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/None.h"
17 #include "llvm/ADT/STLExtras.h"
18 #include "llvm/ADT/SmallSet.h"
19 #include "llvm/ADT/SmallString.h"
20 #include "llvm/ADT/SmallVector.h"
21 #include "llvm/ADT/StringExtras.h"
22 #include "llvm/ADT/StringMap.h"
23 #include "llvm/ADT/StringRef.h"
24 #include "llvm/ADT/Twine.h"
25 #include "llvm/BinaryFormat/Dwarf.h"
26 #include "llvm/DebugInfo/CodeView/SymbolRecord.h"
27 #include "llvm/MC/MCAsmInfo.h"
28 #include "llvm/MC/MCCodeView.h"
29 #include "llvm/MC/MCContext.h"
30 #include "llvm/MC/MCDirectives.h"
31 #include "llvm/MC/MCDwarf.h"
32 #include "llvm/MC/MCExpr.h"
33 #include "llvm/MC/MCInstPrinter.h"
34 #include "llvm/MC/MCInstrDesc.h"
35 #include "llvm/MC/MCInstrInfo.h"
36 #include "llvm/MC/MCObjectFileInfo.h"
37 #include "llvm/MC/MCParser/AsmCond.h"
38 #include "llvm/MC/MCParser/AsmLexer.h"
39 #include "llvm/MC/MCParser/MCAsmLexer.h"
40 #include "llvm/MC/MCParser/MCAsmParser.h"
41 #include "llvm/MC/MCParser/MCAsmParserExtension.h"
42 #include "llvm/MC/MCParser/MCAsmParserUtils.h"
43 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
44 #include "llvm/MC/MCParser/MCTargetAsmParser.h"
45 #include "llvm/MC/MCRegisterInfo.h"
46 #include "llvm/MC/MCSection.h"
47 #include "llvm/MC/MCStreamer.h"
48 #include "llvm/MC/MCSymbol.h"
49 #include "llvm/MC/MCTargetOptions.h"
50 #include "llvm/MC/MCValue.h"
51 #include "llvm/Support/Casting.h"
52 #include "llvm/Support/CommandLine.h"
53 #include "llvm/Support/ErrorHandling.h"
54 #include "llvm/Support/MD5.h"
55 #include "llvm/Support/MathExtras.h"
56 #include "llvm/Support/MemoryBuffer.h"
57 #include "llvm/Support/SMLoc.h"
58 #include "llvm/Support/SourceMgr.h"
59 #include "llvm/Support/raw_ostream.h"
60 #include <algorithm>
61 #include <cassert>
62 #include <cctype>
63 #include <climits>
64 #include <cstddef>
65 #include <cstdint>
66 #include <deque>
67 #include <memory>
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 extern cl::opt<unsigned> AsmMacroMaxNestingDepth;
79
80 namespace {
81
82 /// Helper types for tracking macro definitions.
83 typedef std::vector<AsmToken> MCAsmMacroArgument;
84 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments;
85
86 /// Helper class for storing information about an active macro
87 /// instantiation.
88 struct MacroInstantiation {
89 /// The location of the instantiation.
90 SMLoc InstantiationLoc;
91
92 /// The buffer where parsing should resume upon instantiation completion.
93 unsigned ExitBuffer;
94
95 /// The location where parsing should resume upon instantiation completion.
96 SMLoc ExitLoc;
97
98 /// The depth of TheCondStack at the start of the instantiation.
99 size_t CondStackDepth;
100 };
101
102 struct ParseStatementInfo {
103 /// The parsed operands from the last parsed statement.
104 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands;
105
106 /// The opcode from the last parsed instruction.
107 unsigned Opcode = ~0U;
108
109 /// Was there an error parsing the inline assembly?
110 bool ParseError = false;
111
112 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr;
113
114 ParseStatementInfo() = delete;
ParseStatementInfo__anon092769c30111::ParseStatementInfo115 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
116 : AsmRewrites(rewrites) {}
117 };
118
119 /// The concrete assembly parser instance.
120 class AsmParser : public MCAsmParser {
121 private:
122 AsmLexer Lexer;
123 MCContext &Ctx;
124 MCStreamer &Out;
125 const MCAsmInfo &MAI;
126 SourceMgr &SrcMgr;
127 SourceMgr::DiagHandlerTy SavedDiagHandler;
128 void *SavedDiagContext;
129 std::unique_ptr<MCAsmParserExtension> PlatformParser;
130 SMLoc StartTokLoc;
131
132 /// This is the current buffer index we're lexing from as managed by the
133 /// SourceMgr object.
134 unsigned CurBuffer;
135
136 AsmCond TheCondState;
137 std::vector<AsmCond> TheCondStack;
138
139 /// maps directive names to handler methods in parser
140 /// extensions. Extensions register themselves in this map by calling
141 /// addDirectiveHandler.
142 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap;
143
144 /// Stack of active macro instantiations.
145 std::vector<MacroInstantiation*> ActiveMacros;
146
147 /// List of bodies of anonymous macros.
148 std::deque<MCAsmMacro> MacroLikeBodies;
149
150 /// Boolean tracking whether macro substitution is enabled.
151 unsigned MacrosEnabledFlag : 1;
152
153 /// Keeps track of how many .macro's have been instantiated.
154 unsigned NumOfMacroInstantiations;
155
156 /// The values from the last parsed cpp hash file line comment if any.
157 struct CppHashInfoTy {
158 StringRef Filename;
159 int64_t LineNumber;
160 SMLoc Loc;
161 unsigned Buf;
CppHashInfoTy__anon092769c30111::AsmParser::CppHashInfoTy162 CppHashInfoTy() : Filename(), LineNumber(0), Loc(), Buf(0) {}
163 };
164 CppHashInfoTy CppHashInfo;
165
166 /// The filename from the first cpp hash file line comment, if any.
167 StringRef FirstCppHashFilename;
168
169 /// List of forward directional labels for diagnosis at the end.
170 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels;
171
172 SmallSet<StringRef, 2> LTODiscardSymbols;
173
174 /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
175 unsigned AssemblerDialect = ~0U;
176
177 /// is Darwin compatibility enabled?
178 bool IsDarwin = false;
179
180 /// Are we parsing ms-style inline assembly?
181 bool ParsingMSInlineAsm = false;
182
183 /// Did we already inform the user about inconsistent MD5 usage?
184 bool ReportedInconsistentMD5 = false;
185
186 // Is alt macro mode enabled.
187 bool AltMacroMode = false;
188
189 protected:
190 virtual bool parseStatement(ParseStatementInfo &Info,
191 MCAsmParserSemaCallback *SI);
192
193 /// This routine uses the target specific ParseInstruction function to
194 /// parse an instruction into Operands, and then call the target specific
195 /// MatchAndEmit function to match and emit the instruction.
196 bool parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info,
197 StringRef IDVal, AsmToken ID,
198 SMLoc IDLoc);
199
200 /// Should we emit DWARF describing this assembler source? (Returns false if
201 /// the source has .file directives, which means we don't want to generate
202 /// info describing the assembler source itself.)
203 bool enabledGenDwarfForAssembly();
204
205 public:
206 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
207 const MCAsmInfo &MAI, unsigned CB);
208 AsmParser(const AsmParser &) = delete;
209 AsmParser &operator=(const AsmParser &) = delete;
210 ~AsmParser() override;
211
212 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override;
213
addDirectiveHandler(StringRef Directive,ExtensionDirectiveHandler Handler)214 void addDirectiveHandler(StringRef Directive,
215 ExtensionDirectiveHandler Handler) override {
216 ExtensionDirectiveMap[Directive] = Handler;
217 }
218
addAliasForDirective(StringRef Directive,StringRef Alias)219 void addAliasForDirective(StringRef Directive, StringRef Alias) override {
220 DirectiveKindMap[Directive.lower()] = DirectiveKindMap[Alias.lower()];
221 }
222
223 /// @name MCAsmParser Interface
224 /// {
225
getSourceManager()226 SourceMgr &getSourceManager() override { return SrcMgr; }
getLexer()227 MCAsmLexer &getLexer() override { return Lexer; }
getContext()228 MCContext &getContext() override { return Ctx; }
getStreamer()229 MCStreamer &getStreamer() override { return Out; }
230
getCVContext()231 CodeViewContext &getCVContext() { return Ctx.getCVContext(); }
232
getAssemblerDialect()233 unsigned getAssemblerDialect() override {
234 if (AssemblerDialect == ~0U)
235 return MAI.getAssemblerDialect();
236 else
237 return AssemblerDialect;
238 }
setAssemblerDialect(unsigned i)239 void setAssemblerDialect(unsigned i) override {
240 AssemblerDialect = i;
241 }
242
243 void Note(SMLoc L, const Twine &Msg, SMRange Range = None) override;
244 bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) override;
245 bool printError(SMLoc L, const Twine &Msg, SMRange Range = None) override;
246
247 const AsmToken &Lex() override;
248
setParsingMSInlineAsm(bool V)249 void setParsingMSInlineAsm(bool V) override {
250 ParsingMSInlineAsm = V;
251 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and
252 // hex integer literals.
253 Lexer.setLexMasmIntegers(V);
254 }
isParsingMSInlineAsm()255 bool isParsingMSInlineAsm() override { return ParsingMSInlineAsm; }
256
discardLTOSymbol(StringRef Name) const257 bool discardLTOSymbol(StringRef Name) const override {
258 return LTODiscardSymbols.contains(Name);
259 }
260
261 bool parseMSInlineAsm(std::string &AsmString, unsigned &NumOutputs,
262 unsigned &NumInputs,
263 SmallVectorImpl<std::pair<void *, bool>> &OpDecls,
264 SmallVectorImpl<std::string> &Constraints,
265 SmallVectorImpl<std::string> &Clobbers,
266 const MCInstrInfo *MII, const MCInstPrinter *IP,
267 MCAsmParserSemaCallback &SI) override;
268
269 bool parseExpression(const MCExpr *&Res);
270 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
271 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc,
272 AsmTypeInfo *TypeInfo) override;
273 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
274 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
275 SMLoc &EndLoc) override;
276 bool parseAbsoluteExpression(int64_t &Res) override;
277
278 /// Parse a floating point expression using the float \p Semantics
279 /// and set \p Res to the value.
280 bool parseRealValue(const fltSemantics &Semantics, APInt &Res);
281
282 /// Parse an identifier or string (as a quoted identifier)
283 /// and set \p Res to the identifier contents.
284 bool parseIdentifier(StringRef &Res) override;
285 void eatToEndOfStatement() override;
286
287 bool checkForValidSection() override;
288
289 /// }
290
291 private:
292 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites);
293 bool parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo = true);
294
295 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body,
296 ArrayRef<MCAsmMacroParameter> Parameters);
297 bool expandMacro(raw_svector_ostream &OS, StringRef Body,
298 ArrayRef<MCAsmMacroParameter> Parameters,
299 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable,
300 SMLoc L);
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(const 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=None) const327 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
328 SMRange Range = None) 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 bool parseAssignment(StringRef Name, bool allow_redef,
360 bool NoDeadStrip = false);
361
362 unsigned getBinOpPrecedence(AsmToken::TokenKind K,
363 MCBinaryExpr::Opcode &Kind);
364
365 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
366 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
367 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
368
369 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
370
371 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName);
372 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName);
373
374 // Generic (target and platform independent) directive parsing.
375 enum DirectiveKind {
376 DK_NO_DIRECTIVE, // Placeholder
377 DK_SET,
378 DK_EQU,
379 DK_EQUIV,
380 DK_ASCII,
381 DK_ASCIZ,
382 DK_STRING,
383 DK_BYTE,
384 DK_SHORT,
385 DK_RELOC,
386 DK_VALUE,
387 DK_2BYTE,
388 DK_LONG,
389 DK_INT,
390 DK_4BYTE,
391 DK_QUAD,
392 DK_8BYTE,
393 DK_OCTA,
394 DK_DC,
395 DK_DC_A,
396 DK_DC_B,
397 DK_DC_D,
398 DK_DC_L,
399 DK_DC_S,
400 DK_DC_W,
401 DK_DC_X,
402 DK_DCB,
403 DK_DCB_B,
404 DK_DCB_D,
405 DK_DCB_L,
406 DK_DCB_S,
407 DK_DCB_W,
408 DK_DCB_X,
409 DK_DS,
410 DK_DS_B,
411 DK_DS_D,
412 DK_DS_L,
413 DK_DS_P,
414 DK_DS_S,
415 DK_DS_W,
416 DK_DS_X,
417 DK_SINGLE,
418 DK_FLOAT,
419 DK_DOUBLE,
420 DK_ALIGN,
421 DK_ALIGN32,
422 DK_BALIGN,
423 DK_BALIGNW,
424 DK_BALIGNL,
425 DK_P2ALIGN,
426 DK_P2ALIGNW,
427 DK_P2ALIGNL,
428 DK_ORG,
429 DK_FILL,
430 DK_ENDR,
431 DK_BUNDLE_ALIGN_MODE,
432 DK_BUNDLE_LOCK,
433 DK_BUNDLE_UNLOCK,
434 DK_ZERO,
435 DK_EXTERN,
436 DK_GLOBL,
437 DK_GLOBAL,
438 DK_LAZY_REFERENCE,
439 DK_NO_DEAD_STRIP,
440 DK_SYMBOL_RESOLVER,
441 DK_PRIVATE_EXTERN,
442 DK_REFERENCE,
443 DK_WEAK_DEFINITION,
444 DK_WEAK_REFERENCE,
445 DK_WEAK_DEF_CAN_BE_HIDDEN,
446 DK_COLD,
447 DK_COMM,
448 DK_COMMON,
449 DK_LCOMM,
450 DK_ABORT,
451 DK_INCLUDE,
452 DK_INCBIN,
453 DK_CODE16,
454 DK_CODE16GCC,
455 DK_REPT,
456 DK_IRP,
457 DK_IRPC,
458 DK_IF,
459 DK_IFEQ,
460 DK_IFGE,
461 DK_IFGT,
462 DK_IFLE,
463 DK_IFLT,
464 DK_IFNE,
465 DK_IFB,
466 DK_IFNB,
467 DK_IFC,
468 DK_IFEQS,
469 DK_IFNC,
470 DK_IFNES,
471 DK_IFDEF,
472 DK_IFNDEF,
473 DK_IFNOTDEF,
474 DK_ELSEIF,
475 DK_ELSE,
476 DK_ENDIF,
477 DK_SPACE,
478 DK_SKIP,
479 DK_FILE,
480 DK_LINE,
481 DK_LOC,
482 DK_STABS,
483 DK_CV_FILE,
484 DK_CV_FUNC_ID,
485 DK_CV_INLINE_SITE_ID,
486 DK_CV_LOC,
487 DK_CV_LINETABLE,
488 DK_CV_INLINE_LINETABLE,
489 DK_CV_DEF_RANGE,
490 DK_CV_STRINGTABLE,
491 DK_CV_STRING,
492 DK_CV_FILECHECKSUMS,
493 DK_CV_FILECHECKSUM_OFFSET,
494 DK_CV_FPO_DATA,
495 DK_CFI_SECTIONS,
496 DK_CFI_STARTPROC,
497 DK_CFI_ENDPROC,
498 DK_CFI_DEF_CFA,
499 DK_CFI_DEF_CFA_OFFSET,
500 DK_CFI_ADJUST_CFA_OFFSET,
501 DK_CFI_DEF_CFA_REGISTER,
502 DK_CFI_LLVM_DEF_ASPACE_CFA,
503 DK_CFI_OFFSET,
504 DK_CFI_REL_OFFSET,
505 DK_CFI_PERSONALITY,
506 DK_CFI_LSDA,
507 DK_CFI_REMEMBER_STATE,
508 DK_CFI_RESTORE_STATE,
509 DK_CFI_SAME_VALUE,
510 DK_CFI_RESTORE,
511 DK_CFI_ESCAPE,
512 DK_CFI_RETURN_COLUMN,
513 DK_CFI_SIGNAL_FRAME,
514 DK_CFI_UNDEFINED,
515 DK_CFI_REGISTER,
516 DK_CFI_WINDOW_SAVE,
517 DK_CFI_B_KEY_FRAME,
518 DK_MACROS_ON,
519 DK_MACROS_OFF,
520 DK_ALTMACRO,
521 DK_NOALTMACRO,
522 DK_MACRO,
523 DK_EXITM,
524 DK_ENDM,
525 DK_ENDMACRO,
526 DK_PURGEM,
527 DK_SLEB128,
528 DK_ULEB128,
529 DK_ERR,
530 DK_ERROR,
531 DK_WARNING,
532 DK_PRINT,
533 DK_ADDRSIG,
534 DK_ADDRSIG_SYM,
535 DK_PSEUDO_PROBE,
536 DK_LTO_DISCARD,
537 DK_END
538 };
539
540 /// Maps directive name --> DirectiveKind enum, for
541 /// directives parsed by this class.
542 StringMap<DirectiveKind> DirectiveKindMap;
543
544 // Codeview def_range type parsing.
545 enum CVDefRangeType {
546 CVDR_DEFRANGE = 0, // Placeholder
547 CVDR_DEFRANGE_REGISTER,
548 CVDR_DEFRANGE_FRAMEPOINTER_REL,
549 CVDR_DEFRANGE_SUBFIELD_REGISTER,
550 CVDR_DEFRANGE_REGISTER_REL
551 };
552
553 /// Maps Codeview def_range types --> CVDefRangeType enum, for
554 /// Codeview def_range types parsed by this class.
555 StringMap<CVDefRangeType> CVDefRangeTypeMap;
556
557 // ".ascii", ".asciz", ".string"
558 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
559 bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc"
560 bool parseDirectiveValue(StringRef IDVal,
561 unsigned Size); // ".byte", ".long", ...
562 bool parseDirectiveOctaValue(StringRef IDVal); // ".octa", ...
563 bool parseDirectiveRealValue(StringRef IDVal,
564 const fltSemantics &); // ".single", ...
565 bool parseDirectiveFill(); // ".fill"
566 bool parseDirectiveZero(); // ".zero"
567 // ".set", ".equ", ".equiv"
568 bool parseDirectiveSet(StringRef IDVal, bool allow_redef);
569 bool parseDirectiveOrg(); // ".org"
570 // ".align{,32}", ".p2align{,w,l}"
571 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize);
572
573 // ".file", ".line", ".loc", ".stabs"
574 bool parseDirectiveFile(SMLoc DirectiveLoc);
575 bool parseDirectiveLine();
576 bool parseDirectiveLoc();
577 bool parseDirectiveStabs();
578
579 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable",
580 // ".cv_inline_linetable", ".cv_def_range", ".cv_string"
581 bool parseDirectiveCVFile();
582 bool parseDirectiveCVFuncId();
583 bool parseDirectiveCVInlineSiteId();
584 bool parseDirectiveCVLoc();
585 bool parseDirectiveCVLinetable();
586 bool parseDirectiveCVInlineLinetable();
587 bool parseDirectiveCVDefRange();
588 bool parseDirectiveCVString();
589 bool parseDirectiveCVStringTable();
590 bool parseDirectiveCVFileChecksums();
591 bool parseDirectiveCVFileChecksumOffset();
592 bool parseDirectiveCVFPOData();
593
594 // .cfi directives
595 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc);
596 bool parseDirectiveCFIWindowSave();
597 bool parseDirectiveCFISections();
598 bool parseDirectiveCFIStartProc();
599 bool parseDirectiveCFIEndProc();
600 bool parseDirectiveCFIDefCfaOffset();
601 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc);
602 bool parseDirectiveCFIAdjustCfaOffset();
603 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc);
604 bool parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc);
605 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc);
606 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc);
607 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality);
608 bool parseDirectiveCFIRememberState();
609 bool parseDirectiveCFIRestoreState();
610 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc);
611 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc);
612 bool parseDirectiveCFIEscape();
613 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc);
614 bool parseDirectiveCFISignalFrame();
615 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc);
616
617 // macro directives
618 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc);
619 bool parseDirectiveExitMacro(StringRef Directive);
620 bool parseDirectiveEndMacro(StringRef Directive);
621 bool parseDirectiveMacro(SMLoc DirectiveLoc);
622 bool parseDirectiveMacrosOnOff(StringRef Directive);
623 // alternate macro mode directives
624 bool parseDirectiveAltmacro(StringRef Directive);
625 // ".bundle_align_mode"
626 bool parseDirectiveBundleAlignMode();
627 // ".bundle_lock"
628 bool parseDirectiveBundleLock();
629 // ".bundle_unlock"
630 bool parseDirectiveBundleUnlock();
631
632 // ".space", ".skip"
633 bool parseDirectiveSpace(StringRef IDVal);
634
635 // ".dcb"
636 bool parseDirectiveDCB(StringRef IDVal, unsigned Size);
637 bool parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &);
638 // ".ds"
639 bool parseDirectiveDS(StringRef IDVal, unsigned Size);
640
641 // .sleb128 (Signed=true) and .uleb128 (Signed=false)
642 bool parseDirectiveLEB128(bool Signed);
643
644 /// Parse a directive like ".globl" which
645 /// accepts a single symbol (which should be a label or an external).
646 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr);
647
648 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
649
650 bool parseDirectiveAbort(); // ".abort"
651 bool parseDirectiveInclude(); // ".include"
652 bool parseDirectiveIncbin(); // ".incbin"
653
654 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne"
655 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
656 // ".ifb" or ".ifnb", depending on ExpectBlank.
657 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
658 // ".ifc" or ".ifnc", depending on ExpectEqual.
659 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
660 // ".ifeqs" or ".ifnes", depending on ExpectEqual.
661 bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual);
662 // ".ifdef" or ".ifndef", depending on expect_defined
663 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
664 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
665 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else"
666 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
667 bool parseEscapedString(std::string &Data) override;
668 bool parseAngleBracketString(std::string &Data) override;
669
670 const MCExpr *applyModifierToExpr(const MCExpr *E,
671 MCSymbolRefExpr::VariantKind Variant);
672
673 // Macro-like directives
674 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
675 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
676 raw_svector_ostream &OS);
677 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive);
678 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp"
679 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
680 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
681
682 // "_emit" or "__emit"
683 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info,
684 size_t Len);
685
686 // "align"
687 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info);
688
689 // "end"
690 bool parseDirectiveEnd(SMLoc DirectiveLoc);
691
692 // ".err" or ".error"
693 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage);
694
695 // ".warning"
696 bool parseDirectiveWarning(SMLoc DirectiveLoc);
697
698 // .print <double-quotes-string>
699 bool parseDirectivePrint(SMLoc DirectiveLoc);
700
701 // .pseudoprobe
702 bool parseDirectivePseudoProbe();
703
704 // ".lto_discard"
705 bool parseDirectiveLTODiscard();
706
707 // Directives to support address-significance tables.
708 bool parseDirectiveAddrsig();
709 bool parseDirectiveAddrsigSym();
710
711 void initializeDirectiveKindMap();
712 void initializeCVDefRangeTypeMap();
713 };
714
715 class HLASMAsmParser final : public AsmParser {
716 private:
717 MCAsmLexer &Lexer;
718 MCStreamer &Out;
719
lexLeadingSpaces()720 void lexLeadingSpaces() {
721 while (Lexer.is(AsmToken::Space))
722 Lexer.Lex();
723 }
724
725 bool parseAsHLASMLabel(ParseStatementInfo &Info, MCAsmParserSemaCallback *SI);
726 bool parseAsMachineInstruction(ParseStatementInfo &Info,
727 MCAsmParserSemaCallback *SI);
728
729 public:
HLASMAsmParser(SourceMgr & SM,MCContext & Ctx,MCStreamer & Out,const MCAsmInfo & MAI,unsigned CB=0)730 HLASMAsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
731 const MCAsmInfo &MAI, unsigned CB = 0)
732 : AsmParser(SM, Ctx, Out, MAI, CB), Lexer(getLexer()), Out(Out) {
733 Lexer.setSkipSpace(false);
734 Lexer.setAllowHashInIdentifier(true);
735 Lexer.setLexHLASMIntegers(true);
736 Lexer.setLexHLASMStrings(true);
737 }
738
~HLASMAsmParser()739 ~HLASMAsmParser() { Lexer.setSkipSpace(true); }
740
741 bool parseStatement(ParseStatementInfo &Info,
742 MCAsmParserSemaCallback *SI) override;
743 };
744
745 } // end anonymous namespace
746
747 namespace llvm {
748
749 extern MCAsmParserExtension *createDarwinAsmParser();
750 extern MCAsmParserExtension *createELFAsmParser();
751 extern MCAsmParserExtension *createCOFFAsmParser();
752 extern MCAsmParserExtension *createXCOFFAsmParser();
753 extern MCAsmParserExtension *createWasmAsmParser();
754
755 } // end namespace llvm
756
757 enum { DEFAULT_ADDRSPACE = 0 };
758
AsmParser(SourceMgr & SM,MCContext & Ctx,MCStreamer & Out,const MCAsmInfo & MAI,unsigned CB=0)759 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
760 const MCAsmInfo &MAI, unsigned CB = 0)
761 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM),
762 CurBuffer(CB ? CB : SM.getMainFileID()), MacrosEnabledFlag(true) {
763 HadError = false;
764 // Save the old handler.
765 SavedDiagHandler = SrcMgr.getDiagHandler();
766 SavedDiagContext = SrcMgr.getDiagContext();
767 // Set our own handler which calls the saved handler.
768 SrcMgr.setDiagHandler(DiagHandler, this);
769 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
770 // Make MCStreamer aware of the StartTokLoc for locations in diagnostics.
771 Out.setStartTokLocPtr(&StartTokLoc);
772
773 // Initialize the platform / file format parser.
774 switch (Ctx.getObjectFileType()) {
775 case MCContext::IsCOFF:
776 PlatformParser.reset(createCOFFAsmParser());
777 break;
778 case MCContext::IsMachO:
779 PlatformParser.reset(createDarwinAsmParser());
780 IsDarwin = true;
781 break;
782 case MCContext::IsELF:
783 PlatformParser.reset(createELFAsmParser());
784 break;
785 case MCContext::IsGOFF:
786 report_fatal_error("GOFFAsmParser support not implemented yet");
787 case MCContext::IsWasm:
788 PlatformParser.reset(createWasmAsmParser());
789 break;
790 case MCContext::IsXCOFF:
791 PlatformParser.reset(createXCOFFAsmParser());
792 break;
793 }
794
795 PlatformParser->Initialize(*this);
796 initializeDirectiveKindMap();
797 initializeCVDefRangeTypeMap();
798
799 NumOfMacroInstantiations = 0;
800 }
801
~AsmParser()802 AsmParser::~AsmParser() {
803 assert((HadError || ActiveMacros.empty()) &&
804 "Unexpected active macro instantiation!");
805
806 // Remove MCStreamer's reference to the parser SMLoc.
807 Out.setStartTokLocPtr(nullptr);
808 // Restore the saved diagnostics handler and context for use during
809 // finalization.
810 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext);
811 }
812
printMacroInstantiations()813 void AsmParser::printMacroInstantiations() {
814 // Print the active macro instantiation stack.
815 for (std::vector<MacroInstantiation *>::const_reverse_iterator
816 it = ActiveMacros.rbegin(),
817 ie = ActiveMacros.rend();
818 it != ie; ++it)
819 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
820 "while in macro instantiation");
821 }
822
Note(SMLoc L,const Twine & Msg,SMRange Range)823 void AsmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) {
824 printPendingErrors();
825 printMessage(L, SourceMgr::DK_Note, Msg, Range);
826 printMacroInstantiations();
827 }
828
Warning(SMLoc L,const Twine & Msg,SMRange Range)829 bool AsmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) {
830 if(getTargetParser().getTargetOptions().MCNoWarn)
831 return false;
832 if (getTargetParser().getTargetOptions().MCFatalWarnings)
833 return Error(L, Msg, Range);
834 printMessage(L, SourceMgr::DK_Warning, Msg, Range);
835 printMacroInstantiations();
836 return false;
837 }
838
printError(SMLoc L,const Twine & Msg,SMRange Range)839 bool AsmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) {
840 HadError = true;
841 printMessage(L, SourceMgr::DK_Error, Msg, Range);
842 printMacroInstantiations();
843 return true;
844 }
845
enterIncludeFile(const std::string & Filename)846 bool AsmParser::enterIncludeFile(const std::string &Filename) {
847 std::string IncludedFile;
848 unsigned NewBuf =
849 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
850 if (!NewBuf)
851 return true;
852
853 CurBuffer = NewBuf;
854 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
855 return false;
856 }
857
858 /// Process the specified .incbin file by searching for it in the include paths
859 /// then just emitting the byte contents of the file to the streamer. This
860 /// returns true on failure.
processIncbinFile(const std::string & Filename,int64_t Skip,const MCExpr * Count,SMLoc Loc)861 bool AsmParser::processIncbinFile(const std::string &Filename, int64_t Skip,
862 const MCExpr *Count, SMLoc Loc) {
863 std::string IncludedFile;
864 unsigned NewBuf =
865 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
866 if (!NewBuf)
867 return true;
868
869 // Pick up the bytes from the file and emit them.
870 StringRef Bytes = SrcMgr.getMemoryBuffer(NewBuf)->getBuffer();
871 Bytes = Bytes.drop_front(Skip);
872 if (Count) {
873 int64_t Res;
874 if (!Count->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr()))
875 return Error(Loc, "expected absolute expression");
876 if (Res < 0)
877 return Warning(Loc, "negative count has no effect");
878 Bytes = Bytes.take_front(Res);
879 }
880 getStreamer().emitBytes(Bytes);
881 return false;
882 }
883
jumpToLoc(SMLoc Loc,unsigned InBuffer)884 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) {
885 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc);
886 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(),
887 Loc.getPointer());
888 }
889
Lex()890 const AsmToken &AsmParser::Lex() {
891 if (Lexer.getTok().is(AsmToken::Error))
892 Error(Lexer.getErrLoc(), Lexer.getErr());
893
894 // if it's a end of statement with a comment in it
895 if (getTok().is(AsmToken::EndOfStatement)) {
896 // if this is a line comment output it.
