1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This program is a utility that works like binutils "objdump", that is, it 11 // dumps out a plethora of information about an object file depending on the 12 // flags. 13 // 14 // The flags and output of this program should be near identical to those of 15 // binutils objdump. 16 // 17 //===----------------------------------------------------------------------===// 18 19 #include "llvm-objdump.h" 20 #include "llvm/ADT/Optional.h" 21 #include "llvm/ADT/STLExtras.h" 22 #include "llvm/ADT/StringExtras.h" 23 #include "llvm/ADT/StringSet.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/CodeGen/FaultMaps.h" 26 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 27 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 28 #include "llvm/MC/MCAsmInfo.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 32 #include "llvm/MC/MCInst.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrAnalysis.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCRegisterInfo.h" 38 #include "llvm/MC/MCSubtargetInfo.h" 39 #include "llvm/Object/Archive.h" 40 #include "llvm/Object/COFF.h" 41 #include "llvm/Object/COFFImportFile.h" 42 #include "llvm/Object/ELFObjectFile.h" 43 #include "llvm/Object/MachO.h" 44 #include "llvm/Object/ObjectFile.h" 45 #include "llvm/Object/Wasm.h" 46 #include "llvm/Support/Casting.h" 47 #include "llvm/Support/CommandLine.h" 48 #include "llvm/Support/Debug.h" 49 #include "llvm/Support/Errc.h" 50 #include "llvm/Support/FileSystem.h" 51 #include "llvm/Support/Format.h" 52 #include "llvm/Support/GraphWriter.h" 53 #include "llvm/Support/Host.h" 54 #include "llvm/Support/InitLLVM.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 #include "llvm/Support/SourceMgr.h" 57 #include "llvm/Support/TargetRegistry.h" 58 #include "llvm/Support/TargetSelect.h" 59 #include "llvm/Support/raw_ostream.h" 60 #include <algorithm> 61 #include <cctype> 62 #include <cstring> 63 #include <system_error> 64 #include <unordered_map> 65 #include <utility> 66 67 using namespace llvm; 68 using namespace object; 69 70 static cl::list<std::string> 71 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 72 73 cl::opt<bool> 74 llvm::Disassemble("disassemble", 75 cl::desc("Display assembler mnemonics for the machine instructions")); 76 static cl::alias 77 Disassembled("d", cl::desc("Alias for --disassemble"), 78 cl::aliasopt(Disassemble)); 79 80 cl::opt<bool> 81 llvm::DisassembleAll("disassemble-all", 82 cl::desc("Display assembler mnemonics for the machine instructions")); 83 static cl::alias 84 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 85 cl::aliasopt(DisassembleAll)); 86 87 static cl::list<std::string> 88 DisassembleFunctions("df", 89 cl::CommaSeparated, 90 cl::desc("List of functions to disassemble")); 91 static StringSet<> DisasmFuncsSet; 92 93 cl::opt<bool> 94 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 95 96 cl::opt<bool> 97 llvm::SectionContents("s", cl::desc("Display the content of each section")); 98 99 cl::opt<bool> 100 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 101 102 cl::opt<bool> 103 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 104 105 cl::opt<bool> 106 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 107 108 cl::opt<bool> 109 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 110 111 cl::opt<bool> 112 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 113 114 cl::opt<bool> 115 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 116 117 cl::opt<bool> 118 llvm::RawClangAST("raw-clang-ast", 119 cl::desc("Dump the raw binary contents of the clang AST section")); 120 121 static cl::opt<bool> 122 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 123 static cl::alias 124 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 125 126 cl::opt<std::string> 127 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 128 "see -version for available targets")); 129 130 cl::opt<std::string> 131 llvm::MCPU("mcpu", 132 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 133 cl::value_desc("cpu-name"), 134 cl::init("")); 135 136 cl::opt<std::string> 137 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 138 "see -version for available targets")); 139 140 cl::opt<bool> 141 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 142 "headers for each section.")); 143 static cl::alias 144 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 145 cl::aliasopt(SectionHeaders)); 146 static cl::alias 147 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 148 cl::aliasopt(SectionHeaders)); 149 150 cl::list<std::string> 151 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 152 "With -macho dump segment,section")); 153 cl::alias 154 static FilterSectionsj("j", cl::desc("Alias for --section"), 155 cl::aliasopt(llvm::FilterSections)); 156 157 cl::list<std::string> 158 llvm::MAttrs("mattr", 159 cl::CommaSeparated, 160 cl::desc("Target specific attributes"), 161 cl::value_desc("a1,+a2,-a3,...")); 162 163 cl::opt<bool> 164 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 165 "instructions, do not print " 166 "the instruction bytes.")); 167 cl::opt<bool> 168 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 169 170 cl::opt<bool> 171 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 172 173 static cl::alias 174 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 175 cl::aliasopt(UnwindInfo)); 176 177 cl::opt<bool> 178 llvm::PrivateHeaders("private-headers", 179 cl::desc("Display format specific file headers")); 180 181 cl::opt<bool> 182 llvm::FirstPrivateHeader("private-header", 183 cl::desc("Display only the first format specific file " 184 "header")); 185 186 static cl::alias 187 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 188 cl::aliasopt(PrivateHeaders)); 189 190 cl::opt<bool> 191 llvm::PrintImmHex("print-imm-hex", 192 cl::desc("Use hex format for immediate values")); 193 194 cl::opt<bool> PrintFaultMaps("fault-map-section", 195 cl::desc("Display contents of faultmap section")); 196 197 cl::opt<DIDumpType> llvm::DwarfDumpType( 198 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 199 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 200 201 cl::opt<bool> PrintSource( 202 "source", 203 cl::desc( 204 "Display source inlined with disassembly. Implies disassemble object")); 205 206 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 207 cl::aliasopt(PrintSource)); 208 209 cl::opt<bool> PrintLines("line-numbers", 210 cl::desc("Display source line numbers with " 211 "disassembly. Implies disassemble object")); 212 213 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 214 cl::aliasopt(PrintLines)); 215 216 cl::opt<unsigned long long> 217 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 218 cl::value_desc("address"), cl::init(0)); 219 cl::opt<unsigned long long> 220 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 221 cl::value_desc("address"), cl::init(UINT64_MAX)); 222 static StringRef ToolName; 223 224 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 225 226 namespace { 227 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 228 229 class SectionFilterIterator { 230 public: 231 SectionFilterIterator(FilterPredicate P, 232 llvm::object::section_iterator const &I, 233 llvm::object::section_iterator const &E) 234 : Predicate(std::move(P)), Iterator(I), End(E) { 235 ScanPredicate(); 236 } 237 const llvm::object::SectionRef &operator*() const { return *Iterator; } 238 SectionFilterIterator &operator++() { 239 ++Iterator; 240 ScanPredicate(); 241 return *this; 242 } 243 bool operator!=(SectionFilterIterator const &Other) const { 244 return Iterator != Other.Iterator; 245 } 246 247 private: 248 void ScanPredicate() { 249 while (Iterator != End && !Predicate(*Iterator)) { 250 ++Iterator; 251 } 252 } 253 FilterPredicate Predicate; 254 llvm::object::section_iterator Iterator; 255 llvm::object::section_iterator End; 256 }; 257 258 class SectionFilter { 259 public: 260 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 261 : Predicate(std::move(P)), Object(O) {} 262 SectionFilterIterator begin() { 263 return SectionFilterIterator(Predicate, Object.section_begin(), 264 Object.section_end()); 265 } 266 SectionFilterIterator end() { 267 return SectionFilterIterator(Predicate, Object.section_end(), 268 Object.section_end()); 269 } 270 271 private: 272 FilterPredicate Predicate; 273 llvm::object::ObjectFile const &Object; 274 }; 275 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 276 return SectionFilter( 277 [](llvm::object::SectionRef const &S) { 278 if (FilterSections.empty()) 279 return true; 280 llvm::StringRef String; 281 std::error_code error = S.getName(String); 282 if (error) 283 return false; 284 return is_contained(FilterSections, String); 285 }, 286 O); 287 } 288 } 289 290 void llvm::error(std::error_code EC) { 291 if (!EC) 292 return; 293 294 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 295 errs().flush(); 296 exit(1); 297 } 298 299 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 300 errs() << ToolName << ": " << Message << ".