1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This class implements the parser for assembly files. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "llvm/ADT/APFloat.h" 14 #include "llvm/ADT/APInt.h" 15 #include "llvm/ADT/ArrayRef.h" 16 #include "llvm/ADT/None.h" 17 #include "llvm/ADT/STLExtras.h" 18 #include "llvm/ADT/SmallString.h" 19 #include "llvm/ADT/SmallVector.h" 20 #include "llvm/ADT/StringExtras.h" 21 #include "llvm/ADT/StringMap.h" 22 #include "llvm/ADT/StringRef.h" 23 #include "llvm/ADT/Twine.h" 24 #include "llvm/BinaryFormat/Dwarf.h" 25 #include "llvm/MC/MCAsmInfo.h" 26 #include "llvm/MC/MCCodeView.h" 27 #include "llvm/MC/MCContext.h" 28 #include "llvm/MC/MCDirectives.h" 29 #include "llvm/MC/MCDwarf.h" 30 #include "llvm/MC/MCExpr.h" 31 #include "llvm/MC/MCInstPrinter.h" 32 #include "llvm/MC/MCInstrDesc.h" 33 #include "llvm/MC/MCInstrInfo.h" 34 #include "llvm/MC/MCObjectFileInfo.h" 35 #include "llvm/MC/MCParser/AsmCond.h" 36 #include "llvm/MC/MCParser/AsmLexer.h" 37 #include "llvm/MC/MCParser/MCAsmLexer.h" 38 #include "llvm/MC/MCParser/MCAsmParser.h" 39 #include "llvm/MC/MCParser/MCAsmParserExtension.h" 40 #include "llvm/MC/MCParser/MCAsmParserUtils.h" 41 #include "llvm/MC/MCParser/MCParsedAsmOperand.h" 42 #include "llvm/MC/MCParser/MCTargetAsmParser.h" 43 #include "llvm/MC/MCRegisterInfo.h" 44 #include "llvm/MC/MCSection.h" 45 #include "llvm/MC/MCStreamer.h" 46 #include "llvm/MC/MCSymbol.h" 47 #include "llvm/MC/MCTargetOptions.h" 48 #include "llvm/MC/MCValue.h" 49 #include "llvm/Support/Casting.h" 50 #include "llvm/Support/CommandLine.h" 51 #include "llvm/Support/ErrorHandling.h" 52 #include "llvm/Support/MD5.h" 53 #include "llvm/Support/MathExtras.h" 54 #include "llvm/Support/MemoryBuffer.h" 55 #include "llvm/Support/SMLoc.h" 56 #include "llvm/Support/SourceMgr.h" 57 #include "llvm/Support/raw_ostream.h" 58 #include <algorithm> 59 #include <cassert> 60 #include <cctype> 61 #include <climits> 62 #include <cstddef> 63 #include <cstdint> 64 #include <deque> 65 #include <memory> 66 #include <sstream> 67 #include <string> 68 #include <tuple> 69 #include <utility> 70 #include <vector> 71 72 using namespace llvm; 73 74 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() = default; 75 76 static cl::opt<unsigned> AsmMacroMaxNestingDepth( 77 "asm-macro-max-nesting-depth", cl::init(20), cl::Hidden, 78 cl::desc("The maximum nesting depth allowed for assembly macros.")); 79 80 namespace { 81 82 /// Helper types for tracking macro definitions. 83 typedef std::vector<AsmToken> MCAsmMacroArgument; 84 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; 85 86 /// Helper class for storing information about an active macro 87 /// instantiation. 88 struct MacroInstantiation { 89 /// The location of the instantiation. 90 SMLoc InstantiationLoc; 91 92 /// The buffer where parsing should resume upon instantiation completion. 93 int ExitBuffer; 94 95 /// The location where parsing should resume upon instantiation completion. 96 SMLoc ExitLoc; 97 98 /// The depth of TheCondStack at the start of the instantiation. 99 size_t CondStackDepth; 100 101 public: 102 MacroInstantiation(SMLoc IL, int EB, SMLoc EL, size_t CondStackDepth); 103 }; 104 105 struct ParseStatementInfo { 106 /// The parsed operands from the last parsed statement. 107 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; 108 109 /// The opcode from the last parsed instruction. 110 unsigned Opcode = ~0U; 111 112 /// Was there an error parsing the inline assembly? 113 bool ParseError = false; 114 115 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr; 116 117 ParseStatementInfo() = delete; 118 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites) 119 : AsmRewrites(rewrites) {} 120 }; 121 122 /// The concrete assembly parser instance. 123 class AsmParser : public MCAsmParser { 124 private: 125 AsmLexer Lexer; 126 MCContext &Ctx; 127 MCStreamer &Out; 128 const MCAsmInfo &MAI; 129 SourceMgr &SrcMgr; 130 SourceMgr::DiagHandlerTy SavedDiagHandler; 131 void *SavedDiagContext; 132 std::unique_ptr<MCAsmParserExtension> PlatformParser; 133 134 /// This is the current buffer index we're lexing from as managed by the 135 /// SourceMgr object. 136 unsigned CurBuffer; 137 138 AsmCond TheCondState; 139 std::vector<AsmCond> TheCondStack; 140 141 /// maps directive names to handler methods in parser 142 /// extensions. Extensions register themselves in this map by calling 143 /// addDirectiveHandler. 144 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap; 145 146 /// Stack of active macro instantiations. 147 std::vector<MacroInstantiation*> ActiveMacros; 148 149 /// List of bodies of anonymous macros. 150 std::deque<MCAsmMacro> MacroLikeBodies; 151 152 /// Boolean tracking whether macro substitution is enabled. 153 unsigned MacrosEnabledFlag : 1; 154 155 /// Keeps track of how many .macro's have been instantiated. 156 unsigned NumOfMacroInstantiations; 157 158 /// The values from the last parsed cpp hash file line comment if any. 159 struct CppHashInfoTy { 160 StringRef Filename; 161 int64_t LineNumber; 162 SMLoc Loc; 163 unsigned Buf; 164 CppHashInfoTy() : Filename(), LineNumber(0), Loc(), Buf(0) {} 165 }; 166 CppHashInfoTy CppHashInfo; 167 168 /// The filename from the first cpp hash file line comment, if any. 169 StringRef FirstCppHashFilename; 170 171 /// List of forward directional labels for diagnosis at the end. 172 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; 173 174 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 175 unsigned AssemblerDialect = ~0U; 176 177 /// is Darwin compatibility enabled? 178 bool IsDarwin = false; 179 180 /// Are we parsing ms-style inline assembly? 181 bool ParsingInlineAsm = false; 182 183 /// Did we already inform the user about inconsistent MD5 usage? 184 bool ReportedInconsistentMD5 = false; 185 186 // Is alt macro mode enabled. 187 bool AltMacroMode = false; 188 189 public: 190 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 191 const MCAsmInfo &MAI, unsigned CB); 192 AsmParser(const AsmParser &) = delete; 193 AsmParser &operator=(const AsmParser &) = delete; 194 ~AsmParser() override; 195 196 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override; 197 198 void addDirectiveHandler(StringRef Directive, 199 ExtensionDirectiveHandler Handler) override { 200 ExtensionDirectiveMap[Directive] = Handler; 201 } 202 203 void addAliasForDirective(StringRef Directive, StringRef Alias) override { 204 DirectiveKindMap[Directive] = DirectiveKindMap[Alias]; 205 } 206 207 /// @name MCAsmParser Interface 208 /// { 209 210 SourceMgr &getSourceManager() override { return SrcMgr; } 211 MCAsmLexer &getLexer() override { return Lexer; } 212 MCContext &getContext() override { return Ctx; } 213 MCStreamer &getStreamer() override { return Out; } 214 215 CodeViewContext &getCVContext() { return Ctx.getCVContext(); } 216 217 unsigned getAssemblerDialect() override { 218 if (AssemblerDialect == ~0U) 219 return MAI.getAssemblerDialect(); 220 else 221 return AssemblerDialect; 222 } 223 void setAssemblerDialect(unsigned i) override { 224 AssemblerDialect = i; 225 } 226 227 void Note(SMLoc L, const Twine &Msg, SMRange Range = None) override; 228 bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) override; 229 bool printError(SMLoc L, const Twine &Msg, SMRange Range = None) override; 230 231 const AsmToken &Lex() override; 232 233 void setParsingInlineAsm(bool V) override { 234 ParsingInlineAsm = V; 235 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and 236 // hex integer literals. 237 Lexer.setLexMasmIntegers(V); 238 } 239 bool isParsingInlineAsm() override { return ParsingInlineAsm; } 240 241 bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString, 242 unsigned &NumOutputs, unsigned &NumInputs, 243 SmallVectorImpl<std::pair<void *,bool>> &OpDecls, 244 SmallVectorImpl<std::string> &Constraints, 245 SmallVectorImpl<std::string> &Clobbers, 246 const MCInstrInfo *MII, const MCInstPrinter *IP, 247 MCAsmParserSemaCallback &SI) override; 248 249 bool parseExpression(const MCExpr *&Res); 250 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 251 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override; 252 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 253 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 254 SMLoc &EndLoc) override; 255 bool parseAbsoluteExpression(int64_t &Res) override; 256 257 /// Parse a floating point expression using the float \p Semantics 258 /// and set \p Res to the value. 259 bool parseRealValue(const fltSemantics &Semantics, APInt &Res); 260 261 /// Parse an identifier or string (as a quoted identifier) 262 /// and set \p Res to the identifier contents. 263 bool parseIdentifier(StringRef &Res) override; 264 void eatToEndOfStatement() override; 265 266 bool checkForValidSection() override; 267 268 /// } 269 270 private: 271 bool parseStatement(ParseStatementInfo &Info, 272 MCAsmParserSemaCallback *SI); 273 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); 274 bool parseCppHashLineFilenameComment(SMLoc L); 275 276 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body, 277 ArrayRef<MCAsmMacroParameter> Parameters); 278 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 279 ArrayRef<MCAsmMacroParameter> Parameters, 280 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable, 281 SMLoc L); 282 283 /// Are macros enabled in the parser? 284 bool areMacrosEnabled() {return MacrosEnabledFlag;} 285 286 /// Control a flag in the parser that enables or disables macros. 287 void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;} 288 289 /// Are we inside a macro instantiation? 290 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();} 291 292 /// Handle entry to macro instantiation. 293 /// 294 /// \param M The macro. 295 /// \param NameLoc Instantiation location. 296 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc); 297 298 /// Handle exit from macro instantiation. 299 void handleMacroExit(); 300 301 /// Extract AsmTokens for a macro argument. 302 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg); 303 304 /// Parse all macro arguments for a given macro. 305 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A); 306 307 void printMacroInstantiations(); 308 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 309 SMRange Range = None) const { 310 ArrayRef<SMRange> Ranges(Range); 311 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 312 } 313 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 314 315 /// Should we emit DWARF describing this assembler source? (Returns false if 316 /// the source has .file directives, which means we don't want to generate 317 /// info describing the assembler source itself.) 318 bool enabledGenDwarfForAssembly(); 319 320 /// Enter the specified file. This returns true on failure. 321 bool enterIncludeFile(const std::string &Filename); 322 323 /// Process the specified file for the .incbin directive. 324 /// This returns true on failure. 325 bool processIncbinFile(const std::string &Filename, int64_t Skip = 0, 326 const MCExpr *Count = nullptr, SMLoc Loc = SMLoc()); 327 328 /// Reset the current lexer position to that given by \p Loc. The 329 /// current token is not set; clients should ensure Lex() is called 330 /// subsequently. 331 /// 332 /// \param InBuffer If not 0, should be the known buffer id that contains the 333 /// location. 334 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0); 335 336 /// Parse up to the end of statement and a return the contents from the 337 /// current token until the end of the statement; the current token on exit 338 /// will be either the EndOfStatement or EOF. 339 StringRef parseStringToEndOfStatement() override; 340 341 /// Parse until the end of a statement or a comma is encountered, 342 /// return the contents from the current token up to the end or comma. 343 StringRef parseStringToComma(); 344 345 bool parseAssignment(StringRef Name, bool allow_redef, 346 bool NoDeadStrip = false); 347 348 unsigned getBinOpPrecedence(AsmToken::TokenKind K, 349 MCBinaryExpr::Opcode &Kind); 350 351 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 352 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 353 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 354 355 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 356 357 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName); 358 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName); 359 360 // Generic (target and platform independent) directive parsing. 361 enum DirectiveKind { 362 DK_NO_DIRECTIVE, // Placeholder 363 DK_SET, 364 DK_EQU, 365 DK_EQUIV, 366 DK_ASCII, 367 DK_ASCIZ, 368 DK_STRING, 369 DK_BYTE, 370 DK_SHORT, 371 DK_RELOC, 372 DK_VALUE, 373 DK_2BYTE, 374 DK_LONG, 375 DK_INT, 376 DK_4BYTE, 377 DK_QUAD, 378 DK_8BYTE, 379 DK_OCTA, 380 DK_DC, 381 DK_DC_A, 382 DK_DC_B, 383 DK_DC_D, 384 DK_DC_L, 385 DK_DC_S, 386 DK_DC_W, 387 DK_DC_X, 388 DK_DCB, 389 DK_DCB_B, 390 DK_DCB_D, 391 DK_DCB_L, 392 DK_DCB_S, 393 DK_DCB_W, 394 DK_DCB_X, 395 DK_DS, 396 DK_DS_B, 397 DK_DS_D, 398 DK_DS_L, 399 DK_DS_P, 400 DK_DS_S, 401 DK_DS_W, 402 DK_DS_X, 403 DK_SINGLE, 404 DK_FLOAT, 405 DK_DOUBLE, 406 DK_ALIGN, 407 DK_ALIGN32, 408 DK_BALIGN, 409 DK_BALIGNW, 410 DK_BALIGNL, 411 DK_P2ALIGN, 412 DK_P2ALIGNW, 413 DK_P2ALIGNL, 414 DK_ORG, 415 DK_FILL, 416 DK_ENDR, 417 DK_BUNDLE_ALIGN_MODE, 418 DK_BUNDLE_LOCK, 419 DK_BUNDLE_UNLOCK, 420 DK_ZERO, 421 DK_EXTERN, 422 DK_GLOBL, 423 DK_GLOBAL, 424 DK_LAZY_REFERENCE, 425 DK_NO_DEAD_STRIP, 426 DK_SYMBOL_RESOLVER, 427 DK_PRIVATE_EXTERN, 428 DK_REFERENCE, 429 DK_WEAK_DEFINITION, 430 DK_WEAK_REFERENCE, 431 DK_WEAK_DEF_CAN_BE_HIDDEN, 432 DK_COLD, 433 DK_COMM, 434 DK_COMMON, 435 DK_LCOMM, 436 DK_ABORT, 437 DK_INCLUDE, 438 DK_INCBIN, 439 DK_CODE16, 440 DK_CODE16GCC, 441 DK_REPT, 442 DK_IRP, 443 DK_IRPC, 444 DK_IF, 445 DK_IFEQ, 446 DK_IFGE, 447 DK_IFGT, 448 DK_IFLE, 449 DK_IFLT, 450 DK_IFNE, 451 DK_IFB, 452 DK_IFNB, 453 DK_IFC, 454 DK_IFEQS, 455 DK_IFNC, 456 DK_IFNES, 457 DK_IFDEF, 458 DK_IFNDEF, 459 DK_IFNOTDEF, 460 DK_ELSEIF, 461 DK_ELSE, 462 DK_ENDIF, 463 DK_SPACE, 464 DK_SKIP, 465 DK_FILE, 466 DK_LINE, 467 DK_LOC, 468 DK_STABS, 469 DK_CV_FILE, 470 DK_CV_FUNC_ID, 471 DK_CV_INLINE_SITE_ID, 472 DK_CV_LOC, 473 DK_CV_LINETABLE, 474 DK_CV_INLINE_LINETABLE, 475 DK_CV_DEF_RANGE, 476 DK_CV_STRINGTABLE, 477 DK_CV_STRING, 478 DK_CV_FILECHECKSUMS, 479 DK_CV_FILECHECKSUM_OFFSET, 480 DK_CV_FPO_DATA, 481 DK_CFI_SECTIONS, 482 DK_CFI_STARTPROC, 483 DK_CFI_ENDPROC, 484 DK_CFI_DEF_CFA, 485 DK_CFI_DEF_CFA_OFFSET, 486 DK_CFI_ADJUST_CFA_OFFSET, 487 DK_CFI_DEF_CFA_REGISTER, 488 DK_CFI_OFFSET, 489 DK_CFI_REL_OFFSET, 490 DK_CFI_PERSONALITY, 491 DK_CFI_LSDA, 492 DK_CFI_REMEMBER_STATE, 493 DK_CFI_RESTORE_STATE, 494 DK_CFI_SAME_VALUE, 495 DK_CFI_RESTORE, 496 DK_CFI_ESCAPE, 497 DK_CFI_RETURN_COLUMN, 498 DK_CFI_SIGNAL_FRAME, 499 DK_CFI_UNDEFINED, 500 DK_CFI_REGISTER, 501 DK_CFI_WINDOW_SAVE, 502 DK_CFI_B_KEY_FRAME, 503 DK_MACROS_ON, 504 DK_MACROS_OFF, 505 DK_ALTMACRO, 506 DK_NOALTMACRO, 507 DK_MACRO, 508 DK_EXITM, 509 DK_ENDM, 510 DK_ENDMACRO, 511 DK_PURGEM, 512 DK_SLEB128, 513 DK_ULEB128, 514 DK_ERR, 515 DK_ERROR, 516 DK_WARNING, 517 DK_PRINT, 518 DK_ADDRSIG, 519 DK_ADDRSIG_SYM, 520 DK_END 521 }; 522 523 /// Maps directive name --> DirectiveKind enum, for 524 /// directives parsed by this class. 525 StringMap<DirectiveKind> DirectiveKindMap; 526 527 // ".ascii", ".asciz", ".string" 528 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 529 bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc" 530 bool parseDirectiveValue(StringRef IDVal, 531 unsigned Size); // ".byte", ".long", ... 532 bool parseDirectiveOctaValue(StringRef IDVal); // ".octa", ... 533 bool parseDirectiveRealValue(StringRef IDVal, 534 const fltSemantics &); // ".single", ... 535 bool parseDirectiveFill(); // ".fill" 536 bool parseDirectiveZero(); // ".zero" 537 // ".set", ".equ", ".equiv" 538 bool parseDirectiveSet(StringRef IDVal, bool allow_redef); 539 bool parseDirectiveOrg(); // ".org" 540 // ".align{,32}", ".p2align{,w,l}" 541 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize); 542 543 // ".file", ".line", ".loc", ".stabs" 544 bool parseDirectiveFile(SMLoc DirectiveLoc); 545 bool parseDirectiveLine(); 546 bool parseDirectiveLoc(); 547 bool parseDirectiveStabs(); 548 549 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable", 550 // ".cv_inline_linetable", ".cv_def_range", ".cv_string" 551 bool parseDirectiveCVFile(); 552 bool parseDirectiveCVFuncId(); 553 bool parseDirectiveCVInlineSiteId(); 554 bool parseDirectiveCVLoc(); 555 bool parseDirectiveCVLinetable(); 556 bool parseDirectiveCVInlineLinetable(); 557 bool parseDirectiveCVDefRange(); 558 bool parseDirectiveCVString(); 559 bool parseDirectiveCVStringTable(); 560 bool parseDirectiveCVFileChecksums(); 561 bool parseDirectiveCVFileChecksumOffset(); 562 bool parseDirectiveCVFPOData(); 563 564 // .cfi directives 565 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc); 566 bool parseDirectiveCFIWindowSave(); 567 bool parseDirectiveCFISections(); 568 bool parseDirectiveCFIStartProc(); 569 bool parseDirectiveCFIEndProc(); 570 bool parseDirectiveCFIDefCfaOffset(); 571 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc); 572 bool parseDirectiveCFIAdjustCfaOffset(); 573 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc); 574 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc); 575 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc); 576 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality); 577 bool parseDirectiveCFIRememberState(); 578 bool parseDirectiveCFIRestoreState(); 579 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc); 580 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc); 581 bool parseDirectiveCFIEscape(); 582 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc); 583 bool parseDirectiveCFISignalFrame(); 584 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc); 585 586 // macro directives 587 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc); 588 bool parseDirectiveExitMacro(StringRef Directive); 589 bool parseDirectiveEndMacro(StringRef Directive); 590 bool parseDirectiveMacro(SMLoc DirectiveLoc); 591 bool parseDirectiveMacrosOnOff(StringRef Directive); 592 // alternate macro mode directives 593 bool parseDirectiveAltmacro(StringRef Directive); 594 // ".bundle_align_mode" 595 bool parseDirectiveBundleAlignMode(); 596 // ".bundle_lock" 597 bool parseDirectiveBundleLock(); 598 // ".bundle_unlock" 599 bool parseDirectiveBundleUnlock(); 600 601 // ".space", ".skip" 602 bool parseDirectiveSpace(StringRef IDVal); 603 604 // ".dcb" 605 bool parseDirectiveDCB(StringRef IDVal, unsigned Size); 606 bool parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &); 607 // ".ds" 608 bool parseDirectiveDS(StringRef IDVal, unsigned Size); 609 610 // .sleb128 (Signed=true) and .uleb128 (Signed=false) 611 bool parseDirectiveLEB128(bool Signed); 612 613 /// Parse a directive like ".globl" which 614 /// accepts a single symbol (which should be a label or an external). 615 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr); 616 617 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 618 619 bool parseDirectiveAbort(); // ".abort" 620 bool parseDirectiveInclude(); // ".include" 621 bool parseDirectiveIncbin(); // ".incbin" 622 623 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne" 624 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 625 // ".ifb" or ".ifnb", depending on ExpectBlank. 626 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 627 // ".ifc" or ".ifnc", depending on ExpectEqual. 628 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual); 629 // ".ifeqs" or ".ifnes", depending on ExpectEqual. 630 bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual); 631 // ".ifdef" or ".ifndef", depending on expect_defined 632 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 633 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 634 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else" 635 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 636 bool parseEscapedString(std::string &Data) override; 637 638 const MCExpr *applyModifierToExpr(const MCExpr *E, 639 MCSymbolRefExpr::VariantKind Variant); 640 641 // Macro-like directives 642 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc); 643 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 644 raw_svector_ostream &OS); 645 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive); 646 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 647 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc" 648 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 649 650 // "_emit" or "__emit" 651 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info, 652 size_t Len); 653 654 // "align" 655 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info); 656 657 // "end" 658 bool parseDirectiveEnd(SMLoc DirectiveLoc); 659 660 // ".err" or ".error" 661 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage); 662 663 // ".warning" 664 bool parseDirectiveWarning(SMLoc DirectiveLoc); 665 666 // .print <double-quotes-string> 667 bool parseDirectivePrint(SMLoc DirectiveLoc); 668 669 // Directives to support address-significance tables. 670 bool parseDirectiveAddrsig(); 671 bool parseDirectiveAddrsigSym(); 672 673 void initializeDirectiveKindMap(); 674 }; 675 676 } // end anonymous namespace 677 678 namespace llvm { 679 680 extern MCAsmParserExtension *createDarwinAsmParser(); 681 extern MCAsmParserExtension *createELFAsmParser(); 682 extern MCAsmParserExtension *createCOFFAsmParser(); 683 extern MCAsmParserExtension *createWasmAsmParser(); 684 685 } // end namespace llvm 686 687 enum { DEFAULT_ADDRSPACE = 0 }; 688 689 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 690 const MCAsmInfo &MAI, unsigned CB = 0) 691 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM), 692 CurBuffer(CB ? CB : SM.getMainFileID()), MacrosEnabledFlag(true) { 693 HadError = false; 694 // Save the old handler. 695 SavedDiagHandler = SrcMgr.getDiagHandler(); 696 SavedDiagContext = SrcMgr.getDiagContext(); 697 // Set our own handler which calls the saved handler. 698 SrcMgr.setDiagHandler(DiagHandler, this); 699 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 700 701 // Initialize the platform / file format parser. 702 switch (Ctx.getObjectFileInfo()->getObjectFileType()) { 703 case MCObjectFileInfo::IsCOFF: 704 PlatformParser.reset(createCOFFAsmParser()); 705 break; 706 case MCObjectFileInfo::IsMachO: 707 PlatformParser.reset(createDarwinAsmParser()); 708 IsDarwin = true; 709 break; 710 case MCObjectFileInfo::IsELF: 711 PlatformParser.reset(createELFAsmParser()); 712 break; 713 case MCObjectFileInfo::IsWasm: 714 PlatformParser.reset(createWasmAsmParser()); 715 break; 716 case MCObjectFileInfo::IsXCOFF: 717 // TODO: Need to implement createXCOFFAsmParser for XCOFF format. 718 break; 719 } 720 721 PlatformParser->Initialize(*this); 722 initializeDirectiveKindMap(); 723 724 NumOfMacroInstantiations = 0; 725 } 726 727 AsmParser::~AsmParser() { 728 assert((HadError || ActiveMacros.empty()) && 729 "Unexpected active macro instantiation!"); 730 731 // Restore the saved diagnostics handler and context for use during 732 // finalization. 733 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext); 734 } 735 736 void AsmParser::printMacroInstantiations() { 737 // Print the active macro instantiation stack. 738 for (std::vector<MacroInstantiation *>::const_reverse_iterator 739 it = ActiveMacros.rbegin(), 740 ie = ActiveMacros.rend(); 741 it != ie; ++it) 742 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 743 "while in macro instantiation"); 744 } 745 746 void AsmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { 747 printPendingErrors(); 748 printMessage(L, SourceMgr::DK_Note, Msg, Range); 749 printMacroInstantiations(); 750 } 751 752 bool AsmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) { 753 if(getTargetParser().getTargetOptions().MCNoWarn) 754 return false; 755 if (getTargetParser().getTargetOptions().MCFatalWarnings) 756 return Error(L, Msg, Range); 757 printMessage(L, SourceMgr::DK_Warning, Msg, Range); 758 printMacroInstantiations(); 759 return false; 760 } 761 762 bool AsmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { 763 HadError = true; 764 printMessage(L, SourceMgr::DK_Error, Msg, Range); 765 printMacroInstantiations(); 766 return true; 767 } 768 769 bool AsmParser::enterIncludeFile(const std::string &Filename) { 770 std::string IncludedFile; 771 unsigned NewBuf = 772 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 773 if (!NewBuf) 774 return true; 775 776 CurBuffer = NewBuf; 777 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 778 return false; 779 } 780 781 /// Process the specified .incbin file by searching for it in the include paths 782 /// then just emitting the byte contents of the file to the streamer. This 783 /// returns true on failure. 