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 a parser for assembly files similar to gas syntax. 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/STLExtras.h" 17 #include "llvm/ADT/SmallSet.h" 18 #include "llvm/ADT/SmallString.h" 19 #include "llvm/ADT/SmallVector.h" 20 #include "llvm/ADT/StringExtras.h" 21 #include "llvm/ADT/StringMap.h" 22 #include "llvm/ADT/StringRef.h" 23 #include "llvm/ADT/Twine.h" 24 #include "llvm/BinaryFormat/Dwarf.h" 25 #include "llvm/DebugInfo/CodeView/SymbolRecord.h" 26 #include "llvm/MC/MCAsmInfo.h" 27 #include "llvm/MC/MCCodeView.h" 28 #include "llvm/MC/MCContext.h" 29 #include "llvm/MC/MCDirectives.h" 30 #include "llvm/MC/MCDwarf.h" 31 #include "llvm/MC/MCExpr.h" 32 #include "llvm/MC/MCInstPrinter.h" 33 #include "llvm/MC/MCInstrDesc.h" 34 #include "llvm/MC/MCInstrInfo.h" 35 #include "llvm/MC/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 <optional> 67 #include <sstream> 68 #include <string> 69 #include <tuple> 70 #include <utility> 71 #include <vector> 72 73 using namespace llvm; 74 75 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() = default; 76 77 namespace { 78 79 /// Helper types for tracking macro definitions. 80 typedef std::vector<AsmToken> MCAsmMacroArgument; 81 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; 82 83 /// Helper class for storing information about an active macro 84 /// instantiation. 85 struct MacroInstantiation { 86 /// The location of the instantiation. 87 SMLoc InstantiationLoc; 88 89 /// The buffer where parsing should resume upon instantiation completion. 90 unsigned ExitBuffer; 91 92 /// The location where parsing should resume upon instantiation completion. 93 SMLoc ExitLoc; 94 95 /// The depth of TheCondStack at the start of the instantiation. 96 size_t CondStackDepth; 97 }; 98 99 struct ParseStatementInfo { 100 /// The parsed operands from the last parsed statement. 101 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; 102 103 /// The opcode from the last parsed instruction. 104 unsigned Opcode = ~0U; 105 106 /// Was there an error parsing the inline assembly? 107 bool ParseError = false; 108 109 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr; 110 111 ParseStatementInfo() = delete; 112 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites) 113 : AsmRewrites(rewrites) {} 114 }; 115 116 /// The concrete assembly parser instance. 117 class AsmParser : public MCAsmParser { 118 private: 119 AsmLexer Lexer; 120 MCContext &Ctx; 121 MCStreamer &Out; 122 const MCAsmInfo &MAI; 123 SourceMgr &SrcMgr; 124 SourceMgr::DiagHandlerTy SavedDiagHandler; 125 void *SavedDiagContext; 126 std::unique_ptr<MCAsmParserExtension> PlatformParser; 127 SMLoc StartTokLoc; 128 std::optional<SMLoc> CFIStartProcLoc; 129 130 /// This is the current buffer index we're lexing from as managed by the 131 /// SourceMgr object. 132 unsigned CurBuffer; 133 134 AsmCond TheCondState; 135 std::vector<AsmCond> TheCondStack; 136 137 /// maps directive names to handler methods in parser 138 /// extensions. Extensions register themselves in this map by calling 139 /// addDirectiveHandler. 140 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap; 141 142 /// Stack of active macro instantiations. 143 std::vector<MacroInstantiation*> ActiveMacros; 144 145 /// List of bodies of anonymous macros. 146 std::deque<MCAsmMacro> MacroLikeBodies; 147 148 /// Boolean tracking whether macro substitution is enabled. 149 unsigned MacrosEnabledFlag : 1; 150 151 /// Keeps track of how many .macro's have been instantiated. 152 unsigned NumOfMacroInstantiations; 153 154 /// The values from the last parsed cpp hash file line comment if any. 155 struct CppHashInfoTy { 156 StringRef Filename; 157 int64_t LineNumber; 158 SMLoc Loc; 159 unsigned Buf; 160 CppHashInfoTy() : LineNumber(0), Buf(0) {} 161 }; 162 CppHashInfoTy CppHashInfo; 163 164 /// The filename from the first cpp hash file line comment, if any. 165 StringRef FirstCppHashFilename; 166 167 /// List of forward directional labels for diagnosis at the end. 168 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; 169 170 SmallSet<StringRef, 2> LTODiscardSymbols; 171 172 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 173 unsigned AssemblerDialect = ~0U; 174 175 /// is Darwin compatibility enabled? 176 bool IsDarwin = false; 177 178 /// Are we parsing ms-style inline assembly? 179 bool ParsingMSInlineAsm = false; 180 181 /// Did we already inform the user about inconsistent MD5 usage? 182 bool ReportedInconsistentMD5 = false; 183 184 // Is alt macro mode enabled. 185 bool AltMacroMode = false; 186 187 protected: 188 virtual bool parseStatement(ParseStatementInfo &Info, 189 MCAsmParserSemaCallback *SI); 190 191 /// This routine uses the target specific ParseInstruction function to 192 /// parse an instruction into Operands, and then call the target specific 193 /// MatchAndEmit function to match and emit the instruction. 194 bool parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info, 195 StringRef IDVal, AsmToken ID, 196 SMLoc IDLoc); 197 198 /// Should we emit DWARF describing this assembler source? (Returns false if 199 /// the source has .file directives, which means we don't want to generate 200 /// info describing the assembler source itself.) 201 bool enabledGenDwarfForAssembly(); 202 203 public: 204 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 205 const MCAsmInfo &MAI, unsigned CB); 206 AsmParser(const AsmParser &) = delete; 207 AsmParser &operator=(const AsmParser &) = delete; 208 ~AsmParser() override; 209 210 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override; 211 212 void addDirectiveHandler(StringRef Directive, 213 ExtensionDirectiveHandler Handler) override { 214 ExtensionDirectiveMap[Directive] = Handler; 215 } 216 217 void addAliasForDirective(StringRef Directive, StringRef Alias) override { 218 DirectiveKindMap[Directive.lower()] = DirectiveKindMap[Alias.lower()]; 219 } 220 221 /// @name MCAsmParser Interface 222 /// { 223 224 SourceMgr &getSourceManager() override { return SrcMgr; } 225 MCAsmLexer &getLexer() override { return Lexer; } 226 MCContext &getContext() override { return Ctx; } 227 MCStreamer &getStreamer() override { return Out; } 228 229 CodeViewContext &getCVContext() { return Ctx.getCVContext(); } 230 231 unsigned getAssemblerDialect() override { 232 if (AssemblerDialect == ~0U) 233 return MAI.getAssemblerDialect(); 234 else 235 return AssemblerDialect; 236 } 237 void setAssemblerDialect(unsigned i) override { 238 AssemblerDialect = i; 239 } 240 241 void Note(SMLoc L, const Twine &Msg, SMRange Range = std::nullopt) override; 242 bool Warning(SMLoc L, const Twine &Msg, 243 SMRange Range = std::nullopt) override; 244 bool printError(SMLoc L, const Twine &Msg, 245 SMRange Range = std::nullopt) override; 246 247 const AsmToken &Lex() override; 248 249 void setParsingMSInlineAsm(bool V) override { 250 ParsingMSInlineAsm = V; 251 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and 252 // hex integer literals. 253 Lexer.setLexMasmIntegers(V); 254 } 255 bool isParsingMSInlineAsm() override { return ParsingMSInlineAsm; } 256 257 bool discardLTOSymbol(StringRef Name) const override { 258 return LTODiscardSymbols.contains(Name); 259 } 260 261 bool parseMSInlineAsm(std::string &AsmString, unsigned &NumOutputs, 262 unsigned &NumInputs, 263 SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 264 SmallVectorImpl<std::string> &Constraints, 265 SmallVectorImpl<std::string> &Clobbers, 266 const MCInstrInfo *MII, const MCInstPrinter *IP, 267 MCAsmParserSemaCallback &SI) override; 268 269 bool parseExpression(const MCExpr *&Res); 270 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 271 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc, 272 AsmTypeInfo *TypeInfo) override; 273 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 274 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 275 SMLoc &EndLoc) override; 276 bool parseAbsoluteExpression(int64_t &Res) override; 277 278 /// Parse a floating point expression using the float \p Semantics 279 /// and set \p Res to the value. 280 bool parseRealValue(const fltSemantics &Semantics, APInt &Res); 281 282 /// Parse an identifier or string (as a quoted identifier) 283 /// and set \p Res to the identifier contents. 284 bool parseIdentifier(StringRef &Res) override; 285 void eatToEndOfStatement() override; 286 287 bool checkForValidSection() override; 288 289 /// } 290 291 private: 292 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); 293 bool parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo = true); 294 295 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body, 296 ArrayRef<MCAsmMacroParameter> Parameters); 297 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 298 ArrayRef<MCAsmMacroParameter> Parameters, 299 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable, 300 SMLoc L); 301 302 /// Are macros enabled in the parser? 303 bool areMacrosEnabled() {return MacrosEnabledFlag;} 304 305 /// Control a flag in the parser that enables or disables macros. 306 void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;} 307 308 /// Are we inside a macro instantiation? 309 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();} 310 311 /// Handle entry to macro instantiation. 312 /// 313 /// \param M The macro. 314 /// \param NameLoc Instantiation location. 315 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc); 316 317 /// Handle exit from macro instantiation. 318 void handleMacroExit(); 319 320 /// Extract AsmTokens for a macro argument. 321 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg); 322 323 /// Parse all macro arguments for a given macro. 324 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A); 325 326 void printMacroInstantiations(); 327 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 328 SMRange Range = std::nullopt) const { 329 ArrayRef<SMRange> Ranges(Range); 330 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 331 } 332 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 333 334 /// Enter the specified file. This returns true on failure. 335 bool enterIncludeFile(const std::string &Filename); 336 337 /// Process the specified file for the .incbin directive. 338 /// This returns true on failure. 339 bool processIncbinFile(const std::string &Filename, int64_t Skip = 0, 340 const MCExpr *Count = nullptr, SMLoc Loc = SMLoc()); 341 342 /// Reset the current lexer position to that given by \p Loc. The 343 /// current token is not set; clients should ensure Lex() is called 344 /// subsequently. 345 /// 346 /// \param InBuffer If not 0, should be the known buffer id that contains the 347 /// location. 348 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0); 349 350 /// Parse up to the end of statement and a return the contents from the 351 /// current token until the end of the statement; the current token on exit 352 /// will be either the EndOfStatement or EOF. 353 StringRef parseStringToEndOfStatement() override; 354 355 /// Parse until the end of a statement or a comma is encountered, 356 /// return the contents from the current token up to the end or comma. 357 StringRef parseStringToComma(); 358 359 enum class AssignmentKind { 360 Set, 361 Equiv, 362 Equal, 363 LTOSetConditional, 364 }; 365 366 bool parseAssignment(StringRef Name, AssignmentKind Kind); 367 368 unsigned getBinOpPrecedence(AsmToken::TokenKind K, 369 MCBinaryExpr::Opcode &Kind); 370 371 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 372 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 373 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 374 375 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 376 377 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName); 378 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName); 379 380 // Generic (target and platform independent) directive parsing. 381 enum DirectiveKind { 382 DK_NO_DIRECTIVE, // Placeholder 383 DK_SET, 384 DK_EQU, 385 DK_EQUIV, 386 DK_ASCII, 387 DK_ASCIZ, 388 DK_STRING, 389 DK_BYTE, 390 DK_SHORT, 391 DK_RELOC, 392 DK_VALUE, 393 DK_2BYTE, 394 DK_LONG, 395 DK_INT, 396 DK_4BYTE, 397 DK_QUAD, 398 DK_8BYTE, 399 DK_OCTA, 400 DK_DC, 401 DK_DC_A, 402 DK_DC_B, 403 DK_DC_D, 404 DK_DC_L, 405 DK_DC_S, 406 DK_DC_W, 407 DK_DC_X, 408 DK_DCB, 409 DK_DCB_B, 410 DK_DCB_D, 411 DK_DCB_L, 412 DK_DCB_S, 413 DK_DCB_W, 414 DK_DCB_X, 415 DK_DS, 416 DK_DS_B, 417 DK_DS_D, 418 DK_DS_L, 419 DK_DS_P, 420 DK_DS_S, 421 DK_DS_W, 422 DK_DS_X, 423 DK_SINGLE, 424 DK_FLOAT, 425 DK_DOUBLE, 426 DK_ALIGN, 427 DK_ALIGN32, 428 DK_BALIGN, 429 DK_BALIGNW, 430 DK_BALIGNL, 431 DK_P2ALIGN, 432 DK_P2ALIGNW, 433 DK_P2ALIGNL, 434 DK_ORG, 435 DK_FILL, 436 DK_ENDR, 437 DK_BUNDLE_ALIGN_MODE, 438 DK_BUNDLE_LOCK, 439 DK_BUNDLE_UNLOCK, 440 DK_ZERO, 441 DK_EXTERN, 442 DK_GLOBL, 443 DK_GLOBAL, 444 DK_LAZY_REFERENCE, 445 DK_NO_DEAD_STRIP, 446 DK_SYMBOL_RESOLVER, 447 DK_PRIVATE_EXTERN, 448 DK_REFERENCE, 449 DK_WEAK_DEFINITION, 450 DK_WEAK_REFERENCE, 451 DK_WEAK_DEF_CAN_BE_HIDDEN, 452 DK_COLD, 453 DK_COMM, 454 DK_COMMON, 455 DK_LCOMM, 456 DK_ABORT, 457 DK_INCLUDE, 458 DK_INCBIN, 459 DK_CODE16, 460 DK_CODE16GCC, 461 DK_REPT, 462 DK_IRP, 463 DK_IRPC, 464 DK_IF, 465 DK_IFEQ, 466 DK_IFGE, 467 DK_IFGT, 468 DK_IFLE, 469 DK_IFLT, 470 DK_IFNE, 471 DK_IFB, 472 DK_IFNB, 473 DK_IFC, 474 DK_IFEQS, 475 DK_IFNC, 476 DK_IFNES, 477 DK_IFDEF, 478 DK_IFNDEF, 479 DK_IFNOTDEF, 480 DK_ELSEIF, 481 DK_ELSE, 482 DK_ENDIF, 483 DK_SPACE, 484 DK_SKIP, 485 DK_FILE, 486 DK_LINE, 487 DK_LOC, 488 DK_STABS, 489 DK_CV_FILE, 490 DK_CV_FUNC_ID, 491 DK_CV_INLINE_SITE_ID, 492 DK_CV_LOC, 493 DK_CV_LINETABLE, 494 DK_CV_INLINE_LINETABLE, 495 DK_CV_DEF_RANGE, 496 DK_CV_STRINGTABLE, 497 DK_CV_STRING, 498 DK_CV_FILECHECKSUMS, 499 DK_CV_FILECHECKSUM_OFFSET, 500 DK_CV_FPO_DATA, 501 DK_CFI_SECTIONS, 502 DK_CFI_STARTPROC, 503 DK_CFI_ENDPROC, 504 DK_CFI_DEF_CFA, 505 DK_CFI_DEF_CFA_OFFSET, 506 DK_CFI_ADJUST_CFA_OFFSET, 507 DK_CFI_DEF_CFA_REGISTER, 508 DK_CFI_LLVM_DEF_ASPACE_CFA, 509 DK_CFI_OFFSET, 510 DK_CFI_REL_OFFSET, 511 DK_CFI_PERSONALITY, 512 DK_CFI_LSDA, 513 DK_CFI_REMEMBER_STATE, 514 DK_CFI_RESTORE_STATE, 515 DK_CFI_SAME_VALUE, 516 DK_CFI_RESTORE, 517 DK_CFI_ESCAPE, 518 DK_CFI_RETURN_COLUMN, 519 DK_CFI_SIGNAL_FRAME, 520 DK_CFI_UNDEFINED, 521 DK_CFI_REGISTER, 522 DK_CFI_WINDOW_SAVE, 523 DK_CFI_B_KEY_FRAME, 524 DK_MACROS_ON, 525 DK_MACROS_OFF, 526 DK_ALTMACRO, 527 DK_NOALTMACRO, 528 DK_MACRO, 529 DK_EXITM, 530 DK_ENDM, 531 DK_ENDMACRO, 532 DK_PURGEM, 533 DK_SLEB128, 534 DK_ULEB128, 535 DK_ERR, 536 DK_ERROR, 537 DK_WARNING, 538 DK_PRINT, 539 DK_ADDRSIG, 540 DK_ADDRSIG_SYM, 541 DK_PSEUDO_PROBE, 542 DK_LTO_DISCARD, 543 DK_LTO_SET_CONDITIONAL, 544 DK_CFI_MTE_TAGGED_FRAME, 545 DK_MEMTAG, 546 DK_END 547 }; 548 549 /// Maps directive name --> DirectiveKind enum, for 550 /// directives parsed by this class. 551 StringMap<DirectiveKind> DirectiveKindMap; 552 553 // Codeview def_range type parsing. 554 enum CVDefRangeType { 555 CVDR_DEFRANGE = 0, // Placeholder 556 CVDR_DEFRANGE_REGISTER, 557 CVDR_DEFRANGE_FRAMEPOINTER_REL, 558 CVDR_DEFRANGE_SUBFIELD_REGISTER, 559 CVDR_DEFRANGE_REGISTER_REL 560 }; 561 562 /// Maps Codeview def_range types --> CVDefRangeType enum, for 563 /// Codeview def_range types parsed by this class. 564 StringMap<CVDefRangeType> CVDefRangeTypeMap; 565 566 // ".ascii", ".asciz", ".string" 567 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 568 bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc" 569 bool parseDirectiveValue(StringRef IDVal, 570 unsigned Size); // ".byte", ".long", ... 571 bool parseDirectiveOctaValue(StringRef IDVal); // ".octa", ... 572 bool parseDirectiveRealValue(StringRef IDVal, 573 const fltSemantics &); // ".single", ... 574 bool parseDirectiveFill(); // ".fill" 575 bool parseDirectiveZero(); // ".zero" 576 // ".set", ".equ", ".equiv", ".lto_set_conditional" 577 bool parseDirectiveSet(StringRef IDVal, AssignmentKind Kind); 578 bool parseDirectiveOrg(); // ".org" 579 // ".align{,32}", ".p2align{,w,l}" 580 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize); 581 582 // ".file", ".line", ".loc", ".stabs" 583 bool parseDirectiveFile(SMLoc DirectiveLoc); 584 bool parseDirectiveLine(); 585 bool parseDirectiveLoc(); 586 bool parseDirectiveStabs(); 587 588 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable", 589 // ".cv_inline_linetable", ".cv_def_range", ".cv_string" 590 bool parseDirectiveCVFile(); 591 bool parseDirectiveCVFuncId(); 592 bool parseDirectiveCVInlineSiteId(); 593 bool parseDirectiveCVLoc(); 594 bool parseDirectiveCVLinetable(); 595 bool parseDirectiveCVInlineLinetable(); 596 bool parseDirectiveCVDefRange(); 597 bool parseDirectiveCVString(); 598 bool parseDirectiveCVStringTable(); 599 bool parseDirectiveCVFileChecksums(); 600 bool parseDirectiveCVFileChecksumOffset(); 601 bool parseDirectiveCVFPOData(); 602 603 // .cfi directives 604 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc); 605 bool parseDirectiveCFIWindowSave(SMLoc DirectiveLoc); 606 bool parseDirectiveCFISections(); 607 bool parseDirectiveCFIStartProc(); 608 bool parseDirectiveCFIEndProc(); 609 bool parseDirectiveCFIDefCfaOffset(SMLoc DirectiveLoc); 610 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc); 611 bool parseDirectiveCFIAdjustCfaOffset(SMLoc DirectiveLoc); 612 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc); 613 bool parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc); 614 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc); 615 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc); 616 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality); 617 bool parseDirectiveCFIRememberState(SMLoc DirectiveLoc); 618 bool parseDirectiveCFIRestoreState(SMLoc DirectiveLoc); 619 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc); 620 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc); 621 bool parseDirectiveCFIEscape(SMLoc DirectiveLoc); 622 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc); 623 bool parseDirectiveCFISignalFrame(SMLoc DirectiveLoc); 624 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc); 625 626 // macro directives 627 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc); 628 bool parseDirectiveExitMacro(StringRef Directive); 629 bool parseDirectiveEndMacro(StringRef Directive); 630 bool parseDirectiveMacro(SMLoc DirectiveLoc); 631 bool parseDirectiveMacrosOnOff(StringRef Directive); 632 // alternate macro mode directives 633 bool parseDirectiveAltmacro(StringRef Directive); 634 // ".bundle_align_mode" 635 bool parseDirectiveBundleAlignMode(); 636 // ".bundle_lock" 637 bool parseDirectiveBundleLock(); 638 // ".bundle_unlock" 639 bool parseDirectiveBundleUnlock(); 640 641 // ".space", ".skip" 642 bool parseDirectiveSpace(StringRef IDVal); 643 644 // ".dcb" 645 bool parseDirectiveDCB(StringRef IDVal, unsigned Size); 646 bool parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &); 647 // ".ds" 648 bool parseDirectiveDS(StringRef IDVal, unsigned Size); 649 650 // .sleb128 (Signed=true) and .uleb128 (Signed=false) 651 bool parseDirectiveLEB128(bool Signed); 652 653 /// Parse a directive like ".globl" which 654 /// accepts a single symbol (which should be a label or an external). 655 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr); 656 657 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 658 659 bool parseDirectiveAbort(); // ".abort" 660 bool parseDirectiveInclude(); // ".include" 661 bool parseDirectiveIncbin(); // ".incbin" 662 663 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne" 664 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 665 // ".ifb" or ".ifnb", depending on ExpectBlank. 666 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 667 // ".ifc" or ".ifnc", depending on ExpectEqual. 668 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual); 669 // ".ifeqs" or ".ifnes", depending on ExpectEqual. 670 bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual); 671 // ".ifdef" or ".ifndef", depending on expect_defined 672 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 673 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 674 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else" 675 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 676 bool parseEscapedString(std::string &Data) override; 677 bool parseAngleBracketString(std::string &Data) override; 678 679 const MCExpr *applyModifierToExpr(const MCExpr *E, 680 MCSymbolRefExpr::VariantKind Variant); 681 682 // Macro-like directives 683 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc); 684 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 685 raw_svector_ostream &OS); 686 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive); 687 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 688 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc" 689 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 690 691 // "_emit" or "__emit" 692 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info, 693 size_t Len); 694 695 // "align" 696 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info); 697 698 // "end" 699 bool parseDirectiveEnd(SMLoc DirectiveLoc); 700 701 // ".err" or ".error" 702 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage); 703 704 // ".warning" 705 bool parseDirectiveWarning(SMLoc DirectiveLoc); 706 707 // .print <double-quotes-string> 708 bool parseDirectivePrint(SMLoc DirectiveLoc); 709 710 // .pseudoprobe 711 bool parseDirectivePseudoProbe(); 712 713 // ".lto_discard" 714 bool parseDirectiveLTODiscard(); 715 716 // Directives to support address-significance tables. 717 bool parseDirectiveAddrsig(); 718 bool parseDirectiveAddrsigSym(); 719 720 void initializeDirectiveKindMap(); 721 void initializeCVDefRangeTypeMap(); 722 }; 723 724 class HLASMAsmParser final : public AsmParser { 725 private: 726 MCAsmLexer &Lexer; 727 MCStreamer &Out; 728 729 void lexLeadingSpaces() { 730 while (Lexer.is(AsmToken::Space)) 731 Lexer.Lex(); 732 } 733 734 bool parseAsHLASMLabel(ParseStatementInfo &Info, MCAsmParserSemaCallback *SI); 735 bool parseAsMachineInstruction(ParseStatementInfo &Info, 736 MCAsmParserSemaCallback *SI); 737 738 public: 739 HLASMAsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 740 const MCAsmInfo &MAI, unsigned CB = 0) 741 : AsmParser(SM, Ctx, Out, MAI, CB), Lexer(getLexer()), Out(Out) { 742 Lexer.setSkipSpace(false); 743 Lexer.setAllowHashInIdentifier(true); 744 Lexer.setLexHLASMIntegers(true); 745 Lexer.setLexHLASMStrings(true); 746 } 747 748 ~HLASMAsmParser() { Lexer.setSkipSpace(true); } 749 750 bool parseStatement(ParseStatementInfo &Info, 751 MCAsmParserSemaCallback *SI) override; 752 }; 753 754 } // end anonymous namespace 755 756 namespace llvm { 757 758 extern cl::opt<unsigned> AsmMacroMaxNestingDepth; 759 760 extern MCAsmParserExtension *createDarwinAsmParser(); 761 extern MCAsmParserExtension *createELFAsmParser(); 762 extern MCAsmParserExtension *createCOFFAsmParser(); 763 extern MCAsmParserExtension *createGOFFAsmParser(); 764 extern MCAsmParserExtension *createXCOFFAsmParser(); 765 extern MCAsmParserExtension *createWasmAsmParser(); 766 767 } // end namespace llvm 768 769 enum { DEFAULT_ADDRSPACE = 0 }; 770 771 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 772 const MCAsmInfo &MAI, unsigned CB = 0) 773 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM), 774 CurBuffer(CB ? CB : SM.getMainFileID()), MacrosEnabledFlag(true) { 775 HadError = false; 776 // Save the old handler. 777 SavedDiagHandler = SrcMgr.getDiagHandler(); 778 SavedDiagContext = SrcMgr.getDiagContext(); 779 // Set our own handler which calls the saved handler. 780 SrcMgr.setDiagHandler(DiagHandler, this); 781 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 782 // Make MCStreamer aware of the StartTokLoc for locations in diagnostics. 783 Out.setStartTokLocPtr(&StartTokLoc); 784 785 // Initialize the platform / file format parser. 786 switch (Ctx.getObjectFileType()) { 787 case MCContext::IsCOFF: 788 PlatformParser.reset(createCOFFAsmParser()); 789 break; 790 case MCContext::IsMachO: 791 PlatformParser.reset(createDarwinAsmParser()); 792 IsDarwin = true; 793 break; 794 case MCContext::IsELF: 795 PlatformParser.reset(createELFAsmParser()); 796 break; 797 case MCContext::IsGOFF: 798 PlatformParser.reset(createGOFFAsmParser()); 799 break; 800 case MCContext::IsSPIRV: 801 report_fatal_error( 802 "Need to implement createSPIRVAsmParser for SPIRV format."); 803 break; 804 case MCContext::IsWasm: 805 PlatformParser.reset(createWasmAsmParser()); 806 break; 807 case MCContext::IsXCOFF: 808 PlatformParser.reset(createXCOFFAsmParser()); 809 break; 810 case MCContext::IsDXContainer: 811 report_fatal_error("DXContainer is not supported yet"); 812 break; 813 } 814 815 PlatformParser->Initialize(*this); 816 initializeDirectiveKindMap(); 817 initializeCVDefRangeTypeMap(); 818 819 NumOfMacroInstantiations = 0; 820 } 821 822 AsmParser::~AsmParser() { 823 assert((HadError || ActiveMacros.empty()) && 824 "Unexpected active macro instantiation!"); 825 826 // Remove MCStreamer's reference to the parser SMLoc. 827 Out.setStartTokLocPtr(nullptr); 828 // Restore the saved diagnostics handler and context for use during 829 // finalization. 830 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext); 831 } 832 833 void AsmParser::printMacroInstantiations() { 834 // Print the active macro instantiation stack. 835 for (std::vector<MacroInstantiation *>::const_reverse_iterator 836 it = ActiveMacros.rbegin(), 837 ie = ActiveMacros.rend(); 838 it != ie; ++it) 839 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 840 "while in macro instantiation"); 841 } 842 843 void AsmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { 844 printPendingErrors(); 845 printMessage(L, SourceMgr::DK_Note, Msg, Range); 846 printMacroInstantiations(); 847 } 848 849 bool AsmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) { 850 if(getTargetParser().getTargetOptions().MCNoWarn) 851 return false; 852 if (getTargetParser().getTargetOptions().MCFatalWarnings) 853 return Error(L, Msg, Range); 854 printMessage(L, SourceMgr::DK_Warning, Msg, Range); 855 printMacroInstantiations(); 856 return false; 857 } 858 859 bool AsmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { 860 HadError = true; 861 printMessage(L, SourceMgr::DK_Error, Msg, Range); 862 printMacroInstantiations(); 863 return true; 864 } 865 866 bool AsmParser::enterIncludeFile(const std::string &Filename) { 867 std::string IncludedFile; 868 unsigned NewBuf = 869 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 870 if (!NewBuf) 871 return true; 872 873 CurBuffer = NewBuf; 874 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 875 return false; 876 } 877 878 /// Process the specified .incbin file by searching for it in the include paths 879 /// then just emitting the byte contents of the file to the streamer. This 880 /// returns true on failure. 