1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This program is a utility that works like binutils "objdump", that is, it
10 // dumps out a plethora of information about an object file depending on the
11 // flags.
12 //
13 // The flags and output of this program should be near identical to those of
14 // binutils objdump.
15 //
16 //===----------------------------------------------------------------------===//
17
18 #include "llvm-objdump.h"
19 #include "COFFDump.h"
20 #include "ELFDump.h"
21 #include "MachODump.h"
22 #include "ObjdumpOptID.h"
23 #include "SourcePrinter.h"
24 #include "WasmDump.h"
25 #include "XCOFFDump.h"
26 #include "llvm/ADT/IndexedMap.h"
27 #include "llvm/ADT/Optional.h"
28 #include "llvm/ADT/STLExtras.h"
29 #include "llvm/ADT/SetOperations.h"
30 #include "llvm/ADT/SmallSet.h"
31 #include "llvm/ADT/StringExtras.h"
32 #include "llvm/ADT/StringSet.h"
33 #include "llvm/ADT/Triple.h"
34 #include "llvm/ADT/Twine.h"
35 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
36 #include "llvm/DebugInfo/Symbolize/Symbolize.h"
37 #include "llvm/Demangle/Demangle.h"
38 #include "llvm/MC/MCAsmInfo.h"
39 #include "llvm/MC/MCContext.h"
40 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
41 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h"
42 #include "llvm/MC/MCInst.h"
43 #include "llvm/MC/MCInstPrinter.h"
44 #include "llvm/MC/MCInstrAnalysis.h"
45 #include "llvm/MC/MCInstrInfo.h"
46 #include "llvm/MC/MCObjectFileInfo.h"
47 #include "llvm/MC/MCRegisterInfo.h"
48 #include "llvm/MC/MCSubtargetInfo.h"
49 #include "llvm/MC/MCTargetOptions.h"
50 #include "llvm/Object/Archive.h"
51 #include "llvm/Object/COFF.h"
52 #include "llvm/Object/COFFImportFile.h"
53 #include "llvm/Object/ELFObjectFile.h"
54 #include "llvm/Object/FaultMapParser.h"
55 #include "llvm/Object/MachO.h"
56 #include "llvm/Object/MachOUniversal.h"
57 #include "llvm/Object/ObjectFile.h"
58 #include "llvm/Object/Wasm.h"
59 #include "llvm/Option/Arg.h"
60 #include "llvm/Option/ArgList.h"
61 #include "llvm/Option/Option.h"
62 #include "llvm/Support/Casting.h"
63 #include "llvm/Support/Debug.h"
64 #include "llvm/Support/Errc.h"
65 #include "llvm/Support/FileSystem.h"
66 #include "llvm/Support/Format.h"
67 #include "llvm/Support/FormatVariadic.h"
68 #include "llvm/Support/GraphWriter.h"
69 #include "llvm/Support/Host.h"
70 #include "llvm/Support/InitLLVM.h"
71 #include "llvm/Support/MemoryBuffer.h"
72 #include "llvm/Support/SourceMgr.h"
73 #include "llvm/Support/StringSaver.h"
74 #include "llvm/Support/TargetRegistry.h"
75 #include "llvm/Support/TargetSelect.h"
76 #include "llvm/Support/WithColor.h"
77 #include "llvm/Support/raw_ostream.h"
78 #include <algorithm>
79 #include <cctype>
80 #include <cstring>
81 #include <system_error>
82 #include <unordered_map>
83 #include <utility>
84
85 using namespace llvm;
86 using namespace llvm::object;
87 using namespace llvm::objdump;
88 using namespace llvm::opt;
89
90 namespace {
91
92 class CommonOptTable : public opt::OptTable {
93 public:
CommonOptTable(ArrayRef<Info> OptionInfos,const char * Usage,const char * Description)94 CommonOptTable(ArrayRef<Info> OptionInfos, const char *Usage,
95 const char *Description)
96 : OptTable(OptionInfos), Usage(Usage), Description(Description) {
97 setGroupedShortOptions(true);
98 }
99
printHelp(StringRef Argv0,bool ShowHidden=false) const100 void printHelp(StringRef Argv0, bool ShowHidden = false) const {
101 Argv0 = sys::path::filename(Argv0);
102 PrintHelp(outs(), (Argv0 + Usage).str().c_str(), Description, ShowHidden,
103 ShowHidden);
104 // TODO Replace this with OptTable API once it adds extrahelp support.
105 outs() << "\nPass @FILE as argument to read options from FILE.\n";
106 }
107
108 private:
109 const char *Usage;
110 const char *Description;
111 };
112
113 // ObjdumpOptID is in ObjdumpOptID.h
114
115 #define PREFIX(NAME, VALUE) const char *const OBJDUMP_##NAME[] = VALUE;
116 #include "ObjdumpOpts.inc"
117 #undef PREFIX
118
119 static constexpr opt::OptTable::Info ObjdumpInfoTable[] = {
120 #define OBJDUMP_nullptr nullptr
121 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM, \
122 HELPTEXT, METAVAR, VALUES) \
123 {OBJDUMP_##PREFIX, NAME, HELPTEXT, \
124 METAVAR, OBJDUMP_##ID, opt::Option::KIND##Class, \
125 PARAM, FLAGS, OBJDUMP_##GROUP, \
126 OBJDUMP_##ALIAS, ALIASARGS, VALUES},
127 #include "ObjdumpOpts.inc"
128 #undef OPTION
129 #undef OBJDUMP_nullptr
130 };
131
132 class ObjdumpOptTable : public CommonOptTable {
133 public:
ObjdumpOptTable()134 ObjdumpOptTable()
135 : CommonOptTable(ObjdumpInfoTable, " [options] <input object files>",
136 "llvm object file dumper") {}
137 };
138
139 enum OtoolOptID {
140 OTOOL_INVALID = 0, // This is not an option ID.
141 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM, \
142 HELPTEXT, METAVAR, VALUES) \
143 OTOOL_##ID,
144 #include "OtoolOpts.inc"
145 #undef OPTION
146 };
147
148 #define PREFIX(NAME, VALUE) const char *const OTOOL_##NAME[] = VALUE;
149 #include "OtoolOpts.inc"
150 #undef PREFIX
151
152 static constexpr opt::OptTable::Info OtoolInfoTable[] = {
153 #define OTOOL_nullptr nullptr
154 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM, \
155 HELPTEXT, METAVAR, VALUES) \
156 {OTOOL_##PREFIX, NAME, HELPTEXT, \
157 METAVAR, OTOOL_##ID, opt::Option::KIND##Class, \
158 PARAM, FLAGS, OTOOL_##GROUP, \
159 OTOOL_##ALIAS, ALIASARGS, VALUES},
160 #include "OtoolOpts.inc"
161 #undef OPTION
162 #undef OTOOL_nullptr
163 };
164
165 class OtoolOptTable : public CommonOptTable {
166 public:
OtoolOptTable()167 OtoolOptTable()
168 : CommonOptTable(OtoolInfoTable, " [option...] [file...]",
169 "Mach-O object file displaying tool") {}
170 };
171
172 } // namespace
173
174 #define DEBUG_TYPE "objdump"
175
176 static uint64_t AdjustVMA;
177 static bool AllHeaders;
178 static std::string ArchName;
179 bool objdump::ArchiveHeaders;
180 bool objdump::Demangle;
181 bool objdump::Disassemble;
182 bool objdump::DisassembleAll;
183 bool objdump::SymbolDescription;
184 static std::vector<std::string> DisassembleSymbols;
185 static bool DisassembleZeroes;
186 static std::vector<std::string> DisassemblerOptions;
187 DIDumpType objdump::DwarfDumpType;
188 static bool DynamicRelocations;
189 static bool FaultMapSection;
190 static bool FileHeaders;
191 bool objdump::SectionContents;
192 static std::vector<std::string> InputFilenames;
193 bool objdump::PrintLines;
194 static bool MachOOpt;
195 std::string objdump::MCPU;
196 std::vector<std::string> objdump::MAttrs;
197 bool objdump::ShowRawInsn;
198 bool objdump::LeadingAddr;
199 static bool RawClangAST;
200 bool objdump::Relocations;
201 bool objdump::PrintImmHex;
202 bool objdump::PrivateHeaders;
203 std::vector<std::string> objdump::FilterSections;
204 bool objdump::SectionHeaders;
205 static bool ShowLMA;
206 bool objdump::PrintSource;
207
208 static uint64_t StartAddress;
209 static bool HasStartAddressFlag;
210 static uint64_t StopAddress = UINT64_MAX;
211 static bool HasStopAddressFlag;
212
213 bool objdump::SymbolTable;
214 static bool SymbolizeOperands;
215 static bool DynamicSymbolTable;
216 std::string objdump::TripleName;
217 bool objdump::UnwindInfo;
218 static bool Wide;
219 std::string objdump::Prefix;
220 uint32_t objdump::PrefixStrip;
221
222 DebugVarsFormat objdump::DbgVariables = DVDisabled;
223
224 int objdump::DbgIndent = 40;
225
226 static StringSet<> DisasmSymbolSet;
227 StringSet<> objdump::FoundSectionSet;
228 static StringRef ToolName;
229
230 namespace {
231 struct FilterResult {
232 // True if the section should not be skipped.
233 bool Keep;
234
235 // True if the index counter should be incremented, even if the section should
236 // be skipped. For example, sections may be skipped if they are not included
237 // in the --section flag, but we still want those to count toward the section
238 // count.
239 bool IncrementIndex;
240 };
241 } // namespace
242
checkSectionFilter(object::SectionRef S)243 static FilterResult checkSectionFilter(object::SectionRef S) {
244 if (FilterSections.empty())
245 return {/*Keep=*/true, /*IncrementIndex=*/true};
246
247 Expected<StringRef> SecNameOrErr = S.getName();
248 if (!SecNameOrErr) {
249 consumeError(SecNameOrErr.takeError());
250 return {/*Keep=*/false, /*IncrementIndex=*/false};
251 }
252 StringRef SecName = *SecNameOrErr;
253
254 // StringSet does not allow empty key so avoid adding sections with
255 // no name (such as the section with index 0) here.
256 if (!SecName.empty())
257 FoundSectionSet.insert(SecName);
258
259 // Only show the section if it's in the FilterSections list, but always
260 // increment so the indexing is stable.
261 return {/*Keep=*/is_contained(FilterSections, SecName),
262 /*IncrementIndex=*/true};
263 }
264
ToolSectionFilter(object::ObjectFile const & O,uint64_t * Idx)265 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O,
266 uint64_t *Idx) {
267 // Start at UINT64_MAX so that the first index returned after an increment is
268 // zero (after the unsigned wrap).
269 if (Idx)
270 *Idx = UINT64_MAX;
271 return SectionFilter(
272 [Idx](object::SectionRef S) {
273 FilterResult Result = checkSectionFilter(S);
274 if (Idx != nullptr && Result.IncrementIndex)
275 *Idx += 1;
276 return Result.Keep;
277 },
278 O);
279 }
280
getFileNameForError(const object::Archive::Child & C,unsigned Index)281 std::string objdump::getFileNameForError(const object::Archive::Child &C,
282 unsigned Index) {
283 Expected<StringRef> NameOrErr = C.getName();
284 if (NameOrErr)
285 return std::string(NameOrErr.get());
286 // If we have an error getting the name then we print the index of the archive
287 // member. Since we are already in an error state, we just ignore this error.
288 consumeError(NameOrErr.takeError());
289 return "<file index: " + std::to_string(Index) + ">";
290 }
291
reportWarning(const Twine & Message,StringRef File)292 void objdump::reportWarning(const Twine &Message, StringRef File) {
293 // Output order between errs() and outs() matters especially for archive
294 // files where the output is per member object.
295 outs().flush();
296 WithColor::warning(errs(), ToolName)
297 << "'" << File << "': " << Message << "\n";
298 }
299
reportError(StringRef File,const Twine & Message)300 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(StringRef File,
301 const Twine &Message) {
302 outs().flush();
303 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n";
304 exit(1);
305 }
306
reportError(Error E,StringRef FileName,StringRef ArchiveName,StringRef ArchitectureName)307 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(Error E, StringRef FileName,
308 StringRef ArchiveName,
309 StringRef ArchitectureName) {
310 assert(E);
311 outs().flush();
312 WithColor::error(errs(), ToolName);
313 if (ArchiveName != "")
314 errs() << ArchiveName << "(" << FileName << ")";
315 else
316 errs() << "'" << FileName << "'";
317 if (!ArchitectureName.empty())
318 errs() << " (for architecture " << ArchitectureName << ")";
319 errs() << ": ";
320 logAllUnhandledErrors(std::move(E), errs());
321 exit(1);
322 }
323
reportCmdLineWarning(const Twine & Message)324 static void reportCmdLineWarning(const Twine &Message) {
325 WithColor::warning(errs(), ToolName) << Message << "\n";
326 }
327
reportCmdLineError(const Twine & Message)328 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(const Twine &Message) {
329 WithColor::error(errs(), ToolName) << Message << "\n";
330 exit(1);
331 }
332
warnOnNoMatchForSections()333 static void warnOnNoMatchForSections() {
334 SetVector<StringRef> MissingSections;
335 for (StringRef S : FilterSections) {
336 if (FoundSectionSet.count(S))
337 return;
338 // User may specify a unnamed section. Don't warn for it.