784 bool AsmParser::processIncbinFile(const std::string &Filename, int64_t Skip, 785 const MCExpr *Count, SMLoc Loc) { 786 std::string IncludedFile; 787 unsigned NewBuf = 788 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 789 if (!NewBuf) 790 return true; 791 792 // Pick up the bytes from the file and emit them. 793 StringRef Bytes = SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(); 794 Bytes = Bytes.drop_front(Skip); 795 if (Count) { 796 int64_t Res; 797 if (!Count->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 798 return Error(Loc, "expected absolute expression"); 799 if (Res < 0) 800 return Warning(Loc, "negative count has no effect"); 801 Bytes = Bytes.take_front(Res); 802 } 803 getStreamer().EmitBytes(Bytes); 804 return false; 805 } 806 807 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) { 808 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); 809 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), 810 Loc.getPointer()); 811 } 812 813 const AsmToken &AsmParser::Lex() { 814 if (Lexer.getTok().is(AsmToken::Error)) 815 Error(Lexer.getErrLoc(), Lexer.getErr()); 816 817 // if it's a end of statement with a comment in it 818 if (getTok().is(AsmToken::EndOfStatement)) { 819 // if this is a line comment output it. 820 if (!getTok().getString().empty() && getTok().getString().front() != '\n' && 821 getTok().getString().front() != '\r' && MAI.preserveAsmComments()) 822 Out.addExplicitComment(Twine(getTok().getString())); 823 } 824 825 const AsmToken *tok = &Lexer.Lex(); 826 827 // Parse comments here to be deferred until end of next statement. 828 while (tok->is(AsmToken::Comment)) { 829 if (MAI.preserveAsmComments()) 830 Out.addExplicitComment(Twine(tok->getString())); 831 tok = &Lexer.Lex(); 832 } 833 834 if (tok->is(AsmToken::Eof)) { 835 // If this is the end of an included file, pop the parent file off the 836 // include stack. 837 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 838 if (ParentIncludeLoc != SMLoc()) { 839 jumpToLoc(ParentIncludeLoc); 840 return Lex(); 841 } 842 } 843 844 return *tok; 845 } 846 847 bool AsmParser::enabledGenDwarfForAssembly() { 848 // Check whether the user specified -g. 849 if (!getContext().getGenDwarfForAssembly()) 850 return false; 851 // If we haven't encountered any .file directives (which would imply that 852 // the assembler source was produced with debug info already) then emit one 853 // describing the assembler source file itself. 854 if (getContext().getGenDwarfFileNumber() == 0) { 855 // Use the first #line directive for this, if any. It's preprocessed, so 856 // there is no checksum, and of course no source directive. 857 if (!FirstCppHashFilename.empty()) 858 getContext().setMCLineTableRootFile(/*CUID=*/0, 859 getContext().getCompilationDir(), 860 FirstCppHashFilename, 861 /*Cksum=*/None, /*Source=*/None); 862 const MCDwarfFile &RootFile = 863 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); 864 getContext().setGenDwarfFileNumber(getStreamer().EmitDwarfFileDirective( 865 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name, 866 RootFile.Checksum, RootFile.Source)); 867 } 868 return true; 869 } 870 871 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 872 // Create the initial section, if requested. 873 if (!NoInitialTextSection) 874 Out.InitSections(false); 875 876 // Prime the lexer. 877 Lex(); 878 879 HadError = false; 880 AsmCond StartingCondState = TheCondState; 881 SmallVector<AsmRewrite, 4> AsmStrRewrites; 882 883 // If we are generating dwarf for assembly source files save the initial text 884 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't 885 // emitting any actual debug info yet and haven't had a chance to parse any 886 // embedded .file directives.) 887 if (getContext().getGenDwarfForAssembly()) { 888 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 889 if (!Sec->getBeginSymbol()) { 890 MCSymbol *SectionStartSym = getContext().createTempSymbol(); 891 getStreamer().EmitLabel(SectionStartSym); 892 Sec->setBeginSymbol(SectionStartSym); 893 } 894 bool InsertResult = getContext().addGenDwarfSection(Sec); 895 assert(InsertResult && ".text section should not have debug info yet"); 896 (void)InsertResult; 897 } 898 899 // While we have input, parse each statement. 900 while (Lexer.isNot(AsmToken::Eof)) { 901 ParseStatementInfo Info(&AsmStrRewrites); 902 if (!parseStatement(Info, nullptr)) 903 continue; 904 905 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error 906 // for printing ErrMsg via Lex() only if no (presumably better) parser error 907 // exists. 908 if (!hasPendingError() && Lexer.getTok().is(AsmToken::Error)) { 909 Lex(); 910 } 911 912 // parseStatement returned true so may need to emit an error. 913 printPendingErrors(); 914 915 // Skipping to the next line if needed. 916 if (!getLexer().isAtStartOfStatement()) 917 eatToEndOfStatement(); 918 } 919 920 getTargetParser().onEndOfFile(); 921 printPendingErrors(); 922 923 // All errors should have been emitted. 924 assert(!hasPendingError() && "unexpected error from parseStatement"); 925 926 getTargetParser().flushPendingInstructions(getStreamer()); 927 928 if (TheCondState.TheCond != StartingCondState.TheCond || 929 TheCondState.Ignore != StartingCondState.Ignore) 930 printError(getTok().getLoc(), "unmatched .ifs or .elses"); 931 // Check to see there are no empty DwarfFile slots. 932 const auto &LineTables = getContext().getMCDwarfLineTables(); 933 if (!LineTables.empty()) { 934 unsigned Index = 0; 935 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) { 936 if (File.Name.empty() && Index != 0) 937 printError(getTok().getLoc(), "unassigned file number: " + 938 Twine(Index) + 939 " for .file directives"); 940 ++Index; 941 } 942 } 943 944 // Check to see that all assembler local symbols were actually defined. 945 // Targets that don't do subsections via symbols may not want this, though, 946 // so conservatively exclude them. Only do this if we're finalizing, though, 947 // as otherwise we won't necessarilly have seen everything yet. 948 if (!NoFinalize) { 949 if (MAI.hasSubsectionsViaSymbols()) { 950 for (const auto &TableEntry : getContext().getSymbols()) { 951 MCSymbol *Sym = TableEntry.getValue(); 952 // Variable symbols may not be marked as defined, so check those 953 // explicitly. If we know it's a variable, we have a definition for 954 // the purposes of this check. 955 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 956 // FIXME: We would really like to refer back to where the symbol was 957 // first referenced for a source location. We need to add something 958 // to track that. Currently, we just point to the end of the file. 959 printError(getTok().getLoc(), "assembler local symbol '" + 960 Sym->getName() + "' not defined"); 961 } 962 } 963 964 // Temporary symbols like the ones for directional jumps don't go in the 965 // symbol table. They also need to be diagnosed in all (final) cases. 966 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { 967 if (std::get<2>(LocSym)->isUndefined()) { 968 // Reset the state of any "# line file" directives we've seen to the 969 // context as it was at the diagnostic site. 970 CppHashInfo = std::get<1>(LocSym); 971 printError(std::get<0>(LocSym), "directional label undefined"); 972 } 973 } 974 } 975 976 // Finalize the output stream if there are no errors and if the client wants 977 // us to. 978 if (!HadError && !NoFinalize) 979 Out.Finish(); 980 981 return HadError || getContext().hadError(); 982 } 983 984 bool AsmParser::checkForValidSection() { 985 if (!ParsingInlineAsm && !getStreamer().getCurrentSectionOnly()) { 986 Out.InitSections(false); 987 return Error(getTok().getLoc(), 988 "expected section directive before assembly directive"); 989 } 990 return false; 991 } 992 993 /// Throw away the rest of the line for testing purposes. 994 void AsmParser::eatToEndOfStatement() { 995 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 996 Lexer.Lex(); 997 998 // Eat EOL. 999 if (Lexer.is(AsmToken::EndOfStatement)) 1000 Lexer.Lex(); 1001 } 1002 1003 StringRef AsmParser::parseStringToEndOfStatement() { 1004 const char *Start = getTok().getLoc().getPointer(); 1005 1006 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1007 Lexer.Lex(); 1008 1009 const char *End = getTok().getLoc().getPointer(); 1010 return StringRef(Start, End - Start); 1011 } 1012 1013 StringRef AsmParser::parseStringToComma() { 1014 const char *Start = getTok().getLoc().getPointer(); 1015 1016 while (Lexer.isNot(AsmToken::EndOfStatement) && 1017 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof)) 1018 Lexer.Lex(); 1019 1020 const char *End = getTok().getLoc().getPointer(); 1021 return StringRef(Start, End - Start); 1022 } 1023 1024 /// Parse a paren expression and return it. 1025 /// NOTE: This assumes the leading '(' has already been consumed. 1026 /// 1027 /// parenexpr ::= expr) 1028 /// 1029 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1030 if (parseExpression(Res)) 1031 return true; 1032 if (Lexer.isNot(AsmToken::RParen)) 1033 return TokError("expected ')' in parentheses expression"); 1034 EndLoc = Lexer.getTok().getEndLoc(); 1035 Lex(); 1036 return false; 1037 } 1038 1039 /// Parse a bracket expression and return it. 1040 /// NOTE: This assumes the leading '[' has already been consumed. 1041 /// 1042 /// bracketexpr ::= expr] 1043 /// 1044 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1045 if (parseExpression(Res)) 1046 return true; 1047 EndLoc = getTok().getEndLoc(); 1048 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression")) 1049 return true; 1050 return false; 1051 } 1052 1053 /// Parse a primary expression and return it. 1054 /// primaryexpr ::= (parenexpr 1055 /// primaryexpr ::= symbol 1056 /// primaryexpr ::= number 1057 /// primaryexpr ::= '.' 1058 /// primaryexpr ::= ~,+,- primaryexpr 1059 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1060 SMLoc FirstTokenLoc = getLexer().getLoc(); 1061 AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); 1062 switch (FirstTokenKind) { 1063 default: 1064 return TokError("unknown token in expression"); 1065 // If we have an error assume that we've already handled it. 1066 case AsmToken::Error: 1067 return true; 1068 case AsmToken::Exclaim: 1069 Lex(); // Eat the operator. 1070 if (parsePrimaryExpr(Res, EndLoc)) 1071 return true; 1072 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc); 1073 return false; 1074 case AsmToken::Dollar: 1075 case AsmToken::At: 1076 case AsmToken::String: 1077 case AsmToken::Identifier: { 1078 StringRef Identifier; 1079 if (parseIdentifier(Identifier)) { 1080 // We may have failed but $ may be a valid token. 1081 if (getTok().is(AsmToken::Dollar)) { 1082 if (Lexer.getMAI().getDollarIsPC()) { 1083 Lex(); 1084 // This is a '$' reference, which references the current PC. Emit a 1085 // temporary label to the streamer and refer to it. 1086 MCSymbol *Sym = Ctx.createTempSymbol(); 1087 Out.EmitLabel(Sym); 1088 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, 1089 getContext()); 1090 EndLoc = FirstTokenLoc; 1091 return false; 1092 } 1093 return Error(FirstTokenLoc, "invalid token in expression"); 1094 } 1095 } 1096 // Parse symbol variant 1097 std::pair<StringRef, StringRef> Split; 1098 if (!MAI.useParensForSymbolVariant()) { 1099 if (FirstTokenKind == AsmToken::String) { 1100 if (Lexer.is(AsmToken::At)) { 1101 Lex(); // eat @ 1102 SMLoc AtLoc = getLexer().getLoc(); 1103 StringRef VName; 1104 if (parseIdentifier(VName)) 1105 return Error(AtLoc, "expected symbol variant after '@'"); 1106 1107 Split = std::make_pair(Identifier, VName); 1108 } 1109 } else { 1110 Split = Identifier.split('@'); 1111 } 1112 } else if (Lexer.is(AsmToken::LParen)) { 1113 Lex(); // eat '('. 1114 StringRef VName; 1115 parseIdentifier(VName); 1116 // eat ')'. 1117 if (parseToken(AsmToken::RParen, 1118 "unexpected token in variant, expected ')'")) 1119 return true; 1120 Split = std::make_pair(Identifier, VName); 1121 } 1122 1123 EndLoc = SMLoc::getFromPointer(Identifier.end()); 1124 1125 // This is a symbol reference. 1126 StringRef SymbolName = Identifier; 1127 if (SymbolName.empty()) 1128 return Error(getLexer().getLoc(), "expected a symbol reference"); 1129 1130 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1131 1132 // Lookup the symbol variant if used. 1133 if (!Split.second.empty()) { 1134 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1135 if (Variant != MCSymbolRefExpr::VK_Invalid) { 1136 SymbolName = Split.first; 1137 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { 1138 Variant = MCSymbolRefExpr::VK_None; 1139 } else { 1140 return Error(SMLoc::getFromPointer(Split.second.begin()), 1141 "invalid variant '" + Split.second + "'"); 1142 } 1143 } 1144 1145 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName); 1146 if (!Sym) 1147 Sym = getContext().getOrCreateSymbol(SymbolName); 1148 1149 // If this is an absolute variable reference, substitute it now to preserve 1150 // semantics in the face of reassignment. 1151 if (Sym->isVariable()) { 1152 auto V = Sym->getVariableValue(/*SetUsed*/ false); 1153 bool DoInline = isa<MCConstantExpr>(V) && !Variant; 1154 if (auto TV = dyn_cast<MCTargetExpr>(V)) 1155 DoInline = TV->inlineAssignedExpr(); 1156 if (DoInline) { 1157 if (Variant) 1158 return Error(EndLoc, "unexpected modifier on variable reference"); 1159 Res = Sym->getVariableValue(/*SetUsed*/ false); 1160 return false; 1161 } 1162 } 1163 1164 // Otherwise create a symbol ref. 1165 Res = MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc); 1166 return false; 1167 } 1168 case AsmToken::BigNum: 1169 return TokError("literal value out of range for directive"); 1170 case AsmToken::Integer: { 1171 SMLoc Loc = getTok().getLoc(); 1172 int64_t IntVal = getTok().getIntVal(); 1173 Res = MCConstantExpr::create(IntVal, getContext()); 1174 EndLoc = Lexer.getTok().getEndLoc(); 1175 Lex(); // Eat token. 1176 // Look for 'b' or 'f' following an Integer as a directional label 1177 if (Lexer.getKind() == AsmToken::Identifier) { 1178 StringRef IDVal = getTok().getString(); 1179 // Lookup the symbol variant if used. 1180 std::pair<StringRef, StringRef> Split = IDVal.split('@'); 1181 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1182 if (Split.first.size() != IDVal.size()) { 1183 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1184 if (Variant == MCSymbolRefExpr::VK_Invalid) 1185 return TokError("invalid variant '" + Split.second + "'"); 1186 IDVal = Split.first; 1187 } 1188 if (IDVal == "f" || IDVal == "b") { 1189 MCSymbol *Sym = 1190 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b"); 1191 Res = MCSymbolRefExpr::create(Sym, Variant, getContext()); 1192 if (IDVal == "b" && Sym->isUndefined()) 1193 return Error(Loc, "directional label undefined"); 1194 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym)); 1195 EndLoc = Lexer.getTok().getEndLoc(); 1196 Lex(); // Eat identifier. 1197 } 1198 } 1199 return false; 1200 } 1201 case AsmToken::Real: { 1202 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); 1203 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 1204 Res = MCConstantExpr::create(IntVal, getContext()); 1205 EndLoc = Lexer.getTok().getEndLoc(); 1206 Lex(); // Eat token. 1207 return false; 1208 } 1209 case AsmToken::Dot: { 1210 // This is a '.' reference, which references the current PC. Emit a 1211 // temporary label to the streamer and refer to it. 1212 MCSymbol *Sym = Ctx.createTempSymbol(); 1213 Out.EmitLabel(Sym); 1214 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext()); 1215 EndLoc = Lexer.getTok().getEndLoc(); 1216 Lex(); // Eat identifier. 1217 return false; 1218 } 1219 case AsmToken::LParen: 1220 Lex(); // Eat the '('. 1221 return parseParenExpr(Res, EndLoc); 1222 case AsmToken::LBrac: 1223 if (!PlatformParser->HasBracketExpressions()) 1224 return TokError("brackets expression not supported on this target"); 1225 Lex(); // Eat the '['. 1226 return parseBracketExpr(Res, EndLoc); 1227 case AsmToken::Minus: 1228 Lex(); // Eat the operator. 1229 if (parsePrimaryExpr(Res, EndLoc)) 1230 return true; 1231 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc); 1232 return false; 1233 case AsmToken::Plus: 1234 Lex(); // Eat the operator. 1235 if (parsePrimaryExpr(Res, EndLoc)) 1236 return true; 1237 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc); 1238 return false; 1239 case AsmToken::Tilde: 1240 Lex(); // Eat the operator. 1241 if (parsePrimaryExpr(Res, EndLoc)) 1242 return true; 1243 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1244 return false; 1245 // MIPS unary expression operators. The lexer won't generate these tokens if 1246 // MCAsmInfo::HasMipsExpressions is false for the target. 1247 case AsmToken::PercentCall16: 1248 case AsmToken::PercentCall_Hi: 1249 case AsmToken::PercentCall_Lo: 1250 case AsmToken::PercentDtprel_Hi: 1251 case AsmToken::PercentDtprel_Lo: 1252 case AsmToken::PercentGot: 1253 case AsmToken::PercentGot_Disp: 1254 case AsmToken::PercentGot_Hi: 1255 case AsmToken::PercentGot_Lo: 1256 case AsmToken::PercentGot_Ofst: 1257 case AsmToken::PercentGot_Page: 1258 case AsmToken::PercentGottprel: 1259 case AsmToken::PercentGp_Rel: 1260 case AsmToken::PercentHi: 1261 case AsmToken::PercentHigher: 1262 case AsmToken::PercentHighest: 1263 case AsmToken::PercentLo: 1264 case AsmToken::PercentNeg: 1265 case AsmToken::PercentPcrel_Hi: 1266 case AsmToken::PercentPcrel_Lo: 1267 case AsmToken::PercentTlsgd: 1268 case AsmToken::PercentTlsldm: 1269 case AsmToken::PercentTprel_Hi: 1270 case AsmToken::PercentTprel_Lo: 1271 Lex(); // Eat the operator. 1272 if (Lexer.isNot(AsmToken::LParen)) 1273 return TokError("expected '(' after operator"); 1274 Lex(); // Eat the operator. 1275 if (parseExpression(Res, EndLoc)) 1276 return true; 1277 if (Lexer.isNot(AsmToken::RParen)) 1278 return TokError("expected ')'"); 1279 Lex(); // Eat the operator. 1280 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx); 1281 return !Res; 1282 } 1283 } 1284 1285 bool AsmParser::parseExpression(const MCExpr *&Res) { 1286 SMLoc EndLoc; 1287 return parseExpression(Res, EndLoc); 1288 } 1289 1290 const MCExpr * 1291 AsmParser::applyModifierToExpr(const MCExpr *E, 1292 MCSymbolRefExpr::VariantKind Variant) { 1293 // Ask the target implementation about this expression first. 1294 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx); 1295 if (NewE) 1296 return NewE; 1297 // Recurse over the given expression, rebuilding it to apply the given variant 1298 // if there is exactly one symbol. 1299 switch (E->getKind()) { 1300 case MCExpr::Target: 1301 case MCExpr::Constant: 1302 return nullptr; 1303 1304 case MCExpr::SymbolRef: { 1305 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 1306 1307 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 1308 TokError("invalid variant on expression '" + getTok().getIdentifier() + 1309 "' (already modified)"); 1310 return E; 1311 } 1312 1313 return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext()); 1314 } 1315 1316 case MCExpr::Unary: { 1317 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 1318 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant); 1319 if (!Sub) 1320 return nullptr; 1321 return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext()); 1322 } 1323 1324 case MCExpr::Binary: { 1325 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 1326 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant); 1327 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant); 1328 1329 if (!LHS && !RHS) 1330 return nullptr; 1331 1332 if (!LHS) 1333 LHS = BE->getLHS(); 1334 if (!RHS) 1335 RHS = BE->getRHS(); 1336 1337 return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext()); 1338 } 1339 } 1340 1341 llvm_unreachable("Invalid expression kind!"); 1342 } 1343 1344 /// This function checks if the next token is <string> type or arithmetic. 1345 /// string that begin with character '<' must end with character '>'. 1346 /// otherwise it is arithmetics. 1347 /// If the function returns a 'true' value, 1348 /// the End argument will be filled with the last location pointed to the '>' 1349 /// character. 1350 1351 /// There is a gap between the AltMacro's documentation and the single quote 1352 /// implementation. GCC does not fully support this feature and so we will not 1353 /// support it. 1354 /// TODO: Adding single quote as a string. 1355 static bool isAltmacroString(SMLoc &StrLoc, SMLoc &EndLoc) { 1356 assert((StrLoc.getPointer() != nullptr) && 1357 "Argument to the function cannot be a NULL value"); 1358 const char *CharPtr = StrLoc.getPointer(); 1359 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && 1360 (*CharPtr != '\0')) { 1361 if (*CharPtr == '!') 1362 CharPtr++; 1363 CharPtr++; 1364 } 1365 if (*CharPtr == '>') { 1366 EndLoc = StrLoc.getFromPointer(CharPtr + 1); 1367 return true; 1368 } 1369 return false; 1370 } 1371 1372 /// creating a string without the escape characters '!'. 1373 static std::string altMacroString(StringRef AltMacroStr) { 1374 std::string Res; 1375 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) { 1376 if (AltMacroStr[Pos] == '!') 1377 Pos++; 1378 Res += AltMacroStr[Pos]; 1379 } 1380 return Res; 1381 } 1382 1383 /// Parse an expression and return it. 1384 /// 1385 /// expr ::= expr &&,|| expr -> lowest. 1386 /// expr ::= expr |,^,&,! expr 1387 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 1388 /// expr ::= expr <<,>> expr 1389 /// expr ::= expr +,- expr 1390 /// expr ::= expr *,/,% expr -> highest. 1391 /// expr ::= primaryexpr 1392 /// 1393 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1394 // Parse the expression. 1395 Res = nullptr; 1396 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || 1397 parseBinOpRHS(1, Res, EndLoc)) 1398 return true; 1399 1400 // As a special case, we support 'a op b @ modifier' by rewriting the 1401 // expression to include the modifier. This is inefficient, but in general we 1402 // expect users to use 'a@modifier op b'. 1403 if (Lexer.getKind() == AsmToken::At) { 1404 Lex(); 1405 1406 if (Lexer.isNot(AsmToken::Identifier)) 1407 return TokError("unexpected symbol modifier following '@'"); 1408 1409 MCSymbolRefExpr::VariantKind Variant = 1410 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 1411 if (Variant == MCSymbolRefExpr::VK_Invalid) 1412 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 1413 1414 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant); 1415 if (!ModifiedRes) { 1416 return TokError("invalid modifier '" + getTok().getIdentifier() + 1417 "' (no symbols present)"); 1418 } 1419 1420 Res = ModifiedRes; 1421 Lex(); 1422 } 1423 1424 // Try to constant fold it up front, if possible. Do not exploit 1425 // assembler here. 1426 int64_t Value; 1427 if (Res->evaluateAsAbsolute(Value)) 1428 Res = MCConstantExpr::create(Value, getContext()); 1429 1430 return false; 1431 } 1432 1433 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1434 Res = nullptr; 1435 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc); 1436 } 1437 1438 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 1439 SMLoc &EndLoc) { 1440 if (parseParenExpr(Res, EndLoc)) 1441 return true; 1442 1443 for (; ParenDepth > 0; --ParenDepth) { 1444 if (parseBinOpRHS(1, Res, EndLoc)) 1445 return true; 1446 1447 // We don't Lex() the last RParen. 1448 // This is the same behavior as parseParenExpression(). 1449 if (ParenDepth - 1 > 0) { 1450 EndLoc = getTok().getEndLoc(); 1451 if (parseToken(AsmToken::RParen, 1452 "expected ')' in parentheses expression")) 1453 return true; 1454 } 1455 } 1456 return false; 1457 } 1458 1459 bool AsmParser::parseAbsoluteExpression(int64_t &Res) { 1460 const MCExpr *Expr; 1461 1462 SMLoc StartLoc = Lexer.getLoc(); 1463 if (parseExpression(Expr)) 1464 return true; 1465 1466 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 1467 return Error(StartLoc, "expected absolute expression"); 1468 1469 return false; 1470 } 1471 1472 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K, 1473 MCBinaryExpr::Opcode &Kind, 1474 bool ShouldUseLogicalShr) { 1475 switch (K) { 1476 default: 1477 return 0; // not a binop. 1478 1479 // Lowest Precedence: &&, || 1480 case AsmToken::AmpAmp: 1481 Kind = MCBinaryExpr::LAnd; 1482 return 1; 1483 case AsmToken::PipePipe: 1484 Kind = MCBinaryExpr::LOr; 1485 return 1; 1486 1487 // Low Precedence: |, &, ^ 1488 // 1489 // FIXME: gas seems to support '!' as an infix operator? 