xref: /llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp (revision 9ca8db352d22444feabd859380252f13826a8aff)
1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===//
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 program is a utility that works like binutils "objdump", that is, it
10 // dumps out a plethora of information about an object file depending on the
11 // flags.
12 //
13 // The flags and output of this program should be near identical to those of
14 // binutils objdump.
15 //
16 //===----------------------------------------------------------------------===//
17 
18 #include "llvm-objdump.h"
19 #include "COFFDump.h"
20 #include "ELFDump.h"
21 #include "MachODump.h"
22 #include "ObjdumpOptID.h"
23 #include "OffloadDump.h"
24 #include "SourcePrinter.h"
25 #include "WasmDump.h"
26 #include "XCOFFDump.h"
27 #include "llvm/ADT/STLExtras.h"
28 #include "llvm/ADT/SetOperations.h"
29 #include "llvm/ADT/StringExtras.h"
30 #include "llvm/ADT/StringSet.h"
31 #include "llvm/ADT/Twine.h"
32 #include "llvm/BinaryFormat/Wasm.h"
33 #include "llvm/DebugInfo/BTF/BTFParser.h"
34 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
35 #include "llvm/DebugInfo/Symbolize/SymbolizableModule.h"
36 #include "llvm/DebugInfo/Symbolize/Symbolize.h"
37 #include "llvm/Debuginfod/BuildIDFetcher.h"
38 #include "llvm/Debuginfod/Debuginfod.h"
39 #include "llvm/Debuginfod/HTTPClient.h"
40 #include "llvm/Demangle/Demangle.h"
41 #include "llvm/MC/MCAsmInfo.h"
42 #include "llvm/MC/MCContext.h"
43 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
44 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h"
45 #include "llvm/MC/MCInst.h"
46 #include "llvm/MC/MCInstPrinter.h"
47 #include "llvm/MC/MCInstrAnalysis.h"
48 #include "llvm/MC/MCInstrInfo.h"
49 #include "llvm/MC/MCObjectFileInfo.h"
50 #include "llvm/MC/MCRegisterInfo.h"
51 #include "llvm/MC/MCTargetOptions.h"
52 #include "llvm/MC/TargetRegistry.h"
53 #include "llvm/Object/Archive.h"
54 #include "llvm/Object/BuildID.h"
55 #include "llvm/Object/COFF.h"
56 #include "llvm/Object/COFFImportFile.h"
57 #include "llvm/Object/ELFObjectFile.h"
58 #include "llvm/Object/ELFTypes.h"
59 #include "llvm/Object/FaultMapParser.h"
60 #include "llvm/Object/MachO.h"
61 #include "llvm/Object/MachOUniversal.h"
62 #include "llvm/Object/ObjectFile.h"
63 #include "llvm/Object/OffloadBinary.h"
64 #include "llvm/Object/Wasm.h"
65 #include "llvm/Option/Arg.h"
66 #include "llvm/Option/ArgList.h"
67 #include "llvm/Option/Option.h"
68 #include "llvm/Support/Casting.h"
69 #include "llvm/Support/Debug.h"
70 #include "llvm/Support/Errc.h"
71 #include "llvm/Support/FileSystem.h"
72 #include "llvm/Support/Format.h"
73 #include "llvm/Support/FormatVariadic.h"
74 #include "llvm/Support/GraphWriter.h"
75 #include "llvm/Support/LLVMDriver.h"
76 #include "llvm/Support/MemoryBuffer.h"
77 #include "llvm/Support/SourceMgr.h"
78 #include "llvm/Support/StringSaver.h"
79 #include "llvm/Support/TargetSelect.h"
80 #include "llvm/Support/WithColor.h"
81 #include "llvm/Support/raw_ostream.h"
82 #include "llvm/TargetParser/Host.h"
83 #include "llvm/TargetParser/Triple.h"
84 #include <algorithm>
85 #include <cctype>
86 #include <cstring>
87 #include <optional>
88 #include <set>
89 #include <system_error>
90 #include <unordered_map>
91 #include <utility>
92 
93 using namespace llvm;
94 using namespace llvm::object;
95 using namespace llvm::objdump;
96 using namespace llvm::opt;
97 
98 namespace {
99 
100 class CommonOptTable : public opt::GenericOptTable {
101 public:
102   CommonOptTable(ArrayRef<Info> OptionInfos, const char *Usage,
103                  const char *Description)
104       : opt::GenericOptTable(OptionInfos), Usage(Usage),
105         Description(Description) {
106     setGroupedShortOptions(true);
107   }
108 
109   void printHelp(StringRef Argv0, bool ShowHidden = false) const {
110     Argv0 = sys::path::filename(Argv0);
111     opt::GenericOptTable::printHelp(outs(), (Argv0 + Usage).str().c_str(),
112                                     Description, ShowHidden, ShowHidden);
113     // TODO Replace this with OptTable API once it adds extrahelp support.
114     outs() << "\nPass @FILE as argument to read options from FILE.\n";
115   }
116 
117 private:
118   const char *Usage;
119   const char *Description;
120 };
121 
122 // ObjdumpOptID is in ObjdumpOptID.h
123 namespace objdump_opt {
124 #define PREFIX(NAME, VALUE)                                                    \
125   static constexpr StringLiteral NAME##_init[] = VALUE;                        \
126   static constexpr ArrayRef<StringLiteral> NAME(NAME##_init,                   \
127                                                 std::size(NAME##_init) - 1);
128 #include "ObjdumpOpts.inc"
129 #undef PREFIX
130 
131 static constexpr opt::OptTable::Info ObjdumpInfoTable[] = {
132 #define OPTION(...)                                                            \
133   LLVM_CONSTRUCT_OPT_INFO_WITH_ID_PREFIX(OBJDUMP_, __VA_ARGS__),
134 #include "ObjdumpOpts.inc"
135 #undef OPTION
136 };
137 } // namespace objdump_opt
138 
139 class ObjdumpOptTable : public CommonOptTable {
140 public:
141   ObjdumpOptTable()
142       : CommonOptTable(objdump_opt::ObjdumpInfoTable,
143                        " [options] <input object files>",
144                        "llvm object file dumper") {}
145 };
146 
147 enum OtoolOptID {
148   OTOOL_INVALID = 0, // This is not an option ID.
149 #define OPTION(...) LLVM_MAKE_OPT_ID_WITH_ID_PREFIX(OTOOL_, __VA_ARGS__),
150 #include "OtoolOpts.inc"
151 #undef OPTION
152 };
153 
154 namespace otool {
155 #define PREFIX(NAME, VALUE)                                                    \
156   static constexpr StringLiteral NAME##_init[] = VALUE;                        \
157   static constexpr ArrayRef<StringLiteral> NAME(NAME##_init,                   \
158                                                 std::size(NAME##_init) - 1);
159 #include "OtoolOpts.inc"
160 #undef PREFIX
161 
162 static constexpr opt::OptTable::Info OtoolInfoTable[] = {
163 #define OPTION(...) LLVM_CONSTRUCT_OPT_INFO_WITH_ID_PREFIX(OTOOL_, __VA_ARGS__),
164 #include "OtoolOpts.inc"
165 #undef OPTION
166 };
167 } // namespace otool
168 
169 class OtoolOptTable : public CommonOptTable {
170 public:
171   OtoolOptTable()
172       : CommonOptTable(otool::OtoolInfoTable, " [option...] [file...]",
173                        "Mach-O object file displaying tool") {}
174 };
175 
176 struct BBAddrMapLabel {
177   std::string BlockLabel;
178   std::string PGOAnalysis;
179 };
180 
181 // This class represents the BBAddrMap and PGOMap associated with a single
182 // function.
183 class BBAddrMapFunctionEntry {
184 public:
185   BBAddrMapFunctionEntry(BBAddrMap AddrMap, PGOAnalysisMap PGOMap)
186       : AddrMap(std::move(AddrMap)), PGOMap(std::move(PGOMap)) {}
187 
188   const BBAddrMap &getAddrMap() const { return AddrMap; }
189 
190   // Returns the PGO string associated with the entry of index `PGOBBEntryIndex`
191   // in `PGOMap`. If PrettyPGOAnalysis is true, prints BFI as relative frequency
192   // and BPI as percentage. Otherwise raw values are displayed.
193   std::string constructPGOLabelString(size_t PGOBBEntryIndex,
194                                       bool PrettyPGOAnalysis) const {
195     if (!PGOMap.FeatEnable.hasPGOAnalysis())
196       return "";
197     std::string PGOString;
198     raw_string_ostream PGOSS(PGOString);
199 
200     PGOSS << " (";
201     if (PGOMap.FeatEnable.FuncEntryCount && PGOBBEntryIndex == 0) {
202       PGOSS << "Entry count: " << Twine(PGOMap.FuncEntryCount);
203       if (PGOMap.FeatEnable.hasPGOAnalysisBBData()) {
204         PGOSS << ", ";
205       }
206     }
207 
208     if (PGOMap.FeatEnable.hasPGOAnalysisBBData()) {
209 
210       assert(PGOBBEntryIndex < PGOMap.BBEntries.size() &&
211              "Expected PGOAnalysisMap and BBAddrMap to have the same entries");
212       const PGOAnalysisMap::PGOBBEntry &PGOBBEntry =
213           PGOMap.BBEntries[PGOBBEntryIndex];
214 
215       if (PGOMap.FeatEnable.BBFreq) {
216         PGOSS << "Frequency: ";
217         if (PrettyPGOAnalysis)
218           printRelativeBlockFreq(PGOSS, PGOMap.BBEntries.front().BlockFreq,
219                                  PGOBBEntry.BlockFreq);
220         else
221           PGOSS << Twine(PGOBBEntry.BlockFreq.getFrequency());
222         if (PGOMap.FeatEnable.BrProb && PGOBBEntry.Successors.size() > 0) {
223           PGOSS << ", ";
224         }
225       }
226       if (PGOMap.FeatEnable.BrProb && PGOBBEntry.Successors.size() > 0) {
227         PGOSS << "Successors: ";
228         interleaveComma(
229             PGOBBEntry.Successors, PGOSS,
230             [&](const PGOAnalysisMap::PGOBBEntry::SuccessorEntry &SE) {
231               PGOSS << "BB" << SE.ID << ":";
232               if (PrettyPGOAnalysis)
233                 PGOSS << "[" << SE.Prob << "]";
234               else
235                 PGOSS.write_hex(SE.Prob.getNumerator());
236             });
237       }
238     }
239     PGOSS << ")";
240 
241     return PGOString;
242   }
243 
244 private:
245   const BBAddrMap AddrMap;
246   const PGOAnalysisMap PGOMap;
247 };
248 
249 // This class represents the BBAddrMap and PGOMap of potentially multiple
250 // functions in a section.
251 class BBAddrMapInfo {
252 public:
253   void clear() {
254     FunctionAddrToMap.clear();
255     RangeBaseAddrToFunctionAddr.clear();
256   }
257 
258   bool empty() const { return FunctionAddrToMap.empty(); }
259 
260   void AddFunctionEntry(BBAddrMap AddrMap, PGOAnalysisMap PGOMap) {
261     uint64_t FunctionAddr = AddrMap.getFunctionAddress();
262     for (size_t I = 1; I < AddrMap.BBRanges.size(); ++I)
263       RangeBaseAddrToFunctionAddr.emplace(AddrMap.BBRanges[I].BaseAddress,
264                                           FunctionAddr);
265     [[maybe_unused]] auto R = FunctionAddrToMap.try_emplace(
266         FunctionAddr, std::move(AddrMap), std::move(PGOMap));
267     assert(R.second && "duplicate function address");
268   }
269 
270   // Returns the BBAddrMap entry for the function associated with `BaseAddress`.
271   // `BaseAddress` could be the function address or the address of a range
272   // associated with that function. Returns `nullptr` if `BaseAddress` is not
273   // mapped to any entry.
274   const BBAddrMapFunctionEntry *getEntryForAddress(uint64_t BaseAddress) const {
275     uint64_t FunctionAddr = BaseAddress;
276     auto S = RangeBaseAddrToFunctionAddr.find(BaseAddress);
277     if (S != RangeBaseAddrToFunctionAddr.end())
278       FunctionAddr = S->second;
279     auto R = FunctionAddrToMap.find(FunctionAddr);
280     if (R == FunctionAddrToMap.end())
281       return nullptr;
282     return &R->second;
283   }
284 
285 private:
286   std::unordered_map<uint64_t, BBAddrMapFunctionEntry> FunctionAddrToMap;
287   std::unordered_map<uint64_t, uint64_t> RangeBaseAddrToFunctionAddr;
288 };
289 
290 } // namespace
291 
292 #define DEBUG_TYPE "objdump"
293 
294 enum class ColorOutput {
295   Auto,
296   Enable,
297   Disable,
298   Invalid,
299 };
300 
301 static uint64_t AdjustVMA;
302 static bool AllHeaders;
303 static std::string ArchName;
304 bool objdump::ArchiveHeaders;
305 bool objdump::Demangle;
306 bool objdump::Disassemble;
307 bool objdump::DisassembleAll;
308 bool objdump::SymbolDescription;
309 bool objdump::TracebackTable;
310 static std::vector<std::string> DisassembleSymbols;
311 static bool DisassembleZeroes;
312 static std::vector<std::string> DisassemblerOptions;
313 static ColorOutput DisassemblyColor;
314 DIDumpType objdump::DwarfDumpType;
315 static bool DynamicRelocations;
316 static bool FaultMapSection;
317 static bool FileHeaders;
318 bool objdump::SectionContents;
319 static std::vector<std::string> InputFilenames;
320 bool objdump::PrintLines;
321 static bool MachOOpt;
322 std::string objdump::MCPU;
323 std::vector<std::string> objdump::MAttrs;
324 bool objdump::ShowRawInsn;
325 bool objdump::LeadingAddr;
326 static bool Offloading;
327 static bool RawClangAST;
328 bool objdump::Relocations;
329 bool objdump::PrintImmHex;
330 bool objdump::PrivateHeaders;
331 std::vector<std::string> objdump::FilterSections;
332 bool objdump::SectionHeaders;
333 static bool ShowAllSymbols;
334 static bool ShowLMA;
335 bool objdump::PrintSource;
336 
337 static uint64_t StartAddress;
338 static bool HasStartAddressFlag;
339 static uint64_t StopAddress = UINT64_MAX;
340 static bool HasStopAddressFlag;
341 
342 bool objdump::SymbolTable;
343 static bool SymbolizeOperands;
344 static bool PrettyPGOAnalysisMap;
345 static bool DynamicSymbolTable;
346 std::string objdump::TripleName;
347 bool objdump::UnwindInfo;
348 static bool Wide;
349 std::string objdump::Prefix;
350 uint32_t objdump::PrefixStrip;
351 
352 DebugVarsFormat objdump::DbgVariables = DVDisabled;
353 
354 int objdump::DbgIndent = 52;
355 
356 static StringSet<> DisasmSymbolSet;
357 StringSet<> objdump::FoundSectionSet;
358 static StringRef ToolName;
359 
360 std::unique_ptr<BuildIDFetcher> BIDFetcher;
361 
362 Dumper::Dumper(const object::ObjectFile &O) : O(O) {
363   WarningHandler = [this](const Twine &Msg) {
364     if (Warnings.insert(Msg.str()).second)
365       reportWarning(Msg, this->O.getFileName());
366     return Error::success();
367   };
368 }
369 
370 void Dumper::reportUniqueWarning(Error Err) {
371   reportUniqueWarning(toString(std::move(Err)));
372 }
373 
374 void Dumper::reportUniqueWarning(const Twine &Msg) {
375   cantFail(WarningHandler(Msg));
376 }
377 
378 static Expected<std::unique_ptr<Dumper>> createDumper(const ObjectFile &Obj) {
379   if (const auto *O = dyn_cast<COFFObjectFile>(&Obj))
380     return createCOFFDumper(*O);
381   if (const auto *O = dyn_cast<ELFObjectFileBase>(&Obj))
382     return createELFDumper(*O);
383   if (const auto *O = dyn_cast<MachOObjectFile>(&Obj))
384     return createMachODumper(*O);
385   if (const auto *O = dyn_cast<WasmObjectFile>(&Obj))
386     return createWasmDumper(*O);
387   if (const auto *O = dyn_cast<XCOFFObjectFile>(&Obj))
388     return createXCOFFDumper(*O);
389 
390   return createStringError(errc::invalid_argument,
391                            "unsupported object file format");
392 }
393 
394 namespace {
395 struct FilterResult {
396   // True if the section should not be skipped.
397   bool Keep;
398 
399   // True if the index counter should be incremented, even if the section should
400   // be skipped. For example, sections may be skipped if they are not included
401   // in the --section flag, but we still want those to count toward the section
402   // count.
403   bool IncrementIndex;
404 };
405 } // namespace
406 
407 static FilterResult checkSectionFilter(object::SectionRef S) {
408   if (FilterSections.empty())
409     return {/*Keep=*/true, /*IncrementIndex=*/true};
410 
411   Expected<StringRef> SecNameOrErr = S.getName();
412   if (!SecNameOrErr) {
413     consumeError(SecNameOrErr.takeError());
414     return {/*Keep=*/false, /*IncrementIndex=*/false};
415   }
416   StringRef SecName = *SecNameOrErr;
417 
418   // StringSet does not allow empty key so avoid adding sections with
419   // no name (such as the section with index 0) here.
420   if (!SecName.empty())
421     FoundSectionSet.insert(SecName);
422 
423   // Only show the section if it's in the FilterSections list, but always
424   // increment so the indexing is stable.
425   return {/*Keep=*/is_contained(FilterSections, SecName),
426           /*IncrementIndex=*/true};
427 }
428 
429 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O,
430                                          uint64_t *Idx) {
431   // Start at UINT64_MAX so that the first index returned after an increment is
432   // zero (after the unsigned wrap).
433   if (Idx)
434     *Idx = UINT64_MAX;
435   return SectionFilter(
436       [Idx](object::SectionRef S) {
437         FilterResult Result = checkSectionFilter(S);
438         if (Idx != nullptr && Result.IncrementIndex)
439           *Idx += 1;
440         return Result.Keep;
441       },
442       O);
443 }
444 
445 std::string objdump::getFileNameForError(const object::Archive::Child &C,
446                                          unsigned Index) {
447   Expected<StringRef> NameOrErr = C.getName();
448   if (NameOrErr)
449     return std::string(NameOrErr.get());
450   // If we have an error getting the name then we print the index of the archive
451   // member. Since we are already in an error state, we just ignore this error.
452   consumeError(NameOrErr.takeError());
453   return "<file index: " + std::to_string(Index) + ">";
454 }
455 
456 void objdump::reportWarning(const Twine &Message, StringRef File) {
457   // Output order between errs() and outs() matters especially for archive
458   // files where the output is per member object.
459   outs().flush();
460   WithColor::warning(errs(), ToolName)
461       << "'" << File << "': " << Message << "\n";
462 }
463 
464 [[noreturn]] void objdump::reportError(StringRef File, const Twine &Message) {
465   outs().flush();
466   WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n";
467   exit(1);
468 }
469 
470 [[noreturn]] void objdump::reportError(Error E, StringRef FileName,
471                                        StringRef ArchiveName,
472                                        StringRef ArchitectureName) {
473   assert(E);
474   outs().flush();
475   WithColor::error(errs(), ToolName);
476   if (ArchiveName != "")
477     errs() << ArchiveName << "(" << FileName << ")";
478   else
479     errs() << "'" << FileName << "'";
480   if (!ArchitectureName.empty())
481     errs() << " (for architecture " << ArchitectureName << ")";
482   errs() << ": ";
483   logAllUnhandledErrors(std::move(E), errs());
484   exit(1);
485 }
486 
487 static void reportCmdLineWarning(const Twine &Message) {
488   WithColor::warning(errs(), ToolName) << Message << "\n";
489 }
490 
491 [[noreturn]] static void reportCmdLineError(const Twine &Message) {
492   WithColor::error(errs(), ToolName) << Message << "\n";
493   exit(1);
494 }
495 
496 static void warnOnNoMatchForSections() {
497   SetVector<StringRef> MissingSections;
498   for (StringRef S : FilterSections) {
499     if (FoundSectionSet.count(S))
500       return;
501     // User may specify a unnamed section. Don't warn for it.
