1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This class implements the parser for assembly files. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "llvm/ADT/APFloat.h" 14 #include "llvm/ADT/APInt.h" 15 #include "llvm/ADT/ArrayRef.h" 16 #include "llvm/ADT/None.h" 17 #include "llvm/ADT/STLExtras.h" 18 #include "llvm/ADT/SmallString.h" 19 #include "llvm/ADT/SmallVector.h" 20 #include "llvm/ADT/StringExtras.h" 21 #include "llvm/ADT/StringMap.h" 22 #include "llvm/ADT/StringRef.h" 23 #include "llvm/ADT/Twine.h" 24 #include "llvm/BinaryFormat/Dwarf.h" 25 #include "llvm/DebugInfo/CodeView/SymbolRecord.h" 26 #include "llvm/MC/MCAsmInfo.h" 27 #include "llvm/MC/MCCodeView.h" 28 #include "llvm/MC/MCContext.h" 29 #include "llvm/MC/MCDirectives.h" 30 #include "llvm/MC/MCDwarf.h" 31 #include "llvm/MC/MCExpr.h" 32 #include "llvm/MC/MCInstPrinter.h" 33 #include "llvm/MC/MCInstrDesc.h" 34 #include "llvm/MC/MCInstrInfo.h" 35 #include "llvm/MC/MCObjectFileInfo.h" 36 #include "llvm/MC/MCParser/AsmCond.h" 37 #include "llvm/MC/MCParser/AsmLexer.h" 38 #include "llvm/MC/MCParser/MCAsmLexer.h" 39 #include "llvm/MC/MCParser/MCAsmParser.h" 40 #include "llvm/MC/MCParser/MCAsmParserExtension.h" 41 #include "llvm/MC/MCParser/MCAsmParserUtils.h" 42 #include "llvm/MC/MCParser/MCParsedAsmOperand.h" 43 #include "llvm/MC/MCParser/MCTargetAsmParser.h" 44 #include "llvm/MC/MCRegisterInfo.h" 45 #include "llvm/MC/MCSection.h" 46 #include "llvm/MC/MCStreamer.h" 47 #include "llvm/MC/MCSymbol.h" 48 #include "llvm/MC/MCTargetOptions.h" 49 #include "llvm/MC/MCValue.h" 50 #include "llvm/Support/Casting.h" 51 #include "llvm/Support/CommandLine.h" 52 #include "llvm/Support/ErrorHandling.h" 53 #include "llvm/Support/MD5.h" 54 #include "llvm/Support/MathExtras.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 #include "llvm/Support/SMLoc.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/raw_ostream.h" 59 #include <algorithm> 60 #include <cassert> 61 #include <cctype> 62 #include <climits> 63 #include <cstddef> 64 #include <cstdint> 65 #include <deque> 66 #include <memory> 67 #include <sstream> 68 #include <string> 69 #include <tuple> 70 #include <utility> 71 #include <vector> 72 73 using namespace llvm; 74 75 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() = default; 76 77 static cl::opt<unsigned> AsmMacroMaxNestingDepth( 78 "asm-macro-max-nesting-depth", cl::init(20), cl::Hidden, 79 cl::desc("The maximum nesting depth allowed for assembly macros.")); 80 81 namespace { 82 83 /// Helper types for tracking macro definitions. 84 typedef std::vector<AsmToken> MCAsmMacroArgument; 85 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; 86 87 /// Helper class for storing information about an active macro 88 /// instantiation. 89 struct MacroInstantiation { 90 /// The location of the instantiation. 91 SMLoc InstantiationLoc; 92 93 /// The buffer where parsing should resume upon instantiation completion. 94 unsigned ExitBuffer; 95 96 /// The location where parsing should resume upon instantiation completion. 97 SMLoc ExitLoc; 98 99 /// The depth of TheCondStack at the start of the instantiation. 100 size_t CondStackDepth; 101 }; 102 103 struct ParseStatementInfo { 104 /// The parsed operands from the last parsed statement. 105 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; 106 107 /// The opcode from the last parsed instruction. 108 unsigned Opcode = ~0U; 109 110 /// Was there an error parsing the inline assembly? 111 bool ParseError = false; 112 113 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr; 114 115 ParseStatementInfo() = delete; 116 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites) 117 : AsmRewrites(rewrites) {} 118 }; 119 120 /// The concrete assembly parser instance. 121 class AsmParser : public MCAsmParser { 122 private: 123 AsmLexer Lexer; 124 MCContext &Ctx; 125 MCStreamer &Out; 126 const MCAsmInfo &MAI; 127 SourceMgr &SrcMgr; 128 SourceMgr::DiagHandlerTy SavedDiagHandler; 129 void *SavedDiagContext; 130 std::unique_ptr<MCAsmParserExtension> PlatformParser; 131 132 /// This is the current buffer index we're lexing from as managed by the 133 /// SourceMgr object. 134 unsigned CurBuffer; 135 136 AsmCond TheCondState; 137 std::vector<AsmCond> TheCondStack; 138 139 /// maps directive names to handler methods in parser 140 /// extensions. Extensions register themselves in this map by calling 141 /// addDirectiveHandler. 142 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap; 143 144 /// Stack of active macro instantiations. 145 std::vector<MacroInstantiation*> ActiveMacros; 146 147 /// List of bodies of anonymous macros. 148 std::deque<MCAsmMacro> MacroLikeBodies; 149 150 /// Boolean tracking whether macro substitution is enabled. 151 unsigned MacrosEnabledFlag : 1; 152 153 /// Keeps track of how many .macro's have been instantiated. 154 unsigned NumOfMacroInstantiations; 155 156 /// The values from the last parsed cpp hash file line comment if any. 157 struct CppHashInfoTy { 158 StringRef Filename; 159 int64_t LineNumber; 160 SMLoc Loc; 161 unsigned Buf; 162 CppHashInfoTy() : Filename(), LineNumber(0), Loc(), Buf(0) {} 163 }; 164 CppHashInfoTy CppHashInfo; 165 166 /// The filename from the first cpp hash file line comment, if any. 167 StringRef FirstCppHashFilename; 168 169 /// List of forward directional labels for diagnosis at the end. 170 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; 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 ParsingInlineAsm = 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 public: 188 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 189 const MCAsmInfo &MAI, unsigned CB); 190 AsmParser(const AsmParser &) = delete; 191 AsmParser &operator=(const AsmParser &) = delete; 192 ~AsmParser() override; 193 194 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override; 195 196 void addDirectiveHandler(StringRef Directive, 197 ExtensionDirectiveHandler Handler) override { 198 ExtensionDirectiveMap[Directive] = Handler; 199 } 200 201 void addAliasForDirective(StringRef Directive, StringRef Alias) override { 202 DirectiveKindMap[Directive] = DirectiveKindMap[Alias]; 203 } 204 205 /// @name MCAsmParser Interface 206 /// { 207 208 SourceMgr &getSourceManager() override { return SrcMgr; } 209 MCAsmLexer &getLexer() override { return Lexer; } 210 MCContext &getContext() override { return Ctx; } 211 MCStreamer &getStreamer() override { return Out; } 212 213 CodeViewContext &getCVContext() { return Ctx.getCVContext(); } 214 215 unsigned getAssemblerDialect() override { 216 if (AssemblerDialect == ~0U) 217 return MAI.getAssemblerDialect(); 218 else 219 return AssemblerDialect; 220 } 221 void setAssemblerDialect(unsigned i) override { 222 AssemblerDialect = i; 223 } 224 225 void Note(SMLoc L, const Twine &Msg, SMRange Range = None) override; 226 bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) override; 227 bool printError(SMLoc L, const Twine &Msg, SMRange Range = None) override; 228 229 const AsmToken &Lex() override; 230 231 void setParsingInlineAsm(bool V) override { 232 ParsingInlineAsm = V; 233 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and 234 // hex integer literals. 235 Lexer.setLexMasmIntegers(V); 236 } 237 bool isParsingInlineAsm() override { return ParsingInlineAsm; } 238 239 bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString, 240 unsigned &NumOutputs, unsigned &NumInputs, 241 SmallVectorImpl<std::pair<void *,bool>> &OpDecls, 242 SmallVectorImpl<std::string> &Constraints, 243 SmallVectorImpl<std::string> &Clobbers, 244 const MCInstrInfo *MII, const MCInstPrinter *IP, 245 MCAsmParserSemaCallback &SI) override; 246 247 bool parseExpression(const MCExpr *&Res); 248 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 249 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override; 250 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 251 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 252 SMLoc &EndLoc) override; 253 bool parseAbsoluteExpression(int64_t &Res) override; 254 255 /// Parse a floating point expression using the float \p Semantics 256 /// and set \p Res to the value. 257 bool parseRealValue(const fltSemantics &Semantics, APInt &Res); 258 259 /// Parse an identifier or string (as a quoted identifier) 260 /// and set \p Res to the identifier contents. 261 bool parseIdentifier(StringRef &Res) override; 262 void eatToEndOfStatement() override; 263 264 bool checkForValidSection() override; 265 266 /// } 267 268 private: 269 bool parseStatement(ParseStatementInfo &Info, 270 MCAsmParserSemaCallback *SI); 271 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); 272 bool parseCppHashLineFilenameComment(SMLoc L); 273 274 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body, 275 ArrayRef<MCAsmMacroParameter> Parameters); 276 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 277 ArrayRef<MCAsmMacroParameter> Parameters, 278 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable, 279 SMLoc L); 280 281 /// Are macros enabled in the parser? 282 bool areMacrosEnabled() {return MacrosEnabledFlag;} 283 284 /// Control a flag in the parser that enables or disables macros. 285 void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;} 286 287 /// Are we inside a macro instantiation? 288 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();} 289 290 /// Handle entry to macro instantiation. 291 /// 292 /// \param M The macro. 293 /// \param NameLoc Instantiation location. 294 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc); 295 296 /// Handle exit from macro instantiation. 297 void handleMacroExit(); 298 299 /// Extract AsmTokens for a macro argument. 300 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg); 301 302 /// Parse all macro arguments for a given macro. 303 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A); 304 305 void printMacroInstantiations(); 306 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 307 SMRange Range = None) const { 308 ArrayRef<SMRange> Ranges(Range); 309 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 310 } 311 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 312 313 /// Should we emit DWARF describing this assembler source? (Returns false if 314 /// the source has .file directives, which means we don't want to generate 315 /// info describing the assembler source itself.) 316 bool enabledGenDwarfForAssembly(); 317 318 /// Enter the specified file. This returns true on failure. 319 bool enterIncludeFile(const std::string &Filename); 320 321 /// Process the specified file for the .incbin directive. 322 /// This returns true on failure. 323 bool processIncbinFile(const std::string &Filename, int64_t Skip = 0, 324 const MCExpr *Count = nullptr, SMLoc Loc = SMLoc()); 325 326 /// Reset the current lexer position to that given by \p Loc. The 327 /// current token is not set; clients should ensure Lex() is called 328 /// subsequently. 329 /// 330 /// \param InBuffer If not 0, should be the known buffer id that contains the 331 /// location. 332 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0); 333 334 /// Parse up to the end of statement and a return the contents from the 335 /// current token until the end of the statement; the current token on exit 336 /// will be either the EndOfStatement or EOF. 337 StringRef parseStringToEndOfStatement() override; 338 339 /// Parse until the end of a statement or a comma is encountered, 340 /// return the contents from the current token up to the end or comma. 341 StringRef parseStringToComma(); 342 343 bool parseAssignment(StringRef Name, bool allow_redef, 344 bool NoDeadStrip = false); 345 346 unsigned getBinOpPrecedence(AsmToken::TokenKind K, 347 MCBinaryExpr::Opcode &Kind); 348 349 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 350 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 351 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 352 353 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 354 355 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName); 356 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName); 357 358 // Generic (target and platform independent) directive parsing. 359 enum DirectiveKind { 360 DK_NO_DIRECTIVE, // Placeholder 361 DK_SET, 362 DK_EQU, 363 DK_EQUIV, 364 DK_ASCII, 365 DK_ASCIZ, 366 DK_STRING, 367 DK_BYTE, 368 DK_SHORT, 369 DK_RELOC, 370 DK_VALUE, 371 DK_2BYTE, 372 DK_LONG, 373 DK_INT, 374 DK_4BYTE, 375 DK_QUAD, 376 DK_8BYTE, 377 DK_OCTA, 378 DK_DC, 379 DK_DC_A, 380 DK_DC_B, 381 DK_DC_D, 382 DK_DC_L, 383 DK_DC_S, 384 DK_DC_W, 385 DK_DC_X, 386 DK_DCB, 387 DK_DCB_B, 388 DK_DCB_D, 389 DK_DCB_L, 390 DK_DCB_S, 391 DK_DCB_W, 392 DK_DCB_X, 393 DK_DS, 394 DK_DS_B, 395 DK_DS_D, 396 DK_DS_L, 397 DK_DS_P, 398 DK_DS_S, 399 DK_DS_W, 400 DK_DS_X, 401 DK_SINGLE, 402 DK_FLOAT, 403 DK_DOUBLE, 404 DK_ALIGN, 405 DK_ALIGN32, 406 DK_BALIGN, 407 DK_BALIGNW, 408 DK_BALIGNL, 409 DK_P2ALIGN, 410 DK_P2ALIGNW, 411 DK_P2ALIGNL, 412 DK_ORG, 413 DK_FILL, 414 DK_ENDR, 415 DK_BUNDLE_ALIGN_MODE, 416 DK_BUNDLE_LOCK, 417 DK_BUNDLE_UNLOCK, 418 DK_ZERO, 419 DK_EXTERN, 420 DK_GLOBL, 421 DK_GLOBAL, 422 DK_LAZY_REFERENCE, 423 DK_NO_DEAD_STRIP, 424 DK_SYMBOL_RESOLVER, 425 DK_PRIVATE_EXTERN, 426 DK_REFERENCE, 427 DK_WEAK_DEFINITION, 428 DK_WEAK_REFERENCE, 429 DK_WEAK_DEF_CAN_BE_HIDDEN, 430 DK_COLD, 431 DK_COMM, 432 DK_COMMON, 433 DK_LCOMM, 434 DK_ABORT, 435 DK_INCLUDE, 436 DK_INCBIN, 437 DK_CODE16, 438 DK_CODE16GCC, 439 DK_REPT, 440 DK_IRP, 441 DK_IRPC, 442 DK_IF, 443 DK_IFEQ, 444 DK_IFGE, 445 DK_IFGT, 446 DK_IFLE, 447 DK_IFLT, 448 DK_IFNE, 449 DK_IFB, 450 DK_IFNB, 451 DK_IFC, 452 DK_IFEQS, 453 DK_IFNC, 454 DK_IFNES, 455 DK_IFDEF, 456 DK_IFNDEF, 457 DK_IFNOTDEF, 458 DK_ELSEIF, 459 DK_ELSE, 460 DK_ENDIF, 461 DK_SPACE, 462 DK_SKIP, 463 DK_FILE, 464 DK_LINE, 465 DK_LOC, 466 DK_STABS, 467 DK_CV_FILE, 468 DK_CV_FUNC_ID, 469 DK_CV_INLINE_SITE_ID, 470 DK_CV_LOC, 471 DK_CV_LINETABLE, 472 DK_CV_INLINE_LINETABLE, 473 DK_CV_DEF_RANGE, 474 DK_CV_STRINGTABLE, 475 DK_CV_STRING, 476 DK_CV_FILECHECKSUMS, 477 DK_CV_FILECHECKSUM_OFFSET, 478 DK_CV_FPO_DATA, 479 DK_CFI_SECTIONS, 480 DK_CFI_STARTPROC, 481 DK_CFI_ENDPROC, 482 DK_CFI_DEF_CFA, 483 DK_CFI_DEF_CFA_OFFSET, 484 DK_CFI_ADJUST_CFA_OFFSET, 485 DK_CFI_DEF_CFA_REGISTER, 486 DK_CFI_OFFSET, 487 DK_CFI_REL_OFFSET, 488 DK_CFI_PERSONALITY, 489 DK_CFI_LSDA, 490 DK_CFI_REMEMBER_STATE, 491 DK_CFI_RESTORE_STATE, 492 DK_CFI_SAME_VALUE, 493 DK_CFI_RESTORE, 494 DK_CFI_ESCAPE, 495 DK_CFI_RETURN_COLUMN, 496 DK_CFI_SIGNAL_FRAME, 497 DK_CFI_UNDEFINED, 498 DK_CFI_REGISTER, 499 DK_CFI_WINDOW_SAVE, 500 DK_CFI_B_KEY_FRAME, 501 DK_MACROS_ON, 502 DK_MACROS_OFF, 503 DK_ALTMACRO, 504 DK_NOALTMACRO, 505 DK_MACRO, 506 DK_EXITM, 507 DK_ENDM, 508 DK_ENDMACRO, 509 DK_PURGEM, 510 DK_SLEB128, 511 DK_ULEB128, 512 DK_ERR, 513 DK_ERROR, 514 DK_WARNING, 515 DK_PRINT, 516 DK_ADDRSIG, 517 DK_ADDRSIG_SYM, 518 DK_END 519 }; 520 521 /// Maps directive name --> DirectiveKind enum, for 522 /// directives parsed by this class. 523 StringMap<DirectiveKind> DirectiveKindMap; 524 525 // Codeview def_range type parsing. 526 enum CVDefRangeType { 527 CVDR_DEFRANGE = 0, // Placeholder 528 CVDR_DEFRANGE_REGISTER, 529 CVDR_DEFRANGE_FRAMEPOINTER_REL, 530 CVDR_DEFRANGE_SUBFIELD_REGISTER, 531 CVDR_DEFRANGE_REGISTER_REL 532 }; 533 534 /// Maps Codeview def_range types --> CVDefRangeType enum, for 535 /// Codeview def_range types parsed by this class. 536 StringMap<CVDefRangeType> CVDefRangeTypeMap; 537 538 // ".ascii", ".asciz", ".string" 539 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 540 bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc" 541 bool parseDirectiveValue(StringRef IDVal, 542 unsigned Size); // ".byte", ".long", ... 543 bool parseDirectiveOctaValue(StringRef IDVal); // ".octa", ... 544 bool parseDirectiveRealValue(StringRef IDVal, 545 const fltSemantics &); // ".single", ... 546 bool parseDirectiveFill(); // ".fill" 547 bool parseDirectiveZero(); // ".zero" 548 // ".set", ".equ", ".equiv" 549 bool parseDirectiveSet(StringRef IDVal, bool allow_redef); 550 bool parseDirectiveOrg(); // ".org" 551 // ".align{,32}", ".p2align{,w,l}" 552 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize); 553 554 // ".file", ".line", ".loc", ".stabs" 555 bool parseDirectiveFile(SMLoc DirectiveLoc); 556 bool parseDirectiveLine(); 557 bool parseDirectiveLoc(); 558 bool parseDirectiveStabs(); 559 560 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable", 561 // ".cv_inline_linetable", ".cv_def_range", ".cv_string" 562 bool parseDirectiveCVFile(); 563 bool parseDirectiveCVFuncId(); 564 bool parseDirectiveCVInlineSiteId(); 565 bool parseDirectiveCVLoc(); 566 bool parseDirectiveCVLinetable(); 567 bool parseDirectiveCVInlineLinetable(); 568 bool parseDirectiveCVDefRange(); 569 bool parseDirectiveCVString(); 570 bool parseDirectiveCVStringTable(); 571 bool parseDirectiveCVFileChecksums(); 572 bool parseDirectiveCVFileChecksumOffset(); 573 bool parseDirectiveCVFPOData(); 574 575 // .cfi directives 576 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc); 577 bool parseDirectiveCFIWindowSave(); 578 bool parseDirectiveCFISections(); 579 bool parseDirectiveCFIStartProc(); 580 bool parseDirectiveCFIEndProc(); 581 bool parseDirectiveCFIDefCfaOffset(); 582 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc); 583 bool parseDirectiveCFIAdjustCfaOffset(); 584 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc); 585 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc); 586 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc); 587 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality); 588 bool parseDirectiveCFIRememberState(); 589 bool parseDirectiveCFIRestoreState(); 590 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc); 591 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc); 592 bool parseDirectiveCFIEscape(); 593 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc); 594 bool parseDirectiveCFISignalFrame(); 595 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc); 596 597 // macro directives 598 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc); 599 bool parseDirectiveExitMacro(StringRef Directive); 600 bool parseDirectiveEndMacro(StringRef Directive); 601 bool parseDirectiveMacro(SMLoc DirectiveLoc); 602 bool parseDirectiveMacrosOnOff(StringRef Directive); 603 // alternate macro mode directives 604 bool parseDirectiveAltmacro(StringRef Directive); 605 // ".bundle_align_mode" 606 bool parseDirectiveBundleAlignMode(); 607 // ".bundle_lock" 608 bool parseDirectiveBundleLock(); 609 // ".bundle_unlock" 610 bool parseDirectiveBundleUnlock(); 611 612 // ".space", ".skip" 613 bool parseDirectiveSpace(StringRef IDVal); 614 615 // ".dcb" 616 bool parseDirectiveDCB(StringRef IDVal, unsigned Size); 617 bool parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &); 618 // ".ds" 619 bool parseDirectiveDS(StringRef IDVal, unsigned Size); 620 621 // .sleb128 (Signed=true) and .uleb128 (Signed=false) 622 bool parseDirectiveLEB128(bool Signed); 623 624 /// Parse a directive like ".globl" which 625 /// accepts a single symbol (which should be a label or an external). 626 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr); 627 628 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 629 630 bool parseDirectiveAbort(); // ".abort" 631 bool parseDirectiveInclude(); // ".include" 632 bool parseDirectiveIncbin(); // ".incbin" 633 634 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne" 635 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 636 // ".ifb" or ".ifnb", depending on ExpectBlank. 637 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 638 // ".ifc" or ".ifnc", depending on ExpectEqual. 639 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual); 640 // ".ifeqs" or ".ifnes", depending on ExpectEqual. 641 bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual); 642 // ".ifdef" or ".ifndef", depending on expect_defined 643 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 644 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 645 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else" 646 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 647 bool parseEscapedString(std::string &Data) override; 648 649 const MCExpr *applyModifierToExpr(const MCExpr *E, 650 MCSymbolRefExpr::VariantKind Variant); 651 652 // Macro-like directives 653 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc); 654 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 655 raw_svector_ostream &OS); 656 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive); 657 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 658 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc" 659 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 660 661 // "_emit" or "__emit" 662 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info, 663 size_t Len); 664 665 // "align" 666 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info); 667 668 // "end" 669 bool parseDirectiveEnd(SMLoc DirectiveLoc); 670 671 // ".err" or ".error" 672 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage); 673 674 // ".warning" 675 bool parseDirectiveWarning(SMLoc DirectiveLoc); 676 677 // .print <double-quotes-string> 678 bool parseDirectivePrint(SMLoc DirectiveLoc); 679 680 // Directives to support address-significance tables. 681 bool parseDirectiveAddrsig(); 682 bool parseDirectiveAddrsigSym(); 683 684 void initializeDirectiveKindMap(); 685 void initializeCVDefRangeTypeMap(); 686 }; 687 688 } // end anonymous namespace 689 690 namespace llvm { 691 692 extern MCAsmParserExtension *createDarwinAsmParser(); 693 extern MCAsmParserExtension *createELFAsmParser(); 694 extern MCAsmParserExtension *createCOFFAsmParser(); 695 extern MCAsmParserExtension *createWasmAsmParser(); 696 697 } // end namespace llvm 698 699 enum { DEFAULT_ADDRSPACE = 0 }; 700 701 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 702 const MCAsmInfo &MAI, unsigned CB = 0) 703 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM), 704 CurBuffer(CB ? CB : SM.getMainFileID()), MacrosEnabledFlag(true) { 705 HadError = false; 706 // Save the old handler. 707 SavedDiagHandler = SrcMgr.getDiagHandler(); 708 SavedDiagContext = SrcMgr.getDiagContext(); 709 // Set our own handler which calls the saved handler. 710 SrcMgr.setDiagHandler(DiagHandler, this); 711 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 712 713 // Initialize the platform / file format parser. 714 switch (Ctx.getObjectFileInfo()->getObjectFileType()) { 715 case MCObjectFileInfo::IsCOFF: 716 PlatformParser.reset(createCOFFAsmParser()); 717 break; 718 case MCObjectFileInfo::IsMachO: 719 PlatformParser.reset(createDarwinAsmParser()); 720 IsDarwin = true; 721 break; 722 case MCObjectFileInfo::IsELF: 723 PlatformParser.reset(createELFAsmParser()); 724 break; 725 case MCObjectFileInfo::IsWasm: 726 PlatformParser.reset(createWasmAsmParser()); 727 break; 728 case MCObjectFileInfo::IsXCOFF: 729 report_fatal_error( 730 "Need to implement createXCOFFAsmParser for XCOFF format."); 731 break; 732 } 733 734 PlatformParser->Initialize(*this); 735 initializeDirectiveKindMap(); 736 initializeCVDefRangeTypeMap(); 737 738 NumOfMacroInstantiations = 0; 739 } 740 741 AsmParser::~AsmParser() { 742 assert((HadError || ActiveMacros.empty()) && 743 "Unexpected active macro instantiation!"); 744 745 // Restore the saved diagnostics handler and context for use during 746 // finalization. 747 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext); 748 } 749 750 void AsmParser::printMacroInstantiations() { 751 // Print the active macro instantiation stack. 752 for (std::vector<MacroInstantiation *>::const_reverse_iterator 753 it = ActiveMacros.rbegin(), 754 ie = ActiveMacros.rend(); 755 it != ie; ++it) 756 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 757 "while in macro instantiation"); 758 } 759 760 void AsmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { 761 printPendingErrors(); 762 printMessage(L, SourceMgr::DK_Note, Msg, Range); 763 printMacroInstantiations(); 764 } 765 766 bool AsmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) { 767 if(getTargetParser().getTargetOptions().MCNoWarn) 768 return false; 769 if (getTargetParser().getTargetOptions().MCFatalWarnings) 770 return Error(L, Msg, Range); 771 printMessage(L, SourceMgr::DK_Warning, Msg, Range); 772 printMacroInstantiations(); 773 return false; 774 } 775 776 bool AsmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { 777 HadError = true; 778 printMessage(L, SourceMgr::DK_Error, Msg, Range); 779 printMacroInstantiations(); 780 return true; 781 } 782 783 bool AsmParser::enterIncludeFile(const std::string &Filename) { 784 std::string IncludedFile; 785 unsigned NewBuf = 786 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 787 if (!NewBuf) 788 return true; 789 790 CurBuffer = NewBuf; 791 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 792 return false; 793 } 794 795 /// Process the specified .incbin file by searching for it in the include paths 796 /// then just emitting the byte contents of the file to the streamer. This 797 /// returns true on failure. 