• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This class implements the parser for assembly files.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "llvm/ADT/APFloat.h"
15 #include "llvm/ADT/STLExtras.h"
16 #include "llvm/ADT/SmallString.h"
17 #include "llvm/ADT/StringMap.h"
18 #include "llvm/ADT/Twine.h"
19 #include "llvm/MC/MCAsmInfo.h"
20 #include "llvm/MC/MCContext.h"
21 #include "llvm/MC/MCDwarf.h"
22 #include "llvm/MC/MCExpr.h"
23 #include "llvm/MC/MCInstPrinter.h"
24 #include "llvm/MC/MCInstrInfo.h"
25 #include "llvm/MC/MCObjectFileInfo.h"
26 #include "llvm/MC/MCParser/AsmCond.h"
27 #include "llvm/MC/MCParser/AsmLexer.h"
28 #include "llvm/MC/MCParser/MCAsmParser.h"
29 #include "llvm/MC/MCParser/MCAsmParserUtils.h"
30 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
31 #include "llvm/MC/MCParser/MCTargetAsmParser.h"
32 #include "llvm/MC/MCRegisterInfo.h"
33 #include "llvm/MC/MCSectionMachO.h"
34 #include "llvm/MC/MCStreamer.h"
35 #include "llvm/MC/MCSymbol.h"
36 #include "llvm/MC/MCValue.h"
37 #include "llvm/Support/ErrorHandling.h"
38 #include "llvm/Support/MathExtras.h"
39 #include "llvm/Support/MemoryBuffer.h"
40 #include "llvm/Support/SourceMgr.h"
41 #include "llvm/Support/raw_ostream.h"
42 #include <cctype>
43 #include <deque>
44 #include <string>
45 #include <vector>
46 using namespace llvm;
47 
~MCAsmParserSemaCallback()48 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() {}
49 
50 namespace {
51 /// \brief Helper types for tracking macro definitions.
52 typedef std::vector<AsmToken> MCAsmMacroArgument;
53 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments;
54 
55 struct MCAsmMacroParameter {
56   StringRef Name;
57   MCAsmMacroArgument Value;
58   bool Required;
59   bool Vararg;
60 
MCAsmMacroParameter__anone16e2b900111::MCAsmMacroParameter61   MCAsmMacroParameter() : Required(false), Vararg(false) {}
62 };
63 
64 typedef std::vector<MCAsmMacroParameter> MCAsmMacroParameters;
65 
66 struct MCAsmMacro {
67   StringRef Name;
68   StringRef Body;
69   MCAsmMacroParameters Parameters;
70 
71 public:
MCAsmMacro__anone16e2b900111::MCAsmMacro72   MCAsmMacro(StringRef N, StringRef B, MCAsmMacroParameters P)
73       : Name(N), Body(B), Parameters(std::move(P)) {}
74 };
75 
76 /// \brief Helper class for storing information about an active macro
77 /// instantiation.
78 struct MacroInstantiation {
79   /// The location of the instantiation.
80   SMLoc InstantiationLoc;
81 
82   /// The buffer where parsing should resume upon instantiation completion.
83   int ExitBuffer;
84 
85   /// The location where parsing should resume upon instantiation completion.
86   SMLoc ExitLoc;
87 
88   /// The depth of TheCondStack at the start of the instantiation.
89   size_t CondStackDepth;
90 
91 public:
92   MacroInstantiation(SMLoc IL, int EB, SMLoc EL, size_t CondStackDepth);
93 };
94 
95 struct ParseStatementInfo {
96   /// \brief The parsed operands from the last parsed statement.
97   SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands;
98 
99   /// \brief The opcode from the last parsed instruction.
100   unsigned Opcode;
101 
102   /// \brief Was there an error parsing the inline assembly?
103   bool ParseError;
104 
105   SmallVectorImpl<AsmRewrite> *AsmRewrites;
106 
ParseStatementInfo__anone16e2b900111::ParseStatementInfo107   ParseStatementInfo() : Opcode(~0U), ParseError(false), AsmRewrites(nullptr) {}
ParseStatementInfo__anone16e2b900111::ParseStatementInfo108   ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
109     : Opcode(~0), ParseError(false), AsmRewrites(rewrites) {}
110 };
111 
112 /// \brief The concrete assembly parser instance.
113 class AsmParser : public MCAsmParser {
114   AsmParser(const AsmParser &) = delete;
115   void operator=(const AsmParser &) = delete;
116 private:
117   AsmLexer Lexer;
118   MCContext &Ctx;
119   MCStreamer &Out;
120   const MCAsmInfo &MAI;
121   SourceMgr &SrcMgr;
122   SourceMgr::DiagHandlerTy SavedDiagHandler;
123   void *SavedDiagContext;
124   std::unique_ptr<MCAsmParserExtension> PlatformParser;
125 
126   /// This is the current buffer index we're lexing from as managed by the
127   /// SourceMgr object.
128   unsigned CurBuffer;
129 
130   AsmCond TheCondState;
131   std::vector<AsmCond> TheCondStack;
132 
133   /// \brief maps directive names to handler methods in parser
134   /// extensions. Extensions register themselves in this map by calling
135   /// addDirectiveHandler.
136   StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap;
137 
138   /// \brief Map of currently defined macros.
139   StringMap<MCAsmMacro> MacroMap;
140 
141   /// \brief Stack of active macro instantiations.
142   std::vector<MacroInstantiation*> ActiveMacros;
143 
144   /// \brief List of bodies of anonymous macros.
145   std::deque<MCAsmMacro> MacroLikeBodies;
146 
147   /// Boolean tracking whether macro substitution is enabled.
148   unsigned MacrosEnabledFlag : 1;
149 
150   /// \brief Keeps track of how many .macro's have been instantiated.
151   unsigned NumOfMacroInstantiations;
152 
153   /// Flag tracking whether any errors have been encountered.
154   unsigned HadError : 1;
155 
156   /// The values from the last parsed cpp hash file line comment if any.
157   struct CppHashInfoTy {
158     StringRef Filename;
159     int64_t LineNumber = 0;
160     SMLoc Loc;
161     unsigned Buf = 0;
162   };
163   CppHashInfoTy CppHashInfo;
164 
165   /// \brief List of forward directional labels for diagnosis at the end.
166   SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels;
167 
168   /// When generating dwarf for assembly source files we need to calculate the
169   /// logical line number based on the last parsed cpp hash file line comment
170   /// and current line. Since this is slow and messes up the SourceMgr's
171   /// cache we save the last info we queried with SrcMgr.FindLineNumber().
172   SMLoc LastQueryIDLoc;
173   unsigned LastQueryBuffer;
174   unsigned LastQueryLine;
175 
176   /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
177   unsigned AssemblerDialect;
178 
179   /// \brief is Darwin compatibility enabled?
180   bool IsDarwin;
181 
182   /// \brief Are we parsing ms-style inline assembly?
183   bool ParsingInlineAsm;
184 
185 public:
186   AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
187             const MCAsmInfo &MAI);
188   ~AsmParser() override;
189 
190   bool Run(bool NoInitialTextSection, bool NoFinalize = false) override;
191 
addDirectiveHandler(StringRef Directive,ExtensionDirectiveHandler Handler)192   void addDirectiveHandler(StringRef Directive,
193                            ExtensionDirectiveHandler Handler) override {
194     ExtensionDirectiveMap[Directive] = Handler;
195   }
196 
addAliasForDirective(StringRef Directive,StringRef Alias)197   void addAliasForDirective(StringRef Directive, StringRef Alias) override {
198     DirectiveKindMap[Directive] = DirectiveKindMap[Alias];
199   }
200 
201 public:
202   /// @name MCAsmParser Interface
203   /// {
204 
getSourceManager()205   SourceMgr &getSourceManager() override { return SrcMgr; }
getLexer()206   MCAsmLexer &getLexer() override { return Lexer; }
getContext()207   MCContext &getContext() override { return Ctx; }
getStreamer()208   MCStreamer &getStreamer() override { return Out; }
getAssemblerDialect()209   unsigned getAssemblerDialect() override {
210     if (AssemblerDialect == ~0U)
211       return MAI.getAssemblerDialect();
212     else
213       return AssemblerDialect;
214   }
setAssemblerDialect(unsigned i)215   void setAssemblerDialect(unsigned i) override {
216     AssemblerDialect = i;
217   }
218 
219   void Note(SMLoc L, const Twine &Msg,
220             ArrayRef<SMRange> Ranges = None) override;
221   bool Warning(SMLoc L, const Twine &Msg,
222                ArrayRef<SMRange> Ranges = None) override;
223   bool Error(SMLoc L, const Twine &Msg,
224              ArrayRef<SMRange> Ranges = None) override;
225 
226   const AsmToken &Lex() override;
227 
setParsingInlineAsm(bool V)228   void setParsingInlineAsm(bool V) override { ParsingInlineAsm = V; }
isParsingInlineAsm()229   bool isParsingInlineAsm() override { return ParsingInlineAsm; }
230 
231   bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString,
232                         unsigned &NumOutputs, unsigned &NumInputs,
233                         SmallVectorImpl<std::pair<void *,bool> > &OpDecls,
234                         SmallVectorImpl<std::string> &Constraints,
235                         SmallVectorImpl<std::string> &Clobbers,
236                         const MCInstrInfo *MII, const MCInstPrinter *IP,
237                         MCAsmParserSemaCallback &SI) override;
238 
239   bool parseExpression(const MCExpr *&Res);
240   bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
241   bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override;
242   bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
243   bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
244                              SMLoc &EndLoc) override;
245   bool parseAbsoluteExpression(int64_t &Res) override;
246 
247   /// \brief Parse an identifier or string (as a quoted identifier)
248   /// and set \p Res to the identifier contents.
249   bool parseIdentifier(StringRef &Res) override;
250   void eatToEndOfStatement() override;
251 
252   void checkForValidSection() override;
253 
getTokenLoc(SMLoc & Loc)254   bool getTokenLoc(SMLoc &Loc) {
255     Loc = getTok().getLoc();
256     return false;
257   }
258 
parseEOL(const Twine & ErrMsg)259   bool parseEOL(const Twine &ErrMsg) {
260     if (getTok().getKind() == AsmToken::Hash) {
261       StringRef CommentStr = parseStringToEndOfStatement();
262       Lexer.Lex();
263       Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr));
264     }
265     if (getTok().getKind() != AsmToken::EndOfStatement)
266       return TokError(ErrMsg);
267     Lex();
268     return false;
269   }
270 
271   /// parseToken - If current token has the specified kind, eat it and
272   /// return success.  Otherwise, emit the specified error and return failure.
parseToken(AsmToken::TokenKind T,const Twine & ErrMsg)273   bool parseToken(AsmToken::TokenKind T, const Twine &ErrMsg) {
274     if (T == AsmToken::EndOfStatement)
275       return parseEOL(ErrMsg);
276     if (getTok().getKind() != T)
277       return TokError(ErrMsg);
278     Lex();
279     return false;
280   }
281 
parseIntToken(int64_t & V,const Twine & ErrMsg)282   bool parseIntToken(int64_t &V, const Twine &ErrMsg) {
283     if (getTok().getKind() != AsmToken::Integer)
284       return TokError(ErrMsg);
285     V = getTok().getIntVal();
286     Lex();
287     return false;
288   }
289 
290   /// }
291 
292 private:
293 
294   bool parseStatement(ParseStatementInfo &Info,
295                       MCAsmParserSemaCallback *SI);
296   bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites);
297   bool parseCppHashLineFilenameComment(SMLoc L);
298 
299   void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body,
300                         ArrayRef<MCAsmMacroParameter> Parameters);
301   bool expandMacro(raw_svector_ostream &OS, StringRef Body,
302                    ArrayRef<MCAsmMacroParameter> Parameters,
303                    ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable,
304                    SMLoc L);
305 
306   /// \brief Are macros enabled in the parser?
areMacrosEnabled()307   bool areMacrosEnabled() {return MacrosEnabledFlag;}
308 
309   /// \brief Control a flag in the parser that enables or disables macros.
setMacrosEnabled(bool Flag)310   void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;}
311 
312   /// \brief Lookup a previously defined macro.
313   /// \param Name Macro name.
314   /// \returns Pointer to macro. NULL if no such macro was defined.
315   const MCAsmMacro* lookupMacro(StringRef Name);
316 
317   /// \brief Define a new macro with the given name and information.
318   void defineMacro(StringRef Name, MCAsmMacro Macro);
319 
320   /// \brief Undefine a macro. If no such macro was defined, it's a no-op.
321   void undefineMacro(StringRef Name);
322 
323   /// \brief Are we inside a macro instantiation?
isInsideMacroInstantiation()324   bool isInsideMacroInstantiation() {return !ActiveMacros.empty();}
325 
326   /// \brief Handle entry to macro instantiation.
327   ///
328   /// \param M The macro.
329   /// \param NameLoc Instantiation location.
330   bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc);
331 
332   /// \brief Handle exit from macro instantiation.
333   void handleMacroExit();
334 
335   /// \brief Extract AsmTokens for a macro argument.
336   bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg);
337 
338   /// \brief Parse all macro arguments for a given macro.
339   bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A);
340 
341   void printMacroInstantiations();
printMessage(SMLoc Loc,SourceMgr::DiagKind Kind,const Twine & Msg,ArrayRef<SMRange> Ranges=None) const342   void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
343                     ArrayRef<SMRange> Ranges = None) const {
344     SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
345   }
346   static void DiagHandler(const SMDiagnostic &Diag, void *Context);
347 
348   /// \brief Enter the specified file. This returns true on failure.
349   bool enterIncludeFile(const std::string &Filename);
350 
351   /// \brief Process the specified file for the .incbin directive.
352   /// This returns true on failure.
353   bool processIncbinFile(const std::string &Filename);
354 
355   /// \brief Reset the current lexer position to that given by \p Loc. The
356   /// current token is not set; clients should ensure Lex() is called
357   /// subsequently.
358   ///
359   /// \param InBuffer If not 0, should be the known buffer id that contains the
360   /// location.
361   void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0);
362 
363   /// \brief Parse up to the end of statement and a return the contents from the
364   /// current token until the end of the statement; the current token on exit
365   /// will be either the EndOfStatement or EOF.
366   StringRef parseStringToEndOfStatement() override;
367 
368   /// \brief Parse until the end of a statement or a comma is encountered,
369   /// return the contents from the current token up to the end or comma.
370   StringRef parseStringToComma();
371 
372   bool parseAssignment(StringRef Name, bool allow_redef,
373                        bool NoDeadStrip = false);
374 
375   unsigned getBinOpPrecedence(AsmToken::TokenKind K,
376                               MCBinaryExpr::Opcode &Kind);
377 
378   bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
379   bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
380   bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
381 
382   bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
383 
384   // Generic (target and platform independent) directive parsing.
385   enum DirectiveKind {
386     DK_NO_DIRECTIVE, // Placeholder
387     DK_SET, DK_EQU, DK_EQUIV, DK_ASCII, DK_ASCIZ, DK_STRING, DK_BYTE, DK_SHORT,
388     DK_RELOC,
389     DK_VALUE, DK_2BYTE, DK_LONG, DK_INT, DK_4BYTE, DK_QUAD, DK_8BYTE, DK_OCTA,
390     DK_SINGLE, DK_FLOAT, DK_DOUBLE, DK_ALIGN, DK_ALIGN32, DK_BALIGN, DK_BALIGNW,
391     DK_BALIGNL, DK_P2ALIGN, DK_P2ALIGNW, DK_P2ALIGNL, DK_ORG, DK_FILL, DK_ENDR,
392     DK_BUNDLE_ALIGN_MODE, DK_BUNDLE_LOCK, DK_BUNDLE_UNLOCK,
393     DK_ZERO, DK_EXTERN, DK_GLOBL, DK_GLOBAL,
394     DK_LAZY_REFERENCE, DK_NO_DEAD_STRIP, DK_SYMBOL_RESOLVER,
395     DK_PRIVATE_EXTERN, DK_REFERENCE, DK_WEAK_DEFINITION, DK_WEAK_REFERENCE,
396     DK_WEAK_DEF_CAN_BE_HIDDEN, DK_COMM, DK_COMMON, DK_LCOMM, DK_ABORT,
397     DK_INCLUDE, DK_INCBIN, DK_CODE16, DK_CODE16GCC, DK_REPT, DK_IRP, DK_IRPC,
398     DK_IF, DK_IFEQ, DK_IFGE, DK_IFGT, DK_IFLE, DK_IFLT, DK_IFNE, DK_IFB,
399     DK_IFNB, DK_IFC, DK_IFEQS, DK_IFNC, DK_IFNES, DK_IFDEF, DK_IFNDEF,
400     DK_IFNOTDEF, DK_ELSEIF, DK_ELSE, DK_ENDIF,
401     DK_SPACE, DK_SKIP, DK_FILE, DK_LINE, DK_LOC, DK_STABS,
402     DK_CV_FILE, DK_CV_LOC, DK_CV_LINETABLE, DK_CV_INLINE_LINETABLE,
403     DK_CV_DEF_RANGE, DK_CV_STRINGTABLE, DK_CV_FILECHECKSUMS,
404     DK_CFI_SECTIONS, DK_CFI_STARTPROC, DK_CFI_ENDPROC, DK_CFI_DEF_CFA,
405     DK_CFI_DEF_CFA_OFFSET, DK_CFI_ADJUST_CFA_OFFSET, DK_CFI_DEF_CFA_REGISTER,
406     DK_CFI_OFFSET, DK_CFI_REL_OFFSET, DK_CFI_PERSONALITY, DK_CFI_LSDA,
407     DK_CFI_REMEMBER_STATE, DK_CFI_RESTORE_STATE, DK_CFI_SAME_VALUE,
408     DK_CFI_RESTORE, DK_CFI_ESCAPE, DK_CFI_SIGNAL_FRAME, DK_CFI_UNDEFINED,
409     DK_CFI_REGISTER, DK_CFI_WINDOW_SAVE,
410     DK_MACROS_ON, DK_MACROS_OFF,
411     DK_MACRO, DK_EXITM, DK_ENDM, DK_ENDMACRO, DK_PURGEM,
412     DK_SLEB128, DK_ULEB128,
413     DK_ERR, DK_ERROR, DK_WARNING,
414     DK_END
415   };
416 
417   /// \brief Maps directive name --> DirectiveKind enum, for
418   /// directives parsed by this class.
419   StringMap<DirectiveKind> DirectiveKindMap;
420 
421   // ".ascii", ".asciz", ".string"
422   bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
423   bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc"
424   bool parseDirectiveValue(unsigned Size); // ".byte", ".long", ...
425   bool parseDirectiveOctaValue(); // ".octa"
426   bool parseDirectiveRealValue(const fltSemantics &); // ".single", ...
427   bool parseDirectiveFill(); // ".fill"
428   bool parseDirectiveZero(); // ".zero"
429   // ".set", ".equ", ".equiv"
430   bool parseDirectiveSet(StringRef IDVal, bool allow_redef);
431   bool parseDirectiveOrg(); // ".org"
432   // ".align{,32}", ".p2align{,w,l}"
433   bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize);
434 
435   // ".file", ".line", ".loc", ".stabs"
436   bool parseDirectiveFile(SMLoc DirectiveLoc);
437   bool parseDirectiveLine();
438   bool parseDirectiveLoc();
439   bool parseDirectiveStabs();
440 
441   // ".cv_file", ".cv_loc", ".cv_linetable", "cv_inline_linetable",
442   // ".cv_def_range"
443   bool parseDirectiveCVFile();
444   bool parseDirectiveCVLoc();
445   bool parseDirectiveCVLinetable();
446   bool parseDirectiveCVInlineLinetable();
447   bool parseDirectiveCVDefRange();
448   bool parseDirectiveCVStringTable();
449   bool parseDirectiveCVFileChecksums();
450 
451   // .cfi directives
452   bool parseDirectiveCFIRegister(SMLoc DirectiveLoc);
453   bool parseDirectiveCFIWindowSave();
454   bool parseDirectiveCFISections();
455   bool parseDirectiveCFIStartProc();
456   bool parseDirectiveCFIEndProc();
457   bool parseDirectiveCFIDefCfaOffset();
458   bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc);
459   bool parseDirectiveCFIAdjustCfaOffset();
460   bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc);
461   bool parseDirectiveCFIOffset(SMLoc DirectiveLoc);
462   bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc);
463   bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality);
464   bool parseDirectiveCFIRememberState();
465   bool parseDirectiveCFIRestoreState();
466   bool parseDirectiveCFISameValue(SMLoc DirectiveLoc);
467   bool parseDirectiveCFIRestore(SMLoc DirectiveLoc);
468   bool parseDirectiveCFIEscape();
469   bool parseDirectiveCFISignalFrame();
470   bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc);
471 
472   // macro directives
473   bool parseDirectivePurgeMacro(SMLoc DirectiveLoc);
474   bool parseDirectiveExitMacro(StringRef Directive);
475   bool parseDirectiveEndMacro(StringRef Directive);
476   bool parseDirectiveMacro(SMLoc DirectiveLoc);
477   bool parseDirectiveMacrosOnOff(StringRef Directive);
478 
479   // ".bundle_align_mode"
480   bool parseDirectiveBundleAlignMode();
481   // ".bundle_lock"
482   bool parseDirectiveBundleLock();
483   // ".bundle_unlock"
484   bool parseDirectiveBundleUnlock();
485 
486   // ".space", ".skip"
487   bool parseDirectiveSpace(StringRef IDVal);
488 
489   // .sleb128 (Signed=true) and .uleb128 (Signed=false)
490   bool parseDirectiveLEB128(bool Signed);
491 
492   /// \brief Parse a directive like ".globl" which
493   /// accepts a single symbol (which should be a label or an external).
494   bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr);
495 
496   bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
497 
498   bool parseDirectiveAbort(); // ".abort"
499   bool parseDirectiveInclude(); // ".include"
500   bool parseDirectiveIncbin(); // ".incbin"
501 
502   // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne"
503   bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
504   // ".ifb" or ".ifnb", depending on ExpectBlank.
505   bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
506   // ".ifc" or ".ifnc", depending on ExpectEqual.
507   bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
508   // ".ifeqs" or ".ifnes", depending on ExpectEqual.
509   bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual);
510   // ".ifdef" or ".ifndef", depending on expect_defined
511   bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
512   bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
513   bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else"
514   bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
515   bool parseEscapedString(std::string &Data) override;
516 
517   const MCExpr *applyModifierToExpr(const MCExpr *E,
518                                     MCSymbolRefExpr::VariantKind Variant);
519 
520   // Macro-like directives
521   MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
522   void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
523                                 raw_svector_ostream &OS);
524   bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive);
525   bool parseDirectiveIrp(SMLoc DirectiveLoc);  // ".irp"
526   bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
527   bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
528 
529   // "_emit" or "__emit"
530   bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info,
531                             size_t Len);
532 
533   // "align"
534   bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info);
535 
536   // "end"
537   bool parseDirectiveEnd(SMLoc DirectiveLoc);
538 
539   // ".err" or ".error"
540   bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage);
541 
542   // ".warning"
543   bool parseDirectiveWarning(SMLoc DirectiveLoc);
544 
545   void initializeDirectiveKindMap();
546 };
547 }
548 
549 namespace llvm {
550 
551 extern MCAsmParserExtension *createDarwinAsmParser();
552 extern MCAsmParserExtension *createELFAsmParser();
553 extern MCAsmParserExtension *createCOFFAsmParser();
554 
555 }
556 
557 enum { DEFAULT_ADDRSPACE = 0 };
558 
AsmParser(SourceMgr & SM,MCContext & Ctx,MCStreamer & Out,const MCAsmInfo & MAI)559 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
560                      const MCAsmInfo &MAI)
561     : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM),
562       PlatformParser(nullptr), CurBuffer(SM.getMainFileID()),
563       MacrosEnabledFlag(true), HadError(false), CppHashInfo(),
564       AssemblerDialect(~0U), IsDarwin(false), ParsingInlineAsm(false) {
565   // Save the old handler.
566   SavedDiagHandler = SrcMgr.getDiagHandler();
567   SavedDiagContext = SrcMgr.getDiagContext();
568   // Set our own handler which calls the saved handler.
569   SrcMgr.setDiagHandler(DiagHandler, this);
570   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
571 
572   // Initialize the platform / file format parser.
573   switch (Ctx.getObjectFileInfo()->getObjectFileType()) {
574   case MCObjectFileInfo::IsCOFF:
575     PlatformParser.reset(createCOFFAsmParser());
576     break;
577   case MCObjectFileInfo::IsMachO:
578     PlatformParser.reset(createDarwinAsmParser());
579     IsDarwin = true;
580     break;
581   case MCObjectFileInfo::IsELF:
582     PlatformParser.reset(createELFAsmParser());
583     break;
584   }
585 
586   PlatformParser->Initialize(*this);
587   initializeDirectiveKindMap();
588 
589   NumOfMacroInstantiations = 0;
590 }
591 
~AsmParser()592 AsmParser::~AsmParser() {
593   assert((HadError || ActiveMacros.empty()) &&
594          "Unexpected active macro instantiation!");
595 }
596 
printMacroInstantiations()597 void AsmParser::printMacroInstantiations() {
598   // Print the active macro instantiation stack.