502     if (!S.empty())
503       MissingSections.insert(S);
504   }
505 
506   // Warn only if no section in FilterSections is matched.
507   for (StringRef S : MissingSections)
508     reportCmdLineWarning("section '" + S +
509                          "' mentioned in a -j/--section option, but not "
510                          "found in any input file");
511 }
512 
513 static const Target *getTarget(const ObjectFile *Obj) {
514   // Figure out the target triple.
515   Triple TheTriple("unknown-unknown-unknown");
516   if (TripleName.empty()) {
517     TheTriple = Obj->makeTriple();
518   } else {
519     TheTriple.setTriple(Triple::normalize(TripleName));
520     auto Arch = Obj->getArch();
521     if (Arch == Triple::arm || Arch == Triple::armeb)
522       Obj->setARMSubArch(TheTriple);
523   }
524 
525   // Get the target specific parser.
526   std::string Error;
527   const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple,
528                                                          Error);
529   if (!TheTarget)
530     reportError(Obj->getFileName(), "can't find target: " + Error);
531 
532   // Update the triple name and return the found target.
533   TripleName = TheTriple.getTriple();
534   return TheTarget;
535 }
536 
537 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) {
538   return A.getOffset() < B.getOffset();
539 }
540 
541 static Error getRelocationValueString(const RelocationRef &Rel,
542                                       bool SymbolDescription,
543                                       SmallVectorImpl<char> &Result) {
544   const ObjectFile *Obj = Rel.getObject();
545   if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj))
546     return getELFRelocationValueString(ELF, Rel, Result);
547   if (auto *COFF = dyn_cast<COFFObjectFile>(Obj))
548     return getCOFFRelocationValueString(COFF, Rel, Result);
549   if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj))
550     return getWasmRelocationValueString(Wasm, Rel, Result);
551   if (auto *MachO = dyn_cast<MachOObjectFile>(Obj))
552     return getMachORelocationValueString(MachO, Rel, Result);
553   if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj))
554     return getXCOFFRelocationValueString(*XCOFF, Rel, SymbolDescription,
555                                          Result);
556   llvm_unreachable("unknown object file format");
557 }
558 
559 /// Indicates whether this relocation should hidden when listing
560 /// relocations, usually because it is the trailing part of a multipart
561 /// relocation that will be printed as part of the leading relocation.
562 static bool getHidden(RelocationRef RelRef) {
563   auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject());
564   if (!MachO)
565     return false;
566 
567   unsigned Arch = MachO->getArch();
568   DataRefImpl Rel = RelRef.getRawDataRefImpl();
569   uint64_t Type = MachO->getRelocationType(Rel);
570 
571   // On arches that use the generic relocations, GENERIC_RELOC_PAIR
572   // is always hidden.
573   if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc)
574     return Type == MachO::GENERIC_RELOC_PAIR;
575 
576   if (Arch == Triple::x86_64) {
577     // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
578     // an X86_64_RELOC_SUBTRACTOR.
579     if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) {
580       DataRefImpl RelPrev = Rel;
581       RelPrev.d.a--;
582       uint64_t PrevType = MachO->getRelocationType(RelPrev);
583       if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR)
584         return true;
585     }
586   }
587 
588   return false;
589 }
590 
591 /// Get the column at which we want to start printing the instruction
592 /// disassembly, taking into account anything which appears to the left of it.
593 unsigned objdump::getInstStartColumn(const MCSubtargetInfo &STI) {
594   return !ShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24;
595 }
596 
597 static void AlignToInstStartColumn(size_t Start, const MCSubtargetInfo &STI,
598                                    raw_ostream &OS) {
599   // The output of printInst starts with a tab. Print some spaces so that
600   // the tab has 1 column and advances to the target tab stop.
601   unsigned TabStop = getInstStartColumn(STI);
602   unsigned Column = OS.tell() - Start;
603   OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8);
604 }
605 
606 void objdump::printRawData(ArrayRef<uint8_t> Bytes, uint64_t Address,
607                            formatted_raw_ostream &OS,
608                            MCSubtargetInfo const &STI) {
609   size_t Start = OS.tell();
610   if (LeadingAddr)
611     OS << format("%8" PRIx64 ":", Address);
612   if (ShowRawInsn) {
613     OS << ' ';
614     dumpBytes(Bytes, OS);
615   }
616   AlignToInstStartColumn(Start, STI, OS);
617 }
618 
619 namespace {
620 
621 static bool isAArch64Elf(const ObjectFile &Obj) {
622   const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj);
623   return Elf && Elf->getEMachine() == ELF::EM_AARCH64;
624 }
625 
626 static bool isArmElf(const ObjectFile &Obj) {
627   const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj);
628   return Elf && Elf->getEMachine() == ELF::EM_ARM;
629 }
630 
631 static bool isCSKYElf(const ObjectFile &Obj) {
632   const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj);
633   return Elf && Elf->getEMachine() == ELF::EM_CSKY;
634 }
635 
636 static bool hasMappingSymbols(const ObjectFile &Obj) {
637   return isArmElf(Obj) || isAArch64Elf(Obj) || isCSKYElf(Obj) ;
638 }
639 
640 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName,
641                             const RelocationRef &Rel, uint64_t Address,
642                             bool Is64Bits) {
643   StringRef Fmt = Is64Bits ? "%016" PRIx64 ":  " : "%08" PRIx64 ":  ";
644   SmallString<16> Name;
645   SmallString<32> Val;
646   Rel.getTypeName(Name);
647   if (Error E = getRelocationValueString(Rel, SymbolDescription, Val))
648     reportError(std::move(E), FileName);
649   OS << (Is64Bits || !LeadingAddr ? "\t\t" : "\t\t\t");
650   if (LeadingAddr)
651     OS << format(Fmt.data(), Address);
652   OS << Name << "\t" << Val;
653 }
654 
655 static void printBTFRelocation(formatted_raw_ostream &FOS, llvm::BTFParser &BTF,
656                                object::SectionedAddress Address,
657                                LiveVariablePrinter &LVP) {
658   const llvm::BTF::BPFFieldReloc *Reloc = BTF.findFieldReloc(Address);
659   if (!Reloc)
660     return;
661 
662   SmallString<64> Val;
663   BTF.symbolize(Reloc, Val);
664   FOS << "\t\t";
665   if (LeadingAddr)
666     FOS << format("%016" PRIx64 ":  ", Address.Address + AdjustVMA);
667   FOS << "CO-RE " << Val;
668   LVP.printAfterOtherLine(FOS, true);
669 }
670 
671 class PrettyPrinter {
672 public:
673   virtual ~PrettyPrinter() = default;
674   virtual void
675   printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
676             object::SectionedAddress Address, formatted_raw_ostream &OS,
677             StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
678             StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
679             LiveVariablePrinter &LVP) {
680     if (SP && (PrintSource || PrintLines))
681       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
682     LVP.printBetweenInsts(OS, false);
683 
684     printRawData(Bytes, Address.Address, OS, STI);
685 
686     if (MI) {
687       // See MCInstPrinter::printInst. On targets where a PC relative immediate
688       // is relative to the next instruction and the length of a MCInst is
689       // difficult to measure (x86), this is the address of the next
690       // instruction.
691       uint64_t Addr =
692           Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0);
693       IP.printInst(MI, Addr, "", STI, OS);
694     } else
695       OS << "\t<unknown>";
696   }
697 };
698 PrettyPrinter PrettyPrinterInst;
699 
700 class HexagonPrettyPrinter : public PrettyPrinter {
701 public:
702   void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
703                  formatted_raw_ostream &OS) {
704     uint32_t opcode =
705       (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
706     if (LeadingAddr)
707       OS << format("%8" PRIx64 ":", Address);
708     if (ShowRawInsn) {
709       OS << "\t";
710       dumpBytes(Bytes.slice(0, 4), OS);
711       OS << format("\t%08" PRIx32, opcode);
712     }
713   }
714   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
715                  object::SectionedAddress Address, formatted_raw_ostream &OS,
716                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
717                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
718                  LiveVariablePrinter &LVP) override {
719     if (SP && (PrintSource || PrintLines))
720       SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
721     if (!MI) {
722       printLead(Bytes, Address.Address, OS);
723       OS << " <unknown>";
724       return;
725     }
726     std::string Buffer;
727     {
728       raw_string_ostream TempStream(Buffer);
729       IP.printInst(MI, Address.Address, "", STI, TempStream);
730     }
731     StringRef Contents(Buffer);
732     // Split off bundle attributes
733     auto PacketBundle = Contents.rsplit('\n');
734     // Split off first instruction from the rest
735     auto HeadTail = PacketBundle.first.split('\n');
736     auto Preamble = " { ";
737     auto Separator = "";
738 
739     // Hexagon's packets require relocations to be inline rather than
740     // clustered at the end of the packet.
741     std::vector<RelocationRef>::const_iterator RelCur = Rels->begin();
742     std::vector<RelocationRef>::const_iterator RelEnd = Rels->end();
743     auto PrintReloc = [&]() -> void {
744       while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) {
745         if (RelCur->getOffset() == Address.Address) {
746           printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false);
747           return;
748         }
749         ++RelCur;
750       }
751     };
752 
753     while (!HeadTail.first.empty()) {
754       OS << Separator;
755       Separator = "\n";
756       if (SP && (PrintSource || PrintLines))
757         SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
758       printLead(Bytes, Address.Address, OS);
759       OS << Preamble;
760       Preamble = "   ";
761       StringRef Inst;
762       auto Duplex = HeadTail.first.split('\v');
763       if (!Duplex.second.empty()) {
764         OS << Duplex.first;
765         OS << "; ";
766         Inst = Duplex.second;
767       }
768       else
769         Inst = HeadTail.first;
770       OS << Inst;
771       HeadTail = HeadTail.second.split('\n');
772       if (HeadTail.first.empty())
773         OS << " } " << PacketBundle.second;
774       PrintReloc();
775       Bytes = Bytes.slice(4);
776       Address.Address += 4;
777     }
778   }
779 };
780 HexagonPrettyPrinter HexagonPrettyPrinterInst;
781 
782 class AMDGCNPrettyPrinter : public PrettyPrinter {
783 public:
784   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
785                  object::SectionedAddress Address, formatted_raw_ostream &OS,
786                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
787                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
788                  LiveVariablePrinter &LVP) override {
789     if (SP && (PrintSource || PrintLines))
790       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
791 
792     if (MI) {
793       SmallString<40> InstStr;
794       raw_svector_ostream IS(InstStr);
795 
796       IP.printInst(MI, Address.Address, "", STI, IS);
797 
798       OS << left_justify(IS.str(), 60);
799     } else {
800       // an unrecognized encoding - this is probably data so represent it
801       // using the .long directive, or .byte directive if fewer than 4 bytes
802       // remaining
803       if (Bytes.size() >= 4) {
804         OS << format(
805             "\t.long 0x%08" PRIx32 " ",
806             support::endian::read32<llvm::endianness::little>(Bytes.data()));
807         OS.indent(42);
808       } else {
809           OS << format("\t.byte 0x%02" PRIx8, Bytes[0]);
810           for (unsigned int i = 1; i < Bytes.size(); i++)
811             OS << format(", 0x%02" PRIx8, Bytes[i]);
812           OS.indent(55 - (6 * Bytes.size()));
813       }
814     }
815 
816     OS << format("// %012" PRIX64 ":", Address.Address);
817     if (Bytes.size() >= 4) {
818       // D should be casted to uint32_t here as it is passed by format to
819       // snprintf as vararg.
820       for (uint32_t D :
821            ArrayRef(reinterpret_cast<const support::little32_t *>(Bytes.data()),
822                     Bytes.size() / 4))
823           OS << format(" %08" PRIX32, D);
824     } else {
825       for (unsigned char B : Bytes)
826         OS << format(" %02" PRIX8, B);
827     }
828 
829     if (!Annot.empty())
830       OS << " // " << Annot;
831   }
832 };
833 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst;
834 
835 class BPFPrettyPrinter : public PrettyPrinter {
836 public:
837   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
838                  object::SectionedAddress Address, formatted_raw_ostream &OS,
839                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
840                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
841                  LiveVariablePrinter &LVP) override {
842     if (SP && (PrintSource || PrintLines))
843       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
844     if (LeadingAddr)
845       OS << format("%8" PRId64 ":", Address.Address / 8);
846     if (ShowRawInsn) {
847       OS << "\t";
848       dumpBytes(Bytes, OS);
849     }
850     if (MI)
851       IP.printInst(MI, Address.Address, "", STI, OS);
852     else
853       OS << "\t<unknown>";
854   }
855 };
856 BPFPrettyPrinter BPFPrettyPrinterInst;
857 
858 class ARMPrettyPrinter : public PrettyPrinter {
859 public:
860   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
861                  object::SectionedAddress Address, formatted_raw_ostream &OS,
862                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
863                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
864                  LiveVariablePrinter &LVP) override {
865     if (SP && (PrintSource || PrintLines))
866       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
867     LVP.printBetweenInsts(OS, false);
868 
869     size_t Start = OS.tell();
870     if (LeadingAddr)
871       OS << format("%8" PRIx64 ":", Address.Address);
872     if (ShowRawInsn) {
873       size_t Pos = 0, End = Bytes.size();
874       if (STI.checkFeatures("+thumb-mode")) {
875         for (; Pos + 2 <= End; Pos += 2)
876           OS << ' '
877              << format_hex_no_prefix(
878                     llvm::support::endian::read<uint16_t>(
879                         Bytes.data() + Pos, InstructionEndianness),
880                     4);
881       } else {
882         for (; Pos + 4 <= End; Pos += 4)
883           OS << ' '
884              << format_hex_no_prefix(
885                     llvm::support::endian::read<uint32_t>(
886                         Bytes.data() + Pos, InstructionEndianness),
887                     8);
888       }
889       if (Pos < End) {
890         OS << ' ';
891         dumpBytes(Bytes.slice(Pos), OS);
892       }
893     }
894 
895     AlignToInstStartColumn(Start, STI, OS);
896 
897     if (MI) {
898       IP.printInst(MI, Address.Address, "", STI, OS);
899     } else
900       OS << "\t<unknown>";
901   }
902 
903   void setInstructionEndianness(llvm::endianness Endianness) {
904     InstructionEndianness = Endianness;
905   }
906 
907 private:
908   llvm::endianness InstructionEndianness = llvm::endianness::little;
909 };
910 ARMPrettyPrinter ARMPrettyPrinterInst;
911 
912 class AArch64PrettyPrinter : public PrettyPrinter {
913 public:
914   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
915                  object::SectionedAddress Address, formatted_raw_ostream &OS,
916                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
917                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
918                  LiveVariablePrinter &LVP) override {
919     if (SP && (PrintSource || PrintLines))
920       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
921     LVP.printBetweenInsts(OS, false);
922 
923     size_t Start = OS.tell();
924     if (LeadingAddr)
925       OS << format("%8" PRIx64 ":", Address.Address);
926     if (ShowRawInsn) {
927       size_t Pos = 0, End = Bytes.size();
928       for (; Pos + 4 <= End; Pos += 4)
929         OS << ' '
930            << format_hex_no_prefix(
931                   llvm::support::endian::read<uint32_t>(
932                       Bytes.data() + Pos, llvm::endianness::little),
933                   8);
934       if (Pos < End) {
935         OS << ' ';
936         dumpBytes(Bytes.slice(Pos), OS);
937       }
938     }
939 
940     AlignToInstStartColumn(Start, STI, OS);
941 
942     if (MI) {
943       IP.printInst(MI, Address.Address, "", STI, OS);
944     } else
945       OS << "\t<unknown>";
946   }
947 };
948 AArch64PrettyPrinter AArch64PrettyPrinterInst;
949 
950 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) {
951   switch(Triple.getArch()) {
952   default:
953     return PrettyPrinterInst;
954   case Triple::hexagon:
955     return HexagonPrettyPrinterInst;
956   case Triple::amdgcn:
957     return AMDGCNPrettyPrinterInst;
958   case Triple::bpfel:
959   case Triple::bpfeb:
960     return BPFPrettyPrinterInst;
961   case Triple::arm:
962   case Triple::armeb:
963   case Triple::thumb:
964   case Triple::thumbeb:
965     return ARMPrettyPrinterInst;
966   case Triple::aarch64:
967   case Triple::aarch64_be:
968   case Triple::aarch64_32:
969     return AArch64PrettyPrinterInst;
970   }
971 }
972 
973 class DisassemblerTarget {
974 public:
975   const Target *TheTarget;
976   std::unique_ptr<const MCSubtargetInfo> SubtargetInfo;
977   std::shared_ptr<MCContext> Context;
978   std::unique_ptr<MCDisassembler> DisAsm;
979   std::shared_ptr<MCInstrAnalysis> InstrAnalysis;
980   std::shared_ptr<MCInstPrinter> InstPrinter;
981   PrettyPrinter *Printer;
982 
983   DisassemblerTarget(const Target *TheTarget, ObjectFile &Obj,
984                      StringRef TripleName, StringRef MCPU,
985                      SubtargetFeatures &Features);
986   DisassemblerTarget(DisassemblerTarget &Other, SubtargetFeatures &Features);
987 
988 private:
989   MCTargetOptions Options;
990   std::shared_ptr<const MCRegisterInfo> RegisterInfo;
991   std::shared_ptr<const MCAsmInfo> AsmInfo;
992   std::shared_ptr<const MCInstrInfo> InstrInfo;
993   std::shared_ptr<MCObjectFileInfo> ObjectFileInfo;
994 };
995 
996 DisassemblerTarget::DisassemblerTarget(const Target *TheTarget, ObjectFile &Obj,
997                                        StringRef TripleName, StringRef MCPU,
998                                        SubtargetFeatures &Features)
999     : TheTarget(TheTarget),
1000       Printer(&selectPrettyPrinter(Triple(TripleName))),
1001       RegisterInfo(TheTarget->createMCRegInfo(TripleName)) {
1002   if (!RegisterInfo)
1003     reportError(Obj.getFileName(), "no register info for target " + TripleName);
1004 
1005   // Set up disassembler.
