xref: /freebsd/contrib/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp (revision 0fca6ea1d4eea4c934cfff25ac9ee8ad6fe95583)
1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This program is a utility that works like binutils "objdump", that is, it
10 // dumps out a plethora of information about an object file depending on the
11 // flags.
12 //
13 // The flags and output of this program should be near identical to those of
14 // binutils objdump.
15 //
16 //===----------------------------------------------------------------------===//
17 
18 #include "llvm-objdump.h"
19 #include "COFFDump.h"
20 #include "ELFDump.h"
21 #include "MachODump.h"
22 #include "ObjdumpOptID.h"
23 #include "OffloadDump.h"
24 #include "SourcePrinter.h"
25 #include "WasmDump.h"
26 #include "XCOFFDump.h"
27 #include "llvm/ADT/STLExtras.h"
28 #include "llvm/ADT/SetOperations.h"
29 #include "llvm/ADT/StringExtras.h"
30 #include "llvm/ADT/StringSet.h"
31 #include "llvm/ADT/Twine.h"
32 #include "llvm/BinaryFormat/Wasm.h"
33 #include "llvm/DebugInfo/BTF/BTFParser.h"
34 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
35 #include "llvm/DebugInfo/Symbolize/SymbolizableModule.h"
36 #include "llvm/DebugInfo/Symbolize/Symbolize.h"
37 #include "llvm/Debuginfod/BuildIDFetcher.h"
38 #include "llvm/Debuginfod/Debuginfod.h"
39 #include "llvm/Debuginfod/HTTPClient.h"
40 #include "llvm/Demangle/Demangle.h"
41 #include "llvm/MC/MCAsmInfo.h"
42 #include "llvm/MC/MCContext.h"
43 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
44 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h"
45 #include "llvm/MC/MCInst.h"
46 #include "llvm/MC/MCInstPrinter.h"
47 #include "llvm/MC/MCInstrAnalysis.h"
48 #include "llvm/MC/MCInstrInfo.h"
49 #include "llvm/MC/MCObjectFileInfo.h"
50 #include "llvm/MC/MCRegisterInfo.h"
51 #include "llvm/MC/MCTargetOptions.h"
52 #include "llvm/MC/TargetRegistry.h"
53 #include "llvm/Object/Archive.h"
54 #include "llvm/Object/BuildID.h"
55 #include "llvm/Object/COFF.h"
56 #include "llvm/Object/COFFImportFile.h"
57 #include "llvm/Object/ELFObjectFile.h"
58 #include "llvm/Object/ELFTypes.h"
59 #include "llvm/Object/FaultMapParser.h"
60 #include "llvm/Object/MachO.h"
61 #include "llvm/Object/MachOUniversal.h"
62 #include "llvm/Object/ObjectFile.h"
63 #include "llvm/Object/OffloadBinary.h"
64 #include "llvm/Object/Wasm.h"
65 #include "llvm/Option/Arg.h"
66 #include "llvm/Option/ArgList.h"
67 #include "llvm/Option/Option.h"
68 #include "llvm/Support/Casting.h"
69 #include "llvm/Support/Debug.h"
70 #include "llvm/Support/Errc.h"
71 #include "llvm/Support/FileSystem.h"
72 #include "llvm/Support/Format.h"
73 #include "llvm/Support/FormatVariadic.h"
74 #include "llvm/Support/GraphWriter.h"
75 #include "llvm/Support/LLVMDriver.h"
76 #include "llvm/Support/MemoryBuffer.h"
77 #include "llvm/Support/SourceMgr.h"
78 #include "llvm/Support/StringSaver.h"
79 #include "llvm/Support/TargetSelect.h"
80 #include "llvm/Support/WithColor.h"
81 #include "llvm/Support/raw_ostream.h"
82 #include "llvm/TargetParser/Host.h"
83 #include "llvm/TargetParser/Triple.h"
84 #include <algorithm>
85 #include <cctype>
86 #include <cstring>
87 #include <optional>
88 #include <set>
89 #include <system_error>
90 #include <unordered_map>
91 #include <utility>
92 
93 using namespace llvm;
94 using namespace llvm::object;
95 using namespace llvm::objdump;
96 using namespace llvm::opt;
97 
98 namespace {
99 
100 class CommonOptTable : public opt::GenericOptTable {
101 public:
CommonOptTable(ArrayRef<Info> OptionInfos,const char * Usage,const char * Description)102   CommonOptTable(ArrayRef<Info> OptionInfos, const char *Usage,
103                  const char *Description)
104       : opt::GenericOptTable(OptionInfos), Usage(Usage),
105         Description(Description) {
106     setGroupedShortOptions(true);
107   }
108 
printHelp(StringRef Argv0,bool ShowHidden=false) const109   void printHelp(StringRef Argv0, bool ShowHidden = false) const {
110     Argv0 = sys::path::filename(Argv0);
111     opt::GenericOptTable::printHelp(outs(), (Argv0 + Usage).str().c_str(),
112                                     Description, ShowHidden, ShowHidden);
113     // TODO Replace this with OptTable API once it adds extrahelp support.
114     outs() << "\nPass @FILE as argument to read options from FILE.\n";
115   }
116 
117 private:
118   const char *Usage;
119   const char *Description;
120 };
121 
122 // ObjdumpOptID is in ObjdumpOptID.h
123 namespace objdump_opt {
124 #define PREFIX(NAME, VALUE)                                                    \
125   static constexpr StringLiteral NAME##_init[] = VALUE;                        \
126   static constexpr ArrayRef<StringLiteral> NAME(NAME##_init,                   \
127                                                 std::size(NAME##_init) - 1);
128 #include "ObjdumpOpts.inc"
129 #undef PREFIX
130 
131 static constexpr opt::OptTable::Info ObjdumpInfoTable[] = {
132 #define OPTION(...)                                                            \
133   LLVM_CONSTRUCT_OPT_INFO_WITH_ID_PREFIX(OBJDUMP_, __VA_ARGS__),
134 #include "ObjdumpOpts.inc"
135 #undef OPTION
136 };
137 } // namespace objdump_opt
138 
139 class ObjdumpOptTable : public CommonOptTable {
140 public:
ObjdumpOptTable()141   ObjdumpOptTable()
142       : CommonOptTable(objdump_opt::ObjdumpInfoTable,
143                        " [options] <input object files>",
144                        "llvm object file dumper") {}
145 };
146 
147 enum OtoolOptID {
148   OTOOL_INVALID = 0, // This is not an option ID.
149 #define OPTION(...) LLVM_MAKE_OPT_ID_WITH_ID_PREFIX(OTOOL_, __VA_ARGS__),
150 #include "OtoolOpts.inc"
151 #undef OPTION
152 };
153 
154 namespace otool {
155 #define PREFIX(NAME, VALUE)                                                    \
156   static constexpr StringLiteral NAME##_init[] = VALUE;                        \
157   static constexpr ArrayRef<StringLiteral> NAME(NAME##_init,                   \
158                                                 std::size(NAME##_init) - 1);
159 #include "OtoolOpts.inc"
160 #undef PREFIX
161 
162 static constexpr opt::OptTable::Info OtoolInfoTable[] = {
163 #define OPTION(...) LLVM_CONSTRUCT_OPT_INFO_WITH_ID_PREFIX(OTOOL_, __VA_ARGS__),
164 #include "OtoolOpts.inc"
165 #undef OPTION
166 };
167 } // namespace otool
168 
169 class OtoolOptTable : public CommonOptTable {
170 public:
OtoolOptTable()171   OtoolOptTable()
172       : CommonOptTable(otool::OtoolInfoTable, " [option...] [file...]",
173                        "Mach-O object file displaying tool") {}
174 };
175 
176 struct BBAddrMapLabel {
177   std::string BlockLabel;
178   std::string PGOAnalysis;
179 };
180 
181 // This class represents the BBAddrMap and PGOMap associated with a single
182 // function.
183 class BBAddrMapFunctionEntry {
184 public:
BBAddrMapFunctionEntry(BBAddrMap AddrMap,PGOAnalysisMap PGOMap)185   BBAddrMapFunctionEntry(BBAddrMap AddrMap, PGOAnalysisMap PGOMap)
186       : AddrMap(std::move(AddrMap)), PGOMap(std::move(PGOMap)) {}
187 
getAddrMap() const188   const BBAddrMap &getAddrMap() const { return AddrMap; }
189 
190   // Returns the PGO string associated with the entry of index `PGOBBEntryIndex`
191   // in `PGOMap`. If PrettyPGOAnalysis is true, prints BFI as relative frequency
192   // and BPI as percentage. Otherwise raw values are displayed.
constructPGOLabelString(size_t PGOBBEntryIndex,bool PrettyPGOAnalysis) const193   std::string constructPGOLabelString(size_t PGOBBEntryIndex,
194                                       bool PrettyPGOAnalysis) const {
195     if (!PGOMap.FeatEnable.hasPGOAnalysis())
196       return "";
197     std::string PGOString;
198     raw_string_ostream PGOSS(PGOString);
199 
200     PGOSS << " (";
201     if (PGOMap.FeatEnable.FuncEntryCount && PGOBBEntryIndex == 0) {
202       PGOSS << "Entry count: " << Twine(PGOMap.FuncEntryCount);
203       if (PGOMap.FeatEnable.hasPGOAnalysisBBData()) {
204         PGOSS << ", ";
205       }
206     }
207 
208     if (PGOMap.FeatEnable.hasPGOAnalysisBBData()) {
209 
210       assert(PGOBBEntryIndex < PGOMap.BBEntries.size() &&
211              "Expected PGOAnalysisMap and BBAddrMap to have the same entries");
212       const PGOAnalysisMap::PGOBBEntry &PGOBBEntry =
213           PGOMap.BBEntries[PGOBBEntryIndex];
214 
215       if (PGOMap.FeatEnable.BBFreq) {
216         PGOSS << "Frequency: ";
217         if (PrettyPGOAnalysis)
218           printRelativeBlockFreq(PGOSS, PGOMap.BBEntries.front().BlockFreq,
219                                  PGOBBEntry.BlockFreq);
220         else
221           PGOSS << Twine(PGOBBEntry.BlockFreq.getFrequency());
222         if (PGOMap.FeatEnable.BrProb && PGOBBEntry.Successors.size() > 0) {
223           PGOSS << ", ";
224         }
225       }
226       if (PGOMap.FeatEnable.BrProb && PGOBBEntry.Successors.size() > 0) {
227         PGOSS << "Successors: ";
228         interleaveComma(
229             PGOBBEntry.Successors, PGOSS,
230             [&](const PGOAnalysisMap::PGOBBEntry::SuccessorEntry &SE) {
231               PGOSS << "BB" << SE.ID << ":";
232               if (PrettyPGOAnalysis)
233                 PGOSS << "[" << SE.Prob << "]";
234               else
235                 PGOSS.write_hex(SE.Prob.getNumerator());
236             });
237       }
238     }
239     PGOSS << ")";
240 
241     return PGOString;
242   }
243 
244 private:
245   const BBAddrMap AddrMap;
246   const PGOAnalysisMap PGOMap;
247 };
248 
249 // This class represents the BBAddrMap and PGOMap of potentially multiple
250 // functions in a section.
251 class BBAddrMapInfo {
252 public:
clear()253   void clear() {
254     FunctionAddrToMap.clear();
255     RangeBaseAddrToFunctionAddr.clear();
256   }
257 
empty() const258   bool empty() const { return FunctionAddrToMap.empty(); }
259 
AddFunctionEntry(BBAddrMap AddrMap,PGOAnalysisMap PGOMap)260   void AddFunctionEntry(BBAddrMap AddrMap, PGOAnalysisMap PGOMap) {
261     uint64_t FunctionAddr = AddrMap.getFunctionAddress();
262     for (size_t I = 1; I < AddrMap.BBRanges.size(); ++I)
263       RangeBaseAddrToFunctionAddr.emplace(AddrMap.BBRanges[I].BaseAddress,
264                                           FunctionAddr);
265     [[maybe_unused]] auto R = FunctionAddrToMap.try_emplace(
266         FunctionAddr, std::move(AddrMap), std::move(PGOMap));
267     assert(R.second && "duplicate function address");
268   }
269 
270   // Returns the BBAddrMap entry for the function associated with `BaseAddress`.
271   // `BaseAddress` could be the function address or the address of a range
272   // associated with that function. Returns `nullptr` if `BaseAddress` is not
273   // mapped to any entry.
getEntryForAddress(uint64_t BaseAddress) const274   const BBAddrMapFunctionEntry *getEntryForAddress(uint64_t BaseAddress) const {
275     uint64_t FunctionAddr = BaseAddress;
276     auto S = RangeBaseAddrToFunctionAddr.find(BaseAddress);
277     if (S != RangeBaseAddrToFunctionAddr.end())
278       FunctionAddr = S->second;
279     auto R = FunctionAddrToMap.find(FunctionAddr);
280     if (R == FunctionAddrToMap.end())
281       return nullptr;
282     return &R->second;
283   }
284 
285 private:
286   std::unordered_map<uint64_t, BBAddrMapFunctionEntry> FunctionAddrToMap;
287   std::unordered_map<uint64_t, uint64_t> RangeBaseAddrToFunctionAddr;
288 };
289 
290 } // namespace
291 
292 #define DEBUG_TYPE "objdump"
293 
294 enum class ColorOutput {
295   Auto,
296   Enable,
297   Disable,
298   Invalid,
299 };
300 
301 static uint64_t AdjustVMA;
302 static bool AllHeaders;
303 static std::string ArchName;
304 bool objdump::ArchiveHeaders;
305 bool objdump::Demangle;
306 bool objdump::Disassemble;
307 bool objdump::DisassembleAll;
308 bool objdump::SymbolDescription;
309 bool objdump::TracebackTable;
310 static std::vector<std::string> DisassembleSymbols;
311 static bool DisassembleZeroes;
312 static std::vector<std::string> DisassemblerOptions;
313 static ColorOutput DisassemblyColor;
314 DIDumpType objdump::DwarfDumpType;
315 static bool DynamicRelocations;
316 static bool FaultMapSection;
317 static bool FileHeaders;
318 bool objdump::SectionContents;
319 static std::vector<std::string> InputFilenames;
320 bool objdump::PrintLines;
321 static bool MachOOpt;
322 std::string objdump::MCPU;
323 std::vector<std::string> objdump::MAttrs;
324 bool objdump::ShowRawInsn;
325 bool objdump::LeadingAddr;
326 static bool Offloading;
327 static bool RawClangAST;
328 bool objdump::Relocations;
329 bool objdump::PrintImmHex;
330 bool objdump::PrivateHeaders;
331 std::vector<std::string> objdump::FilterSections;
332 bool objdump::SectionHeaders;
333 static bool ShowAllSymbols;
334 static bool ShowLMA;
335 bool objdump::PrintSource;
336 
337 static uint64_t StartAddress;
338 static bool HasStartAddressFlag;
339 static uint64_t StopAddress = UINT64_MAX;
340 static bool HasStopAddressFlag;
341 
342 bool objdump::SymbolTable;
343 static bool SymbolizeOperands;
344 static bool PrettyPGOAnalysisMap;
345 static bool DynamicSymbolTable;
346 std::string objdump::TripleName;
347 bool objdump::UnwindInfo;
348 static bool Wide;
349 std::string objdump::Prefix;
350 uint32_t objdump::PrefixStrip;
351 
352 DebugVarsFormat objdump::DbgVariables = DVDisabled;
353 
354 int objdump::DbgIndent = 52;
355 
356 static StringSet<> DisasmSymbolSet;
357 StringSet<> objdump::FoundSectionSet;
358 static StringRef ToolName;
359 
360 std::unique_ptr<BuildIDFetcher> BIDFetcher;
361 
Dumper(const object::ObjectFile & O)362 Dumper::Dumper(const object::ObjectFile &O) : O(O) {
363   WarningHandler = [this](const Twine &Msg) {
364     if (Warnings.insert(Msg.str()).second)
365       reportWarning(Msg, this->O.getFileName());
366     return Error::success();
367   };
368 }
369 
reportUniqueWarning(Error Err)370 void Dumper::reportUniqueWarning(Error Err) {
371   reportUniqueWarning(toString(std::move(Err)));
372 }
373 
reportUniqueWarning(const Twine & Msg)374 void Dumper::reportUniqueWarning(const Twine &Msg) {
375   cantFail(WarningHandler(Msg));
376 }
377 
createDumper(const ObjectFile & Obj)378 static Expected<std::unique_ptr<Dumper>> createDumper(const ObjectFile &Obj) {
379   if (const auto *O = dyn_cast<COFFObjectFile>(&Obj))
380     return createCOFFDumper(*O);
381   if (const auto *O = dyn_cast<ELFObjectFileBase>(&Obj))
382     return createELFDumper(*O);
383   if (const auto *O = dyn_cast<MachOObjectFile>(&Obj))
384     return createMachODumper(*O);
385   if (const auto *O = dyn_cast<WasmObjectFile>(&Obj))
386     return createWasmDumper(*O);
387   if (const auto *O = dyn_cast<XCOFFObjectFile>(&Obj))
388     return createXCOFFDumper(*O);
389 
390   return createStringError(errc::invalid_argument,
391                            "unsupported object file format");
392 }
393 
394 namespace {
395 struct FilterResult {
396   // True if the section should not be skipped.
397   bool Keep;
398 
399   // True if the index counter should be incremented, even if the section should
400   // be skipped. For example, sections may be skipped if they are not included
401   // in the --section flag, but we still want those to count toward the section
402   // count.
403   bool IncrementIndex;
404 };
405 } // namespace
406 
checkSectionFilter(object::SectionRef S)407 static FilterResult checkSectionFilter(object::SectionRef S) {
408   if (FilterSections.empty())
409     return {/*Keep=*/true, /*IncrementIndex=*/true};
410 
411   Expected<StringRef> SecNameOrErr = S.getName();
412   if (!SecNameOrErr) {
413     consumeError(SecNameOrErr.takeError());
414     return {/*Keep=*/false, /*IncrementIndex=*/false};
415   }
416   StringRef SecName = *SecNameOrErr;
417 
418   // StringSet does not allow empty key so avoid adding sections with
419   // no name (such as the section with index 0) here.
420   if (!SecName.empty())
421     FoundSectionSet.insert(SecName);
422 
423   // Only show the section if it's in the FilterSections list, but always
424   // increment so the indexing is stable.
425   return {/*Keep=*/is_contained(FilterSections, SecName),
426           /*IncrementIndex=*/true};
427 }
428 
ToolSectionFilter(object::ObjectFile const & O,uint64_t * Idx)429 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O,
430                                          uint64_t *Idx) {
431   // Start at UINT64_MAX so that the first index returned after an increment is
432   // zero (after the unsigned wrap).
433   if (Idx)
434     *Idx = UINT64_MAX;
435   return SectionFilter(
436       [Idx](object::SectionRef S) {
437         FilterResult Result = checkSectionFilter(S);
438         if (Idx != nullptr && Result.IncrementIndex)
439           *Idx += 1;
440         return Result.Keep;
441       },
442       O);
443 }
444 
getFileNameForError(const object::Archive::Child & C,unsigned Index)445 std::string objdump::getFileNameForError(const object::Archive::Child &C,
446                                          unsigned Index) {
447   Expected<StringRef> NameOrErr = C.getName();
448   if (NameOrErr)
449     return std::string(NameOrErr.get());
450   // If we have an error getting the name then we print the index of the archive
451   // member. Since we are already in an error state, we just ignore this error.
452   consumeError(NameOrErr.takeError());
453   return "<file index: " + std::to_string(Index) + ">";
454 }
455 
reportWarning(const Twine & Message,StringRef File)456 void objdump::reportWarning(const Twine &Message, StringRef File) {
457   // Output order between errs() and outs() matters especially for archive
458   // files where the output is per member object.
459   outs().flush();
460   WithColor::warning(errs(), ToolName)
461       << "'" << File << "': " << Message << "\n";
462 }
463 
reportError(StringRef File,const Twine & Message)464 [[noreturn]] void objdump::reportError(StringRef File, const Twine &Message) {
465   outs().flush();
466   WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n";
467   exit(1);
468 }
469 
reportError(Error E,StringRef FileName,StringRef ArchiveName,StringRef ArchitectureName)470 [[noreturn]] void objdump::reportError(Error E, StringRef FileName,
471                                        StringRef ArchiveName,
472                                        StringRef ArchitectureName) {
473   assert(E);
474   outs().flush();
475   WithColor::error(errs(), ToolName);
476   if (ArchiveName != "")
477     errs() << ArchiveName << "(" << FileName << ")";
478   else
479     errs() << "'" << FileName << "'";
480   if (!ArchitectureName.empty())
481     errs() << " (for architecture " << ArchitectureName << ")";
482   errs() << ": ";
483   logAllUnhandledErrors(std::move(E), errs());
484   exit(1);
485 }
486 
reportCmdLineWarning(const Twine & Message)487 static void reportCmdLineWarning(const Twine &Message) {
488   WithColor::warning(errs(), ToolName) << Message << "\n";
489 }
490 
reportCmdLineError(const Twine & Message)491 [[noreturn]] static void reportCmdLineError(const Twine &Message) {
492   WithColor::error(errs(), ToolName) << Message << "\n";
493   exit(1);
494 }
495 
warnOnNoMatchForSections()496 static void warnOnNoMatchForSections() {
497   SetVector<StringRef> MissingSections;
498   for (StringRef S : FilterSections) {
499     if (FoundSectionSet.count(S))
500       return;
501     // User may specify a unnamed section. Don't warn for it.