798 bool AsmParser::processIncbinFile(const std::string &Filename, int64_t Skip, 799 const MCExpr *Count, SMLoc Loc) { 800 std::string IncludedFile; 801 unsigned NewBuf = 802 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 803 if (!NewBuf) 804 return true; 805 806 // Pick up the bytes from the file and emit them. 807 StringRef Bytes = SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(); 808 Bytes = Bytes.drop_front(Skip); 809 if (Count) { 810 int64_t Res; 811 if (!Count->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 812 return Error(Loc, "expected absolute expression"); 813 if (Res < 0) 814 return Warning(Loc, "negative count has no effect"); 815 Bytes = Bytes.take_front(Res); 816 } 817 getStreamer().EmitBytes(Bytes); 818 return false; 819 } 820 821 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) { 822 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); 823 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), 824 Loc.getPointer()); 825 } 826 827 const AsmToken &AsmParser::Lex() { 828 if (Lexer.getTok().is(AsmToken::Error)) 829 Error(Lexer.getErrLoc(), Lexer.getErr()); 830 831 // if it's a end of statement with a comment in it 832 if (getTok().is(AsmToken::EndOfStatement)) { 833 // if this is a line comment output it. 834 if (!getTok().getString().empty() && getTok().getString().front() != '\n' && 835 getTok().getString().front() != '\r' && MAI.preserveAsmComments()) 836 Out.addExplicitComment(Twine(getTok().getString())); 837 } 838 839 const AsmToken *tok = &Lexer.Lex(); 840 841 // Parse comments here to be deferred until end of next statement. 842 while (tok->is(AsmToken::Comment)) { 843 if (MAI.preserveAsmComments()) 844 Out.addExplicitComment(Twine(tok->getString())); 845 tok = &Lexer.Lex(); 846 } 847 848 if (tok->is(AsmToken::Eof)) { 849 // If this is the end of an included file, pop the parent file off the 850 // include stack. 851 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 852 if (ParentIncludeLoc != SMLoc()) { 853 jumpToLoc(ParentIncludeLoc); 854 return Lex(); 855 } 856 } 857 858 return *tok; 859 } 860 861 bool AsmParser::enabledGenDwarfForAssembly() { 862 // Check whether the user specified -g. 863 if (!getContext().getGenDwarfForAssembly()) 864 return false; 865 // If we haven't encountered any .file directives (which would imply that 866 // the assembler source was produced with debug info already) then emit one 867 // describing the assembler source file itself. 868 if (getContext().getGenDwarfFileNumber() == 0) { 869 // Use the first #line directive for this, if any. It's preprocessed, so 870 // there is no checksum, and of course no source directive. 871 if (!FirstCppHashFilename.empty()) 872 getContext().setMCLineTableRootFile(/*CUID=*/0, 873 getContext().getCompilationDir(), 874 FirstCppHashFilename, 875 /*Cksum=*/None, /*Source=*/None); 876 const MCDwarfFile &RootFile = 877 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); 878 getContext().setGenDwarfFileNumber(getStreamer().EmitDwarfFileDirective( 879 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name, 880 RootFile.Checksum, RootFile.Source)); 881 } 882 return true; 883 } 884 885 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 886 // Create the initial section, if requested. 887 if (!NoInitialTextSection) 888 Out.InitSections(false); 889 890 // Prime the lexer. 891 Lex(); 892 893 HadError = false; 894 AsmCond StartingCondState = TheCondState; 895 SmallVector<AsmRewrite, 4> AsmStrRewrites; 896 897 // If we are generating dwarf for assembly source files save the initial text 898 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't 899 // emitting any actual debug info yet and haven't had a chance to parse any 900 // embedded .file directives.) 901 if (getContext().getGenDwarfForAssembly()) { 902 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 903 if (!Sec->getBeginSymbol()) { 904 MCSymbol *SectionStartSym = getContext().createTempSymbol(); 905 getStreamer().EmitLabel(SectionStartSym); 906 Sec->setBeginSymbol(SectionStartSym); 907 } 908 bool InsertResult = getContext().addGenDwarfSection(Sec); 909 assert(InsertResult && ".text section should not have debug info yet"); 910 (void)InsertResult; 911 } 912 913 StringRef Filename = getContext().getMainFileName(); 914 if (!Filename.empty() && (Filename.compare(StringRef("-")) != 0)) 915 Out.EmitFileDirective(Filename); 916 917 // While we have input, parse each statement. 918 while (Lexer.isNot(AsmToken::Eof)) { 919 ParseStatementInfo Info(&AsmStrRewrites); 920 bool Parsed = parseStatement(Info, nullptr); 921 922 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error 923 // for printing ErrMsg via Lex() only if no (presumably better) parser error 924 // exists. 925 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) { 926 Lex(); 927 } 928 929 // parseStatement returned true so may need to emit an error. 930 printPendingErrors(); 931 932 // Skipping to the next line if needed. 933 if (Parsed && !getLexer().isAtStartOfStatement()) 934 eatToEndOfStatement(); 935 } 936 937 getTargetParser().onEndOfFile(); 938 printPendingErrors(); 939 940 // All errors should have been emitted. 941 assert(!hasPendingError() && "unexpected error from parseStatement"); 942 943 getTargetParser().flushPendingInstructions(getStreamer()); 944 945 if (TheCondState.TheCond != StartingCondState.TheCond || 946 TheCondState.Ignore != StartingCondState.Ignore) 947 printError(getTok().getLoc(), "unmatched .ifs or .elses"); 948 // Check to see there are no empty DwarfFile slots. 949 const auto &LineTables = getContext().getMCDwarfLineTables(); 950 if (!LineTables.empty()) { 951 unsigned Index = 0; 952 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) { 953 if (File.Name.empty() && Index != 0) 954 printError(getTok().getLoc(), "unassigned file number: " + 955 Twine(Index) + 956 " for .file directives"); 957 ++Index; 958 } 959 } 960 961 // Check to see that all assembler local symbols were actually defined. 962 // Targets that don't do subsections via symbols may not want this, though, 963 // so conservatively exclude them. Only do this if we're finalizing, though, 964 // as otherwise we won't necessarilly have seen everything yet. 965 if (!NoFinalize) { 966 if (MAI.hasSubsectionsViaSymbols()) { 967 for (const auto &TableEntry : getContext().getSymbols()) { 968 MCSymbol *Sym = TableEntry.getValue(); 969 // Variable symbols may not be marked as defined, so check those 970 // explicitly. If we know it's a variable, we have a definition for 971 // the purposes of this check. 972 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 973 // FIXME: We would really like to refer back to where the symbol was 974 // first referenced for a source location. We need to add something 975 // to track that. Currently, we just point to the end of the file. 976 printError(getTok().getLoc(), "assembler local symbol '" + 977 Sym->getName() + "' not defined"); 978 } 979 } 980 981 // Temporary symbols like the ones for directional jumps don't go in the 982 // symbol table. They also need to be diagnosed in all (final) cases. 983 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { 984 if (std::get<2>(LocSym)->isUndefined()) { 985 // Reset the state of any "# line file" directives we've seen to the 986 // context as it was at the diagnostic site. 987 CppHashInfo = std::get<1>(LocSym); 988 printError(std::get<0>(LocSym), "directional label undefined"); 989 } 990 } 991 } 992 993 // Finalize the output stream if there are no errors and if the client wants 994 // us to. 995 if (!HadError && !NoFinalize) 996 Out.Finish(); 997 998 return HadError || getContext().hadError(); 999 } 1000 1001 bool AsmParser::checkForValidSection() { 1002 if (!ParsingInlineAsm && !getStreamer().getCurrentSectionOnly()) { 1003 Out.InitSections(false); 1004 return Error(getTok().getLoc(), 1005 "expected section directive before assembly directive"); 1006 } 1007 return false; 1008 } 1009 1010 /// Throw away the rest of the line for testing purposes. 1011 void AsmParser::eatToEndOfStatement() { 1012 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1013 Lexer.Lex(); 1014 1015 // Eat EOL. 1016 if (Lexer.is(AsmToken::EndOfStatement)) 1017 Lexer.Lex(); 1018 } 1019 1020 StringRef AsmParser::parseStringToEndOfStatement() { 1021 const char *Start = getTok().getLoc().getPointer(); 1022 1023 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1024 Lexer.Lex(); 1025 1026 const char *End = getTok().getLoc().getPointer(); 1027 return StringRef(Start, End - Start); 1028 } 1029 1030 StringRef AsmParser::parseStringToComma() { 1031 const char *Start = getTok().getLoc().getPointer(); 1032 1033 while (Lexer.isNot(AsmToken::EndOfStatement) && 1034 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof)) 1035 Lexer.Lex(); 1036 1037 const char *End = getTok().getLoc().getPointer(); 1038 return StringRef(Start, End - Start); 1039 } 1040 1041 /// Parse a paren expression and return it. 1042 /// NOTE: This assumes the leading '(' has already been consumed. 1043 /// 1044 /// parenexpr ::= expr) 1045 /// 1046 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1047 if (parseExpression(Res)) 1048 return true; 1049 if (Lexer.isNot(AsmToken::RParen)) 1050 return TokError("expected ')' in parentheses expression"); 1051 EndLoc = Lexer.getTok().getEndLoc(); 1052 Lex(); 1053 return false; 1054 } 1055 1056 /// Parse a bracket expression and return it. 1057 /// NOTE: This assumes the leading '[' has already been consumed. 1058 /// 1059 /// bracketexpr ::= expr] 1060 /// 1061 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1062 if (parseExpression(Res)) 1063 return true; 1064 EndLoc = getTok().getEndLoc(); 1065 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression")) 1066 return true; 1067 return false; 1068 } 1069 1070 /// Parse a primary expression and return it. 1071 /// primaryexpr ::= (parenexpr 1072 /// primaryexpr ::= symbol 1073 /// primaryexpr ::= number 1074 /// primaryexpr ::= '.' 1075 /// primaryexpr ::= ~,+,- primaryexpr 1076 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1077 SMLoc FirstTokenLoc = getLexer().getLoc(); 1078 AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); 1079 switch (FirstTokenKind) { 1080 default: 1081 return TokError("unknown token in expression"); 1082 // If we have an error assume that we've already handled it. 1083 case AsmToken::Error: 1084 return true; 1085 case AsmToken::Exclaim: 1086 Lex(); // Eat the operator. 1087 if (parsePrimaryExpr(Res, EndLoc)) 1088 return true; 1089 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc); 1090 return false; 1091 case AsmToken::Dollar: 1092 case AsmToken::At: 1093 case AsmToken::String: 1094 case AsmToken::Identifier: { 1095 StringRef Identifier; 1096 if (parseIdentifier(Identifier)) { 1097 // We may have failed but $ may be a valid token. 1098 if (getTok().is(AsmToken::Dollar)) { 1099 if (Lexer.getMAI().getDollarIsPC()) { 1100 Lex(); 1101 // This is a '$' reference, which references the current PC. Emit a 1102 // temporary label to the streamer and refer to it. 1103 MCSymbol *Sym = Ctx.createTempSymbol(); 1104 Out.EmitLabel(Sym); 1105 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, 1106 getContext()); 1107 EndLoc = FirstTokenLoc; 1108 return false; 1109 } 1110 return Error(FirstTokenLoc, "invalid token in expression"); 1111 } 1112 } 1113 // Parse symbol variant 1114 std::pair<StringRef, StringRef> Split; 1115 if (!MAI.useParensForSymbolVariant()) { 1116 if (FirstTokenKind == AsmToken::String) { 1117 if (Lexer.is(AsmToken::At)) { 1118 Lex(); // eat @ 1119 SMLoc AtLoc = getLexer().getLoc(); 1120 StringRef VName; 1121 if (parseIdentifier(VName)) 1122 return Error(AtLoc, "expected symbol variant after '@'"); 1123 1124 Split = std::make_pair(Identifier, VName); 1125 } 1126 } else { 1127 Split = Identifier.split('@'); 1128 } 1129 } else if (Lexer.is(AsmToken::LParen)) { 1130 Lex(); // eat '('. 1131 StringRef VName; 1132 parseIdentifier(VName); 1133 // eat ')'. 1134 if (parseToken(AsmToken::RParen, 1135 "unexpected token in variant, expected ')'")) 1136 return true; 1137 Split = std::make_pair(Identifier, VName); 1138 } 1139 1140 EndLoc = SMLoc::getFromPointer(Identifier.end()); 1141 1142 // This is a symbol reference. 1143 StringRef SymbolName = Identifier; 1144 if (SymbolName.empty()) 1145 return Error(getLexer().getLoc(), "expected a symbol reference"); 1146 1147 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1148 1149 // Lookup the symbol variant if used. 1150 if (!Split.second.empty()) { 1151 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1152 if (Variant != MCSymbolRefExpr::VK_Invalid) { 1153 SymbolName = Split.first; 1154 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { 1155 Variant = MCSymbolRefExpr::VK_None; 1156 } else { 1157 return Error(SMLoc::getFromPointer(Split.second.begin()), 1158 "invalid variant '" + Split.second + "'"); 1159 } 1160 } 1161 1162 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName); 1163 if (!Sym) 1164 Sym = getContext().getOrCreateSymbol(SymbolName); 1165 1166 // If this is an absolute variable reference, substitute it now to preserve 1167 // semantics in the face of reassignment. 1168 if (Sym->isVariable()) { 1169 auto V = Sym->getVariableValue(/*SetUsed*/ false); 1170 bool DoInline = isa<MCConstantExpr>(V) && !Variant; 1171 if (auto TV = dyn_cast<MCTargetExpr>(V)) 1172 DoInline = TV->inlineAssignedExpr(); 1173 if (DoInline) { 1174 if (Variant) 1175 return Error(EndLoc, "unexpected modifier on variable reference"); 1176 Res = Sym->getVariableValue(/*SetUsed*/ false); 1177 return false; 1178 } 1179 } 1180 1181 // Otherwise create a symbol ref. 1182 Res = MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc); 1183 return false; 1184 } 1185 case AsmToken::BigNum: 1186 return TokError("literal value out of range for directive"); 1187 case AsmToken::Integer: { 1188 SMLoc Loc = getTok().getLoc(); 1189 int64_t IntVal = getTok().getIntVal(); 1190 Res = MCConstantExpr::create(IntVal, getContext()); 1191 EndLoc = Lexer.getTok().getEndLoc(); 1192 Lex(); // Eat token. 1193 // Look for 'b' or 'f' following an Integer as a directional label 1194 if (Lexer.getKind() == AsmToken::Identifier) { 1195 StringRef IDVal = getTok().getString(); 1196 // Lookup the symbol variant if used. 1197 std::pair<StringRef, StringRef> Split = IDVal.split('@'); 1198 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1199 if (Split.first.size() != IDVal.size()) { 1200 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1201 if (Variant == MCSymbolRefExpr::VK_Invalid) 1202 return TokError("invalid variant '" + Split.second + "'"); 1203 IDVal = Split.first; 1204 } 1205 if (IDVal == "f" || IDVal == "b") { 1206 MCSymbol *Sym = 1207 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b"); 1208 Res = MCSymbolRefExpr::create(Sym, Variant, getContext()); 1209 if (IDVal == "b" && Sym->isUndefined()) 1210 return Error(Loc, "directional label undefined"); 1211 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym)); 1212 EndLoc = Lexer.getTok().getEndLoc(); 1213 Lex(); // Eat identifier. 1214 } 1215 } 1216 return false; 1217 } 1218 case AsmToken::Real: { 1219 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); 1220 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 1221 Res = MCConstantExpr::create(IntVal, getContext()); 1222 EndLoc = Lexer.getTok().getEndLoc(); 1223 Lex(); // Eat token. 1224 return false; 1225 } 1226 case AsmToken::Dot: { 1227 // This is a '.' reference, which references the current PC. Emit a 1228 // temporary label to the streamer and refer to it. 1229 MCSymbol *Sym = Ctx.createTempSymbol(); 1230 Out.EmitLabel(Sym); 1231 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext()); 1232 EndLoc = Lexer.getTok().getEndLoc(); 1233 Lex(); // Eat identifier. 1234 return false; 1235 } 1236 case AsmToken::LParen: 1237 Lex(); // Eat the '('. 1238 return parseParenExpr(Res, EndLoc); 1239 case AsmToken::LBrac: 1240 if (!PlatformParser->HasBracketExpressions()) 1241 return TokError("brackets expression not supported on this target"); 1242 Lex(); // Eat the '['. 1243 return parseBracketExpr(Res, EndLoc); 1244 case AsmToken::Minus: 1245 Lex(); // Eat the operator. 1246 if (parsePrimaryExpr(Res, EndLoc)) 1247 return true; 1248 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc); 1249 return false; 1250 case AsmToken::Plus: 1251 Lex(); // Eat the operator. 1252 if (parsePrimaryExpr(Res, EndLoc)) 1253 return true; 1254 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc); 1255 return false; 1256 case AsmToken::Tilde: 1257 Lex(); // Eat the operator. 1258 if (parsePrimaryExpr(Res, EndLoc)) 1259 return true; 1260 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1261 return false; 1262 // MIPS unary expression operators. The lexer won't generate these tokens if 1263 // MCAsmInfo::HasMipsExpressions is false for the target. 1264 case AsmToken::PercentCall16: 1265 case AsmToken::PercentCall_Hi: 1266 case AsmToken::PercentCall_Lo: 1267 case AsmToken::PercentDtprel_Hi: 1268 case AsmToken::PercentDtprel_Lo: 1269 case AsmToken::PercentGot: 1270 case AsmToken::PercentGot_Disp: 1271 case AsmToken::PercentGot_Hi: 1272 case AsmToken::PercentGot_Lo: 1273 case AsmToken::PercentGot_Ofst: 1274 case AsmToken::PercentGot_Page: 1275 case AsmToken::PercentGottprel: 1276 case AsmToken::PercentGp_Rel: 1277 case AsmToken::PercentHi: 1278 case AsmToken::PercentHigher: 1279 case AsmToken::PercentHighest: 1280 case AsmToken::PercentLo: 1281 case AsmToken::PercentNeg: 1282 case AsmToken::PercentPcrel_Hi: 1283 case AsmToken::PercentPcrel_Lo: 1284 case AsmToken::PercentTlsgd: 1285 case AsmToken::PercentTlsldm: 1286 case AsmToken::PercentTprel_Hi: 1287 case AsmToken::PercentTprel_Lo: 1288 Lex(); // Eat the operator. 1289 if (Lexer.isNot(AsmToken::LParen)) 1290 return TokError("expected '(' after operator"); 1291 Lex(); // Eat the operator. 1292 if (parseExpression(Res, EndLoc)) 1293 return true; 1294 if (Lexer.isNot(AsmToken::RParen)) 1295 return TokError("expected ')'"); 1296 Lex(); // Eat the operator. 1297 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx); 1298 return !Res; 1299 } 1300 } 1301 1302 bool AsmParser::parseExpression(const MCExpr *&Res) { 1303 SMLoc EndLoc; 1304 return parseExpression(Res, EndLoc); 1305 } 1306 1307 const MCExpr * 1308 AsmParser::applyModifierToExpr(const MCExpr *E, 1309 MCSymbolRefExpr::VariantKind Variant) { 1310 // Ask the target implementation about this expression first. 1311 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx); 1312 if (NewE) 1313 return NewE; 1314 // Recurse over the given expression, rebuilding it to apply the given variant 1315 // if there is exactly one symbol. 1316 switch (E->getKind()) { 1317 case MCExpr::Target: 1318 case MCExpr::Constant: 1319 return nullptr; 1320 1321 case MCExpr::SymbolRef: { 1322 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 1323 1324 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 1325 TokError("invalid variant on expression '" + getTok().getIdentifier() + 1326 "' (already modified)"); 1327 return E; 1328 } 1329 1330 return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext()); 1331 } 1332 1333 case MCExpr::Unary: { 1334 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 1335 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant); 1336 if (!Sub) 1337 return nullptr; 1338 return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext()); 1339 } 1340 1341 case MCExpr::Binary: { 1342 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 1343 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant); 1344 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant); 1345 1346 if (!LHS && !RHS) 1347 return nullptr; 1348 1349 if (!LHS) 1350 LHS = BE->getLHS(); 1351 if (!RHS) 1352 RHS = BE->getRHS(); 1353 1354 return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext()); 1355 } 1356 } 1357 1358 llvm_unreachable("Invalid expression kind!"); 1359 } 1360 1361 /// This function checks if the next token is <string> type or arithmetic. 1362 /// string that begin with character '<' must end with character '>'. 1363 /// otherwise it is arithmetics. 1364 /// If the function returns a 'true' value, 1365 /// the End argument will be filled with the last location pointed to the '>' 1366 /// character. 1367 1368 /// There is a gap between the AltMacro's documentation and the single quote 1369 /// implementation. GCC does not fully support this feature and so we will not 1370 /// support it. 1371 /// TODO: Adding single quote as a string. 1372 static bool isAltmacroString(SMLoc &StrLoc, SMLoc &EndLoc) { 1373 assert((StrLoc.getPointer() != nullptr) && 1374 "Argument to the function cannot be a NULL value"); 1375 const char *CharPtr = StrLoc.getPointer(); 1376 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && 1377 (*CharPtr != '\0')) { 1378 if (*CharPtr == '!') 1379 CharPtr++; 1380 CharPtr++; 1381 } 1382 if (*CharPtr == '>') { 1383 EndLoc = StrLoc.getFromPointer(CharPtr + 1); 1384 return true; 1385 } 1386 return false; 1387 } 1388 1389 /// creating a string without the escape characters '!'. 1390 static std::string altMacroString(StringRef AltMacroStr) { 1391 std::string Res; 1392 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) { 1393 if (AltMacroStr[Pos] == '!') 1394 Pos++; 1395 Res += AltMacroStr[Pos]; 1396 } 1397 return Res; 1398 } 1399 1400 /// Parse an expression and return it. 1401 /// 1402 /// expr ::= expr &&,|| expr -> lowest. 1403 /// expr ::= expr |,^,&,! expr 1404 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 1405 /// expr ::= expr <<,>> expr 1406 /// expr ::= expr +,- expr 1407 /// expr ::= expr *,/,% expr -> highest. 1408 /// expr ::= primaryexpr 1409 /// 1410 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1411 // Parse the expression. 1412 Res = nullptr; 1413 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || 1414 parseBinOpRHS(1, Res, EndLoc)) 1415 return true; 1416 1417 // As a special case, we support 'a op b @ modifier' by rewriting the 1418 // expression to include the modifier. This is inefficient, but in general we 1419 // expect users to use 'a@modifier op b'. 1420 if (Lexer.getKind() == AsmToken::At) { 1421 Lex(); 1422 1423 if (Lexer.isNot(AsmToken::Identifier)) 1424 return TokError("unexpected symbol modifier following '@'"); 1425 1426 MCSymbolRefExpr::VariantKind Variant = 1427 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 1428 if (Variant == MCSymbolRefExpr::VK_Invalid) 1429 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 1430 1431 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant); 1432 if (!ModifiedRes) { 1433 return TokError("invalid modifier '" + getTok().getIdentifier() + 1434 "' (no symbols present)"); 1435 } 1436 1437 Res = ModifiedRes; 1438 Lex(); 1439 } 1440 1441 // Try to constant fold it up front, if possible. Do not exploit 1442 // assembler here. 1443 int64_t Value; 1444 if (Res->evaluateAsAbsolute(Value)) 1445 Res = MCConstantExpr::create(Value, getContext()); 1446 1447 return false; 1448 } 1449 1450 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1451 Res = nullptr; 1452 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc); 1453 } 1454 1455 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 1456 SMLoc &EndLoc) { 1457 if (parseParenExpr(Res, EndLoc)) 1458 return true; 1459 1460 for (; ParenDepth > 0; --ParenDepth) { 1461 if (parseBinOpRHS(1, Res, EndLoc)) 1462 return true; 1463 1464 // We don't Lex() the last RParen. 1465 // This is the same behavior as parseParenExpression(). 1466 if (ParenDepth - 1 > 0) { 1467 EndLoc = getTok().getEndLoc(); 1468 if (parseToken(AsmToken::RParen, 1469 "expected ')' in parentheses expression")) 1470 return true; 1471 } 1472 } 1473 return false; 1474 } 1475 1476 bool AsmParser::parseAbsoluteExpression(int64_t &Res) { 1477 const MCExpr *Expr; 1478 1479 SMLoc StartLoc = Lexer.getLoc(); 1480 if (parseExpression(Expr)) 1481 return true; 1482 1483 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 1484 return Error(StartLoc, "expected absolute expression"); 1485 1486 return false; 1487 } 1488 1489 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K, 1490 MCBinaryExpr::Opcode &Kind, 1491 bool ShouldUseLogicalShr) { 1492 switch (K) { 1493 default: 1494 return 0; // not a binop. 1495 1496 // Lowest Precedence: &&, || 1497 case AsmToken::AmpAmp: 1498 Kind = MCBinaryExpr::LAnd; 1499 return 1; 1500 case AsmToken::PipePipe: 1501 Kind = MCBinaryExpr::LOr; 1502 return 1; 1503 1504 // Low Precedence: |, &, ^ 1505 // 1506 // FIXME: gas seems to support '!' as an infix operator? 1507 case AsmToken::Pipe: 1508 Kind = MCBinaryExpr::Or; 1509 return 2; 1510 case AsmToken::Caret: 1511 Kind = MCBinaryExpr::Xor; 1512 return 2; 1513 case AsmToken::Amp: 1514 Kind = MCBinaryExpr::And; 1515 return 2; 1516 1517 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 1518 case AsmToken::EqualEqual: 1519 Kind = MCBinaryExpr::EQ; 1520 return 3; 1521 case AsmToken::ExclaimEqual: 1522 case AsmToken::LessGreater: 1523 Kind = MCBinaryExpr::NE; 1524 return 3; 1525 case AsmToken::Less: 1526 Kind = MCBinaryExpr::LT; 1527 return 3; 1528 case AsmToken::LessEqual: 1529 Kind = MCBinaryExpr::LTE; 1530 return 3; 1531 case AsmToken::Greater: 1532 Kind = MCBinaryExpr::GT; 1533 return 3; 1534 case AsmToken::GreaterEqual: 1535 Kind = MCBinaryExpr::GTE; 1536 return 3; 1537 1538 // Intermediate Precedence: <<, >> 1539 case AsmToken::LessLess: 1540 Kind = MCBinaryExpr::Shl; 1541 return 4; 1542 case AsmToken::GreaterGreater: 1543 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1544 return 4; 1545 1546 // High Intermediate Precedence: +, - 1547 case AsmToken::Plus: 1548 Kind = MCBinaryExpr::Add; 1549 return 5; 1550 case AsmToken::Minus: 1551 Kind = MCBinaryExpr::Sub; 1552 return 5; 1553 1554 // Highest Precedence: *, /, % 1555 case AsmToken::Star: 1556 Kind = MCBinaryExpr::Mul; 1557 return 6; 1558 case AsmToken::Slash: 1559 Kind = MCBinaryExpr::Div; 1560 return 6; 1561 case AsmToken::Percent: 1562 Kind = MCBinaryExpr::Mod; 1563 return 6; 1564 } 1565 } 1566 1567 static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K, 1568 MCBinaryExpr::Opcode &Kind, 1569 bool ShouldUseLogicalShr) { 1570 switch (K) { 1571 default: 1572 return 0; // not a binop. 