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/Triple.h" 24#include "llvm/CodeGen/FaultMaps.h" 25#include "llvm/MC/MCAsmInfo.h" 26#include "llvm/MC/MCContext.h" 27#include "llvm/MC/MCDisassembler.h" 28#include "llvm/MC/MCInst.h" 29#include "llvm/MC/MCInstPrinter.h" 30#include "llvm/MC/MCInstrAnalysis.h" 31#include "llvm/MC/MCInstrInfo.h" 32#include "llvm/MC/MCObjectFileInfo.h" 33#include "llvm/MC/MCRegisterInfo.h" 34#include "llvm/MC/MCRelocationInfo.h" 35#include "llvm/MC/MCSubtargetInfo.h" 36#include "llvm/Object/Archive.h" 37#include "llvm/Object/ELFObjectFile.h" 38#include "llvm/Object/COFF.h" 39#include "llvm/Object/MachO.h" 40#include "llvm/Object/ObjectFile.h" 41#include "llvm/Support/Casting.h" 42#include "llvm/Support/CommandLine.h" 43#include "llvm/Support/Debug.h" 44#include "llvm/Support/Errc.h" 45#include "llvm/Support/FileSystem.h" 46#include "llvm/Support/Format.h" 47#include "llvm/Support/GraphWriter.h" 48#include "llvm/Support/Host.h" 49#include "llvm/Support/ManagedStatic.h" 50#include "llvm/Support/MemoryBuffer.h" 51#include "llvm/Support/PrettyStackTrace.h" 52#include "llvm/Support/Signals.h" 53#include "llvm/Support/SourceMgr.h" 54#include "llvm/Support/TargetRegistry.h" 55#include "llvm/Support/TargetSelect.h" 56#include "llvm/Support/raw_ostream.h" 57#include <algorithm> 58#include <cctype> 59#include <cstring> 60#include <system_error> 61 62using namespace llvm; 63using namespace object; 64 65static cl::list<std::string> 66InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 67 68cl::opt<bool> 69llvm::Disassemble("disassemble", 70 cl::desc("Display assembler mnemonics for the machine instructions")); 71static cl::alias 72Disassembled("d", cl::desc("Alias for --disassemble"), 73 cl::aliasopt(Disassemble)); 74 75cl::opt<bool> 76llvm::DisassembleAll("disassemble-all", 77 cl::desc("Display assembler mnemonics for the machine instructions")); 78static cl::alias 79DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 80 cl::aliasopt(DisassembleAll)); 81 82cl::opt<bool> 83llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 84 85cl::opt<bool> 86llvm::SectionContents("s", cl::desc("Display the content of each section")); 87 88cl::opt<bool> 89llvm::SymbolTable("t", cl::desc("Display the symbol table")); 90 91cl::opt<bool> 92llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 93 94cl::opt<bool> 95llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 96 97cl::opt<bool> 98llvm::Bind("bind", cl::desc("Display mach-o binding info")); 99 100cl::opt<bool> 101llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 102 103cl::opt<bool> 104llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 105 106cl::opt<bool> 107llvm::RawClangAST("raw-clang-ast", 108 cl::desc("Dump the raw binary contents of the clang AST section")); 109 110static cl::opt<bool> 111MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 112static cl::alias 113MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 114 115cl::opt<std::string> 116llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 117 "see -version for available targets")); 118 119cl::opt<std::string> 120llvm::MCPU("mcpu", 121 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 122 cl::value_desc("cpu-name"), 123 cl::init("")); 124 125cl::opt<std::string> 126llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 127 "see -version for available targets")); 128 129cl::opt<bool> 130llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 131 "headers for each section.")); 132static cl::alias 133SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 134 cl::aliasopt(SectionHeaders)); 135static cl::alias 136SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 137 cl::aliasopt(SectionHeaders)); 138 139cl::list<std::string> 140llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 141 "With -macho dump segment,section")); 142cl::alias 143static FilterSectionsj("j", cl::desc("Alias for --section"), 144 cl::aliasopt(llvm::FilterSections)); 145 146cl::list<std::string> 147llvm::MAttrs("mattr", 148 cl::CommaSeparated, 149 cl::desc("Target specific attributes"), 150 cl::value_desc("a1,+a2,-a3,...")); 151 152cl::opt<bool> 153llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 154 "instructions, do not print " 155 "the instruction bytes.")); 156 157cl::opt<bool> 158llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 159 160static cl::alias 161UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 162 cl::aliasopt(UnwindInfo)); 163 164cl::opt<bool> 165llvm::PrivateHeaders("private-headers", 166 cl::desc("Display format specific file headers")); 167 168cl::opt<bool> 169llvm::FirstPrivateHeader("private-header", 170 cl::desc("Display only the first format specific file " 171 "header")); 172 173static cl::alias 174PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 175 cl::aliasopt(PrivateHeaders)); 176 177cl::opt<bool> 178 llvm::PrintImmHex("print-imm-hex", 179 cl::desc("Use hex format for immediate values")); 180 181cl::opt<bool> PrintFaultMaps("fault-map-section", 182 cl::desc("Display contents of faultmap section")); 183 184static StringRef ToolName; 185 186namespace { 187typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 188 189class SectionFilterIterator { 190public: 191 SectionFilterIterator(FilterPredicate P, 192 llvm::object::section_iterator const &I, 193 llvm::object::section_iterator const &E) 194 : Predicate(P), Iterator(I), End(E) { 195 ScanPredicate(); 196 } 197 const llvm::object::SectionRef &operator*() const { return *Iterator; } 198 SectionFilterIterator &operator++() { 199 ++Iterator; 200 ScanPredicate(); 201 return *this; 202 } 203 bool operator!