339 if (!S.empty())
340 MissingSections.insert(S);
341 }
342
343 // Warn only if no section in FilterSections is matched.
344 for (StringRef S : MissingSections)
345 reportCmdLineWarning("section '" + S +
346 "' mentioned in a -j/--section option, but not "
347 "found in any input file");
348 }
349
getTarget(const ObjectFile * Obj)350 static const Target *getTarget(const ObjectFile *Obj) {
351 // Figure out the target triple.
352 Triple TheTriple("unknown-unknown-unknown");
353 if (TripleName.empty()) {
354 TheTriple = Obj->makeTriple();
355 } else {
356 TheTriple.setTriple(Triple::normalize(TripleName));
357 auto Arch = Obj->getArch();
358 if (Arch == Triple::arm || Arch == Triple::armeb)
359 Obj->setARMSubArch(TheTriple);
360 }
361
362 // Get the target specific parser.
363 std::string Error;
364 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple,
365 Error);
366 if (!TheTarget)
367 reportError(Obj->getFileName(), "can't find target: " + Error);
368
369 // Update the triple name and return the found target.
370 TripleName = TheTriple.getTriple();
371 return TheTarget;
372 }
373
isRelocAddressLess(RelocationRef A,RelocationRef B)374 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) {
375 return A.getOffset() < B.getOffset();
376 }
377
getRelocationValueString(const RelocationRef & Rel,SmallVectorImpl<char> & Result)378 static Error getRelocationValueString(const RelocationRef &Rel,
379 SmallVectorImpl<char> &Result) {
380 const ObjectFile *Obj = Rel.getObject();
381 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj))
382 return getELFRelocationValueString(ELF, Rel, Result);
383 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj))
384 return getCOFFRelocationValueString(COFF, Rel, Result);
385 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj))
386 return getWasmRelocationValueString(Wasm, Rel, Result);
387 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj))
388 return getMachORelocationValueString(MachO, Rel, Result);
389 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj))
390 return getXCOFFRelocationValueString(XCOFF, Rel, Result);
391 llvm_unreachable("unknown object file format");
392 }
393
394 /// Indicates whether this relocation should hidden when listing
395 /// relocations, usually because it is the trailing part of a multipart
396 /// relocation that will be printed as part of the leading relocation.
getHidden(RelocationRef RelRef)397 static bool getHidden(RelocationRef RelRef) {
398 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject());
399 if (!MachO)
400 return false;
401
402 unsigned Arch = MachO->getArch();
403 DataRefImpl Rel = RelRef.getRawDataRefImpl();
404 uint64_t Type = MachO->getRelocationType(Rel);
405
406 // On arches that use the generic relocations, GENERIC_RELOC_PAIR
407 // is always hidden.
408 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc)
409 return Type == MachO::GENERIC_RELOC_PAIR;
410
411 if (Arch == Triple::x86_64) {
412 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows
413 // an X86_64_RELOC_SUBTRACTOR.
414 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) {
415 DataRefImpl RelPrev = Rel;
416 RelPrev.d.a--;
417 uint64_t PrevType = MachO->getRelocationType(RelPrev);
418 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR)
419 return true;
420 }
421 }
422
423 return false;
424 }
425
426 namespace {
427
428 /// Get the column at which we want to start printing the instruction
429 /// disassembly, taking into account anything which appears to the left of it.
getInstStartColumn(const MCSubtargetInfo & STI)430 unsigned getInstStartColumn(const MCSubtargetInfo &STI) {
431 return !ShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24;
432 }
433
isAArch64Elf(const ObjectFile * Obj)434 static bool isAArch64Elf(const ObjectFile *Obj) {
435 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
436 return Elf && Elf->getEMachine() == ELF::EM_AARCH64;
437 }
438
isArmElf(const ObjectFile * Obj)439 static bool isArmElf(const ObjectFile *Obj) {
440 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
441 return Elf && Elf->getEMachine() == ELF::EM_ARM;
442 }
443
hasMappingSymbols(const ObjectFile * Obj)444 static bool hasMappingSymbols(const ObjectFile *Obj) {
445 return isArmElf(Obj) || isAArch64Elf(Obj);
446 }
447
printRelocation(formatted_raw_ostream & OS,StringRef FileName,const RelocationRef & Rel,uint64_t Address,bool Is64Bits)448 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName,
449 const RelocationRef &Rel, uint64_t Address,
450 bool Is64Bits) {
451 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": ";
452 SmallString<16> Name;
453 SmallString<32> Val;
454 Rel.getTypeName(Name);
455 if (Error E = getRelocationValueString(Rel, Val))
456 reportError(std::move(E), FileName);
457 OS << format(Fmt.data(), Address) << Name << "\t" << Val;
458 }
459
460 class PrettyPrinter {
461 public:
462 virtual ~PrettyPrinter() = default;
463 virtual void
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)464 printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
465 object::SectionedAddress Address, formatted_raw_ostream &OS,
466 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
467 StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
468 LiveVariablePrinter &LVP) {
469 if (SP && (PrintSource || PrintLines))
470 SP->printSourceLine(OS, Address, ObjectFilename, LVP);
471 LVP.printBetweenInsts(OS, false);
472
473 size_t Start = OS.tell();
474 if (LeadingAddr)
475 OS << format("%8" PRIx64 ":", Address.Address);
476 if (ShowRawInsn) {
477 OS << ' ';
478 dumpBytes(Bytes, OS);
479 }
480
481 // The output of printInst starts with a tab. Print some spaces so that
482 // the tab has 1 column and advances to the target tab stop.
483 unsigned TabStop = getInstStartColumn(STI);
484 unsigned Column = OS.tell() - Start;
485 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8);
486
487 if (MI) {
488 // See MCInstPrinter::printInst. On targets where a PC relative immediate
489 // is relative to the next instruction and the length of a MCInst is
490 // difficult to measure (x86), this is the address of the next
491 // instruction.
492 uint64_t Addr =
493 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0);
494 IP.printInst(MI, Addr, "", STI, OS);
495 } else
496 OS << "\t<unknown>";
497 }
498 };
499 PrettyPrinter PrettyPrinterInst;
500
501 class HexagonPrettyPrinter : public PrettyPrinter {
502 public:
printLead(ArrayRef<uint8_t> Bytes,uint64_t Address,formatted_raw_ostream & OS)503 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
504 formatted_raw_ostream &OS) {
505 uint32_t opcode =
506 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
507 if (LeadingAddr)
508 OS << format("%8" PRIx64 ":", Address);
509 if (ShowRawInsn) {
510 OS << "\t";
511 dumpBytes(Bytes.slice(0, 4), OS);
512 OS << format("\t%08" PRIx32, opcode);
513 }
514 }
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)515 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
516 object::SectionedAddress Address, formatted_raw_ostream &OS,
517 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
518 StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
519 LiveVariablePrinter &LVP) override {
520 if (SP && (PrintSource || PrintLines))
521 SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
522 if (!MI) {
523 printLead(Bytes, Address.Address, OS);
524 OS << " <unknown>";
525 return;
526 }
527 std::string Buffer;
528 {
529 raw_string_ostream TempStream(Buffer);
530 IP.printInst(MI, Address.Address, "", STI, TempStream);
531 }
532 StringRef Contents(Buffer);
533 // Split off bundle attributes
534 auto PacketBundle = Contents.rsplit('\n');
535 // Split off first instruction from the rest
536 auto HeadTail = PacketBundle.first.split('\n');
537 auto Preamble = " { ";
538 auto Separator = "";
539
540 // Hexagon's packets require relocations to be inline rather than
541 // clustered at the end of the packet.
542 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin();
543 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end();
544 auto PrintReloc = [&]() -> void {
545 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) {
546 if (RelCur->getOffset() == Address.Address) {
547 printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false);
548 return;
549 }
550 ++RelCur;
551 }
552 };
553
554 while (!HeadTail.first.empty()) {
555 OS << Separator;
556 Separator = "\n";
557 if (SP && (PrintSource || PrintLines))
558 SP->printSourceLine(OS, Address, ObjectFilename, LVP, "");
559 printLead(Bytes, Address.Address, OS);
560 OS << Preamble;
561 Preamble = " ";
562 StringRef Inst;
563 auto Duplex = HeadTail.first.split('\v');
564 if (!Duplex.second.empty()) {
565 OS << Duplex.first;
566 OS << "; ";
567 Inst = Duplex.second;
568 }
569 else
570 Inst = HeadTail.first;
571 OS << Inst;
572 HeadTail = HeadTail.second.split('\n');
573 if (HeadTail.first.empty())
574 OS << " } " << PacketBundle.second;
575 PrintReloc();
576 Bytes = Bytes.slice(4);
577 Address.Address += 4;
578 }
579 }
580 };
581 HexagonPrettyPrinter HexagonPrettyPrinterInst;
582
583 class AMDGCNPrettyPrinter : public PrettyPrinter {
584 public:
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)585 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
586 object::SectionedAddress Address, formatted_raw_ostream &OS,
587 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
588 StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
589 LiveVariablePrinter &LVP) override {
590 if (SP && (PrintSource || PrintLines))
591 SP->printSourceLine(OS, Address, ObjectFilename, LVP);
592
593 if (MI) {
594 SmallString<40> InstStr;
595 raw_svector_ostream IS(InstStr);
596
597 IP.printInst(MI, Address.Address, "", STI, IS);
598
599 OS << left_justify(IS.str(), 60);
600 } else {
601 // an unrecognized encoding - this is probably data so represent it
602 // using the .long directive, or .byte directive if fewer than 4 bytes
603 // remaining
604 if (Bytes.size() >= 4) {
605 OS << format("\t.long 0x%08" PRIx32 " ",
606 support::endian::read32<support::little>(Bytes.data()));
607 OS.indent(42);
608 } else {
609 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]);
610 for (unsigned int i = 1; i < Bytes.size(); i++)
611 OS << format(", 0x%02" PRIx8, Bytes[i]);
612 OS.indent(55 - (6 * Bytes.size()));
613 }
614 }
615
616 OS << format("// %012" PRIX64 ":", Address.Address);
617 if (Bytes.size() >= 4) {
618 // D should be casted to uint32_t here as it is passed by format to
619 // snprintf as vararg.
620 for (uint32_t D : makeArrayRef(
621 reinterpret_cast<const support::little32_t *>(Bytes.data()),
622 Bytes.size() / 4))
623 OS << format(" %08" PRIX32, D);
624 } else {
625 for (unsigned char B : Bytes)
626 OS << format(" %02" PRIX8, B);
627 }
628
629 if (!Annot.empty())
630 OS << " // " << Annot;
631 }
632 };
633 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst;
634
635 class BPFPrettyPrinter : public PrettyPrinter {
636 public:
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)637 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes,
638 object::SectionedAddress Address, formatted_raw_ostream &OS,
639 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP,
640 StringRef ObjectFilename, std::vector<RelocationRef> *Rels,
641 LiveVariablePrinter &LVP) override {
642 if (SP && (PrintSource || PrintLines))
643 SP->printSourceLine(OS, Address, ObjectFilename, LVP);
644 if (LeadingAddr)
645 OS << format("%8" PRId64 ":", Address.Address / 8);
646 if (ShowRawInsn) {
647 OS << "\t";
648 dumpBytes(Bytes, OS);
649 }
650 if (MI)
651 IP.printInst(MI, Address.Address, "", STI, OS);
652 else
653 OS << "\t<unknown>";
654 }
655 };
656 BPFPrettyPrinter BPFPrettyPrinterInst;
657
selectPrettyPrinter(Triple const & Triple)658 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) {
659 switch(Triple.getArch()) {
660 default:
661 return PrettyPrinterInst;
662 case Triple::hexagon:
663 return HexagonPrettyPrinterInst;
664 case Triple::amdgcn:
665 return AMDGCNPrettyPrinterInst;
666 case Triple::bpfel:
667 case Triple::bpfeb:
668 return BPFPrettyPrinterInst;
669 }
670 }
671 }
672
getElfSymbolType(const ObjectFile * Obj,const SymbolRef & Sym)673 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) {
674 assert(Obj->isELF());
675 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
676 return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()),
677 Obj->getFileName())
678 ->getType();
679 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
680 return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()),
681 Obj->getFileName())
682 ->getType();
683 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
684 return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()),
685 Obj->getFileName())
686 ->getType();
687 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
688 return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()),
689 Obj->getFileName())
690 ->getType();
691 llvm_unreachable("Unsupported binary format");
692 }
693
694 template <class ELFT> static void
addDynamicElfSymbols(const ELFObjectFile<ELFT> * Obj,std::map<SectionRef,SectionSymbolsTy> & AllSymbols)695 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj,
696 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
697 for (auto Symbol : Obj->getDynamicSymbolIterators()) {
698 uint8_t SymbolType = Symbol.getELFType();
699 if (SymbolType == ELF::STT_SECTION)
700 continue;
701
702 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName());
703 // ELFSymbolRef::getAddress() returns size instead of value for common
704 // symbols which is not desirable for disassembly output. Overriding.