1573 1574 // Lowest Precedence: &&, || 1575 case AsmToken::AmpAmp: 1576 Kind = MCBinaryExpr::LAnd; 1577 return 2; 1578 case AsmToken::PipePipe: 1579 Kind = MCBinaryExpr::LOr; 1580 return 1; 1581 1582 // Low Precedence: ==, !=, <>, <, <=, >, >= 1583 case AsmToken::EqualEqual: 1584 Kind = MCBinaryExpr::EQ; 1585 return 3; 1586 case AsmToken::ExclaimEqual: 1587 case AsmToken::LessGreater: 1588 Kind = MCBinaryExpr::NE; 1589 return 3; 1590 case AsmToken::Less: 1591 Kind = MCBinaryExpr::LT; 1592 return 3; 1593 case AsmToken::LessEqual: 1594 Kind = MCBinaryExpr::LTE; 1595 return 3; 1596 case AsmToken::Greater: 1597 Kind = MCBinaryExpr::GT; 1598 return 3; 1599 case AsmToken::GreaterEqual: 1600 Kind = MCBinaryExpr::GTE; 1601 return 3; 1602 1603 // Low Intermediate Precedence: +, - 1604 case AsmToken::Plus: 1605 Kind = MCBinaryExpr::Add; 1606 return 4; 1607 case AsmToken::Minus: 1608 Kind = MCBinaryExpr::Sub; 1609 return 4; 1610 1611 // High Intermediate Precedence: |, &, ^ 1612 // 1613 // FIXME: gas seems to support '!' as an infix operator? 1614 case AsmToken::Pipe: 1615 Kind = MCBinaryExpr::Or; 1616 return 5; 1617 case AsmToken::Caret: 1618 Kind = MCBinaryExpr::Xor; 1619 return 5; 1620 case AsmToken::Amp: 1621 Kind = MCBinaryExpr::And; 1622 return 5; 1623 1624 // Highest Precedence: *, /, %, <<, >> 1625 case AsmToken::Star: 1626 Kind = MCBinaryExpr::Mul; 1627 return 6; 1628 case AsmToken::Slash: 1629 Kind = MCBinaryExpr::Div; 1630 return 6; 1631 case AsmToken::Percent: 1632 Kind = MCBinaryExpr::Mod; 1633 return 6; 1634 case AsmToken::LessLess: 1635 Kind = MCBinaryExpr::Shl; 1636 return 6; 1637 case AsmToken::GreaterGreater: 1638 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1639 return 6; 1640 } 1641 } 1642 1643 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K, 1644 MCBinaryExpr::Opcode &Kind) { 1645 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); 1646 return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr) 1647 : getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr); 1648 } 1649 1650 /// Parse all binary operators with precedence >= 'Precedence'. 1651 /// Res contains the LHS of the expression on input. 1652 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 1653 SMLoc &EndLoc) { 1654 SMLoc StartLoc = Lexer.getLoc(); 1655 while (true) { 1656 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 1657 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 1658 1659 // If the next token is lower precedence than we are allowed to eat, return 1660 // successfully with what we ate already. 1661 if (TokPrec < Precedence) 1662 return false; 1663 1664 Lex(); 1665 1666 // Eat the next primary expression. 1667 const MCExpr *RHS; 1668 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc)) 1669 return true; 1670 1671 // If BinOp binds less tightly with RHS than the operator after RHS, let 1672 // the pending operator take RHS as its LHS. 1673 MCBinaryExpr::Opcode Dummy; 1674 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1675 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc)) 1676 return true; 1677 1678 // Merge LHS and RHS according to operator. 1679 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc); 1680 } 1681 } 1682 1683 /// ParseStatement: 1684 /// ::= EndOfStatement 1685 /// ::= Label* Directive ...Operands... EndOfStatement 1686 /// ::= Label* Identifier OperandList* EndOfStatement 1687 bool AsmParser::parseStatement(ParseStatementInfo &Info, 1688 MCAsmParserSemaCallback *SI) { 1689 assert(!hasPendingError() && "parseStatement started with pending error"); 1690 // Eat initial spaces and comments 1691 while (Lexer.is(AsmToken::Space)) 1692 Lex(); 1693 if (Lexer.is(AsmToken::EndOfStatement)) { 1694 // if this is a line comment we can drop it safely 1695 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 1696 getTok().getString().front() == '\n') 1697 Out.AddBlankLine(); 1698 Lex(); 1699 return false; 1700 } 1701 // Statements always start with an identifier. 1702 AsmToken ID = getTok(); 1703 SMLoc IDLoc = ID.getLoc(); 1704 StringRef IDVal; 1705 int64_t LocalLabelVal = -1; 1706 if (Lexer.is(AsmToken::HashDirective)) 1707 return parseCppHashLineFilenameComment(IDLoc); 1708 // Allow an integer followed by a ':' as a directional local label. 1709 if (Lexer.is(AsmToken::Integer)) { 1710 LocalLabelVal = getTok().getIntVal(); 1711 if (LocalLabelVal < 0) { 1712 if (!TheCondState.Ignore) { 1713 Lex(); // always eat a token 1714 return Error(IDLoc, "unexpected token at start of statement"); 1715 } 1716 IDVal = ""; 1717 } else { 1718 IDVal = getTok().getString(); 1719 Lex(); // Consume the integer token to be used as an identifier token. 1720 if (Lexer.getKind() != AsmToken::Colon) { 1721 if (!TheCondState.Ignore) { 1722 Lex(); // always eat a token 1723 return Error(IDLoc, "unexpected token at start of statement"); 1724 } 1725 } 1726 } 1727 } else if (Lexer.is(AsmToken::Dot)) { 1728 // Treat '.' as a valid identifier in this context. 1729 Lex(); 1730 IDVal = "."; 1731 } else if (Lexer.is(AsmToken::LCurly)) { 1732 // Treat '{' as a valid identifier in this context. 1733 Lex(); 1734 IDVal = "{"; 1735 1736 } else if (Lexer.is(AsmToken::RCurly)) { 1737 // Treat '}' as a valid identifier in this context. 1738 Lex(); 1739 IDVal = "}"; 1740 } else if (Lexer.is(AsmToken::Star) && 1741 getTargetParser().starIsStartOfStatement()) { 1742 // Accept '*' as a valid start of statement. 1743 Lex(); 1744 IDVal = "*"; 1745 } else if (parseIdentifier(IDVal)) { 1746 if (!TheCondState.Ignore) { 1747 Lex(); // always eat a token 1748 return Error(IDLoc, "unexpected token at start of statement"); 1749 } 1750 IDVal = ""; 1751 } 1752 1753 // Handle conditional assembly here before checking for skipping. We 1754 // have to do this so that .endif isn't skipped in a ".if 0" block for 1755 // example. 1756 StringMap<DirectiveKind>::const_iterator DirKindIt = 1757 DirectiveKindMap.find(IDVal); 1758 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) 1759 1760 ? DK_NO_DIRECTIVE 1761 : DirKindIt->getValue(); 1762 switch (DirKind) { 1763 default: 1764 break; 1765 case DK_IF: 1766 case DK_IFEQ: 1767 case DK_IFGE: 1768 case DK_IFGT: 1769 case DK_IFLE: 1770 case DK_IFLT: 1771 case DK_IFNE: 1772 return parseDirectiveIf(IDLoc, DirKind); 1773 case DK_IFB: 1774 return parseDirectiveIfb(IDLoc, true); 1775 case DK_IFNB: 1776 return parseDirectiveIfb(IDLoc, false); 1777 case DK_IFC: 1778 return parseDirectiveIfc(IDLoc, true); 1779 case DK_IFEQS: 1780 return parseDirectiveIfeqs(IDLoc, true); 1781 case DK_IFNC: 1782 return parseDirectiveIfc(IDLoc, false); 1783 case DK_IFNES: 1784 return parseDirectiveIfeqs(IDLoc, false); 1785 case DK_IFDEF: 1786 return parseDirectiveIfdef(IDLoc, true); 1787 case DK_IFNDEF: 1788 case DK_IFNOTDEF: 1789 return parseDirectiveIfdef(IDLoc, false); 1790 case DK_ELSEIF: 1791 return parseDirectiveElseIf(IDLoc); 1792 case DK_ELSE: 1793 return parseDirectiveElse(IDLoc); 1794 case DK_ENDIF: 1795 return parseDirectiveEndIf(IDLoc); 1796 } 1797 1798 // Ignore the statement if in the middle of inactive conditional 1799 // (e.g. ".if 0"). 1800 if (TheCondState.Ignore) { 1801 eatToEndOfStatement(); 1802 return false; 1803 } 1804 1805 // FIXME: Recurse on local labels? 1806 1807 // See what kind of statement we have. 1808 switch (Lexer.getKind()) { 1809 case AsmToken::Colon: { 1810 if (!getTargetParser().isLabel(ID)) 1811 break; 1812 if (checkForValidSection()) 1813 return true; 1814 1815 // identifier ':' -> Label. 1816 Lex(); 1817 1818 // Diagnose attempt to use '.' as a label. 1819 if (IDVal == ".") 1820 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1821 1822 // Diagnose attempt to use a variable as a label. 1823 // 1824 // FIXME: Diagnostics. Note the location of the definition as a label. 1825 // FIXME: This doesn't diagnose assignment to a symbol which has been 1826 // implicitly marked as external. 1827 MCSymbol *Sym; 1828 if (LocalLabelVal == -1) { 1829 if (ParsingInlineAsm && SI) { 1830 StringRef RewrittenLabel = 1831 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true); 1832 assert(!RewrittenLabel.empty() && 1833 "We should have an internal name here."); 1834 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(), 1835 RewrittenLabel); 1836 IDVal = RewrittenLabel; 1837 } 1838 Sym = getContext().getOrCreateSymbol(IDVal); 1839 } else 1840 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal); 1841 // End of Labels should be treated as end of line for lexing 1842 // purposes but that information is not available to the Lexer who 1843 // does not understand Labels. This may cause us to see a Hash 1844 // here instead of a preprocessor line comment. 1845 if (getTok().is(AsmToken::Hash)) { 1846 StringRef CommentStr = parseStringToEndOfStatement(); 1847 Lexer.Lex(); 1848 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr)); 1849 } 1850 1851 // Consume any end of statement token, if present, to avoid spurious 1852 // AddBlankLine calls(). 1853 if (getTok().is(AsmToken::EndOfStatement)) { 1854 Lex(); 1855 } 1856 1857 getTargetParser().doBeforeLabelEmit(Sym); 1858 1859 // Emit the label. 1860 if (!getTargetParser().isParsingInlineAsm()) 1861 Out.EmitLabel(Sym, IDLoc); 1862 1863 // If we are generating dwarf for assembly source files then gather the 1864 // info to make a dwarf label entry for this label if needed. 1865 if (enabledGenDwarfForAssembly()) 1866 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 1867 IDLoc); 1868 1869 getTargetParser().onLabelParsed(Sym); 1870 1871 return false; 1872 } 1873 1874 case AsmToken::Equal: 1875 if (!getTargetParser().equalIsAsmAssignment()) 1876 break; 1877 // identifier '=' ... -> assignment statement 1878 Lex(); 1879 1880 return parseAssignment(IDVal, true); 1881 1882 default: // Normal instruction or directive. 1883 break; 1884 } 1885 1886 // If macros are enabled, check to see if this is a macro instantiation. 1887 if (areMacrosEnabled()) 1888 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) { 1889 return handleMacroEntry(M, IDLoc); 1890 } 1891 1892 // Otherwise, we have a normal instruction or directive. 1893 1894 // Directives start with "." 1895 if (IDVal.startswith(".") && IDVal != ".") { 1896 // There are several entities interested in parsing directives: 1897 // 1898 // 1. The target-specific assembly parser. Some directives are target 1899 // specific or may potentially behave differently on certain targets. 1900 // 2. Asm parser extensions. For example, platform-specific parsers 1901 // (like the ELF parser) register themselves as extensions. 1902 // 3. The generic directive parser implemented by this class. These are 1903 // all the directives that behave in a target and platform independent 1904 // manner, or at least have a default behavior that's shared between 1905 // all targets and platforms. 1906 1907 getTargetParser().flushPendingInstructions(getStreamer()); 1908 1909 SMLoc StartTokLoc = getTok().getLoc(); 1910 bool TPDirectiveReturn = getTargetParser().ParseDirective(ID); 1911 1912 if (hasPendingError()) 1913 return true; 1914 // Currently the return value should be true if we are 1915 // uninterested but as this is at odds with the standard parsing 1916 // convention (return true = error) we have instances of a parsed 1917 // directive that fails returning true as an error. Catch these 1918 // cases as best as possible errors here. 1919 if (TPDirectiveReturn && StartTokLoc != getTok().getLoc()) 1920 return true; 1921 // Return if we did some parsing or believe we succeeded. 1922 if (!TPDirectiveReturn || StartTokLoc != getTok().getLoc()) 1923 return false; 1924 1925 // Next, check the extension directive map to see if any extension has 1926 // registered itself to parse this directive. 1927 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 1928 ExtensionDirectiveMap.lookup(IDVal); 1929 if (Handler.first) 1930 return (*Handler.second)(Handler.first, IDVal, IDLoc); 1931 1932 // Finally, if no one else is interested in this directive, it must be 1933 // generic and familiar to this class. 1934 switch (DirKind) { 1935 default: 1936 break; 1937 case DK_SET: 1938 case DK_EQU: 1939 return parseDirectiveSet(IDVal, true); 1940 case DK_EQUIV: 1941 return parseDirectiveSet(IDVal, false); 1942 case DK_ASCII: 1943 return parseDirectiveAscii(IDVal, false); 1944 case DK_ASCIZ: 1945 case DK_STRING: 1946 return parseDirectiveAscii(IDVal, true); 1947 case DK_BYTE: 1948 case DK_DC_B: 1949 return parseDirectiveValue(IDVal, 1); 1950 case DK_DC: 1951 case DK_DC_W: 1952 case DK_SHORT: 1953 case DK_VALUE: 1954 case DK_2BYTE: 1955 return parseDirectiveValue(IDVal, 2); 1956 case DK_LONG: 1957 case DK_INT: 1958 case DK_4BYTE: 1959 case DK_DC_L: 1960 return parseDirectiveValue(IDVal, 4); 1961 case DK_QUAD: 1962 case DK_8BYTE: 1963 return parseDirectiveValue(IDVal, 8); 1964 case DK_DC_A: 1965 return parseDirectiveValue( 1966 IDVal, getContext().getAsmInfo()->getCodePointerSize()); 1967 case DK_OCTA: 1968 return parseDirectiveOctaValue(IDVal); 1969 case DK_SINGLE: 1970 case DK_FLOAT: 1971 case DK_DC_S: 1972 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle()); 1973 case DK_DOUBLE: 1974 case DK_DC_D: 1975 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble()); 1976 case DK_ALIGN: { 1977 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 1978 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1); 1979 } 1980 case DK_ALIGN32: { 1981 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 1982 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4); 1983 } 1984 case DK_BALIGN: 1985 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 1986 case DK_BALIGNW: 1987 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 1988 case DK_BALIGNL: 1989 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 1990 case DK_P2ALIGN: 1991 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 1992 case DK_P2ALIGNW: 1993 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 1994 case DK_P2ALIGNL: 1995 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 1996 case DK_ORG: 1997 return parseDirectiveOrg(); 1998 case DK_FILL: 1999 return parseDirectiveFill(); 2000 case DK_ZERO: 2001 return parseDirectiveZero(); 2002 case DK_EXTERN: 2003 eatToEndOfStatement(); // .extern is the default, ignore it. 2004 return false; 2005 case DK_GLOBL: 2006 case DK_GLOBAL: 2007 return parseDirectiveSymbolAttribute(MCSA_Global); 2008 case DK_LAZY_REFERENCE: 2009 return parseDirectiveSymbolAttribute(MCSA_LazyReference); 2010 case DK_NO_DEAD_STRIP: 2011 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 2012 case DK_SYMBOL_RESOLVER: 2013 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver); 2014 case DK_PRIVATE_EXTERN: 2015 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern); 2016 case DK_REFERENCE: 2017 return parseDirectiveSymbolAttribute(MCSA_Reference); 2018 case DK_WEAK_DEFINITION: 2019 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition); 2020 case DK_WEAK_REFERENCE: 2021 return parseDirectiveSymbolAttribute(MCSA_WeakReference); 2022 case DK_WEAK_DEF_CAN_BE_HIDDEN: 2023 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 2024 case DK_COLD: 2025 return parseDirectiveSymbolAttribute(MCSA_Cold); 2026 case DK_COMM: 2027 case DK_COMMON: 2028 return parseDirectiveComm(/*IsLocal=*/false); 2029 case DK_LCOMM: 2030 return parseDirectiveComm(/*IsLocal=*/true); 2031 case DK_ABORT: 2032 return parseDirectiveAbort(); 2033 case DK_INCLUDE: 2034 return parseDirectiveInclude(); 2035 case DK_INCBIN: 2036 return parseDirectiveIncbin(); 2037 case DK_CODE16: 2038 case DK_CODE16GCC: 2039 return TokError(Twine(IDVal) + 2040 " not currently supported for this target"); 2041 case DK_REPT: 2042 return parseDirectiveRept(IDLoc, IDVal); 2043 case DK_IRP: 2044 return parseDirectiveIrp(IDLoc); 2045 case DK_IRPC: 2046 return parseDirectiveIrpc(IDLoc); 2047 case DK_ENDR: 2048 return parseDirectiveEndr(IDLoc); 2049 case DK_BUNDLE_ALIGN_MODE: 2050 return parseDirectiveBundleAlignMode(); 2051 case DK_BUNDLE_LOCK: 2052 return parseDirectiveBundleLock(); 2053 case DK_BUNDLE_UNLOCK: 2054 return parseDirectiveBundleUnlock(); 2055 case DK_SLEB128: 2056 return parseDirectiveLEB128(true); 2057 case DK_ULEB128: 2058 return parseDirectiveLEB128(false); 2059 case DK_SPACE: 2060 case DK_SKIP: 2061 return parseDirectiveSpace(IDVal); 2062 case DK_FILE: 2063 return parseDirectiveFile(IDLoc); 2064 case DK_LINE: 2065 return parseDirectiveLine(); 2066 case DK_LOC: 2067 return parseDirectiveLoc(); 2068 case DK_STABS: 2069 return parseDirectiveStabs(); 2070 case DK_CV_FILE: 2071 return parseDirectiveCVFile(); 2072 case DK_CV_FUNC_ID: 2073 return parseDirectiveCVFuncId(); 2074 case DK_CV_INLINE_SITE_ID: 2075 return parseDirectiveCVInlineSiteId(); 2076 case DK_CV_LOC: 2077 return parseDirectiveCVLoc(); 2078 case DK_CV_LINETABLE: 2079 return parseDirectiveCVLinetable(); 2080 case DK_CV_INLINE_LINETABLE: 2081 return parseDirectiveCVInlineLinetable(); 2082 case DK_CV_DEF_RANGE: 2083 return parseDirectiveCVDefRange(); 2084 case DK_CV_STRING: 2085 return parseDirectiveCVString(); 2086 case DK_CV_STRINGTABLE: 2087 return parseDirectiveCVStringTable(); 2088 case DK_CV_FILECHECKSUMS: 2089 return parseDirectiveCVFileChecksums(); 2090 case DK_CV_FILECHECKSUM_OFFSET: 2091 return parseDirectiveCVFileChecksumOffset(); 2092 case DK_CV_FPO_DATA: 2093 return parseDirectiveCVFPOData(); 2094 case DK_CFI_SECTIONS: 2095 return parseDirectiveCFISections(); 2096 case DK_CFI_STARTPROC: 2097 return parseDirectiveCFIStartProc(); 2098 case DK_CFI_ENDPROC: 2099 return parseDirectiveCFIEndProc(); 2100 case DK_CFI_DEF_CFA: 2101 return parseDirectiveCFIDefCfa(IDLoc); 2102 case DK_CFI_DEF_CFA_OFFSET: 2103 return parseDirectiveCFIDefCfaOffset(); 2104 case DK_CFI_ADJUST_CFA_OFFSET: 2105 return parseDirectiveCFIAdjustCfaOffset(); 2106 case DK_CFI_DEF_CFA_REGISTER: 2107 return parseDirectiveCFIDefCfaRegister(IDLoc); 2108 case DK_CFI_OFFSET: 2109 return parseDirectiveCFIOffset(IDLoc); 2110 case DK_CFI_REL_OFFSET: 2111 return parseDirectiveCFIRelOffset(IDLoc); 2112 case DK_CFI_PERSONALITY: 2113 return parseDirectiveCFIPersonalityOrLsda(true); 2114 case DK_CFI_LSDA: 2115 return parseDirectiveCFIPersonalityOrLsda(false); 2116 case DK_CFI_REMEMBER_STATE: 2117 return parseDirectiveCFIRememberState(); 2118 case DK_CFI_RESTORE_STATE: 2119 return parseDirectiveCFIRestoreState(); 2120 case DK_CFI_SAME_VALUE: 2121 return parseDirectiveCFISameValue(IDLoc); 2122 case DK_CFI_RESTORE: 2123 return parseDirectiveCFIRestore(IDLoc); 2124 case DK_CFI_ESCAPE: 2125 return parseDirectiveCFIEscape(); 2126 case DK_CFI_RETURN_COLUMN: 2127 return parseDirectiveCFIReturnColumn(IDLoc); 2128 case DK_CFI_SIGNAL_FRAME: 2129 return parseDirectiveCFISignalFrame(); 2130 case DK_CFI_UNDEFINED: 2131 return parseDirectiveCFIUndefined(IDLoc); 2132 case DK_CFI_REGISTER: 2133 return parseDirectiveCFIRegister(IDLoc); 2134 case DK_CFI_WINDOW_SAVE: 2135 return parseDirectiveCFIWindowSave(); 2136 case DK_MACROS_ON: 2137 case DK_MACROS_OFF: 2138 return parseDirectiveMacrosOnOff(IDVal); 2139 case DK_MACRO: 2140 return parseDirectiveMacro(IDLoc); 2141 case DK_ALTMACRO: 2142 case DK_NOALTMACRO: 2143 return parseDirectiveAltmacro(IDVal); 2144 case DK_EXITM: 2145 return parseDirectiveExitMacro(IDVal); 2146 case DK_ENDM: 2147 case DK_ENDMACRO: 2148 return parseDirectiveEndMacro(IDVal); 2149 case DK_PURGEM: 2150 return parseDirectivePurgeMacro(IDLoc); 2151 case DK_END: 2152 return parseDirectiveEnd(IDLoc); 2153 case DK_ERR: 2154 return parseDirectiveError(IDLoc, false); 2155 case DK_ERROR: 2156 return parseDirectiveError(IDLoc, true); 2157 case DK_WARNING: 2158 return parseDirectiveWarning(IDLoc); 2159 case DK_RELOC: 2160 return parseDirectiveReloc(IDLoc); 2161 case DK_DCB: 2162 case DK_DCB_W: 2163 return parseDirectiveDCB(IDVal, 2); 2164 case DK_DCB_B: 2165 return parseDirectiveDCB(IDVal, 1); 2166 case DK_DCB_D: 2167 return parseDirectiveRealDCB(IDVal, APFloat::IEEEdouble()); 2168 case DK_DCB_L: 2169 return parseDirectiveDCB(IDVal, 4); 2170 case DK_DCB_S: 2171 return parseDirectiveRealDCB(IDVal, APFloat::IEEEsingle()); 2172 case DK_DC_X: 2173 case DK_DCB_X: 2174 return TokError(Twine(IDVal) + 2175 " not currently supported for this target"); 2176 case DK_DS: 2177 case DK_DS_W: 2178 return parseDirectiveDS(IDVal, 2); 2179 case DK_DS_B: 2180 return parseDirectiveDS(IDVal, 1); 2181 case DK_DS_D: 2182 return parseDirectiveDS(IDVal, 8); 2183 case DK_DS_L: 2184 case DK_DS_S: 2185 return parseDirectiveDS(IDVal, 4); 2186 case DK_DS_P: 2187 case DK_DS_X: 2188 return parseDirectiveDS(IDVal, 12); 2189 case DK_PRINT: 2190 return parseDirectivePrint(IDLoc); 2191 case DK_ADDRSIG: 2192 return parseDirectiveAddrsig(); 2193 case DK_ADDRSIG_SYM: 2194 return parseDirectiveAddrsigSym(); 2195 } 2196 2197 return Error(IDLoc, "unknown directive"); 2198 } 2199 2200 // __asm _emit or __asm __emit 2201 if (ParsingInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || 2202 IDVal == "_EMIT" || IDVal == "__EMIT")) 2203 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size()); 2204 2205 // __asm align 2206 if (ParsingInlineAsm && (IDVal == "align" || IDVal == "ALIGN")) 2207 return parseDirectiveMSAlign(IDLoc, Info); 2208 2209 if (ParsingInlineAsm && (IDVal == "even" || IDVal == "EVEN")) 2210 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4); 2211 if (checkForValidSection()) 2212 return true; 2213 2214 // Canonicalize the opcode to lower case. 2215 std::string OpcodeStr = IDVal.lower(); 2216 ParseInstructionInfo IInfo(Info.AsmRewrites); 2217 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID, 2218 Info.ParsedOperands); 2219 Info.ParseError = ParseHadError; 2220 2221 // Dump the parsed representation, if requested. 2222 if (getShowParsedOperands()) { 2223 SmallString<256> Str; 2224 raw_svector_ostream OS(Str); 2225 OS << "parsed instruction: ["; 2226 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { 2227 if (i != 0) 2228 OS << ", "; 2229 Info.ParsedOperands[i]->print(OS); 2230 } 2231 OS << "]"; 2232 2233 printMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 2234 } 2235 2236 // Fail even if ParseInstruction erroneously returns false. 2237 if (hasPendingError() || ParseHadError) 2238 return true; 2239 2240 // If we are generating dwarf for the current section then generate a .loc 2241 // directive for the instruction. 2242 if (!ParseHadError && enabledGenDwarfForAssembly() && 2243 getContext().getGenDwarfSectionSyms().count( 2244 getStreamer().getCurrentSectionOnly())) { 2245 unsigned Line; 2246 if (ActiveMacros.empty()) 2247 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer); 2248 else 2249 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc, 2250 ActiveMacros.front()->ExitBuffer); 2251 2252 // If we previously parsed a cpp hash file line comment then make sure the 2253 // current Dwarf File is for the CppHashFilename if not then emit the 2254 // Dwarf File table for it and adjust the line number for the .loc. 2255 if (!CppHashInfo.Filename.empty()) { 2256 unsigned FileNumber = getStreamer().EmitDwarfFileDirective( 2257 0, StringRef(), CppHashInfo.Filename); 2258 getContext().setGenDwarfFileNumber(FileNumber); 2259 2260 unsigned CppHashLocLineNo = 2261 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf); 2262 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); 2263 } 2264 2265 getStreamer().EmitDwarfLocDirective( 2266 getContext().getGenDwarfFileNumber(), Line, 0, 2267 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0, 2268 StringRef()); 2269 } 2270 2271 // If parsing succeeded, match the instruction. 2272 if (!ParseHadError) { 2273 uint64_t ErrorInfo; 2274 if (getTargetParser().MatchAndEmitInstruction( 2275 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo, 2276 getTargetParser().isParsingInlineAsm())) 2277 return true; 2278 } 2279 return false; 2280 } 2281 2282 // Parse and erase curly braces marking block start/end 2283 bool 2284 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { 2285 // Identify curly brace marking block start/end 2286 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly)) 2287 return false; 2288 2289 SMLoc StartLoc = Lexer.getLoc(); 2290 Lex(); // Eat the brace 2291 if (Lexer.is(AsmToken::EndOfStatement)) 2292 Lex(); // Eat EndOfStatement following the brace 2293 2294 // Erase the block start/end brace from the output asm string 2295 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() - 2296 StartLoc.getPointer()); 2297 return true; 2298 } 2299 2300 /// parseCppHashLineFilenameComment as this: 2301 /// ::= # number "filename" 2302 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L) { 2303 Lex(); // Eat the hash token. 2304 // Lexer only ever emits HashDirective if it fully formed if it's 2305 // done the checking already so this is an internal error. 2306 assert(getTok().is(AsmToken::Integer) && 2307 "Lexing Cpp line comment: Expected Integer"); 2308 int64_t LineNumber = getTok().getIntVal(); 2309 Lex(); 2310 assert(getTok().is(AsmToken::String) && 2311 "Lexing Cpp line comment: Expected String"); 2312 StringRef Filename = getTok().getString(); 2313 Lex(); 2314 2315 // Get rid of the enclosing quotes. 