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 "llvm/ADT/Optional.h"
20 #include "llvm/ADT/STLExtras.h"
21 #include "llvm/ADT/StringExtras.h"
22 #include "llvm/ADT/StringSet.h"
23 #include "llvm/ADT/Triple.h"
24 #include "llvm/CodeGen/FaultMaps.h"
25 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
26 #include "llvm/DebugInfo/Symbolize/Symbolize.h"
27 #include "llvm/Demangle/Demangle.h"
28 #include "llvm/MC/MCAsmInfo.h"
29 #include "llvm/MC/MCContext.h"
30 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h"
32 #include "llvm/MC/MCInst.h"
33 #include "llvm/MC/MCInstPrinter.h"
34 #include "llvm/MC/MCInstrAnalysis.h"
35 #include "llvm/MC/MCInstrInfo.h"
36 #include "llvm/MC/MCObjectFileInfo.h"
37 #include "llvm/MC/MCRegisterInfo.h"
38 #include "llvm/MC/MCSubtargetInfo.h"
39 #include "llvm/Object/Archive.h"
40 #include "llvm/Object/COFF.h"
41 #include "llvm/Object/COFFImportFile.h"
42 #include "llvm/Object/ELFObjectFile.h"
43 #include "llvm/Object/MachO.h"
44 #include "llvm/Object/MachOUniversal.h"
45 #include "llvm/Object/ObjectFile.h"
46 #include "llvm/Object/Wasm.h"
47 #include "llvm/Support/Casting.h"
48 #include "llvm/Support/CommandLine.h"
49 #include "llvm/Support/Debug.h"
50 #include "llvm/Support/Errc.h"
51 #include "llvm/Support/FileSystem.h"
52 #include "llvm/Support/Format.h"
53 #include "llvm/Support/GraphWriter.h"
54 #include "llvm/Support/Host.h"
55 #include "llvm/Support/InitLLVM.h"
56 #include "llvm/Support/MemoryBuffer.h"
57 #include "llvm/Support/SourceMgr.h"
58 #include "llvm/Support/StringSaver.h"
59 #include "llvm/Support/TargetRegistry.h"
60 #include "llvm/Support/TargetSelect.h"
61 #include "llvm/Support/WithColor.h"
62 #include "llvm/Support/raw_ostream.h"
63 #include <algorithm>
64 #include <cctype>
65 #include <cstring>
66 #include <system_error>
67 #include <unordered_map>
68 #include <utility>
69 
70 using namespace llvm;
71 using namespace object;
72 
73 cl::opt<unsigned long long> AdjustVMA(
74     "adjust-vma",
75     cl::desc("Increase the displayed address by the specified offset"),
76     cl::value_desc("offset"), cl::init(0));
77 
78 cl::opt<bool>
79     llvm::AllHeaders("all-headers",
80                      cl::desc("Display all available header information"));
81 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"),
82                                  cl::NotHidden, cl::Grouping,
83                                  cl::aliasopt(AllHeaders));
84 
85 static cl::list<std::string>
86 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore);
87 
88 cl::opt<bool>
89 llvm::Disassemble("disassemble",
90   cl::desc("Display assembler mnemonics for the machine instructions"));
91 static cl::alias Disassembled("d", cl::desc("Alias for --disassemble"),
92                               cl::NotHidden, cl::Grouping,
93                               cl::aliasopt(Disassemble));
94 
95 cl::opt<bool>
96 llvm::DisassembleAll("disassemble-all",
97   cl::desc("Display assembler mnemonics for the machine instructions"));
98 static cl::alias DisassembleAlld("D", cl::desc("Alias for --disassemble-all"),
99                                  cl::NotHidden, cl::Grouping,
100                                  cl::aliasopt(DisassembleAll));
101 
102 cl::opt<bool> llvm::Demangle("demangle", cl::desc("Demangle symbols names"),
103                              cl::init(false));
104 
105 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"),
106                                cl::NotHidden, cl::Grouping,
107                                cl::aliasopt(llvm::Demangle));
108 
109 static cl::list<std::string>
110 DisassembleFunctions("disassemble-functions",
111                      cl::CommaSeparated,
112                      cl::desc("List of functions to disassemble"));
113 static StringSet<> DisasmFuncsSet;
114 
115 cl::opt<bool>
116 llvm::Relocations("reloc",
117                   cl::desc("Display the relocation entries in the file"));
118 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"),
119                                   cl::NotHidden, cl::Grouping,
120                                   cl::aliasopt(llvm::Relocations));
121 
122 cl::opt<bool>
123 llvm::DynamicRelocations("dynamic-reloc",
124   cl::desc("Display the dynamic relocation entries in the file"));
125 static cl::alias DynamicRelocationsd("R", cl::desc("Alias for --dynamic-reloc"),
126                                      cl::NotHidden, cl::Grouping,
127                                      cl::aliasopt(DynamicRelocations));
128 
129 cl::opt<bool>
130     llvm::SectionContents("full-contents",
131                           cl::desc("Display the content of each section"));
132 static cl::alias SectionContentsShort("s",
133                                       cl::desc("Alias for --full-contents"),
134                                       cl::NotHidden, cl::Grouping,
135                                       cl::aliasopt(SectionContents));
136 
137 cl::opt<bool> llvm::SymbolTable("syms", cl::desc("Display the symbol table"));
138 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"),
139                                   cl::NotHidden, cl::Grouping,
140                                   cl::aliasopt(llvm::SymbolTable));
141 
142 cl::opt<bool>
143 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols"));
144 
145 cl::opt<bool>
146 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info"));
147 
148 cl::opt<bool>
149 llvm::Bind("bind", cl::desc("Display mach-o binding info"));
150 
151 cl::opt<bool>
152 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info"));
153 
154 cl::opt<bool>
155 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info"));
156 
157 cl::opt<bool>
158 llvm::RawClangAST("raw-clang-ast",
159     cl::desc("Dump the raw binary contents of the clang AST section"));
160 
161 static cl::opt<bool>
162 MachOOpt("macho", cl::desc("Use MachO specific object file parser"));
163 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden,
164                         cl::Grouping, cl::aliasopt(MachOOpt));
165 
166 cl::opt<std::string>
167 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, "
168                                     "see -version for available targets"));
169 
170 cl::opt<std::string>
171 llvm::MCPU("mcpu",
172      cl::desc("Target a specific cpu type (-mcpu=help for details)"),
173      cl::value_desc("cpu-name"),
174      cl::init(""));
175 
176 cl::opt<std::string>
177 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, "
178                                 "see -version for available targets"));
179 
180 cl::opt<bool>
181 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the "
182                                                  "headers for each section."));
183 static cl::alias SectionHeadersShort("headers",
184                                      cl::desc("Alias for --section-headers"),
185                                      cl::NotHidden,
186                                      cl::aliasopt(SectionHeaders));
187 static cl::alias SectionHeadersShorter("h",
188                                        cl::desc("Alias for --section-headers"),
189                                        cl::NotHidden, cl::Grouping,
190                                        cl::aliasopt(SectionHeaders));
191 
192 static cl::opt<bool>
193     ShowLMA("show-lma",
194             cl::desc("Display LMA column when dumping ELF section headers"));
195 
196 cl::list<std::string>
197 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. "
198                                          "With -macho dump segment,section"));
199 cl::alias static FilterSectionsj("j", cl::desc("Alias for --section"),
200                                  cl::NotHidden,
201                                  cl::aliasopt(llvm::FilterSections));
202 
203 cl::list<std::string>
204 llvm::MAttrs("mattr",
205   cl::CommaSeparated,
206   cl::desc("Target specific attributes"),
207   cl::value_desc("a1,+a2,-a3,..."));
208 
209 cl::opt<bool>
210 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling "
211                                                  "instructions, do not print "
212                                                  "the instruction bytes."));
213 cl::opt<bool>
214 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address"));
215 
216 cl::opt<bool>
217 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information"));
218 
219 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"),
220                                  cl::NotHidden, cl::Grouping,
221                                  cl::aliasopt(UnwindInfo));
222 
223 cl::opt<bool>
224 llvm::PrivateHeaders("private-headers",
225                      cl::desc("Display format specific file headers"));
226 
227 cl::opt<bool>
228 llvm::FirstPrivateHeader("private-header",
229                          cl::desc("Display only the first format specific file "
230                                   "header"));
231 
232 static cl::alias PrivateHeadersShort("p",
233                                      cl::desc("Alias for --private-headers"),
234                                      cl::NotHidden, cl::Grouping,
235                                      cl::aliasopt(PrivateHeaders));
236 
237 cl::opt<bool> llvm::FileHeaders(
238     "file-headers",
239     cl::desc("Display the contents of the overall file header"));
240 
241 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"),
242                                   cl::NotHidden, cl::Grouping,
243                                   cl::aliasopt(FileHeaders));
244 
245 cl::opt<bool>
246     llvm::ArchiveHeaders("archive-headers",
247                          cl::desc("Display archive header information"));
248 
249 cl::alias ArchiveHeadersShort("a", cl::desc("Alias for --archive-headers"),
250                               cl::NotHidden, cl::Grouping,
251                               cl::aliasopt(ArchiveHeaders));
252 
253 cl::opt<bool>
254     llvm::PrintImmHex("print-imm-hex",
255                       cl::desc("Use hex format for immediate values"));
256 
257 cl::opt<bool> PrintFaultMaps("fault-map-section",
258                              cl::desc("Display contents of faultmap section"));
259 
260 cl::opt<DIDumpType> llvm::DwarfDumpType(
261     "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"),
262     cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")));
263 
264 cl::opt<bool> PrintSource(
265     "source",
266     cl::desc(
267         "Display source inlined with disassembly. Implies disassemble object"));
268 
269 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), cl::NotHidden,
270                            cl::Grouping, cl::aliasopt(PrintSource));
271 
272 cl::opt<bool> PrintLines("line-numbers",
273                          cl::desc("Display source line numbers with "
274                                   "disassembly. Implies disassemble object"));
275 
276 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"),
277                           cl::NotHidden, cl::Grouping,
278                           cl::aliasopt(PrintLines));
279 
280 cl::opt<unsigned long long>
281     StartAddress("start-address", cl::desc("Disassemble beginning at address"),
282                  cl::value_desc("address"), cl::init(0));
283 cl::opt<unsigned long long>
284     StopAddress("stop-address",
285                 cl::desc("Stop disassembly at address"),
286                 cl::value_desc("address"), cl::init(UINT64_MAX));
287 