897 if (!getTok().getString().empty() && getTok().getString().front() != '\n' &&
898 getTok().getString().front() != '\r' && MAI.preserveAsmComments())
899 Out.addExplicitComment(Twine(getTok().getString()));
900 }
901
902 const AsmToken *tok = &Lexer.Lex();
903
904 // Parse comments here to be deferred until end of next statement.
905 while (tok->is(AsmToken::Comment)) {
906 if (MAI.preserveAsmComments())
907 Out.addExplicitComment(Twine(tok->getString()));
908 tok = &Lexer.Lex();
909 }
910
911 if (tok->is(AsmToken::Eof)) {
912 // If this is the end of an included file, pop the parent file off the
913 // include stack.
914 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
915 if (ParentIncludeLoc != SMLoc()) {
916 jumpToLoc(ParentIncludeLoc);
917 return Lex();
918 }
919 }
920
921 return *tok;
922 }
923
enabledGenDwarfForAssembly()924 bool AsmParser::enabledGenDwarfForAssembly() {
925 // Check whether the user specified -g.
926 if (!getContext().getGenDwarfForAssembly())
927 return false;
928 // If we haven't encountered any .file directives (which would imply that
929 // the assembler source was produced with debug info already) then emit one
930 // describing the assembler source file itself.
931 if (getContext().getGenDwarfFileNumber() == 0) {
932 // Use the first #line directive for this, if any. It's preprocessed, so
933 // there is no checksum, and of course no source directive.
934 if (!FirstCppHashFilename.empty())
935 getContext().setMCLineTableRootFile(/*CUID=*/0,
936 getContext().getCompilationDir(),
937 FirstCppHashFilename,
938 /*Cksum=*/None, /*Source=*/None);
939 const MCDwarfFile &RootFile =
940 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile();
941 getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective(
942 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name,
943 RootFile.Checksum, RootFile.Source));
944 }
945 return true;
946 }
947
Run(bool NoInitialTextSection,bool NoFinalize)948 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
949 LTODiscardSymbols.clear();
950
951 // Create the initial section, if requested.
952 if (!NoInitialTextSection)
953 Out.InitSections(false);
954
955 // Prime the lexer.
956 Lex();
957
958 HadError = false;
959 AsmCond StartingCondState = TheCondState;
960 SmallVector<AsmRewrite, 4> AsmStrRewrites;
961
962 // If we are generating dwarf for assembly source files save the initial text
963 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't
964 // emitting any actual debug info yet and haven't had a chance to parse any
965 // embedded .file directives.)
966 if (getContext().getGenDwarfForAssembly()) {
967 MCSection *Sec = getStreamer().getCurrentSectionOnly();
968 if (!Sec->getBeginSymbol()) {
969 MCSymbol *SectionStartSym = getContext().createTempSymbol();
970 getStreamer().emitLabel(SectionStartSym);
971 Sec->setBeginSymbol(SectionStartSym);
972 }
973 bool InsertResult = getContext().addGenDwarfSection(Sec);
974 assert(InsertResult && ".text section should not have debug info yet");
975 (void)InsertResult;
976 }
977
978 getTargetParser().onBeginOfFile();
979
980 // While we have input, parse each statement.
981 while (Lexer.isNot(AsmToken::Eof)) {
982 ParseStatementInfo Info(&AsmStrRewrites);
983 bool Parsed = parseStatement(Info, nullptr);
984
985 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error
986 // for printing ErrMsg via Lex() only if no (presumably better) parser error
987 // exists.
988 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) {
989 Lex();
990 }
991
992 // parseStatement returned true so may need to emit an error.
993 printPendingErrors();
994
995 // Skipping to the next line if needed.
996 if (Parsed && !getLexer().isAtStartOfStatement())
997 eatToEndOfStatement();
998 }
999
1000 getTargetParser().onEndOfFile();
1001 printPendingErrors();
1002
1003 // All errors should have been emitted.
1004 assert(!hasPendingError() && "unexpected error from parseStatement");
1005
1006 getTargetParser().flushPendingInstructions(getStreamer());
1007
1008 if (TheCondState.TheCond != StartingCondState.TheCond ||
1009 TheCondState.Ignore != StartingCondState.Ignore)
1010 printError(getTok().getLoc(), "unmatched .ifs or .elses");
1011 // Check to see there are no empty DwarfFile slots.
1012 const auto &LineTables = getContext().getMCDwarfLineTables();
1013 if (!LineTables.empty()) {
1014 unsigned Index = 0;
1015 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) {
1016 if (File.Name.empty() && Index != 0)
1017 printError(getTok().getLoc(), "unassigned file number: " +
1018 Twine(Index) +
1019 " for .file directives");
1020 ++Index;
1021 }
1022 }
1023
1024 // Check to see that all assembler local symbols were actually defined.
1025 // Targets that don't do subsections via symbols may not want this, though,
1026 // so conservatively exclude them. Only do this if we're finalizing, though,
1027 // as otherwise we won't necessarilly have seen everything yet.
1028 if (!NoFinalize) {
1029 if (MAI.hasSubsectionsViaSymbols()) {
1030 for (const auto &TableEntry : getContext().getSymbols()) {
1031 MCSymbol *Sym = TableEntry.getValue();
1032 // Variable symbols may not be marked as defined, so check those
1033 // explicitly. If we know it's a variable, we have a definition for
1034 // the purposes of this check.
1035 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
1036 // FIXME: We would really like to refer back to where the symbol was
1037 // first referenced for a source location. We need to add something
1038 // to track that. Currently, we just point to the end of the file.
1039 printError(getTok().getLoc(), "assembler local symbol '" +
1040 Sym->getName() + "' not defined");
1041 }
1042 }
1043
1044 // Temporary symbols like the ones for directional jumps don't go in the
1045 // symbol table. They also need to be diagnosed in all (final) cases.
1046 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) {
1047 if (std::get<2>(LocSym)->isUndefined()) {
1048 // Reset the state of any "# line file" directives we've seen to the
1049 // context as it was at the diagnostic site.
1050 CppHashInfo = std::get<1>(LocSym);
1051 printError(std::get<0>(LocSym), "directional label undefined");
1052 }
1053 }
1054 }
1055
1056 // Finalize the output stream if there are no errors and if the client wants
1057 // us to.
1058 if (!HadError && !NoFinalize)
1059 Out.Finish(Lexer.getLoc());
1060
1061 return HadError || getContext().hadError();
1062 }
1063
checkForValidSection()1064 bool AsmParser::checkForValidSection() {
1065 if (!ParsingMSInlineAsm && !getStreamer().getCurrentSectionOnly()) {
1066 Out.InitSections(false);
1067 return Error(getTok().getLoc(),
1068 "expected section directive before assembly directive");
1069 }
1070 return false;
1071 }
1072
1073 /// Throw away the rest of the line for testing purposes.
eatToEndOfStatement()1074 void AsmParser::eatToEndOfStatement() {
1075 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
1076 Lexer.Lex();
1077
1078 // Eat EOL.
1079 if (Lexer.is(AsmToken::EndOfStatement))
1080 Lexer.Lex();
1081 }
1082
parseStringToEndOfStatement()1083 StringRef AsmParser::parseStringToEndOfStatement() {
1084 const char *Start = getTok().getLoc().getPointer();
1085
1086 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
1087 Lexer.Lex();
1088
1089 const char *End = getTok().getLoc().getPointer();
1090 return StringRef(Start, End - Start);
1091 }
1092
parseStringToComma()1093 StringRef AsmParser::parseStringToComma() {
1094 const char *Start = getTok().getLoc().getPointer();
1095
1096 while (Lexer.isNot(AsmToken::EndOfStatement) &&
1097 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof))
1098 Lexer.Lex();
1099
1100 const char *End = getTok().getLoc().getPointer();
1101 return StringRef(Start, End - Start);
1102 }
1103
1104 /// Parse a paren expression and return it.
1105 /// NOTE: This assumes the leading '(' has already been consumed.
1106 ///
1107 /// parenexpr ::= expr)
1108 ///
parseParenExpr(const MCExpr * & Res,SMLoc & EndLoc)1109 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
1110 if (parseExpression(Res))
1111 return true;
1112 if (Lexer.isNot(AsmToken::RParen))
1113 return TokError("expected ')' in parentheses expression");
1114 EndLoc = Lexer.getTok().getEndLoc();
1115 Lex();
1116 return false;
1117 }
1118
1119 /// Parse a bracket expression and return it.
1120 /// NOTE: This assumes the leading '[' has already been consumed.
1121 ///
1122 /// bracketexpr ::= expr]
1123 ///
parseBracketExpr(const MCExpr * & Res,SMLoc & EndLoc)1124 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
1125 if (parseExpression(Res))
1126 return true;
1127 EndLoc = getTok().getEndLoc();
1128 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression"))
1129 return true;
1130 return false;
1131 }
1132
1133 /// Parse a primary expression and return it.
1134 /// primaryexpr ::= (parenexpr
1135 /// primaryexpr ::= symbol
1136 /// primaryexpr ::= number
1137 /// primaryexpr ::= '.'
1138 /// primaryexpr ::= ~,+,- primaryexpr
parsePrimaryExpr(const MCExpr * & Res,SMLoc & EndLoc,AsmTypeInfo * TypeInfo)1139 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc,
1140 AsmTypeInfo *TypeInfo) {
1141 SMLoc FirstTokenLoc = getLexer().getLoc();
1142 AsmToken::TokenKind FirstTokenKind = Lexer.getKind();
1143 switch (FirstTokenKind) {
1144 default:
1145 return TokError("unknown token in expression");
1146 // If we have an error assume that we've already handled it.
1147 case AsmToken::Error:
1148 return true;
1149 case AsmToken::Exclaim:
1150 Lex(); // Eat the operator.
1151 if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1152 return true;
1153 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc);
1154 return false;
1155 case AsmToken::Dollar:
1156 case AsmToken::Star:
1157 case AsmToken::At:
1158 case AsmToken::String:
1159 case AsmToken::Identifier: {
1160 StringRef Identifier;
1161 if (parseIdentifier(Identifier)) {
1162 // We may have failed but '$'|'*' may be a valid token in context of
1163 // the current PC.
1164 if (getTok().is(AsmToken::Dollar) || getTok().is(AsmToken::Star)) {
1165 bool ShouldGenerateTempSymbol = false;
1166 if ((getTok().is(AsmToken::Dollar) && MAI.getDollarIsPC()) ||
1167 (getTok().is(AsmToken::Star) && MAI.getStarIsPC()))
1168 ShouldGenerateTempSymbol = true;
1169
1170 if (!ShouldGenerateTempSymbol)
1171 return Error(FirstTokenLoc, "invalid token in expression");
1172
1173 // Eat the '$'|'*' token.
1174 Lex();
1175 // This is either a '$'|'*' reference, which references the current PC.
1176 // Emit a temporary label to the streamer and refer to it.
1177 MCSymbol *Sym = Ctx.createTempSymbol();
1178 Out.emitLabel(Sym);
1179 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None,
1180 getContext());
1181 EndLoc = FirstTokenLoc;
1182 return false;
1183 }
1184 }
1185 // Parse symbol variant
1186 std::pair<StringRef, StringRef> Split;
1187 if (!MAI.useParensForSymbolVariant()) {
1188 if (FirstTokenKind == AsmToken::String) {
1189 if (Lexer.is(AsmToken::At)) {
1190 Lex(); // eat @
1191 SMLoc AtLoc = getLexer().getLoc();
1192 StringRef VName;
1193 if (parseIdentifier(VName))
1194 return Error(AtLoc, "expected symbol variant after '@'");
1195
1196 Split = std::make_pair(Identifier, VName);
1197 }
1198 } else {
1199 Split = Identifier.split('@');
1200 }
1201 } else if (Lexer.is(AsmToken::LParen)) {
1202 Lex(); // eat '('.
1203 StringRef VName;
1204 parseIdentifier(VName);
1205 // eat ')'.
1206 if (parseToken(AsmToken::RParen,
1207 "unexpected token in variant, expected ')'"))
1208 return true;
1209 Split = std::make_pair(Identifier, VName);
1210 }
1211
1212 EndLoc = SMLoc::getFromPointer(Identifier.end());
1213
1214 // This is a symbol reference.
1215 StringRef SymbolName = Identifier;
1216 if (SymbolName.empty())
1217 return Error(getLexer().getLoc(), "expected a symbol reference");
1218
1219 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
1220
1221 // Lookup the symbol variant if used.
1222 if (!Split.second.empty()) {
1223 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
1224 if (Variant != MCSymbolRefExpr::VK_Invalid) {
1225 SymbolName = Split.first;
1226 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) {
1227 Variant = MCSymbolRefExpr::VK_None;
1228 } else {
1229 return Error(SMLoc::getFromPointer(Split.second.begin()),
1230 "invalid variant '" + Split.second + "'");
1231 }
1232 }
1233
1234 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName);
1235 if (!Sym)
1236 Sym = getContext().getOrCreateSymbol(
1237 MAI.shouldEmitLabelsInUpperCase() ? SymbolName.upper() : SymbolName);
1238
1239 // If this is an absolute variable reference, substitute it now to preserve
1240 // semantics in the face of reassignment.
1241 if (Sym->isVariable()) {
1242 auto V = Sym->getVariableValue(/*SetUsed*/ false);
1243 bool DoInline = isa<MCConstantExpr>(V) && !Variant;
1244 if (auto TV = dyn_cast<MCTargetExpr>(V))
1245 DoInline = TV->inlineAssignedExpr();
1246 if (DoInline) {
1247 if (Variant)
1248 return Error(EndLoc, "unexpected modifier on variable reference");
1249 Res = Sym->getVariableValue(/*SetUsed*/ false);
1250 return false;
1251 }
1252 }
1253
1254 // Otherwise create a symbol ref.
1255 Res = MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc);
1256 return false;
1257 }
1258 case AsmToken::BigNum:
1259 return TokError("literal value out of range for directive");
1260 case AsmToken::Integer: {
1261 SMLoc Loc = getTok().getLoc();
1262 int64_t IntVal = getTok().getIntVal();
1263 Res = MCConstantExpr::create(IntVal, getContext());
1264 EndLoc = Lexer.getTok().getEndLoc();
1265 Lex(); // Eat token.
1266 // Look for 'b' or 'f' following an Integer as a directional label
1267 if (Lexer.getKind() == AsmToken::Identifier) {
1268 StringRef IDVal = getTok().getString();
1269 // Lookup the symbol variant if used.
1270 std::pair<StringRef, StringRef> Split = IDVal.split('@');
1271 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
1272 if (Split.first.size() != IDVal.size()) {
1273 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
1274 if (Variant == MCSymbolRefExpr::VK_Invalid)
1275 return TokError("invalid variant '" + Split.second + "'");
1276 IDVal = Split.first;
1277 }
1278 if (IDVal == "f" || IDVal == "b") {
1279 MCSymbol *Sym =
1280 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b");
1281 Res = MCSymbolRefExpr::create(Sym, Variant, getContext());
1282 if (IDVal == "b" && Sym->isUndefined())
1283 return Error(Loc, "directional label undefined");
1284 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym));
1285 EndLoc = Lexer.getTok().getEndLoc();
1286 Lex(); // Eat identifier.
1287 }
1288 }
1289 return false;
1290 }
1291 case AsmToken::Real: {
1292 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString());
1293 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
1294 Res = MCConstantExpr::create(IntVal, getContext());
1295 EndLoc = Lexer.getTok().getEndLoc();
1296 Lex(); // Eat token.
1297 return false;
1298 }
1299 case AsmToken::Dot: {
1300 if (!MAI.getDotIsPC())
1301 return TokError("cannot use . as current PC");
1302
1303 // This is a '.' reference, which references the current PC. Emit a
1304 // temporary label to the streamer and refer to it.
1305 MCSymbol *Sym = Ctx.createTempSymbol();
1306 Out.emitLabel(Sym);
1307 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext());
1308 EndLoc = Lexer.getTok().getEndLoc();
1309 Lex(); // Eat identifier.
1310 return false;
1311 }
1312 case AsmToken::LParen:
1313 Lex(); // Eat the '('.
1314 return parseParenExpr(Res, EndLoc);
1315 case AsmToken::LBrac:
1316 if (!PlatformParser->HasBracketExpressions())
1317 return TokError("brackets expression not supported on this target");
1318 Lex(); // Eat the '['.
1319 return parseBracketExpr(Res, EndLoc);
1320 case AsmToken::Minus:
1321 Lex(); // Eat the operator.
1322 if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1323 return true;
1324 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc);
1325 return false;
1326 case AsmToken::Plus:
1327 Lex(); // Eat the operator.
1328 if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1329 return true;
1330 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc);
1331 return false;
1332 case AsmToken::Tilde:
1333 Lex(); // Eat the operator.
1334 if (parsePrimaryExpr(Res, EndLoc, TypeInfo))
1335 return true;
1336 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc);
1337 return false;
1338 // MIPS unary expression operators. The lexer won't generate these tokens if
1339 // MCAsmInfo::HasMipsExpressions is false for the target.
1340 case AsmToken::PercentCall16:
1341 case AsmToken::PercentCall_Hi:
1342 case AsmToken::PercentCall_Lo:
1343 case AsmToken::PercentDtprel_Hi:
1344 case AsmToken::PercentDtprel_Lo:
1345 case AsmToken::PercentGot:
1346 case AsmToken::PercentGot_Disp:
1347 case AsmToken::PercentGot_Hi:
1348 case AsmToken::PercentGot_Lo:
1349 case AsmToken::PercentGot_Ofst:
1350 case AsmToken::PercentGot_Page:
1351 case AsmToken::PercentGottprel:
1352 case AsmToken::PercentGp_Rel:
1353 case AsmToken::PercentHi:
1354 case AsmToken::PercentHigher:
1355 case AsmToken::PercentHighest:
1356 case AsmToken::PercentLo:
1357 case AsmToken::PercentNeg:
1358 case AsmToken::PercentPcrel_Hi:
1359 case AsmToken::PercentPcrel_Lo:
1360 case AsmToken::PercentTlsgd:
1361 case AsmToken::PercentTlsldm:
1362 case AsmToken::PercentTprel_Hi:
1363 case AsmToken::PercentTprel_Lo:
1364 Lex(); // Eat the operator.
1365 if (Lexer.isNot(AsmToken::LParen))
1366 return TokError("expected '(' after operator");
1367 Lex(); // Eat the operator.
1368 if (parseExpression(Res, EndLoc))
1369 return true;
1370 if (Lexer.isNot(AsmToken::RParen))
1371 return TokError("expected ')'");
1372 Lex(); // Eat the operator.
1373 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx);
1374 return !Res;
1375 }
1376 }
1377
parseExpression(const MCExpr * & Res)1378 bool AsmParser::parseExpression(const MCExpr *&Res) {
1379 SMLoc EndLoc;
1380 return parseExpression(Res, EndLoc);
1381 }
1382
1383 const MCExpr *
applyModifierToExpr(const MCExpr * E,MCSymbolRefExpr::VariantKind Variant)1384 AsmParser::applyModifierToExpr(const MCExpr *E,
1385 MCSymbolRefExpr::VariantKind Variant) {
1386 // Ask the target implementation about this expression first.
1387 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx);
1388 if (NewE)
1389 return NewE;
1390 // Recurse over the given expression, rebuilding it to apply the given variant
1391 // if there is exactly one symbol.
1392 switch (E->getKind()) {
1393 case MCExpr::Target:
1394 case MCExpr::Constant:
1395 return nullptr;
1396
1397 case MCExpr::SymbolRef: {
1398 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
1399
1400 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
1401 TokError("invalid variant on expression '" + getTok().getIdentifier() +
1402 "' (already modified)");
1403 return E;
1404 }
1405
1406 return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext());
1407 }
1408
1409 case MCExpr::Unary: {
1410 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
1411 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant);
1412 if (!Sub)
1413 return nullptr;
1414 return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext());
1415 }
1416
1417 case MCExpr::Binary: {
1418 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
1419 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant);
1420 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant);
1421
1422 if (!LHS && !RHS)
1423 return nullptr;
1424
1425 if (!LHS)
1426 LHS = BE->getLHS();
1427 if (!RHS)
1428 RHS = BE->getRHS();
1429
1430 return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext());
1431 }
1432 }
1433
1434 llvm_unreachable("Invalid expression kind!");
1435 }
1436
1437 /// This function checks if the next token is <string> type or arithmetic.
1438 /// string that begin with character '<' must end with character '>'.
1439 /// otherwise it is arithmetics.
1440 /// If the function returns a 'true' value,
1441 /// the End argument will be filled with the last location pointed to the '>'
1442 /// character.
1443
1444 /// There is a gap between the AltMacro's documentation and the single quote
1445 /// implementation. GCC does not fully support this feature and so we will not
1446 /// support it.
1447 /// TODO: Adding single quote as a string.
isAngleBracketString(SMLoc & StrLoc,SMLoc & EndLoc)1448 static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) {
1449 assert((StrLoc.getPointer() != nullptr) &&
1450 "Argument to the function cannot be a NULL value");
1451 const char *CharPtr = StrLoc.getPointer();
1452 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') &&
1453 (*CharPtr != '\0')) {
1454 if (*CharPtr == '!')
1455 CharPtr++;
1456 CharPtr++;
1457 }
1458 if (*CharPtr == '>') {
1459 EndLoc = StrLoc.getFromPointer(CharPtr + 1);
1460 return true;
1461 }
1462 return false;
1463 }
1464
1465 /// creating a string without the escape characters '!'.
angleBracketString(StringRef AltMacroStr)1466 static std::string angleBracketString(StringRef AltMacroStr) {
1467 std::string Res;
1468 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) {
1469 if (AltMacroStr[Pos] == '!')
1470 Pos++;
1471 Res += AltMacroStr[Pos];
1472 }
1473 return Res;
1474 }
1475
1476 /// Parse an expression and return it.
1477 ///
1478 /// expr ::= expr &&,|| expr -> lowest.
1479 /// expr ::= expr |,^,&,! expr
1480 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr
1481 /// expr ::= expr <<,>> expr
1482 /// expr ::= expr +,- expr
1483 /// expr ::= expr *,/,% expr -> highest.
1484 /// expr ::= primaryexpr
1485 ///
parseExpression(const MCExpr * & Res,SMLoc & EndLoc)1486 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1487 // Parse the expression.
1488 Res = nullptr;
1489 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) ||
1490 parseBinOpRHS(1, Res, EndLoc))
1491 return true;
1492
1493 // As a special case, we support 'a op b @ modifier' by rewriting the
1494 // expression to include the modifier. This is inefficient, but in general we
1495 // expect users to use 'a@modifier op b'.
1496 if (Lexer.getKind() == AsmToken::At) {
1497 Lex();
1498
1499 if (Lexer.isNot(AsmToken::Identifier))
1500 return TokError("unexpected symbol modifier following '@'");
1501
1502 MCSymbolRefExpr::VariantKind Variant =
1503 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
1504 if (Variant == MCSymbolRefExpr::VK_Invalid)
1505 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
1506
1507 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant);
1508 if (!ModifiedRes) {
1509 return TokError("invalid modifier '" + getTok().getIdentifier() +
1510 "' (no symbols present)");
1511 }
1512
1513 Res = ModifiedRes;
1514 Lex();
1515 }
1516
1517 // Try to constant fold it up front, if possible. Do not exploit
1518 // assembler here.
1519 int64_t Value;
1520 if (Res->evaluateAsAbsolute(Value))
1521 Res = MCConstantExpr::create(Value, getContext());
1522
1523 return false;
1524 }
1525
parseParenExpression(const MCExpr * & Res,SMLoc & EndLoc)1526 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1527 Res = nullptr;
1528 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc);
1529 }
1530
parseParenExprOfDepth(unsigned ParenDepth,const MCExpr * & Res,SMLoc & EndLoc)1531 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
1532 SMLoc &EndLoc) {
1533 if (parseParenExpr(Res, EndLoc))
1534 return true;
1535
1536 for (; ParenDepth > 0; --ParenDepth) {
1537 if (parseBinOpRHS(1, Res, EndLoc))
1538 return true;
1539
1540 // We don't Lex() the last RParen.
1541 // This is the same behavior as parseParenExpression().
1542 if (ParenDepth - 1 > 0) {
1543 EndLoc = getTok().getEndLoc();
1544 if (parseToken(AsmToken::RParen,
1545 "expected ')' in parentheses expression"))
1546 return true;
1547 }
1548 }
1549 return false;
1550 }
1551
parseAbsoluteExpression(int64_t & Res)1552 bool AsmParser::parseAbsoluteExpression(int64_t &Res) {
1553 const MCExpr *Expr;
1554
1555 SMLoc StartLoc = Lexer.getLoc();
1556 if (parseExpression(Expr))
1557 return true;
1558
1559 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr()))
1560 return Error(StartLoc, "expected absolute expression");
1561
1562 return false;
1563 }
1564
getDarwinBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1565 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K,
1566 MCBinaryExpr::Opcode &Kind,
1567 bool ShouldUseLogicalShr) {
1568 switch (K) {
1569 default:
1570 return 0; // not a binop.
1571
1572 // Lowest Precedence: &&, ||
1573 case AsmToken::AmpAmp:
1574 Kind = MCBinaryExpr::LAnd;
1575 return 1;
1576 case AsmToken::PipePipe:
1577 Kind = MCBinaryExpr::LOr;
1578 return 1;
1579
1580 // Low Precedence: |, &, ^
1581 case AsmToken::Pipe:
1582 Kind = MCBinaryExpr::Or;
1583 return 2;
1584 case AsmToken::Caret:
1585 Kind = MCBinaryExpr::Xor;
1586 return 2;
1587 case AsmToken::Amp:
1588 Kind = MCBinaryExpr::And;
1589 return 2;
1590
1591 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
1592 case AsmToken::EqualEqual:
1593 Kind = MCBinaryExpr::EQ;
1594 return 3;
1595 case AsmToken::ExclaimEqual:
1596 case AsmToken::LessGreater:
1597 Kind = MCBinaryExpr::NE;
1598 return 3;
1599 case AsmToken::Less:
1600 Kind = MCBinaryExpr::LT;
1601 return 3;
1602 case AsmToken::LessEqual:
1603 Kind = MCBinaryExpr::LTE;
1604 return 3;
1605 case AsmToken::Greater:
1606 Kind = MCBinaryExpr::GT;
1607 return 3;
1608 case AsmToken::GreaterEqual:
1609 Kind = MCBinaryExpr::GTE;
1610 return 3;
1611
1612 // Intermediate Precedence: <<, >>
1613 case AsmToken::LessLess:
1614 Kind = MCBinaryExpr::Shl;
1615 return 4;
1616 case AsmToken::GreaterGreater:
1617 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1618 return 4;
1619
1620 // High Intermediate Precedence: +, -
1621 case AsmToken::Plus:
1622 Kind = MCBinaryExpr::Add;
1623 return 5;
1624 case AsmToken::Minus:
1625 Kind = MCBinaryExpr::Sub;
1626 return 5;
1627
1628 // Highest Precedence: *, /, %
1629 case AsmToken::Star:
1630 Kind = MCBinaryExpr::Mul;
1631 return 6;
1632 case AsmToken::Slash:
1633 Kind = MCBinaryExpr::Div;
1634 return 6;
1635 case AsmToken::Percent:
1636 Kind = MCBinaryExpr::Mod;
1637 return 6;
1638 }
1639 }
1640
getGNUBinOpPrecedence(const MCAsmInfo & MAI,AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind,bool ShouldUseLogicalShr)1641 static unsigned getGNUBinOpPrecedence(const MCAsmInfo &MAI,
1642 AsmToken::TokenKind K,
1643 MCBinaryExpr::Opcode &Kind,
1644 bool ShouldUseLogicalShr) {
1645 switch (K) {
1646 default:
1647 return 0; // not a binop.
1648
1649 // Lowest Precedence: &&, ||
1650 case AsmToken::AmpAmp:
1651 Kind = MCBinaryExpr::LAnd;
1652 return 2;
1653 case AsmToken::PipePipe:
1654 Kind = MCBinaryExpr::LOr;
1655 return 1;
1656
1657 // Low Precedence: ==, !=, <>, <, <=, >, >=
1658 case AsmToken::EqualEqual:
1659 Kind = MCBinaryExpr::EQ;
1660 return 3;
1661 case AsmToken::ExclaimEqual:
1662 case AsmToken::LessGreater:
1663 Kind = MCBinaryExpr::NE;
1664 return 3;
1665 case AsmToken::Less:
1666 Kind = MCBinaryExpr::LT;
1667 return 3;
1668 case AsmToken::LessEqual:
1669 Kind = MCBinaryExpr::LTE;
1670 return 3;
1671 case AsmToken::Greater:
1672 Kind = MCBinaryExpr::GT;
1673 return 3;
1674 case AsmToken::GreaterEqual:
1675 Kind = MCBinaryExpr::GTE;
1676 return 3;
1677
1678 // Low Intermediate Precedence: +, -
1679 case AsmToken::Plus:
1680 Kind = MCBinaryExpr::Add;
1681 return 4;
1682 case AsmToken::Minus:
1683 Kind = MCBinaryExpr::Sub;
1684 return 4;
1685
1686 // High Intermediate Precedence: |, !, &, ^
1687 //
1688 case AsmToken::Pipe:
1689 Kind = MCBinaryExpr::Or;
1690 return 5;
1691 case AsmToken::Exclaim:
1692 // Hack to support ARM compatible aliases (implied 'sp' operand in 'srs*'
1693 // instructions like 'srsda #31!') and not parse ! as an infix operator.