502     if (!S.empty())
503       MissingSections.insert(S);
504   }
505 
506   // Warn only if no section in FilterSections is matched.
507   for (StringRef S : MissingSections)
508     reportCmdLineWarning("section '" + S +
509                          "' mentioned in a -j/--section option, but not "
510                          "found in any input file");
511 }
512 
getTarget(const ObjectFile * Obj)513 static const Target *getTarget(const ObjectFile *Obj) {
514   // Figure out the target triple.
515   Triple TheTriple("unknown-unknown-unknown");
516   if (TripleName.empty()) {
517     TheTriple = Obj->makeTriple();
518   } else {
519     TheTriple.setTriple(Triple::normalize(TripleName));
520     auto Arch = Obj->getArch();
521     if (Arch == Triple::arm || Arch == Triple::armeb)
522       Obj->setARMSubArch(TheTriple);
523   }
524 
525   // Get the target specific parser.
526   std::string Error;
527   const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple,
528                                                          Error);
529   if (!TheTarget)
530     reportError(Obj->getFileName(), "can't find target: " + Error);
531 
532   // Update the triple name and return the found target.
533   TripleName = TheTriple.getTriple();
534   return TheTarget;
535 }
536 
isRelocAddressLess(RelocationRef A,RelocationRef B)537 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) {
538   return A.getOffset() < B.getOffset();
539 }
540 
getRelocationValueString(const RelocationRef & Rel,bool SymbolDescription,SmallVectorImpl<char> & Result)541 static Error getRelocationValueString(const RelocationRef &Rel,
542                                       bool SymbolDescription,
543                                       SmallVectorImpl<char> &Result) {
544   const ObjectFile *Obj = Rel.getObject();
545   if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj))
546     return getELFRelocationValueString(ELF, Rel, Result);
547   if (auto *COFF = dyn_cast<COFFObjectFile>(Obj))
548     return getCOFFRelocationValueString(COFF, Rel, Result);
549   if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj))
550     return getWasmRelocationValueString(Wasm, Rel, Result);
551   if (auto *MachO = dyn_cast<MachOObjectFile>(Obj))
552     return getMachORelocationValueString(MachO, Rel, Result);
553   if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj))
554     return getXCOFFRelocationValueString(*XCOFF, Rel, SymbolDescription,
555                                          Result);
556   llvm_unreachable("unknown object file format");
557 }
558 
559 /// Indicates whether this relocation should hidden when listing
560 /// relocations, usually because it is the trailing part of a multipart
561 /// relocation that will be printed as part of the leading relocation.
getHidden(RelocationRef RelRef)562 static bool getHidden(RelocationRef RelRef) {
563   auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject());
564   if (!MachO)
565     return false;
566 
567   unsigned Arch = MachO->getArch();
568   DataRefImpl Rel = RelRef.getRawDataRefImpl();
569   uint64_t Type = MachO->getRelocationType(Rel);
570 
571   // On arches that use the generic relocations, GENERIC_RELOC_PAIR
572   // is always hidden.
573   if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc)
574     return Type == MachO::GENERIC_RELOC_PAIR;
575 
576   if (Arch == Triple::x86_64) {
577     // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
578     // an X86_64_RELOC_SUBTRACTOR.
579     if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) {
580       DataRefImpl RelPrev = Rel;
581       RelPrev.d.a--;
582       uint64_t PrevType = MachO->getRelocationType(RelPrev);
583       if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR)
584         return true;
585     }
586   }
587 
588   return false;
589 }
590 
591 /// Get the column at which we want to start printing the instruction
592 /// disassembly, taking into account anything which appears to the left of it.
getInstStartColumn(const MCSubtargetInfo & STI)593 unsigned objdump::getInstStartColumn(const MCSubtargetInfo &STI) {
594   return !ShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24;
595 }
596 
AlignToInstStartColumn(size_t Start,const MCSubtargetInfo & STI,raw_ostream & OS)597 static void AlignToInstStartColumn(size_t Start, const MCSubtargetInfo &STI,
598                                    raw_ostream &OS) {
599   // The output of printInst starts with a tab. Print some spaces so that
600   // the tab has 1 column and advances to the target tab stop.
601   unsigned TabStop = getInstStartColumn(STI);
602   unsigned Column = OS.tell() - Start;
603   OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8);
604 }
605 
printRawData(ArrayRef<uint8_t> Bytes,uint64_t Address,formatted_raw_ostream & OS,MCSubtargetInfo const & STI)606 void objdump::printRawData(ArrayRef<uint8_t> Bytes, uint64_t Address,
607                            formatted_raw_ostream &OS,
608                            MCSubtargetInfo const &STI) {
609   size_t Start = OS.tell();
610   if (LeadingAddr)
611     OS << format("%8" PRIx64 ":", Address);
612   if (ShowRawInsn) {
613     OS << ' ';
614     dumpBytes(Bytes, OS);
615   }
616   AlignToInstStartColumn(Start, STI, OS);
617 }
618 
619 namespace {
620 
isAArch64Elf(const ObjectFile & Obj)621 static bool isAArch64Elf(const ObjectFile &Obj) {
622   const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj);
623   return Elf && Elf->getEMachine() == ELF::EM_AARCH64;
624 }
625 
isArmElf(const ObjectFile & Obj)626 static bool isArmElf(const ObjectFile &Obj) {
627   const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj);
628   return Elf && Elf->getEMachine() == ELF::EM_ARM;
629 }
630 
isCSKYElf(const ObjectFile & Obj)631 static bool isCSKYElf(const ObjectFile &Obj) {
632   const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj);
633   return Elf && Elf->getEMachine() == ELF::EM_CSKY;
634 }
635 
hasMappingSymbols(const ObjectFile & Obj)636 static bool hasMappingSymbols(const ObjectFile &Obj) {
637   return isArmElf(Obj) || isAArch64Elf(Obj) || isCSKYElf(Obj) ;
638 }
639 
printRelocation(formatted_raw_ostream & OS,StringRef FileName,const RelocationRef & Rel,uint64_t Address,bool Is64Bits)640 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName,
641                             const RelocationRef &Rel, uint64_t Address,
642                             bool Is64Bits) {
643   StringRef Fmt = Is64Bits ? "%016" PRIx64 ":  " : "%08" PRIx64 ":  ";
644   SmallString<16> Name;
645   SmallString<32> Val;
646   Rel.getTypeName(Name);
647   if (Error E = getRelocationValueString(Rel, SymbolDescription, Val))
648     reportError(std::move(E), FileName);
649   OS << (Is64Bits || !LeadingAddr ? "\t\t" : "\t\t\t");
650   if (LeadingAddr)
651     OS << format(Fmt.data(), Address);
652   OS << Name << "\t" << Val;
653 }
654 
printBTFRelocation(formatted_raw_ostream & FOS,llvm::BTFParser & BTF,object::SectionedAddress Address,LiveVariablePrinter & LVP)655 static void printBTFRelocation(formatted_raw_ostream &FOS, llvm::BTFParser &BTF,
656                                object::SectionedAddress Address,
657                                LiveVariablePrinter &LVP) {
658   const llvm::BTF::BPFFieldReloc *Reloc = BTF.findFieldReloc(Address);
659   if (!Reloc)
660     return;
661 
662   SmallString<64> Val;
663   BTF.symbolize(Reloc, Val);
664   FOS << "\t\t";
665   if (LeadingAddr)
666     FOS << format("%016" PRIx64 ":  ", Address.Address + AdjustVMA);
667   FOS << "CO-RE " << Val;
668   LVP.printAfterOtherLine(FOS, true);
669 }
670 
671 class PrettyPrinter {
672 public:
673   virtual ~PrettyPrinter() = default;
674   virtual void
printInst(MCInstPrinter & IP,const MCInst * MI,ArrayRef<uint8_t> Bytes,object::SectionedAddress Address,formatted_raw_ostream & OS,StringRef Annot,MCSubtargetInfo const & STI,SourcePrinter * SP,StringRef ObjectFilename,std::vector<RelocationRef> * Rels,LiveVariablePrinter & LVP)675   printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
676             object::SectionedAddress Address, formatted_raw_ostream &OS,
677             StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
678             StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
679             LiveVariablePrinter &LVP) {
680     if (SP && (PrintSource || PrintLines))
681       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
682     LVP.printBetweenInsts(OS, false);
683 
684     printRawData(Bytes, Address.Address, OS, STI);
685 
686     if (MI) {
687       // See MCInstPrinter::printInst. On targets where a PC relative immediate
688       // is relative to the next instruction and the length of a MCInst is
689       // difficult to measure (x86), this is the address of the next
690       // instruction.
691       uint64_t Addr =
692           Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0);
693       IP.printInst(MI, Addr, "", STI, OS);
694     } else
695       OS << "\t<unknown>";
696   }
697 };
698 PrettyPrinter PrettyPrinterInst;
699 
700 class HexagonPrettyPrinter : public PrettyPrinter {
701 public:
printLead(ArrayRef<uint8_t> Bytes,uint64_t Address,formatted_raw_ostream & OS)702   void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
703                  formatted_raw_ostream &OS) {
704     uint32_t opcode =
705       (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
706     if (LeadingAddr)
707       OS << format("%8" PRIx64 ":", Address);
708     if (ShowRawInsn) {
709       OS << "\t";
710       dumpBytes(Bytes.slice(0, 4), OS);
711       OS << format("\t%08" PRIx32, opcode);
712     }
713   }
printInst(MCInstPrinter & IP,const MCInst * MI,ArrayRef<uint8_t> Bytes,object::SectionedAddress Address,formatted_raw_ostream & OS,StringRef Annot,MCSubtargetInfo const & STI,SourcePrinter * SP,StringRef ObjectFilename,std::vector<RelocationRef> * Rels,LiveVariablePrinter & LVP)714   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
715                  object::SectionedAddress Address, formatted_raw_ostream &OS,
716                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
717                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
718                  LiveVariablePrinter &LVP) override {
719     if (SP && (PrintSource || PrintLines))
720       SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
721     if (!MI) {
722       printLead(Bytes, Address.Address, OS);
723       OS << " <unknown>";
724       return;
725     }
726     std::string Buffer;
727     {
728       raw_string_ostream TempStream(Buffer);
729       IP.printInst(MI, Address.Address, "", STI, TempStream);
730     }
731     StringRef Contents(Buffer);
732     // Split off bundle attributes
733     auto PacketBundle = Contents.rsplit('\n');
734     // Split off first instruction from the rest
735     auto HeadTail = PacketBundle.first.split('\n');
736     auto Preamble = " { ";
737     auto Separator = "";
738 
739     // Hexagon's packets require relocations to be inline rather than
740     // clustered at the end of the packet.
741     std::vector<RelocationRef>::const_iterator RelCur = Rels->begin();
742     std::vector<RelocationRef>::const_iterator RelEnd = Rels->end();
743     auto PrintReloc = [&]() -> void {
744       while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) {
745         if (RelCur->getOffset() == Address.Address) {
746           printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false);
747           return;
748         }
749         ++RelCur;
750       }
751     };
752 
753     while (!HeadTail.first.empty()) {
754       OS << Separator;
755       Separator = "\n";
756       if (SP && (PrintSource || PrintLines))
757         SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
758       printLead(Bytes, Address.Address, OS);
759       OS << Preamble;
760       Preamble = "   ";
761       StringRef Inst;
762       auto Duplex = HeadTail.first.split('\v');
763       if (!Duplex.second.empty()) {
764         OS << Duplex.first;
765         OS << "; ";
766         Inst = Duplex.second;
767       }
768       else
769         Inst = HeadTail.first;
770       OS << Inst;
771       HeadTail = HeadTail.second.split('\n');
772       if (HeadTail.first.empty())
773         OS << " } " << PacketBundle.second;
774       PrintReloc();
775       Bytes = Bytes.slice(4);
776       Address.Address += 4;
777     }
778   }
779 };
780 HexagonPrettyPrinter HexagonPrettyPrinterInst;
781 
782 class AMDGCNPrettyPrinter : public PrettyPrinter {
783 public:
printInst(MCInstPrinter & IP,const MCInst * MI,ArrayRef<uint8_t> Bytes,object::SectionedAddress Address,formatted_raw_ostream & OS,StringRef Annot,MCSubtargetInfo const & STI,SourcePrinter * SP,StringRef ObjectFilename,std::vector<RelocationRef> * Rels,LiveVariablePrinter & LVP)784   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
785                  object::SectionedAddress Address, formatted_raw_ostream &OS,
786                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
787                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
788                  LiveVariablePrinter &LVP) override {
789     if (SP && (PrintSource || PrintLines))
790       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
791 
792     if (MI) {
793       SmallString<40> InstStr;
794       raw_svector_ostream IS(InstStr);
795 
796       IP.printInst(MI, Address.Address, "", STI, IS);
797 
798       OS << left_justify(IS.str(), 60);
799     } else {
800       // an unrecognized encoding - this is probably data so represent it
801       // using the .long directive, or .byte directive if fewer than 4 bytes
802       // remaining
803       if (Bytes.size() >= 4) {
804         OS << format(
805             "\t.long 0x%08" PRIx32 " ",
806             support::endian::read32<llvm::endianness::little>(Bytes.data()));
807         OS.indent(42);
808       } else {
809           OS << format("\t.byte 0x%02" PRIx8, Bytes[0]);
810           for (unsigned int i = 1; i < Bytes.size(); i++)
811             OS << format(", 0x%02" PRIx8, Bytes[i]);
812           OS.indent(55 - (6 * Bytes.size()));
813       }
814     }
815 
816     OS << format("// %012" PRIX64 ":", Address.Address);
817     if (Bytes.size() >= 4) {
818       // D should be casted to uint32_t here as it is passed by format to
819       // snprintf as vararg.
820       for (uint32_t D :
821            ArrayRef(reinterpret_cast<const support::little32_t *>(Bytes.data()),
822                     Bytes.size() / 4))
823           OS << format(" %08" PRIX32, D);
824     } else {
825       for (unsigned char B : Bytes)
826         OS << format(" %02" PRIX8, B);
827     }
828 
829     if (!Annot.empty())
830       OS << " // " << Annot;
831   }
832 };
833 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst;
834 
835 class BPFPrettyPrinter : public PrettyPrinter {
836 public:
printInst(MCInstPrinter & IP,const MCInst * MI,ArrayRef<uint8_t> Bytes,object::SectionedAddress Address,formatted_raw_ostream & OS,StringRef Annot,MCSubtargetInfo const & STI,SourcePrinter * SP,StringRef ObjectFilename,std::vector<RelocationRef> * Rels,LiveVariablePrinter & LVP)837   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
838                  object::SectionedAddress Address, formatted_raw_ostream &OS,
839                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
840                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
841                  LiveVariablePrinter &LVP) override {
842     if (SP && (PrintSource || PrintLines))
843       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
844     if (LeadingAddr)
845       OS << format("%8" PRId64 ":", Address.Address / 8);
846     if (ShowRawInsn) {
847       OS << "\t";
848       dumpBytes(Bytes, OS);
849     }
850     if (MI)
851       IP.printInst(MI, Address.Address, "", STI, OS);
852     else
853       OS << "\t<unknown>";
854   }
855 };
856 BPFPrettyPrinter BPFPrettyPrinterInst;
857 
858 class ARMPrettyPrinter : public PrettyPrinter {
859 public:
printInst(MCInstPrinter & IP,const MCInst * MI,ArrayRef<uint8_t> Bytes,object::SectionedAddress Address,formatted_raw_ostream & OS,StringRef Annot,MCSubtargetInfo const & STI,SourcePrinter * SP,StringRef ObjectFilename,std::vector<RelocationRef> * Rels,LiveVariablePrinter & LVP)860   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
861                  object::SectionedAddress Address, formatted_raw_ostream &OS,
862                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
863                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
864                  LiveVariablePrinter &LVP) override {
865     if (SP && (PrintSource || PrintLines))
866       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
867     LVP.printBetweenInsts(OS, false);
868 
869     size_t Start = OS.tell();
870     if (LeadingAddr)
871       OS << format("%8" PRIx64 ":", Address.Address);
872     if (ShowRawInsn) {
873       size_t Pos = 0, End = Bytes.size();
874       if (STI.checkFeatures("+thumb-mode")) {
875         for (; Pos + 2 <= End; Pos += 2)
876           OS << ' '
877              << format_hex_no_prefix(
878                     llvm::support::endian::read<uint16_t>(
879                         Bytes.data() + Pos, InstructionEndianness),
880                     4);
881       } else {
882         for (; Pos + 4 <= End; Pos += 4)
883           OS << ' '
884              << format_hex_no_prefix(
885                     llvm::support::endian::read<uint32_t>(
886                         Bytes.data() + Pos, InstructionEndianness),
887                     8);
888       }
889       if (Pos < End) {
890         OS << ' ';
891         dumpBytes(Bytes.slice(Pos), OS);
892       }
893     }
894 
895     AlignToInstStartColumn(Start, STI, OS);
896 
897     if (MI) {
898       IP.printInst(MI, Address.Address, "", STI, OS);
899     } else
900       OS << "\t<unknown>";
901   }
902 
setInstructionEndianness(llvm::endianness Endianness)903   void setInstructionEndianness(llvm::endianness Endianness) {
904     InstructionEndianness = Endianness;
905   }
906 
907 private:
908   llvm::endianness InstructionEndianness = llvm::endianness::little;
909 };
910 ARMPrettyPrinter ARMPrettyPrinterInst;
911 
912 class AArch64PrettyPrinter : public PrettyPrinter {
913 public:
printInst(MCInstPrinter & IP,const MCInst * MI,ArrayRef<uint8_t> Bytes,object::SectionedAddress Address,formatted_raw_ostream & OS,StringRef Annot,MCSubtargetInfo const & STI,SourcePrinter * SP,StringRef ObjectFilename,std::vector<RelocationRef> * Rels,LiveVariablePrinter & LVP)914   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
915                  object::SectionedAddress Address, formatted_raw_ostream &OS,
916                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
917                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
918                  LiveVariablePrinter &LVP) override {
919     if (SP && (PrintSource || PrintLines))
920       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
921     LVP.printBetweenInsts(OS, false);
922 
923     size_t Start = OS.tell();
924     if (LeadingAddr)
925       OS << format("%8" PRIx64 ":", Address.Address);
926     if (ShowRawInsn) {
927       size_t Pos = 0, End = Bytes.size();
928       for (; Pos + 4 <= End; Pos += 4)
929         OS << ' '
930            << format_hex_no_prefix(
931                   llvm::support::endian::read<uint32_t>(
932                       Bytes.data() + Pos, llvm::endianness::little),
933                   8);
934       if (Pos < End) {
935         OS << ' ';
936         dumpBytes(Bytes.slice(Pos), OS);
937       }
938     }
939 
940     AlignToInstStartColumn(Start, STI, OS);
941 
942     if (MI) {
943       IP.printInst(MI, Address.Address, "", STI, OS);
944     } else
945       OS << "\t<unknown>";
946   }
947 };
948 AArch64PrettyPrinter AArch64PrettyPrinterInst;
949 
950 class RISCVPrettyPrinter : public PrettyPrinter {
951 public:
printInst(MCInstPrinter & IP,const MCInst * MI,ArrayRef<uint8_t> Bytes,object::SectionedAddress Address,formatted_raw_ostream & OS,StringRef Annot,MCSubtargetInfo const & STI,SourcePrinter * SP,StringRef ObjectFilename,std::vector<RelocationRef> * Rels,LiveVariablePrinter & LVP)952   void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
953                  object::SectionedAddress Address, formatted_raw_ostream &OS,
954                  StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
955                  StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
956                  LiveVariablePrinter &LVP) override {
957     if (SP && (PrintSource || PrintLines))
958       SP->printSourceLine(OS, Address, ObjectFilename, LVP);
959     LVP.printBetweenInsts(OS, false);
960 
961     size_t Start = OS.tell();
962     if (LeadingAddr)
963       OS << format("%8" PRIx64 ":", Address.Address);
964     if (ShowRawInsn) {
965       size_t Pos = 0, End = Bytes.size();
966       if (End % 4 == 0) {
967         // 32-bit and 64-bit instructions.
968         for (; Pos + 4 <= End; Pos += 4)
969           OS << ' '
970              << format_hex_no_prefix(
971                     llvm::support::endian::read<uint32_t>(
972                         Bytes.data() + Pos, llvm::endianness::little),
973                     8);
974       } else if (End % 2 == 0) {
975         // 16-bit and 48-bits instructions.