288 cl::opt<bool> DisassembleZeroes(
289                 "disassemble-zeroes",
290                 cl::desc("Do not skip blocks of zeroes when disassembling"));
291 cl::alias DisassembleZeroesShort("z",
292                                  cl::desc("Alias for --disassemble-zeroes"),
293                                  cl::NotHidden, cl::Grouping,
294                                  cl::aliasopt(DisassembleZeroes));
295 
296 static cl::list<std::string>
297     DisassemblerOptions("disassembler-options",
298                         cl::desc("Pass target specific disassembler options"),
299                         cl::value_desc("options"), cl::CommaSeparated);
300 static cl::alias
301     DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"),
302                              cl::NotHidden, cl::Prefix, cl::CommaSeparated,
303                              cl::aliasopt(DisassemblerOptions));
304 
305 static StringRef ToolName;
306 
307 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy;
308 
309 SectionFilter llvm::ToolSectionFilter(llvm::object::ObjectFile const &O) {
310   return SectionFilter(
311       [](llvm::object::SectionRef const &S) {
312         if (FilterSections.empty())
313           return true;
314         llvm::StringRef String;
315         std::error_code error = S.getName(String);
316         if (error)
317           return false;
318         return is_contained(FilterSections, String);
319       },
320       O);
321 }
322 
323 void llvm::error(std::error_code EC) {
324   if (!EC)
325     return;
326   WithColor::error(errs(), ToolName)
327       << "reading file: " << EC.message() << ".\n";
328   errs().flush();
329   exit(1);
330 }
331 
332 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) {
333   WithColor::error(errs(), ToolName) << Message << ".\n";
334   errs().flush();
335   exit(1);
336 }
337 
338 void llvm::warn(StringRef Message) {
339   WithColor::warning(errs(), ToolName) << Message << ".\n";
340   errs().flush();
341 }
342 
343 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File,
344                                                 Twine Message) {
345   WithColor::error(errs(), ToolName)
346       << "'" << File << "': " << Message << ".\n";
347   exit(1);
348 }
349 
350 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File,
351                                                 std::error_code EC) {
352   assert(EC);
353   WithColor::error(errs(), ToolName)
354       << "'" << File << "': " << EC.message() << ".\n";
355   exit(1);
356 }
357 
358 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File,
359                                                 llvm::Error E) {
360   assert(E);
361   std::string Buf;
362   raw_string_ostream OS(Buf);
363   logAllUnhandledErrors(std::move(E), OS);
364   OS.flush();
365   WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf;
366   exit(1);
367 }
368 
369 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName,
370                                                 StringRef FileName,
371                                                 llvm::Error E,
372                                                 StringRef ArchitectureName) {
373   assert(E);
374   WithColor::error(errs(), ToolName);
375   if (ArchiveName != "")
376     errs() << ArchiveName << "(" << FileName << ")";
377   else
378     errs() << "'" << FileName << "'";
379   if (!ArchitectureName.empty())
380     errs() << " (for architecture " << ArchitectureName << ")";
381   std::string Buf;
382   raw_string_ostream OS(Buf);
383   logAllUnhandledErrors(std::move(E), OS);
384   OS.flush();
385   errs() << ": " << Buf;
386   exit(1);
387 }
388 
389 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName,
390                                                 const object::Archive::Child &C,
391                                                 llvm::Error E,
392                                                 StringRef ArchitectureName) {
393   Expected<StringRef> NameOrErr = C.getName();
394   // TODO: if we have a error getting the name then it would be nice to print
395   // the index of which archive member this is and or its offset in the
396   // archive instead of "???" as the name.
397   if (!NameOrErr) {
398     consumeError(NameOrErr.takeError());
399     llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName);
400   } else
401     llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E),
402                        ArchitectureName);
403 }
404 
405 static const Target *getTarget(const ObjectFile *Obj = nullptr) {
406   // Figure out the target triple.
407   llvm::Triple TheTriple("unknown-unknown-unknown");
408   if (TripleName.empty()) {
409     if (Obj)
410       TheTriple = Obj->makeTriple();
411   } else {
412     TheTriple.setTriple(Triple::normalize(TripleName));
413 
414     // Use the triple, but also try to combine with ARM build attributes.
415     if (Obj) {
416       auto Arch = Obj->getArch();
417       if (Arch == Triple::arm || Arch == Triple::armeb)
418         Obj->setARMSubArch(TheTriple);
419     }
420   }
421 
422   // Get the target specific parser.
423   std::string Error;
424   const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple,
425                                                          Error);
426   if (!TheTarget) {
427     if (Obj)
428       report_error(Obj->getFileName(), "can't find target: " + Error);
429     else
430       error("can't find target: " + Error);
431   }
432 
433   // Update the triple name and return the found target.
434   TripleName = TheTriple.getTriple();
435   return TheTarget;
436 }
437 
438 bool llvm::isRelocAddressLess(RelocationRef A, RelocationRef B) {
439   return A.getOffset() < B.getOffset();
440 }
441 
442 static std::error_code getRelocationValueString(const RelocationRef &Rel,
443                                                 SmallVectorImpl<char> &Result) {
444   const ObjectFile *Obj = Rel.getObject();
445   if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj))
446     return getELFRelocationValueString(ELF, Rel, Result);
447   if (auto *COFF = dyn_cast<COFFObjectFile>(Obj))
448     return getCOFFRelocationValueString(COFF, Rel, Result);
449   if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj))
450     return getWasmRelocationValueString(Wasm, Rel, Result);
451   if (auto *MachO = dyn_cast<MachOObjectFile>(Obj))
452     return getMachORelocationValueString(MachO, Rel, Result);
453   llvm_unreachable("unknown object file format");
454 }
455 
456 /// Indicates whether this relocation should hidden when listing
457 /// relocations, usually because it is the trailing part of a multipart
458 /// relocation that will be printed as part of the leading relocation.
459 static bool getHidden(RelocationRef RelRef) {
460   auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject());
461   if (!MachO)
462     return false;
463 
464   unsigned Arch = MachO->getArch();
465   DataRefImpl Rel = RelRef.getRawDataRefImpl();
466   uint64_t Type = MachO->getRelocationType(Rel);
467 
468   // On arches that use the generic relocations, GENERIC_RELOC_PAIR
469   // is always hidden.
470   if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc)
471     return Type == MachO::GENERIC_RELOC_PAIR;
472 
473   if (Arch == Triple::x86_64) {
474     // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
475     // an X86_64_RELOC_SUBTRACTOR.
476     if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) {
477       DataRefImpl RelPrev = Rel;
478       RelPrev.d.a--;
479       uint64_t PrevType = MachO->getRelocationType(RelPrev);
480       if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR)
481         return true;
482     }
483   }
484 
485   return false;
486 }
487 
488 namespace {
489 class SourcePrinter {
490 protected:
491   DILineInfo OldLineInfo;
492   const ObjectFile *Obj = nullptr;
493   std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer;
494   // File name to file contents of source
495   std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache;
496   // Mark the line endings of the cached source
497   std::unordered_map<std::string, std::vector<StringRef>> LineCache;
498 
499 private:
500   bool cacheSource(const DILineInfo& LineInfoFile);
501 
502 public:
503   SourcePrinter() = default;
504   SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) {
505     symbolize::LLVMSymbolizer::Options SymbolizerOpts(
506         DILineInfoSpecifier::FunctionNameKind::None, true, false, false,
507         DefaultArch);
508     Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts));
509   }
510   virtual ~SourcePrinter() = default;
511   virtual void printSourceLine(raw_ostream &OS,
512                                object::SectionedAddress Address,
513                                StringRef Delimiter = "; ");
514 };
515 
516 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) {
517   std::unique_ptr<MemoryBuffer> Buffer;
518   if (LineInfo.Source) {
519     Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source);
520   } else {
521     auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName);
522     if (!BufferOrError)
523       return false;
524     Buffer = std::move(*BufferOrError);
525   }
526   // Chomp the file to get lines
527   size_t BufferSize = Buffer->getBufferSize();
528   const char *BufferStart = Buffer->getBufferStart();
529   for (const char *Start = BufferStart, *End = BufferStart;
530        End < BufferStart + BufferSize; End++)
531     if (*End == '\n' || End == BufferStart + BufferSize - 1 ||
532         (*End == '\r' && *(End + 1) == '\n')) {
533       LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start));
534       if (*End == '\r')
535         End++;
536       Start = End + 1;
537     }
538   SourceCache[LineInfo.FileName] = std::move(Buffer);
539   return true;
540 }
541 
542 void SourcePrinter::printSourceLine(raw_ostream &OS,
543                                     object::SectionedAddress Address,
544                                     StringRef Delimiter) {
545   if (!Symbolizer)
546     return;
547   DILineInfo LineInfo = DILineInfo();
548   auto ExpectecLineInfo =
549       Symbolizer->symbolizeCode(Obj->getFileName(), Address);
550   if (!ExpectecLineInfo)
551     consumeError(ExpectecLineInfo.takeError());
552   else
553     LineInfo = *ExpectecLineInfo;
554 
555   if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line ||
556       LineInfo.Line == 0)
557     return;
558 
559   if (PrintLines)
560     OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n";
561   if (PrintSource) {
562     if (SourceCache.find(LineInfo.FileName) == SourceCache.end())
563       if (!cacheSource(LineInfo))
564         return;
565     auto FileBuffer = SourceCache.find(LineInfo.FileName);
566     if (FileBuffer != SourceCache.end()) {
567       auto LineBuffer = LineCache.find(LineInfo.FileName);
568       if (LineBuffer != LineCache.end()) {
569         if (LineInfo.Line > LineBuffer->second.size())
570           return;
571         // Vector begins at 0, line numbers are non-zero
572         OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim()
573            << "\n";
574       }
575     }
576   }
577   OldLineInfo = LineInfo;
578 }
579 
580 static bool isArmElf(const ObjectFile *Obj) {
581   return (Obj->isELF() &&
582           (Obj->getArch() == Triple::aarch64 ||
583            Obj->getArch() == Triple::aarch64_be ||