\n"; 301 errs().flush(); 302 exit(1); 303 } 304 305 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 306 Twine Message) { 307 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 308 exit(1); 309 } 310 311 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 312 std::error_code EC) { 313 assert(EC); 314 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 315 exit(1); 316 } 317 318 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 319 llvm::Error E) { 320 assert(E); 321 std::string Buf; 322 raw_string_ostream OS(Buf); 323 logAllUnhandledErrors(std::move(E), OS, ""); 324 OS.flush(); 325 errs() << ToolName << ": '" << File << "': " << Buf; 326 exit(1); 327 } 328 329 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 330 StringRef FileName, 331 llvm::Error E, 332 StringRef ArchitectureName) { 333 assert(E); 334 errs() << ToolName << ": "; 335 if (ArchiveName != "") 336 errs() << ArchiveName << "(" << FileName << ")"; 337 else 338 errs() << "'" << FileName << "'"; 339 if (!ArchitectureName.empty()) 340 errs() << " (for architecture " << ArchitectureName << ")"; 341 std::string Buf; 342 raw_string_ostream OS(Buf); 343 logAllUnhandledErrors(std::move(E), OS, ""); 344 OS.flush(); 345 errs() << ": " << Buf; 346 exit(1); 347 } 348 349 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 350 const object::Archive::Child &C, 351 llvm::Error E, 352 StringRef ArchitectureName) { 353 Expected<StringRef> NameOrErr = C.getName(); 354 // TODO: if we have a error getting the name then it would be nice to print 355 // the index of which archive member this is and or its offset in the 356 // archive instead of "???" as the name. 357 if (!NameOrErr) { 358 consumeError(NameOrErr.takeError()); 359 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 360 } else 361 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 362 ArchitectureName); 363 } 364 365 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 366 // Figure out the target triple. 367 llvm::Triple TheTriple("unknown-unknown-unknown"); 368 if (TripleName.empty()) { 369 if (Obj) { 370 TheTriple = Obj->makeTriple(); 371 } 372 } else { 373 TheTriple.setTriple(Triple::normalize(TripleName)); 374 375 // Use the triple, but also try to combine with ARM build attributes. 376 if (Obj) { 377 auto Arch = Obj->getArch(); 378 if (Arch == Triple::arm || Arch == Triple::armeb) { 379 Obj->setARMSubArch(TheTriple); 380 } 381 } 382 } 383 384 // Get the target specific parser. 385 std::string Error; 386 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 387 Error); 388 if (!TheTarget) { 389 if (Obj) 390 report_error(Obj->getFileName(), "can't find target: " + Error); 391 else 392 error("can't find target: " + Error); 393 } 394 395 // Update the triple name and return the found target. 396 TripleName = TheTriple.getTriple(); 397 return TheTarget; 398 } 399 400 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 401 return a.getOffset() < b.getOffset(); 402 } 403 404 template <class ELFT> 405 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 406 const RelocationRef &RelRef, 407 SmallVectorImpl<char> &Result) { 408 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 409 410 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 411 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 412 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 413 414 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 415 416 auto SecOrErr = EF.getSection(Rel.d.a); 417 if (!SecOrErr) 418 return errorToErrorCode(SecOrErr.takeError()); 419 const Elf_Shdr *Sec = *SecOrErr; 420 auto SymTabOrErr = EF.getSection(Sec->sh_link); 421 if (!SymTabOrErr) 422 return errorToErrorCode(SymTabOrErr.takeError()); 423 const Elf_Shdr *SymTab = *SymTabOrErr; 424 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 425 SymTab->sh_type == ELF::SHT_DYNSYM); 426 auto StrTabSec = EF.getSection(SymTab->sh_link); 427 if (!StrTabSec) 428 return errorToErrorCode(StrTabSec.takeError()); 429 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 430 if (!StrTabOrErr) 431 return errorToErrorCode(StrTabOrErr.takeError()); 432 StringRef StrTab = *StrTabOrErr; 433 uint8_t type = RelRef.getType(); 434 StringRef res; 435 int64_t addend = 0; 436 switch (Sec->sh_type) { 437 default: 438 return object_error::parse_failed; 439 case ELF::SHT_REL: { 440 // TODO: Read implicit addend from section data. 441 break; 442 } 443 case ELF::SHT_RELA: { 444 const Elf_Rela *ERela = Obj->getRela(Rel); 445 addend = ERela->r_addend; 446 break; 447 } 448 } 449 symbol_iterator SI = RelRef.getSymbol(); 450 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 451 StringRef Target; 452 if (symb->getType() == ELF::STT_SECTION) { 453 Expected<section_iterator> SymSI = SI->getSection(); 454 if (!SymSI) 455 return errorToErrorCode(SymSI.takeError()); 456 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 457 auto SecName = EF.getSectionName(SymSec); 458 if (!SecName) 459 return errorToErrorCode(SecName.takeError()); 460 Target = *SecName; 461 } else { 462 Expected<StringRef> SymName = symb->getName(StrTab); 463 if (!SymName) 464 return errorToErrorCode(SymName.takeError()); 465 Target = *SymName; 466 } 467 switch (EF.getHeader()->e_machine) { 468 case ELF::EM_X86_64: 469 switch (type) { 470 case ELF::R_X86_64_PC8: 471 case ELF::R_X86_64_PC16: 472 case ELF::R_X86_64_PC32: { 473 std::string fmtbuf; 474 raw_string_ostream fmt(fmtbuf); 475 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 476 fmt.flush(); 477 Result.append(fmtbuf.begin(), fmtbuf.end()); 478 } break; 479 case ELF::R_X86_64_8: 480 case ELF::R_X86_64_16: 481 case ELF::R_X86_64_32: 482 case ELF::R_X86_64_32S: 483 case ELF::R_X86_64_64: { 484 std::string fmtbuf; 485 raw_string_ostream fmt(fmtbuf); 486 fmt << Target << (addend < 0 ? "" : "+") << addend; 487 fmt.flush(); 488 Result.append(fmtbuf.begin(), fmtbuf.end()); 489 } break; 490 default: 491 res = "Unknown"; 492 } 493 break; 494 case ELF::EM_LANAI: 495 case ELF::EM_AVR: 496 case ELF::EM_AARCH64: { 497 std::string fmtbuf; 498 raw_string_ostream fmt(fmtbuf); 499 fmt << Target; 500 if (addend != 0) 501 fmt << (addend < 0 ? "" : "+") << addend; 502 fmt.flush(); 503 Result.append(fmtbuf.begin(), fmtbuf.end()); 504 break; 505 } 506 case ELF::EM_386: 507 case ELF::EM_IAMCU: 508 case ELF::EM_ARM: 509 case ELF::EM_HEXAGON: 510 case ELF::EM_MIPS: 511 case ELF::EM_BPF: 512 case ELF::EM_RISCV: 513 res = Target; 514 break; 515 case ELF::EM_WEBASSEMBLY: 516 switch (type) { 517 case ELF::R_WEBASSEMBLY_DATA: { 518 std::string fmtbuf; 519 raw_string_ostream fmt(fmtbuf); 520 fmt << Target << (addend < 0 ? "" : "+") << addend; 521 fmt.flush(); 522 Result.append(fmtbuf.begin(), fmtbuf.end()); 523 break; 524 } 525 case ELF::R_WEBASSEMBLY_FUNCTION: 526 res = Target; 527 break; 528 default: 529 res = "Unknown"; 530 } 531 break; 532 default: 533 res = "Unknown"; 534 } 535 if (Result.empty()) 536 Result.append(res.begin(), res.end()); 537 return std::error_code(); 538 } 539 540 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 541 const RelocationRef &Rel, 542 SmallVectorImpl<char> &Result) { 543 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 544 return getRelocationValueString(ELF32LE, Rel, Result); 545 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 546 return getRelocationValueString(ELF64LE, Rel, Result); 547 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 548 return getRelocationValueString(ELF32BE, Rel, Result); 549 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 550 return getRelocationValueString(ELF64BE, Rel, Result); 551 } 552 553 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 554 const RelocationRef &Rel, 555 SmallVectorImpl<char> &Result) { 556 symbol_iterator SymI = Rel.getSymbol(); 557 Expected<StringRef> SymNameOrErr = SymI->getName(); 558 if (!SymNameOrErr) 559 return errorToErrorCode(SymNameOrErr.takeError()); 560 StringRef SymName = *SymNameOrErr; 561 Result.append(SymName.begin(), SymName.end()); 562 return std::error_code(); 563 } 564 565 static void printRelocationTargetName(const MachOObjectFile *O, 566 const MachO::any_relocation_info &RE, 567 raw_string_ostream &fmt) { 568 bool IsScattered = O->isRelocationScattered(RE); 569 570 // Target of a scattered relocation is an address. In the interest of 571 // generating pretty output, scan through the symbol table looking for a 572 // symbol that aligns with that address. If we find one, print it. 573 // Otherwise, we just print the hex address of the target. 574 if (IsScattered) { 575 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 576 577 for (const SymbolRef &Symbol : O->symbols()) { 578 std::error_code ec; 579 Expected<uint64_t> Addr = Symbol.getAddress(); 580 if (!Addr) 581 report_error(O->getFileName(), Addr.takeError()); 582 if (*Addr != Val) 583 continue; 584 Expected<StringRef> Name = Symbol.getName(); 585 if (!Name) 586 report_error(O->getFileName(), Name.takeError()); 587 fmt << *Name; 588 return; 589 } 590 591 // If we couldn't find a symbol that this relocation refers to, try 592 // to find a section beginning instead. 