 2388 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID, 2389 Info.ParsedOperands); 2390 Info.ParseError = ParseHadError; 2391 2392 // Dump the parsed representation, if requested. 2393 if (getShowParsedOperands()) { 2394 SmallString<256> Str; 2395 raw_svector_ostream OS(Str); 2396 OS << "parsed instruction: ["; 2397 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { 2398 if (i != 0) 2399 OS << ", "; 2400 Info.ParsedOperands[i]->print(OS); 2401 } 2402 OS << "]"; 2403 2404 printMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 2405 } 2406 2407 // Fail even if ParseInstruction erroneously returns false. 2408 if (hasPendingError() || ParseHadError) 2409 return true; 2410 2411 // If we are generating dwarf for the current section then generate a .loc 2412 // directive for the instruction. 2413 if (!ParseHadError && enabledGenDwarfForAssembly() && 2414 getContext().getGenDwarfSectionSyms().count( 2415 getStreamer().getCurrentSectionOnly())) { 2416 unsigned Line; 2417 if (ActiveMacros.empty()) 2418 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer); 2419 else 2420 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc, 2421 ActiveMacros.front()->ExitBuffer); 2422 2423 // If we previously parsed a cpp hash file line comment then make sure the 2424 // current Dwarf File is for the CppHashFilename if not then emit the 2425 // Dwarf File table for it and adjust the line number for the .loc. 2426 if (!CppHashInfo.Filename.empty()) { 2427 unsigned FileNumber = getStreamer().emitDwarfFileDirective( 2428 0, StringRef(), CppHashInfo.Filename); 2429 getContext().setGenDwarfFileNumber(FileNumber); 2430 2431 unsigned CppHashLocLineNo = 2432 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf); 2433 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); 2434 } 2435 2436 getStreamer().emitDwarfLocDirective( 2437 getContext().getGenDwarfFileNumber(), Line, 0, 2438 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0, 2439 StringRef()); 2440 } 2441 2442 // If parsing succeeded, match the instruction. 2443 if (!ParseHadError) { 2444 uint64_t ErrorInfo; 2445 if (getTargetParser().MatchAndEmitInstruction( 2446 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo, 2447 getTargetParser().isParsingMSInlineAsm())) 2448 return true; 2449 } 2450 return false; 2451 } 2452 2453 // Parse and erase curly braces marking block start/end. 2454 bool MasmParser::parseCurlyBlockScope( 2455 SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { 2456 // Identify curly brace marking block start/end. 2457 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly)) 2458 return false; 2459 2460 SMLoc StartLoc = Lexer.getLoc(); 2461 Lex(); // Eat the brace. 2462 if (Lexer.is(AsmToken::EndOfStatement)) 2463 Lex(); // Eat EndOfStatement following the brace. 2464 2465 // Erase the block start/end brace from the output asm string. 2466 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() - 2467 StartLoc.getPointer()); 2468 return true; 2469 } 2470 2471 /// parseCppHashLineFilenameComment as this: 2472 /// ::= # number "filename" 2473 bool MasmParser::parseCppHashLineFilenameComment(SMLoc L) { 2474 Lex(); // Eat the hash token. 2475 // Lexer only ever emits HashDirective if it fully formed if it's 2476 // done the checking already so this is an internal error. 2477 assert(getTok().is(AsmToken::Integer) && 2478 "Lexing Cpp line comment: Expected Integer"); 2479 int64_t LineNumber = getTok().getIntVal(); 2480 Lex(); 2481 assert(getTok().is(AsmToken::String) && 2482 "Lexing Cpp line comment: Expected String"); 2483 StringRef Filename = getTok().getString(); 2484 Lex(); 2485 2486 // Get rid of the enclosing quotes. 2487 Filename = Filename.substr(1, Filename.size() - 2); 2488 2489 // Save the SMLoc, Filename and LineNumber for later use by diagnostics 2490 // and possibly DWARF file info. 2491 CppHashInfo.Loc = L; 2492 CppHashInfo.Filename = Filename; 2493 CppHashInfo.LineNumber = LineNumber; 2494 CppHashInfo.Buf = CurBuffer; 2495 if (FirstCppHashFilename.empty()) 2496 FirstCppHashFilename = Filename; 2497 return false; 2498 } 2499 2500 /// will use the last parsed cpp hash line filename comment 2501 /// for the Filename and LineNo if any in the diagnostic. 2502 void MasmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 2503 const MasmParser *Parser = static_cast<const MasmParser *>(Context); 2504 raw_ostream &OS = errs(); 2505 2506 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 2507 SMLoc DiagLoc = Diag.getLoc(); 2508 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2509 unsigned CppHashBuf = 2510 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc); 2511 2512 // Like SourceMgr::printMessage() we need to print the include stack if any 2513 // before printing the message. 2514 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2515 if (!Parser->SavedDiagHandler && DiagCurBuffer && 2516 DiagCurBuffer != DiagSrcMgr.getMainFileID()) { 2517 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 2518 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 2519 } 2520 2521 // If we have not parsed a cpp hash line filename comment or the source 2522 // manager changed or buffer changed (like in a nested include) then just 2523 // print the normal diagnostic using its Filename and LineNo. 2524 if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr || 2525 DiagBuf != CppHashBuf) { 2526 if (Parser->SavedDiagHandler) 2527 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2528 else 2529 Diag.print(nullptr, OS); 2530 return; 2531 } 2532 2533 // Use the CppHashFilename and calculate a line number based on the 2534 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc 2535 // for the diagnostic. 2536 const std::string &Filename = std::string(Parser->CppHashInfo.Filename); 2537 2538 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 2539 int CppHashLocLineNo = 2540 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf); 2541 int LineNo = 2542 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); 2543 2544 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, 2545 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), 2546 Diag.getLineContents(), Diag.getRanges()); 2547 2548 if (Parser->SavedDiagHandler) 2549 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); 2550 else 2551 NewDiag.print(nullptr, OS); 2552 } 2553 2554 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The 2555 // difference being that that function accepts '@' as part of identifiers and 2556 // we can't do that. AsmLexer.cpp should probably be changed to handle 2557 // '@' as a special case when needed. 2558 static bool isIdentifierChar(char c) { 2559 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' || 2560 c == '.'; 2561 } 2562 2563 bool MasmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 2564 ArrayRef<MCAsmMacroParameter> Parameters, 2565 ArrayRef<MCAsmMacroArgument> A, 2566 bool EnableAtPseudoVariable, SMLoc L) { 2567 unsigned NParameters = Parameters.size(); 2568 bool HasVararg = NParameters ? Parameters.back().Vararg : false; 2569 if ((!IsDarwin || NParameters != 0) && NParameters != A.size()) 2570 return Error(L, "Wrong number of arguments"); 2571 2572 // A macro without parameters is handled differently on Darwin: 2573 // gas accepts no arguments and does no substitutions 2574 while (!Body.empty()) { 2575 // Scan for the next substitution. 2576 std::size_t End = Body.size(), Pos = 0; 2577 for (; Pos != End; ++Pos) { 2578 // Check for a substitution or escape. 2579 if (IsDarwin && !NParameters) { 2580 // This macro has no parameters, look for $0, $1, etc. 2581 if (Body[Pos] != '$' || Pos + 1 == End) 2582 continue; 2583 2584 char Next = Body[Pos + 1]; 2585 if (Next == '$' || Next == 'n' || 2586 isdigit(static_cast<unsigned char>(Next))) 2587 break; 2588 } else { 2589 // This macro has parameters, look for \foo, \bar, etc. 2590 if (Body[Pos] == '\\' && Pos + 1 != End) 2591 break; 2592 } 2593 } 2594 2595 // Add the prefix. 2596 OS << Body.slice(0, Pos); 2597 2598 // Check if we reached the end. 2599 if (Pos == End) 2600 break; 2601 2602 if (IsDarwin && !NParameters) { 2603 switch (Body[Pos + 1]) { 2604 // $$ => $ 2605 case '$': 2606 OS << '$'; 2607 break; 2608 2609 // $n => number of arguments 2610 case 'n': 2611 OS << A.size(); 2612 break; 2613 2614 // $[0-9] => argument 2615 default: { 2616 // Missing arguments are ignored. 2617 unsigned Index = Body[Pos + 1] - '0'; 2618 if (Index >= A.size()) 2619 break; 2620 2621 // Otherwise substitute with the token values, with spaces eliminated. 2622 for (const AsmToken &Token : A[Index]) 2623 OS << Token.getString(); 2624 break; 2625 } 2626 } 2627 Pos += 2; 2628 } else { 2629 unsigned I = Pos + 1; 2630 2631 // Check for the \@ pseudo-variable. 2632 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End) 2633 ++I; 2634 else 2635 while (isIdentifierChar(Body[I]) && I + 1 != End) 2636 ++I; 2637 2638 const char *Begin = Body.data() + Pos + 1; 2639 StringRef Argument(Begin, I - (Pos + 1)); 2640 unsigned Index = 0; 2641 2642 if (Argument == "@") { 2643 OS << NumOfMacroInstantiations; 2644 Pos += 2; 2645 } else { 2646 for (; Index < NParameters; ++Index) 2647 if (Parameters[Index].Name == Argument) 2648 break; 2649 2650 if (Index == NParameters) { 2651 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 2652 Pos += 3; 2653 else { 2654 OS << '\\' << Argument; 2655 Pos = I; 2656 } 2657 } else { 2658 bool VarargParameter = HasVararg && Index == (NParameters - 1); 2659 for (const AsmToken &Token : A[Index]) 2660 // For altmacro mode, you can write '%expr'. 2661 // The prefix '%' evaluates the expression 'expr' 2662 // and uses the result as a string (e.g. replace %(1+2) with the 2663 // string "3"). 2664 // Here, we identify the integer token which is the result of the 2665 // absolute expression evaluation and replace it with its string 2666 // representation. 2667 if (AltMacroMode && Token.getString().front() == '%' && 2668 Token.is(AsmToken::Integer)) 2669 // Emit an integer value to the buffer. 2670 OS << Token.getIntVal(); 2671 // Only Token that was validated as a string and begins with '<' 2672 // is considered altMacroString!!! 2673 else if (AltMacroMode && Token.getString().front() == '<' && 2674 Token.is(AsmToken::String)) { 2675 OS << angleBracketString(Token.getStringContents()); 2676 } 2677 // We expect no quotes around the string's contents when 2678 // parsing for varargs. 2679 else if (Token.isNot(AsmToken::String) || VarargParameter) 2680 OS << Token.getString(); 2681 else 2682 OS << Token.getStringContents(); 2683 2684 Pos += 1 + Argument.size(); 2685 } 2686 } 2687 } 2688 // Update the scan point. 2689 Body = Body.substr(Pos); 2690 } 2691 2692 return false; 2693 } 2694 2695 static bool isOperator(AsmToken::TokenKind kind) { 2696 switch (kind) { 2697 default: 2698 return false; 2699 case AsmToken::Plus: 2700 case AsmToken::Minus: 2701 case AsmToken::Tilde: 2702 case AsmToken::Slash: 2703 case AsmToken::Star: 2704 case AsmToken::Dot: 2705 case AsmToken::Equal: 2706 case AsmToken::EqualEqual: 2707 case AsmToken::Pipe: 2708 case AsmToken::PipePipe: 2709 case AsmToken::Caret: 2710 case AsmToken::Amp: 2711 case AsmToken::AmpAmp: 2712 case AsmToken::Exclaim: 2713 case AsmToken::ExclaimEqual: 2714 case AsmToken::Less: 2715 case AsmToken::LessEqual: 2716 case AsmToken::LessLess: 2717 case AsmToken::LessGreater: 2718 case AsmToken::Greater: 2719 case AsmToken::GreaterEqual: 2720 case AsmToken::GreaterGreater: 2721 return true; 2722 } 2723 } 2724 2725 namespace { 2726 2727 class AsmLexerSkipSpaceRAII { 2728 public: 2729 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { 2730 Lexer.setSkipSpace(SkipSpace); 2731 } 2732 2733 ~AsmLexerSkipSpaceRAII() { 2734 Lexer.setSkipSpace(true); 2735 } 2736 2737 private: 2738 AsmLexer &Lexer; 2739 }; 2740 2741 } // end anonymous namespace 2742 2743 bool MasmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) { 2744 2745 if (Vararg) { 2746 if (Lexer.isNot(AsmToken::EndOfStatement)) { 2747 StringRef Str = parseStringToEndOfStatement(); 2748 MA.emplace_back(AsmToken::String, Str); 2749 } 2750 return false; 2751 } 2752 2753 unsigned ParenLevel = 0; 2754 2755 // Darwin doesn't use spaces to delmit arguments. 2756 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); 2757 2758 bool SpaceEaten; 2759 2760 while (true) { 2761 SpaceEaten = false; 2762 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 2763 return TokError("unexpected token in macro instantiation"); 2764 2765 if (ParenLevel == 0) { 2766 2767 if (Lexer.is(AsmToken::Comma)) 2768 break; 2769 2770 if (Lexer.is(AsmToken::Space)) { 2771 SpaceEaten = true; 2772 Lexer.Lex(); // Eat spaces. 2773 } 2774 2775 // Spaces can delimit parameters, but could also be part an expression. 2776 // If the token after a space is an operator, add the token and the next 2777 // one into this argument 2778 if (!IsDarwin) { 2779 if (isOperator(Lexer.getKind())) { 2780 MA.push_back(getTok()); 2781 Lexer.Lex(); 2782 2783 // Whitespace after an operator can be ignored. 2784 if (Lexer.is(AsmToken::Space)) 2785 Lexer.Lex(); 2786 2787 continue; 2788 } 2789 } 2790 if (SpaceEaten) 2791 break; 2792 } 2793 2794 // handleMacroEntry relies on not advancing the lexer here 2795 // to be able to fill in the remaining default parameter values 2796 if (Lexer.is(AsmToken::EndOfStatement)) 2797 break; 2798 2799 // Adjust the current parentheses level. 2800 if (Lexer.is(AsmToken::LParen)) 2801 ++ParenLevel; 2802 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 2803 --ParenLevel; 2804 2805 // Append the token to the current argument list. 2806 MA.push_back(getTok()); 2807 Lexer.Lex(); 2808 } 2809 2810 if (ParenLevel != 0) 2811 return TokError("unbalanced parentheses in macro argument"); 2812 return false; 2813 } 2814 2815 // Parse the macro instantiation arguments. 2816 bool MasmParser::parseMacroArguments(const MCAsmMacro *M, 2817 MCAsmMacroArguments &A) { 2818 const unsigned NParameters = M ? M->Parameters.size() : 0; 2819 bool NamedParametersFound = false; 2820 SmallVector<SMLoc, 4> FALocs; 2821 2822 A.resize(NParameters); 2823 FALocs.resize(NParameters); 2824 2825 // Parse two kinds of macro invocations: 2826 // - macros defined without any parameters accept an arbitrary number of them 2827 // - macros defined with parameters accept at most that many of them 2828 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false; 2829 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 2830 ++Parameter) { 2831 SMLoc IDLoc = Lexer.getLoc(); 2832 MCAsmMacroParameter FA; 2833 2834 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) { 2835 if (parseIdentifier(FA.Name)) 2836 return Error(IDLoc, "invalid argument identifier for formal argument"); 2837 2838 if (Lexer.isNot(AsmToken::Equal)) 2839 return TokError("expected '=' after formal parameter identifier"); 2840 2841 Lex(); 2842 2843 NamedParametersFound = true; 2844 } 2845 bool Vararg = HasVararg && Parameter == (NParameters - 1); 2846 2847 if (NamedParametersFound && FA.Name.empty()) 2848 return Error(IDLoc, "cannot mix positional and keyword arguments"); 2849 2850 SMLoc StrLoc = Lexer.getLoc(); 2851 SMLoc EndLoc; 2852 if (AltMacroMode && Lexer.is(AsmToken::Percent)) { 2853 const MCExpr *AbsoluteExp; 2854 int64_t Value; 2855 /// Eat '%'. 2856 Lex(); 2857 if (parseExpression(AbsoluteExp, EndLoc)) 2858 return false; 2859 if (!AbsoluteExp->evaluateAsAbsolute(Value, 2860 getStreamer().getAssemblerPtr())) 2861 return Error(StrLoc, "expected absolute expression"); 2862 const char *StrChar = StrLoc.getPointer(); 2863 const char *EndChar = EndLoc.getPointer(); 2864 AsmToken newToken(AsmToken::Integer, 2865 StringRef(StrChar, EndChar - StrChar), Value); 2866 FA.Value.push_back(newToken); 2867 } else if (AltMacroMode && Lexer.is(AsmToken::Less) && 2868 isAngleBracketString(StrLoc, EndLoc)) { 2869 const char *StrChar = StrLoc.getPointer(); 2870 const char *EndChar = EndLoc.getPointer(); 2871 jumpToLoc(EndLoc, CurBuffer); 2872 /// Eat from '<' to '>'. 2873 Lex(); 2874 AsmToken newToken(AsmToken::String, 2875 StringRef(StrChar, EndChar - StrChar)); 2876 FA.Value.push_back(newToken); 2877 } else if(parseMacroArgument(FA.Value, Vararg)) 2878 return true; 2879 2880 unsigned PI = Parameter; 2881 if (!FA.Name.empty()) { 2882 unsigned FAI = 0; 2883 for (FAI = 0; FAI < NParameters; ++FAI) 2884 if (M->Parameters[FAI].Name == FA.Name) 2885 break; 2886 2887 if (FAI >= NParameters) { 2888 assert(M && "expected macro to be defined"); 2889 return Error(IDLoc, "parameter named '" + FA.Name + 2890 "' does not exist for macro '" + M->Name + "'"); 2891 } 2892 PI = FAI; 2893 } 2894 2895 if (!FA.Value.empty()) { 2896 if (A.size() <= PI) 2897 A.resize(PI + 1); 2898 A[PI] = FA.Value; 2899 2900 if (FALocs.size() <= PI) 2901 FALocs.resize(PI + 1); 2902 2903 FALocs[PI] = Lexer.getLoc(); 2904 } 2905 2906 // At the end of the statement, fill in remaining arguments that have 2907 // default values. If there aren't any, then the next argument is 2908 // required but missing 2909 if (Lexer.is(AsmToken::EndOfStatement)) { 2910 bool Failure = false; 2911 for (unsigned FAI = 0; FAI < NParameters; ++FAI) { 2912 if (A[FAI].empty()) { 2913 if (M->Parameters[FAI].Required) { 2914 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), 2915 "missing value for required parameter " 2916 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'"); 2917 Failure = true; 2918 } 2919 2920 if (!M->Parameters[FAI].Value.empty()) 2921 A[FAI] = M->Parameters[FAI].Value; 2922 } 2923 } 2924 return Failure; 2925 } 2926 2927 if (Lexer.is(AsmToken::Comma)) 2928 Lex(); 2929 } 2930 2931 return TokError("too many positional arguments"); 2932 } 2933 2934 bool MasmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) { 2935 // Arbitrarily limit macro nesting depth (default matches 'as'). We can 2936 // eliminate this, although we should protect against infinite loops. 2937 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; 2938 if (ActiveMacros.size() == MaxNestingDepth) { 2939 std::ostringstream MaxNestingDepthError; 2940 MaxNestingDepthError << "macros cannot be nested more than " 2941 << MaxNestingDepth << " levels deep." 2942 << " Use -asm-macro-max-nesting-depth to increase " 2943 "this limit."; 2944 return TokError(MaxNestingDepthError.str()); 2945 } 2946 2947 MCAsmMacroArguments A; 2948 if (parseMacroArguments(M, A)) 2949 return true; 2950 2951 // Macro instantiation is lexical, unfortunately. We construct a new buffer 2952 // to hold the macro body with substitutions. 2953 SmallString<256> Buf; 2954 StringRef Body = M->Body; 2955 raw_svector_ostream OS(Buf); 2956 2957 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc())) 2958 return true; 2959 2960 // We include the .endmacro in the buffer as our cue to exit the macro 2961 // instantiation. 2962 OS << ".endmacro\n"; 2963 2964 std::unique_ptr<MemoryBuffer> Instantiation = 2965 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 2966 2967 // Create the macro instantiation object and add to the current macro 2968 // instantiation stack. 2969 MacroInstantiation *MI = new MacroInstantiation{ 2970 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 2971 ActiveMacros.push_back(MI); 2972 2973 ++NumOfMacroInstantiations; 2974 2975 // Jump to the macro instantiation and prime the lexer. 