584            Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb ||
585            Obj->getArch() == Triple::thumb ||
586            Obj->getArch() == Triple::thumbeb));
587 }
588 
589 static void printRelocation(const RelocationRef &Rel, uint64_t Address,
590                             uint8_t AddrSize) {
591   StringRef Fmt =
592       AddrSize > 4 ? "\t\t%016" PRIx64 ":  " : "\t\t\t%08" PRIx64 ":  ";
593   SmallString<16> Name;
594   SmallString<32> Val;
595   Rel.getTypeName(Name);
596   error(getRelocationValueString(Rel, Val));
597   outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n";
598 }
599 
600 class PrettyPrinter {
601 public:
602   virtual ~PrettyPrinter() = default;
603   virtual void printInst(MCInstPrinter &IP, const MCInst *MI,
604                          ArrayRef<uint8_t> Bytes,
605                          object::SectionedAddress Address, raw_ostream &OS,
606                          StringRef Annot, MCSubtargetInfo const &STI,
607                          SourcePrinter *SP,
608                          std::vector<RelocationRef> *Rels = nullptr) {
609     if (SP && (PrintSource || PrintLines))
610       SP->printSourceLine(OS, Address);
611     if (!NoLeadingAddr)
612       OS << format("%8" PRIx64 ":", Address.Address);
613     if (!NoShowRawInsn) {
614       OS << "\t";
615       dumpBytes(Bytes, OS);
616     }
617     if (MI)
618       IP.printInst(MI, OS, "", STI);
619     else
620       OS << " <unknown>";
621   }
622 };
623 PrettyPrinter PrettyPrinterInst;
624 class HexagonPrettyPrinter : public PrettyPrinter {
625 public:
626   void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
627                  raw_ostream &OS) {
628     uint32_t opcode =
629       (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
630     if (!NoLeadingAddr)
631       OS << format("%8" PRIx64 ":", Address);
632     if (!NoShowRawInsn) {
633       OS << "\t";
634       dumpBytes(Bytes.slice(0, 4), OS);
635       OS << format("%08" PRIx32, opcode);
636     }
637   }
638   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
639                  object::SectionedAddress Address, raw_ostream &OS,
640                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
641                  std::vector<RelocationRef> *Rels) override {
642     if (SP && (PrintSource || PrintLines))
643       SP->printSourceLine(OS, Address, "");
644     if (!MI) {
645       printLead(Bytes, Address.Address, OS);
646       OS << " <unknown>";
647       return;
648     }
649     std::string Buffer;
650     {
651       raw_string_ostream TempStream(Buffer);
652       IP.printInst(MI, TempStream, "", STI);
653     }
654     StringRef Contents(Buffer);
655     // Split off bundle attributes
656     auto PacketBundle = Contents.rsplit('\n');
657     // Split off first instruction from the rest
658     auto HeadTail = PacketBundle.first.split('\n');
659     auto Preamble = " { ";
660     auto Separator = "";
661 
662     // Hexagon's packets require relocations to be inline rather than
663     // clustered at the end of the packet.
664     std::vector<RelocationRef>::const_iterator RelCur = Rels->begin();
665     std::vector<RelocationRef>::const_iterator RelEnd = Rels->end();
666     auto PrintReloc = [&]() -> void {
667       while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) {
668         if (RelCur->getOffset() == Address.Address) {
669           printRelocation(*RelCur, Address.Address, 4);
670           return;
671         }
672         ++RelCur;
673       }
674     };
675 
676     while (!HeadTail.first.empty()) {
677       OS << Separator;
678       Separator = "\n";
679       if (SP && (PrintSource || PrintLines))
680         SP->printSourceLine(OS, Address, "");
681       printLead(Bytes, Address.Address, OS);
682       OS << Preamble;
683       Preamble = "   ";
684       StringRef Inst;
685       auto Duplex = HeadTail.first.split('\v');
686       if (!Duplex.second.empty()) {
687         OS << Duplex.first;
688         OS << "; ";
689         Inst = Duplex.second;
690       }
691       else
692         Inst = HeadTail.first;
693       OS << Inst;
694       HeadTail = HeadTail.second.split('\n');
695       if (HeadTail.first.empty())
696         OS << " } " << PacketBundle.second;
697       PrintReloc();
698       Bytes = Bytes.slice(4);
699       Address.Address += 4;
700     }
701   }
702 };
703 HexagonPrettyPrinter HexagonPrettyPrinterInst;
704 
705 class AMDGCNPrettyPrinter : public PrettyPrinter {
706 public:
707   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
708                  object::SectionedAddress Address, raw_ostream &OS,
709                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
710                  std::vector<RelocationRef> *Rels) override {
711     if (SP && (PrintSource || PrintLines))
712       SP->printSourceLine(OS, Address);
713 
714     typedef support::ulittle32_t U32;
715 
716     if (MI) {
717       SmallString<40> InstStr;
718       raw_svector_ostream IS(InstStr);
719 
720       IP.printInst(MI, IS, "", STI);
721 
722       OS << left_justify(IS.str(), 60);
723     } else {
724       // an unrecognized encoding - this is probably data so represent it
725       // using the .long directive, or .byte directive if fewer than 4 bytes
726       // remaining
727       if (Bytes.size() >= 4) {
728         OS << format("\t.long 0x%08" PRIx32 " ",
729                      static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data())));
730         OS.indent(42);
731       } else {
732           OS << format("\t.byte 0x%02" PRIx8, Bytes[0]);
733           for (unsigned int i = 1; i < Bytes.size(); i++)
734             OS << format(", 0x%02" PRIx8, Bytes[i]);
735           OS.indent(55 - (6 * Bytes.size()));
736       }
737     }
738 
739     OS << format("// %012" PRIX64 ": ", Address.Address);
740     if (Bytes.size() >=4) {
741       for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()),
742                                  Bytes.size() / sizeof(U32)))
743         // D should be explicitly casted to uint32_t here as it is passed
744         // by format to snprintf as vararg.
745         OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D));
746     } else {
747       for (unsigned int i = 0; i < Bytes.size(); i++)
748         OS << format("%02" PRIX8 " ", Bytes[i]);
749     }
750 
751     if (!Annot.empty())
752       OS << "// " << Annot;
753   }
754 };
755 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst;
756 
757 class BPFPrettyPrinter : public PrettyPrinter {
758 public:
759   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
760                  object::SectionedAddress Address, raw_ostream &OS,
761                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
762                  std::vector<RelocationRef> *Rels) override {
763     if (SP && (PrintSource || PrintLines))
764       SP->printSourceLine(OS, Address);
765     if (!NoLeadingAddr)
766       OS << format("%8" PRId64 ":", Address.Address / 8);
767     if (!NoShowRawInsn) {
768       OS << "\t";
769       dumpBytes(Bytes, OS);
770     }
771     if (MI)
772       IP.printInst(MI, OS, "", STI);
773     else
774       OS << " <unknown>";
775   }
776 };
777 BPFPrettyPrinter BPFPrettyPrinterInst;
778 
779 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) {
780   switch(Triple.getArch()) {
781   default:
782     return PrettyPrinterInst;
783   case Triple::hexagon:
784     return HexagonPrettyPrinterInst;
785   case Triple::amdgcn:
786     return AMDGCNPrettyPrinterInst;
787   case Triple::bpfel:
788   case Triple::bpfeb:
789     return BPFPrettyPrinterInst;
790   }
791 }
792 }
793 
794 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) {
795   assert(Obj->isELF());
796   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
797     return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
798   if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
799     return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
800   if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
801     return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
802   if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
803     return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType();
804   llvm_unreachable("Unsupported binary format");
805 }
806 
807 template <class ELFT> static void
808 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj,
809                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
810   for (auto Symbol : Obj->getDynamicSymbolIterators()) {
811     uint8_t SymbolType = Symbol.getELFType();
812     if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0)
813       continue;
814 
815     Expected<uint64_t> AddressOrErr = Symbol.getAddress();
816     if (!AddressOrErr)
817       report_error(Obj->getFileName(), AddressOrErr.takeError());
818 
819     Expected<StringRef> Name = Symbol.getName();
820     if (!Name)
821       report_error(Obj->getFileName(), Name.takeError());
822     if (Name->empty())
823       continue;
824 
825     Expected<section_iterator> SectionOrErr = Symbol.getSection();
826     if (!SectionOrErr)
827       report_error(Obj->getFileName(), SectionOrErr.takeError());
828     section_iterator SecI = *SectionOrErr;
829     if (SecI == Obj->section_end())
830       continue;
831 
832     AllSymbols[*SecI].emplace_back(*AddressOrErr, *Name, SymbolType);
833   }
834 }
835 
836 static void
837 addDynamicElfSymbols(const ObjectFile *Obj,
838                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
839   assert(Obj->isELF());
840   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
841     addDynamicElfSymbols(Elf32LEObj, AllSymbols);
842   else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
843     addDynamicElfSymbols(Elf64LEObj, AllSymbols);
844   else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
845     addDynamicElfSymbols(Elf32BEObj, AllSymbols);
846   else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
847     addDynamicElfSymbols(Elf64BEObj, AllSymbols);
848   else
849     llvm_unreachable("Unsupported binary format");
850 }
851 
852 static void addPltEntries(const ObjectFile *Obj,
853                           std::map<SectionRef, SectionSymbolsTy> &AllSymbols,
854                           StringSaver &Saver) {
855   Optional<SectionRef> Plt = None;
856   for (const SectionRef &Section : Obj->sections()) {
857     StringRef Name;
858     if (Section.getName(Name))
859       continue;
860     if (Name == ".plt")
861       Plt = Section;
862   }
863   if (!Plt)
864     return;
865   if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) {
866     for (auto PltEntry : ElfObj->getPltAddresses()) {
867       SymbolRef Symbol(PltEntry.first, ElfObj);
868       uint8_t SymbolType = getElfSymbolType(Obj, Symbol);
869 
870       Expected<StringRef> NameOrErr = Symbol.getName();
871       if (!NameOrErr)
872         report_error(Obj->getFileName(), NameOrErr.takeError());
873       if (NameOrErr->empty())
874         continue;
875       StringRef Name = Saver.save((*NameOrErr + "@plt").str());
876 
877       AllSymbols[*Plt].emplace_back(PltEntry.second, Name, SymbolType);
878     }
879   }
880 }
881 
882 // Normally the disassembly output will skip blocks of zeroes. This function
883 // returns the number of zero bytes that can be skipped when dumping the
884 // disassembly of the instructions in Buf.