705 if (SymbolType == ELF::STT_COMMON)
706 Address = unwrapOrError(Obj->getSymbol(Symbol.getRawDataRefImpl()),
707 Obj->getFileName())
708 ->st_value;
709
710 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName());
711 if (Name.empty())
712 continue;
713
714 section_iterator SecI =
715 unwrapOrError(Symbol.getSection(), Obj->getFileName());
716 if (SecI == Obj->section_end())
717 continue;
718
719 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType);
720 }
721 }
722
723 static void
addDynamicElfSymbols(const ObjectFile * Obj,std::map<SectionRef,SectionSymbolsTy> & AllSymbols)724 addDynamicElfSymbols(const ObjectFile *Obj,
725 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) {
726 assert(Obj->isELF());
727 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj))
728 addDynamicElfSymbols(Elf32LEObj, AllSymbols);
729 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj))
730 addDynamicElfSymbols(Elf64LEObj, AllSymbols);
731 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj))
732 addDynamicElfSymbols(Elf32BEObj, AllSymbols);
733 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj))
734 addDynamicElfSymbols(Elf64BEObj, AllSymbols);
735 else
736 llvm_unreachable("Unsupported binary format");
737 }
738
addPltEntries(const ObjectFile * Obj,std::map<SectionRef,SectionSymbolsTy> & AllSymbols,StringSaver & Saver)739 static void addPltEntries(const ObjectFile *Obj,
740 std::map<SectionRef, SectionSymbolsTy> &AllSymbols,
741 StringSaver &Saver) {
742 Optional<SectionRef> Plt = None;
743 for (const SectionRef &Section : Obj->sections()) {
744 Expected<StringRef> SecNameOrErr = Section.getName();
745 if (!SecNameOrErr) {
746 consumeError(SecNameOrErr.takeError());
747 continue;
748 }
749 if (*SecNameOrErr == ".plt")
750 Plt = Section;
751 }
752 if (!Plt)
753 return;
754 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) {
755 for (auto PltEntry : ElfObj->getPltAddresses()) {
756 if (PltEntry.first) {
757 SymbolRef Symbol(*PltEntry.first, ElfObj);
758 uint8_t SymbolType = getElfSymbolType(Obj, Symbol);
759 if (Expected<StringRef> NameOrErr = Symbol.getName()) {
760 if (!NameOrErr->empty())
761 AllSymbols[*Plt].emplace_back(
762 PltEntry.second, Saver.save((*NameOrErr + "@plt").str()),
763 SymbolType);
764 continue;
765 } else {
766 // The warning has been reported in disassembleObject().
767 consumeError(NameOrErr.takeError());
768 }
769 }
770 reportWarning("PLT entry at 0x" + Twine::utohexstr(PltEntry.second) +
771 " references an invalid symbol",
772 Obj->getFileName());
773 }
774 }
775 }
776
777 // Normally the disassembly output will skip blocks of zeroes. This function
778 // returns the number of zero bytes that can be skipped when dumping the
779 // disassembly of the instructions in Buf.
countSkippableZeroBytes(ArrayRef<uint8_t> Buf)780 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) {
781 // Find the number of leading zeroes.
782 size_t N = 0;
783 while (N < Buf.size() && !Buf[N])
784 ++N;
785
786 // We may want to skip blocks of zero bytes, but unless we see
787 // at least 8 of them in a row.
788 if (N < 8)
789 return 0;
790
791 // We skip zeroes in multiples of 4 because do not want to truncate an
792 // instruction if it starts with a zero byte.
793 return N & ~0x3;
794 }
795
796 // Returns a map from sections to their relocations.
797 static std::map<SectionRef, std::vector<RelocationRef>>
getRelocsMap(object::ObjectFile const & Obj)798 getRelocsMap(object::ObjectFile const &Obj) {
799 std::map<SectionRef, std::vector<RelocationRef>> Ret;
800 uint64_t I = (uint64_t)-1;
801 for (SectionRef Sec : Obj.sections()) {
802 ++I;
803 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection();
804 if (!RelocatedOrErr)
805 reportError(Obj.getFileName(),
806 "section (" + Twine(I) +
807 "): failed to get a relocated section: " +
808 toString(RelocatedOrErr.takeError()));
809
810 section_iterator Relocated = *RelocatedOrErr;
811 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep)
812 continue;
813 std::vector<RelocationRef> &V = Ret[*Relocated];
814 append_range(V, Sec.relocations());
815 // Sort relocations by address.
816 llvm::stable_sort(V, isRelocAddressLess);
817 }
818 return Ret;
819 }
820
821 // Used for --adjust-vma to check if address should be adjusted by the
822 // specified value for a given section.
823 // For ELF we do not adjust non-allocatable sections like debug ones,
824 // because they are not loadable.
825 // TODO: implement for other file formats.
shouldAdjustVA(const SectionRef & Section)826 static bool shouldAdjustVA(const SectionRef &Section) {
827 const ObjectFile *Obj = Section.getObject();
828 if (Obj->isELF())
829 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC;
830 return false;
831 }
832
833
834 typedef std::pair<uint64_t, char> MappingSymbolPair;
getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols,uint64_t Address)835 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols,
836 uint64_t Address) {
837 auto It =
838 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) {
839 return Val.first <= Address;
840 });
841 // Return zero for any address before the first mapping symbol; this means
842 // we should use the default disassembly mode, depending on the target.
843 if (It == MappingSymbols.begin())
844 return '\x00';
845 return (It - 1)->second;
846 }
847
dumpARMELFData(uint64_t SectionAddr,uint64_t Index,uint64_t End,const ObjectFile * Obj,ArrayRef<uint8_t> Bytes,ArrayRef<MappingSymbolPair> MappingSymbols,raw_ostream & OS)848 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index,
849 uint64_t End, const ObjectFile *Obj,
850 ArrayRef<uint8_t> Bytes,
851 ArrayRef<MappingSymbolPair> MappingSymbols,
852 raw_ostream &OS) {
853 support::endianness Endian =
854 Obj->isLittleEndian() ? support::little : support::big;
855 OS << format("%8" PRIx64 ":\t", SectionAddr + Index);
856 if (Index + 4 <= End) {
857 dumpBytes(Bytes.slice(Index, 4), OS);
858 OS << "\t.word\t"
859 << format_hex(support::endian::read32(Bytes.data() + Index, Endian),
860 10);
861 return 4;
862 }
863 if (Index + 2 <= End) {
864 dumpBytes(Bytes.slice(Index, 2), OS);
865 OS << "\t\t.short\t"
866 << format_hex(support::endian::read16(Bytes.data() + Index, Endian),
867 6);
868 return 2;
869 }
870 dumpBytes(Bytes.slice(Index, 1), OS);
871 OS << "\t\t.byte\t" << format_hex(Bytes[0], 4);
872 return 1;
873 }
874
dumpELFData(uint64_t SectionAddr,uint64_t Index,uint64_t End,ArrayRef<uint8_t> Bytes)875 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End,
876 ArrayRef<uint8_t> Bytes) {
877 // print out data up to 8 bytes at a time in hex and ascii
878 uint8_t AsciiData[9] = {'\0'};
879 uint8_t Byte;
880 int NumBytes = 0;
881
882 for (; Index < End; ++Index) {
883 if (NumBytes == 0)
884 outs() << format("%8" PRIx64 ":", SectionAddr + Index);
885 Byte = Bytes.slice(Index)[0];
886 outs() << format(" %02x", Byte);
887 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.';
888
889 uint8_t IndentOffset = 0;
890 NumBytes++;
891 if (Index == End - 1 || NumBytes > 8) {
892 // Indent the space for less than 8 bytes data.
893 // 2 spaces for byte and one for space between bytes
894 IndentOffset = 3 * (8 - NumBytes);
895 for (int Excess = NumBytes; Excess < 8; Excess++)
896 AsciiData[Excess] = '\0';
897 NumBytes = 8;
898 }
899 if (NumBytes == 8) {
900 AsciiData[8] = '\0';
901 outs() << std::string(IndentOffset, ' ') << " ";
902 outs() << reinterpret_cast<char *>(AsciiData);
903 outs() << '\n';
904 NumBytes = 0;
905 }
906 }
907 }
908
createSymbolInfo(const ObjectFile * Obj,const SymbolRef & Symbol)909 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile *Obj,
910 const SymbolRef &Symbol) {
911 const StringRef FileName = Obj->getFileName();
912 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName);
913 const StringRef Name = unwrapOrError(Symbol.getName(), FileName);
914
915 if (Obj->isXCOFF() && SymbolDescription) {
916 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj);
917 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl();
918
919 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p);
920 Optional<XCOFF::StorageMappingClass> Smc =
921 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol);
922 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex,
923 isLabel(XCOFFObj, Symbol));
924 } else
925 return SymbolInfoTy(Addr, Name,
926 Obj->isELF() ? getElfSymbolType(Obj, Symbol)
927 : (uint8_t)ELF::STT_NOTYPE);
928 }
929
createDummySymbolInfo(const ObjectFile * Obj,const uint64_t Addr,StringRef & Name,uint8_t Type)930 static SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj,
931 const uint64_t Addr, StringRef &Name,
932 uint8_t Type) {
933 if (Obj->isXCOFF() && SymbolDescription)
934 return SymbolInfoTy(Addr, Name, None, None, false);
935 else
936 return SymbolInfoTy(Addr, Name, Type);
937 }
938
939 static void
collectLocalBranchTargets(ArrayRef<uint8_t> Bytes,const 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)940 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA,
941 MCDisassembler *DisAsm, MCInstPrinter *IP,
942 const MCSubtargetInfo *STI, uint64_t SectionAddr,
943 uint64_t Start, uint64_t End,
944 std::unordered_map<uint64_t, std::string> &Labels) {
945 // So far only supports X86.
946 if (!STI->getTargetTriple().isX86())
947 return;
948
949 Labels.clear();
950 unsigned LabelCount = 0;
951 Start += SectionAddr;
952 End += SectionAddr;
953 uint64_t Index = Start;
954 while (Index < End) {
955 // Disassemble a real instruction and record function-local branch labels.
956 MCInst Inst;
957 uint64_t Size;
958 bool Disassembled = DisAsm->getInstruction(
959 Inst, Size, Bytes.slice(Index - SectionAddr), Index, nulls());
960 if (Size == 0)
961 Size = 1;
962
963 if (Disassembled && MIA) {
964 uint64_t Target;
965 bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target);
966 if (TargetKnown && (Target >= Start && Target < End) &&
967 !Labels.count(Target))
968 Labels[Target] = ("L" + Twine(LabelCount++)).str();
969 }
970
971 Index += Size;
972 }
973 }
974
975 // Create an MCSymbolizer for the target and add it to the MCDisassembler.
976 // This is currently only used on AMDGPU, and assumes the format of the
977 // void * argument passed to AMDGPU's createMCSymbolizer.
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)978 static void addSymbolizer(
979 MCContext &Ctx, const Target *Target, StringRef TripleName,
980 MCDisassembler *DisAsm, uint64_t SectionAddr, ArrayRef<uint8_t> Bytes,
981 SectionSymbolsTy &Symbols,
982 std::vector<std::unique_ptr<std::string>> &SynthesizedLabelNames) {
983
984 std::unique_ptr<MCRelocationInfo> RelInfo(
985 Target->createMCRelocationInfo(TripleName, Ctx));
986 if (!RelInfo)
987 return;
988 std::unique_ptr<MCSymbolizer> Symbolizer(Target->createMCSymbolizer(
989 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
990 MCSymbolizer *SymbolizerPtr = &*Symbolizer;
991 DisAsm->setSymbolizer(std::move(Symbolizer));
992
993 if (!SymbolizeOperands)
994 return;
995
996 // Synthesize labels referenced by branch instructions by
997 // disassembling, discarding the output, and collecting the referenced
998 // addresses from the symbolizer.
999 for (size_t Index = 0; Index != Bytes.size();) {
1000 MCInst Inst;
1001 uint64_t Size;
1002 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), SectionAddr + Index,
1003 nulls());
1004 if (Size == 0)
1005 Size = 1;
1006 Index += Size;
1007 }
1008 ArrayRef<uint64_t> LabelAddrsRef = SymbolizerPtr->getReferencedAddresses();
1009 // Copy and sort to remove duplicates.
1010 std::vector<uint64_t> LabelAddrs;
1011 LabelAddrs.insert(LabelAddrs.end(), LabelAddrsRef.begin(),
1012 LabelAddrsRef.end());
1013 llvm::sort(LabelAddrs);
1014 LabelAddrs.resize(std::unique(LabelAddrs.begin(), LabelAddrs.end()) -
1015 LabelAddrs.begin());
1016 // Add the labels.