976         for (; Pos + 2 <= End; Pos += 2)
977           OS << ' '
978              << format_hex_no_prefix(
979                     llvm::support::endian::read<uint16_t>(
980                         Bytes.data() + Pos, llvm::endianness::little),
981                     4);
982       }
983       if (Pos < End) {
984         OS << ' ';
985         dumpBytes(Bytes.slice(Pos), OS);
986       }
987     }
988 
989     AlignToInstStartColumn(Start, STI, OS);
990 
991     if (MI) {
992       IP.printInst(MI, Address.Address, "", STI, OS);
993     } else
994       OS << "\t<unknown>";
995   }
996 };
997 RISCVPrettyPrinter RISCVPrettyPrinterInst;
998 
selectPrettyPrinter(Triple const & Triple)999 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) {
1000   switch(Triple.getArch()) {
1001   default:
1002     return PrettyPrinterInst;
1003   case Triple::hexagon:
1004     return HexagonPrettyPrinterInst;
1005   case Triple::amdgcn:
1006     return AMDGCNPrettyPrinterInst;
1007   case Triple::bpfel:
1008   case Triple::bpfeb:
1009     return BPFPrettyPrinterInst;
1010   case Triple::arm:
1011   case Triple::armeb:
1012   case Triple::thumb:
1013   case Triple::thumbeb:
1014     return ARMPrettyPrinterInst;
1015   case Triple::aarch64:
1016   case Triple::aarch64_be:
1017   case Triple::aarch64_32:
1018     return AArch64PrettyPrinterInst;
1019   case Triple::riscv32:
1020   case Triple::riscv64:
1021     return RISCVPrettyPrinterInst;
1022   }
1023 }
1024 
1025 class DisassemblerTarget {
1026 public:
1027   const Target *TheTarget;
1028   std::unique_ptr<const MCSubtargetInfo> SubtargetInfo;
1029   std::shared_ptr<MCContext> Context;
1030   std::unique_ptr<MCDisassembler> DisAsm;
1031   std::shared_ptr<MCInstrAnalysis> InstrAnalysis;
1032   std::shared_ptr<MCInstPrinter> InstPrinter;
1033   PrettyPrinter *Printer;
1034 
1035   DisassemblerTarget(const Target *TheTarget, ObjectFile &Obj,
1036                      StringRef TripleName, StringRef MCPU,
1037                      SubtargetFeatures &Features);
1038   DisassemblerTarget(DisassemblerTarget &Other, SubtargetFeatures &Features);
1039 
1040 private:
1041   MCTargetOptions Options;
1042   std::shared_ptr<const MCRegisterInfo> RegisterInfo;
1043   std::shared_ptr<const MCAsmInfo> AsmInfo;
1044   std::shared_ptr<const MCInstrInfo> InstrInfo;
1045   std::shared_ptr<MCObjectFileInfo> ObjectFileInfo;
1046 };
1047 
DisassemblerTarget(const Target * TheTarget,ObjectFile & Obj,StringRef TripleName,StringRef MCPU,SubtargetFeatures & Features)1048 DisassemblerTarget::DisassemblerTarget(const Target *TheTarget, ObjectFile &Obj,
1049                                        StringRef TripleName, StringRef MCPU,
1050                                        SubtargetFeatures &Features)
1051     : TheTarget(TheTarget),
1052       Printer(&selectPrettyPrinter(Triple(TripleName))),
1053       RegisterInfo(TheTarget->createMCRegInfo(TripleName)) {
1054   if (!RegisterInfo)
1055     reportError(Obj.getFileName(), "no register info for target " + TripleName);
1056 
1057   // Set up disassembler.
1058   AsmInfo.reset(TheTarget->createMCAsmInfo(*RegisterInfo, TripleName, Options));
1059   if (!AsmInfo)
1060     reportError(Obj.getFileName(), "no assembly info for target " + TripleName);
1061 
1062   SubtargetInfo.reset(
1063       TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString()));
1064   if (!SubtargetInfo)
1065     reportError(Obj.getFileName(),
1066                 "no subtarget info for target " + TripleName);
1067   InstrInfo.reset(TheTarget->createMCInstrInfo());
1068   if (!InstrInfo)
1069     reportError(Obj.getFileName(),
1070                 "no instruction info for target " + TripleName);
1071   Context =
1072       std::make_shared<MCContext>(Triple(TripleName), AsmInfo.get(),
1073                                   RegisterInfo.get(), SubtargetInfo.get());
1074 
1075   // FIXME: for now initialize MCObjectFileInfo with default values
1076   ObjectFileInfo.reset(
1077       TheTarget->createMCObjectFileInfo(*Context, /*PIC=*/false));
1078   Context->setObjectFileInfo(ObjectFileInfo.get());
1079 
1080   DisAsm.reset(TheTarget->createMCDisassembler(*SubtargetInfo, *Context));
1081   if (!DisAsm)
1082     reportError(Obj.getFileName(), "no disassembler for target " + TripleName);
1083 
1084   if (auto *ELFObj = dyn_cast<ELFObjectFileBase>(&Obj))
1085     DisAsm->setABIVersion(ELFObj->getEIdentABIVersion());
1086 
1087   InstrAnalysis.reset(TheTarget->createMCInstrAnalysis(InstrInfo.get()));
1088 
1089   int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1090   InstPrinter.reset(TheTarget->createMCInstPrinter(Triple(TripleName),
1091                                                    AsmPrinterVariant, *AsmInfo,
1092                                                    *InstrInfo, *RegisterInfo));
1093   if (!InstPrinter)
1094     reportError(Obj.getFileName(),
1095                 "no instruction printer for target " + TripleName);
1096   InstPrinter->setPrintImmHex(PrintImmHex);
1097   InstPrinter->setPrintBranchImmAsAddress(true);
1098   InstPrinter->setSymbolizeOperands(SymbolizeOperands);
1099   InstPrinter->setMCInstrAnalysis(InstrAnalysis.get());
1100 
1101   switch (DisassemblyColor) {
1102   case ColorOutput::Enable:
1103     InstPrinter->setUseColor(true);
1104     break;
1105   case ColorOutput::Auto:
1106     InstPrinter->setUseColor(outs().has_colors());
1107     break;
1108   case ColorOutput::Disable:
1109   case ColorOutput::Invalid:
1110     InstPrinter->setUseColor(false);
1111     break;
1112   };
1113 }
1114 
DisassemblerTarget(DisassemblerTarget & Other,SubtargetFeatures & Features)1115 DisassemblerTarget::DisassemblerTarget(DisassemblerTarget &Other,
1116                                        SubtargetFeatures &Features)
1117     : TheTarget(Other.TheTarget),
1118       SubtargetInfo(TheTarget->createMCSubtargetInfo(TripleName, MCPU,
1119                                                      Features.getString())),
1120       Context(Other.Context),
1121       DisAsm(TheTarget->createMCDisassembler(*SubtargetInfo, *Context)),
1122       InstrAnalysis(Other.InstrAnalysis), InstPrinter(Other.InstPrinter),
1123       Printer(Other.Printer), RegisterInfo(Other.RegisterInfo),
1124       AsmInfo(Other.AsmInfo), InstrInfo(Other.InstrInfo),
1125       ObjectFileInfo(Other.ObjectFileInfo) {}
1126 } // namespace
1127 
getElfSymbolType(const ObjectFile & Obj,const SymbolRef & Sym)1128 static uint8_t getElfSymbolType(const ObjectFile &Obj, const SymbolRef &Sym) {
1129   assert(Obj.isELF());
1130   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj))
1131     return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()),
1132                          Obj.getFileName())
1133         ->getType();
1134   if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj))
1135     return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()),
1136                          Obj.getFileName())
1137         ->getType();
1138   if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj))
1139     return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()),
1140                          Obj.getFileName())
1141         ->getType();
1142   if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj))
1143     return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()),
1144                          Obj.getFileName())
1145         ->getType();
1146   llvm_unreachable("Unsupported binary format");
1147 }
1148 
1149 template <class ELFT>
1150 static void
addDynamicElfSymbols(const ELFObjectFile<ELFT> & Obj,std::map<SectionRef,SectionSymbolsTy> & AllSymbols)1151 addDynamicElfSymbols(const ELFObjectFile<ELFT> &Obj,
1152                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
1153   for (auto Symbol : Obj.getDynamicSymbolIterators()) {
1154     uint8_t SymbolType = Symbol.getELFType();
1155     if (SymbolType == ELF::STT_SECTION)
1156       continue;
1157 
1158     uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj.getFileName());
1159     // ELFSymbolRef::getAddress() returns size instead of value for common
1160     // symbols which is not desirable for disassembly output. Overriding.
1161     if (SymbolType == ELF::STT_COMMON)
1162       Address = unwrapOrError(Obj.getSymbol(Symbol.getRawDataRefImpl()),
1163                               Obj.getFileName())
1164                     ->st_value;
1165 
1166     StringRef Name = unwrapOrError(Symbol.getName(), Obj.getFileName());
1167     if (Name.empty())
1168       continue;
1169 
1170     section_iterator SecI =
1171         unwrapOrError(Symbol.getSection(), Obj.getFileName());
1172     if (SecI == Obj.section_end())
1173       continue;
1174 
1175     AllSymbols[*SecI].emplace_back(Address, Name, SymbolType);
1176   }
1177 }
1178 
1179 static void
addDynamicElfSymbols(const ELFObjectFileBase & Obj,std::map<SectionRef,SectionSymbolsTy> & AllSymbols)1180 addDynamicElfSymbols(const ELFObjectFileBase &Obj,
1181                      std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
1182   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj))
1183     addDynamicElfSymbols(*Elf32LEObj, AllSymbols);
1184   else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj))
1185     addDynamicElfSymbols(*Elf64LEObj, AllSymbols);
1186   else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj))
1187     addDynamicElfSymbols(*Elf32BEObj, AllSymbols);
1188   else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj))
1189     addDynamicElfSymbols(*Elf64BEObj, AllSymbols);
1190   else
1191     llvm_unreachable("Unsupported binary format");
1192 }
1193 
getWasmCodeSection(const WasmObjectFile & Obj)1194 static std::optional<SectionRef> getWasmCodeSection(const WasmObjectFile &Obj) {
1195   for (auto SecI : Obj.sections()) {
1196     const WasmSection &Section = Obj.getWasmSection(SecI);
1197     if (Section.Type == wasm::WASM_SEC_CODE)
1198       return SecI;
1199   }
1200   return std::nullopt;
1201 }
1202 
1203 static void
addMissingWasmCodeSymbols(const WasmObjectFile & Obj,std::map<SectionRef,SectionSymbolsTy> & AllSymbols)1204 addMissingWasmCodeSymbols(const WasmObjectFile &Obj,
1205                           std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
1206   std::optional<SectionRef> Section = getWasmCodeSection(Obj);
1207   if (!Section)
1208     return;
1209   SectionSymbolsTy &Symbols = AllSymbols[*Section];
1210 
1211   std::set<uint64_t> SymbolAddresses;
1212   for (const auto &Sym : Symbols)
1213     SymbolAddresses.insert(Sym.Addr);
1214 
1215   for (const wasm::WasmFunction &Function : Obj.functions()) {
1216     // This adjustment mirrors the one in WasmObjectFile::getSymbolAddress.
1217     uint32_t Adjustment = Obj.isRelocatableObject() || Obj.isSharedObject()
1218                               ? 0
1219                               : Section->getAddress();
1220     uint64_t Address = Function.CodeSectionOffset + Adjustment;
1221     // Only add fallback symbols for functions not already present in the symbol
1222     // table.
1223     if (SymbolAddresses.count(Address))
1224       continue;
1225     // This function has no symbol, so it should have no SymbolName.
1226     assert(Function.SymbolName.empty());
1227     // We use DebugName for the name, though it may be empty if there is no
1228     // "name" custom section, or that section is missing a name for this
1229     // function.
1230     StringRef Name = Function.DebugName;
1231     Symbols.emplace_back(Address, Name, ELF::STT_NOTYPE);
1232   }
1233 }
1234 
addPltEntries(const ObjectFile & Obj,std::map<SectionRef,SectionSymbolsTy> & AllSymbols,StringSaver & Saver)1235 static void addPltEntries(const ObjectFile &Obj,
1236                           std::map<SectionRef, SectionSymbolsTy> &AllSymbols,
1237                           StringSaver &Saver) {
1238   auto *ElfObj = dyn_cast<ELFObjectFileBase>(&Obj);
1239   if (!ElfObj)
1240     return;
1241   DenseMap<StringRef, SectionRef> Sections;
1242   for (SectionRef Section : Obj.sections()) {
1243     Expected<StringRef> SecNameOrErr = Section.getName();
1244     if (!SecNameOrErr) {
1245       consumeError(SecNameOrErr.takeError());
1246       continue;
1247     }
1248     Sections[*SecNameOrErr] = Section;
1249   }
1250   for (auto Plt : ElfObj->getPltEntries()) {
1251     if (Plt.Symbol) {
1252       SymbolRef Symbol(*Plt.Symbol, ElfObj);
1253       uint8_t SymbolType = getElfSymbolType(Obj, Symbol);
1254       if (Expected<StringRef> NameOrErr = Symbol.getName()) {
1255         if (!NameOrErr->empty())
1256           AllSymbols[Sections[Plt.Section]].emplace_back(
1257               Plt.Address, Saver.save((*NameOrErr + "@plt").str()), SymbolType);
1258         continue;
1259       } else {
1260         // The warning has been reported in disassembleObject().
1261         consumeError(NameOrErr.takeError());
1262       }
1263     }
1264     reportWarning("PLT entry at 0x" + Twine::utohexstr(Plt.Address) +
1265                       " references an invalid symbol",
1266                   Obj.getFileName());
1267   }
1268 }
1269 
1270 // Normally the disassembly output will skip blocks of zeroes. This function
1271 // returns the number of zero bytes that can be skipped when dumping the
1272 // disassembly of the instructions in Buf.
countSkippableZeroBytes(ArrayRef<uint8_t> Buf)1273 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) {
1274   // Find the number of leading zeroes.
1275   size_t N = 0;
1276   while (N < Buf.size() && !Buf[N])
1277     ++N;
1278 
1279   // We may want to skip blocks of zero bytes, but unless we see
1280   // at least 8 of them in a row.
1281   if (N < 8)
1282     return 0;
1283 
1284   // We skip zeroes in multiples of 4 because do not want to truncate an
1285   // instruction if it starts with a zero byte.
1286   return N & ~0x3;
1287 }
1288 
1289 // Returns a map from sections to their relocations.
1290 static std::map<SectionRef, std::vector<RelocationRef>>
getRelocsMap(object::ObjectFile const & Obj)1291 getRelocsMap(object::ObjectFile const &Obj) {
1292   std::map<SectionRef, std::vector<RelocationRef>> Ret;
1293   uint64_t I = (uint64_t)-1;
1294   for (SectionRef Sec : Obj.sections()) {
1295     ++I;
1296     Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection();
1297     if (!RelocatedOrErr)
1298       reportError(Obj.getFileName(),
1299                   "section (" + Twine(I) +
1300                       "): failed to get a relocated section: " +
1301                       toString(RelocatedOrErr.takeError()));
1302 
1303     section_iterator Relocated = *RelocatedOrErr;
1304     if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep)
1305       continue;
1306     std::vector<RelocationRef> &V = Ret[*Relocated];
1307     append_range(V, Sec.relocations());
1308     // Sort relocations by address.
1309     llvm::stable_sort(V, isRelocAddressLess);
1310   }
1311   return Ret;
1312 }
1313 
1314 // Used for --adjust-vma to check if address should be adjusted by the
1315 // specified value for a given section.
1316 // For ELF we do not adjust non-allocatable sections like debug ones,
1317 // because they are not loadable.
1318 // TODO: implement for other file formats.
shouldAdjustVA(const SectionRef & Section)1319 static bool shouldAdjustVA(const SectionRef &Section) {
1320   const ObjectFile *Obj = Section.getObject();
1321   if (Obj->isELF())
1322     return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC;
1323   return false;
1324 }
1325 
1326 
1327 typedef std::pair<uint64_t, char> MappingSymbolPair;
getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols,uint64_t Address)1328 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols,
1329                                  uint64_t Address) {
1330   auto It =
1331       partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) {
1332         return Val.first <= Address;
1333       });
1334   // Return zero for any address before the first mapping symbol; this means
1335   // we should use the default disassembly mode, depending on the target.
1336   if (It == MappingSymbols.begin())
1337     return '\x00';
1338   return (It - 1)->second;
1339 }
1340 
dumpARMELFData(uint64_t SectionAddr,uint64_t Index,uint64_t End,const ObjectFile & Obj,ArrayRef<uint8_t> Bytes,ArrayRef<MappingSymbolPair> MappingSymbols,const MCSubtargetInfo & STI,raw_ostream & OS)1341 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index,
1342                                uint64_t End, const ObjectFile &Obj,
1343                                ArrayRef<uint8_t> Bytes,
1344                                ArrayRef<MappingSymbolPair> MappingSymbols,
1345                                const MCSubtargetInfo &STI, raw_ostream &OS) {
1346   llvm::endianness Endian =
1347       Obj.isLittleEndian() ? llvm::endianness::little : llvm::endianness::big;
1348   size_t Start = OS.tell();
1349   OS << format("%8" PRIx64 ": ", SectionAddr + Index);
1350   if (Index + 4 <= End) {
1351     dumpBytes(Bytes.slice(Index, 4), OS);
1352     AlignToInstStartColumn(Start, STI, OS);
1353     OS << "\t.word\t"
1354            << format_hex(support::endian::read32(Bytes.data() + Index, Endian),
1355                          10);
1356     return 4;
1357   }
1358   if (Index + 2 <= End) {
1359     dumpBytes(Bytes.slice(Index, 2), OS);
1360     AlignToInstStartColumn(Start, STI, OS);
1361     OS << "\t.short\t"
1362        << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 6);
1363     return 2;
1364   }
1365   dumpBytes(Bytes.slice(Index, 1), OS);
1366   AlignToInstStartColumn(Start, STI, OS);
1367   OS << "\t.byte\t" << format_hex(Bytes[Index], 4);
1368   return 1;
1369 }
1370 
dumpELFData(uint64_t SectionAddr,uint64_t Index,uint64_t End,ArrayRef<uint8_t> Bytes)1371 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End,
1372                         ArrayRef<uint8_t> Bytes) {
1373   // print out data up to 8 bytes at a time in hex and ascii
1374   uint8_t AsciiData[9] = {'\0'};
1375   uint8_t Byte;
1376   int NumBytes = 0;
1377 
1378   for (; Index < End; ++Index) {
1379     if (NumBytes == 0)
1380       outs() << format("%8" PRIx64 ":", SectionAddr + Index);
1381     Byte = Bytes.slice(Index)[0];
1382     outs() << format(" %02x", Byte);
1383     AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.';
1384 
1385     uint8_t IndentOffset = 0;
1386     NumBytes++;
1387     if (Index == End - 1 || NumBytes > 8) {
1388       // Indent the space for less than 8 bytes data.