2976 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 2977 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 2978 Lex(); 2979 2980 return false; 2981 } 2982 2983 void MasmParser::handleMacroExit() { 2984 // Jump to the EndOfStatement we should return to, and consume it. 2985 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer); 2986 Lex(); 2987 2988 // Pop the instantiation entry. 2989 delete ActiveMacros.back(); 2990 ActiveMacros.pop_back(); 2991 } 2992 2993 /// parseIdentifier: 2994 /// ::= identifier 2995 /// ::= string 2996 bool MasmParser::parseIdentifier(StringRef &Res) { 2997 // The assembler has relaxed rules for accepting identifiers, in particular we 2998 // allow things like '.globl $foo' and '.def @feat.00', which would normally 2999 // be separate tokens. At this level, we have already lexed so we cannot 3000 // (currently) handle this as a context dependent token, instead we detect 3001 // adjacent tokens and return the combined identifier. 3002 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) { 3003 SMLoc PrefixLoc = getLexer().getLoc(); 3004 3005 // Consume the prefix character, and check for a following identifier. 3006 3007 AsmToken Buf[1]; 3008 Lexer.peekTokens(Buf, false); 3009 3010 if (Buf[0].isNot(AsmToken::Identifier)) 3011 return true; 3012 3013 // We have a '$' or '@' followed by an identifier, make sure they are adjacent. 3014 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer()) 3015 return true; 3016 3017 // eat $ or @ 3018 Lexer.Lex(); // Lexer's Lex guarantees consecutive token. 3019 // Construct the joined identifier and consume the token. 3020 Res = 3021 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1); 3022 Lex(); // Parser Lex to maintain invariants. 3023 return false; 3024 } 3025 3026 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String)) 3027 return true; 3028 3029 Res = getTok().getIdentifier(); 3030 3031 Lex(); // Consume the identifier token. 3032 3033 return false; 3034 } 3035 3036 /// parseDirectiveEquate: 3037 /// ::= name "=" expression 3038 /// | name "equ" expression (not redefinable) 3039 /// | name "equ" text-list 3040 /// | name "textequ" text-list 3041 bool MasmParser::parseDirectiveEquate(StringRef IDVal, StringRef Name, 3042 DirectiveKind DirKind) { 3043 Variable &Var = Variables[Name]; 3044 if (Var.Name.empty()) { 3045 Var.Name = Name; 3046 } else if (!Var.Redefinable) { 3047 return TokError("invalid variable redefinition"); 3048 } 3049 Var.Redefinable = (DirKind != DK_EQU); 3050 3051 if (DirKind == DK_EQU || DirKind == DK_TEXTEQU) { 3052 // "equ" and "textequ" both allow text expressions. 3053 std::string Value; 3054 if (!parseTextItem(Value)) { 3055 Var.IsText = true; 3056 Var.TextValue = Value; 3057 3058 // Accept a text-list, not just one text-item. 3059 auto parseItem = [&]() -> bool { 3060 if (parseTextItem(Value)) 3061 return true; 3062 Var.TextValue += Value; 3063 return false; 3064 }; 3065 if (parseOptionalToken(AsmToken::Comma) && parseMany(parseItem)) 3066 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3067 3068 return false; 3069 } 3070 } 3071 if (DirKind == DK_TEXTEQU) 3072 return TokError("expected <text> in '" + Twine(IDVal) + "' directive"); 3073 3074 // Parse as expression assignment. 3075 const MCExpr *Expr; 3076 SMLoc EndLoc, StartLoc = Lexer.getLoc(); 3077 if (parseExpression(Expr, EndLoc)) 3078 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3079 MCSymbol *Sym = getContext().getOrCreateSymbol(Var.Name); 3080 Sym->setRedefinable(Var.Redefinable); 3081 Sym->setVariableValue(Expr); 3082 Sym->setExternal(false); 3083 3084 if (Expr->evaluateAsAbsolute(Var.NumericValue, 3085 getStreamer().getAssemblerPtr())) 3086 return false; 3087 3088 // Not an absolute expression; define as a text replacement. 3089 Var.IsText = true; 3090 Var.TextValue = StringRef(StartLoc.getPointer(), 3091 EndLoc.getPointer() - StartLoc.getPointer()).str(); 3092 return false; 3093 } 3094 3095 bool MasmParser::parseEscapedString(std::string &Data) { 3096 if (check(getTok().isNot(AsmToken::String), "expected string")) 3097 return true; 3098 3099 Data = ""; 3100 StringRef Str = getTok().getStringContents(); 3101 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 3102 if (Str[i] != '\\') { 3103 Data += Str[i]; 3104 continue; 3105 } 3106 3107 // Recognize escaped characters. Note that this escape semantics currently 3108 // loosely follows Darwin 'as'. 3109 ++i; 3110 if (i == e) 3111 return TokError("unexpected backslash at end of string"); 3112 3113 // Recognize hex sequences similarly to GNU 'as'. 3114 if (Str[i] == 'x' || Str[i] == 'X') { 3115 size_t length = Str.size(); 3116 if (i + 1 >= length || !isHexDigit(Str[i + 1])) 3117 return TokError("invalid hexadecimal escape sequence"); 3118 3119 // Consume hex characters. GNU 'as' reads all hexadecimal characters and 3120 // then truncates to the lower 16 bits. Seems reasonable. 3121 unsigned Value = 0; 3122 while (i + 1 < length && isHexDigit(Str[i + 1])) 3123 Value = Value * 16 + hexDigitValue(Str[++i]); 3124 3125 Data += (unsigned char)(Value & 0xFF); 3126 continue; 3127 } 3128 3129 // Recognize octal sequences. 3130 if ((unsigned)(Str[i] - '0') <= 7) { 3131 // Consume up to three octal characters. 3132 unsigned Value = Str[i] - '0'; 3133 3134 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3135 ++i; 3136 Value = Value * 8 + (Str[i] - '0'); 3137 3138 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3139 ++i; 3140 Value = Value * 8 + (Str[i] - '0'); 3141 } 3142 } 3143 3144 if (Value > 255) 3145 return TokError("invalid octal escape sequence (out of range)"); 3146 3147 Data += (unsigned char)Value; 3148 continue; 3149 } 3150 3151 // Otherwise recognize individual escapes. 3152 switch (Str[i]) { 3153 default: 3154 // Just reject invalid escape sequences for now. 3155 return TokError("invalid escape sequence (unrecognized character)"); 3156 3157 case 'b': Data += '\b'; break; 3158 case 'f': Data += '\f'; break; 3159 case 'n': Data += '\n'; break; 3160 case 'r': Data += '\r'; break; 3161 case 't': Data += '\t'; break; 3162 case '"': Data += '"'; break; 3163 case '\\': Data += '\\'; break; 3164 } 3165 } 3166 3167 Lex(); 3168 return false; 3169 } 3170 3171 bool MasmParser::parseAngleBracketString(std::string &Data) { 3172 SMLoc EndLoc, StartLoc = getTok().getLoc(); 3173 if (isAngleBracketString(StartLoc, EndLoc)) { 3174 const char *StartChar = StartLoc.getPointer() + 1; 3175 const char *EndChar = EndLoc.getPointer() - 1; 3176 jumpToLoc(EndLoc, CurBuffer); 3177 // Eat from '<' to '>'. 3178 Lex(); 3179 3180 Data = angleBracketString(StringRef(StartChar, EndChar - StartChar)); 3181 return false; 3182 } 3183 return true; 3184 } 3185 3186 /// textItem ::= textLiteral | textMacroID | % constExpr 3187 bool MasmParser::parseTextItem(std::string &Data) { 3188 // TODO(epastor): Support textMacroID and % expansion of expressions. 3189 return parseAngleBracketString(Data); 3190 } 3191 3192 /// parseDirectiveAscii: 3193 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 3194 bool MasmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 3195 auto parseOp = [&]() -> bool { 3196 std::string Data; 3197 if (checkForValidSection() || parseEscapedString(Data)) 3198 return true; 3199 getStreamer().emitBytes(Data); 3200 if (ZeroTerminated) 3201 getStreamer().emitBytes(StringRef("\0", 1)); 3202 return false; 3203 }; 3204 3205 if (parseMany(parseOp)) 3206 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3207 return false; 3208 } 3209 3210 bool MasmParser::emitIntValue(const MCExpr *Value, unsigned Size) { 3211 // Special case constant expressions to match code generator. 3212 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3213 assert(Size <= 8 && "Invalid size"); 3214 int64_t IntValue = MCE->getValue(); 3215 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 3216 return Error(MCE->getLoc(), "out of range literal value"); 3217 getStreamer().emitIntValue(IntValue, Size); 3218 } else { 3219 const MCSymbolRefExpr *MSE = dyn_cast<MCSymbolRefExpr>(Value); 3220 if (MSE && MSE->getSymbol().getName() == "?") { 3221 // ? initializer; treat as 0. 3222 getStreamer().emitIntValue(0, Size); 3223 } else { 3224 getStreamer().emitValue(Value, Size, Value->getLoc()); 3225 } 3226 } 3227 return false; 3228 } 3229 3230 bool MasmParser::parseScalarInitializer(unsigned Size, 3231 SmallVectorImpl<const MCExpr *> &Values, 3232 unsigned StringPadLength) { 3233 if (getTok().is(AsmToken::String)) { 3234 StringRef Value = getTok().getStringContents(); 3235 if (Size == 1) { 3236 // Treat each character as an initializer. 3237 for (const char CharVal : Value) 3238 Values.push_back(MCConstantExpr::create(CharVal, getContext())); 3239 3240 // Pad the string with spaces to the specified length. 3241 for (size_t i = Value.size(); i < StringPadLength; ++i) 3242 Values.push_back(MCConstantExpr::create(' ', getContext())); 3243 } else { 3244 // Treat the string as an initial value in big-endian representation. 3245 if (Value.size() > Size) 3246 return Error(getTok().getLoc(), "out of range literal value"); 3247 3248 uint64_t IntValue = 0; 3249 for (const unsigned char CharVal : Value.bytes()) 3250 IntValue = (IntValue << 8) | CharVal; 3251 Values.push_back(MCConstantExpr::create(IntValue, getContext())); 3252 } 3253 Lex(); 3254 } else { 3255 const MCExpr *Value; 3256 if (parseExpression(Value)) 3257 return true; 3258 if (getTok().is(AsmToken::Identifier) && 3259 getTok().getString().equals_lower("dup")) { 3260 Lex(); // Eat 'dup'. 3261 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 3262 if (!MCE) 3263 return Error(Value->getLoc(), 3264 "cannot repeat value a non-constant number of times"); 3265 const int64_t Repetitions = MCE->getValue(); 3266 if (Repetitions < 0) 3267 return Error(Value->getLoc(), 3268 "cannot repeat value a negative number of times"); 3269 3270 SmallVector<const MCExpr *, 1> DuplicatedValues; 3271 if (parseToken(AsmToken::LParen, 3272 "parentheses required for 'dup' contents") || 3273 parseScalarInstList(Size, DuplicatedValues) || 3274 parseToken(AsmToken::RParen, "unmatched parentheses")) 3275 return true; 3276 3277 for (int i = 0; i < Repetitions; ++i) 3278 Values.append(DuplicatedValues.begin(), DuplicatedValues.end()); 3279 } else { 3280 Values.push_back(Value); 3281 } 3282 } 3283 return false; 3284 } 3285 3286 bool MasmParser::parseScalarInstList(unsigned Size, 3287 SmallVectorImpl<const MCExpr *> &Values, 3288 const AsmToken::TokenKind EndToken) { 3289 while (getTok().isNot(EndToken) && 3290 (EndToken != AsmToken::Greater || 3291 getTok().isNot(AsmToken::GreaterGreater))) { 3292 parseScalarInitializer(Size, Values); 3293 3294 // If we see a comma, continue, and allow line continuation. 3295 if (!parseOptionalToken(AsmToken::Comma)) 3296 break; 3297 parseOptionalToken(AsmToken::EndOfStatement); 3298 } 3299 return false; 3300 } 3301 3302 bool MasmParser::emitIntegralValues(unsigned Size) { 3303 SmallVector<const MCExpr *, 1> Values; 3304 if (checkForValidSection() || parseScalarInstList(Size, Values)) 3305 return true; 3306 3307 for (auto Value : Values) { 3308 emitIntValue(Value, Size); 3309 } 3310 return false; 3311 } 3312 3313 // Add a field to the current structure. 3314 bool MasmParser::addIntegralField(StringRef Name, unsigned Size) { 3315 StructInfo &Struct = StructInProgress.back(); 3316 FieldInfo &Field = Struct.addField(Name, FT_INTEGRAL, Size); 3317 IntFieldInfo &IntInfo = Field.Contents.IntInfo; 3318 3319 Field.Type = Size; 3320 3321 if (parseScalarInstList(Size, IntInfo.Values)) 3322 return true; 3323 3324 Field.SizeOf = Field.Type * IntInfo.Values.size(); 3325 Field.LengthOf = IntInfo.Values.size(); 3326 if (Struct.IsUnion) 3327 Struct.Size = std::max(Struct.Size, Field.SizeOf); 3328 else 3329 Struct.Size += Field.SizeOf; 3330 return false; 3331 } 3332 3333 /// parseDirectiveValue 3334 /// ::= (byte | word | ... ) [ expression (, expression)* ] 3335 bool MasmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { 3336 if (StructInProgress.empty()) { 3337 // Initialize data value. 3338 if (emitIntegralValues(Size)) 3339 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3340 } else if (addIntegralField("", Size)) { 3341 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3342 } 3343 3344 return false; 3345 } 3346 3347 /// parseDirectiveNamedValue 3348 /// ::= name (byte | word | ... ) [ expression (, expression)* ] 3349 bool MasmParser::parseDirectiveNamedValue(StringRef IDVal, unsigned Size, 3350 StringRef Name, SMLoc NameLoc) { 3351 if (StructInProgress.empty()) { 3352 // Initialize named data value. 3353 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 3354 getStreamer().emitLabel(Sym); 3355 if (emitIntegralValues(Size)) 3356 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3357 } else if (addIntegralField(Name, Size)) { 3358 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3359 } 3360 3361 return false; 3362 } 3363 3364 static bool parseHexOcta(MasmParser &Asm, uint64_t &hi, uint64_t &lo) { 3365 if (Asm.getTok().isNot(AsmToken::Integer) && 3366 Asm.getTok().isNot(AsmToken::BigNum)) 3367 return Asm.TokError("unknown token in expression"); 3368 SMLoc ExprLoc = Asm.getTok().getLoc(); 3369 APInt IntValue = Asm.getTok().getAPIntVal(); 3370 Asm.Lex(); 3371 if (!IntValue.isIntN(128)) 3372 return Asm.Error(ExprLoc, "out of range literal value"); 3373 if (!IntValue.isIntN(64)) { 3374 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue(); 3375 lo = IntValue.getLoBits(64).getZExtValue(); 3376 } else { 3377 hi = 0; 3378 lo = IntValue.getZExtValue(); 3379 } 3380 return false; 3381 } 3382 3383 bool MasmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { 3384 // We don't truly support arithmetic on floating point expressions, so we 3385 // have to manually parse unary prefixes. 3386 bool IsNeg = false; 3387 if (getLexer().is(AsmToken::Minus)) { 3388 Lexer.Lex(); 3389 IsNeg = true; 3390 } else if (getLexer().is(AsmToken::Plus)) { 3391 Lexer.Lex(); 3392 } 3393 3394 if (Lexer.is(AsmToken::Error)) 3395 return TokError(Lexer.getErr()); 3396 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) && 3397 Lexer.isNot(AsmToken::Identifier)) 3398 return TokError("unexpected token in directive"); 3399 3400 // Convert to an APFloat. 3401 APFloat Value(Semantics); 3402 StringRef IDVal = getTok().getString(); 3403 if (getLexer().is(AsmToken::Identifier)) { 3404 if (IDVal.equals_lower("infinity") || IDVal.equals_lower("inf")) 3405 Value = APFloat::getInf(Semantics); 3406 else if (IDVal.equals_lower("nan")) 3407 Value = APFloat::getNaN(Semantics, false, ~0); 3408 else if (IDVal.equals_lower("?")) 3409 Value = APFloat::getZero(Semantics); 3410 else 3411 return TokError("invalid floating point literal"); 3412 } else if (errorToBool( 3413 Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) 3414 .takeError())) { 3415 return TokError("invalid floating point literal"); 3416 } 3417 if (IsNeg) 3418 Value.changeSign(); 3419 3420 // Consume the numeric token. 3421 Lex(); 3422 3423 Res = Value.bitcastToAPInt(); 3424 3425 return false; 3426 } 3427 3428 bool MasmParser::parseRealInstList(const fltSemantics &Semantics, 3429 SmallVectorImpl<APInt> &ValuesAsInt, 3430 const AsmToken::TokenKind EndToken) { 3431 while (getTok().isNot(EndToken) || 3432 (EndToken == AsmToken::Greater && 3433 getTok().isNot(AsmToken::GreaterGreater))) { 3434 const AsmToken NextTok = Lexer.peekTok(); 3435 if (NextTok.is(AsmToken::Identifier) && 3436 NextTok.getString().equals_lower("dup")) { 3437 const MCExpr *Value; 3438 if (parseExpression(Value) || parseToken(AsmToken::Identifier)) 3439 return true; 3440 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 3441 if (!MCE) 3442 return Error(Value->getLoc(), 3443 "cannot repeat value a non-constant number of times"); 3444 const int64_t Repetitions = MCE->getValue(); 3445 if (Repetitions < 0) 3446 return Error(Value->getLoc(), 3447 "cannot repeat value a negative number of times"); 3448 3449 SmallVector<APInt, 1> DuplicatedValues; 3450 if (parseToken(AsmToken::LParen, 3451 "parentheses required for 'dup' contents") || 3452 parseRealInstList(Semantics, DuplicatedValues) || 3453 parseToken(AsmToken::RParen, "unmatched parentheses")) 3454 return true; 3455 3456 for (int i = 0; i < Repetitions; ++i) 3457 ValuesAsInt.append(DuplicatedValues.begin(), DuplicatedValues.end()); 3458 } else { 3459 APInt AsInt; 3460 if (parseRealValue(Semantics, AsInt)) 3461 return true; 3462 ValuesAsInt.push_back(AsInt); 3463 } 3464 3465 // Continue if we see a comma. (Also, allow line continuation.) 3466 if (!parseOptionalToken(AsmToken::Comma)) 3467 break; 3468 parseOptionalToken(AsmToken::EndOfStatement); 3469 } 3470 3471 return false; 3472 } 3473 3474 // Initialize real data values. 3475 bool MasmParser::emitRealValues(const fltSemantics &Semantics) { 3476 if (checkForValidSection()) 3477 return true; 3478 3479 SmallVector<APInt, 1> ValuesAsInt; 3480 if (parseRealInstList(Semantics, ValuesAsInt)) 3481 return true; 3482 3483 for (const APInt &AsInt : ValuesAsInt) { 3484 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3485 AsInt.getBitWidth() / 8); 3486 } 3487 return false; 3488 } 3489 3490 // Add a real field to the current struct. 3491 bool MasmParser::addRealField(StringRef Name, const fltSemantics &Semantics, 3492 size_t Size) { 3493 StructInfo &Struct = StructInProgress.back(); 3494 FieldInfo &Field = Struct.addField(Name, FT_REAL, Size); 3495 RealFieldInfo &RealInfo = Field.Contents.RealInfo; 3496 3497 Field.SizeOf = 0; 3498 3499 if (parseRealInstList(Semantics, RealInfo.AsIntValues)) 3500 return true; 3501 3502 Field.Type = RealInfo.AsIntValues.back().getBitWidth() / 8; 3503 Field.LengthOf = RealInfo.AsIntValues.size(); 3504 Field.SizeOf = Field.Type * Field.LengthOf; 3505 if (Struct.IsUnion) 3506 Struct.Size = std::max(Struct.Size, Field.SizeOf); 3507 else 3508 Struct.Size += Field.SizeOf; 3509 return false; 3510 } 3511 3512 /// parseDirectiveRealValue 3513 /// ::= (real4 | real8) [ expression (, expression)* ] 3514 bool MasmParser::parseDirectiveRealValue(StringRef IDVal, 3515 const fltSemantics &Semantics, 3516 size_t Size) { 3517 if (StructInProgress.empty()) { 3518 // Initialize data value. 3519 if (emitRealValues(Semantics)) 3520 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3521 } else if (addRealField("", Semantics, Size)) { 3522 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3523 } 3524 return false; 3525 } 3526 3527 /// parseDirectiveNamedRealValue 3528 /// ::= name (real4 | real8) [ expression (, expression)* ] 3529 bool MasmParser::parseDirectiveNamedRealValue(StringRef IDVal, 3530 const fltSemantics &Semantics, 3531 size_t Size, StringRef Name, 3532 SMLoc NameLoc) { 3533 if (StructInProgress.empty()) { 3534 // Initialize named data value. 