xref: /freebsd/contrib/llvm-project/llvm/lib/MC/MCParser/AsmParser.cpp (revision 0fca6ea1d4eea4c934cfff25ac9ee8ad6fe95583)
1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This class implements a parser for assembly files similar to gas syntax.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #include "llvm/ADT/APFloat.h"
14 #include "llvm/ADT/APInt.h"
15 #include "llvm/ADT/ArrayRef.h"
16 #include "llvm/ADT/STLExtras.h"
17 #include "llvm/ADT/SmallSet.h"
18 #include "llvm/ADT/SmallString.h"
19 #include "llvm/ADT/SmallVector.h"
20 #include "llvm/ADT/StringExtras.h"
21 #include "llvm/ADT/StringMap.h"
22 #include "llvm/ADT/StringRef.h"
23 #include "llvm/ADT/Twine.h"
24 #include "llvm/BinaryFormat/Dwarf.h"
25 #include "llvm/DebugInfo/CodeView/SymbolRecord.h"
26 #include "llvm/MC/MCAsmInfo.h"
27 #include "llvm/MC/MCCodeView.h"
28 #include "llvm/MC/MCContext.h"
29 #include "llvm/MC/MCDirectives.h"
30 #include "llvm/MC/MCDwarf.h"
31 #include "llvm/MC/MCExpr.h"
32 #include "llvm/MC/MCInstPrinter.h"
33 #include "llvm/MC/MCInstrDesc.h"
34 #include "llvm/MC/MCInstrInfo.h"
35 #include "llvm/MC/MCParser/AsmCond.h"
36 #include "llvm/MC/MCParser/AsmLexer.h"
37 #include "llvm/MC/MCParser/MCAsmLexer.h"
38 #include "llvm/MC/MCParser/MCAsmParser.h"
39 #include "llvm/MC/MCParser/MCAsmParserExtension.h"
40 #include "llvm/MC/MCParser/MCAsmParserUtils.h"
41 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
42 #include "llvm/MC/MCParser/MCTargetAsmParser.h"
43 #include "llvm/MC/MCRegisterInfo.h"
44 #include "llvm/MC/MCSection.h"
45 #include "llvm/MC/MCStreamer.h"
46 #include "llvm/MC/MCSymbol.h"
47 #include "llvm/MC/MCSymbolMachO.h"
48 #include "llvm/MC/MCTargetOptions.h"
49 #include "llvm/MC/MCValue.h"
50 #include "llvm/Support/Casting.h"
51 #include "llvm/Support/CommandLine.h"
52 #include "llvm/Support/ErrorHandling.h"
53 #include "llvm/Support/MD5.h"
54 #include "llvm/Support/MathExtras.h"
55 #include "llvm/Support/MemoryBuffer.h"
56 #include "llvm/Support/SMLoc.h"
57 #include "llvm/Support/SourceMgr.h"
58 #include "llvm/Support/raw_ostream.h"
59 #include <algorithm>
60 #include <cassert>
61 #include <cctype>
62 #include <climits>
63 #include <cstddef>
64 #include <cstdint>
65 #include <deque>
66 #include <memory>
67 #include <optional>
68 #include <sstream>
69 #include <string>
70 #include <tuple>
71 #include <utility>
72 #include <vector>
73 
74 using namespace llvm;
75 
76 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() = default;
77 
78 namespace {
79 
80 /// Helper types for tracking macro definitions.
81 typedef std::vector<AsmToken> MCAsmMacroArgument;
82 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments;
83 
84 /// Helper class for storing information about an active macro
85 /// instantiation.
86 struct MacroInstantiation {
87   /// The location of the instantiation.
88   SMLoc InstantiationLoc;
89 
90   /// The buffer where parsing should resume upon instantiation completion.
91   unsigned ExitBuffer;
92 
93   /// The location where parsing should resume upon instantiation completion.
94   SMLoc ExitLoc;
95 
96   /// The depth of TheCondStack at the start of the instantiation.
97   size_t CondStackDepth;
98 };
99 
100 struct ParseStatementInfo {
101   /// The parsed operands from the last parsed statement.
102   SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands;
103 
104   /// The opcode from the last parsed instruction.
105   unsigned Opcode = ~0U;
106 
107   /// Was there an error parsing the inline assembly?
108   bool ParseError = false;
109 
110   SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr;
111 
112   ParseStatementInfo() = delete;
ParseStatementInfo__anon4b9616090111::ParseStatementInfo113   ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
114     : AsmRewrites(rewrites) {}
115 };
116 
117 /// The concrete assembly parser instance.
118 class AsmParser : public MCAsmParser {
119 private:
120   AsmLexer Lexer;
121   MCContext &Ctx;
122   MCStreamer &Out;
123   const MCAsmInfo &MAI;
124   SourceMgr &SrcMgr;
125   SourceMgr::DiagHandlerTy SavedDiagHandler;
126   void *SavedDiagContext;
127   std::unique_ptr<MCAsmParserExtension> PlatformParser;
128   SMLoc StartTokLoc;
129   std::optional<SMLoc> CFIStartProcLoc;
130 
131   /// This is the current buffer index we're lexing from as managed by the
132   /// SourceMgr object.
133   unsigned CurBuffer;
134 
135   AsmCond TheCondState;
136   std::vector<AsmCond> TheCondStack;
137 
138   /// maps directive names to handler methods in parser
139   /// extensions. Extensions register themselves in this map by calling
140   /// addDirectiveHandler.
141   StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap;
142 
143   /// Stack of active macro instantiations.
144   std::vector<MacroInstantiation*> ActiveMacros;
145 
146   /// 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   /// Keeps track of how many .macro's have been instantiated.
153   unsigned NumOfMacroInstantiations;
154 
155   /// The values from the last parsed cpp hash file line comment if any.
156   struct CppHashInfoTy {
157     StringRef Filename;
158     int64_t LineNumber;
159     SMLoc Loc;
160     unsigned Buf;
CppHashInfoTy__anon4b9616090111::AsmParser::CppHashInfoTy161     CppHashInfoTy() : LineNumber(0), Buf(0) {}
162   };
163   CppHashInfoTy CppHashInfo;
164 
165   /// The filename from the first cpp hash file line comment, if any.
166   StringRef FirstCppHashFilename;
167 
168   /// List of forward directional labels for diagnosis at the end.
169   SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels;
170 
171   SmallSet<StringRef, 2> LTODiscardSymbols;
172 
173   /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
174   unsigned AssemblerDialect = ~0U;
175 
176   /// is Darwin compatibility enabled?
177   bool IsDarwin = false;
178 
179   /// Are we parsing ms-style inline assembly?
180   bool ParsingMSInlineAsm = false;
181 
182   /// Did we already inform the user about inconsistent MD5 usage?
183   bool ReportedInconsistentMD5 = false;
184 
185   // Is alt macro mode enabled.
186   bool AltMacroMode = false;
187 
188 protected:
189   virtual bool parseStatement(ParseStatementInfo &Info,
190                               MCAsmParserSemaCallback *SI);
191 
192   /// This routine uses the target specific ParseInstruction function to
193   /// parse an instruction into Operands, and then call the target specific
194   /// MatchAndEmit function to match and emit the instruction.
195   bool parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info,
196                                              StringRef IDVal, AsmToken ID,
197                                              SMLoc IDLoc);
198 
199   /// Should we emit DWARF describing this assembler source?  (Returns false if
200   /// the source has .file directives, which means we don't want to generate
201   /// info describing the assembler source itself.)
202   bool enabledGenDwarfForAssembly();
203 
204 public:
205   AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
206             const MCAsmInfo &MAI, unsigned CB);
207   AsmParser(const AsmParser &) = delete;
208   AsmParser &operator=(const AsmParser &) = delete;
209   ~AsmParser() override;
210 
211   bool Run(bool NoInitialTextSection, bool NoFinalize = false) override;
212 
addDirectiveHandler(StringRef Directive,ExtensionDirectiveHandler Handler)213   void addDirectiveHandler(StringRef Directive,
214                            ExtensionDirectiveHandler Handler) override {
215     ExtensionDirectiveMap[Directive] = Handler;
216   }
217 
addAliasForDirective(StringRef Directive,StringRef Alias)218   void addAliasForDirective(StringRef Directive, StringRef Alias) override {
219     DirectiveKindMap[Directive.lower()] = DirectiveKindMap[Alias.lower()];
220   }
221 
222   /// @name MCAsmParser Interface
223   /// {
224 
getSourceManager()225   SourceMgr &getSourceManager() override { return SrcMgr; }
getLexer()226   MCAsmLexer &getLexer() override { return Lexer; }
getContext()227   MCContext &getContext() override { return Ctx; }
getStreamer()228   MCStreamer &getStreamer() override { return Out; }
229 
getCVContext()230   CodeViewContext &getCVContext() { return Ctx.getCVContext(); }
231 
getAssemblerDialect()232   unsigned getAssemblerDialect() override {
233     if (AssemblerDialect == ~0U)
234       return MAI.getAssemblerDialect();
235     else
236       return AssemblerDialect;
237   }
setAssemblerDialect(unsigned i)238   void setAssemblerDialect(unsigned i) override {
239     AssemblerDialect = i;
240   }
241 
242   void Note(SMLoc L, const Twine &Msg, SMRange Range = std::nullopt) override;
243   bool Warning(SMLoc L, const Twine &Msg,
244                SMRange Range = std::nullopt) override;
245   bool printError(SMLoc L, const Twine &Msg,
246                   SMRange Range = std::nullopt) override;
247 
248   const AsmToken &Lex() override;
249 
setParsingMSInlineAsm(bool V)250   void setParsingMSInlineAsm(bool V) override {
251     ParsingMSInlineAsm = V;
252     // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and
253     // hex integer literals.
254     Lexer.setLexMasmIntegers(V);
255   }
isParsingMSInlineAsm()256   bool isParsingMSInlineAsm() override { return ParsingMSInlineAsm; }
257 
discardLTOSymbol(StringRef Name) const258   bool discardLTOSymbol(StringRef Name) const override {
259     return LTODiscardSymbols.contains(Name);
260   }
261 
262   bool parseMSInlineAsm(std::string &AsmString, unsigned &NumOutputs,
263                         unsigned &NumInputs,
264                         SmallVectorImpl<std::pair<void *, bool>> &OpDecls,
265                         SmallVectorImpl<std::string> &Constraints,
266                         SmallVectorImpl<std::string> &Clobbers,
267                         const MCInstrInfo *MII, const MCInstPrinter *IP,
268                         MCAsmParserSemaCallback &SI) override;
269 
270   bool parseExpression(const MCExpr *&Res);
271   bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
272   bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc,
273                         AsmTypeInfo *TypeInfo) override;
274   bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
275   bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
276                              SMLoc &EndLoc) override;
277   bool parseAbsoluteExpression(int64_t &Res) override;
278 
279   /// Parse a floating point expression using the float \p Semantics
280   /// and set \p Res to the value.
281   bool parseRealValue(const fltSemantics &Semantics, APInt &Res);
282 
283   /// Parse an identifier or string (as a quoted identifier)
284   /// and set \p Res to the identifier contents.
285   bool parseIdentifier(StringRef &Res) override;
286   void eatToEndOfStatement() override;
287 
288   bool checkForValidSection() override;
289 
290   /// }
291 
292 private:
293   bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites);
294   bool parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo = true);
295 
296   void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body,
297                         ArrayRef<MCAsmMacroParameter> Parameters);
298   bool expandMacro(raw_svector_ostream &OS, MCAsmMacro &Macro,
299                    ArrayRef<MCAsmMacroParameter> Parameters,
300                    ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable);
301 
302   /// Are macros enabled in the parser?
areMacrosEnabled()303   bool areMacrosEnabled() {return MacrosEnabledFlag;}
304 
305   /// Control a flag in the parser that enables or disables macros.
setMacrosEnabled(bool Flag)306   void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;}
307 
308   /// Are we inside a macro instantiation?
isInsideMacroInstantiation()309   bool isInsideMacroInstantiation() {return !ActiveMacros.empty();}
310 
311   /// Handle entry to macro instantiation.
312   ///
313   /// \param M The macro.
314   /// \param NameLoc Instantiation location.
315   bool handleMacroEntry(MCAsmMacro *M, SMLoc NameLoc);
316 
317   /// Handle exit from macro instantiation.
318   void handleMacroExit();
319 
320   /// Extract AsmTokens for a macro argument.
321   bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg);
322 
323   /// Parse all macro arguments for a given macro.
324   bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A);
325 
326   void printMacroInstantiations();
printMessage(SMLoc Loc,SourceMgr::DiagKind Kind,const Twine & Msg,SMRange Range=std::nullopt) const327   void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
328                     SMRange Range = std::nullopt) const {
329     ArrayRef<SMRange> Ranges(Range);
330     SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
331   }
332   static void DiagHandler(const SMDiagnostic &Diag, void *Context);
333 
334   /// Enter the specified file. This returns true on failure.
335   bool enterIncludeFile(const std::string &Filename);
336 
337   /// Process the specified file for the .incbin directive.
338   /// This returns true on failure.
339   bool processIncbinFile(const std::string &Filename, int64_t Skip = 0,
340                          const MCExpr *Count = nullptr, SMLoc Loc = SMLoc());
341 
342   /// Reset the current lexer position to that given by \p Loc. The
343   /// current token is not set; clients should ensure Lex() is called
344   /// subsequently.
345   ///
346   /// \param InBuffer If not 0, should be the known buffer id that contains the
347   /// location.
348   void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0);
349 
350   /// Parse up to the end of statement and a return the contents from the
351   /// current token until the end of the statement; the current token on exit
352   /// will be either the EndOfStatement or EOF.
353   StringRef parseStringToEndOfStatement() override;
354 
355   /// Parse until the end of a statement or a comma is encountered,
356   /// return the contents from the current token up to the end or comma.
357   StringRef parseStringToComma();
358 
359   enum class AssignmentKind {
360     Set,
361     Equiv,
362     Equal,
363     LTOSetConditional,
364   };
365 
366   bool parseAssignment(StringRef Name, AssignmentKind Kind);
367 
368   unsigned getBinOpPrecedence(AsmToken::TokenKind K,
369                               MCBinaryExpr::Opcode &Kind);
370 
371   bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
372   bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
373   bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
374 
375   bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
376 
377   bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName);
378   bool parseCVFileId(int64_t &FileId, StringRef DirectiveName);
379 
380   // Generic (target and platform independent) directive parsing.
381   enum DirectiveKind {
382     DK_NO_DIRECTIVE, // Placeholder
383     DK_SET,
384     DK_EQU,
385     DK_EQUIV,
386     DK_ASCII,
387     DK_ASCIZ,
388     DK_STRING,
389     DK_BYTE,
390     DK_SHORT,
391     DK_RELOC,
392     DK_VALUE,
393     DK_2BYTE,
394     DK_LONG,
395     DK_INT,
396     DK_4BYTE,
397     DK_QUAD,
398     DK_8BYTE,
399     DK_OCTA,
400     DK_DC,
401     DK_DC_A,
402     DK_DC_B,
403     DK_DC_D,
404     DK_DC_L,
405     DK_DC_S,
406     DK_DC_W,
407     DK_DC_X,
408     DK_DCB,
409     DK_DCB_B,
410     DK_DCB_D,
411     DK_DCB_L,
412     DK_DCB_S,
413     DK_DCB_W,
414     DK_DCB_X,
415     DK_DS,
416     DK_DS_B,
417     DK_DS_D,
418     DK_DS_L,
419     DK_DS_P,
420     DK_DS_S,
421     DK_DS_W,
422     DK_DS_X,
423     DK_SINGLE,
424     DK_FLOAT,
425     DK_DOUBLE,
426     DK_ALIGN,
427     DK_ALIGN32,
428     DK_BALIGN,
429     DK_BALIGNW,
430     DK_BALIGNL,
431     DK_P2ALIGN,
432     DK_P2ALIGNW,
433     DK_P2ALIGNL,
434     DK_ORG,
435     DK_FILL,
436     DK_ENDR,
437     DK_BUNDLE_ALIGN_MODE,
438     DK_BUNDLE_LOCK,
439     DK_BUNDLE_UNLOCK,
440     DK_ZERO,
441     DK_EXTERN,
442     DK_GLOBL,
443     DK_GLOBAL,
444     DK_LAZY_REFERENCE,
445     DK_NO_DEAD_STRIP,
446     DK_SYMBOL_RESOLVER,
447     DK_PRIVATE_EXTERN,
448     DK_REFERENCE,
449     DK_WEAK_DEFINITION,
450     DK_WEAK_REFERENCE,
451     DK_WEAK_DEF_CAN_BE_HIDDEN,
452     DK_COLD,
453     DK_COMM,
454     DK_COMMON,
455     DK_LCOMM,
456     DK_ABORT,
457     DK_INCLUDE,
458     DK_INCBIN,
459     DK_CODE16,
460     DK_CODE16GCC,
461     DK_REPT,
462     DK_IRP,
463     DK_IRPC,
464     DK_IF,
465     DK_IFEQ,
466     DK_IFGE,
467     DK_IFGT,
468     DK_IFLE,
469     DK_IFLT,
470     DK_IFNE,
471     DK_IFB,
472     DK_IFNB,
473     DK_IFC,
474     DK_IFEQS,
475     DK_IFNC,
476     DK_IFNES,
477     DK_IFDEF,
478     DK_IFNDEF,
479     DK_IFNOTDEF,
480     DK_ELSEIF,
481     DK_ELSE,
482     DK_ENDIF,
483     DK_SPACE,
484     DK_SKIP,
485     DK_FILE,
486     DK_LINE,
487     DK_LOC,
488     DK_STABS,
489     DK_CV_FILE,
490     DK_CV_FUNC_ID,
491     DK_CV_INLINE_SITE_ID,
492     DK_CV_LOC,
493     DK_CV_LINETABLE,
494     DK_CV_INLINE_LINETABLE,
495     DK_CV_DEF_RANGE,
496     DK_CV_STRINGTABLE,
497     DK_CV_STRING,
498     DK_CV_FILECHECKSUMS,
499     DK_CV_FILECHECKSUM_OFFSET,
500     DK_CV_FPO_DATA,
501     DK_CFI_SECTIONS,
502     DK_CFI_STARTPROC,
503     DK_CFI_ENDPROC,
504     DK_CFI_DEF_CFA,
505     DK_CFI_DEF_CFA_OFFSET,
506     DK_CFI_ADJUST_CFA_OFFSET,
507     DK_CFI_DEF_CFA_REGISTER,
508     DK_CFI_LLVM_DEF_ASPACE_CFA,
509     DK_CFI_OFFSET,
510     DK_CFI_REL_OFFSET,
511     DK_CFI_PERSONALITY,
512     DK_CFI_LSDA,
513     DK_CFI_REMEMBER_STATE,
514     DK_CFI_RESTORE_STATE,
515     DK_CFI_SAME_VALUE,
516     DK_CFI_RESTORE,
517     DK_CFI_ESCAPE,
518     DK_CFI_RETURN_COLUMN,
519     DK_CFI_SIGNAL_FRAME,
520     DK_CFI_UNDEFINED,
521     DK_CFI_REGISTER,
522     DK_CFI_WINDOW_SAVE,
523     DK_CFI_LABEL,
524     DK_CFI_B_KEY_FRAME,
525     DK_MACROS_ON,
526     DK_MACROS_OFF,
527     DK_ALTMACRO,
528     DK_NOALTMACRO,
529     DK_MACRO,
530     DK_EXITM,
531     DK_ENDM,
532     DK_ENDMACRO,
533     DK_PURGEM,
534     DK_SLEB128,
535     DK_ULEB128,
536     DK_ERR,
537     DK_ERROR,
538     DK_WARNING,
539     DK_PRINT,
540     DK_ADDRSIG,
541     DK_ADDRSIG_SYM,
542     DK_PSEUDO_PROBE,
543     DK_LTO_DISCARD,
544     DK_LTO_SET_CONDITIONAL,
545     DK_CFI_MTE_TAGGED_FRAME,
546     DK_MEMTAG,
547     DK_END
548   };
549 
550   /// Maps directive name --> DirectiveKind enum, for
551   /// directives parsed by this class.
552   StringMap<DirectiveKind> DirectiveKindMap;
553 
554   // Codeview def_range type parsing.
555   enum CVDefRangeType {
556     CVDR_DEFRANGE = 0, // Placeholder
557     CVDR_DEFRANGE_REGISTER,
558     CVDR_DEFRANGE_FRAMEPOINTER_REL,
559     CVDR_DEFRANGE_SUBFIELD_REGISTER,
560     CVDR_DEFRANGE_REGISTER_REL
561   };
562 
563   /// Maps Codeview def_range types --> CVDefRangeType enum, for
564   /// Codeview def_range types parsed by this class.
565   StringMap<CVDefRangeType> CVDefRangeTypeMap;
566 
567   // ".ascii", ".asciz", ".string"
568   bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
569   bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc"
570   bool parseDirectiveValue(StringRef IDVal,
571                            unsigned Size);       // ".byte", ".long", ...
572   bool parseDirectiveOctaValue(StringRef IDVal); // ".octa", ...
573   bool parseDirectiveRealValue(StringRef IDVal,
574                                const fltSemantics &); // ".single", ...
575   bool parseDirectiveFill(); // ".fill"
576   bool parseDirectiveZero(); // ".zero"
577   // ".set", ".equ", ".equiv", ".lto_set_conditional"
578   bool parseDirectiveSet(StringRef IDVal, AssignmentKind Kind);
579   bool parseDirectiveOrg(); // ".org"
580   // ".align{,32}", ".p2align{,w,l}"
581   bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize);
582 
583   // ".file", ".line", ".loc", ".stabs"
584   bool parseDirectiveFile(SMLoc DirectiveLoc);
585   bool parseDirectiveLine();
586   bool parseDirectiveLoc();
587   bool parseDirectiveStabs();
588 
589   // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable",
590   // ".cv_inline_linetable", ".cv_def_range", ".cv_string"
591   bool parseDirectiveCVFile();
592   bool parseDirectiveCVFuncId();
593   bool parseDirectiveCVInlineSiteId();
594   bool parseDirectiveCVLoc();
595   bool parseDirectiveCVLinetable();
596   bool parseDirectiveCVInlineLinetable();
597   bool parseDirectiveCVDefRange();
598   bool parseDirectiveCVString();
599   bool parseDirectiveCVStringTable();
600   bool parseDirectiveCVFileChecksums();
601   bool parseDirectiveCVFileChecksumOffset();
602   bool parseDirectiveCVFPOData();
603 
604   // .cfi directives
605   bool parseDirectiveCFIRegister(SMLoc DirectiveLoc);
606   bool parseDirectiveCFIWindowSave(SMLoc DirectiveLoc);
607   bool parseDirectiveCFISections();
608   bool parseDirectiveCFIStartProc();
609   bool parseDirectiveCFIEndProc();
610   bool parseDirectiveCFIDefCfaOffset(SMLoc DirectiveLoc);
611   bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc);
612   bool parseDirectiveCFIAdjustCfaOffset(SMLoc DirectiveLoc);
613   bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc);
614   bool parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc);
615   bool parseDirectiveCFIOffset(SMLoc DirectiveLoc);
616   bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc);
617   bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality);
618   bool parseDirectiveCFIRememberState(SMLoc DirectiveLoc);
619   bool parseDirectiveCFIRestoreState(SMLoc DirectiveLoc);
620   bool parseDirectiveCFISameValue(SMLoc DirectiveLoc);
621   bool parseDirectiveCFIRestore(SMLoc DirectiveLoc);
622   bool parseDirectiveCFIEscape(SMLoc DirectiveLoc);
623   bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc);
624   bool parseDirectiveCFISignalFrame(SMLoc DirectiveLoc);
625   bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc);
626   bool parseDirectiveCFILabel(SMLoc DirectiveLoc);
627 
628   // macro directives
629   bool parseDirectivePurgeMacro(SMLoc DirectiveLoc);
630   bool parseDirectiveExitMacro(StringRef Directive);
631   bool parseDirectiveEndMacro(StringRef Directive);
632   bool parseDirectiveMacro(SMLoc DirectiveLoc);
633   bool parseDirectiveMacrosOnOff(StringRef Directive);
634   // alternate macro mode directives
635   bool parseDirectiveAltmacro(StringRef Directive);
636   // ".bundle_align_mode"
637   bool parseDirectiveBundleAlignMode();
638   // ".bundle_lock"
639   bool parseDirectiveBundleLock();
640   // ".bundle_unlock"
641   bool parseDirectiveBundleUnlock();
642 
643   // ".space", ".skip"
644   bool parseDirectiveSpace(StringRef IDVal);
645 
646   // ".dcb"
647   bool parseDirectiveDCB(StringRef IDVal, unsigned Size);
648   bool parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &);
649   // ".ds"
650   bool parseDirectiveDS(StringRef IDVal, unsigned Size);
651 
652   // .sleb128 (Signed=true) and .uleb128 (Signed=false)
653   bool parseDirectiveLEB128(bool Signed);
654 
655   /// Parse a directive like ".globl" which
656   /// accepts a single symbol (which should be a label or an external).
657   bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr);
658 
659   bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
660 
661   bool parseDirectiveAbort(SMLoc DirectiveLoc); // ".abort"
662   bool parseDirectiveInclude(); // ".include"
663   bool parseDirectiveIncbin(); // ".incbin"
664 
665   // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne"
666   bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
667   // ".ifb" or ".ifnb", depending on ExpectBlank.
668   bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
669   // ".ifc" or ".ifnc", depending on ExpectEqual.
670   bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
671   // ".ifeqs" or ".ifnes", depending on ExpectEqual.
672   bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual);
673   // ".ifdef" or ".ifndef", depending on expect_defined
674   bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
675   bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
676   bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else"
677   bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
678   bool parseEscapedString(std::string &Data) override;
679   bool parseAngleBracketString(std::string &Data) override;
680 
681   const MCExpr *applyModifierToExpr(const MCExpr *E,
682                                     MCSymbolRefExpr::VariantKind Variant);
683 
684   // Macro-like directives
685   MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
686   void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
687                                 raw_svector_ostream &OS);
688   bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive);
689   bool parseDirectiveIrp(SMLoc DirectiveLoc);  // ".irp"
690   bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
691   bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
692 
693   // "_emit" or "__emit"
694   bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info,
695                             size_t Len);
696 
697   // "align"
698   bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info);
699 
700   // "end"
701   bool parseDirectiveEnd(SMLoc DirectiveLoc);
702 
703   // ".err" or ".error"
704   bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage);
705 
706   // ".warning"
707   bool parseDirectiveWarning(SMLoc DirectiveLoc);
708 
709   // .print <double-quotes-string>
710   bool parseDirectivePrint(SMLoc DirectiveLoc);
711 
712   // .pseudoprobe
713   bool parseDirectivePseudoProbe();
714 
715   // ".lto_discard"
716   bool parseDirectiveLTODiscard();
717 
718   // Directives to support address-significance tables.
719   bool parseDirectiveAddrsig();
720   bool parseDirectiveAddrsigSym();
721 
722   void initializeDirectiveKindMap();
723   void initializeCVDefRangeTypeMap();
724 };
725 
726 class HLASMAsmParser final : public AsmParser {
727 private:
728   MCAsmLexer &Lexer;
729   MCStreamer &Out;
730 
lexLeadingSpaces()731   void lexLeadingSpaces() {
732     while (Lexer.is(AsmToken::Space))
733       Lexer.Lex();
734   }
735 
736   bool parseAsHLASMLabel(ParseStatementInfo &Info, MCAsmParserSemaCallback *SI);
737   bool parseAsMachineInstruction(ParseStatementInfo &Info,
738                                  MCAsmParserSemaCallback *SI);
739 
740 public:
HLASMAsmParser(SourceMgr & SM,MCContext & Ctx,MCStreamer & Out,const MCAsmInfo & MAI,unsigned CB=0)741   HLASMAsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
742                  const MCAsmInfo &MAI, unsigned CB = 0)
743       : AsmParser(SM, Ctx, Out, MAI, CB), Lexer(getLexer()), Out(Out) {
744     Lexer.setSkipSpace(false);
745     Lexer.setAllowHashInIdentifier(true);
746     Lexer.setLexHLASMIntegers(true);
747     Lexer.setLexHLASMStrings(true);
748   }
749 
~HLASMAsmParser()750   ~HLASMAsmParser() { Lexer.setSkipSpace(true); }
751 
752   bool parseStatement(ParseStatementInfo &Info,
753                       MCAsmParserSemaCallback *SI) override;
754 };
755 
756 } // end anonymous namespace
757 
758 namespace llvm {
759 
760 extern cl::opt<unsigned> AsmMacroMaxNestingDepth;
761 
762 extern MCAsmParserExtension *createDarwinAsmParser();
763 extern MCAsmParserExtension *createELFAsmParser();
764 extern MCAsmParserExtension *createCOFFAsmParser();
765 extern MCAsmParserExtension *createGOFFAsmParser();
766 extern MCAsmParserExtension *createXCOFFAsmParser();
767 extern MCAsmParserExtension *createWasmAsmParser();
768 
769 } // end namespace llvm
770 
771 enum { DEFAULT_ADDRSPACE = 0 };
772 
AsmParser(SourceMgr & SM,MCContext & Ctx,MCStreamer & Out,const MCAsmInfo & MAI,unsigned CB=0)773 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
774                      const MCAsmInfo &MAI, unsigned CB = 0)
775     : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM),
776       CurBuffer(CB ? CB : SM.getMainFileID()), MacrosEnabledFlag(true) {
777   HadError = false;
778   // Save the old handler.
779   SavedDiagHandler = SrcMgr.getDiagHandler();
780   SavedDiagContext = SrcMgr.getDiagContext();
781   // Set our own handler which calls the saved handler.
782   SrcMgr.setDiagHandler(DiagHandler, this);
783   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
784   // Make MCStreamer aware of the StartTokLoc for locations in diagnostics.
785   Out.setStartTokLocPtr(&StartTokLoc);
786 
787   // Initialize the platform / file format parser.
788   switch (Ctx.getObjectFileType()) {
789   case MCContext::IsCOFF:
790     PlatformParser.reset(createCOFFAsmParser());
791     break;
792   case MCContext::IsMachO:
793     PlatformParser.reset(createDarwinAsmParser());
794     IsDarwin = true;
795     break;
796   case MCContext::IsELF:
797     PlatformParser.reset(createELFAsmParser());
798     break;
799   case MCContext::IsGOFF:
800     PlatformParser.reset(createGOFFAsmParser());
801     break;
802   case MCContext::IsSPIRV:
803     report_fatal_error(
804         "Need to implement createSPIRVAsmParser for SPIRV format.");
805     break;
806   case MCContext::IsWasm:
807     PlatformParser.reset(createWasmAsmParser());
808     break;
809   case MCContext::IsXCOFF:
810     PlatformParser.reset(createXCOFFAsmParser());
811     break;
812   case MCContext::IsDXContainer:
813     report_fatal_error("DXContainer is not supported yet");
814     break;
815   }
816 
817   PlatformParser->Initialize(*this);
818   initializeDirectiveKindMap();
819   initializeCVDefRangeTypeMap();
820 
821   NumOfMacroInstantiations = 0;
822 }
823 
~AsmParser()824 AsmParser::~AsmParser() {
825   assert((HadError || ActiveMacros.empty()) &&
826          "Unexpected active macro instantiation!");
827 
828   // Remove MCStreamer's reference to the parser SMLoc.
829   Out.setStartTokLocPtr(nullptr);
830   // Restore the saved diagnostics handler and context for use during
831   // finalization.