599   for (std::vector<MacroInstantiation *>::const_reverse_iterator
600            it = ActiveMacros.rbegin(),
601            ie = ActiveMacros.rend();
602        it != ie; ++it)
603     printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
604                  "while in macro instantiation");
605 }
606 
Note(SMLoc L,const Twine & Msg,ArrayRef<SMRange> Ranges)607 void AsmParser::Note(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
608   printMessage(L, SourceMgr::DK_Note, Msg, Ranges);
609   printMacroInstantiations();
610 }
611 
Warning(SMLoc L,const Twine & Msg,ArrayRef<SMRange> Ranges)612 bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
613   if(getTargetParser().getTargetOptions().MCNoWarn)
614     return false;
615   if (getTargetParser().getTargetOptions().MCFatalWarnings)
616     return Error(L, Msg, Ranges);
617   printMessage(L, SourceMgr::DK_Warning, Msg, Ranges);
618   printMacroInstantiations();
619   return false;
620 }
621 
Error(SMLoc L,const Twine & Msg,ArrayRef<SMRange> Ranges)622 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
623   HadError = true;
624   printMessage(L, SourceMgr::DK_Error, Msg, Ranges);
625   printMacroInstantiations();
626   return true;
627 }
628 
enterIncludeFile(const std::string & Filename)629 bool AsmParser::enterIncludeFile(const std::string &Filename) {
630   std::string IncludedFile;
631   unsigned NewBuf =
632       SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
633   if (!NewBuf)
634     return true;
635 
636   CurBuffer = NewBuf;
637   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
638   return false;
639 }
640 
641 /// Process the specified .incbin file by searching for it in the include paths
642 /// then just emitting the byte contents of the file to the streamer. This
643 /// returns true on failure.
processIncbinFile(const std::string & Filename)644 bool AsmParser::processIncbinFile(const std::string &Filename) {
645   std::string IncludedFile;
646   unsigned NewBuf =
647       SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
648   if (!NewBuf)
649     return true;
650 
651   // Pick up the bytes from the file and emit them.
652   getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer());
653   return false;
654 }
655 
jumpToLoc(SMLoc Loc,unsigned InBuffer)656 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) {
657   CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc);
658   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(),
659                   Loc.getPointer());
660 }
661 
Lex()662 const AsmToken &AsmParser::Lex() {
663   if (Lexer.getTok().is(AsmToken::Error))
664     Error(Lexer.getErrLoc(), Lexer.getErr());
665 
666   // if it's a end of statement with a comment in it
667   if (getTok().is(AsmToken::EndOfStatement)) {
668     // if this is a line comment output it.
669     if (getTok().getString().front() != '\n' &&
670         getTok().getString().front() != '\r' && MAI.preserveAsmComments())
671       Out.addExplicitComment(Twine(getTok().getString()));
672   }
673 
674   const AsmToken *tok = &Lexer.Lex();
675 
676   // Parse comments here to be deferred until end of next statement.
677   while (tok->is(AsmToken::Comment)) {
678     if (MAI.preserveAsmComments())
679       Out.addExplicitComment(Twine(tok->getString()));
680     tok = &Lexer.Lex();
681   }
682 
683   if (tok->is(AsmToken::Eof)) {
684     // If this is the end of an included file, pop the parent file off the
685     // include stack.
686     SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
687     if (ParentIncludeLoc != SMLoc()) {
688       jumpToLoc(ParentIncludeLoc);
689       return Lex();
690     }
691   }
692 
693 
694   return *tok;
695 }
696 
Run(bool NoInitialTextSection,bool NoFinalize)697 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
698   // Create the initial section, if requested.
699   if (!NoInitialTextSection)
700     Out.InitSections(false);
701 
702   // Prime the lexer.
703   Lex();
704 
705   HadError = false;
706   AsmCond StartingCondState = TheCondState;
707 
708   // If we are generating dwarf for assembly source files save the initial text
709   // section and generate a .file directive.
710   if (getContext().getGenDwarfForAssembly()) {
711     MCSection *Sec = getStreamer().getCurrentSection().first;
712     if (!Sec->getBeginSymbol()) {
713       MCSymbol *SectionStartSym = getContext().createTempSymbol();
714       getStreamer().EmitLabel(SectionStartSym);
715       Sec->setBeginSymbol(SectionStartSym);
716     }
717     bool InsertResult = getContext().addGenDwarfSection(Sec);
718     assert(InsertResult && ".text section should not have debug info yet");
719     (void)InsertResult;
720     getContext().setGenDwarfFileNumber(getStreamer().EmitDwarfFileDirective(
721         0, StringRef(), getContext().getMainFileName()));
722   }
723 
724   // While we have input, parse each statement.
725   while (Lexer.isNot(AsmToken::Eof)) {
726     ParseStatementInfo Info;
727     if (!parseStatement(Info, nullptr))
728       continue;
729 
730     // If we've failed, but on a Error Token, but did not consume it in
731     // favor of a better message, emit it now.
732     if (Lexer.getTok().is(AsmToken::Error)) {
733       Lex();
734     }
735 
736     // We had an error, validate that one was emitted and recover by skipping to
737     // the next line.
738     assert(HadError && "Parse statement returned an error, but none emitted!");
739     eatToEndOfStatement();
740   }
741 
742   if (TheCondState.TheCond != StartingCondState.TheCond ||
743       TheCondState.Ignore != StartingCondState.Ignore)
744     return TokError("unmatched .ifs or .elses");
745 
746   // Check to see there are no empty DwarfFile slots.
747   const auto &LineTables = getContext().getMCDwarfLineTables();
748   if (!LineTables.empty()) {
749     unsigned Index = 0;
750     for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) {
751       if (File.Name.empty() && Index != 0)
752         TokError("unassigned file number: " + Twine(Index) +
753                  " for .file directives");
754       ++Index;
755     }
756   }
757 
758   // Check to see that all assembler local symbols were actually defined.
759   // Targets that don't do subsections via symbols may not want this, though,
760   // so conservatively exclude them. Only do this if we're finalizing, though,
761   // as otherwise we won't necessarilly have seen everything yet.
762   if (!NoFinalize) {
763     if (MAI.hasSubsectionsViaSymbols()) {
764       for (const auto &TableEntry : getContext().getSymbols()) {
765         MCSymbol *Sym = TableEntry.getValue();
766         // Variable symbols may not be marked as defined, so check those
767         // explicitly. If we know it's a variable, we have a definition for
768         // the purposes of this check.
769         if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
770           // FIXME: We would really like to refer back to where the symbol was
771           // first referenced for a source location. We need to add something
772           // to track that. Currently, we just point to the end of the file.
773           HadError |=
774               Error(getTok().getLoc(), "assembler local symbol '" +
775                                            Sym->getName() + "' not defined");
776       }
777     }
778 
779     // Temporary symbols like the ones for directional jumps don't go in the
780     // symbol table. They also need to be diagnosed in all (final) cases.
781     for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) {
782       if (std::get<2>(LocSym)->isUndefined()) {
783         // Reset the state of any "# line file" directives we've seen to the
784         // context as it was at the diagnostic site.
785         CppHashInfo = std::get<1>(LocSym);
786         HadError |= Error(std::get<0>(LocSym), "directional label undefined");
787       }
788     }
789   }
790 
791   // Finalize the output stream if there are no errors and if the client wants
792   // us to.
793   if (!HadError && !NoFinalize)
794     Out.Finish();
795 
796   return HadError || getContext().hadError();
797 }
798 
checkForValidSection()799 void AsmParser::checkForValidSection() {
800   if (!ParsingInlineAsm && !getStreamer().getCurrentSection().first) {
801     TokError("expected section directive before assembly directive");
802     Out.InitSections(false);
803   }
804 }
805 
806 /// \brief Throw away the rest of the line for testing purposes.
eatToEndOfStatement()807 void AsmParser::eatToEndOfStatement() {
808   while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
809     Lexer.Lex();
810 
811   // Eat EOL.
812   if (Lexer.is(AsmToken::EndOfStatement))
813     Lexer.Lex();
814 }
815 
parseStringToEndOfStatement()816 StringRef AsmParser::parseStringToEndOfStatement() {
817   const char *Start = getTok().getLoc().getPointer();
818 
819   while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
820     Lexer.Lex();
821 
822   const char *End = getTok().getLoc().getPointer();
823   return StringRef(Start, End - Start);
824 }
825 
parseStringToComma()826 StringRef AsmParser::parseStringToComma() {
827   const char *Start = getTok().getLoc().getPointer();
828 
829   while (Lexer.isNot(AsmToken::EndOfStatement) &&
830          Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof))
831     Lexer.Lex();
832 
833   const char *End = getTok().getLoc().getPointer();
834   return StringRef(Start, End - Start);
835 }
836 
837 /// \brief Parse a paren expression and return it.
838 /// NOTE: This assumes the leading '(' has already been consumed.
839 ///
840 /// parenexpr ::= expr)
841 ///
parseParenExpr(const MCExpr * & Res,SMLoc & EndLoc)842 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
843   if (parseExpression(Res))
844     return true;
845   if (Lexer.isNot(AsmToken::RParen))
846     return TokError("expected ')' in parentheses expression");
847   EndLoc = Lexer.getTok().getEndLoc();
848   Lex();
849   return false;
850 }
851 
852 /// \brief Parse a bracket expression and return it.
853 /// NOTE: This assumes the leading '[' has already been consumed.
854 ///
855 /// bracketexpr ::= expr]
856 ///
parseBracketExpr(const MCExpr * & Res,SMLoc & EndLoc)857 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
858   if (parseExpression(Res))
859     return true;
860   if (Lexer.isNot(AsmToken::RBrac))
861     return TokError("expected ']' in brackets expression");
862   EndLoc = Lexer.getTok().getEndLoc();
863   Lex();
864   return false;
865 }
866 
867 /// \brief Parse a primary expression and return it.
868 ///  primaryexpr ::= (parenexpr
869 ///  primaryexpr ::= symbol
870 ///  primaryexpr ::= number
871 ///  primaryexpr ::= '.'
872 ///  primaryexpr ::= ~,+,- primaryexpr
parsePrimaryExpr(const MCExpr * & Res,SMLoc & EndLoc)873 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
874   SMLoc FirstTokenLoc = getLexer().getLoc();
875   AsmToken::TokenKind FirstTokenKind = Lexer.getKind();
876   switch (FirstTokenKind) {
877   default:
878     return TokError("unknown token in expression");
879   // If we have an error assume that we've already handled it.
880   case AsmToken::Error:
881     return true;
882   case AsmToken::Exclaim:
883     Lex(); // Eat the operator.
884     if (parsePrimaryExpr(Res, EndLoc))
885       return true;
886     Res = MCUnaryExpr::createLNot(Res, getContext());
887     return false;
888   case AsmToken::Dollar:
889   case AsmToken::At:
890   case AsmToken::String:
891   case AsmToken::Identifier: {
892     StringRef Identifier;
893     if (parseIdentifier(Identifier)) {
894       if (FirstTokenKind == AsmToken::Dollar) {
895         if (Lexer.getMAI().getDollarIsPC()) {
896           // This is a '$' reference, which references the current PC.  Emit a
897           // temporary label to the streamer and refer to it.
898           MCSymbol *Sym = Ctx.createTempSymbol();
899           Out.EmitLabel(Sym);
900           Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None,
901                                         getContext());
902           EndLoc = FirstTokenLoc;
903           return false;
904         }
905         return Error(FirstTokenLoc, "invalid token in expression");
906       }
907     }
908     // Parse symbol variant
909     std::pair<StringRef, StringRef> Split;
910     if (!MAI.useParensForSymbolVariant()) {
911       if (FirstTokenKind == AsmToken::String) {
912         if (Lexer.is(AsmToken::At)) {
913           Lex(); // eat @
914           SMLoc AtLoc = getLexer().getLoc();
915           StringRef VName;
916           if (parseIdentifier(VName))
917             return Error(AtLoc, "expected symbol variant after '@'");
918 
919           Split = std::make_pair(Identifier, VName);
920         }
921       } else {
922         Split = Identifier.split('@');
923       }
924     } else if (Lexer.is(AsmToken::LParen)) {
925       Lex(); // eat '('.
926       StringRef VName;
927       parseIdentifier(VName);
928       if (Lexer.isNot(AsmToken::RParen)) {
929           return Error(Lexer.getTok().getLoc(),
930                        "unexpected token in variant, expected ')'");
931       }
932       Lex(); // eat ')'.
933       Split = std::make_pair(Identifier, VName);
934     }
935 
936     EndLoc = SMLoc::getFromPointer(Identifier.end());
937 
938     // This is a symbol reference.
939     StringRef SymbolName = Identifier;
940     MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
941 
942     // Lookup the symbol variant if used.
943     if (Split.second.size()) {
944       Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
945       if (Variant != MCSymbolRefExpr::VK_Invalid) {
946         SymbolName = Split.first;
947       } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) {
948         Variant = MCSymbolRefExpr::VK_None;
949       } else {
950         return Error(SMLoc::getFromPointer(Split.second.begin()),
951                      "invalid variant '" + Split.second + "'");
952       }
953     }
954 
955     MCSymbol *Sym = getContext().getOrCreateSymbol(SymbolName);
956 
957     // If this is an absolute variable reference, substitute it now to preserve
958     // semantics in the face of reassignment.
959     if (Sym->isVariable() &&
960         isa<MCConstantExpr>(Sym->getVariableValue(/*SetUsed*/ false))) {
961       if (Variant)
962         return Error(EndLoc, "unexpected modifier on variable reference");
963 
964       Res = Sym->getVariableValue(/*SetUsed*/ false);
965       return false;
966     }
967 
968     // Otherwise create a symbol ref.
969     Res = MCSymbolRefExpr::create(Sym, Variant, getContext());
970     return false;
971   }
972   case AsmToken::BigNum:
973     return TokError("literal value out of range for directive");
974   case AsmToken::Integer: {
975     SMLoc Loc = getTok().getLoc();
976     int64_t IntVal = getTok().getIntVal();
977     Res = MCConstantExpr::create(IntVal, getContext());
978     EndLoc = Lexer.getTok().getEndLoc();
979     Lex(); // Eat token.
980     // Look for 'b' or 'f' following an Integer as a directional label
981     if (Lexer.getKind() == AsmToken::Identifier) {
982       StringRef IDVal = getTok().getString();
983       // Lookup the symbol variant if used.
984       std::pair<StringRef, StringRef> Split = IDVal.split('@');
985       MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
986       if (Split.first.size() != IDVal.size()) {
987         Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
988         if (Variant == MCSymbolRefExpr::VK_Invalid)
989           return TokError("invalid variant '" + Split.second + "'");
990         IDVal = Split.first;
991       }
992       if (IDVal == "f" || IDVal == "b") {
993         MCSymbol *Sym =
994             Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b");
995         Res = MCSymbolRefExpr::create(Sym, Variant, getContext());
996         if (IDVal == "b" && Sym->isUndefined())
997           return Error(Loc, "directional label undefined");
998         DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym));
999         EndLoc = Lexer.getTok().getEndLoc();
1000         Lex(); // Eat identifier.
1001       }
1002     }
1003     return false;
1004   }
1005   case AsmToken::Real: {
1006     APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
1007     uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
1008     Res = MCConstantExpr::create(IntVal, getContext());
1009     EndLoc = Lexer.getTok().getEndLoc();
1010     Lex(); // Eat token.
1011     return false;
1012   }
1013   case AsmToken::Dot: {
1014     // This is a '.' reference, which references the current PC.  Emit a
1015     // temporary label to the streamer and refer to it.
1016     MCSymbol *Sym = Ctx.createTempSymbol();
1017     Out.EmitLabel(Sym);
1018     Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext());
1019     EndLoc = Lexer.getTok().getEndLoc();
1020     Lex(); // Eat identifier.
1021     return false;
1022   }
1023   case AsmToken::LParen:
1024     Lex(); // Eat the '('.
1025     return parseParenExpr(Res, EndLoc);
1026   case AsmToken::LBrac:
1027     if (!PlatformParser->HasBracketExpressions())
1028       return TokError("brackets expression not supported on this target");
1029     Lex(); // Eat the '['.
1030     return parseBracketExpr(Res, EndLoc);
1031   case AsmToken::Minus:
1032     Lex(); // Eat the operator.
1033     if (parsePrimaryExpr(Res, EndLoc))
1034       return true;
1035     Res = MCUnaryExpr::createMinus(Res, getContext());
1036     return false;
1037   case AsmToken::Plus:
1038     Lex(); // Eat the operator.
1039     if (parsePrimaryExpr(Res, EndLoc))
1040       return true;
1041     Res = MCUnaryExpr::createPlus(Res, getContext());
1042     return false;
1043   case AsmToken::Tilde:
1044     Lex(); // Eat the operator.
1045     if (parsePrimaryExpr(Res, EndLoc))
1046       return true;
1047     Res = MCUnaryExpr::createNot(Res, getContext());
1048     return false;
1049   }
1050 }
1051 
parseExpression(const MCExpr * & Res)1052 bool AsmParser::parseExpression(const MCExpr *&Res) {
1053   SMLoc EndLoc;
1054   return parseExpression(Res, EndLoc);
1055 }
1056 
1057 const MCExpr *
applyModifierToExpr(const MCExpr * E,MCSymbolRefExpr::VariantKind Variant)1058 AsmParser::applyModifierToExpr(const MCExpr *E,
1059                                MCSymbolRefExpr::VariantKind Variant) {
1060   // Ask the target implementation about this expression first.
1061   const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx);
1062   if (NewE)
1063     return NewE;
1064   // Recurse over the given expression, rebuilding it to apply the given variant
1065   // if there is exactly one symbol.
1066   switch (E->getKind()) {
1067   case MCExpr::Target:
1068   case MCExpr::Constant:
1069     return nullptr;
1070 
1071   case MCExpr::SymbolRef: {
1072     const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
1073 
1074     if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
1075       TokError("invalid variant on expression '" + getTok().getIdentifier() +
1076                "' (already modified)");
1077       return E;
1078     }
1079 
1080     return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext());
1081   }
1082 
1083   case MCExpr::Unary: {
1084     const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
1085     const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant);
1086     if (!Sub)
1087       return nullptr;
1088     return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext());
1089   }
1090 
1091   case MCExpr::Binary: {
1092     const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
1093     const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant);
1094     const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant);
1095 
1096     if (!LHS && !RHS)
1097       return nullptr;
1098 
1099     if (!LHS)
1100       LHS = BE->getLHS();
1101     if (!RHS)
1102       RHS = BE->getRHS();
1103 
1104     return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext());
1105   }
1106   }
1107 
1108   llvm_unreachable("Invalid expression kind!");
1109 }
1110 
1111 /// \brief Parse an expression and return it.
1112 ///
1113 ///  expr ::= expr &&,|| expr               -> lowest.
1114 ///  expr ::= expr |,^,&,! expr
1115 ///  expr ::= expr ==,!=,<>,<,<=,>,>= expr
1116 ///  expr ::= expr <<,>> expr
1117 ///  expr ::= expr +,- expr
1118 ///  expr ::= expr *,/,% expr               -> highest.
1119 ///  expr ::= primaryexpr
1120 ///
parseExpression(const MCExpr * & Res,SMLoc & EndLoc)1121 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1122   // Parse the expression.
1123   Res = nullptr;
1124   if (parsePrimaryExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc))
1125     return true;
1126 
1127   // As a special case, we support 'a op b @ modifier' by rewriting the
1128   // expression to include the modifier. This is inefficient, but in general we
1129   // expect users to use 'a@modifier op b'.
1130   if (Lexer.getKind() == AsmToken::At) {
1131     Lex();
1132 
1133     if (Lexer.isNot(AsmToken::Identifier))
1134       return TokError("unexpected symbol modifier following '@'");
1135 
1136     MCSymbolRefExpr::VariantKind Variant =
1137         MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
1138     if (Variant == MCSymbolRefExpr::VK_Invalid)
1139       return TokError("invalid variant '" + getTok().getIdentifier() + "'");
1140 
1141     const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant);
1142     if (!ModifiedRes) {
1143       return TokError("invalid modifier '" + getTok().getIdentifier() +
1144                       "' (no symbols present)");
1145     }
1146 
1147     Res = ModifiedRes;
1148     Lex();
1149   }
1150 
1151   // Try to constant fold it up front, if possible.
1152   int64_t Value;
1153   if (Res->evaluateAsAbsolute(Value))
1154     Res = MCConstantExpr::create(Value, getContext());
1155 
1156   return false;
1157 }
1158 
parseParenExpression(const MCExpr * & Res,SMLoc & EndLoc)1159 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1160   Res = nullptr;
1161   return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc);
1162 }
1163 
parseParenExprOfDepth(unsigned ParenDepth,const MCExpr * & Res,SMLoc & EndLoc)1164 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
1165                                       SMLoc &EndLoc) {
1166   if (parseParenExpr(Res, EndLoc))
1167     return true;
1168 
1169   for (; ParenDepth > 0; --ParenDepth) {
1170     if (parseBinOpRHS(1, Res, EndLoc))
1171       return true;
1172 
1173     // We don't Lex() the last RParen.
1174     // This is the same behavior as parseParenExpression().
1175     if (ParenDepth - 1 > 0) {
1176       if (Lexer.isNot(AsmToken::RParen))
1177         return TokError("expected ')' in parentheses expression");
1178       EndLoc = Lexer.getTok().getEndLoc();
1179       Lex();
1180     }
1181   }
1182   return false;
1183 }
1184 
parseAbsoluteExpression(int64_t & Res)1185 bool AsmParser::parseAbsoluteExpression(int64_t &Res) {
1186   const MCExpr *Expr;
1187 
1188   SMLoc StartLoc = Lexer.getLoc();
1189   if (parseExpression(Expr))
1190     return true;
1191 
1192   if (!Expr->evaluateAsAbsolute(Res))
1193     return Error(StartLoc, "expected absolute expression");
1194 
1195   return false;
1196 }
1197 
getDarwinBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1198 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K,
1199                                          MCBinaryExpr::Opcode &Kind,
1200                                          bool ShouldUseLogicalShr) {
1201   switch (K) {
1202   default:
1203     return 0; // not a binop.
1204 
1205   // Lowest Precedence: &&, ||
1206   case AsmToken::AmpAmp:
1207     Kind = MCBinaryExpr::LAnd;
1208     return 1;
1209   case AsmToken::PipePipe:
1210     Kind = MCBinaryExpr::LOr;
1211     return 1;
1212 
1213   // Low Precedence: |, &, ^
1214   //
1215   // FIXME: gas seems to support '!' as an infix operator?
1216   case AsmToken::Pipe:
1217     Kind = MCBinaryExpr::Or;
1218     return 2;
1219   case AsmToken::Caret:
1220     Kind = MCBinaryExpr::Xor;
1221     return 2;
1222   case AsmToken::Amp:
1223     Kind = MCBinaryExpr::And;
1224     return 2;
1225 
1226   // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
1227   case AsmToken::EqualEqual:
1228     Kind = MCBinaryExpr::EQ;
1229     return 3;
1230   case AsmToken::ExclaimEqual:
1231   case AsmToken::LessGreater:
1232     Kind = MCBinaryExpr::NE;
1233     return 3;
1234   case AsmToken::Less:
1235     Kind = MCBinaryExpr::LT;
1236     return 3;
1237   case AsmToken::LessEqual:
1238     Kind = MCBinaryExpr::LTE;
1239     return 3;
1240   case AsmToken::Greater:
1241     Kind = MCBinaryExpr::GT;
1242     return 3;
1243   case AsmToken::GreaterEqual:
1244     Kind = MCBinaryExpr::GTE;
1245     return 3;
1246 
1247   // Intermediate Precedence: <<, >>
1248   case AsmToken::LessLess:
1249     Kind = MCBinaryExpr::Shl;
1250     return 4;
1251   case AsmToken::GreaterGreater:
1252     Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1253     return 4;
1254 
1255   // High Intermediate Precedence: +, -
1256   case AsmToken::Plus:
1257     Kind = MCBinaryExpr::Add;
1258     return 5;
1259   case AsmToken::Minus:
1260     Kind = MCBinaryExpr::Sub;
1261     return 5;
1262 
1263   // Highest Precedence: *, /, %
1264   case AsmToken::Star:
1265     Kind = MCBinaryExpr::Mul;
1266     return 6;
1267   case AsmToken::Slash:
1268     Kind = MCBinaryExpr::Div;
1269     return 6;
1270   case AsmToken::Percent:
1271     Kind = MCBinaryExpr::Mod;
1272     return 6;
1273   }
1274 }
1275 
getGNUBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1276 static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K,
1277                                       MCBinaryExpr::Opcode &Kind,
1278                                       bool ShouldUseLogicalShr) {
1279   switch (K) {
1280   default:
1281     return 0; // not a binop.