1017 for (unsigned LabelNum = 0; LabelNum != LabelAddrs.size(); ++LabelNum) {
1018 auto Name = std::make_unique<std::string>();
1019 *Name = (Twine("L") + Twine(LabelNum)).str();
1020 SynthesizedLabelNames.push_back(std::move(Name));
1021 Symbols.push_back(SymbolInfoTy(
1022 LabelAddrs[LabelNum], *SynthesizedLabelNames.back(), ELF::STT_NOTYPE));
1023 }
1024 llvm::stable_sort(Symbols);
1025 // Recreate the symbolizer with the new symbols list.
1026 RelInfo.reset(Target->createMCRelocationInfo(TripleName, Ctx));
1027 Symbolizer.reset(Target->createMCSymbolizer(
1028 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo)));
1029 DisAsm->setSymbolizer(std::move(Symbolizer));
1030 }
1031
getSegmentName(const MachOObjectFile * MachO,const SectionRef & Section)1032 static StringRef getSegmentName(const MachOObjectFile *MachO,
1033 const SectionRef &Section) {
1034 if (MachO) {
1035 DataRefImpl DR = Section.getRawDataRefImpl();
1036 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR);
1037 return SegmentName;
1038 }
1039 return "";
1040 }
1041
disassembleObject(const Target * TheTarget,const ObjectFile * Obj,MCContext & Ctx,MCDisassembler * PrimaryDisAsm,MCDisassembler * SecondaryDisAsm,const MCInstrAnalysis * MIA,MCInstPrinter * IP,const MCSubtargetInfo * PrimarySTI,const MCSubtargetInfo * SecondarySTI,PrettyPrinter & PIP,SourcePrinter & SP,bool InlineRelocs)1042 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj,
1043 MCContext &Ctx, MCDisassembler *PrimaryDisAsm,
1044 MCDisassembler *SecondaryDisAsm,
1045 const MCInstrAnalysis *MIA, MCInstPrinter *IP,
1046 const MCSubtargetInfo *PrimarySTI,
1047 const MCSubtargetInfo *SecondarySTI,
1048 PrettyPrinter &PIP,
1049 SourcePrinter &SP, bool InlineRelocs) {
1050 const MCSubtargetInfo *STI = PrimarySTI;
1051 MCDisassembler *DisAsm = PrimaryDisAsm;
1052 bool PrimaryIsThumb = false;
1053 if (isArmElf(Obj))
1054 PrimaryIsThumb = STI->checkFeatures("+thumb-mode");
1055
1056 std::map<SectionRef, std::vector<RelocationRef>> RelocMap;
1057 if (InlineRelocs)
1058 RelocMap = getRelocsMap(*Obj);
1059 bool Is64Bits = Obj->getBytesInAddress() > 4;
1060
1061 // Create a mapping from virtual address to symbol name. This is used to
1062 // pretty print the symbols while disassembling.
1063 std::map<SectionRef, SectionSymbolsTy> AllSymbols;
1064 SectionSymbolsTy AbsoluteSymbols;
1065 const StringRef FileName = Obj->getFileName();
1066 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
1067 for (const SymbolRef &Symbol : Obj->symbols()) {
1068 Expected<StringRef> NameOrErr = Symbol.getName();
1069 if (!NameOrErr) {
1070 reportWarning(toString(NameOrErr.takeError()), FileName);
1071 continue;
1072 }
1073 if (NameOrErr->empty() && !(Obj->isXCOFF() && SymbolDescription))
1074 continue;
1075
1076 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION)
1077 continue;
1078
1079 if (MachO) {
1080 // __mh_(execute|dylib|dylinker|bundle|preload|object)_header are special
1081 // symbols that support MachO header introspection. They do not bind to
1082 // code locations and are irrelevant for disassembly.
1083 if (NameOrErr->startswith("__mh_") && NameOrErr->endswith("_header"))
1084 continue;
1085 // Don't ask a Mach-O STAB symbol for its section unless you know that
1086 // STAB symbol's section field refers to a valid section index. Otherwise
1087 // the symbol may error trying to load a section that does not exist.
1088 DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1089 uint8_t NType = (MachO->is64Bit() ?
1090 MachO->getSymbol64TableEntry(SymDRI).n_type:
1091 MachO->getSymbolTableEntry(SymDRI).n_type);
1092 if (NType & MachO::N_STAB)
1093 continue;
1094 }
1095
1096 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName);
1097 if (SecI != Obj->section_end())
1098 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol));
1099 else
1100 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol));
1101 }
1102
1103 if (AllSymbols.empty() && Obj->isELF())
1104 addDynamicElfSymbols(Obj, AllSymbols);
1105
1106 BumpPtrAllocator A;
1107 StringSaver Saver(A);
1108 addPltEntries(Obj, AllSymbols, Saver);
1109
1110 // Create a mapping from virtual address to section. An empty section can
1111 // cause more than one section at the same address. Sort such sections to be
1112 // before same-addressed non-empty sections so that symbol lookups prefer the
1113 // non-empty section.
1114 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses;
1115 for (SectionRef Sec : Obj->sections())
1116 SectionAddresses.emplace_back(Sec.getAddress(), Sec);
1117 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) {
1118 if (LHS.first != RHS.first)
1119 return LHS.first < RHS.first;
1120 return LHS.second.getSize() < RHS.second.getSize();
1121 });
1122
1123 // Linked executables (.exe and .dll files) typically don't include a real
1124 // symbol table but they might contain an export table.
1125 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) {
1126 for (const auto &ExportEntry : COFFObj->export_directories()) {
1127 StringRef Name;
1128 if (Error E = ExportEntry.getSymbolName(Name))
1129 reportError(std::move(E), Obj->getFileName());
1130 if (Name.empty())
1131 continue;
1132
1133 uint32_t RVA;
1134 if (Error E = ExportEntry.getExportRVA(RVA))
1135 reportError(std::move(E), Obj->getFileName());
1136
1137 uint64_t VA = COFFObj->getImageBase() + RVA;
1138 auto Sec = partition_point(
1139 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) {
1140 return O.first <= VA;
1141 });
1142 if (Sec != SectionAddresses.begin()) {
1143 --Sec;
1144 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE);
1145 } else
1146 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE);
1147 }
1148 }
1149
1150 // Sort all the symbols, this allows us to use a simple binary search to find
1151 // Multiple symbols can have the same address. Use a stable sort to stabilize
1152 // the output.
1153 StringSet<> FoundDisasmSymbolSet;
1154 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols)
1155 llvm::stable_sort(SecSyms.second);
1156 llvm::stable_sort(AbsoluteSymbols);
1157
1158 std::unique_ptr<DWARFContext> DICtx;
1159 LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI);
1160
1161 if (DbgVariables != DVDisabled) {
1162 DICtx = DWARFContext::create(*Obj);
1163 for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units())
1164 LVP.addCompileUnit(CU->getUnitDIE(false));
1165 }
1166
1167 LLVM_DEBUG(LVP.dump());
1168
1169 for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1170 if (FilterSections.empty() && !DisassembleAll &&
1171 (!Section.isText() || Section.isVirtual()))
1172 continue;
1173
1174 uint64_t SectionAddr = Section.getAddress();
1175 uint64_t SectSize = Section.getSize();
1176 if (!SectSize)
1177 continue;
1178
1179 // Get the list of all the symbols in this section.
1180 SectionSymbolsTy &Symbols = AllSymbols[Section];
1181 std::vector<MappingSymbolPair> MappingSymbols;
1182 if (hasMappingSymbols(Obj)) {
1183 for (const auto &Symb : Symbols) {
1184 uint64_t Address = Symb.Addr;
1185 StringRef Name = Symb.Name;
1186 if (Name.startswith("$d"))
1187 MappingSymbols.emplace_back(Address - SectionAddr, 'd');
1188 if (Name.startswith("$x"))
1189 MappingSymbols.emplace_back(Address - SectionAddr, 'x');
1190 if (Name.startswith("$a"))
1191 MappingSymbols.emplace_back(Address - SectionAddr, 'a');
1192 if (Name.startswith("$t"))
1193 MappingSymbols.emplace_back(Address - SectionAddr, 't');
1194 }
1195 }
1196
1197 llvm::sort(MappingSymbols);
1198
1199 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef(
1200 unwrapOrError(Section.getContents(), Obj->getFileName()));
1201
1202 std::vector<std::unique_ptr<std::string>> SynthesizedLabelNames;
1203 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) {
1204 // AMDGPU disassembler uses symbolizer for printing labels
1205 addSymbolizer(Ctx, TheTarget, TripleName, DisAsm, SectionAddr, Bytes,
1206 Symbols, SynthesizedLabelNames);
1207 }
1208
1209 StringRef SegmentName = getSegmentName(MachO, Section);
1210 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName());
1211 // If the section has no symbol at the start, just insert a dummy one.
1212 if (Symbols.empty() || Symbols[0].Addr != 0) {
1213 Symbols.insert(Symbols.begin(),
1214 createDummySymbolInfo(Obj, SectionAddr, SectionName,
1215 Section.isText() ? ELF::STT_FUNC
1216 : ELF::STT_OBJECT));
1217 }
1218
1219 SmallString<40> Comments;
1220 raw_svector_ostream CommentStream(Comments);
1221
1222 uint64_t VMAAdjustment = 0;
1223 if (shouldAdjustVA(Section))
1224 VMAAdjustment = AdjustVMA;
1225
1226 uint64_t Size;
1227 uint64_t Index;
1228 bool PrintedSection = false;
1229 std::vector<RelocationRef> Rels = RelocMap[Section];
1230 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin();
1231 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end();
1232 // Disassemble symbol by symbol.
1233 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) {
1234 std::string SymbolName = Symbols[SI].Name.str();
1235 if (Demangle)
1236 SymbolName = demangle(SymbolName);
1237
1238 // Skip if --disassemble-symbols is not empty and the symbol is not in
1239 // the list.
1240 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName))
1241 continue;
1242
1243 uint64_t Start = Symbols[SI].Addr;
1244 if (Start < SectionAddr || StopAddress <= Start)
1245 continue;
1246 else
1247 FoundDisasmSymbolSet.insert(SymbolName);
1248
1249 // The end is the section end, the beginning of the next symbol, or
1250 // --stop-address.
1251 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress);
1252 if (SI + 1 < SE)
1253 End = std::min(End, Symbols[SI + 1].Addr);
1254 if (Start >= End || End <= StartAddress)
1255 continue;
1256 Start -= SectionAddr;
1257 End -= SectionAddr;
1258
1259 if (!PrintedSection) {
1260 PrintedSection = true;
1261 outs() << "\nDisassembly of section ";
1262 if (!SegmentName.empty())
1263 outs() << SegmentName << ",";
1264 outs() << SectionName << ":\n";
1265 }
1266
1267 outs() << '\n';
1268 if (LeadingAddr)
1269 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ",
1270 SectionAddr + Start + VMAAdjustment);
1271 if (Obj->isXCOFF() && SymbolDescription) {
1272 outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n";
1273 } else
1274 outs() << '<' << SymbolName << ">:\n";
1275
1276 // Don't print raw contents of a virtual section. A virtual section
1277 // doesn't have any contents in the file.
1278 if (Section.isVirtual()) {
1279 outs() << "...\n";
1280 continue;
1281 }
1282
1283 auto Status = DisAsm->onSymbolStart(Symbols[SI], Size,
1284 Bytes.slice(Start, End - Start),
1285 SectionAddr + Start, CommentStream);
1286 // To have round trippable disassembly, we fall back to decoding the
1287 // remaining bytes as instructions.
1288 //
1289 // If there is a failure, we disassemble the failed region as bytes before
1290 // falling back. The target is expected to print nothing in this case.
1291 //
1292 // If there is Success or SoftFail i.e no 'real' failure, we go ahead by
1293 // Size bytes before falling back.
1294 // So if the entire symbol is 'eaten' by the target:
1295 // Start += Size // Now Start = End and we will never decode as
1296 // // instructions
1297 //
1298 // Right now, most targets return None i.e ignore to treat a symbol
1299 // separately. But WebAssembly decodes preludes for some symbols.
1300 //
1301 if (Status.hasValue()) {
1302 if (Status.getValue() == MCDisassembler::Fail) {
1303 outs() << "// Error in decoding " << SymbolName
1304 << " : Decoding failed region as bytes.\n";
1305 for (uint64_t I = 0; I < Size; ++I) {
1306 outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true)
1307 << "\n";
1308 }
1309 }
1310 } else {
1311 Size = 0;
1312 }
1313
1314 Start += Size;
1315
1316 Index = Start;
1317 if (SectionAddr < StartAddress)
1318 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr);
1319
1320 // If there is a data/common symbol inside an ELF text section and we are
1321 // only disassembling text (applicable all architectures), we are in a
1322 // situation where we must print the data and not disassemble it.
