xref: /llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp (revision c623945d707c8d50d0cae79b5c95a57166cd066c)
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 static StringRef getSegmentName(const MachOObjectFile *MachO,
976                                 const SectionRef &Section) {
977   if (MachO) {
978     DataRefImpl DR = Section.getRawDataRefImpl();
979     StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
980     return SegmentName;
981   }
982   return "";
983 }
984 
985 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj,
986                               MCContext &Ctx, MCDisassembler *PrimaryDisAsm,
987                               MCDisassembler *SecondaryDisAsm,
988                               const MCInstrAnalysis *MIA, MCInstPrinter *IP,
989                               const MCSubtargetInfo *PrimarySTI,
990                               const MCSubtargetInfo *SecondarySTI,
991                               PrettyPrinter &PIP,
992                               SourcePrinter &SP, bool InlineRelocs) {
993   const MCSubtargetInfo *STI = PrimarySTI;
994   MCDisassembler *DisAsm = PrimaryDisAsm;
995   bool PrimaryIsThumb = false;
996   if (isArmElf(Obj))
997     PrimaryIsThumb = STI->checkFeatures("+thumb-mode");
998 
999   std::map<SectionRef, std::vector<RelocationRef>> RelocMap;
1000   if (InlineRelocs)
1001     RelocMap = getRelocsMap(*Obj);
1002   bool Is64Bits = Obj->getBytesInAddress() > 4;
1003 
1004   // Create a mapping from virtual address to symbol name.  This is used to
1005   // pretty print the symbols while disassembling.
1006   std::map<SectionRef, SectionSymbolsTy> AllSymbols;
1007   SectionSymbolsTy AbsoluteSymbols;
1008   const StringRef FileName = Obj->getFileName();
1009   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
1010   for (const SymbolRef &Symbol : Obj->symbols()) {
1011     Expected<StringRef> NameOrErr = Symbol.getName();
1012     if (!NameOrErr) {
1013       reportWarning(toString(NameOrErr.takeError()), FileName);
1014       continue;
1015     }
1016     if (NameOrErr->empty() && !(Obj->isXCOFF() && SymbolDescription))
1017       continue;
1018 
1019     if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION)
1020       continue;
1021 
1022     // Don't ask a Mach-O STAB symbol for its section unless you know that
1023     // STAB symbol's section field refers to a valid section index. Otherwise
1024     // the symbol may error trying to load a section that does not exist.
1025     if (MachO) {
1026       DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1027       uint8_t NType = (MachO->is64Bit() ?
1028                        MachO->getSymbol64TableEntry(SymDRI).n_type:
1029                        MachO->getSymbolTableEntry(SymDRI).n_type);
1030       if (NType & MachO::N_STAB)
1031         continue;
1032     }
1033 
1034     section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1035     if (SecI != Obj->section_end())
1036       AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol));
1037     else
1038       AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol));
1039   }
1040 
1041   if (AllSymbols.empty() && Obj->isELF())
1042     addDynamicElfSymbols(Obj, AllSymbols);
1043 
1044   BumpPtrAllocator A;
1045   StringSaver Saver(A);
1046   addPltEntries(Obj, AllSymbols, Saver);
1047 
1048   // Create a mapping from virtual address to section. An empty section can
1049   // cause more than one section at the same address. Sort such sections to be
1050   // before same-addressed non-empty sections so that symbol lookups prefer the
1051   // non-empty section.
1052   std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses;
1053   for (SectionRef Sec : Obj->sections())
1054     SectionAddresses.emplace_back(Sec.getAddress(), Sec);
1055   llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) {
1056     if (LHS.first != RHS.first)
1057       return LHS.first < RHS.first;
1058     return LHS.second.getSize() < RHS.second.getSize();
1059   });
1060 
1061   // Linked executables (.exe and .dll files) typically don't include a real
1062   // symbol table but they might contain an export table.
1063   if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) {
1064     for (const auto &ExportEntry : COFFObj->export_directories()) {
1065       StringRef Name;
1066       if (Error E = ExportEntry.getSymbolName(Name))
1067         reportError(std::move(E), Obj->getFileName());
1068       if (Name.empty())
1069         continue;
1070 
1071       uint32_t RVA;
1072       if (Error E = ExportEntry.getExportRVA(RVA))
1073         reportError(std::move(E), Obj->getFileName());
1074 
1075       uint64_t VA = COFFObj->getImageBase() + RVA;
1076       auto Sec = partition_point(
1077           SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) {
1078             return O.first <= VA;
1079           });
1080       if (Sec != SectionAddresses.begin()) {
1081         --Sec;
1082         AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE);
1083       } else
1084         AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE);
1085     }
1086   }
1087 
1088   // Sort all the symbols, this allows us to use a simple binary search to find
1089   // Multiple symbols can have the same address. Use a stable sort to stabilize
1090   // the output.
1091   StringSet<> FoundDisasmSymbolSet;
1092   for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols)
1093     llvm::stable_sort(SecSyms.second);
1094   llvm::stable_sort(AbsoluteSymbols);
1095 
1096   std::unique_ptr<DWARFContext> DICtx;
1097   LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI);
1098 
1099   if (DbgVariables != DVDisabled) {
1100     DICtx = DWARFContext::create(*Obj);
1101     for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units())
1102       LVP.addCompileUnit(CU->getUnitDIE(false));
1103   }
1104 
1105   LLVM_DEBUG(LVP.dump());
1106 
1107   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1108     if (FilterSections.empty() && !DisassembleAll &&
1109         (!Section.isText() || Section.isVirtual()))
1110       continue;
1111 
1112     uint64_t SectionAddr = Section.getAddress();
1113     uint64_t SectSize = Section.getSize();
1114     if (!SectSize)
1115       continue;
1116 
1117     // Get the list of all the symbols in this section.
1118     SectionSymbolsTy &Symbols = AllSymbols[Section];
1119     std::vector<MappingSymbolPair> MappingSymbols;
1120     if (hasMappingSymbols(Obj)) {
1121       for (const auto &Symb : Symbols) {
1122         uint64_t Address = Symb.Addr;
1123         StringRef Name = Symb.Name;
1124         if (Name.startswith("$d"))
1125           MappingSymbols.emplace_back(Address - SectionAddr, 'd');
1126         if (Name.startswith("$x"))
1127           MappingSymbols.emplace_back(Address - SectionAddr, 'x');
1128         if (Name.startswith("$a"))
1129           MappingSymbols.emplace_back(Address - SectionAddr, 'a');
1130         if (Name.startswith("$t"))
1131           MappingSymbols.emplace_back(Address - SectionAddr, 't');
1132       }
1133     }
1134 
1135     llvm::sort(MappingSymbols);
1136 
1137     if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) {
1138       // AMDGPU disassembler uses symbolizer for printing labels
1139       std::unique_ptr<MCRelocationInfo> RelInfo(
1140         TheTarget->createMCRelocationInfo(TripleName, Ctx));
1141       if (RelInfo) {
1142         std::unique_ptr<MCSymbolizer> Symbolizer(
1143           TheTarget->createMCSymbolizer(
1144             TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1145         DisAsm->setSymbolizer(std::move(Symbolizer));
1146       }
1147     }
1148 
1149     StringRef SegmentName = getSegmentName(MachO, Section);
1150     StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName());
1151     // If the section has no symbol at the start, just insert a dummy one.
1152     if (Symbols.empty() || Symbols[0].Addr != 0) {
1153       Symbols.insert(Symbols.begin(),
1154                      createDummySymbolInfo(Obj, SectionAddr, SectionName,
1155                                            Section.isText() ? ELF::STT_FUNC
1156                                                             : ELF::STT_OBJECT));
1157     }
1158 
1159     SmallString<40> Comments;
1160     raw_svector_ostream CommentStream(Comments);
1161 
1162     ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(
1163         unwrapOrError(Section.getContents(), Obj->getFileName()));
1164 
1165     uint64_t VMAAdjustment = 0;
1166     if (shouldAdjustVA(Section))
1167       VMAAdjustment = AdjustVMA;
1168 
1169     uint64_t Size;
1170     uint64_t Index;
1171     bool PrintedSection = false;
1172     std::vector<RelocationRef> Rels = RelocMap[Section];
1173     std::vector<RelocationRef>::const_iterator RelCur = Rels.begin();
1174     std::vector<RelocationRef>::const_iterator RelEnd = Rels.end();
1175     // Disassemble symbol by symbol.