1282 
1283   // Lowest Precedence: &&, ||
1284   case AsmToken::AmpAmp:
1285     Kind = MCBinaryExpr::LAnd;
1286     return 2;
1287   case AsmToken::PipePipe:
1288     Kind = MCBinaryExpr::LOr;
1289     return 1;
1290 
1291   // Low Precedence: ==, !=, <>, <, <=, >, >=
1292   case AsmToken::EqualEqual:
1293     Kind = MCBinaryExpr::EQ;
1294     return 3;
1295   case AsmToken::ExclaimEqual:
1296   case AsmToken::LessGreater:
1297     Kind = MCBinaryExpr::NE;
1298     return 3;
1299   case AsmToken::Less:
1300     Kind = MCBinaryExpr::LT;
1301     return 3;
1302   case AsmToken::LessEqual:
1303     Kind = MCBinaryExpr::LTE;
1304     return 3;
1305   case AsmToken::Greater:
1306     Kind = MCBinaryExpr::GT;
1307     return 3;
1308   case AsmToken::GreaterEqual:
1309     Kind = MCBinaryExpr::GTE;
1310     return 3;
1311 
1312   // Low Intermediate Precedence: +, -
1313   case AsmToken::Plus:
1314     Kind = MCBinaryExpr::Add;
1315     return 4;
1316   case AsmToken::Minus:
1317     Kind = MCBinaryExpr::Sub;
1318     return 4;
1319 
1320   // High Intermediate Precedence: |, &, ^
1321   //
1322   // FIXME: gas seems to support '!' as an infix operator?
1323   case AsmToken::Pipe:
1324     Kind = MCBinaryExpr::Or;
1325     return 5;
1326   case AsmToken::Caret:
1327     Kind = MCBinaryExpr::Xor;
1328     return 5;
1329   case AsmToken::Amp:
1330     Kind = MCBinaryExpr::And;
1331     return 5;
1332 
1333   // Highest Precedence: *, /, %, <<, >>
1334   case AsmToken::Star:
1335     Kind = MCBinaryExpr::Mul;
1336     return 6;
1337   case AsmToken::Slash:
1338     Kind = MCBinaryExpr::Div;
1339     return 6;
1340   case AsmToken::Percent:
1341     Kind = MCBinaryExpr::Mod;
1342     return 6;
1343   case AsmToken::LessLess:
1344     Kind = MCBinaryExpr::Shl;
1345     return 6;
1346   case AsmToken::GreaterGreater:
1347     Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1348     return 6;
1349   }
1350 }
1351 
getBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind)1352 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K,
1353                                        MCBinaryExpr::Opcode &Kind) {
1354   bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr();
1355   return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr)
1356                   : getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr);
1357 }
1358 
1359 /// \brief Parse all binary operators with precedence >= 'Precedence'.
1360 /// Res contains the LHS of the expression on input.
parseBinOpRHS(unsigned Precedence,const MCExpr * & Res,SMLoc & EndLoc)1361 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1362                               SMLoc &EndLoc) {
1363   while (1) {
1364     MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1365     unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1366 
1367     // If the next token is lower precedence than we are allowed to eat, return
1368     // successfully with what we ate already.
1369     if (TokPrec < Precedence)
1370       return false;
1371 
1372     Lex();
1373 
1374     // Eat the next primary expression.
1375     const MCExpr *RHS;
1376     if (parsePrimaryExpr(RHS, EndLoc))
1377       return true;
1378 
1379     // If BinOp binds less tightly with RHS than the operator after RHS, let
1380     // the pending operator take RHS as its LHS.
1381     MCBinaryExpr::Opcode Dummy;
1382     unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1383     if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc))
1384       return true;
1385 
1386     // Merge LHS and RHS according to operator.
1387     Res = MCBinaryExpr::create(Kind, Res, RHS, getContext());
1388   }
1389 }
1390 
1391 /// ParseStatement:
1392 ///   ::= EndOfStatement
1393 ///   ::= Label* Directive ...Operands... EndOfStatement
1394 ///   ::= Label* Identifier OperandList* EndOfStatement
parseStatement(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)1395 bool AsmParser::parseStatement(ParseStatementInfo &Info,
1396                                MCAsmParserSemaCallback *SI) {
1397   // Eat initial spaces and comments
1398   while (Lexer.is(AsmToken::Space))
1399     Lex();
1400   if (Lexer.is(AsmToken::EndOfStatement)) {
1401     // if this is a line comment we can drop it safely
1402     if (getTok().getString().front() == '\r' ||
1403         getTok().getString().front() == '\n')
1404       Out.AddBlankLine();
1405     Lex();
1406     return false;
1407   }
1408   if (Lexer.is(AsmToken::Hash)) {
1409     // Seeing a hash here means that it was an end-of-line comment in
1410     // an asm syntax where hash's are not comment and the previous
1411     // statement parser did not check the end of statement. Relex as
1412     // EndOfStatement.
1413     StringRef CommentStr = parseStringToEndOfStatement();
1414     Lexer.Lex();
1415     Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr));
1416     return false;
1417   }
1418   // Statements always start with an identifier.
1419   AsmToken ID = getTok();
1420   SMLoc IDLoc = ID.getLoc();
1421   StringRef IDVal;
1422   int64_t LocalLabelVal = -1;
1423   if (Lexer.is(AsmToken::HashDirective))
1424     return parseCppHashLineFilenameComment(IDLoc);
1425   // Allow an integer followed by a ':' as a directional local label.
1426   if (Lexer.is(AsmToken::Integer)) {
1427     LocalLabelVal = getTok().getIntVal();
1428     if (LocalLabelVal < 0) {
1429       if (!TheCondState.Ignore)
1430         return TokError("unexpected token at start of statement");
1431       IDVal = "";
1432     } else {
1433       IDVal = getTok().getString();
1434       Lex(); // Consume the integer token to be used as an identifier token.
1435       if (Lexer.getKind() != AsmToken::Colon) {
1436         if (!TheCondState.Ignore)
1437           return TokError("unexpected token at start of statement");
1438       }
1439     }
1440   } else if (Lexer.is(AsmToken::Dot)) {
1441     // Treat '.' as a valid identifier in this context.
1442     Lex();
1443     IDVal = ".";
1444   } else if (Lexer.is(AsmToken::LCurly)) {
1445     // Treat '{' as a valid identifier in this context.
1446     Lex();
1447     IDVal = "{";
1448 
1449   } else if (Lexer.is(AsmToken::RCurly)) {
1450     // Treat '}' as a valid identifier in this context.
1451     Lex();
1452     IDVal = "}";
1453   } else if (parseIdentifier(IDVal)) {
1454     if (!TheCondState.Ignore)
1455       return TokError("unexpected token at start of statement");
1456     IDVal = "";
1457   }
1458 
1459   // Handle conditional assembly here before checking for skipping.  We
1460   // have to do this so that .endif isn't skipped in a ".if 0" block for
1461   // example.
1462   StringMap<DirectiveKind>::const_iterator DirKindIt =
1463       DirectiveKindMap.find(IDVal);
1464   DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end())
1465                               ? DK_NO_DIRECTIVE
1466                               : DirKindIt->getValue();
1467   switch (DirKind) {
1468   default:
1469     break;
1470   case DK_IF:
1471   case DK_IFEQ:
1472   case DK_IFGE:
1473   case DK_IFGT:
1474   case DK_IFLE:
1475   case DK_IFLT:
1476   case DK_IFNE:
1477     return parseDirectiveIf(IDLoc, DirKind);
1478   case DK_IFB:
1479     return parseDirectiveIfb(IDLoc, true);
1480   case DK_IFNB:
1481     return parseDirectiveIfb(IDLoc, false);
1482   case DK_IFC:
1483     return parseDirectiveIfc(IDLoc, true);
1484   case DK_IFEQS:
1485     return parseDirectiveIfeqs(IDLoc, true);
1486   case DK_IFNC:
1487     return parseDirectiveIfc(IDLoc, false);
1488   case DK_IFNES:
1489     return parseDirectiveIfeqs(IDLoc, false);
1490   case DK_IFDEF:
1491     return parseDirectiveIfdef(IDLoc, true);
1492   case DK_IFNDEF:
1493   case DK_IFNOTDEF:
1494     return parseDirectiveIfdef(IDLoc, false);
1495   case DK_ELSEIF:
1496     return parseDirectiveElseIf(IDLoc);
1497   case DK_ELSE:
1498     return parseDirectiveElse(IDLoc);
1499   case DK_ENDIF:
1500     return parseDirectiveEndIf(IDLoc);
1501   }
1502 
1503   // Ignore the statement if in the middle of inactive conditional
1504   // (e.g. ".if 0").
1505   if (TheCondState.Ignore) {
1506     eatToEndOfStatement();
1507     return false;
1508   }
1509 
1510   // FIXME: Recurse on local labels?
1511 
1512   // See what kind of statement we have.
1513   switch (Lexer.getKind()) {
1514   case AsmToken::Colon: {
1515     if (!getTargetParser().isLabel(ID))
1516       break;
1517     checkForValidSection();
1518 
1519     // identifier ':'   -> Label.
1520     Lex();
1521 
1522     // Diagnose attempt to use '.' as a label.
1523     if (IDVal == ".")
1524       return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1525 
1526     // Diagnose attempt to use a variable as a label.
1527     //
1528     // FIXME: Diagnostics. Note the location of the definition as a label.
1529     // FIXME: This doesn't diagnose assignment to a symbol which has been
1530     // implicitly marked as external.
1531     MCSymbol *Sym;
1532     if (LocalLabelVal == -1) {
1533       if (ParsingInlineAsm && SI) {
1534         StringRef RewrittenLabel =
1535             SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true);
1536         assert(RewrittenLabel.size() &&
1537                "We should have an internal name here.");
1538         Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(),
1539                                        RewrittenLabel);
1540         IDVal = RewrittenLabel;
1541       }
1542       Sym = getContext().getOrCreateSymbol(IDVal);
1543     } else
1544       Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal);
1545 
1546     Sym->redefineIfPossible();
1547 
1548     if (!Sym->isUndefined() || Sym->isVariable())
1549       return Error(IDLoc, "invalid symbol redefinition");
1550 
1551     // End of Labels should be treated as end of line for lexing
1552     // purposes but that information is not available to the Lexer who
1553     // does not understand Labels. This may cause us to see a Hash
1554     // here instead of a preprocessor line comment.
1555     if (getTok().is(AsmToken::Hash)) {
1556       StringRef CommentStr = parseStringToEndOfStatement();
1557       Lexer.Lex();
1558       Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr));
1559     }
1560 
1561     // Consume any end of statement token, if present, to avoid spurious
1562     // AddBlankLine calls().
1563     if (getTok().is(AsmToken::EndOfStatement)) {
1564       Lex();
1565     }
1566 
1567     // Emit the label.
1568     if (!ParsingInlineAsm)
1569       Out.EmitLabel(Sym);
1570 
1571     // If we are generating dwarf for assembly source files then gather the
1572     // info to make a dwarf label entry for this label if needed.
1573     if (getContext().getGenDwarfForAssembly())
1574       MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1575                                  IDLoc);
1576 
1577     getTargetParser().onLabelParsed(Sym);
1578 
1579 
1580 
1581     return false;
1582   }
1583 
1584   case AsmToken::Equal:
1585     if (!getTargetParser().equalIsAsmAssignment())
1586       break;
1587     // identifier '=' ... -> assignment statement
1588     Lex();
1589 
1590     return parseAssignment(IDVal, true);
1591 
1592   default: // Normal instruction or directive.
1593     break;
1594   }
1595 
1596   // If macros are enabled, check to see if this is a macro instantiation.
1597   if (areMacrosEnabled())
1598     if (const MCAsmMacro *M = lookupMacro(IDVal)) {
1599       return handleMacroEntry(M, IDLoc);
1600     }
1601 
1602   // Otherwise, we have a normal instruction or directive.
1603 
1604   // Directives start with "."
1605   if (IDVal[0] == '.' && IDVal != ".") {
1606     // There are several entities interested in parsing directives:
1607     //
1608     // 1. The target-specific assembly parser. Some directives are target
1609     //    specific or may potentially behave differently on certain targets.
1610     // 2. Asm parser extensions. For example, platform-specific parsers
1611     //    (like the ELF parser) register themselves as extensions.
1612     // 3. The generic directive parser implemented by this class. These are
1613     //    all the directives that behave in a target and platform independent
1614     //    manner, or at least have a default behavior that's shared between
1615     //    all targets and platforms.
1616 
1617     // First query the target-specific parser. It will return 'true' if it
1618     // isn't interested in this directive.
1619     if (!getTargetParser().ParseDirective(ID))
1620       return false;
1621 
1622     // Next, check the extension directive map to see if any extension has
1623     // registered itself to parse this directive.
1624     std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
1625         ExtensionDirectiveMap.lookup(IDVal);
1626     if (Handler.first)
1627       return (*Handler.second)(Handler.first, IDVal, IDLoc);
1628 
1629     // Finally, if no one else is interested in this directive, it must be
1630     // generic and familiar to this class.
1631     switch (DirKind) {
1632     default:
1633       break;
1634     case DK_SET:
1635     case DK_EQU:
1636       return parseDirectiveSet(IDVal, true);
1637     case DK_EQUIV:
1638       return parseDirectiveSet(IDVal, false);
1639     case DK_ASCII:
1640       return parseDirectiveAscii(IDVal, false);
1641     case DK_ASCIZ:
1642     case DK_STRING:
1643       return parseDirectiveAscii(IDVal, true);
1644     case DK_BYTE:
1645       return parseDirectiveValue(1);
1646     case DK_SHORT:
1647     case DK_VALUE:
1648     case DK_2BYTE:
1649       return parseDirectiveValue(2);
1650     case DK_LONG:
1651     case DK_INT:
1652     case DK_4BYTE:
1653       return parseDirectiveValue(4);
1654     case DK_QUAD:
1655     case DK_8BYTE:
1656       return parseDirectiveValue(8);
1657     case DK_OCTA:
1658       return parseDirectiveOctaValue();
1659     case DK_SINGLE:
1660     case DK_FLOAT:
1661       return parseDirectiveRealValue(APFloat::IEEEsingle);
1662     case DK_DOUBLE:
1663       return parseDirectiveRealValue(APFloat::IEEEdouble);
1664     case DK_ALIGN: {
1665       bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1666       return parseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1667     }
1668     case DK_ALIGN32: {
1669       bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1670       return parseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1671     }
1672     case DK_BALIGN:
1673       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1674     case DK_BALIGNW:
1675       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1676     case DK_BALIGNL:
1677       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1678     case DK_P2ALIGN:
1679       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1680     case DK_P2ALIGNW:
1681       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1682     case DK_P2ALIGNL:
1683       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1684     case DK_ORG:
1685       return parseDirectiveOrg();
1686     case DK_FILL:
1687       return parseDirectiveFill();
1688     case DK_ZERO:
1689       return parseDirectiveZero();
1690     case DK_EXTERN:
1691       eatToEndOfStatement(); // .extern is the default, ignore it.
1692       return false;
1693     case DK_GLOBL:
1694     case DK_GLOBAL:
1695       return parseDirectiveSymbolAttribute(MCSA_Global);
1696     case DK_LAZY_REFERENCE:
1697       return parseDirectiveSymbolAttribute(MCSA_LazyReference);
1698     case DK_NO_DEAD_STRIP:
1699       return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1700     case DK_SYMBOL_RESOLVER:
1701       return parseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1702     case DK_PRIVATE_EXTERN:
1703       return parseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1704     case DK_REFERENCE:
1705       return parseDirectiveSymbolAttribute(MCSA_Reference);
1706     case DK_WEAK_DEFINITION:
1707       return parseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1708     case DK_WEAK_REFERENCE:
1709       return parseDirectiveSymbolAttribute(MCSA_WeakReference);
1710     case DK_WEAK_DEF_CAN_BE_HIDDEN:
1711       return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1712     case DK_COMM:
1713     case DK_COMMON:
1714       return parseDirectiveComm(/*IsLocal=*/false);
1715     case DK_LCOMM:
1716       return parseDirectiveComm(/*IsLocal=*/true);
1717     case DK_ABORT:
1718       return parseDirectiveAbort();
1719     case DK_INCLUDE:
1720       return parseDirectiveInclude();
1721     case DK_INCBIN:
1722       return parseDirectiveIncbin();
1723     case DK_CODE16:
1724     case DK_CODE16GCC:
1725       return TokError(Twine(IDVal) +
1726                       " not currently supported for this target");
1727     case DK_REPT:
1728       return parseDirectiveRept(IDLoc, IDVal);
1729     case DK_IRP:
1730       return parseDirectiveIrp(IDLoc);
1731     case DK_IRPC:
1732       return parseDirectiveIrpc(IDLoc);
1733     case DK_ENDR:
1734       return parseDirectiveEndr(IDLoc);
1735     case DK_BUNDLE_ALIGN_MODE:
1736       return parseDirectiveBundleAlignMode();
1737     case DK_BUNDLE_LOCK:
1738       return parseDirectiveBundleLock();
1739     case DK_BUNDLE_UNLOCK:
1740       return parseDirectiveBundleUnlock();
1741     case DK_SLEB128:
1742       return parseDirectiveLEB128(true);
1743     case DK_ULEB128:
1744       return parseDirectiveLEB128(false);
1745     case DK_SPACE:
1746     case DK_SKIP:
1747       return parseDirectiveSpace(IDVal);
1748     case DK_FILE:
1749       return parseDirectiveFile(IDLoc);
1750     case DK_LINE:
1751       return parseDirectiveLine();
1752     case DK_LOC:
1753       return parseDirectiveLoc();
1754     case DK_STABS:
1755       return parseDirectiveStabs();
1756     case DK_CV_FILE:
1757       return parseDirectiveCVFile();
1758     case DK_CV_LOC:
1759       return parseDirectiveCVLoc();
1760     case DK_CV_LINETABLE:
1761       return parseDirectiveCVLinetable();
1762     case DK_CV_INLINE_LINETABLE:
1763       return parseDirectiveCVInlineLinetable();
1764     case DK_CV_DEF_RANGE:
1765       return parseDirectiveCVDefRange();
1766     case DK_CV_STRINGTABLE:
1767       return parseDirectiveCVStringTable();
1768     case DK_CV_FILECHECKSUMS:
1769       return parseDirectiveCVFileChecksums();
1770     case DK_CFI_SECTIONS:
1771       return parseDirectiveCFISections();
1772     case DK_CFI_STARTPROC:
1773       return parseDirectiveCFIStartProc();
1774     case DK_CFI_ENDPROC:
1775       return parseDirectiveCFIEndProc();
1776     case DK_CFI_DEF_CFA:
1777       return parseDirectiveCFIDefCfa(IDLoc);
1778     case DK_CFI_DEF_CFA_OFFSET:
1779       return parseDirectiveCFIDefCfaOffset();
1780     case DK_CFI_ADJUST_CFA_OFFSET:
1781       return parseDirectiveCFIAdjustCfaOffset();
1782     case DK_CFI_DEF_CFA_REGISTER:
1783       return parseDirectiveCFIDefCfaRegister(IDLoc);
1784     case DK_CFI_OFFSET:
1785       return parseDirectiveCFIOffset(IDLoc);
1786     case DK_CFI_REL_OFFSET:
1787       return parseDirectiveCFIRelOffset(IDLoc);
1788     case DK_CFI_PERSONALITY:
1789       return parseDirectiveCFIPersonalityOrLsda(true);
1790     case DK_CFI_LSDA:
1791       return parseDirectiveCFIPersonalityOrLsda(false);
1792     case DK_CFI_REMEMBER_STATE:
1793       return parseDirectiveCFIRememberState();
1794     case DK_CFI_RESTORE_STATE:
1795       return parseDirectiveCFIRestoreState();
1796     case DK_CFI_SAME_VALUE:
1797       return parseDirectiveCFISameValue(IDLoc);
1798     case DK_CFI_RESTORE:
1799       return parseDirectiveCFIRestore(IDLoc);
1800     case DK_CFI_ESCAPE:
1801       return parseDirectiveCFIEscape();
1802     case DK_CFI_SIGNAL_FRAME:
1803       return parseDirectiveCFISignalFrame();
1804     case DK_CFI_UNDEFINED:
1805       return parseDirectiveCFIUndefined(IDLoc);
1806     case DK_CFI_REGISTER:
1807       return parseDirectiveCFIRegister(IDLoc);
1808     case DK_CFI_WINDOW_SAVE:
1809       return parseDirectiveCFIWindowSave();
1810     case DK_MACROS_ON:
1811     case DK_MACROS_OFF:
1812       return parseDirectiveMacrosOnOff(IDVal);
1813     case DK_MACRO:
1814       return parseDirectiveMacro(IDLoc);
1815     case DK_EXITM:
1816       return parseDirectiveExitMacro(IDVal);
1817     case DK_ENDM:
1818     case DK_ENDMACRO:
1819       return parseDirectiveEndMacro(IDVal);
1820     case DK_PURGEM:
1821       return parseDirectivePurgeMacro(IDLoc);
1822     case DK_END:
1823       return parseDirectiveEnd(IDLoc);
1824     case DK_ERR:
1825       return parseDirectiveError(IDLoc, false);
1826     case DK_ERROR:
1827       return parseDirectiveError(IDLoc, true);
1828     case DK_WARNING:
1829       return parseDirectiveWarning(IDLoc);
1830     case DK_RELOC:
1831       return parseDirectiveReloc(IDLoc);
1832     }
1833 
1834     return Error(IDLoc, "unknown directive");
1835   }
1836 
1837   // __asm _emit or __asm __emit
1838   if (ParsingInlineAsm && (IDVal == "_emit" || IDVal == "__emit" ||
1839                            IDVal == "_EMIT" || IDVal == "__EMIT"))
1840     return parseDirectiveMSEmit(IDLoc, Info, IDVal.size());
1841 
1842   // __asm align
1843   if (ParsingInlineAsm && (IDVal == "align" || IDVal == "ALIGN"))
1844     return parseDirectiveMSAlign(IDLoc, Info);
1845 
1846   if (ParsingInlineAsm && (IDVal == "even"))
1847     Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4);
1848   checkForValidSection();
1849 
1850   // Canonicalize the opcode to lower case.
1851   std::string OpcodeStr = IDVal.lower();
1852   ParseInstructionInfo IInfo(Info.AsmRewrites);
1853   bool HadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID,
1854                                                      Info.ParsedOperands);
1855   Info.ParseError = HadError;
1856 
1857   // Dump the parsed representation, if requested.
1858   if (getShowParsedOperands()) {
1859     SmallString<256> Str;
1860     raw_svector_ostream OS(Str);
1861     OS << "parsed instruction: [";
1862     for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
1863       if (i != 0)
1864         OS << ", ";
1865       Info.ParsedOperands[i]->print(OS);
1866     }
1867     OS << "]";
1868 
1869     printMessage(IDLoc, SourceMgr::DK_Note, OS.str());
1870   }
1871 
1872   // If we are generating dwarf for the current section then generate a .loc
1873   // directive for the instruction.
1874   if (!HadError && getContext().getGenDwarfForAssembly() &&
1875       getContext().getGenDwarfSectionSyms().count(
1876           getStreamer().getCurrentSection().first)) {
1877     unsigned Line;
1878     if (ActiveMacros.empty())
1879       Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
1880     else
1881       Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc,
1882                                    ActiveMacros.front()->ExitBuffer);
1883 
1884     // If we previously parsed a cpp hash file line comment then make sure the
1885     // current Dwarf File is for the CppHashFilename if not then emit the
1886     // Dwarf File table for it and adjust the line number for the .loc.
1887     if (CppHashInfo.Filename.size()) {
1888       unsigned FileNumber = getStreamer().EmitDwarfFileDirective(
1889           0, StringRef(), CppHashInfo.Filename);
1890       getContext().setGenDwarfFileNumber(FileNumber);
1891 
1892       // Since SrcMgr.FindLineNumber() is slow and messes up the SourceMgr's
1893       // cache with the different Loc from the call above we save the last
1894       // info we queried here with SrcMgr.FindLineNumber().
1895       unsigned CppHashLocLineNo;
1896       if (LastQueryIDLoc == CppHashInfo.Loc &&
1897           LastQueryBuffer == CppHashInfo.Buf)
1898         CppHashLocLineNo = LastQueryLine;
1899       else {
1900         CppHashLocLineNo =
1901             SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf);
1902         LastQueryLine = CppHashLocLineNo;
1903         LastQueryIDLoc = CppHashInfo.Loc;
1904         LastQueryBuffer = CppHashInfo.Buf;
1905       }
1906       Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo);
1907     }
1908 
1909     getStreamer().EmitDwarfLocDirective(
1910         getContext().getGenDwarfFileNumber(), Line, 0,
1911         DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0,
1912         StringRef());
1913   }
1914 
1915   // If parsing succeeded, match the instruction.
1916   if (!HadError) {
1917     uint64_t ErrorInfo;
1918     getTargetParser().MatchAndEmitInstruction(IDLoc, Info.Opcode,
1919                                               Info.ParsedOperands, Out,
1920                                               ErrorInfo, ParsingInlineAsm);
1921   }
1922 
1923   // Don't skip the rest of the line, the instruction parser is responsible for
1924   // that.