1323 if (Obj->isELF() && !DisassembleAll && Section.isText()) {
1324 uint8_t SymTy = Symbols[SI].Type;
1325 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) {
1326 dumpELFData(SectionAddr, Index, End, Bytes);
1327 Index = End;
1328 }
1329 }
1330
1331 bool CheckARMELFData = hasMappingSymbols(Obj) &&
1332 Symbols[SI].Type != ELF::STT_OBJECT &&
1333 !DisassembleAll;
1334 bool DumpARMELFData = false;
1335 formatted_raw_ostream FOS(outs());
1336
1337 std::unordered_map<uint64_t, std::string> AllLabels;
1338 if (SymbolizeOperands)
1339 collectLocalBranchTargets(Bytes, MIA, DisAsm, IP, PrimarySTI,
1340 SectionAddr, Index, End, AllLabels);
1341
1342 while (Index < End) {
1343 // ARM and AArch64 ELF binaries can interleave data and text in the
1344 // same section. We rely on the markers introduced to understand what
1345 // we need to dump. If the data marker is within a function, it is
1346 // denoted as a word/short etc.
1347 if (CheckARMELFData) {
1348 char Kind = getMappingSymbolKind(MappingSymbols, Index);
1349 DumpARMELFData = Kind == 'd';
1350 if (SecondarySTI) {
1351 if (Kind == 'a') {
1352 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI;
1353 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm;
1354 } else if (Kind == 't') {
1355 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI;
1356 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm;
1357 }
1358 }
1359 }
1360
1361 if (DumpARMELFData) {
1362 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes,
1363 MappingSymbols, FOS);
1364 } else {
1365 // When -z or --disassemble-zeroes are given we always dissasemble
1366 // them. Otherwise we might want to skip zero bytes we see.
1367 if (!DisassembleZeroes) {
1368 uint64_t MaxOffset = End - Index;
1369 // For --reloc: print zero blocks patched by relocations, so that
1370 // relocations can be shown in the dump.
1371 if (RelCur != RelEnd)
1372 MaxOffset = RelCur->getOffset() - Index;
1373
1374 if (size_t N =
1375 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) {
1376 FOS << "\t\t..." << '\n';
1377 Index += N;
1378 continue;
1379 }
1380 }
1381
1382 // Print local label if there's any.
1383 auto Iter = AllLabels.find(SectionAddr + Index);
1384 if (Iter != AllLabels.end())
1385 FOS << "<" << Iter->second << ">:\n";
1386
1387 // Disassemble a real instruction or a data when disassemble all is
1388 // provided
1389 MCInst Inst;
1390 bool Disassembled =
1391 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
1392 SectionAddr + Index, CommentStream);
1393 if (Size == 0)
1394 Size = 1;
1395
1396 LVP.update({Index, Section.getIndex()},
1397 {Index + Size, Section.getIndex()}, Index + Size != End);
1398
1399 PIP.printInst(
1400 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size),
1401 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS,
1402 "", *STI, &SP, Obj->getFileName(), &Rels, LVP);
1403 FOS << CommentStream.str();
1404 Comments.clear();
1405
1406 // If disassembly has failed, avoid analysing invalid/incomplete
1407 // instruction information. Otherwise, try to resolve the target
1408 // address (jump target or memory operand address) and print it on the
1409 // right of the instruction.
1410 if (Disassembled && MIA) {
1411 uint64_t Target;
1412 bool PrintTarget =
1413 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target);
1414 if (!PrintTarget)
1415 if (Optional<uint64_t> MaybeTarget =
1416 MIA->evaluateMemoryOperandAddress(
1417 Inst, SectionAddr + Index, Size)) {
1418 Target = *MaybeTarget;
1419 PrintTarget = true;
1420 // Do not print real address when symbolizing.
1421 if (!SymbolizeOperands)
1422 FOS << " # " << Twine::utohexstr(Target);
1423 }
1424 if (PrintTarget) {
1425 // In a relocatable object, the target's section must reside in
1426 // the same section as the call instruction or it is accessed
1427 // through a relocation.
1428 //
1429 // In a non-relocatable object, the target may be in any section.
1430 // In that case, locate the section(s) containing the target
1431 // address and find the symbol in one of those, if possible.
1432 //
1433 // N.B. We don't walk the relocations in the relocatable case yet.
1434 std::vector<const SectionSymbolsTy *> TargetSectionSymbols;
1435 if (!Obj->isRelocatableObject()) {
1436 auto It = llvm::partition_point(
1437 SectionAddresses,
1438 [=](const std::pair<uint64_t, SectionRef> &O) {
1439 return O.first <= Target;
1440 });
1441 uint64_t TargetSecAddr = 0;
1442 while (It != SectionAddresses.begin()) {
1443 --It;
1444 if (TargetSecAddr == 0)
1445 TargetSecAddr = It->first;
1446 if (It->first != TargetSecAddr)
1447 break;
1448 TargetSectionSymbols.push_back(&AllSymbols[It->second]);
1449 }
1450 } else {
1451 TargetSectionSymbols.push_back(&Symbols);
1452 }
1453 TargetSectionSymbols.push_back(&AbsoluteSymbols);
1454
1455 // Find the last symbol in the first candidate section whose
1456 // offset is less than or equal to the target. If there are no
1457 // such symbols, try in the next section and so on, before finally
1458 // using the nearest preceding absolute symbol (if any), if there
1459 // are no other valid symbols.
1460 const SymbolInfoTy *TargetSym = nullptr;
1461 for (const SectionSymbolsTy *TargetSymbols :
1462 TargetSectionSymbols) {
1463 auto It = llvm::partition_point(
1464 *TargetSymbols,
1465 [=](const SymbolInfoTy &O) { return O.Addr <= Target; });
1466 if (It != TargetSymbols->begin()) {
1467 TargetSym = &*(It - 1);
1468 break;
1469 }
1470 }
1471
1472 // Print the labels corresponding to the target if there's any.
1473 bool LabelAvailable = AllLabels.count(Target);
1474 if (TargetSym != nullptr) {
1475 uint64_t TargetAddress = TargetSym->Addr;
1476 uint64_t Disp = Target - TargetAddress;
1477 std::string TargetName = TargetSym->Name.str();
1478 if (Demangle)
1479 TargetName = demangle(TargetName);
1480
1481 FOS << " <";
1482 if (!Disp) {
1483 // Always Print the binary symbol precisely corresponding to
1484 // the target address.
1485 FOS << TargetName;
1486 } else if (!LabelAvailable) {
1487 // Always Print the binary symbol plus an offset if there's no
1488 // local label corresponding to the target address.
1489 FOS << TargetName << "+0x" << Twine::utohexstr(Disp);
1490 } else {
1491 FOS << AllLabels[Target];
1492 }
1493 FOS << ">";
1494 } else if (LabelAvailable) {
1495 FOS << " <" << AllLabels[Target] << ">";
1496 }
1497 }
1498 }
1499 }
1500
1501 LVP.printAfterInst(FOS);
1502 FOS << "\n";
1503
1504 // Hexagon does this in pretty printer
1505 if (Obj->getArch() != Triple::hexagon) {
1506 // Print relocation for instruction and data.
1507 while (RelCur != RelEnd) {
1508 uint64_t Offset = RelCur->getOffset();
1509 // If this relocation is hidden, skip it.
1510 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) {
1511 ++RelCur;
1512 continue;
1513 }
1514
1515 // Stop when RelCur's offset is past the disassembled
1516 // instruction/data. Note that it's possible the disassembled data
1517 // is not the complete data: we might see the relocation printed in
1518 // the middle of the data, but this matches the binutils objdump
1519 // output.
1520 if (Offset >= Index + Size)
1521 break;
1522
1523 // When --adjust-vma is used, update the address printed.
1524 if (RelCur->getSymbol() != Obj->symbol_end()) {
1525 Expected<section_iterator> SymSI =
1526 RelCur->getSymbol()->getSection();
1527 if (SymSI && *SymSI != Obj->section_end() &&
1528 shouldAdjustVA(**SymSI))
1529 Offset += AdjustVMA;
1530 }
1531
1532 printRelocation(FOS, Obj->getFileName(), *RelCur,
1533 SectionAddr + Offset, Is64Bits);
1534 LVP.printAfterOtherLine(FOS, true);
1535 ++RelCur;
1536 }
1537 }
1538
1539 Index += Size;
1540 }
1541 }
1542 }
1543 StringSet<> MissingDisasmSymbolSet =
1544 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet);
1545 for (StringRef Sym : MissingDisasmSymbolSet.keys())
1546 reportWarning("failed to disassemble missing symbol " + Sym, FileName);
1547 }
1548
disassembleObject(const ObjectFile * Obj,bool InlineRelocs)1549 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) {
1550 const Target *TheTarget = getTarget(Obj);
1551
1552 // Package up features to be passed to target/subtarget
1553 SubtargetFeatures Features = Obj->getFeatures();
1554 if (!MAttrs.empty())
1555 for (unsigned I = 0; I != MAttrs.size(); ++I)
1556 Features.AddFeature(MAttrs[I]);
1557
1558 std::unique_ptr<const MCRegisterInfo> MRI(
1559 TheTarget->createMCRegInfo(TripleName));
1560 if (!MRI)
1561 reportError(Obj->getFileName(),
1562 "no register info for target " + TripleName);
1563
1564 // Set up disassembler.
1565 MCTargetOptions MCOptions;
1566 std::unique_ptr<const MCAsmInfo> AsmInfo(
1567 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions));
1568 if (!AsmInfo)
1569 reportError(Obj->getFileName(),
1570 "no assembly info for target " + TripleName);
1571
1572 if (MCPU.empty())
1573 MCPU = Obj->tryGetCPUName().getValueOr("").str();
1574
1575 std::unique_ptr<const MCSubtargetInfo> STI(
1576 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString()));
1577 if (!STI)
1578 reportError(Obj->getFileName(),
1579 "no subtarget info for target " + TripleName);
1580 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo());
1581 if (!MII)
1582 reportError(Obj->getFileName(),
1583 "no instruction info for target " + TripleName);
1584 MCContext Ctx(Triple(TripleName), AsmInfo.get(), MRI.get(), STI.get());
1585 // FIXME: for now initialize MCObjectFileInfo with default values
1586 std::unique_ptr<MCObjectFileInfo> MOFI(
1587 TheTarget->createMCObjectFileInfo(Ctx, /*PIC=*/false));
1588 Ctx.setObjectFileInfo(MOFI.get());
1589
1590 std::unique_ptr<MCDisassembler> DisAsm(
1591 TheTarget->createMCDisassembler(*STI, Ctx));
1592 if (!DisAsm)
1593 reportError(Obj->getFileName(), "no disassembler for target " + TripleName);
1594
1595 // If we have an ARM object file, we need a second disassembler, because
1596 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode.
1597 // We use mapping symbols to switch between the two assemblers, where
1598 // appropriate.
1599 std::unique_ptr<MCDisassembler> SecondaryDisAsm;
1600 std::unique_ptr<const MCSubtargetInfo> SecondarySTI;
1601 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) {
1602 if (STI->checkFeatures("+thumb-mode"))
1603 Features.AddFeature("-thumb-mode");
1604 else
1605 Features.AddFeature("+thumb-mode");
1606 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU,
1607 Features.getString()));
1608 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx));
1609 }
1610
1611 std::unique_ptr<const MCInstrAnalysis> MIA(
1612 TheTarget->createMCInstrAnalysis(MII.get()));
1613
1614 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
1615 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
1616 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI));
1617 if (!IP)
1618 reportError(Obj->getFileName(),
1619 "no instruction printer for target " + TripleName);
1620 IP->setPrintImmHex(PrintImmHex);
1621 IP->setPrintBranchImmAsAddress(true);
1622 IP->setSymbolizeOperands(SymbolizeOperands);
1623 IP->setMCInstrAnalysis(MIA.get());
1624
1625 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName));
1626 SourcePrinter SP(Obj, TheTarget->getName());
1627
1628 for (StringRef Opt : DisassemblerOptions)
1629 if (!IP->applyTargetSpecificCLOption(Opt))
1630 reportError(Obj->getFileName(),
1631 "Unrecognized disassembler option: " + Opt);
1632
1633 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(),
1634 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP,
1635 SP, InlineRelocs);
1636 }
1637
printRelocations(const ObjectFile * Obj)1638 void objdump::printRelocations(const ObjectFile *Obj) {
1639 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 :
1640 "%08" PRIx64;
1641 // Regular objdump doesn't print relocations in non-relocatable object
1642 // files.
1643 if (!Obj->isRelocatableObject())
1644 return;
1645
1646 // Build a mapping from relocation target to a vector of relocation
1647 // sections. Usually, there is an only one relocation section for
1648 // each relocated section.