1694 if (MAI.getCommentString() == "@")
1695 return 0;
1696 Kind = MCBinaryExpr::OrNot;
1697 return 5;
1698 case AsmToken::Caret:
1699 Kind = MCBinaryExpr::Xor;
1700 return 5;
1701 case AsmToken::Amp:
1702 Kind = MCBinaryExpr::And;
1703 return 5;
1704
1705 // Highest Precedence: *, /, %, <<, >>
1706 case AsmToken::Star:
1707 Kind = MCBinaryExpr::Mul;
1708 return 6;
1709 case AsmToken::Slash:
1710 Kind = MCBinaryExpr::Div;
1711 return 6;
1712 case AsmToken::Percent:
1713 Kind = MCBinaryExpr::Mod;
1714 return 6;
1715 case AsmToken::LessLess:
1716 Kind = MCBinaryExpr::Shl;
1717 return 6;
1718 case AsmToken::GreaterGreater:
1719 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
1720 return 6;
1721 }
1722 }
1723
getBinOpPrecedence(AsmToken::TokenKind K,MCBinaryExpr::Opcode & Kind)1724 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K,
1725 MCBinaryExpr::Opcode &Kind) {
1726 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr();
1727 return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr)
1728 : getGNUBinOpPrecedence(MAI, K, Kind, ShouldUseLogicalShr);
1729 }
1730
1731 /// Parse all binary operators with precedence >= 'Precedence'.
1732 /// Res contains the LHS of the expression on input.
parseBinOpRHS(unsigned Precedence,const MCExpr * & Res,SMLoc & EndLoc)1733 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1734 SMLoc &EndLoc) {
1735 SMLoc StartLoc = Lexer.getLoc();
1736 while (true) {
1737 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1738 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1739
1740 // If the next token is lower precedence than we are allowed to eat, return
1741 // successfully with what we ate already.
1742 if (TokPrec < Precedence)
1743 return false;
1744
1745 Lex();
1746
1747 // Eat the next primary expression.
1748 const MCExpr *RHS;
1749 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc))
1750 return true;
1751
1752 // If BinOp binds less tightly with RHS than the operator after RHS, let
1753 // the pending operator take RHS as its LHS.
1754 MCBinaryExpr::Opcode Dummy;
1755 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1756 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc))
1757 return true;
1758
1759 // Merge LHS and RHS according to operator.
1760 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc);
1761 }
1762 }
1763
1764 /// ParseStatement:
1765 /// ::= EndOfStatement
1766 /// ::= Label* Directive ...Operands... EndOfStatement
1767 /// ::= Label* Identifier OperandList* EndOfStatement
parseStatement(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)1768 bool AsmParser::parseStatement(ParseStatementInfo &Info,
1769 MCAsmParserSemaCallback *SI) {
1770 assert(!hasPendingError() && "parseStatement started with pending error");
1771 // Eat initial spaces and comments
1772 while (Lexer.is(AsmToken::Space))
1773 Lex();
1774 if (Lexer.is(AsmToken::EndOfStatement)) {
1775 // if this is a line comment we can drop it safely
1776 if (getTok().getString().empty() || getTok().getString().front() == '\r' ||
1777 getTok().getString().front() == '\n')
1778 Out.AddBlankLine();
1779 Lex();
1780 return false;
1781 }
1782 // Statements always start with an identifier.
1783 AsmToken ID = getTok();
1784 SMLoc IDLoc = ID.getLoc();
1785 StringRef IDVal;
1786 int64_t LocalLabelVal = -1;
1787 StartTokLoc = ID.getLoc();
1788 if (Lexer.is(AsmToken::HashDirective))
1789 return parseCppHashLineFilenameComment(IDLoc,
1790 !isInsideMacroInstantiation());
1791
1792 // Allow an integer followed by a ':' as a directional local label.
1793 if (Lexer.is(AsmToken::Integer)) {
1794 LocalLabelVal = getTok().getIntVal();
1795 if (LocalLabelVal < 0) {
1796 if (!TheCondState.Ignore) {
1797 Lex(); // always eat a token
1798 return Error(IDLoc, "unexpected token at start of statement");
1799 }
1800 IDVal = "";
1801 } else {
1802 IDVal = getTok().getString();
1803 Lex(); // Consume the integer token to be used as an identifier token.
1804 if (Lexer.getKind() != AsmToken::Colon) {
1805 if (!TheCondState.Ignore) {
1806 Lex(); // always eat a token
1807 return Error(IDLoc, "unexpected token at start of statement");
1808 }
1809 }
1810 }
1811 } else if (Lexer.is(AsmToken::Dot)) {
1812 // Treat '.' as a valid identifier in this context.
1813 Lex();
1814 IDVal = ".";
1815 } else if (Lexer.is(AsmToken::LCurly)) {
1816 // Treat '{' as a valid identifier in this context.
1817 Lex();
1818 IDVal = "{";
1819
1820 } else if (Lexer.is(AsmToken::RCurly)) {
1821 // Treat '}' as a valid identifier in this context.
1822 Lex();
1823 IDVal = "}";
1824 } else if (Lexer.is(AsmToken::Star) &&
1825 getTargetParser().starIsStartOfStatement()) {
1826 // Accept '*' as a valid start of statement.
1827 Lex();
1828 IDVal = "*";
1829 } else if (parseIdentifier(IDVal)) {
1830 if (!TheCondState.Ignore) {
1831 Lex(); // always eat a token
1832 return Error(IDLoc, "unexpected token at start of statement");
1833 }
1834 IDVal = "";
1835 }
1836
1837 // Handle conditional assembly here before checking for skipping. We
1838 // have to do this so that .endif isn't skipped in a ".if 0" block for
1839 // example.
1840 StringMap<DirectiveKind>::const_iterator DirKindIt =
1841 DirectiveKindMap.find(IDVal.lower());
1842 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end())
1843 ? DK_NO_DIRECTIVE
1844 : DirKindIt->getValue();
1845 switch (DirKind) {
1846 default:
1847 break;
1848 case DK_IF:
1849 case DK_IFEQ:
1850 case DK_IFGE:
1851 case DK_IFGT:
1852 case DK_IFLE:
1853 case DK_IFLT:
1854 case DK_IFNE:
1855 return parseDirectiveIf(IDLoc, DirKind);
1856 case DK_IFB:
1857 return parseDirectiveIfb(IDLoc, true);
1858 case DK_IFNB:
1859 return parseDirectiveIfb(IDLoc, false);
1860 case DK_IFC:
1861 return parseDirectiveIfc(IDLoc, true);
1862 case DK_IFEQS:
1863 return parseDirectiveIfeqs(IDLoc, true);
1864 case DK_IFNC:
1865 return parseDirectiveIfc(IDLoc, false);
1866 case DK_IFNES:
1867 return parseDirectiveIfeqs(IDLoc, false);
1868 case DK_IFDEF:
1869 return parseDirectiveIfdef(IDLoc, true);
1870 case DK_IFNDEF:
1871 case DK_IFNOTDEF:
1872 return parseDirectiveIfdef(IDLoc, false);
1873 case DK_ELSEIF:
1874 return parseDirectiveElseIf(IDLoc);
1875 case DK_ELSE:
1876 return parseDirectiveElse(IDLoc);
1877 case DK_ENDIF:
1878 return parseDirectiveEndIf(IDLoc);
1879 }
1880
1881 // Ignore the statement if in the middle of inactive conditional
1882 // (e.g. ".if 0").
1883 if (TheCondState.Ignore) {
1884 eatToEndOfStatement();
1885 return false;
1886 }
1887
1888 // FIXME: Recurse on local labels?
1889
1890 // See what kind of statement we have.
1891 switch (Lexer.getKind()) {
1892 case AsmToken::Colon: {
1893 if (!getTargetParser().isLabel(ID))
1894 break;
1895 if (checkForValidSection())
1896 return true;
1897
1898 // identifier ':' -> Label.
1899 Lex();
1900
1901 // Diagnose attempt to use '.' as a label.
1902 if (IDVal == ".")
1903 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1904
1905 // Diagnose attempt to use a variable as a label.
1906 //
1907 // FIXME: Diagnostics. Note the location of the definition as a label.
1908 // FIXME: This doesn't diagnose assignment to a symbol which has been
1909 // implicitly marked as external.
1910 MCSymbol *Sym;
1911 if (LocalLabelVal == -1) {
1912 if (ParsingMSInlineAsm && SI) {
1913 StringRef RewrittenLabel =
1914 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true);
1915 assert(!RewrittenLabel.empty() &&
1916 "We should have an internal name here.");
1917 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(),
1918 RewrittenLabel);
1919 IDVal = RewrittenLabel;
1920 }
1921 Sym = getContext().getOrCreateSymbol(IDVal);
1922 } else
1923 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal);
1924 // End of Labels should be treated as end of line for lexing
1925 // purposes but that information is not available to the Lexer who
1926 // does not understand Labels. This may cause us to see a Hash
1927 // here instead of a preprocessor line comment.
1928 if (getTok().is(AsmToken::Hash)) {
1929 StringRef CommentStr = parseStringToEndOfStatement();
1930 Lexer.Lex();
1931 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr));
1932 }
1933
1934 // Consume any end of statement token, if present, to avoid spurious
1935 // AddBlankLine calls().
1936 if (getTok().is(AsmToken::EndOfStatement)) {
1937 Lex();
1938 }
1939
1940 if (discardLTOSymbol(IDVal))
1941 return false;
1942
1943 getTargetParser().doBeforeLabelEmit(Sym);
1944
1945 // Emit the label.
1946 if (!getTargetParser().isParsingMSInlineAsm())
1947 Out.emitLabel(Sym, IDLoc);
1948
1949 // If we are generating dwarf for assembly source files then gather the
1950 // info to make a dwarf label entry for this label if needed.
1951 if (enabledGenDwarfForAssembly())
1952 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1953 IDLoc);
1954
1955 getTargetParser().onLabelParsed(Sym);
1956
1957 return false;
1958 }
1959
1960 case AsmToken::Equal:
1961 if (!getTargetParser().equalIsAsmAssignment())
1962 break;
1963 // identifier '=' ... -> assignment statement
1964 Lex();
1965
1966 return parseAssignment(IDVal, true);
1967
1968 default: // Normal instruction or directive.
1969 break;
1970 }
1971
1972 // If macros are enabled, check to see if this is a macro instantiation.
1973 if (areMacrosEnabled())
1974 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) {
1975 return handleMacroEntry(M, IDLoc);
1976 }
1977
1978 // Otherwise, we have a normal instruction or directive.
1979
1980 // Directives start with "."
1981 if (IDVal.startswith(".") && IDVal != ".") {
1982 // There are several entities interested in parsing directives:
1983 //
1984 // 1. The target-specific assembly parser. Some directives are target
1985 // specific or may potentially behave differently on certain targets.
1986 // 2. Asm parser extensions. For example, platform-specific parsers
1987 // (like the ELF parser) register themselves as extensions.
1988 // 3. The generic directive parser implemented by this class. These are
1989 // all the directives that behave in a target and platform independent
1990 // manner, or at least have a default behavior that's shared between
1991 // all targets and platforms.
1992
1993 getTargetParser().flushPendingInstructions(getStreamer());
1994
1995 SMLoc StartTokLoc = getTok().getLoc();
1996 bool TPDirectiveReturn = getTargetParser().ParseDirective(ID);
1997
1998 if (hasPendingError())
1999 return true;
2000 // Currently the return value should be true if we are
2001 // uninterested but as this is at odds with the standard parsing
2002 // convention (return true = error) we have instances of a parsed
2003 // directive that fails returning true as an error. Catch these
2004 // cases as best as possible errors here.
2005 if (TPDirectiveReturn && StartTokLoc != getTok().getLoc())
2006 return true;
2007 // Return if we did some parsing or believe we succeeded.
2008 if (!TPDirectiveReturn || StartTokLoc != getTok().getLoc())
2009 return false;
2010
2011 // Next, check the extension directive map to see if any extension has
2012 // registered itself to parse this directive.
2013 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
2014 ExtensionDirectiveMap.lookup(IDVal);
2015 if (Handler.first)
2016 return (*Handler.second)(Handler.first, IDVal, IDLoc);
2017
2018 // Finally, if no one else is interested in this directive, it must be
2019 // generic and familiar to this class.
2020 switch (DirKind) {
2021 default:
2022 break;
2023 case DK_SET:
2024 case DK_EQU:
2025 return parseDirectiveSet(IDVal, true);
2026 case DK_EQUIV:
2027 return parseDirectiveSet(IDVal, false);
2028 case DK_ASCII:
2029 return parseDirectiveAscii(IDVal, false);
2030 case DK_ASCIZ:
2031 case DK_STRING:
2032 return parseDirectiveAscii(IDVal, true);
2033 case DK_BYTE:
2034 case DK_DC_B:
2035 return parseDirectiveValue(IDVal, 1);
2036 case DK_DC:
2037 case DK_DC_W:
2038 case DK_SHORT:
2039 case DK_VALUE:
2040 case DK_2BYTE:
2041 return parseDirectiveValue(IDVal, 2);
2042 case DK_LONG:
2043 case DK_INT:
2044 case DK_4BYTE:
2045 case DK_DC_L:
2046 return parseDirectiveValue(IDVal, 4);
2047 case DK_QUAD:
2048 case DK_8BYTE:
2049 return parseDirectiveValue(IDVal, 8);
2050 case DK_DC_A:
2051 return parseDirectiveValue(
2052 IDVal, getContext().getAsmInfo()->getCodePointerSize());
2053 case DK_OCTA:
2054 return parseDirectiveOctaValue(IDVal);
2055 case DK_SINGLE:
2056 case DK_FLOAT:
2057 case DK_DC_S:
2058 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle());
2059 case DK_DOUBLE:
2060 case DK_DC_D:
2061 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble());
2062 case DK_ALIGN: {
2063 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
2064 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1);
2065 }
2066 case DK_ALIGN32: {
2067 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
2068 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4);
2069 }
2070 case DK_BALIGN:
2071 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
2072 case DK_BALIGNW:
2073 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
2074 case DK_BALIGNL:
2075 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
2076 case DK_P2ALIGN:
2077 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
2078 case DK_P2ALIGNW:
2079 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
2080 case DK_P2ALIGNL:
2081 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
2082 case DK_ORG:
2083 return parseDirectiveOrg();
2084 case DK_FILL:
2085 return parseDirectiveFill();
2086 case DK_ZERO:
2087 return parseDirectiveZero();
2088 case DK_EXTERN:
2089 eatToEndOfStatement(); // .extern is the default, ignore it.
2090 return false;
2091 case DK_GLOBL:
2092 case DK_GLOBAL:
2093 return parseDirectiveSymbolAttribute(MCSA_Global);
2094 case DK_LAZY_REFERENCE:
2095 return parseDirectiveSymbolAttribute(MCSA_LazyReference);
2096 case DK_NO_DEAD_STRIP:
2097 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
2098 case DK_SYMBOL_RESOLVER:
2099 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver);
2100 case DK_PRIVATE_EXTERN:
2101 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern);
2102 case DK_REFERENCE:
2103 return parseDirectiveSymbolAttribute(MCSA_Reference);
2104 case DK_WEAK_DEFINITION:
2105 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition);
2106 case DK_WEAK_REFERENCE:
2107 return parseDirectiveSymbolAttribute(MCSA_WeakReference);
2108 case DK_WEAK_DEF_CAN_BE_HIDDEN:
2109 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
2110 case DK_COLD:
2111 return parseDirectiveSymbolAttribute(MCSA_Cold);
2112 case DK_COMM:
2113 case DK_COMMON:
2114 return parseDirectiveComm(/*IsLocal=*/false);
2115 case DK_LCOMM:
2116 return parseDirectiveComm(/*IsLocal=*/true);
2117 case DK_ABORT:
2118 return parseDirectiveAbort();
2119 case DK_INCLUDE:
2120 return parseDirectiveInclude();
2121 case DK_INCBIN:
2122 return parseDirectiveIncbin();
2123 case DK_CODE16:
2124 case DK_CODE16GCC:
2125 return TokError(Twine(IDVal) +
2126 " not currently supported for this target");
2127 case DK_REPT:
2128 return parseDirectiveRept(IDLoc, IDVal);
2129 case DK_IRP:
2130 return parseDirectiveIrp(IDLoc);
2131 case DK_IRPC:
2132 return parseDirectiveIrpc(IDLoc);
2133 case DK_ENDR:
2134 return parseDirectiveEndr(IDLoc);
2135 case DK_BUNDLE_ALIGN_MODE:
2136 return parseDirectiveBundleAlignMode();
2137 case DK_BUNDLE_LOCK:
2138 return parseDirectiveBundleLock();
2139 case DK_BUNDLE_UNLOCK:
2140 return parseDirectiveBundleUnlock();
2141 case DK_SLEB128:
2142 return parseDirectiveLEB128(true);
2143 case DK_ULEB128:
2144 return parseDirectiveLEB128(false);
2145 case DK_SPACE:
2146 case DK_SKIP:
2147 return parseDirectiveSpace(IDVal);
2148 case DK_FILE:
2149 return parseDirectiveFile(IDLoc);
2150 case DK_LINE:
2151 return parseDirectiveLine();
2152 case DK_LOC:
2153 return parseDirectiveLoc();
2154 case DK_STABS:
2155 return parseDirectiveStabs();
2156 case DK_CV_FILE:
2157 return parseDirectiveCVFile();
2158 case DK_CV_FUNC_ID:
2159 return parseDirectiveCVFuncId();
2160 case DK_CV_INLINE_SITE_ID:
2161 return parseDirectiveCVInlineSiteId();
2162 case DK_CV_LOC:
2163 return parseDirectiveCVLoc();
2164 case DK_CV_LINETABLE:
2165 return parseDirectiveCVLinetable();
2166 case DK_CV_INLINE_LINETABLE:
2167 return parseDirectiveCVInlineLinetable();
2168 case DK_CV_DEF_RANGE:
2169 return parseDirectiveCVDefRange();
2170 case DK_CV_STRING:
2171 return parseDirectiveCVString();
2172 case DK_CV_STRINGTABLE:
2173 return parseDirectiveCVStringTable();
2174 case DK_CV_FILECHECKSUMS:
2175 return parseDirectiveCVFileChecksums();
2176 case DK_CV_FILECHECKSUM_OFFSET:
2177 return parseDirectiveCVFileChecksumOffset();
2178 case DK_CV_FPO_DATA:
2179 return parseDirectiveCVFPOData();
2180 case DK_CFI_SECTIONS:
2181 return parseDirectiveCFISections();
2182 case DK_CFI_STARTPROC:
2183 return parseDirectiveCFIStartProc();
2184 case DK_CFI_ENDPROC:
2185 return parseDirectiveCFIEndProc();
2186 case DK_CFI_DEF_CFA:
2187 return parseDirectiveCFIDefCfa(IDLoc);
2188 case DK_CFI_DEF_CFA_OFFSET:
2189 return parseDirectiveCFIDefCfaOffset();
2190 case DK_CFI_ADJUST_CFA_OFFSET:
2191 return parseDirectiveCFIAdjustCfaOffset();
2192 case DK_CFI_DEF_CFA_REGISTER:
2193 return parseDirectiveCFIDefCfaRegister(IDLoc);
2194 case DK_CFI_LLVM_DEF_ASPACE_CFA:
2195 return parseDirectiveCFILLVMDefAspaceCfa(IDLoc);
2196 case DK_CFI_OFFSET:
2197 return parseDirectiveCFIOffset(IDLoc);
2198 case DK_CFI_REL_OFFSET:
2199 return parseDirectiveCFIRelOffset(IDLoc);
2200 case DK_CFI_PERSONALITY:
2201 return parseDirectiveCFIPersonalityOrLsda(true);
2202 case DK_CFI_LSDA:
2203 return parseDirectiveCFIPersonalityOrLsda(false);
2204 case DK_CFI_REMEMBER_STATE:
2205 return parseDirectiveCFIRememberState();
2206 case DK_CFI_RESTORE_STATE:
2207 return parseDirectiveCFIRestoreState();
2208 case DK_CFI_SAME_VALUE:
2209 return parseDirectiveCFISameValue(IDLoc);
2210 case DK_CFI_RESTORE:
2211 return parseDirectiveCFIRestore(IDLoc);
2212 case DK_CFI_ESCAPE:
2213 return parseDirectiveCFIEscape();
2214 case DK_CFI_RETURN_COLUMN:
2215 return parseDirectiveCFIReturnColumn(IDLoc);
2216 case DK_CFI_SIGNAL_FRAME:
2217 return parseDirectiveCFISignalFrame();
2218 case DK_CFI_UNDEFINED:
2219 return parseDirectiveCFIUndefined(IDLoc);
2220 case DK_CFI_REGISTER:
2221 return parseDirectiveCFIRegister(IDLoc);
2222 case DK_CFI_WINDOW_SAVE:
2223 return parseDirectiveCFIWindowSave();
2224 case DK_MACROS_ON:
2225 case DK_MACROS_OFF:
2226 return parseDirectiveMacrosOnOff(IDVal);
2227 case DK_MACRO:
2228 return parseDirectiveMacro(IDLoc);
2229 case DK_ALTMACRO:
2230 case DK_NOALTMACRO:
2231 return parseDirectiveAltmacro(IDVal);
2232 case DK_EXITM:
2233 return parseDirectiveExitMacro(IDVal);
2234 case DK_ENDM:
2235 case DK_ENDMACRO:
2236 return parseDirectiveEndMacro(IDVal);
2237 case DK_PURGEM:
2238 return parseDirectivePurgeMacro(IDLoc);
2239 case DK_END:
2240 return parseDirectiveEnd(IDLoc);
2241 case DK_ERR:
2242 return parseDirectiveError(IDLoc, false);
2243 case DK_ERROR:
2244 return parseDirectiveError(IDLoc, true);
2245 case DK_WARNING:
2246 return parseDirectiveWarning(IDLoc);
2247 case DK_RELOC:
2248 return parseDirectiveReloc(IDLoc);
2249 case DK_DCB:
2250 case DK_DCB_W:
2251 return parseDirectiveDCB(IDVal, 2);
2252 case DK_DCB_B:
2253 return parseDirectiveDCB(IDVal, 1);
2254 case DK_DCB_D:
2255 return parseDirectiveRealDCB(IDVal, APFloat::IEEEdouble());
2256 case DK_DCB_L:
2257 return parseDirectiveDCB(IDVal, 4);
2258 case DK_DCB_S:
2259 return parseDirectiveRealDCB(IDVal, APFloat::IEEEsingle());
2260 case DK_DC_X:
2261 case DK_DCB_X:
2262 return TokError(Twine(IDVal) +
2263 " not currently supported for this target");
2264 case DK_DS:
2265 case DK_DS_W:
2266 return parseDirectiveDS(IDVal, 2);
2267 case DK_DS_B:
2268 return parseDirectiveDS(IDVal, 1);
2269 case DK_DS_D:
2270 return parseDirectiveDS(IDVal, 8);
2271 case DK_DS_L:
2272 case DK_DS_S:
2273 return parseDirectiveDS(IDVal, 4);
2274 case DK_DS_P:
2275 case DK_DS_X:
2276 return parseDirectiveDS(IDVal, 12);
2277 case DK_PRINT:
2278 return parseDirectivePrint(IDLoc);
2279 case DK_ADDRSIG:
2280 return parseDirectiveAddrsig();
2281 case DK_ADDRSIG_SYM:
2282 return parseDirectiveAddrsigSym();
2283 case DK_PSEUDO_PROBE:
2284 return parseDirectivePseudoProbe();
2285 case DK_LTO_DISCARD:
2286 return parseDirectiveLTODiscard();
2287 }
2288
2289 return Error(IDLoc, "unknown directive");
2290 }
2291
2292 // __asm _emit or __asm __emit
2293 if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" ||
2294 IDVal == "_EMIT" || IDVal == "__EMIT"))
2295 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size());
2296
2297 // __asm align
2298 if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN"))
2299 return parseDirectiveMSAlign(IDLoc, Info);
2300
2301 if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN"))
2302 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4);
2303 if (checkForValidSection())
2304 return true;
2305
2306 return parseAndMatchAndEmitTargetInstruction(Info, IDVal, ID, IDLoc);
2307 }
2308
parseAndMatchAndEmitTargetInstruction(ParseStatementInfo & Info,StringRef IDVal,AsmToken ID,SMLoc IDLoc)2309 bool AsmParser::parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info,
2310 StringRef IDVal,
2311 AsmToken ID,
2312 SMLoc IDLoc) {
2313 // Canonicalize the opcode to lower case.
2314 std::string OpcodeStr = IDVal.lower();
2315 ParseInstructionInfo IInfo(Info.AsmRewrites);
2316 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID,
2317 Info.ParsedOperands);
2318 Info.ParseError = ParseHadError;
2319
2320 // Dump the parsed representation, if requested.
2321 if (getShowParsedOperands()) {
2322 SmallString<256> Str;
2323 raw_svector_ostream OS(Str);
2324 OS << "parsed instruction: [";
2325 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
2326 if (i != 0)
2327 OS << ", ";
2328 Info.ParsedOperands[i]->print(OS);
2329 }
2330 OS << "]";
2331
2332 printMessage(IDLoc, SourceMgr::DK_Note, OS.str());
2333 }
2334
2335 // Fail even if ParseInstruction erroneously returns false.
2336 if (hasPendingError() || ParseHadError)
2337 return true;
2338
2339 // If we are generating dwarf for the current section then generate a .loc
2340 // directive for the instruction.
2341 if (!ParseHadError && enabledGenDwarfForAssembly() &&
2342 getContext().getGenDwarfSectionSyms().count(
2343 getStreamer().getCurrentSectionOnly())) {
2344 unsigned Line;
2345 if (ActiveMacros.empty())
2346 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
2347 else
2348 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc,
2349 ActiveMacros.front()->ExitBuffer);
2350
2351 // If we previously parsed a cpp hash file line comment then make sure the
2352 // current Dwarf File is for the CppHashFilename if not then emit the
2353 // Dwarf File table for it and adjust the line number for the .loc.
2354 if (!CppHashInfo.Filename.empty()) {
2355 unsigned FileNumber = getStreamer().emitDwarfFileDirective(
2356 0, StringRef(), CppHashInfo.Filename);
2357 getContext().setGenDwarfFileNumber(FileNumber);
2358
2359 unsigned CppHashLocLineNo =
2360 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf);
2361 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo);
2362 }
2363
2364 getStreamer().emitDwarfLocDirective(
2365 getContext().getGenDwarfFileNumber(), Line, 0,
2366 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0,
2367 StringRef());
2368 }
2369
2370 // If parsing succeeded, match the instruction.
2371 if (!ParseHadError) {
2372 uint64_t ErrorInfo;
2373 if (getTargetParser().MatchAndEmitInstruction(
2374 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo,
2375 getTargetParser().isParsingMSInlineAsm()))
2376 return true;
2377 }
2378 return false;
2379 }
2380
2381 // Parse and erase curly braces marking block start/end
2382 bool
parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> & AsmStrRewrites)2383 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) {
2384 // Identify curly brace marking block start/end
2385 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly))
2386 return false;
2387
2388 SMLoc StartLoc = Lexer.getLoc();
2389 Lex(); // Eat the brace
2390 if (Lexer.is(AsmToken::EndOfStatement))
2391 Lex(); // Eat EndOfStatement following the brace
2392
2393 // Erase the block start/end brace from the output asm string
2394 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() -
2395 StartLoc.getPointer());
2396 return true;
2397 }
2398
2399 /// parseCppHashLineFilenameComment as this:
2400 /// ::= # number "filename"
parseCppHashLineFilenameComment(SMLoc L,bool SaveLocInfo)2401 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo) {
2402 Lex(); // Eat the hash token.
2403 // Lexer only ever emits HashDirective if it fully formed if it's
2404 // done the checking already so this is an internal error.
2405 assert(getTok().is(AsmToken::Integer) &&
2406 "Lexing Cpp line comment: Expected Integer");
2407 int64_t LineNumber = getTok().getIntVal();
2408 Lex();
2409 assert(getTok().is(AsmToken::String) &&
2410 "Lexing Cpp line comment: Expected String");
2411 StringRef Filename = getTok().getString();
2412 Lex();
2413
2414 if (!SaveLocInfo)
2415 return false;
2416
2417 // Get rid of the enclosing quotes.
2418 Filename = Filename.substr(1, Filename.size() - 2);
2419
2420 // Save the SMLoc, Filename and LineNumber for later use by diagnostics
2421 // and possibly DWARF file info.
2422 CppHashInfo.Loc = L;
2423 CppHashInfo.Filename = Filename;
2424 CppHashInfo.LineNumber = LineNumber;
2425 CppHashInfo.Buf = CurBuffer;
2426 if (FirstCppHashFilename.empty())
2427 FirstCppHashFilename = Filename;
2428 return false;
2429 }
2430
2431 /// will use the last parsed cpp hash line filename comment
2432 /// for the Filename and LineNo if any in the diagnostic.
DiagHandler(const SMDiagnostic & Diag,void * Context)2433 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
2434 auto *Parser = static_cast<AsmParser *>(Context);
2435 raw_ostream &OS = errs();
2436
2437 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
2438 SMLoc DiagLoc = Diag.getLoc();
2439 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
2440 unsigned CppHashBuf =
2441 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc);
2442
2443 // Like SourceMgr::printMessage() we need to print the include stack if any
2444 // before printing the message.
