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/MC/MCAsmInfo.h" 26 #include "llvm/MC/MCCodeView.h" 27 #include "llvm/MC/MCContext.h" 28 #include "llvm/MC/MCDirectives.h" 29 #include "llvm/MC/MCDwarf.h" 30 #include "llvm/MC/MCExpr.h" 31 #include "llvm/MC/MCInstPrinter.h" 32 #include "llvm/MC/MCInstrDesc.h" 33 #include "llvm/MC/MCInstrInfo.h" 34 #include "llvm/MC/MCObjectFileInfo.h" 35 #include "llvm/MC/MCParser/AsmCond.h" 36 #include "llvm/MC/MCParser/AsmLexer.h" 37 #include "llvm/MC/MCParser/MCAsmLexer.h" 38 #include "llvm/MC/MCParser/MCAsmParser.h" 39 #include "llvm/MC/MCParser/MCAsmParserExtension.h" 40 #include "llvm/MC/MCParser/MCAsmParserUtils.h" 41 #include "llvm/MC/MCParser/MCParsedAsmOperand.h" 42 #include "llvm/MC/MCParser/MCTargetAsmParser.h" 43 #include "llvm/MC/MCRegisterInfo.h" 44 #include "llvm/MC/MCSection.h" 45 #include "llvm/MC/MCStreamer.h" 46 #include "llvm/MC/MCSymbol.h" 47 #include "llvm/MC/MCTargetOptions.h" 48 #include "llvm/MC/MCValue.h" 49 #include "llvm/Support/Casting.h" 50 #include "llvm/Support/CommandLine.h" 51 #include "llvm/Support/ErrorHandling.h" 52 #include "llvm/Support/MD5.h" 53 #include "llvm/Support/MathExtras.h" 54 #include "llvm/Support/MemoryBuffer.h" 55 #include "llvm/Support/SMLoc.h" 56 #include "llvm/Support/SourceMgr.h" 57 #include "llvm/Support/raw_ostream.h" 58 #include <algorithm> 59 #include <cassert> 60 #include <cctype> 61 #include <climits> 62 #include <cstddef> 63 #include <cstdint> 64 #include <deque> 65 #include <memory> 66 #include <sstream> 67 #include <string> 68 #include <tuple> 69 #include <utility> 70 #include <vector> 71 72 using namespace llvm; 73 74 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() = default; 75 76 static cl::opt<unsigned> AsmMacroMaxNestingDepth( 77 "asm-macro-max-nesting-depth", cl::init(20), cl::Hidden, 78 cl::desc("The maximum nesting depth allowed for assembly macros.")); 79 80 namespace { 81 82 /// Helper types for tracking macro definitions. 83 typedef std::vector<AsmToken> MCAsmMacroArgument; 84 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; 85 86 /// Helper class for storing information about an active macro 87 /// instantiation. 88 struct MacroInstantiation { 89 /// The location of the instantiation. 90 SMLoc InstantiationLoc; 91 92 /// The buffer where parsing should resume upon instantiation completion. 93 int ExitBuffer; 94 95 /// The location where parsing should resume upon instantiation completion. 96 SMLoc ExitLoc; 97 98 /// The depth of TheCondStack at the start of the instantiation. 99 size_t CondStackDepth; 100 101 public: 102 MacroInstantiation(SMLoc IL, int EB, SMLoc EL, size_t CondStackDepth); 103 }; 104 105 struct ParseStatementInfo { 106 /// The parsed operands from the last parsed statement. 107 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; 108 109 /// The opcode from the last parsed instruction. 110 unsigned Opcode = ~0U; 111 112 /// Was there an error parsing the inline assembly? 113 bool ParseError = false; 114 115 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr; 116 117 ParseStatementInfo() = delete; 118 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites) 119 : AsmRewrites(rewrites) {} 120 }; 121 122 /// The concrete assembly parser instance. 123 class AsmParser : public MCAsmParser { 124 private: 125 AsmLexer Lexer; 126 MCContext &Ctx; 127 MCStreamer &Out; 128 const MCAsmInfo &MAI; 129 SourceMgr &SrcMgr; 130 SourceMgr::DiagHandlerTy SavedDiagHandler; 131 void *SavedDiagContext; 132 std::unique_ptr<MCAsmParserExtension> PlatformParser; 133 134 /// This is the current buffer index we're lexing from as managed by the 135 /// SourceMgr object. 136 unsigned CurBuffer; 137 138 AsmCond TheCondState; 139 std::vector<AsmCond> TheCondStack; 140 141 /// maps directive names to handler methods in parser 142 /// extensions. Extensions register themselves in this map by calling 143 /// addDirectiveHandler. 144 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap; 145 146 /// Stack of active macro instantiations. 147 std::vector<MacroInstantiation*> ActiveMacros; 148 149 /// List of bodies of anonymous macros. 150 std::deque<MCAsmMacro> MacroLikeBodies; 151 152 /// Boolean tracking whether macro substitution is enabled. 153 unsigned MacrosEnabledFlag : 1; 154 155 /// Keeps track of how many .macro's have been instantiated. 156 unsigned NumOfMacroInstantiations; 157 158 /// The values from the last parsed cpp hash file line comment if any. 159 struct CppHashInfoTy { 160 StringRef Filename; 161 int64_t LineNumber; 162 SMLoc Loc; 163 unsigned Buf; 164 CppHashInfoTy() : Filename(), LineNumber(0), Loc(), Buf(0) {} 165 }; 166 CppHashInfoTy CppHashInfo; 167 168 /// The filename from the first cpp hash file line comment, if any. 169 StringRef FirstCppHashFilename; 170 171 /// List of forward directional labels for diagnosis at the end. 172 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; 173 174 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 175 unsigned AssemblerDialect = ~0U; 176 177 /// is Darwin compatibility enabled? 178 bool IsDarwin = false; 179 180 /// Are we parsing ms-style inline assembly? 181 bool ParsingInlineAsm = false; 182 183 /// Did we already inform the user about inconsistent MD5 usage? 184 bool ReportedInconsistentMD5 = false; 185 186 // Is alt macro mode enabled. 187 bool AltMacroMode = false; 188 189 public: 190 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 191 const MCAsmInfo &MAI, unsigned CB); 192 AsmParser(const AsmParser &) = delete; 193 AsmParser &operator=(const AsmParser &) = delete; 194 ~AsmParser() override; 195 196 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override; 197 198 void addDirectiveHandler(StringRef Directive, 199 ExtensionDirectiveHandler Handler) override { 200 ExtensionDirectiveMap[Directive] = Handler; 201 } 202 203 void addAliasForDirective(StringRef Directive, StringRef Alias) override { 204 DirectiveKindMap[Directive] = DirectiveKindMap[Alias]; 205 } 206 207 /// @name MCAsmParser Interface 208 /// { 209 210 SourceMgr &getSourceManager() override { return SrcMgr; } 211 MCAsmLexer &getLexer() override { return Lexer; } 212 MCContext &getContext() override { return Ctx; } 213 MCStreamer &getStreamer() override { return Out; } 214 215 CodeViewContext &getCVContext() { return Ctx.getCVContext(); } 216 217 unsigned getAssemblerDialect() override { 218 if (AssemblerDialect == ~0U) 219 return MAI.getAssemblerDialect(); 220 else 221 return AssemblerDialect; 222 } 223 void setAssemblerDialect(unsigned i) override { 224 AssemblerDialect = i; 225 } 226 227 void Note(SMLoc L, const Twine &Msg, SMRange Range = None) override; 228 bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) override; 229 bool printError(SMLoc L, const Twine &Msg, SMRange Range = None) override; 230 231 const AsmToken &Lex() override; 232 233 void setParsingInlineAsm(bool V) override { 234 ParsingInlineAsm = V; 235 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and 236 // hex integer literals. 237 Lexer.setLexMasmIntegers(V); 238 } 239 bool isParsingInlineAsm() override { return ParsingInlineAsm; } 240 241 bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString, 242 unsigned &NumOutputs, unsigned &NumInputs, 243 SmallVectorImpl<std::pair<void *,bool>> &OpDecls, 244 SmallVectorImpl<std::string> &Constraints, 245 SmallVectorImpl<std::string> &Clobbers, 246 const MCInstrInfo *MII, const MCInstPrinter *IP, 247 MCAsmParserSemaCallback &SI) override; 248 249 bool parseExpression(const MCExpr *&Res); 250 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 251 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override; 252 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 253 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 254 SMLoc &EndLoc) override; 255 bool parseAbsoluteExpression(int64_t &Res) override; 256 257 /// Parse a floating point expression using the float \p Semantics 258 /// and set \p Res to the value. 259 bool parseRealValue(const fltSemantics &Semantics, APInt &Res); 260 261 /// Parse an identifier or string (as a quoted identifier) 262 /// and set \p Res to the identifier contents. 263 bool parseIdentifier(StringRef &Res) override; 264 void eatToEndOfStatement() override; 265 266 bool checkForValidSection() override; 267 268 /// } 269 270 private: 271 bool parseStatement(ParseStatementInfo &Info, 272 MCAsmParserSemaCallback *SI); 273 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); 274 bool parseCppHashLineFilenameComment(SMLoc L); 275 276 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body, 277 ArrayRef<MCAsmMacroParameter> Parameters); 278 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 279 ArrayRef<MCAsmMacroParameter> Parameters, 280 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable, 281 SMLoc L); 282 283 /// Are macros enabled in the parser? 284 bool areMacrosEnabled() {return MacrosEnabledFlag;} 285 286 /// Control a flag in the parser that enables or disables macros. 287 void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;} 288 289 /// Are we inside a macro instantiation? 290 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();} 291 292 /// Handle entry to macro instantiation. 293 /// 294 /// \param M The macro. 295 /// \param NameLoc Instantiation location. 296 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc); 297 298 /// Handle exit from macro instantiation. 299 void handleMacroExit(); 300 301 /// Extract AsmTokens for a macro argument. 302 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg); 303 304 /// Parse all macro arguments for a given macro. 305 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A); 306 307 void printMacroInstantiations(); 308 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 309 SMRange Range = None) const { 310 ArrayRef<SMRange> Ranges(Range); 311 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 312 } 313 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 314 315 /// Should we emit DWARF describing this assembler source? (Returns false if 316 /// the source has .file directives, which means we don't want to generate 317 /// info describing the assembler source itself.) 318 bool enabledGenDwarfForAssembly(); 319 320 /// Enter the specified file. This returns true on failure. 321 bool enterIncludeFile(const std::string &Filename); 322 323 /// Process the specified file for the .incbin directive. 324 /// This returns true on failure. 325 bool processIncbinFile(const std::string &Filename, int64_t Skip = 0, 326 const MCExpr *Count = nullptr, SMLoc Loc = SMLoc()); 327 328 /// Reset the current lexer position to that given by \p Loc. The 329 /// current token is not set; clients should ensure Lex() is called 330 /// subsequently. 331 /// 332 /// \param InBuffer If not 0, should be the known buffer id that contains the 333 /// location. 334 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0); 335 336 /// Parse up to the end of statement and a return the contents from the 337 /// current token until the end of the statement; the current token on exit 338 /// will be either the EndOfStatement or EOF. 339 StringRef parseStringToEndOfStatement() override; 340 341 /// Parse until the end of a statement or a comma is encountered, 342 /// return the contents from the current token up to the end or comma. 343 StringRef parseStringToComma(); 344 345 bool parseAssignment(StringRef Name, bool allow_redef, 346 bool NoDeadStrip = false); 347 348 unsigned getBinOpPrecedence(AsmToken::TokenKind K, 349 MCBinaryExpr::Opcode &Kind); 350 351 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 352 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 353 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 354 355 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 356 357 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName); 358 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName); 359 360 // Generic (target and platform independent) directive parsing. 361 enum DirectiveKind { 362 DK_NO_DIRECTIVE, // Placeholder 363 DK_SET, 364 DK_EQU, 365 DK_EQUIV, 366 DK_ASCII, 367 DK_ASCIZ, 368 DK_STRING, 369 DK_BYTE, 370 DK_SHORT, 371 DK_RELOC, 372 DK_VALUE, 373 DK_2BYTE, 374 DK_LONG, 375 DK_INT, 376 DK_4BYTE, 377 DK_QUAD, 378 DK_8BYTE, 379 DK_OCTA, 380 DK_DC, 381 DK_DC_A, 382 DK_DC_B, 383 DK_DC_D, 384 DK_DC_L, 385 DK_DC_S, 386 DK_DC_W, 387 DK_DC_X, 388 DK_DCB, 389 DK_DCB_B, 390 DK_DCB_D, 391 DK_DCB_L, 392 DK_DCB_S, 393 DK_DCB_W, 394 DK_DCB_X, 395 DK_DS, 396 DK_DS_B, 397 DK_DS_D, 398 DK_DS_L, 399 DK_DS_P, 400 DK_DS_S, 401 DK_DS_W, 402 DK_DS_X, 403 DK_SINGLE, 404 DK_FLOAT, 405 DK_DOUBLE, 406 DK_ALIGN, 407 DK_ALIGN32, 408 DK_BALIGN, 409 DK_BALIGNW, 410 DK_BALIGNL, 411 DK_P2ALIGN, 412 DK_P2ALIGNW, 413 DK_P2ALIGNL, 414 DK_ORG, 415 DK_FILL, 416 DK_ENDR, 417 DK_BUNDLE_ALIGN_MODE, 418 DK_BUNDLE_LOCK, 419 DK_BUNDLE_UNLOCK, 420 DK_ZERO, 421 DK_EXTERN, 422 DK_GLOBL, 423 DK_GLOBAL, 424 DK_LAZY_REFERENCE, 425 DK_NO_DEAD_STRIP, 426 DK_SYMBOL_RESOLVER, 427 DK_PRIVATE_EXTERN, 428 DK_REFERENCE, 429 DK_WEAK_DEFINITION, 430 DK_WEAK_REFERENCE, 431 DK_WEAK_DEF_CAN_BE_HIDDEN, 432 DK_COLD, 433 DK_COMM, 434 DK_COMMON, 435 DK_LCOMM, 436 DK_ABORT, 437 DK_INCLUDE, 438 DK_INCBIN, 439 DK_CODE16, 440 DK_CODE16GCC, 441 DK_REPT, 442 DK_IRP, 443 DK_IRPC, 444 DK_IF, 445 DK_IFEQ, 446 DK_IFGE, 447 DK_IFGT, 448 DK_IFLE, 449 DK_IFLT, 450 DK_IFNE, 451 DK_IFB, 452 DK_IFNB, 453 DK_IFC, 454 DK_IFEQS, 455 DK_IFNC, 456 DK_IFNES, 457 DK_IFDEF, 458 DK_IFNDEF, 459 DK_IFNOTDEF, 460 DK_ELSEIF, 461 DK_ELSE, 462 DK_ENDIF, 463 DK_SPACE, 464 DK_SKIP, 465 DK_FILE, 466 DK_LINE, 467 DK_LOC, 468 DK_STABS, 469 DK_CV_FILE, 470 DK_CV_FUNC_ID, 471 DK_CV_INLINE_SITE_ID, 472 DK_CV_LOC, 473 DK_CV_LINETABLE, 474 DK_CV_INLINE_LINETABLE, 475 DK_CV_DEF_RANGE, 476 DK_CV_STRINGTABLE, 477 DK_CV_STRING, 478 DK_CV_FILECHECKSUMS, 479 DK_CV_FILECHECKSUM_OFFSET, 480 DK_CV_FPO_DATA, 481 DK_CFI_SECTIONS, 482 DK_CFI_STARTPROC, 483 DK_CFI_ENDPROC, 484 DK_CFI_DEF_CFA, 485 DK_CFI_DEF_CFA_OFFSET, 486 DK_CFI_ADJUST_CFA_OFFSET, 487 DK_CFI_DEF_CFA_REGISTER, 488 DK_CFI_OFFSET, 489 DK_CFI_REL_OFFSET, 490 DK_CFI_PERSONALITY, 491 DK_CFI_LSDA, 492 DK_CFI_REMEMBER_STATE, 493 DK_CFI_RESTORE_STATE, 494 DK_CFI_SAME_VALUE, 495 DK_CFI_RESTORE, 496 DK_CFI_ESCAPE, 497 DK_CFI_RETURN_COLUMN, 498 DK_CFI_SIGNAL_FRAME, 499 DK_CFI_UNDEFINED, 500 DK_CFI_REGISTER, 501 DK_CFI_WINDOW_SAVE, 502 DK_CFI_B_KEY_FRAME, 503 DK_MACROS_ON, 504 DK_MACROS_OFF, 505 DK_ALTMACRO, 506 DK_NOALTMACRO, 507 DK_MACRO, 508 DK_EXITM, 509 DK_ENDM, 510 DK_ENDMACRO, 511 DK_PURGEM, 512 DK_SLEB128, 513 DK_ULEB128, 514 DK_ERR, 515 DK_ERROR, 516 DK_WARNING, 517 DK_PRINT, 518 DK_ADDRSIG, 519 DK_ADDRSIG_SYM, 520 DK_END 521 }; 522 523 /// Maps directive name --> DirectiveKind enum, for 524 /// directives parsed by this class. 525 StringMap<DirectiveKind> DirectiveKindMap; 526 527 // Codeview def_range type parsing. 528 enum CVDefRangeType { 529 CVDR_DEFRANGE = 0, // Placeholder 530 CVDR_DEFRANGE_REGISTER, 531 CVDR_DEFRANGE_FRAMEPOINTER_REL, 532 CVDR_DEFRANGE_SUBFIELD_REGISTER, 533 CVDR_DEFRANGE_REGISTER_REL 534 }; 535 536 /// Maps Codeview def_range types --> CVDefRangeType enum, for 537 /// Codeview def_range types parsed by this class. 538 StringMap<CVDefRangeType> CVDefRangeTypeMap; 539 540 // ".ascii", ".asciz", ".string" 541 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 542 bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc" 543 bool parseDirectiveValue(StringRef IDVal, 544 unsigned Size); // ".byte", ".long", ... 545 bool parseDirectiveOctaValue(StringRef IDVal); // ".octa", ... 546 bool parseDirectiveRealValue(StringRef IDVal, 547 const fltSemantics &); // ".single", ... 548 bool parseDirectiveFill(); // ".fill" 549 bool parseDirectiveZero(); // ".zero" 550 // ".set", ".equ", ".equiv" 551 bool parseDirectiveSet(StringRef IDVal, bool allow_redef); 552 bool parseDirectiveOrg(); // ".org" 553 // ".align{,32}", ".p2align{,w,l}" 554 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize); 555 556 // ".file", ".line", ".loc", ".stabs" 557 bool parseDirectiveFile(SMLoc DirectiveLoc); 558 bool parseDirectiveLine(); 559 bool parseDirectiveLoc(); 560 bool parseDirectiveStabs(); 561 562 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable", 563 // ".cv_inline_linetable", ".cv_def_range", ".cv_string" 564 bool parseDirectiveCVFile(); 565 bool parseDirectiveCVFuncId(); 566 bool parseDirectiveCVInlineSiteId(); 567 bool parseDirectiveCVLoc(); 568 bool parseDirectiveCVLinetable(); 569 bool parseDirectiveCVInlineLinetable(); 570 bool parseDirectiveCVDefRange(); 571 bool parseDirectiveCVString(); 572 bool parseDirectiveCVStringTable(); 573 bool parseDirectiveCVFileChecksums(); 574 bool parseDirectiveCVFileChecksumOffset(); 575 bool parseDirectiveCVFPOData(); 576 577 // .cfi directives 578 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc); 579 bool parseDirectiveCFIWindowSave(); 580 bool parseDirectiveCFISections(); 581 bool parseDirectiveCFIStartProc(); 582 bool parseDirectiveCFIEndProc(); 583 bool parseDirectiveCFIDefCfaOffset(); 584 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc); 585 bool parseDirectiveCFIAdjustCfaOffset(); 586 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc); 587 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc); 588 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc); 589 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality); 590 bool parseDirectiveCFIRememberState(); 591 bool parseDirectiveCFIRestoreState(); 592 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc); 593 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc); 594 bool parseDirectiveCFIEscape(); 595 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc); 596 bool parseDirectiveCFISignalFrame(); 597 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc); 598 599 // macro directives 600 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc); 601 bool parseDirectiveExitMacro(StringRef Directive); 602 bool parseDirectiveEndMacro(StringRef Directive); 603 bool parseDirectiveMacro(SMLoc DirectiveLoc); 604 bool parseDirectiveMacrosOnOff(StringRef Directive); 605 // alternate macro mode directives 606 bool parseDirectiveAltmacro(StringRef Directive); 607 // ".bundle_align_mode" 608 bool parseDirectiveBundleAlignMode(); 609 // ".bundle_lock" 610 bool parseDirectiveBundleLock(); 611 // ".bundle_unlock" 612 bool parseDirectiveBundleUnlock(); 613 614 // ".space", ".skip" 615 bool parseDirectiveSpace(StringRef IDVal); 616 617 // ".dcb" 618 bool parseDirectiveDCB(StringRef IDVal, unsigned Size); 619 bool parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &); 620 // ".ds" 621 bool parseDirectiveDS(StringRef IDVal, unsigned Size); 622 623 // .sleb128 (Signed=true) and .uleb128 (Signed=false) 624 bool parseDirectiveLEB128(bool Signed); 625 626 /// Parse a directive like ".globl" which 627 /// accepts a single symbol (which should be a label or an external). 628 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr); 629 630 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 631 632 bool parseDirectiveAbort(); // ".abort" 633 bool parseDirectiveInclude(); // ".include" 634 bool parseDirectiveIncbin(); // ".incbin" 635 636 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne" 637 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 638 // ".ifb" or ".ifnb", depending on ExpectBlank. 639 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 640 // ".ifc" or ".ifnc", depending on ExpectEqual. 641 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual); 642 // ".ifeqs" or ".ifnes", depending on ExpectEqual. 643 bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual); 644 // ".ifdef" or ".ifndef", depending on expect_defined 645 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 646 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 647 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else" 648 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 649 bool parseEscapedString(std::string &Data) override; 650 651 const MCExpr *applyModifierToExpr(const MCExpr *E, 652 MCSymbolRefExpr::VariantKind Variant); 653 654 // Macro-like directives 655 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc); 656 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 657 raw_svector_ostream &OS); 658 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive); 659 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 660 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc" 661 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 662 663 // "_emit" or "__emit" 664 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info, 665 size_t Len); 666 667 // "align" 668 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info); 669 670 // "end" 671 bool parseDirectiveEnd(SMLoc DirectiveLoc); 672 673 // ".err" or ".error" 674 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage); 675 676 // ".warning" 677 bool parseDirectiveWarning(SMLoc DirectiveLoc); 678 679 // .print <double-quotes-string> 680 bool parseDirectivePrint(SMLoc DirectiveLoc); 681 682 // Directives to support address-significance tables. 683 bool parseDirectiveAddrsig(); 684 bool parseDirectiveAddrsigSym(); 685 686 void initializeDirectiveKindMap(); 687 void initializeCVDefRangeTypeMap(); 688 }; 689 690 } // end anonymous namespace 691 692 namespace llvm { 693 694 extern MCAsmParserExtension *createDarwinAsmParser(); 695 extern MCAsmParserExtension *createELFAsmParser(); 696 extern MCAsmParserExtension *createCOFFAsmParser(); 697 extern MCAsmParserExtension *createWasmAsmParser(); 698 699 } // end namespace llvm 700 701 enum { DEFAULT_ADDRSPACE = 0 }; 702 703 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 704 const MCAsmInfo &MAI, unsigned CB = 0) 705 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM), 706 CurBuffer(CB ? CB : SM.getMainFileID()), MacrosEnabledFlag(true) { 707 HadError = false; 708 // Save the old handler. 709 SavedDiagHandler = SrcMgr.getDiagHandler(); 710 SavedDiagContext = SrcMgr.getDiagContext(); 711 // Set our own handler which calls the saved handler. 712 SrcMgr.setDiagHandler(DiagHandler, this); 713 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 714 715 // Initialize the platform / file format parser. 716 switch (Ctx.getObjectFileInfo()->getObjectFileType()) { 717 case MCObjectFileInfo::IsCOFF: 718 PlatformParser.reset(createCOFFAsmParser()); 719 break; 720 case MCObjectFileInfo::IsMachO: 721 PlatformParser.reset(createDarwinAsmParser()); 722 IsDarwin = true; 723 break; 724 case MCObjectFileInfo::IsELF: 725 PlatformParser.reset(createELFAsmParser()); 726 break; 727 case MCObjectFileInfo::IsWasm: 728 PlatformParser.reset(createWasmAsmParser()); 729 break; 730 case MCObjectFileInfo::IsXCOFF: 731 report_fatal_error( 732 "Need to implement createXCOFFAsmParser for XCOFF format."); 733 break; 734 } 735 736 PlatformParser->Initialize(*this); 737 initializeDirectiveKindMap(); 738 initializeCVDefRangeTypeMap(); 739 740 NumOfMacroInstantiations = 0; 741 } 742 743 AsmParser::~AsmParser() { 744 assert((HadError || ActiveMacros.empty()) && 745 "Unexpected active macro instantiation!"); 746 747 // Restore the saved diagnostics handler and context for use during 748 // finalization. 749 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext); 750 } 751 752 void AsmParser::printMacroInstantiations() { 753 // Print the active macro instantiation stack. 754 for (std::vector<MacroInstantiation *>::const_reverse_iterator 755 it = ActiveMacros.rbegin(), 756 ie = ActiveMacros.rend(); 757 it != ie; ++it) 758 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 759 "while in macro instantiation"); 760 } 761 762 void AsmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { 763 printPendingErrors(); 764 printMessage(L, SourceMgr::DK_Note, Msg, Range); 765 printMacroInstantiations(); 766 } 767 768 bool AsmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) { 769 if(getTargetParser().getTargetOptions().MCNoWarn) 770 return false; 771 if (getTargetParser().getTargetOptions().MCFatalWarnings) 772 return Error(L, Msg, Range); 773 printMessage(L, SourceMgr::DK_Warning, Msg, Range); 774 printMacroInstantiations(); 775 return false; 776 } 777 778 bool AsmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { 779 HadError = true; 780 printMessage(L, SourceMgr::DK_Error, Msg, Range); 781 printMacroInstantiations(); 782 return true; 783 } 784 785 bool AsmParser::enterIncludeFile(const std::string &Filename) { 786 std::string IncludedFile; 787 unsigned NewBuf = 788 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 789 if (!NewBuf) 790 return true; 791 792 CurBuffer = NewBuf; 793 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 794 return false; 795 } 796 797 /// Process the specified .incbin file by searching for it in the include paths 798 /// then just emitting the byte contents of the file to the streamer. This 799 /// returns true on failure. 