xref: /llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp (revision 1128311a19179ceca799ff0fbc4dd206ab56e560)
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 "SourcePrinter.h"
24 #include "WasmDump.h"
25 #include "XCOFFDump.h"
26 #include "llvm/ADT/IndexedMap.h"
27 #include "llvm/ADT/Optional.h"
28 #include "llvm/ADT/STLExtras.h"
29 #include "llvm/ADT/SetOperations.h"
30 #include "llvm/ADT/SmallSet.h"
31 #include "llvm/ADT/StringExtras.h"
32 #include "llvm/ADT/StringSet.h"
33 #include "llvm/ADT/Triple.h"
34 #include "llvm/ADT/Twine.h"
35 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
36 #include "llvm/DebugInfo/Symbolize/Symbolize.h"
37 #include "llvm/Demangle/Demangle.h"
38 #include "llvm/MC/MCAsmInfo.h"
39 #include "llvm/MC/MCContext.h"
40 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
41 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h"
42 #include "llvm/MC/MCInst.h"
43 #include "llvm/MC/MCInstPrinter.h"
44 #include "llvm/MC/MCInstrAnalysis.h"
45 #include "llvm/MC/MCInstrInfo.h"
46 #include "llvm/MC/MCObjectFileInfo.h"
47 #include "llvm/MC/MCRegisterInfo.h"
48 #include "llvm/MC/MCSubtargetInfo.h"
49 #include "llvm/MC/MCTargetOptions.h"
50 #include "llvm/Object/Archive.h"
51 #include "llvm/Object/COFF.h"
52 #include "llvm/Object/COFFImportFile.h"
53 #include "llvm/Object/ELFObjectFile.h"
54 #include "llvm/Object/FaultMapParser.h"
55 #include "llvm/Object/MachO.h"
56 #include "llvm/Object/MachOUniversal.h"
57 #include "llvm/Object/ObjectFile.h"
58 #include "llvm/Object/Wasm.h"
59 #include "llvm/Option/Arg.h"
60 #include "llvm/Option/ArgList.h"
61 #include "llvm/Option/Option.h"
62 #include "llvm/Support/Casting.h"
63 #include "llvm/Support/Debug.h"
64 #include "llvm/Support/Errc.h"
65 #include "llvm/Support/FileSystem.h"
66 #include "llvm/Support/Format.h"
67 #include "llvm/Support/FormatVariadic.h"
68 #include "llvm/Support/GraphWriter.h"
69 #include "llvm/Support/Host.h"
70 #include "llvm/Support/InitLLVM.h"
71 #include "llvm/Support/MemoryBuffer.h"
72 #include "llvm/Support/SourceMgr.h"
73 #include "llvm/Support/StringSaver.h"
74 #include "llvm/Support/TargetRegistry.h"
75 #include "llvm/Support/TargetSelect.h"
76 #include "llvm/Support/WithColor.h"
77 #include "llvm/Support/raw_ostream.h"
78 #include <algorithm>
79 #include <cctype>
80 #include <cstring>
81 #include <system_error>
82 #include <unordered_map>
83 #include <utility>
84 
85 using namespace llvm;
86 using namespace llvm::object;
87 using namespace llvm::objdump;
88 using namespace llvm::opt;
89 
90 namespace {
91 
92 class CommonOptTable : public opt::OptTable {
93 public:
94   CommonOptTable(ArrayRef<Info> OptionInfos, const char *Usage,
95                  const char *Description)
96       : OptTable(OptionInfos), Usage(Usage), Description(Description) {
97     setGroupedShortOptions(true);
98   }
99 
100   void printHelp(StringRef Argv0, bool ShowHidden = false) const {
101     Argv0 = sys::path::filename(Argv0);
102     PrintHelp(outs(), (Argv0 + Usage).str().c_str(), Description, ShowHidden,
103               ShowHidden);
104     // TODO Replace this with OptTable API once it adds extrahelp support.
105     outs() << "\nPass @FILE as argument to read options from FILE.\n";
106   }
107 
108 private:
109   const char *Usage;
110   const char *Description;
111 };
112 
113 // ObjdumpOptID is in ObjdumpOptID.h
114 
115 #define PREFIX(NAME, VALUE) const char *const OBJDUMP_##NAME[] = VALUE;
116 #include "ObjdumpOpts.inc"
117 #undef PREFIX
118 
119 static constexpr opt::OptTable::Info ObjdumpInfoTable[] = {
120 #define OBJDUMP_nullptr nullptr
121 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM,  \
122                HELPTEXT, METAVAR, VALUES)                                      \
123   {OBJDUMP_##PREFIX, NAME,         HELPTEXT,                                   \
124    METAVAR,          OBJDUMP_##ID, opt::Option::KIND##Class,                   \
125    PARAM,            FLAGS,        OBJDUMP_##GROUP,                            \
126    OBJDUMP_##ALIAS,  ALIASARGS,    VALUES},
127 #include "ObjdumpOpts.inc"
128 #undef OPTION
129 #undef OBJDUMP_nullptr
130 };
131 
132 class ObjdumpOptTable : public CommonOptTable {
133 public:
134   ObjdumpOptTable()
135       : CommonOptTable(ObjdumpInfoTable, " [options] <input object files>",
136                        "llvm object file dumper") {}
137 };
138 
139 enum OtoolOptID {
140   OTOOL_INVALID = 0, // This is not an option ID.
141 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM,  \
142                HELPTEXT, METAVAR, VALUES)                                      \
143   OTOOL_##ID,
144 #include "OtoolOpts.inc"
145 #undef OPTION
146 };
147 
148 #define PREFIX(NAME, VALUE) const char *const OTOOL_##NAME[] = VALUE;
149 #include "OtoolOpts.inc"
150 #undef PREFIX
151 
152 static constexpr opt::OptTable::Info OtoolInfoTable[] = {
153 #define OTOOL_nullptr nullptr
154 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM,  \
155                HELPTEXT, METAVAR, VALUES)                                      \
156   {OTOOL_##PREFIX, NAME,       HELPTEXT,                                       \
157    METAVAR,        OTOOL_##ID, opt::Option::KIND##Class,                       \
158    PARAM,          FLAGS,      OTOOL_##GROUP,                                  \
159    OTOOL_##ALIAS,  ALIASARGS,  VALUES},
160 #include "OtoolOpts.inc"
161 #undef OPTION
162 #undef OTOOL_nullptr
163 };
164 
165 class OtoolOptTable : public CommonOptTable {
166 public:
167   OtoolOptTable()
168       : CommonOptTable(OtoolInfoTable, " [option...] [file...]",
169                        "Mach-O object file displaying tool") {}
170 };
171 
172 } // namespace
173 
174 #define DEBUG_TYPE "objdump"
175 
176 static uint64_t AdjustVMA;
177 static bool AllHeaders;
178 static std::string ArchName;
179 bool objdump::ArchiveHeaders;
180 bool objdump::Demangle;
181 bool objdump::Disassemble;
182 bool objdump::DisassembleAll;
183 bool objdump::SymbolDescription;
184 static std::vector<std::string> DisassembleSymbols;
185 static bool DisassembleZeroes;
186 static std::vector<std::string> DisassemblerOptions;
187 DIDumpType objdump::DwarfDumpType;
188 static bool DynamicRelocations;
189 static bool FaultMapSection;
190 static bool FileHeaders;
191 bool objdump::SectionContents;
192 static std::vector<std::string> InputFilenames;
193 bool objdump::PrintLines;
194 static bool MachOOpt;
195 std::string objdump::MCPU;
196 std::vector<std::string> objdump::MAttrs;
197 bool objdump::ShowRawInsn;
198 bool objdump::LeadingAddr;
199 static bool RawClangAST;
200 bool objdump::Relocations;
201 bool objdump::PrintImmHex;
202 bool objdump::PrivateHeaders;
203 std::vector<std::string> objdump::FilterSections;
204 bool objdump::SectionHeaders;
205 static bool ShowLMA;
206 bool objdump::PrintSource;
207 
208 static uint64_t StartAddress;
209 static bool HasStartAddressFlag;
210 static uint64_t StopAddress = UINT64_MAX;
211 static bool HasStopAddressFlag;
212 
213 bool objdump::SymbolTable;
214 static bool SymbolizeOperands;
215 static bool DynamicSymbolTable;
216 std::string objdump::TripleName;
217 bool objdump::UnwindInfo;
218 static bool Wide;
219 std::string objdump::Prefix;
220 uint32_t objdump::PrefixStrip;
221 
222 DebugVarsFormat objdump::DbgVariables = DVDisabled;
223 
224 int objdump::DbgIndent = 40;
225 
226 static StringSet<> DisasmSymbolSet;
227 StringSet<> objdump::FoundSectionSet;
228 static StringRef ToolName;
229 
230 namespace {
231 struct FilterResult {
232   // True if the section should not be skipped.
233   bool Keep;
234 
235   // True if the index counter should be incremented, even if the section should
236   // be skipped. For example, sections may be skipped if they are not included
237   // in the --section flag, but we still want those to count toward the section
238   // count.
239   bool IncrementIndex;
240 };
241 } // namespace
242 
243 static FilterResult checkSectionFilter(object::SectionRef S) {
244   if (FilterSections.empty())
245     return {/*Keep=*/true, /*IncrementIndex=*/true};
246 
247   Expected<StringRef> SecNameOrErr = S.getName();
248   if (!SecNameOrErr) {
249     consumeError(SecNameOrErr.takeError());
250     return {/*Keep=*/false, /*IncrementIndex=*/false};
251   }
252   StringRef SecName = *SecNameOrErr;
253 
254   // StringSet does not allow empty key so avoid adding sections with
255   // no name (such as the section with index 0) here.
256   if (!SecName.empty())
257     FoundSectionSet.insert(SecName);
258 
259   // Only show the section if it's in the FilterSections list, but always
260   // increment so the indexing is stable.
261   return {/*Keep=*/is_contained(FilterSections, SecName),
262           /*IncrementIndex=*/true};
263 }
264 
265 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O,
266                                          uint64_t *Idx) {
267   // Start at UINT64_MAX so that the first index returned after an increment is
268   // zero (after the unsigned wrap).
269   if (Idx)
270     *Idx = UINT64_MAX;
271   return SectionFilter(
272       [Idx](object::SectionRef S) {
273         FilterResult Result = checkSectionFilter(S);
274         if (Idx != nullptr && Result.IncrementIndex)
275           *Idx += 1;
276         return Result.Keep;
277       },
278       O);
279 }
280 
281 std::string objdump::getFileNameForError(const object::Archive::Child &C,
282                                          unsigned Index) {
283   Expected<StringRef> NameOrErr = C.getName();
284   if (NameOrErr)
285     return std::string(NameOrErr.get());
286   // If we have an error getting the name then we print the index of the archive
287   // member. Since we are already in an error state, we just ignore this error.
288   consumeError(NameOrErr.takeError());
289   return "<file index: " + std::to_string(Index) + ">";
290 }
291 
292 void objdump::reportWarning(const Twine &Message, StringRef File) {
293   // Output order between errs() and outs() matters especially for archive
294   // files where the output is per member object.
295   outs().flush();
296   WithColor::warning(errs(), ToolName)
297       << "'" << File << "': " << Message << "\n";
298 }
299 
300 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(StringRef File,
301                                                   const Twine &Message) {
302   outs().flush();
303   WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n";
304   exit(1);
305 }
306 
307 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(Error E, StringRef FileName,
308                                                   StringRef ArchiveName,
309                                                   StringRef ArchitectureName) {
310   assert(E);
311   outs().flush();
312   WithColor::error(errs(), ToolName);
313   if (ArchiveName != "")
314     errs() << ArchiveName << "(" << FileName << ")";
315   else
316     errs() << "'" << FileName << "'";
317   if (!ArchitectureName.empty())
318     errs() << " (for architecture " << ArchitectureName << ")";
319   errs() << ": ";
320   logAllUnhandledErrors(std::move(E), errs());
321   exit(1);
322 }
323 
324 static void reportCmdLineWarning(const Twine &Message) {
325   WithColor::warning(errs(), ToolName) << Message << "\n";
326 }
327 
328 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(const Twine &Message) {
329   WithColor::error(errs(), ToolName) << Message << "\n";
330   exit(1);
331 }
332 
333 static void warnOnNoMatchForSections() {
334   SetVector<StringRef> MissingSections;
335   for (StringRef S : FilterSections) {
336     if (FoundSectionSet.count(S))
337       return;
338     // User may specify a unnamed section. Don't warn for it.
339     if (!S.empty())
340       MissingSections.insert(S);
341   }
342 
343   // Warn only if no section in FilterSections is matched.