2445 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
2446 if (!Parser->SavedDiagHandler && DiagCurBuffer &&
2447 DiagCurBuffer != DiagSrcMgr.getMainFileID()) {
2448 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
2449 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
2450 }
2451
2452 // If we have not parsed a cpp hash line filename comment or the source
2453 // manager changed or buffer changed (like in a nested include) then just
2454 // print the normal diagnostic using its Filename and LineNo.
2455 if (!Parser->CppHashInfo.LineNumber || DiagBuf != CppHashBuf) {
2456 if (Parser->SavedDiagHandler)
2457 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
2458 else
2459 Parser->getContext().diagnose(Diag);
2460 return;
2461 }
2462
2463 // Use the CppHashFilename and calculate a line number based on the
2464 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc
2465 // for the diagnostic.
2466 const std::string &Filename = std::string(Parser->CppHashInfo.Filename);
2467
2468 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
2469 int CppHashLocLineNo =
2470 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf);
2471 int LineNo =
2472 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo);
2473
2474 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo,
2475 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(),
2476 Diag.getLineContents(), Diag.getRanges());
2477
2478 if (Parser->SavedDiagHandler)
2479 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
2480 else
2481 Parser->getContext().diagnose(NewDiag);
2482 }
2483
2484 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
2485 // difference being that that function accepts '@' as part of identifiers and
2486 // we can't do that. AsmLexer.cpp should probably be changed to handle
2487 // '@' as a special case when needed.
isIdentifierChar(char c)2488 static bool isIdentifierChar(char c) {
2489 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' ||
2490 c == '.';
2491 }
2492
expandMacro(raw_svector_ostream & OS,StringRef Body,ArrayRef<MCAsmMacroParameter> Parameters,ArrayRef<MCAsmMacroArgument> A,bool EnableAtPseudoVariable,SMLoc L)2493 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
2494 ArrayRef<MCAsmMacroParameter> Parameters,
2495 ArrayRef<MCAsmMacroArgument> A,
2496 bool EnableAtPseudoVariable, SMLoc L) {
2497 unsigned NParameters = Parameters.size();
2498 bool HasVararg = NParameters ? Parameters.back().Vararg : false;
2499 if ((!IsDarwin || NParameters != 0) && NParameters != A.size())
2500 return Error(L, "Wrong number of arguments");
2501
2502 // A macro without parameters is handled differently on Darwin:
2503 // gas accepts no arguments and does no substitutions
2504 while (!Body.empty()) {
2505 // Scan for the next substitution.
2506 std::size_t End = Body.size(), Pos = 0;
2507 for (; Pos != End; ++Pos) {
2508 // Check for a substitution or escape.
2509 if (IsDarwin && !NParameters) {
2510 // This macro has no parameters, look for $0, $1, etc.
2511 if (Body[Pos] != '$' || Pos + 1 == End)
2512 continue;
2513
2514 char Next = Body[Pos + 1];
2515 if (Next == '$' || Next == 'n' ||
2516 isdigit(static_cast<unsigned char>(Next)))
2517 break;
2518 } else {
2519 // This macro has parameters, look for \foo, \bar, etc.
2520 if (Body[Pos] == '\\' && Pos + 1 != End)
2521 break;
2522 }
2523 }
2524
2525 // Add the prefix.
2526 OS << Body.slice(0, Pos);
2527
2528 // Check if we reached the end.
2529 if (Pos == End)
2530 break;
2531
2532 if (IsDarwin && !NParameters) {
2533 switch (Body[Pos + 1]) {
2534 // $$ => $
2535 case '$':
2536 OS << '$';
2537 break;
2538
2539 // $n => number of arguments
2540 case 'n':
2541 OS << A.size();
2542 break;
2543
2544 // $[0-9] => argument
2545 default: {
2546 // Missing arguments are ignored.
2547 unsigned Index = Body[Pos + 1] - '0';
2548 if (Index >= A.size())
2549 break;
2550
2551 // Otherwise substitute with the token values, with spaces eliminated.
2552 for (const AsmToken &Token : A[Index])
2553 OS << Token.getString();
2554 break;
2555 }
2556 }
2557 Pos += 2;
2558 } else {
2559 unsigned I = Pos + 1;
2560
2561 // Check for the \@ pseudo-variable.
2562 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End)
2563 ++I;
2564 else
2565 while (isIdentifierChar(Body[I]) && I + 1 != End)
2566 ++I;
2567
2568 const char *Begin = Body.data() + Pos + 1;
2569 StringRef Argument(Begin, I - (Pos + 1));
2570 unsigned Index = 0;
2571
2572 if (Argument == "@") {
2573 OS << NumOfMacroInstantiations;
2574 Pos += 2;
2575 } else {
2576 for (; Index < NParameters; ++Index)
2577 if (Parameters[Index].Name == Argument)
2578 break;
2579
2580 if (Index == NParameters) {
2581 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
2582 Pos += 3;
2583 else {
2584 OS << '\\' << Argument;
2585 Pos = I;
2586 }
2587 } else {
2588 bool VarargParameter = HasVararg && Index == (NParameters - 1);
2589 for (const AsmToken &Token : A[Index])
2590 // For altmacro mode, you can write '%expr'.
2591 // The prefix '%' evaluates the expression 'expr'
2592 // and uses the result as a string (e.g. replace %(1+2) with the
2593 // string "3").
2594 // Here, we identify the integer token which is the result of the
2595 // absolute expression evaluation and replace it with its string
2596 // representation.
2597 if (AltMacroMode && Token.getString().front() == '%' &&
2598 Token.is(AsmToken::Integer))
2599 // Emit an integer value to the buffer.
2600 OS << Token.getIntVal();
2601 // Only Token that was validated as a string and begins with '<'
2602 // is considered altMacroString!!!
2603 else if (AltMacroMode && Token.getString().front() == '<' &&
2604 Token.is(AsmToken::String)) {
2605 OS << angleBracketString(Token.getStringContents());
2606 }
2607 // We expect no quotes around the string's contents when
2608 // parsing for varargs.
2609 else if (Token.isNot(AsmToken::String) || VarargParameter)
2610 OS << Token.getString();
2611 else
2612 OS << Token.getStringContents();
2613
2614 Pos += 1 + Argument.size();
2615 }
2616 }
2617 }
2618 // Update the scan point.
2619 Body = Body.substr(Pos);
2620 }
2621
2622 return false;
2623 }
2624
isOperator(AsmToken::TokenKind kind)2625 static bool isOperator(AsmToken::TokenKind kind) {
2626 switch (kind) {
2627 default:
2628 return false;
2629 case AsmToken::Plus:
2630 case AsmToken::Minus:
2631 case AsmToken::Tilde:
2632 case AsmToken::Slash:
2633 case AsmToken::Star:
2634 case AsmToken::Dot:
2635 case AsmToken::Equal:
2636 case AsmToken::EqualEqual:
2637 case AsmToken::Pipe:
2638 case AsmToken::PipePipe:
2639 case AsmToken::Caret:
2640 case AsmToken::Amp:
2641 case AsmToken::AmpAmp:
2642 case AsmToken::Exclaim:
2643 case AsmToken::ExclaimEqual:
2644 case AsmToken::Less:
2645 case AsmToken::LessEqual:
2646 case AsmToken::LessLess:
2647 case AsmToken::LessGreater:
2648 case AsmToken::Greater:
2649 case AsmToken::GreaterEqual:
2650 case AsmToken::GreaterGreater:
2651 return true;
2652 }
2653 }
2654
2655 namespace {
2656
2657 class AsmLexerSkipSpaceRAII {
2658 public:
AsmLexerSkipSpaceRAII(AsmLexer & Lexer,bool SkipSpace)2659 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) {
2660 Lexer.setSkipSpace(SkipSpace);
2661 }
2662
~AsmLexerSkipSpaceRAII()2663 ~AsmLexerSkipSpaceRAII() {
2664 Lexer.setSkipSpace(true);
2665 }
2666
2667 private:
2668 AsmLexer &Lexer;
2669 };
2670
2671 } // end anonymous namespace
2672
parseMacroArgument(MCAsmMacroArgument & MA,bool Vararg)2673 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) {
2674
2675 if (Vararg) {
2676 if (Lexer.isNot(AsmToken::EndOfStatement)) {
2677 StringRef Str = parseStringToEndOfStatement();
2678 MA.emplace_back(AsmToken::String, Str);
2679 }
2680 return false;
2681 }
2682
2683 unsigned ParenLevel = 0;
2684
2685 // Darwin doesn't use spaces to delmit arguments.
2686 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin);
2687
2688 bool SpaceEaten;
2689
2690 while (true) {
2691 SpaceEaten = false;
2692 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
2693 return TokError("unexpected token in macro instantiation");
2694
2695 if (ParenLevel == 0) {
2696
2697 if (Lexer.is(AsmToken::Comma))
2698 break;
2699
2700 if (Lexer.is(AsmToken::Space)) {
2701 SpaceEaten = true;
2702 Lexer.Lex(); // Eat spaces
2703 }
2704
2705 // Spaces can delimit parameters, but could also be part an expression.
2706 // If the token after a space is an operator, add the token and the next
2707 // one into this argument
2708 if (!IsDarwin) {
2709 if (isOperator(Lexer.getKind())) {
2710 MA.push_back(getTok());
2711 Lexer.Lex();
2712
2713 // Whitespace after an operator can be ignored.
2714 if (Lexer.is(AsmToken::Space))
2715 Lexer.Lex();
2716
2717 continue;
2718 }
2719 }
2720 if (SpaceEaten)
2721 break;
2722 }
2723
2724 // handleMacroEntry relies on not advancing the lexer here
2725 // to be able to fill in the remaining default parameter values
2726 if (Lexer.is(AsmToken::EndOfStatement))
2727 break;
2728
2729 // Adjust the current parentheses level.
2730 if (Lexer.is(AsmToken::LParen))
2731 ++ParenLevel;
2732 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
2733 --ParenLevel;
2734
2735 // Append the token to the current argument list.
2736 MA.push_back(getTok());
2737 Lexer.Lex();
2738 }
2739
2740 if (ParenLevel != 0)
2741 return TokError("unbalanced parentheses in macro argument");
2742 return false;
2743 }
2744
2745 // Parse the macro instantiation arguments.
parseMacroArguments(const MCAsmMacro * M,MCAsmMacroArguments & A)2746 bool AsmParser::parseMacroArguments(const MCAsmMacro *M,
2747 MCAsmMacroArguments &A) {
2748 const unsigned NParameters = M ? M->Parameters.size() : 0;
2749 bool NamedParametersFound = false;
2750 SmallVector<SMLoc, 4> FALocs;
2751
2752 A.resize(NParameters);
2753 FALocs.resize(NParameters);
2754
2755 // Parse two kinds of macro invocations:
2756 // - macros defined without any parameters accept an arbitrary number of them
2757 // - macros defined with parameters accept at most that many of them
2758 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false;
2759 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
2760 ++Parameter) {
2761 SMLoc IDLoc = Lexer.getLoc();
2762 MCAsmMacroParameter FA;
2763
2764 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) {
2765 if (parseIdentifier(FA.Name))
2766 return Error(IDLoc, "invalid argument identifier for formal argument");
2767
2768 if (Lexer.isNot(AsmToken::Equal))
2769 return TokError("expected '=' after formal parameter identifier");
2770
2771 Lex();
2772
2773 NamedParametersFound = true;
2774 }
2775 bool Vararg = HasVararg && Parameter == (NParameters - 1);
2776
2777 if (NamedParametersFound && FA.Name.empty())
2778 return Error(IDLoc, "cannot mix positional and keyword arguments");
2779
2780 SMLoc StrLoc = Lexer.getLoc();
2781 SMLoc EndLoc;
2782 if (AltMacroMode && Lexer.is(AsmToken::Percent)) {
2783 const MCExpr *AbsoluteExp;
2784 int64_t Value;
2785 /// Eat '%'
2786 Lex();
2787 if (parseExpression(AbsoluteExp, EndLoc))
2788 return false;
2789 if (!AbsoluteExp->evaluateAsAbsolute(Value,
2790 getStreamer().getAssemblerPtr()))
2791 return Error(StrLoc, "expected absolute expression");
2792 const char *StrChar = StrLoc.getPointer();
2793 const char *EndChar = EndLoc.getPointer();
2794 AsmToken newToken(AsmToken::Integer,
2795 StringRef(StrChar, EndChar - StrChar), Value);
2796 FA.Value.push_back(newToken);
2797 } else if (AltMacroMode && Lexer.is(AsmToken::Less) &&
2798 isAngleBracketString(StrLoc, EndLoc)) {
2799 const char *StrChar = StrLoc.getPointer();
2800 const char *EndChar = EndLoc.getPointer();
2801 jumpToLoc(EndLoc, CurBuffer);
2802 /// Eat from '<' to '>'
2803 Lex();
2804 AsmToken newToken(AsmToken::String,
2805 StringRef(StrChar, EndChar - StrChar));
2806 FA.Value.push_back(newToken);
2807 } else if(parseMacroArgument(FA.Value, Vararg))
2808 return true;
2809
2810 unsigned PI = Parameter;
2811 if (!FA.Name.empty()) {
2812 unsigned FAI = 0;
2813 for (FAI = 0; FAI < NParameters; ++FAI)
2814 if (M->Parameters[FAI].Name == FA.Name)
2815 break;
2816
2817 if (FAI >= NParameters) {
2818 assert(M && "expected macro to be defined");
2819 return Error(IDLoc, "parameter named '" + FA.Name +
2820 "' does not exist for macro '" + M->Name + "'");
2821 }
2822 PI = FAI;
2823 }
2824
2825 if (!FA.Value.empty()) {
2826 if (A.size() <= PI)
2827 A.resize(PI + 1);
2828 A[PI] = FA.Value;
2829
2830 if (FALocs.size() <= PI)
2831 FALocs.resize(PI + 1);
2832
2833 FALocs[PI] = Lexer.getLoc();
2834 }
2835
2836 // At the end of the statement, fill in remaining arguments that have
2837 // default values. If there aren't any, then the next argument is
2838 // required but missing
2839 if (Lexer.is(AsmToken::EndOfStatement)) {
2840 bool Failure = false;
2841 for (unsigned FAI = 0; FAI < NParameters; ++FAI) {
2842 if (A[FAI].empty()) {
2843 if (M->Parameters[FAI].Required) {
2844 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(),
2845 "missing value for required parameter "
2846 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'");
2847 Failure = true;
2848 }
2849
2850 if (!M->Parameters[FAI].Value.empty())
2851 A[FAI] = M->Parameters[FAI].Value;
2852 }
2853 }
2854 return Failure;
2855 }
2856
2857 if (Lexer.is(AsmToken::Comma))
2858 Lex();
2859 }
2860
2861 return TokError("too many positional arguments");
2862 }
2863
handleMacroEntry(const MCAsmMacro * M,SMLoc NameLoc)2864 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) {
2865 // Arbitrarily limit macro nesting depth (default matches 'as'). We can
2866 // eliminate this, although we should protect against infinite loops.
2867 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth;
2868 if (ActiveMacros.size() == MaxNestingDepth) {
2869 std::ostringstream MaxNestingDepthError;
2870 MaxNestingDepthError << "macros cannot be nested more than "
2871 << MaxNestingDepth << " levels deep."
2872 << " Use -asm-macro-max-nesting-depth to increase "
2873 "this limit.";
2874 return TokError(MaxNestingDepthError.str());
2875 }
2876
2877 MCAsmMacroArguments A;
2878 if (parseMacroArguments(M, A))
2879 return true;
2880
2881 // Macro instantiation is lexical, unfortunately. We construct a new buffer
2882 // to hold the macro body with substitutions.
2883 SmallString<256> Buf;
2884 StringRef Body = M->Body;
2885 raw_svector_ostream OS(Buf);
2886
2887 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc()))
2888 return true;
2889
2890 // We include the .endmacro in the buffer as our cue to exit the macro
2891 // instantiation.
2892 OS << ".endmacro\n";
2893
2894 std::unique_ptr<MemoryBuffer> Instantiation =
2895 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
2896
2897 // Create the macro instantiation object and add to the current macro
2898 // instantiation stack.
2899 MacroInstantiation *MI = new MacroInstantiation{
2900 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()};
2901 ActiveMacros.push_back(MI);
2902
2903 ++NumOfMacroInstantiations;
2904
2905 // Jump to the macro instantiation and prime the lexer.
2906 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
2907 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
2908 Lex();
2909
2910 return false;
2911 }
2912
handleMacroExit()2913 void AsmParser::handleMacroExit() {
2914 // Jump to the EndOfStatement we should return to, and consume it.
2915 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer);
2916 Lex();
2917
2918 // Pop the instantiation entry.
2919 delete ActiveMacros.back();
2920 ActiveMacros.pop_back();
2921 }
2922
parseAssignment(StringRef Name,bool allow_redef,bool NoDeadStrip)2923 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef,
2924 bool NoDeadStrip) {
2925 MCSymbol *Sym;
2926 const MCExpr *Value;
2927 if (MCParserUtils::parseAssignmentExpression(Name, allow_redef, *this, Sym,
2928 Value))
2929 return true;
2930
2931 if (!Sym) {
2932 // In the case where we parse an expression starting with a '.', we will
2933 // not generate an error, nor will we create a symbol. In this case we
2934 // should just return out.
2935 return false;
2936 }
2937
2938 if (discardLTOSymbol(Name))
2939 return false;
2940
2941 // Do the assignment.
2942 Out.emitAssignment(Sym, Value);
2943 if (NoDeadStrip)
2944 Out.emitSymbolAttribute(Sym, MCSA_NoDeadStrip);
2945
2946 return false;
2947 }
2948
2949 /// parseIdentifier:
2950 /// ::= identifier
2951 /// ::= string
parseIdentifier(StringRef & Res)2952 bool AsmParser::parseIdentifier(StringRef &Res) {
2953 // The assembler has relaxed rules for accepting identifiers, in particular we
2954 // allow things like '.globl $foo' and '.def @feat.00', which would normally be
2955 // separate tokens. At this level, we have already lexed so we cannot (currently)
2956 // handle this as a context dependent token, instead we detect adjacent tokens
2957 // and return the combined identifier.
2958 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) {
2959 SMLoc PrefixLoc = getLexer().getLoc();
2960
2961 // Consume the prefix character, and check for a following identifier.
2962
2963 AsmToken Buf[1];
2964 Lexer.peekTokens(Buf, false);
2965
2966 if (Buf[0].isNot(AsmToken::Identifier) && Buf[0].isNot(AsmToken::Integer))
2967 return true;
2968
2969 // We have a '$' or '@' followed by an identifier or integer token, make
2970 // sure they are adjacent.
2971 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer())
2972 return true;
2973
2974 // eat $ or @
2975 Lexer.Lex(); // Lexer's Lex guarantees consecutive token.
2976 // Construct the joined identifier and consume the token.
2977 Res = StringRef(PrefixLoc.getPointer(), getTok().getString().size() + 1);
2978 Lex(); // Parser Lex to maintain invariants.
2979 return false;
2980 }
2981
2982 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String))
2983 return true;
2984
2985 Res = getTok().getIdentifier();
2986
2987 Lex(); // Consume the identifier token.
2988
2989 return false;
2990 }
2991
2992 /// parseDirectiveSet:
2993 /// ::= .equ identifier ',' expression
2994 /// ::= .equiv identifier ',' expression
2995 /// ::= .set identifier ',' expression
parseDirectiveSet(StringRef IDVal,bool allow_redef)2996 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) {
2997 StringRef Name;
2998 if (check(parseIdentifier(Name), "expected identifier") || parseComma() ||
2999 parseAssignment(Name, allow_redef, true))
3000 return true;
3001 return false;
3002 }
3003
parseEscapedString(std::string & Data)3004 bool AsmParser::parseEscapedString(std::string &Data) {
3005 if (check(getTok().isNot(AsmToken::String), "expected string"))
3006 return true;
3007
3008 Data = "";
3009 StringRef Str = getTok().getStringContents();
3010 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
3011 if (Str[i] != '\\') {
3012 Data += Str[i];
3013 continue;
3014 }
3015
3016 // Recognize escaped characters. Note that this escape semantics currently
3017 // loosely follows Darwin 'as'.
3018 ++i;
3019 if (i == e)
3020 return TokError("unexpected backslash at end of string");
3021
3022 // Recognize hex sequences similarly to GNU 'as'.
3023 if (Str[i] == 'x' || Str[i] == 'X') {
3024 size_t length = Str.size();
3025 if (i + 1 >= length || !isHexDigit(Str[i + 1]))
3026 return TokError("invalid hexadecimal escape sequence");
3027
3028 // Consume hex characters. GNU 'as' reads all hexadecimal characters and
3029 // then truncates to the lower 16 bits. Seems reasonable.
3030 unsigned Value = 0;
3031 while (i + 1 < length && isHexDigit(Str[i + 1]))
3032 Value = Value * 16 + hexDigitValue(Str[++i]);
3033
3034 Data += (unsigned char)(Value & 0xFF);
3035 continue;
3036 }
3037
3038 // Recognize octal sequences.
3039 if ((unsigned)(Str[i] - '0') <= 7) {
3040 // Consume up to three octal characters.
3041 unsigned Value = Str[i] - '0';
3042
3043 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
3044 ++i;
3045 Value = Value * 8 + (Str[i] - '0');
3046
3047 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
3048 ++i;
3049 Value = Value * 8 + (Str[i] - '0');
3050 }
3051 }
3052
3053 if (Value > 255)
3054 return TokError("invalid octal escape sequence (out of range)");
3055
3056 Data += (unsigned char)Value;
3057 continue;
3058 }
3059
3060 // Otherwise recognize individual escapes.
3061 switch (Str[i]) {
3062 default:
3063 // Just reject invalid escape sequences for now.
3064 return TokError("invalid escape sequence (unrecognized character)");
3065
3066 case 'b': Data += '\b'; break;
3067 case 'f': Data += '\f'; break;
3068 case 'n': Data += '\n'; break;
3069 case 'r': Data += '\r'; break;
3070 case 't': Data += '\t'; break;
3071 case '"': Data += '"'; break;
3072 case '\\': Data += '\\'; break;
3073 }
3074 }
3075
3076 Lex();
3077 return false;
3078 }
3079
parseAngleBracketString(std::string & Data)3080 bool AsmParser::parseAngleBracketString(std::string &Data) {
3081 SMLoc EndLoc, StartLoc = getTok().getLoc();
3082 if (isAngleBracketString(StartLoc, EndLoc)) {
3083 const char *StartChar = StartLoc.getPointer() + 1;
3084 const char *EndChar = EndLoc.getPointer() - 1;
3085 jumpToLoc(EndLoc, CurBuffer);
3086 /// Eat from '<' to '>'
3087 Lex();
3088
3089 Data = angleBracketString(StringRef(StartChar, EndChar - StartChar));
3090 return false;
3091 }
3092 return true;
3093 }
3094
3095 /// parseDirectiveAscii:
3096 // ::= .ascii [ "string"+ ( , "string"+ )* ]
3097 /// ::= ( .asciz | .string ) [ "string" ( , "string" )* ]
parseDirectiveAscii(StringRef IDVal,bool ZeroTerminated)3098 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
3099 auto parseOp = [&]() -> bool {
3100 std::string Data;
3101 if (checkForValidSection())
3102 return true;
3103 // Only support spaces as separators for .ascii directive for now. See the
3104 // discusssion at https://reviews.llvm.org/D91460 for more details.
3105 do {
3106 if (parseEscapedString(Data))
3107 return true;
3108 getStreamer().emitBytes(Data);
3109 } while (!ZeroTerminated && getTok().is(AsmToken::String));
3110 if (ZeroTerminated)
3111 getStreamer().emitBytes(StringRef("\0", 1));
3112 return false;
3113 };
3114
3115 return parseMany(parseOp);
3116 }
3117
3118 /// parseDirectiveReloc
3119 /// ::= .reloc expression , identifier [ , expression ]
parseDirectiveReloc(SMLoc DirectiveLoc)3120 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) {
3121 const MCExpr *Offset;
3122 const MCExpr *Expr = nullptr;
3123 SMLoc OffsetLoc = Lexer.getTok().getLoc();
3124
3125 if (parseExpression(Offset))
3126 return true;
3127 if (parseComma() ||
3128 check(getTok().isNot(AsmToken::Identifier), "expected relocation name"))
3129 return true;
3130
3131 SMLoc NameLoc = Lexer.getTok().getLoc();
3132 StringRef Name = Lexer.getTok().getIdentifier();
3133 Lex();
3134
3135 if (Lexer.is(AsmToken::Comma)) {
3136 Lex();
3137 SMLoc ExprLoc = Lexer.getLoc();
3138 if (parseExpression(Expr))
3139 return true;
3140
3141 MCValue Value;
3142 if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr))
3143 return Error(ExprLoc, "expression must be relocatable");
3144 }
3145
3146 if (parseEOL())
3147 return true;
3148
3149 const MCTargetAsmParser &MCT = getTargetParser();
3150 const MCSubtargetInfo &STI = MCT.getSTI();
3151 if (Optional<std::pair<bool, std::string>> Err =
3152 getStreamer().emitRelocDirective(*Offset, Name, Expr, DirectiveLoc,
3153 STI))
3154 return Error(Err->first ? NameLoc : OffsetLoc, Err->second);
3155
3156 return false;
3157 }
3158
3159 /// parseDirectiveValue
3160 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
parseDirectiveValue(StringRef IDVal,unsigned Size)3161 bool AsmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) {
3162 auto parseOp = [&]() -> bool {
3163 const MCExpr *Value;
3164 SMLoc ExprLoc = getLexer().getLoc();
3165 if (checkForValidSection() || parseExpression(Value))
3166 return true;
3167 // Special case constant expressions to match code generator.
3168 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3169 assert(Size <= 8 && "Invalid size");
3170 uint64_t IntValue = MCE->getValue();
3171 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
3172 return Error(ExprLoc, "out of range literal value");
3173 getStreamer().emitIntValue(IntValue, Size);
3174 } else
3175 getStreamer().emitValue(Value, Size, ExprLoc);
3176 return false;
3177 };
3178
3179 return parseMany(parseOp);
3180 }
3181
parseHexOcta(AsmParser & Asm,uint64_t & hi,uint64_t & lo)3182 static bool parseHexOcta(AsmParser &Asm, uint64_t &hi, uint64_t &lo) {
3183 if (Asm.getTok().isNot(AsmToken::Integer) &&
3184 Asm.getTok().isNot(AsmToken::BigNum))
3185 return Asm.TokError("unknown token in expression");
3186 SMLoc ExprLoc = Asm.getTok().getLoc();
3187 APInt IntValue = Asm.getTok().getAPIntVal();
3188 Asm.Lex();
3189 if (!IntValue.isIntN(128))
3190 return Asm.Error(ExprLoc, "out of range literal value");
3191 if (!IntValue.isIntN(64)) {
3192 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue();
3193 lo = IntValue.getLoBits(64).getZExtValue();
3194 } else {
3195 hi = 0;
3196 lo = IntValue.getZExtValue();
3197 }
3198 return false;
3199 }
3200
3201 /// ParseDirectiveOctaValue
3202 /// ::= .octa [ hexconstant (, hexconstant)* ]
3203
parseDirectiveOctaValue(StringRef IDVal)3204 bool AsmParser::parseDirectiveOctaValue(StringRef IDVal) {
3205 auto parseOp = [&]() -> bool {
3206 if (checkForValidSection())
3207 return true;
3208 uint64_t hi, lo;
3209 if (parseHexOcta(*this, hi, lo))
3210 return true;
3211 if (MAI.isLittleEndian()) {
3212 getStreamer().emitInt64(lo);
3213 getStreamer().emitInt64(hi);
3214 } else {
3215 getStreamer().emitInt64(hi);
3216 getStreamer().emitInt64(lo);
3217 }
3218 return false;
3219 };
3220
3221 return parseMany(parseOp);
3222 }
3223
parseRealValue(const fltSemantics & Semantics,APInt & Res)3224 bool AsmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) {
3225 // We don't truly support arithmetic on floating point expressions, so we
3226 // have to manually parse unary prefixes.
3227 bool IsNeg = false;
3228 if (getLexer().is(AsmToken::Minus)) {
3229 Lexer.Lex();
3230 IsNeg = true;
3231 } else if (getLexer().is(AsmToken::Plus))
3232 Lexer.Lex();
3233
3234 if (Lexer.is(AsmToken::Error))
3235 return TokError(Lexer.getErr());
3236 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) &&
3237 Lexer.isNot(AsmToken::Identifier))
3238 return TokError("unexpected token in directive");
3239
3240 // Convert to an APFloat.
3241 APFloat Value(Semantics);
3242 StringRef IDVal = getTok().getString();
3243 if (getLexer().is(AsmToken::Identifier)) {
3244 if (!IDVal.compare_insensitive("infinity") ||
3245 !IDVal.compare_insensitive("inf"))
3246 Value = APFloat::getInf(Semantics);
3247 else if (!IDVal.compare_insensitive("nan"))
3248 Value = APFloat::getNaN(Semantics, false, ~0);
3249 else
3250 return TokError("invalid floating point literal");
3251 } else if (errorToBool(
3252 Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven)
3253 .takeError()))
3254 return TokError("invalid floating point literal");
3255 if (IsNeg)
3256 Value.changeSign();
3257
3258 // Consume the numeric token.