885 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) {
886   // Find the number of leading zeroes.
887   size_t N = 0;
888   while (N < Buf.size() && !Buf[N])
889     ++N;
890 
891   // We may want to skip blocks of zero bytes, but unless we see
892   // at least 8 of them in a row.
893   if (N < 8)
894     return 0;
895 
896   // We skip zeroes in multiples of 4 because do not want to truncate an
897   // instruction if it starts with a zero byte.
898   return N & ~0x3;
899 }
900 
901 // Returns a map from sections to their relocations.
902 static std::map<SectionRef, std::vector<RelocationRef>>
903 getRelocsMap(llvm::object::ObjectFile const &Obj) {
904   std::map<SectionRef, std::vector<RelocationRef>> Ret;
905   for (const SectionRef &Section : ToolSectionFilter(Obj)) {
906     section_iterator RelSec = Section.getRelocatedSection();
907     if (RelSec == Obj.section_end())
908       continue;
909     std::vector<RelocationRef> &V = Ret[*RelSec];
910     for (const RelocationRef &R : Section.relocations())
911       V.push_back(R);
912     // Sort relocations by address.
913     llvm::sort(V, isRelocAddressLess);
914   }
915   return Ret;
916 }
917 
918 // Used for --adjust-vma to check if address should be adjusted by the
919 // specified value for a given section.
920 // For ELF we do not adjust non-allocatable sections like debug ones,
921 // because they are not loadable.
922 // TODO: implement for other file formats.
923 static bool shouldAdjustVA(const SectionRef &Section) {
924   const ObjectFile *Obj = Section.getObject();
925   if (isa<object::ELFObjectFileBase>(Obj))
926     return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC;
927   return false;
928 }
929 
930 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj,
931                               MCContext &Ctx, MCDisassembler *DisAsm,
932                               const MCInstrAnalysis *MIA, MCInstPrinter *IP,
933                               const MCSubtargetInfo *STI, PrettyPrinter &PIP,
934                               SourcePrinter &SP, bool InlineRelocs) {
935   std::map<SectionRef, std::vector<RelocationRef>> RelocMap;
936   if (InlineRelocs)
937     RelocMap = getRelocsMap(*Obj);
938 
939   // Create a mapping from virtual address to symbol name.  This is used to
940   // pretty print the symbols while disassembling.
941   std::map<SectionRef, SectionSymbolsTy> AllSymbols;
942   SectionSymbolsTy AbsoluteSymbols;
943   for (const SymbolRef &Symbol : Obj->symbols()) {
944     Expected<uint64_t> AddressOrErr = Symbol.getAddress();
945     if (!AddressOrErr)
946       report_error(Obj->getFileName(), AddressOrErr.takeError());
947     uint64_t Address = *AddressOrErr;
948 
949     Expected<StringRef> Name = Symbol.getName();
950     if (!Name)
951       report_error(Obj->getFileName(), Name.takeError());
952     if (Name->empty())
953       continue;
954 
955     Expected<section_iterator> SectionOrErr = Symbol.getSection();
956     if (!SectionOrErr)
957       report_error(Obj->getFileName(), SectionOrErr.takeError());
958 
959     uint8_t SymbolType = ELF::STT_NOTYPE;
960     if (Obj->isELF()) {
961       SymbolType = getElfSymbolType(Obj, Symbol);
962       if (SymbolType == ELF::STT_SECTION)
963         continue;
964     }
965 
966     section_iterator SecI = *SectionOrErr;
967     if (SecI != Obj->section_end())
968       AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType);
969     else
970       AbsoluteSymbols.emplace_back(Address, *Name, SymbolType);
971 
972 
973   }
974   if (AllSymbols.empty() && Obj->isELF())
975     addDynamicElfSymbols(Obj, AllSymbols);
976 
977   BumpPtrAllocator A;
978   StringSaver Saver(A);
979   addPltEntries(Obj, AllSymbols, Saver);
980 
981   // Create a mapping from virtual address to section.
982   std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses;
983   for (SectionRef Sec : Obj->sections())
984     SectionAddresses.emplace_back(Sec.getAddress(), Sec);
985   array_pod_sort(SectionAddresses.begin(), SectionAddresses.end());
986 
987   // Linked executables (.exe and .dll files) typically don't include a real
988   // symbol table but they might contain an export table.
989   if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) {
990     for (const auto &ExportEntry : COFFObj->export_directories()) {
991       StringRef Name;
992       error(ExportEntry.getSymbolName(Name));
993       if (Name.empty())
994         continue;
995       uint32_t RVA;
996       error(ExportEntry.getExportRVA(RVA));
997 
998       uint64_t VA = COFFObj->getImageBase() + RVA;
999       auto Sec = std::upper_bound(
1000           SectionAddresses.begin(), SectionAddresses.end(), VA,
1001           [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) {
1002             return LHS < RHS.first;
1003           });
1004       if (Sec != SectionAddresses.begin())
1005         --Sec;
1006       else
1007         Sec = SectionAddresses.end();
1008 
1009       if (Sec != SectionAddresses.end())
1010         AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE);
1011       else
1012         AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE);
1013     }
1014   }
1015 
1016   // Sort all the symbols, this allows us to use a simple binary search to find
1017   // a symbol near an address.
1018   for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols)
1019     array_pod_sort(SecSyms.second.begin(), SecSyms.second.end());
1020   array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end());
1021 
1022   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1023     if (!DisassembleAll && (!Section.isText() || Section.isVirtual()))
1024       continue;
1025 
1026     uint64_t SectionAddr = Section.getAddress();
1027     uint64_t SectSize = Section.getSize();
1028     if (!SectSize)
1029       continue;
1030 
1031     // Get the list of all the symbols in this section.
1032     SectionSymbolsTy &Symbols = AllSymbols[Section];
1033     std::vector<uint64_t> DataMappingSymsAddr;
1034     std::vector<uint64_t> TextMappingSymsAddr;
1035     if (isArmElf(Obj)) {
1036       for (const auto &Symb : Symbols) {
1037         uint64_t Address = std::get<0>(Symb);
1038         StringRef Name = std::get<1>(Symb);
1039         if (Name.startswith("$d"))
1040           DataMappingSymsAddr.push_back(Address - SectionAddr);
1041         if (Name.startswith("$x"))
1042           TextMappingSymsAddr.push_back(Address - SectionAddr);
1043         if (Name.startswith("$a"))
1044           TextMappingSymsAddr.push_back(Address - SectionAddr);
1045         if (Name.startswith("$t"))
1046           TextMappingSymsAddr.push_back(Address - SectionAddr);
1047       }
1048     }
1049 
1050     llvm::sort(DataMappingSymsAddr);
1051     llvm::sort(TextMappingSymsAddr);
1052 
1053     if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) {
1054       // AMDGPU disassembler uses symbolizer for printing labels
1055       std::unique_ptr<MCRelocationInfo> RelInfo(
1056         TheTarget->createMCRelocationInfo(TripleName, Ctx));
1057       if (RelInfo) {
1058         std::unique_ptr<MCSymbolizer> Symbolizer(
1059           TheTarget->createMCSymbolizer(
1060             TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1061         DisAsm->setSymbolizer(std::move(Symbolizer));
1062       }
1063     }
1064 
1065     StringRef SegmentName = "";
1066     if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) {
1067       DataRefImpl DR = Section.getRawDataRefImpl();
1068       SegmentName = MachO->getSectionFinalSegmentName(DR);
1069     }
1070     StringRef SectionName;
1071     error(Section.getName(SectionName));
1072 
1073     // If the section has no symbol at the start, just insert a dummy one.
1074     if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) {
1075       Symbols.insert(
1076           Symbols.begin(),
1077           std::make_tuple(SectionAddr, SectionName,
1078                           Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT));
1079     }
1080 
1081     SmallString<40> Comments;
1082     raw_svector_ostream CommentStream(Comments);
1083 
1084     StringRef BytesStr;
1085     error(Section.getContents(BytesStr));
1086     ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
1087                             BytesStr.size());
1088 
1089     uint64_t VMAAdjustment = 0;
1090     if (shouldAdjustVA(Section))
1091       VMAAdjustment = AdjustVMA;
1092 
1093     uint64_t Size;
1094     uint64_t Index;
1095     bool PrintedSection = false;
1096     std::vector<RelocationRef> Rels = RelocMap[Section];
1097     std::vector<RelocationRef>::const_iterator RelCur = Rels.begin();
1098     std::vector<RelocationRef>::const_iterator RelEnd = Rels.end();
1099     // Disassemble symbol by symbol.
1100     for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) {
1101       uint64_t Start = std::get<0>(Symbols[SI]) - SectionAddr;
1102       // The end is either the section end or the beginning of the next
1103       // symbol.
1104       uint64_t End = (SI == SE - 1)
1105                          ? SectSize
1106                          : std::get<0>(Symbols[SI + 1]) - SectionAddr;
1107       // Don't try to disassemble beyond the end of section contents.
1108       if (End > SectSize)
1109         End = SectSize;
1110       // If this symbol has the same address as the next symbol, then skip it.