=(SectionFilterIterator const &Other) const { 204 return Iterator != Other.Iterator; 205 } 206 207private: 208 void ScanPredicate() { 209 while (Iterator != End && !Predicate(*Iterator)) { 210 ++Iterator; 211 } 212 } 213 FilterPredicate Predicate; 214 llvm::object::section_iterator Iterator; 215 llvm::object::section_iterator End; 216}; 217 218class SectionFilter { 219public: 220 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 221 : Predicate(P), Object(O) {} 222 SectionFilterIterator begin() { 223 return SectionFilterIterator(Predicate, Object.section_begin(), 224 Object.section_end()); 225 } 226 SectionFilterIterator end() { 227 return SectionFilterIterator(Predicate, Object.section_end(), 228 Object.section_end()); 229 } 230 231private: 232 FilterPredicate Predicate; 233 llvm::object::ObjectFile const &Object; 234}; 235SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 236 return SectionFilter([](llvm::object::SectionRef const &S) { 237 if(FilterSections.empty()) 238 return true; 239 llvm::StringRef String; 240 std::error_code error = S.getName(String); 241 if (error) 242 return false; 243 return std::find(FilterSections.begin(), 244 FilterSections.end(), 245 String) != FilterSections.end(); 246 }, 247 O); 248} 249} 250 251void llvm::error(std::error_code EC) { 252 if (!EC) 253 return; 254 255 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 256 errs().flush(); 257 exit(1); 258} 259 260LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 261 std::error_code EC) { 262 assert(EC); 263 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 264 exit(1); 265} 266 267static const Target *getTarget(const ObjectFile *Obj = nullptr) { 268 // Figure out the target triple. 269 llvm::Triple TheTriple("unknown-unknown-unknown"); 270 if (TripleName.empty()) { 271 if (Obj) { 272 TheTriple.setArch(Triple::ArchType(Obj->getArch())); 273 // TheTriple defaults to ELF, and COFF doesn't have an environment: 274 // the best we can do here is indicate that it is mach-o. 275 if (Obj->isMachO()) 276 TheTriple.setObjectFormat(Triple::MachO); 277 278 if (Obj->isCOFF()) { 279 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 280 if (COFFObj->getArch() == Triple::thumb) 281 TheTriple.setTriple("thumbv7-windows"); 282 } 283 } 284 } else 285 TheTriple.setTriple(Triple::normalize(TripleName)); 286 287 // Get the target specific parser. 288 std::string Error; 289 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 290 Error); 291 if (!TheTarget) 292 report_fatal_error("can't find target: " + Error); 293 294 // Update the triple name and return the found target. 295 TripleName = TheTriple.getTriple(); 296 return TheTarget; 297} 298 299bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 300 return a.getOffset() < b.getOffset(); 301} 302 303namespace { 304class PrettyPrinter { 305public: 306 virtual ~PrettyPrinter(){} 307 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 308 ArrayRef<uint8_t> Bytes, uint64_t Address, 309 raw_ostream &OS, StringRef Annot, 310 MCSubtargetInfo const &STI) { 311 outs() << format("%8" PRIx64 ":", Address); 312 if (!NoShowRawInsn) { 313 outs() << "\t"; 314 dumpBytes(Bytes, outs()); 315 } 316 IP.printInst(MI, outs(), "", STI); 317 } 318}; 319PrettyPrinter PrettyPrinterInst; 320class HexagonPrettyPrinter : public PrettyPrinter { 321public: 322 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 323 raw_ostream &OS) { 324 uint32_t opcode = 325 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 326 OS << format("%8" PRIx64 ":", Address); 327 if (!NoShowRawInsn) { 328 OS << "\t"; 329 dumpBytes(Bytes.slice(0, 4), OS); 330 OS << format("%08" PRIx32, opcode); 331 } 332 } 333 void printInst(MCInstPrinter &IP, const MCInst *MI, 334 ArrayRef<uint8_t> Bytes, uint64_t Address, 335 raw_ostream &OS, StringRef Annot, 336 MCSubtargetInfo const &STI) override { 337 std::string Buffer; 338 { 339 raw_string_ostream TempStream(Buffer); 340 IP.printInst(MI, TempStream, "", STI); 341 } 342 StringRef Contents(Buffer); 343 // Split off bundle attributes 344 auto PacketBundle = Contents.rsplit('\n'); 345 // Split off first instruction from the rest 346 auto HeadTail = PacketBundle.first.split('\n'); 347 auto Preamble = " { "; 348 auto Separator = ""; 349 while(!HeadTail.first.empty()) { 350 OS << Separator; 351 Separator = "\n"; 352 printLead(Bytes, Address, OS); 353 OS << Preamble; 354 Preamble = " "; 355 StringRef Inst; 356 auto Duplex = HeadTail.first.split('\v'); 357 if(!Duplex.second.empty()){ 358 OS << Duplex.first; 359 OS << "; "; 360 Inst = Duplex.second; 361 } 362 else 363 Inst = HeadTail.first; 364 OS << Inst; 365 Bytes = Bytes.slice(4); 366 Address += 4; 367 HeadTail = HeadTail.second.split('\n'); 368 } 369 OS << " } " << PacketBundle.second; 370 } 371}; 372HexagonPrettyPrinter HexagonPrettyPrinterInst; 373PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 374 switch(Triple.getArch()) { 375 default: 376 return PrettyPrinterInst; 377 case Triple::hexagon: 378 return HexagonPrettyPrinterInst; 379 } 380} 381} 382 383template <class ELFT> 384static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 385 const RelocationRef &RelRef, 386 SmallVectorImpl<char> &Result) { 387 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 388 389 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 390 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 391 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 392 393 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 394 395 ErrorOr<const Elf_Shdr *> SecOrErr = EF.getSection(Rel.d.a); 396 if (std::error_code EC = SecOrErr.getError()) 397 return EC; 398 const Elf_Shdr *Sec = *SecOrErr; 399 ErrorOr<const Elf_Shdr *> SymTabOrErr = EF.getSection(Sec->sh_link); 400 if (std::error_code EC = SymTabOrErr.getError()) 401 return EC; 402 const Elf_Shdr *SymTab = *SymTabOrErr; 403 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 404 SymTab->sh_type == ELF::SHT_DYNSYM); 405 ErrorOr<const Elf_Shdr *> StrTabSec = EF.getSection(SymTab->sh_link); 406 if (std::error_code EC = StrTabSec.getError()) 407 return EC; 408 ErrorOr<StringRef> StrTabOrErr = EF.getStringTable(*StrTabSec); 409 if (std::error_code EC = StrTabOrErr.getError()) 410 return EC; 411 StringRef StrTab = *StrTabOrErr; 412 uint8_t type = RelRef.getType(); 413 StringRef res; 414 int64_t addend = 0; 415 switch (Sec->sh_type) { 416 default: 417 return object_error::parse_failed; 418 case ELF::SHT_REL: { 419 // TODO: Read implicit addend from section data. 420 break; 421 } 422 case ELF::SHT_RELA: { 423 const Elf_Rela *ERela = Obj->getRela(Rel); 424 addend = ERela->r_addend; 425 break; 426 } 427 } 428 symbol_iterator SI = RelRef.getSymbol(); 429 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 430 StringRef Target; 431 if (symb->getType() == ELF::STT_SECTION) { 432 ErrorOr<section_iterator> SymSI = SI->getSection(); 433 if (std::error_code EC = SymSI.getError()) 434 return EC; 435 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 436 ErrorOr<StringRef> SecName = EF.getSectionName(SymSec); 437 if (std::error_code EC = SecName.getError()) 438 return EC; 439 Target = *SecName; 440 } else { 441 ErrorOr<StringRef> SymName = symb->getName(StrTab); 442 if (!SymName) 443 return SymName.getError(); 444 Target = *SymName; 445 } 446 switch (EF.getHeader()->e_machine) { 447 case ELF::EM_X86_64: 448 switch (type) { 449 case ELF::R_X86_64_PC8: 450 case ELF::R_X86_64_PC16: 451 case ELF::R_X86_64_PC32: { 452 std::string fmtbuf; 453 raw_string_ostream fmt(fmtbuf); 454 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 455 fmt.flush(); 456 Result.append(fmtbuf.begin(), fmtbuf.end()); 457 } break; 458 case ELF::R_X86_64_8: 459 case ELF::R_X86_64_16: 460 case ELF::R_X86_64_32: 461 case ELF::R_X86_64_32S: 462 case ELF::R_X86_64_64: { 463 std::string fmtbuf; 464 raw_string_ostream fmt(fmtbuf); 465 fmt << Target << (addend < 0 ? "" : "+") << addend; 466 fmt.flush(); 467 Result.append(fmtbuf.begin(), fmtbuf.end()); 468 } break; 469 default: 470 res = "Unknown"; 471 } 472 break; 473 case ELF::EM_AARCH64: { 474 std::string fmtbuf; 475 raw_string_ostream fmt(fmtbuf); 476 fmt << Target; 477 if (addend != 0) 478 fmt << (addend < 0 ? "" : "+") << addend; 479 fmt.flush(); 480 Result.append(fmtbuf.begin(), fmtbuf.end()); 481 break; 482 } 483 case ELF::EM_386: 484 case ELF::EM_IAMCU: 485 case ELF::EM_ARM: 486 case ELF::EM_HEXAGON: 487 case ELF::EM_MIPS: 488 res = Target; 489 break; 490 case ELF::EM_WEBASSEMBLY: 491 switch (type) { 492 case ELF::R_WEBASSEMBLY_DATA: { 493 std::string fmtbuf; 494 raw_string_ostream fmt(fmtbuf); 495 fmt << Target << (addend < 0 ? "" : "+") << addend; 496 fmt.flush(); 497 Result.append(fmtbuf.begin(), fmtbuf.end()); 498 break; 499 } 500 case ELF::R_WEBASSEMBLY_FUNCTION: 501 res = Target; 502 break; 503 default: 504 res = "Unknown"; 505 } 506 break; 507 default: 508 res = "Unknown"; 509 } 510 if (Result.empty()) 511 Result.append(res.begin(), res.end()); 512 return std::error_code(); 513} 514 515static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 516 const RelocationRef &Rel, 517 SmallVectorImpl<char> &Result) { 518 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 519 return getRelocationValueString(ELF32LE, Rel, Result); 520 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 521 return getRelocationValueString(ELF64LE, Rel, Result); 522 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 523 return getRelocationValueString(ELF32BE, Rel, Result); 524 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 525 return getRelocationValueString(ELF64BE, Rel, Result); 526} 527 528static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 529 const RelocationRef &Rel, 530 SmallVectorImpl<char> &Result) { 531 symbol_iterator SymI = Rel.getSymbol(); 532 ErrorOr<StringRef> SymNameOrErr = SymI->getName(); 533 if (std::error_code EC = SymNameOrErr.getError()) 534 return EC; 535 StringRef SymName = *SymNameOrErr; 536 Result.append(SymName.begin(), SymName.end()); 537 return std::error_code(); 538} 539 540static void printRelocationTargetName(const MachOObjectFile *O, 541 const MachO::any_relocation_info &RE, 542 raw_string_ostream &fmt) { 543 bool IsScattered = O->isRelocationScattered(RE); 544 545 // Target of a scattered relocation is an address. In the interest of 546 // generating pretty output, scan through the symbol table looking for a 547 // symbol that aligns with that address. If we find one, print it. 548 // Otherwise, we just print the hex address of the target. 549 if (IsScattered) { 550 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 551 552 for (const SymbolRef &Symbol : O->symbols()) { 553 std::error_code ec; 554 ErrorOr<uint64_t> Addr = Symbol.getAddress(); 555 if ((ec = Addr.getError())) 556 report_fatal_error(ec.message()); 557 if (*Addr != Val) 558 continue; 559 ErrorOr<StringRef> Name = Symbol.getName(); 560 if (std::error_code EC = Name.getError()) 561 report_fatal_error(EC.message()); 562 fmt << *Name; 563 return; 564 } 565 566 // If we couldn't find a symbol that this relocation refers to, try 567 // to find a section beginning instead. 568 for (const SectionRef &Section : ToolSectionFilter(*O)) { 569 std::error_code ec; 570 571 StringRef Name; 572 uint64_t Addr = Section.getAddress(); 573 if (Addr != Val) 574 continue; 575 if ((ec = Section.getName(Name))) 576 report_fatal_error(ec.message()); 577 fmt << Name; 578 return; 579 } 580 581 fmt << format("0x%x", Val); 582 return; 583 } 584 585 StringRef S; 586 bool isExtern = O->getPlainRelocationExternal(RE); 587 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 588 589 if (isExtern) { 590 symbol_iterator SI = O->symbol_begin(); 591 advance(SI, Val); 592 ErrorOr<StringRef> SOrErr = SI->getName(); 593 error(SOrErr.getError()); 594 S = *SOrErr; 595 } else { 596 section_iterator SI = O->section_begin(); 597 // Adjust for the fact that sections are 1-indexed. 598 advance(SI, Val - 1); 599 SI->getName(S); 600 } 601 602 fmt << S; 603} 604 605static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 606 const RelocationRef &RelRef, 607 SmallVectorImpl<char> &Result) { 608 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 609 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 610 611 unsigned Arch = Obj->getArch(); 612 613 std::string fmtbuf; 614 raw_string_ostream fmt(fmtbuf); 615 unsigned Type = Obj->getAnyRelocationType(RE); 616 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 617 618 // Determine any addends that should be displayed with the relocation. 619 // These require decoding the relocation type, which is triple-specific. 620 621 // X86_64 has entirely custom relocation types. 622 if (Arch == Triple::x86_64) { 623 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 624 625 switch (Type) { 626 case MachO::X86_64_RELOC_GOT_LOAD: 627 case MachO::X86_64_RELOC_GOT: { 628 printRelocationTargetName(Obj, RE, fmt); 629 fmt << "@GOT"; 630 if (isPCRel) 631 fmt << "PCREL"; 632 break; 633 } 634 case MachO::X86_64_RELOC_SUBTRACTOR: { 635 DataRefImpl RelNext = Rel; 636 Obj->moveRelocationNext(RelNext); 637 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 638 639 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 640 // X86_64_RELOC_UNSIGNED. 