344   for (StringRef S : MissingSections)
345     reportCmdLineWarning("section '" + S +
346                          "' mentioned in a -j/--section option, but not "
347                          "found in any input file");
348 }
349 
350 static const Target *getTarget(const ObjectFile *Obj) {
351   // Figure out the target triple.
352   Triple TheTriple("unknown-unknown-unknown");
353   if (TripleName.empty()) {
354     TheTriple = Obj->makeTriple();
355   } else {
356     TheTriple.setTriple(Triple::normalize(TripleName));
357     auto Arch = Obj->getArch();
358     if (Arch == Triple::arm || Arch == Triple::armeb)
359       Obj->setARMSubArch(TheTriple);
360   }
361 
362   // Get the target specific parser.
363   std::string Error;
364   const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple,
365                                                          Error);
366   if (!TheTarget)
367     reportError(Obj->getFileName(), "can't find target: " + Error);
368 
369   // Update the triple name and return the found target.
370   TripleName = TheTriple.getTriple();
371   return TheTarget;
372 }
373 
374 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) {
375   return A.getOffset() < B.getOffset();
376 }
377 
378 static Error getRelocationValueString(const RelocationRef &Rel,
379                                       SmallVectorImpl<char> &Result) {
380   const ObjectFile *Obj = Rel.getObject();
381   if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj))
382     return getELFRelocationValueString(ELF, Rel, Result);
383   if (auto *COFF = dyn_cast<COFFObjectFile>(Obj))
384     return getCOFFRelocationValueString(COFF, Rel, Result);
385   if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj))
386     return getWasmRelocationValueString(Wasm, Rel, Result);
387   if (auto *MachO = dyn_cast<MachOObjectFile>(Obj))
388     return getMachORelocationValueString(MachO, Rel, Result);
389   if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj))
390     return getXCOFFRelocationValueString(XCOFF, Rel, Result);
391   llvm_unreachable("unknown object file format");
392 }
393 
394 /// Indicates whether this relocation should hidden when listing
395 /// relocations, usually because it is the trailing part of a multipart
396 /// relocation that will be printed as part of the leading relocation.
397 static bool getHidden(RelocationRef RelRef) {
398   auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject());
399   if (!MachO)
400     return false;
401 
402   unsigned Arch = MachO->getArch();
403   DataRefImpl Rel = RelRef.getRawDataRefImpl();
404   uint64_t Type = MachO->getRelocationType(Rel);
405 
406   // On arches that use the generic relocations, GENERIC_RELOC_PAIR
407   // is always hidden.
408   if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc)
409     return Type == MachO::GENERIC_RELOC_PAIR;
410 
411   if (Arch == Triple::x86_64) {
412     // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
413     // an X86_64_RELOC_SUBTRACTOR.
414     if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) {
415       DataRefImpl RelPrev = Rel;
416       RelPrev.d.a--;
417       uint64_t PrevType = MachO->getRelocationType(RelPrev);
418       if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR)
419         return true;
420     }
421   }
422 
423   return false;
424 }
425 
426 namespace {
427 
428 /// Get the column at which we want to start printing the instruction
429 /// disassembly, taking into account anything which appears to the left of it.
430 unsigned getInstStartColumn(const MCSubtargetInfo &STI) {
431   return !ShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24;
432 }
433 
434 static bool isAArch64Elf(const ObjectFile *Obj) {
435   const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
436   return Elf && Elf->getEMachine() == ELF::EM_AARCH64;
437 }
438 
439 static bool isArmElf(const ObjectFile *Obj) {
440   const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
441   return Elf && Elf->getEMachine() == ELF::EM_ARM;
442 }
443 
444 static bool hasMappingSymbols(const ObjectFile *Obj) {
445   return isArmElf(Obj) || isAArch64Elf(Obj);
446 }
447 
448 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName,
449                             const RelocationRef &Rel, uint64_t Address,
450                             bool Is64Bits) {
451   StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ":  " : "\t\t\t%08" PRIx64 ":  ";
452   SmallString<16> Name;
453   SmallString<32> Val;
454   Rel.getTypeName(Name);
455   if (Error E = getRelocationValueString(Rel, Val))
456     reportError(std::move(E), FileName);
457   OS << format(Fmt.data(), Address) << Name << "\t" << Val;
458 }
459 
460 class PrettyPrinter {
461 public:
462   virtual ~PrettyPrinter() = default;
463   virtual void
464   printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
465             object::SectionedAddress Address, formatted_raw_ostream &OS,
466             StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
467             StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
468             LiveVariablePrinter &LVP) {
469     if (SP && (PrintSource || PrintLines))
470       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
471     LVP.printBetweenInsts(OS, false);
472 
473     size_t Start = OS.tell();
474     if (LeadingAddr)
475       OS << format("%8" PRIx64 ":", Address.Address);
476     if (ShowRawInsn) {
477       OS << ' ';
478       dumpBytes(Bytes, OS);
479     }
480 
481     // The output of printInst starts with a tab. Print some spaces so that
482     // the tab has 1 column and advances to the target tab stop.
483     unsigned TabStop = getInstStartColumn(STI);
484     unsigned Column = OS.tell() - Start;
485     OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8);
486 
487     if (MI) {
488       // See MCInstPrinter::printInst. On targets where a PC relative immediate
489       // is relative to the next instruction and the length of a MCInst is
490       // difficult to measure (x86), this is the address of the next
491       // instruction.
492       uint64_t Addr =
493           Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0);
494       IP.printInst(MI, Addr, "", STI, OS);
495     } else
496       OS << "\t<unknown>";
497   }
498 };
499 PrettyPrinter PrettyPrinterInst;
500 
501 class HexagonPrettyPrinter : public PrettyPrinter {
502 public:
503   void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
504                  formatted_raw_ostream &OS) {
505     uint32_t opcode =
506       (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
507     if (LeadingAddr)
508       OS << format("%8" PRIx64 ":", Address);
509     if (ShowRawInsn) {
510       OS << "\t";
511       dumpBytes(Bytes.slice(0, 4), OS);
512       OS << format("\t%08" PRIx32, opcode);
513     }
514   }
515   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
516                  object::SectionedAddress Address, formatted_raw_ostream &OS,
517                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
518                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
519                  LiveVariablePrinter &LVP) override {
520     if (SP && (PrintSource || PrintLines))
521       SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
522     if (!MI) {
523       printLead(Bytes, Address.Address, OS);
524       OS << " <unknown>";
525       return;
526     }
527     std::string Buffer;
528     {
529       raw_string_ostream TempStream(Buffer);
530       IP.printInst(MI, Address.Address, "", STI, TempStream);
531     }
532     StringRef Contents(Buffer);
533     // Split off bundle attributes
534     auto PacketBundle = Contents.rsplit('\n');
535     // Split off first instruction from the rest
536     auto HeadTail = PacketBundle.first.split('\n');
537     auto Preamble = " { ";
538     auto Separator = "";
539 
540     // Hexagon's packets require relocations to be inline rather than
541     // clustered at the end of the packet.
542     std::vector<RelocationRef>::const_iterator RelCur = Rels->begin();
543     std::vector<RelocationRef>::const_iterator RelEnd = Rels->end();
544     auto PrintReloc = [&]() -> void {
545       while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) {
546         if (RelCur->getOffset() == Address.Address) {
547           printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false);
548           return;
549         }
550         ++RelCur;
551       }
552     };
553 
554     while (!HeadTail.first.empty()) {
555       OS << Separator;
556       Separator = "\n";
557       if (SP && (PrintSource || PrintLines))
558         SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
559       printLead(Bytes, Address.Address, OS);
560       OS << Preamble;
561       Preamble = "   ";
562       StringRef Inst;
563       auto Duplex = HeadTail.first.split('\v');
564       if (!Duplex.second.empty()) {
565         OS << Duplex.first;
566         OS << "; ";
567         Inst = Duplex.second;
568       }
569       else
570         Inst = HeadTail.first;
571       OS << Inst;
572       HeadTail = HeadTail.second.split('\n');
573       if (HeadTail.first.empty())
574         OS << " } " << PacketBundle.second;
575       PrintReloc();
576       Bytes = Bytes.slice(4);
577       Address.Address += 4;
578     }
579   }
580 };
581 HexagonPrettyPrinter HexagonPrettyPrinterInst;
582 
583 class AMDGCNPrettyPrinter : public PrettyPrinter {
584 public:
585   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
586                  object::SectionedAddress Address, formatted_raw_ostream &OS,
587                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
588                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
589                  LiveVariablePrinter &LVP) override {
590     if (SP && (PrintSource || PrintLines))
591       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
592 
593     if (MI) {
594       SmallString<40> InstStr;
595       raw_svector_ostream IS(InstStr);
596 
597       IP.printInst(MI, Address.Address, "", STI, IS);
598 
599       OS << left_justify(IS.str(), 60);
600     } else {
601       // an unrecognized encoding - this is probably data so represent it
602       // using the .long directive, or .byte directive if fewer than 4 bytes
603       // remaining
604       if (Bytes.size() >= 4) {
605         OS << format("\t.long 0x%08" PRIx32 " ",
606                      support::endian::read32<support::little>(Bytes.data()));
607         OS.indent(42);
608       } else {
609           OS << format("\t.byte 0x%02" PRIx8, Bytes[0]);
610           for (unsigned int i = 1; i < Bytes.size(); i++)
611             OS << format(", 0x%02" PRIx8, Bytes[i]);
612           OS.indent(55 - (6 * Bytes.size()));
613       }
614     }
615 
616     OS << format("// %012" PRIX64 ":", Address.Address);
617     if (Bytes.size() >= 4) {
618       // D should be casted to uint32_t here as it is passed by format to
619       // snprintf as vararg.
620       for (uint32_t D : makeArrayRef(
621                reinterpret_cast<const support::little32_t *>(Bytes.data()),
622                Bytes.size() / 4))
623         OS << format(" %08" PRIX32, D);
624     } else {
625       for (unsigned char B : Bytes)
626         OS << format(" %02" PRIX8, B);
627     }
628 
629     if (!Annot.empty())
630       OS << " // " << Annot;
631   }
632 };
633 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst;
634 
635 class BPFPrettyPrinter : public PrettyPrinter {
636 public:
637   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
638                  object::SectionedAddress Address, formatted_raw_ostream &OS,
639                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
640                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
641                  LiveVariablePrinter &LVP) override {
642     if (SP && (PrintSource || PrintLines))
643       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
644     if (LeadingAddr)
645       OS << format("%8" PRId64 ":", Address.Address / 8);
646     if (ShowRawInsn) {
647       OS << "\t";
648       dumpBytes(Bytes, OS);
649     }
650     if (MI)
651       IP.printInst(MI, Address.Address, "", STI, OS);
652     else
653       OS << "\t<unknown>";
654   }
655 };
656 BPFPrettyPrinter BPFPrettyPrinterInst;
657 
658 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) {
659   switch(Triple.getArch()) {
660   default:
661     return PrettyPrinterInst;
662   case Triple::hexagon:
663     return HexagonPrettyPrinterInst;
664   case Triple::amdgcn:
665     return AMDGCNPrettyPrinterInst;
666   case Triple::bpfel:
667   case Triple::bpfeb:
668     return BPFPrettyPrinterInst;
669   }
670 }
671 }
672 
673 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) {
674   assert(Obj->isELF());
675   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
676     return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()),
677                          Obj->getFileName())
678         ->getType();
679   if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
680     return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()),
681                          Obj->getFileName())
682         ->getType();
683   if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
684     return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()),
685                          Obj->getFileName())
686         ->getType();
687   if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
688     return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()),
689                          Obj->getFileName())
690         ->getType();
691   llvm_unreachable("Unsupported binary format");
692 }
693 
694 template <class ELFT> static void
695 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj,
696                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
697   for (auto Symbol : Obj->getDynamicSymbolIterators()) {
698     uint8_t SymbolType = Symbol.getELFType();
699     if (SymbolType == ELF::STT_SECTION)
700       continue;
701 
702     uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName());
703     // ELFSymbolRef::getAddress() returns size instead of value for common
704     // symbols which is not desirable for disassembly output. Overriding.