1490 case AsmToken::Pipe: 1491 Kind = MCBinaryExpr::Or; 1492 return 2; 1493 case AsmToken::Caret: 1494 Kind = MCBinaryExpr::Xor; 1495 return 2; 1496 case AsmToken::Amp: 1497 Kind = MCBinaryExpr::And; 1498 return 2; 1499 1500 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 1501 case AsmToken::EqualEqual: 1502 Kind = MCBinaryExpr::EQ; 1503 return 3; 1504 case AsmToken::ExclaimEqual: 1505 case AsmToken::LessGreater: 1506 Kind = MCBinaryExpr::NE; 1507 return 3; 1508 case AsmToken::Less: 1509 Kind = MCBinaryExpr::LT; 1510 return 3; 1511 case AsmToken::LessEqual: 1512 Kind = MCBinaryExpr::LTE; 1513 return 3; 1514 case AsmToken::Greater: 1515 Kind = MCBinaryExpr::GT; 1516 return 3; 1517 case AsmToken::GreaterEqual: 1518 Kind = MCBinaryExpr::GTE; 1519 return 3; 1520 1521 // Intermediate Precedence: <<, >> 1522 case AsmToken::LessLess: 1523 Kind = MCBinaryExpr::Shl; 1524 return 4; 1525 case AsmToken::GreaterGreater: 1526 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1527 return 4; 1528 1529 // High Intermediate Precedence: +, - 1530 case AsmToken::Plus: 1531 Kind = MCBinaryExpr::Add; 1532 return 5; 1533 case AsmToken::Minus: 1534 Kind = MCBinaryExpr::Sub; 1535 return 5; 1536 1537 // Highest Precedence: *, /, % 1538 case AsmToken::Star: 1539 Kind = MCBinaryExpr::Mul; 1540 return 6; 1541 case AsmToken::Slash: 1542 Kind = MCBinaryExpr::Div; 1543 return 6; 1544 case AsmToken::Percent: 1545 Kind = MCBinaryExpr::Mod; 1546 return 6; 1547 } 1548 } 1549 1550 static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K, 1551 MCBinaryExpr::Opcode &Kind, 1552 bool ShouldUseLogicalShr) { 1553 switch (K) { 1554 default: 1555 return 0; // not a binop. 1556 1557 // Lowest Precedence: &&, || 1558 case AsmToken::AmpAmp: 1559 Kind = MCBinaryExpr::LAnd; 1560 return 2; 1561 case AsmToken::PipePipe: 1562 Kind = MCBinaryExpr::LOr; 1563 return 1; 1564 1565 // Low Precedence: ==, !=, <>, <, <=, >, >= 1566 case AsmToken::EqualEqual: 1567 Kind = MCBinaryExpr::EQ; 1568 return 3; 1569 case AsmToken::ExclaimEqual: 1570 case AsmToken::LessGreater: 1571 Kind = MCBinaryExpr::NE; 1572 return 3; 1573 case AsmToken::Less: 1574 Kind = MCBinaryExpr::LT; 1575 return 3; 1576 case AsmToken::LessEqual: 1577 Kind = MCBinaryExpr::LTE; 1578 return 3; 1579 case AsmToken::Greater: 1580 Kind = MCBinaryExpr::GT; 1581 return 3; 1582 case AsmToken::GreaterEqual: 1583 Kind = MCBinaryExpr::GTE; 1584 return 3; 1585 1586 // Low Intermediate Precedence: +, - 1587 case AsmToken::Plus: 1588 Kind = MCBinaryExpr::Add; 1589 return 4; 1590 case AsmToken::Minus: 1591 Kind = MCBinaryExpr::Sub; 1592 return 4; 1593 1594 // High Intermediate Precedence: |, &, ^ 1595 // 1596 // FIXME: gas seems to support '!' as an infix operator? 1597 case AsmToken::Pipe: 1598 Kind = MCBinaryExpr::Or; 1599 return 5; 1600 case AsmToken::Caret: 1601 Kind = MCBinaryExpr::Xor; 1602 return 5; 1603 case AsmToken::Amp: 1604 Kind = MCBinaryExpr::And; 1605 return 5; 1606 1607 // Highest Precedence: *, /, %, <<, >> 1608 case AsmToken::Star: 1609 Kind = MCBinaryExpr::Mul; 1610 return 6; 1611 case AsmToken::Slash: 1612 Kind = MCBinaryExpr::Div; 1613 return 6; 1614 case AsmToken::Percent: 1615 Kind = MCBinaryExpr::Mod; 1616 return 6; 1617 case AsmToken::LessLess: 1618 Kind = MCBinaryExpr::Shl; 1619 return 6; 1620 case AsmToken::GreaterGreater: 1621 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1622 return 6; 1623 } 1624 } 1625 1626 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K, 1627 MCBinaryExpr::Opcode &Kind) { 1628 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); 1629 return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr) 1630 : getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr); 1631 } 1632 1633 /// Parse all binary operators with precedence >= 'Precedence'. 1634 /// Res contains the LHS of the expression on input. 1635 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 1636 SMLoc &EndLoc) { 1637 SMLoc StartLoc = Lexer.getLoc(); 1638 while (true) { 1639 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 1640 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 1641 1642 // If the next token is lower precedence than we are allowed to eat, return 1643 // successfully with what we ate already. 1644 if (TokPrec < Precedence) 1645 return false; 1646 1647 Lex(); 1648 1649 // Eat the next primary expression. 1650 const MCExpr *RHS; 1651 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc)) 1652 return true; 1653 1654 // If BinOp binds less tightly with RHS than the operator after RHS, let 1655 // the pending operator take RHS as its LHS. 1656 MCBinaryExpr::Opcode Dummy; 1657 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1658 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc)) 1659 return true; 1660 1661 // Merge LHS and RHS according to operator. 1662 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc); 1663 } 1664 } 1665 1666 /// ParseStatement: 1667 /// ::= EndOfStatement 1668 /// ::= Label* Directive ...Operands... EndOfStatement 1669 /// ::= Label* Identifier OperandList* EndOfStatement 1670 bool AsmParser::parseStatement(ParseStatementInfo &Info, 1671 MCAsmParserSemaCallback *SI) { 1672 assert(!hasPendingError() && "parseStatement started with pending error"); 1673 // Eat initial spaces and comments 1674 while (Lexer.is(AsmToken::Space)) 1675 Lex(); 1676 if (Lexer.is(AsmToken::EndOfStatement)) { 1677 // if this is a line comment we can drop it safely 1678 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 1679 getTok().getString().front() == '\n') 1680 Out.AddBlankLine(); 1681 Lex(); 1682 return false; 1683 } 1684 // Statements always start with an identifier. 1685 AsmToken ID = getTok(); 1686 SMLoc IDLoc = ID.getLoc(); 1687 StringRef IDVal; 1688 int64_t LocalLabelVal = -1; 1689 if (Lexer.is(AsmToken::HashDirective)) 1690 return parseCppHashLineFilenameComment(IDLoc); 1691 // Allow an integer followed by a ':' as a directional local label. 1692 if (Lexer.is(AsmToken::Integer)) { 1693 LocalLabelVal = getTok().getIntVal(); 1694 if (LocalLabelVal < 0) { 1695 if (!TheCondState.Ignore) { 1696 Lex(); // always eat a token 1697 return Error(IDLoc, "unexpected token at start of statement"); 1698 } 1699 IDVal = ""; 1700 } else { 1701 IDVal = getTok().getString(); 1702 Lex(); // Consume the integer token to be used as an identifier token. 1703 if (Lexer.getKind() != AsmToken::Colon) { 1704 if (!TheCondState.Ignore) { 1705 Lex(); // always eat a token 1706 return Error(IDLoc, "unexpected token at start of statement"); 1707 } 1708 } 1709 } 1710 } else if (Lexer.is(AsmToken::Dot)) { 1711 // Treat '.' as a valid identifier in this context. 1712 Lex(); 1713 IDVal = "."; 1714 } else if (Lexer.is(AsmToken::LCurly)) { 1715 // Treat '{' as a valid identifier in this context. 1716 Lex(); 1717 IDVal = "{"; 1718 1719 } else if (Lexer.is(AsmToken::RCurly)) { 1720 // Treat '}' as a valid identifier in this context. 1721 Lex(); 1722 IDVal = "}"; 1723 } else if (Lexer.is(AsmToken::Star) && 1724 getTargetParser().starIsStartOfStatement()) { 1725 // Accept '*' as a valid start of statement. 1726 Lex(); 1727 IDVal = "*"; 1728 } else if (parseIdentifier(IDVal)) { 1729 if (!TheCondState.Ignore) { 1730 Lex(); // always eat a token 1731 return Error(IDLoc, "unexpected token at start of statement"); 1732 } 1733 IDVal = ""; 1734 } 1735 1736 // Handle conditional assembly here before checking for skipping. We 1737 // have to do this so that .endif isn't skipped in a ".if 0" block for 1738 // example. 1739 StringMap<DirectiveKind>::const_iterator DirKindIt = 1740 DirectiveKindMap.find(IDVal); 1741 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) 1742 ? DK_NO_DIRECTIVE 1743 : DirKindIt->getValue(); 1744 switch (DirKind) { 1745 default: 1746 break; 1747 case DK_IF: 1748 case DK_IFEQ: 1749 case DK_IFGE: 1750 case DK_IFGT: 1751 case DK_IFLE: 1752 case DK_IFLT: 1753 case DK_IFNE: 1754 return parseDirectiveIf(IDLoc, DirKind); 1755 case DK_IFB: 1756 return parseDirectiveIfb(IDLoc, true); 1757 case DK_IFNB: 1758 return parseDirectiveIfb(IDLoc, false); 1759 case DK_IFC: 1760 return parseDirectiveIfc(IDLoc, true); 1761 case DK_IFEQS: 1762 return parseDirectiveIfeqs(IDLoc, true); 1763 case DK_IFNC: 1764 return parseDirectiveIfc(IDLoc, false); 1765 case DK_IFNES: 1766 return parseDirectiveIfeqs(IDLoc, false); 1767 case DK_IFDEF: 1768 return parseDirectiveIfdef(IDLoc, true); 1769 case DK_IFNDEF: 1770 case DK_IFNOTDEF: 1771 return parseDirectiveIfdef(IDLoc, false); 1772 case DK_ELSEIF: 1773 return parseDirectiveElseIf(IDLoc); 1774 case DK_ELSE: 1775 return parseDirectiveElse(IDLoc); 1776 case DK_ENDIF: 1777 return parseDirectiveEndIf(IDLoc); 1778 } 1779 1780 // Ignore the statement if in the middle of inactive conditional 1781 // (e.g. ".if 0"). 1782 if (TheCondState.Ignore) { 1783 eatToEndOfStatement(); 1784 return false; 1785 } 1786 1787 // FIXME: Recurse on local labels? 1788 1789 // See what kind of statement we have. 1790 switch (Lexer.getKind()) { 1791 case AsmToken::Colon: { 1792 if (!getTargetParser().isLabel(ID)) 1793 break; 1794 if (checkForValidSection()) 1795 return true; 1796 1797 // identifier ':' -> Label. 1798 Lex(); 1799 1800 // Diagnose attempt to use '.' as a label. 1801 if (IDVal == ".") 1802 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1803 1804 // Diagnose attempt to use a variable as a label. 1805 // 1806 // FIXME: Diagnostics. Note the location of the definition as a label. 1807 // FIXME: This doesn't diagnose assignment to a symbol which has been 1808 // implicitly marked as external. 1809 MCSymbol *Sym; 1810 if (LocalLabelVal == -1) { 1811 if (ParsingInlineAsm && SI) { 1812 StringRef RewrittenLabel = 1813 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true); 1814 assert(!RewrittenLabel.empty() && 1815 "We should have an internal name here."); 1816 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(), 1817 RewrittenLabel); 1818 IDVal = RewrittenLabel; 1819 } 1820 Sym = getContext().getOrCreateSymbol(IDVal); 1821 } else 1822 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal); 1823 // End of Labels should be treated as end of line for lexing 1824 // purposes but that information is not available to the Lexer who 1825 // does not understand Labels. This may cause us to see a Hash 1826 // here instead of a preprocessor line comment. 1827 if (getTok().is(AsmToken::Hash)) { 1828 StringRef CommentStr = parseStringToEndOfStatement(); 1829 Lexer.Lex(); 1830 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr)); 1831 } 1832 1833 // Consume any end of statement token, if present, to avoid spurious 1834 // AddBlankLine calls(). 1835 if (getTok().is(AsmToken::EndOfStatement)) { 1836 Lex(); 1837 } 1838 1839 getTargetParser().doBeforeLabelEmit(Sym); 1840 1841 // Emit the label. 1842 if (!getTargetParser().isParsingInlineAsm()) 1843 Out.EmitLabel(Sym, IDLoc); 1844 1845 // If we are generating dwarf for assembly source files then gather the 1846 // info to make a dwarf label entry for this label if needed. 1847 if (enabledGenDwarfForAssembly()) 1848 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 1849 IDLoc); 1850 1851 getTargetParser().onLabelParsed(Sym); 1852 1853 return false; 1854 } 1855 1856 case AsmToken::Equal: 1857 if (!getTargetParser().equalIsAsmAssignment()) 1858 break; 1859 // identifier '=' ... -> assignment statement 1860 Lex(); 1861 1862 return parseAssignment(IDVal, true); 1863 1864 default: // Normal instruction or directive. 1865 break; 1866 } 1867 1868 // If macros are enabled, check to see if this is a macro instantiation. 1869 if (areMacrosEnabled()) 1870 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) { 1871 return handleMacroEntry(M, IDLoc); 1872 } 1873 1874 // Otherwise, we have a normal instruction or directive. 1875 1876 // Directives start with "." 1877 if (IDVal.startswith(".") && IDVal != ".") { 1878 // There are several entities interested in parsing directives: 1879 // 1880 // 1. The target-specific assembly parser. Some directives are target 1881 // specific or may potentially behave differently on certain targets. 1882 // 2. Asm parser extensions. For example, platform-specific parsers 1883 // (like the ELF parser) register themselves as extensions. 1884 // 3. The generic directive parser implemented by this class. These are 1885 // all the directives that behave in a target and platform independent 1886 // manner, or at least have a default behavior that's shared between 1887 // all targets and platforms. 1888 1889 getTargetParser().flushPendingInstructions(getStreamer()); 1890 1891 SMLoc StartTokLoc = getTok().getLoc(); 1892 bool TPDirectiveReturn = getTargetParser().ParseDirective(ID); 1893 1894 if (hasPendingError()) 1895 return true; 1896 // Currently the return value should be true if we are 1897 // uninterested but as this is at odds with the standard parsing 1898 // convention (return true = error) we have instances of a parsed 1899 // directive that fails returning true as an error. Catch these 1900 // cases as best as possible errors here. 1901 if (TPDirectiveReturn && StartTokLoc != getTok().getLoc()) 1902 return true; 1903 // Return if we did some parsing or believe we succeeded. 1904 if (!TPDirectiveReturn || StartTokLoc != getTok().getLoc()) 1905 return false; 1906 1907 // Next, check the extension directive map to see if any extension has 1908 // registered itself to parse this directive. 1909 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 1910 ExtensionDirectiveMap.lookup(IDVal); 1911 if (Handler.first) 1912 return (*Handler.second)(Handler.first, IDVal, IDLoc); 1913 1914 // Finally, if no one else is interested in this directive, it must be 1915 // generic and familiar to this class. 1916 switch (DirKind) { 1917 default: 1918 break; 1919 case DK_SET: 1920 case DK_EQU: 1921 return parseDirectiveSet(IDVal, true); 1922 case DK_EQUIV: 1923 return parseDirectiveSet(IDVal, false); 1924 case DK_ASCII: 1925 return parseDirectiveAscii(IDVal, false); 1926 case DK_ASCIZ: 1927 case DK_STRING: 1928 return parseDirectiveAscii(IDVal, true); 1929 case DK_BYTE: 1930 case DK_DC_B: 1931 return parseDirectiveValue(IDVal, 1); 1932 case DK_DC: 1933 case DK_DC_W: 1934 case DK_SHORT: 1935 case DK_VALUE: 1936 case DK_2BYTE: 1937 return parseDirectiveValue(IDVal, 2); 1938 case DK_LONG: 1939 case DK_INT: 1940 case DK_4BYTE: 1941 case DK_DC_L: 1942 return parseDirectiveValue(IDVal, 4); 1943 case DK_QUAD: 1944 case DK_8BYTE: 1945 return parseDirectiveValue(IDVal, 8); 1946 case DK_DC_A: 1947 return parseDirectiveValue( 1948 IDVal, getContext().getAsmInfo()->getCodePointerSize()); 1949 case DK_OCTA: 1950 return parseDirectiveOctaValue(IDVal); 1951 case DK_SINGLE: 1952 case DK_FLOAT: 1953 case DK_DC_S: 1954 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle()); 1955 case DK_DOUBLE: 1956 case DK_DC_D: 1957 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble()); 1958 case DK_ALIGN: { 1959 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 1960 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1); 1961 } 1962 case DK_ALIGN32: { 1963 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 1964 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4); 1965 } 1966 case DK_BALIGN: 1967 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 1968 case DK_BALIGNW: 1969 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 1970 case DK_BALIGNL: 1971 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 1972 case DK_P2ALIGN: 1973 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 1974 case DK_P2ALIGNW: 1975 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 1976 case DK_P2ALIGNL: 1977 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 1978 case DK_ORG: 1979 return parseDirectiveOrg(); 1980 case DK_FILL: 1981 return parseDirectiveFill(); 1982 case DK_ZERO: 1983 return parseDirectiveZero(); 1984 case DK_EXTERN: 1985 eatToEndOfStatement(); // .extern is the default, ignore it. 1986 return false; 1987 case DK_GLOBL: 1988 case DK_GLOBAL: 1989 return parseDirectiveSymbolAttribute(MCSA_Global); 1990 case DK_LAZY_REFERENCE: 1991 return parseDirectiveSymbolAttribute(MCSA_LazyReference); 1992 case DK_NO_DEAD_STRIP: 1993 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 1994 case DK_SYMBOL_RESOLVER: 1995 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver); 1996 case DK_PRIVATE_EXTERN: 1997 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern); 1998 case DK_REFERENCE: 1999 return parseDirectiveSymbolAttribute(MCSA_Reference); 2000 case DK_WEAK_DEFINITION: 2001 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition); 2002 case DK_WEAK_REFERENCE: 2003 return parseDirectiveSymbolAttribute(MCSA_WeakReference); 2004 case DK_WEAK_DEF_CAN_BE_HIDDEN: 2005 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 2006 case DK_COLD: 2007 return parseDirectiveSymbolAttribute(MCSA_Cold); 2008 case DK_COMM: 2009 case DK_COMMON: 2010 return parseDirectiveComm(/*IsLocal=*/false); 2011 case DK_LCOMM: 2012 return parseDirectiveComm(/*IsLocal=*/true); 2013 case DK_ABORT: 2014 return parseDirectiveAbort(); 2015 case DK_INCLUDE: 2016 return parseDirectiveInclude(); 2017 case DK_INCBIN: 2018 return parseDirectiveIncbin(); 2019 case DK_CODE16: 2020 case DK_CODE16GCC: 2021 return TokError(Twine(IDVal) + 2022 " not currently supported for this target"); 2023 case DK_REPT: 2024 return parseDirectiveRept(IDLoc, IDVal); 2025 case DK_IRP: 2026 return parseDirectiveIrp(IDLoc); 2027 case DK_IRPC: 2028 return parseDirectiveIrpc(IDLoc); 2029 case DK_ENDR: 2030 return parseDirectiveEndr(IDLoc); 2031 case DK_BUNDLE_ALIGN_MODE: 2032 return parseDirectiveBundleAlignMode(); 2033 case DK_BUNDLE_LOCK: 2034 return parseDirectiveBundleLock(); 2035 case DK_BUNDLE_UNLOCK: 2036 return parseDirectiveBundleUnlock(); 2037 case DK_SLEB128: 2038 return parseDirectiveLEB128(true); 2039 case DK_ULEB128: 2040 return parseDirectiveLEB128(false); 2041 case DK_SPACE: 2042 case DK_SKIP: 2043 return parseDirectiveSpace(IDVal); 2044 case DK_FILE: 2045 return parseDirectiveFile(IDLoc); 2046 case DK_LINE: 2047 return parseDirectiveLine(); 2048 case DK_LOC: 2049 return parseDirectiveLoc(); 2050 case DK_STABS: 2051 return parseDirectiveStabs(); 2052 case DK_CV_FILE: 2053 return parseDirectiveCVFile(); 2054 case DK_CV_FUNC_ID: 2055 return parseDirectiveCVFuncId(); 2056 case DK_CV_INLINE_SITE_ID: 2057 return parseDirectiveCVInlineSiteId(); 2058 case DK_CV_LOC: 2059 return parseDirectiveCVLoc(); 2060 case DK_CV_LINETABLE: 2061 return parseDirectiveCVLinetable(); 2062 case DK_CV_INLINE_LINETABLE: 2063 return parseDirectiveCVInlineLinetable(); 2064 case DK_CV_DEF_RANGE: 2065 return parseDirectiveCVDefRange(); 2066 case DK_CV_STRING: 2067 return parseDirectiveCVString(); 2068 case DK_CV_STRINGTABLE: 2069 return parseDirectiveCVStringTable(); 2070 case DK_CV_FILECHECKSUMS: 2071 return parseDirectiveCVFileChecksums(); 2072 case DK_CV_FILECHECKSUM_OFFSET: 2073 return parseDirectiveCVFileChecksumOffset(); 2074 case DK_CV_FPO_DATA: 2075 return parseDirectiveCVFPOData(); 2076 case DK_CFI_SECTIONS: 2077 return parseDirectiveCFISections(); 2078 case DK_CFI_STARTPROC: 2079 return parseDirectiveCFIStartProc(); 2080 case DK_CFI_ENDPROC: 2081 return parseDirectiveCFIEndProc(); 2082 case DK_CFI_DEF_CFA: 2083 return parseDirectiveCFIDefCfa(IDLoc); 2084 case DK_CFI_DEF_CFA_OFFSET: 2085 return parseDirectiveCFIDefCfaOffset(); 2086 case DK_CFI_ADJUST_CFA_OFFSET: 2087 return parseDirectiveCFIAdjustCfaOffset(); 2088 case DK_CFI_DEF_CFA_REGISTER: 2089 return parseDirectiveCFIDefCfaRegister(IDLoc); 2090 case DK_CFI_OFFSET: 2091 return parseDirectiveCFIOffset(IDLoc); 2092 case DK_CFI_REL_OFFSET: 2093 return parseDirectiveCFIRelOffset(IDLoc); 2094 case DK_CFI_PERSONALITY: 2095 return parseDirectiveCFIPersonalityOrLsda(true); 2096 case DK_CFI_LSDA: 2097 return parseDirectiveCFIPersonalityOrLsda(false); 2098 case DK_CFI_REMEMBER_STATE: 2099 return parseDirectiveCFIRememberState(); 2100 case DK_CFI_RESTORE_STATE: 2101 return parseDirectiveCFIRestoreState(); 2102 case DK_CFI_SAME_VALUE: 2103 return parseDirectiveCFISameValue(IDLoc); 2104 case DK_CFI_RESTORE: 2105 return parseDirectiveCFIRestore(IDLoc); 2106 case DK_CFI_ESCAPE: 2107 return parseDirectiveCFIEscape(); 2108 case DK_CFI_RETURN_COLUMN: 2109 return parseDirectiveCFIReturnColumn(IDLoc); 2110 case DK_CFI_SIGNAL_FRAME: 2111 return parseDirectiveCFISignalFrame(); 2112 case DK_CFI_UNDEFINED: 2113 return parseDirectiveCFIUndefined(IDLoc); 2114 case DK_CFI_REGISTER: 2115 return parseDirectiveCFIRegister(IDLoc); 2116 case DK_CFI_WINDOW_SAVE: 2117 return parseDirectiveCFIWindowSave(); 2118 case DK_MACROS_ON: 2119 case DK_MACROS_OFF: 2120 return parseDirectiveMacrosOnOff(IDVal); 2121 case DK_MACRO: 2122 return parseDirectiveMacro(IDLoc); 2123 case DK_ALTMACRO: 2124 case DK_NOALTMACRO: 2125 return parseDirectiveAltmacro(IDVal); 2126 case DK_EXITM: 2127 return parseDirectiveExitMacro(IDVal); 2128 case DK_ENDM: 2129 case DK_ENDMACRO: 2130 return parseDirectiveEndMacro(IDVal); 2131 case DK_PURGEM: 2132 return parseDirectivePurgeMacro(IDLoc); 2133 case DK_END: 2134 return parseDirectiveEnd(IDLoc); 2135 case DK_ERR: 2136 return parseDirectiveError(IDLoc, false); 2137 case DK_ERROR: 2138 return parseDirectiveError(IDLoc, true); 2139 case DK_WARNING: 2140 return parseDirectiveWarning(IDLoc); 2141 case DK_RELOC: 2142 return parseDirectiveReloc(IDLoc); 2143 case DK_DCB: 2144 case DK_DCB_W: 2145 return parseDirectiveDCB(IDVal, 2); 2146 case DK_DCB_B: 2147 return parseDirectiveDCB(IDVal, 1); 2148 case DK_DCB_D: 2149 return parseDirectiveRealDCB(IDVal, APFloat::IEEEdouble()); 2150 case DK_DCB_L: 2151 return parseDirectiveDCB(IDVal, 4); 2152 case DK_DCB_S: 2153 return parseDirectiveRealDCB(IDVal, APFloat::IEEEsingle()); 2154 case DK_DC_X: 2155 case DK_DCB_X: 2156 return TokError(Twine(IDVal) + 2157 " not currently supported for this target"); 2158 case DK_DS: 2159 case DK_DS_W: 2160 return parseDirectiveDS(IDVal, 2); 2161 case DK_DS_B: 2162 return parseDirectiveDS(IDVal, 1); 2163 case DK_DS_D: 2164 return parseDirectiveDS(IDVal, 8); 2165 case DK_DS_L: 2166 case DK_DS_S: 2167 return parseDirectiveDS(IDVal, 4); 2168 case DK_DS_P: 2169 case DK_DS_X: 2170 return parseDirectiveDS(IDVal, 12); 2171 case DK_PRINT: 2172 return parseDirectivePrint(IDLoc); 2173 case DK_ADDRSIG: 2174 return parseDirectiveAddrsig(); 2175 case DK_ADDRSIG_SYM: 2176 return parseDirectiveAddrsigSym(); 2177 } 2178 2179 return Error(IDLoc, "unknown directive"); 2180 } 2181 2182 // __asm _emit or __asm __emit 2183 if (ParsingInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || 2184 IDVal == "_EMIT" || IDVal == "__EMIT")) 2185 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size()); 2186 2187 // __asm align 2188 if (ParsingInlineAsm && (IDVal == "align" || IDVal == "ALIGN")) 2189 return parseDirectiveMSAlign(IDLoc, Info); 2190 2191 if (ParsingInlineAsm && (IDVal == "even" || IDVal == "EVEN")) 2192 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4); 2193 if (checkForValidSection()) 2194 return true; 2195 2196 // Canonicalize the opcode to lower case. 2197 std::string OpcodeStr = IDVal.lower(); 2198 ParseInstructionInfo IInfo(Info.AsmRewrites); 2199 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID, 2200 Info.ParsedOperands); 2201 Info.ParseError = ParseHadError; 2202 2203 // Dump the parsed representation, if requested. 2204 if (getShowParsedOperands()) { 2205 SmallString<256> Str; 2206 raw_svector_ostream OS(Str); 2207 OS << "parsed instruction: ["; 2208 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { 2209 if (i != 0) 2210 OS << ", "; 2211 Info.ParsedOperands[i]->print(OS); 2212 } 2213 OS << "]"; 2214 2215 printMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 2216 } 2217 2218 // Fail even if ParseInstruction erroneously returns false. 2219 if (hasPendingError() || ParseHadError) 2220 return true; 2221 2222 // If we are generating dwarf for the current section then generate a .loc 2223 // directive for the instruction. 2224 if (!ParseHadError && enabledGenDwarfForAssembly() && 2225 getContext().getGenDwarfSectionSyms().count( 2226 getStreamer().getCurrentSectionOnly())) { 2227 unsigned Line; 2228 if (ActiveMacros.empty()) 2229 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer); 2230 else 2231 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc, 2232 ActiveMacros.front()->ExitBuffer); 2233 2234 // If we previously parsed a cpp hash file line comment then make sure the 2235 // current Dwarf File is for the CppHashFilename if not then emit the 2236 // Dwarf File table for it and adjust the line number for the .loc. 2237 if (!CppHashInfo.Filename.empty()) { 2238 unsigned FileNumber = getStreamer().EmitDwarfFileDirective( 2239 0, StringRef(), CppHashInfo.Filename); 2240 getContext().setGenDwarfFileNumber(FileNumber); 2241 2242 unsigned CppHashLocLineNo = 2243 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf); 2244 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); 2245 } 2246 2247 getStreamer().EmitDwarfLocDirective( 2248 getContext().getGenDwarfFileNumber(), Line, 0, 2249 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0, 2250 StringRef()); 2251 } 2252 2253 // If parsing succeeded, match the instruction. 