881 bool AsmParser::processIncbinFile(const std::string &Filename, int64_t Skip, 882 const MCExpr *Count, SMLoc Loc) { 883 std::string IncludedFile; 884 unsigned NewBuf = 885 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 886 if (!NewBuf) 887 return true; 888 889 // Pick up the bytes from the file and emit them. 890 StringRef Bytes = SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(); 891 Bytes = Bytes.drop_front(Skip); 892 if (Count) { 893 int64_t Res; 894 if (!Count->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 895 return Error(Loc, "expected absolute expression"); 896 if (Res < 0) 897 return Warning(Loc, "negative count has no effect"); 898 Bytes = Bytes.take_front(Res); 899 } 900 getStreamer().emitBytes(Bytes); 901 return false; 902 } 903 904 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) { 905 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); 906 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), 907 Loc.getPointer()); 908 } 909 910 const AsmToken &AsmParser::Lex() { 911 if (Lexer.getTok().is(AsmToken::Error)) 912 Error(Lexer.getErrLoc(), Lexer.getErr()); 913 914 // if it's a end of statement with a comment in it 915 if (getTok().is(AsmToken::EndOfStatement)) { 916 // if this is a line comment output it. 917 if (!getTok().getString().empty() && getTok().getString().front() != '\n' && 918 getTok().getString().front() != '\r' && MAI.preserveAsmComments()) 919 Out.addExplicitComment(Twine(getTok().getString())); 920 } 921 922 const AsmToken *tok = &Lexer.Lex(); 923 924 // Parse comments here to be deferred until end of next statement. 925 while (tok->is(AsmToken::Comment)) { 926 if (MAI.preserveAsmComments()) 927 Out.addExplicitComment(Twine(tok->getString())); 928 tok = &Lexer.Lex(); 929 } 930 931 if (tok->is(AsmToken::Eof)) { 932 // If this is the end of an included file, pop the parent file off the 933 // include stack. 934 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 935 if (ParentIncludeLoc != SMLoc()) { 936 jumpToLoc(ParentIncludeLoc); 937 return Lex(); 938 } 939 } 940 941 return *tok; 942 } 943 944 bool AsmParser::enabledGenDwarfForAssembly() { 945 // Check whether the user specified -g. 946 if (!getContext().getGenDwarfForAssembly()) 947 return false; 948 // If we haven't encountered any .file directives (which would imply that 949 // the assembler source was produced with debug info already) then emit one 950 // describing the assembler source file itself. 951 if (getContext().getGenDwarfFileNumber() == 0) { 952 // Use the first #line directive for this, if any. It's preprocessed, so 953 // there is no checksum, and of course no source directive. 954 if (!FirstCppHashFilename.empty()) 955 getContext().setMCLineTableRootFile( 956 /*CUID=*/0, getContext().getCompilationDir(), FirstCppHashFilename, 957 /*Cksum=*/std::nullopt, /*Source=*/std::nullopt); 958 const MCDwarfFile &RootFile = 959 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); 960 getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective( 961 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name, 962 RootFile.Checksum, RootFile.Source)); 963 } 964 return true; 965 } 966 967 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 968 LTODiscardSymbols.clear(); 969 970 // Create the initial section, if requested. 971 if (!NoInitialTextSection) 972 Out.initSections(false, getTargetParser().getSTI()); 973 974 // Prime the lexer. 975 Lex(); 976 977 HadError = false; 978 AsmCond StartingCondState = TheCondState; 979 SmallVector<AsmRewrite, 4> AsmStrRewrites; 980 981 // If we are generating dwarf for assembly source files save the initial text 982 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't 983 // emitting any actual debug info yet and haven't had a chance to parse any 984 // embedded .file directives.) 985 if (getContext().getGenDwarfForAssembly()) { 986 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 987 if (!Sec->getBeginSymbol()) { 988 MCSymbol *SectionStartSym = getContext().createTempSymbol(); 989 getStreamer().emitLabel(SectionStartSym); 990 Sec->setBeginSymbol(SectionStartSym); 991 } 992 bool InsertResult = getContext().addGenDwarfSection(Sec); 993 assert(InsertResult && ".text section should not have debug info yet"); 994 (void)InsertResult; 995 } 996 997 getTargetParser().onBeginOfFile(); 998 999 // While we have input, parse each statement. 1000 while (Lexer.isNot(AsmToken::Eof)) { 1001 ParseStatementInfo Info(&AsmStrRewrites); 1002 bool Parsed = parseStatement(Info, nullptr); 1003 1004 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error 1005 // for printing ErrMsg via Lex() only if no (presumably better) parser error 1006 // exists. 1007 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) { 1008 Lex(); 1009 } 1010 1011 // parseStatement returned true so may need to emit an error. 1012 printPendingErrors(); 1013 1014 // Skipping to the next line if needed. 1015 if (Parsed && !getLexer().isAtStartOfStatement()) 1016 eatToEndOfStatement(); 1017 } 1018 1019 getTargetParser().onEndOfFile(); 1020 printPendingErrors(); 1021 1022 // All errors should have been emitted. 1023 assert(!hasPendingError() && "unexpected error from parseStatement"); 1024 1025 getTargetParser().flushPendingInstructions(getStreamer()); 1026 1027 if (TheCondState.TheCond != StartingCondState.TheCond || 1028 TheCondState.Ignore != StartingCondState.Ignore) 1029 printError(getTok().getLoc(), "unmatched .ifs or .elses"); 1030 // Check to see there are no empty DwarfFile slots. 1031 const auto &LineTables = getContext().getMCDwarfLineTables(); 1032 if (!LineTables.empty()) { 1033 unsigned Index = 0; 1034 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) { 1035 if (File.Name.empty() && Index != 0) 1036 printError(getTok().getLoc(), "unassigned file number: " + 1037 Twine(Index) + 1038 " for .file directives"); 1039 ++Index; 1040 } 1041 } 1042 1043 // Check to see that all assembler local symbols were actually defined. 1044 // Targets that don't do subsections via symbols may not want this, though, 1045 // so conservatively exclude them. Only do this if we're finalizing, though, 1046 // as otherwise we won't necessarilly have seen everything yet. 1047 if (!NoFinalize) { 1048 if (MAI.hasSubsectionsViaSymbols()) { 1049 for (const auto &TableEntry : getContext().getSymbols()) { 1050 MCSymbol *Sym = TableEntry.getValue(); 1051 // Variable symbols may not be marked as defined, so check those 1052 // explicitly. If we know it's a variable, we have a definition for 1053 // the purposes of this check. 1054 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 1055 // FIXME: We would really like to refer back to where the symbol was 1056 // first referenced for a source location. We need to add something 1057 // to track that. Currently, we just point to the end of the file. 1058 printError(getTok().getLoc(), "assembler local symbol '" + 1059 Sym->getName() + "' not defined"); 1060 } 1061 } 1062 1063 // Temporary symbols like the ones for directional jumps don't go in the 1064 // symbol table. They also need to be diagnosed in all (final) cases. 1065 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { 1066 if (std::get<2>(LocSym)->isUndefined()) { 1067 // Reset the state of any "# line file" directives we've seen to the 1068 // context as it was at the diagnostic site. 1069 CppHashInfo = std::get<1>(LocSym); 1070 printError(std::get<0>(LocSym), "directional label undefined"); 1071 } 1072 } 1073 } 1074 // Finalize the output stream if there are no errors and if the client wants 1075 // us to. 1076 if (!HadError && !NoFinalize) { 1077 if (auto *TS = Out.getTargetStreamer()) 1078 TS->emitConstantPools(); 1079 1080 Out.finish(Lexer.getLoc()); 1081 } 1082 1083 return HadError || getContext().hadError(); 1084 } 1085 1086 bool AsmParser::checkForValidSection() { 1087 if (!ParsingMSInlineAsm && !getStreamer().getCurrentSectionOnly()) { 1088 Out.initSections(false, getTargetParser().getSTI()); 1089 return Error(getTok().getLoc(), 1090 "expected section directive before assembly directive"); 1091 } 1092 return false; 1093 } 1094 1095 /// Throw away the rest of the line for testing purposes. 1096 void AsmParser::eatToEndOfStatement() { 1097 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1098 Lexer.Lex(); 1099 1100 // Eat EOL. 1101 if (Lexer.is(AsmToken::EndOfStatement)) 1102 Lexer.Lex(); 1103 } 1104 1105 StringRef AsmParser::parseStringToEndOfStatement() { 1106 const char *Start = getTok().getLoc().getPointer(); 1107 1108 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1109 Lexer.Lex(); 1110 1111 const char *End = getTok().getLoc().getPointer(); 1112 return StringRef(Start, End - Start); 1113 } 1114 1115 StringRef AsmParser::parseStringToComma() { 1116 const char *Start = getTok().getLoc().getPointer(); 1117 1118 while (Lexer.isNot(AsmToken::EndOfStatement) && 1119 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof)) 1120 Lexer.Lex(); 1121 1122 const char *End = getTok().getLoc().getPointer(); 1123 return StringRef(Start, End - Start); 1124 } 1125 1126 /// Parse a paren expression and return it. 1127 /// NOTE: This assumes the leading '(' has already been consumed. 1128 /// 1129 /// parenexpr ::= expr) 1130 /// 1131 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1132 if (parseExpression(Res)) 1133 return true; 1134 EndLoc = Lexer.getTok().getEndLoc(); 1135 return parseRParen(); 1136 } 1137 1138 /// Parse a bracket expression and return it. 1139 /// NOTE: This assumes the leading '[' has already been consumed. 1140 /// 1141 /// bracketexpr ::= expr] 1142 /// 1143 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1144 if (parseExpression(Res)) 1145 return true; 1146 EndLoc = getTok().getEndLoc(); 1147 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression")) 1148 return true; 1149 return false; 1150 } 1151 1152 /// Parse a primary expression and return it. 1153 /// primaryexpr ::= (parenexpr 1154 /// primaryexpr ::= symbol 1155 /// primaryexpr ::= number 1156 /// primaryexpr ::= '.' 1157 /// primaryexpr ::= ~,+,- primaryexpr 1158 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc, 1159 AsmTypeInfo *TypeInfo) { 1160 SMLoc FirstTokenLoc = getLexer().getLoc(); 1161 AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); 1162 switch (FirstTokenKind) { 1163 default: 1164 return TokError("unknown token in expression"); 1165 // If we have an error assume that we've already handled it. 1166 case AsmToken::Error: 1167 return true; 1168 case AsmToken::Exclaim: 1169 Lex(); // Eat the operator. 1170 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1171 return true; 1172 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc); 1173 return false; 1174 case AsmToken::Dollar: 1175 case AsmToken::Star: 1176 case AsmToken::At: 1177 case AsmToken::String: 1178 case AsmToken::Identifier: { 1179 StringRef Identifier; 1180 if (parseIdentifier(Identifier)) { 1181 // We may have failed but '$'|'*' may be a valid token in context of 1182 // the current PC. 1183 if (getTok().is(AsmToken::Dollar) || getTok().is(AsmToken::Star)) { 1184 bool ShouldGenerateTempSymbol = false; 1185 if ((getTok().is(AsmToken::Dollar) && MAI.getDollarIsPC()) || 1186 (getTok().is(AsmToken::Star) && MAI.getStarIsPC())) 1187 ShouldGenerateTempSymbol = true; 1188 1189 if (!ShouldGenerateTempSymbol) 1190 return Error(FirstTokenLoc, "invalid token in expression"); 1191 1192 // Eat the '$'|'*' token. 1193 Lex(); 1194 // This is either a '$'|'*' reference, which references the current PC. 1195 // Emit a temporary label to the streamer and refer to it. 1196 MCSymbol *Sym = Ctx.createTempSymbol(); 1197 Out.emitLabel(Sym); 1198 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, 1199 getContext()); 1200 EndLoc = FirstTokenLoc; 1201 return false; 1202 } 1203 } 1204 // Parse symbol variant 1205 std::pair<StringRef, StringRef> Split; 1206 if (!MAI.useParensForSymbolVariant()) { 1207 if (FirstTokenKind == AsmToken::String) { 1208 if (Lexer.is(AsmToken::At)) { 1209 Lex(); // eat @ 1210 SMLoc AtLoc = getLexer().getLoc(); 1211 StringRef VName; 1212 if (parseIdentifier(VName)) 1213 return Error(AtLoc, "expected symbol variant after '@'"); 1214 1215 Split = std::make_pair(Identifier, VName); 1216 } 1217 } else { 1218 Split = Identifier.split('@'); 1219 } 1220 } else if (Lexer.is(AsmToken::LParen)) { 1221 Lex(); // eat '('. 1222 StringRef VName; 1223 parseIdentifier(VName); 1224 if (parseRParen()) 1225 return true; 1226 Split = std::make_pair(Identifier, VName); 1227 } 1228 1229 EndLoc = SMLoc::getFromPointer(Identifier.end()); 1230 1231 // This is a symbol reference. 1232 StringRef SymbolName = Identifier; 1233 if (SymbolName.empty()) 1234 return Error(getLexer().getLoc(), "expected a symbol reference"); 1235 1236 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1237 1238 // Lookup the symbol variant if used. 1239 if (!Split.second.empty()) { 1240 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1241 if (Variant != MCSymbolRefExpr::VK_Invalid) { 1242 SymbolName = Split.first; 1243 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { 1244 Variant = MCSymbolRefExpr::VK_None; 1245 } else { 1246 return Error(SMLoc::getFromPointer(Split.second.begin()), 1247 "invalid variant '" + Split.second + "'"); 1248 } 1249 } 1250 1251 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName); 1252 if (!Sym) 1253 Sym = getContext().getOrCreateSymbol( 1254 MAI.shouldEmitLabelsInUpperCase() ? SymbolName.upper() : SymbolName); 1255 1256 // If this is an absolute variable reference, substitute it now to preserve 1257 // semantics in the face of reassignment. 1258 if (Sym->isVariable()) { 1259 auto V = Sym->getVariableValue(/*SetUsed*/ false); 1260 bool DoInline = isa<MCConstantExpr>(V) && !Variant; 1261 if (auto TV = dyn_cast<MCTargetExpr>(V)) 1262 DoInline = TV->inlineAssignedExpr(); 1263 if (DoInline) { 1264 if (Variant) 1265 return Error(EndLoc, "unexpected modifier on variable reference"); 1266 Res = Sym->getVariableValue(/*SetUsed*/ false); 1267 return false; 1268 } 1269 } 1270 1271 // Otherwise create a symbol ref. 1272 Res = MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc); 1273 return false; 1274 } 1275 case AsmToken::BigNum: 1276 return TokError("literal value out of range for directive"); 1277 case AsmToken::Integer: { 1278 SMLoc Loc = getTok().getLoc(); 1279 int64_t IntVal = getTok().getIntVal(); 1280 Res = MCConstantExpr::create(IntVal, getContext()); 1281 EndLoc = Lexer.getTok().getEndLoc(); 1282 Lex(); // Eat token. 1283 // Look for 'b' or 'f' following an Integer as a directional label 1284 if (Lexer.getKind() == AsmToken::Identifier) { 1285 StringRef IDVal = getTok().getString(); 1286 // Lookup the symbol variant if used. 1287 std::pair<StringRef, StringRef> Split = IDVal.split('@'); 1288 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1289 if (Split.first.size() != IDVal.size()) { 1290 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1291 if (Variant == MCSymbolRefExpr::VK_Invalid) 1292 return TokError("invalid variant '" + Split.second + "'"); 1293 IDVal = Split.first; 1294 } 1295 if (IDVal == "f" || IDVal == "b") { 1296 MCSymbol *Sym = 1297 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b"); 1298 Res = MCSymbolRefExpr::create(Sym, Variant, getContext()); 1299 if (IDVal == "b" && Sym->isUndefined()) 1300 return Error(Loc, "directional label undefined"); 1301 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym)); 1302 EndLoc = Lexer.getTok().getEndLoc(); 1303 Lex(); // Eat identifier. 1304 } 1305 } 1306 return false; 1307 } 1308 case AsmToken::Real: { 1309 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); 1310 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 1311 Res = MCConstantExpr::create(IntVal, getContext()); 1312 EndLoc = Lexer.getTok().getEndLoc(); 1313 Lex(); // Eat token. 1314 return false; 1315 } 1316 case AsmToken::Dot: { 1317 if (!MAI.getDotIsPC()) 1318 return TokError("cannot use . as current PC"); 1319 1320 // This is a '.' reference, which references the current PC. Emit a 1321 // temporary label to the streamer and refer to it. 1322 MCSymbol *Sym = Ctx.createTempSymbol(); 1323 Out.emitLabel(Sym); 1324 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext()); 1325 EndLoc = Lexer.getTok().getEndLoc(); 1326 Lex(); // Eat identifier. 1327 return false; 1328 } 1329 case AsmToken::LParen: 1330 Lex(); // Eat the '('. 1331 return parseParenExpr(Res, EndLoc); 1332 case AsmToken::LBrac: 1333 if (!PlatformParser->HasBracketExpressions()) 1334 return TokError("brackets expression not supported on this target"); 1335 Lex(); // Eat the '['. 1336 return parseBracketExpr(Res, EndLoc); 1337 case AsmToken::Minus: 1338 Lex(); // Eat the operator. 1339 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1340 return true; 1341 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc); 1342 return false; 1343 case AsmToken::Plus: 1344 Lex(); // Eat the operator. 1345 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1346 return true; 1347 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc); 1348 return false; 1349 case AsmToken::Tilde: 1350 Lex(); // Eat the operator. 1351 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1352 return true; 1353 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1354 return false; 1355 // MIPS unary expression operators. The lexer won't generate these tokens if 1356 // MCAsmInfo::HasMipsExpressions is false for the target. 1357 case AsmToken::PercentCall16: 1358 case AsmToken::PercentCall_Hi: 1359 case AsmToken::PercentCall_Lo: 1360 case AsmToken::PercentDtprel_Hi: 1361 case AsmToken::PercentDtprel_Lo: 1362 case AsmToken::PercentGot: 1363 case AsmToken::PercentGot_Disp: 1364 case AsmToken::PercentGot_Hi: 1365 case AsmToken::PercentGot_Lo: 1366 case AsmToken::PercentGot_Ofst: 1367 case AsmToken::PercentGot_Page: 1368 case AsmToken::PercentGottprel: 1369 case AsmToken::PercentGp_Rel: 1370 case AsmToken::PercentHi: 1371 case AsmToken::PercentHigher: 1372 case AsmToken::PercentHighest: 1373 case AsmToken::PercentLo: 1374 case AsmToken::PercentNeg: 1375 case AsmToken::PercentPcrel_Hi: 1376 case AsmToken::PercentPcrel_Lo: 1377 case AsmToken::PercentTlsgd: 1378 case AsmToken::PercentTlsldm: 1379 case AsmToken::PercentTprel_Hi: 1380 case AsmToken::PercentTprel_Lo: 1381 Lex(); // Eat the operator. 1382 if (Lexer.isNot(AsmToken::LParen)) 1383 return TokError("expected '(' after operator"); 1384 Lex(); // Eat the operator. 1385 if (parseExpression(Res, EndLoc)) 1386 return true; 1387 if (parseRParen()) 1388 return true; 1389 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx); 1390 return !Res; 1391 } 1392 } 1393 1394 bool AsmParser::parseExpression(const MCExpr *&Res) { 1395 SMLoc EndLoc; 1396 return parseExpression(Res, EndLoc); 1397 } 1398 1399 const MCExpr * 1400 AsmParser::applyModifierToExpr(const MCExpr *E, 1401 MCSymbolRefExpr::VariantKind Variant) { 1402 // Ask the target implementation about this expression first. 1403 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx); 1404 if (NewE) 1405 return NewE; 1406 // Recurse over the given expression, rebuilding it to apply the given variant 1407 // if there is exactly one symbol. 1408 switch (E->getKind()) { 1409 case MCExpr::Target: 1410 case MCExpr::Constant: 1411 return nullptr; 1412 1413 case MCExpr::SymbolRef: { 1414 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 1415 1416 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 1417 TokError("invalid variant on expression '" + getTok().getIdentifier() + 1418 "' (already modified)"); 1419 return E; 1420 } 1421 1422 return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext()); 1423 } 1424 1425 case MCExpr::Unary: { 1426 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 1427 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant); 1428 if (!Sub) 1429 return nullptr; 1430 return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext()); 1431 } 1432 1433 case MCExpr::Binary: { 1434 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 1435 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant); 1436 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant); 1437 1438 if (!LHS && !RHS) 1439 return nullptr; 1440 1441 if (!LHS) 1442 LHS = BE->getLHS(); 1443 if (!RHS) 1444 RHS = BE->getRHS(); 1445 1446 return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext()); 1447 } 1448 } 1449 1450 llvm_unreachable("Invalid expression kind!"); 1451 } 1452 1453 /// This function checks if the next token is <string> type or arithmetic. 1454 /// string that begin with character '<' must end with character '>'. 1455 /// otherwise it is arithmetics. 1456 /// If the function returns a 'true' value, 1457 /// the End argument will be filled with the last location pointed to the '>' 1458 /// character. 1459 1460 /// There is a gap between the AltMacro's documentation and the single quote 1461 /// implementation. GCC does not fully support this feature and so we will not 1462 /// support it. 1463 /// TODO: Adding single quote as a string. 1464 static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) { 1465 assert((StrLoc.getPointer() != nullptr) && 1466 "Argument to the function cannot be a NULL value"); 1467 const char *CharPtr = StrLoc.getPointer(); 1468 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && 1469 (*CharPtr != '\0')) { 1470 if (*CharPtr == '!') 1471 CharPtr++; 1472 CharPtr++; 1473 } 1474 if (*CharPtr == '>') { 1475 EndLoc = StrLoc.getFromPointer(CharPtr + 1); 1476 return true; 1477 } 1478 return false; 1479 } 1480 1481 /// creating a string without the escape characters '!'. 1482 static std::string angleBracketString(StringRef AltMacroStr) { 1483 std::string Res; 1484 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) { 1485 if (AltMacroStr[Pos] == '!') 1486 Pos++; 1487 Res += AltMacroStr[Pos]; 1488 } 1489 return Res; 1490 } 1491 1492 /// Parse an expression and return it. 1493 /// 1494 /// expr ::= expr &&,|| expr -> lowest. 1495 /// expr ::= expr |,^,&,! expr 1496 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 1497 /// expr ::= expr <<,>> expr 1498 /// expr ::= expr +,- expr 1499 /// expr ::= expr *,/,% expr -> highest. 1500 /// expr ::= primaryexpr 1501 /// 1502 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1503 // Parse the expression. 1504 Res = nullptr; 1505 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || 1506 parseBinOpRHS(1, Res, EndLoc)) 1507 return true; 1508 1509 // As a special case, we support 'a op b @ modifier' by rewriting the 1510 // expression to include the modifier. This is inefficient, but in general we 1511 // expect users to use 'a@modifier op b'. 1512 if (Lexer.getKind() == AsmToken::At) { 1513 Lex(); 1514 1515 if (Lexer.isNot(AsmToken::Identifier)) 1516 return TokError("unexpected symbol modifier following '@'"); 1517 1518 MCSymbolRefExpr::VariantKind Variant = 1519 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 1520 if (Variant == MCSymbolRefExpr::VK_Invalid) 1521 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 1522 1523 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant); 1524 if (!ModifiedRes) { 1525 return TokError("invalid modifier '" + getTok().getIdentifier() + 1526 "' (no symbols present)"); 1527 } 1528 1529 Res = ModifiedRes; 1530 Lex(); 1531 } 1532 1533 // Try to constant fold it up front, if possible. Do not exploit 1534 // assembler here. 1535 int64_t Value; 1536 if (Res->evaluateAsAbsolute(Value)) 1537 Res = MCConstantExpr::create(Value, getContext()); 1538 1539 return false; 1540 } 1541 1542 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1543 Res = nullptr; 1544 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc); 1545 } 1546 1547 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 1548 SMLoc &EndLoc) { 1549 if (parseParenExpr(Res, EndLoc)) 1550 return true; 1551 1552 for (; ParenDepth > 0; --ParenDepth) { 1553 if (parseBinOpRHS(1, Res, EndLoc)) 1554 return true; 1555 1556 // We don't Lex() the last RParen. 1557 // This is the same behavior as parseParenExpression(). 1558 if (ParenDepth - 1 > 0) { 1559 EndLoc = getTok().getEndLoc(); 1560 if (parseRParen()) 1561 return true; 1562 } 1563 } 1564 return false; 1565 } 1566 1567 bool AsmParser::parseAbsoluteExpression(int64_t &Res) { 1568 const MCExpr *Expr; 1569 1570 SMLoc StartLoc = Lexer.getLoc(); 1571 if (parseExpression(Expr)) 1572 return true; 1573 1574 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 1575 return Error(StartLoc, "expected absolute expression"); 1576 1577 return false; 1578 } 1579 1580 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K, 1581 MCBinaryExpr::Opcode &Kind, 1582 bool ShouldUseLogicalShr) { 1583 switch (K) { 1584 default: 1585 return 0; // not a binop. 1586 1587 // Lowest Precedence: &&, || 1588 case AsmToken::AmpAmp: 1589 Kind = MCBinaryExpr::LAnd; 1590 return 1; 1591 case AsmToken::PipePipe: 1592 Kind = MCBinaryExpr::LOr; 1593 return 1; 1594 1595 // Low Precedence: |, &, ^ 1596 case AsmToken::Pipe: 1597 Kind = MCBinaryExpr::Or; 1598 return 2; 1599 case AsmToken::Caret: 1600 Kind = MCBinaryExpr::Xor; 1601 return 2; 1602 case AsmToken::Amp: 1603 Kind = MCBinaryExpr::And; 1604 return 2; 1605 1606 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 1607 case AsmToken::EqualEqual: 1608 Kind = MCBinaryExpr::EQ; 1609 return 3; 1610 case AsmToken::ExclaimEqual: 1611 case AsmToken::LessGreater: 1612 Kind = MCBinaryExpr::NE; 1613 return 3; 1614 case AsmToken::Less: 1615 Kind = MCBinaryExpr::LT; 1616 return 3; 1617 case AsmToken::LessEqual: 1618 Kind = MCBinaryExpr::LTE; 1619 return 3; 1620 case AsmToken::Greater: 1621 Kind = MCBinaryExpr::GT; 1622 return 3; 1623 case AsmToken::GreaterEqual: 1624 Kind = MCBinaryExpr::GTE; 1625 return 3; 1626 1627 // Intermediate Precedence: <<, >> 1628 case AsmToken::LessLess: 1629 Kind = MCBinaryExpr::Shl; 1630 return 4; 1631 case AsmToken::GreaterGreater: 1632 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1633 return 4; 1634 1635 // High Intermediate Precedence: +, - 1636 case AsmToken::Plus: 1637 Kind = MCBinaryExpr::Add; 1638 return 5; 1639 case AsmToken::Minus: 1640 Kind = MCBinaryExpr::Sub; 1641 return 5; 1642 1643 // Highest Precedence: *, /, % 1644 case AsmToken::Star: 1645 Kind = MCBinaryExpr::Mul; 1646 return 6; 1647 case AsmToken::Slash: 1648 Kind = MCBinaryExpr::Div; 1649 return 6; 1650 case AsmToken::Percent: 1651 Kind = MCBinaryExpr::Mod; 1652 return 6; 1653 } 1654 } 1655 1656 static unsigned getGNUBinOpPrecedence(const MCAsmInfo &MAI, 1657 AsmToken::TokenKind K, 1658 MCBinaryExpr::Opcode &Kind, 1659 bool ShouldUseLogicalShr) { 1660 switch (K) { 1661 default: 1662 return 0; // not a binop. 