705     if (SymbolType == ELF::STT_COMMON)
706       Address = unwrapOrError(Obj->getSymbol(Symbol.getRawDataRefImpl()),
707                               Obj->getFileName())
708                     ->st_value;
709 
710     StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName());
711     if (Name.empty())
712       continue;
713 
714     section_iterator SecI =
715         unwrapOrError(Symbol.getSection(), Obj->getFileName());
716     if (SecI == Obj->section_end())
717       continue;
718 
719     AllSymbols[*SecI].emplace_back(Address, Name, SymbolType);
720   }
721 }
722 
723 static void
724 addDynamicElfSymbols(const ObjectFile *Obj,
725                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
726   assert(Obj->isELF());
727   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
728     addDynamicElfSymbols(Elf32LEObj, AllSymbols);
729   else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
730     addDynamicElfSymbols(Elf64LEObj, AllSymbols);
731   else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
732     addDynamicElfSymbols(Elf32BEObj, AllSymbols);
733   else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
734     addDynamicElfSymbols(Elf64BEObj, AllSymbols);
735   else
736     llvm_unreachable("Unsupported binary format");
737 }
738 
739 static void addPltEntries(const ObjectFile *Obj,
740                           std::map<SectionRef, SectionSymbolsTy> &AllSymbols,
741                           StringSaver &Saver) {
742   Optional<SectionRef> Plt = None;
743   for (const SectionRef &Section : Obj->sections()) {
744     Expected<StringRef> SecNameOrErr = Section.getName();
745     if (!SecNameOrErr) {
746       consumeError(SecNameOrErr.takeError());
747       continue;
748     }
749     if (*SecNameOrErr == ".plt")
750       Plt = Section;
751   }
752   if (!Plt)
753     return;
754   if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) {
755     for (auto PltEntry : ElfObj->getPltAddresses()) {
756       if (PltEntry.first) {
757         SymbolRef Symbol(*PltEntry.first, ElfObj);
758         uint8_t SymbolType = getElfSymbolType(Obj, Symbol);
759         if (Expected<StringRef> NameOrErr = Symbol.getName()) {
760           if (!NameOrErr->empty())
761             AllSymbols[*Plt].emplace_back(
762                 PltEntry.second, Saver.save((*NameOrErr + "@plt").str()),
763                 SymbolType);
764           continue;
765         } else {
766           // The warning has been reported in disassembleObject().
767           consumeError(NameOrErr.takeError());
768         }
769       }
770       reportWarning("PLT entry at 0x" + Twine::utohexstr(PltEntry.second) +
771                         " references an invalid symbol",
772                     Obj->getFileName());
773     }
774   }
775 }
776 
777 // Normally the disassembly output will skip blocks of zeroes. This function
778 // returns the number of zero bytes that can be skipped when dumping the
779 // disassembly of the instructions in Buf.
780 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) {
781   // Find the number of leading zeroes.
782   size_t N = 0;
783   while (N < Buf.size() && !Buf[N])
784     ++N;
785 
786   // We may want to skip blocks of zero bytes, but unless we see
787   // at least 8 of them in a row.
788   if (N < 8)
789     return 0;
790 
791   // We skip zeroes in multiples of 4 because do not want to truncate an
792   // instruction if it starts with a zero byte.
793   return N & ~0x3;
794 }
795 
796 // Returns a map from sections to their relocations.
797 static std::map<SectionRef, std::vector<RelocationRef>>
798 getRelocsMap(object::ObjectFile const &Obj) {
799   std::map<SectionRef, std::vector<RelocationRef>> Ret;
800   uint64_t I = (uint64_t)-1;
801   for (SectionRef Sec : Obj.sections()) {
802     ++I;
803     Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection();
804     if (!RelocatedOrErr)
805       reportError(Obj.getFileName(),
806                   "section (" + Twine(I) +
807                       "): failed to get a relocated section: " +
808                       toString(RelocatedOrErr.takeError()));
809 
810     section_iterator Relocated = *RelocatedOrErr;
811     if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep)
812       continue;
813     std::vector<RelocationRef> &V = Ret[*Relocated];
814     append_range(V, Sec.relocations());
815     // Sort relocations by address.
816     llvm::stable_sort(V, isRelocAddressLess);
817   }
818   return Ret;
819 }
820 
821 // Used for --adjust-vma to check if address should be adjusted by the
822 // specified value for a given section.
823 // For ELF we do not adjust non-allocatable sections like debug ones,
824 // because they are not loadable.
825 // TODO: implement for other file formats.
826 static bool shouldAdjustVA(const SectionRef &Section) {
827   const ObjectFile *Obj = Section.getObject();
828   if (Obj->isELF())
829     return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC;
830   return false;
831 }
832 
833 
834 typedef std::pair<uint64_t, char> MappingSymbolPair;
835 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols,
836                                  uint64_t Address) {
837   auto It =
838       partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) {
839         return Val.first <= Address;
840       });
841   // Return zero for any address before the first mapping symbol; this means
842   // we should use the default disassembly mode, depending on the target.
843   if (It == MappingSymbols.begin())
844     return '\x00';
845   return (It - 1)->second;
846 }
847 
848 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index,
849                                uint64_t End, const ObjectFile *Obj,
850                                ArrayRef<uint8_t> Bytes,
851                                ArrayRef<MappingSymbolPair> MappingSymbols,
852                                raw_ostream &OS) {
853   support::endianness Endian =
854       Obj->isLittleEndian() ? support::little : support::big;
855   OS << format("%8" PRIx64 ":\t", SectionAddr + Index);
856   if (Index + 4 <= End) {
857     dumpBytes(Bytes.slice(Index, 4), OS);
858     OS << "\t.word\t"
859            << format_hex(support::endian::read32(Bytes.data() + Index, Endian),
860                          10);
861     return 4;
862   }
863   if (Index + 2 <= End) {
864     dumpBytes(Bytes.slice(Index, 2), OS);
865     OS << "\t\t.short\t"
866            << format_hex(support::endian::read16(Bytes.data() + Index, Endian),
867                          6);
868     return 2;
869   }
870   dumpBytes(Bytes.slice(Index, 1), OS);
871   OS << "\t\t.byte\t" << format_hex(Bytes[0], 4);
872   return 1;
873 }
874 
875 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End,
876                         ArrayRef<uint8_t> Bytes) {
877   // print out data up to 8 bytes at a time in hex and ascii
878   uint8_t AsciiData[9] = {'\0'};
879   uint8_t Byte;
880   int NumBytes = 0;
881 
882   for (; Index < End; ++Index) {
883     if (NumBytes == 0)
884       outs() << format("%8" PRIx64 ":", SectionAddr + Index);
885     Byte = Bytes.slice(Index)[0];
886     outs() << format(" %02x", Byte);
887     AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.';
888 
889     uint8_t IndentOffset = 0;
890     NumBytes++;
891     if (Index == End - 1 || NumBytes > 8) {
892       // Indent the space for less than 8 bytes data.
893       // 2 spaces for byte and one for space between bytes
894       IndentOffset = 3 * (8 - NumBytes);
895       for (int Excess = NumBytes; Excess < 8; Excess++)
896         AsciiData[Excess] = '\0';
897       NumBytes = 8;
898     }
899     if (NumBytes == 8) {
900       AsciiData[8] = '\0';
901       outs() << std::string(IndentOffset, ' ') << "         ";
902       outs() << reinterpret_cast<char *>(AsciiData);
903       outs() << '\n';
904       NumBytes = 0;
905     }
906   }
907 }
908 
909 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile *Obj,
910                                        const SymbolRef &Symbol) {
911   const StringRef FileName = Obj->getFileName();
912   const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName);
913   const StringRef Name = unwrapOrError(Symbol.getName(), FileName);
914 
915   if (Obj->isXCOFF() && SymbolDescription) {
916     const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj);
917     DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl();
918 
919     const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p);
920     Optional<XCOFF::StorageMappingClass> Smc =
921         getXCOFFSymbolCsectSMC(XCOFFObj, Symbol);
922     return SymbolInfoTy(Addr, Name, Smc, SymbolIndex,
923                         isLabel(XCOFFObj, Symbol));
924   } else
925     return SymbolInfoTy(Addr, Name,
926                         Obj->isELF() ? getElfSymbolType(Obj, Symbol)
927                                      : (uint8_t)ELF::STT_NOTYPE);
928 }
929 
930 static SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj,
931                                           const uint64_t Addr, StringRef &Name,
932                                           uint8_t Type) {
933   if (Obj->isXCOFF() && SymbolDescription)
934     return SymbolInfoTy(Addr, Name, None, None, false);
935   else
936     return SymbolInfoTy(Addr, Name, Type);
937 }
938 
939 static void
940 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA,
941                           MCDisassembler *DisAsm, MCInstPrinter *IP,
942                           const MCSubtargetInfo *STI, uint64_t SectionAddr,
943                           uint64_t Start, uint64_t End,
944                           std::unordered_map<uint64_t, std::string> &Labels) {
945   // So far only supports X86.
946   if (!STI->getTargetTriple().isX86())
947     return;
948 
949   Labels.clear();
950   unsigned LabelCount = 0;
951   Start += SectionAddr;
952   End += SectionAddr;
953   uint64_t Index = Start;
954   while (Index < End) {
955     // Disassemble a real instruction and record function-local branch labels.
956     MCInst Inst;
957     uint64_t Size;
958     bool Disassembled = DisAsm->getInstruction(
959         Inst, Size, Bytes.slice(Index - SectionAddr), Index, nulls());
960     if (Size == 0)
961       Size = 1;
962 
963     if (Disassembled && MIA) {
964       uint64_t Target;
965       bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target);
966       if (TargetKnown && (Target >= Start && Target < End) &&
967           !Labels.count(Target))
968         Labels[Target] = ("L" + Twine(LabelCount++)).str();
969     }
970 
971     Index += Size;
972   }
973 }
974 
975 // Create an MCSymbolizer for the target and add it to the MCDisassembler.
976 // This is currently only used on AMDGPU, and assumes the format of the
977 // void * argument passed to AMDGPU's createMCSymbolizer.
978 static void addSymbolizer(
979     MCContext &Ctx, const Target *Target, StringRef TripleName,
980     MCDisassembler *DisAsm, uint64_t SectionAddr, ArrayRef<uint8_t> Bytes,
981     SectionSymbolsTy &Symbols,
982     std::vector<std::unique_ptr<std::string>> &SynthesizedLabelNames) {
983 
984   std::unique_ptr<MCRelocationInfo> RelInfo(
985       Target->createMCRelocationInfo(TripleName, Ctx));
986   if (!RelInfo)
987     return;
988   std::unique_ptr<MCSymbolizer> Symbolizer(Target->createMCSymbolizer(
989       TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
990   MCSymbolizer *SymbolizerPtr = &*Symbolizer;
991   DisAsm->setSymbolizer(std::move(Symbolizer));
992 
993   if (!SymbolizeOperands)
994     return;
995 
996   // Synthesize labels referenced by branch instructions by
997   // disassembling, discarding the output, and collecting the referenced
998   // addresses from the symbolizer.
999   for (size_t Index = 0; Index != Bytes.size();) {
1000     MCInst Inst;
1001     uint64_t Size;
1002     DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), SectionAddr + Index,
1003                            nulls());
1004     if (Size == 0)
1005       Size = 1;
1006     Index += Size;
1007   }
1008   ArrayRef<uint64_t> LabelAddrsRef = SymbolizerPtr->getReferencedAddresses();
1009   // Copy and sort to remove duplicates.
1010   std::vector<uint64_t> LabelAddrs;
1011   LabelAddrs.insert(LabelAddrs.end(), LabelAddrsRef.begin(),
1012                     LabelAddrsRef.end());
1013   llvm::sort(LabelAddrs);
1014   LabelAddrs.resize(std::unique(LabelAddrs.begin(), LabelAddrs.end()) -
1015                     LabelAddrs.begin());
1016   // Add the labels.
1017   for (unsigned LabelNum = 0; LabelNum != LabelAddrs.size(); ++LabelNum) {
1018     std::unique_ptr<std::string> Name(new std::string);
1019     *Name = (Twine("L") + Twine(LabelNum)).str();
1020     SynthesizedLabelNames.push_back(std::move(Name));
1021     Symbols.push_back(SymbolInfoTy(
1022         LabelAddrs[LabelNum], *SynthesizedLabelNames.back(), ELF::STT_NOTYPE));
1023   }
1024   llvm::stable_sort(Symbols);
1025   // Recreate the symbolizer with the new symbols list.
1026   RelInfo.reset(Target->createMCRelocationInfo(TripleName, Ctx));
1027   Symbolizer.reset(Target->createMCSymbolizer(
1028       TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1029   DisAsm->setSymbolizer(std::move(Symbolizer));
1030 }
1031 
1032 static StringRef getSegmentName(const MachOObjectFile *MachO,
1033                                 const SectionRef &Section) {
1034   if (MachO) {
1035     DataRefImpl DR = Section.getRawDataRefImpl();
1036     StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
1037     return SegmentName;
1038   }
1039   return "";
1040 }
1041 
1042 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj,
1043                               MCContext &Ctx, MCDisassembler *PrimaryDisAsm,
1044                               MCDisassembler *SecondaryDisAsm,
1045                               const MCInstrAnalysis *MIA, MCInstPrinter *IP,
1046                               const MCSubtargetInfo *PrimarySTI,
1047                               const MCSubtargetInfo *SecondarySTI,
1048                               PrettyPrinter &PIP,
1049                               SourcePrinter &SP, bool InlineRelocs) {
1050   const MCSubtargetInfo *STI = PrimarySTI;
1051   MCDisassembler *DisAsm = PrimaryDisAsm;
1052   bool PrimaryIsThumb = false;
1053   if (isArmElf(Obj))
1054     PrimaryIsThumb = STI->checkFeatures("+thumb-mode");
1055 
1056   std::map<SectionRef, std::vector<RelocationRef>> RelocMap;
1057   if (InlineRelocs)
1058     RelocMap = getRelocsMap(*Obj);
1059   bool Is64Bits = Obj->getBytesInAddress() > 4;
1060 
1061   // Create a mapping from virtual address to symbol name.  This is used to
1062   // pretty print the symbols while disassembling.