1006   AsmInfo.reset(TheTarget->createMCAsmInfo(*RegisterInfo, TripleName, Options));
1007   if (!AsmInfo)
1008     reportError(Obj.getFileName(), "no assembly info for target " + TripleName);
1009 
1010   SubtargetInfo.reset(
1011       TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString()));
1012   if (!SubtargetInfo)
1013     reportError(Obj.getFileName(),
1014                 "no subtarget info for target " + TripleName);
1015   InstrInfo.reset(TheTarget->createMCInstrInfo());
1016   if (!InstrInfo)
1017     reportError(Obj.getFileName(),
1018                 "no instruction info for target " + TripleName);
1019   Context =
1020       std::make_shared<MCContext>(Triple(TripleName), AsmInfo.get(),
1021                                   RegisterInfo.get(), SubtargetInfo.get());
1022 
1023   // FIXME: for now initialize MCObjectFileInfo with default values
1024   ObjectFileInfo.reset(
1025       TheTarget->createMCObjectFileInfo(*Context, /*PIC=*/false));
1026   Context->setObjectFileInfo(ObjectFileInfo.get());
1027 
1028   DisAsm.reset(TheTarget->createMCDisassembler(*SubtargetInfo, *Context));
1029   if (!DisAsm)
1030     reportError(Obj.getFileName(), "no disassembler for target " + TripleName);
1031 
1032   if (auto *ELFObj = dyn_cast<ELFObjectFileBase>(&Obj))
1033     DisAsm->setABIVersion(ELFObj->getEIdentABIVersion());
1034 
1035   InstrAnalysis.reset(TheTarget->createMCInstrAnalysis(InstrInfo.get()));
1036 
1037   int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1038   InstPrinter.reset(TheTarget->createMCInstPrinter(Triple(TripleName),
1039                                                    AsmPrinterVariant, *AsmInfo,
1040                                                    *InstrInfo, *RegisterInfo));
1041   if (!InstPrinter)
1042     reportError(Obj.getFileName(),
1043                 "no instruction printer for target " + TripleName);
1044   InstPrinter->setPrintImmHex(PrintImmHex);
1045   InstPrinter->setPrintBranchImmAsAddress(true);
1046   InstPrinter->setSymbolizeOperands(SymbolizeOperands);
1047   InstPrinter->setMCInstrAnalysis(InstrAnalysis.get());
1048 
1049   switch (DisassemblyColor) {
1050   case ColorOutput::Enable:
1051     InstPrinter->setUseColor(true);
1052     break;
1053   case ColorOutput::Auto:
1054     InstPrinter->setUseColor(outs().has_colors());
1055     break;
1056   case ColorOutput::Disable:
1057   case ColorOutput::Invalid:
1058     InstPrinter->setUseColor(false);
1059     break;
1060   };
1061 }
1062 
1063 DisassemblerTarget::DisassemblerTarget(DisassemblerTarget &Other,
1064                                        SubtargetFeatures &Features)
1065     : TheTarget(Other.TheTarget),
1066       SubtargetInfo(TheTarget->createMCSubtargetInfo(TripleName, MCPU,
1067                                                      Features.getString())),
1068       Context(Other.Context),
1069       DisAsm(TheTarget->createMCDisassembler(*SubtargetInfo, *Context)),
1070       InstrAnalysis(Other.InstrAnalysis), InstPrinter(Other.InstPrinter),
1071       Printer(Other.Printer), RegisterInfo(Other.RegisterInfo),
1072       AsmInfo(Other.AsmInfo), InstrInfo(Other.InstrInfo),
1073       ObjectFileInfo(Other.ObjectFileInfo) {}
1074 } // namespace
1075 
1076 static uint8_t getElfSymbolType(const ObjectFile &Obj, const SymbolRef &Sym) {
1077   assert(Obj.isELF());
1078   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj))
1079     return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()),
1080                          Obj.getFileName())
1081         ->getType();
1082   if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj))
1083     return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()),
1084                          Obj.getFileName())
1085         ->getType();
1086   if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj))
1087     return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()),
1088                          Obj.getFileName())
1089         ->getType();
1090   if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj))
1091     return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()),
1092                          Obj.getFileName())
1093         ->getType();
1094   llvm_unreachable("Unsupported binary format");
1095 }
1096 
1097 template <class ELFT>
1098 static void
1099 addDynamicElfSymbols(const ELFObjectFile<ELFT> &Obj,
1100                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
1101   for (auto Symbol : Obj.getDynamicSymbolIterators()) {
1102     uint8_t SymbolType = Symbol.getELFType();
1103     if (SymbolType == ELF::STT_SECTION)
1104       continue;
1105 
1106     uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj.getFileName());
1107     // ELFSymbolRef::getAddress() returns size instead of value for common
1108     // symbols which is not desirable for disassembly output. Overriding.
1109     if (SymbolType == ELF::STT_COMMON)
1110       Address = unwrapOrError(Obj.getSymbol(Symbol.getRawDataRefImpl()),
1111                               Obj.getFileName())
1112                     ->st_value;
1113 
1114     StringRef Name = unwrapOrError(Symbol.getName(), Obj.getFileName());
1115     if (Name.empty())
1116       continue;
1117 
1118     section_iterator SecI =
1119         unwrapOrError(Symbol.getSection(), Obj.getFileName());
1120     if (SecI == Obj.section_end())
1121       continue;
1122 
1123     AllSymbols[*SecI].emplace_back(Address, Name, SymbolType);
1124   }
1125 }
1126 
1127 static void
1128 addDynamicElfSymbols(const ELFObjectFileBase &Obj,
1129                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
1130   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj))
1131     addDynamicElfSymbols(*Elf32LEObj, AllSymbols);
1132   else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj))
1133     addDynamicElfSymbols(*Elf64LEObj, AllSymbols);
1134   else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj))
1135     addDynamicElfSymbols(*Elf32BEObj, AllSymbols);
1136   else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj))
1137     addDynamicElfSymbols(*Elf64BEObj, AllSymbols);
1138   else
1139     llvm_unreachable("Unsupported binary format");
1140 }
1141 
1142 static std::optional<SectionRef> getWasmCodeSection(const WasmObjectFile &Obj) {
1143   for (auto SecI : Obj.sections()) {
1144     const WasmSection &Section = Obj.getWasmSection(SecI);
1145     if (Section.Type == wasm::WASM_SEC_CODE)
1146       return SecI;
1147   }
1148   return std::nullopt;
1149 }
1150 
1151 static void
1152 addMissingWasmCodeSymbols(const WasmObjectFile &Obj,
1153                           std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
1154   std::optional<SectionRef> Section = getWasmCodeSection(Obj);
1155   if (!Section)
1156     return;
1157   SectionSymbolsTy &Symbols = AllSymbols[*Section];
1158 
1159   std::set<uint64_t> SymbolAddresses;
1160   for (const auto &Sym : Symbols)
1161     SymbolAddresses.insert(Sym.Addr);
1162 
1163   for (const wasm::WasmFunction &Function : Obj.functions()) {
1164     // This adjustment mirrors the one in WasmObjectFile::getSymbolAddress.
1165     uint32_t Adjustment = Obj.isRelocatableObject() || Obj.isSharedObject()
1166                               ? 0
1167                               : Section->getAddress();
1168     uint64_t Address = Function.CodeSectionOffset + Adjustment;
1169     // Only add fallback symbols for functions not already present in the symbol
1170     // table.
1171     if (SymbolAddresses.count(Address))
1172       continue;
1173     // This function has no symbol, so it should have no SymbolName.
1174     assert(Function.SymbolName.empty());
1175     // We use DebugName for the name, though it may be empty if there is no
1176     // "name" custom section, or that section is missing a name for this
1177     // function.
1178     StringRef Name = Function.DebugName;
1179     Symbols.emplace_back(Address, Name, ELF::STT_NOTYPE);
1180   }
1181 }
1182 
1183 static void addPltEntries(const ObjectFile &Obj,
1184                           std::map<SectionRef, SectionSymbolsTy> &AllSymbols,
1185                           StringSaver &Saver) {
1186   auto *ElfObj = dyn_cast<ELFObjectFileBase>(&Obj);
1187   if (!ElfObj)
1188     return;
1189   DenseMap<StringRef, SectionRef> Sections;
1190   for (SectionRef Section : Obj.sections()) {
1191     Expected<StringRef> SecNameOrErr = Section.getName();
1192     if (!SecNameOrErr) {
1193       consumeError(SecNameOrErr.takeError());
1194       continue;
1195     }
1196     Sections[*SecNameOrErr] = Section;
1197   }
1198   for (auto Plt : ElfObj->getPltEntries()) {
1199     if (Plt.Symbol) {
1200       SymbolRef Symbol(*Plt.Symbol, ElfObj);
1201       uint8_t SymbolType = getElfSymbolType(Obj, Symbol);
1202       if (Expected<StringRef> NameOrErr = Symbol.getName()) {
1203         if (!NameOrErr->empty())
1204           AllSymbols[Sections[Plt.Section]].emplace_back(
1205               Plt.Address, Saver.save((*NameOrErr + "@plt").str()), SymbolType);
1206         continue;
1207       } else {
1208         // The warning has been reported in disassembleObject().
1209         consumeError(NameOrErr.takeError());
1210       }
1211     }
1212     reportWarning("PLT entry at 0x" + Twine::utohexstr(Plt.Address) +
1213                       " references an invalid symbol",
1214                   Obj.getFileName());
1215   }
1216 }
1217 
1218 // Normally the disassembly output will skip blocks of zeroes. This function
1219 // returns the number of zero bytes that can be skipped when dumping the
1220 // disassembly of the instructions in Buf.
1221 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) {
1222   // Find the number of leading zeroes.
1223   size_t N = 0;
1224   while (N < Buf.size() && !Buf[N])
1225     ++N;
1226 
1227   // We may want to skip blocks of zero bytes, but unless we see
1228   // at least 8 of them in a row.
1229   if (N < 8)
1230     return 0;
1231 
1232   // We skip zeroes in multiples of 4 because do not want to truncate an
1233   // instruction if it starts with a zero byte.
1234   return N & ~0x3;
1235 }
1236 
1237 // Returns a map from sections to their relocations.
1238 static std::map<SectionRef, std::vector<RelocationRef>>
1239 getRelocsMap(object::ObjectFile const &Obj) {
1240   std::map<SectionRef, std::vector<RelocationRef>> Ret;
1241   uint64_t I = (uint64_t)-1;
1242   for (SectionRef Sec : Obj.sections()) {
1243     ++I;
1244     Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection();
1245     if (!RelocatedOrErr)
1246       reportError(Obj.getFileName(),
1247                   "section (" + Twine(I) +
1248                       "): failed to get a relocated section: " +
1249                       toString(RelocatedOrErr.takeError()));
1250 
1251     section_iterator Relocated = *RelocatedOrErr;
1252     if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep)
1253       continue;
1254     std::vector<RelocationRef> &V = Ret[*Relocated];
1255     append_range(V, Sec.relocations());
1256     // Sort relocations by address.
1257     llvm::stable_sort(V, isRelocAddressLess);
1258   }
1259   return Ret;
1260 }
1261 
1262 // Used for --adjust-vma to check if address should be adjusted by the
1263 // specified value for a given section.
1264 // For ELF we do not adjust non-allocatable sections like debug ones,
1265 // because they are not loadable.
1266 // TODO: implement for other file formats.
1267 static bool shouldAdjustVA(const SectionRef &Section) {
1268   const ObjectFile *Obj = Section.getObject();
1269   if (Obj->isELF())
1270     return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC;
1271   return false;
1272 }
1273 
1274 
1275 typedef std::pair<uint64_t, char> MappingSymbolPair;
1276 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols,
1277                                  uint64_t Address) {
1278   auto It =
1279       partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) {
1280         return Val.first <= Address;
1281       });
1282   // Return zero for any address before the first mapping symbol; this means
1283   // we should use the default disassembly mode, depending on the target.
1284   if (It == MappingSymbols.begin())
1285     return '\x00';
1286   return (It - 1)->second;
1287 }
1288 
1289 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index,
1290                                uint64_t End, const ObjectFile &Obj,
1291                                ArrayRef<uint8_t> Bytes,
1292                                ArrayRef<MappingSymbolPair> MappingSymbols,
1293                                const MCSubtargetInfo &STI, raw_ostream &OS) {
1294   llvm::endianness Endian =
1295       Obj.isLittleEndian() ? llvm::endianness::little : llvm::endianness::big;
1296   size_t Start = OS.tell();
1297   OS << format("%8" PRIx64 ": ", SectionAddr + Index);
1298   if (Index + 4 <= End) {
1299     dumpBytes(Bytes.slice(Index, 4), OS);
1300     AlignToInstStartColumn(Start, STI, OS);
1301     OS << "\t.word\t"
1302            << format_hex(support::endian::read32(Bytes.data() + Index, Endian),
1303                          10);
1304     return 4;
1305   }
1306   if (Index + 2 <= End) {
1307     dumpBytes(Bytes.slice(Index, 2), OS);
1308     AlignToInstStartColumn(Start, STI, OS);
1309     OS << "\t.short\t"
1310        << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 6);
1311     return 2;
1312   }
1313   dumpBytes(Bytes.slice(Index, 1), OS);
1314   AlignToInstStartColumn(Start, STI, OS);
1315   OS << "\t.byte\t" << format_hex(Bytes[Index], 4);
1316   return 1;
1317 }
1318 
1319 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End,
1320                         ArrayRef<uint8_t> Bytes) {
1321   // print out data up to 8 bytes at a time in hex and ascii
1322   uint8_t AsciiData[9] = {'\0'};
1323   uint8_t Byte;
1324   int NumBytes = 0;
1325 
1326   for (; Index < End; ++Index) {
1327     if (NumBytes == 0)
1328       outs() << format("%8" PRIx64 ":", SectionAddr + Index);
1329     Byte = Bytes.slice(Index)[0];
1330     outs() << format(" %02x", Byte);
1331     AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.';
1332 
1333     uint8_t IndentOffset = 0;
1334     NumBytes++;
1335     if (Index == End - 1 || NumBytes > 8) {
1336       // Indent the space for less than 8 bytes data.
1337       // 2 spaces for byte and one for space between bytes
1338       IndentOffset = 3 * (8 - NumBytes);
1339       for (int Excess = NumBytes; Excess < 8; Excess++)
1340         AsciiData[Excess] = '\0';
1341       NumBytes = 8;
1342     }
1343     if (NumBytes == 8) {
1344       AsciiData[8] = '\0';
1345       outs() << std::string(IndentOffset, ' ') << "         ";
1346       outs() << reinterpret_cast<char *>(AsciiData);
1347       outs() << '\n';
1348       NumBytes = 0;
1349     }
1350   }
1351 }
1352 
1353 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile &Obj,
1354                                        const SymbolRef &Symbol,
1355                                        bool IsMappingSymbol) {
1356   const StringRef FileName = Obj.getFileName();
1357   const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName);
1358   const StringRef Name = unwrapOrError(Symbol.getName(), FileName);
1359 
1360   if (Obj.isXCOFF() && (SymbolDescription || TracebackTable)) {
1361     const auto &XCOFFObj = cast<XCOFFObjectFile>(Obj);
1362     DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl();
1363 
1364     const uint32_t SymbolIndex = XCOFFObj.getSymbolIndex(SymbolDRI.p);
1365     std::optional<XCOFF::StorageMappingClass> Smc =
1366         getXCOFFSymbolCsectSMC(XCOFFObj, Symbol);
1367     return SymbolInfoTy(Smc, Addr, Name, SymbolIndex,
1368                         isLabel(XCOFFObj, Symbol));
1369   } else if (Obj.isXCOFF()) {
1370     const SymbolRef::Type SymType = unwrapOrError(Symbol.getType(), FileName);
1371     return SymbolInfoTy(Addr, Name, SymType, /*IsMappingSymbol=*/false,
1372                         /*IsXCOFF=*/true);
1373   } else if (Obj.isWasm()) {
1374     uint8_t SymType =
1375         cast<WasmObjectFile>(&Obj)->getWasmSymbol(Symbol).Info.Kind;
1376     return SymbolInfoTy(Addr, Name, SymType, false);
1377   } else {
1378     uint8_t Type =
1379         Obj.isELF() ? getElfSymbolType(Obj, Symbol) : (uint8_t)ELF::STT_NOTYPE;
1380     return SymbolInfoTy(Addr, Name, Type, IsMappingSymbol);
1381   }
1382 }
1383 
1384 static SymbolInfoTy createDummySymbolInfo(const ObjectFile &Obj,
1385                                           const uint64_t Addr, StringRef &Name,
1386                                           uint8_t Type) {
1387   if (Obj.isXCOFF() && (SymbolDescription || TracebackTable))
1388     return SymbolInfoTy(std::nullopt, Addr, Name, std::nullopt, false);
1389   if (Obj.isWasm())
1390     return SymbolInfoTy(Addr, Name, wasm::WASM_SYMBOL_TYPE_SECTION);
1391   return SymbolInfoTy(Addr, Name, Type);
1392 }
1393 
1394 static void collectBBAddrMapLabels(
1395     const BBAddrMapInfo &FullAddrMap, uint64_t SectionAddr, uint64_t Start,
1396     uint64_t End,
1397     std::unordered_map<uint64_t, std::vector<BBAddrMapLabel>> &Labels) {
1398   if (FullAddrMap.empty())
1399     return;
1400   Labels.clear();
1401   uint64_t StartAddress = SectionAddr + Start;
1402   uint64_t EndAddress = SectionAddr + End;
1403   const BBAddrMapFunctionEntry *FunctionMap =
1404       FullAddrMap.getEntryForAddress(StartAddress);
1405   if (!FunctionMap)
1406     return;
1407   std::optional<size_t> BBRangeIndex =
1408       FunctionMap->getAddrMap().getBBRangeIndexForBaseAddress(StartAddress);
1409   if (!BBRangeIndex)
1410     return;
1411   size_t NumBBEntriesBeforeRange = 0;
1412   for (size_t I = 0; I < *BBRangeIndex; ++I)
1413     NumBBEntriesBeforeRange +=
1414         FunctionMap->getAddrMap().BBRanges[I].BBEntries.size();
1415   const auto &BBRange = FunctionMap->getAddrMap().BBRanges[*BBRangeIndex];
1416   for (size_t I = 0; I < BBRange.BBEntries.size(); ++I) {
1417     const BBAddrMap::BBEntry &BBEntry = BBRange.BBEntries[I];
1418     uint64_t BBAddress = BBEntry.Offset + BBRange.BaseAddress;
1419     if (BBAddress >= EndAddress)
1420       continue;
1421 
1422     std::string LabelString = ("BB" + Twine(BBEntry.ID)).str();
1423     Labels[BBAddress].push_back(
1424         {LabelString, FunctionMap->constructPGOLabelString(
1425                           NumBBEntriesBeforeRange + I, PrettyPGOAnalysisMap)});
1426   }
1427 }
1428 
1429 static void
1430 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, MCInstrAnalysis *MIA,
1431                           MCDisassembler *DisAsm, MCInstPrinter *IP,
1432                           const MCSubtargetInfo *STI, uint64_t SectionAddr,
1433                           uint64_t Start, uint64_t End,
1434                           std::unordered_map<uint64_t, std::string> &Labels) {
1435   // So far only supports PowerPC and X86.
1436   const bool isPPC = STI->getTargetTriple().isPPC();
1437   if (!isPPC && !STI->getTargetTriple().isX86())
1438     return;
1439 
1440   if (MIA)
1441     MIA->resetState();
1442 
1443   Labels.clear();
1444   unsigned LabelCount = 0;
1445   Start += SectionAddr;
1446   End += SectionAddr;
1447   const bool isXCOFF = STI->getTargetTriple().isOSBinFormatXCOFF();
1448   for (uint64_t Index = Start; Index < End;) {
1449     // Disassemble a real instruction and record function-local branch labels.
1450     MCInst Inst;
1451     uint64_t Size;
1452     ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index - SectionAddr);
1453     bool Disassembled =
1454         DisAsm->getInstruction(Inst, Size, ThisBytes, Index, nulls());
1455     if (Size == 0)
1456       Size = std::min<uint64_t>(ThisBytes.size(),
1457                                 DisAsm->suggestBytesToSkip(ThisBytes, Index));
1458 
1459     if (MIA) {
1460       if (Disassembled) {
1461         uint64_t Target;
1462         bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target);
1463         if (TargetKnown && (Target >= Start && Target < End) &&
1464             !Labels.count(Target)) {
1465           // On PowerPC and AIX, a function call is encoded as a branch to 0.
1466           // On other PowerPC platforms (ELF), a function call is encoded as
1467           // a branch to self. Do not add a label for these cases.