2316 Filename = Filename.substr(1, Filename.size() - 2); 2317 2318 // Save the SMLoc, Filename and LineNumber for later use by diagnostics 2319 // and possibly DWARF file info. 2320 CppHashInfo.Loc = L; 2321 CppHashInfo.Filename = Filename; 2322 CppHashInfo.LineNumber = LineNumber; 2323 CppHashInfo.Buf = CurBuffer; 2324 if (FirstCppHashFilename.empty()) 2325 FirstCppHashFilename = Filename; 2326 return false; 2327 } 2328 2329 /// will use the last parsed cpp hash line filename comment 2330 /// for the Filename and LineNo if any in the diagnostic. 2331 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 2332 const AsmParser *Parser = static_cast<const AsmParser *>(Context); 2333 raw_ostream &OS = errs(); 2334 2335 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 2336 SMLoc DiagLoc = Diag.getLoc(); 2337 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2338 unsigned CppHashBuf = 2339 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc); 2340 2341 // Like SourceMgr::printMessage() we need to print the include stack if any 2342 // before printing the message. 2343 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2344 if (!Parser->SavedDiagHandler && DiagCurBuffer && 2345 DiagCurBuffer != DiagSrcMgr.getMainFileID()) { 2346 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 2347 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 2348 } 2349 2350 // If we have not parsed a cpp hash line filename comment or the source 2351 // manager changed or buffer changed (like in a nested include) then just 2352 // print the normal diagnostic using its Filename and LineNo. 2353 if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr || 2354 DiagBuf != CppHashBuf) { 2355 if (Parser->SavedDiagHandler) 2356 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2357 else 2358 Diag.print(nullptr, OS); 2359 return; 2360 } 2361 2362 // Use the CppHashFilename and calculate a line number based on the 2363 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc 2364 // for the diagnostic. 2365 const std::string &Filename = Parser->CppHashInfo.Filename; 2366 2367 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 2368 int CppHashLocLineNo = 2369 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf); 2370 int LineNo = 2371 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); 2372 2373 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, 2374 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), 2375 Diag.getLineContents(), Diag.getRanges()); 2376 2377 if (Parser->SavedDiagHandler) 2378 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); 2379 else 2380 NewDiag.print(nullptr, OS); 2381 } 2382 2383 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The 2384 // difference being that that function accepts '@' as part of identifiers and 2385 // we can't do that. AsmLexer.cpp should probably be changed to handle 2386 // '@' as a special case when needed. 2387 static bool isIdentifierChar(char c) { 2388 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' || 2389 c == '.'; 2390 } 2391 2392 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 2393 ArrayRef<MCAsmMacroParameter> Parameters, 2394 ArrayRef<MCAsmMacroArgument> A, 2395 bool EnableAtPseudoVariable, SMLoc L) { 2396 unsigned NParameters = Parameters.size(); 2397 bool HasVararg = NParameters ? Parameters.back().Vararg : false; 2398 if ((!IsDarwin || NParameters != 0) && NParameters != A.size()) 2399 return Error(L, "Wrong number of arguments"); 2400 2401 // A macro without parameters is handled differently on Darwin: 2402 // gas accepts no arguments and does no substitutions 2403 while (!Body.empty()) { 2404 // Scan for the next substitution. 2405 std::size_t End = Body.size(), Pos = 0; 2406 for (; Pos != End; ++Pos) { 2407 // Check for a substitution or escape. 2408 if (IsDarwin && !NParameters) { 2409 // This macro has no parameters, look for $0, $1, etc. 2410 if (Body[Pos] != '$' || Pos + 1 == End) 2411 continue; 2412 2413 char Next = Body[Pos + 1]; 2414 if (Next == '$' || Next == 'n' || 2415 isdigit(static_cast<unsigned char>(Next))) 2416 break; 2417 } else { 2418 // This macro has parameters, look for \foo, \bar, etc. 2419 if (Body[Pos] == '\\' && Pos + 1 != End) 2420 break; 2421 } 2422 } 2423 2424 // Add the prefix. 2425 OS << Body.slice(0, Pos); 2426 2427 // Check if we reached the end. 2428 if (Pos == End) 2429 break; 2430 2431 if (IsDarwin && !NParameters) { 2432 switch (Body[Pos + 1]) { 2433 // $$ => $ 2434 case '$': 2435 OS << '$'; 2436 break; 2437 2438 // $n => number of arguments 2439 case 'n': 2440 OS << A.size(); 2441 break; 2442 2443 // $[0-9] => argument 2444 default: { 2445 // Missing arguments are ignored. 2446 unsigned Index = Body[Pos + 1] - '0'; 2447 if (Index >= A.size()) 2448 break; 2449 2450 // Otherwise substitute with the token values, with spaces eliminated. 2451 for (const AsmToken &Token : A[Index]) 2452 OS << Token.getString(); 2453 break; 2454 } 2455 } 2456 Pos += 2; 2457 } else { 2458 unsigned I = Pos + 1; 2459 2460 // Check for the \@ pseudo-variable. 2461 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End) 2462 ++I; 2463 else 2464 while (isIdentifierChar(Body[I]) && I + 1 != End) 2465 ++I; 2466 2467 const char *Begin = Body.data() + Pos + 1; 2468 StringRef Argument(Begin, I - (Pos + 1)); 2469 unsigned Index = 0; 2470 2471 if (Argument == "@") { 2472 OS << NumOfMacroInstantiations; 2473 Pos += 2; 2474 } else { 2475 for (; Index < NParameters; ++Index) 2476 if (Parameters[Index].Name == Argument) 2477 break; 2478 2479 if (Index == NParameters) { 2480 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 2481 Pos += 3; 2482 else { 2483 OS << '\\' << Argument; 2484 Pos = I; 2485 } 2486 } else { 2487 bool VarargParameter = HasVararg && Index == (NParameters - 1); 2488 for (const AsmToken &Token : A[Index]) 2489 // For altmacro mode, you can write '%expr'. 2490 // The prefix '%' evaluates the expression 'expr' 2491 // and uses the result as a string (e.g. replace %(1+2) with the 2492 // string "3"). 2493 // Here, we identify the integer token which is the result of the 2494 // absolute expression evaluation and replace it with its string 2495 // representation. 2496 if (AltMacroMode && Token.getString().front() == '%' && 2497 Token.is(AsmToken::Integer)) 2498 // Emit an integer value to the buffer. 2499 OS << Token.getIntVal(); 2500 // Only Token that was validated as a string and begins with '<' 2501 // is considered altMacroString!!! 2502 else if (AltMacroMode && Token.getString().front() == '<' && 2503 Token.is(AsmToken::String)) { 2504 OS << altMacroString(Token.getStringContents()); 2505 } 2506 // We expect no quotes around the string's contents when 2507 // parsing for varargs. 2508 else if (Token.isNot(AsmToken::String) || VarargParameter) 2509 OS << Token.getString(); 2510 else 2511 OS << Token.getStringContents(); 2512 2513 Pos += 1 + Argument.size(); 2514 } 2515 } 2516 } 2517 // Update the scan point. 2518 Body = Body.substr(Pos); 2519 } 2520 2521 return false; 2522 } 2523 2524 static bool isOperator(AsmToken::TokenKind kind) { 2525 switch (kind) { 2526 default: 2527 return false; 2528 case AsmToken::Plus: 2529 case AsmToken::Minus: 2530 case AsmToken::Tilde: 2531 case AsmToken::Slash: 2532 case AsmToken::Star: 2533 case AsmToken::Dot: 2534 case AsmToken::Equal: 2535 case AsmToken::EqualEqual: 2536 case AsmToken::Pipe: 2537 case AsmToken::PipePipe: 2538 case AsmToken::Caret: 2539 case AsmToken::Amp: 2540 case AsmToken::AmpAmp: 2541 case AsmToken::Exclaim: 2542 case AsmToken::ExclaimEqual: 2543 case AsmToken::Less: 2544 case AsmToken::LessEqual: 2545 case AsmToken::LessLess: 2546 case AsmToken::LessGreater: 2547 case AsmToken::Greater: 2548 case AsmToken::GreaterEqual: 2549 case AsmToken::GreaterGreater: 2550 return true; 2551 } 2552 } 2553 2554 namespace { 2555 2556 class AsmLexerSkipSpaceRAII { 2557 public: 2558 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { 2559 Lexer.setSkipSpace(SkipSpace); 2560 } 2561 2562 ~AsmLexerSkipSpaceRAII() { 2563 Lexer.setSkipSpace(true); 2564 } 2565 2566 private: 2567 AsmLexer &Lexer; 2568 }; 2569 2570 } // end anonymous namespace 2571 2572 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) { 2573 2574 if (Vararg) { 2575 if (Lexer.isNot(AsmToken::EndOfStatement)) { 2576 StringRef Str = parseStringToEndOfStatement(); 2577 MA.emplace_back(AsmToken::String, Str); 2578 } 2579 return false; 2580 } 2581 2582 unsigned ParenLevel = 0; 2583 2584 // Darwin doesn't use spaces to delmit arguments. 2585 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); 2586 2587 bool SpaceEaten; 2588 2589 while (true) { 2590 SpaceEaten = false; 2591 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 2592 return TokError("unexpected token in macro instantiation"); 2593 2594 if (ParenLevel == 0) { 2595 2596 if (Lexer.is(AsmToken::Comma)) 2597 break; 2598 2599 if (Lexer.is(AsmToken::Space)) { 2600 SpaceEaten = true; 2601 Lexer.Lex(); // Eat spaces 2602 } 2603 2604 // Spaces can delimit parameters, but could also be part an expression. 2605 // If the token after a space is an operator, add the token and the next 2606 // one into this argument 2607 if (!IsDarwin) { 2608 if (isOperator(Lexer.getKind())) { 2609 MA.push_back(getTok()); 2610 Lexer.Lex(); 2611 2612 // Whitespace after an operator can be ignored. 2613 if (Lexer.is(AsmToken::Space)) 2614 Lexer.Lex(); 2615 2616 continue; 2617 } 2618 } 2619 if (SpaceEaten) 2620 break; 2621 } 2622 2623 // handleMacroEntry relies on not advancing the lexer here 2624 // to be able to fill in the remaining default parameter values 2625 if (Lexer.is(AsmToken::EndOfStatement)) 2626 break; 2627 2628 // Adjust the current parentheses level. 2629 if (Lexer.is(AsmToken::LParen)) 2630 ++ParenLevel; 2631 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 2632 --ParenLevel; 2633 2634 // Append the token to the current argument list. 2635 MA.push_back(getTok()); 2636 Lexer.Lex(); 2637 } 2638 2639 if (ParenLevel != 0) 2640 return TokError("unbalanced parentheses in macro argument"); 2641 return false; 2642 } 2643 2644 // Parse the macro instantiation arguments. 2645 bool AsmParser::parseMacroArguments(const MCAsmMacro *M, 2646 MCAsmMacroArguments &A) { 2647 const unsigned NParameters = M ? M->Parameters.size() : 0; 2648 bool NamedParametersFound = false; 2649 SmallVector<SMLoc, 4> FALocs; 2650 2651 A.resize(NParameters); 2652 FALocs.resize(NParameters); 2653 2654 // Parse two kinds of macro invocations: 2655 // - macros defined without any parameters accept an arbitrary number of them 2656 // - macros defined with parameters accept at most that many of them 2657 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false; 2658 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 2659 ++Parameter) { 2660 SMLoc IDLoc = Lexer.getLoc(); 2661 MCAsmMacroParameter FA; 2662 2663 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) { 2664 if (parseIdentifier(FA.Name)) 2665 return Error(IDLoc, "invalid argument identifier for formal argument"); 2666 2667 if (Lexer.isNot(AsmToken::Equal)) 2668 return TokError("expected '=' after formal parameter identifier"); 2669 2670 Lex(); 2671 2672 NamedParametersFound = true; 2673 } 2674 bool Vararg = HasVararg && Parameter == (NParameters - 1); 2675 2676 if (NamedParametersFound && FA.Name.empty()) 2677 return Error(IDLoc, "cannot mix positional and keyword arguments"); 2678 2679 SMLoc StrLoc = Lexer.getLoc(); 2680 SMLoc EndLoc; 2681 if (AltMacroMode && Lexer.is(AsmToken::Percent)) { 2682 const MCExpr *AbsoluteExp; 2683 int64_t Value; 2684 /// Eat '%' 2685 Lex(); 2686 if (parseExpression(AbsoluteExp, EndLoc)) 2687 return false; 2688 if (!AbsoluteExp->evaluateAsAbsolute(Value, 2689 getStreamer().getAssemblerPtr())) 2690 return Error(StrLoc, "expected absolute expression"); 2691 const char *StrChar = StrLoc.getPointer(); 2692 const char *EndChar = EndLoc.getPointer(); 2693 AsmToken newToken(AsmToken::Integer, 2694 StringRef(StrChar, EndChar - StrChar), Value); 2695 FA.Value.push_back(newToken); 2696 } else if (AltMacroMode && Lexer.is(AsmToken::Less) && 2697 isAltmacroString(StrLoc, EndLoc)) { 2698 const char *StrChar = StrLoc.getPointer(); 2699 const char *EndChar = EndLoc.getPointer(); 2700 jumpToLoc(EndLoc, CurBuffer); 2701 /// Eat from '<' to '>' 2702 Lex(); 2703 AsmToken newToken(AsmToken::String, 2704 StringRef(StrChar, EndChar - StrChar)); 2705 FA.Value.push_back(newToken); 2706 } else if(parseMacroArgument(FA.Value, Vararg)) 2707 return true; 2708 2709 unsigned PI = Parameter; 2710 if (!FA.Name.empty()) { 2711 unsigned FAI = 0; 2712 for (FAI = 0; FAI < NParameters; ++FAI) 2713 if (M->Parameters[FAI].Name == FA.Name) 2714 break; 2715 2716 if (FAI >= NParameters) { 2717 assert(M && "expected macro to be defined"); 2718 return Error(IDLoc, "parameter named '" + FA.Name + 2719 "' does not exist for macro '" + M->Name + "'"); 2720 } 2721 PI = FAI; 2722 } 2723 2724 if (!FA.Value.empty()) { 2725 if (A.size() <= PI) 2726 A.resize(PI + 1); 2727 A[PI] = FA.Value; 2728 2729 if (FALocs.size() <= PI) 2730 FALocs.resize(PI + 1); 2731 2732 FALocs[PI] = Lexer.getLoc(); 2733 } 2734 2735 // At the end of the statement, fill in remaining arguments that have 2736 // default values. If there aren't any, then the next argument is 2737 // required but missing 2738 if (Lexer.is(AsmToken::EndOfStatement)) { 2739 bool Failure = false; 2740 for (unsigned FAI = 0; FAI < NParameters; ++FAI) { 2741 if (A[FAI].empty()) { 2742 if (M->Parameters[FAI].Required) { 2743 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), 2744 "missing value for required parameter " 2745 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'"); 2746 Failure = true; 2747 } 2748 2749 if (!M->Parameters[FAI].Value.empty()) 2750 A[FAI] = M->Parameters[FAI].Value; 2751 } 2752 } 2753 return Failure; 2754 } 2755 2756 if (Lexer.is(AsmToken::Comma)) 2757 Lex(); 2758 } 2759 2760 return TokError("too many positional arguments"); 2761 } 2762 2763 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) { 2764 // Arbitrarily limit macro nesting depth (default matches 'as'). We can 2765 // eliminate this, although we should protect against infinite loops. 2766 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; 2767 if (ActiveMacros.size() == MaxNestingDepth) { 2768 std::ostringstream MaxNestingDepthError; 2769 MaxNestingDepthError << "macros cannot be nested more than " 2770 << MaxNestingDepth << " levels deep." 2771 << " Use -asm-macro-max-nesting-depth to increase " 2772 "this limit."; 2773 return TokError(MaxNestingDepthError.str()); 2774 } 2775 2776 MCAsmMacroArguments A; 2777 if (parseMacroArguments(M, A)) 2778 return true; 2779 2780 // Macro instantiation is lexical, unfortunately. We construct a new buffer 2781 // to hold the macro body with substitutions. 2782 SmallString<256> Buf; 2783 StringRef Body = M->Body; 2784 raw_svector_ostream OS(Buf); 2785 2786 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc())) 2787 return true; 2788 2789 // We include the .endmacro in the buffer as our cue to exit the macro 2790 // instantiation. 2791 OS << ".endmacro\n"; 2792 2793 std::unique_ptr<MemoryBuffer> Instantiation = 2794 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 2795 2796 // Create the macro instantiation object and add to the current macro 2797 // instantiation stack. 2798 MacroInstantiation *MI = new MacroInstantiation{ 2799 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 2800 ActiveMacros.push_back(MI); 2801 2802 ++NumOfMacroInstantiations; 2803 2804 // Jump to the macro instantiation and prime the lexer. 2805 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 2806 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 2807 Lex(); 2808 2809 return false; 2810 } 2811 2812 void AsmParser::handleMacroExit() { 2813 // Jump to the EndOfStatement we should return to, and consume it. 2814 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer); 2815 Lex(); 2816 2817 // Pop the instantiation entry. 2818 delete ActiveMacros.back(); 2819 ActiveMacros.pop_back(); 2820 } 2821 2822 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef, 2823 bool NoDeadStrip) { 2824 MCSymbol *Sym; 2825 const MCExpr *Value; 2826 if (MCParserUtils::parseAssignmentExpression(Name, allow_redef, *this, Sym, 2827 Value)) 2828 return true; 2829 2830 if (!Sym) { 2831 // In the case where we parse an expression starting with a '.', we will 2832 // not generate an error, nor will we create a symbol. In this case we 2833 // should just return out. 2834 return false; 2835 } 2836 2837 // Do the assignment. 2838 Out.EmitAssignment(Sym, Value); 2839 if (NoDeadStrip) 2840 Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip); 2841 2842 return false; 2843 } 2844 2845 /// parseIdentifier: 2846 /// ::= identifier 2847 /// ::= string 2848 bool AsmParser::parseIdentifier(StringRef &Res) { 2849 // The assembler has relaxed rules for accepting identifiers, in particular we 2850 // allow things like '.globl $foo' and '.def @feat.00', which would normally be 2851 // separate tokens. At this level, we have already lexed so we cannot (currently) 2852 // handle this as a context dependent token, instead we detect adjacent tokens 2853 // and return the combined identifier. 2854 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) { 2855 SMLoc PrefixLoc = getLexer().getLoc(); 2856 2857 // Consume the prefix character, and check for a following identifier. 2858 2859 AsmToken Buf[1]; 2860 Lexer.peekTokens(Buf, false); 2861 2862 if (Buf[0].isNot(AsmToken::Identifier)) 2863 return true; 2864 2865 // We have a '$' or '@' followed by an identifier, make sure they are adjacent. 2866 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer()) 2867 return true; 2868 2869 // eat $ or @ 2870 Lexer.Lex(); // Lexer's Lex guarantees consecutive token. 2871 // Construct the joined identifier and consume the token. 2872 Res = 2873 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1); 2874 Lex(); // Parser Lex to maintain invariants. 2875 return false; 2876 } 2877 2878 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String)) 2879 return true; 2880 2881 Res = getTok().getIdentifier(); 2882 2883 Lex(); // Consume the identifier token. 2884 2885 return false; 2886 } 2887 2888 /// parseDirectiveSet: 2889 /// ::= .equ identifier ',' expression 2890 /// ::= .equiv identifier ',' expression 2891 /// ::= .set identifier ',' expression 2892 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) { 2893 StringRef Name; 2894 if (check(parseIdentifier(Name), "expected identifier") || 2895 parseToken(AsmToken::Comma) || parseAssignment(Name, allow_redef, true)) 2896 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 2897 return false; 2898 } 2899 2900 bool AsmParser::parseEscapedString(std::string &Data) { 2901 if (check(getTok().isNot(AsmToken::String), "expected string")) 2902 return true; 2903 2904 Data = ""; 2905 StringRef Str = getTok().getStringContents(); 2906 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 2907 if (Str[i] != '\\') { 2908 Data += Str[i]; 2909 continue; 2910 } 2911 2912 // Recognize escaped characters. Note that this escape semantics currently 2913 // loosely follows Darwin 'as'. 2914 ++i; 2915 if (i == e) 2916 return TokError("unexpected backslash at end of string"); 2917 2918 // Recognize hex sequences similarly to GNU 'as'. 2919 if (Str[i] == 'x' || Str[i] == 'X') { 2920 size_t length = Str.size(); 2921 if (i + 1 >= length || !isHexDigit(Str[i + 1])) 2922 return TokError("invalid hexadecimal escape sequence"); 2923 2924 // Consume hex characters. GNU 'as' reads all hexadecimal characters and 2925 // then truncates to the lower 16 bits. Seems reasonable. 2926 unsigned Value = 0; 2927 while (i + 1 < length && isHexDigit(Str[i + 1])) 2928 Value = Value * 16 + hexDigitValue(Str[++i]); 2929 2930 Data += (unsigned char)(Value & 0xFF); 2931 continue; 2932 } 2933 2934 // Recognize octal sequences. 2935 if ((unsigned)(Str[i] - '0') <= 7) { 2936 // Consume up to three octal characters. 2937 unsigned Value = Str[i] - '0'; 2938 2939 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 2940 ++i; 2941 Value = Value * 8 + (Str[i] - '0'); 2942 2943 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 2944 ++i; 2945 Value = Value * 8 + (Str[i] - '0'); 2946 } 2947 } 2948 2949 if (Value > 255) 2950 return TokError("invalid octal escape sequence (out of range)"); 2951 2952 Data += (unsigned char)Value; 2953 continue; 2954 } 2955 2956 // Otherwise recognize individual escapes. 2957 switch (Str[i]) { 2958 default: 2959 // Just reject invalid escape sequences for now. 2960 return TokError("invalid escape sequence (unrecognized character)"); 2961 2962 case 'b': Data += '\b'; break; 2963 case 'f': Data += '\f'; break; 2964 case 'n': Data += '\n'; break; 2965 case 'r': Data += '\r'; break; 2966 case 't': Data += '\t'; break; 2967 case '"': Data += '"'; break; 2968 case '\\': Data += '\\'; break; 2969 } 2970 } 2971 2972 Lex(); 2973 return false; 2974 } 2975 2976 /// parseDirectiveAscii: 2977 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 2978 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 2979 auto parseOp = [&]() -> bool { 2980 std::string Data; 2981 if (checkForValidSection() || parseEscapedString(Data)) 2982 return true; 2983 getStreamer().EmitBytes(Data); 2984 if (ZeroTerminated) 2985 getStreamer().EmitBytes(StringRef("\0", 1)); 2986 return false; 2987 }; 2988 2989 if (parseMany(parseOp)) 2990 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 2991 return false; 2992 } 2993 2994 /// parseDirectiveReloc 2995 /// ::= .reloc expression , identifier [ , expression ] 2996 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) { 2997 const MCExpr *Offset; 2998 const MCExpr *Expr = nullptr; 2999 int64_t OffsetValue; 3000 SMLoc OffsetLoc = Lexer.getTok().getLoc(); 3001 3002 if (parseExpression(Offset)) 3003 return true; 3004 3005 if ((Offset->evaluateAsAbsolute(OffsetValue, 3006 getStreamer().getAssemblerPtr()) && 3007 check(OffsetValue < 0, OffsetLoc, "expression is negative")) || 3008 (check(Offset->getKind() != llvm::MCExpr::Constant && 3009 Offset->getKind() != llvm::MCExpr::SymbolRef, 3010 OffsetLoc, "expected non-negative number or a label")) || 3011 (parseToken(AsmToken::Comma, "expected comma") || 3012 check(getTok().isNot(AsmToken::Identifier), "expected relocation name"))) 3013 return true; 3014 3015 SMLoc NameLoc = Lexer.getTok().getLoc(); 3016 StringRef Name = Lexer.getTok().getIdentifier(); 3017 Lex(); 3018 3019 if (Lexer.is(AsmToken::Comma)) { 3020 Lex(); 3021 SMLoc ExprLoc = Lexer.getLoc(); 3022 if (parseExpression(Expr)) 3023 return true; 3024 3025 MCValue Value; 3026 if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr)) 3027 return Error(ExprLoc, "expression must be relocatable"); 3028 } 3029 3030 if (parseToken(AsmToken::EndOfStatement, 3031 "unexpected token in .reloc directive")) 3032 return true; 3033 3034 const MCTargetAsmParser &MCT = getTargetParser(); 3035 const MCSubtargetInfo &STI = MCT.getSTI(); 3036 if (getStreamer().EmitRelocDirective(*Offset, Name, Expr, DirectiveLoc, STI)) 3037 return Error(NameLoc, "unknown relocation name"); 3038 3039 return false; 3040 } 3041 3042 /// parseDirectiveValue 3043 /// ::= (.byte | .short | ... ) [ expression (, expression)* ] 3044 bool AsmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { 3045 auto parseOp = [&]() -> bool { 3046 const MCExpr *Value; 3047 SMLoc ExprLoc = getLexer().getLoc(); 3048 if (checkForValidSection() || parseExpression(Value)) 3049 return true; 3050 // Special case constant expressions to match code generator. 3051 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3052 assert(Size <= 8 && "Invalid size"); 3053 uint64_t IntValue = MCE->getValue(); 3054 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 3055 return Error(ExprLoc, "out of range literal value"); 3056 getStreamer().EmitIntValue(IntValue, Size); 3057 } else 3058 getStreamer().EmitValue(Value, Size, ExprLoc); 3059 return false; 3060 }; 3061 3062 if (parseMany(parseOp)) 3063 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3064 return false; 3065 } 3066 3067 static bool parseHexOcta(AsmParser &Asm, uint64_t &hi, uint64_t &lo) { 3068 if (Asm.getTok().isNot(AsmToken::Integer) && 3069 Asm.getTok().isNot(AsmToken::BigNum)) 3070 return Asm.TokError("unknown token in expression"); 3071 SMLoc ExprLoc = Asm.getTok().getLoc(); 3072 APInt IntValue = Asm.getTok().getAPIntVal(); 3073 Asm.Lex(); 3074 if (!IntValue.isIntN(128)) 3075 return Asm.Error(ExprLoc, "out of range literal value"); 3076 if (!IntValue.isIntN(64)) { 3077 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue(); 3078 lo = IntValue.getLoBits(64).getZExtValue(); 3079 } else { 3080 hi = 0; 3081 lo = IntValue.getZExtValue(); 3082 } 3083 return false; 3084 } 3085 3086 /// ParseDirectiveOctaValue 3087 /// ::= .octa [ hexconstant (, hexconstant)* ] 3088 3089 bool AsmParser::parseDirectiveOctaValue(StringRef IDVal) { 3090 auto parseOp = [&]() -> bool { 3091 if (checkForValidSection()) 3092 return true; 3093 uint64_t hi, lo; 3094 if (parseHexOcta(*this, hi, lo)) 3095 return true; 3096 if (MAI.isLittleEndian()) { 3097 getStreamer().EmitIntValue(lo, 8); 3098 getStreamer().EmitIntValue(hi, 8); 3099 } else { 3100 getStreamer().EmitIntValue(hi, 8); 3101 getStreamer().EmitIntValue(lo, 8); 3102 } 3103 return false; 3104 }; 3105 3106 if (parseMany(parseOp)) 3107 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3108 return false; 3109 } 3110 3111 bool AsmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { 3112 // We don't truly support arithmetic on floating point expressions, so we 3113 // have to manually parse unary prefixes. 