2254 if (!ParseHadError) { 2255 uint64_t ErrorInfo; 2256 if (getTargetParser().MatchAndEmitInstruction( 2257 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo, 2258 getTargetParser().isParsingInlineAsm())) 2259 return true; 2260 } 2261 return false; 2262 } 2263 2264 // Parse and erase curly braces marking block start/end 2265 bool 2266 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { 2267 // Identify curly brace marking block start/end 2268 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly)) 2269 return false; 2270 2271 SMLoc StartLoc = Lexer.getLoc(); 2272 Lex(); // Eat the brace 2273 if (Lexer.is(AsmToken::EndOfStatement)) 2274 Lex(); // Eat EndOfStatement following the brace 2275 2276 // Erase the block start/end brace from the output asm string 2277 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() - 2278 StartLoc.getPointer()); 2279 return true; 2280 } 2281 2282 /// parseCppHashLineFilenameComment as this: 2283 /// ::= # number "filename" 2284 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L) { 2285 Lex(); // Eat the hash token. 2286 // Lexer only ever emits HashDirective if it fully formed if it's 2287 // done the checking already so this is an internal error. 2288 assert(getTok().is(AsmToken::Integer) && 2289 "Lexing Cpp line comment: Expected Integer"); 2290 int64_t LineNumber = getTok().getIntVal(); 2291 Lex(); 2292 assert(getTok().is(AsmToken::String) && 2293 "Lexing Cpp line comment: Expected String"); 2294 StringRef Filename = getTok().getString(); 2295 Lex(); 2296 2297 // Get rid of the enclosing quotes. 2298 Filename = Filename.substr(1, Filename.size() - 2); 2299 2300 // Save the SMLoc, Filename and LineNumber for later use by diagnostics 2301 // and possibly DWARF file info. 2302 CppHashInfo.Loc = L; 2303 CppHashInfo.Filename = Filename; 2304 CppHashInfo.LineNumber = LineNumber; 2305 CppHashInfo.Buf = CurBuffer; 2306 if (FirstCppHashFilename.empty()) 2307 FirstCppHashFilename = Filename; 2308 return false; 2309 } 2310 2311 /// will use the last parsed cpp hash line filename comment 2312 /// for the Filename and LineNo if any in the diagnostic. 2313 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 2314 const AsmParser *Parser = static_cast<const AsmParser *>(Context); 2315 raw_ostream &OS = errs(); 2316 2317 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 2318 SMLoc DiagLoc = Diag.getLoc(); 2319 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2320 unsigned CppHashBuf = 2321 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc); 2322 2323 // Like SourceMgr::printMessage() we need to print the include stack if any 2324 // before printing the message. 2325 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2326 if (!Parser->SavedDiagHandler && DiagCurBuffer && 2327 DiagCurBuffer != DiagSrcMgr.getMainFileID()) { 2328 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 2329 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 2330 } 2331 2332 // If we have not parsed a cpp hash line filename comment or the source 2333 // manager changed or buffer changed (like in a nested include) then just 2334 // print the normal diagnostic using its Filename and LineNo. 2335 if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr || 2336 DiagBuf != CppHashBuf) { 2337 if (Parser->SavedDiagHandler) 2338 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2339 else 2340 Diag.print(nullptr, OS); 2341 return; 2342 } 2343 2344 // Use the CppHashFilename and calculate a line number based on the 2345 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc 2346 // for the diagnostic. 2347 const std::string &Filename = Parser->CppHashInfo.Filename; 2348 2349 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 2350 int CppHashLocLineNo = 2351 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf); 2352 int LineNo = 2353 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); 2354 2355 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, 2356 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), 2357 Diag.getLineContents(), Diag.getRanges()); 2358 2359 if (Parser->SavedDiagHandler) 2360 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); 2361 else 2362 NewDiag.print(nullptr, OS); 2363 } 2364 2365 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The 2366 // difference being that that function accepts '@' as part of identifiers and 2367 // we can't do that. AsmLexer.cpp should probably be changed to handle 2368 // '@' as a special case when needed. 2369 static bool isIdentifierChar(char c) { 2370 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' || 2371 c == '.'; 2372 } 2373 2374 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 2375 ArrayRef<MCAsmMacroParameter> Parameters, 2376 ArrayRef<MCAsmMacroArgument> A, 2377 bool EnableAtPseudoVariable, SMLoc L) { 2378 unsigned NParameters = Parameters.size(); 2379 bool HasVararg = NParameters ? Parameters.back().Vararg : false; 2380 if ((!IsDarwin || NParameters != 0) && NParameters != A.size()) 2381 return Error(L, "Wrong number of arguments"); 2382 2383 // A macro without parameters is handled differently on Darwin: 2384 // gas accepts no arguments and does no substitutions 2385 while (!Body.empty()) { 2386 // Scan for the next substitution. 2387 std::size_t End = Body.size(), Pos = 0; 2388 for (; Pos != End; ++Pos) { 2389 // Check for a substitution or escape. 2390 if (IsDarwin && !NParameters) { 2391 // This macro has no parameters, look for $0, $1, etc. 2392 if (Body[Pos] != '$' || Pos + 1 == End) 2393 continue; 2394 2395 char Next = Body[Pos + 1]; 2396 if (Next == '$' || Next == 'n' || 2397 isdigit(static_cast<unsigned char>(Next))) 2398 break; 2399 } else { 2400 // This macro has parameters, look for \foo, \bar, etc. 2401 if (Body[Pos] == '\\' && Pos + 1 != End) 2402 break; 2403 } 2404 } 2405 2406 // Add the prefix. 2407 OS << Body.slice(0, Pos); 2408 2409 // Check if we reached the end. 2410 if (Pos == End) 2411 break; 2412 2413 if (IsDarwin && !NParameters) { 2414 switch (Body[Pos + 1]) { 2415 // $$ => $ 2416 case '$': 2417 OS << '$'; 2418 break; 2419 2420 // $n => number of arguments 2421 case 'n': 2422 OS << A.size(); 2423 break; 2424 2425 // $[0-9] => argument 2426 default: { 2427 // Missing arguments are ignored. 2428 unsigned Index = Body[Pos + 1] - '0'; 2429 if (Index >= A.size()) 2430 break; 2431 2432 // Otherwise substitute with the token values, with spaces eliminated. 2433 for (const AsmToken &Token : A[Index]) 2434 OS << Token.getString(); 2435 break; 2436 } 2437 } 2438 Pos += 2; 2439 } else { 2440 unsigned I = Pos + 1; 2441 2442 // Check for the \@ pseudo-variable. 2443 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End) 2444 ++I; 2445 else 2446 while (isIdentifierChar(Body[I]) && I + 1 != End) 2447 ++I; 2448 2449 const char *Begin = Body.data() + Pos + 1; 2450 StringRef Argument(Begin, I - (Pos + 1)); 2451 unsigned Index = 0; 2452 2453 if (Argument == "@") { 2454 OS << NumOfMacroInstantiations; 2455 Pos += 2; 2456 } else { 2457 for (; Index < NParameters; ++Index) 2458 if (Parameters[Index].Name == Argument) 2459 break; 2460 2461 if (Index == NParameters) { 2462 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 2463 Pos += 3; 2464 else { 2465 OS << '\\' << Argument; 2466 Pos = I; 2467 } 2468 } else { 2469 bool VarargParameter = HasVararg && Index == (NParameters - 1); 2470 for (const AsmToken &Token : A[Index]) 2471 // For altmacro mode, you can write '%expr'. 2472 // The prefix '%' evaluates the expression 'expr' 2473 // and uses the result as a string (e.g. replace %(1+2) with the 2474 // string "3"). 2475 // Here, we identify the integer token which is the result of the 2476 // absolute expression evaluation and replace it with its string 2477 // representation. 2478 if (AltMacroMode && Token.getString().front() == '%' && 2479 Token.is(AsmToken::Integer)) 2480 // Emit an integer value to the buffer. 2481 OS << Token.getIntVal(); 2482 // Only Token that was validated as a string and begins with '<' 2483 // is considered altMacroString!!! 2484 else if (AltMacroMode && Token.getString().front() == '<' && 2485 Token.is(AsmToken::String)) { 2486 OS << altMacroString(Token.getStringContents()); 2487 } 2488 // We expect no quotes around the string's contents when 2489 // parsing for varargs. 2490 else if (Token.isNot(AsmToken::String) || VarargParameter) 2491 OS << Token.getString(); 2492 else 2493 OS << Token.getStringContents(); 2494 2495 Pos += 1 + Argument.size(); 2496 } 2497 } 2498 } 2499 // Update the scan point. 2500 Body = Body.substr(Pos); 2501 } 2502 2503 return false; 2504 } 2505 2506 MacroInstantiation::MacroInstantiation(SMLoc IL, int EB, SMLoc EL, 2507 size_t CondStackDepth) 2508 : InstantiationLoc(IL), ExitBuffer(EB), ExitLoc(EL), 2509 CondStackDepth(CondStackDepth) {} 2510 2511 static bool isOperator(AsmToken::TokenKind kind) { 2512 switch (kind) { 2513 default: 2514 return false; 2515 case AsmToken::Plus: 2516 case AsmToken::Minus: 2517 case AsmToken::Tilde: 2518 case AsmToken::Slash: 2519 case AsmToken::Star: 2520 case AsmToken::Dot: 2521 case AsmToken::Equal: 2522 case AsmToken::EqualEqual: 2523 case AsmToken::Pipe: 2524 case AsmToken::PipePipe: 2525 case AsmToken::Caret: 2526 case AsmToken::Amp: 2527 case AsmToken::AmpAmp: 2528 case AsmToken::Exclaim: 2529 case AsmToken::ExclaimEqual: 2530 case AsmToken::Less: 2531 case AsmToken::LessEqual: 2532 case AsmToken::LessLess: 2533 case AsmToken::LessGreater: 2534 case AsmToken::Greater: 2535 case AsmToken::GreaterEqual: 2536 case AsmToken::GreaterGreater: 2537 return true; 2538 } 2539 } 2540 2541 namespace { 2542 2543 class AsmLexerSkipSpaceRAII { 2544 public: 2545 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { 2546 Lexer.setSkipSpace(SkipSpace); 2547 } 2548 2549 ~AsmLexerSkipSpaceRAII() { 2550 Lexer.setSkipSpace(true); 2551 } 2552 2553 private: 2554 AsmLexer &Lexer; 2555 }; 2556 2557 } // end anonymous namespace 2558 2559 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) { 2560 2561 if (Vararg) { 2562 if (Lexer.isNot(AsmToken::EndOfStatement)) { 2563 StringRef Str = parseStringToEndOfStatement(); 2564 MA.emplace_back(AsmToken::String, Str); 2565 } 2566 return false; 2567 } 2568 2569 unsigned ParenLevel = 0; 2570 2571 // Darwin doesn't use spaces to delmit arguments. 2572 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); 2573 2574 bool SpaceEaten; 2575 2576 while (true) { 2577 SpaceEaten = false; 2578 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 2579 return TokError("unexpected token in macro instantiation"); 2580 2581 if (ParenLevel == 0) { 2582 2583 if (Lexer.is(AsmToken::Comma)) 2584 break; 2585 2586 if (Lexer.is(AsmToken::Space)) { 2587 SpaceEaten = true; 2588 Lexer.Lex(); // Eat spaces 2589 } 2590 2591 // Spaces can delimit parameters, but could also be part an expression. 2592 // If the token after a space is an operator, add the token and the next 2593 // one into this argument 2594 if (!IsDarwin) { 2595 if (isOperator(Lexer.getKind())) { 2596 MA.push_back(getTok()); 2597 Lexer.Lex(); 2598 2599 // Whitespace after an operator can be ignored. 2600 if (Lexer.is(AsmToken::Space)) 2601 Lexer.Lex(); 2602 2603 continue; 2604 } 2605 } 2606 if (SpaceEaten) 2607 break; 2608 } 2609 2610 // handleMacroEntry relies on not advancing the lexer here 2611 // to be able to fill in the remaining default parameter values 2612 if (Lexer.is(AsmToken::EndOfStatement)) 2613 break; 2614 2615 // Adjust the current parentheses level. 2616 if (Lexer.is(AsmToken::LParen)) 2617 ++ParenLevel; 2618 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 2619 --ParenLevel; 2620 2621 // Append the token to the current argument list. 2622 MA.push_back(getTok()); 2623 Lexer.Lex(); 2624 } 2625 2626 if (ParenLevel != 0) 2627 return TokError("unbalanced parentheses in macro argument"); 2628 return false; 2629 } 2630 2631 // Parse the macro instantiation arguments. 2632 bool AsmParser::parseMacroArguments(const MCAsmMacro *M, 2633 MCAsmMacroArguments &A) { 2634 const unsigned NParameters = M ? M->Parameters.size() : 0; 2635 bool NamedParametersFound = false; 2636 SmallVector<SMLoc, 4> FALocs; 2637 2638 A.resize(NParameters); 2639 FALocs.resize(NParameters); 2640 2641 // Parse two kinds of macro invocations: 2642 // - macros defined without any parameters accept an arbitrary number of them 2643 // - macros defined with parameters accept at most that many of them 2644 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false; 2645 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 2646 ++Parameter) { 2647 SMLoc IDLoc = Lexer.getLoc(); 2648 MCAsmMacroParameter FA; 2649 2650 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) { 2651 if (parseIdentifier(FA.Name)) 2652 return Error(IDLoc, "invalid argument identifier for formal argument"); 2653 2654 if (Lexer.isNot(AsmToken::Equal)) 2655 return TokError("expected '=' after formal parameter identifier"); 2656 2657 Lex(); 2658 2659 NamedParametersFound = true; 2660 } 2661 bool Vararg = HasVararg && Parameter == (NParameters - 1); 2662 2663 if (NamedParametersFound && FA.Name.empty()) 2664 return Error(IDLoc, "cannot mix positional and keyword arguments"); 2665 2666 SMLoc StrLoc = Lexer.getLoc(); 2667 SMLoc EndLoc; 2668 if (AltMacroMode && Lexer.is(AsmToken::Percent)) { 2669 const MCExpr *AbsoluteExp; 2670 int64_t Value; 2671 /// Eat '%' 2672 Lex(); 2673 if (parseExpression(AbsoluteExp, EndLoc)) 2674 return false; 2675 if (!AbsoluteExp->evaluateAsAbsolute(Value, 2676 getStreamer().getAssemblerPtr())) 2677 return Error(StrLoc, "expected absolute expression"); 2678 const char *StrChar = StrLoc.getPointer(); 2679 const char *EndChar = EndLoc.getPointer(); 2680 AsmToken newToken(AsmToken::Integer, 2681 StringRef(StrChar, EndChar - StrChar), Value); 2682 FA.Value.push_back(newToken); 2683 } else if (AltMacroMode && Lexer.is(AsmToken::Less) && 2684 isAltmacroString(StrLoc, EndLoc)) { 2685 const char *StrChar = StrLoc.getPointer(); 2686 const char *EndChar = EndLoc.getPointer(); 2687 jumpToLoc(EndLoc, CurBuffer); 2688 /// Eat from '<' to '>' 2689 Lex(); 2690 AsmToken newToken(AsmToken::String, 2691 StringRef(StrChar, EndChar - StrChar)); 2692 FA.Value.push_back(newToken); 2693 } else if(parseMacroArgument(FA.Value, Vararg)) 2694 return true; 2695 2696 unsigned PI = Parameter; 2697 if (!FA.Name.empty()) { 2698 unsigned FAI = 0; 2699 for (FAI = 0; FAI < NParameters; ++FAI) 2700 if (M->Parameters[FAI].Name == FA.Name) 2701 break; 2702 2703 if (FAI >= NParameters) { 2704 assert(M && "expected macro to be defined"); 2705 return Error(IDLoc, "parameter named '" + FA.Name + 2706 "' does not exist for macro '" + M->Name + "'"); 2707 } 2708 PI = FAI; 2709 } 2710 2711 if (!FA.Value.empty()) { 2712 if (A.size() <= PI) 2713 A.resize(PI + 1); 2714 A[PI] = FA.Value; 2715 2716 if (FALocs.size() <= PI) 2717 FALocs.resize(PI + 1); 2718 2719 FALocs[PI] = Lexer.getLoc(); 2720 } 2721 2722 // At the end of the statement, fill in remaining arguments that have 2723 // default values. If there aren't any, then the next argument is 2724 // required but missing 2725 if (Lexer.is(AsmToken::EndOfStatement)) { 2726 bool Failure = false; 2727 for (unsigned FAI = 0; FAI < NParameters; ++FAI) { 2728 if (A[FAI].empty()) { 2729 if (M->Parameters[FAI].Required) { 2730 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), 2731 "missing value for required parameter " 2732 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'"); 2733 Failure = true; 2734 } 2735 2736 if (!M->Parameters[FAI].Value.empty()) 2737 A[FAI] = M->Parameters[FAI].Value; 2738 } 2739 } 2740 return Failure; 2741 } 2742 2743 if (Lexer.is(AsmToken::Comma)) 2744 Lex(); 2745 } 2746 2747 return TokError("too many positional arguments"); 2748 } 2749 2750 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) { 2751 // Arbitrarily limit macro nesting depth (default matches 'as'). We can 2752 // eliminate this, although we should protect against infinite loops. 2753 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; 2754 if (ActiveMacros.size() == MaxNestingDepth) { 2755 std::ostringstream MaxNestingDepthError; 2756 MaxNestingDepthError << "macros cannot be nested more than " 2757 << MaxNestingDepth << " levels deep." 2758 << " Use -asm-macro-max-nesting-depth to increase " 2759 "this limit."; 2760 return TokError(MaxNestingDepthError.str()); 2761 } 2762 2763 MCAsmMacroArguments A; 2764 if (parseMacroArguments(M, A)) 2765 return true; 2766 2767 // Macro instantiation is lexical, unfortunately. We construct a new buffer 2768 // to hold the macro body with substitutions. 2769 SmallString<256> Buf; 2770 StringRef Body = M->Body; 2771 raw_svector_ostream OS(Buf); 2772 2773 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc())) 2774 return true; 2775 2776 // We include the .endmacro in the buffer as our cue to exit the macro 2777 // instantiation. 2778 OS << ".endmacro\n"; 2779 2780 std::unique_ptr<MemoryBuffer> Instantiation = 2781 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 2782 2783 // Create the macro instantiation object and add to the current macro 2784 // instantiation stack. 2785 MacroInstantiation *MI = new MacroInstantiation( 2786 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()); 2787 ActiveMacros.push_back(MI); 2788 2789 ++NumOfMacroInstantiations; 2790 2791 // Jump to the macro instantiation and prime the lexer. 2792 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 2793 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 2794 Lex(); 2795 2796 return false; 2797 } 2798 2799 void AsmParser::handleMacroExit() { 2800 // Jump to the EndOfStatement we should return to, and consume it. 2801 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer); 2802 Lex(); 2803 2804 // Pop the instantiation entry. 2805 delete ActiveMacros.back(); 2806 ActiveMacros.pop_back(); 2807 } 2808 2809 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef, 2810 bool NoDeadStrip) { 2811 MCSymbol *Sym; 2812 const MCExpr *Value; 2813 if (MCParserUtils::parseAssignmentExpression(Name, allow_redef, *this, Sym, 2814 Value)) 2815 return true; 2816 2817 if (!Sym) { 2818 // In the case where we parse an expression starting with a '.', we will 2819 // not generate an error, nor will we create a symbol. In this case we 2820 // should just return out. 2821 return false; 2822 } 2823 2824 // Do the assignment. 2825 Out.EmitAssignment(Sym, Value); 2826 if (NoDeadStrip) 2827 Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip); 2828 2829 return false; 2830 } 2831 2832 /// parseIdentifier: 2833 /// ::= identifier 2834 /// ::= string 2835 bool AsmParser::parseIdentifier(StringRef &Res) { 2836 // The assembler has relaxed rules for accepting identifiers, in particular we 2837 // allow things like '.globl $foo' and '.def @feat.00', which would normally be 2838 // separate tokens. At this level, we have already lexed so we cannot (currently) 2839 // handle this as a context dependent token, instead we detect adjacent tokens 2840 // and return the combined identifier. 2841 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) { 2842 SMLoc PrefixLoc = getLexer().getLoc(); 2843 2844 // Consume the prefix character, and check for a following identifier. 2845 2846 AsmToken Buf[1]; 2847 Lexer.peekTokens(Buf, false); 2848 2849 if (Buf[0].isNot(AsmToken::Identifier)) 2850 return true; 2851 2852 // We have a '$' or '@' followed by an identifier, make sure they are adjacent. 2853 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer()) 2854 return true; 2855 2856 // eat $ or @ 2857 Lexer.Lex(); // Lexer's Lex guarantees consecutive token. 2858 // Construct the joined identifier and consume the token. 2859 Res = 2860 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1); 2861 Lex(); // Parser Lex to maintain invariants. 2862 return false; 2863 } 2864 2865 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String)) 2866 return true; 2867 2868 Res = getTok().getIdentifier(); 2869 2870 Lex(); // Consume the identifier token. 2871 2872 return false; 2873 } 2874 2875 /// parseDirectiveSet: 2876 /// ::= .equ identifier ',' expression 2877 /// ::= .equiv identifier ',' expression 2878 /// ::= .set identifier ',' expression 2879 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) { 2880 StringRef Name; 2881 if (check(parseIdentifier(Name), "expected identifier") || 2882 parseToken(AsmToken::Comma) || parseAssignment(Name, allow_redef, true)) 2883 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 2884 return false; 2885 } 2886 2887 bool AsmParser::parseEscapedString(std::string &Data) { 2888 if (check(getTok().isNot(AsmToken::String), "expected string")) 2889 return true; 2890 2891 Data = ""; 2892 StringRef Str = getTok().getStringContents(); 2893 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 2894 if (Str[i] != '\\') { 2895 Data += Str[i]; 2896 continue; 2897 } 2898 2899 // Recognize escaped characters. Note that this escape semantics currently 2900 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes. 2901 ++i; 2902 if (i == e) 2903 return TokError("unexpected backslash at end of string"); 2904 2905 // Recognize octal sequences. 2906 if ((unsigned)(Str[i] - '0') <= 7) { 2907 // Consume up to three octal characters. 2908 unsigned Value = Str[i] - '0'; 2909 2910 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 2911 ++i; 2912 Value = Value * 8 + (Str[i] - '0'); 2913 2914 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 2915 ++i; 2916 Value = Value * 8 + (Str[i] - '0'); 2917 } 2918 } 2919 2920 if (Value > 255) 2921 return TokError("invalid octal escape sequence (out of range)"); 2922 2923 Data += (unsigned char)Value; 2924 continue; 2925 } 2926 2927 // Otherwise recognize individual escapes. 2928 switch (Str[i]) { 2929 default: 2930 // Just reject invalid escape sequences for now. 2931 return TokError("invalid escape sequence (unrecognized character)"); 2932 2933 case 'b': Data += '\b'; break; 2934 case 'f': Data += '\f'; break; 2935 case 'n': Data += '\n'; break; 2936 case 'r': Data += '\r'; break; 2937 case 't': Data += '\t'; break; 2938 case '"': Data += '"'; break; 2939 case '\\': Data += '\\'; break; 2940 } 2941 } 2942 2943 Lex(); 2944 return false; 2945 } 2946 2947 /// parseDirectiveAscii: 2948 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 2949 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 2950 auto parseOp = [&]() -> bool { 2951 std::string Data; 2952 if (checkForValidSection() || parseEscapedString(Data)) 2953 return true; 2954 getStreamer().EmitBytes(Data); 2955 if (ZeroTerminated) 2956 getStreamer().EmitBytes(StringRef("\0", 1)); 2957 return false; 2958 }; 2959 2960 if (parseMany(parseOp)) 2961 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 2962 return false; 2963 } 2964 2965 /// parseDirectiveReloc 2966 /// ::= .reloc expression , identifier [ , expression ] 2967 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) { 2968 const MCExpr *Offset; 2969 const MCExpr *Expr = nullptr; 2970 int64_t OffsetValue; 2971 SMLoc OffsetLoc = Lexer.getTok().getLoc(); 2972 2973 if (parseExpression(Offset)) 2974 return true; 2975 2976 if ((Offset->evaluateAsAbsolute(OffsetValue, 2977 getStreamer().getAssemblerPtr()) && 2978 check(OffsetValue < 0, OffsetLoc, "expression is negative")) || 2979 (check(Offset->getKind() != llvm::MCExpr::Constant && 2980 Offset->getKind() != llvm::MCExpr::SymbolRef, 2981 OffsetLoc, "expected non-negative number or a label")) || 2982 (parseToken(AsmToken::Comma, "expected comma") || 2983 check(getTok().isNot(AsmToken::Identifier), "expected relocation name"))) 2984 return true; 2985 2986 SMLoc NameLoc = Lexer.getTok().getLoc(); 2987 StringRef Name = Lexer.getTok().getIdentifier(); 2988 Lex(); 2989 2990 if (Lexer.is(AsmToken::Comma)) { 2991 Lex(); 2992 SMLoc ExprLoc = Lexer.getLoc(); 2993 if (parseExpression(Expr)) 2994 return true; 2995 2996 MCValue Value; 2997 if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr)) 2998 return Error(ExprLoc, "expression must be relocatable"); 2999 } 3000 3001 if (parseToken(AsmToken::EndOfStatement, 3002 "unexpected token in .reloc directive")) 3003 return true; 3004 3005 const MCTargetAsmParser &MCT = getTargetParser(); 3006 const MCSubtargetInfo &STI = MCT.getSTI(); 3007 if (getStreamer().EmitRelocDirective(*Offset, Name, Expr, DirectiveLoc, STI)) 3008 return Error(NameLoc, "unknown relocation name"); 3009 3010 return false; 3011 } 3012 3013 /// parseDirectiveValue 3014 /// ::= (.byte | .short | ... ) [ expression (, expression)* ] 3015 bool AsmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { 3016 auto parseOp = [&]() -> bool { 3017 const MCExpr *Value; 3018 SMLoc ExprLoc = getLexer().getLoc(); 3019 if (checkForValidSection() || parseExpression(Value)) 3020 return true; 3021 // Special case constant expressions to match code generator. 