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/MCRegisterInfo.h"
32 #include "llvm/MC/MCSectionMachO.h"
33 #include "llvm/MC/MCStreamer.h"
34 #include "llvm/MC/MCSymbol.h"
35 #include "llvm/MC/MCTargetAsmParser.h"
36 #include "llvm/MC/MCValue.h"
37 #include "llvm/Support/CommandLine.h"
38 #include "llvm/Support/ErrorHandling.h"
39 #include "llvm/Support/MathExtras.h"
40 #include "llvm/Support/MemoryBuffer.h"
41 #include "llvm/Support/SourceMgr.h"
42 #include "llvm/Support/raw_ostream.h"
43 #include <cctype>
44 #include <deque>
45 #include <set>
46 #include <string>
47 #include <vector>
48 using namespace llvm;
49 
~MCAsmParserSemaCallback()50 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() {}
51 
52 namespace {
53 /// \brief Helper types for tracking macro definitions.
54 typedef std::vector<AsmToken> MCAsmMacroArgument;
55 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments;
56 
57 struct MCAsmMacroParameter {
58   StringRef Name;
59   MCAsmMacroArgument Value;
60   bool Required;
61   bool Vararg;
62 
MCAsmMacroParameter__anon13c8422b0111::MCAsmMacroParameter63   MCAsmMacroParameter() : Required(false), Vararg(false) {}
64 };
65 
66 typedef std::vector<MCAsmMacroParameter> MCAsmMacroParameters;
67 
68 struct MCAsmMacro {
69   StringRef Name;
70   StringRef Body;
71   MCAsmMacroParameters Parameters;
72 
73 public:
MCAsmMacro__anon13c8422b0111::MCAsmMacro74   MCAsmMacro(StringRef N, StringRef B, MCAsmMacroParameters P)
75       : Name(N), Body(B), Parameters(std::move(P)) {}
76 };
77 
78 /// \brief Helper class for storing information about an active macro
79 /// instantiation.
80 struct MacroInstantiation {
81   /// The location of the instantiation.
82   SMLoc InstantiationLoc;
83 
84   /// The buffer where parsing should resume upon instantiation completion.
85   int ExitBuffer;
86 
87   /// The location where parsing should resume upon instantiation completion.
88   SMLoc ExitLoc;
89 
90   /// The depth of TheCondStack at the start of the instantiation.
91   size_t CondStackDepth;
92 
93 public:
94   MacroInstantiation(SMLoc IL, int EB, SMLoc EL, size_t CondStackDepth);
95 };
96 
97 struct ParseStatementInfo {
98   /// \brief The parsed operands from the last parsed statement.
99   SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands;
100 
101   /// \brief The opcode from the last parsed instruction.
102   unsigned Opcode;
103 
104   /// \brief Was there an error parsing the inline assembly?
105   bool ParseError;
106 
107   SmallVectorImpl<AsmRewrite> *AsmRewrites;
108 
ParseStatementInfo__anon13c8422b0111::ParseStatementInfo109   ParseStatementInfo() : Opcode(~0U), ParseError(false), AsmRewrites(nullptr) {}
ParseStatementInfo__anon13c8422b0111::ParseStatementInfo110   ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
111     : Opcode(~0), ParseError(false), AsmRewrites(rewrites) {}
112 };
113 
114 /// \brief The concrete assembly parser instance.
115 class AsmParser : public MCAsmParser {
116   AsmParser(const AsmParser &) = delete;
117   void operator=(const AsmParser &) = delete;
118 private:
119   AsmLexer Lexer;
120   MCContext &Ctx;
121   MCStreamer &Out;
122   const MCAsmInfo &MAI;
123   SourceMgr &SrcMgr;
124   SourceMgr::DiagHandlerTy SavedDiagHandler;
125   void *SavedDiagContext;
126   std::unique_ptr<MCAsmParserExtension> PlatformParser;
127 
128   /// This is the current buffer index we're lexing from as managed by the
129   /// SourceMgr object.
130   unsigned CurBuffer;
131 
132   AsmCond TheCondState;
133   std::vector<AsmCond> TheCondStack;
134 
135   /// \brief maps directive names to handler methods in parser
136   /// extensions. Extensions register themselves in this map by calling
137   /// addDirectiveHandler.
138   StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap;
139 
140   /// \brief Map of currently defined macros.
141   StringMap<MCAsmMacro> MacroMap;
142 
143   /// \brief Stack of active macro instantiations.
144   std::vector<MacroInstantiation*> ActiveMacros;
145 
146   /// \brief List of bodies of anonymous macros.
147   std::deque<MCAsmMacro> MacroLikeBodies;
148 
149   /// Boolean tracking whether macro substitution is enabled.
150   unsigned MacrosEnabledFlag : 1;
151 
152   /// \brief Keeps track of how many .macro's have been instantiated.
153   unsigned NumOfMacroInstantiations;
154 
155   /// Flag tracking whether any errors have been encountered.
156   unsigned HadError : 1;
157 
158   /// The values from the last parsed cpp hash file line comment if any.
159   StringRef CppHashFilename;
160   int64_t CppHashLineNumber;
161   SMLoc CppHashLoc;
162   unsigned CppHashBuf;
163   /// When generating dwarf for assembly source files we need to calculate the
164   /// logical line number based on the last parsed cpp hash file line comment
165   /// and current line. Since this is slow and messes up the SourceMgr's
166   /// cache we save the last info we queried with SrcMgr.FindLineNumber().
167   SMLoc LastQueryIDLoc;
168   unsigned LastQueryBuffer;
169   unsigned LastQueryLine;
170 
171   /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
172   unsigned AssemblerDialect;
173 
174   /// \brief is Darwin compatibility enabled?
175   bool IsDarwin;
176 
177   /// \brief Are we parsing ms-style inline assembly?
178   bool ParsingInlineAsm;
179 
180 public:
181   AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
182             const MCAsmInfo &MAI);
183   ~AsmParser() override;
184 
185   bool Run(bool NoInitialTextSection, bool NoFinalize = false) override;
186 
addDirectiveHandler(StringRef Directive,ExtensionDirectiveHandler Handler)187   void addDirectiveHandler(StringRef Directive,
188                            ExtensionDirectiveHandler Handler) override {
189     ExtensionDirectiveMap[Directive] = Handler;
190   }
191 
addAliasForDirective(StringRef Directive,StringRef Alias)192   void addAliasForDirective(StringRef Directive, StringRef Alias) override {
193     DirectiveKindMap[Directive] = DirectiveKindMap[Alias];
194   }
195 
196 public:
197   /// @name MCAsmParser Interface
198   /// {
199 
getSourceManager()200   SourceMgr &getSourceManager() override { return SrcMgr; }
getLexer()201   MCAsmLexer &getLexer() override { return Lexer; }
getContext()202   MCContext &getContext() override { return Ctx; }
getStreamer()203   MCStreamer &getStreamer() override { return Out; }
getAssemblerDialect()204   unsigned getAssemblerDialect() override {
205     if (AssemblerDialect == ~0U)
206       return MAI.getAssemblerDialect();
207     else
208       return AssemblerDialect;
209   }
setAssemblerDialect(unsigned i)210   void setAssemblerDialect(unsigned i) override {
211     AssemblerDialect = i;
212   }
213 
214   void Note(SMLoc L, const Twine &Msg,
215             ArrayRef<SMRange> Ranges = None) override;
216   bool Warning(SMLoc L, const Twine &Msg,
217                ArrayRef<SMRange> Ranges = None) override;
218   bool Error(SMLoc L, const Twine &Msg,
219              ArrayRef<SMRange> Ranges = None) override;
220 
221   const AsmToken &Lex() override;
222 
setParsingInlineAsm(bool V)223   void setParsingInlineAsm(bool V) override { ParsingInlineAsm = V; }
isParsingInlineAsm()224   bool isParsingInlineAsm() override { return ParsingInlineAsm; }
225 
226   bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString,
227                         unsigned &NumOutputs, unsigned &NumInputs,
228                         SmallVectorImpl<std::pair<void *,bool> > &OpDecls,
229                         SmallVectorImpl<std::string> &Constraints,
230                         SmallVectorImpl<std::string> &Clobbers,
231                         const MCInstrInfo *MII, const MCInstPrinter *IP,
232                         MCAsmParserSemaCallback &SI) override;
233 
234   bool parseExpression(const MCExpr *&Res);
235   bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
236   bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override;
237   bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
238   bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
239                              SMLoc &EndLoc) override;
240   bool parseAbsoluteExpression(int64_t &Res) override;
241 
242   /// \brief Parse an identifier or string (as a quoted identifier)
243   /// and set \p Res to the identifier contents.
244   bool parseIdentifier(StringRef &Res) override;
245   void eatToEndOfStatement() override;
246 
247   void checkForValidSection() override;
248   /// }
249 
250 private:
251 
252   bool parseStatement(ParseStatementInfo &Info,
253                       MCAsmParserSemaCallback *SI);
254   void eatToEndOfLine();
255   bool parseCppHashLineFilenameComment(SMLoc L);
256 
257   void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body,
258                         ArrayRef<MCAsmMacroParameter> Parameters);
259   bool expandMacro(raw_svector_ostream &OS, StringRef Body,
260                    ArrayRef<MCAsmMacroParameter> Parameters,
261                    ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable,
262                    SMLoc L);
263 
264   /// \brief Are macros enabled in the parser?
areMacrosEnabled()265   bool areMacrosEnabled() {return MacrosEnabledFlag;}
266 
267   /// \brief Control a flag in the parser that enables or disables macros.
setMacrosEnabled(bool Flag)268   void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;}
269 
270   /// \brief Lookup a previously defined macro.
271   /// \param Name Macro name.
272   /// \returns Pointer to macro. NULL if no such macro was defined.
273   const MCAsmMacro* lookupMacro(StringRef Name);
274 
275   /// \brief Define a new macro with the given name and information.
276   void defineMacro(StringRef Name, MCAsmMacro Macro);
277 
278   /// \brief Undefine a macro. If no such macro was defined, it's a no-op.
279   void undefineMacro(StringRef Name);
280 
281   /// \brief Are we inside a macro instantiation?
isInsideMacroInstantiation()282   bool isInsideMacroInstantiation() {return !ActiveMacros.empty();}
283 
284   /// \brief Handle entry to macro instantiation.
285   ///
286   /// \param M The macro.
287   /// \param NameLoc Instantiation location.
288   bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc);
289 
290   /// \brief Handle exit from macro instantiation.
291   void handleMacroExit();
292 
293   /// \brief Extract AsmTokens for a macro argument.
294   bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg);
295 
296   /// \brief Parse all macro arguments for a given macro.
297   bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A);
298 
299   void printMacroInstantiations();
printMessage(SMLoc Loc,SourceMgr::DiagKind Kind,const Twine & Msg,ArrayRef<SMRange> Ranges=None) const300   void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
301                     ArrayRef<SMRange> Ranges = None) const {
302     SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
303   }
304   static void DiagHandler(const SMDiagnostic &Diag, void *Context);
305 
306   /// \brief Enter the specified file. This returns true on failure.
307   bool enterIncludeFile(const std::string &Filename);
308 
309   /// \brief Process the specified file for the .incbin directive.
310   /// This returns true on failure.
311   bool processIncbinFile(const std::string &Filename);
312 
313   /// \brief Reset the current lexer position to that given by \p Loc. The
314   /// current token is not set; clients should ensure Lex() is called
315   /// subsequently.
316   ///
317   /// \param InBuffer If not 0, should be the known buffer id that contains the
318   /// location.
319   void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0);
320 
321   /// \brief Parse up to the end of statement and a return the contents from the
322   /// current token until the end of the statement; the current token on exit
323   /// will be either the EndOfStatement or EOF.
324   StringRef parseStringToEndOfStatement() override;
325 
326   /// \brief Parse until the end of a statement or a comma is encountered,
327   /// return the contents from the current token up to the end or comma.
328   StringRef parseStringToComma();
329 
330   bool parseAssignment(StringRef Name, bool allow_redef,
331                        bool NoDeadStrip = false);
332 
333   unsigned getBinOpPrecedence(AsmToken::TokenKind K,
334                               MCBinaryExpr::Opcode &Kind);
335 
336   bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
337   bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
338   bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
339 
340   bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
341 
342   // Generic (target and platform independent) directive parsing.
343   enum DirectiveKind {
344     DK_NO_DIRECTIVE, // Placeholder
345     DK_SET, DK_EQU, DK_EQUIV, DK_ASCII, DK_ASCIZ, DK_STRING, DK_BYTE, DK_SHORT,
346     DK_RELOC,
347     DK_VALUE, DK_2BYTE, DK_LONG, DK_INT, DK_4BYTE, DK_QUAD, DK_8BYTE, DK_OCTA,
348     DK_SINGLE, DK_FLOAT, DK_DOUBLE, DK_ALIGN, DK_ALIGN32, DK_BALIGN, DK_BALIGNW,
349     DK_BALIGNL, DK_P2ALIGN, DK_P2ALIGNW, DK_P2ALIGNL, DK_ORG, DK_FILL, DK_ENDR,
350     DK_BUNDLE_ALIGN_MODE, DK_BUNDLE_LOCK, DK_BUNDLE_UNLOCK,
351     DK_ZERO, DK_EXTERN, DK_GLOBL, DK_GLOBAL,
352     DK_LAZY_REFERENCE, DK_NO_DEAD_STRIP, DK_SYMBOL_RESOLVER, DK_PRIVATE_EXTERN,
353     DK_REFERENCE, DK_WEAK_DEFINITION, DK_WEAK_REFERENCE,
354     DK_WEAK_DEF_CAN_BE_HIDDEN, DK_COMM, DK_COMMON, DK_LCOMM, DK_ABORT,
355     DK_INCLUDE, DK_INCBIN, DK_CODE16, DK_CODE16GCC, DK_REPT, DK_IRP, DK_IRPC,
356     DK_IF, DK_IFEQ, DK_IFGE, DK_IFGT, DK_IFLE, DK_IFLT, DK_IFNE, DK_IFB,
357     DK_IFNB, DK_IFC, DK_IFEQS, DK_IFNC, DK_IFNES, DK_IFDEF, DK_IFNDEF,
358     DK_IFNOTDEF, DK_ELSEIF, DK_ELSE, DK_ENDIF,
359     DK_SPACE, DK_SKIP, DK_FILE, DK_LINE, DK_LOC, DK_STABS,
360     DK_CFI_SECTIONS, DK_CFI_STARTPROC, DK_CFI_ENDPROC, DK_CFI_DEF_CFA,
361     DK_CFI_DEF_CFA_OFFSET, DK_CFI_ADJUST_CFA_OFFSET, DK_CFI_DEF_CFA_REGISTER,
362     DK_CFI_OFFSET, DK_CFI_REL_OFFSET, DK_CFI_PERSONALITY, DK_CFI_LSDA,
363     DK_CFI_REMEMBER_STATE, DK_CFI_RESTORE_STATE, DK_CFI_SAME_VALUE,
364     DK_CFI_RESTORE, DK_CFI_ESCAPE, DK_CFI_SIGNAL_FRAME, DK_CFI_UNDEFINED,
365     DK_CFI_REGISTER, DK_CFI_WINDOW_SAVE,
366     DK_MACROS_ON, DK_MACROS_OFF,
367     DK_MACRO, DK_EXITM, DK_ENDM, DK_ENDMACRO, DK_PURGEM,
368     DK_SLEB128, DK_ULEB128,
369     DK_ERR, DK_ERROR, DK_WARNING,
370     DK_END
371   };
372 
373   /// \brief Maps directive name --> DirectiveKind enum, for
374   /// directives parsed by this class.
375   StringMap<DirectiveKind> DirectiveKindMap;
376 
377   // ".ascii", ".asciz", ".string"
378   bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
379   bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc"
380   bool parseDirectiveValue(unsigned Size); // ".byte", ".long", ...
381   bool parseDirectiveOctaValue(); // ".octa"
382   bool parseDirectiveRealValue(const fltSemantics &); // ".single", ...
383   bool parseDirectiveFill(); // ".fill"
384   bool parseDirectiveZero(); // ".zero"
385   // ".set", ".equ", ".equiv"
386   bool parseDirectiveSet(StringRef IDVal, bool allow_redef);
387   bool parseDirectiveOrg(); // ".org"
388   // ".align{,32}", ".p2align{,w,l}"
389   bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize);
390 
391   // ".file", ".line", ".loc", ".stabs"
392   bool parseDirectiveFile(SMLoc DirectiveLoc);
393   bool parseDirectiveLine();
394   bool parseDirectiveLoc();
395   bool parseDirectiveStabs();
396 
397   // .cfi directives
398   bool parseDirectiveCFIRegister(SMLoc DirectiveLoc);
399   bool parseDirectiveCFIWindowSave();
400   bool parseDirectiveCFISections();
401   bool parseDirectiveCFIStartProc();
402   bool parseDirectiveCFIEndProc();
403   bool parseDirectiveCFIDefCfaOffset();
404   bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc);
405   bool parseDirectiveCFIAdjustCfaOffset();
406   bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc);
407   bool parseDirectiveCFIOffset(SMLoc DirectiveLoc);
408   bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc);
409   bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality);
410   bool parseDirectiveCFIRememberState();
411   bool parseDirectiveCFIRestoreState();
412   bool parseDirectiveCFISameValue(SMLoc DirectiveLoc);
413   bool parseDirectiveCFIRestore(SMLoc DirectiveLoc);
414   bool parseDirectiveCFIEscape();
415   bool parseDirectiveCFISignalFrame();
416   bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc);
417 
418   // macro directives
419   bool parseDirectivePurgeMacro(SMLoc DirectiveLoc);
420   bool parseDirectiveExitMacro(StringRef Directive);
421   bool parseDirectiveEndMacro(StringRef Directive);
422   bool parseDirectiveMacro(SMLoc DirectiveLoc);
423   bool parseDirectiveMacrosOnOff(StringRef Directive);
424 
425   // ".bundle_align_mode"
426   bool parseDirectiveBundleAlignMode();
427   // ".bundle_lock"
428   bool parseDirectiveBundleLock();
429   // ".bundle_unlock"
430   bool parseDirectiveBundleUnlock();
431 
432   // ".space", ".skip"
433   bool parseDirectiveSpace(StringRef IDVal);
434 
435   // .sleb128 (Signed=true) and .uleb128 (Signed=false)
436   bool parseDirectiveLEB128(bool Signed);
437 
438   /// \brief Parse a directive like ".globl" which
439   /// accepts a single symbol (which should be a label or an external).
440   bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr);
441 
442   bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
443 
444   bool parseDirectiveAbort(); // ".abort"
445   bool parseDirectiveInclude(); // ".include"
446   bool parseDirectiveIncbin(); // ".incbin"
447 
448   // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne"
449   bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
450   // ".ifb" or ".ifnb", depending on ExpectBlank.
451   bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
452   // ".ifc" or ".ifnc", depending on ExpectEqual.
453   bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
454   // ".ifeqs" or ".ifnes", depending on ExpectEqual.
455   bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual);
456   // ".ifdef" or ".ifndef", depending on expect_defined
457   bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
458   bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
459   bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else"
460   bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
461   bool parseEscapedString(std::string &Data) override;
462 
463   const MCExpr *applyModifierToExpr(const MCExpr *E,
464                                     MCSymbolRefExpr::VariantKind Variant);
465 
466   // Macro-like directives
467   MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
468   void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
469                                 raw_svector_ostream &OS);
470   bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive);
471   bool parseDirectiveIrp(SMLoc DirectiveLoc);  // ".irp"
472   bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
473   bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
474 
475   // "_emit" or "__emit"
476   bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info,
477                             size_t Len);
478 
479   // "align"
480   bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info);
481 
482   // "end"
483   bool parseDirectiveEnd(SMLoc DirectiveLoc);
484 
485   // ".err" or ".error"
486   bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage);
487 
488   // ".warning"
489   bool parseDirectiveWarning(SMLoc DirectiveLoc);
490 
491   void initializeDirectiveKindMap();
492 };
493 }
494 
495 namespace llvm {
496 
497 extern MCAsmParserExtension *createDarwinAsmParser();
498 extern MCAsmParserExtension *createELFAsmParser();
499 extern MCAsmParserExtension *createCOFFAsmParser();
500 
501 }
502 
503 enum { DEFAULT_ADDRSPACE = 0 };
504 
AsmParser(SourceMgr & SM,MCContext & Ctx,MCStreamer & Out,const MCAsmInfo & MAI)505 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
506                      const MCAsmInfo &MAI)
507     : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM),
508       PlatformParser(nullptr), CurBuffer(SM.getMainFileID()),
509       MacrosEnabledFlag(true), HadError(false), CppHashLineNumber(0),
510       AssemblerDialect(~0U), IsDarwin(false), ParsingInlineAsm(false) {
511   // Save the old handler.
512   SavedDiagHandler = SrcMgr.getDiagHandler();
513   SavedDiagContext = SrcMgr.getDiagContext();
514   // Set our own handler which calls the saved handler.
515   SrcMgr.setDiagHandler(DiagHandler, this);
516   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
517 
518   // Initialize the platform / file format parser.
519   switch (Ctx.getObjectFileInfo()->getObjectFileType()) {
520   case MCObjectFileInfo::IsCOFF:
521     PlatformParser.reset(createCOFFAsmParser());
522     break;
523   case MCObjectFileInfo::IsMachO:
524     PlatformParser.reset(createDarwinAsmParser());
525     IsDarwin = true;
526     break;
527   case MCObjectFileInfo::IsELF:
528     PlatformParser.reset(createELFAsmParser());
529     break;
530   }
531 
532   PlatformParser->Initialize(*this);
533   initializeDirectiveKindMap();
534 
535   NumOfMacroInstantiations = 0;
536 }
537 
~AsmParser()538 AsmParser::~AsmParser() {
539   assert((HadError || ActiveMacros.empty()) &&
540          "Unexpected active macro instantiation!");
541 }
542 
printMacroInstantiations()543 void AsmParser::printMacroInstantiations() {
544   // Print the active macro instantiation stack.
545   for (std::vector<MacroInstantiation *>::const_reverse_iterator
546            it = ActiveMacros.rbegin(),
547            ie = ActiveMacros.rend();
548        it != ie; ++it)
549     printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
550                  "while in macro instantiation");
551 }
552 
Note(SMLoc L,const Twine & Msg,ArrayRef<SMRange> Ranges)553 void AsmParser::Note(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
554   printMessage(L, SourceMgr::DK_Note, Msg, Ranges);
555   printMacroInstantiations();
556 }
557 
Warning(SMLoc L,const Twine & Msg,ArrayRef<SMRange> Ranges)558 bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
559   if(getTargetParser().getTargetOptions().MCNoWarn)
560     return false;
561   if (getTargetParser().getTargetOptions().MCFatalWarnings)
562     return Error(L, Msg, Ranges);
563   printMessage(L, SourceMgr::DK_Warning, Msg, Ranges);
564   printMacroInstantiations();
565   return false;
566 }
567 
Error(SMLoc L,const Twine & Msg,ArrayRef<SMRange> Ranges)568 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
569   HadError = true;
570   printMessage(L, SourceMgr::DK_Error, Msg, Ranges);
571   printMacroInstantiations();
572   return true;
573 }
574 
enterIncludeFile(const std::string & Filename)575 bool AsmParser::enterIncludeFile(const std::string &Filename) {
576   std::string IncludedFile;
577   unsigned NewBuf =
578       SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
579   if (!NewBuf)
580     return true;
581 
582   CurBuffer = NewBuf;
583   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
584   return false;
585 }
586 
587 /// Process the specified .incbin file by searching for it in the include paths
588 /// then just emitting the byte contents of the file to the streamer. This
589 /// returns true on failure.
processIncbinFile(const std::string & Filename)590 bool AsmParser::processIncbinFile(const std::string &Filename) {
591   std::string IncludedFile;
592   unsigned NewBuf =
593       SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
594   if (!NewBuf)
595     return true;
596 
597   // Pick up the bytes from the file and emit them.