1468           if (!(isPPC &&
1469                 ((Target == 0 && isXCOFF) || (Target == Index && !isXCOFF))))
1470             Labels[Target] = ("L" + Twine(LabelCount++)).str();
1471         }
1472         MIA->updateState(Inst, Index);
1473       } else
1474         MIA->resetState();
1475     }
1476     Index += Size;
1477   }
1478 }
1479 
1480 // Create an MCSymbolizer for the target and add it to the MCDisassembler.
1481 // This is currently only used on AMDGPU, and assumes the format of the
1482 // void * argument passed to AMDGPU's createMCSymbolizer.
1483 static void addSymbolizer(
1484     MCContext &Ctx, const Target *Target, StringRef TripleName,
1485     MCDisassembler *DisAsm, uint64_t SectionAddr, ArrayRef<uint8_t> Bytes,
1486     SectionSymbolsTy &Symbols,
1487     std::vector<std::unique_ptr<std::string>> &SynthesizedLabelNames) {
1488 
1489   std::unique_ptr<MCRelocationInfo> RelInfo(
1490       Target->createMCRelocationInfo(TripleName, Ctx));
1491   if (!RelInfo)
1492     return;
1493   std::unique_ptr<MCSymbolizer> Symbolizer(Target->createMCSymbolizer(
1494       TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1495   MCSymbolizer *SymbolizerPtr = &*Symbolizer;
1496   DisAsm->setSymbolizer(std::move(Symbolizer));
1497 
1498   if (!SymbolizeOperands)
1499     return;
1500 
1501   // Synthesize labels referenced by branch instructions by
1502   // disassembling, discarding the output, and collecting the referenced
1503   // addresses from the symbolizer.
1504   for (size_t Index = 0; Index != Bytes.size();) {
1505     MCInst Inst;
1506     uint64_t Size;
1507     ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index);
1508     const uint64_t ThisAddr = SectionAddr + Index;
1509     DisAsm->getInstruction(Inst, Size, ThisBytes, ThisAddr, nulls());
1510     if (Size == 0)
1511       Size = std::min<uint64_t>(ThisBytes.size(),
1512                                 DisAsm->suggestBytesToSkip(ThisBytes, Index));
1513     Index += Size;
1514   }
1515   ArrayRef<uint64_t> LabelAddrsRef = SymbolizerPtr->getReferencedAddresses();
1516   // Copy and sort to remove duplicates.
1517   std::vector<uint64_t> LabelAddrs;
1518   LabelAddrs.insert(LabelAddrs.end(), LabelAddrsRef.begin(),
1519                     LabelAddrsRef.end());
1520   llvm::sort(LabelAddrs);
1521   LabelAddrs.resize(std::unique(LabelAddrs.begin(), LabelAddrs.end()) -
1522                     LabelAddrs.begin());
1523   // Add the labels.
1524   for (unsigned LabelNum = 0; LabelNum != LabelAddrs.size(); ++LabelNum) {
1525     auto Name = std::make_unique<std::string>();
1526     *Name = (Twine("L") + Twine(LabelNum)).str();
1527     SynthesizedLabelNames.push_back(std::move(Name));
1528     Symbols.push_back(SymbolInfoTy(
1529         LabelAddrs[LabelNum], *SynthesizedLabelNames.back(), ELF::STT_NOTYPE));
1530   }
1531   llvm::stable_sort(Symbols);
1532   // Recreate the symbolizer with the new symbols list.
1533   RelInfo.reset(Target->createMCRelocationInfo(TripleName, Ctx));
1534   Symbolizer.reset(Target->createMCSymbolizer(
1535       TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1536   DisAsm->setSymbolizer(std::move(Symbolizer));
1537 }
1538 
1539 static StringRef getSegmentName(const MachOObjectFile *MachO,
1540                                 const SectionRef &Section) {
1541   if (MachO) {
1542     DataRefImpl DR = Section.getRawDataRefImpl();
1543     StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
1544     return SegmentName;
1545   }
1546   return "";
1547 }
1548 
1549 static void emitPostInstructionInfo(formatted_raw_ostream &FOS,
1550                                     const MCAsmInfo &MAI,
1551                                     const MCSubtargetInfo &STI,
1552                                     StringRef Comments,
1553                                     LiveVariablePrinter &LVP) {
1554   do {
1555     if (!Comments.empty()) {
1556       // Emit a line of comments.
1557       StringRef Comment;
1558       std::tie(Comment, Comments) = Comments.split('\n');
1559       // MAI.getCommentColumn() assumes that instructions are printed at the
1560       // position of 8, while getInstStartColumn() returns the actual position.
1561       unsigned CommentColumn =
1562           MAI.getCommentColumn() - 8 + getInstStartColumn(STI);
1563       FOS.PadToColumn(CommentColumn);
1564       FOS << MAI.getCommentString() << ' ' << Comment;
1565     }
1566     LVP.printAfterInst(FOS);
1567     FOS << '\n';
1568   } while (!Comments.empty());
1569   FOS.flush();
1570 }
1571 
1572 static void createFakeELFSections(ObjectFile &Obj) {
1573   assert(Obj.isELF());
1574   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj))
1575     Elf32LEObj->createFakeSections();
1576   else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj))
1577     Elf64LEObj->createFakeSections();
1578   else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj))
1579     Elf32BEObj->createFakeSections();
1580   else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj))
1581     Elf64BEObj->createFakeSections();
1582   else
1583     llvm_unreachable("Unsupported binary format");
1584 }
1585 
1586 // Tries to fetch a more complete version of the given object file using its
1587 // Build ID. Returns std::nullopt if nothing was found.
1588 static std::optional<OwningBinary<Binary>>
1589 fetchBinaryByBuildID(const ObjectFile &Obj) {
1590   object::BuildIDRef BuildID = getBuildID(&Obj);
1591   if (BuildID.empty())
1592     return std::nullopt;
1593   std::optional<std::string> Path = BIDFetcher->fetch(BuildID);
1594   if (!Path)
1595     return std::nullopt;
1596   Expected<OwningBinary<Binary>> DebugBinary = createBinary(*Path);
1597   if (!DebugBinary) {
1598     reportWarning(toString(DebugBinary.takeError()), *Path);
1599     return std::nullopt;
1600   }
1601   return std::move(*DebugBinary);
1602 }
1603 
1604 static void
1605 disassembleObject(ObjectFile &Obj, const ObjectFile &DbgObj,
1606                   DisassemblerTarget &PrimaryTarget,
1607                   std::optional<DisassemblerTarget> &SecondaryTarget,
1608                   SourcePrinter &SP, bool InlineRelocs) {
1609   DisassemblerTarget *DT = &PrimaryTarget;
1610   bool PrimaryIsThumb = false;
1611   SmallVector<std::pair<uint64_t, uint64_t>, 0> CHPECodeMap;
1612 
1613   if (SecondaryTarget) {
1614     if (isArmElf(Obj)) {
1615       PrimaryIsThumb =
1616           PrimaryTarget.SubtargetInfo->checkFeatures("+thumb-mode");
1617     } else if (const auto *COFFObj = dyn_cast<COFFObjectFile>(&Obj)) {
1618       const chpe_metadata *CHPEMetadata = COFFObj->getCHPEMetadata();
1619       if (CHPEMetadata && CHPEMetadata->CodeMapCount) {
1620         uintptr_t CodeMapInt;
1621         cantFail(COFFObj->getRvaPtr(CHPEMetadata->CodeMap, CodeMapInt));
1622         auto CodeMap = reinterpret_cast<const chpe_range_entry *>(CodeMapInt);
1623 
1624         for (uint32_t i = 0; i < CHPEMetadata->CodeMapCount; ++i) {
1625           if (CodeMap[i].getType() == chpe_range_type::Amd64 &&
1626               CodeMap[i].Length) {
1627             // Store x86_64 CHPE code ranges.
1628             uint64_t Start = CodeMap[i].getStart() + COFFObj->getImageBase();
1629             CHPECodeMap.emplace_back(Start, Start + CodeMap[i].Length);
1630           }
1631         }
1632         llvm::sort(CHPECodeMap);
1633       }
1634     }
1635   }
1636 
1637   std::map<SectionRef, std::vector<RelocationRef>> RelocMap;
1638   if (InlineRelocs || Obj.isXCOFF())
1639     RelocMap = getRelocsMap(Obj);
1640   bool Is64Bits = Obj.getBytesInAddress() > 4;
1641 
1642   // Create a mapping from virtual address to symbol name.  This is used to
1643   // pretty print the symbols while disassembling.
1644   std::map<SectionRef, SectionSymbolsTy> AllSymbols;
1645   std::map<SectionRef, SmallVector<MappingSymbolPair, 0>> AllMappingSymbols;
1646   SectionSymbolsTy AbsoluteSymbols;
1647   const StringRef FileName = Obj.getFileName();
1648   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(&Obj);
1649   for (const SymbolRef &Symbol : Obj.symbols()) {
1650     Expected<StringRef> NameOrErr = Symbol.getName();
1651     if (!NameOrErr) {
1652       reportWarning(toString(NameOrErr.takeError()), FileName);
1653       continue;
1654     }
1655     if (NameOrErr->empty() && !(Obj.isXCOFF() && SymbolDescription))
1656       continue;
1657 
1658     if (Obj.isELF() &&
1659         (cantFail(Symbol.getFlags()) & SymbolRef::SF_FormatSpecific)) {
1660       // Symbol is intended not to be displayed by default (STT_FILE,
1661       // STT_SECTION, or a mapping symbol). Ignore STT_SECTION symbols. We will
1662       // synthesize a section symbol if no symbol is defined at offset 0.
1663       //
1664       // For a mapping symbol, store it within both AllSymbols and
1665       // AllMappingSymbols. If --show-all-symbols is unspecified, its label will
1666       // not be printed in disassembly listing.
1667       if (getElfSymbolType(Obj, Symbol) != ELF::STT_SECTION &&
1668           hasMappingSymbols(Obj)) {
1669         section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1670         if (SecI != Obj.section_end()) {
1671           uint64_t SectionAddr = SecI->getAddress();
1672           uint64_t Address = cantFail(Symbol.getAddress());
1673           StringRef Name = *NameOrErr;
1674           if (Name.consume_front("$") && Name.size() &&
1675               strchr("adtx", Name[0])) {
1676             AllMappingSymbols[*SecI].emplace_back(Address - SectionAddr,
1677                                                   Name[0]);
1678             AllSymbols[*SecI].push_back(
1679                 createSymbolInfo(Obj, Symbol, /*MappingSymbol=*/true));
1680           }
1681         }
1682       }
1683       continue;
1684     }
1685 
1686     if (MachO) {
1687       // __mh_(execute|dylib|dylinker|bundle|preload|object)_header are special
1688       // symbols that support MachO header introspection. They do not bind to
1689       // code locations and are irrelevant for disassembly.
1690       if (NameOrErr->starts_with("__mh_") && NameOrErr->ends_with("_header"))
1691         continue;
1692       // Don't ask a Mach-O STAB symbol for its section unless you know that
1693       // STAB symbol's section field refers to a valid section index. Otherwise
1694       // the symbol may error trying to load a section that does not exist.
1695       DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1696       uint8_t NType = (MachO->is64Bit() ?
1697                        MachO->getSymbol64TableEntry(SymDRI).n_type:
1698                        MachO->getSymbolTableEntry(SymDRI).n_type);
1699       if (NType & MachO::N_STAB)
1700         continue;
1701     }
1702 
1703     section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1704     if (SecI != Obj.section_end())
1705       AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol));
1706     else
1707       AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol));
1708   }
1709 
1710   if (AllSymbols.empty() && Obj.isELF())
1711     addDynamicElfSymbols(cast<ELFObjectFileBase>(Obj), AllSymbols);
1712 
1713   if (Obj.isWasm())
1714     addMissingWasmCodeSymbols(cast<WasmObjectFile>(Obj), AllSymbols);
1715 
1716   if (Obj.isELF() && Obj.sections().empty())
1717     createFakeELFSections(Obj);
1718 
1719   BumpPtrAllocator A;
1720   StringSaver Saver(A);
1721   addPltEntries(Obj, AllSymbols, Saver);
1722 
1723   // Create a mapping from virtual address to section. An empty section can
1724   // cause more than one section at the same address. Sort such sections to be
1725   // before same-addressed non-empty sections so that symbol lookups prefer the
1726   // non-empty section.
1727   std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses;
1728   for (SectionRef Sec : Obj.sections())
1729     SectionAddresses.emplace_back(Sec.getAddress(), Sec);
1730   llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) {
1731     if (LHS.first != RHS.first)
1732       return LHS.first < RHS.first;
1733     return LHS.second.getSize() < RHS.second.getSize();
1734   });
1735 
1736   // Linked executables (.exe and .dll files) typically don't include a real
1737   // symbol table but they might contain an export table.
1738   if (const auto *COFFObj = dyn_cast<COFFObjectFile>(&Obj)) {
1739     for (const auto &ExportEntry : COFFObj->export_directories()) {
1740       StringRef Name;
1741       if (Error E = ExportEntry.getSymbolName(Name))
1742         reportError(std::move(E), Obj.getFileName());
1743       if (Name.empty())
1744         continue;
1745 
1746       uint32_t RVA;
1747       if (Error E = ExportEntry.getExportRVA(RVA))
1748         reportError(std::move(E), Obj.getFileName());
1749 
1750       uint64_t VA = COFFObj->getImageBase() + RVA;
1751       auto Sec = partition_point(
1752           SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) {
1753             return O.first <= VA;
1754           });
1755       if (Sec != SectionAddresses.begin()) {
1756         --Sec;
1757         AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE);
1758       } else
1759         AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE);
1760     }
1761   }
1762 
1763   // Sort all the symbols, this allows us to use a simple binary search to find
1764   // Multiple symbols can have the same address. Use a stable sort to stabilize
1765   // the output.
1766   StringSet<> FoundDisasmSymbolSet;
1767   for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols)
1768     llvm::stable_sort(SecSyms.second);
1769   llvm::stable_sort(AbsoluteSymbols);
1770 
1771   std::unique_ptr<DWARFContext> DICtx;
1772   LiveVariablePrinter LVP(*DT->Context->getRegisterInfo(), *DT->SubtargetInfo);
1773 
1774   if (DbgVariables != DVDisabled) {
1775     DICtx = DWARFContext::create(DbgObj);
1776     for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units())
1777       LVP.addCompileUnit(CU->getUnitDIE(false));
1778   }
1779 
1780   LLVM_DEBUG(LVP.dump());
1781 
1782   BBAddrMapInfo FullAddrMap;
1783   auto ReadBBAddrMap = [&](std::optional<unsigned> SectionIndex =
1784                                std::nullopt) {
1785     FullAddrMap.clear();
1786     if (const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj)) {
1787       std::vector<PGOAnalysisMap> PGOAnalyses;
1788       auto BBAddrMapsOrErr = Elf->readBBAddrMap(SectionIndex, &PGOAnalyses);
1789       if (!BBAddrMapsOrErr) {
1790         reportWarning(toString(BBAddrMapsOrErr.takeError()), Obj.getFileName());
1791         return;
1792       }
1793       for (auto &&[FunctionBBAddrMap, FunctionPGOAnalysis] :
1794            zip_equal(*std::move(BBAddrMapsOrErr), std::move(PGOAnalyses))) {
1795         FullAddrMap.AddFunctionEntry(std::move(FunctionBBAddrMap),
1796                                      std::move(FunctionPGOAnalysis));
1797       }
1798     }
1799   };
1800 
1801   // For non-relocatable objects, Read all LLVM_BB_ADDR_MAP sections into a
1802   // single mapping, since they don't have any conflicts.
1803   if (SymbolizeOperands && !Obj.isRelocatableObject())
1804     ReadBBAddrMap();
1805 
1806   std::optional<llvm::BTFParser> BTF;
1807   if (InlineRelocs && BTFParser::hasBTFSections(Obj)) {
1808     BTF.emplace();
1809     BTFParser::ParseOptions Opts = {};
1810     Opts.LoadTypes = true;
1811     Opts.LoadRelocs = true;
1812     if (Error E = BTF->parse(Obj, Opts))
1813       WithColor::defaultErrorHandler(std::move(E));
1814   }
1815 
1816   for (const SectionRef &Section : ToolSectionFilter(Obj)) {
1817     if (FilterSections.empty() && !DisassembleAll &&
1818         (!Section.isText() || Section.isVirtual()))
1819       continue;
1820 
1821     uint64_t SectionAddr = Section.getAddress();
1822     uint64_t SectSize = Section.getSize();
1823     if (!SectSize)
1824       continue;
1825 
1826     // For relocatable object files, read the LLVM_BB_ADDR_MAP section
1827     // corresponding to this section, if present.
1828     if (SymbolizeOperands && Obj.isRelocatableObject())
1829       ReadBBAddrMap(Section.getIndex());
1830 
1831     // Get the list of all the symbols in this section.
1832     SectionSymbolsTy &Symbols = AllSymbols[Section];
1833     auto &MappingSymbols = AllMappingSymbols[Section];
1834     llvm::sort(MappingSymbols);
1835 
1836     ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(
1837         unwrapOrError(Section.getContents(), Obj.getFileName()));
1838 
1839     std::vector<std::unique_ptr<std::string>> SynthesizedLabelNames;
1840     if (Obj.isELF() && Obj.getArch() == Triple::amdgcn) {
1841       // AMDGPU disassembler uses symbolizer for printing labels
1842       addSymbolizer(*DT->Context, DT->TheTarget, TripleName, DT->DisAsm.get(),
1843                     SectionAddr, Bytes, Symbols, SynthesizedLabelNames);
1844     }
1845 
1846     StringRef SegmentName = getSegmentName(MachO, Section);
1847     StringRef SectionName = unwrapOrError(Section.getName(), Obj.getFileName());
1848     // If the section has no symbol at the start, just insert a dummy one.
1849     // Without --show-all-symbols, also insert one if all symbols at the start
1850     // are mapping symbols.
1851     bool CreateDummy = Symbols.empty();
1852     if (!CreateDummy) {
1853       CreateDummy = true;
1854       for (auto &Sym : Symbols) {
1855         if (Sym.Addr != SectionAddr)
1856           break;
1857         if (!Sym.IsMappingSymbol || ShowAllSymbols)
1858           CreateDummy = false;
1859       }
1860     }
1861     if (CreateDummy) {
1862       SymbolInfoTy Sym = createDummySymbolInfo(
1863           Obj, SectionAddr, SectionName,
1864           Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT);
1865       if (Obj.isXCOFF())
1866         Symbols.insert(Symbols.begin(), Sym);
1867       else
1868         Symbols.insert(llvm::lower_bound(Symbols, Sym), Sym);
1869     }
1870 
1871     SmallString<40> Comments;
1872     raw_svector_ostream CommentStream(Comments);
1873 
1874     uint64_t VMAAdjustment = 0;
1875     if (shouldAdjustVA(Section))
1876       VMAAdjustment = AdjustVMA;
1877 
1878     // In executable and shared objects, r_offset holds a virtual address.
1879     // Subtract SectionAddr from the r_offset field of a relocation to get
1880     // the section offset.
1881     uint64_t RelAdjustment = Obj.isRelocatableObject() ? 0 : SectionAddr;
1882     uint64_t Size;
1883     uint64_t Index;
1884     bool PrintedSection = false;
1885     std::vector<RelocationRef> Rels = RelocMap[Section];
1886     std::vector<RelocationRef>::const_iterator RelCur = Rels.begin();
1887     std::vector<RelocationRef>::const_iterator RelEnd = Rels.end();
1888 
1889     // Loop over each chunk of code between two points where at least
1890     // one symbol is defined.