1663 1664 // Lowest Precedence: &&, || 1665 case AsmToken::AmpAmp: 1666 Kind = MCBinaryExpr::LAnd; 1667 return 2; 1668 case AsmToken::PipePipe: 1669 Kind = MCBinaryExpr::LOr; 1670 return 1; 1671 1672 // Low Precedence: ==, !=, <>, <, <=, >, >= 1673 case AsmToken::EqualEqual: 1674 Kind = MCBinaryExpr::EQ; 1675 return 3; 1676 case AsmToken::ExclaimEqual: 1677 case AsmToken::LessGreater: 1678 Kind = MCBinaryExpr::NE; 1679 return 3; 1680 case AsmToken::Less: 1681 Kind = MCBinaryExpr::LT; 1682 return 3; 1683 case AsmToken::LessEqual: 1684 Kind = MCBinaryExpr::LTE; 1685 return 3; 1686 case AsmToken::Greater: 1687 Kind = MCBinaryExpr::GT; 1688 return 3; 1689 case AsmToken::GreaterEqual: 1690 Kind = MCBinaryExpr::GTE; 1691 return 3; 1692 1693 // Low Intermediate Precedence: +, - 1694 case AsmToken::Plus: 1695 Kind = MCBinaryExpr::Add; 1696 return 4; 1697 case AsmToken::Minus: 1698 Kind = MCBinaryExpr::Sub; 1699 return 4; 1700 1701 // High Intermediate Precedence: |, !, &, ^ 1702 // 1703 case AsmToken::Pipe: 1704 Kind = MCBinaryExpr::Or; 1705 return 5; 1706 case AsmToken::Exclaim: 1707 // Hack to support ARM compatible aliases (implied 'sp' operand in 'srs*' 1708 // instructions like 'srsda #31!') and not parse ! as an infix operator. 1709 if (MAI.getCommentString() == "@") 1710 return 0; 1711 Kind = MCBinaryExpr::OrNot; 1712 return 5; 1713 case AsmToken::Caret: 1714 Kind = MCBinaryExpr::Xor; 1715 return 5; 1716 case AsmToken::Amp: 1717 Kind = MCBinaryExpr::And; 1718 return 5; 1719 1720 // Highest Precedence: *, /, %, <<, >> 1721 case AsmToken::Star: 1722 Kind = MCBinaryExpr::Mul; 1723 return 6; 1724 case AsmToken::Slash: 1725 Kind = MCBinaryExpr::Div; 1726 return 6; 1727 case AsmToken::Percent: 1728 Kind = MCBinaryExpr::Mod; 1729 return 6; 1730 case AsmToken::LessLess: 1731 Kind = MCBinaryExpr::Shl; 1732 return 6; 1733 case AsmToken::GreaterGreater: 1734 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1735 return 6; 1736 } 1737 } 1738 1739 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K, 1740 MCBinaryExpr::Opcode &Kind) { 1741 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); 1742 return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr) 1743 : getGNUBinOpPrecedence(MAI, K, Kind, ShouldUseLogicalShr); 1744 } 1745 1746 /// Parse all binary operators with precedence >= 'Precedence'. 1747 /// Res contains the LHS of the expression on input. 1748 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 1749 SMLoc &EndLoc) { 1750 SMLoc StartLoc = Lexer.getLoc(); 1751 while (true) { 1752 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 1753 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 1754 1755 // If the next token is lower precedence than we are allowed to eat, return 1756 // successfully with what we ate already. 1757 if (TokPrec < Precedence) 1758 return false; 1759 1760 Lex(); 1761 1762 // Eat the next primary expression. 1763 const MCExpr *RHS; 1764 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc)) 1765 return true; 1766 1767 // If BinOp binds less tightly with RHS than the operator after RHS, let 1768 // the pending operator take RHS as its LHS. 1769 MCBinaryExpr::Opcode Dummy; 1770 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1771 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc)) 1772 return true; 1773 1774 // Merge LHS and RHS according to operator. 1775 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc); 1776 } 1777 } 1778 1779 /// ParseStatement: 1780 /// ::= EndOfStatement 1781 /// ::= Label* Directive ...Operands... EndOfStatement 1782 /// ::= Label* Identifier OperandList* EndOfStatement 1783 bool AsmParser::parseStatement(ParseStatementInfo &Info, 1784 MCAsmParserSemaCallback *SI) { 1785 assert(!hasPendingError() && "parseStatement started with pending error"); 1786 // Eat initial spaces and comments 1787 while (Lexer.is(AsmToken::Space)) 1788 Lex(); 1789 if (Lexer.is(AsmToken::EndOfStatement)) { 1790 // if this is a line comment we can drop it safely 1791 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 1792 getTok().getString().front() == '\n') 1793 Out.addBlankLine(); 1794 Lex(); 1795 return false; 1796 } 1797 // Statements always start with an identifier. 1798 AsmToken ID = getTok(); 1799 SMLoc IDLoc = ID.getLoc(); 1800 StringRef IDVal; 1801 int64_t LocalLabelVal = -1; 1802 StartTokLoc = ID.getLoc(); 1803 if (Lexer.is(AsmToken::HashDirective)) 1804 return parseCppHashLineFilenameComment(IDLoc, 1805 !isInsideMacroInstantiation()); 1806 1807 // Allow an integer followed by a ':' as a directional local label. 1808 if (Lexer.is(AsmToken::Integer)) { 1809 LocalLabelVal = getTok().getIntVal(); 1810 if (LocalLabelVal < 0) { 1811 if (!TheCondState.Ignore) { 1812 Lex(); // always eat a token 1813 return Error(IDLoc, "unexpected token at start of statement"); 1814 } 1815 IDVal = ""; 1816 } else { 1817 IDVal = getTok().getString(); 1818 Lex(); // Consume the integer token to be used as an identifier token. 1819 if (Lexer.getKind() != AsmToken::Colon) { 1820 if (!TheCondState.Ignore) { 1821 Lex(); // always eat a token 1822 return Error(IDLoc, "unexpected token at start of statement"); 1823 } 1824 } 1825 } 1826 } else if (Lexer.is(AsmToken::Dot)) { 1827 // Treat '.' as a valid identifier in this context. 1828 Lex(); 1829 IDVal = "."; 1830 } else if (Lexer.is(AsmToken::LCurly)) { 1831 // Treat '{' as a valid identifier in this context. 1832 Lex(); 1833 IDVal = "{"; 1834 1835 } else if (Lexer.is(AsmToken::RCurly)) { 1836 // Treat '}' as a valid identifier in this context. 1837 Lex(); 1838 IDVal = "}"; 1839 } else if (Lexer.is(AsmToken::Star) && 1840 getTargetParser().starIsStartOfStatement()) { 1841 // Accept '*' as a valid start of statement. 1842 Lex(); 1843 IDVal = "*"; 1844 } else if (parseIdentifier(IDVal)) { 1845 if (!TheCondState.Ignore) { 1846 Lex(); // always eat a token 1847 return Error(IDLoc, "unexpected token at start of statement"); 1848 } 1849 IDVal = ""; 1850 } 1851 1852 // Handle conditional assembly here before checking for skipping. We 1853 // have to do this so that .endif isn't skipped in a ".if 0" block for 1854 // example. 1855 StringMap<DirectiveKind>::const_iterator DirKindIt = 1856 DirectiveKindMap.find(IDVal.lower()); 1857 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) 1858 ? DK_NO_DIRECTIVE 1859 : DirKindIt->getValue(); 1860 switch (DirKind) { 1861 default: 1862 break; 1863 case DK_IF: 1864 case DK_IFEQ: 1865 case DK_IFGE: 1866 case DK_IFGT: 1867 case DK_IFLE: 1868 case DK_IFLT: 1869 case DK_IFNE: 1870 return parseDirectiveIf(IDLoc, DirKind); 1871 case DK_IFB: 1872 return parseDirectiveIfb(IDLoc, true); 1873 case DK_IFNB: 1874 return parseDirectiveIfb(IDLoc, false); 1875 case DK_IFC: 1876 return parseDirectiveIfc(IDLoc, true); 1877 case DK_IFEQS: 1878 return parseDirectiveIfeqs(IDLoc, true); 1879 case DK_IFNC: 1880 return parseDirectiveIfc(IDLoc, false); 1881 case DK_IFNES: 1882 return parseDirectiveIfeqs(IDLoc, false); 1883 case DK_IFDEF: 1884 return parseDirectiveIfdef(IDLoc, true); 1885 case DK_IFNDEF: 1886 case DK_IFNOTDEF: 1887 return parseDirectiveIfdef(IDLoc, false); 1888 case DK_ELSEIF: 1889 return parseDirectiveElseIf(IDLoc); 1890 case DK_ELSE: 1891 return parseDirectiveElse(IDLoc); 1892 case DK_ENDIF: 1893 return parseDirectiveEndIf(IDLoc); 1894 } 1895 1896 // Ignore the statement if in the middle of inactive conditional 1897 // (e.g. ".if 0"). 1898 if (TheCondState.Ignore) { 1899 eatToEndOfStatement(); 1900 return false; 1901 } 1902 1903 // FIXME: Recurse on local labels? 1904 1905 // Check for a label. 1906 // ::= identifier ':' 1907 // ::= number ':' 1908 if (Lexer.is(AsmToken::Colon) && getTargetParser().isLabel(ID)) { 1909 if (checkForValidSection()) 1910 return true; 1911 1912 Lex(); // Consume the ':'. 1913 1914 // Diagnose attempt to use '.' as a label. 1915 if (IDVal == ".") 1916 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1917 1918 // Diagnose attempt to use a variable as a label. 1919 // 1920 // FIXME: Diagnostics. Note the location of the definition as a label. 1921 // FIXME: This doesn't diagnose assignment to a symbol which has been 1922 // implicitly marked as external. 1923 MCSymbol *Sym; 1924 if (LocalLabelVal == -1) { 1925 if (ParsingMSInlineAsm && SI) { 1926 StringRef RewrittenLabel = 1927 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true); 1928 assert(!RewrittenLabel.empty() && 1929 "We should have an internal name here."); 1930 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(), 1931 RewrittenLabel); 1932 IDVal = RewrittenLabel; 1933 } 1934 Sym = getContext().getOrCreateSymbol(IDVal); 1935 } else 1936 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal); 1937 // End of Labels should be treated as end of line for lexing 1938 // purposes but that information is not available to the Lexer who 1939 // does not understand Labels. This may cause us to see a Hash 1940 // here instead of a preprocessor line comment. 1941 if (getTok().is(AsmToken::Hash)) { 1942 StringRef CommentStr = parseStringToEndOfStatement(); 1943 Lexer.Lex(); 1944 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr)); 1945 } 1946 1947 // Consume any end of statement token, if present, to avoid spurious 1948 // addBlankLine calls(). 1949 if (getTok().is(AsmToken::EndOfStatement)) { 1950 Lex(); 1951 } 1952 1953 if (MAI.hasSubsectionsViaSymbols() && CFIStartProcLoc && Sym->isExternal()) 1954 return Error(StartTokLoc, "non-private labels cannot appear between " 1955 ".cfi_startproc / .cfi_endproc pairs") && 1956 Error(*CFIStartProcLoc, "previous .cfi_startproc was here"); 1957 1958 if (discardLTOSymbol(IDVal)) 1959 return false; 1960 1961 getTargetParser().doBeforeLabelEmit(Sym, IDLoc); 1962 1963 // Emit the label. 1964 if (!getTargetParser().isParsingMSInlineAsm()) 1965 Out.emitLabel(Sym, IDLoc); 1966 1967 // If we are generating dwarf for assembly source files then gather the 1968 // info to make a dwarf label entry for this label if needed. 1969 if (enabledGenDwarfForAssembly()) 1970 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 1971 IDLoc); 1972 1973 getTargetParser().onLabelParsed(Sym); 1974 1975 return false; 1976 } 1977 1978 // Check for an assignment statement. 1979 // ::= identifier '=' 1980 if (Lexer.is(AsmToken::Equal) && getTargetParser().equalIsAsmAssignment()) { 1981 Lex(); 1982 return parseAssignment(IDVal, AssignmentKind::Equal); 1983 } 1984 1985 // If macros are enabled, check to see if this is a macro instantiation. 1986 if (areMacrosEnabled()) 1987 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) { 1988 return handleMacroEntry(M, IDLoc); 1989 } 1990 1991 // Otherwise, we have a normal instruction or directive. 1992 1993 // Directives start with "." 1994 if (IDVal.starts_with(".") && IDVal != ".") { 1995 // There are several entities interested in parsing directives: 1996 // 1997 // 1. The target-specific assembly parser. Some directives are target 1998 // specific or may potentially behave differently on certain targets. 1999 // 2. Asm parser extensions. For example, platform-specific parsers 2000 // (like the ELF parser) register themselves as extensions. 2001 // 3. The generic directive parser implemented by this class. These are 2002 // all the directives that behave in a target and platform independent 2003 // manner, or at least have a default behavior that's shared between 2004 // all targets and platforms. 2005 2006 getTargetParser().flushPendingInstructions(getStreamer()); 2007 2008 ParseStatus TPDirectiveReturn = getTargetParser().parseDirective(ID); 2009 assert(TPDirectiveReturn.isFailure() == hasPendingError() && 2010 "Should only return Failure iff there was an error"); 2011 if (TPDirectiveReturn.isFailure()) 2012 return true; 2013 if (TPDirectiveReturn.isSuccess()) 2014 return false; 2015 2016 // Next, check the extension directive map to see if any extension has 2017 // registered itself to parse this directive. 2018 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 2019 ExtensionDirectiveMap.lookup(IDVal); 2020 if (Handler.first) 2021 return (*Handler.second)(Handler.first, IDVal, IDLoc); 2022 2023 // Finally, if no one else is interested in this directive, it must be 2024 // generic and familiar to this class. 2025 switch (DirKind) { 2026 default: 2027 break; 2028 case DK_SET: 2029 case DK_EQU: 2030 return parseDirectiveSet(IDVal, AssignmentKind::Set); 2031 case DK_EQUIV: 2032 return parseDirectiveSet(IDVal, AssignmentKind::Equiv); 2033 case DK_LTO_SET_CONDITIONAL: 2034 return parseDirectiveSet(IDVal, AssignmentKind::LTOSetConditional); 2035 case DK_ASCII: 2036 return parseDirectiveAscii(IDVal, false); 2037 case DK_ASCIZ: 2038 case DK_STRING: 2039 return parseDirectiveAscii(IDVal, true); 2040 case DK_BYTE: 2041 case DK_DC_B: 2042 return parseDirectiveValue(IDVal, 1); 2043 case DK_DC: 2044 case DK_DC_W: 2045 case DK_SHORT: 2046 case DK_VALUE: 2047 case DK_2BYTE: 2048 return parseDirectiveValue(IDVal, 2); 2049 case DK_LONG: 2050 case DK_INT: 2051 case DK_4BYTE: 2052 case DK_DC_L: 2053 return parseDirectiveValue(IDVal, 4); 2054 case DK_QUAD: 2055 case DK_8BYTE: 2056 return parseDirectiveValue(IDVal, 8); 2057 case DK_DC_A: 2058 return parseDirectiveValue( 2059 IDVal, getContext().getAsmInfo()->getCodePointerSize()); 2060 case DK_OCTA: 2061 return parseDirectiveOctaValue(IDVal); 2062 case DK_SINGLE: 2063 case DK_FLOAT: 2064 case DK_DC_S: 2065 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle()); 2066 case DK_DOUBLE: 2067 case DK_DC_D: 2068 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble()); 2069 case DK_ALIGN: { 2070 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 2071 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1); 2072 } 2073 case DK_ALIGN32: { 2074 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 2075 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4); 2076 } 2077 case DK_BALIGN: 2078 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 2079 case DK_BALIGNW: 2080 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 2081 case DK_BALIGNL: 2082 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 2083 case DK_P2ALIGN: 2084 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 2085 case DK_P2ALIGNW: 2086 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 2087 case DK_P2ALIGNL: 2088 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 2089 case DK_ORG: 2090 return parseDirectiveOrg(); 2091 case DK_FILL: 2092 return parseDirectiveFill(); 2093 case DK_ZERO: 2094 return parseDirectiveZero(); 2095 case DK_EXTERN: 2096 eatToEndOfStatement(); // .extern is the default, ignore it. 2097 return false; 2098 case DK_GLOBL: 2099 case DK_GLOBAL: 2100 return parseDirectiveSymbolAttribute(MCSA_Global); 2101 case DK_LAZY_REFERENCE: 2102 return parseDirectiveSymbolAttribute(MCSA_LazyReference); 2103 case DK_NO_DEAD_STRIP: 2104 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 2105 case DK_SYMBOL_RESOLVER: 2106 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver); 2107 case DK_PRIVATE_EXTERN: 2108 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern); 2109 case DK_REFERENCE: 2110 return parseDirectiveSymbolAttribute(MCSA_Reference); 2111 case DK_WEAK_DEFINITION: 2112 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition); 2113 case DK_WEAK_REFERENCE: 2114 return parseDirectiveSymbolAttribute(MCSA_WeakReference); 2115 case DK_WEAK_DEF_CAN_BE_HIDDEN: 2116 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 2117 case DK_COLD: 2118 return parseDirectiveSymbolAttribute(MCSA_Cold); 2119 case DK_COMM: 2120 case DK_COMMON: 2121 return parseDirectiveComm(/*IsLocal=*/false); 2122 case DK_LCOMM: 2123 return parseDirectiveComm(/*IsLocal=*/true); 2124 case DK_ABORT: 2125 return parseDirectiveAbort(); 2126 case DK_INCLUDE: 2127 return parseDirectiveInclude(); 2128 case DK_INCBIN: 2129 return parseDirectiveIncbin(); 2130 case DK_CODE16: 2131 case DK_CODE16GCC: 2132 return TokError(Twine(IDVal) + 2133 " not currently supported for this target"); 2134 case DK_REPT: 2135 return parseDirectiveRept(IDLoc, IDVal); 2136 case DK_IRP: 2137 return parseDirectiveIrp(IDLoc); 2138 case DK_IRPC: 2139 return parseDirectiveIrpc(IDLoc); 2140 case DK_ENDR: 2141 return parseDirectiveEndr(IDLoc); 2142 case DK_BUNDLE_ALIGN_MODE: 2143 return parseDirectiveBundleAlignMode(); 2144 case DK_BUNDLE_LOCK: 2145 return parseDirectiveBundleLock(); 2146 case DK_BUNDLE_UNLOCK: 2147 return parseDirectiveBundleUnlock(); 2148 case DK_SLEB128: 2149 return parseDirectiveLEB128(true); 2150 case DK_ULEB128: 2151 return parseDirectiveLEB128(false); 2152 case DK_SPACE: 2153 case DK_SKIP: 2154 return parseDirectiveSpace(IDVal); 2155 case DK_FILE: 2156 return parseDirectiveFile(IDLoc); 2157 case DK_LINE: 2158 return parseDirectiveLine(); 2159 case DK_LOC: 2160 return parseDirectiveLoc(); 2161 case DK_STABS: 2162 return parseDirectiveStabs(); 2163 case DK_CV_FILE: 2164 return parseDirectiveCVFile(); 2165 case DK_CV_FUNC_ID: 2166 return parseDirectiveCVFuncId(); 2167 case DK_CV_INLINE_SITE_ID: 2168 return parseDirectiveCVInlineSiteId(); 2169 case DK_CV_LOC: 2170 return parseDirectiveCVLoc(); 2171 case DK_CV_LINETABLE: 2172 return parseDirectiveCVLinetable(); 2173 case DK_CV_INLINE_LINETABLE: 2174 return parseDirectiveCVInlineLinetable(); 2175 case DK_CV_DEF_RANGE: 2176 return parseDirectiveCVDefRange(); 2177 case DK_CV_STRING: 2178 return parseDirectiveCVString(); 2179 case DK_CV_STRINGTABLE: 2180 return parseDirectiveCVStringTable(); 2181 case DK_CV_FILECHECKSUMS: 2182 return parseDirectiveCVFileChecksums(); 2183 case DK_CV_FILECHECKSUM_OFFSET: 2184 return parseDirectiveCVFileChecksumOffset(); 2185 case DK_CV_FPO_DATA: 2186 return parseDirectiveCVFPOData(); 2187 case DK_CFI_SECTIONS: 2188 return parseDirectiveCFISections(); 2189 case DK_CFI_STARTPROC: 2190 return parseDirectiveCFIStartProc(); 2191 case DK_CFI_ENDPROC: 2192 return parseDirectiveCFIEndProc(); 2193 case DK_CFI_DEF_CFA: 2194 return parseDirectiveCFIDefCfa(IDLoc); 2195 case DK_CFI_DEF_CFA_OFFSET: 2196 return parseDirectiveCFIDefCfaOffset(IDLoc); 2197 case DK_CFI_ADJUST_CFA_OFFSET: 2198 return parseDirectiveCFIAdjustCfaOffset(IDLoc); 2199 case DK_CFI_DEF_CFA_REGISTER: 2200 return parseDirectiveCFIDefCfaRegister(IDLoc); 2201 case DK_CFI_LLVM_DEF_ASPACE_CFA: 2202 return parseDirectiveCFILLVMDefAspaceCfa(IDLoc); 2203 case DK_CFI_OFFSET: 2204 return parseDirectiveCFIOffset(IDLoc); 2205 case DK_CFI_REL_OFFSET: 2206 return parseDirectiveCFIRelOffset(IDLoc); 2207 case DK_CFI_PERSONALITY: 2208 return parseDirectiveCFIPersonalityOrLsda(true); 2209 case DK_CFI_LSDA: 2210 return parseDirectiveCFIPersonalityOrLsda(false); 2211 case DK_CFI_REMEMBER_STATE: 2212 return parseDirectiveCFIRememberState(IDLoc); 2213 case DK_CFI_RESTORE_STATE: 2214 return parseDirectiveCFIRestoreState(IDLoc); 2215 case DK_CFI_SAME_VALUE: 2216 return parseDirectiveCFISameValue(IDLoc); 2217 case DK_CFI_RESTORE: 2218 return parseDirectiveCFIRestore(IDLoc); 2219 case DK_CFI_ESCAPE: 2220 return parseDirectiveCFIEscape(IDLoc); 2221 case DK_CFI_RETURN_COLUMN: 2222 return parseDirectiveCFIReturnColumn(IDLoc); 2223 case DK_CFI_SIGNAL_FRAME: 2224 return parseDirectiveCFISignalFrame(IDLoc); 2225 case DK_CFI_UNDEFINED: 2226 return parseDirectiveCFIUndefined(IDLoc); 2227 case DK_CFI_REGISTER: 2228 return parseDirectiveCFIRegister(IDLoc); 2229 case DK_CFI_WINDOW_SAVE: 2230 return parseDirectiveCFIWindowSave(IDLoc); 2231 case DK_MACROS_ON: 2232 case DK_MACROS_OFF: 2233 return parseDirectiveMacrosOnOff(IDVal); 2234 case DK_MACRO: 2235 return parseDirectiveMacro(IDLoc); 2236 case DK_ALTMACRO: 2237 case DK_NOALTMACRO: 2238 return parseDirectiveAltmacro(IDVal); 2239 case DK_EXITM: 2240 return parseDirectiveExitMacro(IDVal); 2241 case DK_ENDM: 2242 case DK_ENDMACRO: 2243 return parseDirectiveEndMacro(IDVal); 2244 case DK_PURGEM: 2245 return parseDirectivePurgeMacro(IDLoc); 2246 case DK_END: 2247 return parseDirectiveEnd(IDLoc); 2248 case DK_ERR: 2249 return parseDirectiveError(IDLoc, false); 2250 case DK_ERROR: 2251 return parseDirectiveError(IDLoc, true); 2252 case DK_WARNING: 2253 return parseDirectiveWarning(IDLoc); 2254 case DK_RELOC: 2255 return parseDirectiveReloc(IDLoc); 2256 case DK_DCB: 2257 case DK_DCB_W: 2258 return parseDirectiveDCB(IDVal, 2); 2259 case DK_DCB_B: 2260 return parseDirectiveDCB(IDVal, 1); 2261 case DK_DCB_D: 2262 return parseDirectiveRealDCB(IDVal, APFloat::IEEEdouble()); 2263 case DK_DCB_L: 2264 return parseDirectiveDCB(IDVal, 4); 2265 case DK_DCB_S: 2266 return parseDirectiveRealDCB(IDVal, APFloat::IEEEsingle()); 2267 case DK_DC_X: 2268 case DK_DCB_X: 2269 return TokError(Twine(IDVal) + 2270 " not currently supported for this target"); 2271 case DK_DS: 2272 case DK_DS_W: 2273 return parseDirectiveDS(IDVal, 2); 2274 case DK_DS_B: 2275 return parseDirectiveDS(IDVal, 1); 2276 case DK_DS_D: 2277 return parseDirectiveDS(IDVal, 8); 2278 case DK_DS_L: 2279 case DK_DS_S: 2280 return parseDirectiveDS(IDVal, 4); 2281 case DK_DS_P: 2282 case DK_DS_X: 2283 return parseDirectiveDS(IDVal, 12); 2284 case DK_PRINT: 2285 return parseDirectivePrint(IDLoc); 2286 case DK_ADDRSIG: 2287 return parseDirectiveAddrsig(); 2288 case DK_ADDRSIG_SYM: 2289 return parseDirectiveAddrsigSym(); 2290 case DK_PSEUDO_PROBE: 2291 return parseDirectivePseudoProbe(); 2292 case DK_LTO_DISCARD: 2293 return parseDirectiveLTODiscard(); 2294 case DK_MEMTAG: 2295 return parseDirectiveSymbolAttribute(MCSA_Memtag); 2296 } 2297 2298 return Error(IDLoc, "unknown directive"); 2299 } 2300 2301 // __asm _emit or __asm __emit 2302 if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || 2303 IDVal == "_EMIT" || IDVal == "__EMIT")) 2304 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size()); 2305 2306 // __asm align 2307 if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN")) 2308 return parseDirectiveMSAlign(IDLoc, Info); 2309 2310 if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN")) 2311 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4); 2312 if (checkForValidSection()) 2313 return true; 2314 2315 return parseAndMatchAndEmitTargetInstruction(Info, IDVal, ID, IDLoc); 2316 } 2317 2318 bool AsmParser::parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info, 2319 StringRef IDVal, 2320 AsmToken ID, 2321 SMLoc IDLoc) { 2322 // Canonicalize the opcode to lower case. 2323 std::string OpcodeStr = IDVal.lower(); 2324 ParseInstructionInfo IInfo(Info.AsmRewrites); 2325 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID, 2326 Info.ParsedOperands); 2327 Info.ParseError = ParseHadError; 2328 2329 // Dump the parsed representation, if requested. 2330 if (getShowParsedOperands()) { 2331 SmallString<256> Str; 2332 raw_svector_ostream OS(Str); 2333 OS << "parsed instruction: ["; 2334 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { 2335 if (i != 0) 2336 OS << ", "; 2337 Info.ParsedOperands[i]->print(OS); 2338 } 2339 OS << "]"; 2340 2341 printMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 2342 } 2343 2344 // Fail even if ParseInstruction erroneously returns false. 2345 if (hasPendingError() || ParseHadError) 2346 return true; 2347 2348 // If we are generating dwarf for the current section then generate a .loc 2349 // directive for the instruction. 2350 if (!ParseHadError && enabledGenDwarfForAssembly() && 2351 getContext().getGenDwarfSectionSyms().count( 2352 getStreamer().getCurrentSectionOnly())) { 2353 unsigned Line; 2354 if (ActiveMacros.empty()) 2355 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer); 2356 else 2357 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc, 2358 ActiveMacros.front()->ExitBuffer); 2359 2360 // If we previously parsed a cpp hash file line comment then make sure the 2361 // current Dwarf File is for the CppHashFilename if not then emit the 2362 // Dwarf File table for it and adjust the line number for the .loc. 2363 if (!CppHashInfo.Filename.empty()) { 2364 unsigned FileNumber = getStreamer().emitDwarfFileDirective( 2365 0, StringRef(), CppHashInfo.Filename); 2366 getContext().setGenDwarfFileNumber(FileNumber); 2367 2368 unsigned CppHashLocLineNo = 2369 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf); 2370 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); 2371 } 2372 2373 getStreamer().emitDwarfLocDirective( 2374 getContext().getGenDwarfFileNumber(), Line, 0, 2375 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0, 2376 StringRef()); 2377 } 2378 2379 // If parsing succeeded, match the instruction. 2380 if (!ParseHadError) { 2381 uint64_t ErrorInfo; 2382 if (getTargetParser().MatchAndEmitInstruction( 2383 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo, 2384 getTargetParser().isParsingMSInlineAsm())) 2385 return true; 2386 } 2387 return false; 2388 } 2389 2390 // Parse and erase curly braces marking block start/end 2391 bool 2392 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { 2393 // Identify curly brace marking block start/end 2394 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly)) 2395 return false; 2396 2397 SMLoc StartLoc = Lexer.getLoc(); 2398 Lex(); // Eat the brace 2399 if (Lexer.is(AsmToken::EndOfStatement)) 2400 Lex(); // Eat EndOfStatement following the brace 2401 2402 // Erase the block start/end brace from the output asm string 2403 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() - 2404 StartLoc.getPointer()); 2405 return true; 2406 } 2407 2408 /// parseCppHashLineFilenameComment as this: 2409 /// ::= # number "filename" 2410 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo) { 2411 Lex(); // Eat the hash token. 2412 // Lexer only ever emits HashDirective if it fully formed if it's 2413 // done the checking already so this is an internal error. 2414 assert(getTok().is(AsmToken::Integer) && 2415 "Lexing Cpp line comment: Expected Integer"); 2416 int64_t LineNumber = getTok().getIntVal(); 2417 Lex(); 2418 assert(getTok().is(AsmToken::String) && 2419 "Lexing Cpp line comment: Expected String"); 2420 StringRef Filename = getTok().getString(); 2421 Lex(); 2422 2423 if (!SaveLocInfo) 2424 return false; 2425 2426 // Get rid of the enclosing quotes. 2427 Filename = Filename.substr(1, Filename.size() - 2); 2428 2429 // Save the SMLoc, Filename and LineNumber for later use by diagnostics 2430 // and possibly DWARF file info. 2431 CppHashInfo.Loc = L; 2432 CppHashInfo.Filename = Filename; 2433 CppHashInfo.LineNumber = LineNumber; 2434 CppHashInfo.Buf = CurBuffer; 2435 if (FirstCppHashFilename.empty()) 2436 FirstCppHashFilename = Filename; 2437 return false; 2438 } 2439 2440 /// will use the last parsed cpp hash line filename comment 2441 /// for the Filename and LineNo if any in the diagnostic. 