832   SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext);
833 }
834 
printMacroInstantiations()835 void AsmParser::printMacroInstantiations() {
836   // Print the active macro instantiation stack.
837   for (MacroInstantiation *M : reverse(ActiveMacros))
838     printMessage(M->InstantiationLoc, SourceMgr::DK_Note,
839                  "while in macro instantiation");
840 }
841 
Note(SMLoc L,const Twine & Msg,SMRange Range)842 void AsmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) {
843   printPendingErrors();
844   printMessage(L, SourceMgr::DK_Note, Msg, Range);
845   printMacroInstantiations();
846 }
847 
Warning(SMLoc L,const Twine & Msg,SMRange Range)848 bool AsmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) {
849   if(getTargetParser().getTargetOptions().MCNoWarn)
850     return false;
851   if (getTargetParser().getTargetOptions().MCFatalWarnings)
852     return Error(L, Msg, Range);
853   printMessage(L, SourceMgr::DK_Warning, Msg, Range);
854   printMacroInstantiations();
855   return false;
856 }
857 
printError(SMLoc L,const Twine & Msg,SMRange Range)858 bool AsmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) {
859   HadError = true;
860   printMessage(L, SourceMgr::DK_Error, Msg, Range);
861   printMacroInstantiations();
862   return true;
863 }
864 
enterIncludeFile(const std::string & Filename)865 bool AsmParser::enterIncludeFile(const std::string &Filename) {
866   std::string IncludedFile;
867   unsigned NewBuf =
868       SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
869   if (!NewBuf)
870     return true;
871 
872   CurBuffer = NewBuf;
873   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
874   return false;
875 }
876 
877 /// Process the specified .incbin file by searching for it in the include paths
878 /// then just emitting the byte contents of the file to the streamer. This
879 /// returns true on failure.
processIncbinFile(const std::string & Filename,int64_t Skip,const MCExpr * Count,SMLoc Loc)880 bool AsmParser::processIncbinFile(const std::string &Filename, int64_t Skip,
881                                   const MCExpr *Count, SMLoc Loc) {
882   std::string IncludedFile;
883   unsigned NewBuf =
884       SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
885   if (!NewBuf)
886     return true;
887 
888   // Pick up the bytes from the file and emit them.
889   StringRef Bytes = SrcMgr.getMemoryBuffer(NewBuf)->getBuffer();
890   Bytes = Bytes.drop_front(Skip);
891   if (Count) {
892     int64_t Res;
893     if (!Count->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr()))
894       return Error(Loc, "expected absolute expression");
895     if (Res < 0)
896       return Warning(Loc, "negative count has no effect");
897     Bytes = Bytes.take_front(Res);
898   }
899   getStreamer().emitBytes(Bytes);
900   return false;
901 }
902 
jumpToLoc(SMLoc Loc,unsigned InBuffer)903 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) {
904   CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc);
905   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(),
906                   Loc.getPointer());
907 }
908 
Lex()909 const AsmToken &AsmParser::Lex() {
910   if (Lexer.getTok().is(AsmToken::Error))
911     Error(Lexer.getErrLoc(), Lexer.getErr());
912 
913   // if it's a end of statement with a comment in it
914   if (getTok().is(AsmToken::EndOfStatement)) {
915     // if this is a line comment output it.
916     if (!getTok().getString().empty() && getTok().getString().front() != '\n' &&
917         getTok().getString().front() != '\r' && MAI.preserveAsmComments())
918       Out.addExplicitComment(Twine(getTok().getString()));
919   }
920 
921   const AsmToken *tok = &Lexer.Lex();
922 
923   // Parse comments here to be deferred until end of next statement.
924   while (tok->is(AsmToken::Comment)) {
925     if (MAI.preserveAsmComments())
926       Out.addExplicitComment(Twine(tok->getString()));
927     tok = &Lexer.Lex();
928   }
929 
930   if (tok->is(AsmToken::Eof)) {
931     // If this is the end of an included file, pop the parent file off the
932     // include stack.
933     SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
934     if (ParentIncludeLoc != SMLoc()) {
935       jumpToLoc(ParentIncludeLoc);
936       return Lex();
937     }
938   }
939 
940   return *tok;
941 }
942 
enabledGenDwarfForAssembly()943 bool AsmParser::enabledGenDwarfForAssembly() {
944   // Check whether the user specified -g.
945   if (!getContext().getGenDwarfForAssembly())
946     return false;
947   // If we haven't encountered any .file directives (which would imply that
948   // the assembler source was produced with debug info already) then emit one
949   // describing the assembler source file itself.
950   if (getContext().getGenDwarfFileNumber() == 0) {
951     // Use the first #line directive for this, if any. It's preprocessed, so
952     // there is no checksum, and of course no source directive.
953     if (!FirstCppHashFilename.empty())
954       getContext().setMCLineTableRootFile(
955           /*CUID=*/0, getContext().getCompilationDir(), FirstCppHashFilename,
956           /*Cksum=*/std::nullopt, /*Source=*/std::nullopt);
957     const MCDwarfFile &RootFile =
958         getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile();
959     getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective(
960         /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name,
961         RootFile.Checksum, RootFile.Source));
962   }
963   return true;
964 }
965 
Run(bool NoInitialTextSection,bool NoFinalize)966 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
967   LTODiscardSymbols.clear();
968 
969   // Create the initial section, if requested.
970   if (!NoInitialTextSection)
971     Out.initSections(false, getTargetParser().getSTI());
972 
973   // Prime the lexer.
974   Lex();
975 
976   HadError = false;
977   AsmCond StartingCondState = TheCondState;
978   SmallVector<AsmRewrite, 4> AsmStrRewrites;
979 
980   // If we are generating dwarf for assembly source files save the initial text
981   // section.  (Don't use enabledGenDwarfForAssembly() here, as we aren't
982   // emitting any actual debug info yet and haven't had a chance to parse any
983   // embedded .file directives.)
984   if (getContext().getGenDwarfForAssembly()) {
985     MCSection *Sec = getStreamer().getCurrentSectionOnly();
986     if (!Sec->getBeginSymbol()) {
987       MCSymbol *SectionStartSym = getContext().createTempSymbol();
988       getStreamer().emitLabel(SectionStartSym);
989       Sec->setBeginSymbol(SectionStartSym);
990     }
991     bool InsertResult = getContext().addGenDwarfSection(Sec);
992     assert(InsertResult && ".text section should not have debug info yet");
993     (void)InsertResult;
994   }
995 
996   getTargetParser().onBeginOfFile();
997 
998   // While we have input, parse each statement.
999   while (Lexer.isNot(AsmToken::Eof)) {
1000     ParseStatementInfo Info(&AsmStrRewrites);
1001     bool Parsed = parseStatement(Info, nullptr);
1002 
1003     // If we have a Lexer Error we are on an Error Token. Load in Lexer Error
1004     // for printing ErrMsg via Lex() only if no (presumably better) parser error
1005     // exists.
1006     if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) {
1007       Lex();
1008     }
1009 
1010     // parseStatement returned true so may need to emit an error.
1011     printPendingErrors();
1012 
1013     // Skipping to the next line if needed.
1014     if (Parsed && !getLexer().isAtStartOfStatement())
1015       eatToEndOfStatement();
1016   }
1017 
1018   getTargetParser().onEndOfFile();
1019   printPendingErrors();
1020 
1021   // All errors should have been emitted.
1022   assert(!hasPendingError() && "unexpected error from parseStatement");
1023 
1024   getTargetParser().flushPendingInstructions(getStreamer());
1025 
1026   if (TheCondState.TheCond != StartingCondState.TheCond ||
1027       TheCondState.Ignore != StartingCondState.Ignore)
1028     printError(getTok().getLoc(), "unmatched .ifs or .elses");
1029   // Check to see there are no empty DwarfFile slots.
1030   const auto &LineTables = getContext().getMCDwarfLineTables();
1031   if (!LineTables.empty()) {
1032     unsigned Index = 0;
1033     for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) {
1034       if (File.Name.empty() && Index != 0)
1035         printError(getTok().getLoc(), "unassigned file number: " +
1036                                           Twine(Index) +
1037                                           " for .file directives");
1038       ++Index;
1039     }
1040   }
1041 
1042   // Check to see that all assembler local symbols were actually defined.
1043   // Targets that don't do subsections via symbols may not want this, though,
1044   // so conservatively exclude them. Only do this if we're finalizing, though,
1045   // as otherwise we won't necessarilly have seen everything yet.
1046   if (!NoFinalize) {
1047     if (MAI.hasSubsectionsViaSymbols()) {
1048       for (const auto &TableEntry : getContext().getSymbols()) {
1049         MCSymbol *Sym = TableEntry.getValue().Symbol;
1050         // Variable symbols may not be marked as defined, so check those
1051         // explicitly. If we know it's a variable, we have a definition for
1052         // the purposes of this check.
1053         if (Sym && Sym->isTemporary() && !Sym->isVariable() &&
1054             !Sym->isDefined())
1055           // FIXME: We would really like to refer back to where the symbol was
1056           // first referenced for a source location. We need to add something
1057           // to track that. Currently, we just point to the end of the file.
1058           printError(getTok().getLoc(), "assembler local symbol '" +
1059                                             Sym->getName() + "' not defined");
1060       }
1061     }
1062 
1063     // Temporary symbols like the ones for directional jumps don't go in the
1064     // symbol table. They also need to be diagnosed in all (final) cases.
1065     for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) {
1066       if (std::get<2>(LocSym)->isUndefined()) {
1067         // Reset the state of any "# line file" directives we've seen to the
1068         // context as it was at the diagnostic site.
1069         CppHashInfo = std::get<1>(LocSym);
1070         printError(std::get<0>(LocSym), "directional label undefined");
1071       }
1072     }
1073   }
1074   // Finalize the output stream if there are no errors and if the client wants
1075   // us to.
1076   if (!HadError && !NoFinalize) {
1077     if (auto *TS = Out.getTargetStreamer())
1078       TS->emitConstantPools();
1079 
1080     Out.finish(Lexer.getLoc());
1081   }
1082 
1083   return HadError || getContext().hadError();
1084 }
1085 
checkForValidSection()1086 bool AsmParser::checkForValidSection() {
1087   if (!ParsingMSInlineAsm && !getStreamer().getCurrentFragment()) {
1088     Out.initSections(false, getTargetParser().getSTI());
1089     return Error(getTok().getLoc(),
1090                  "expected section directive before assembly directive");
1091   }
1092   return false;
1093 }
1094 
1095 /// Throw away the rest of the line for testing purposes.
eatToEndOfStatement()1096 void AsmParser::eatToEndOfStatement() {
1097   while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
1098     Lexer.Lex();
1099 
1100   // Eat EOL.
1101   if (Lexer.is(AsmToken::EndOfStatement))
1102     Lexer.Lex();
1103 }
1104 
parseStringToEndOfStatement()1105 StringRef AsmParser::parseStringToEndOfStatement() {
1106   const char *Start = getTok().getLoc().getPointer();
1107 
1108   while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
1109     Lexer.Lex();
1110 
1111   const char *End = getTok().getLoc().getPointer();
1112   return StringRef(Start, End - Start);
1113 }
1114 
parseStringToComma()1115 StringRef AsmParser::parseStringToComma() {
1116   const char *Start = getTok().getLoc().getPointer();
1117 
1118   while (Lexer.isNot(AsmToken::EndOfStatement) &&
1119          Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof))
1120     Lexer.Lex();
1121 
1122   const char *End = getTok().getLoc().getPointer();
1123   return StringRef(Start, End - Start);
1124 }
1125 
1126 /// Parse a paren expression and return it.
1127 /// NOTE: This assumes the leading '(' has already been consumed.
1128 ///
1129 /// parenexpr ::= expr)
1130 ///
parseParenExpr(const MCExpr * & Res,SMLoc & EndLoc)1131 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
1132   if (parseExpression(Res))
1133     return true;
1134   EndLoc = Lexer.getTok().getEndLoc();
1135   return parseRParen();
1136 }
1137 
1138 /// Parse a bracket expression and return it.
1139 /// NOTE: This assumes the leading '[' has already been consumed.
1140 ///
1141 /// bracketexpr ::= expr]
1142 ///
parseBracketExpr(const MCExpr * & Res,SMLoc & EndLoc)1143 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
1144   if (parseExpression(Res))
1145     return true;
1146   EndLoc = getTok().getEndLoc();
1147   if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression"))
1148     return true;
1149   return false;
1150 }
1151 
1152 /// Parse a primary expression and return it.
1153 ///  primaryexpr ::= (parenexpr
1154 ///  primaryexpr ::= symbol
1155 ///  primaryexpr ::= number
1156 ///  primaryexpr ::= '.'
1157 ///  primaryexpr ::= ~,+,- primaryexpr
parsePrimaryExpr(const MCExpr * & Res,SMLoc & EndLoc,AsmTypeInfo * TypeInfo)1158 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc,
1159                                  AsmTypeInfo *TypeInfo) {
1160   SMLoc FirstTokenLoc = getLexer().getLoc();
1161   AsmToken::TokenKind FirstTokenKind = Lexer.getKind();
1162   switch (FirstTokenKind) {
1163   default:
1164     return TokError("unknown token in expression");
1165   // If we have an error assume that we've already handled it.
1166   case AsmToken::Error:
1167     return true;
1168   case AsmToken::Exclaim:
1169     Lex(); // Eat the operator.
1170     if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1171       return true;
1172     Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc);
1173     return false;
1174   case AsmToken::Dollar:
1175   case AsmToken::Star:
1176   case AsmToken::At:
1177   case AsmToken::String:
1178   case AsmToken::Identifier: {
1179     StringRef Identifier;
1180     if (parseIdentifier(Identifier)) {
1181       // We may have failed but '$'|'*' may be a valid token in context of
1182       // the current PC.
1183       if (getTok().is(AsmToken::Dollar) || getTok().is(AsmToken::Star)) {
1184         bool ShouldGenerateTempSymbol = false;
1185         if ((getTok().is(AsmToken::Dollar) && MAI.getDollarIsPC()) ||
1186             (getTok().is(AsmToken::Star) && MAI.getStarIsPC()))
1187           ShouldGenerateTempSymbol = true;
1188 
1189         if (!ShouldGenerateTempSymbol)
1190           return Error(FirstTokenLoc, "invalid token in expression");
1191 
1192         // Eat the '$'|'*' token.
1193         Lex();
1194         // This is either a '$'|'*' reference, which references the current PC.
1195         // Emit a temporary label to the streamer and refer to it.
1196         MCSymbol *Sym = Ctx.createTempSymbol();
1197         Out.emitLabel(Sym);
1198         Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None,
1199                                       getContext());
1200         EndLoc = FirstTokenLoc;
1201         return false;
1202       }
1203     }
1204     // Parse symbol variant
1205     std::pair<StringRef, StringRef> Split;
1206     if (!MAI.useParensForSymbolVariant()) {
1207       if (FirstTokenKind == AsmToken::String) {
1208         if (Lexer.is(AsmToken::At)) {
1209           Lex(); // eat @
1210           SMLoc AtLoc = getLexer().getLoc();
1211           StringRef VName;
1212           if (parseIdentifier(VName))
1213             return Error(AtLoc, "expected symbol variant after '@'");
1214 
1215           Split = std::make_pair(Identifier, VName);
1216         }
1217       } else {
1218         Split = Identifier.split('@');
1219       }
1220     } else if (Lexer.is(AsmToken::LParen)) {
1221       Lex(); // eat '('.
1222       StringRef VName;
1223       parseIdentifier(VName);
1224       if (parseRParen())
1225         return true;
1226       Split = std::make_pair(Identifier, VName);
1227     }
1228 
1229     EndLoc = SMLoc::getFromPointer(Identifier.end());
1230 
1231     // This is a symbol reference.
1232     StringRef SymbolName = Identifier;
1233     if (SymbolName.empty())
1234       return Error(getLexer().getLoc(), "expected a symbol reference");
1235 
1236     MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
1237 
1238     // Lookup the symbol variant if used.
1239     if (!Split.second.empty()) {
1240       Variant = getTargetParser().getVariantKindForName(Split.second);
1241       if (Variant != MCSymbolRefExpr::VK_Invalid) {
1242         SymbolName = Split.first;
1243       } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) {
1244         Variant = MCSymbolRefExpr::VK_None;
1245       } else {
1246         return Error(SMLoc::getFromPointer(Split.second.begin()),
1247                      "invalid variant '" + Split.second + "'");
1248       }
1249     }
1250 
1251     MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName);
1252     if (!Sym)
1253       Sym = getContext().getOrCreateSymbol(
1254           MAI.shouldEmitLabelsInUpperCase() ? SymbolName.upper() : SymbolName);
1255 
1256     // If this is an absolute variable reference, substitute it now to preserve
1257     // semantics in the face of reassignment.
1258     if (Sym->isVariable()) {
1259       auto V = Sym->getVariableValue(/*SetUsed*/ false);
1260       bool DoInline = isa<MCConstantExpr>(V) && !Variant;
1261       if (auto TV = dyn_cast<MCTargetExpr>(V))
1262         DoInline = TV->inlineAssignedExpr();
1263       if (DoInline) {
1264         if (Variant)
1265           return Error(EndLoc, "unexpected modifier on variable reference");
1266         Res = Sym->getVariableValue(/*SetUsed*/ false);
1267         return false;
1268       }
1269     }
1270 
1271     // Otherwise create a symbol ref.
1272     Res = MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc);
1273     return false;
1274   }
1275   case AsmToken::BigNum:
1276     return TokError("literal value out of range for directive");
1277   case AsmToken::Integer: {
1278     SMLoc Loc = getTok().getLoc();
1279     int64_t IntVal = getTok().getIntVal();
1280     Res = MCConstantExpr::create(IntVal, getContext());
1281     EndLoc = Lexer.getTok().getEndLoc();
1282     Lex(); // Eat token.
1283     // Look for 'b' or 'f' following an Integer as a directional label
1284     if (Lexer.getKind() == AsmToken::Identifier) {
1285       StringRef IDVal = getTok().getString();
1286       // Lookup the symbol variant if used.
1287       std::pair<StringRef, StringRef> Split = IDVal.split('@');
1288       MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
1289       if (Split.first.size() != IDVal.size()) {
1290         Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
1291         if (Variant == MCSymbolRefExpr::VK_Invalid)
1292           return TokError("invalid variant '" + Split.second + "'");
1293         IDVal = Split.first;
1294       }
1295       if (IDVal == "f" || IDVal == "b") {
1296         MCSymbol *Sym =
1297             Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b");
1298         Res = MCSymbolRefExpr::create(Sym, Variant, getContext());
1299         if (IDVal == "b" && Sym->isUndefined())
1300           return Error(Loc, "directional label undefined");
1301         DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym));
1302         EndLoc = Lexer.getTok().getEndLoc();
1303         Lex(); // Eat identifier.
1304       }
1305     }
1306     return false;
1307   }
1308   case AsmToken::Real: {
1309     APFloat RealVal(APFloat::IEEEdouble(), getTok().getString());
1310     uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
1311     Res = MCConstantExpr::create(IntVal, getContext());
1312     EndLoc = Lexer.getTok().getEndLoc();
1313     Lex(); // Eat token.
1314     return false;
1315   }
1316   case AsmToken::Dot: {
1317     if (!MAI.getDotIsPC())
1318       return TokError("cannot use . as current PC");
1319 
1320     // This is a '.' reference, which references the current PC.  Emit a
1321     // temporary label to the streamer and refer to it.
1322     MCSymbol *Sym = Ctx.createTempSymbol();
1323     Out.emitLabel(Sym);
1324     Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext());
1325     EndLoc = Lexer.getTok().getEndLoc();
1326     Lex(); // Eat identifier.
1327     return false;
1328   }
1329   case AsmToken::LParen:
1330     Lex(); // Eat the '('.
1331     return parseParenExpr(Res, EndLoc);
1332   case AsmToken::LBrac:
1333     if (!PlatformParser->HasBracketExpressions())
1334       return TokError("brackets expression not supported on this target");
1335     Lex(); // Eat the '['.
1336     return parseBracketExpr(Res, EndLoc);
1337   case AsmToken::Minus:
1338     Lex(); // Eat the operator.
1339     if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1340       return true;
1341     Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc);
1342     return false;
1343   case AsmToken::Plus:
1344     Lex(); // Eat the operator.
1345     if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1346       return true;
1347     Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc);
1348     return false;
1349   case AsmToken::Tilde:
1350     Lex(); // Eat the operator.
1351     if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1352       return true;
1353     Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc);
1354     return false;
1355   // MIPS unary expression operators. The lexer won't generate these tokens if
1356   // MCAsmInfo::HasMipsExpressions is false for the target.
1357   case AsmToken::PercentCall16:
1358   case AsmToken::PercentCall_Hi:
1359   case AsmToken::PercentCall_Lo:
1360   case AsmToken::PercentDtprel_Hi:
1361   case AsmToken::PercentDtprel_Lo:
1362   case AsmToken::PercentGot:
1363   case AsmToken::PercentGot_Disp:
1364   case AsmToken::PercentGot_Hi:
1365   case AsmToken::PercentGot_Lo:
1366   case AsmToken::PercentGot_Ofst:
1367   case AsmToken::PercentGot_Page:
1368   case AsmToken::PercentGottprel:
1369   case AsmToken::PercentGp_Rel:
1370   case AsmToken::PercentHi:
1371   case AsmToken::PercentHigher:
1372   case AsmToken::PercentHighest:
1373   case AsmToken::PercentLo:
1374   case AsmToken::PercentNeg:
1375   case AsmToken::PercentPcrel_Hi:
1376   case AsmToken::PercentPcrel_Lo:
1377   case AsmToken::PercentTlsgd:
1378   case AsmToken::PercentTlsldm:
1379   case AsmToken::PercentTprel_Hi:
1380   case AsmToken::PercentTprel_Lo:
1381     Lex(); // Eat the operator.
1382     if (Lexer.isNot(AsmToken::LParen))
1383       return TokError("expected '(' after operator");
1384     Lex(); // Eat the operator.
1385     if (parseExpression(Res, EndLoc))
1386       return true;
1387     if (parseRParen())
1388       return true;
1389     Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx);
1390     return !Res;
1391   }
1392 }
1393 
parseExpression(const MCExpr * & Res)1394 bool AsmParser::parseExpression(const MCExpr *&Res) {
1395   SMLoc EndLoc;
1396   return parseExpression(Res, EndLoc);
1397 }
1398 
1399 const MCExpr *
applyModifierToExpr(const MCExpr * E,MCSymbolRefExpr::VariantKind Variant)1400 AsmParser::applyModifierToExpr(const MCExpr *E,
1401                                MCSymbolRefExpr::VariantKind Variant) {
1402   // Ask the target implementation about this expression first.
1403   const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx);
1404   if (NewE)
1405     return NewE;
1406   // Recurse over the given expression, rebuilding it to apply the given variant
1407   // if there is exactly one symbol.
1408   switch (E->getKind()) {
1409   case MCExpr::Target:
1410   case MCExpr::Constant:
1411     return nullptr;
1412 
1413   case MCExpr::SymbolRef: {
1414     const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
1415 
1416     if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
1417       TokError("invalid variant on expression '" + getTok().getIdentifier() +
1418                "' (already modified)");
1419       return E;
1420     }
1421 
1422     return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext());
1423   }
1424 
1425   case MCExpr::Unary: {
1426     const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
1427     const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant);
1428     if (!Sub)
1429       return nullptr;
1430     return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext());
1431   }
1432 
1433   case MCExpr::Binary: {
1434     const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
1435     const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant);
1436     const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant);
1437 
1438     if (!LHS && !RHS)
1439       return nullptr;
1440 
1441     if (!LHS)
1442       LHS = BE->getLHS();
1443     if (!RHS)
1444       RHS = BE->getRHS();
1445 
1446     return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext());
1447   }
1448   }
1449 
1450   llvm_unreachable("Invalid expression kind!");
1451 }
1452 
1453 /// This function checks if the next token is <string> type or arithmetic.
1454 /// string that begin with character '<' must end with character '>'.
1455 /// otherwise it is arithmetics.
1456 /// If the function returns a 'true' value,
1457 /// the End argument will be filled with the last location pointed to the '>'
1458 /// character.
1459 
1460 /// There is a gap between the AltMacro's documentation and the single quote
1461 /// implementation. GCC does not fully support this feature and so we will not
1462 /// support it.
1463 /// TODO: Adding single quote as a string.
isAngleBracketString(SMLoc & StrLoc,SMLoc & EndLoc)1464 static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) {
1465   assert((StrLoc.getPointer() != nullptr) &&
1466          "Argument to the function cannot be a NULL value");
1467   const char *CharPtr = StrLoc.getPointer();
1468   while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') &&
1469          (*CharPtr != '\0')) {
1470     if (*CharPtr == '!')
1471       CharPtr++;
1472     CharPtr++;
1473   }
1474   if (*CharPtr == '>') {
1475     EndLoc = StrLoc.getFromPointer(CharPtr + 1);
1476     return true;
1477   }
1478   return false;
1479 }
1480 
1481 /// creating a string without the escape characters '!'.
angleBracketString(StringRef AltMacroStr)1482 static std::string angleBracketString(StringRef AltMacroStr) {
1483   std::string Res;
1484   for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) {
1485     if (AltMacroStr[Pos] == '!')
1486       Pos++;
1487     Res += AltMacroStr[Pos];
1488   }
1489   return Res;
1490 }
1491 
1492 /// Parse an expression and return it.
1493 ///
1494 ///  expr ::= expr &&,|| expr               -> lowest.
1495 ///  expr ::= expr |,^,&,! expr
1496 ///  expr ::= expr ==,!=,<>,<,<=,>,>= expr
1497 ///  expr ::= expr <<,>> expr
1498 ///  expr ::= expr +,- expr
1499 ///  expr ::= expr *,/,% expr               -> highest.
1500 ///  expr ::= primaryexpr
1501 ///
parseExpression(const MCExpr * & Res,SMLoc & EndLoc)1502 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1503   // Parse the expression.
1504   Res = nullptr;
1505   if (getTargetParser().parsePrimaryExpr(Res, EndLoc) ||
1506       parseBinOpRHS(1, Res, EndLoc))
1507     return true;
1508 
1509   // As a special case, we support 'a op b @ modifier' by rewriting the
1510   // expression to include the modifier. This is inefficient, but in general we
1511   // expect users to use 'a@modifier op b'.
1512   if (parseOptionalToken(AsmToken::At)) {
1513     if (Lexer.isNot(AsmToken::Identifier))
1514       return TokError("unexpected symbol modifier following '@'");
1515 
1516     MCSymbolRefExpr::VariantKind Variant =
1517         MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
1518     if (Variant == MCSymbolRefExpr::VK_Invalid)
1519       return TokError("invalid variant '" + getTok().getIdentifier() + "'");
1520 
1521     const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant);
1522     if (!ModifiedRes) {
1523       return TokError("invalid modifier '" + getTok().getIdentifier() +
1524                       "' (no symbols present)");
1525     }
1526 
1527     Res = ModifiedRes;
1528     Lex();
1529   }
1530 
1531   // Try to constant fold it up front, if possible. Do not exploit
1532   // assembler here.
1533   int64_t Value;
1534   if (Res->evaluateAsAbsolute(Value))
1535     Res = MCConstantExpr::create(Value, getContext());
1536 
1537   return false;
1538 }
1539 
parseParenExpression(const MCExpr * & Res,SMLoc & EndLoc)1540 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1541   Res = nullptr;
1542   return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc);
1543 }
1544 
parseParenExprOfDepth(unsigned ParenDepth,const MCExpr * & Res,SMLoc & EndLoc)1545 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
1546                                       SMLoc &EndLoc) {
1547   if (parseParenExpr(Res, EndLoc))
1548     return true;
1549 
1550   for (; ParenDepth > 0; --ParenDepth) {
1551     if (parseBinOpRHS(1, Res, EndLoc))
1552       return true;
1553 
1554     // We don't Lex() the last RParen.
1555     // This is the same behavior as parseParenExpression().
1556     if (ParenDepth - 1 > 0) {
1557       EndLoc = getTok().getEndLoc();
1558       if (parseRParen())
1559         return true;
1560     }
1561   }
1562   return false;
1563 }
1564 
parseAbsoluteExpression(int64_t & Res)1565 bool AsmParser::parseAbsoluteExpression(int64_t &Res) {
1566   const MCExpr *Expr;
1567 
1568   SMLoc StartLoc = Lexer.getLoc();
1569   if (parseExpression(Expr))
1570     return true;
1571 
1572   if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr()))
1573     return Error(StartLoc, "expected absolute expression");
1574 
1575   return false;
1576 }
1577 
getDarwinBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1578 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K,
1579                                          MCBinaryExpr::Opcode &Kind,
1580                                          bool ShouldUseLogicalShr) {
1581   switch (K) {
1582   default:
1583     return 0; // not a binop.
1584 
1585   // Lowest Precedence: &&, ||
1586   case AsmToken::AmpAmp:
1587     Kind = MCBinaryExpr::LAnd;
1588     return 1;
1589   case AsmToken::PipePipe:
1590     Kind = MCBinaryExpr::LOr;
1591     return 1;
1592 
1593   // Low Precedence: |, &, ^
1594   case AsmToken::Pipe:
1595     Kind = MCBinaryExpr::Or;
1596     return 2;
1597   case AsmToken::Caret:
1598     Kind = MCBinaryExpr::Xor;
1599     return 2;
1600   case AsmToken::Amp:
1601     Kind = MCBinaryExpr::And;
1602     return 2;
1603 
1604   // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
1605   case AsmToken::EqualEqual:
1606     Kind = MCBinaryExpr::EQ;
1607     return 3;
1608   case AsmToken::ExclaimEqual:
1609   case AsmToken::LessGreater:
1610     Kind = MCBinaryExpr::NE;
1611     return 3;
1612   case AsmToken::Less:
1613     Kind = MCBinaryExpr::LT;
1614     return 3;
1615   case AsmToken::LessEqual:
1616     Kind = MCBinaryExpr::LTE;
1617     return 3;
1618   case AsmToken::Greater:
1619     Kind = MCBinaryExpr::GT;
1620     return 3;
1621   case AsmToken::GreaterEqual:
1622     Kind = MCBinaryExpr::GTE;
1623     return 3;
1624 
1625   // Intermediate Precedence: <<, >>
1626   case AsmToken::LessLess:
1627     Kind = MCBinaryExpr::Shl;
1628     return 4;
1629   case AsmToken::GreaterGreater:
1630     Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1631     return 4;
1632 
1633   // High Intermediate Precedence: +, -
1634   case AsmToken::Plus:
1635     Kind = MCBinaryExpr::Add;
1636     return 5;
1637   case AsmToken::Minus:
1638     Kind = MCBinaryExpr::Sub;
1639     return 5;
1640 
1641   // Highest Precedence: *, /, %
1642   case AsmToken::Star:
1643     Kind = MCBinaryExpr::Mul;
1644     return 6;
1645   case AsmToken::Slash:
1646     Kind = MCBinaryExpr::Div;
1647     return 6;
1648   case AsmToken::Percent:
1649     Kind = MCBinaryExpr::Mod;
1650     return 6;
1651   }
1652 }
1653 
getGNUBinOpPrecedence(const MCAsmInfo & MAI,AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1654 static unsigned getGNUBinOpPrecedence(const MCAsmInfo &MAI,
1655                                       AsmToken::TokenKind K,
1656                                       MCBinaryExpr::Opcode &Kind,
1657                                       bool ShouldUseLogicalShr) {
1658   switch (K) {
1659   default:
1660     return 0; // not a binop.