1063   std::map<SectionRef, SectionSymbolsTy> AllSymbols;
1064   SectionSymbolsTy AbsoluteSymbols;
1065   const StringRef FileName = Obj->getFileName();
1066   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
1067   for (const SymbolRef &Symbol : Obj->symbols()) {
1068     Expected<StringRef> NameOrErr = Symbol.getName();
1069     if (!NameOrErr) {
1070       reportWarning(toString(NameOrErr.takeError()), FileName);
1071       continue;
1072     }
1073     if (NameOrErr->empty() && !(Obj->isXCOFF() && SymbolDescription))
1074       continue;
1075 
1076     if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION)
1077       continue;
1078 
1079     // Don't ask a Mach-O STAB symbol for its section unless you know that
1080     // STAB symbol's section field refers to a valid section index. Otherwise
1081     // the symbol may error trying to load a section that does not exist.
1082     if (MachO) {
1083       DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1084       uint8_t NType = (MachO->is64Bit() ?
1085                        MachO->getSymbol64TableEntry(SymDRI).n_type:
1086                        MachO->getSymbolTableEntry(SymDRI).n_type);
1087       if (NType & MachO::N_STAB)
1088         continue;
1089     }
1090 
1091     section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1092     if (SecI != Obj->section_end())
1093       AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol));
1094     else
1095       AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol));
1096   }
1097 
1098   if (AllSymbols.empty() && Obj->isELF())
1099     addDynamicElfSymbols(Obj, AllSymbols);
1100 
1101   BumpPtrAllocator A;
1102   StringSaver Saver(A);
1103   addPltEntries(Obj, AllSymbols, Saver);
1104 
1105   // Create a mapping from virtual address to section. An empty section can
1106   // cause more than one section at the same address. Sort such sections to be
1107   // before same-addressed non-empty sections so that symbol lookups prefer the
1108   // non-empty section.
1109   std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses;
1110   for (SectionRef Sec : Obj->sections())
1111     SectionAddresses.emplace_back(Sec.getAddress(), Sec);
1112   llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) {
1113     if (LHS.first != RHS.first)
1114       return LHS.first < RHS.first;
1115     return LHS.second.getSize() < RHS.second.getSize();
1116   });
1117 
1118   // Linked executables (.exe and .dll files) typically don't include a real
1119   // symbol table but they might contain an export table.
1120   if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) {
1121     for (const auto &ExportEntry : COFFObj->export_directories()) {
1122       StringRef Name;
1123       if (Error E = ExportEntry.getSymbolName(Name))
1124         reportError(std::move(E), Obj->getFileName());
1125       if (Name.empty())
1126         continue;
1127 
1128       uint32_t RVA;
1129       if (Error E = ExportEntry.getExportRVA(RVA))
1130         reportError(std::move(E), Obj->getFileName());
1131 
1132       uint64_t VA = COFFObj->getImageBase() + RVA;
1133       auto Sec = partition_point(
1134           SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) {
1135             return O.first <= VA;
1136           });
1137       if (Sec != SectionAddresses.begin()) {
1138         --Sec;
1139         AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE);
1140       } else
1141         AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE);
1142     }
1143   }
1144 
1145   // Sort all the symbols, this allows us to use a simple binary search to find
1146   // Multiple symbols can have the same address. Use a stable sort to stabilize
1147   // the output.
1148   StringSet<> FoundDisasmSymbolSet;
1149   for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols)
1150     llvm::stable_sort(SecSyms.second);
1151   llvm::stable_sort(AbsoluteSymbols);
1152 
1153   std::unique_ptr<DWARFContext> DICtx;
1154   LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI);
1155 
1156   if (DbgVariables != DVDisabled) {
1157     DICtx = DWARFContext::create(*Obj);
1158     for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units())
1159       LVP.addCompileUnit(CU->getUnitDIE(false));
1160   }
1161 
1162   LLVM_DEBUG(LVP.dump());
1163 
1164   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1165     if (FilterSections.empty() && !DisassembleAll &&
1166         (!Section.isText() || Section.isVirtual()))
1167       continue;
1168 
1169     uint64_t SectionAddr = Section.getAddress();
1170     uint64_t SectSize = Section.getSize();
1171     if (!SectSize)
1172       continue;
1173 
1174     // Get the list of all the symbols in this section.
1175     SectionSymbolsTy &Symbols = AllSymbols[Section];
1176     std::vector<MappingSymbolPair> MappingSymbols;
1177     if (hasMappingSymbols(Obj)) {
1178       for (const auto &Symb : Symbols) {
1179         uint64_t Address = Symb.Addr;
1180         StringRef Name = Symb.Name;
1181         if (Name.startswith("$d"))
1182           MappingSymbols.emplace_back(Address - SectionAddr, 'd');
1183         if (Name.startswith("$x"))
1184           MappingSymbols.emplace_back(Address - SectionAddr, 'x');
1185         if (Name.startswith("$a"))
1186           MappingSymbols.emplace_back(Address - SectionAddr, 'a');
1187         if (Name.startswith("$t"))
1188           MappingSymbols.emplace_back(Address - SectionAddr, 't');
1189       }
1190     }
1191 
1192     llvm::sort(MappingSymbols);
1193 
1194     ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(
1195         unwrapOrError(Section.getContents(), Obj->getFileName()));
1196 
1197     std::vector<std::unique_ptr<std::string>> SynthesizedLabelNames;
1198     if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) {
1199       // AMDGPU disassembler uses symbolizer for printing labels
1200       addSymbolizer(Ctx, TheTarget, TripleName, DisAsm, SectionAddr, Bytes,
1201                     Symbols, SynthesizedLabelNames);
1202     }
1203 
1204     StringRef SegmentName = getSegmentName(MachO, Section);
1205     StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName());
1206     // If the section has no symbol at the start, just insert a dummy one.
1207     if (Symbols.empty() || Symbols[0].Addr != 0) {
1208       Symbols.insert(Symbols.begin(),
1209                      createDummySymbolInfo(Obj, SectionAddr, SectionName,
1210                                            Section.isText() ? ELF::STT_FUNC
1211                                                             : ELF::STT_OBJECT));
1212     }
1213 
1214     SmallString<40> Comments;
1215     raw_svector_ostream CommentStream(Comments);
1216 
1217     uint64_t VMAAdjustment = 0;
1218     if (shouldAdjustVA(Section))
1219       VMAAdjustment = AdjustVMA;
1220 
1221     uint64_t Size;
1222     uint64_t Index;
1223     bool PrintedSection = false;
1224     std::vector<RelocationRef> Rels = RelocMap[Section];
1225     std::vector<RelocationRef>::const_iterator RelCur = Rels.begin();
1226     std::vector<RelocationRef>::const_iterator RelEnd = Rels.end();
1227     // Disassemble symbol by symbol.
1228     for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) {
1229       std::string SymbolName = Symbols[SI].Name.str();
1230       if (Demangle)
1231         SymbolName = demangle(SymbolName);
1232 
1233       // Skip if --disassemble-symbols is not empty and the symbol is not in
1234       // the list.
1235       if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName))
1236         continue;
1237 
1238       uint64_t Start = Symbols[SI].Addr;
1239       if (Start < SectionAddr || StopAddress <= Start)
1240         continue;
1241       else
1242         FoundDisasmSymbolSet.insert(SymbolName);
1243 
1244       // The end is the section end, the beginning of the next symbol, or
1245       // --stop-address.
1246       uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress);
1247       if (SI + 1 < SE)
1248         End = std::min(End, Symbols[SI + 1].Addr);
1249       if (Start >= End || End <= StartAddress)
1250         continue;
1251       Start -= SectionAddr;
1252       End -= SectionAddr;
1253 
1254       if (!PrintedSection) {
1255         PrintedSection = true;
1256         outs() << "\nDisassembly of section ";
1257         if (!SegmentName.empty())
1258           outs() << SegmentName << ",";
1259         outs() << SectionName << ":\n";
1260       }
1261 
1262       outs() << '\n';
1263       if (LeadingAddr)
1264         outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ",
1265                          SectionAddr + Start + VMAAdjustment);
1266       if (Obj->isXCOFF() && SymbolDescription) {
1267         outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n";
1268       } else
1269         outs() << '<' << SymbolName << ">:\n";
1270 
1271       // Don't print raw contents of a virtual section. A virtual section
1272       // doesn't have any contents in the file.
1273       if (Section.isVirtual()) {
1274         outs() << "...\n";
1275         continue;
1276       }
1277 
1278       auto Status = DisAsm->onSymbolStart(Symbols[SI], Size,
1279                                           Bytes.slice(Start, End - Start),
1280                                           SectionAddr + Start, CommentStream);
1281       // To have round trippable disassembly, we fall back to decoding the
1282       // remaining bytes as instructions.
1283       //
1284       // If there is a failure, we disassemble the failed region as bytes before
1285       // falling back. The target is expected to print nothing in this case.
1286       //
1287       // If there is Success or SoftFail i.e no 'real' failure, we go ahead by
1288       // Size bytes before falling back.
1289       // So if the entire symbol is 'eaten' by the target:
1290       //   Start += Size  // Now Start = End and we will never decode as
1291       //                  // instructions
1292       //
1293       // Right now, most targets return None i.e ignore to treat a symbol
1294       // separately. But WebAssembly decodes preludes for some symbols.
1295       //
1296       if (Status.hasValue()) {
1297         if (Status.getValue() == MCDisassembler::Fail) {
1298           outs() << "// Error in decoding " << SymbolName
1299                  << " : Decoding failed region as bytes.\n";
1300           for (uint64_t I = 0; I < Size; ++I) {
1301             outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true)
1302                    << "\n";
1303           }
1304         }
1305       } else {
1306         Size = 0;
1307       }
1308 
1309       Start += Size;
1310 
1311       Index = Start;
1312       if (SectionAddr < StartAddress)
1313         Index = std::max<uint64_t>(Index, StartAddress - SectionAddr);
1314 
1315       // If there is a data/common symbol inside an ELF text section and we are
1316       // only disassembling text (applicable all architectures), we are in a
1317       // situation where we must print the data and not disassemble it.
1318       if (Obj->isELF() && !DisassembleAll && Section.isText()) {
1319         uint8_t SymTy = Symbols[SI].Type;
1320         if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) {
1321           dumpELFData(SectionAddr, Index, End, Bytes);
1322           Index = End;
1323         }
1324       }
1325 
1326       bool CheckARMELFData = hasMappingSymbols(Obj) &&
1327                              Symbols[SI].Type != ELF::STT_OBJECT &&
1328                              !DisassembleAll;
1329       bool DumpARMELFData = false;
1330       formatted_raw_ostream FOS(outs());
1331 
1332       std::unordered_map<uint64_t, std::string> AllLabels;
1333       if (SymbolizeOperands)
1334         collectLocalBranchTargets(Bytes, MIA, DisAsm, IP, PrimarySTI,
1335                                   SectionAddr, Index, End, AllLabels);
1336 
1337       while (Index < End) {
1338         // ARM and AArch64 ELF binaries can interleave data and text in the
1339         // same section. We rely on the markers introduced to understand what
1340         // we need to dump. If the data marker is within a function, it is
1341         // denoted as a word/short etc.
1342         if (CheckARMELFData) {
1343           char Kind = getMappingSymbolKind(MappingSymbols, Index);
1344           DumpARMELFData = Kind == 'd';
1345           if (SecondarySTI) {
1346             if (Kind == 'a') {
1347               STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI;
1348               DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm;
1349             } else if (Kind == 't') {
1350               STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI;
1351               DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm;
1352             }
1353           }
1354         }
1355 
1356         if (DumpARMELFData) {
1357           Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes,
1358                                 MappingSymbols, FOS);
1359         } else {
1360           // When -z or --disassemble-zeroes are given we always dissasemble
1361           // them. Otherwise we might want to skip zero bytes we see.