1176     for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) {
1177       std::string SymbolName = Symbols[SI].Name.str();
1178       if (Demangle)
1179         SymbolName = demangle(SymbolName);
1180 
1181       // Skip if --disassemble-symbols is not empty and the symbol is not in
1182       // the list.
1183       if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName))
1184         continue;
1185 
1186       uint64_t Start = Symbols[SI].Addr;
1187       if (Start < SectionAddr || StopAddress <= Start)
1188         continue;
1189       else
1190         FoundDisasmSymbolSet.insert(SymbolName);
1191 
1192       // The end is the section end, the beginning of the next symbol, or
1193       // --stop-address.
1194       uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress);
1195       if (SI + 1 < SE)
1196         End = std::min(End, Symbols[SI + 1].Addr);
1197       if (Start >= End || End <= StartAddress)
1198         continue;
1199       Start -= SectionAddr;
1200       End -= SectionAddr;
1201 
1202       if (!PrintedSection) {
1203         PrintedSection = true;
1204         outs() << "\nDisassembly of section ";
1205         if (!SegmentName.empty())
1206           outs() << SegmentName << ",";
1207         outs() << SectionName << ":\n";
1208       }
1209 
1210       outs() << '\n';
1211       if (LeadingAddr)
1212         outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ",
1213                          SectionAddr + Start + VMAAdjustment);
1214       if (Obj->isXCOFF() && SymbolDescription) {
1215         outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n";
1216       } else
1217         outs() << '<' << SymbolName << ">:\n";
1218 
1219       // Don't print raw contents of a virtual section. A virtual section
1220       // doesn't have any contents in the file.
1221       if (Section.isVirtual()) {
1222         outs() << "...\n";
1223         continue;
1224       }
1225 
1226       auto Status = DisAsm->onSymbolStart(Symbols[SI], Size,
1227                                           Bytes.slice(Start, End - Start),
1228                                           SectionAddr + Start, CommentStream);
1229       // To have round trippable disassembly, we fall back to decoding the
1230       // remaining bytes as instructions.
1231       //
1232       // If there is a failure, we disassemble the failed region as bytes before
1233       // falling back. The target is expected to print nothing in this case.
1234       //
1235       // If there is Success or SoftFail i.e no 'real' failure, we go ahead by
1236       // Size bytes before falling back.
1237       // So if the entire symbol is 'eaten' by the target:
1238       //   Start += Size  // Now Start = End and we will never decode as
1239       //                  // instructions
1240       //
1241       // Right now, most targets return None i.e ignore to treat a symbol
1242       // separately. But WebAssembly decodes preludes for some symbols.
1243       //
1244       if (Status.hasValue()) {
1245         if (Status.getValue() == MCDisassembler::Fail) {
1246           outs() << "// Error in decoding " << SymbolName
1247                  << " : Decoding failed region as bytes.\n";
1248           for (uint64_t I = 0; I < Size; ++I) {
1249             outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true)
1250                    << "\n";
1251           }
1252         }
1253       } else {
1254         Size = 0;
1255       }
1256 
1257       Start += Size;
1258 
1259       Index = Start;
1260       if (SectionAddr < StartAddress)
1261         Index = std::max<uint64_t>(Index, StartAddress - SectionAddr);
1262 
1263       // If there is a data/common symbol inside an ELF text section and we are
1264       // only disassembling text (applicable all architectures), we are in a
1265       // situation where we must print the data and not disassemble it.
1266       if (Obj->isELF() && !DisassembleAll && Section.isText()) {
1267         uint8_t SymTy = Symbols[SI].Type;
1268         if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) {
1269           dumpELFData(SectionAddr, Index, End, Bytes);
1270           Index = End;
1271         }
1272       }
1273 
1274       bool CheckARMELFData = hasMappingSymbols(Obj) &&
1275                              Symbols[SI].Type != ELF::STT_OBJECT &&
1276                              !DisassembleAll;
1277       bool DumpARMELFData = false;
1278       formatted_raw_ostream FOS(outs());
1279 
1280       std::unordered_map<uint64_t, std::string> AllLabels;
1281       if (SymbolizeOperands)
1282         collectLocalBranchTargets(Bytes, MIA, DisAsm, IP, PrimarySTI,
1283                                   SectionAddr, Index, End, AllLabels);
1284 
1285       while (Index < End) {
1286         // ARM and AArch64 ELF binaries can interleave data and text in the
1287         // same section. We rely on the markers introduced to understand what
1288         // we need to dump. If the data marker is within a function, it is
1289         // denoted as a word/short etc.
1290         if (CheckARMELFData) {
1291           char Kind = getMappingSymbolKind(MappingSymbols, Index);
1292           DumpARMELFData = Kind == 'd';
1293           if (SecondarySTI) {
1294             if (Kind == 'a') {
1295               STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI;
1296               DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm;
1297             } else if (Kind == 't') {
1298               STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI;
1299               DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm;
1300             }
1301           }
1302         }
1303 
1304         if (DumpARMELFData) {
1305           Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes,
1306                                 MappingSymbols, FOS);
1307         } else {
1308           // When -z or --disassemble-zeroes are given we always dissasemble
1309           // them. Otherwise we might want to skip zero bytes we see.
1310           if (!DisassembleZeroes) {
1311             uint64_t MaxOffset = End - Index;
1312             // For --reloc: print zero blocks patched by relocations, so that
1313             // relocations can be shown in the dump.
1314             if (RelCur != RelEnd)
1315               MaxOffset = RelCur->getOffset() - Index;
1316 
1317             if (size_t N =
1318                     countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) {
1319               FOS << "\t\t..." << '\n';
1320               Index += N;
1321               continue;
1322             }
1323           }
1324 
1325           // Print local label if there's any.
1326           auto Iter = AllLabels.find(SectionAddr + Index);
1327           if (Iter != AllLabels.end())
1328             FOS << "<" << Iter->second << ">:\n";
1329 
1330           // Disassemble a real instruction or a data when disassemble all is
1331           // provided
1332           MCInst Inst;
1333           bool Disassembled =
1334               DisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
1335                                      SectionAddr + Index, CommentStream);
1336           if (Size == 0)
1337             Size = 1;
1338 
1339           LVP.update({Index, Section.getIndex()},
1340                      {Index + Size, Section.getIndex()}, Index + Size != End);
1341 
1342           PIP.printInst(
1343               *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size),
1344               {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS,
1345               "", *STI, &SP, Obj->getFileName(), &Rels, LVP);
1346           FOS << CommentStream.str();
1347           Comments.clear();
1348 
1349           // If disassembly has failed, avoid analysing invalid/incomplete
1350           // instruction information. Otherwise, try to resolve the target
1351           // address (jump target or memory operand address) and print it on the
1352           // right of the instruction.
1353           if (Disassembled && MIA) {
1354             uint64_t Target;
1355             bool PrintTarget =
1356                 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target);
1357             if (!PrintTarget)
1358               if (Optional<uint64_t> MaybeTarget =
1359                       MIA->evaluateMemoryOperandAddress(
1360                           Inst, SectionAddr + Index, Size)) {
1361                 Target = *MaybeTarget;
1362                 PrintTarget = true;
1363                 // Do not print real address when symbolizing.
1364                 if (!SymbolizeOperands)
1365                   FOS << "  # " << Twine::utohexstr(Target);
1366               }
1367             if (PrintTarget) {
1368               // In a relocatable object, the target's section must reside in
1369               // the same section as the call instruction or it is accessed
1370               // through a relocation.
1371               //
1372               // In a non-relocatable object, the target may be in any section.
1373               // In that case, locate the section(s) containing the target
1374               // address and find the symbol in one of those, if possible.
1375               //
1376               // N.B. We don't walk the relocations in the relocatable case yet.