3535 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 3536 getStreamer().emitLabel(Sym); 3537 if (emitRealValues(Semantics)) 3538 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3539 } else if (addRealField(Name, Semantics, Size)) { 3540 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3541 } 3542 return false; 3543 } 3544 3545 bool MasmParser::parseOptionalAngleBracketOpen() { 3546 const AsmToken Tok = getTok(); 3547 if (parseOptionalToken(AsmToken::LessLess)) { 3548 AngleBracketDepth++; 3549 Lexer.UnLex(AsmToken(AsmToken::Less, Tok.getString().substr(1))); 3550 return true; 3551 } else if (parseOptionalToken(AsmToken::LessGreater)) { 3552 AngleBracketDepth++; 3553 Lexer.UnLex(AsmToken(AsmToken::Greater, Tok.getString().substr(1))); 3554 return true; 3555 } else if (parseOptionalToken(AsmToken::Less)) { 3556 AngleBracketDepth++; 3557 return true; 3558 } 3559 3560 return false; 3561 } 3562 3563 bool MasmParser::parseAngleBracketClose(const Twine &Msg) { 3564 const AsmToken Tok = getTok(); 3565 if (parseOptionalToken(AsmToken::GreaterGreater)) { 3566 Lexer.UnLex(AsmToken(AsmToken::Greater, Tok.getString().substr(1))); 3567 } else if (parseToken(AsmToken::Greater, Msg)) { 3568 return true; 3569 } 3570 AngleBracketDepth--; 3571 return false; 3572 } 3573 3574 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3575 const IntFieldInfo &Contents, 3576 FieldInitializer &Initializer) { 3577 SMLoc Loc = getTok().getLoc(); 3578 3579 SmallVector<const MCExpr *, 1> Values; 3580 if (parseOptionalToken(AsmToken::LCurly)) { 3581 if (Field.LengthOf == 1 && Field.Type > 1) 3582 return Error(Loc, "Cannot initialize scalar field with array value"); 3583 if (parseScalarInstList(Field.Type, Values, AsmToken::RCurly) || 3584 parseToken(AsmToken::RCurly)) 3585 return true; 3586 } else if (parseOptionalAngleBracketOpen()) { 3587 if (Field.LengthOf == 1 && Field.Type > 1) 3588 return Error(Loc, "Cannot initialize scalar field with array value"); 3589 if (parseScalarInstList(Field.Type, Values, AsmToken::Greater) || 3590 parseAngleBracketClose()) 3591 return true; 3592 } else if (Field.LengthOf > 1 && Field.Type > 1) { 3593 return Error(Loc, "Cannot initialize array field with scalar value"); 3594 } else if (parseScalarInitializer(Field.Type, Values, 3595 /*StringPadLength=*/Field.LengthOf)) { 3596 return true; 3597 } 3598 3599 if (Values.size() > Field.LengthOf) { 3600 return Error(Loc, "Initializer too long for field; expected at most " + 3601 std::to_string(Field.LengthOf) + " elements, got " + 3602 std::to_string(Values.size())); 3603 } 3604 // Default-initialize all remaining values. 3605 Values.append(Contents.Values.begin() + Values.size(), Contents.Values.end()); 3606 3607 Initializer = FieldInitializer(std::move(Values)); 3608 return false; 3609 } 3610 3611 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3612 const RealFieldInfo &Contents, 3613 FieldInitializer &Initializer) { 3614 const fltSemantics &Semantics = 3615 (Field.Type == 4) ? APFloat::IEEEsingle() : APFloat::IEEEdouble(); 3616 3617 SMLoc Loc = getTok().getLoc(); 3618 3619 SmallVector<APInt, 1> AsIntValues; 3620 if (parseOptionalToken(AsmToken::LCurly)) { 3621 if (Field.LengthOf == 1) 3622 return Error(Loc, "Cannot initialize scalar field with array value"); 3623 if (parseRealInstList(Semantics, AsIntValues, AsmToken::RCurly) || 3624 parseToken(AsmToken::RCurly)) 3625 return true; 3626 } else if (parseOptionalAngleBracketOpen()) { 3627 if (Field.LengthOf == 1) 3628 return Error(Loc, "Cannot initialize scalar field with array value"); 3629 if (parseRealInstList(Semantics, AsIntValues, AsmToken::Greater) || 3630 parseAngleBracketClose()) 3631 return true; 3632 } else if (Field.LengthOf > 1) { 3633 return Error(Loc, "Cannot initialize array field with scalar value"); 3634 } else { 3635 AsIntValues.emplace_back(); 3636 if (parseRealValue(Semantics, AsIntValues.back())) 3637 return true; 3638 } 3639 3640 if (AsIntValues.size() > Field.LengthOf) { 3641 return Error(Loc, "Initializer too long for field; expected at most " + 3642 std::to_string(Field.LengthOf) + " elements, got " + 3643 std::to_string(AsIntValues.size())); 3644 } 3645 // Default-initialize all remaining values. 3646 AsIntValues.append(Contents.AsIntValues.begin() + AsIntValues.size(), 3647 Contents.AsIntValues.end()); 3648 3649 Initializer = FieldInitializer(std::move(AsIntValues)); 3650 return false; 3651 } 3652 3653 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3654 const StructFieldInfo &Contents, 3655 FieldInitializer &Initializer) { 3656 SMLoc Loc = getTok().getLoc(); 3657 3658 std::vector<StructInitializer> Initializers; 3659 if (Field.LengthOf > 1) { 3660 if (parseOptionalToken(AsmToken::LCurly)) { 3661 if (parseStructInstList(Contents.Structure, Initializers, 3662 AsmToken::RCurly) || 3663 parseToken(AsmToken::RCurly)) 3664 return true; 3665 } else if (parseOptionalAngleBracketOpen()) { 3666 if (parseStructInstList(Contents.Structure, Initializers, 3667 AsmToken::Greater) || 3668 parseAngleBracketClose()) 3669 return true; 3670 } else { 3671 return Error(Loc, "Cannot initialize array field with scalar value"); 3672 } 3673 } else { 3674 Initializers.emplace_back(); 3675 if (parseStructInitializer(Contents.Structure, Initializers.back())) 3676 return true; 3677 } 3678 3679 if (Initializers.size() > Field.LengthOf) { 3680 return Error(Loc, "Initializer too long for field; expected at most " + 3681 std::to_string(Field.LengthOf) + " elements, got " + 3682 std::to_string(Initializers.size())); 3683 } 3684 // Default-initialize all remaining values. 3685 Initializers.insert(Initializers.end(), 3686 Contents.Initializers.begin() + Initializers.size(), 3687 Contents.Initializers.end()); 3688 3689 Initializer = FieldInitializer(std::move(Initializers), Contents.Structure); 3690 return false; 3691 } 3692 3693 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3694 FieldInitializer &Initializer) { 3695 switch (Field.Contents.FT) { 3696 case FT_INTEGRAL: 3697 return parseFieldInitializer(Field, Field.Contents.IntInfo, Initializer); 3698 case FT_REAL: 3699 return parseFieldInitializer(Field, Field.Contents.RealInfo, Initializer); 3700 case FT_STRUCT: 3701 return parseFieldInitializer(Field, Field.Contents.StructInfo, Initializer); 3702 } 3703 llvm_unreachable("Unhandled FieldType enum"); 3704 } 3705 3706 bool MasmParser::parseStructInitializer(const StructInfo &Structure, 3707 StructInitializer &Initializer) { 3708 const AsmToken FirstToken = getTok(); 3709 3710 Optional<AsmToken::TokenKind> EndToken; 3711 if (parseOptionalToken(AsmToken::LCurly)) { 3712 EndToken = AsmToken::RCurly; 3713 } else if (parseOptionalAngleBracketOpen()) { 3714 EndToken = AsmToken::Greater; 3715 AngleBracketDepth++; 3716 } else if (FirstToken.is(AsmToken::Identifier) && 3717 FirstToken.getString() == "?") { 3718 // ? initializer; leave EndToken uninitialized to treat as empty. 3719 if (parseToken(AsmToken::Identifier)) 3720 return true; 3721 } else { 3722 return Error(FirstToken.getLoc(), "Expected struct initializer"); 3723 } 3724 3725 auto &FieldInitializers = Initializer.FieldInitializers; 3726 size_t FieldIndex = 0; 3727 if (EndToken.hasValue()) { 3728 // Initialize all fields with given initializers. 3729 while (getTok().isNot(EndToken.getValue()) && 3730 FieldIndex < Structure.Fields.size()) { 3731 const FieldInfo &Field = Structure.Fields[FieldIndex++]; 3732 if (parseOptionalToken(AsmToken::Comma)) { 3733 // Empty initializer; use the default and continue. (Also, allow line 3734 // continuation.) 3735 FieldInitializers.push_back(Field.Contents); 3736 parseOptionalToken(AsmToken::EndOfStatement); 3737 continue; 3738 } 3739 FieldInitializers.emplace_back(Field.Contents.FT); 3740 if (parseFieldInitializer(Field, FieldInitializers.back())) 3741 return true; 3742 3743 // Continue if we see a comma. (Also, allow line continuation.) 3744 SMLoc CommaLoc = getTok().getLoc(); 3745 if (!parseOptionalToken(AsmToken::Comma)) 3746 break; 3747 if (FieldIndex == Structure.Fields.size()) 3748 return Error(CommaLoc, "'" + Structure.Name + 3749 "' initializer initializes too many fields"); 3750 parseOptionalToken(AsmToken::EndOfStatement); 3751 } 3752 } 3753 // Default-initialize all remaining fields. 3754 for (auto It = Structure.Fields.begin() + FieldIndex; 3755 It != Structure.Fields.end(); ++It) { 3756 const FieldInfo &Field = *It; 3757 FieldInitializers.push_back(Field.Contents); 3758 } 3759 3760 if (EndToken.hasValue()) { 3761 if (EndToken.getValue() == AsmToken::Greater) 3762 return parseAngleBracketClose(); 3763 3764 return parseToken(EndToken.getValue()); 3765 } 3766 3767 return false; 3768 } 3769 3770 bool MasmParser::parseStructInstList( 3771 const StructInfo &Structure, std::vector<StructInitializer> &Initializers, 3772 const AsmToken::TokenKind EndToken) { 3773 while (getTok().isNot(EndToken) || 3774 (EndToken == AsmToken::Greater && 3775 getTok().isNot(AsmToken::GreaterGreater))) { 3776 const AsmToken NextTok = Lexer.peekTok(); 3777 if (NextTok.is(AsmToken::Identifier) && 3778 NextTok.getString().equals_lower("dup")) { 3779 const MCExpr *Value; 3780 if (parseExpression(Value) || parseToken(AsmToken::Identifier)) 3781 return true; 3782 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 3783 if (!MCE) 3784 return Error(Value->getLoc(), 3785 "cannot repeat value a non-constant number of times"); 3786 const int64_t Repetitions = MCE->getValue(); 3787 if (Repetitions < 0) 3788 return Error(Value->getLoc(), 3789 "cannot repeat value a negative number of times"); 3790 3791 std::vector<StructInitializer> DuplicatedValues; 3792 if (parseToken(AsmToken::LParen, 3793 "parentheses required for 'dup' contents") || 3794 parseStructInstList(Structure, DuplicatedValues) || 3795 parseToken(AsmToken::RParen, "unmatched parentheses")) 3796 return true; 3797 3798 for (int i = 0; i < Repetitions; ++i) 3799 Initializers.insert(Initializers.end(), DuplicatedValues.begin(), 3800 DuplicatedValues.end()); 3801 } else { 3802 Initializers.emplace_back(); 3803 if (parseStructInitializer(Structure, Initializers.back())) 3804 return true; 3805 } 3806 3807 // Continue if we see a comma. (Also, allow line continuation.) 3808 if (!parseOptionalToken(AsmToken::Comma)) 3809 break; 3810 parseOptionalToken(AsmToken::EndOfStatement); 3811 } 3812 3813 return false; 3814 } 3815 3816 bool MasmParser::emitFieldValue(const FieldInfo &Field, 3817 const IntFieldInfo &Contents) { 3818 // Default-initialize all values. 3819 for (const MCExpr *Value : Contents.Values) { 3820 if (emitIntValue(Value, Field.Type)) 3821 return true; 3822 } 3823 return false; 3824 } 3825 3826 bool MasmParser::emitFieldValue(const FieldInfo &Field, 3827 const RealFieldInfo &Contents) { 3828 for (const APInt &AsInt : Contents.AsIntValues) { 3829 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3830 AsInt.getBitWidth() / 8); 3831 } 3832 return false; 3833 } 3834 3835 bool MasmParser::emitFieldValue(const FieldInfo &Field, 3836 const StructFieldInfo &Contents) { 3837 for (const auto &Initializer : Contents.Initializers) { 3838 size_t Index = 0, Offset = 0; 3839 for (const auto &SubField : Contents.Structure.Fields) { 3840 getStreamer().emitZeros(SubField.Offset - Offset); 3841 Offset = SubField.Offset + SubField.SizeOf; 3842 emitFieldInitializer(SubField, Initializer.FieldInitializers[Index++]); 3843 } 3844 } 3845 return false; 3846 } 3847 3848 bool MasmParser::emitFieldValue(const FieldInfo &Field) { 3849 switch (Field.Contents.FT) { 3850 case FT_INTEGRAL: 3851 return emitFieldValue(Field, Field.Contents.IntInfo); 3852 case FT_REAL: 3853 return emitFieldValue(Field, Field.Contents.RealInfo); 3854 case FT_STRUCT: 3855 return emitFieldValue(Field, Field.Contents.StructInfo); 3856 } 3857 llvm_unreachable("Unhandled FieldType enum"); 3858 } 3859 3860 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3861 const IntFieldInfo &Contents, 3862 const IntFieldInfo &Initializer) { 3863 for (const auto &Value : Initializer.Values) { 3864 if (emitIntValue(Value, Field.Type)) 3865 return true; 3866 } 3867 // Default-initialize all remaining values. 3868 for (auto it = Contents.Values.begin() + Initializer.Values.size(); 3869 it != Contents.Values.end(); ++it) { 3870 const auto &Value = *it; 3871 if (emitIntValue(Value, Field.Type)) 3872 return true; 3873 } 3874 return false; 3875 } 3876 3877 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3878 const RealFieldInfo &Contents, 3879 const RealFieldInfo &Initializer) { 3880 for (const auto &AsInt : Initializer.AsIntValues) { 3881 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3882 AsInt.getBitWidth() / 8); 3883 } 3884 // Default-initialize all remaining values. 3885 for (auto It = Contents.AsIntValues.begin() + Initializer.AsIntValues.size(); 3886 It != Contents.AsIntValues.end(); ++It) { 3887 const auto &AsInt = *It; 3888 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3889 AsInt.getBitWidth() / 8); 3890 } 3891 return false; 3892 } 3893 3894 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3895 const StructFieldInfo &Contents, 3896 const StructFieldInfo &Initializer) { 3897 for (const auto &Init : Initializer.Initializers) { 3898 emitStructInitializer(Contents.Structure, Init); 3899 } 3900 // Default-initialize all remaining values. 3901 for (auto It = 3902 Contents.Initializers.begin() + Initializer.Initializers.size(); 3903 It != Contents.Initializers.end(); ++It) { 3904 const auto &Init = *It; 3905 emitStructInitializer(Contents.Structure, Init); 3906 } 3907 return false; 3908 } 3909 3910 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3911 const FieldInitializer &Initializer) { 3912 switch (Field.Contents.FT) { 3913 case FT_INTEGRAL: 3914 return emitFieldInitializer(Field, Field.Contents.IntInfo, 3915 Initializer.IntInfo); 3916 case FT_REAL: 3917 return emitFieldInitializer(Field, Field.Contents.RealInfo, 3918 Initializer.RealInfo); 3919 case FT_STRUCT: 3920 return emitFieldInitializer(Field, Field.Contents.StructInfo, 3921 Initializer.StructInfo); 3922 } 3923 llvm_unreachable("Unhandled FieldType enum"); 3924 } 3925 3926 bool MasmParser::emitStructInitializer(const StructInfo &Structure, 3927 const StructInitializer &Initializer) { 3928 size_t Index = 0, Offset = 0; 3929 for (const auto &Init : Initializer.FieldInitializers) { 3930 const auto &Field = Structure.Fields[Index++]; 3931 getStreamer().emitZeros(Field.Offset - Offset); 3932 Offset = Field.Offset + Field.SizeOf; 3933 if (emitFieldInitializer(Field, Init)) 3934 return true; 3935 } 3936 // Default-initialize all remaining fields. 3937 for (auto It = 3938 Structure.Fields.begin() + Initializer.FieldInitializers.size(); 3939 It != Structure.Fields.end(); ++It) { 3940 const auto &Field = *It; 3941 getStreamer().emitZeros(Field.Offset - Offset); 3942 Offset = Field.Offset + Field.SizeOf; 3943 if (emitFieldValue(Field)) 3944 return true; 3945 } 3946 // Add final padding. 3947 if (Offset != Structure.Size) 3948 getStreamer().emitZeros(Structure.Size - Offset); 3949 return false; 3950 } 3951 3952 // Set data values from initializers. 3953 bool MasmParser::emitStructValues(const StructInfo &Structure) { 3954 std::vector<StructInitializer> Initializers; 3955 if (parseStructInstList(Structure, Initializers)) 3956 return true; 3957 3958 for (const auto &Initializer : Initializers) { 3959 if (emitStructInitializer(Structure, Initializer)) 3960 return true; 3961 } 3962 3963 return false; 3964 } 3965 3966 // Declare a field in the current struct. 3967 bool MasmParser::addStructField(StringRef Name, const StructInfo &Structure) { 3968 StructInfo &OwningStruct = StructInProgress.back(); 3969 FieldInfo &Field = OwningStruct.addField(Name, FT_STRUCT, Structure.Size); 3970 StructFieldInfo &StructInfo = Field.Contents.StructInfo; 3971 3972 StructInfo.Structure = Structure; 3973 Field.Type = Structure.Size; 3974 3975 if (parseStructInstList(Structure, StructInfo.Initializers)) 3976 return true; 3977 3978 Field.LengthOf = StructInfo.Initializers.size(); 3979 Field.SizeOf = Field.Type * Field.LengthOf; 3980 if (OwningStruct.IsUnion) 3981 OwningStruct.Size = std::max(OwningStruct.Size, Field.SizeOf); 3982 else 3983 OwningStruct.Size += Field.SizeOf; 3984 3985 return false; 3986 } 3987 3988 /// parseDirectiveStructValue 3989 /// ::= struct-id (<struct-initializer> | {struct-initializer}) 3990 /// [, (<struct-initializer> | {struct-initializer})]* 3991 bool MasmParser::parseDirectiveStructValue(const StructInfo &Structure, 3992 StringRef Directive, SMLoc DirLoc) { 3993 if (StructInProgress.empty()) { 3994 if (emitStructValues(Structure)) 3995 return true; 3996 } else if (addStructField("", Structure)) { 3997 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 3998 } 3999 4000 return false; 4001 } 4002 4003 /// parseDirectiveNamedValue 4004 /// ::= name (byte | word | ... ) [ expression (, expression)* ] 4005 bool MasmParser::parseDirectiveNamedStructValue(const StructInfo &Structure, 4006 StringRef Directive, 4007 SMLoc DirLoc, StringRef Name) { 4008 if (StructInProgress.empty()) { 4009 // Initialize named data value. 4010 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4011 getStreamer().emitLabel(Sym); 4012 KnownType[Name] = &Structure; 4013 if (emitStructValues(Structure)) 4014 return true; 4015 } else if (addStructField(Name, Structure)) { 4016 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4017 } 4018 4019 return false; 4020 } 4021 4022 /// parseDirectiveStruct 4023 /// ::= <name> (STRUC | STRUCT | UNION) [fieldAlign] [, NONUNIQUE] 4024 /// (dataDir | generalDir | offsetDir | nestedStruct)+ 4025 /// <name> ENDS 4026 ////// dataDir = data declaration 4027 ////// offsetDir = EVEN, ORG, ALIGN 4028 bool MasmParser::parseDirectiveStruct(StringRef Directive, 4029 DirectiveKind DirKind, StringRef Name, 4030 SMLoc NameLoc) { 4031 // We ignore NONUNIQUE; we do not support OPTION M510 or OPTION OLDSTRUCTS 4032 // anyway, so all field accesses must be qualified. 4033 AsmToken NextTok = getTok(); 4034 int64_t AlignmentValue = 1; 4035 if (NextTok.isNot(AsmToken::Comma) && 4036 NextTok.isNot(AsmToken::EndOfStatement) && 4037 parseAbsoluteExpression(AlignmentValue)) { 4038 return addErrorSuffix(" in alignment value for '" + Twine(Directive) + 4039 "' directive"); 4040 } 4041 if (!isPowerOf2_64(AlignmentValue)) { 4042 return Error(NextTok.getLoc(), "alignment must be a power of two; was " + 4043 std::to_string(AlignmentValue)); 4044 } 4045 4046 StringRef Qualifier; 4047 SMLoc QualifierLoc; 4048 if (parseOptionalToken(AsmToken::Comma)) { 4049 QualifierLoc = getTok().getLoc(); 4050 if (parseIdentifier(Qualifier)) 4051 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4052 if (!Qualifier.equals_lower("nonunique")) 4053 return Error(QualifierLoc, "Unrecognized qualifier for '" + 4054 Twine(Directive) + 4055 "' directive; expected none or NONUNIQUE"); 4056 } 4057 4058 if (parseToken(AsmToken::EndOfStatement)) 4059 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4060 4061 StructInProgress.emplace_back(Name, DirKind == DK_UNION, AlignmentValue); 4062 return false; 4063 } 4064 4065 /// parseDirectiveNestedStruct 4066 /// ::= (STRUC | STRUCT | UNION) [name] 4067 /// (dataDir | generalDir | offsetDir | nestedStruct)+ 4068 /// ENDS 4069 bool MasmParser::parseDirectiveNestedStruct(StringRef Directive, 4070 DirectiveKind DirKind) { 4071 if (StructInProgress.empty()) 4072 return TokError("missing name in top-level '" + Twine(Directive) + 4073 "' directive"); 4074 4075 StringRef Name; 4076 if (getTok().is(AsmToken::Identifier)) { 4077 Name = getTok().getIdentifier(); 4078 parseToken(AsmToken::Identifier); 4079 } 4080 if (parseToken(AsmToken::EndOfStatement)) 4081 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4082 4083 StructInProgress.emplace_back(Name, DirKind == DK_UNION, 4084 StructInProgress.back().Alignment); 4085 return false; 4086 } 4087 4088 bool MasmParser::parseDirectiveEnds(StringRef Name, SMLoc NameLoc) { 4089 if (StructInProgress.empty()) 4090 return Error(NameLoc, "ENDS directive without matching STRUC/STRUCT/UNION"); 4091 if (StructInProgress.size() > 1) 4092 return Error(NameLoc, "unexpected name in nested ENDS directive"); 4093 if (StructInProgress.back().Name.compare_lower(Name)) 4094 return Error(NameLoc, "mismatched name in ENDS directive; expected '" + 4095 StructInProgress.back().Name + "'"); 4096 StructInfo Structure = StructInProgress.pop_back_val(); 4097 // Pad to make the structure's size divisible by its alignment. 