800 bool AsmParser::processIncbinFile(const std::string &Filename, int64_t Skip, 801 const MCExpr *Count, SMLoc Loc) { 802 std::string IncludedFile; 803 unsigned NewBuf = 804 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 805 if (!NewBuf) 806 return true; 807 808 // Pick up the bytes from the file and emit them. 809 StringRef Bytes = SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(); 810 Bytes = Bytes.drop_front(Skip); 811 if (Count) { 812 int64_t Res; 813 if (!Count->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 814 return Error(Loc, "expected absolute expression"); 815 if (Res < 0) 816 return Warning(Loc, "negative count has no effect"); 817 Bytes = Bytes.take_front(Res); 818 } 819 getStreamer().EmitBytes(Bytes); 820 return false; 821 } 822 823 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) { 824 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); 825 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), 826 Loc.getPointer()); 827 } 828 829 const AsmToken &AsmParser::Lex() { 830 if (Lexer.getTok().is(AsmToken::Error)) 831 Error(Lexer.getErrLoc(), Lexer.getErr()); 832 833 // if it's a end of statement with a comment in it 834 if (getTok().is(AsmToken::EndOfStatement)) { 835 // if this is a line comment output it. 836 if (!getTok().getString().empty() && getTok().getString().front() != '\n' && 837 getTok().getString().front() != '\r' && MAI.preserveAsmComments()) 838 Out.addExplicitComment(Twine(getTok().getString())); 839 } 840 841 const AsmToken *tok = &Lexer.Lex(); 842 843 // Parse comments here to be deferred until end of next statement. 844 while (tok->is(AsmToken::Comment)) { 845 if (MAI.preserveAsmComments()) 846 Out.addExplicitComment(Twine(tok->getString())); 847 tok = &Lexer.Lex(); 848 } 849 850 if (tok->is(AsmToken::Eof)) { 851 // If this is the end of an included file, pop the parent file off the 852 // include stack. 853 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 854 if (ParentIncludeLoc != SMLoc()) { 855 jumpToLoc(ParentIncludeLoc); 856 return Lex(); 857 } 858 } 859 860 return *tok; 861 } 862 863 bool AsmParser::enabledGenDwarfForAssembly() { 864 // Check whether the user specified -g. 865 if (!getContext().getGenDwarfForAssembly()) 866 return false; 867 // If we haven't encountered any .file directives (which would imply that 868 // the assembler source was produced with debug info already) then emit one 869 // describing the assembler source file itself. 870 if (getContext().getGenDwarfFileNumber() == 0) { 871 // Use the first #line directive for this, if any. It's preprocessed, so 872 // there is no checksum, and of course no source directive. 873 if (!FirstCppHashFilename.empty()) 874 getContext().setMCLineTableRootFile(/*CUID=*/0, 875 getContext().getCompilationDir(), 876 FirstCppHashFilename, 877 /*Cksum=*/None, /*Source=*/None); 878 const MCDwarfFile &RootFile = 879 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); 880 getContext().setGenDwarfFileNumber(getStreamer().EmitDwarfFileDirective( 881 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name, 882 RootFile.Checksum, RootFile.Source)); 883 } 884 return true; 885 } 886 887 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 888 // Create the initial section, if requested. 889 if (!NoInitialTextSection) 890 Out.InitSections(false); 891 892 // Prime the lexer. 893 Lex(); 894 895 HadError = false; 896 AsmCond StartingCondState = TheCondState; 897 SmallVector<AsmRewrite, 4> AsmStrRewrites; 898 899 // If we are generating dwarf for assembly source files save the initial text 900 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't 901 // emitting any actual debug info yet and haven't had a chance to parse any 902 // embedded .file directives.) 903 if (getContext().getGenDwarfForAssembly()) { 904 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 905 if (!Sec->getBeginSymbol()) { 906 MCSymbol *SectionStartSym = getContext().createTempSymbol(); 907 getStreamer().EmitLabel(SectionStartSym); 908 Sec->setBeginSymbol(SectionStartSym); 909 } 910 bool InsertResult = getContext().addGenDwarfSection(Sec); 911 assert(InsertResult && ".text section should not have debug info yet"); 912 (void)InsertResult; 913 } 914 915 // While we have input, parse each statement. 916 while (Lexer.isNot(AsmToken::Eof)) { 917 ParseStatementInfo Info(&AsmStrRewrites); 918 if (!parseStatement(Info, nullptr)) 919 continue; 920 921 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error 922 // for printing ErrMsg via Lex() only if no (presumably better) parser error 923 // exists. 924 if (!hasPendingError() && Lexer.getTok().is(AsmToken::Error)) { 925 Lex(); 926 } 927 928 // parseStatement returned true so may need to emit an error. 929 printPendingErrors(); 930 931 // Skipping to the next line if needed. 932 if (!getLexer().isAtStartOfStatement()) 933 eatToEndOfStatement(); 934 } 935 936 getTargetParser().onEndOfFile(); 937 printPendingErrors(); 938 939 // All errors should have been emitted. 940 assert(!hasPendingError() && "unexpected error from parseStatement"); 941 942 getTargetParser().flushPendingInstructions(getStreamer()); 943 944 if (TheCondState.TheCond != StartingCondState.TheCond || 945 TheCondState.Ignore != StartingCondState.Ignore) 946 printError(getTok().getLoc(), "unmatched .ifs or .elses"); 947 // Check to see there are no empty DwarfFile slots. 948 const auto &LineTables = getContext().getMCDwarfLineTables(); 949 if (!LineTables.empty()) { 950 unsigned Index = 0; 951 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) { 952 if (File.Name.empty() && Index != 0) 953 printError(getTok().getLoc(), "unassigned file number: " + 954 Twine(Index) + 955 " for .file directives"); 956 ++Index; 957 } 958 } 959 960 // Check to see that all assembler local symbols were actually defined. 961 // Targets that don't do subsections via symbols may not want this, though, 962 // so conservatively exclude them. Only do this if we're finalizing, though, 963 // as otherwise we won't necessarilly have seen everything yet. 964 if (!NoFinalize) { 965 if (MAI.hasSubsectionsViaSymbols()) { 966 for (const auto &TableEntry : getContext().getSymbols()) { 967 MCSymbol *Sym = TableEntry.getValue(); 968 // Variable symbols may not be marked as defined, so check those 969 // explicitly. If we know it's a variable, we have a definition for 970 // the purposes of this check. 971 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 972 // FIXME: We would really like to refer back to where the symbol was 973 // first referenced for a source location. We need to add something 974 // to track that. Currently, we just point to the end of the file. 975 printError(getTok().getLoc(), "assembler local symbol '" + 976 Sym->getName() + "' not defined"); 977 } 978 } 979 980 // Temporary symbols like the ones for directional jumps don't go in the 981 // symbol table. They also need to be diagnosed in all (final) cases. 982 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { 983 if (std::get<2>(LocSym)->isUndefined()) { 984 // Reset the state of any "# line file" directives we've seen to the 985 // context as it was at the diagnostic site. 986 CppHashInfo = std::get<1>(LocSym); 987 printError(std::get<0>(LocSym), "directional label undefined"); 988 } 989 } 990 } 991 992 // Finalize the output stream if there are no errors and if the client wants 993 // us to. 994 if (!HadError && !NoFinalize) 995 Out.Finish(); 996 997 return HadError || getContext().hadError(); 998 } 999 1000 bool AsmParser::checkForValidSection() { 1001 if (!ParsingInlineAsm && !getStreamer().getCurrentSectionOnly()) { 1002 Out.InitSections(false); 1003 return Error(getTok().getLoc(), 1004 "expected section directive before assembly directive"); 1005 } 1006 return false; 1007 } 1008 1009 /// Throw away the rest of the line for testing purposes. 1010 void AsmParser::eatToEndOfStatement() { 1011 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1012 Lexer.Lex(); 1013 1014 // Eat EOL. 1015 if (Lexer.is(AsmToken::EndOfStatement)) 1016 Lexer.Lex(); 1017 } 1018 1019 StringRef AsmParser::parseStringToEndOfStatement() { 1020 const char *Start = getTok().getLoc().getPointer(); 1021 1022 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1023 Lexer.Lex(); 1024 1025 const char *End = getTok().getLoc().getPointer(); 1026 return StringRef(Start, End - Start); 1027 } 1028 1029 StringRef AsmParser::parseStringToComma() { 1030 const char *Start = getTok().getLoc().getPointer(); 1031 1032 while (Lexer.isNot(AsmToken::EndOfStatement) && 1033 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof)) 1034 Lexer.Lex(); 1035 1036 const char *End = getTok().getLoc().getPointer(); 1037 return StringRef(Start, End - Start); 1038 } 1039 1040 /// Parse a paren expression and return it. 1041 /// NOTE: This assumes the leading '(' has already been consumed. 1042 /// 1043 /// parenexpr ::= expr) 1044 /// 1045 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1046 if (parseExpression(Res)) 1047 return true; 1048 if (Lexer.isNot(AsmToken::RParen)) 1049 return TokError("expected ')' in parentheses expression"); 1050 EndLoc = Lexer.getTok().getEndLoc(); 1051 Lex(); 1052 return false; 1053 } 1054 1055 /// Parse a bracket expression and return it. 1056 /// NOTE: This assumes the leading '[' has already been consumed. 1057 /// 1058 /// bracketexpr ::= expr] 1059 /// 1060 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1061 if (parseExpression(Res)) 1062 return true; 1063 EndLoc = getTok().getEndLoc(); 1064 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression")) 1065 return true; 1066 return false; 1067 } 1068 1069 /// Parse a primary expression and return it. 1070 /// primaryexpr ::= (parenexpr 1071 /// primaryexpr ::= symbol 1072 /// primaryexpr ::= number 1073 /// primaryexpr ::= '.' 1074 /// primaryexpr ::= ~,+,- primaryexpr 1075 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1076 SMLoc FirstTokenLoc = getLexer().getLoc(); 1077 AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); 1078 switch (FirstTokenKind) { 1079 default: 1080 return TokError("unknown token in expression"); 1081 // If we have an error assume that we've already handled it. 1082 case AsmToken::Error: 1083 return true; 1084 case AsmToken::Exclaim: 1085 Lex(); // Eat the operator. 1086 if (parsePrimaryExpr(Res, EndLoc)) 1087 return true; 1088 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc); 1089 return false; 1090 case AsmToken::Dollar: 1091 case AsmToken::At: 1092 case AsmToken::String: 1093 case AsmToken::Identifier: { 1094 StringRef Identifier; 1095 if (parseIdentifier(Identifier)) { 1096 // We may have failed but $ may be a valid token. 1097 if (getTok().is(AsmToken::Dollar)) { 1098 if (Lexer.getMAI().getDollarIsPC()) { 1099 Lex(); 1100 // This is a '$' reference, which references the current PC. Emit a 1101 // temporary label to the streamer and refer to it. 1102 MCSymbol *Sym = Ctx.createTempSymbol(); 1103 Out.EmitLabel(Sym); 1104 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, 1105 getContext()); 1106 EndLoc = FirstTokenLoc; 1107 return false; 1108 } 1109 return Error(FirstTokenLoc, "invalid token in expression"); 1110 } 1111 } 1112 // Parse symbol variant 1113 std::pair<StringRef, StringRef> Split; 1114 if (!MAI.useParensForSymbolVariant()) { 1115 if (FirstTokenKind == AsmToken::String) { 1116 if (Lexer.is(AsmToken::At)) { 1117 Lex(); // eat @ 1118 SMLoc AtLoc = getLexer().getLoc(); 1119 StringRef VName; 1120 if (parseIdentifier(VName)) 1121 return Error(AtLoc, "expected symbol variant after '@'"); 1122 1123 Split = std::make_pair(Identifier, VName); 1124 } 1125 } else { 1126 Split = Identifier.split('@'); 1127 } 1128 } else if (Lexer.is(AsmToken::LParen)) { 1129 Lex(); // eat '('. 1130 StringRef VName; 1131 parseIdentifier(VName); 1132 // eat ')'. 1133 if (parseToken(AsmToken::RParen, 1134 "unexpected token in variant, expected ')'")) 1135 return true; 1136 Split = std::make_pair(Identifier, VName); 1137 } 1138 1139 EndLoc = SMLoc::getFromPointer(Identifier.end()); 1140 1141 // This is a symbol reference. 1142 StringRef SymbolName = Identifier; 1143 if (SymbolName.empty()) 1144 return Error(getLexer().getLoc(), "expected a symbol reference"); 1145 1146 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1147 1148 // Lookup the symbol variant if used. 1149 if (!Split.second.empty()) { 1150 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1151 if (Variant != MCSymbolRefExpr::VK_Invalid) { 1152 SymbolName = Split.first; 1153 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { 1154 Variant = MCSymbolRefExpr::VK_None; 1155 } else { 1156 return Error(SMLoc::getFromPointer(Split.second.begin()), 1157 "invalid variant '" + Split.second + "'"); 1158 } 1159 } 1160 1161 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName); 1162 if (!Sym) 1163 Sym = getContext().getOrCreateSymbol(SymbolName); 1164 1165 // If this is an absolute variable reference, substitute it now to preserve 1166 // semantics in the face of reassignment. 1167 if (Sym->isVariable()) { 1168 auto V = Sym->getVariableValue(/*SetUsed*/ false); 1169 bool DoInline = isa<MCConstantExpr>(V) && !Variant; 1170 if (auto TV = dyn_cast<MCTargetExpr>(V)) 1171 DoInline = TV->inlineAssignedExpr(); 1172 if (DoInline) { 1173 if (Variant) 1174 return Error(EndLoc, "unexpected modifier on variable reference"); 1175 Res = Sym->getVariableValue(/*SetUsed*/ false); 1176 return false; 1177 } 1178 } 1179 1180 // Otherwise create a symbol ref. 1181 Res = MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc); 1182 return false; 1183 } 1184 case AsmToken::BigNum: 1185 return TokError("literal value out of range for directive"); 1186 case AsmToken::Integer: { 1187 SMLoc Loc = getTok().getLoc(); 1188 int64_t IntVal = getTok().getIntVal(); 1189 Res = MCConstantExpr::create(IntVal, getContext()); 1190 EndLoc = Lexer.getTok().getEndLoc(); 1191 Lex(); // Eat token. 1192 // Look for 'b' or 'f' following an Integer as a directional label 1193 if (Lexer.getKind() == AsmToken::Identifier) { 1194 StringRef IDVal = getTok().getString(); 1195 // Lookup the symbol variant if used. 1196 std::pair<StringRef, StringRef> Split = IDVal.split('@'); 1197 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1198 if (Split.first.size() != IDVal.size()) { 1199 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1200 if (Variant == MCSymbolRefExpr::VK_Invalid) 1201 return TokError("invalid variant '" + Split.second + "'"); 1202 IDVal = Split.first; 1203 } 1204 if (IDVal == "f" || IDVal == "b") { 1205 MCSymbol *Sym = 1206 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b"); 1207 Res = MCSymbolRefExpr::create(Sym, Variant, getContext()); 1208 if (IDVal == "b" && Sym->isUndefined()) 1209 return Error(Loc, "directional label undefined"); 1210 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym)); 1211 EndLoc = Lexer.getTok().getEndLoc(); 1212 Lex(); // Eat identifier. 1213 } 1214 } 1215 return false; 1216 } 1217 case AsmToken::Real: { 1218 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); 1219 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 1220 Res = MCConstantExpr::create(IntVal, getContext()); 1221 EndLoc = Lexer.getTok().getEndLoc(); 1222 Lex(); // Eat token. 1223 return false; 1224 } 1225 case AsmToken::Dot: { 1226 // This is a '.' reference, which references the current PC. Emit a 1227 // temporary label to the streamer and refer to it. 1228 MCSymbol *Sym = Ctx.createTempSymbol(); 1229 Out.EmitLabel(Sym); 1230 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext()); 1231 EndLoc = Lexer.getTok().getEndLoc(); 1232 Lex(); // Eat identifier. 1233 return false; 1234 } 1235 case AsmToken::LParen: 1236 Lex(); // Eat the '('. 1237 return parseParenExpr(Res, EndLoc); 1238 case AsmToken::LBrac: 1239 if (!PlatformParser->HasBracketExpressions()) 1240 return TokError("brackets expression not supported on this target"); 1241 Lex(); // Eat the '['. 1242 return parseBracketExpr(Res, EndLoc); 1243 case AsmToken::Minus: 1244 Lex(); // Eat the operator. 1245 if (parsePrimaryExpr(Res, EndLoc)) 1246 return true; 1247 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc); 1248 return false; 1249 case AsmToken::Plus: 1250 Lex(); // Eat the operator. 1251 if (parsePrimaryExpr(Res, EndLoc)) 1252 return true; 1253 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc); 1254 return false; 1255 case AsmToken::Tilde: 1256 Lex(); // Eat the operator. 1257 if (parsePrimaryExpr(Res, EndLoc)) 1258 return true; 1259 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1260 return false; 1261 // MIPS unary expression operators. The lexer won't generate these tokens if 1262 // MCAsmInfo::HasMipsExpressions is false for the target. 1263 case AsmToken::PercentCall16: 1264 case AsmToken::PercentCall_Hi: 1265 case AsmToken::PercentCall_Lo: 1266 case AsmToken::PercentDtprel_Hi: 1267 case AsmToken::PercentDtprel_Lo: 1268 case AsmToken::PercentGot: 1269 case AsmToken::PercentGot_Disp: 1270 case AsmToken::PercentGot_Hi: 1271 case AsmToken::PercentGot_Lo: 1272 case AsmToken::PercentGot_Ofst: 1273 case AsmToken::PercentGot_Page: 1274 case AsmToken::PercentGottprel: 1275 case AsmToken::PercentGp_Rel: 1276 case AsmToken::PercentHi: 1277 case AsmToken::PercentHigher: 1278 case AsmToken::PercentHighest: 1279 case AsmToken::PercentLo: 1280 case AsmToken::PercentNeg: 1281 case AsmToken::PercentPcrel_Hi: 1282 case AsmToken::PercentPcrel_Lo: 1283 case AsmToken::PercentTlsgd: 1284 case AsmToken::PercentTlsldm: 1285 case AsmToken::PercentTprel_Hi: 1286 case AsmToken::PercentTprel_Lo: 1287 Lex(); // Eat the operator. 1288 if (Lexer.isNot(AsmToken::LParen)) 1289 return TokError("expected '(' after operator"); 1290 Lex(); // Eat the operator. 1291 if (parseExpression(Res, EndLoc)) 1292 return true; 1293 if (Lexer.isNot(AsmToken::RParen)) 1294 return TokError("expected ')'"); 1295 Lex(); // Eat the operator. 1296 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx); 1297 return !Res; 1298 } 1299 } 1300 1301 bool AsmParser::parseExpression(const MCExpr *&Res) { 1302 SMLoc EndLoc; 1303 return parseExpression(Res, EndLoc); 1304 } 1305 1306 const MCExpr * 1307 AsmParser::applyModifierToExpr(const MCExpr *E, 1308 MCSymbolRefExpr::VariantKind Variant) { 1309 // Ask the target implementation about this expression first. 1310 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx); 1311 if (NewE) 1312 return NewE; 1313 // Recurse over the given expression, rebuilding it to apply the given variant 1314 // if there is exactly one symbol. 1315 switch (E->getKind()) { 1316 case MCExpr::Target: 1317 case MCExpr::Constant: 1318 return nullptr; 1319 1320 case MCExpr::SymbolRef: { 1321 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 1322 1323 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 1324 TokError("invalid variant on expression '" + getTok().getIdentifier() + 1325 "' (already modified)"); 1326 return E; 1327 } 1328 1329 return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext()); 1330 } 1331 1332 case MCExpr::Unary: { 1333 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 1334 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant); 1335 if (!Sub) 1336 return nullptr; 1337 return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext()); 1338 } 1339 1340 case MCExpr::Binary: { 1341 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 1342 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant); 1343 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant); 1344 1345 if (!LHS && !RHS) 1346 return nullptr; 1347 1348 if (!LHS) 1349 LHS = BE->getLHS(); 1350 if (!RHS) 1351 RHS = BE->getRHS(); 1352 1353 return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext()); 1354 } 1355 } 1356 1357 llvm_unreachable("Invalid expression kind!"); 1358 } 1359 1360 /// This function checks if the next token is <string> type or arithmetic. 1361 /// string that begin with character '<' must end with character '>'. 1362 /// otherwise it is arithmetics. 1363 /// If the function returns a 'true' value, 1364 /// the End argument will be filled with the last location pointed to the '>' 1365 /// character. 1366 1367 /// There is a gap between the AltMacro's documentation and the single quote 1368 /// implementation. GCC does not fully support this feature and so we will not 1369 /// support it. 1370 /// TODO: Adding single quote as a string. 1371 static bool isAltmacroString(SMLoc &StrLoc, SMLoc &EndLoc) { 1372 assert((StrLoc.getPointer() != nullptr) && 1373 "Argument to the function cannot be a NULL value"); 1374 const char *CharPtr = StrLoc.getPointer(); 1375 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && 1376 (*CharPtr != '\0')) { 1377 if (*CharPtr == '!') 1378 CharPtr++; 1379 CharPtr++; 1380 } 1381 if (*CharPtr == '>') { 1382 EndLoc = StrLoc.getFromPointer(CharPtr + 1); 1383 return true; 1384 } 1385 return false; 1386 } 1387 1388 /// creating a string without the escape characters '!'. 1389 static std::string altMacroString(StringRef AltMacroStr) { 1390 std::string Res; 1391 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) { 1392 if (AltMacroStr[Pos] == '!') 1393 Pos++; 1394 Res += AltMacroStr[Pos]; 1395 } 1396 return Res; 1397 } 1398 1399 /// Parse an expression and return it. 1400 /// 1401 /// expr ::= expr &&,|| expr -> lowest. 1402 /// expr ::= expr |,^,&,! expr 1403 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 1404 /// expr ::= expr <<,>> expr 1405 /// expr ::= expr +,- expr 1406 /// expr ::= expr *,/,% expr -> highest. 1407 /// expr ::= primaryexpr 1408 /// 1409 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1410 // Parse the expression. 1411 Res = nullptr; 1412 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || 1413 parseBinOpRHS(1, Res, EndLoc)) 1414 return true; 1415 1416 // As a special case, we support 'a op b @ modifier' by rewriting the 1417 // expression to include the modifier. This is inefficient, but in general we 1418 // expect users to use 'a@modifier op b'. 1419 if (Lexer.getKind() == AsmToken::At) { 1420 Lex(); 1421 1422 if (Lexer.isNot(AsmToken::Identifier)) 1423 return TokError("unexpected symbol modifier following '@'"); 1424 1425 MCSymbolRefExpr::VariantKind Variant = 1426 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 1427 if (Variant == MCSymbolRefExpr::VK_Invalid) 1428 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 1429 1430 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant); 1431 if (!ModifiedRes) { 1432 return TokError("invalid modifier '" + getTok().getIdentifier() + 1433 "' (no symbols present)"); 1434 } 1435 1436 Res = ModifiedRes; 1437 Lex(); 1438 } 1439 1440 // Try to constant fold it up front, if possible. Do not exploit 1441 // assembler here. 1442 int64_t Value; 1443 if (Res->evaluateAsAbsolute(Value)) 1444 Res = MCConstantExpr::create(Value, getContext()); 1445 1446 return false; 1447 } 1448 1449 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1450 Res = nullptr; 1451 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc); 1452 } 1453 1454 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 1455 SMLoc &EndLoc) { 1456 if (parseParenExpr(Res, EndLoc)) 1457 return true; 1458 1459 for (; ParenDepth > 0; --ParenDepth) { 1460 if (parseBinOpRHS(1, Res, EndLoc)) 1461 return true; 1462 1463 // We don't Lex() the last RParen. 1464 // This is the same behavior as parseParenExpression(). 1465 if (ParenDepth - 1 > 0) { 1466 EndLoc = getTok().getEndLoc(); 1467 if (parseToken(AsmToken::RParen, 1468 "expected ')' in parentheses expression")) 1469 return true; 1470 } 1471 } 1472 return false; 1473 } 1474 1475 bool AsmParser::parseAbsoluteExpression(int64_t &Res) { 1476 const MCExpr *Expr; 1477 1478 SMLoc StartLoc = Lexer.getLoc(); 1479 if (parseExpression(Expr)) 1480 return true; 1481 1482 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 1483 return Error(StartLoc, "expected absolute expression"); 1484 1485 return false; 1486 } 1487 1488 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K, 1489 MCBinaryExpr::Opcode &Kind, 1490 bool ShouldUseLogicalShr) { 1491 switch (K) { 1492 default: 1493 return 0; // not a binop. 1494 1495 // Lowest Precedence: &&, || 1496 case AsmToken::AmpAmp: 1497 Kind = MCBinaryExpr::LAnd; 1498 return 1; 1499 case AsmToken::PipePipe: 1500 Kind = MCBinaryExpr::LOr; 1501 return 1; 1502 1503 // Low Precedence: |, &, ^ 1504 // 1505 // FIXME: gas seems to support '!' as an infix operator? 1506 case AsmToken::Pipe: 1507 Kind = MCBinaryExpr::Or; 1508 return 2; 1509 case AsmToken::Caret: 1510 Kind = MCBinaryExpr::Xor; 1511 return 2; 1512 case AsmToken::Amp: 1513 Kind = MCBinaryExpr::And; 1514 return 2; 1515 1516 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 1517 case AsmToken::EqualEqual: 1518 Kind = MCBinaryExpr::EQ; 1519 return 3; 1520 case AsmToken::ExclaimEqual: 1521 case AsmToken::LessGreater: 1522 Kind = MCBinaryExpr::NE; 1523 return 3; 1524 case AsmToken::Less: 1525 Kind = MCBinaryExpr::LT; 1526 return 3; 1527 case AsmToken::LessEqual: 1528 Kind = MCBinaryExpr::LTE; 1529 return 3; 1530 case AsmToken::Greater: 1531 Kind = MCBinaryExpr::GT; 1532 return 3; 1533 case AsmToken::GreaterEqual: 1534 Kind = MCBinaryExpr::GTE; 1535 return 3; 1536 1537 // Intermediate Precedence: <<, >> 1538 case AsmToken::LessLess: 1539 Kind = MCBinaryExpr::Shl; 1540 return 4; 1541 case AsmToken::GreaterGreater: 1542 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1543 return 4; 1544 1545 // High Intermediate Precedence: +, - 1546 case AsmToken::Plus: 1547 Kind = MCBinaryExpr::Add; 1548 return 5; 1549 case AsmToken::Minus: 1550 Kind = MCBinaryExpr::Sub; 1551 return 5; 1552 1553 // Highest Precedence: *, /, % 1554 case AsmToken::Star: 1555 Kind = MCBinaryExpr::Mul; 1556 return 6; 1557 case AsmToken::Slash: 1558 Kind = MCBinaryExpr::Div; 1559 return 6; 1560 case AsmToken::Percent: 1561 Kind = MCBinaryExpr::Mod; 1562 return 6; 1563 } 1564 } 1565 1566 static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K, 1567 MCBinaryExpr::Opcode &Kind, 1568 bool ShouldUseLogicalShr) { 1569 switch (K) { 1570 default: 1571 return 0; // not a binop. 