1362           if (!DisassembleZeroes) {
1363             uint64_t MaxOffset = End - Index;
1364             // For --reloc: print zero blocks patched by relocations, so that
1365             // relocations can be shown in the dump.
1366             if (RelCur != RelEnd)
1367               MaxOffset = RelCur->getOffset() - Index;
1368 
1369             if (size_t N =
1370                     countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) {
1371               FOS << "\t\t..." << '\n';
1372               Index += N;
1373               continue;
1374             }
1375           }
1376 
1377           // Print local label if there's any.
1378           auto Iter = AllLabels.find(SectionAddr + Index);
1379           if (Iter != AllLabels.end())
1380             FOS << "<" << Iter->second << ">:\n";
1381 
1382           // Disassemble a real instruction or a data when disassemble all is
1383           // provided
1384           MCInst Inst;
1385           bool Disassembled =
1386               DisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
1387                                      SectionAddr + Index, CommentStream);
1388           if (Size == 0)
1389             Size = 1;
1390 
1391           LVP.update({Index, Section.getIndex()},
1392                      {Index + Size, Section.getIndex()}, Index + Size != End);
1393 
1394           PIP.printInst(
1395               *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size),
1396               {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS,
1397               "", *STI, &SP, Obj->getFileName(), &Rels, LVP);
1398           FOS << CommentStream.str();
1399           Comments.clear();
1400 
1401           // If disassembly has failed, avoid analysing invalid/incomplete
1402           // instruction information. Otherwise, try to resolve the target
1403           // address (jump target or memory operand address) and print it on the
1404           // right of the instruction.
1405           if (Disassembled && MIA) {
1406             uint64_t Target;
1407             bool PrintTarget =
1408                 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target);
1409             if (!PrintTarget)
1410               if (Optional<uint64_t> MaybeTarget =
1411                       MIA->evaluateMemoryOperandAddress(
1412                           Inst, SectionAddr + Index, Size)) {
1413                 Target = *MaybeTarget;
1414                 PrintTarget = true;
1415                 // Do not print real address when symbolizing.
1416                 if (!SymbolizeOperands)
1417                   FOS << "  # " << Twine::utohexstr(Target);
1418               }
1419             if (PrintTarget) {
1420               // In a relocatable object, the target's section must reside in
1421               // the same section as the call instruction or it is accessed
1422               // through a relocation.
1423               //
1424               // In a non-relocatable object, the target may be in any section.
1425               // In that case, locate the section(s) containing the target
1426               // address and find the symbol in one of those, if possible.
1427               //
1428               // N.B. We don't walk the relocations in the relocatable case yet.
1429               std::vector<const SectionSymbolsTy *> TargetSectionSymbols;
1430               if (!Obj->isRelocatableObject()) {
1431                 auto It = llvm::partition_point(
1432                     SectionAddresses,
1433                     [=](const std::pair<uint64_t, SectionRef> &O) {
1434                       return O.first <= Target;
1435                     });
1436                 uint64_t TargetSecAddr = 0;
1437                 while (It != SectionAddresses.begin()) {
1438                   --It;
1439                   if (TargetSecAddr == 0)
1440                     TargetSecAddr = It->first;
1441                   if (It->first != TargetSecAddr)
1442                     break;
1443                   TargetSectionSymbols.push_back(&AllSymbols[It->second]);
1444                 }
1445               } else {
1446                 TargetSectionSymbols.push_back(&Symbols);
1447               }
1448               TargetSectionSymbols.push_back(&AbsoluteSymbols);
1449 
1450               // Find the last symbol in the first candidate section whose
1451               // offset is less than or equal to the target. If there are no
1452               // such symbols, try in the next section and so on, before finally
1453               // using the nearest preceding absolute symbol (if any), if there
1454               // are no other valid symbols.
1455               const SymbolInfoTy *TargetSym = nullptr;
1456               for (const SectionSymbolsTy *TargetSymbols :
1457                    TargetSectionSymbols) {
1458                 auto It = llvm::partition_point(
1459                     *TargetSymbols,
1460                     [=](const SymbolInfoTy &O) { return O.Addr <= Target; });
1461                 if (It != TargetSymbols->begin()) {
1462                   TargetSym = &*(It - 1);
1463                   break;
1464                 }
1465               }
1466 
1467               // Print the labels corresponding to the target if there's any.
1468               bool LabelAvailable = AllLabels.count(Target);
1469               if (TargetSym != nullptr) {
1470                 uint64_t TargetAddress = TargetSym->Addr;
1471                 uint64_t Disp = Target - TargetAddress;
1472                 std::string TargetName = TargetSym->Name.str();
1473                 if (Demangle)
1474                   TargetName = demangle(TargetName);
1475 
1476                 FOS << " <";
1477                 if (!Disp) {
1478                   // Always Print the binary symbol precisely corresponding to
1479                   // the target address.
1480                   FOS << TargetName;
1481                 } else if (!LabelAvailable) {
1482                   // Always Print the binary symbol plus an offset if there's no
1483                   // local label corresponding to the target address.
1484                   FOS << TargetName << "+0x" << Twine::utohexstr(Disp);
1485                 } else {
1486                   FOS << AllLabels[Target];
1487                 }
1488                 FOS << ">";
1489               } else if (LabelAvailable) {
1490                 FOS << " <" << AllLabels[Target] << ">";
1491               }
1492             }
1493           }
1494         }
1495 
1496         LVP.printAfterInst(FOS);
1497         FOS << "\n";
1498 
1499         // Hexagon does this in pretty printer
1500         if (Obj->getArch() != Triple::hexagon) {
1501           // Print relocation for instruction and data.
1502           while (RelCur != RelEnd) {
1503             uint64_t Offset = RelCur->getOffset();
1504             // If this relocation is hidden, skip it.
1505             if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) {
1506               ++RelCur;
1507               continue;
1508             }
1509 
1510             // Stop when RelCur's offset is past the disassembled
1511             // instruction/data. Note that it's possible the disassembled data
1512             // is not the complete data: we might see the relocation printed in
1513             // the middle of the data, but this matches the binutils objdump
1514             // output.
1515             if (Offset >= Index + Size)
1516               break;
1517 
1518             // When --adjust-vma is used, update the address printed.
1519             if (RelCur->getSymbol() != Obj->symbol_end()) {
1520               Expected<section_iterator> SymSI =
1521                   RelCur->getSymbol()->getSection();
1522               if (SymSI && *SymSI != Obj->section_end() &&
1523                   shouldAdjustVA(**SymSI))
1524                 Offset += AdjustVMA;
1525             }
1526 
1527             printRelocation(FOS, Obj->getFileName(), *RelCur,
1528                             SectionAddr + Offset, Is64Bits);
1529             LVP.printAfterOtherLine(FOS, true);
1530             ++RelCur;
1531           }
1532         }
1533 
1534         Index += Size;
1535       }
1536     }
1537   }
1538   StringSet<> MissingDisasmSymbolSet =
1539       set_difference(DisasmSymbolSet, FoundDisasmSymbolSet);
1540   for (StringRef Sym : MissingDisasmSymbolSet.keys())
1541     reportWarning("failed to disassemble missing symbol " + Sym, FileName);
1542 }
1543 
1544 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) {
1545   const Target *TheTarget = getTarget(Obj);
1546 
1547   // Package up features to be passed to target/subtarget
1548   SubtargetFeatures Features = Obj->getFeatures();
1549   if (!MAttrs.empty())
1550     for (unsigned I = 0; I != MAttrs.size(); ++I)
1551       Features.AddFeature(MAttrs[I]);
1552 
1553   std::unique_ptr<const MCRegisterInfo> MRI(
1554       TheTarget->createMCRegInfo(TripleName));
1555   if (!MRI)
1556     reportError(Obj->getFileName(),
1557                 "no register info for target " + TripleName);
1558 
1559   // Set up disassembler.
1560   MCTargetOptions MCOptions;
1561   std::unique_ptr<const MCAsmInfo> AsmInfo(
1562       TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions));
1563   if (!AsmInfo)
1564     reportError(Obj->getFileName(),
1565                 "no assembly info for target " + TripleName);
1566 
1567   if (MCPU.empty())
1568     MCPU = Obj->tryGetCPUName().getValueOr("").str();
1569 
1570   std::unique_ptr<const MCSubtargetInfo> STI(
1571       TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString()));
1572   if (!STI)
1573     reportError(Obj->getFileName(),
1574                 "no subtarget info for target " + TripleName);
1575   std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo());
1576   if (!MII)
1577     reportError(Obj->getFileName(),
1578                 "no instruction info for target " + TripleName);
1579   MCObjectFileInfo MOFI;
1580   MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI);
1581   // FIXME: for now initialize MCObjectFileInfo with default values
1582   MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx);
1583 
1584   std::unique_ptr<MCDisassembler> DisAsm(
1585       TheTarget->createMCDisassembler(*STI, Ctx));
1586   if (!DisAsm)
1587     reportError(Obj->getFileName(), "no disassembler for target " + TripleName);
1588 
1589   // If we have an ARM object file, we need a second disassembler, because
1590   // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode.
1591   // We use mapping symbols to switch between the two assemblers, where
1592   // appropriate.
1593   std::unique_ptr<MCDisassembler> SecondaryDisAsm;
1594   std::unique_ptr<const MCSubtargetInfo> SecondarySTI;
1595   if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) {
1596     if (STI->checkFeatures("+thumb-mode"))
1597       Features.AddFeature("-thumb-mode");
1598     else
1599       Features.AddFeature("+thumb-mode");
1600     SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU,
1601                                                         Features.getString()));
1602     SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx));
1603   }
1604 
1605   std::unique_ptr<const MCInstrAnalysis> MIA(
1606       TheTarget->createMCInstrAnalysis(MII.get()));
1607 
1608   int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1609   std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
1610       Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI));
1611   if (!IP)
1612     reportError(Obj->getFileName(),
1613                 "no instruction printer for target " + TripleName);
1614   IP->setPrintImmHex(PrintImmHex);
1615   IP->setPrintBranchImmAsAddress(true);
1616   IP->setSymbolizeOperands(SymbolizeOperands);
1617   IP->setMCInstrAnalysis(MIA.get());
1618 
1619   PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName));
1620   SourcePrinter SP(Obj, TheTarget->getName());
1621 
1622   for (StringRef Opt : DisassemblerOptions)
1623     if (!IP->applyTargetSpecificCLOption(Opt))
1624       reportError(Obj->getFileName(),
1625                   "Unrecognized disassembler option: " + Opt);
1626 
1627   disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(),
1628                     MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP,
1629                     SP, InlineRelocs);
1630 }
1631 
1632 void objdump::printRelocations(const ObjectFile *Obj) {
1633   StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 :
1634                                                  "%08" PRIx64;
1635   // Regular objdump doesn't print relocations in non-relocatable object
1636   // files.
1637   if (!Obj->isRelocatableObject())
1638     return;
1639 
1640   // Build a mapping from relocation target to a vector of relocation
1641   // sections. Usually, there is an only one relocation section for
1642   // each relocated section.
1643   MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec;
1644   uint64_t Ndx;
1645   for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) {
1646     if (Section.relocation_begin() == Section.relocation_end())
1647       continue;
1648     Expected<section_iterator> SecOrErr = Section.getRelocatedSection();
1649     if (!SecOrErr)
1650       reportError(Obj->getFileName(),
1651                   "section (" + Twine(Ndx) +
1652                       "): unable to get a relocation target: " +
1653                       toString(SecOrErr.takeError()));
1654     SecToRelSec[**SecOrErr].push_back(Section);
1655   }
1656 
1657   for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) {
1658     StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName());
1659     outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n";
1660     uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8);
1661     uint32_t TypePadding = 24;
1662     outs() << left_justify("OFFSET", OffsetPadding) << " "
1663            << left_justify("TYPE", TypePadding) << " "
1664            << "VALUE\n";
1665 
1666     for (SectionRef Section : P.second) {
1667       for (const RelocationRef &Reloc : Section.relocations()) {
1668         uint64_t Address = Reloc.getOffset();
1669         SmallString<32> RelocName;
1670         SmallString<32> ValueStr;
1671         if (Address < StartAddress || Address > StopAddress || getHidden(Reloc))
1672           continue;
1673         Reloc.getTypeName(RelocName);
1674         if (Error E = getRelocationValueString(Reloc, ValueStr))
1675           reportError(std::move(E), Obj->getFileName());
1676 
1677         outs() << format(Fmt.data(), Address) << " "
1678                << left_justify(RelocName, TypePadding) << " " << ValueStr
1679                << "\n";
1680       }
1681     }
1682     outs() << "\n";
1683   }
1684 }
1685 
1686 void objdump::printDynamicRelocations(const ObjectFile *Obj) {
1687   // For the moment, this option is for ELF only
1688   if (!Obj->isELF())
1689     return;
1690 
1691   const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
1692   if (!Elf || Elf->getEType() != ELF::ET_DYN) {
1693     reportError(Obj->getFileName(), "not a dynamic object");
1694     return;
1695   }
1696 
1697   std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections();
1698   if (DynRelSec.empty())
1699     return;
1700 
1701   outs() << "DYNAMIC RELOCATION RECORDS\n";
1702   StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1703   for (const SectionRef &Section : DynRelSec)
1704     for (const RelocationRef &Reloc : Section.relocations()) {
1705       uint64_t Address = Reloc.getOffset();
1706       SmallString<32> RelocName;
1707       SmallString<32> ValueStr;
1708       Reloc.getTypeName(RelocName);
1709       if (Error E = getRelocationValueString(Reloc, ValueStr))
1710         reportError(std::move(E), Obj->getFileName());
1711       outs() << format(Fmt.data(), Address) << " " << RelocName << " "
1712              << ValueStr << "\n";
1713     }
1714 }
1715 
1716 // Returns true if we need to show LMA column when dumping section headers. We
1717 // show it only when the platform is ELF and either we have at least one section
1718 // whose VMA and LMA are different and/or when --show-lma flag is used.