1661 
1662   // Lowest Precedence: &&, ||
1663   case AsmToken::AmpAmp:
1664     Kind = MCBinaryExpr::LAnd;
1665     return 2;
1666   case AsmToken::PipePipe:
1667     Kind = MCBinaryExpr::LOr;
1668     return 1;
1669 
1670   // Low Precedence: ==, !=, <>, <, <=, >, >=
1671   case AsmToken::EqualEqual:
1672     Kind = MCBinaryExpr::EQ;
1673     return 3;
1674   case AsmToken::ExclaimEqual:
1675   case AsmToken::LessGreater:
1676     Kind = MCBinaryExpr::NE;
1677     return 3;
1678   case AsmToken::Less:
1679     Kind = MCBinaryExpr::LT;
1680     return 3;
1681   case AsmToken::LessEqual:
1682     Kind = MCBinaryExpr::LTE;
1683     return 3;
1684   case AsmToken::Greater:
1685     Kind = MCBinaryExpr::GT;
1686     return 3;
1687   case AsmToken::GreaterEqual:
1688     Kind = MCBinaryExpr::GTE;
1689     return 3;
1690 
1691   // Low Intermediate Precedence: +, -
1692   case AsmToken::Plus:
1693     Kind = MCBinaryExpr::Add;
1694     return 4;
1695   case AsmToken::Minus:
1696     Kind = MCBinaryExpr::Sub;
1697     return 4;
1698 
1699   // High Intermediate Precedence: |, !, &, ^
1700   //
1701   case AsmToken::Pipe:
1702     Kind = MCBinaryExpr::Or;
1703     return 5;
1704   case AsmToken::Exclaim:
1705     // Hack to support ARM compatible aliases (implied 'sp' operand in 'srs*'
1706     // instructions like 'srsda #31!') and not parse ! as an infix operator.
1707     if (MAI.getCommentString() == "@")
1708       return 0;
1709     Kind = MCBinaryExpr::OrNot;
1710     return 5;
1711   case AsmToken::Caret:
1712     Kind = MCBinaryExpr::Xor;
1713     return 5;
1714   case AsmToken::Amp:
1715     Kind = MCBinaryExpr::And;
1716     return 5;
1717 
1718   // Highest Precedence: *, /, %, <<, >>
1719   case AsmToken::Star:
1720     Kind = MCBinaryExpr::Mul;
1721     return 6;
1722   case AsmToken::Slash:
1723     Kind = MCBinaryExpr::Div;
1724     return 6;
1725   case AsmToken::Percent:
1726     Kind = MCBinaryExpr::Mod;
1727     return 6;
1728   case AsmToken::LessLess:
1729     Kind = MCBinaryExpr::Shl;
1730     return 6;
1731   case AsmToken::GreaterGreater:
1732     Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1733     return 6;
1734   }
1735 }
1736 
getBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind)1737 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K,
1738                                        MCBinaryExpr::Opcode &Kind) {
1739   bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr();
1740   return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr)
1741                   : getGNUBinOpPrecedence(MAI, K, Kind, ShouldUseLogicalShr);
1742 }
1743 
1744 /// Parse all binary operators with precedence >= 'Precedence'.
1745 /// Res contains the LHS of the expression on input.
parseBinOpRHS(unsigned Precedence,const MCExpr * & Res,SMLoc & EndLoc)1746 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1747                               SMLoc &EndLoc) {
1748   SMLoc StartLoc = Lexer.getLoc();
1749   while (true) {
1750     MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1751     unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1752 
1753     // If the next token is lower precedence than we are allowed to eat, return
1754     // successfully with what we ate already.
1755     if (TokPrec < Precedence)
1756       return false;
1757 
1758     Lex();
1759 
1760     // Eat the next primary expression.
1761     const MCExpr *RHS;
1762     if (getTargetParser().parsePrimaryExpr(RHS, EndLoc))
1763       return true;
1764 
1765     // If BinOp binds less tightly with RHS than the operator after RHS, let
1766     // the pending operator take RHS as its LHS.
1767     MCBinaryExpr::Opcode Dummy;
1768     unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1769     if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc))
1770       return true;
1771 
1772     // Merge LHS and RHS according to operator.
1773     Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc);
1774   }
1775 }
1776 
1777 /// ParseStatement:
1778 ///   ::= EndOfStatement
1779 ///   ::= Label* Directive ...Operands... EndOfStatement
1780 ///   ::= Label* Identifier OperandList* EndOfStatement
parseStatement(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)1781 bool AsmParser::parseStatement(ParseStatementInfo &Info,
1782                                MCAsmParserSemaCallback *SI) {
1783   assert(!hasPendingError() && "parseStatement started with pending error");
1784   // Eat initial spaces and comments
1785   while (Lexer.is(AsmToken::Space))
1786     Lex();
1787   if (Lexer.is(AsmToken::EndOfStatement)) {
1788     // if this is a line comment we can drop it safely
1789     if (getTok().getString().empty() || getTok().getString().front() == '\r' ||
1790         getTok().getString().front() == '\n')
1791       Out.addBlankLine();
1792     Lex();
1793     return false;
1794   }
1795   // Statements always start with an identifier.
1796   AsmToken ID = getTok();
1797   SMLoc IDLoc = ID.getLoc();
1798   StringRef IDVal;
1799   int64_t LocalLabelVal = -1;
1800   StartTokLoc = ID.getLoc();
1801   if (Lexer.is(AsmToken::HashDirective))
1802     return parseCppHashLineFilenameComment(IDLoc,
1803                                            !isInsideMacroInstantiation());
1804 
1805   // Allow an integer followed by a ':' as a directional local label.
1806   if (Lexer.is(AsmToken::Integer)) {
1807     LocalLabelVal = getTok().getIntVal();
1808     if (LocalLabelVal < 0) {
1809       if (!TheCondState.Ignore) {
1810         Lex(); // always eat a token
1811         return Error(IDLoc, "unexpected token at start of statement");
1812       }
1813       IDVal = "";
1814     } else {
1815       IDVal = getTok().getString();
1816       Lex(); // Consume the integer token to be used as an identifier token.
1817       if (Lexer.getKind() != AsmToken::Colon) {
1818         if (!TheCondState.Ignore) {
1819           Lex(); // always eat a token
1820           return Error(IDLoc, "unexpected token at start of statement");
1821         }
1822       }
1823     }
1824   } else if (Lexer.is(AsmToken::Dot)) {
1825     // Treat '.' as a valid identifier in this context.
1826     Lex();
1827     IDVal = ".";
1828   } else if (Lexer.is(AsmToken::LCurly)) {
1829     // Treat '{' as a valid identifier in this context.
1830     Lex();
1831     IDVal = "{";
1832 
1833   } else if (Lexer.is(AsmToken::RCurly)) {
1834     // Treat '}' as a valid identifier in this context.
1835     Lex();
1836     IDVal = "}";
1837   } else if (Lexer.is(AsmToken::Star) &&
1838              getTargetParser().starIsStartOfStatement()) {
1839     // Accept '*' as a valid start of statement.
1840     Lex();
1841     IDVal = "*";
1842   } else if (parseIdentifier(IDVal)) {
1843     if (!TheCondState.Ignore) {
1844       Lex(); // always eat a token
1845       return Error(IDLoc, "unexpected token at start of statement");
1846     }
1847     IDVal = "";
1848   }
1849 
1850   // Handle conditional assembly here before checking for skipping.  We
1851   // have to do this so that .endif isn't skipped in a ".if 0" block for
1852   // example.
1853   StringMap<DirectiveKind>::const_iterator DirKindIt =
1854       DirectiveKindMap.find(IDVal.lower());
1855   DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end())
1856                               ? DK_NO_DIRECTIVE
1857                               : DirKindIt->getValue();
1858   switch (DirKind) {
1859   default:
1860     break;
1861   case DK_IF:
1862   case DK_IFEQ:
1863   case DK_IFGE:
1864   case DK_IFGT:
1865   case DK_IFLE:
1866   case DK_IFLT:
1867   case DK_IFNE:
1868     return parseDirectiveIf(IDLoc, DirKind);
1869   case DK_IFB:
1870     return parseDirectiveIfb(IDLoc, true);
1871   case DK_IFNB:
1872     return parseDirectiveIfb(IDLoc, false);
1873   case DK_IFC:
1874     return parseDirectiveIfc(IDLoc, true);
1875   case DK_IFEQS:
1876     return parseDirectiveIfeqs(IDLoc, true);
1877   case DK_IFNC:
1878     return parseDirectiveIfc(IDLoc, false);
1879   case DK_IFNES:
1880     return parseDirectiveIfeqs(IDLoc, false);
1881   case DK_IFDEF:
1882     return parseDirectiveIfdef(IDLoc, true);
1883   case DK_IFNDEF:
1884   case DK_IFNOTDEF:
1885     return parseDirectiveIfdef(IDLoc, false);
1886   case DK_ELSEIF:
1887     return parseDirectiveElseIf(IDLoc);
1888   case DK_ELSE:
1889     return parseDirectiveElse(IDLoc);
1890   case DK_ENDIF:
1891     return parseDirectiveEndIf(IDLoc);
1892   }
1893 
1894   // Ignore the statement if in the middle of inactive conditional
1895   // (e.g. ".if 0").
1896   if (TheCondState.Ignore) {
1897     eatToEndOfStatement();
1898     return false;
1899   }
1900 
1901   // FIXME: Recurse on local labels?
1902 
1903   // Check for a label.
1904   //   ::= identifier ':'
1905   //   ::= number ':'
1906   if (Lexer.is(AsmToken::Colon) && getTargetParser().isLabel(ID)) {
1907     if (checkForValidSection())
1908       return true;
1909 
1910     Lex(); // Consume the ':'.
1911 
1912     // Diagnose attempt to use '.' as a label.
1913     if (IDVal == ".")
1914       return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1915 
1916     // Diagnose attempt to use a variable as a label.
1917     //
1918     // FIXME: Diagnostics. Note the location of the definition as a label.
1919     // FIXME: This doesn't diagnose assignment to a symbol which has been
1920     // implicitly marked as external.
1921     MCSymbol *Sym;
1922     if (LocalLabelVal == -1) {
1923       if (ParsingMSInlineAsm && SI) {
1924         StringRef RewrittenLabel =
1925             SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true);
1926         assert(!RewrittenLabel.empty() &&
1927                "We should have an internal name here.");
1928         Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(),
1929                                        RewrittenLabel);
1930         IDVal = RewrittenLabel;
1931       }
1932       Sym = getContext().getOrCreateSymbol(IDVal);
1933     } else
1934       Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal);
1935     // End of Labels should be treated as end of line for lexing
1936     // purposes but that information is not available to the Lexer who
1937     // does not understand Labels. This may cause us to see a Hash
1938     // here instead of a preprocessor line comment.
1939     if (getTok().is(AsmToken::Hash)) {
1940       StringRef CommentStr = parseStringToEndOfStatement();
1941       Lexer.Lex();
1942       Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr));
1943     }
1944 
1945     // Consume any end of statement token, if present, to avoid spurious
1946     // addBlankLine calls().
1947     if (getTok().is(AsmToken::EndOfStatement)) {
1948       Lex();
1949     }
1950 
1951     if (MAI.hasSubsectionsViaSymbols() && CFIStartProcLoc &&
1952         Sym->isExternal() && !cast<MCSymbolMachO>(Sym)->isAltEntry())
1953       return Error(StartTokLoc, "non-private labels cannot appear between "
1954                                 ".cfi_startproc / .cfi_endproc pairs") &&
1955              Error(*CFIStartProcLoc, "previous .cfi_startproc was here");
1956 
1957     if (discardLTOSymbol(IDVal))
1958       return false;
1959 
1960     getTargetParser().doBeforeLabelEmit(Sym, IDLoc);
1961 
1962     // Emit the label.
1963     if (!getTargetParser().isParsingMSInlineAsm())
1964       Out.emitLabel(Sym, IDLoc);
1965 
1966     // If we are generating dwarf for assembly source files then gather the
1967     // info to make a dwarf label entry for this label if needed.
1968     if (enabledGenDwarfForAssembly())
1969       MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1970                                  IDLoc);
1971 
1972     getTargetParser().onLabelParsed(Sym);
1973 
1974     return false;
1975   }
1976 
1977   // Check for an assignment statement.
1978   //   ::= identifier '='
1979   if (Lexer.is(AsmToken::Equal) && getTargetParser().equalIsAsmAssignment()) {
1980     Lex();
1981     return parseAssignment(IDVal, AssignmentKind::Equal);
1982   }
1983 
1984   // If macros are enabled, check to see if this is a macro instantiation.
1985   if (areMacrosEnabled())
1986     if (MCAsmMacro *M = getContext().lookupMacro(IDVal))
1987       return handleMacroEntry(M, IDLoc);
1988 
1989   // Otherwise, we have a normal instruction or directive.
1990 
1991   // Directives start with "."
1992   if (IDVal.starts_with(".") && IDVal != ".") {
1993     // There are several entities interested in parsing directives:
1994     //
1995     // 1. The target-specific assembly parser. Some directives are target
1996     //    specific or may potentially behave differently on certain targets.
1997     // 2. Asm parser extensions. For example, platform-specific parsers
1998     //    (like the ELF parser) register themselves as extensions.
1999     // 3. The generic directive parser implemented by this class. These are
2000     //    all the directives that behave in a target and platform independent
2001     //    manner, or at least have a default behavior that's shared between
2002     //    all targets and platforms.
2003 
2004     getTargetParser().flushPendingInstructions(getStreamer());
2005 
2006     ParseStatus TPDirectiveReturn = getTargetParser().parseDirective(ID);
2007     assert(TPDirectiveReturn.isFailure() == hasPendingError() &&
2008            "Should only return Failure iff there was an error");
2009     if (TPDirectiveReturn.isFailure())
2010       return true;
2011     if (TPDirectiveReturn.isSuccess())
2012       return false;
2013 
2014     // Next, check the extension directive map to see if any extension has
2015     // registered itself to parse this directive.
2016     std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
2017         ExtensionDirectiveMap.lookup(IDVal);
2018     if (Handler.first)
2019       return (*Handler.second)(Handler.first, IDVal, IDLoc);
2020 
2021     // Finally, if no one else is interested in this directive, it must be
2022     // generic and familiar to this class.
2023     switch (DirKind) {
2024     default:
2025       break;
2026     case DK_SET:
2027     case DK_EQU:
2028       return parseDirectiveSet(IDVal, AssignmentKind::Set);
2029     case DK_EQUIV:
2030       return parseDirectiveSet(IDVal, AssignmentKind::Equiv);
2031     case DK_LTO_SET_CONDITIONAL:
2032       return parseDirectiveSet(IDVal, AssignmentKind::LTOSetConditional);
2033     case DK_ASCII:
2034       return parseDirectiveAscii(IDVal, false);
2035     case DK_ASCIZ:
2036     case DK_STRING:
2037       return parseDirectiveAscii(IDVal, true);
2038     case DK_BYTE:
2039     case DK_DC_B:
2040       return parseDirectiveValue(IDVal, 1);
2041     case DK_DC:
2042     case DK_DC_W:
2043     case DK_SHORT:
2044     case DK_VALUE:
2045     case DK_2BYTE:
2046       return parseDirectiveValue(IDVal, 2);
2047     case DK_LONG:
2048     case DK_INT:
2049     case DK_4BYTE:
2050     case DK_DC_L:
2051       return parseDirectiveValue(IDVal, 4);
2052     case DK_QUAD:
2053     case DK_8BYTE:
2054       return parseDirectiveValue(IDVal, 8);
2055     case DK_DC_A:
2056       return parseDirectiveValue(
2057           IDVal, getContext().getAsmInfo()->getCodePointerSize());
2058     case DK_OCTA:
2059       return parseDirectiveOctaValue(IDVal);
2060     case DK_SINGLE:
2061     case DK_FLOAT:
2062     case DK_DC_S:
2063       return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle());
2064     case DK_DOUBLE:
2065     case DK_DC_D:
2066       return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble());
2067     case DK_ALIGN: {
2068       bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
2069       return parseDirectiveAlign(IsPow2, /*ExprSize=*/1);
2070     }
2071     case DK_ALIGN32: {
2072       bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
2073       return parseDirectiveAlign(IsPow2, /*ExprSize=*/4);
2074     }
2075     case DK_BALIGN:
2076       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
2077     case DK_BALIGNW:
2078       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
2079     case DK_BALIGNL:
2080       return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
2081     case DK_P2ALIGN:
2082       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
2083     case DK_P2ALIGNW:
2084       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
2085     case DK_P2ALIGNL:
2086       return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
2087     case DK_ORG:
2088       return parseDirectiveOrg();
2089     case DK_FILL:
2090       return parseDirectiveFill();
2091     case DK_ZERO:
2092       return parseDirectiveZero();
2093     case DK_EXTERN:
2094       eatToEndOfStatement(); // .extern is the default, ignore it.
2095       return false;
2096     case DK_GLOBL:
2097     case DK_GLOBAL:
2098       return parseDirectiveSymbolAttribute(MCSA_Global);
2099     case DK_LAZY_REFERENCE:
2100       return parseDirectiveSymbolAttribute(MCSA_LazyReference);
2101     case DK_NO_DEAD_STRIP:
2102       return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
2103     case DK_SYMBOL_RESOLVER:
2104       return parseDirectiveSymbolAttribute(MCSA_SymbolResolver);
2105     case DK_PRIVATE_EXTERN:
2106       return parseDirectiveSymbolAttribute(MCSA_PrivateExtern);
2107     case DK_REFERENCE:
2108       return parseDirectiveSymbolAttribute(MCSA_Reference);
2109     case DK_WEAK_DEFINITION:
2110       return parseDirectiveSymbolAttribute(MCSA_WeakDefinition);
2111     case DK_WEAK_REFERENCE:
2112       return parseDirectiveSymbolAttribute(MCSA_WeakReference);
2113     case DK_WEAK_DEF_CAN_BE_HIDDEN:
2114       return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
2115     case DK_COLD:
2116       return parseDirectiveSymbolAttribute(MCSA_Cold);
2117     case DK_COMM:
2118     case DK_COMMON:
2119       return parseDirectiveComm(/*IsLocal=*/false);
2120     case DK_LCOMM:
2121       return parseDirectiveComm(/*IsLocal=*/true);
2122     case DK_ABORT:
2123       return parseDirectiveAbort(IDLoc);
2124     case DK_INCLUDE:
2125       return parseDirectiveInclude();
2126     case DK_INCBIN:
2127       return parseDirectiveIncbin();
2128     case DK_CODE16:
2129     case DK_CODE16GCC:
2130       return TokError(Twine(IDVal) +
2131                       " not currently supported for this target");
2132     case DK_REPT:
2133       return parseDirectiveRept(IDLoc, IDVal);
2134     case DK_IRP:
2135       return parseDirectiveIrp(IDLoc);
2136     case DK_IRPC:
2137       return parseDirectiveIrpc(IDLoc);
2138     case DK_ENDR:
2139       return parseDirectiveEndr(IDLoc);
2140     case DK_BUNDLE_ALIGN_MODE:
2141       return parseDirectiveBundleAlignMode();
2142     case DK_BUNDLE_LOCK:
2143       return parseDirectiveBundleLock();
2144     case DK_BUNDLE_UNLOCK:
2145       return parseDirectiveBundleUnlock();
2146     case DK_SLEB128:
2147       return parseDirectiveLEB128(true);
2148     case DK_ULEB128:
2149       return parseDirectiveLEB128(false);
2150     case DK_SPACE:
2151     case DK_SKIP:
2152       return parseDirectiveSpace(IDVal);
2153     case DK_FILE:
2154       return parseDirectiveFile(IDLoc);
2155     case DK_LINE:
2156       return parseDirectiveLine();
2157     case DK_LOC:
2158       return parseDirectiveLoc();
2159     case DK_STABS:
2160       return parseDirectiveStabs();
2161     case DK_CV_FILE:
2162       return parseDirectiveCVFile();
2163     case DK_CV_FUNC_ID:
2164       return parseDirectiveCVFuncId();
2165     case DK_CV_INLINE_SITE_ID:
2166       return parseDirectiveCVInlineSiteId();
2167     case DK_CV_LOC:
2168       return parseDirectiveCVLoc();
2169     case DK_CV_LINETABLE:
2170       return parseDirectiveCVLinetable();
2171     case DK_CV_INLINE_LINETABLE:
2172       return parseDirectiveCVInlineLinetable();
2173     case DK_CV_DEF_RANGE:
2174       return parseDirectiveCVDefRange();
2175     case DK_CV_STRING:
2176       return parseDirectiveCVString();
2177     case DK_CV_STRINGTABLE:
2178       return parseDirectiveCVStringTable();
2179     case DK_CV_FILECHECKSUMS:
2180       return parseDirectiveCVFileChecksums();
2181     case DK_CV_FILECHECKSUM_OFFSET:
2182       return parseDirectiveCVFileChecksumOffset();
2183     case DK_CV_FPO_DATA:
2184       return parseDirectiveCVFPOData();
2185     case DK_CFI_SECTIONS:
2186       return parseDirectiveCFISections();
2187     case DK_CFI_STARTPROC:
2188       return parseDirectiveCFIStartProc();
2189     case DK_CFI_ENDPROC:
2190       return parseDirectiveCFIEndProc();
2191     case DK_CFI_DEF_CFA:
2192       return parseDirectiveCFIDefCfa(IDLoc);
2193     case DK_CFI_DEF_CFA_OFFSET:
2194       return parseDirectiveCFIDefCfaOffset(IDLoc);
2195     case DK_CFI_ADJUST_CFA_OFFSET:
2196       return parseDirectiveCFIAdjustCfaOffset(IDLoc);
2197     case DK_CFI_DEF_CFA_REGISTER:
2198       return parseDirectiveCFIDefCfaRegister(IDLoc);
2199     case DK_CFI_LLVM_DEF_ASPACE_CFA:
2200       return parseDirectiveCFILLVMDefAspaceCfa(IDLoc);
2201     case DK_CFI_OFFSET:
2202       return parseDirectiveCFIOffset(IDLoc);
2203     case DK_CFI_REL_OFFSET:
2204       return parseDirectiveCFIRelOffset(IDLoc);
2205     case DK_CFI_PERSONALITY:
2206       return parseDirectiveCFIPersonalityOrLsda(true);
2207     case DK_CFI_LSDA:
2208       return parseDirectiveCFIPersonalityOrLsda(false);
2209     case DK_CFI_REMEMBER_STATE:
2210       return parseDirectiveCFIRememberState(IDLoc);
2211     case DK_CFI_RESTORE_STATE:
2212       return parseDirectiveCFIRestoreState(IDLoc);
2213     case DK_CFI_SAME_VALUE:
2214       return parseDirectiveCFISameValue(IDLoc);
2215     case DK_CFI_RESTORE:
2216       return parseDirectiveCFIRestore(IDLoc);
2217     case DK_CFI_ESCAPE:
2218       return parseDirectiveCFIEscape(IDLoc);
2219     case DK_CFI_RETURN_COLUMN:
2220       return parseDirectiveCFIReturnColumn(IDLoc);
2221     case DK_CFI_SIGNAL_FRAME:
2222       return parseDirectiveCFISignalFrame(IDLoc);
2223     case DK_CFI_UNDEFINED:
2224       return parseDirectiveCFIUndefined(IDLoc);
2225     case DK_CFI_REGISTER:
2226       return parseDirectiveCFIRegister(IDLoc);
2227     case DK_CFI_WINDOW_SAVE:
2228       return parseDirectiveCFIWindowSave(IDLoc);
2229     case DK_CFI_LABEL:
2230       return parseDirectiveCFILabel(IDLoc);
2231     case DK_MACROS_ON:
2232     case DK_MACROS_OFF:
2233       return parseDirectiveMacrosOnOff(IDVal);
2234     case DK_MACRO:
2235       return parseDirectiveMacro(IDLoc);
2236     case DK_ALTMACRO:
2237     case DK_NOALTMACRO:
2238       return parseDirectiveAltmacro(IDVal);
2239     case DK_EXITM:
2240       return parseDirectiveExitMacro(IDVal);
2241     case DK_ENDM:
2242     case DK_ENDMACRO:
2243       return parseDirectiveEndMacro(IDVal);
2244     case DK_PURGEM:
2245       return parseDirectivePurgeMacro(IDLoc);
2246     case DK_END:
2247       return parseDirectiveEnd(IDLoc);
2248     case DK_ERR:
2249       return parseDirectiveError(IDLoc, false);
2250     case DK_ERROR:
2251       return parseDirectiveError(IDLoc, true);
2252     case DK_WARNING:
2253       return parseDirectiveWarning(IDLoc);
2254     case DK_RELOC:
2255       return parseDirectiveReloc(IDLoc);
2256     case DK_DCB:
2257     case DK_DCB_W:
2258       return parseDirectiveDCB(IDVal, 2);
2259     case DK_DCB_B:
2260       return parseDirectiveDCB(IDVal, 1);
2261     case DK_DCB_D:
2262       return parseDirectiveRealDCB(IDVal, APFloat::IEEEdouble());
2263     case DK_DCB_L:
2264       return parseDirectiveDCB(IDVal, 4);
2265     case DK_DCB_S:
2266       return parseDirectiveRealDCB(IDVal, APFloat::IEEEsingle());
2267     case DK_DC_X:
2268     case DK_DCB_X:
2269       return TokError(Twine(IDVal) +
2270                       " not currently supported for this target");
2271     case DK_DS:
2272     case DK_DS_W:
2273       return parseDirectiveDS(IDVal, 2);
2274     case DK_DS_B:
2275       return parseDirectiveDS(IDVal, 1);
2276     case DK_DS_D:
2277       return parseDirectiveDS(IDVal, 8);
2278     case DK_DS_L:
2279     case DK_DS_S:
2280       return parseDirectiveDS(IDVal, 4);
2281     case DK_DS_P:
2282     case DK_DS_X:
2283       return parseDirectiveDS(IDVal, 12);
2284     case DK_PRINT:
2285       return parseDirectivePrint(IDLoc);
2286     case DK_ADDRSIG:
2287       return parseDirectiveAddrsig();
2288     case DK_ADDRSIG_SYM:
2289       return parseDirectiveAddrsigSym();
2290     case DK_PSEUDO_PROBE:
2291       return parseDirectivePseudoProbe();
2292     case DK_LTO_DISCARD:
2293       return parseDirectiveLTODiscard();
2294     case DK_MEMTAG:
2295       return parseDirectiveSymbolAttribute(MCSA_Memtag);
2296     }
2297 
2298     return Error(IDLoc, "unknown directive");
2299   }
2300 
2301   // __asm _emit or __asm __emit
2302   if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" ||
2303                              IDVal == "_EMIT" || IDVal == "__EMIT"))
2304     return parseDirectiveMSEmit(IDLoc, Info, IDVal.size());
2305 
2306   // __asm align
2307   if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN"))
2308     return parseDirectiveMSAlign(IDLoc, Info);
2309 
2310   if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN"))
2311     Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4);
2312   if (checkForValidSection())
2313     return true;
2314 
2315   return parseAndMatchAndEmitTargetInstruction(Info, IDVal, ID, IDLoc);
2316 }
2317 
parseAndMatchAndEmitTargetInstruction(ParseStatementInfo & Info,StringRef IDVal,AsmToken ID,SMLoc IDLoc)2318 bool AsmParser::parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info,
2319                                                       StringRef IDVal,
2320                                                       AsmToken ID,
2321                                                       SMLoc IDLoc) {
2322   // Canonicalize the opcode to lower case.
2323   std::string OpcodeStr = IDVal.lower();
2324   ParseInstructionInfo IInfo(Info.AsmRewrites);
2325   bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID,
2326                                                           Info.ParsedOperands);
2327   Info.ParseError = ParseHadError;
2328 
2329   // Dump the parsed representation, if requested.
2330   if (getShowParsedOperands()) {
2331     SmallString<256> Str;
2332     raw_svector_ostream OS(Str);
2333     OS << "parsed instruction: [";
2334     for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
2335       if (i != 0)
2336         OS << ", ";
2337       Info.ParsedOperands[i]->print(OS);
2338     }
2339     OS << "]";
2340 
2341     printMessage(IDLoc, SourceMgr::DK_Note, OS.str());
2342   }
2343 
2344   // Fail even if ParseInstruction erroneously returns false.
2345   if (hasPendingError() || ParseHadError)
2346     return true;
2347 
2348   // If we are generating dwarf for the current section then generate a .loc
2349   // directive for the instruction.
2350   if (!ParseHadError && enabledGenDwarfForAssembly() &&
2351       getContext().getGenDwarfSectionSyms().count(
2352           getStreamer().getCurrentSectionOnly())) {
2353     unsigned Line;
2354     if (ActiveMacros.empty())
2355       Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
2356     else
2357       Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc,
2358                                    ActiveMacros.front()->ExitBuffer);
2359 
2360     // If we previously parsed a cpp hash file line comment then make sure the
2361     // current Dwarf File is for the CppHashFilename if not then emit the
2362     // Dwarf File table for it and adjust the line number for the .loc.
2363     if (!CppHashInfo.Filename.empty()) {
2364       unsigned FileNumber = getStreamer().emitDwarfFileDirective(
2365           0, StringRef(), CppHashInfo.Filename);
2366       getContext().setGenDwarfFileNumber(FileNumber);
2367 
2368       unsigned CppHashLocLineNo =
2369         SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf);
2370       Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo);
2371     }
2372 
2373     getStreamer().emitDwarfLocDirective(
2374         getContext().getGenDwarfFileNumber(), Line, 0,
2375         DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0,
2376         StringRef());
2377   }
2378 
2379   // If parsing succeeded, match the instruction.
2380   if (!ParseHadError) {
2381     uint64_t ErrorInfo;
2382     if (getTargetParser().MatchAndEmitInstruction(
2383             IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo,
2384             getTargetParser().isParsingMSInlineAsm()))
2385       return true;
2386   }
2387   return false;
2388 }
2389 
2390 // Parse and erase curly braces marking block start/end
2391 bool
parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> & AsmStrRewrites)2392 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) {
2393   // Identify curly brace marking block start/end
2394   if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly))
2395     return false;
2396 
2397   SMLoc StartLoc = Lexer.getLoc();
2398   Lex(); // Eat the brace
2399   if (Lexer.is(AsmToken::EndOfStatement))
2400     Lex(); // Eat EndOfStatement following the brace
2401 
2402   // Erase the block start/end brace from the output asm string
2403   AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() -
2404                                                   StartLoc.getPointer());
2405   return true;
2406 }
2407 
2408 /// parseCppHashLineFilenameComment as this:
2409 ///   ::= # number "filename"
parseCppHashLineFilenameComment(SMLoc L,bool SaveLocInfo)2410 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo) {
2411   Lex(); // Eat the hash token.
2412   // Lexer only ever emits HashDirective if it fully formed if it's
2413   // done the checking already so this is an internal error.
2414   assert(getTok().is(AsmToken::Integer) &&
2415          "Lexing Cpp line comment: Expected Integer");
2416   int64_t LineNumber = getTok().getIntVal();
2417   Lex();
2418   assert(getTok().is(AsmToken::String) &&
2419          "Lexing Cpp line comment: Expected String");
2420   StringRef Filename = getTok().getString();
2421   Lex();
2422 
2423   if (!SaveLocInfo)
2424     return false;
2425 
2426   // Get rid of the enclosing quotes.
2427   Filename = Filename.substr(1, Filename.size() - 2);
2428 
2429   // Save the SMLoc, Filename and LineNumber for later use by diagnostics
2430   // and possibly DWARF file info.