641 // NOTE: Scattered relocations don't exist on x86_64. 642 unsigned RType = Obj->getAnyRelocationType(RENext); 643 if (RType != MachO::X86_64_RELOC_UNSIGNED) 644 report_fatal_error("Expected X86_64_RELOC_UNSIGNED after " 645 "X86_64_RELOC_SUBTRACTOR."); 646 647 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 648 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 649 printRelocationTargetName(Obj, RENext, fmt); 650 fmt << "-"; 651 printRelocationTargetName(Obj, RE, fmt); 652 break; 653 } 654 case MachO::X86_64_RELOC_TLV: 655 printRelocationTargetName(Obj, RE, fmt); 656 fmt << "@TLV"; 657 if (isPCRel) 658 fmt << "P"; 659 break; 660 case MachO::X86_64_RELOC_SIGNED_1: 661 printRelocationTargetName(Obj, RE, fmt); 662 fmt << "-1"; 663 break; 664 case MachO::X86_64_RELOC_SIGNED_2: 665 printRelocationTargetName(Obj, RE, fmt); 666 fmt << "-2"; 667 break; 668 case MachO::X86_64_RELOC_SIGNED_4: 669 printRelocationTargetName(Obj, RE, fmt); 670 fmt << "-4"; 671 break; 672 default: 673 printRelocationTargetName(Obj, RE, fmt); 674 break; 675 } 676 // X86 and ARM share some relocation types in common. 677 } else if (Arch == Triple::x86 || Arch == Triple::arm || 678 Arch == Triple::ppc) { 679 // Generic relocation types... 680 switch (Type) { 681 case MachO::GENERIC_RELOC_PAIR: // prints no info 682 return std::error_code(); 683 case MachO::GENERIC_RELOC_SECTDIFF: { 684 DataRefImpl RelNext = Rel; 685 Obj->moveRelocationNext(RelNext); 686 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 687 688 // X86 sect diff's must be followed by a relocation of type 689 // GENERIC_RELOC_PAIR. 690 unsigned RType = Obj->getAnyRelocationType(RENext); 691 692 if (RType != MachO::GENERIC_RELOC_PAIR) 693 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 694 "GENERIC_RELOC_SECTDIFF."); 695 696 printRelocationTargetName(Obj, RE, fmt); 697 fmt << "-"; 698 printRelocationTargetName(Obj, RENext, fmt); 699 break; 700 } 701 } 702 703 if (Arch == Triple::x86 || Arch == Triple::ppc) { 704 switch (Type) { 705 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 706 DataRefImpl RelNext = Rel; 707 Obj->moveRelocationNext(RelNext); 708 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 709 710 // X86 sect diff's must be followed by a relocation of type 711 // GENERIC_RELOC_PAIR. 712 unsigned RType = Obj->getAnyRelocationType(RENext); 713 if (RType != MachO::GENERIC_RELOC_PAIR) 714 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 715 "GENERIC_RELOC_LOCAL_SECTDIFF."); 716 717 printRelocationTargetName(Obj, RE, fmt); 718 fmt << "-"; 719 printRelocationTargetName(Obj, RENext, fmt); 720 break; 721 } 722 case MachO::GENERIC_RELOC_TLV: { 723 printRelocationTargetName(Obj, RE, fmt); 724 fmt << "@TLV"; 725 if (IsPCRel) 726 fmt << "P"; 727 break; 728 } 729 default: 730 printRelocationTargetName(Obj, RE, fmt); 731 } 732 } else { // ARM-specific relocations 733 switch (Type) { 734 case MachO::ARM_RELOC_HALF: 735 case MachO::ARM_RELOC_HALF_SECTDIFF: { 736 // Half relocations steal a bit from the length field to encode 737 // whether this is an upper16 or a lower16 relocation. 738 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 739 740 if (isUpper) 741 fmt << ":upper16:("; 742 else 743 fmt << ":lower16:("; 744 printRelocationTargetName(Obj, RE, fmt); 745 746 DataRefImpl RelNext = Rel; 747 Obj->moveRelocationNext(RelNext); 748 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 749 750 // ARM half relocs must be followed by a relocation of type 751 // ARM_RELOC_PAIR. 752 unsigned RType = Obj->getAnyRelocationType(RENext); 753 if (RType != MachO::ARM_RELOC_PAIR) 754 report_fatal_error("Expected ARM_RELOC_PAIR after " 755 "ARM_RELOC_HALF"); 756 757 // NOTE: The half of the target virtual address is stashed in the 758 // address field of the secondary relocation, but we can't reverse 759 // engineer the constant offset from it without decoding the movw/movt 760 // instruction to find the other half in its immediate field. 761 762 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 763 // symbol/section pointer of the follow-on relocation. 764 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 765 fmt << "-"; 766 printRelocationTargetName(Obj, RENext, fmt); 767 } 768 769 fmt << ")"; 770 break; 771 } 772 default: { printRelocationTargetName(Obj, RE, fmt); } 773 } 774 } 775 } else 776 printRelocationTargetName(Obj, RE, fmt); 777 778 fmt.flush(); 779 Result.append(fmtbuf.begin(), fmtbuf.end()); 780 return std::error_code(); 781} 782 783static std::error_code getRelocationValueString(const RelocationRef &Rel, 784 SmallVectorImpl<char> &Result) { 785 const ObjectFile *Obj = Rel.getObject(); 786 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 787 return getRelocationValueString(ELF, Rel, Result); 788 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 789 return getRelocationValueString(COFF, Rel, Result); 790 auto *MachO = cast<MachOObjectFile>(Obj); 791 return getRelocationValueString(MachO, Rel, Result); 792} 793 794/// @brief Indicates whether this relocation should hidden when listing 795/// relocations, usually because it is the trailing part of a multipart 796/// relocation that will be printed as part of the leading relocation. 797static bool getHidden(RelocationRef RelRef) { 798 const ObjectFile *Obj = RelRef.getObject(); 799 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 800 if (!MachO) 801 return false; 802 803 unsigned Arch = MachO->getArch(); 804 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 805 uint64_t Type = MachO->getRelocationType(Rel); 806 807 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 808 // is always hidden. 809 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 810 if (Type == MachO::GENERIC_RELOC_PAIR) 811 return true; 812 } else if (Arch == Triple::x86_64) { 813 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 814 // an X86_64_RELOC_SUBTRACTOR. 