1111       if (Start >= End)
1112         continue;
1113 
1114       // Check if we need to skip symbol
1115       // Skip if the symbol's data is not between StartAddress and StopAddress
1116       if (End + SectionAddr < StartAddress ||
1117           Start + SectionAddr > StopAddress) {
1118         continue;
1119       }
1120 
1121       /// Skip if user requested specific symbols and this is not in the list
1122       if (!DisasmFuncsSet.empty() &&
1123           !DisasmFuncsSet.count(std::get<1>(Symbols[SI])))
1124         continue;
1125 
1126       if (!PrintedSection) {
1127         PrintedSection = true;
1128         outs() << "Disassembly of section ";
1129         if (!SegmentName.empty())
1130           outs() << SegmentName << ",";
1131         outs() << SectionName << ':';
1132       }
1133 
1134       // Stop disassembly at the stop address specified
1135       if (End + SectionAddr > StopAddress)
1136         End = StopAddress - SectionAddr;
1137 
1138       if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) {
1139         if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) {
1140           // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes)
1141           Start += 256;
1142         }
1143         if (SI == SE - 1 ||
1144             std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) {
1145           // cut trailing zeroes at the end of kernel
1146           // cut up to 256 bytes
1147           const uint64_t EndAlign = 256;
1148           const auto Limit = End - (std::min)(EndAlign, End - Start);
1149           while (End > Limit &&
1150             *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0)
1151             End -= 4;
1152         }
1153       }
1154 
1155       outs() << '\n';
1156       if (!NoLeadingAddr)
1157         outs() << format("%016" PRIx64 " ",
1158                          SectionAddr + Start + VMAAdjustment);
1159 
1160       StringRef SymbolName = std::get<1>(Symbols[SI]);
1161       if (Demangle)
1162         outs() << demangle(SymbolName) << ":\n";
1163       else
1164         outs() << SymbolName << ":\n";
1165 
1166       // Don't print raw contents of a virtual section. A virtual section
1167       // doesn't have any contents in the file.
1168       if (Section.isVirtual()) {
1169         outs() << "...\n";
1170         continue;
1171       }
1172 
1173 #ifndef NDEBUG
1174       raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
1175 #else
1176       raw_ostream &DebugOut = nulls();
1177 #endif
1178 
1179       // Some targets (like WebAssembly) have a special prelude at the start
1180       // of each symbol.
1181       DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start),
1182                             SectionAddr + Start, DebugOut, CommentStream);
1183       Start += Size;
1184 
1185       for (Index = Start; Index < End; Index += Size) {
1186         MCInst Inst;
1187 
1188         if (Index + SectionAddr < StartAddress ||
1189             Index + SectionAddr > StopAddress) {
1190           // skip byte by byte till StartAddress is reached
1191           Size = 1;
1192           continue;
1193         }
1194         // AArch64 ELF binaries can interleave data and text in the
1195         // same section. We rely on the markers introduced to
1196         // understand what we need to dump. If the data marker is within a
1197         // function, it is denoted as a word/short etc
1198         if (isArmElf(Obj) && std::get<2>(Symbols[SI]) != ELF::STT_OBJECT &&
1199             !DisassembleAll) {
1200           uint64_t Stride = 0;
1201 
1202           auto DAI = std::lower_bound(DataMappingSymsAddr.begin(),
1203                                       DataMappingSymsAddr.end(), Index);
1204           if (DAI != DataMappingSymsAddr.end() && *DAI == Index) {
1205             // Switch to data.
1206             while (Index < End) {
1207               outs() << format("%8" PRIx64 ":", SectionAddr + Index);
1208               outs() << "\t";
1209               if (Index + 4 <= End) {
1210                 Stride = 4;
1211                 dumpBytes(Bytes.slice(Index, 4), outs());
1212                 outs() << "\t.word\t";
1213                 uint32_t Data = 0;
1214                 if (Obj->isLittleEndian()) {
1215                   const auto Word =
1216                       reinterpret_cast<const support::ulittle32_t *>(
1217                           Bytes.data() + Index);
1218                   Data = *Word;
1219                 } else {
1220                   const auto Word = reinterpret_cast<const support::ubig32_t *>(
1221                       Bytes.data() + Index);
1222                   Data = *Word;
1223                 }
1224                 outs() << "0x" << format("%08" PRIx32, Data);
1225               } else if (Index + 2 <= End) {
1226                 Stride = 2;
1227                 dumpBytes(Bytes.slice(Index, 2), outs());
1228                 outs() << "\t\t.short\t";
1229                 uint16_t Data = 0;
1230                 if (Obj->isLittleEndian()) {
1231                   const auto Short =
1232                       reinterpret_cast<const support::ulittle16_t *>(
1233                           Bytes.data() + Index);
1234                   Data = *Short;
1235                 } else {
1236                   const auto Short =
1237                       reinterpret_cast<const support::ubig16_t *>(Bytes.data() +
1238                                                                   Index);
1239                   Data = *Short;
1240                 }
1241                 outs() << "0x" << format("%04" PRIx16, Data);
1242               } else {
1243                 Stride = 1;
1244                 dumpBytes(Bytes.slice(Index, 1), outs());
1245                 outs() << "\t\t.byte\t";
1246                 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]);
1247               }
1248               Index += Stride;
1249               outs() << "\n";
1250               auto TAI = std::lower_bound(TextMappingSymsAddr.begin(),
1251                                           TextMappingSymsAddr.end(), Index);
1252               if (TAI != TextMappingSymsAddr.end() && *TAI == Index)
1253                 break;
1254             }
1255           }
1256         }
1257 
1258         // If there is a data symbol inside an ELF text section and we are only
1259         // disassembling text (applicable all architectures),
1260         // we are in a situation where we must print the data and not
1261         // disassemble it.
1262         if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT &&
1263             !DisassembleAll && Section.isText()) {
1264           // print out data up to 8 bytes at a time in hex and ascii
1265           uint8_t AsciiData[9] = {'\0'};
1266           uint8_t Byte;
1267           int NumBytes = 0;
1268 
1269           for (Index = Start; Index < End; Index += 1) {
1270             if (((SectionAddr + Index) < StartAddress) ||
1271                 ((SectionAddr + Index) > StopAddress))
1272               continue;
1273             if (NumBytes == 0) {
1274               outs() << format("%8" PRIx64 ":", SectionAddr + Index);
1275               outs() << "\t";
1276             }
1277             Byte = Bytes.slice(Index)[0];
1278             outs() << format(" %02x", Byte);
1279             AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.';
1280 
1281             uint8_t IndentOffset = 0;
1282             NumBytes++;
1283             if (Index == End - 1 || NumBytes > 8) {
1284               // Indent the space for less than 8 bytes data.
1285               // 2 spaces for byte and one for space between bytes
1286               IndentOffset = 3 * (8 - NumBytes);
1287               for (int Excess = NumBytes; Excess < 8; Excess++)
1288                 AsciiData[Excess] = '\0';
1289               NumBytes = 8;
1290             }
1291             if (NumBytes == 8) {
1292               AsciiData[8] = '\0';
1293               outs() << std::string(IndentOffset, ' ') << "         ";
1294               outs() << reinterpret_cast<char *>(AsciiData);
1295               outs() << '\n';
1296               NumBytes = 0;
1297             }
1298           }
1299         }
1300         if (Index >= End)
1301           break;
1302 
1303         // When -z or --disassemble-zeroes are given we always dissasemble them.
1304         // Otherwise we might want to skip zero bytes we see.
1305         if (!DisassembleZeroes) {
1306           uint64_t MaxOffset = End - Index;
1307           // For -reloc: print zero blocks patched by relocations, so that
1308           // relocations can be shown in the dump.
1309           if (RelCur != RelEnd)
1310             MaxOffset = RelCur->getOffset() - Index;
1311 
1312           if (size_t N =
1313                   countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) {
1314             outs() << "\t\t..." << '\n';
1315             Index += N;
1316             if (Index >= End)
1317               break;
1318           }
1319         }
1320 
1321         // Disassemble a real instruction or a data when disassemble all is
1322         // provided
1323         bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
1324                                                    SectionAddr + Index, DebugOut,
1325                                                    CommentStream);
1326         if (Size == 0)
1327           Size = 1;
1328 
1329         PIP.printInst(*IP, Disassembled ? &Inst : nullptr,
1330                       Bytes.slice(Index, Size),
1331                       {SectionAddr + Index + VMAAdjustment, Section.getIndex()},
1332                       outs(), "", *STI, &SP, &Rels);
1333         outs() << CommentStream.str();
1334         Comments.clear();
1335 
1336         // Try to resolve the target of a call, tail call, etc. to a specific
1337         // symbol.
1338         if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) ||
1339                     MIA->isConditionalBranch(Inst))) {
1340           uint64_t Target;
1341           if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) {
1342             // In a relocatable object, the target's section must reside in
1343             // the same section as the call instruction or it is accessed
1344             // through a relocation.
1345             //
1346             // In a non-relocatable object, the target may be in any section.
1347             //
1348             // N.B. We don't walk the relocations in the relocatable case yet.
1349             auto *TargetSectionSymbols = &Symbols;
1350             if (!Obj->isRelocatableObject()) {
1351               auto SectionAddress = std::upper_bound(
1352                   SectionAddresses.begin(), SectionAddresses.end(), Target,
1353                   [](uint64_t LHS,
1354                       const std::pair<uint64_t, SectionRef> &RHS) {
1355                     return LHS < RHS.first;
1356                   });
1357               if (SectionAddress != SectionAddresses.begin()) {
1358                 --SectionAddress;
1359                 TargetSectionSymbols = &AllSymbols[SectionAddress->second];
1360               } else {
1361                 TargetSectionSymbols = &AbsoluteSymbols;
1362               }
1363             }
1364 
1365             // Find the first symbol in the section whose offset is less than
1366             // or equal to the target. If there isn't a section that contains
1367             // the target, find the nearest preceding absolute symbol.