2431   CppHashInfo.Loc = L;
2432   CppHashInfo.Filename = Filename;
2433   CppHashInfo.LineNumber = LineNumber;
2434   CppHashInfo.Buf = CurBuffer;
2435   if (FirstCppHashFilename.empty())
2436     FirstCppHashFilename = Filename;
2437   return false;
2438 }
2439 
2440 /// will use the last parsed cpp hash line filename comment
2441 /// for the Filename and LineNo if any in the diagnostic.
DiagHandler(const SMDiagnostic & Diag,void * Context)2442 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
2443   auto *Parser = static_cast<AsmParser *>(Context);
2444   raw_ostream &OS = errs();
2445 
2446   const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
2447   SMLoc DiagLoc = Diag.getLoc();
2448   unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
2449   unsigned CppHashBuf =
2450       Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc);
2451 
2452   // Like SourceMgr::printMessage() we need to print the include stack if any
2453   // before printing the message.
2454   unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
2455   if (!Parser->SavedDiagHandler && DiagCurBuffer &&
2456       DiagCurBuffer != DiagSrcMgr.getMainFileID()) {
2457     SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
2458     DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
2459   }
2460 
2461   // If we have not parsed a cpp hash line filename comment or the source
2462   // manager changed or buffer changed (like in a nested include) then just
2463   // print the normal diagnostic using its Filename and LineNo.
2464   if (!Parser->CppHashInfo.LineNumber || DiagBuf != CppHashBuf) {
2465     if (Parser->SavedDiagHandler)
2466       Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
2467     else
2468       Parser->getContext().diagnose(Diag);
2469     return;
2470   }
2471 
2472   // Use the CppHashFilename and calculate a line number based on the
2473   // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc
2474   // for the diagnostic.
2475   const std::string &Filename = std::string(Parser->CppHashInfo.Filename);
2476 
2477   int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
2478   int CppHashLocLineNo =
2479       Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf);
2480   int LineNo =
2481       Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo);
2482 
2483   SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo,
2484                        Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(),
2485                        Diag.getLineContents(), Diag.getRanges());
2486 
2487   if (Parser->SavedDiagHandler)
2488     Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
2489   else
2490     Parser->getContext().diagnose(NewDiag);
2491 }
2492 
2493 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
2494 // difference being that that function accepts '@' as part of identifiers and
2495 // we can't do that. AsmLexer.cpp should probably be changed to handle
2496 // '@' as a special case when needed.
isIdentifierChar(char c)2497 static bool isIdentifierChar(char c) {
2498   return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' ||
2499          c == '.';
2500 }
2501 
expandMacro(raw_svector_ostream & OS,MCAsmMacro & Macro,ArrayRef<MCAsmMacroParameter> Parameters,ArrayRef<MCAsmMacroArgument> A,bool EnableAtPseudoVariable)2502 bool AsmParser::expandMacro(raw_svector_ostream &OS, MCAsmMacro &Macro,
2503                             ArrayRef<MCAsmMacroParameter> Parameters,
2504                             ArrayRef<MCAsmMacroArgument> A,
2505                             bool EnableAtPseudoVariable) {
2506   unsigned NParameters = Parameters.size();
2507   auto expandArg = [&](unsigned Index) {
2508     bool HasVararg = NParameters ? Parameters.back().Vararg : false;
2509     bool VarargParameter = HasVararg && Index == (NParameters - 1);
2510     for (const AsmToken &Token : A[Index])
2511       // For altmacro mode, you can write '%expr'.
2512       // The prefix '%' evaluates the expression 'expr'
2513       // and uses the result as a string (e.g. replace %(1+2) with the
2514       // string "3").
2515       // Here, we identify the integer token which is the result of the
2516       // absolute expression evaluation and replace it with its string
2517       // representation.
2518       if (AltMacroMode && Token.getString().front() == '%' &&
2519           Token.is(AsmToken::Integer))
2520         // Emit an integer value to the buffer.
2521         OS << Token.getIntVal();
2522       // Only Token that was validated as a string and begins with '<'
2523       // is considered altMacroString!!!
2524       else if (AltMacroMode && Token.getString().front() == '<' &&
2525                Token.is(AsmToken::String)) {
2526         OS << angleBracketString(Token.getStringContents());
2527       }
2528       // We expect no quotes around the string's contents when
2529       // parsing for varargs.
2530       else if (Token.isNot(AsmToken::String) || VarargParameter)
2531         OS << Token.getString();
2532       else
2533         OS << Token.getStringContents();
2534   };
2535 
2536   // A macro without parameters is handled differently on Darwin:
2537   // gas accepts no arguments and does no substitutions
2538   StringRef Body = Macro.Body;
2539   size_t I = 0, End = Body.size();
2540   while (I != End) {
2541     if (Body[I] == '\\' && I + 1 != End) {
2542       // Check for \@ and \+ pseudo variables.
2543       if (EnableAtPseudoVariable && Body[I + 1] == '@') {
2544         OS << NumOfMacroInstantiations;
2545         I += 2;
2546         continue;
2547       }
2548       if (Body[I + 1] == '+') {
2549         OS << Macro.Count;
2550         I += 2;
2551         continue;
2552       }
2553       if (Body[I + 1] == '(' && Body[I + 2] == ')') {
2554         I += 3;
2555         continue;
2556       }
2557 
2558       size_t Pos = ++I;
2559       while (I != End && isIdentifierChar(Body[I]))
2560         ++I;
2561       StringRef Argument(Body.data() + Pos, I - Pos);
2562       if (AltMacroMode && I != End && Body[I] == '&')
2563         ++I;
2564       unsigned Index = 0;
2565       for (; Index < NParameters; ++Index)
2566         if (Parameters[Index].Name == Argument)
2567           break;
2568       if (Index == NParameters)
2569         OS << '\\' << Argument;
2570       else
2571         expandArg(Index);
2572       continue;
2573     }
2574 
2575     // In Darwin mode, $ is used for macro expansion, not considered an
2576     // identifier char.
2577     if (Body[I] == '$' && I + 1 != End && IsDarwin && !NParameters) {
2578       // This macro has no parameters, look for $0, $1, etc.
2579       switch (Body[I + 1]) {
2580       // $$ => $
2581       case '$':
2582         OS << '$';
2583         I += 2;
2584         continue;
2585       // $n => number of arguments
2586       case 'n':
2587         OS << A.size();
2588         I += 2;
2589         continue;
2590       default: {
2591         if (!isDigit(Body[I + 1]))
2592           break;
2593         // $[0-9] => argument
2594         // Missing arguments are ignored.
2595         unsigned Index = Body[I + 1] - '0';
2596         if (Index < A.size())
2597           for (const AsmToken &Token : A[Index])
2598             OS << Token.getString();
2599         I += 2;
2600         continue;
2601       }
2602       }
2603     }
2604 
2605     if (!isIdentifierChar(Body[I]) || IsDarwin) {
2606       OS << Body[I++];
2607       continue;
2608     }
2609 
2610     const size_t Start = I;
2611     while (++I && isIdentifierChar(Body[I])) {
2612     }
2613     StringRef Token(Body.data() + Start, I - Start);
2614     if (AltMacroMode) {
2615       unsigned Index = 0;
2616       for (; Index != NParameters; ++Index)
2617         if (Parameters[Index].Name == Token)
2618           break;
2619       if (Index != NParameters) {
2620         expandArg(Index);
2621         if (I != End && Body[I] == '&')
2622           ++I;
2623         continue;
2624       }
2625     }
2626     OS << Token;
2627   }
2628 
2629   ++Macro.Count;
2630   return false;
2631 }
2632 
isOperator(AsmToken::TokenKind kind)2633 static bool isOperator(AsmToken::TokenKind kind) {
2634   switch (kind) {
2635   default:
2636     return false;
2637   case AsmToken::Plus:
2638   case AsmToken::Minus:
2639   case AsmToken::Tilde:
2640   case AsmToken::Slash:
2641   case AsmToken::Star:
2642   case AsmToken::Dot:
2643   case AsmToken::Equal:
2644   case AsmToken::EqualEqual:
2645   case AsmToken::Pipe:
2646   case AsmToken::PipePipe:
2647   case AsmToken::Caret:
2648   case AsmToken::Amp:
2649   case AsmToken::AmpAmp:
2650   case AsmToken::Exclaim:
2651   case AsmToken::ExclaimEqual:
2652   case AsmToken::Less:
2653   case AsmToken::LessEqual:
2654   case AsmToken::LessLess:
2655   case AsmToken::LessGreater:
2656   case AsmToken::Greater:
2657   case AsmToken::GreaterEqual:
2658   case AsmToken::GreaterGreater:
2659     return true;
2660   }
2661 }
2662 
2663 namespace {
2664 
2665 class AsmLexerSkipSpaceRAII {
2666 public:
AsmLexerSkipSpaceRAII(AsmLexer & Lexer,bool SkipSpace)2667   AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) {
2668     Lexer.setSkipSpace(SkipSpace);
2669   }
2670 
~AsmLexerSkipSpaceRAII()2671   ~AsmLexerSkipSpaceRAII() {
2672     Lexer.setSkipSpace(true);
2673   }
2674 
2675 private:
2676   AsmLexer &Lexer;
2677 };
2678 
2679 } // end anonymous namespace
2680 
parseMacroArgument(MCAsmMacroArgument & MA,bool Vararg)2681 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) {
2682 
2683   if (Vararg) {
2684     if (Lexer.isNot(AsmToken::EndOfStatement)) {
2685       StringRef Str = parseStringToEndOfStatement();
2686       MA.emplace_back(AsmToken::String, Str);
2687     }
2688     return false;
2689   }
2690 
2691   unsigned ParenLevel = 0;
2692 
2693   // Darwin doesn't use spaces to delmit arguments.
2694   AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin);
2695 
2696   bool SpaceEaten;
2697 
2698   while (true) {
2699     SpaceEaten = false;
2700     if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
2701       return TokError("unexpected token in macro instantiation");
2702 
2703     if (ParenLevel == 0) {
2704 
2705       if (Lexer.is(AsmToken::Comma))
2706         break;
2707 
2708       if (parseOptionalToken(AsmToken::Space))
2709         SpaceEaten = true;
2710 
2711       // Spaces can delimit parameters, but could also be part an expression.
2712       // If the token after a space is an operator, add the token and the next
2713       // one into this argument
2714       if (!IsDarwin) {
2715         if (isOperator(Lexer.getKind())) {
2716           MA.push_back(getTok());
2717           Lexer.Lex();
2718 
2719           // Whitespace after an operator can be ignored.
2720           parseOptionalToken(AsmToken::Space);
2721           continue;
2722         }
2723       }
2724       if (SpaceEaten)
2725         break;
2726     }
2727 
2728     // handleMacroEntry relies on not advancing the lexer here
2729     // to be able to fill in the remaining default parameter values
2730     if (Lexer.is(AsmToken::EndOfStatement))
2731       break;
2732 
2733     // Adjust the current parentheses level.
2734     if (Lexer.is(AsmToken::LParen))
2735       ++ParenLevel;
2736     else if (Lexer.is(AsmToken::RParen) && ParenLevel)
2737       --ParenLevel;
2738 
2739     // Append the token to the current argument list.
2740     MA.push_back(getTok());
2741     Lexer.Lex();
2742   }
2743 
2744   if (ParenLevel != 0)
2745     return TokError("unbalanced parentheses in macro argument");
2746   return false;
2747 }
2748 
2749 // Parse the macro instantiation arguments.
parseMacroArguments(const MCAsmMacro * M,MCAsmMacroArguments & A)2750 bool AsmParser::parseMacroArguments(const MCAsmMacro *M,
2751                                     MCAsmMacroArguments &A) {
2752   const unsigned NParameters = M ? M->Parameters.size() : 0;
2753   bool NamedParametersFound = false;
2754   SmallVector<SMLoc, 4> FALocs;
2755 
2756   A.resize(NParameters);
2757   FALocs.resize(NParameters);
2758 
2759   // Parse two kinds of macro invocations:
2760   // - macros defined without any parameters accept an arbitrary number of them
2761   // - macros defined with parameters accept at most that many of them
2762   bool HasVararg = NParameters ? M->Parameters.back().Vararg : false;
2763   for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
2764        ++Parameter) {
2765     SMLoc IDLoc = Lexer.getLoc();
2766     MCAsmMacroParameter FA;
2767 
2768     if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) {
2769       if (parseIdentifier(FA.Name))
2770         return Error(IDLoc, "invalid argument identifier for formal argument");
2771 
2772       if (Lexer.isNot(AsmToken::Equal))
2773         return TokError("expected '=' after formal parameter identifier");
2774 
2775       Lex();
2776 
2777       NamedParametersFound = true;
2778     }
2779     bool Vararg = HasVararg && Parameter == (NParameters - 1);
2780 
2781     if (NamedParametersFound && FA.Name.empty())
2782       return Error(IDLoc, "cannot mix positional and keyword arguments");
2783 
2784     SMLoc StrLoc = Lexer.getLoc();
2785     SMLoc EndLoc;
2786     if (AltMacroMode && Lexer.is(AsmToken::Percent)) {
2787       const MCExpr *AbsoluteExp;
2788       int64_t Value;
2789       /// Eat '%'
2790       Lex();
2791       if (parseExpression(AbsoluteExp, EndLoc))
2792         return false;
2793       if (!AbsoluteExp->evaluateAsAbsolute(Value,
2794                                            getStreamer().getAssemblerPtr()))
2795         return Error(StrLoc, "expected absolute expression");
2796       const char *StrChar = StrLoc.getPointer();
2797       const char *EndChar = EndLoc.getPointer();
2798       AsmToken newToken(AsmToken::Integer,
2799                         StringRef(StrChar, EndChar - StrChar), Value);
2800       FA.Value.push_back(newToken);
2801     } else if (AltMacroMode && Lexer.is(AsmToken::Less) &&
2802                isAngleBracketString(StrLoc, EndLoc)) {
2803       const char *StrChar = StrLoc.getPointer();
2804       const char *EndChar = EndLoc.getPointer();
2805       jumpToLoc(EndLoc, CurBuffer);
2806       /// Eat from '<' to '>'
2807       Lex();
2808       AsmToken newToken(AsmToken::String,
2809                         StringRef(StrChar, EndChar - StrChar));
2810       FA.Value.push_back(newToken);
2811     } else if(parseMacroArgument(FA.Value, Vararg))
2812       return true;
2813 
2814     unsigned PI = Parameter;
2815     if (!FA.Name.empty()) {
2816       unsigned FAI = 0;
2817       for (FAI = 0; FAI < NParameters; ++FAI)
2818         if (M->Parameters[FAI].Name == FA.Name)
2819           break;
2820 
2821       if (FAI >= NParameters) {
2822         assert(M && "expected macro to be defined");
2823         return Error(IDLoc, "parameter named '" + FA.Name +
2824                                 "' does not exist for macro '" + M->Name + "'");
2825       }
2826       PI = FAI;
2827     }
2828 
2829     if (!FA.Value.empty()) {
2830       if (A.size() <= PI)
2831         A.resize(PI + 1);
2832       A[PI] = FA.Value;
2833 
2834       if (FALocs.size() <= PI)
2835         FALocs.resize(PI + 1);
2836 
2837       FALocs[PI] = Lexer.getLoc();
2838     }
2839 
2840     // At the end of the statement, fill in remaining arguments that have
2841     // default values. If there aren't any, then the next argument is
2842     // required but missing
2843     if (Lexer.is(AsmToken::EndOfStatement)) {
2844       bool Failure = false;
2845       for (unsigned FAI = 0; FAI < NParameters; ++FAI) {
2846         if (A[FAI].empty()) {
2847           if (M->Parameters[FAI].Required) {
2848             Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(),
2849                   "missing value for required parameter "
2850                   "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'");
2851             Failure = true;
2852           }
2853 
2854           if (!M->Parameters[FAI].Value.empty())
2855             A[FAI] = M->Parameters[FAI].Value;
2856         }
2857       }
2858       return Failure;
2859     }
2860 
2861     parseOptionalToken(AsmToken::Comma);
2862   }
2863 
2864   return TokError("too many positional arguments");
2865 }
2866 
handleMacroEntry(MCAsmMacro * M,SMLoc NameLoc)2867 bool AsmParser::handleMacroEntry(MCAsmMacro *M, SMLoc NameLoc) {
2868   // Arbitrarily limit macro nesting depth (default matches 'as'). We can
2869   // eliminate this, although we should protect against infinite loops.
2870   unsigned MaxNestingDepth = AsmMacroMaxNestingDepth;
2871   if (ActiveMacros.size() == MaxNestingDepth) {
2872     std::ostringstream MaxNestingDepthError;
2873     MaxNestingDepthError << "macros cannot be nested more than "
2874                          << MaxNestingDepth << " levels deep."
2875                          << " Use -asm-macro-max-nesting-depth to increase "
2876                             "this limit.";
2877     return TokError(MaxNestingDepthError.str());
2878   }
2879 
2880   MCAsmMacroArguments A;
2881   if (parseMacroArguments(M, A))
2882     return true;
2883 
2884   // Macro instantiation is lexical, unfortunately. We construct a new buffer
2885   // to hold the macro body with substitutions.
2886   SmallString<256> Buf;
2887   raw_svector_ostream OS(Buf);
2888 
2889   if ((!IsDarwin || M->Parameters.size()) && M->Parameters.size() != A.size())
2890     return Error(getTok().getLoc(), "Wrong number of arguments");
2891   if (expandMacro(OS, *M, M->Parameters, A, true))
2892     return true;
2893 
2894   // We include the .endmacro in the buffer as our cue to exit the macro
2895   // instantiation.
2896   OS << ".endmacro\n";
2897 
2898   std::unique_ptr<MemoryBuffer> Instantiation =
2899       MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
2900 
2901   // Create the macro instantiation object and add to the current macro
2902   // instantiation stack.
2903   MacroInstantiation *MI = new MacroInstantiation{
2904       NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()};
2905   ActiveMacros.push_back(MI);
2906 
2907   ++NumOfMacroInstantiations;
2908 
2909   // Jump to the macro instantiation and prime the lexer.
2910   CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
2911   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
2912   Lex();
2913 
2914   return false;
2915 }
2916 
handleMacroExit()2917 void AsmParser::handleMacroExit() {
2918   // Jump to the EndOfStatement we should return to, and consume it.
2919   jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer);
2920   Lex();
2921   // If .endm/.endr is followed by \n instead of a comment, consume it so that
2922   // we don't print an excess \n.
2923   if (getTok().is(AsmToken::EndOfStatement))
2924     Lex();
2925 
2926   // Pop the instantiation entry.
2927   delete ActiveMacros.back();
2928   ActiveMacros.pop_back();
2929 }
2930 
parseAssignment(StringRef Name,AssignmentKind Kind)2931 bool AsmParser::parseAssignment(StringRef Name, AssignmentKind Kind) {
2932   MCSymbol *Sym;
2933   const MCExpr *Value;
2934   SMLoc ExprLoc = getTok().getLoc();
2935   bool AllowRedef =
2936       Kind == AssignmentKind::Set || Kind == AssignmentKind::Equal;
2937   if (MCParserUtils::parseAssignmentExpression(Name, AllowRedef, *this, Sym,
2938                                                Value))
2939     return true;
2940 
2941   if (!Sym) {
2942     // In the case where we parse an expression starting with a '.', we will
2943     // not generate an error, nor will we create a symbol.  In this case we
2944     // should just return out.
2945     return false;
2946   }
2947 
2948   if (discardLTOSymbol(Name))
2949     return false;
2950 
2951   // Do the assignment.
2952   switch (Kind) {
2953   case AssignmentKind::Equal:
2954     Out.emitAssignment(Sym, Value);
2955     break;
2956   case AssignmentKind::Set:
2957   case AssignmentKind::Equiv:
2958     Out.emitAssignment(Sym, Value);
2959     Out.emitSymbolAttribute(Sym, MCSA_NoDeadStrip);
2960     break;
2961   case AssignmentKind::LTOSetConditional:
2962     if (Value->getKind() != MCExpr::SymbolRef)
2963       return Error(ExprLoc, "expected identifier");
2964 
2965     Out.emitConditionalAssignment(Sym, Value);
2966     break;
2967   }
2968 
2969   return false;
2970 }
2971 
2972 /// parseIdentifier:
2973 ///   ::= identifier
2974 ///   ::= string
parseIdentifier(StringRef & Res)2975 bool AsmParser::parseIdentifier(StringRef &Res) {
2976   // The assembler has relaxed rules for accepting identifiers, in particular we
2977   // allow things like '.globl $foo' and '.def @feat.00', which would normally be
2978   // separate tokens. At this level, we have already lexed so we cannot (currently)
2979   // handle this as a context dependent token, instead we detect adjacent tokens
2980   // and return the combined identifier.
2981   if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) {
2982     SMLoc PrefixLoc = getLexer().getLoc();
2983 
2984     // Consume the prefix character, and check for a following identifier.
2985 
2986     AsmToken Buf[1];
2987     Lexer.peekTokens(Buf, false);
2988 
2989     if (Buf[0].isNot(AsmToken::Identifier) && Buf[0].isNot(AsmToken::Integer))
2990       return true;
2991 
2992     // We have a '$' or '@' followed by an identifier or integer token, make
2993     // sure they are adjacent.
2994     if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer())
2995       return true;
2996 
2997     // eat $ or @
2998     Lexer.Lex(); // Lexer's Lex guarantees consecutive token.
2999     // Construct the joined identifier and consume the token.
3000     Res = StringRef(PrefixLoc.getPointer(), getTok().getString().size() + 1);
3001     Lex(); // Parser Lex to maintain invariants.
3002     return false;
3003   }
3004 
3005   if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String))
3006     return true;
3007 
3008   Res = getTok().getIdentifier();
3009 
3010   Lex(); // Consume the identifier token.
3011 
3012   return false;
3013 }
3014 
3015 /// parseDirectiveSet:
3016 ///   ::= .equ identifier ',' expression
3017 ///   ::= .equiv identifier ',' expression
3018 ///   ::= .set identifier ',' expression
3019 ///   ::= .lto_set_conditional identifier ',' expression
parseDirectiveSet(StringRef IDVal,AssignmentKind Kind)3020 bool AsmParser::parseDirectiveSet(StringRef IDVal, AssignmentKind Kind) {
3021   StringRef Name;
3022   if (check(parseIdentifier(Name), "expected identifier") || parseComma() ||
3023       parseAssignment(Name, Kind))
3024     return true;
3025   return false;
3026 }
3027 
parseEscapedString(std::string & Data)3028 bool AsmParser::parseEscapedString(std::string &Data) {
3029   if (check(getTok().isNot(AsmToken::String), "expected string"))
3030     return true;
3031 
3032   Data = "";
3033   StringRef Str = getTok().getStringContents();
3034   for (unsigned i = 0, e = Str.size(); i != e; ++i) {
3035     if (Str[i] != '\\') {
3036       if (Str[i] == '\n') {
3037         SMLoc NewlineLoc = SMLoc::getFromPointer(Str.data() + i);
3038         if (Warning(NewlineLoc, "unterminated string; newline inserted"))
3039           return true;
3040       }
3041       Data += Str[i];
3042       continue;
3043     }
3044 
3045     // Recognize escaped characters. Note that this escape semantics currently
3046     // loosely follows Darwin 'as'.
3047     ++i;
3048     if (i == e)
3049       return TokError("unexpected backslash at end of string");
3050 
3051     // Recognize hex sequences similarly to GNU 'as'.
3052     if (Str[i] == 'x' || Str[i] == 'X') {
3053       size_t length = Str.size();
3054       if (i + 1 >= length || !isHexDigit(Str[i + 1]))
3055         return TokError("invalid hexadecimal escape sequence");
3056 
3057       // Consume hex characters. GNU 'as' reads all hexadecimal characters and
3058       // then truncates to the lower 16 bits. Seems reasonable.
3059       unsigned Value = 0;
3060       while (i + 1 < length && isHexDigit(Str[i + 1]))
3061         Value = Value * 16 + hexDigitValue(Str[++i]);
3062 
3063       Data += (unsigned char)(Value & 0xFF);
3064       continue;
3065     }
3066 
3067     // Recognize octal sequences.
3068     if ((unsigned)(Str[i] - '0') <= 7) {
3069       // Consume up to three octal characters.
3070       unsigned Value = Str[i] - '0';
3071 
3072       if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
3073         ++i;
3074         Value = Value * 8 + (Str[i] - '0');
3075 
3076         if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
3077           ++i;
3078           Value = Value * 8 + (Str[i] - '0');
3079         }
3080       }
3081 
3082       if (Value > 255)
3083         return TokError("invalid octal escape sequence (out of range)");
3084 
3085       Data += (unsigned char)Value;
3086       continue;
3087     }
3088 
3089     // Otherwise recognize individual escapes.
3090     switch (Str[i]) {
3091     default:
3092       // Just reject invalid escape sequences for now.
3093       return TokError("invalid escape sequence (unrecognized character)");
3094 
3095     case 'b': Data += '\b'; break;
3096     case 'f': Data += '\f'; break;
3097     case 'n': Data += '\n'; break;
3098     case 'r': Data += '\r'; break;
3099     case 't': Data += '\t'; break;
3100     case '"': Data += '"'; break;
3101     case '\\': Data += '\\'; break;
3102     }
3103   }
3104 
3105   Lex();
3106   return false;
3107 }
3108 
parseAngleBracketString(std::string & Data)3109 bool AsmParser::parseAngleBracketString(std::string &Data) {
3110   SMLoc EndLoc, StartLoc = getTok().getLoc();
3111   if (isAngleBracketString(StartLoc, EndLoc)) {
3112     const char *StartChar = StartLoc.getPointer() + 1;
3113     const char *EndChar = EndLoc.getPointer() - 1;
3114     jumpToLoc(EndLoc, CurBuffer);
3115     /// Eat from '<' to '>'
3116     Lex();
3117 
3118     Data = angleBracketString(StringRef(StartChar, EndChar - StartChar));
3119     return false;
3120   }
3121   return true;
3122 }
3123 
3124 /// parseDirectiveAscii:
3125 //    ::= .ascii [ "string"+ ( , "string"+ )* ]
3126 ///   ::= ( .asciz | .string ) [ "string" ( , "string" )* ]
parseDirectiveAscii(StringRef IDVal,bool ZeroTerminated)3127 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
3128   auto parseOp = [&]() -> bool {
3129     std::string Data;
3130     if (checkForValidSection())
3131       return true;
3132     // Only support spaces as separators for .ascii directive for now. See the
3133     // discusssion at https://reviews.llvm.org/D91460 for more details.
3134     do {
3135       if (parseEscapedString(Data))
3136         return true;
3137       getStreamer().emitBytes(Data);
3138     } while (!ZeroTerminated && getTok().is(AsmToken::String));
3139     if (ZeroTerminated)
3140       getStreamer().emitBytes(StringRef("\0", 1));
3141     return false;
3142   };
3143 
3144   return parseMany(parseOp);
3145 }
3146 
3147 /// parseDirectiveReloc
3148 ///  ::= .reloc expression , identifier [ , expression ]
parseDirectiveReloc(SMLoc DirectiveLoc)3149 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) {
3150   const MCExpr *Offset;
3151   const MCExpr *Expr = nullptr;
3152   SMLoc OffsetLoc = Lexer.getTok().getLoc();
3153 
3154   if (parseExpression(Offset))
3155     return true;
3156   if (parseComma() ||
3157       check(getTok().isNot(AsmToken::Identifier), "expected relocation name"))
3158     return true;
3159 
3160   SMLoc NameLoc = Lexer.getTok().getLoc();
3161   StringRef Name = Lexer.getTok().getIdentifier();
3162   Lex();
3163 
3164   if (Lexer.is(AsmToken::Comma)) {
3165     Lex();
3166     SMLoc ExprLoc = Lexer.getLoc();
3167     if (parseExpression(Expr))
3168       return true;
3169 
3170     MCValue Value;
3171     if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr))
3172       return Error(ExprLoc, "expression must be relocatable");
3173   }
3174 
3175   if (parseEOL())
3176     return true;
3177 
3178   const MCTargetAsmParser &MCT = getTargetParser();
3179   const MCSubtargetInfo &STI = MCT.getSTI();
3180   if (std::optional<std::pair<bool, std::string>> Err =
3181           getStreamer().emitRelocDirective(*Offset, Name, Expr, DirectiveLoc,
3182                                            STI))
3183     return Error(Err->first ? NameLoc : OffsetLoc, Err->second);
3184 
3185   return false;
3186 }
3187 
3188 /// parseDirectiveValue
3189 ///  ::= (.byte | .short | ... ) [ expression (, expression)* ]
parseDirectiveValue(StringRef IDVal,unsigned Size)3190 bool AsmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) {
3191   auto parseOp = [&]() -> bool {
3192     const MCExpr *Value;
3193     SMLoc ExprLoc = getLexer().getLoc();
3194     if (checkForValidSection() || parseExpression(Value))
3195       return true;
3196     // Special case constant expressions to match code generator.
3197     if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3198       assert(Size <= 8 && "Invalid size");
3199       uint64_t IntValue = MCE->getValue();
3200       if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
3201         return Error(ExprLoc, "out of range literal value");
3202       getStreamer().emitIntValue(IntValue, Size);
3203     } else
3204       getStreamer().emitValue(Value, Size, ExprLoc);
3205     return false;
3206   };
3207 
3208   return parseMany(parseOp);
3209 }
3210 
parseHexOcta(AsmParser & Asm,uint64_t & hi,uint64_t & lo)3211 static bool parseHexOcta(AsmParser &Asm, uint64_t &hi, uint64_t &lo) {
3212   if (Asm.getTok().isNot(AsmToken::Integer) &&
3213       Asm.getTok().isNot(AsmToken::BigNum))
3214     return Asm.TokError("unknown token in expression");
3215   SMLoc ExprLoc = Asm.getTok().getLoc();
3216   APInt IntValue = Asm.getTok().getAPIntVal();
3217   Asm.Lex();
3218   if (!IntValue.isIntN(128))
3219     return Asm.Error(ExprLoc, "out of range literal value");
3220   if (!IntValue.isIntN(64)) {
3221     hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue();
3222     lo = IntValue.getLoBits(64).getZExtValue();
3223   } else {
3224     hi = 0;
3225     lo = IntValue.getZExtValue();
3226   }
3227   return false;
3228 }
3229 
3230 /// ParseDirectiveOctaValue
3231 ///  ::= .octa [ hexconstant (, hexconstant)* ]
3232 
parseDirectiveOctaValue(StringRef IDVal)3233 bool AsmParser::parseDirectiveOctaValue(StringRef IDVal) {
3234   auto parseOp = [&]() -> bool {
3235     if (checkForValidSection())
3236       return true;
3237     uint64_t hi, lo;
3238     if (parseHexOcta(*this, hi, lo))
3239       return true;
3240     if (MAI.isLittleEndian()) {
3241       getStreamer().emitInt64(lo);
3242       getStreamer().emitInt64(hi);
3243     } else {
3244       getStreamer().emitInt64(hi);
3245       getStreamer().emitInt64(lo);
3246     }
3247     return false;
3248   };
3249 
3250   return parseMany(parseOp);
3251 }
3252 
parseRealValue(const fltSemantics & Semantics,APInt & Res)3253 bool AsmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) {
3254   // We don't truly support arithmetic on floating point expressions, so we
3255   // have to manually parse unary prefixes.
3256   bool IsNeg = false;
3257   if (getLexer().is(AsmToken::Minus)) {
3258     Lexer.Lex();
3259     IsNeg = true;
3260   } else if (getLexer().is(AsmToken::Plus))
3261     Lexer.Lex();
3262 
3263   if (Lexer.is(AsmToken::Error))
3264     return TokError(Lexer.getErr());
3265   if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) &&
3266       Lexer.isNot(AsmToken::Identifier))
3267     return TokError("unexpected token in directive");
3268 
3269   // Convert to an APFloat.