3022 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3023 assert(Size <= 8 && "Invalid size"); 3024 uint64_t IntValue = MCE->getValue(); 3025 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 3026 return Error(ExprLoc, "out of range literal value"); 3027 getStreamer().EmitIntValue(IntValue, Size); 3028 } else 3029 getStreamer().EmitValue(Value, Size, ExprLoc); 3030 return false; 3031 }; 3032 3033 if (parseMany(parseOp)) 3034 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3035 return false; 3036 } 3037 3038 static bool parseHexOcta(AsmParser &Asm, uint64_t &hi, uint64_t &lo) { 3039 if (Asm.getTok().isNot(AsmToken::Integer) && 3040 Asm.getTok().isNot(AsmToken::BigNum)) 3041 return Asm.TokError("unknown token in expression"); 3042 SMLoc ExprLoc = Asm.getTok().getLoc(); 3043 APInt IntValue = Asm.getTok().getAPIntVal(); 3044 Asm.Lex(); 3045 if (!IntValue.isIntN(128)) 3046 return Asm.Error(ExprLoc, "out of range literal value"); 3047 if (!IntValue.isIntN(64)) { 3048 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue(); 3049 lo = IntValue.getLoBits(64).getZExtValue(); 3050 } else { 3051 hi = 0; 3052 lo = IntValue.getZExtValue(); 3053 } 3054 return false; 3055 } 3056 3057 /// ParseDirectiveOctaValue 3058 /// ::= .octa [ hexconstant (, hexconstant)* ] 3059 3060 bool AsmParser::parseDirectiveOctaValue(StringRef IDVal) { 3061 auto parseOp = [&]() -> bool { 3062 if (checkForValidSection()) 3063 return true; 3064 uint64_t hi, lo; 3065 if (parseHexOcta(*this, hi, lo)) 3066 return true; 3067 if (MAI.isLittleEndian()) { 3068 getStreamer().EmitIntValue(lo, 8); 3069 getStreamer().EmitIntValue(hi, 8); 3070 } else { 3071 getStreamer().EmitIntValue(hi, 8); 3072 getStreamer().EmitIntValue(lo, 8); 3073 } 3074 return false; 3075 }; 3076 3077 if (parseMany(parseOp)) 3078 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3079 return false; 3080 } 3081 3082 bool AsmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { 3083 // We don't truly support arithmetic on floating point expressions, so we 3084 // have to manually parse unary prefixes. 3085 bool IsNeg = false; 3086 if (getLexer().is(AsmToken::Minus)) { 3087 Lexer.Lex(); 3088 IsNeg = true; 3089 } else if (getLexer().is(AsmToken::Plus)) 3090 Lexer.Lex(); 3091 3092 if (Lexer.is(AsmToken::Error)) 3093 return TokError(Lexer.getErr()); 3094 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) && 3095 Lexer.isNot(AsmToken::Identifier)) 3096 return TokError("unexpected token in directive"); 3097 3098 // Convert to an APFloat. 3099 APFloat Value(Semantics); 3100 StringRef IDVal = getTok().getString(); 3101 if (getLexer().is(AsmToken::Identifier)) { 3102 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf")) 3103 Value = APFloat::getInf(Semantics); 3104 else if (!IDVal.compare_lower("nan")) 3105 Value = APFloat::getNaN(Semantics, false, ~0); 3106 else 3107 return TokError("invalid floating point literal"); 3108 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) == 3109 APFloat::opInvalidOp) 3110 return TokError("invalid floating point literal"); 3111 if (IsNeg) 3112 Value.changeSign(); 3113 3114 // Consume the numeric token. 3115 Lex(); 3116 3117 Res = Value.bitcastToAPInt(); 3118 3119 return false; 3120 } 3121 3122 /// parseDirectiveRealValue 3123 /// ::= (.single | .double) [ expression (, expression)* ] 3124 bool AsmParser::parseDirectiveRealValue(StringRef IDVal, 3125 const fltSemantics &Semantics) { 3126 auto parseOp = [&]() -> bool { 3127 APInt AsInt; 3128 if (checkForValidSection() || parseRealValue(Semantics, AsInt)) 3129 return true; 3130 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 3131 AsInt.getBitWidth() / 8); 3132 return false; 3133 }; 3134 3135 if (parseMany(parseOp)) 3136 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3137 return false; 3138 } 3139 3140 /// parseDirectiveZero 3141 /// ::= .zero expression 3142 bool AsmParser::parseDirectiveZero() { 3143 SMLoc NumBytesLoc = Lexer.getLoc(); 3144 const MCExpr *NumBytes; 3145 if (checkForValidSection() || parseExpression(NumBytes)) 3146 return true; 3147 3148 int64_t Val = 0; 3149 if (getLexer().is(AsmToken::Comma)) { 3150 Lex(); 3151 if (parseAbsoluteExpression(Val)) 3152 return true; 3153 } 3154 3155 if (parseToken(AsmToken::EndOfStatement, 3156 "unexpected token in '.zero' directive")) 3157 return true; 3158 getStreamer().emitFill(*NumBytes, Val, NumBytesLoc); 3159 3160 return false; 3161 } 3162 3163 /// parseDirectiveFill 3164 /// ::= .fill expression [ , expression [ , expression ] ] 3165 bool AsmParser::parseDirectiveFill() { 3166 SMLoc NumValuesLoc = Lexer.getLoc(); 3167 const MCExpr *NumValues; 3168 if (checkForValidSection() || parseExpression(NumValues)) 3169 return true; 3170 3171 int64_t FillSize = 1; 3172 int64_t FillExpr = 0; 3173 3174 SMLoc SizeLoc, ExprLoc; 3175 3176 if (parseOptionalToken(AsmToken::Comma)) { 3177 SizeLoc = getTok().getLoc(); 3178 if (parseAbsoluteExpression(FillSize)) 3179 return true; 3180 if (parseOptionalToken(AsmToken::Comma)) { 3181 ExprLoc = getTok().getLoc(); 3182 if (parseAbsoluteExpression(FillExpr)) 3183 return true; 3184 } 3185 } 3186 if (parseToken(AsmToken::EndOfStatement, 3187 "unexpected token in '.fill' directive")) 3188 return true; 3189 3190 if (FillSize < 0) { 3191 Warning(SizeLoc, "'.fill' directive with negative size has no effect"); 3192 return false; 3193 } 3194 if (FillSize > 8) { 3195 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8"); 3196 FillSize = 8; 3197 } 3198 3199 if (!isUInt<32>(FillExpr) && FillSize > 4) 3200 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits"); 3201 3202 getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc); 3203 3204 return false; 3205 } 3206 3207 /// parseDirectiveOrg 3208 /// ::= .org expression [ , expression ] 3209 bool AsmParser::parseDirectiveOrg() { 3210 const MCExpr *Offset; 3211 SMLoc OffsetLoc = Lexer.getLoc(); 3212 if (checkForValidSection() || parseExpression(Offset)) 3213 return true; 3214 3215 // Parse optional fill expression. 3216 int64_t FillExpr = 0; 3217 if (parseOptionalToken(AsmToken::Comma)) 3218 if (parseAbsoluteExpression(FillExpr)) 3219 return addErrorSuffix(" in '.org' directive"); 3220 if (parseToken(AsmToken::EndOfStatement)) 3221 return addErrorSuffix(" in '.org' directive"); 3222 3223 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc); 3224 return false; 3225 } 3226 3227 /// parseDirectiveAlign 3228 /// ::= {.align, ...} expression [ , expression [ , expression ]] 3229 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 3230 SMLoc AlignmentLoc = getLexer().getLoc(); 3231 int64_t Alignment; 3232 SMLoc MaxBytesLoc; 3233 bool HasFillExpr = false; 3234 int64_t FillExpr = 0; 3235 int64_t MaxBytesToFill = 0; 3236 3237 auto parseAlign = [&]() -> bool { 3238 if (parseAbsoluteExpression(Alignment)) 3239 return true; 3240 if (parseOptionalToken(AsmToken::Comma)) { 3241 // The fill expression can be omitted while specifying a maximum number of 3242 // alignment bytes, e.g: 3243 // .align 3,,4 3244 if (getTok().isNot(AsmToken::Comma)) { 3245 HasFillExpr = true; 3246 if (parseAbsoluteExpression(FillExpr)) 3247 return true; 3248 } 3249 if (parseOptionalToken(AsmToken::Comma)) 3250 if (parseTokenLoc(MaxBytesLoc) || 3251 parseAbsoluteExpression(MaxBytesToFill)) 3252 return true; 3253 } 3254 return parseToken(AsmToken::EndOfStatement); 3255 }; 3256 3257 if (checkForValidSection()) 3258 return addErrorSuffix(" in directive"); 3259 // Ignore empty '.p2align' directives for GNU-as compatibility 3260 if (IsPow2 && (ValueSize == 1) && getTok().is(AsmToken::EndOfStatement)) { 3261 Warning(AlignmentLoc, "p2align directive with no operand(s) is ignored"); 3262 return parseToken(AsmToken::EndOfStatement); 3263 } 3264 if (parseAlign()) 3265 return addErrorSuffix(" in directive"); 3266 3267 // Always emit an alignment here even if we thrown an error. 3268 bool ReturnVal = false; 3269 3270 // Compute alignment in bytes. 3271 if (IsPow2) { 3272 // FIXME: Diagnose overflow. 3273 if (Alignment >= 32) { 3274 ReturnVal |= Error(AlignmentLoc, "invalid alignment value"); 3275 Alignment = 31; 3276 } 3277 3278 Alignment = 1ULL << Alignment; 3279 } else { 3280 // Reject alignments that aren't either a power of two or zero, 3281 // for gas compatibility. Alignment of zero is silently rounded 3282 // up to one. 3283 if (Alignment == 0) 3284 Alignment = 1; 3285 if (!isPowerOf2_64(Alignment)) 3286 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2"); 3287 } 3288 3289 // Diagnose non-sensical max bytes to align. 3290 if (MaxBytesLoc.isValid()) { 3291 if (MaxBytesToFill < 1) { 3292 ReturnVal |= Error(MaxBytesLoc, 3293 "alignment directive can never be satisfied in this " 3294 "many bytes, ignoring maximum bytes expression"); 3295 MaxBytesToFill = 0; 3296 } 3297 3298 if (MaxBytesToFill >= Alignment) { 3299 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 3300 "has no effect"); 3301 MaxBytesToFill = 0; 3302 } 3303 } 3304 3305 // Check whether we should use optimal code alignment for this .align 3306 // directive. 3307 const MCSection *Section = getStreamer().getCurrentSectionOnly(); 3308 assert(Section && "must have section to emit alignment"); 3309 bool UseCodeAlign = Section->UseCodeAlign(); 3310 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 3311 ValueSize == 1 && UseCodeAlign) { 3312 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill); 3313 } else { 3314 // FIXME: Target specific behavior about how the "extra" bytes are filled. 3315 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize, 3316 MaxBytesToFill); 3317 } 3318 3319 return ReturnVal; 3320 } 3321 3322 /// parseDirectiveFile 3323 /// ::= .file filename 3324 /// ::= .file number [directory] filename [md5 checksum] [source source-text] 3325 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) { 3326 // FIXME: I'm not sure what this is. 3327 int64_t FileNumber = -1; 3328 if (getLexer().is(AsmToken::Integer)) { 3329 FileNumber = getTok().getIntVal(); 3330 Lex(); 3331 3332 if (FileNumber < 0) 3333 return TokError("negative file number"); 3334 } 3335 3336 std::string Path; 3337 3338 // Usually the directory and filename together, otherwise just the directory. 3339 // Allow the strings to have escaped octal character sequence. 3340 if (check(getTok().isNot(AsmToken::String), 3341 "unexpected token in '.file' directive") || 3342 parseEscapedString(Path)) 3343 return true; 3344 3345 StringRef Directory; 3346 StringRef Filename; 3347 std::string FilenameData; 3348 if (getLexer().is(AsmToken::String)) { 3349 if (check(FileNumber == -1, 3350 "explicit path specified, but no file number") || 3351 parseEscapedString(FilenameData)) 3352 return true; 3353 Filename = FilenameData; 3354 Directory = Path; 3355 } else { 3356 Filename = Path; 3357 } 3358 3359 uint64_t MD5Hi, MD5Lo; 3360 bool HasMD5 = false; 3361 3362 Optional<StringRef> Source; 3363 bool HasSource = false; 3364 std::string SourceString; 3365 3366 while (!parseOptionalToken(AsmToken::EndOfStatement)) { 3367 StringRef Keyword; 3368 if (check(getTok().isNot(AsmToken::Identifier), 3369 "unexpected token in '.file' directive") || 3370 parseIdentifier(Keyword)) 3371 return true; 3372 if (Keyword == "md5") { 3373 HasMD5 = true; 3374 if (check(FileNumber == -1, 3375 "MD5 checksum specified, but no file number") || 3376 parseHexOcta(*this, MD5Hi, MD5Lo)) 3377 return true; 3378 } else if (Keyword == "source") { 3379 HasSource = true; 3380 if (check(FileNumber == -1, 3381 "source specified, but no file number") || 3382 check(getTok().isNot(AsmToken::String), 3383 "unexpected token in '.file' directive") || 3384 parseEscapedString(SourceString)) 3385 return true; 3386 } else { 3387 return TokError("unexpected token in '.file' directive"); 3388 } 3389 } 3390 3391 if (FileNumber == -1) { 3392 // Ignore the directive if there is no number and the target doesn't support 3393 // numberless .file directives. This allows some portability of assembler 3394 // between different object file formats. 3395 if (getContext().getAsmInfo()->hasSingleParameterDotFile()) 3396 getStreamer().EmitFileDirective(Filename); 3397 } else { 3398 // In case there is a -g option as well as debug info from directive .file, 3399 // we turn off the -g option, directly use the existing debug info instead. 3400 // Throw away any implicit file table for the assembler source. 3401 if (Ctx.getGenDwarfForAssembly()) { 3402 Ctx.getMCDwarfLineTable(0).resetFileTable(); 3403 Ctx.setGenDwarfForAssembly(false); 3404 } 3405 3406 Optional<MD5::MD5Result> CKMem; 3407 if (HasMD5) { 3408 MD5::MD5Result Sum; 3409 for (unsigned i = 0; i != 8; ++i) { 3410 Sum.Bytes[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); 3411 Sum.Bytes[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); 3412 } 3413 CKMem = Sum; 3414 } 3415 if (HasSource) { 3416 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size())); 3417 memcpy(SourceBuf, SourceString.data(), SourceString.size()); 3418 Source = StringRef(SourceBuf, SourceString.size()); 3419 } 3420 if (FileNumber == 0) { 3421 if (Ctx.getDwarfVersion() < 5) 3422 return Warning(DirectiveLoc, "file 0 not supported prior to DWARF-5"); 3423 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source); 3424 } else { 3425 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( 3426 FileNumber, Directory, Filename, CKMem, Source); 3427 if (!FileNumOrErr) 3428 return Error(DirectiveLoc, toString(FileNumOrErr.takeError())); 3429 } 3430 // Alert the user if there are some .file directives with MD5 and some not. 3431 // But only do that once. 3432 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) { 3433 ReportedInconsistentMD5 = true; 3434 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums"); 3435 } 3436 } 3437 3438 return false; 3439 } 3440 3441 /// parseDirectiveLine 3442 /// ::= .line [number] 3443 bool AsmParser::parseDirectiveLine() { 3444 int64_t LineNumber; 3445 if (getLexer().is(AsmToken::Integer)) { 3446 if (parseIntToken(LineNumber, "unexpected token in '.line' directive")) 3447 return true; 3448 (void)LineNumber; 3449 // FIXME: Do something with the .line. 3450 } 3451 if (parseToken(AsmToken::EndOfStatement, 3452 "unexpected token in '.line' directive")) 3453 return true; 3454 3455 return false; 3456 } 3457 3458 /// parseDirectiveLoc 3459 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 3460 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 3461 /// The first number is a file number, must have been previously assigned with 3462 /// a .file directive, the second number is the line number and optionally the 3463 /// third number is a column position (zero if not specified). The remaining 3464 /// optional items are .loc sub-directives. 3465 bool AsmParser::parseDirectiveLoc() { 3466 int64_t FileNumber = 0, LineNumber = 0; 3467 SMLoc Loc = getTok().getLoc(); 3468 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") || 3469 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, 3470 "file number less than one in '.loc' directive") || 3471 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc, 3472 "unassigned file number in '.loc' directive")) 3473 return true; 3474 3475 // optional 3476 if (getLexer().is(AsmToken::Integer)) { 3477 LineNumber = getTok().getIntVal(); 3478 if (LineNumber < 0) 3479 return TokError("line number less than zero in '.loc' directive"); 3480 Lex(); 3481 } 3482 3483 int64_t ColumnPos = 0; 3484 if (getLexer().is(AsmToken::Integer)) { 3485 ColumnPos = getTok().getIntVal(); 3486 if (ColumnPos < 0) 3487 return TokError("column position less than zero in '.loc' directive"); 3488 Lex(); 3489 } 3490 3491 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0; 3492 unsigned Isa = 0; 3493 int64_t Discriminator = 0; 3494 3495 auto parseLocOp = [&]() -> bool { 3496 StringRef Name; 3497 SMLoc Loc = getTok().getLoc(); 3498 if (parseIdentifier(Name)) 3499 return TokError("unexpected token in '.loc' directive"); 3500 3501 if (Name == "basic_block") 3502 Flags |= DWARF2_FLAG_BASIC_BLOCK; 3503 else if (Name == "prologue_end") 3504 Flags |= DWARF2_FLAG_PROLOGUE_END; 3505 else if (Name == "epilogue_begin") 3506 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 3507 else if (Name == "is_stmt") { 3508 Loc = getTok().getLoc(); 3509 const MCExpr *Value; 3510 if (parseExpression(Value)) 3511 return true; 3512 // The expression must be the constant 0 or 1. 3513 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3514 int Value = MCE->getValue(); 3515 if (Value == 0) 3516 Flags &= ~DWARF2_FLAG_IS_STMT; 3517 else if (Value == 1) 3518 Flags |= DWARF2_FLAG_IS_STMT; 3519 else 3520 return Error(Loc, "is_stmt value not 0 or 1"); 3521 } else { 3522 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 3523 } 3524 } else if (Name == "isa") { 3525 Loc = getTok().getLoc(); 3526 const MCExpr *Value; 3527 if (parseExpression(Value)) 3528 return true; 3529 // The expression must be a constant greater or equal to 0. 3530 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3531 int Value = MCE->getValue(); 3532 if (Value < 0) 3533 return Error(Loc, "isa number less than zero"); 3534 Isa = Value; 3535 } else { 3536 return Error(Loc, "isa number not a constant value"); 3537 } 3538 } else if (Name == "discriminator") { 3539 if (parseAbsoluteExpression(Discriminator)) 3540 return true; 3541 } else { 3542 return Error(Loc, "unknown sub-directive in '.loc' directive"); 3543 } 3544 return false; 3545 }; 3546 3547 if (parseMany(parseLocOp, false /*hasComma*/)) 3548 return true; 3549 3550 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 3551 Isa, Discriminator, StringRef()); 3552 3553 return false; 3554 } 3555 3556 /// parseDirectiveStabs 3557 /// ::= .stabs string, number, number, number 3558 bool AsmParser::parseDirectiveStabs() { 3559 return TokError("unsupported directive '.stabs'"); 3560 } 3561 3562 /// parseDirectiveCVFile 3563 /// ::= .cv_file number filename [checksum] [checksumkind] 3564 bool AsmParser::parseDirectiveCVFile() { 3565 SMLoc FileNumberLoc = getTok().getLoc(); 3566 int64_t FileNumber; 3567 std::string Filename; 3568 std::string Checksum; 3569 int64_t ChecksumKind = 0; 3570 3571 if (parseIntToken(FileNumber, 3572 "expected file number in '.cv_file' directive") || 3573 check(FileNumber < 1, FileNumberLoc, "file number less than one") || 3574 check(getTok().isNot(AsmToken::String), 3575 "unexpected token in '.cv_file' directive") || 3576 parseEscapedString(Filename)) 3577 return true; 3578 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 3579 if (check(getTok().isNot(AsmToken::String), 3580 "unexpected token in '.cv_file' directive") || 3581 parseEscapedString(Checksum) || 3582 parseIntToken(ChecksumKind, 3583 "expected checksum kind in '.cv_file' directive") || 3584 parseToken(AsmToken::EndOfStatement, 3585 "unexpected token in '.cv_file' directive")) 3586 return true; 3587 } 3588 3589 Checksum = fromHex(Checksum); 3590 void *CKMem = Ctx.allocate(Checksum.size(), 1); 3591 memcpy(CKMem, Checksum.data(), Checksum.size()); 3592 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), 3593 Checksum.size()); 3594 3595 if (!getStreamer().EmitCVFileDirective(FileNumber, Filename, ChecksumAsBytes, 3596 static_cast<uint8_t>(ChecksumKind))) 3597 return Error(FileNumberLoc, "file number already allocated"); 3598 3599 return false; 3600 } 3601 3602 bool AsmParser::parseCVFunctionId(int64_t &FunctionId, 3603 StringRef DirectiveName) { 3604 SMLoc Loc; 3605 return parseTokenLoc(Loc) || 3606 parseIntToken(FunctionId, "expected function id in '" + DirectiveName + 3607 "' directive") || 3608 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc, 3609 "expected function id within range [0, UINT_MAX)"); 3610 } 3611 3612 bool AsmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { 3613 SMLoc Loc; 3614 return parseTokenLoc(Loc) || 3615 parseIntToken(FileNumber, "expected integer in '" + DirectiveName + 3616 "' directive") || 3617 check(FileNumber < 1, Loc, "file number less than one in '" + 3618 DirectiveName + "' directive") || 3619 check(!getCVContext().isValidFileNumber(FileNumber), Loc, 3620 "unassigned file number in '" + DirectiveName + "' directive"); 3621 } 3622 3623 /// parseDirectiveCVFuncId 3624 /// ::= .cv_func_id FunctionId 3625 /// 3626 /// Introduces a function ID that can be used with .cv_loc. 3627 bool AsmParser::parseDirectiveCVFuncId() { 3628 SMLoc FunctionIdLoc = getTok().getLoc(); 3629 int64_t FunctionId; 3630 3631 if (parseCVFunctionId(FunctionId, ".cv_func_id") || 3632 parseToken(AsmToken::EndOfStatement, 3633 "unexpected token in '.cv_func_id' directive")) 3634 return true; 3635 3636 if (!getStreamer().EmitCVFuncIdDirective(FunctionId)) 3637 return Error(FunctionIdLoc, "function id already allocated"); 3638 3639 return false; 3640 } 3641 3642 /// parseDirectiveCVInlineSiteId 3643 /// ::= .cv_inline_site_id FunctionId 3644 /// "within" IAFunc 3645 /// "inlined_at" IAFile IALine [IACol] 3646 /// 3647 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined 3648 /// at" source location information for use in the line table of the caller, 3649 /// whether the caller is a real function or another inlined call site. 3650 bool AsmParser::parseDirectiveCVInlineSiteId() { 3651 SMLoc FunctionIdLoc = getTok().getLoc(); 3652 int64_t FunctionId; 3653 int64_t IAFunc; 3654 int64_t IAFile; 3655 int64_t IALine; 3656 int64_t IACol = 0; 3657 3658 // FunctionId 3659 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id")) 3660 return true; 3661 3662 // "within" 3663 if (check((getLexer().isNot(AsmToken::Identifier) || 3664 getTok().getIdentifier() != "within"), 3665 "expected 'within' identifier in '.cv_inline_site_id' directive")) 3666 return true; 3667 Lex(); 3668 3669 // IAFunc 3670 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id")) 3671 return true; 3672 3673 // "inlined_at" 3674 if (check((getLexer().isNot(AsmToken::Identifier) || 3675 getTok().getIdentifier() != "inlined_at"), 3676 "expected 'inlined_at' identifier in '.cv_inline_site_id' " 3677 "directive") ) 3678 return true; 3679 Lex(); 3680 3681 // IAFile IALine 3682 if (parseCVFileId(IAFile, ".cv_inline_site_id") || 3683 parseIntToken(IALine, "expected line number after 'inlined_at'")) 3684 return true; 3685 3686 // [IACol] 3687 if (getLexer().is(AsmToken::Integer)) { 3688 IACol = getTok().getIntVal(); 3689 Lex(); 3690 } 3691 3692 if (parseToken(AsmToken::EndOfStatement, 3693 "unexpected token in '.cv_inline_site_id' directive")) 3694 return true; 3695 3696 if (!getStreamer().EmitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, 3697 IALine, IACol, FunctionIdLoc)) 3698 return Error(FunctionIdLoc, "function id already allocated"); 3699 3700 return false; 3701 } 3702 3703 /// parseDirectiveCVLoc 3704 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] 3705 /// [is_stmt VALUE] 3706 /// The first number is a file number, must have been previously assigned with 3707 /// a .file directive, the second number is the line number and optionally the 3708 /// third number is a column position (zero if not specified). The remaining 3709 /// optional items are .loc sub-directives. 3710 bool AsmParser::parseDirectiveCVLoc() { 3711 SMLoc DirectiveLoc = getTok().getLoc(); 3712 int64_t FunctionId, FileNumber; 3713 if (parseCVFunctionId(FunctionId, ".cv_loc") || 3714 parseCVFileId(FileNumber, ".cv_loc")) 3715 return true; 3716 3717 int64_t LineNumber = 0; 3718 if (getLexer().is(AsmToken::Integer)) { 3719 LineNumber = getTok().getIntVal(); 3720 if (LineNumber < 0) 3721 return TokError("line number less than zero in '.cv_loc' directive"); 3722 Lex(); 3723 } 3724 3725 int64_t ColumnPos = 0; 3726 if (getLexer().is(AsmToken::Integer)) { 3727 ColumnPos = getTok().getIntVal(); 3728 if (ColumnPos < 0) 3729 return TokError("column position less than zero in '.cv_loc' directive"); 3730 Lex(); 3731 } 3732 3733 bool PrologueEnd = false; 3734 uint64_t IsStmt = 0; 3735 3736 auto parseOp = [&]() -> bool { 3737 StringRef Name; 3738 SMLoc Loc = getTok().getLoc(); 3739 if (parseIdentifier(Name)) 3740 return TokError("unexpected token in '.cv_loc' directive"); 3741 if (Name == "prologue_end") 3742 PrologueEnd = true; 3743 else if (Name == "is_stmt") { 3744 Loc = getTok().getLoc(); 3745 const MCExpr *Value; 3746 if (parseExpression(Value)) 3747 return true; 3748 // The expression must be the constant 0 or 1. 