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