4098 Structure.Size = llvm::alignTo(Structure.Size, Structure.Alignment); 4099 Structs[Name.lower()] = Structure; 4100 4101 if (parseToken(AsmToken::EndOfStatement)) 4102 return addErrorSuffix(" in ENDS directive"); 4103 4104 return false; 4105 } 4106 4107 bool MasmParser::parseDirectiveNestedEnds() { 4108 if (StructInProgress.empty()) 4109 return TokError("ENDS directive without matching STRUC/STRUCT/UNION"); 4110 if (StructInProgress.size() == 1) 4111 return TokError("missing name in top-level ENDS directive"); 4112 4113 if (parseToken(AsmToken::EndOfStatement)) 4114 return addErrorSuffix(" in nested ENDS directive"); 4115 4116 StructInfo Structure = StructInProgress.pop_back_val(); 4117 // Pad to make the structure's size divisible by its alignment. 4118 Structure.Size = llvm::alignTo(Structure.Size, Structure.Alignment); 4119 4120 StructInfo &ParentStruct = StructInProgress.back(); 4121 if (Structure.Name.empty()) { 4122 const size_t OldFields = ParentStruct.Fields.size(); 4123 ParentStruct.Fields.insert( 4124 ParentStruct.Fields.end(), 4125 std::make_move_iterator(Structure.Fields.begin()), 4126 std::make_move_iterator(Structure.Fields.end())); 4127 for (const auto &FieldByName : Structure.FieldsByName) { 4128 ParentStruct.FieldsByName[FieldByName.getKey()] = 4129 FieldByName.getValue() + OldFields; 4130 } 4131 if (!ParentStruct.IsUnion) { 4132 for (auto FieldIter = ParentStruct.Fields.begin() + OldFields; 4133 FieldIter != ParentStruct.Fields.end(); ++FieldIter) { 4134 FieldIter->Offset += ParentStruct.Size; 4135 } 4136 } 4137 4138 if (ParentStruct.IsUnion) 4139 ParentStruct.Size = std::max(ParentStruct.Size, Structure.Size); 4140 else 4141 ParentStruct.Size += Structure.Size; 4142 } else { 4143 FieldInfo &Field = 4144 ParentStruct.addField(Structure.Name, FT_STRUCT, Structure.Size); 4145 StructFieldInfo &StructInfo = Field.Contents.StructInfo; 4146 Field.Type = Structure.Size; 4147 Field.LengthOf = 1; 4148 Field.SizeOf = Structure.Size; 4149 4150 if (ParentStruct.IsUnion) 4151 ParentStruct.Size = std::max(ParentStruct.Size, Field.SizeOf); 4152 else 4153 ParentStruct.Size += Field.SizeOf; 4154 4155 StructInfo.Structure = Structure; 4156 StructInfo.Initializers.emplace_back(); 4157 auto &FieldInitializers = StructInfo.Initializers.back().FieldInitializers; 4158 for (const auto &SubField : Structure.Fields) { 4159 FieldInitializers.push_back(SubField.Contents); 4160 } 4161 } 4162 4163 return false; 4164 } 4165 4166 /// parseDirectiveOrg 4167 /// ::= .org expression [ , expression ] 4168 bool MasmParser::parseDirectiveOrg() { 4169 const MCExpr *Offset; 4170 SMLoc OffsetLoc = Lexer.getLoc(); 4171 if (checkForValidSection() || parseExpression(Offset)) 4172 return true; 4173 4174 // Parse optional fill expression. 4175 int64_t FillExpr = 0; 4176 if (parseOptionalToken(AsmToken::Comma)) 4177 if (parseAbsoluteExpression(FillExpr)) 4178 return addErrorSuffix(" in '.org' directive"); 4179 if (parseToken(AsmToken::EndOfStatement)) 4180 return addErrorSuffix(" in '.org' directive"); 4181 4182 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc); 4183 return false; 4184 } 4185 4186 /// parseDirectiveAlign 4187 /// ::= align expression 4188 bool MasmParser::parseDirectiveAlign() { 4189 SMLoc AlignmentLoc = getLexer().getLoc(); 4190 int64_t Alignment; 4191 4192 if (checkForValidSection()) 4193 return addErrorSuffix(" in align directive"); 4194 // Ignore empty 'align' directives. 4195 if (getTok().is(AsmToken::EndOfStatement)) { 4196 Warning(AlignmentLoc, "align directive with no operand is ignored"); 4197 return parseToken(AsmToken::EndOfStatement); 4198 } 4199 if (parseAbsoluteExpression(Alignment) || 4200 parseToken(AsmToken::EndOfStatement)) 4201 return addErrorSuffix(" in align directive"); 4202 4203 // Always emit an alignment here even if we thrown an error. 4204 bool ReturnVal = false; 4205 4206 // Reject alignments that aren't either a power of two or zero, for gas 4207 // compatibility. Alignment of zero is silently rounded up to one. 4208 if (Alignment == 0) 4209 Alignment = 1; 4210 if (!isPowerOf2_64(Alignment)) 4211 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2"); 4212 4213 // Check whether we should use optimal code alignment for this align 4214 // directive. 4215 const MCSection *Section = getStreamer().getCurrentSectionOnly(); 4216 assert(Section && "must have section to emit alignment"); 4217 if (Section->UseCodeAlign()) { 4218 getStreamer().emitCodeAlignment(Alignment, /*MaxBytesToEmit=*/0); 4219 } else { 4220 // FIXME: Target specific behavior about how the "extra" bytes are filled. 4221 getStreamer().emitValueToAlignment(Alignment, /*Value=*/0, /*ValueSize=*/1, 4222 /*MaxBytesToEmit=*/0); 4223 } 4224 4225 return ReturnVal; 4226 } 4227 4228 /// parseDirectiveFile 4229 /// ::= .file filename 4230 /// ::= .file number [directory] filename [md5 checksum] [source source-text] 4231 bool MasmParser::parseDirectiveFile(SMLoc DirectiveLoc) { 4232 // FIXME: I'm not sure what this is. 4233 int64_t FileNumber = -1; 4234 if (getLexer().is(AsmToken::Integer)) { 4235 FileNumber = getTok().getIntVal(); 4236 Lex(); 4237 4238 if (FileNumber < 0) 4239 return TokError("negative file number"); 4240 } 4241 4242 std::string Path; 4243 4244 // Usually the directory and filename together, otherwise just the directory. 4245 // Allow the strings to have escaped octal character sequence. 4246 if (check(getTok().isNot(AsmToken::String), 4247 "unexpected token in '.file' directive") || 4248 parseEscapedString(Path)) 4249 return true; 4250 4251 StringRef Directory; 4252 StringRef Filename; 4253 std::string FilenameData; 4254 if (getLexer().is(AsmToken::String)) { 4255 if (check(FileNumber == -1, 4256 "explicit path specified, but no file number") || 4257 parseEscapedString(FilenameData)) 4258 return true; 4259 Filename = FilenameData; 4260 Directory = Path; 4261 } else { 4262 Filename = Path; 4263 } 4264 4265 uint64_t MD5Hi, MD5Lo; 4266 bool HasMD5 = false; 4267 4268 Optional<StringRef> Source; 4269 bool HasSource = false; 4270 std::string SourceString; 4271 4272 while (!parseOptionalToken(AsmToken::EndOfStatement)) { 4273 StringRef Keyword; 4274 if (check(getTok().isNot(AsmToken::Identifier), 4275 "unexpected token in '.file' directive") || 4276 parseIdentifier(Keyword)) 4277 return true; 4278 if (Keyword == "md5") { 4279 HasMD5 = true; 4280 if (check(FileNumber == -1, 4281 "MD5 checksum specified, but no file number") || 4282 parseHexOcta(*this, MD5Hi, MD5Lo)) 4283 return true; 4284 } else if (Keyword == "source") { 4285 HasSource = true; 4286 if (check(FileNumber == -1, 4287 "source specified, but no file number") || 4288 check(getTok().isNot(AsmToken::String), 4289 "unexpected token in '.file' directive") || 4290 parseEscapedString(SourceString)) 4291 return true; 4292 } else { 4293 return TokError("unexpected token in '.file' directive"); 4294 } 4295 } 4296 4297 if (FileNumber == -1) { 4298 // Ignore the directive if there is no number and the target doesn't support 4299 // numberless .file directives. This allows some portability of assembler 4300 // between different object file formats. 4301 if (getContext().getAsmInfo()->hasSingleParameterDotFile()) 4302 getStreamer().emitFileDirective(Filename); 4303 } else { 4304 // In case there is a -g option as well as debug info from directive .file, 4305 // we turn off the -g option, directly use the existing debug info instead. 4306 // Throw away any implicit file table for the assembler source. 4307 if (Ctx.getGenDwarfForAssembly()) { 4308 Ctx.getMCDwarfLineTable(0).resetFileTable(); 4309 Ctx.setGenDwarfForAssembly(false); 4310 } 4311 4312 Optional<MD5::MD5Result> CKMem; 4313 if (HasMD5) { 4314 MD5::MD5Result Sum; 4315 for (unsigned i = 0; i != 8; ++i) { 4316 Sum.Bytes[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); 4317 Sum.Bytes[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); 4318 } 4319 CKMem = Sum; 4320 } 4321 if (HasSource) { 4322 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size())); 4323 memcpy(SourceBuf, SourceString.data(), SourceString.size()); 4324 Source = StringRef(SourceBuf, SourceString.size()); 4325 } 4326 if (FileNumber == 0) { 4327 if (Ctx.getDwarfVersion() < 5) 4328 return Warning(DirectiveLoc, "file 0 not supported prior to DWARF-5"); 4329 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source); 4330 } else { 4331 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( 4332 FileNumber, Directory, Filename, CKMem, Source); 4333 if (!FileNumOrErr) 4334 return Error(DirectiveLoc, toString(FileNumOrErr.takeError())); 4335 } 4336 // Alert the user if there are some .file directives with MD5 and some not. 4337 // But only do that once. 4338 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) { 4339 ReportedInconsistentMD5 = true; 4340 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums"); 4341 } 4342 } 4343 4344 return false; 4345 } 4346 4347 /// parseDirectiveLine 4348 /// ::= .line [number] 4349 bool MasmParser::parseDirectiveLine() { 4350 int64_t LineNumber; 4351 if (getLexer().is(AsmToken::Integer)) { 4352 if (parseIntToken(LineNumber, "unexpected token in '.line' directive")) 4353 return true; 4354 (void)LineNumber; 4355 // FIXME: Do something with the .line. 4356 } 4357 if (parseToken(AsmToken::EndOfStatement, 4358 "unexpected token in '.line' directive")) 4359 return true; 4360 4361 return false; 4362 } 4363 4364 /// parseDirectiveLoc 4365 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 4366 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 4367 /// The first number is a file number, must have been previously assigned with 4368 /// a .file directive, the second number is the line number and optionally the 4369 /// third number is a column position (zero if not specified). The remaining 4370 /// optional items are .loc sub-directives. 4371 bool MasmParser::parseDirectiveLoc() { 4372 int64_t FileNumber = 0, LineNumber = 0; 4373 SMLoc Loc = getTok().getLoc(); 4374 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") || 4375 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, 4376 "file number less than one in '.loc' directive") || 4377 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc, 4378 "unassigned file number in '.loc' directive")) 4379 return true; 4380 4381 // optional 4382 if (getLexer().is(AsmToken::Integer)) { 4383 LineNumber = getTok().getIntVal(); 4384 if (LineNumber < 0) 4385 return TokError("line number less than zero in '.loc' directive"); 4386 Lex(); 4387 } 4388 4389 int64_t ColumnPos = 0; 4390 if (getLexer().is(AsmToken::Integer)) { 4391 ColumnPos = getTok().getIntVal(); 4392 if (ColumnPos < 0) 4393 return TokError("column position less than zero in '.loc' directive"); 4394 Lex(); 4395 } 4396 4397 auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags(); 4398 unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT; 4399 unsigned Isa = 0; 4400 int64_t Discriminator = 0; 4401 4402 auto parseLocOp = [&]() -> bool { 4403 StringRef Name; 4404 SMLoc Loc = getTok().getLoc(); 4405 if (parseIdentifier(Name)) 4406 return TokError("unexpected token in '.loc' directive"); 4407 4408 if (Name == "basic_block") 4409 Flags |= DWARF2_FLAG_BASIC_BLOCK; 4410 else if (Name == "prologue_end") 4411 Flags |= DWARF2_FLAG_PROLOGUE_END; 4412 else if (Name == "epilogue_begin") 4413 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 4414 else if (Name == "is_stmt") { 4415 Loc = getTok().getLoc(); 4416 const MCExpr *Value; 4417 if (parseExpression(Value)) 4418 return true; 4419 // The expression must be the constant 0 or 1. 4420 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4421 int Value = MCE->getValue(); 4422 if (Value == 0) 4423 Flags &= ~DWARF2_FLAG_IS_STMT; 4424 else if (Value == 1) 4425 Flags |= DWARF2_FLAG_IS_STMT; 4426 else 4427 return Error(Loc, "is_stmt value not 0 or 1"); 4428 } else { 4429 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 4430 } 4431 } else if (Name == "isa") { 4432 Loc = getTok().getLoc(); 4433 const MCExpr *Value; 4434 if (parseExpression(Value)) 4435 return true; 4436 // The expression must be a constant greater or equal to 0. 4437 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4438 int Value = MCE->getValue(); 4439 if (Value < 0) 4440 return Error(Loc, "isa number less than zero"); 4441 Isa = Value; 4442 } else { 4443 return Error(Loc, "isa number not a constant value"); 4444 } 4445 } else if (Name == "discriminator") { 4446 if (parseAbsoluteExpression(Discriminator)) 4447 return true; 4448 } else { 4449 return Error(Loc, "unknown sub-directive in '.loc' directive"); 4450 } 4451 return false; 4452 }; 4453 4454 if (parseMany(parseLocOp, false /*hasComma*/)) 4455 return true; 4456 4457 getStreamer().emitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 4458 Isa, Discriminator, StringRef()); 4459 4460 return false; 4461 } 4462 4463 /// parseDirectiveStabs 4464 /// ::= .stabs string, number, number, number 4465 bool MasmParser::parseDirectiveStabs() { 4466 return TokError("unsupported directive '.stabs'"); 4467 } 4468 4469 /// parseDirectiveCVFile 4470 /// ::= .cv_file number filename [checksum] [checksumkind] 4471 bool MasmParser::parseDirectiveCVFile() { 4472 SMLoc FileNumberLoc = getTok().getLoc(); 4473 int64_t FileNumber; 4474 std::string Filename; 4475 std::string Checksum; 4476 int64_t ChecksumKind = 0; 4477 4478 if (parseIntToken(FileNumber, 4479 "expected file number in '.cv_file' directive") || 4480 check(FileNumber < 1, FileNumberLoc, "file number less than one") || 4481 check(getTok().isNot(AsmToken::String), 4482 "unexpected token in '.cv_file' directive") || 4483 parseEscapedString(Filename)) 4484 return true; 4485 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4486 if (check(getTok().isNot(AsmToken::String), 4487 "unexpected token in '.cv_file' directive") || 4488 parseEscapedString(Checksum) || 4489 parseIntToken(ChecksumKind, 4490 "expected checksum kind in '.cv_file' directive") || 4491 parseToken(AsmToken::EndOfStatement, 4492 "unexpected token in '.cv_file' directive")) 4493 return true; 4494 } 4495 4496 Checksum = fromHex(Checksum); 4497 void *CKMem = Ctx.allocate(Checksum.size(), 1); 4498 memcpy(CKMem, Checksum.data(), Checksum.size()); 4499 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), 4500 Checksum.size()); 4501 4502 if (!getStreamer().EmitCVFileDirective(FileNumber, Filename, ChecksumAsBytes, 4503 static_cast<uint8_t>(ChecksumKind))) 4504 return Error(FileNumberLoc, "file number already allocated"); 4505 4506 return false; 4507 } 4508 4509 bool MasmParser::parseCVFunctionId(int64_t &FunctionId, 4510 StringRef DirectiveName) { 4511 SMLoc Loc; 4512 return parseTokenLoc(Loc) || 4513 parseIntToken(FunctionId, "expected function id in '" + DirectiveName + 4514 "' directive") || 4515 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc, 4516 "expected function id within range [0, UINT_MAX)"); 4517 } 4518 4519 bool MasmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { 4520 SMLoc Loc; 4521 return parseTokenLoc(Loc) || 4522 parseIntToken(FileNumber, "expected integer in '" + DirectiveName + 4523 "' directive") || 4524 check(FileNumber < 1, Loc, "file number less than one in '" + 4525 DirectiveName + "' directive") || 4526 check(!getCVContext().isValidFileNumber(FileNumber), Loc, 4527 "unassigned file number in '" + DirectiveName + "' directive"); 4528 } 4529 4530 /// parseDirectiveCVFuncId 4531 /// ::= .cv_func_id FunctionId 4532 /// 4533 /// Introduces a function ID that can be used with .cv_loc. 4534 bool MasmParser::parseDirectiveCVFuncId() { 4535 SMLoc FunctionIdLoc = getTok().getLoc(); 4536 int64_t FunctionId; 4537 4538 if (parseCVFunctionId(FunctionId, ".cv_func_id") || 4539 parseToken(AsmToken::EndOfStatement, 4540 "unexpected token in '.cv_func_id' directive")) 4541 return true; 4542 4543 if (!getStreamer().EmitCVFuncIdDirective(FunctionId)) 4544 return Error(FunctionIdLoc, "function id already allocated"); 4545 4546 return false; 4547 } 4548 4549 /// parseDirectiveCVInlineSiteId 4550 /// ::= .cv_inline_site_id FunctionId 4551 /// "within" IAFunc 4552 /// "inlined_at" IAFile IALine [IACol] 4553 /// 4554 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined 4555 /// at" source location information for use in the line table of the caller, 4556 /// whether the caller is a real function or another inlined call site. 4557 bool MasmParser::parseDirectiveCVInlineSiteId() { 4558 SMLoc FunctionIdLoc = getTok().getLoc(); 4559 int64_t FunctionId; 4560 int64_t IAFunc; 4561 int64_t IAFile; 4562 int64_t IALine; 4563 int64_t IACol = 0; 4564 4565 // FunctionId 4566 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id")) 4567 return true; 4568 4569 // "within" 4570 if (check((getLexer().isNot(AsmToken::Identifier) || 4571 getTok().getIdentifier() != "within"), 4572 "expected 'within' identifier in '.cv_inline_site_id' directive")) 4573 return true; 4574 Lex(); 4575 4576 // IAFunc 4577 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id")) 4578 return true; 4579 4580 // "inlined_at" 4581 if (check((getLexer().isNot(AsmToken::Identifier) || 4582 getTok().getIdentifier() != "inlined_at"), 4583 "expected 'inlined_at' identifier in '.cv_inline_site_id' " 4584 "directive") ) 4585 return true; 4586 Lex(); 4587 4588 // IAFile IALine 4589 if (parseCVFileId(IAFile, ".cv_inline_site_id") || 4590 parseIntToken(IALine, "expected line number after 'inlined_at'")) 4591 return true; 4592 4593 // [IACol] 4594 if (getLexer().is(AsmToken::Integer)) { 4595 IACol = getTok().getIntVal(); 4596 Lex(); 4597 } 4598 4599 if (parseToken(AsmToken::EndOfStatement, 4600 "unexpected token in '.cv_inline_site_id' directive")) 4601 return true; 4602 4603 if (!getStreamer().EmitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, 4604 IALine, IACol, FunctionIdLoc)) 4605 return Error(FunctionIdLoc, "function id already allocated"); 4606 4607 return false; 4608 } 4609 4610 /// parseDirectiveCVLoc 4611 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] 4612 /// [is_stmt VALUE] 4613 /// The first number is a file number, must have been previously assigned with 4614 /// a .file directive, the second number is the line number and optionally the 4615 /// third number is a column position (zero if not specified). The remaining 4616 /// optional items are .loc sub-directives. 