1891     for (size_t SI = 0, SE = Symbols.size(); SI != SE;) {
1892       // Advance SI past all the symbols starting at the same address,
1893       // and make an ArrayRef of them.
1894       unsigned FirstSI = SI;
1895       uint64_t Start = Symbols[SI].Addr;
1896       ArrayRef<SymbolInfoTy> SymbolsHere;
1897       while (SI != SE && Symbols[SI].Addr == Start)
1898         ++SI;
1899       SymbolsHere = ArrayRef<SymbolInfoTy>(&Symbols[FirstSI], SI - FirstSI);
1900 
1901       // Get the demangled names of all those symbols. We end up with a vector
1902       // of StringRef that holds the names we're going to use, and a vector of
1903       // std::string that stores the new strings returned by demangle(), if
1904       // any. If we don't call demangle() then that vector can stay empty.
1905       std::vector<StringRef> SymNamesHere;
1906       std::vector<std::string> DemangledSymNamesHere;
1907       if (Demangle) {
1908         // Fetch the demangled names and store them locally.
1909         for (const SymbolInfoTy &Symbol : SymbolsHere)
1910           DemangledSymNamesHere.push_back(demangle(Symbol.Name));
1911         // Now we've finished modifying that vector, it's safe to make
1912         // a vector of StringRefs pointing into it.
1913         SymNamesHere.insert(SymNamesHere.begin(), DemangledSymNamesHere.begin(),
1914                             DemangledSymNamesHere.end());
1915       } else {
1916         for (const SymbolInfoTy &Symbol : SymbolsHere)
1917           SymNamesHere.push_back(Symbol.Name);
1918       }
1919 
1920       // Distinguish ELF data from code symbols, which will be used later on to
1921       // decide whether to 'disassemble' this chunk as a data declaration via
1922       // dumpELFData(), or whether to treat it as code.
1923       //
1924       // If data _and_ code symbols are defined at the same address, the code
1925       // takes priority, on the grounds that disassembling code is our main
1926       // purpose here, and it would be a worse failure to _not_ interpret
1927       // something that _was_ meaningful as code than vice versa.
1928       //
1929       // Any ELF symbol type that is not clearly data will be regarded as code.
1930       // In particular, one of the uses of STT_NOTYPE is for branch targets
1931       // inside functions, for which STT_FUNC would be inaccurate.
1932       //
1933       // So here, we spot whether there's any non-data symbol present at all,
1934       // and only set the DisassembleAsELFData flag if there isn't. Also, we use
1935       // this distinction to inform the decision of which symbol to print at
1936       // the head of the section, so that if we're printing code, we print a
1937       // code-related symbol name to go with it.
1938       bool DisassembleAsELFData = false;
1939       size_t DisplaySymIndex = SymbolsHere.size() - 1;
1940       if (Obj.isELF() && !DisassembleAll && Section.isText()) {
1941         DisassembleAsELFData = true; // unless we find a code symbol below
1942 
1943         for (size_t i = 0; i < SymbolsHere.size(); ++i) {
1944           uint8_t SymTy = SymbolsHere[i].Type;
1945           if (SymTy != ELF::STT_OBJECT && SymTy != ELF::STT_COMMON) {
1946             DisassembleAsELFData = false;
1947             DisplaySymIndex = i;
1948           }
1949         }
1950       }
1951 
1952       // Decide which symbol(s) from this collection we're going to print.
1953       std::vector<bool> SymsToPrint(SymbolsHere.size(), false);
1954       // If the user has given the --disassemble-symbols option, then we must
1955       // display every symbol in that set, and no others.
1956       if (!DisasmSymbolSet.empty()) {
1957         bool FoundAny = false;
1958         for (size_t i = 0; i < SymbolsHere.size(); ++i) {
1959           if (DisasmSymbolSet.count(SymNamesHere[i])) {
1960             SymsToPrint[i] = true;
1961             FoundAny = true;
1962           }
1963         }
1964 
1965         // And if none of the symbols here is one that the user asked for, skip
1966         // disassembling this entire chunk of code.
1967         if (!FoundAny)
1968           continue;
1969       } else if (!SymbolsHere[DisplaySymIndex].IsMappingSymbol) {
1970         // Otherwise, print whichever symbol at this location is last in the
1971         // Symbols array, because that array is pre-sorted in a way intended to
1972         // correlate with priority of which symbol to display.
1973         SymsToPrint[DisplaySymIndex] = true;
1974       }
1975 
1976       // Now that we know we're disassembling this section, override the choice
1977       // of which symbols to display by printing _all_ of them at this address
1978       // if the user asked for all symbols.
1979       //
1980       // That way, '--show-all-symbols --disassemble-symbol=foo' will print
1981       // only the chunk of code headed by 'foo', but also show any other
1982       // symbols defined at that address, such as aliases for 'foo', or the ARM
1983       // mapping symbol preceding its code.
1984       if (ShowAllSymbols) {
1985         for (size_t i = 0; i < SymbolsHere.size(); ++i)
1986           SymsToPrint[i] = true;
1987       }
1988 
1989       if (Start < SectionAddr || StopAddress <= Start)
1990         continue;
1991 
1992       for (size_t i = 0; i < SymbolsHere.size(); ++i)
1993         FoundDisasmSymbolSet.insert(SymNamesHere[i]);
1994 
1995       // The end is the section end, the beginning of the next symbol, or
1996       // --stop-address.
1997       uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress);
1998       if (SI < SE)
1999         End = std::min(End, Symbols[SI].Addr);
2000       if (Start >= End || End <= StartAddress)
2001         continue;
2002       Start -= SectionAddr;
2003       End -= SectionAddr;
2004 
2005       if (!PrintedSection) {
2006         PrintedSection = true;
2007         outs() << "\nDisassembly of section ";
2008         if (!SegmentName.empty())
2009           outs() << SegmentName << ",";
2010         outs() << SectionName << ":\n";
2011       }
2012 
2013       bool PrintedLabel = false;
2014       for (size_t i = 0; i < SymbolsHere.size(); ++i) {
2015         if (!SymsToPrint[i])
2016           continue;
2017 
2018         const SymbolInfoTy &Symbol = SymbolsHere[i];
2019         const StringRef SymbolName = SymNamesHere[i];
2020 
2021         if (!PrintedLabel) {
2022           outs() << '\n';
2023           PrintedLabel = true;
2024         }
2025         if (LeadingAddr)
2026           outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ",
2027                            SectionAddr + Start + VMAAdjustment);
2028         if (Obj.isXCOFF() && SymbolDescription) {
2029           outs() << getXCOFFSymbolDescription(Symbol, SymbolName) << ":\n";
2030         } else
2031           outs() << '<' << SymbolName << ">:\n";
2032       }
2033 
2034       // Don't print raw contents of a virtual section. A virtual section
2035       // doesn't have any contents in the file.
2036       if (Section.isVirtual()) {
2037         outs() << "...\n";
2038         continue;
2039       }
2040 
2041       // See if any of the symbols defined at this location triggers target-
2042       // specific disassembly behavior, e.g. of special descriptors or function
2043       // prelude information.
2044       //
2045       // We stop this loop at the first symbol that triggers some kind of
2046       // interesting behavior (if any), on the assumption that if two symbols
2047       // defined at the same address trigger two conflicting symbol handlers,
2048       // the object file is probably confused anyway, and it would make even
2049       // less sense to present the output of _both_ handlers, because that
2050       // would describe the same data twice.
2051       for (size_t SHI = 0; SHI < SymbolsHere.size(); ++SHI) {
2052         SymbolInfoTy Symbol = SymbolsHere[SHI];
2053 
2054         auto Status = DT->DisAsm->onSymbolStart(
2055             Symbol, Size, Bytes.slice(Start, End - Start), SectionAddr + Start,
2056             CommentStream);
2057 
2058         if (!Status) {
2059           // If onSymbolStart returns std::nullopt, that means it didn't trigger
2060           // any interesting handling for this symbol. Try the other symbols
2061           // defined at this address.
2062           continue;
2063         }
2064 
2065         if (*Status == MCDisassembler::Fail) {
2066           // If onSymbolStart returns Fail, that means it identified some kind
2067           // of special data at this address, but wasn't able to disassemble it
2068           // meaningfully. So we fall back to disassembling the failed region
2069           // as bytes, assuming that the target detected the failure before
2070           // printing anything.
2071           //
2072           // Return values Success or SoftFail (i.e no 'real' failure) are
2073           // expected to mean that the target has emitted its own output.
2074           //
2075           // Either way, 'Size' will have been set to the amount of data
2076           // covered by whatever prologue the target identified. So we advance
2077           // our own position to beyond that. Sometimes that will be the entire
2078           // distance to the next symbol, and sometimes it will be just a
2079           // prologue and we should start disassembling instructions from where
2080           // it left off.
2081           outs() << DT->Context->getAsmInfo()->getCommentString()
2082                  << " error in decoding " << SymNamesHere[SHI]
2083                  << " : decoding failed region as bytes.\n";
2084           for (uint64_t I = 0; I < Size; ++I) {
2085             outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true)
2086                    << "\n";
2087           }
2088         }
2089         Start += Size;
2090         break;
2091       }
2092 
2093       Index = Start;
2094       if (SectionAddr < StartAddress)
2095         Index = std::max<uint64_t>(Index, StartAddress - SectionAddr);
2096 
2097       if (DisassembleAsELFData) {
2098         dumpELFData(SectionAddr, Index, End, Bytes);
2099         Index = End;
2100         continue;
2101       }
2102 
2103       // Skip relocations from symbols that are not dumped.
2104       for (; RelCur != RelEnd; ++RelCur) {
2105         uint64_t Offset = RelCur->getOffset() - RelAdjustment;
2106         if (Index <= Offset)
2107           break;
2108       }
2109 
2110       bool DumpARMELFData = false;
2111       bool DumpTracebackTableForXCOFFFunction =
2112           Obj.isXCOFF() && Section.isText() && TracebackTable &&
2113           Symbols[SI - 1].XCOFFSymInfo.StorageMappingClass &&
2114           (*Symbols[SI - 1].XCOFFSymInfo.StorageMappingClass == XCOFF::XMC_PR);
2115 
2116       formatted_raw_ostream FOS(outs());
2117 
2118       // FIXME: Workaround for bug in formatted_raw_ostream. Color escape codes
2119       // are (incorrectly) written directly to the unbuffered raw_ostream
2120       // wrapped by the formatted_raw_ostream.
2121       if (DisassemblyColor == ColorOutput::Enable ||
2122           DisassemblyColor == ColorOutput::Auto)
2123         FOS.SetUnbuffered();
2124 
2125       std::unordered_map<uint64_t, std::string> AllLabels;
2126       std::unordered_map<uint64_t, std::vector<BBAddrMapLabel>> BBAddrMapLabels;
2127       if (SymbolizeOperands) {
2128         collectLocalBranchTargets(Bytes, DT->InstrAnalysis.get(),
2129                                   DT->DisAsm.get(), DT->InstPrinter.get(),
2130                                   PrimaryTarget.SubtargetInfo.get(),
2131                                   SectionAddr, Index, End, AllLabels);
2132         collectBBAddrMapLabels(FullAddrMap, SectionAddr, Index, End,
2133                                BBAddrMapLabels);
2134       }
2135 
2136       if (DT->InstrAnalysis)
2137         DT->InstrAnalysis->resetState();
2138 
2139       while (Index < End) {
2140         uint64_t RelOffset;
2141 
2142         // ARM and AArch64 ELF binaries can interleave data and text in the
2143         // same section. We rely on the markers introduced to understand what
2144         // we need to dump. If the data marker is within a function, it is
2145         // denoted as a word/short etc.
2146         if (!MappingSymbols.empty()) {
2147           char Kind = getMappingSymbolKind(MappingSymbols, Index);
2148           DumpARMELFData = Kind == 'd';
2149           if (SecondaryTarget) {
2150             if (Kind == 'a') {
2151               DT = PrimaryIsThumb ? &*SecondaryTarget : &PrimaryTarget;
2152             } else if (Kind == 't') {
2153               DT = PrimaryIsThumb ? &PrimaryTarget : &*SecondaryTarget;
2154             }
2155           }
2156         } else if (!CHPECodeMap.empty()) {
2157           uint64_t Address = SectionAddr + Index;
2158           auto It = partition_point(
2159               CHPECodeMap,
2160               [Address](const std::pair<uint64_t, uint64_t> &Entry) {
2161                 return Entry.first <= Address;
2162               });
2163           if (It != CHPECodeMap.begin() && Address < (It - 1)->second) {
2164             DT = &*SecondaryTarget;
2165           } else {
2166             DT = &PrimaryTarget;
2167             // X64 disassembler range may have left Index unaligned, so
2168             // make sure that it's aligned when we switch back to ARM64
2169             // code.
2170             Index = llvm::alignTo(Index, 4);
2171             if (Index >= End)
2172               break;
2173           }
2174         }
2175 
2176         auto findRel = [&]() {
2177           while (RelCur != RelEnd) {
2178             RelOffset = RelCur->getOffset() - RelAdjustment;
2179             // If this relocation is hidden, skip it.
2180             if (getHidden(*RelCur) || SectionAddr + RelOffset < StartAddress) {
2181               ++RelCur;
2182               continue;
2183             }
2184 
2185             // Stop when RelCur's offset is past the disassembled
2186             // instruction/data.
2187             if (RelOffset >= Index + Size)
2188               return false;
2189             if (RelOffset >= Index)
2190               return true;
2191             ++RelCur;
2192           }
2193           return false;
2194         };
2195 
2196         if (DumpARMELFData) {
2197           Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes,
2198                                 MappingSymbols, *DT->SubtargetInfo, FOS);
2199         } else {
2200           // When -z or --disassemble-zeroes are given we always dissasemble
2201           // them. Otherwise we might want to skip zero bytes we see.
2202           if (!DisassembleZeroes) {
2203             uint64_t MaxOffset = End - Index;
2204             // For --reloc: print zero blocks patched by relocations, so that
2205             // relocations can be shown in the dump.
2206             if (InlineRelocs && RelCur != RelEnd)
2207               MaxOffset = std::min(RelCur->getOffset() - RelAdjustment - Index,
2208                                    MaxOffset);
2209 
2210             if (size_t N =
2211                     countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) {
2212               FOS << "\t\t..." << '\n';
2213               Index += N;
2214               continue;
2215             }
2216           }
2217 
2218           if (DumpTracebackTableForXCOFFFunction &&
2219               doesXCOFFTracebackTableBegin(Bytes.slice(Index, 4))) {
2220             dumpTracebackTable(Bytes.slice(Index),
2221                                SectionAddr + Index + VMAAdjustment, FOS,
2222                                SectionAddr + End + VMAAdjustment,
2223                                *DT->SubtargetInfo, cast<XCOFFObjectFile>(&Obj));
2224             Index = End;
2225             continue;
2226           }
2227 
2228           // Print local label if there's any.
2229           auto Iter1 = BBAddrMapLabels.find(SectionAddr + Index);
2230           if (Iter1 != BBAddrMapLabels.end()) {
2231             for (const auto &BBLabel : Iter1->second)
2232               FOS << "<" << BBLabel.BlockLabel << ">" << BBLabel.PGOAnalysis
2233                   << ":\n";
2234           } else {
2235             auto Iter2 = AllLabels.find(SectionAddr + Index);
2236             if (Iter2 != AllLabels.end())
2237               FOS << "<" << Iter2->second << ">:\n";
2238           }
2239 
2240           // Disassemble a real instruction or a data when disassemble all is
2241           // provided
2242           MCInst Inst;
2243           ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index);
2244           uint64_t ThisAddr = SectionAddr + Index;
2245           bool Disassembled = DT->DisAsm->getInstruction(
2246               Inst, Size, ThisBytes, ThisAddr, CommentStream);
2247           if (Size == 0)
2248             Size = std::min<uint64_t>(
2249                 ThisBytes.size(),
2250                 DT->DisAsm->suggestBytesToSkip(ThisBytes, ThisAddr));
2251 
2252           LVP.update({Index, Section.getIndex()},
2253                      {Index + Size, Section.getIndex()}, Index + Size != End);
2254 
2255           DT->InstPrinter->setCommentStream(CommentStream);
2256 
2257           DT->Printer->printInst(
2258               *DT->InstPrinter, Disassembled ? &Inst : nullptr,
2259               Bytes.slice(Index, Size),
2260               {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS,
2261               "", *DT->SubtargetInfo, &SP, Obj.getFileName(), &Rels, LVP);
2262 
2263           DT->InstPrinter->setCommentStream(llvm::nulls());
2264 
2265           // If disassembly succeeds, we try to resolve the target address
2266           // (jump target or memory operand address) and print it to the
2267           // right of the instruction.
2268           //
2269           // Otherwise, we don't print anything else so that we avoid
2270           // analyzing invalid or incomplete instruction information.
2271           if (Disassembled && DT->InstrAnalysis) {
2272             llvm::raw_ostream *TargetOS = &FOS;
2273             uint64_t Target;
2274             bool PrintTarget = DT->InstrAnalysis->evaluateBranch(
2275                 Inst, SectionAddr + Index, Size, Target);
2276 
2277             if (!PrintTarget) {
2278               if (std::optional<uint64_t> MaybeTarget =
2279                       DT->InstrAnalysis->evaluateMemoryOperandAddress(
2280                           Inst, DT->SubtargetInfo.get(), SectionAddr + Index,
2281                           Size)) {
2282                 Target = *MaybeTarget;
2283                 PrintTarget = true;
2284                 // Do not print real address when symbolizing.
2285                 if (!SymbolizeOperands) {
2286                   // Memory operand addresses are printed as comments.
2287                   TargetOS = &CommentStream;
2288                   *TargetOS << "0x" << Twine::utohexstr(Target);
2289                 }
2290               }
2291             }
2292 
2293             if (PrintTarget) {
2294               // In a relocatable object, the target's section must reside in
2295               // the same section as the call instruction or it is accessed
2296               // through a relocation.
2297               //
2298               // In a non-relocatable object, the target may be in any section.
2299               // In that case, locate the section(s) containing the target
2300               // address and find the symbol in one of those, if possible.
2301               //
2302               // N.B. Except for XCOFF, we don't walk the relocations in the
2303               // relocatable case yet.
2304               std::vector<const SectionSymbolsTy *> TargetSectionSymbols;
2305               if (!Obj.isRelocatableObject()) {
2306                 auto It = llvm::partition_point(
2307                     SectionAddresses,
2308                     [=](const std::pair<uint64_t, SectionRef> &O) {
2309                       return O.first <= Target;
2310                     });
2311                 uint64_t TargetSecAddr = 0;
2312                 while (It != SectionAddresses.begin()) {
2313                   --It;
2314                   if (TargetSecAddr == 0)
2315                     TargetSecAddr = It->first;
2316                   if (It->first != TargetSecAddr)
2317                     break;
2318                   TargetSectionSymbols.push_back(&AllSymbols[It->second]);
2319                 }
2320               } else {
2321                 TargetSectionSymbols.push_back(&Symbols);
2322               }
2323               TargetSectionSymbols.push_back(&AbsoluteSymbols);
2324 
2325               // Find the last symbol in the first candidate section whose
2326               // offset is less than or equal to the target. If there are no
2327               // such symbols, try in the next section and so on, before finally
2328               // using the nearest preceding absolute symbol (if any), if there
2329               // are no other valid symbols.
2330               const SymbolInfoTy *TargetSym = nullptr;
2331               for (const SectionSymbolsTy *TargetSymbols :
2332                    TargetSectionSymbols) {
2333                 auto It = llvm::partition_point(
2334                     *TargetSymbols,
2335                     [=](const SymbolInfoTy &O) { return O.Addr <= Target; });
2336                 while (It != TargetSymbols->begin()) {
2337                   --It;
2338                   // Skip mapping symbols to avoid possible ambiguity as they
2339                   // do not allow uniquely identifying the target address.