1719 static bool shouldDisplayLMA(const ObjectFile *Obj) {
1720   if (!Obj->isELF())
1721     return false;
1722   for (const SectionRef &S : ToolSectionFilter(*Obj))
1723     if (S.getAddress() != getELFSectionLMA(S))
1724       return true;
1725   return ShowLMA;
1726 }
1727 
1728 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) {
1729   // Default column width for names is 13 even if no names are that long.
1730   size_t MaxWidth = 13;
1731   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1732     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1733     MaxWidth = std::max(MaxWidth, Name.size());
1734   }
1735   return MaxWidth;
1736 }
1737 
1738 void objdump::printSectionHeaders(const ObjectFile *Obj) {
1739   size_t NameWidth = getMaxSectionNameWidth(Obj);
1740   size_t AddressWidth = 2 * Obj->getBytesInAddress();
1741   bool HasLMAColumn = shouldDisplayLMA(Obj);
1742   if (HasLMAColumn)
1743     outs() << "Sections:\n"
1744               "Idx "
1745            << left_justify("Name", NameWidth) << " Size     "
1746            << left_justify("VMA", AddressWidth) << " "
1747            << left_justify("LMA", AddressWidth) << " Type\n";
1748   else
1749     outs() << "Sections:\n"
1750               "Idx "
1751            << left_justify("Name", NameWidth) << " Size     "
1752            << left_justify("VMA", AddressWidth) << " Type\n";
1753 
1754   uint64_t Idx;
1755   for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) {
1756     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1757     uint64_t VMA = Section.getAddress();
1758     if (shouldAdjustVA(Section))
1759       VMA += AdjustVMA;
1760 
1761     uint64_t Size = Section.getSize();
1762 
1763     std::string Type = Section.isText() ? "TEXT" : "";
1764     if (Section.isData())
1765       Type += Type.empty() ? "DATA" : " DATA";
1766     if (Section.isBSS())
1767       Type += Type.empty() ? "BSS" : " BSS";
1768 
1769     if (HasLMAColumn)
1770       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1771                        Name.str().c_str(), Size)
1772              << format_hex_no_prefix(VMA, AddressWidth) << " "
1773              << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth)
1774              << " " << Type << "\n";
1775     else
1776       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1777                        Name.str().c_str(), Size)
1778              << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n";
1779   }
1780   outs() << "\n";
1781 }
1782 
1783 void objdump::printSectionContents(const ObjectFile *Obj) {
1784   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
1785 
1786   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1787     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1788     uint64_t BaseAddr = Section.getAddress();
1789     uint64_t Size = Section.getSize();
1790     if (!Size)
1791       continue;
1792 
1793     outs() << "Contents of section ";
1794     StringRef SegmentName = getSegmentName(MachO, Section);
1795     if (!SegmentName.empty())
1796       outs() << SegmentName << ",";
1797     outs() << Name << ":\n";
1798     if (Section.isBSS()) {
1799       outs() << format("<skipping contents of bss section at [%04" PRIx64
1800                        ", %04" PRIx64 ")>\n",
1801                        BaseAddr, BaseAddr + Size);
1802       continue;
1803     }
1804 
1805     StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName());
1806 
1807     // Dump out the content as hex and printable ascii characters.
1808     for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) {
1809       outs() << format(" %04" PRIx64 " ", BaseAddr + Addr);
1810       // Dump line of hex.
1811       for (std::size_t I = 0; I < 16; ++I) {
1812         if (I != 0 && I % 4 == 0)
1813           outs() << ' ';
1814         if (Addr + I < End)
1815           outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true)
1816                  << hexdigit(Contents[Addr + I] & 0xF, true);
1817         else
1818           outs() << "  ";
1819       }
1820       // Print ascii.
1821       outs() << "  ";
1822       for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) {
1823         if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF))
1824           outs() << Contents[Addr + I];
1825         else
1826           outs() << ".";
1827       }
1828       outs() << "\n";
1829     }
1830   }
1831 }
1832 
1833 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName,
1834                                StringRef ArchitectureName, bool DumpDynamic) {
1835   if (O->isCOFF() && !DumpDynamic) {
1836     outs() << "SYMBOL TABLE:\n";
1837     printCOFFSymbolTable(cast<const COFFObjectFile>(O));
1838     return;
1839   }
1840 
1841   const StringRef FileName = O->getFileName();
1842 
1843   if (!DumpDynamic) {
1844     outs() << "SYMBOL TABLE:\n";
1845     for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I)
1846       printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic);
1847     return;
1848   }
1849 
1850   outs() << "DYNAMIC SYMBOL TABLE:\n";
1851   if (!O->isELF()) {
1852     reportWarning(
1853         "this operation is not currently supported for this file format",
1854         FileName);
1855     return;
1856   }
1857 
1858   const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O);
1859   for (auto I = ELF->getDynamicSymbolIterators().begin();
1860        I != ELF->getDynamicSymbolIterators().end(); ++I)
1861     printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic);
1862 }
1863 
1864 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol,
1865                           StringRef FileName, StringRef ArchiveName,
1866                           StringRef ArchitectureName, bool DumpDynamic) {
1867   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O);
1868   uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName,
1869                                    ArchitectureName);
1870   if ((Address < StartAddress) || (Address > StopAddress))
1871     return;
1872   SymbolRef::Type Type =
1873       unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName);
1874   uint32_t Flags =
1875       unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName);
1876 
1877   // Don't ask a Mach-O STAB symbol for its section unless you know that
1878   // STAB symbol's section field refers to a valid section index. Otherwise
1879   // the symbol may error trying to load a section that does not exist.
1880   bool IsSTAB = false;
1881   if (MachO) {
1882     DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1883     uint8_t NType =
1884         (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type
1885                           : MachO->getSymbolTableEntry(SymDRI).n_type);
1886     if (NType & MachO::N_STAB)
1887       IsSTAB = true;
1888   }
1889   section_iterator Section = IsSTAB
1890                                  ? O->section_end()
1891                                  : unwrapOrError(Symbol.getSection(), FileName,
1892                                                  ArchiveName, ArchitectureName);
1893 
1894   StringRef Name;
1895   if (Type == SymbolRef::ST_Debug && Section != O->section_end()) {
1896     if (Expected<StringRef> NameOrErr = Section->getName())
1897       Name = *NameOrErr;
1898     else
1899       consumeError(NameOrErr.takeError());
1900 
1901   } else {
1902     Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName,
1903                          ArchitectureName);
1904   }
1905 
1906   bool Global = Flags & SymbolRef::SF_Global;
1907   bool Weak = Flags & SymbolRef::SF_Weak;
1908   bool Absolute = Flags & SymbolRef::SF_Absolute;
1909   bool Common = Flags & SymbolRef::SF_Common;
1910   bool Hidden = Flags & SymbolRef::SF_Hidden;
1911 
1912   char GlobLoc = ' ';
1913   if ((Section != O->section_end() || Absolute) && !Weak)
1914     GlobLoc = Global ? 'g' : 'l';
1915   char IFunc = ' ';
1916   if (O->isELF()) {
1917     if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC)
1918       IFunc = 'i';
1919     if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE)
1920       GlobLoc = 'u';
1921   }
1922 
1923   char Debug = ' ';
1924   if (DumpDynamic)
1925     Debug = 'D';
1926   else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
1927     Debug = 'd';
1928 
1929   char FileFunc = ' ';
1930   if (Type == SymbolRef::ST_File)
1931     FileFunc = 'f';
1932   else if (Type == SymbolRef::ST_Function)
1933     FileFunc = 'F';
1934   else if (Type == SymbolRef::ST_Data)
1935     FileFunc = 'O';
1936 
1937   const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1938 
1939   outs() << format(Fmt, Address) << " "
1940          << GlobLoc            // Local -> 'l', Global -> 'g', Neither -> ' '
1941          << (Weak ? 'w' : ' ') // Weak?
1942          << ' '                // Constructor. Not supported yet.
1943          << ' '                // Warning. Not supported yet.
1944          << IFunc              // Indirect reference to another symbol.
1945          << Debug              // Debugging (d) or dynamic (D) symbol.
1946          << FileFunc           // Name of function (F), file (f) or object (O).
1947          << ' ';
1948   if (Absolute) {
1949     outs() << "*ABS*";
1950   } else if (Common) {
1951     outs() << "*COM*";
1952   } else if (Section == O->section_end()) {
1953     outs() << "*UND*";
1954   } else {
1955     StringRef SegmentName = getSegmentName(MachO, *Section);
1956     if (!SegmentName.empty())
1957       outs() << SegmentName << ",";
1958     StringRef SectionName = unwrapOrError(Section->getName(), FileName);
1959     outs() << SectionName;
1960   }
1961 
1962   if (Common || O->isELF()) {
1963     uint64_t Val =
1964         Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize();
1965     outs() << '\t' << format(Fmt, Val);
1966   }
1967 
1968   if (O->isELF()) {
1969     uint8_t Other = ELFSymbolRef(Symbol).getOther();
1970     switch (Other) {
1971     case ELF::STV_DEFAULT:
1972       break;
1973     case ELF::STV_INTERNAL:
1974       outs() << " .internal";
1975       break;
1976     case ELF::STV_HIDDEN:
1977       outs() << " .hidden";
1978       break;
1979     case ELF::STV_PROTECTED:
1980       outs() << " .protected";
1981       break;
1982     default:
1983       outs() << format(" 0x%02x", Other);
1984       break;
1985     }
1986   } else if (Hidden) {
1987     outs() << " .hidden";
1988   }
1989 
1990   if (Demangle)
1991     outs() << ' ' << demangle(std::string(Name)) << '\n';
1992   else
1993     outs() << ' ' << Name << '\n';
1994 }
1995 
1996 static void printUnwindInfo(const ObjectFile *O) {
1997   outs() << "Unwind info:\n\n";
1998 
1999   if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O))
2000     printCOFFUnwindInfo(Coff);
2001   else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O))
2002     printMachOUnwindInfo(MachO);
2003   else
2004     // TODO: Extract DWARF dump tool to objdump.
2005     WithColor::error(errs(), ToolName)
2006         << "This operation is only currently supported "
2007            "for COFF and MachO object files.\n";
2008 }
2009 
2010 /// Dump the raw contents of the __clangast section so the output can be piped
2011 /// into llvm-bcanalyzer.