4617 bool MasmParser::parseDirectiveCVLoc() { 4618 SMLoc DirectiveLoc = getTok().getLoc(); 4619 int64_t FunctionId, FileNumber; 4620 if (parseCVFunctionId(FunctionId, ".cv_loc") || 4621 parseCVFileId(FileNumber, ".cv_loc")) 4622 return true; 4623 4624 int64_t LineNumber = 0; 4625 if (getLexer().is(AsmToken::Integer)) { 4626 LineNumber = getTok().getIntVal(); 4627 if (LineNumber < 0) 4628 return TokError("line number less than zero in '.cv_loc' directive"); 4629 Lex(); 4630 } 4631 4632 int64_t ColumnPos = 0; 4633 if (getLexer().is(AsmToken::Integer)) { 4634 ColumnPos = getTok().getIntVal(); 4635 if (ColumnPos < 0) 4636 return TokError("column position less than zero in '.cv_loc' directive"); 4637 Lex(); 4638 } 4639 4640 bool PrologueEnd = false; 4641 uint64_t IsStmt = 0; 4642 4643 auto parseOp = [&]() -> bool { 4644 StringRef Name; 4645 SMLoc Loc = getTok().getLoc(); 4646 if (parseIdentifier(Name)) 4647 return TokError("unexpected token in '.cv_loc' directive"); 4648 if (Name == "prologue_end") 4649 PrologueEnd = true; 4650 else if (Name == "is_stmt") { 4651 Loc = getTok().getLoc(); 4652 const MCExpr *Value; 4653 if (parseExpression(Value)) 4654 return true; 4655 // The expression must be the constant 0 or 1. 4656 IsStmt = ~0ULL; 4657 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value)) 4658 IsStmt = MCE->getValue(); 4659 4660 if (IsStmt > 1) 4661 return Error(Loc, "is_stmt value not 0 or 1"); 4662 } else { 4663 return Error(Loc, "unknown sub-directive in '.cv_loc' directive"); 4664 } 4665 return false; 4666 }; 4667 4668 if (parseMany(parseOp, false /*hasComma*/)) 4669 return true; 4670 4671 getStreamer().emitCVLocDirective(FunctionId, FileNumber, LineNumber, 4672 ColumnPos, PrologueEnd, IsStmt, StringRef(), 4673 DirectiveLoc); 4674 return false; 4675 } 4676 4677 /// parseDirectiveCVLinetable 4678 /// ::= .cv_linetable FunctionId, FnStart, FnEnd 4679 bool MasmParser::parseDirectiveCVLinetable() { 4680 int64_t FunctionId; 4681 StringRef FnStartName, FnEndName; 4682 SMLoc Loc = getTok().getLoc(); 4683 if (parseCVFunctionId(FunctionId, ".cv_linetable") || 4684 parseToken(AsmToken::Comma, 4685 "unexpected token in '.cv_linetable' directive") || 4686 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 4687 "expected identifier in directive") || 4688 parseToken(AsmToken::Comma, 4689 "unexpected token in '.cv_linetable' directive") || 4690 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 4691 "expected identifier in directive")) 4692 return true; 4693 4694 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 4695 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 4696 4697 getStreamer().emitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym); 4698 return false; 4699 } 4700 4701 /// parseDirectiveCVInlineLinetable 4702 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd 4703 bool MasmParser::parseDirectiveCVInlineLinetable() { 4704 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; 4705 StringRef FnStartName, FnEndName; 4706 SMLoc Loc = getTok().getLoc(); 4707 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") || 4708 parseTokenLoc(Loc) || 4709 parseIntToken( 4710 SourceFileId, 4711 "expected SourceField in '.cv_inline_linetable' directive") || 4712 check(SourceFileId <= 0, Loc, 4713 "File id less than zero in '.cv_inline_linetable' directive") || 4714 parseTokenLoc(Loc) || 4715 parseIntToken( 4716 SourceLineNum, 4717 "expected SourceLineNum in '.cv_inline_linetable' directive") || 4718 check(SourceLineNum < 0, Loc, 4719 "Line number less than zero in '.cv_inline_linetable' directive") || 4720 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 4721 "expected identifier in directive") || 4722 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 4723 "expected identifier in directive")) 4724 return true; 4725 4726 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 4727 return true; 4728 4729 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 4730 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 4731 getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, 4732 SourceLineNum, FnStartSym, 4733 FnEndSym); 4734 return false; 4735 } 4736 4737 void MasmParser::initializeCVDefRangeTypeMap() { 4738 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER; 4739 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL; 4740 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER; 4741 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL; 4742 } 4743 4744 /// parseDirectiveCVDefRange 4745 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* 4746 bool MasmParser::parseDirectiveCVDefRange() { 4747 SMLoc Loc; 4748 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; 4749 while (getLexer().is(AsmToken::Identifier)) { 4750 Loc = getLexer().getLoc(); 4751 StringRef GapStartName; 4752 if (parseIdentifier(GapStartName)) 4753 return Error(Loc, "expected identifier in directive"); 4754 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName); 4755 4756 Loc = getLexer().getLoc(); 4757 StringRef GapEndName; 4758 if (parseIdentifier(GapEndName)) 4759 return Error(Loc, "expected identifier in directive"); 4760 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName); 4761 4762 Ranges.push_back({GapStartSym, GapEndSym}); 4763 } 4764 4765 StringRef CVDefRangeTypeStr; 4766 if (parseToken( 4767 AsmToken::Comma, 4768 "expected comma before def_range type in .cv_def_range directive") || 4769 parseIdentifier(CVDefRangeTypeStr)) 4770 return Error(Loc, "expected def_range type in directive"); 4771 4772 StringMap<CVDefRangeType>::const_iterator CVTypeIt = 4773 CVDefRangeTypeMap.find(CVDefRangeTypeStr); 4774 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end()) 4775 ? CVDR_DEFRANGE 4776 : CVTypeIt->getValue(); 4777 switch (CVDRType) { 4778 case CVDR_DEFRANGE_REGISTER: { 4779 int64_t DRRegister; 4780 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4781 ".cv_def_range directive") || 4782 parseAbsoluteExpression(DRRegister)) 4783 return Error(Loc, "expected register number"); 4784 4785 codeview::DefRangeRegisterHeader DRHdr; 4786 DRHdr.Register = DRRegister; 4787 DRHdr.MayHaveNoName = 0; 4788 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4789 break; 4790 } 4791 case CVDR_DEFRANGE_FRAMEPOINTER_REL: { 4792 int64_t DROffset; 4793 if (parseToken(AsmToken::Comma, 4794 "expected comma before offset in .cv_def_range directive") || 4795 parseAbsoluteExpression(DROffset)) 4796 return Error(Loc, "expected offset value"); 4797 4798 codeview::DefRangeFramePointerRelHeader DRHdr; 4799 DRHdr.Offset = DROffset; 4800 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4801 break; 4802 } 4803 case CVDR_DEFRANGE_SUBFIELD_REGISTER: { 4804 int64_t DRRegister; 4805 int64_t DROffsetInParent; 4806 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4807 ".cv_def_range directive") || 4808 parseAbsoluteExpression(DRRegister)) 4809 return Error(Loc, "expected register number"); 4810 if (parseToken(AsmToken::Comma, 4811 "expected comma before offset in .cv_def_range directive") || 4812 parseAbsoluteExpression(DROffsetInParent)) 4813 return Error(Loc, "expected offset value"); 4814 4815 codeview::DefRangeSubfieldRegisterHeader DRHdr; 4816 DRHdr.Register = DRRegister; 4817 DRHdr.MayHaveNoName = 0; 4818 DRHdr.OffsetInParent = DROffsetInParent; 4819 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4820 break; 4821 } 4822 case CVDR_DEFRANGE_REGISTER_REL: { 4823 int64_t DRRegister; 4824 int64_t DRFlags; 4825 int64_t DRBasePointerOffset; 4826 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4827 ".cv_def_range directive") || 4828 parseAbsoluteExpression(DRRegister)) 4829 return Error(Loc, "expected register value"); 4830 if (parseToken( 4831 AsmToken::Comma, 4832 "expected comma before flag value in .cv_def_range directive") || 4833 parseAbsoluteExpression(DRFlags)) 4834 return Error(Loc, "expected flag value"); 4835 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset " 4836 "in .cv_def_range directive") || 4837 parseAbsoluteExpression(DRBasePointerOffset)) 4838 return Error(Loc, "expected base pointer offset value"); 4839 4840 codeview::DefRangeRegisterRelHeader DRHdr; 4841 DRHdr.Register = DRRegister; 4842 DRHdr.Flags = DRFlags; 4843 DRHdr.BasePointerOffset = DRBasePointerOffset; 4844 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4845 break; 4846 } 4847 default: 4848 return Error(Loc, "unexpected def_range type in .cv_def_range directive"); 4849 } 4850 return true; 4851 } 4852 4853 /// parseDirectiveCVString 4854 /// ::= .cv_stringtable "string" 4855 bool MasmParser::parseDirectiveCVString() { 4856 std::string Data; 4857 if (checkForValidSection() || parseEscapedString(Data)) 4858 return addErrorSuffix(" in '.cv_string' directive"); 4859 4860 // Put the string in the table and emit the offset. 4861 std::pair<StringRef, unsigned> Insertion = 4862 getCVContext().addToStringTable(Data); 4863 getStreamer().emitIntValue(Insertion.second, 4); 4864 return false; 4865 } 4866 4867 /// parseDirectiveCVStringTable 4868 /// ::= .cv_stringtable 4869 bool MasmParser::parseDirectiveCVStringTable() { 4870 getStreamer().emitCVStringTableDirective(); 4871 return false; 4872 } 4873 4874 /// parseDirectiveCVFileChecksums 4875 /// ::= .cv_filechecksums 4876 bool MasmParser::parseDirectiveCVFileChecksums() { 4877 getStreamer().emitCVFileChecksumsDirective(); 4878 return false; 4879 } 4880 4881 /// parseDirectiveCVFileChecksumOffset 4882 /// ::= .cv_filechecksumoffset fileno 4883 bool MasmParser::parseDirectiveCVFileChecksumOffset() { 4884 int64_t FileNo; 4885 if (parseIntToken(FileNo, "expected identifier in directive")) 4886 return true; 4887 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 4888 return true; 4889 getStreamer().emitCVFileChecksumOffsetDirective(FileNo); 4890 return false; 4891 } 4892 4893 /// parseDirectiveCVFPOData 4894 /// ::= .cv_fpo_data procsym 4895 bool MasmParser::parseDirectiveCVFPOData() { 4896 SMLoc DirLoc = getLexer().getLoc(); 4897 StringRef ProcName; 4898 if (parseIdentifier(ProcName)) 4899 return TokError("expected symbol name"); 4900 if (parseEOL("unexpected tokens")) 4901 return addErrorSuffix(" in '.cv_fpo_data' directive"); 4902 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName); 4903 getStreamer().EmitCVFPOData(ProcSym, DirLoc); 4904 return false; 4905 } 4906 4907 /// parseDirectiveCFISections 4908 /// ::= .cfi_sections section [, section] 4909 bool MasmParser::parseDirectiveCFISections() { 4910 StringRef Name; 4911 bool EH = false; 4912 bool Debug = false; 4913 4914 if (parseIdentifier(Name)) 4915 return TokError("Expected an identifier"); 4916 4917 if (Name == ".eh_frame") 4918 EH = true; 4919 else if (Name == ".debug_frame") 4920 Debug = true; 4921 4922 if (getLexer().is(AsmToken::Comma)) { 4923 Lex(); 4924 4925 if (parseIdentifier(Name)) 4926 return TokError("Expected an identifier"); 4927 4928 if (Name == ".eh_frame") 4929 EH = true; 4930 else if (Name == ".debug_frame") 4931 Debug = true; 4932 } 4933 4934 getStreamer().emitCFISections(EH, Debug); 4935 return false; 4936 } 4937 4938 /// parseDirectiveCFIStartProc 4939 /// ::= .cfi_startproc [simple] 4940 bool MasmParser::parseDirectiveCFIStartProc() { 4941 StringRef Simple; 4942 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4943 if (check(parseIdentifier(Simple) || Simple != "simple", 4944 "unexpected token") || 4945 parseToken(AsmToken::EndOfStatement)) 4946 return addErrorSuffix(" in '.cfi_startproc' directive"); 4947 } 4948 4949 // TODO(kristina): Deal with a corner case of incorrect diagnostic context 4950 // being produced if this directive is emitted as part of preprocessor macro 4951 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. 4952 // Tools like llvm-mc on the other hand are not affected by it, and report 4953 // correct context information. 4954 getStreamer().emitCFIStartProc(!Simple.empty(), Lexer.getLoc()); 4955 return false; 4956 } 4957 4958 /// parseDirectiveCFIEndProc 4959 /// ::= .cfi_endproc 4960 bool MasmParser::parseDirectiveCFIEndProc() { 4961 getStreamer().emitCFIEndProc(); 4962 return false; 4963 } 4964 4965 /// parse register name or number. 4966 bool MasmParser::parseRegisterOrRegisterNumber(int64_t &Register, 4967 SMLoc DirectiveLoc) { 4968 unsigned RegNo; 4969 4970 if (getLexer().isNot(AsmToken::Integer)) { 4971 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc)) 4972 return true; 4973 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true); 4974 } else 4975 return parseAbsoluteExpression(Register); 4976 4977 return false; 4978 } 4979 4980 /// parseDirectiveCFIDefCfa 4981 /// ::= .cfi_def_cfa register, offset 4982 bool MasmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { 4983 int64_t Register = 0, Offset = 0; 4984 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4985 parseToken(AsmToken::Comma, "unexpected token in directive") || 4986 parseAbsoluteExpression(Offset)) 4987 return true; 4988 4989 getStreamer().emitCFIDefCfa(Register, Offset); 4990 return false; 4991 } 4992 4993 /// parseDirectiveCFIDefCfaOffset 4994 /// ::= .cfi_def_cfa_offset offset 4995 bool MasmParser::parseDirectiveCFIDefCfaOffset() { 4996 int64_t Offset = 0; 4997 if (parseAbsoluteExpression(Offset)) 4998 return true; 4999 5000 getStreamer().emitCFIDefCfaOffset(Offset); 5001 return false; 5002 } 5003 5004 /// parseDirectiveCFIRegister 5005 /// ::= .cfi_register register, register 5006 bool MasmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { 5007 int64_t Register1 = 0, Register2 = 0; 5008 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || 5009 parseToken(AsmToken::Comma, "unexpected token in directive") || 5010 parseRegisterOrRegisterNumber(Register2, DirectiveLoc)) 5011 return true; 5012 5013 getStreamer().emitCFIRegister(Register1, Register2); 5014 return false; 5015 } 5016 5017 /// parseDirectiveCFIWindowSave 5018 /// ::= .cfi_window_save 5019 bool MasmParser::parseDirectiveCFIWindowSave() { 5020 getStreamer().emitCFIWindowSave(); 5021 return false; 5022 } 5023 5024 /// parseDirectiveCFIAdjustCfaOffset 5025 /// ::= .cfi_adjust_cfa_offset adjustment 5026 bool MasmParser::parseDirectiveCFIAdjustCfaOffset() { 5027 int64_t Adjustment = 0; 5028 if (parseAbsoluteExpression(Adjustment)) 5029 return true; 5030 5031 getStreamer().emitCFIAdjustCfaOffset(Adjustment); 5032 return false; 5033 } 5034 5035 /// parseDirectiveCFIDefCfaRegister 5036 /// ::= .cfi_def_cfa_register register 5037 bool MasmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { 5038 int64_t Register = 0; 5039 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5040 return true; 5041 5042 getStreamer().emitCFIDefCfaRegister(Register); 5043 return false; 5044 } 5045 5046 /// parseDirectiveCFIOffset 5047 /// ::= .cfi_offset register, offset 5048 bool MasmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { 5049 int64_t Register = 0; 5050 int64_t Offset = 0; 5051 5052 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 5053 parseToken(AsmToken::Comma, "unexpected token in directive") || 5054 parseAbsoluteExpression(Offset)) 5055 return true; 5056 5057 getStreamer().emitCFIOffset(Register, Offset); 5058 return false; 5059 } 5060 5061 /// parseDirectiveCFIRelOffset 5062 /// ::= .cfi_rel_offset register, offset 5063 bool MasmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { 5064 int64_t Register = 0, Offset = 0; 5065 5066 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 5067 parseToken(AsmToken::Comma, "unexpected token in directive") || 5068 parseAbsoluteExpression(Offset)) 5069 return true; 5070 5071 getStreamer().emitCFIRelOffset(Register, Offset); 5072 return false; 5073 } 5074 5075 static bool isValidEncoding(int64_t Encoding) { 5076 if (Encoding & ~0xff) 5077 return false; 5078 5079 if (Encoding == dwarf::DW_EH_PE_omit) 5080 return true; 5081 5082 const unsigned Format = Encoding & 0xf; 5083 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 5084 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 5085 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 5086 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 5087 return false; 5088 5089 const unsigned Application = Encoding & 0x70; 5090 if (Application != dwarf::DW_EH_PE_absptr && 5091 Application != dwarf::DW_EH_PE_pcrel) 5092 return false; 5093 5094 return true; 5095 } 5096 5097 /// parseDirectiveCFIPersonalityOrLsda 5098 /// IsPersonality true for cfi_personality, false for cfi_lsda 5099 /// ::= .cfi_personality encoding, [symbol_name] 5100 /// ::= .cfi_lsda encoding, [symbol_name] 5101 bool MasmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { 5102 int64_t Encoding = 0; 5103 if (parseAbsoluteExpression(Encoding)) 5104 return true; 5105 if (Encoding == dwarf::DW_EH_PE_omit) 5106 return false; 5107 5108 StringRef Name; 5109 if (check(!isValidEncoding(Encoding), "unsupported encoding.") || 5110 parseToken(AsmToken::Comma, "unexpected token in directive") || 5111 check(parseIdentifier(Name), "expected identifier in directive")) 5112 return true; 5113 5114 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5115 5116 if (IsPersonality) 5117 getStreamer().emitCFIPersonality(Sym, Encoding); 5118 else 5119 getStreamer().emitCFILsda(Sym, Encoding); 5120 return false; 5121 } 5122 5123 /// parseDirectiveCFIRememberState 5124 /// ::= .cfi_remember_state 5125 bool MasmParser::parseDirectiveCFIRememberState() { 5126 getStreamer().emitCFIRememberState(); 5127 return false; 5128 } 5129 5130 /// parseDirectiveCFIRestoreState 5131 /// ::= .cfi_remember_state 5132 bool MasmParser::parseDirectiveCFIRestoreState() { 5133 getStreamer().emitCFIRestoreState(); 5134 return false; 5135 } 5136 5137 /// parseDirectiveCFISameValue 5138 /// ::= .cfi_same_value register 5139 bool MasmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { 5140 int64_t Register = 0; 5141 5142 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5143 return true; 5144 5145 getStreamer().emitCFISameValue(Register); 5146 return false; 5147 } 5148 5149 /// parseDirectiveCFIRestore 5150 /// ::= .cfi_restore register 5151 bool MasmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { 5152 int64_t Register = 0; 5153 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5154 return true; 5155 5156 getStreamer().emitCFIRestore(Register); 5157 return false; 5158 } 5159 5160 /// parseDirectiveCFIEscape 5161 /// ::= .cfi_escape expression[,...] 