2340                   if (!It->IsMappingSymbol) {
2341                     TargetSym = &*It;
2342                     break;
2343                   }
2344                 }
2345                 if (TargetSym)
2346                   break;
2347               }
2348 
2349               // Branch targets are printed just after the instructions.
2350               // Print the labels corresponding to the target if there's any.
2351               bool BBAddrMapLabelAvailable = BBAddrMapLabels.count(Target);
2352               bool LabelAvailable = AllLabels.count(Target);
2353 
2354               if (TargetSym != nullptr) {
2355                 uint64_t TargetAddress = TargetSym->Addr;
2356                 uint64_t Disp = Target - TargetAddress;
2357                 std::string TargetName = Demangle ? demangle(TargetSym->Name)
2358                                                   : TargetSym->Name.str();
2359                 bool RelFixedUp = false;
2360                 SmallString<32> Val;
2361 
2362                 *TargetOS << " <";
2363                 // On XCOFF, we use relocations, even without -r, so we
2364                 // can print the correct name for an extern function call.
2365                 if (Obj.isXCOFF() && findRel()) {
2366                   // Check for possible branch relocations and
2367                   // branches to fixup code.
2368                   bool BranchRelocationType = true;
2369                   XCOFF::RelocationType RelocType;
2370                   if (Obj.is64Bit()) {
2371                     const XCOFFRelocation64 *Reloc =
2372                         reinterpret_cast<XCOFFRelocation64 *>(
2373                             RelCur->getRawDataRefImpl().p);
2374                     RelFixedUp = Reloc->isFixupIndicated();
2375                     RelocType = Reloc->Type;
2376                   } else {
2377                     const XCOFFRelocation32 *Reloc =
2378                         reinterpret_cast<XCOFFRelocation32 *>(
2379                             RelCur->getRawDataRefImpl().p);
2380                     RelFixedUp = Reloc->isFixupIndicated();
2381                     RelocType = Reloc->Type;
2382                   }
2383                   BranchRelocationType =
2384                       RelocType == XCOFF::R_BA || RelocType == XCOFF::R_BR ||
2385                       RelocType == XCOFF::R_RBA || RelocType == XCOFF::R_RBR;
2386 
2387                   // If we have a valid relocation, try to print its
2388                   // corresponding symbol name. Multiple relocations on the
2389                   // same instruction are not handled.
2390                   // Branches to fixup code will have the RelFixedUp flag set in
2391                   // the RLD. For these instructions, we print the correct
2392                   // branch target, but print the referenced symbol as a
2393                   // comment.
2394                   if (Error E = getRelocationValueString(*RelCur, false, Val)) {
2395                     // If -r was used, this error will be printed later.
2396                     // Otherwise, we ignore the error and print what
2397                     // would have been printed without using relocations.
2398                     consumeError(std::move(E));
2399                     *TargetOS << TargetName;
2400                     RelFixedUp = false; // Suppress comment for RLD sym name
2401                   } else if (BranchRelocationType && !RelFixedUp)
2402                     *TargetOS << Val;
2403                   else
2404                     *TargetOS << TargetName;
2405                   if (Disp)
2406                     *TargetOS << "+0x" << Twine::utohexstr(Disp);
2407                 } else if (!Disp) {
2408                   *TargetOS << TargetName;
2409                 } else if (BBAddrMapLabelAvailable) {
2410                   *TargetOS << BBAddrMapLabels[Target].front().BlockLabel;
2411                 } else if (LabelAvailable) {
2412                   *TargetOS << AllLabels[Target];
2413                 } else {
2414                   // Always Print the binary symbol plus an offset if there's no
2415                   // local label corresponding to the target address.
2416                   *TargetOS << TargetName << "+0x" << Twine::utohexstr(Disp);
2417                 }
2418                 *TargetOS << ">";
2419                 if (RelFixedUp && !InlineRelocs) {
2420                   // We have fixup code for a relocation. We print the
2421                   // referenced symbol as a comment.
2422                   *TargetOS << "\t# " << Val;
2423                 }
2424 
2425               } else if (BBAddrMapLabelAvailable) {
2426                 *TargetOS << " <" << BBAddrMapLabels[Target].front().BlockLabel
2427                           << ">";
2428               } else if (LabelAvailable) {
2429                 *TargetOS << " <" << AllLabels[Target] << ">";
2430               }
2431               // By convention, each record in the comment stream should be
2432               // terminated.
2433               if (TargetOS == &CommentStream)
2434                 *TargetOS << "\n";
2435             }
2436 
2437             DT->InstrAnalysis->updateState(Inst, SectionAddr + Index);
2438           } else if (!Disassembled && DT->InstrAnalysis) {
2439             DT->InstrAnalysis->resetState();
2440           }
2441         }
2442 
2443         assert(DT->Context->getAsmInfo());
2444         emitPostInstructionInfo(FOS, *DT->Context->getAsmInfo(),
2445                                 *DT->SubtargetInfo, CommentStream.str(), LVP);
2446         Comments.clear();
2447 
2448         if (BTF)
2449           printBTFRelocation(FOS, *BTF, {Index, Section.getIndex()}, LVP);
2450 
2451         // Hexagon handles relocs in pretty printer
2452         if (InlineRelocs && Obj.getArch() != Triple::hexagon) {
2453           while (findRel()) {
2454             // When --adjust-vma is used, update the address printed.
2455             if (RelCur->getSymbol() != Obj.symbol_end()) {
2456               Expected<section_iterator> SymSI =
2457                   RelCur->getSymbol()->getSection();
2458               if (SymSI && *SymSI != Obj.section_end() &&
2459                   shouldAdjustVA(**SymSI))
2460                 RelOffset += AdjustVMA;
2461             }
2462 
2463             printRelocation(FOS, Obj.getFileName(), *RelCur,
2464                             SectionAddr + RelOffset, Is64Bits);
2465             LVP.printAfterOtherLine(FOS, true);
2466             ++RelCur;
2467           }
2468         }
2469 
2470         Index += Size;
2471       }
2472     }
2473   }
2474   StringSet<> MissingDisasmSymbolSet =
2475       set_difference(DisasmSymbolSet, FoundDisasmSymbolSet);
2476   for (StringRef Sym : MissingDisasmSymbolSet.keys())
2477     reportWarning("failed to disassemble missing symbol " + Sym, FileName);
2478 }
2479 
2480 static void disassembleObject(ObjectFile *Obj, bool InlineRelocs) {
2481   // If information useful for showing the disassembly is missing, try to find a
2482   // more complete binary and disassemble that instead.
2483   OwningBinary<Binary> FetchedBinary;
2484   if (Obj->symbols().empty()) {
2485     if (std::optional<OwningBinary<Binary>> FetchedBinaryOpt =
2486             fetchBinaryByBuildID(*Obj)) {
2487       if (auto *O = dyn_cast<ObjectFile>(FetchedBinaryOpt->getBinary())) {
2488         if (!O->symbols().empty() ||
2489             (!O->sections().empty() && Obj->sections().empty())) {
2490           FetchedBinary = std::move(*FetchedBinaryOpt);
2491           Obj = O;
2492         }
2493       }
2494     }
2495   }
2496 
2497   const Target *TheTarget = getTarget(Obj);
2498 
2499   // Package up features to be passed to target/subtarget
2500   Expected<SubtargetFeatures> FeaturesValue = Obj->getFeatures();
2501   if (!FeaturesValue)
2502     reportError(FeaturesValue.takeError(), Obj->getFileName());
2503   SubtargetFeatures Features = *FeaturesValue;
2504   if (!MAttrs.empty()) {
2505     for (unsigned I = 0; I != MAttrs.size(); ++I)
2506       Features.AddFeature(MAttrs[I]);
2507   } else if (MCPU.empty() && Obj->getArch() == llvm::Triple::aarch64) {
2508     Features.AddFeature("+all");
2509   }
2510 
2511   if (MCPU.empty())
2512     MCPU = Obj->tryGetCPUName().value_or("").str();
2513 
2514   if (isArmElf(*Obj)) {
2515     // When disassembling big-endian Arm ELF, the instruction endianness is
2516     // determined in a complex way. In relocatable objects, AAELF32 mandates
2517     // that instruction endianness matches the ELF file endianness; in
2518     // executable images, that's true unless the file header has the EF_ARM_BE8
2519     // flag, in which case instructions are little-endian regardless of data
2520     // endianness.
2521     //
2522     // We must set the big-endian-instructions SubtargetFeature to make the
2523     // disassembler read the instructions the right way round, and also tell
2524     // our own prettyprinter to retrieve the encodings the same way to print in
2525     // hex.
2526     const auto *Elf32BE = dyn_cast<ELF32BEObjectFile>(Obj);
2527 
2528     if (Elf32BE && (Elf32BE->isRelocatableObject() ||
2529                     !(Elf32BE->getPlatformFlags() & ELF::EF_ARM_BE8))) {
2530       Features.AddFeature("+big-endian-instructions");
2531       ARMPrettyPrinterInst.setInstructionEndianness(llvm::endianness::big);
2532     } else {
2533       ARMPrettyPrinterInst.setInstructionEndianness(llvm::endianness::little);
2534     }
2535   }
2536 
2537   DisassemblerTarget PrimaryTarget(TheTarget, *Obj, TripleName, MCPU, Features);
2538 
2539   // If we have an ARM object file, we need a second disassembler, because
2540   // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode.
2541   // We use mapping symbols to switch between the two assemblers, where
2542   // appropriate.
2543   std::optional<DisassemblerTarget> SecondaryTarget;
2544 
2545   if (isArmElf(*Obj)) {
2546     if (!PrimaryTarget.SubtargetInfo->checkFeatures("+mclass")) {
2547       if (PrimaryTarget.SubtargetInfo->checkFeatures("+thumb-mode"))
2548         Features.AddFeature("-thumb-mode");
2549       else
2550         Features.AddFeature("+thumb-mode");
2551       SecondaryTarget.emplace(PrimaryTarget, Features);
2552     }
2553   } else if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) {
2554     const chpe_metadata *CHPEMetadata = COFFObj->getCHPEMetadata();
2555     if (CHPEMetadata && CHPEMetadata->CodeMapCount) {
2556       // Set up x86_64 disassembler for ARM64EC binaries.
2557       Triple X64Triple(TripleName);
2558       X64Triple.setArch(Triple::ArchType::x86_64);
2559 
2560       std::string Error;
2561       const Target *X64Target =
2562           TargetRegistry::lookupTarget("", X64Triple, Error);
2563       if (X64Target) {
2564         SubtargetFeatures X64Features;
2565         SecondaryTarget.emplace(X64Target, *Obj, X64Triple.getTriple(), "",
2566                                 X64Features);
2567       } else {
2568         reportWarning(Error, Obj->getFileName());
2569       }
2570     }
2571   }
2572 
2573   const ObjectFile *DbgObj = Obj;
2574   if (!FetchedBinary.getBinary() && !Obj->hasDebugInfo()) {
2575     if (std::optional<OwningBinary<Binary>> DebugBinaryOpt =
2576             fetchBinaryByBuildID(*Obj)) {
2577       if (auto *FetchedObj =
2578               dyn_cast<const ObjectFile>(DebugBinaryOpt->getBinary())) {
2579         if (FetchedObj->hasDebugInfo()) {
2580           FetchedBinary = std::move(*DebugBinaryOpt);
2581           DbgObj = FetchedObj;
2582         }
2583       }
2584     }
2585   }
2586 
2587   std::unique_ptr<object::Binary> DSYMBinary;
2588   std::unique_ptr<MemoryBuffer> DSYMBuf;
2589   if (!DbgObj->hasDebugInfo()) {
2590     if (const MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*Obj)) {
2591       DbgObj = objdump::getMachODSymObject(MachOOF, Obj->getFileName(),
2592                                            DSYMBinary, DSYMBuf);
2593       if (!DbgObj)
2594         return;
2595     }
2596   }
2597 
2598   SourcePrinter SP(DbgObj, TheTarget->getName());
2599 
2600   for (StringRef Opt : DisassemblerOptions)
2601     if (!PrimaryTarget.InstPrinter->applyTargetSpecificCLOption(Opt))
2602       reportError(Obj->getFileName(),
2603                   "Unrecognized disassembler option: " + Opt);
2604 
2605   disassembleObject(*Obj, *DbgObj, PrimaryTarget, SecondaryTarget, SP,
2606                     InlineRelocs);
2607 }
2608 
2609 void Dumper::printRelocations() {
2610   StringRef Fmt = O.getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2611 
2612   // Build a mapping from relocation target to a vector of relocation
2613   // sections. Usually, there is an only one relocation section for
2614   // each relocated section.
2615   MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec;
2616   uint64_t Ndx;
2617   for (const SectionRef &Section : ToolSectionFilter(O, &Ndx)) {
2618     if (O.isELF() && (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC))
2619       continue;
2620     if (Section.relocation_begin() == Section.relocation_end())
2621       continue;
2622     Expected<section_iterator> SecOrErr = Section.getRelocatedSection();
2623     if (!SecOrErr)
2624       reportError(O.getFileName(),
2625                   "section (" + Twine(Ndx) +
2626                       "): unable to get a relocation target: " +
2627                       toString(SecOrErr.takeError()));
2628     SecToRelSec[**SecOrErr].push_back(Section);
2629   }
2630 
2631   for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) {
2632     StringRef SecName = unwrapOrError(P.first.getName(), O.getFileName());
2633     outs() << "\nRELOCATION RECORDS FOR [" << SecName << "]:\n";
2634     uint32_t OffsetPadding = (O.getBytesInAddress() > 4 ? 16 : 8);
2635     uint32_t TypePadding = 24;
2636     outs() << left_justify("OFFSET", OffsetPadding) << " "
2637            << left_justify("TYPE", TypePadding) << " "
2638            << "VALUE\n";
2639 
2640     for (SectionRef Section : P.second) {
2641       for (const RelocationRef &Reloc : Section.relocations()) {
2642         uint64_t Address = Reloc.getOffset();
2643         SmallString<32> RelocName;
2644         SmallString<32> ValueStr;
2645         if (Address < StartAddress || Address > StopAddress || getHidden(Reloc))
2646           continue;
2647         Reloc.getTypeName(RelocName);
2648         if (Error E =
2649                 getRelocationValueString(Reloc, SymbolDescription, ValueStr))
2650           reportUniqueWarning(std::move(E));
2651 
2652         outs() << format(Fmt.data(), Address) << " "
2653                << left_justify(RelocName, TypePadding) << " " << ValueStr
2654                << "\n";
2655       }
2656     }
2657   }
2658 }
2659 
2660 // Returns true if we need to show LMA column when dumping section headers. We
2661 // show it only when the platform is ELF and either we have at least one section
2662 // whose VMA and LMA are different and/or when --show-lma flag is used.
2663 static bool shouldDisplayLMA(const ObjectFile &Obj) {
2664   if (!Obj.isELF())
2665     return false;
2666   for (const SectionRef &S : ToolSectionFilter(Obj))
2667     if (S.getAddress() != getELFSectionLMA(S))
2668       return true;
2669   return ShowLMA;
2670 }
2671 
2672 static size_t getMaxSectionNameWidth(const ObjectFile &Obj) {
2673   // Default column width for names is 13 even if no names are that long.
2674   size_t MaxWidth = 13;
2675   for (const SectionRef &Section : ToolSectionFilter(Obj)) {
2676     StringRef Name = unwrapOrError(Section.getName(), Obj.getFileName());
2677     MaxWidth = std::max(MaxWidth, Name.size());
2678   }
2679   return MaxWidth;
2680 }
2681 
2682 void objdump::printSectionHeaders(ObjectFile &Obj) {
2683   if (Obj.isELF() && Obj.sections().empty())
2684     createFakeELFSections(Obj);
2685 
2686   size_t NameWidth = getMaxSectionNameWidth(Obj);
2687   size_t AddressWidth = 2 * Obj.getBytesInAddress();
2688   bool HasLMAColumn = shouldDisplayLMA(Obj);
2689   outs() << "\nSections:\n";
2690   if (HasLMAColumn)
2691     outs() << "Idx " << left_justify("Name", NameWidth) << " Size     "
2692            << left_justify("VMA", AddressWidth) << " "
2693            << left_justify("LMA", AddressWidth) << " Type\n";
2694   else
2695     outs() << "Idx " << left_justify("Name", NameWidth) << " Size     "
2696            << left_justify("VMA", AddressWidth) << " Type\n";
2697 
2698   uint64_t Idx;
2699   for (const SectionRef &Section : ToolSectionFilter(Obj, &Idx)) {
2700     StringRef Name = unwrapOrError(Section.getName(), Obj.getFileName());
2701     uint64_t VMA = Section.getAddress();
2702     if (shouldAdjustVA(Section))
2703       VMA += AdjustVMA;
2704 
2705     uint64_t Size = Section.getSize();
2706 
2707     std::string Type = Section.isText() ? "TEXT" : "";
2708     if (Section.isData())
2709       Type += Type.empty() ? "DATA" : ", DATA";
2710     if (Section.isBSS())
2711       Type += Type.empty() ? "BSS" : ", BSS";
2712     if (Section.isDebugSection())
2713       Type += Type.empty() ? "DEBUG" : ", DEBUG";
2714 
2715     if (HasLMAColumn)
2716       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
2717                        Name.str().c_str(), Size)
2718              << format_hex_no_prefix(VMA, AddressWidth) << " "
2719              << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth)
2720              << " " << Type << "\n";
2721     else
2722       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
2723                        Name.str().c_str(), Size)
2724              << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n";
2725   }
2726 }
2727 
2728 void objdump::printSectionContents(const ObjectFile *Obj) {
2729   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
2730 
2731   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
2732     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
2733     uint64_t BaseAddr = Section.getAddress();
2734     uint64_t Size = Section.getSize();
2735     if (!Size)
2736       continue;
2737 
2738     outs() << "Contents of section ";
2739     StringRef SegmentName = getSegmentName(MachO, Section);
2740     if (!SegmentName.empty())
2741       outs() << SegmentName << ",";
2742     outs() << Name << ":\n";
2743     if (Section.isBSS()) {
2744       outs() << format("<skipping contents of bss section at [%04" PRIx64
2745                        ", %04" PRIx64 ")>\n",
2746                        BaseAddr, BaseAddr + Size);
2747       continue;
2748     }
2749 
2750     StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName());
2751 
2752     // Dump out the content as hex and printable ascii characters.
2753     for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) {
2754       outs() << format(" %04" PRIx64 " ", BaseAddr + Addr);
2755       // Dump line of hex.
2756       for (std::size_t I = 0; I < 16; ++I) {
2757         if (I != 0 && I % 4 == 0)
2758           outs() << ' ';
2759         if (Addr + I < End)
2760           outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true)
2761                  << hexdigit(Contents[Addr + I] & 0xF, true);
2762         else
2763           outs() << "  ";
2764       }
2765       // Print ascii.