1377               std::vector<const SectionSymbolsTy *> TargetSectionSymbols;
1378               if (!Obj->isRelocatableObject()) {
1379                 auto It = llvm::partition_point(
1380                     SectionAddresses,
1381                     [=](const std::pair<uint64_t, SectionRef> &O) {
1382                       return O.first <= Target;
1383                     });
1384                 uint64_t TargetSecAddr = 0;
1385                 while (It != SectionAddresses.begin()) {
1386                   --It;
1387                   if (TargetSecAddr == 0)
1388                     TargetSecAddr = It->first;
1389                   if (It->first != TargetSecAddr)
1390                     break;
1391                   TargetSectionSymbols.push_back(&AllSymbols[It->second]);
1392                 }
1393               } else {
1394                 TargetSectionSymbols.push_back(&Symbols);
1395               }
1396               TargetSectionSymbols.push_back(&AbsoluteSymbols);
1397 
1398               // Find the last symbol in the first candidate section whose
1399               // offset is less than or equal to the target. If there are no
1400               // such symbols, try in the next section and so on, before finally
1401               // using the nearest preceding absolute symbol (if any), if there
1402               // are no other valid symbols.
1403               const SymbolInfoTy *TargetSym = nullptr;
1404               for (const SectionSymbolsTy *TargetSymbols :
1405                    TargetSectionSymbols) {
1406                 auto It = llvm::partition_point(
1407                     *TargetSymbols,
1408                     [=](const SymbolInfoTy &O) { return O.Addr <= Target; });
1409                 if (It != TargetSymbols->begin()) {
1410                   TargetSym = &*(It - 1);
1411                   break;
1412                 }
1413               }
1414 
1415               // Print the labels corresponding to the target if there's any.
1416               bool LabelAvailable = AllLabels.count(Target);
1417               if (TargetSym != nullptr) {
1418                 uint64_t TargetAddress = TargetSym->Addr;
1419                 uint64_t Disp = Target - TargetAddress;
1420                 std::string TargetName = TargetSym->Name.str();
1421                 if (Demangle)
1422                   TargetName = demangle(TargetName);
1423 
1424                 FOS << " <";
1425                 if (!Disp) {
1426                   // Always Print the binary symbol precisely corresponding to
1427                   // the target address.
1428                   FOS << TargetName;
1429                 } else if (!LabelAvailable) {
1430                   // Always Print the binary symbol plus an offset if there's no
1431                   // local label corresponding to the target address.
1432                   FOS << TargetName << "+0x" << Twine::utohexstr(Disp);
1433                 } else {
1434                   FOS << AllLabels[Target];
1435                 }
1436                 FOS << ">";
1437               } else if (LabelAvailable) {
1438                 FOS << " <" << AllLabels[Target] << ">";
1439               }
1440             }
1441           }
1442         }
1443 
1444         LVP.printAfterInst(FOS);
1445         FOS << "\n";
1446 
1447         // Hexagon does this in pretty printer
1448         if (Obj->getArch() != Triple::hexagon) {
1449           // Print relocation for instruction and data.
1450           while (RelCur != RelEnd) {
1451             uint64_t Offset = RelCur->getOffset();
1452             // If this relocation is hidden, skip it.
1453             if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) {
1454               ++RelCur;
1455               continue;
1456             }
1457 
1458             // Stop when RelCur's offset is past the disassembled
1459             // instruction/data. Note that it's possible the disassembled data
1460             // is not the complete data: we might see the relocation printed in
1461             // the middle of the data, but this matches the binutils objdump
1462             // output.
1463             if (Offset >= Index + Size)
1464               break;
1465 
1466             // When --adjust-vma is used, update the address printed.
1467             if (RelCur->getSymbol() != Obj->symbol_end()) {
1468               Expected<section_iterator> SymSI =
1469                   RelCur->getSymbol()->getSection();
1470               if (SymSI && *SymSI != Obj->section_end() &&
1471                   shouldAdjustVA(**SymSI))
1472                 Offset += AdjustVMA;
1473             }
1474 
1475             printRelocation(FOS, Obj->getFileName(), *RelCur,
1476                             SectionAddr + Offset, Is64Bits);
1477             LVP.printAfterOtherLine(FOS, true);
1478             ++RelCur;
1479           }
1480         }
1481 
1482         Index += Size;
1483       }
1484     }
1485   }
1486   StringSet<> MissingDisasmSymbolSet =
1487       set_difference(DisasmSymbolSet, FoundDisasmSymbolSet);
1488   for (StringRef Sym : MissingDisasmSymbolSet.keys())
1489     reportWarning("failed to disassemble missing symbol " + Sym, FileName);
1490 }
1491 
1492 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) {
1493   const Target *TheTarget = getTarget(Obj);
1494 
1495   // Package up features to be passed to target/subtarget
1496   SubtargetFeatures Features = Obj->getFeatures();
1497   if (!MAttrs.empty())
1498     for (unsigned I = 0; I != MAttrs.size(); ++I)
1499       Features.AddFeature(MAttrs[I]);
1500 
1501   std::unique_ptr<const MCRegisterInfo> MRI(
1502       TheTarget->createMCRegInfo(TripleName));
1503   if (!MRI)
1504     reportError(Obj->getFileName(),
1505                 "no register info for target " + TripleName);
1506 
1507   // Set up disassembler.
1508   MCTargetOptions MCOptions;
1509   std::unique_ptr<const MCAsmInfo> AsmInfo(
1510       TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions));
1511   if (!AsmInfo)
1512     reportError(Obj->getFileName(),
1513                 "no assembly info for target " + TripleName);
1514 
1515   if (MCPU.empty())
1516     MCPU = Obj->tryGetCPUName().getValueOr("").str();
1517 
1518   std::unique_ptr<const MCSubtargetInfo> STI(
1519       TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString()));
1520   if (!STI)
1521     reportError(Obj->getFileName(),
1522                 "no subtarget info for target " + TripleName);
1523   std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo());
1524   if (!MII)
1525     reportError(Obj->getFileName(),
1526                 "no instruction info for target " + TripleName);
1527   MCObjectFileInfo MOFI;
1528   MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI);
1529   // FIXME: for now initialize MCObjectFileInfo with default values
1530   MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx);
1531 
1532   std::unique_ptr<MCDisassembler> DisAsm(
1533       TheTarget->createMCDisassembler(*STI, Ctx));
1534   if (!DisAsm)
1535     reportError(Obj->getFileName(), "no disassembler for target " + TripleName);
1536 
1537   // If we have an ARM object file, we need a second disassembler, because
1538   // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode.
1539   // We use mapping symbols to switch between the two assemblers, where
1540   // appropriate.
1541   std::unique_ptr<MCDisassembler> SecondaryDisAsm;
1542   std::unique_ptr<const MCSubtargetInfo> SecondarySTI;
1543   if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) {
1544     if (STI->checkFeatures("+thumb-mode"))
1545       Features.AddFeature("-thumb-mode");
1546     else
1547       Features.AddFeature("+thumb-mode");
1548     SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU,
1549                                                         Features.getString()));
1550     SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx));
1551   }
1552 
1553   std::unique_ptr<const MCInstrAnalysis> MIA(
1554       TheTarget->createMCInstrAnalysis(MII.get()));
1555 
1556   int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1557   std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
1558       Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI));
1559   if (!IP)
1560     reportError(Obj->getFileName(),
1561                 "no instruction printer for target " + TripleName);
1562   IP->setPrintImmHex(PrintImmHex);
1563   IP->setPrintBranchImmAsAddress(true);
1564   IP->setSymbolizeOperands(SymbolizeOperands);
1565   IP->setMCInstrAnalysis(MIA.get());
1566 
1567   PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName));
1568   SourcePrinter SP(Obj, TheTarget->getName());
1569 
1570   for (StringRef Opt : DisassemblerOptions)
1571     if (!IP->applyTargetSpecificCLOption(Opt))
1572       reportError(Obj->getFileName(),
1573                   "Unrecognized disassembler option: " + Opt);
1574 
1575   disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(),
1576                     MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP,
1577                     SP, InlineRelocs);
1578 }
1579 
1580 void objdump::printRelocations(const ObjectFile *Obj) {
1581   StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 :
1582                                                  "%08" PRIx64;
1583   // Regular objdump doesn't print relocations in non-relocatable object
1584   // files.