3259 Lex();
3260
3261 Res = Value.bitcastToAPInt();
3262
3263 return false;
3264 }
3265
3266 /// parseDirectiveRealValue
3267 /// ::= (.single | .double) [ expression (, expression)* ]
parseDirectiveRealValue(StringRef IDVal,const fltSemantics & Semantics)3268 bool AsmParser::parseDirectiveRealValue(StringRef IDVal,
3269 const fltSemantics &Semantics) {
3270 auto parseOp = [&]() -> bool {
3271 APInt AsInt;
3272 if (checkForValidSection() || parseRealValue(Semantics, AsInt))
3273 return true;
3274 getStreamer().emitIntValue(AsInt.getLimitedValue(),
3275 AsInt.getBitWidth() / 8);
3276 return false;
3277 };
3278
3279 return parseMany(parseOp);
3280 }
3281
3282 /// parseDirectiveZero
3283 /// ::= .zero expression
parseDirectiveZero()3284 bool AsmParser::parseDirectiveZero() {
3285 SMLoc NumBytesLoc = Lexer.getLoc();
3286 const MCExpr *NumBytes;
3287 if (checkForValidSection() || parseExpression(NumBytes))
3288 return true;
3289
3290 int64_t Val = 0;
3291 if (getLexer().is(AsmToken::Comma)) {
3292 Lex();
3293 if (parseAbsoluteExpression(Val))
3294 return true;
3295 }
3296
3297 if (parseEOL())
3298 return true;
3299 getStreamer().emitFill(*NumBytes, Val, NumBytesLoc);
3300
3301 return false;
3302 }
3303
3304 /// parseDirectiveFill
3305 /// ::= .fill expression [ , expression [ , expression ] ]
parseDirectiveFill()3306 bool AsmParser::parseDirectiveFill() {
3307 SMLoc NumValuesLoc = Lexer.getLoc();
3308 const MCExpr *NumValues;
3309 if (checkForValidSection() || parseExpression(NumValues))
3310 return true;
3311
3312 int64_t FillSize = 1;
3313 int64_t FillExpr = 0;
3314
3315 SMLoc SizeLoc, ExprLoc;
3316
3317 if (parseOptionalToken(AsmToken::Comma)) {
3318 SizeLoc = getTok().getLoc();
3319 if (parseAbsoluteExpression(FillSize))
3320 return true;
3321 if (parseOptionalToken(AsmToken::Comma)) {
3322 ExprLoc = getTok().getLoc();
3323 if (parseAbsoluteExpression(FillExpr))
3324 return true;
3325 }
3326 }
3327 if (parseEOL())
3328 return true;
3329
3330 if (FillSize < 0) {
3331 Warning(SizeLoc, "'.fill' directive with negative size has no effect");
3332 return false;
3333 }
3334 if (FillSize > 8) {
3335 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8");
3336 FillSize = 8;
3337 }
3338
3339 if (!isUInt<32>(FillExpr) && FillSize > 4)
3340 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits");
3341
3342 getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc);
3343
3344 return false;
3345 }
3346
3347 /// parseDirectiveOrg
3348 /// ::= .org expression [ , expression ]
parseDirectiveOrg()3349 bool AsmParser::parseDirectiveOrg() {
3350 const MCExpr *Offset;
3351 SMLoc OffsetLoc = Lexer.getLoc();
3352 if (checkForValidSection() || parseExpression(Offset))
3353 return true;
3354
3355 // Parse optional fill expression.
3356 int64_t FillExpr = 0;
3357 if (parseOptionalToken(AsmToken::Comma))
3358 if (parseAbsoluteExpression(FillExpr))
3359 return true;
3360 if (parseEOL())
3361 return true;
3362
3363 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc);
3364 return false;
3365 }
3366
3367 /// parseDirectiveAlign
3368 /// ::= {.align, ...} expression [ , expression [ , expression ]]
parseDirectiveAlign(bool IsPow2,unsigned ValueSize)3369 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
3370 SMLoc AlignmentLoc = getLexer().getLoc();
3371 int64_t Alignment;
3372 SMLoc MaxBytesLoc;
3373 bool HasFillExpr = false;
3374 int64_t FillExpr = 0;
3375 int64_t MaxBytesToFill = 0;
3376
3377 auto parseAlign = [&]() -> bool {
3378 if (parseAbsoluteExpression(Alignment))
3379 return true;
3380 if (parseOptionalToken(AsmToken::Comma)) {
3381 // The fill expression can be omitted while specifying a maximum number of
3382 // alignment bytes, e.g:
3383 // .align 3,,4
3384 if (getTok().isNot(AsmToken::Comma)) {
3385 HasFillExpr = true;
3386 if (parseAbsoluteExpression(FillExpr))
3387 return true;
3388 }
3389 if (parseOptionalToken(AsmToken::Comma))
3390 if (parseTokenLoc(MaxBytesLoc) ||
3391 parseAbsoluteExpression(MaxBytesToFill))
3392 return true;
3393 }
3394 return parseEOL();
3395 };
3396
3397 if (checkForValidSection())
3398 return true;
3399 // Ignore empty '.p2align' directives for GNU-as compatibility
3400 if (IsPow2 && (ValueSize == 1) && getTok().is(AsmToken::EndOfStatement)) {
3401 Warning(AlignmentLoc, "p2align directive with no operand(s) is ignored");
3402 return parseEOL();
3403 }
3404 if (parseAlign())
3405 return true;
3406
3407 // Always emit an alignment here even if we thrown an error.
3408 bool ReturnVal = false;
3409
3410 // Compute alignment in bytes.
3411 if (IsPow2) {
3412 // FIXME: Diagnose overflow.
3413 if (Alignment >= 32) {
3414 ReturnVal |= Error(AlignmentLoc, "invalid alignment value");
3415 Alignment = 31;
3416 }
3417
3418 Alignment = 1ULL << Alignment;
3419 } else {
3420 // Reject alignments that aren't either a power of two or zero,
3421 // for gas compatibility. Alignment of zero is silently rounded
3422 // up to one.
3423 if (Alignment == 0)
3424 Alignment = 1;
3425 if (!isPowerOf2_64(Alignment))
3426 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2");
3427 if (!isUInt<32>(Alignment))
3428 ReturnVal |= Error(AlignmentLoc, "alignment must be smaller than 2**32");
3429 }
3430
3431 // Diagnose non-sensical max bytes to align.
3432 if (MaxBytesLoc.isValid()) {
3433 if (MaxBytesToFill < 1) {
3434 ReturnVal |= Error(MaxBytesLoc,
3435 "alignment directive can never be satisfied in this "
3436 "many bytes, ignoring maximum bytes expression");
3437 MaxBytesToFill = 0;
3438 }
3439
3440 if (MaxBytesToFill >= Alignment) {
3441 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
3442 "has no effect");
3443 MaxBytesToFill = 0;
3444 }
3445 }
3446
3447 // Check whether we should use optimal code alignment for this .align
3448 // directive.
3449 const MCSection *Section = getStreamer().getCurrentSectionOnly();
3450 assert(Section && "must have section to emit alignment");
3451 bool UseCodeAlign = Section->UseCodeAlign();
3452 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
3453 ValueSize == 1 && UseCodeAlign) {
3454 getStreamer().emitCodeAlignment(Alignment, MaxBytesToFill);
3455 } else {
3456 // FIXME: Target specific behavior about how the "extra" bytes are filled.
3457 getStreamer().emitValueToAlignment(Alignment, FillExpr, ValueSize,
3458 MaxBytesToFill);
3459 }
3460
3461 return ReturnVal;
3462 }
3463
3464 /// parseDirectiveFile
3465 /// ::= .file filename
3466 /// ::= .file number [directory] filename [md5 checksum] [source source-text]
parseDirectiveFile(SMLoc DirectiveLoc)3467 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) {
3468 // FIXME: I'm not sure what this is.
3469 int64_t FileNumber = -1;
3470 if (getLexer().is(AsmToken::Integer)) {
3471 FileNumber = getTok().getIntVal();
3472 Lex();
3473
3474 if (FileNumber < 0)
3475 return TokError("negative file number");
3476 }
3477
3478 std::string Path;
3479
3480 // Usually the directory and filename together, otherwise just the directory.
3481 // Allow the strings to have escaped octal character sequence.
3482 if (parseEscapedString(Path))
3483 return true;
3484
3485 StringRef Directory;
3486 StringRef Filename;
3487 std::string FilenameData;
3488 if (getLexer().is(AsmToken::String)) {
3489 if (check(FileNumber == -1,
3490 "explicit path specified, but no file number") ||
3491 parseEscapedString(FilenameData))
3492 return true;
3493 Filename = FilenameData;
3494 Directory = Path;
3495 } else {
3496 Filename = Path;
3497 }
3498
3499 uint64_t MD5Hi, MD5Lo;
3500 bool HasMD5 = false;
3501
3502 Optional<StringRef> Source;
3503 bool HasSource = false;
3504 std::string SourceString;
3505
3506 while (!parseOptionalToken(AsmToken::EndOfStatement)) {
3507 StringRef Keyword;
3508 if (check(getTok().isNot(AsmToken::Identifier),
3509 "unexpected token in '.file' directive") ||
3510 parseIdentifier(Keyword))
3511 return true;
3512 if (Keyword == "md5") {
3513 HasMD5 = true;
3514 if (check(FileNumber == -1,
3515 "MD5 checksum specified, but no file number") ||
3516 parseHexOcta(*this, MD5Hi, MD5Lo))
3517 return true;
3518 } else if (Keyword == "source") {
3519 HasSource = true;
3520 if (check(FileNumber == -1,
3521 "source specified, but no file number") ||
3522 check(getTok().isNot(AsmToken::String),
3523 "unexpected token in '.file' directive") ||
3524 parseEscapedString(SourceString))
3525 return true;
3526 } else {
3527 return TokError("unexpected token in '.file' directive");
3528 }
3529 }
3530
3531 if (FileNumber == -1) {
3532 // Ignore the directive if there is no number and the target doesn't support
3533 // numberless .file directives. This allows some portability of assembler
3534 // between different object file formats.
3535 if (getContext().getAsmInfo()->hasSingleParameterDotFile())
3536 getStreamer().emitFileDirective(Filename);
3537 } else {
3538 // In case there is a -g option as well as debug info from directive .file,
3539 // we turn off the -g option, directly use the existing debug info instead.
3540 // Throw away any implicit file table for the assembler source.
3541 if (Ctx.getGenDwarfForAssembly()) {
3542 Ctx.getMCDwarfLineTable(0).resetFileTable();
3543 Ctx.setGenDwarfForAssembly(false);
3544 }
3545
3546 Optional<MD5::MD5Result> CKMem;
3547 if (HasMD5) {
3548 MD5::MD5Result Sum;
3549 for (unsigned i = 0; i != 8; ++i) {
3550 Sum.Bytes[i] = uint8_t(MD5Hi >> ((7 - i) * 8));
3551 Sum.Bytes[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8));
3552 }
3553 CKMem = Sum;
3554 }
3555 if (HasSource) {
3556 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size()));
3557 memcpy(SourceBuf, SourceString.data(), SourceString.size());
3558 Source = StringRef(SourceBuf, SourceString.size());
3559 }
3560 if (FileNumber == 0) {
3561 // Upgrade to Version 5 for assembly actions like clang -c a.s.
3562 if (Ctx.getDwarfVersion() < 5)
3563 Ctx.setDwarfVersion(5);
3564 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source);
3565 } else {
3566 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective(
3567 FileNumber, Directory, Filename, CKMem, Source);
3568 if (!FileNumOrErr)
3569 return Error(DirectiveLoc, toString(FileNumOrErr.takeError()));
3570 }
3571 // Alert the user if there are some .file directives with MD5 and some not.
3572 // But only do that once.
3573 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) {
3574 ReportedInconsistentMD5 = true;
3575 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums");
3576 }
3577 }
3578
3579 return false;
3580 }
3581
3582 /// parseDirectiveLine
3583 /// ::= .line [number]
parseDirectiveLine()3584 bool AsmParser::parseDirectiveLine() {
3585 int64_t LineNumber;
3586 if (getLexer().is(AsmToken::Integer)) {
3587 if (parseIntToken(LineNumber, "unexpected token in '.line' directive"))
3588 return true;
3589 (void)LineNumber;
3590 // FIXME: Do something with the .line.
3591 }
3592 return parseEOL();
3593 }
3594
3595 /// parseDirectiveLoc
3596 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
3597 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
3598 /// The first number is a file number, must have been previously assigned with
3599 /// a .file directive, the second number is the line number and optionally the
3600 /// third number is a column position (zero if not specified). The remaining
3601 /// optional items are .loc sub-directives.
parseDirectiveLoc()3602 bool AsmParser::parseDirectiveLoc() {
3603 int64_t FileNumber = 0, LineNumber = 0;
3604 SMLoc Loc = getTok().getLoc();
3605 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") ||
3606 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc,
3607 "file number less than one in '.loc' directive") ||
3608 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc,
3609 "unassigned file number in '.loc' directive"))
3610 return true;
3611
3612 // optional
3613 if (getLexer().is(AsmToken::Integer)) {
3614 LineNumber = getTok().getIntVal();
3615 if (LineNumber < 0)
3616 return TokError("line number less than zero in '.loc' directive");
3617 Lex();
3618 }
3619
3620 int64_t ColumnPos = 0;
3621 if (getLexer().is(AsmToken::Integer)) {
3622 ColumnPos = getTok().getIntVal();
3623 if (ColumnPos < 0)
3624 return TokError("column position less than zero in '.loc' directive");
3625 Lex();
3626 }
3627
3628 auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags();
3629 unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT;
3630 unsigned Isa = 0;
3631 int64_t Discriminator = 0;
3632
3633 auto parseLocOp = [&]() -> bool {
3634 StringRef Name;
3635 SMLoc Loc = getTok().getLoc();
3636 if (parseIdentifier(Name))
3637 return TokError("unexpected token in '.loc' directive");
3638
3639 if (Name == "basic_block")
3640 Flags |= DWARF2_FLAG_BASIC_BLOCK;
3641 else if (Name == "prologue_end")
3642 Flags |= DWARF2_FLAG_PROLOGUE_END;
3643 else if (Name == "epilogue_begin")
3644 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
3645 else if (Name == "is_stmt") {
3646 Loc = getTok().getLoc();
3647 const MCExpr *Value;
3648 if (parseExpression(Value))
3649 return true;
3650 // The expression must be the constant 0 or 1.
3651 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3652 int Value = MCE->getValue();
3653 if (Value == 0)
3654 Flags &= ~DWARF2_FLAG_IS_STMT;
3655 else if (Value == 1)
3656 Flags |= DWARF2_FLAG_IS_STMT;
3657 else
3658 return Error(Loc, "is_stmt value not 0 or 1");
3659 } else {
3660 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
3661 }
3662 } else if (Name == "isa") {
3663 Loc = getTok().getLoc();
3664 const MCExpr *Value;
3665 if (parseExpression(Value))
3666 return true;
3667 // The expression must be a constant greater or equal to 0.
3668 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3669 int Value = MCE->getValue();
3670 if (Value < 0)
3671 return Error(Loc, "isa number less than zero");
3672 Isa = Value;
3673 } else {
3674 return Error(Loc, "isa number not a constant value");
3675 }
3676 } else if (Name == "discriminator") {
3677 if (parseAbsoluteExpression(Discriminator))
3678 return true;
3679 } else {
3680 return Error(Loc, "unknown sub-directive in '.loc' directive");
3681 }
3682 return false;
3683 };
3684
3685 if (parseMany(parseLocOp, false /*hasComma*/))
3686 return true;
3687
3688 getStreamer().emitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
3689 Isa, Discriminator, StringRef());
3690
3691 return false;
3692 }
3693
3694 /// parseDirectiveStabs
3695 /// ::= .stabs string, number, number, number
parseDirectiveStabs()3696 bool AsmParser::parseDirectiveStabs() {
3697 return TokError("unsupported directive '.stabs'");
3698 }
3699
3700 /// parseDirectiveCVFile
3701 /// ::= .cv_file number filename [checksum] [checksumkind]
parseDirectiveCVFile()3702 bool AsmParser::parseDirectiveCVFile() {
3703 SMLoc FileNumberLoc = getTok().getLoc();
3704 int64_t FileNumber;
3705 std::string Filename;
3706 std::string Checksum;
3707 int64_t ChecksumKind = 0;
3708
3709 if (parseIntToken(FileNumber,
3710 "expected file number in '.cv_file' directive") ||
3711 check(FileNumber < 1, FileNumberLoc, "file number less than one") ||
3712 check(getTok().isNot(AsmToken::String),
3713 "unexpected token in '.cv_file' directive") ||
3714 parseEscapedString(Filename))
3715 return true;
3716 if (!parseOptionalToken(AsmToken::EndOfStatement)) {
3717 if (check(getTok().isNot(AsmToken::String),
3718 "unexpected token in '.cv_file' directive") ||
3719 parseEscapedString(Checksum) ||
3720 parseIntToken(ChecksumKind,
3721 "expected checksum kind in '.cv_file' directive") ||
3722 parseToken(AsmToken::EndOfStatement,
3723 "unexpected token in '.cv_file' directive"))
3724 return true;
3725 }
3726
3727 Checksum = fromHex(Checksum);
3728 void *CKMem = Ctx.allocate(Checksum.size(), 1);
3729 memcpy(CKMem, Checksum.data(), Checksum.size());
3730 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem),
3731 Checksum.size());
3732
3733 if (!getStreamer().EmitCVFileDirective(FileNumber, Filename, ChecksumAsBytes,
3734 static_cast<uint8_t>(ChecksumKind)))
3735 return Error(FileNumberLoc, "file number already allocated");
3736
3737 return false;
3738 }
3739
parseCVFunctionId(int64_t & FunctionId,StringRef DirectiveName)3740 bool AsmParser::parseCVFunctionId(int64_t &FunctionId,
3741 StringRef DirectiveName) {
3742 SMLoc Loc;
3743 return parseTokenLoc(Loc) ||
3744 parseIntToken(FunctionId, "expected function id in '" + DirectiveName +
3745 "' directive") ||
3746 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc,
3747 "expected function id within range [0, UINT_MAX)");
3748 }
3749
parseCVFileId(int64_t & FileNumber,StringRef DirectiveName)3750 bool AsmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) {
3751 SMLoc Loc;
3752 return parseTokenLoc(Loc) ||
3753 parseIntToken(FileNumber, "expected integer in '" + DirectiveName +
3754 "' directive") ||
3755 check(FileNumber < 1, Loc, "file number less than one in '" +
3756 DirectiveName + "' directive") ||
3757 check(!getCVContext().isValidFileNumber(FileNumber), Loc,
3758 "unassigned file number in '" + DirectiveName + "' directive");
3759 }
3760
3761 /// parseDirectiveCVFuncId
3762 /// ::= .cv_func_id FunctionId
3763 ///
3764 /// Introduces a function ID that can be used with .cv_loc.
parseDirectiveCVFuncId()3765 bool AsmParser::parseDirectiveCVFuncId() {
3766 SMLoc FunctionIdLoc = getTok().getLoc();
3767 int64_t FunctionId;
3768
3769 if (parseCVFunctionId(FunctionId, ".cv_func_id") ||
3770 parseToken(AsmToken::EndOfStatement,
3771 "unexpected token in '.cv_func_id' directive"))
3772 return true;
3773
3774 if (!getStreamer().EmitCVFuncIdDirective(FunctionId))
3775 return Error(FunctionIdLoc, "function id already allocated");
3776
3777 return false;
3778 }
3779
3780 /// parseDirectiveCVInlineSiteId
3781 /// ::= .cv_inline_site_id FunctionId
3782 /// "within" IAFunc
3783 /// "inlined_at" IAFile IALine [IACol]
3784 ///
3785 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined
3786 /// at" source location information for use in the line table of the caller,
3787 /// whether the caller is a real function or another inlined call site.
parseDirectiveCVInlineSiteId()3788 bool AsmParser::parseDirectiveCVInlineSiteId() {
3789 SMLoc FunctionIdLoc = getTok().getLoc();
3790 int64_t FunctionId;
3791 int64_t IAFunc;
3792 int64_t IAFile;
3793 int64_t IALine;
3794 int64_t IACol = 0;
3795
3796 // FunctionId
3797 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id"))
3798 return true;
3799
3800 // "within"
3801 if (check((getLexer().isNot(AsmToken::Identifier) ||
3802 getTok().getIdentifier() != "within"),
3803 "expected 'within' identifier in '.cv_inline_site_id' directive"))
3804 return true;
3805 Lex();
3806
3807 // IAFunc
3808 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id"))
3809 return true;
3810
3811 // "inlined_at"
3812 if (check((getLexer().isNot(AsmToken::Identifier) ||
3813 getTok().getIdentifier() != "inlined_at"),
3814 "expected 'inlined_at' identifier in '.cv_inline_site_id' "
3815 "directive") )
3816 return true;
3817 Lex();
3818
3819 // IAFile IALine
3820 if (parseCVFileId(IAFile, ".cv_inline_site_id") ||
3821 parseIntToken(IALine, "expected line number after 'inlined_at'"))
3822 return true;
3823
3824 // [IACol]
3825 if (getLexer().is(AsmToken::Integer)) {
3826 IACol = getTok().getIntVal();
3827 Lex();
3828 }
3829
3830 if (parseToken(AsmToken::EndOfStatement,
3831 "unexpected token in '.cv_inline_site_id' directive"))
3832 return true;
3833
3834 if (!getStreamer().EmitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile,
3835 IALine, IACol, FunctionIdLoc))
3836 return Error(FunctionIdLoc, "function id already allocated");
3837
3838 return false;
3839 }
3840
3841 /// parseDirectiveCVLoc
3842 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end]
3843 /// [is_stmt VALUE]
3844 /// The first number is a file number, must have been previously assigned with
3845 /// a .file directive, the second number is the line number and optionally the
3846 /// third number is a column position (zero if not specified). The remaining
3847 /// optional items are .loc sub-directives.
parseDirectiveCVLoc()3848 bool AsmParser::parseDirectiveCVLoc() {
3849 SMLoc DirectiveLoc = getTok().getLoc();
3850 int64_t FunctionId, FileNumber;
3851 if (parseCVFunctionId(FunctionId, ".cv_loc") ||
3852 parseCVFileId(FileNumber, ".cv_loc"))
3853 return true;
3854
3855 int64_t LineNumber = 0;
3856 if (getLexer().is(AsmToken::Integer)) {
3857 LineNumber = getTok().getIntVal();
3858 if (LineNumber < 0)
3859 return TokError("line number less than zero in '.cv_loc' directive");
3860 Lex();
3861 }
3862
3863 int64_t ColumnPos = 0;
3864 if (getLexer().is(AsmToken::Integer)) {
3865 ColumnPos = getTok().getIntVal();
3866 if (ColumnPos < 0)
3867 return TokError("column position less than zero in '.cv_loc' directive");
3868 Lex();
3869 }
3870
3871 bool PrologueEnd = false;
3872 uint64_t IsStmt = 0;
3873
3874 auto parseOp = [&]() -> bool {
3875 StringRef Name;
3876 SMLoc Loc = getTok().getLoc();
3877 if (parseIdentifier(Name))
3878 return TokError("unexpected token in '.cv_loc' directive");
3879 if (Name == "prologue_end")
3880 PrologueEnd = true;
3881 else if (Name == "is_stmt") {
3882 Loc = getTok().getLoc();
3883 const MCExpr *Value;
3884 if (parseExpression(Value))
3885 return true;
3886 // The expression must be the constant 0 or 1.
3887 IsStmt = ~0ULL;
3888 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value))
3889 IsStmt = MCE->getValue();
3890
3891 if (IsStmt > 1)
3892 return Error(Loc, "is_stmt value not 0 or 1");
3893 } else {
3894 return Error(Loc, "unknown sub-directive in '.cv_loc' directive");
3895 }
3896 return false;
3897 };
3898
3899 if (parseMany(parseOp, false /*hasComma*/))
3900 return true;
3901
3902 getStreamer().emitCVLocDirective(FunctionId, FileNumber, LineNumber,
3903 ColumnPos, PrologueEnd, IsStmt, StringRef(),
3904 DirectiveLoc);
3905 return false;
3906 }
3907
3908 /// parseDirectiveCVLinetable
3909 /// ::= .cv_linetable FunctionId, FnStart, FnEnd
parseDirectiveCVLinetable()3910 bool AsmParser::parseDirectiveCVLinetable() {
3911 int64_t FunctionId;
3912 StringRef FnStartName, FnEndName;
3913 SMLoc Loc = getTok().getLoc();
3914 if (parseCVFunctionId(FunctionId, ".cv_linetable") || parseComma() ||
3915 parseTokenLoc(Loc) ||
3916 check(parseIdentifier(FnStartName), Loc,
3917 "expected identifier in directive") ||
3918 parseComma() || parseTokenLoc(Loc) ||
3919 check(parseIdentifier(FnEndName), Loc,
3920 "expected identifier in directive"))
3921 return true;
3922
3923 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
3924 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
3925
3926 getStreamer().emitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym);
3927 return false;
3928 }
3929
3930 /// parseDirectiveCVInlineLinetable
3931 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd
parseDirectiveCVInlineLinetable()3932 bool AsmParser::parseDirectiveCVInlineLinetable() {
3933 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum;
3934 StringRef FnStartName, FnEndName;
3935 SMLoc Loc = getTok().getLoc();
3936 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") ||
3937 parseTokenLoc(Loc) ||
3938 parseIntToken(
3939 SourceFileId,
3940 "expected SourceField in '.cv_inline_linetable' directive") ||
3941 check(SourceFileId <= 0, Loc,
3942 "File id less than zero in '.cv_inline_linetable' directive") ||
3943 parseTokenLoc(Loc) ||
3944 parseIntToken(
3945 SourceLineNum,
3946 "expected SourceLineNum in '.cv_inline_linetable' directive") ||
3947 check(SourceLineNum < 0, Loc,
3948 "Line number less than zero in '.cv_inline_linetable' directive") ||
3949 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc,
3950 "expected identifier in directive") ||
3951 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc,
3952 "expected identifier in directive"))
3953 return true;
3954
3955 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement"))
3956 return true;
3957
3958 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
3959 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
3960 getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId,
3961 SourceLineNum, FnStartSym,
3962 FnEndSym);
3963 return false;
3964 }
3965
initializeCVDefRangeTypeMap()3966 void AsmParser::initializeCVDefRangeTypeMap() {
3967 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER;
3968 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL;
3969 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER;
3970 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL;
3971 }
3972
3973 /// parseDirectiveCVDefRange
3974 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes*
parseDirectiveCVDefRange()3975 bool AsmParser::parseDirectiveCVDefRange() {
3976 SMLoc Loc;
3977 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges;
3978 while (getLexer().is(AsmToken::Identifier)) {
3979 Loc = getLexer().getLoc();
3980 StringRef GapStartName;
3981 if (parseIdentifier(GapStartName))
3982 return Error(Loc, "expected identifier in directive");
3983 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName);
3984
3985 Loc = getLexer().getLoc();
3986 StringRef GapEndName;
3987 if (parseIdentifier(GapEndName))
3988 return Error(Loc, "expected identifier in directive");
3989 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName);
3990
3991 Ranges.push_back({GapStartSym, GapEndSym});
3992 }
3993
3994 StringRef CVDefRangeTypeStr;
3995 if (parseToken(
3996 AsmToken::Comma,
3997 "expected comma before def_range type in .cv_def_range directive") ||
3998 parseIdentifier(CVDefRangeTypeStr))
3999 return Error(Loc, "expected def_range type in directive");
4000
4001 StringMap<CVDefRangeType>::const_iterator CVTypeIt =
4002 CVDefRangeTypeMap.find(CVDefRangeTypeStr);
4003 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end())
4004 ? CVDR_DEFRANGE
4005 : CVTypeIt->getValue();
4006 switch (CVDRType) {
4007 case CVDR_DEFRANGE_REGISTER: {
4008 int64_t DRRegister;
4009 if (parseToken(AsmToken::Comma, "expected comma before register number in "
4010 ".cv_def_range directive") ||
4011 parseAbsoluteExpression(DRRegister))
4012 return Error(Loc, "expected register number");
4013
4014 codeview::DefRangeRegisterHeader DRHdr;
4015 DRHdr.Register = DRRegister;
4016 DRHdr.MayHaveNoName = 0;
4017 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4018 break;
4019 }
4020 case CVDR_DEFRANGE_FRAMEPOINTER_REL: {
4021 int64_t DROffset;
4022 if (parseToken(AsmToken::Comma,
4023 "expected comma before offset in .cv_def_range directive") ||
4024 parseAbsoluteExpression(DROffset))
4025 return Error(Loc, "expected offset value");
4026
4027 codeview::DefRangeFramePointerRelHeader DRHdr;
4028 DRHdr.Offset = DROffset;
4029 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4030 break;
4031 }
4032 case CVDR_DEFRANGE_SUBFIELD_REGISTER: {
4033 int64_t DRRegister;
4034 int64_t DROffsetInParent;
4035 if (parseToken(AsmToken::Comma, "expected comma before register number in "
4036 ".cv_def_range directive") ||
4037 parseAbsoluteExpression(DRRegister))
4038 return Error(Loc, "expected register number");
4039 if (parseToken(AsmToken::Comma,
4040 "expected comma before offset in .cv_def_range directive") ||
4041 parseAbsoluteExpression(DROffsetInParent))
4042 return Error(Loc, "expected offset value");
4043
4044 codeview::DefRangeSubfieldRegisterHeader DRHdr;
4045 DRHdr.Register = DRRegister;
4046 DRHdr.MayHaveNoName = 0;
4047 DRHdr.OffsetInParent = DROffsetInParent;
4048 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4049 break;
4050 }
4051 case CVDR_DEFRANGE_REGISTER_REL: {
4052 int64_t DRRegister;
4053 int64_t DRFlags;
4054 int64_t DRBasePointerOffset;
4055 if (parseToken(AsmToken::Comma, "expected comma before register number in "
4056 ".cv_def_range directive") ||
4057 parseAbsoluteExpression(DRRegister))
4058 return Error(Loc, "expected register value");
4059 if (parseToken(
4060 AsmToken::Comma,
4061 "expected comma before flag value in .cv_def_range directive") ||
4062 parseAbsoluteExpression(DRFlags))
4063 return Error(Loc, "expected flag value");
4064 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset "
4065 "in .cv_def_range directive") ||
4066 parseAbsoluteExpression(DRBasePointerOffset))
4067 return Error(Loc, "expected base pointer offset value");
4068
4069 codeview::DefRangeRegisterRelHeader DRHdr;
4070 DRHdr.Register = DRRegister;
4071 DRHdr.Flags = DRFlags;
4072 DRHdr.BasePointerOffset = DRBasePointerOffset;
4073 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
4074 break;
4075 }
4076 default:
4077 return Error(Loc, "unexpected def_range type in .cv_def_range directive");
4078 }
4079 return true;
4080 }
4081
4082 /// parseDirectiveCVString
4083 /// ::= .cv_stringtable "string"
parseDirectiveCVString()4084 bool AsmParser::parseDirectiveCVString() {
4085 std::string Data;
4086 if (checkForValidSection() || parseEscapedString(Data))
4087 return true;
4088
4089 // Put the string in the table and emit the offset.