2766       outs() << "  ";
2767       for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) {
2768         if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF))
2769           outs() << Contents[Addr + I];
2770         else
2771           outs() << ".";
2772       }
2773       outs() << "\n";
2774     }
2775   }
2776 }
2777 
2778 void Dumper::printSymbolTable(StringRef ArchiveName, StringRef ArchitectureName,
2779                               bool DumpDynamic) {
2780   if (O.isCOFF() && !DumpDynamic) {
2781     outs() << "\nSYMBOL TABLE:\n";
2782     printCOFFSymbolTable(cast<const COFFObjectFile>(O));
2783     return;
2784   }
2785 
2786   const StringRef FileName = O.getFileName();
2787 
2788   if (!DumpDynamic) {
2789     outs() << "\nSYMBOL TABLE:\n";
2790     for (auto I = O.symbol_begin(); I != O.symbol_end(); ++I)
2791       printSymbol(*I, {}, FileName, ArchiveName, ArchitectureName, DumpDynamic);
2792     return;
2793   }
2794 
2795   outs() << "\nDYNAMIC SYMBOL TABLE:\n";
2796   if (!O.isELF()) {
2797     reportWarning(
2798         "this operation is not currently supported for this file format",
2799         FileName);
2800     return;
2801   }
2802 
2803   const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(&O);
2804   auto Symbols = ELF->getDynamicSymbolIterators();
2805   Expected<std::vector<VersionEntry>> SymbolVersionsOrErr =
2806       ELF->readDynsymVersions();
2807   if (!SymbolVersionsOrErr) {
2808     reportWarning(toString(SymbolVersionsOrErr.takeError()), FileName);
2809     SymbolVersionsOrErr = std::vector<VersionEntry>();
2810     (void)!SymbolVersionsOrErr;
2811   }
2812   for (auto &Sym : Symbols)
2813     printSymbol(Sym, *SymbolVersionsOrErr, FileName, ArchiveName,
2814                 ArchitectureName, DumpDynamic);
2815 }
2816 
2817 void Dumper::printSymbol(const SymbolRef &Symbol,
2818                          ArrayRef<VersionEntry> SymbolVersions,
2819                          StringRef FileName, StringRef ArchiveName,
2820                          StringRef ArchitectureName, bool DumpDynamic) {
2821   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(&O);
2822   Expected<uint64_t> AddrOrErr = Symbol.getAddress();
2823   if (!AddrOrErr) {
2824     reportUniqueWarning(AddrOrErr.takeError());
2825     return;
2826   }
2827   uint64_t Address = *AddrOrErr;
2828   section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
2829   if (SecI != O.section_end() && shouldAdjustVA(*SecI))
2830     Address += AdjustVMA;
2831   if ((Address < StartAddress) || (Address > StopAddress))
2832     return;
2833   SymbolRef::Type Type =
2834       unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName);
2835   uint32_t Flags =
2836       unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName);
2837 
2838   // Don't ask a Mach-O STAB symbol for its section unless you know that
2839   // STAB symbol's section field refers to a valid section index. Otherwise
2840   // the symbol may error trying to load a section that does not exist.
2841   bool IsSTAB = false;
2842   if (MachO) {
2843     DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
2844     uint8_t NType =
2845         (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type
2846                           : MachO->getSymbolTableEntry(SymDRI).n_type);
2847     if (NType & MachO::N_STAB)
2848       IsSTAB = true;
2849   }
2850   section_iterator Section = IsSTAB
2851                                  ? O.section_end()
2852                                  : unwrapOrError(Symbol.getSection(), FileName,
2853                                                  ArchiveName, ArchitectureName);
2854 
2855   StringRef Name;
2856   if (Type == SymbolRef::ST_Debug && Section != O.section_end()) {
2857     if (Expected<StringRef> NameOrErr = Section->getName())
2858       Name = *NameOrErr;
2859     else
2860       consumeError(NameOrErr.takeError());
2861 
2862   } else {
2863     Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName,
2864                          ArchitectureName);
2865   }
2866 
2867   bool Global = Flags & SymbolRef::SF_Global;
2868   bool Weak = Flags & SymbolRef::SF_Weak;
2869   bool Absolute = Flags & SymbolRef::SF_Absolute;
2870   bool Common = Flags & SymbolRef::SF_Common;
2871   bool Hidden = Flags & SymbolRef::SF_Hidden;
2872 
2873   char GlobLoc = ' ';
2874   if ((Section != O.section_end() || Absolute) && !Weak)
2875     GlobLoc = Global ? 'g' : 'l';
2876   char IFunc = ' ';
2877   if (O.isELF()) {
2878     if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC)
2879       IFunc = 'i';
2880     if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE)
2881       GlobLoc = 'u';
2882   }
2883 
2884   char Debug = ' ';
2885   if (DumpDynamic)
2886     Debug = 'D';
2887   else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
2888     Debug = 'd';
2889 
2890   char FileFunc = ' ';
2891   if (Type == SymbolRef::ST_File)
2892     FileFunc = 'f';
2893   else if (Type == SymbolRef::ST_Function)
2894     FileFunc = 'F';
2895   else if (Type == SymbolRef::ST_Data)
2896     FileFunc = 'O';
2897 
2898   const char *Fmt = O.getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2899 
2900   outs() << format(Fmt, Address) << " "
2901          << GlobLoc            // Local -> 'l', Global -> 'g', Neither -> ' '
2902          << (Weak ? 'w' : ' ') // Weak?
2903          << ' '                // Constructor. Not supported yet.
2904          << ' '                // Warning. Not supported yet.
2905          << IFunc              // Indirect reference to another symbol.
2906          << Debug              // Debugging (d) or dynamic (D) symbol.
2907          << FileFunc           // Name of function (F), file (f) or object (O).
2908          << ' ';
2909   if (Absolute) {
2910     outs() << "*ABS*";
2911   } else if (Common) {
2912     outs() << "*COM*";
2913   } else if (Section == O.section_end()) {
2914     if (O.isXCOFF()) {
2915       XCOFFSymbolRef XCOFFSym = cast<const XCOFFObjectFile>(O).toSymbolRef(
2916           Symbol.getRawDataRefImpl());
2917       if (XCOFF::N_DEBUG == XCOFFSym.getSectionNumber())
2918         outs() << "*DEBUG*";
2919       else
2920         outs() << "*UND*";
2921     } else
2922       outs() << "*UND*";
2923   } else {
2924     StringRef SegmentName = getSegmentName(MachO, *Section);
2925     if (!SegmentName.empty())
2926       outs() << SegmentName << ",";
2927     StringRef SectionName = unwrapOrError(Section->getName(), FileName);
2928     outs() << SectionName;
2929     if (O.isXCOFF()) {
2930       std::optional<SymbolRef> SymRef =
2931           getXCOFFSymbolContainingSymbolRef(cast<XCOFFObjectFile>(O), Symbol);
2932       if (SymRef) {
2933 
2934         Expected<StringRef> NameOrErr = SymRef->getName();
2935 
2936         if (NameOrErr) {
2937           outs() << " (csect:";
2938           std::string SymName =
2939               Demangle ? demangle(*NameOrErr) : NameOrErr->str();
2940 
2941           if (SymbolDescription)
2942             SymName = getXCOFFSymbolDescription(createSymbolInfo(O, *SymRef),
2943                                                 SymName);
2944 
2945           outs() << ' ' << SymName;
2946           outs() << ") ";
2947         } else
2948           reportWarning(toString(NameOrErr.takeError()), FileName);
2949       }
2950     }
2951   }
2952 
2953   if (Common)
2954     outs() << '\t' << format(Fmt, static_cast<uint64_t>(Symbol.getAlignment()));
2955   else if (O.isXCOFF())
2956     outs() << '\t'
2957            << format(Fmt, cast<XCOFFObjectFile>(O).getSymbolSize(
2958                               Symbol.getRawDataRefImpl()));
2959   else if (O.isELF())
2960     outs() << '\t' << format(Fmt, ELFSymbolRef(Symbol).getSize());
2961   else if (O.isWasm())
2962     outs() << '\t'
2963            << format(Fmt, static_cast<uint64_t>(
2964                               cast<WasmObjectFile>(O).getSymbolSize(Symbol)));
2965 
2966   if (O.isELF()) {
2967     if (!SymbolVersions.empty()) {
2968       const VersionEntry &Ver =
2969           SymbolVersions[Symbol.getRawDataRefImpl().d.b - 1];
2970       std::string Str;
2971       if (!Ver.Name.empty())
2972         Str = Ver.IsVerDef ? ' ' + Ver.Name : '(' + Ver.Name + ')';
2973       outs() << ' ' << left_justify(Str, 12);
2974     }
2975 
2976     uint8_t Other = ELFSymbolRef(Symbol).getOther();
2977     switch (Other) {
2978     case ELF::STV_DEFAULT:
2979       break;
2980     case ELF::STV_INTERNAL:
2981       outs() << " .internal";
2982       break;
2983     case ELF::STV_HIDDEN:
2984       outs() << " .hidden";
2985       break;
2986     case ELF::STV_PROTECTED:
2987       outs() << " .protected";
2988       break;
2989     default:
2990       outs() << format(" 0x%02x", Other);
2991       break;
2992     }
2993   } else if (Hidden) {
2994     outs() << " .hidden";
2995   }
2996 
2997   std::string SymName = Demangle ? demangle(Name) : Name.str();
2998   if (O.isXCOFF() && SymbolDescription)
2999     SymName = getXCOFFSymbolDescription(createSymbolInfo(O, Symbol), SymName);
3000 
3001   outs() << ' ' << SymName << '\n';
3002 }
3003 
3004 static void printUnwindInfo(const ObjectFile *O) {
3005   outs() << "Unwind info:\n\n";
3006 
3007   if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O))
3008     printCOFFUnwindInfo(Coff);
3009   else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O))
3010     printMachOUnwindInfo(MachO);
3011   else
3012     // TODO: Extract DWARF dump tool to objdump.
3013     WithColor::error(errs(), ToolName)
3014         << "This operation is only currently supported "
3015            "for COFF and MachO object files.\n";
3016 }
3017 
3018 /// Dump the raw contents of the __clangast section so the output can be piped
3019 /// into llvm-bcanalyzer.
3020 static void printRawClangAST(const ObjectFile *Obj) {
3021   if (outs().is_displayed()) {
3022     WithColor::error(errs(), ToolName)
3023         << "The -raw-clang-ast option will dump the raw binary contents of "
3024            "the clang ast section.\n"
3025            "Please redirect the output to a file or another program such as "
3026            "llvm-bcanalyzer.\n";
3027     return;
3028   }
3029 
3030   StringRef ClangASTSectionName("__clangast");
3031   if (Obj->isCOFF()) {
3032     ClangASTSectionName = "clangast";
3033   }
3034 
3035   std::optional<object::SectionRef> ClangASTSection;
3036   for (auto Sec : ToolSectionFilter(*Obj)) {
3037     StringRef Name;
3038     if (Expected<StringRef> NameOrErr = Sec.getName())
3039       Name = *NameOrErr;
3040     else
3041       consumeError(NameOrErr.takeError());
3042 
3043     if (Name == ClangASTSectionName) {
3044       ClangASTSection = Sec;
3045       break;
3046     }
3047   }
3048   if (!ClangASTSection)
3049     return;
3050 
3051   StringRef ClangASTContents =
3052       unwrapOrError(ClangASTSection->getContents(), Obj->getFileName());
3053   outs().write(ClangASTContents.data(), ClangASTContents.size());
3054 }
3055 
3056 static void printFaultMaps(const ObjectFile *Obj) {
3057   StringRef FaultMapSectionName;
3058 
3059   if (Obj->isELF()) {
3060     FaultMapSectionName = ".llvm_faultmaps";
3061   } else if (Obj->isMachO()) {
3062     FaultMapSectionName = "__llvm_faultmaps";
3063   } else {
3064     WithColor::error(errs(), ToolName)
3065         << "This operation is only currently supported "
3066            "for ELF and Mach-O executable files.\n";
3067     return;
3068   }
3069 
3070   std::optional<object::SectionRef> FaultMapSection;
3071 
3072   for (auto Sec : ToolSectionFilter(*Obj)) {
3073     StringRef Name;
3074     if (Expected<StringRef> NameOrErr = Sec.getName())
3075       Name = *NameOrErr;
3076     else
3077       consumeError(NameOrErr.takeError());
3078 
3079     if (Name == FaultMapSectionName) {
3080       FaultMapSection = Sec;
3081       break;
3082     }
3083   }
3084 
3085   outs() << "FaultMap table:\n";
3086 
3087   if (!FaultMapSection) {
3088     outs() << "<not found>\n";
3089     return;
3090   }
3091 
3092   StringRef FaultMapContents =
3093       unwrapOrError(FaultMapSection->getContents(), Obj->getFileName());
3094   FaultMapParser FMP(FaultMapContents.bytes_begin(),
3095                      FaultMapContents.bytes_end());
3096 
3097   outs() << FMP;
3098 }
3099 
3100 void Dumper::printPrivateHeaders() {
3101   reportError(O.getFileName(), "Invalid/Unsupported object file format");
3102 }
3103 
3104 static void printFileHeaders(const ObjectFile *O) {
3105   if (!O->isELF() && !O->isCOFF())
3106     reportError(O->getFileName(), "Invalid/Unsupported object file format");
3107 
3108   Triple::ArchType AT = O->getArch();
3109   outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n";
3110   uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName());
3111 
3112   StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
3113   outs() << "start address: "
3114          << "0x" << format(Fmt.data(), Address) << "\n";
3115 }
3116 
3117 static void printArchiveChild(StringRef Filename, const Archive::Child &C) {
3118   Expected<sys::fs::perms> ModeOrErr = C.getAccessMode();
3119   if (!ModeOrErr) {
3120     WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n";
3121     consumeError(ModeOrErr.takeError());
3122     return;
3123   }
3124   sys::fs::perms Mode = ModeOrErr.get();
3125   outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
3126   outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
3127   outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
3128   outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
3129   outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
3130   outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
3131   outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
3132   outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
3133   outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
3134 
3135   outs() << " ";
3136 
3137   outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename),
3138                    unwrapOrError(C.getGID(), Filename),
3139                    unwrapOrError(C.getRawSize(), Filename));
3140 
3141   StringRef RawLastModified = C.getRawLastModified();
3142   unsigned Seconds;
3143   if (RawLastModified.getAsInteger(10, Seconds))
3144     outs() << "(date: \"" << RawLastModified
3145            << "\" contains non-decimal chars) ";
3146   else {
3147     // Since ctime(3) returns a 26 character string of the form:
3148     // "Sun Sep 16 01:03:52 1973\n\0"
3149     // just print 24 characters.
3150     time_t t = Seconds;
3151     outs() << format("%.24s ", ctime(&t));
3152   }
3153 
3154   StringRef Name = "";
3155   Expected<StringRef> NameOrErr = C.getName();
3156   if (!NameOrErr) {
3157     consumeError(NameOrErr.takeError());
3158     Name = unwrapOrError(C.getRawName(), Filename);
3159   } else {
3160     Name = NameOrErr.get();
3161   }
3162   outs() << Name << "\n";
3163 }
3164 
3165 // For ELF only now.
3166 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) {
3167   if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) {
3168     if (Elf->getEType() != ELF::ET_REL)
3169       return true;
3170   }
3171   return false;
3172 }
3173 
3174 static void checkForInvalidStartStopAddress(ObjectFile *Obj,
3175                                             uint64_t Start, uint64_t Stop) {
3176   if (!shouldWarnForInvalidStartStopAddress(Obj))
3177     return;
3178 
3179   for (const SectionRef &Section : Obj->sections())
3180     if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) {
3181       uint64_t BaseAddr = Section.getAddress();
3182       uint64_t Size = Section.getSize();
3183       if ((Start < BaseAddr + Size) && Stop > BaseAddr)
3184         return;
3185     }
3186 
3187   if (!HasStartAddressFlag)
3188     reportWarning("no section has address less than 0x" +
3189                       Twine::utohexstr(Stop) + " specified by --stop-address",
3190                   Obj->getFileName());
3191   else if (!HasStopAddressFlag)
3192     reportWarning("no section has address greater than or equal to 0x" +
3193                       Twine::utohexstr(Start) + " specified by --start-address",
3194                   Obj->getFileName());
3195   else
3196     reportWarning("no section overlaps the range [0x" +
3197                       Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) +
3198                       ") specified by --start-address/--stop-address",
3199                   Obj->getFileName());
3200 }
3201 
3202 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
3203                        const Archive::Child *C = nullptr) {
3204   Expected<std::unique_ptr<Dumper>> DumperOrErr = createDumper(*O);
3205   if (!DumperOrErr) {
3206     reportError(DumperOrErr.takeError(), O->getFileName(),
3207                 A ? A->getFileName() : "");
3208     return;
3209   }
3210   Dumper &D = **DumperOrErr;
3211 
3212   // Avoid other output when using a raw option.
3213   if (!RawClangAST) {
3214     outs() << '\n';
3215     if (A)
3216       outs() << A->getFileName() << "(" << O->getFileName() << ")";
3217     else
3218       outs() << O->getFileName();
3219     outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n";
3220   }
3221 
3222   if (HasStartAddressFlag || HasStopAddressFlag)
3223     checkForInvalidStartStopAddress(O, StartAddress, StopAddress);
3224 
3225   // TODO: Change print* free functions to Dumper member functions to utilitize
3226   // stateful functions like reportUniqueWarning.
3227 
3228   // Note: the order here matches GNU objdump for compatability.
3229   StringRef ArchiveName = A ? A->getFileName() : "";
3230   if (ArchiveHeaders && !MachOOpt && C)
3231     printArchiveChild(ArchiveName, *C);
3232   if (FileHeaders)
3233     printFileHeaders(O);
3234   if (PrivateHeaders || FirstPrivateHeader)
3235     D.printPrivateHeaders();
3236   if (SectionHeaders)
3237     printSectionHeaders(*O);
3238   if (SymbolTable)
3239     D.printSymbolTable(ArchiveName);
3240   if (DynamicSymbolTable)
3241     D.printSymbolTable(ArchiveName, /*ArchitectureName=*/"",
3242                        /*DumpDynamic=*/true);
3243   if (DwarfDumpType != DIDT_Null) {
3244     std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O);
3245     // Dump the complete DWARF structure.
3246     DIDumpOptions DumpOpts;
3247     DumpOpts.DumpType = DwarfDumpType;
3248     DICtx->dump(outs(), DumpOpts);
3249   }
3250   if (Relocations && !Disassemble)
3251     D.printRelocations();
3252   if (DynamicRelocations)
3253     D.printDynamicRelocations();
3254   if (SectionContents)
3255     printSectionContents(O);
3256   if (Disassemble)
3257     disassembleObject(O, Relocations);
3258   if (UnwindInfo)
3259     printUnwindInfo(O);
3260 
3261   // Mach-O specific options:
3262   if (ExportsTrie)
3263     printExportsTrie(O);
3264   if (Rebase)
3265     printRebaseTable(O);
3266   if (Bind)
3267     printBindTable(O);
3268   if (LazyBind)
3269     printLazyBindTable(O);
3270   if (WeakBind)
3271     printWeakBindTable(O);
3272 
3273   // Other special sections:
3274   if (RawClangAST)
3275     printRawClangAST(O);
3276   if (FaultMapSection)
3277     printFaultMaps(O);
3278   if (Offloading)
3279     dumpOffloadBinary(*O);
3280 }
3281 
3282 static void dumpObject(const COFFImportFile *I, const Archive *A,
3283                        const Archive::Child *C = nullptr) {
3284   StringRef ArchiveName = A ? A->getFileName() : "";
3285 
3286   // Avoid other output when using a raw option.
3287   if (!RawClangAST)
3288     outs() << '\n'
3289            << ArchiveName << "(" << I->getFileName() << ")"
3290            << ":\tfile format COFF-import-file"
3291            << "\n\n";
3292 
3293   if (ArchiveHeaders && !MachOOpt && C)
3294     printArchiveChild(ArchiveName, *C);
3295   if (SymbolTable)
3296     printCOFFSymbolTable(*I);
3297 }
3298 
3299 /// Dump each object file in \a a;
3300 static void dumpArchive(const Archive *A) {
3301   Error Err = Error::success();
3302   unsigned I = -1;
3303   for (auto &C : A->children(Err)) {
3304     ++I;
3305     Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
3306     if (!ChildOrErr) {
3307       if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError()))
3308         reportError(std::move(E), getFileNameForError(C, I), A->getFileName());
3309       continue;
3310     }
3311     if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
3312       dumpObject(O, A, &C);
3313     else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get()))
3314       dumpObject(I, A, &C);
3315     else
3316       reportError(errorCodeToError(object_error::invalid_file_type),
3317                   A->getFileName());
3318   }
3319   if (Err)
3320     reportError(std::move(Err), A->getFileName());
3321 }
3322 
3323 /// Open file and figure out how to dump it.