1389       // 2 spaces for byte and one for space between bytes
1390       IndentOffset = 3 * (8 - NumBytes);
1391       for (int Excess = NumBytes; Excess < 8; Excess++)
1392         AsciiData[Excess] = '\0';
1393       NumBytes = 8;
1394     }
1395     if (NumBytes == 8) {
1396       AsciiData[8] = '\0';
1397       outs() << std::string(IndentOffset, ' ') << "         ";
1398       outs() << reinterpret_cast<char *>(AsciiData);
1399       outs() << '\n';
1400       NumBytes = 0;
1401     }
1402   }
1403 }
1404 
createSymbolInfo(const ObjectFile & Obj,const SymbolRef & Symbol,bool IsMappingSymbol)1405 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile &Obj,
1406                                        const SymbolRef &Symbol,
1407                                        bool IsMappingSymbol) {
1408   const StringRef FileName = Obj.getFileName();
1409   const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName);
1410   const StringRef Name = unwrapOrError(Symbol.getName(), FileName);
1411 
1412   if (Obj.isXCOFF() && (SymbolDescription || TracebackTable)) {
1413     const auto &XCOFFObj = cast<XCOFFObjectFile>(Obj);
1414     DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl();
1415 
1416     const uint32_t SymbolIndex = XCOFFObj.getSymbolIndex(SymbolDRI.p);
1417     std::optional<XCOFF::StorageMappingClass> Smc =
1418         getXCOFFSymbolCsectSMC(XCOFFObj, Symbol);
1419     return SymbolInfoTy(Smc, Addr, Name, SymbolIndex,
1420                         isLabel(XCOFFObj, Symbol));
1421   } else if (Obj.isXCOFF()) {
1422     const SymbolRef::Type SymType = unwrapOrError(Symbol.getType(), FileName);
1423     return SymbolInfoTy(Addr, Name, SymType, /*IsMappingSymbol=*/false,
1424                         /*IsXCOFF=*/true);
1425   } else if (Obj.isWasm()) {
1426     uint8_t SymType =
1427         cast<WasmObjectFile>(&Obj)->getWasmSymbol(Symbol).Info.Kind;
1428     return SymbolInfoTy(Addr, Name, SymType, false);
1429   } else {
1430     uint8_t Type =
1431         Obj.isELF() ? getElfSymbolType(Obj, Symbol) : (uint8_t)ELF::STT_NOTYPE;
1432     return SymbolInfoTy(Addr, Name, Type, IsMappingSymbol);
1433   }
1434 }
1435 
createDummySymbolInfo(const ObjectFile & Obj,const uint64_t Addr,StringRef & Name,uint8_t Type)1436 static SymbolInfoTy createDummySymbolInfo(const ObjectFile &Obj,
1437                                           const uint64_t Addr, StringRef &Name,
1438                                           uint8_t Type) {
1439   if (Obj.isXCOFF() && (SymbolDescription || TracebackTable))
1440     return SymbolInfoTy(std::nullopt, Addr, Name, std::nullopt, false);
1441   if (Obj.isWasm())
1442     return SymbolInfoTy(Addr, Name, wasm::WASM_SYMBOL_TYPE_SECTION);
1443   return SymbolInfoTy(Addr, Name, Type);
1444 }
1445 
collectBBAddrMapLabels(const BBAddrMapInfo & FullAddrMap,uint64_t SectionAddr,uint64_t Start,uint64_t End,std::unordered_map<uint64_t,std::vector<BBAddrMapLabel>> & Labels)1446 static void collectBBAddrMapLabels(
1447     const BBAddrMapInfo &FullAddrMap, uint64_t SectionAddr, uint64_t Start,
1448     uint64_t End,
1449     std::unordered_map<uint64_t, std::vector<BBAddrMapLabel>> &Labels) {
1450   if (FullAddrMap.empty())
1451     return;
1452   Labels.clear();
1453   uint64_t StartAddress = SectionAddr + Start;
1454   uint64_t EndAddress = SectionAddr + End;
1455   const BBAddrMapFunctionEntry *FunctionMap =
1456       FullAddrMap.getEntryForAddress(StartAddress);
1457   if (!FunctionMap)
1458     return;
1459   std::optional<size_t> BBRangeIndex =
1460       FunctionMap->getAddrMap().getBBRangeIndexForBaseAddress(StartAddress);
1461   if (!BBRangeIndex)
1462     return;
1463   size_t NumBBEntriesBeforeRange = 0;
1464   for (size_t I = 0; I < *BBRangeIndex; ++I)
1465     NumBBEntriesBeforeRange +=
1466         FunctionMap->getAddrMap().BBRanges[I].BBEntries.size();
1467   const auto &BBRange = FunctionMap->getAddrMap().BBRanges[*BBRangeIndex];
1468   for (size_t I = 0; I < BBRange.BBEntries.size(); ++I) {
1469     const BBAddrMap::BBEntry &BBEntry = BBRange.BBEntries[I];
1470     uint64_t BBAddress = BBEntry.Offset + BBRange.BaseAddress;
1471     if (BBAddress >= EndAddress)
1472       continue;
1473 
1474     std::string LabelString = ("BB" + Twine(BBEntry.ID)).str();
1475     Labels[BBAddress].push_back(
1476         {LabelString, FunctionMap->constructPGOLabelString(
1477                           NumBBEntriesBeforeRange + I, PrettyPGOAnalysisMap)});
1478   }
1479 }
1480 
1481 static void
collectLocalBranchTargets(ArrayRef<uint8_t> Bytes,MCInstrAnalysis * MIA,MCDisassembler * DisAsm,MCInstPrinter * IP,const MCSubtargetInfo * STI,uint64_t SectionAddr,uint64_t Start,uint64_t End,std::unordered_map<uint64_t,std::string> & Labels)1482 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, MCInstrAnalysis *MIA,
1483                           MCDisassembler *DisAsm, MCInstPrinter *IP,
1484                           const MCSubtargetInfo *STI, uint64_t SectionAddr,
1485                           uint64_t Start, uint64_t End,
1486                           std::unordered_map<uint64_t, std::string> &Labels) {
1487   // So far only supports PowerPC and X86.
1488   const bool isPPC = STI->getTargetTriple().isPPC();
1489   if (!isPPC && !STI->getTargetTriple().isX86())
1490     return;
1491 
1492   if (MIA)
1493     MIA->resetState();
1494 
1495   Labels.clear();
1496   unsigned LabelCount = 0;
1497   Start += SectionAddr;
1498   End += SectionAddr;
1499   const bool isXCOFF = STI->getTargetTriple().isOSBinFormatXCOFF();
1500   for (uint64_t Index = Start; Index < End;) {
1501     // Disassemble a real instruction and record function-local branch labels.
1502     MCInst Inst;
1503     uint64_t Size;
1504     ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index - SectionAddr);
1505     bool Disassembled =
1506         DisAsm->getInstruction(Inst, Size, ThisBytes, Index, nulls());
1507     if (Size == 0)
1508       Size = std::min<uint64_t>(ThisBytes.size(),
1509                                 DisAsm->suggestBytesToSkip(ThisBytes, Index));
1510 
1511     if (MIA) {
1512       if (Disassembled) {
1513         uint64_t Target;
1514         bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target);
1515         if (TargetKnown && (Target >= Start && Target < End) &&
1516             !Labels.count(Target)) {
1517           // On PowerPC and AIX, a function call is encoded as a branch to 0.
1518           // On other PowerPC platforms (ELF), a function call is encoded as
1519           // a branch to self. Do not add a label for these cases.
1520           if (!(isPPC &&
1521                 ((Target == 0 && isXCOFF) || (Target == Index && !isXCOFF))))
1522             Labels[Target] = ("L" + Twine(LabelCount++)).str();
1523         }
1524         MIA->updateState(Inst, Index);
1525       } else
1526         MIA->resetState();
1527     }
1528     Index += Size;
1529   }
1530 }
1531 
1532 // Create an MCSymbolizer for the target and add it to the MCDisassembler.
1533 // This is currently only used on AMDGPU, and assumes the format of the
1534 // void * argument passed to AMDGPU's createMCSymbolizer.
addSymbolizer(MCContext & Ctx,const Target * Target,StringRef TripleName,MCDisassembler * DisAsm,uint64_t SectionAddr,ArrayRef<uint8_t> Bytes,SectionSymbolsTy & Symbols,std::vector<std::unique_ptr<std::string>> & SynthesizedLabelNames)1535 static void addSymbolizer(
1536     MCContext &Ctx, const Target *Target, StringRef TripleName,
1537     MCDisassembler *DisAsm, uint64_t SectionAddr, ArrayRef<uint8_t> Bytes,
1538     SectionSymbolsTy &Symbols,
1539     std::vector<std::unique_ptr<std::string>> &SynthesizedLabelNames) {
1540 
1541   std::unique_ptr<MCRelocationInfo> RelInfo(
1542       Target->createMCRelocationInfo(TripleName, Ctx));
1543   if (!RelInfo)
1544     return;
1545   std::unique_ptr<MCSymbolizer> Symbolizer(Target->createMCSymbolizer(
1546       TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1547   MCSymbolizer *SymbolizerPtr = &*Symbolizer;
1548   DisAsm->setSymbolizer(std::move(Symbolizer));
1549 
1550   if (!SymbolizeOperands)
1551     return;
1552 
1553   // Synthesize labels referenced by branch instructions by
1554   // disassembling, discarding the output, and collecting the referenced
1555   // addresses from the symbolizer.
1556   for (size_t Index = 0; Index != Bytes.size();) {
1557     MCInst Inst;
1558     uint64_t Size;
1559     ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index);
1560     const uint64_t ThisAddr = SectionAddr + Index;
1561     DisAsm->getInstruction(Inst, Size, ThisBytes, ThisAddr, nulls());
1562     if (Size == 0)
1563       Size = std::min<uint64_t>(ThisBytes.size(),
1564                                 DisAsm->suggestBytesToSkip(ThisBytes, Index));
1565     Index += Size;
1566   }
1567   ArrayRef<uint64_t> LabelAddrsRef = SymbolizerPtr->getReferencedAddresses();
1568   // Copy and sort to remove duplicates.
1569   std::vector<uint64_t> LabelAddrs;
1570   LabelAddrs.insert(LabelAddrs.end(), LabelAddrsRef.begin(),
1571                     LabelAddrsRef.end());
1572   llvm::sort(LabelAddrs);
1573   LabelAddrs.resize(llvm::unique(LabelAddrs) - LabelAddrs.begin());
1574   // Add the labels.
1575   for (unsigned LabelNum = 0; LabelNum != LabelAddrs.size(); ++LabelNum) {
1576     auto Name = std::make_unique<std::string>();
1577     *Name = (Twine("L") + Twine(LabelNum)).str();
1578     SynthesizedLabelNames.push_back(std::move(Name));
1579     Symbols.push_back(SymbolInfoTy(
1580         LabelAddrs[LabelNum], *SynthesizedLabelNames.back(), ELF::STT_NOTYPE));
1581   }
1582   llvm::stable_sort(Symbols);
1583   // Recreate the symbolizer with the new symbols list.
1584   RelInfo.reset(Target->createMCRelocationInfo(TripleName, Ctx));
1585   Symbolizer.reset(Target->createMCSymbolizer(
1586       TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1587   DisAsm->setSymbolizer(std::move(Symbolizer));
1588 }
1589 
getSegmentName(const MachOObjectFile * MachO,const SectionRef & Section)1590 static StringRef getSegmentName(const MachOObjectFile *MachO,
1591                                 const SectionRef &Section) {
1592   if (MachO) {
1593     DataRefImpl DR = Section.getRawDataRefImpl();
1594     StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
1595     return SegmentName;
1596   }
1597   return "";
1598 }
1599 
emitPostInstructionInfo(formatted_raw_ostream & FOS,const MCAsmInfo & MAI,const MCSubtargetInfo & STI,StringRef Comments,LiveVariablePrinter & LVP)1600 static void emitPostInstructionInfo(formatted_raw_ostream &FOS,
1601                                     const MCAsmInfo &MAI,
1602                                     const MCSubtargetInfo &STI,
1603                                     StringRef Comments,
1604                                     LiveVariablePrinter &LVP) {
1605   do {
1606     if (!Comments.empty()) {
1607       // Emit a line of comments.
1608       StringRef Comment;
1609       std::tie(Comment, Comments) = Comments.split('\n');
1610       // MAI.getCommentColumn() assumes that instructions are printed at the
1611       // position of 8, while getInstStartColumn() returns the actual position.
1612       unsigned CommentColumn =
1613           MAI.getCommentColumn() - 8 + getInstStartColumn(STI);
1614       FOS.PadToColumn(CommentColumn);
1615       FOS << MAI.getCommentString() << ' ' << Comment;
1616     }
1617     LVP.printAfterInst(FOS);
1618     FOS << '\n';
1619   } while (!Comments.empty());
1620   FOS.flush();
1621 }
1622 
createFakeELFSections(ObjectFile & Obj)1623 static void createFakeELFSections(ObjectFile &Obj) {
1624   assert(Obj.isELF());
1625   if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(&Obj))
1626     Elf32LEObj->createFakeSections();
1627   else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(&Obj))
1628     Elf64LEObj->createFakeSections();
1629   else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(&Obj))
1630     Elf32BEObj->createFakeSections();
1631   else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(&Obj))
1632     Elf64BEObj->createFakeSections();
1633   else
1634     llvm_unreachable("Unsupported binary format");
1635 }
1636 
1637 // Tries to fetch a more complete version of the given object file using its
1638 // Build ID. Returns std::nullopt if nothing was found.
1639 static std::optional<OwningBinary<Binary>>
fetchBinaryByBuildID(const ObjectFile & Obj)1640 fetchBinaryByBuildID(const ObjectFile &Obj) {
1641   object::BuildIDRef BuildID = getBuildID(&Obj);
1642   if (BuildID.empty())
1643     return std::nullopt;
1644   std::optional<std::string> Path = BIDFetcher->fetch(BuildID);
1645   if (!Path)
1646     return std::nullopt;
1647   Expected<OwningBinary<Binary>> DebugBinary = createBinary(*Path);
1648   if (!DebugBinary) {
1649     reportWarning(toString(DebugBinary.takeError()), *Path);
1650     return std::nullopt;
1651   }
1652   return std::move(*DebugBinary);
1653 }
1654 
1655 static void
disassembleObject(ObjectFile & Obj,const ObjectFile & DbgObj,DisassemblerTarget & PrimaryTarget,std::optional<DisassemblerTarget> & SecondaryTarget,SourcePrinter & SP,bool InlineRelocs)1656 disassembleObject(ObjectFile &Obj, const ObjectFile &DbgObj,
1657                   DisassemblerTarget &PrimaryTarget,
1658                   std::optional<DisassemblerTarget> &SecondaryTarget,
1659                   SourcePrinter &SP, bool InlineRelocs) {
1660   DisassemblerTarget *DT = &PrimaryTarget;
1661   bool PrimaryIsThumb = false;
1662   SmallVector<std::pair<uint64_t, uint64_t>, 0> CHPECodeMap;
1663 
1664   if (SecondaryTarget) {
1665     if (isArmElf(Obj)) {
1666       PrimaryIsThumb =
1667           PrimaryTarget.SubtargetInfo->checkFeatures("+thumb-mode");
1668     } else if (const auto *COFFObj = dyn_cast<COFFObjectFile>(&Obj)) {
1669       const chpe_metadata *CHPEMetadata = COFFObj->getCHPEMetadata();
1670       if (CHPEMetadata && CHPEMetadata->CodeMapCount) {
1671         uintptr_t CodeMapInt;
1672         cantFail(COFFObj->getRvaPtr(CHPEMetadata->CodeMap, CodeMapInt));
1673         auto CodeMap = reinterpret_cast<const chpe_range_entry *>(CodeMapInt);
1674 
1675         for (uint32_t i = 0; i < CHPEMetadata->CodeMapCount; ++i) {
1676           if (CodeMap[i].getType() == chpe_range_type::Amd64 &&
1677               CodeMap[i].Length) {
1678             // Store x86_64 CHPE code ranges.
1679             uint64_t Start = CodeMap[i].getStart() + COFFObj->getImageBase();
1680             CHPECodeMap.emplace_back(Start, Start + CodeMap[i].Length);
1681           }
1682         }
1683         llvm::sort(CHPECodeMap);
1684       }
1685     }
1686   }
1687 
1688   std::map<SectionRef, std::vector<RelocationRef>> RelocMap;
1689   if (InlineRelocs || Obj.isXCOFF())
1690     RelocMap = getRelocsMap(Obj);
1691   bool Is64Bits = Obj.getBytesInAddress() > 4;
1692 
1693   // Create a mapping from virtual address to symbol name.  This is used to
1694   // pretty print the symbols while disassembling.
1695   std::map<SectionRef, SectionSymbolsTy> AllSymbols;
1696   std::map<SectionRef, SmallVector<MappingSymbolPair, 0>> AllMappingSymbols;
1697   SectionSymbolsTy AbsoluteSymbols;
1698   const StringRef FileName = Obj.getFileName();
1699   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(&Obj);
1700   for (const SymbolRef &Symbol : Obj.symbols()) {
1701     Expected<StringRef> NameOrErr = Symbol.getName();
1702     if (!NameOrErr) {
1703       reportWarning(toString(NameOrErr.takeError()), FileName);
1704       continue;
1705     }
1706     if (NameOrErr->empty() && !(Obj.isXCOFF() && SymbolDescription))
1707       continue;
1708 
1709     if (Obj.isELF() &&
1710         (cantFail(Symbol.getFlags()) & SymbolRef::SF_FormatSpecific)) {
1711       // Symbol is intended not to be displayed by default (STT_FILE,
1712       // STT_SECTION, or a mapping symbol). Ignore STT_SECTION symbols. We will
1713       // synthesize a section symbol if no symbol is defined at offset 0.
1714       //
1715       // For a mapping symbol, store it within both AllSymbols and
1716       // AllMappingSymbols. If --show-all-symbols is unspecified, its label will
1717       // not be printed in disassembly listing.
1718       if (getElfSymbolType(Obj, Symbol) != ELF::STT_SECTION &&
1719           hasMappingSymbols(Obj)) {
1720         section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1721         if (SecI != Obj.section_end()) {
1722           uint64_t SectionAddr = SecI->getAddress();
1723           uint64_t Address = cantFail(Symbol.getAddress());
1724           StringRef Name = *NameOrErr;
1725           if (Name.consume_front("$") && Name.size() &&
1726               strchr("adtx", Name[0])) {
1727             AllMappingSymbols[*SecI].emplace_back(Address - SectionAddr,
1728                                                   Name[0]);
1729             AllSymbols[*SecI].push_back(
1730                 createSymbolInfo(Obj, Symbol, /*MappingSymbol=*/true));
1731           }
1732         }
1733       }
1734       continue;
1735     }
1736 
1737     if (MachO) {
1738       // __mh_(execute|dylib|dylinker|bundle|preload|object)_header are special
1739       // symbols that support MachO header introspection. They do not bind to
1740       // code locations and are irrelevant for disassembly.
1741       if (NameOrErr->starts_with("__mh_") && NameOrErr->ends_with("_header"))
1742         continue;
1743       // Don't ask a Mach-O STAB symbol for its section unless you know that
1744       // STAB symbol's section field refers to a valid section index. Otherwise
1745       // the symbol may error trying to load a section that does not exist.
1746       DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1747       uint8_t NType = (MachO->is64Bit() ?
1748                        MachO->getSymbol64TableEntry(SymDRI).n_type:
1749                        MachO->getSymbolTableEntry(SymDRI).n_type);
1750       if (NType & MachO::N_STAB)
1751         continue;
1752     }
1753 
1754     section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1755     if (SecI != Obj.section_end())
1756       AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol));
1757     else
1758       AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol));
1759   }
1760 
1761   if (AllSymbols.empty() && Obj.isELF())
1762     addDynamicElfSymbols(cast<ELFObjectFileBase>(Obj), AllSymbols);
1763 
1764   if (Obj.isWasm())
1765     addMissingWasmCodeSymbols(cast<WasmObjectFile>(Obj), AllSymbols);
1766 
1767   if (Obj.isELF() && Obj.sections().empty())
1768     createFakeELFSections(Obj);
1769 
1770   BumpPtrAllocator A;
1771   StringSaver Saver(A);
1772   addPltEntries(Obj, AllSymbols, Saver);
1773 
1774   // Create a mapping from virtual address to section. An empty section can
1775   // cause more than one section at the same address. Sort such sections to be
1776   // before same-addressed non-empty sections so that symbol lookups prefer the
1777   // non-empty section.
1778   std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses;
1779   for (SectionRef Sec : Obj.sections())
1780     SectionAddresses.emplace_back(Sec.getAddress(), Sec);
1781   llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) {
1782     if (LHS.first != RHS.first)
1783       return LHS.first < RHS.first;
1784     return LHS.second.getSize() < RHS.second.getSize();
1785   });
1786 
1787   // Linked executables (.exe and .dll files) typically don't include a real
1788   // symbol table but they might contain an export table.
1789   if (const auto *COFFObj = dyn_cast<COFFObjectFile>(&Obj)) {
1790     for (const auto &ExportEntry : COFFObj->export_directories()) {
1791       StringRef Name;
1792       if (Error E = ExportEntry.getSymbolName(Name))
1793         reportError(std::move(E), Obj.getFileName());
1794       if (Name.empty())
1795         continue;
1796 
1797       uint32_t RVA;
1798       if (Error E = ExportEntry.getExportRVA(RVA))
1799         reportError(std::move(E), Obj.getFileName());
1800 
1801       uint64_t VA = COFFObj->getImageBase() + RVA;
1802       auto Sec = partition_point(
1803           SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) {
1804             return O.first <= VA;
1805           });
1806       if (Sec != SectionAddresses.begin()) {
1807         --Sec;
1808         AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE);
1809       } else
1810         AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE);
1811     }
1812   }
1813 
1814   // Sort all the symbols, this allows us to use a simple binary search to find
1815   // Multiple symbols can have the same address. Use a stable sort to stabilize
1816   // the output.
1817   StringSet<> FoundDisasmSymbolSet;
1818   for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols)
1819     llvm::stable_sort(SecSyms.second);
1820   llvm::stable_sort(AbsoluteSymbols);
1821 
1822   std::unique_ptr<DWARFContext> DICtx;
1823   LiveVariablePrinter LVP(*DT->Context->getRegisterInfo(), *DT->SubtargetInfo);
1824 
1825   if (DbgVariables != DVDisabled) {
1826     DICtx = DWARFContext::create(DbgObj);
1827     for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units())
1828       LVP.addCompileUnit(CU->getUnitDIE(false));
1829   }
1830 
1831   LLVM_DEBUG(LVP.dump());
1832 
1833   BBAddrMapInfo FullAddrMap;
1834   auto ReadBBAddrMap = [&](std::optional<unsigned> SectionIndex =
1835                                std::nullopt) {
1836     FullAddrMap.clear();
1837     if (const auto *Elf = dyn_cast<ELFObjectFileBase>(&Obj)) {
1838       std::vector<PGOAnalysisMap> PGOAnalyses;
1839       auto BBAddrMapsOrErr = Elf->readBBAddrMap(SectionIndex, &PGOAnalyses);
1840       if (!BBAddrMapsOrErr) {
1841         reportWarning(toString(BBAddrMapsOrErr.takeError()), Obj.getFileName());
1842         return;
1843       }
1844       for (auto &&[FunctionBBAddrMap, FunctionPGOAnalysis] :
1845            zip_equal(*std::move(BBAddrMapsOrErr), std::move(PGOAnalyses))) {
1846         FullAddrMap.AddFunctionEntry(std::move(FunctionBBAddrMap),
1847                                      std::move(FunctionPGOAnalysis));
1848       }
1849     }
1850   };
1851 
1852   // For non-relocatable objects, Read all LLVM_BB_ADDR_MAP sections into a
1853   // single mapping, since they don't have any conflicts.