593 for (const SectionRef &Section : ToolSectionFilter(*O)) { 594 std::error_code ec; 595 596 StringRef Name; 597 uint64_t Addr = Section.getAddress(); 598 if (Addr != Val) 599 continue; 600 if ((ec = Section.getName(Name))) 601 report_error(O->getFileName(), ec); 602 fmt << Name; 603 return; 604 } 605 606 fmt << format("0x%x", Val); 607 return; 608 } 609 610 StringRef S; 611 bool isExtern = O->getPlainRelocationExternal(RE); 612 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 613 614 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 615 fmt << format("0x%0" PRIx64, Val); 616 return; 617 } else if (isExtern) { 618 symbol_iterator SI = O->symbol_begin(); 619 advance(SI, Val); 620 Expected<StringRef> SOrErr = SI->getName(); 621 if (!SOrErr) 622 report_error(O->getFileName(), SOrErr.takeError()); 623 S = *SOrErr; 624 } else { 625 section_iterator SI = O->section_begin(); 626 // Adjust for the fact that sections are 1-indexed. 627 if (Val == 0) { 628 fmt << "0 (?,?)"; 629 return; 630 } 631 uint32_t i = Val - 1; 632 while (i != 0 && SI != O->section_end()) { 633 i--; 634 advance(SI, 1); 635 } 636 if (SI == O->section_end()) 637 fmt << Val << " (?,?)"; 638 else 639 SI->getName(S); 640 } 641 642 fmt << S; 643 } 644 645 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 646 const RelocationRef &RelRef, 647 SmallVectorImpl<char> &Result) { 648 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 649 symbol_iterator SI = RelRef.getSymbol(); 650 std::string fmtbuf; 651 raw_string_ostream fmt(fmtbuf); 652 if (SI == Obj->symbol_end()) { 653 // Not all wasm relocations have symbols associated with them. 654 // In particular R_WEBASSEMBLY_TYPE_INDEX_LEB. 655 fmt << Rel.Index; 656 } else { 657 Expected<StringRef> SymNameOrErr = SI->getName(); 658 if (!SymNameOrErr) 659 return errorToErrorCode(SymNameOrErr.takeError()); 660 StringRef SymName = *SymNameOrErr; 661 Result.append(SymName.begin(), SymName.end()); 662 } 663 fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 664 fmt.flush(); 665 Result.append(fmtbuf.begin(), fmtbuf.end()); 666 return std::error_code(); 667 } 668 669 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 670 const RelocationRef &RelRef, 671 SmallVectorImpl<char> &Result) { 672 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 673 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 674 675 unsigned Arch = Obj->getArch(); 676 677 std::string fmtbuf; 678 raw_string_ostream fmt(fmtbuf); 679 unsigned Type = Obj->getAnyRelocationType(RE); 680 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 681 682 // Determine any addends that should be displayed with the relocation. 683 // These require decoding the relocation type, which is triple-specific. 684 685 // X86_64 has entirely custom relocation types. 686 if (Arch == Triple::x86_64) { 687 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 688 689 switch (Type) { 690 case MachO::X86_64_RELOC_GOT_LOAD: 691 case MachO::X86_64_RELOC_GOT: { 692 printRelocationTargetName(Obj, RE, fmt); 693 fmt << "@GOT"; 694 if (isPCRel) 695 fmt << "PCREL"; 696 break; 697 } 698 case MachO::X86_64_RELOC_SUBTRACTOR: { 699 DataRefImpl RelNext = Rel; 700 Obj->moveRelocationNext(RelNext); 701 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 702 703 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 704 // X86_64_RELOC_UNSIGNED. 705 // NOTE: Scattered relocations don't exist on x86_64. 706 unsigned RType = Obj->getAnyRelocationType(RENext); 707 if (RType != MachO::X86_64_RELOC_UNSIGNED) 708 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 709 "X86_64_RELOC_SUBTRACTOR."); 710 711 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 712 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 713 printRelocationTargetName(Obj, RENext, fmt); 714 fmt << "-"; 715 printRelocationTargetName(Obj, RE, fmt); 716 break; 717 } 718 case MachO::X86_64_RELOC_TLV: 719 printRelocationTargetName(Obj, RE, fmt); 720 fmt << "@TLV"; 721 if (isPCRel) 722 fmt << "P"; 723 break; 724 case MachO::X86_64_RELOC_SIGNED_1: 725 printRelocationTargetName(Obj, RE, fmt); 726 fmt << "-1"; 727 break; 728 case MachO::X86_64_RELOC_SIGNED_2: 729 printRelocationTargetName(Obj, RE, fmt); 730 fmt << "-2"; 731 break; 732 case MachO::X86_64_RELOC_SIGNED_4: 733 printRelocationTargetName(Obj, RE, fmt); 734 fmt << "-4"; 735 break; 736 default: 737 printRelocationTargetName(Obj, RE, fmt); 738 break; 739 } 740 // X86 and ARM share some relocation types in common. 741 } else if (Arch == Triple::x86 || Arch == Triple::arm || 742 Arch == Triple::ppc) { 743 // Generic relocation types... 744 switch (Type) { 745 case MachO::GENERIC_RELOC_PAIR: // prints no info 746 return std::error_code(); 747 case MachO::GENERIC_RELOC_SECTDIFF: { 748 DataRefImpl RelNext = Rel; 749 Obj->moveRelocationNext(RelNext); 750 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 751 752 // X86 sect diff's must be followed by a relocation of type 753 // GENERIC_RELOC_PAIR. 754 unsigned RType = Obj->getAnyRelocationType(RENext); 755 756 if (RType != MachO::GENERIC_RELOC_PAIR) 757 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 758 "GENERIC_RELOC_SECTDIFF."); 759 760 printRelocationTargetName(Obj, RE, fmt); 761 fmt << "-"; 762 printRelocationTargetName(Obj, RENext, fmt); 763 break; 764 } 765 } 766 767 if (Arch == Triple::x86 || Arch == Triple::ppc) { 768 switch (Type) { 769 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 770 DataRefImpl RelNext = Rel; 771 Obj->moveRelocationNext(RelNext); 772 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 773 774 // X86 sect diff's must be followed by a relocation of type 775 // GENERIC_RELOC_PAIR. 776 unsigned RType = Obj->getAnyRelocationType(RENext); 777 if (RType != MachO::GENERIC_RELOC_PAIR) 778 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 779 "GENERIC_RELOC_LOCAL_SECTDIFF."); 780 781 printRelocationTargetName(Obj, RE, fmt); 782 fmt << "-"; 783 printRelocationTargetName(Obj, RENext, fmt); 784 break; 785 } 786 case MachO::GENERIC_RELOC_TLV: { 787 printRelocationTargetName(Obj, RE, fmt); 788 fmt << "@TLV"; 789 if (IsPCRel) 790 fmt << "P"; 791 break; 792 } 793 default: 794 printRelocationTargetName(Obj, RE, fmt); 795 } 796 } else { // ARM-specific relocations 797 switch (Type) { 798 case MachO::ARM_RELOC_HALF: 799 case MachO::ARM_RELOC_HALF_SECTDIFF: { 800 // Half relocations steal a bit from the length field to encode 801 // whether this is an upper16 or a lower16 relocation. 802 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 803 804 if (isUpper) 805 fmt << ":upper16:("; 806 else 807 fmt << ":lower16:("; 808 printRelocationTargetName(Obj, RE, fmt); 809 810 DataRefImpl RelNext = Rel; 811 Obj->moveRelocationNext(RelNext); 812 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 813 814 // ARM half relocs must be followed by a relocation of type 815 // ARM_RELOC_PAIR. 816 unsigned RType = Obj->getAnyRelocationType(RENext); 817 if (RType != MachO::ARM_RELOC_PAIR) 818 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 819 "ARM_RELOC_HALF"); 820 821 // NOTE: The half of the target virtual address is stashed in the 822 // address field of the secondary relocation, but we can't reverse 823 // engineer the constant offset from it without decoding the movw/movt 824 // instruction to find the other half in its immediate field. 825 826 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 827 // symbol/section pointer of the follow-on relocation. 828 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 829 fmt << "-"; 830 printRelocationTargetName(Obj, RENext, fmt); 831 } 832 833 fmt << ")"; 834 break; 835 } 836 default: { printRelocationTargetName(Obj, RE, fmt); } 837 } 838 } 839 } else 840 printRelocationTargetName(Obj, RE, fmt); 841 842 fmt.flush(); 843 Result.append(fmtbuf.begin(), fmtbuf.end()); 844 return std::error_code(); 845 } 846 847 static std::error_code getRelocationValueString(const RelocationRef &Rel, 848 SmallVectorImpl<char> &Result) { 849 const ObjectFile *Obj = Rel.getObject(); 850 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 851 return getRelocationValueString(ELF, Rel, Result); 852 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 853 return getRelocationValueString(COFF, Rel, Result); 854 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 855 return getRelocationValueString(Wasm, Rel, Result); 856 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 857 return getRelocationValueString(MachO, Rel, Result); 858 llvm_unreachable("unknown object file format"); 859 } 860 861 /// Indicates whether this relocation should hidden when listing 862 /// relocations, usually because it is the trailing part of a multipart 863 /// relocation that will be printed as part of the leading relocation. 864 static bool getHidden(RelocationRef RelRef) { 865 const ObjectFile *Obj = RelRef.getObject(); 866 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 867 if (!MachO) 868 return false; 869 870 unsigned Arch = MachO->getArch(); 871 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 872 uint64_t Type = MachO->getRelocationType(Rel); 873 874 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 875 // is always hidden. 876 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 877 if (Type == MachO::GENERIC_RELOC_PAIR) 878 return true; 879 } else if (Arch == Triple::x86_64) { 880 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 881 // an X86_64_RELOC_SUBTRACTOR. 