3114 bool IsNeg = false; 3115 if (getLexer().is(AsmToken::Minus)) { 3116 Lexer.Lex(); 3117 IsNeg = true; 3118 } else if (getLexer().is(AsmToken::Plus)) 3119 Lexer.Lex(); 3120 3121 if (Lexer.is(AsmToken::Error)) 3122 return TokError(Lexer.getErr()); 3123 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) && 3124 Lexer.isNot(AsmToken::Identifier)) 3125 return TokError("unexpected token in directive"); 3126 3127 // Convert to an APFloat. 3128 APFloat Value(Semantics); 3129 StringRef IDVal = getTok().getString(); 3130 if (getLexer().is(AsmToken::Identifier)) { 3131 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf")) 3132 Value = APFloat::getInf(Semantics); 3133 else if (!IDVal.compare_lower("nan")) 3134 Value = APFloat::getNaN(Semantics, false, ~0); 3135 else 3136 return TokError("invalid floating point literal"); 3137 } else if (errorToBool( 3138 Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) 3139 .takeError())) 3140 return TokError("invalid floating point literal"); 3141 if (IsNeg) 3142 Value.changeSign(); 3143 3144 // Consume the numeric token. 3145 Lex(); 3146 3147 Res = Value.bitcastToAPInt(); 3148 3149 return false; 3150 } 3151 3152 /// parseDirectiveRealValue 3153 /// ::= (.single | .double) [ expression (, expression)* ] 3154 bool AsmParser::parseDirectiveRealValue(StringRef IDVal, 3155 const fltSemantics &Semantics) { 3156 auto parseOp = [&]() -> bool { 3157 APInt AsInt; 3158 if (checkForValidSection() || parseRealValue(Semantics, AsInt)) 3159 return true; 3160 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 3161 AsInt.getBitWidth() / 8); 3162 return false; 3163 }; 3164 3165 if (parseMany(parseOp)) 3166 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3167 return false; 3168 } 3169 3170 /// parseDirectiveZero 3171 /// ::= .zero expression 3172 bool AsmParser::parseDirectiveZero() { 3173 SMLoc NumBytesLoc = Lexer.getLoc(); 3174 const MCExpr *NumBytes; 3175 if (checkForValidSection() || parseExpression(NumBytes)) 3176 return true; 3177 3178 int64_t Val = 0; 3179 if (getLexer().is(AsmToken::Comma)) { 3180 Lex(); 3181 if (parseAbsoluteExpression(Val)) 3182 return true; 3183 } 3184 3185 if (parseToken(AsmToken::EndOfStatement, 3186 "unexpected token in '.zero' directive")) 3187 return true; 3188 getStreamer().emitFill(*NumBytes, Val, NumBytesLoc); 3189 3190 return false; 3191 } 3192 3193 /// parseDirectiveFill 3194 /// ::= .fill expression [ , expression [ , expression ] ] 3195 bool AsmParser::parseDirectiveFill() { 3196 SMLoc NumValuesLoc = Lexer.getLoc(); 3197 const MCExpr *NumValues; 3198 if (checkForValidSection() || parseExpression(NumValues)) 3199 return true; 3200 3201 int64_t FillSize = 1; 3202 int64_t FillExpr = 0; 3203 3204 SMLoc SizeLoc, ExprLoc; 3205 3206 if (parseOptionalToken(AsmToken::Comma)) { 3207 SizeLoc = getTok().getLoc(); 3208 if (parseAbsoluteExpression(FillSize)) 3209 return true; 3210 if (parseOptionalToken(AsmToken::Comma)) { 3211 ExprLoc = getTok().getLoc(); 3212 if (parseAbsoluteExpression(FillExpr)) 3213 return true; 3214 } 3215 } 3216 if (parseToken(AsmToken::EndOfStatement, 3217 "unexpected token in '.fill' directive")) 3218 return true; 3219 3220 if (FillSize < 0) { 3221 Warning(SizeLoc, "'.fill' directive with negative size has no effect"); 3222 return false; 3223 } 3224 if (FillSize > 8) { 3225 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8"); 3226 FillSize = 8; 3227 } 3228 3229 if (!isUInt<32>(FillExpr) && FillSize > 4) 3230 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits"); 3231 3232 getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc); 3233 3234 return false; 3235 } 3236 3237 /// parseDirectiveOrg 3238 /// ::= .org expression [ , expression ] 3239 bool AsmParser::parseDirectiveOrg() { 3240 const MCExpr *Offset; 3241 SMLoc OffsetLoc = Lexer.getLoc(); 3242 if (checkForValidSection() || parseExpression(Offset)) 3243 return true; 3244 3245 // Parse optional fill expression. 3246 int64_t FillExpr = 0; 3247 if (parseOptionalToken(AsmToken::Comma)) 3248 if (parseAbsoluteExpression(FillExpr)) 3249 return addErrorSuffix(" in '.org' directive"); 3250 if (parseToken(AsmToken::EndOfStatement)) 3251 return addErrorSuffix(" in '.org' directive"); 3252 3253 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc); 3254 return false; 3255 } 3256 3257 /// parseDirectiveAlign 3258 /// ::= {.align, ...} expression [ , expression [ , expression ]] 3259 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 3260 SMLoc AlignmentLoc = getLexer().getLoc(); 3261 int64_t Alignment; 3262 SMLoc MaxBytesLoc; 3263 bool HasFillExpr = false; 3264 int64_t FillExpr = 0; 3265 int64_t MaxBytesToFill = 0; 3266 3267 auto parseAlign = [&]() -> bool { 3268 if (parseAbsoluteExpression(Alignment)) 3269 return true; 3270 if (parseOptionalToken(AsmToken::Comma)) { 3271 // The fill expression can be omitted while specifying a maximum number of 3272 // alignment bytes, e.g: 3273 // .align 3,,4 3274 if (getTok().isNot(AsmToken::Comma)) { 3275 HasFillExpr = true; 3276 if (parseAbsoluteExpression(FillExpr)) 3277 return true; 3278 } 3279 if (parseOptionalToken(AsmToken::Comma)) 3280 if (parseTokenLoc(MaxBytesLoc) || 3281 parseAbsoluteExpression(MaxBytesToFill)) 3282 return true; 3283 } 3284 return parseToken(AsmToken::EndOfStatement); 3285 }; 3286 3287 if (checkForValidSection()) 3288 return addErrorSuffix(" in directive"); 3289 // Ignore empty '.p2align' directives for GNU-as compatibility 3290 if (IsPow2 && (ValueSize == 1) && getTok().is(AsmToken::EndOfStatement)) { 3291 Warning(AlignmentLoc, "p2align directive with no operand(s) is ignored"); 3292 return parseToken(AsmToken::EndOfStatement); 3293 } 3294 if (parseAlign()) 3295 return addErrorSuffix(" in directive"); 3296 3297 // Always emit an alignment here even if we thrown an error. 3298 bool ReturnVal = false; 3299 3300 // Compute alignment in bytes. 3301 if (IsPow2) { 3302 // FIXME: Diagnose overflow. 3303 if (Alignment >= 32) { 3304 ReturnVal |= Error(AlignmentLoc, "invalid alignment value"); 3305 Alignment = 31; 3306 } 3307 3308 Alignment = 1ULL << Alignment; 3309 } else { 3310 // Reject alignments that aren't either a power of two or zero, 3311 // for gas compatibility. Alignment of zero is silently rounded 3312 // up to one. 3313 if (Alignment == 0) 3314 Alignment = 1; 3315 if (!isPowerOf2_64(Alignment)) 3316 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2"); 3317 } 3318 3319 // Diagnose non-sensical max bytes to align. 3320 if (MaxBytesLoc.isValid()) { 3321 if (MaxBytesToFill < 1) { 3322 ReturnVal |= Error(MaxBytesLoc, 3323 "alignment directive can never be satisfied in this " 3324 "many bytes, ignoring maximum bytes expression"); 3325 MaxBytesToFill = 0; 3326 } 3327 3328 if (MaxBytesToFill >= Alignment) { 3329 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 3330 "has no effect"); 3331 MaxBytesToFill = 0; 3332 } 3333 } 3334 3335 // Check whether we should use optimal code alignment for this .align 3336 // directive. 3337 const MCSection *Section = getStreamer().getCurrentSectionOnly(); 3338 assert(Section && "must have section to emit alignment"); 3339 bool UseCodeAlign = Section->UseCodeAlign(); 3340 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 3341 ValueSize == 1 && UseCodeAlign) { 3342 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill); 3343 } else { 3344 // FIXME: Target specific behavior about how the "extra" bytes are filled. 3345 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize, 3346 MaxBytesToFill); 3347 } 3348 3349 return ReturnVal; 3350 } 3351 3352 /// parseDirectiveFile 3353 /// ::= .file filename 3354 /// ::= .file number [directory] filename [md5 checksum] [source source-text] 3355 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) { 3356 // FIXME: I'm not sure what this is. 3357 int64_t FileNumber = -1; 3358 if (getLexer().is(AsmToken::Integer)) { 3359 FileNumber = getTok().getIntVal(); 3360 Lex(); 3361 3362 if (FileNumber < 0) 3363 return TokError("negative file number"); 3364 } 3365 3366 std::string Path; 3367 3368 // Usually the directory and filename together, otherwise just the directory. 3369 // Allow the strings to have escaped octal character sequence. 3370 if (check(getTok().isNot(AsmToken::String), 3371 "unexpected token in '.file' directive") || 3372 parseEscapedString(Path)) 3373 return true; 3374 3375 StringRef Directory; 3376 StringRef Filename; 3377 std::string FilenameData; 3378 if (getLexer().is(AsmToken::String)) { 3379 if (check(FileNumber == -1, 3380 "explicit path specified, but no file number") || 3381 parseEscapedString(FilenameData)) 3382 return true; 3383 Filename = FilenameData; 3384 Directory = Path; 3385 } else { 3386 Filename = Path; 3387 } 3388 3389 uint64_t MD5Hi, MD5Lo; 3390 bool HasMD5 = false; 3391 3392 Optional<StringRef> Source; 3393 bool HasSource = false; 3394 std::string SourceString; 3395 3396 while (!parseOptionalToken(AsmToken::EndOfStatement)) { 3397 StringRef Keyword; 3398 if (check(getTok().isNot(AsmToken::Identifier), 3399 "unexpected token in '.file' directive") || 3400 parseIdentifier(Keyword)) 3401 return true; 3402 if (Keyword == "md5") { 3403 HasMD5 = true; 3404 if (check(FileNumber == -1, 3405 "MD5 checksum specified, but no file number") || 3406 parseHexOcta(*this, MD5Hi, MD5Lo)) 3407 return true; 3408 } else if (Keyword == "source") { 3409 HasSource = true; 3410 if (check(FileNumber == -1, 3411 "source specified, but no file number") || 3412 check(getTok().isNot(AsmToken::String), 3413 "unexpected token in '.file' directive") || 3414 parseEscapedString(SourceString)) 3415 return true; 3416 } else { 3417 return TokError("unexpected token in '.file' directive"); 3418 } 3419 } 3420 3421 if (FileNumber == -1) { 3422 // Ignore the directive if there is no number and the target doesn't support 3423 // numberless .file directives. This allows some portability of assembler 3424 // between different object file formats. 3425 if (getContext().getAsmInfo()->hasSingleParameterDotFile()) 3426 getStreamer().EmitFileDirective(Filename); 3427 } else { 3428 // In case there is a -g option as well as debug info from directive .file, 3429 // we turn off the -g option, directly use the existing debug info instead. 3430 // Throw away any implicit file table for the assembler source. 3431 if (Ctx.getGenDwarfForAssembly()) { 3432 Ctx.getMCDwarfLineTable(0).resetFileTable(); 3433 Ctx.setGenDwarfForAssembly(false); 3434 } 3435 3436 Optional<MD5::MD5Result> CKMem; 3437 if (HasMD5) { 3438 MD5::MD5Result Sum; 3439 for (unsigned i = 0; i != 8; ++i) { 3440 Sum.Bytes[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); 3441 Sum.Bytes[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); 3442 } 3443 CKMem = Sum; 3444 } 3445 if (HasSource) { 3446 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size())); 3447 memcpy(SourceBuf, SourceString.data(), SourceString.size()); 3448 Source = StringRef(SourceBuf, SourceString.size()); 3449 } 3450 if (FileNumber == 0) { 3451 if (Ctx.getDwarfVersion() < 5) 3452 return Warning(DirectiveLoc, "file 0 not supported prior to DWARF-5"); 3453 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source); 3454 } else { 3455 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( 3456 FileNumber, Directory, Filename, CKMem, Source); 3457 if (!FileNumOrErr) 3458 return Error(DirectiveLoc, toString(FileNumOrErr.takeError())); 3459 } 3460 // Alert the user if there are some .file directives with MD5 and some not. 3461 // But only do that once. 3462 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) { 3463 ReportedInconsistentMD5 = true; 3464 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums"); 3465 } 3466 } 3467 3468 return false; 3469 } 3470 3471 /// parseDirectiveLine 3472 /// ::= .line [number] 3473 bool AsmParser::parseDirectiveLine() { 3474 int64_t LineNumber; 3475 if (getLexer().is(AsmToken::Integer)) { 3476 if (parseIntToken(LineNumber, "unexpected token in '.line' directive")) 3477 return true; 3478 (void)LineNumber; 3479 // FIXME: Do something with the .line. 3480 } 3481 if (parseToken(AsmToken::EndOfStatement, 3482 "unexpected token in '.line' directive")) 3483 return true; 3484 3485 return false; 3486 } 3487 3488 /// parseDirectiveLoc 3489 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 3490 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 3491 /// The first number is a file number, must have been previously assigned with 3492 /// a .file directive, the second number is the line number and optionally the 3493 /// third number is a column position (zero if not specified). The remaining 3494 /// optional items are .loc sub-directives. 3495 bool AsmParser::parseDirectiveLoc() { 3496 int64_t FileNumber = 0, LineNumber = 0; 3497 SMLoc Loc = getTok().getLoc(); 3498 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") || 3499 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, 3500 "file number less than one in '.loc' directive") || 3501 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc, 3502 "unassigned file number in '.loc' directive")) 3503 return true; 3504 3505 // optional 3506 if (getLexer().is(AsmToken::Integer)) { 3507 LineNumber = getTok().getIntVal(); 3508 if (LineNumber < 0) 3509 return TokError("line number less than zero in '.loc' directive"); 3510 Lex(); 3511 } 3512 3513 int64_t ColumnPos = 0; 3514 if (getLexer().is(AsmToken::Integer)) { 3515 ColumnPos = getTok().getIntVal(); 3516 if (ColumnPos < 0) 3517 return TokError("column position less than zero in '.loc' directive"); 3518 Lex(); 3519 } 3520 3521 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0; 3522 unsigned Isa = 0; 3523 int64_t Discriminator = 0; 3524 3525 auto parseLocOp = [&]() -> bool { 3526 StringRef Name; 3527 SMLoc Loc = getTok().getLoc(); 3528 if (parseIdentifier(Name)) 3529 return TokError("unexpected token in '.loc' directive"); 3530 3531 if (Name == "basic_block") 3532 Flags |= DWARF2_FLAG_BASIC_BLOCK; 3533 else if (Name == "prologue_end") 3534 Flags |= DWARF2_FLAG_PROLOGUE_END; 3535 else if (Name == "epilogue_begin") 3536 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 3537 else if (Name == "is_stmt") { 3538 Loc = getTok().getLoc(); 3539 const MCExpr *Value; 3540 if (parseExpression(Value)) 3541 return true; 3542 // The expression must be the constant 0 or 1. 3543 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3544 int Value = MCE->getValue(); 3545 if (Value == 0) 3546 Flags &= ~DWARF2_FLAG_IS_STMT; 3547 else if (Value == 1) 3548 Flags |= DWARF2_FLAG_IS_STMT; 3549 else 3550 return Error(Loc, "is_stmt value not 0 or 1"); 3551 } else { 3552 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 3553 } 3554 } else if (Name == "isa") { 3555 Loc = getTok().getLoc(); 3556 const MCExpr *Value; 3557 if (parseExpression(Value)) 3558 return true; 3559 // The expression must be a constant greater or equal to 0. 3560 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3561 int Value = MCE->getValue(); 3562 if (Value < 0) 3563 return Error(Loc, "isa number less than zero"); 3564 Isa = Value; 3565 } else { 3566 return Error(Loc, "isa number not a constant value"); 3567 } 3568 } else if (Name == "discriminator") { 3569 if (parseAbsoluteExpression(Discriminator)) 3570 return true; 3571 } else { 3572 return Error(Loc, "unknown sub-directive in '.loc' directive"); 3573 } 3574 return false; 3575 }; 3576 3577 if (parseMany(parseLocOp, false /*hasComma*/)) 3578 return true; 3579 3580 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 3581 Isa, Discriminator, StringRef()); 3582 3583 return false; 3584 } 3585 3586 /// parseDirectiveStabs 3587 /// ::= .stabs string, number, number, number 3588 bool AsmParser::parseDirectiveStabs() { 3589 return TokError("unsupported directive '.stabs'"); 3590 } 3591 3592 /// parseDirectiveCVFile 3593 /// ::= .cv_file number filename [checksum] [checksumkind] 3594 bool AsmParser::parseDirectiveCVFile() { 3595 SMLoc FileNumberLoc = getTok().getLoc(); 3596 int64_t FileNumber; 3597 std::string Filename; 3598 std::string Checksum; 3599 int64_t ChecksumKind = 0; 3600 3601 if (parseIntToken(FileNumber, 3602 "expected file number in '.cv_file' directive") || 3603 check(FileNumber < 1, FileNumberLoc, "file number less than one") || 3604 check(getTok().isNot(AsmToken::String), 3605 "unexpected token in '.cv_file' directive") || 3606 parseEscapedString(Filename)) 3607 return true; 3608 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 3609 if (check(getTok().isNot(AsmToken::String), 3610 "unexpected token in '.cv_file' directive") || 3611 parseEscapedString(Checksum) || 3612 parseIntToken(ChecksumKind, 3613 "expected checksum kind in '.cv_file' directive") || 3614 parseToken(AsmToken::EndOfStatement, 3615 "unexpected token in '.cv_file' directive")) 3616 return true; 3617 } 3618 3619 Checksum = fromHex(Checksum); 3620 void *CKMem = Ctx.allocate(Checksum.size(), 1); 3621 memcpy(CKMem, Checksum.data(), Checksum.size()); 3622 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), 3623 Checksum.size()); 3624 3625 if (!getStreamer().EmitCVFileDirective(FileNumber, Filename, ChecksumAsBytes, 3626 static_cast<uint8_t>(ChecksumKind))) 3627 return Error(FileNumberLoc, "file number already allocated"); 3628 3629 return false; 3630 } 3631 3632 bool AsmParser::parseCVFunctionId(int64_t &FunctionId, 3633 StringRef DirectiveName) { 3634 SMLoc Loc; 3635 return parseTokenLoc(Loc) || 3636 parseIntToken(FunctionId, "expected function id in '" + DirectiveName + 3637 "' directive") || 3638 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc, 3639 "expected function id within range [0, UINT_MAX)"); 3640 } 3641 3642 bool AsmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { 3643 SMLoc Loc; 3644 return parseTokenLoc(Loc) || 3645 parseIntToken(FileNumber, "expected integer in '" + DirectiveName + 3646 "' directive") || 3647 check(FileNumber < 1, Loc, "file number less than one in '" + 3648 DirectiveName + "' directive") || 3649 check(!getCVContext().isValidFileNumber(FileNumber), Loc, 3650 "unassigned file number in '" + DirectiveName + "' directive"); 3651 } 3652 3653 /// parseDirectiveCVFuncId 3654 /// ::= .cv_func_id FunctionId 3655 /// 3656 /// Introduces a function ID that can be used with .cv_loc. 3657 bool AsmParser::parseDirectiveCVFuncId() { 3658 SMLoc FunctionIdLoc = getTok().getLoc(); 3659 int64_t FunctionId; 3660 3661 if (parseCVFunctionId(FunctionId, ".cv_func_id") || 3662 parseToken(AsmToken::EndOfStatement, 3663 "unexpected token in '.cv_func_id' directive")) 3664 return true; 3665 3666 if (!getStreamer().EmitCVFuncIdDirective(FunctionId)) 3667 return Error(FunctionIdLoc, "function id already allocated"); 3668 3669 return false; 3670 } 3671 3672 /// parseDirectiveCVInlineSiteId 3673 /// ::= .cv_inline_site_id FunctionId 3674 /// "within" IAFunc 3675 /// "inlined_at" IAFile IALine [IACol] 3676 /// 3677 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined 3678 /// at" source location information for use in the line table of the caller, 3679 /// whether the caller is a real function or another inlined call site. 3680 bool AsmParser::parseDirectiveCVInlineSiteId() { 3681 SMLoc FunctionIdLoc = getTok().getLoc(); 3682 int64_t FunctionId; 3683 int64_t IAFunc; 3684 int64_t IAFile; 3685 int64_t IALine; 3686 int64_t IACol = 0; 3687 3688 // FunctionId 3689 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id")) 3690 return true; 3691 3692 // "within" 3693 if (check((getLexer().isNot(AsmToken::Identifier) || 3694 getTok().getIdentifier() != "within"), 3695 "expected 'within' identifier in '.cv_inline_site_id' directive")) 3696 return true; 3697 Lex(); 3698 3699 // IAFunc 3700 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id")) 3701 return true; 3702 3703 // "inlined_at" 3704 if (check((getLexer().isNot(AsmToken::Identifier) || 3705 getTok().getIdentifier() != "inlined_at"), 3706 "expected 'inlined_at' identifier in '.cv_inline_site_id' " 3707 "directive") ) 3708 return true; 3709 Lex(); 3710 3711 // IAFile IALine 3712 if (parseCVFileId(IAFile, ".cv_inline_site_id") || 3713 parseIntToken(IALine, "expected line number after 'inlined_at'")) 3714 return true; 3715 3716 // [IACol] 3717 if (getLexer().is(AsmToken::Integer)) { 3718 IACol = getTok().getIntVal(); 3719 Lex(); 3720 } 3721 3722 if (parseToken(AsmToken::EndOfStatement, 3723 "unexpected token in '.cv_inline_site_id' directive")) 3724 return true; 3725 3726 if (!getStreamer().EmitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, 3727 IALine, IACol, FunctionIdLoc)) 3728 return Error(FunctionIdLoc, "function id already allocated"); 3729 3730 return false; 3731 } 3732 3733 /// parseDirectiveCVLoc 3734 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] 3735 /// [is_stmt VALUE] 3736 /// The first number is a file number, must have been previously assigned with 3737 /// a .file directive, the second number is the line number and optionally the 3738 /// third number is a column position (zero if not specified). The remaining 3739 /// optional items are .loc sub-directives. 3740 bool AsmParser::parseDirectiveCVLoc() { 3741 SMLoc DirectiveLoc = getTok().getLoc(); 3742 int64_t FunctionId, FileNumber; 3743 if (parseCVFunctionId(FunctionId, ".cv_loc") || 3744 parseCVFileId(FileNumber, ".cv_loc")) 3745 return true; 3746 3747 int64_t LineNumber = 0; 3748 if (getLexer().is(AsmToken::Integer)) { 3749 LineNumber = getTok().getIntVal(); 3750 if (LineNumber < 0) 3751 return TokError("line number less than zero in '.cv_loc' directive"); 3752 Lex(); 3753 } 3754 3755 int64_t ColumnPos = 0; 3756 if (getLexer().is(AsmToken::Integer)) { 3757 ColumnPos = getTok().getIntVal(); 3758 if (ColumnPos < 0) 3759 return TokError("column position less than zero in '.cv_loc' directive"); 3760 Lex(); 3761 } 3762 3763 bool PrologueEnd = false; 3764 uint64_t IsStmt = 0; 3765 3766 auto parseOp = [&]() -> bool { 3767 StringRef Name; 3768 SMLoc Loc = getTok().getLoc(); 3769 if (parseIdentifier(Name)) 3770 return TokError("unexpected token in '.cv_loc' directive"); 3771 if (Name == "prologue_end") 3772 PrologueEnd = true; 3773 else if (Name == "is_stmt") { 3774 Loc = getTok().getLoc(); 3775 const MCExpr *Value; 3776 if (parseExpression(Value)) 3777 return true; 3778 // The expression must be the constant 0 or 1. 