815 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 816 DataRefImpl RelPrev = Rel; 817 RelPrev.d.a--; 818 uint64_t PrevType = MachO->getRelocationType(RelPrev); 819 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 820 return true; 821 } 822 } 823 824 return false; 825} 826 827static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 828 const Target *TheTarget = getTarget(Obj); 829 830 // Package up features to be passed to target/subtarget 831 std::string FeaturesStr; 832 if (MAttrs.size()) { 833 SubtargetFeatures Features; 834 for (unsigned i = 0; i != MAttrs.size(); ++i) 835 Features.AddFeature(MAttrs[i]); 836 FeaturesStr = Features.getString(); 837 } 838 839 std::unique_ptr<const MCRegisterInfo> MRI( 840 TheTarget->createMCRegInfo(TripleName)); 841 if (!MRI) 842 report_fatal_error("error: no register info for target " + TripleName); 843 844 // Set up disassembler. 845 std::unique_ptr<const MCAsmInfo> AsmInfo( 846 TheTarget->createMCAsmInfo(*MRI, TripleName)); 847 if (!AsmInfo) 848 report_fatal_error("error: no assembly info for target " + TripleName); 849 std::unique_ptr<const MCSubtargetInfo> STI( 850 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 851 if (!STI) 852 report_fatal_error("error: no subtarget info for target " + TripleName); 853 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 854 if (!MII) 855 report_fatal_error("error: no instruction info for target " + TripleName); 856 std::unique_ptr<const MCObjectFileInfo> MOFI(new MCObjectFileInfo); 857 MCContext Ctx(AsmInfo.get(), MRI.get(), MOFI.get()); 858 859 std::unique_ptr<MCDisassembler> DisAsm( 860 TheTarget->createMCDisassembler(*STI, Ctx)); 861 if (!DisAsm) 862 report_fatal_error("error: no disassembler for target " + TripleName); 863 864 std::unique_ptr<const MCInstrAnalysis> MIA( 865 TheTarget->createMCInstrAnalysis(MII.get())); 866 867 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 868 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 869 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 870 if (!IP) 871 report_fatal_error("error: no instruction printer for target " + 872 TripleName); 873 IP->setPrintImmHex(PrintImmHex); 874 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 875 876 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 877 "\t\t\t%08" PRIx64 ": "; 878 879 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 880 // in RelocSecs contain the relocations for section S. 881 std::error_code EC; 882 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 883 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 884 section_iterator Sec2 = Section.getRelocatedSection(); 885 if (Sec2 != Obj->section_end()) 886 SectionRelocMap[*Sec2].push_back(Section); 887 } 888 889 // Create a mapping from virtual address to symbol name. This is used to 890 // pretty print the symbols while disassembling. 891 typedef std::vector<std::pair<uint64_t, StringRef>> SectionSymbolsTy; 892 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 893 for (const SymbolRef &Symbol : Obj->symbols()) { 894 ErrorOr<uint64_t> AddressOrErr = Symbol.getAddress(); 895 error(AddressOrErr.getError()); 896 uint64_t Address = *AddressOrErr; 897 898 ErrorOr<StringRef> Name = Symbol.getName(); 899 error(Name.getError()); 900 if (Name->empty()) 901 continue; 902 903 ErrorOr<section_iterator> SectionOrErr = Symbol.getSection(); 904 error(SectionOrErr.getError()); 905 section_iterator SecI = *SectionOrErr; 906 if (SecI == Obj->section_end()) 907 continue; 908 909 AllSymbols[*SecI].emplace_back(Address, *Name); 910 } 911 912 // Create a mapping from virtual address to section. 913 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 914 for (SectionRef Sec : Obj->sections()) 915 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 916 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 917 918 // Linked executables (.exe and .dll files) typically don't include a real 919 // symbol table but they might contain an export table. 920 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 921 for (const auto &ExportEntry : COFFObj->export_directories()) { 922 StringRef Name; 923 error(ExportEntry.getSymbolName(Name)); 924 if (Name.empty()) 925 continue; 926 uint32_t RVA; 927 error(ExportEntry.getExportRVA(RVA)); 928 929 uint64_t VA = COFFObj->getImageBase() + RVA; 930 auto Sec = std::upper_bound( 931 SectionAddresses.begin(), SectionAddresses.end(), VA, 932 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 933 return LHS < RHS.first; 934 }); 935 if (Sec != SectionAddresses.begin()) 936 --Sec; 937 else 938 Sec = SectionAddresses.end(); 939 940 if (Sec != SectionAddresses.end()) 941 AllSymbols[Sec->second].emplace_back(VA, Name); 942 } 943 } 944 945 // Sort all the symbols, this allows us to use a simple binary search to find 946 // a symbol near an address. 947 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 948 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 949 950 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 951 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 952 continue; 953 954 uint64_t SectionAddr = Section.getAddress(); 955 uint64_t SectSize = Section.getSize(); 956 if (!SectSize) 957 continue; 958 959 // Get the list of all the symbols in this section. 960 SectionSymbolsTy &Symbols = AllSymbols[Section]; 961 std::vector<uint64_t> DataMappingSymsAddr; 962 std::vector<uint64_t> TextMappingSymsAddr; 963 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 964 for (const auto &Symb : Symbols) { 965 uint64_t Address = Symb.first; 966 StringRef Name = Symb.second; 967 if (Name.startswith("$d")) 968 DataMappingSymsAddr.push_back(Address - SectionAddr); 969 if (Name.startswith("$x")) 970 TextMappingSymsAddr.push_back(Address - SectionAddr); 971 } 972 } 973 974 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 975 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 976 977 // Make a list of all the relocations for this section. 978 std::vector<RelocationRef> Rels; 979 if (InlineRelocs) { 980 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 981 for (const RelocationRef &Reloc : RelocSec.relocations()) { 982 Rels.push_back(Reloc); 983 } 984 } 985 } 986 987 // Sort relocations by address. 