882 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 883 DataRefImpl RelPrev = Rel; 884 RelPrev.d.a--; 885 uint64_t PrevType = MachO->getRelocationType(RelPrev); 886 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 887 return true; 888 } 889 } 890 891 return false; 892 } 893 894 namespace { 895 class SourcePrinter { 896 protected: 897 DILineInfo OldLineInfo; 898 const ObjectFile *Obj = nullptr; 899 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 900 // File name to file contents of source 901 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 902 // Mark the line endings of the cached source 903 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 904 905 private: 906 bool cacheSource(const DILineInfo& LineInfoFile); 907 908 public: 909 SourcePrinter() = default; 910 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 911 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 912 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 913 DefaultArch); 914 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 915 } 916 virtual ~SourcePrinter() = default; 917 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 918 StringRef Delimiter = "; "); 919 }; 920 921 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 922 std::unique_ptr<MemoryBuffer> Buffer; 923 if (LineInfo.Source) { 924 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 925 } else { 926 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 927 if (!BufferOrError) 928 return false; 929 Buffer = std::move(*BufferOrError); 930 } 931 // Chomp the file to get lines 932 size_t BufferSize = Buffer->getBufferSize(); 933 const char *BufferStart = Buffer->getBufferStart(); 934 for (const char *Start = BufferStart, *End = BufferStart; 935 End < BufferStart + BufferSize; End++) 936 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 937 (*End == '\r' && *(End + 1) == '\n')) { 938 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 939 if (*End == '\r') 940 End++; 941 Start = End + 1; 942 } 943 SourceCache[LineInfo.FileName] = std::move(Buffer); 944 return true; 945 } 946 947 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 948 StringRef Delimiter) { 949 if (!Symbolizer) 950 return; 951 DILineInfo LineInfo = DILineInfo(); 952 auto ExpectecLineInfo = 953 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 954 if (!ExpectecLineInfo) 955 consumeError(ExpectecLineInfo.takeError()); 956 else 957 LineInfo = *ExpectecLineInfo; 958 959 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 960 LineInfo.Line == 0) 961 return; 962 963 if (PrintLines) 964 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 965 if (PrintSource) { 966 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 967 if (!cacheSource(LineInfo)) 968 return; 969 auto FileBuffer = SourceCache.find(LineInfo.FileName); 970 if (FileBuffer != SourceCache.end()) { 971 auto LineBuffer = LineCache.find(LineInfo.FileName); 972 if (LineBuffer != LineCache.end()) { 973 if (LineInfo.Line > LineBuffer->second.size()) 974 return; 975 // Vector begins at 0, line numbers are non-zero 976 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 977 << "\n"; 978 } 979 } 980 } 981 OldLineInfo = LineInfo; 982 } 983 984 static bool isArmElf(const ObjectFile *Obj) { 985 return (Obj->isELF() && 986 (Obj->getArch() == Triple::aarch64 || 987 Obj->getArch() == Triple::aarch64_be || 988 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 989 Obj->getArch() == Triple::thumb || 990 Obj->getArch() == Triple::thumbeb)); 991 } 992 993 class PrettyPrinter { 994 public: 995 virtual ~PrettyPrinter() = default; 996 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 997 ArrayRef<uint8_t> Bytes, uint64_t Address, 998 raw_ostream &OS, StringRef Annot, 999 MCSubtargetInfo const &STI, SourcePrinter *SP, 1000 std::vector<RelocationRef> *Rels = nullptr) { 1001 if (SP && (PrintSource || PrintLines)) 1002 SP->printSourceLine(OS, Address); 1003 if (!NoLeadingAddr) 1004 OS << format("%8" PRIx64 ":", Address); 1005 if (!NoShowRawInsn) { 1006 OS << "\t"; 1007 dumpBytes(Bytes, OS); 1008 } 1009 if (MI) 1010 IP.printInst(MI, OS, "", STI); 1011 else 1012 OS << " <unknown>"; 1013 } 1014 }; 1015 PrettyPrinter PrettyPrinterInst; 1016 class HexagonPrettyPrinter : public PrettyPrinter { 1017 public: 1018 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1019 raw_ostream &OS) { 1020 uint32_t opcode = 1021 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1022 if (!NoLeadingAddr) 1023 OS << format("%8" PRIx64 ":", Address); 1024 if (!NoShowRawInsn) { 1025 OS << "\t"; 1026 dumpBytes(Bytes.slice(0, 4), OS); 1027 OS << format("%08" PRIx32, opcode); 1028 } 1029 } 1030 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1031 uint64_t Address, raw_ostream &OS, StringRef Annot, 1032 MCSubtargetInfo const &STI, SourcePrinter *SP, 1033 std::vector<RelocationRef> *Rels) override { 1034 if (SP && (PrintSource || PrintLines)) 1035 SP->printSourceLine(OS, Address, ""); 1036 if (!MI) { 1037 printLead(Bytes, Address, OS); 1038 OS << " <unknown>"; 1039 return; 1040 } 1041 std::string Buffer; 1042 { 1043 raw_string_ostream TempStream(Buffer); 1044 IP.printInst(MI, TempStream, "", STI); 1045 } 1046 StringRef Contents(Buffer); 1047 // Split off bundle attributes 1048 auto PacketBundle = Contents.rsplit('\n'); 1049 // Split off first instruction from the rest 1050 auto HeadTail = PacketBundle.first.split('\n'); 1051 auto Preamble = " { "; 1052 auto Separator = ""; 1053 StringRef Fmt = "\t\t\t%08" PRIx64 ": "; 1054 std::vector<RelocationRef>::const_iterator rel_cur = Rels->begin(); 1055 std::vector<RelocationRef>::const_iterator rel_end = Rels->end(); 1056 1057 // Hexagon's packets require relocations to be inline rather than 1058 // clustered at the end of the packet. 1059 auto PrintReloc = [&]() -> void { 1060 while ((rel_cur != rel_end) && (rel_cur->getOffset() <= Address)) { 1061 if (rel_cur->getOffset() == Address) { 1062 SmallString<16> name; 1063 SmallString<32> val; 1064 rel_cur->getTypeName(name); 1065 error(getRelocationValueString(*rel_cur, val)); 1066 OS << Separator << format(Fmt.data(), Address) << name << "\t" << val 1067 << "\n"; 1068 return; 1069 } 1070 rel_cur++; 1071 } 1072 }; 1073 1074 while(!HeadTail.first.empty()) { 1075 OS << Separator; 1076 Separator = "\n"; 1077 if (SP && (PrintSource || PrintLines)) 1078 SP->printSourceLine(OS, Address, ""); 1079 printLead(Bytes, Address, OS); 1080 OS << Preamble; 1081 Preamble = " "; 1082 StringRef Inst; 1083 auto Duplex = HeadTail.first.split('\v'); 1084 if(!Duplex.second.empty()){ 1085 OS << Duplex.first; 1086 OS << "; "; 1087 Inst = Duplex.second; 1088 } 1089 else 1090 Inst = HeadTail.first; 1091 OS << Inst; 1092 HeadTail = HeadTail.second.split('\n'); 1093 if (HeadTail.first.empty()) 1094 OS << " } " << PacketBundle.second; 1095 PrintReloc(); 1096 Bytes = Bytes.slice(4); 1097 Address += 4; 1098 } 1099 } 1100 }; 1101 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1102 1103 class AMDGCNPrettyPrinter : public PrettyPrinter { 1104 public: 1105 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1106 uint64_t Address, raw_ostream &OS, StringRef Annot, 1107 MCSubtargetInfo const &STI, SourcePrinter *SP, 1108 std::vector<RelocationRef> *Rels) override { 1109 if (SP && (PrintSource || PrintLines)) 1110 SP->printSourceLine(OS, Address); 1111 1112 typedef support::ulittle32_t U32; 1113 1114 if (MI) { 1115 SmallString<40> InstStr; 1116 raw_svector_ostream IS(InstStr); 1117 1118 IP.printInst(MI, IS, "", STI); 1119 1120 OS << left_justify(IS.str(), 60); 1121 } else { 1122 // an unrecognized encoding - this is probably data so represent it 1123 // using the .long directive, or .byte directive if fewer than 4 bytes 1124 // remaining 1125 if (Bytes.size() >= 4) { 1126 OS << format("\t.long 0x%08" PRIx32 " ", 1127 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 1128 OS.indent(42); 1129 } else { 1130 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1131 for (unsigned int i = 1; i < Bytes.size(); i++) 1132 OS << format(", 0x%02" PRIx8, Bytes[i]); 1133 OS.indent(55 - (6 * Bytes.size())); 1134 } 1135 } 1136 1137 OS << format("// %012" PRIX64 ": ", Address); 1138 if (Bytes.size() >=4) { 1139 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 1140 Bytes.size() / sizeof(U32))) 1141 // D should be explicitly casted to uint32_t here as it is passed 1142 // by format to snprintf as vararg. 