4090 std::pair<StringRef, unsigned> Insertion =
4091 getCVContext().addToStringTable(Data);
4092 getStreamer().emitInt32(Insertion.second);
4093 return false;
4094 }
4095
4096 /// parseDirectiveCVStringTable
4097 /// ::= .cv_stringtable
parseDirectiveCVStringTable()4098 bool AsmParser::parseDirectiveCVStringTable() {
4099 getStreamer().emitCVStringTableDirective();
4100 return false;
4101 }
4102
4103 /// parseDirectiveCVFileChecksums
4104 /// ::= .cv_filechecksums
parseDirectiveCVFileChecksums()4105 bool AsmParser::parseDirectiveCVFileChecksums() {
4106 getStreamer().emitCVFileChecksumsDirective();
4107 return false;
4108 }
4109
4110 /// parseDirectiveCVFileChecksumOffset
4111 /// ::= .cv_filechecksumoffset fileno
parseDirectiveCVFileChecksumOffset()4112 bool AsmParser::parseDirectiveCVFileChecksumOffset() {
4113 int64_t FileNo;
4114 if (parseIntToken(FileNo, "expected identifier in directive"))
4115 return true;
4116 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement"))
4117 return true;
4118 getStreamer().emitCVFileChecksumOffsetDirective(FileNo);
4119 return false;
4120 }
4121
4122 /// parseDirectiveCVFPOData
4123 /// ::= .cv_fpo_data procsym
parseDirectiveCVFPOData()4124 bool AsmParser::parseDirectiveCVFPOData() {
4125 SMLoc DirLoc = getLexer().getLoc();
4126 StringRef ProcName;
4127 if (parseIdentifier(ProcName))
4128 return TokError("expected symbol name");
4129 if (parseEOL())
4130 return true;
4131 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName);
4132 getStreamer().EmitCVFPOData(ProcSym, DirLoc);
4133 return false;
4134 }
4135
4136 /// parseDirectiveCFISections
4137 /// ::= .cfi_sections section [, section]
parseDirectiveCFISections()4138 bool AsmParser::parseDirectiveCFISections() {
4139 StringRef Name;
4140 bool EH = false;
4141 bool Debug = false;
4142
4143 if (!parseOptionalToken(AsmToken::EndOfStatement)) {
4144 for (;;) {
4145 if (parseIdentifier(Name))
4146 return TokError("expected .eh_frame or .debug_frame");
4147 if (Name == ".eh_frame")
4148 EH = true;
4149 else if (Name == ".debug_frame")
4150 Debug = true;
4151 if (parseOptionalToken(AsmToken::EndOfStatement))
4152 break;
4153 if (parseComma())
4154 return true;
4155 }
4156 }
4157 getStreamer().emitCFISections(EH, Debug);
4158 return false;
4159 }
4160
4161 /// parseDirectiveCFIStartProc
4162 /// ::= .cfi_startproc [simple]
parseDirectiveCFIStartProc()4163 bool AsmParser::parseDirectiveCFIStartProc() {
4164 StringRef Simple;
4165 if (!parseOptionalToken(AsmToken::EndOfStatement)) {
4166 if (check(parseIdentifier(Simple) || Simple != "simple",
4167 "unexpected token") ||
4168 parseEOL())
4169 return true;
4170 }
4171
4172 // TODO(kristina): Deal with a corner case of incorrect diagnostic context
4173 // being produced if this directive is emitted as part of preprocessor macro
4174 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer.
4175 // Tools like llvm-mc on the other hand are not affected by it, and report
4176 // correct context information.
4177 getStreamer().emitCFIStartProc(!Simple.empty(), Lexer.getLoc());
4178 return false;
4179 }
4180
4181 /// parseDirectiveCFIEndProc
4182 /// ::= .cfi_endproc
parseDirectiveCFIEndProc()4183 bool AsmParser::parseDirectiveCFIEndProc() {
4184 if (parseEOL())
4185 return true;
4186 getStreamer().emitCFIEndProc();
4187 return false;
4188 }
4189
4190 /// parse register name or number.
parseRegisterOrRegisterNumber(int64_t & Register,SMLoc DirectiveLoc)4191 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register,
4192 SMLoc DirectiveLoc) {
4193 unsigned RegNo;
4194
4195 if (getLexer().isNot(AsmToken::Integer)) {
4196 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc))
4197 return true;
4198 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true);
4199 } else
4200 return parseAbsoluteExpression(Register);
4201
4202 return false;
4203 }
4204
4205 /// parseDirectiveCFIDefCfa
4206 /// ::= .cfi_def_cfa register, offset
parseDirectiveCFIDefCfa(SMLoc DirectiveLoc)4207 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) {
4208 int64_t Register = 0, Offset = 0;
4209 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4210 parseAbsoluteExpression(Offset) || parseEOL())
4211 return true;
4212
4213 getStreamer().emitCFIDefCfa(Register, Offset);
4214 return false;
4215 }
4216
4217 /// parseDirectiveCFIDefCfaOffset
4218 /// ::= .cfi_def_cfa_offset offset
parseDirectiveCFIDefCfaOffset()4219 bool AsmParser::parseDirectiveCFIDefCfaOffset() {
4220 int64_t Offset = 0;
4221 if (parseAbsoluteExpression(Offset) || parseEOL())
4222 return true;
4223
4224 getStreamer().emitCFIDefCfaOffset(Offset);
4225 return false;
4226 }
4227
4228 /// parseDirectiveCFIRegister
4229 /// ::= .cfi_register register, register
parseDirectiveCFIRegister(SMLoc DirectiveLoc)4230 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) {
4231 int64_t Register1 = 0, Register2 = 0;
4232 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || parseComma() ||
4233 parseRegisterOrRegisterNumber(Register2, DirectiveLoc) || parseEOL())
4234 return true;
4235
4236 getStreamer().emitCFIRegister(Register1, Register2);
4237 return false;
4238 }
4239
4240 /// parseDirectiveCFIWindowSave
4241 /// ::= .cfi_window_save
parseDirectiveCFIWindowSave()4242 bool AsmParser::parseDirectiveCFIWindowSave() {
4243 if (parseEOL())
4244 return true;
4245 getStreamer().emitCFIWindowSave();
4246 return false;
4247 }
4248
4249 /// parseDirectiveCFIAdjustCfaOffset
4250 /// ::= .cfi_adjust_cfa_offset adjustment
parseDirectiveCFIAdjustCfaOffset()4251 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() {
4252 int64_t Adjustment = 0;
4253 if (parseAbsoluteExpression(Adjustment) || parseEOL())
4254 return true;
4255
4256 getStreamer().emitCFIAdjustCfaOffset(Adjustment);
4257 return false;
4258 }
4259
4260 /// parseDirectiveCFIDefCfaRegister
4261 /// ::= .cfi_def_cfa_register register
parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc)4262 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) {
4263 int64_t Register = 0;
4264 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4265 return true;
4266
4267 getStreamer().emitCFIDefCfaRegister(Register);
4268 return false;
4269 }
4270
4271 /// parseDirectiveCFILLVMDefAspaceCfa
4272 /// ::= .cfi_llvm_def_aspace_cfa register, offset, address_space
parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc)4273 bool AsmParser::parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc) {
4274 int64_t Register = 0, Offset = 0, AddressSpace = 0;
4275 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4276 parseAbsoluteExpression(Offset) || parseComma() ||
4277 parseAbsoluteExpression(AddressSpace) || parseEOL())
4278 return true;
4279
4280 getStreamer().emitCFILLVMDefAspaceCfa(Register, Offset, AddressSpace);
4281 return false;
4282 }
4283
4284 /// parseDirectiveCFIOffset
4285 /// ::= .cfi_offset register, offset
parseDirectiveCFIOffset(SMLoc DirectiveLoc)4286 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) {
4287 int64_t Register = 0;
4288 int64_t Offset = 0;
4289
4290 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4291 parseAbsoluteExpression(Offset) || parseEOL())
4292 return true;
4293
4294 getStreamer().emitCFIOffset(Register, Offset);
4295 return false;
4296 }
4297
4298 /// parseDirectiveCFIRelOffset
4299 /// ::= .cfi_rel_offset register, offset
parseDirectiveCFIRelOffset(SMLoc DirectiveLoc)4300 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) {
4301 int64_t Register = 0, Offset = 0;
4302
4303 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() ||
4304 parseAbsoluteExpression(Offset) || parseEOL())
4305 return true;
4306
4307 getStreamer().emitCFIRelOffset(Register, Offset);
4308 return false;
4309 }
4310
isValidEncoding(int64_t Encoding)4311 static bool isValidEncoding(int64_t Encoding) {
4312 if (Encoding & ~0xff)
4313 return false;
4314
4315 if (Encoding == dwarf::DW_EH_PE_omit)
4316 return true;
4317
4318 const unsigned Format = Encoding & 0xf;
4319 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
4320 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
4321 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
4322 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
4323 return false;
4324
4325 const unsigned Application = Encoding & 0x70;
4326 if (Application != dwarf::DW_EH_PE_absptr &&
4327 Application != dwarf::DW_EH_PE_pcrel)
4328 return false;
4329
4330 return true;
4331 }
4332
4333 /// parseDirectiveCFIPersonalityOrLsda
4334 /// IsPersonality true for cfi_personality, false for cfi_lsda
4335 /// ::= .cfi_personality encoding, [symbol_name]
4336 /// ::= .cfi_lsda encoding, [symbol_name]
parseDirectiveCFIPersonalityOrLsda(bool IsPersonality)4337 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) {
4338 int64_t Encoding = 0;
4339 if (parseAbsoluteExpression(Encoding))
4340 return true;
4341 if (Encoding == dwarf::DW_EH_PE_omit)
4342 return false;
4343
4344 StringRef Name;
4345 if (check(!isValidEncoding(Encoding), "unsupported encoding.") ||
4346 parseComma() ||
4347 check(parseIdentifier(Name), "expected identifier in directive") ||
4348 parseEOL())
4349 return true;
4350
4351 MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
4352
4353 if (IsPersonality)
4354 getStreamer().emitCFIPersonality(Sym, Encoding);
4355 else
4356 getStreamer().emitCFILsda(Sym, Encoding);
4357 return false;
4358 }
4359
4360 /// parseDirectiveCFIRememberState
4361 /// ::= .cfi_remember_state
parseDirectiveCFIRememberState()4362 bool AsmParser::parseDirectiveCFIRememberState() {
4363 if (parseEOL())
4364 return true;
4365 getStreamer().emitCFIRememberState();
4366 return false;
4367 }
4368
4369 /// parseDirectiveCFIRestoreState
4370 /// ::= .cfi_remember_state
parseDirectiveCFIRestoreState()4371 bool AsmParser::parseDirectiveCFIRestoreState() {
4372 if (parseEOL())
4373 return true;
4374 getStreamer().emitCFIRestoreState();
4375 return false;
4376 }
4377
4378 /// parseDirectiveCFISameValue
4379 /// ::= .cfi_same_value register
parseDirectiveCFISameValue(SMLoc DirectiveLoc)4380 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) {
4381 int64_t Register = 0;
4382
4383 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4384 return true;
4385
4386 getStreamer().emitCFISameValue(Register);
4387 return false;
4388 }
4389
4390 /// parseDirectiveCFIRestore
4391 /// ::= .cfi_restore register
parseDirectiveCFIRestore(SMLoc DirectiveLoc)4392 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) {
4393 int64_t Register = 0;
4394 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4395 return true;
4396
4397 getStreamer().emitCFIRestore(Register);
4398 return false;
4399 }
4400
4401 /// parseDirectiveCFIEscape
4402 /// ::= .cfi_escape expression[,...]
parseDirectiveCFIEscape()4403 bool AsmParser::parseDirectiveCFIEscape() {
4404 std::string Values;
4405 int64_t CurrValue;
4406 if (parseAbsoluteExpression(CurrValue))
4407 return true;
4408
4409 Values.push_back((uint8_t)CurrValue);
4410
4411 while (getLexer().is(AsmToken::Comma)) {
4412 Lex();
4413
4414 if (parseAbsoluteExpression(CurrValue))
4415 return true;
4416
4417 Values.push_back((uint8_t)CurrValue);
4418 }
4419
4420 getStreamer().emitCFIEscape(Values);
4421 return false;
4422 }
4423
4424 /// parseDirectiveCFIReturnColumn
4425 /// ::= .cfi_return_column register
parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc)4426 bool AsmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) {
4427 int64_t Register = 0;
4428 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4429 return true;
4430 getStreamer().emitCFIReturnColumn(Register);
4431 return false;
4432 }
4433
4434 /// parseDirectiveCFISignalFrame
4435 /// ::= .cfi_signal_frame
parseDirectiveCFISignalFrame()4436 bool AsmParser::parseDirectiveCFISignalFrame() {
4437 if (parseEOL())
4438 return true;
4439
4440 getStreamer().emitCFISignalFrame();
4441 return false;
4442 }
4443
4444 /// parseDirectiveCFIUndefined
4445 /// ::= .cfi_undefined register
parseDirectiveCFIUndefined(SMLoc DirectiveLoc)4446 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) {
4447 int64_t Register = 0;
4448
4449 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL())
4450 return true;
4451
4452 getStreamer().emitCFIUndefined(Register);
4453 return false;
4454 }
4455
4456 /// parseDirectiveAltmacro
4457 /// ::= .altmacro
4458 /// ::= .noaltmacro
parseDirectiveAltmacro(StringRef Directive)4459 bool AsmParser::parseDirectiveAltmacro(StringRef Directive) {
4460 if (parseEOL())
4461 return true;
4462 AltMacroMode = (Directive == ".altmacro");
4463 return false;
4464 }
4465
4466 /// parseDirectiveMacrosOnOff
4467 /// ::= .macros_on
4468 /// ::= .macros_off
parseDirectiveMacrosOnOff(StringRef Directive)4469 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) {
4470 if (parseEOL())
4471 return true;
4472 setMacrosEnabled(Directive == ".macros_on");
4473 return false;
4474 }
4475
4476 /// parseDirectiveMacro
4477 /// ::= .macro name[,] [parameters]
parseDirectiveMacro(SMLoc DirectiveLoc)4478 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) {
4479 StringRef Name;
4480 if (parseIdentifier(Name))
4481 return TokError("expected identifier in '.macro' directive");
4482
4483 if (getLexer().is(AsmToken::Comma))
4484 Lex();
4485
4486 MCAsmMacroParameters Parameters;
4487 while (getLexer().isNot(AsmToken::EndOfStatement)) {
4488
4489 if (!Parameters.empty() && Parameters.back().Vararg)
4490 return Error(Lexer.getLoc(), "vararg parameter '" +
4491 Parameters.back().Name +
4492 "' should be the last parameter");
4493
4494 MCAsmMacroParameter Parameter;
4495 if (parseIdentifier(Parameter.Name))
4496 return TokError("expected identifier in '.macro' directive");
4497
4498 // Emit an error if two (or more) named parameters share the same name
4499 for (const MCAsmMacroParameter& CurrParam : Parameters)
4500 if (CurrParam.Name.equals(Parameter.Name))
4501 return TokError("macro '" + Name + "' has multiple parameters"
4502 " named '" + Parameter.Name + "'");
4503
4504 if (Lexer.is(AsmToken::Colon)) {
4505 Lex(); // consume ':'
4506
4507 SMLoc QualLoc;
4508 StringRef Qualifier;
4509
4510 QualLoc = Lexer.getLoc();
4511 if (parseIdentifier(Qualifier))
4512 return Error(QualLoc, "missing parameter qualifier for "
4513 "'" + Parameter.Name + "' in macro '" + Name + "'");
4514
4515 if (Qualifier == "req")
4516 Parameter.Required = true;
4517 else if (Qualifier == "vararg")
4518 Parameter.Vararg = true;
4519 else
4520 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier "
4521 "for '" + Parameter.Name + "' in macro '" + Name + "'");
4522 }
4523
4524 if (getLexer().is(AsmToken::Equal)) {
4525 Lex();
4526
4527 SMLoc ParamLoc;
4528
4529 ParamLoc = Lexer.getLoc();
4530 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false ))
4531 return true;
4532
4533 if (Parameter.Required)
4534 Warning(ParamLoc, "pointless default value for required parameter "
4535 "'" + Parameter.Name + "' in macro '" + Name + "'");
4536 }
4537
4538 Parameters.push_back(std::move(Parameter));
4539
4540 if (getLexer().is(AsmToken::Comma))
4541 Lex();
4542 }
4543
4544 // Eat just the end of statement.
4545 Lexer.Lex();
4546
4547 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors
4548 AsmToken EndToken, StartToken = getTok();
4549 unsigned MacroDepth = 0;
4550 // Lex the macro definition.
4551 while (true) {
4552 // Ignore Lexing errors in macros.
4553 while (Lexer.is(AsmToken::Error)) {
4554 Lexer.Lex();
4555 }
4556
4557 // Check whether we have reached the end of the file.
4558 if (getLexer().is(AsmToken::Eof))
4559 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
4560
4561 // Otherwise, check whether we have reach the .endmacro or the start of a
4562 // preprocessor line marker.
4563 if (getLexer().is(AsmToken::Identifier)) {
4564 if (getTok().getIdentifier() == ".endm" ||
4565 getTok().getIdentifier() == ".endmacro") {
4566 if (MacroDepth == 0) { // Outermost macro.
4567 EndToken = getTok();
4568 Lexer.Lex();
4569 if (getLexer().isNot(AsmToken::EndOfStatement))
4570 return TokError("unexpected token in '" + EndToken.getIdentifier() +
4571 "' directive");
4572 break;
4573 } else {
4574 // Otherwise we just found the end of an inner macro.
4575 --MacroDepth;
4576 }
4577 } else if (getTok().getIdentifier() == ".macro") {
4578 // We allow nested macros. Those aren't instantiated until the outermost
4579 // macro is expanded so just ignore them for now.
4580 ++MacroDepth;
4581 }
4582 } else if (Lexer.is(AsmToken::HashDirective)) {
4583 (void)parseCppHashLineFilenameComment(getLexer().getLoc());
4584 }
4585
4586 // Otherwise, scan til the end of the statement.
4587 eatToEndOfStatement();
4588 }
4589
4590 if (getContext().lookupMacro(Name)) {
4591 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
4592 }
4593
4594 const char *BodyStart = StartToken.getLoc().getPointer();
4595 const char *BodyEnd = EndToken.getLoc().getPointer();
4596 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
4597 checkForBadMacro(DirectiveLoc, Name, Body, Parameters);
4598 MCAsmMacro Macro(Name, Body, std::move(Parameters));
4599 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n";
4600 Macro.dump());
4601 getContext().defineMacro(Name, std::move(Macro));
4602 return false;
4603 }
4604
4605 /// checkForBadMacro
4606 ///
4607 /// With the support added for named parameters there may be code out there that
4608 /// is transitioning from positional parameters. In versions of gas that did
4609 /// not support named parameters they would be ignored on the macro definition.
4610 /// But to support both styles of parameters this is not possible so if a macro
4611 /// definition has named parameters but does not use them and has what appears
4612 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a
4613 /// warning that the positional parameter found in body which have no effect.
4614 /// Hoping the developer will either remove the named parameters from the macro
4615 /// definition so the positional parameters get used if that was what was
4616 /// intended or change the macro to use the named parameters. It is possible
4617 /// this warning will trigger when the none of the named parameters are used
4618 /// and the strings like $1 are infact to simply to be passed trough unchanged.
checkForBadMacro(SMLoc DirectiveLoc,StringRef Name,StringRef Body,ArrayRef<MCAsmMacroParameter> Parameters)4619 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name,
4620 StringRef Body,
4621 ArrayRef<MCAsmMacroParameter> Parameters) {
4622 // If this macro is not defined with named parameters the warning we are
4623 // checking for here doesn't apply.
4624 unsigned NParameters = Parameters.size();
4625 if (NParameters == 0)
4626 return;
4627
4628 bool NamedParametersFound = false;
4629 bool PositionalParametersFound = false;
4630
4631 // Look at the body of the macro for use of both the named parameters and what
4632 // are likely to be positional parameters. This is what expandMacro() is
4633 // doing when it finds the parameters in the body.
4634 while (!Body.empty()) {
4635 // Scan for the next possible parameter.
4636 std::size_t End = Body.size(), Pos = 0;
4637 for (; Pos != End; ++Pos) {
4638 // Check for a substitution or escape.
4639 // This macro is defined with parameters, look for \foo, \bar, etc.
4640 if (Body[Pos] == '\\' && Pos + 1 != End)
4641 break;
4642
4643 // This macro should have parameters, but look for $0, $1, ..., $n too.
4644 if (Body[Pos] != '$' || Pos + 1 == End)
4645 continue;
4646 char Next = Body[Pos + 1];
4647 if (Next == '$' || Next == 'n' ||
4648 isdigit(static_cast<unsigned char>(Next)))
4649 break;
4650 }
4651
4652 // Check if we reached the end.
4653 if (Pos == End)
4654 break;
4655
4656 if (Body[Pos] == '$') {
4657 switch (Body[Pos + 1]) {
4658 // $$ => $
4659 case '$':
4660 break;
4661
4662 // $n => number of arguments
4663 case 'n':
4664 PositionalParametersFound = true;
4665 break;
4666
4667 // $[0-9] => argument
4668 default: {
4669 PositionalParametersFound = true;
4670 break;
4671 }
4672 }
4673 Pos += 2;
4674 } else {
4675 unsigned I = Pos + 1;
4676 while (isIdentifierChar(Body[I]) && I + 1 != End)
4677 ++I;
4678
4679 const char *Begin = Body.data() + Pos + 1;
4680 StringRef Argument(Begin, I - (Pos + 1));
4681 unsigned Index = 0;
4682 for (; Index < NParameters; ++Index)
4683 if (Parameters[Index].Name == Argument)
4684 break;
4685
4686 if (Index == NParameters) {
4687 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
4688 Pos += 3;
4689 else {
4690 Pos = I;
4691 }
4692 } else {
4693 NamedParametersFound = true;
4694 Pos += 1 + Argument.size();
4695 }
4696 }
4697 // Update the scan point.
4698 Body = Body.substr(Pos);
4699 }
4700
4701 if (!NamedParametersFound && PositionalParametersFound)
4702 Warning(DirectiveLoc, "macro defined with named parameters which are not "
4703 "used in macro body, possible positional parameter "
4704 "found in body which will have no effect");
4705 }
4706
4707 /// parseDirectiveExitMacro
4708 /// ::= .exitm
parseDirectiveExitMacro(StringRef Directive)4709 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) {
4710 if (parseEOL())
4711 return true;
4712
4713 if (!isInsideMacroInstantiation())
4714 return TokError("unexpected '" + Directive + "' in file, "
4715 "no current macro definition");
4716
4717 // Exit all conditionals that are active in the current macro.
4718 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) {
4719 TheCondState = TheCondStack.back();
4720 TheCondStack.pop_back();
4721 }
4722
4723 handleMacroExit();
4724 return false;
4725 }
4726
4727 /// parseDirectiveEndMacro
4728 /// ::= .endm
4729 /// ::= .endmacro
parseDirectiveEndMacro(StringRef Directive)4730 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) {
4731 if (getLexer().isNot(AsmToken::EndOfStatement))
4732 return TokError("unexpected token in '" + Directive + "' directive");
4733
4734 // If we are inside a macro instantiation, terminate the current
4735 // instantiation.
4736 if (isInsideMacroInstantiation()) {
4737 handleMacroExit();
4738 return false;
4739 }
4740
4741 // Otherwise, this .endmacro is a stray entry in the file; well formed
4742 // .endmacro directives are handled during the macro definition parsing.
4743 return TokError("unexpected '" + Directive + "' in file, "
4744 "no current macro definition");
4745 }
4746
4747 /// parseDirectivePurgeMacro
4748 /// ::= .purgem name
parseDirectivePurgeMacro(SMLoc DirectiveLoc)4749 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) {
4750 StringRef Name;
4751 SMLoc Loc;
4752 if (parseTokenLoc(Loc) ||
4753 check(parseIdentifier(Name), Loc,
4754 "expected identifier in '.purgem' directive") ||
4755 parseEOL())
4756 return true;
4757
4758 if (!getContext().lookupMacro(Name))
4759 return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
4760
4761 getContext().undefineMacro(Name);
4762 DEBUG_WITH_TYPE("asm-macros", dbgs()
4763 << "Un-defining macro: " << Name << "\n");
4764 return false;
4765 }
4766
4767 /// parseDirectiveBundleAlignMode
4768 /// ::= {.bundle_align_mode} expression
parseDirectiveBundleAlignMode()4769 bool AsmParser::parseDirectiveBundleAlignMode() {
4770 // Expect a single argument: an expression that evaluates to a constant
4771 // in the inclusive range 0-30.
4772 SMLoc ExprLoc = getLexer().getLoc();
4773 int64_t AlignSizePow2;
4774 if (checkForValidSection() || parseAbsoluteExpression(AlignSizePow2) ||
4775 parseEOL() ||
4776 check(AlignSizePow2 < 0 || AlignSizePow2 > 30, ExprLoc,
4777 "invalid bundle alignment size (expected between 0 and 30)"))
4778 return true;
4779
4780 // Because of AlignSizePow2's verified range we can safely truncate it to
4781 // unsigned.
4782 getStreamer().emitBundleAlignMode(static_cast<unsigned>(AlignSizePow2));
4783 return false;
4784 }
4785
4786 /// parseDirectiveBundleLock
4787 /// ::= {.bundle_lock} [align_to_end]
parseDirectiveBundleLock()4788 bool AsmParser::parseDirectiveBundleLock() {
4789 if (checkForValidSection())
4790 return true;
4791 bool AlignToEnd = false;
4792
4793 StringRef Option;
4794 SMLoc Loc = getTok().getLoc();
4795 const char *kInvalidOptionError =
4796 "invalid option for '.bundle_lock' directive";
4797
4798 if (!parseOptionalToken(AsmToken::EndOfStatement)) {
4799 if (check(parseIdentifier(Option), Loc, kInvalidOptionError) ||
4800 check(Option != "align_to_end", Loc, kInvalidOptionError) || parseEOL())
4801 return true;
4802 AlignToEnd = true;
4803 }
4804
4805 getStreamer().emitBundleLock(AlignToEnd);
4806 return false;
4807 }
4808
4809 /// parseDirectiveBundleLock
4810 /// ::= {.bundle_lock}
parseDirectiveBundleUnlock()4811 bool AsmParser::parseDirectiveBundleUnlock() {
4812 if (checkForValidSection() || parseEOL())
4813 return true;
4814
4815 getStreamer().emitBundleUnlock();
4816 return false;
4817 }
4818
4819 /// parseDirectiveSpace
4820 /// ::= (.skip | .space) expression [ , expression ]
parseDirectiveSpace(StringRef IDVal)4821 bool AsmParser::parseDirectiveSpace(StringRef IDVal) {
4822 SMLoc NumBytesLoc = Lexer.getLoc();
4823 const MCExpr *NumBytes;
4824 if (checkForValidSection() || parseExpression(NumBytes))
4825 return true;
4826
4827 int64_t FillExpr = 0;
4828 if (parseOptionalToken(AsmToken::Comma))
4829 if (parseAbsoluteExpression(FillExpr))
4830 return true;
4831 if (parseEOL())
4832 return true;
4833
4834 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
4835 getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc);
4836
4837 return false;
4838 }
4839
4840 /// parseDirectiveDCB
4841 /// ::= .dcb.{b, l, w} expression, expression
parseDirectiveDCB(StringRef IDVal,unsigned Size)4842 bool AsmParser::parseDirectiveDCB(StringRef IDVal, unsigned Size) {
4843 SMLoc NumValuesLoc = Lexer.getLoc();
4844 int64_t NumValues;
4845 if (checkForValidSection() || parseAbsoluteExpression(NumValues))
4846 return true;
4847
4848 if (NumValues < 0) {
4849 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect");
4850 return false;
4851 }
4852
4853 if (parseComma())
4854 return true;
4855
4856 const MCExpr *Value;
4857 SMLoc ExprLoc = getLexer().getLoc();
4858 if (parseExpression(Value))
4859 return true;
4860
4861 // Special case constant expressions to match code generator.