3779 IsStmt = ~0ULL; 3780 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value)) 3781 IsStmt = MCE->getValue(); 3782 3783 if (IsStmt > 1) 3784 return Error(Loc, "is_stmt value not 0 or 1"); 3785 } else { 3786 return Error(Loc, "unknown sub-directive in '.cv_loc' directive"); 3787 } 3788 return false; 3789 }; 3790 3791 if (parseMany(parseOp, false /*hasComma*/)) 3792 return true; 3793 3794 getStreamer().EmitCVLocDirective(FunctionId, FileNumber, LineNumber, 3795 ColumnPos, PrologueEnd, IsStmt, StringRef(), 3796 DirectiveLoc); 3797 return false; 3798 } 3799 3800 /// parseDirectiveCVLinetable 3801 /// ::= .cv_linetable FunctionId, FnStart, FnEnd 3802 bool AsmParser::parseDirectiveCVLinetable() { 3803 int64_t FunctionId; 3804 StringRef FnStartName, FnEndName; 3805 SMLoc Loc = getTok().getLoc(); 3806 if (parseCVFunctionId(FunctionId, ".cv_linetable") || 3807 parseToken(AsmToken::Comma, 3808 "unexpected token in '.cv_linetable' directive") || 3809 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3810 "expected identifier in directive") || 3811 parseToken(AsmToken::Comma, 3812 "unexpected token in '.cv_linetable' directive") || 3813 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3814 "expected identifier in directive")) 3815 return true; 3816 3817 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3818 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3819 3820 getStreamer().EmitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym); 3821 return false; 3822 } 3823 3824 /// parseDirectiveCVInlineLinetable 3825 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd 3826 bool AsmParser::parseDirectiveCVInlineLinetable() { 3827 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; 3828 StringRef FnStartName, FnEndName; 3829 SMLoc Loc = getTok().getLoc(); 3830 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") || 3831 parseTokenLoc(Loc) || 3832 parseIntToken( 3833 SourceFileId, 3834 "expected SourceField in '.cv_inline_linetable' directive") || 3835 check(SourceFileId <= 0, Loc, 3836 "File id less than zero in '.cv_inline_linetable' directive") || 3837 parseTokenLoc(Loc) || 3838 parseIntToken( 3839 SourceLineNum, 3840 "expected SourceLineNum in '.cv_inline_linetable' directive") || 3841 check(SourceLineNum < 0, Loc, 3842 "Line number less than zero in '.cv_inline_linetable' directive") || 3843 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3844 "expected identifier in directive") || 3845 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3846 "expected identifier in directive")) 3847 return true; 3848 3849 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 3850 return true; 3851 3852 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3853 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3854 getStreamer().EmitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, 3855 SourceLineNum, FnStartSym, 3856 FnEndSym); 3857 return false; 3858 } 3859 3860 void AsmParser::initializeCVDefRangeTypeMap() { 3861 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER; 3862 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL; 3863 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER; 3864 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL; 3865 } 3866 3867 /// parseDirectiveCVDefRange 3868 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* 3869 bool AsmParser::parseDirectiveCVDefRange() { 3870 SMLoc Loc; 3871 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; 3872 while (getLexer().is(AsmToken::Identifier)) { 3873 Loc = getLexer().getLoc(); 3874 StringRef GapStartName; 3875 if (parseIdentifier(GapStartName)) 3876 return Error(Loc, "expected identifier in directive"); 3877 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName); 3878 3879 Loc = getLexer().getLoc(); 3880 StringRef GapEndName; 3881 if (parseIdentifier(GapEndName)) 3882 return Error(Loc, "expected identifier in directive"); 3883 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName); 3884 3885 Ranges.push_back({GapStartSym, GapEndSym}); 3886 } 3887 3888 StringRef CVDefRangeTypeStr; 3889 if (parseToken( 3890 AsmToken::Comma, 3891 "expected comma before def_range type in .cv_def_range directive") || 3892 parseIdentifier(CVDefRangeTypeStr)) 3893 return Error(Loc, "expected def_range type in directive"); 3894 3895 StringMap<CVDefRangeType>::const_iterator CVTypeIt = 3896 CVDefRangeTypeMap.find(CVDefRangeTypeStr); 3897 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end()) 3898 ? CVDR_DEFRANGE 3899 : CVTypeIt->getValue(); 3900 switch (CVDRType) { 3901 case CVDR_DEFRANGE_REGISTER: { 3902 int64_t DRRegister; 3903 if (parseToken(AsmToken::Comma, "expected comma before register number in " 3904 ".cv_def_range directive") || 3905 parseAbsoluteExpression(DRRegister)) 3906 return Error(Loc, "expected register number"); 3907 3908 codeview::DefRangeRegisterHeader DRHdr; 3909 DRHdr.Register = DRRegister; 3910 DRHdr.MayHaveNoName = 0; 3911 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3912 break; 3913 } 3914 case CVDR_DEFRANGE_FRAMEPOINTER_REL: { 3915 int64_t DROffset; 3916 if (parseToken(AsmToken::Comma, 3917 "expected comma before offset in .cv_def_range directive") || 3918 parseAbsoluteExpression(DROffset)) 3919 return Error(Loc, "expected offset value"); 3920 3921 codeview::DefRangeFramePointerRelHeader DRHdr; 3922 DRHdr.Offset = DROffset; 3923 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3924 break; 3925 } 3926 case CVDR_DEFRANGE_SUBFIELD_REGISTER: { 3927 int64_t DRRegister; 3928 int64_t DROffsetInParent; 3929 if (parseToken(AsmToken::Comma, "expected comma before register number in " 3930 ".cv_def_range directive") || 3931 parseAbsoluteExpression(DRRegister)) 3932 return Error(Loc, "expected register number"); 3933 if (parseToken(AsmToken::Comma, 3934 "expected comma before offset in .cv_def_range directive") || 3935 parseAbsoluteExpression(DROffsetInParent)) 3936 return Error(Loc, "expected offset value"); 3937 3938 codeview::DefRangeSubfieldRegisterHeader DRHdr; 3939 DRHdr.Register = DRRegister; 3940 DRHdr.MayHaveNoName = 0; 3941 DRHdr.OffsetInParent = DROffsetInParent; 3942 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3943 break; 3944 } 3945 case CVDR_DEFRANGE_REGISTER_REL: { 3946 int64_t DRRegister; 3947 int64_t DRFlags; 3948 int64_t DRBasePointerOffset; 3949 if (parseToken(AsmToken::Comma, "expected comma before register number in " 3950 ".cv_def_range directive") || 3951 parseAbsoluteExpression(DRRegister)) 3952 return Error(Loc, "expected register value"); 3953 if (parseToken( 3954 AsmToken::Comma, 3955 "expected comma before flag value in .cv_def_range directive") || 3956 parseAbsoluteExpression(DRFlags)) 3957 return Error(Loc, "expected flag value"); 3958 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset " 3959 "in .cv_def_range directive") || 3960 parseAbsoluteExpression(DRBasePointerOffset)) 3961 return Error(Loc, "expected base pointer offset value"); 3962 3963 codeview::DefRangeRegisterRelHeader DRHdr; 3964 DRHdr.Register = DRRegister; 3965 DRHdr.Flags = DRFlags; 3966 DRHdr.BasePointerOffset = DRBasePointerOffset; 3967 getStreamer().EmitCVDefRangeDirective(Ranges, DRHdr); 3968 break; 3969 } 3970 default: 3971 return Error(Loc, "unexpected def_range type in .cv_def_range directive"); 3972 } 3973 return true; 3974 } 3975 3976 /// parseDirectiveCVString 3977 /// ::= .cv_stringtable "string" 3978 bool AsmParser::parseDirectiveCVString() { 3979 std::string Data; 3980 if (checkForValidSection() || parseEscapedString(Data)) 3981 return addErrorSuffix(" in '.cv_string' directive"); 3982 3983 // Put the string in the table and emit the offset. 3984 std::pair<StringRef, unsigned> Insertion = 3985 getCVContext().addToStringTable(Data); 3986 getStreamer().EmitIntValue(Insertion.second, 4); 3987 return false; 3988 } 3989 3990 /// parseDirectiveCVStringTable 3991 /// ::= .cv_stringtable 3992 bool AsmParser::parseDirectiveCVStringTable() { 3993 getStreamer().EmitCVStringTableDirective(); 3994 return false; 3995 } 3996 3997 /// parseDirectiveCVFileChecksums 3998 /// ::= .cv_filechecksums 3999 bool AsmParser::parseDirectiveCVFileChecksums() { 4000 getStreamer().EmitCVFileChecksumsDirective(); 4001 return false; 4002 } 4003 4004 /// parseDirectiveCVFileChecksumOffset 4005 /// ::= .cv_filechecksumoffset fileno 4006 bool AsmParser::parseDirectiveCVFileChecksumOffset() { 4007 int64_t FileNo; 4008 if (parseIntToken(FileNo, "expected identifier in directive")) 4009 return true; 4010 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 4011 return true; 4012 getStreamer().EmitCVFileChecksumOffsetDirective(FileNo); 4013 return false; 4014 } 4015 4016 /// parseDirectiveCVFPOData 4017 /// ::= .cv_fpo_data procsym 4018 bool AsmParser::parseDirectiveCVFPOData() { 4019 SMLoc DirLoc = getLexer().getLoc(); 4020 StringRef ProcName; 4021 if (parseIdentifier(ProcName)) 4022 return TokError("expected symbol name"); 4023 if (parseEOL("unexpected tokens")) 4024 return addErrorSuffix(" in '.cv_fpo_data' directive"); 4025 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName); 4026 getStreamer().EmitCVFPOData(ProcSym, DirLoc); 4027 return false; 4028 } 4029 4030 /// parseDirectiveCFISections 4031 /// ::= .cfi_sections section [, section] 4032 bool AsmParser::parseDirectiveCFISections() { 4033 StringRef Name; 4034 bool EH = false; 4035 bool Debug = false; 4036 4037 if (parseIdentifier(Name)) 4038 return TokError("Expected an identifier"); 4039 4040 if (Name == ".eh_frame") 4041 EH = true; 4042 else if (Name == ".debug_frame") 4043 Debug = true; 4044 4045 if (getLexer().is(AsmToken::Comma)) { 4046 Lex(); 4047 4048 if (parseIdentifier(Name)) 4049 return TokError("Expected an identifier"); 4050 4051 if (Name == ".eh_frame") 4052 EH = true; 4053 else if (Name == ".debug_frame") 4054 Debug = true; 4055 } 4056 4057 getStreamer().EmitCFISections(EH, Debug); 4058 return false; 4059 } 4060 4061 /// parseDirectiveCFIStartProc 4062 /// ::= .cfi_startproc [simple] 4063 bool AsmParser::parseDirectiveCFIStartProc() { 4064 StringRef Simple; 4065 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4066 if (check(parseIdentifier(Simple) || Simple != "simple", 4067 "unexpected token") || 4068 parseToken(AsmToken::EndOfStatement)) 4069 return addErrorSuffix(" in '.cfi_startproc' directive"); 4070 } 4071 4072 // TODO(kristina): Deal with a corner case of incorrect diagnostic context 4073 // being produced if this directive is emitted as part of preprocessor macro 4074 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. 4075 // Tools like llvm-mc on the other hand are not affected by it, and report 4076 // correct context information. 4077 getStreamer().EmitCFIStartProc(!Simple.empty(), Lexer.getLoc()); 4078 return false; 4079 } 4080 4081 /// parseDirectiveCFIEndProc 4082 /// ::= .cfi_endproc 4083 bool AsmParser::parseDirectiveCFIEndProc() { 4084 getStreamer().EmitCFIEndProc(); 4085 return false; 4086 } 4087 4088 /// parse register name or number. 4089 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register, 4090 SMLoc DirectiveLoc) { 4091 unsigned RegNo; 4092 4093 if (getLexer().isNot(AsmToken::Integer)) { 4094 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc)) 4095 return true; 4096 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true); 4097 } else 4098 return parseAbsoluteExpression(Register); 4099 4100 return false; 4101 } 4102 4103 /// parseDirectiveCFIDefCfa 4104 /// ::= .cfi_def_cfa register, offset 4105 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { 4106 int64_t Register = 0, Offset = 0; 4107 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4108 parseToken(AsmToken::Comma, "unexpected token in directive") || 4109 parseAbsoluteExpression(Offset)) 4110 return true; 4111 4112 getStreamer().EmitCFIDefCfa(Register, Offset); 4113 return false; 4114 } 4115 4116 /// parseDirectiveCFIDefCfaOffset 4117 /// ::= .cfi_def_cfa_offset offset 4118 bool AsmParser::parseDirectiveCFIDefCfaOffset() { 4119 int64_t Offset = 0; 4120 if (parseAbsoluteExpression(Offset)) 4121 return true; 4122 4123 getStreamer().EmitCFIDefCfaOffset(Offset); 4124 return false; 4125 } 4126 4127 /// parseDirectiveCFIRegister 4128 /// ::= .cfi_register register, register 4129 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { 4130 int64_t Register1 = 0, Register2 = 0; 4131 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || 4132 parseToken(AsmToken::Comma, "unexpected token in directive") || 4133 parseRegisterOrRegisterNumber(Register2, DirectiveLoc)) 4134 return true; 4135 4136 getStreamer().EmitCFIRegister(Register1, Register2); 4137 return false; 4138 } 4139 4140 /// parseDirectiveCFIWindowSave 4141 /// ::= .cfi_window_save 4142 bool AsmParser::parseDirectiveCFIWindowSave() { 4143 getStreamer().EmitCFIWindowSave(); 4144 return false; 4145 } 4146 4147 /// parseDirectiveCFIAdjustCfaOffset 4148 /// ::= .cfi_adjust_cfa_offset adjustment 4149 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() { 4150 int64_t Adjustment = 0; 4151 if (parseAbsoluteExpression(Adjustment)) 4152 return true; 4153 4154 getStreamer().EmitCFIAdjustCfaOffset(Adjustment); 4155 return false; 4156 } 4157 4158 /// parseDirectiveCFIDefCfaRegister 4159 /// ::= .cfi_def_cfa_register register 4160 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { 4161 int64_t Register = 0; 4162 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4163 return true; 4164 4165 getStreamer().EmitCFIDefCfaRegister(Register); 4166 return false; 4167 } 4168 4169 /// parseDirectiveCFIOffset 4170 /// ::= .cfi_offset register, offset 4171 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { 4172 int64_t Register = 0; 4173 int64_t Offset = 0; 4174 4175 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4176 parseToken(AsmToken::Comma, "unexpected token in directive") || 4177 parseAbsoluteExpression(Offset)) 4178 return true; 4179 4180 getStreamer().EmitCFIOffset(Register, Offset); 4181 return false; 4182 } 4183 4184 /// parseDirectiveCFIRelOffset 4185 /// ::= .cfi_rel_offset register, offset 4186 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { 4187 int64_t Register = 0, Offset = 0; 4188 4189 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4190 parseToken(AsmToken::Comma, "unexpected token in directive") || 4191 parseAbsoluteExpression(Offset)) 4192 return true; 4193 4194 getStreamer().EmitCFIRelOffset(Register, Offset); 4195 return false; 4196 } 4197 4198 static bool isValidEncoding(int64_t Encoding) { 4199 if (Encoding & ~0xff) 4200 return false; 4201 4202 if (Encoding == dwarf::DW_EH_PE_omit) 4203 return true; 4204 4205 const unsigned Format = Encoding & 0xf; 4206 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 4207 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 4208 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 4209 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 4210 return false; 4211 4212 const unsigned Application = Encoding & 0x70; 4213 if (Application != dwarf::DW_EH_PE_absptr && 4214 Application != dwarf::DW_EH_PE_pcrel) 4215 return false; 4216 4217 return true; 4218 } 4219 4220 /// parseDirectiveCFIPersonalityOrLsda 4221 /// IsPersonality true for cfi_personality, false for cfi_lsda 4222 /// ::= .cfi_personality encoding, [symbol_name] 4223 /// ::= .cfi_lsda encoding, [symbol_name] 4224 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { 4225 int64_t Encoding = 0; 4226 if (parseAbsoluteExpression(Encoding)) 4227 return true; 4228 if (Encoding == dwarf::DW_EH_PE_omit) 4229 return false; 4230 4231 StringRef Name; 4232 if (check(!isValidEncoding(Encoding), "unsupported encoding.") || 4233 parseToken(AsmToken::Comma, "unexpected token in directive") || 4234 check(parseIdentifier(Name), "expected identifier in directive")) 4235 return true; 4236 4237 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4238 4239 if (IsPersonality) 4240 getStreamer().EmitCFIPersonality(Sym, Encoding); 4241 else 4242 getStreamer().EmitCFILsda(Sym, Encoding); 4243 return false; 4244 } 4245 4246 /// parseDirectiveCFIRememberState 4247 /// ::= .cfi_remember_state 4248 bool AsmParser::parseDirectiveCFIRememberState() { 4249 getStreamer().EmitCFIRememberState(); 4250 return false; 4251 } 4252 4253 /// parseDirectiveCFIRestoreState 4254 /// ::= .cfi_remember_state 4255 bool AsmParser::parseDirectiveCFIRestoreState() { 4256 getStreamer().EmitCFIRestoreState(); 4257 return false; 4258 } 4259 4260 /// parseDirectiveCFISameValue 4261 /// ::= .cfi_same_value register 4262 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { 4263 int64_t Register = 0; 4264 4265 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4266 return true; 4267 4268 getStreamer().EmitCFISameValue(Register); 4269 return false; 4270 } 4271 4272 /// parseDirectiveCFIRestore 4273 /// ::= .cfi_restore register 4274 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { 4275 int64_t Register = 0; 4276 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4277 return true; 4278 4279 getStreamer().EmitCFIRestore(Register); 4280 return false; 4281 } 4282 4283 /// parseDirectiveCFIEscape 4284 /// ::= .cfi_escape expression[,...] 4285 bool AsmParser::parseDirectiveCFIEscape() { 4286 std::string Values; 4287 int64_t CurrValue; 4288 if (parseAbsoluteExpression(CurrValue)) 4289 return true; 4290 4291 Values.push_back((uint8_t)CurrValue); 4292 4293 while (getLexer().is(AsmToken::Comma)) { 4294 Lex(); 4295 4296 if (parseAbsoluteExpression(CurrValue)) 4297 return true; 4298 4299 Values.push_back((uint8_t)CurrValue); 4300 } 4301 4302 getStreamer().EmitCFIEscape(Values); 4303 return false; 4304 } 4305 4306 /// parseDirectiveCFIReturnColumn 4307 /// ::= .cfi_return_column register 4308 bool AsmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { 4309 int64_t Register = 0; 4310 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4311 return true; 4312 getStreamer().EmitCFIReturnColumn(Register); 4313 return false; 4314 } 4315 4316 /// parseDirectiveCFISignalFrame 4317 /// ::= .cfi_signal_frame 4318 bool AsmParser::parseDirectiveCFISignalFrame() { 4319 if (parseToken(AsmToken::EndOfStatement, 4320 "unexpected token in '.cfi_signal_frame'")) 4321 return true; 4322 4323 getStreamer().EmitCFISignalFrame(); 4324 return false; 4325 } 4326 4327 /// parseDirectiveCFIUndefined 4328 /// ::= .cfi_undefined register 4329 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { 4330 int64_t Register = 0; 4331 4332 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4333 return true; 4334 4335 getStreamer().EmitCFIUndefined(Register); 4336 return false; 4337 } 4338 4339 /// parseDirectiveAltmacro 4340 /// ::= .altmacro 4341 /// ::= .noaltmacro 4342 bool AsmParser::parseDirectiveAltmacro(StringRef Directive) { 4343 if (getLexer().isNot(AsmToken::EndOfStatement)) 4344 return TokError("unexpected token in '" + Directive + "' directive"); 4345 AltMacroMode = (Directive == ".altmacro"); 4346 return false; 4347 } 4348 4349 /// parseDirectiveMacrosOnOff 4350 /// ::= .macros_on 4351 /// ::= .macros_off 4352 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) { 4353 if (parseToken(AsmToken::EndOfStatement, 4354 "unexpected token in '" + Directive + "' directive")) 4355 return true; 4356 4357 setMacrosEnabled(Directive == ".macros_on"); 4358 return false; 4359 } 4360 4361 /// parseDirectiveMacro 4362 /// ::= .macro name[,] [parameters] 4363 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) { 4364 StringRef Name; 4365 if (parseIdentifier(Name)) 4366 return TokError("expected identifier in '.macro' directive"); 4367 4368 if (getLexer().is(AsmToken::Comma)) 4369 Lex(); 4370 4371 MCAsmMacroParameters Parameters; 4372 while (getLexer().isNot(AsmToken::EndOfStatement)) { 4373 4374 if (!Parameters.empty() && Parameters.back().Vararg) 4375 return Error(Lexer.getLoc(), 4376 "Vararg parameter '" + Parameters.back().Name + 4377 "' should be last one in the list of parameters."); 4378 4379 MCAsmMacroParameter Parameter; 4380 if (parseIdentifier(Parameter.Name)) 4381 return TokError("expected identifier in '.macro' directive"); 4382 4383 // Emit an error if two (or more) named parameters share the same name 4384 for (const MCAsmMacroParameter& CurrParam : Parameters) 4385 if (CurrParam.Name.equals(Parameter.Name)) 4386 return TokError("macro '" + Name + "' has multiple parameters" 4387 " named '" + Parameter.Name + "'"); 4388 4389 if (Lexer.is(AsmToken::Colon)) { 4390 Lex(); // consume ':' 4391 4392 SMLoc QualLoc; 4393 StringRef Qualifier; 4394 4395 QualLoc = Lexer.getLoc(); 4396 if (parseIdentifier(Qualifier)) 4397 return Error(QualLoc, "missing parameter qualifier for " 4398 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4399 4400 if (Qualifier == "req") 4401 Parameter.Required = true; 4402 else if (Qualifier == "vararg") 4403 Parameter.Vararg = true; 4404 else 4405 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier " 4406 "for '" + Parameter.Name + "' in macro '" + Name + "'"); 4407 } 4408 4409 if (getLexer().is(AsmToken::Equal)) { 4410 Lex(); 4411 4412 SMLoc ParamLoc; 4413 4414 ParamLoc = Lexer.getLoc(); 4415 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false )) 4416 return true; 4417 4418 if (Parameter.Required) 4419 Warning(ParamLoc, "pointless default value for required parameter " 4420 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4421 } 4422 4423 Parameters.push_back(std::move(Parameter)); 4424 4425 if (getLexer().is(AsmToken::Comma)) 4426 Lex(); 4427 } 4428 4429 // Eat just the end of statement. 4430 Lexer.Lex(); 4431 4432 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors 4433 AsmToken EndToken, StartToken = getTok(); 4434 unsigned MacroDepth = 0; 4435 // Lex the macro definition. 4436 while (true) { 4437 // Ignore Lexing errors in macros. 4438 while (Lexer.is(AsmToken::Error)) { 4439 Lexer.Lex(); 4440 } 4441 4442 // Check whether we have reached the end of the file. 4443 if (getLexer().is(AsmToken::Eof)) 4444 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 4445 4446 // Otherwise, check whether we have reach the .endmacro. 4447 if (getLexer().is(AsmToken::Identifier)) { 4448 if (getTok().getIdentifier() == ".endm" || 4449 getTok().getIdentifier() == ".endmacro") { 4450 if (MacroDepth == 0) { // Outermost macro. 4451 EndToken = getTok(); 4452 Lexer.Lex(); 4453 if (getLexer().isNot(AsmToken::EndOfStatement)) 4454 return TokError("unexpected token in '" + EndToken.getIdentifier() + 4455 "' directive"); 4456 break; 4457 } else { 4458 // Otherwise we just found the end of an inner macro. 4459 --MacroDepth; 4460 } 4461 } else if (getTok().getIdentifier() == ".macro") { 4462 // We allow nested macros. Those aren't instantiated until the outermost 4463 // macro is expanded so just ignore them for now. 4464 ++MacroDepth; 4465 } 4466 } 4467 4468 // Otherwise, scan til the end of the statement. 4469 eatToEndOfStatement(); 4470 } 4471 4472 if (getContext().lookupMacro(Name)) { 4473 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 4474 } 4475 4476 const char *BodyStart = StartToken.getLoc().getPointer(); 4477 const char *BodyEnd = EndToken.getLoc().getPointer(); 4478 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 4479 checkForBadMacro(DirectiveLoc, Name, Body, Parameters); 4480 MCAsmMacro Macro(Name, Body, std::move(Parameters)); 4481 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n"; 4482 Macro.dump()); 4483 getContext().defineMacro(Name, std::move(Macro)); 4484 return false; 4485 } 4486 4487 /// checkForBadMacro 4488 /// 4489 /// With the support added for named parameters there may be code out there that 4490 /// is transitioning from positional parameters. In versions of gas that did 4491 /// not support named parameters they would be ignored on the macro definition. 4492 /// But to support both styles of parameters this is not possible so if a macro 4493 /// definition has named parameters but does not use them and has what appears 4494 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a 4495 /// warning that the positional parameter found in body which have no effect. 4496 /// Hoping the developer will either remove the named parameters from the macro 4497 /// definition so the positional parameters get used if that was what was 4498 /// intended or change the macro to use the named parameters. It is possible 4499 /// this warning will trigger when the none of the named parameters are used 4500 /// and the strings like $1 are infact to simply to be passed trough unchanged. 4501 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, 4502 StringRef Body, 4503 ArrayRef<MCAsmMacroParameter> Parameters) { 4504 // If this macro is not defined with named parameters the warning we are 4505 // checking for here doesn't apply. 4506 unsigned NParameters = Parameters.size(); 4507 if (NParameters == 0) 4508 return; 4509 4510 bool NamedParametersFound = false; 4511 bool PositionalParametersFound = false; 4512 4513 // Look at the body of the macro for use of both the named parameters and what 4514 // are likely to be positional parameters. This is what expandMacro() is 4515 // doing when it finds the parameters in the body. 4516 while (!Body.empty()) { 4517 // Scan for the next possible parameter. 4518 std::size_t End = Body.size(), Pos = 0; 4519 for (; Pos != End; ++Pos) { 4520 // Check for a substitution or escape. 4521 // This macro is defined with parameters, look for \foo, \bar, etc. 4522 if (Body[Pos] == '\\' && Pos + 1 != End) 4523 break; 4524 4525 // This macro should have parameters, but look for $0, $1, ..., $n too. 4526 if (Body[Pos] != '$' || Pos + 1 == End) 4527 continue; 4528 char Next = Body[Pos + 1]; 4529 if (Next == '$' || Next == 'n' || 4530 isdigit(static_cast<unsigned char>(Next))) 4531 break; 4532 } 4533 4534 // Check if we reached the end. 4535 if (Pos == End) 4536 break; 4537 4538 if (Body[Pos] == '$') { 4539 switch (Body[Pos + 1]) { 4540 // $$ => $ 4541 case '$': 4542 break; 4543 4544 // $n => number of arguments 4545 case 'n': 4546 PositionalParametersFound = true; 4547 break; 4548 4549 // $[0-9] => argument 4550 default: { 4551 PositionalParametersFound = true; 4552 break; 4553 } 4554 } 4555 Pos += 2; 4556 } else { 4557 unsigned I = Pos + 1; 4558 while (isIdentifierChar(Body[I]) && I + 1 != End) 4559 ++I; 4560 4561 const char *Begin = Body.data() + Pos + 1; 4562 StringRef Argument(Begin, I - (Pos + 1)); 4563 unsigned Index = 0; 4564 for (; Index < NParameters; ++Index) 4565 if (Parameters[Index].Name == Argument) 4566 break; 4567 4568 if (Index == NParameters) { 4569 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 4570 Pos += 3; 4571 else { 4572 Pos = I; 4573 } 4574 } else { 4575 NamedParametersFound = true; 4576 Pos += 1 + Argument.size(); 4577 } 4578 } 4579 // Update the scan point. 4580 Body = Body.substr(Pos); 4581 } 4582 4583 if (!NamedParametersFound && PositionalParametersFound) 4584 Warning(DirectiveLoc, "macro defined with named parameters which are not " 4585 "used in macro body, possible positional parameter " 4586 "found in body which will have no effect"); 4587 } 4588 4589 /// parseDirectiveExitMacro 4590 /// ::= .exitm 4591 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) { 4592 if (parseToken(AsmToken::EndOfStatement, 4593 "unexpected token in '" + Directive + "' directive")) 4594 return true; 4595 4596 if (!isInsideMacroInstantiation()) 4597 return TokError("unexpected '" + Directive + "' in file, " 4598 "no current macro definition"); 4599 4600 // Exit all conditionals that are active in the current macro. 4601 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { 4602 TheCondState = TheCondStack.back(); 4603 TheCondStack.pop_back(); 4604 } 4605 4606 handleMacroExit(); 4607 return false; 4608 } 4609 4610 /// parseDirectiveEndMacro 4611 /// ::= .endm 4612 /// ::= .endmacro 4613 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) { 4614 if (getLexer().isNot(AsmToken::EndOfStatement)) 4615 return TokError("unexpected token in '" + Directive + "' directive"); 4616 4617 // If we are inside a macro instantiation, terminate the current 4618 // instantiation. 4619 if (isInsideMacroInstantiation()) { 4620 handleMacroExit(); 4621 return false; 4622 } 4623 4624 // Otherwise, this .endmacro is a stray entry in the file; well formed 4625 // .endmacro directives are handled during the macro definition parsing. 