1854   if (SymbolizeOperands && !Obj.isRelocatableObject())
1855     ReadBBAddrMap();
1856 
1857   std::optional<llvm::BTFParser> BTF;
1858   if (InlineRelocs && BTFParser::hasBTFSections(Obj)) {
1859     BTF.emplace();
1860     BTFParser::ParseOptions Opts = {};
1861     Opts.LoadTypes = true;
1862     Opts.LoadRelocs = true;
1863     if (Error E = BTF->parse(Obj, Opts))
1864       WithColor::defaultErrorHandler(std::move(E));
1865   }
1866 
1867   for (const SectionRef &Section : ToolSectionFilter(Obj)) {
1868     if (FilterSections.empty() && !DisassembleAll &&
1869         (!Section.isText() || Section.isVirtual()))
1870       continue;
1871 
1872     uint64_t SectionAddr = Section.getAddress();
1873     uint64_t SectSize = Section.getSize();
1874     if (!SectSize)
1875       continue;
1876 
1877     // For relocatable object files, read the LLVM_BB_ADDR_MAP section
1878     // corresponding to this section, if present.
1879     if (SymbolizeOperands && Obj.isRelocatableObject())
1880       ReadBBAddrMap(Section.getIndex());
1881 
1882     // Get the list of all the symbols in this section.
1883     SectionSymbolsTy &Symbols = AllSymbols[Section];
1884     auto &MappingSymbols = AllMappingSymbols[Section];
1885     llvm::sort(MappingSymbols);
1886 
1887     ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(
1888         unwrapOrError(Section.getContents(), Obj.getFileName()));
1889 
1890     std::vector<std::unique_ptr<std::string>> SynthesizedLabelNames;
1891     if (Obj.isELF() && Obj.getArch() == Triple::amdgcn) {
1892       // AMDGPU disassembler uses symbolizer for printing labels
1893       addSymbolizer(*DT->Context, DT->TheTarget, TripleName, DT->DisAsm.get(),
1894                     SectionAddr, Bytes, Symbols, SynthesizedLabelNames);
1895     }
1896 
1897     StringRef SegmentName = getSegmentName(MachO, Section);
1898     StringRef SectionName = unwrapOrError(Section.getName(), Obj.getFileName());
1899     // If the section has no symbol at the start, just insert a dummy one.
1900     // Without --show-all-symbols, also insert one if all symbols at the start
1901     // are mapping symbols.
1902     bool CreateDummy = Symbols.empty();
1903     if (!CreateDummy) {
1904       CreateDummy = true;
1905       for (auto &Sym : Symbols) {
1906         if (Sym.Addr != SectionAddr)
1907           break;
1908         if (!Sym.IsMappingSymbol || ShowAllSymbols)
1909           CreateDummy = false;
1910       }
1911     }
1912     if (CreateDummy) {
1913       SymbolInfoTy Sym = createDummySymbolInfo(
1914           Obj, SectionAddr, SectionName,
1915           Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT);
1916       if (Obj.isXCOFF())
1917         Symbols.insert(Symbols.begin(), Sym);
1918       else
1919         Symbols.insert(llvm::lower_bound(Symbols, Sym), Sym);
1920     }
1921 
1922     SmallString<40> Comments;
1923     raw_svector_ostream CommentStream(Comments);
1924 
1925     uint64_t VMAAdjustment = 0;
1926     if (shouldAdjustVA(Section))
1927       VMAAdjustment = AdjustVMA;
1928 
1929     // In executable and shared objects, r_offset holds a virtual address.
1930     // Subtract SectionAddr from the r_offset field of a relocation to get
1931     // the section offset.
1932     uint64_t RelAdjustment = Obj.isRelocatableObject() ? 0 : SectionAddr;
1933     uint64_t Size;
1934     uint64_t Index;
1935     bool PrintedSection = false;
1936     std::vector<RelocationRef> Rels = RelocMap[Section];
1937     std::vector<RelocationRef>::const_iterator RelCur = Rels.begin();
1938     std::vector<RelocationRef>::const_iterator RelEnd = Rels.end();
1939 
1940     // Loop over each chunk of code between two points where at least
1941     // one symbol is defined.
1942     for (size_t SI = 0, SE = Symbols.size(); SI != SE;) {
1943       // Advance SI past all the symbols starting at the same address,
1944       // and make an ArrayRef of them.
1945       unsigned FirstSI = SI;
1946       uint64_t Start = Symbols[SI].Addr;
1947       ArrayRef<SymbolInfoTy> SymbolsHere;
1948       while (SI != SE && Symbols[SI].Addr == Start)
1949         ++SI;
1950       SymbolsHere = ArrayRef<SymbolInfoTy>(&Symbols[FirstSI], SI - FirstSI);
1951 
1952       // Get the demangled names of all those symbols. We end up with a vector
1953       // of StringRef that holds the names we're going to use, and a vector of
1954       // std::string that stores the new strings returned by demangle(), if
1955       // any. If we don't call demangle() then that vector can stay empty.
1956       std::vector<StringRef> SymNamesHere;
1957       std::vector<std::string> DemangledSymNamesHere;
1958       if (Demangle) {
1959         // Fetch the demangled names and store them locally.
1960         for (const SymbolInfoTy &Symbol : SymbolsHere)
1961           DemangledSymNamesHere.push_back(demangle(Symbol.Name));
1962         // Now we've finished modifying that vector, it's safe to make
1963         // a vector of StringRefs pointing into it.
1964         SymNamesHere.insert(SymNamesHere.begin(), DemangledSymNamesHere.begin(),
1965                             DemangledSymNamesHere.end());
1966       } else {
1967         for (const SymbolInfoTy &Symbol : SymbolsHere)
1968           SymNamesHere.push_back(Symbol.Name);
1969       }
1970 
1971       // Distinguish ELF data from code symbols, which will be used later on to
1972       // decide whether to 'disassemble' this chunk as a data declaration via
1973       // dumpELFData(), or whether to treat it as code.
1974       //
1975       // If data _and_ code symbols are defined at the same address, the code
1976       // takes priority, on the grounds that disassembling code is our main
1977       // purpose here, and it would be a worse failure to _not_ interpret
1978       // something that _was_ meaningful as code than vice versa.
1979       //
1980       // Any ELF symbol type that is not clearly data will be regarded as code.
1981       // In particular, one of the uses of STT_NOTYPE is for branch targets
1982       // inside functions, for which STT_FUNC would be inaccurate.
1983       //
1984       // So here, we spot whether there's any non-data symbol present at all,
1985       // and only set the DisassembleAsELFData flag if there isn't. Also, we use
1986       // this distinction to inform the decision of which symbol to print at
1987       // the head of the section, so that if we're printing code, we print a
1988       // code-related symbol name to go with it.
1989       bool DisassembleAsELFData = false;
1990       size_t DisplaySymIndex = SymbolsHere.size() - 1;
1991       if (Obj.isELF() && !DisassembleAll && Section.isText()) {
1992         DisassembleAsELFData = true; // unless we find a code symbol below
1993 
1994         for (size_t i = 0; i < SymbolsHere.size(); ++i) {
1995           uint8_t SymTy = SymbolsHere[i].Type;
1996           if (SymTy != ELF::STT_OBJECT && SymTy != ELF::STT_COMMON) {
1997             DisassembleAsELFData = false;
1998             DisplaySymIndex = i;
1999           }
2000         }
2001       }
2002 
2003       // Decide which symbol(s) from this collection we're going to print.
2004       std::vector<bool> SymsToPrint(SymbolsHere.size(), false);
2005       // If the user has given the --disassemble-symbols option, then we must
2006       // display every symbol in that set, and no others.
2007       if (!DisasmSymbolSet.empty()) {
2008         bool FoundAny = false;
2009         for (size_t i = 0; i < SymbolsHere.size(); ++i) {
2010           if (DisasmSymbolSet.count(SymNamesHere[i])) {
2011             SymsToPrint[i] = true;
2012             FoundAny = true;
2013           }
2014         }
2015 
2016         // And if none of the symbols here is one that the user asked for, skip
2017         // disassembling this entire chunk of code.
2018         if (!FoundAny)
2019           continue;
2020       } else if (!SymbolsHere[DisplaySymIndex].IsMappingSymbol) {
2021         // Otherwise, print whichever symbol at this location is last in the
2022         // Symbols array, because that array is pre-sorted in a way intended to
2023         // correlate with priority of which symbol to display.
2024         SymsToPrint[DisplaySymIndex] = true;
2025       }
2026 
2027       // Now that we know we're disassembling this section, override the choice
2028       // of which symbols to display by printing _all_ of them at this address
2029       // if the user asked for all symbols.
2030       //
2031       // That way, '--show-all-symbols --disassemble-symbol=foo' will print
2032       // only the chunk of code headed by 'foo', but also show any other
2033       // symbols defined at that address, such as aliases for 'foo', or the ARM
2034       // mapping symbol preceding its code.
2035       if (ShowAllSymbols) {
2036         for (size_t i = 0; i < SymbolsHere.size(); ++i)
2037           SymsToPrint[i] = true;
2038       }
2039 
2040       if (Start < SectionAddr || StopAddress <= Start)
2041         continue;
2042 
2043       for (size_t i = 0; i < SymbolsHere.size(); ++i)
2044         FoundDisasmSymbolSet.insert(SymNamesHere[i]);
2045 
2046       // The end is the section end, the beginning of the next symbol, or
2047       // --stop-address.
2048       uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress);
2049       if (SI < SE)
2050         End = std::min(End, Symbols[SI].Addr);
2051       if (Start >= End || End <= StartAddress)
2052         continue;
2053       Start -= SectionAddr;
2054       End -= SectionAddr;
2055 
2056       if (!PrintedSection) {
2057         PrintedSection = true;
2058         outs() << "\nDisassembly of section ";
2059         if (!SegmentName.empty())
2060           outs() << SegmentName << ",";
2061         outs() << SectionName << ":\n";
2062       }
2063 
2064       bool PrintedLabel = false;
2065       for (size_t i = 0; i < SymbolsHere.size(); ++i) {
2066         if (!SymsToPrint[i])
2067           continue;
2068 
2069         const SymbolInfoTy &Symbol = SymbolsHere[i];
2070         const StringRef SymbolName = SymNamesHere[i];
2071 
2072         if (!PrintedLabel) {
2073           outs() << '\n';
2074           PrintedLabel = true;
2075         }
2076         if (LeadingAddr)
2077           outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ",
2078                            SectionAddr + Start + VMAAdjustment);
2079         if (Obj.isXCOFF() && SymbolDescription) {
2080           outs() << getXCOFFSymbolDescription(Symbol, SymbolName) << ":\n";
2081         } else
2082           outs() << '<' << SymbolName << ">:\n";
2083       }
2084 
2085       // Don't print raw contents of a virtual section. A virtual section
2086       // doesn't have any contents in the file.
2087       if (Section.isVirtual()) {
2088         outs() << "...\n";
2089         continue;
2090       }
2091 
2092       // See if any of the symbols defined at this location triggers target-
2093       // specific disassembly behavior, e.g. of special descriptors or function
2094       // prelude information.
2095       //
2096       // We stop this loop at the first symbol that triggers some kind of
2097       // interesting behavior (if any), on the assumption that if two symbols
2098       // defined at the same address trigger two conflicting symbol handlers,
2099       // the object file is probably confused anyway, and it would make even
2100       // less sense to present the output of _both_ handlers, because that
2101       // would describe the same data twice.
2102       for (size_t SHI = 0; SHI < SymbolsHere.size(); ++SHI) {
2103         SymbolInfoTy Symbol = SymbolsHere[SHI];
2104 
2105         Expected<bool> RespondedOrErr = DT->DisAsm->onSymbolStart(
2106             Symbol, Size, Bytes.slice(Start, End - Start), SectionAddr + Start);
2107 
2108         if (RespondedOrErr && !*RespondedOrErr) {
2109           // This symbol didn't trigger any interesting handling. Try the other
2110           // symbols defined at this address.
2111           continue;
2112         }
2113 
2114         // If onSymbolStart returned an Error, that means it identified some
2115         // kind of special data at this address, but wasn't able to disassemble
2116         // it meaningfully. So we fall back to printing the error out and
2117         // disassembling the failed region as bytes, assuming that the target
2118         // detected the failure before printing anything.
2119         if (!RespondedOrErr) {
2120           std::string ErrMsgStr = toString(RespondedOrErr.takeError());
2121           StringRef ErrMsg = ErrMsgStr;
2122           do {
2123             StringRef Line;
2124             std::tie(Line, ErrMsg) = ErrMsg.split('\n');
2125             outs() << DT->Context->getAsmInfo()->getCommentString()
2126                    << " error decoding " << SymNamesHere[SHI] << ": " << Line
2127                    << '\n';
2128           } while (!ErrMsg.empty());
2129 
2130           if (Size) {
2131             outs() << DT->Context->getAsmInfo()->getCommentString()
2132                    << " decoding failed region as bytes\n";
2133             for (uint64_t I = 0; I < Size; ++I)
2134               outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true)
2135                      << '\n';
2136           }
2137         }
2138 
2139         // Regardless of whether onSymbolStart returned an Error or true, 'Size'
2140         // will have been set to the amount of data covered by whatever prologue
2141         // the target identified. So we advance our own position to beyond that.
2142         // Sometimes that will be the entire distance to the next symbol, and
2143         // sometimes it will be just a prologue and we should start
2144         // disassembling instructions from where it left off.
2145         Start += Size;
2146         break;
2147       }
2148 
2149       Index = Start;
2150       if (SectionAddr < StartAddress)
2151         Index = std::max<uint64_t>(Index, StartAddress - SectionAddr);
2152 
2153       if (DisassembleAsELFData) {
2154         dumpELFData(SectionAddr, Index, End, Bytes);
2155         Index = End;
2156         continue;
2157       }
2158 
2159       // Skip relocations from symbols that are not dumped.
2160       for (; RelCur != RelEnd; ++RelCur) {
2161         uint64_t Offset = RelCur->getOffset() - RelAdjustment;
2162         if (Index <= Offset)
2163           break;
2164       }
2165 
2166       bool DumpARMELFData = false;
2167       bool DumpTracebackTableForXCOFFFunction =
2168           Obj.isXCOFF() && Section.isText() && TracebackTable &&
2169           Symbols[SI - 1].XCOFFSymInfo.StorageMappingClass &&
2170           (*Symbols[SI - 1].XCOFFSymInfo.StorageMappingClass == XCOFF::XMC_PR);
2171 
2172       formatted_raw_ostream FOS(outs());
2173 
2174       std::unordered_map<uint64_t, std::string> AllLabels;
2175       std::unordered_map<uint64_t, std::vector<BBAddrMapLabel>> BBAddrMapLabels;
2176       if (SymbolizeOperands) {
2177         collectLocalBranchTargets(Bytes, DT->InstrAnalysis.get(),
2178                                   DT->DisAsm.get(), DT->InstPrinter.get(),
2179                                   PrimaryTarget.SubtargetInfo.get(),
2180                                   SectionAddr, Index, End, AllLabels);
2181         collectBBAddrMapLabels(FullAddrMap, SectionAddr, Index, End,
2182                                BBAddrMapLabels);
2183       }
2184 
2185       if (DT->InstrAnalysis)
2186         DT->InstrAnalysis->resetState();
2187 
2188       while (Index < End) {
2189         uint64_t RelOffset;
2190 
2191         // ARM and AArch64 ELF binaries can interleave data and text in the
2192         // same section. We rely on the markers introduced to understand what
2193         // we need to dump. If the data marker is within a function, it is
2194         // denoted as a word/short etc.
2195         if (!MappingSymbols.empty()) {
2196           char Kind = getMappingSymbolKind(MappingSymbols, Index);
2197           DumpARMELFData = Kind == 'd';
2198           if (SecondaryTarget) {
2199             if (Kind == 'a') {
2200               DT = PrimaryIsThumb ? &*SecondaryTarget : &PrimaryTarget;
2201             } else if (Kind == 't') {
2202               DT = PrimaryIsThumb ? &PrimaryTarget : &*SecondaryTarget;
2203             }
2204           }
2205         } else if (!CHPECodeMap.empty()) {
2206           uint64_t Address = SectionAddr + Index;
2207           auto It = partition_point(
2208               CHPECodeMap,
2209               [Address](const std::pair<uint64_t, uint64_t> &Entry) {
2210                 return Entry.first <= Address;
2211               });
2212           if (It != CHPECodeMap.begin() && Address < (It - 1)->second) {
2213             DT = &*SecondaryTarget;
2214           } else {
2215             DT = &PrimaryTarget;
2216             // X64 disassembler range may have left Index unaligned, so
2217             // make sure that it's aligned when we switch back to ARM64
2218             // code.
2219             Index = llvm::alignTo(Index, 4);
2220             if (Index >= End)
2221               break;
2222           }
2223         }
2224 
2225         auto findRel = [&]() {
2226           while (RelCur != RelEnd) {
2227             RelOffset = RelCur->getOffset() - RelAdjustment;
2228             // If this relocation is hidden, skip it.
2229             if (getHidden(*RelCur) || SectionAddr + RelOffset < StartAddress) {
2230               ++RelCur;
2231               continue;
2232             }
2233 
2234             // Stop when RelCur's offset is past the disassembled
2235             // instruction/data.
2236             if (RelOffset >= Index + Size)
2237               return false;
2238             if (RelOffset >= Index)
2239               return true;
2240             ++RelCur;
2241           }
2242           return false;
2243         };
2244 
2245         if (DumpARMELFData) {
2246           Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes,
2247                                 MappingSymbols, *DT->SubtargetInfo, FOS);
2248         } else {
2249           // When -z or --disassemble-zeroes are given we always dissasemble
2250           // them. Otherwise we might want to skip zero bytes we see.
2251           if (!DisassembleZeroes) {
2252             uint64_t MaxOffset = End - Index;
2253             // For --reloc: print zero blocks patched by relocations, so that
2254             // relocations can be shown in the dump.
2255             if (InlineRelocs && RelCur != RelEnd)
2256               MaxOffset = std::min(RelCur->getOffset() - RelAdjustment - Index,
2257                                    MaxOffset);
2258 
2259             if (size_t N =
2260                     countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) {
2261               FOS << "\t\t..." << '\n';
2262               Index += N;
2263               continue;
2264             }
2265           }
2266 
2267           if (DumpTracebackTableForXCOFFFunction &&
2268               doesXCOFFTracebackTableBegin(Bytes.slice(Index, 4))) {
2269             dumpTracebackTable(Bytes.slice(Index),
2270                                SectionAddr + Index + VMAAdjustment, FOS,
2271                                SectionAddr + End + VMAAdjustment,
2272                                *DT->SubtargetInfo, cast<XCOFFObjectFile>(&Obj));
2273             Index = End;
2274             continue;
2275           }
2276 
2277           // Print local label if there's any.
2278           auto Iter1 = BBAddrMapLabels.find(SectionAddr + Index);
2279           if (Iter1 != BBAddrMapLabels.end()) {
2280             for (const auto &BBLabel : Iter1->second)
2281               FOS << "<" << BBLabel.BlockLabel << ">" << BBLabel.PGOAnalysis
2282                   << ":\n";
2283           } else {
2284             auto Iter2 = AllLabels.find(SectionAddr + Index);
2285             if (Iter2 != AllLabels.end())
2286               FOS << "<" << Iter2->second << ">:\n";
2287           }
2288 
2289           // Disassemble a real instruction or a data when disassemble all is
2290           // provided
2291           MCInst Inst;
2292           ArrayRef<uint8_t> ThisBytes = Bytes.slice(Index);
2293           uint64_t ThisAddr = SectionAddr + Index;
2294           bool Disassembled = DT->DisAsm->getInstruction(
2295               Inst, Size, ThisBytes, ThisAddr, CommentStream);
2296           if (Size == 0)
2297             Size = std::min<uint64_t>(
2298                 ThisBytes.size(),
2299                 DT->DisAsm->suggestBytesToSkip(ThisBytes, ThisAddr));
2300 
2301           LVP.update({Index, Section.getIndex()},
2302                      {Index + Size, Section.getIndex()}, Index + Size != End);
2303 
2304           DT->InstPrinter->setCommentStream(CommentStream);
2305 
2306           DT->Printer->printInst(
2307               *DT->InstPrinter, Disassembled ? &Inst : nullptr,
2308               Bytes.slice(Index, Size),
2309               {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS,
2310               "", *DT->SubtargetInfo, &SP, Obj.getFileName(), &Rels, LVP);
2311 
2312           DT->InstPrinter->setCommentStream(llvm::nulls());
2313 
2314           // If disassembly succeeds, we try to resolve the target address
2315           // (jump target or memory operand address) and print it to the
2316           // right of the instruction.
2317           //
2318           // Otherwise, we don't print anything else so that we avoid
2319           // analyzing invalid or incomplete instruction information.
2320           if (Disassembled && DT->InstrAnalysis) {
2321             llvm::raw_ostream *TargetOS = &FOS;
2322             uint64_t Target;
2323             bool PrintTarget = DT->InstrAnalysis->evaluateBranch(
2324                 Inst, SectionAddr + Index, Size, Target);
2325 
2326             if (!PrintTarget) {
2327               if (std::optional<uint64_t> MaybeTarget =
2328                       DT->InstrAnalysis->evaluateMemoryOperandAddress(
2329                           Inst, DT->SubtargetInfo.get(), SectionAddr + Index,
2330                           Size)) {
2331                 Target = *MaybeTarget;
2332                 PrintTarget = true;
2333                 // Do not print real address when symbolizing.