1925   return false;
1926 }
1927 
1928 // Parse and erase curly braces marking block start/end
1929 bool
parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> & AsmStrRewrites)1930 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) {
1931   // Identify curly brace marking block start/end
1932   if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly))
1933     return false;
1934 
1935   SMLoc StartLoc = Lexer.getLoc();
1936   Lex(); // Eat the brace
1937   if (Lexer.is(AsmToken::EndOfStatement))
1938     Lex(); // Eat EndOfStatement following the brace
1939 
1940   // Erase the block start/end brace from the output asm string
1941   AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() -
1942                                                   StartLoc.getPointer());
1943   return true;
1944 }
1945 
1946 /// parseCppHashLineFilenameComment as this:
1947 ///   ::= # number "filename"
parseCppHashLineFilenameComment(SMLoc L)1948 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L) {
1949   Lex(); // Eat the hash token.
1950   // Lexer only ever emits HashDirective if it fully formed if it's
1951   // done the checking already so this is an internal error.
1952   assert(getTok().is(AsmToken::Integer) &&
1953          "Lexing Cpp line comment: Expected Integer");
1954   int64_t LineNumber = getTok().getIntVal();
1955   Lex();
1956   assert(getTok().is(AsmToken::String) &&
1957          "Lexing Cpp line comment: Expected String");
1958   StringRef Filename = getTok().getString();
1959   Lex();
1960   // Get rid of the enclosing quotes.
1961   Filename = Filename.substr(1, Filename.size() - 2);
1962 
1963   // Save the SMLoc, Filename and LineNumber for later use by diagnostics.
1964   CppHashInfo.Loc = L;
1965   CppHashInfo.Filename = Filename;
1966   CppHashInfo.LineNumber = LineNumber;
1967   CppHashInfo.Buf = CurBuffer;
1968   return false;
1969 }
1970 
1971 /// \brief will use the last parsed cpp hash line filename comment
1972 /// for the Filename and LineNo if any in the diagnostic.
DiagHandler(const SMDiagnostic & Diag,void * Context)1973 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
1974   const AsmParser *Parser = static_cast<const AsmParser *>(Context);
1975   raw_ostream &OS = errs();
1976 
1977   const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
1978   SMLoc DiagLoc = Diag.getLoc();
1979   unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1980   unsigned CppHashBuf =
1981       Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc);
1982 
1983   // Like SourceMgr::printMessage() we need to print the include stack if any
1984   // before printing the message.
1985   unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1986   if (!Parser->SavedDiagHandler && DiagCurBuffer &&
1987       DiagCurBuffer != DiagSrcMgr.getMainFileID()) {
1988     SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
1989     DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
1990   }
1991 
1992   // If we have not parsed a cpp hash line filename comment or the source
1993   // manager changed or buffer changed (like in a nested include) then just
1994   // print the normal diagnostic using its Filename and LineNo.
1995   if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr ||
1996       DiagBuf != CppHashBuf) {
1997     if (Parser->SavedDiagHandler)
1998       Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
1999     else
2000       Diag.print(nullptr, OS);
2001     return;
2002   }
2003 
2004   // Use the CppHashFilename and calculate a line number based on the
2005   // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc
2006   // for the diagnostic.
2007   const std::string &Filename = Parser->CppHashInfo.Filename;
2008 
2009   int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
2010   int CppHashLocLineNo =
2011       Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf);
2012   int LineNo =
2013       Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo);
2014 
2015   SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo,
2016                        Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(),
2017                        Diag.getLineContents(), Diag.getRanges());
2018 
2019   if (Parser->SavedDiagHandler)
2020     Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext);
2021   else
2022     NewDiag.print(nullptr, OS);
2023 }
2024 
2025 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
2026 // difference being that that function accepts '@' as part of identifiers and
2027 // we can't do that. AsmLexer.cpp should probably be changed to handle
2028 // '@' as a special case when needed.
isIdentifierChar(char c)2029 static bool isIdentifierChar(char c) {
2030   return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' ||
2031          c == '.';
2032 }
2033 
expandMacro(raw_svector_ostream & OS,StringRef Body,ArrayRef<MCAsmMacroParameter> Parameters,ArrayRef<MCAsmMacroArgument> A,bool EnableAtPseudoVariable,SMLoc L)2034 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
2035                             ArrayRef<MCAsmMacroParameter> Parameters,
2036                             ArrayRef<MCAsmMacroArgument> A,
2037                             bool EnableAtPseudoVariable, SMLoc L) {
2038   unsigned NParameters = Parameters.size();
2039   bool HasVararg = NParameters ? Parameters.back().Vararg : false;
2040   if ((!IsDarwin || NParameters != 0) && NParameters != A.size())
2041     return Error(L, "Wrong number of arguments");
2042 
2043   // A macro without parameters is handled differently on Darwin:
2044   // gas accepts no arguments and does no substitutions
2045   while (!Body.empty()) {
2046     // Scan for the next substitution.
2047     std::size_t End = Body.size(), Pos = 0;
2048     for (; Pos != End; ++Pos) {
2049       // Check for a substitution or escape.
2050       if (IsDarwin && !NParameters) {
2051         // This macro has no parameters, look for $0, $1, etc.
2052         if (Body[Pos] != '$' || Pos + 1 == End)
2053           continue;
2054 
2055         char Next = Body[Pos + 1];
2056         if (Next == '$' || Next == 'n' ||
2057             isdigit(static_cast<unsigned char>(Next)))
2058           break;
2059       } else {
2060         // This macro has parameters, look for \foo, \bar, etc.
2061         if (Body[Pos] == '\\' && Pos + 1 != End)
2062           break;
2063       }
2064     }
2065 
2066     // Add the prefix.
2067     OS << Body.slice(0, Pos);
2068 
2069     // Check if we reached the end.
2070     if (Pos == End)
2071       break;
2072 
2073     if (IsDarwin && !NParameters) {
2074       switch (Body[Pos + 1]) {
2075       // $$ => $
2076       case '$':
2077         OS << '$';
2078         break;
2079 
2080       // $n => number of arguments
2081       case 'n':
2082         OS << A.size();
2083         break;
2084 
2085       // $[0-9] => argument
2086       default: {
2087         // Missing arguments are ignored.
2088         unsigned Index = Body[Pos + 1] - '0';
2089         if (Index >= A.size())
2090           break;
2091 
2092         // Otherwise substitute with the token values, with spaces eliminated.
2093         for (const AsmToken &Token : A[Index])
2094           OS << Token.getString();
2095         break;
2096       }
2097       }
2098       Pos += 2;
2099     } else {
2100       unsigned I = Pos + 1;
2101 
2102       // Check for the \@ pseudo-variable.
2103       if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End)
2104         ++I;
2105       else
2106         while (isIdentifierChar(Body[I]) && I + 1 != End)
2107           ++I;
2108 
2109       const char *Begin = Body.data() + Pos + 1;
2110       StringRef Argument(Begin, I - (Pos + 1));
2111       unsigned Index = 0;
2112 
2113       if (Argument == "@") {
2114         OS << NumOfMacroInstantiations;
2115         Pos += 2;
2116       } else {
2117         for (; Index < NParameters; ++Index)
2118           if (Parameters[Index].Name == Argument)
2119             break;
2120 
2121         if (Index == NParameters) {
2122           if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
2123             Pos += 3;
2124           else {
2125             OS << '\\' << Argument;
2126             Pos = I;
2127           }
2128         } else {
2129           bool VarargParameter = HasVararg && Index == (NParameters - 1);
2130           for (const AsmToken &Token : A[Index])
2131             // We expect no quotes around the string's contents when
2132             // parsing for varargs.
2133             if (Token.getKind() != AsmToken::String || VarargParameter)
2134               OS << Token.getString();
2135             else
2136               OS << Token.getStringContents();
2137 
2138           Pos += 1 + Argument.size();
2139         }
2140       }
2141     }
2142     // Update the scan point.
2143     Body = Body.substr(Pos);
2144   }
2145 
2146   return false;
2147 }
2148 
MacroInstantiation(SMLoc IL,int EB,SMLoc EL,size_t CondStackDepth)2149 MacroInstantiation::MacroInstantiation(SMLoc IL, int EB, SMLoc EL,
2150                                        size_t CondStackDepth)
2151     : InstantiationLoc(IL), ExitBuffer(EB), ExitLoc(EL),
2152       CondStackDepth(CondStackDepth) {}
2153 
isOperator(AsmToken::TokenKind kind)2154 static bool isOperator(AsmToken::TokenKind kind) {
2155   switch (kind) {
2156   default:
2157     return false;
2158   case AsmToken::Plus:
2159   case AsmToken::Minus:
2160   case AsmToken::Tilde:
2161   case AsmToken::Slash:
2162   case AsmToken::Star:
2163   case AsmToken::Dot:
2164   case AsmToken::Equal:
2165   case AsmToken::EqualEqual:
2166   case AsmToken::Pipe:
2167   case AsmToken::PipePipe:
2168   case AsmToken::Caret:
2169   case AsmToken::Amp:
2170   case AsmToken::AmpAmp:
2171   case AsmToken::Exclaim:
2172   case AsmToken::ExclaimEqual:
2173   case AsmToken::Less:
2174   case AsmToken::LessEqual:
2175   case AsmToken::LessLess:
2176   case AsmToken::LessGreater:
2177   case AsmToken::Greater:
2178   case AsmToken::GreaterEqual:
2179   case AsmToken::GreaterGreater:
2180     return true;
2181   }
2182 }
2183 
2184 namespace {
2185 class AsmLexerSkipSpaceRAII {
2186 public:
AsmLexerSkipSpaceRAII(AsmLexer & Lexer,bool SkipSpace)2187   AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) {
2188     Lexer.setSkipSpace(SkipSpace);
2189   }
2190 
~AsmLexerSkipSpaceRAII()2191   ~AsmLexerSkipSpaceRAII() {
2192     Lexer.setSkipSpace(true);
2193   }
2194 
2195 private:
2196   AsmLexer &Lexer;
2197 };
2198 }
2199 
parseMacroArgument(MCAsmMacroArgument & MA,bool Vararg)2200 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) {
2201 
2202   if (Vararg) {
2203     if (Lexer.isNot(AsmToken::EndOfStatement)) {
2204       StringRef Str = parseStringToEndOfStatement();
2205       MA.emplace_back(AsmToken::String, Str);
2206     }
2207     return false;
2208   }
2209 
2210   unsigned ParenLevel = 0;
2211 
2212   // Darwin doesn't use spaces to delmit arguments.
2213   AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin);
2214 
2215   bool SpaceEaten;
2216 
2217   for (;;) {
2218     SpaceEaten = false;
2219     if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
2220       return TokError("unexpected token in macro instantiation");
2221 
2222     if (ParenLevel == 0) {
2223 
2224       if (Lexer.is(AsmToken::Comma))
2225         break;
2226 
2227       if (Lexer.is(AsmToken::Space)) {
2228         SpaceEaten = true;
2229         Lexer.Lex(); // Eat spaces
2230       }
2231 
2232       // Spaces can delimit parameters, but could also be part an expression.
2233       // If the token after a space is an operator, add the token and the next
2234       // one into this argument
2235       if (!IsDarwin) {
2236         if (isOperator(Lexer.getKind())) {
2237           MA.push_back(getTok());
2238           Lexer.Lex();
2239 
2240           // Whitespace after an operator can be ignored.
2241           if (Lexer.is(AsmToken::Space))
2242             Lexer.Lex();
2243 
2244           continue;
2245         }
2246       }
2247       if (SpaceEaten)
2248         break;
2249     }
2250 
2251     // handleMacroEntry relies on not advancing the lexer here
2252     // to be able to fill in the remaining default parameter values
2253     if (Lexer.is(AsmToken::EndOfStatement))
2254       break;
2255 
2256     // Adjust the current parentheses level.
2257     if (Lexer.is(AsmToken::LParen))
2258       ++ParenLevel;
2259     else if (Lexer.is(AsmToken::RParen) && ParenLevel)
2260       --ParenLevel;
2261 
2262     // Append the token to the current argument list.
2263     MA.push_back(getTok());
2264     Lexer.Lex();
2265   }
2266 
2267   if (ParenLevel != 0)
2268     return TokError("unbalanced parentheses in macro argument");
2269   return false;
2270 }
2271 
2272 // Parse the macro instantiation arguments.
parseMacroArguments(const MCAsmMacro * M,MCAsmMacroArguments & A)2273 bool AsmParser::parseMacroArguments(const MCAsmMacro *M,
2274                                     MCAsmMacroArguments &A) {
2275   const unsigned NParameters = M ? M->Parameters.size() : 0;
2276   bool NamedParametersFound = false;
2277   SmallVector<SMLoc, 4> FALocs;
2278 
2279   A.resize(NParameters);
2280   FALocs.resize(NParameters);
2281 
2282   // Parse two kinds of macro invocations:
2283   // - macros defined without any parameters accept an arbitrary number of them
2284   // - macros defined with parameters accept at most that many of them
2285   bool HasVararg = NParameters ? M->Parameters.back().Vararg : false;
2286   for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
2287        ++Parameter) {
2288     SMLoc IDLoc = Lexer.getLoc();
2289     MCAsmMacroParameter FA;
2290 
2291     if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) {
2292       if (parseIdentifier(FA.Name)) {
2293         Error(IDLoc, "invalid argument identifier for formal argument");
2294         eatToEndOfStatement();
2295         return true;
2296       }
2297 
2298       if (!Lexer.is(AsmToken::Equal)) {
2299         TokError("expected '=' after formal parameter identifier");
2300         eatToEndOfStatement();
2301         return true;
2302       }
2303       Lex();
2304 
2305       NamedParametersFound = true;
2306     }
2307 
2308     if (NamedParametersFound && FA.Name.empty()) {
2309       Error(IDLoc, "cannot mix positional and keyword arguments");
2310       eatToEndOfStatement();
2311       return true;
2312     }
2313 
2314     bool Vararg = HasVararg && Parameter == (NParameters - 1);
2315     if (parseMacroArgument(FA.Value, Vararg))
2316       return true;
2317 
2318     unsigned PI = Parameter;
2319     if (!FA.Name.empty()) {
2320       unsigned FAI = 0;
2321       for (FAI = 0; FAI < NParameters; ++FAI)
2322         if (M->Parameters[FAI].Name == FA.Name)
2323           break;
2324 
2325       if (FAI >= NParameters) {
2326         assert(M && "expected macro to be defined");
2327         Error(IDLoc,
2328               "parameter named '" + FA.Name + "' does not exist for macro '" +
2329               M->Name + "'");
2330         return true;
2331       }
2332       PI = FAI;
2333     }
2334 
2335     if (!FA.Value.empty()) {
2336       if (A.size() <= PI)
2337         A.resize(PI + 1);
2338       A[PI] = FA.Value;
2339 
2340       if (FALocs.size() <= PI)
2341         FALocs.resize(PI + 1);
2342 
2343       FALocs[PI] = Lexer.getLoc();
2344     }
2345 
2346     // At the end of the statement, fill in remaining arguments that have
2347     // default values. If there aren't any, then the next argument is
2348     // required but missing
2349     if (Lexer.is(AsmToken::EndOfStatement)) {
2350       bool Failure = false;
2351       for (unsigned FAI = 0; FAI < NParameters; ++FAI) {
2352         if (A[FAI].empty()) {
2353           if (M->Parameters[FAI].Required) {
2354             Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(),
2355                   "missing value for required parameter "
2356                   "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'");
2357             Failure = true;
2358           }
2359 
2360           if (!M->Parameters[FAI].Value.empty())
2361             A[FAI] = M->Parameters[FAI].Value;
2362         }
2363       }
2364       return Failure;
2365     }
2366 
2367     if (Lexer.is(AsmToken::Comma))
2368       Lex();
2369   }
2370 
2371   return TokError("too many positional arguments");
2372 }
2373 
lookupMacro(StringRef Name)2374 const MCAsmMacro *AsmParser::lookupMacro(StringRef Name) {
2375   StringMap<MCAsmMacro>::iterator I = MacroMap.find(Name);
2376   return (I == MacroMap.end()) ? nullptr : &I->getValue();
2377 }
2378 
defineMacro(StringRef Name,MCAsmMacro Macro)2379 void AsmParser::defineMacro(StringRef Name, MCAsmMacro Macro) {
2380   MacroMap.insert(std::make_pair(Name, std::move(Macro)));
2381 }
2382 
undefineMacro(StringRef Name)2383 void AsmParser::undefineMacro(StringRef Name) { MacroMap.erase(Name); }
2384 
handleMacroEntry(const MCAsmMacro * M,SMLoc NameLoc)2385 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) {
2386   // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
2387   // this, although we should protect against infinite loops.
2388   if (ActiveMacros.size() == 20)
2389     return TokError("macros cannot be nested more than 20 levels deep");
2390 
2391   MCAsmMacroArguments A;
2392   if (parseMacroArguments(M, A))
2393     return true;
2394 
2395   // Macro instantiation is lexical, unfortunately. We construct a new buffer
2396   // to hold the macro body with substitutions.
2397   SmallString<256> Buf;
2398   StringRef Body = M->Body;
2399   raw_svector_ostream OS(Buf);
2400 
2401   if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc()))
2402     return true;
2403 
2404   // We include the .endmacro in the buffer as our cue to exit the macro
2405   // instantiation.
2406   OS << ".endmacro\n";
2407 
2408   std::unique_ptr<MemoryBuffer> Instantiation =
2409       MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
2410 
2411   // Create the macro instantiation object and add to the current macro
2412   // instantiation stack.
2413   MacroInstantiation *MI = new MacroInstantiation(
2414       NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size());
2415   ActiveMacros.push_back(MI);
2416 
2417   ++NumOfMacroInstantiations;
2418 
2419   // Jump to the macro instantiation and prime the lexer.
2420   CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
2421   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
2422   Lex();
2423 
2424   return false;
2425 }
2426 
handleMacroExit()2427 void AsmParser::handleMacroExit() {
2428   // Jump to the EndOfStatement we should return to, and consume it.
2429   jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer);
2430   Lex();
2431 
2432   // Pop the instantiation entry.
2433   delete ActiveMacros.back();
2434   ActiveMacros.pop_back();
2435 }
2436 
parseAssignment(StringRef Name,bool allow_redef,bool NoDeadStrip)2437 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef,
2438                                 bool NoDeadStrip) {
2439   MCSymbol *Sym;
2440   const MCExpr *Value;
2441   if (MCParserUtils::parseAssignmentExpression(Name, allow_redef, *this, Sym,
2442                                                Value))
2443     return true;
2444 
2445   if (!Sym) {
2446     // In the case where we parse an expression starting with a '.', we will
2447     // not generate an error, nor will we create a symbol.  In this case we
2448     // should just return out.
2449     return false;
2450   }
2451 
2452   // Do the assignment.
2453   Out.EmitAssignment(Sym, Value);
2454   if (NoDeadStrip)
2455     Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip);
2456 
2457   return false;
2458 }
2459 
2460 /// parseIdentifier:
2461 ///   ::= identifier
2462 ///   ::= string
parseIdentifier(StringRef & Res)2463 bool AsmParser::parseIdentifier(StringRef &Res) {
2464   // The assembler has relaxed rules for accepting identifiers, in particular we
2465   // allow things like '.globl $foo' and '.def @feat.00', which would normally be
2466   // separate tokens. At this level, we have already lexed so we cannot (currently)
2467   // handle this as a context dependent token, instead we detect adjacent tokens
2468   // and return the combined identifier.
2469   if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) {
2470     SMLoc PrefixLoc = getLexer().getLoc();
2471 
2472     // Consume the prefix character, and check for a following identifier.
2473     Lexer.Lex(); // Lexer's Lex guarantees consecutive token.
2474     if (Lexer.isNot(AsmToken::Identifier))
2475       return true;
2476 
2477     // We have a '$' or '@' followed by an identifier, make sure they are adjacent.
2478     if (PrefixLoc.getPointer() + 1 != getTok().getLoc().getPointer())
2479       return true;
2480 
2481     // Construct the joined identifier and consume the token.
2482     Res =
2483         StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1);
2484     Lex(); // Parser Lex to maintain invariants.
2485     return false;
2486   }
2487 
2488   if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String))
2489     return true;
2490 
2491   Res = getTok().getIdentifier();
2492 
2493   Lex(); // Consume the identifier token.
2494 
2495   return false;
2496 }
2497 
2498 /// parseDirectiveSet:
2499 ///   ::= .equ identifier ',' expression
2500 ///   ::= .equiv identifier ',' expression
2501 ///   ::= .set identifier ',' expression
parseDirectiveSet(StringRef IDVal,bool allow_redef)2502 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) {
2503   StringRef Name;
2504 
2505   if (parseIdentifier(Name))
2506     return TokError("expected identifier after '" + Twine(IDVal) + "'");
2507 
2508   if (getLexer().isNot(AsmToken::Comma))
2509     return TokError("unexpected token in '" + Twine(IDVal) + "'");
2510   Lex();
2511 
2512   return parseAssignment(Name, allow_redef, true);
2513 }
2514 
parseEscapedString(std::string & Data)2515 bool AsmParser::parseEscapedString(std::string &Data) {
2516   assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
2517 
2518   Data = "";
2519   StringRef Str = getTok().getStringContents();
2520   for (unsigned i = 0, e = Str.size(); i != e; ++i) {
2521     if (Str[i] != '\\') {
2522       Data += Str[i];
2523       continue;
2524     }
2525 
2526     // Recognize escaped characters. Note that this escape semantics currently
2527     // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
2528     ++i;
2529     if (i == e)
2530       return TokError("unexpected backslash at end of string");
2531 
2532     // Recognize octal sequences.
2533     if ((unsigned)(Str[i] - '0') <= 7) {
2534       // Consume up to three octal characters.
2535       unsigned Value = Str[i] - '0';
2536 
2537       if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2538         ++i;
2539         Value = Value * 8 + (Str[i] - '0');
2540 
2541         if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2542           ++i;
2543           Value = Value * 8 + (Str[i] - '0');
2544         }
2545       }
2546 
2547       if (Value > 255)
2548         return TokError("invalid octal escape sequence (out of range)");
2549 
2550       Data += (unsigned char)Value;
2551       continue;
2552     }
2553 
2554     // Otherwise recognize individual escapes.
2555     switch (Str[i]) {
2556     default:
2557       // Just reject invalid escape sequences for now.
2558       return TokError("invalid escape sequence (unrecognized character)");
2559 
2560     case 'b': Data += '\b'; break;
2561     case 'f': Data += '\f'; break;
2562     case 'n': Data += '\n'; break;
2563     case 'r': Data += '\r'; break;
2564     case 't': Data += '\t'; break;
2565     case '"': Data += '"'; break;
2566     case '\\': Data += '\\'; break;
2567     }
2568   }
2569 
2570   return false;
2571 }
2572 
2573 /// parseDirectiveAscii:
2574 ///   ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
parseDirectiveAscii(StringRef IDVal,bool ZeroTerminated)2575 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
2576   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2577     checkForValidSection();
2578 
2579     for (;;) {
2580       if (getLexer().isNot(AsmToken::String))
2581         return TokError("expected string in '" + Twine(IDVal) + "' directive");
2582 
2583       std::string Data;
2584       if (parseEscapedString(Data))
2585         return true;
2586 
2587       getStreamer().EmitBytes(Data);
2588       if (ZeroTerminated)
2589         getStreamer().EmitBytes(StringRef("\0", 1));
2590 
2591       Lex();
2592 
2593       if (getLexer().is(AsmToken::EndOfStatement))
2594         break;
2595 
2596       if (getLexer().isNot(AsmToken::Comma))
2597         return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
2598       Lex();
2599     }
2600   }
2601 
2602   Lex();
2603   return false;
2604 }
2605 
2606 /// parseDirectiveReloc
2607 ///  ::= .reloc expression , identifier [ , expression ]
parseDirectiveReloc(SMLoc DirectiveLoc)2608 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) {
2609   const MCExpr *Offset;
2610   const MCExpr *Expr = nullptr;
2611 
2612   SMLoc OffsetLoc = Lexer.getTok().getLoc();
2613   if (parseExpression(Offset))
2614     return true;
2615 
2616   // We can only deal with constant expressions at the moment.
2617   int64_t OffsetValue;
2618   if (!Offset->evaluateAsAbsolute(OffsetValue))
2619     return Error(OffsetLoc, "expression is not a constant value");
2620 
2621   if (OffsetValue < 0)
2622     return Error(OffsetLoc, "expression is negative");
2623 
2624   if (Lexer.isNot(AsmToken::Comma))
2625     return TokError("expected comma");
2626   Lex();
2627 
2628   if (Lexer.isNot(AsmToken::Identifier))
2629     return TokError("expected relocation name");
2630   SMLoc NameLoc = Lexer.getTok().getLoc();
2631   StringRef Name = Lexer.getTok().getIdentifier();
2632   Lex();
2633 
2634   if (Lexer.is(AsmToken::Comma)) {
2635     Lex();
2636     SMLoc ExprLoc = Lexer.getLoc();
2637     if (parseExpression(Expr))
2638       return true;
2639 
2640     MCValue Value;
2641     if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr))
2642       return Error(ExprLoc, "expression must be relocatable");
2643   }
2644 
2645   if (Lexer.isNot(AsmToken::EndOfStatement))
2646     return TokError("unexpected token in .reloc directive");
2647 
2648   if (getStreamer().EmitRelocDirective(*Offset, Name, Expr, DirectiveLoc))
2649     return Error(NameLoc, "unknown relocation name");
2650 
2651   return false;
2652 }
2653 
2654 /// parseDirectiveValue
2655 ///  ::= (.byte | .short | ... ) [ expression (, expression)* ]
parseDirectiveValue(unsigned Size)2656 bool AsmParser::parseDirectiveValue(unsigned Size) {
2657   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2658     checkForValidSection();
2659 
2660     for (;;) {
2661       const MCExpr *Value;
2662       SMLoc ExprLoc = getLexer().getLoc();
2663       if (parseExpression(Value))
2664         return true;
2665 
2666       // Special case constant expressions to match code generator.