4862 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
4863 assert(Size <= 8 && "Invalid size");
4864 uint64_t IntValue = MCE->getValue();
4865 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
4866 return Error(ExprLoc, "literal value out of range for directive");
4867 for (uint64_t i = 0, e = NumValues; i != e; ++i)
4868 getStreamer().emitIntValue(IntValue, Size);
4869 } else {
4870 for (uint64_t i = 0, e = NumValues; i != e; ++i)
4871 getStreamer().emitValue(Value, Size, ExprLoc);
4872 }
4873
4874 return parseEOL();
4875 }
4876
4877 /// parseDirectiveRealDCB
4878 /// ::= .dcb.{d, s} expression, expression
parseDirectiveRealDCB(StringRef IDVal,const fltSemantics & Semantics)4879 bool AsmParser::parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &Semantics) {
4880 SMLoc NumValuesLoc = Lexer.getLoc();
4881 int64_t NumValues;
4882 if (checkForValidSection() || parseAbsoluteExpression(NumValues))
4883 return true;
4884
4885 if (NumValues < 0) {
4886 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect");
4887 return false;
4888 }
4889
4890 if (parseComma())
4891 return true;
4892
4893 APInt AsInt;
4894 if (parseRealValue(Semantics, AsInt) || parseEOL())
4895 return true;
4896
4897 for (uint64_t i = 0, e = NumValues; i != e; ++i)
4898 getStreamer().emitIntValue(AsInt.getLimitedValue(),
4899 AsInt.getBitWidth() / 8);
4900
4901 return false;
4902 }
4903
4904 /// parseDirectiveDS
4905 /// ::= .ds.{b, d, l, p, s, w, x} expression
parseDirectiveDS(StringRef IDVal,unsigned Size)4906 bool AsmParser::parseDirectiveDS(StringRef IDVal, unsigned Size) {
4907 SMLoc NumValuesLoc = Lexer.getLoc();
4908 int64_t NumValues;
4909 if (checkForValidSection() || parseAbsoluteExpression(NumValues) ||
4910 parseEOL())
4911 return true;
4912
4913 if (NumValues < 0) {
4914 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect");
4915 return false;
4916 }
4917
4918 for (uint64_t i = 0, e = NumValues; i != e; ++i)
4919 getStreamer().emitFill(Size, 0);
4920
4921 return false;
4922 }
4923
4924 /// parseDirectiveLEB128
4925 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ]
parseDirectiveLEB128(bool Signed)4926 bool AsmParser::parseDirectiveLEB128(bool Signed) {
4927 if (checkForValidSection())
4928 return true;
4929
4930 auto parseOp = [&]() -> bool {
4931 const MCExpr *Value;
4932 if (parseExpression(Value))
4933 return true;
4934 if (Signed)
4935 getStreamer().emitSLEB128Value(Value);
4936 else
4937 getStreamer().emitULEB128Value(Value);
4938 return false;
4939 };
4940
4941 return parseMany(parseOp);
4942 }
4943
4944 /// parseDirectiveSymbolAttribute
4945 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
parseDirectiveSymbolAttribute(MCSymbolAttr Attr)4946 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
4947 auto parseOp = [&]() -> bool {
4948 StringRef Name;
4949 SMLoc Loc = getTok().getLoc();
4950 if (parseIdentifier(Name))
4951 return Error(Loc, "expected identifier");
4952
4953 if (discardLTOSymbol(Name))
4954 return false;
4955
4956 MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
4957
4958 // Assembler local symbols don't make any sense here. Complain loudly.
4959 if (Sym->isTemporary())
4960 return Error(Loc, "non-local symbol required");
4961
4962 if (!getStreamer().emitSymbolAttribute(Sym, Attr))
4963 return Error(Loc, "unable to emit symbol attribute");
4964 return false;
4965 };
4966
4967 return parseMany(parseOp);
4968 }
4969
4970 /// parseDirectiveComm
4971 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
parseDirectiveComm(bool IsLocal)4972 bool AsmParser::parseDirectiveComm(bool IsLocal) {
4973 if (checkForValidSection())
4974 return true;
4975
4976 SMLoc IDLoc = getLexer().getLoc();
4977 StringRef Name;
4978 if (parseIdentifier(Name))
4979 return TokError("expected identifier in directive");
4980
4981 // Handle the identifier as the key symbol.
4982 MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
4983
4984 if (parseComma())
4985 return true;
4986
4987 int64_t Size;
4988 SMLoc SizeLoc = getLexer().getLoc();
4989 if (parseAbsoluteExpression(Size))
4990 return true;
4991
4992 int64_t Pow2Alignment = 0;
4993 SMLoc Pow2AlignmentLoc;
4994 if (getLexer().is(AsmToken::Comma)) {
4995 Lex();
4996 Pow2AlignmentLoc = getLexer().getLoc();
4997 if (parseAbsoluteExpression(Pow2Alignment))
4998 return true;
4999
5000 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType();
5001 if (IsLocal && LCOMM == LCOMM::NoAlignment)
5002 return Error(Pow2AlignmentLoc, "alignment not supported on this target");
5003
5004 // If this target takes alignments in bytes (not log) validate and convert.
5005 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) ||
5006 (IsLocal && LCOMM == LCOMM::ByteAlignment)) {
5007 if (!isPowerOf2_64(Pow2Alignment))
5008 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
5009 Pow2Alignment = Log2_64(Pow2Alignment);
5010 }
5011 }
5012
5013 if (parseEOL())
5014 return true;
5015
5016 // NOTE: a size of zero for a .comm should create a undefined symbol
5017 // but a size of .lcomm creates a bss symbol of size zero.
5018 if (Size < 0)
5019 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
5020 "be less than zero");
5021
5022 // NOTE: The alignment in the directive is a power of 2 value, the assembler
5023 // may internally end up wanting an alignment in bytes.
5024 // FIXME: Diagnose overflow.
5025 if (Pow2Alignment < 0)
5026 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
5027 "alignment, can't be less than zero");
5028
5029 Sym->redefineIfPossible();
5030 if (!Sym->isUndefined())
5031 return Error(IDLoc, "invalid symbol redefinition");
5032
5033 // Create the Symbol as a common or local common with Size and Pow2Alignment
5034 if (IsLocal) {
5035 getStreamer().emitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment);
5036 return false;
5037 }
5038
5039 getStreamer().emitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
5040 return false;
5041 }
5042
5043 /// parseDirectiveAbort
5044 /// ::= .abort [... message ...]
parseDirectiveAbort()5045 bool AsmParser::parseDirectiveAbort() {
5046 // FIXME: Use loc from directive.
5047 SMLoc Loc = getLexer().getLoc();
5048
5049 StringRef Str = parseStringToEndOfStatement();
5050 if (parseEOL())
5051 return true;
5052
5053 if (Str.empty())
5054 return Error(Loc, ".abort detected. Assembly stopping.");
5055 else
5056 return Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
5057 // FIXME: Actually abort assembly here.
5058
5059 return false;
5060 }
5061
5062 /// parseDirectiveInclude
5063 /// ::= .include "filename"
parseDirectiveInclude()5064 bool AsmParser::parseDirectiveInclude() {
5065 // Allow the strings to have escaped octal character sequence.
5066 std::string Filename;
5067 SMLoc IncludeLoc = getTok().getLoc();
5068
5069 if (check(getTok().isNot(AsmToken::String),
5070 "expected string in '.include' directive") ||
5071 parseEscapedString(Filename) ||
5072 check(getTok().isNot(AsmToken::EndOfStatement),
5073 "unexpected token in '.include' directive") ||
5074 // Attempt to switch the lexer to the included file before consuming the
5075 // end of statement to avoid losing it when we switch.
5076 check(enterIncludeFile(Filename), IncludeLoc,
5077 "Could not find include file '" + Filename + "'"))
5078 return true;
5079
5080 return false;
5081 }
5082
5083 /// parseDirectiveIncbin
5084 /// ::= .incbin "filename" [ , skip [ , count ] ]
parseDirectiveIncbin()5085 bool AsmParser::parseDirectiveIncbin() {
5086 // Allow the strings to have escaped octal character sequence.
5087 std::string Filename;
5088 SMLoc IncbinLoc = getTok().getLoc();
5089 if (check(getTok().isNot(AsmToken::String),
5090 "expected string in '.incbin' directive") ||
5091 parseEscapedString(Filename))
5092 return true;
5093
5094 int64_t Skip = 0;
5095 const MCExpr *Count = nullptr;
5096 SMLoc SkipLoc, CountLoc;
5097 if (parseOptionalToken(AsmToken::Comma)) {
5098 // The skip expression can be omitted while specifying the count, e.g:
5099 // .incbin "filename",,4
5100 if (getTok().isNot(AsmToken::Comma)) {
5101 if (parseTokenLoc(SkipLoc) || parseAbsoluteExpression(Skip))
5102 return true;
5103 }
5104 if (parseOptionalToken(AsmToken::Comma)) {
5105 CountLoc = getTok().getLoc();
5106 if (parseExpression(Count))
5107 return true;
5108 }
5109 }
5110
5111 if (parseEOL())
5112 return true;
5113
5114 if (check(Skip < 0, SkipLoc, "skip is negative"))
5115 return true;
5116
5117 // Attempt to process the included file.
5118 if (processIncbinFile(Filename, Skip, Count, CountLoc))
5119 return Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
5120 return false;
5121 }
5122
5123 /// parseDirectiveIf
5124 /// ::= .if{,eq,ge,gt,le,lt,ne} expression
parseDirectiveIf(SMLoc DirectiveLoc,DirectiveKind DirKind)5125 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) {
5126 TheCondStack.push_back(TheCondState);
5127 TheCondState.TheCond = AsmCond::IfCond;
5128 if (TheCondState.Ignore) {
5129 eatToEndOfStatement();
5130 } else {
5131 int64_t ExprValue;
5132 if (parseAbsoluteExpression(ExprValue) || parseEOL())
5133 return true;
5134
5135 switch (DirKind) {
5136 default:
5137 llvm_unreachable("unsupported directive");
5138 case DK_IF:
5139 case DK_IFNE:
5140 break;
5141 case DK_IFEQ:
5142 ExprValue = ExprValue == 0;
5143 break;
5144 case DK_IFGE:
5145 ExprValue = ExprValue >= 0;
5146 break;
5147 case DK_IFGT:
5148 ExprValue = ExprValue > 0;
5149 break;
5150 case DK_IFLE:
5151 ExprValue = ExprValue <= 0;
5152 break;
5153 case DK_IFLT:
5154 ExprValue = ExprValue < 0;
5155 break;
5156 }
5157
5158 TheCondState.CondMet = ExprValue;
5159 TheCondState.Ignore = !TheCondState.CondMet;
5160 }
5161
5162 return false;
5163 }
5164
5165 /// parseDirectiveIfb
5166 /// ::= .ifb string
parseDirectiveIfb(SMLoc DirectiveLoc,bool ExpectBlank)5167 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
5168 TheCondStack.push_back(TheCondState);
5169 TheCondState.TheCond = AsmCond::IfCond;
5170
5171 if (TheCondState.Ignore) {
5172 eatToEndOfStatement();
5173 } else {
5174 StringRef Str = parseStringToEndOfStatement();
5175
5176 if (parseEOL())
5177 return true;
5178
5179 TheCondState.CondMet = ExpectBlank == Str.empty();
5180 TheCondState.Ignore = !TheCondState.CondMet;
5181 }
5182
5183 return false;
5184 }
5185
5186 /// parseDirectiveIfc
5187 /// ::= .ifc string1, string2
5188 /// ::= .ifnc string1, string2
parseDirectiveIfc(SMLoc DirectiveLoc,bool ExpectEqual)5189 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
5190 TheCondStack.push_back(TheCondState);
5191 TheCondState.TheCond = AsmCond::IfCond;
5192
5193 if (TheCondState.Ignore) {
5194 eatToEndOfStatement();
5195 } else {
5196 StringRef Str1 = parseStringToComma();
5197
5198 if (parseComma())
5199 return true;
5200
5201 StringRef Str2 = parseStringToEndOfStatement();
5202
5203 if (parseEOL())
5204 return true;
5205
5206 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim());
5207 TheCondState.Ignore = !TheCondState.CondMet;
5208 }
5209
5210 return false;
5211 }
5212
5213 /// parseDirectiveIfeqs
5214 /// ::= .ifeqs string1, string2
parseDirectiveIfeqs(SMLoc DirectiveLoc,bool ExpectEqual)5215 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) {
5216 if (Lexer.isNot(AsmToken::String)) {
5217 if (ExpectEqual)
5218 return TokError("expected string parameter for '.ifeqs' directive");
5219 return TokError("expected string parameter for '.ifnes' directive");
5220 }
5221
5222 StringRef String1 = getTok().getStringContents();
5223 Lex();
5224
5225 if (Lexer.isNot(AsmToken::Comma)) {
5226 if (ExpectEqual)
5227 return TokError(
5228 "expected comma after first string for '.ifeqs' directive");
5229 return TokError("expected comma after first string for '.ifnes' directive");
5230 }
5231
5232 Lex();
5233
5234 if (Lexer.isNot(AsmToken::String)) {
5235 if (ExpectEqual)
5236 return TokError("expected string parameter for '.ifeqs' directive");
5237 return TokError("expected string parameter for '.ifnes' directive");
5238 }
5239
5240 StringRef String2 = getTok().getStringContents();
5241 Lex();
5242
5243 TheCondStack.push_back(TheCondState);
5244 TheCondState.TheCond = AsmCond::IfCond;
5245 TheCondState.CondMet = ExpectEqual == (String1 == String2);
5246 TheCondState.Ignore = !TheCondState.CondMet;
5247
5248 return false;
5249 }
5250
5251 /// parseDirectiveIfdef
5252 /// ::= .ifdef symbol
parseDirectiveIfdef(SMLoc DirectiveLoc,bool expect_defined)5253 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
5254 StringRef Name;
5255 TheCondStack.push_back(TheCondState);
5256 TheCondState.TheCond = AsmCond::IfCond;
5257
5258 if (TheCondState.Ignore) {
5259 eatToEndOfStatement();
5260 } else {
5261 if (check(parseIdentifier(Name), "expected identifier after '.ifdef'") ||
5262 parseEOL())
5263 return true;
5264
5265 MCSymbol *Sym = getContext().lookupSymbol(Name);
5266
5267 if (expect_defined)
5268 TheCondState.CondMet = (Sym && !Sym->isUndefined(false));
5269 else
5270 TheCondState.CondMet = (!Sym || Sym->isUndefined(false));
5271 TheCondState.Ignore = !TheCondState.CondMet;
5272 }
5273
5274 return false;
5275 }
5276
5277 /// parseDirectiveElseIf
5278 /// ::= .elseif expression
parseDirectiveElseIf(SMLoc DirectiveLoc)5279 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) {
5280 if (TheCondState.TheCond != AsmCond::IfCond &&
5281 TheCondState.TheCond != AsmCond::ElseIfCond)
5282 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an"
5283 " .if or an .elseif");
5284 TheCondState.TheCond = AsmCond::ElseIfCond;
5285
5286 bool LastIgnoreState = false;
5287 if (!TheCondStack.empty())
5288 LastIgnoreState = TheCondStack.back().Ignore;
5289 if (LastIgnoreState || TheCondState.CondMet) {
5290 TheCondState.Ignore = true;
5291 eatToEndOfStatement();
5292 } else {
5293 int64_t ExprValue;
5294 if (parseAbsoluteExpression(ExprValue))
5295 return true;
5296
5297 if (parseEOL())
5298 return true;
5299
5300 TheCondState.CondMet = ExprValue;
5301 TheCondState.Ignore = !TheCondState.CondMet;
5302 }
5303
5304 return false;
5305 }
5306
5307 /// parseDirectiveElse
5308 /// ::= .else
parseDirectiveElse(SMLoc DirectiveLoc)5309 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) {
5310 if (parseEOL())
5311 return true;
5312
5313 if (TheCondState.TheCond != AsmCond::IfCond &&
5314 TheCondState.TheCond != AsmCond::ElseIfCond)
5315 return Error(DirectiveLoc, "Encountered a .else that doesn't follow "
5316 " an .if or an .elseif");
5317 TheCondState.TheCond = AsmCond::ElseCond;
5318 bool LastIgnoreState = false;
5319 if (!TheCondStack.empty())
5320 LastIgnoreState = TheCondStack.back().Ignore;
5321 if (LastIgnoreState || TheCondState.CondMet)
5322 TheCondState.Ignore = true;
5323 else
5324 TheCondState.Ignore = false;
5325
5326 return false;
5327 }
5328
5329 /// parseDirectiveEnd
5330 /// ::= .end
parseDirectiveEnd(SMLoc DirectiveLoc)5331 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) {
5332 if (parseEOL())
5333 return true;
5334
5335 while (Lexer.isNot(AsmToken::Eof))
5336 Lexer.Lex();
5337
5338 return false;
5339 }
5340
5341 /// parseDirectiveError
5342 /// ::= .err
5343 /// ::= .error [string]
parseDirectiveError(SMLoc L,bool WithMessage)5344 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) {
5345 if (!TheCondStack.empty()) {
5346 if (TheCondStack.back().Ignore) {
5347 eatToEndOfStatement();
5348 return false;
5349 }
5350 }
5351
5352 if (!WithMessage)
5353 return Error(L, ".err encountered");
5354
5355 StringRef Message = ".error directive invoked in source file";
5356 if (Lexer.isNot(AsmToken::EndOfStatement)) {
5357 if (Lexer.isNot(AsmToken::String))
5358 return TokError(".error argument must be a string");
5359
5360 Message = getTok().getStringContents();
5361 Lex();
5362 }
5363
5364 return Error(L, Message);
5365 }
5366
5367 /// parseDirectiveWarning
5368 /// ::= .warning [string]
parseDirectiveWarning(SMLoc L)5369 bool AsmParser::parseDirectiveWarning(SMLoc L) {
5370 if (!TheCondStack.empty()) {
5371 if (TheCondStack.back().Ignore) {
5372 eatToEndOfStatement();
5373 return false;
5374 }
5375 }
5376
5377 StringRef Message = ".warning directive invoked in source file";
5378
5379 if (!parseOptionalToken(AsmToken::EndOfStatement)) {
5380 if (Lexer.isNot(AsmToken::String))
5381 return TokError(".warning argument must be a string");
5382
5383 Message = getTok().getStringContents();
5384 Lex();
5385 if (parseEOL())
5386 return true;
5387 }
5388
5389 return Warning(L, Message);
5390 }
5391
5392 /// parseDirectiveEndIf
5393 /// ::= .endif
parseDirectiveEndIf(SMLoc DirectiveLoc)5394 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) {
5395 if (parseEOL())
5396 return true;
5397
5398 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty())
5399 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow "
5400 "an .if or .else");
5401 if (!TheCondStack.empty()) {
5402 TheCondState = TheCondStack.back();
5403 TheCondStack.pop_back();
5404 }
5405
5406 return false;
5407 }
5408
initializeDirectiveKindMap()5409 void AsmParser::initializeDirectiveKindMap() {
5410 /* Lookup will be done with the directive
5411 * converted to lower case, so all these
5412 * keys should be lower case.
5413 * (target specific directives are handled
5414 * elsewhere)
5415 */
5416 DirectiveKindMap[".set"] = DK_SET;
5417 DirectiveKindMap[".equ"] = DK_EQU;
5418 DirectiveKindMap[".equiv"] = DK_EQUIV;
5419 DirectiveKindMap[".ascii"] = DK_ASCII;
5420 DirectiveKindMap[".asciz"] = DK_ASCIZ;
5421 DirectiveKindMap[".string"] = DK_STRING;
5422 DirectiveKindMap[".byte"] = DK_BYTE;
5423 DirectiveKindMap[".short"] = DK_SHORT;
5424 DirectiveKindMap[".value"] = DK_VALUE;
5425 DirectiveKindMap[".2byte"] = DK_2BYTE;
5426 DirectiveKindMap[".long"] = DK_LONG;
5427 DirectiveKindMap[".int"] = DK_INT;
5428 DirectiveKindMap[".4byte"] = DK_4BYTE;
5429 DirectiveKindMap[".quad"] = DK_QUAD;
5430 DirectiveKindMap[".8byte"] = DK_8BYTE;
5431 DirectiveKindMap[".octa"] = DK_OCTA;
5432 DirectiveKindMap[".single"] = DK_SINGLE;
5433 DirectiveKindMap[".float"] = DK_FLOAT;
5434 DirectiveKindMap[".double"] = DK_DOUBLE;
5435 DirectiveKindMap[".align"] = DK_ALIGN;
5436 DirectiveKindMap[".align32"] = DK_ALIGN32;
5437 DirectiveKindMap[".balign"] = DK_BALIGN;
5438 DirectiveKindMap[".balignw"] = DK_BALIGNW;
5439 DirectiveKindMap[".balignl"] = DK_BALIGNL;
5440 DirectiveKindMap[".p2align"] = DK_P2ALIGN;
5441 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW;
5442 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL;
5443 DirectiveKindMap[".org"] = DK_ORG;
5444 DirectiveKindMap[".fill"] = DK_FILL;
5445 DirectiveKindMap[".zero"] = DK_ZERO;
5446 DirectiveKindMap[".extern"] = DK_EXTERN;
5447 DirectiveKindMap[".globl"] = DK_GLOBL;
5448 DirectiveKindMap[".global"] = DK_GLOBAL;
5449 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE;
5450 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP;
5451 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER;
5452 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN;
5453 DirectiveKindMap[".reference"] = DK_REFERENCE;
5454 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION;
5455 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE;
5456 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN;
5457 DirectiveKindMap[".cold"] = DK_COLD;
5458 DirectiveKindMap[".comm"] = DK_COMM;
5459 DirectiveKindMap[".common"] = DK_COMMON;
5460 DirectiveKindMap[".lcomm"] = DK_LCOMM;
5461 DirectiveKindMap[".abort"] = DK_ABORT;
5462 DirectiveKindMap[".include"] = DK_INCLUDE;
5463 DirectiveKindMap[".incbin"] = DK_INCBIN;
5464 DirectiveKindMap[".code16"] = DK_CODE16;
5465 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC;
5466 DirectiveKindMap[".rept"] = DK_REPT;
5467 DirectiveKindMap[".rep"] = DK_REPT;
5468 DirectiveKindMap[".irp"] = DK_IRP;
5469 DirectiveKindMap[".irpc"] = DK_IRPC;
5470 DirectiveKindMap[".endr"] = DK_ENDR;
5471 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE;
5472 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK;
5473 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK;
5474 DirectiveKindMap[".if"] = DK_IF;
5475 DirectiveKindMap[".ifeq"] = DK_IFEQ;
5476 DirectiveKindMap[".ifge"] = DK_IFGE;
5477 DirectiveKindMap[".ifgt"] = DK_IFGT;
5478 DirectiveKindMap[".ifle"] = DK_IFLE;
5479 DirectiveKindMap[".iflt"] = DK_IFLT;
5480 DirectiveKindMap[".ifne"] = DK_IFNE;
5481 DirectiveKindMap[".ifb"] = DK_IFB;
5482 DirectiveKindMap[".ifnb"] = DK_IFNB;
5483 DirectiveKindMap[".ifc"] = DK_IFC;
5484 DirectiveKindMap[".ifeqs"] = DK_IFEQS;
5485 DirectiveKindMap[".ifnc"] = DK_IFNC;
5486 DirectiveKindMap[".ifnes"] = DK_IFNES;
5487 DirectiveKindMap[".ifdef"] = DK_IFDEF;
5488 DirectiveKindMap[".ifndef"] = DK_IFNDEF;
5489 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF;
5490 DirectiveKindMap[".elseif"] = DK_ELSEIF;
5491 DirectiveKindMap[".else"] = DK_ELSE;
5492 DirectiveKindMap[".end"] = DK_END;
5493 DirectiveKindMap[".endif"] = DK_ENDIF;
5494 DirectiveKindMap[".skip"] = DK_SKIP;
5495 DirectiveKindMap[".space"] = DK_SPACE;
5496 DirectiveKindMap[".file"] = DK_FILE;
5497 DirectiveKindMap[".line"] = DK_LINE;
5498 DirectiveKindMap[".loc"] = DK_LOC;
5499 DirectiveKindMap[".stabs"] = DK_STABS;
5500 DirectiveKindMap[".cv_file"] = DK_CV_FILE;
5501 DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID;
5502 DirectiveKindMap[".cv_loc"] = DK_CV_LOC;
5503 DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE;
5504 DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE;
5505 DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID;
5506 DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE;
5507 DirectiveKindMap[".cv_string"] = DK_CV_STRING;
5508 DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE;
5509 DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS;
5510 DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET;
5511 DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA;
5512 DirectiveKindMap[".sleb128"] = DK_SLEB128;
5513 DirectiveKindMap[".uleb128"] = DK_ULEB128;
5514 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS;
5515 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC;
5516 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC;
5517 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA;
5518 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET;
5519 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET;
5520 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER;
5521 DirectiveKindMap[".cfi_llvm_def_aspace_cfa"] = DK_CFI_LLVM_DEF_ASPACE_CFA;
5522 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET;
5523 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET;
5524 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY;
5525 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA;
5526 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE;
5527 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE;
5528 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE;
5529 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE;
5530 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE;
5531 DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN;
5532 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME;
5533 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED;
5534 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER;
5535 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE;
5536 DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME;
5537 DirectiveKindMap[".macros_on"] = DK_MACROS_ON;
5538 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF;
5539 DirectiveKindMap[".macro"] = DK_MACRO;
5540 DirectiveKindMap[".exitm"] = DK_EXITM;
5541 DirectiveKindMap[".endm"] = DK_ENDM;
5542 DirectiveKindMap[".endmacro"] = DK_ENDMACRO;
5543 DirectiveKindMap[".purgem"] = DK_PURGEM;
5544 DirectiveKindMap[".err"] = DK_ERR;
5545 DirectiveKindMap[".error"] = DK_ERROR;
5546 DirectiveKindMap[".warning"] = DK_WARNING;
5547 DirectiveKindMap[".altmacro"] = DK_ALTMACRO;
5548 DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO;
5549 DirectiveKindMap[".reloc"] = DK_RELOC;
5550 DirectiveKindMap[".dc"] = DK_DC;
5551 DirectiveKindMap[".dc.a"] = DK_DC_A;
5552 DirectiveKindMap[".dc.b"] = DK_DC_B;
5553 DirectiveKindMap[".dc.d"] = DK_DC_D;
5554 DirectiveKindMap[".dc.l"] = DK_DC_L;
5555 DirectiveKindMap[".dc.s"] = DK_DC_S;
5556 DirectiveKindMap[".dc.w"] = DK_DC_W;
5557 DirectiveKindMap[".dc.x"] = DK_DC_X;
5558 DirectiveKindMap[".dcb"] = DK_DCB;
5559 DirectiveKindMap[".dcb.b"] = DK_DCB_B;
5560 DirectiveKindMap[".dcb.d"] = DK_DCB_D;
5561 DirectiveKindMap[".dcb.l"] = DK_DCB_L;
5562 DirectiveKindMap[".dcb.s"] = DK_DCB_S;
5563 DirectiveKindMap[".dcb.w"] = DK_DCB_W;
5564 DirectiveKindMap[".dcb.x"] = DK_DCB_X;
5565 DirectiveKindMap[".ds"] = DK_DS;
5566 DirectiveKindMap[".ds.b"] = DK_DS_B;
5567 DirectiveKindMap[".ds.d"] = DK_DS_D;
5568 DirectiveKindMap[".ds.l"] = DK_DS_L;
5569 DirectiveKindMap[".ds.p"] = DK_DS_P;
5570 DirectiveKindMap[".ds.s"] = DK_DS_S;
5571 DirectiveKindMap[".ds.w"] = DK_DS_W;
5572 DirectiveKindMap[".ds.x"] = DK_DS_X;
5573 DirectiveKindMap[".print"] = DK_PRINT;
5574 DirectiveKindMap[".addrsig"] = DK_ADDRSIG;
5575 DirectiveKindMap[".addrsig_sym"] = DK_ADDRSIG_SYM;
5576 DirectiveKindMap[".pseudoprobe"] = DK_PSEUDO_PROBE;
5577 DirectiveKindMap[".lto_discard"] = DK_LTO_DISCARD;
5578 }
5579
parseMacroLikeBody(SMLoc DirectiveLoc)5580 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) {
5581 AsmToken EndToken, StartToken = getTok();
5582
5583 unsigned NestLevel = 0;
5584 while (true) {
5585 // Check whether we have reached the end of the file.
5586 if (getLexer().is(AsmToken::Eof)) {
5587 printError(DirectiveLoc, "no matching '.endr' in definition");
5588 return nullptr;
5589 }
5590
5591 if (Lexer.is(AsmToken::Identifier) &&
5592 (getTok().getIdentifier() == ".rep" ||
5593 getTok().getIdentifier() == ".rept" ||
5594 getTok().getIdentifier() == ".irp" ||
5595 getTok().getIdentifier() == ".irpc")) {
5596 ++NestLevel;
5597 }
5598
5599 // Otherwise, check whether we have reached the .endr.
5600 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") {
5601 if (NestLevel == 0) {
5602 EndToken = getTok();
5603 Lex();
5604 if (Lexer.isNot(AsmToken::EndOfStatement)) {
5605 printError(getTok().getLoc(),
5606 "unexpected token in '.endr' directive");
5607 return nullptr;
5608 }
5609 break;
5610 }
5611 --NestLevel;
5612 }
5613
5614 // Otherwise, scan till the end of the statement.
5615 eatToEndOfStatement();
5616 }
5617
5618 const char *BodyStart = StartToken.getLoc().getPointer();
5619 const char *BodyEnd = EndToken.getLoc().getPointer();
5620 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
5621
5622 // We Are Anonymous.