1585   if (!Obj->isRelocatableObject())
1586     return;
1587 
1588   // Build a mapping from relocation target to a vector of relocation
1589   // sections. Usually, there is an only one relocation section for
1590   // each relocated section.
1591   MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec;
1592   uint64_t Ndx;
1593   for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) {
1594     if (Section.relocation_begin() == Section.relocation_end())
1595       continue;
1596     Expected<section_iterator> SecOrErr = Section.getRelocatedSection();
1597     if (!SecOrErr)
1598       reportError(Obj->getFileName(),
1599                   "section (" + Twine(Ndx) +
1600                       "): unable to get a relocation target: " +
1601                       toString(SecOrErr.takeError()));
1602     SecToRelSec[**SecOrErr].push_back(Section);
1603   }
1604 
1605   for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) {
1606     StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName());
1607     outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n";
1608     uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8);
1609     uint32_t TypePadding = 24;
1610     outs() << left_justify("OFFSET", OffsetPadding) << " "
1611            << left_justify("TYPE", TypePadding) << " "
1612            << "VALUE\n";
1613 
1614     for (SectionRef Section : P.second) {
1615       for (const RelocationRef &Reloc : Section.relocations()) {
1616         uint64_t Address = Reloc.getOffset();
1617         SmallString<32> RelocName;
1618         SmallString<32> ValueStr;
1619         if (Address < StartAddress || Address > StopAddress || getHidden(Reloc))
1620           continue;
1621         Reloc.getTypeName(RelocName);
1622         if (Error E = getRelocationValueString(Reloc, ValueStr))
1623           reportError(std::move(E), Obj->getFileName());
1624 
1625         outs() << format(Fmt.data(), Address) << " "
1626                << left_justify(RelocName, TypePadding) << " " << ValueStr
1627                << "\n";
1628       }
1629     }
1630     outs() << "\n";
1631   }
1632 }
1633 
1634 void objdump::printDynamicRelocations(const ObjectFile *Obj) {
1635   // For the moment, this option is for ELF only
1636   if (!Obj->isELF())
1637     return;
1638 
1639   const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
1640   if (!Elf || Elf->getEType() != ELF::ET_DYN) {
1641     reportError(Obj->getFileName(), "not a dynamic object");
1642     return;
1643   }
1644 
1645   std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections();
1646   if (DynRelSec.empty())
1647     return;
1648 
1649   outs() << "DYNAMIC RELOCATION RECORDS\n";
1650   StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1651   for (const SectionRef &Section : DynRelSec)
1652     for (const RelocationRef &Reloc : Section.relocations()) {
1653       uint64_t Address = Reloc.getOffset();
1654       SmallString<32> RelocName;
1655       SmallString<32> ValueStr;
1656       Reloc.getTypeName(RelocName);
1657       if (Error E = getRelocationValueString(Reloc, ValueStr))
1658         reportError(std::move(E), Obj->getFileName());
1659       outs() << format(Fmt.data(), Address) << " " << RelocName << " "
1660              << ValueStr << "\n";
1661     }
1662 }
1663 
1664 // Returns true if we need to show LMA column when dumping section headers. We
1665 // show it only when the platform is ELF and either we have at least one section
1666 // whose VMA and LMA are different and/or when --show-lma flag is used.
1667 static bool shouldDisplayLMA(const ObjectFile *Obj) {
1668   if (!Obj->isELF())
1669     return false;
1670   for (const SectionRef &S : ToolSectionFilter(*Obj))
1671     if (S.getAddress() != getELFSectionLMA(S))
1672       return true;
1673   return ShowLMA;
1674 }
1675 
1676 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) {
1677   // Default column width for names is 13 even if no names are that long.
1678   size_t MaxWidth = 13;
1679   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1680     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1681     MaxWidth = std::max(MaxWidth, Name.size());
1682   }
1683   return MaxWidth;
1684 }
1685 
1686 void objdump::printSectionHeaders(const ObjectFile *Obj) {
1687   size_t NameWidth = getMaxSectionNameWidth(Obj);
1688   size_t AddressWidth = 2 * Obj->getBytesInAddress();
1689   bool HasLMAColumn = shouldDisplayLMA(Obj);
1690   if (HasLMAColumn)
1691     outs() << "Sections:\n"
1692               "Idx "
1693            << left_justify("Name", NameWidth) << " Size     "
1694            << left_justify("VMA", AddressWidth) << " "
1695            << left_justify("LMA", AddressWidth) << " Type\n";
1696   else
1697     outs() << "Sections:\n"
1698               "Idx "
1699            << left_justify("Name", NameWidth) << " Size     "
1700            << left_justify("VMA", AddressWidth) << " Type\n";
1701 
1702   uint64_t Idx;
1703   for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) {
1704     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1705     uint64_t VMA = Section.getAddress();
1706     if (shouldAdjustVA(Section))
1707       VMA += AdjustVMA;
1708 
1709     uint64_t Size = Section.getSize();
1710 
1711     std::string Type = Section.isText() ? "TEXT" : "";
1712     if (Section.isData())
1713       Type += Type.empty() ? "DATA" : " DATA";
1714     if (Section.isBSS())
1715       Type += Type.empty() ? "BSS" : " BSS";
1716 
1717     if (HasLMAColumn)
1718       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1719                        Name.str().c_str(), Size)
1720              << format_hex_no_prefix(VMA, AddressWidth) << " "
1721              << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth)
1722              << " " << Type << "\n";
1723     else
1724       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1725                        Name.str().c_str(), Size)
1726              << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n";
1727   }
1728   outs() << "\n";
1729 }
1730 
1731 void objdump::printSectionContents(const ObjectFile *Obj) {
1732   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
1733 
1734   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1735     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1736     uint64_t BaseAddr = Section.getAddress();
1737     uint64_t Size = Section.getSize();
1738     if (!Size)
1739       continue;
1740 
1741     outs() << "Contents of section ";
1742     StringRef SegmentName = getSegmentName(MachO, Section);
1743     if (!SegmentName.empty())
1744       outs() << SegmentName << ",";
1745     outs() << Name << ":\n";
1746     if (Section.isBSS()) {
1747       outs() << format("<skipping contents of bss section at [%04" PRIx64
1748                        ", %04" PRIx64 ")>\n",
1749                        BaseAddr, BaseAddr + Size);
1750       continue;
1751     }
1752 
1753     StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName());
1754 
1755     // Dump out the content as hex and printable ascii characters.
1756     for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) {
1757       outs() << format(" %04" PRIx64 " ", BaseAddr + Addr);
1758       // Dump line of hex.
1759       for (std::size_t I = 0; I < 16; ++I) {
1760         if (I != 0 && I % 4 == 0)
1761           outs() << ' ';
1762         if (Addr + I < End)
1763           outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true)
1764                  << hexdigit(Contents[Addr + I] & 0xF, true);
1765         else
1766           outs() << "  ";
1767       }
1768       // Print ascii.
1769       outs() << "  ";
1770       for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) {
1771         if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF))
1772           outs() << Contents[Addr + I];
1773         else
1774           outs() << ".";
1775       }
1776       outs() << "\n";
1777     }
1778   }
1779 }
1780 
1781 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName,
1782                                StringRef ArchitectureName, bool DumpDynamic) {
1783   if (O->isCOFF() && !DumpDynamic) {
1784     outs() << "SYMBOL TABLE:\n";
1785     printCOFFSymbolTable(cast<const COFFObjectFile>(O));
1786     return;
1787   }
1788 
1789   const StringRef FileName = O->getFileName();
1790 
1791   if (!DumpDynamic) {
1792     outs() << "SYMBOL TABLE:\n";
1793     for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I)
1794       printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic);
1795     return;
1796   }
1797 
1798   outs() << "DYNAMIC SYMBOL TABLE:\n";
1799   if (!O->isELF()) {
1800     reportWarning(
1801         "this operation is not currently supported for this file format",
1802         FileName);
1803     return;
1804   }
1805 
1806   const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O);
1807   for (auto I = ELF->getDynamicSymbolIterators().begin();
1808        I != ELF->getDynamicSymbolIterators().end(); ++I)
1809     printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic);
1810 }
1811 
1812 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol,
1813                           StringRef FileName, StringRef ArchiveName,
1814                           StringRef ArchitectureName, bool DumpDynamic) {
1815   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O);
1816   uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName,
1817                                    ArchitectureName);
1818   if ((Address < StartAddress) || (Address > StopAddress))
1819     return;
1820   SymbolRef::Type Type =
1821       unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName);
1822   uint32_t Flags =
1823       unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName);
1824 
1825   // Don't ask a Mach-O STAB symbol for its section unless you know that
1826   // STAB symbol's section field refers to a valid section index. Otherwise
1827   // the symbol may error trying to load a section that does not exist.