3749 IsStmt = ~0ULL; 3750 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value)) 3751 IsStmt = MCE->getValue(); 3752 3753 if (IsStmt > 1) 3754 return Error(Loc, "is_stmt value not 0 or 1"); 3755 } else { 3756 return Error(Loc, "unknown sub-directive in '.cv_loc' directive"); 3757 } 3758 return false; 3759 }; 3760 3761 if (parseMany(parseOp, false /*hasComma*/)) 3762 return true; 3763 3764 getStreamer().EmitCVLocDirective(FunctionId, FileNumber, LineNumber, 3765 ColumnPos, PrologueEnd, IsStmt, StringRef(), 3766 DirectiveLoc); 3767 return false; 3768 } 3769 3770 /// parseDirectiveCVLinetable 3771 /// ::= .cv_linetable FunctionId, FnStart, FnEnd 3772 bool AsmParser::parseDirectiveCVLinetable() { 3773 int64_t FunctionId; 3774 StringRef FnStartName, FnEndName; 3775 SMLoc Loc = getTok().getLoc(); 3776 if (parseCVFunctionId(FunctionId, ".cv_linetable") || 3777 parseToken(AsmToken::Comma, 3778 "unexpected token in '.cv_linetable' directive") || 3779 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3780 "expected identifier in directive") || 3781 parseToken(AsmToken::Comma, 3782 "unexpected token in '.cv_linetable' directive") || 3783 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3784 "expected identifier in directive")) 3785 return true; 3786 3787 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3788 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3789 3790 getStreamer().EmitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym); 3791 return false; 3792 } 3793 3794 /// parseDirectiveCVInlineLinetable 3795 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd 3796 bool AsmParser::parseDirectiveCVInlineLinetable() { 3797 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; 3798 StringRef FnStartName, FnEndName; 3799 SMLoc Loc = getTok().getLoc(); 3800 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") || 3801 parseTokenLoc(Loc) || 3802 parseIntToken( 3803 SourceFileId, 3804 "expected SourceField in '.cv_inline_linetable' directive") || 3805 check(SourceFileId <= 0, Loc, 3806 "File id less than zero in '.cv_inline_linetable' directive") || 3807 parseTokenLoc(Loc) || 3808 parseIntToken( 3809 SourceLineNum, 3810 "expected SourceLineNum in '.cv_inline_linetable' directive") || 3811 check(SourceLineNum < 0, Loc, 3812 "Line number less than zero in '.cv_inline_linetable' directive") || 3813 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3814 "expected identifier in directive") || 3815 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3816 "expected identifier in directive")) 3817 return true; 3818 3819 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 3820 return true; 3821 3822 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3823 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3824 getStreamer().EmitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, 3825 SourceLineNum, FnStartSym, 3826 FnEndSym); 3827 return false; 3828 } 3829 3830 /// parseDirectiveCVDefRange 3831 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* 3832 bool AsmParser::parseDirectiveCVDefRange() { 3833 SMLoc Loc; 3834 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; 3835 while (getLexer().is(AsmToken::Identifier)) { 3836 Loc = getLexer().getLoc(); 3837 StringRef GapStartName; 3838 if (parseIdentifier(GapStartName)) 3839 return Error(Loc, "expected identifier in directive"); 3840 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName); 3841 3842 Loc = getLexer().getLoc(); 3843 StringRef GapEndName; 3844 if (parseIdentifier(GapEndName)) 3845 return Error(Loc, "expected identifier in directive"); 3846 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName); 3847 3848 Ranges.push_back({GapStartSym, GapEndSym}); 3849 } 3850 3851 std::string FixedSizePortion; 3852 if (parseToken(AsmToken::Comma, "unexpected token in directive") || 3853 parseEscapedString(FixedSizePortion)) 3854 return true; 3855 3856 getStreamer().EmitCVDefRangeDirective(Ranges, FixedSizePortion); 3857 return false; 3858 } 3859 3860 /// parseDirectiveCVString 3861 /// ::= .cv_stringtable "string" 3862 bool AsmParser::parseDirectiveCVString() { 3863 std::string Data; 3864 if (checkForValidSection() || parseEscapedString(Data)) 3865 return addErrorSuffix(" in '.cv_string' directive"); 3866 3867 // Put the string in the table and emit the offset. 3868 std::pair<StringRef, unsigned> Insertion = 3869 getCVContext().addToStringTable(Data); 3870 getStreamer().EmitIntValue(Insertion.second, 4); 3871 return false; 3872 } 3873 3874 /// parseDirectiveCVStringTable 3875 /// ::= .cv_stringtable 3876 bool AsmParser::parseDirectiveCVStringTable() { 3877 getStreamer().EmitCVStringTableDirective(); 3878 return false; 3879 } 3880 3881 /// parseDirectiveCVFileChecksums 3882 /// ::= .cv_filechecksums 3883 bool AsmParser::parseDirectiveCVFileChecksums() { 3884 getStreamer().EmitCVFileChecksumsDirective(); 3885 return false; 3886 } 3887 3888 /// parseDirectiveCVFileChecksumOffset 3889 /// ::= .cv_filechecksumoffset fileno 3890 bool AsmParser::parseDirectiveCVFileChecksumOffset() { 3891 int64_t FileNo; 3892 if (parseIntToken(FileNo, "expected identifier in directive")) 3893 return true; 3894 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 3895 return true; 3896 getStreamer().EmitCVFileChecksumOffsetDirective(FileNo); 3897 return false; 3898 } 3899 3900 /// parseDirectiveCVFPOData 3901 /// ::= .cv_fpo_data procsym 3902 bool AsmParser::parseDirectiveCVFPOData() { 3903 SMLoc DirLoc = getLexer().getLoc(); 3904 StringRef ProcName; 3905 if (parseIdentifier(ProcName)) 3906 return TokError("expected symbol name"); 3907 if (parseEOL("unexpected tokens")) 3908 return addErrorSuffix(" in '.cv_fpo_data' directive"); 3909 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName); 3910 getStreamer().EmitCVFPOData(ProcSym, DirLoc); 3911 return false; 3912 } 3913 3914 /// parseDirectiveCFISections 3915 /// ::= .cfi_sections section [, section] 3916 bool AsmParser::parseDirectiveCFISections() { 3917 StringRef Name; 3918 bool EH = false; 3919 bool Debug = false; 3920 3921 if (parseIdentifier(Name)) 3922 return TokError("Expected an identifier"); 3923 3924 if (Name == ".eh_frame") 3925 EH = true; 3926 else if (Name == ".debug_frame") 3927 Debug = true; 3928 3929 if (getLexer().is(AsmToken::Comma)) { 3930 Lex(); 3931 3932 if (parseIdentifier(Name)) 3933 return TokError("Expected an identifier"); 3934 3935 if (Name == ".eh_frame") 3936 EH = true; 3937 else if (Name == ".debug_frame") 3938 Debug = true; 3939 } 3940 3941 getStreamer().EmitCFISections(EH, Debug); 3942 return false; 3943 } 3944 3945 /// parseDirectiveCFIStartProc 3946 /// ::= .cfi_startproc [simple] 3947 bool AsmParser::parseDirectiveCFIStartProc() { 3948 StringRef Simple; 3949 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 3950 if (check(parseIdentifier(Simple) || Simple != "simple", 3951 "unexpected token") || 3952 parseToken(AsmToken::EndOfStatement)) 3953 return addErrorSuffix(" in '.cfi_startproc' directive"); 3954 } 3955 3956 // TODO(kristina): Deal with a corner case of incorrect diagnostic context 3957 // being produced if this directive is emitted as part of preprocessor macro 3958 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. 3959 // Tools like llvm-mc on the other hand are not affected by it, and report 3960 // correct context information. 3961 getStreamer().EmitCFIStartProc(!Simple.empty(), Lexer.getLoc()); 3962 return false; 3963 } 3964 3965 /// parseDirectiveCFIEndProc 3966 /// ::= .cfi_endproc 3967 bool AsmParser::parseDirectiveCFIEndProc() { 3968 getStreamer().EmitCFIEndProc(); 3969 return false; 3970 } 3971 3972 /// parse register name or number. 3973 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register, 3974 SMLoc DirectiveLoc) { 3975 unsigned RegNo; 3976 3977 if (getLexer().isNot(AsmToken::Integer)) { 3978 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc)) 3979 return true; 3980 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true); 3981 } else 3982 return parseAbsoluteExpression(Register); 3983 3984 return false; 3985 } 3986 3987 /// parseDirectiveCFIDefCfa 3988 /// ::= .cfi_def_cfa register, offset 3989 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { 3990 int64_t Register = 0, Offset = 0; 3991 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 3992 parseToken(AsmToken::Comma, "unexpected token in directive") || 3993 parseAbsoluteExpression(Offset)) 3994 return true; 3995 3996 getStreamer().EmitCFIDefCfa(Register, Offset); 3997 return false; 3998 } 3999 4000 /// parseDirectiveCFIDefCfaOffset 4001 /// ::= .cfi_def_cfa_offset offset 4002 bool AsmParser::parseDirectiveCFIDefCfaOffset() { 4003 int64_t Offset = 0; 4004 if (parseAbsoluteExpression(Offset)) 4005 return true; 4006 4007 getStreamer().EmitCFIDefCfaOffset(Offset); 4008 return false; 4009 } 4010 4011 /// parseDirectiveCFIRegister 4012 /// ::= .cfi_register register, register 4013 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { 4014 int64_t Register1 = 0, Register2 = 0; 4015 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || 4016 parseToken(AsmToken::Comma, "unexpected token in directive") || 4017 parseRegisterOrRegisterNumber(Register2, DirectiveLoc)) 4018 return true; 4019 4020 getStreamer().EmitCFIRegister(Register1, Register2); 4021 return false; 4022 } 4023 4024 /// parseDirectiveCFIWindowSave 4025 /// ::= .cfi_window_save 4026 bool AsmParser::parseDirectiveCFIWindowSave() { 4027 getStreamer().EmitCFIWindowSave(); 4028 return false; 4029 } 4030 4031 /// parseDirectiveCFIAdjustCfaOffset 4032 /// ::= .cfi_adjust_cfa_offset adjustment 4033 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() { 4034 int64_t Adjustment = 0; 4035 if (parseAbsoluteExpression(Adjustment)) 4036 return true; 4037 4038 getStreamer().EmitCFIAdjustCfaOffset(Adjustment); 4039 return false; 4040 } 4041 4042 /// parseDirectiveCFIDefCfaRegister 4043 /// ::= .cfi_def_cfa_register register 4044 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { 4045 int64_t Register = 0; 4046 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4047 return true; 4048 4049 getStreamer().EmitCFIDefCfaRegister(Register); 4050 return false; 4051 } 4052 4053 /// parseDirectiveCFIOffset 4054 /// ::= .cfi_offset register, offset 4055 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { 4056 int64_t Register = 0; 4057 int64_t Offset = 0; 4058 4059 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4060 parseToken(AsmToken::Comma, "unexpected token in directive") || 4061 parseAbsoluteExpression(Offset)) 4062 return true; 4063 4064 getStreamer().EmitCFIOffset(Register, Offset); 4065 return false; 4066 } 4067 4068 /// parseDirectiveCFIRelOffset 4069 /// ::= .cfi_rel_offset register, offset 4070 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { 4071 int64_t Register = 0, Offset = 0; 4072 4073 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4074 parseToken(AsmToken::Comma, "unexpected token in directive") || 4075 parseAbsoluteExpression(Offset)) 4076 return true; 4077 4078 getStreamer().EmitCFIRelOffset(Register, Offset); 4079 return false; 4080 } 4081 4082 static bool isValidEncoding(int64_t Encoding) { 4083 if (Encoding & ~0xff) 4084 return false; 4085 4086 if (Encoding == dwarf::DW_EH_PE_omit) 4087 return true; 4088 4089 const unsigned Format = Encoding & 0xf; 4090 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 4091 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 4092 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 4093 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 4094 return false; 4095 4096 const unsigned Application = Encoding & 0x70; 4097 if (Application != dwarf::DW_EH_PE_absptr && 4098 Application != dwarf::DW_EH_PE_pcrel) 4099 return false; 4100 4101 return true; 4102 } 4103 4104 /// parseDirectiveCFIPersonalityOrLsda 4105 /// IsPersonality true for cfi_personality, false for cfi_lsda 4106 /// ::= .cfi_personality encoding, [symbol_name] 4107 /// ::= .cfi_lsda encoding, [symbol_name] 4108 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { 4109 int64_t Encoding = 0; 4110 if (parseAbsoluteExpression(Encoding)) 4111 return true; 4112 if (Encoding == dwarf::DW_EH_PE_omit) 4113 return false; 4114 4115 StringRef Name; 4116 if (check(!isValidEncoding(Encoding), "unsupported encoding.") || 4117 parseToken(AsmToken::Comma, "unexpected token in directive") || 4118 check(parseIdentifier(Name), "expected identifier in directive")) 4119 return true; 4120 4121 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4122 4123 if (IsPersonality) 4124 getStreamer().EmitCFIPersonality(Sym, Encoding); 4125 else 4126 getStreamer().EmitCFILsda(Sym, Encoding); 4127 return false; 4128 } 4129 4130 /// parseDirectiveCFIRememberState 4131 /// ::= .cfi_remember_state 4132 bool AsmParser::parseDirectiveCFIRememberState() { 4133 getStreamer().EmitCFIRememberState(); 4134 return false; 4135 } 4136 4137 /// parseDirectiveCFIRestoreState 4138 /// ::= .cfi_remember_state 4139 bool AsmParser::parseDirectiveCFIRestoreState() { 4140 getStreamer().EmitCFIRestoreState(); 4141 return false; 4142 } 4143 4144 /// parseDirectiveCFISameValue 4145 /// ::= .cfi_same_value register 4146 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { 4147 int64_t Register = 0; 4148 4149 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4150 return true; 4151 4152 getStreamer().EmitCFISameValue(Register); 4153 return false; 4154 } 4155 4156 /// parseDirectiveCFIRestore 4157 /// ::= .cfi_restore register 4158 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { 4159 int64_t Register = 0; 4160 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4161 return true; 4162 4163 getStreamer().EmitCFIRestore(Register); 4164 return false; 4165 } 4166 4167 /// parseDirectiveCFIEscape 4168 /// ::= .cfi_escape expression[,...] 4169 bool AsmParser::parseDirectiveCFIEscape() { 4170 std::string Values; 4171 int64_t CurrValue; 4172 if (parseAbsoluteExpression(CurrValue)) 4173 return true; 4174 4175 Values.push_back((uint8_t)CurrValue); 4176 4177 while (getLexer().is(AsmToken::Comma)) { 4178 Lex(); 4179 4180 if (parseAbsoluteExpression(CurrValue)) 4181 return true; 4182 4183 Values.push_back((uint8_t)CurrValue); 4184 } 4185 4186 getStreamer().EmitCFIEscape(Values); 4187 return false; 4188 } 4189 4190 /// parseDirectiveCFIReturnColumn 4191 /// ::= .cfi_return_column register 4192 bool AsmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { 4193 int64_t Register = 0; 4194 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4195 return true; 4196 getStreamer().EmitCFIReturnColumn(Register); 4197 return false; 4198 } 4199 4200 /// parseDirectiveCFISignalFrame 4201 /// ::= .cfi_signal_frame 4202 bool AsmParser::parseDirectiveCFISignalFrame() { 4203 if (parseToken(AsmToken::EndOfStatement, 4204 "unexpected token in '.cfi_signal_frame'")) 4205 return true; 4206 4207 getStreamer().EmitCFISignalFrame(); 4208 return false; 4209 } 4210 4211 /// parseDirectiveCFIUndefined 4212 /// ::= .cfi_undefined register 4213 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { 4214 int64_t Register = 0; 4215 4216 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 4217 return true; 4218 4219 getStreamer().EmitCFIUndefined(Register); 4220 return false; 4221 } 4222 4223 /// parseDirectiveAltmacro 4224 /// ::= .altmacro 4225 /// ::= .noaltmacro 4226 bool AsmParser::parseDirectiveAltmacro(StringRef Directive) { 4227 if (getLexer().isNot(AsmToken::EndOfStatement)) 4228 return TokError("unexpected token in '" + Directive + "' directive"); 4229 AltMacroMode = (Directive == ".altmacro"); 4230 return false; 4231 } 4232 4233 /// parseDirectiveMacrosOnOff 4234 /// ::= .macros_on 4235 /// ::= .macros_off 4236 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) { 4237 if (parseToken(AsmToken::EndOfStatement, 4238 "unexpected token in '" + Directive + "' directive")) 4239 return true; 4240 4241 setMacrosEnabled(Directive == ".macros_on"); 4242 return false; 4243 } 4244 4245 /// parseDirectiveMacro 4246 /// ::= .macro name[,] [parameters] 4247 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) { 4248 StringRef Name; 4249 if (parseIdentifier(Name)) 4250 return TokError("expected identifier in '.macro' directive"); 4251 4252 if (getLexer().is(AsmToken::Comma)) 4253 Lex(); 4254 4255 MCAsmMacroParameters Parameters; 4256 while (getLexer().isNot(AsmToken::EndOfStatement)) { 4257 4258 if (!Parameters.empty() && Parameters.back().Vararg) 4259 return Error(Lexer.getLoc(), 4260 "Vararg parameter '" + Parameters.back().Name + 4261 "' should be last one in the list of parameters."); 4262 4263 MCAsmMacroParameter Parameter; 4264 if (parseIdentifier(Parameter.Name)) 4265 return TokError("expected identifier in '.macro' directive"); 4266 4267 // Emit an error if two (or more) named parameters share the same name 4268 for (const MCAsmMacroParameter& CurrParam : Parameters) 4269 if (CurrParam.Name.equals(Parameter.Name)) 4270 return TokError("macro '" + Name + "' has multiple parameters" 4271 " named '" + Parameter.Name + "'"); 4272 4273 if (Lexer.is(AsmToken::Colon)) { 4274 Lex(); // consume ':' 4275 4276 SMLoc QualLoc; 4277 StringRef Qualifier; 4278 4279 QualLoc = Lexer.getLoc(); 4280 if (parseIdentifier(Qualifier)) 4281 return Error(QualLoc, "missing parameter qualifier for " 4282 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4283 4284 if (Qualifier == "req") 4285 Parameter.Required = true; 4286 else if (Qualifier == "vararg") 4287 Parameter.Vararg = true; 4288 else 4289 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier " 4290 "for '" + Parameter.Name + "' in macro '" + Name + "'"); 4291 } 4292 4293 if (getLexer().is(AsmToken::Equal)) { 4294 Lex(); 4295 4296 SMLoc ParamLoc; 4297 4298 ParamLoc = Lexer.getLoc(); 4299 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false )) 4300 return true; 4301 4302 if (Parameter.Required) 4303 Warning(ParamLoc, "pointless default value for required parameter " 4304 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4305 } 4306 4307 Parameters.push_back(std::move(Parameter)); 4308 4309 if (getLexer().is(AsmToken::Comma)) 4310 Lex(); 4311 } 4312 4313 // Eat just the end of statement. 4314 Lexer.Lex(); 4315 4316 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors 4317 AsmToken EndToken, StartToken = getTok(); 4318 unsigned MacroDepth = 0; 4319 // Lex the macro definition. 4320 while (true) { 4321 // Ignore Lexing errors in macros. 4322 while (Lexer.is(AsmToken::Error)) { 4323 Lexer.Lex(); 4324 } 4325 4326 // Check whether we have reached the end of the file. 4327 if (getLexer().is(AsmToken::Eof)) 4328 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 4329 4330 // Otherwise, check whether we have reach the .endmacro. 4331 if (getLexer().is(AsmToken::Identifier)) { 4332 if (getTok().getIdentifier() == ".endm" || 4333 getTok().getIdentifier() == ".endmacro") { 4334 if (MacroDepth == 0) { // Outermost macro. 4335 EndToken = getTok(); 4336 Lexer.Lex(); 4337 if (getLexer().isNot(AsmToken::EndOfStatement)) 4338 return TokError("unexpected token in '" + EndToken.getIdentifier() + 4339 "' directive"); 4340 break; 4341 } else { 4342 // Otherwise we just found the end of an inner macro. 4343 --MacroDepth; 4344 } 4345 } else if (getTok().getIdentifier() == ".macro") { 4346 // We allow nested macros. Those aren't instantiated until the outermost 4347 // macro is expanded so just ignore them for now. 4348 ++MacroDepth; 4349 } 4350 } 4351 4352 // Otherwise, scan til the end of the statement. 4353 eatToEndOfStatement(); 4354 } 4355 4356 if (getContext().lookupMacro(Name)) { 4357 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 4358 } 4359 4360 const char *BodyStart = StartToken.getLoc().getPointer(); 4361 const char *BodyEnd = EndToken.getLoc().getPointer(); 4362 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 4363 checkForBadMacro(DirectiveLoc, Name, Body, Parameters); 4364 MCAsmMacro Macro(Name, Body, std::move(Parameters)); 4365 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n"; 4366 Macro.dump()); 4367 getContext().defineMacro(Name, std::move(Macro)); 4368 return false; 4369 } 4370 4371 /// checkForBadMacro 4372 /// 4373 /// With the support added for named parameters there may be code out there that 4374 /// is transitioning from positional parameters. In versions of gas that did 4375 /// not support named parameters they would be ignored on the macro definition. 4376 /// But to support both styles of parameters this is not possible so if a macro 4377 /// definition has named parameters but does not use them and has what appears 4378 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a 4379 /// warning that the positional parameter found in body which have no effect. 4380 /// Hoping the developer will either remove the named parameters from the macro 4381 /// definition so the positional parameters get used if that was what was 4382 /// intended or change the macro to use the named parameters. It is possible 4383 /// this warning will trigger when the none of the named parameters are used 4384 /// and the strings like $1 are infact to simply to be passed trough unchanged. 4385 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, 4386 StringRef Body, 4387 ArrayRef<MCAsmMacroParameter> Parameters) { 4388 // If this macro is not defined with named parameters the warning we are 4389 // checking for here doesn't apply. 4390 unsigned NParameters = Parameters.size(); 4391 if (NParameters == 0) 4392 return; 4393 4394 bool NamedParametersFound = false; 4395 bool PositionalParametersFound = false; 4396 4397 // Look at the body of the macro for use of both the named parameters and what 4398 // are likely to be positional parameters. This is what expandMacro() is 4399 // doing when it finds the parameters in the body. 4400 while (!Body.empty()) { 4401 // Scan for the next possible parameter. 4402 std::size_t End = Body.size(), Pos = 0; 4403 for (; Pos != End; ++Pos) { 4404 // Check for a substitution or escape. 4405 // This macro is defined with parameters, look for \foo, \bar, etc. 4406 if (Body[Pos] == '\\' && Pos + 1 != End) 4407 break; 4408 4409 // This macro should have parameters, but look for $0, $1, ..., $n too. 4410 if (Body[Pos] != '$' || Pos + 1 == End) 4411 continue; 4412 char Next = Body[Pos + 1]; 4413 if (Next == '$' || Next == 'n' || 4414 isdigit(static_cast<unsigned char>(Next))) 4415 break; 4416 } 4417 4418 // Check if we reached the end. 4419 if (Pos == End) 4420 break; 4421 4422 if (Body[Pos] == '$') { 4423 switch (Body[Pos + 1]) { 4424 // $$ => $ 4425 case '$': 4426 break; 4427 4428 // $n => number of arguments 4429 case 'n': 4430 PositionalParametersFound = true; 4431 break; 4432 4433 // $[0-9] => argument 4434 default: { 4435 PositionalParametersFound = true; 4436 break; 4437 } 4438 } 4439 Pos += 2; 4440 } else { 4441 unsigned I = Pos + 1; 4442 while (isIdentifierChar(Body[I]) && I + 1 != End) 4443 ++I; 4444 4445 const char *Begin = Body.data() + Pos + 1; 4446 StringRef Argument(Begin, I - (Pos + 1)); 4447 unsigned Index = 0; 4448 for (; Index < NParameters; ++Index) 4449 if (Parameters[Index].Name == Argument) 4450 break; 4451 4452 if (Index == NParameters) { 4453 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 4454 Pos += 3; 4455 else { 4456 Pos = I; 4457 } 4458 } else { 4459 NamedParametersFound = true; 4460 Pos += 1 + Argument.size(); 4461 } 4462 } 4463 // Update the scan point. 4464 Body = Body.substr(Pos); 4465 } 4466 4467 if (!NamedParametersFound && PositionalParametersFound) 4468 Warning(DirectiveLoc, "macro defined with named parameters which are not " 4469 "used in macro body, possible positional parameter " 4470 "found in body which will have no effect"); 4471 } 4472 4473 /// parseDirectiveExitMacro 4474 /// ::= .exitm 4475 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) { 4476 if (parseToken(AsmToken::EndOfStatement, 4477 "unexpected token in '" + Directive + "' directive")) 4478 return true; 4479 4480 if (!isInsideMacroInstantiation()) 4481 return TokError("unexpected '" + Directive + "' in file, " 4482 "no current macro definition"); 4483 4484 // Exit all conditionals that are active in the current macro. 4485 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { 4486 TheCondState = TheCondStack.back(); 4487 TheCondStack.pop_back(); 4488 } 4489 4490 handleMacroExit(); 4491 return false; 4492 } 4493 4494 /// parseDirectiveEndMacro 4495 /// ::= .endm 4496 /// ::= .endmacro 4497 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) { 4498 if (getLexer().isNot(AsmToken::EndOfStatement)) 4499 return TokError("unexpected token in '" + Directive + "' directive"); 4500 4501 // If we are inside a macro instantiation, terminate the current 4502 // instantiation. 4503 if (isInsideMacroInstantiation()) { 4504 handleMacroExit(); 4505 return false; 4506 } 4507 4508 // Otherwise, this .endmacro is a stray entry in the file; well formed 4509 // .endmacro directives are handled during the macro definition parsing. 