3270   APFloat Value(Semantics);
3271   StringRef IDVal = getTok().getString();
3272   if (getLexer().is(AsmToken::Identifier)) {
3273     if (!IDVal.compare_insensitive("infinity") ||
3274         !IDVal.compare_insensitive("inf"))
3275       Value = APFloat::getInf(Semantics);
3276     else if (!IDVal.compare_insensitive("nan"))
3277       Value = APFloat::getNaN(Semantics, false, ~0);
3278     else
3279       return TokError("invalid floating point literal");
3280   } else if (errorToBool(
3281                  Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven)
3282                      .takeError()))
3283     return TokError("invalid floating point literal");
3284   if (IsNeg)
3285     Value.changeSign();
3286 
3287   // Consume the numeric token.
3288   Lex();
3289 
3290   Res = Value.bitcastToAPInt();
3291 
3292   return false;
3293 }
3294 
3295 /// parseDirectiveRealValue
3296 ///  ::= (.single | .double) [ expression (, expression)* ]
parseDirectiveRealValue(StringRef IDVal,const fltSemantics & Semantics)3297 bool AsmParser::parseDirectiveRealValue(StringRef IDVal,
3298                                         const fltSemantics &Semantics) {
3299   auto parseOp = [&]() -> bool {
3300     APInt AsInt;
3301     if (checkForValidSection() || parseRealValue(Semantics, AsInt))
3302       return true;
3303     getStreamer().emitIntValue(AsInt.getLimitedValue(),
3304                                AsInt.getBitWidth() / 8);
3305     return false;
3306   };
3307 
3308   return parseMany(parseOp);
3309 }
3310 
3311 /// parseDirectiveZero
3312 ///  ::= .zero expression
parseDirectiveZero()3313 bool AsmParser::parseDirectiveZero() {
3314   SMLoc NumBytesLoc = Lexer.getLoc();
3315   const MCExpr *NumBytes;
3316   if (checkForValidSection() || parseExpression(NumBytes))
3317     return true;
3318 
3319   int64_t Val = 0;
3320   if (getLexer().is(AsmToken::Comma)) {
3321     Lex();
3322     if (parseAbsoluteExpression(Val))
3323       return true;
3324   }
3325 
3326   if (parseEOL())
3327     return true;
3328   getStreamer().emitFill(*NumBytes, Val, NumBytesLoc);
3329 
3330   return false;
3331 }
3332 
3333 /// parseDirectiveFill
3334 ///  ::= .fill expression [ , expression [ , expression ] ]
parseDirectiveFill()3335 bool AsmParser::parseDirectiveFill() {
3336   SMLoc NumValuesLoc = Lexer.getLoc();
3337   const MCExpr *NumValues;
3338   if (checkForValidSection() || parseExpression(NumValues))
3339     return true;
3340 
3341   int64_t FillSize = 1;
3342   int64_t FillExpr = 0;
3343 
3344   SMLoc SizeLoc, ExprLoc;
3345 
3346   if (parseOptionalToken(AsmToken::Comma)) {
3347     SizeLoc = getTok().getLoc();
3348     if (parseAbsoluteExpression(FillSize))
3349       return true;
3350     if (parseOptionalToken(AsmToken::Comma)) {
3351       ExprLoc = getTok().getLoc();
3352       if (parseAbsoluteExpression(FillExpr))
3353         return true;
3354     }
3355   }
3356   if (parseEOL())
3357     return true;
3358 
3359   if (FillSize < 0) {
3360     Warning(SizeLoc, "'.fill' directive with negative size has no effect");
3361     return false;
3362   }
3363   if (FillSize > 8) {
3364     Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8");
3365     FillSize = 8;
3366   }
3367 
3368   if (!isUInt<32>(FillExpr) && FillSize > 4)
3369     Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits");
3370 
3371   getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc);
3372 
3373   return false;
3374 }
3375 
3376 /// parseDirectiveOrg
3377 ///  ::= .org expression [ , expression ]
parseDirectiveOrg()3378 bool AsmParser::parseDirectiveOrg() {
3379   const MCExpr *Offset;
3380   SMLoc OffsetLoc = Lexer.getLoc();
3381   if (checkForValidSection() || parseExpression(Offset))
3382     return true;
3383 
3384   // Parse optional fill expression.
3385   int64_t FillExpr = 0;
3386   if (parseOptionalToken(AsmToken::Comma))
3387     if (parseAbsoluteExpression(FillExpr))
3388       return true;
3389   if (parseEOL())
3390     return true;
3391 
3392   getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc);
3393   return false;
3394 }
3395 
3396 /// parseDirectiveAlign
3397 ///  ::= {.align, ...} expression [ , expression [ , expression ]]
parseDirectiveAlign(bool IsPow2,unsigned ValueSize)3398 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
3399   SMLoc AlignmentLoc = getLexer().getLoc();
3400   int64_t Alignment;
3401   SMLoc MaxBytesLoc;
3402   bool HasFillExpr = false;
3403   int64_t FillExpr = 0;
3404   int64_t MaxBytesToFill = 0;
3405   SMLoc FillExprLoc;
3406 
3407   auto parseAlign = [&]() -> bool {
3408     if (parseAbsoluteExpression(Alignment))
3409       return true;
3410     if (parseOptionalToken(AsmToken::Comma)) {
3411       // The fill expression can be omitted while specifying a maximum number of
3412       // alignment bytes, e.g:
3413       //  .align 3,,4
3414       if (getTok().isNot(AsmToken::Comma)) {
3415         HasFillExpr = true;
3416         if (parseTokenLoc(FillExprLoc) || parseAbsoluteExpression(FillExpr))
3417           return true;
3418       }
3419       if (parseOptionalToken(AsmToken::Comma))
3420         if (parseTokenLoc(MaxBytesLoc) ||
3421             parseAbsoluteExpression(MaxBytesToFill))
3422           return true;
3423     }
3424     return parseEOL();
3425   };
3426 
3427   if (checkForValidSection())
3428     return true;
3429   // Ignore empty '.p2align' directives for GNU-as compatibility
3430   if (IsPow2 && (ValueSize == 1) && getTok().is(AsmToken::EndOfStatement)) {
3431     Warning(AlignmentLoc, "p2align directive with no operand(s) is ignored");
3432     return parseEOL();
3433   }
3434   if (parseAlign())
3435     return true;
3436 
3437   // Always emit an alignment here even if we thrown an error.
3438   bool ReturnVal = false;
3439 
3440   // Compute alignment in bytes.
3441   if (IsPow2) {
3442     // FIXME: Diagnose overflow.
3443     if (Alignment >= 32) {
3444       ReturnVal |= Error(AlignmentLoc, "invalid alignment value");
3445       Alignment = 31;
3446     }
3447 
3448     Alignment = 1ULL << Alignment;
3449   } else {
3450     // Reject alignments that aren't either a power of two or zero,
3451     // for gas compatibility. Alignment of zero is silently rounded
3452     // up to one.
3453     if (Alignment == 0)
3454       Alignment = 1;
3455     else if (!isPowerOf2_64(Alignment)) {
3456       ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2");
3457       Alignment = llvm::bit_floor<uint64_t>(Alignment);
3458     }
3459     if (!isUInt<32>(Alignment)) {
3460       ReturnVal |= Error(AlignmentLoc, "alignment must be smaller than 2**32");
3461       Alignment = 1u << 31;
3462     }
3463   }
3464 
3465   if (HasFillExpr && FillExpr != 0) {
3466     MCSection *Sec = getStreamer().getCurrentSectionOnly();
3467     if (Sec && Sec->isVirtualSection()) {
3468       ReturnVal |=
3469           Warning(FillExprLoc, "ignoring non-zero fill value in " +
3470                                    Sec->getVirtualSectionKind() + " section '" +
3471                                    Sec->getName() + "'");
3472       FillExpr = 0;
3473     }
3474   }
3475 
3476   // Diagnose non-sensical max bytes to align.
3477   if (MaxBytesLoc.isValid()) {
3478     if (MaxBytesToFill < 1) {
3479       ReturnVal |= Error(MaxBytesLoc,
3480                          "alignment directive can never be satisfied in this "
3481                          "many bytes, ignoring maximum bytes expression");
3482       MaxBytesToFill = 0;
3483     }
3484 
3485     if (MaxBytesToFill >= Alignment) {
3486       Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
3487                            "has no effect");
3488       MaxBytesToFill = 0;
3489     }
3490   }
3491 
3492   // Check whether we should use optimal code alignment for this .align
3493   // directive.
3494   const MCSection *Section = getStreamer().getCurrentSectionOnly();
3495   assert(Section && "must have section to emit alignment");
3496   bool useCodeAlign = Section->useCodeAlign();
3497   if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
3498       ValueSize == 1 && useCodeAlign) {
3499     getStreamer().emitCodeAlignment(
3500         Align(Alignment), &getTargetParser().getSTI(), MaxBytesToFill);
3501   } else {
3502     // FIXME: Target specific behavior about how the "extra" bytes are filled.
3503     getStreamer().emitValueToAlignment(Align(Alignment), FillExpr, ValueSize,
3504                                        MaxBytesToFill);
3505   }
3506 
3507   return ReturnVal;
3508 }
3509 
3510 /// parseDirectiveFile
3511 /// ::= .file filename
3512 /// ::= .file number [directory] filename [md5 checksum] [source source-text]
parseDirectiveFile(SMLoc DirectiveLoc)3513 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) {
3514   // FIXME: I'm not sure what this is.
3515   int64_t FileNumber = -1;
3516   if (getLexer().is(AsmToken::Integer)) {
3517     FileNumber = getTok().getIntVal();
3518     Lex();
3519 
3520     if (FileNumber < 0)
3521       return TokError("negative file number");
3522   }
3523 
3524   std::string Path;
3525 
3526   // Usually the directory and filename together, otherwise just the directory.
3527   // Allow the strings to have escaped octal character sequence.
3528   if (parseEscapedString(Path))
3529     return true;
3530 
3531   StringRef Directory;
3532   StringRef Filename;
3533   std::string FilenameData;
3534   if (getLexer().is(AsmToken::String)) {
3535     if (check(FileNumber == -1,
3536               "explicit path specified, but no file number") ||
3537         parseEscapedString(FilenameData))
3538       return true;
3539     Filename = FilenameData;
3540     Directory = Path;
3541   } else {
3542     Filename = Path;
3543   }
3544 
3545   uint64_t MD5Hi, MD5Lo;
3546   bool HasMD5 = false;
3547 
3548   std::optional<StringRef> Source;
3549   bool HasSource = false;
3550   std::string SourceString;
3551 
3552   while (!parseOptionalToken(AsmToken::EndOfStatement)) {
3553     StringRef Keyword;
3554     if (check(getTok().isNot(AsmToken::Identifier),
3555               "unexpected token in '.file' directive") ||
3556         parseIdentifier(Keyword))
3557       return true;
3558     if (Keyword == "md5") {
3559       HasMD5 = true;
3560       if (check(FileNumber == -1,
3561                 "MD5 checksum specified, but no file number") ||
3562           parseHexOcta(*this, MD5Hi, MD5Lo))
3563         return true;
3564     } else if (Keyword == "source") {
3565       HasSource = true;
3566       if (check(FileNumber == -1,
3567                 "source specified, but no file number") ||
3568           check(getTok().isNot(AsmToken::String),
3569                 "unexpected token in '.file' directive") ||
3570           parseEscapedString(SourceString))
3571         return true;
3572     } else {
3573       return TokError("unexpected token in '.file' directive");
3574     }
3575   }
3576 
3577   if (FileNumber == -1) {
3578     // Ignore the directive if there is no number and the target doesn't support
3579     // numberless .file directives. This allows some portability of assembler
3580     // between different object file formats.
3581     if (getContext().getAsmInfo()->hasSingleParameterDotFile())
3582       getStreamer().emitFileDirective(Filename);
3583   } else {
3584     // In case there is a -g option as well as debug info from directive .file,
3585     // we turn off the -g option, directly use the existing debug info instead.
3586     // Throw away any implicit file table for the assembler source.
3587     if (Ctx.getGenDwarfForAssembly()) {
3588       Ctx.getMCDwarfLineTable(0).resetFileTable();
3589       Ctx.setGenDwarfForAssembly(false);
3590     }
3591 
3592     std::optional<MD5::MD5Result> CKMem;
3593     if (HasMD5) {
3594       MD5::MD5Result Sum;
3595       for (unsigned i = 0; i != 8; ++i) {
3596         Sum[i] = uint8_t(MD5Hi >> ((7 - i) * 8));
3597         Sum[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8));
3598       }
3599       CKMem = Sum;
3600     }
3601     if (HasSource) {
3602       char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size()));
3603       memcpy(SourceBuf, SourceString.data(), SourceString.size());
3604       Source = StringRef(SourceBuf, SourceString.size());
3605     }
3606     if (FileNumber == 0) {
3607       // Upgrade to Version 5 for assembly actions like clang -c a.s.
3608       if (Ctx.getDwarfVersion() < 5)
3609         Ctx.setDwarfVersion(5);
3610       getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source);
3611     } else {
3612       Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective(
3613           FileNumber, Directory, Filename, CKMem, Source);
3614       if (!FileNumOrErr)
3615         return Error(DirectiveLoc, toString(FileNumOrErr.takeError()));
3616     }
3617     // Alert the user if there are some .file directives with MD5 and some not.
3618     // But only do that once.
3619     if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) {
3620       ReportedInconsistentMD5 = true;
3621       return Warning(DirectiveLoc, "inconsistent use of MD5 checksums");
3622     }
3623   }
3624 
3625   return false;
3626 }
3627 
3628 /// parseDirectiveLine
3629 /// ::= .line [number]
parseDirectiveLine()3630 bool AsmParser::parseDirectiveLine() {
3631   int64_t LineNumber;
3632   if (getLexer().is(AsmToken::Integer)) {
3633     if (parseIntToken(LineNumber, "unexpected token in '.line' directive"))
3634       return true;
3635     (void)LineNumber;
3636     // FIXME: Do something with the .line.
3637   }
3638   return parseEOL();
3639 }
3640 
3641 /// parseDirectiveLoc
3642 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
3643 ///                                [epilogue_begin] [is_stmt VALUE] [isa VALUE]
3644 /// The first number is a file number, must have been previously assigned with
3645 /// a .file directive, the second number is the line number and optionally the
3646 /// third number is a column position (zero if not specified).  The remaining
3647 /// optional items are .loc sub-directives.
parseDirectiveLoc()3648 bool AsmParser::parseDirectiveLoc() {
3649   int64_t FileNumber = 0, LineNumber = 0;
3650   SMLoc Loc = getTok().getLoc();
3651   if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") ||
3652       check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc,
3653             "file number less than one in '.loc' directive") ||
3654       check(!getContext().isValidDwarfFileNumber(FileNumber), Loc,
3655             "unassigned file number in '.loc' directive"))
3656     return true;
3657 
3658   // optional
3659   if (getLexer().is(AsmToken::Integer)) {
3660     LineNumber = getTok().getIntVal();
3661     if (LineNumber < 0)
3662       return TokError("line number less than zero in '.loc' directive");
3663     Lex();
3664   }
3665 
3666   int64_t ColumnPos = 0;
3667   if (getLexer().is(AsmToken::Integer)) {
3668     ColumnPos = getTok().getIntVal();
3669     if (ColumnPos < 0)
3670       return TokError("column position less than zero in '.loc' directive");
3671     Lex();
3672   }
3673 
3674   auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags();
3675   unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT;
3676   unsigned Isa = 0;
3677   int64_t Discriminator = 0;
3678 
3679   auto parseLocOp = [&]() -> bool {
3680     StringRef Name;
3681     SMLoc Loc = getTok().getLoc();
3682     if (parseIdentifier(Name))
3683       return TokError("unexpected token in '.loc' directive");
3684 
3685     if (Name == "basic_block")
3686       Flags |= DWARF2_FLAG_BASIC_BLOCK;
3687     else if (Name == "prologue_end")
3688       Flags |= DWARF2_FLAG_PROLOGUE_END;
3689     else if (Name == "epilogue_begin")
3690       Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
3691     else if (Name == "is_stmt") {
3692       Loc = getTok().getLoc();
3693       const MCExpr *Value;
3694       if (parseExpression(Value))
3695         return true;
3696       // The expression must be the constant 0 or 1.
3697       if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3698         int Value = MCE->getValue();
3699         if (Value == 0)
3700           Flags &= ~DWARF2_FLAG_IS_STMT;
3701         else if (Value == 1)
3702           Flags |= DWARF2_FLAG_IS_STMT;
3703         else
3704           return Error(Loc, "is_stmt value not 0 or 1");
3705       } else {
3706         return Error(Loc, "is_stmt value not the constant value of 0 or 1");
3707       }
3708     } else if (Name == "isa") {
3709       Loc = getTok().getLoc();
3710       const MCExpr *Value;
3711       if (parseExpression(Value))
3712         return true;
3713       // The expression must be a constant greater or equal to 0.
3714       if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3715         int Value = MCE->getValue();
3716         if (Value < 0)
3717           return Error(Loc, "isa number less than zero");
3718         Isa = Value;
3719       } else {
3720         return Error(Loc, "isa number not a constant value");
3721       }
3722     } else if (Name == "discriminator") {
3723       if (parseAbsoluteExpression(Discriminator))
3724         return true;
3725     } else {
3726       return Error(Loc, "unknown sub-directive in '.loc' directive");
3727     }
3728     return false;
3729   };
3730 
3731   if (parseMany(parseLocOp, false /*hasComma*/))
3732     return true;
3733 
3734   getStreamer().emitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
3735                                       Isa, Discriminator, StringRef());
3736 
3737   return false;
3738 }
3739 
3740 /// parseDirectiveStabs
3741 /// ::= .stabs string, number, number, number
parseDirectiveStabs()3742 bool AsmParser::parseDirectiveStabs() {
3743   return TokError("unsupported directive '.stabs'");
3744 }
3745 
3746 /// parseDirectiveCVFile
3747 /// ::= .cv_file number filename [checksum] [checksumkind]
parseDirectiveCVFile()3748 bool AsmParser::parseDirectiveCVFile() {
3749   SMLoc FileNumberLoc = getTok().getLoc();
3750   int64_t FileNumber;
3751   std::string Filename;
3752   std::string Checksum;
3753   int64_t ChecksumKind = 0;
3754 
3755   if (parseIntToken(FileNumber,
3756                     "expected file number in '.cv_file' directive") ||
3757       check(FileNumber < 1, FileNumberLoc, "file number less than one") ||
3758       check(getTok().isNot(AsmToken::String),
3759             "unexpected token in '.cv_file' directive") ||
3760       parseEscapedString(Filename))
3761     return true;
3762   if (!parseOptionalToken(AsmToken::EndOfStatement)) {
3763     if (check(getTok().isNot(AsmToken::String),
3764               "unexpected token in '.cv_file' directive") ||
3765         parseEscapedString(Checksum) ||
3766         parseIntToken(ChecksumKind,
3767                       "expected checksum kind in '.cv_file' directive") ||
3768         parseEOL())
3769       return true;
3770   }
3771 
3772   Checksum = fromHex(Checksum);
3773   void *CKMem = Ctx.allocate(Checksum.size(), 1);
3774   memcpy(CKMem, Checksum.data(), Checksum.size());
3775   ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem),
3776                                     Checksum.size());
3777 
3778   if (!getStreamer().emitCVFileDirective(FileNumber, Filename, ChecksumAsBytes,
3779                                          static_cast<uint8_t>(ChecksumKind)))
3780     return Error(FileNumberLoc, "file number already allocated");
3781 
3782   return false;
3783 }
3784 
parseCVFunctionId(int64_t & FunctionId,StringRef DirectiveName)3785 bool AsmParser::parseCVFunctionId(int64_t &FunctionId,
3786                                   StringRef DirectiveName) {
3787   SMLoc Loc;
3788   return parseTokenLoc(Loc) ||
3789          parseIntToken(FunctionId, "expected function id in '" + DirectiveName +
3790                                        "' directive") ||
3791          check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc,
3792                "expected function id within range [0, UINT_MAX)");
3793 }
3794 
parseCVFileId(int64_t & FileNumber,StringRef DirectiveName)3795 bool AsmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) {
3796   SMLoc Loc;
3797   return parseTokenLoc(Loc) ||
3798          parseIntToken(FileNumber, "expected integer in '" + DirectiveName +
3799                                        "' directive") ||
3800          check(FileNumber < 1, Loc, "file number less than one in '" +
3801                                         DirectiveName + "' directive") ||
3802          check(!getCVContext().isValidFileNumber(FileNumber), Loc,
3803                "unassigned file number in '" + DirectiveName + "' directive");
3804 }
3805 
3806 /// parseDirectiveCVFuncId
3807 /// ::= .cv_func_id FunctionId
3808 ///
3809 /// Introduces a function ID that can be used with .cv_loc.
parseDirectiveCVFuncId()3810 bool AsmParser::parseDirectiveCVFuncId() {
3811   SMLoc FunctionIdLoc = getTok().getLoc();
3812   int64_t FunctionId;
3813 
3814   if (parseCVFunctionId(FunctionId, ".cv_func_id") || parseEOL())
3815     return true;
3816 
3817   if (!getStreamer().emitCVFuncIdDirective(FunctionId))
3818     return Error(FunctionIdLoc, "function id already allocated");
3819 
3820   return false;
3821 }
3822 
3823 /// parseDirectiveCVInlineSiteId
3824 /// ::= .cv_inline_site_id FunctionId
3825 ///         "within" IAFunc
3826 ///         "inlined_at" IAFile IALine [IACol]
3827 ///
3828 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined
3829 /// at" source location information for use in the line table of the caller,
3830 /// whether the caller is a real function or another inlined call site.
parseDirectiveCVInlineSiteId()3831 bool AsmParser::parseDirectiveCVInlineSiteId() {
3832   SMLoc FunctionIdLoc = getTok().getLoc();
3833   int64_t FunctionId;
3834   int64_t IAFunc;
3835   int64_t IAFile;
3836   int64_t IALine;
3837   int64_t IACol = 0;
3838 
3839   // FunctionId
3840   if (parseCVFunctionId(FunctionId, ".cv_inline_site_id"))
3841     return true;
3842 
3843   // "within"
3844   if (check((getLexer().isNot(AsmToken::Identifier) ||
3845              getTok().getIdentifier() != "within"),
3846             "expected 'within' identifier in '.cv_inline_site_id' directive"))
3847     return true;
3848   Lex();
3849 
3850   // IAFunc
3851   if (parseCVFunctionId(IAFunc, ".cv_inline_site_id"))
3852     return true;
3853 
3854   // "inlined_at"
3855   if (check((getLexer().isNot(AsmToken::Identifier) ||
3856              getTok().getIdentifier() != "inlined_at"),
3857             "expected 'inlined_at' identifier in '.cv_inline_site_id' "
3858             "directive") )
3859     return true;
3860   Lex();
3861 
3862   // IAFile IALine
3863   if (parseCVFileId(IAFile, ".cv_inline_site_id") ||
3864       parseIntToken(IALine, "expected line number after 'inlined_at'"))
3865     return true;
3866 
3867   // [IACol]
3868   if (getLexer().is(AsmToken::Integer)) {
3869     IACol = getTok().getIntVal();
3870     Lex();
3871   }
3872 
3873   if (parseEOL())
3874     return true;
3875 
3876   if (!getStreamer().emitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile,
3877                                                  IALine, IACol, FunctionIdLoc))
3878     return Error(FunctionIdLoc, "function id already allocated");
3879 
3880   return false;
3881 }
3882 
3883 /// parseDirectiveCVLoc
3884 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end]
3885 ///                                [is_stmt VALUE]
3886 /// The first number is a file number, must have been previously assigned with
3887 /// a .file directive, the second number is the line number and optionally the
3888 /// third number is a column position (zero if not specified).  The remaining
3889 /// optional items are .loc sub-directives.
parseDirectiveCVLoc()3890 bool AsmParser::parseDirectiveCVLoc() {
3891   SMLoc DirectiveLoc = getTok().getLoc();
3892   int64_t FunctionId, FileNumber;
3893   if (parseCVFunctionId(FunctionId, ".cv_loc") ||
3894       parseCVFileId(FileNumber, ".cv_loc"))
3895     return true;
3896 
3897   int64_t LineNumber = 0;
3898   if (getLexer().is(AsmToken::Integer)) {
3899     LineNumber = getTok().getIntVal();
3900     if (LineNumber < 0)
3901       return TokError("line number less than zero in '.cv_loc' directive");
3902     Lex();
3903   }
3904 
3905   int64_t ColumnPos = 0;
3906   if (getLexer().is(AsmToken::Integer)) {
3907     ColumnPos = getTok().getIntVal();
3908     if (ColumnPos < 0)
3909       return TokError("column position less than zero in '.cv_loc' directive");
3910     Lex();
3911   }
3912 
3913   bool PrologueEnd = false;
3914   uint64_t IsStmt = 0;
3915 
3916   auto parseOp = [&]() -> bool {
3917     StringRef Name;
3918     SMLoc Loc = getTok().getLoc();
3919     if (parseIdentifier(Name))
3920       return TokError("unexpected token in '.cv_loc' directive");
3921     if (Name == "prologue_end")
3922       PrologueEnd = true;
3923     else if (Name == "is_stmt") {
3924       Loc = getTok().getLoc();
3925       const MCExpr *Value;
3926       if (parseExpression(Value))
3927         return true;
3928       // The expression must be the constant 0 or 1.
3929       IsStmt = ~0ULL;
3930       if (const auto *MCE = dyn_cast<MCConstantExpr>(Value))
3931         IsStmt = MCE->getValue();
3932 
3933       if (IsStmt > 1)
3934         return Error(Loc, "is_stmt value not 0 or 1");
3935     } else {
3936       return Error(Loc, "unknown sub-directive in '.cv_loc' directive");
3937     }
3938     return false;
3939   };
3940 
3941   if (parseMany(parseOp, false /*hasComma*/))
3942     return true;
3943 
3944   getStreamer().emitCVLocDirective(FunctionId, FileNumber, LineNumber,
3945                                    ColumnPos, PrologueEnd, IsStmt, StringRef(),
3946                                    DirectiveLoc);
3947   return false;
3948 }
3949 
3950 /// parseDirectiveCVLinetable
3951 /// ::= .cv_linetable FunctionId, FnStart, FnEnd
parseDirectiveCVLinetable()3952 bool AsmParser::parseDirectiveCVLinetable() {
3953   int64_t FunctionId;
3954   StringRef FnStartName, FnEndName;
3955   SMLoc Loc = getTok().getLoc();
3956   if (parseCVFunctionId(FunctionId, ".cv_linetable") || parseComma() ||
3957       parseTokenLoc(Loc) ||
3958       check(parseIdentifier(FnStartName), Loc,
3959             "expected identifier in directive") ||
3960       parseComma() || parseTokenLoc(Loc) ||
3961       check(parseIdentifier(FnEndName), Loc,
3962             "expected identifier in directive"))
3963     return true;
3964 
3965   MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
3966   MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
3967 
3968   getStreamer().emitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym);
3969   return false;
3970 }
3971 
3972 /// parseDirectiveCVInlineLinetable
3973 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd
parseDirectiveCVInlineLinetable()3974 bool AsmParser::parseDirectiveCVInlineLinetable() {
3975   int64_t PrimaryFunctionId, SourceFileId, SourceLineNum;
3976   StringRef FnStartName, FnEndName;
3977   SMLoc Loc = getTok().getLoc();
3978   if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") ||
3979       parseTokenLoc(Loc) ||
3980       parseIntToken(
3981           SourceFileId,
3982           "expected SourceField in '.cv_inline_linetable' directive") ||
3983       check(SourceFileId <= 0, Loc,
3984             "File id less than zero in '.cv_inline_linetable' directive") ||
3985       parseTokenLoc(Loc) ||
3986       parseIntToken(
3987           SourceLineNum,
3988           "expected SourceLineNum in '.cv_inline_linetable' directive") ||
3989       check(SourceLineNum < 0, Loc,
3990             "Line number less than zero in '.cv_inline_linetable' directive") ||
3991       parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc,
3992                                   "expected identifier in directive") ||
3993       parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc,
3994                                   "expected identifier in directive"))
3995     return true;
3996 
3997   if (parseEOL())
3998     return true;
3999 
4000   MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
4001   MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
4002   getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId,
4003                                                SourceLineNum, FnStartSym,
4004                                                FnEndSym);
4005   return false;
4006 }
4007 
initializeCVDefRangeTypeMap()4008 void AsmParser::initializeCVDefRangeTypeMap() {
4009   CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER;
4010   CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL;
4011   CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER;
4012   CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL;
4013 }
4014 
4015 /// parseDirectiveCVDefRange
4016 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes*
parseDirectiveCVDefRange()4017 bool AsmParser::parseDirectiveCVDefRange() {
4018   SMLoc Loc;
4019   std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges;
4020   while (getLexer().is(AsmToken::Identifier)) {
4021     Loc = getLexer().getLoc();
4022     StringRef GapStartName;
4023     if (parseIdentifier(GapStartName))
4024       return Error(Loc, "expected identifier in directive");
4025     MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName);
4026 
4027     Loc = getLexer().getLoc();
4028     StringRef GapEndName;
4029     if (parseIdentifier(GapEndName))
4030       return Error(Loc, "expected identifier in directive");
4031     MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName);
4032 
4033     Ranges.push_back({GapStartSym, GapEndSym});
4034   }
4035 
4036   StringRef CVDefRangeTypeStr;
4037   if (parseToken(
4038           AsmToken::Comma,
4039           "expected comma before def_range type in .cv_def_range directive") ||
4040       parseIdentifier(CVDefRangeTypeStr))
4041     return Error(Loc, "expected def_range type in directive");
4042 
4043   StringMap<CVDefRangeType>::const_iterator CVTypeIt =
4044       CVDefRangeTypeMap.find(CVDefRangeTypeStr);
4045   CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end())
4046                                 ? CVDR_DEFRANGE
4047                                 : CVTypeIt->getValue();
4048   switch (CVDRType) {
4049   case CVDR_DEFRANGE_REGISTER: {
4050     int64_t DRRegister;
4051     if (parseToken(AsmToken::Comma, "expected comma before register number in "
4052                                     ".cv_def_range directive") ||
4053         parseAbsoluteExpression(DRRegister))
4054       return Error(Loc, "expected register number");
4055 
4056     codeview::DefRangeRegisterHeader DRHdr;
4057     DRHdr.Register = DRRegister;
4058     DRHdr.MayHaveNoName = 0;
4059     getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4060     break;
4061   }
4062   case CVDR_DEFRANGE_FRAMEPOINTER_REL: {
4063     int64_t DROffset;
4064     if (parseToken(AsmToken::Comma,
4065                    "expected comma before offset in .cv_def_range directive") ||
4066         parseAbsoluteExpression(DROffset))
4067       return Error(Loc, "expected offset value");
4068 
4069     codeview::DefRangeFramePointerRelHeader DRHdr;
4070     DRHdr.Offset = DROffset;
4071     getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4072     break;
4073   }
4074   case CVDR_DEFRANGE_SUBFIELD_REGISTER: {
4075     int64_t DRRegister;
4076     int64_t DROffsetInParent;
4077     if (parseToken(AsmToken::Comma, "expected comma before register number in "
4078                                     ".cv_def_range directive") ||
4079         parseAbsoluteExpression(DRRegister))
4080       return Error(Loc, "expected register number");
4081     if (parseToken(AsmToken::Comma,
4082                    "expected comma before offset in .cv_def_range directive") ||
4083         parseAbsoluteExpression(DROffsetInParent))
4084       return Error(Loc, "expected offset value");
4085 
4086     codeview::DefRangeSubfieldRegisterHeader DRHdr;
4087     DRHdr.Register = DRRegister;
4088     DRHdr.MayHaveNoName = 0;
4089     DRHdr.OffsetInParent = DROffsetInParent;
4090     getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4091     break;
4092   }
4093   case CVDR_DEFRANGE_REGISTER_REL: {
4094     int64_t DRRegister;
4095     int64_t DRFlags;
4096     int64_t DRBasePointerOffset;
4097     if (parseToken(AsmToken::Comma, "expected comma before register number in "
4098                                     ".cv_def_range directive") ||
4099         parseAbsoluteExpression(DRRegister))
4100       return Error(Loc, "expected register value");
4101     if (parseToken(
4102             AsmToken::Comma,
4103             "expected comma before flag value in .cv_def_range directive") ||
4104         parseAbsoluteExpression(DRFlags))
4105       return Error(Loc, "expected flag value");
4106     if (parseToken(AsmToken::Comma, "expected comma before base pointer offset "
4107                                     "in .cv_def_range directive") ||
4108         parseAbsoluteExpression(DRBasePointerOffset))
4109       return Error(Loc, "expected base pointer offset value");
4110 
4111     codeview::DefRangeRegisterRelHeader DRHdr;
4112     DRHdr.Register = DRRegister;
4113     DRHdr.Flags = DRFlags;
4114     DRHdr.BasePointerOffset = DRBasePointerOffset;
4115     getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4116     break;
4117   }
4118   default:
4119     return Error(Loc, "unexpected def_range type in .cv_def_range directive");
4120   }
4121   return true;
4122 }
4123 
4124 /// parseDirectiveCVString
4125 /// ::= .cv_stringtable "string"
parseDirectiveCVString()4126 bool AsmParser::parseDirectiveCVString() {
4127   std::string Data;
4128   if (checkForValidSection() || parseEscapedString(Data))
4129     return true;
4130 
4131   // Put the string in the table and emit the offset.