2667       if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2668         assert(Size <= 8 && "Invalid size");
2669         uint64_t IntValue = MCE->getValue();
2670         if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
2671           return Error(ExprLoc, "literal value out of range for directive");
2672         getStreamer().EmitIntValue(IntValue, Size);
2673       } else
2674         getStreamer().EmitValue(Value, Size, ExprLoc);
2675 
2676       if (getLexer().is(AsmToken::EndOfStatement))
2677         break;
2678 
2679       // FIXME: Improve diagnostic.
2680       if (getLexer().isNot(AsmToken::Comma))
2681         return TokError("unexpected token in directive");
2682       Lex();
2683     }
2684   }
2685 
2686   Lex();
2687   return false;
2688 }
2689 
2690 /// ParseDirectiveOctaValue
2691 ///  ::= .octa [ hexconstant (, hexconstant)* ]
parseDirectiveOctaValue()2692 bool AsmParser::parseDirectiveOctaValue() {
2693   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2694     checkForValidSection();
2695 
2696     for (;;) {
2697       if (Lexer.getKind() == AsmToken::Error)
2698         return true;
2699       if (Lexer.getKind() != AsmToken::Integer &&
2700           Lexer.getKind() != AsmToken::BigNum)
2701         return TokError("unknown token in expression");
2702 
2703       SMLoc ExprLoc = getLexer().getLoc();
2704       APInt IntValue = getTok().getAPIntVal();
2705       Lex();
2706 
2707       uint64_t hi, lo;
2708       if (IntValue.isIntN(64)) {
2709         hi = 0;
2710         lo = IntValue.getZExtValue();
2711       } else if (IntValue.isIntN(128)) {
2712         // It might actually have more than 128 bits, but the top ones are zero.
2713         hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue();
2714         lo = IntValue.getLoBits(64).getZExtValue();
2715       } else
2716         return Error(ExprLoc, "literal value out of range for directive");
2717 
2718       if (MAI.isLittleEndian()) {
2719         getStreamer().EmitIntValue(lo, 8);
2720         getStreamer().EmitIntValue(hi, 8);
2721       } else {
2722         getStreamer().EmitIntValue(hi, 8);
2723         getStreamer().EmitIntValue(lo, 8);
2724       }
2725 
2726       if (getLexer().is(AsmToken::EndOfStatement))
2727         break;
2728 
2729       // FIXME: Improve diagnostic.
2730       if (getLexer().isNot(AsmToken::Comma))
2731         return TokError("unexpected token in directive");
2732       Lex();
2733     }
2734   }
2735 
2736   Lex();
2737   return false;
2738 }
2739 
2740 /// parseDirectiveRealValue
2741 ///  ::= (.single | .double) [ expression (, expression)* ]
parseDirectiveRealValue(const fltSemantics & Semantics)2742 bool AsmParser::parseDirectiveRealValue(const fltSemantics &Semantics) {
2743   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2744     checkForValidSection();
2745 
2746     for (;;) {
2747       // We don't truly support arithmetic on floating point expressions, so we
2748       // have to manually parse unary prefixes.
2749       bool IsNeg = false;
2750       if (getLexer().is(AsmToken::Minus)) {
2751         Lexer.Lex();
2752         IsNeg = true;
2753       } else if (getLexer().is(AsmToken::Plus))
2754         Lexer.Lex();
2755 
2756       if (Lexer.is(AsmToken::Error))
2757         return TokError(Lexer.getErr());
2758       if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) &&
2759           Lexer.isNot(AsmToken::Identifier))
2760         return TokError("unexpected token in directive");
2761 
2762       // Convert to an APFloat.
2763       APFloat Value(Semantics);
2764       StringRef IDVal = getTok().getString();
2765       if (getLexer().is(AsmToken::Identifier)) {
2766         if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
2767           Value = APFloat::getInf(Semantics);
2768         else if (!IDVal.compare_lower("nan"))
2769           Value = APFloat::getNaN(Semantics, false, ~0);
2770         else
2771           return TokError("invalid floating point literal");
2772       } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
2773                  APFloat::opInvalidOp)
2774         return TokError("invalid floating point literal");
2775       if (IsNeg)
2776         Value.changeSign();
2777 
2778       // Consume the numeric token.
2779       Lex();
2780 
2781       // Emit the value as an integer.
2782       APInt AsInt = Value.bitcastToAPInt();
2783       getStreamer().EmitIntValue(AsInt.getLimitedValue(),
2784                                  AsInt.getBitWidth() / 8);
2785 
2786       if (Lexer.is(AsmToken::EndOfStatement))
2787         break;
2788 
2789       if (Lexer.isNot(AsmToken::Comma))
2790         return TokError("unexpected token in directive");
2791       Lex();
2792     }
2793   }
2794 
2795   Lex();
2796   return false;
2797 }
2798 
2799 /// parseDirectiveZero
2800 ///  ::= .zero expression
parseDirectiveZero()2801 bool AsmParser::parseDirectiveZero() {
2802   checkForValidSection();
2803 
2804   SMLoc NumBytesLoc = Lexer.getLoc();
2805   const MCExpr *NumBytes;
2806   if (parseExpression(NumBytes))
2807     return true;
2808 
2809   int64_t Val = 0;
2810   if (getLexer().is(AsmToken::Comma)) {
2811     Lex();
2812     if (parseAbsoluteExpression(Val))
2813       return true;
2814   }
2815 
2816   if (getLexer().isNot(AsmToken::EndOfStatement))
2817     return TokError("unexpected token in '.zero' directive");
2818 
2819   Lex();
2820 
2821   getStreamer().emitFill(*NumBytes, Val, NumBytesLoc);
2822 
2823   return false;
2824 }
2825 
2826 /// parseDirectiveFill
2827 ///  ::= .fill expression [ , expression [ , expression ] ]
parseDirectiveFill()2828 bool AsmParser::parseDirectiveFill() {
2829   checkForValidSection();
2830 
2831   SMLoc NumValuesLoc = Lexer.getLoc();
2832   const MCExpr *NumValues;
2833   if (parseExpression(NumValues))
2834     return true;
2835 
2836   int64_t FillSize = 1;
2837   int64_t FillExpr = 0;
2838 
2839   SMLoc SizeLoc, ExprLoc;
2840   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2841     if (getLexer().isNot(AsmToken::Comma))
2842       return TokError("unexpected token in '.fill' directive");
2843     Lex();
2844 
2845     SizeLoc = getLexer().getLoc();
2846     if (parseAbsoluteExpression(FillSize))
2847       return true;
2848 
2849     if (getLexer().isNot(AsmToken::EndOfStatement)) {
2850       if (getLexer().isNot(AsmToken::Comma))
2851         return TokError("unexpected token in '.fill' directive");
2852       Lex();
2853 
2854       ExprLoc = getLexer().getLoc();
2855       if (parseAbsoluteExpression(FillExpr))
2856         return true;
2857 
2858       if (getLexer().isNot(AsmToken::EndOfStatement))
2859         return TokError("unexpected token in '.fill' directive");
2860 
2861       Lex();
2862     }
2863   }
2864 
2865   if (FillSize < 0) {
2866     Warning(SizeLoc, "'.fill' directive with negative size has no effect");
2867     return false;
2868   }
2869   if (FillSize > 8) {
2870     Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8");
2871     FillSize = 8;
2872   }
2873 
2874   if (!isUInt<32>(FillExpr) && FillSize > 4)
2875     Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits");
2876 
2877   getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc);
2878 
2879   return false;
2880 }
2881 
2882 /// parseDirectiveOrg
2883 ///  ::= .org expression [ , expression ]
parseDirectiveOrg()2884 bool AsmParser::parseDirectiveOrg() {
2885   checkForValidSection();
2886 
2887   const MCExpr *Offset;
2888   if (parseExpression(Offset))
2889     return true;
2890 
2891   // Parse optional fill expression.
2892   int64_t FillExpr = 0;
2893   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2894     if (getLexer().isNot(AsmToken::Comma))
2895       return TokError("unexpected token in '.org' directive");
2896     Lex();
2897 
2898     if (parseAbsoluteExpression(FillExpr))
2899       return true;
2900 
2901     if (getLexer().isNot(AsmToken::EndOfStatement))
2902       return TokError("unexpected token in '.org' directive");
2903   }
2904 
2905   Lex();
2906   getStreamer().emitValueToOffset(Offset, FillExpr);
2907   return false;
2908 }
2909 
2910 /// parseDirectiveAlign
2911 ///  ::= {.align, ...} expression [ , expression [ , expression ]]
parseDirectiveAlign(bool IsPow2,unsigned ValueSize)2912 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
2913   checkForValidSection();
2914 
2915   SMLoc AlignmentLoc = getLexer().getLoc();
2916   int64_t Alignment;
2917   if (parseAbsoluteExpression(Alignment))
2918     return true;
2919 
2920   SMLoc MaxBytesLoc;
2921   bool HasFillExpr = false;
2922   int64_t FillExpr = 0;
2923   int64_t MaxBytesToFill = 0;
2924   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2925     if (getLexer().isNot(AsmToken::Comma))
2926       return TokError("unexpected token in directive");
2927     Lex();
2928 
2929     // The fill expression can be omitted while specifying a maximum number of
2930     // alignment bytes, e.g:
2931     //  .align 3,,4
2932     if (getLexer().isNot(AsmToken::Comma)) {
2933       HasFillExpr = true;
2934       if (parseAbsoluteExpression(FillExpr))
2935         return true;
2936     }
2937 
2938     if (getLexer().isNot(AsmToken::EndOfStatement)) {
2939       if (getLexer().isNot(AsmToken::Comma))
2940         return TokError("unexpected token in directive");
2941       Lex();
2942 
2943       MaxBytesLoc = getLexer().getLoc();
2944       if (parseAbsoluteExpression(MaxBytesToFill))
2945         return true;
2946 
2947       if (getLexer().isNot(AsmToken::EndOfStatement))
2948         return TokError("unexpected token in directive");
2949     }
2950   }
2951 
2952   Lex();
2953 
2954   if (!HasFillExpr)
2955     FillExpr = 0;
2956 
2957   // Compute alignment in bytes.
2958   if (IsPow2) {
2959     // FIXME: Diagnose overflow.
2960     if (Alignment >= 32) {
2961       Error(AlignmentLoc, "invalid alignment value");
2962       Alignment = 31;
2963     }
2964 
2965     Alignment = 1ULL << Alignment;
2966   } else {
2967     // Reject alignments that aren't either a power of two or zero,
2968     // for gas compatibility. Alignment of zero is silently rounded
2969     // up to one.
2970     if (Alignment == 0)
2971       Alignment = 1;
2972     if (!isPowerOf2_64(Alignment))
2973       Error(AlignmentLoc, "alignment must be a power of 2");
2974   }
2975 
2976   // Diagnose non-sensical max bytes to align.
2977   if (MaxBytesLoc.isValid()) {
2978     if (MaxBytesToFill < 1) {
2979       Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
2980                          "many bytes, ignoring maximum bytes expression");
2981       MaxBytesToFill = 0;
2982     }
2983 
2984     if (MaxBytesToFill >= Alignment) {
2985       Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2986                            "has no effect");
2987       MaxBytesToFill = 0;
2988     }
2989   }
2990 
2991   // Check whether we should use optimal code alignment for this .align
2992   // directive.
2993   const MCSection *Section = getStreamer().getCurrentSection().first;
2994   assert(Section && "must have section to emit alignment");
2995   bool UseCodeAlign = Section->UseCodeAlign();
2996   if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2997       ValueSize == 1 && UseCodeAlign) {
2998     getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2999   } else {
3000     // FIXME: Target specific behavior about how the "extra" bytes are filled.
3001     getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
3002                                        MaxBytesToFill);
3003   }
3004 
3005   return false;
3006 }
3007 
3008 /// parseDirectiveFile
3009 /// ::= .file [number] filename
3010 /// ::= .file number directory filename
parseDirectiveFile(SMLoc DirectiveLoc)3011 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) {
3012   // FIXME: I'm not sure what this is.
3013   int64_t FileNumber = -1;
3014   SMLoc FileNumberLoc = getLexer().getLoc();
3015   if (getLexer().is(AsmToken::Integer)) {
3016     FileNumber = getTok().getIntVal();
3017     Lex();
3018 
3019     if (FileNumber < 1)
3020       return TokError("file number less than one");
3021   }
3022 
3023   if (getLexer().isNot(AsmToken::String))
3024     return TokError("unexpected token in '.file' directive");
3025 
3026   // Usually the directory and filename together, otherwise just the directory.
3027   // Allow the strings to have escaped octal character sequence.
3028   std::string Path = getTok().getString();
3029   if (parseEscapedString(Path))
3030     return true;
3031   Lex();
3032 
3033   StringRef Directory;
3034   StringRef Filename;
3035   std::string FilenameData;
3036   if (getLexer().is(AsmToken::String)) {
3037     if (FileNumber == -1)
3038       return TokError("explicit path specified, but no file number");
3039     if (parseEscapedString(FilenameData))
3040       return true;
3041     Filename = FilenameData;
3042     Directory = Path;
3043     Lex();
3044   } else {
3045     Filename = Path;
3046   }
3047 
3048   if (getLexer().isNot(AsmToken::EndOfStatement))
3049     return TokError("unexpected token in '.file' directive");
3050 
3051   if (FileNumber == -1)
3052     getStreamer().EmitFileDirective(Filename);
3053   else {
3054     // If there is -g option as well as debug info from directive file,
3055     // we turn off -g option, directly use the existing debug info instead.
3056     if (getContext().getGenDwarfForAssembly())
3057       getContext().setGenDwarfForAssembly(false);
3058     else if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename) ==
3059         0)
3060       Error(FileNumberLoc, "file number already allocated");
3061   }
3062 
3063   return false;
3064 }
3065 
3066 /// parseDirectiveLine
3067 /// ::= .line [number]
parseDirectiveLine()3068 bool AsmParser::parseDirectiveLine() {
3069   if (getLexer().isNot(AsmToken::EndOfStatement)) {
3070     if (getLexer().isNot(AsmToken::Integer))
3071       return TokError("unexpected token in '.line' directive");
3072 
3073     int64_t LineNumber = getTok().getIntVal();
3074     (void)LineNumber;
3075     Lex();
3076 
3077     // FIXME: Do something with the .line.
3078   }
3079 
3080   if (getLexer().isNot(AsmToken::EndOfStatement))
3081     return TokError("unexpected token in '.line' directive");
3082 
3083   return false;
3084 }
3085 
3086 /// parseDirectiveLoc
3087 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
3088 ///                                [epilogue_begin] [is_stmt VALUE] [isa VALUE]
3089 /// The first number is a file number, must have been previously assigned with
3090 /// a .file directive, the second number is the line number and optionally the
3091 /// third number is a column position (zero if not specified).  The remaining
3092 /// optional items are .loc sub-directives.
parseDirectiveLoc()3093 bool AsmParser::parseDirectiveLoc() {
3094   if (getLexer().isNot(AsmToken::Integer))
3095     return TokError("unexpected token in '.loc' directive");
3096   int64_t FileNumber = getTok().getIntVal();
3097   if (FileNumber < 1)
3098     return TokError("file number less than one in '.loc' directive");
3099   if (!getContext().isValidDwarfFileNumber(FileNumber))
3100     return TokError("unassigned file number in '.loc' directive");
3101   Lex();
3102 
3103   int64_t LineNumber = 0;
3104   if (getLexer().is(AsmToken::Integer)) {
3105     LineNumber = getTok().getIntVal();
3106     if (LineNumber < 0)
3107       return TokError("line number less than zero in '.loc' directive");
3108     Lex();
3109   }
3110 
3111   int64_t ColumnPos = 0;
3112   if (getLexer().is(AsmToken::Integer)) {
3113     ColumnPos = getTok().getIntVal();
3114     if (ColumnPos < 0)
3115       return TokError("column position less than zero in '.loc' directive");
3116     Lex();
3117   }
3118 
3119   unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
3120   unsigned Isa = 0;
3121   int64_t Discriminator = 0;
3122   if (getLexer().isNot(AsmToken::EndOfStatement)) {
3123     for (;;) {
3124       if (getLexer().is(AsmToken::EndOfStatement))
3125         break;
3126 
3127       StringRef Name;
3128       SMLoc Loc = getTok().getLoc();
3129       if (parseIdentifier(Name))
3130         return TokError("unexpected token in '.loc' directive");
3131 
3132       if (Name == "basic_block")
3133         Flags |= DWARF2_FLAG_BASIC_BLOCK;
3134       else if (Name == "prologue_end")
3135         Flags |= DWARF2_FLAG_PROLOGUE_END;
3136       else if (Name == "epilogue_begin")
3137         Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
3138       else if (Name == "is_stmt") {
3139         Loc = getTok().getLoc();
3140         const MCExpr *Value;
3141         if (parseExpression(Value))
3142           return true;
3143         // The expression must be the constant 0 or 1.
3144         if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3145           int Value = MCE->getValue();
3146           if (Value == 0)
3147             Flags &= ~DWARF2_FLAG_IS_STMT;
3148           else if (Value == 1)
3149             Flags |= DWARF2_FLAG_IS_STMT;
3150           else
3151             return Error(Loc, "is_stmt value not 0 or 1");
3152         } else {
3153           return Error(Loc, "is_stmt value not the constant value of 0 or 1");
3154         }
3155       } else if (Name == "isa") {
3156         Loc = getTok().getLoc();
3157         const MCExpr *Value;
3158         if (parseExpression(Value))
3159           return true;
3160         // The expression must be a constant greater or equal to 0.
3161         if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3162           int Value = MCE->getValue();
3163           if (Value < 0)
3164             return Error(Loc, "isa number less than zero");
3165           Isa = Value;
3166         } else {
3167           return Error(Loc, "isa number not a constant value");
3168         }
3169       } else if (Name == "discriminator") {
3170         if (parseAbsoluteExpression(Discriminator))
3171           return true;
3172       } else {
3173         return Error(Loc, "unknown sub-directive in '.loc' directive");
3174       }
3175 
3176       if (getLexer().is(AsmToken::EndOfStatement))
3177         break;
3178     }
3179   }
3180 
3181   getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
3182                                       Isa, Discriminator, StringRef());
3183 
3184   return false;
3185 }
3186 
3187 /// parseDirectiveStabs
3188 /// ::= .stabs string, number, number, number
parseDirectiveStabs()3189 bool AsmParser::parseDirectiveStabs() {
3190   return TokError("unsupported directive '.stabs'");
3191 }
3192 
3193 /// parseDirectiveCVFile
3194 /// ::= .cv_file number filename
parseDirectiveCVFile()3195 bool AsmParser::parseDirectiveCVFile() {
3196   SMLoc FileNumberLoc = getLexer().getLoc();
3197   if (getLexer().isNot(AsmToken::Integer))
3198     return TokError("expected file number in '.cv_file' directive");
3199 
3200   int64_t FileNumber = getTok().getIntVal();
3201   Lex();
3202 
3203   if (FileNumber < 1)
3204     return TokError("file number less than one");
3205 
3206   if (getLexer().isNot(AsmToken::String))
3207     return TokError("unexpected token in '.cv_file' directive");
3208 
3209   // Usually the directory and filename together, otherwise just the directory.
3210   // Allow the strings to have escaped octal character sequence.
3211   std::string Filename;
3212   if (parseEscapedString(Filename))
3213     return true;
3214   Lex();
3215 
3216   if (getLexer().isNot(AsmToken::EndOfStatement))
3217     return TokError("unexpected token in '.cv_file' directive");
3218 
3219   if (getStreamer().EmitCVFileDirective(FileNumber, Filename) == 0)
3220     Error(FileNumberLoc, "file number already allocated");
3221 
3222   return false;
3223 }
3224 
3225 /// parseDirectiveCVLoc
3226 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end]
3227 ///                                [is_stmt VALUE]
3228 /// The first number is a file number, must have been previously assigned with
3229 /// a .file directive, the second number is the line number and optionally the
3230 /// third number is a column position (zero if not specified).  The remaining
3231 /// optional items are .loc sub-directives.
parseDirectiveCVLoc()3232 bool AsmParser::parseDirectiveCVLoc() {
3233   if (getLexer().isNot(AsmToken::Integer))
3234     return TokError("unexpected token in '.cv_loc' directive");
3235 
3236   int64_t FunctionId = getTok().getIntVal();
3237   if (FunctionId < 0)
3238     return TokError("function id less than zero in '.cv_loc' directive");
3239   Lex();
3240 
3241   int64_t FileNumber = getTok().getIntVal();
3242   if (FileNumber < 1)
3243     return TokError("file number less than one in '.cv_loc' directive");
3244   if (!getContext().isValidCVFileNumber(FileNumber))
3245     return TokError("unassigned file number in '.cv_loc' directive");
3246   Lex();
3247 
3248   int64_t LineNumber = 0;
3249   if (getLexer().is(AsmToken::Integer)) {
3250     LineNumber = getTok().getIntVal();
3251     if (LineNumber < 0)
3252       return TokError("line number less than zero in '.cv_loc' directive");
3253     Lex();
3254   }
3255 
3256   int64_t ColumnPos = 0;
3257   if (getLexer().is(AsmToken::Integer)) {
3258     ColumnPos = getTok().getIntVal();
3259     if (ColumnPos < 0)
3260       return TokError("column position less than zero in '.cv_loc' directive");
3261     Lex();
3262   }
3263 
3264   bool PrologueEnd = false;
3265   uint64_t IsStmt = 0;
3266   while (getLexer().isNot(AsmToken::EndOfStatement)) {
3267     StringRef Name;
3268     SMLoc Loc = getTok().getLoc();
3269     if (parseIdentifier(Name))
3270       return TokError("unexpected token in '.cv_loc' directive");
3271 
3272     if (Name == "prologue_end")
3273       PrologueEnd = true;
3274     else if (Name == "is_stmt") {
3275       Loc = getTok().getLoc();
3276       const MCExpr *Value;
3277       if (parseExpression(Value))
3278         return true;
3279       // The expression must be the constant 0 or 1.