2442 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 2443 auto *Parser = static_cast<AsmParser *>(Context); 2444 raw_ostream &OS = errs(); 2445 2446 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 2447 SMLoc DiagLoc = Diag.getLoc(); 2448 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2449 unsigned CppHashBuf = 2450 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc); 2451 2452 // Like SourceMgr::printMessage() we need to print the include stack if any 2453 // before printing the message. 2454 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2455 if (!Parser->SavedDiagHandler && DiagCurBuffer && 2456 DiagCurBuffer != DiagSrcMgr.getMainFileID()) { 2457 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 2458 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 2459 } 2460 2461 // If we have not parsed a cpp hash line filename comment or the source 2462 // manager changed or buffer changed (like in a nested include) then just 2463 // print the normal diagnostic using its Filename and LineNo. 2464 if (!Parser->CppHashInfo.LineNumber || DiagBuf != CppHashBuf) { 2465 if (Parser->SavedDiagHandler) 2466 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2467 else 2468 Parser->getContext().diagnose(Diag); 2469 return; 2470 } 2471 2472 // Use the CppHashFilename and calculate a line number based on the 2473 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc 2474 // for the diagnostic. 2475 const std::string &Filename = std::string(Parser->CppHashInfo.Filename); 2476 2477 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 2478 int CppHashLocLineNo = 2479 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf); 2480 int LineNo = 2481 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); 2482 2483 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, 2484 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), 2485 Diag.getLineContents(), Diag.getRanges()); 2486 2487 if (Parser->SavedDiagHandler) 2488 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2489 else 2490 Parser->getContext().diagnose(NewDiag); 2491 } 2492 2493 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The 2494 // difference being that that function accepts '@' as part of identifiers and 2495 // we can't do that. AsmLexer.cpp should probably be changed to handle 2496 // '@' as a special case when needed. 2497 static bool isIdentifierChar(char c) { 2498 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' || 2499 c == '.'; 2500 } 2501 2502 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 2503 ArrayRef<MCAsmMacroParameter> Parameters, 2504 ArrayRef<MCAsmMacroArgument> A, 2505 bool EnableAtPseudoVariable, SMLoc L) { 2506 unsigned NParameters = Parameters.size(); 2507 bool HasVararg = NParameters ? Parameters.back().Vararg : false; 2508 if ((!IsDarwin || NParameters != 0) && NParameters != A.size()) 2509 return Error(L, "Wrong number of arguments"); 2510 2511 // A macro without parameters is handled differently on Darwin: 2512 // gas accepts no arguments and does no substitutions 2513 while (!Body.empty()) { 2514 // Scan for the next substitution. 2515 std::size_t End = Body.size(), Pos = 0; 2516 for (; Pos != End; ++Pos) { 2517 // Check for a substitution or escape. 2518 if (IsDarwin && !NParameters) { 2519 // This macro has no parameters, look for $0, $1, etc. 2520 if (Body[Pos] != '$' || Pos + 1 == End) 2521 continue; 2522 2523 char Next = Body[Pos + 1]; 2524 if (Next == '$' || Next == 'n' || 2525 isdigit(static_cast<unsigned char>(Next))) 2526 break; 2527 } else { 2528 // This macro has parameters, look for \foo, \bar, etc. 2529 if (Body[Pos] == '\\' && Pos + 1 != End) 2530 break; 2531 } 2532 } 2533 2534 // Add the prefix. 2535 OS << Body.slice(0, Pos); 2536 2537 // Check if we reached the end. 2538 if (Pos == End) 2539 break; 2540 2541 if (IsDarwin && !NParameters) { 2542 switch (Body[Pos + 1]) { 2543 // $$ => $ 2544 case '$': 2545 OS << '$'; 2546 break; 2547 2548 // $n => number of arguments 2549 case 'n': 2550 OS << A.size(); 2551 break; 2552 2553 // $[0-9] => argument 2554 default: { 2555 // Missing arguments are ignored. 2556 unsigned Index = Body[Pos + 1] - '0'; 2557 if (Index >= A.size()) 2558 break; 2559 2560 // Otherwise substitute with the token values, with spaces eliminated. 2561 for (const AsmToken &Token : A[Index]) 2562 OS << Token.getString(); 2563 break; 2564 } 2565 } 2566 Pos += 2; 2567 } else { 2568 unsigned I = Pos + 1; 2569 2570 // Check for the \@ pseudo-variable. 2571 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End) 2572 ++I; 2573 else 2574 while (isIdentifierChar(Body[I]) && I + 1 != End) 2575 ++I; 2576 2577 const char *Begin = Body.data() + Pos + 1; 2578 StringRef Argument(Begin, I - (Pos + 1)); 2579 unsigned Index = 0; 2580 2581 if (Argument == "@") { 2582 OS << NumOfMacroInstantiations; 2583 Pos += 2; 2584 } else { 2585 for (; Index < NParameters; ++Index) 2586 if (Parameters[Index].Name == Argument) 2587 break; 2588 2589 if (Index == NParameters) { 2590 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 2591 Pos += 3; 2592 else { 2593 OS << '\\' << Argument; 2594 Pos = I; 2595 } 2596 } else { 2597 bool VarargParameter = HasVararg && Index == (NParameters - 1); 2598 for (const AsmToken &Token : A[Index]) 2599 // For altmacro mode, you can write '%expr'. 2600 // The prefix '%' evaluates the expression 'expr' 2601 // and uses the result as a string (e.g. replace %(1+2) with the 2602 // string "3"). 2603 // Here, we identify the integer token which is the result of the 2604 // absolute expression evaluation and replace it with its string 2605 // representation. 2606 if (AltMacroMode && Token.getString().front() == '%' && 2607 Token.is(AsmToken::Integer)) 2608 // Emit an integer value to the buffer. 2609 OS << Token.getIntVal(); 2610 // Only Token that was validated as a string and begins with '<' 2611 // is considered altMacroString!!! 2612 else if (AltMacroMode && Token.getString().front() == '<' && 2613 Token.is(AsmToken::String)) { 2614 OS << angleBracketString(Token.getStringContents()); 2615 } 2616 // We expect no quotes around the string's contents when 2617 // parsing for varargs. 2618 else if (Token.isNot(AsmToken::String) || VarargParameter) 2619 OS << Token.getString(); 2620 else 2621 OS << Token.getStringContents(); 2622 2623 Pos += 1 + Argument.size(); 2624 } 2625 } 2626 } 2627 // Update the scan point. 2628 Body = Body.substr(Pos); 2629 } 2630 2631 return false; 2632 } 2633 2634 static bool isOperator(AsmToken::TokenKind kind) { 2635 switch (kind) { 2636 default: 2637 return false; 2638 case AsmToken::Plus: 2639 case AsmToken::Minus: 2640 case AsmToken::Tilde: 2641 case AsmToken::Slash: 2642 case AsmToken::Star: 2643 case AsmToken::Dot: 2644 case AsmToken::Equal: 2645 case AsmToken::EqualEqual: 2646 case AsmToken::Pipe: 2647 case AsmToken::PipePipe: 2648 case AsmToken::Caret: 2649 case AsmToken::Amp: 2650 case AsmToken::AmpAmp: 2651 case AsmToken::Exclaim: 2652 case AsmToken::ExclaimEqual: 2653 case AsmToken::Less: 2654 case AsmToken::LessEqual: 2655 case AsmToken::LessLess: 2656 case AsmToken::LessGreater: 2657 case AsmToken::Greater: 2658 case AsmToken::GreaterEqual: 2659 case AsmToken::GreaterGreater: 2660 return true; 2661 } 2662 } 2663 2664 namespace { 2665 2666 class AsmLexerSkipSpaceRAII { 2667 public: 2668 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { 2669 Lexer.setSkipSpace(SkipSpace); 2670 } 2671 2672 ~AsmLexerSkipSpaceRAII() { 2673 Lexer.setSkipSpace(true); 2674 } 2675 2676 private: 2677 AsmLexer &Lexer; 2678 }; 2679 2680 } // end anonymous namespace 2681 2682 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) { 2683 2684 if (Vararg) { 2685 if (Lexer.isNot(AsmToken::EndOfStatement)) { 2686 StringRef Str = parseStringToEndOfStatement(); 2687 MA.emplace_back(AsmToken::String, Str); 2688 } 2689 return false; 2690 } 2691 2692 unsigned ParenLevel = 0; 2693 2694 // Darwin doesn't use spaces to delmit arguments. 2695 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); 2696 2697 bool SpaceEaten; 2698 2699 while (true) { 2700 SpaceEaten = false; 2701 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 2702 return TokError("unexpected token in macro instantiation"); 2703 2704 if (ParenLevel == 0) { 2705 2706 if (Lexer.is(AsmToken::Comma)) 2707 break; 2708 2709 if (Lexer.is(AsmToken::Space)) { 2710 SpaceEaten = true; 2711 Lexer.Lex(); // Eat spaces 2712 } 2713 2714 // Spaces can delimit parameters, but could also be part an expression. 2715 // If the token after a space is an operator, add the token and the next 2716 // one into this argument 2717 if (!IsDarwin) { 2718 if (isOperator(Lexer.getKind())) { 2719 MA.push_back(getTok()); 2720 Lexer.Lex(); 2721 2722 // Whitespace after an operator can be ignored. 2723 if (Lexer.is(AsmToken::Space)) 2724 Lexer.Lex(); 2725 2726 continue; 2727 } 2728 } 2729 if (SpaceEaten) 2730 break; 2731 } 2732 2733 // handleMacroEntry relies on not advancing the lexer here 2734 // to be able to fill in the remaining default parameter values 2735 if (Lexer.is(AsmToken::EndOfStatement)) 2736 break; 2737 2738 // Adjust the current parentheses level. 2739 if (Lexer.is(AsmToken::LParen)) 2740 ++ParenLevel; 2741 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 2742 --ParenLevel; 2743 2744 // Append the token to the current argument list. 2745 MA.push_back(getTok()); 2746 Lexer.Lex(); 2747 } 2748 2749 if (ParenLevel != 0) 2750 return TokError("unbalanced parentheses in macro argument"); 2751 return false; 2752 } 2753 2754 // Parse the macro instantiation arguments. 2755 bool AsmParser::parseMacroArguments(const MCAsmMacro *M, 2756 MCAsmMacroArguments &A) { 2757 const unsigned NParameters = M ? M->Parameters.size() : 0; 2758 bool NamedParametersFound = false; 2759 SmallVector<SMLoc, 4> FALocs; 2760 2761 A.resize(NParameters); 2762 FALocs.resize(NParameters); 2763 2764 // Parse two kinds of macro invocations: 2765 // - macros defined without any parameters accept an arbitrary number of them 2766 // - macros defined with parameters accept at most that many of them 2767 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false; 2768 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 2769 ++Parameter) { 2770 SMLoc IDLoc = Lexer.getLoc(); 2771 MCAsmMacroParameter FA; 2772 2773 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) { 2774 if (parseIdentifier(FA.Name)) 2775 return Error(IDLoc, "invalid argument identifier for formal argument"); 2776 2777 if (Lexer.isNot(AsmToken::Equal)) 2778 return TokError("expected '=' after formal parameter identifier"); 2779 2780 Lex(); 2781 2782 NamedParametersFound = true; 2783 } 2784 bool Vararg = HasVararg && Parameter == (NParameters - 1); 2785 2786 if (NamedParametersFound && FA.Name.empty()) 2787 return Error(IDLoc, "cannot mix positional and keyword arguments"); 2788 2789 SMLoc StrLoc = Lexer.getLoc(); 2790 SMLoc EndLoc; 2791 if (AltMacroMode && Lexer.is(AsmToken::Percent)) { 2792 const MCExpr *AbsoluteExp; 2793 int64_t Value; 2794 /// Eat '%' 2795 Lex(); 2796 if (parseExpression(AbsoluteExp, EndLoc)) 2797 return false; 2798 if (!AbsoluteExp->evaluateAsAbsolute(Value, 2799 getStreamer().getAssemblerPtr())) 2800 return Error(StrLoc, "expected absolute expression"); 2801 const char *StrChar = StrLoc.getPointer(); 2802 const char *EndChar = EndLoc.getPointer(); 2803 AsmToken newToken(AsmToken::Integer, 2804 StringRef(StrChar, EndChar - StrChar), Value); 2805 FA.Value.push_back(newToken); 2806 } else if (AltMacroMode && Lexer.is(AsmToken::Less) && 2807 isAngleBracketString(StrLoc, EndLoc)) { 2808 const char *StrChar = StrLoc.getPointer(); 2809 const char *EndChar = EndLoc.getPointer(); 2810 jumpToLoc(EndLoc, CurBuffer); 2811 /// Eat from '<' to '>' 2812 Lex(); 2813 AsmToken newToken(AsmToken::String, 2814 StringRef(StrChar, EndChar - StrChar)); 2815 FA.Value.push_back(newToken); 2816 } else if(parseMacroArgument(FA.Value, Vararg)) 2817 return true; 2818 2819 unsigned PI = Parameter; 2820 if (!FA.Name.empty()) { 2821 unsigned FAI = 0; 2822 for (FAI = 0; FAI < NParameters; ++FAI) 2823 if (M->Parameters[FAI].Name == FA.Name) 2824 break; 2825 2826 if (FAI >= NParameters) { 2827 assert(M && "expected macro to be defined"); 2828 return Error(IDLoc, "parameter named '" + FA.Name + 2829 "' does not exist for macro '" + M->Name + "'"); 2830 } 2831 PI = FAI; 2832 } 2833 2834 if (!FA.Value.empty()) { 2835 if (A.size() <= PI) 2836 A.resize(PI + 1); 2837 A[PI] = FA.Value; 2838 2839 if (FALocs.size() <= PI) 2840 FALocs.resize(PI + 1); 2841 2842 FALocs[PI] = Lexer.getLoc(); 2843 } 2844 2845 // At the end of the statement, fill in remaining arguments that have 2846 // default values. If there aren't any, then the next argument is 2847 // required but missing 2848 if (Lexer.is(AsmToken::EndOfStatement)) { 2849 bool Failure = false; 2850 for (unsigned FAI = 0; FAI < NParameters; ++FAI) { 2851 if (A[FAI].empty()) { 2852 if (M->Parameters[FAI].Required) { 2853 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), 2854 "missing value for required parameter " 2855 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'"); 2856 Failure = true; 2857 } 2858 2859 if (!M->Parameters[FAI].Value.empty()) 2860 A[FAI] = M->Parameters[FAI].Value; 2861 } 2862 } 2863 return Failure; 2864 } 2865 2866 if (Lexer.is(AsmToken::Comma)) 2867 Lex(); 2868 } 2869 2870 return TokError("too many positional arguments"); 2871 } 2872 2873 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) { 2874 // Arbitrarily limit macro nesting depth (default matches 'as'). We can 2875 // eliminate this, although we should protect against infinite loops. 2876 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; 2877 if (ActiveMacros.size() == MaxNestingDepth) { 2878 std::ostringstream MaxNestingDepthError; 2879 MaxNestingDepthError << "macros cannot be nested more than " 2880 << MaxNestingDepth << " levels deep." 2881 << " Use -asm-macro-max-nesting-depth to increase " 2882 "this limit."; 2883 return TokError(MaxNestingDepthError.str()); 2884 } 2885 2886 MCAsmMacroArguments A; 2887 if (parseMacroArguments(M, A)) 2888 return true; 2889 2890 // Macro instantiation is lexical, unfortunately. We construct a new buffer 2891 // to hold the macro body with substitutions. 2892 SmallString<256> Buf; 2893 StringRef Body = M->Body; 2894 raw_svector_ostream OS(Buf); 2895 2896 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc())) 2897 return true; 2898 2899 // We include the .endmacro in the buffer as our cue to exit the macro 2900 // instantiation. 2901 OS << ".endmacro\n"; 2902 2903 std::unique_ptr<MemoryBuffer> Instantiation = 2904 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 2905 2906 // Create the macro instantiation object and add to the current macro 2907 // instantiation stack. 2908 MacroInstantiation *MI = new MacroInstantiation{ 2909 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 2910 ActiveMacros.push_back(MI); 2911 2912 ++NumOfMacroInstantiations; 2913 2914 // Jump to the macro instantiation and prime the lexer. 2915 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 2916 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 2917 Lex(); 2918 2919 return false; 2920 } 2921 2922 void AsmParser::handleMacroExit() { 2923 // Jump to the EndOfStatement we should return to, and consume it. 2924 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer); 2925 Lex(); 2926 2927 // Pop the instantiation entry. 2928 delete ActiveMacros.back(); 2929 ActiveMacros.pop_back(); 2930 } 2931 2932 bool AsmParser::parseAssignment(StringRef Name, AssignmentKind Kind) { 2933 MCSymbol *Sym; 2934 const MCExpr *Value; 2935 SMLoc ExprLoc = getTok().getLoc(); 2936 bool AllowRedef = 2937 Kind == AssignmentKind::Set || Kind == AssignmentKind::Equal; 2938 if (MCParserUtils::parseAssignmentExpression(Name, AllowRedef, *this, Sym, 2939 Value)) 2940 return true; 2941 2942 if (!Sym) { 2943 // In the case where we parse an expression starting with a '.', we will 2944 // not generate an error, nor will we create a symbol. In this case we 2945 // should just return out. 2946 return false; 2947 } 2948 2949 if (discardLTOSymbol(Name)) 2950 return false; 2951 2952 // Do the assignment. 2953 switch (Kind) { 2954 case AssignmentKind::Equal: 2955 Out.emitAssignment(Sym, Value); 2956 break; 2957 case AssignmentKind::Set: 2958 case AssignmentKind::Equiv: 2959 Out.emitAssignment(Sym, Value); 2960 Out.emitSymbolAttribute(Sym, MCSA_NoDeadStrip); 2961 break; 2962 case AssignmentKind::LTOSetConditional: 2963 if (Value->getKind() != MCExpr::SymbolRef) 2964 return Error(ExprLoc, "expected identifier"); 2965 2966 Out.emitConditionalAssignment(Sym, Value); 2967 break; 2968 } 2969 2970 return false; 2971 } 2972 2973 /// parseIdentifier: 2974 /// ::= identifier 2975 /// ::= string 2976 bool AsmParser::parseIdentifier(StringRef &Res) { 2977 // The assembler has relaxed rules for accepting identifiers, in particular we 2978 // allow things like '.globl $foo' and '.def @feat.00', which would normally be 2979 // separate tokens. At this level, we have already lexed so we cannot (currently) 2980 // handle this as a context dependent token, instead we detect adjacent tokens 2981 // and return the combined identifier. 2982 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) { 2983 SMLoc PrefixLoc = getLexer().getLoc(); 2984 2985 // Consume the prefix character, and check for a following identifier. 2986 2987 AsmToken Buf[1]; 2988 Lexer.peekTokens(Buf, false); 2989 2990 if (Buf[0].isNot(AsmToken::Identifier) && Buf[0].isNot(AsmToken::Integer)) 2991 return true; 2992 2993 // We have a '$' or '@' followed by an identifier or integer token, make 2994 // sure they are adjacent. 2995 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer()) 2996 return true; 2997 2998 // eat $ or @ 2999 Lexer.Lex(); // Lexer's Lex guarantees consecutive token. 3000 // Construct the joined identifier and consume the token. 3001 Res = StringRef(PrefixLoc.getPointer(), getTok().getString().size() + 1); 3002 Lex(); // Parser Lex to maintain invariants. 3003 return false; 3004 } 3005 3006 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String)) 3007 return true; 3008 3009 Res = getTok().getIdentifier(); 3010 3011 Lex(); // Consume the identifier token. 3012 3013 return false; 3014 } 3015 3016 /// parseDirectiveSet: 3017 /// ::= .equ identifier ',' expression 3018 /// ::= .equiv identifier ',' expression 3019 /// ::= .set identifier ',' expression 3020 /// ::= .lto_set_conditional identifier ',' expression 3021 bool AsmParser::parseDirectiveSet(StringRef IDVal, AssignmentKind Kind) { 3022 StringRef Name; 3023 if (check(parseIdentifier(Name), "expected identifier") || parseComma() || 3024 parseAssignment(Name, Kind)) 3025 return true; 3026 return false; 3027 } 3028 3029 bool AsmParser::parseEscapedString(std::string &Data) { 3030 if (check(getTok().isNot(AsmToken::String), "expected string")) 3031 return true; 3032 3033 Data = ""; 3034 StringRef Str = getTok().getStringContents(); 3035 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 3036 if (Str[i] != '\\') { 3037 Data += Str[i]; 3038 continue; 3039 } 3040 3041 // Recognize escaped characters. Note that this escape semantics currently 3042 // loosely follows Darwin 'as'. 3043 ++i; 3044 if (i == e) 3045 return TokError("unexpected backslash at end of string"); 3046 3047 // Recognize hex sequences similarly to GNU 'as'. 3048 if (Str[i] == 'x' || Str[i] == 'X') { 3049 size_t length = Str.size(); 3050 if (i + 1 >= length || !isHexDigit(Str[i + 1])) 3051 return TokError("invalid hexadecimal escape sequence"); 3052 3053 // Consume hex characters. GNU 'as' reads all hexadecimal characters and 3054 // then truncates to the lower 16 bits. Seems reasonable. 3055 unsigned Value = 0; 3056 while (i + 1 < length && isHexDigit(Str[i + 1])) 3057 Value = Value * 16 + hexDigitValue(Str[++i]); 3058 3059 Data += (unsigned char)(Value & 0xFF); 3060 continue; 3061 } 3062 3063 // Recognize octal sequences. 3064 if ((unsigned)(Str[i] - '0') <= 7) { 3065 // Consume up to three octal characters. 3066 unsigned Value = Str[i] - '0'; 3067 3068 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3069 ++i; 3070 Value = Value * 8 + (Str[i] - '0'); 3071 3072 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3073 ++i; 3074 Value = Value * 8 + (Str[i] - '0'); 3075 } 3076 } 3077 3078 if (Value > 255) 3079 return TokError("invalid octal escape sequence (out of range)"); 3080 3081 Data += (unsigned char)Value; 3082 continue; 3083 } 3084 3085 // Otherwise recognize individual escapes. 3086 switch (Str[i]) { 3087 default: 3088 // Just reject invalid escape sequences for now. 3089 return TokError("invalid escape sequence (unrecognized character)"); 3090 3091 case 'b': Data += '\b'; break; 3092 case 'f': Data += '\f'; break; 3093 case 'n': Data += '\n'; break; 3094 case 'r': Data += '\r'; break; 3095 case 't': Data += '\t'; break; 3096 case '"': Data += '"'; break; 3097 case '\\': Data += '\\'; break; 3098 } 3099 } 3100 3101 Lex(); 3102 return false; 3103 } 3104 3105 bool AsmParser::parseAngleBracketString(std::string &Data) { 3106 SMLoc EndLoc, StartLoc = getTok().getLoc(); 3107 if (isAngleBracketString(StartLoc, EndLoc)) { 3108 const char *StartChar = StartLoc.getPointer() + 1; 3109 const char *EndChar = EndLoc.getPointer() - 1; 3110 jumpToLoc(EndLoc, CurBuffer); 3111 /// Eat from '<' to '>' 3112 Lex(); 3113 3114 Data = angleBracketString(StringRef(StartChar, EndChar - StartChar)); 3115 return false; 3116 } 3117 return true; 3118 } 3119 3120 /// parseDirectiveAscii: 3121 // ::= .ascii [ "string"+ ( , "string"+ )* ] 3122 /// ::= ( .asciz | .string ) [ "string" ( , "string" )* ] 3123 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 3124 auto parseOp = [&]() -> bool { 3125 std::string Data; 3126 if (checkForValidSection()) 3127 return true; 3128 // Only support spaces as separators for .ascii directive for now. See the 3129 // discusssion at https://reviews.llvm.org/D91460 for more details. 3130 do { 3131 if (parseEscapedString(Data)) 3132 return true; 3133 getStreamer().emitBytes(Data); 3134 } while (!ZeroTerminated && getTok().is(AsmToken::String)); 3135 if (ZeroTerminated) 3136 getStreamer().emitBytes(StringRef("\0", 1)); 3137 return false; 3138 }; 3139 3140 return parseMany(parseOp); 3141 } 3142 3143 /// parseDirectiveReloc 3144 /// ::= .reloc expression , identifier [ , expression ] 3145 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) { 3146 const MCExpr *Offset; 3147 const MCExpr *Expr = nullptr; 3148 SMLoc OffsetLoc = Lexer.getTok().getLoc(); 3149 3150 if (parseExpression(Offset)) 3151 return true; 3152 if (parseComma() || 3153 check(getTok().isNot(AsmToken::Identifier), "expected relocation name")) 3154 return true; 3155 3156 SMLoc NameLoc = Lexer.getTok().getLoc(); 3157 StringRef Name = Lexer.getTok().getIdentifier(); 3158 Lex(); 3159 3160 if (Lexer.is(AsmToken::Comma)) { 3161 Lex(); 3162 SMLoc ExprLoc = Lexer.getLoc(); 3163 if (parseExpression(Expr)) 3164 return true; 3165 3166 MCValue Value; 3167 if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr)) 3168 return Error(ExprLoc, "expression must be relocatable"); 3169 } 3170 3171 if (parseEOL()) 3172 return true; 3173 3174 const MCTargetAsmParser &MCT = getTargetParser(); 3175 const MCSubtargetInfo &STI = MCT.getSTI(); 3176 if (std::optional<std::pair<bool, std::string>> Err = 3177 getStreamer().emitRelocDirective(*Offset, Name, Expr, DirectiveLoc, 3178 STI)) 3179 return Error(Err->first ? NameLoc : OffsetLoc, Err->second); 3180 3181 return false; 3182 } 3183 3184 /// parseDirectiveValue 3185 /// ::= (.byte | .short | ... ) [ expression (, expression)* ] 3186 bool AsmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { 3187 auto parseOp = [&]() -> bool { 3188 const MCExpr *Value; 3189 SMLoc ExprLoc = getLexer().getLoc(); 3190 if (checkForValidSection() || parseExpression(Value)) 3191 return true; 3192 // Special case constant expressions to match code generator. 3193 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3194 assert(Size <= 8 && "Invalid size"); 3195 uint64_t IntValue = MCE->getValue(); 3196 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 3197 return Error(ExprLoc, "out of range literal value"); 3198 getStreamer().emitIntValue(IntValue, Size); 3199 } else 3200 getStreamer().emitValue(Value, Size, ExprLoc); 3201 return false; 3202 }; 3203 3204 return parseMany(parseOp); 3205 } 3206 3207 static bool parseHexOcta(AsmParser &Asm, uint64_t &hi, uint64_t &lo) { 3208 if (Asm.getTok().isNot(AsmToken::Integer) && 3209 Asm.getTok().isNot(AsmToken::BigNum)) 3210 return Asm.TokError("unknown token in expression"); 3211 SMLoc ExprLoc = Asm.getTok().getLoc(); 3212 APInt IntValue = Asm.getTok().getAPIntVal(); 3213 Asm.Lex(); 3214 if (!IntValue.isIntN(128)) 3215 return Asm.Error(ExprLoc, "out of range literal value"); 3216 if (!IntValue.isIntN(64)) { 3217 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue(); 3218 lo = IntValue.getLoBits(64).getZExtValue(); 3219 } else { 3220 hi = 0; 3221 lo = IntValue.getZExtValue(); 3222 } 3223 return false; 3224 } 3225 3226 /// ParseDirectiveOctaValue 3227 /// ::= .octa [ hexconstant (, hexconstant)* ] 3228 3229 bool AsmParser::parseDirectiveOctaValue(StringRef IDVal) { 3230 auto parseOp = [&]() -> bool { 3231 if (checkForValidSection()) 3232 return true; 3233 uint64_t hi, lo; 3234 if (parseHexOcta(*this, hi, lo)) 3235 return true; 3236 if (MAI.isLittleEndian()) { 3237 getStreamer().emitInt64(lo); 3238 getStreamer().emitInt64(hi); 3239 } else { 3240 getStreamer().emitInt64(hi); 3241 getStreamer().emitInt64(lo); 3242 } 3243 return false; 3244 }; 3245 3246 return parseMany(parseOp); 3247 } 3248 3249 bool AsmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { 3250 // We don't truly support arithmetic on floating point expressions, so we 3251 // have to manually parse unary prefixes. 3252 bool IsNeg = false; 3253 if (getLexer().is(AsmToken::Minus)) { 3254 Lexer.Lex(); 3255 IsNeg = true; 3256 } else if (getLexer().is(AsmToken::Plus)) 3257 Lexer.Lex(); 3258 3259 if (Lexer.is(AsmToken::Error)) 3260 return TokError(Lexer.getErr()); 3261 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) && 3262 Lexer.isNot(AsmToken::Identifier)) 3263 return TokError("unexpected token in directive"); 3264 3265 // Convert to an APFloat. 