2012 static void printRawClangAST(const ObjectFile *Obj) {
2013   if (outs().is_displayed()) {
2014     WithColor::error(errs(), ToolName)
2015         << "The -raw-clang-ast option will dump the raw binary contents of "
2016            "the clang ast section.\n"
2017            "Please redirect the output to a file or another program such as "
2018            "llvm-bcanalyzer.\n";
2019     return;
2020   }
2021 
2022   StringRef ClangASTSectionName("__clangast");
2023   if (Obj->isCOFF()) {
2024     ClangASTSectionName = "clangast";
2025   }
2026 
2027   Optional<object::SectionRef> ClangASTSection;
2028   for (auto Sec : ToolSectionFilter(*Obj)) {
2029     StringRef Name;
2030     if (Expected<StringRef> NameOrErr = Sec.getName())
2031       Name = *NameOrErr;
2032     else
2033       consumeError(NameOrErr.takeError());
2034 
2035     if (Name == ClangASTSectionName) {
2036       ClangASTSection = Sec;
2037       break;
2038     }
2039   }
2040   if (!ClangASTSection)
2041     return;
2042 
2043   StringRef ClangASTContents = unwrapOrError(
2044       ClangASTSection.getValue().getContents(), Obj->getFileName());
2045   outs().write(ClangASTContents.data(), ClangASTContents.size());
2046 }
2047 
2048 static void printFaultMaps(const ObjectFile *Obj) {
2049   StringRef FaultMapSectionName;
2050 
2051   if (Obj->isELF()) {
2052     FaultMapSectionName = ".llvm_faultmaps";
2053   } else if (Obj->isMachO()) {
2054     FaultMapSectionName = "__llvm_faultmaps";
2055   } else {
2056     WithColor::error(errs(), ToolName)
2057         << "This operation is only currently supported "
2058            "for ELF and Mach-O executable files.\n";
2059     return;
2060   }
2061 
2062   Optional<object::SectionRef> FaultMapSection;
2063 
2064   for (auto Sec : ToolSectionFilter(*Obj)) {
2065     StringRef Name;
2066     if (Expected<StringRef> NameOrErr = Sec.getName())
2067       Name = *NameOrErr;
2068     else
2069       consumeError(NameOrErr.takeError());
2070 
2071     if (Name == FaultMapSectionName) {
2072       FaultMapSection = Sec;
2073       break;
2074     }
2075   }
2076 
2077   outs() << "FaultMap table:\n";
2078 
2079   if (!FaultMapSection.hasValue()) {
2080     outs() << "<not found>\n";
2081     return;
2082   }
2083 
2084   StringRef FaultMapContents =
2085       unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName());
2086   FaultMapParser FMP(FaultMapContents.bytes_begin(),
2087                      FaultMapContents.bytes_end());
2088 
2089   outs() << FMP;
2090 }
2091 
2092 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) {
2093   if (O->isELF()) {
2094     printELFFileHeader(O);
2095     printELFDynamicSection(O);
2096     printELFSymbolVersionInfo(O);
2097     return;
2098   }
2099   if (O->isCOFF())
2100     return printCOFFFileHeader(O);
2101   if (O->isWasm())
2102     return printWasmFileHeader(O);
2103   if (O->isMachO()) {
2104     printMachOFileHeader(O);
2105     if (!OnlyFirst)
2106       printMachOLoadCommands(O);
2107     return;
2108   }
2109   reportError(O->getFileName(), "Invalid/Unsupported object file format");
2110 }
2111 
2112 static void printFileHeaders(const ObjectFile *O) {
2113   if (!O->isELF() && !O->isCOFF())
2114     reportError(O->getFileName(), "Invalid/Unsupported object file format");
2115 
2116   Triple::ArchType AT = O->getArch();
2117   outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n";
2118   uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName());
2119 
2120   StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2121   outs() << "start address: "
2122          << "0x" << format(Fmt.data(), Address) << "\n\n";
2123 }
2124 
2125 static void printArchiveChild(StringRef Filename, const Archive::Child &C) {
2126   Expected<sys::fs::perms> ModeOrErr = C.getAccessMode();
2127   if (!ModeOrErr) {
2128     WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n";
2129     consumeError(ModeOrErr.takeError());
2130     return;
2131   }
2132   sys::fs::perms Mode = ModeOrErr.get();
2133   outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
2134   outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
2135   outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
2136   outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
2137   outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
2138   outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
2139   outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
2140   outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
2141   outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
2142 
2143   outs() << " ";
2144 
2145   outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename),
2146                    unwrapOrError(C.getGID(), Filename),
2147                    unwrapOrError(C.getRawSize(), Filename));
2148 
2149   StringRef RawLastModified = C.getRawLastModified();
2150   unsigned Seconds;
2151   if (RawLastModified.getAsInteger(10, Seconds))
2152     outs() << "(date: \"" << RawLastModified
2153            << "\" contains non-decimal chars) ";
2154   else {
2155     // Since ctime(3) returns a 26 character string of the form:
2156     // "Sun Sep 16 01:03:52 1973\n\0"
2157     // just print 24 characters.
2158     time_t t = Seconds;
2159     outs() << format("%.24s ", ctime(&t));
2160   }
2161 
2162   StringRef Name = "";
2163   Expected<StringRef> NameOrErr = C.getName();
2164   if (!NameOrErr) {
2165     consumeError(NameOrErr.takeError());
2166     Name = unwrapOrError(C.getRawName(), Filename);
2167   } else {
2168     Name = NameOrErr.get();
2169   }
2170   outs() << Name << "\n";
2171 }
2172 
2173 // For ELF only now.
2174 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) {
2175   if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) {
2176     if (Elf->getEType() != ELF::ET_REL)
2177       return true;
2178   }
2179   return false;
2180 }
2181 
2182 static void checkForInvalidStartStopAddress(ObjectFile *Obj,
2183                                             uint64_t Start, uint64_t Stop) {
2184   if (!shouldWarnForInvalidStartStopAddress(Obj))
2185     return;
2186 
2187   for (const SectionRef &Section : Obj->sections())
2188     if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) {
2189       uint64_t BaseAddr = Section.getAddress();
2190       uint64_t Size = Section.getSize();
2191       if ((Start < BaseAddr + Size) && Stop > BaseAddr)
2192         return;
2193     }
2194 
2195   if (!HasStartAddressFlag)
2196     reportWarning("no section has address less than 0x" +
2197                       Twine::utohexstr(Stop) + " specified by --stop-address",
2198                   Obj->getFileName());
2199   else if (!HasStopAddressFlag)
2200     reportWarning("no section has address greater than or equal to 0x" +
2201                       Twine::utohexstr(Start) + " specified by --start-address",
2202                   Obj->getFileName());
2203   else
2204     reportWarning("no section overlaps the range [0x" +
2205                       Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) +
2206                       ") specified by --start-address/--stop-address",
2207                   Obj->getFileName());
2208 }
2209 
2210 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
2211                        const Archive::Child *C = nullptr) {
2212   // Avoid other output when using a raw option.
2213   if (!RawClangAST) {
2214     outs() << '\n';
2215     if (A)
2216       outs() << A->getFileName() << "(" << O->getFileName() << ")";
2217     else
2218       outs() << O->getFileName();
2219     outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n";
2220   }
2221 
2222   if (HasStartAddressFlag || HasStopAddressFlag)
2223     checkForInvalidStartStopAddress(O, StartAddress, StopAddress);
2224 
2225   // Note: the order here matches GNU objdump for compatability.
2226   StringRef ArchiveName = A ? A->getFileName() : "";
2227   if (ArchiveHeaders && !MachOOpt && C)
2228     printArchiveChild(ArchiveName, *C);
2229   if (FileHeaders)
2230     printFileHeaders(O);
2231   if (PrivateHeaders || FirstPrivateHeader)
2232     printPrivateFileHeaders(O, FirstPrivateHeader);
2233   if (SectionHeaders)
2234     printSectionHeaders(O);
2235   if (SymbolTable)
2236     printSymbolTable(O, ArchiveName);
2237   if (DynamicSymbolTable)
2238     printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"",
2239                      /*DumpDynamic=*/true);
2240   if (DwarfDumpType != DIDT_Null) {
2241     std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O);
2242     // Dump the complete DWARF structure.
2243     DIDumpOptions DumpOpts;
2244     DumpOpts.DumpType = DwarfDumpType;
2245     DICtx->dump(outs(), DumpOpts);
2246   }
2247   if (Relocations && !Disassemble)
2248     printRelocations(O);
2249   if (DynamicRelocations)
2250     printDynamicRelocations(O);
2251   if (SectionContents)
2252     printSectionContents(O);
2253   if (Disassemble)
2254     disassembleObject(O, Relocations);
2255   if (UnwindInfo)
2256     printUnwindInfo(O);
2257 
2258   // Mach-O specific options:
2259   if (ExportsTrie)
2260     printExportsTrie(O);
2261   if (Rebase)
2262     printRebaseTable(O);
2263   if (Bind)
2264     printBindTable(O);
2265   if (LazyBind)
2266     printLazyBindTable(O);
2267   if (WeakBind)
2268     printWeakBindTable(O);
2269 
2270   // Other special sections:
2271   if (RawClangAST)
2272     printRawClangAST(O);
2273   if (FaultMapSection)
2274     printFaultMaps(O);
2275 }
2276 
2277 static void dumpObject(const COFFImportFile *I, const Archive *A,
2278                        const Archive::Child *C = nullptr) {
2279   StringRef ArchiveName = A ? A->getFileName() : "";
2280 
2281   // Avoid other output when using a raw option.
2282   if (!RawClangAST)
2283     outs() << '\n'
2284            << ArchiveName << "(" << I->getFileName() << ")"
2285            << ":\tfile format COFF-import-file"
2286            << "\n\n";
2287 
2288   if (ArchiveHeaders && !MachOOpt && C)
2289     printArchiveChild(ArchiveName, *C);
2290   if (SymbolTable)
2291     printCOFFSymbolTable(I);
2292 }
2293 
2294 /// Dump each object file in \a a;
2295 static void dumpArchive(const Archive *A) {
2296   Error Err = Error::success();
2297   unsigned I = -1;
2298   for (auto &C : A->children(Err)) {
2299     ++I;
2300     Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
2301     if (!ChildOrErr) {
2302       if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError()))
2303         reportError(std::move(E), getFileNameForError(C, I), A->getFileName());
2304       continue;
2305     }
2306     if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
2307       dumpObject(O, A, &C);
2308     else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get()))
2309       dumpObject(I, A, &C);
2310     else
2311       reportError(errorCodeToError(object_error::invalid_file_type),
2312                   A->getFileName());
2313   }
2314   if (Err)
2315     reportError(std::move(Err), A->getFileName());
2316 }
2317 
2318 /// Open file and figure out how to dump it.
2319 static void dumpInput(StringRef file) {
2320   // If we are using the Mach-O specific object file parser, then let it parse
2321   // the file and process the command line options.  So the -arch flags can
2322   // be used to select specific slices, etc.
2323   if (MachOOpt) {
2324     parseInputMachO(file);
2325     return;
2326   }
2327 
2328   // Attempt to open the binary.