598   getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer());
599   return false;
600 }
601 
jumpToLoc(SMLoc Loc,unsigned InBuffer)602 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) {
603   CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc);
604   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(),
605                   Loc.getPointer());
606 }
607 
Lex()608 const AsmToken &AsmParser::Lex() {
609   const AsmToken *tok = &Lexer.Lex();
610 
611   if (tok->is(AsmToken::Eof)) {
612     // If this is the end of an included file, pop the parent file off the
613     // include stack.
614     SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
615     if (ParentIncludeLoc != SMLoc()) {
616       jumpToLoc(ParentIncludeLoc);
617       tok = &Lexer.Lex();
618     }
619   }
620 
621   if (tok->is(AsmToken::Error))
622     Error(Lexer.getErrLoc(), Lexer.getErr());
623 
624   return *tok;
625 }
626 
Run(bool NoInitialTextSection,bool NoFinalize)627 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
628   // Create the initial section, if requested.
629   if (!NoInitialTextSection)
630     Out.InitSections(false);
631 
632   // Prime the lexer.
633   Lex();
634 
635   HadError = false;
636   AsmCond StartingCondState = TheCondState;
637 
638   // If we are generating dwarf for assembly source files save the initial text
639   // section and generate a .file directive.
640   if (getContext().getGenDwarfForAssembly()) {
641     MCSection *Sec = getStreamer().getCurrentSection().first;
642     if (!Sec->getBeginSymbol()) {
643       MCSymbol *SectionStartSym = getContext().createTempSymbol();
644       getStreamer().EmitLabel(SectionStartSym);
645       Sec->setBeginSymbol(SectionStartSym);
646     }
647     bool InsertResult = getContext().addGenDwarfSection(Sec);
648     assert(InsertResult && ".text section should not have debug info yet");
649     (void)InsertResult;
650     getContext().setGenDwarfFileNumber(getStreamer().EmitDwarfFileDirective(
651         0, StringRef(), getContext().getMainFileName()));
652   }
653 
654   // While we have input, parse each statement.
655   while (Lexer.isNot(AsmToken::Eof)) {
656     ParseStatementInfo Info;
657     if (!parseStatement(Info, nullptr))
658       continue;
659 
660     // We had an error, validate that one was emitted and recover by skipping to
661     // the next line.
662     assert(HadError && "Parse statement returned an error, but none emitted!");
663     eatToEndOfStatement();
664   }
665 
666   if (TheCondState.TheCond != StartingCondState.TheCond ||
667       TheCondState.Ignore != StartingCondState.Ignore)
668     return TokError("unmatched .ifs or .elses");
669 
670   // Check to see there are no empty DwarfFile slots.
671   const auto &LineTables = getContext().getMCDwarfLineTables();
672   if (!LineTables.empty()) {
673     unsigned Index = 0;
674     for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) {
675       if (File.Name.empty() && Index != 0)
676         TokError("unassigned file number: " + Twine(Index) +
677                  " for .file directives");
678       ++Index;
679     }
680   }
681 
682   // Check to see that all assembler local symbols were actually defined.
683   // Targets that don't do subsections via symbols may not want this, though,
684   // so conservatively exclude them. Only do this if we're finalizing, though,
685   // as otherwise we won't necessarilly have seen everything yet.
686   if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) {
687     for (const auto &TableEntry : getContext().getSymbols()) {
688       MCSymbol *Sym = TableEntry.getValue();
689       // Variable symbols may not be marked as defined, so check those
690       // explicitly. If we know it's a variable, we have a definition for
691       // the purposes of this check.
692       if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
693         // FIXME: We would really like to refer back to where the symbol was
694         // first referenced for a source location. We need to add something
695         // to track that. Currently, we just point to the end of the file.
696         return Error(getLexer().getLoc(), "assembler local symbol '" +
697                                               Sym->getName() + "' not defined");
698     }
699   }
700 
701   // Finalize the output stream if there are no errors and if the client wants
702   // us to.
703   if (!HadError && !NoFinalize)
704     Out.Finish();
705 
706   return HadError || getContext().hadError();
707 }
708 
checkForValidSection()709 void AsmParser::checkForValidSection() {
710   if (!ParsingInlineAsm && !getStreamer().getCurrentSection().first) {
711     TokError("expected section directive before assembly directive");
712     Out.InitSections(false);
713   }
714 }
715 
716 /// \brief Throw away the rest of the line for testing purposes.
eatToEndOfStatement()717 void AsmParser::eatToEndOfStatement() {
718   while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
719     Lex();
720 
721   // Eat EOL.
722   if (Lexer.is(AsmToken::EndOfStatement))
723     Lex();
724 }
725 
parseStringToEndOfStatement()726 StringRef AsmParser::parseStringToEndOfStatement() {
727   const char *Start = getTok().getLoc().getPointer();
728 
729   while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
730     Lex();
731 
732   const char *End = getTok().getLoc().getPointer();
733   return StringRef(Start, End - Start);
734 }
735 
parseStringToComma()736 StringRef AsmParser::parseStringToComma() {
737   const char *Start = getTok().getLoc().getPointer();
738 
739   while (Lexer.isNot(AsmToken::EndOfStatement) &&
740          Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof))
741     Lex();
742 
743   const char *End = getTok().getLoc().getPointer();
744   return StringRef(Start, End - Start);
745 }
746 
747 /// \brief Parse a paren expression and return it.
748 /// NOTE: This assumes the leading '(' has already been consumed.
749 ///
750 /// parenexpr ::= expr)
751 ///
parseParenExpr(const MCExpr * & Res,SMLoc & EndLoc)752 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
753   if (parseExpression(Res))
754     return true;
755   if (Lexer.isNot(AsmToken::RParen))
756     return TokError("expected ')' in parentheses expression");
757   EndLoc = Lexer.getTok().getEndLoc();
758   Lex();
759   return false;
760 }
761 
762 /// \brief Parse a bracket expression and return it.
763 /// NOTE: This assumes the leading '[' has already been consumed.
764 ///
765 /// bracketexpr ::= expr]
766 ///
parseBracketExpr(const MCExpr * & Res,SMLoc & EndLoc)767 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
768   if (parseExpression(Res))
769     return true;
770   if (Lexer.isNot(AsmToken::RBrac))
771     return TokError("expected ']' in brackets expression");
772   EndLoc = Lexer.getTok().getEndLoc();
773   Lex();
774   return false;
775 }
776 
777 /// \brief Parse a primary expression and return it.
778 ///  primaryexpr ::= (parenexpr
779 ///  primaryexpr ::= symbol
780 ///  primaryexpr ::= number
781 ///  primaryexpr ::= '.'
782 ///  primaryexpr ::= ~,+,- primaryexpr
parsePrimaryExpr(const MCExpr * & Res,SMLoc & EndLoc)783 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
784   SMLoc FirstTokenLoc = getLexer().getLoc();
785   AsmToken::TokenKind FirstTokenKind = Lexer.getKind();
786   switch (FirstTokenKind) {
787   default:
788     return TokError("unknown token in expression");
789   // If we have an error assume that we've already handled it.
790   case AsmToken::Error:
791     return true;
792   case AsmToken::Exclaim:
793     Lex(); // Eat the operator.
794     if (parsePrimaryExpr(Res, EndLoc))
795       return true;
796     Res = MCUnaryExpr::createLNot(Res, getContext());
797     return false;
798   case AsmToken::Dollar:
799   case AsmToken::At:
800   case AsmToken::String:
801   case AsmToken::Identifier: {
802     StringRef Identifier;
803     if (parseIdentifier(Identifier)) {
804       if (FirstTokenKind == AsmToken::Dollar) {
805         if (Lexer.getMAI().getDollarIsPC()) {
806           // This is a '$' reference, which references the current PC.  Emit a
807           // temporary label to the streamer and refer to it.
808           MCSymbol *Sym = Ctx.createTempSymbol();
809           Out.EmitLabel(Sym);
810           Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None,
811                                         getContext());
812           EndLoc = FirstTokenLoc;
813           return false;
814         }
815         return Error(FirstTokenLoc, "invalid token in expression");
816       }
817     }
818     // Parse symbol variant
819     std::pair<StringRef, StringRef> Split;
820     if (!MAI.useParensForSymbolVariant()) {
821       if (FirstTokenKind == AsmToken::String) {
822         if (Lexer.is(AsmToken::At)) {
823           Lexer.Lex(); // eat @
824           SMLoc AtLoc = getLexer().getLoc();
825           StringRef VName;
826           if (parseIdentifier(VName))
827             return Error(AtLoc, "expected symbol variant after '@'");
828 
829           Split = std::make_pair(Identifier, VName);
830         }
831       } else {
832         Split = Identifier.split('@');
833       }
834     } else if (Lexer.is(AsmToken::LParen)) {
835       Lexer.Lex(); // eat (
836       StringRef VName;
837       parseIdentifier(VName);
838       if (Lexer.isNot(AsmToken::RParen)) {
839           return Error(Lexer.getTok().getLoc(),
840                        "unexpected token in variant, expected ')'");
841       }
842       Lexer.Lex(); // eat )
843       Split = std::make_pair(Identifier, VName);
844     }
845 
846     EndLoc = SMLoc::getFromPointer(Identifier.end());
847 
848     // This is a symbol reference.
849     StringRef SymbolName = Identifier;
850     MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
851 
852     // Lookup the symbol variant if used.
853     if (Split.second.size()) {
854       Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
855       if (Variant != MCSymbolRefExpr::VK_Invalid) {
856         SymbolName = Split.first;
857       } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) {
858         Variant = MCSymbolRefExpr::VK_None;
859       } else {
860         return Error(SMLoc::getFromPointer(Split.second.begin()),
861                      "invalid variant '" + Split.second + "'");
862       }
863     }
864 
865     MCSymbol *Sym = getContext().getOrCreateSymbol(SymbolName);
866 
867     // If this is an absolute variable reference, substitute it now to preserve
868     // semantics in the face of reassignment.
869     if (Sym->isVariable() &&
870         isa<MCConstantExpr>(Sym->getVariableValue(/*SetUsed*/ false))) {
871       if (Variant)
872         return Error(EndLoc, "unexpected modifier on variable reference");
873 
874       Res = Sym->getVariableValue(/*SetUsed*/ false);
875       return false;
876     }
877 
878     // Otherwise create a symbol ref.
879     Res = MCSymbolRefExpr::create(Sym, Variant, getContext());
880     return false;
881   }
882   case AsmToken::BigNum:
883     return TokError("literal value out of range for directive");
884   case AsmToken::Integer: {
885     SMLoc Loc = getTok().getLoc();
886     int64_t IntVal = getTok().getIntVal();
887     Res = MCConstantExpr::create(IntVal, getContext());
888     EndLoc = Lexer.getTok().getEndLoc();
889     Lex(); // Eat token.
890     // Look for 'b' or 'f' following an Integer as a directional label
891     if (Lexer.getKind() == AsmToken::Identifier) {
892       StringRef IDVal = getTok().getString();
893       // Lookup the symbol variant if used.
894       std::pair<StringRef, StringRef> Split = IDVal.split('@');
895       MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
896       if (Split.first.size() != IDVal.size()) {
897         Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
898         if (Variant == MCSymbolRefExpr::VK_Invalid)
899           return TokError("invalid variant '" + Split.second + "'");
900         IDVal = Split.first;
901       }
902       if (IDVal == "f" || IDVal == "b") {
903         MCSymbol *Sym =
904             Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b");
905         Res = MCSymbolRefExpr::create(Sym, Variant, getContext());
906         if (IDVal == "b" && Sym->isUndefined())
907           return Error(Loc, "invalid reference to undefined symbol");
908         EndLoc = Lexer.getTok().getEndLoc();
909         Lex(); // Eat identifier.
910       }
911     }
912     return false;
913   }
914   case AsmToken::Real: {
915     APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
916     uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
917     Res = MCConstantExpr::create(IntVal, getContext());
918     EndLoc = Lexer.getTok().getEndLoc();
919     Lex(); // Eat token.
920     return false;
921   }
922   case AsmToken::Dot: {
923     // This is a '.' reference, which references the current PC.  Emit a
924     // temporary label to the streamer and refer to it.
925     MCSymbol *Sym = Ctx.createTempSymbol();
926     Out.EmitLabel(Sym);
927     Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext());
928     EndLoc = Lexer.getTok().getEndLoc();
929     Lex(); // Eat identifier.
930     return false;
931   }
932   case AsmToken::LParen:
933     Lex(); // Eat the '('.
934     return parseParenExpr(Res, EndLoc);
935   case AsmToken::LBrac:
936     if (!PlatformParser->HasBracketExpressions())
937       return TokError("brackets expression not supported on this target");
938     Lex(); // Eat the '['.
939     return parseBracketExpr(Res, EndLoc);
940   case AsmToken::Minus:
941     Lex(); // Eat the operator.
942     if (parsePrimaryExpr(Res, EndLoc))
943       return true;
944     Res = MCUnaryExpr::createMinus(Res, getContext());
945     return false;
946   case AsmToken::Plus:
947     Lex(); // Eat the operator.
948     if (parsePrimaryExpr(Res, EndLoc))
949       return true;
950     Res = MCUnaryExpr::createPlus(Res, getContext());
951     return false;
952   case AsmToken::Tilde:
953     Lex(); // Eat the operator.
954     if (parsePrimaryExpr(Res, EndLoc))
955       return true;
956     Res = MCUnaryExpr::createNot(Res, getContext());
957     return false;
958   }
959 }
960 
parseExpression(const MCExpr * & Res)961 bool AsmParser::parseExpression(const MCExpr *&Res) {
962   SMLoc EndLoc;
963   return parseExpression(Res, EndLoc);
964 }
965 
966 const MCExpr *
applyModifierToExpr(const MCExpr * E,MCSymbolRefExpr::VariantKind Variant)967 AsmParser::applyModifierToExpr(const MCExpr *E,
968                                MCSymbolRefExpr::VariantKind Variant) {
969   // Ask the target implementation about this expression first.
970   const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx);
971   if (NewE)
972     return NewE;
973   // Recurse over the given expression, rebuilding it to apply the given variant
974   // if there is exactly one symbol.
975   switch (E->getKind()) {
976   case MCExpr::Target:
977   case MCExpr::Constant:
978     return nullptr;
979 
980   case MCExpr::SymbolRef: {
981     const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
982 
983     if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
984       TokError("invalid variant on expression '" + getTok().getIdentifier() +
985                "' (already modified)");
986       return E;
987     }
988 
989     return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext());
990   }
991 
992   case MCExpr::Unary: {
993     const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
994     const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant);
995     if (!Sub)
996       return nullptr;
997     return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext());
998   }
999 
1000   case MCExpr::Binary: {
1001     const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
1002     const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant);
1003     const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant);
1004 
1005     if (!LHS && !RHS)
1006       return nullptr;
1007 
1008     if (!LHS)
1009       LHS = BE->getLHS();
1010     if (!RHS)
1011       RHS = BE->getRHS();
1012 
1013     return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext());
1014   }
1015   }
1016 
1017   llvm_unreachable("Invalid expression kind!");
1018 }
1019 
1020 /// \brief Parse an expression and return it.
1021 ///
1022 ///  expr ::= expr &&,|| expr               -> lowest.
1023 ///  expr ::= expr |,^,&,! expr
1024 ///  expr ::= expr ==,!=,<>,<,<=,>,>= expr
1025 ///  expr ::= expr <<,>> expr
1026 ///  expr ::= expr +,- expr
1027 ///  expr ::= expr *,/,% expr               -> highest.
1028 ///  expr ::= primaryexpr
1029 ///
parseExpression(const MCExpr * & Res,SMLoc & EndLoc)1030 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1031   // Parse the expression.
1032   Res = nullptr;
1033   if (parsePrimaryExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc))
1034     return true;
1035 
1036   // As a special case, we support 'a op b @ modifier' by rewriting the
1037   // expression to include the modifier. This is inefficient, but in general we
1038   // expect users to use 'a@modifier op b'.
1039   if (Lexer.getKind() == AsmToken::At) {
1040     Lex();
1041 
1042     if (Lexer.isNot(AsmToken::Identifier))
1043       return TokError("unexpected symbol modifier following '@'");
1044 
1045     MCSymbolRefExpr::VariantKind Variant =
1046         MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
1047     if (Variant == MCSymbolRefExpr::VK_Invalid)
1048       return TokError("invalid variant '" + getTok().getIdentifier() + "'");
1049 
1050     const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant);
1051     if (!ModifiedRes) {
1052       return TokError("invalid modifier '" + getTok().getIdentifier() +
1053                       "' (no symbols present)");
1054     }
1055 
1056     Res = ModifiedRes;
1057     Lex();
1058   }
1059 
1060   // Try to constant fold it up front, if possible.
1061   int64_t Value;
1062   if (Res->evaluateAsAbsolute(Value))
1063     Res = MCConstantExpr::create(Value, getContext());
1064 
1065   return false;
1066 }
1067 
parseParenExpression(const MCExpr * & Res,SMLoc & EndLoc)1068 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1069   Res = nullptr;
1070   return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc);
1071 }
1072 
parseParenExprOfDepth(unsigned ParenDepth,const MCExpr * & Res,SMLoc & EndLoc)1073 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
1074                                       SMLoc &EndLoc) {
1075   if (parseParenExpr(Res, EndLoc))
1076     return true;
1077 
1078   for (; ParenDepth > 0; --ParenDepth) {
1079     if (parseBinOpRHS(1, Res, EndLoc))
1080       return true;
1081 
1082     // We don't Lex() the last RParen.
1083     // This is the same behavior as parseParenExpression().
1084     if (ParenDepth - 1 > 0) {
1085       if (Lexer.isNot(AsmToken::RParen))
1086         return TokError("expected ')' in parentheses expression");
1087       EndLoc = Lexer.getTok().getEndLoc();
1088       Lex();
1089     }
1090   }
1091   return false;
1092 }
1093 
parseAbsoluteExpression(int64_t & Res)1094 bool AsmParser::parseAbsoluteExpression(int64_t &Res) {
1095   const MCExpr *Expr;
1096 
1097   SMLoc StartLoc = Lexer.getLoc();
1098   if (parseExpression(Expr))
1099     return true;
1100 
1101   if (!Expr->evaluateAsAbsolute(Res))
1102     return Error(StartLoc, "expected absolute expression");
1103 
1104   return false;
1105 }
1106 
getDarwinBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1107 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K,
1108                                          MCBinaryExpr::Opcode &Kind,
1109                                          bool ShouldUseLogicalShr) {
1110   switch (K) {
1111   default:
1112     return 0; // not a binop.
1113 
1114   // Lowest Precedence: &&, ||
1115   case AsmToken::AmpAmp:
1116     Kind = MCBinaryExpr::LAnd;
1117     return 1;
1118   case AsmToken::PipePipe:
1119     Kind = MCBinaryExpr::LOr;
1120     return 1;
1121 
1122   // Low Precedence: |, &, ^
1123   //
1124   // FIXME: gas seems to support '!' as an infix operator?
1125   case AsmToken::Pipe:
1126     Kind = MCBinaryExpr::Or;
1127     return 2;
1128   case AsmToken::Caret:
1129     Kind = MCBinaryExpr::Xor;
1130     return 2;
1131   case AsmToken::Amp:
1132     Kind = MCBinaryExpr::And;
1133     return 2;
1134 
1135   // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
1136   case AsmToken::EqualEqual:
1137     Kind = MCBinaryExpr::EQ;
1138     return 3;
1139   case AsmToken::ExclaimEqual:
1140   case AsmToken::LessGreater:
1141     Kind = MCBinaryExpr::NE;
1142     return 3;
1143   case AsmToken::Less:
1144     Kind = MCBinaryExpr::LT;
1145     return 3;
1146   case AsmToken::LessEqual:
1147     Kind = MCBinaryExpr::LTE;
1148     return 3;
1149   case AsmToken::Greater:
1150     Kind = MCBinaryExpr::GT;
1151     return 3;
1152   case AsmToken::GreaterEqual:
1153     Kind = MCBinaryExpr::GTE;
1154     return 3;
1155 
1156   // Intermediate Precedence: <<, >>
1157   case AsmToken::LessLess:
1158     Kind = MCBinaryExpr::Shl;
1159     return 4;
1160   case AsmToken::GreaterGreater:
1161     Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1162     return 4;
1163 
1164   // High Intermediate Precedence: +, -
1165   case AsmToken::Plus:
1166     Kind = MCBinaryExpr::Add;
1167     return 5;
1168   case AsmToken::Minus:
1169     Kind = MCBinaryExpr::Sub;
1170     return 5;
1171 
1172   // Highest Precedence: *, /, %
1173   case AsmToken::Star:
1174     Kind = MCBinaryExpr::Mul;
1175     return 6;
1176   case AsmToken::Slash:
1177     Kind = MCBinaryExpr::Div;
1178     return 6;
1179   case AsmToken::Percent:
1180     Kind = MCBinaryExpr::Mod;
1181     return 6;
1182   }
1183 }
1184 
getGNUBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1185 static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K,
1186                                       MCBinaryExpr::Opcode &Kind,
1187                                       bool ShouldUseLogicalShr) {
1188   switch (K) {
1189   default:
1190     return 0; // not a binop.
1191 
1192   // Lowest Precedence: &&, ||
1193   case AsmToken::AmpAmp:
1194     Kind = MCBinaryExpr::LAnd;
1195     return 2;
1196   case AsmToken::PipePipe:
1197     Kind = MCBinaryExpr::LOr;
1198     return 1;
1199 
1200   // Low Precedence: ==, !=, <>, <, <=, >, >=
1201   case AsmToken::EqualEqual:
1202     Kind = MCBinaryExpr::EQ;
1203     return 3;
1204   case AsmToken::ExclaimEqual:
1205   case AsmToken::LessGreater:
1206     Kind = MCBinaryExpr::NE;
1207     return 3;
1208   case AsmToken::Less:
1209     Kind = MCBinaryExpr::LT;
1210     return 3;
1211   case AsmToken::LessEqual:
1212     Kind = MCBinaryExpr::LTE;
1213     return 3;
1214   case AsmToken::Greater:
1215     Kind = MCBinaryExpr::GT;
1216     return 3;
1217   case AsmToken::GreaterEqual:
1218     Kind = MCBinaryExpr::GTE;
1219     return 3;
1220 
1221   // Low Intermediate Precedence: +, -
1222   case AsmToken::Plus:
1223     Kind = MCBinaryExpr::Add;
1224     return 4;
1225   case AsmToken::Minus:
1226     Kind = MCBinaryExpr::Sub;
1227     return 4;
1228 
1229   // High Intermediate Precedence: |, &, ^
1230   //
1231   // FIXME: gas seems to support '!' as an infix operator?