2334                 if (!SymbolizeOperands) {
2335                   // Memory operand addresses are printed as comments.
2336                   TargetOS = &CommentStream;
2337                   *TargetOS << "0x" << Twine::utohexstr(Target);
2338                 }
2339               }
2340             }
2341 
2342             if (PrintTarget) {
2343               // In a relocatable object, the target's section must reside in
2344               // the same section as the call instruction or it is accessed
2345               // through a relocation.
2346               //
2347               // In a non-relocatable object, the target may be in any section.
2348               // In that case, locate the section(s) containing the target
2349               // address and find the symbol in one of those, if possible.
2350               //
2351               // N.B. Except for XCOFF, we don't walk the relocations in the
2352               // relocatable case yet.
2353               std::vector<const SectionSymbolsTy *> TargetSectionSymbols;
2354               if (!Obj.isRelocatableObject()) {
2355                 auto It = llvm::partition_point(
2356                     SectionAddresses,
2357                     [=](const std::pair<uint64_t, SectionRef> &O) {
2358                       return O.first <= Target;
2359                     });
2360                 uint64_t TargetSecAddr = 0;
2361                 while (It != SectionAddresses.begin()) {
2362                   --It;
2363                   if (TargetSecAddr == 0)
2364                     TargetSecAddr = It->first;
2365                   if (It->first != TargetSecAddr)
2366                     break;
2367                   TargetSectionSymbols.push_back(&AllSymbols[It->second]);
2368                 }
2369               } else {
2370                 TargetSectionSymbols.push_back(&Symbols);
2371               }
2372               TargetSectionSymbols.push_back(&AbsoluteSymbols);
2373 
2374               // Find the last symbol in the first candidate section whose
2375               // offset is less than or equal to the target. If there are no
2376               // such symbols, try in the next section and so on, before finally
2377               // using the nearest preceding absolute symbol (if any), if there
2378               // are no other valid symbols.
2379               const SymbolInfoTy *TargetSym = nullptr;
2380               for (const SectionSymbolsTy *TargetSymbols :
2381                    TargetSectionSymbols) {
2382                 auto It = llvm::partition_point(
2383                     *TargetSymbols,
2384                     [=](const SymbolInfoTy &O) { return O.Addr <= Target; });
2385                 while (It != TargetSymbols->begin()) {
2386                   --It;
2387                   // Skip mapping symbols to avoid possible ambiguity as they
2388                   // do not allow uniquely identifying the target address.
2389                   if (!It->IsMappingSymbol) {
2390                     TargetSym = &*It;
2391                     break;
2392                   }
2393                 }
2394                 if (TargetSym)
2395                   break;
2396               }
2397 
2398               // Branch targets are printed just after the instructions.
2399               // Print the labels corresponding to the target if there's any.
2400               bool BBAddrMapLabelAvailable = BBAddrMapLabels.count(Target);
2401               bool LabelAvailable = AllLabels.count(Target);
2402 
2403               if (TargetSym != nullptr) {
2404                 uint64_t TargetAddress = TargetSym->Addr;
2405                 uint64_t Disp = Target - TargetAddress;
2406                 std::string TargetName = Demangle ? demangle(TargetSym->Name)
2407                                                   : TargetSym->Name.str();
2408                 bool RelFixedUp = false;
2409                 SmallString<32> Val;
2410 
2411                 *TargetOS << " <";
2412                 // On XCOFF, we use relocations, even without -r, so we
2413                 // can print the correct name for an extern function call.
2414                 if (Obj.isXCOFF() && findRel()) {
2415                   // Check for possible branch relocations and
2416                   // branches to fixup code.
2417                   bool BranchRelocationType = true;
2418                   XCOFF::RelocationType RelocType;
2419                   if (Obj.is64Bit()) {
2420                     const XCOFFRelocation64 *Reloc =
2421                         reinterpret_cast<XCOFFRelocation64 *>(
2422                             RelCur->getRawDataRefImpl().p);
2423                     RelFixedUp = Reloc->isFixupIndicated();
2424                     RelocType = Reloc->Type;
2425                   } else {
2426                     const XCOFFRelocation32 *Reloc =
2427                         reinterpret_cast<XCOFFRelocation32 *>(
2428                             RelCur->getRawDataRefImpl().p);
2429                     RelFixedUp = Reloc->isFixupIndicated();
2430                     RelocType = Reloc->Type;
2431                   }
2432                   BranchRelocationType =
2433                       RelocType == XCOFF::R_BA || RelocType == XCOFF::R_BR ||
2434                       RelocType == XCOFF::R_RBA || RelocType == XCOFF::R_RBR;
2435 
2436                   // If we have a valid relocation, try to print its
2437                   // corresponding symbol name. Multiple relocations on the
2438                   // same instruction are not handled.
2439                   // Branches to fixup code will have the RelFixedUp flag set in
2440                   // the RLD. For these instructions, we print the correct
2441                   // branch target, but print the referenced symbol as a
2442                   // comment.
2443                   if (Error E = getRelocationValueString(*RelCur, false, Val)) {
2444                     // If -r was used, this error will be printed later.
2445                     // Otherwise, we ignore the error and print what
2446                     // would have been printed without using relocations.
2447                     consumeError(std::move(E));
2448                     *TargetOS << TargetName;
2449                     RelFixedUp = false; // Suppress comment for RLD sym name
2450                   } else if (BranchRelocationType && !RelFixedUp)
2451                     *TargetOS << Val;
2452                   else
2453                     *TargetOS << TargetName;
2454                   if (Disp)
2455                     *TargetOS << "+0x" << Twine::utohexstr(Disp);
2456                 } else if (!Disp) {
2457                   *TargetOS << TargetName;
2458                 } else if (BBAddrMapLabelAvailable) {
2459                   *TargetOS << BBAddrMapLabels[Target].front().BlockLabel;
2460                 } else if (LabelAvailable) {
2461                   *TargetOS << AllLabels[Target];
2462                 } else {
2463                   // Always Print the binary symbol plus an offset if there's no
2464                   // local label corresponding to the target address.
2465                   *TargetOS << TargetName << "+0x" << Twine::utohexstr(Disp);
2466                 }
2467                 *TargetOS << ">";
2468                 if (RelFixedUp && !InlineRelocs) {
2469                   // We have fixup code for a relocation. We print the
2470                   // referenced symbol as a comment.
2471                   *TargetOS << "\t# " << Val;
2472                 }
2473 
2474               } else if (BBAddrMapLabelAvailable) {
2475                 *TargetOS << " <" << BBAddrMapLabels[Target].front().BlockLabel
2476                           << ">";
2477               } else if (LabelAvailable) {
2478                 *TargetOS << " <" << AllLabels[Target] << ">";
2479               }
2480               // By convention, each record in the comment stream should be
2481               // terminated.
2482               if (TargetOS == &CommentStream)
2483                 *TargetOS << "\n";
2484             }
2485 
2486             DT->InstrAnalysis->updateState(Inst, SectionAddr + Index);
2487           } else if (!Disassembled && DT->InstrAnalysis) {
2488             DT->InstrAnalysis->resetState();
2489           }
2490         }
2491 
2492         assert(DT->Context->getAsmInfo());
2493         emitPostInstructionInfo(FOS, *DT->Context->getAsmInfo(),
2494                                 *DT->SubtargetInfo, CommentStream.str(), LVP);
2495         Comments.clear();
2496 
2497         if (BTF)
2498           printBTFRelocation(FOS, *BTF, {Index, Section.getIndex()}, LVP);
2499 
2500         // Hexagon handles relocs in pretty printer
2501         if (InlineRelocs && Obj.getArch() != Triple::hexagon) {
2502           while (findRel()) {
2503             // When --adjust-vma is used, update the address printed.
2504             if (RelCur->getSymbol() != Obj.symbol_end()) {
2505               Expected<section_iterator> SymSI =
2506                   RelCur->getSymbol()->getSection();
2507               if (SymSI && *SymSI != Obj.section_end() &&
2508                   shouldAdjustVA(**SymSI))
2509                 RelOffset += AdjustVMA;
2510             }
2511 
2512             printRelocation(FOS, Obj.getFileName(), *RelCur,
2513                             SectionAddr + RelOffset, Is64Bits);
2514             LVP.printAfterOtherLine(FOS, true);
2515             ++RelCur;
2516           }
2517         }
2518 
2519         Index += Size;
2520       }
2521     }
2522   }
2523   StringSet<> MissingDisasmSymbolSet =
2524       set_difference(DisasmSymbolSet, FoundDisasmSymbolSet);
2525   for (StringRef Sym : MissingDisasmSymbolSet.keys())
2526     reportWarning("failed to disassemble missing symbol " + Sym, FileName);
2527 }
2528 
disassembleObject(ObjectFile * Obj,bool InlineRelocs)2529 static void disassembleObject(ObjectFile *Obj, bool InlineRelocs) {
2530   // If information useful for showing the disassembly is missing, try to find a
2531   // more complete binary and disassemble that instead.
2532   OwningBinary<Binary> FetchedBinary;
2533   if (Obj->symbols().empty()) {
2534     if (std::optional<OwningBinary<Binary>> FetchedBinaryOpt =
2535             fetchBinaryByBuildID(*Obj)) {
2536       if (auto *O = dyn_cast<ObjectFile>(FetchedBinaryOpt->getBinary())) {
2537         if (!O->symbols().empty() ||
2538             (!O->sections().empty() && Obj->sections().empty())) {
2539           FetchedBinary = std::move(*FetchedBinaryOpt);
2540           Obj = O;
2541         }
2542       }
2543     }
2544   }
2545 
2546   const Target *TheTarget = getTarget(Obj);
2547 
2548   // Package up features to be passed to target/subtarget
2549   Expected<SubtargetFeatures> FeaturesValue = Obj->getFeatures();
2550   if (!FeaturesValue)
2551     reportError(FeaturesValue.takeError(), Obj->getFileName());
2552   SubtargetFeatures Features = *FeaturesValue;
2553   if (!MAttrs.empty()) {
2554     for (unsigned I = 0; I != MAttrs.size(); ++I)
2555       Features.AddFeature(MAttrs[I]);
2556   } else if (MCPU.empty() && Obj->getArch() == llvm::Triple::aarch64) {
2557     Features.AddFeature("+all");
2558   }
2559 
2560   if (MCPU.empty())
2561     MCPU = Obj->tryGetCPUName().value_or("").str();
2562 
2563   if (isArmElf(*Obj)) {
2564     // When disassembling big-endian Arm ELF, the instruction endianness is
2565     // determined in a complex way. In relocatable objects, AAELF32 mandates
2566     // that instruction endianness matches the ELF file endianness; in
2567     // executable images, that's true unless the file header has the EF_ARM_BE8
2568     // flag, in which case instructions are little-endian regardless of data
2569     // endianness.
2570     //
2571     // We must set the big-endian-instructions SubtargetFeature to make the
2572     // disassembler read the instructions the right way round, and also tell
2573     // our own prettyprinter to retrieve the encodings the same way to print in
2574     // hex.
2575     const auto *Elf32BE = dyn_cast<ELF32BEObjectFile>(Obj);
2576 
2577     if (Elf32BE && (Elf32BE->isRelocatableObject() ||
2578                     !(Elf32BE->getPlatformFlags() & ELF::EF_ARM_BE8))) {
2579       Features.AddFeature("+big-endian-instructions");
2580       ARMPrettyPrinterInst.setInstructionEndianness(llvm::endianness::big);
2581     } else {
2582       ARMPrettyPrinterInst.setInstructionEndianness(llvm::endianness::little);
2583     }
2584   }
2585 
2586   DisassemblerTarget PrimaryTarget(TheTarget, *Obj, TripleName, MCPU, Features);
2587 
2588   // If we have an ARM object file, we need a second disassembler, because
2589   // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode.
2590   // We use mapping symbols to switch between the two assemblers, where
2591   // appropriate.
2592   std::optional<DisassemblerTarget> SecondaryTarget;
2593 
2594   if (isArmElf(*Obj)) {
2595     if (!PrimaryTarget.SubtargetInfo->checkFeatures("+mclass")) {
2596       if (PrimaryTarget.SubtargetInfo->checkFeatures("+thumb-mode"))
2597         Features.AddFeature("-thumb-mode");
2598       else
2599         Features.AddFeature("+thumb-mode");
2600       SecondaryTarget.emplace(PrimaryTarget, Features);
2601     }
2602   } else if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) {
2603     const chpe_metadata *CHPEMetadata = COFFObj->getCHPEMetadata();
2604     if (CHPEMetadata && CHPEMetadata->CodeMapCount) {
2605       // Set up x86_64 disassembler for ARM64EC binaries.
2606       Triple X64Triple(TripleName);
2607       X64Triple.setArch(Triple::ArchType::x86_64);
2608 
2609       std::string Error;
2610       const Target *X64Target =
2611           TargetRegistry::lookupTarget("", X64Triple, Error);
2612       if (X64Target) {
2613         SubtargetFeatures X64Features;
2614         SecondaryTarget.emplace(X64Target, *Obj, X64Triple.getTriple(), "",
2615                                 X64Features);
2616       } else {
2617         reportWarning(Error, Obj->getFileName());
2618       }
2619     }
2620   }
2621 
2622   const ObjectFile *DbgObj = Obj;
2623   if (!FetchedBinary.getBinary() && !Obj->hasDebugInfo()) {
2624     if (std::optional<OwningBinary<Binary>> DebugBinaryOpt =
2625             fetchBinaryByBuildID(*Obj)) {
2626       if (auto *FetchedObj =
2627               dyn_cast<const ObjectFile>(DebugBinaryOpt->getBinary())) {
2628         if (FetchedObj->hasDebugInfo()) {
2629           FetchedBinary = std::move(*DebugBinaryOpt);
2630           DbgObj = FetchedObj;
2631         }
2632       }
2633     }
2634   }
2635 
2636   std::unique_ptr<object::Binary> DSYMBinary;
2637   std::unique_ptr<MemoryBuffer> DSYMBuf;
2638   if (!DbgObj->hasDebugInfo()) {
2639     if (const MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*Obj)) {
2640       DbgObj = objdump::getMachODSymObject(MachOOF, Obj->getFileName(),
2641                                            DSYMBinary, DSYMBuf);
2642       if (!DbgObj)
2643         return;
2644     }
2645   }
2646 
2647   SourcePrinter SP(DbgObj, TheTarget->getName());
2648 
2649   for (StringRef Opt : DisassemblerOptions)
2650     if (!PrimaryTarget.InstPrinter->applyTargetSpecificCLOption(Opt))
2651       reportError(Obj->getFileName(),
2652                   "Unrecognized disassembler option: " + Opt);
2653 
2654   disassembleObject(*Obj, *DbgObj, PrimaryTarget, SecondaryTarget, SP,
2655                     InlineRelocs);
2656 }
2657 
printRelocations()2658 void Dumper::printRelocations() {
2659   StringRef Fmt = O.getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2660 
2661   // Build a mapping from relocation target to a vector of relocation
2662   // sections. Usually, there is an only one relocation section for
2663   // each relocated section.
2664   MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec;
2665   uint64_t Ndx;
2666   for (const SectionRef &Section : ToolSectionFilter(O, &Ndx)) {
2667     if (O.isELF() && (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC))
2668       continue;
2669     if (Section.relocation_begin() == Section.relocation_end())
2670       continue;
2671     Expected<section_iterator> SecOrErr = Section.getRelocatedSection();
2672     if (!SecOrErr)
2673       reportError(O.getFileName(),
2674                   "section (" + Twine(Ndx) +
2675                       "): unable to get a relocation target: " +
2676                       toString(SecOrErr.takeError()));
2677     SecToRelSec[**SecOrErr].push_back(Section);
2678   }
2679 
2680   for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) {
2681     StringRef SecName = unwrapOrError(P.first.getName(), O.getFileName());
2682     outs() << "\nRELOCATION RECORDS FOR [" << SecName << "]:\n";
2683     uint32_t OffsetPadding = (O.getBytesInAddress() > 4 ? 16 : 8);
2684     uint32_t TypePadding = 24;
2685     outs() << left_justify("OFFSET", OffsetPadding) << " "
2686            << left_justify("TYPE", TypePadding) << " "
2687            << "VALUE\n";
2688 
2689     for (SectionRef Section : P.second) {
2690       // CREL sections require decoding, each section may have its own specific
2691       // decode problems.
2692       if (O.isELF() && ELFSectionRef(Section).getType() == ELF::SHT_CREL) {
2693         StringRef Err =
2694             cast<const ELFObjectFileBase>(O).getCrelDecodeProblem(Section);
2695         if (!Err.empty()) {
2696           reportUniqueWarning(Err);
2697           continue;
2698         }
2699       }
2700       for (const RelocationRef &Reloc : Section.relocations()) {
2701         uint64_t Address = Reloc.getOffset();
2702         SmallString<32> RelocName;
2703         SmallString<32> ValueStr;
2704         if (Address < StartAddress || Address > StopAddress || getHidden(Reloc))
2705           continue;
2706         Reloc.getTypeName(RelocName);
2707         if (Error E =
2708                 getRelocationValueString(Reloc, SymbolDescription, ValueStr))
2709           reportUniqueWarning(std::move(E));
2710 
2711         outs() << format(Fmt.data(), Address) << " "
2712                << left_justify(RelocName, TypePadding) << " " << ValueStr
2713                << "\n";
2714       }
2715     }
2716   }
2717 }
2718 
2719 // Returns true if we need to show LMA column when dumping section headers. We
2720 // show it only when the platform is ELF and either we have at least one section
2721 // whose VMA and LMA are different and/or when --show-lma flag is used.
shouldDisplayLMA(const ObjectFile & Obj)2722 static bool shouldDisplayLMA(const ObjectFile &Obj) {
2723   if (!Obj.isELF())
2724     return false;
2725   for (const SectionRef &S : ToolSectionFilter(Obj))
2726     if (S.getAddress() != getELFSectionLMA(S))
2727       return true;
2728   return ShowLMA;
2729 }
2730 
getMaxSectionNameWidth(const ObjectFile & Obj)2731 static size_t getMaxSectionNameWidth(const ObjectFile &Obj) {
2732   // Default column width for names is 13 even if no names are that long.
2733   size_t MaxWidth = 13;
2734   for (const SectionRef &Section : ToolSectionFilter(Obj)) {
2735     StringRef Name = unwrapOrError(Section.getName(), Obj.getFileName());
2736     MaxWidth = std::max(MaxWidth, Name.size());
2737   }
2738   return MaxWidth;
2739 }
2740 
printSectionHeaders(ObjectFile & Obj)2741 void objdump::printSectionHeaders(ObjectFile &Obj) {
2742   if (Obj.isELF() && Obj.sections().empty())
2743     createFakeELFSections(Obj);
2744 
2745   size_t NameWidth = getMaxSectionNameWidth(Obj);
2746   size_t AddressWidth = 2 * Obj.getBytesInAddress();
2747   bool HasLMAColumn = shouldDisplayLMA(Obj);
2748   outs() << "\nSections:\n";
2749   if (HasLMAColumn)
2750     outs() << "Idx " << left_justify("Name", NameWidth) << " Size     "
2751            << left_justify("VMA", AddressWidth) << " "
2752            << left_justify("LMA", AddressWidth) << " Type\n";
2753   else
2754     outs() << "Idx " << left_justify("Name", NameWidth) << " Size     "
2755            << left_justify("VMA", AddressWidth) << " Type\n";
2756 
2757   uint64_t Idx;
2758   for (const SectionRef &Section : ToolSectionFilter(Obj, &Idx)) {
2759     StringRef Name = unwrapOrError(Section.getName(), Obj.getFileName());
2760     uint64_t VMA = Section.getAddress();
2761     if (shouldAdjustVA(Section))
2762       VMA += AdjustVMA;
2763 
2764     uint64_t Size = Section.getSize();
2765 
2766     std::string Type = Section.isText() ? "TEXT" : "";
2767     if (Section.isData())
2768       Type += Type.empty() ? "DATA" : ", DATA";
2769     if (Section.isBSS())
2770       Type += Type.empty() ? "BSS" : ", BSS";
2771     if (Section.isDebugSection())
2772       Type += Type.empty() ? "DEBUG" : ", DEBUG";
2773 
2774     if (HasLMAColumn)
2775       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
2776                        Name.str().c_str(), Size)
2777              << format_hex_no_prefix(VMA, AddressWidth) << " "
2778              << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth)
2779              << " " << Type << "\n";
2780     else
2781       outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
2782                        Name.str().c_str(), Size)
2783              << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n";
2784   }
2785 }
2786 
printSectionContents(const ObjectFile * Obj)2787 void objdump::printSectionContents(const ObjectFile *Obj) {
2788   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
2789 
2790   for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
2791     StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
2792     uint64_t BaseAddr = Section.getAddress();
2793     uint64_t Size = Section.getSize();
2794     if (!Size)
2795       continue;
2796 
2797     outs() << "Contents of section ";
2798     StringRef SegmentName = getSegmentName(MachO, Section);
2799     if (!SegmentName.empty())
2800       outs() << SegmentName << ",";
2801     outs() << Name << ":\n";
2802     if (Section.isBSS()) {
2803       outs() << format("<skipping contents of bss section at [%04" PRIx64
2804                        ", %04" PRIx64 ")>\n",
2805                        BaseAddr, BaseAddr + Size);
2806       continue;
2807     }
2808 
2809     StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName());
2810 
2811     // Dump out the content as hex and printable ascii characters.