3280       IsStmt = ~0ULL;
3281       if (const auto *MCE = dyn_cast<MCConstantExpr>(Value))
3282         IsStmt = MCE->getValue();
3283 
3284       if (IsStmt > 1)
3285         return Error(Loc, "is_stmt value not 0 or 1");
3286     } else {
3287       return Error(Loc, "unknown sub-directive in '.cv_loc' directive");
3288     }
3289   }
3290 
3291   getStreamer().EmitCVLocDirective(FunctionId, FileNumber, LineNumber,
3292                                    ColumnPos, PrologueEnd, IsStmt, StringRef());
3293   return false;
3294 }
3295 
3296 /// parseDirectiveCVLinetable
3297 /// ::= .cv_linetable FunctionId, FnStart, FnEnd
parseDirectiveCVLinetable()3298 bool AsmParser::parseDirectiveCVLinetable() {
3299   int64_t FunctionId = getTok().getIntVal();
3300   if (FunctionId < 0)
3301     return TokError("function id less than zero in '.cv_linetable' directive");
3302   Lex();
3303 
3304   if (Lexer.isNot(AsmToken::Comma))
3305     return TokError("unexpected token in '.cv_linetable' directive");
3306   Lex();
3307 
3308   SMLoc Loc = getLexer().getLoc();
3309   StringRef FnStartName;
3310   if (parseIdentifier(FnStartName))
3311     return Error(Loc, "expected identifier in directive");
3312 
3313   if (Lexer.isNot(AsmToken::Comma))
3314     return TokError("unexpected token in '.cv_linetable' directive");
3315   Lex();
3316 
3317   Loc = getLexer().getLoc();
3318   StringRef FnEndName;
3319   if (parseIdentifier(FnEndName))
3320     return Error(Loc, "expected identifier in directive");
3321 
3322   MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
3323   MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
3324 
3325   getStreamer().EmitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym);
3326   return false;
3327 }
3328 
3329 /// parseDirectiveCVInlineLinetable
3330 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd
3331 ///          ("contains" SecondaryFunctionId+)?
parseDirectiveCVInlineLinetable()3332 bool AsmParser::parseDirectiveCVInlineLinetable() {
3333   int64_t PrimaryFunctionId = getTok().getIntVal();
3334   if (PrimaryFunctionId < 0)
3335     return TokError(
3336         "function id less than zero in '.cv_inline_linetable' directive");
3337   Lex();
3338 
3339   int64_t SourceFileId = getTok().getIntVal();
3340   if (SourceFileId <= 0)
3341     return TokError(
3342         "File id less than zero in '.cv_inline_linetable' directive");
3343   Lex();
3344 
3345   int64_t SourceLineNum = getTok().getIntVal();
3346   if (SourceLineNum < 0)
3347     return TokError(
3348         "Line number less than zero in '.cv_inline_linetable' directive");
3349   Lex();
3350 
3351   SMLoc Loc = getLexer().getLoc();
3352   StringRef FnStartName;
3353   if (parseIdentifier(FnStartName))
3354     return Error(Loc, "expected identifier in directive");
3355   MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
3356 
3357   Loc = getLexer().getLoc();
3358   StringRef FnEndName;
3359   if (parseIdentifier(FnEndName))
3360     return Error(Loc, "expected identifier in directive");
3361   MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
3362 
3363   SmallVector<unsigned, 8> SecondaryFunctionIds;
3364   if (getLexer().is(AsmToken::Identifier)) {
3365     if (getTok().getIdentifier() != "contains")
3366       return TokError(
3367           "unexpected identifier in '.cv_inline_linetable' directive");
3368     Lex();
3369 
3370     while (getLexer().isNot(AsmToken::EndOfStatement)) {
3371       int64_t SecondaryFunctionId = getTok().getIntVal();
3372       if (SecondaryFunctionId < 0)
3373         return TokError(
3374             "function id less than zero in '.cv_inline_linetable' directive");
3375       Lex();
3376 
3377       SecondaryFunctionIds.push_back(SecondaryFunctionId);
3378     }
3379   }
3380 
3381   getStreamer().EmitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId,
3382                                                SourceLineNum, FnStartSym,
3383                                                FnEndSym, SecondaryFunctionIds);
3384   return false;
3385 }
3386 
3387 /// parseDirectiveCVDefRange
3388 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes*
parseDirectiveCVDefRange()3389 bool AsmParser::parseDirectiveCVDefRange() {
3390   SMLoc Loc;
3391   std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges;
3392   while (getLexer().is(AsmToken::Identifier)) {
3393     Loc = getLexer().getLoc();
3394     StringRef GapStartName;
3395     if (parseIdentifier(GapStartName))
3396       return Error(Loc, "expected identifier in directive");
3397     MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName);
3398 
3399     Loc = getLexer().getLoc();
3400     StringRef GapEndName;
3401     if (parseIdentifier(GapEndName))
3402       return Error(Loc, "expected identifier in directive");
3403     MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName);
3404 
3405     Ranges.push_back({GapStartSym, GapEndSym});
3406   }
3407 
3408   if (getLexer().isNot(AsmToken::Comma))
3409     return TokError("unexpected token in directive");
3410   Lex();
3411 
3412   std::string FixedSizePortion;
3413   if (parseEscapedString(FixedSizePortion))
3414     return true;
3415   Lex();
3416 
3417   getStreamer().EmitCVDefRangeDirective(Ranges, FixedSizePortion);
3418   return false;
3419 }
3420 
3421 /// parseDirectiveCVStringTable
3422 /// ::= .cv_stringtable
parseDirectiveCVStringTable()3423 bool AsmParser::parseDirectiveCVStringTable() {
3424   getStreamer().EmitCVStringTableDirective();
3425   return false;
3426 }
3427 
3428 /// parseDirectiveCVFileChecksums
3429 /// ::= .cv_filechecksums
parseDirectiveCVFileChecksums()3430 bool AsmParser::parseDirectiveCVFileChecksums() {
3431   getStreamer().EmitCVFileChecksumsDirective();
3432   return false;
3433 }
3434 
3435 /// parseDirectiveCFISections
3436 /// ::= .cfi_sections section [, section]
parseDirectiveCFISections()3437 bool AsmParser::parseDirectiveCFISections() {
3438   StringRef Name;
3439   bool EH = false;
3440   bool Debug = false;
3441 
3442   if (parseIdentifier(Name))
3443     return TokError("Expected an identifier");
3444 
3445   if (Name == ".eh_frame")
3446     EH = true;
3447   else if (Name == ".debug_frame")
3448     Debug = true;
3449 
3450   if (getLexer().is(AsmToken::Comma)) {
3451     Lex();
3452 
3453     if (parseIdentifier(Name))
3454       return TokError("Expected an identifier");
3455 
3456     if (Name == ".eh_frame")
3457       EH = true;
3458     else if (Name == ".debug_frame")
3459       Debug = true;
3460   }
3461 
3462   getStreamer().EmitCFISections(EH, Debug);
3463   return false;
3464 }
3465 
3466 /// parseDirectiveCFIStartProc
3467 /// ::= .cfi_startproc [simple]
parseDirectiveCFIStartProc()3468 bool AsmParser::parseDirectiveCFIStartProc() {
3469   StringRef Simple;
3470   if (getLexer().isNot(AsmToken::EndOfStatement))
3471     if (parseIdentifier(Simple) || Simple != "simple")
3472       return TokError("unexpected token in .cfi_startproc directive");
3473 
3474   getStreamer().EmitCFIStartProc(!Simple.empty());
3475   return false;
3476 }
3477 
3478 /// parseDirectiveCFIEndProc
3479 /// ::= .cfi_endproc
parseDirectiveCFIEndProc()3480 bool AsmParser::parseDirectiveCFIEndProc() {
3481   getStreamer().EmitCFIEndProc();
3482   return false;
3483 }
3484 
3485 /// \brief parse register name or number.
parseRegisterOrRegisterNumber(int64_t & Register,SMLoc DirectiveLoc)3486 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register,
3487                                               SMLoc DirectiveLoc) {
3488   unsigned RegNo;
3489 
3490   if (getLexer().isNot(AsmToken::Integer)) {
3491     if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc))
3492       return true;
3493     Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true);
3494   } else
3495     return parseAbsoluteExpression(Register);
3496 
3497   return false;
3498 }
3499 
3500 /// parseDirectiveCFIDefCfa
3501 /// ::= .cfi_def_cfa register,  offset
parseDirectiveCFIDefCfa(SMLoc DirectiveLoc)3502 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) {
3503   int64_t Register = 0;
3504   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3505     return true;
3506 
3507   if (getLexer().isNot(AsmToken::Comma))
3508     return TokError("unexpected token in directive");
3509   Lex();
3510 
3511   int64_t Offset = 0;
3512   if (parseAbsoluteExpression(Offset))
3513     return true;
3514 
3515   getStreamer().EmitCFIDefCfa(Register, Offset);
3516   return false;
3517 }
3518 
3519 /// parseDirectiveCFIDefCfaOffset
3520 /// ::= .cfi_def_cfa_offset offset
parseDirectiveCFIDefCfaOffset()3521 bool AsmParser::parseDirectiveCFIDefCfaOffset() {
3522   int64_t Offset = 0;
3523   if (parseAbsoluteExpression(Offset))
3524     return true;
3525 
3526   getStreamer().EmitCFIDefCfaOffset(Offset);
3527   return false;
3528 }
3529 
3530 /// parseDirectiveCFIRegister
3531 /// ::= .cfi_register register, register
parseDirectiveCFIRegister(SMLoc DirectiveLoc)3532 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) {
3533   int64_t Register1 = 0;
3534   if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc))
3535     return true;
3536 
3537   if (getLexer().isNot(AsmToken::Comma))
3538     return TokError("unexpected token in directive");
3539   Lex();
3540 
3541   int64_t Register2 = 0;
3542   if (parseRegisterOrRegisterNumber(Register2, DirectiveLoc))
3543     return true;
3544 
3545   getStreamer().EmitCFIRegister(Register1, Register2);
3546   return false;
3547 }
3548 
3549 /// parseDirectiveCFIWindowSave
3550 /// ::= .cfi_window_save
parseDirectiveCFIWindowSave()3551 bool AsmParser::parseDirectiveCFIWindowSave() {
3552   getStreamer().EmitCFIWindowSave();
3553   return false;
3554 }
3555 
3556 /// parseDirectiveCFIAdjustCfaOffset
3557 /// ::= .cfi_adjust_cfa_offset adjustment
parseDirectiveCFIAdjustCfaOffset()3558 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() {
3559   int64_t Adjustment = 0;
3560   if (parseAbsoluteExpression(Adjustment))
3561     return true;
3562 
3563   getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
3564   return false;
3565 }
3566 
3567 /// parseDirectiveCFIDefCfaRegister
3568 /// ::= .cfi_def_cfa_register register
parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc)3569 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) {
3570   int64_t Register = 0;
3571   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3572     return true;
3573 
3574   getStreamer().EmitCFIDefCfaRegister(Register);
3575   return false;
3576 }
3577 
3578 /// parseDirectiveCFIOffset
3579 /// ::= .cfi_offset register, offset
parseDirectiveCFIOffset(SMLoc DirectiveLoc)3580 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) {
3581   int64_t Register = 0;
3582   int64_t Offset = 0;
3583 
3584   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3585     return true;
3586 
3587   if (getLexer().isNot(AsmToken::Comma))
3588     return TokError("unexpected token in directive");
3589   Lex();
3590 
3591   if (parseAbsoluteExpression(Offset))
3592     return true;
3593 
3594   getStreamer().EmitCFIOffset(Register, Offset);
3595   return false;
3596 }
3597 
3598 /// parseDirectiveCFIRelOffset
3599 /// ::= .cfi_rel_offset register, offset
parseDirectiveCFIRelOffset(SMLoc DirectiveLoc)3600 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) {
3601   int64_t Register = 0;
3602 
3603   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3604     return true;
3605 
3606   if (getLexer().isNot(AsmToken::Comma))
3607     return TokError("unexpected token in directive");
3608   Lex();
3609 
3610   int64_t Offset = 0;
3611   if (parseAbsoluteExpression(Offset))
3612     return true;
3613 
3614   getStreamer().EmitCFIRelOffset(Register, Offset);
3615   return false;
3616 }
3617 
isValidEncoding(int64_t Encoding)3618 static bool isValidEncoding(int64_t Encoding) {
3619   if (Encoding & ~0xff)
3620     return false;
3621 
3622   if (Encoding == dwarf::DW_EH_PE_omit)
3623     return true;
3624 
3625   const unsigned Format = Encoding & 0xf;
3626   if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
3627       Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
3628       Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
3629       Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
3630     return false;
3631 
3632   const unsigned Application = Encoding & 0x70;
3633   if (Application != dwarf::DW_EH_PE_absptr &&
3634       Application != dwarf::DW_EH_PE_pcrel)
3635     return false;
3636 
3637   return true;
3638 }
3639 
3640 /// parseDirectiveCFIPersonalityOrLsda
3641 /// IsPersonality true for cfi_personality, false for cfi_lsda
3642 /// ::= .cfi_personality encoding, [symbol_name]
3643 /// ::= .cfi_lsda encoding, [symbol_name]
parseDirectiveCFIPersonalityOrLsda(bool IsPersonality)3644 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) {
3645   int64_t Encoding = 0;
3646   if (parseAbsoluteExpression(Encoding))
3647     return true;
3648   if (Encoding == dwarf::DW_EH_PE_omit)
3649     return false;
3650 
3651   if (!isValidEncoding(Encoding))
3652     return TokError("unsupported encoding.");
3653 
3654   if (getLexer().isNot(AsmToken::Comma))
3655     return TokError("unexpected token in directive");
3656   Lex();
3657 
3658   StringRef Name;
3659   if (parseIdentifier(Name))
3660     return TokError("expected identifier in directive");
3661 
3662   MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
3663 
3664   if (IsPersonality)
3665     getStreamer().EmitCFIPersonality(Sym, Encoding);
3666   else
3667     getStreamer().EmitCFILsda(Sym, Encoding);
3668   return false;
3669 }
3670 
3671 /// parseDirectiveCFIRememberState
3672 /// ::= .cfi_remember_state
parseDirectiveCFIRememberState()3673 bool AsmParser::parseDirectiveCFIRememberState() {
3674   getStreamer().EmitCFIRememberState();
3675   return false;
3676 }
3677 
3678 /// parseDirectiveCFIRestoreState
3679 /// ::= .cfi_remember_state
parseDirectiveCFIRestoreState()3680 bool AsmParser::parseDirectiveCFIRestoreState() {
3681   getStreamer().EmitCFIRestoreState();
3682   return false;
3683 }
3684 
3685 /// parseDirectiveCFISameValue
3686 /// ::= .cfi_same_value register
parseDirectiveCFISameValue(SMLoc DirectiveLoc)3687 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) {
3688   int64_t Register = 0;
3689 
3690   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3691     return true;
3692 
3693   getStreamer().EmitCFISameValue(Register);
3694   return false;
3695 }
3696 
3697 /// parseDirectiveCFIRestore
3698 /// ::= .cfi_restore register
parseDirectiveCFIRestore(SMLoc DirectiveLoc)3699 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) {
3700   int64_t Register = 0;
3701   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3702     return true;
3703 
3704   getStreamer().EmitCFIRestore(Register);
3705   return false;
3706 }
3707 
3708 /// parseDirectiveCFIEscape
3709 /// ::= .cfi_escape expression[,...]
parseDirectiveCFIEscape()3710 bool AsmParser::parseDirectiveCFIEscape() {
3711   std::string Values;
3712   int64_t CurrValue;
3713   if (parseAbsoluteExpression(CurrValue))
3714     return true;
3715 
3716   Values.push_back((uint8_t)CurrValue);
3717 
3718   while (getLexer().is(AsmToken::Comma)) {
3719     Lex();
3720 
3721     if (parseAbsoluteExpression(CurrValue))
3722       return true;
3723 
3724     Values.push_back((uint8_t)CurrValue);
3725   }
3726 
3727   getStreamer().EmitCFIEscape(Values);
3728   return false;
3729 }
3730 
3731 /// parseDirectiveCFISignalFrame
3732 /// ::= .cfi_signal_frame
parseDirectiveCFISignalFrame()3733 bool AsmParser::parseDirectiveCFISignalFrame() {
3734   if (getLexer().isNot(AsmToken::EndOfStatement))
3735     return Error(getLexer().getLoc(),
3736                  "unexpected token in '.cfi_signal_frame'");
3737 
3738   getStreamer().EmitCFISignalFrame();
3739   return false;
3740 }
3741 
3742 /// parseDirectiveCFIUndefined
3743 /// ::= .cfi_undefined register
parseDirectiveCFIUndefined(SMLoc DirectiveLoc)3744 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) {
3745   int64_t Register = 0;
3746 
3747   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3748     return true;
3749 
3750   getStreamer().EmitCFIUndefined(Register);
3751   return false;
3752 }
3753 
3754 /// parseDirectiveMacrosOnOff
3755 /// ::= .macros_on
3756 /// ::= .macros_off
parseDirectiveMacrosOnOff(StringRef Directive)3757 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) {
3758   if (getLexer().isNot(AsmToken::EndOfStatement))
3759     return Error(getLexer().getLoc(),
3760                  "unexpected token in '" + Directive + "' directive");
3761 
3762   setMacrosEnabled(Directive == ".macros_on");
3763   return false;
3764 }
3765 
3766 /// parseDirectiveMacro
3767 /// ::= .macro name[,] [parameters]
parseDirectiveMacro(SMLoc DirectiveLoc)3768 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) {
3769   StringRef Name;
3770   if (parseIdentifier(Name))
3771     return TokError("expected identifier in '.macro' directive");
3772 
3773   if (getLexer().is(AsmToken::Comma))
3774     Lex();
3775 
3776   MCAsmMacroParameters Parameters;
3777   while (getLexer().isNot(AsmToken::EndOfStatement)) {
3778 
3779     if (!Parameters.empty() && Parameters.back().Vararg)
3780       return Error(Lexer.getLoc(),
3781                    "Vararg parameter '" + Parameters.back().Name +
3782                    "' should be last one in the list of parameters.");
3783 
3784     MCAsmMacroParameter Parameter;
3785     if (parseIdentifier(Parameter.Name))
3786       return TokError("expected identifier in '.macro' directive");
3787 
3788     if (Lexer.is(AsmToken::Colon)) {
3789       Lex();  // consume ':'
3790 
3791       SMLoc QualLoc;
3792       StringRef Qualifier;
3793 
3794       QualLoc = Lexer.getLoc();
3795       if (parseIdentifier(Qualifier))
3796         return Error(QualLoc, "missing parameter qualifier for "
3797                      "'" + Parameter.Name + "' in macro '" + Name + "'");
3798 
3799       if (Qualifier == "req")
3800         Parameter.Required = true;
3801       else if (Qualifier == "vararg")
3802         Parameter.Vararg = true;
3803       else
3804         return Error(QualLoc, Qualifier + " is not a valid parameter qualifier "
3805                      "for '" + Parameter.Name + "' in macro '" + Name + "'");
3806     }
3807 
3808     if (getLexer().is(AsmToken::Equal)) {
3809       Lex();
3810 
3811       SMLoc ParamLoc;
3812 
3813       ParamLoc = Lexer.getLoc();
3814       if (parseMacroArgument(Parameter.Value, /*Vararg=*/false ))
3815         return true;
3816 
3817       if (Parameter.Required)
3818         Warning(ParamLoc, "pointless default value for required parameter "
3819                 "'" + Parameter.Name + "' in macro '" + Name + "'");
3820     }
3821 
3822     Parameters.push_back(std::move(Parameter));
3823 
3824     if (getLexer().is(AsmToken::Comma))
3825       Lex();
3826   }
3827 
3828   // Eat just the end of statement.
3829   Lexer.Lex();
3830 
3831   // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors
3832   AsmToken EndToken, StartToken = getTok();
3833   unsigned MacroDepth = 0;
3834   // Lex the macro definition.
3835   for (;;) {
3836     // Ignore Lexing errors in macros.
3837     while (Lexer.is(AsmToken::Error)) {
3838       Lexer.Lex();
3839     }
3840 
3841     // Check whether we have reached the end of the file.
3842     if (getLexer().is(AsmToken::Eof))
3843       return Error(DirectiveLoc, "no matching '.endmacro' in definition");
3844 
3845     // Otherwise, check whether we have reach the .endmacro.
3846     if (getLexer().is(AsmToken::Identifier)) {
3847       if (getTok().getIdentifier() == ".endm" ||
3848           getTok().getIdentifier() == ".endmacro") {
3849         if (MacroDepth == 0) { // Outermost macro.
3850           EndToken = getTok();
3851           Lexer.Lex();
3852           if (getLexer().isNot(AsmToken::EndOfStatement))
3853             return TokError("unexpected token in '" + EndToken.getIdentifier() +
3854                             "' directive");
3855           break;
3856         } else {
3857           // Otherwise we just found the end of an inner macro.
3858           --MacroDepth;
3859         }
3860       } else if (getTok().getIdentifier() == ".macro") {
3861         // We allow nested macros. Those aren't instantiated until the outermost
3862         // macro is expanded so just ignore them for now.
3863         ++MacroDepth;
3864       }
3865     }
3866 
3867     // Otherwise, scan til the end of the statement.
3868     eatToEndOfStatement();
3869   }
3870 
3871   if (lookupMacro(Name)) {
3872     return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
3873   }
3874 
3875   const char *BodyStart = StartToken.getLoc().getPointer();
3876   const char *BodyEnd = EndToken.getLoc().getPointer();
3877   StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3878   checkForBadMacro(DirectiveLoc, Name, Body, Parameters);
3879   defineMacro(Name, MCAsmMacro(Name, Body, std::move(Parameters)));
3880   return false;
3881 }
3882 
3883 /// checkForBadMacro
3884 ///
3885 /// With the support added for named parameters there may be code out there that
3886 /// is transitioning from positional parameters.  In versions of gas that did
3887 /// not support named parameters they would be ignored on the macro definition.
3888 /// But to support both styles of parameters this is not possible so if a macro
3889 /// definition has named parameters but does not use them and has what appears
3890 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a
3891 /// warning that the positional parameter found in body which have no effect.
3892 /// Hoping the developer will either remove the named parameters from the macro
3893 /// definition so the positional parameters get used if that was what was
3894 /// intended or change the macro to use the named parameters.  It is possible
3895 /// this warning will trigger when the none of the named parameters are used
3896 /// and the strings like $1 are infact to simply to be passed trough unchanged.
checkForBadMacro(SMLoc DirectiveLoc,StringRef Name,StringRef Body,ArrayRef<MCAsmMacroParameter> Parameters)3897 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name,
3898                                  StringRef Body,
3899                                  ArrayRef<MCAsmMacroParameter> Parameters) {
3900   // If this macro is not defined with named parameters the warning we are
3901   // checking for here doesn't apply.
3902   unsigned NParameters = Parameters.size();
3903   if (NParameters == 0)
3904     return;
3905 
3906   bool NamedParametersFound = false;
3907   bool PositionalParametersFound = false;
3908 
3909   // Look at the body of the macro for use of both the named parameters and what
3910   // are likely to be positional parameters.  This is what expandMacro() is
3911   // doing when it finds the parameters in the body.
3912   while (!Body.empty()) {
3913     // Scan for the next possible parameter.
3914     std::size_t End = Body.size(), Pos = 0;
3915     for (; Pos != End; ++Pos) {
3916       // Check for a substitution or escape.
3917       // This macro is defined with parameters, look for \foo, \bar, etc.
3918       if (Body[Pos] == '\\' && Pos + 1 != End)
3919         break;
3920 
3921       // This macro should have parameters, but look for $0, $1, ..., $n too.
3922       if (Body[Pos] != '$' || Pos + 1 == End)
3923         continue;
3924       char Next = Body[Pos + 1];
3925       if (Next == '$' || Next == 'n' ||
3926           isdigit(static_cast<unsigned char>(Next)))
3927         break;
3928     }
3929 
3930     // Check if we reached the end.
3931     if (Pos == End)
3932       break;
3933 
3934     if (Body[Pos] == '$') {
3935       switch (Body[Pos + 1]) {
3936       // $$ => $
3937       case '$':
3938         break;
3939 
3940       // $n => number of arguments
3941       case 'n':
3942         PositionalParametersFound = true;
3943         break;
3944 
3945       // $[0-9] => argument
3946       default: {
3947         PositionalParametersFound = true;
3948         break;
3949       }
3950       }
3951       Pos += 2;
3952     } else {
3953       unsigned I = Pos + 1;
3954       while (isIdentifierChar(Body[I]) && I + 1 != End)
3955         ++I;
3956 
3957       const char *Begin = Body.data() + Pos + 1;
3958       StringRef Argument(Begin, I - (Pos + 1));
3959       unsigned Index = 0;
3960       for (; Index < NParameters; ++Index)
3961         if (Parameters[Index].Name == Argument)
3962           break;
3963 
3964       if (Index == NParameters) {
3965         if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
3966           Pos += 3;
3967         else {
3968           Pos = I;
3969         }
3970       } else {
3971         NamedParametersFound = true;
3972         Pos += 1 + Argument.size();
3973       }
3974     }
3975     // Update the scan point.
3976     Body = Body.substr(Pos);
3977   }
3978 
3979   if (!NamedParametersFound && PositionalParametersFound)
3980     Warning(DirectiveLoc, "macro defined with named parameters which are not "
3981                           "used in macro body, possible positional parameter "
3982                           "found in body which will have no effect");
3983 }
3984 
3985 /// parseDirectiveExitMacro
3986 /// ::= .exitm
parseDirectiveExitMacro(StringRef Directive)3987 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) {
3988   if (getLexer().isNot(AsmToken::EndOfStatement))
3989     return TokError("unexpected token in '" + Directive + "' directive");
3990 
3991   if (!isInsideMacroInstantiation())
3992     return TokError("unexpected '" + Directive + "' in file, "
3993                                                  "no current macro definition");
3994 
3995   // Exit all conditionals that are active in the current macro.
3996   while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) {
3997     TheCondState = TheCondStack.back();
3998     TheCondStack.pop_back();
3999   }
4000 
4001   handleMacroExit();
4002   return false;
4003 }
4004 
4005 /// parseDirectiveEndMacro
4006 /// ::= .endm
4007 /// ::= .endmacro
parseDirectiveEndMacro(StringRef Directive)4008 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) {
4009   if (getLexer().isNot(AsmToken::EndOfStatement))
4010     return TokError("unexpected token in '" + Directive + "' directive");
4011 
4012   // If we are inside a macro instantiation, terminate the current
4013   // instantiation.
4014   if (isInsideMacroInstantiation()) {
4015     handleMacroExit();
4016     return false;
4017   }
4018 
4019   // Otherwise, this .endmacro is a stray entry in the file; well formed
4020   // .endmacro directives are handled during the macro definition parsing.
4021   return TokError("unexpected '" + Directive + "' in file, "
4022                                                "no current macro definition");
4023 }
4024 
4025 /// parseDirectivePurgeMacro
4026 /// ::= .purgem
parseDirectivePurgeMacro(SMLoc DirectiveLoc)4027 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) {
4028   StringRef Name;
4029   if (parseIdentifier(Name))
4030     return TokError("expected identifier in '.purgem' directive");
4031 
4032   if (getLexer().isNot(AsmToken::EndOfStatement))
4033     return TokError("unexpected token in '.purgem' directive");
4034 
4035   if (!lookupMacro(Name))
4036     return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
4037 
4038   undefineMacro(Name);
4039   return false;
4040 }
4041 
4042 /// parseDirectiveBundleAlignMode
4043 /// ::= {.bundle_align_mode} expression
parseDirectiveBundleAlignMode()4044 bool AsmParser::parseDirectiveBundleAlignMode() {
4045   checkForValidSection();
4046 
4047   // Expect a single argument: an expression that evaluates to a constant
4048   // in the inclusive range 0-30.