1232   case AsmToken::Pipe:
1233     Kind = MCBinaryExpr::Or;
1234     return 5;
1235   case AsmToken::Caret:
1236     Kind = MCBinaryExpr::Xor;
1237     return 5;
1238   case AsmToken::Amp:
1239     Kind = MCBinaryExpr::And;
1240     return 5;
1241 
1242   // Highest Precedence: *, /, %, <<, >>
1243   case AsmToken::Star:
1244     Kind = MCBinaryExpr::Mul;
1245     return 6;
1246   case AsmToken::Slash:
1247     Kind = MCBinaryExpr::Div;
1248     return 6;
1249   case AsmToken::Percent:
1250     Kind = MCBinaryExpr::Mod;
1251     return 6;
1252   case AsmToken::LessLess:
1253     Kind = MCBinaryExpr::Shl;
1254     return 6;
1255   case AsmToken::GreaterGreater:
1256     Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1257     return 6;
1258   }
1259 }
1260 
getBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind)1261 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K,
1262                                        MCBinaryExpr::Opcode &Kind) {
1263   bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr();
1264   return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr)
1265                   : getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr);
1266 }
1267 
1268 /// \brief Parse all binary operators with precedence >= 'Precedence'.
1269 /// Res contains the LHS of the expression on input.
parseBinOpRHS(unsigned Precedence,const MCExpr * & Res,SMLoc & EndLoc)1270 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1271                               SMLoc &EndLoc) {
1272   while (1) {
1273     MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1274     unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1275 
1276     // If the next token is lower precedence than we are allowed to eat, return
1277     // successfully with what we ate already.
1278     if (TokPrec < Precedence)
1279       return false;
1280 
1281     Lex();
1282 
1283     // Eat the next primary expression.
1284     const MCExpr *RHS;
1285     if (parsePrimaryExpr(RHS, EndLoc))
1286       return true;
1287 
1288     // If BinOp binds less tightly with RHS than the operator after RHS, let
1289     // the pending operator take RHS as its LHS.
1290     MCBinaryExpr::Opcode Dummy;
1291     unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1292     if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc))
1293       return true;
1294 
1295     // Merge LHS and RHS according to operator.
1296     Res = MCBinaryExpr::create(Kind, Res, RHS, getContext());
1297   }
1298 }
1299 
1300 /// ParseStatement:
1301 ///   ::= EndOfStatement
1302 ///   ::= Label* Directive ...Operands... EndOfStatement
1303 ///   ::= Label* Identifier OperandList* EndOfStatement
parseStatement(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)1304 bool AsmParser::parseStatement(ParseStatementInfo &Info,
1305                                MCAsmParserSemaCallback *SI) {
1306   if (Lexer.is(AsmToken::EndOfStatement)) {
1307     Out.AddBlankLine();
1308     Lex();
1309     return false;
1310   }
1311 
1312   // Statements always start with an identifier or are a full line comment.
1313   AsmToken ID = getTok();
1314   SMLoc IDLoc = ID.getLoc();
1315   StringRef IDVal;
1316   int64_t LocalLabelVal = -1;
1317   // A full line comment is a '#' as the first token.
1318   if (Lexer.is(AsmToken::Hash))
1319     return parseCppHashLineFilenameComment(IDLoc);
1320 
1321   // Allow an integer followed by a ':' as a directional local label.
1322   if (Lexer.is(AsmToken::Integer)) {
1323     LocalLabelVal = getTok().getIntVal();
1324     if (LocalLabelVal < 0) {
1325       if (!TheCondState.Ignore)
1326         return TokError("unexpected token at start of statement");
1327       IDVal = "";
1328     } else {
1329       IDVal = getTok().getString();
1330       Lex(); // Consume the integer token to be used as an identifier token.
1331       if (Lexer.getKind() != AsmToken::Colon) {
1332         if (!TheCondState.Ignore)
1333           return TokError("unexpected token at start of statement");
1334       }
1335     }
1336   } else if (Lexer.is(AsmToken::Dot)) {
1337     // Treat '.' as a valid identifier in this context.
1338     Lex();
1339     IDVal = ".";
1340   } else if (Lexer.is(AsmToken::LCurly)) {
1341     // Treat '{' as a valid identifier in this context.
1342     Lex();
1343     IDVal = "{";
1344 
1345   } else if (Lexer.is(AsmToken::RCurly)) {
1346     // Treat '}' as a valid identifier in this context.
1347     Lex();
1348     IDVal = "}";
1349   } else if (parseIdentifier(IDVal)) {
1350     if (!TheCondState.Ignore)
1351       return TokError("unexpected token at start of statement");
1352     IDVal = "";
1353   }
1354 
1355   // Handle conditional assembly here before checking for skipping.  We
1356   // have to do this so that .endif isn't skipped in a ".if 0" block for
1357   // example.
1358   StringMap<DirectiveKind>::const_iterator DirKindIt =
1359       DirectiveKindMap.find(IDVal);
1360   DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end())
1361                               ? DK_NO_DIRECTIVE
1362                               : DirKindIt->getValue();
1363   switch (DirKind) {
1364   default:
1365     break;
1366   case DK_IF:
1367   case DK_IFEQ:
1368   case DK_IFGE:
1369   case DK_IFGT:
1370   case DK_IFLE:
1371   case DK_IFLT:
1372   case DK_IFNE:
1373     return parseDirectiveIf(IDLoc, DirKind);
1374   case DK_IFB:
1375     return parseDirectiveIfb(IDLoc, true);
1376   case DK_IFNB:
1377     return parseDirectiveIfb(IDLoc, false);
1378   case DK_IFC:
1379     return parseDirectiveIfc(IDLoc, true);
1380   case DK_IFEQS:
1381     return parseDirectiveIfeqs(IDLoc, true);
1382   case DK_IFNC:
1383     return parseDirectiveIfc(IDLoc, false);
1384   case DK_IFNES:
1385     return parseDirectiveIfeqs(IDLoc, false);
1386   case DK_IFDEF:
1387     return parseDirectiveIfdef(IDLoc, true);
1388   case DK_IFNDEF:
1389   case DK_IFNOTDEF:
1390     return parseDirectiveIfdef(IDLoc, false);
1391   case DK_ELSEIF:
1392     return parseDirectiveElseIf(IDLoc);
1393   case DK_ELSE:
1394     return parseDirectiveElse(IDLoc);
1395   case DK_ENDIF:
1396     return parseDirectiveEndIf(IDLoc);
1397   }
1398 
1399   // Ignore the statement if in the middle of inactive conditional
1400   // (e.g. ".if 0").
1401   if (TheCondState.Ignore) {
1402     eatToEndOfStatement();
1403     return false;
1404   }
1405 
1406   // FIXME: Recurse on local labels?
1407 
1408   // See what kind of statement we have.
1409   switch (Lexer.getKind()) {
1410   case AsmToken::Colon: {
1411     if (!getTargetParser().isLabel(ID))
1412       break;
1413     checkForValidSection();
1414 
1415     // identifier ':'   -> Label.
1416     Lex();
1417 
1418     // Diagnose attempt to use '.' as a label.
1419     if (IDVal == ".")
1420       return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1421 
1422     // Diagnose attempt to use a variable as a label.
1423     //
1424     // FIXME: Diagnostics. Note the location of the definition as a label.
1425     // FIXME: This doesn't diagnose assignment to a symbol which has been
1426     // implicitly marked as external.
1427     MCSymbol *Sym;
1428     if (LocalLabelVal == -1) {
1429       if (ParsingInlineAsm && SI) {
1430         StringRef RewrittenLabel =
1431             SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true);
1432         assert(RewrittenLabel.size() &&
1433                "We should have an internal name here.");
1434         Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(),
1435                                        RewrittenLabel);
1436         IDVal = RewrittenLabel;
1437       }
1438       Sym = getContext().getOrCreateSymbol(IDVal);
1439     } else
1440       Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal);
1441 
1442     Sym->redefineIfPossible();
1443 
1444     if (!Sym->isUndefined() || Sym->isVariable())
1445       return Error(IDLoc, "invalid symbol redefinition");
1446 
1447     // Emit the label.
1448     if (!ParsingInlineAsm)
1449       Out.EmitLabel(Sym);
1450 
1451     // If we are generating dwarf for assembly source files then gather the
1452     // info to make a dwarf label entry for this label if needed.
1453     if (getContext().getGenDwarfForAssembly())
1454       MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1455                                  IDLoc);
1456 
1457     getTargetParser().onLabelParsed(Sym);
1458 
1459     // Consume any end of statement token, if present, to avoid spurious
1460     // AddBlankLine calls().
1461     if (Lexer.is(AsmToken::EndOfStatement)) {
1462       Lex();
1463       if (Lexer.is(AsmToken::Eof))
1464         return false;
1465     }
1466 
1467     return false;
1468   }
1469 
1470   case AsmToken::Equal:
1471     if (!getTargetParser().equalIsAsmAssignment())
1472       break;
1473     // identifier '=' ... -> assignment statement
1474     Lex();
1475 
1476     return parseAssignment(IDVal, true);
1477 
1478   default: // Normal instruction or directive.
1479     break;
1480   }
1481 
1482   // If macros are enabled, check to see if this is a macro instantiation.
1483   if (areMacrosEnabled())
1484     if (const MCAsmMacro *M = lookupMacro(IDVal)) {
1485       return handleMacroEntry(M, IDLoc);
1486     }
1487 
1488   // Otherwise, we have a normal instruction or directive.
1489 
1490   // Directives start with "."
1491   if (IDVal[0] == '.' && IDVal != ".") {
1492     // There are several entities interested in parsing directives:
1493     //
1494     // 1. The target-specific assembly parser. Some directives are target
1495     //    specific or may potentially behave differently on certain targets.
1496     // 2. Asm parser extensions. For example, platform-specific parsers
1497     //    (like the ELF parser) register themselves as extensions.
1498     // 3. The generic directive parser implemented by this class. These are
1499     //    all the directives that behave in a target and platform independent
1500     //    manner, or at least have a default behavior that's shared between
1501     //    all targets and platforms.
1502 
1503     // First query the target-specific parser. It will return 'true' if it
1504     // isn't interested in this directive.
1505     if (!getTargetParser().ParseDirective(ID))
1506       return false;
1507 
1508     // Next, check the extension directive map to see if any extension has
1509     // registered itself to parse this directive.
1510     std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
1511         ExtensionDirectiveMap.lookup(IDVal);
1512     if (Handler.first)
1513       return (*Handler.second)(Handler.first, IDVal, IDLoc);
1514 
1515     // Finally, if no one else is interested in this directive, it must be
1516     // generic and familiar to this class.
1517     switch (DirKind) {
1518     default:
1519       break;
1520     case DK_SET:
1521     case DK_EQU:
1522       return parseDirectiveSet(IDVal, true);
1523     case DK_EQUIV:
1524       return parseDirectiveSet(IDVal, false);
1525     case DK_ASCII:
1526       return parseDirectiveAscii(IDVal, false);
1527     case DK_ASCIZ:
1528     case DK_STRING:
1529       return parseDirectiveAscii(IDVal, true);
1530     case DK_BYTE:
1531       return parseDirectiveValue(1);
1532     case DK_SHORT:
1533     case DK_VALUE:
1534     case DK_2BYTE:
1535       return parseDirectiveValue(2);
1536     case DK_LONG:
1537     case DK_INT:
1538     case DK_4BYTE:
1539       return parseDirectiveValue(4);
1540     case DK_QUAD:
1541     case DK_8BYTE:
1542       return parseDirectiveValue(8);
1543     case DK_OCTA:
1544       return parseDirectiveOctaValue();
1545     case DK_SINGLE:
1546     case DK_FLOAT:
1547       return parseDirectiveRealValue(APFloat::IEEEsingle);
1548     case DK_DOUBLE:
1549       return parseDirectiveRealValue(APFloat::IEEEdouble);
1550     case DK_ALIGN: {
1551       bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1552       return parseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1553     }
1554     case DK_ALIGN32: {
1555       bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1556       return parseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1557     }
1558     case DK_BALIGN:
1559       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1560     case DK_BALIGNW:
1561       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1562     case DK_BALIGNL:
1563       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1564     case DK_P2ALIGN:
1565       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1566     case DK_P2ALIGNW:
1567       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1568     case DK_P2ALIGNL:
1569       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1570     case DK_ORG:
1571       return parseDirectiveOrg();
1572     case DK_FILL:
1573       return parseDirectiveFill();
1574     case DK_ZERO:
1575       return parseDirectiveZero();
1576     case DK_EXTERN:
1577       eatToEndOfStatement(); // .extern is the default, ignore it.
1578       return false;
1579     case DK_GLOBL:
1580     case DK_GLOBAL:
1581       return parseDirectiveSymbolAttribute(MCSA_Global);
1582     case DK_LAZY_REFERENCE:
1583       return parseDirectiveSymbolAttribute(MCSA_LazyReference);
1584     case DK_NO_DEAD_STRIP:
1585       return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1586     case DK_SYMBOL_RESOLVER:
1587       return parseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1588     case DK_PRIVATE_EXTERN:
1589       return parseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1590     case DK_REFERENCE:
1591       return parseDirectiveSymbolAttribute(MCSA_Reference);
1592     case DK_WEAK_DEFINITION:
1593       return parseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1594     case DK_WEAK_REFERENCE:
1595       return parseDirectiveSymbolAttribute(MCSA_WeakReference);
1596     case DK_WEAK_DEF_CAN_BE_HIDDEN:
1597       return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1598     case DK_COMM:
1599     case DK_COMMON:
1600       return parseDirectiveComm(/*IsLocal=*/false);
1601     case DK_LCOMM:
1602       return parseDirectiveComm(/*IsLocal=*/true);
1603     case DK_ABORT:
1604       return parseDirectiveAbort();
1605     case DK_INCLUDE:
1606       return parseDirectiveInclude();
1607     case DK_INCBIN:
1608       return parseDirectiveIncbin();
1609     case DK_CODE16:
1610     case DK_CODE16GCC:
1611       return TokError(Twine(IDVal) + " not supported yet");
1612     case DK_REPT:
1613       return parseDirectiveRept(IDLoc, IDVal);
1614     case DK_IRP:
1615       return parseDirectiveIrp(IDLoc);
1616     case DK_IRPC:
1617       return parseDirectiveIrpc(IDLoc);
1618     case DK_ENDR:
1619       return parseDirectiveEndr(IDLoc);
1620     case DK_BUNDLE_ALIGN_MODE:
1621       return parseDirectiveBundleAlignMode();
1622     case DK_BUNDLE_LOCK:
1623       return parseDirectiveBundleLock();
1624     case DK_BUNDLE_UNLOCK:
1625       return parseDirectiveBundleUnlock();
1626     case DK_SLEB128:
1627       return parseDirectiveLEB128(true);
1628     case DK_ULEB128:
1629       return parseDirectiveLEB128(false);
1630     case DK_SPACE:
1631     case DK_SKIP:
1632       return parseDirectiveSpace(IDVal);
1633     case DK_FILE:
1634       return parseDirectiveFile(IDLoc);
1635     case DK_LINE:
1636       return parseDirectiveLine();
1637     case DK_LOC:
1638       return parseDirectiveLoc();
1639     case DK_STABS:
1640       return parseDirectiveStabs();
1641     case DK_CFI_SECTIONS:
1642       return parseDirectiveCFISections();
1643     case DK_CFI_STARTPROC:
1644       return parseDirectiveCFIStartProc();
1645     case DK_CFI_ENDPROC:
1646       return parseDirectiveCFIEndProc();
1647     case DK_CFI_DEF_CFA:
1648       return parseDirectiveCFIDefCfa(IDLoc);
1649     case DK_CFI_DEF_CFA_OFFSET:
1650       return parseDirectiveCFIDefCfaOffset();
1651     case DK_CFI_ADJUST_CFA_OFFSET:
1652       return parseDirectiveCFIAdjustCfaOffset();
1653     case DK_CFI_DEF_CFA_REGISTER:
1654       return parseDirectiveCFIDefCfaRegister(IDLoc);
1655     case DK_CFI_OFFSET:
1656       return parseDirectiveCFIOffset(IDLoc);
1657     case DK_CFI_REL_OFFSET:
1658       return parseDirectiveCFIRelOffset(IDLoc);
1659     case DK_CFI_PERSONALITY:
1660       return parseDirectiveCFIPersonalityOrLsda(true);
1661     case DK_CFI_LSDA:
1662       return parseDirectiveCFIPersonalityOrLsda(false);
1663     case DK_CFI_REMEMBER_STATE:
1664       return parseDirectiveCFIRememberState();
1665     case DK_CFI_RESTORE_STATE:
1666       return parseDirectiveCFIRestoreState();
1667     case DK_CFI_SAME_VALUE:
1668       return parseDirectiveCFISameValue(IDLoc);
1669     case DK_CFI_RESTORE:
1670       return parseDirectiveCFIRestore(IDLoc);
1671     case DK_CFI_ESCAPE:
1672       return parseDirectiveCFIEscape();
1673     case DK_CFI_SIGNAL_FRAME:
1674       return parseDirectiveCFISignalFrame();
1675     case DK_CFI_UNDEFINED:
1676       return parseDirectiveCFIUndefined(IDLoc);
1677     case DK_CFI_REGISTER:
1678       return parseDirectiveCFIRegister(IDLoc);
1679     case DK_CFI_WINDOW_SAVE:
1680       return parseDirectiveCFIWindowSave();
1681     case DK_MACROS_ON:
1682     case DK_MACROS_OFF:
1683       return parseDirectiveMacrosOnOff(IDVal);
1684     case DK_MACRO:
1685       return parseDirectiveMacro(IDLoc);
1686     case DK_EXITM:
1687       return parseDirectiveExitMacro(IDVal);
1688     case DK_ENDM:
1689     case DK_ENDMACRO:
1690       return parseDirectiveEndMacro(IDVal);
1691     case DK_PURGEM:
1692       return parseDirectivePurgeMacro(IDLoc);
1693     case DK_END:
1694       return parseDirectiveEnd(IDLoc);
1695     case DK_ERR:
1696       return parseDirectiveError(IDLoc, false);
1697     case DK_ERROR:
1698       return parseDirectiveError(IDLoc, true);
1699     case DK_WARNING:
1700       return parseDirectiveWarning(IDLoc);
1701     case DK_RELOC:
1702       return parseDirectiveReloc(IDLoc);
1703     }
1704 
1705     return Error(IDLoc, "unknown directive");
1706   }
1707 
1708   // __asm _emit or __asm __emit
1709   if (ParsingInlineAsm && (IDVal == "_emit" || IDVal == "__emit" ||
1710                            IDVal == "_EMIT" || IDVal == "__EMIT"))
1711     return parseDirectiveMSEmit(IDLoc, Info, IDVal.size());
1712 
1713   // __asm align
1714   if (ParsingInlineAsm && (IDVal == "align" || IDVal == "ALIGN"))
1715     return parseDirectiveMSAlign(IDLoc, Info);
1716 
1717   if (ParsingInlineAsm && (IDVal == "even"))
1718     Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4);
1719   checkForValidSection();
1720 
1721   // Canonicalize the opcode to lower case.
1722   std::string OpcodeStr = IDVal.lower();
1723   ParseInstructionInfo IInfo(Info.AsmRewrites);
1724   bool HadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID,
1725                                                      Info.ParsedOperands);
1726   Info.ParseError = HadError;
1727 
1728   // Dump the parsed representation, if requested.
1729   if (getShowParsedOperands()) {
1730     SmallString<256> Str;
1731     raw_svector_ostream OS(Str);
1732     OS << "parsed instruction: [";
1733     for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
1734       if (i != 0)
1735         OS << ", ";
1736       Info.ParsedOperands[i]->print(OS);
1737     }
1738     OS << "]";
1739 
1740     printMessage(IDLoc, SourceMgr::DK_Note, OS.str());
1741   }
1742 
1743   // If we are generating dwarf for the current section then generate a .loc
1744   // directive for the instruction.
1745   if (!HadError && getContext().getGenDwarfForAssembly() &&
1746       getContext().getGenDwarfSectionSyms().count(
1747           getStreamer().getCurrentSection().first)) {
1748     unsigned Line;
1749     if (ActiveMacros.empty())
1750       Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
1751     else
1752       Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc,
1753                                    ActiveMacros.front()->ExitBuffer);
1754 
1755     // If we previously parsed a cpp hash file line comment then make sure the
1756     // current Dwarf File is for the CppHashFilename if not then emit the
1757     // Dwarf File table for it and adjust the line number for the .loc.
1758     if (CppHashFilename.size()) {
1759       unsigned FileNumber = getStreamer().EmitDwarfFileDirective(
1760           0, StringRef(), CppHashFilename);
1761       getContext().setGenDwarfFileNumber(FileNumber);
1762 
1763       // Since SrcMgr.FindLineNumber() is slow and messes up the SourceMgr's
1764       // cache with the different Loc from the call above we save the last
1765       // info we queried here with SrcMgr.FindLineNumber().
1766       unsigned CppHashLocLineNo;
1767       if (LastQueryIDLoc == CppHashLoc && LastQueryBuffer == CppHashBuf)
1768         CppHashLocLineNo = LastQueryLine;
1769       else {
1770         CppHashLocLineNo = SrcMgr.FindLineNumber(CppHashLoc, CppHashBuf);
1771         LastQueryLine = CppHashLocLineNo;
1772         LastQueryIDLoc = CppHashLoc;
1773         LastQueryBuffer = CppHashBuf;
1774       }
1775       Line = CppHashLineNumber - 1 + (Line - CppHashLocLineNo);
1776     }
1777 
1778     getStreamer().EmitDwarfLocDirective(
1779         getContext().getGenDwarfFileNumber(), Line, 0,
1780         DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0,
1781         StringRef());
1782   }
1783 
1784   // If parsing succeeded, match the instruction.
1785   if (!HadError) {
1786     uint64_t ErrorInfo;
1787     getTargetParser().MatchAndEmitInstruction(IDLoc, Info.Opcode,
1788                                               Info.ParsedOperands, Out,
1789                                               ErrorInfo, ParsingInlineAsm);
1790   }
1791 
1792   // Don't skip the rest of the line, the instruction parser is responsible for
1793   // that.
1794   return false;
1795 }
1796 
1797 /// eatToEndOfLine uses the Lexer to eat the characters to the end of the line
1798 /// since they may not be able to be tokenized to get to the end of line token.
eatToEndOfLine()1799 void AsmParser::eatToEndOfLine() {
1800   if (!Lexer.is(AsmToken::EndOfStatement))
1801     Lexer.LexUntilEndOfLine();
1802   // Eat EOL.