1649 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec;
1650 uint64_t Ndx;
1651 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) {
1652 if (Section.relocation_begin() == Section.relocation_end())
1653 continue;
1654 Expected<section_iterator> SecOrErr = Section.getRelocatedSection();
1655 if (!SecOrErr)
1656 reportError(Obj->getFileName(),
1657 "section (" + Twine(Ndx) +
1658 "): unable to get a relocation target: " +
1659 toString(SecOrErr.takeError()));
1660 SecToRelSec[**SecOrErr].push_back(Section);
1661 }
1662
1663 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) {
1664 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName());
1665 outs() << "\nRELOCATION RECORDS FOR [" << SecName << "]:\n";
1666 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8);
1667 uint32_t TypePadding = 24;
1668 outs() << left_justify("OFFSET", OffsetPadding) << " "
1669 << left_justify("TYPE", TypePadding) << " "
1670 << "VALUE\n";
1671
1672 for (SectionRef Section : P.second) {
1673 for (const RelocationRef &Reloc : Section.relocations()) {
1674 uint64_t Address = Reloc.getOffset();
1675 SmallString<32> RelocName;
1676 SmallString<32> ValueStr;
1677 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc))
1678 continue;
1679 Reloc.getTypeName(RelocName);
1680 if (Error E = getRelocationValueString(Reloc, ValueStr))
1681 reportError(std::move(E), Obj->getFileName());
1682
1683 outs() << format(Fmt.data(), Address) << " "
1684 << left_justify(RelocName, TypePadding) << " " << ValueStr
1685 << "\n";
1686 }
1687 }
1688 }
1689 }
1690
printDynamicRelocations(const ObjectFile * Obj)1691 void objdump::printDynamicRelocations(const ObjectFile *Obj) {
1692 // For the moment, this option is for ELF only
1693 if (!Obj->isELF())
1694 return;
1695
1696 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj);
1697 if (!Elf || Elf->getEType() != ELF::ET_DYN) {
1698 reportError(Obj->getFileName(), "not a dynamic object");
1699 return;
1700 }
1701
1702 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections();
1703 if (DynRelSec.empty())
1704 return;
1705
1706 outs() << "DYNAMIC RELOCATION RECORDS\n";
1707 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1708 for (const SectionRef &Section : DynRelSec)
1709 for (const RelocationRef &Reloc : Section.relocations()) {
1710 uint64_t Address = Reloc.getOffset();
1711 SmallString<32> RelocName;
1712 SmallString<32> ValueStr;
1713 Reloc.getTypeName(RelocName);
1714 if (Error E = getRelocationValueString(Reloc, ValueStr))
1715 reportError(std::move(E), Obj->getFileName());
1716 outs() << format(Fmt.data(), Address) << " " << RelocName << " "
1717 << ValueStr << "\n";
1718 }
1719 }
1720
1721 // Returns true if we need to show LMA column when dumping section headers. We
1722 // show it only when the platform is ELF and either we have at least one section
1723 // whose VMA and LMA are different and/or when --show-lma flag is used.
shouldDisplayLMA(const ObjectFile * Obj)1724 static bool shouldDisplayLMA(const ObjectFile *Obj) {
1725 if (!Obj->isELF())
1726 return false;
1727 for (const SectionRef &S : ToolSectionFilter(*Obj))
1728 if (S.getAddress() != getELFSectionLMA(S))
1729 return true;
1730 return ShowLMA;
1731 }
1732
getMaxSectionNameWidth(const ObjectFile * Obj)1733 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) {
1734 // Default column width for names is 13 even if no names are that long.
1735 size_t MaxWidth = 13;
1736 for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1737 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1738 MaxWidth = std::max(MaxWidth, Name.size());
1739 }
1740 return MaxWidth;
1741 }
1742
printSectionHeaders(const ObjectFile * Obj)1743 void objdump::printSectionHeaders(const ObjectFile *Obj) {
1744 size_t NameWidth = getMaxSectionNameWidth(Obj);
1745 size_t AddressWidth = 2 * Obj->getBytesInAddress();
1746 bool HasLMAColumn = shouldDisplayLMA(Obj);
1747 outs() << "\nSections:\n";
1748 if (HasLMAColumn)
1749 outs() << "Idx " << left_justify("Name", NameWidth) << " Size "
1750 << left_justify("VMA", AddressWidth) << " "
1751 << left_justify("LMA", AddressWidth) << " Type\n";
1752 else
1753 outs() << "Idx " << left_justify("Name", NameWidth) << " Size "
1754 << left_justify("VMA", AddressWidth) << " Type\n";
1755
1756 uint64_t Idx;
1757 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) {
1758 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1759 uint64_t VMA = Section.getAddress();
1760 if (shouldAdjustVA(Section))
1761 VMA += AdjustVMA;
1762
1763 uint64_t Size = Section.getSize();
1764
1765 std::string Type = Section.isText() ? "TEXT" : "";
1766 if (Section.isData())
1767 Type += Type.empty() ? "DATA" : " DATA";
1768 if (Section.isBSS())
1769 Type += Type.empty() ? "BSS" : " BSS";
1770
1771 if (HasLMAColumn)
1772 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1773 Name.str().c_str(), Size)
1774 << format_hex_no_prefix(VMA, AddressWidth) << " "
1775 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth)
1776 << " " << Type << "\n";
1777 else
1778 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth,
1779 Name.str().c_str(), Size)
1780 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n";
1781 }
1782 }
1783
printSectionContents(const ObjectFile * Obj)1784 void objdump::printSectionContents(const ObjectFile *Obj) {
1785 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj);
1786
1787 for (const SectionRef &Section : ToolSectionFilter(*Obj)) {
1788 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName());
1789 uint64_t BaseAddr = Section.getAddress();
1790 uint64_t Size = Section.getSize();
1791 if (!Size)
1792 continue;
1793
1794 outs() << "Contents of section ";
1795 StringRef SegmentName = getSegmentName(MachO, Section);
1796 if (!SegmentName.empty())
1797 outs() << SegmentName << ",";
1798 outs() << Name << ":\n";
1799 if (Section.isBSS()) {
1800 outs() << format("<skipping contents of bss section at [%04" PRIx64
1801 ", %04" PRIx64 ")>\n",
1802 BaseAddr, BaseAddr + Size);
1803 continue;
1804 }
1805
1806 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName());
1807
1808 // Dump out the content as hex and printable ascii characters.
1809 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) {
1810 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr);
1811 // Dump line of hex.
1812 for (std::size_t I = 0; I < 16; ++I) {
1813 if (I != 0 && I % 4 == 0)
1814 outs() << ' ';
1815 if (Addr + I < End)
1816 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true)
1817 << hexdigit(Contents[Addr + I] & 0xF, true);
1818 else
1819 outs() << " ";
1820 }
1821 // Print ascii.
1822 outs() << " ";
1823 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) {
1824 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF))
1825 outs() << Contents[Addr + I];
1826 else
1827 outs() << ".";
1828 }
1829 outs() << "\n";
1830 }
1831 }
1832 }
1833
printSymbolTable(const ObjectFile * O,StringRef ArchiveName,StringRef ArchitectureName,bool DumpDynamic)1834 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName,
1835 StringRef ArchitectureName, bool DumpDynamic) {
1836 if (O->isCOFF() && !DumpDynamic) {
1837 outs() << "\nSYMBOL TABLE:\n";
1838 printCOFFSymbolTable(cast<const COFFObjectFile>(O));
1839 return;
1840 }
1841
1842 const StringRef FileName = O->getFileName();
1843
1844 if (!DumpDynamic) {
1845 outs() << "\nSYMBOL TABLE:\n";
1846 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I)
1847 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic);
1848 return;
1849 }
1850
1851 outs() << "\nDYNAMIC SYMBOL TABLE:\n";
1852 if (!O->isELF()) {
1853 reportWarning(
1854 "this operation is not currently supported for this file format",
1855 FileName);
1856 return;
1857 }
1858
1859 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O);
1860 for (auto I = ELF->getDynamicSymbolIterators().begin();
1861 I != ELF->getDynamicSymbolIterators().end(); ++I)
1862 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic);
1863 }
1864
printSymbol(const ObjectFile * O,const SymbolRef & Symbol,StringRef FileName,StringRef ArchiveName,StringRef ArchitectureName,bool DumpDynamic)1865 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol,
1866 StringRef FileName, StringRef ArchiveName,
1867 StringRef ArchitectureName, bool DumpDynamic) {
1868 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O);
1869 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName,
1870 ArchitectureName);
1871 if ((Address < StartAddress) || (Address > StopAddress))
1872 return;
1873 SymbolRef::Type Type =
1874 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName);
1875 uint32_t Flags =
1876 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName);
1877
1878 // Don't ask a Mach-O STAB symbol for its section unless you know that
1879 // STAB symbol's section field refers to a valid section index. Otherwise
1880 // the symbol may error trying to load a section that does not exist.
1881 bool IsSTAB = false;
1882 if (MachO) {
1883 DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1884 uint8_t NType =
1885 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type
1886 : MachO->getSymbolTableEntry(SymDRI).n_type);
1887 if (NType & MachO::N_STAB)
1888 IsSTAB = true;
1889 }
1890 section_iterator Section = IsSTAB
1891 ? O->section_end()
1892 : unwrapOrError(Symbol.getSection(), FileName,
1893 ArchiveName, ArchitectureName);
1894
1895 StringRef Name;
1896 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) {
1897 if (Expected<StringRef> NameOrErr = Section->getName())
1898 Name = *NameOrErr;
1899 else
1900 consumeError(NameOrErr.takeError());
1901
1902 } else {
1903 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName,
1904 ArchitectureName);
1905 }
1906
1907 bool Global = Flags & SymbolRef::SF_Global;
1908 bool Weak = Flags & SymbolRef::SF_Weak;
1909 bool Absolute = Flags & SymbolRef::SF_Absolute;
1910 bool Common = Flags & SymbolRef::SF_Common;
1911 bool Hidden = Flags & SymbolRef::SF_Hidden;
1912
1913 char GlobLoc = ' ';
1914 if ((Section != O->section_end() || Absolute) && !Weak)
1915 GlobLoc = Global ? 'g' : 'l';
1916 char IFunc = ' ';
1917 if (O->isELF()) {
1918 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC)
1919 IFunc = 'i';
1920 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE)
1921 GlobLoc = 'u';
1922 }
1923
1924 char Debug = ' ';
1925 if (DumpDynamic)
1926 Debug = 'D';
1927 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File)
1928 Debug = 'd';
1929
1930 char FileFunc = ' ';
1931 if (Type == SymbolRef::ST_File)
1932 FileFunc = 'f';
1933 else if (Type == SymbolRef::ST_Function)
1934 FileFunc = 'F';
1935 else if (Type == SymbolRef::ST_Data)
1936 FileFunc = 'O';
1937
1938 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
1939
1940 outs() << format(Fmt, Address) << " "
1941 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' '
1942 << (Weak ? 'w' : ' ') // Weak?
1943 << ' ' // Constructor. Not supported yet.
1944 << ' ' // Warning. Not supported yet.
1945 << IFunc // Indirect reference to another symbol.
1946 << Debug // Debugging (d) or dynamic (D) symbol.
1947 << FileFunc // Name of function (F), file (f) or object (O).
1948 << ' ';
1949 if (Absolute) {
1950 outs() << "*ABS*";
1951 } else if (Common) {
1952 outs() << "*COM*";
1953 } else if (Section == O->section_end()) {
1954 outs() << "*UND*";
1955 } else {
1956 StringRef SegmentName = getSegmentName(MachO, *Section);
1957 if (!SegmentName.empty())
1958 outs() << SegmentName << ",";
1959 StringRef SectionName = unwrapOrError(Section->getName(), FileName);
1960 outs() << SectionName;
1961 }
1962
1963 if (Common || O->isELF()) {
1964 uint64_t Val =
1965 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize();
1966 outs() << '\t' << format(Fmt, Val);
1967 }
1968
1969 if (O->isELF()) {
1970 uint8_t Other = ELFSymbolRef(Symbol).getOther();
1971 switch (Other) {
1972 case ELF::STV_DEFAULT:
1973 break;
1974 case ELF::STV_INTERNAL:
1975 outs() << " .internal";
1976 break;
1977 case ELF::STV_HIDDEN:
1978 outs() << " .hidden";
1979 break;
1980 case ELF::STV_PROTECTED:
1981 outs() << " .protected";
1982 break;
1983 default:
1984 outs() << format(" 0x%02x", Other);
1985 break;
1986 }
1987 } else if (Hidden) {
1988 outs() << " .hidden";
1989 }
1990
1991 if (Demangle)
1992 outs() << ' ' << demangle(std::string(Name)) << '\n';
1993 else
1994 outs() << ' ' << Name << '\n';
1995 }
1996
printUnwindInfo(const ObjectFile * O)1997 static void printUnwindInfo(const ObjectFile *O) {
1998 outs() << "Unwind info:\n\n";
1999
2000 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O))
2001 printCOFFUnwindInfo(Coff);
2002 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O))
2003 printMachOUnwindInfo(MachO);
2004 else
2005 // TODO: Extract DWARF dump tool to objdump.