5162 bool MasmParser::parseDirectiveCFIEscape() { 5163 std::string Values; 5164 int64_t CurrValue; 5165 if (parseAbsoluteExpression(CurrValue)) 5166 return true; 5167 5168 Values.push_back((uint8_t)CurrValue); 5169 5170 while (getLexer().is(AsmToken::Comma)) { 5171 Lex(); 5172 5173 if (parseAbsoluteExpression(CurrValue)) 5174 return true; 5175 5176 Values.push_back((uint8_t)CurrValue); 5177 } 5178 5179 getStreamer().emitCFIEscape(Values); 5180 return false; 5181 } 5182 5183 /// parseDirectiveCFIReturnColumn 5184 /// ::= .cfi_return_column register 5185 bool MasmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { 5186 int64_t Register = 0; 5187 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5188 return true; 5189 getStreamer().emitCFIReturnColumn(Register); 5190 return false; 5191 } 5192 5193 /// parseDirectiveCFISignalFrame 5194 /// ::= .cfi_signal_frame 5195 bool MasmParser::parseDirectiveCFISignalFrame() { 5196 if (parseToken(AsmToken::EndOfStatement, 5197 "unexpected token in '.cfi_signal_frame'")) 5198 return true; 5199 5200 getStreamer().emitCFISignalFrame(); 5201 return false; 5202 } 5203 5204 /// parseDirectiveCFIUndefined 5205 /// ::= .cfi_undefined register 5206 bool MasmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { 5207 int64_t Register = 0; 5208 5209 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5210 return true; 5211 5212 getStreamer().emitCFIUndefined(Register); 5213 return false; 5214 } 5215 5216 /// parseDirectiveAltmacro 5217 /// ::= .altmacro 5218 /// ::= .noaltmacro 5219 bool MasmParser::parseDirectiveAltmacro(StringRef Directive) { 5220 if (getLexer().isNot(AsmToken::EndOfStatement)) 5221 return TokError("unexpected token in '" + Directive + "' directive"); 5222 AltMacroMode = (Directive == ".altmacro"); 5223 return false; 5224 } 5225 5226 /// parseDirectiveMacro 5227 /// ::= .macro name[,] [parameters] 5228 bool MasmParser::parseDirectiveMacro(SMLoc DirectiveLoc) { 5229 StringRef Name; 5230 if (parseIdentifier(Name)) 5231 return TokError("expected identifier in '.macro' directive"); 5232 5233 if (getLexer().is(AsmToken::Comma)) 5234 Lex(); 5235 5236 MCAsmMacroParameters Parameters; 5237 while (getLexer().isNot(AsmToken::EndOfStatement)) { 5238 5239 if (!Parameters.empty() && Parameters.back().Vararg) 5240 return Error(Lexer.getLoc(), 5241 "Vararg parameter '" + Parameters.back().Name + 5242 "' should be last one in the list of parameters."); 5243 5244 MCAsmMacroParameter Parameter; 5245 if (parseIdentifier(Parameter.Name)) 5246 return TokError("expected identifier in '.macro' directive"); 5247 5248 // Emit an error if two (or more) named parameters share the same name. 5249 for (const MCAsmMacroParameter& CurrParam : Parameters) 5250 if (CurrParam.Name.equals(Parameter.Name)) 5251 return TokError("macro '" + Name + "' has multiple parameters" 5252 " named '" + Parameter.Name + "'"); 5253 5254 if (Lexer.is(AsmToken::Colon)) { 5255 Lex(); // consume ':' 5256 5257 SMLoc QualLoc; 5258 StringRef Qualifier; 5259 5260 QualLoc = Lexer.getLoc(); 5261 if (parseIdentifier(Qualifier)) 5262 return Error(QualLoc, "missing parameter qualifier for " 5263 "'" + Parameter.Name + "' in macro '" + Name + "'"); 5264 5265 if (Qualifier == "req") 5266 Parameter.Required = true; 5267 else if (Qualifier == "vararg") 5268 Parameter.Vararg = true; 5269 else 5270 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier " 5271 "for '" + Parameter.Name + "' in macro '" + Name + "'"); 5272 } 5273 5274 if (getLexer().is(AsmToken::Equal)) { 5275 Lex(); 5276 5277 SMLoc ParamLoc; 5278 5279 ParamLoc = Lexer.getLoc(); 5280 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false )) 5281 return true; 5282 5283 if (Parameter.Required) 5284 Warning(ParamLoc, "pointless default value for required parameter " 5285 "'" + Parameter.Name + "' in macro '" + Name + "'"); 5286 } 5287 5288 Parameters.push_back(std::move(Parameter)); 5289 5290 if (getLexer().is(AsmToken::Comma)) 5291 Lex(); 5292 } 5293 5294 // Eat just the end of statement. 5295 Lexer.Lex(); 5296 5297 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors. 5298 AsmToken EndToken, StartToken = getTok(); 5299 unsigned MacroDepth = 0; 5300 // Lex the macro definition. 5301 while (true) { 5302 // Ignore Lexing errors in macros. 5303 while (Lexer.is(AsmToken::Error)) { 5304 Lexer.Lex(); 5305 } 5306 5307 // Check whether we have reached the end of the file. 5308 if (getLexer().is(AsmToken::Eof)) 5309 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 5310 5311 // Otherwise, check whether we have reach the .endmacro. 5312 if (getLexer().is(AsmToken::Identifier)) { 5313 if (getTok().getIdentifier() == ".endm" || 5314 getTok().getIdentifier() == ".endmacro") { 5315 if (MacroDepth == 0) { // Outermost macro. 5316 EndToken = getTok(); 5317 Lexer.Lex(); 5318 if (getLexer().isNot(AsmToken::EndOfStatement)) 5319 return TokError("unexpected token in '" + EndToken.getIdentifier() + 5320 "' directive"); 5321 break; 5322 } else { 5323 // Otherwise we just found the end of an inner macro. 5324 --MacroDepth; 5325 } 5326 } else if (getTok().getIdentifier() == ".macro") { 5327 // We allow nested macros. Those aren't instantiated until the outermost 5328 // macro is expanded so just ignore them for now. 5329 ++MacroDepth; 5330 } 5331 } 5332 5333 // Otherwise, scan til the end of the statement. 5334 eatToEndOfStatement(); 5335 } 5336 5337 if (getContext().lookupMacro(Name)) { 5338 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 5339 } 5340 5341 const char *BodyStart = StartToken.getLoc().getPointer(); 5342 const char *BodyEnd = EndToken.getLoc().getPointer(); 5343 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 5344 checkForBadMacro(DirectiveLoc, Name, Body, Parameters); 5345 MCAsmMacro Macro(Name, Body, std::move(Parameters)); 5346 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n"; 5347 Macro.dump()); 5348 getContext().defineMacro(Name, std::move(Macro)); 5349 return false; 5350 } 5351 5352 /// checkForBadMacro 5353 /// 5354 /// With the support added for named parameters there may be code out there that 5355 /// is transitioning from positional parameters. In versions of gas that did 5356 /// not support named parameters they would be ignored on the macro definition. 5357 /// But to support both styles of parameters this is not possible so if a macro 5358 /// definition has named parameters but does not use them and has what appears 5359 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a 5360 /// warning that the positional parameter found in body which have no effect. 5361 /// Hoping the developer will either remove the named parameters from the macro 5362 /// definition so the positional parameters get used if that was what was 5363 /// intended or change the macro to use the named parameters. It is possible 5364 /// this warning will trigger when the none of the named parameters are used 5365 /// and the strings like $1 are infact to simply to be passed trough unchanged. 5366 void MasmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, 5367 StringRef Body, 5368 ArrayRef<MCAsmMacroParameter> Parameters) { 5369 // If this macro is not defined with named parameters the warning we are 5370 // checking for here doesn't apply. 5371 unsigned NParameters = Parameters.size(); 5372 if (NParameters == 0) 5373 return; 5374 5375 bool NamedParametersFound = false; 5376 bool PositionalParametersFound = false; 5377 5378 // Look at the body of the macro for use of both the named parameters and what 5379 // are likely to be positional parameters. This is what expandMacro() is 5380 // doing when it finds the parameters in the body. 5381 while (!Body.empty()) { 5382 // Scan for the next possible parameter. 5383 std::size_t End = Body.size(), Pos = 0; 5384 for (; Pos != End; ++Pos) { 5385 // Check for a substitution or escape. 5386 // This macro is defined with parameters, look for \foo, \bar, etc. 5387 if (Body[Pos] == '\\' && Pos + 1 != End) 5388 break; 5389 5390 // This macro should have parameters, but look for $0, $1, ..., $n too. 5391 if (Body[Pos] != '$' || Pos + 1 == End) 5392 continue; 5393 char Next = Body[Pos + 1]; 5394 if (Next == '$' || Next == 'n' || 5395 isdigit(static_cast<unsigned char>(Next))) 5396 break; 5397 } 5398 5399 // Check if we reached the end. 5400 if (Pos == End) 5401 break; 5402 5403 if (Body[Pos] == '$') { 5404 switch (Body[Pos + 1]) { 5405 // $$ => $ 5406 case '$': 5407 break; 5408 5409 // $n => number of arguments 5410 case 'n': 5411 PositionalParametersFound = true; 5412 break; 5413 5414 // $[0-9] => argument 5415 default: { 5416 PositionalParametersFound = true; 5417 break; 5418 } 5419 } 5420 Pos += 2; 5421 } else { 5422 unsigned I = Pos + 1; 5423 while (isIdentifierChar(Body[I]) && I + 1 != End) 5424 ++I; 5425 5426 const char *Begin = Body.data() + Pos + 1; 5427 StringRef Argument(Begin, I - (Pos + 1)); 5428 unsigned Index = 0; 5429 for (; Index < NParameters; ++Index) 5430 if (Parameters[Index].Name == Argument) 5431 break; 5432 5433 if (Index == NParameters) { 5434 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 5435 Pos += 3; 5436 else { 5437 Pos = I; 5438 } 5439 } else { 5440 NamedParametersFound = true; 5441 Pos += 1 + Argument.size(); 5442 } 5443 } 5444 // Update the scan point. 5445 Body = Body.substr(Pos); 5446 } 5447 5448 if (!NamedParametersFound && PositionalParametersFound) 5449 Warning(DirectiveLoc, "macro defined with named parameters which are not " 5450 "used in macro body, possible positional parameter " 5451 "found in body which will have no effect"); 5452 } 5453 5454 /// parseDirectiveExitMacro 5455 /// ::= .exitm 5456 bool MasmParser::parseDirectiveExitMacro(StringRef Directive) { 5457 if (parseToken(AsmToken::EndOfStatement, 5458 "unexpected token in '" + Directive + "' directive")) 5459 return true; 5460 5461 if (!isInsideMacroInstantiation()) 5462 return TokError("unexpected '" + Directive + "' in file, " 5463 "no current macro definition"); 5464 5465 // Exit all conditionals that are active in the current macro. 5466 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { 5467 TheCondState = TheCondStack.back(); 5468 TheCondStack.pop_back(); 5469 } 5470 5471 handleMacroExit(); 5472 return false; 5473 } 5474 5475 /// parseDirectiveEndMacro 5476 /// ::= .endm 5477 /// ::= .endmacro 5478 bool MasmParser::parseDirectiveEndMacro(StringRef Directive) { 5479 if (getLexer().isNot(AsmToken::EndOfStatement)) 5480 return TokError("unexpected token in '" + Directive + "' directive"); 5481 5482 // If we are inside a macro instantiation, terminate the current 5483 // instantiation. 5484 if (isInsideMacroInstantiation()) { 5485 handleMacroExit(); 5486 return false; 5487 } 5488 5489 // Otherwise, this .endmacro is a stray entry in the file; well formed 5490 // .endmacro directives are handled during the macro definition parsing. 5491 return TokError("unexpected '" + Directive + "' in file, " 5492 "no current macro definition"); 5493 } 5494 5495 /// parseDirectivePurgeMacro 5496 /// ::= .purgem 5497 bool MasmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { 5498 StringRef Name; 5499 SMLoc Loc; 5500 if (parseTokenLoc(Loc) || 5501 check(parseIdentifier(Name), Loc, 5502 "expected identifier in '.purgem' directive") || 5503 parseToken(AsmToken::EndOfStatement, 5504 "unexpected token in '.purgem' directive")) 5505 return true; 5506 5507 if (!getContext().lookupMacro(Name)) 5508 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 5509 5510 getContext().undefineMacro(Name); 5511 DEBUG_WITH_TYPE("asm-macros", dbgs() 5512 << "Un-defining macro: " << Name << "\n"); 5513 return false; 5514 } 5515 5516 /// parseDirectiveSymbolAttribute 5517 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 5518 bool MasmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 5519 auto parseOp = [&]() -> bool { 5520 StringRef Name; 5521 SMLoc Loc = getTok().getLoc(); 5522 if (parseIdentifier(Name)) 5523 return Error(Loc, "expected identifier"); 5524 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5525 5526 // Assembler local symbols don't make any sense here. Complain loudly. 5527 if (Sym->isTemporary()) 5528 return Error(Loc, "non-local symbol required"); 5529 5530 if (!getStreamer().emitSymbolAttribute(Sym, Attr)) 5531 return Error(Loc, "unable to emit symbol attribute"); 5532 return false; 5533 }; 5534 5535 if (parseMany(parseOp)) 5536 return addErrorSuffix(" in directive"); 5537 return false; 5538 } 5539 5540 /// parseDirectiveComm 5541 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 5542 bool MasmParser::parseDirectiveComm(bool IsLocal) { 5543 if (checkForValidSection()) 5544 return true; 5545 5546 SMLoc IDLoc = getLexer().getLoc(); 5547 StringRef Name; 5548 if (parseIdentifier(Name)) 5549 return TokError("expected identifier in directive"); 5550 5551 // Handle the identifier as the key symbol. 5552 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5553 5554 if (getLexer().isNot(AsmToken::Comma)) 5555 return TokError("unexpected token in directive"); 5556 Lex(); 5557 5558 int64_t Size; 5559 SMLoc SizeLoc = getLexer().getLoc(); 5560 if (parseAbsoluteExpression(Size)) 5561 return true; 5562 5563 int64_t Pow2Alignment = 0; 5564 SMLoc Pow2AlignmentLoc; 5565 if (getLexer().is(AsmToken::Comma)) { 5566 Lex(); 5567 Pow2AlignmentLoc = getLexer().getLoc(); 5568 if (parseAbsoluteExpression(Pow2Alignment)) 5569 return true; 5570 5571 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); 5572 if (IsLocal && LCOMM == LCOMM::NoAlignment) 5573 return Error(Pow2AlignmentLoc, "alignment not supported on this target"); 5574 5575 // If this target takes alignments in bytes (not log) validate and convert. 5576 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || 5577 (IsLocal && LCOMM == LCOMM::ByteAlignment)) { 5578 if (!isPowerOf2_64(Pow2Alignment)) 5579 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 5580 Pow2Alignment = Log2_64(Pow2Alignment); 5581 } 5582 } 5583 5584 if (parseToken(AsmToken::EndOfStatement, 5585 "unexpected token in '.comm' or '.lcomm' directive")) 5586 return true; 5587 5588 // NOTE: a size of zero for a .comm should create a undefined symbol 5589 // but a size of .lcomm creates a bss symbol of size zero. 5590 if (Size < 0) 5591 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 5592 "be less than zero"); 5593 5594 // NOTE: The alignment in the directive is a power of 2 value, the assembler 5595 // may internally end up wanting an alignment in bytes. 5596 // FIXME: Diagnose overflow. 5597 if (Pow2Alignment < 0) 5598 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 5599 "alignment, can't be less than zero"); 5600 5601 Sym->redefineIfPossible(); 5602 if (!Sym->isUndefined()) 5603 return Error(IDLoc, "invalid symbol redefinition"); 5604 5605 // Create the Symbol as a common or local common with Size and Pow2Alignment. 5606 if (IsLocal) { 5607 getStreamer().emitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment); 5608 return false; 5609 } 5610 5611 getStreamer().emitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 5612 return false; 5613 } 5614 5615 /// parseDirectiveComment 5616 /// ::= comment delimiter [[text]] 5617 /// [[text]] 5618 /// [[text]] delimiter [[text]] 5619 bool MasmParser::parseDirectiveComment(SMLoc DirectiveLoc) { 5620 StringRef FirstLine = parseStringToEndOfStatement(); 5621 size_t DelimiterEnd = FirstLine.find_first_of("\b\t\v\f\r\x1A "); 5622 StringRef Delimiter = FirstLine.take_front(DelimiterEnd); 5623 if (Delimiter.empty()) 5624 return Error(DirectiveLoc, "no delimiter in 'comment' directive"); 5625 do { 5626 if (getTok().is(AsmToken::Eof)) 5627 return Error(DirectiveLoc, "unmatched delimiter in 'comment' directive"); 5628 Lex(); // eat end of statement 5629 } while (!parseStringToEndOfStatement().contains(Delimiter)); 5630 return parseToken(AsmToken::EndOfStatement, 5631 "unexpected token in 'comment' directive"); 5632 } 5633 5634 /// parseDirectiveInclude 5635 /// ::= include <filename> 5636 /// | include filename 5637 bool MasmParser::parseDirectiveInclude() { 5638 // Allow the strings to have escaped octal character sequence. 5639 std::string Filename; 5640 SMLoc IncludeLoc = getTok().getLoc(); 5641 5642 if (!parseAngleBracketString(Filename)) 5643 Filename = parseStringToEndOfStatement().str(); 5644 if (check(!Filename.empty(), "missing filename in 'include' directive") || 5645 check(getTok().isNot(AsmToken::EndOfStatement), 5646 "unexpected token in 'include' directive") || 5647 // Attempt to switch the lexer to the included file before consuming the 5648 // end of statement to avoid losing it when we switch. 5649 check(enterIncludeFile(Filename), IncludeLoc, 5650 "Could not find include file '" + Filename + "'")) 5651 return true; 5652 5653 return false; 5654 } 5655 5656 /// parseDirectiveIf 5657 /// ::= .if{,eq,ge,gt,le,lt,ne} expression 5658 bool MasmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { 5659 TheCondStack.push_back(TheCondState); 5660 TheCondState.TheCond = AsmCond::IfCond; 5661 if (TheCondState.Ignore) { 5662 eatToEndOfStatement(); 5663 } else { 5664 int64_t ExprValue; 5665 if (parseAbsoluteExpression(ExprValue) || 5666 parseToken(AsmToken::EndOfStatement, 5667 "unexpected token in '.if' directive")) 5668 return true; 5669 5670 switch (DirKind) { 5671 default: 5672 llvm_unreachable("unsupported directive"); 5673 case DK_IF: 5674 break; 5675 case DK_IFE: 5676 ExprValue = ExprValue == 0; 5677 break; 5678 } 5679 5680 TheCondState.CondMet = ExprValue; 5681 TheCondState.Ignore = !TheCondState.CondMet; 5682 } 5683 5684 return false; 5685 } 5686 5687 /// parseDirectiveIfb 5688 /// ::= .ifb string 5689 bool MasmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5690 TheCondStack.push_back(TheCondState); 5691 TheCondState.TheCond = AsmCond::IfCond; 5692 5693 if (TheCondState.Ignore) { 5694 eatToEndOfStatement(); 5695 } else { 5696 std::string Str; 5697 if (parseTextItem(Str)) 5698 return TokError("expected string parameter for 'ifb' directive"); 5699 5700 if (parseToken(AsmToken::EndOfStatement, 5701 "unexpected token in 'ifb' directive")) 5702 return true; 5703 5704 TheCondState.CondMet = ExpectBlank == Str.empty(); 5705 TheCondState.Ignore = !TheCondState.CondMet; 5706 } 5707 5708 return false; 5709 } 5710 5711 /// parseDirectiveIfidn 5712 /// ::= ifidn string1, string2 5713 bool MasmParser::parseDirectiveIfidn(SMLoc DirectiveLoc, bool ExpectEqual, bool CaseInsensitive) { 5714 std::string String1, String2; 5715 5716 if (parseTextItem(String1)) { 5717 if (ExpectEqual) 5718 return TokError("expected string parameter for 'ifidn' directive"); 5719 return TokError("expected string parameter for 'ifdif' directive"); 5720 } 5721 5722 if (Lexer.isNot(AsmToken::Comma)) { 5723 if (ExpectEqual) 5724 return TokError( 5725 "expected comma after first string for 'ifidn' directive"); 5726 return TokError("expected comma after first string for 'ifdif' directive"); 5727 } 5728 Lex(); 5729 5730 if (parseTextItem(String2)) { 5731 if (ExpectEqual) 5732 return TokError("expected string parameter for 'ifidn' directive"); 5733 return TokError("expected string parameter for 'ifdif' directive"); 5734 } 5735 5736 TheCondStack.push_back(TheCondState); 5737 TheCondState.TheCond = AsmCond::IfCond; 5738 if (CaseInsensitive) 5739 TheCondState.CondMet = 5740 ExpectEqual == (StringRef(String1).equals_lower(String2)); 5741 else 5742 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5743 TheCondState.Ignore = !TheCondState.CondMet; 5744 5745 return false; 5746 } 5747 5748 /// parseDirectiveIfdef 5749 /// ::= ifdef symbol 5750 /// | ifdef variable 5751 bool MasmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 5752 TheCondStack.push_back(TheCondState); 5753 TheCondState.TheCond = AsmCond::IfCond; 5754 5755 if (TheCondState.Ignore) { 5756 eatToEndOfStatement(); 5757 } else { 5758 bool is_defined = false; 5759 unsigned RegNo; 5760 SMLoc StartLoc, EndLoc; 5761 is_defined = (getTargetParser().tryParseRegister( 5762 RegNo, StartLoc, EndLoc) == MatchOperand_Success); 5763 if (!is_defined) { 5764 StringRef Name; 5765 if (check(parseIdentifier(Name), "expected identifier after 'ifdef'") || 5766 parseToken(AsmToken::EndOfStatement, "unexpected