1143 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 1144 } else { 1145 for (unsigned int i = 0; i < Bytes.size(); i++) 1146 OS << format("%02" PRIX8 " ", Bytes[i]); 1147 } 1148 1149 if (!Annot.empty()) 1150 OS << "// " << Annot; 1151 } 1152 }; 1153 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1154 1155 class BPFPrettyPrinter : public PrettyPrinter { 1156 public: 1157 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1158 uint64_t Address, raw_ostream &OS, StringRef Annot, 1159 MCSubtargetInfo const &STI, SourcePrinter *SP, 1160 std::vector<RelocationRef> *Rels) override { 1161 if (SP && (PrintSource || PrintLines)) 1162 SP->printSourceLine(OS, Address); 1163 if (!NoLeadingAddr) 1164 OS << format("%8" PRId64 ":", Address / 8); 1165 if (!NoShowRawInsn) { 1166 OS << "\t"; 1167 dumpBytes(Bytes, OS); 1168 } 1169 if (MI) 1170 IP.printInst(MI, OS, "", STI); 1171 else 1172 OS << " <unknown>"; 1173 } 1174 }; 1175 BPFPrettyPrinter BPFPrettyPrinterInst; 1176 1177 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1178 switch(Triple.getArch()) { 1179 default: 1180 return PrettyPrinterInst; 1181 case Triple::hexagon: 1182 return HexagonPrettyPrinterInst; 1183 case Triple::amdgcn: 1184 return AMDGCNPrettyPrinterInst; 1185 case Triple::bpfel: 1186 case Triple::bpfeb: 1187 return BPFPrettyPrinterInst; 1188 } 1189 } 1190 } 1191 1192 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1193 assert(Obj->isELF()); 1194 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1195 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1196 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1197 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1198 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1199 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1200 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1201 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1202 llvm_unreachable("Unsupported binary format"); 1203 } 1204 1205 template <class ELFT> static void 1206 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1207 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1208 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1209 uint8_t SymbolType = Symbol.getELFType(); 1210 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1211 continue; 1212 1213 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1214 if (!AddressOrErr) 1215 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1216 uint64_t Address = *AddressOrErr; 1217 1218 Expected<StringRef> Name = Symbol.getName(); 1219 if (!Name) 1220 report_error(Obj->getFileName(), Name.takeError()); 1221 if (Name->empty()) 1222 continue; 1223 1224 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1225 if (!SectionOrErr) 1226 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1227 section_iterator SecI = *SectionOrErr; 1228 if (SecI == Obj->section_end()) 1229 continue; 1230 1231 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1232 } 1233 } 1234 1235 static void 1236 addDynamicElfSymbols(const ObjectFile *Obj, 1237 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1238 assert(Obj->isELF()); 1239 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1240 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1241 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1242 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1243 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1244 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1245 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1246 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1247 else 1248 llvm_unreachable("Unsupported binary format"); 1249 } 1250 1251 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1252 if (StartAddress > StopAddress) 1253 error("Start address should be less than stop address"); 1254 1255 const Target *TheTarget = getTarget(Obj); 1256 1257 // Package up features to be passed to target/subtarget 1258 SubtargetFeatures Features = Obj->getFeatures(); 1259 if (MAttrs.size()) { 1260 for (unsigned i = 0; i != MAttrs.size(); ++i) 1261 Features.AddFeature(MAttrs[i]); 1262 } 1263 1264 std::unique_ptr<const MCRegisterInfo> MRI( 1265 TheTarget->createMCRegInfo(TripleName)); 1266 if (!MRI) 1267 report_error(Obj->getFileName(), "no register info for target " + 1268 TripleName); 1269 1270 // Set up disassembler. 1271 std::unique_ptr<const MCAsmInfo> AsmInfo( 1272 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1273 if (!AsmInfo) 1274 report_error(Obj->getFileName(), "no assembly info for target " + 1275 TripleName); 1276 std::unique_ptr<const MCSubtargetInfo> STI( 1277 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1278 if (!STI) 1279 report_error(Obj->getFileName(), "no subtarget info for target " + 1280 TripleName); 1281 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1282 if (!MII) 1283 report_error(Obj->getFileName(), "no instruction info for target " + 1284 TripleName); 1285 MCObjectFileInfo MOFI; 1286 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1287 // FIXME: for now initialize MCObjectFileInfo with default values 1288 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1289 1290 std::unique_ptr<MCDisassembler> DisAsm( 1291 TheTarget->createMCDisassembler(*STI, Ctx)); 1292 if (!DisAsm) 1293 report_error(Obj->getFileName(), "no disassembler for target " + 1294 TripleName); 1295 1296 std::unique_ptr<const MCInstrAnalysis> MIA( 1297 TheTarget->createMCInstrAnalysis(MII.get())); 1298 1299 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1300 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1301 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1302 if (!IP) 1303 report_error(Obj->getFileName(), "no instruction printer for target " + 1304 TripleName); 1305 IP->setPrintImmHex(PrintImmHex); 1306 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1307 1308 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1309 "\t\t\t%08" PRIx64 ": "; 1310 1311 SourcePrinter SP(Obj, TheTarget->getName()); 1312 1313 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1314 // in RelocSecs contain the relocations for section S. 1315 std::error_code EC; 1316 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1317 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1318 section_iterator Sec2 = Section.getRelocatedSection(); 1319 if (Sec2 != Obj->section_end()) 1320 SectionRelocMap[*Sec2].push_back(Section); 1321 } 1322 1323 // Create a mapping from virtual address to symbol name. This is used to 1324 // pretty print the symbols while disassembling. 1325 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1326 for (const SymbolRef &Symbol : Obj->symbols()) { 1327 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1328 if (!AddressOrErr) 1329 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1330 uint64_t Address = *AddressOrErr; 1331 1332 Expected<StringRef> Name = Symbol.getName(); 1333 if (!Name) 1334 report_error(Obj->getFileName(), Name.takeError()); 1335 if (Name->empty()) 1336 continue; 1337 1338 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1339 if (!SectionOrErr) 1340 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1341 section_iterator SecI = *SectionOrErr; 1342 if (SecI == Obj->section_end()) 1343 continue; 1344 1345 uint8_t SymbolType = ELF::STT_NOTYPE; 1346 if (Obj->isELF()) 1347 SymbolType = getElfSymbolType(Obj, Symbol); 1348 1349 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1350 1351 } 1352 if (AllSymbols.empty() && Obj->isELF()) 1353 addDynamicElfSymbols(Obj, AllSymbols); 1354 1355 // Create a mapping from virtual address to section. 1356 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1357 for (SectionRef Sec : Obj->sections()) 1358 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1359 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1360 1361 // Linked executables (.exe and .dll files) typically don't include a real 1362 // symbol table but they might contain an export table. 1363 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1364 for (const auto &ExportEntry : COFFObj->export_directories()) { 1365 StringRef Name; 1366 error(ExportEntry.getSymbolName(Name)); 1367 if (Name.empty()) 1368 continue; 1369 uint32_t RVA; 1370 error(ExportEntry.getExportRVA(RVA)); 1371 1372 uint64_t VA = COFFObj->getImageBase() + RVA; 1373 auto Sec = std::upper_bound( 1374 SectionAddresses.begin(), SectionAddresses.end(), VA, 1375 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1376 return LHS < RHS.first; 1377 }); 1378 if (Sec != SectionAddresses.begin()) 1379 --Sec; 1380 else 1381 Sec = SectionAddresses.end(); 1382 1383 if (Sec != SectionAddresses.end()) 1384 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1385 } 1386 } 1387 1388 // Sort all the symbols, this allows us to use a simple binary search to find 1389 // a symbol near an address. 1390 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1391 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1392 1393 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1394 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1395 continue; 1396 1397 uint64_t SectionAddr = Section.getAddress(); 1398 uint64_t SectSize = Section.getSize(); 1399 if (!SectSize) 1400 continue; 1401 1402 // Get the list of all the symbols in this section. 