1572 1573 // Lowest Precedence: &&, || 1574 case AsmToken::AmpAmp: 1575 Kind = MCBinaryExpr::LAnd; 1576 return 2; 1577 case AsmToken::PipePipe: 1578 Kind = MCBinaryExpr::LOr; 1579 return 1; 1580 1581 // Low Precedence: ==, !=, <>, <, <=, >, >= 1582 case AsmToken::EqualEqual: 1583 Kind = MCBinaryExpr::EQ; 1584 return 3; 1585 case AsmToken::ExclaimEqual: 1586 case AsmToken::LessGreater: 1587 Kind = MCBinaryExpr::NE; 1588 return 3; 1589 case AsmToken::Less: 1590 Kind = MCBinaryExpr::LT; 1591 return 3; 1592 case AsmToken::LessEqual: 1593 Kind = MCBinaryExpr::LTE; 1594 return 3; 1595 case AsmToken::Greater: 1596 Kind = MCBinaryExpr::GT; 1597 return 3; 1598 case AsmToken::GreaterEqual: 1599 Kind = MCBinaryExpr::GTE; 1600 return 3; 1601 1602 // Low Intermediate Precedence: +, - 1603 case AsmToken::Plus: 1604 Kind = MCBinaryExpr::Add; 1605 return 4; 1606 case AsmToken::Minus: 1607 Kind = MCBinaryExpr::Sub; 1608 return 4; 1609 1610 // High Intermediate Precedence: |, &, ^ 1611 // 1612 // FIXME: gas seems to support '!' as an infix operator? 1613 case AsmToken::Pipe: 1614 Kind = MCBinaryExpr::Or; 1615 return 5; 1616 case AsmToken::Caret: 1617 Kind = MCBinaryExpr::Xor; 1618 return 5; 1619 case AsmToken::Amp: 1620 Kind = MCBinaryExpr::And; 1621 return 5; 1622 1623 // Highest Precedence: *, /, %, <<, >> 1624 case AsmToken::Star: 1625 Kind = MCBinaryExpr::Mul; 1626 return 6; 1627 case AsmToken::Slash: 1628 Kind = MCBinaryExpr::Div; 1629 return 6; 1630 case AsmToken::Percent: 1631 Kind = MCBinaryExpr::Mod; 1632 return 6; 1633 case AsmToken::LessLess: 1634 Kind = MCBinaryExpr::Shl; 1635 return 6; 1636 case AsmToken::GreaterGreater: 1637 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1638 return 6; 1639 } 1640 } 1641 1642 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K, 1643 MCBinaryExpr::Opcode &Kind) { 1644 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); 1645 return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr) 1646 : getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr); 1647 } 1648 1649 /// Parse all binary operators with precedence >= 'Precedence'. 1650 /// Res contains the LHS of the expression on input. 1651 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 1652 SMLoc &EndLoc) { 1653 SMLoc StartLoc = Lexer.getLoc(); 1654 while (true) { 1655 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 1656 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 1657 1658 // If the next token is lower precedence than we are allowed to eat, return 1659 // successfully with what we ate already. 1660 if (TokPrec < Precedence) 1661 return false; 1662 1663 Lex(); 1664 1665 // Eat the next primary expression. 1666 const MCExpr *RHS; 1667 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc)) 1668 return true; 1669 1670 // If BinOp binds less tightly with RHS than the operator after RHS, let 1671 // the pending operator take RHS as its LHS. 1672 MCBinaryExpr::Opcode Dummy; 1673 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1674 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc)) 1675 return true; 1676 1677 // Merge LHS and RHS according to operator. 1678 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc); 1679 } 1680 } 1681 1682 /// ParseStatement: 1683 /// ::= EndOfStatement 1684 /// ::= Label* Directive ...Operands... EndOfStatement 1685 /// ::= Label* Identifier OperandList* EndOfStatement 1686 bool AsmParser::parseStatement(ParseStatementInfo &Info, 1687 MCAsmParserSemaCallback *SI) { 1688 assert(!hasPendingError() && "parseStatement started with pending error"); 1689 // Eat initial spaces and comments 1690 while (Lexer.is(AsmToken::Space)) 1691 Lex(); 1692 if (Lexer.is(AsmToken::EndOfStatement)) { 1693 // if this is a line comment we can drop it safely 1694 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 1695 getTok().getString().front() == '\n') 1696 Out.AddBlankLine(); 1697 Lex(); 1698 return false; 1699 } 1700 // Statements always start with an identifier. 1701 AsmToken ID = getTok(); 1702 SMLoc IDLoc = ID.getLoc(); 1703 StringRef IDVal; 1704 int64_t LocalLabelVal = -1; 1705 if (Lexer.is(AsmToken::HashDirective)) 1706 return parseCppHashLineFilenameComment(IDLoc); 1707 // Allow an integer followed by a ':' as a directional local label. 1708 if (Lexer.is(AsmToken::Integer)) { 1709 LocalLabelVal = getTok().getIntVal(); 1710 if (LocalLabelVal < 0) { 1711 if (!TheCondState.Ignore) { 1712 Lex(); // always eat a token 1713 return Error(IDLoc, "unexpected token at start of statement"); 1714 } 1715 IDVal = ""; 1716 } else { 1717 IDVal = getTok().getString(); 1718 Lex(); // Consume the integer token to be used as an identifier token. 1719 if (Lexer.getKind() != AsmToken::Colon) { 1720 if (!TheCondState.Ignore) { 1721 Lex(); // always eat a token 1722 return Error(IDLoc, "unexpected token at start of statement"); 1723 } 1724 } 1725 } 1726 } else if (Lexer.is(AsmToken::Dot)) { 1727 // Treat '.' as a valid identifier in this context. 1728 Lex(); 1729 IDVal = "."; 1730 } else if (Lexer.is(AsmToken::LCurly)) { 1731 // Treat '{' as a valid identifier in this context. 1732 Lex(); 1733 IDVal = "{"; 1734 1735 } else if (Lexer.is(AsmToken::RCurly)) { 1736 // Treat '}' as a valid identifier in this context. 1737 Lex(); 1738 IDVal = "}"; 1739 } else if (Lexer.is(AsmToken::Star) && 1740 getTargetParser().starIsStartOfStatement()) { 1741 // Accept '*' as a valid start of statement. 1742 Lex(); 1743 IDVal = "*"; 1744 } else if (parseIdentifier(IDVal)) { 1745 if (!TheCondState.Ignore) { 1746 Lex(); // always eat a token 1747 return Error(IDLoc, "unexpected token at start of statement"); 1748 } 1749 IDVal = ""; 1750 } 1751 1752 // Handle conditional assembly here before checking for skipping. We 1753 // have to do this so that .endif isn't skipped in a ".if 0" block for 1754 // example. 1755 StringMap<DirectiveKind>::const_iterator DirKindIt = 1756 DirectiveKindMap.find(IDVal); 1757 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) 1758 1759 ? DK_NO_DIRECTIVE 1760 : DirKindIt->getValue(); 1761 switch (DirKind) { 1762 default: 1763 break; 1764 case DK_IF: 1765 case DK_IFEQ: 1766 case DK_IFGE: 1767 case DK_IFGT: 1768 case DK_IFLE: 1769 case DK_IFLT: 1770 case DK_IFNE: 1771 return parseDirectiveIf(IDLoc, DirKind); 1772 case DK_IFB: 1773 return parseDirectiveIfb(IDLoc, true); 1774 case DK_IFNB: 1775 return parseDirectiveIfb(IDLoc, false); 1776 case DK_IFC: 1777 return parseDirectiveIfc(IDLoc, true); 1778 case DK_IFEQS: 1779 return parseDirectiveIfeqs(IDLoc, true); 1780 case DK_IFNC: 1781 return parseDirectiveIfc(IDLoc, false); 1782 case DK_IFNES: 1783 return parseDirectiveIfeqs(IDLoc, false); 1784 case DK_IFDEF: 1785 return parseDirectiveIfdef(IDLoc, true); 1786 case DK_IFNDEF: 1787 case DK_IFNOTDEF: 1788 return parseDirectiveIfdef(IDLoc, false); 1789 case DK_ELSEIF: 1790 return parseDirectiveElseIf(IDLoc); 1791 case DK_ELSE: 1792 return parseDirectiveElse(IDLoc); 1793 case DK_ENDIF: 1794 return parseDirectiveEndIf(IDLoc); 1795 } 1796 1797 // Ignore the statement if in the middle of inactive conditional 1798 // (e.g. ".if 0"). 1799 if (TheCondState.Ignore) { 1800 eatToEndOfStatement(); 1801 return false; 1802 } 1803 1804 // FIXME: Recurse on local labels? 1805 1806 // See what kind of statement we have. 1807 switch (Lexer.getKind()) { 1808 case AsmToken::Colon: { 1809 if (!getTargetParser().isLabel(ID)) 1810 break; 1811 if (checkForValidSection()) 1812 return true; 1813 1814 // identifier ':' -> Label. 1815 Lex(); 1816 1817 // Diagnose attempt to use '.' as a label. 1818 if (IDVal == ".") 1819 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1820 1821 // Diagnose attempt to use a variable as a label. 1822 // 1823 // FIXME: Diagnostics. Note the location of the definition as a label. 1824 // FIXME: This doesn't diagnose assignment to a symbol which has been 1825 // implicitly marked as external. 1826 MCSymbol *Sym; 1827 if (LocalLabelVal == -1) { 1828 if (ParsingInlineAsm && SI) { 1829 StringRef RewrittenLabel = 1830 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true); 1831 assert(!RewrittenLabel.empty() && 1832 "We should have an internal name here."); 1833 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(), 1834 RewrittenLabel); 1835 IDVal = RewrittenLabel; 1836 } 1837 Sym = getContext().getOrCreateSymbol(IDVal); 1838 } else 1839 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal); 1840 // End of Labels should be treated as end of line for lexing 1841 // purposes but that information is not available to the Lexer who 1842 // does not understand Labels. This may cause us to see a Hash 1843 // here instead of a preprocessor line comment. 1844 if (getTok().is(AsmToken::Hash)) { 1845 StringRef CommentStr = parseStringToEndOfStatement(); 1846 Lexer.Lex(); 1847 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr)); 1848 } 1849 1850 // Consume any end of statement token, if present, to avoid spurious 1851 // AddBlankLine calls(). 1852 if (getTok().is(AsmToken::EndOfStatement)) { 1853 Lex(); 1854 } 1855 1856 getTargetParser().doBeforeLabelEmit(Sym); 1857 1858 // Emit the label. 1859 if (!getTargetParser().isParsingInlineAsm()) 1860 Out.EmitLabel(Sym, IDLoc); 1861 1862 // If we are generating dwarf for assembly source files then gather the 1863 // info to make a dwarf label entry for this label if needed. 1864 if (enabledGenDwarfForAssembly()) 1865 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 1866 IDLoc); 1867 1868 getTargetParser().onLabelParsed(Sym); 1869 1870 return false; 1871 } 1872 1873 case AsmToken::Equal: 1874 if (!getTargetParser().equalIsAsmAssignment()) 1875 break; 1876 // identifier '=' ... -> assignment statement 1877 Lex(); 1878 1879 return parseAssignment(IDVal, true); 1880 1881 default: // Normal instruction or directive. 1882 break; 1883 } 1884 1885 // If macros are enabled, check to see if this is a macro instantiation. 1886 if (areMacrosEnabled()) 1887 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) { 1888 return handleMacroEntry(M, IDLoc); 1889 } 1890 1891 // Otherwise, we have a normal instruction or directive. 1892 1893 // Directives start with "." 1894 if (IDVal.startswith(".") && IDVal != ".") { 1895 // There are several entities interested in parsing directives: 1896 // 1897 // 1. The target-specific assembly parser. Some directives are target 1898 // specific or may potentially behave differently on certain targets. 1899 // 2. Asm parser extensions. For example, platform-specific parsers 1900 // (like the ELF parser) register themselves as extensions. 1901 // 3. The generic directive parser implemented by this class. These are 1902 // all the directives that behave in a target and platform independent 1903 // manner, or at least have a default behavior that's shared between 1904 // all targets and platforms. 1905 1906 getTargetParser().flushPendingInstructions(getStreamer()); 1907 1908 SMLoc StartTokLoc = getTok().getLoc(); 1909 bool TPDirectiveReturn = getTargetParser().ParseDirective(ID); 1910 1911 if (hasPendingError()) 1912 return true; 1913 // Currently the return value should be true if we are 1914 // uninterested but as this is at odds with the standard parsing 1915 // convention (return true = error) we have instances of a parsed 1916 // directive that fails returning true as an error. Catch these 1917 // cases as best as possible errors here. 1918 if (TPDirectiveReturn && StartTokLoc != getTok().getLoc()) 1919 return true; 1920 // Return if we did some parsing or believe we succeeded. 1921 if (!TPDirectiveReturn || StartTokLoc != getTok().getLoc()) 1922 return false; 1923 1924 // Next, check the extension directive map to see if any extension has 1925 // registered itself to parse this directive. 1926 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 1927 ExtensionDirectiveMap.lookup(IDVal); 1928 if (Handler.first) 1929 return (*Handler.second)(Handler.first, IDVal, IDLoc); 1930 1931 // Finally, if no one else is interested in this directive, it must be 1932 // generic and familiar to this class. 1933 switch (DirKind) { 1934 default: 1935 break; 1936 case DK_SET: 1937 case DK_EQU: 1938 return parseDirectiveSet(IDVal, true); 1939 case DK_EQUIV: 1940 return parseDirectiveSet(IDVal, false); 1941 case DK_ASCII: 1942 return parseDirectiveAscii(IDVal, false); 1943 case DK_ASCIZ: 1944 case DK_STRING: 1945 return parseDirectiveAscii(IDVal, true); 1946 case DK_BYTE: 1947 case DK_DC_B: 1948 return parseDirectiveValue(IDVal, 1); 1949 case DK_DC: 1950 case DK_DC_W: 1951 case DK_SHORT: 1952 case DK_VALUE: 1953 case DK_2BYTE: 1954 return parseDirectiveValue(IDVal, 2); 1955 case DK_LONG: 1956 case DK_INT: 1957 case DK_4BYTE: 1958 case DK_DC_L: 1959 return parseDirectiveValue(IDVal, 4); 1960 case DK_QUAD: 1961 case DK_8BYTE: 1962 return parseDirectiveValue(IDVal, 8); 1963 case DK_DC_A: 1964 return parseDirectiveValue( 1965 IDVal, getContext().getAsmInfo()->getCodePointerSize()); 1966 case DK_OCTA: 1967 return parseDirectiveOctaValue(IDVal); 1968 case DK_SINGLE: 1969 case DK_FLOAT: 1970 case DK_DC_S: 1971 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle()); 1972 case DK_DOUBLE: 1973 case DK_DC_D: 1974 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble()); 1975 case DK_ALIGN: { 1976 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 1977 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1); 1978 } 1979 case DK_ALIGN32: { 1980 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 1981 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4); 1982 } 1983 case DK_BALIGN: 1984 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 1985 case DK_BALIGNW: 1986 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 1987 case DK_BALIGNL: 1988 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 1989 case DK_P2ALIGN: 1990 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 1991 case DK_P2ALIGNW: 1992 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 1993 case DK_P2ALIGNL: 1994 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 1995 case DK_ORG: 1996 return parseDirectiveOrg(); 1997 case DK_FILL: 1998 return parseDirectiveFill(); 1999 case DK_ZERO: 2000 return parseDirectiveZero(); 2001 case DK_EXTERN: 2002 eatToEndOfStatement(); // .extern is the default, ignore it. 2003 return false; 2004 case DK_GLOBL: 2005 case DK_GLOBAL: 2006 return parseDirectiveSymbolAttribute(MCSA_Global); 2007 case DK_LAZY_REFERENCE: 2008 return parseDirectiveSymbolAttribute(MCSA_LazyReference); 2009 case DK_NO_DEAD_STRIP: 2010 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 2011 case DK_SYMBOL_RESOLVER: 2012 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver); 2013 case DK_PRIVATE_EXTERN: 2014 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern); 2015 case DK_REFERENCE: 2016 return parseDirectiveSymbolAttribute(MCSA_Reference); 2017 case DK_WEAK_DEFINITION: 2018 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition); 2019 case DK_WEAK_REFERENCE: 2020 return parseDirectiveSymbolAttribute(MCSA_WeakReference); 2021 case DK_WEAK_DEF_CAN_BE_HIDDEN: 2022 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 2023 case DK_COLD: 2024 return parseDirectiveSymbolAttribute(MCSA_Cold); 2025 case DK_COMM: 2026 case DK_COMMON: 2027 return parseDirectiveComm(/*IsLocal=*/false); 2028 case DK_LCOMM: 2029 return parseDirectiveComm(/*IsLocal=*/true); 2030 case DK_ABORT: 2031 return parseDirectiveAbort(); 2032 case DK_INCLUDE: 2033 return parseDirectiveInclude(); 2034 case DK_INCBIN: 2035 return parseDirectiveIncbin(); 2036 case DK_CODE16: 2037 case DK_CODE16GCC: 2038 return TokError(Twine(IDVal) + 2039 " not currently supported for this target"); 2040 case DK_REPT: 2041 return parseDirectiveRept(IDLoc, IDVal); 2042 case DK_IRP: 2043 return parseDirectiveIrp(IDLoc); 2044 case DK_IRPC: 2045 return parseDirectiveIrpc(IDLoc); 2046 case DK_ENDR: 2047 return parseDirectiveEndr(IDLoc); 2048 case DK_BUNDLE_ALIGN_MODE: 2049 return parseDirectiveBundleAlignMode(); 2050 case DK_BUNDLE_LOCK: 2051 return parseDirectiveBundleLock(); 2052 case DK_BUNDLE_UNLOCK: 2053 return parseDirectiveBundleUnlock(); 2054 case DK_SLEB128: 2055 return parseDirectiveLEB128(true); 2056 case DK_ULEB128: 2057 return parseDirectiveLEB128(false); 2058 case DK_SPACE: 2059 case DK_SKIP: 2060 return parseDirectiveSpace(IDVal); 2061 case DK_FILE: 2062 return parseDirectiveFile(IDLoc); 2063 case DK_LINE: 2064 return parseDirectiveLine(); 2065 case DK_LOC: 2066 return parseDirectiveLoc(); 2067 case DK_STABS: 2068 return parseDirectiveStabs(); 2069 case DK_CV_FILE: 2070 return parseDirectiveCVFile(); 2071 case DK_CV_FUNC_ID: 2072 return parseDirectiveCVFuncId(); 2073 case DK_CV_INLINE_SITE_ID: 2074 return parseDirectiveCVInlineSiteId(); 2075 case DK_CV_LOC: 2076 return parseDirectiveCVLoc(); 2077 case DK_CV_LINETABLE: 2078 return parseDirectiveCVLinetable(); 2079 case DK_CV_INLINE_LINETABLE: 2080 return parseDirectiveCVInlineLinetable(); 2081 case DK_CV_DEF_RANGE: 2082 return parseDirectiveCVDefRange(); 2083 case DK_CV_STRING: 2084 return parseDirectiveCVString(); 2085 case DK_CV_STRINGTABLE: 2086 return parseDirectiveCVStringTable(); 2087 case DK_CV_FILECHECKSUMS: 2088 return parseDirectiveCVFileChecksums(); 2089 case DK_CV_FILECHECKSUM_OFFSET: 2090 return parseDirectiveCVFileChecksumOffset(); 2091 case DK_CV_FPO_DATA: 2092 return parseDirectiveCVFPOData(); 2093 case DK_CFI_SECTIONS: 2094 return parseDirectiveCFISections(); 2095 case DK_CFI_STARTPROC: 2096 return parseDirectiveCFIStartProc(); 2097 case DK_CFI_ENDPROC: 2098 return parseDirectiveCFIEndProc(); 2099 case DK_CFI_DEF_CFA: 2100 return parseDirectiveCFIDefCfa(IDLoc); 2101 case DK_CFI_DEF_CFA_OFFSET: 2102 return parseDirectiveCFIDefCfaOffset(); 2103 case DK_CFI_ADJUST_CFA_OFFSET: 2104 return parseDirectiveCFIAdjustCfaOffset(); 2105 case DK_CFI_DEF_CFA_REGISTER: 2106 return parseDirectiveCFIDefCfaRegister(IDLoc); 2107 case DK_CFI_OFFSET: 2108 return parseDirectiveCFIOffset(IDLoc); 2109 case DK_CFI_REL_OFFSET: 2110 return parseDirectiveCFIRelOffset(IDLoc); 2111 case DK_CFI_PERSONALITY: 2112 return parseDirectiveCFIPersonalityOrLsda(true); 2113 case DK_CFI_LSDA: 2114 return parseDirectiveCFIPersonalityOrLsda(false); 2115 case DK_CFI_REMEMBER_STATE: 2116 return parseDirectiveCFIRememberState(); 2117 case DK_CFI_RESTORE_STATE: 2118 return parseDirectiveCFIRestoreState(); 2119 case DK_CFI_SAME_VALUE: 2120 return parseDirectiveCFISameValue(IDLoc); 2121 case DK_CFI_RESTORE: 2122 return parseDirectiveCFIRestore(IDLoc); 2123 case DK_CFI_ESCAPE: 2124 return parseDirectiveCFIEscape(); 2125 case DK_CFI_RETURN_COLUMN: 2126 return parseDirectiveCFIReturnColumn(IDLoc); 2127 case DK_CFI_SIGNAL_FRAME: 2128 return parseDirectiveCFISignalFrame(); 2129 case DK_CFI_UNDEFINED: 2130 return parseDirectiveCFIUndefined(IDLoc); 2131 case DK_CFI_REGISTER: 2132 return parseDirectiveCFIRegister(IDLoc); 2133 case DK_CFI_WINDOW_SAVE: 2134 return parseDirectiveCFIWindowSave(); 2135 case DK_MACROS_ON: 2136 case DK_MACROS_OFF: 2137 return parseDirectiveMacrosOnOff(IDVal); 2138 case DK_MACRO: 2139 return parseDirectiveMacro(IDLoc); 2140 case DK_ALTMACRO: 2141 case DK_NOALTMACRO: 2142 return parseDirectiveAltmacro(IDVal); 2143 case DK_EXITM: 2144 return parseDirectiveExitMacro(IDVal); 2145 case DK_ENDM: 2146 case DK_ENDMACRO: 2147 return parseDirectiveEndMacro(IDVal); 2148 case DK_PURGEM: 2149 return parseDirectivePurgeMacro(IDLoc); 2150 case DK_END: 2151 return parseDirectiveEnd(IDLoc); 2152 case DK_ERR: 2153 return parseDirectiveError(IDLoc, false); 2154 case DK_ERROR: 2155 return parseDirectiveError(IDLoc, true); 2156 case DK_WARNING: 2157 return parseDirectiveWarning(IDLoc); 2158 case DK_RELOC: 2159 return parseDirectiveReloc(IDLoc); 2160 case DK_DCB: 2161 case DK_DCB_W: 2162 return parseDirectiveDCB(IDVal, 2); 2163 case DK_DCB_B: 2164 return parseDirectiveDCB(IDVal, 1); 2165 case DK_DCB_D: 2166 return parseDirectiveRealDCB(IDVal, APFloat::IEEEdouble()); 2167 case DK_DCB_L: 2168 return parseDirectiveDCB(IDVal, 4); 2169 case DK_DCB_S: 2170 return parseDirectiveRealDCB(IDVal, APFloat::IEEEsingle()); 2171 case DK_DC_X: 2172 case DK_DCB_X: 2173 return TokError(Twine(IDVal) + 2174 " not currently supported for this target"); 2175 case DK_DS: 2176 case DK_DS_W: 2177 return parseDirectiveDS(IDVal, 2); 2178 case DK_DS_B: 2179 return parseDirectiveDS(IDVal, 1); 2180 case DK_DS_D: 2181 return parseDirectiveDS(IDVal, 8); 2182 case DK_DS_L: 2183 case DK_DS_S: 2184 return parseDirectiveDS(IDVal, 4); 2185 case DK_DS_P: 2186 case DK_DS_X: 2187 return parseDirectiveDS(IDVal, 12); 2188 case DK_PRINT: 2189 return parseDirectivePrint(IDLoc); 2190 case DK_ADDRSIG: 2191 return parseDirectiveAddrsig(); 2192 case DK_ADDRSIG_SYM: 2193 return parseDirectiveAddrsigSym(); 2194 } 2195 2196 return Error(IDLoc, "unknown directive"); 2197 } 2198 2199 // __asm _emit or __asm __emit 2200 if (ParsingInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || 2201 IDVal == "_EMIT" || IDVal == "__EMIT")) 2202 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size()); 2203 2204 // __asm align 2205 if (ParsingInlineAsm && (IDVal == "align" || IDVal == "ALIGN")) 2206 return parseDirectiveMSAlign(IDLoc, Info); 2207 2208 if (ParsingInlineAsm && (IDVal == "even" || IDVal == "EVEN")) 2209 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4); 2210 if (checkForValidSection()) 2211 return true; 2212 2213 // Canonicalize the opcode to lower case. 2214 std::string OpcodeStr = IDVal.lower(); 2215 ParseInstructionInfo IInfo(Info.AsmRewrites); 2216 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID, 2217 Info.ParsedOperands); 2218 Info.ParseError = ParseHadError; 2219 2220 // Dump the parsed representation, if requested. 2221 if (getShowParsedOperands()) { 2222 SmallString<256> Str; 2223 raw_svector_ostream OS(Str); 2224 OS << "parsed instruction: ["; 2225 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { 2226 if (i != 0) 2227 OS << ", "; 2228 Info.ParsedOperands[i]->print(OS); 2229 } 2230 OS << "]"; 2231 2232 printMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 2233 } 2234 2235 // Fail even if ParseInstruction erroneously returns false. 2236 if (hasPendingError() || ParseHadError) 2237 return true; 2238 2239 // If we are generating dwarf for the current section then generate a .loc 2240 // directive for the instruction. 2241 if (!ParseHadError && enabledGenDwarfForAssembly() && 2242 getContext().getGenDwarfSectionSyms().count( 2243 getStreamer().getCurrentSectionOnly())) { 2244 unsigned Line; 2245 if (ActiveMacros.empty()) 2246 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer); 2247 else 2248 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc, 2249 ActiveMacros.front()->ExitBuffer); 2250 2251 // If we previously parsed a cpp hash file line comment then make sure the 2252 // current Dwarf File is for the CppHashFilename if not then emit the 2253 // Dwarf File table for it and adjust the line number for the .loc. 2254 if (!CppHashInfo.Filename.empty()) { 2255 unsigned FileNumber = getStreamer().EmitDwarfFileDirective( 2256 0, StringRef(), CppHashInfo.Filename); 2257 getContext().setGenDwarfFileNumber(FileNumber); 2258 2259 unsigned CppHashLocLineNo = 2260 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf); 2261 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); 2262 } 2263 2264 getStreamer().EmitDwarfLocDirective( 2265 getContext().getGenDwarfFileNumber(), Line, 0, 2266 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0, 2267 StringRef()); 2268 } 2269 2270 // If parsing succeeded, match the instruction. 2271 if (!ParseHadError) { 2272 uint64_t ErrorInfo; 2273 if (getTargetParser().MatchAndEmitInstruction( 2274 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo, 2275 getTargetParser().isParsingInlineAsm())) 2276 return true; 2277 } 2278 return false; 2279 } 2280 2281 // Parse and erase curly braces marking block start/end 2282 bool 2283 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { 2284 // Identify curly brace marking block start/end 2285 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly)) 2286 return false; 2287 2288 SMLoc StartLoc = Lexer.getLoc(); 2289 Lex(); // Eat the brace 2290 if (Lexer.is(AsmToken::EndOfStatement)) 2291 Lex(); // Eat EndOfStatement following the brace 2292 2293 // Erase the block start/end brace from the output asm string 2294 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() - 2295 StartLoc.getPointer()); 2296 return true; 2297 } 2298 2299 /// parseCppHashLineFilenameComment as this: 2300 /// ::= # number "filename" 2301 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L) { 2302 Lex(); // Eat the hash token. 