1803   Lex();
1804 }
1805 
1806 /// parseCppHashLineFilenameComment as this:
1807 ///   ::= # number "filename"
1808 /// or just as a full line comment if it doesn't have a number and a string.
parseCppHashLineFilenameComment(SMLoc L)1809 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L) {
1810   Lex(); // Eat the hash token.
1811 
1812   if (getLexer().isNot(AsmToken::Integer)) {
1813     // Consume the line since in cases it is not a well-formed line directive,
1814     // as if were simply a full line comment.
1815     eatToEndOfLine();
1816     return false;
1817   }
1818 
1819   int64_t LineNumber = getTok().getIntVal();
1820   Lex();
1821 
1822   if (getLexer().isNot(AsmToken::String)) {
1823     eatToEndOfLine();
1824     return false;
1825   }
1826 
1827   StringRef Filename = getTok().getString();
1828   // Get rid of the enclosing quotes.
1829   Filename = Filename.substr(1, Filename.size() - 2);
1830 
1831   // Save the SMLoc, Filename and LineNumber for later use by diagnostics.
1832   CppHashLoc = L;
1833   CppHashFilename = Filename;
1834   CppHashLineNumber = LineNumber;
1835   CppHashBuf = CurBuffer;
1836 
1837   // Ignore any trailing characters, they're just comment.
1838   eatToEndOfLine();
1839   return false;
1840 }
1841 
1842 /// \brief will use the last parsed cpp hash line filename comment
1843 /// for the Filename and LineNo if any in the diagnostic.
DiagHandler(const SMDiagnostic & Diag,void * Context)1844 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
1845   const AsmParser *Parser = static_cast<const AsmParser *>(Context);
1846   raw_ostream &OS = errs();
1847 
1848   const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
1849   SMLoc DiagLoc = Diag.getLoc();
1850   unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1851   unsigned CppHashBuf =
1852       Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc);
1853 
1854   // Like SourceMgr::printMessage() we need to print the include stack if any
1855   // before printing the message.
1856   unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1857   if (!Parser->SavedDiagHandler && DiagCurBuffer &&
1858       DiagCurBuffer != DiagSrcMgr.getMainFileID()) {
1859     SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
1860     DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
1861   }
1862 
1863   // If we have not parsed a cpp hash line filename comment or the source
1864   // manager changed or buffer changed (like in a nested include) then just
1865   // print the normal diagnostic using its Filename and LineNo.
1866   if (!Parser->CppHashLineNumber || &DiagSrcMgr != &Parser->SrcMgr ||
1867       DiagBuf != CppHashBuf) {
1868     if (Parser->SavedDiagHandler)
1869       Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
1870     else
1871       Diag.print(nullptr, OS);
1872     return;
1873   }
1874 
1875   // Use the CppHashFilename and calculate a line number based on the
1876   // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for
1877   // the diagnostic.
1878   const std::string &Filename = Parser->CppHashFilename;
1879 
1880   int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
1881   int CppHashLocLineNo =
1882       Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf);
1883   int LineNo =
1884       Parser->CppHashLineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo);
1885 
1886   SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo,
1887                        Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(),
1888                        Diag.getLineContents(), Diag.getRanges());
1889 
1890   if (Parser->SavedDiagHandler)
1891     Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext);
1892   else
1893     NewDiag.print(nullptr, OS);
1894 }
1895 
1896 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
1897 // difference being that that function accepts '@' as part of identifiers and
1898 // we can't do that. AsmLexer.cpp should probably be changed to handle
1899 // '@' as a special case when needed.
isIdentifierChar(char c)1900 static bool isIdentifierChar(char c) {
1901   return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' ||
1902          c == '.';
1903 }
1904 
expandMacro(raw_svector_ostream & OS,StringRef Body,ArrayRef<MCAsmMacroParameter> Parameters,ArrayRef<MCAsmMacroArgument> A,bool EnableAtPseudoVariable,SMLoc L)1905 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
1906                             ArrayRef<MCAsmMacroParameter> Parameters,
1907                             ArrayRef<MCAsmMacroArgument> A,
1908                             bool EnableAtPseudoVariable, SMLoc L) {
1909   unsigned NParameters = Parameters.size();
1910   bool HasVararg = NParameters ? Parameters.back().Vararg : false;
1911   if ((!IsDarwin || NParameters != 0) && NParameters != A.size())
1912     return Error(L, "Wrong number of arguments");
1913 
1914   // A macro without parameters is handled differently on Darwin:
1915   // gas accepts no arguments and does no substitutions
1916   while (!Body.empty()) {
1917     // Scan for the next substitution.
1918     std::size_t End = Body.size(), Pos = 0;
1919     for (; Pos != End; ++Pos) {
1920       // Check for a substitution or escape.
1921       if (IsDarwin && !NParameters) {
1922         // This macro has no parameters, look for $0, $1, etc.
1923         if (Body[Pos] != '$' || Pos + 1 == End)
1924           continue;
1925 
1926         char Next = Body[Pos + 1];
1927         if (Next == '$' || Next == 'n' ||
1928             isdigit(static_cast<unsigned char>(Next)))
1929           break;
1930       } else {
1931         // This macro has parameters, look for \foo, \bar, etc.
1932         if (Body[Pos] == '\\' && Pos + 1 != End)
1933           break;
1934       }
1935     }
1936 
1937     // Add the prefix.
1938     OS << Body.slice(0, Pos);
1939 
1940     // Check if we reached the end.
1941     if (Pos == End)
1942       break;
1943 
1944     if (IsDarwin && !NParameters) {
1945       switch (Body[Pos + 1]) {
1946       // $$ => $
1947       case '$':
1948         OS << '$';
1949         break;
1950 
1951       // $n => number of arguments
1952       case 'n':
1953         OS << A.size();
1954         break;
1955 
1956       // $[0-9] => argument
1957       default: {
1958         // Missing arguments are ignored.
1959         unsigned Index = Body[Pos + 1] - '0';
1960         if (Index >= A.size())
1961           break;
1962 
1963         // Otherwise substitute with the token values, with spaces eliminated.
1964         for (const AsmToken &Token : A[Index])
1965           OS << Token.getString();
1966         break;
1967       }
1968       }
1969       Pos += 2;
1970     } else {
1971       unsigned I = Pos + 1;
1972 
1973       // Check for the \@ pseudo-variable.
1974       if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End)
1975         ++I;
1976       else
1977         while (isIdentifierChar(Body[I]) && I + 1 != End)
1978           ++I;
1979 
1980       const char *Begin = Body.data() + Pos + 1;
1981       StringRef Argument(Begin, I - (Pos + 1));
1982       unsigned Index = 0;
1983 
1984       if (Argument == "@") {
1985         OS << NumOfMacroInstantiations;
1986         Pos += 2;
1987       } else {
1988         for (; Index < NParameters; ++Index)
1989           if (Parameters[Index].Name == Argument)
1990             break;
1991 
1992         if (Index == NParameters) {
1993           if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
1994             Pos += 3;
1995           else {
1996             OS << '\\' << Argument;
1997             Pos = I;
1998           }
1999         } else {
2000           bool VarargParameter = HasVararg && Index == (NParameters - 1);
2001           for (const AsmToken &Token : A[Index])
2002             // We expect no quotes around the string's contents when
2003             // parsing for varargs.
2004             if (Token.getKind() != AsmToken::String || VarargParameter)
2005               OS << Token.getString();
2006             else
2007               OS << Token.getStringContents();
2008 
2009           Pos += 1 + Argument.size();
2010         }
2011       }
2012     }
2013     // Update the scan point.
2014     Body = Body.substr(Pos);
2015   }
2016 
2017   return false;
2018 }
2019 
MacroInstantiation(SMLoc IL,int EB,SMLoc EL,size_t CondStackDepth)2020 MacroInstantiation::MacroInstantiation(SMLoc IL, int EB, SMLoc EL,
2021                                        size_t CondStackDepth)
2022     : InstantiationLoc(IL), ExitBuffer(EB), ExitLoc(EL),
2023       CondStackDepth(CondStackDepth) {}
2024 
isOperator(AsmToken::TokenKind kind)2025 static bool isOperator(AsmToken::TokenKind kind) {
2026   switch (kind) {
2027   default:
2028     return false;
2029   case AsmToken::Plus:
2030   case AsmToken::Minus:
2031   case AsmToken::Tilde:
2032   case AsmToken::Slash:
2033   case AsmToken::Star:
2034   case AsmToken::Dot:
2035   case AsmToken::Equal:
2036   case AsmToken::EqualEqual:
2037   case AsmToken::Pipe:
2038   case AsmToken::PipePipe:
2039   case AsmToken::Caret:
2040   case AsmToken::Amp:
2041   case AsmToken::AmpAmp:
2042   case AsmToken::Exclaim:
2043   case AsmToken::ExclaimEqual:
2044   case AsmToken::Percent:
2045   case AsmToken::Less:
2046   case AsmToken::LessEqual:
2047   case AsmToken::LessLess:
2048   case AsmToken::LessGreater:
2049   case AsmToken::Greater:
2050   case AsmToken::GreaterEqual:
2051   case AsmToken::GreaterGreater:
2052     return true;
2053   }
2054 }
2055 
2056 namespace {
2057 class AsmLexerSkipSpaceRAII {
2058 public:
AsmLexerSkipSpaceRAII(AsmLexer & Lexer,bool SkipSpace)2059   AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) {
2060     Lexer.setSkipSpace(SkipSpace);
2061   }
2062 
~AsmLexerSkipSpaceRAII()2063   ~AsmLexerSkipSpaceRAII() {
2064     Lexer.setSkipSpace(true);
2065   }
2066 
2067 private:
2068   AsmLexer &Lexer;
2069 };
2070 }
2071 
parseMacroArgument(MCAsmMacroArgument & MA,bool Vararg)2072 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) {
2073 
2074   if (Vararg) {
2075     if (Lexer.isNot(AsmToken::EndOfStatement)) {
2076       StringRef Str = parseStringToEndOfStatement();
2077       MA.emplace_back(AsmToken::String, Str);
2078     }
2079     return false;
2080   }
2081 
2082   unsigned ParenLevel = 0;
2083   unsigned AddTokens = 0;
2084 
2085   // Darwin doesn't use spaces to delmit arguments.
2086   AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin);
2087 
2088   for (;;) {
2089     if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
2090       return TokError("unexpected token in macro instantiation");
2091 
2092     if (ParenLevel == 0 && Lexer.is(AsmToken::Comma))
2093       break;
2094 
2095     if (Lexer.is(AsmToken::Space)) {
2096       Lex(); // Eat spaces
2097 
2098       // Spaces can delimit parameters, but could also be part an expression.
2099       // If the token after a space is an operator, add the token and the next
2100       // one into this argument
2101       if (!IsDarwin) {
2102         if (isOperator(Lexer.getKind())) {
2103           // Check to see whether the token is used as an operator,
2104           // or part of an identifier
2105           const char *NextChar = getTok().getEndLoc().getPointer();
2106           if (*NextChar == ' ')
2107             AddTokens = 2;
2108         }
2109 
2110         if (!AddTokens && ParenLevel == 0) {
2111           break;
2112         }
2113       }
2114     }
2115 
2116     // handleMacroEntry relies on not advancing the lexer here
2117     // to be able to fill in the remaining default parameter values
2118     if (Lexer.is(AsmToken::EndOfStatement))
2119       break;
2120 
2121     // Adjust the current parentheses level.
2122     if (Lexer.is(AsmToken::LParen))
2123       ++ParenLevel;
2124     else if (Lexer.is(AsmToken::RParen) && ParenLevel)
2125       --ParenLevel;
2126 
2127     // Append the token to the current argument list.
2128     MA.push_back(getTok());
2129     if (AddTokens)
2130       AddTokens--;
2131     Lex();
2132   }
2133 
2134   if (ParenLevel != 0)
2135     return TokError("unbalanced parentheses in macro argument");
2136   return false;
2137 }
2138 
2139 // Parse the macro instantiation arguments.
parseMacroArguments(const MCAsmMacro * M,MCAsmMacroArguments & A)2140 bool AsmParser::parseMacroArguments(const MCAsmMacro *M,
2141                                     MCAsmMacroArguments &A) {
2142   const unsigned NParameters = M ? M->Parameters.size() : 0;
2143   bool NamedParametersFound = false;
2144   SmallVector<SMLoc, 4> FALocs;
2145 
2146   A.resize(NParameters);
2147   FALocs.resize(NParameters);
2148 
2149   // Parse two kinds of macro invocations:
2150   // - macros defined without any parameters accept an arbitrary number of them
2151   // - macros defined with parameters accept at most that many of them
2152   bool HasVararg = NParameters ? M->Parameters.back().Vararg : false;
2153   for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
2154        ++Parameter) {
2155     SMLoc IDLoc = Lexer.getLoc();
2156     MCAsmMacroParameter FA;
2157 
2158     if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) {
2159       if (parseIdentifier(FA.Name)) {
2160         Error(IDLoc, "invalid argument identifier for formal argument");
2161         eatToEndOfStatement();
2162         return true;
2163       }
2164 
2165       if (!Lexer.is(AsmToken::Equal)) {
2166         TokError("expected '=' after formal parameter identifier");
2167         eatToEndOfStatement();
2168         return true;
2169       }
2170       Lex();
2171 
2172       NamedParametersFound = true;
2173     }
2174 
2175     if (NamedParametersFound && FA.Name.empty()) {
2176       Error(IDLoc, "cannot mix positional and keyword arguments");
2177       eatToEndOfStatement();
2178       return true;
2179     }
2180 
2181     bool Vararg = HasVararg && Parameter == (NParameters - 1);
2182     if (parseMacroArgument(FA.Value, Vararg))
2183       return true;
2184 
2185     unsigned PI = Parameter;
2186     if (!FA.Name.empty()) {
2187       unsigned FAI = 0;
2188       for (FAI = 0; FAI < NParameters; ++FAI)
2189         if (M->Parameters[FAI].Name == FA.Name)
2190           break;
2191 
2192       if (FAI >= NParameters) {
2193     assert(M && "expected macro to be defined");
2194         Error(IDLoc,
2195               "parameter named '" + FA.Name + "' does not exist for macro '" +
2196               M->Name + "'");
2197         return true;
2198       }
2199       PI = FAI;
2200     }
2201 
2202     if (!FA.Value.empty()) {
2203       if (A.size() <= PI)
2204         A.resize(PI + 1);
2205       A[PI] = FA.Value;
2206 
2207       if (FALocs.size() <= PI)
2208         FALocs.resize(PI + 1);
2209 
2210       FALocs[PI] = Lexer.getLoc();
2211     }
2212 
2213     // At the end of the statement, fill in remaining arguments that have
2214     // default values. If there aren't any, then the next argument is
2215     // required but missing
2216     if (Lexer.is(AsmToken::EndOfStatement)) {
2217       bool Failure = false;
2218       for (unsigned FAI = 0; FAI < NParameters; ++FAI) {
2219         if (A[FAI].empty()) {
2220           if (M->Parameters[FAI].Required) {
2221             Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(),
2222                   "missing value for required parameter "
2223                   "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'");
2224             Failure = true;
2225           }
2226 
2227           if (!M->Parameters[FAI].Value.empty())
2228             A[FAI] = M->Parameters[FAI].Value;
2229         }
2230       }
2231       return Failure;
2232     }
2233 
2234     if (Lexer.is(AsmToken::Comma))
2235       Lex();
2236   }
2237 
2238   return TokError("too many positional arguments");
2239 }
2240 
lookupMacro(StringRef Name)2241 const MCAsmMacro *AsmParser::lookupMacro(StringRef Name) {
2242   StringMap<MCAsmMacro>::iterator I = MacroMap.find(Name);
2243   return (I == MacroMap.end()) ? nullptr : &I->getValue();
2244 }
2245 
defineMacro(StringRef Name,MCAsmMacro Macro)2246 void AsmParser::defineMacro(StringRef Name, MCAsmMacro Macro) {
2247   MacroMap.insert(std::make_pair(Name, std::move(Macro)));
2248 }
2249 
undefineMacro(StringRef Name)2250 void AsmParser::undefineMacro(StringRef Name) { MacroMap.erase(Name); }
2251 
handleMacroEntry(const MCAsmMacro * M,SMLoc NameLoc)2252 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) {
2253   // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
2254   // this, although we should protect against infinite loops.
2255   if (ActiveMacros.size() == 20)
2256     return TokError("macros cannot be nested more than 20 levels deep");
2257 
2258   MCAsmMacroArguments A;
2259   if (parseMacroArguments(M, A))
2260     return true;
2261 
2262   // Macro instantiation is lexical, unfortunately. We construct a new buffer
2263   // to hold the macro body with substitutions.
2264   SmallString<256> Buf;
2265   StringRef Body = M->Body;
2266   raw_svector_ostream OS(Buf);
2267 
2268   if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc()))
2269     return true;
2270 
2271   // We include the .endmacro in the buffer as our cue to exit the macro
2272   // instantiation.
2273   OS << ".endmacro\n";
2274 
2275   std::unique_ptr<MemoryBuffer> Instantiation =
2276       MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
2277 
2278   // Create the macro instantiation object and add to the current macro
2279   // instantiation stack.
2280   MacroInstantiation *MI = new MacroInstantiation(
2281       NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size());
2282   ActiveMacros.push_back(MI);
2283 
2284   ++NumOfMacroInstantiations;
2285 
2286   // Jump to the macro instantiation and prime the lexer.
2287   CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
2288   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
2289   Lex();
2290 
2291   return false;
2292 }
2293 
handleMacroExit()2294 void AsmParser::handleMacroExit() {
2295   // Jump to the EndOfStatement we should return to, and consume it.
2296   jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer);
2297   Lex();
2298 
2299   // Pop the instantiation entry.
2300   delete ActiveMacros.back();
2301   ActiveMacros.pop_back();
2302 }
2303 
parseAssignment(StringRef Name,bool allow_redef,bool NoDeadStrip)2304 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef,
2305                                 bool NoDeadStrip) {
2306   MCSymbol *Sym;
2307   const MCExpr *Value;
2308   if (MCParserUtils::parseAssignmentExpression(Name, allow_redef, *this, Sym,
2309                                                Value))
2310     return true;
2311 
2312   if (!Sym) {
2313     // In the case where we parse an expression starting with a '.', we will
2314     // not generate an error, nor will we create a symbol.  In this case we
2315     // should just return out.
2316     return false;
2317   }
2318 
2319   // Do the assignment.
2320   Out.EmitAssignment(Sym, Value);
2321   if (NoDeadStrip)
2322     Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip);
2323 
2324   return false;
2325 }
2326 
2327 /// parseIdentifier:
2328 ///   ::= identifier
2329 ///   ::= string
parseIdentifier(StringRef & Res)2330 bool AsmParser::parseIdentifier(StringRef &Res) {
2331   // The assembler has relaxed rules for accepting identifiers, in particular we
2332   // allow things like '.globl $foo' and '.def @feat.00', which would normally be
2333   // separate tokens. At this level, we have already lexed so we cannot (currently)
2334   // handle this as a context dependent token, instead we detect adjacent tokens
2335   // and return the combined identifier.
2336   if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) {
2337     SMLoc PrefixLoc = getLexer().getLoc();
2338 
2339     // Consume the prefix character, and check for a following identifier.
2340     Lex();
2341     if (Lexer.isNot(AsmToken::Identifier))
2342       return true;
2343 
2344     // We have a '$' or '@' followed by an identifier, make sure they are adjacent.
2345     if (PrefixLoc.getPointer() + 1 != getTok().getLoc().getPointer())
2346       return true;
2347 
2348     // Construct the joined identifier and consume the token.
2349     Res =
2350         StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1);
2351     Lex();
2352     return false;
2353   }
2354 
2355   if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String))
2356     return true;
2357 
2358   Res = getTok().getIdentifier();
2359 
2360   Lex(); // Consume the identifier token.
2361 
2362   return false;
2363 }
2364 
2365 /// parseDirectiveSet:
2366 ///   ::= .equ identifier ',' expression
2367 ///   ::= .equiv identifier ',' expression
2368 ///   ::= .set identifier ',' expression
parseDirectiveSet(StringRef IDVal,bool allow_redef)2369 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) {
2370   StringRef Name;
2371 
2372   if (parseIdentifier(Name))
2373     return TokError("expected identifier after '" + Twine(IDVal) + "'");
2374 
2375   if (getLexer().isNot(AsmToken::Comma))
2376     return TokError("unexpected token in '" + Twine(IDVal) + "'");
2377   Lex();
2378 
2379   return parseAssignment(Name, allow_redef, true);
2380 }
2381 
parseEscapedString(std::string & Data)2382 bool AsmParser::parseEscapedString(std::string &Data) {
2383   assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
2384 
2385   Data = "";
2386   StringRef Str = getTok().getStringContents();
2387   for (unsigned i = 0, e = Str.size(); i != e; ++i) {
2388     if (Str[i] != '\\') {
2389       Data += Str[i];
2390       continue;
2391     }
2392 
2393     // Recognize escaped characters. Note that this escape semantics currently
2394     // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
2395     ++i;
2396     if (i == e)
2397       return TokError("unexpected backslash at end of string");
2398 
2399     // Recognize octal sequences.
2400     if ((unsigned)(Str[i] - '0') <= 7) {
2401       // Consume up to three octal characters.
2402       unsigned Value = Str[i] - '0';
2403 
2404       if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2405         ++i;
2406         Value = Value * 8 + (Str[i] - '0');
2407 
2408         if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2409           ++i;
2410           Value = Value * 8 + (Str[i] - '0');
2411         }
2412       }
2413 
2414       if (Value > 255)
2415         return TokError("invalid octal escape sequence (out of range)");
2416 
2417       Data += (unsigned char)Value;
2418       continue;
2419     }
2420 
2421     // Otherwise recognize individual escapes.
2422     switch (Str[i]) {
2423     default:
2424       // Just reject invalid escape sequences for now.
2425       return TokError("invalid escape sequence (unrecognized character)");
2426 
2427     case 'b': Data += '\b'; break;
2428     case 'f': Data += '\f'; break;
2429     case 'n': Data += '\n'; break;
2430     case 'r': Data += '\r'; break;
2431     case 't': Data += '\t'; break;
2432     case '"': Data += '"'; break;
2433     case '\\': Data += '\\'; break;
2434     }
2435   }
2436 
2437   return false;
2438 }
2439 
2440 /// parseDirectiveAscii:
2441 ///   ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
parseDirectiveAscii(StringRef IDVal,bool ZeroTerminated)2442 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
2443   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2444     checkForValidSection();
2445 
2446     for (;;) {
2447       if (getLexer().isNot(AsmToken::String))
2448         return TokError("expected string in '" + Twine(IDVal) + "' directive");
2449 
2450       std::string Data;
2451       if (parseEscapedString(Data))
2452         return true;
2453 
2454       getStreamer().EmitBytes(Data);
2455       if (ZeroTerminated)
2456         getStreamer().EmitBytes(StringRef("\0", 1));
2457 
2458       Lex();
2459 
2460       if (getLexer().is(AsmToken::EndOfStatement))
2461         break;
2462 
2463       if (getLexer().isNot(AsmToken::Comma))
2464         return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
2465       Lex();
2466     }
2467   }
2468 
2469   Lex();
2470   return false;
2471 }
2472 
2473 /// parseDirectiveReloc
2474 ///  ::= .reloc expression , identifier [ , expression ]
parseDirectiveReloc(SMLoc DirectiveLoc)2475 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) {
2476   const MCExpr *Offset;
2477   const MCExpr *Expr = nullptr;
2478 
2479   SMLoc OffsetLoc = Lexer.getTok().getLoc();
2480   if (parseExpression(Offset))
2481     return true;
2482 
2483   // We can only deal with constant expressions at the moment.