4049   SMLoc ExprLoc = getLexer().getLoc();
4050   int64_t AlignSizePow2;
4051   if (parseAbsoluteExpression(AlignSizePow2))
4052     return true;
4053   else if (getLexer().isNot(AsmToken::EndOfStatement))
4054     return TokError("unexpected token after expression in"
4055                     " '.bundle_align_mode' directive");
4056   else if (AlignSizePow2 < 0 || AlignSizePow2 > 30)
4057     return Error(ExprLoc,
4058                  "invalid bundle alignment size (expected between 0 and 30)");
4059 
4060   Lex();
4061 
4062   // Because of AlignSizePow2's verified range we can safely truncate it to
4063   // unsigned.
4064   getStreamer().EmitBundleAlignMode(static_cast<unsigned>(AlignSizePow2));
4065   return false;
4066 }
4067 
4068 /// parseDirectiveBundleLock
4069 /// ::= {.bundle_lock} [align_to_end]
parseDirectiveBundleLock()4070 bool AsmParser::parseDirectiveBundleLock() {
4071   checkForValidSection();
4072   bool AlignToEnd = false;
4073 
4074   if (getLexer().isNot(AsmToken::EndOfStatement)) {
4075     StringRef Option;
4076     SMLoc Loc = getTok().getLoc();
4077     const char *kInvalidOptionError =
4078         "invalid option for '.bundle_lock' directive";
4079 
4080     if (parseIdentifier(Option))
4081       return Error(Loc, kInvalidOptionError);
4082 
4083     if (Option != "align_to_end")
4084       return Error(Loc, kInvalidOptionError);
4085     else if (getLexer().isNot(AsmToken::EndOfStatement))
4086       return Error(Loc,
4087                    "unexpected token after '.bundle_lock' directive option");
4088     AlignToEnd = true;
4089   }
4090 
4091   Lex();
4092 
4093   getStreamer().EmitBundleLock(AlignToEnd);
4094   return false;
4095 }
4096 
4097 /// parseDirectiveBundleLock
4098 /// ::= {.bundle_lock}
parseDirectiveBundleUnlock()4099 bool AsmParser::parseDirectiveBundleUnlock() {
4100   checkForValidSection();
4101 
4102   if (getLexer().isNot(AsmToken::EndOfStatement))
4103     return TokError("unexpected token in '.bundle_unlock' directive");
4104   Lex();
4105 
4106   getStreamer().EmitBundleUnlock();
4107   return false;
4108 }
4109 
4110 /// parseDirectiveSpace
4111 /// ::= (.skip | .space) expression [ , expression ]
parseDirectiveSpace(StringRef IDVal)4112 bool AsmParser::parseDirectiveSpace(StringRef IDVal) {
4113   checkForValidSection();
4114 
4115   SMLoc NumBytesLoc = Lexer.getLoc();
4116   const MCExpr *NumBytes;
4117   if (parseExpression(NumBytes))
4118     return true;
4119 
4120   int64_t FillExpr = 0;
4121   if (getLexer().isNot(AsmToken::EndOfStatement)) {
4122     if (getLexer().isNot(AsmToken::Comma))
4123       return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
4124     Lex();
4125 
4126     if (parseAbsoluteExpression(FillExpr))
4127       return true;
4128 
4129     if (getLexer().isNot(AsmToken::EndOfStatement))
4130       return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
4131   }
4132 
4133   Lex();
4134 
4135   // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
4136   getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc);
4137 
4138   return false;
4139 }
4140 
4141 /// parseDirectiveLEB128
4142 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ]
parseDirectiveLEB128(bool Signed)4143 bool AsmParser::parseDirectiveLEB128(bool Signed) {
4144   checkForValidSection();
4145   const MCExpr *Value;
4146 
4147   for (;;) {
4148     if (parseExpression(Value))
4149       return true;
4150 
4151     if (Signed)
4152       getStreamer().EmitSLEB128Value(Value);
4153     else
4154       getStreamer().EmitULEB128Value(Value);
4155 
4156     if (getLexer().is(AsmToken::EndOfStatement))
4157       break;
4158 
4159     if (getLexer().isNot(AsmToken::Comma))
4160       return TokError("unexpected token in directive");
4161     Lex();
4162   }
4163 
4164   return false;
4165 }
4166 
4167 /// parseDirectiveSymbolAttribute
4168 ///  ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
parseDirectiveSymbolAttribute(MCSymbolAttr Attr)4169 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
4170   if (getLexer().isNot(AsmToken::EndOfStatement)) {
4171     for (;;) {
4172       StringRef Name;
4173       SMLoc Loc = getTok().getLoc();
4174 
4175       if (parseIdentifier(Name))
4176         return Error(Loc, "expected identifier in directive");
4177 
4178       MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
4179 
4180       // Assembler local symbols don't make any sense here. Complain loudly.
4181       if (Sym->isTemporary())
4182         return Error(Loc, "non-local symbol required in directive");
4183 
4184       if (!getStreamer().EmitSymbolAttribute(Sym, Attr))
4185         return Error(Loc, "unable to emit symbol attribute");
4186 
4187       if (getLexer().is(AsmToken::EndOfStatement))
4188         break;
4189 
4190       if (getLexer().isNot(AsmToken::Comma))
4191         return TokError("unexpected token in directive");
4192       Lex();
4193     }
4194   }
4195 
4196   Lex();
4197   return false;
4198 }
4199 
4200 /// parseDirectiveComm
4201 ///  ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
parseDirectiveComm(bool IsLocal)4202 bool AsmParser::parseDirectiveComm(bool IsLocal) {
4203   checkForValidSection();
4204 
4205   SMLoc IDLoc = getLexer().getLoc();
4206   StringRef Name;
4207   if (parseIdentifier(Name))
4208     return TokError("expected identifier in directive");
4209 
4210   // Handle the identifier as the key symbol.
4211   MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
4212 
4213   if (getLexer().isNot(AsmToken::Comma))
4214     return TokError("unexpected token in directive");
4215   Lex();
4216 
4217   int64_t Size;
4218   SMLoc SizeLoc = getLexer().getLoc();
4219   if (parseAbsoluteExpression(Size))
4220     return true;
4221 
4222   int64_t Pow2Alignment = 0;
4223   SMLoc Pow2AlignmentLoc;
4224   if (getLexer().is(AsmToken::Comma)) {
4225     Lex();
4226     Pow2AlignmentLoc = getLexer().getLoc();
4227     if (parseAbsoluteExpression(Pow2Alignment))
4228       return true;
4229 
4230     LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType();
4231     if (IsLocal && LCOMM == LCOMM::NoAlignment)
4232       return Error(Pow2AlignmentLoc, "alignment not supported on this target");
4233 
4234     // If this target takes alignments in bytes (not log) validate and convert.
4235     if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) ||
4236         (IsLocal && LCOMM == LCOMM::ByteAlignment)) {
4237       if (!isPowerOf2_64(Pow2Alignment))
4238         return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
4239       Pow2Alignment = Log2_64(Pow2Alignment);
4240     }
4241   }
4242 
4243   if (getLexer().isNot(AsmToken::EndOfStatement))
4244     return TokError("unexpected token in '.comm' or '.lcomm' directive");
4245 
4246   Lex();
4247 
4248   // NOTE: a size of zero for a .comm should create a undefined symbol
4249   // but a size of .lcomm creates a bss symbol of size zero.
4250   if (Size < 0)
4251     return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
4252                           "be less than zero");
4253 
4254   // NOTE: The alignment in the directive is a power of 2 value, the assembler
4255   // may internally end up wanting an alignment in bytes.
4256   // FIXME: Diagnose overflow.
4257   if (Pow2Alignment < 0)
4258     return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
4259                                    "alignment, can't be less than zero");
4260 
4261   if (!Sym->isUndefined())
4262     return Error(IDLoc, "invalid symbol redefinition");
4263 
4264   // Create the Symbol as a common or local common with Size and Pow2Alignment
4265   if (IsLocal) {
4266     getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment);
4267     return false;
4268   }
4269 
4270   getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
4271   return false;
4272 }
4273 
4274 /// parseDirectiveAbort
4275 ///  ::= .abort [... message ...]
parseDirectiveAbort()4276 bool AsmParser::parseDirectiveAbort() {
4277   // FIXME: Use loc from directive.
4278   SMLoc Loc = getLexer().getLoc();
4279 
4280   StringRef Str = parseStringToEndOfStatement();
4281   if (getLexer().isNot(AsmToken::EndOfStatement))
4282     return TokError("unexpected token in '.abort' directive");
4283 
4284   Lex();
4285 
4286   if (Str.empty())
4287     Error(Loc, ".abort detected. Assembly stopping.");
4288   else
4289     Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
4290   // FIXME: Actually abort assembly here.
4291 
4292   return false;
4293 }
4294 
4295 /// parseDirectiveInclude
4296 ///  ::= .include "filename"
parseDirectiveInclude()4297 bool AsmParser::parseDirectiveInclude() {
4298   if (getLexer().isNot(AsmToken::String))
4299     return TokError("expected string in '.include' directive");
4300 
4301   // Allow the strings to have escaped octal character sequence.
4302   std::string Filename;
4303   if (parseEscapedString(Filename))
4304     return true;
4305   SMLoc IncludeLoc = getLexer().getLoc();
4306   Lex();
4307 
4308   if (getLexer().isNot(AsmToken::EndOfStatement))
4309     return TokError("unexpected token in '.include' directive");
4310 
4311   // Attempt to switch the lexer to the included file before consuming the end
4312   // of statement to avoid losing it when we switch.
4313   if (enterIncludeFile(Filename)) {
4314     Error(IncludeLoc, "Could not find include file '" + Filename + "'");
4315     return true;
4316   }
4317 
4318   return false;
4319 }
4320 
4321 /// parseDirectiveIncbin
4322 ///  ::= .incbin "filename"
parseDirectiveIncbin()4323 bool AsmParser::parseDirectiveIncbin() {
4324   if (getLexer().isNot(AsmToken::String))
4325     return TokError("expected string in '.incbin' directive");
4326 
4327   // Allow the strings to have escaped octal character sequence.
4328   std::string Filename;
4329   if (parseEscapedString(Filename))
4330     return true;
4331   SMLoc IncbinLoc = getLexer().getLoc();
4332   Lex();
4333 
4334   if (getLexer().isNot(AsmToken::EndOfStatement))
4335     return TokError("unexpected token in '.incbin' directive");
4336 
4337   // Attempt to process the included file.
4338   if (processIncbinFile(Filename)) {
4339     Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
4340     return true;
4341   }
4342 
4343   return false;
4344 }
4345 
4346 /// parseDirectiveIf
4347 /// ::= .if{,eq,ge,gt,le,lt,ne} expression
parseDirectiveIf(SMLoc DirectiveLoc,DirectiveKind DirKind)4348 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) {
4349   TheCondStack.push_back(TheCondState);
4350   TheCondState.TheCond = AsmCond::IfCond;
4351   if (TheCondState.Ignore) {
4352     eatToEndOfStatement();
4353   } else {
4354     int64_t ExprValue;
4355     if (parseAbsoluteExpression(ExprValue))
4356       return true;
4357 
4358     if (getLexer().isNot(AsmToken::EndOfStatement))
4359       return TokError("unexpected token in '.if' directive");
4360 
4361     Lex();
4362 
4363     switch (DirKind) {
4364     default:
4365       llvm_unreachable("unsupported directive");
4366     case DK_IF:
4367     case DK_IFNE:
4368       break;
4369     case DK_IFEQ:
4370       ExprValue = ExprValue == 0;
4371       break;
4372     case DK_IFGE:
4373       ExprValue = ExprValue >= 0;
4374       break;
4375     case DK_IFGT:
4376       ExprValue = ExprValue > 0;
4377       break;
4378     case DK_IFLE:
4379       ExprValue = ExprValue <= 0;
4380       break;
4381     case DK_IFLT:
4382       ExprValue = ExprValue < 0;
4383       break;
4384     }
4385 
4386     TheCondState.CondMet = ExprValue;
4387     TheCondState.Ignore = !TheCondState.CondMet;
4388   }
4389 
4390   return false;
4391 }
4392 
4393 /// parseDirectiveIfb
4394 /// ::= .ifb string
parseDirectiveIfb(SMLoc DirectiveLoc,bool ExpectBlank)4395 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
4396   TheCondStack.push_back(TheCondState);
4397   TheCondState.TheCond = AsmCond::IfCond;
4398 
4399   if (TheCondState.Ignore) {
4400     eatToEndOfStatement();
4401   } else {
4402     StringRef Str = parseStringToEndOfStatement();
4403 
4404     if (getLexer().isNot(AsmToken::EndOfStatement))
4405       return TokError("unexpected token in '.ifb' directive");
4406 
4407     Lex();
4408 
4409     TheCondState.CondMet = ExpectBlank == Str.empty();
4410     TheCondState.Ignore = !TheCondState.CondMet;
4411   }
4412 
4413   return false;
4414 }
4415 
4416 /// parseDirectiveIfc
4417 /// ::= .ifc string1, string2
4418 /// ::= .ifnc string1, string2
parseDirectiveIfc(SMLoc DirectiveLoc,bool ExpectEqual)4419 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
4420   TheCondStack.push_back(TheCondState);
4421   TheCondState.TheCond = AsmCond::IfCond;
4422 
4423   if (TheCondState.Ignore) {
4424     eatToEndOfStatement();
4425   } else {
4426     StringRef Str1 = parseStringToComma();
4427 
4428     if (getLexer().isNot(AsmToken::Comma))
4429       return TokError("unexpected token in '.ifc' directive");
4430 
4431     Lex();
4432 
4433     StringRef Str2 = parseStringToEndOfStatement();
4434 
4435     if (getLexer().isNot(AsmToken::EndOfStatement))
4436       return TokError("unexpected token in '.ifc' directive");
4437 
4438     Lex();
4439 
4440     TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim());
4441     TheCondState.Ignore = !TheCondState.CondMet;
4442   }
4443 
4444   return false;
4445 }
4446 
4447 /// parseDirectiveIfeqs
4448 ///   ::= .ifeqs string1, string2
parseDirectiveIfeqs(SMLoc DirectiveLoc,bool ExpectEqual)4449 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) {
4450   if (Lexer.isNot(AsmToken::String)) {
4451     if (ExpectEqual)
4452       TokError("expected string parameter for '.ifeqs' directive");
4453     else
4454       TokError("expected string parameter for '.ifnes' directive");
4455     eatToEndOfStatement();
4456     return true;
4457   }
4458 
4459   StringRef String1 = getTok().getStringContents();
4460   Lex();
4461 
4462   if (Lexer.isNot(AsmToken::Comma)) {
4463     if (ExpectEqual)
4464       TokError("expected comma after first string for '.ifeqs' directive");
4465     else
4466       TokError("expected comma after first string for '.ifnes' directive");
4467     eatToEndOfStatement();
4468     return true;
4469   }
4470 
4471   Lex();
4472 
4473   if (Lexer.isNot(AsmToken::String)) {
4474     if (ExpectEqual)
4475       TokError("expected string parameter for '.ifeqs' directive");
4476     else
4477       TokError("expected string parameter for '.ifnes' directive");
4478     eatToEndOfStatement();
4479     return true;
4480   }
4481 
4482   StringRef String2 = getTok().getStringContents();
4483   Lex();
4484 
4485   TheCondStack.push_back(TheCondState);
4486   TheCondState.TheCond = AsmCond::IfCond;
4487   TheCondState.CondMet = ExpectEqual == (String1 == String2);
4488   TheCondState.Ignore = !TheCondState.CondMet;
4489 
4490   return false;
4491 }
4492 
4493 /// parseDirectiveIfdef
4494 /// ::= .ifdef symbol
parseDirectiveIfdef(SMLoc DirectiveLoc,bool expect_defined)4495 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
4496   StringRef Name;
4497   TheCondStack.push_back(TheCondState);
4498   TheCondState.TheCond = AsmCond::IfCond;
4499 
4500   if (TheCondState.Ignore) {
4501     eatToEndOfStatement();
4502   } else {
4503     if (parseIdentifier(Name))
4504       return TokError("expected identifier after '.ifdef'");
4505 
4506     Lex();
4507 
4508     MCSymbol *Sym = getContext().lookupSymbol(Name);
4509 
4510     if (expect_defined)
4511       TheCondState.CondMet = (Sym && !Sym->isUndefined());
4512     else
4513       TheCondState.CondMet = (!Sym || Sym->isUndefined());
4514     TheCondState.Ignore = !TheCondState.CondMet;
4515   }
4516 
4517   return false;
4518 }
4519 
4520 /// parseDirectiveElseIf
4521 /// ::= .elseif expression
parseDirectiveElseIf(SMLoc DirectiveLoc)4522 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) {
4523   if (TheCondState.TheCond != AsmCond::IfCond &&
4524       TheCondState.TheCond != AsmCond::ElseIfCond)
4525     Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
4526                         " an .elseif");
4527   TheCondState.TheCond = AsmCond::ElseIfCond;
4528 
4529   bool LastIgnoreState = false;
4530   if (!TheCondStack.empty())
4531     LastIgnoreState = TheCondStack.back().Ignore;
4532   if (LastIgnoreState || TheCondState.CondMet) {
4533     TheCondState.Ignore = true;
4534     eatToEndOfStatement();
4535   } else {
4536     int64_t ExprValue;
4537     if (parseAbsoluteExpression(ExprValue))
4538       return true;
4539 
4540     if (getLexer().isNot(AsmToken::EndOfStatement))
4541       return TokError("unexpected token in '.elseif' directive");
4542 
4543     Lex();
4544     TheCondState.CondMet = ExprValue;
4545     TheCondState.Ignore = !TheCondState.CondMet;
4546   }
4547 
4548   return false;
4549 }
4550 
4551 /// parseDirectiveElse
4552 /// ::= .else
parseDirectiveElse(SMLoc DirectiveLoc)4553 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) {
4554   if (getLexer().isNot(AsmToken::EndOfStatement))
4555     return TokError("unexpected token in '.else' directive");
4556 
4557   Lex();
4558 
4559   if (TheCondState.TheCond != AsmCond::IfCond &&
4560       TheCondState.TheCond != AsmCond::ElseIfCond)
4561     Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
4562                         ".elseif");
4563   TheCondState.TheCond = AsmCond::ElseCond;
4564   bool LastIgnoreState = false;
4565   if (!TheCondStack.empty())
4566     LastIgnoreState = TheCondStack.back().Ignore;
4567   if (LastIgnoreState || TheCondState.CondMet)
4568     TheCondState.Ignore = true;
4569   else
4570     TheCondState.Ignore = false;
4571 
4572   return false;
4573 }
4574 
4575 /// parseDirectiveEnd
4576 /// ::= .end
parseDirectiveEnd(SMLoc DirectiveLoc)4577 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) {
4578   if (getLexer().isNot(AsmToken::EndOfStatement))
4579     return TokError("unexpected token in '.end' directive");
4580 
4581   Lex();
4582 
4583   while (Lexer.isNot(AsmToken::Eof))
4584     Lex();
4585 
4586   return false;
4587 }
4588 
4589 /// parseDirectiveError
4590 ///   ::= .err
4591 ///   ::= .error [string]
parseDirectiveError(SMLoc L,bool WithMessage)4592 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) {
4593   if (!TheCondStack.empty()) {
4594     if (TheCondStack.back().Ignore) {
4595       eatToEndOfStatement();
4596       return false;
4597     }
4598   }
4599 
4600   if (!WithMessage)
4601     return Error(L, ".err encountered");
4602 
4603   StringRef Message = ".error directive invoked in source file";
4604   if (Lexer.isNot(AsmToken::EndOfStatement)) {
4605     if (Lexer.isNot(AsmToken::String)) {
4606       TokError(".error argument must be a string");
4607       eatToEndOfStatement();
4608       return true;
4609     }
4610 
4611     Message = getTok().getStringContents();
4612     Lex();
4613   }
4614 
4615   Error(L, Message);
4616   return true;
4617 }
4618 
4619 /// parseDirectiveWarning
4620 ///   ::= .warning [string]
parseDirectiveWarning(SMLoc L)4621 bool AsmParser::parseDirectiveWarning(SMLoc L) {
4622   if (!TheCondStack.empty()) {
4623     if (TheCondStack.back().Ignore) {
4624       eatToEndOfStatement();
4625       return false;
4626     }
4627   }
4628 
4629   StringRef Message = ".warning directive invoked in source file";
4630   if (Lexer.isNot(AsmToken::EndOfStatement)) {
4631     if (Lexer.isNot(AsmToken::String)) {
4632       TokError(".warning argument must be a string");
4633       eatToEndOfStatement();
4634       return true;
4635     }
4636 
4637     Message = getTok().getStringContents();
4638     Lex();
4639   }
4640 
4641   Warning(L, Message);
4642   return false;
4643 }
4644 
4645 /// parseDirectiveEndIf
4646 /// ::= .endif
parseDirectiveEndIf(SMLoc DirectiveLoc)4647 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) {
4648   if (getLexer().isNot(AsmToken::EndOfStatement))
4649     return TokError("unexpected token in '.endif' directive");
4650 
4651   Lex();
4652 
4653   if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty())
4654     Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
4655                         ".else");
4656   if (!TheCondStack.empty()) {
4657     TheCondState = TheCondStack.back();
4658     TheCondStack.pop_back();
4659   }
4660 
4661   return false;
4662 }
4663 
initializeDirectiveKindMap()4664 void AsmParser::initializeDirectiveKindMap() {
4665   DirectiveKindMap[".set"] = DK_SET;
4666   DirectiveKindMap[".equ"] = DK_EQU;
4667   DirectiveKindMap[".equiv"] = DK_EQUIV;
4668   DirectiveKindMap[".ascii"] = DK_ASCII;
4669   DirectiveKindMap[".asciz"] = DK_ASCIZ;
4670   DirectiveKindMap[".string"] = DK_STRING;
4671   DirectiveKindMap[".byte"] = DK_BYTE;
4672   DirectiveKindMap[".short"] = DK_SHORT;
4673   DirectiveKindMap[".value"] = DK_VALUE;
4674   DirectiveKindMap[".2byte"] = DK_2BYTE;
4675   DirectiveKindMap[".long"] = DK_LONG;
4676   DirectiveKindMap[".int"] = DK_INT;
4677   DirectiveKindMap[".4byte"] = DK_4BYTE;
4678   DirectiveKindMap[".quad"] = DK_QUAD;
4679   DirectiveKindMap[".8byte"] = DK_8BYTE;
4680   DirectiveKindMap[".octa"] = DK_OCTA;
4681   DirectiveKindMap[".single"] = DK_SINGLE;
4682   DirectiveKindMap[".float"] = DK_FLOAT;
4683   DirectiveKindMap[".double"] = DK_DOUBLE;
4684   DirectiveKindMap[".align"] = DK_ALIGN;
4685   DirectiveKindMap[".align32"] = DK_ALIGN32;
4686   DirectiveKindMap[".balign"] = DK_BALIGN;
4687   DirectiveKindMap[".balignw"] = DK_BALIGNW;
4688   DirectiveKindMap[".balignl"] = DK_BALIGNL;
4689   DirectiveKindMap[".p2align"] = DK_P2ALIGN;
4690   DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW;
4691   DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL;
4692   DirectiveKindMap[".org"] = DK_ORG;
4693   DirectiveKindMap[".fill"] = DK_FILL;
4694   DirectiveKindMap[".zero"] = DK_ZERO;
4695   DirectiveKindMap[".extern"] = DK_EXTERN;
4696   DirectiveKindMap[".globl"] = DK_GLOBL;
4697   DirectiveKindMap[".global"] = DK_GLOBAL;
4698   DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE;
4699   DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP;
4700   DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER;
4701   DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN;
4702   DirectiveKindMap[".reference"] = DK_REFERENCE;
4703   DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION;
4704   DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE;
4705   DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN;
4706   DirectiveKindMap[".comm"] = DK_COMM;
4707   DirectiveKindMap[".common"] = DK_COMMON;
4708   DirectiveKindMap[".lcomm"] = DK_LCOMM;
4709   DirectiveKindMap[".abort"] = DK_ABORT;
4710   DirectiveKindMap[".include"] = DK_INCLUDE;
4711   DirectiveKindMap[".incbin"] = DK_INCBIN;
4712   DirectiveKindMap[".code16"] = DK_CODE16;
4713   DirectiveKindMap[".code16gcc"] = DK_CODE16GCC;
4714   DirectiveKindMap[".rept"] = DK_REPT;
4715   DirectiveKindMap[".rep"] = DK_REPT;
4716   DirectiveKindMap[".irp"] = DK_IRP;
4717   DirectiveKindMap[".irpc"] = DK_IRPC;
4718   DirectiveKindMap[".endr"] = DK_ENDR;
4719   DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE;
4720   DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK;
4721   DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK;
4722   DirectiveKindMap[".if"] = DK_IF;
4723   DirectiveKindMap[".ifeq"] = DK_IFEQ;
4724   DirectiveKindMap[".ifge"] = DK_IFGE;
4725   DirectiveKindMap[".ifgt"] = DK_IFGT;
4726   DirectiveKindMap[".ifle"] = DK_IFLE;
4727   DirectiveKindMap[".iflt"] = DK_IFLT;
4728   DirectiveKindMap[".ifne"] = DK_IFNE;
4729   DirectiveKindMap[".ifb"] = DK_IFB;
4730   DirectiveKindMap[".ifnb"] = DK_IFNB;
4731   DirectiveKindMap[".ifc"] = DK_IFC;
4732   DirectiveKindMap[".ifeqs"] = DK_IFEQS;
4733   DirectiveKindMap[".ifnc"] = DK_IFNC;
4734   DirectiveKindMap[".ifnes"] = DK_IFNES;
4735   DirectiveKindMap[".ifdef"] = DK_IFDEF;
4736   DirectiveKindMap[".ifndef"] = DK_IFNDEF;
4737   DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF;
4738   DirectiveKindMap[".elseif"] = DK_ELSEIF;
4739   DirectiveKindMap[".else"] = DK_ELSE;
4740   DirectiveKindMap[".end"] = DK_END;
4741   DirectiveKindMap[".endif"] = DK_ENDIF;
4742   DirectiveKindMap[".skip"] = DK_SKIP;
4743   DirectiveKindMap[".space"] = DK_SPACE;
4744   DirectiveKindMap[".file"] = DK_FILE;
4745   DirectiveKindMap[".line"] = DK_LINE;
4746   DirectiveKindMap[".loc"] = DK_LOC;
4747   DirectiveKindMap[".stabs"] = DK_STABS;
4748   DirectiveKindMap[".cv_file"] = DK_CV_FILE;
4749   DirectiveKindMap[".cv_loc"] = DK_CV_LOC;
4750   DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE;
4751   DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE;
4752   DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE;
4753   DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE;
4754   DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS;
4755   DirectiveKindMap[".sleb128"] = DK_SLEB128;
4756   DirectiveKindMap[".uleb128"] = DK_ULEB128;
4757   DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS;
4758   DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC;
4759   DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC;
4760   DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA;
4761   DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET;
4762   DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET;
4763   DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER;
4764   DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET;
4765   DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET;
4766   DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY;
4767   DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA;
4768   DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE;
4769   DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE;
4770   DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE;
4771   DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE;
4772   DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE;
4773   DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME;
4774   DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED;
4775   DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER;
4776   DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE;
4777   DirectiveKindMap[".macros_on"] = DK_MACROS_ON;
4778   DirectiveKindMap[".macros_off"] = DK_MACROS_OFF;
4779   DirectiveKindMap[".macro"] = DK_MACRO;
4780   DirectiveKindMap[".exitm"] = DK_EXITM;
4781   DirectiveKindMap[".endm"] = DK_ENDM;
4782   DirectiveKindMap[".endmacro"] = DK_ENDMACRO;
4783   DirectiveKindMap[".purgem"] = DK_PURGEM;
4784   DirectiveKindMap[".err"] = DK_ERR;
4785   DirectiveKindMap[".error"] = DK_ERROR;
4786   DirectiveKindMap[".warning"] = DK_WARNING;
4787   DirectiveKindMap[".reloc"] = DK_RELOC;
4788 }
4789 
parseMacroLikeBody(SMLoc DirectiveLoc)4790 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) {
4791   AsmToken EndToken, StartToken = getTok();
4792 
4793   unsigned NestLevel = 0;
4794   for (;;) {
4795     // Check whether we have reached the end of the file.