1368             auto TargetSym = std::upper_bound(
1369                 TargetSectionSymbols->begin(), TargetSectionSymbols->end(),
1370                 Target, [](uint64_t LHS,
1371                            const std::tuple<uint64_t, StringRef, uint8_t> &RHS) {
1372                   return LHS < std::get<0>(RHS);
1373                 });
1374             if (TargetSym == TargetSectionSymbols->begin()) {
1375               TargetSectionSymbols = &AbsoluteSymbols;
1376               TargetSym = std::upper_bound(
1377                   AbsoluteSymbols.begin(), AbsoluteSymbols.end(),
1378                   Target, [](uint64_t LHS,
1379                              const std::tuple<uint64_t, StringRef, uint8_t> &RHS) {
1380                             return LHS < std::get<0>(RHS);
1381                           });
1382             }
1383             if (TargetSym != TargetSectionSymbols->begin()) {
1384               --TargetSym;
1385               uint64_t TargetAddress = std::get<0>(*TargetSym);
1386               StringRef TargetName = std::get<1>(*TargetSym);
1387               outs() << " <" << TargetName;
1388               uint64_t Disp = Target - TargetAddress;
1389               if (Disp)
1390                 outs() << "+0x" << Twine::utohexstr(Disp);
1391               outs() << '>';
1392             }
1393           }
1394         }
1395         outs() << "\n";
1396 
1397         // Hexagon does this in pretty printer
1398         if (Obj->getArch() != Triple::hexagon) {
1399           // Print relocation for instruction.
1400           while (RelCur != RelEnd) {
1401             uint64_t Offset = RelCur->getOffset();
1402             // If this relocation is hidden, skip it.
1403             if (getHidden(*RelCur) || ((SectionAddr + Offset) < StartAddress)) {
1404               ++RelCur;
1405               continue;
1406             }
1407 
1408             // Stop when RelCur's offset is past the current instruction.
1409             if (Offset >= Index + Size)
1410               break;
1411 
1412             // When --adjust-vma is used, update the address printed.
1413             if (RelCur->getSymbol() != Obj->symbol_end()) {
1414               Expected<section_iterator> SymSI =
1415                   RelCur->getSymbol()->getSection();
1416               if (SymSI && *SymSI != Obj->section_end() &&
1417                   (shouldAdjustVA(**SymSI)))
1418                 Offset += AdjustVMA;
1419             }
1420 
1421             printRelocation(*RelCur, SectionAddr + Offset,
1422                             Obj->getBytesInAddress());
1423             ++RelCur;
1424           }
1425         }
1426       }
1427     }
1428   }
1429 }
1430 
1431 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) {
1432   if (StartAddress > StopAddress)
1433     error("Start address should be less than stop address");
1434 
1435   const Target *TheTarget = getTarget(Obj);
1436 
1437   // Package up features to be passed to target/subtarget
1438   SubtargetFeatures Features = Obj->getFeatures();
1439   if (!MAttrs.empty())
1440     for (unsigned I = 0; I != MAttrs.size(); ++I)
1441       Features.AddFeature(MAttrs[I]);
1442 
1443   std::unique_ptr<const MCRegisterInfo> MRI(
1444       TheTarget->createMCRegInfo(TripleName));
1445   if (!MRI)
1446     report_error(Obj->getFileName(),
1447                  "no register info for target " + TripleName);
1448 
1449   // Set up disassembler.
1450   std::unique_ptr<const MCAsmInfo> AsmInfo(
1451       TheTarget->createMCAsmInfo(*MRI, TripleName));
1452   if (!AsmInfo)
1453     report_error(Obj->getFileName(),
1454                  "no assembly info for target " + TripleName);
1455   std::unique_ptr<const MCSubtargetInfo> STI(
1456       TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString()));
1457   if (!STI)
1458     report_error(Obj->getFileName(),
1459                  "no subtarget info for target " + TripleName);
1460   std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo());
1461   if (!MII)
1462     report_error(Obj->getFileName(),
1463                  "no instruction info for target " + TripleName);
1464   MCObjectFileInfo MOFI;
1465   MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI);
1466   // FIXME: for now initialize MCObjectFileInfo with default values
1467   MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx);
1468 
1469   std::unique_ptr<MCDisassembler> DisAsm(
1470       TheTarget->createMCDisassembler(*STI, Ctx));
1471   if (!DisAsm)
1472     report_error(Obj->getFileName(),
1473                  "no disassembler for target " + TripleName);
1474 
1475   std::unique_ptr<const MCInstrAnalysis> MIA(
1476       TheTarget->createMCInstrAnalysis(MII.get()));
1477 
1478   int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1479   std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
1480       Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI));
1481   if (!IP)
1482     report_error(Obj->getFileName(),
1483                  "no instruction printer for target " + TripleName);
1484   IP->setPrintImmHex(PrintImmHex);
1485 
1486   PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName));
1487   SourcePrinter SP(Obj, TheTarget->getName());
1488 
1489   for (StringRef Opt : DisassemblerOptions)
1490     if (!IP->applyTargetSpecificCLOption(Opt))
1491       error("Unrecognized disassembler option: " + Opt);
1492 
1493   disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(),
1494                     STI.get(), PIP, SP, InlineRelocs);
1495 }
1496 
1497 void llvm::printRelocations(const ObjectFile *Obj) {
1498   StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 :
1499                                                  "%08" PRIx64;
1500   // Regular objdump doesn't print relocations in non-relocatable object
1501   // files.
1502   if (!Obj->isRelocatableObject())
1503     return;
1504 
1505   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1506     if (Section.relocation_begin() == Section.relocation_end())
1507       continue;
1508     StringRef SecName;
1509     error(Section.getName(SecName));
1510     outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n";
1511     for (const RelocationRef &Reloc : Section.relocations()) {
1512       uint64_t Address = Reloc.getOffset();
1513       SmallString<32> RelocName;
1514       SmallString<32> ValueStr;
1515       if (Address < StartAddress || Address > StopAddress || getHidden(Reloc))
1516         continue;
1517       Reloc.getTypeName(RelocName);
1518       error(getRelocationValueString(Reloc, ValueStr));
1519       outs() << format(Fmt.data(), Address) << " " << RelocName << " "
1520              << ValueStr << "\n";
1521     }
1522     outs() << "\n";
1523   }
1524 }
1525 
1526 void llvm::printDynamicRelocations(const ObjectFile *Obj) {
1527   // For the moment, this option is for ELF only
1528   if (!Obj->isELF())
1529     return;
1530 
1531   const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
1532   if (!Elf || Elf->getEType() != ELF::ET_DYN) {
1533     error("not a dynamic object");
1534     return;
1535   }
1536 
1537   std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections();
1538   if (DynRelSec.empty())
1539     return;
1540 
1541   outs() << "DYNAMIC RELOCATION RECORDS\n";
1542   StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1543   for (const SectionRef &Section : DynRelSec) {
1544     if (Section.relocation_begin() == Section.relocation_end())
1545       continue;
1546     for (const RelocationRef &Reloc : Section.relocations()) {
1547       uint64_t Address = Reloc.getOffset();
1548       SmallString<32> RelocName;
1549       SmallString<32> ValueStr;
1550       Reloc.getTypeName(RelocName);
1551       error(getRelocationValueString(Reloc, ValueStr));
1552       outs() << format(Fmt.data(), Address) << " " << RelocName << " "
1553              << ValueStr << "\n";
1554     }
1555   }
1556 }
1557 
1558 // Returns true if we need to show LMA column when dumping section headers. We
1559 // show it only when the platform is ELF and either we have at least one section
1560 // whose VMA and LMA are different and/or when --show-lma flag is used.
1561 static bool shouldDisplayLMA(const ObjectFile *Obj) {
1562   if (!Obj->isELF())
1563     return false;
1564   for (const SectionRef &S : ToolSectionFilter(*Obj))
1565     if (S.getAddress() != getELFSectionLMA(S))
1566       return true;
1567   return ShowLMA;
1568 }
1569 
1570 void llvm::printSectionHeaders(const ObjectFile *Obj) {
1571   bool HasLMAColumn = shouldDisplayLMA(Obj);
1572   if (HasLMAColumn)
1573     outs() << "Sections:\n"
1574               "Idx Name          Size     VMA              LMA              "
1575               "Type\n";
1576   else
1577     outs() << "Sections:\n"
1578               "Idx Name          Size     VMA          Type\n";
1579 
1580   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1581     StringRef Name;
1582     error(Section.getName(Name));
1583     uint64_t VMA = Section.getAddress();
1584     if (shouldAdjustVA(Section))
1585       VMA += AdjustVMA;
1586 
1587     uint64_t Size = Section.getSize();
1588     bool Text = Section.isText();
1589     bool Data = Section.isData();
1590     bool BSS = Section.isBSS();
1591     std::string Type = (std::string(Text ? "TEXT " : "") +
1592                         (Data ? "DATA " : "") + (BSS ? "BSS" : ""));
1593 
1594     if (HasLMAColumn)
1595       outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64
1596                        " %s\n",
1597                        (unsigned)Section.getIndex(), Name.str().c_str(), Size,
1598                        VMA, getELFSectionLMA(Section), Type.c_str());
1599     else
1600       outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n",
1601                        (unsigned)Section.getIndex(), Name.str().c_str(), Size,
1602                        VMA, Type.c_str());
1603   }
1604   outs() << "\n";
1605 }
1606 
1607 void llvm::printSectionContents(const ObjectFile *Obj) {
1608   std::error_code EC;
1609   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1610     StringRef Name;
1611     StringRef Contents;
1612     error(Section.getName(Name));
1613     uint64_t BaseAddr = Section.getAddress();
1614     uint64_t Size = Section.getSize();
1615     if (!Size)
1616       continue;
1617 
1618     outs() << "Contents of section " << Name << ":\n";
1619     if (Section.isBSS()) {
1620       outs() << format("<skipping contents of bss section at [%04" PRIx64
1621                        ", %04" PRIx64 ")>\n",
1622                        BaseAddr, BaseAddr + Size);
1623       continue;
1624     }
1625 
1626     error(Section.getContents(Contents));
1627 
1628     // Dump out the content as hex and printable ascii characters.
1629     for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) {
1630       outs() << format(" %04" PRIx64 " ", BaseAddr + Addr);
1631       // Dump line of hex.
1632       for (std::size_t I = 0; I < 16; ++I) {
1633         if (I != 0 && I % 4 == 0)
1634           outs() << ' ';
1635         if (Addr + I < End)
1636           outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true)
1637                  << hexdigit(Contents[Addr + I] & 0xF, true);
1638         else
1639           outs() << "  ";
1640       }
1641       // Print ascii.