2484   int64_t OffsetValue;
2485   if (!Offset->evaluateAsAbsolute(OffsetValue))
2486     return Error(OffsetLoc, "expression is not a constant value");
2487 
2488   if (Lexer.isNot(AsmToken::Comma))
2489     return TokError("expected comma");
2490   Lexer.Lex();
2491 
2492   if (Lexer.isNot(AsmToken::Identifier))
2493     return TokError("expected relocation name");
2494   SMLoc NameLoc = Lexer.getTok().getLoc();
2495   StringRef Name = Lexer.getTok().getIdentifier();
2496   Lexer.Lex();
2497 
2498   if (Lexer.is(AsmToken::Comma)) {
2499     Lexer.Lex();
2500     SMLoc ExprLoc = Lexer.getLoc();
2501     if (parseExpression(Expr))
2502       return true;
2503 
2504     MCValue Value;
2505     if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr))
2506       return Error(ExprLoc, "expression must be relocatable");
2507   }
2508 
2509   if (Lexer.isNot(AsmToken::EndOfStatement))
2510     return TokError("unexpected token in .reloc directive");
2511 
2512   if (getStreamer().EmitRelocDirective(*Offset, Name, Expr, DirectiveLoc))
2513     return Error(NameLoc, "unknown relocation name");
2514 
2515   return false;
2516 }
2517 
2518 /// parseDirectiveValue
2519 ///  ::= (.byte | .short | ... ) [ expression (, expression)* ]
parseDirectiveValue(unsigned Size)2520 bool AsmParser::parseDirectiveValue(unsigned Size) {
2521   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2522     checkForValidSection();
2523 
2524     for (;;) {
2525       const MCExpr *Value;
2526       SMLoc ExprLoc = getLexer().getLoc();
2527       if (parseExpression(Value))
2528         return true;
2529 
2530       // Special case constant expressions to match code generator.
2531       if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2532         assert(Size <= 8 && "Invalid size");
2533         uint64_t IntValue = MCE->getValue();
2534         if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
2535           return Error(ExprLoc, "literal value out of range for directive");
2536         getStreamer().EmitIntValue(IntValue, Size);
2537       } else
2538         getStreamer().EmitValue(Value, Size, ExprLoc);
2539 
2540       if (getLexer().is(AsmToken::EndOfStatement))
2541         break;
2542 
2543       // FIXME: Improve diagnostic.
2544       if (getLexer().isNot(AsmToken::Comma))
2545         return TokError("unexpected token in directive");
2546       Lex();
2547     }
2548   }
2549 
2550   Lex();
2551   return false;
2552 }
2553 
2554 /// ParseDirectiveOctaValue
2555 ///  ::= .octa [ hexconstant (, hexconstant)* ]
parseDirectiveOctaValue()2556 bool AsmParser::parseDirectiveOctaValue() {
2557   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2558     checkForValidSection();
2559 
2560     for (;;) {
2561       if (Lexer.getKind() == AsmToken::Error)
2562         return true;
2563       if (Lexer.getKind() != AsmToken::Integer &&
2564           Lexer.getKind() != AsmToken::BigNum)
2565         return TokError("unknown token in expression");
2566 
2567       SMLoc ExprLoc = getLexer().getLoc();
2568       APInt IntValue = getTok().getAPIntVal();
2569       Lex();
2570 
2571       uint64_t hi, lo;
2572       if (IntValue.isIntN(64)) {
2573         hi = 0;
2574         lo = IntValue.getZExtValue();
2575       } else if (IntValue.isIntN(128)) {
2576         // It might actually have more than 128 bits, but the top ones are zero.
2577         hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue();
2578         lo = IntValue.getLoBits(64).getZExtValue();
2579       } else
2580         return Error(ExprLoc, "literal value out of range for directive");
2581 
2582       if (MAI.isLittleEndian()) {
2583         getStreamer().EmitIntValue(lo, 8);
2584         getStreamer().EmitIntValue(hi, 8);
2585       } else {
2586         getStreamer().EmitIntValue(hi, 8);
2587         getStreamer().EmitIntValue(lo, 8);
2588       }
2589 
2590       if (getLexer().is(AsmToken::EndOfStatement))
2591         break;
2592 
2593       // FIXME: Improve diagnostic.
2594       if (getLexer().isNot(AsmToken::Comma))
2595         return TokError("unexpected token in directive");
2596       Lex();
2597     }
2598   }
2599 
2600   Lex();
2601   return false;
2602 }
2603 
2604 /// parseDirectiveRealValue
2605 ///  ::= (.single | .double) [ expression (, expression)* ]
parseDirectiveRealValue(const fltSemantics & Semantics)2606 bool AsmParser::parseDirectiveRealValue(const fltSemantics &Semantics) {
2607   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2608     checkForValidSection();
2609 
2610     for (;;) {
2611       // We don't truly support arithmetic on floating point expressions, so we
2612       // have to manually parse unary prefixes.
2613       bool IsNeg = false;
2614       if (getLexer().is(AsmToken::Minus)) {
2615         Lex();
2616         IsNeg = true;
2617       } else if (getLexer().is(AsmToken::Plus))
2618         Lex();
2619 
2620       if (getLexer().isNot(AsmToken::Integer) &&
2621           getLexer().isNot(AsmToken::Real) &&
2622           getLexer().isNot(AsmToken::Identifier))
2623         return TokError("unexpected token in directive");
2624 
2625       // Convert to an APFloat.
2626       APFloat Value(Semantics);
2627       StringRef IDVal = getTok().getString();
2628       if (getLexer().is(AsmToken::Identifier)) {
2629         if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
2630           Value = APFloat::getInf(Semantics);
2631         else if (!IDVal.compare_lower("nan"))
2632           Value = APFloat::getNaN(Semantics, false, ~0);
2633         else
2634           return TokError("invalid floating point literal");
2635       } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
2636                  APFloat::opInvalidOp)
2637         return TokError("invalid floating point literal");
2638       if (IsNeg)
2639         Value.changeSign();
2640 
2641       // Consume the numeric token.
2642       Lex();
2643 
2644       // Emit the value as an integer.
2645       APInt AsInt = Value.bitcastToAPInt();
2646       getStreamer().EmitIntValue(AsInt.getLimitedValue(),
2647                                  AsInt.getBitWidth() / 8);
2648 
2649       if (getLexer().is(AsmToken::EndOfStatement))
2650         break;
2651 
2652       if (getLexer().isNot(AsmToken::Comma))
2653         return TokError("unexpected token in directive");
2654       Lex();
2655     }
2656   }
2657 
2658   Lex();
2659   return false;
2660 }
2661 
2662 /// parseDirectiveZero
2663 ///  ::= .zero expression
parseDirectiveZero()2664 bool AsmParser::parseDirectiveZero() {
2665   checkForValidSection();
2666 
2667   int64_t NumBytes;
2668   if (parseAbsoluteExpression(NumBytes))
2669     return true;
2670 
2671   int64_t Val = 0;
2672   if (getLexer().is(AsmToken::Comma)) {
2673     Lex();
2674     if (parseAbsoluteExpression(Val))
2675       return true;
2676   }
2677 
2678   if (getLexer().isNot(AsmToken::EndOfStatement))
2679     return TokError("unexpected token in '.zero' directive");
2680 
2681   Lex();
2682 
2683   getStreamer().EmitFill(NumBytes, Val);
2684 
2685   return false;
2686 }
2687 
2688 /// parseDirectiveFill
2689 ///  ::= .fill expression [ , expression [ , expression ] ]
parseDirectiveFill()2690 bool AsmParser::parseDirectiveFill() {
2691   checkForValidSection();
2692 
2693   SMLoc RepeatLoc = getLexer().getLoc();
2694   int64_t NumValues;
2695   if (parseAbsoluteExpression(NumValues))
2696     return true;
2697 
2698   if (NumValues < 0) {
2699     Warning(RepeatLoc,
2700             "'.fill' directive with negative repeat count has no effect");
2701     NumValues = 0;
2702   }
2703 
2704   int64_t FillSize = 1;
2705   int64_t FillExpr = 0;
2706 
2707   SMLoc SizeLoc, ExprLoc;
2708   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2709     if (getLexer().isNot(AsmToken::Comma))
2710       return TokError("unexpected token in '.fill' directive");
2711     Lex();
2712 
2713     SizeLoc = getLexer().getLoc();
2714     if (parseAbsoluteExpression(FillSize))
2715       return true;
2716 
2717     if (getLexer().isNot(AsmToken::EndOfStatement)) {
2718       if (getLexer().isNot(AsmToken::Comma))
2719         return TokError("unexpected token in '.fill' directive");
2720       Lex();
2721 
2722       ExprLoc = getLexer().getLoc();
2723       if (parseAbsoluteExpression(FillExpr))
2724         return true;
2725 
2726       if (getLexer().isNot(AsmToken::EndOfStatement))
2727         return TokError("unexpected token in '.fill' directive");
2728 
2729       Lex();
2730     }
2731   }
2732 
2733   if (FillSize < 0) {
2734     Warning(SizeLoc, "'.fill' directive with negative size has no effect");
2735     NumValues = 0;
2736   }
2737   if (FillSize > 8) {
2738     Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8");
2739     FillSize = 8;
2740   }
2741 
2742   if (!isUInt<32>(FillExpr) && FillSize > 4)
2743     Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits");
2744 
2745   if (NumValues > 0) {
2746     int64_t NonZeroFillSize = FillSize > 4 ? 4 : FillSize;
2747     FillExpr &= ~0ULL >> (64 - NonZeroFillSize * 8);
2748     for (uint64_t i = 0, e = NumValues; i != e; ++i) {
2749       getStreamer().EmitIntValue(FillExpr, NonZeroFillSize);
2750       if (NonZeroFillSize < FillSize)
2751         getStreamer().EmitIntValue(0, FillSize - NonZeroFillSize);
2752     }
2753   }
2754 
2755   return false;
2756 }
2757 
2758 /// parseDirectiveOrg
2759 ///  ::= .org expression [ , expression ]
parseDirectiveOrg()2760 bool AsmParser::parseDirectiveOrg() {
2761   checkForValidSection();
2762 
2763   const MCExpr *Offset;
2764   if (parseExpression(Offset))
2765     return true;
2766 
2767   // Parse optional fill expression.
2768   int64_t FillExpr = 0;
2769   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2770     if (getLexer().isNot(AsmToken::Comma))
2771       return TokError("unexpected token in '.org' directive");
2772     Lex();
2773 
2774     if (parseAbsoluteExpression(FillExpr))
2775       return true;
2776 
2777     if (getLexer().isNot(AsmToken::EndOfStatement))
2778       return TokError("unexpected token in '.org' directive");
2779   }
2780 
2781   Lex();
2782   getStreamer().emitValueToOffset(Offset, FillExpr);
2783   return false;
2784 }
2785 
2786 /// parseDirectiveAlign
2787 ///  ::= {.align, ...} expression [ , expression [ , expression ]]
parseDirectiveAlign(bool IsPow2,unsigned ValueSize)2788 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
2789   checkForValidSection();
2790 
2791   SMLoc AlignmentLoc = getLexer().getLoc();
2792   int64_t Alignment;
2793   if (parseAbsoluteExpression(Alignment))
2794     return true;
2795 
2796   SMLoc MaxBytesLoc;
2797   bool HasFillExpr = false;
2798   int64_t FillExpr = 0;
2799   int64_t MaxBytesToFill = 0;
2800   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2801     if (getLexer().isNot(AsmToken::Comma))
2802       return TokError("unexpected token in directive");
2803     Lex();
2804 
2805     // The fill expression can be omitted while specifying a maximum number of
2806     // alignment bytes, e.g:
2807     //  .align 3,,4
2808     if (getLexer().isNot(AsmToken::Comma)) {
2809       HasFillExpr = true;
2810       if (parseAbsoluteExpression(FillExpr))
2811         return true;
2812     }
2813 
2814     if (getLexer().isNot(AsmToken::EndOfStatement)) {
2815       if (getLexer().isNot(AsmToken::Comma))
2816         return TokError("unexpected token in directive");
2817       Lex();
2818 
2819       MaxBytesLoc = getLexer().getLoc();
2820       if (parseAbsoluteExpression(MaxBytesToFill))
2821         return true;
2822 
2823       if (getLexer().isNot(AsmToken::EndOfStatement))
2824         return TokError("unexpected token in directive");
2825     }
2826   }
2827 
2828   Lex();
2829 
2830   if (!HasFillExpr)
2831     FillExpr = 0;
2832 
2833   // Compute alignment in bytes.
2834   if (IsPow2) {
2835     // FIXME: Diagnose overflow.
2836     if (Alignment >= 32) {
2837       Error(AlignmentLoc, "invalid alignment value");
2838       Alignment = 31;
2839     }
2840 
2841     Alignment = 1ULL << Alignment;
2842   } else {
2843     // Reject alignments that aren't either a power of two or zero,
2844     // for gas compatibility. Alignment of zero is silently rounded
2845     // up to one.
2846     if (Alignment == 0)
2847       Alignment = 1;
2848     if (!isPowerOf2_64(Alignment))
2849       Error(AlignmentLoc, "alignment must be a power of 2");
2850   }
2851 
2852   // Diagnose non-sensical max bytes to align.
2853   if (MaxBytesLoc.isValid()) {
2854     if (MaxBytesToFill < 1) {
2855       Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
2856                          "many bytes, ignoring maximum bytes expression");
2857       MaxBytesToFill = 0;
2858     }
2859 
2860     if (MaxBytesToFill >= Alignment) {
2861       Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2862                            "has no effect");
2863       MaxBytesToFill = 0;
2864     }
2865   }
2866 
2867   // Check whether we should use optimal code alignment for this .align
2868   // directive.
2869   const MCSection *Section = getStreamer().getCurrentSection().first;
2870   assert(Section && "must have section to emit alignment");
2871   bool UseCodeAlign = Section->UseCodeAlign();
2872   if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2873       ValueSize == 1 && UseCodeAlign) {
2874     getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2875   } else {
2876     // FIXME: Target specific behavior about how the "extra" bytes are filled.
2877     getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
2878                                        MaxBytesToFill);
2879   }
2880 
2881   return false;
2882 }
2883 
2884 /// parseDirectiveFile
2885 /// ::= .file [number] filename
2886 /// ::= .file number directory filename
parseDirectiveFile(SMLoc DirectiveLoc)2887 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) {
2888   // FIXME: I'm not sure what this is.
2889   int64_t FileNumber = -1;
2890   SMLoc FileNumberLoc = getLexer().getLoc();
2891   if (getLexer().is(AsmToken::Integer)) {
2892     FileNumber = getTok().getIntVal();
2893     Lex();
2894 
2895     if (FileNumber < 1)
2896       return TokError("file number less than one");
2897   }
2898 
2899   if (getLexer().isNot(AsmToken::String))
2900     return TokError("unexpected token in '.file' directive");
2901 
2902   // Usually the directory and filename together, otherwise just the directory.
2903   // Allow the strings to have escaped octal character sequence.
2904   std::string Path = getTok().getString();
2905   if (parseEscapedString(Path))
2906     return true;
2907   Lex();
2908 
2909   StringRef Directory;
2910   StringRef Filename;
2911   std::string FilenameData;
2912   if (getLexer().is(AsmToken::String)) {
2913     if (FileNumber == -1)
2914       return TokError("explicit path specified, but no file number");
2915     if (parseEscapedString(FilenameData))
2916       return true;
2917     Filename = FilenameData;
2918     Directory = Path;
2919     Lex();
2920   } else {
2921     Filename = Path;
2922   }
2923 
2924   if (getLexer().isNot(AsmToken::EndOfStatement))
2925     return TokError("unexpected token in '.file' directive");
2926 
2927   if (FileNumber == -1)
2928     getStreamer().EmitFileDirective(Filename);
2929   else {
2930     if (getContext().getGenDwarfForAssembly())
2931       Error(DirectiveLoc,
2932             "input can't have .file dwarf directives when -g is "
2933             "used to generate dwarf debug info for assembly code");
2934 
2935     if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename) ==
2936         0)
2937       Error(FileNumberLoc, "file number already allocated");
2938   }
2939 
2940   return false;
2941 }
2942 
2943 /// parseDirectiveLine
2944 /// ::= .line [number]
parseDirectiveLine()2945 bool AsmParser::parseDirectiveLine() {
2946   if (getLexer().isNot(AsmToken::EndOfStatement)) {
2947     if (getLexer().isNot(AsmToken::Integer))
2948       return TokError("unexpected token in '.line' directive");
2949 
2950     int64_t LineNumber = getTok().getIntVal();
2951     (void)LineNumber;
2952     Lex();
2953 
2954     // FIXME: Do something with the .line.
2955   }
2956 
2957   if (getLexer().isNot(AsmToken::EndOfStatement))
2958     return TokError("unexpected token in '.line' directive");
2959 
2960   return false;
2961 }
2962 
2963 /// parseDirectiveLoc
2964 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2965 ///                                [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2966 /// The first number is a file number, must have been previously assigned with
2967 /// a .file directive, the second number is the line number and optionally the
2968 /// third number is a column position (zero if not specified).  The remaining
2969 /// optional items are .loc sub-directives.
parseDirectiveLoc()2970 bool AsmParser::parseDirectiveLoc() {
2971   if (getLexer().isNot(AsmToken::Integer))
2972     return TokError("unexpected token in '.loc' directive");
2973   int64_t FileNumber = getTok().getIntVal();
2974   if (FileNumber < 1)
2975     return TokError("file number less than one in '.loc' directive");
2976   if (!getContext().isValidDwarfFileNumber(FileNumber))
2977     return TokError("unassigned file number in '.loc' directive");
2978   Lex();
2979 
2980   int64_t LineNumber = 0;
2981   if (getLexer().is(AsmToken::Integer)) {
2982     LineNumber = getTok().getIntVal();
2983     if (LineNumber < 0)
2984       return TokError("line number less than zero in '.loc' directive");
2985     Lex();
2986   }
2987 
2988   int64_t ColumnPos = 0;
2989   if (getLexer().is(AsmToken::Integer)) {
2990     ColumnPos = getTok().getIntVal();
2991     if (ColumnPos < 0)
2992       return TokError("column position less than zero in '.loc' directive");
2993     Lex();
2994   }
2995 
2996   unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2997   unsigned Isa = 0;
2998   int64_t Discriminator = 0;
2999   if (getLexer().isNot(AsmToken::EndOfStatement)) {
3000     for (;;) {
3001       if (getLexer().is(AsmToken::EndOfStatement))
3002         break;
3003 
3004       StringRef Name;
3005       SMLoc Loc = getTok().getLoc();
3006       if (parseIdentifier(Name))
3007         return TokError("unexpected token in '.loc' directive");
3008 
3009       if (Name == "basic_block")
3010         Flags |= DWARF2_FLAG_BASIC_BLOCK;
3011       else if (Name == "prologue_end")
3012         Flags |= DWARF2_FLAG_PROLOGUE_END;
3013       else if (Name == "epilogue_begin")
3014         Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
3015       else if (Name == "is_stmt") {
3016         Loc = getTok().getLoc();
3017         const MCExpr *Value;
3018         if (parseExpression(Value))
3019           return true;
3020         // The expression must be the constant 0 or 1.
3021         if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3022           int Value = MCE->getValue();
3023           if (Value == 0)
3024             Flags &= ~DWARF2_FLAG_IS_STMT;
3025           else if (Value == 1)
3026             Flags |= DWARF2_FLAG_IS_STMT;
3027           else
3028             return Error(Loc, "is_stmt value not 0 or 1");
3029         } else {
3030           return Error(Loc, "is_stmt value not the constant value of 0 or 1");
3031         }
3032       } else if (Name == "isa") {
3033         Loc = getTok().getLoc();
3034         const MCExpr *Value;
3035         if (parseExpression(Value))
3036           return true;
3037         // The expression must be a constant greater or equal to 0.