4626 return TokError("unexpected '" + Directive + "' in file, " 4627 "no current macro definition"); 4628 } 4629 4630 /// parseDirectivePurgeMacro 4631 /// ::= .purgem 4632 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { 4633 StringRef Name; 4634 SMLoc Loc; 4635 if (parseTokenLoc(Loc) || 4636 check(parseIdentifier(Name), Loc, 4637 "expected identifier in '.purgem' directive") || 4638 parseToken(AsmToken::EndOfStatement, 4639 "unexpected token in '.purgem' directive")) 4640 return true; 4641 4642 if (!getContext().lookupMacro(Name)) 4643 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 4644 4645 getContext().undefineMacro(Name); 4646 DEBUG_WITH_TYPE("asm-macros", dbgs() 4647 << "Un-defining macro: " << Name << "\n"); 4648 return false; 4649 } 4650 4651 /// parseDirectiveBundleAlignMode 4652 /// ::= {.bundle_align_mode} expression 4653 bool AsmParser::parseDirectiveBundleAlignMode() { 4654 // Expect a single argument: an expression that evaluates to a constant 4655 // in the inclusive range 0-30. 4656 SMLoc ExprLoc = getLexer().getLoc(); 4657 int64_t AlignSizePow2; 4658 if (checkForValidSection() || parseAbsoluteExpression(AlignSizePow2) || 4659 parseToken(AsmToken::EndOfStatement, "unexpected token after expression " 4660 "in '.bundle_align_mode' " 4661 "directive") || 4662 check(AlignSizePow2 < 0 || AlignSizePow2 > 30, ExprLoc, 4663 "invalid bundle alignment size (expected between 0 and 30)")) 4664 return true; 4665 4666 // Because of AlignSizePow2's verified range we can safely truncate it to 4667 // unsigned. 4668 getStreamer().EmitBundleAlignMode(static_cast<unsigned>(AlignSizePow2)); 4669 return false; 4670 } 4671 4672 /// parseDirectiveBundleLock 4673 /// ::= {.bundle_lock} [align_to_end] 4674 bool AsmParser::parseDirectiveBundleLock() { 4675 if (checkForValidSection()) 4676 return true; 4677 bool AlignToEnd = false; 4678 4679 StringRef Option; 4680 SMLoc Loc = getTok().getLoc(); 4681 const char *kInvalidOptionError = 4682 "invalid option for '.bundle_lock' directive"; 4683 4684 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4685 if (check(parseIdentifier(Option), Loc, kInvalidOptionError) || 4686 check(Option != "align_to_end", Loc, kInvalidOptionError) || 4687 parseToken(AsmToken::EndOfStatement, 4688 "unexpected token after '.bundle_lock' directive option")) 4689 return true; 4690 AlignToEnd = true; 4691 } 4692 4693 getStreamer().EmitBundleLock(AlignToEnd); 4694 return false; 4695 } 4696 4697 /// parseDirectiveBundleLock 4698 /// ::= {.bundle_lock} 4699 bool AsmParser::parseDirectiveBundleUnlock() { 4700 if (checkForValidSection() || 4701 parseToken(AsmToken::EndOfStatement, 4702 "unexpected token in '.bundle_unlock' directive")) 4703 return true; 4704 4705 getStreamer().EmitBundleUnlock(); 4706 return false; 4707 } 4708 4709 /// parseDirectiveSpace 4710 /// ::= (.skip | .space) expression [ , expression ] 4711 bool AsmParser::parseDirectiveSpace(StringRef IDVal) { 4712 SMLoc NumBytesLoc = Lexer.getLoc(); 4713 const MCExpr *NumBytes; 4714 if (checkForValidSection() || parseExpression(NumBytes)) 4715 return true; 4716 4717 int64_t FillExpr = 0; 4718 if (parseOptionalToken(AsmToken::Comma)) 4719 if (parseAbsoluteExpression(FillExpr)) 4720 return addErrorSuffix("in '" + Twine(IDVal) + "' directive"); 4721 if (parseToken(AsmToken::EndOfStatement)) 4722 return addErrorSuffix("in '" + Twine(IDVal) + "' directive"); 4723 4724 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 4725 getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc); 4726 4727 return false; 4728 } 4729 4730 /// parseDirectiveDCB 4731 /// ::= .dcb.{b, l, w} expression, expression 4732 bool AsmParser::parseDirectiveDCB(StringRef IDVal, unsigned Size) { 4733 SMLoc NumValuesLoc = Lexer.getLoc(); 4734 int64_t NumValues; 4735 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4736 return true; 4737 4738 if (NumValues < 0) { 4739 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4740 return false; 4741 } 4742 4743 if (parseToken(AsmToken::Comma, 4744 "unexpected token in '" + Twine(IDVal) + "' directive")) 4745 return true; 4746 4747 const MCExpr *Value; 4748 SMLoc ExprLoc = getLexer().getLoc(); 4749 if (parseExpression(Value)) 4750 return true; 4751 4752 // Special case constant expressions to match code generator. 4753 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4754 assert(Size <= 8 && "Invalid size"); 4755 uint64_t IntValue = MCE->getValue(); 4756 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 4757 return Error(ExprLoc, "literal value out of range for directive"); 4758 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4759 getStreamer().EmitIntValue(IntValue, Size); 4760 } else { 4761 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4762 getStreamer().EmitValue(Value, Size, ExprLoc); 4763 } 4764 4765 if (parseToken(AsmToken::EndOfStatement, 4766 "unexpected token in '" + Twine(IDVal) + "' directive")) 4767 return true; 4768 4769 return false; 4770 } 4771 4772 /// parseDirectiveRealDCB 4773 /// ::= .dcb.{d, s} expression, expression 4774 bool AsmParser::parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &Semantics) { 4775 SMLoc NumValuesLoc = Lexer.getLoc(); 4776 int64_t NumValues; 4777 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4778 return true; 4779 4780 if (NumValues < 0) { 4781 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4782 return false; 4783 } 4784 4785 if (parseToken(AsmToken::Comma, 4786 "unexpected token in '" + Twine(IDVal) + "' directive")) 4787 return true; 4788 4789 APInt AsInt; 4790 if (parseRealValue(Semantics, AsInt)) 4791 return true; 4792 4793 if (parseToken(AsmToken::EndOfStatement, 4794 "unexpected token in '" + Twine(IDVal) + "' directive")) 4795 return true; 4796 4797 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4798 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 4799 AsInt.getBitWidth() / 8); 4800 4801 return false; 4802 } 4803 4804 /// parseDirectiveDS 4805 /// ::= .ds.{b, d, l, p, s, w, x} expression 4806 bool AsmParser::parseDirectiveDS(StringRef IDVal, unsigned Size) { 4807 SMLoc NumValuesLoc = Lexer.getLoc(); 4808 int64_t NumValues; 4809 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4810 return true; 4811 4812 if (NumValues < 0) { 4813 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4814 return false; 4815 } 4816 4817 if (parseToken(AsmToken::EndOfStatement, 4818 "unexpected token in '" + Twine(IDVal) + "' directive")) 4819 return true; 4820 4821 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4822 getStreamer().emitFill(Size, 0); 4823 4824 return false; 4825 } 4826 4827 /// parseDirectiveLEB128 4828 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ] 4829 bool AsmParser::parseDirectiveLEB128(bool Signed) { 4830 if (checkForValidSection()) 4831 return true; 4832 4833 auto parseOp = [&]() -> bool { 4834 const MCExpr *Value; 4835 if (parseExpression(Value)) 4836 return true; 4837 if (Signed) 4838 getStreamer().EmitSLEB128Value(Value); 4839 else 4840 getStreamer().EmitULEB128Value(Value); 4841 return false; 4842 }; 4843 4844 if (parseMany(parseOp)) 4845 return addErrorSuffix(" in directive"); 4846 4847 return false; 4848 } 4849 4850 /// parseDirectiveSymbolAttribute 4851 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 4852 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 4853 auto parseOp = [&]() -> bool { 4854 StringRef Name; 4855 SMLoc Loc = getTok().getLoc(); 4856 if (parseIdentifier(Name)) 4857 return Error(Loc, "expected identifier"); 4858 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4859 4860 // Assembler local symbols don't make any sense here. Complain loudly. 4861 if (Sym->isTemporary()) 4862 return Error(Loc, "non-local symbol required"); 4863 4864 if (!getStreamer().EmitSymbolAttribute(Sym, Attr)) 4865 return Error(Loc, "unable to emit symbol attribute"); 4866 return false; 4867 }; 4868 4869 if (parseMany(parseOp)) 4870 return addErrorSuffix(" in directive"); 4871 return false; 4872 } 4873 4874 /// parseDirectiveComm 4875 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 4876 bool AsmParser::parseDirectiveComm(bool IsLocal) { 4877 if (checkForValidSection()) 4878 return true; 4879 4880 SMLoc IDLoc = getLexer().getLoc(); 4881 StringRef Name; 4882 if (parseIdentifier(Name)) 4883 return TokError("expected identifier in directive"); 4884 4885 // Handle the identifier as the key symbol. 4886 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4887 4888 if (getLexer().isNot(AsmToken::Comma)) 4889 return TokError("unexpected token in directive"); 4890 Lex(); 4891 4892 int64_t Size; 4893 SMLoc SizeLoc = getLexer().getLoc(); 4894 if (parseAbsoluteExpression(Size)) 4895 return true; 4896 4897 int64_t Pow2Alignment = 0; 4898 SMLoc Pow2AlignmentLoc; 4899 if (getLexer().is(AsmToken::Comma)) { 4900 Lex(); 4901 Pow2AlignmentLoc = getLexer().getLoc(); 4902 if (parseAbsoluteExpression(Pow2Alignment)) 4903 return true; 4904 4905 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); 4906 if (IsLocal && LCOMM == LCOMM::NoAlignment) 4907 return Error(Pow2AlignmentLoc, "alignment not supported on this target"); 4908 4909 // If this target takes alignments in bytes (not log) validate and convert. 4910 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || 4911 (IsLocal && LCOMM == LCOMM::ByteAlignment)) { 4912 if (!isPowerOf2_64(Pow2Alignment)) 4913 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 4914 Pow2Alignment = Log2_64(Pow2Alignment); 4915 } 4916 } 4917 4918 if (parseToken(AsmToken::EndOfStatement, 4919 "unexpected token in '.comm' or '.lcomm' directive")) 4920 return true; 4921 4922 // NOTE: a size of zero for a .comm should create a undefined symbol 4923 // but a size of .lcomm creates a bss symbol of size zero. 4924 if (Size < 0) 4925 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 4926 "be less than zero"); 4927 4928 // NOTE: The alignment in the directive is a power of 2 value, the assembler 4929 // may internally end up wanting an alignment in bytes. 4930 // FIXME: Diagnose overflow. 4931 if (Pow2Alignment < 0) 4932 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 4933 "alignment, can't be less than zero"); 4934 4935 Sym->redefineIfPossible(); 4936 if (!Sym->isUndefined()) 4937 return Error(IDLoc, "invalid symbol redefinition"); 4938 4939 // Create the Symbol as a common or local common with Size and Pow2Alignment 4940 if (IsLocal) { 4941 getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment); 4942 return false; 4943 } 4944 4945 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 4946 return false; 4947 } 4948 4949 /// parseDirectiveAbort 4950 /// ::= .abort [... message ...] 4951 bool AsmParser::parseDirectiveAbort() { 4952 // FIXME: Use loc from directive. 4953 SMLoc Loc = getLexer().getLoc(); 4954 4955 StringRef Str = parseStringToEndOfStatement(); 4956 if (parseToken(AsmToken::EndOfStatement, 4957 "unexpected token in '.abort' directive")) 4958 return true; 4959 4960 if (Str.empty()) 4961 return Error(Loc, ".abort detected. Assembly stopping."); 4962 else 4963 return Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 4964 // FIXME: Actually abort assembly here. 4965 4966 return false; 4967 } 4968 4969 /// parseDirectiveInclude 4970 /// ::= .include "filename" 4971 bool AsmParser::parseDirectiveInclude() { 4972 // Allow the strings to have escaped octal character sequence. 4973 std::string Filename; 4974 SMLoc IncludeLoc = getTok().getLoc(); 4975 4976 if (check(getTok().isNot(AsmToken::String), 4977 "expected string in '.include' directive") || 4978 parseEscapedString(Filename) || 4979 check(getTok().isNot(AsmToken::EndOfStatement), 4980 "unexpected token in '.include' directive") || 4981 // Attempt to switch the lexer to the included file before consuming the 4982 // end of statement to avoid losing it when we switch. 4983 check(enterIncludeFile(Filename), IncludeLoc, 4984 "Could not find include file '" + Filename + "'")) 4985 return true; 4986 4987 return false; 4988 } 4989 4990 /// parseDirectiveIncbin 4991 /// ::= .incbin "filename" [ , skip [ , count ] ] 4992 bool AsmParser::parseDirectiveIncbin() { 4993 // Allow the strings to have escaped octal character sequence. 4994 std::string Filename; 4995 SMLoc IncbinLoc = getTok().getLoc(); 4996 if (check(getTok().isNot(AsmToken::String), 4997 "expected string in '.incbin' directive") || 4998 parseEscapedString(Filename)) 4999 return true; 5000 5001 int64_t Skip = 0; 5002 const MCExpr *Count = nullptr; 5003 SMLoc SkipLoc, CountLoc; 5004 if (parseOptionalToken(AsmToken::Comma)) { 5005 // The skip expression can be omitted while specifying the count, e.g: 5006 // .incbin "filename",,4 5007 if (getTok().isNot(AsmToken::Comma)) { 5008 if (parseTokenLoc(SkipLoc) || parseAbsoluteExpression(Skip)) 5009 return true; 5010 } 5011 if (parseOptionalToken(AsmToken::Comma)) { 5012 CountLoc = getTok().getLoc(); 5013 if (parseExpression(Count)) 5014 return true; 5015 } 5016 } 5017 5018 if (parseToken(AsmToken::EndOfStatement, 5019 "unexpected token in '.incbin' directive")) 5020 return true; 5021 5022 if (check(Skip < 0, SkipLoc, "skip is negative")) 5023 return true; 5024 5025 // Attempt to process the included file. 5026 if (processIncbinFile(Filename, Skip, Count, CountLoc)) 5027 return Error(IncbinLoc, "Could not find incbin file '" + Filename + "'"); 5028 return false; 5029 } 5030 5031 /// parseDirectiveIf 5032 /// ::= .if{,eq,ge,gt,le,lt,ne} expression 5033 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { 5034 TheCondStack.push_back(TheCondState); 5035 TheCondState.TheCond = AsmCond::IfCond; 5036 if (TheCondState.Ignore) { 5037 eatToEndOfStatement(); 5038 } else { 5039 int64_t ExprValue; 5040 if (parseAbsoluteExpression(ExprValue) || 5041 parseToken(AsmToken::EndOfStatement, 5042 "unexpected token in '.if' directive")) 5043 return true; 5044 5045 switch (DirKind) { 5046 default: 5047 llvm_unreachable("unsupported directive"); 5048 case DK_IF: 5049 case DK_IFNE: 5050 break; 5051 case DK_IFEQ: 5052 ExprValue = ExprValue == 0; 5053 break; 5054 case DK_IFGE: 5055 ExprValue = ExprValue >= 0; 5056 break; 5057 case DK_IFGT: 5058 ExprValue = ExprValue > 0; 5059 break; 5060 case DK_IFLE: 5061 ExprValue = ExprValue <= 0; 5062 break; 5063 case DK_IFLT: 5064 ExprValue = ExprValue < 0; 5065 break; 5066 } 5067 5068 TheCondState.CondMet = ExprValue; 5069 TheCondState.Ignore = !TheCondState.CondMet; 5070 } 5071 5072 return false; 5073 } 5074 5075 /// parseDirectiveIfb 5076 /// ::= .ifb string 5077 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5078 TheCondStack.push_back(TheCondState); 5079 TheCondState.TheCond = AsmCond::IfCond; 5080 5081 if (TheCondState.Ignore) { 5082 eatToEndOfStatement(); 5083 } else { 5084 StringRef Str = parseStringToEndOfStatement(); 5085 5086 if (parseToken(AsmToken::EndOfStatement, 5087 "unexpected token in '.ifb' directive")) 5088 return true; 5089 5090 TheCondState.CondMet = ExpectBlank == Str.empty(); 5091 TheCondState.Ignore = !TheCondState.CondMet; 5092 } 5093 5094 return false; 5095 } 5096 5097 /// parseDirectiveIfc 5098 /// ::= .ifc string1, string2 5099 /// ::= .ifnc string1, string2 5100 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) { 5101 TheCondStack.push_back(TheCondState); 5102 TheCondState.TheCond = AsmCond::IfCond; 5103 5104 if (TheCondState.Ignore) { 5105 eatToEndOfStatement(); 5106 } else { 5107 StringRef Str1 = parseStringToComma(); 5108 5109 if (parseToken(AsmToken::Comma, "unexpected token in '.ifc' directive")) 5110 return true; 5111 5112 StringRef Str2 = parseStringToEndOfStatement(); 5113 5114 if (parseToken(AsmToken::EndOfStatement, 5115 "unexpected token in '.ifc' directive")) 5116 return true; 5117 5118 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim()); 5119 TheCondState.Ignore = !TheCondState.CondMet; 5120 } 5121 5122 return false; 5123 } 5124 5125 /// parseDirectiveIfeqs 5126 /// ::= .ifeqs string1, string2 5127 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) { 5128 if (Lexer.isNot(AsmToken::String)) { 5129 if (ExpectEqual) 5130 return TokError("expected string parameter for '.ifeqs' directive"); 5131 return TokError("expected string parameter for '.ifnes' directive"); 5132 } 5133 5134 StringRef String1 = getTok().getStringContents(); 5135 Lex(); 5136 5137 if (Lexer.isNot(AsmToken::Comma)) { 5138 if (ExpectEqual) 5139 return TokError( 5140 "expected comma after first string for '.ifeqs' directive"); 5141 return TokError("expected comma after first string for '.ifnes' directive"); 5142 } 5143 5144 Lex(); 5145 5146 if (Lexer.isNot(AsmToken::String)) { 5147 if (ExpectEqual) 5148 return TokError("expected string parameter for '.ifeqs' directive"); 5149 return TokError("expected string parameter for '.ifnes' directive"); 5150 } 5151 5152 StringRef String2 = getTok().getStringContents(); 5153 Lex(); 5154 5155 TheCondStack.push_back(TheCondState); 5156 TheCondState.TheCond = AsmCond::IfCond; 5157 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5158 TheCondState.Ignore = !TheCondState.CondMet; 5159 5160 return false; 5161 } 5162 5163 /// parseDirectiveIfdef 5164 /// ::= .ifdef symbol 5165 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 5166 StringRef Name; 5167 TheCondStack.push_back(TheCondState); 5168 TheCondState.TheCond = AsmCond::IfCond; 5169 5170 if (TheCondState.Ignore) { 5171 eatToEndOfStatement(); 5172 } else { 5173 if (check(parseIdentifier(Name), "expected identifier after '.ifdef'") || 5174 parseToken(AsmToken::EndOfStatement, "unexpected token in '.ifdef'")) 5175 return true; 5176 5177 MCSymbol *Sym = getContext().lookupSymbol(Name); 5178 5179 if (expect_defined) 5180 TheCondState.CondMet = (Sym && !Sym->isUndefined(false)); 5181 else 5182 TheCondState.CondMet = (!Sym || Sym->isUndefined(false)); 5183 TheCondState.Ignore = !TheCondState.CondMet; 5184 } 5185 5186 return false; 5187 } 5188 5189 /// parseDirectiveElseIf 5190 /// ::= .elseif expression 5191 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) { 5192 if (TheCondState.TheCond != AsmCond::IfCond && 5193 TheCondState.TheCond != AsmCond::ElseIfCond) 5194 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an" 5195 " .if or an .elseif"); 5196 TheCondState.TheCond = AsmCond::ElseIfCond; 5197 5198 bool LastIgnoreState = false; 5199 if (!TheCondStack.empty()) 5200 LastIgnoreState = TheCondStack.back().Ignore; 5201 if (LastIgnoreState || TheCondState.CondMet) { 5202 TheCondState.Ignore = true; 5203 eatToEndOfStatement(); 5204 } else { 5205 int64_t ExprValue; 5206 if (parseAbsoluteExpression(ExprValue)) 5207 return true; 5208 5209 if (parseToken(AsmToken::EndOfStatement, 5210 "unexpected token in '.elseif' directive")) 5211 return true; 5212 5213 TheCondState.CondMet = ExprValue; 5214 TheCondState.Ignore = !TheCondState.CondMet; 5215 } 5216 5217 return false; 5218 } 5219 5220 /// parseDirectiveElse 5221 /// ::= .else 5222 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) { 5223 if (parseToken(AsmToken::EndOfStatement, 5224 "unexpected token in '.else' directive")) 5225 return true; 5226 5227 if (TheCondState.TheCond != AsmCond::IfCond && 5228 TheCondState.TheCond != AsmCond::ElseIfCond) 5229 return Error(DirectiveLoc, "Encountered a .else that doesn't follow " 5230 " an .if or an .elseif"); 5231 TheCondState.TheCond = AsmCond::ElseCond; 5232 bool LastIgnoreState = false; 5233 if (!TheCondStack.empty()) 5234 LastIgnoreState = TheCondStack.back().Ignore; 5235 if (LastIgnoreState || TheCondState.CondMet) 5236 TheCondState.Ignore = true; 5237 else 5238 TheCondState.Ignore = false; 5239 5240 return false; 5241 } 5242 5243 /// parseDirectiveEnd 5244 /// ::= .end 5245 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { 5246 if (parseToken(AsmToken::EndOfStatement, 5247 "unexpected token in '.end' directive")) 5248 return true; 5249 5250 while (Lexer.isNot(AsmToken::Eof)) 5251 Lexer.Lex(); 5252 5253 return false; 5254 } 5255 5256 /// parseDirectiveError 5257 /// ::= .err 5258 /// ::= .error [string] 5259 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) { 5260 if (!TheCondStack.empty()) { 5261 if (TheCondStack.back().Ignore) { 5262 eatToEndOfStatement(); 5263 return false; 5264 } 5265 } 5266 5267 if (!WithMessage) 5268 return Error(L, ".err encountered"); 5269 5270 StringRef Message = ".error directive invoked in source file"; 5271 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5272 if (Lexer.isNot(AsmToken::String)) 5273 return TokError(".error argument must be a string"); 5274 5275 Message = getTok().getStringContents(); 5276 Lex(); 5277 } 5278 5279 return Error(L, Message); 5280 } 5281 5282 /// parseDirectiveWarning 5283 /// ::= .warning [string] 5284 bool AsmParser::parseDirectiveWarning(SMLoc L) { 5285 if (!TheCondStack.empty()) { 5286 if (TheCondStack.back().Ignore) { 5287 eatToEndOfStatement(); 5288 return false; 5289 } 5290 } 5291 5292 StringRef Message = ".warning directive invoked in source file"; 5293 5294 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 5295 if (Lexer.isNot(AsmToken::String)) 5296 return TokError(".warning argument must be a string"); 5297 5298 Message = getTok().getStringContents(); 5299 Lex(); 5300 if (parseToken(AsmToken::EndOfStatement, 5301 "expected end of statement in '.warning' directive")) 5302 return true; 5303 } 5304 5305 return Warning(L, Message); 5306 } 5307 5308 /// parseDirectiveEndIf 5309 /// ::= .endif 5310 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { 5311 if (parseToken(AsmToken::EndOfStatement, 5312 "unexpected token in '.endif' directive")) 5313 return true; 5314 5315 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) 5316 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow " 5317 "an .if or .else"); 5318 if (!TheCondStack.empty()) { 5319 TheCondState = TheCondStack.back(); 5320 TheCondStack.pop_back(); 5321 } 5322 5323 return false; 5324 } 5325 5326 void AsmParser::initializeDirectiveKindMap() { 5327 DirectiveKindMap[".set"] = DK_SET; 5328 DirectiveKindMap[".equ"] = DK_EQU; 5329 DirectiveKindMap[".equiv"] = DK_EQUIV; 5330 DirectiveKindMap[".ascii"] = DK_ASCII; 5331 DirectiveKindMap[".asciz"] = DK_ASCIZ; 5332 DirectiveKindMap[".string"] = DK_STRING; 5333 DirectiveKindMap[".byte"] = DK_BYTE; 5334 DirectiveKindMap[".short"] = DK_SHORT; 5335 DirectiveKindMap[".value"] = DK_VALUE; 5336 DirectiveKindMap[".2byte"] = DK_2BYTE; 5337 DirectiveKindMap[".long"] = DK_LONG; 5338 DirectiveKindMap[".int"] = DK_INT; 5339 DirectiveKindMap[".4byte"] = DK_4BYTE; 5340 DirectiveKindMap[".quad"] = DK_QUAD; 5341 DirectiveKindMap[".8byte"] = DK_8BYTE; 5342 DirectiveKindMap[".octa"] = DK_OCTA; 5343 DirectiveKindMap[".single"] = DK_SINGLE; 5344 DirectiveKindMap[".float"] = DK_FLOAT; 5345 DirectiveKindMap[".double"] = DK_DOUBLE; 5346 DirectiveKindMap[".align"] = DK_ALIGN; 5347 DirectiveKindMap[".align32"] = DK_ALIGN32; 5348 DirectiveKindMap[".balign"] = DK_BALIGN; 5349 DirectiveKindMap[".balignw"] = DK_BALIGNW; 5350 DirectiveKindMap[".balignl"] = DK_BALIGNL; 5351 DirectiveKindMap[".p2align"] = DK_P2ALIGN; 5352 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW; 5353 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL; 5354 DirectiveKindMap[".org"] = DK_ORG; 5355 DirectiveKindMap[".fill"] = DK_FILL; 5356 DirectiveKindMap[".zero"] = DK_ZERO; 5357 DirectiveKindMap[".extern"] = DK_EXTERN; 5358 DirectiveKindMap[".globl"] = DK_GLOBL; 5359 DirectiveKindMap[".global"] = DK_GLOBAL; 5360 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE; 5361 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP; 5362 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER; 5363 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN; 5364 DirectiveKindMap[".reference"] = DK_REFERENCE; 5365 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION; 5366 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE; 5367 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN; 5368 DirectiveKindMap[".cold"] = DK_COLD; 5369 DirectiveKindMap[".comm"] = DK_COMM; 5370 DirectiveKindMap[".common"] = DK_COMMON; 5371 DirectiveKindMap[".lcomm"] = DK_LCOMM; 5372 DirectiveKindMap[".abort"] = DK_ABORT; 5373 DirectiveKindMap[".include"] = DK_INCLUDE; 5374 DirectiveKindMap[".incbin"] = DK_INCBIN; 5375 DirectiveKindMap[".code16"] = DK_CODE16; 5376 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC; 5377 DirectiveKindMap[".rept"] = DK_REPT; 5378 DirectiveKindMap[".rep"] = DK_REPT; 5379 DirectiveKindMap[".irp"] = DK_IRP; 5380 DirectiveKindMap[".irpc"] = DK_IRPC; 5381 DirectiveKindMap[".endr"] = DK_ENDR; 5382 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE; 5383 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK; 5384 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK; 5385 DirectiveKindMap[".if"] = DK_IF; 5386 DirectiveKindMap[".ifeq"] = DK_IFEQ; 5387 DirectiveKindMap[".ifge"] = DK_IFGE; 5388 DirectiveKindMap[".ifgt"] = DK_IFGT; 5389 DirectiveKindMap[".ifle"] = DK_IFLE; 5390 DirectiveKindMap[".iflt"] = DK_IFLT; 5391 DirectiveKindMap[".ifne"] = DK_IFNE; 5392 DirectiveKindMap[".ifb"] = DK_IFB; 5393 DirectiveKindMap[".ifnb"] = DK_IFNB; 5394 DirectiveKindMap[".ifc"] = DK_IFC; 5395 DirectiveKindMap[".ifeqs"] = DK_IFEQS; 5396 DirectiveKindMap[".ifnc"] = DK_IFNC; 5397 DirectiveKindMap[".ifnes"] = DK_IFNES; 5398 DirectiveKindMap[".ifdef"] = DK_IFDEF; 5399 DirectiveKindMap[".ifndef"] = DK_IFNDEF; 5400 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF; 5401 DirectiveKindMap[".elseif"] = DK_ELSEIF; 5402 DirectiveKindMap[".else"] = DK_ELSE; 5403 DirectiveKindMap[".end"] = DK_END; 5404 DirectiveKindMap[".endif"] = DK_ENDIF; 5405 DirectiveKindMap[".skip"] = DK_SKIP; 5406 DirectiveKindMap[".space"] = DK_SPACE; 