1642       outs() << "  ";
1643       for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) {
1644         if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF))
1645           outs() << Contents[Addr + I];
1646         else
1647           outs() << ".";
1648       }
1649       outs() << "\n";
1650     }
1651   }
1652 }
1653 
1654 void llvm::printSymbolTable(const ObjectFile *O, StringRef ArchiveName,
1655                             StringRef ArchitectureName) {
1656   outs() << "SYMBOL TABLE:\n";
1657 
1658   if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) {
1659     printCOFFSymbolTable(Coff);
1660     return;
1661   }
1662 
1663   for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) {
1664     // Skip printing the special zero symbol when dumping an ELF file.
1665     // This makes the output consistent with the GNU objdump.
1666     if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O))
1667       continue;
1668 
1669     const SymbolRef &Symbol = *I;
1670     Expected<uint64_t> AddressOrError = Symbol.getAddress();
1671     if (!AddressOrError)
1672       report_error(ArchiveName, O->getFileName(), AddressOrError.takeError(),
1673                    ArchitectureName);
1674     uint64_t Address = *AddressOrError;
1675     if ((Address < StartAddress) || (Address > StopAddress))
1676       continue;
1677     Expected<SymbolRef::Type> TypeOrError = Symbol.getType();
1678     if (!TypeOrError)
1679       report_error(ArchiveName, O->getFileName(), TypeOrError.takeError(),
1680                    ArchitectureName);
1681     SymbolRef::Type Type = *TypeOrError;
1682     uint32_t Flags = Symbol.getFlags();
1683     Expected<section_iterator> SectionOrErr = Symbol.getSection();
1684     if (!SectionOrErr)
1685       report_error(ArchiveName, O->getFileName(), SectionOrErr.takeError(),
1686                    ArchitectureName);
1687     section_iterator Section = *SectionOrErr;
1688     StringRef Name;
1689     if (Type == SymbolRef::ST_Debug && Section != O->section_end()) {
1690       Section->getName(Name);
1691     } else {
1692       Expected<StringRef> NameOrErr = Symbol.getName();
1693       if (!NameOrErr)
1694         report_error(ArchiveName, O->getFileName(), NameOrErr.takeError(),
1695                      ArchitectureName);
1696       Name = *NameOrErr;
1697     }
1698 
1699     bool Global = Flags & SymbolRef::SF_Global;
1700     bool Weak = Flags & SymbolRef::SF_Weak;
1701     bool Absolute = Flags & SymbolRef::SF_Absolute;
1702     bool Common = Flags & SymbolRef::SF_Common;
1703     bool Hidden = Flags & SymbolRef::SF_Hidden;
1704 
1705     char GlobLoc = ' ';
1706     if (Type != SymbolRef::ST_Unknown)
1707       GlobLoc = Global ? 'g' : 'l';
1708     char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
1709                  ? 'd' : ' ';
1710     char FileFunc = ' ';
1711     if (Type == SymbolRef::ST_File)
1712       FileFunc = 'f';
1713     else if (Type == SymbolRef::ST_Function)
1714       FileFunc = 'F';
1715     else if (Type == SymbolRef::ST_Data)
1716       FileFunc = 'O';
1717 
1718     const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 :
1719                                                    "%08" PRIx64;
1720 
1721     outs() << format(Fmt, Address) << " "
1722            << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' '
1723            << (Weak ? 'w' : ' ') // Weak?
1724            << ' ' // Constructor. Not supported yet.
1725            << ' ' // Warning. Not supported yet.
1726            << ' ' // Indirect reference to another symbol.
1727            << Debug // Debugging (d) or dynamic (D) symbol.
1728            << FileFunc // Name of function (F), file (f) or object (O).
1729            << ' ';
1730     if (Absolute) {
1731       outs() << "*ABS*";
1732     } else if (Common) {
1733       outs() << "*COM*";
1734     } else if (Section == O->section_end()) {
1735       outs() << "*UND*";
1736     } else {
1737       if (const MachOObjectFile *MachO =
1738           dyn_cast<const MachOObjectFile>(O)) {
1739         DataRefImpl DR = Section->getRawDataRefImpl();
1740         StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
1741         outs() << SegmentName << ",";
1742       }
1743       StringRef SectionName;
1744       error(Section->getName(SectionName));
1745       outs() << SectionName;
1746     }
1747 
1748     outs() << '\t';
1749     if (Common || isa<ELFObjectFileBase>(O)) {
1750       uint64_t Val =
1751           Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize();
1752       outs() << format("\t %08" PRIx64 " ", Val);
1753     }
1754 
1755     if (Hidden)
1756       outs() << ".hidden ";
1757 
1758     if (Demangle)
1759       outs() << demangle(Name) << '\n';
1760     else
1761       outs() << Name << '\n';
1762   }
1763 }
1764 
1765 static void printUnwindInfo(const ObjectFile *O) {
1766   outs() << "Unwind info:\n\n";
1767 
1768   if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O))
1769     printCOFFUnwindInfo(Coff);
1770   else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O))
1771     printMachOUnwindInfo(MachO);
1772   else
1773     // TODO: Extract DWARF dump tool to objdump.
1774     WithColor::error(errs(), ToolName)
1775         << "This operation is only currently supported "
1776            "for COFF and MachO object files.\n";
1777 }
1778 
1779 void llvm::printExportsTrie(const ObjectFile *o) {
1780   outs() << "Exports trie:\n";
1781   if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1782     printMachOExportsTrie(MachO);
1783   else
1784     WithColor::error(errs(), ToolName)
1785         << "This operation is only currently supported "
1786            "for Mach-O executable files.\n";
1787 }
1788 
1789 void llvm::printRebaseTable(ObjectFile *o) {
1790   outs() << "Rebase table:\n";
1791   if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1792     printMachORebaseTable(MachO);
1793   else
1794     WithColor::error(errs(), ToolName)
1795         << "This operation is only currently supported "
1796            "for Mach-O executable files.\n";
1797 }
1798 
1799 void llvm::printBindTable(ObjectFile *o) {
1800   outs() << "Bind table:\n";
1801   if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1802     printMachOBindTable(MachO);
1803   else
1804     WithColor::error(errs(), ToolName)
1805         << "This operation is only currently supported "
1806            "for Mach-O executable files.\n";
1807 }
1808 
1809 void llvm::printLazyBindTable(ObjectFile *o) {
1810   outs() << "Lazy bind table:\n";
1811   if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1812     printMachOLazyBindTable(MachO);
1813   else
1814     WithColor::error(errs(), ToolName)
1815         << "This operation is only currently supported "
1816            "for Mach-O executable files.\n";
1817 }
1818 
1819 void llvm::printWeakBindTable(ObjectFile *o) {
1820   outs() << "Weak bind table:\n";
1821   if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o))
1822     printMachOWeakBindTable(MachO);
1823   else
1824     WithColor::error(errs(), ToolName)
1825         << "This operation is only currently supported "
1826            "for Mach-O executable files.\n";
1827 }
1828 
1829 /// Dump the raw contents of the __clangast section so the output can be piped
1830 /// into llvm-bcanalyzer.
1831 void llvm::printRawClangAST(const ObjectFile *Obj) {
1832   if (outs().is_displayed()) {
1833     WithColor::error(errs(), ToolName)
1834         << "The -raw-clang-ast option will dump the raw binary contents of "
1835            "the clang ast section.\n"
1836            "Please redirect the output to a file or another program such as "
1837            "llvm-bcanalyzer.\n";
1838     return;
1839   }
1840 
1841   StringRef ClangASTSectionName("__clangast");
1842   if (isa<COFFObjectFile>(Obj)) {
1843     ClangASTSectionName = "clangast";
1844   }
1845 
1846   Optional<object::SectionRef> ClangASTSection;
1847   for (auto Sec : ToolSectionFilter(*Obj)) {
1848     StringRef Name;
1849     Sec.getName(Name);
1850     if (Name == ClangASTSectionName) {
1851       ClangASTSection = Sec;
1852       break;
1853     }
1854   }
1855   if (!ClangASTSection)
1856     return;
1857 
1858   StringRef ClangASTContents;
1859   error(ClangASTSection.getValue().getContents(ClangASTContents));
1860   outs().write(ClangASTContents.data(), ClangASTContents.size());
1861 }
1862 
1863 static void printFaultMaps(const ObjectFile *Obj) {
1864   StringRef FaultMapSectionName;
1865 
1866   if (isa<ELFObjectFileBase>(Obj)) {
1867     FaultMapSectionName = ".llvm_faultmaps";
1868   } else if (isa<MachOObjectFile>(Obj)) {
1869     FaultMapSectionName = "__llvm_faultmaps";
1870   } else {
1871     WithColor::error(errs(), ToolName)
1872         << "This operation is only currently supported "
1873            "for ELF and Mach-O executable files.\n";
1874     return;
1875   }
1876 
1877   Optional<object::SectionRef> FaultMapSection;
1878 
1879   for (auto Sec : ToolSectionFilter(*Obj)) {
1880     StringRef Name;
1881     Sec.getName(Name);
1882     if (Name == FaultMapSectionName) {
1883       FaultMapSection = Sec;
1884       break;
1885     }
1886   }
1887 
1888   outs() << "FaultMap table:\n";
1889 
1890   if (!FaultMapSection.hasValue()) {
1891     outs() << "<not found>\n";
1892     return;
1893   }
1894 
1895   StringRef FaultMapContents;
1896   error(FaultMapSection.getValue().getContents(FaultMapContents));
1897 
1898   FaultMapParser FMP(FaultMapContents.bytes_begin(),
1899                      FaultMapContents.bytes_end());
1900 
1901   outs() << FMP;
1902 }
1903 
1904 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) {
1905   if (O->isELF()) {
1906     printELFFileHeader(O);
1907     printELFDynamicSection(O);
1908     printELFSymbolVersionInfo(O);
1909     return;
1910   }
1911   if (O->isCOFF())
1912     return printCOFFFileHeader(O);
1913   if (O->isWasm())
1914     return printWasmFileHeader(O);
1915   if (O->isMachO()) {
1916     printMachOFileHeader(O);
1917     if (!OnlyFirst)
1918       printMachOLoadCommands(O);
1919     return;
1920   }
1921   report_error(O->getFileName(), "Invalid/Unsupported object file format");
1922 }
1923 
1924 static void printFileHeaders(const ObjectFile *O) {
1925   if (!O->isELF() && !O->isCOFF())
1926     report_error(O->getFileName(), "Invalid/Unsupported object file format");
1927 
1928   Triple::ArchType AT = O->getArch();
1929   outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n";
1930   Expected<uint64_t> StartAddrOrErr = O->getStartAddress();
1931   if (!StartAddrOrErr)
1932     report_error(O->getFileName(), StartAddrOrErr.takeError());
1933 
1934   StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1935   uint64_t Address = StartAddrOrErr.get();
1936   outs() << "start address: "
1937          << "0x" << format(Fmt.data(), Address) << "\n\n";
1938 }
1939 
1940 static void printArchiveChild(StringRef Filename, const Archive::Child &C) {
1941   Expected<sys::fs::perms> ModeOrErr = C.getAccessMode();
1942   if (!ModeOrErr) {
1943     WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n";
1944     consumeError(ModeOrErr.takeError());
1945     return;
1946   }
1947   sys::fs::perms Mode = ModeOrErr.get();
1948   outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
1949   outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
1950   outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
1951   outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
1952   outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
1953   outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
1954   outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
1955   outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
1956   outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
1957 
1958   outs() << " ";
1959 
1960   Expected<unsigned> UIDOrErr = C.getUID();
1961   if (!UIDOrErr)
1962     report_error(Filename, UIDOrErr.takeError());
1963   unsigned UID = UIDOrErr.get();
1964   outs() << format("%d/", UID);
1965 
1966   Expected<unsigned> GIDOrErr = C.getGID();
1967   if (!GIDOrErr)
1968     report_error(Filename, GIDOrErr.takeError());
1969   unsigned GID = GIDOrErr.get();
1970   outs() << format("%-d ", GID);
1971 
1972   Expected<uint64_t> Size = C.getRawSize();
1973   if (!Size)
1974     report_error(Filename, Size.takeError());
1975   outs() << format("%6" PRId64, Size.get()) << " ";
1976 
1977   StringRef RawLastModified = C.getRawLastModified();
1978   unsigned Seconds;
1979   if (RawLastModified.getAsInteger(10, Seconds))
1980     outs() << "(date: \"" << RawLastModified
1981            << "\" contains non-decimal chars) ";
1982   else {
1983     // Since ctime(3) returns a 26 character string of the form:
1984     // "Sun Sep 16 01:03:52 1973\n\0"
1985     // just print 24 characters.