1828   bool IsSTAB = false;
1829   if (MachO) {
1830     DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1831     uint8_t NType =
1832         (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type
1833                           : MachO->getSymbolTableEntry(SymDRI).n_type);
1834     if (NType & MachO::N_STAB)
1835       IsSTAB = true;
1836   }
1837   section_iterator Section = IsSTAB
1838                                  ? O->section_end()
1839                                  : unwrapOrError(Symbol.getSection(), FileName,
1840                                                  ArchiveName, ArchitectureName);
1841 
1842   StringRef Name;
1843   if (Type == SymbolRef::ST_Debug && Section != O->section_end()) {
1844     if (Expected<StringRef> NameOrErr = Section->getName())
1845       Name = *NameOrErr;
1846     else
1847       consumeError(NameOrErr.takeError());
1848 
1849   } else {
1850     Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName,
1851                          ArchitectureName);
1852   }
1853 
1854   bool Global = Flags & SymbolRef::SF_Global;
1855   bool Weak = Flags & SymbolRef::SF_Weak;
1856   bool Absolute = Flags & SymbolRef::SF_Absolute;
1857   bool Common = Flags & SymbolRef::SF_Common;
1858   bool Hidden = Flags & SymbolRef::SF_Hidden;
1859 
1860   char GlobLoc = ' ';
1861   if ((Section != O->section_end() || Absolute) && !Weak)
1862     GlobLoc = Global ? 'g' : 'l';
1863   char IFunc = ' ';
1864   if (O->isELF()) {
1865     if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC)
1866       IFunc = 'i';
1867     if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE)
1868       GlobLoc = 'u';
1869   }
1870 
1871   char Debug = ' ';
1872   if (DumpDynamic)
1873     Debug = 'D';
1874   else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
1875     Debug = 'd';
1876 
1877   char FileFunc = ' ';
1878   if (Type == SymbolRef::ST_File)
1879     FileFunc = 'f';
1880   else if (Type == SymbolRef::ST_Function)
1881     FileFunc = 'F';
1882   else if (Type == SymbolRef::ST_Data)
1883     FileFunc = 'O';
1884 
1885   const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1886 
1887   outs() << format(Fmt, Address) << " "
1888          << GlobLoc            // Local -> 'l', Global -> 'g', Neither -> ' '
1889          << (Weak ? 'w' : ' ') // Weak?
1890          << ' '                // Constructor. Not supported yet.
1891          << ' '                // Warning. Not supported yet.
1892          << IFunc              // Indirect reference to another symbol.
1893          << Debug              // Debugging (d) or dynamic (D) symbol.
1894          << FileFunc           // Name of function (F), file (f) or object (O).
1895          << ' ';
1896   if (Absolute) {
1897     outs() << "*ABS*";
1898   } else if (Common) {
1899     outs() << "*COM*";
1900   } else if (Section == O->section_end()) {
1901     outs() << "*UND*";
1902   } else {
1903     StringRef SegmentName = getSegmentName(MachO, *Section);
1904     if (!SegmentName.empty())
1905       outs() << SegmentName << ",";
1906     StringRef SectionName = unwrapOrError(Section->getName(), FileName);
1907     outs() << SectionName;
1908   }
1909 
1910   if (Common || O->isELF()) {
1911     uint64_t Val =
1912         Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize();
1913     outs() << '\t' << format(Fmt, Val);
1914   }
1915 
1916   if (O->isELF()) {
1917     uint8_t Other = ELFSymbolRef(Symbol).getOther();
1918     switch (Other) {
1919     case ELF::STV_DEFAULT:
1920       break;
1921     case ELF::STV_INTERNAL:
1922       outs() << " .internal";
1923       break;
1924     case ELF::STV_HIDDEN:
1925       outs() << " .hidden";
1926       break;
1927     case ELF::STV_PROTECTED:
1928       outs() << " .protected";
1929       break;
1930     default:
1931       outs() << format(" 0x%02x", Other);
1932       break;
1933     }
1934   } else if (Hidden) {
1935     outs() << " .hidden";
1936   }
1937 
1938   if (Demangle)
1939     outs() << ' ' << demangle(std::string(Name)) << '\n';
1940   else
1941     outs() << ' ' << Name << '\n';
1942 }
1943 
1944 static void printUnwindInfo(const ObjectFile *O) {
1945   outs() << "Unwind info:\n\n";
1946 
1947   if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O))
1948     printCOFFUnwindInfo(Coff);
1949   else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O))
1950     printMachOUnwindInfo(MachO);
1951   else
1952     // TODO: Extract DWARF dump tool to objdump.
1953     WithColor::error(errs(), ToolName)
1954         << "This operation is only currently supported "
1955            "for COFF and MachO object files.\n";
1956 }
1957 
1958 /// Dump the raw contents of the __clangast section so the output can be piped
1959 /// into llvm-bcanalyzer.
1960 static void printRawClangAST(const ObjectFile *Obj) {
1961   if (outs().is_displayed()) {
1962     WithColor::error(errs(), ToolName)
1963         << "The -raw-clang-ast option will dump the raw binary contents of "
1964            "the clang ast section.\n"
1965            "Please redirect the output to a file or another program such as "
1966            "llvm-bcanalyzer.\n";
1967     return;
1968   }
1969 
1970   StringRef ClangASTSectionName("__clangast");
1971   if (Obj->isCOFF()) {
1972     ClangASTSectionName = "clangast";
1973   }
1974 
1975   Optional<object::SectionRef> ClangASTSection;
1976   for (auto Sec : ToolSectionFilter(*Obj)) {
1977     StringRef Name;
1978     if (Expected<StringRef> NameOrErr = Sec.getName())
1979       Name = *NameOrErr;
1980     else
1981       consumeError(NameOrErr.takeError());
1982 
1983     if (Name == ClangASTSectionName) {
1984       ClangASTSection = Sec;
1985       break;
1986     }
1987   }
1988   if (!ClangASTSection)
1989     return;
1990 
1991   StringRef ClangASTContents = unwrapOrError(
1992       ClangASTSection.getValue().getContents(), Obj->getFileName());
1993   outs().write(ClangASTContents.data(), ClangASTContents.size());
1994 }
1995 
1996 static void printFaultMaps(const ObjectFile *Obj) {
1997   StringRef FaultMapSectionName;
1998 
1999   if (Obj->isELF()) {
2000     FaultMapSectionName = ".llvm_faultmaps";
2001   } else if (Obj->isMachO()) {
2002     FaultMapSectionName = "__llvm_faultmaps";
2003   } else {
2004     WithColor::error(errs(), ToolName)
2005         << "This operation is only currently supported "
2006            "for ELF and Mach-O executable files.\n";
2007     return;
2008   }
2009 
2010   Optional<object::SectionRef> FaultMapSection;
2011 
2012   for (auto Sec : ToolSectionFilter(*Obj)) {
2013     StringRef Name;
2014     if (Expected<StringRef> NameOrErr = Sec.getName())
2015       Name = *NameOrErr;
2016     else
2017       consumeError(NameOrErr.takeError());
2018 
2019     if (Name == FaultMapSectionName) {
2020       FaultMapSection = Sec;
2021       break;
2022     }
2023   }
2024 
2025   outs() << "FaultMap table:\n";
2026 
2027   if (!FaultMapSection.hasValue()) {
2028     outs() << "<not found>\n";
2029     return;
2030   }
2031 
2032   StringRef FaultMapContents =
2033       unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName());
2034   FaultMapParser FMP(FaultMapContents.bytes_begin(),
2035                      FaultMapContents.bytes_end());
2036 
2037   outs() << FMP;
2038 }
2039 
2040 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) {
2041   if (O->isELF()) {
2042     printELFFileHeader(O);
2043     printELFDynamicSection(O);
2044     printELFSymbolVersionInfo(O);
2045     return;
2046   }
2047   if (O->isCOFF())
2048     return printCOFFFileHeader(O);
2049   if (O->isWasm())
2050     return printWasmFileHeader(O);
2051   if (O->isMachO()) {
2052     printMachOFileHeader(O);
2053     if (!OnlyFirst)
2054       printMachOLoadCommands(O);
2055     return;
2056   }
2057   reportError(O->getFileName(), "Invalid/Unsupported object file format");
2058 }
2059 
2060 static void printFileHeaders(const ObjectFile *O) {