4132   std::pair<StringRef, unsigned> Insertion =
4133       getCVContext().addToStringTable(Data);
4134   getStreamer().emitInt32(Insertion.second);
4135   return false;
4136 }
4137 
4138 /// parseDirectiveCVStringTable
4139 /// ::= .cv_stringtable
parseDirectiveCVStringTable()4140 bool AsmParser::parseDirectiveCVStringTable() {
4141   getStreamer().emitCVStringTableDirective();
4142   return false;
4143 }
4144 
4145 /// parseDirectiveCVFileChecksums
4146 /// ::= .cv_filechecksums
parseDirectiveCVFileChecksums()4147 bool AsmParser::parseDirectiveCVFileChecksums() {
4148   getStreamer().emitCVFileChecksumsDirective();
4149   return false;
4150 }
4151 
4152 /// parseDirectiveCVFileChecksumOffset
4153 /// ::= .cv_filechecksumoffset fileno
parseDirectiveCVFileChecksumOffset()4154 bool AsmParser::parseDirectiveCVFileChecksumOffset() {
4155   int64_t FileNo;
4156   if (parseIntToken(FileNo, "expected identifier in directive"))
4157     return true;
4158   if (parseEOL())
4159     return true;
4160   getStreamer().emitCVFileChecksumOffsetDirective(FileNo);
4161   return false;
4162 }
4163 
4164 /// parseDirectiveCVFPOData
4165 /// ::= .cv_fpo_data procsym
parseDirectiveCVFPOData()4166 bool AsmParser::parseDirectiveCVFPOData() {
4167   SMLoc DirLoc = getLexer().getLoc();
4168   StringRef ProcName;
4169   if (parseIdentifier(ProcName))
4170     return TokError("expected symbol name");
4171   if (parseEOL())
4172     return true;
4173   MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName);
4174   getStreamer().emitCVFPOData(ProcSym, DirLoc);
4175   return false;
4176 }
4177 
4178 /// parseDirectiveCFISections
4179 /// ::= .cfi_sections section [, section]
parseDirectiveCFISections()4180 bool AsmParser::parseDirectiveCFISections() {
4181   StringRef Name;
4182   bool EH = false;
4183   bool Debug = false;
4184 
4185   if (!parseOptionalToken(AsmToken::EndOfStatement)) {
4186     for (;;) {
4187       if (parseIdentifier(Name))
4188         return TokError("expected .eh_frame or .debug_frame");
4189       if (Name == ".eh_frame")
4190         EH = true;
4191       else if (Name == ".debug_frame")
4192         Debug = true;
4193       if (parseOptionalToken(AsmToken::EndOfStatement))
4194         break;
4195       if (parseComma())
4196         return true;
4197     }
4198   }
4199   getStreamer().emitCFISections(EH, Debug);
4200   return false;
4201 }
4202 
4203 /// parseDirectiveCFIStartProc
4204 /// ::= .cfi_startproc [simple]
parseDirectiveCFIStartProc()4205 bool AsmParser::parseDirectiveCFIStartProc() {
4206   CFIStartProcLoc = StartTokLoc;
4207 
4208   StringRef Simple;
4209   if (!parseOptionalToken(AsmToken::EndOfStatement)) {
4210     if (check(parseIdentifier(Simple) || Simple != "simple",
4211               "unexpected token") ||
4212         parseEOL())
4213       return true;
4214   }
4215 
4216   // TODO(kristina): Deal with a corner case of incorrect diagnostic context
4217   // being produced if this directive is emitted as part of preprocessor macro
4218   // expansion which can *ONLY* happen if Clang's cc1as is the API consumer.
4219   // Tools like llvm-mc on the other hand are not affected by it, and report
4220   // correct context information.
4221   getStreamer().emitCFIStartProc(!Simple.empty(), Lexer.getLoc());
4222   return false;
4223 }
4224 
4225 /// parseDirectiveCFIEndProc
4226 /// ::= .cfi_endproc
parseDirectiveCFIEndProc()4227 bool AsmParser::parseDirectiveCFIEndProc() {
4228   CFIStartProcLoc = std::nullopt;
4229 
4230   if (parseEOL())
4231     return true;
4232 
4233   getStreamer().emitCFIEndProc();
4234   return false;
4235 }
4236 
4237 /// parse register name or number.
parseRegisterOrRegisterNumber(int64_t & Register,SMLoc DirectiveLoc)4238 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register,
4239                                               SMLoc DirectiveLoc) {
4240   MCRegister RegNo;
4241 
4242   if (getLexer().isNot(AsmToken::Integer)) {
4243     if (getTargetParser().parseRegister(RegNo, DirectiveLoc, DirectiveLoc))
4244       return true;
4245     Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true);
4246   } else
4247     return parseAbsoluteExpression(Register);
4248 
4249   return false;
4250 }
4251 
4252 /// parseDirectiveCFIDefCfa
4253 /// ::= .cfi_def_cfa register,  offset
parseDirectiveCFIDefCfa(SMLoc DirectiveLoc)4254 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) {
4255   int64_t Register = 0, Offset = 0;
4256   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4257       parseAbsoluteExpression(Offset) || parseEOL())
4258     return true;
4259 
4260   getStreamer().emitCFIDefCfa(Register, Offset, DirectiveLoc);
4261   return false;
4262 }
4263 
4264 /// parseDirectiveCFIDefCfaOffset
4265 /// ::= .cfi_def_cfa_offset offset
parseDirectiveCFIDefCfaOffset(SMLoc DirectiveLoc)4266 bool AsmParser::parseDirectiveCFIDefCfaOffset(SMLoc DirectiveLoc) {
4267   int64_t Offset = 0;
4268   if (parseAbsoluteExpression(Offset) || parseEOL())
4269     return true;
4270 
4271   getStreamer().emitCFIDefCfaOffset(Offset, DirectiveLoc);
4272   return false;
4273 }
4274 
4275 /// parseDirectiveCFIRegister
4276 /// ::= .cfi_register register, register
parseDirectiveCFIRegister(SMLoc DirectiveLoc)4277 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) {
4278   int64_t Register1 = 0, Register2 = 0;
4279   if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || parseComma() ||
4280       parseRegisterOrRegisterNumber(Register2, DirectiveLoc) || parseEOL())
4281     return true;
4282 
4283   getStreamer().emitCFIRegister(Register1, Register2, DirectiveLoc);
4284   return false;
4285 }
4286 
4287 /// parseDirectiveCFIWindowSave
4288 /// ::= .cfi_window_save
parseDirectiveCFIWindowSave(SMLoc DirectiveLoc)4289 bool AsmParser::parseDirectiveCFIWindowSave(SMLoc DirectiveLoc) {
4290   if (parseEOL())
4291     return true;
4292   getStreamer().emitCFIWindowSave(DirectiveLoc);
4293   return false;
4294 }
4295 
4296 /// parseDirectiveCFIAdjustCfaOffset
4297 /// ::= .cfi_adjust_cfa_offset adjustment
parseDirectiveCFIAdjustCfaOffset(SMLoc DirectiveLoc)4298 bool AsmParser::parseDirectiveCFIAdjustCfaOffset(SMLoc DirectiveLoc) {
4299   int64_t Adjustment = 0;
4300   if (parseAbsoluteExpression(Adjustment) || parseEOL())
4301     return true;
4302 
4303   getStreamer().emitCFIAdjustCfaOffset(Adjustment, DirectiveLoc);
4304   return false;
4305 }
4306 
4307 /// parseDirectiveCFIDefCfaRegister
4308 /// ::= .cfi_def_cfa_register register
parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc)4309 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) {
4310   int64_t Register = 0;
4311   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4312     return true;
4313 
4314   getStreamer().emitCFIDefCfaRegister(Register, DirectiveLoc);
4315   return false;
4316 }
4317 
4318 /// parseDirectiveCFILLVMDefAspaceCfa
4319 /// ::= .cfi_llvm_def_aspace_cfa register, offset, address_space
parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc)4320 bool AsmParser::parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc) {
4321   int64_t Register = 0, Offset = 0, AddressSpace = 0;
4322   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4323       parseAbsoluteExpression(Offset) || parseComma() ||
4324       parseAbsoluteExpression(AddressSpace) || parseEOL())
4325     return true;
4326 
4327   getStreamer().emitCFILLVMDefAspaceCfa(Register, Offset, AddressSpace,
4328                                         DirectiveLoc);
4329   return false;
4330 }
4331 
4332 /// parseDirectiveCFIOffset
4333 /// ::= .cfi_offset register, offset
parseDirectiveCFIOffset(SMLoc DirectiveLoc)4334 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) {
4335   int64_t Register = 0;
4336   int64_t Offset = 0;
4337 
4338   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4339       parseAbsoluteExpression(Offset) || parseEOL())
4340     return true;
4341 
4342   getStreamer().emitCFIOffset(Register, Offset, DirectiveLoc);
4343   return false;
4344 }
4345 
4346 /// parseDirectiveCFIRelOffset
4347 /// ::= .cfi_rel_offset register, offset
parseDirectiveCFIRelOffset(SMLoc DirectiveLoc)4348 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) {
4349   int64_t Register = 0, Offset = 0;
4350 
4351   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4352       parseAbsoluteExpression(Offset) || parseEOL())
4353     return true;
4354 
4355   getStreamer().emitCFIRelOffset(Register, Offset, DirectiveLoc);
4356   return false;
4357 }
4358 
isValidEncoding(int64_t Encoding)4359 static bool isValidEncoding(int64_t Encoding) {
4360   if (Encoding & ~0xff)
4361     return false;
4362 
4363   if (Encoding == dwarf::DW_EH_PE_omit)
4364     return true;
4365 
4366   const unsigned Format = Encoding & 0xf;
4367   if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
4368       Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
4369       Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
4370       Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
4371     return false;
4372 
4373   const unsigned Application = Encoding & 0x70;
4374   if (Application != dwarf::DW_EH_PE_absptr &&
4375       Application != dwarf::DW_EH_PE_pcrel)
4376     return false;
4377 
4378   return true;
4379 }
4380 
4381 /// parseDirectiveCFIPersonalityOrLsda
4382 /// IsPersonality true for cfi_personality, false for cfi_lsda
4383 /// ::= .cfi_personality encoding, [symbol_name]
4384 /// ::= .cfi_lsda encoding, [symbol_name]
parseDirectiveCFIPersonalityOrLsda(bool IsPersonality)4385 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) {
4386   int64_t Encoding = 0;
4387   if (parseAbsoluteExpression(Encoding))
4388     return true;
4389   if (Encoding == dwarf::DW_EH_PE_omit)
4390     return false;
4391 
4392   StringRef Name;
4393   if (check(!isValidEncoding(Encoding), "unsupported encoding.") ||
4394       parseComma() ||
4395       check(parseIdentifier(Name), "expected identifier in directive") ||
4396       parseEOL())
4397     return true;
4398 
4399   MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
4400 
4401   if (IsPersonality)
4402     getStreamer().emitCFIPersonality(Sym, Encoding);
4403   else
4404     getStreamer().emitCFILsda(Sym, Encoding);
4405   return false;
4406 }
4407 
4408 /// parseDirectiveCFIRememberState
4409 /// ::= .cfi_remember_state
parseDirectiveCFIRememberState(SMLoc DirectiveLoc)4410 bool AsmParser::parseDirectiveCFIRememberState(SMLoc DirectiveLoc) {
4411   if (parseEOL())
4412     return true;
4413   getStreamer().emitCFIRememberState(DirectiveLoc);
4414   return false;
4415 }
4416 
4417 /// parseDirectiveCFIRestoreState
4418 /// ::= .cfi_remember_state
parseDirectiveCFIRestoreState(SMLoc DirectiveLoc)4419 bool AsmParser::parseDirectiveCFIRestoreState(SMLoc DirectiveLoc) {
4420   if (parseEOL())
4421     return true;
4422   getStreamer().emitCFIRestoreState(DirectiveLoc);
4423   return false;
4424 }
4425 
4426 /// parseDirectiveCFISameValue
4427 /// ::= .cfi_same_value register
parseDirectiveCFISameValue(SMLoc DirectiveLoc)4428 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) {
4429   int64_t Register = 0;
4430 
4431   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4432     return true;
4433 
4434   getStreamer().emitCFISameValue(Register, DirectiveLoc);
4435   return false;
4436 }
4437 
4438 /// parseDirectiveCFIRestore
4439 /// ::= .cfi_restore register
parseDirectiveCFIRestore(SMLoc DirectiveLoc)4440 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) {
4441   int64_t Register = 0;
4442   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4443     return true;
4444 
4445   getStreamer().emitCFIRestore(Register, DirectiveLoc);
4446   return false;
4447 }
4448 
4449 /// parseDirectiveCFIEscape
4450 /// ::= .cfi_escape expression[,...]
parseDirectiveCFIEscape(SMLoc DirectiveLoc)4451 bool AsmParser::parseDirectiveCFIEscape(SMLoc DirectiveLoc) {
4452   std::string Values;
4453   int64_t CurrValue;
4454   if (parseAbsoluteExpression(CurrValue))
4455     return true;
4456 
4457   Values.push_back((uint8_t)CurrValue);
4458 
4459   while (getLexer().is(AsmToken::Comma)) {
4460     Lex();
4461 
4462     if (parseAbsoluteExpression(CurrValue))
4463       return true;
4464 
4465     Values.push_back((uint8_t)CurrValue);
4466   }
4467 
4468   getStreamer().emitCFIEscape(Values, DirectiveLoc);
4469   return false;
4470 }
4471 
4472 /// parseDirectiveCFIReturnColumn
4473 /// ::= .cfi_return_column register
parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc)4474 bool AsmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) {
4475   int64_t Register = 0;
4476   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4477     return true;
4478   getStreamer().emitCFIReturnColumn(Register);
4479   return false;
4480 }
4481 
4482 /// parseDirectiveCFISignalFrame
4483 /// ::= .cfi_signal_frame
parseDirectiveCFISignalFrame(SMLoc DirectiveLoc)4484 bool AsmParser::parseDirectiveCFISignalFrame(SMLoc DirectiveLoc) {
4485   if (parseEOL())
4486     return true;
4487 
4488   getStreamer().emitCFISignalFrame();
4489   return false;
4490 }
4491 
4492 /// parseDirectiveCFIUndefined
4493 /// ::= .cfi_undefined register
parseDirectiveCFIUndefined(SMLoc DirectiveLoc)4494 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) {
4495   int64_t Register = 0;
4496 
4497   if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4498     return true;
4499 
4500   getStreamer().emitCFIUndefined(Register, DirectiveLoc);
4501   return false;
4502 }
4503 
4504 /// parseDirectiveCFILabel
4505 /// ::= .cfi_label label
parseDirectiveCFILabel(SMLoc Loc)4506 bool AsmParser::parseDirectiveCFILabel(SMLoc Loc) {
4507   StringRef Name;
4508   Loc = Lexer.getLoc();
4509   if (parseIdentifier(Name))
4510     return TokError("expected identifier");
4511   if (parseEOL())
4512     return true;
4513   getStreamer().emitCFILabelDirective(Loc, Name);
4514   return false;
4515 }
4516 
4517 /// parseDirectiveAltmacro
4518 /// ::= .altmacro
4519 /// ::= .noaltmacro
parseDirectiveAltmacro(StringRef Directive)4520 bool AsmParser::parseDirectiveAltmacro(StringRef Directive) {
4521   if (parseEOL())
4522     return true;
4523   AltMacroMode = (Directive == ".altmacro");
4524   return false;
4525 }
4526 
4527 /// parseDirectiveMacrosOnOff
4528 /// ::= .macros_on
4529 /// ::= .macros_off
parseDirectiveMacrosOnOff(StringRef Directive)4530 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) {
4531   if (parseEOL())
4532     return true;
4533   setMacrosEnabled(Directive == ".macros_on");
4534   return false;
4535 }
4536 
4537 /// parseDirectiveMacro
4538 /// ::= .macro name[,] [parameters]
parseDirectiveMacro(SMLoc DirectiveLoc)4539 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) {
4540   StringRef Name;
4541   if (parseIdentifier(Name))
4542     return TokError("expected identifier in '.macro' directive");
4543 
4544   if (getLexer().is(AsmToken::Comma))
4545     Lex();
4546 
4547   MCAsmMacroParameters Parameters;
4548   while (getLexer().isNot(AsmToken::EndOfStatement)) {
4549 
4550     if (!Parameters.empty() && Parameters.back().Vararg)
4551       return Error(Lexer.getLoc(), "vararg parameter '" +
4552                                        Parameters.back().Name +
4553                                        "' should be the last parameter");
4554 
4555     MCAsmMacroParameter Parameter;
4556     if (parseIdentifier(Parameter.Name))
4557       return TokError("expected identifier in '.macro' directive");
4558 
4559     // Emit an error if two (or more) named parameters share the same name
4560     for (const MCAsmMacroParameter& CurrParam : Parameters)
4561       if (CurrParam.Name == Parameter.Name)
4562         return TokError("macro '" + Name + "' has multiple parameters"
4563                         " named '" + Parameter.Name + "'");
4564 
4565     if (Lexer.is(AsmToken::Colon)) {
4566       Lex();  // consume ':'
4567 
4568       SMLoc QualLoc;
4569       StringRef Qualifier;
4570 
4571       QualLoc = Lexer.getLoc();
4572       if (parseIdentifier(Qualifier))
4573         return Error(QualLoc, "missing parameter qualifier for "
4574                      "'" + Parameter.Name + "' in macro '" + Name + "'");
4575 
4576       if (Qualifier == "req")
4577         Parameter.Required = true;
4578       else if (Qualifier == "vararg")
4579         Parameter.Vararg = true;
4580       else
4581         return Error(QualLoc, Qualifier + " is not a valid parameter qualifier "
4582                      "for '" + Parameter.Name + "' in macro '" + Name + "'");
4583     }
4584 
4585     if (getLexer().is(AsmToken::Equal)) {
4586       Lex();
4587 
4588       SMLoc ParamLoc;
4589 
4590       ParamLoc = Lexer.getLoc();
4591       if (parseMacroArgument(Parameter.Value, /*Vararg=*/false ))
4592         return true;
4593 
4594       if (Parameter.Required)
4595         Warning(ParamLoc, "pointless default value for required parameter "
4596                 "'" + Parameter.Name + "' in macro '" + Name + "'");
4597     }
4598 
4599     Parameters.push_back(std::move(Parameter));
4600 
4601     if (getLexer().is(AsmToken::Comma))
4602       Lex();
4603   }
4604 
4605   // Eat just the end of statement.
4606   Lexer.Lex();
4607 
4608   // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors
4609   AsmToken EndToken, StartToken = getTok();
4610   unsigned MacroDepth = 0;
4611   // Lex the macro definition.
4612   while (true) {
4613     // Ignore Lexing errors in macros.
4614     while (Lexer.is(AsmToken::Error)) {
4615       Lexer.Lex();
4616     }
4617 
4618     // Check whether we have reached the end of the file.
4619     if (getLexer().is(AsmToken::Eof))
4620       return Error(DirectiveLoc, "no matching '.endmacro' in definition");
4621 
4622     // Otherwise, check whether we have reach the .endmacro or the start of a
4623     // preprocessor line marker.
4624     if (getLexer().is(AsmToken::Identifier)) {
4625       if (getTok().getIdentifier() == ".endm" ||
4626           getTok().getIdentifier() == ".endmacro") {
4627         if (MacroDepth == 0) { // Outermost macro.
4628           EndToken = getTok();
4629           Lexer.Lex();
4630           if (getLexer().isNot(AsmToken::EndOfStatement))
4631             return TokError("unexpected token in '" + EndToken.getIdentifier() +
4632                             "' directive");
4633           break;
4634         } else {
4635           // Otherwise we just found the end of an inner macro.
4636           --MacroDepth;
4637         }
4638       } else if (getTok().getIdentifier() == ".macro") {
4639         // We allow nested macros. Those aren't instantiated until the outermost
4640         // macro is expanded so just ignore them for now.
4641         ++MacroDepth;
4642       }
4643     } else if (Lexer.is(AsmToken::HashDirective)) {
4644       (void)parseCppHashLineFilenameComment(getLexer().getLoc());
4645     }
4646 
4647     // Otherwise, scan til the end of the statement.
4648     eatToEndOfStatement();
4649   }
4650 
4651   if (getContext().lookupMacro(Name)) {
4652     return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
4653   }
4654 
4655   const char *BodyStart = StartToken.getLoc().getPointer();
4656   const char *BodyEnd = EndToken.getLoc().getPointer();
4657   StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
4658   checkForBadMacro(DirectiveLoc, Name, Body, Parameters);
4659   MCAsmMacro Macro(Name, Body, std::move(Parameters));
4660   DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n";
4661                   Macro.dump());
4662   getContext().defineMacro(Name, std::move(Macro));
4663   return false;
4664 }
4665 
4666 /// checkForBadMacro
4667 ///
4668 /// With the support added for named parameters there may be code out there that
4669 /// is transitioning from positional parameters.  In versions of gas that did
4670 /// not support named parameters they would be ignored on the macro definition.
4671 /// But to support both styles of parameters this is not possible so if a macro
4672 /// definition has named parameters but does not use them and has what appears
4673 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a
4674 /// warning that the positional parameter found in body which have no effect.
4675 /// Hoping the developer will either remove the named parameters from the macro
4676 /// definition so the positional parameters get used if that was what was
4677 /// intended or change the macro to use the named parameters.  It is possible
4678 /// this warning will trigger when the none of the named parameters are used
4679 /// and the strings like $1 are infact to simply to be passed trough unchanged.
checkForBadMacro(SMLoc DirectiveLoc,StringRef Name,StringRef Body,ArrayRef<MCAsmMacroParameter> Parameters)4680 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name,
4681                                  StringRef Body,
4682                                  ArrayRef<MCAsmMacroParameter> Parameters) {
4683   // If this macro is not defined with named parameters the warning we are
4684   // checking for here doesn't apply.
4685   unsigned NParameters = Parameters.size();
4686   if (NParameters == 0)
4687     return;
4688 
4689   bool NamedParametersFound = false;
4690   bool PositionalParametersFound = false;
4691 
4692   // Look at the body of the macro for use of both the named parameters and what
4693   // are likely to be positional parameters.  This is what expandMacro() is
4694   // doing when it finds the parameters in the body.
4695   while (!Body.empty()) {
4696     // Scan for the next possible parameter.
4697     std::size_t End = Body.size(), Pos = 0;
4698     for (; Pos != End; ++Pos) {
4699       // Check for a substitution or escape.
4700       // This macro is defined with parameters, look for \foo, \bar, etc.
4701       if (Body[Pos] == '\\' && Pos + 1 != End)
4702         break;
4703 
4704       // This macro should have parameters, but look for $0, $1, ..., $n too.
4705       if (Body[Pos] != '$' || Pos + 1 == End)
4706         continue;
4707       char Next = Body[Pos + 1];
4708       if (Next == '$' || Next == 'n' ||
4709           isdigit(static_cast<unsigned char>(Next)))
4710         break;
4711     }
4712 
4713     // Check if we reached the end.
4714     if (Pos == End)
4715       break;
4716 
4717     if (Body[Pos] == '$') {
4718       switch (Body[Pos + 1]) {
4719       // $$ => $
4720       case '$':
4721         break;
4722 
4723       // $n => number of arguments
4724       case 'n':
4725         PositionalParametersFound = true;
4726         break;
4727 
4728       // $[0-9] => argument
4729       default: {
4730         PositionalParametersFound = true;
4731         break;
4732       }
4733       }
4734       Pos += 2;
4735     } else {
4736       unsigned I = Pos + 1;
4737       while (isIdentifierChar(Body[I]) && I + 1 != End)
4738         ++I;
4739 
4740       const char *Begin = Body.data() + Pos + 1;
4741       StringRef Argument(Begin, I - (Pos + 1));
4742       unsigned Index = 0;
4743       for (; Index < NParameters; ++Index)
4744         if (Parameters[Index].Name == Argument)
4745           break;
4746 
4747       if (Index == NParameters) {
4748         if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
4749           Pos += 3;
4750         else {
4751           Pos = I;
4752         }
4753       } else {
4754         NamedParametersFound = true;
4755         Pos += 1 + Argument.size();
4756       }
4757     }
4758     // Update the scan point.
4759     Body = Body.substr(Pos);
4760   }
4761 
4762   if (!NamedParametersFound && PositionalParametersFound)
4763     Warning(DirectiveLoc, "macro defined with named parameters which are not "
4764                           "used in macro body, possible positional parameter "
4765                           "found in body which will have no effect");
4766 }
4767 
4768 /// parseDirectiveExitMacro
4769 /// ::= .exitm
parseDirectiveExitMacro(StringRef Directive)4770 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) {
4771   if (parseEOL())
4772     return true;
4773 
4774   if (!isInsideMacroInstantiation())
4775     return TokError("unexpected '" + Directive + "' in file, "
4776                                                  "no current macro definition");
4777 
4778   // Exit all conditionals that are active in the current macro.
4779   while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) {
4780     TheCondState = TheCondStack.back();
4781     TheCondStack.pop_back();
4782   }
4783 
4784   handleMacroExit();
4785   return false;
4786 }
4787 
4788 /// parseDirectiveEndMacro
4789 /// ::= .endm
4790 /// ::= .endmacro
parseDirectiveEndMacro(StringRef Directive)4791 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) {
4792   if (getLexer().isNot(AsmToken::EndOfStatement))
4793     return TokError("unexpected token in '" + Directive + "' directive");
4794 
4795   // If we are inside a macro instantiation, terminate the current
4796   // instantiation.
4797   if (isInsideMacroInstantiation()) {
4798     handleMacroExit();
4799     return false;
4800   }
4801 
4802   // Otherwise, this .endmacro is a stray entry in the file; well formed
4803   // .endmacro directives are handled during the macro definition parsing.
4804   return TokError("unexpected '" + Directive + "' in file, "
4805                                                "no current macro definition");
4806 }
4807 
4808 /// parseDirectivePurgeMacro
4809 /// ::= .purgem name
parseDirectivePurgeMacro(SMLoc DirectiveLoc)4810 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) {
4811   StringRef Name;
4812   SMLoc Loc;
4813   if (parseTokenLoc(Loc) ||
4814       check(parseIdentifier(Name), Loc,
4815             "expected identifier in '.purgem' directive") ||
4816       parseEOL())
4817     return true;
4818 
4819   if (!getContext().lookupMacro(Name))
4820     return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
4821 
4822   getContext().undefineMacro(Name);
4823   DEBUG_WITH_TYPE("asm-macros", dbgs()
4824                                     << "Un-defining macro: " << Name << "\n");
4825   return false;
4826 }
4827 
4828 /// parseDirectiveBundleAlignMode
4829 /// ::= {.bundle_align_mode} expression
parseDirectiveBundleAlignMode()4830 bool AsmParser::parseDirectiveBundleAlignMode() {
4831   // Expect a single argument: an expression that evaluates to a constant
4832   // in the inclusive range 0-30.
4833   SMLoc ExprLoc = getLexer().getLoc();
4834   int64_t AlignSizePow2;
4835   if (checkForValidSection() || parseAbsoluteExpression(AlignSizePow2) ||
4836       parseEOL() ||
4837       check(AlignSizePow2 < 0 || AlignSizePow2 > 30, ExprLoc,
4838             "invalid bundle alignment size (expected between 0 and 30)"))
4839     return true;
4840 
4841   getStreamer().emitBundleAlignMode(Align(1ULL << AlignSizePow2));
4842   return false;
4843 }
4844 
4845 /// parseDirectiveBundleLock
4846 /// ::= {.bundle_lock} [align_to_end]
parseDirectiveBundleLock()4847 bool AsmParser::parseDirectiveBundleLock() {
4848   if (checkForValidSection())
4849     return true;
4850   bool AlignToEnd = false;
4851 
4852   StringRef Option;
4853   SMLoc Loc = getTok().getLoc();
4854   const char *kInvalidOptionError =
4855       "invalid option for '.bundle_lock' directive";
4856 
4857   if (!parseOptionalToken(AsmToken::EndOfStatement)) {
4858     if (check(parseIdentifier(Option), Loc, kInvalidOptionError) ||
4859         check(Option != "align_to_end", Loc, kInvalidOptionError) || parseEOL())
4860       return true;
4861     AlignToEnd = true;
4862   }
4863 
4864   getStreamer().emitBundleLock(AlignToEnd);
4865   return false;
4866 }
4867 
4868 /// parseDirectiveBundleLock
4869 /// ::= {.bundle_lock}
parseDirectiveBundleUnlock()4870 bool AsmParser::parseDirectiveBundleUnlock() {
4871   if (checkForValidSection() || parseEOL())
4872     return true;
4873 
4874   getStreamer().emitBundleUnlock();
4875   return false;
4876 }
4877 
4878 /// parseDirectiveSpace
4879 /// ::= (.skip | .space) expression [ , expression ]
parseDirectiveSpace(StringRef IDVal)4880 bool AsmParser::parseDirectiveSpace(StringRef IDVal) {
4881   SMLoc NumBytesLoc = Lexer.getLoc();
4882   const MCExpr *NumBytes;
4883   if (checkForValidSection() || parseExpression(NumBytes))
4884     return true;
4885 
4886   int64_t FillExpr = 0;
4887   if (parseOptionalToken(AsmToken::Comma))
4888     if (parseAbsoluteExpression(FillExpr))
4889       return true;
4890   if (parseEOL())
4891     return true;
4892 
4893   // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
4894   getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc);
4895 
4896   return false;
4897 }
4898 
4899 /// parseDirectiveDCB
4900 /// ::= .dcb.{b, l, w} expression, expression
parseDirectiveDCB(StringRef IDVal,unsigned Size)4901 bool AsmParser::parseDirectiveDCB(StringRef IDVal, unsigned Size) {
4902   SMLoc NumValuesLoc = Lexer.getLoc();
4903   int64_t NumValues;
4904   if (checkForValidSection() || parseAbsoluteExpression(NumValues))
4905     return true;
4906 
4907   if (NumValues < 0) {
4908     Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect");
4909     return false;
4910   }
4911 
4912   if (parseComma())
4913     return true;
4914 
4915   const MCExpr *Value;
4916   SMLoc ExprLoc = getLexer().getLoc();
4917   if (parseExpression(Value))
4918     return true;
4919 
4920   // Special case constant expressions to match code generator.