3038         if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3039           int Value = MCE->getValue();
3040           if (Value < 0)
3041             return Error(Loc, "isa number less than zero");
3042           Isa = Value;
3043         } else {
3044           return Error(Loc, "isa number not a constant value");
3045         }
3046       } else if (Name == "discriminator") {
3047         if (parseAbsoluteExpression(Discriminator))
3048           return true;
3049       } else {
3050         return Error(Loc, "unknown sub-directive in '.loc' directive");
3051       }
3052 
3053       if (getLexer().is(AsmToken::EndOfStatement))
3054         break;
3055     }
3056   }
3057 
3058   getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
3059                                       Isa, Discriminator, StringRef());
3060 
3061   return false;
3062 }
3063 
3064 /// parseDirectiveStabs
3065 /// ::= .stabs string, number, number, number
parseDirectiveStabs()3066 bool AsmParser::parseDirectiveStabs() {
3067   return TokError("unsupported directive '.stabs'");
3068 }
3069 
3070 /// parseDirectiveCFISections
3071 /// ::= .cfi_sections section [, section]
parseDirectiveCFISections()3072 bool AsmParser::parseDirectiveCFISections() {
3073   StringRef Name;
3074   bool EH = false;
3075   bool Debug = false;
3076 
3077   if (parseIdentifier(Name))
3078     return TokError("Expected an identifier");
3079 
3080   if (Name == ".eh_frame")
3081     EH = true;
3082   else if (Name == ".debug_frame")
3083     Debug = true;
3084 
3085   if (getLexer().is(AsmToken::Comma)) {
3086     Lex();
3087 
3088     if (parseIdentifier(Name))
3089       return TokError("Expected an identifier");
3090 
3091     if (Name == ".eh_frame")
3092       EH = true;
3093     else if (Name == ".debug_frame")
3094       Debug = true;
3095   }
3096 
3097   getStreamer().EmitCFISections(EH, Debug);
3098   return false;
3099 }
3100 
3101 /// parseDirectiveCFIStartProc
3102 /// ::= .cfi_startproc [simple]
parseDirectiveCFIStartProc()3103 bool AsmParser::parseDirectiveCFIStartProc() {
3104   StringRef Simple;
3105   if (getLexer().isNot(AsmToken::EndOfStatement))
3106     if (parseIdentifier(Simple) || Simple != "simple")
3107       return TokError("unexpected token in .cfi_startproc directive");
3108 
3109   getStreamer().EmitCFIStartProc(!Simple.empty());
3110   return false;
3111 }
3112 
3113 /// parseDirectiveCFIEndProc
3114 /// ::= .cfi_endproc
parseDirectiveCFIEndProc()3115 bool AsmParser::parseDirectiveCFIEndProc() {
3116   getStreamer().EmitCFIEndProc();
3117   return false;
3118 }
3119 
3120 /// \brief parse register name or number.
parseRegisterOrRegisterNumber(int64_t & Register,SMLoc DirectiveLoc)3121 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register,
3122                                               SMLoc DirectiveLoc) {
3123   unsigned RegNo;
3124 
3125   if (getLexer().isNot(AsmToken::Integer)) {
3126     if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc))
3127       return true;
3128     Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true);
3129   } else
3130     return parseAbsoluteExpression(Register);
3131 
3132   return false;
3133 }
3134 
3135 /// parseDirectiveCFIDefCfa
3136 /// ::= .cfi_def_cfa register,  offset
parseDirectiveCFIDefCfa(SMLoc DirectiveLoc)3137 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) {
3138   int64_t Register = 0;
3139   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3140     return true;
3141 
3142   if (getLexer().isNot(AsmToken::Comma))
3143     return TokError("unexpected token in directive");
3144   Lex();
3145 
3146   int64_t Offset = 0;
3147   if (parseAbsoluteExpression(Offset))
3148     return true;
3149 
3150   getStreamer().EmitCFIDefCfa(Register, Offset);
3151   return false;
3152 }
3153 
3154 /// parseDirectiveCFIDefCfaOffset
3155 /// ::= .cfi_def_cfa_offset offset
parseDirectiveCFIDefCfaOffset()3156 bool AsmParser::parseDirectiveCFIDefCfaOffset() {
3157   int64_t Offset = 0;
3158   if (parseAbsoluteExpression(Offset))
3159     return true;
3160 
3161   getStreamer().EmitCFIDefCfaOffset(Offset);
3162   return false;
3163 }
3164 
3165 /// parseDirectiveCFIRegister
3166 /// ::= .cfi_register register, register
parseDirectiveCFIRegister(SMLoc DirectiveLoc)3167 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) {
3168   int64_t Register1 = 0;
3169   if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc))
3170     return true;
3171 
3172   if (getLexer().isNot(AsmToken::Comma))
3173     return TokError("unexpected token in directive");
3174   Lex();
3175 
3176   int64_t Register2 = 0;
3177   if (parseRegisterOrRegisterNumber(Register2, DirectiveLoc))
3178     return true;
3179 
3180   getStreamer().EmitCFIRegister(Register1, Register2);
3181   return false;
3182 }
3183 
3184 /// parseDirectiveCFIWindowSave
3185 /// ::= .cfi_window_save
parseDirectiveCFIWindowSave()3186 bool AsmParser::parseDirectiveCFIWindowSave() {
3187   getStreamer().EmitCFIWindowSave();
3188   return false;
3189 }
3190 
3191 /// parseDirectiveCFIAdjustCfaOffset
3192 /// ::= .cfi_adjust_cfa_offset adjustment
parseDirectiveCFIAdjustCfaOffset()3193 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() {
3194   int64_t Adjustment = 0;
3195   if (parseAbsoluteExpression(Adjustment))
3196     return true;
3197 
3198   getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
3199   return false;
3200 }
3201 
3202 /// parseDirectiveCFIDefCfaRegister
3203 /// ::= .cfi_def_cfa_register register
parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc)3204 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) {
3205   int64_t Register = 0;
3206   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3207     return true;
3208 
3209   getStreamer().EmitCFIDefCfaRegister(Register);
3210   return false;
3211 }
3212 
3213 /// parseDirectiveCFIOffset
3214 /// ::= .cfi_offset register, offset
parseDirectiveCFIOffset(SMLoc DirectiveLoc)3215 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) {
3216   int64_t Register = 0;
3217   int64_t Offset = 0;
3218 
3219   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3220     return true;
3221 
3222   if (getLexer().isNot(AsmToken::Comma))
3223     return TokError("unexpected token in directive");
3224   Lex();
3225 
3226   if (parseAbsoluteExpression(Offset))
3227     return true;
3228 
3229   getStreamer().EmitCFIOffset(Register, Offset);
3230   return false;
3231 }
3232 
3233 /// parseDirectiveCFIRelOffset
3234 /// ::= .cfi_rel_offset register, offset
parseDirectiveCFIRelOffset(SMLoc DirectiveLoc)3235 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) {
3236   int64_t Register = 0;
3237 
3238   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3239     return true;
3240 
3241   if (getLexer().isNot(AsmToken::Comma))
3242     return TokError("unexpected token in directive");
3243   Lex();
3244 
3245   int64_t Offset = 0;
3246   if (parseAbsoluteExpression(Offset))
3247     return true;
3248 
3249   getStreamer().EmitCFIRelOffset(Register, Offset);
3250   return false;
3251 }
3252 
isValidEncoding(int64_t Encoding)3253 static bool isValidEncoding(int64_t Encoding) {
3254   if (Encoding & ~0xff)
3255     return false;
3256 
3257   if (Encoding == dwarf::DW_EH_PE_omit)
3258     return true;
3259 
3260   const unsigned Format = Encoding & 0xf;
3261   if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
3262       Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
3263       Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
3264       Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
3265     return false;
3266 
3267   const unsigned Application = Encoding & 0x70;
3268   if (Application != dwarf::DW_EH_PE_absptr &&
3269       Application != dwarf::DW_EH_PE_pcrel)
3270     return false;
3271 
3272   return true;
3273 }
3274 
3275 /// parseDirectiveCFIPersonalityOrLsda
3276 /// IsPersonality true for cfi_personality, false for cfi_lsda
3277 /// ::= .cfi_personality encoding, [symbol_name]
3278 /// ::= .cfi_lsda encoding, [symbol_name]
parseDirectiveCFIPersonalityOrLsda(bool IsPersonality)3279 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) {
3280   int64_t Encoding = 0;
3281   if (parseAbsoluteExpression(Encoding))
3282     return true;
3283   if (Encoding == dwarf::DW_EH_PE_omit)
3284     return false;
3285 
3286   if (!isValidEncoding(Encoding))
3287     return TokError("unsupported encoding.");
3288 
3289   if (getLexer().isNot(AsmToken::Comma))
3290     return TokError("unexpected token in directive");
3291   Lex();
3292 
3293   StringRef Name;
3294   if (parseIdentifier(Name))
3295     return TokError("expected identifier in directive");
3296 
3297   MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
3298 
3299   if (IsPersonality)
3300     getStreamer().EmitCFIPersonality(Sym, Encoding);
3301   else
3302     getStreamer().EmitCFILsda(Sym, Encoding);
3303   return false;
3304 }
3305 
3306 /// parseDirectiveCFIRememberState
3307 /// ::= .cfi_remember_state
parseDirectiveCFIRememberState()3308 bool AsmParser::parseDirectiveCFIRememberState() {
3309   getStreamer().EmitCFIRememberState();
3310   return false;
3311 }
3312 
3313 /// parseDirectiveCFIRestoreState
3314 /// ::= .cfi_remember_state
parseDirectiveCFIRestoreState()3315 bool AsmParser::parseDirectiveCFIRestoreState() {
3316   getStreamer().EmitCFIRestoreState();
3317   return false;
3318 }
3319 
3320 /// parseDirectiveCFISameValue
3321 /// ::= .cfi_same_value register
parseDirectiveCFISameValue(SMLoc DirectiveLoc)3322 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) {
3323   int64_t Register = 0;
3324 
3325   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3326     return true;
3327 
3328   getStreamer().EmitCFISameValue(Register);
3329   return false;
3330 }
3331 
3332 /// parseDirectiveCFIRestore
3333 /// ::= .cfi_restore register
parseDirectiveCFIRestore(SMLoc DirectiveLoc)3334 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) {
3335   int64_t Register = 0;
3336   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3337     return true;
3338 
3339   getStreamer().EmitCFIRestore(Register);
3340   return false;
3341 }
3342 
3343 /// parseDirectiveCFIEscape
3344 /// ::= .cfi_escape expression[,...]
parseDirectiveCFIEscape()3345 bool AsmParser::parseDirectiveCFIEscape() {
3346   std::string Values;
3347   int64_t CurrValue;
3348   if (parseAbsoluteExpression(CurrValue))
3349     return true;
3350 
3351   Values.push_back((uint8_t)CurrValue);
3352 
3353   while (getLexer().is(AsmToken::Comma)) {
3354     Lex();
3355 
3356     if (parseAbsoluteExpression(CurrValue))
3357       return true;
3358 
3359     Values.push_back((uint8_t)CurrValue);
3360   }
3361 
3362   getStreamer().EmitCFIEscape(Values);
3363   return false;
3364 }
3365 
3366 /// parseDirectiveCFISignalFrame
3367 /// ::= .cfi_signal_frame
parseDirectiveCFISignalFrame()3368 bool AsmParser::parseDirectiveCFISignalFrame() {
3369   if (getLexer().isNot(AsmToken::EndOfStatement))
3370     return Error(getLexer().getLoc(),
3371                  "unexpected token in '.cfi_signal_frame'");
3372 
3373   getStreamer().EmitCFISignalFrame();
3374   return false;
3375 }
3376 
3377 /// parseDirectiveCFIUndefined
3378 /// ::= .cfi_undefined register
parseDirectiveCFIUndefined(SMLoc DirectiveLoc)3379 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) {
3380   int64_t Register = 0;
3381 
3382   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3383     return true;
3384 
3385   getStreamer().EmitCFIUndefined(Register);
3386   return false;
3387 }
3388 
3389 /// parseDirectiveMacrosOnOff
3390 /// ::= .macros_on
3391 /// ::= .macros_off
parseDirectiveMacrosOnOff(StringRef Directive)3392 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) {
3393   if (getLexer().isNot(AsmToken::EndOfStatement))
3394     return Error(getLexer().getLoc(),
3395                  "unexpected token in '" + Directive + "' directive");
3396 
3397   setMacrosEnabled(Directive == ".macros_on");
3398   return false;
3399 }
3400 
3401 /// parseDirectiveMacro
3402 /// ::= .macro name[,] [parameters]
parseDirectiveMacro(SMLoc DirectiveLoc)3403 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) {
3404   StringRef Name;
3405   if (parseIdentifier(Name))
3406     return TokError("expected identifier in '.macro' directive");
3407 
3408   if (getLexer().is(AsmToken::Comma))
3409     Lex();
3410 
3411   MCAsmMacroParameters Parameters;
3412   while (getLexer().isNot(AsmToken::EndOfStatement)) {
3413 
3414     if (!Parameters.empty() && Parameters.back().Vararg)
3415       return Error(Lexer.getLoc(),
3416                    "Vararg parameter '" + Parameters.back().Name +
3417                    "' should be last one in the list of parameters.");
3418 
3419     MCAsmMacroParameter Parameter;
3420     if (parseIdentifier(Parameter.Name))
3421       return TokError("expected identifier in '.macro' directive");
3422 
3423     if (Lexer.is(AsmToken::Colon)) {
3424       Lex();  // consume ':'
3425 
3426       SMLoc QualLoc;
3427       StringRef Qualifier;
3428 
3429       QualLoc = Lexer.getLoc();
3430       if (parseIdentifier(Qualifier))
3431         return Error(QualLoc, "missing parameter qualifier for "
3432                      "'" + Parameter.Name + "' in macro '" + Name + "'");
3433 
3434       if (Qualifier == "req")
3435         Parameter.Required = true;
3436       else if (Qualifier == "vararg")
3437         Parameter.Vararg = true;
3438       else
3439         return Error(QualLoc, Qualifier + " is not a valid parameter qualifier "
3440                      "for '" + Parameter.Name + "' in macro '" + Name + "'");
3441     }
3442 
3443     if (getLexer().is(AsmToken::Equal)) {
3444       Lex();
3445 
3446       SMLoc ParamLoc;
3447 
3448       ParamLoc = Lexer.getLoc();
3449       if (parseMacroArgument(Parameter.Value, /*Vararg=*/false ))
3450         return true;
3451 
3452       if (Parameter.Required)
3453         Warning(ParamLoc, "pointless default value for required parameter "
3454                 "'" + Parameter.Name + "' in macro '" + Name + "'");
3455     }
3456 
3457     Parameters.push_back(std::move(Parameter));
3458 
3459     if (getLexer().is(AsmToken::Comma))
3460       Lex();
3461   }
3462 
3463   // Eat the end of statement.
3464   Lex();
3465 
3466   AsmToken EndToken, StartToken = getTok();
3467   unsigned MacroDepth = 0;
3468 
3469   // Lex the macro definition.
3470   for (;;) {
3471     // Check whether we have reached the end of the file.
3472     if (getLexer().is(AsmToken::Eof))
3473       return Error(DirectiveLoc, "no matching '.endmacro' in definition");
3474 
3475     // Otherwise, check whether we have reach the .endmacro.
3476     if (getLexer().is(AsmToken::Identifier)) {
3477       if (getTok().getIdentifier() == ".endm" ||
3478           getTok().getIdentifier() == ".endmacro") {
3479         if (MacroDepth == 0) { // Outermost macro.
3480           EndToken = getTok();
3481           Lex();
3482           if (getLexer().isNot(AsmToken::EndOfStatement))
3483             return TokError("unexpected token in '" + EndToken.getIdentifier() +
3484                             "' directive");
3485           break;
3486         } else {
3487           // Otherwise we just found the end of an inner macro.
3488           --MacroDepth;
3489         }
3490       } else if (getTok().getIdentifier() == ".macro") {
3491         // We allow nested macros. Those aren't instantiated until the outermost
3492         // macro is expanded so just ignore them for now.
3493         ++MacroDepth;
3494       }
3495     }
3496 
3497     // Otherwise, scan til the end of the statement.
3498     eatToEndOfStatement();
3499   }
3500 
3501   if (lookupMacro(Name)) {
3502     return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
3503   }
3504 
3505   const char *BodyStart = StartToken.getLoc().getPointer();
3506   const char *BodyEnd = EndToken.getLoc().getPointer();
3507   StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3508   checkForBadMacro(DirectiveLoc, Name, Body, Parameters);
3509   defineMacro(Name, MCAsmMacro(Name, Body, std::move(Parameters)));
3510   return false;
3511 }
3512 
3513 /// checkForBadMacro
3514 ///
3515 /// With the support added for named parameters there may be code out there that
3516 /// is transitioning from positional parameters.  In versions of gas that did
3517 /// not support named parameters they would be ignored on the macro definition.
3518 /// But to support both styles of parameters this is not possible so if a macro
3519 /// definition has named parameters but does not use them and has what appears
3520 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a
3521 /// warning that the positional parameter found in body which have no effect.
3522 /// Hoping the developer will either remove the named parameters from the macro
3523 /// definition so the positional parameters get used if that was what was
3524 /// intended or change the macro to use the named parameters.  It is possible
3525 /// this warning will trigger when the none of the named parameters are used
3526 /// and the strings like $1 are infact to simply to be passed trough unchanged.
checkForBadMacro(SMLoc DirectiveLoc,StringRef Name,StringRef Body,ArrayRef<MCAsmMacroParameter> Parameters)3527 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name,
3528                                  StringRef Body,
3529                                  ArrayRef<MCAsmMacroParameter> Parameters) {
3530   // If this macro is not defined with named parameters the warning we are
3531   // checking for here doesn't apply.
3532   unsigned NParameters = Parameters.size();
3533   if (NParameters == 0)
3534     return;
3535 
3536   bool NamedParametersFound = false;
3537   bool PositionalParametersFound = false;
3538 
3539   // Look at the body of the macro for use of both the named parameters and what
3540   // are likely to be positional parameters.  This is what expandMacro() is
3541   // doing when it finds the parameters in the body.
3542   while (!Body.empty()) {
3543     // Scan for the next possible parameter.
3544     std::size_t End = Body.size(), Pos = 0;
3545     for (; Pos != End; ++Pos) {
3546       // Check for a substitution or escape.
3547       // This macro is defined with parameters, look for \foo, \bar, etc.
3548       if (Body[Pos] == '\\' && Pos + 1 != End)
3549         break;
3550 
3551       // This macro should have parameters, but look for $0, $1, ..., $n too.
3552       if (Body[Pos] != '$' || Pos + 1 == End)
3553         continue;
3554       char Next = Body[Pos + 1];
3555       if (Next == '$' || Next == 'n' ||
3556           isdigit(static_cast<unsigned char>(Next)))
3557         break;
3558     }
3559 
3560     // Check if we reached the end.
3561     if (Pos == End)
3562       break;
3563 
3564     if (Body[Pos] == '$') {
3565       switch (Body[Pos + 1]) {
3566       // $$ => $
3567       case '$':
3568         break;
3569 
3570       // $n => number of arguments
3571       case 'n':
3572         PositionalParametersFound = true;
3573         break;
3574 
3575       // $[0-9] => argument
3576       default: {
3577         PositionalParametersFound = true;
3578         break;
3579       }
3580       }
3581       Pos += 2;
3582     } else {
3583       unsigned I = Pos + 1;
3584       while (isIdentifierChar(Body[I]) && I + 1 != End)
3585         ++I;
3586 
3587       const char *Begin = Body.data() + Pos + 1;
3588       StringRef Argument(Begin, I - (Pos + 1));
3589       unsigned Index = 0;
3590       for (; Index < NParameters; ++Index)
3591         if (Parameters[Index].Name == Argument)
3592           break;
3593 
3594       if (Index == NParameters) {
3595         if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
3596           Pos += 3;
3597         else {
3598           Pos = I;
3599         }
3600       } else {
3601         NamedParametersFound = true;
3602         Pos += 1 + Argument.size();
3603       }
3604     }
3605     // Update the scan point.
3606     Body = Body.substr(Pos);
3607   }
3608 
3609   if (!NamedParametersFound && PositionalParametersFound)
3610     Warning(DirectiveLoc, "macro defined with named parameters which are not "
3611                           "used in macro body, possible positional parameter "
3612                           "found in body which will have no effect");
3613 }
3614 
3615 /// parseDirectiveExitMacro
3616 /// ::= .exitm
parseDirectiveExitMacro(StringRef Directive)3617 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) {
3618   if (getLexer().isNot(AsmToken::EndOfStatement))
3619     return TokError("unexpected token in '" + Directive + "' directive");
3620 
3621   if (!isInsideMacroInstantiation())
3622     return TokError("unexpected '" + Directive + "' in file, "
3623                                                  "no current macro definition");
3624 
3625   // Exit all conditionals that are active in the current macro.
3626   while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) {
3627     TheCondState = TheCondStack.back();
3628     TheCondStack.pop_back();
3629   }
3630 
3631   handleMacroExit();
3632   return false;
3633 }
3634 
3635 /// parseDirectiveEndMacro
3636 /// ::= .endm
3637 /// ::= .endmacro
parseDirectiveEndMacro(StringRef Directive)3638 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) {
3639   if (getLexer().isNot(AsmToken::EndOfStatement))
3640     return TokError("unexpected token in '" + Directive + "' directive");
3641 
3642   // If we are inside a macro instantiation, terminate the current
3643   // instantiation.
3644   if (isInsideMacroInstantiation()) {
3645     handleMacroExit();
3646     return false;
3647   }
3648 
3649   // Otherwise, this .endmacro is a stray entry in the file; well formed
3650   // .endmacro directives are handled during the macro definition parsing.
3651   return TokError("unexpected '" + Directive + "' in file, "
3652                                                "no current macro definition");
3653 }
3654 
3655 /// parseDirectivePurgeMacro
3656 /// ::= .purgem
parseDirectivePurgeMacro(SMLoc DirectiveLoc)3657 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) {
3658   StringRef Name;
3659   if (parseIdentifier(Name))
3660     return TokError("expected identifier in '.purgem' directive");
3661 
3662   if (getLexer().isNot(AsmToken::EndOfStatement))
3663     return TokError("unexpected token in '.purgem' directive");
3664 
3665   if (!lookupMacro(Name))
3666     return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
3667 
3668   undefineMacro(Name);
3669   return false;
3670 }
3671 
3672 /// parseDirectiveBundleAlignMode
3673 /// ::= {.bundle_align_mode} expression
parseDirectiveBundleAlignMode()3674 bool AsmParser::parseDirectiveBundleAlignMode() {
3675   checkForValidSection();
3676 
3677   // Expect a single argument: an expression that evaluates to a constant
3678   // in the inclusive range 0-30.
3679   SMLoc ExprLoc = getLexer().getLoc();
3680   int64_t AlignSizePow2;
3681   if (parseAbsoluteExpression(AlignSizePow2))
3682     return true;
3683   else if (getLexer().isNot(AsmToken::EndOfStatement))
3684     return TokError("unexpected token after expression in"
3685                     " '.bundle_align_mode' directive");
3686   else if (AlignSizePow2 < 0 || AlignSizePow2 > 30)
3687     return Error(ExprLoc,
3688                  "invalid bundle alignment size (expected between 0 and 30)");
3689 
3690   Lex();
3691 
3692   // Because of AlignSizePow2's verified range we can safely truncate it to
3693   // unsigned.
3694   getStreamer().EmitBundleAlignMode(static_cast<unsigned>(AlignSizePow2));
3695   return false;
3696 }
3697 
3698 /// parseDirectiveBundleLock
3699 /// ::= {.bundle_lock} [align_to_end]
parseDirectiveBundleLock()3700 bool AsmParser::parseDirectiveBundleLock() {
3701   checkForValidSection();
3702   bool AlignToEnd = false;
3703 
3704   if (getLexer().isNot(AsmToken::EndOfStatement)) {
3705     StringRef Option;
3706     SMLoc Loc = getTok().getLoc();
3707     const char *kInvalidOptionError =
3708         "invalid option for '.bundle_lock' directive";
3709 
3710     if (parseIdentifier(Option))
3711       return Error(Loc, kInvalidOptionError);
3712 
3713     if (Option != "align_to_end")
3714       return Error(Loc, kInvalidOptionError);
3715     else if (getLexer().isNot(AsmToken::EndOfStatement))
3716       return Error(Loc,
3717                    "unexpected token after '.bundle_lock' directive option");
3718     AlignToEnd = true;
3719   }
3720 
3721   Lex();
3722 
3723   getStreamer().EmitBundleLock(AlignToEnd);
3724   return false;
3725 }
3726 
3727 /// parseDirectiveBundleLock
3728 /// ::= {.bundle_lock}
parseDirectiveBundleUnlock()3729 bool AsmParser::parseDirectiveBundleUnlock() {
3730   checkForValidSection();
3731 
3732   if (getLexer().isNot(AsmToken::EndOfStatement))
3733     return TokError("unexpected token in '.bundle_unlock' directive");
3734   Lex();
3735 
3736   getStreamer().EmitBundleUnlock();
3737   return false;
3738 }
3739 
3740 /// parseDirectiveSpace
3741 /// ::= (.skip | .space) expression [ , expression ]
parseDirectiveSpace(StringRef IDVal)3742 bool AsmParser::parseDirectiveSpace(StringRef IDVal) {
3743   checkForValidSection();
3744 
3745   int64_t NumBytes;
3746   if (parseAbsoluteExpression(NumBytes))
3747     return true;
3748 
3749   int64_t FillExpr = 0;
3750   if (getLexer().isNot(AsmToken::EndOfStatement)) {
3751     if (getLexer().isNot(AsmToken::Comma))
3752       return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
3753     Lex();
3754 
3755     if (parseAbsoluteExpression(FillExpr))
3756       return true;
3757 
3758     if (getLexer().isNot(AsmToken::EndOfStatement))
3759       return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
3760   }
3761 
3762   Lex();
3763 
3764   if (NumBytes <= 0)
3765     return TokError("invalid number of bytes in '" + Twine(IDVal) +
3766                     "' directive");
3767 
3768   // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
3769   getStreamer().EmitFill(NumBytes, FillExpr);
3770 
3771   return false;
3772 }
3773 
3774 /// parseDirectiveLEB128
3775 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ]
parseDirectiveLEB128(bool Signed)3776 bool AsmParser::parseDirectiveLEB128(bool Signed) {
3777   checkForValidSection();
3778   const MCExpr *Value;
3779 
3780   for (;;) {
3781     if (parseExpression(Value))
3782       return true;
3783 
3784     if (Signed)
3785       getStreamer().EmitSLEB128Value(Value);
3786     else
3787       getStreamer().EmitULEB128Value(Value);
3788 
3789     if (getLexer().is(AsmToken::EndOfStatement))
3790       break;
3791 
3792     if (getLexer().isNot(AsmToken::Comma))
3793       return TokError("unexpected token in directive");
3794     Lex();
3795   }
3796 
3797   return false;
3798 }
3799 
3800 /// parseDirectiveSymbolAttribute
3801 ///  ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
parseDirectiveSymbolAttribute(MCSymbolAttr Attr)3802 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
3803   if (getLexer().isNot(AsmToken::EndOfStatement)) {
3804     for (;;) {
3805       StringRef Name;
3806       SMLoc Loc = getTok().getLoc();
3807 
3808       if (parseIdentifier(Name))
3809         return Error(Loc, "expected identifier in directive");
3810 
3811       MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
3812 
3813       // Assembler local symbols don't make any sense here. Complain loudly.