1403 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1404 std::vector<uint64_t> DataMappingSymsAddr; 1405 std::vector<uint64_t> TextMappingSymsAddr; 1406 if (isArmElf(Obj)) { 1407 for (const auto &Symb : Symbols) { 1408 uint64_t Address = std::get<0>(Symb); 1409 StringRef Name = std::get<1>(Symb); 1410 if (Name.startswith("$d")) 1411 DataMappingSymsAddr.push_back(Address - SectionAddr); 1412 if (Name.startswith("$x")) 1413 TextMappingSymsAddr.push_back(Address - SectionAddr); 1414 if (Name.startswith("$a")) 1415 TextMappingSymsAddr.push_back(Address - SectionAddr); 1416 if (Name.startswith("$t")) 1417 TextMappingSymsAddr.push_back(Address - SectionAddr); 1418 } 1419 } 1420 1421 llvm::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1422 llvm::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1423 1424 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1425 // AMDGPU disassembler uses symbolizer for printing labels 1426 std::unique_ptr<MCRelocationInfo> RelInfo( 1427 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1428 if (RelInfo) { 1429 std::unique_ptr<MCSymbolizer> Symbolizer( 1430 TheTarget->createMCSymbolizer( 1431 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1432 DisAsm->setSymbolizer(std::move(Symbolizer)); 1433 } 1434 } 1435 1436 // Make a list of all the relocations for this section. 1437 std::vector<RelocationRef> Rels; 1438 if (InlineRelocs) { 1439 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1440 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1441 Rels.push_back(Reloc); 1442 } 1443 } 1444 } 1445 1446 // Sort relocations by address. 1447 llvm::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1448 1449 StringRef SegmentName = ""; 1450 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1451 DataRefImpl DR = Section.getRawDataRefImpl(); 1452 SegmentName = MachO->getSectionFinalSegmentName(DR); 1453 } 1454 StringRef SectionName; 1455 error(Section.getName(SectionName)); 1456 1457 // If the section has no symbol at the start, just insert a dummy one. 1458 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1459 Symbols.insert( 1460 Symbols.begin(), 1461 std::make_tuple(SectionAddr, SectionName, 1462 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1463 } 1464 1465 SmallString<40> Comments; 1466 raw_svector_ostream CommentStream(Comments); 1467 1468 StringRef BytesStr; 1469 error(Section.getContents(BytesStr)); 1470 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1471 BytesStr.size()); 1472 1473 uint64_t Size; 1474 uint64_t Index; 1475 bool PrintedSection = false; 1476 1477 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1478 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1479 // Disassemble symbol by symbol. 1480 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1481 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1482 // The end is either the section end or the beginning of the next 1483 // symbol. 1484 uint64_t End = 1485 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1486 // Don't try to disassemble beyond the end of section contents. 1487 if (End > SectSize) 1488 End = SectSize; 1489 // If this symbol has the same address as the next symbol, then skip it. 1490 if (Start >= End) 1491 continue; 1492 1493 // Check if we need to skip symbol 1494 // Skip if the symbol's data is not between StartAddress and StopAddress 1495 if (End + SectionAddr < StartAddress || 1496 Start + SectionAddr > StopAddress) { 1497 continue; 1498 } 1499 1500 /// Skip if user requested specific symbols and this is not in the list 1501 if (!DisasmFuncsSet.empty() && 1502 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1503 continue; 1504 1505 if (!PrintedSection) { 1506 PrintedSection = true; 1507 outs() << "Disassembly of section "; 1508 if (!SegmentName.empty()) 1509 outs() << SegmentName << ","; 1510 outs() << SectionName << ':'; 1511 } 1512 1513 // Stop disassembly at the stop address specified 1514 if (End + SectionAddr > StopAddress) 1515 End = StopAddress - SectionAddr; 1516 1517 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1518 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1519 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1520 Start += 256; 1521 } 1522 if (si == se - 1 || 1523 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1524 // cut trailing zeroes at the end of kernel 1525 // cut up to 256 bytes 1526 const uint64_t EndAlign = 256; 1527 const auto Limit = End - (std::min)(EndAlign, End - Start); 1528 while (End > Limit && 1529 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1530 End -= 4; 1531 } 1532 } 1533 1534 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1535 1536 // Don't print raw contents of a virtual section. A virtual section 1537 // doesn't have any contents in the file. 1538 if (Section.isVirtual()) { 1539 outs() << "...\n"; 1540 continue; 1541 } 1542 1543 #ifndef NDEBUG 1544 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1545 #else 1546 raw_ostream &DebugOut = nulls(); 1547 #endif 1548 1549 for (Index = Start; Index < End; Index += Size) { 1550 MCInst Inst; 1551 1552 if (Index + SectionAddr < StartAddress || 1553 Index + SectionAddr > StopAddress) { 1554 // skip byte by byte till StartAddress is reached 1555 Size = 1; 1556 continue; 1557 } 1558 // AArch64 ELF binaries can interleave data and text in the 1559 // same section. We rely on the markers introduced to 1560 // understand what we need to dump. If the data marker is within a 1561 // function, it is denoted as a word/short etc 1562 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1563 !DisassembleAll) { 1564 uint64_t Stride = 0; 1565 1566 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1567 DataMappingSymsAddr.end(), Index); 1568 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1569 // Switch to data. 1570 while (Index < End) { 1571 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1572 outs() << "\t"; 1573 if (Index + 4 <= End) { 1574 Stride = 4; 1575 dumpBytes(Bytes.slice(Index, 4), outs()); 1576 outs() << "\t.word\t"; 1577 uint32_t Data = 0; 1578 if (Obj->isLittleEndian()) { 1579 const auto Word = 1580 reinterpret_cast<const support::ulittle32_t *>( 1581 Bytes.data() + Index); 1582 Data = *Word; 1583 } else { 1584 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1585 Bytes.data() + Index); 1586 Data = *Word; 1587 } 1588 outs() << "0x" << format("%08" PRIx32, Data); 1589 } else if (Index + 2 <= End) { 1590 Stride = 2; 1591 dumpBytes(Bytes.slice(Index, 2), outs()); 1592 outs() << "\t\t.short\t"; 1593 uint16_t Data = 0; 1594 if (Obj->isLittleEndian()) { 1595 const auto Short = 1596 reinterpret_cast<const support::ulittle16_t *>( 1597 Bytes.data() + Index); 1598 Data = *Short; 1599 } else { 1600 const auto Short = 1601 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1602 Index); 1603 Data = *Short; 1604 } 1605 outs() << "0x" << format("%04" PRIx16, Data); 1606 } else { 1607 Stride = 1; 1608 dumpBytes(Bytes.slice(Index, 1), outs()); 1609 outs() << "\t\t.byte\t"; 1610 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1611 } 1612 Index += Stride; 1613 outs() << "\n"; 1614 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1615 TextMappingSymsAddr.end(), Index); 1616 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1617 break; 1618 } 1619 } 1620 } 1621 1622 // If there is a data symbol inside an ELF text section and we are only 1623 // disassembling text (applicable all architectures), 1624 // we are in a situation where we must print the data and not 1625 // disassemble it. 1626 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1627 !DisassembleAll && Section.isText()) { 1628 // print out data up to 8 bytes at a time in hex and ascii 1629 uint8_t AsciiData[9] = {'\0'}; 1630 uint8_t Byte; 1631 int NumBytes = 0; 1632 1633 for (Index = Start; Index < End; Index += 1) { 1634 if (((SectionAddr + Index) < StartAddress) || 1635 ((SectionAddr + Index) > StopAddress)) 1636 continue; 1637 if (NumBytes == 0) { 1638 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1639 outs() << "\t"; 1640 } 1641 Byte = Bytes.slice(Index)[0]; 1642 outs() << format(" %02x", Byte); 1643 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1644 1645 uint8_t IndentOffset = 0; 1646 NumBytes++; 1647 if (Index == End - 1 || NumBytes > 8) { 1648 // Indent the space for less than 8 bytes data. 1649 // 2 spaces for byte and one for space between bytes 1650 IndentOffset = 3 * (8 - NumBytes); 1651 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1652 AsciiData[Excess] = '\0'; 1653 NumBytes = 8; 1654 } 1655 if (NumBytes == 8) { 1656 AsciiData[8] = '\0'; 1657 outs() << std::string(IndentOffset, ' ') << " "; 1658 outs() << reinterpret_cast<char *>(AsciiData); 1659 outs() << '\n'; 1660 NumBytes = 0; 1661 } 1662 } 1663 } 1664 if (Index >= End) 1665 break; 1666 1667 // Disassemble a real instruction or a data when disassemble all is 1668 // provided 1669 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1670 SectionAddr + Index, DebugOut, 1671 CommentStream); 1672 if (Size == 0) 1673 Size = 1; 1674 1675 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1676 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1677 *STI, &SP, &Rels); 1678 outs() << CommentStream.str(); 1679 Comments.clear(); 1680 1681 // Try to resolve the target of a call, tail call, etc. to a specific 1682 // symbol. 1683 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1684 MIA->isConditionalBranch(Inst))) { 1685 uint64_t Target; 1686 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1687 // In a relocatable object, the target's section must reside in 1688 // the same section as the call instruction or it is accessed 1689 // through a relocation. 