4510 return TokError("unexpected '" + Directive + "' in file, " 4511 "no current macro definition"); 4512 } 4513 4514 /// parseDirectivePurgeMacro 4515 /// ::= .purgem 4516 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { 4517 StringRef Name; 4518 SMLoc Loc; 4519 if (parseTokenLoc(Loc) || 4520 check(parseIdentifier(Name), Loc, 4521 "expected identifier in '.purgem' directive") || 4522 parseToken(AsmToken::EndOfStatement, 4523 "unexpected token in '.purgem' directive")) 4524 return true; 4525 4526 if (!getContext().lookupMacro(Name)) 4527 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 4528 4529 getContext().undefineMacro(Name); 4530 DEBUG_WITH_TYPE("asm-macros", dbgs() 4531 << "Un-defining macro: " << Name << "\n"); 4532 return false; 4533 } 4534 4535 /// parseDirectiveBundleAlignMode 4536 /// ::= {.bundle_align_mode} expression 4537 bool AsmParser::parseDirectiveBundleAlignMode() { 4538 // Expect a single argument: an expression that evaluates to a constant 4539 // in the inclusive range 0-30. 4540 SMLoc ExprLoc = getLexer().getLoc(); 4541 int64_t AlignSizePow2; 4542 if (checkForValidSection() || parseAbsoluteExpression(AlignSizePow2) || 4543 parseToken(AsmToken::EndOfStatement, "unexpected token after expression " 4544 "in '.bundle_align_mode' " 4545 "directive") || 4546 check(AlignSizePow2 < 0 || AlignSizePow2 > 30, ExprLoc, 4547 "invalid bundle alignment size (expected between 0 and 30)")) 4548 return true; 4549 4550 // Because of AlignSizePow2's verified range we can safely truncate it to 4551 // unsigned. 4552 getStreamer().EmitBundleAlignMode(static_cast<unsigned>(AlignSizePow2)); 4553 return false; 4554 } 4555 4556 /// parseDirectiveBundleLock 4557 /// ::= {.bundle_lock} [align_to_end] 4558 bool AsmParser::parseDirectiveBundleLock() { 4559 if (checkForValidSection()) 4560 return true; 4561 bool AlignToEnd = false; 4562 4563 StringRef Option; 4564 SMLoc Loc = getTok().getLoc(); 4565 const char *kInvalidOptionError = 4566 "invalid option for '.bundle_lock' directive"; 4567 4568 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4569 if (check(parseIdentifier(Option), Loc, kInvalidOptionError) || 4570 check(Option != "align_to_end", Loc, kInvalidOptionError) || 4571 parseToken(AsmToken::EndOfStatement, 4572 "unexpected token after '.bundle_lock' directive option")) 4573 return true; 4574 AlignToEnd = true; 4575 } 4576 4577 getStreamer().EmitBundleLock(AlignToEnd); 4578 return false; 4579 } 4580 4581 /// parseDirectiveBundleLock 4582 /// ::= {.bundle_lock} 4583 bool AsmParser::parseDirectiveBundleUnlock() { 4584 if (checkForValidSection() || 4585 parseToken(AsmToken::EndOfStatement, 4586 "unexpected token in '.bundle_unlock' directive")) 4587 return true; 4588 4589 getStreamer().EmitBundleUnlock(); 4590 return false; 4591 } 4592 4593 /// parseDirectiveSpace 4594 /// ::= (.skip | .space) expression [ , expression ] 4595 bool AsmParser::parseDirectiveSpace(StringRef IDVal) { 4596 SMLoc NumBytesLoc = Lexer.getLoc(); 4597 const MCExpr *NumBytes; 4598 if (checkForValidSection() || parseExpression(NumBytes)) 4599 return true; 4600 4601 int64_t FillExpr = 0; 4602 if (parseOptionalToken(AsmToken::Comma)) 4603 if (parseAbsoluteExpression(FillExpr)) 4604 return addErrorSuffix("in '" + Twine(IDVal) + "' directive"); 4605 if (parseToken(AsmToken::EndOfStatement)) 4606 return addErrorSuffix("in '" + Twine(IDVal) + "' directive"); 4607 4608 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 4609 getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc); 4610 4611 return false; 4612 } 4613 4614 /// parseDirectiveDCB 4615 /// ::= .dcb.{b, l, w} expression, expression 4616 bool AsmParser::parseDirectiveDCB(StringRef IDVal, unsigned Size) { 4617 SMLoc NumValuesLoc = Lexer.getLoc(); 4618 int64_t NumValues; 4619 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4620 return true; 4621 4622 if (NumValues < 0) { 4623 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4624 return false; 4625 } 4626 4627 if (parseToken(AsmToken::Comma, 4628 "unexpected token in '" + Twine(IDVal) + "' directive")) 4629 return true; 4630 4631 const MCExpr *Value; 4632 SMLoc ExprLoc = getLexer().getLoc(); 4633 if (parseExpression(Value)) 4634 return true; 4635 4636 // Special case constant expressions to match code generator. 4637 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4638 assert(Size <= 8 && "Invalid size"); 4639 uint64_t IntValue = MCE->getValue(); 4640 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 4641 return Error(ExprLoc, "literal value out of range for directive"); 4642 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4643 getStreamer().EmitIntValue(IntValue, Size); 4644 } else { 4645 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4646 getStreamer().EmitValue(Value, Size, ExprLoc); 4647 } 4648 4649 if (parseToken(AsmToken::EndOfStatement, 4650 "unexpected token in '" + Twine(IDVal) + "' directive")) 4651 return true; 4652 4653 return false; 4654 } 4655 4656 /// parseDirectiveRealDCB 4657 /// ::= .dcb.{d, s} expression, expression 4658 bool AsmParser::parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &Semantics) { 4659 SMLoc NumValuesLoc = Lexer.getLoc(); 4660 int64_t NumValues; 4661 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4662 return true; 4663 4664 if (NumValues < 0) { 4665 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4666 return false; 4667 } 4668 4669 if (parseToken(AsmToken::Comma, 4670 "unexpected token in '" + Twine(IDVal) + "' directive")) 4671 return true; 4672 4673 APInt AsInt; 4674 if (parseRealValue(Semantics, AsInt)) 4675 return true; 4676 4677 if (parseToken(AsmToken::EndOfStatement, 4678 "unexpected token in '" + Twine(IDVal) + "' directive")) 4679 return true; 4680 4681 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4682 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 4683 AsInt.getBitWidth() / 8); 4684 4685 return false; 4686 } 4687 4688 /// parseDirectiveDS 4689 /// ::= .ds.{b, d, l, p, s, w, x} expression 4690 bool AsmParser::parseDirectiveDS(StringRef IDVal, unsigned Size) { 4691 SMLoc NumValuesLoc = Lexer.getLoc(); 4692 int64_t NumValues; 4693 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4694 return true; 4695 4696 if (NumValues < 0) { 4697 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4698 return false; 4699 } 4700 4701 if (parseToken(AsmToken::EndOfStatement, 4702 "unexpected token in '" + Twine(IDVal) + "' directive")) 4703 return true; 4704 4705 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4706 getStreamer().emitFill(Size, 0); 4707 4708 return false; 4709 } 4710 4711 /// parseDirectiveLEB128 4712 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ] 4713 bool AsmParser::parseDirectiveLEB128(bool Signed) { 4714 if (checkForValidSection()) 4715 return true; 4716 4717 auto parseOp = [&]() -> bool { 4718 const MCExpr *Value; 4719 if (parseExpression(Value)) 4720 return true; 4721 if (Signed) 4722 getStreamer().EmitSLEB128Value(Value); 4723 else 4724 getStreamer().EmitULEB128Value(Value); 4725 return false; 4726 }; 4727 4728 if (parseMany(parseOp)) 4729 return addErrorSuffix(" in directive"); 4730 4731 return false; 4732 } 4733 4734 /// parseDirectiveSymbolAttribute 4735 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 4736 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 4737 auto parseOp = [&]() -> bool { 4738 StringRef Name; 4739 SMLoc Loc = getTok().getLoc(); 4740 if (parseIdentifier(Name)) 4741 return Error(Loc, "expected identifier"); 4742 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4743 4744 // Assembler local symbols don't make any sense here. Complain loudly. 4745 if (Sym->isTemporary()) 4746 return Error(Loc, "non-local symbol required"); 4747 4748 if (!getStreamer().EmitSymbolAttribute(Sym, Attr)) 4749 return Error(Loc, "unable to emit symbol attribute"); 4750 return false; 4751 }; 4752 4753 if (parseMany(parseOp)) 4754 return addErrorSuffix(" in directive"); 4755 return false; 4756 } 4757 4758 /// parseDirectiveComm 4759 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 4760 bool AsmParser::parseDirectiveComm(bool IsLocal) { 4761 if (checkForValidSection()) 4762 return true; 4763 4764 SMLoc IDLoc = getLexer().getLoc(); 4765 StringRef Name; 4766 if (parseIdentifier(Name)) 4767 return TokError("expected identifier in directive"); 4768 4769 // Handle the identifier as the key symbol. 4770 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4771 4772 if (getLexer().isNot(AsmToken::Comma)) 4773 return TokError("unexpected token in directive"); 4774 Lex(); 4775 4776 int64_t Size; 4777 SMLoc SizeLoc = getLexer().getLoc(); 4778 if (parseAbsoluteExpression(Size)) 4779 return true; 4780 4781 int64_t Pow2Alignment = 0; 4782 SMLoc Pow2AlignmentLoc; 4783 if (getLexer().is(AsmToken::Comma)) { 4784 Lex(); 4785 Pow2AlignmentLoc = getLexer().getLoc(); 4786 if (parseAbsoluteExpression(Pow2Alignment)) 4787 return true; 4788 4789 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); 4790 if (IsLocal && LCOMM == LCOMM::NoAlignment) 4791 return Error(Pow2AlignmentLoc, "alignment not supported on this target"); 4792 4793 // If this target takes alignments in bytes (not log) validate and convert. 4794 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || 4795 (IsLocal && LCOMM == LCOMM::ByteAlignment)) { 4796 if (!isPowerOf2_64(Pow2Alignment)) 4797 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 4798 Pow2Alignment = Log2_64(Pow2Alignment); 4799 } 4800 } 4801 4802 if (parseToken(AsmToken::EndOfStatement, 4803 "unexpected token in '.comm' or '.lcomm' directive")) 4804 return true; 4805 4806 // NOTE: a size of zero for a .comm should create a undefined symbol 4807 // but a size of .lcomm creates a bss symbol of size zero. 4808 if (Size < 0) 4809 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 4810 "be less than zero"); 4811 4812 // NOTE: The alignment in the directive is a power of 2 value, the assembler 4813 // may internally end up wanting an alignment in bytes. 4814 // FIXME: Diagnose overflow. 4815 if (Pow2Alignment < 0) 4816 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 4817 "alignment, can't be less than zero"); 4818 4819 Sym->redefineIfPossible(); 4820 if (!Sym->isUndefined()) 4821 return Error(IDLoc, "invalid symbol redefinition"); 4822 4823 // Create the Symbol as a common or local common with Size and Pow2Alignment 4824 if (IsLocal) { 4825 getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment); 4826 return false; 4827 } 4828 4829 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 4830 return false; 4831 } 4832 4833 /// parseDirectiveAbort 4834 /// ::= .abort [... message ...] 4835 bool AsmParser::parseDirectiveAbort() { 4836 // FIXME: Use loc from directive. 4837 SMLoc Loc = getLexer().getLoc(); 4838 4839 StringRef Str = parseStringToEndOfStatement(); 4840 if (parseToken(AsmToken::EndOfStatement, 4841 "unexpected token in '.abort' directive")) 4842 return true; 4843 4844 if (Str.empty()) 4845 return Error(Loc, ".abort detected. Assembly stopping."); 4846 else 4847 return Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 4848 // FIXME: Actually abort assembly here. 4849 4850 return false; 4851 } 4852 4853 /// parseDirectiveInclude 4854 /// ::= .include "filename" 4855 bool AsmParser::parseDirectiveInclude() { 4856 // Allow the strings to have escaped octal character sequence. 4857 std::string Filename; 4858 SMLoc IncludeLoc = getTok().getLoc(); 4859 4860 if (check(getTok().isNot(AsmToken::String), 4861 "expected string in '.include' directive") || 4862 parseEscapedString(Filename) || 4863 check(getTok().isNot(AsmToken::EndOfStatement), 4864 "unexpected token in '.include' directive") || 4865 // Attempt to switch the lexer to the included file before consuming the 4866 // end of statement to avoid losing it when we switch. 4867 check(enterIncludeFile(Filename), IncludeLoc, 4868 "Could not find include file '" + Filename + "'")) 4869 return true; 4870 4871 return false; 4872 } 4873 4874 /// parseDirectiveIncbin 4875 /// ::= .incbin "filename" [ , skip [ , count ] ] 4876 bool AsmParser::parseDirectiveIncbin() { 4877 // Allow the strings to have escaped octal character sequence. 4878 std::string Filename; 4879 SMLoc IncbinLoc = getTok().getLoc(); 4880 if (check(getTok().isNot(AsmToken::String), 4881 "expected string in '.incbin' directive") || 4882 parseEscapedString(Filename)) 4883 return true; 4884 4885 int64_t Skip = 0; 4886 const MCExpr *Count = nullptr; 4887 SMLoc SkipLoc, CountLoc; 4888 if (parseOptionalToken(AsmToken::Comma)) { 4889 // The skip expression can be omitted while specifying the count, e.g: 4890 // .incbin "filename",,4 4891 if (getTok().isNot(AsmToken::Comma)) { 4892 if (parseTokenLoc(SkipLoc) || parseAbsoluteExpression(Skip)) 4893 return true; 4894 } 4895 if (parseOptionalToken(AsmToken::Comma)) { 4896 CountLoc = getTok().getLoc(); 4897 if (parseExpression(Count)) 4898 return true; 4899 } 4900 } 4901 4902 if (parseToken(AsmToken::EndOfStatement, 4903 "unexpected token in '.incbin' directive")) 4904 return true; 4905 4906 if (check(Skip < 0, SkipLoc, "skip is negative")) 4907 return true; 4908 4909 // Attempt to process the included file. 4910 if (processIncbinFile(Filename, Skip, Count, CountLoc)) 4911 return Error(IncbinLoc, "Could not find incbin file '" + Filename + "'"); 4912 return false; 4913 } 4914 4915 /// parseDirectiveIf 4916 /// ::= .if{,eq,ge,gt,le,lt,ne} expression 4917 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { 4918 TheCondStack.push_back(TheCondState); 4919 TheCondState.TheCond = AsmCond::IfCond; 4920 if (TheCondState.Ignore) { 4921 eatToEndOfStatement(); 4922 } else { 4923 int64_t ExprValue; 4924 if (parseAbsoluteExpression(ExprValue) || 4925 parseToken(AsmToken::EndOfStatement, 4926 "unexpected token in '.if' directive")) 4927 return true; 4928 4929 switch (DirKind) { 4930 default: 4931 llvm_unreachable("unsupported directive"); 4932 case DK_IF: 4933 case DK_IFNE: 4934 break; 4935 case DK_IFEQ: 4936 ExprValue = ExprValue == 0; 4937 break; 4938 case DK_IFGE: 4939 ExprValue = ExprValue >= 0; 4940 break; 4941 case DK_IFGT: 4942 ExprValue = ExprValue > 0; 4943 break; 4944 case DK_IFLE: 4945 ExprValue = ExprValue <= 0; 4946 break; 4947 case DK_IFLT: 4948 ExprValue = ExprValue < 0; 4949 break; 4950 } 4951 4952 TheCondState.CondMet = ExprValue; 4953 TheCondState.Ignore = !TheCondState.CondMet; 4954 } 4955 4956 return false; 4957 } 4958 4959 /// parseDirectiveIfb 4960 /// ::= .ifb string 4961 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 4962 TheCondStack.push_back(TheCondState); 4963 TheCondState.TheCond = AsmCond::IfCond; 4964 4965 if (TheCondState.Ignore) { 4966 eatToEndOfStatement(); 4967 } else { 4968 StringRef Str = parseStringToEndOfStatement(); 4969 4970 if (parseToken(AsmToken::EndOfStatement, 4971 "unexpected token in '.ifb' directive")) 4972 return true; 4973 4974 TheCondState.CondMet = ExpectBlank == Str.empty(); 4975 TheCondState.Ignore = !TheCondState.CondMet; 4976 } 4977 4978 return false; 4979 } 4980 4981 /// parseDirectiveIfc 4982 /// ::= .ifc string1, string2 4983 /// ::= .ifnc string1, string2 4984 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) { 4985 TheCondStack.push_back(TheCondState); 4986 TheCondState.TheCond = AsmCond::IfCond; 4987 4988 if (TheCondState.Ignore) { 4989 eatToEndOfStatement(); 4990 } else { 4991 StringRef Str1 = parseStringToComma(); 4992 4993 if (parseToken(AsmToken::Comma, "unexpected token in '.ifc' directive")) 4994 return true; 4995 4996 StringRef Str2 = parseStringToEndOfStatement(); 4997 4998 if (parseToken(AsmToken::EndOfStatement, 4999 "unexpected token in '.ifc' directive")) 5000 return true; 5001 5002 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim()); 5003 TheCondState.Ignore = !TheCondState.CondMet; 5004 } 5005 5006 return false; 5007 } 5008 5009 /// parseDirectiveIfeqs 5010 /// ::= .ifeqs string1, string2 5011 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) { 5012 if (Lexer.isNot(AsmToken::String)) { 5013 if (ExpectEqual) 5014 return TokError("expected string parameter for '.ifeqs' directive"); 5015 return TokError("expected string parameter for '.ifnes' directive"); 5016 } 5017 5018 StringRef String1 = getTok().getStringContents(); 5019 Lex(); 5020 5021 if (Lexer.isNot(AsmToken::Comma)) { 5022 if (ExpectEqual) 5023 return TokError( 5024 "expected comma after first string for '.ifeqs' directive"); 5025 return TokError("expected comma after first string for '.ifnes' directive"); 5026 } 5027 5028 Lex(); 5029 5030 if (Lexer.isNot(AsmToken::String)) { 5031 if (ExpectEqual) 5032 return TokError("expected string parameter for '.ifeqs' directive"); 5033 return TokError("expected string parameter for '.ifnes' directive"); 5034 } 5035 5036 StringRef String2 = getTok().getStringContents(); 5037 Lex(); 5038 5039 TheCondStack.push_back(TheCondState); 5040 TheCondState.TheCond = AsmCond::IfCond; 5041 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5042 TheCondState.Ignore = !TheCondState.CondMet; 5043 5044 return false; 5045 } 5046 5047 /// parseDirectiveIfdef 5048 /// ::= .ifdef symbol 5049 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 5050 StringRef Name; 5051 TheCondStack.push_back(TheCondState); 5052 TheCondState.TheCond = AsmCond::IfCond; 5053 5054 if (TheCondState.Ignore) { 5055 eatToEndOfStatement(); 5056 } else { 5057 if (check(parseIdentifier(Name), "expected identifier after '.ifdef'") || 5058 parseToken(AsmToken::EndOfStatement, "unexpected token in '.ifdef'")) 5059 return true; 5060 5061 MCSymbol *Sym = getContext().lookupSymbol(Name); 5062 5063 if (expect_defined) 5064 TheCondState.CondMet = (Sym && !Sym->isUndefined(false)); 5065 else 5066 TheCondState.CondMet = (!Sym || Sym->isUndefined(false)); 5067 TheCondState.Ignore = !TheCondState.CondMet; 5068 } 5069 5070 return false; 5071 } 5072 5073 /// parseDirectiveElseIf 5074 /// ::= .elseif expression 5075 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) { 5076 if (TheCondState.TheCond != AsmCond::IfCond && 5077 TheCondState.TheCond != AsmCond::ElseIfCond) 5078 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an" 5079 " .if or an .elseif"); 5080 TheCondState.TheCond = AsmCond::ElseIfCond; 5081 5082 bool LastIgnoreState = false; 5083 if (!TheCondStack.empty()) 5084 LastIgnoreState = TheCondStack.back().Ignore; 5085 if (LastIgnoreState || TheCondState.CondMet) { 5086 TheCondState.Ignore = true; 5087 eatToEndOfStatement(); 5088 } else { 5089 int64_t ExprValue; 5090 if (parseAbsoluteExpression(ExprValue)) 5091 return true; 5092 5093 if (parseToken(AsmToken::EndOfStatement, 5094 "unexpected token in '.elseif' directive")) 5095 return true; 5096 5097 TheCondState.CondMet = ExprValue; 5098 TheCondState.Ignore = !TheCondState.CondMet; 5099 } 5100 5101 return false; 5102 } 5103 5104 /// parseDirectiveElse 5105 /// ::= .else 5106 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) { 5107 if (parseToken(AsmToken::EndOfStatement, 5108 "unexpected token in '.else' directive")) 5109 return true; 5110 5111 if (TheCondState.TheCond != AsmCond::IfCond && 5112 TheCondState.TheCond != AsmCond::ElseIfCond) 5113 return Error(DirectiveLoc, "Encountered a .else that doesn't follow " 5114 " an .if or an .elseif"); 5115 TheCondState.TheCond = AsmCond::ElseCond; 5116 bool LastIgnoreState = false; 5117 if (!TheCondStack.empty()) 5118 LastIgnoreState = TheCondStack.back().Ignore; 5119 if (LastIgnoreState || TheCondState.CondMet) 5120 TheCondState.Ignore = true; 5121 else 5122 TheCondState.Ignore = false; 5123 5124 return false; 5125 } 5126 5127 /// parseDirectiveEnd 5128 /// ::= .end 5129 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { 5130 if (parseToken(AsmToken::EndOfStatement, 5131 "unexpected token in '.end' directive")) 5132 return true; 5133 5134 while (Lexer.isNot(AsmToken::Eof)) 5135 Lexer.Lex(); 5136 5137 return false; 5138 } 5139 5140 /// parseDirectiveError 5141 /// ::= .err 5142 /// ::= .error [string] 5143 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) { 5144 if (!TheCondStack.empty()) { 5145 if (TheCondStack.back().Ignore) { 5146 eatToEndOfStatement(); 5147 return false; 5148 } 5149 } 5150 5151 if (!WithMessage) 5152 return Error(L, ".err encountered"); 5153 5154 StringRef Message = ".error directive invoked in source file"; 5155 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5156 if (Lexer.isNot(AsmToken::String)) 5157 return TokError(".error argument must be a string"); 5158 5159 Message = getTok().getStringContents(); 5160 Lex(); 5161 } 5162 5163 return Error(L, Message); 5164 } 5165 5166 /// parseDirectiveWarning 5167 /// ::= .warning [string] 5168 bool AsmParser::parseDirectiveWarning(SMLoc L) { 5169 if (!TheCondStack.empty()) { 5170 if (TheCondStack.back().Ignore) { 5171 eatToEndOfStatement(); 5172 return false; 5173 } 5174 } 5175 5176 StringRef Message = ".warning directive invoked in source file"; 5177 5178 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 5179 if (Lexer.isNot(AsmToken::String)) 5180 return TokError(".warning argument must be a string"); 5181 5182 Message = getTok().getStringContents(); 5183 Lex(); 5184 if (parseToken(AsmToken::EndOfStatement, 5185 "expected end of statement in '.warning' directive")) 5186 return true; 5187 } 5188 5189 return Warning(L, Message); 5190 } 5191 5192 /// parseDirectiveEndIf 5193 /// ::= .endif 5194 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { 5195 if (parseToken(AsmToken::EndOfStatement, 5196 "unexpected token in '.endif' directive")) 5197 return true; 5198 5199 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) 5200 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow " 5201 "an .if or .else"); 5202 if (!TheCondStack.empty()) { 5203 TheCondState = TheCondStack.back(); 5204 TheCondStack.pop_back(); 5205 } 5206 5207 return false; 5208 } 5209 5210 void AsmParser::initializeDirectiveKindMap() { 5211 DirectiveKindMap[".set"] = DK_SET; 5212 DirectiveKindMap[".equ"] = DK_EQU; 5213 DirectiveKindMap[".equiv"] = DK_EQUIV; 5214 DirectiveKindMap[".ascii"] = DK_ASCII; 5215 DirectiveKindMap[".asciz"] = DK_ASCIZ; 5216 DirectiveKindMap[".string"] = DK_STRING; 5217 DirectiveKindMap[".byte"] = DK_BYTE; 5218 DirectiveKindMap[".short"] = DK_SHORT; 5219 DirectiveKindMap[".value"] = DK_VALUE; 5220 DirectiveKindMap[".2byte"] = DK_2BYTE; 5221 DirectiveKindMap[".long"] = DK_LONG; 5222 DirectiveKindMap[".int"] = DK_INT; 5223 DirectiveKindMap[".4byte"] = DK_4BYTE; 5224 DirectiveKindMap[".quad"] = DK_QUAD; 5225 DirectiveKindMap[".8byte"] = DK_8BYTE; 5226 DirectiveKindMap[".octa"] = DK_OCTA; 5227 DirectiveKindMap[".single"] = DK_SINGLE; 5228 DirectiveKindMap[".float"] = DK_FLOAT; 5229 DirectiveKindMap[".double"] = DK_DOUBLE; 5230 DirectiveKindMap[".align"] = DK_ALIGN; 5231 DirectiveKindMap[".align32"] = DK_ALIGN32; 5232 DirectiveKindMap[".balign"] = DK_BALIGN; 5233 DirectiveKindMap[".balignw"] = DK_BALIGNW; 5234 DirectiveKindMap[".balignl"] = DK_BALIGNL; 5235 DirectiveKindMap[".p2align"] = DK_P2ALIGN; 5236 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW; 5237 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL; 5238 DirectiveKindMap[".org"] = DK_ORG; 5239 DirectiveKindMap[".fill"] = DK_FILL; 5240 DirectiveKindMap[".zero"] = DK_ZERO; 5241 DirectiveKindMap[".extern"] = DK_EXTERN; 5242 DirectiveKindMap[".globl"] = DK_GLOBL; 5243 DirectiveKindMap[".global"] = DK_GLOBAL; 5244 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE; 5245 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP; 5246 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER; 5247 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN; 5248 DirectiveKindMap[".reference"] = DK_REFERENCE; 5249 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION; 5250 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE; 5251 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN; 5252 DirectiveKindMap[".cold"] = DK_COLD; 5253 DirectiveKindMap[".comm"] = DK_COMM; 5254 DirectiveKindMap[".common"] = DK_COMMON; 5255 DirectiveKindMap[".lcomm"] = DK_LCOMM; 5256 DirectiveKindMap[".abort"] = DK_ABORT; 5257 DirectiveKindMap[".include"] = DK_INCLUDE; 5258 DirectiveKindMap[".incbin"] = DK_INCBIN; 5259 DirectiveKindMap[".code16"] = DK_CODE16; 5260 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC; 5261 DirectiveKindMap[".rept"] = DK_REPT; 5262 DirectiveKindMap[".rep"] = DK_REPT; 5263 DirectiveKindMap[".irp"] = DK_IRP; 5264 DirectiveKindMap[".irpc"] = DK_IRPC; 5265 DirectiveKindMap[".endr"] = DK_ENDR; 5266 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE; 5267 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK; 5268 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK; 5269 DirectiveKindMap[".if"] = DK_IF; 5270 DirectiveKindMap[".ifeq"] = DK_IFEQ; 5271 DirectiveKindMap[".ifge"] = DK_IFGE; 5272 DirectiveKindMap[".ifgt"] = DK_IFGT; 5273 DirectiveKindMap[".ifle"] = DK_IFLE; 5274 DirectiveKindMap[".iflt"] = DK_IFLT; 5275 