988 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 989 990 StringRef SegmentName = ""; 991 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 992 DataRefImpl DR = Section.getRawDataRefImpl(); 993 SegmentName = MachO->getSectionFinalSegmentName(DR); 994 } 995 StringRef name; 996 error(Section.getName(name)); 997 outs() << "Disassembly of section "; 998 if (!SegmentName.empty()) 999 outs() << SegmentName << ","; 1000 outs() << name << ':'; 1001 1002 // If the section has no symbol at the start, just insert a dummy one. 1003 if (Symbols.empty() || Symbols[0].first != 0) 1004 Symbols.insert(Symbols.begin(), std::make_pair(SectionAddr, name)); 1005 1006 SmallString<40> Comments; 1007 raw_svector_ostream CommentStream(Comments); 1008 1009 StringRef BytesStr; 1010 error(Section.getContents(BytesStr)); 1011 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1012 BytesStr.size()); 1013 1014 uint64_t Size; 1015 uint64_t Index; 1016 1017 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1018 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1019 // Disassemble symbol by symbol. 1020 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1021 1022 uint64_t Start = Symbols[si].first - SectionAddr; 1023 // The end is either the section end or the beginning of the next 1024 // symbol. 1025 uint64_t End = 1026 (si == se - 1) ? SectSize : Symbols[si + 1].first - SectionAddr; 1027 // Don't try to disassemble beyond the end of section contents. 1028 if (End > SectSize) 1029 End = SectSize; 1030 // If this symbol has the same address as the next symbol, then skip it. 1031 if (Start >= End) 1032 continue; 1033 1034 outs() << '\n' << Symbols[si].second << ":\n"; 1035 1036#ifndef NDEBUG 1037 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1038#else 1039 raw_ostream &DebugOut = nulls(); 1040#endif 1041 1042 for (Index = Start; Index < End; Index += Size) { 1043 MCInst Inst; 1044 1045 // AArch64 ELF binaries can interleave data and text in the 1046 // same section. We rely on the markers introduced to 1047 // understand what we need to dump. 1048 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 1049 uint64_t Stride = 0; 1050 1051 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1052 DataMappingSymsAddr.end(), Index); 1053 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1054 // Switch to data. 1055 while (Index < End) { 1056 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1057 outs() << "\t"; 1058 if (Index + 4 <= End) { 1059 Stride = 4; 1060 dumpBytes(Bytes.slice(Index, 4), outs()); 1061 outs() << "\t.word"; 1062 } else if (Index + 2 <= End) { 1063 Stride = 2; 1064 dumpBytes(Bytes.slice(Index, 2), outs()); 1065 outs() << "\t.short"; 1066 } else { 1067 Stride = 1; 1068 dumpBytes(Bytes.slice(Index, 1), outs()); 1069 outs() << "\t.byte"; 1070 } 1071 Index += Stride; 1072 outs() << "\n"; 1073 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1074 TextMappingSymsAddr.end(), Index); 1075 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1076 break; 1077 } 1078 } 1079 } 1080 1081 if (Index >= End) 1082 break; 1083 1084 if (DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1085 SectionAddr + Index, DebugOut, 1086 CommentStream)) { 1087 PIP.printInst(*IP, &Inst, 1088 Bytes.slice(Index, Size), 1089 SectionAddr + Index, outs(), "", *STI); 1090 outs() << CommentStream.str(); 1091 Comments.clear(); 1092 1093 // Try to resolve the target of a call, tail call, etc. to a specific 1094 // symbol. 1095 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1096 MIA->isConditionalBranch(Inst))) { 1097 uint64_t Target; 1098 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1099 // In a relocatable object, the target's section must reside in 1100 // the same section as the call instruction or it is accessed 1101 // through a relocation. 1102 // 1103 // In a non-relocatable object, the target may be in any section. 1104 // 1105 // N.B. We don't walk the relocations in the relocatable case yet. 1106 auto *TargetSectionSymbols = &Symbols; 1107 if (!Obj->isRelocatableObject()) { 1108 auto SectionAddress = std::upper_bound( 1109 SectionAddresses.begin(), SectionAddresses.end(), Target, 1110 [](uint64_t LHS, 1111 const std::pair<uint64_t, SectionRef> &RHS) { 1112 return LHS < RHS.first; 1113 }); 1114 if (SectionAddress != SectionAddresses.begin()) { 1115 --SectionAddress; 1116 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1117 } else { 1118 TargetSectionSymbols = nullptr; 1119 } 1120 } 1121 1122 // Find the first symbol in the section whose offset is less than 1123 // or equal to the target. 1124 if (TargetSectionSymbols) { 1125 auto TargetSym = std::upper_bound( 1126 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1127 Target, [](uint64_t LHS, 1128 const std::pair<uint64_t, StringRef> &RHS) { 1129 return LHS < RHS.first; 1130 }); 1131 if (TargetSym != TargetSectionSymbols->begin()) { 1132 --TargetSym; 1133 uint64_t TargetAddress = std::get<0>(*TargetSym); 1134 StringRef TargetName = std::get<1>(*TargetSym); 1135 outs() << " <" << TargetName; 1136 uint64_t Disp = Target - TargetAddress; 1137 if (Disp) 1138 outs() << '+' << utohexstr(Disp); 1139 outs() << '>'; 1140 } 1141 } 1142 } 1143 } 1144 outs() << "\n"; 1145 } else { 1146 errs() << ToolName << ": warning: invalid instruction encoding\n"; 1147 if (Size == 0) 1148 Size = 1; // skip illegible bytes 1149 } 1150 1151 // Print relocation for instruction. 1152 while (rel_cur != rel_end) { 1153 bool hidden = getHidden(*rel_cur); 1154 uint64_t addr = rel_cur->getOffset(); 1155 SmallString<16> name; 1156 SmallString<32> val; 1157 1158 // If this relocation is hidden, skip it. 1159 if (hidden) goto skip_print_rel; 1160 1161 // Stop when rel_cur's address is past the current instruction. 1162 if (addr >= Index + Size) break; 1163 rel_cur->getTypeName(name); 1164 error(getRelocationValueString(*rel_cur, val)); 1165 outs() << format(Fmt.data(), SectionAddr + addr) << name 1166 << "\t" << val << "\n"; 1167 1168 skip_print_rel: 1169 ++rel_cur; 1170 } 1171 } 1172 } 1173 } 1174} 1175 1176void llvm::PrintRelocations(const ObjectFile *Obj) { 1177 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1178 "%08" PRIx64; 1179 // Regular objdump doesn't print relocations in non-relocatable object 1180 // files. 