1690 // 1691 // In a non-relocatable object, the target may be in any section. 1692 // 1693 // N.B. We don't walk the relocations in the relocatable case yet. 1694 auto *TargetSectionSymbols = &Symbols; 1695 if (!Obj->isRelocatableObject()) { 1696 auto SectionAddress = std::upper_bound( 1697 SectionAddresses.begin(), SectionAddresses.end(), Target, 1698 [](uint64_t LHS, 1699 const std::pair<uint64_t, SectionRef> &RHS) { 1700 return LHS < RHS.first; 1701 }); 1702 if (SectionAddress != SectionAddresses.begin()) { 1703 --SectionAddress; 1704 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1705 } else { 1706 TargetSectionSymbols = nullptr; 1707 } 1708 } 1709 1710 // Find the first symbol in the section whose offset is less than 1711 // or equal to the target. 1712 if (TargetSectionSymbols) { 1713 auto TargetSym = std::upper_bound( 1714 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1715 Target, [](uint64_t LHS, 1716 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1717 return LHS < std::get<0>(RHS); 1718 }); 1719 if (TargetSym != TargetSectionSymbols->begin()) { 1720 --TargetSym; 1721 uint64_t TargetAddress = std::get<0>(*TargetSym); 1722 StringRef TargetName = std::get<1>(*TargetSym); 1723 outs() << " <" << TargetName; 1724 uint64_t Disp = Target - TargetAddress; 1725 if (Disp) 1726 outs() << "+0x" << Twine::utohexstr(Disp); 1727 outs() << '>'; 1728 } 1729 } 1730 } 1731 } 1732 outs() << "\n"; 1733 1734 // Hexagon does this in pretty printer 1735 if (Obj->getArch() != Triple::hexagon) 1736 // Print relocation for instruction. 1737 while (rel_cur != rel_end) { 1738 bool hidden = getHidden(*rel_cur); 1739 uint64_t addr = rel_cur->getOffset(); 1740 SmallString<16> name; 1741 SmallString<32> val; 1742 1743 // If this relocation is hidden, skip it. 1744 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1745 ++rel_cur; 1746 continue; 1747 } 1748 1749 // Stop when rel_cur's address is past the current instruction. 1750 if (addr >= Index + Size) break; 1751 rel_cur->getTypeName(name); 1752 error(getRelocationValueString(*rel_cur, val)); 1753 outs() << format(Fmt.data(), SectionAddr + addr) << name 1754 << "\t" << val << "\n"; 1755 ++rel_cur; 1756 } 1757 } 1758 } 1759 } 1760 } 1761 1762 void llvm::PrintRelocations(const ObjectFile *Obj) { 1763 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1764 "%08" PRIx64; 1765 // Regular objdump doesn't print relocations in non-relocatable object 1766 // files. 1767 if (!Obj->isRelocatableObject()) 1768 return; 1769 1770 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1771 if (Section.relocation_begin() == Section.relocation_end()) 1772 continue; 1773 StringRef secname; 1774 error(Section.getName(secname)); 1775 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1776 for (const RelocationRef &Reloc : Section.relocations()) { 1777 bool hidden = getHidden(Reloc); 1778 uint64_t address = Reloc.getOffset(); 1779 SmallString<32> relocname; 1780 SmallString<32> valuestr; 1781 if (address < StartAddress || address > StopAddress || hidden) 1782 continue; 1783 Reloc.getTypeName(relocname); 1784 error(getRelocationValueString(Reloc, valuestr)); 1785 outs() << format(Fmt.data(), address) << " " << relocname << " " 1786 << valuestr << "\n"; 1787 } 1788 outs() << "\n"; 1789 } 1790 } 1791 1792 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1793 outs() << "Sections:\n" 1794 "Idx Name Size Address Type\n"; 1795 unsigned i = 0; 1796 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1797 StringRef Name; 1798 error(Section.getName(Name)); 1799 uint64_t Address = Section.getAddress(); 1800 uint64_t Size = Section.getSize(); 1801 bool Text = Section.isText(); 1802 bool Data = Section.isData(); 1803 bool BSS = Section.isBSS(); 1804 std::string Type = (std::string(Text ? "TEXT " : "") + 1805 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1806 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1807 Name.str().c_str(), Size, Address, Type.c_str()); 1808 ++i; 1809 } 1810 } 1811 1812 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1813 std::error_code EC; 1814 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1815 StringRef Name; 1816 StringRef Contents; 1817 error(Section.getName(Name)); 1818 uint64_t BaseAddr = Section.getAddress(); 1819 uint64_t Size = Section.getSize(); 1820 if (!Size) 1821 continue; 1822 1823 outs() << "Contents of section " << Name << ":\n"; 1824 if (Section.isBSS()) { 1825 outs() << format("<skipping contents of bss section at [%04" PRIx64 1826 ", %04" PRIx64 ")>\n", 1827 BaseAddr, BaseAddr + Size); 1828 continue; 1829 } 1830 1831 error(Section.getContents(Contents)); 1832 1833 // Dump out the content as hex and printable ascii characters. 1834 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1835 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1836 // Dump line of hex. 1837 for (std::size_t i = 0; i < 16; ++i) { 1838 if (i != 0 && i % 4 == 0) 1839 outs() << ' '; 1840 if (addr + i < end) 1841 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1842 << hexdigit(Contents[addr + i] & 0xF, true); 1843 else 1844 outs() << " "; 1845 } 1846 // Print ascii. 1847 outs() << " "; 1848 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1849 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1850 outs() << Contents[addr + i]; 1851 else 1852 outs() << "."; 1853 } 1854 outs() << "\n"; 1855 } 1856 } 1857 } 1858 1859 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1860 StringRef ArchitectureName) { 1861 outs() << "SYMBOL TABLE:\n"; 1862 1863 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1864 printCOFFSymbolTable(coff); 1865 return; 1866 } 1867 for (const SymbolRef &Symbol : o->symbols()) { 1868 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1869 if (!AddressOrError) 1870 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1871 ArchitectureName); 1872 uint64_t Address = *AddressOrError; 1873 if ((Address < StartAddress) || (Address > StopAddress)) 1874 continue; 1875 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1876 if (!TypeOrError) 1877 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1878 ArchitectureName); 1879 SymbolRef::Type Type = *TypeOrError; 1880 uint32_t Flags = Symbol.getFlags(); 1881 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1882 if (!SectionOrErr) 1883 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1884 ArchitectureName); 1885 section_iterator Section = *SectionOrErr; 1886 StringRef Name; 1887 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1888 Section->getName(Name); 1889 } else { 1890 Expected<StringRef> NameOrErr = Symbol.getName(); 1891 if (!NameOrErr) 1892 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1893 ArchitectureName); 1894 Name = *NameOrErr; 1895 } 1896 1897 bool Global = Flags & SymbolRef::SF_Global; 1898 bool Weak = Flags & SymbolRef::SF_Weak; 1899 bool Absolute = Flags & SymbolRef::SF_Absolute; 1900 bool Common = Flags & SymbolRef::SF_Common; 1901 bool Hidden = Flags & SymbolRef::SF_Hidden; 1902 1903 char GlobLoc = ' '; 1904 if (Type != SymbolRef::ST_Unknown) 1905 GlobLoc = Global ? 'g' : 'l'; 1906 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1907 ? 'd' : ' '; 1908 char FileFunc = ' '; 1909 if (Type == SymbolRef::ST_File) 1910 FileFunc = 'f'; 1911 else if (Type == SymbolRef::ST_Function) 1912 FileFunc = 'F'; 1913 1914 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1915 "%08" PRIx64; 1916 1917 outs() << format(Fmt, Address) << " " 1918 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1919 << (Weak ? 'w' : ' ') // Weak? 1920 << ' ' // Constructor. Not supported yet. 1921 << ' ' // Warning. Not supported yet. 1922 << ' ' // Indirect reference to another symbol. 1923 << Debug // Debugging (d) or dynamic (D) symbol. 1924 << FileFunc // Name of function (F), file (f) or object (O). 1925 << ' '; 1926 if (Absolute) { 1927 outs() << "*ABS*"; 1928 } else if (Common) { 1929 outs() << "*COM*"; 1930 } else if (Section == o->section_end()) { 1931 outs() << "*UND*"; 1932 } else { 1933 if (const MachOObjectFile *MachO = 1934 dyn_cast<const MachOObjectFile>(o)) { 1935 DataRefImpl DR = Section->getRawDataRefImpl(); 1936 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1937 outs() << SegmentName << ","; 1938 } 1939 StringRef SectionName; 1940 error(Section->getName(SectionName)); 1941 outs() << SectionName; 1942 } 1943 1944 outs() << '\t'; 1945 if (Common || isa<ELFObjectFileBase>(o)) { 1946 uint64_t Val = 1947 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1948 outs() << format("\t %08" PRIx64 " ", Val); 1949 } 1950 1951 if (Hidden) { 1952 outs() << ".hidden "; 1953 } 1954 outs() << Name 1955 << '\n'; 1956 } 1957 } 1958 1959 static void PrintUnwindInfo(const ObjectFile *o) { 1960 outs() << "Unwind info:\n\n"; 1961 1962 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1963 printCOFFUnwindInfo(coff); 1964 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1965 printMachOUnwindInfo(MachO); 1966 else { 1967 // TODO: Extract DWARF dump tool to objdump. 1968 errs() << "This operation is only currently supported " 1969 "for COFF and MachO object files.