2303 // Lexer only ever emits HashDirective if it fully formed if it's 2304 // done the checking already so this is an internal error. 2305 assert(getTok().is(AsmToken::Integer) && 2306 "Lexing Cpp line comment: Expected Integer"); 2307 int64_t LineNumber = getTok().getIntVal(); 2308 Lex(); 2309 assert(getTok().is(AsmToken::String) && 2310 "Lexing Cpp line comment: Expected String"); 2311 StringRef Filename = getTok().getString(); 2312 Lex(); 2313 2314 // Get rid of the enclosing quotes. 2315 Filename = Filename.substr(1, Filename.size() - 2); 2316 2317 // Save the SMLoc, Filename and LineNumber for later use by diagnostics 2318 // and possibly DWARF file info. 2319 CppHashInfo.Loc = L; 2320 CppHashInfo.Filename = Filename; 2321 CppHashInfo.LineNumber = LineNumber; 2322 CppHashInfo.Buf = CurBuffer; 2323 if (FirstCppHashFilename.empty()) 2324 FirstCppHashFilename = Filename; 2325 return false; 2326 } 2327 2328 /// will use the last parsed cpp hash line filename comment 2329 /// for the Filename and LineNo if any in the diagnostic. 2330 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 2331 const AsmParser *Parser = static_cast<const AsmParser *>(Context); 2332 raw_ostream &OS = errs(); 2333 2334 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 2335 SMLoc DiagLoc = Diag.getLoc(); 2336 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2337 unsigned CppHashBuf = 2338 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc); 2339 2340 // Like SourceMgr::printMessage() we need to print the include stack if any 2341 // before printing the message. 2342 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2343 if (!Parser->SavedDiagHandler && DiagCurBuffer && 2344 DiagCurBuffer != DiagSrcMgr.getMainFileID()) { 2345 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 2346 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 2347 } 2348 2349 // If we have not parsed a cpp hash line filename comment or the source 2350 // manager changed or buffer changed (like in a nested include) then just 2351 // print the normal diagnostic using its Filename and LineNo. 2352 if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr || 2353 DiagBuf != CppHashBuf) { 2354 if (Parser->SavedDiagHandler) 2355 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2356 else 2357 Diag.print(nullptr, OS); 2358 return; 2359 } 2360 2361 // Use the CppHashFilename and calculate a line number based on the 2362 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc 2363 // for the diagnostic. 2364 const std::string &Filename = Parser->CppHashInfo.Filename; 2365 2366 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 2367 int CppHashLocLineNo = 2368 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf); 2369 int LineNo = 2370 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); 2371 2372 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, 2373 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), 2374 Diag.getLineContents(), Diag.getRanges()); 2375 2376 if (Parser->SavedDiagHandler) 2377 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); 2378 else 2379 NewDiag.print(nullptr, OS); 2380 } 2381 2382 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The 2383 // difference being that that function accepts '@' as part of identifiers and 2384 // we can't do that. AsmLexer.cpp should probably be changed to handle 2385 // '@' as a special case when needed. 2386 static bool isIdentifierChar(char c) { 2387 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' || 2388 c == '.'; 2389 } 2390 2391 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 2392 ArrayRef<MCAsmMacroParameter> Parameters, 2393 ArrayRef<MCAsmMacroArgument> A, 2394 bool EnableAtPseudoVariable, SMLoc L) { 2395 unsigned NParameters = Parameters.size(); 2396 bool HasVararg = NParameters ? Parameters.back().Vararg : false; 2397 if ((!IsDarwin || NParameters != 0) && NParameters != A.size()) 2398 return Error(L, "Wrong number of arguments"); 2399 2400 // A macro without parameters is handled differently on Darwin: 2401 // gas accepts no arguments and does no substitutions 2402 while (!Body.empty()) { 2403 // Scan for the next substitution. 2404 std::size_t End = Body.size(), Pos = 0; 2405 for (; Pos != End; ++Pos) { 2406 // Check for a substitution or escape. 2407 if (IsDarwin && !NParameters) { 2408 // This macro has no parameters, look for $0, $1, etc. 2409 if (Body[Pos] != '$' || Pos + 1 == End) 2410 continue; 2411 2412 char Next = Body[Pos + 1]; 2413 if (Next == '$' || Next == 'n' || 2414 isdigit(static_cast<unsigned char>(Next))) 2415 break; 2416 } else { 2417 // This macro has parameters, look for \foo, \bar, etc. 2418 if (Body[Pos] == '\\' && Pos + 1 != End) 2419 break; 2420 } 2421 } 2422 2423 // Add the prefix. 2424 OS << Body.slice(0, Pos); 2425 2426 // Check if we reached the end. 2427 if (Pos == End) 2428 break; 2429 2430 if (IsDarwin && !NParameters) { 2431 switch (Body[Pos + 1]) { 2432 // $$ => $ 2433 case '$': 2434 OS << '$'; 2435 break; 2436 2437 // $n => number of arguments 2438 case 'n': 2439 OS << A.size(); 2440 break; 2441 2442 // $[0-9] => argument 2443 default: { 2444 // Missing arguments are ignored. 2445 unsigned Index = Body[Pos + 1] - '0'; 2446 if (Index >= A.size()) 2447 break; 2448 2449 // Otherwise substitute with the token values, with spaces eliminated. 2450 for (const AsmToken &Token : A[Index]) 2451 OS << Token.getString(); 2452 break; 2453 } 2454 } 2455 Pos += 2; 2456 } else { 2457 unsigned I = Pos + 1; 2458 2459 // Check for the \@ pseudo-variable. 2460 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End) 2461 ++I; 2462 else 2463 while (isIdentifierChar(Body[I]) && I + 1 != End) 2464 ++I; 2465 2466 const char *Begin = Body.data() + Pos + 1; 2467 StringRef Argument(Begin, I - (Pos + 1)); 2468 unsigned Index = 0; 2469 2470 if (Argument == "@") { 2471 OS << NumOfMacroInstantiations; 2472 Pos += 2; 2473 } else { 2474 for (; Index < NParameters; ++Index) 2475 if (Parameters[Index].Name == Argument) 2476 break; 2477 2478 if (Index == NParameters) { 2479 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 2480 Pos += 3; 2481 else { 2482 OS << '\\' << Argument; 2483 Pos = I; 2484 } 2485 } else { 2486 bool VarargParameter = HasVararg && Index == (NParameters - 1); 2487 for (const AsmToken &Token : A[Index]) 2488 // For altmacro mode, you can write '%expr'. 2489 // The prefix '%' evaluates the expression 'expr' 2490 // and uses the result as a string (e.g. replace %(1+2) with the 2491 // string "3"). 2492 // Here, we identify the integer token which is the result of the 2493 // absolute expression evaluation and replace it with its string 2494 // representation. 2495 if (AltMacroMode && Token.getString().front() == '%' && 2496 Token.is(AsmToken::Integer)) 2497 // Emit an integer value to the buffer. 2498 OS << Token.getIntVal(); 2499 // Only Token that was validated as a string and begins with '<' 2500 // is considered altMacroString!!! 2501 else if (AltMacroMode && Token.getString().front() == '<' && 2502 Token.is(AsmToken::String)) { 2503 OS << altMacroString(Token.getStringContents()); 2504 } 2505 // We expect no quotes around the string's contents when 2506 // parsing for varargs. 2507 else if (Token.isNot(AsmToken::String) || VarargParameter) 2508 OS << Token.getString(); 2509 else 2510 OS << Token.getStringContents(); 2511 2512 Pos += 1 + Argument.size(); 2513 } 2514 } 2515 } 2516 // Update the scan point. 2517 Body = Body.substr(Pos); 2518 } 2519 2520 return false; 2521 } 2522 2523 MacroInstantiation::MacroInstantiation(SMLoc IL, int EB, SMLoc EL, 2524 size_t CondStackDepth) 2525 : InstantiationLoc(IL), ExitBuffer(EB), ExitLoc(EL), 2526 CondStackDepth(CondStackDepth) {} 2527 2528 static bool isOperator(AsmToken::TokenKind kind) { 2529 switch (kind) { 2530 default: 2531 return false; 2532 case AsmToken::Plus: 2533 case AsmToken::Minus: 2534 case AsmToken::Tilde: 2535 case AsmToken::Slash: 2536 case AsmToken::Star: 2537 case AsmToken::Dot: 2538 case AsmToken::Equal: 2539 case AsmToken::EqualEqual: 2540 case AsmToken::Pipe: 2541 case AsmToken::PipePipe: 2542 case AsmToken::Caret: 2543 case AsmToken::Amp: 2544 case AsmToken::AmpAmp: 2545 case AsmToken::Exclaim: 2546 case AsmToken::ExclaimEqual: 2547 case AsmToken::Less: 2548 case AsmToken::LessEqual: 2549 case AsmToken::LessLess: 2550 case AsmToken::LessGreater: 2551 case AsmToken::Greater: 2552 case AsmToken::GreaterEqual: 2553 case AsmToken::GreaterGreater: 2554 return true; 2555 } 2556 } 2557 2558 namespace { 2559 2560 class AsmLexerSkipSpaceRAII { 2561 public: 2562 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { 2563 Lexer.setSkipSpace(SkipSpace); 2564 } 2565 2566 ~AsmLexerSkipSpaceRAII() { 2567 Lexer.setSkipSpace(true); 2568 } 2569 2570 private: 2571 AsmLexer &Lexer; 2572 }; 2573 2574 } // end anonymous namespace 2575 2576 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) { 2577 2578 if (Vararg) { 2579 if (Lexer.isNot(AsmToken::EndOfStatement)) { 2580 StringRef Str = parseStringToEndOfStatement(); 2581 MA.emplace_back(AsmToken::String, Str); 2582 } 2583 return false; 2584 } 2585 2586 unsigned ParenLevel = 0; 2587 2588 // Darwin doesn't use spaces to delmit arguments. 2589 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); 2590 2591 bool SpaceEaten; 2592 2593 while (true) { 2594 SpaceEaten = false; 2595 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 2596 return TokError("unexpected token in macro instantiation"); 2597 2598 if (ParenLevel == 0) { 2599 2600 if (Lexer.is(AsmToken::Comma)) 2601 break; 2602 2603 if (Lexer.is(AsmToken::Space)) { 2604 SpaceEaten = true; 2605 Lexer.Lex(); // Eat spaces 2606 } 2607 2608 // Spaces can delimit parameters, but could also be part an expression. 2609 // If the token after a space is an operator, add the token and the next 2610 // one into this argument 2611 if (!IsDarwin) { 2612 if (isOperator(Lexer.getKind())) { 2613 MA.push_back(getTok()); 2614 Lexer.Lex(); 2615 2616 // Whitespace after an operator can be ignored. 2617 if (Lexer.is(AsmToken::Space)) 2618 Lexer.Lex(); 2619 2620 continue; 2621 } 2622 } 2623 if (SpaceEaten) 2624 break; 2625 } 2626 2627 // handleMacroEntry relies on not advancing the lexer here 2628 // to be able to fill in the remaining default parameter values 2629 if (Lexer.is(AsmToken::EndOfStatement)) 2630 break; 2631 2632 // Adjust the current parentheses level. 2633 if (Lexer.is(AsmToken::LParen)) 2634 ++ParenLevel; 2635 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 2636 --ParenLevel; 2637 2638 // Append the token to the current argument list. 2639 MA.push_back(getTok()); 2640 Lexer.Lex(); 2641 } 2642 2643 if (ParenLevel != 0) 2644 return TokError("unbalanced parentheses in macro argument"); 2645 return false; 2646 } 2647 2648 // Parse the macro instantiation arguments. 2649 bool AsmParser::parseMacroArguments(const MCAsmMacro *M, 2650 MCAsmMacroArguments &A) { 2651 const unsigned NParameters = M ? M->Parameters.size() : 0; 2652 bool NamedParametersFound = false; 2653 SmallVector<SMLoc, 4> FALocs; 2654 2655 A.resize(NParameters); 2656 FALocs.resize(NParameters); 2657 2658 // Parse two kinds of macro invocations: 2659 // - macros defined without any parameters accept an arbitrary number of them 2660 // - macros defined with parameters accept at most that many of them 2661 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false; 2662 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 2663 ++Parameter) { 2664 SMLoc IDLoc = Lexer.getLoc(); 2665 MCAsmMacroParameter FA; 2666 2667 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) { 2668 if (parseIdentifier(FA.Name)) 2669 return Error(IDLoc, "invalid argument identifier for formal argument"); 2670 2671 if (Lexer.isNot(AsmToken::Equal)) 2672 return TokError("expected '=' after formal parameter identifier"); 2673 2674 Lex(); 2675 2676 NamedParametersFound = true; 2677 } 2678 bool Vararg = HasVararg && Parameter == (NParameters - 1); 2679 2680 if (NamedParametersFound && FA.Name.empty()) 2681 return Error(IDLoc, "cannot mix positional and keyword arguments"); 2682 2683 SMLoc StrLoc = Lexer.getLoc(); 2684 SMLoc EndLoc; 2685 if (AltMacroMode && Lexer.is(AsmToken::Percent)) { 2686 const MCExpr *AbsoluteExp; 2687 int64_t Value; 2688 /// Eat '%' 2689 Lex(); 2690 if (parseExpression(AbsoluteExp, EndLoc)) 2691 return false; 2692 if (!AbsoluteExp->evaluateAsAbsolute(Value, 2693 getStreamer().getAssemblerPtr())) 2694 return Error(StrLoc, "expected absolute expression"); 2695 const char *StrChar = StrLoc.getPointer(); 2696 const char *EndChar = EndLoc.getPointer(); 2697 AsmToken newToken(AsmToken::Integer, 2698 StringRef(StrChar, EndChar - StrChar), Value); 2699 FA.Value.push_back(newToken); 2700 } else if (AltMacroMode && Lexer.is(AsmToken::Less) && 2701 isAltmacroString(StrLoc, EndLoc)) { 2702 const char *StrChar = StrLoc.getPointer(); 2703 const char *EndChar = EndLoc.getPointer(); 2704 jumpToLoc(EndLoc, CurBuffer); 2705 /// Eat from '<' to '>' 2706 Lex(); 2707 AsmToken newToken(AsmToken::String, 2708 StringRef(StrChar, EndChar - StrChar)); 2709 FA.Value.push_back(newToken); 2710 } else if(parseMacroArgument(FA.Value, Vararg)) 2711 return true; 2712 2713 unsigned PI = Parameter; 2714 if (!FA.Name.empty()) { 2715 unsigned FAI = 0; 2716 for (FAI = 0; FAI < NParameters; ++FAI) 2717 if (M->Parameters[FAI].Name == FA.Name) 2718 break; 2719 2720 if (FAI >= NParameters) { 2721 assert(M && "expected macro to be defined"); 2722 return Error(IDLoc, "parameter named '" + FA.Name + 2723 "' does not exist for macro '" + M->Name + "'"); 2724 } 2725 PI = FAI; 2726 } 2727 2728 if (!FA.Value.empty()) { 2729 if (A.size() <= PI) 2730 A.resize(PI + 1); 2731 A[PI] = FA.Value; 2732 2733 if (FALocs.size() <= PI) 2734 FALocs.resize(PI + 1); 2735 2736 FALocs[PI] = Lexer.getLoc(); 2737 } 2738 2739 // At the end of the statement, fill in remaining arguments that have 2740 // default values. If there aren't any, then the next argument is 2741 // required but missing 2742 if (Lexer.is(AsmToken::EndOfStatement)) { 2743 bool Failure = false; 2744 for (unsigned FAI = 0; FAI < NParameters; ++FAI) { 2745 if (A[FAI].empty()) { 2746 if (M->Parameters[FAI].Required) { 2747 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), 2748 "missing value for required parameter " 2749 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'"); 2750 Failure = true; 2751 } 2752 2753 if (!M->Parameters[FAI].Value.empty()) 2754 A[FAI] = M->Parameters[FAI].Value; 2755 } 2756 } 2757 return Failure; 2758 } 2759 2760 if (Lexer.is(AsmToken::Comma)) 2761 Lex(); 2762 } 2763 2764 return TokError("too many positional arguments"); 2765 } 2766 2767 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) { 2768 // Arbitrarily limit macro nesting depth (default matches 'as'). We can 2769 // eliminate this, although we should protect against infinite loops. 2770 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; 2771 if (ActiveMacros.size() == MaxNestingDepth) { 2772 std::ostringstream MaxNestingDepthError; 2773 MaxNestingDepthError << "macros cannot be nested more than " 2774 << MaxNestingDepth << " levels deep." 2775 << " Use -asm-macro-max-nesting-depth to increase " 2776 "this limit."; 2777 return TokError(MaxNestingDepthError.str()); 2778 } 2779 2780 MCAsmMacroArguments A; 2781 if (parseMacroArguments(M, A)) 2782 return true; 2783 2784 // Macro instantiation is lexical, unfortunately. We construct a new buffer 2785 // to hold the macro body with substitutions. 2786 SmallString<256> Buf; 2787 StringRef Body = M->Body; 2788 raw_svector_ostream OS(Buf); 2789 2790 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc())) 2791 return true; 2792 2793 // We include the .endmacro in the buffer as our cue to exit the macro 2794 // instantiation. 2795 OS << ".endmacro\n"; 2796 2797 std::unique_ptr<MemoryBuffer> Instantiation = 2798 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 2799 2800 // Create the macro instantiation object and add to the current macro 2801 // instantiation stack. 2802 MacroInstantiation *MI = new MacroInstantiation( 2803 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()); 2804 ActiveMacros.push_back(MI); 2805 2806 ++NumOfMacroInstantiations; 2807 2808 // Jump to the macro instantiation and prime the lexer. 2809 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 2810 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 2811 Lex(); 2812 2813 return false; 2814 } 2815 2816 void AsmParser::handleMacroExit() { 2817 // Jump to the EndOfStatement we should return to, and consume it. 2818 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer); 2819 Lex(); 2820 2821 // Pop the instantiation entry. 2822 delete ActiveMacros.back(); 2823 ActiveMacros.pop_back(); 2824 } 2825 2826 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef, 2827 bool NoDeadStrip) { 2828 MCSymbol *Sym; 2829 const MCExpr *Value; 2830 if (MCParserUtils::parseAssignmentExpression(Name, allow_redef, *this, Sym, 2831 Value)) 2832 return true; 2833 2834 if (!Sym) { 2835 // In the case where we parse an expression starting with a '.', we will 2836 // not generate an error, nor will we create a symbol. In this case we 2837 // should just return out. 2838 return false; 2839 } 2840 2841 // Do the assignment. 2842 Out.EmitAssignment(Sym, Value); 2843 if (NoDeadStrip) 2844 Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip); 2845 2846 return false; 2847 } 2848 2849 /// parseIdentifier: 2850 /// ::= identifier 2851 /// ::= string 2852 bool AsmParser::parseIdentifier(StringRef &Res) { 2853 // The assembler has relaxed rules for accepting identifiers, in particular we 2854 // allow things like '.globl $foo' and '.def @feat.00', which would normally be 2855 // separate tokens. At this level, we have already lexed so we cannot (currently) 2856 // handle this as a context dependent token, instead we detect adjacent tokens 2857 // and return the combined identifier. 2858 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) { 2859 SMLoc PrefixLoc = getLexer().getLoc(); 2860 2861 // Consume the prefix character, and check for a following identifier. 2862 2863 AsmToken Buf[1]; 2864 Lexer.peekTokens(Buf, false); 2865 2866 if (Buf[0].isNot(AsmToken::Identifier)) 2867 return true; 2868 2869 // We have a '$' or '@' followed by an identifier, make sure they are adjacent. 2870 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer()) 2871 return true; 2872 2873 // eat $ or @ 2874 Lexer.Lex(); // Lexer's Lex guarantees consecutive token. 2875 // Construct the joined identifier and consume the token. 2876 Res = 2877 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1); 2878 Lex(); // Parser Lex to maintain invariants. 2879 return false; 2880 } 2881 2882 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String)) 2883 return true; 2884 2885 Res = getTok().getIdentifier(); 2886 2887 Lex(); // Consume the identifier token. 2888 2889 return false; 2890 } 2891 2892 /// parseDirectiveSet: 2893 /// ::= .equ identifier ',' expression 2894 /// ::= .equiv identifier ',' expression 2895 /// ::= .set identifier ',' expression 2896 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) { 2897 StringRef Name; 2898 if (check(parseIdentifier(Name), "expected identifier") || 2899 parseToken(AsmToken::Comma) || parseAssignment(Name, allow_redef, true)) 2900 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 2901 return false; 2902 } 2903 2904 bool AsmParser::parseEscapedString(std::string &Data) { 2905 if (check(getTok().isNot(AsmToken::String), "expected string")) 2906 return true; 2907 2908 Data = ""; 2909 StringRef Str = getTok().getStringContents(); 2910 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 2911 if (Str[i] != '\\') { 2912 Data += Str[i]; 2913 continue; 2914 } 2915 2916 // Recognize escaped characters. Note that this escape semantics currently 2917 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes. 2918 ++i; 2919 if (i == e) 2920 return TokError("unexpected backslash at end of string"); 2921 2922 // Recognize octal sequences. 2923 if ((unsigned)(Str[i] - '0') <= 7) { 2924 // Consume up to three octal characters. 2925 unsigned Value = Str[i] - '0'; 2926 2927 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 2928 ++i; 2929 Value = Value * 8 + (Str[i] - '0'); 2930 2931 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 2932 ++i; 2933 Value = Value * 8 + (Str[i] - '0'); 2934 } 2935 } 2936 2937 if (Value > 255) 2938 return TokError("invalid octal escape sequence (out of range)"); 2939 2940 Data += (unsigned char)Value; 2941 continue; 2942 } 2943 2944 // Otherwise recognize individual escapes. 2945 switch (Str[i]) { 2946 default: 2947 // Just reject invalid escape sequences for now. 2948 return TokError("invalid escape sequence (unrecognized character)"); 2949 2950 case 'b': Data += '\b'; break; 2951 case 'f': Data += '\f'; break; 2952 case 'n': Data += '\n'; break; 2953 case 'r': Data += '\r'; break; 2954 case 't': Data += '\t'; break; 2955 case '"': Data += '"'; break; 2956 case '\\': Data += '\\'; break; 2957 } 2958 } 2959 2960 Lex(); 2961 return false; 2962 } 2963 2964 /// parseDirectiveAscii: 2965 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 2966 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 2967 auto parseOp = [&]() -> bool { 2968 std::string Data; 2969 if (checkForValidSection() || parseEscapedString(Data)) 2970 return true; 2971 getStreamer().EmitBytes(Data); 2972 if (ZeroTerminated) 2973 getStreamer().EmitBytes(StringRef("\0", 1)); 2974 return false; 2975 }; 2976 2977 if (parseMany(parseOp)) 2978 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 2979 return false; 2980 } 2981 2982 /// parseDirectiveReloc 2983 /// ::= .reloc expression , identifier [ , expression ] 2984 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) { 2985 const MCExpr *Offset; 2986 const MCExpr *Expr = nullptr; 2987 int64_t OffsetValue; 2988 SMLoc OffsetLoc = Lexer.getTok().getLoc(); 2989 2990 if (parseExpression(Offset)) 2991 return true; 2992 2993 if ((Offset->evaluateAsAbsolute(OffsetValue, 2994 getStreamer().getAssemblerPtr()) && 2995 check(OffsetValue < 0, OffsetLoc, "expression is negative")) || 2996 (check(Offset->getKind() != llvm::MCExpr::Constant && 2997 Offset->getKind() != llvm::MCExpr::SymbolRef, 2998 OffsetLoc, "expected non-negative number or a label")) || 2999 (parseToken(AsmToken::Comma, "expected comma") || 3000 check(getTok().isNot(AsmToken::Identifier), "expected relocation name"))) 3001 return true; 3002 3003 SMLoc NameLoc = Lexer.getTok().getLoc(); 3004 StringRef Name = Lexer.getTok().getIdentifier(); 3005 Lex(); 3006 3007 if (Lexer.is(AsmToken::Comma)) { 3008 Lex(); 3009 SMLoc ExprLoc = Lexer.getLoc(); 3010 if (parseExpression(Expr)) 3011 return true; 3012 3013 MCValue Value; 3014 if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr)) 3015 return Error(ExprLoc, "expression must be relocatable"); 3016 } 3017 3018 if (parseToken(AsmToken::EndOfStatement, 3019 "unexpected token in .reloc directive")) 3020 return true; 3021 3022 const MCTargetAsmParser &MCT = getTargetParser(); 3023 const MCSubtargetInfo &STI = MCT.getSTI(); 3024 if (getStreamer().EmitRelocDirective(*Offset, Name, Expr, DirectiveLoc, STI)) 3025 return Error(NameLoc, "unknown relocation name"); 3026 3027 return false; 3028 } 3029 3030 /// parseDirectiveValue 3031 /// ::= (.byte | .short | ... ) [ expression (, expression)* ] 3032 bool AsmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { 3033 auto parseOp = [&]() -> bool { 3034 const MCExpr *Value; 3035 SMLoc ExprLoc = getLexer().getLoc(); 3036 if (checkForValidSection() || parseExpression(Value)) 3037 return true; 3038 // Special case constant expressions to match code generator. 3039 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3040 assert(Size <= 8 && "Invalid size"); 3041 uint64_t IntValue = MCE->getValue(); 3042 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 3043 return Error(ExprLoc, "out of range literal value"); 3044 getStreamer().EmitIntValue(IntValue, Size); 3045 } else 3046 getStreamer().EmitValue(Value, Size, ExprLoc); 3047 return false; 3048 }; 3049 3050 if (parseMany(parseOp)) 3051 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3052 return false; 3053 } 3054 3055 static bool parseHexOcta(AsmParser &Asm, uint64_t &hi, uint64_t &lo) { 3056 if (Asm.getTok().isNot(AsmToken::Integer) && 3057 Asm.getTok().isNot(AsmToken::BigNum)) 3058 return Asm.TokError("unknown token in expression"); 3059 SMLoc ExprLoc = Asm.getTok().getLoc(); 3060 APInt IntValue = Asm.getTok().getAPIntVal(); 3061 Asm.Lex(); 3062 if (!IntValue.isIntN(128)) 3063 return Asm.Error(ExprLoc, "out of range literal value"); 3064 if (!IntValue.isIntN(64)) { 3065 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue(); 3066 lo = IntValue.getLoBits(64).getZExtValue(); 3067 } else { 3068 hi = 0; 3069 lo = IntValue.getZExtValue(); 3070 } 3071 return false; 3072 } 3073 3074 /// ParseDirectiveOctaValue 3075 /// ::= .octa [ hexconstant (, hexconstant)* ] 3076 3077 bool AsmParser::parseDirectiveOctaValue(StringRef IDVal) { 3078 auto parseOp = [&]() -> bool { 3079 if (checkForValidSection()) 3080 return true; 3081 uint64_t hi, lo; 3082 if (parseHexOcta(*this, hi, lo)) 3083 return true; 3084 if (MAI.isLittleEndian()) { 3085 getStreamer().EmitIntValue(lo, 8); 3086 getStreamer().EmitIntValue(hi, 8); 3087 } else { 3088 getStreamer().EmitIntValue(hi, 8); 3089 getStreamer().EmitIntValue(lo, 8); 3090 } 3091 return false; 3092 }; 3093 3094 if (parseMany(parseOp)) 3095 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3096 return false; 3097 } 3098 3099 bool AsmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { 3100 // We don't truly support arithmetic on floating point expressions, so we 3101 // have to manually parse unary prefixes. 