2061   if (!O->isELF() && !O->isCOFF())
2062     reportError(O->getFileName(), "Invalid/Unsupported object file format");
2063 
2064   Triple::ArchType AT = O->getArch();
2065   outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n";
2066   uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName());
2067 
2068   StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2069   outs() << "start address: "
2070          << "0x" << format(Fmt.data(), Address) << "\n\n";
2071 }
2072 
2073 static void printArchiveChild(StringRef Filename, const Archive::Child &C) {
2074   Expected<sys::fs::perms> ModeOrErr = C.getAccessMode();
2075   if (!ModeOrErr) {
2076     WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n";
2077     consumeError(ModeOrErr.takeError());
2078     return;
2079   }
2080   sys::fs::perms Mode = ModeOrErr.get();
2081   outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
2082   outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
2083   outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
2084   outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
2085   outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
2086   outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
2087   outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
2088   outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
2089   outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
2090 
2091   outs() << " ";
2092 
2093   outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename),
2094                    unwrapOrError(C.getGID(), Filename),
2095                    unwrapOrError(C.getRawSize(), Filename));
2096 
2097   StringRef RawLastModified = C.getRawLastModified();
2098   unsigned Seconds;
2099   if (RawLastModified.getAsInteger(10, Seconds))
2100     outs() << "(date: \"" << RawLastModified
2101            << "\" contains non-decimal chars) ";
2102   else {
2103     // Since ctime(3) returns a 26 character string of the form:
2104     // "Sun Sep 16 01:03:52 1973\n\0"
2105     // just print 24 characters.
2106     time_t t = Seconds;
2107     outs() << format("%.24s ", ctime(&t));
2108   }
2109 
2110   StringRef Name = "";
2111   Expected<StringRef> NameOrErr = C.getName();
2112   if (!NameOrErr) {
2113     consumeError(NameOrErr.takeError());
2114     Name = unwrapOrError(C.getRawName(), Filename);
2115   } else {
2116     Name = NameOrErr.get();
2117   }
2118   outs() << Name << "\n";
2119 }
2120 
2121 // For ELF only now.
2122 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) {
2123   if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) {
2124     if (Elf->getEType() != ELF::ET_REL)
2125       return true;
2126   }
2127   return false;
2128 }
2129 
2130 static void checkForInvalidStartStopAddress(ObjectFile *Obj,
2131                                             uint64_t Start, uint64_t Stop) {
2132   if (!shouldWarnForInvalidStartStopAddress(Obj))
2133     return;
2134 
2135   for (const SectionRef &Section : Obj->sections())
2136     if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) {
2137       uint64_t BaseAddr = Section.getAddress();
2138       uint64_t Size = Section.getSize();
2139       if ((Start < BaseAddr + Size) && Stop > BaseAddr)
2140         return;
2141     }
2142 
2143   if (!HasStartAddressFlag)
2144     reportWarning("no section has address less than 0x" +
2145                       Twine::utohexstr(Stop) + " specified by --stop-address",
2146                   Obj->getFileName());
2147   else if (!HasStopAddressFlag)
2148     reportWarning("no section has address greater than or equal to 0x" +
2149                       Twine::utohexstr(Start) + " specified by --start-address",
2150                   Obj->getFileName());
2151   else
2152     reportWarning("no section overlaps the range [0x" +
2153                       Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) +
2154                       ") specified by --start-address/--stop-address",
2155                   Obj->getFileName());
2156 }
2157 
2158 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
2159                        const Archive::Child *C = nullptr) {
2160   // Avoid other output when using a raw option.
2161   if (!RawClangAST) {
2162     outs() << '\n';
2163     if (A)
2164       outs() << A->getFileName() << "(" << O->getFileName() << ")";
2165     else
2166       outs() << O->getFileName();
2167     outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n";
2168   }
2169 
2170   if (HasStartAddressFlag || HasStopAddressFlag)
2171     checkForInvalidStartStopAddress(O, StartAddress, StopAddress);
2172 
2173   // Note: the order here matches GNU objdump for compatability.
2174   StringRef ArchiveName = A ? A->getFileName() : "";
2175   if (ArchiveHeaders && !MachOOpt && C)
2176     printArchiveChild(ArchiveName, *C);
2177   if (FileHeaders)
2178     printFileHeaders(O);
2179   if (PrivateHeaders || FirstPrivateHeader)
2180     printPrivateFileHeaders(O, FirstPrivateHeader);
2181   if (SectionHeaders)
2182     printSectionHeaders(O);
2183   if (SymbolTable)
2184     printSymbolTable(O, ArchiveName);
2185   if (DynamicSymbolTable)
2186     printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"",
2187                      /*DumpDynamic=*/true);
2188   if (DwarfDumpType != DIDT_Null) {
2189     std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O);
2190     // Dump the complete DWARF structure.
2191     DIDumpOptions DumpOpts;
2192     DumpOpts.DumpType = DwarfDumpType;
2193     DICtx->dump(outs(), DumpOpts);
2194   }
2195   if (Relocations && !Disassemble)
2196     printRelocations(O);
2197   if (DynamicRelocations)
2198     printDynamicRelocations(O);
2199   if (SectionContents)
2200     printSectionContents(O);
2201   if (Disassemble)
2202     disassembleObject(O, Relocations);
2203   if (UnwindInfo)
2204     printUnwindInfo(O);
2205 
2206   // Mach-O specific options:
2207   if (ExportsTrie)
2208     printExportsTrie(O);
2209   if (Rebase)
2210     printRebaseTable(O);
2211   if (Bind)
2212     printBindTable(O);
2213   if (LazyBind)
2214     printLazyBindTable(O);
2215   if (WeakBind)
2216     printWeakBindTable(O);
2217 
2218   // Other special sections:
2219   if (RawClangAST)
2220     printRawClangAST(O);
2221   if (FaultMapSection)
2222     printFaultMaps(O);
2223 }
2224 
2225 static void dumpObject(const COFFImportFile *I, const Archive *A,
2226                        const Archive::Child *C = nullptr) {
2227   StringRef ArchiveName = A ? A->getFileName() : "";
2228 
2229   // Avoid other output when using a raw option.
2230   if (!RawClangAST)
2231     outs() << '\n'
2232            << ArchiveName << "(" << I->getFileName() << ")"
2233            << ":\tfile format COFF-import-file"
2234            << "\n\n";
2235 
2236   if (ArchiveHeaders && !MachOOpt && C)
2237     printArchiveChild(ArchiveName, *C);
2238   if (SymbolTable)
2239     printCOFFSymbolTable(I);
2240 }
2241 
2242 /// Dump each object file in \a a;
2243 static void dumpArchive(const Archive *A) {
2244   Error Err = Error::success();
2245   unsigned I = -1;
2246   for (auto &C : A->children(Err)) {
2247     ++I;
2248     Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
2249     if (!ChildOrErr) {
2250       if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError()))
2251         reportError(std::move(E), getFileNameForError(C, I), A->getFileName());
2252       continue;
2253     }
2254     if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
2255       dumpObject(O, A, &C);
2256     else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get()))
2257       dumpObject(I, A, &C);
2258     else
2259       reportError(errorCodeToError(object_error::invalid_file_type),
2260                   A->getFileName());
2261   }
2262   if (Err)
2263     reportError(std::move(Err), A->getFileName());
2264 }
2265 
2266 /// Open file and figure out how to dump it.