1181 if (!Obj->isRelocatableObject()) 1182 return; 1183 1184 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1185 if (Section.relocation_begin() == Section.relocation_end()) 1186 continue; 1187 StringRef secname; 1188 error(Section.getName(secname)); 1189 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1190 for (const RelocationRef &Reloc : Section.relocations()) { 1191 bool hidden = getHidden(Reloc); 1192 uint64_t address = Reloc.getOffset(); 1193 SmallString<32> relocname; 1194 SmallString<32> valuestr; 1195 if (hidden) 1196 continue; 1197 Reloc.getTypeName(relocname); 1198 error(getRelocationValueString(Reloc, valuestr)); 1199 outs() << format(Fmt.data(), address) << " " << relocname << " " 1200 << valuestr << "\n"; 1201 } 1202 outs() << "\n"; 1203 } 1204} 1205 1206void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1207 outs() << "Sections:\n" 1208 "Idx Name Size Address Type\n"; 1209 unsigned i = 0; 1210 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1211 StringRef Name; 1212 error(Section.getName(Name)); 1213 uint64_t Address = Section.getAddress(); 1214 uint64_t Size = Section.getSize(); 1215 bool Text = Section.isText(); 1216 bool Data = Section.isData(); 1217 bool BSS = Section.isBSS(); 1218 std::string Type = (std::string(Text ? "TEXT " : "") + 1219 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1220 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1221 Name.str().c_str(), Size, Address, Type.c_str()); 1222 ++i; 1223 } 1224} 1225 1226void llvm::PrintSectionContents(const ObjectFile *Obj) { 1227 std::error_code EC; 1228 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1229 StringRef Name; 1230 StringRef Contents; 1231 error(Section.getName(Name)); 1232 uint64_t BaseAddr = Section.getAddress(); 1233 uint64_t Size = Section.getSize(); 1234 if (!Size) 1235 continue; 1236 1237 outs() << "Contents of section " << Name << ":\n"; 1238 if (Section.isBSS()) { 1239 outs() << format("<skipping contents of bss section at [%04" PRIx64 1240 ", %04" PRIx64 ")>\n", 1241 BaseAddr, BaseAddr + Size); 1242 continue; 1243 } 1244 1245 error(Section.getContents(Contents)); 1246 1247 // Dump out the content as hex and printable ascii characters. 1248 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1249 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1250 // Dump line of hex. 1251 for (std::size_t i = 0; i < 16; ++i) { 1252 if (i != 0 && i % 4 == 0) 1253 outs() << ' '; 1254 if (addr + i < end) 1255 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1256 << hexdigit(Contents[addr + i] & 0xF, true); 1257 else 1258 outs() << " "; 1259 } 1260 // Print ascii. 1261 outs() << " "; 1262 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1263 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1264 outs() << Contents[addr + i]; 1265 else 1266 outs() << "."; 1267 } 1268 outs() << "\n"; 1269 } 1270 } 1271} 1272 1273void llvm::PrintSymbolTable(const ObjectFile *o) { 1274 outs() << "SYMBOL TABLE:\n"; 1275 1276 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1277 printCOFFSymbolTable(coff); 1278 return; 1279 } 1280 for (const SymbolRef &Symbol : o->symbols()) { 1281 ErrorOr<uint64_t> AddressOrError = Symbol.getAddress(); 1282 error(AddressOrError.getError()); 1283 uint64_t Address = *AddressOrError; 1284 SymbolRef::Type Type = Symbol.getType(); 1285 uint32_t Flags = Symbol.getFlags(); 1286 ErrorOr<section_iterator> SectionOrErr = Symbol.getSection(); 1287 error(SectionOrErr.getError()); 1288 section_iterator Section = *SectionOrErr; 1289 StringRef Name; 1290 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1291 Section->getName(Name); 1292 } else { 1293 ErrorOr<StringRef> NameOrErr = Symbol.getName(); 1294 error(NameOrErr.getError()); 1295 Name = *NameOrErr; 1296 } 1297 1298 bool Global = Flags & SymbolRef::SF_Global; 1299 bool Weak = Flags & SymbolRef::SF_Weak; 1300 bool Absolute = Flags & SymbolRef::SF_Absolute; 1301 bool Common = Flags & SymbolRef::SF_Common; 1302 bool Hidden = Flags & SymbolRef::SF_Hidden; 1303 1304 char GlobLoc = ' '; 1305 if (Type != SymbolRef::ST_Unknown) 1306 GlobLoc = Global ? 'g' : 'l'; 1307 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1308 ? 'd' : ' '; 1309 char FileFunc = ' '; 1310 if (Type == SymbolRef::ST_File) 1311 FileFunc = 'f'; 1312 else if (Type == SymbolRef::ST_Function) 1313 FileFunc = 'F'; 1314 1315 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1316 "%08" PRIx64; 1317 1318 outs() << format(Fmt, Address) << " " 1319 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1320 << (Weak ? 'w' : ' ') // Weak? 1321 << ' ' // Constructor. Not supported yet. 1322 << ' ' // Warning. Not supported yet. 1323 << ' ' // Indirect reference to another symbol. 1324 << Debug // Debugging (d) or dynamic (D) symbol. 1325 << FileFunc // Name of function (F), file (f) or object (O). 1326 << ' '; 1327 if (Absolute) { 1328 outs() << "*ABS*"; 1329 } else if (Common) { 1330 outs() << "*COM*"; 1331 } else if (Section == o->section_end()) { 1332 outs() << "*UND*"; 1333 } else { 1334 if (const MachOObjectFile *MachO = 1335 dyn_cast<const MachOObjectFile>(o)) { 1336 DataRefImpl DR = Section->getRawDataRefImpl(); 1337 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1338 outs() << SegmentName << ","; 1339 } 1340 StringRef SectionName; 1341 error(Section->getName(SectionName)); 1342 outs() << SectionName; 1343 } 1344 1345 outs() << '\t'; 1346 if (Common || isa<ELFObjectFileBase>(o)) { 1347 uint64_t Val = 1348 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1349 outs() << format("\t %08" PRIx64 " ", Val); 1350 } 1351 1352 if (Hidden) { 1353 outs() << ".hidden "; 1354 } 1355 outs() << Name 1356 << '\n'; 1357 } 1358} 1359 1360static void PrintUnwindInfo(const ObjectFile *o) { 1361 outs() << "Unwind info:\n\n"; 1362 1363 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1364 printCOFFUnwindInfo(coff); 1365 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1366 printMachOUnwindInfo(MachO); 1367 else { 1368 // TODO: Extract DWARF dump tool to objdump. 1369 errs() << "This operation is only currently supported " 1370 "for COFF and MachO object files.\n"; 1371 return; 1372 } 1373} 1374 1375void llvm::printExportsTrie(const ObjectFile *o) { 1376 outs() << "Exports trie:\n"; 1377 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1378 printMachOExportsTrie(MachO); 1379 else { 1380 errs() << "This operation is only currently supported " 1381 "for Mach-O executable files.