token in 'ifdef'")) 5767 return true; 5768 5769 if (Variables.find(Name) != Variables.end()) { 5770 is_defined = true; 5771 } else { 5772 MCSymbol *Sym = getContext().lookupSymbol(Name); 5773 is_defined = (Sym && !Sym->isUndefined(false)); 5774 } 5775 } 5776 5777 TheCondState.CondMet = (is_defined == expect_defined); 5778 TheCondState.Ignore = !TheCondState.CondMet; 5779 } 5780 5781 return false; 5782 } 5783 5784 /// parseDirectiveElseIf 5785 /// ::= elseif expression 5786 bool MasmParser::parseDirectiveElseIf(SMLoc DirectiveLoc, 5787 DirectiveKind DirKind) { 5788 if (TheCondState.TheCond != AsmCond::IfCond && 5789 TheCondState.TheCond != AsmCond::ElseIfCond) 5790 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an" 5791 " .if or an .elseif"); 5792 TheCondState.TheCond = AsmCond::ElseIfCond; 5793 5794 bool LastIgnoreState = false; 5795 if (!TheCondStack.empty()) 5796 LastIgnoreState = TheCondStack.back().Ignore; 5797 if (LastIgnoreState || TheCondState.CondMet) { 5798 TheCondState.Ignore = true; 5799 eatToEndOfStatement(); 5800 } else { 5801 int64_t ExprValue; 5802 if (parseAbsoluteExpression(ExprValue)) 5803 return true; 5804 5805 if (parseToken(AsmToken::EndOfStatement, 5806 "unexpected token in '.elseif' directive")) 5807 return true; 5808 5809 switch (DirKind) { 5810 default: 5811 llvm_unreachable("unsupported directive"); 5812 case DK_ELSEIF: 5813 break; 5814 case DK_ELSEIFE: 5815 ExprValue = ExprValue == 0; 5816 break; 5817 } 5818 5819 TheCondState.CondMet = ExprValue; 5820 TheCondState.Ignore = !TheCondState.CondMet; 5821 } 5822 5823 return false; 5824 } 5825 5826 /// parseDirectiveElseIfb 5827 /// ::= elseifb expression 5828 bool MasmParser::parseDirectiveElseIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5829 if (TheCondState.TheCond != AsmCond::IfCond && 5830 TheCondState.TheCond != AsmCond::ElseIfCond) 5831 return Error(DirectiveLoc, "Encountered an elseif that doesn't follow an" 5832 " if or an elseif"); 5833 TheCondState.TheCond = AsmCond::ElseIfCond; 5834 5835 bool LastIgnoreState = false; 5836 if (!TheCondStack.empty()) 5837 LastIgnoreState = TheCondStack.back().Ignore; 5838 if (LastIgnoreState || TheCondState.CondMet) { 5839 TheCondState.Ignore = true; 5840 eatToEndOfStatement(); 5841 } else { 5842 std::string Str; 5843 if (parseTextItem(Str)) 5844 return TokError("expected string parameter for 'elseifb' directive"); 5845 5846 if (parseToken(AsmToken::EndOfStatement, 5847 "unexpected token in 'elseifb' directive")) 5848 return true; 5849 5850 TheCondState.CondMet = ExpectBlank == Str.empty(); 5851 TheCondState.Ignore = !TheCondState.CondMet; 5852 } 5853 5854 return false; 5855 } 5856 5857 /// parseDirectiveElseIfdef 5858 /// ::= elseifdef symbol 5859 /// | elseifdef variable 5860 bool MasmParser::parseDirectiveElseIfdef(SMLoc DirectiveLoc, 5861 bool expect_defined) { 5862 if (TheCondState.TheCond != AsmCond::IfCond && 5863 TheCondState.TheCond != AsmCond::ElseIfCond) 5864 return Error(DirectiveLoc, "Encountered an elseif that doesn't follow an" 5865 " if or an elseif"); 5866 TheCondState.TheCond = AsmCond::ElseIfCond; 5867 5868 bool LastIgnoreState = false; 5869 if (!TheCondStack.empty()) 5870 LastIgnoreState = TheCondStack.back().Ignore; 5871 if (LastIgnoreState || TheCondState.CondMet) { 5872 TheCondState.Ignore = true; 5873 eatToEndOfStatement(); 5874 } else { 5875 bool is_defined = false; 5876 unsigned RegNo; 5877 SMLoc StartLoc, EndLoc; 5878 is_defined = (getTargetParser().tryParseRegister(RegNo, StartLoc, EndLoc) == 5879 MatchOperand_Success); 5880 if (!is_defined) { 5881 StringRef Name; 5882 if (check(parseIdentifier(Name), 5883 "expected identifier after 'elseifdef'") || 5884 parseToken(AsmToken::EndOfStatement, 5885 "unexpected token in 'elseifdef'")) 5886 return true; 5887 5888 if (Variables.find(Name) != Variables.end()) { 5889 is_defined = true; 5890 } else { 5891 MCSymbol *Sym = getContext().lookupSymbol(Name); 5892 is_defined = (Sym && !Sym->isUndefined(false)); 5893 } 5894 } 5895 5896 TheCondState.CondMet = (is_defined == expect_defined); 5897 TheCondState.Ignore = !TheCondState.CondMet; 5898 } 5899 5900 return false; 5901 } 5902 5903 /// parseDirectiveElseIfidn 5904 /// ::= elseifidn string1, string2 5905 bool MasmParser::parseDirectiveElseIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 5906 bool CaseInsensitive) { 5907 if (TheCondState.TheCond != AsmCond::IfCond && 5908 TheCondState.TheCond != AsmCond::ElseIfCond) 5909 return Error(DirectiveLoc, "Encountered an elseif that doesn't follow an" 5910 " if or an elseif"); 5911 TheCondState.TheCond = AsmCond::ElseIfCond; 5912 5913 bool LastIgnoreState = false; 5914 if (!TheCondStack.empty()) 5915 LastIgnoreState = TheCondStack.back().Ignore; 5916 if (LastIgnoreState || TheCondState.CondMet) { 5917 TheCondState.Ignore = true; 5918 eatToEndOfStatement(); 5919 } else { 5920 std::string String1, String2; 5921 5922 if (parseTextItem(String1)) { 5923 if (ExpectEqual) 5924 return TokError("expected string parameter for 'elseifidn' directive"); 5925 return TokError("expected string parameter for 'elseifdif' directive"); 5926 } 5927 5928 if (Lexer.isNot(AsmToken::Comma)) { 5929 if (ExpectEqual) 5930 return TokError( 5931 "expected comma after first string for 'elseifidn' directive"); 5932 return TokError( 5933 "expected comma after first string for 'elseifdif' directive"); 5934 } 5935 Lex(); 5936 5937 if (parseTextItem(String2)) { 5938 if (ExpectEqual) 5939 return TokError("expected string parameter for 'elseifidn' directive"); 5940 return TokError("expected string parameter for 'elseifdif' directive"); 5941 } 5942 5943 if (CaseInsensitive) 5944 TheCondState.CondMet = 5945 ExpectEqual == (StringRef(String1).equals_lower(String2)); 5946 else 5947 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5948 TheCondState.Ignore = !TheCondState.CondMet; 5949 } 5950 5951 return false; 5952 } 5953 5954 /// parseDirectiveElse 5955 /// ::= else 5956 bool MasmParser::parseDirectiveElse(SMLoc DirectiveLoc) { 5957 if (parseToken(AsmToken::EndOfStatement, 5958 "unexpected token in 'else' directive")) 5959 return true; 5960 5961 if (TheCondState.TheCond != AsmCond::IfCond && 5962 TheCondState.TheCond != AsmCond::ElseIfCond) 5963 return Error(DirectiveLoc, "Encountered an else that doesn't follow an if" 5964 " or an elseif"); 5965 TheCondState.TheCond = AsmCond::ElseCond; 5966 bool LastIgnoreState = false; 5967 if (!TheCondStack.empty()) 5968 LastIgnoreState = TheCondStack.back().Ignore; 5969 if (LastIgnoreState || TheCondState.CondMet) 5970 TheCondState.Ignore = true; 5971 else 5972 TheCondState.Ignore = false; 5973 5974 return false; 5975 } 5976 5977 /// parseDirectiveEnd 5978 /// ::= end 5979 bool MasmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { 5980 if (parseToken(AsmToken::EndOfStatement, 5981 "unexpected token in 'end' directive")) 5982 return true; 5983 5984 while (Lexer.isNot(AsmToken::Eof)) 5985 Lexer.Lex(); 5986 5987 return false; 5988 } 5989 5990 /// parseDirectiveError 5991 /// ::= .err [message] 5992 bool MasmParser::parseDirectiveError(SMLoc DirectiveLoc) { 5993 if (!TheCondStack.empty()) { 5994 if (TheCondStack.back().Ignore) { 5995 eatToEndOfStatement(); 5996 return false; 5997 } 5998 } 5999 6000 StringRef Message = ".err directive invoked in source file"; 6001 if (Lexer.isNot(AsmToken::EndOfStatement)) 6002 Message = parseStringToEndOfStatement(); 6003 Lex(); 6004 6005 return Error(DirectiveLoc, Message); 6006 } 6007 6008 /// parseDirectiveErrorIfb 6009 /// ::= .errb textitem[, message] 6010 bool MasmParser::parseDirectiveErrorIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 6011 if (!TheCondStack.empty()) { 6012 if (TheCondStack.back().Ignore) { 6013 eatToEndOfStatement(); 6014 return false; 6015 } 6016 } 6017 6018 std::string Text; 6019 if (parseTextItem(Text)) 6020 return Error(getTok().getLoc(), "missing text item in '.errb' directive"); 6021 6022 StringRef Message = ".errb directive invoked in source file"; 6023 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6024 if (parseToken(AsmToken::Comma)) 6025 return addErrorSuffix(" in '.errb' directive"); 6026 Message = parseStringToEndOfStatement(); 6027 } 6028 Lex(); 6029 6030 if (Text.empty() == ExpectBlank) 6031 return Error(DirectiveLoc, Message); 6032 return false; 6033 } 6034 6035 /// parseDirectiveErrorIfdef 6036 /// ::= .errdef name[, message] 6037 bool MasmParser::parseDirectiveErrorIfdef(SMLoc DirectiveLoc, 6038 bool ExpectDefined) { 6039 if (!TheCondStack.empty()) { 6040 if (TheCondStack.back().Ignore) { 6041 eatToEndOfStatement(); 6042 return false; 6043 } 6044 } 6045 6046 bool IsDefined = false; 6047 unsigned RegNo; 6048 SMLoc StartLoc, EndLoc; 6049 IsDefined = (getTargetParser().tryParseRegister(RegNo, StartLoc, EndLoc) == 6050 MatchOperand_Success); 6051 if (!IsDefined) { 6052 StringRef Name; 6053 if (check(parseIdentifier(Name), "expected identifier after '.errdef'")) 6054 return true; 6055 6056 if (Variables.find(Name) != Variables.end()) { 6057 IsDefined = true; 6058 } else { 6059 MCSymbol *Sym = getContext().lookupSymbol(Name); 6060 IsDefined = (Sym && !Sym->isUndefined(false)); 6061 } 6062 } 6063 6064 StringRef Message = ".errdef directive invoked in source file"; 6065 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6066 if (parseToken(AsmToken::Comma)) 6067 return addErrorSuffix(" in '.errdef' directive"); 6068 Message = parseStringToEndOfStatement(); 6069 } 6070 Lex(); 6071 6072 if (IsDefined == ExpectDefined) 6073 return Error(DirectiveLoc, Message); 6074 return false; 6075 } 6076 6077 /// parseDirectiveErrorIfidn 6078 /// ::= .erridn textitem1, textitem2[, message] 6079 bool MasmParser::parseDirectiveErrorIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 6080 bool CaseInsensitive) { 6081 if (!TheCondStack.empty()) { 6082 if (TheCondStack.back().Ignore) { 6083 eatToEndOfStatement(); 6084 return false; 6085 } 6086 } 6087 6088 std::string String1, String2; 6089 6090 if (parseTextItem(String1)) { 6091 if (ExpectEqual) 6092 return TokError("expected string parameter for '.erridn' directive"); 6093 return TokError("expected string parameter for '.errdif' directive"); 6094 } 6095 6096 if (Lexer.isNot(AsmToken::Comma)) { 6097 if (ExpectEqual) 6098 return TokError( 6099 "expected comma after first string for '.erridn' directive"); 6100 return TokError( 6101 "expected comma after first string for '.errdif' directive"); 6102 } 6103 Lex(); 6104 6105 if (parseTextItem(String2)) { 6106 if (ExpectEqual) 6107 return TokError("expected string parameter for '.erridn' directive"); 6108 return TokError("expected string parameter for '.errdif' directive"); 6109 } 6110 6111 StringRef Message; 6112 if (ExpectEqual) 6113 Message = ".erridn directive invoked in source file"; 6114 else 6115 Message = ".errdif directive invoked in source file"; 6116 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6117 if (parseToken(AsmToken::Comma)) 6118 return addErrorSuffix(" in '.erridn' directive"); 6119 Message = parseStringToEndOfStatement(); 6120 } 6121 Lex(); 6122 6123 if (CaseInsensitive) 6124 TheCondState.CondMet = 6125 ExpectEqual == (StringRef(String1).equals_lower(String2)); 6126 else 6127 TheCondState.CondMet = ExpectEqual == (String1 == String2); 6128 TheCondState.Ignore = !TheCondState.CondMet; 6129 6130 if ((CaseInsensitive && 6131 ExpectEqual == StringRef(String1).equals_lower(String2)) || 6132 (ExpectEqual == (String1 == String2))) 6133 return Error(DirectiveLoc, Message); 6134 return false; 6135 } 6136 6137 /// parseDirectiveErrorIfe 6138 /// ::= .erre expression[, message] 6139 bool MasmParser::parseDirectiveErrorIfe(SMLoc DirectiveLoc, bool ExpectZero) { 6140 if (!TheCondStack.empty()) { 6141 if (TheCondStack.back().Ignore) { 6142 eatToEndOfStatement(); 6143 return false; 6144 } 6145 } 6146 6147 int64_t ExprValue; 6148 if (parseAbsoluteExpression(ExprValue)) 6149 return addErrorSuffix(" in '.erre' directive"); 6150 6151 StringRef Message = ".erre directive invoked in source file"; 6152 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6153 if (parseToken(AsmToken::Comma)) 6154 return addErrorSuffix(" in '.erre' directive"); 6155 Message = parseStringToEndOfStatement(); 6156 } 6157 Lex(); 6158 6159 if ((ExprValue == 0) == ExpectZero) 6160 return Error(DirectiveLoc, Message); 6161 return false; 6162 } 6163 6164 /// parseDirectiveEndIf 6165 /// ::= .endif 6166 bool MasmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { 6167 if (parseToken(AsmToken::EndOfStatement, 6168 "unexpected token in '.endif' directive")) 6169 return true; 6170 6171 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) 6172 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow " 6173 "an .if or .else"); 6174 if (!TheCondStack.empty()) { 6175 TheCondState = TheCondStack.back(); 6176 TheCondStack.pop_back(); 6177 } 6178 6179 return false; 6180 } 6181 6182 void MasmParser::initializeDirectiveKindMap() { 6183 DirectiveKindMap["="] = DK_ASSIGN; 6184 DirectiveKindMap["equ"] = DK_EQU; 6185 DirectiveKindMap["textequ"] = DK_TEXTEQU; 6186 // DirectiveKindMap[".ascii"] = DK_ASCII; 6187 // DirectiveKindMap[".asciz"] = DK_ASCIZ; 6188 // DirectiveKindMap[".string"] = DK_STRING; 6189 DirectiveKindMap["byte"] = DK_BYTE; 6190 DirectiveKindMap["sbyte"] = DK_SBYTE; 6191 DirectiveKindMap["word"] = DK_WORD; 6192 DirectiveKindMap["sword"] = DK_SWORD; 6193 DirectiveKindMap["dword"] = DK_DWORD; 6194 DirectiveKindMap["sdword"] = DK_SDWORD; 6195 DirectiveKindMap["fword"] = DK_FWORD; 6196 DirectiveKindMap["qword"] = DK_QWORD; 6197 DirectiveKindMap["sqword"] = DK_SQWORD; 6198 DirectiveKindMap["real4"] = DK_REAL4; 6199 DirectiveKindMap["real8"] = DK_REAL8; 6200 DirectiveKindMap["align"] = DK_ALIGN; 6201 // DirectiveKindMap[".org"] = DK_ORG; 6202 DirectiveKindMap["extern"] = DK_EXTERN; 6203 DirectiveKindMap["public"] = DK_PUBLIC; 6204 // DirectiveKindMap[".comm"] = DK_COMM; 6205 DirectiveKindMap["comment"] = DK_COMMENT; 6206 DirectiveKindMap["include"] = DK_INCLUDE; 6207 // DirectiveKindMap[".rept"] = DK_REPT; 6208 // DirectiveKindMap[".rep"] = DK_REPT; 6209 // DirectiveKindMap[".irp"] = DK_IRP; 6210 // DirectiveKindMap[".irpc"] = DK_IRPC; 6211 // DirectiveKindMap[".endr"] = DK_ENDR; 6212 DirectiveKindMap["if"] = DK_IF; 6213 DirectiveKindMap["ife"] = DK_IFE; 6214 DirectiveKindMap["ifb"] = DK_IFB; 6215 DirectiveKindMap["ifnb"] = DK_IFNB; 6216 DirectiveKindMap["ifdef"] = DK_IFDEF; 6217 DirectiveKindMap["ifndef"] = DK_IFNDEF; 6218 DirectiveKindMap["ifdif"] = DK_IFDIF; 6219 DirectiveKindMap["ifdifi"] = DK_IFDIFI; 6220 DirectiveKindMap["ifidn"] = DK_IFIDN; 6221 DirectiveKindMap["ifidni"] = DK_IFIDNI; 6222 DirectiveKindMap["elseif"] = DK_ELSEIF; 6223 DirectiveKindMap["elseifdef"] = DK_ELSEIFDEF; 6224 DirectiveKindMap["elseifndef"] = DK_ELSEIFNDEF; 6225 DirectiveKindMap["elseifdif"] = DK_ELSEIFDIF; 6226 DirectiveKindMap["elseifidn"] = DK_ELSEIFIDN; 6227 DirectiveKindMap["else"] = DK_ELSE; 6228 DirectiveKindMap["end"] = DK_END; 6229 DirectiveKindMap["endif"] = DK_ENDIF; 6230 // DirectiveKindMap[".file"] = DK_FILE; 6231 // DirectiveKindMap[".line"] = DK_LINE; 6232 // DirectiveKindMap[".loc"] = DK_LOC; 6233 // DirectiveKindMap[".stabs"] = DK_STABS; 6234 // DirectiveKindMap[".cv_file"] = DK_CV_FILE; 6235 // DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; 6236 // DirectiveKindMap[".cv_loc"] = DK_CV_LOC; 6237 // DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; 6238 // DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; 6239 // DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; 6240 // DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; 6241 // DirectiveKindMap[".cv_string"] = DK_CV_STRING; 6242 // DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; 6243 // DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; 6244 // DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; 6245 // DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; 6246 // DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; 6247 // DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; 6248 // DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; 6249 // DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; 6250 // DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; 6251 // DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; 6252 // DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; 6253 // DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; 6254 // DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; 6255 // DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; 6256 // DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; 6257 // DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; 6258 // DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; 6259 // DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; 6260 // DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; 6261 // DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; 6262 // DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; 6263 // DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; 6264 // DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; 6265 // DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; 6266 // DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; 6267 // DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; 6268 // DirectiveKindMap[".macro"] = DK_MACRO; 6269 // DirectiveKindMap[".exitm"] = DK_EXITM; 6270 // DirectiveKindMap[".endm"] = DK_ENDM; 6271 // DirectiveKindMap[".purgem"] = DK_PURGEM; 6272 DirectiveKindMap[".err"] = DK_ERR; 6273 DirectiveKindMap[".errb"] = DK_ERRB; 6274 DirectiveKindMap[".errnb"] = DK_ERRNB; 6275 DirectiveKindMap[".errdef"] = DK_ERRDEF; 6276 DirectiveKindMap[".errndef"] = DK_ERRNDEF; 6277 DirectiveKindMap[".errdif"] = DK_ERRDIF; 6278 DirectiveKindMap[".errdifi"] = DK_ERRDIFI; 6279 DirectiveKindMap[".erridn"] = DK_ERRIDN; 6280 DirectiveKindMap[".erridni"] = DK_ERRIDNI; 6281 DirectiveKindMap[".erre"] = DK_ERRE; 6282 DirectiveKindMap[".errnz"] = DK_ERRNZ; 6283 // DirectiveKindMap[".altmacro"] = DK_ALTMACRO; 6284 // DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO; 6285 DirectiveKindMap["db"] = DK_DB; 6286 DirectiveKindMap["dd"] = DK_DD; 6287 DirectiveKindMap["dq"] = DK_DQ; 6288 DirectiveKindMap["dw"] = DK_DW; 6289 DirectiveKindMap["echo"] = DK_ECHO; 6290 DirectiveKindMap["struc"] = DK_STRUCT; 6291 DirectiveKindMap["struct"] = DK_STRUCT; 6292 DirectiveKindMap["union"] = DK_UNION; 6293 DirectiveKindMap["ends"] = DK_ENDS; 6294 } 6295 6296 MCAsmMacro *MasmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { 6297 AsmToken EndToken, StartToken = getTok(); 6298 6299 unsigned NestLevel = 0; 6300 while (true) { 6301 // Check whether we have reached the end of the file. 6302 if (getLexer().is(AsmToken::Eof)) { 6303 printError(DirectiveLoc, "no matching '.endr' in definition"); 6304 return nullptr; 6305 } 6306 6307 if (Lexer.is(AsmToken::Identifier) && 6308 (getTok().getIdentifier() == ".rep" || 6309 getTok().getIdentifier() == ".rept" || 6310 getTok().getIdentifier() == ".irp" || 6311 getTok().getIdentifier() == ".irpc")) { 6312 ++NestLevel; 6313 } 6314 6315 // Otherwise, check whether we have reached the .endr. 6316 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") { 6317 if (NestLevel == 0) { 6318 EndToken = getTok(); 6319 Lex(); 6320 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6321 printError(getTok().getLoc(), 6322 "unexpected token in '.endr' directive"); 6323 return nullptr; 6324 } 6325 break; 6326 } 6327 --NestLevel; 6328 } 6329 6330 // Otherwise, scan till the end of the statement. 