5623 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters());
5624 return &MacroLikeBodies.back();
5625 }
5626
instantiateMacroLikeBody(MCAsmMacro * M,SMLoc DirectiveLoc,raw_svector_ostream & OS)5627 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
5628 raw_svector_ostream &OS) {
5629 OS << ".endr\n";
5630
5631 std::unique_ptr<MemoryBuffer> Instantiation =
5632 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
5633
5634 // Create the macro instantiation object and add to the current macro
5635 // instantiation stack.
5636 MacroInstantiation *MI = new MacroInstantiation{
5637 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()};
5638 ActiveMacros.push_back(MI);
5639
5640 // Jump to the macro instantiation and prime the lexer.
5641 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
5642 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
5643 Lex();
5644 }
5645
5646 /// parseDirectiveRept
5647 /// ::= .rep | .rept count
parseDirectiveRept(SMLoc DirectiveLoc,StringRef Dir)5648 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) {
5649 const MCExpr *CountExpr;
5650 SMLoc CountLoc = getTok().getLoc();
5651 if (parseExpression(CountExpr))
5652 return true;
5653
5654 int64_t Count;
5655 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) {
5656 return Error(CountLoc, "unexpected token in '" + Dir + "' directive");
5657 }
5658
5659 if (check(Count < 0, CountLoc, "Count is negative") || parseEOL())
5660 return true;
5661
5662 // Lex the rept definition.
5663 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
5664 if (!M)
5665 return true;
5666
5667 // Macro instantiation is lexical, unfortunately. We construct a new buffer
5668 // to hold the macro body with substitutions.
5669 SmallString<256> Buf;
5670 raw_svector_ostream OS(Buf);
5671 while (Count--) {
5672 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t).
5673 if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc()))
5674 return true;
5675 }
5676 instantiateMacroLikeBody(M, DirectiveLoc, OS);
5677
5678 return false;
5679 }
5680
5681 /// parseDirectiveIrp
5682 /// ::= .irp symbol,values
parseDirectiveIrp(SMLoc DirectiveLoc)5683 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) {
5684 MCAsmMacroParameter Parameter;
5685 MCAsmMacroArguments A;
5686 if (check(parseIdentifier(Parameter.Name),
5687 "expected identifier in '.irp' directive") ||
5688 parseComma() || parseMacroArguments(nullptr, A) || parseEOL())
5689 return true;
5690
5691 // Lex the irp definition.
5692 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
5693 if (!M)
5694 return true;
5695
5696 // Macro instantiation is lexical, unfortunately. We construct a new buffer
5697 // to hold the macro body with substitutions.
5698 SmallString<256> Buf;
5699 raw_svector_ostream OS(Buf);
5700
5701 for (const MCAsmMacroArgument &Arg : A) {
5702 // Note that the AtPseudoVariable is enabled for instantiations of .irp.
5703 // This is undocumented, but GAS seems to support it.
5704 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc()))
5705 return true;
5706 }
5707
5708 instantiateMacroLikeBody(M, DirectiveLoc, OS);
5709
5710 return false;
5711 }
5712
5713 /// parseDirectiveIrpc
5714 /// ::= .irpc symbol,values
parseDirectiveIrpc(SMLoc DirectiveLoc)5715 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) {
5716 MCAsmMacroParameter Parameter;
5717 MCAsmMacroArguments A;
5718
5719 if (check(parseIdentifier(Parameter.Name),
5720 "expected identifier in '.irpc' directive") ||
5721 parseComma() || parseMacroArguments(nullptr, A))
5722 return true;
5723
5724 if (A.size() != 1 || A.front().size() != 1)
5725 return TokError("unexpected token in '.irpc' directive");
5726 if (parseEOL())
5727 return true;
5728
5729 // Lex the irpc definition.
5730 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
5731 if (!M)
5732 return true;
5733
5734 // Macro instantiation is lexical, unfortunately. We construct a new buffer
5735 // to hold the macro body with substitutions.
5736 SmallString<256> Buf;
5737 raw_svector_ostream OS(Buf);
5738
5739 StringRef Values = A.front().front().getString();
5740 for (std::size_t I = 0, End = Values.size(); I != End; ++I) {
5741 MCAsmMacroArgument Arg;
5742 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1));
5743
5744 // Note that the AtPseudoVariable is enabled for instantiations of .irpc.
5745 // This is undocumented, but GAS seems to support it.
5746 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc()))
5747 return true;
5748 }
5749
5750 instantiateMacroLikeBody(M, DirectiveLoc, OS);
5751
5752 return false;
5753 }
5754
parseDirectiveEndr(SMLoc DirectiveLoc)5755 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) {
5756 if (ActiveMacros.empty())
5757 return TokError("unmatched '.endr' directive");
5758
5759 // The only .repl that should get here are the ones created by
5760 // instantiateMacroLikeBody.
5761 assert(getLexer().is(AsmToken::EndOfStatement));
5762
5763 handleMacroExit();
5764 return false;
5765 }
5766
parseDirectiveMSEmit(SMLoc IDLoc,ParseStatementInfo & Info,size_t Len)5767 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info,
5768 size_t Len) {
5769 const MCExpr *Value;
5770 SMLoc ExprLoc = getLexer().getLoc();
5771 if (parseExpression(Value))
5772 return true;
5773 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
5774 if (!MCE)
5775 return Error(ExprLoc, "unexpected expression in _emit");
5776 uint64_t IntValue = MCE->getValue();
5777 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue))
5778 return Error(ExprLoc, "literal value out of range for directive");
5779
5780 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len);
5781 return false;
5782 }
5783
parseDirectiveMSAlign(SMLoc IDLoc,ParseStatementInfo & Info)5784 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) {
5785 const MCExpr *Value;
5786 SMLoc ExprLoc = getLexer().getLoc();
5787 if (parseExpression(Value))
5788 return true;
5789 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
5790 if (!MCE)
5791 return Error(ExprLoc, "unexpected expression in align");
5792 uint64_t IntValue = MCE->getValue();
5793 if (!isPowerOf2_64(IntValue))
5794 return Error(ExprLoc, "literal value not a power of two greater then zero");
5795
5796 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue));
5797 return false;
5798 }
5799
parseDirectivePrint(SMLoc DirectiveLoc)5800 bool AsmParser::parseDirectivePrint(SMLoc DirectiveLoc) {
5801 const AsmToken StrTok = getTok();
5802 Lex();
5803 if (StrTok.isNot(AsmToken::String) || StrTok.getString().front() != '"')
5804 return Error(DirectiveLoc, "expected double quoted string after .print");
5805 if (parseEOL())
5806 return true;
5807 llvm::outs() << StrTok.getStringContents() << '\n';
5808 return false;
5809 }
5810
parseDirectiveAddrsig()5811 bool AsmParser::parseDirectiveAddrsig() {
5812 if (parseEOL())
5813 return true;
5814 getStreamer().emitAddrsig();
5815 return false;
5816 }
5817
parseDirectiveAddrsigSym()5818 bool AsmParser::parseDirectiveAddrsigSym() {
5819 StringRef Name;
5820 if (check(parseIdentifier(Name), "expected identifier") || parseEOL())
5821 return true;
5822 MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
5823 getStreamer().emitAddrsigSym(Sym);
5824 return false;
5825 }
5826
parseDirectivePseudoProbe()5827 bool AsmParser::parseDirectivePseudoProbe() {
5828 int64_t Guid;
5829 int64_t Index;
5830 int64_t Type;
5831 int64_t Attr;
5832
5833 if (getLexer().is(AsmToken::Integer)) {
5834 if (parseIntToken(Guid, "unexpected token in '.pseudoprobe' directive"))
5835 return true;
5836 }
5837
5838 if (getLexer().is(AsmToken::Integer)) {
5839 if (parseIntToken(Index, "unexpected token in '.pseudoprobe' directive"))
5840 return true;
5841 }
5842
5843 if (getLexer().is(AsmToken::Integer)) {
5844 if (parseIntToken(Type, "unexpected token in '.pseudoprobe' directive"))
5845 return true;
5846 }
5847
5848 if (getLexer().is(AsmToken::Integer)) {
5849 if (parseIntToken(Attr, "unexpected token in '.pseudoprobe' directive"))
5850 return true;
5851 }
5852
5853 // Parse inline stack like @ GUID:11:12 @ GUID:1:11 @ GUID:3:21
5854 MCPseudoProbeInlineStack InlineStack;
5855
5856 while (getLexer().is(AsmToken::At)) {
5857 // eat @
5858 Lex();
5859
5860 int64_t CallerGuid = 0;
5861 if (getLexer().is(AsmToken::Integer)) {
5862 if (parseIntToken(CallerGuid,
5863 "unexpected token in '.pseudoprobe' directive"))
5864 return true;
5865 }
5866
5867 // eat colon
5868 if (getLexer().is(AsmToken::Colon))
5869 Lex();
5870
5871 int64_t CallerProbeId = 0;
5872 if (getLexer().is(AsmToken::Integer)) {
5873 if (parseIntToken(CallerProbeId,
5874 "unexpected token in '.pseudoprobe' directive"))
5875 return true;
5876 }
5877
5878 InlineSite Site(CallerGuid, CallerProbeId);
5879 InlineStack.push_back(Site);
5880 }
5881
5882 if (parseEOL())
5883 return true;
5884
5885 getStreamer().emitPseudoProbe(Guid, Index, Type, Attr, InlineStack);
5886 return false;
5887 }
5888
5889 /// parseDirectiveLTODiscard
5890 /// ::= ".lto_discard" [ identifier ( , identifier )* ]
5891 /// The LTO library emits this directive to discard non-prevailing symbols.
5892 /// We ignore symbol assignments and attribute changes for the specified
5893 /// symbols.
parseDirectiveLTODiscard()5894 bool AsmParser::parseDirectiveLTODiscard() {
5895 auto ParseOp = [&]() -> bool {
5896 StringRef Name;
5897 SMLoc Loc = getTok().getLoc();
5898 if (parseIdentifier(Name))
5899 return Error(Loc, "expected identifier");
5900 LTODiscardSymbols.insert(Name);
5901 return false;
5902 };
5903
5904 LTODiscardSymbols.clear();
5905 return parseMany(ParseOp);
5906 }
5907
5908 // We are comparing pointers, but the pointers are relative to a single string.
5909 // Thus, this should always be deterministic.
rewritesSort(const AsmRewrite * AsmRewriteA,const AsmRewrite * AsmRewriteB)5910 static int rewritesSort(const AsmRewrite *AsmRewriteA,
5911 const AsmRewrite *AsmRewriteB) {
5912 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer())
5913 return -1;
5914 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer())
5915 return 1;
5916
5917 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output
5918 // rewrite to the same location. Make sure the SizeDirective rewrite is
5919 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This
5920 // ensures the sort algorithm is stable.
5921 if (AsmRewritePrecedence[AsmRewriteA->Kind] >
5922 AsmRewritePrecedence[AsmRewriteB->Kind])
5923 return -1;
5924
5925 if (AsmRewritePrecedence[AsmRewriteA->Kind] <
5926 AsmRewritePrecedence[AsmRewriteB->Kind])
5927 return 1;
5928 llvm_unreachable("Unstable rewrite sort.");
5929 }
5930
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)5931 bool AsmParser::parseMSInlineAsm(
5932 std::string &AsmString, unsigned &NumOutputs, unsigned &NumInputs,
5933 SmallVectorImpl<std::pair<void *, bool>> &OpDecls,
5934 SmallVectorImpl<std::string> &Constraints,
5935 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII,
5936 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) {
5937 SmallVector<void *, 4> InputDecls;
5938 SmallVector<void *, 4> OutputDecls;
5939 SmallVector<bool, 4> InputDeclsAddressOf;
5940 SmallVector<bool, 4> OutputDeclsAddressOf;
5941 SmallVector<std::string, 4> InputConstraints;
5942 SmallVector<std::string, 4> OutputConstraints;
5943 SmallVector<unsigned, 4> ClobberRegs;
5944
5945 SmallVector<AsmRewrite, 4> AsmStrRewrites;
5946
5947 // Prime the lexer.
5948 Lex();
5949
5950 // While we have input, parse each statement.
5951 unsigned InputIdx = 0;
5952 unsigned OutputIdx = 0;
5953 while (getLexer().isNot(AsmToken::Eof)) {
5954 // Parse curly braces marking block start/end
5955 if (parseCurlyBlockScope(AsmStrRewrites))
5956 continue;
5957
5958 ParseStatementInfo Info(&AsmStrRewrites);
5959 bool StatementErr = parseStatement(Info, &SI);
5960
5961 if (StatementErr || Info.ParseError) {
5962 // Emit pending errors if any exist.
5963 printPendingErrors();
5964 return true;
5965 }
5966
5967 // No pending error should exist here.
5968 assert(!hasPendingError() && "unexpected error from parseStatement");
5969
5970 if (Info.Opcode == ~0U)
5971 continue;
5972
5973 const MCInstrDesc &Desc = MII->get(Info.Opcode);
5974
5975 // Build the list of clobbers, outputs and inputs.
5976 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
5977 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i];
5978
5979 // Register operand.
5980 if (Operand.isReg() && !Operand.needAddressOf() &&
5981 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) {
5982 unsigned NumDefs = Desc.getNumDefs();
5983 // Clobber.
5984 if (NumDefs && Operand.getMCOperandNum() < NumDefs)
5985 ClobberRegs.push_back(Operand.getReg());
5986 continue;
5987 }
5988
5989 // Expr/Input or Output.
5990 StringRef SymName = Operand.getSymName();
5991 if (SymName.empty())
5992 continue;
5993
5994 void *OpDecl = Operand.getOpDecl();
5995 if (!OpDecl)
5996 continue;
5997
5998 StringRef Constraint = Operand.getConstraint();
5999 if (Operand.isImm()) {
6000 // Offset as immediate
6001 if (Operand.isOffsetOfLocal())
6002 Constraint = "r";
6003 else
6004 Constraint = "i";
6005 }
6006
6007 bool isOutput = (i == 1) && Desc.mayStore();
6008 SMLoc Start = SMLoc::getFromPointer(SymName.data());
6009 if (isOutput) {
6010 ++InputIdx;
6011 OutputDecls.push_back(OpDecl);
6012 OutputDeclsAddressOf.push_back(Operand.needAddressOf());
6013 OutputConstraints.push_back(("=" + Constraint).str());
6014 AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size());
6015 } else {
6016 InputDecls.push_back(OpDecl);
6017 InputDeclsAddressOf.push_back(Operand.needAddressOf());
6018 InputConstraints.push_back(Constraint.str());
6019 if (Desc.OpInfo[i - 1].isBranchTarget())
6020 AsmStrRewrites.emplace_back(AOK_CallInput, Start, SymName.size());
6021 else
6022 AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size());
6023 }
6024 }
6025
6026 // Consider implicit defs to be clobbers. Think of cpuid and push.
6027 ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(),
6028 Desc.getNumImplicitDefs());
6029 llvm::append_range(ClobberRegs, ImpDefs);
6030 }
6031
6032 // Set the number of Outputs and Inputs.
6033 NumOutputs = OutputDecls.size();
6034 NumInputs = InputDecls.size();
6035
6036 // Set the unique clobbers.
6037 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end());
6038 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()),
6039 ClobberRegs.end());
6040 Clobbers.assign(ClobberRegs.size(), std::string());
6041 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) {
6042 raw_string_ostream OS(Clobbers[I]);
6043 IP->printRegName(OS, ClobberRegs[I]);
6044 }
6045
6046 // Merge the various outputs and inputs. Output are expected first.
6047 if (NumOutputs || NumInputs) {
6048 unsigned NumExprs = NumOutputs + NumInputs;
6049 OpDecls.resize(NumExprs);
6050 Constraints.resize(NumExprs);
6051 for (unsigned i = 0; i < NumOutputs; ++i) {
6052 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]);
6053 Constraints[i] = OutputConstraints[i];
6054 }
6055 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
6056 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]);
6057 Constraints[j] = InputConstraints[i];
6058 }
6059 }
6060
6061 // Build the IR assembly string.
6062 std::string AsmStringIR;
6063 raw_string_ostream OS(AsmStringIR);
6064 StringRef ASMString =
6065 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer();
6066 const char *AsmStart = ASMString.begin();
6067 const char *AsmEnd = ASMString.end();
6068 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort);
6069 for (auto it = AsmStrRewrites.begin(); it != AsmStrRewrites.end(); ++it) {
6070 const AsmRewrite &AR = *it;
6071 // Check if this has already been covered by another rewrite...
6072 if (AR.Done)
6073 continue;
6074 AsmRewriteKind Kind = AR.Kind;
6075
6076 const char *Loc = AR.Loc.getPointer();
6077 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!");
6078
6079 // Emit everything up to the immediate/expression.
6080 if (unsigned Len = Loc - AsmStart)
6081 OS << StringRef(AsmStart, Len);
6082
6083 // Skip the original expression.
6084 if (Kind == AOK_Skip) {
6085 AsmStart = Loc + AR.Len;
6086 continue;
6087 }
6088
6089 unsigned AdditionalSkip = 0;
6090 // Rewrite expressions in $N notation.
6091 switch (Kind) {
6092 default:
6093 break;
6094 case AOK_IntelExpr:
6095 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression");
6096 if (AR.IntelExp.NeedBracs)
6097 OS << "[";
6098 if (AR.IntelExp.hasBaseReg())
6099 OS << AR.IntelExp.BaseReg;
6100 if (AR.IntelExp.hasIndexReg())
6101 OS << (AR.IntelExp.hasBaseReg() ? " + " : "")
6102 << AR.IntelExp.IndexReg;
6103 if (AR.IntelExp.Scale > 1)
6104 OS << " * $$" << AR.IntelExp.Scale;
6105 if (AR.IntelExp.hasOffset()) {
6106 if (AR.IntelExp.hasRegs())
6107 OS << " + ";
6108 // Fuse this rewrite with a rewrite of the offset name, if present.
6109 StringRef OffsetName = AR.IntelExp.OffsetName;
6110 SMLoc OffsetLoc = SMLoc::getFromPointer(AR.IntelExp.OffsetName.data());
6111 size_t OffsetLen = OffsetName.size();
6112 auto rewrite_it = std::find_if(
6113 it, AsmStrRewrites.end(), [&](const AsmRewrite &FusingAR) {
6114 return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen &&
6115 (FusingAR.Kind == AOK_Input ||
6116 FusingAR.Kind == AOK_CallInput);
6117 });
6118 if (rewrite_it == AsmStrRewrites.end()) {
6119 OS << "offset " << OffsetName;
6120 } else if (rewrite_it->Kind == AOK_CallInput) {
6121 OS << "${" << InputIdx++ << ":P}";
6122 rewrite_it->Done = true;
6123 } else {
6124 OS << '$' << InputIdx++;
6125 rewrite_it->Done = true;
6126 }
6127 }
6128 if (AR.IntelExp.Imm || AR.IntelExp.emitImm())
6129 OS << (AR.IntelExp.emitImm() ? "$$" : " + $$") << AR.IntelExp.Imm;
6130 if (AR.IntelExp.NeedBracs)
6131 OS << "]";
6132 break;
6133 case AOK_Label:
6134 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label;
6135 break;
6136 case AOK_Input:
6137 OS << '$' << InputIdx++;
6138 break;
6139 case AOK_CallInput:
6140 OS << "${" << InputIdx++ << ":P}";
6141 break;
6142 case AOK_Output:
6143 OS << '$' << OutputIdx++;
6144 break;
6145 case AOK_SizeDirective:
6146 switch (AR.Val) {
6147 default: break;
6148 case 8: OS << "byte ptr "; break;
6149 case 16: OS << "word ptr "; break;
6150 case 32: OS << "dword ptr "; break;
6151 case 64: OS << "qword ptr "; break;
6152 case 80: OS << "xword ptr "; break;
6153 case 128: OS << "xmmword ptr "; break;
6154 case 256: OS << "ymmword ptr "; break;
6155 }
6156 break;
6157 case AOK_Emit:
6158 OS << ".byte";
6159 break;
6160 case AOK_Align: {
6161 // MS alignment directives are measured in bytes. If the native assembler
6162 // measures alignment in bytes, we can pass it straight through.
6163 OS << ".align";
6164 if (getContext().getAsmInfo()->getAlignmentIsInBytes())
6165 break;
6166
6167 // Alignment is in log2 form, so print that instead and skip the original
6168 // immediate.
6169 unsigned Val = AR.Val;
6170 OS << ' ' << Val;
6171 assert(Val < 10 && "Expected alignment less then 2^10.");
6172 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4;
6173 break;
6174 }
6175 case AOK_EVEN:
6176 OS << ".even";
6177 break;
6178 case AOK_EndOfStatement:
6179 OS << "\n\t";
6180 break;
6181 }
6182
6183 // Skip the original expression.
6184 AsmStart = Loc + AR.Len + AdditionalSkip;
6185 }
6186
6187 // Emit the remainder of the asm string.
6188 if (AsmStart != AsmEnd)
6189 OS << StringRef(AsmStart, AsmEnd - AsmStart);
6190
6191 AsmString = OS.str();
6192 return false;
6193 }
6194
parseAsHLASMLabel(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)6195 bool HLASMAsmParser::parseAsHLASMLabel(ParseStatementInfo &Info,
6196 MCAsmParserSemaCallback *SI) {
6197 AsmToken LabelTok = getTok();
6198 SMLoc LabelLoc = LabelTok.getLoc();
6199 StringRef LabelVal;
6200
6201 if (parseIdentifier(LabelVal))
6202 return Error(LabelLoc, "The HLASM Label has to be an Identifier");
6203
6204 // We have validated whether the token is an Identifier.
6205 // Now we have to validate whether the token is a
6206 // valid HLASM Label.
6207 if (!getTargetParser().isLabel(LabelTok) || checkForValidSection())
6208 return true;
6209
6210 // Lex leading spaces to get to the next operand.
6211 lexLeadingSpaces();
6212
6213 // We shouldn't emit the label if there is nothing else after the label.
6214 // i.e asm("<token>\n")
6215 if (getTok().is(AsmToken::EndOfStatement))
6216 return Error(LabelLoc,
6217 "Cannot have just a label for an HLASM inline asm statement");
6218
6219 MCSymbol *Sym = getContext().getOrCreateSymbol(
6220 getContext().getAsmInfo()->shouldEmitLabelsInUpperCase()
6221 ? LabelVal.upper()
6222 : LabelVal);
6223
6224 getTargetParser().doBeforeLabelEmit(Sym);
6225
6226 // Emit the label.
6227 Out.emitLabel(Sym, LabelLoc);
6228
6229 // If we are generating dwarf for assembly source files then gather the
6230 // info to make a dwarf label entry for this label if needed.
6231 if (enabledGenDwarfForAssembly())
6232 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
6233 LabelLoc);
6234
6235 getTargetParser().onLabelParsed(Sym);
6236
6237 return false;
6238 }
6239
parseAsMachineInstruction(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)6240 bool HLASMAsmParser::parseAsMachineInstruction(ParseStatementInfo &Info,
6241 MCAsmParserSemaCallback *SI) {
6242 AsmToken OperationEntryTok = Lexer.getTok();
6243 SMLoc OperationEntryLoc = OperationEntryTok.getLoc();
6244 StringRef OperationEntryVal;
6245
6246 // Attempt to parse the first token as an Identifier
6247 if (parseIdentifier(OperationEntryVal))
6248 return Error(OperationEntryLoc, "unexpected token at start of statement");
6249
6250 // Once we've parsed the operation entry successfully, lex
6251 // any spaces to get to the OperandEntries.
6252 lexLeadingSpaces();
6253
6254 return parseAndMatchAndEmitTargetInstruction(
6255 Info, OperationEntryVal, OperationEntryTok, OperationEntryLoc);
6256 }
6257
parseStatement(ParseStatementInfo & Info,MCAsmParserSemaCallback * SI)6258 bool HLASMAsmParser::parseStatement(ParseStatementInfo &Info,
6259 MCAsmParserSemaCallback *SI) {
6260 assert(!hasPendingError() && "parseStatement started with pending error");
6261
6262 // Should the first token be interpreted as a HLASM Label.
6263 bool ShouldParseAsHLASMLabel = false;
6264
6265 // If a Name Entry exists, it should occur at the very
6266 // start of the string. In this case, we should parse the
6267 // first non-space token as a Label.
6268 // If the Name entry is missing (i.e. there's some other
6269 // token), then we attempt to parse the first non-space
6270 // token as a Machine Instruction.
6271 if (getTok().isNot(AsmToken::Space))
6272 ShouldParseAsHLASMLabel = true;
6273
6274 // If we have an EndOfStatement (which includes the target's comment
6275 // string) we can appropriately lex it early on)
6276 if (Lexer.is(AsmToken::EndOfStatement)) {
6277 // if this is a line comment we can drop it safely
6278 if (getTok().getString().empty() || getTok().getString().front() == '\r' ||
6279 getTok().getString().front() == '\n')
6280 Out.AddBlankLine();
6281 Lex();
6282 return false;
6283 }
6284
6285 // We have established how to parse the inline asm statement.
6286 // Now we can safely lex any leading spaces to get to the
6287 // first token.
6288 lexLeadingSpaces();
6289
6290 // If we see a new line or carriage return as the first operand,
6291 // after lexing leading spaces, emit the new line and lex the
6292 // EndOfStatement token.
6293 if (Lexer.is(AsmToken::EndOfStatement)) {
6294 if (getTok().getString().front() == '\n' ||
6295 getTok().getString().front() == '\r') {
6296 Out.AddBlankLine();
6297 Lex();
6298 return false;
6299 }
6300 }
6301
6302 // Handle the label first if we have to before processing the rest
6303 // of the tokens as a machine instruction.
6304 if (ShouldParseAsHLASMLabel) {
6305 // If there were any errors while handling and emitting the label,
6306 // early return.
6307 if (parseAsHLASMLabel(Info, SI)) {
6308 // If we know we've failed in parsing, simply eat until end of the
6309 // statement. This ensures that we don't process any other statements.
6310 eatToEndOfStatement();
6311 return true;
6312 }
6313 }
6314
6315 return parseAsMachineInstruction(Info, SI);
6316 }
6317
6318 namespace llvm {
6319 namespace MCParserUtils {
6320
6321 /// Returns whether the given symbol is used anywhere in the given expression,
6322 /// or subexpressions.
isSymbolUsedInExpression(const MCSymbol * Sym,const MCExpr * Value)6323 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) {
6324 switch (Value->getKind()) {
6325 case MCExpr::Binary: {
6326 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value);
6327 return isSymbolUsedInExpression(Sym, BE->getLHS()) ||
6328 isSymbolUsedInExpression(Sym, BE->getRHS());
6329 }
6330 case MCExpr::Target:
6331 case MCExpr::Constant:
6332 return false;
6333 case MCExpr::SymbolRef: {
6334 const MCSymbol &S =
6335 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol();
6336 if (S.isVariable())
6337 return isSymbolUsedInExpression(Sym, S.getVariableValue());
6338 return &S == Sym;
6339 }
6340 case MCExpr::Unary:
6341 return isSymbolUsedInExpression(
6342 Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr());
6343 }
6344
6345 llvm_unreachable("Unknown expr kind!");
6346 }
6347
parseAssignmentExpression(StringRef Name,bool allow_redef,MCAsmParser & Parser,MCSymbol * & Sym,const MCExpr * & Value)6348 bool parseAssignmentExpression(StringRef Name, bool allow_redef,
6349 MCAsmParser &Parser, MCSymbol *&Sym,
6350 const MCExpr *&Value) {
6351
6352 // FIXME: Use better location, we should use proper tokens.
6353 SMLoc EqualLoc = Parser.getTok().getLoc();
6354 if (Parser.parseExpression(Value))
6355 return Parser.TokError("missing expression");
6356
6357 // Note: we don't count b as used in "a = b". This is to allow
6358 // a = b
6359 // b = c
6360
6361 if (Parser.parseEOL())
6362 return true;
6363
6364 // Validate that the LHS is allowed to be a variable (either it has not been
6365 // used as a symbol, or it is an absolute symbol).
6366 Sym = Parser.getContext().lookupSymbol(Name);
6367 if (Sym) {
6368 // Diagnose assignment to a label.
6369 //
6370 // FIXME: Diagnostics. Note the location of the definition as a label.
6371 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
6372 if (isSymbolUsedInExpression(Sym, Value))
6373 return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'");
6374 else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() &&
6375 !Sym->isVariable())
6376 ; // Allow redefinitions of undefined symbols only used in directives.
6377 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
6378 ; // Allow redefinitions of variables that haven't yet been used.
6379 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
6380 return Parser.Error(EqualLoc, "redefinition of '" + Name + "'");
6381 else if (!Sym->isVariable())
6382 return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'");
6383 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
6384 return Parser.Error(EqualLoc,
6385 "invalid reassignment of non-absolute variable '" +
6386 Name + "'");
6387 } else if (Name == ".") {
6388 Parser.getStreamer().emitValueToOffset(Value, 0, EqualLoc);
6389 return false;
6390 } else
6391 Sym = Parser.getContext().getOrCreateSymbol(Name);
6392
6393 Sym->setRedefinable(allow_redef);
6394
6395 return false;
6396 }
6397
6398 } // end namespace MCParserUtils
6399 } // end namespace llvm
6400
6401 /// Create an MCAsmParser instance.
createMCAsmParser(SourceMgr & SM,MCContext & C,MCStreamer & Out,const MCAsmInfo & MAI,unsigned CB)6402 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C,
6403 MCStreamer &Out, const MCAsmInfo &MAI,
6404 unsigned CB) {
6405 if (C.getTargetTriple().isSystemZ() && C.getTargetTriple().isOSzOS())
6406 return new HLASMAsmParser(SM, C, Out, MAI, CB);
6407
6408 return new AsmParser(SM, C, Out, MAI, CB);
6409 }
6410