2267 static void dumpInput(StringRef file) {
2268   // If we are using the Mach-O specific object file parser, then let it parse
2269   // the file and process the command line options.  So the -arch flags can
2270   // be used to select specific slices, etc.
2271   if (MachOOpt) {
2272     parseInputMachO(file);
2273     return;
2274   }
2275 
2276   // Attempt to open the binary.
2277   OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file);
2278   Binary &Binary = *OBinary.getBinary();
2279 
2280   if (Archive *A = dyn_cast<Archive>(&Binary))
2281     dumpArchive(A);
2282   else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary))
2283     dumpObject(O);
2284   else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary))
2285     parseInputMachO(UB);
2286   else
2287     reportError(errorCodeToError(object_error::invalid_file_type), file);
2288 }
2289 
2290 template <typename T>
2291 static void parseIntArg(const llvm::opt::InputArgList &InputArgs, int ID,
2292                         T &Value) {
2293   if (const opt::Arg *A = InputArgs.getLastArg(ID)) {
2294     StringRef V(A->getValue());
2295     if (!llvm::to_integer(V, Value, 0)) {
2296       reportCmdLineError(A->getSpelling() +
2297                          ": expected a non-negative integer, but got '" + V +
2298                          "'");
2299     }
2300   }
2301 }
2302 
2303 static std::vector<std::string>
2304 commaSeparatedValues(const llvm::opt::InputArgList &InputArgs, int ID) {
2305   std::vector<std::string> Values;
2306   for (StringRef Value : InputArgs.getAllArgValues(ID)) {
2307     llvm::SmallVector<StringRef, 2> SplitValues;
2308     llvm::SplitString(Value, SplitValues, ",");
2309     for (StringRef SplitValue : SplitValues)
2310       Values.push_back(SplitValue.str());
2311   }
2312   return Values;
2313 }
2314 
2315 static void parseOtoolOptions(const llvm::opt::InputArgList &InputArgs) {
2316   MachOOpt = true;
2317   FullLeadingAddr = true;
2318   PrintImmHex = true;
2319 
2320   ArchName = InputArgs.getLastArgValue(OTOOL_arch).str();
2321   LinkOptHints = InputArgs.hasArg(OTOOL_C);
2322   if (InputArgs.hasArg(OTOOL_d))
2323     FilterSections.push_back("__DATA,__data");
2324   DylibId = InputArgs.hasArg(OTOOL_D);
2325   UniversalHeaders = InputArgs.hasArg(OTOOL_f);
2326   DataInCode = InputArgs.hasArg(OTOOL_G);
2327   FirstPrivateHeader = InputArgs.hasArg(OTOOL_h);
2328   IndirectSymbols = InputArgs.hasArg(OTOOL_I);
2329   ShowRawInsn = InputArgs.hasArg(OTOOL_j);
2330   PrivateHeaders = InputArgs.hasArg(OTOOL_l);
2331   DylibsUsed = InputArgs.hasArg(OTOOL_L);
2332   MCPU = InputArgs.getLastArgValue(OTOOL_mcpu_EQ).str();
2333   ObjcMetaData = InputArgs.hasArg(OTOOL_o);
2334   DisSymName = InputArgs.getLastArgValue(OTOOL_p).str();
2335   InfoPlist = InputArgs.hasArg(OTOOL_P);
2336   Relocations = InputArgs.hasArg(OTOOL_r);
2337   if (const Arg *A = InputArgs.getLastArg(OTOOL_s)) {
2338     auto Filter = (A->getValue(0) + StringRef(",") + A->getValue(1)).str();
2339     FilterSections.push_back(Filter);
2340   }
2341   if (InputArgs.hasArg(OTOOL_t))
2342     FilterSections.push_back("__TEXT,__text");
2343   Verbose = InputArgs.hasArg(OTOOL_v) || InputArgs.hasArg(OTOOL_V) ||
2344             InputArgs.hasArg(OTOOL_o);
2345   SymbolicOperands = InputArgs.hasArg(OTOOL_V);
2346   if (InputArgs.hasArg(OTOOL_x))
2347     FilterSections.push_back(",__text");
2348   LeadingAddr = LeadingHeaders = !InputArgs.hasArg(OTOOL_X);
2349 
2350   InputFilenames = InputArgs.getAllArgValues(OTOOL_INPUT);
2351   if (InputFilenames.empty())
2352     reportCmdLineError("no input file");
2353 
2354   for (const Arg *A : InputArgs) {
2355     const Option &O = A->getOption();
2356     if (O.getGroup().isValid() && O.getGroup().getID() == OTOOL_grp_obsolete) {
2357       reportCmdLineWarning(O.getPrefixedName() +
2358                            " is obsolete and not implemented");
2359     }
2360   }
2361 }
2362 
2363 static void parseObjdumpOptions(const llvm::opt::InputArgList &InputArgs) {
2364   parseIntArg(InputArgs, OBJDUMP_adjust_vma_EQ, AdjustVMA);
2365   AllHeaders = InputArgs.hasArg(OBJDUMP_all_headers);
2366   ArchName = InputArgs.getLastArgValue(OBJDUMP_arch_name_EQ).str();
2367   ArchiveHeaders = InputArgs.hasArg(OBJDUMP_archive_headers);
2368   Demangle = InputArgs.hasArg(OBJDUMP_demangle);
2369   Disassemble = InputArgs.hasArg(OBJDUMP_disassemble);
2370   DisassembleAll = InputArgs.hasArg(OBJDUMP_disassemble_all);
2371   SymbolDescription = InputArgs.hasArg(OBJDUMP_symbol_description);
2372   DisassembleSymbols =
2373       commaSeparatedValues(InputArgs, OBJDUMP_disassemble_symbols_EQ);
2374   DisassembleZeroes = InputArgs.hasArg(OBJDUMP_disassemble_zeroes);
2375   DisassemblerOptions =
2376       commaSeparatedValues(InputArgs, OBJDUMP_disassembler_options_EQ);
2377   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_dwarf_EQ)) {
2378     DwarfDumpType =
2379         StringSwitch<DIDumpType>(A->getValue()).Case("frames", DIDT_DebugFrame);
2380   }
2381   DynamicRelocations = InputArgs.hasArg(OBJDUMP_dynamic_reloc);
2382   FaultMapSection = InputArgs.hasArg(OBJDUMP_fault_map_section);
2383   FileHeaders = InputArgs.hasArg(OBJDUMP_file_headers);
2384   SectionContents = InputArgs.hasArg(OBJDUMP_full_contents);
2385   PrintLines = InputArgs.hasArg(OBJDUMP_line_numbers);
2386   InputFilenames = InputArgs.getAllArgValues(OBJDUMP_INPUT);
2387   MachOOpt = InputArgs.hasArg(OBJDUMP_macho);
2388   MCPU = InputArgs.getLastArgValue(OBJDUMP_mcpu_EQ).str();
2389   MAttrs = commaSeparatedValues(InputArgs, OBJDUMP_mattr_EQ);
2390   ShowRawInsn = !InputArgs.hasArg(OBJDUMP_no_show_raw_insn);
2391   LeadingAddr = !InputArgs.hasArg(OBJDUMP_no_leading_addr);
2392   RawClangAST = InputArgs.hasArg(OBJDUMP_raw_clang_ast);
2393   Relocations = InputArgs.hasArg(OBJDUMP_reloc);
2394   PrintImmHex =
2395       InputArgs.hasFlag(OBJDUMP_print_imm_hex, OBJDUMP_no_print_imm_hex, false);
2396   PrivateHeaders = InputArgs.hasArg(OBJDUMP_private_headers);
2397   FilterSections = InputArgs.getAllArgValues(OBJDUMP_section_EQ);
2398   SectionHeaders = InputArgs.hasArg(OBJDUMP_section_headers);
2399   ShowLMA = InputArgs.hasArg(OBJDUMP_show_lma);
2400   PrintSource = InputArgs.hasArg(OBJDUMP_source);
2401   parseIntArg(InputArgs, OBJDUMP_start_address_EQ, StartAddress);
2402   HasStartAddressFlag = InputArgs.hasArg(OBJDUMP_start_address_EQ);
2403   parseIntArg(InputArgs, OBJDUMP_stop_address_EQ, StopAddress);
2404   HasStopAddressFlag = InputArgs.hasArg(OBJDUMP_stop_address_EQ);
2405   SymbolTable = InputArgs.hasArg(OBJDUMP_syms);
2406   SymbolizeOperands = InputArgs.hasArg(OBJDUMP_symbolize_operands);
2407   DynamicSymbolTable = InputArgs.hasArg(OBJDUMP_dynamic_syms);
2408   TripleName = InputArgs.getLastArgValue(OBJDUMP_triple_EQ).str();
2409   UnwindInfo = InputArgs.hasArg(OBJDUMP_unwind_info);
2410   Wide = InputArgs.hasArg(OBJDUMP_wide);
2411   Prefix = InputArgs.getLastArgValue(OBJDUMP_prefix).str();
2412   parseIntArg(InputArgs, OBJDUMP_prefix_strip, PrefixStrip);
2413   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_debug_vars_EQ)) {
2414     DbgVariables = StringSwitch<DebugVarsFormat>(A->getValue())
2415                        .Case("ascii", DVASCII)
2416                        .Case("unicode", DVUnicode);
2417   }
2418   parseIntArg(InputArgs, OBJDUMP_debug_vars_indent_EQ, DbgIndent);
2419 
2420   parseMachOOptions(InputArgs);
2421 
2422   // Handle options that get forwarded to cl::opt<>s in libraries.