3266 APFloat Value(Semantics); 3267 StringRef IDVal = getTok().getString(); 3268 if (getLexer().is(AsmToken::Identifier)) { 3269 if (!IDVal.compare_insensitive("infinity") || 3270 !IDVal.compare_insensitive("inf")) 3271 Value = APFloat::getInf(Semantics); 3272 else if (!IDVal.compare_insensitive("nan")) 3273 Value = APFloat::getNaN(Semantics, false, ~0); 3274 else 3275 return TokError("invalid floating point literal"); 3276 } else if (errorToBool( 3277 Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) 3278 .takeError())) 3279 return TokError("invalid floating point literal"); 3280 if (IsNeg) 3281 Value.changeSign(); 3282 3283 // Consume the numeric token. 3284 Lex(); 3285 3286 Res = Value.bitcastToAPInt(); 3287 3288 return false; 3289 } 3290 3291 /// parseDirectiveRealValue 3292 /// ::= (.single | .double) [ expression (, expression)* ] 3293 bool AsmParser::parseDirectiveRealValue(StringRef IDVal, 3294 const fltSemantics &Semantics) { 3295 auto parseOp = [&]() -> bool { 3296 APInt AsInt; 3297 if (checkForValidSection() || parseRealValue(Semantics, AsInt)) 3298 return true; 3299 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3300 AsInt.getBitWidth() / 8); 3301 return false; 3302 }; 3303 3304 return parseMany(parseOp); 3305 } 3306 3307 /// parseDirectiveZero 3308 /// ::= .zero expression 3309 bool AsmParser::parseDirectiveZero() { 3310 SMLoc NumBytesLoc = Lexer.getLoc(); 3311 const MCExpr *NumBytes; 3312 if (checkForValidSection() || parseExpression(NumBytes)) 3313 return true; 3314 3315 int64_t Val = 0; 3316 if (getLexer().is(AsmToken::Comma)) { 3317 Lex(); 3318 if (parseAbsoluteExpression(Val)) 3319 return true; 3320 } 3321 3322 if (parseEOL()) 3323 return true; 3324 getStreamer().emitFill(*NumBytes, Val, NumBytesLoc); 3325 3326 return false; 3327 } 3328 3329 /// parseDirectiveFill 3330 /// ::= .fill expression [ , expression [ , expression ] ] 3331 bool AsmParser::parseDirectiveFill() { 3332 SMLoc NumValuesLoc = Lexer.getLoc(); 3333 const MCExpr *NumValues; 3334 if (checkForValidSection() || parseExpression(NumValues)) 3335 return true; 3336 3337 int64_t FillSize = 1; 3338 int64_t FillExpr = 0; 3339 3340 SMLoc SizeLoc, ExprLoc; 3341 3342 if (parseOptionalToken(AsmToken::Comma)) { 3343 SizeLoc = getTok().getLoc(); 3344 if (parseAbsoluteExpression(FillSize)) 3345 return true; 3346 if (parseOptionalToken(AsmToken::Comma)) { 3347 ExprLoc = getTok().getLoc(); 3348 if (parseAbsoluteExpression(FillExpr)) 3349 return true; 3350 } 3351 } 3352 if (parseEOL()) 3353 return true; 3354 3355 if (FillSize < 0) { 3356 Warning(SizeLoc, "'.fill' directive with negative size has no effect"); 3357 return false; 3358 } 3359 if (FillSize > 8) { 3360 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8"); 3361 FillSize = 8; 3362 } 3363 3364 if (!isUInt<32>(FillExpr) && FillSize > 4) 3365 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits"); 3366 3367 getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc); 3368 3369 return false; 3370 } 3371 3372 /// parseDirectiveOrg 3373 /// ::= .org expression [ , expression ] 3374 bool AsmParser::parseDirectiveOrg() { 3375 const MCExpr *Offset; 3376 SMLoc OffsetLoc = Lexer.getLoc(); 3377 if (checkForValidSection() || parseExpression(Offset)) 3378 return true; 3379 3380 // Parse optional fill expression. 3381 int64_t FillExpr = 0; 3382 if (parseOptionalToken(AsmToken::Comma)) 3383 if (parseAbsoluteExpression(FillExpr)) 3384 return true; 3385 if (parseEOL()) 3386 return true; 3387 3388 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc); 3389 return false; 3390 } 3391 3392 /// parseDirectiveAlign 3393 /// ::= {.align, ...} expression [ , expression [ , expression ]] 3394 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 3395 SMLoc AlignmentLoc = getLexer().getLoc(); 3396 int64_t Alignment; 3397 SMLoc MaxBytesLoc; 3398 bool HasFillExpr = false; 3399 int64_t FillExpr = 0; 3400 int64_t MaxBytesToFill = 0; 3401 SMLoc FillExprLoc; 3402 3403 auto parseAlign = [&]() -> bool { 3404 if (parseAbsoluteExpression(Alignment)) 3405 return true; 3406 if (parseOptionalToken(AsmToken::Comma)) { 3407 // The fill expression can be omitted while specifying a maximum number of 3408 // alignment bytes, e.g: 3409 // .align 3,,4 3410 if (getTok().isNot(AsmToken::Comma)) { 3411 HasFillExpr = true; 3412 if (parseTokenLoc(FillExprLoc) || parseAbsoluteExpression(FillExpr)) 3413 return true; 3414 } 3415 if (parseOptionalToken(AsmToken::Comma)) 3416 if (parseTokenLoc(MaxBytesLoc) || 3417 parseAbsoluteExpression(MaxBytesToFill)) 3418 return true; 3419 } 3420 return parseEOL(); 3421 }; 3422 3423 if (checkForValidSection()) 3424 return true; 3425 // Ignore empty '.p2align' directives for GNU-as compatibility 3426 if (IsPow2 && (ValueSize == 1) && getTok().is(AsmToken::EndOfStatement)) { 3427 Warning(AlignmentLoc, "p2align directive with no operand(s) is ignored"); 3428 return parseEOL(); 3429 } 3430 if (parseAlign()) 3431 return true; 3432 3433 // Always emit an alignment here even if we thrown an error. 3434 bool ReturnVal = false; 3435 3436 // Compute alignment in bytes. 3437 if (IsPow2) { 3438 // FIXME: Diagnose overflow. 3439 if (Alignment >= 32) { 3440 ReturnVal |= Error(AlignmentLoc, "invalid alignment value"); 3441 Alignment = 31; 3442 } 3443 3444 Alignment = 1ULL << Alignment; 3445 } else { 3446 // Reject alignments that aren't either a power of two or zero, 3447 // for gas compatibility. Alignment of zero is silently rounded 3448 // up to one. 3449 if (Alignment == 0) 3450 Alignment = 1; 3451 else if (!isPowerOf2_64(Alignment)) { 3452 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2"); 3453 Alignment = llvm::bit_floor<uint64_t>(Alignment); 3454 } 3455 if (!isUInt<32>(Alignment)) { 3456 ReturnVal |= Error(AlignmentLoc, "alignment must be smaller than 2**32"); 3457 Alignment = 1u << 31; 3458 } 3459 } 3460 3461 if (HasFillExpr && FillExpr != 0) { 3462 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 3463 if (Sec && Sec->isVirtualSection()) { 3464 ReturnVal |= 3465 Warning(FillExprLoc, "ignoring non-zero fill value in " + 3466 Sec->getVirtualSectionKind() + " section '" + 3467 Sec->getName() + "'"); 3468 FillExpr = 0; 3469 } 3470 } 3471 3472 // Diagnose non-sensical max bytes to align. 3473 if (MaxBytesLoc.isValid()) { 3474 if (MaxBytesToFill < 1) { 3475 ReturnVal |= Error(MaxBytesLoc, 3476 "alignment directive can never be satisfied in this " 3477 "many bytes, ignoring maximum bytes expression"); 3478 MaxBytesToFill = 0; 3479 } 3480 3481 if (MaxBytesToFill >= Alignment) { 3482 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 3483 "has no effect"); 3484 MaxBytesToFill = 0; 3485 } 3486 } 3487 3488 // Check whether we should use optimal code alignment for this .align 3489 // directive. 3490 const MCSection *Section = getStreamer().getCurrentSectionOnly(); 3491 assert(Section && "must have section to emit alignment"); 3492 bool useCodeAlign = Section->useCodeAlign(); 3493 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 3494 ValueSize == 1 && useCodeAlign) { 3495 getStreamer().emitCodeAlignment( 3496 Align(Alignment), &getTargetParser().getSTI(), MaxBytesToFill); 3497 } else { 3498 // FIXME: Target specific behavior about how the "extra" bytes are filled. 3499 getStreamer().emitValueToAlignment(Align(Alignment), FillExpr, ValueSize, 3500 MaxBytesToFill); 3501 } 3502 3503 return ReturnVal; 3504 } 3505 3506 /// parseDirectiveFile 3507 /// ::= .file filename 3508 /// ::= .file number [directory] filename [md5 checksum] [source source-text] 3509 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) { 3510 // FIXME: I'm not sure what this is. 3511 int64_t FileNumber = -1; 3512 if (getLexer().is(AsmToken::Integer)) { 3513 FileNumber = getTok().getIntVal(); 3514 Lex(); 3515 3516 if (FileNumber < 0) 3517 return TokError("negative file number"); 3518 } 3519 3520 std::string Path; 3521 3522 // Usually the directory and filename together, otherwise just the directory. 3523 // Allow the strings to have escaped octal character sequence. 3524 if (parseEscapedString(Path)) 3525 return true; 3526 3527 StringRef Directory; 3528 StringRef Filename; 3529 std::string FilenameData; 3530 if (getLexer().is(AsmToken::String)) { 3531 if (check(FileNumber == -1, 3532 "explicit path specified, but no file number") || 3533 parseEscapedString(FilenameData)) 3534 return true; 3535 Filename = FilenameData; 3536 Directory = Path; 3537 } else { 3538 Filename = Path; 3539 } 3540 3541 uint64_t MD5Hi, MD5Lo; 3542 bool HasMD5 = false; 3543 3544 std::optional<StringRef> Source; 3545 bool HasSource = false; 3546 std::string SourceString; 3547 3548 while (!parseOptionalToken(AsmToken::EndOfStatement)) { 3549 StringRef Keyword; 3550 if (check(getTok().isNot(AsmToken::Identifier), 3551 "unexpected token in '.file' directive") || 3552 parseIdentifier(Keyword)) 3553 return true; 3554 if (Keyword == "md5") { 3555 HasMD5 = true; 3556 if (check(FileNumber == -1, 3557 "MD5 checksum specified, but no file number") || 3558 parseHexOcta(*this, MD5Hi, MD5Lo)) 3559 return true; 3560 } else if (Keyword == "source") { 3561 HasSource = true; 3562 if (check(FileNumber == -1, 3563 "source specified, but no file number") || 3564 check(getTok().isNot(AsmToken::String), 3565 "unexpected token in '.file' directive") || 3566 parseEscapedString(SourceString)) 3567 return true; 3568 } else { 3569 return TokError("unexpected token in '.file' directive"); 3570 } 3571 } 3572 3573 if (FileNumber == -1) { 3574 // Ignore the directive if there is no number and the target doesn't support 3575 // numberless .file directives. This allows some portability of assembler 3576 // between different object file formats. 3577 if (getContext().getAsmInfo()->hasSingleParameterDotFile()) 3578 getStreamer().emitFileDirective(Filename); 3579 } else { 3580 // In case there is a -g option as well as debug info from directive .file, 3581 // we turn off the -g option, directly use the existing debug info instead. 3582 // Throw away any implicit file table for the assembler source. 3583 if (Ctx.getGenDwarfForAssembly()) { 3584 Ctx.getMCDwarfLineTable(0).resetFileTable(); 3585 Ctx.setGenDwarfForAssembly(false); 3586 } 3587 3588 std::optional<MD5::MD5Result> CKMem; 3589 if (HasMD5) { 3590 MD5::MD5Result Sum; 3591 for (unsigned i = 0; i != 8; ++i) { 3592 Sum[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); 3593 Sum[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); 3594 } 3595 CKMem = Sum; 3596 } 3597 if (HasSource) { 3598 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size())); 3599 memcpy(SourceBuf, SourceString.data(), SourceString.size()); 3600 Source = StringRef(SourceBuf, SourceString.size()); 3601 } 3602 if (FileNumber == 0) { 3603 // Upgrade to Version 5 for assembly actions like clang -c a.s. 3604 if (Ctx.getDwarfVersion() < 5) 3605 Ctx.setDwarfVersion(5); 3606 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source); 3607 } else { 3608 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( 3609 FileNumber, Directory, Filename, CKMem, Source); 3610 if (!FileNumOrErr) 3611 return Error(DirectiveLoc, toString(FileNumOrErr.takeError())); 3612 } 3613 // Alert the user if there are some .file directives with MD5 and some not. 3614 // But only do that once. 3615 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) { 3616 ReportedInconsistentMD5 = true; 3617 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums"); 3618 } 3619 } 3620 3621 return false; 3622 } 3623 3624 /// parseDirectiveLine 3625 /// ::= .line [number] 3626 bool AsmParser::parseDirectiveLine() { 3627 int64_t LineNumber; 3628 if (getLexer().is(AsmToken::Integer)) { 3629 if (parseIntToken(LineNumber, "unexpected token in '.line' directive")) 3630 return true; 3631 (void)LineNumber; 3632 // FIXME: Do something with the .line. 3633 } 3634 return parseEOL(); 3635 } 3636 3637 /// parseDirectiveLoc 3638 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 3639 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 3640 /// The first number is a file number, must have been previously assigned with 3641 /// a .file directive, the second number is the line number and optionally the 3642 /// third number is a column position (zero if not specified). The remaining 3643 /// optional items are .loc sub-directives. 3644 bool AsmParser::parseDirectiveLoc() { 3645 int64_t FileNumber = 0, LineNumber = 0; 3646 SMLoc Loc = getTok().getLoc(); 3647 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") || 3648 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, 3649 "file number less than one in '.loc' directive") || 3650 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc, 3651 "unassigned file number in '.loc' directive")) 3652 return true; 3653 3654 // optional 3655 if (getLexer().is(AsmToken::Integer)) { 3656 LineNumber = getTok().getIntVal(); 3657 if (LineNumber < 0) 3658 return TokError("line number less than zero in '.loc' directive"); 3659 Lex(); 3660 } 3661 3662 int64_t ColumnPos = 0; 3663 if (getLexer().is(AsmToken::Integer)) { 3664 ColumnPos = getTok().getIntVal(); 3665 if (ColumnPos < 0) 3666 return TokError("column position less than zero in '.loc' directive"); 3667 Lex(); 3668 } 3669 3670 auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags(); 3671 unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT; 3672 unsigned Isa = 0; 3673 int64_t Discriminator = 0; 3674 3675 auto parseLocOp = [&]() -> bool { 3676 StringRef Name; 3677 SMLoc Loc = getTok().getLoc(); 3678 if (parseIdentifier(Name)) 3679 return TokError("unexpected token in '.loc' directive"); 3680 3681 if (Name == "basic_block") 3682 Flags |= DWARF2_FLAG_BASIC_BLOCK; 3683 else if (Name == "prologue_end") 3684 Flags |= DWARF2_FLAG_PROLOGUE_END; 3685 else if (Name == "epilogue_begin") 3686 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 3687 else if (Name == "is_stmt") { 3688 Loc = getTok().getLoc(); 3689 const MCExpr *Value; 3690 if (parseExpression(Value)) 3691 return true; 3692 // The expression must be the constant 0 or 1. 3693 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3694 int Value = MCE->getValue(); 3695 if (Value == 0) 3696 Flags &= ~DWARF2_FLAG_IS_STMT; 3697 else if (Value == 1) 3698 Flags |= DWARF2_FLAG_IS_STMT; 3699 else 3700 return Error(Loc, "is_stmt value not 0 or 1"); 3701 } else { 3702 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 3703 } 3704 } else if (Name == "isa") { 3705 Loc = getTok().getLoc(); 3706 const MCExpr *Value; 3707 if (parseExpression(Value)) 3708 return true; 3709 // The expression must be a constant greater or equal to 0. 3710 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3711 int Value = MCE->getValue(); 3712 if (Value < 0) 3713 return Error(Loc, "isa number less than zero"); 3714 Isa = Value; 3715 } else { 3716 return Error(Loc, "isa number not a constant value"); 3717 } 3718 } else if (Name == "discriminator") { 3719 if (parseAbsoluteExpression(Discriminator)) 3720 return true; 3721 } else { 3722 return Error(Loc, "unknown sub-directive in '.loc' directive"); 3723 } 3724 return false; 3725 }; 3726 3727 if (parseMany(parseLocOp, false /*hasComma*/)) 3728 return true; 3729 3730 getStreamer().emitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 3731 Isa, Discriminator, StringRef()); 3732 3733 return false; 3734 } 3735 3736 /// parseDirectiveStabs 3737 /// ::= .stabs string, number, number, number 3738 bool AsmParser::parseDirectiveStabs() { 3739 return TokError("unsupported directive '.stabs'"); 3740 } 3741 3742 /// parseDirectiveCVFile 3743 /// ::= .cv_file number filename [checksum] [checksumkind] 3744 bool AsmParser::parseDirectiveCVFile() { 3745 SMLoc FileNumberLoc = getTok().getLoc(); 3746 int64_t FileNumber; 3747 std::string Filename; 3748 std::string Checksum; 3749 int64_t ChecksumKind = 0; 3750 3751 if (parseIntToken(FileNumber, 3752 "expected file number in '.cv_file' directive") || 3753 check(FileNumber < 1, FileNumberLoc, "file number less than one") || 3754 check(getTok().isNot(AsmToken::String), 3755 "unexpected token in '.cv_file' directive") || 3756 parseEscapedString(Filename)) 3757 return true; 3758 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 3759 if (check(getTok().isNot(AsmToken::String), 3760 "unexpected token in '.cv_file' directive") || 3761 parseEscapedString(Checksum) || 3762 parseIntToken(ChecksumKind, 3763 "expected checksum kind in '.cv_file' directive") || 3764 parseEOL()) 3765 return true; 3766 } 3767 3768 Checksum = fromHex(Checksum); 3769 void *CKMem = Ctx.allocate(Checksum.size(), 1); 3770 memcpy(CKMem, Checksum.data(), Checksum.size()); 3771 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), 3772 Checksum.size()); 3773 3774 if (!getStreamer().emitCVFileDirective(FileNumber, Filename, ChecksumAsBytes, 3775 static_cast<uint8_t>(ChecksumKind))) 3776 return Error(FileNumberLoc, "file number already allocated"); 3777 3778 return false; 3779 } 3780 3781 bool AsmParser::parseCVFunctionId(int64_t &FunctionId, 3782 StringRef DirectiveName) { 3783 SMLoc Loc; 3784 return parseTokenLoc(Loc) || 3785 parseIntToken(FunctionId, "expected function id in '" + DirectiveName + 3786 "' directive") || 3787 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc, 3788 "expected function id within range [0, UINT_MAX)"); 3789 } 3790 3791 bool AsmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { 3792 SMLoc Loc; 3793 return parseTokenLoc(Loc) || 3794 parseIntToken(FileNumber, "expected integer in '" + DirectiveName + 3795 "' directive") || 3796 check(FileNumber < 1, Loc, "file number less than one in '" + 3797 DirectiveName + "' directive") || 3798 check(!getCVContext().isValidFileNumber(FileNumber), Loc, 3799 "unassigned file number in '" + DirectiveName + "' directive"); 3800 } 3801 3802 /// parseDirectiveCVFuncId 3803 /// ::= .cv_func_id FunctionId 3804 /// 3805 /// Introduces a function ID that can be used with .cv_loc. 3806 bool AsmParser::parseDirectiveCVFuncId() { 3807 SMLoc FunctionIdLoc = getTok().getLoc(); 3808 int64_t FunctionId; 3809 3810 if (parseCVFunctionId(FunctionId, ".cv_func_id") || parseEOL()) 3811 return true; 3812 3813 if (!getStreamer().emitCVFuncIdDirective(FunctionId)) 3814 return Error(FunctionIdLoc, "function id already allocated"); 3815 3816 return false; 3817 } 3818 3819 /// parseDirectiveCVInlineSiteId 3820 /// ::= .cv_inline_site_id FunctionId 3821 /// "within" IAFunc 3822 /// "inlined_at" IAFile IALine [IACol] 3823 /// 3824 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined 3825 /// at" source location information for use in the line table of the caller, 3826 /// whether the caller is a real function or another inlined call site. 3827 bool AsmParser::parseDirectiveCVInlineSiteId() { 3828 SMLoc FunctionIdLoc = getTok().getLoc(); 3829 int64_t FunctionId; 3830 int64_t IAFunc; 3831 int64_t IAFile; 3832 int64_t IALine; 3833 int64_t IACol = 0; 3834 3835 // FunctionId 3836 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id")) 3837 return true; 3838 3839 // "within" 3840 if (check((getLexer().isNot(AsmToken::Identifier) || 3841 getTok().getIdentifier() != "within"), 3842 "expected 'within' identifier in '.cv_inline_site_id' directive")) 3843 return true; 3844 Lex(); 3845 3846 // IAFunc 3847 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id")) 3848 return true; 3849 3850 // "inlined_at" 3851 if (check((getLexer().isNot(AsmToken::Identifier) || 3852 getTok().getIdentifier() != "inlined_at"), 3853 "expected 'inlined_at' identifier in '.cv_inline_site_id' " 3854 "directive") ) 3855 return true; 3856 Lex(); 3857 3858 // IAFile IALine 3859 if (parseCVFileId(IAFile, ".cv_inline_site_id") || 3860 parseIntToken(IALine, "expected line number after 'inlined_at'")) 3861 return true; 3862 3863 // [IACol] 3864 if (getLexer().is(AsmToken::Integer)) { 3865 IACol = getTok().getIntVal(); 3866 Lex(); 3867 } 3868 3869 if (parseEOL()) 3870 return true; 3871 3872 if (!getStreamer().emitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, 3873 IALine, IACol, FunctionIdLoc)) 3874 return Error(FunctionIdLoc, "function id already allocated"); 3875 3876 return false; 3877 } 3878 3879 /// parseDirectiveCVLoc 3880 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] 3881 /// [is_stmt VALUE] 3882 /// The first number is a file number, must have been previously assigned with 3883 /// a .file directive, the second number is the line number and optionally the 3884 /// third number is a column position (zero if not specified). The remaining 3885 /// optional items are .loc sub-directives. 3886 bool AsmParser::parseDirectiveCVLoc() { 3887 SMLoc DirectiveLoc = getTok().getLoc(); 3888 int64_t FunctionId, FileNumber; 3889 if (parseCVFunctionId(FunctionId, ".cv_loc") || 3890 parseCVFileId(FileNumber, ".cv_loc")) 3891 return true; 3892 3893 int64_t LineNumber = 0; 3894 if (getLexer().is(AsmToken::Integer)) { 3895 LineNumber = getTok().getIntVal(); 3896 if (LineNumber < 0) 3897 return TokError("line number less than zero in '.cv_loc' directive"); 3898 Lex(); 3899 } 3900 3901 int64_t ColumnPos = 0; 3902 if (getLexer().is(AsmToken::Integer)) { 3903 ColumnPos = getTok().getIntVal(); 3904 if (ColumnPos < 0) 3905 return TokError("column position less than zero in '.cv_loc' directive"); 3906 Lex(); 3907 } 3908 3909 bool PrologueEnd = false; 3910 uint64_t IsStmt = 0; 3911 3912 auto parseOp = [&]() -> bool { 3913 StringRef Name; 3914 SMLoc Loc = getTok().getLoc(); 3915 if (parseIdentifier(Name)) 3916 return TokError("unexpected token in '.cv_loc' directive"); 3917 if (Name == "prologue_end") 3918 PrologueEnd = true; 3919 else if (Name == "is_stmt") { 3920 Loc = getTok().getLoc(); 3921 const MCExpr *Value; 3922 if (parseExpression(Value)) 3923 return true; 3924 // The expression must be the constant 0 or 1. 3925 IsStmt = ~0ULL; 3926 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value)) 3927 IsStmt = MCE->getValue(); 3928 3929 if (IsStmt > 1) 3930 return Error(Loc, "is_stmt value not 0 or 1"); 3931 } else { 3932 return Error(Loc, "unknown sub-directive in '.cv_loc' directive"); 3933 } 3934 return false; 3935 }; 3936 3937 if (parseMany(parseOp, false /*hasComma*/)) 3938 return true; 3939 3940 getStreamer().emitCVLocDirective(FunctionId, FileNumber, LineNumber, 3941 ColumnPos, PrologueEnd, IsStmt, StringRef(), 3942 DirectiveLoc); 3943 return false; 3944 } 3945 3946 /// parseDirectiveCVLinetable 3947 /// ::= .cv_linetable FunctionId, FnStart, FnEnd 3948 bool AsmParser::parseDirectiveCVLinetable() { 3949 int64_t FunctionId; 3950 StringRef FnStartName, FnEndName; 3951 SMLoc Loc = getTok().getLoc(); 3952 if (parseCVFunctionId(FunctionId, ".cv_linetable") || parseComma() || 3953 parseTokenLoc(Loc) || 3954 check(parseIdentifier(FnStartName), Loc, 3955 "expected identifier in directive") || 3956 parseComma() || parseTokenLoc(Loc) || 3957 check(parseIdentifier(FnEndName), Loc, 3958 "expected identifier in directive")) 3959 return true; 3960 3961 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3962 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3963 3964 getStreamer().emitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym); 3965 return false; 3966 } 3967 3968 /// parseDirectiveCVInlineLinetable 3969 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd 3970 bool AsmParser::parseDirectiveCVInlineLinetable() { 3971 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; 3972 StringRef FnStartName, FnEndName; 3973 SMLoc Loc = getTok().getLoc(); 3974 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") || 3975 parseTokenLoc(Loc) || 3976 parseIntToken( 3977 SourceFileId, 3978 "expected SourceField in '.cv_inline_linetable' directive") || 3979 check(SourceFileId <= 0, Loc, 3980 "File id less than zero in '.cv_inline_linetable' directive") || 3981 parseTokenLoc(Loc) || 3982 parseIntToken( 3983 SourceLineNum, 3984 "expected SourceLineNum in '.cv_inline_linetable' directive") || 3985 check(SourceLineNum < 0, Loc, 3986 "Line number less than zero in '.cv_inline_linetable' directive") || 3987 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3988 "expected identifier in directive") || 3989 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3990 "expected identifier in directive")) 3991 return true; 3992 3993 if (parseEOL()) 3994 return true; 3995 3996 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3997 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3998 getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, 3999 SourceLineNum, FnStartSym, 4000 FnEndSym); 4001 return false; 4002 } 4003 4004 void AsmParser::initializeCVDefRangeTypeMap() { 4005 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER; 4006 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL; 4007 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER; 4008 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL; 4009 } 4010 4011 /// parseDirectiveCVDefRange 4012 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* 4013 bool AsmParser::parseDirectiveCVDefRange() { 4014 SMLoc Loc; 4015 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; 4016 while (getLexer().is(AsmToken::Identifier)) { 4017 Loc = getLexer().getLoc(); 4018 StringRef GapStartName; 4019 if (parseIdentifier(GapStartName)) 4020 return Error(Loc, "expected identifier in directive"); 4021 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName); 4022 4023 Loc = getLexer().getLoc(); 4024 StringRef GapEndName; 4025 if (parseIdentifier(GapEndName)) 4026 return Error(Loc, "expected identifier in directive"); 4027 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName); 4028 4029 Ranges.push_back({GapStartSym, GapEndSym}); 4030 } 4031 4032 StringRef CVDefRangeTypeStr; 4033 if (parseToken( 4034 AsmToken::Comma, 4035 "expected comma before def_range type in .cv_def_range directive") || 4036 parseIdentifier(CVDefRangeTypeStr)) 4037 return Error(Loc, "expected def_range type in directive"); 4038 4039 StringMap<CVDefRangeType>::const_iterator CVTypeIt = 4040 CVDefRangeTypeMap.find(CVDefRangeTypeStr); 4041 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end()) 4042 ? CVDR_DEFRANGE 4043 : CVTypeIt->getValue(); 4044 switch (CVDRType) { 4045 case CVDR_DEFRANGE_REGISTER: { 4046 int64_t DRRegister; 4047 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4048 ".cv_def_range directive") || 4049 parseAbsoluteExpression(DRRegister)) 4050 return Error(Loc, "expected register number"); 4051 4052 codeview::DefRangeRegisterHeader DRHdr; 4053 DRHdr.Register = DRRegister; 4054 DRHdr.MayHaveNoName = 0; 4055 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4056 break; 4057 } 4058 case CVDR_DEFRANGE_FRAMEPOINTER_REL: { 4059 int64_t DROffset; 4060 if (parseToken(AsmToken::Comma, 4061 "expected comma before offset in .cv_def_range directive") || 4062 parseAbsoluteExpression(DROffset)) 4063 return Error(Loc, "expected offset value"); 4064 4065 codeview::DefRangeFramePointerRelHeader DRHdr; 4066 DRHdr.Offset = DROffset; 4067 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4068 break; 4069 } 4070 case CVDR_DEFRANGE_SUBFIELD_REGISTER: { 4071 int64_t DRRegister; 4072 int64_t DROffsetInParent; 4073 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4074 ".cv_def_range directive") || 4075 parseAbsoluteExpression(DRRegister)) 4076 return Error(Loc, "expected register number"); 4077 if (parseToken(AsmToken::Comma, 4078 "expected comma before offset in .cv_def_range directive") || 4079 parseAbsoluteExpression(DROffsetInParent)) 4080 return Error(Loc, "expected offset value"); 4081 4082 codeview::DefRangeSubfieldRegisterHeader DRHdr; 4083 DRHdr.Register = DRRegister; 4084 DRHdr.MayHaveNoName = 0; 4085 DRHdr.OffsetInParent = DROffsetInParent; 4086 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4087 break; 4088 } 4089 case CVDR_DEFRANGE_REGISTER_REL: { 4090 int64_t DRRegister; 4091 int64_t DRFlags; 4092 int64_t DRBasePointerOffset; 4093 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4094 ".cv_def_range directive") || 4095 parseAbsoluteExpression(DRRegister)) 4096 return Error(Loc, "expected register value"); 4097 if (parseToken( 4098 AsmToken::Comma, 4099 "expected comma before flag value in .cv_def_range directive") || 4100 parseAbsoluteExpression(DRFlags)) 4101 return Error(Loc, "expected flag value"); 4102 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset " 4103 "in .cv_def_range directive") || 4104 parseAbsoluteExpression(DRBasePointerOffset)) 4105 return Error(Loc, "expected base pointer offset value"); 4106 4107 codeview::DefRangeRegisterRelHeader DRHdr; 4108 DRHdr.Register = DRRegister; 4109 DRHdr.Flags = DRFlags; 4110 DRHdr.BasePointerOffset = DRBasePointerOffset; 4111 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4112 break; 4113 } 4114 default: 4115 return Error(Loc, "unexpected def_range type in .cv_def_range directive"); 4116 } 4117 return true; 4118 } 4119 4120 /// parseDirectiveCVString 4121 /// ::= .cv_stringtable "string" 4122 bool AsmParser::parseDirectiveCVString() { 4123 std::string Data; 4124 if (checkForValidSection() || parseEscapedString(Data)) 4125 return true; 4126 4127 // Put the string in the table and emit the offset. 