4921   if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
4922     assert(Size <= 8 && "Invalid size");
4923     uint64_t IntValue = MCE->getValue();
4924     if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
4925       return Error(ExprLoc, "literal value out of range for directive");
4926     for (uint64_t i = 0, e = NumValues; i != e; ++i)
4927       getStreamer().emitIntValue(IntValue, Size);
4928   } else {
4929     for (uint64_t i = 0, e = NumValues; i != e; ++i)
4930       getStreamer().emitValue(Value, Size, ExprLoc);
4931   }
4932 
4933   return parseEOL();
4934 }
4935 
4936 /// parseDirectiveRealDCB
4937 /// ::= .dcb.{d, s} expression, expression
parseDirectiveRealDCB(StringRef IDVal,const fltSemantics & Semantics)4938 bool AsmParser::parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &Semantics) {
4939   SMLoc NumValuesLoc = Lexer.getLoc();
4940   int64_t NumValues;
4941   if (checkForValidSection() || parseAbsoluteExpression(NumValues))
4942     return true;
4943 
4944   if (NumValues < 0) {
4945     Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect");
4946     return false;
4947   }
4948 
4949   if (parseComma())
4950     return true;
4951 
4952   APInt AsInt;
4953   if (parseRealValue(Semantics, AsInt) || parseEOL())
4954     return true;
4955 
4956   for (uint64_t i = 0, e = NumValues; i != e; ++i)
4957     getStreamer().emitIntValue(AsInt.getLimitedValue(),
4958                                AsInt.getBitWidth() / 8);
4959 
4960   return false;
4961 }
4962 
4963 /// parseDirectiveDS
4964 /// ::= .ds.{b, d, l, p, s, w, x} expression
parseDirectiveDS(StringRef IDVal,unsigned Size)4965 bool AsmParser::parseDirectiveDS(StringRef IDVal, unsigned Size) {
4966   SMLoc NumValuesLoc = Lexer.getLoc();
4967   int64_t NumValues;
4968   if (checkForValidSection() || parseAbsoluteExpression(NumValues) ||
4969       parseEOL())
4970     return true;
4971 
4972   if (NumValues < 0) {
4973     Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect");
4974     return false;
4975   }
4976 
4977   for (uint64_t i = 0, e = NumValues; i != e; ++i)
4978     getStreamer().emitFill(Size, 0);
4979 
4980   return false;
4981 }
4982 
4983 /// parseDirectiveLEB128
4984 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ]
parseDirectiveLEB128(bool Signed)4985 bool AsmParser::parseDirectiveLEB128(bool Signed) {
4986   if (checkForValidSection())
4987     return true;
4988 
4989   auto parseOp = [&]() -> bool {
4990     const MCExpr *Value;
4991     if (parseExpression(Value))
4992       return true;
4993     if (Signed)
4994       getStreamer().emitSLEB128Value(Value);
4995     else
4996       getStreamer().emitULEB128Value(Value);
4997     return false;
4998   };
4999 
5000   return parseMany(parseOp);
5001 }
5002 
5003 /// parseDirectiveSymbolAttribute
5004 ///  ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
parseDirectiveSymbolAttribute(MCSymbolAttr Attr)5005 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
5006   auto parseOp = [&]() -> bool {
5007     StringRef Name;
5008     SMLoc Loc = getTok().getLoc();
5009     if (parseIdentifier(Name))
5010       return Error(Loc, "expected identifier");
5011 
5012     if (discardLTOSymbol(Name))
5013       return false;
5014 
5015     MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
5016 
5017     // Assembler local symbols don't make any sense here, except for directives
5018     // that the symbol should be tagged.
5019     if (Sym->isTemporary() && Attr != MCSA_Memtag)
5020       return Error(Loc, "non-local symbol required");
5021 
5022     if (!getStreamer().emitSymbolAttribute(Sym, Attr))
5023       return Error(Loc, "unable to emit symbol attribute");
5024     return false;
5025   };
5026 
5027   return parseMany(parseOp);
5028 }
5029 
5030 /// parseDirectiveComm
5031 ///  ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
parseDirectiveComm(bool IsLocal)5032 bool AsmParser::parseDirectiveComm(bool IsLocal) {
5033   if (checkForValidSection())
5034     return true;
5035 
5036   SMLoc IDLoc = getLexer().getLoc();
5037   StringRef Name;
5038   if (parseIdentifier(Name))
5039     return TokError("expected identifier in directive");
5040 
5041   // Handle the identifier as the key symbol.
5042   MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
5043 
5044   if (parseComma())
5045     return true;
5046 
5047   int64_t Size;
5048   SMLoc SizeLoc = getLexer().getLoc();
5049   if (parseAbsoluteExpression(Size))
5050     return true;
5051 
5052   int64_t Pow2Alignment = 0;
5053   SMLoc Pow2AlignmentLoc;
5054   if (getLexer().is(AsmToken::Comma)) {
5055     Lex();
5056     Pow2AlignmentLoc = getLexer().getLoc();
5057     if (parseAbsoluteExpression(Pow2Alignment))
5058       return true;
5059 
5060     LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType();
5061     if (IsLocal && LCOMM == LCOMM::NoAlignment)
5062       return Error(Pow2AlignmentLoc, "alignment not supported on this target");
5063 
5064     // If this target takes alignments in bytes (not log) validate and convert.
5065     if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) ||
5066         (IsLocal && LCOMM == LCOMM::ByteAlignment)) {
5067       if (!isPowerOf2_64(Pow2Alignment))
5068         return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
5069       Pow2Alignment = Log2_64(Pow2Alignment);
5070     }
5071   }
5072 
5073   if (parseEOL())
5074     return true;
5075 
5076   // NOTE: a size of zero for a .comm should create a undefined symbol
5077   // but a size of .lcomm creates a bss symbol of size zero.
5078   if (Size < 0)
5079     return Error(SizeLoc, "size must be non-negative");
5080 
5081   Sym->redefineIfPossible();
5082   if (!Sym->isUndefined())
5083     return Error(IDLoc, "invalid symbol redefinition");
5084 
5085   // Create the Symbol as a common or local common with Size and Pow2Alignment
5086   if (IsLocal) {
5087     getStreamer().emitLocalCommonSymbol(Sym, Size,
5088                                         Align(1ULL << Pow2Alignment));
5089     return false;
5090   }
5091 
5092   getStreamer().emitCommonSymbol(Sym, Size, Align(1ULL << Pow2Alignment));
5093   return false;
5094 }
5095 
5096 /// parseDirectiveAbort
5097 ///  ::= .abort [... message ...]
parseDirectiveAbort(SMLoc DirectiveLoc)5098 bool AsmParser::parseDirectiveAbort(SMLoc DirectiveLoc) {
5099   StringRef Str = parseStringToEndOfStatement();
5100   if (parseEOL())
5101     return true;
5102 
5103   if (Str.empty())
5104     return Error(DirectiveLoc, ".abort detected. Assembly stopping");
5105 
5106   // FIXME: Actually abort assembly here.
5107   return Error(DirectiveLoc,
5108                ".abort '" + Str + "' detected. Assembly stopping");
5109 }
5110 
5111 /// parseDirectiveInclude
5112 ///  ::= .include "filename"
parseDirectiveInclude()5113 bool AsmParser::parseDirectiveInclude() {
5114   // Allow the strings to have escaped octal character sequence.
5115   std::string Filename;
5116   SMLoc IncludeLoc = getTok().getLoc();
5117 
5118   if (check(getTok().isNot(AsmToken::String),
5119             "expected string in '.include' directive") ||
5120       parseEscapedString(Filename) ||
5121       check(getTok().isNot(AsmToken::EndOfStatement),
5122             "unexpected token in '.include' directive") ||
5123       // Attempt to switch the lexer to the included file before consuming the
5124       // end of statement to avoid losing it when we switch.
5125       check(enterIncludeFile(Filename), IncludeLoc,
5126             "Could not find include file '" + Filename + "'"))
5127     return true;
5128 
5129   return false;
5130 }
5131 
5132 /// parseDirectiveIncbin
5133 ///  ::= .incbin "filename" [ , skip [ , count ] ]
parseDirectiveIncbin()5134 bool AsmParser::parseDirectiveIncbin() {
5135   // Allow the strings to have escaped octal character sequence.
5136   std::string Filename;
5137   SMLoc IncbinLoc = getTok().getLoc();
5138   if (check(getTok().isNot(AsmToken::String),
5139             "expected string in '.incbin' directive") ||
5140       parseEscapedString(Filename))
5141     return true;
5142 
5143   int64_t Skip = 0;
5144   const MCExpr *Count = nullptr;
5145   SMLoc SkipLoc, CountLoc;
5146   if (parseOptionalToken(AsmToken::Comma)) {
5147     // The skip expression can be omitted while specifying the count, e.g:
5148     //  .incbin "filename",,4
5149     if (getTok().isNot(AsmToken::Comma)) {
5150       if (parseTokenLoc(SkipLoc) || parseAbsoluteExpression(Skip))
5151         return true;
5152     }
5153     if (parseOptionalToken(AsmToken::Comma)) {
5154       CountLoc = getTok().getLoc();
5155       if (parseExpression(Count))
5156         return true;
5157     }
5158   }
5159 
5160   if (parseEOL())
5161     return true;
5162 
5163   if (check(Skip < 0, SkipLoc, "skip is negative"))
5164     return true;
5165 
5166   // Attempt to process the included file.
5167   if (processIncbinFile(Filename, Skip, Count, CountLoc))
5168     return Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
5169   return false;
5170 }
5171 
5172 /// parseDirectiveIf
5173 /// ::= .if{,eq,ge,gt,le,lt,ne} expression
parseDirectiveIf(SMLoc DirectiveLoc,DirectiveKind DirKind)5174 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) {
5175   TheCondStack.push_back(TheCondState);
5176   TheCondState.TheCond = AsmCond::IfCond;
5177   if (TheCondState.Ignore) {
5178     eatToEndOfStatement();
5179   } else {
5180     int64_t ExprValue;
5181     if (parseAbsoluteExpression(ExprValue) || parseEOL())
5182       return true;
5183 
5184     switch (DirKind) {
5185     default:
5186       llvm_unreachable("unsupported directive");
5187     case DK_IF:
5188     case DK_IFNE:
5189       break;
5190     case DK_IFEQ:
5191       ExprValue = ExprValue == 0;
5192       break;
5193     case DK_IFGE:
5194       ExprValue = ExprValue >= 0;
5195       break;
5196     case DK_IFGT:
5197       ExprValue = ExprValue > 0;
5198       break;
5199     case DK_IFLE:
5200       ExprValue = ExprValue <= 0;
5201       break;
5202     case DK_IFLT:
5203       ExprValue = ExprValue < 0;
5204       break;
5205     }
5206 
5207     TheCondState.CondMet = ExprValue;
5208     TheCondState.Ignore = !TheCondState.CondMet;
5209   }
5210 
5211   return false;
5212 }
5213 
5214 /// parseDirectiveIfb
5215 /// ::= .ifb string
parseDirectiveIfb(SMLoc DirectiveLoc,bool ExpectBlank)5216 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
5217   TheCondStack.push_back(TheCondState);
5218   TheCondState.TheCond = AsmCond::IfCond;
5219 
5220   if (TheCondState.Ignore) {
5221     eatToEndOfStatement();
5222   } else {
5223     StringRef Str = parseStringToEndOfStatement();
5224 
5225     if (parseEOL())
5226       return true;
5227 
5228     TheCondState.CondMet = ExpectBlank == Str.empty();
5229     TheCondState.Ignore = !TheCondState.CondMet;
5230   }
5231 
5232   return false;
5233 }
5234 
5235 /// parseDirectiveIfc
5236 /// ::= .ifc string1, string2
5237 /// ::= .ifnc string1, string2
parseDirectiveIfc(SMLoc DirectiveLoc,bool ExpectEqual)5238 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
5239   TheCondStack.push_back(TheCondState);
5240   TheCondState.TheCond = AsmCond::IfCond;
5241 
5242   if (TheCondState.Ignore) {
5243     eatToEndOfStatement();
5244   } else {
5245     StringRef Str1 = parseStringToComma();
5246 
5247     if (parseComma())
5248       return true;
5249 
5250     StringRef Str2 = parseStringToEndOfStatement();
5251 
5252     if (parseEOL())
5253       return true;
5254 
5255     TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim());
5256     TheCondState.Ignore = !TheCondState.CondMet;
5257   }
5258 
5259   return false;
5260 }
5261 
5262 /// parseDirectiveIfeqs
5263 ///   ::= .ifeqs string1, string2
parseDirectiveIfeqs(SMLoc DirectiveLoc,bool ExpectEqual)5264 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) {
5265   if (Lexer.isNot(AsmToken::String)) {
5266     if (ExpectEqual)
5267       return TokError("expected string parameter for '.ifeqs' directive");
5268     return TokError("expected string parameter for '.ifnes' directive");
5269   }
5270 
5271   StringRef String1 = getTok().getStringContents();
5272   Lex();
5273 
5274   if (Lexer.isNot(AsmToken::Comma)) {
5275     if (ExpectEqual)
5276       return TokError(
5277           "expected comma after first string for '.ifeqs' directive");
5278     return TokError("expected comma after first string for '.ifnes' directive");
5279   }
5280 
5281   Lex();
5282 
5283   if (Lexer.isNot(AsmToken::String)) {
5284     if (ExpectEqual)
5285       return TokError("expected string parameter for '.ifeqs' directive");
5286     return TokError("expected string parameter for '.ifnes' directive");
5287   }
5288 
5289   StringRef String2 = getTok().getStringContents();
5290   Lex();
5291 
5292   TheCondStack.push_back(TheCondState);
5293   TheCondState.TheCond = AsmCond::IfCond;
5294   TheCondState.CondMet = ExpectEqual == (String1 == String2);
5295   TheCondState.Ignore = !TheCondState.CondMet;
5296 
5297   return false;
5298 }
5299 
5300 /// parseDirectiveIfdef
5301 /// ::= .ifdef symbol
parseDirectiveIfdef(SMLoc DirectiveLoc,bool expect_defined)5302 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
5303   StringRef Name;
5304   TheCondStack.push_back(TheCondState);
5305   TheCondState.TheCond = AsmCond::IfCond;
5306 
5307   if (TheCondState.Ignore) {
5308     eatToEndOfStatement();
5309   } else {
5310     if (check(parseIdentifier(Name), "expected identifier after '.ifdef'") ||
5311         parseEOL())
5312       return true;
5313 
5314     MCSymbol *Sym = getContext().lookupSymbol(Name);
5315 
5316     if (expect_defined)
5317       TheCondState.CondMet = (Sym && !Sym->isUndefined(false));
5318     else
5319       TheCondState.CondMet = (!Sym || Sym->isUndefined(false));
5320     TheCondState.Ignore = !TheCondState.CondMet;
5321   }
5322 
5323   return false;
5324 }
5325 
5326 /// parseDirectiveElseIf
5327 /// ::= .elseif expression
parseDirectiveElseIf(SMLoc DirectiveLoc)5328 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) {
5329   if (TheCondState.TheCond != AsmCond::IfCond &&
5330       TheCondState.TheCond != AsmCond::ElseIfCond)
5331     return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an"
5332                                " .if or  an .elseif");
5333   TheCondState.TheCond = AsmCond::ElseIfCond;
5334 
5335   bool LastIgnoreState = false;
5336   if (!TheCondStack.empty())
5337     LastIgnoreState = TheCondStack.back().Ignore;
5338   if (LastIgnoreState || TheCondState.CondMet) {
5339     TheCondState.Ignore = true;
5340     eatToEndOfStatement();
5341   } else {
5342     int64_t ExprValue;
5343     if (parseAbsoluteExpression(ExprValue))
5344       return true;
5345 
5346     if (parseEOL())
5347       return true;
5348 
5349     TheCondState.CondMet = ExprValue;
5350     TheCondState.Ignore = !TheCondState.CondMet;
5351   }
5352 
5353   return false;
5354 }
5355 
5356 /// parseDirectiveElse
5357 /// ::= .else
parseDirectiveElse(SMLoc DirectiveLoc)5358 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) {
5359   if (parseEOL())
5360     return true;
5361 
5362   if (TheCondState.TheCond != AsmCond::IfCond &&
5363       TheCondState.TheCond != AsmCond::ElseIfCond)
5364     return Error(DirectiveLoc, "Encountered a .else that doesn't follow "
5365                                " an .if or an .elseif");
5366   TheCondState.TheCond = AsmCond::ElseCond;
5367   bool LastIgnoreState = false;
5368   if (!TheCondStack.empty())
5369     LastIgnoreState = TheCondStack.back().Ignore;
5370   if (LastIgnoreState || TheCondState.CondMet)
5371     TheCondState.Ignore = true;
5372   else
5373     TheCondState.Ignore = false;
5374 
5375   return false;
5376 }
5377 
5378 /// parseDirectiveEnd
5379 /// ::= .end
parseDirectiveEnd(SMLoc DirectiveLoc)5380 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) {
5381   if (parseEOL())
5382     return true;
5383 
5384   while (Lexer.isNot(AsmToken::Eof))
5385     Lexer.Lex();
5386 
5387   return false;
5388 }
5389 
5390 /// parseDirectiveError
5391 ///   ::= .err
5392 ///   ::= .error [string]
parseDirectiveError(SMLoc L,bool WithMessage)5393 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) {
5394   if (!TheCondStack.empty()) {
5395     if (TheCondStack.back().Ignore) {
5396       eatToEndOfStatement();
5397       return false;
5398     }
5399   }
5400 
5401   if (!WithMessage)
5402     return Error(L, ".err encountered");
5403 
5404   StringRef Message = ".error directive invoked in source file";
5405   if (Lexer.isNot(AsmToken::EndOfStatement)) {
5406     if (Lexer.isNot(AsmToken::String))
5407       return TokError(".error argument must be a string");
5408 
5409     Message = getTok().getStringContents();
5410     Lex();
5411   }
5412 
5413   return Error(L, Message);
5414 }
5415 
5416 /// parseDirectiveWarning
5417 ///   ::= .warning [string]
parseDirectiveWarning(SMLoc L)5418 bool AsmParser::parseDirectiveWarning(SMLoc L) {
5419   if (!TheCondStack.empty()) {
5420     if (TheCondStack.back().Ignore) {
5421       eatToEndOfStatement();
5422       return false;
5423     }
5424   }
5425 
5426   StringRef Message = ".warning directive invoked in source file";
5427 
5428   if (!parseOptionalToken(AsmToken::EndOfStatement)) {
5429     if (Lexer.isNot(AsmToken::String))
5430       return TokError(".warning argument must be a string");
5431 
5432     Message = getTok().getStringContents();
5433     Lex();
5434     if (parseEOL())
5435       return true;
5436   }
5437 
5438   return Warning(L, Message);
5439 }
5440 
5441 /// parseDirectiveEndIf
5442 /// ::= .endif
parseDirectiveEndIf(SMLoc DirectiveLoc)5443 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) {
5444   if (parseEOL())
5445     return true;
5446 
5447   if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty())
5448     return Error(DirectiveLoc, "Encountered a .endif that doesn't follow "
5449                                "an .if or .else");
5450   if (!TheCondStack.empty()) {
5451     TheCondState = TheCondStack.back();
5452     TheCondStack.pop_back();
5453   }
5454 
5455   return false;
5456 }
5457 
initializeDirectiveKindMap()5458 void AsmParser::initializeDirectiveKindMap() {
5459   /* Lookup will be done with the directive
5460    * converted to lower case, so all these
5461    * keys should be lower case.
5462    * (target specific directives are handled
5463    *  elsewhere)
5464    */
5465   DirectiveKindMap[".set"] = DK_SET;
5466   DirectiveKindMap[".equ"] = DK_EQU;
5467   DirectiveKindMap[".equiv"] = DK_EQUIV;
5468   DirectiveKindMap[".ascii"] = DK_ASCII;
5469   DirectiveKindMap[".asciz"] = DK_ASCIZ;
5470   DirectiveKindMap[".string"] = DK_STRING;
5471   DirectiveKindMap[".byte"] = DK_BYTE;
5472   DirectiveKindMap[".short"] = DK_SHORT;
5473   DirectiveKindMap[".value"] = DK_VALUE;
5474   DirectiveKindMap[".2byte"] = DK_2BYTE;
5475   DirectiveKindMap[".long"] = DK_LONG;
5476   DirectiveKindMap[".int"] = DK_INT;
5477   DirectiveKindMap[".4byte"] = DK_4BYTE;
5478   DirectiveKindMap[".quad"] = DK_QUAD;
5479   DirectiveKindMap[".8byte"] = DK_8BYTE;
5480   DirectiveKindMap[".octa"] = DK_OCTA;
5481   DirectiveKindMap[".single"] = DK_SINGLE;
5482   DirectiveKindMap[".float"] = DK_FLOAT;
5483   DirectiveKindMap[".double"] = DK_DOUBLE;
5484   DirectiveKindMap[".align"] = DK_ALIGN;
5485   DirectiveKindMap[".align32"] = DK_ALIGN32;
5486   DirectiveKindMap[".balign"] = DK_BALIGN;
5487   DirectiveKindMap[".balignw"] = DK_BALIGNW;
5488   DirectiveKindMap[".balignl"] = DK_BALIGNL;
5489   DirectiveKindMap[".p2align"] = DK_P2ALIGN;
5490   DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW;
5491   DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL;
5492   DirectiveKindMap[".org"] = DK_ORG;
5493   DirectiveKindMap[".fill"] = DK_FILL;
5494   DirectiveKindMap[".zero"] = DK_ZERO;
5495   DirectiveKindMap[".extern"] = DK_EXTERN;
5496   DirectiveKindMap[".globl"] = DK_GLOBL;
5497   DirectiveKindMap[".global"] = DK_GLOBAL;
5498   DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE;
5499   DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP;
5500   DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER;
5501   DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN;
5502   DirectiveKindMap[".reference"] = DK_REFERENCE;
5503   DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION;
5504   DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE;
5505   DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN;
5506   DirectiveKindMap[".cold"] = DK_COLD;
5507   DirectiveKindMap[".comm"] = DK_COMM;
5508   DirectiveKindMap[".common"] = DK_COMMON;
5509   DirectiveKindMap[".lcomm"] = DK_LCOMM;
5510   DirectiveKindMap[".abort"] = DK_ABORT;
5511   DirectiveKindMap[".include"] = DK_INCLUDE;
5512   DirectiveKindMap[".incbin"] = DK_INCBIN;
5513   DirectiveKindMap[".code16"] = DK_CODE16;
5514   DirectiveKindMap[".code16gcc"] = DK_CODE16GCC;
5515   DirectiveKindMap[".rept"] = DK_REPT;
5516   DirectiveKindMap[".rep"] = DK_REPT;
5517   DirectiveKindMap[".irp"] = DK_IRP;
5518   DirectiveKindMap[".irpc"] = DK_IRPC;
5519   DirectiveKindMap[".endr"] = DK_ENDR;
5520   DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE;
5521   DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK;
5522   DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK;
5523   DirectiveKindMap[".if"] = DK_IF;
5524   DirectiveKindMap[".ifeq"] = DK_IFEQ;
5525   DirectiveKindMap[".ifge"] = DK_IFGE;
5526   DirectiveKindMap[".ifgt"] = DK_IFGT;
5527   DirectiveKindMap[".ifle"] = DK_IFLE;
5528   DirectiveKindMap[".iflt"] = DK_IFLT;
5529   DirectiveKindMap[".ifne"] = DK_IFNE;
5530   DirectiveKindMap[".ifb"] = DK_IFB;
5531   DirectiveKindMap[".ifnb"] = DK_IFNB;
5532   DirectiveKindMap[".ifc"] = DK_IFC;
5533   DirectiveKindMap[".ifeqs"] = DK_IFEQS;
5534   DirectiveKindMap[".ifnc"] = DK_IFNC;
5535   DirectiveKindMap[".ifnes"] = DK_IFNES;
5536   DirectiveKindMap[".ifdef"] = DK_IFDEF;
5537   DirectiveKindMap[".ifndef"] = DK_IFNDEF;
5538   DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF;
5539   DirectiveKindMap[".elseif"] = DK_ELSEIF;
5540   DirectiveKindMap[".else"] = DK_ELSE;
5541   DirectiveKindMap[".end"] = DK_END;
5542   DirectiveKindMap[".endif"] = DK_ENDIF;
5543   DirectiveKindMap[".skip"] = DK_SKIP;
5544   DirectiveKindMap[".space"] = DK_SPACE;
5545   DirectiveKindMap[".file"] = DK_FILE;
5546   DirectiveKindMap[".line"] = DK_LINE;
5547   DirectiveKindMap[".loc"] = DK_LOC;
5548   DirectiveKindMap[".stabs"] = DK_STABS;
5549   DirectiveKindMap[".cv_file"] = DK_CV_FILE;
5550   DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID;
5551   DirectiveKindMap[".cv_loc"] = DK_CV_LOC;
5552   DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE;
5553   DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE;
5554   DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID;
5555   DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE;
5556   DirectiveKindMap[".cv_string"] = DK_CV_STRING;
5557   DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE;
5558   DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS;
5559   DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET;
5560   DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA;
5561   DirectiveKindMap[".sleb128"] = DK_SLEB128;
5562   DirectiveKindMap[".uleb128"] = DK_ULEB128;
5563   DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS;
5564   DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC;
5565   DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC;
5566   DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA;
5567   DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET;
5568   DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET;
5569   DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER;
5570   DirectiveKindMap[".cfi_llvm_def_aspace_cfa"] = DK_CFI_LLVM_DEF_ASPACE_CFA;
5571   DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET;
5572   DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET;
5573   DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY;
5574   DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA;
5575   DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE;
5576   DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE;
5577   DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE;
5578   DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE;
5579   DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE;
5580   DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN;
5581   DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME;
5582   DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED;
5583   DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER;
5584   DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE;
5585   DirectiveKindMap[".cfi_label"] = DK_CFI_LABEL;
5586   DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME;
5587   DirectiveKindMap[".cfi_mte_tagged_frame"] = DK_CFI_MTE_TAGGED_FRAME;
5588   DirectiveKindMap[".macros_on"] = DK_MACROS_ON;
5589   DirectiveKindMap[".macros_off"] = DK_MACROS_OFF;
5590   DirectiveKindMap[".macro"] = DK_MACRO;
5591   DirectiveKindMap[".exitm"] = DK_EXITM;
5592   DirectiveKindMap[".endm"] = DK_ENDM;
5593   DirectiveKindMap[".endmacro"] = DK_ENDMACRO;
5594   DirectiveKindMap[".purgem"] = DK_PURGEM;
5595   DirectiveKindMap[".err"] = DK_ERR;
5596   DirectiveKindMap[".error"] = DK_ERROR;
5597   DirectiveKindMap[".warning"] = DK_WARNING;
5598   DirectiveKindMap[".altmacro"] = DK_ALTMACRO;
5599   DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO;
5600   DirectiveKindMap[".reloc"] = DK_RELOC;
5601   DirectiveKindMap[".dc"] = DK_DC;
5602   DirectiveKindMap[".dc.a"] = DK_DC_A;
5603   DirectiveKindMap[".dc.b"] = DK_DC_B;
5604   DirectiveKindMap[".dc.d"] = DK_DC_D;
5605   DirectiveKindMap[".dc.l"] = DK_DC_L;
5606   DirectiveKindMap[".dc.s"] = DK_DC_S;
5607   DirectiveKindMap[".dc.w"] = DK_DC_W;
5608   DirectiveKindMap[".dc.x"] = DK_DC_X;
5609   DirectiveKindMap[".dcb"] = DK_DCB;
5610   DirectiveKindMap[".dcb.b"] = DK_DCB_B;
5611   DirectiveKindMap[".dcb.d"] = DK_DCB_D;
5612   DirectiveKindMap[".dcb.l"] = DK_DCB_L;
5613   DirectiveKindMap[".dcb.s"] = DK_DCB_S;
5614   DirectiveKindMap[".dcb.w"] = DK_DCB_W;
5615   DirectiveKindMap[".dcb.x"] = DK_DCB_X;
5616   DirectiveKindMap[".ds"] = DK_DS;
5617   DirectiveKindMap[".ds.b"] = DK_DS_B;
5618   DirectiveKindMap[".ds.d"] = DK_DS_D;
5619   DirectiveKindMap[".ds.l"] = DK_DS_L;
5620   DirectiveKindMap[".ds.p"] = DK_DS_P;
5621   DirectiveKindMap[".ds.s"] = DK_DS_S;
5622   DirectiveKindMap[".ds.w"] = DK_DS_W;
5623   DirectiveKindMap[".ds.x"] = DK_DS_X;
5624   DirectiveKindMap[".print"] = DK_PRINT;
5625   DirectiveKindMap[".addrsig"] = DK_ADDRSIG;
5626   DirectiveKindMap[".addrsig_sym"] = DK_ADDRSIG_SYM;
5627   DirectiveKindMap[".pseudoprobe"] = DK_PSEUDO_PROBE;
5628   DirectiveKindMap[".lto_discard"] = DK_LTO_DISCARD;
5629   DirectiveKindMap[".lto_set_conditional"] = DK_LTO_SET_CONDITIONAL;
5630   DirectiveKindMap[".memtag"] = DK_MEMTAG;
5631 }
5632 
parseMacroLikeBody(SMLoc DirectiveLoc)5633 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) {
5634   AsmToken EndToken, StartToken = getTok();
5635 
5636   unsigned NestLevel = 0;
5637   while (true) {
5638     // Check whether we have reached the end of the file.
5639     if (getLexer().is(AsmToken::Eof)) {
5640       printError(DirectiveLoc, "no matching '.endr' in definition");
5641       return nullptr;
5642     }
5643 
5644     if (Lexer.is(AsmToken::Identifier)) {
5645       StringRef Ident = getTok().getIdentifier();
5646       if (Ident == ".rep" || Ident == ".rept" || Ident == ".irp" ||
5647           Ident == ".irpc") {
5648         ++NestLevel;
5649       } else if (Ident == ".endr") {
5650         if (NestLevel == 0) {
5651           EndToken = getTok();
5652           Lex();
5653           if (Lexer.is(AsmToken::EndOfStatement))
5654             break;
5655           printError(getTok().getLoc(), "expected newline");
5656           return nullptr;
5657         }
5658         --NestLevel;
5659       }
5660     }
5661 
5662     // Otherwise, scan till the end of the statement.
5663     eatToEndOfStatement();
5664   }
5665 
5666   const char *BodyStart = StartToken.getLoc().getPointer();
5667   const char *BodyEnd = EndToken.getLoc().getPointer();
5668   StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
5669 
5670   // We Are Anonymous.