2423   // FIXME: Depending on https://reviews.llvm.org/D84191#inline-946075 ,
2424   // hopefully remove this again.
2425   std::vector<const char *> LLVMArgs;
2426   LLVMArgs.push_back("llvm-objdump (LLVM option parsing)");
2427   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_x86_asm_syntax_att,
2428                                                OBJDUMP_x86_asm_syntax_intel)) {
2429     switch (A->getOption().getID()) {
2430     case OBJDUMP_x86_asm_syntax_att:
2431       LLVMArgs.push_back("--x86-asm-syntax=att");
2432       break;
2433     case OBJDUMP_x86_asm_syntax_intel:
2434       LLVMArgs.push_back("--x86-asm-syntax=intel");
2435       break;
2436     }
2437   }
2438   if (InputArgs.hasArg(OBJDUMP_mhvx))
2439     LLVMArgs.push_back("--mhvx");
2440   if (InputArgs.hasArg(OBJDUMP_mhvx_v66))
2441     LLVMArgs.push_back("--mhvx=v66");
2442   if (InputArgs.hasArg(OBJDUMP_mv60))
2443     LLVMArgs.push_back("--mv60");
2444   if (InputArgs.hasArg(OBJDUMP_mv65))
2445     LLVMArgs.push_back("--mv65");
2446   if (InputArgs.hasArg(OBJDUMP_mv66))
2447     LLVMArgs.push_back("--mv66");
2448   if (InputArgs.hasArg(OBJDUMP_mv67))
2449     LLVMArgs.push_back("--mv67");
2450   if (InputArgs.hasArg(OBJDUMP_mv67t))
2451     LLVMArgs.push_back("--mv67t");
2452   if (InputArgs.hasArg(OBJDUMP_riscv_no_aliases))
2453     LLVMArgs.push_back("--riscv-no-aliases");
2454   LLVMArgs.push_back(nullptr);
2455   llvm::cl::ParseCommandLineOptions(LLVMArgs.size() - 1, LLVMArgs.data());
2456 
2457   // objdump defaults to a.out if no filenames specified.
2458   if (InputFilenames.empty())
2459     InputFilenames.push_back("a.out");
2460 }
2461 
2462 int main(int argc, char **argv) {
2463   using namespace llvm;
2464   InitLLVM X(argc, argv);
2465 
2466   ToolName = argv[0];
2467   std::unique_ptr<CommonOptTable> T;
2468   OptSpecifier Unknown, HelpFlag, HelpHiddenFlag, VersionFlag;
2469 
2470   StringRef Stem = sys::path::stem(ToolName);
2471   auto Is = [=](StringRef Tool) {
2472     // We need to recognize the following filenames:
2473     //
2474     // llvm-objdump -> objdump
2475     // llvm-otool-10.exe -> otool
2476     // powerpc64-unknown-freebsd13-objdump -> objdump
2477     auto I = Stem.rfind_lower(Tool);
2478     return I != StringRef::npos &&
2479            (I + Tool.size() == Stem.size() || !isAlnum(Stem[I + Tool.size()]));
2480   };
2481   if (Is("otool")) {
2482     T = std::make_unique<OtoolOptTable>();
2483     Unknown = OTOOL_UNKNOWN;
2484     HelpFlag = OTOOL_help;
2485     HelpHiddenFlag = OTOOL_help_hidden;
2486     VersionFlag = OTOOL_version;
2487   } else {
2488     T = std::make_unique<ObjdumpOptTable>();
2489     Unknown = OBJDUMP_UNKNOWN;
2490     HelpFlag = OBJDUMP_help;
2491     HelpHiddenFlag = OBJDUMP_help_hidden;
2492     VersionFlag = OBJDUMP_version;
2493   }
2494 
2495   BumpPtrAllocator A;
2496   StringSaver Saver(A);
2497   opt::InputArgList InputArgs =
2498       T->parseArgs(argc, argv, Unknown, Saver,
2499                    [&](StringRef Msg) { reportCmdLineError(Msg); });
2500 
2501   if (InputArgs.size() == 0 || InputArgs.hasArg(HelpFlag)) {
2502     T->printHelp(ToolName);
2503     return 0;
2504   }
2505   if (InputArgs.hasArg(HelpHiddenFlag)) {
2506     T->printHelp(ToolName, /*show_hidden=*/true);
2507     return 0;
2508   }
2509 
2510   // Initialize targets and assembly printers/parsers.
2511   InitializeAllTargetInfos();
2512   InitializeAllTargetMCs();
2513   InitializeAllDisassemblers();
2514 
2515   if (InputArgs.hasArg(VersionFlag)) {
2516     cl::PrintVersionMessage();
2517     if (!Is("otool")) {
2518       outs() << '\n';
2519       TargetRegistry::printRegisteredTargetsForVersion(outs());
2520     }
2521     return 0;
2522   }
2523 
2524   if (Is("otool"))
2525     parseOtoolOptions(InputArgs);
2526   else
2527     parseObjdumpOptions(InputArgs);
2528 
2529   if (StartAddress >= StopAddress)
2530     reportCmdLineError("start address should be less than stop address");
2531 
2532   // Removes trailing separators from prefix.
2533   while (!Prefix.empty() && sys::path::is_separator(Prefix.back()))
2534     Prefix.pop_back();
2535 
2536   if (AllHeaders)
2537     ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations =
2538         SectionHeaders = SymbolTable = true;
2539 
2540   if (DisassembleAll || PrintSource || PrintLines ||
2541       !DisassembleSymbols.empty())
2542     Disassemble = true;
2543 
2544   if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null &&
2545       !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST &&
2546       !Relocations && !SectionHeaders && !SectionContents && !SymbolTable &&
2547       !DynamicSymbolTable && !UnwindInfo && !FaultMapSection &&
2548       !(MachOOpt &&
2549         (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie ||
2550          FirstPrivateHeader || FunctionStarts || IndirectSymbols || InfoPlist ||
2551          LazyBind || LinkOptHints || ObjcMetaData || Rebase || Rpaths ||
2552          UniversalHeaders || WeakBind || !FilterSections.empty()))) {
2553     T->printHelp(ToolName);
2554     return 2;
2555   }
2556 
2557   DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end());
2558 
2559   llvm::for_each(InputFilenames, dumpInput);
2560 
2561   warnOnNoMatchForSections();
2562 
2563   return EXIT_SUCCESS;
2564 }
2565