3324 static void dumpInput(StringRef file) {
3325   // If we are using the Mach-O specific object file parser, then let it parse
3326   // the file and process the command line options.  So the -arch flags can
3327   // be used to select specific slices, etc.
3328   if (MachOOpt) {
3329     parseInputMachO(file);
3330     return;
3331   }
3332 
3333   // Attempt to open the binary.
3334   OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file);
3335   Binary &Binary = *OBinary.getBinary();
3336 
3337   if (Archive *A = dyn_cast<Archive>(&Binary))
3338     dumpArchive(A);
3339   else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary))
3340     dumpObject(O);
3341   else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary))
3342     parseInputMachO(UB);
3343   else if (OffloadBinary *OB = dyn_cast<OffloadBinary>(&Binary))
3344     dumpOffloadSections(*OB);
3345   else
3346     reportError(errorCodeToError(object_error::invalid_file_type), file);
3347 }
3348 
3349 template <typename T>
3350 static void parseIntArg(const llvm::opt::InputArgList &InputArgs, int ID,
3351                         T &Value) {
3352   if (const opt::Arg *A = InputArgs.getLastArg(ID)) {
3353     StringRef V(A->getValue());
3354     if (!llvm::to_integer(V, Value, 0)) {
3355       reportCmdLineError(A->getSpelling() +
3356                          ": expected a non-negative integer, but got '" + V +
3357                          "'");
3358     }
3359   }
3360 }
3361 
3362 static object::BuildID parseBuildIDArg(const opt::Arg *A) {
3363   StringRef V(A->getValue());
3364   object::BuildID BID = parseBuildID(V);
3365   if (BID.empty())
3366     reportCmdLineError(A->getSpelling() + ": expected a build ID, but got '" +
3367                        V + "'");
3368   return BID;
3369 }
3370 
3371 void objdump::invalidArgValue(const opt::Arg *A) {
3372   reportCmdLineError("'" + StringRef(A->getValue()) +
3373                      "' is not a valid value for '" + A->getSpelling() + "'");
3374 }
3375 
3376 static std::vector<std::string>
3377 commaSeparatedValues(const llvm::opt::InputArgList &InputArgs, int ID) {
3378   std::vector<std::string> Values;
3379   for (StringRef Value : InputArgs.getAllArgValues(ID)) {
3380     llvm::SmallVector<StringRef, 2> SplitValues;
3381     llvm::SplitString(Value, SplitValues, ",");
3382     for (StringRef SplitValue : SplitValues)
3383       Values.push_back(SplitValue.str());
3384   }
3385   return Values;
3386 }
3387 
3388 static void parseOtoolOptions(const llvm::opt::InputArgList &InputArgs) {
3389   MachOOpt = true;
3390   FullLeadingAddr = true;
3391   PrintImmHex = true;
3392 
3393   ArchName = InputArgs.getLastArgValue(OTOOL_arch).str();
3394   LinkOptHints = InputArgs.hasArg(OTOOL_C);
3395   if (InputArgs.hasArg(OTOOL_d))
3396     FilterSections.push_back("__DATA,__data");
3397   DylibId = InputArgs.hasArg(OTOOL_D);
3398   UniversalHeaders = InputArgs.hasArg(OTOOL_f);
3399   DataInCode = InputArgs.hasArg(OTOOL_G);
3400   FirstPrivateHeader = InputArgs.hasArg(OTOOL_h);
3401   IndirectSymbols = InputArgs.hasArg(OTOOL_I);
3402   ShowRawInsn = InputArgs.hasArg(OTOOL_j);
3403   PrivateHeaders = InputArgs.hasArg(OTOOL_l);
3404   DylibsUsed = InputArgs.hasArg(OTOOL_L);
3405   MCPU = InputArgs.getLastArgValue(OTOOL_mcpu_EQ).str();
3406   ObjcMetaData = InputArgs.hasArg(OTOOL_o);
3407   DisSymName = InputArgs.getLastArgValue(OTOOL_p).str();
3408   InfoPlist = InputArgs.hasArg(OTOOL_P);
3409   Relocations = InputArgs.hasArg(OTOOL_r);
3410   if (const Arg *A = InputArgs.getLastArg(OTOOL_s)) {
3411     auto Filter = (A->getValue(0) + StringRef(",") + A->getValue(1)).str();
3412     FilterSections.push_back(Filter);
3413   }
3414   if (InputArgs.hasArg(OTOOL_t))
3415     FilterSections.push_back("__TEXT,__text");
3416   Verbose = InputArgs.hasArg(OTOOL_v) || InputArgs.hasArg(OTOOL_V) ||
3417             InputArgs.hasArg(OTOOL_o);
3418   SymbolicOperands = InputArgs.hasArg(OTOOL_V);
3419   if (InputArgs.hasArg(OTOOL_x))
3420     FilterSections.push_back(",__text");
3421   LeadingAddr = LeadingHeaders = !InputArgs.hasArg(OTOOL_X);
3422 
3423   ChainedFixups = InputArgs.hasArg(OTOOL_chained_fixups);
3424   DyldInfo = InputArgs.hasArg(OTOOL_dyld_info);
3425 
3426   InputFilenames = InputArgs.getAllArgValues(OTOOL_INPUT);
3427   if (InputFilenames.empty())
3428     reportCmdLineError("no input file");
3429 
3430   for (const Arg *A : InputArgs) {
3431     const Option &O = A->getOption();
3432     if (O.getGroup().isValid() && O.getGroup().getID() == OTOOL_grp_obsolete) {
3433       reportCmdLineWarning(O.getPrefixedName() +
3434                            " is obsolete and not implemented");
3435     }
3436   }
3437 }
3438 
3439 static void parseObjdumpOptions(const llvm::opt::InputArgList &InputArgs) {
3440   parseIntArg(InputArgs, OBJDUMP_adjust_vma_EQ, AdjustVMA);
3441   AllHeaders = InputArgs.hasArg(OBJDUMP_all_headers);
3442   ArchName = InputArgs.getLastArgValue(OBJDUMP_arch_name_EQ).str();
3443   ArchiveHeaders = InputArgs.hasArg(OBJDUMP_archive_headers);
3444   Demangle = InputArgs.hasArg(OBJDUMP_demangle);
3445   Disassemble = InputArgs.hasArg(OBJDUMP_disassemble);
3446   DisassembleAll = InputArgs.hasArg(OBJDUMP_disassemble_all);
3447   SymbolDescription = InputArgs.hasArg(OBJDUMP_symbol_description);
3448   TracebackTable = InputArgs.hasArg(OBJDUMP_traceback_table);
3449   DisassembleSymbols =
3450       commaSeparatedValues(InputArgs, OBJDUMP_disassemble_symbols_EQ);
3451   DisassembleZeroes = InputArgs.hasArg(OBJDUMP_disassemble_zeroes);
3452   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_dwarf_EQ)) {
3453     DwarfDumpType = StringSwitch<DIDumpType>(A->getValue())
3454                         .Case("frames", DIDT_DebugFrame)
3455                         .Default(DIDT_Null);
3456     if (DwarfDumpType == DIDT_Null)
3457       invalidArgValue(A);
3458   }
3459   DynamicRelocations = InputArgs.hasArg(OBJDUMP_dynamic_reloc);
3460   FaultMapSection = InputArgs.hasArg(OBJDUMP_fault_map_section);
3461   Offloading = InputArgs.hasArg(OBJDUMP_offloading);
3462   FileHeaders = InputArgs.hasArg(OBJDUMP_file_headers);
3463   SectionContents = InputArgs.hasArg(OBJDUMP_full_contents);
3464   PrintLines = InputArgs.hasArg(OBJDUMP_line_numbers);
3465   InputFilenames = InputArgs.getAllArgValues(OBJDUMP_INPUT);
3466   MachOOpt = InputArgs.hasArg(OBJDUMP_macho);
3467   MCPU = InputArgs.getLastArgValue(OBJDUMP_mcpu_EQ).str();
3468   MAttrs = commaSeparatedValues(InputArgs, OBJDUMP_mattr_EQ);
3469   ShowRawInsn = !InputArgs.hasArg(OBJDUMP_no_show_raw_insn);
3470   LeadingAddr = !InputArgs.hasArg(OBJDUMP_no_leading_addr);
3471   RawClangAST = InputArgs.hasArg(OBJDUMP_raw_clang_ast);
3472   Relocations = InputArgs.hasArg(OBJDUMP_reloc);
3473   PrintImmHex =
3474       InputArgs.hasFlag(OBJDUMP_print_imm_hex, OBJDUMP_no_print_imm_hex, true);
3475   PrivateHeaders = InputArgs.hasArg(OBJDUMP_private_headers);
3476   FilterSections = InputArgs.getAllArgValues(OBJDUMP_section_EQ);
3477   SectionHeaders = InputArgs.hasArg(OBJDUMP_section_headers);
3478   ShowAllSymbols = InputArgs.hasArg(OBJDUMP_show_all_symbols);
3479   ShowLMA = InputArgs.hasArg(OBJDUMP_show_lma);
3480   PrintSource = InputArgs.hasArg(OBJDUMP_source);
3481   parseIntArg(InputArgs, OBJDUMP_start_address_EQ, StartAddress);
3482   HasStartAddressFlag = InputArgs.hasArg(OBJDUMP_start_address_EQ);
3483   parseIntArg(InputArgs, OBJDUMP_stop_address_EQ, StopAddress);
3484   HasStopAddressFlag = InputArgs.hasArg(OBJDUMP_stop_address_EQ);
3485   SymbolTable = InputArgs.hasArg(OBJDUMP_syms);
3486   SymbolizeOperands = InputArgs.hasArg(OBJDUMP_symbolize_operands);
3487   PrettyPGOAnalysisMap = InputArgs.hasArg(OBJDUMP_pretty_pgo_analysis_map);
3488   if (PrettyPGOAnalysisMap && !SymbolizeOperands)
3489     reportCmdLineWarning("--symbolize-operands must be enabled for "
3490                          "--pretty-pgo-analysis-map to have an effect");
3491   DynamicSymbolTable = InputArgs.hasArg(OBJDUMP_dynamic_syms);
3492   TripleName = InputArgs.getLastArgValue(OBJDUMP_triple_EQ).str();
3493   UnwindInfo = InputArgs.hasArg(OBJDUMP_unwind_info);
3494   Wide = InputArgs.hasArg(OBJDUMP_wide);
3495   Prefix = InputArgs.getLastArgValue(OBJDUMP_prefix).str();
3496   parseIntArg(InputArgs, OBJDUMP_prefix_strip, PrefixStrip);
3497   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_debug_vars_EQ)) {
3498     DbgVariables = StringSwitch<DebugVarsFormat>(A->getValue())
3499                        .Case("ascii", DVASCII)
3500                        .Case("unicode", DVUnicode)
3501                        .Default(DVInvalid);
3502     if (DbgVariables == DVInvalid)
3503       invalidArgValue(A);
3504   }
3505   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_disassembler_color_EQ)) {
3506     DisassemblyColor = StringSwitch<ColorOutput>(A->getValue())
3507                            .Case("on", ColorOutput::Enable)
3508                            .Case("off", ColorOutput::Disable)
3509                            .Case("terminal", ColorOutput::Auto)
3510                            .Default(ColorOutput::Invalid);
3511     if (DisassemblyColor == ColorOutput::Invalid)
3512       invalidArgValue(A);
3513   }
3514 
3515   parseIntArg(InputArgs, OBJDUMP_debug_vars_indent_EQ, DbgIndent);
3516 
3517   parseMachOOptions(InputArgs);
3518 
3519   // Parse -M (--disassembler-options) and deprecated
3520   // --x86-asm-syntax={att,intel}.
3521   //
3522   // Note, for x86, the asm dialect (AssemblerDialect) is initialized when the
3523   // MCAsmInfo is constructed. MCInstPrinter::applyTargetSpecificCLOption is
3524   // called too late. For now we have to use the internal cl::opt option.
3525   const char *AsmSyntax = nullptr;
3526   for (const auto *A : InputArgs.filtered(OBJDUMP_disassembler_options_EQ,
3527                                           OBJDUMP_x86_asm_syntax_att,
3528                                           OBJDUMP_x86_asm_syntax_intel)) {
3529     switch (A->getOption().getID()) {
3530     case OBJDUMP_x86_asm_syntax_att:
3531       AsmSyntax = "--x86-asm-syntax=att";
3532       continue;
3533     case OBJDUMP_x86_asm_syntax_intel:
3534       AsmSyntax = "--x86-asm-syntax=intel";
3535       continue;
3536     }
3537 
3538     SmallVector<StringRef, 2> Values;
3539     llvm::SplitString(A->getValue(), Values, ",");
3540     for (StringRef V : Values) {
3541       if (V == "att")
3542         AsmSyntax = "--x86-asm-syntax=att";
3543       else if (V == "intel")
3544         AsmSyntax = "--x86-asm-syntax=intel";
3545       else
3546         DisassemblerOptions.push_back(V.str());
3547     }
3548   }
3549   SmallVector<const char *> Args = {"llvm-objdump"};
3550   for (const opt::Arg *A : InputArgs.filtered(OBJDUMP_mllvm))
3551     Args.push_back(A->getValue());
3552   if (AsmSyntax)
3553     Args.push_back(AsmSyntax);
3554   if (Args.size() > 1)
3555     llvm::cl::ParseCommandLineOptions(Args.size(), Args.data());
3556 
3557   // Look up any provided build IDs, then append them to the input filenames.
3558   for (const opt::Arg *A : InputArgs.filtered(OBJDUMP_build_id)) {
3559     object::BuildID BuildID = parseBuildIDArg(A);
3560     std::optional<std::string> Path = BIDFetcher->fetch(BuildID);
3561     if (!Path) {
3562       reportCmdLineError(A->getSpelling() + ": could not find build ID '" +
3563                          A->getValue() + "'");
3564     }
3565     InputFilenames.push_back(std::move(*Path));
3566   }
3567 
3568   // objdump defaults to a.out if no filenames specified.
3569   if (InputFilenames.empty())
3570     InputFilenames.push_back("a.out");
3571 }
3572 
3573 int llvm_objdump_main(int argc, char **argv, const llvm::ToolContext &) {
3574   using namespace llvm;
3575 
3576   ToolName = argv[0];
3577   std::unique_ptr<CommonOptTable> T;
3578   OptSpecifier Unknown, HelpFlag, HelpHiddenFlag, VersionFlag;
3579 
3580   StringRef Stem = sys::path::stem(ToolName);
3581   auto Is = [=](StringRef Tool) {
3582     // We need to recognize the following filenames:
3583     //
3584     // llvm-objdump -> objdump
3585     // llvm-otool-10.exe -> otool
3586     // powerpc64-unknown-freebsd13-objdump -> objdump
3587     auto I = Stem.rfind_insensitive(Tool);
3588     return I != StringRef::npos &&
3589            (I + Tool.size() == Stem.size() || !isAlnum(Stem[I + Tool.size()]));
3590   };
3591   if (Is("otool")) {
3592     T = std::make_unique<OtoolOptTable>();
3593     Unknown = OTOOL_UNKNOWN;
3594     HelpFlag = OTOOL_help;
3595     HelpHiddenFlag = OTOOL_help_hidden;
3596     VersionFlag = OTOOL_version;
3597   } else {
3598     T = std::make_unique<ObjdumpOptTable>();
3599     Unknown = OBJDUMP_UNKNOWN;
3600     HelpFlag = OBJDUMP_help;
3601     HelpHiddenFlag = OBJDUMP_help_hidden;
3602     VersionFlag = OBJDUMP_version;
3603   }
3604 
3605   BumpPtrAllocator A;
3606   StringSaver Saver(A);
3607   opt::InputArgList InputArgs =
3608       T->parseArgs(argc, argv, Unknown, Saver,
3609                    [&](StringRef Msg) { reportCmdLineError(Msg); });
3610 
3611   if (InputArgs.size() == 0 || InputArgs.hasArg(HelpFlag)) {
3612     T->printHelp(ToolName);
3613     return 0;
3614   }
3615   if (InputArgs.hasArg(HelpHiddenFlag)) {
3616     T->printHelp(ToolName, /*ShowHidden=*/true);
3617     return 0;
3618   }
3619 
3620   // Initialize targets and assembly printers/parsers.
3621   InitializeAllTargetInfos();
3622   InitializeAllTargetMCs();
3623   InitializeAllDisassemblers();
3624 
3625   if (InputArgs.hasArg(VersionFlag)) {
3626     cl::PrintVersionMessage();
3627     if (!Is("otool")) {
3628       outs() << '\n';
3629       TargetRegistry::printRegisteredTargetsForVersion(outs());
3630     }
3631     return 0;
3632   }
3633 
3634   // Initialize debuginfod.
3635   const bool ShouldUseDebuginfodByDefault =
3636       InputArgs.hasArg(OBJDUMP_build_id) || canUseDebuginfod();
3637   std::vector<std::string> DebugFileDirectories =
3638       InputArgs.getAllArgValues(OBJDUMP_debug_file_directory);
3639   if (InputArgs.hasFlag(OBJDUMP_debuginfod, OBJDUMP_no_debuginfod,
3640                         ShouldUseDebuginfodByDefault)) {
3641     HTTPClient::initialize();
3642     BIDFetcher =
3643         std::make_unique<DebuginfodFetcher>(std::move(DebugFileDirectories));
3644   } else {
3645     BIDFetcher =
3646         std::make_unique<BuildIDFetcher>(std::move(DebugFileDirectories));
3647   }
3648 
3649   if (Is("otool"))
3650     parseOtoolOptions(InputArgs);
3651   else
3652     parseObjdumpOptions(InputArgs);
3653 
3654   if (StartAddress >= StopAddress)
3655     reportCmdLineError("start address should be less than stop address");
3656 
3657   // Removes trailing separators from prefix.
3658   while (!Prefix.empty() && sys::path::is_separator(Prefix.back()))
3659     Prefix.pop_back();
3660 
3661   if (AllHeaders)
3662     ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations =
3663         SectionHeaders = SymbolTable = true;
3664 
3665   if (DisassembleAll || PrintSource || PrintLines || TracebackTable ||
3666       !DisassembleSymbols.empty())
3667     Disassemble = true;
3668 
3669   if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null &&
3670       !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST &&
3671       !Relocations && !SectionHeaders && !SectionContents && !SymbolTable &&
3672       !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && !Offloading &&
3673       !(MachOOpt &&
3674         (Bind || DataInCode || ChainedFixups || DyldInfo || DylibId ||
3675          DylibsUsed || ExportsTrie || FirstPrivateHeader ||
3676          FunctionStartsType != FunctionStartsMode::None || IndirectSymbols ||
3677          InfoPlist || LazyBind || LinkOptHints || ObjcMetaData || Rebase ||
3678          Rpaths || UniversalHeaders || WeakBind || !FilterSections.empty()))) {
3679     T->printHelp(ToolName);
3680     return 2;
3681   }
3682 
3683   DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end());
3684 
3685   llvm::for_each(InputFilenames, dumpInput);
3686 
3687   warnOnNoMatchForSections();
3688 
3689   return EXIT_SUCCESS;
3690 }
3691