2812     for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) {
2813       outs() << format(" %04" PRIx64 " ", BaseAddr + Addr);
2814       // Dump line of hex.
2815       for (std::size_t I = 0; I < 16; ++I) {
2816         if (I != 0 && I % 4 == 0)
2817           outs() << ' ';
2818         if (Addr + I < End)
2819           outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true)
2820                  << hexdigit(Contents[Addr + I] & 0xF, true);
2821         else
2822           outs() << "  ";
2823       }
2824       // Print ascii.
2825       outs() << "  ";
2826       for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) {
2827         if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF))
2828           outs() << Contents[Addr + I];
2829         else
2830           outs() << ".";
2831       }
2832       outs() << "\n";
2833     }
2834   }
2835 }
2836 
printSymbolTable(StringRef ArchiveName,StringRef ArchitectureName,bool DumpDynamic)2837 void Dumper::printSymbolTable(StringRef ArchiveName, StringRef ArchitectureName,
2838                               bool DumpDynamic) {
2839   if (O.isCOFF() && !DumpDynamic) {
2840     outs() << "\nSYMBOL TABLE:\n";
2841     printCOFFSymbolTable(cast<const COFFObjectFile>(O));
2842     return;
2843   }
2844 
2845   const StringRef FileName = O.getFileName();
2846 
2847   if (!DumpDynamic) {
2848     outs() << "\nSYMBOL TABLE:\n";
2849     for (auto I = O.symbol_begin(); I != O.symbol_end(); ++I)
2850       printSymbol(*I, {}, FileName, ArchiveName, ArchitectureName, DumpDynamic);
2851     return;
2852   }
2853 
2854   outs() << "\nDYNAMIC SYMBOL TABLE:\n";
2855   if (!O.isELF()) {
2856     reportWarning(
2857         "this operation is not currently supported for this file format",
2858         FileName);
2859     return;
2860   }
2861 
2862   const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(&O);
2863   auto Symbols = ELF->getDynamicSymbolIterators();
2864   Expected<std::vector<VersionEntry>> SymbolVersionsOrErr =
2865       ELF->readDynsymVersions();
2866   if (!SymbolVersionsOrErr) {
2867     reportWarning(toString(SymbolVersionsOrErr.takeError()), FileName);
2868     SymbolVersionsOrErr = std::vector<VersionEntry>();
2869     (void)!SymbolVersionsOrErr;
2870   }
2871   for (auto &Sym : Symbols)
2872     printSymbol(Sym, *SymbolVersionsOrErr, FileName, ArchiveName,
2873                 ArchitectureName, DumpDynamic);
2874 }
2875 
printSymbol(const SymbolRef & Symbol,ArrayRef<VersionEntry> SymbolVersions,StringRef FileName,StringRef ArchiveName,StringRef ArchitectureName,bool DumpDynamic)2876 void Dumper::printSymbol(const SymbolRef &Symbol,
2877                          ArrayRef<VersionEntry> SymbolVersions,
2878                          StringRef FileName, StringRef ArchiveName,
2879                          StringRef ArchitectureName, bool DumpDynamic) {
2880   const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(&O);
2881   Expected<uint64_t> AddrOrErr = Symbol.getAddress();
2882   if (!AddrOrErr) {
2883     reportUniqueWarning(AddrOrErr.takeError());
2884     return;
2885   }
2886   uint64_t Address = *AddrOrErr;
2887   section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
2888   if (SecI != O.section_end() && shouldAdjustVA(*SecI))
2889     Address += AdjustVMA;
2890   if ((Address < StartAddress) || (Address > StopAddress))
2891     return;
2892   SymbolRef::Type Type =
2893       unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName);
2894   uint32_t Flags =
2895       unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName);
2896 
2897   // Don't ask a Mach-O STAB symbol for its section unless you know that
2898   // STAB symbol's section field refers to a valid section index. Otherwise
2899   // the symbol may error trying to load a section that does not exist.
2900   bool IsSTAB = false;
2901   if (MachO) {
2902     DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
2903     uint8_t NType =
2904         (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type
2905                           : MachO->getSymbolTableEntry(SymDRI).n_type);
2906     if (NType & MachO::N_STAB)
2907       IsSTAB = true;
2908   }
2909   section_iterator Section = IsSTAB
2910                                  ? O.section_end()
2911                                  : unwrapOrError(Symbol.getSection(), FileName,
2912                                                  ArchiveName, ArchitectureName);
2913 
2914   StringRef Name;
2915   if (Type == SymbolRef::ST_Debug && Section != O.section_end()) {
2916     if (Expected<StringRef> NameOrErr = Section->getName())
2917       Name = *NameOrErr;
2918     else
2919       consumeError(NameOrErr.takeError());
2920 
2921   } else {
2922     Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName,
2923                          ArchitectureName);
2924   }
2925 
2926   bool Global = Flags & SymbolRef::SF_Global;
2927   bool Weak = Flags & SymbolRef::SF_Weak;
2928   bool Absolute = Flags & SymbolRef::SF_Absolute;
2929   bool Common = Flags & SymbolRef::SF_Common;
2930   bool Hidden = Flags & SymbolRef::SF_Hidden;
2931 
2932   char GlobLoc = ' ';
2933   if ((Section != O.section_end() || Absolute) && !Weak)
2934     GlobLoc = Global ? 'g' : 'l';
2935   char IFunc = ' ';
2936   if (O.isELF()) {
2937     if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC)
2938       IFunc = 'i';
2939     if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE)
2940       GlobLoc = 'u';
2941   }
2942 
2943   char Debug = ' ';
2944   if (DumpDynamic)
2945     Debug = 'D';
2946   else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
2947     Debug = 'd';
2948 
2949   char FileFunc = ' ';
2950   if (Type == SymbolRef::ST_File)
2951     FileFunc = 'f';
2952   else if (Type == SymbolRef::ST_Function)
2953     FileFunc = 'F';
2954   else if (Type == SymbolRef::ST_Data)
2955     FileFunc = 'O';
2956 
2957   const char *Fmt = O.getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2958 
2959   outs() << format(Fmt, Address) << " "
2960          << GlobLoc            // Local -> 'l', Global -> 'g', Neither -> ' '
2961          << (Weak ? 'w' : ' ') // Weak?
2962          << ' '                // Constructor. Not supported yet.
2963          << ' '                // Warning. Not supported yet.
2964          << IFunc              // Indirect reference to another symbol.
2965          << Debug              // Debugging (d) or dynamic (D) symbol.
2966          << FileFunc           // Name of function (F), file (f) or object (O).
2967          << ' ';
2968   if (Absolute) {
2969     outs() << "*ABS*";
2970   } else if (Common) {
2971     outs() << "*COM*";
2972   } else if (Section == O.section_end()) {
2973     if (O.isXCOFF()) {
2974       XCOFFSymbolRef XCOFFSym = cast<const XCOFFObjectFile>(O).toSymbolRef(
2975           Symbol.getRawDataRefImpl());
2976       if (XCOFF::N_DEBUG == XCOFFSym.getSectionNumber())
2977         outs() << "*DEBUG*";
2978       else
2979         outs() << "*UND*";
2980     } else
2981       outs() << "*UND*";
2982   } else {
2983     StringRef SegmentName = getSegmentName(MachO, *Section);
2984     if (!SegmentName.empty())
2985       outs() << SegmentName << ",";
2986     StringRef SectionName = unwrapOrError(Section->getName(), FileName);
2987     outs() << SectionName;
2988     if (O.isXCOFF()) {
2989       std::optional<SymbolRef> SymRef =
2990           getXCOFFSymbolContainingSymbolRef(cast<XCOFFObjectFile>(O), Symbol);
2991       if (SymRef) {
2992 
2993         Expected<StringRef> NameOrErr = SymRef->getName();
2994 
2995         if (NameOrErr) {
2996           outs() << " (csect:";
2997           std::string SymName =
2998               Demangle ? demangle(*NameOrErr) : NameOrErr->str();
2999 
3000           if (SymbolDescription)
3001             SymName = getXCOFFSymbolDescription(createSymbolInfo(O, *SymRef),
3002                                                 SymName);
3003 
3004           outs() << ' ' << SymName;
3005           outs() << ") ";
3006         } else
3007           reportWarning(toString(NameOrErr.takeError()), FileName);
3008       }
3009     }
3010   }
3011 
3012   if (Common)
3013     outs() << '\t' << format(Fmt, static_cast<uint64_t>(Symbol.getAlignment()));
3014   else if (O.isXCOFF())
3015     outs() << '\t'
3016            << format(Fmt, cast<XCOFFObjectFile>(O).getSymbolSize(
3017                               Symbol.getRawDataRefImpl()));
3018   else if (O.isELF())
3019     outs() << '\t' << format(Fmt, ELFSymbolRef(Symbol).getSize());
3020   else if (O.isWasm())
3021     outs() << '\t'
3022            << format(Fmt, static_cast<uint64_t>(
3023                               cast<WasmObjectFile>(O).getSymbolSize(Symbol)));
3024 
3025   if (O.isELF()) {
3026     if (!SymbolVersions.empty()) {
3027       const VersionEntry &Ver =
3028           SymbolVersions[Symbol.getRawDataRefImpl().d.b - 1];
3029       std::string Str;
3030       if (!Ver.Name.empty())
3031         Str = Ver.IsVerDef ? ' ' + Ver.Name : '(' + Ver.Name + ')';
3032       outs() << ' ' << left_justify(Str, 12);
3033     }
3034 
3035     uint8_t Other = ELFSymbolRef(Symbol).getOther();
3036     switch (Other) {
3037     case ELF::STV_DEFAULT:
3038       break;
3039     case ELF::STV_INTERNAL:
3040       outs() << " .internal";
3041       break;
3042     case ELF::STV_HIDDEN:
3043       outs() << " .hidden";
3044       break;
3045     case ELF::STV_PROTECTED:
3046       outs() << " .protected";
3047       break;
3048     default:
3049       outs() << format(" 0x%02x", Other);
3050       break;
3051     }
3052   } else if (Hidden) {
3053     outs() << " .hidden";
3054   }
3055 
3056   std::string SymName = Demangle ? demangle(Name) : Name.str();
3057   if (O.isXCOFF() && SymbolDescription)
3058     SymName = getXCOFFSymbolDescription(createSymbolInfo(O, Symbol), SymName);
3059 
3060   outs() << ' ' << SymName << '\n';
3061 }
3062 
printUnwindInfo(const ObjectFile * O)3063 static void printUnwindInfo(const ObjectFile *O) {
3064   outs() << "Unwind info:\n\n";
3065 
3066   if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O))
3067     printCOFFUnwindInfo(Coff);
3068   else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O))
3069     printMachOUnwindInfo(MachO);
3070   else
3071     // TODO: Extract DWARF dump tool to objdump.
3072     WithColor::error(errs(), ToolName)
3073         << "This operation is only currently supported "
3074            "for COFF and MachO object files.\n";
3075 }
3076 
3077 /// Dump the raw contents of the __clangast section so the output can be piped
3078 /// into llvm-bcanalyzer.
printRawClangAST(const ObjectFile * Obj)3079 static void printRawClangAST(const ObjectFile *Obj) {
3080   if (outs().is_displayed()) {
3081     WithColor::error(errs(), ToolName)
3082         << "The -raw-clang-ast option will dump the raw binary contents of "
3083            "the clang ast section.\n"
3084            "Please redirect the output to a file or another program such as "
3085            "llvm-bcanalyzer.\n";
3086     return;
3087   }
3088 
3089   StringRef ClangASTSectionName("__clangast");
3090   if (Obj->isCOFF()) {
3091     ClangASTSectionName = "clangast";
3092   }
3093 
3094   std::optional<object::SectionRef> ClangASTSection;
3095   for (auto Sec : ToolSectionFilter(*Obj)) {
3096     StringRef Name;
3097     if (Expected<StringRef> NameOrErr = Sec.getName())
3098       Name = *NameOrErr;
3099     else
3100       consumeError(NameOrErr.takeError());
3101 
3102     if (Name == ClangASTSectionName) {
3103       ClangASTSection = Sec;
3104       break;
3105     }
3106   }
3107   if (!ClangASTSection)
3108     return;
3109 
3110   StringRef ClangASTContents =
3111       unwrapOrError(ClangASTSection->getContents(), Obj->getFileName());
3112   outs().write(ClangASTContents.data(), ClangASTContents.size());
3113 }
3114 
printFaultMaps(const ObjectFile * Obj)3115 static void printFaultMaps(const ObjectFile *Obj) {
3116   StringRef FaultMapSectionName;
3117 
3118   if (Obj->isELF()) {
3119     FaultMapSectionName = ".llvm_faultmaps";
3120   } else if (Obj->isMachO()) {
3121     FaultMapSectionName = "__llvm_faultmaps";
3122   } else {
3123     WithColor::error(errs(), ToolName)
3124         << "This operation is only currently supported "
3125            "for ELF and Mach-O executable files.\n";
3126     return;
3127   }
3128 
3129   std::optional<object::SectionRef> FaultMapSection;
3130 
3131   for (auto Sec : ToolSectionFilter(*Obj)) {
3132     StringRef Name;
3133     if (Expected<StringRef> NameOrErr = Sec.getName())
3134       Name = *NameOrErr;
3135     else
3136       consumeError(NameOrErr.takeError());
3137 
3138     if (Name == FaultMapSectionName) {
3139       FaultMapSection = Sec;
3140       break;
3141     }
3142   }
3143 
3144   outs() << "FaultMap table:\n";
3145 
3146   if (!FaultMapSection) {
3147     outs() << "<not found>\n";
3148     return;
3149   }
3150 
3151   StringRef FaultMapContents =
3152       unwrapOrError(FaultMapSection->getContents(), Obj->getFileName());
3153   FaultMapParser FMP(FaultMapContents.bytes_begin(),
3154                      FaultMapContents.bytes_end());
3155 
3156   outs() << FMP;
3157 }
3158 
printPrivateHeaders()3159 void Dumper::printPrivateHeaders() {
3160   reportError(O.getFileName(), "Invalid/Unsupported object file format");
3161 }
3162 
printFileHeaders(const ObjectFile * O)3163 static void printFileHeaders(const ObjectFile *O) {
3164   if (!O->isELF() && !O->isCOFF() && !O->isXCOFF())
3165     reportError(O->getFileName(), "Invalid/Unsupported object file format");
3166 
3167   Triple::ArchType AT = O->getArch();
3168   outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n";
3169   uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName());
3170 
3171   StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
3172   outs() << "start address: "
3173          << "0x" << format(Fmt.data(), Address) << "\n";
3174 }
3175 
printArchiveChild(StringRef Filename,const Archive::Child & C)3176 static void printArchiveChild(StringRef Filename, const Archive::Child &C) {
3177   Expected<sys::fs::perms> ModeOrErr = C.getAccessMode();
3178   if (!ModeOrErr) {
3179     WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n";
3180     consumeError(ModeOrErr.takeError());
3181     return;
3182   }
3183   sys::fs::perms Mode = ModeOrErr.get();
3184   outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
3185   outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
3186   outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
3187   outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
3188   outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
3189   outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
3190   outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
3191   outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
3192   outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
3193 
3194   outs() << " ";
3195 
3196   outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename),
3197                    unwrapOrError(C.getGID(), Filename),
3198                    unwrapOrError(C.getRawSize(), Filename));
3199 
3200   StringRef RawLastModified = C.getRawLastModified();
3201   unsigned Seconds;
3202   if (RawLastModified.getAsInteger(10, Seconds))
3203     outs() << "(date: \"" << RawLastModified
3204            << "\" contains non-decimal chars) ";
3205   else {
3206     // Since ctime(3) returns a 26 character string of the form:
3207     // "Sun Sep 16 01:03:52 1973\n\0"
3208     // just print 24 characters.
3209     time_t t = Seconds;
3210     outs() << format("%.24s ", ctime(&t));
3211   }
3212 
3213   StringRef Name = "";
3214   Expected<StringRef> NameOrErr = C.getName();
3215   if (!NameOrErr) {
3216     consumeError(NameOrErr.takeError());
3217     Name = unwrapOrError(C.getRawName(), Filename);
3218   } else {
3219     Name = NameOrErr.get();
3220   }
3221   outs() << Name << "\n";
3222 }
3223 
3224 // For ELF only now.
shouldWarnForInvalidStartStopAddress(ObjectFile * Obj)3225 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) {
3226   if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) {
3227     if (Elf->getEType() != ELF::ET_REL)
3228       return true;
3229   }
3230   return false;
3231 }
3232 
checkForInvalidStartStopAddress(ObjectFile * Obj,uint64_t Start,uint64_t Stop)3233 static void checkForInvalidStartStopAddress(ObjectFile *Obj,
3234                                             uint64_t Start, uint64_t Stop) {
3235   if (!shouldWarnForInvalidStartStopAddress(Obj))
3236     return;
3237 
3238   for (const SectionRef &Section : Obj->sections())
3239     if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) {
3240       uint64_t BaseAddr = Section.getAddress();
3241       uint64_t Size = Section.getSize();
3242       if ((Start < BaseAddr + Size) && Stop > BaseAddr)
3243         return;
3244     }
3245 
3246   if (!HasStartAddressFlag)
3247     reportWarning("no section has address less than 0x" +
3248                       Twine::utohexstr(Stop) + " specified by --stop-address",
3249                   Obj->getFileName());
3250   else if (!HasStopAddressFlag)
3251     reportWarning("no section has address greater than or equal to 0x" +
3252                       Twine::utohexstr(Start) + " specified by --start-address",
3253                   Obj->getFileName());
3254   else
3255     reportWarning("no section overlaps the range [0x" +
3256                       Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) +
3257                       ") specified by --start-address/--stop-address",
3258                   Obj->getFileName());
3259 }
3260 
dumpObject(ObjectFile * O,const Archive * A=nullptr,const Archive::Child * C=nullptr)3261 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
3262                        const Archive::Child *C = nullptr) {
3263   Expected<std::unique_ptr<Dumper>> DumperOrErr = createDumper(*O);
3264   if (!DumperOrErr) {
3265     reportError(DumperOrErr.takeError(), O->getFileName(),
3266                 A ? A->getFileName() : "");
3267     return;
3268   }
3269   Dumper &D = **DumperOrErr;
3270 
3271   // Avoid other output when using a raw option.
3272   if (!RawClangAST) {
3273     outs() << '\n';
3274     if (A)
3275       outs() << A->getFileName() << "(" << O->getFileName() << ")";
3276     else
3277       outs() << O->getFileName();
3278     outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n";
3279   }
3280 
3281   if (HasStartAddressFlag || HasStopAddressFlag)
3282     checkForInvalidStartStopAddress(O, StartAddress, StopAddress);
3283 
3284   // TODO: Change print* free functions to Dumper member functions to utilitize
3285   // stateful functions like reportUniqueWarning.
3286 
3287   // Note: the order here matches GNU objdump for compatability.
3288   StringRef ArchiveName = A ? A->getFileName() : "";
3289   if (ArchiveHeaders && !MachOOpt && C)
3290     printArchiveChild(ArchiveName, *C);
3291   if (FileHeaders)
3292     printFileHeaders(O);
3293   if (PrivateHeaders || FirstPrivateHeader)
3294     D.printPrivateHeaders();
3295   if (SectionHeaders)
3296     printSectionHeaders(*O);
3297   if (SymbolTable)
3298     D.printSymbolTable(ArchiveName);
3299   if (DynamicSymbolTable)
3300     D.printSymbolTable(ArchiveName, /*ArchitectureName=*/"",
3301                        /*DumpDynamic=*/true);
3302   if (DwarfDumpType != DIDT_Null) {
3303     std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O);
3304     // Dump the complete DWARF structure.
3305     DIDumpOptions DumpOpts;
3306     DumpOpts.DumpType = DwarfDumpType;
3307     DICtx->dump(outs(), DumpOpts);
3308   }
3309   if (Relocations && !Disassemble)
3310     D.printRelocations();
3311   if (DynamicRelocations)
3312     D.printDynamicRelocations();
3313   if (SectionContents)
3314     printSectionContents(O);
3315   if (Disassemble)
3316     disassembleObject(O, Relocations);
3317   if (UnwindInfo)
3318     printUnwindInfo(O);
3319 
3320   // Mach-O specific options:
3321   if (ExportsTrie)
3322     printExportsTrie(O);
3323   if (Rebase)
3324     printRebaseTable(O);
3325   if (Bind)
3326     printBindTable(O);
3327   if (LazyBind)
3328     printLazyBindTable(O);
3329   if (WeakBind)
3330     printWeakBindTable(O);
3331 
3332   // Other special sections:
3333   if (RawClangAST)
3334     printRawClangAST(O);
3335   if (FaultMapSection)
3336     printFaultMaps(O);
3337   if (Offloading)
3338     dumpOffloadBinary(*O);
3339 }
3340 
dumpObject(const COFFImportFile * I,const Archive * A,const Archive::Child * C=nullptr)3341 static void dumpObject(const COFFImportFile *I, const Archive *A,
3342                        const Archive::Child *C = nullptr) {
3343   StringRef ArchiveName = A ? A->getFileName() : "";
3344 
3345   // Avoid other output when using a raw option.