4128 std::pair<StringRef, unsigned> Insertion = 4129 getCVContext().addToStringTable(Data); 4130 getStreamer().emitInt32(Insertion.second); 4131 return false; 4132 } 4133 4134 /// parseDirectiveCVStringTable 4135 /// ::= .cv_stringtable 4136 bool AsmParser::parseDirectiveCVStringTable() { 4137 getStreamer().emitCVStringTableDirective(); 4138 return false; 4139 } 4140 4141 /// parseDirectiveCVFileChecksums 4142 /// ::= .cv_filechecksums 4143 bool AsmParser::parseDirectiveCVFileChecksums() { 4144 getStreamer().emitCVFileChecksumsDirective(); 4145 return false; 4146 } 4147 4148 /// parseDirectiveCVFileChecksumOffset 4149 /// ::= .cv_filechecksumoffset fileno 4150 bool AsmParser::parseDirectiveCVFileChecksumOffset() { 4151 int64_t FileNo; 4152 if (parseIntToken(FileNo, "expected identifier in directive")) 4153 return true; 4154 if (parseEOL()) 4155 return true; 4156 getStreamer().emitCVFileChecksumOffsetDirective(FileNo); 4157 return false; 4158 } 4159 4160 /// parseDirectiveCVFPOData 4161 /// ::= .cv_fpo_data procsym 4162 bool AsmParser::parseDirectiveCVFPOData() { 4163 SMLoc DirLoc = getLexer().getLoc(); 4164 StringRef ProcName; 4165 if (parseIdentifier(ProcName)) 4166 return TokError("expected symbol name"); 4167 if (parseEOL()) 4168 return true; 4169 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName); 4170 getStreamer().emitCVFPOData(ProcSym, DirLoc); 4171 return false; 4172 } 4173 4174 /// parseDirectiveCFISections 4175 /// ::= .cfi_sections section [, section] 4176 bool AsmParser::parseDirectiveCFISections() { 4177 StringRef Name; 4178 bool EH = false; 4179 bool Debug = false; 4180 4181 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4182 for (;;) { 4183 if (parseIdentifier(Name)) 4184 return TokError("expected .eh_frame or .debug_frame"); 4185 if (Name == ".eh_frame") 4186 EH = true; 4187 else if (Name == ".debug_frame") 4188 Debug = true; 4189 if (parseOptionalToken(AsmToken::EndOfStatement)) 4190 break; 4191 if (parseComma()) 4192 return true; 4193 } 4194 } 4195 getStreamer().emitCFISections(EH, Debug); 4196 return false; 4197 } 4198 4199 /// parseDirectiveCFIStartProc 4200 /// ::= .cfi_startproc [simple] 4201 bool AsmParser::parseDirectiveCFIStartProc() { 4202 CFIStartProcLoc = StartTokLoc; 4203 4204 StringRef Simple; 4205 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4206 if (check(parseIdentifier(Simple) || Simple != "simple", 4207 "unexpected token") || 4208 parseEOL()) 4209 return true; 4210 } 4211 4212 // TODO(kristina): Deal with a corner case of incorrect diagnostic context 4213 // being produced if this directive is emitted as part of preprocessor macro 4214 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. 4215 // Tools like llvm-mc on the other hand are not affected by it, and report 4216 // correct context information. 4217 getStreamer().emitCFIStartProc(!Simple.empty(), Lexer.getLoc()); 4218 return false; 4219 } 4220 4221 /// parseDirectiveCFIEndProc 4222 /// ::= .cfi_endproc 4223 bool AsmParser::parseDirectiveCFIEndProc() { 4224 CFIStartProcLoc = std::nullopt; 4225 4226 if (parseEOL()) 4227 return true; 4228 4229 getStreamer().emitCFIEndProc(); 4230 return false; 4231 } 4232 4233 /// parse register name or number. 4234 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register, 4235 SMLoc DirectiveLoc) { 4236 MCRegister RegNo; 4237 4238 if (getLexer().isNot(AsmToken::Integer)) { 4239 if (getTargetParser().parseRegister(RegNo, DirectiveLoc, DirectiveLoc)) 4240 return true; 4241 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true); 4242 } else 4243 return parseAbsoluteExpression(Register); 4244 4245 return false; 4246 } 4247 4248 /// parseDirectiveCFIDefCfa 4249 /// ::= .cfi_def_cfa register, offset 4250 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { 4251 int64_t Register = 0, Offset = 0; 4252 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4253 parseAbsoluteExpression(Offset) || parseEOL()) 4254 return true; 4255 4256 getStreamer().emitCFIDefCfa(Register, Offset, DirectiveLoc); 4257 return false; 4258 } 4259 4260 /// parseDirectiveCFIDefCfaOffset 4261 /// ::= .cfi_def_cfa_offset offset 4262 bool AsmParser::parseDirectiveCFIDefCfaOffset(SMLoc DirectiveLoc) { 4263 int64_t Offset = 0; 4264 if (parseAbsoluteExpression(Offset) || parseEOL()) 4265 return true; 4266 4267 getStreamer().emitCFIDefCfaOffset(Offset, DirectiveLoc); 4268 return false; 4269 } 4270 4271 /// parseDirectiveCFIRegister 4272 /// ::= .cfi_register register, register 4273 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { 4274 int64_t Register1 = 0, Register2 = 0; 4275 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || parseComma() || 4276 parseRegisterOrRegisterNumber(Register2, DirectiveLoc) || parseEOL()) 4277 return true; 4278 4279 getStreamer().emitCFIRegister(Register1, Register2, DirectiveLoc); 4280 return false; 4281 } 4282 4283 /// parseDirectiveCFIWindowSave 4284 /// ::= .cfi_window_save 4285 bool AsmParser::parseDirectiveCFIWindowSave(SMLoc DirectiveLoc) { 4286 if (parseEOL()) 4287 return true; 4288 getStreamer().emitCFIWindowSave(DirectiveLoc); 4289 return false; 4290 } 4291 4292 /// parseDirectiveCFIAdjustCfaOffset 4293 /// ::= .cfi_adjust_cfa_offset adjustment 4294 bool AsmParser::parseDirectiveCFIAdjustCfaOffset(SMLoc DirectiveLoc) { 4295 int64_t Adjustment = 0; 4296 if (parseAbsoluteExpression(Adjustment) || parseEOL()) 4297 return true; 4298 4299 getStreamer().emitCFIAdjustCfaOffset(Adjustment, DirectiveLoc); 4300 return false; 4301 } 4302 4303 /// parseDirectiveCFIDefCfaRegister 4304 /// ::= .cfi_def_cfa_register register 4305 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { 4306 int64_t Register = 0; 4307 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4308 return true; 4309 4310 getStreamer().emitCFIDefCfaRegister(Register, DirectiveLoc); 4311 return false; 4312 } 4313 4314 /// parseDirectiveCFILLVMDefAspaceCfa 4315 /// ::= .cfi_llvm_def_aspace_cfa register, offset, address_space 4316 bool AsmParser::parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc) { 4317 int64_t Register = 0, Offset = 0, AddressSpace = 0; 4318 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4319 parseAbsoluteExpression(Offset) || parseComma() || 4320 parseAbsoluteExpression(AddressSpace) || parseEOL()) 4321 return true; 4322 4323 getStreamer().emitCFILLVMDefAspaceCfa(Register, Offset, AddressSpace, 4324 DirectiveLoc); 4325 return false; 4326 } 4327 4328 /// parseDirectiveCFIOffset 4329 /// ::= .cfi_offset register, offset 4330 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { 4331 int64_t Register = 0; 4332 int64_t Offset = 0; 4333 4334 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4335 parseAbsoluteExpression(Offset) || parseEOL()) 4336 return true; 4337 4338 getStreamer().emitCFIOffset(Register, Offset, DirectiveLoc); 4339 return false; 4340 } 4341 4342 /// parseDirectiveCFIRelOffset 4343 /// ::= .cfi_rel_offset register, offset 4344 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { 4345 int64_t Register = 0, Offset = 0; 4346 4347 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4348 parseAbsoluteExpression(Offset) || parseEOL()) 4349 return true; 4350 4351 getStreamer().emitCFIRelOffset(Register, Offset, DirectiveLoc); 4352 return false; 4353 } 4354 4355 static bool isValidEncoding(int64_t Encoding) { 4356 if (Encoding & ~0xff) 4357 return false; 4358 4359 if (Encoding == dwarf::DW_EH_PE_omit) 4360 return true; 4361 4362 const unsigned Format = Encoding & 0xf; 4363 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 4364 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 4365 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 4366 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 4367 return false; 4368 4369 const unsigned Application = Encoding & 0x70; 4370 if (Application != dwarf::DW_EH_PE_absptr && 4371 Application != dwarf::DW_EH_PE_pcrel) 4372 return false; 4373 4374 return true; 4375 } 4376 4377 /// parseDirectiveCFIPersonalityOrLsda 4378 /// IsPersonality true for cfi_personality, false for cfi_lsda 4379 /// ::= .cfi_personality encoding, [symbol_name] 4380 /// ::= .cfi_lsda encoding, [symbol_name] 4381 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { 4382 int64_t Encoding = 0; 4383 if (parseAbsoluteExpression(Encoding)) 4384 return true; 4385 if (Encoding == dwarf::DW_EH_PE_omit) 4386 return false; 4387 4388 StringRef Name; 4389 if (check(!isValidEncoding(Encoding), "unsupported encoding.") || 4390 parseComma() || 4391 check(parseIdentifier(Name), "expected identifier in directive") || 4392 parseEOL()) 4393 return true; 4394 4395 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4396 4397 if (IsPersonality) 4398 getStreamer().emitCFIPersonality(Sym, Encoding); 4399 else 4400 getStreamer().emitCFILsda(Sym, Encoding); 4401 return false; 4402 } 4403 4404 /// parseDirectiveCFIRememberState 4405 /// ::= .cfi_remember_state 4406 bool AsmParser::parseDirectiveCFIRememberState(SMLoc DirectiveLoc) { 4407 if (parseEOL()) 4408 return true; 4409 getStreamer().emitCFIRememberState(DirectiveLoc); 4410 return false; 4411 } 4412 4413 /// parseDirectiveCFIRestoreState 4414 /// ::= .cfi_remember_state 4415 bool AsmParser::parseDirectiveCFIRestoreState(SMLoc DirectiveLoc) { 4416 if (parseEOL()) 4417 return true; 4418 getStreamer().emitCFIRestoreState(DirectiveLoc); 4419 return false; 4420 } 4421 4422 /// parseDirectiveCFISameValue 4423 /// ::= .cfi_same_value register 4424 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { 4425 int64_t Register = 0; 4426 4427 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4428 return true; 4429 4430 getStreamer().emitCFISameValue(Register, DirectiveLoc); 4431 return false; 4432 } 4433 4434 /// parseDirectiveCFIRestore 4435 /// ::= .cfi_restore register 4436 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { 4437 int64_t Register = 0; 4438 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4439 return true; 4440 4441 getStreamer().emitCFIRestore(Register, DirectiveLoc); 4442 return false; 4443 } 4444 4445 /// parseDirectiveCFIEscape 4446 /// ::= .cfi_escape expression[,...] 4447 bool AsmParser::parseDirectiveCFIEscape(SMLoc DirectiveLoc) { 4448 std::string Values; 4449 int64_t CurrValue; 4450 if (parseAbsoluteExpression(CurrValue)) 4451 return true; 4452 4453 Values.push_back((uint8_t)CurrValue); 4454 4455 while (getLexer().is(AsmToken::Comma)) { 4456 Lex(); 4457 4458 if (parseAbsoluteExpression(CurrValue)) 4459 return true; 4460 4461 Values.push_back((uint8_t)CurrValue); 4462 } 4463 4464 getStreamer().emitCFIEscape(Values, DirectiveLoc); 4465 return false; 4466 } 4467 4468 /// parseDirectiveCFIReturnColumn 4469 /// ::= .cfi_return_column register 4470 bool AsmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { 4471 int64_t Register = 0; 4472 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4473 return true; 4474 getStreamer().emitCFIReturnColumn(Register); 4475 return false; 4476 } 4477 4478 /// parseDirectiveCFISignalFrame 4479 /// ::= .cfi_signal_frame 4480 bool AsmParser::parseDirectiveCFISignalFrame(SMLoc DirectiveLoc) { 4481 if (parseEOL()) 4482 return true; 4483 4484 getStreamer().emitCFISignalFrame(); 4485 return false; 4486 } 4487 4488 /// parseDirectiveCFIUndefined 4489 /// ::= .cfi_undefined register 4490 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { 4491 int64_t Register = 0; 4492 4493 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4494 return true; 4495 4496 getStreamer().emitCFIUndefined(Register, DirectiveLoc); 4497 return false; 4498 } 4499 4500 /// parseDirectiveAltmacro 4501 /// ::= .altmacro 4502 /// ::= .noaltmacro 4503 bool AsmParser::parseDirectiveAltmacro(StringRef Directive) { 4504 if (parseEOL()) 4505 return true; 4506 AltMacroMode = (Directive == ".altmacro"); 4507 return false; 4508 } 4509 4510 /// parseDirectiveMacrosOnOff 4511 /// ::= .macros_on 4512 /// ::= .macros_off 4513 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) { 4514 if (parseEOL()) 4515 return true; 4516 setMacrosEnabled(Directive == ".macros_on"); 4517 return false; 4518 } 4519 4520 /// parseDirectiveMacro 4521 /// ::= .macro name[,] [parameters] 4522 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) { 4523 StringRef Name; 4524 if (parseIdentifier(Name)) 4525 return TokError("expected identifier in '.macro' directive"); 4526 4527 if (getLexer().is(AsmToken::Comma)) 4528 Lex(); 4529 4530 MCAsmMacroParameters Parameters; 4531 while (getLexer().isNot(AsmToken::EndOfStatement)) { 4532 4533 if (!Parameters.empty() && Parameters.back().Vararg) 4534 return Error(Lexer.getLoc(), "vararg parameter '" + 4535 Parameters.back().Name + 4536 "' should be the last parameter"); 4537 4538 MCAsmMacroParameter Parameter; 4539 if (parseIdentifier(Parameter.Name)) 4540 return TokError("expected identifier in '.macro' directive"); 4541 4542 // Emit an error if two (or more) named parameters share the same name 4543 for (const MCAsmMacroParameter& CurrParam : Parameters) 4544 if (CurrParam.Name.equals(Parameter.Name)) 4545 return TokError("macro '" + Name + "' has multiple parameters" 4546 " named '" + Parameter.Name + "'"); 4547 4548 if (Lexer.is(AsmToken::Colon)) { 4549 Lex(); // consume ':' 4550 4551 SMLoc QualLoc; 4552 StringRef Qualifier; 4553 4554 QualLoc = Lexer.getLoc(); 4555 if (parseIdentifier(Qualifier)) 4556 return Error(QualLoc, "missing parameter qualifier for " 4557 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4558 4559 if (Qualifier == "req") 4560 Parameter.Required = true; 4561 else if (Qualifier == "vararg") 4562 Parameter.Vararg = true; 4563 else 4564 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier " 4565 "for '" + Parameter.Name + "' in macro '" + Name + "'"); 4566 } 4567 4568 if (getLexer().is(AsmToken::Equal)) { 4569 Lex(); 4570 4571 SMLoc ParamLoc; 4572 4573 ParamLoc = Lexer.getLoc(); 4574 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false )) 4575 return true; 4576 4577 if (Parameter.Required) 4578 Warning(ParamLoc, "pointless default value for required parameter " 4579 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4580 } 4581 4582 Parameters.push_back(std::move(Parameter)); 4583 4584 if (getLexer().is(AsmToken::Comma)) 4585 Lex(); 4586 } 4587 4588 // Eat just the end of statement. 4589 Lexer.Lex(); 4590 4591 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors 4592 AsmToken EndToken, StartToken = getTok(); 4593 unsigned MacroDepth = 0; 4594 // Lex the macro definition. 4595 while (true) { 4596 // Ignore Lexing errors in macros. 4597 while (Lexer.is(AsmToken::Error)) { 4598 Lexer.Lex(); 4599 } 4600 4601 // Check whether we have reached the end of the file. 4602 if (getLexer().is(AsmToken::Eof)) 4603 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 4604 4605 // Otherwise, check whether we have reach the .endmacro or the start of a 4606 // preprocessor line marker. 4607 if (getLexer().is(AsmToken::Identifier)) { 4608 if (getTok().getIdentifier() == ".endm" || 4609 getTok().getIdentifier() == ".endmacro") { 4610 if (MacroDepth == 0) { // Outermost macro. 4611 EndToken = getTok(); 4612 Lexer.Lex(); 4613 if (getLexer().isNot(AsmToken::EndOfStatement)) 4614 return TokError("unexpected token in '" + EndToken.getIdentifier() + 4615 "' directive"); 4616 break; 4617 } else { 4618 // Otherwise we just found the end of an inner macro. 4619 --MacroDepth; 4620 } 4621 } else if (getTok().getIdentifier() == ".macro") { 4622 // We allow nested macros. Those aren't instantiated until the outermost 4623 // macro is expanded so just ignore them for now. 4624 ++MacroDepth; 4625 } 4626 } else if (Lexer.is(AsmToken::HashDirective)) { 4627 (void)parseCppHashLineFilenameComment(getLexer().getLoc()); 4628 } 4629 4630 // Otherwise, scan til the end of the statement. 4631 eatToEndOfStatement(); 4632 } 4633 4634 if (getContext().lookupMacro(Name)) { 4635 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 4636 } 4637 4638 const char *BodyStart = StartToken.getLoc().getPointer(); 4639 const char *BodyEnd = EndToken.getLoc().getPointer(); 4640 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 4641 checkForBadMacro(DirectiveLoc, Name, Body, Parameters); 4642 MCAsmMacro Macro(Name, Body, std::move(Parameters)); 4643 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n"; 4644 Macro.dump()); 4645 getContext().defineMacro(Name, std::move(Macro)); 4646 return false; 4647 } 4648 4649 /// checkForBadMacro 4650 /// 4651 /// With the support added for named parameters there may be code out there that 4652 /// is transitioning from positional parameters. In versions of gas that did 4653 /// not support named parameters they would be ignored on the macro definition. 4654 /// But to support both styles of parameters this is not possible so if a macro 4655 /// definition has named parameters but does not use them and has what appears 4656 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a 4657 /// warning that the positional parameter found in body which have no effect. 4658 /// Hoping the developer will either remove the named parameters from the macro 4659 /// definition so the positional parameters get used if that was what was 4660 /// intended or change the macro to use the named parameters. It is possible 4661 /// this warning will trigger when the none of the named parameters are used 4662 /// and the strings like $1 are infact to simply to be passed trough unchanged. 4663 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, 4664 StringRef Body, 4665 ArrayRef<MCAsmMacroParameter> Parameters) { 4666 // If this macro is not defined with named parameters the warning we are 4667 // checking for here doesn't apply. 4668 unsigned NParameters = Parameters.size(); 4669 if (NParameters == 0) 4670 return; 4671 4672 bool NamedParametersFound = false; 4673 bool PositionalParametersFound = false; 4674 4675 // Look at the body of the macro for use of both the named parameters and what 4676 // are likely to be positional parameters. This is what expandMacro() is 4677 // doing when it finds the parameters in the body. 4678 while (!Body.empty()) { 4679 // Scan for the next possible parameter. 4680 std::size_t End = Body.size(), Pos = 0; 4681 for (; Pos != End; ++Pos) { 4682 // Check for a substitution or escape. 4683 // This macro is defined with parameters, look for \foo, \bar, etc. 4684 if (Body[Pos] == '\\' && Pos + 1 != End) 4685 break; 4686 4687 // This macro should have parameters, but look for $0, $1, ..., $n too. 4688 if (Body[Pos] != '$' || Pos + 1 == End) 4689 continue; 4690 char Next = Body[Pos + 1]; 4691 if (Next == '$' || Next == 'n' || 4692 isdigit(static_cast<unsigned char>(Next))) 4693 break; 4694 } 4695 4696 // Check if we reached the end. 4697 if (Pos == End) 4698 break; 4699 4700 if (Body[Pos] == '$') { 4701 switch (Body[Pos + 1]) { 4702 // $$ => $ 4703 case '$': 4704 break; 4705 4706 // $n => number of arguments 4707 case 'n': 4708 PositionalParametersFound = true; 4709 break; 4710 4711 // $[0-9] => argument 4712 default: { 4713 PositionalParametersFound = true; 4714 break; 4715 } 4716 } 4717 Pos += 2; 4718 } else { 4719 unsigned I = Pos + 1; 4720 while (isIdentifierChar(Body[I]) && I + 1 != End) 4721 ++I; 4722 4723 const char *Begin = Body.data() + Pos + 1; 4724 StringRef Argument(Begin, I - (Pos + 1)); 4725 unsigned Index = 0; 4726 for (; Index < NParameters; ++Index) 4727 if (Parameters[Index].Name == Argument) 4728 break; 4729 4730 if (Index == NParameters) { 4731 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 4732 Pos += 3; 4733 else { 4734 Pos = I; 4735 } 4736 } else { 4737 NamedParametersFound = true; 4738 Pos += 1 + Argument.size(); 4739 } 4740 } 4741 // Update the scan point. 4742 Body = Body.substr(Pos); 4743 } 4744 4745 if (!NamedParametersFound && PositionalParametersFound) 4746 Warning(DirectiveLoc, "macro defined with named parameters which are not " 4747 "used in macro body, possible positional parameter " 4748 "found in body which will have no effect"); 4749 } 4750 4751 /// parseDirectiveExitMacro 4752 /// ::= .exitm 4753 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) { 4754 if (parseEOL()) 4755 return true; 4756 4757 if (!isInsideMacroInstantiation()) 4758 return TokError("unexpected '" + Directive + "' in file, " 4759 "no current macro definition"); 4760 4761 // Exit all conditionals that are active in the current macro. 4762 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { 4763 TheCondState = TheCondStack.back(); 4764 TheCondStack.pop_back(); 4765 } 4766 4767 handleMacroExit(); 4768 return false; 4769 } 4770 4771 /// parseDirectiveEndMacro 4772 /// ::= .endm 4773 /// ::= .endmacro 4774 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) { 4775 if (getLexer().isNot(AsmToken::EndOfStatement)) 4776 return TokError("unexpected token in '" + Directive + "' directive"); 4777 4778 // If we are inside a macro instantiation, terminate the current 4779 // instantiation. 4780 if (isInsideMacroInstantiation()) { 4781 handleMacroExit(); 4782 return false; 4783 } 4784 4785 // Otherwise, this .endmacro is a stray entry in the file; well formed 4786 // .endmacro directives are handled during the macro definition parsing. 4787 return TokError("unexpected '" + Directive + "' in file, " 4788 "no current macro definition"); 4789 } 4790 4791 /// parseDirectivePurgeMacro 4792 /// ::= .purgem name 4793 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { 4794 StringRef Name; 4795 SMLoc Loc; 4796 if (parseTokenLoc(Loc) || 4797 check(parseIdentifier(Name), Loc, 4798 "expected identifier in '.purgem' directive") || 4799 parseEOL()) 4800 return true; 4801 4802 if (!getContext().lookupMacro(Name)) 4803 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 4804 4805 getContext().undefineMacro(Name); 4806 DEBUG_WITH_TYPE("asm-macros", dbgs() 4807 << "Un-defining macro: " << Name << "\n"); 4808 return false; 4809 } 4810 4811 /// parseDirectiveBundleAlignMode 4812 /// ::= {.bundle_align_mode} expression 4813 bool AsmParser::parseDirectiveBundleAlignMode() { 4814 // Expect a single argument: an expression that evaluates to a constant 4815 // in the inclusive range 0-30. 4816 SMLoc ExprLoc = getLexer().getLoc(); 4817 int64_t AlignSizePow2; 4818 if (checkForValidSection() || parseAbsoluteExpression(AlignSizePow2) || 4819 parseEOL() || 4820 check(AlignSizePow2 < 0 || AlignSizePow2 > 30, ExprLoc, 4821 "invalid bundle alignment size (expected between 0 and 30)")) 4822 return true; 4823 4824 getStreamer().emitBundleAlignMode(Align(1ULL << AlignSizePow2)); 4825 return false; 4826 } 4827 4828 /// parseDirectiveBundleLock 4829 /// ::= {.bundle_lock} [align_to_end] 4830 bool AsmParser::parseDirectiveBundleLock() { 4831 if (checkForValidSection()) 4832 return true; 4833 bool AlignToEnd = false; 4834 4835 StringRef Option; 4836 SMLoc Loc = getTok().getLoc(); 4837 const char *kInvalidOptionError = 4838 "invalid option for '.bundle_lock' directive"; 4839 4840 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4841 if (check(parseIdentifier(Option), Loc, kInvalidOptionError) || 4842 check(Option != "align_to_end", Loc, kInvalidOptionError) || parseEOL()) 4843 return true; 4844 AlignToEnd = true; 4845 } 4846 4847 getStreamer().emitBundleLock(AlignToEnd); 4848 return false; 4849 } 4850 4851 /// parseDirectiveBundleLock 4852 /// ::= {.bundle_lock} 4853 bool AsmParser::parseDirectiveBundleUnlock() { 4854 if (checkForValidSection() || parseEOL()) 4855 return true; 4856 4857 getStreamer().emitBundleUnlock(); 4858 return false; 4859 } 4860 4861 /// parseDirectiveSpace 4862 /// ::= (.skip | .space) expression [ , expression ] 4863 bool AsmParser::parseDirectiveSpace(StringRef IDVal) { 4864 SMLoc NumBytesLoc = Lexer.getLoc(); 4865 const MCExpr *NumBytes; 4866 if (checkForValidSection() || parseExpression(NumBytes)) 4867 return true; 4868 4869 int64_t FillExpr = 0; 4870 if (parseOptionalToken(AsmToken::Comma)) 4871 if (parseAbsoluteExpression(FillExpr)) 4872 return true; 4873 if (parseEOL()) 4874 return true; 4875 4876 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 4877 getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc); 4878 4879 return false; 4880 } 4881 4882 /// parseDirectiveDCB 4883 /// ::= .dcb.