5671   MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters());
5672   return &MacroLikeBodies.back();
5673 }
5674 
instantiateMacroLikeBody(MCAsmMacro * M,SMLoc DirectiveLoc,raw_svector_ostream & OS)5675 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
5676                                          raw_svector_ostream &OS) {
5677   OS << ".endr\n";
5678 
5679   std::unique_ptr<MemoryBuffer> Instantiation =
5680       MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
5681 
5682   // Create the macro instantiation object and add to the current macro
5683   // instantiation stack.
5684   MacroInstantiation *MI = new MacroInstantiation{
5685       DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()};
5686   ActiveMacros.push_back(MI);
5687 
5688   // Jump to the macro instantiation and prime the lexer.
5689   CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
5690   Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
5691   Lex();
5692 }
5693 
5694 /// parseDirectiveRept
5695 ///   ::= .rep | .rept count
parseDirectiveRept(SMLoc DirectiveLoc,StringRef Dir)5696 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) {
5697   const MCExpr *CountExpr;
5698   SMLoc CountLoc = getTok().getLoc();
5699   if (parseExpression(CountExpr))
5700     return true;
5701 
5702   int64_t Count;
5703   if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) {
5704     return Error(CountLoc, "unexpected token in '" + Dir + "' directive");
5705   }
5706 
5707   if (check(Count < 0, CountLoc, "Count is negative") || parseEOL())
5708     return true;
5709 
5710   // Lex the rept definition.
5711   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
5712   if (!M)
5713     return true;
5714 
5715   // Macro instantiation is lexical, unfortunately. We construct a new buffer
5716   // to hold the macro body with substitutions.
5717   SmallString<256> Buf;
5718   raw_svector_ostream OS(Buf);
5719   while (Count--) {
5720     // Note that the AtPseudoVariable is disabled for instantiations of .rep(t).
5721     if (expandMacro(OS, *M, std::nullopt, std::nullopt, false))
5722       return true;
5723   }
5724   instantiateMacroLikeBody(M, DirectiveLoc, OS);
5725 
5726   return false;
5727 }
5728 
5729 /// parseDirectiveIrp
5730 /// ::= .irp symbol,values
parseDirectiveIrp(SMLoc DirectiveLoc)5731 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) {
5732   MCAsmMacroParameter Parameter;
5733   MCAsmMacroArguments A;
5734   if (check(parseIdentifier(Parameter.Name),
5735             "expected identifier in '.irp' directive") ||
5736       parseComma() || parseMacroArguments(nullptr, A) || parseEOL())
5737     return true;
5738 
5739   // Lex the irp definition.
5740   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
5741   if (!M)
5742     return true;
5743 
5744   // Macro instantiation is lexical, unfortunately. We construct a new buffer
5745   // to hold the macro body with substitutions.
5746   SmallString<256> Buf;
5747   raw_svector_ostream OS(Buf);
5748 
5749   for (const MCAsmMacroArgument &Arg : A) {
5750     // Note that the AtPseudoVariable is enabled for instantiations of .irp.
5751     // This is undocumented, but GAS seems to support it.
5752     if (expandMacro(OS, *M, Parameter, Arg, true))
5753       return true;
5754   }
5755 
5756   instantiateMacroLikeBody(M, DirectiveLoc, OS);
5757 
5758   return false;
5759 }
5760 
5761 /// parseDirectiveIrpc
5762 /// ::= .irpc symbol,values
parseDirectiveIrpc(SMLoc DirectiveLoc)5763 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) {
5764   MCAsmMacroParameter Parameter;
5765   MCAsmMacroArguments A;
5766 
5767   if (check(parseIdentifier(Parameter.Name),
5768             "expected identifier in '.irpc' directive") ||
5769       parseComma() || parseMacroArguments(nullptr, A))
5770     return true;
5771 
5772   if (A.size() != 1 || A.front().size() != 1)
5773     return TokError("unexpected token in '.irpc' directive");
5774   if (parseEOL())
5775     return true;
5776 
5777   // Lex the irpc definition.
5778   MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
5779   if (!M)
5780     return true;
5781 
5782   // Macro instantiation is lexical, unfortunately. We construct a new buffer
5783   // to hold the macro body with substitutions.
5784   SmallString<256> Buf;
5785   raw_svector_ostream OS(Buf);
5786 
5787   StringRef Values = A.front().front().getString();
5788   for (std::size_t I = 0, End = Values.size(); I != End; ++I) {
5789     MCAsmMacroArgument Arg;
5790     Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1));
5791 
5792     // Note that the AtPseudoVariable is enabled for instantiations of .irpc.
5793     // This is undocumented, but GAS seems to support it.
5794     if (expandMacro(OS, *M, Parameter, Arg, true))
5795       return true;
5796   }
5797 
5798   instantiateMacroLikeBody(M, DirectiveLoc, OS);
5799 
5800   return false;
5801 }
5802 
parseDirectiveEndr(SMLoc DirectiveLoc)5803 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) {
5804   if (ActiveMacros.empty())
5805     return TokError("unmatched '.endr' directive");
5806 
5807   // The only .repl that should get here are the ones created by
5808   // instantiateMacroLikeBody.
5809   assert(getLexer().is(AsmToken::EndOfStatement));
5810 
5811   handleMacroExit();
5812   return false;
5813 }
5814 
parseDirectiveMSEmit(SMLoc IDLoc,ParseStatementInfo & Info,size_t Len)5815 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info,
5816                                      size_t Len) {
5817   const MCExpr *Value;
5818   SMLoc ExprLoc = getLexer().getLoc();
5819   if (parseExpression(Value))
5820     return true;
5821   const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
5822   if (!MCE)
5823     return Error(ExprLoc, "unexpected expression in _emit");
5824   uint64_t IntValue = MCE->getValue();
5825   if (!isUInt<8>(IntValue) && !isInt<8>(IntValue))
5826     return Error(ExprLoc, "literal value out of range for directive");
5827 
5828   Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len);
5829   return false;
5830 }
5831 
parseDirectiveMSAlign(SMLoc IDLoc,ParseStatementInfo & Info)5832 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) {
5833   const MCExpr *Value;
5834   SMLoc ExprLoc = getLexer().getLoc();
5835   if (parseExpression(Value))
5836     return true;
5837   const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
5838   if (!MCE)
5839     return Error(ExprLoc, "unexpected expression in align");
5840   uint64_t IntValue = MCE->getValue();
5841   if (!isPowerOf2_64(IntValue))
5842     return Error(ExprLoc, "literal value not a power of two greater then zero");
5843 
5844   Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue));
5845   return false;
5846 }
5847 
parseDirectivePrint(SMLoc DirectiveLoc)5848 bool AsmParser::parseDirectivePrint(SMLoc DirectiveLoc) {
5849   const AsmToken StrTok = getTok();
5850   Lex();
5851   if (StrTok.isNot(AsmToken::String) || StrTok.getString().front() != '"')
5852     return Error(DirectiveLoc, "expected double quoted string after .print");
5853   if (parseEOL())
5854     return true;
5855   llvm::outs() << StrTok.getStringContents() << '\n';
5856   return false;
5857 }
5858 
parseDirectiveAddrsig()5859 bool AsmParser::parseDirectiveAddrsig() {
5860   if (parseEOL())
5861     return true;
5862   getStreamer().emitAddrsig();
5863   return false;
5864 }
5865 
parseDirectiveAddrsigSym()5866 bool AsmParser::parseDirectiveAddrsigSym() {
5867   StringRef Name;
5868   if (check(parseIdentifier(Name), "expected identifier") || parseEOL())
5869     return true;
5870   MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
5871   getStreamer().emitAddrsigSym(Sym);
5872   return false;
5873 }
5874 
parseDirectivePseudoProbe()5875 bool AsmParser::parseDirectivePseudoProbe() {
5876   int64_t Guid;
5877   int64_t Index;
5878   int64_t Type;
5879   int64_t Attr;
5880   int64_t Discriminator = 0;
5881 
5882   if (parseIntToken(Guid, "unexpected token in '.pseudoprobe' directive"))
5883     return true;
5884 
5885   if (parseIntToken(Index, "unexpected token in '.pseudoprobe' directive"))
5886     return true;
5887 
5888   if (parseIntToken(Type, "unexpected token in '.pseudoprobe' directive"))
5889     return true;
5890 
5891   if (parseIntToken(Attr, "unexpected token in '.pseudoprobe' directive"))
5892     return true;
5893 
5894   if (hasDiscriminator(Attr)) {
5895     if (parseIntToken(Discriminator,
5896                       "unexpected token in '.pseudoprobe' directive"))
5897       return true;
5898   }
5899 
5900   // Parse inline stack like @ GUID:11:12 @ GUID:1:11 @ GUID:3:21
5901   MCPseudoProbeInlineStack InlineStack;
5902 
5903   while (getLexer().is(AsmToken::At)) {
5904     // eat @
5905     Lex();
5906 
5907     int64_t CallerGuid = 0;
5908     if (getLexer().is(AsmToken::Integer)) {
5909       if (parseIntToken(CallerGuid,
5910                         "unexpected token in '.pseudoprobe' directive"))
5911         return true;
5912     }
5913 
5914     // eat colon
5915     if (getLexer().is(AsmToken::Colon))
5916       Lex();
5917 
5918     int64_t CallerProbeId = 0;
5919     if (getLexer().is(AsmToken::Integer)) {
5920       if (parseIntToken(CallerProbeId,
5921                         "unexpected token in '.pseudoprobe' directive"))
5922         return true;
5923     }
5924 
5925     InlineSite Site(CallerGuid, CallerProbeId);
5926     InlineStack.push_back(Site);
5927   }
5928 
5929   // Parse function entry name
5930   StringRef FnName;
5931   if (parseIdentifier(FnName))
5932     return Error(getLexer().getLoc(), "unexpected token in '.pseudoprobe' directive");
5933   MCSymbol *FnSym = getContext().lookupSymbol(FnName);
5934 
5935   if (parseEOL())
5936     return true;
5937 
5938   getStreamer().emitPseudoProbe(Guid, Index, Type, Attr, Discriminator,
5939                                 InlineStack, FnSym);
5940   return false;
5941 }
5942 
5943 /// parseDirectiveLTODiscard
5944 ///  ::= ".lto_discard" [ identifier ( , identifier )* ]
5945 /// The LTO library emits this directive to discard non-prevailing symbols.
5946 /// We ignore symbol assignments and attribute changes for the specified
5947 /// symbols.
parseDirectiveLTODiscard()5948 bool AsmParser::parseDirectiveLTODiscard() {
5949   auto ParseOp = [&]() -> bool {
5950     StringRef Name;
5951     SMLoc Loc = getTok().getLoc();
5952     if (parseIdentifier(Name))
5953       return Error(Loc, "expected identifier");
5954     LTODiscardSymbols.insert(Name);
5955     return false;
5956   };
5957 
5958   LTODiscardSymbols.clear();
5959   return parseMany(ParseOp);
5960 }
5961 
5962 // We are comparing pointers, but the pointers are relative to a single string.
5963 // Thus, this should always be deterministic.
rewritesSort(const AsmRewrite * AsmRewriteA,const AsmRewrite * AsmRewriteB)5964 static int rewritesSort(const AsmRewrite *AsmRewriteA,
5965                         const AsmRewrite *AsmRewriteB) {
5966   if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer())
5967     return -1;
5968   if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer())
5969     return 1;
5970 
5971   // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output
5972   // rewrite to the same location.  Make sure the SizeDirective rewrite is
5973   // performed first, then the Imm/ImmPrefix and finally the Input/Output.  This
5974   // ensures the sort algorithm is stable.
5975   if (AsmRewritePrecedence[AsmRewriteA->Kind] >
5976       AsmRewritePrecedence[AsmRewriteB->Kind])
5977     return -1;
5978 
5979   if (AsmRewritePrecedence[AsmRewriteA->Kind] <
5980       AsmRewritePrecedence[AsmRewriteB->Kind])
5981     return 1;
5982   llvm_unreachable("Unstable rewrite sort.");
5983 }
5984 
parseMSInlineAsm(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)5985 bool AsmParser::parseMSInlineAsm(
5986     std::string &AsmString, unsigned &NumOutputs, unsigned &NumInputs,
5987     SmallVectorImpl<std::pair<void *, bool>> &OpDecls,
5988     SmallVectorImpl<std::string> &Constraints,
5989     SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII,
5990     const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) {
5991   SmallVector<void *, 4> InputDecls;
5992   SmallVector<void *, 4> OutputDecls;
5993   SmallVector<bool, 4> InputDeclsAddressOf;
5994   SmallVector<bool, 4> OutputDeclsAddressOf;
5995   SmallVector<std::string, 4> InputConstraints;
5996   SmallVector<std::string, 4> OutputConstraints;
5997   SmallVector<unsigned, 4> ClobberRegs;
5998 
5999   SmallVector<AsmRewrite, 4> AsmStrRewrites;
6000 
6001   // Prime the lexer.
6002   Lex();
6003 
6004   // While we have input, parse each statement.
6005   unsigned InputIdx = 0;
6006   unsigned OutputIdx = 0;
6007   while (getLexer().isNot(AsmToken::Eof)) {
6008     // Parse curly braces marking block start/end
6009     if (parseCurlyBlockScope(AsmStrRewrites))
6010       continue;
6011 
6012     ParseStatementInfo Info(&AsmStrRewrites);
6013     bool StatementErr = parseStatement(Info, &SI);
6014 
6015     if (StatementErr || Info.ParseError) {
6016       // Emit pending errors if any exist.
6017       printPendingErrors();
6018       return true;
6019     }
6020 
6021     // No pending error should exist here.
6022     assert(!hasPendingError() && "unexpected error from parseStatement");
6023 
6024     if (Info.Opcode == ~0U)
6025       continue;
6026 
6027     const MCInstrDesc &Desc = MII->get(Info.Opcode);
6028 
6029     // Build the list of clobbers, outputs and inputs.
6030     for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
6031       MCParsedAsmOperand &Operand = *Info.ParsedOperands[i];
6032 
6033       // Register operand.
6034       if (Operand.isReg() && !Operand.needAddressOf() &&
6035           !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) {
6036         unsigned NumDefs = Desc.getNumDefs();
6037         // Clobber.
6038         if (NumDefs && Operand.getMCOperandNum() < NumDefs)
6039           ClobberRegs.push_back(Operand.getReg());
6040         continue;
6041       }
6042 
6043       // Expr/Input or Output.
6044       StringRef SymName = Operand.getSymName();
6045       if (SymName.empty())
6046         continue;
6047 
6048       void *OpDecl = Operand.getOpDecl();
6049       if (!OpDecl)
6050         continue;
6051 
6052       StringRef Constraint = Operand.getConstraint();
6053       if (Operand.isImm()) {
6054         // Offset as immediate
6055         if (Operand.isOffsetOfLocal())
6056           Constraint = "r";
6057         else
6058           Constraint = "i";
6059       }
6060 
6061       bool isOutput = (i == 1) && Desc.mayStore();
6062       bool Restricted = Operand.isMemUseUpRegs();
6063       SMLoc Start = SMLoc::getFromPointer(SymName.data());
6064       if (isOutput) {
6065         ++InputIdx;
6066         OutputDecls.push_back(OpDecl);
6067         OutputDeclsAddressOf.push_back(Operand.needAddressOf());
6068         OutputConstraints.push_back(("=" + Constraint).str());
6069         AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size(), 0,
6070                                     Restricted);
6071       } else {
6072         InputDecls.push_back(OpDecl);
6073         InputDeclsAddressOf.push_back(Operand.needAddressOf());
6074         InputConstraints.push_back(Constraint.str());
6075         if (Desc.operands()[i - 1].isBranchTarget())
6076           AsmStrRewrites.emplace_back(AOK_CallInput, Start, SymName.size(), 0,
6077                                       Restricted);
6078         else
6079           AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size(), 0,
6080                                       Restricted);
6081       }
6082     }
6083 
6084     // Consider implicit defs to be clobbers.  Think of cpuid and push.
6085     llvm::append_range(ClobberRegs, Desc.implicit_defs());
6086   }
6087 
6088   // Set the number of Outputs and Inputs.
6089   NumOutputs = OutputDecls.size();
6090   NumInputs = InputDecls.size();
6091 
6092   // Set the unique clobbers.
6093   array_pod_sort(ClobberRegs.begin(), ClobberRegs.end());
6094   ClobberRegs.erase(llvm::unique(ClobberRegs), ClobberRegs.end());
6095   Clobbers.assign(ClobberRegs.size(), std::string());
6096   for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) {
6097     raw_string_ostream OS(Clobbers[I]);
6098     IP->printRegName(OS, ClobberRegs[I]);
6099   }
6100 
6101   // Merge the various outputs and inputs.  Output are expected first.
6102   if (NumOutputs || NumInputs) {
6103     unsigned NumExprs = NumOutputs + NumInputs;
6104     OpDecls.resize(NumExprs);
6105     Constraints.resize(NumExprs);
6106     for (unsigned i = 0; i < NumOutputs; ++i) {
6107       OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]);
6108       Constraints[i] = OutputConstraints[i];
6109     }
6110     for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
6111       OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]);
6112       Constraints[j] = InputConstraints[i];
6113     }
6114   }
6115 
6116   // Build the IR assembly string.
6117   std::string AsmStringIR;
6118   raw_string_ostream OS(AsmStringIR);
6119   StringRef ASMString =
6120       SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer();
6121   const char *AsmStart = ASMString.begin();
6122   const char *AsmEnd = ASMString.end();
6123   array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort);
6124   for (auto I = AsmStrRewrites.begin(), E = AsmStrRewrites.end(); I != E; ++I) {
6125     const AsmRewrite &AR = *I;
6126     // Check if this has already been covered by another rewrite...
6127     if (AR.Done)
6128       continue;
6129     AsmRewriteKind Kind = AR.Kind;
6130 
6131     const char *Loc = AR.Loc.getPointer();
6132     assert(Loc >= AsmStart && "Expected Loc to be at or after Start!");
6133 
6134     // Emit everything up to the immediate/expression.
6135     if (unsigned Len = Loc - AsmStart)
6136       OS << StringRef(AsmStart, Len);
6137 
6138     // Skip the original expression.
6139     if (Kind == AOK_Skip) {
6140       AsmStart = Loc + AR.Len;
6141       continue;
6142     }
6143 
6144     unsigned AdditionalSkip = 0;
6145     // Rewrite expressions in $N notation.
6146     switch (Kind) {
6147     default:
6148       break;
6149     case AOK_IntelExpr:
6150       assert(AR.IntelExp.isValid() && "cannot write invalid intel expression");
6151       if (AR.IntelExp.NeedBracs)
6152         OS << "[";
6153       if (AR.IntelExp.hasBaseReg())
6154         OS << AR.IntelExp.BaseReg;
6155       if (AR.IntelExp.hasIndexReg())
6156         OS << (AR.IntelExp.hasBaseReg() ? " + " : "")
6157            << AR.IntelExp.IndexReg;
6158       if (AR.IntelExp.Scale > 1)
6159         OS << " * $$" << AR.IntelExp.Scale;
6160       if (AR.IntelExp.hasOffset()) {
6161         if (AR.IntelExp.hasRegs())
6162           OS << " + ";
6163         // Fuse this rewrite with a rewrite of the offset name, if present.
6164         StringRef OffsetName = AR.IntelExp.OffsetName;
6165         SMLoc OffsetLoc = SMLoc::getFromPointer(AR.IntelExp.OffsetName.data());
6166         size_t OffsetLen = OffsetName.size();
6167         auto rewrite_it = std::find_if(
6168             I, AsmStrRewrites.end(), [&](const AsmRewrite &FusingAR) {
6169               return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen &&
6170                      (FusingAR.Kind == AOK_Input ||
6171                       FusingAR.Kind == AOK_CallInput);
6172             });
6173         if (rewrite_it == AsmStrRewrites.end()) {
6174           OS << "offset " << OffsetName;
6175         } else if (rewrite_it->Kind == AOK_CallInput) {
6176           OS << "${" << InputIdx++ << ":P}";
6177           rewrite_it->Done = true;
6178         } else {
6179           OS << '$' << InputIdx++;
6180           rewrite_it->Done = true;
6181         }
6182       }
6183       if (AR.IntelExp.Imm || AR.IntelExp.emitImm())
6184         OS << (AR.IntelExp.emitImm() ? "$$" : " + $$") << AR.IntelExp.Imm;
6185       if (AR.IntelExp.NeedBracs)
6186         OS << "]";
6187       break;
6188     case AOK_Label:
6189       OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label;
6190       break;
6191     case AOK_Input:
6192       if (AR.IntelExpRestricted)
6193         OS << "${" << InputIdx++ << ":P}";
6194       else
6195         OS << '$' << InputIdx++;
6196       break;
6197     case AOK_CallInput:
6198       OS << "${" << InputIdx++ << ":P}";
6199       break;
6200     case AOK_Output:
6201       if (AR.IntelExpRestricted)
6202         OS << "${" << OutputIdx++ << ":P}";
6203       else
6204         OS << '$' << OutputIdx++;
6205       break;
6206     case AOK_SizeDirective:
6207       switch (AR.Val) {
6208       default: break;
6209       case 8:  OS << "byte ptr "; break;
6210       case 16: OS << "word ptr "; break;
6211       case 32: OS << "dword ptr "; break;
6212       case 64: OS << "qword ptr "; break;
6213       case 80: OS << "xword ptr "; break;
6214       case 128: OS << "xmmword ptr "; break;
6215       case 256: OS << "ymmword ptr "; break;
6216       }
6217       break;
6218     case AOK_Emit:
6219       OS << ".byte";
6220       break;
6221     case AOK_Align: {
6222       // MS alignment directives are measured in bytes. If the native assembler
6223       // measures alignment in bytes, we can pass it straight through.
6224       OS << ".align";
6225       if (getContext().getAsmInfo()->getAlignmentIsInBytes())
6226         break;
6227 
6228       // Alignment is in log2 form, so print that instead and skip the original
6229       // immediate.
6230       unsigned Val = AR.Val;
6231       OS << ' ' << Val;
6232       assert(Val < 10 && "Expected alignment less then 2^10.");
6233       AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4;
6234       break;
6235     }
6236     case AOK_EVEN:
6237       OS << ".even";
6238       break;
6239     case AOK_EndOfStatement:
6240       OS << "\n\t";
6241       break;
6242     }
6243 
6244     // Skip the original expression.
6245     AsmStart = Loc + AR.Len + AdditionalSkip;
6246   }
6247 
6248   // Emit the remainder of the asm string.
6249   if (AsmStart != AsmEnd)
6250     OS << StringRef(AsmStart, AsmEnd - AsmStart);
6251 
6252   AsmString = AsmStringIR;
6253   return false;
6254 }
6255 
parseAsHLASMLabel(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)6256 bool HLASMAsmParser::parseAsHLASMLabel(ParseStatementInfo &Info,
6257                                        MCAsmParserSemaCallback *SI) {
6258   AsmToken LabelTok = getTok();
6259   SMLoc LabelLoc = LabelTok.getLoc();
6260   StringRef LabelVal;
6261 
6262   if (parseIdentifier(LabelVal))
6263     return Error(LabelLoc, "The HLASM Label has to be an Identifier");
6264 
6265   // We have validated whether the token is an Identifier.
6266   // Now we have to validate whether the token is a
6267   // valid HLASM Label.
6268   if (!getTargetParser().isLabel(LabelTok) || checkForValidSection())
6269     return true;
6270 
6271   // Lex leading spaces to get to the next operand.
6272   lexLeadingSpaces();
6273 
6274   // We shouldn't emit the label if there is nothing else after the label.
6275   // i.e asm("<token>\n")
6276   if (getTok().is(AsmToken::EndOfStatement))
6277     return Error(LabelLoc,
6278                  "Cannot have just a label for an HLASM inline asm statement");
6279 
6280   MCSymbol *Sym = getContext().getOrCreateSymbol(
6281       getContext().getAsmInfo()->shouldEmitLabelsInUpperCase()
6282           ? LabelVal.upper()
6283           : LabelVal);
6284 
6285   getTargetParser().doBeforeLabelEmit(Sym, LabelLoc);
6286 
6287   // Emit the label.
6288   Out.emitLabel(Sym, LabelLoc);
6289 
6290   // If we are generating dwarf for assembly source files then gather the
6291   // info to make a dwarf label entry for this label if needed.
6292   if (enabledGenDwarfForAssembly())
6293     MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
6294                                LabelLoc);
6295 
6296   getTargetParser().onLabelParsed(Sym);
6297 
6298   return false;
6299 }
6300 
parseAsMachineInstruction(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)6301 bool HLASMAsmParser::parseAsMachineInstruction(ParseStatementInfo &Info,
6302                                                MCAsmParserSemaCallback *SI) {
6303   AsmToken OperationEntryTok = Lexer.getTok();
6304   SMLoc OperationEntryLoc = OperationEntryTok.getLoc();
6305   StringRef OperationEntryVal;
6306 
6307   // Attempt to parse the first token as an Identifier
6308   if (parseIdentifier(OperationEntryVal))
6309     return Error(OperationEntryLoc, "unexpected token at start of statement");
6310 
6311   // Once we've parsed the operation entry successfully, lex
6312   // any spaces to get to the OperandEntries.
6313   lexLeadingSpaces();
6314 
6315   return parseAndMatchAndEmitTargetInstruction(
6316       Info, OperationEntryVal, OperationEntryTok, OperationEntryLoc);
6317 }
6318 
parseStatement(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)6319 bool HLASMAsmParser::parseStatement(ParseStatementInfo &Info,
6320                                     MCAsmParserSemaCallback *SI) {
6321   assert(!hasPendingError() && "parseStatement started with pending error");
6322 
6323   // Should the first token be interpreted as a HLASM Label.
6324   bool ShouldParseAsHLASMLabel = false;
6325 
6326   // If a Name Entry exists, it should occur at the very
6327   // start of the string. In this case, we should parse the
6328   // first non-space token as a Label.
6329   // If the Name entry is missing (i.e. there's some other
6330   // token), then we attempt to parse the first non-space
6331   // token as a Machine Instruction.
6332   if (getTok().isNot(AsmToken::Space))
6333     ShouldParseAsHLASMLabel = true;
6334 
6335   // If we have an EndOfStatement (which includes the target's comment
6336   // string) we can appropriately lex it early on)
6337   if (Lexer.is(AsmToken::EndOfStatement)) {
6338     // if this is a line comment we can drop it safely
6339     if (getTok().getString().empty() || getTok().getString().front() == '\r' ||
6340         getTok().getString().front() == '\n')
6341       Out.addBlankLine();
6342     Lex();
6343     return false;
6344   }
6345 
6346   // We have established how to parse the inline asm statement.
6347   // Now we can safely lex any leading spaces to get to the
6348   // first token.
6349   lexLeadingSpaces();
6350 
6351   // If we see a new line or carriage return as the first operand,
6352   // after lexing leading spaces, emit the new line and lex the
6353   // EndOfStatement token.
6354   if (Lexer.is(AsmToken::EndOfStatement)) {
6355     if (getTok().getString().front() == '\n' ||
6356         getTok().getString().front() == '\r') {
6357       Out.addBlankLine();
6358       Lex();
6359       return false;
6360     }
6361   }
6362 
6363   // Handle the label first if we have to before processing the rest
6364   // of the tokens as a machine instruction.
6365   if (ShouldParseAsHLASMLabel) {
6366     // If there were any errors while handling and emitting the label,
6367     // early return.
6368     if (parseAsHLASMLabel(Info, SI)) {
6369       // If we know we've failed in parsing, simply eat until end of the
6370       // statement. This ensures that we don't process any other statements.
6371       eatToEndOfStatement();
6372       return true;
6373     }
6374   }
6375 
6376   return parseAsMachineInstruction(Info, SI);
6377 }
6378 
6379 namespace llvm {
6380 namespace MCParserUtils {
6381 
6382 /// Returns whether the given symbol is used anywhere in the given expression,
6383 /// or subexpressions.
isSymbolUsedInExpression(const MCSymbol * Sym,const MCExpr * Value)6384 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) {
6385   switch (Value->getKind()) {
6386   case MCExpr::Binary: {
6387     const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value);
6388     return isSymbolUsedInExpression(Sym, BE->getLHS()) ||
6389            isSymbolUsedInExpression(Sym, BE->getRHS());
6390   }
6391   case MCExpr::Target:
6392   case MCExpr::Constant:
6393     return false;
6394   case MCExpr::SymbolRef: {
6395     const MCSymbol &S =
6396         static_cast<const MCSymbolRefExpr *>(Value)->getSymbol();
6397     if (S.isVariable() && !S.isWeakExternal())
6398       return isSymbolUsedInExpression(Sym, S.getVariableValue());
6399     return &S == Sym;
6400   }
6401   case MCExpr::Unary:
6402     return isSymbolUsedInExpression(
6403         Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr());
6404   }
6405 
6406   llvm_unreachable("Unknown expr kind!");
6407 }
6408 
parseAssignmentExpression(StringRef Name,bool allow_redef,MCAsmParser & Parser,MCSymbol * & Sym,const MCExpr * & Value)6409 bool parseAssignmentExpression(StringRef Name, bool allow_redef,
6410                                MCAsmParser &Parser, MCSymbol *&Sym,
6411                                const MCExpr *&Value) {
6412 
6413   // FIXME: Use better location, we should use proper tokens.
6414   SMLoc EqualLoc = Parser.getTok().getLoc();
6415   if (Parser.parseExpression(Value))
6416     return Parser.TokError("missing expression");
6417 
6418   // Note: we don't count b as used in "a = b". This is to allow
6419   // a = b
6420   // b = c
6421 
6422   if (Parser.parseEOL())
6423     return true;
6424 
6425   // Validate that the LHS is allowed to be a variable (either it has not been
6426   // used as a symbol, or it is an absolute symbol).
6427   Sym = Parser.getContext().lookupSymbol(Name);
6428   if (Sym) {
6429     // Diagnose assignment to a label.
6430     //
6431     // FIXME: Diagnostics. Note the location of the definition as a label.
6432     // FIXME: Diagnose assignment to protected identifier (e.g., register name).
6433     if (isSymbolUsedInExpression(Sym, Value))
6434       return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'");
6435     else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() &&
6436              !Sym->isVariable())
6437       ; // Allow redefinitions of undefined symbols only used in directives.
6438     else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
6439       ; // Allow redefinitions of variables that haven't yet been used.
6440     else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
6441       return Parser.Error(EqualLoc, "redefinition of '" + Name + "'");
6442     else if (!Sym->isVariable())
6443       return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'");
6444     else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
6445       return Parser.Error(EqualLoc,
6446                           "invalid reassignment of non-absolute variable '" +
6447                               Name + "'");
6448   } else if (Name == ".") {
6449     Parser.getStreamer().emitValueToOffset(Value, 0, EqualLoc);
6450     return false;
6451   } else
6452     Sym = Parser.getContext().getOrCreateSymbol(Name);
6453 
6454   Sym->setRedefinable(allow_redef);
6455 
6456   return false;
6457 }
6458 
6459 } // end namespace MCParserUtils
6460 } // end namespace llvm
6461 
6462 /// Create an MCAsmParser instance.
createMCAsmParser(SourceMgr & SM,MCContext & C,MCStreamer & Out,const MCAsmInfo & MAI,unsigned CB)6463 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C,
6464                                      MCStreamer &Out, const MCAsmInfo &MAI,
6465                                      unsigned CB) {
6466   if (C.getTargetTriple().isSystemZ() && C.getTargetTriple().isOSzOS())
6467     return new HLASMAsmParser(SM, C, Out, MAI, CB);
6468 
6469   return new AsmParser(SM, C, Out, MAI, CB);
6470 }
6471