3346   if (!RawClangAST)
3347     outs() << '\n'
3348            << ArchiveName << "(" << I->getFileName() << ")"
3349            << ":\tfile format COFF-import-file"
3350            << "\n\n";
3351 
3352   if (ArchiveHeaders && !MachOOpt && C)
3353     printArchiveChild(ArchiveName, *C);
3354   if (SymbolTable)
3355     printCOFFSymbolTable(*I);
3356 }
3357 
3358 /// Dump each object file in \a a;
dumpArchive(const Archive * A)3359 static void dumpArchive(const Archive *A) {
3360   Error Err = Error::success();
3361   unsigned I = -1;
3362   for (auto &C : A->children(Err)) {
3363     ++I;
3364     Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
3365     if (!ChildOrErr) {
3366       if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError()))
3367         reportError(std::move(E), getFileNameForError(C, I), A->getFileName());
3368       continue;
3369     }
3370     if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
3371       dumpObject(O, A, &C);
3372     else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get()))
3373       dumpObject(I, A, &C);
3374     else
3375       reportError(errorCodeToError(object_error::invalid_file_type),
3376                   A->getFileName());
3377   }
3378   if (Err)
3379     reportError(std::move(Err), A->getFileName());
3380 }
3381 
3382 /// Open file and figure out how to dump it.
dumpInput(StringRef file)3383 static void dumpInput(StringRef file) {
3384   // If we are using the Mach-O specific object file parser, then let it parse
3385   // the file and process the command line options.  So the -arch flags can
3386   // be used to select specific slices, etc.
3387   if (MachOOpt) {
3388     parseInputMachO(file);
3389     return;
3390   }
3391 
3392   // Attempt to open the binary.
3393   OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file);
3394   Binary &Binary = *OBinary.getBinary();
3395 
3396   if (Archive *A = dyn_cast<Archive>(&Binary))
3397     dumpArchive(A);
3398   else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary))
3399     dumpObject(O);
3400   else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary))
3401     parseInputMachO(UB);
3402   else if (OffloadBinary *OB = dyn_cast<OffloadBinary>(&Binary))
3403     dumpOffloadSections(*OB);
3404   else
3405     reportError(errorCodeToError(object_error::invalid_file_type), file);
3406 }
3407 
3408 template <typename T>
parseIntArg(const llvm::opt::InputArgList & InputArgs,int ID,T & Value)3409 static void parseIntArg(const llvm::opt::InputArgList &InputArgs, int ID,
3410                         T &Value) {
3411   if (const opt::Arg *A = InputArgs.getLastArg(ID)) {
3412     StringRef V(A->getValue());
3413     if (!llvm::to_integer(V, Value, 0)) {
3414       reportCmdLineError(A->getSpelling() +
3415                          ": expected a non-negative integer, but got '" + V +
3416                          "'");
3417     }
3418   }
3419 }
3420 
parseBuildIDArg(const opt::Arg * A)3421 static object::BuildID parseBuildIDArg(const opt::Arg *A) {
3422   StringRef V(A->getValue());
3423   object::BuildID BID = parseBuildID(V);
3424   if (BID.empty())
3425     reportCmdLineError(A->getSpelling() + ": expected a build ID, but got '" +
3426                        V + "'");
3427   return BID;
3428 }
3429 
invalidArgValue(const opt::Arg * A)3430 void objdump::invalidArgValue(const opt::Arg *A) {
3431   reportCmdLineError("'" + StringRef(A->getValue()) +
3432                      "' is not a valid value for '" + A->getSpelling() + "'");
3433 }
3434 
3435 static std::vector<std::string>
commaSeparatedValues(const llvm::opt::InputArgList & InputArgs,int ID)3436 commaSeparatedValues(const llvm::opt::InputArgList &InputArgs, int ID) {
3437   std::vector<std::string> Values;
3438   for (StringRef Value : InputArgs.getAllArgValues(ID)) {
3439     llvm::SmallVector<StringRef, 2> SplitValues;
3440     llvm::SplitString(Value, SplitValues, ",");
3441     for (StringRef SplitValue : SplitValues)
3442       Values.push_back(SplitValue.str());
3443   }
3444   return Values;
3445 }
3446 
parseOtoolOptions(const llvm::opt::InputArgList & InputArgs)3447 static void parseOtoolOptions(const llvm::opt::InputArgList &InputArgs) {
3448   MachOOpt = true;
3449   FullLeadingAddr = true;
3450   PrintImmHex = true;
3451 
3452   ArchName = InputArgs.getLastArgValue(OTOOL_arch).str();
3453   LinkOptHints = InputArgs.hasArg(OTOOL_C);
3454   if (InputArgs.hasArg(OTOOL_d))
3455     FilterSections.push_back("__DATA,__data");
3456   DylibId = InputArgs.hasArg(OTOOL_D);
3457   UniversalHeaders = InputArgs.hasArg(OTOOL_f);
3458   DataInCode = InputArgs.hasArg(OTOOL_G);
3459   FirstPrivateHeader = InputArgs.hasArg(OTOOL_h);
3460   IndirectSymbols = InputArgs.hasArg(OTOOL_I);
3461   ShowRawInsn = InputArgs.hasArg(OTOOL_j);
3462   PrivateHeaders = InputArgs.hasArg(OTOOL_l);
3463   DylibsUsed = InputArgs.hasArg(OTOOL_L);
3464   MCPU = InputArgs.getLastArgValue(OTOOL_mcpu_EQ).str();
3465   ObjcMetaData = InputArgs.hasArg(OTOOL_o);
3466   DisSymName = InputArgs.getLastArgValue(OTOOL_p).str();
3467   InfoPlist = InputArgs.hasArg(OTOOL_P);
3468   Relocations = InputArgs.hasArg(OTOOL_r);
3469   if (const Arg *A = InputArgs.getLastArg(OTOOL_s)) {
3470     auto Filter = (A->getValue(0) + StringRef(",") + A->getValue(1)).str();
3471     FilterSections.push_back(Filter);
3472   }
3473   if (InputArgs.hasArg(OTOOL_t))
3474     FilterSections.push_back("__TEXT,__text");
3475   Verbose = InputArgs.hasArg(OTOOL_v) || InputArgs.hasArg(OTOOL_V) ||
3476             InputArgs.hasArg(OTOOL_o);
3477   SymbolicOperands = InputArgs.hasArg(OTOOL_V);
3478   if (InputArgs.hasArg(OTOOL_x))
3479     FilterSections.push_back(",__text");
3480   LeadingAddr = LeadingHeaders = !InputArgs.hasArg(OTOOL_X);
3481 
3482   ChainedFixups = InputArgs.hasArg(OTOOL_chained_fixups);
3483   DyldInfo = InputArgs.hasArg(OTOOL_dyld_info);
3484 
3485   InputFilenames = InputArgs.getAllArgValues(OTOOL_INPUT);
3486   if (InputFilenames.empty())
3487     reportCmdLineError("no input file");
3488 
3489   for (const Arg *A : InputArgs) {
3490     const Option &O = A->getOption();
3491     if (O.getGroup().isValid() && O.getGroup().getID() == OTOOL_grp_obsolete) {
3492       reportCmdLineWarning(O.getPrefixedName() +
3493                            " is obsolete and not implemented");
3494     }
3495   }
3496 }
3497 
parseObjdumpOptions(const llvm::opt::InputArgList & InputArgs)3498 static void parseObjdumpOptions(const llvm::opt::InputArgList &InputArgs) {
3499   parseIntArg(InputArgs, OBJDUMP_adjust_vma_EQ, AdjustVMA);
3500   AllHeaders = InputArgs.hasArg(OBJDUMP_all_headers);
3501   ArchName = InputArgs.getLastArgValue(OBJDUMP_arch_name_EQ).str();
3502   ArchiveHeaders = InputArgs.hasArg(OBJDUMP_archive_headers);
3503   Demangle = InputArgs.hasArg(OBJDUMP_demangle);
3504   Disassemble = InputArgs.hasArg(OBJDUMP_disassemble);
3505   DisassembleAll = InputArgs.hasArg(OBJDUMP_disassemble_all);
3506   SymbolDescription = InputArgs.hasArg(OBJDUMP_symbol_description);
3507   TracebackTable = InputArgs.hasArg(OBJDUMP_traceback_table);
3508   DisassembleSymbols =
3509       commaSeparatedValues(InputArgs, OBJDUMP_disassemble_symbols_EQ);
3510   DisassembleZeroes = InputArgs.hasArg(OBJDUMP_disassemble_zeroes);
3511   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_dwarf_EQ)) {
3512     DwarfDumpType = StringSwitch<DIDumpType>(A->getValue())
3513                         .Case("frames", DIDT_DebugFrame)
3514                         .Default(DIDT_Null);
3515     if (DwarfDumpType == DIDT_Null)
3516       invalidArgValue(A);
3517   }
3518   DynamicRelocations = InputArgs.hasArg(OBJDUMP_dynamic_reloc);
3519   FaultMapSection = InputArgs.hasArg(OBJDUMP_fault_map_section);
3520   Offloading = InputArgs.hasArg(OBJDUMP_offloading);
3521   FileHeaders = InputArgs.hasArg(OBJDUMP_file_headers);
3522   SectionContents = InputArgs.hasArg(OBJDUMP_full_contents);
3523   PrintLines = InputArgs.hasArg(OBJDUMP_line_numbers);
3524   InputFilenames = InputArgs.getAllArgValues(OBJDUMP_INPUT);
3525   MachOOpt = InputArgs.hasArg(OBJDUMP_macho);
3526   MCPU = InputArgs.getLastArgValue(OBJDUMP_mcpu_EQ).str();
3527   MAttrs = commaSeparatedValues(InputArgs, OBJDUMP_mattr_EQ);
3528   ShowRawInsn = !InputArgs.hasArg(OBJDUMP_no_show_raw_insn);
3529   LeadingAddr = !InputArgs.hasArg(OBJDUMP_no_leading_addr);
3530   RawClangAST = InputArgs.hasArg(OBJDUMP_raw_clang_ast);
3531   Relocations = InputArgs.hasArg(OBJDUMP_reloc);
3532   PrintImmHex =
3533       InputArgs.hasFlag(OBJDUMP_print_imm_hex, OBJDUMP_no_print_imm_hex, true);
3534   PrivateHeaders = InputArgs.hasArg(OBJDUMP_private_headers);
3535   FilterSections = InputArgs.getAllArgValues(OBJDUMP_section_EQ);
3536   SectionHeaders = InputArgs.hasArg(OBJDUMP_section_headers);
3537   ShowAllSymbols = InputArgs.hasArg(OBJDUMP_show_all_symbols);
3538   ShowLMA = InputArgs.hasArg(OBJDUMP_show_lma);
3539   PrintSource = InputArgs.hasArg(OBJDUMP_source);
3540   parseIntArg(InputArgs, OBJDUMP_start_address_EQ, StartAddress);
3541   HasStartAddressFlag = InputArgs.hasArg(OBJDUMP_start_address_EQ);
3542   parseIntArg(InputArgs, OBJDUMP_stop_address_EQ, StopAddress);
3543   HasStopAddressFlag = InputArgs.hasArg(OBJDUMP_stop_address_EQ);
3544   SymbolTable = InputArgs.hasArg(OBJDUMP_syms);
3545   SymbolizeOperands = InputArgs.hasArg(OBJDUMP_symbolize_operands);
3546   PrettyPGOAnalysisMap = InputArgs.hasArg(OBJDUMP_pretty_pgo_analysis_map);
3547   if (PrettyPGOAnalysisMap && !SymbolizeOperands)
3548     reportCmdLineWarning("--symbolize-operands must be enabled for "
3549                          "--pretty-pgo-analysis-map to have an effect");
3550   DynamicSymbolTable = InputArgs.hasArg(OBJDUMP_dynamic_syms);
3551   TripleName = InputArgs.getLastArgValue(OBJDUMP_triple_EQ).str();
3552   UnwindInfo = InputArgs.hasArg(OBJDUMP_unwind_info);
3553   Wide = InputArgs.hasArg(OBJDUMP_wide);
3554   Prefix = InputArgs.getLastArgValue(OBJDUMP_prefix).str();
3555   parseIntArg(InputArgs, OBJDUMP_prefix_strip, PrefixStrip);
3556   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_debug_vars_EQ)) {
3557     DbgVariables = StringSwitch<DebugVarsFormat>(A->getValue())
3558                        .Case("ascii", DVASCII)
3559                        .Case("unicode", DVUnicode)
3560                        .Default(DVInvalid);
3561     if (DbgVariables == DVInvalid)
3562       invalidArgValue(A);
3563   }
3564   if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_disassembler_color_EQ)) {
3565     DisassemblyColor = StringSwitch<ColorOutput>(A->getValue())
3566                            .Case("on", ColorOutput::Enable)
3567                            .Case("off", ColorOutput::Disable)
3568                            .Case("terminal", ColorOutput::Auto)
3569                            .Default(ColorOutput::Invalid);
3570     if (DisassemblyColor == ColorOutput::Invalid)
3571       invalidArgValue(A);
3572   }
3573 
3574   parseIntArg(InputArgs, OBJDUMP_debug_vars_indent_EQ, DbgIndent);
3575 
3576   parseMachOOptions(InputArgs);
3577 
3578   // Parse -M (--disassembler-options) and deprecated
3579   // --x86-asm-syntax={att,intel}.
3580   //
3581   // Note, for x86, the asm dialect (AssemblerDialect) is initialized when the
3582   // MCAsmInfo is constructed. MCInstPrinter::applyTargetSpecificCLOption is
3583   // called too late. For now we have to use the internal cl::opt option.
3584   const char *AsmSyntax = nullptr;
3585   for (const auto *A : InputArgs.filtered(OBJDUMP_disassembler_options_EQ,
3586                                           OBJDUMP_x86_asm_syntax_att,
3587                                           OBJDUMP_x86_asm_syntax_intel)) {
3588     switch (A->getOption().getID()) {
3589     case OBJDUMP_x86_asm_syntax_att:
3590       AsmSyntax = "--x86-asm-syntax=att";
3591       continue;
3592     case OBJDUMP_x86_asm_syntax_intel:
3593       AsmSyntax = "--x86-asm-syntax=intel";
3594       continue;
3595     }
3596 
3597     SmallVector<StringRef, 2> Values;
3598     llvm::SplitString(A->getValue(), Values, ",");
3599     for (StringRef V : Values) {
3600       if (V == "att")
3601         AsmSyntax = "--x86-asm-syntax=att";
3602       else if (V == "intel")
3603         AsmSyntax = "--x86-asm-syntax=intel";
3604       else
3605         DisassemblerOptions.push_back(V.str());
3606     }
3607   }
3608   SmallVector<const char *> Args = {"llvm-objdump"};
3609   for (const opt::Arg *A : InputArgs.filtered(OBJDUMP_mllvm))
3610     Args.push_back(A->getValue());
3611   if (AsmSyntax)
3612     Args.push_back(AsmSyntax);
3613   if (Args.size() > 1)
3614     llvm::cl::ParseCommandLineOptions(Args.size(), Args.data());
3615 
3616   // Look up any provided build IDs, then append them to the input filenames.
3617   for (const opt::Arg *A : InputArgs.filtered(OBJDUMP_build_id)) {
3618     object::BuildID BuildID = parseBuildIDArg(A);
3619     std::optional<std::string> Path = BIDFetcher->fetch(BuildID);
3620     if (!Path) {
3621       reportCmdLineError(A->getSpelling() + ": could not find build ID '" +
3622                          A->getValue() + "'");
3623     }
3624     InputFilenames.push_back(std::move(*Path));
3625   }
3626 
3627   // objdump defaults to a.out if no filenames specified.
3628   if (InputFilenames.empty())
3629     InputFilenames.push_back("a.out");
3630 }
3631 
llvm_objdump_main(int argc,char ** argv,const llvm::ToolContext &)3632 int llvm_objdump_main(int argc, char **argv, const llvm::ToolContext &) {
3633   using namespace llvm;
3634 
3635   ToolName = argv[0];
3636   std::unique_ptr<CommonOptTable> T;
3637   OptSpecifier Unknown, HelpFlag, HelpHiddenFlag, VersionFlag;
3638 
3639   StringRef Stem = sys::path::stem(ToolName);
3640   auto Is = [=](StringRef Tool) {
3641     // We need to recognize the following filenames:
3642     //
3643     // llvm-objdump -> objdump
3644     // llvm-otool-10.exe -> otool
3645     // powerpc64-unknown-freebsd13-objdump -> objdump
3646     auto I = Stem.rfind_insensitive(Tool);
3647     return I != StringRef::npos &&
3648            (I + Tool.size() == Stem.size() || !isAlnum(Stem[I + Tool.size()]));
3649   };
3650   if (Is("otool")) {
3651     T = std::make_unique<OtoolOptTable>();
3652     Unknown = OTOOL_UNKNOWN;
3653     HelpFlag = OTOOL_help;
3654     HelpHiddenFlag = OTOOL_help_hidden;
3655     VersionFlag = OTOOL_version;
3656   } else {
3657     T = std::make_unique<ObjdumpOptTable>();
3658     Unknown = OBJDUMP_UNKNOWN;
3659     HelpFlag = OBJDUMP_help;
3660     HelpHiddenFlag = OBJDUMP_help_hidden;
3661     VersionFlag = OBJDUMP_version;
3662   }
3663 
3664   BumpPtrAllocator A;
3665   StringSaver Saver(A);
3666   opt::InputArgList InputArgs =
3667       T->parseArgs(argc, argv, Unknown, Saver,
3668                    [&](StringRef Msg) { reportCmdLineError(Msg); });
3669 
3670   if (InputArgs.size() == 0 || InputArgs.hasArg(HelpFlag)) {
3671     T->printHelp(ToolName);
3672     return 0;
3673   }
3674   if (InputArgs.hasArg(HelpHiddenFlag)) {
3675     T->printHelp(ToolName, /*ShowHidden=*/true);
3676     return 0;
3677   }
3678 
3679   // Initialize targets and assembly printers/parsers.
3680   InitializeAllTargetInfos();
3681   InitializeAllTargetMCs();
3682   InitializeAllDisassemblers();
3683 
3684   if (InputArgs.hasArg(VersionFlag)) {
3685     cl::PrintVersionMessage();
3686     if (!Is("otool")) {
3687       outs() << '\n';
3688       TargetRegistry::printRegisteredTargetsForVersion(outs());
3689     }
3690     return 0;
3691   }
3692 
3693   // Initialize debuginfod.
3694   const bool ShouldUseDebuginfodByDefault =
3695       InputArgs.hasArg(OBJDUMP_build_id) || canUseDebuginfod();
3696   std::vector<std::string> DebugFileDirectories =
3697       InputArgs.getAllArgValues(OBJDUMP_debug_file_directory);
3698   if (InputArgs.hasFlag(OBJDUMP_debuginfod, OBJDUMP_no_debuginfod,
3699                         ShouldUseDebuginfodByDefault)) {
3700     HTTPClient::initialize();
3701     BIDFetcher =
3702         std::make_unique<DebuginfodFetcher>(std::move(DebugFileDirectories));
3703   } else {
3704     BIDFetcher =
3705         std::make_unique<BuildIDFetcher>(std::move(DebugFileDirectories));
3706   }
3707 
3708   if (Is("otool"))
3709     parseOtoolOptions(InputArgs);
3710   else
3711     parseObjdumpOptions(InputArgs);
3712 
3713   if (StartAddress >= StopAddress)
3714     reportCmdLineError("start address should be less than stop address");
3715 
3716   // Removes trailing separators from prefix.
3717   while (!Prefix.empty() && sys::path::is_separator(Prefix.back()))
3718     Prefix.pop_back();
3719 
3720   if (AllHeaders)
3721     ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations =
3722         SectionHeaders = SymbolTable = true;
3723 
3724   if (DisassembleAll || PrintSource || PrintLines || TracebackTable ||
3725       !DisassembleSymbols.empty())
3726     Disassemble = true;
3727 
3728   if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null &&
3729       !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST &&
3730       !Relocations && !SectionHeaders && !SectionContents && !SymbolTable &&
3731       !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && !Offloading &&
3732       !(MachOOpt &&
3733         (Bind || DataInCode || ChainedFixups || DyldInfo || DylibId ||
3734          DylibsUsed || ExportsTrie || FirstPrivateHeader ||
3735          FunctionStartsType != FunctionStartsMode::None || IndirectSymbols ||
3736          InfoPlist || LazyBind || LinkOptHints || ObjcMetaData || Rebase ||
3737          Rpaths || UniversalHeaders || WeakBind || !FilterSections.empty()))) {
3738     T->printHelp(ToolName);
3739     return 2;
3740   }
3741 
3742   DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end());
3743 
3744   llvm::for_each(InputFilenames, dumpInput);
3745 
3746   warnOnNoMatchForSections();
3747 
3748   return EXIT_SUCCESS;
3749 }
3750