{b, l, w} expression, expression 4884 bool AsmParser::parseDirectiveDCB(StringRef IDVal, unsigned Size) { 4885 SMLoc NumValuesLoc = Lexer.getLoc(); 4886 int64_t NumValues; 4887 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4888 return true; 4889 4890 if (NumValues < 0) { 4891 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4892 return false; 4893 } 4894 4895 if (parseComma()) 4896 return true; 4897 4898 const MCExpr *Value; 4899 SMLoc ExprLoc = getLexer().getLoc(); 4900 if (parseExpression(Value)) 4901 return true; 4902 4903 // Special case constant expressions to match code generator. 4904 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4905 assert(Size <= 8 && "Invalid size"); 4906 uint64_t IntValue = MCE->getValue(); 4907 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 4908 return Error(ExprLoc, "literal value out of range for directive"); 4909 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4910 getStreamer().emitIntValue(IntValue, Size); 4911 } else { 4912 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4913 getStreamer().emitValue(Value, Size, ExprLoc); 4914 } 4915 4916 return parseEOL(); 4917 } 4918 4919 /// parseDirectiveRealDCB 4920 /// ::= .dcb.{d, s} expression, expression 4921 bool AsmParser::parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &Semantics) { 4922 SMLoc NumValuesLoc = Lexer.getLoc(); 4923 int64_t NumValues; 4924 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4925 return true; 4926 4927 if (NumValues < 0) { 4928 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4929 return false; 4930 } 4931 4932 if (parseComma()) 4933 return true; 4934 4935 APInt AsInt; 4936 if (parseRealValue(Semantics, AsInt) || parseEOL()) 4937 return true; 4938 4939 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4940 getStreamer().emitIntValue(AsInt.getLimitedValue(), 4941 AsInt.getBitWidth() / 8); 4942 4943 return false; 4944 } 4945 4946 /// parseDirectiveDS 4947 /// ::= .ds.{b, d, l, p, s, w, x} expression 4948 bool AsmParser::parseDirectiveDS(StringRef IDVal, unsigned Size) { 4949 SMLoc NumValuesLoc = Lexer.getLoc(); 4950 int64_t NumValues; 4951 if (checkForValidSection() || parseAbsoluteExpression(NumValues) || 4952 parseEOL()) 4953 return true; 4954 4955 if (NumValues < 0) { 4956 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4957 return false; 4958 } 4959 4960 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4961 getStreamer().emitFill(Size, 0); 4962 4963 return false; 4964 } 4965 4966 /// parseDirectiveLEB128 4967 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ] 4968 bool AsmParser::parseDirectiveLEB128(bool Signed) { 4969 if (checkForValidSection()) 4970 return true; 4971 4972 auto parseOp = [&]() -> bool { 4973 const MCExpr *Value; 4974 if (parseExpression(Value)) 4975 return true; 4976 if (Signed) 4977 getStreamer().emitSLEB128Value(Value); 4978 else 4979 getStreamer().emitULEB128Value(Value); 4980 return false; 4981 }; 4982 4983 return parseMany(parseOp); 4984 } 4985 4986 /// parseDirectiveSymbolAttribute 4987 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 4988 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 4989 auto parseOp = [&]() -> bool { 4990 StringRef Name; 4991 SMLoc Loc = getTok().getLoc(); 4992 if (parseIdentifier(Name)) 4993 return Error(Loc, "expected identifier"); 4994 4995 if (discardLTOSymbol(Name)) 4996 return false; 4997 4998 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4999 5000 // Assembler local symbols don't make any sense here, except for directives 5001 // that the symbol should be tagged. 5002 if (Sym->isTemporary() && Attr != MCSA_Memtag) 5003 return Error(Loc, "non-local symbol required"); 5004 5005 if (!getStreamer().emitSymbolAttribute(Sym, Attr)) 5006 return Error(Loc, "unable to emit symbol attribute"); 5007 return false; 5008 }; 5009 5010 return parseMany(parseOp); 5011 } 5012 5013 /// parseDirectiveComm 5014 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 5015 bool AsmParser::parseDirectiveComm(bool IsLocal) { 5016 if (checkForValidSection()) 5017 return true; 5018 5019 SMLoc IDLoc = getLexer().getLoc(); 5020 StringRef Name; 5021 if (parseIdentifier(Name)) 5022 return TokError("expected identifier in directive"); 5023 5024 // Handle the identifier as the key symbol. 5025 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5026 5027 if (parseComma()) 5028 return true; 5029 5030 int64_t Size; 5031 SMLoc SizeLoc = getLexer().getLoc(); 5032 if (parseAbsoluteExpression(Size)) 5033 return true; 5034 5035 int64_t Pow2Alignment = 0; 5036 SMLoc Pow2AlignmentLoc; 5037 if (getLexer().is(AsmToken::Comma)) { 5038 Lex(); 5039 Pow2AlignmentLoc = getLexer().getLoc(); 5040 if (parseAbsoluteExpression(Pow2Alignment)) 5041 return true; 5042 5043 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); 5044 if (IsLocal && LCOMM == LCOMM::NoAlignment) 5045 return Error(Pow2AlignmentLoc, "alignment not supported on this target"); 5046 5047 // If this target takes alignments in bytes (not log) validate and convert. 5048 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || 5049 (IsLocal && LCOMM == LCOMM::ByteAlignment)) { 5050 if (!isPowerOf2_64(Pow2Alignment)) 5051 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 5052 Pow2Alignment = Log2_64(Pow2Alignment); 5053 } 5054 } 5055 5056 if (parseEOL()) 5057 return true; 5058 5059 // NOTE: a size of zero for a .comm should create a undefined symbol 5060 // but a size of .lcomm creates a bss symbol of size zero. 5061 if (Size < 0) 5062 return Error(SizeLoc, "size must be non-negative"); 5063 5064 Sym->redefineIfPossible(); 5065 if (!Sym->isUndefined()) 5066 return Error(IDLoc, "invalid symbol redefinition"); 5067 5068 // Create the Symbol as a common or local common with Size and Pow2Alignment 5069 if (IsLocal) { 5070 getStreamer().emitLocalCommonSymbol(Sym, Size, 5071 Align(1ULL << Pow2Alignment)); 5072 return false; 5073 } 5074 5075 getStreamer().emitCommonSymbol(Sym, Size, Align(1ULL << Pow2Alignment)); 5076 return false; 5077 } 5078 5079 /// parseDirectiveAbort 5080 /// ::= .abort [... message ...] 5081 bool AsmParser::parseDirectiveAbort() { 5082 // FIXME: Use loc from directive. 5083 SMLoc Loc = getLexer().getLoc(); 5084 5085 StringRef Str = parseStringToEndOfStatement(); 5086 if (parseEOL()) 5087 return true; 5088 5089 if (Str.empty()) 5090 return Error(Loc, ".abort detected. Assembly stopping."); 5091 else 5092 return Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 5093 // FIXME: Actually abort assembly here. 5094 5095 return false; 5096 } 5097 5098 /// parseDirectiveInclude 5099 /// ::= .include "filename" 5100 bool AsmParser::parseDirectiveInclude() { 5101 // Allow the strings to have escaped octal character sequence. 5102 std::string Filename; 5103 SMLoc IncludeLoc = getTok().getLoc(); 5104 5105 if (check(getTok().isNot(AsmToken::String), 5106 "expected string in '.include' directive") || 5107 parseEscapedString(Filename) || 5108 check(getTok().isNot(AsmToken::EndOfStatement), 5109 "unexpected token in '.include' directive") || 5110 // Attempt to switch the lexer to the included file before consuming the 5111 // end of statement to avoid losing it when we switch. 5112 check(enterIncludeFile(Filename), IncludeLoc, 5113 "Could not find include file '" + Filename + "'")) 5114 return true; 5115 5116 return false; 5117 } 5118 5119 /// parseDirectiveIncbin 5120 /// ::= .incbin "filename" [ , skip [ , count ] ] 5121 bool AsmParser::parseDirectiveIncbin() { 5122 // Allow the strings to have escaped octal character sequence. 5123 std::string Filename; 5124 SMLoc IncbinLoc = getTok().getLoc(); 5125 if (check(getTok().isNot(AsmToken::String), 5126 "expected string in '.incbin' directive") || 5127 parseEscapedString(Filename)) 5128 return true; 5129 5130 int64_t Skip = 0; 5131 const MCExpr *Count = nullptr; 5132 SMLoc SkipLoc, CountLoc; 5133 if (parseOptionalToken(AsmToken::Comma)) { 5134 // The skip expression can be omitted while specifying the count, e.g: 5135 // .incbin "filename",,4 5136 if (getTok().isNot(AsmToken::Comma)) { 5137 if (parseTokenLoc(SkipLoc) || parseAbsoluteExpression(Skip)) 5138 return true; 5139 } 5140 if (parseOptionalToken(AsmToken::Comma)) { 5141 CountLoc = getTok().getLoc(); 5142 if (parseExpression(Count)) 5143 return true; 5144 } 5145 } 5146 5147 if (parseEOL()) 5148 return true; 5149 5150 if (check(Skip < 0, SkipLoc, "skip is negative")) 5151 return true; 5152 5153 // Attempt to process the included file. 5154 if (processIncbinFile(Filename, Skip, Count, CountLoc)) 5155 return Error(IncbinLoc, "Could not find incbin file '" + Filename + "'"); 5156 return false; 5157 } 5158 5159 /// parseDirectiveIf 5160 /// ::= .if{,eq,ge,gt,le,lt,ne} expression 5161 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { 5162 TheCondStack.push_back(TheCondState); 5163 TheCondState.TheCond = AsmCond::IfCond; 5164 if (TheCondState.Ignore) { 5165 eatToEndOfStatement(); 5166 } else { 5167 int64_t ExprValue; 5168 if (parseAbsoluteExpression(ExprValue) || parseEOL()) 5169 return true; 5170 5171 switch (DirKind) { 5172 default: 5173 llvm_unreachable("unsupported directive"); 5174 case DK_IF: 5175 case DK_IFNE: 5176 break; 5177 case DK_IFEQ: 5178 ExprValue = ExprValue == 0; 5179 break; 5180 case DK_IFGE: 5181 ExprValue = ExprValue >= 0; 5182 break; 5183 case DK_IFGT: 5184 ExprValue = ExprValue > 0; 5185 break; 5186 case DK_IFLE: 5187 ExprValue = ExprValue <= 0; 5188 break; 5189 case DK_IFLT: 5190 ExprValue = ExprValue < 0; 5191 break; 5192 } 5193 5194 TheCondState.CondMet = ExprValue; 5195 TheCondState.Ignore = !TheCondState.CondMet; 5196 } 5197 5198 return false; 5199 } 5200 5201 /// parseDirectiveIfb 5202 /// ::= .ifb string 5203 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5204 TheCondStack.push_back(TheCondState); 5205 TheCondState.TheCond = AsmCond::IfCond; 5206 5207 if (TheCondState.Ignore) { 5208 eatToEndOfStatement(); 5209 } else { 5210 StringRef Str = parseStringToEndOfStatement(); 5211 5212 if (parseEOL()) 5213 return true; 5214 5215 TheCondState.CondMet = ExpectBlank == Str.empty(); 5216 TheCondState.Ignore = !TheCondState.CondMet; 5217 } 5218 5219 return false; 5220 } 5221 5222 /// parseDirectiveIfc 5223 /// ::= .ifc string1, string2 5224 /// ::= .ifnc string1, string2 5225 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) { 5226 TheCondStack.push_back(TheCondState); 5227 TheCondState.TheCond = AsmCond::IfCond; 5228 5229 if (TheCondState.Ignore) { 5230 eatToEndOfStatement(); 5231 } else { 5232 StringRef Str1 = parseStringToComma(); 5233 5234 if (parseComma()) 5235 return true; 5236 5237 StringRef Str2 = parseStringToEndOfStatement(); 5238 5239 if (parseEOL()) 5240 return true; 5241 5242 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim()); 5243 TheCondState.Ignore = !TheCondState.CondMet; 5244 } 5245 5246 return false; 5247 } 5248 5249 /// parseDirectiveIfeqs 5250 /// ::= .ifeqs string1, string2 5251 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) { 5252 if (Lexer.isNot(AsmToken::String)) { 5253 if (ExpectEqual) 5254 return TokError("expected string parameter for '.ifeqs' directive"); 5255 return TokError("expected string parameter for '.ifnes' directive"); 5256 } 5257 5258 StringRef String1 = getTok().getStringContents(); 5259 Lex(); 5260 5261 if (Lexer.isNot(AsmToken::Comma)) { 5262 if (ExpectEqual) 5263 return TokError( 5264 "expected comma after first string for '.ifeqs' directive"); 5265 return TokError("expected comma after first string for '.ifnes' directive"); 5266 } 5267 5268 Lex(); 5269 5270 if (Lexer.isNot(AsmToken::String)) { 5271 if (ExpectEqual) 5272 return TokError("expected string parameter for '.ifeqs' directive"); 5273 return TokError("expected string parameter for '.ifnes' directive"); 5274 } 5275 5276 StringRef String2 = getTok().getStringContents(); 5277 Lex(); 5278 5279 TheCondStack.push_back(TheCondState); 5280 TheCondState.TheCond = AsmCond::IfCond; 5281 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5282 TheCondState.Ignore = !TheCondState.CondMet; 5283 5284 return false; 5285 } 5286 5287 /// parseDirectiveIfdef 5288 /// ::= .ifdef symbol 5289 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 5290 StringRef Name; 5291 TheCondStack.push_back(TheCondState); 5292 TheCondState.TheCond = AsmCond::IfCond; 5293 5294 if (TheCondState.Ignore) { 5295 eatToEndOfStatement(); 5296 } else { 5297 if (check(parseIdentifier(Name), "expected identifier after '.ifdef'") || 5298 parseEOL()) 5299 return true; 5300 5301 MCSymbol *Sym = getContext().lookupSymbol(Name); 5302 5303 if (expect_defined) 5304 TheCondState.CondMet = (Sym && !Sym->isUndefined(false)); 5305 else 5306 TheCondState.CondMet = (!Sym || Sym->isUndefined(false)); 5307 TheCondState.Ignore = !TheCondState.CondMet; 5308 } 5309 5310 return false; 5311 } 5312 5313 /// parseDirectiveElseIf 5314 /// ::= .elseif expression 5315 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) { 5316 if (TheCondState.TheCond != AsmCond::IfCond && 5317 TheCondState.TheCond != AsmCond::ElseIfCond) 5318 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an" 5319 " .if or an .elseif"); 5320 TheCondState.TheCond = AsmCond::ElseIfCond; 5321 5322 bool LastIgnoreState = false; 5323 if (!TheCondStack.empty()) 5324 LastIgnoreState = TheCondStack.back().Ignore; 5325 if (LastIgnoreState || TheCondState.CondMet) { 5326 TheCondState.Ignore = true; 5327 eatToEndOfStatement(); 5328 } else { 5329 int64_t ExprValue; 5330 if (parseAbsoluteExpression(ExprValue)) 5331 return true; 5332 5333 if (parseEOL()) 5334 return true; 5335 5336 TheCondState.CondMet = ExprValue; 5337 TheCondState.Ignore = !TheCondState.CondMet; 5338 } 5339 5340 return false; 5341 } 5342 5343 /// parseDirectiveElse 5344 /// ::= .else 5345 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) { 5346 if (parseEOL()) 5347 return true; 5348 5349 if (TheCondState.TheCond != AsmCond::IfCond && 5350 TheCondState.TheCond != AsmCond::ElseIfCond) 5351 return Error(DirectiveLoc, "Encountered a .else that doesn't follow " 5352 " an .if or an .elseif"); 5353 TheCondState.TheCond = AsmCond::ElseCond; 5354 bool LastIgnoreState = false; 5355 if (!TheCondStack.empty()) 5356 LastIgnoreState = TheCondStack.back().Ignore; 5357 if (LastIgnoreState || TheCondState.CondMet) 5358 TheCondState.Ignore = true; 5359 else 5360 TheCondState.Ignore = false; 5361 5362 return false; 5363 } 5364 5365 /// parseDirectiveEnd 5366 /// ::= .end 5367 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { 5368 if (parseEOL()) 5369 return true; 5370 5371 while (Lexer.isNot(AsmToken::Eof)) 5372 Lexer.Lex(); 5373 5374 return false; 5375 } 5376 5377 /// parseDirectiveError 5378 /// ::= .err 5379 /// ::= .error [string] 5380 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) { 5381 if (!TheCondStack.empty()) { 5382 if (TheCondStack.back().Ignore) { 5383 eatToEndOfStatement(); 5384 return false; 5385 } 5386 } 5387 5388 if (!WithMessage) 5389 return Error(L, ".err encountered"); 5390 5391 StringRef Message = ".error directive invoked in source file"; 5392 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5393 if (Lexer.isNot(AsmToken::String)) 5394 return TokError(".error argument must be a string"); 5395 5396 Message = getTok().getStringContents(); 5397 Lex(); 5398 } 5399 5400 return Error(L, Message); 5401 } 5402 5403 /// parseDirectiveWarning 5404 /// ::= .warning [string] 5405 bool AsmParser::parseDirectiveWarning(SMLoc L) { 5406 if (!TheCondStack.empty()) { 5407 if (TheCondStack.back().Ignore) { 5408 eatToEndOfStatement(); 5409 return false; 5410 } 5411 } 5412 5413 StringRef Message = ".warning directive invoked in source file"; 5414 5415 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 5416 if (Lexer.isNot(AsmToken::String)) 5417 return TokError(".warning argument must be a string"); 5418 5419 Message = getTok().getStringContents(); 5420 Lex(); 5421 if (parseEOL()) 5422 return true; 5423 } 5424 5425 return Warning(L, Message); 5426 } 5427 5428 /// parseDirectiveEndIf 5429 /// ::= .endif 5430 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { 5431 if (parseEOL()) 5432 return true; 5433 5434 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) 5435 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow " 5436 "an .if or .else"); 5437 if (!TheCondStack.empty()) { 5438 TheCondState = TheCondStack.back(); 5439 TheCondStack.pop_back(); 5440 } 5441 5442 return false; 5443 } 5444 5445 void AsmParser::initializeDirectiveKindMap() { 5446 /* Lookup will be done with the directive 5447 * converted to lower case, so all these 5448 * keys should be lower case. 5449 * (target specific directives are handled 5450 * elsewhere) 5451 */ 5452 DirectiveKindMap[".set"] = DK_SET; 5453 DirectiveKindMap[".equ"] = DK_EQU; 5454 DirectiveKindMap[".equiv"] = DK_EQUIV; 5455 DirectiveKindMap[".ascii"] = DK_ASCII; 5456 DirectiveKindMap[".asciz"] = DK_ASCIZ; 5457 DirectiveKindMap[".string"] = DK_STRING; 5458 DirectiveKindMap[".byte"] = DK_BYTE; 5459 DirectiveKindMap[".short"] = DK_SHORT; 5460 DirectiveKindMap[".value"] = DK_VALUE; 5461 DirectiveKindMap[".2byte"] = DK_2BYTE; 5462 DirectiveKindMap[".long"] = DK_LONG; 5463 DirectiveKindMap[".int"] = DK_INT; 5464 DirectiveKindMap[".4byte"] = DK_4BYTE; 5465 DirectiveKindMap[".quad"] = DK_QUAD; 5466 DirectiveKindMap[".8byte"] = DK_8BYTE; 5467 DirectiveKindMap[".octa"] = DK_OCTA; 5468 DirectiveKindMap[".single"] = DK_SINGLE; 5469 DirectiveKindMap[".float"] = DK_FLOAT; 5470 DirectiveKindMap[".double"] = DK_DOUBLE; 5471 DirectiveKindMap[".align"] = DK_ALIGN; 5472 DirectiveKindMap[".align32"] = DK_ALIGN32; 5473 DirectiveKindMap[".balign"] = DK_BALIGN; 5474 DirectiveKindMap[".balignw"] = DK_BALIGNW; 5475 DirectiveKindMap[".balignl"] = DK_BALIGNL; 5476 DirectiveKindMap[".p2align"] = DK_P2ALIGN; 5477 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW; 5478 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL; 5479 DirectiveKindMap[".org"] = DK_ORG; 5480 DirectiveKindMap[".fill"] = DK_FILL; 5481 DirectiveKindMap[".zero"] = DK_ZERO; 5482 DirectiveKindMap[".extern"] = DK_EXTERN; 5483 DirectiveKindMap[".globl"] = DK_GLOBL; 5484 DirectiveKindMap[".global"] = DK_GLOBAL; 5485 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE; 5486 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP; 5487 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER; 5488 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN; 5489 DirectiveKindMap[".reference"] = DK_REFERENCE; 5490 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION; 5491 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE; 5492 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN; 5493 DirectiveKindMap[".cold"] = DK_COLD; 5494 DirectiveKindMap[".comm"] = DK_COMM; 5495 DirectiveKindMap[".common"] = DK_COMMON; 5496 DirectiveKindMap[".lcomm"] = DK_LCOMM; 5497 DirectiveKindMap[".abort"] = DK_ABORT; 5498 DirectiveKindMap[".include"] = DK_INCLUDE; 5499 DirectiveKindMap[".incbin"] = DK_INCBIN; 5500 DirectiveKindMap[".code16"] = DK_CODE16; 5501 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC; 5502 DirectiveKindMap[".rept"] = DK_REPT; 5503 DirectiveKindMap[".rep"] = DK_REPT; 5504 DirectiveKindMap[".irp"] = DK_IRP; 5505 DirectiveKindMap[".irpc"] = DK_IRPC; 5506 DirectiveKindMap[".endr"] = DK_ENDR; 5507 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE; 5508 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK; 5509 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK; 5510 DirectiveKindMap[".if"] = DK_IF; 5511 DirectiveKindMap[".ifeq"] = DK_IFEQ; 5512 DirectiveKindMap[".ifge"] = DK_IFGE; 5513 DirectiveKindMap[".ifgt"] = DK_IFGT; 5514 DirectiveKindMap[".ifle"] = DK_IFLE; 5515 DirectiveKindMap[".iflt"] = DK_IFLT; 5516 DirectiveKindMap[".ifne"] = DK_IFNE; 5517 DirectiveKindMap[".ifb"] = DK_IFB; 5518 DirectiveKindMap[".ifnb"] = DK_IFNB; 5519 DirectiveKindMap[".ifc"] = DK_IFC; 5520 DirectiveKindMap[".ifeqs"] = DK_IFEQS; 5521 DirectiveKindMap[".ifnc"] = DK_IFNC; 5522 DirectiveKindMap[".ifnes"] = DK_IFNES; 5523 DirectiveKindMap[".ifdef"] = DK_IFDEF; 5524 DirectiveKindMap[".ifndef"] = DK_IFNDEF; 5525 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF; 5526 DirectiveKindMap[".elseif"] = DK_ELSEIF; 5527 DirectiveKindMap[".else"] = DK_ELSE; 5528 DirectiveKindMap[".end"] = DK_END; 5529 DirectiveKindMap[".endif"] = DK_ENDIF; 5530 DirectiveKindMap[".skip"] = DK_SKIP; 5531 DirectiveKindMap[".space"] = DK_SPACE; 5532 DirectiveKindMap[".file"] = DK_FILE; 5533 DirectiveKindMap[".line"] = DK_LINE; 5534 DirectiveKindMap[".loc"] = DK_LOC; 5535 DirectiveKindMap[".stabs"] = DK_STABS; 5536 DirectiveKindMap[".cv_file"] = DK_CV_FILE; 5537 DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; 5538 DirectiveKindMap[".cv_loc"] = DK_CV_LOC; 5539 DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; 5540 DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; 5541 DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; 5542 DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; 5543 DirectiveKindMap[".cv_string"] = DK_CV_STRING; 5544 DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; 5545 DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; 5546 DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; 5547 DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; 5548 DirectiveKindMap[".sleb128"] = DK_SLEB128; 5549 DirectiveKindMap[".uleb128"] = DK_ULEB128; 5550 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; 5551 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; 5552 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; 5553 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; 5554 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; 5555 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; 5556 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; 5557 DirectiveKindMap[".cfi_llvm_def_aspace_cfa"] = DK_CFI_LLVM_DEF_ASPACE_CFA; 5558 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; 5559 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; 5560 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; 5561 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; 5562 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; 5563 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; 5564 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; 5565 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; 5566 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; 5567 DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; 5568 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; 5569 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; 5570 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; 5571 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; 5572 DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; 5573 DirectiveKindMap[".cfi_mte_tagged_frame"] = DK_CFI_MTE_TAGGED_FRAME; 5574 DirectiveKindMap[".macros_on"] = DK_MACROS_ON; 5575 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF; 5576 DirectiveKindMap[".macro"] = DK_MACRO; 5577 DirectiveKindMap[".exitm"] = DK_EXITM; 5578 DirectiveKindMap[".endm"] = DK_ENDM; 5579 DirectiveKindMap[".endmacro"] = DK_ENDMACRO; 5580 DirectiveKindMap[".purgem"] = DK_PURGEM; 5581 DirectiveKindMap[".err"] = DK_ERR; 5582 DirectiveKindMap[".error"] = DK_ERROR; 5583 DirectiveKindMap[".warning"] = DK_WARNING; 5584 DirectiveKindMap[".altmacro"] = DK_ALTMACRO; 5585 DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO; 5586 DirectiveKindMap[".reloc"] = DK_RELOC; 5587 DirectiveKindMap[".dc"] = DK_DC; 5588 DirectiveKindMap[".dc.a"] = DK_DC_A; 5589 DirectiveKindMap[".dc.b"] = DK_DC_B; 5590 DirectiveKindMap[".dc.d"] = DK_DC_D; 5591 DirectiveKindMap[".dc.l"] = DK_DC_L; 5592 DirectiveKindMap[".dc.s"] = DK_DC_S; 5593 DirectiveKindMap[".dc.w"] = DK_DC_W; 5594 DirectiveKindMap[".dc.x"] = DK_DC_X; 5595 DirectiveKindMap[".dcb"] = DK_DCB; 5596 DirectiveKindMap[".dcb.b"] = DK_DCB_B; 5597 DirectiveKindMap[".dcb.d"] = DK_DCB_D; 5598 DirectiveKindMap[".dcb.l"] = DK_DCB_L; 5599 DirectiveKindMap[".dcb.s"] = DK_DCB_S; 5600 DirectiveKindMap[".dcb.w"] = DK_DCB_W; 5601 DirectiveKindMap[".dcb.x"] = DK_DCB_X; 5602 DirectiveKindMap[".ds"] = DK_DS; 5603 DirectiveKindMap[".ds.b"] = DK_DS_B; 5604 DirectiveKindMap[".ds.d"] = DK_DS_D; 5605 DirectiveKindMap[".ds.l"] = DK_DS_L; 5606 DirectiveKindMap[".ds.p"] = DK_DS_P; 5607 DirectiveKindMap[".ds.s"] = DK_DS_S; 5608 DirectiveKindMap[".ds.w"] = DK_DS_W; 5609 DirectiveKindMap[".ds.x"] = DK_DS_X; 5610 DirectiveKindMap[".print"] = DK_PRINT; 5611 DirectiveKindMap[".addrsig"] = DK_ADDRSIG; 5612 DirectiveKindMap[".addrsig_sym"] = DK_ADDRSIG_SYM; 5613 DirectiveKindMap[".pseudoprobe"] = DK_PSEUDO_PROBE; 5614 DirectiveKindMap[".lto_discard"] = DK_LTO_DISCARD; 5615 DirectiveKindMap[".lto_set_conditional"] = DK_LTO_SET_CONDITIONAL; 5616 DirectiveKindMap[".memtag"] = DK_MEMTAG; 5617 } 5618 5619 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { 5620 AsmToken EndToken, StartToken = getTok(); 5621 5622 unsigned NestLevel = 0; 5623 while (true) { 5624 // Check whether we have reached the end of the file. 5625 if (getLexer().is(AsmToken::Eof)) { 5626 printError(DirectiveLoc, "no matching '.endr' in definition"); 5627 return nullptr; 5628 } 5629 5630 if (Lexer.is(AsmToken::Identifier) && 5631 (getTok().getIdentifier() == ".rep" || 5632 getTok().getIdentifier() == ".rept" || 5633 getTok().getIdentifier() == ".irp" || 5634 getTok().getIdentifier() == ".irpc")) { 5635 ++NestLevel; 5636 } 5637 5638 // Otherwise, check whether we have reached the .endr. 5639 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") { 5640 if (NestLevel == 0) { 5641 EndToken = getTok(); 5642 Lex(); 5643 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5644 printError(getTok().getLoc(), 5645 "unexpected token in '.endr' directive"); 5646 return nullptr; 5647 } 5648 break; 5649 } 5650 --NestLevel; 5651 } 5652 5653 // Otherwise, scan till the end of the statement. 5654 eatToEndOfStatement(); 5655 } 5656 5657 const char *BodyStart = StartToken.getLoc().getPointer(); 5658 const char *BodyEnd = EndToken.getLoc().getPointer(); 5659 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 5660 5661 // We Are Anonymous. 5662 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters()); 5663 return &MacroLikeBodies.back(); 5664 } 5665 5666 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 5667 raw_svector_ostream &OS) { 5668 OS << ".endr\n"; 5669 5670 std::unique_ptr<MemoryBuffer> Instantiation = 5671 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 5672 5673 // Create the macro instantiation object and add to the current macro 5674 // instantiation stack. 