2006 WithColor::error(errs(), ToolName)
2007 << "This operation is only currently supported "
2008 "for COFF and MachO object files.\n";
2009 }
2010
2011 /// Dump the raw contents of the __clangast section so the output can be piped
2012 /// into llvm-bcanalyzer.
printRawClangAST(const ObjectFile * Obj)2013 static void printRawClangAST(const ObjectFile *Obj) {
2014 if (outs().is_displayed()) {
2015 WithColor::error(errs(), ToolName)
2016 << "The -raw-clang-ast option will dump the raw binary contents of "
2017 "the clang ast section.\n"
2018 "Please redirect the output to a file or another program such as "
2019 "llvm-bcanalyzer.\n";
2020 return;
2021 }
2022
2023 StringRef ClangASTSectionName("__clangast");
2024 if (Obj->isCOFF()) {
2025 ClangASTSectionName = "clangast";
2026 }
2027
2028 Optional<object::SectionRef> ClangASTSection;
2029 for (auto Sec : ToolSectionFilter(*Obj)) {
2030 StringRef Name;
2031 if (Expected<StringRef> NameOrErr = Sec.getName())
2032 Name = *NameOrErr;
2033 else
2034 consumeError(NameOrErr.takeError());
2035
2036 if (Name == ClangASTSectionName) {
2037 ClangASTSection = Sec;
2038 break;
2039 }
2040 }
2041 if (!ClangASTSection)
2042 return;
2043
2044 StringRef ClangASTContents = unwrapOrError(
2045 ClangASTSection.getValue().getContents(), Obj->getFileName());
2046 outs().write(ClangASTContents.data(), ClangASTContents.size());
2047 }
2048
printFaultMaps(const ObjectFile * Obj)2049 static void printFaultMaps(const ObjectFile *Obj) {
2050 StringRef FaultMapSectionName;
2051
2052 if (Obj->isELF()) {
2053 FaultMapSectionName = ".llvm_faultmaps";
2054 } else if (Obj->isMachO()) {
2055 FaultMapSectionName = "__llvm_faultmaps";
2056 } else {
2057 WithColor::error(errs(), ToolName)
2058 << "This operation is only currently supported "
2059 "for ELF and Mach-O executable files.\n";
2060 return;
2061 }
2062
2063 Optional<object::SectionRef> FaultMapSection;
2064
2065 for (auto Sec : ToolSectionFilter(*Obj)) {
2066 StringRef Name;
2067 if (Expected<StringRef> NameOrErr = Sec.getName())
2068 Name = *NameOrErr;
2069 else
2070 consumeError(NameOrErr.takeError());
2071
2072 if (Name == FaultMapSectionName) {
2073 FaultMapSection = Sec;
2074 break;
2075 }
2076 }
2077
2078 outs() << "FaultMap table:\n";
2079
2080 if (!FaultMapSection.hasValue()) {
2081 outs() << "<not found>\n";
2082 return;
2083 }
2084
2085 StringRef FaultMapContents =
2086 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName());
2087 FaultMapParser FMP(FaultMapContents.bytes_begin(),
2088 FaultMapContents.bytes_end());
2089
2090 outs() << FMP;
2091 }
2092
printPrivateFileHeaders(const ObjectFile * O,bool OnlyFirst)2093 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) {
2094 if (O->isELF()) {
2095 printELFFileHeader(O);
2096 printELFDynamicSection(O);
2097 printELFSymbolVersionInfo(O);
2098 return;
2099 }
2100 if (O->isCOFF())
2101 return printCOFFFileHeader(O);
2102 if (O->isWasm())
2103 return printWasmFileHeader(O);
2104 if (O->isMachO()) {
2105 printMachOFileHeader(O);
2106 if (!OnlyFirst)
2107 printMachOLoadCommands(O);
2108 return;
2109 }
2110 reportError(O->getFileName(), "Invalid/Unsupported object file format");
2111 }
2112
printFileHeaders(const ObjectFile * O)2113 static void printFileHeaders(const ObjectFile *O) {
2114 if (!O->isELF() && !O->isCOFF())
2115 reportError(O->getFileName(), "Invalid/Unsupported object file format");
2116
2117 Triple::ArchType AT = O->getArch();
2118 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n";
2119 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName());
2120
2121 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64;
2122 outs() << "start address: "
2123 << "0x" << format(Fmt.data(), Address) << "\n";
2124 }
2125
printArchiveChild(StringRef Filename,const Archive::Child & C)2126 static void printArchiveChild(StringRef Filename, const Archive::Child &C) {
2127 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode();
2128 if (!ModeOrErr) {
2129 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n";
2130 consumeError(ModeOrErr.takeError());
2131 return;
2132 }
2133 sys::fs::perms Mode = ModeOrErr.get();
2134 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
2135 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
2136 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
2137 outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
2138 outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
2139 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
2140 outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
2141 outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
2142 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
2143
2144 outs() << " ";
2145
2146 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename),
2147 unwrapOrError(C.getGID(), Filename),
2148 unwrapOrError(C.getRawSize(), Filename));
2149
2150 StringRef RawLastModified = C.getRawLastModified();
2151 unsigned Seconds;
2152 if (RawLastModified.getAsInteger(10, Seconds))
2153 outs() << "(date: \"" << RawLastModified
2154 << "\" contains non-decimal chars) ";
2155 else {
2156 // Since ctime(3) returns a 26 character string of the form:
2157 // "Sun Sep 16 01:03:52 1973\n\0"
2158 // just print 24 characters.
2159 time_t t = Seconds;
2160 outs() << format("%.24s ", ctime(&t));
2161 }
2162
2163 StringRef Name = "";
2164 Expected<StringRef> NameOrErr = C.getName();
2165 if (!NameOrErr) {
2166 consumeError(NameOrErr.takeError());
2167 Name = unwrapOrError(C.getRawName(), Filename);
2168 } else {
2169 Name = NameOrErr.get();
2170 }
2171 outs() << Name << "\n";
2172 }
2173
2174 // For ELF only now.
shouldWarnForInvalidStartStopAddress(ObjectFile * Obj)2175 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) {
2176 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) {
2177 if (Elf->getEType() != ELF::ET_REL)
2178 return true;
2179 }
2180 return false;
2181 }
2182
checkForInvalidStartStopAddress(ObjectFile * Obj,uint64_t Start,uint64_t Stop)2183 static void checkForInvalidStartStopAddress(ObjectFile *Obj,
2184 uint64_t Start, uint64_t Stop) {
2185 if (!shouldWarnForInvalidStartStopAddress(Obj))
2186 return;
2187
2188 for (const SectionRef &Section : Obj->sections())
2189 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) {
2190 uint64_t BaseAddr = Section.getAddress();
2191 uint64_t Size = Section.getSize();
2192 if ((Start < BaseAddr + Size) && Stop > BaseAddr)
2193 return;
2194 }
2195
2196 if (!HasStartAddressFlag)
2197 reportWarning("no section has address less than 0x" +
2198 Twine::utohexstr(Stop) + " specified by --stop-address",
2199 Obj->getFileName());
2200 else if (!HasStopAddressFlag)
2201 reportWarning("no section has address greater than or equal to 0x" +
2202 Twine::utohexstr(Start) + " specified by --start-address",
2203 Obj->getFileName());
2204 else
2205 reportWarning("no section overlaps the range [0x" +
2206 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) +
2207 ") specified by --start-address/--stop-address",
2208 Obj->getFileName());
2209 }
2210
dumpObject(ObjectFile * O,const Archive * A=nullptr,const Archive::Child * C=nullptr)2211 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
2212 const Archive::Child *C = nullptr) {
2213 // Avoid other output when using a raw option.
2214 if (!RawClangAST) {
2215 outs() << '\n';
2216 if (A)
2217 outs() << A->getFileName() << "(" << O->getFileName() << ")";
2218 else
2219 outs() << O->getFileName();
2220 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n";
2221 }
2222
2223 if (HasStartAddressFlag || HasStopAddressFlag)
2224 checkForInvalidStartStopAddress(O, StartAddress, StopAddress);
2225
2226 // Note: the order here matches GNU objdump for compatability.
2227 StringRef ArchiveName = A ? A->getFileName() : "";
2228 if (ArchiveHeaders && !MachOOpt && C)
2229 printArchiveChild(ArchiveName, *C);
2230 if (FileHeaders)
2231 printFileHeaders(O);
2232 if (PrivateHeaders || FirstPrivateHeader)
2233 printPrivateFileHeaders(O, FirstPrivateHeader);
2234 if (SectionHeaders)
2235 printSectionHeaders(O);
2236 if (SymbolTable)
2237 printSymbolTable(O, ArchiveName);
2238 if (DynamicSymbolTable)
2239 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"",
2240 /*DumpDynamic=*/true);
2241 if (DwarfDumpType != DIDT_Null) {
2242 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O);
2243 // Dump the complete DWARF structure.
2244 DIDumpOptions DumpOpts;
2245 DumpOpts.DumpType = DwarfDumpType;
2246 DICtx->dump(outs(), DumpOpts);
2247 }
2248 if (Relocations && !Disassemble)
2249 printRelocations(O);
2250 if (DynamicRelocations)
2251 printDynamicRelocations(O);
2252 if (SectionContents)
2253 printSectionContents(O);
2254 if (Disassemble)
2255 disassembleObject(O, Relocations);
2256 if (UnwindInfo)
2257 printUnwindInfo(O);
2258
2259 // Mach-O specific options:
2260 if (ExportsTrie)
2261 printExportsTrie(O);
2262 if (Rebase)
2263 printRebaseTable(O);
2264 if (Bind)
2265 printBindTable(O);
2266 if (LazyBind)
2267 printLazyBindTable(O);
2268 if (WeakBind)
2269 printWeakBindTable(O);
2270
2271 // Other special sections:
2272 if (RawClangAST)
2273 printRawClangAST(O);
2274 if (FaultMapSection)
2275 printFaultMaps(O);
2276 }
2277
dumpObject(const COFFImportFile * I,const Archive * A,const Archive::Child * C=nullptr)2278 static void dumpObject(const COFFImportFile *I, const Archive *A,
2279 const Archive::Child *C = nullptr) {
2280 StringRef ArchiveName = A ? A->getFileName() : "";
2281
2282 // Avoid other output when using a raw option.
2283 if (!RawClangAST)
2284 outs() << '\n'
2285 << ArchiveName << "(" << I->getFileName() << ")"
2286 << ":\tfile format COFF-import-file"
2287 << "\n\n";
2288
2289 if (ArchiveHeaders && !MachOOpt && C)
2290 printArchiveChild(ArchiveName, *C);
2291 if (SymbolTable)
2292 printCOFFSymbolTable(I);
2293 }
2294
2295 /// Dump each object file in \a a;
dumpArchive(const Archive * A)2296 static void dumpArchive(const Archive *A) {
2297 Error Err = Error::success();
2298 unsigned I = -1;
2299 for (auto &C : A->children(Err)) {
2300 ++I;
2301 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
2302 if (!ChildOrErr) {
2303 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError()))
2304 reportError(std::move(E), getFileNameForError(C, I), A->getFileName());
2305 continue;
2306 }
2307 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get()))
2308 dumpObject(O, A, &C);
2309 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get()))
2310 dumpObject(I, A, &C);
2311 else
2312 reportError(errorCodeToError(object_error::invalid_file_type),
2313 A->getFileName());
2314 }
2315 if (Err)
2316 reportError(std::move(Err), A->getFileName());
2317 }
2318
2319 /// Open file and figure out how to dump it.
dumpInput(StringRef file)2320 static void dumpInput(StringRef file) {
2321 // If we are using the Mach-O specific object file parser, then let it parse
2322 // the file and process the command line options. So the -arch flags can
2323 // be used to select specific slices, etc.
2324 if (MachOOpt) {
2325 parseInputMachO(file);
2326 return;
2327 }
2328
2329 // Attempt to open the binary.