3814       if (Sym->isTemporary())
3815         return Error(Loc, "non-local symbol required in directive");
3816 
3817       if (!getStreamer().EmitSymbolAttribute(Sym, Attr))
3818         return Error(Loc, "unable to emit symbol attribute");
3819 
3820       if (getLexer().is(AsmToken::EndOfStatement))
3821         break;
3822 
3823       if (getLexer().isNot(AsmToken::Comma))
3824         return TokError("unexpected token in directive");
3825       Lex();
3826     }
3827   }
3828 
3829   Lex();
3830   return false;
3831 }
3832 
3833 /// parseDirectiveComm
3834 ///  ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
parseDirectiveComm(bool IsLocal)3835 bool AsmParser::parseDirectiveComm(bool IsLocal) {
3836   checkForValidSection();
3837 
3838   SMLoc IDLoc = getLexer().getLoc();
3839   StringRef Name;
3840   if (parseIdentifier(Name))
3841     return TokError("expected identifier in directive");
3842 
3843   // Handle the identifier as the key symbol.
3844   MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
3845 
3846   if (getLexer().isNot(AsmToken::Comma))
3847     return TokError("unexpected token in directive");
3848   Lex();
3849 
3850   int64_t Size;
3851   SMLoc SizeLoc = getLexer().getLoc();
3852   if (parseAbsoluteExpression(Size))
3853     return true;
3854 
3855   int64_t Pow2Alignment = 0;
3856   SMLoc Pow2AlignmentLoc;
3857   if (getLexer().is(AsmToken::Comma)) {
3858     Lex();
3859     Pow2AlignmentLoc = getLexer().getLoc();
3860     if (parseAbsoluteExpression(Pow2Alignment))
3861       return true;
3862 
3863     LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType();
3864     if (IsLocal && LCOMM == LCOMM::NoAlignment)
3865       return Error(Pow2AlignmentLoc, "alignment not supported on this target");
3866 
3867     // If this target takes alignments in bytes (not log) validate and convert.
3868     if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) ||
3869         (IsLocal && LCOMM == LCOMM::ByteAlignment)) {
3870       if (!isPowerOf2_64(Pow2Alignment))
3871         return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
3872       Pow2Alignment = Log2_64(Pow2Alignment);
3873     }
3874   }
3875 
3876   if (getLexer().isNot(AsmToken::EndOfStatement))
3877     return TokError("unexpected token in '.comm' or '.lcomm' directive");
3878 
3879   Lex();
3880 
3881   // NOTE: a size of zero for a .comm should create a undefined symbol
3882   // but a size of .lcomm creates a bss symbol of size zero.
3883   if (Size < 0)
3884     return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
3885                           "be less than zero");
3886 
3887   // NOTE: The alignment in the directive is a power of 2 value, the assembler
3888   // may internally end up wanting an alignment in bytes.
3889   // FIXME: Diagnose overflow.
3890   if (Pow2Alignment < 0)
3891     return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
3892                                    "alignment, can't be less than zero");
3893 
3894   if (!Sym->isUndefined())
3895     return Error(IDLoc, "invalid symbol redefinition");
3896 
3897   // Create the Symbol as a common or local common with Size and Pow2Alignment
3898   if (IsLocal) {
3899     getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment);
3900     return false;
3901   }
3902 
3903   getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
3904   return false;
3905 }
3906 
3907 /// parseDirectiveAbort
3908 ///  ::= .abort [... message ...]
parseDirectiveAbort()3909 bool AsmParser::parseDirectiveAbort() {
3910   // FIXME: Use loc from directive.
3911   SMLoc Loc = getLexer().getLoc();
3912 
3913   StringRef Str = parseStringToEndOfStatement();
3914   if (getLexer().isNot(AsmToken::EndOfStatement))
3915     return TokError("unexpected token in '.abort' directive");
3916 
3917   Lex();
3918 
3919   if (Str.empty())
3920     Error(Loc, ".abort detected. Assembly stopping.");
3921   else
3922     Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
3923   // FIXME: Actually abort assembly here.
3924 
3925   return false;
3926 }
3927 
3928 /// parseDirectiveInclude
3929 ///  ::= .include "filename"
parseDirectiveInclude()3930 bool AsmParser::parseDirectiveInclude() {
3931   if (getLexer().isNot(AsmToken::String))
3932     return TokError("expected string in '.include' directive");
3933 
3934   // Allow the strings to have escaped octal character sequence.
3935   std::string Filename;
3936   if (parseEscapedString(Filename))
3937     return true;
3938   SMLoc IncludeLoc = getLexer().getLoc();
3939   Lex();
3940 
3941   if (getLexer().isNot(AsmToken::EndOfStatement))
3942     return TokError("unexpected token in '.include' directive");
3943 
3944   // Attempt to switch the lexer to the included file before consuming the end
3945   // of statement to avoid losing it when we switch.
3946   if (enterIncludeFile(Filename)) {
3947     Error(IncludeLoc, "Could not find include file '" + Filename + "'");
3948     return true;
3949   }
3950 
3951   return false;
3952 }
3953 
3954 /// parseDirectiveIncbin
3955 ///  ::= .incbin "filename"
parseDirectiveIncbin()3956 bool AsmParser::parseDirectiveIncbin() {
3957   if (getLexer().isNot(AsmToken::String))
3958     return TokError("expected string in '.incbin' directive");
3959 
3960   // Allow the strings to have escaped octal character sequence.
3961   std::string Filename;
3962   if (parseEscapedString(Filename))
3963     return true;
3964   SMLoc IncbinLoc = getLexer().getLoc();
3965   Lex();
3966 
3967   if (getLexer().isNot(AsmToken::EndOfStatement))
3968     return TokError("unexpected token in '.incbin' directive");
3969 
3970   // Attempt to process the included file.
3971   if (processIncbinFile(Filename)) {
3972     Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
3973     return true;
3974   }
3975 
3976   return false;
3977 }
3978 
3979 /// parseDirectiveIf
3980 /// ::= .if{,eq,ge,gt,le,lt,ne} expression
parseDirectiveIf(SMLoc DirectiveLoc,DirectiveKind DirKind)3981 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) {
3982   TheCondStack.push_back(TheCondState);
3983   TheCondState.TheCond = AsmCond::IfCond;
3984   if (TheCondState.Ignore) {
3985     eatToEndOfStatement();
3986   } else {
3987     int64_t ExprValue;
3988     if (parseAbsoluteExpression(ExprValue))
3989       return true;
3990 
3991     if (getLexer().isNot(AsmToken::EndOfStatement))
3992       return TokError("unexpected token in '.if' directive");
3993 
3994     Lex();
3995 
3996     switch (DirKind) {
3997     default:
3998       llvm_unreachable("unsupported directive");
3999     case DK_IF:
4000     case DK_IFNE:
4001       break;
4002     case DK_IFEQ:
4003       ExprValue = ExprValue == 0;
4004       break;
4005     case DK_IFGE:
4006       ExprValue = ExprValue >= 0;
4007       break;
4008     case DK_IFGT:
4009       ExprValue = ExprValue > 0;
4010       break;
4011     case DK_IFLE:
4012       ExprValue = ExprValue <= 0;
4013       break;
4014     case DK_IFLT:
4015       ExprValue = ExprValue < 0;
4016       break;
4017     }
4018 
4019     TheCondState.CondMet = ExprValue;
4020     TheCondState.Ignore = !TheCondState.CondMet;
4021   }
4022 
4023   return false;
4024 }
4025 
4026 /// parseDirectiveIfb
4027 /// ::= .ifb string
parseDirectiveIfb(SMLoc DirectiveLoc,bool ExpectBlank)4028 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
4029   TheCondStack.push_back(TheCondState);
4030   TheCondState.TheCond = AsmCond::IfCond;
4031 
4032   if (TheCondState.Ignore) {
4033     eatToEndOfStatement();
4034   } else {
4035     StringRef Str = parseStringToEndOfStatement();
4036 
4037     if (getLexer().isNot(AsmToken::EndOfStatement))
4038       return TokError("unexpected token in '.ifb' directive");
4039 
4040     Lex();
4041 
4042     TheCondState.CondMet = ExpectBlank == Str.empty();
4043     TheCondState.Ignore = !TheCondState.CondMet;
4044   }
4045 
4046   return false;
4047 }
4048 
4049 /// parseDirectiveIfc
4050 /// ::= .ifc string1, string2
4051 /// ::= .ifnc string1, string2
parseDirectiveIfc(SMLoc DirectiveLoc,bool ExpectEqual)4052 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
4053   TheCondStack.push_back(TheCondState);
4054   TheCondState.TheCond = AsmCond::IfCond;
4055 
4056   if (TheCondState.Ignore) {
4057     eatToEndOfStatement();
4058   } else {
4059     StringRef Str1 = parseStringToComma();
4060 
4061     if (getLexer().isNot(AsmToken::Comma))
4062       return TokError("unexpected token in '.ifc' directive");
4063 
4064     Lex();
4065 
4066     StringRef Str2 = parseStringToEndOfStatement();
4067 
4068     if (getLexer().isNot(AsmToken::EndOfStatement))
4069       return TokError("unexpected token in '.ifc' directive");
4070 
4071     Lex();
4072 
4073     TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim());
4074     TheCondState.Ignore = !TheCondState.CondMet;
4075   }
4076 
4077   return false;
4078 }
4079 
4080 /// parseDirectiveIfeqs
4081 ///   ::= .ifeqs string1, string2
parseDirectiveIfeqs(SMLoc DirectiveLoc,bool ExpectEqual)4082 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) {
4083   if (Lexer.isNot(AsmToken::String)) {
4084     if (ExpectEqual)
4085       TokError("expected string parameter for '.ifeqs' directive");
4086     else
4087       TokError("expected string parameter for '.ifnes' directive");
4088     eatToEndOfStatement();
4089     return true;
4090   }
4091 
4092   StringRef String1 = getTok().getStringContents();
4093   Lex();
4094 
4095   if (Lexer.isNot(AsmToken::Comma)) {
4096     if (ExpectEqual)
4097       TokError("expected comma after first string for '.ifeqs' directive");
4098     else
4099       TokError("expected comma after first string for '.ifnes' directive");
4100     eatToEndOfStatement();
4101     return true;
4102   }
4103 
4104   Lex();
4105 
4106   if (Lexer.isNot(AsmToken::String)) {
4107     if (ExpectEqual)
4108       TokError("expected string parameter for '.ifeqs' directive");
4109     else
4110       TokError("expected string parameter for '.ifnes' directive");
4111     eatToEndOfStatement();
4112     return true;
4113   }
4114 
4115   StringRef String2 = getTok().getStringContents();
4116   Lex();
4117 
4118   TheCondStack.push_back(TheCondState);
4119   TheCondState.TheCond = AsmCond::IfCond;
4120   TheCondState.CondMet = ExpectEqual == (String1 == String2);
4121   TheCondState.Ignore = !TheCondState.CondMet;
4122 
4123   return false;
4124 }
4125 
4126 /// parseDirectiveIfdef
4127 /// ::= .ifdef symbol
parseDirectiveIfdef(SMLoc DirectiveLoc,bool expect_defined)4128 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
4129   StringRef Name;
4130   TheCondStack.push_back(TheCondState);
4131   TheCondState.TheCond = AsmCond::IfCond;
4132 
4133   if (TheCondState.Ignore) {
4134     eatToEndOfStatement();
4135   } else {
4136     if (parseIdentifier(Name))
4137       return TokError("expected identifier after '.ifdef'");
4138 
4139     Lex();
4140 
4141     MCSymbol *Sym = getContext().lookupSymbol(Name);
4142 
4143     if (expect_defined)
4144       TheCondState.CondMet = (Sym && !Sym->isUndefined());
4145     else
4146       TheCondState.CondMet = (!Sym || Sym->isUndefined());
4147     TheCondState.Ignore = !TheCondState.CondMet;
4148   }
4149 
4150   return false;
4151 }
4152 
4153 /// parseDirectiveElseIf
4154 /// ::= .elseif expression
parseDirectiveElseIf(SMLoc DirectiveLoc)4155 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) {
4156   if (TheCondState.TheCond != AsmCond::IfCond &&
4157       TheCondState.TheCond != AsmCond::ElseIfCond)
4158     Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
4159                         " an .elseif");
4160   TheCondState.TheCond = AsmCond::ElseIfCond;
4161 
4162   bool LastIgnoreState = false;
4163   if (!TheCondStack.empty())
4164     LastIgnoreState = TheCondStack.back().Ignore;
4165   if (LastIgnoreState || TheCondState.CondMet) {
4166     TheCondState.Ignore = true;
4167     eatToEndOfStatement();
4168   } else {
4169     int64_t ExprValue;
4170     if (parseAbsoluteExpression(ExprValue))
4171       return true;
4172 
4173     if (getLexer().isNot(AsmToken::EndOfStatement))
4174       return TokError("unexpected token in '.elseif' directive");
4175 
4176     Lex();
4177     TheCondState.CondMet = ExprValue;
4178     TheCondState.Ignore = !TheCondState.CondMet;
4179   }
4180 
4181   return false;
4182 }
4183 
4184 /// parseDirectiveElse
4185 /// ::= .else
parseDirectiveElse(SMLoc DirectiveLoc)4186 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) {
4187   if (getLexer().isNot(AsmToken::EndOfStatement))
4188     return TokError("unexpected token in '.else' directive");
4189 
4190   Lex();
4191 
4192   if (TheCondState.TheCond != AsmCond::IfCond &&
4193       TheCondState.TheCond != AsmCond::ElseIfCond)
4194     Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
4195                         ".elseif");
4196   TheCondState.TheCond = AsmCond::ElseCond;
4197   bool LastIgnoreState = false;
4198   if (!TheCondStack.empty())
4199     LastIgnoreState = TheCondStack.back().Ignore;
4200   if (LastIgnoreState || TheCondState.CondMet)
4201     TheCondState.Ignore = true;
4202   else
4203     TheCondState.Ignore = false;
4204 
4205   return false;
4206 }
4207 
4208 /// parseDirectiveEnd
4209 /// ::= .end
parseDirectiveEnd(SMLoc DirectiveLoc)4210 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) {
4211   if (getLexer().isNot(AsmToken::EndOfStatement))
4212     return TokError("unexpected token in '.end' directive");
4213 
4214   Lex();
4215 
4216   while (Lexer.isNot(AsmToken::Eof))
4217     Lex();
4218 
4219   return false;
4220 }
4221 
4222 /// parseDirectiveError
4223 ///   ::= .err
4224 ///   ::= .error [string]
parseDirectiveError(SMLoc L,bool WithMessage)4225 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) {
4226   if (!TheCondStack.empty()) {
4227     if (TheCondStack.back().Ignore) {
4228       eatToEndOfStatement();
4229       return false;
4230     }
4231   }
4232 
4233   if (!WithMessage)
4234     return Error(L, ".err encountered");
4235 
4236   StringRef Message = ".error directive invoked in source file";
4237   if (Lexer.isNot(AsmToken::EndOfStatement)) {
4238     if (Lexer.isNot(AsmToken::String)) {
4239       TokError(".error argument must be a string");
4240       eatToEndOfStatement();
4241       return true;
4242     }
4243 
4244     Message = getTok().getStringContents();
4245     Lex();
4246   }
4247 
4248   Error(L, Message);
4249   return true;
4250 }
4251 
4252 /// parseDirectiveWarning
4253 ///   ::= .warning [string]
parseDirectiveWarning(SMLoc L)4254 bool AsmParser::parseDirectiveWarning(SMLoc L) {
4255   if (!TheCondStack.empty()) {
4256     if (TheCondStack.back().Ignore) {
4257       eatToEndOfStatement();
4258       return false;
4259     }
4260   }
4261 
4262   StringRef Message = ".warning directive invoked in source file";
4263   if (Lexer.isNot(AsmToken::EndOfStatement)) {
4264     if (Lexer.isNot(AsmToken::String)) {
4265       TokError(".warning argument must be a string");
4266       eatToEndOfStatement();
4267       return true;
4268     }
4269 
4270     Message = getTok().getStringContents();
4271     Lex();
4272   }
4273 
4274   Warning(L, Message);
4275   return false;
4276 }
4277 
4278 /// parseDirectiveEndIf
4279 /// ::= .endif
parseDirectiveEndIf(SMLoc DirectiveLoc)4280 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) {
4281   if (getLexer().isNot(AsmToken::EndOfStatement))
4282     return TokError("unexpected token in '.endif' directive");
4283 
4284   Lex();
4285 
4286   if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty())
4287     Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
4288                         ".else");
4289   if (!TheCondStack.empty()) {
4290     TheCondState = TheCondStack.back();
4291     TheCondStack.pop_back();
4292   }
4293 
4294   return false;
4295 }
4296 
initializeDirectiveKindMap()4297 void AsmParser::initializeDirectiveKindMap() {
4298   DirectiveKindMap[".set"] = DK_SET;
4299   DirectiveKindMap[".equ"] = DK_EQU;
4300   DirectiveKindMap[".equiv"] = DK_EQUIV;
4301   DirectiveKindMap[".ascii"] = DK_ASCII;
4302   DirectiveKindMap[".asciz"] = DK_ASCIZ;
4303   DirectiveKindMap[".string"] = DK_STRING;
4304   DirectiveKindMap[".byte"] = DK_BYTE;
4305   DirectiveKindMap[".short"] = DK_SHORT;
4306   DirectiveKindMap[".value"] = DK_VALUE;
4307   DirectiveKindMap[".2byte"] = DK_2BYTE;
4308   DirectiveKindMap[".long"] = DK_LONG;
4309   DirectiveKindMap[".int"] = DK_INT;
4310   DirectiveKindMap[".4byte"] = DK_4BYTE;
4311   DirectiveKindMap[".quad"] = DK_QUAD;
4312   DirectiveKindMap[".8byte"] = DK_8BYTE;
4313   DirectiveKindMap[".octa"] = DK_OCTA;
4314   DirectiveKindMap[".single"] = DK_SINGLE;
4315   DirectiveKindMap[".float"] = DK_FLOAT;
4316   DirectiveKindMap[".double"] = DK_DOUBLE;
4317   DirectiveKindMap[".align"] = DK_ALIGN;
4318   DirectiveKindMap[".align32"] = DK_ALIGN32;
4319   DirectiveKindMap[".balign"] = DK_BALIGN;
4320   DirectiveKindMap[".balignw"] = DK_BALIGNW;
4321   DirectiveKindMap[".balignl"] = DK_BALIGNL;
4322   DirectiveKindMap[".p2align"] = DK_P2ALIGN;
4323   DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW;
4324   DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL;
4325   DirectiveKindMap[".org"] = DK_ORG;
4326   DirectiveKindMap[".fill"] = DK_FILL;
4327   DirectiveKindMap[".zero"] = DK_ZERO;
4328   DirectiveKindMap[".extern"] = DK_EXTERN;
4329   DirectiveKindMap[".globl"] = DK_GLOBL;
4330   DirectiveKindMap[".global"] = DK_GLOBAL;
4331   DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE;
4332   DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP;
4333   DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER;
4334   DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN;
4335   DirectiveKindMap[".reference"] = DK_REFERENCE;
4336   DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION;
4337   DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE;
4338   DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN;
4339   DirectiveKindMap[".comm"] = DK_COMM;
4340   DirectiveKindMap[".common"] = DK_COMMON;
4341   DirectiveKindMap[".lcomm"] = DK_LCOMM;
4342   DirectiveKindMap[".abort"] = DK_ABORT;
4343   DirectiveKindMap[".include"] = DK_INCLUDE;
4344   DirectiveKindMap[".incbin"] = DK_INCBIN;
4345   DirectiveKindMap[".code16"] = DK_CODE16;
4346   DirectiveKindMap[".code16gcc"] = DK_CODE16GCC;
4347   DirectiveKindMap[".rept"] = DK_REPT;
4348   DirectiveKindMap[".rep"] = DK_REPT;
4349   DirectiveKindMap[".irp"] = DK_IRP;
4350   DirectiveKindMap[".irpc"] = DK_IRPC;
4351   DirectiveKindMap[".endr"] = DK_ENDR;
4352   DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE;
4353   DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK;
4354   DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK;
4355   DirectiveKindMap[".if"] = DK_IF;
4356   DirectiveKindMap[".ifeq"] = DK_IFEQ;
4357   DirectiveKindMap[".ifge"] = DK_IFGE;
4358   DirectiveKindMap[".ifgt"] = DK_IFGT;
4359   DirectiveKindMap[".ifle"] = DK_IFLE;
4360   DirectiveKindMap[".iflt"] = DK_IFLT;
4361   DirectiveKindMap[".ifne"] = DK_IFNE;
4362   DirectiveKindMap[".ifb"] = DK_IFB;
4363   DirectiveKindMap[".ifnb"] = DK_IFNB;
4364   DirectiveKindMap[".ifc"] = DK_IFC;
4365   DirectiveKindMap[".ifeqs"] = DK_IFEQS;
4366   DirectiveKindMap[".ifnc"] = DK_IFNC;
4367   DirectiveKindMap[".ifnes"] = DK_IFNES;
4368   DirectiveKindMap[".ifdef"] = DK_IFDEF;
4369   DirectiveKindMap[".ifndef"] = DK_IFNDEF;
4370   DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF;
4371   DirectiveKindMap[".elseif"] = DK_ELSEIF;
4372   DirectiveKindMap[".else"] = DK_ELSE;
4373   DirectiveKindMap[".end"] = DK_END;
4374   DirectiveKindMap[".endif"] = DK_ENDIF;
4375   DirectiveKindMap[".skip"] = DK_SKIP;
4376   DirectiveKindMap[".space"] = DK_SPACE;
4377   DirectiveKindMap[".file"] = DK_FILE;
4378   DirectiveKindMap[".line"] = DK_LINE;
4379   DirectiveKindMap[".loc"] = DK_LOC;
4380   DirectiveKindMap[".stabs"] = DK_STABS;
4381   DirectiveKindMap[".sleb128"] = DK_SLEB128;
4382   DirectiveKindMap[".uleb128"] = DK_ULEB128;
4383   DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS;
4384   DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC;
4385   DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC;
4386   DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA;
4387   DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET;
4388   DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET;
4389   DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER;
4390   DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET;
4391   DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET;
4392   DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY;
4393   DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA;
4394   DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE;
4395   DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE;
4396   DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE;
4397   DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE;
4398   DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE;
4399   DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME;
4400   DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED;
4401   DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER;
4402   DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE;
4403   DirectiveKindMap[".macros_on"] = DK_MACROS_ON;
4404   DirectiveKindMap[".macros_off"] = DK_MACROS_OFF;
4405   DirectiveKindMap[".macro"] = DK_MACRO;
4406   DirectiveKindMap[".exitm"] = DK_EXITM;
4407   DirectiveKindMap[".endm"] = DK_ENDM;
4408   DirectiveKindMap[".endmacro"] = DK_ENDMACRO;
4409   DirectiveKindMap[".purgem"] = DK_PURGEM;
4410   DirectiveKindMap[".err"] = DK_ERR;
4411   DirectiveKindMap[".error"] = DK_ERROR;
4412   DirectiveKindMap[".warning"] = DK_WARNING;
4413   DirectiveKindMap[".reloc"] = DK_RELOC;
4414 }
4415 
parseMacroLikeBody(SMLoc DirectiveLoc)4416 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) {
4417   AsmToken EndToken, StartToken = getTok();
4418 
4419   unsigned NestLevel = 0;
4420   for (;;) {
4421     // Check whether we have reached the end of the file.