3102 bool IsNeg = false; 3103 if (getLexer().is(AsmToken::Minus)) { 3104 Lexer.Lex(); 3105 IsNeg = true; 3106 } else if (getLexer().is(AsmToken::Plus)) 3107 Lexer.Lex(); 3108 3109 if (Lexer.is(AsmToken::Error)) 3110 return TokError(Lexer.getErr()); 3111 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) && 3112 Lexer.isNot(AsmToken::Identifier)) 3113 return TokError("unexpected token in directive"); 3114 3115 // Convert to an APFloat. 3116 APFloat Value(Semantics); 3117 StringRef IDVal = getTok().getString(); 3118 if (getLexer().is(AsmToken::Identifier)) { 3119 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf")) 3120 Value = APFloat::getInf(Semantics); 3121 else if (!IDVal.compare_lower("nan")) 3122 Value = APFloat::getNaN(Semantics, false, ~0); 3123 else 3124 return TokError("invalid floating point literal"); 3125 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) == 3126 APFloat::opInvalidOp) 3127 return TokError("invalid floating point literal"); 3128 if (IsNeg) 3129 Value.changeSign(); 3130 3131 // Consume the numeric token. 3132 Lex(); 3133 3134 Res = Value.bitcastToAPInt(); 3135 3136 return false; 3137 } 3138 3139 /// parseDirectiveRealValue 3140 /// ::= (.single | .double) [ expression (, expression)* ] 3141 bool AsmParser::parseDirectiveRealValue(StringRef IDVal, 3142 const fltSemantics &Semantics) { 3143 auto parseOp = [&]() -> bool { 3144 APInt AsInt; 3145 if (checkForValidSection() || parseRealValue(Semantics, AsInt)) 3146 return true; 3147 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 3148 AsInt.getBitWidth() / 8); 3149 return false; 3150 }; 3151 3152 if (parseMany(parseOp)) 3153 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3154 return false; 3155 } 3156 3157 /// parseDirectiveZero 3158 /// ::= .zero expression 3159 bool AsmParser::parseDirectiveZero() { 3160 SMLoc NumBytesLoc = Lexer.getLoc(); 3161 const MCExpr *NumBytes; 3162 if (checkForValidSection() || parseExpression(NumBytes)) 3163 return true; 3164 3165 int64_t Val = 0; 3166 if (getLexer().is(AsmToken::Comma)) { 3167 Lex(); 3168 if (parseAbsoluteExpression(Val)) 3169 return true; 3170 } 3171 3172 if (parseToken(AsmToken::EndOfStatement, 3173 "unexpected token in '.zero' directive")) 3174 return true; 3175 getStreamer().emitFill(*NumBytes, Val, NumBytesLoc); 3176 3177 return false; 3178 } 3179 3180 /// parseDirectiveFill 3181 /// ::= .fill expression [ , expression [ , expression ] ] 3182 bool AsmParser::parseDirectiveFill() { 3183 SMLoc NumValuesLoc = Lexer.getLoc(); 3184 const MCExpr *NumValues; 3185 if (checkForValidSection() || parseExpression(NumValues)) 3186 return true; 3187 3188 int64_t FillSize = 1; 3189 int64_t FillExpr = 0; 3190 3191 SMLoc SizeLoc, ExprLoc; 3192 3193 if (parseOptionalToken(AsmToken::Comma)) { 3194 SizeLoc = getTok().getLoc(); 3195 if (parseAbsoluteExpression(FillSize)) 3196 return true; 3197 if (parseOptionalToken(AsmToken::Comma)) { 3198 ExprLoc = getTok().getLoc(); 3199 if (parseAbsoluteExpression(FillExpr)) 3200 return true; 3201 } 3202 } 3203 if (parseToken(AsmToken::EndOfStatement, 3204 "unexpected token in '.fill' directive")) 3205 return true; 3206 3207 if (FillSize < 0) { 3208 Warning(SizeLoc, "'.fill' directive with negative size has no effect"); 3209 return false; 3210 } 3211 if (FillSize > 8) { 3212 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8"); 3213 FillSize = 8; 3214 } 3215 3216 if (!isUInt<32>(FillExpr) && FillSize > 4) 3217 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits"); 3218 3219 getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc); 3220 3221 return false; 3222 } 3223 3224 /// parseDirectiveOrg 3225 /// ::= .org expression [ , expression ] 3226 bool AsmParser::parseDirectiveOrg() { 3227 const MCExpr *Offset; 3228 SMLoc OffsetLoc = Lexer.getLoc(); 3229 if (checkForValidSection() || parseExpression(Offset)) 3230 return true; 3231 3232 // Parse optional fill expression. 3233 int64_t FillExpr = 0; 3234 if (parseOptionalToken(AsmToken::Comma)) 3235 if (parseAbsoluteExpression(FillExpr)) 3236 return addErrorSuffix(" in '.org' directive"); 3237 if (parseToken(AsmToken::EndOfStatement)) 3238 return addErrorSuffix(" in '.org' directive"); 3239 3240 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc); 3241 return false; 3242 } 3243 3244 /// parseDirectiveAlign 3245 /// ::= {.align, ...} expression [ , expression [ , expression ]] 3246 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 3247 SMLoc AlignmentLoc = getLexer().getLoc(); 3248 int64_t Alignment; 3249 SMLoc MaxBytesLoc; 3250 bool HasFillExpr = false; 3251 int64_t FillExpr = 0; 3252 int64_t MaxBytesToFill = 0; 3253 3254 auto parseAlign = [&]() -> bool { 3255 if (parseAbsoluteExpression(Alignment)) 3256 return true; 3257 if (parseOptionalToken(AsmToken::Comma)) { 3258 // The fill expression can be omitted while specifying a maximum number of 3259 // alignment bytes, e.g: 3260 // .align 3,,4 3261 if (getTok().isNot(AsmToken::Comma)) { 3262 HasFillExpr = true; 3263 if (parseAbsoluteExpression(FillExpr)) 3264 return true; 3265 } 3266 if (parseOptionalToken(AsmToken::Comma)) 3267 if (parseTokenLoc(MaxBytesLoc) || 3268 parseAbsoluteExpression(MaxBytesToFill)) 3269 return true; 3270 } 3271 return parseToken(AsmToken::EndOfStatement); 3272 }; 3273 3274 if (checkForValidSection()) 3275 return addErrorSuffix(" in directive"); 3276 // Ignore empty '.p2align' directives for GNU-as compatibility 3277 if (IsPow2 && (ValueSize == 1) && getTok().is(AsmToken::EndOfStatement)) { 3278 Warning(AlignmentLoc, "p2align directive with no operand(s) is ignored"); 3279 return parseToken(AsmToken::EndOfStatement); 3280 } 3281 if (parseAlign()) 3282 return addErrorSuffix(" in directive"); 3283 3284 // Always emit an alignment here even if we thrown an error. 3285 bool ReturnVal = false; 3286 3287 // Compute alignment in bytes. 3288 if (IsPow2) { 3289 // FIXME: Diagnose overflow. 3290 if (Alignment >= 32) { 3291 ReturnVal |= Error(AlignmentLoc, "invalid alignment value"); 3292 Alignment = 31; 3293 } 3294 3295 Alignment = 1ULL << Alignment; 3296 } else { 3297 // Reject alignments that aren't either a power of two or zero, 3298 // for gas compatibility. Alignment of zero is silently rounded 3299 // up to one. 3300 if (Alignment == 0) 3301 Alignment = 1; 3302 if (!isPowerOf2_64(Alignment)) 3303 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2"); 3304 } 3305 3306 // Diagnose non-sensical max bytes to align. 3307 if (MaxBytesLoc.isValid()) { 3308 if (MaxBytesToFill < 1) { 3309 ReturnVal |= Error(MaxBytesLoc, 3310 "alignment directive can never be satisfied in this " 3311 "many bytes, ignoring maximum bytes expression"); 3312 MaxBytesToFill = 0; 3313 } 3314 3315 if (MaxBytesToFill >= Alignment) { 3316 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 3317 "has no effect"); 3318 MaxBytesToFill = 0; 3319 } 3320 } 3321 3322 // Check whether we should use optimal code alignment for this .align 3323 // directive. 3324 const MCSection *Section = getStreamer().getCurrentSectionOnly(); 3325 assert(Section && "must have section to emit alignment"); 3326 bool UseCodeAlign = Section->UseCodeAlign(); 3327 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 3328 ValueSize == 1 && UseCodeAlign) { 3329 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill); 3330 } else { 3331 // FIXME: Target specific behavior about how the "extra" bytes are filled. 3332 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize, 3333 MaxBytesToFill); 3334 } 3335 3336 return ReturnVal; 3337 } 3338 3339 /// parseDirectiveFile 3340 /// ::= .file filename 3341 /// ::= .file number [directory] filename [md5 checksum] [source source-text] 3342 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) { 3343 // FIXME: I'm not sure what this is. 3344 int64_t FileNumber = -1; 3345 if (getLexer().is(AsmToken::Integer)) { 3346 FileNumber = getTok().getIntVal(); 3347 Lex(); 3348 3349 if (FileNumber < 0) 3350 return TokError("negative file number"); 3351 } 3352 3353 std::string Path; 3354 3355 // Usually the directory and filename together, otherwise just the directory. 3356 // Allow the strings to have escaped octal character sequence. 3357 if (check(getTok().isNot(AsmToken::String), 3358 "unexpected token in '.file' directive") || 3359 parseEscapedString(Path)) 3360 return true; 3361 3362 StringRef Directory; 3363 StringRef Filename; 3364 std::string FilenameData; 3365 if (getLexer().is(AsmToken::String)) { 3366 if (check(FileNumber == -1, 3367 "explicit path specified, but no file number") || 3368 parseEscapedString(FilenameData)) 3369 return true; 3370 Filename = FilenameData; 3371 Directory = Path; 3372 } else { 3373 Filename = Path; 3374 } 3375 3376 uint64_t MD5Hi, MD5Lo; 3377 bool HasMD5 = false; 3378 3379 Optional<StringRef> Source; 3380 bool HasSource = false; 3381 std::string SourceString; 3382 3383 while (!parseOptionalToken(AsmToken::EndOfStatement)) { 3384 StringRef Keyword; 3385 if (check(getTok().isNot(AsmToken::Identifier), 3386 "unexpected token in '.file' directive") || 3387 parseIdentifier(Keyword)) 3388 return true; 3389 if (Keyword == "md5") { 3390 HasMD5 = true; 3391 if (check(FileNumber == -1, 3392 "MD5 checksum specified, but no file number") || 3393 parseHexOcta(*this, MD5Hi, MD5Lo)) 3394 return true; 3395 } else if (Keyword == "source") { 3396 HasSource = true; 3397 if (check(FileNumber == -1, 3398 "source specified, but no file number") || 3399 check(getTok().isNot(AsmToken::String), 3400 "unexpected token in '.file' directive") || 3401 parseEscapedString(SourceString)) 3402 return true; 3403 } else { 3404 return TokError("unexpected token in '.file' directive"); 3405 } 3406 } 3407 3408 if (FileNumber == -1) { 3409 // Ignore the directive if there is no number and the target doesn't support 3410 // numberless .file directives. This allows some portability of assembler 3411 // between different object file formats. 3412 if (getContext().getAsmInfo()->hasSingleParameterDotFile()) 3413 getStreamer().EmitFileDirective(Filename); 3414 } else { 3415 // In case there is a -g option as well as debug info from directive .file, 3416 // we turn off the -g option, directly use the existing debug info instead. 3417 // Throw away any implicit file table for the assembler source. 3418 if (Ctx.getGenDwarfForAssembly()) { 3419 Ctx.getMCDwarfLineTable(0).resetFileTable(); 3420 Ctx.setGenDwarfForAssembly(false); 3421 } 3422 3423 Optional<MD5::MD5Result> CKMem; 3424 if (HasMD5) { 3425 MD5::MD5Result Sum; 3426 for (unsigned i = 0; i != 8; ++i) { 3427 Sum.Bytes[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); 3428 Sum.Bytes[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); 3429 } 3430 CKMem = Sum; 3431 } 3432 if (HasSource) { 3433 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size())); 3434 memcpy(SourceBuf, SourceString.data(), SourceString.size()); 3435 Source = StringRef(SourceBuf, SourceString.size()); 3436 } 3437 if (FileNumber == 0) { 3438 if (Ctx.getDwarfVersion() < 5) 3439 return Warning(DirectiveLoc, "file 0 not supported prior to DWARF-5"); 3440 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source); 3441 } else { 3442 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( 3443 FileNumber, Directory, Filename, CKMem, Source); 3444 if (!FileNumOrErr) 3445 return Error(DirectiveLoc, toString(FileNumOrErr.takeError())); 3446 } 3447 // Alert the user if there are some .file directives with MD5 and some not. 3448 // But only do that once. 3449 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) { 3450 ReportedInconsistentMD5 = true; 3451 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums"); 3452 } 3453 } 3454 3455 return false; 3456 } 3457 3458 /// parseDirectiveLine 3459 /// ::= .line [number] 3460 bool AsmParser::parseDirectiveLine() { 3461 int64_t LineNumber; 3462 if (getLexer().is(AsmToken::Integer)) { 3463 if (parseIntToken(LineNumber, "unexpected token in '.line' directive")) 3464 return true; 3465 (void)LineNumber; 3466 // FIXME: Do something with the .line. 3467 } 3468 if (parseToken(AsmToken::EndOfStatement, 3469 "unexpected token in '.line' directive")) 3470 return true; 3471 3472 return false; 3473 } 3474 3475 /// parseDirectiveLoc 3476 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 3477 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 3478 /// The first number is a file number, must have been previously assigned with 3479 /// a .file directive, the second number is the line number and optionally the 3480 /// third number is a column position (zero if not specified). The remaining 3481 /// optional items are .loc sub-directives. 3482 bool AsmParser::parseDirectiveLoc() { 3483 int64_t FileNumber = 0, LineNumber = 0; 3484 SMLoc Loc = getTok().getLoc(); 3485 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") || 3486 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, 3487 "file number less than one in '.loc' directive") || 3488 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc, 3489 "unassigned file number in '.loc' directive")) 3490 return true; 3491 3492 // optional 3493 if (getLexer().is(AsmToken::Integer)) { 3494 LineNumber = getTok().getIntVal(); 3495 if (LineNumber < 0) 3496 return TokError("line number less than zero in '.loc' directive"); 3497 Lex(); 3498 } 3499 3500 int64_t ColumnPos = 0; 3501 if (getLexer().is(AsmToken::Integer)) { 3502 ColumnPos = getTok().getIntVal(); 3503 if (ColumnPos < 0) 3504 return TokError("column position less than zero in '.loc' directive"); 3505 Lex(); 3506 } 3507 3508 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0; 3509 unsigned Isa = 0; 3510 int64_t Discriminator = 0; 3511 3512 auto parseLocOp = [&]() -> bool { 3513 StringRef Name; 3514 SMLoc Loc = getTok().getLoc(); 3515 if (parseIdentifier(Name)) 3516 return TokError("unexpected token in '.loc' directive"); 3517 3518 if (Name == "basic_block") 3519 Flags |= DWARF2_FLAG_BASIC_BLOCK; 3520 else if (Name == "prologue_end") 3521 Flags |= DWARF2_FLAG_PROLOGUE_END; 3522 else if (Name == "epilogue_begin") 3523 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 3524 else if (Name == "is_stmt") { 3525 Loc = getTok().getLoc(); 3526 const MCExpr *Value; 3527 if (parseExpression(Value)) 3528 return true; 3529 // The expression must be the constant 0 or 1. 3530 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3531 int Value = MCE->getValue(); 3532 if (Value == 0) 3533 Flags &= ~DWARF2_FLAG_IS_STMT; 3534 else if (Value == 1) 3535 Flags |= DWARF2_FLAG_IS_STMT; 3536 else 3537 return Error(Loc, "is_stmt value not 0 or 1"); 3538 } else { 3539 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 3540 } 3541 } else if (Name == "isa") { 3542 Loc = getTok().getLoc(); 3543 const MCExpr *Value; 3544 if (parseExpression(Value)) 3545 return true; 3546 // The expression must be a constant greater or equal to 0. 3547 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3548 int Value = MCE->getValue(); 3549 if (Value < 0) 3550 return Error(Loc, "isa number less than zero"); 3551 Isa = Value; 3552 } else { 3553 return Error(Loc, "isa number not a constant value"); 3554 } 3555 } else if (Name == "discriminator") { 3556 if (parseAbsoluteExpression(Discriminator)) 3557 return true; 3558 } else { 3559 return Error(Loc, "unknown sub-directive in '.loc' directive"); 3560 } 3561 return false; 3562 }; 3563 3564 if (parseMany(parseLocOp, false /*hasComma*/)) 3565 return true; 3566 3567 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 3568 Isa, Discriminator, StringRef()); 3569 3570 return false; 3571 } 3572 3573 /// parseDirectiveStabs 3574 /// ::= .stabs string, number, number, number 3575 bool AsmParser::parseDirectiveStabs() { 3576 return TokError("unsupported directive '.stabs'"); 3577 } 3578 3579 /// parseDirectiveCVFile 3580 /// ::= .cv_file number filename [checksum] [checksumkind] 3581 bool AsmParser::parseDirectiveCVFile() { 3582 SMLoc FileNumberLoc = getTok().getLoc(); 3583 int64_t FileNumber; 3584 std::string Filename; 3585 std::string Checksum; 3586 int64_t ChecksumKind = 0; 3587 3588 if (parseIntToken(FileNumber, 3589 "expected file number in '.cv_file' directive") || 3590 check(FileNumber < 1, FileNumberLoc, "file number less than one") || 3591 check(getTok().isNot(AsmToken::String), 3592 "unexpected token in '.cv_file' directive") || 3593 parseEscapedString(Filename)) 3594 return true; 3595 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 3596 if (check(getTok().isNot(AsmToken::String), 3597 "unexpected token in '.cv_file' directive") || 3598 parseEscapedString(Checksum) || 3599 parseIntToken(ChecksumKind, 3600 "expected checksum kind in '.cv_file' directive") || 3601 parseToken(AsmToken::EndOfStatement, 3602 "unexpected token in '.cv_file' directive")) 3603 return true; 3604 } 3605 3606 Checksum = fromHex(Checksum); 3607 void *CKMem = Ctx.allocate(Checksum.size(), 1); 3608 memcpy(CKMem, Checksum.data(), Checksum.size()); 3609 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), 3610 Checksum.size()); 3611 3612 if (!getStreamer().EmitCVFileDirective(FileNumber, Filename, ChecksumAsBytes, 3613 static_cast<uint8_t>(ChecksumKind))) 3614 return Error(FileNumberLoc, "file number already allocated"); 3615 3616 return false; 3617 } 3618 3619 bool AsmParser::parseCVFunctionId(int64_t &FunctionId, 3620 StringRef DirectiveName) { 3621 SMLoc Loc; 3622 return parseTokenLoc(Loc) || 3623 parseIntToken(FunctionId, "expected function id in '" + DirectiveName + 3624 "' directive") || 3625 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc, 3626 "expected function id within range [0, UINT_MAX)"); 3627 } 3628 3629 bool AsmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { 3630 SMLoc Loc; 3631 return parseTokenLoc(Loc) || 3632 parseIntToken(FileNumber, "expected integer in '" + DirectiveName + 3633 "' directive") || 3634 check(FileNumber < 1, Loc, "file number less than one in '" + 3635 DirectiveName + "' directive") || 3636 check(!getCVContext().isValidFileNumber(FileNumber), Loc, 3637 "unassigned file number in '" + DirectiveName + "' directive"); 3638 } 3639 3640 /// parseDirectiveCVFuncId 3641 /// ::= .cv_func_id FunctionId 3642 /// 3643 /// Introduces a function ID that can be used with .cv_loc. 3644 bool AsmParser::parseDirectiveCVFuncId() { 3645 SMLoc FunctionIdLoc = getTok().getLoc(); 3646 int64_t FunctionId; 3647 3648 if (parseCVFunctionId(FunctionId, ".cv_func_id") || 3649 parseToken(AsmToken::EndOfStatement, 3650 "unexpected token in '.cv_func_id' directive")) 3651 return true; 3652 3653 if (!getStreamer().EmitCVFuncIdDirective(FunctionId)) 3654 return Error(FunctionIdLoc, "function id already allocated"); 3655 3656 return false; 3657 } 3658 3659 /// parseDirectiveCVInlineSiteId 3660 /// ::= .cv_inline_site_id FunctionId 3661 /// "within" IAFunc 3662 /// "inlined_at" IAFile IALine [IACol] 3663 /// 3664 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined 3665 /// at" source location information for use in the line table of the caller, 3666 /// whether the caller is a real function or another inlined call site. 3667 bool AsmParser::parseDirectiveCVInlineSiteId() { 3668 SMLoc FunctionIdLoc = getTok().getLoc(); 3669 int64_t FunctionId; 3670 int64_t IAFunc; 3671 int64_t IAFile; 3672 int64_t IALine; 3673 int64_t IACol = 0; 3674 3675 // FunctionId 3676 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id")) 3677 return true; 3678 3679 // "within" 3680 if (check((getLexer().isNot(AsmToken::Identifier) || 3681 getTok().getIdentifier() != "within"), 3682 "expected 'within' identifier in '.cv_inline_site_id' directive")) 3683 return true; 3684 Lex(); 3685 3686 // IAFunc 3687 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id")) 3688 return true; 3689 3690 // "inlined_at" 3691 if (check((getLexer().isNot(AsmToken::Identifier) || 3692 getTok().getIdentifier() != "inlined_at"), 3693 "expected 'inlined_at' identifier in '.cv_inline_site_id' " 3694 "directive") ) 3695 return true; 3696 Lex(); 3697 3698 // IAFile IALine 3699 if (parseCVFileId(IAFile, ".cv_inline_site_id") || 3700 parseIntToken(IALine, "expected line number after 'inlined_at'")) 3701 return true; 3702 3703 // [IACol] 3704 if (getLexer().is(AsmToken::Integer)) { 3705 IACol = getTok().getIntVal(); 3706 Lex(); 3707 } 3708 3709 if (parseToken(AsmToken::EndOfStatement, 3710 "unexpected token in '.cv_inline_site_id' directive")) 3711 return true; 3712 3713 if (!getStreamer().EmitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, 3714 IALine, IACol, FunctionIdLoc)) 3715 return Error(FunctionIdLoc, "function id already allocated"); 3716 3717 return false; 3718 } 3719 3720 /// parseDirectiveCVLoc 3721 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] 3722 /// [is_stmt VALUE] 3723 /// The first number is a file number, must have been previously assigned with 3724 /// a .file directive, the second number is the line number and optionally the 3725 /// third number is a column position (zero if not specified). The remaining 3726 /// optional items are .loc sub-directives. 3727 bool AsmParser::parseDirectiveCVLoc() { 3728 SMLoc DirectiveLoc = getTok().getLoc(); 3729 int64_t FunctionId, FileNumber; 3730 if (parseCVFunctionId(FunctionId, ".cv_loc") || 3731 parseCVFileId(FileNumber, ".cv_loc")) 3732 return true; 3733 3734 int64_t LineNumber = 0; 3735 if (getLexer().is(AsmToken::Integer)) { 3736 LineNumber = getTok().getIntVal(); 3737 if (LineNumber < 0) 3738 return TokError("line number less than zero in '.cv_loc' directive"); 3739 Lex(); 3740 } 3741 3742 int64_t ColumnPos = 0; 3743 if (getLexer().is(AsmToken::Integer)) { 3744 ColumnPos = getTok().getIntVal(); 3745 if (ColumnPos < 0) 3746 return TokError("column position less than zero in '.cv_loc' directive"); 3747 Lex(); 3748 } 3749 3750 bool PrologueEnd = false; 3751 uint64_t IsStmt = 0; 3752 3753 auto parseOp = [&]() -> bool { 3754 StringRef Name; 3755 SMLoc Loc = getTok().getLoc(); 3756 if (parseIdentifier(Name)) 3757 return TokError("unexpected token in '.cv_loc' directive"); 3758 if (Name == "prologue_end") 3759 PrologueEnd = true; 3760 else if (Name == "is_stmt") { 3761 Loc = getTok().getLoc(); 3762 const MCExpr *Value; 3763 if (parseExpression(Value)) 3764 return true; 3765 // The expression must be the constant 0 or 1. 