DirectiveKindMap[".ifne"] = DK_IFNE; 5276 DirectiveKindMap[".ifb"] = DK_IFB; 5277 DirectiveKindMap[".ifnb"] = DK_IFNB; 5278 DirectiveKindMap[".ifc"] = DK_IFC; 5279 DirectiveKindMap[".ifeqs"] = DK_IFEQS; 5280 DirectiveKindMap[".ifnc"] = DK_IFNC; 5281 DirectiveKindMap[".ifnes"] = DK_IFNES; 5282 DirectiveKindMap[".ifdef"] = DK_IFDEF; 5283 DirectiveKindMap[".ifndef"] = DK_IFNDEF; 5284 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF; 5285 DirectiveKindMap[".elseif"] = DK_ELSEIF; 5286 DirectiveKindMap[".else"] = DK_ELSE; 5287 DirectiveKindMap[".end"] = DK_END; 5288 DirectiveKindMap[".endif"] = DK_ENDIF; 5289 DirectiveKindMap[".skip"] = DK_SKIP; 5290 DirectiveKindMap[".space"] = DK_SPACE; 5291 DirectiveKindMap[".file"] = DK_FILE; 5292 DirectiveKindMap[".line"] = DK_LINE; 5293 DirectiveKindMap[".loc"] = DK_LOC; 5294 DirectiveKindMap[".stabs"] = DK_STABS; 5295 DirectiveKindMap[".cv_file"] = DK_CV_FILE; 5296 DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; 5297 DirectiveKindMap[".cv_loc"] = DK_CV_LOC; 5298 DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; 5299 DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; 5300 DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; 5301 DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; 5302 DirectiveKindMap[".cv_string"] = DK_CV_STRING; 5303 DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; 5304 DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; 5305 DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; 5306 DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; 5307 DirectiveKindMap[".sleb128"] = DK_SLEB128; 5308 DirectiveKindMap[".uleb128"] = DK_ULEB128; 5309 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; 5310 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; 5311 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; 5312 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; 5313 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; 5314 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; 5315 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; 5316 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; 5317 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; 5318 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; 5319 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; 5320 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; 5321 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; 5322 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; 5323 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; 5324 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; 5325 DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; 5326 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; 5327 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; 5328 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; 5329 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; 5330 DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; 5331 DirectiveKindMap[".macros_on"] = DK_MACROS_ON; 5332 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF; 5333 DirectiveKindMap[".macro"] = DK_MACRO; 5334 DirectiveKindMap[".exitm"] = DK_EXITM; 5335 DirectiveKindMap[".endm"] = DK_ENDM; 5336 DirectiveKindMap[".endmacro"] = DK_ENDMACRO; 5337 DirectiveKindMap[".purgem"] = DK_PURGEM; 5338 DirectiveKindMap[".err"] = DK_ERR; 5339 DirectiveKindMap[".error"] = DK_ERROR; 5340 DirectiveKindMap[".warning"] = DK_WARNING; 5341 DirectiveKindMap[".altmacro"] = DK_ALTMACRO; 5342 DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO; 5343 DirectiveKindMap[".reloc"] = DK_RELOC; 5344 DirectiveKindMap[".dc"] = DK_DC; 5345 DirectiveKindMap[".dc.a"] = DK_DC_A; 5346 DirectiveKindMap[".dc.b"] = DK_DC_B; 5347 DirectiveKindMap[".dc.d"] = DK_DC_D; 5348 DirectiveKindMap[".dc.l"] = DK_DC_L; 5349 DirectiveKindMap[".dc.s"] = DK_DC_S; 5350 DirectiveKindMap[".dc.w"] = DK_DC_W; 5351 DirectiveKindMap[".dc.x"] = DK_DC_X; 5352 DirectiveKindMap[".dcb"] = DK_DCB; 5353 DirectiveKindMap[".dcb.b"] = DK_DCB_B; 5354 DirectiveKindMap[".dcb.d"] = DK_DCB_D; 5355 DirectiveKindMap[".dcb.l"] = DK_DCB_L; 5356 DirectiveKindMap[".dcb.s"] = DK_DCB_S; 5357 DirectiveKindMap[".dcb.w"] = DK_DCB_W; 5358 DirectiveKindMap[".dcb.x"] = DK_DCB_X; 5359 DirectiveKindMap[".ds"] = DK_DS; 5360 DirectiveKindMap[".ds.b"] = DK_DS_B; 5361 DirectiveKindMap[".ds.d"] = DK_DS_D; 5362 DirectiveKindMap[".ds.l"] = DK_DS_L; 5363 DirectiveKindMap[".ds.p"] = DK_DS_P; 5364 DirectiveKindMap[".ds.s"] = DK_DS_S; 5365 DirectiveKindMap[".ds.w"] = DK_DS_W; 5366 DirectiveKindMap[".ds.x"] = DK_DS_X; 5367 DirectiveKindMap[".print"] = DK_PRINT; 5368 DirectiveKindMap[".addrsig"] = DK_ADDRSIG; 5369 DirectiveKindMap[".addrsig_sym"] = DK_ADDRSIG_SYM; 5370 } 5371 5372 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { 5373 AsmToken EndToken, StartToken = getTok(); 5374 5375 unsigned NestLevel = 0; 5376 while (true) { 5377 // Check whether we have reached the end of the file. 5378 if (getLexer().is(AsmToken::Eof)) { 5379 printError(DirectiveLoc, "no matching '.endr' in definition"); 5380 return nullptr; 5381 } 5382 5383 if (Lexer.is(AsmToken::Identifier) && 5384 (getTok().getIdentifier() == ".rep" || 5385 getTok().getIdentifier() == ".rept" || 5386 getTok().getIdentifier() == ".irp" || 5387 getTok().getIdentifier() == ".irpc")) { 5388 ++NestLevel; 5389 } 5390 5391 // Otherwise, check whether we have reached the .endr. 5392 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") { 5393 if (NestLevel == 0) { 5394 EndToken = getTok(); 5395 Lex(); 5396 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5397 printError(getTok().getLoc(), 5398 "unexpected token in '.endr' directive"); 5399 return nullptr; 5400 } 5401 break; 5402 } 5403 --NestLevel; 5404 } 5405 5406 // Otherwise, scan till the end of the statement. 5407 eatToEndOfStatement(); 5408 } 5409 5410 const char *BodyStart = StartToken.getLoc().getPointer(); 5411 const char *BodyEnd = EndToken.getLoc().getPointer(); 5412 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 5413 5414 // We Are Anonymous. 5415 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters()); 5416 return &MacroLikeBodies.back(); 5417 } 5418 5419 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 5420 raw_svector_ostream &OS) { 5421 OS << ".endr\n"; 5422 5423 std::unique_ptr<MemoryBuffer> Instantiation = 5424 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 5425 5426 // Create the macro instantiation object and add to the current macro 5427 // instantiation stack. 5428 MacroInstantiation *MI = new MacroInstantiation( 5429 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()); 5430 ActiveMacros.push_back(MI); 5431 5432 // Jump to the macro instantiation and prime the lexer. 5433 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 5434 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 5435 Lex(); 5436 } 5437 5438 /// parseDirectiveRept 5439 /// ::= .rep | .rept count 5440 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) { 5441 const MCExpr *CountExpr; 5442 SMLoc CountLoc = getTok().getLoc(); 5443 if (parseExpression(CountExpr)) 5444 return true; 5445 5446 int64_t Count; 5447 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) { 5448 return Error(CountLoc, "unexpected token in '" + Dir + "' directive"); 5449 } 5450 5451 if (check(Count < 0, CountLoc, "Count is negative") || 5452 parseToken(AsmToken::EndOfStatement, 5453 "unexpected token in '" + Dir + "' directive")) 5454 return true; 5455 5456 // Lex the rept definition. 5457 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5458 if (!M) 5459 return true; 5460 5461 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5462 // to hold the macro body with substitutions. 5463 SmallString<256> Buf; 5464 raw_svector_ostream OS(Buf); 5465 while (Count--) { 5466 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t). 5467 if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc())) 5468 return true; 5469 } 5470 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5471 5472 return false; 5473 } 5474 5475 /// parseDirectiveIrp 5476 /// ::= .irp symbol,values 5477 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) { 5478 MCAsmMacroParameter Parameter; 5479 MCAsmMacroArguments A; 5480 if (check(parseIdentifier(Parameter.Name), 5481 "expected identifier in '.irp' directive") || 5482 parseToken(AsmToken::Comma, "expected comma in '.irp' directive") || 5483 parseMacroArguments(nullptr, A) || 5484 parseToken(AsmToken::EndOfStatement, "expected End of Statement")) 5485 return true; 5486 5487 // Lex the irp definition. 5488 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5489 if (!M) 5490 return true; 5491 5492 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5493 // to hold the macro body with substitutions. 5494 SmallString<256> Buf; 5495 raw_svector_ostream OS(Buf); 5496 5497 for (const MCAsmMacroArgument &Arg : A) { 5498 // Note that the AtPseudoVariable is enabled for instantiations of .irp. 5499 // This is undocumented, but GAS seems to support it. 5500 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5501 return true; 5502 } 5503 5504 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5505 5506 return false; 5507 } 5508 5509 /// parseDirectiveIrpc 5510 /// ::= .irpc symbol,values 5511 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) { 5512 MCAsmMacroParameter Parameter; 5513 MCAsmMacroArguments A; 5514 5515 if (check(parseIdentifier(Parameter.Name), 5516 "expected identifier in '.irpc' directive") || 5517 parseToken(AsmToken::Comma, "expected comma in '.irpc' directive") || 5518 parseMacroArguments(nullptr, A)) 5519 return true; 5520 5521 if (A.size() != 1 || A.front().size() != 1) 5522 return TokError("unexpected token in '.irpc' directive"); 5523 5524 // Eat the end of statement. 5525 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 5526 return true; 5527 5528 // Lex the irpc definition. 5529 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5530 if (!M) 5531 return true; 5532 5533 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5534 // to hold the macro body with substitutions. 5535 SmallString<256> Buf; 5536 raw_svector_ostream OS(Buf); 5537 5538 StringRef Values = A.front().front().getString(); 5539 for (std::size_t I = 0, End = Values.size(); I != End; ++I) { 5540 MCAsmMacroArgument Arg; 5541 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1)); 5542 5543 // Note that the AtPseudoVariable is enabled for instantiations of .irpc. 5544 // This is undocumented, but GAS seems to support it. 5545 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5546 return true; 5547 } 5548 5549 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5550 5551 return false; 5552 } 5553 5554 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) { 5555 if (ActiveMacros.empty()) 5556 return TokError("unmatched '.endr' directive"); 5557 5558 // The only .repl that should get here are the ones created by 5559 // instantiateMacroLikeBody. 5560 assert(getLexer().is(AsmToken::EndOfStatement)); 5561 5562 handleMacroExit(); 5563 return false; 5564 } 5565 5566 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, 5567 size_t Len) { 5568 const MCExpr *Value; 5569 SMLoc ExprLoc = getLexer().getLoc(); 5570 if (parseExpression(Value)) 5571 return true; 5572 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5573 if (!MCE) 5574 return Error(ExprLoc, "unexpected expression in _emit"); 5575 uint64_t IntValue = MCE->getValue(); 5576 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue)) 5577 return Error(ExprLoc, "literal value out of range for directive"); 5578 5579 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len); 5580 return false; 5581 } 5582 5583 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { 5584 const MCExpr *Value; 5585 SMLoc ExprLoc = getLexer().getLoc(); 5586 if (parseExpression(Value)) 5587 return true; 5588 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5589 if (!MCE) 5590 return Error(ExprLoc, "unexpected expression in align"); 5591 uint64_t IntValue = MCE->getValue(); 5592 if (!isPowerOf2_64(IntValue)) 5593 return Error(ExprLoc, "literal value not a power of two greater then zero"); 5594 5595 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue)); 5596 return false; 5597 } 5598 5599 bool AsmParser::parseDirectivePrint(SMLoc DirectiveLoc) { 5600 const AsmToken StrTok = getTok(); 5601 Lex(); 5602 if (StrTok.isNot(AsmToken::String) || StrTok.getString().front() != '"') 5603 return Error(DirectiveLoc, "expected double quoted string after .print"); 5604 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 5605 return true; 5606 llvm::outs() << StrTok.getStringContents() << '\n'; 5607 return false; 5608 } 5609 5610 bool AsmParser::parseDirectiveAddrsig() { 5611 getStreamer().EmitAddrsig(); 5612 return false; 5613 } 5614 5615 bool AsmParser::parseDirectiveAddrsigSym() { 5616 StringRef Name; 5617 if (check(parseIdentifier(Name), 5618 "expected identifier in '.addrsig_sym' directive")) 5619 return true; 5620 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5621 getStreamer().EmitAddrsigSym(Sym); 5622 return false; 5623 } 5624 5625 // We are comparing pointers, but the pointers are relative to a single string. 5626 // Thus, this should always be deterministic. 5627 static int rewritesSort(const AsmRewrite *AsmRewriteA, 5628 const AsmRewrite *AsmRewriteB) { 5629 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) 5630 return -1; 5631 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) 5632 return 1; 5633 5634 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output 5635 // rewrite to the same location. Make sure the SizeDirective rewrite is 5636 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This 5637 // ensures the sort algorithm is stable. 5638 if (AsmRewritePrecedence[AsmRewriteA->Kind] > 5639 AsmRewritePrecedence[AsmRewriteB->Kind]) 5640 return -1; 5641 5642 if (AsmRewritePrecedence[AsmRewriteA->Kind] < 5643 AsmRewritePrecedence[AsmRewriteB->Kind]) 5644 return 1; 5645 llvm_unreachable("Unstable rewrite sort."); 5646 } 5647 5648 bool AsmParser::parseMSInlineAsm( 5649 void *AsmLoc, std::string &AsmString, unsigned &NumOutputs, 5650 unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 5651 SmallVectorImpl<std::string> &Constraints, 5652 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, 5653 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { 5654 SmallVector<void *, 4> InputDecls; 5655 SmallVector<void *, 4> OutputDecls; 5656 SmallVector<bool, 4> InputDeclsAddressOf; 5657 SmallVector<bool, 4> OutputDeclsAddressOf; 5658 SmallVector<std::string, 4> InputConstraints; 5659 SmallVector<std::string, 4> OutputConstraints; 5660 SmallVector<unsigned, 4> ClobberRegs; 5661 5662 SmallVector<AsmRewrite, 4> AsmStrRewrites; 5663 5664 // Prime the lexer. 5665 Lex(); 5666 5667 // While we have input, parse each statement. 5668 unsigned InputIdx = 0; 5669 unsigned OutputIdx = 0; 5670 while (getLexer().isNot(AsmToken::Eof)) { 5671 // Parse curly braces marking block start/end 5672 if (parseCurlyBlockScope(AsmStrRewrites)) 5673 continue; 5674 5675 ParseStatementInfo Info(&AsmStrRewrites); 5676 bool StatementErr = parseStatement(Info, &SI); 5677 5678 if (StatementErr || Info.ParseError) { 5679 // Emit pending errors if any exist. 5680 printPendingErrors(); 5681 return true; 5682 } 5683 5684 // No pending error should exist here. 5685 assert(!hasPendingError() && "unexpected error from parseStatement"); 5686 5687 if (Info.Opcode == ~0U) 5688 continue; 5689 5690 const MCInstrDesc &Desc = MII->get(Info.Opcode); 5691 5692 // Build the list of clobbers, outputs and inputs. 5693 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { 5694 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; 5695 5696 // Immediate. 5697 if (Operand.isImm()) 5698 continue; 5699 5700 // Register operand. 5701 if (Operand.isReg() && !Operand.needAddressOf() && 5702 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) { 5703 unsigned NumDefs = Desc.getNumDefs(); 5704 // Clobber. 5705 if (NumDefs && Operand.getMCOperandNum() < NumDefs) 5706 ClobberRegs.push_back(Operand.getReg()); 5707 continue; 5708 } 5709 5710 // Expr/Input or Output. 5711 StringRef SymName = Operand.getSymName(); 5712 if (SymName.empty()) 5713 continue; 5714 5715 void *OpDecl = Operand.getOpDecl(); 5716 if (!OpDecl) 5717 continue; 5718 5719 bool isOutput = (i == 1) && Desc.mayStore(); 5720 SMLoc Start = SMLoc::getFromPointer(SymName.data()); 5721 if (isOutput) { 5722 ++InputIdx; 5723 OutputDecls.push_back(OpDecl); 5724 OutputDeclsAddressOf.push_back(Operand.needAddressOf()); 5725 OutputConstraints.push_back(("=" + Operand.getConstraint()).str()); 5726 AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size()); 5727 } else { 5728 InputDecls.push_back(OpDecl); 5729 InputDeclsAddressOf.push_back(Operand.needAddressOf()); 5730 InputConstraints.push_back(Operand.getConstraint().str()); 5731 AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size()); 5732 } 5733 } 5734 5735 // Consider implicit defs to be clobbers. Think of cpuid and push. 5736 ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(), 5737 Desc.getNumImplicitDefs()); 5738 ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end()); 5739 } 5740 5741 // Set the number of Outputs and Inputs. 5742 NumOutputs = OutputDecls.size(); 5743 NumInputs = InputDecls.size(); 5744 5745 // Set the unique clobbers. 5746 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end()); 5747 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()), 5748 ClobberRegs.end()); 5749 Clobbers.assign(ClobberRegs.size(), std::string()); 5750 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { 5751 raw_string_ostream OS(Clobbers[I]); 5752 IP->printRegName(OS, ClobberRegs[I]); 5753 } 5754 5755 // Merge the various outputs and inputs. Output are expected first. 5756 if (NumOutputs || NumInputs) { 5757 unsigned NumExprs = NumOutputs + NumInputs; 5758 OpDecls.resize(NumExprs); 5759 Constraints.resize(NumExprs); 5760 for (unsigned i = 0; i < NumOutputs; ++i) { 5761 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]); 5762 Constraints[i] = OutputConstraints[i]; 5763 } 5764 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { 5765 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]); 5766 Constraints[j] = InputConstraints[i]; 5767 } 5768 } 5769 5770 // Build the IR assembly string. 5771 std::string AsmStringIR; 5772 raw_string_ostream OS(AsmStringIR); 5773 StringRef ASMString = 5774 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer(); 5775 const char *AsmStart = ASMString.begin(); 5776 const char *AsmEnd = ASMString.end(); 5777 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort); 5778 for (const AsmRewrite &AR : AsmStrRewrites) { 5779 AsmRewriteKind Kind = AR.Kind; 5780 5781 const char *Loc = AR.Loc.getPointer(); 5782 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!"); 5783 5784 // Emit everything up to the immediate/expression. 5785 if (unsigned Len = Loc - AsmStart) 5786 OS << StringRef(AsmStart, Len); 5787 5788 // Skip the original expression. 5789 if (Kind == AOK_Skip) { 5790 AsmStart = Loc + AR.Len; 5791 continue; 5792 } 5793 5794 unsigned AdditionalSkip = 0; 5795 // Rewrite expressions in $N notation. 5796 switch (Kind) { 5797 default: 5798 break; 5799 case AOK_IntelExpr: 5800 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression"); 5801 if (AR.IntelExp.NeedBracs) 5802 OS << "["; 5803 if (AR.IntelExp.hasBaseReg()) 5804 OS << AR.IntelExp.BaseReg; 5805 if (AR.IntelExp.hasIndexReg()) 5806 OS << (AR.IntelExp.hasBaseReg() ? " + " : "") 5807 << AR.IntelExp.IndexReg; 5808 if (AR.IntelExp.Scale > 1) 5809 OS << " * $$" << AR.IntelExp.Scale; 5810 if (AR.IntelExp.Imm || !AR.IntelExp.hasRegs()) 5811 OS << (AR.IntelExp.hasRegs() ? " + $$" : "$$") << AR.IntelExp.Imm; 5812 if (AR.IntelExp.NeedBracs) 5813 OS << "]"; 5814 break; 5815 case AOK_Label: 5816 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; 5817 break; 5818 case AOK_Input: 5819 OS << '$' << InputIdx++; 5820 break; 5821 case AOK_Output: 5822 OS << '$' << OutputIdx++; 5823 break; 5824 case AOK_SizeDirective: 5825 switch (AR.Val) { 5826 default: break; 5827 case 8: OS << "byte ptr "; break; 5828 case 16: OS << "word ptr "; break; 5829 case 32: OS << "dword ptr "; break; 5830 case 64: OS << "qword ptr "; break; 5831 case 80: OS << "xword ptr "; break; 5832 case 128: OS << "xmmword ptr "; break; 5833 case 256: OS << "ymmword ptr "; break; 5834 } 5835 break; 5836 case AOK_Emit: 5837 OS << ".byte"; 5838 break; 5839 case AOK_Align: { 5840 // MS alignment directives are measured in bytes. If the native assembler 5841 // measures alignment in bytes, we can pass it straight through. 5842 OS << ".align"; 5843 if (getContext().getAsmInfo()->getAlignmentIsInBytes()) 5844 break; 5845 5846 // Alignment is in log2 form, so print that instead and skip the original 5847 // immediate. 5848 unsigned Val = AR.Val; 5849 OS << ' ' << Val; 5850 assert(Val < 10 && "Expected alignment less then 2^10."); 5851 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; 5852 break; 5853 } 5854 case AOK_EVEN: 5855 OS << ".even"; 5856 break; 5857 case AOK_EndOfStatement: 5858 OS << "\n\t"; 5859 break; 5860 } 5861 5862 // Skip the original expression. 5863 AsmStart = Loc + AR.Len + AdditionalSkip; 5864 } 5865 5866 // Emit the remainder of the asm string. 5867 if (AsmStart != AsmEnd) 5868 OS << StringRef(AsmStart, AsmEnd - AsmStart); 5869 5870 AsmString = OS.str(); 5871 return false; 5872 } 5873 5874 namespace llvm { 5875 namespace MCParserUtils { 5876 5877 /// Returns whether the given symbol is used anywhere in the given expression, 5878 /// or subexpressions. 5879 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) { 5880 switch (Value->getKind()) { 5881 case MCExpr::Binary: { 5882 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value); 5883 return isSymbolUsedInExpression(Sym, BE->getLHS()) || 5884 isSymbolUsedInExpression(Sym, BE->getRHS()); 5885 } 5886 case MCExpr::Target: 5887 case MCExpr::Constant: 5888 return false; 5889 case MCExpr::SymbolRef: { 5890 const MCSymbol &S = 5891 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol(); 5892 if (S.isVariable()) 5893 return isSymbolUsedInExpression(Sym, S.getVariableValue()); 5894 return &S == Sym; 5895 } 5896 case MCExpr::Unary: 5897 return isSymbolUsedInExpression( 5898 Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr()); 5899 } 5900 5901 llvm_unreachable("Unknown expr kind!"); 5902 } 5903 5904 bool parseAssignmentExpression(StringRef Name, bool allow_redef, 5905 MCAsmParser &Parser, MCSymbol *&Sym, 5906 const MCExpr *&Value) { 5907 5908 // FIXME: Use better location, we should use proper tokens. 5909 SMLoc EqualLoc = Parser.getTok().getLoc(); 5910 if (Parser.parseExpression(Value)) 5911 return Parser.TokError("missing expression"); 5912 5913 // Note: we don't count b as used in "a = b". This is to allow 5914 // a = b 5915 // b = c 5916 5917 if (Parser.parseToken(AsmToken::EndOfStatement)) 5918 return true; 5919 5920 // Validate that the LHS is allowed to be a variable (either it has not been 5921 // used as a symbol, or it is an absolute symbol). 5922 Sym = Parser.getContext().lookupSymbol(Name); 5923 if (Sym) { 5924 // Diagnose assignment to a label. 5925 // 5926 // FIXME: Diagnostics. Note the location of the definition as a label. 5927 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 5928 if (isSymbolUsedInExpression(Sym, Value)) 5929 return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'"); 5930 else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() && 5931 !Sym->isVariable()) 5932 ; // Allow redefinitions of undefined symbols only used in directives. 5933 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef) 5934 ; // Allow redefinitions of variables that haven't yet been used. 5935 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 5936 return Parser.Error(EqualLoc, "redefinition of '" + Name + "'"); 5937 else if (!Sym->isVariable()) 5938 return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'"); 5939 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 5940 return Parser.Error(EqualLoc, 5941 "invalid reassignment of non-absolute variable '" + 5942 Name + "'"); 5943 } else if (Name == ".") { 5944 Parser.getStreamer().emitValueToOffset(Value, 0, EqualLoc); 5945 return false; 5946 } else 5947 Sym = Parser.getContext().getOrCreateSymbol(Name); 5948 5949 Sym->setRedefinable(allow_redef); 5950 5951 return false; 5952 } 5953 5954 } // end namespace MCParserUtils 5955 } // end namespace llvm 5956 5957 /// Create an MCAsmParser instance. 5958 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C, 5959 MCStreamer &Out, const MCAsmInfo &MAI, 5960 unsigned CB) { 5961 return new AsmParser(SM, C, Out, MAI, CB); 5962 } 5963