\n"; 1970 return; 1971 } 1972 } 1973 1974 void llvm::printExportsTrie(const ObjectFile *o) { 1975 outs() << "Exports trie:\n"; 1976 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1977 printMachOExportsTrie(MachO); 1978 else { 1979 errs() << "This operation is only currently supported " 1980 "for Mach-O executable files.\n"; 1981 return; 1982 } 1983 } 1984 1985 void llvm::printRebaseTable(ObjectFile *o) { 1986 outs() << "Rebase table:\n"; 1987 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1988 printMachORebaseTable(MachO); 1989 else { 1990 errs() << "This operation is only currently supported " 1991 "for Mach-O executable files.\n"; 1992 return; 1993 } 1994 } 1995 1996 void llvm::printBindTable(ObjectFile *o) { 1997 outs() << "Bind table:\n"; 1998 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1999 printMachOBindTable(MachO); 2000 else { 2001 errs() << "This operation is only currently supported " 2002 "for Mach-O executable files.\n"; 2003 return; 2004 } 2005 } 2006 2007 void llvm::printLazyBindTable(ObjectFile *o) { 2008 outs() << "Lazy bind table:\n"; 2009 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2010 printMachOLazyBindTable(MachO); 2011 else { 2012 errs() << "This operation is only currently supported " 2013 "for Mach-O executable files.\n"; 2014 return; 2015 } 2016 } 2017 2018 void llvm::printWeakBindTable(ObjectFile *o) { 2019 outs() << "Weak bind table:\n"; 2020 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 2021 printMachOWeakBindTable(MachO); 2022 else { 2023 errs() << "This operation is only currently supported " 2024 "for Mach-O executable files.\n"; 2025 return; 2026 } 2027 } 2028 2029 /// Dump the raw contents of the __clangast section so the output can be piped 2030 /// into llvm-bcanalyzer. 2031 void llvm::printRawClangAST(const ObjectFile *Obj) { 2032 if (outs().is_displayed()) { 2033 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 2034 "the clang ast section.\n" 2035 "Please redirect the output to a file or another program such as " 2036 "llvm-bcanalyzer.\n"; 2037 return; 2038 } 2039 2040 StringRef ClangASTSectionName("__clangast"); 2041 if (isa<COFFObjectFile>(Obj)) { 2042 ClangASTSectionName = "clangast"; 2043 } 2044 2045 Optional<object::SectionRef> ClangASTSection; 2046 for (auto Sec : ToolSectionFilter(*Obj)) { 2047 StringRef Name; 2048 Sec.getName(Name); 2049 if (Name == ClangASTSectionName) { 2050 ClangASTSection = Sec; 2051 break; 2052 } 2053 } 2054 if (!ClangASTSection) 2055 return; 2056 2057 StringRef ClangASTContents; 2058 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2059 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2060 } 2061 2062 static void printFaultMaps(const ObjectFile *Obj) { 2063 const char *FaultMapSectionName = nullptr; 2064 2065 if (isa<ELFObjectFileBase>(Obj)) { 2066 FaultMapSectionName = ".llvm_faultmaps"; 2067 } else if (isa<MachOObjectFile>(Obj)) { 2068 FaultMapSectionName = "__llvm_faultmaps"; 2069 } else { 2070 errs() << "This operation is only currently supported " 2071 "for ELF and Mach-O executable files.\n"; 2072 return; 2073 } 2074 2075 Optional<object::SectionRef> FaultMapSection; 2076 2077 for (auto Sec : ToolSectionFilter(*Obj)) { 2078 StringRef Name; 2079 Sec.getName(Name); 2080 if (Name == FaultMapSectionName) { 2081 FaultMapSection = Sec; 2082 break; 2083 } 2084 } 2085 2086 outs() << "FaultMap table:\n"; 2087 2088 if (!FaultMapSection.hasValue()) { 2089 outs() << "<not found>\n"; 2090 return; 2091 } 2092 2093 StringRef FaultMapContents; 2094 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2095 2096 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2097 FaultMapContents.bytes_end()); 2098 2099 outs() << FMP; 2100 } 2101 2102 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2103 if (o->isELF()) 2104 return printELFFileHeader(o); 2105 if (o->isCOFF()) 2106 return printCOFFFileHeader(o); 2107 if (o->isWasm()) 2108 return printWasmFileHeader(o); 2109 if (o->isMachO()) { 2110 printMachOFileHeader(o); 2111 if (!onlyFirst) 2112 printMachOLoadCommands(o); 2113 return; 2114 } 2115 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2116 } 2117 2118 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2119 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2120 // Avoid other output when using a raw option. 2121 if (!RawClangAST) { 2122 outs() << '\n'; 2123 if (a) 2124 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2125 else 2126 outs() << o->getFileName(); 2127 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2128 } 2129 2130 if (Disassemble) 2131 DisassembleObject(o, Relocations); 2132 if (Relocations && !Disassemble) 2133 PrintRelocations(o); 2134 if (SectionHeaders) 2135 PrintSectionHeaders(o); 2136 if (SectionContents) 2137 PrintSectionContents(o); 2138 if (SymbolTable) 2139 PrintSymbolTable(o, ArchiveName); 2140 if (UnwindInfo) 2141 PrintUnwindInfo(o); 2142 if (PrivateHeaders || FirstPrivateHeader) 2143 printPrivateFileHeaders(o, FirstPrivateHeader); 2144 if (ExportsTrie) 2145 printExportsTrie(o); 2146 if (Rebase) 2147 printRebaseTable(o); 2148 if (Bind) 2149 printBindTable(o); 2150 if (LazyBind) 2151 printLazyBindTable(o); 2152 if (WeakBind) 2153 printWeakBindTable(o); 2154 if (RawClangAST) 2155 printRawClangAST(o); 2156 if (PrintFaultMaps) 2157 printFaultMaps(o); 2158 if (DwarfDumpType != DIDT_Null) { 2159 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2160 // Dump the complete DWARF structure. 2161 DIDumpOptions DumpOpts; 2162 DumpOpts.DumpType = DwarfDumpType; 2163 DICtx->dump(outs(), DumpOpts); 2164 } 2165 } 2166 2167 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2168 StringRef ArchiveName = A ? A->getFileName() : ""; 2169 2170 // Avoid other output when using a raw option. 2171 if (!RawClangAST) 2172 outs() << '\n' 2173 << ArchiveName << "(" << I->getFileName() << ")" 2174 << ":\tfile format COFF-import-file" 2175 << "\n\n"; 2176 2177 if (SymbolTable) 2178 printCOFFSymbolTable(I); 2179 } 2180 2181 /// Dump each object file in \a a; 2182 static void DumpArchive(const Archive *a) { 2183 Error Err = Error::success(); 2184 for (auto &C : a->children(Err)) { 2185 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2186 if (!ChildOrErr) { 2187 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2188 report_error(a->getFileName(), C, std::move(E)); 2189 continue; 2190 } 2191 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2192 DumpObject(o, a); 2193 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2194 DumpObject(I, a); 2195 else 2196 report_error(a->getFileName(), object_error::invalid_file_type); 2197 } 2198 if (Err) 2199 report_error(a->getFileName(), std::move(Err)); 2200 } 2201 2202 /// Open file and figure out how to dump it. 2203 static void DumpInput(StringRef file) { 2204 2205 // If we are using the Mach-O specific object file parser, then let it parse 2206 // the file and process the command line options. So the -arch flags can 2207 // be used to select specific slices, etc. 2208 if (MachOOpt) { 2209 ParseInputMachO(file); 2210 return; 2211 } 2212 2213 // Attempt to open the binary. 2214 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2215 if (!BinaryOrErr) 2216 report_error(file, BinaryOrErr.takeError()); 2217 Binary &Binary = *BinaryOrErr.get().getBinary(); 2218 2219 if (Archive *a = dyn_cast<Archive>(&Binary)) 2220 DumpArchive(a); 2221 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2222 DumpObject(o); 2223 else 2224 report_error(file, object_error::invalid_file_type); 2225 } 2226 2227 int main(int argc, char **argv) { 2228 InitLLVM X(argc, argv); 2229 2230 // Initialize targets and assembly printers/parsers. 2231 llvm::InitializeAllTargetInfos(); 2232 llvm::InitializeAllTargetMCs(); 2233 llvm::InitializeAllDisassemblers(); 2234 2235 // Register the target printer for --version. 2236 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2237 2238 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2239 TripleName = Triple::normalize(TripleName); 2240 2241 ToolName = argv[0]; 2242 2243 // Defaults to a.out if no filenames specified. 2244 if (InputFilenames.size() == 0) 2245 InputFilenames.push_back("a.out"); 2246 2247 if (DisassembleAll || PrintSource || PrintLines) 2248 Disassemble = true; 2249 if (!Disassemble 2250 && !Relocations 2251 && !SectionHeaders 2252 && !SectionContents 2253 && !SymbolTable 2254 && !UnwindInfo 2255 && !PrivateHeaders 2256 && !FirstPrivateHeader 2257 && !ExportsTrie 2258 && !Rebase 2259 && !Bind 2260 && !LazyBind 2261 && !WeakBind 2262 && !RawClangAST 2263 && !(UniversalHeaders && MachOOpt) 2264 && !(ArchiveHeaders && MachOOpt) 2265 && !(IndirectSymbols && MachOOpt) 2266 && !(DataInCode && MachOOpt) 2267 && !(LinkOptHints && MachOOpt) 2268 && !(InfoPlist && MachOOpt) 2269 && !(DylibsUsed && MachOOpt) 2270 && !(DylibId && MachOOpt) 2271 && !(ObjcMetaData && MachOOpt) 2272 && !(FilterSections.size() != 0 && MachOOpt) 2273 && !PrintFaultMaps 2274 && DwarfDumpType == DIDT_Null) { 2275 cl::PrintHelpMessage(); 2276 return 2; 2277 } 2278 2279 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2280 DisassembleFunctions.end()); 2281 2282 llvm::for_each(InputFilenames, DumpInput); 2283 2284 return EXIT_SUCCESS; 2285 } 2286