3766 IsStmt = ~0ULL; 3767 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value)) 3768 IsStmt = MCE->getValue(); 3769 3770 if (IsStmt > 1) 3771 return Error(Loc, "is_stmt value not 0 or 1"); 3772 } else { 3773 return Error(Loc, "unknown sub-directive in '.cv_loc' directive"); 3774 } 3775 return false; 3776 }; 3777 3778 if (parseMany(parseOp, false /*hasComma*/)) 3779 return true; 3780 3781 getStreamer().EmitCVLocDirective(FunctionId, FileNumber, LineNumber, 3782 ColumnPos, PrologueEnd, IsStmt, StringRef(), 3783 DirectiveLoc); 3784 return false; 3785 } 3786 3787 /// parseDirectiveCVLinetable 3788 /// ::= .cv_linetable FunctionId, FnStart, FnEnd 3789 bool AsmParser::parseDirectiveCVLinetable() { 3790 int64_t FunctionId; 3791 StringRef FnStartName, FnEndName; 3792 SMLoc Loc = getTok().getLoc(); 3793 if (parseCVFunctionId(FunctionId, ".cv_linetable") || 3794 parseToken(AsmToken::Comma, 3795 "unexpected token in '.cv_linetable' directive") || 3796 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3797 "expected identifier in directive") || 3798 parseToken(AsmToken::Comma, 3799 "unexpected token in '.cv_linetable' directive") || 3800 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3801 "expected identifier in directive")) 3802 return true; 3803 3804 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3805 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3806 3807 getStreamer().EmitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym); 3808 return false; 3809 } 3810 3811 /// parseDirectiveCVInlineLinetable 3812 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd 3813 bool AsmParser::parseDirectiveCVInlineLinetable() { 3814 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; 3815 StringRef FnStartName, FnEndName; 3816 SMLoc Loc = getTok().getLoc(); 3817 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") || 3818 parseTokenLoc(Loc) || 3819 parseIntToken( 3820 SourceFileId, 3821 "expected SourceField in '.cv_inline_linetable' directive") || 3822 check(SourceFileId <= 0, Loc, 3823 "File id less than zero in '.cv_inline_linetable' directive") || 3824 parseTokenLoc(Loc) || 3825 parseIntToken( 3826 SourceLineNum, 3827 "expected SourceLineNum in '.cv_inline_linetable' directive") || 3828 check(SourceLineNum < 0, Loc, 3829 "Line number less than zero in '.cv_inline_linetable' directive") || 3830 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3831 "expected identifier in directive") || 3832 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3833 "expected identifier in directive")) 3834 return true; 3835 3836 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 3837 return true; 3838 3839 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3840 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3841 getStreamer().EmitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, 3842 SourceLineNum, FnStartSym, 3843 FnEndSym); 3844 return false; 3845 } 3846 3847 void AsmParser::initializeCVDefRangeTypeMap() { 3848 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER; 3849 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL; 3850 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER; 3851 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL; 3852 } 3853 3854 /// parseDirectiveCVDefRange 3855 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* 3856 bool AsmParser::parseDirectiveCVDefRange() { 3857 SMLoc Loc; 3858 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; 3859 while (getLexer().is(AsmToken::Identifier)) { 3860 Loc = getLexer().getLoc(); 3861 StringRef GapStartName; 3862 if (parseIdentifier(GapStartName)) 3863 return Error(Loc, "expected identifier in directive"); 3864 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName); 3865 3866 Loc = getLexer().getLoc(); 3867 StringRef GapEndName; 3868 if (parseIdentifier(GapEndName)) 3869 return Error(Loc, "expected identifier in directive"); 3870 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName); 3871 3872 Ranges.push_back({GapStartSym, GapEndSym}); 3873 } 3874 3875 StringRef CVDefRangeTypeStr; 3876 if (parseToken( 3877 AsmToken::Comma, 3878 "expected comma before def_range type in .cv_def_range directive") || 3879 parseIdentifier(CVDefRangeTypeStr)) 3880 return Error(Loc, "expected def_range type in directive"); 3881 3882 StringMap<CVDefRangeType>::const_iterator CVTypeIt = 3883 CVDefRangeTypeMap.find(CVDefRangeTypeStr); 3884 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end()) 3885 ? CVDR_DEFRANGE 3886 : CVTypeIt->getValue(); 3887 switch (CVDRType) { 3888 case CVDR_DEFRANGE_REGISTER: { 3889 int64_t DRRegister; 3890 if (parseToken(AsmToken::Comma, "expected comma before register number in " 3891 ".cv_def_range directive") || 3892 parseAbsoluteExpression(DRRegister)) 3893 return Error(Loc, "expected register number"); 3894 3895 codeview::DefRangeRegisterSym::Header DRHdr; 3896 DRHdr.Register = DRRegister; 3897 DRHdr.MayHaveNoName = 0; 3898 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3899 break; 3900 } 3901 case CVDR_DEFRANGE_FRAMEPOINTER_REL: { 3902 int64_t DROffset; 3903 if (parseToken(AsmToken::Comma, 3904 "expected comma before offset in .cv_def_range directive") || 3905 parseAbsoluteExpression(DROffset)) 3906 return Error(Loc, "expected offset value"); 3907 3908 codeview::DefRangeFramePointerRelSym::Header DRHdr; 3909 DRHdr.Offset = DROffset; 3910 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3911 break; 3912 } 3913 case CVDR_DEFRANGE_SUBFIELD_REGISTER: { 3914 int64_t DRRegister; 3915 int64_t DROffsetInParent; 3916 if (parseToken(AsmToken::Comma, "expected comma before register number in " 3917 ".cv_def_range directive") || 3918 parseAbsoluteExpression(DRRegister)) 3919 return Error(Loc, "expected register number"); 3920 if (parseToken(AsmToken::Comma, 3921 "expected comma before offset in .cv_def_range directive") || 3922 parseAbsoluteExpression(DROffsetInParent)) 3923 return Error(Loc, "expected offset value"); 3924 3925 codeview::DefRangeSubfieldRegisterSym::Header DRHdr; 3926 DRHdr.Register = DRRegister; 3927 DRHdr.MayHaveNoName = 0; 3928 DRHdr.OffsetInParent = DROffsetInParent; 3929 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3930 break; 3931 } 3932 case CVDR_DEFRANGE_REGISTER_REL: { 3933 int64_t DRRegister; 3934 int64_t DRFlags; 3935 int64_t DRBasePointerOffset; 3936 if (parseToken(AsmToken::Comma, "expected comma before register number in " 3937 ".cv_def_range directive") || 3938 parseAbsoluteExpression(DRRegister)) 3939 return Error(Loc, "expected register value"); 3940 if (parseToken( 3941 AsmToken::Comma, 3942 "expected comma before flag value in .cv_def_range directive") || 3943 parseAbsoluteExpression(DRFlags)) 3944 return Error(Loc, "expected flag value"); 3945 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset " 3946 "in .cv_def_range directive") || 3947 parseAbsoluteExpression(DRBasePointerOffset)) 3948 return Error(Loc, "expected base pointer offset value"); 3949 3950 codeview::DefRangeRegisterRelSym::Header DRHdr; 3951 DRHdr.Register = DRRegister; 3952 DRHdr.Flags = DRFlags; 3953 DRHdr.BasePointerOffset = DRBasePointerOffset; 3954 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3955 break; 3956 } 3957 default: 3958 return Error(Loc, "unexpected def_range type in .cv_def_range directive"); 3959 } 3960 return true; 3961 } 3962 3963 /// parseDirectiveCVString 3964 /// ::= .cv_stringtable "string" 3965 bool AsmParser::parseDirectiveCVString() { 3966 std::string Data; 3967 if (checkForValidSection() || parseEscapedString(Data)) 3968 return addErrorSuffix(" in '.cv_string' directive"); 3969 3970 // Put the string in the table and emit the offset. 3971 std::pair<StringRef, unsigned> Insertion = 3972 getCVContext().addToStringTable(Data); 3973 getStreamer().EmitIntValue(Insertion.second, 4); 3974 return false; 3975 } 3976 3977 /// parseDirectiveCVStringTable 3978 /// ::= .cv_stringtable 3979 bool AsmParser::parseDirectiveCVStringTable() { 3980 getStreamer().EmitCVStringTableDirective(); 3981 return false; 3982 } 3983 3984 /// parseDirectiveCVFileChecksums 3985 /// ::= .cv_filechecksums 3986 bool AsmParser::parseDirectiveCVFileChecksums() { 3987 getStreamer().EmitCVFileChecksumsDirective(); 3988 return false; 3989 } 3990 3991 /// parseDirectiveCVFileChecksumOffset 3992 /// ::= .cv_filechecksumoffset fileno 3993 bool AsmParser::parseDirectiveCVFileChecksumOffset() { 3994 int64_t FileNo; 3995 if (parseIntToken(FileNo, "expected identifier in directive")) 3996 return true; 3997 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 3998 return true; 3999 getStreamer().EmitCVFileChecksumOffsetDirective(FileNo); 4000 return false; 4001 } 4002 4003 /// parseDirectiveCVFPOData 4004 /// ::= .cv_fpo_data procsym 4005 bool AsmParser::parseDirectiveCVFPOData() { 4006 SMLoc DirLoc = getLexer().getLoc(); 4007 StringRef ProcName; 4008 if (parseIdentifier(ProcName)) 4009 return TokError("expected symbol name"); 4010 if (parseEOL("unexpected tokens")) 4011 return addErrorSuffix(" in '.cv_fpo_data' directive"); 4012 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName); 4013 getStreamer().EmitCVFPOData(ProcSym, DirLoc); 4014 return false; 4015 } 4016 4017 /// parseDirectiveCFISections 4018 /// ::= .cfi_sections section [, section] 4019 bool AsmParser::parseDirectiveCFISections() { 4020 StringRef Name; 4021 bool EH = false; 4022 bool Debug = false; 4023 4024 if (parseIdentifier(Name)) 4025 return TokError("Expected an identifier"); 4026 4027 if (Name == ".eh_frame") 4028 EH = true; 4029 else if (Name == ".debug_frame") 4030 Debug = true; 4031 4032 if (getLexer().is(AsmToken::Comma)) { 4033 Lex(); 4034 4035 if (parseIdentifier(Name)) 4036 return TokError("Expected an identifier"); 4037 4038 if (Name == ".eh_frame") 4039 EH = true; 4040 else if (Name == ".debug_frame") 4041 Debug = true; 4042 } 4043 4044 getStreamer().EmitCFISections(EH, Debug); 4045 return false; 4046 } 4047 4048 /// parseDirectiveCFIStartProc 4049 /// ::= .cfi_startproc [simple] 4050 bool AsmParser::parseDirectiveCFIStartProc() { 4051 StringRef Simple; 4052 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4053 if (check(parseIdentifier(Simple) || Simple != "simple", 4054 "unexpected token") || 4055 parseToken(AsmToken::EndOfStatement)) 4056 return addErrorSuffix(" in '.cfi_startproc' directive"); 4057 } 4058 4059 // TODO(kristina): Deal with a corner case of incorrect diagnostic context 4060 // being produced if this directive is emitted as part of preprocessor macro 4061 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. 4062 // Tools like llvm-mc on the other hand are not affected by it, and report 4063 // correct context information. 4064 getStreamer().EmitCFIStartProc(!Simple.empty(), Lexer.getLoc()); 4065 return false; 4066 } 4067 4068 /// parseDirectiveCFIEndProc 4069 /// ::= .cfi_endproc 4070 bool AsmParser::parseDirectiveCFIEndProc() { 4071 getStreamer().EmitCFIEndProc(); 4072 return false; 4073 } 4074 4075 /// parse register name or number. 4076 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register, 4077 SMLoc DirectiveLoc) { 4078 unsigned RegNo; 4079 4080 if (getLexer().isNot(AsmToken::Integer)) { 4081 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc)) 4082 return true; 4083 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true); 4084 } else 4085 return parseAbsoluteExpression(Register); 4086 4087 return false; 4088 } 4089 4090 /// parseDirectiveCFIDefCfa 4091 /// ::= .cfi_def_cfa register, offset 4092 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { 4093 int64_t Register = 0, Offset = 0; 4094 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4095 parseToken(AsmToken::Comma, "unexpected token in directive") || 4096 parseAbsoluteExpression(Offset)) 4097 return true; 4098 4099 getStreamer().EmitCFIDefCfa(Register, Offset); 4100 return false; 4101 } 4102 4103 /// parseDirectiveCFIDefCfaOffset 4104 /// ::= .cfi_def_cfa_offset offset 4105 bool AsmParser::parseDirectiveCFIDefCfaOffset() { 4106 int64_t Offset = 0; 4107 if (parseAbsoluteExpression(Offset)) 4108 return true; 4109 4110 getStreamer().EmitCFIDefCfaOffset(Offset); 4111 return false; 4112 } 4113 4114 /// parseDirectiveCFIRegister 4115 /// ::= .cfi_register register, register 4116 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { 4117 int64_t Register1 = 0, Register2 = 0; 4118 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || 4119 parseToken(AsmToken::Comma, "unexpected token in directive") || 4120 parseRegisterOrRegisterNumber(Register2, DirectiveLoc)) 4121 return true; 4122 4123 getStreamer().EmitCFIRegister(Register1, Register2); 4124 return false; 4125 } 4126 4127 /// parseDirectiveCFIWindowSave 4128 /// ::= .cfi_window_save 4129 bool AsmParser::parseDirectiveCFIWindowSave() { 4130 getStreamer().EmitCFIWindowSave(); 4131 return false; 4132 } 4133 4134 /// parseDirectiveCFIAdjustCfaOffset 4135 /// ::= .cfi_adjust_cfa_offset adjustment 4136 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() { 4137 int64_t Adjustment = 0; 4138 if (parseAbsoluteExpression(Adjustment)) 4139 return true; 4140 4141 getStreamer().EmitCFIAdjustCfaOffset(Adjustment); 4142 return false; 4143 } 4144 4145 /// parseDirectiveCFIDefCfaRegister 4146 /// ::= .cfi_def_cfa_register register 4147 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { 4148 int64_t Register = 0; 4149 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4150 return true; 4151 4152 getStreamer().EmitCFIDefCfaRegister(Register); 4153 return false; 4154 } 4155 4156 /// parseDirectiveCFIOffset 4157 /// ::= .cfi_offset register, offset 4158 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { 4159 int64_t Register = 0; 4160 int64_t Offset = 0; 4161 4162 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4163 parseToken(AsmToken::Comma, "unexpected token in directive") || 4164 parseAbsoluteExpression(Offset)) 4165 return true; 4166 4167 getStreamer().EmitCFIOffset(Register, Offset); 4168 return false; 4169 } 4170 4171 /// parseDirectiveCFIRelOffset 4172 /// ::= .cfi_rel_offset register, offset 4173 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { 4174 int64_t Register = 0, Offset = 0; 4175 4176 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4177 parseToken(AsmToken::Comma, "unexpected token in directive") || 4178 parseAbsoluteExpression(Offset)) 4179 return true; 4180 4181 getStreamer().EmitCFIRelOffset(Register, Offset); 4182 return false; 4183 } 4184 4185 static bool isValidEncoding(int64_t Encoding) { 4186 if (Encoding & ~0xff) 4187 return false; 4188 4189 if (Encoding == dwarf::DW_EH_PE_omit) 4190 return true; 4191 4192 const unsigned Format = Encoding & 0xf; 4193 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 4194 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 4195 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 4196 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 4197 return false; 4198 4199 const unsigned Application = Encoding & 0x70; 4200 if (Application != dwarf::DW_EH_PE_absptr && 4201 Application != dwarf::DW_EH_PE_pcrel) 4202 return false; 4203 4204 return true; 4205 } 4206 4207 /// parseDirectiveCFIPersonalityOrLsda 4208 /// IsPersonality true for cfi_personality, false for cfi_lsda 4209 /// ::= .cfi_personality encoding, [symbol_name] 4210 /// ::= .cfi_lsda encoding, [symbol_name] 4211 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { 4212 int64_t Encoding = 0; 4213 if (parseAbsoluteExpression(Encoding)) 4214 return true; 4215 if (Encoding == dwarf::DW_EH_PE_omit) 4216 return false; 4217 4218 StringRef Name; 4219 if (check(!isValidEncoding(Encoding), "unsupported encoding.") || 4220 parseToken(AsmToken::Comma, "unexpected token in directive") || 4221 check(parseIdentifier(Name), "expected identifier in directive")) 4222 return true; 4223 4224 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4225 4226 if (IsPersonality) 4227 getStreamer().EmitCFIPersonality(Sym, Encoding); 4228 else 4229 getStreamer().EmitCFILsda(Sym, Encoding); 4230 return false; 4231 } 4232 4233 /// parseDirectiveCFIRememberState 4234 /// ::= .cfi_remember_state 4235 bool AsmParser::parseDirectiveCFIRememberState() { 4236 getStreamer().EmitCFIRememberState(); 4237 return false; 4238 } 4239 4240 /// parseDirectiveCFIRestoreState 4241 /// ::= .cfi_remember_state 4242 bool AsmParser::parseDirectiveCFIRestoreState() { 4243 getStreamer().EmitCFIRestoreState(); 4244 return false; 4245 } 4246 4247 /// parseDirectiveCFISameValue 4248 /// ::= .cfi_same_value register 4249 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { 4250 int64_t Register = 0; 4251 4252 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4253 return true; 4254 4255 getStreamer().EmitCFISameValue(Register); 4256 return false; 4257 } 4258 4259 /// parseDirectiveCFIRestore 4260 /// ::= .cfi_restore register 4261 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { 4262 int64_t Register = 0; 4263 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4264 return true; 4265 4266 getStreamer().EmitCFIRestore(Register); 4267 return false; 4268 } 4269 4270 /// parseDirectiveCFIEscape 4271 /// ::= .cfi_escape expression[,...] 4272 bool AsmParser::parseDirectiveCFIEscape() { 4273 std::string Values; 4274 int64_t CurrValue; 4275 if (parseAbsoluteExpression(CurrValue)) 4276 return true; 4277 4278 Values.push_back((uint8_t)CurrValue); 4279 4280 while (getLexer().is(AsmToken::Comma)) { 4281 Lex(); 4282 4283 if (parseAbsoluteExpression(CurrValue)) 4284 return true; 4285 4286 Values.push_back((uint8_t)CurrValue); 4287 } 4288 4289 getStreamer().EmitCFIEscape(Values); 4290 return false; 4291 } 4292 4293 /// parseDirectiveCFIReturnColumn 4294 /// ::= .cfi_return_column register 4295 bool AsmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { 4296 int64_t Register = 0; 4297 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4298 return true; 4299 getStreamer().EmitCFIReturnColumn(Register); 4300 return false; 4301 } 4302 4303 /// parseDirectiveCFISignalFrame 4304 /// ::= .cfi_signal_frame 4305 bool AsmParser::parseDirectiveCFISignalFrame() { 4306 if (parseToken(AsmToken::EndOfStatement, 4307 "unexpected token in '.cfi_signal_frame'")) 4308 return true; 4309 4310 getStreamer().EmitCFISignalFrame(); 4311 return false; 4312 } 4313 4314 /// parseDirectiveCFIUndefined 4315 /// ::= .cfi_undefined register 4316 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { 4317 int64_t Register = 0; 4318 4319 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4320 return true; 4321 4322 getStreamer().EmitCFIUndefined(Register); 4323 return false; 4324 } 4325 4326 /// parseDirectiveAltmacro 4327 /// ::= .altmacro 4328 /// ::= .noaltmacro 4329 bool AsmParser::parseDirectiveAltmacro(StringRef Directive) { 4330 if (getLexer().isNot(AsmToken::EndOfStatement)) 4331 return TokError("unexpected token in '" + Directive + "' directive"); 4332 AltMacroMode = (Directive == ".altmacro"); 4333 return false; 4334 } 4335 4336 /// parseDirectiveMacrosOnOff 4337 /// ::= .macros_on 4338 /// ::= .macros_off 4339 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) { 4340 if (parseToken(AsmToken::EndOfStatement, 4341 "unexpected token in '" + Directive + "' directive")) 4342 return true; 4343 4344 setMacrosEnabled(Directive == ".macros_on"); 4345 return false; 4346 } 4347 4348 /// parseDirectiveMacro 4349 /// ::= .macro name[,] [parameters] 4350 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) { 4351 StringRef Name; 4352 if (parseIdentifier(Name)) 4353 return TokError("expected identifier in '.macro' directive"); 4354 4355 if (getLexer().is(AsmToken::Comma)) 4356 Lex(); 4357 4358 MCAsmMacroParameters Parameters; 4359 while (getLexer().isNot(AsmToken::EndOfStatement)) { 4360 4361 if (!Parameters.empty() && Parameters.back().Vararg) 4362 return Error(Lexer.getLoc(), 4363 "Vararg parameter '" + Parameters.back().Name + 4364 "' should be last one in the list of parameters."); 4365 4366 MCAsmMacroParameter Parameter; 4367 if (parseIdentifier(Parameter.Name)) 4368 return TokError("expected identifier in '.macro' directive"); 4369 4370 // Emit an error if two (or more) named parameters share the same name 4371 for (const MCAsmMacroParameter& CurrParam : Parameters) 4372 if (CurrParam.Name.equals(Parameter.Name)) 4373 return TokError("macro '" + Name + "' has multiple parameters" 4374 " named '" + Parameter.Name + "'"); 4375 4376 if (Lexer.is(AsmToken::Colon)) { 4377 Lex(); // consume ':' 4378 4379 SMLoc QualLoc; 4380 StringRef Qualifier; 4381 4382 QualLoc = Lexer.getLoc(); 4383 if (parseIdentifier(Qualifier)) 4384 return Error(QualLoc, "missing parameter qualifier for " 4385 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4386 4387 if (Qualifier == "req") 4388 Parameter.Required = true; 4389 else if (Qualifier == "vararg") 4390 Parameter.Vararg = true; 4391 else 4392 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier " 4393 "for '" + Parameter.Name + "' in macro '" + Name + "'"); 4394 } 4395 4396 if (getLexer().is(AsmToken::Equal)) { 4397 Lex(); 4398 4399 SMLoc ParamLoc; 4400 4401 ParamLoc = Lexer.getLoc(); 4402 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false )) 4403 return true; 4404 4405 if (Parameter.Required) 4406 Warning(ParamLoc, "pointless default value for required parameter " 4407 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4408 } 4409 4410 Parameters.push_back(std::move(Parameter)); 4411 4412 if (getLexer().is(AsmToken::Comma)) 4413 Lex(); 4414 } 4415 4416 // Eat just the end of statement. 4417 Lexer.Lex(); 4418 4419 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors 4420 AsmToken EndToken, StartToken = getTok(); 4421 unsigned MacroDepth = 0; 4422 // Lex the macro definition. 4423 while (true) { 4424 // Ignore Lexing errors in macros. 4425 while (Lexer.is(AsmToken::Error)) { 4426 Lexer.Lex(); 4427 } 4428 4429 // Check whether we have reached the end of the file. 4430 if (getLexer().is(AsmToken::Eof)) 4431 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 4432 4433 // Otherwise, check whether we have reach the .endmacro. 4434 if (getLexer().is(AsmToken::Identifier)) { 4435 if (getTok().getIdentifier() == ".endm" || 4436 getTok().getIdentifier() == ".endmacro") { 4437 if (MacroDepth == 0) { // Outermost macro. 4438 EndToken = getTok(); 4439 Lexer.Lex(); 4440 if (getLexer().isNot(AsmToken::EndOfStatement)) 4441 return TokError("unexpected token in '" + EndToken.getIdentifier() + 4442 "' directive"); 4443 break; 4444 } else { 4445 // Otherwise we just found the end of an inner macro. 4446 --MacroDepth; 4447 } 4448 } else if (getTok().getIdentifier() == ".macro") { 4449 // We allow nested macros. Those aren't instantiated until the outermost 4450 // macro is expanded so just ignore them for now. 4451 ++MacroDepth; 4452 } 4453 } 4454 4455 // Otherwise, scan til the end of the statement. 4456 eatToEndOfStatement(); 4457 } 4458 4459 if (getContext().lookupMacro(Name)) { 4460 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 4461 } 4462 4463 const char *BodyStart = StartToken.getLoc().getPointer(); 4464 const char *BodyEnd = EndToken.getLoc().getPointer(); 4465 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 4466 checkForBadMacro(DirectiveLoc, Name, Body, Parameters); 4467 MCAsmMacro Macro(Name, Body, std::move(Parameters)); 4468 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n"; 4469 Macro.dump()); 4470 getContext().defineMacro(Name, std::move(Macro)); 4471 return false; 4472 } 4473 4474 /// checkForBadMacro 4475 /// 4476 /// With the support added for named parameters there may be code out there that 4477 /// is transitioning from positional parameters. In versions of gas that did 4478 /// not support named parameters they would be ignored on the macro definition. 4479 /// But to support both styles of parameters this is not possible so if a macro 4480 /// definition has named parameters but does not use them and has what appears 4481 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a 4482 /// warning that the positional parameter found in body which have no effect. 4483 /// Hoping the developer will either remove the named parameters from the macro 4484 /// definition so the positional parameters get used if that was what was 4485 /// intended or change the macro to use the named parameters. It is possible 4486 /// this warning will trigger when the none of the named parameters are used 4487 /// and the strings like $1 are infact to simply to be passed trough unchanged. 4488 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, 4489 StringRef Body, 4490 ArrayRef<MCAsmMacroParameter> Parameters) { 4491 // If this macro is not defined with named parameters the warning we are 4492 // checking for here doesn't apply. 4493 unsigned NParameters = Parameters.size(); 4494 if (NParameters == 0) 4495 return; 4496 4497 bool NamedParametersFound = false; 4498 bool PositionalParametersFound = false; 4499 4500 // Look at the body of the macro for use of both the named parameters and what 4501 // are likely to be positional parameters. This is what expandMacro() is 4502 // doing when it finds the parameters in the body. 4503 while (!Body.empty()) { 4504 // Scan for the next possible parameter. 4505 std::size_t End = Body.size(), Pos = 0; 4506 for (; Pos != End; ++Pos) { 4507 // Check for a substitution or escape. 4508 // This macro is defined with parameters, look for \foo, \bar, etc. 4509 if (Body[Pos] == '\\' && Pos + 1 != End) 4510 break; 4511 4512 // This macro should have parameters, but look for $0, $1, ..., $n too. 4513 if (Body[Pos] != '$' || Pos + 1 == End) 4514 continue; 4515 char Next = Body[Pos + 1]; 4516 if (Next == '$' || Next == 'n' || 4517 isdigit(static_cast<unsigned char>(Next))) 4518 break; 4519 } 4520 4521 // Check if we reached the end. 4522 if (Pos == End) 4523 break; 4524 4525 if (Body[Pos] == '$') { 4526 switch (Body[Pos + 1]) { 4527 // $$ => $ 4528 case '$': 4529 break; 4530 4531 // $n => number of arguments 4532 case 'n': 4533 PositionalParametersFound = true; 4534 break; 4535 4536 // $[0-9] => argument 4537 default: { 4538 PositionalParametersFound = true; 4539 break; 4540 } 4541 } 4542 Pos += 2; 4543 } else { 4544 unsigned I = Pos + 1; 4545 while (isIdentifierChar(Body[I]) && I + 1 != End) 4546 ++I; 4547 4548 const char *Begin = Body.data() + Pos + 1; 4549 StringRef Argument(Begin, I - (Pos + 1)); 4550 unsigned Index = 0; 4551 for (; Index < NParameters; ++Index) 4552 if (Parameters[Index].Name == Argument) 4553 break; 4554 4555 if (Index == NParameters) { 4556 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 4557 Pos += 3; 4558 else { 4559 Pos = I; 4560 } 4561 } else { 4562 NamedParametersFound = true; 4563 Pos += 1 + Argument.size(); 4564 } 4565 } 4566 // Update the scan point. 4567 Body = Body.substr(Pos); 4568 } 4569 4570 if (!NamedParametersFound && PositionalParametersFound) 4571 Warning(DirectiveLoc, "macro defined with named parameters which are not " 4572 "used in macro body, possible positional parameter " 4573 "found in body which will have no effect"); 4574 } 4575 4576 /// parseDirectiveExitMacro 4577 /// ::= .exitm 4578 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) { 4579 if (parseToken(AsmToken::EndOfStatement, 4580 "unexpected token in '" + Directive + "' directive")) 4581 return true; 4582 4583 if (!isInsideMacroInstantiation()) 4584 return TokError("unexpected '" + Directive + "' in file, " 4585 "no current macro definition"); 4586 4587 // Exit all conditionals that are active in the current macro. 