1986     time_t t = Seconds;
1987     outs() << format("%.24s ", ctime(&t));
1988   }
1989 
1990   StringRef Name = "";
1991   Expected<StringRef> NameOrErr = C.getName();
1992   if (!NameOrErr) {
1993     consumeError(NameOrErr.takeError());
1994     Expected<StringRef> RawNameOrErr = C.getRawName();
1995     if (!RawNameOrErr)
1996       report_error(Filename, NameOrErr.takeError());
1997     Name = RawNameOrErr.get();
1998   } else {
1999     Name = NameOrErr.get();
2000   }
2001   outs() << Name << "\n";
2002 }
2003 
2004 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
2005                        const Archive::Child *C = nullptr) {
2006   // Avoid other output when using a raw option.
2007   if (!RawClangAST) {
2008     outs() << '\n';
2009     if (A)
2010       outs() << A->getFileName() << "(" << O->getFileName() << ")";
2011     else
2012       outs() << O->getFileName();
2013     outs() << ":\tfile format " << O->getFileFormatName() << "\n\n";
2014   }
2015 
2016   StringRef ArchiveName = A ? A->getFileName() : "";
2017   if (FileHeaders)
2018     printFileHeaders(O);
2019   if (ArchiveHeaders && !MachOOpt && C)
2020     printArchiveChild(ArchiveName, *C);
2021   if (Disassemble)
2022     disassembleObject(O, Relocations);
2023   if (Relocations && !Disassemble)
2024     printRelocations(O);
2025   if (DynamicRelocations)
2026     printDynamicRelocations(O);
2027   if (SectionHeaders)
2028     printSectionHeaders(O);
2029   if (SectionContents)
2030     printSectionContents(O);
2031   if (SymbolTable)
2032     printSymbolTable(O, ArchiveName);
2033   if (UnwindInfo)
2034     printUnwindInfo(O);
2035   if (PrivateHeaders || FirstPrivateHeader)
2036     printPrivateFileHeaders(O, FirstPrivateHeader);
2037   if (ExportsTrie)
2038     printExportsTrie(O);
2039   if (Rebase)
2040     printRebaseTable(O);
2041   if (Bind)
2042     printBindTable(O);
2043   if (LazyBind)
2044     printLazyBindTable(O);
2045   if (WeakBind)
2046     printWeakBindTable(O);
2047   if (RawClangAST)
2048     printRawClangAST(O);
2049   if (PrintFaultMaps)
2050     printFaultMaps(O);
2051   if (DwarfDumpType != DIDT_Null) {
2052     std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O);
2053     // Dump the complete DWARF structure.
2054     DIDumpOptions DumpOpts;
2055     DumpOpts.DumpType = DwarfDumpType;
2056     DICtx->dump(outs(), DumpOpts);
2057   }
2058 }
2059 
2060 static void dumpObject(const COFFImportFile *I, const Archive *A,
2061                        const Archive::Child *C = nullptr) {
2062   StringRef ArchiveName = A ? A->getFileName() : "";
2063 
2064   // Avoid other output when using a raw option.
2065   if (!RawClangAST)
2066     outs() << '\n'
2067            << ArchiveName << "(" << I->getFileName() << ")"
2068            << ":\tfile format COFF-import-file"
2069            << "\n\n";
2070 
2071   if (ArchiveHeaders && !MachOOpt && C)
2072     printArchiveChild(ArchiveName, *C);
2073   if (SymbolTable)
2074     printCOFFSymbolTable(I);
2075 }
2076 
2077 /// Dump each object file in \a a;
2078 static void dumpArchive(const Archive *A) {
2079   Error Err = Error::success();
2080   for (auto &C : A->children(Err)) {
2081     Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
2082     if (!ChildOrErr) {
2083       if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError()))
2084         report_error(A->getFileName(), C, std::move(E));
2085       continue;
2086     }
2087     if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
2088       dumpObject(O, A, &C);
2089     else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get()))
2090       dumpObject(I, A, &C);
2091     else
2092       report_error(A->getFileName(), object_error::invalid_file_type);
2093   }
2094   if (Err)
2095     report_error(A->getFileName(), std::move(Err));
2096 }
2097 
2098 /// Open file and figure out how to dump it.
2099 static void dumpInput(StringRef file) {
2100   // If we are using the Mach-O specific object file parser, then let it parse
2101   // the file and process the command line options.  So the -arch flags can
2102   // be used to select specific slices, etc.
2103   if (MachOOpt) {
2104     parseInputMachO(file);
2105     return;
2106   }
2107 
2108   // Attempt to open the binary.
2109   Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file);
2110   if (!BinaryOrErr)
2111     report_error(file, BinaryOrErr.takeError());
2112   Binary &Binary = *BinaryOrErr.get().getBinary();
2113 
2114   if (Archive *A = dyn_cast<Archive>(&Binary))
2115     dumpArchive(A);
2116   else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary))
2117     dumpObject(O);
2118   else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary))
2119     parseInputMachO(UB);
2120   else
2121     report_error(file, object_error::invalid_file_type);
2122 }
2123 
2124 int main(int argc, char **argv) {
2125   InitLLVM X(argc, argv);
2126 
2127   // Initialize targets and assembly printers/parsers.
2128   llvm::InitializeAllTargetInfos();
2129   llvm::InitializeAllTargetMCs();
2130   llvm::InitializeAllDisassemblers();
2131 
2132   // Register the target printer for --version.
2133   cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion);
2134 
2135   cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n");
2136 
2137   ToolName = argv[0];
2138 
2139   // Defaults to a.out if no filenames specified.
2140   if (InputFilenames.empty())
2141     InputFilenames.push_back("a.out");
2142 
2143   if (AllHeaders)
2144     ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations =
2145         SectionHeaders = SymbolTable = true;
2146 
2147   if (DisassembleAll || PrintSource || PrintLines)
2148     Disassemble = true;
2149 
2150   if (!Disassemble
2151       && !Relocations
2152       && !DynamicRelocations
2153       && !SectionHeaders
2154       && !SectionContents
2155       && !SymbolTable
2156       && !UnwindInfo
2157       && !PrivateHeaders
2158       && !FileHeaders
2159       && !FirstPrivateHeader
2160       && !ExportsTrie
2161       && !Rebase
2162       && !Bind
2163       && !LazyBind
2164       && !WeakBind
2165       && !RawClangAST
2166       && !(UniversalHeaders && MachOOpt)
2167       && !ArchiveHeaders
2168       && !(IndirectSymbols && MachOOpt)
2169       && !(DataInCode && MachOOpt)
2170       && !(LinkOptHints && MachOOpt)
2171       && !(InfoPlist && MachOOpt)
2172       && !(DylibsUsed && MachOOpt)
2173       && !(DylibId && MachOOpt)
2174       && !(ObjcMetaData && MachOOpt)
2175       && !(!FilterSections.empty() && MachOOpt)
2176       && !PrintFaultMaps
2177       && DwarfDumpType == DIDT_Null) {
2178     cl::PrintHelpMessage();
2179     return 2;
2180   }
2181 
2182   DisasmFuncsSet.insert(DisassembleFunctions.begin(),
2183                         DisassembleFunctions.end());
2184 
2185   llvm::for_each(InputFilenames, dumpInput);
2186 
2187   return EXIT_SUCCESS;
2188 }
2189