4796     if (getLexer().is(AsmToken::Eof)) {
4797       Error(DirectiveLoc, "no matching '.endr' in definition");
4798       return nullptr;
4799     }
4800 
4801     if (Lexer.is(AsmToken::Identifier) &&
4802         (getTok().getIdentifier() == ".rept" ||
4803          getTok().getIdentifier() == ".irp" ||
4804          getTok().getIdentifier() == ".irpc")) {
4805       ++NestLevel;
4806     }
4807 
4808     // Otherwise, check whether we have reached the .endr.
4809     if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") {
4810       if (NestLevel == 0) {
4811         EndToken = getTok();
4812         Lex();
4813         if (Lexer.isNot(AsmToken::EndOfStatement)) {
4814           TokError("unexpected token in '.endr' directive");
4815           return nullptr;
4816         }
4817         break;
4818       }
4819       --NestLevel;
4820     }
4821 
4822     // Otherwise, scan till the end of the statement.
4823     eatToEndOfStatement();
4824   }
4825 
4826   const char *BodyStart = StartToken.getLoc().getPointer();
4827   const char *BodyEnd = EndToken.getLoc().getPointer();
4828   StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
4829 
4830   // We Are Anonymous.
4831   MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters());
4832   return &MacroLikeBodies.back();
4833 }
4834 
instantiateMacroLikeBody(MCAsmMacro * M,SMLoc DirectiveLoc,raw_svector_ostream & OS)4835 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
4836                                          raw_svector_ostream &OS) {
4837   OS << ".endr\n";
4838 
4839   std::unique_ptr<MemoryBuffer> Instantiation =
4840       MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
4841 
4842   // Create the macro instantiation object and add to the current macro
4843   // instantiation stack.
4844   MacroInstantiation *MI = new MacroInstantiation(
4845       DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size());
4846   ActiveMacros.push_back(MI);
4847 
4848   // Jump to the macro instantiation and prime the lexer.
4849   CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
4850   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
4851   Lex();
4852 }
4853 
4854 /// parseDirectiveRept
4855 ///   ::= .rep | .rept count
parseDirectiveRept(SMLoc DirectiveLoc,StringRef Dir)4856 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) {
4857   const MCExpr *CountExpr;
4858   SMLoc CountLoc = getTok().getLoc();
4859   if (parseExpression(CountExpr))
4860     return true;
4861 
4862   int64_t Count;
4863   if (!CountExpr->evaluateAsAbsolute(Count)) {
4864     eatToEndOfStatement();
4865     return Error(CountLoc, "unexpected token in '" + Dir + "' directive");
4866   }
4867 
4868   if (Count < 0)
4869     return Error(CountLoc, "Count is negative");
4870 
4871   if (Lexer.isNot(AsmToken::EndOfStatement))
4872     return TokError("unexpected token in '" + Dir + "' directive");
4873 
4874   // Eat the end of statement.
4875   Lex();
4876 
4877   // Lex the rept definition.
4878   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4879   if (!M)
4880     return true;
4881 
4882   // Macro instantiation is lexical, unfortunately. We construct a new buffer
4883   // to hold the macro body with substitutions.
4884   SmallString<256> Buf;
4885   raw_svector_ostream OS(Buf);
4886   while (Count--) {
4887     // Note that the AtPseudoVariable is disabled for instantiations of .rep(t).
4888     if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc()))
4889       return true;
4890   }
4891   instantiateMacroLikeBody(M, DirectiveLoc, OS);
4892 
4893   return false;
4894 }
4895 
4896 /// parseDirectiveIrp
4897 /// ::= .irp symbol,values
parseDirectiveIrp(SMLoc DirectiveLoc)4898 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) {
4899   MCAsmMacroParameter Parameter;
4900 
4901   if (parseIdentifier(Parameter.Name))
4902     return TokError("expected identifier in '.irp' directive");
4903 
4904   if (Lexer.isNot(AsmToken::Comma))
4905     return TokError("expected comma in '.irp' directive");
4906 
4907   Lex();
4908 
4909   MCAsmMacroArguments A;
4910   if (parseMacroArguments(nullptr, A))
4911     return true;
4912 
4913   // Eat the end of statement.
4914   Lex();
4915 
4916   // Lex the irp definition.
4917   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4918   if (!M)
4919     return true;
4920 
4921   // Macro instantiation is lexical, unfortunately. We construct a new buffer
4922   // to hold the macro body with substitutions.
4923   SmallString<256> Buf;
4924   raw_svector_ostream OS(Buf);
4925 
4926   for (const MCAsmMacroArgument &Arg : A) {
4927     // Note that the AtPseudoVariable is enabled for instantiations of .irp.
4928     // This is undocumented, but GAS seems to support it.
4929     if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc()))
4930       return true;
4931   }
4932 
4933   instantiateMacroLikeBody(M, DirectiveLoc, OS);
4934 
4935   return false;
4936 }
4937 
4938 /// parseDirectiveIrpc
4939 /// ::= .irpc symbol,values
parseDirectiveIrpc(SMLoc DirectiveLoc)4940 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) {
4941   MCAsmMacroParameter Parameter;
4942 
4943   if (parseIdentifier(Parameter.Name))
4944     return TokError("expected identifier in '.irpc' directive");
4945 
4946   if (Lexer.isNot(AsmToken::Comma))
4947     return TokError("expected comma in '.irpc' directive");
4948 
4949   Lex();
4950 
4951   MCAsmMacroArguments A;
4952   if (parseMacroArguments(nullptr, A))
4953     return true;
4954 
4955   if (A.size() != 1 || A.front().size() != 1)
4956     return TokError("unexpected token in '.irpc' directive");
4957 
4958   // Eat the end of statement.
4959   Lex();
4960 
4961   // Lex the irpc definition.
4962   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4963   if (!M)
4964     return true;
4965 
4966   // Macro instantiation is lexical, unfortunately. We construct a new buffer
4967   // to hold the macro body with substitutions.
4968   SmallString<256> Buf;
4969   raw_svector_ostream OS(Buf);
4970 
4971   StringRef Values = A.front().front().getString();
4972   for (std::size_t I = 0, End = Values.size(); I != End; ++I) {
4973     MCAsmMacroArgument Arg;
4974     Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1));
4975 
4976     // Note that the AtPseudoVariable is enabled for instantiations of .irpc.
4977     // This is undocumented, but GAS seems to support it.
4978     if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc()))
4979       return true;
4980   }
4981 
4982   instantiateMacroLikeBody(M, DirectiveLoc, OS);
4983 
4984   return false;
4985 }
4986 
parseDirectiveEndr(SMLoc DirectiveLoc)4987 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) {
4988   if (ActiveMacros.empty())
4989     return TokError("unmatched '.endr' directive");
4990 
4991   // The only .repl that should get here are the ones created by
4992   // instantiateMacroLikeBody.
4993   assert(getLexer().is(AsmToken::EndOfStatement));
4994 
4995   handleMacroExit();
4996   return false;
4997 }
4998 
parseDirectiveMSEmit(SMLoc IDLoc,ParseStatementInfo & Info,size_t Len)4999 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info,
5000                                      size_t Len) {
5001   const MCExpr *Value;
5002   SMLoc ExprLoc = getLexer().getLoc();
5003   if (parseExpression(Value))
5004     return true;
5005   const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
5006   if (!MCE)
5007     return Error(ExprLoc, "unexpected expression in _emit");
5008   uint64_t IntValue = MCE->getValue();
5009   if (!isUInt<8>(IntValue) && !isInt<8>(IntValue))
5010     return Error(ExprLoc, "literal value out of range for directive");
5011 
5012   Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len);
5013   return false;
5014 }
5015 
parseDirectiveMSAlign(SMLoc IDLoc,ParseStatementInfo & Info)5016 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) {
5017   const MCExpr *Value;
5018   SMLoc ExprLoc = getLexer().getLoc();
5019   if (parseExpression(Value))
5020     return true;
5021   const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
5022   if (!MCE)
5023     return Error(ExprLoc, "unexpected expression in align");
5024   uint64_t IntValue = MCE->getValue();
5025   if (!isPowerOf2_64(IntValue))
5026     return Error(ExprLoc, "literal value not a power of two greater then zero");
5027 
5028   Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue));
5029   return false;
5030 }
5031 
5032 // We are comparing pointers, but the pointers are relative to a single string.
5033 // Thus, this should always be deterministic.
rewritesSort(const AsmRewrite * AsmRewriteA,const AsmRewrite * AsmRewriteB)5034 static int rewritesSort(const AsmRewrite *AsmRewriteA,
5035                         const AsmRewrite *AsmRewriteB) {
5036   if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer())
5037     return -1;
5038   if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer())
5039     return 1;
5040 
5041   // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output
5042   // rewrite to the same location.  Make sure the SizeDirective rewrite is
5043   // performed first, then the Imm/ImmPrefix and finally the Input/Output.  This
5044   // ensures the sort algorithm is stable.
5045   if (AsmRewritePrecedence[AsmRewriteA->Kind] >
5046       AsmRewritePrecedence[AsmRewriteB->Kind])
5047     return -1;
5048 
5049   if (AsmRewritePrecedence[AsmRewriteA->Kind] <
5050       AsmRewritePrecedence[AsmRewriteB->Kind])
5051     return 1;
5052   llvm_unreachable("Unstable rewrite sort.");
5053 }
5054 
parseMSInlineAsm(void * AsmLoc,std::string & AsmString,unsigned & NumOutputs,unsigned & NumInputs,SmallVectorImpl<std::pair<void *,bool>> & OpDecls,SmallVectorImpl<std::string> & Constraints,SmallVectorImpl<std::string> & Clobbers,const MCInstrInfo * MII,const MCInstPrinter * IP,MCAsmParserSemaCallback & SI)5055 bool AsmParser::parseMSInlineAsm(
5056     void *AsmLoc, std::string &AsmString, unsigned &NumOutputs,
5057     unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool> > &OpDecls,
5058     SmallVectorImpl<std::string> &Constraints,
5059     SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII,
5060     const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) {
5061   SmallVector<void *, 4> InputDecls;
5062   SmallVector<void *, 4> OutputDecls;
5063   SmallVector<bool, 4> InputDeclsAddressOf;
5064   SmallVector<bool, 4> OutputDeclsAddressOf;
5065   SmallVector<std::string, 4> InputConstraints;
5066   SmallVector<std::string, 4> OutputConstraints;
5067   SmallVector<unsigned, 4> ClobberRegs;
5068 
5069   SmallVector<AsmRewrite, 4> AsmStrRewrites;
5070 
5071   // Prime the lexer.
5072   Lex();
5073 
5074   // While we have input, parse each statement.
5075   unsigned InputIdx = 0;
5076   unsigned OutputIdx = 0;
5077   while (getLexer().isNot(AsmToken::Eof)) {
5078     // Parse curly braces marking block start/end
5079     if (parseCurlyBlockScope(AsmStrRewrites))
5080       continue;
5081 
5082     ParseStatementInfo Info(&AsmStrRewrites);
5083     if (parseStatement(Info, &SI))
5084       return true;
5085 
5086     if (Info.ParseError)
5087       return true;
5088 
5089     if (Info.Opcode == ~0U)
5090       continue;
5091 
5092     const MCInstrDesc &Desc = MII->get(Info.Opcode);
5093 
5094     // Build the list of clobbers, outputs and inputs.
5095     for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
5096       MCParsedAsmOperand &Operand = *Info.ParsedOperands[i];
5097 
5098       // Immediate.
5099       if (Operand.isImm())
5100         continue;
5101 
5102       // Register operand.
5103       if (Operand.isReg() && !Operand.needAddressOf() &&
5104           !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) {
5105         unsigned NumDefs = Desc.getNumDefs();
5106         // Clobber.
5107         if (NumDefs && Operand.getMCOperandNum() < NumDefs)
5108           ClobberRegs.push_back(Operand.getReg());
5109         continue;
5110       }
5111 
5112       // Expr/Input or Output.
5113       StringRef SymName = Operand.getSymName();
5114       if (SymName.empty())
5115         continue;
5116 
5117       void *OpDecl = Operand.getOpDecl();
5118       if (!OpDecl)
5119         continue;
5120 
5121       bool isOutput = (i == 1) && Desc.mayStore();
5122       SMLoc Start = SMLoc::getFromPointer(SymName.data());
5123       if (isOutput) {
5124         ++InputIdx;
5125         OutputDecls.push_back(OpDecl);
5126         OutputDeclsAddressOf.push_back(Operand.needAddressOf());
5127         OutputConstraints.push_back(("=" + Operand.getConstraint()).str());
5128         AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size());
5129       } else {
5130         InputDecls.push_back(OpDecl);
5131         InputDeclsAddressOf.push_back(Operand.needAddressOf());
5132         InputConstraints.push_back(Operand.getConstraint().str());
5133         AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size());
5134       }
5135     }
5136 
5137     // Consider implicit defs to be clobbers.  Think of cpuid and push.
5138     ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(),
5139                                 Desc.getNumImplicitDefs());
5140     ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end());
5141   }
5142 
5143   // Set the number of Outputs and Inputs.
5144   NumOutputs = OutputDecls.size();
5145   NumInputs = InputDecls.size();
5146 
5147   // Set the unique clobbers.
5148   array_pod_sort(ClobberRegs.begin(), ClobberRegs.end());
5149   ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()),
5150                     ClobberRegs.end());
5151   Clobbers.assign(ClobberRegs.size(), std::string());
5152   for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) {
5153     raw_string_ostream OS(Clobbers[I]);
5154     IP->printRegName(OS, ClobberRegs[I]);
5155   }
5156 
5157   // Merge the various outputs and inputs.  Output are expected first.
5158   if (NumOutputs || NumInputs) {
5159     unsigned NumExprs = NumOutputs + NumInputs;
5160     OpDecls.resize(NumExprs);
5161     Constraints.resize(NumExprs);
5162     for (unsigned i = 0; i < NumOutputs; ++i) {
5163       OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]);
5164       Constraints[i] = OutputConstraints[i];
5165     }
5166     for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
5167       OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]);
5168       Constraints[j] = InputConstraints[i];
5169     }
5170   }
5171 
5172   // Build the IR assembly string.
5173   std::string AsmStringIR;
5174   raw_string_ostream OS(AsmStringIR);
5175   StringRef ASMString =
5176       SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer();
5177   const char *AsmStart = ASMString.begin();
5178   const char *AsmEnd = ASMString.end();
5179   array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort);
5180   for (const AsmRewrite &AR : AsmStrRewrites) {
5181     AsmRewriteKind Kind = AR.Kind;
5182     if (Kind == AOK_Delete)
5183       continue;
5184 
5185     const char *Loc = AR.Loc.getPointer();
5186     assert(Loc >= AsmStart && "Expected Loc to be at or after Start!");
5187 
5188     // Emit everything up to the immediate/expression.
5189     if (unsigned Len = Loc - AsmStart)
5190       OS << StringRef(AsmStart, Len);
5191 
5192     // Skip the original expression.
5193     if (Kind == AOK_Skip) {
5194       AsmStart = Loc + AR.Len;
5195       continue;
5196     }
5197 
5198     unsigned AdditionalSkip = 0;
5199     // Rewrite expressions in $N notation.
5200     switch (Kind) {
5201     default:
5202       break;
5203     case AOK_Imm:
5204       OS << "$$" << AR.Val;
5205       break;
5206     case AOK_ImmPrefix:
5207       OS << "$$";
5208       break;
5209     case AOK_Label:
5210       OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label;
5211       break;
5212     case AOK_Input:
5213       OS << '$' << InputIdx++;
5214       break;
5215     case AOK_Output:
5216       OS << '$' << OutputIdx++;
5217       break;
5218     case AOK_SizeDirective:
5219       switch (AR.Val) {
5220       default: break;
5221       case 8:  OS << "byte ptr "; break;
5222       case 16: OS << "word ptr "; break;
5223       case 32: OS << "dword ptr "; break;
5224       case 64: OS << "qword ptr "; break;
5225       case 80: OS << "xword ptr "; break;
5226       case 128: OS << "xmmword ptr "; break;
5227       case 256: OS << "ymmword ptr "; break;
5228       }
5229       break;
5230     case AOK_Emit:
5231       OS << ".byte";
5232       break;
5233     case AOK_Align: {
5234       // MS alignment directives are measured in bytes. If the native assembler
5235       // measures alignment in bytes, we can pass it straight through.
5236       OS << ".align";
5237       if (getContext().getAsmInfo()->getAlignmentIsInBytes())
5238         break;
5239 
5240       // Alignment is in log2 form, so print that instead and skip the original
5241       // immediate.
5242       unsigned Val = AR.Val;
5243       OS << ' ' << Val;
5244       assert(Val < 10 && "Expected alignment less then 2^10.");
5245       AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4;
5246       break;
5247     }
5248     case AOK_EVEN:
5249       OS << ".even";
5250       break;
5251     case AOK_DotOperator:
5252       // Insert the dot if the user omitted it.
5253       OS.flush();
5254       if (AsmStringIR.back() != '.')
5255         OS << '.';
5256       OS << AR.Val;
5257       break;
5258     case AOK_EndOfStatement:
5259       OS << "\n\t";
5260       break;
5261     }
5262 
5263     // Skip the original expression.
5264     AsmStart = Loc + AR.Len + AdditionalSkip;
5265   }
5266 
5267   // Emit the remainder of the asm string.
5268   if (AsmStart != AsmEnd)
5269     OS << StringRef(AsmStart, AsmEnd - AsmStart);
5270 
5271   AsmString = OS.str();
5272   return false;
5273 }
5274 
5275 namespace llvm {
5276 namespace MCParserUtils {
5277 
5278 /// Returns whether the given symbol is used anywhere in the given expression,
5279 /// or subexpressions.
isSymbolUsedInExpression(const MCSymbol * Sym,const MCExpr * Value)5280 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) {
5281   switch (Value->getKind()) {
5282   case MCExpr::Binary: {
5283     const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value);
5284     return isSymbolUsedInExpression(Sym, BE->getLHS()) ||
5285            isSymbolUsedInExpression(Sym, BE->getRHS());
5286   }
5287   case MCExpr::Target:
5288   case MCExpr::Constant:
5289     return false;
5290   case MCExpr::SymbolRef: {
5291     const MCSymbol &S =
5292         static_cast<const MCSymbolRefExpr *>(Value)->getSymbol();
5293     if (S.isVariable())
5294       return isSymbolUsedInExpression(Sym, S.getVariableValue());
5295     return &S == Sym;
5296   }
5297   case MCExpr::Unary:
5298     return isSymbolUsedInExpression(
5299         Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr());
5300   }
5301 
5302   llvm_unreachable("Unknown expr kind!");
5303 }
5304 
parseAssignmentExpression(StringRef Name,bool allow_redef,MCAsmParser & Parser,MCSymbol * & Sym,const MCExpr * & Value)5305 bool parseAssignmentExpression(StringRef Name, bool allow_redef,
5306                                MCAsmParser &Parser, MCSymbol *&Sym,
5307                                const MCExpr *&Value) {
5308 
5309   // FIXME: Use better location, we should use proper tokens.
5310   SMLoc EqualLoc = Parser.getTok().getLoc();
5311 
5312   if (Parser.parseExpression(Value)) {
5313     Parser.TokError("missing expression");
5314     Parser.eatToEndOfStatement();
5315     return true;
5316   }
5317 
5318   // Note: we don't count b as used in "a = b". This is to allow
5319   // a = b
5320   // b = c
5321 
5322   if (Parser.getTok().isNot(AsmToken::EndOfStatement))
5323     return Parser.TokError("unexpected token in assignment");
5324 
5325   // Eat the end of statement marker.
5326   Parser.Lex();
5327 
5328   // Validate that the LHS is allowed to be a variable (either it has not been
5329   // used as a symbol, or it is an absolute symbol).
5330   Sym = Parser.getContext().lookupSymbol(Name);
5331   if (Sym) {
5332     // Diagnose assignment to a label.
5333     //
5334     // FIXME: Diagnostics. Note the location of the definition as a label.
5335     // FIXME: Diagnose assignment to protected identifier (e.g., register name).
5336     if (isSymbolUsedInExpression(Sym, Value))
5337       return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'");
5338     else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() &&
5339              !Sym->isVariable())
5340       ; // Allow redefinitions of undefined symbols only used in directives.
5341     else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
5342       ; // Allow redefinitions of variables that haven't yet been used.
5343     else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
5344       return Parser.Error(EqualLoc, "redefinition of '" + Name + "'");
5345     else if (!Sym->isVariable())
5346       return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'");
5347     else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
5348       return Parser.Error(EqualLoc,
5349                           "invalid reassignment of non-absolute variable '" +
5350                               Name + "'");
5351   } else if (Name == ".") {
5352     Parser.getStreamer().emitValueToOffset(Value, 0);
5353     return false;
5354   } else
5355     Sym = Parser.getContext().getOrCreateSymbol(Name);
5356 
5357   Sym->setRedefinable(allow_redef);
5358 
5359   return false;
5360 }
5361 
5362 } // namespace MCParserUtils
5363 } // namespace llvm
5364 
5365 /// \brief Create an MCAsmParser instance.
createMCAsmParser(SourceMgr & SM,MCContext & C,MCStreamer & Out,const MCAsmInfo & MAI)5366 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C,
5367                                      MCStreamer &Out, const MCAsmInfo &MAI) {
5368   return new AsmParser(SM, C, Out, MAI);
5369 }
5370