4588 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { 4589 TheCondState = TheCondStack.back(); 4590 TheCondStack.pop_back(); 4591 } 4592 4593 handleMacroExit(); 4594 return false; 4595 } 4596 4597 /// parseDirectiveEndMacro 4598 /// ::= .endm 4599 /// ::= .endmacro 4600 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) { 4601 if (getLexer().isNot(AsmToken::EndOfStatement)) 4602 return TokError("unexpected token in '" + Directive + "' directive"); 4603 4604 // If we are inside a macro instantiation, terminate the current 4605 // instantiation. 4606 if (isInsideMacroInstantiation()) { 4607 handleMacroExit(); 4608 return false; 4609 } 4610 4611 // Otherwise, this .endmacro is a stray entry in the file; well formed 4612 // .endmacro directives are handled during the macro definition parsing. 4613 return TokError("unexpected '" + Directive + "' in file, " 4614 "no current macro definition"); 4615 } 4616 4617 /// parseDirectivePurgeMacro 4618 /// ::= .purgem 4619 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { 4620 StringRef Name; 4621 SMLoc Loc; 4622 if (parseTokenLoc(Loc) || 4623 check(parseIdentifier(Name), Loc, 4624 "expected identifier in '.purgem' directive") || 4625 parseToken(AsmToken::EndOfStatement, 4626 "unexpected token in '.purgem' directive")) 4627 return true; 4628 4629 if (!getContext().lookupMacro(Name)) 4630 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 4631 4632 getContext().undefineMacro(Name); 4633 DEBUG_WITH_TYPE("asm-macros", dbgs() 4634 << "Un-defining macro: " << Name << "\n"); 4635 return false; 4636 } 4637 4638 /// parseDirectiveBundleAlignMode 4639 /// ::= {.bundle_align_mode} expression 4640 bool AsmParser::parseDirectiveBundleAlignMode() { 4641 // Expect a single argument: an expression that evaluates to a constant 4642 // in the inclusive range 0-30. 4643 SMLoc ExprLoc = getLexer().getLoc(); 4644 int64_t AlignSizePow2; 4645 if (checkForValidSection() || parseAbsoluteExpression(AlignSizePow2) || 4646 parseToken(AsmToken::EndOfStatement, "unexpected token after expression " 4647 "in '.bundle_align_mode' " 4648 "directive") || 4649 check(AlignSizePow2 < 0 || AlignSizePow2 > 30, ExprLoc, 4650 "invalid bundle alignment size (expected between 0 and 30)")) 4651 return true; 4652 4653 // Because of AlignSizePow2's verified range we can safely truncate it to 4654 // unsigned. 4655 getStreamer().EmitBundleAlignMode(static_cast<unsigned>(AlignSizePow2)); 4656 return false; 4657 } 4658 4659 /// parseDirectiveBundleLock 4660 /// ::= {.bundle_lock} [align_to_end] 4661 bool AsmParser::parseDirectiveBundleLock() { 4662 if (checkForValidSection()) 4663 return true; 4664 bool AlignToEnd = false; 4665 4666 StringRef Option; 4667 SMLoc Loc = getTok().getLoc(); 4668 const char *kInvalidOptionError = 4669 "invalid option for '.bundle_lock' directive"; 4670 4671 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4672 if (check(parseIdentifier(Option), Loc, kInvalidOptionError) || 4673 check(Option != "align_to_end", Loc, kInvalidOptionError) || 4674 parseToken(AsmToken::EndOfStatement, 4675 "unexpected token after '.bundle_lock' directive option")) 4676 return true; 4677 AlignToEnd = true; 4678 } 4679 4680 getStreamer().EmitBundleLock(AlignToEnd); 4681 return false; 4682 } 4683 4684 /// parseDirectiveBundleLock 4685 /// ::= {.bundle_lock} 4686 bool AsmParser::parseDirectiveBundleUnlock() { 4687 if (checkForValidSection() || 4688 parseToken(AsmToken::EndOfStatement, 4689 "unexpected token in '.bundle_unlock' directive")) 4690 return true; 4691 4692 getStreamer().EmitBundleUnlock(); 4693 return false; 4694 } 4695 4696 /// parseDirectiveSpace 4697 /// ::= (.skip | .space) expression [ , expression ] 4698 bool AsmParser::parseDirectiveSpace(StringRef IDVal) { 4699 SMLoc NumBytesLoc = Lexer.getLoc(); 4700 const MCExpr *NumBytes; 4701 if (checkForValidSection() || parseExpression(NumBytes)) 4702 return true; 4703 4704 int64_t FillExpr = 0; 4705 if (parseOptionalToken(AsmToken::Comma)) 4706 if (parseAbsoluteExpression(FillExpr)) 4707 return addErrorSuffix("in '" + Twine(IDVal) + "' directive"); 4708 if (parseToken(AsmToken::EndOfStatement)) 4709 return addErrorSuffix("in '" + Twine(IDVal) + "' directive"); 4710 4711 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 4712 getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc); 4713 4714 return false; 4715 } 4716 4717 /// parseDirectiveDCB 4718 /// ::= .dcb.{b, l, w} expression, expression 4719 bool AsmParser::parseDirectiveDCB(StringRef IDVal, unsigned Size) { 4720 SMLoc NumValuesLoc = Lexer.getLoc(); 4721 int64_t NumValues; 4722 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4723 return true; 4724 4725 if (NumValues < 0) { 4726 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4727 return false; 4728 } 4729 4730 if (parseToken(AsmToken::Comma, 4731 "unexpected token in '" + Twine(IDVal) + "' directive")) 4732 return true; 4733 4734 const MCExpr *Value; 4735 SMLoc ExprLoc = getLexer().getLoc(); 4736 if (parseExpression(Value)) 4737 return true; 4738 4739 // Special case constant expressions to match code generator. 4740 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4741 assert(Size <= 8 && "Invalid size"); 4742 uint64_t IntValue = MCE->getValue(); 4743 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 4744 return Error(ExprLoc, "literal value out of range for directive"); 4745 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4746 getStreamer().EmitIntValue(IntValue, Size); 4747 } else { 4748 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4749 getStreamer().EmitValue(Value, Size, ExprLoc); 4750 } 4751 4752 if (parseToken(AsmToken::EndOfStatement, 4753 "unexpected token in '" + Twine(IDVal) + "' directive")) 4754 return true; 4755 4756 return false; 4757 } 4758 4759 /// parseDirectiveRealDCB 4760 /// ::= .dcb.{d, s} expression, expression 4761 bool AsmParser::parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &Semantics) { 4762 SMLoc NumValuesLoc = Lexer.getLoc(); 4763 int64_t NumValues; 4764 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4765 return true; 4766 4767 if (NumValues < 0) { 4768 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4769 return false; 4770 } 4771 4772 if (parseToken(AsmToken::Comma, 4773 "unexpected token in '" + Twine(IDVal) + "' directive")) 4774 return true; 4775 4776 APInt AsInt; 4777 if (parseRealValue(Semantics, AsInt)) 4778 return true; 4779 4780 if (parseToken(AsmToken::EndOfStatement, 4781 "unexpected token in '" + Twine(IDVal) + "' directive")) 4782 return true; 4783 4784 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4785 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 4786 AsInt.getBitWidth() / 8); 4787 4788 return false; 4789 } 4790 4791 /// parseDirectiveDS 4792 /// ::= .ds.{b, d, l, p, s, w, x} expression 4793 bool AsmParser::parseDirectiveDS(StringRef IDVal, unsigned Size) { 4794 SMLoc NumValuesLoc = Lexer.getLoc(); 4795 int64_t NumValues; 4796 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4797 return true; 4798 4799 if (NumValues < 0) { 4800 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4801 return false; 4802 } 4803 4804 if (parseToken(AsmToken::EndOfStatement, 4805 "unexpected token in '" + Twine(IDVal) + "' directive")) 4806 return true; 4807 4808 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4809 getStreamer().emitFill(Size, 0); 4810 4811 return false; 4812 } 4813 4814 /// parseDirectiveLEB128 4815 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ] 4816 bool AsmParser::parseDirectiveLEB128(bool Signed) { 4817 if (checkForValidSection()) 4818 return true; 4819 4820 auto parseOp = [&]() -> bool { 4821 const MCExpr *Value; 4822 if (parseExpression(Value)) 4823 return true; 4824 if (Signed) 4825 getStreamer().EmitSLEB128Value(Value); 4826 else 4827 getStreamer().EmitULEB128Value(Value); 4828 return false; 4829 }; 4830 4831 if (parseMany(parseOp)) 4832 return addErrorSuffix(" in directive"); 4833 4834 return false; 4835 } 4836 4837 /// parseDirectiveSymbolAttribute 4838 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 4839 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 4840 auto parseOp = [&]() -> bool { 4841 StringRef Name; 4842 SMLoc Loc = getTok().getLoc(); 4843 if (parseIdentifier(Name)) 4844 return Error(Loc, "expected identifier"); 4845 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4846 4847 // Assembler local symbols don't make any sense here. Complain loudly. 4848 if (Sym->isTemporary()) 4849 return Error(Loc, "non-local symbol required"); 4850 4851 if (!getStreamer().EmitSymbolAttribute(Sym, Attr)) 4852 return Error(Loc, "unable to emit symbol attribute"); 4853 return false; 4854 }; 4855 4856 if (parseMany(parseOp)) 4857 return addErrorSuffix(" in directive"); 4858 return false; 4859 } 4860 4861 /// parseDirectiveComm 4862 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 4863 bool AsmParser::parseDirectiveComm(bool IsLocal) { 4864 if (checkForValidSection()) 4865 return true; 4866 4867 SMLoc IDLoc = getLexer().getLoc(); 4868 StringRef Name; 4869 if (parseIdentifier(Name)) 4870 return TokError("expected identifier in directive"); 4871 4872 // Handle the identifier as the key symbol. 4873 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4874 4875 if (getLexer().isNot(AsmToken::Comma)) 4876 return TokError("unexpected token in directive"); 4877 Lex(); 4878 4879 int64_t Size; 4880 SMLoc SizeLoc = getLexer().getLoc(); 4881 if (parseAbsoluteExpression(Size)) 4882 return true; 4883 4884 int64_t Pow2Alignment = 0; 4885 SMLoc Pow2AlignmentLoc; 4886 if (getLexer().is(AsmToken::Comma)) { 4887 Lex(); 4888 Pow2AlignmentLoc = getLexer().getLoc(); 4889 if (parseAbsoluteExpression(Pow2Alignment)) 4890 return true; 4891 4892 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); 4893 if (IsLocal && LCOMM == LCOMM::NoAlignment) 4894 return Error(Pow2AlignmentLoc, "alignment not supported on this target"); 4895 4896 // If this target takes alignments in bytes (not log) validate and convert. 4897 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || 4898 (IsLocal && LCOMM == LCOMM::ByteAlignment)) { 4899 if (!isPowerOf2_64(Pow2Alignment)) 4900 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 4901 Pow2Alignment = Log2_64(Pow2Alignment); 4902 } 4903 } 4904 4905 if (parseToken(AsmToken::EndOfStatement, 4906 "unexpected token in '.comm' or '.lcomm' directive")) 4907 return true; 4908 4909 // NOTE: a size of zero for a .comm should create a undefined symbol 4910 // but a size of .lcomm creates a bss symbol of size zero. 4911 if (Size < 0) 4912 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 4913 "be less than zero"); 4914 4915 // NOTE: The alignment in the directive is a power of 2 value, the assembler 4916 // may internally end up wanting an alignment in bytes. 4917 // FIXME: Diagnose overflow. 4918 if (Pow2Alignment < 0) 4919 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 4920 "alignment, can't be less than zero"); 4921 4922 Sym->redefineIfPossible(); 4923 if (!Sym->isUndefined()) 4924 return Error(IDLoc, "invalid symbol redefinition"); 4925 4926 // Create the Symbol as a common or local common with Size and Pow2Alignment 4927 if (IsLocal) { 4928 getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment); 4929 return false; 4930 } 4931 4932 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 4933 return false; 4934 } 4935 4936 /// parseDirectiveAbort 4937 /// ::= .abort [... message ...] 4938 bool AsmParser::parseDirectiveAbort() { 4939 // FIXME: Use loc from directive. 4940 SMLoc Loc = getLexer().getLoc(); 4941 4942 StringRef Str = parseStringToEndOfStatement(); 4943 if (parseToken(AsmToken::EndOfStatement, 4944 "unexpected token in '.abort' directive")) 4945 return true; 4946 4947 if (Str.empty()) 4948 return Error(Loc, ".abort detected. Assembly stopping."); 4949 else 4950 return Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 4951 // FIXME: Actually abort assembly here. 4952 4953 return false; 4954 } 4955 4956 /// parseDirectiveInclude 4957 /// ::= .include "filename" 4958 bool AsmParser::parseDirectiveInclude() { 4959 // Allow the strings to have escaped octal character sequence. 4960 std::string Filename; 4961 SMLoc IncludeLoc = getTok().getLoc(); 4962 4963 if (check(getTok().isNot(AsmToken::String), 4964 "expected string in '.include' directive") || 4965 parseEscapedString(Filename) || 4966 check(getTok().isNot(AsmToken::EndOfStatement), 4967 "unexpected token in '.include' directive") || 4968 // Attempt to switch the lexer to the included file before consuming the 4969 // end of statement to avoid losing it when we switch. 4970 check(enterIncludeFile(Filename), IncludeLoc, 4971 "Could not find include file '" + Filename + "'")) 4972 return true; 4973 4974 return false; 4975 } 4976 4977 /// parseDirectiveIncbin 4978 /// ::= .incbin "filename" [ , skip [ , count ] ] 4979 bool AsmParser::parseDirectiveIncbin() { 4980 // Allow the strings to have escaped octal character sequence. 4981 std::string Filename; 4982 SMLoc IncbinLoc = getTok().getLoc(); 4983 if (check(getTok().isNot(AsmToken::String), 4984 "expected string in '.incbin' directive") || 4985 parseEscapedString(Filename)) 4986 return true; 4987 4988 int64_t Skip = 0; 4989 const MCExpr *Count = nullptr; 4990 SMLoc SkipLoc, CountLoc; 4991 if (parseOptionalToken(AsmToken::Comma)) { 4992 // The skip expression can be omitted while specifying the count, e.g: 4993 // .incbin "filename",,4 4994 if (getTok().isNot(AsmToken::Comma)) { 4995 if (parseTokenLoc(SkipLoc) || parseAbsoluteExpression(Skip)) 4996 return true; 4997 } 4998 if (parseOptionalToken(AsmToken::Comma)) { 4999 CountLoc = getTok().getLoc(); 5000 if (parseExpression(Count)) 5001 return true; 5002 } 5003 } 5004 5005 if (parseToken(AsmToken::EndOfStatement, 5006 "unexpected token in '.incbin' directive")) 5007 return true; 5008 5009 if (check(Skip < 0, SkipLoc, "skip is negative")) 5010 return true; 5011 5012 // Attempt to process the included file. 5013 if (processIncbinFile(Filename, Skip, Count, CountLoc)) 5014 return Error(IncbinLoc, "Could not find incbin file '" + Filename + "'"); 5015 return false; 5016 } 5017 5018 /// parseDirectiveIf 5019 /// ::= .if{,eq,ge,gt,le,lt,ne} expression 5020 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { 5021 TheCondStack.push_back(TheCondState); 5022 TheCondState.TheCond = AsmCond::IfCond; 5023 if (TheCondState.Ignore) { 5024 eatToEndOfStatement(); 5025 } else { 5026 int64_t ExprValue; 5027 if (parseAbsoluteExpression(ExprValue) || 5028 parseToken(AsmToken::EndOfStatement, 5029 "unexpected token in '.if' directive")) 5030 return true; 5031 5032 switch (DirKind) { 5033 default: 5034 llvm_unreachable("unsupported directive"); 5035 case DK_IF: 5036 case DK_IFNE: 5037 break; 5038 case DK_IFEQ: 5039 ExprValue = ExprValue == 0; 5040 break; 5041 case DK_IFGE: 5042 ExprValue = ExprValue >= 0; 5043 break; 5044 case DK_IFGT: 5045 ExprValue = ExprValue > 0; 5046 break; 5047 case DK_IFLE: 5048 ExprValue = ExprValue <= 0; 5049 break; 5050 case DK_IFLT: 5051 ExprValue = ExprValue < 0; 5052 break; 5053 } 5054 5055 TheCondState.CondMet = ExprValue; 5056 TheCondState.Ignore = !TheCondState.CondMet; 5057 } 5058 5059 return false; 5060 } 5061 5062 /// parseDirectiveIfb 5063 /// ::= .ifb string 5064 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5065 TheCondStack.push_back(TheCondState); 5066 TheCondState.TheCond = AsmCond::IfCond; 5067 5068 if (TheCondState.Ignore) { 5069 eatToEndOfStatement(); 5070 } else { 5071 StringRef Str = parseStringToEndOfStatement(); 5072 5073 if (parseToken(AsmToken::EndOfStatement, 5074 "unexpected token in '.ifb' directive")) 5075 return true; 5076 5077 TheCondState.CondMet = ExpectBlank == Str.empty(); 5078 TheCondState.Ignore = !TheCondState.CondMet; 5079 } 5080 5081 return false; 5082 } 5083 5084 /// parseDirectiveIfc 5085 /// ::= .ifc string1, string2 5086 /// ::= .ifnc string1, string2 5087 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) { 5088 TheCondStack.push_back(TheCondState); 5089 TheCondState.TheCond = AsmCond::IfCond; 5090 5091 if (TheCondState.Ignore) { 5092 eatToEndOfStatement(); 5093 } else { 5094 StringRef Str1 = parseStringToComma(); 5095 5096 if (parseToken(AsmToken::Comma, "unexpected token in '.ifc' directive")) 5097 return true; 5098 5099 StringRef Str2 = parseStringToEndOfStatement(); 5100 5101 if (parseToken(AsmToken::EndOfStatement, 5102 "unexpected token in '.ifc' directive")) 5103 return true; 5104 5105 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim()); 5106 TheCondState.Ignore = !TheCondState.CondMet; 5107 } 5108 5109 return false; 5110 } 5111 5112 /// parseDirectiveIfeqs 5113 /// ::= .ifeqs string1, string2 5114 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) { 5115 if (Lexer.isNot(AsmToken::String)) { 5116 if (ExpectEqual) 5117 return TokError("expected string parameter for '.ifeqs' directive"); 5118 return TokError("expected string parameter for '.ifnes' directive"); 5119 } 5120 5121 StringRef String1 = getTok().getStringContents(); 5122 Lex(); 5123 5124 if (Lexer.isNot(AsmToken::Comma)) { 5125 if (ExpectEqual) 5126 return TokError( 5127 "expected comma after first string for '.ifeqs' directive"); 5128 return TokError("expected comma after first string for '.ifnes' directive"); 5129 } 5130 5131 Lex(); 5132 5133 if (Lexer.isNot(AsmToken::String)) { 5134 if (ExpectEqual) 5135 return TokError("expected string parameter for '.ifeqs' directive"); 5136 return TokError("expected string parameter for '.ifnes' directive"); 5137 } 5138 5139 StringRef String2 = getTok().getStringContents(); 5140 Lex(); 5141 5142 TheCondStack.push_back(TheCondState); 5143 TheCondState.TheCond = AsmCond::IfCond; 5144 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5145 TheCondState.Ignore = !TheCondState.CondMet; 5146 5147 return false; 5148 } 5149 5150 /// parseDirectiveIfdef 5151 /// ::= .ifdef symbol 5152 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 5153 StringRef Name; 5154 TheCondStack.push_back(TheCondState); 5155 TheCondState.TheCond = AsmCond::IfCond; 5156 5157 if (TheCondState.Ignore) { 5158 eatToEndOfStatement(); 5159 } else { 5160 if (check(parseIdentifier(Name), "expected identifier after '.ifdef'") || 5161 parseToken(AsmToken::EndOfStatement, "unexpected token in '.ifdef'")) 5162 return true; 5163 5164 MCSymbol *Sym = getContext().lookupSymbol(Name); 5165 5166 if (expect_defined) 5167 TheCondState.CondMet = (Sym && !Sym->isUndefined(false)); 5168 else 5169 TheCondState.CondMet = (!Sym || Sym->isUndefined(false)); 5170 TheCondState.Ignore = !TheCondState.CondMet; 5171 } 5172 5173 return false; 5174 } 5175 5176 /// parseDirectiveElseIf 5177 /// ::= .elseif expression 5178 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) { 5179 if (TheCondState.TheCond != AsmCond::IfCond && 5180 TheCondState.TheCond != AsmCond::ElseIfCond) 5181 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an" 5182 " .if or an .elseif"); 5183 TheCondState.TheCond = AsmCond::ElseIfCond; 5184 5185 bool LastIgnoreState = false; 5186 if (!TheCondStack.empty()) 5187 LastIgnoreState = TheCondStack.back().Ignore; 5188 if (LastIgnoreState || TheCondState.CondMet) { 5189 TheCondState.Ignore = true; 5190 eatToEndOfStatement(); 5191 } else { 5192 int64_t ExprValue; 5193 if (parseAbsoluteExpression(ExprValue)) 5194 return true; 5195 5196 if (parseToken(AsmToken::EndOfStatement, 5197 "unexpected token in '.elseif' directive")) 5198 return true; 5199 5200 TheCondState.CondMet = ExprValue; 5201 TheCondState.Ignore = !TheCondState.CondMet; 5202 } 5203 5204 return false; 5205 } 5206 5207 /// parseDirectiveElse 5208 /// ::= .else 5209 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) { 5210 if (parseToken(AsmToken::EndOfStatement, 5211 "unexpected token in '.else' directive")) 5212 return true; 5213 5214 if (TheCondState.TheCond != AsmCond::IfCond && 5215 TheCondState.TheCond != AsmCond::ElseIfCond) 5216 return Error(DirectiveLoc, "Encountered a .else that doesn't follow " 5217 " an .if or an .elseif"); 5218 TheCondState.TheCond = AsmCond::ElseCond; 5219 bool LastIgnoreState = false; 5220 if (!TheCondStack.empty()) 5221 LastIgnoreState = TheCondStack.back().Ignore; 5222 if (LastIgnoreState || TheCondState.CondMet) 5223 TheCondState.Ignore = true; 5224 else 5225 TheCondState.Ignore = false; 5226 5227 return false; 5228 } 5229 5230 /// parseDirectiveEnd 5231 /// ::= .end 5232 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { 5233 if (parseToken(AsmToken::EndOfStatement, 5234 "unexpected token in '.end' directive")) 5235 return true; 5236 5237 while (Lexer.isNot(AsmToken::Eof)) 5238 Lexer.Lex(); 5239 5240 return false; 5241 } 5242 5243 /// parseDirectiveError 5244 /// ::= .err 5245 /// ::= .error [string] 5246 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) { 5247 if (!TheCondStack.empty()) { 5248 if (TheCondStack.back().Ignore) { 5249 eatToEndOfStatement(); 5250 return false; 5251 } 5252 } 5253 5254 if (!WithMessage) 5255 return Error(L, ".err encountered"); 5256 5257 StringRef Message = ".error directive invoked in source file"; 5258 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5259 if (Lexer.isNot(AsmToken::String)) 5260 return TokError(".error argument must be a string"); 5261 5262 Message = getTok().getStringContents(); 5263 Lex(); 5264 } 5265 5266 return Error(L, Message); 5267 } 5268 5269 /// parseDirectiveWarning 5270 /// ::= .warning [string] 5271 bool AsmParser::parseDirectiveWarning(SMLoc L) { 5272 if (!TheCondStack.empty()) { 5273 if (TheCondStack.back().Ignore) { 5274 eatToEndOfStatement(); 5275 return false; 5276 } 5277 } 5278 5279 StringRef Message = ".warning directive invoked in source file"; 5280 5281 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 5282 if (Lexer.isNot(AsmToken::String)) 5283 return TokError(".warning argument must be a string"); 5284 5285 Message = getTok().getStringContents(); 5286 Lex(); 5287 if (parseToken(AsmToken::EndOfStatement, 5288 "expected end of statement in '.warning' directive")) 5289 return true; 5290 } 5291 5292 return Warning(L, Message); 5293 } 5294 5295 /// parseDirectiveEndIf 5296 /// ::= .endif 5297 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { 5298 if (parseToken(AsmToken::EndOfStatement, 5299 "unexpected token in '.endif' directive")) 5300 return true; 5301 5302 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) 5303 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow " 5304 "an .if or .else"); 5305 if (!TheCondStack.empty()) { 5306 TheCondState = TheCondStack.back(); 5307 TheCondStack.pop_back(); 5308 } 5309 5310 return false; 5311 } 5312 5313 void AsmParser::initializeDirectiveKindMap() { 5314 DirectiveKindMap[".set"] = DK_SET; 5315 DirectiveKindMap[".equ"] = DK_EQU; 5316 DirectiveKindMap[".equiv"] = DK_EQUIV; 5317 DirectiveKindMap[".ascii"] = DK_ASCII; 5318 DirectiveKindMap[".asciz"] = DK_ASCIZ; 5319 DirectiveKindMap[".string"] = DK_STRING; 5320 DirectiveKindMap[".byte"] = DK_BYTE; 5321 DirectiveKindMap[".short"] = DK_SHORT; 5322 DirectiveKindMap[".value"] = DK_VALUE; 5323 DirectiveKindMap[".2byte"] = DK_2BYTE; 5324 DirectiveKindMap[".long"] = DK_LONG; 5325 DirectiveKindMap[".int"] = DK_INT; 5326 DirectiveKindMap[".4byte"] = DK_4BYTE; 5327 DirectiveKindMap[".quad"] = DK_QUAD; 5328 DirectiveKindMap[".8byte"] = DK_8BYTE; 5329 DirectiveKindMap[".octa"] = DK_OCTA; 5330 DirectiveKindMap[".single"] = DK_SINGLE; 5331 DirectiveKindMap[".float"] = DK_FLOAT; 5332 DirectiveKindMap[".double"] = DK_DOUBLE; 5333 DirectiveKindMap[".align"] = DK_ALIGN; 5334 DirectiveKindMap[".align32"] = DK_ALIGN32; 5335 DirectiveKindMap[".balign"] = DK_BALIGN; 5336 DirectiveKindMap[".balignw"] = DK_BALIGNW; 5337 DirectiveKindMap[".balignl"] = DK_BALIGNL; 5338 DirectiveKindMap[".p2align"] = DK_P2ALIGN; 5339 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW; 5340 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL; 5341 DirectiveKindMap[".org"] = DK_ORG; 5342 DirectiveKindMap[".fill"] = DK_FILL; 5343 DirectiveKindMap[".zero"] = DK_ZERO; 5344 DirectiveKindMap[".extern"] = DK_EXTERN; 5345 DirectiveKindMap[".globl"] = DK_GLOBL; 5346 DirectiveKindMap[".global"] = DK_GLOBAL; 5347 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE; 5348 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP; 5349 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER; 5350 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN; 5351 DirectiveKindMap[".reference"] = DK_REFERENCE; 5352 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION; 5353 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE; 5354 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN; 5355 DirectiveKindMap[".cold"] = DK_COLD; 5356 DirectiveKindMap[".comm"] = DK_COMM; 5357 DirectiveKindMap[".common"] = DK_COMMON; 5358 DirectiveKindMap[".lcomm"] = DK_LCOMM; 5359 DirectiveKindMap[".abort"] = DK_ABORT; 5360 DirectiveKindMap[".include"] = DK_INCLUDE; 5361 DirectiveKindMap[".incbin"] = DK_INCBIN; 5362 DirectiveKindMap[".code16"] = DK_CODE16; 5363 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC; 5364 DirectiveKindMap[".rept"] = DK_REPT; 5365 DirectiveKindMap[".rep"] = DK_REPT; 5366 DirectiveKindMap[".irp"] = DK_IRP; 5367 DirectiveKindMap[".irpc"] = DK_IRPC; 5368 DirectiveKindMap[".endr"] = DK_ENDR; 5369 