2329   OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file);
2330   Binary &Binary = *OBinary.getBinary();
2331 
2332   if (Archive *A = dyn_cast<Archive>(&Binary))
2333     dumpArchive(A);
2334   else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary))
2335     dumpObject(O);
2336   else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary))
2337     parseInputMachO(UB);
2338   else
2339     reportError(errorCodeToError(object_error::invalid_file_type), file);
2340 }
2341 
2342 template <typename T>
2343 static void parseIntArg(const llvm::opt::InputArgList &InputArgs, int ID,
2344                         T &Value) {
2345   if (const opt::Arg *A = InputArgs.getLastArg(ID)) {
2346     StringRef V(A->getValue());
2347     if (!llvm::to_integer(V, Value, 0)) {
2348       reportCmdLineError(A->getSpelling() +
2349                          ": expected a non-negative integer, but got '" + V +
2350                          "'");
2351     }
2352   }
2353 }
2354 
2355 static std::vector<std::string>
2356 commaSeparatedValues(const llvm::opt::InputArgList &InputArgs, int ID) {
2357   std::vector<std::string> Values;
2358   for (StringRef Value : InputArgs.getAllArgValues(ID)) {
2359     llvm::SmallVector<StringRef, 2> SplitValues;
2360     llvm::SplitString(Value, SplitValues, ",");
2361     for (StringRef SplitValue : SplitValues)
2362       Values.push_back(SplitValue.str());
2363   }
2364   return Values;
2365 }
2366 
2367 static void parseOtoolOptions(const llvm::opt::InputArgList &InputArgs) {
2368   MachOOpt = true;
2369   FullLeadingAddr = true;
2370   PrintImmHex = true;
2371 
2372   ArchName = InputArgs.getLastArgValue(OTOOL_arch).str();
2373   LinkOptHints = InputArgs.hasArg(OTOOL_C);
2374   if (InputArgs.hasArg(OTOOL_d))
2375     FilterSections.push_back("__DATA,__data");
2376   DylibId = InputArgs.hasArg(OTOOL_D);
2377   UniversalHeaders = InputArgs.hasArg(OTOOL_f);
2378   DataInCode = InputArgs.hasArg(OTOOL_G);
2379   FirstPrivateHeader = InputArgs.hasArg(OTOOL_h);
2380   IndirectSymbols = InputArgs.hasArg(OTOOL_I);
2381   ShowRawInsn = InputArgs.hasArg(OTOOL_j);
2382   PrivateHeaders = InputArgs.hasArg(OTOOL_l);
2383   DylibsUsed = InputArgs.hasArg(OTOOL_L);
2384   MCPU = InputArgs.getLastArgValue(OTOOL_mcpu_EQ).str();
2385   ObjcMetaData = InputArgs.hasArg(OTOOL_o);
2386   DisSymName = InputArgs.getLastArgValue(OTOOL_p).str();
2387   InfoPlist = InputArgs.hasArg(OTOOL_P);
2388   Relocations = InputArgs.hasArg(OTOOL_r);
2389   if (const Arg *A = InputArgs.getLastArg(OTOOL_s)) {
2390     auto Filter = (A->getValue(0) + StringRef(",") + A->getValue(1)).str();
2391     FilterSections.push_back(Filter);
2392   }
2393   if (InputArgs.hasArg(OTOOL_t))
2394     FilterSections.push_back("__TEXT,__text");
2395   Verbose = InputArgs.hasArg(OTOOL_v) || InputArgs.hasArg(OTOOL_V) ||
2396             InputArgs.hasArg(OTOOL_o);
2397   SymbolicOperands = InputArgs.hasArg(OTOOL_V);
2398   if (InputArgs.hasArg(OTOOL_x))
2399     FilterSections.push_back(",__text");
2400   LeadingAddr = LeadingHeaders = !InputArgs.hasArg(OTOOL_X);
2401 
2402   InputFilenames = InputArgs.getAllArgValues(OTOOL_INPUT);
2403   if (InputFilenames.empty())
2404     reportCmdLineError("no input file");
2405 
2406   for (const Arg *A : InputArgs) {
2407     const Option &O = A->getOption();
2408     if (O.getGroup().isValid() && O.getGroup().getID() == OTOOL_grp_obsolete) {
2409       reportCmdLineWarning(O.getPrefixedName() +
2410                            " is obsolete and not implemented");
2411     }
2412   }
2413 }
2414 
2415 static void parseObjdumpOptions(const llvm::opt::InputArgList &InputArgs) {
2416   parseIntArg(InputArgs, OBJDUMP_adjust_vma_EQ, AdjustVMA);
2417   AllHeaders = InputArgs.hasArg(OBJDUMP_all_headers);
2418   ArchName = InputArgs.getLastArgValue(OBJDUMP_arch_name_EQ).str();
2419   ArchiveHeaders = InputArgs.hasArg(OBJDUMP_archive_headers);
2420   Demangle = InputArgs.hasArg(OBJDUMP_demangle);
2421   Disassemble = InputArgs.hasArg(OBJDUMP_disassemble);
2422   DisassembleAll = InputArgs.hasArg(OBJDUMP_disassemble_all);
2423   SymbolDescription = InputArgs.hasArg(OBJDUMP_symbol_description);
2424   DisassembleSymbols =
2425       commaSeparatedValues(InputArgs, OBJDUMP_disassemble_symbols_EQ);
2426   DisassembleZeroes = InputArgs.hasArg(OBJDUMP_disassemble_zeroes);
2427   DisassemblerOptions =
2428       commaSeparatedValues(InputArgs, OBJDUMP_disassembler_options_EQ);
2429   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_dwarf_EQ)) {
2430     DwarfDumpType =
2431         StringSwitch<DIDumpType>(A->getValue()).Case("frames", DIDT_DebugFrame);
2432   }
2433   DynamicRelocations = InputArgs.hasArg(OBJDUMP_dynamic_reloc);
2434   FaultMapSection = InputArgs.hasArg(OBJDUMP_fault_map_section);
2435   FileHeaders = InputArgs.hasArg(OBJDUMP_file_headers);
2436   SectionContents = InputArgs.hasArg(OBJDUMP_full_contents);
2437   PrintLines = InputArgs.hasArg(OBJDUMP_line_numbers);
2438   InputFilenames = InputArgs.getAllArgValues(OBJDUMP_INPUT);
2439   MachOOpt = InputArgs.hasArg(OBJDUMP_macho);
2440   MCPU = InputArgs.getLastArgValue(OBJDUMP_mcpu_EQ).str();
2441   MAttrs = commaSeparatedValues(InputArgs, OBJDUMP_mattr_EQ);
2442   ShowRawInsn = !InputArgs.hasArg(OBJDUMP_no_show_raw_insn);
2443   LeadingAddr = !InputArgs.hasArg(OBJDUMP_no_leading_addr);
2444   RawClangAST = InputArgs.hasArg(OBJDUMP_raw_clang_ast);
2445   Relocations = InputArgs.hasArg(OBJDUMP_reloc);
2446   PrintImmHex =
2447       InputArgs.hasFlag(OBJDUMP_print_imm_hex, OBJDUMP_no_print_imm_hex, false);
2448   PrivateHeaders = InputArgs.hasArg(OBJDUMP_private_headers);
2449   FilterSections = InputArgs.getAllArgValues(OBJDUMP_section_EQ);
2450   SectionHeaders = InputArgs.hasArg(OBJDUMP_section_headers);
2451   ShowLMA = InputArgs.hasArg(OBJDUMP_show_lma);
2452   PrintSource = InputArgs.hasArg(OBJDUMP_source);
2453   parseIntArg(InputArgs, OBJDUMP_start_address_EQ, StartAddress);
2454   HasStartAddressFlag = InputArgs.hasArg(OBJDUMP_start_address_EQ);
2455   parseIntArg(InputArgs, OBJDUMP_stop_address_EQ, StopAddress);
2456   HasStopAddressFlag = InputArgs.hasArg(OBJDUMP_stop_address_EQ);
2457   SymbolTable = InputArgs.hasArg(OBJDUMP_syms);
2458   SymbolizeOperands = InputArgs.hasArg(OBJDUMP_symbolize_operands);
2459   DynamicSymbolTable = InputArgs.hasArg(OBJDUMP_dynamic_syms);
2460   TripleName = InputArgs.getLastArgValue(OBJDUMP_triple_EQ).str();
2461   UnwindInfo = InputArgs.hasArg(OBJDUMP_unwind_info);
2462   Wide = InputArgs.hasArg(OBJDUMP_wide);
2463   Prefix = InputArgs.getLastArgValue(OBJDUMP_prefix).str();
2464   parseIntArg(InputArgs, OBJDUMP_prefix_strip, PrefixStrip);
2465   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_debug_vars_EQ)) {
2466     DbgVariables = StringSwitch<DebugVarsFormat>(A->getValue())
2467                        .Case("ascii", DVASCII)
2468                        .Case("unicode", DVUnicode);
2469   }
2470   parseIntArg(InputArgs, OBJDUMP_debug_vars_indent_EQ, DbgIndent);
2471 
2472   parseMachOOptions(InputArgs);
2473 
2474   // Handle options that get forwarded to cl::opt<>s in libraries.
2475   // FIXME: Depending on https://reviews.llvm.org/D84191#inline-946075 ,
2476   // hopefully remove this again.
2477   std::vector<const char *> LLVMArgs;
2478   LLVMArgs.push_back("llvm-objdump (LLVM option parsing)");
2479   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_x86_asm_syntax_att,
2480                                                OBJDUMP_x86_asm_syntax_intel)) {
2481     switch (A->getOption().getID()) {
2482     case OBJDUMP_x86_asm_syntax_att:
2483       LLVMArgs.push_back("--x86-asm-syntax=att");
2484       break;
2485     case OBJDUMP_x86_asm_syntax_intel:
2486       LLVMArgs.push_back("--x86-asm-syntax=intel");
2487       break;
2488     }
2489   }
2490   if (InputArgs.hasArg(OBJDUMP_mhvx))
2491     LLVMArgs.push_back("--mhvx");
2492   if (InputArgs.hasArg(OBJDUMP_mhvx_v66))
2493     LLVMArgs.push_back("--mhvx=v66");
2494   if (InputArgs.hasArg(OBJDUMP_mv60))
2495     LLVMArgs.push_back("--mv60");
2496   if (InputArgs.hasArg(OBJDUMP_mv65))
2497     LLVMArgs.push_back("--mv65");
2498   if (InputArgs.hasArg(OBJDUMP_mv66))
2499     LLVMArgs.push_back("--mv66");
2500   if (InputArgs.hasArg(OBJDUMP_mv67))
2501     LLVMArgs.push_back("--mv67");
2502   if (InputArgs.hasArg(OBJDUMP_mv67t))
2503     LLVMArgs.push_back("--mv67t");
2504   if (InputArgs.hasArg(OBJDUMP_riscv_no_aliases))
2505     LLVMArgs.push_back("--riscv-no-aliases");
2506   LLVMArgs.push_back(nullptr);
2507   llvm::cl::ParseCommandLineOptions(LLVMArgs.size() - 1, LLVMArgs.data());
2508 
2509   // objdump defaults to a.out if no filenames specified.
2510   if (InputFilenames.empty())
2511     InputFilenames.push_back("a.out");
2512 }
2513 
2514 int main(int argc, char **argv) {
2515   using namespace llvm;
2516   InitLLVM X(argc, argv);
2517 
2518   ToolName = argv[0];
2519   std::unique_ptr<CommonOptTable> T;
2520   OptSpecifier Unknown, HelpFlag, HelpHiddenFlag, VersionFlag;
2521 
2522   StringRef Stem = sys::path::stem(ToolName);
2523   auto Is = [=](StringRef Tool) {
2524     // We need to recognize the following filenames:
2525     //
2526     // llvm-objdump -> objdump
2527     // llvm-otool-10.exe -> otool
2528     // powerpc64-unknown-freebsd13-objdump -> objdump
2529     auto I = Stem.rfind_lower(Tool);
2530     return I != StringRef::npos &&
2531            (I + Tool.size() == Stem.size() || !isAlnum(Stem[I + Tool.size()]));
2532   };
2533   if (Is("otool")) {
2534     T = std::make_unique<OtoolOptTable>();
2535     Unknown = OTOOL_UNKNOWN;
2536     HelpFlag = OTOOL_help;
2537     HelpHiddenFlag = OTOOL_help_hidden;
2538     VersionFlag = OTOOL_version;
2539   } else {
2540     T = std::make_unique<ObjdumpOptTable>();
2541     Unknown = OBJDUMP_UNKNOWN;
2542     HelpFlag = OBJDUMP_help;
2543     HelpHiddenFlag = OBJDUMP_help_hidden;
2544     VersionFlag = OBJDUMP_version;
2545   }
2546 
2547   BumpPtrAllocator A;
2548   StringSaver Saver(A);
2549   opt::InputArgList InputArgs =
2550       T->parseArgs(argc, argv, Unknown, Saver,
2551                    [&](StringRef Msg) { reportCmdLineError(Msg); });
2552 
2553   if (InputArgs.size() == 0 || InputArgs.hasArg(HelpFlag)) {
2554     T->printHelp(ToolName);
2555     return 0;
2556   }
2557   if (InputArgs.hasArg(HelpHiddenFlag)) {
2558     T->printHelp(ToolName, /*show_hidden=*/true);
2559     return 0;
2560   }
2561 
2562   // Initialize targets and assembly printers/parsers.
2563   InitializeAllTargetInfos();
2564   InitializeAllTargetMCs();
2565   InitializeAllDisassemblers();
2566 
2567   if (InputArgs.hasArg(VersionFlag)) {
2568     cl::PrintVersionMessage();
2569     if (!Is("otool")) {
2570       outs() << '\n';
2571       TargetRegistry::printRegisteredTargetsForVersion(outs());
2572     }
2573     return 0;
2574   }
2575 
2576   if (Is("otool"))
2577     parseOtoolOptions(InputArgs);
2578   else
2579     parseObjdumpOptions(InputArgs);
2580 
2581   if (StartAddress >= StopAddress)
2582     reportCmdLineError("start address should be less than stop address");
2583 
2584   // Removes trailing separators from prefix.
2585   while (!Prefix.empty() && sys::path::is_separator(Prefix.back()))
2586     Prefix.pop_back();
2587 
2588   if (AllHeaders)
2589     ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations =
2590         SectionHeaders = SymbolTable = true;
2591 
2592   if (DisassembleAll || PrintSource || PrintLines ||
2593       !DisassembleSymbols.empty())
2594     Disassemble = true;
2595 
2596   if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null &&
2597       !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST &&
2598       !Relocations && !SectionHeaders && !SectionContents && !SymbolTable &&
2599       !DynamicSymbolTable && !UnwindInfo && !FaultMapSection &&
2600       !(MachOOpt &&
2601         (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie ||
2602          FirstPrivateHeader || FunctionStarts || IndirectSymbols || InfoPlist ||
2603          LazyBind || LinkOptHints || ObjcMetaData || Rebase || Rpaths ||
2604          UniversalHeaders || WeakBind || !FilterSections.empty()))) {
2605     T->printHelp(ToolName);
2606     return 2;
2607   }
2608 
2609   DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end());
2610 
2611   llvm::for_each(InputFilenames, dumpInput);
2612 
2613   warnOnNoMatchForSections();
2614 
2615   return EXIT_SUCCESS;
2616 }
2617