4422     if (getLexer().is(AsmToken::Eof)) {
4423       Error(DirectiveLoc, "no matching '.endr' in definition");
4424       return nullptr;
4425     }
4426 
4427     if (Lexer.is(AsmToken::Identifier) &&
4428         (getTok().getIdentifier() == ".rept")) {
4429       ++NestLevel;
4430     }
4431 
4432     // Otherwise, check whether we have reached the .endr.
4433     if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") {
4434       if (NestLevel == 0) {
4435         EndToken = getTok();
4436         Lex();
4437         if (Lexer.isNot(AsmToken::EndOfStatement)) {
4438           TokError("unexpected token in '.endr' directive");
4439           return nullptr;
4440         }
4441         break;
4442       }
4443       --NestLevel;
4444     }
4445 
4446     // Otherwise, scan till the end of the statement.
4447     eatToEndOfStatement();
4448   }
4449 
4450   const char *BodyStart = StartToken.getLoc().getPointer();
4451   const char *BodyEnd = EndToken.getLoc().getPointer();
4452   StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
4453 
4454   // We Are Anonymous.
4455   MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters());
4456   return &MacroLikeBodies.back();
4457 }
4458 
instantiateMacroLikeBody(MCAsmMacro * M,SMLoc DirectiveLoc,raw_svector_ostream & OS)4459 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
4460                                          raw_svector_ostream &OS) {
4461   OS << ".endr\n";
4462 
4463   std::unique_ptr<MemoryBuffer> Instantiation =
4464       MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
4465 
4466   // Create the macro instantiation object and add to the current macro
4467   // instantiation stack.
4468   MacroInstantiation *MI = new MacroInstantiation(
4469       DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size());
4470   ActiveMacros.push_back(MI);
4471 
4472   // Jump to the macro instantiation and prime the lexer.
4473   CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
4474   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
4475   Lex();
4476 }
4477 
4478 /// parseDirectiveRept
4479 ///   ::= .rep | .rept count
parseDirectiveRept(SMLoc DirectiveLoc,StringRef Dir)4480 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) {
4481   const MCExpr *CountExpr;
4482   SMLoc CountLoc = getTok().getLoc();
4483   if (parseExpression(CountExpr))
4484     return true;
4485 
4486   int64_t Count;
4487   if (!CountExpr->evaluateAsAbsolute(Count)) {
4488     eatToEndOfStatement();
4489     return Error(CountLoc, "unexpected token in '" + Dir + "' directive");
4490   }
4491 
4492   if (Count < 0)
4493     return Error(CountLoc, "Count is negative");
4494 
4495   if (Lexer.isNot(AsmToken::EndOfStatement))
4496     return TokError("unexpected token in '" + Dir + "' directive");
4497 
4498   // Eat the end of statement.
4499   Lex();
4500 
4501   // Lex the rept definition.
4502   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4503   if (!M)
4504     return true;
4505 
4506   // Macro instantiation is lexical, unfortunately. We construct a new buffer
4507   // to hold the macro body with substitutions.
4508   SmallString<256> Buf;
4509   raw_svector_ostream OS(Buf);
4510   while (Count--) {
4511     // Note that the AtPseudoVariable is disabled for instantiations of .rep(t).
4512     if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc()))
4513       return true;
4514   }
4515   instantiateMacroLikeBody(M, DirectiveLoc, OS);
4516 
4517   return false;
4518 }
4519 
4520 /// parseDirectiveIrp
4521 /// ::= .irp symbol,values
parseDirectiveIrp(SMLoc DirectiveLoc)4522 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) {
4523   MCAsmMacroParameter Parameter;
4524 
4525   if (parseIdentifier(Parameter.Name))
4526     return TokError("expected identifier in '.irp' directive");
4527 
4528   if (Lexer.isNot(AsmToken::Comma))
4529     return TokError("expected comma in '.irp' directive");
4530 
4531   Lex();
4532 
4533   MCAsmMacroArguments A;
4534   if (parseMacroArguments(nullptr, A))
4535     return true;
4536 
4537   // Eat the end of statement.
4538   Lex();
4539 
4540   // Lex the irp definition.
4541   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4542   if (!M)
4543     return true;
4544 
4545   // Macro instantiation is lexical, unfortunately. We construct a new buffer
4546   // to hold the macro body with substitutions.
4547   SmallString<256> Buf;
4548   raw_svector_ostream OS(Buf);
4549 
4550   for (const MCAsmMacroArgument &Arg : A) {
4551     // Note that the AtPseudoVariable is enabled for instantiations of .irp.
4552     // This is undocumented, but GAS seems to support it.
4553     if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc()))
4554       return true;
4555   }
4556 
4557   instantiateMacroLikeBody(M, DirectiveLoc, OS);
4558 
4559   return false;
4560 }
4561 
4562 /// parseDirectiveIrpc
4563 /// ::= .irpc symbol,values
parseDirectiveIrpc(SMLoc DirectiveLoc)4564 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) {
4565   MCAsmMacroParameter Parameter;
4566 
4567   if (parseIdentifier(Parameter.Name))
4568     return TokError("expected identifier in '.irpc' directive");
4569 
4570   if (Lexer.isNot(AsmToken::Comma))
4571     return TokError("expected comma in '.irpc' directive");
4572 
4573   Lex();
4574 
4575   MCAsmMacroArguments A;
4576   if (parseMacroArguments(nullptr, A))
4577     return true;
4578 
4579   if (A.size() != 1 || A.front().size() != 1)
4580     return TokError("unexpected token in '.irpc' directive");
4581 
4582   // Eat the end of statement.
4583   Lex();
4584 
4585   // Lex the irpc definition.
4586   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4587   if (!M)
4588     return true;
4589 
4590   // Macro instantiation is lexical, unfortunately. We construct a new buffer
4591   // to hold the macro body with substitutions.
4592   SmallString<256> Buf;
4593   raw_svector_ostream OS(Buf);
4594 
4595   StringRef Values = A.front().front().getString();
4596   for (std::size_t I = 0, End = Values.size(); I != End; ++I) {
4597     MCAsmMacroArgument Arg;
4598     Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1));
4599 
4600     // Note that the AtPseudoVariable is enabled for instantiations of .irpc.
4601     // This is undocumented, but GAS seems to support it.
4602     if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc()))
4603       return true;
4604   }
4605 
4606   instantiateMacroLikeBody(M, DirectiveLoc, OS);
4607 
4608   return false;
4609 }
4610 
parseDirectiveEndr(SMLoc DirectiveLoc)4611 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) {
4612   if (ActiveMacros.empty())
4613     return TokError("unmatched '.endr' directive");
4614 
4615   // The only .repl that should get here are the ones created by
4616   // instantiateMacroLikeBody.
4617   assert(getLexer().is(AsmToken::EndOfStatement));
4618 
4619   handleMacroExit();
4620   return false;
4621 }
4622 
parseDirectiveMSEmit(SMLoc IDLoc,ParseStatementInfo & Info,size_t Len)4623 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info,
4624                                      size_t Len) {
4625   const MCExpr *Value;
4626   SMLoc ExprLoc = getLexer().getLoc();
4627   if (parseExpression(Value))
4628     return true;
4629   const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
4630   if (!MCE)
4631     return Error(ExprLoc, "unexpected expression in _emit");
4632   uint64_t IntValue = MCE->getValue();
4633   if (!isUInt<8>(IntValue) && !isInt<8>(IntValue))
4634     return Error(ExprLoc, "literal value out of range for directive");
4635 
4636   Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len);
4637   return false;
4638 }
4639 
parseDirectiveMSAlign(SMLoc IDLoc,ParseStatementInfo & Info)4640 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) {
4641   const MCExpr *Value;
4642   SMLoc ExprLoc = getLexer().getLoc();
4643   if (parseExpression(Value))
4644     return true;
4645   const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
4646   if (!MCE)
4647     return Error(ExprLoc, "unexpected expression in align");
4648   uint64_t IntValue = MCE->getValue();
4649   if (!isPowerOf2_64(IntValue))
4650     return Error(ExprLoc, "literal value not a power of two greater then zero");
4651 
4652   Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue));
4653   return false;
4654 }
4655 
4656 // We are comparing pointers, but the pointers are relative to a single string.
4657 // Thus, this should always be deterministic.
rewritesSort(const AsmRewrite * AsmRewriteA,const AsmRewrite * AsmRewriteB)4658 static int rewritesSort(const AsmRewrite *AsmRewriteA,
4659                         const AsmRewrite *AsmRewriteB) {
4660   if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer())
4661     return -1;
4662   if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer())
4663     return 1;
4664 
4665   // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output
4666   // rewrite to the same location.  Make sure the SizeDirective rewrite is
4667   // performed first, then the Imm/ImmPrefix and finally the Input/Output.  This
4668   // ensures the sort algorithm is stable.
4669   if (AsmRewritePrecedence[AsmRewriteA->Kind] >
4670       AsmRewritePrecedence[AsmRewriteB->Kind])
4671     return -1;
4672 
4673   if (AsmRewritePrecedence[AsmRewriteA->Kind] <
4674       AsmRewritePrecedence[AsmRewriteB->Kind])
4675     return 1;
4676   llvm_unreachable("Unstable rewrite sort.");
4677 }
4678 
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)4679 bool AsmParser::parseMSInlineAsm(
4680     void *AsmLoc, std::string &AsmString, unsigned &NumOutputs,
4681     unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool> > &OpDecls,
4682     SmallVectorImpl<std::string> &Constraints,
4683     SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII,
4684     const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) {
4685   SmallVector<void *, 4> InputDecls;
4686   SmallVector<void *, 4> OutputDecls;
4687   SmallVector<bool, 4> InputDeclsAddressOf;
4688   SmallVector<bool, 4> OutputDeclsAddressOf;
4689   SmallVector<std::string, 4> InputConstraints;
4690   SmallVector<std::string, 4> OutputConstraints;
4691   SmallVector<unsigned, 4> ClobberRegs;
4692 
4693   SmallVector<AsmRewrite, 4> AsmStrRewrites;
4694 
4695   // Prime the lexer.
4696   Lex();
4697 
4698   // While we have input, parse each statement.
4699   unsigned InputIdx = 0;
4700   unsigned OutputIdx = 0;
4701   while (getLexer().isNot(AsmToken::Eof)) {
4702     ParseStatementInfo Info(&AsmStrRewrites);
4703     if (parseStatement(Info, &SI))
4704       return true;
4705 
4706     if (Info.ParseError)
4707       return true;
4708 
4709     if (Info.Opcode == ~0U)
4710       continue;
4711 
4712     const MCInstrDesc &Desc = MII->get(Info.Opcode);
4713 
4714     // Build the list of clobbers, outputs and inputs.
4715     for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
4716       MCParsedAsmOperand &Operand = *Info.ParsedOperands[i];
4717 
4718       // Immediate.
4719       if (Operand.isImm())
4720         continue;
4721 
4722       // Register operand.
4723       if (Operand.isReg() && !Operand.needAddressOf() &&
4724           !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) {
4725         unsigned NumDefs = Desc.getNumDefs();
4726         // Clobber.
4727         if (NumDefs && Operand.getMCOperandNum() < NumDefs)
4728           ClobberRegs.push_back(Operand.getReg());
4729         continue;
4730       }
4731 
4732       // Expr/Input or Output.
4733       StringRef SymName = Operand.getSymName();
4734       if (SymName.empty())
4735         continue;
4736 
4737       void *OpDecl = Operand.getOpDecl();
4738       if (!OpDecl)
4739         continue;
4740 
4741       bool isOutput = (i == 1) && Desc.mayStore();
4742       SMLoc Start = SMLoc::getFromPointer(SymName.data());
4743       if (isOutput) {
4744         ++InputIdx;
4745         OutputDecls.push_back(OpDecl);
4746         OutputDeclsAddressOf.push_back(Operand.needAddressOf());
4747         OutputConstraints.push_back(("=" + Operand.getConstraint()).str());
4748         AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size());
4749       } else {
4750         InputDecls.push_back(OpDecl);
4751         InputDeclsAddressOf.push_back(Operand.needAddressOf());
4752         InputConstraints.push_back(Operand.getConstraint().str());
4753         AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size());
4754       }
4755     }
4756 
4757     // Consider implicit defs to be clobbers.  Think of cpuid and push.
4758     ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(),
4759                                 Desc.getNumImplicitDefs());
4760     ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end());
4761   }
4762 
4763   // Set the number of Outputs and Inputs.
4764   NumOutputs = OutputDecls.size();
4765   NumInputs = InputDecls.size();
4766 
4767   // Set the unique clobbers.
4768   array_pod_sort(ClobberRegs.begin(), ClobberRegs.end());
4769   ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()),
4770                     ClobberRegs.end());
4771   Clobbers.assign(ClobberRegs.size(), std::string());
4772   for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) {
4773     raw_string_ostream OS(Clobbers[I]);
4774     IP->printRegName(OS, ClobberRegs[I]);
4775   }
4776 
4777   // Merge the various outputs and inputs.  Output are expected first.
4778   if (NumOutputs || NumInputs) {
4779     unsigned NumExprs = NumOutputs + NumInputs;
4780     OpDecls.resize(NumExprs);
4781     Constraints.resize(NumExprs);
4782     for (unsigned i = 0; i < NumOutputs; ++i) {
4783       OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]);
4784       Constraints[i] = OutputConstraints[i];
4785     }
4786     for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
4787       OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]);
4788       Constraints[j] = InputConstraints[i];
4789     }
4790   }
4791 
4792   // Build the IR assembly string.
4793   std::string AsmStringIR;
4794   raw_string_ostream OS(AsmStringIR);
4795   StringRef ASMString =
4796       SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer();
4797   const char *AsmStart = ASMString.begin();
4798   const char *AsmEnd = ASMString.end();
4799   array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort);
4800   for (const AsmRewrite &AR : AsmStrRewrites) {
4801     AsmRewriteKind Kind = AR.Kind;
4802     if (Kind == AOK_Delete)
4803       continue;
4804 
4805     const char *Loc = AR.Loc.getPointer();
4806     assert(Loc >= AsmStart && "Expected Loc to be at or after Start!");
4807 
4808     // Emit everything up to the immediate/expression.
4809     if (unsigned Len = Loc - AsmStart)
4810       OS << StringRef(AsmStart, Len);
4811 
4812     // Skip the original expression.
4813     if (Kind == AOK_Skip) {
4814       AsmStart = Loc + AR.Len;
4815       continue;
4816     }
4817 
4818     unsigned AdditionalSkip = 0;
4819     // Rewrite expressions in $N notation.
4820     switch (Kind) {
4821     default:
4822       break;
4823     case AOK_Imm:
4824       OS << "$$" << AR.Val;
4825       break;
4826     case AOK_ImmPrefix:
4827       OS << "$$";
4828       break;
4829     case AOK_Label:
4830       OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label;
4831       break;
4832     case AOK_Input:
4833       OS << '$' << InputIdx++;
4834       break;
4835     case AOK_Output:
4836       OS << '$' << OutputIdx++;
4837       break;
4838     case AOK_SizeDirective:
4839       switch (AR.Val) {
4840       default: break;
4841       case 8:  OS << "byte ptr "; break;
4842       case 16: OS << "word ptr "; break;
4843       case 32: OS << "dword ptr "; break;
4844       case 64: OS << "qword ptr "; break;
4845       case 80: OS << "xword ptr "; break;
4846       case 128: OS << "xmmword ptr "; break;
4847       case 256: OS << "ymmword ptr "; break;
4848       }
4849       break;
4850     case AOK_Emit:
4851       OS << ".byte";
4852       break;
4853     case AOK_Align: {
4854       // MS alignment directives are measured in bytes. If the native assembler
4855       // measures alignment in bytes, we can pass it straight through.
4856       OS << ".align";
4857       if (getContext().getAsmInfo()->getAlignmentIsInBytes())
4858         break;
4859 
4860       // Alignment is in log2 form, so print that instead and skip the original
4861       // immediate.
4862       unsigned Val = AR.Val;
4863       OS << ' ' << Val;
4864       assert(Val < 10 && "Expected alignment less then 2^10.");
4865       AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4;
4866       break;
4867     }
4868     case AOK_EVEN:
4869       OS << ".even";
4870       break;
4871     case AOK_DotOperator:
4872       // Insert the dot if the user omitted it.
4873       OS.flush();
4874       if (AsmStringIR.back() != '.')
4875         OS << '.';
4876       OS << AR.Val;
4877       break;
4878     }
4879 
4880     // Skip the original expression.
4881     AsmStart = Loc + AR.Len + AdditionalSkip;
4882   }
4883 
4884   // Emit the remainder of the asm string.
4885   if (AsmStart != AsmEnd)
4886     OS << StringRef(AsmStart, AsmEnd - AsmStart);
4887 
4888   AsmString = OS.str();
4889   return false;
4890 }
4891 
4892 namespace llvm {
4893 namespace MCParserUtils {
4894 
4895 /// Returns whether the given symbol is used anywhere in the given expression,
4896 /// or subexpressions.
isSymbolUsedInExpression(const MCSymbol * Sym,const MCExpr * Value)4897 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) {
4898   switch (Value->getKind()) {
4899   case MCExpr::Binary: {
4900     const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value);
4901     return isSymbolUsedInExpression(Sym, BE->getLHS()) ||
4902            isSymbolUsedInExpression(Sym, BE->getRHS());
4903   }
4904   case MCExpr::Target:
4905   case MCExpr::Constant:
4906     return false;
4907   case MCExpr::SymbolRef: {
4908     const MCSymbol &S =
4909         static_cast<const MCSymbolRefExpr *>(Value)->getSymbol();
4910     if (S.isVariable())
4911       return isSymbolUsedInExpression(Sym, S.getVariableValue());
4912     return &S == Sym;
4913   }
4914   case MCExpr::Unary:
4915     return isSymbolUsedInExpression(
4916         Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr());
4917   }
4918 
4919   llvm_unreachable("Unknown expr kind!");
4920 }
4921 
parseAssignmentExpression(StringRef Name,bool allow_redef,MCAsmParser & Parser,MCSymbol * & Sym,const MCExpr * & Value)4922 bool parseAssignmentExpression(StringRef Name, bool allow_redef,
4923                                MCAsmParser &Parser, MCSymbol *&Sym,
4924                                const MCExpr *&Value) {
4925   MCAsmLexer &Lexer = Parser.getLexer();
4926 
4927   // FIXME: Use better location, we should use proper tokens.
4928   SMLoc EqualLoc = Lexer.getLoc();
4929 
4930   if (Parser.parseExpression(Value)) {
4931     Parser.TokError("missing expression");
4932     Parser.eatToEndOfStatement();
4933     return true;
4934   }
4935 
4936   // Note: we don't count b as used in "a = b". This is to allow
4937   // a = b
4938   // b = c
4939 
4940   if (Lexer.isNot(AsmToken::EndOfStatement))
4941     return Parser.TokError("unexpected token in assignment");
4942 
4943   // Eat the end of statement marker.
4944   Parser.Lex();
4945 
4946   // Validate that the LHS is allowed to be a variable (either it has not been
4947   // used as a symbol, or it is an absolute symbol).
4948   Sym = Parser.getContext().lookupSymbol(Name);
4949   if (Sym) {
4950     // Diagnose assignment to a label.
4951     //
4952     // FIXME: Diagnostics. Note the location of the definition as a label.
4953     // FIXME: Diagnose assignment to protected identifier (e.g., register name).
4954     if (isSymbolUsedInExpression(Sym, Value))
4955       return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'");
4956     else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() &&
4957              !Sym->isVariable())
4958       ; // Allow redefinitions of undefined symbols only used in directives.
4959     else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
4960       ; // Allow redefinitions of variables that haven't yet been used.
4961     else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
4962       return Parser.Error(EqualLoc, "redefinition of '" + Name + "'");
4963     else if (!Sym->isVariable())
4964       return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'");
4965     else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
4966       return Parser.Error(EqualLoc,
4967                           "invalid reassignment of non-absolute variable '" +
4968                               Name + "'");
4969   } else if (Name == ".") {
4970     Parser.getStreamer().emitValueToOffset(Value, 0);
4971     return false;
4972   } else
4973     Sym = Parser.getContext().getOrCreateSymbol(Name);
4974 
4975   Sym->setRedefinable(allow_redef);
4976 
4977   return false;
4978 }
4979 
4980 } // namespace MCParserUtils
4981 } // namespace llvm
4982 
4983 /// \brief Create an MCAsmParser instance.
createMCAsmParser(SourceMgr & SM,MCContext & C,MCStreamer & Out,const MCAsmInfo & MAI)4984 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C,
4985                                      MCStreamer &Out, const MCAsmInfo &MAI) {
4986   return new AsmParser(SM, C, Out, MAI);
4987 }
4988