6331 eatToEndOfStatement(); 6332 } 6333 6334 const char *BodyStart = StartToken.getLoc().getPointer(); 6335 const char *BodyEnd = EndToken.getLoc().getPointer(); 6336 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 6337 6338 // We Are Anonymous. 6339 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters()); 6340 return &MacroLikeBodies.back(); 6341 } 6342 6343 void MasmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 6344 raw_svector_ostream &OS) { 6345 OS << ".endr\n"; 6346 6347 std::unique_ptr<MemoryBuffer> Instantiation = 6348 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 6349 6350 // Create the macro instantiation object and add to the current macro 6351 // instantiation stack. 6352 MacroInstantiation *MI = new MacroInstantiation{ 6353 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 6354 ActiveMacros.push_back(MI); 6355 6356 // Jump to the macro instantiation and prime the lexer. 6357 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 6358 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 6359 Lex(); 6360 } 6361 6362 /// parseDirectiveRept 6363 /// ::= .rep | .rept count 6364 bool MasmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) { 6365 const MCExpr *CountExpr; 6366 SMLoc CountLoc = getTok().getLoc(); 6367 if (parseExpression(CountExpr)) 6368 return true; 6369 6370 int64_t Count; 6371 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) { 6372 return Error(CountLoc, "unexpected token in '" + Dir + "' directive"); 6373 } 6374 6375 if (check(Count < 0, CountLoc, "Count is negative") || 6376 parseToken(AsmToken::EndOfStatement, 6377 "unexpected token in '" + Dir + "' directive")) 6378 return true; 6379 6380 // Lex the rept definition. 6381 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 6382 if (!M) 6383 return true; 6384 6385 // Macro instantiation is lexical, unfortunately. We construct a new buffer 6386 // to hold the macro body with substitutions. 6387 SmallString<256> Buf; 6388 raw_svector_ostream OS(Buf); 6389 while (Count--) { 6390 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t). 6391 if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc())) 6392 return true; 6393 } 6394 instantiateMacroLikeBody(M, DirectiveLoc, OS); 6395 6396 return false; 6397 } 6398 6399 /// parseDirectiveIrp 6400 /// ::= .irp symbol,values 6401 bool MasmParser::parseDirectiveIrp(SMLoc DirectiveLoc) { 6402 MCAsmMacroParameter Parameter; 6403 MCAsmMacroArguments A; 6404 if (check(parseIdentifier(Parameter.Name), 6405 "expected identifier in '.irp' directive") || 6406 parseToken(AsmToken::Comma, "expected comma in '.irp' directive") || 6407 parseMacroArguments(nullptr, A) || 6408 parseToken(AsmToken::EndOfStatement, "expected End of Statement")) 6409 return true; 6410 6411 // Lex the irp definition. 6412 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 6413 if (!M) 6414 return true; 6415 6416 // Macro instantiation is lexical, unfortunately. We construct a new buffer 6417 // to hold the macro body with substitutions. 6418 SmallString<256> Buf; 6419 raw_svector_ostream OS(Buf); 6420 6421 for (const MCAsmMacroArgument &Arg : A) { 6422 // Note that the AtPseudoVariable is enabled for instantiations of .irp. 6423 // This is undocumented, but GAS seems to support it. 6424 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 6425 return true; 6426 } 6427 6428 instantiateMacroLikeBody(M, DirectiveLoc, OS); 6429 6430 return false; 6431 } 6432 6433 /// parseDirectiveIrpc 6434 /// ::= .irpc symbol,values 6435 bool MasmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) { 6436 MCAsmMacroParameter Parameter; 6437 MCAsmMacroArguments A; 6438 6439 if (check(parseIdentifier(Parameter.Name), 6440 "expected identifier in '.irpc' directive") || 6441 parseToken(AsmToken::Comma, "expected comma in '.irpc' directive") || 6442 parseMacroArguments(nullptr, A)) 6443 return true; 6444 6445 if (A.size() != 1 || A.front().size() != 1) 6446 return TokError("unexpected token in '.irpc' directive"); 6447 6448 // Eat the end of statement. 6449 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 6450 return true; 6451 6452 // Lex the irpc definition. 6453 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 6454 if (!M) 6455 return true; 6456 6457 // Macro instantiation is lexical, unfortunately. We construct a new buffer 6458 // to hold the macro body with substitutions. 6459 SmallString<256> Buf; 6460 raw_svector_ostream OS(Buf); 6461 6462 StringRef Values = A.front().front().getString(); 6463 for (std::size_t I = 0, End = Values.size(); I != End; ++I) { 6464 MCAsmMacroArgument Arg; 6465 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1)); 6466 6467 // Note that the AtPseudoVariable is enabled for instantiations of .irpc. 6468 // This is undocumented, but GAS seems to support it. 6469 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 6470 return true; 6471 } 6472 6473 instantiateMacroLikeBody(M, DirectiveLoc, OS); 6474 6475 return false; 6476 } 6477 6478 bool MasmParser::parseDirectiveEndr(SMLoc DirectiveLoc) { 6479 if (ActiveMacros.empty()) 6480 return TokError("unmatched '.endr' directive"); 6481 6482 // The only .repl that should get here are the ones created by 6483 // instantiateMacroLikeBody. 6484 assert(getLexer().is(AsmToken::EndOfStatement)); 6485 6486 handleMacroExit(); 6487 return false; 6488 } 6489 6490 bool MasmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, 6491 size_t Len) { 6492 const MCExpr *Value; 6493 SMLoc ExprLoc = getLexer().getLoc(); 6494 if (parseExpression(Value)) 6495 return true; 6496 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 6497 if (!MCE) 6498 return Error(ExprLoc, "unexpected expression in _emit"); 6499 uint64_t IntValue = MCE->getValue(); 6500 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue)) 6501 return Error(ExprLoc, "literal value out of range for directive"); 6502 6503 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len); 6504 return false; 6505 } 6506 6507 bool MasmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { 6508 const MCExpr *Value; 6509 SMLoc ExprLoc = getLexer().getLoc(); 6510 if (parseExpression(Value)) 6511 return true; 6512 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 6513 if (!MCE) 6514 return Error(ExprLoc, "unexpected expression in align"); 6515 uint64_t IntValue = MCE->getValue(); 6516 if (!isPowerOf2_64(IntValue)) 6517 return Error(ExprLoc, "literal value not a power of two greater then zero"); 6518 6519 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue)); 6520 return false; 6521 } 6522 6523 bool MasmParser::parseDirectiveEcho() { 6524 StringRef Message = parseStringToEndOfStatement(); 6525 Lex(); // eat end of statement 6526 llvm::outs() << Message << '\n'; 6527 return false; 6528 } 6529 6530 // We are comparing pointers, but the pointers are relative to a single string. 6531 // Thus, this should always be deterministic. 6532 static int rewritesSort(const AsmRewrite *AsmRewriteA, 6533 const AsmRewrite *AsmRewriteB) { 6534 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) 6535 return -1; 6536 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) 6537 return 1; 6538 6539 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output 6540 // rewrite to the same location. Make sure the SizeDirective rewrite is 6541 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This 6542 // ensures the sort algorithm is stable. 6543 if (AsmRewritePrecedence[AsmRewriteA->Kind] > 6544 AsmRewritePrecedence[AsmRewriteB->Kind]) 6545 return -1; 6546 6547 if (AsmRewritePrecedence[AsmRewriteA->Kind] < 6548 AsmRewritePrecedence[AsmRewriteB->Kind]) 6549 return 1; 6550 llvm_unreachable("Unstable rewrite sort."); 6551 } 6552 6553 bool MasmParser::lookUpField(StringRef Name, StringRef &Type, 6554 unsigned &Offset) const { 6555 const std::pair<StringRef, StringRef> BaseMember = Name.split('.'); 6556 const StringRef Base = BaseMember.first, Member = BaseMember.second; 6557 return lookUpField(Base, Member, Type, Offset); 6558 } 6559 6560 bool MasmParser::lookUpField(StringRef Base, StringRef Member, StringRef &Type, 6561 unsigned &Offset) const { 6562 if (Base.empty()) 6563 return true; 6564 6565 unsigned BaseOffset = 0; 6566 if (Base.contains('.') && !lookUpField(Base, Type, BaseOffset)) 6567 Base = Type; 6568 6569 auto TypeIt = KnownType.find(Base); 6570 if (TypeIt != KnownType.end()) 6571 return lookUpField(*TypeIt->second, Member, Type, Offset); 6572 6573 auto StructIt = Structs.find(Base.lower()); 6574 if (StructIt != Structs.end()) 6575 return lookUpField(StructIt->second, Member, Type, Offset); 6576 6577 return true; 6578 } 6579 6580 bool MasmParser::lookUpField(const StructInfo &Structure, StringRef Member, 6581 StringRef &Type, unsigned &Offset) const { 6582 if (Member.empty()) { 6583 Type = Structure.Name; 6584 return false; 6585 } 6586 6587 std::pair<StringRef, StringRef> Split = Member.split('.'); 6588 const StringRef FieldName = Split.first, FieldMember = Split.second; 6589 6590 auto StructIt = Structs.find(FieldName.lower()); 6591 if (StructIt != Structs.end()) 6592 return lookUpField(StructIt->second, FieldMember, Type, Offset); 6593 6594 auto FieldIt = Structure.FieldsByName.find(FieldName.lower()); 6595 if (FieldIt == Structure.FieldsByName.end()) 6596 return true; 6597 6598 const FieldInfo &Field = Structure.Fields[FieldIt->second]; 6599 if (FieldMember.empty()) { 6600 Offset += Field.Offset; 6601 if (Field.Contents.FT == FT_STRUCT) 6602 Type = Field.Contents.StructInfo.Structure.Name; 6603 return false; 6604 } 6605 6606 if (Field.Contents.FT != FT_STRUCT) 6607 return true; 6608 const StructFieldInfo &StructInfo = Field.Contents.StructInfo; 6609 6610 bool Result = lookUpField(StructInfo.Structure, FieldMember, Type, Offset); 6611 if (Result) 6612 return true; 6613 6614 Offset += Field.Offset; 6615 return false; 6616 } 6617 6618 bool MasmParser::parseMSInlineAsm( 6619 void *AsmLoc, std::string &AsmString, unsigned &NumOutputs, 6620 unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 6621 SmallVectorImpl<std::string> &Constraints, 6622 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, 6623 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { 6624 SmallVector<void *, 4> InputDecls; 6625 SmallVector<void *, 4> OutputDecls; 6626 SmallVector<bool, 4> InputDeclsAddressOf; 6627 SmallVector<bool, 4> OutputDeclsAddressOf; 6628 SmallVector<std::string, 4> InputConstraints; 6629 SmallVector<std::string, 4> OutputConstraints; 6630 SmallVector<unsigned, 4> ClobberRegs; 6631 6632 SmallVector<AsmRewrite, 4> AsmStrRewrites; 6633 6634 // Prime the lexer. 6635 Lex(); 6636 6637 // While we have input, parse each statement. 6638 unsigned InputIdx = 0; 6639 unsigned OutputIdx = 0; 6640 while (getLexer().isNot(AsmToken::Eof)) { 6641 // Parse curly braces marking block start/end. 6642 if (parseCurlyBlockScope(AsmStrRewrites)) 6643 continue; 6644 6645 ParseStatementInfo Info(&AsmStrRewrites); 6646 bool StatementErr = parseStatement(Info, &SI); 6647 6648 if (StatementErr || Info.ParseError) { 6649 // Emit pending errors if any exist. 6650 printPendingErrors(); 6651 return true; 6652 } 6653 6654 // No pending error should exist here. 6655 assert(!hasPendingError() && "unexpected error from parseStatement"); 6656 6657 if (Info.Opcode == ~0U) 6658 continue; 6659 6660 const MCInstrDesc &Desc = MII->get(Info.Opcode); 6661 6662 // Build the list of clobbers, outputs and inputs. 6663 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { 6664 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; 6665 6666 // Register operand. 6667 if (Operand.isReg() && !Operand.needAddressOf() && 6668 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) { 6669 unsigned NumDefs = Desc.getNumDefs(); 6670 // Clobber. 6671 if (NumDefs && Operand.getMCOperandNum() < NumDefs) 6672 ClobberRegs.push_back(Operand.getReg()); 6673 continue; 6674 } 6675 6676 // Expr/Input or Output. 6677 StringRef SymName = Operand.getSymName(); 6678 if (SymName.empty()) 6679 continue; 6680 6681 void *OpDecl = Operand.getOpDecl(); 6682 if (!OpDecl) 6683 continue; 6684 6685 StringRef Constraint = Operand.getConstraint(); 6686 if (Operand.isImm()) { 6687 // Offset as immediate. 6688 if (Operand.isOffsetOfLocal()) 6689 Constraint = "r"; 6690 else 6691 Constraint = "i"; 6692 } 6693 6694 bool isOutput = (i == 1) && Desc.mayStore(); 6695 SMLoc Start = SMLoc::getFromPointer(SymName.data()); 6696 if (isOutput) { 6697 ++InputIdx; 6698 OutputDecls.push_back(OpDecl); 6699 OutputDeclsAddressOf.push_back(Operand.needAddressOf()); 6700 OutputConstraints.push_back(("=" + Constraint).str()); 6701 AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size()); 6702 } else { 6703 InputDecls.push_back(OpDecl); 6704 InputDeclsAddressOf.push_back(Operand.needAddressOf()); 6705 InputConstraints.push_back(Constraint.str()); 6706 if (Desc.OpInfo[i - 1].isBranchTarget()) 6707 AsmStrRewrites.emplace_back(AOK_CallInput, Start, SymName.size()); 6708 else 6709 AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size()); 6710 } 6711 } 6712 6713 // Consider implicit defs to be clobbers. Think of cpuid and push. 6714 ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(), 6715 Desc.getNumImplicitDefs()); 6716 ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end()); 6717 } 6718 6719 // Set the number of Outputs and Inputs. 6720 NumOutputs = OutputDecls.size(); 6721 NumInputs = InputDecls.size(); 6722 6723 // Set the unique clobbers. 6724 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end()); 6725 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()), 6726 ClobberRegs.end()); 6727 Clobbers.assign(ClobberRegs.size(), std::string()); 6728 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { 6729 raw_string_ostream OS(Clobbers[I]); 6730 IP->printRegName(OS, ClobberRegs[I]); 6731 } 6732 6733 // Merge the various outputs and inputs. Output are expected first. 6734 if (NumOutputs || NumInputs) { 6735 unsigned NumExprs = NumOutputs + NumInputs; 6736 OpDecls.resize(NumExprs); 6737 Constraints.resize(NumExprs); 6738 for (unsigned i = 0; i < NumOutputs; ++i) { 6739 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]); 6740 Constraints[i] = OutputConstraints[i]; 6741 } 6742 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { 6743 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]); 6744 Constraints[j] = InputConstraints[i]; 6745 } 6746 } 6747 6748 // Build the IR assembly string. 6749 std::string AsmStringIR; 6750 raw_string_ostream OS(AsmStringIR); 6751 StringRef ASMString = 6752 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer(); 6753 const char *AsmStart = ASMString.begin(); 6754 const char *AsmEnd = ASMString.end(); 6755 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort); 6756 for (auto it = AsmStrRewrites.begin(); it != AsmStrRewrites.end(); ++it) { 6757 const AsmRewrite &AR = *it; 6758 // Check if this has already been covered by another rewrite... 6759 if (AR.Done) 6760 continue; 6761 AsmRewriteKind Kind = AR.Kind; 6762 6763 const char *Loc = AR.Loc.getPointer(); 6764 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!"); 6765 6766 // Emit everything up to the immediate/expression. 6767 if (unsigned Len = Loc - AsmStart) 6768 OS << StringRef(AsmStart, Len); 6769 6770 // Skip the original expression. 6771 if (Kind == AOK_Skip) { 6772 AsmStart = Loc + AR.Len; 6773 continue; 6774 } 6775 6776 unsigned AdditionalSkip = 0; 6777 // Rewrite expressions in $N notation. 6778 switch (Kind) { 6779 default: 6780 break; 6781 case AOK_IntelExpr: 6782 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression"); 6783 if (AR.IntelExp.NeedBracs) 6784 OS << "["; 6785 if (AR.IntelExp.hasBaseReg()) 6786 OS << AR.IntelExp.BaseReg; 6787 if (AR.IntelExp.hasIndexReg()) 6788 OS << (AR.IntelExp.hasBaseReg() ? " + " : "") 6789 << AR.IntelExp.IndexReg; 6790 if (AR.IntelExp.Scale > 1) 6791 OS << " * $$" << AR.IntelExp.Scale; 6792 if (AR.IntelExp.hasOffset()) { 6793 if (AR.IntelExp.hasRegs()) 6794 OS << " + "; 6795 // Fuse this rewrite with a rewrite of the offset name, if present. 6796 StringRef OffsetName = AR.IntelExp.OffsetName; 6797 SMLoc OffsetLoc = SMLoc::getFromPointer(AR.IntelExp.OffsetName.data()); 6798 size_t OffsetLen = OffsetName.size(); 6799 auto rewrite_it = std::find_if( 6800 it, AsmStrRewrites.end(), [&](const AsmRewrite &FusingAR) { 6801 return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen && 6802 (FusingAR.Kind == AOK_Input || 6803 FusingAR.Kind == AOK_CallInput); 6804 }); 6805 if (rewrite_it == AsmStrRewrites.end()) { 6806 OS << "offset " << OffsetName; 6807 } else if (rewrite_it->Kind == AOK_CallInput) { 6808 OS << "${" << InputIdx++ << ":P}"; 6809 rewrite_it->Done = true; 6810 } else { 6811 OS << '$' << InputIdx++; 6812 rewrite_it->Done = true; 6813 } 6814 } 6815 if (AR.IntelExp.Imm || AR.IntelExp.emitImm()) 6816 OS << (AR.IntelExp.emitImm() ? "$$" : " + $$") << AR.IntelExp.Imm; 6817 if (AR.IntelExp.NeedBracs) 6818 OS << "]"; 6819 break; 6820 case AOK_Label: 6821 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; 6822 break; 6823 case AOK_Input: 6824 OS << '$' << InputIdx++; 6825 break; 6826 case AOK_CallInput: 6827 OS << "${" << InputIdx++ << ":P}"; 6828 break; 6829 case AOK_Output: 6830 OS << '$' << OutputIdx++; 6831 break; 6832 case AOK_SizeDirective: 6833 switch (AR.Val) { 6834 default: break; 6835 case 8: OS << "byte ptr "; break; 6836 case 16: OS << "word ptr "; break; 6837 case 32: OS << "dword ptr "; break; 6838 case 64: OS << "qword ptr "; break; 6839 case 80: OS << "xword ptr "; break; 6840 case 128: OS << "xmmword ptr "; break; 6841 case 256: OS << "ymmword ptr "; break; 6842 } 6843 break; 6844 case AOK_Emit: 6845 OS << ".byte"; 6846 break; 6847 case AOK_Align: { 6848 // MS alignment directives are measured in bytes. If the native assembler 6849 // measures alignment in bytes, we can pass it straight through. 6850 OS << ".align"; 6851 if (getContext().getAsmInfo()->getAlignmentIsInBytes()) 6852 break; 6853 6854 // Alignment is in log2 form, so print that instead and skip the original 6855 // immediate. 6856 unsigned Val = AR.Val; 6857 OS << ' ' << Val; 6858 assert(Val < 10 && "Expected alignment less then 2^10."); 6859 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; 6860 break; 6861 } 6862 case AOK_EVEN: 6863 OS << ".even"; 6864 break; 6865 case AOK_EndOfStatement: 6866 OS << "\n\t"; 6867 break; 6868 } 6869 6870 // Skip the original expression. 6871 AsmStart = Loc + AR.Len + AdditionalSkip; 6872 } 6873 6874 // Emit the remainder of the asm string. 6875 if (AsmStart != AsmEnd) 6876 OS << StringRef(AsmStart, AsmEnd - AsmStart); 6877 6878 AsmString = OS.str(); 6879 return false; 6880 } 6881 6882 /// Create an MCAsmParser instance. 6883 MCAsmParser *llvm::createMCMasmParser(SourceMgr &SM, MCContext &C, 6884 MCStreamer &Out, const MCAsmInfo &MAI, 6885 unsigned CB) { 6886 return new MasmParser(SM, C, Out, MAI, CB); 6887 } 6888