5675 MacroInstantiation *MI = new MacroInstantiation{ 5676 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 5677 ActiveMacros.push_back(MI); 5678 5679 // Jump to the macro instantiation and prime the lexer. 5680 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 5681 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 5682 Lex(); 5683 } 5684 5685 /// parseDirectiveRept 5686 /// ::= .rep | .rept count 5687 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) { 5688 const MCExpr *CountExpr; 5689 SMLoc CountLoc = getTok().getLoc(); 5690 if (parseExpression(CountExpr)) 5691 return true; 5692 5693 int64_t Count; 5694 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) { 5695 return Error(CountLoc, "unexpected token in '" + Dir + "' directive"); 5696 } 5697 5698 if (check(Count < 0, CountLoc, "Count is negative") || parseEOL()) 5699 return true; 5700 5701 // Lex the rept definition. 5702 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5703 if (!M) 5704 return true; 5705 5706 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5707 // to hold the macro body with substitutions. 5708 SmallString<256> Buf; 5709 raw_svector_ostream OS(Buf); 5710 while (Count--) { 5711 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t). 5712 if (expandMacro(OS, M->Body, std::nullopt, std::nullopt, false, 5713 getTok().getLoc())) 5714 return true; 5715 } 5716 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5717 5718 return false; 5719 } 5720 5721 /// parseDirectiveIrp 5722 /// ::= .irp symbol,values 5723 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) { 5724 MCAsmMacroParameter Parameter; 5725 MCAsmMacroArguments A; 5726 if (check(parseIdentifier(Parameter.Name), 5727 "expected identifier in '.irp' directive") || 5728 parseComma() || parseMacroArguments(nullptr, A) || parseEOL()) 5729 return true; 5730 5731 // Lex the irp definition. 5732 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5733 if (!M) 5734 return true; 5735 5736 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5737 // to hold the macro body with substitutions. 5738 SmallString<256> Buf; 5739 raw_svector_ostream OS(Buf); 5740 5741 for (const MCAsmMacroArgument &Arg : A) { 5742 // Note that the AtPseudoVariable is enabled for instantiations of .irp. 5743 // This is undocumented, but GAS seems to support it. 5744 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5745 return true; 5746 } 5747 5748 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5749 5750 return false; 5751 } 5752 5753 /// parseDirectiveIrpc 5754 /// ::= .irpc symbol,values 5755 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) { 5756 MCAsmMacroParameter Parameter; 5757 MCAsmMacroArguments A; 5758 5759 if (check(parseIdentifier(Parameter.Name), 5760 "expected identifier in '.irpc' directive") || 5761 parseComma() || parseMacroArguments(nullptr, A)) 5762 return true; 5763 5764 if (A.size() != 1 || A.front().size() != 1) 5765 return TokError("unexpected token in '.irpc' directive"); 5766 if (parseEOL()) 5767 return true; 5768 5769 // Lex the irpc definition. 5770 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5771 if (!M) 5772 return true; 5773 5774 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5775 // to hold the macro body with substitutions. 5776 SmallString<256> Buf; 5777 raw_svector_ostream OS(Buf); 5778 5779 StringRef Values = A.front().front().getString(); 5780 for (std::size_t I = 0, End = Values.size(); I != End; ++I) { 5781 MCAsmMacroArgument Arg; 5782 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1)); 5783 5784 // Note that the AtPseudoVariable is enabled for instantiations of .irpc. 5785 // This is undocumented, but GAS seems to support it. 5786 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5787 return true; 5788 } 5789 5790 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5791 5792 return false; 5793 } 5794 5795 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) { 5796 if (ActiveMacros.empty()) 5797 return TokError("unmatched '.endr' directive"); 5798 5799 // The only .repl that should get here are the ones created by 5800 // instantiateMacroLikeBody. 5801 assert(getLexer().is(AsmToken::EndOfStatement)); 5802 5803 handleMacroExit(); 5804 return false; 5805 } 5806 5807 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, 5808 size_t Len) { 5809 const MCExpr *Value; 5810 SMLoc ExprLoc = getLexer().getLoc(); 5811 if (parseExpression(Value)) 5812 return true; 5813 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5814 if (!MCE) 5815 return Error(ExprLoc, "unexpected expression in _emit"); 5816 uint64_t IntValue = MCE->getValue(); 5817 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue)) 5818 return Error(ExprLoc, "literal value out of range for directive"); 5819 5820 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len); 5821 return false; 5822 } 5823 5824 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { 5825 const MCExpr *Value; 5826 SMLoc ExprLoc = getLexer().getLoc(); 5827 if (parseExpression(Value)) 5828 return true; 5829 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5830 if (!MCE) 5831 return Error(ExprLoc, "unexpected expression in align"); 5832 uint64_t IntValue = MCE->getValue(); 5833 if (!isPowerOf2_64(IntValue)) 5834 return Error(ExprLoc, "literal value not a power of two greater then zero"); 5835 5836 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue)); 5837 return false; 5838 } 5839 5840 bool AsmParser::parseDirectivePrint(SMLoc DirectiveLoc) { 5841 const AsmToken StrTok = getTok(); 5842 Lex(); 5843 if (StrTok.isNot(AsmToken::String) || StrTok.getString().front() != '"') 5844 return Error(DirectiveLoc, "expected double quoted string after .print"); 5845 if (parseEOL()) 5846 return true; 5847 llvm::outs() << StrTok.getStringContents() << '\n'; 5848 return false; 5849 } 5850 5851 bool AsmParser::parseDirectiveAddrsig() { 5852 if (parseEOL()) 5853 return true; 5854 getStreamer().emitAddrsig(); 5855 return false; 5856 } 5857 5858 bool AsmParser::parseDirectiveAddrsigSym() { 5859 StringRef Name; 5860 if (check(parseIdentifier(Name), "expected identifier") || parseEOL()) 5861 return true; 5862 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5863 getStreamer().emitAddrsigSym(Sym); 5864 return false; 5865 } 5866 5867 bool AsmParser::parseDirectivePseudoProbe() { 5868 int64_t Guid; 5869 int64_t Index; 5870 int64_t Type; 5871 int64_t Attr; 5872 int64_t Discriminator = 0; 5873 5874 if (parseIntToken(Guid, "unexpected token in '.pseudoprobe' directive")) 5875 return true; 5876 5877 if (parseIntToken(Index, "unexpected token in '.pseudoprobe' directive")) 5878 return true; 5879 5880 if (parseIntToken(Type, "unexpected token in '.pseudoprobe' directive")) 5881 return true; 5882 5883 if (parseIntToken(Attr, "unexpected token in '.pseudoprobe' directive")) 5884 return true; 5885 5886 if (hasDiscriminator(Attr)) { 5887 if (parseIntToken(Discriminator, 5888 "unexpected token in '.pseudoprobe' directive")) 5889 return true; 5890 } 5891 5892 // Parse inline stack like @ GUID:11:12 @ GUID:1:11 @ GUID:3:21 5893 MCPseudoProbeInlineStack InlineStack; 5894 5895 while (getLexer().is(AsmToken::At)) { 5896 // eat @ 5897 Lex(); 5898 5899 int64_t CallerGuid = 0; 5900 if (getLexer().is(AsmToken::Integer)) { 5901 if (parseIntToken(CallerGuid, 5902 "unexpected token in '.pseudoprobe' directive")) 5903 return true; 5904 } 5905 5906 // eat colon 5907 if (getLexer().is(AsmToken::Colon)) 5908 Lex(); 5909 5910 int64_t CallerProbeId = 0; 5911 if (getLexer().is(AsmToken::Integer)) { 5912 if (parseIntToken(CallerProbeId, 5913 "unexpected token in '.pseudoprobe' directive")) 5914 return true; 5915 } 5916 5917 InlineSite Site(CallerGuid, CallerProbeId); 5918 InlineStack.push_back(Site); 5919 } 5920 5921 // Parse function entry name 5922 StringRef FnName; 5923 if (parseIdentifier(FnName)) 5924 return Error(getLexer().getLoc(), "unexpected token in '.pseudoprobe' directive"); 5925 MCSymbol *FnSym = getContext().lookupSymbol(FnName); 5926 5927 if (parseEOL()) 5928 return true; 5929 5930 getStreamer().emitPseudoProbe(Guid, Index, Type, Attr, Discriminator, 5931 InlineStack, FnSym); 5932 return false; 5933 } 5934 5935 /// parseDirectiveLTODiscard 5936 /// ::= ".lto_discard" [ identifier ( , identifier )* ] 5937 /// The LTO library emits this directive to discard non-prevailing symbols. 5938 /// We ignore symbol assignments and attribute changes for the specified 5939 /// symbols. 5940 bool AsmParser::parseDirectiveLTODiscard() { 5941 auto ParseOp = [&]() -> bool { 5942 StringRef Name; 5943 SMLoc Loc = getTok().getLoc(); 5944 if (parseIdentifier(Name)) 5945 return Error(Loc, "expected identifier"); 5946 LTODiscardSymbols.insert(Name); 5947 return false; 5948 }; 5949 5950 LTODiscardSymbols.clear(); 5951 return parseMany(ParseOp); 5952 } 5953 5954 // We are comparing pointers, but the pointers are relative to a single string. 5955 // Thus, this should always be deterministic. 5956 static int rewritesSort(const AsmRewrite *AsmRewriteA, 5957 const AsmRewrite *AsmRewriteB) { 5958 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) 5959 return -1; 5960 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) 5961 return 1; 5962 5963 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output 5964 // rewrite to the same location. Make sure the SizeDirective rewrite is 5965 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This 5966 // ensures the sort algorithm is stable. 5967 if (AsmRewritePrecedence[AsmRewriteA->Kind] > 5968 AsmRewritePrecedence[AsmRewriteB->Kind]) 5969 return -1; 5970 5971 if (AsmRewritePrecedence[AsmRewriteA->Kind] < 5972 AsmRewritePrecedence[AsmRewriteB->Kind]) 5973 return 1; 5974 llvm_unreachable("Unstable rewrite sort."); 5975 } 5976 5977 bool AsmParser::parseMSInlineAsm( 5978 std::string &AsmString, unsigned &NumOutputs, unsigned &NumInputs, 5979 SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 5980 SmallVectorImpl<std::string> &Constraints, 5981 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, 5982 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { 5983 SmallVector<void *, 4> InputDecls; 5984 SmallVector<void *, 4> OutputDecls; 5985 SmallVector<bool, 4> InputDeclsAddressOf; 5986 SmallVector<bool, 4> OutputDeclsAddressOf; 5987 SmallVector<std::string, 4> InputConstraints; 5988 SmallVector<std::string, 4> OutputConstraints; 5989 SmallVector<unsigned, 4> ClobberRegs; 5990 5991 SmallVector<AsmRewrite, 4> AsmStrRewrites; 5992 5993 // Prime the lexer. 5994 Lex(); 5995 5996 // While we have input, parse each statement. 5997 unsigned InputIdx = 0; 5998 unsigned OutputIdx = 0; 5999 while (getLexer().isNot(AsmToken::Eof)) { 6000 // Parse curly braces marking block start/end 6001 if (parseCurlyBlockScope(AsmStrRewrites)) 6002 continue; 6003 6004 ParseStatementInfo Info(&AsmStrRewrites); 6005 bool StatementErr = parseStatement(Info, &SI); 6006 6007 if (StatementErr || Info.ParseError) { 6008 // Emit pending errors if any exist. 6009 printPendingErrors(); 6010 return true; 6011 } 6012 6013 // No pending error should exist here. 6014 assert(!hasPendingError() && "unexpected error from parseStatement"); 6015 6016 if (Info.Opcode == ~0U) 6017 continue; 6018 6019 const MCInstrDesc &Desc = MII->get(Info.Opcode); 6020 6021 // Build the list of clobbers, outputs and inputs. 6022 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { 6023 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; 6024 6025 // Register operand. 6026 if (Operand.isReg() && !Operand.needAddressOf() && 6027 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) { 6028 unsigned NumDefs = Desc.getNumDefs(); 6029 // Clobber. 6030 if (NumDefs && Operand.getMCOperandNum() < NumDefs) 6031 ClobberRegs.push_back(Operand.getReg()); 6032 continue; 6033 } 6034 6035 // Expr/Input or Output. 6036 StringRef SymName = Operand.getSymName(); 6037 if (SymName.empty()) 6038 continue; 6039 6040 void *OpDecl = Operand.getOpDecl(); 6041 if (!OpDecl) 6042 continue; 6043 6044 StringRef Constraint = Operand.getConstraint(); 6045 if (Operand.isImm()) { 6046 // Offset as immediate 6047 if (Operand.isOffsetOfLocal()) 6048 Constraint = "r"; 6049 else 6050 Constraint = "i"; 6051 } 6052 6053 bool isOutput = (i == 1) && Desc.mayStore(); 6054 bool Restricted = Operand.isMemUseUpRegs(); 6055 SMLoc Start = SMLoc::getFromPointer(SymName.data()); 6056 if (isOutput) { 6057 ++InputIdx; 6058 OutputDecls.push_back(OpDecl); 6059 OutputDeclsAddressOf.push_back(Operand.needAddressOf()); 6060 OutputConstraints.push_back(("=" + Constraint).str()); 6061 AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size(), 0, 6062 Restricted); 6063 } else { 6064 InputDecls.push_back(OpDecl); 6065 InputDeclsAddressOf.push_back(Operand.needAddressOf()); 6066 InputConstraints.push_back(Constraint.str()); 6067 if (Desc.operands()[i - 1].isBranchTarget()) 6068 AsmStrRewrites.emplace_back(AOK_CallInput, Start, SymName.size(), 0, 6069 Restricted); 6070 else 6071 AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size(), 0, 6072 Restricted); 6073 } 6074 } 6075 6076 // Consider implicit defs to be clobbers. Think of cpuid and push. 6077 llvm::append_range(ClobberRegs, Desc.implicit_defs()); 6078 } 6079 6080 // Set the number of Outputs and Inputs. 6081 NumOutputs = OutputDecls.size(); 6082 NumInputs = InputDecls.size(); 6083 6084 // Set the unique clobbers. 6085 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end()); 6086 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()), 6087 ClobberRegs.end()); 6088 Clobbers.assign(ClobberRegs.size(), std::string()); 6089 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { 6090 raw_string_ostream OS(Clobbers[I]); 6091 IP->printRegName(OS, ClobberRegs[I]); 6092 } 6093 6094 // Merge the various outputs and inputs. Output are expected first. 6095 if (NumOutputs || NumInputs) { 6096 unsigned NumExprs = NumOutputs + NumInputs; 6097 OpDecls.resize(NumExprs); 6098 Constraints.resize(NumExprs); 6099 for (unsigned i = 0; i < NumOutputs; ++i) { 6100 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]); 6101 Constraints[i] = OutputConstraints[i]; 6102 } 6103 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { 6104 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]); 6105 Constraints[j] = InputConstraints[i]; 6106 } 6107 } 6108 6109 // Build the IR assembly string. 6110 std::string AsmStringIR; 6111 raw_string_ostream OS(AsmStringIR); 6112 StringRef ASMString = 6113 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer(); 6114 const char *AsmStart = ASMString.begin(); 6115 const char *AsmEnd = ASMString.end(); 6116 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort); 6117 for (auto it = AsmStrRewrites.begin(); it != AsmStrRewrites.end(); ++it) { 6118 const AsmRewrite &AR = *it; 6119 // Check if this has already been covered by another rewrite... 6120 if (AR.Done) 6121 continue; 6122 AsmRewriteKind Kind = AR.Kind; 6123 6124 const char *Loc = AR.Loc.getPointer(); 6125 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!"); 6126 6127 // Emit everything up to the immediate/expression. 6128 if (unsigned Len = Loc - AsmStart) 6129 OS << StringRef(AsmStart, Len); 6130 6131 // Skip the original expression. 6132 if (Kind == AOK_Skip) { 6133 AsmStart = Loc + AR.Len; 6134 continue; 6135 } 6136 6137 unsigned AdditionalSkip = 0; 6138 // Rewrite expressions in $N notation. 6139 switch (Kind) { 6140 default: 6141 break; 6142 case AOK_IntelExpr: 6143 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression"); 6144 if (AR.IntelExp.NeedBracs) 6145 OS << "["; 6146 if (AR.IntelExp.hasBaseReg()) 6147 OS << AR.IntelExp.BaseReg; 6148 if (AR.IntelExp.hasIndexReg()) 6149 OS << (AR.IntelExp.hasBaseReg() ? " + " : "") 6150 << AR.IntelExp.IndexReg; 6151 if (AR.IntelExp.Scale > 1) 6152 OS << " * $$" << AR.IntelExp.Scale; 6153 if (AR.IntelExp.hasOffset()) { 6154 if (AR.IntelExp.hasRegs()) 6155 OS << " + "; 6156 // Fuse this rewrite with a rewrite of the offset name, if present. 6157 StringRef OffsetName = AR.IntelExp.OffsetName; 6158 SMLoc OffsetLoc = SMLoc::getFromPointer(AR.IntelExp.OffsetName.data()); 6159 size_t OffsetLen = OffsetName.size(); 6160 auto rewrite_it = std::find_if( 6161 it, AsmStrRewrites.end(), [&](const AsmRewrite &FusingAR) { 6162 return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen && 6163 (FusingAR.Kind == AOK_Input || 6164 FusingAR.Kind == AOK_CallInput); 6165 }); 6166 if (rewrite_it == AsmStrRewrites.end()) { 6167 OS << "offset " << OffsetName; 6168 } else if (rewrite_it->Kind == AOK_CallInput) { 6169 OS << "${" << InputIdx++ << ":P}"; 6170 rewrite_it->Done = true; 6171 } else { 6172 OS << '$' << InputIdx++; 6173 rewrite_it->Done = true; 6174 } 6175 } 6176 if (AR.IntelExp.Imm || AR.IntelExp.emitImm()) 6177 OS << (AR.IntelExp.emitImm() ? "$$" : " + $$") << AR.IntelExp.Imm; 6178 if (AR.IntelExp.NeedBracs) 6179 OS << "]"; 6180 break; 6181 case AOK_Label: 6182 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; 6183 break; 6184 case AOK_Input: 6185 if (AR.IntelExpRestricted) 6186 OS << "${" << InputIdx++ << ":P}"; 6187 else 6188 OS << '$' << InputIdx++; 6189 break; 6190 case AOK_CallInput: 6191 OS << "${" << InputIdx++ << ":P}"; 6192 break; 6193 case AOK_Output: 6194 if (AR.IntelExpRestricted) 6195 OS << "${" << OutputIdx++ << ":P}"; 6196 else 6197 OS << '$' << OutputIdx++; 6198 break; 6199 case AOK_SizeDirective: 6200 switch (AR.Val) { 6201 default: break; 6202 case 8: OS << "byte ptr "; break; 6203 case 16: OS << "word ptr "; break; 6204 case 32: OS << "dword ptr "; break; 6205 case 64: OS << "qword ptr "; break; 6206 case 80: OS << "xword ptr "; break; 6207 case 128: OS << "xmmword ptr "; break; 6208 case 256: OS << "ymmword ptr "; break; 6209 } 6210 break; 6211 case AOK_Emit: 6212 OS << ".byte"; 6213 break; 6214 case AOK_Align: { 6215 // MS alignment directives are measured in bytes. If the native assembler 6216 // measures alignment in bytes, we can pass it straight through. 6217 OS << ".align"; 6218 if (getContext().getAsmInfo()->getAlignmentIsInBytes()) 6219 break; 6220 6221 // Alignment is in log2 form, so print that instead and skip the original 6222 // immediate. 6223 unsigned Val = AR.Val; 6224 OS << ' ' << Val; 6225 assert(Val < 10 && "Expected alignment less then 2^10."); 6226 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; 6227 break; 6228 } 6229 case AOK_EVEN: 6230 OS << ".even"; 6231 break; 6232 case AOK_EndOfStatement: 6233 OS << "\n\t"; 6234 break; 6235 } 6236 6237 // Skip the original expression. 6238 AsmStart = Loc + AR.Len + AdditionalSkip; 6239 } 6240 6241 // Emit the remainder of the asm string. 6242 if (AsmStart != AsmEnd) 6243 OS << StringRef(AsmStart, AsmEnd - AsmStart); 6244 6245 AsmString = OS.str(); 6246 return false; 6247 } 6248 6249 bool HLASMAsmParser::parseAsHLASMLabel(ParseStatementInfo &Info, 6250 MCAsmParserSemaCallback *SI) { 6251 AsmToken LabelTok = getTok(); 6252 SMLoc LabelLoc = LabelTok.getLoc(); 6253 StringRef LabelVal; 6254 6255 if (parseIdentifier(LabelVal)) 6256 return Error(LabelLoc, "The HLASM Label has to be an Identifier"); 6257 6258 // We have validated whether the token is an Identifier. 6259 // Now we have to validate whether the token is a 6260 // valid HLASM Label. 6261 if (!getTargetParser().isLabel(LabelTok) || checkForValidSection()) 6262 return true; 6263 6264 // Lex leading spaces to get to the next operand. 6265 lexLeadingSpaces(); 6266 6267 // We shouldn't emit the label if there is nothing else after the label. 6268 // i.e asm("<token>\n") 6269 if (getTok().is(AsmToken::EndOfStatement)) 6270 return Error(LabelLoc, 6271 "Cannot have just a label for an HLASM inline asm statement"); 6272 6273 MCSymbol *Sym = getContext().getOrCreateSymbol( 6274 getContext().getAsmInfo()->shouldEmitLabelsInUpperCase() 6275 ? LabelVal.upper() 6276 : LabelVal); 6277 6278 getTargetParser().doBeforeLabelEmit(Sym, LabelLoc); 6279 6280 // Emit the label. 6281 Out.emitLabel(Sym, LabelLoc); 6282 6283 // If we are generating dwarf for assembly source files then gather the 6284 // info to make a dwarf label entry for this label if needed. 6285 if (enabledGenDwarfForAssembly()) 6286 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 6287 LabelLoc); 6288 6289 getTargetParser().onLabelParsed(Sym); 6290 6291 return false; 6292 } 6293 6294 bool HLASMAsmParser::parseAsMachineInstruction(ParseStatementInfo &Info, 6295 MCAsmParserSemaCallback *SI) { 6296 AsmToken OperationEntryTok = Lexer.getTok(); 6297 SMLoc OperationEntryLoc = OperationEntryTok.getLoc(); 6298 StringRef OperationEntryVal; 6299 6300 // Attempt to parse the first token as an Identifier 6301 if (parseIdentifier(OperationEntryVal)) 6302 return Error(OperationEntryLoc, "unexpected token at start of statement"); 6303 6304 // Once we've parsed the operation entry successfully, lex 6305 // any spaces to get to the OperandEntries. 6306 lexLeadingSpaces(); 6307 6308 return parseAndMatchAndEmitTargetInstruction( 6309 Info, OperationEntryVal, OperationEntryTok, OperationEntryLoc); 6310 } 6311 6312 bool HLASMAsmParser::parseStatement(ParseStatementInfo &Info, 6313 MCAsmParserSemaCallback *SI) { 6314 assert(!hasPendingError() && "parseStatement started with pending error"); 6315 6316 // Should the first token be interpreted as a HLASM Label. 6317 bool ShouldParseAsHLASMLabel = false; 6318 6319 // If a Name Entry exists, it should occur at the very 6320 // start of the string. In this case, we should parse the 6321 // first non-space token as a Label. 6322 // If the Name entry is missing (i.e. there's some other 6323 // token), then we attempt to parse the first non-space 6324 // token as a Machine Instruction. 6325 if (getTok().isNot(AsmToken::Space)) 6326 ShouldParseAsHLASMLabel = true; 6327 6328 // If we have an EndOfStatement (which includes the target's comment 6329 // string) we can appropriately lex it early on) 6330 if (Lexer.is(AsmToken::EndOfStatement)) { 6331 // if this is a line comment we can drop it safely 6332 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 6333 getTok().getString().front() == '\n') 6334 Out.addBlankLine(); 6335 Lex(); 6336 return false; 6337 } 6338 6339 // We have established how to parse the inline asm statement. 6340 // Now we can safely lex any leading spaces to get to the 6341 // first token. 6342 lexLeadingSpaces(); 6343 6344 // If we see a new line or carriage return as the first operand, 6345 // after lexing leading spaces, emit the new line and lex the 6346 // EndOfStatement token. 6347 if (Lexer.is(AsmToken::EndOfStatement)) { 6348 if (getTok().getString().front() == '\n' || 6349 getTok().getString().front() == '\r') { 6350 Out.addBlankLine(); 6351 Lex(); 6352 return false; 6353 } 6354 } 6355 6356 // Handle the label first if we have to before processing the rest 6357 // of the tokens as a machine instruction. 6358 if (ShouldParseAsHLASMLabel) { 6359 // If there were any errors while handling and emitting the label, 6360 // early return. 6361 if (parseAsHLASMLabel(Info, SI)) { 6362 // If we know we've failed in parsing, simply eat until end of the 6363 // statement. This ensures that we don't process any other statements. 6364 eatToEndOfStatement(); 6365 return true; 6366 } 6367 } 6368 6369 return parseAsMachineInstruction(Info, SI); 6370 } 6371 6372 namespace llvm { 6373 namespace MCParserUtils { 6374 6375 /// Returns whether the given symbol is used anywhere in the given expression, 6376 /// or subexpressions. 6377 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) { 6378 switch (Value->getKind()) { 6379 case MCExpr::Binary: { 6380 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value); 6381 return isSymbolUsedInExpression(Sym, BE->getLHS()) || 6382 isSymbolUsedInExpression(Sym, BE->getRHS()); 6383 } 6384 case MCExpr::Target: 6385 case MCExpr::Constant: 6386 return false; 6387 case MCExpr::SymbolRef: { 6388 const MCSymbol &S = 6389 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol(); 6390 if (S.isVariable() && !S.isWeakExternal()) 6391 return isSymbolUsedInExpression(Sym, S.getVariableValue()); 6392 return &S == Sym; 6393 } 6394 case MCExpr::Unary: 6395 return isSymbolUsedInExpression( 6396 Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr()); 6397 } 6398 6399 llvm_unreachable("Unknown expr kind!"); 6400 } 6401 6402 bool parseAssignmentExpression(StringRef Name, bool allow_redef, 6403 MCAsmParser &Parser, MCSymbol *&Sym, 6404 const MCExpr *&Value) { 6405 6406 // FIXME: Use better location, we should use proper tokens. 6407 SMLoc EqualLoc = Parser.getTok().getLoc(); 6408 if (Parser.parseExpression(Value)) 6409 return Parser.TokError("missing expression"); 6410 6411 // Note: we don't count b as used in "a = b". This is to allow 6412 // a = b 6413 // b = c 6414 6415 if (Parser.parseEOL()) 6416 return true; 6417 6418 // Validate that the LHS is allowed to be a variable (either it has not been 6419 // used as a symbol, or it is an absolute symbol). 6420 Sym = Parser.getContext().lookupSymbol(Name); 6421 if (Sym) { 6422 // Diagnose assignment to a label. 6423 // 6424 // FIXME: Diagnostics. Note the location of the definition as a label. 6425 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 6426 if (isSymbolUsedInExpression(Sym, Value)) 6427 return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'"); 6428 else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() && 6429 !Sym->isVariable()) 6430 ; // Allow redefinitions of undefined symbols only used in directives. 6431 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef) 6432 ; // Allow redefinitions of variables that haven't yet been used. 6433 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 6434 return Parser.Error(EqualLoc, "redefinition of '" + Name + "'"); 6435 else if (!Sym->isVariable()) 6436 return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'"); 6437 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 6438 return Parser.Error(EqualLoc, 6439 "invalid reassignment of non-absolute variable '" + 6440 Name + "'"); 6441 } else if (Name == ".") { 6442 Parser.getStreamer().emitValueToOffset(Value, 0, EqualLoc); 6443 return false; 6444 } else 6445 Sym = Parser.getContext().getOrCreateSymbol(Name); 6446 6447 Sym->setRedefinable(allow_redef); 6448 6449 return false; 6450 } 6451 6452 } // end namespace MCParserUtils 6453 } // end namespace llvm 6454 6455 /// Create an MCAsmParser instance. 6456 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C, 6457 MCStreamer &Out, const MCAsmInfo &MAI, 6458 unsigned CB) { 6459 if (C.getTargetTriple().isSystemZ() && C.getTargetTriple().isOSzOS()) 6460 return new HLASMAsmParser(SM, C, Out, MAI, CB); 6461 6462 return new AsmParser(SM, C, Out, MAI, CB); 6463 } 6464