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE; 5370 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK; 5371 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK; 5372 DirectiveKindMap[".if"] = DK_IF; 5373 DirectiveKindMap[".ifeq"] = DK_IFEQ; 5374 DirectiveKindMap[".ifge"] = DK_IFGE; 5375 DirectiveKindMap[".ifgt"] = DK_IFGT; 5376 DirectiveKindMap[".ifle"] = DK_IFLE; 5377 DirectiveKindMap[".iflt"] = DK_IFLT; 5378 DirectiveKindMap[".ifne"] = DK_IFNE; 5379 DirectiveKindMap[".ifb"] = DK_IFB; 5380 DirectiveKindMap[".ifnb"] = DK_IFNB; 5381 DirectiveKindMap[".ifc"] = DK_IFC; 5382 DirectiveKindMap[".ifeqs"] = DK_IFEQS; 5383 DirectiveKindMap[".ifnc"] = DK_IFNC; 5384 DirectiveKindMap[".ifnes"] = DK_IFNES; 5385 DirectiveKindMap[".ifdef"] = DK_IFDEF; 5386 DirectiveKindMap[".ifndef"] = DK_IFNDEF; 5387 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF; 5388 DirectiveKindMap[".elseif"] = DK_ELSEIF; 5389 DirectiveKindMap[".else"] = DK_ELSE; 5390 DirectiveKindMap[".end"] = DK_END; 5391 DirectiveKindMap[".endif"] = DK_ENDIF; 5392 DirectiveKindMap[".skip"] = DK_SKIP; 5393 DirectiveKindMap[".space"] = DK_SPACE; 5394 DirectiveKindMap[".file"] = DK_FILE; 5395 DirectiveKindMap[".line"] = DK_LINE; 5396 DirectiveKindMap[".loc"] = DK_LOC; 5397 DirectiveKindMap[".stabs"] = DK_STABS; 5398 DirectiveKindMap[".cv_file"] = DK_CV_FILE; 5399 DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; 5400 DirectiveKindMap[".cv_loc"] = DK_CV_LOC; 5401 DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; 5402 DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; 5403 DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; 5404 DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; 5405 DirectiveKindMap[".cv_string"] = DK_CV_STRING; 5406 DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; 5407 DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; 5408 DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; 5409 DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; 5410 DirectiveKindMap[".sleb128"] = DK_SLEB128; 5411 DirectiveKindMap[".uleb128"] = DK_ULEB128; 5412 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; 5413 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; 5414 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; 5415 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; 5416 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; 5417 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; 5418 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; 5419 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; 5420 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; 5421 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; 5422 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; 5423 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; 5424 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; 5425 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; 5426 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; 5427 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; 5428 DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; 5429 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; 5430 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; 5431 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; 5432 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; 5433 DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; 5434 DirectiveKindMap[".macros_on"] = DK_MACROS_ON; 5435 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF; 5436 DirectiveKindMap[".macro"] = DK_MACRO; 5437 DirectiveKindMap[".exitm"] = DK_EXITM; 5438 DirectiveKindMap[".endm"] = DK_ENDM; 5439 DirectiveKindMap[".endmacro"] = DK_ENDMACRO; 5440 DirectiveKindMap[".purgem"] = DK_PURGEM; 5441 DirectiveKindMap[".err"] = DK_ERR; 5442 DirectiveKindMap[".error"] = DK_ERROR; 5443 DirectiveKindMap[".warning"] = DK_WARNING; 5444 DirectiveKindMap[".altmacro"] = DK_ALTMACRO; 5445 DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO; 5446 DirectiveKindMap[".reloc"] = DK_RELOC; 5447 DirectiveKindMap[".dc"] = DK_DC; 5448 DirectiveKindMap[".dc.a"] = DK_DC_A; 5449 DirectiveKindMap[".dc.b"] = DK_DC_B; 5450 DirectiveKindMap[".dc.d"] = DK_DC_D; 5451 DirectiveKindMap[".dc.l"] = DK_DC_L; 5452 DirectiveKindMap[".dc.s"] = DK_DC_S; 5453 DirectiveKindMap[".dc.w"] = DK_DC_W; 5454 DirectiveKindMap[".dc.x"] = DK_DC_X; 5455 DirectiveKindMap[".dcb"] = DK_DCB; 5456 DirectiveKindMap[".dcb.b"] = DK_DCB_B; 5457 DirectiveKindMap[".dcb.d"] = DK_DCB_D; 5458 DirectiveKindMap[".dcb.l"] = DK_DCB_L; 5459 DirectiveKindMap[".dcb.s"] = DK_DCB_S; 5460 DirectiveKindMap[".dcb.w"] = DK_DCB_W; 5461 DirectiveKindMap[".dcb.x"] = DK_DCB_X; 5462 DirectiveKindMap[".ds"] = DK_DS; 5463 DirectiveKindMap[".ds.b"] = DK_DS_B; 5464 DirectiveKindMap[".ds.d"] = DK_DS_D; 5465 DirectiveKindMap[".ds.l"] = DK_DS_L; 5466 DirectiveKindMap[".ds.p"] = DK_DS_P; 5467 DirectiveKindMap[".ds.s"] = DK_DS_S; 5468 DirectiveKindMap[".ds.w"] = DK_DS_W; 5469 DirectiveKindMap[".ds.x"] = DK_DS_X; 5470 DirectiveKindMap[".print"] = DK_PRINT; 5471 DirectiveKindMap[".addrsig"] = DK_ADDRSIG; 5472 DirectiveKindMap[".addrsig_sym"] = DK_ADDRSIG_SYM; 5473 } 5474 5475 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { 5476 AsmToken EndToken, StartToken = getTok(); 5477 5478 unsigned NestLevel = 0; 5479 while (true) { 5480 // Check whether we have reached the end of the file. 5481 if (getLexer().is(AsmToken::Eof)) { 5482 printError(DirectiveLoc, "no matching '.endr' in definition"); 5483 return nullptr; 5484 } 5485 5486 if (Lexer.is(AsmToken::Identifier) && 5487 (getTok().getIdentifier() == ".rep" || 5488 getTok().getIdentifier() == ".rept" || 5489 getTok().getIdentifier() == ".irp" || 5490 getTok().getIdentifier() == ".irpc")) { 5491 ++NestLevel; 5492 } 5493 5494 // Otherwise, check whether we have reached the .endr. 5495 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") { 5496 if (NestLevel == 0) { 5497 EndToken = getTok(); 5498 Lex(); 5499 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5500 printError(getTok().getLoc(), 5501 "unexpected token in '.endr' directive"); 5502 return nullptr; 5503 } 5504 break; 5505 } 5506 --NestLevel; 5507 } 5508 5509 // Otherwise, scan till the end of the statement. 5510 eatToEndOfStatement(); 5511 } 5512 5513 const char *BodyStart = StartToken.getLoc().getPointer(); 5514 const char *BodyEnd = EndToken.getLoc().getPointer(); 5515 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 5516 5517 // We Are Anonymous. 5518 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters()); 5519 return &MacroLikeBodies.back(); 5520 } 5521 5522 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 5523 raw_svector_ostream &OS) { 5524 OS << ".endr\n"; 5525 5526 std::unique_ptr<MemoryBuffer> Instantiation = 5527 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 5528 5529 // Create the macro instantiation object and add to the current macro 5530 // instantiation stack. 5531 MacroInstantiation *MI = new MacroInstantiation( 5532 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()); 5533 ActiveMacros.push_back(MI); 5534 5535 // Jump to the macro instantiation and prime the lexer. 5536 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 5537 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 5538 Lex(); 5539 } 5540 5541 /// parseDirectiveRept 5542 /// ::= .rep | .rept count 5543 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) { 5544 const MCExpr *CountExpr; 5545 SMLoc CountLoc = getTok().getLoc(); 5546 if (parseExpression(CountExpr)) 5547 return true; 5548 5549 int64_t Count; 5550 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) { 5551 return Error(CountLoc, "unexpected token in '" + Dir + "' directive"); 5552 } 5553 5554 if (check(Count < 0, CountLoc, "Count is negative") || 5555 parseToken(AsmToken::EndOfStatement, 5556 "unexpected token in '" + Dir + "' directive")) 5557 return true; 5558 5559 // Lex the rept definition. 5560 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5561 if (!M) 5562 return true; 5563 5564 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5565 // to hold the macro body with substitutions. 5566 SmallString<256> Buf; 5567 raw_svector_ostream OS(Buf); 5568 while (Count--) { 5569 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t). 5570 if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc())) 5571 return true; 5572 } 5573 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5574 5575 return false; 5576 } 5577 5578 /// parseDirectiveIrp 5579 /// ::= .irp symbol,values 5580 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) { 5581 MCAsmMacroParameter Parameter; 5582 MCAsmMacroArguments A; 5583 if (check(parseIdentifier(Parameter.Name), 5584 "expected identifier in '.irp' directive") || 5585 parseToken(AsmToken::Comma, "expected comma in '.irp' directive") || 5586 parseMacroArguments(nullptr, A) || 5587 parseToken(AsmToken::EndOfStatement, "expected End of Statement")) 5588 return true; 5589 5590 // Lex the irp definition. 5591 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5592 if (!M) 5593 return true; 5594 5595 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5596 // to hold the macro body with substitutions. 5597 SmallString<256> Buf; 5598 raw_svector_ostream OS(Buf); 5599 5600 for (const MCAsmMacroArgument &Arg : A) { 5601 // Note that the AtPseudoVariable is enabled for instantiations of .irp. 5602 // This is undocumented, but GAS seems to support it. 5603 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5604 return true; 5605 } 5606 5607 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5608 5609 return false; 5610 } 5611 5612 /// parseDirectiveIrpc 5613 /// ::= .irpc symbol,values 5614 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) { 5615 MCAsmMacroParameter Parameter; 5616 MCAsmMacroArguments A; 5617 5618 if (check(parseIdentifier(Parameter.Name), 5619 "expected identifier in '.irpc' directive") || 5620 parseToken(AsmToken::Comma, "expected comma in '.irpc' directive") || 5621 parseMacroArguments(nullptr, A)) 5622 return true; 5623 5624 if (A.size() != 1 || A.front().size() != 1) 5625 return TokError("unexpected token in '.irpc' directive"); 5626 5627 // Eat the end of statement. 5628 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 5629 return true; 5630 5631 // Lex the irpc definition. 5632 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5633 if (!M) 5634 return true; 5635 5636 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5637 // to hold the macro body with substitutions. 5638 SmallString<256> Buf; 5639 raw_svector_ostream OS(Buf); 5640 5641 StringRef Values = A.front().front().getString(); 5642 for (std::size_t I = 0, End = Values.size(); I != End; ++I) { 5643 MCAsmMacroArgument Arg; 5644 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1)); 5645 5646 // Note that the AtPseudoVariable is enabled for instantiations of .irpc. 5647 // This is undocumented, but GAS seems to support it. 5648 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5649 return true; 5650 } 5651 5652 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5653 5654 return false; 5655 } 5656 5657 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) { 5658 if (ActiveMacros.empty()) 5659 return TokError("unmatched '.endr' directive"); 5660 5661 // The only .repl that should get here are the ones created by 5662 // instantiateMacroLikeBody. 5663 assert(getLexer().is(AsmToken::EndOfStatement)); 5664 5665 handleMacroExit(); 5666 return false; 5667 } 5668 5669 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, 5670 size_t Len) { 5671 const MCExpr *Value; 5672 SMLoc ExprLoc = getLexer().getLoc(); 5673 if (parseExpression(Value)) 5674 return true; 5675 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5676 if (!MCE) 5677 return Error(ExprLoc, "unexpected expression in _emit"); 5678 uint64_t IntValue = MCE->getValue(); 5679 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue)) 5680 return Error(ExprLoc, "literal value out of range for directive"); 5681 5682 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len); 5683 return false; 5684 } 5685 5686 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { 5687 const MCExpr *Value; 5688 SMLoc ExprLoc = getLexer().getLoc(); 5689 if (parseExpression(Value)) 5690 return true; 5691 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5692 if (!MCE) 5693 return Error(ExprLoc, "unexpected expression in align"); 5694 uint64_t IntValue = MCE->getValue(); 5695 if (!isPowerOf2_64(IntValue)) 5696 return Error(ExprLoc, "literal value not a power of two greater then zero"); 5697 5698 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue)); 5699 return false; 5700 } 5701 5702 bool AsmParser::parseDirectivePrint(SMLoc DirectiveLoc) { 5703 const AsmToken StrTok = getTok(); 5704 Lex(); 5705 if (StrTok.isNot(AsmToken::String) || StrTok.getString().front() != '"') 5706 return Error(DirectiveLoc, "expected double quoted string after .print"); 5707 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 5708 return true; 5709 llvm::outs() << StrTok.getStringContents() << '\n'; 5710 return false; 5711 } 5712 5713 bool AsmParser::parseDirectiveAddrsig() { 5714 getStreamer().EmitAddrsig(); 5715 return false; 5716 } 5717 5718 bool AsmParser::parseDirectiveAddrsigSym() { 5719 StringRef Name; 5720 if (check(parseIdentifier(Name), 5721 "expected identifier in '.addrsig_sym' directive")) 5722 return true; 5723 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5724 getStreamer().EmitAddrsigSym(Sym); 5725 return false; 5726 } 5727 5728 // We are comparing pointers, but the pointers are relative to a single string. 5729 // Thus, this should always be deterministic. 5730 static int rewritesSort(const AsmRewrite *AsmRewriteA, 5731 const AsmRewrite *AsmRewriteB) { 5732 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) 5733 return -1; 5734 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) 5735 return 1; 5736 5737 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output 5738 // rewrite to the same location. Make sure the SizeDirective rewrite is 5739 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This 5740 // ensures the sort algorithm is stable. 5741 if (AsmRewritePrecedence[AsmRewriteA->Kind] > 5742 AsmRewritePrecedence[AsmRewriteB->Kind]) 5743 return -1; 5744 5745 if (AsmRewritePrecedence[AsmRewriteA->Kind] < 5746 AsmRewritePrecedence[AsmRewriteB->Kind]) 5747 return 1; 5748 llvm_unreachable("Unstable rewrite sort."); 5749 } 5750 5751 bool AsmParser::parseMSInlineAsm( 5752 void *AsmLoc, std::string &AsmString, unsigned &NumOutputs, 5753 unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 5754 SmallVectorImpl<std::string> &Constraints, 5755 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, 5756 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { 5757 SmallVector<void *, 4> InputDecls; 5758 SmallVector<void *, 4> OutputDecls; 5759 SmallVector<bool, 4> InputDeclsAddressOf; 5760 SmallVector<bool, 4> OutputDeclsAddressOf; 5761 SmallVector<std::string, 4> InputConstraints; 5762 SmallVector<std::string, 4> OutputConstraints; 5763 SmallVector<unsigned, 4> ClobberRegs; 5764 5765 SmallVector<AsmRewrite, 4> AsmStrRewrites; 5766 5767 // Prime the lexer. 5768 Lex(); 5769 5770 // While we have input, parse each statement. 5771 unsigned InputIdx = 0; 5772 unsigned OutputIdx = 0; 5773 while (getLexer().isNot(AsmToken::Eof)) { 5774 // Parse curly braces marking block start/end 5775 if (parseCurlyBlockScope(AsmStrRewrites)) 5776 continue; 5777 5778 ParseStatementInfo Info(&AsmStrRewrites); 5779 bool StatementErr = parseStatement(Info, &SI); 5780 5781 if (StatementErr || Info.ParseError) { 5782 // Emit pending errors if any exist. 5783 printPendingErrors(); 5784 return true; 5785 } 5786 5787 // No pending error should exist here. 5788 assert(!hasPendingError() && "unexpected error from parseStatement"); 5789 5790 if (Info.Opcode == ~0U) 5791 continue; 5792 5793 const MCInstrDesc &Desc = MII->get(Info.Opcode); 5794 5795 // Build the list of clobbers, outputs and inputs. 5796 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { 5797 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; 5798 5799 // Immediate. 5800 if (Operand.isImm()) 5801 continue; 5802 5803 // Register operand. 5804 if (Operand.isReg() && !Operand.needAddressOf() && 5805 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) { 5806 unsigned NumDefs = Desc.getNumDefs(); 5807 // Clobber. 5808 if (NumDefs && Operand.getMCOperandNum() < NumDefs) 5809 ClobberRegs.push_back(Operand.getReg()); 5810 continue; 5811 } 5812 5813 // Expr/Input or Output. 5814 StringRef SymName = Operand.getSymName(); 5815 if (SymName.empty()) 5816 continue; 5817 5818 void *OpDecl = Operand.getOpDecl(); 5819 if (!OpDecl) 5820 continue; 5821 5822 bool isOutput = (i == 1) && Desc.mayStore(); 5823 SMLoc Start = SMLoc::getFromPointer(SymName.data()); 5824 if (isOutput) { 5825 ++InputIdx; 5826 OutputDecls.push_back(OpDecl); 5827 OutputDeclsAddressOf.push_back(Operand.needAddressOf()); 5828 OutputConstraints.push_back(("=" + Operand.getConstraint()).str()); 5829 AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size()); 5830 } else { 5831 InputDecls.push_back(OpDecl); 5832 InputDeclsAddressOf.push_back(Operand.needAddressOf()); 5833 InputConstraints.push_back(Operand.getConstraint().str()); 5834 AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size()); 5835 } 5836 } 5837 5838 // Consider implicit defs to be clobbers. Think of cpuid and push. 5839 ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(), 5840 Desc.getNumImplicitDefs()); 5841 ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end()); 5842 } 5843 5844 // Set the number of Outputs and Inputs. 5845 NumOutputs = OutputDecls.size(); 5846 NumInputs = InputDecls.size(); 5847 5848 // Set the unique clobbers. 5849 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end()); 5850 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()), 5851 ClobberRegs.end()); 5852 Clobbers.assign(ClobberRegs.size(), std::string()); 5853 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { 5854 raw_string_ostream OS(Clobbers[I]); 5855 IP->printRegName(OS, ClobberRegs[I]); 5856 } 5857 5858 // Merge the various outputs and inputs. Output are expected first. 5859 if (NumOutputs || NumInputs) { 5860 unsigned NumExprs = NumOutputs + NumInputs; 5861 OpDecls.resize(NumExprs); 5862 Constraints.resize(NumExprs); 5863 for (unsigned i = 0; i < NumOutputs; ++i) { 5864 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]); 5865 Constraints[i] = OutputConstraints[i]; 5866 } 5867 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { 5868 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]); 5869 Constraints[j] = InputConstraints[i]; 5870 } 5871 } 5872 5873 // Build the IR assembly string. 5874 std::string AsmStringIR; 5875 raw_string_ostream OS(AsmStringIR); 5876 StringRef ASMString = 5877 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer(); 5878 const char *AsmStart = ASMString.begin(); 5879 const char *AsmEnd = ASMString.end(); 5880 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort); 5881 for (const AsmRewrite &AR : AsmStrRewrites) { 5882 AsmRewriteKind Kind = AR.Kind; 5883 5884 const char *Loc = AR.Loc.getPointer(); 5885 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!"); 5886 5887 // Emit everything up to the immediate/expression. 5888 if (unsigned Len = Loc - AsmStart) 5889 OS << StringRef(AsmStart, Len); 5890 5891 // Skip the original expression. 5892 if (Kind == AOK_Skip) { 5893 AsmStart = Loc + AR.Len; 5894 continue; 5895 } 5896 5897 unsigned AdditionalSkip = 0; 5898 // Rewrite expressions in $N notation. 5899 switch (Kind) { 5900 default: 5901 break; 5902 case AOK_IntelExpr: 5903 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression"); 5904 if (AR.IntelExp.NeedBracs) 5905 OS << "["; 5906 if (AR.IntelExp.hasBaseReg()) 5907 OS << AR.IntelExp.BaseReg; 5908 if (AR.IntelExp.hasIndexReg()) 5909 OS << (AR.IntelExp.hasBaseReg() ? " + " : "") 5910 << AR.IntelExp.IndexReg; 5911 if (AR.IntelExp.Scale > 1) 5912 OS << " * $$" << AR.IntelExp.Scale; 5913 if (AR.IntelExp.Imm || !AR.IntelExp.hasRegs()) 5914 OS << (AR.IntelExp.hasRegs() ? " + $$" : "$$") << AR.IntelExp.Imm; 5915 if (AR.IntelExp.NeedBracs) 5916 OS << "]"; 5917 break; 5918 case AOK_Label: 5919 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; 5920 break; 5921 case AOK_Input: 5922 OS << '$' << InputIdx++; 5923 break; 5924 case AOK_Output: 5925 OS << '$' << OutputIdx++; 5926 break; 5927 case AOK_SizeDirective: 5928 switch (AR.Val) { 5929 default: break; 5930 case 8: OS << "byte ptr "; break; 5931 case 16: OS << "word ptr "; break; 5932 case 32: OS << "dword ptr "; break; 5933 case 64: OS << "qword ptr "; break; 5934 case 80: OS << "xword ptr "; break; 5935 case 128: OS << "xmmword ptr "; break; 5936 case 256: OS << "ymmword ptr "; break; 5937 } 5938 break; 5939 case AOK_Emit: 5940 OS << ".byte"; 5941 break; 5942 case AOK_Align: { 5943 // MS alignment directives are measured in bytes. If the native assembler 5944 // measures alignment in bytes, we can pass it straight through. 5945 OS << ".align"; 5946 if (getContext().getAsmInfo()->getAlignmentIsInBytes()) 5947 break; 5948 5949 // Alignment is in log2 form, so print that instead and skip the original 5950 // immediate. 5951 unsigned Val = AR.Val; 5952 OS << ' ' << Val; 5953 assert(Val < 10 && "Expected alignment less then 2^10."); 5954 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; 5955 break; 5956 } 5957 case AOK_EVEN: 5958 OS << ".even"; 5959 break; 5960 case AOK_EndOfStatement: 5961 OS << "\n\t"; 5962 break; 5963 } 5964 5965 // Skip the original expression. 5966 AsmStart = Loc + AR.Len + AdditionalSkip; 5967 } 5968 5969 // Emit the remainder of the asm string. 5970 if (AsmStart != AsmEnd) 5971 OS << StringRef(AsmStart, AsmEnd - AsmStart); 5972 5973 AsmString = OS.str(); 5974 return false; 5975 } 5976 5977 namespace llvm { 5978 namespace MCParserUtils { 5979 5980 /// Returns whether the given symbol is used anywhere in the given expression, 5981 /// or subexpressions. 5982 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) { 5983 switch (Value->getKind()) { 5984 case MCExpr::Binary: { 5985 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value); 5986 return isSymbolUsedInExpression(Sym, BE->getLHS()) || 5987 isSymbolUsedInExpression(Sym, BE->getRHS()); 5988 } 5989 case MCExpr::Target: 5990 case MCExpr::Constant: 5991 return false; 5992 case MCExpr::SymbolRef: { 5993 const MCSymbol &S = 5994 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol(); 5995 if (S.isVariable()) 5996 return isSymbolUsedInExpression(Sym, S.getVariableValue()); 5997 return &S == Sym; 5998 } 5999 case MCExpr::Unary: 6000 return isSymbolUsedInExpression( 6001 Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr()); 6002 } 6003 6004 llvm_unreachable("Unknown expr kind!"); 6005 } 6006 6007 bool parseAssignmentExpression(StringRef Name, bool allow_redef, 6008 MCAsmParser &Parser, MCSymbol *&Sym, 6009 const MCExpr *&Value) { 6010 6011 // FIXME: Use better location, we should use proper tokens. 6012 SMLoc EqualLoc = Parser.getTok().getLoc(); 6013 if (Parser.parseExpression(Value)) 6014 return Parser.TokError("missing expression"); 6015 6016 // Note: we don't count b as used in "a = b". This is to allow 6017 // a = b 6018 // b = c 6019 6020 if (Parser.parseToken(AsmToken::EndOfStatement)) 6021 return true; 6022 6023 // Validate that the LHS is allowed to be a variable (either it has not been 6024 // used as a symbol, or it is an absolute symbol). 6025 Sym = Parser.getContext().lookupSymbol(Name); 6026 if (Sym) { 6027 // Diagnose assignment to a label. 6028 // 6029 // FIXME: Diagnostics. Note the location of the definition as a label. 6030 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 6031 if (isSymbolUsedInExpression(Sym, Value)) 6032 return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'"); 6033 else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() && 6034 !Sym->isVariable()) 6035 ; // Allow redefinitions of undefined symbols only used in directives. 6036 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef) 6037 ; // Allow redefinitions of variables that haven't yet been used. 6038 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 6039 return Parser.Error(EqualLoc, "redefinition of '" + Name + "'"); 6040 else if (!Sym->isVariable()) 6041 return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'"); 6042 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 6043 return Parser.Error(EqualLoc, 6044 "invalid reassignment of non-absolute variable '" + 6045 Name + "'"); 6046 } else if (Name == ".") { 6047 Parser.getStreamer().emitValueToOffset(Value, 0, EqualLoc); 6048 return false; 6049 } else 6050 Sym = Parser.getContext().getOrCreateSymbol(Name); 6051 6052 Sym->setRedefinable(allow_redef); 6053 6054 return false; 6055 } 6056 6057 } // end namespace MCParserUtils 6058 } // end namespace llvm 6059 6060 /// Create an MCAsmParser instance. 6061 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C, 6062 MCStreamer &Out, const MCAsmInfo &MAI, 6063 unsigned CB) { 6064 return new AsmParser(SM, C, Out, MAI, CB); 6065 } 6066