5407 DirectiveKindMap[".file"] = DK_FILE; 5408 DirectiveKindMap[".line"] = DK_LINE; 5409 DirectiveKindMap[".loc"] = DK_LOC; 5410 DirectiveKindMap[".stabs"] = DK_STABS; 5411 DirectiveKindMap[".cv_file"] = DK_CV_FILE; 5412 DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; 5413 DirectiveKindMap[".cv_loc"] = DK_CV_LOC; 5414 DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; 5415 DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; 5416 DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; 5417 DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; 5418 DirectiveKindMap[".cv_string"] = DK_CV_STRING; 5419 DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; 5420 DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; 5421 DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; 5422 DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; 5423 DirectiveKindMap[".sleb128"] = DK_SLEB128; 5424 DirectiveKindMap[".uleb128"] = DK_ULEB128; 5425 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; 5426 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; 5427 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; 5428 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; 5429 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; 5430 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; 5431 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; 5432 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; 5433 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; 5434 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; 5435 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; 5436 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; 5437 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; 5438 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; 5439 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; 5440 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; 5441 DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; 5442 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; 5443 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; 5444 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; 5445 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; 5446 DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; 5447 DirectiveKindMap[".macros_on"] = DK_MACROS_ON; 5448 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF; 5449 DirectiveKindMap[".macro"] = DK_MACRO; 5450 DirectiveKindMap[".exitm"] = DK_EXITM; 5451 DirectiveKindMap[".endm"] = DK_ENDM; 5452 DirectiveKindMap[".endmacro"] = DK_ENDMACRO; 5453 DirectiveKindMap[".purgem"] = DK_PURGEM; 5454 DirectiveKindMap[".err"] = DK_ERR; 5455 DirectiveKindMap[".error"] = DK_ERROR; 5456 DirectiveKindMap[".warning"] = DK_WARNING; 5457 DirectiveKindMap[".altmacro"] = DK_ALTMACRO; 5458 DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO; 5459 DirectiveKindMap[".reloc"] = DK_RELOC; 5460 DirectiveKindMap[".dc"] = DK_DC; 5461 DirectiveKindMap[".dc.a"] = DK_DC_A; 5462 DirectiveKindMap[".dc.b"] = DK_DC_B; 5463 DirectiveKindMap[".dc.d"] = DK_DC_D; 5464 DirectiveKindMap[".dc.l"] = DK_DC_L; 5465 DirectiveKindMap[".dc.s"] = DK_DC_S; 5466 DirectiveKindMap[".dc.w"] = DK_DC_W; 5467 DirectiveKindMap[".dc.x"] = DK_DC_X; 5468 DirectiveKindMap[".dcb"] = DK_DCB; 5469 DirectiveKindMap[".dcb.b"] = DK_DCB_B; 5470 DirectiveKindMap[".dcb.d"] = DK_DCB_D; 5471 DirectiveKindMap[".dcb.l"] = DK_DCB_L; 5472 DirectiveKindMap[".dcb.s"] = DK_DCB_S; 5473 DirectiveKindMap[".dcb.w"] = DK_DCB_W; 5474 DirectiveKindMap[".dcb.x"] = DK_DCB_X; 5475 DirectiveKindMap[".ds"] = DK_DS; 5476 DirectiveKindMap[".ds.b"] = DK_DS_B; 5477 DirectiveKindMap[".ds.d"] = DK_DS_D; 5478 DirectiveKindMap[".ds.l"] = DK_DS_L; 5479 DirectiveKindMap[".ds.p"] = DK_DS_P; 5480 DirectiveKindMap[".ds.s"] = DK_DS_S; 5481 DirectiveKindMap[".ds.w"] = DK_DS_W; 5482 DirectiveKindMap[".ds.x"] = DK_DS_X; 5483 DirectiveKindMap[".print"] = DK_PRINT; 5484 DirectiveKindMap[".addrsig"] = DK_ADDRSIG; 5485 DirectiveKindMap[".addrsig_sym"] = DK_ADDRSIG_SYM; 5486 } 5487 5488 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { 5489 AsmToken EndToken, StartToken = getTok(); 5490 5491 unsigned NestLevel = 0; 5492 while (true) { 5493 // Check whether we have reached the end of the file. 5494 if (getLexer().is(AsmToken::Eof)) { 5495 printError(DirectiveLoc, "no matching '.endr' in definition"); 5496 return nullptr; 5497 } 5498 5499 if (Lexer.is(AsmToken::Identifier) && 5500 (getTok().getIdentifier() == ".rep" || 5501 getTok().getIdentifier() == ".rept" || 5502 getTok().getIdentifier() == ".irp" || 5503 getTok().getIdentifier() == ".irpc")) { 5504 ++NestLevel; 5505 } 5506 5507 // Otherwise, check whether we have reached the .endr. 5508 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") { 5509 if (NestLevel == 0) { 5510 EndToken = getTok(); 5511 Lex(); 5512 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5513 printError(getTok().getLoc(), 5514 "unexpected token in '.endr' directive"); 5515 return nullptr; 5516 } 5517 break; 5518 } 5519 --NestLevel; 5520 } 5521 5522 // Otherwise, scan till the end of the statement. 5523 eatToEndOfStatement(); 5524 } 5525 5526 const char *BodyStart = StartToken.getLoc().getPointer(); 5527 const char *BodyEnd = EndToken.getLoc().getPointer(); 5528 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 5529 5530 // We Are Anonymous. 5531 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters()); 5532 return &MacroLikeBodies.back(); 5533 } 5534 5535 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 5536 raw_svector_ostream &OS) { 5537 OS << ".endr\n"; 5538 5539 std::unique_ptr<MemoryBuffer> Instantiation = 5540 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 5541 5542 // Create the macro instantiation object and add to the current macro 5543 // instantiation stack. 5544 MacroInstantiation *MI = new MacroInstantiation{ 5545 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 5546 ActiveMacros.push_back(MI); 5547 5548 // Jump to the macro instantiation and prime the lexer. 5549 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 5550 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 5551 Lex(); 5552 } 5553 5554 /// parseDirectiveRept 5555 /// ::= .rep | .rept count 5556 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) { 5557 const MCExpr *CountExpr; 5558 SMLoc CountLoc = getTok().getLoc(); 5559 if (parseExpression(CountExpr)) 5560 return true; 5561 5562 int64_t Count; 5563 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) { 5564 return Error(CountLoc, "unexpected token in '" + Dir + "' directive"); 5565 } 5566 5567 if (check(Count < 0, CountLoc, "Count is negative") || 5568 parseToken(AsmToken::EndOfStatement, 5569 "unexpected token in '" + Dir + "' directive")) 5570 return true; 5571 5572 // Lex the rept definition. 5573 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5574 if (!M) 5575 return true; 5576 5577 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5578 // to hold the macro body with substitutions. 5579 SmallString<256> Buf; 5580 raw_svector_ostream OS(Buf); 5581 while (Count--) { 5582 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t). 5583 if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc())) 5584 return true; 5585 } 5586 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5587 5588 return false; 5589 } 5590 5591 /// parseDirectiveIrp 5592 /// ::= .irp symbol,values 5593 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) { 5594 MCAsmMacroParameter Parameter; 5595 MCAsmMacroArguments A; 5596 if (check(parseIdentifier(Parameter.Name), 5597 "expected identifier in '.irp' directive") || 5598 parseToken(AsmToken::Comma, "expected comma in '.irp' directive") || 5599 parseMacroArguments(nullptr, A) || 5600 parseToken(AsmToken::EndOfStatement, "expected End of Statement")) 5601 return true; 5602 5603 // Lex the irp definition. 5604 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5605 if (!M) 5606 return true; 5607 5608 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5609 // to hold the macro body with substitutions. 5610 SmallString<256> Buf; 5611 raw_svector_ostream OS(Buf); 5612 5613 for (const MCAsmMacroArgument &Arg : A) { 5614 // Note that the AtPseudoVariable is enabled for instantiations of .irp. 5615 // This is undocumented, but GAS seems to support it. 5616 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5617 return true; 5618 } 5619 5620 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5621 5622 return false; 5623 } 5624 5625 /// parseDirectiveIrpc 5626 /// ::= .irpc symbol,values 5627 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) { 5628 MCAsmMacroParameter Parameter; 5629 MCAsmMacroArguments A; 5630 5631 if (check(parseIdentifier(Parameter.Name), 5632 "expected identifier in '.irpc' directive") || 5633 parseToken(AsmToken::Comma, "expected comma in '.irpc' directive") || 5634 parseMacroArguments(nullptr, A)) 5635 return true; 5636 5637 if (A.size() != 1 || A.front().size() != 1) 5638 return TokError("unexpected token in '.irpc' directive"); 5639 5640 // Eat the end of statement. 5641 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 5642 return true; 5643 5644 // Lex the irpc definition. 5645 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5646 if (!M) 5647 return true; 5648 5649 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5650 // to hold the macro body with substitutions. 5651 SmallString<256> Buf; 5652 raw_svector_ostream OS(Buf); 5653 5654 StringRef Values = A.front().front().getString(); 5655 for (std::size_t I = 0, End = Values.size(); I != End; ++I) { 5656 MCAsmMacroArgument Arg; 5657 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1)); 5658 5659 // Note that the AtPseudoVariable is enabled for instantiations of .irpc. 5660 // This is undocumented, but GAS seems to support it. 5661 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5662 return true; 5663 } 5664 5665 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5666 5667 return false; 5668 } 5669 5670 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) { 5671 if (ActiveMacros.empty()) 5672 return TokError("unmatched '.endr' directive"); 5673 5674 // The only .repl that should get here are the ones created by 5675 // instantiateMacroLikeBody. 5676 assert(getLexer().is(AsmToken::EndOfStatement)); 5677 5678 handleMacroExit(); 5679 return false; 5680 } 5681 5682 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, 5683 size_t Len) { 5684 const MCExpr *Value; 5685 SMLoc ExprLoc = getLexer().getLoc(); 5686 if (parseExpression(Value)) 5687 return true; 5688 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5689 if (!MCE) 5690 return Error(ExprLoc, "unexpected expression in _emit"); 5691 uint64_t IntValue = MCE->getValue(); 5692 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue)) 5693 return Error(ExprLoc, "literal value out of range for directive"); 5694 5695 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len); 5696 return false; 5697 } 5698 5699 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { 5700 const MCExpr *Value; 5701 SMLoc ExprLoc = getLexer().getLoc(); 5702 if (parseExpression(Value)) 5703 return true; 5704 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5705 if (!MCE) 5706 return Error(ExprLoc, "unexpected expression in align"); 5707 uint64_t IntValue = MCE->getValue(); 5708 if (!isPowerOf2_64(IntValue)) 5709 return Error(ExprLoc, "literal value not a power of two greater then zero"); 5710 5711 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue)); 5712 return false; 5713 } 5714 5715 bool AsmParser::parseDirectivePrint(SMLoc DirectiveLoc) { 5716 const AsmToken StrTok = getTok(); 5717 Lex(); 5718 if (StrTok.isNot(AsmToken::String) || StrTok.getString().front() != '"') 5719 return Error(DirectiveLoc, "expected double quoted string after .print"); 5720 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 5721 return true; 5722 llvm::outs() << StrTok.getStringContents() << '\n'; 5723 return false; 5724 } 5725 5726 bool AsmParser::parseDirectiveAddrsig() { 5727 getStreamer().EmitAddrsig(); 5728 return false; 5729 } 5730 5731 bool AsmParser::parseDirectiveAddrsigSym() { 5732 StringRef Name; 5733 if (check(parseIdentifier(Name), 5734 "expected identifier in '.addrsig_sym' directive")) 5735 return true; 5736 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5737 getStreamer().EmitAddrsigSym(Sym); 5738 return false; 5739 } 5740 5741 // We are comparing pointers, but the pointers are relative to a single string. 5742 // Thus, this should always be deterministic. 5743 static int rewritesSort(const AsmRewrite *AsmRewriteA, 5744 const AsmRewrite *AsmRewriteB) { 5745 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) 5746 return -1; 5747 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) 5748 return 1; 5749 5750 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output 5751 // rewrite to the same location. Make sure the SizeDirective rewrite is 5752 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This 5753 // ensures the sort algorithm is stable. 5754 if (AsmRewritePrecedence[AsmRewriteA->Kind] > 5755 AsmRewritePrecedence[AsmRewriteB->Kind]) 5756 return -1; 5757 5758 if (AsmRewritePrecedence[AsmRewriteA->Kind] < 5759 AsmRewritePrecedence[AsmRewriteB->Kind]) 5760 return 1; 5761 llvm_unreachable("Unstable rewrite sort."); 5762 } 5763 5764 bool AsmParser::parseMSInlineAsm( 5765 void *AsmLoc, std::string &AsmString, unsigned &NumOutputs, 5766 unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 5767 SmallVectorImpl<std::string> &Constraints, 5768 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, 5769 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { 5770 SmallVector<void *, 4> InputDecls; 5771 SmallVector<void *, 4> OutputDecls; 5772 SmallVector<bool, 4> InputDeclsAddressOf; 5773 SmallVector<bool, 4> OutputDeclsAddressOf; 5774 SmallVector<std::string, 4> InputConstraints; 5775 SmallVector<std::string, 4> OutputConstraints; 5776 SmallVector<unsigned, 4> ClobberRegs; 5777 5778 SmallVector<AsmRewrite, 4> AsmStrRewrites; 5779 5780 // Prime the lexer. 5781 Lex(); 5782 5783 // While we have input, parse each statement. 5784 unsigned InputIdx = 0; 5785 unsigned OutputIdx = 0; 5786 while (getLexer().isNot(AsmToken::Eof)) { 5787 // Parse curly braces marking block start/end 5788 if (parseCurlyBlockScope(AsmStrRewrites)) 5789 continue; 5790 5791 ParseStatementInfo Info(&AsmStrRewrites); 5792 bool StatementErr = parseStatement(Info, &SI); 5793 5794 if (StatementErr || Info.ParseError) { 5795 // Emit pending errors if any exist. 5796 printPendingErrors(); 5797 return true; 5798 } 5799 5800 // No pending error should exist here. 5801 assert(!hasPendingError() && "unexpected error from parseStatement"); 5802 5803 if (Info.Opcode == ~0U) 5804 continue; 5805 5806 const MCInstrDesc &Desc = MII->get(Info.Opcode); 5807 5808 // Build the list of clobbers, outputs and inputs. 5809 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { 5810 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; 5811 5812 // Register operand. 5813 if (Operand.isReg() && !Operand.needAddressOf() && 5814 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) { 5815 unsigned NumDefs = Desc.getNumDefs(); 5816 // Clobber. 5817 if (NumDefs && Operand.getMCOperandNum() < NumDefs) 5818 ClobberRegs.push_back(Operand.getReg()); 5819 continue; 5820 } 5821 5822 // Expr/Input or Output. 5823 StringRef SymName = Operand.getSymName(); 5824 if (SymName.empty()) 5825 continue; 5826 5827 void *OpDecl = Operand.getOpDecl(); 5828 if (!OpDecl) 5829 continue; 5830 5831 StringRef Constraint = Operand.getConstraint(); 5832 if (Operand.isImm()) { 5833 // Offset as immediate 5834 if (Operand.isOffsetOfLocal()) 5835 Constraint = "r"; 5836 else 5837 Constraint = "i"; 5838 } 5839 5840 bool isOutput = (i == 1) && Desc.mayStore(); 5841 SMLoc Start = SMLoc::getFromPointer(SymName.data()); 5842 if (isOutput) { 5843 ++InputIdx; 5844 OutputDecls.push_back(OpDecl); 5845 OutputDeclsAddressOf.push_back(Operand.needAddressOf()); 5846 OutputConstraints.push_back(("=" + Constraint).str()); 5847 AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size()); 5848 } else { 5849 InputDecls.push_back(OpDecl); 5850 InputDeclsAddressOf.push_back(Operand.needAddressOf()); 5851 InputConstraints.push_back(Constraint.str()); 5852 if (Desc.OpInfo[i - 1].isBranchTarget()) 5853 AsmStrRewrites.emplace_back(AOK_CallInput, Start, SymName.size()); 5854 else 5855 AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size()); 5856 } 5857 } 5858 5859 // Consider implicit defs to be clobbers. Think of cpuid and push. 5860 ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(), 5861 Desc.getNumImplicitDefs()); 5862 ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end()); 5863 } 5864 5865 // Set the number of Outputs and Inputs. 5866 NumOutputs = OutputDecls.size(); 5867 NumInputs = InputDecls.size(); 5868 5869 // Set the unique clobbers. 5870 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end()); 5871 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()), 5872 ClobberRegs.end()); 5873 Clobbers.assign(ClobberRegs.size(), std::string()); 5874 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { 5875 raw_string_ostream OS(Clobbers[I]); 5876 IP->printRegName(OS, ClobberRegs[I]); 5877 } 5878 5879 // Merge the various outputs and inputs. Output are expected first. 5880 if (NumOutputs || NumInputs) { 5881 unsigned NumExprs = NumOutputs + NumInputs; 5882 OpDecls.resize(NumExprs); 5883 Constraints.resize(NumExprs); 5884 for (unsigned i = 0; i < NumOutputs; ++i) { 5885 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]); 5886 Constraints[i] = OutputConstraints[i]; 5887 } 5888 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { 5889 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]); 5890 Constraints[j] = InputConstraints[i]; 5891 } 5892 } 5893 5894 // Build the IR assembly string. 5895 std::string AsmStringIR; 5896 raw_string_ostream OS(AsmStringIR); 5897 StringRef ASMString = 5898 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer(); 5899 const char *AsmStart = ASMString.begin(); 5900 const char *AsmEnd = ASMString.end(); 5901 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort); 5902 for (auto it = AsmStrRewrites.begin(); it != AsmStrRewrites.end(); ++it) { 5903 const AsmRewrite &AR = *it; 5904 // Check if this has already been covered by another rewrite... 5905 if (AR.Done) 5906 continue; 5907 AsmRewriteKind Kind = AR.Kind; 5908 5909 const char *Loc = AR.Loc.getPointer(); 5910 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!"); 5911 5912 // Emit everything up to the immediate/expression. 5913 if (unsigned Len = Loc - AsmStart) 5914 OS << StringRef(AsmStart, Len); 5915 5916 // Skip the original expression. 5917 if (Kind == AOK_Skip) { 5918 AsmStart = Loc + AR.Len; 5919 continue; 5920 } 5921 5922 unsigned AdditionalSkip = 0; 5923 // Rewrite expressions in $N notation. 5924 switch (Kind) { 5925 default: 5926 break; 5927 case AOK_IntelExpr: 5928 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression"); 5929 if (AR.IntelExp.NeedBracs) 5930 OS << "["; 5931 if (AR.IntelExp.hasBaseReg()) 5932 OS << AR.IntelExp.BaseReg; 5933 if (AR.IntelExp.hasIndexReg()) 5934 OS << (AR.IntelExp.hasBaseReg() ? " + " : "") 5935 << AR.IntelExp.IndexReg; 5936 if (AR.IntelExp.Scale > 1) 5937 OS << " * $$" << AR.IntelExp.Scale; 5938 if (AR.IntelExp.hasOffset()) { 5939 if (AR.IntelExp.hasRegs()) 5940 OS << " + "; 5941 // Fuse this rewrite with a rewrite of the offset name, if present. 5942 StringRef OffsetName = AR.IntelExp.OffsetName; 5943 SMLoc OffsetLoc = SMLoc::getFromPointer(AR.IntelExp.OffsetName.data()); 5944 size_t OffsetLen = OffsetName.size(); 5945 auto rewrite_it = std::find_if( 5946 it, AsmStrRewrites.end(), [&](const AsmRewrite &FusingAR) { 5947 return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen && 5948 (FusingAR.Kind == AOK_Input || 5949 FusingAR.Kind == AOK_CallInput); 5950 }); 5951 if (rewrite_it == AsmStrRewrites.end()) { 5952 OS << "offset " << OffsetName; 5953 } else if (rewrite_it->Kind == AOK_CallInput) { 5954 OS << "${" << InputIdx++ << ":P}"; 5955 rewrite_it->Done = true; 5956 } else { 5957 OS << '$' << InputIdx++; 5958 rewrite_it->Done = true; 5959 } 5960 } 5961 if (AR.IntelExp.Imm || AR.IntelExp.emitImm()) 5962 OS << (AR.IntelExp.emitImm() ? "$$" : " + $$") << AR.IntelExp.Imm; 5963 if (AR.IntelExp.NeedBracs) 5964 OS << "]"; 5965 break; 5966 case AOK_Label: 5967 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; 5968 break; 5969 case AOK_Input: 5970 OS << '$' << InputIdx++; 5971 break; 5972 case AOK_CallInput: 5973 OS << "${" << InputIdx++ << ":P}"; 5974 break; 5975 case AOK_Output: 5976 OS << '$' << OutputIdx++; 5977 break; 5978 case AOK_SizeDirective: 5979 switch (AR.Val) { 5980 default: break; 5981 case 8: OS << "byte ptr "; break; 5982 case 16: OS << "word ptr "; break; 5983 case 32: OS << "dword ptr "; break; 5984 case 64: OS << "qword ptr "; break; 5985 case 80: OS << "xword ptr "; break; 5986 case 128: OS << "xmmword ptr "; break; 5987 case 256: OS << "ymmword ptr "; break; 5988 } 5989 break; 5990 case AOK_Emit: 5991 OS << ".byte"; 5992 break; 5993 case AOK_Align: { 5994 // MS alignment directives are measured in bytes. If the native assembler 5995 // measures alignment in bytes, we can pass it straight through. 5996 OS << ".align"; 5997 if (getContext().getAsmInfo()->getAlignmentIsInBytes()) 5998 break; 5999 6000 // Alignment is in log2 form, so print that instead and skip the original 6001 // immediate. 6002 unsigned Val = AR.Val; 6003 OS << ' ' << Val; 6004 assert(Val < 10 && "Expected alignment less then 2^10."); 6005 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; 6006 break; 6007 } 6008 case AOK_EVEN: 6009 OS << ".even"; 6010 break; 6011 case AOK_EndOfStatement: 6012 OS << "\n\t"; 6013 break; 6014 } 6015 6016 // Skip the original expression. 6017 AsmStart = Loc + AR.Len + AdditionalSkip; 6018 } 6019 6020 // Emit the remainder of the asm string. 6021 if (AsmStart != AsmEnd) 6022 OS << StringRef(AsmStart, AsmEnd - AsmStart); 6023 6024 AsmString = OS.str(); 6025 return false; 6026 } 6027 6028 namespace llvm { 6029 namespace MCParserUtils { 6030 6031 /// Returns whether the given symbol is used anywhere in the given expression, 6032 /// or subexpressions. 6033 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) { 6034 switch (Value->getKind()) { 6035 case MCExpr::Binary: { 6036 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value); 6037 return isSymbolUsedInExpression(Sym, BE->getLHS()) || 6038 isSymbolUsedInExpression(Sym, BE->getRHS()); 6039 } 6040 case MCExpr::Target: 6041 case MCExpr::Constant: 6042 return false; 6043 case MCExpr::SymbolRef: { 6044 const MCSymbol &S = 6045 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol(); 6046 if (S.isVariable()) 6047 return isSymbolUsedInExpression(Sym, S.getVariableValue()); 6048 return &S == Sym; 6049 } 6050 case MCExpr::Unary: 6051 return isSymbolUsedInExpression( 6052 Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr()); 6053 } 6054 6055 llvm_unreachable("Unknown expr kind!"); 6056 } 6057 6058 bool parseAssignmentExpression(StringRef Name, bool allow_redef, 6059 MCAsmParser &Parser, MCSymbol *&Sym, 6060 const MCExpr *&Value) { 6061 6062 // FIXME: Use better location, we should use proper tokens. 6063 SMLoc EqualLoc = Parser.getTok().getLoc(); 6064 if (Parser.parseExpression(Value)) 6065 return Parser.TokError("missing expression"); 6066 6067 // Note: we don't count b as used in "a = b". This is to allow 6068 // a = b 6069 // b = c 6070 6071 if (Parser.parseToken(AsmToken::EndOfStatement)) 6072 return true; 6073 6074 // Validate that the LHS is allowed to be a variable (either it has not been 6075 // used as a symbol, or it is an absolute symbol). 6076 Sym = Parser.getContext().lookupSymbol(Name); 6077 if (Sym) { 6078 // Diagnose assignment to a label. 6079 // 6080 // FIXME: Diagnostics. Note the location of the definition as a label. 6081 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 6082 if (isSymbolUsedInExpression(Sym, Value)) 6083 return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'"); 6084 else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() && 6085 !Sym->isVariable()) 6086 ; // Allow redefinitions of undefined symbols only used in directives. 6087 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef) 6088 ; // Allow redefinitions of variables that haven't yet been used. 6089 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 6090 return Parser.Error(EqualLoc, "redefinition of '" + Name + "'"); 6091 else if (!Sym->isVariable()) 6092 return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'"); 6093 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 6094 return Parser.Error(EqualLoc, 6095 "invalid reassignment of non-absolute variable '" + 6096 Name + "'"); 6097 } else if (Name == ".") { 6098 Parser.getStreamer().emitValueToOffset(Value, 0, EqualLoc); 6099 return false; 6100 } else 6101 Sym = Parser.getContext().getOrCreateSymbol(Name); 6102 6103 Sym->setRedefinable(allow_redef); 6104 6105 return false; 6106 } 6107 6108 } // end namespace MCParserUtils 6109 } // end namespace llvm 6110 6111 /// Create an MCAsmParser instance. 6112 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C, 6113 MCStreamer &Out, const MCAsmInfo &MAI, 6114 unsigned CB) { 6115 return new AsmParser(SM, C, Out, MAI, CB); 6116 } 6117