2330 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file);
2331 Binary &Binary = *OBinary.getBinary();
2332
2333 if (Archive *A = dyn_cast<Archive>(&Binary))
2334 dumpArchive(A);
2335 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary))
2336 dumpObject(O);
2337 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary))
2338 parseInputMachO(UB);
2339 else
2340 reportError(errorCodeToError(object_error::invalid_file_type), file);
2341 }
2342
2343 template <typename T>
parseIntArg(const llvm::opt::InputArgList & InputArgs,int ID,T & Value)2344 static void parseIntArg(const llvm::opt::InputArgList &InputArgs, int ID,
2345 T &Value) {
2346 if (const opt::Arg *A = InputArgs.getLastArg(ID)) {
2347 StringRef V(A->getValue());
2348 if (!llvm::to_integer(V, Value, 0)) {
2349 reportCmdLineError(A->getSpelling() +
2350 ": expected a non-negative integer, but got '" + V +
2351 "'");
2352 }
2353 }
2354 }
2355
2356 static std::vector<std::string>
commaSeparatedValues(const llvm::opt::InputArgList & InputArgs,int ID)2357 commaSeparatedValues(const llvm::opt::InputArgList &InputArgs, int ID) {
2358 std::vector<std::string> Values;
2359 for (StringRef Value : InputArgs.getAllArgValues(ID)) {
2360 llvm::SmallVector<StringRef, 2> SplitValues;
2361 llvm::SplitString(Value, SplitValues, ",");
2362 for (StringRef SplitValue : SplitValues)
2363 Values.push_back(SplitValue.str());
2364 }
2365 return Values;
2366 }
2367
parseOtoolOptions(const llvm::opt::InputArgList & InputArgs)2368 static void parseOtoolOptions(const llvm::opt::InputArgList &InputArgs) {
2369 MachOOpt = true;
2370 FullLeadingAddr = true;
2371 PrintImmHex = true;
2372
2373 ArchName = InputArgs.getLastArgValue(OTOOL_arch).str();
2374 LinkOptHints = InputArgs.hasArg(OTOOL_C);
2375 if (InputArgs.hasArg(OTOOL_d))
2376 FilterSections.push_back("__DATA,__data");
2377 DylibId = InputArgs.hasArg(OTOOL_D);
2378 UniversalHeaders = InputArgs.hasArg(OTOOL_f);
2379 DataInCode = InputArgs.hasArg(OTOOL_G);
2380 FirstPrivateHeader = InputArgs.hasArg(OTOOL_h);
2381 IndirectSymbols = InputArgs.hasArg(OTOOL_I);
2382 ShowRawInsn = InputArgs.hasArg(OTOOL_j);
2383 PrivateHeaders = InputArgs.hasArg(OTOOL_l);
2384 DylibsUsed = InputArgs.hasArg(OTOOL_L);
2385 MCPU = InputArgs.getLastArgValue(OTOOL_mcpu_EQ).str();
2386 ObjcMetaData = InputArgs.hasArg(OTOOL_o);
2387 DisSymName = InputArgs.getLastArgValue(OTOOL_p).str();
2388 InfoPlist = InputArgs.hasArg(OTOOL_P);
2389 Relocations = InputArgs.hasArg(OTOOL_r);
2390 if (const Arg *A = InputArgs.getLastArg(OTOOL_s)) {
2391 auto Filter = (A->getValue(0) + StringRef(",") + A->getValue(1)).str();
2392 FilterSections.push_back(Filter);
2393 }
2394 if (InputArgs.hasArg(OTOOL_t))
2395 FilterSections.push_back("__TEXT,__text");
2396 Verbose = InputArgs.hasArg(OTOOL_v) || InputArgs.hasArg(OTOOL_V) ||
2397 InputArgs.hasArg(OTOOL_o);
2398 SymbolicOperands = InputArgs.hasArg(OTOOL_V);
2399 if (InputArgs.hasArg(OTOOL_x))
2400 FilterSections.push_back(",__text");
2401 LeadingAddr = LeadingHeaders = !InputArgs.hasArg(OTOOL_X);
2402
2403 InputFilenames = InputArgs.getAllArgValues(OTOOL_INPUT);
2404 if (InputFilenames.empty())
2405 reportCmdLineError("no input file");
2406
2407 for (const Arg *A : InputArgs) {
2408 const Option &O = A->getOption();
2409 if (O.getGroup().isValid() && O.getGroup().getID() == OTOOL_grp_obsolete) {
2410 reportCmdLineWarning(O.getPrefixedName() +
2411 " is obsolete and not implemented");
2412 }
2413 }
2414 }
2415
parseObjdumpOptions(const llvm::opt::InputArgList & InputArgs)2416 static void parseObjdumpOptions(const llvm::opt::InputArgList &InputArgs) {
2417 parseIntArg(InputArgs, OBJDUMP_adjust_vma_EQ, AdjustVMA);
2418 AllHeaders = InputArgs.hasArg(OBJDUMP_all_headers);
2419 ArchName = InputArgs.getLastArgValue(OBJDUMP_arch_name_EQ).str();
2420 ArchiveHeaders = InputArgs.hasArg(OBJDUMP_archive_headers);
2421 Demangle = InputArgs.hasArg(OBJDUMP_demangle);
2422 Disassemble = InputArgs.hasArg(OBJDUMP_disassemble);
2423 DisassembleAll = InputArgs.hasArg(OBJDUMP_disassemble_all);
2424 SymbolDescription = InputArgs.hasArg(OBJDUMP_symbol_description);
2425 DisassembleSymbols =
2426 commaSeparatedValues(InputArgs, OBJDUMP_disassemble_symbols_EQ);
2427 DisassembleZeroes = InputArgs.hasArg(OBJDUMP_disassemble_zeroes);
2428 if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_dwarf_EQ)) {
2429 DwarfDumpType =
2430 StringSwitch<DIDumpType>(A->getValue()).Case("frames", DIDT_DebugFrame);
2431 }
2432 DynamicRelocations = InputArgs.hasArg(OBJDUMP_dynamic_reloc);
2433 FaultMapSection = InputArgs.hasArg(OBJDUMP_fault_map_section);
2434 FileHeaders = InputArgs.hasArg(OBJDUMP_file_headers);
2435 SectionContents = InputArgs.hasArg(OBJDUMP_full_contents);
2436 PrintLines = InputArgs.hasArg(OBJDUMP_line_numbers);
2437 InputFilenames = InputArgs.getAllArgValues(OBJDUMP_INPUT);
2438 MachOOpt = InputArgs.hasArg(OBJDUMP_macho);
2439 MCPU = InputArgs.getLastArgValue(OBJDUMP_mcpu_EQ).str();
2440 MAttrs = commaSeparatedValues(InputArgs, OBJDUMP_mattr_EQ);
2441 ShowRawInsn = !InputArgs.hasArg(OBJDUMP_no_show_raw_insn);
2442 LeadingAddr = !InputArgs.hasArg(OBJDUMP_no_leading_addr);
2443 RawClangAST = InputArgs.hasArg(OBJDUMP_raw_clang_ast);
2444 Relocations = InputArgs.hasArg(OBJDUMP_reloc);
2445 PrintImmHex =
2446 InputArgs.hasFlag(OBJDUMP_print_imm_hex, OBJDUMP_no_print_imm_hex, false);
2447 PrivateHeaders = InputArgs.hasArg(OBJDUMP_private_headers);
2448 FilterSections = InputArgs.getAllArgValues(OBJDUMP_section_EQ);
2449 SectionHeaders = InputArgs.hasArg(OBJDUMP_section_headers);
2450 ShowLMA = InputArgs.hasArg(OBJDUMP_show_lma);
2451 PrintSource = InputArgs.hasArg(OBJDUMP_source);
2452 parseIntArg(InputArgs, OBJDUMP_start_address_EQ, StartAddress);
2453 HasStartAddressFlag = InputArgs.hasArg(OBJDUMP_start_address_EQ);
2454 parseIntArg(InputArgs, OBJDUMP_stop_address_EQ, StopAddress);
2455 HasStopAddressFlag = InputArgs.hasArg(OBJDUMP_stop_address_EQ);
2456 SymbolTable = InputArgs.hasArg(OBJDUMP_syms);
2457 SymbolizeOperands = InputArgs.hasArg(OBJDUMP_symbolize_operands);
2458 DynamicSymbolTable = InputArgs.hasArg(OBJDUMP_dynamic_syms);
2459 TripleName = InputArgs.getLastArgValue(OBJDUMP_triple_EQ).str();
2460 UnwindInfo = InputArgs.hasArg(OBJDUMP_unwind_info);
2461 Wide = InputArgs.hasArg(OBJDUMP_wide);
2462 Prefix = InputArgs.getLastArgValue(OBJDUMP_prefix).str();
2463 parseIntArg(InputArgs, OBJDUMP_prefix_strip, PrefixStrip);
2464 if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_debug_vars_EQ)) {
2465 DbgVariables = StringSwitch<DebugVarsFormat>(A->getValue())
2466 .Case("ascii", DVASCII)
2467 .Case("unicode", DVUnicode);
2468 }
2469 parseIntArg(InputArgs, OBJDUMP_debug_vars_indent_EQ, DbgIndent);
2470
2471 parseMachOOptions(InputArgs);
2472
2473 // Parse -M (--disassembler-options) and deprecated
2474 // --x86-asm-syntax={att,intel}.
2475 //
2476 // Note, for x86, the asm dialect (AssemblerDialect) is initialized when the
2477 // MCAsmInfo is constructed. MCInstPrinter::applyTargetSpecificCLOption is
2478 // called too late. For now we have to use the internal cl::opt option.
2479 const char *AsmSyntax = nullptr;
2480 for (const auto *A : InputArgs.filtered(OBJDUMP_disassembler_options_EQ,
2481 OBJDUMP_x86_asm_syntax_att,
2482 OBJDUMP_x86_asm_syntax_intel)) {
2483 switch (A->getOption().getID()) {
2484 case OBJDUMP_x86_asm_syntax_att:
2485 AsmSyntax = "--x86-asm-syntax=att";
2486 continue;
2487 case OBJDUMP_x86_asm_syntax_intel:
2488 AsmSyntax = "--x86-asm-syntax=intel";
2489 continue;
2490 }
2491
2492 SmallVector<StringRef, 2> Values;
2493 llvm::SplitString(A->getValue(), Values, ",");
2494 for (StringRef V : Values) {
2495 if (V == "att")
2496 AsmSyntax = "--x86-asm-syntax=att";
2497 else if (V == "intel")
2498 AsmSyntax = "--x86-asm-syntax=intel";
2499 else
2500 DisassemblerOptions.push_back(V.str());
2501 }
2502 }
2503 if (AsmSyntax) {
2504 const char *Argv[] = {"llvm-objdump", AsmSyntax};
2505 llvm::cl::ParseCommandLineOptions(2, Argv);
2506 }
2507
2508 // objdump defaults to a.out if no filenames specified.
2509 if (InputFilenames.empty())
2510 InputFilenames.push_back("a.out");
2511 }
2512
main(int argc,char ** argv)2513 int main(int argc, char **argv) {
2514 using namespace llvm;
2515 InitLLVM X(argc, argv);
2516
2517 ToolName = argv[0];
2518 std::unique_ptr<CommonOptTable> T;
2519 OptSpecifier Unknown, HelpFlag, HelpHiddenFlag, VersionFlag;
2520
2521 StringRef Stem = sys::path::stem(ToolName);
2522 auto Is = [=](StringRef Tool) {
2523 // We need to recognize the following filenames:
2524 //
2525 // llvm-objdump -> objdump
2526 // llvm-otool-10.exe -> otool
2527 // powerpc64-unknown-freebsd13-objdump -> objdump
2528 auto I = Stem.rfind_lower(Tool);
2529 return I != StringRef::npos &&
2530 (I + Tool.size() == Stem.size() || !isAlnum(Stem[I + Tool.size()]));
2531 };
2532 if (Is("otool")) {
2533 T = std::make_unique<OtoolOptTable>();
2534 Unknown = OTOOL_UNKNOWN;
2535 HelpFlag = OTOOL_help;
2536 HelpHiddenFlag = OTOOL_help_hidden;
2537 VersionFlag = OTOOL_version;
2538 } else {
2539 T = std::make_unique<ObjdumpOptTable>();
2540 Unknown = OBJDUMP_UNKNOWN;
2541 HelpFlag = OBJDUMP_help;
2542 HelpHiddenFlag = OBJDUMP_help_hidden;
2543 VersionFlag = OBJDUMP_version;
2544 }
2545
2546 BumpPtrAllocator A;
2547 StringSaver Saver(A);
2548 opt::InputArgList InputArgs =
2549 T->parseArgs(argc, argv, Unknown, Saver,
2550 [&](StringRef Msg) { reportCmdLineError(Msg); });
2551
2552 if (InputArgs.size() == 0 || InputArgs.hasArg(HelpFlag)) {
2553 T->printHelp(ToolName);
2554 return 0;
2555 }
2556 if (InputArgs.hasArg(HelpHiddenFlag)) {
2557 T->printHelp(ToolName, /*show_hidden=*/true);
2558 return 0;
2559 }
2560
2561 // Initialize targets and assembly printers/parsers.
2562 InitializeAllTargetInfos();
2563 InitializeAllTargetMCs();
2564 InitializeAllDisassemblers();
2565
2566 if (InputArgs.hasArg(VersionFlag)) {
2567 cl::PrintVersionMessage();
2568 if (!Is("otool")) {
2569 outs() << '\n';
2570 TargetRegistry::printRegisteredTargetsForVersion(outs());
2571 }
2572 return 0;
2573 }
2574
2575 if (Is("otool"))
2576 parseOtoolOptions(InputArgs);
2577 else
2578 parseObjdumpOptions(InputArgs);
2579
2580 if (StartAddress >= StopAddress)
2581 reportCmdLineError("start address should be less than stop address");
2582
2583 // Removes trailing separators from prefix.
2584 while (!Prefix.empty() && sys::path::is_separator(Prefix.back()))
2585 Prefix.pop_back();
2586
2587 if (AllHeaders)
2588 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations =
2589 SectionHeaders = SymbolTable = true;
2590
2591 if (DisassembleAll || PrintSource || PrintLines ||
2592 !DisassembleSymbols.empty())
2593 Disassemble = true;
2594
2595 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null &&
2596 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST &&
2597 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable &&
2598 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection &&
2599 !(MachOOpt &&
2600 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie ||
2601 FirstPrivateHeader || FunctionStarts || IndirectSymbols || InfoPlist ||
2602 LazyBind || LinkOptHints || ObjcMetaData || Rebase || Rpaths ||
2603 UniversalHeaders || WeakBind || !FilterSections.empty()))) {
2604 T->printHelp(ToolName);
2605 return 2;
2606 }
2607
2608 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end());
2609
2610 llvm::for_each(InputFilenames, dumpInput);
2611
2612 warnOnNoMatchForSections();
2613
2614 return EXIT_SUCCESS;
2615 }
2616