\n"; 1382 return; 1383 } 1384} 1385 1386void llvm::printRebaseTable(const ObjectFile *o) { 1387 outs() << "Rebase table:\n"; 1388 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1389 printMachORebaseTable(MachO); 1390 else { 1391 errs() << "This operation is only currently supported " 1392 "for Mach-O executable files.\n"; 1393 return; 1394 } 1395} 1396 1397void llvm::printBindTable(const ObjectFile *o) { 1398 outs() << "Bind table:\n"; 1399 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1400 printMachOBindTable(MachO); 1401 else { 1402 errs() << "This operation is only currently supported " 1403 "for Mach-O executable files.\n"; 1404 return; 1405 } 1406} 1407 1408void llvm::printLazyBindTable(const ObjectFile *o) { 1409 outs() << "Lazy bind table:\n"; 1410 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1411 printMachOLazyBindTable(MachO); 1412 else { 1413 errs() << "This operation is only currently supported " 1414 "for Mach-O executable files.\n"; 1415 return; 1416 } 1417} 1418 1419void llvm::printWeakBindTable(const ObjectFile *o) { 1420 outs() << "Weak bind table:\n"; 1421 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1422 printMachOWeakBindTable(MachO); 1423 else { 1424 errs() << "This operation is only currently supported " 1425 "for Mach-O executable files.\n"; 1426 return; 1427 } 1428} 1429 1430/// Dump the raw contents of the __clangast section so the output can be piped 1431/// into llvm-bcanalyzer. 1432void llvm::printRawClangAST(const ObjectFile *Obj) { 1433 if (outs().is_displayed()) { 1434 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1435 "the clang ast section.\n" 1436 "Please redirect the output to a file or another program such as " 1437 "llvm-bcanalyzer.\n"; 1438 return; 1439 } 1440 1441 StringRef ClangASTSectionName("__clangast"); 1442 if (isa<COFFObjectFile>(Obj)) { 1443 ClangASTSectionName = "clangast"; 1444 } 1445 1446 Optional<object::SectionRef> ClangASTSection; 1447 for (auto Sec : ToolSectionFilter(*Obj)) { 1448 StringRef Name; 1449 Sec.getName(Name); 1450 if (Name == ClangASTSectionName) { 1451 ClangASTSection = Sec; 1452 break; 1453 } 1454 } 1455 if (!ClangASTSection) 1456 return; 1457 1458 StringRef ClangASTContents; 1459 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1460 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1461} 1462 1463static void printFaultMaps(const ObjectFile *Obj) { 1464 const char *FaultMapSectionName = nullptr; 1465 1466 if (isa<ELFObjectFileBase>(Obj)) { 1467 FaultMapSectionName = ".llvm_faultmaps"; 1468 } else if (isa<MachOObjectFile>(Obj)) { 1469 FaultMapSectionName = "__llvm_faultmaps"; 1470 } else { 1471 errs() << "This operation is only currently supported " 1472 "for ELF and Mach-O executable files.\n"; 1473 return; 1474 } 1475 1476 Optional<object::SectionRef> FaultMapSection; 1477 1478 for (auto Sec : ToolSectionFilter(*Obj)) { 1479 StringRef Name; 1480 Sec.getName(Name); 1481 if (Name == FaultMapSectionName) { 1482 FaultMapSection = Sec; 1483 break; 1484 } 1485 } 1486 1487 outs() << "FaultMap table:\n"; 1488 1489 if (!FaultMapSection.hasValue()) { 1490 outs() << "<not found>\n"; 1491 return; 1492 } 1493 1494 StringRef FaultMapContents; 1495 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1496 1497 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1498 FaultMapContents.bytes_end()); 1499 1500 outs() << FMP; 1501} 1502 1503static void printPrivateFileHeaders(const ObjectFile *o) { 1504 if (o->isELF()) 1505 printELFFileHeader(o); 1506 else if (o->isCOFF()) 1507 printCOFFFileHeader(o); 1508 else if (o->isMachO()) { 1509 printMachOFileHeader(o); 1510 printMachOLoadCommands(o); 1511 } else 1512 report_fatal_error("Invalid/Unsupported object file format"); 1513} 1514 1515static void printFirstPrivateFileHeader(const ObjectFile *o) { 1516 if (o->isELF()) 1517 printELFFileHeader(o); 1518 else if (o->isCOFF()) 1519 printCOFFFileHeader(o); 1520 else if (o->isMachO()) 1521 printMachOFileHeader(o); 1522 else 1523 report_fatal_error("Invalid/Unsupported object file format"); 1524} 1525 1526static void DumpObject(const ObjectFile *o) { 1527 // Avoid other output when using a raw option. 1528 if (!RawClangAST) { 1529 outs() << '\n'; 1530 outs() << o->getFileName() 1531 << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1532 } 1533 1534 if (Disassemble) 1535 DisassembleObject(o, Relocations); 1536 if (Relocations && !Disassemble) 1537 PrintRelocations(o); 1538 if (SectionHeaders) 1539 PrintSectionHeaders(o); 1540 if (SectionContents) 1541 PrintSectionContents(o); 1542 if (SymbolTable) 1543 PrintSymbolTable(o); 1544 if (UnwindInfo) 1545 PrintUnwindInfo(o); 1546 if (PrivateHeaders) 1547 printPrivateFileHeaders(o); 1548 if (FirstPrivateHeader) 1549 printFirstPrivateFileHeader(o); 1550 if (ExportsTrie) 1551 printExportsTrie(o); 1552 if (Rebase) 1553 printRebaseTable(o); 1554 if (Bind) 1555 printBindTable(o); 1556 if (LazyBind) 1557 printLazyBindTable(o); 1558 if (WeakBind) 1559 printWeakBindTable(o); 1560 if (RawClangAST) 1561 printRawClangAST(o); 1562 if (PrintFaultMaps) 1563 printFaultMaps(o); 1564} 1565 1566/// @brief Dump each object file in \a a; 1567static void DumpArchive(const Archive *a) { 1568 for (auto &ErrorOrChild : a->children()) { 1569 if (std::error_code EC = ErrorOrChild.getError()) 1570 report_error(a->getFileName(), EC); 1571 const Archive::Child &C = *ErrorOrChild; 1572 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1573 if (std::error_code EC = ChildOrErr.getError()) 1574 if (EC != object_error::invalid_file_type) 1575 report_error(a->getFileName(), EC); 1576 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1577 DumpObject(o); 1578 else 1579 report_error(a->getFileName(), object_error::invalid_file_type); 1580 } 1581} 1582 1583/// @brief Open file and figure out how to dump it. 1584static void DumpInput(StringRef file) { 1585 1586 // If we are using the Mach-O specific object file parser, then let it parse 1587 // the file and process the command line options. So the -arch flags can 1588 // be used to select specific slices, etc. 1589 if (MachOOpt) { 1590 ParseInputMachO(file); 1591 return; 1592 } 1593 1594 // Attempt to open the binary. 1595 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 1596 if (std::error_code EC = BinaryOrErr.getError()) 1597 report_error(file, EC); 1598 Binary &Binary = *BinaryOrErr.get().getBinary(); 1599 1600 if (Archive *a = dyn_cast<Archive>(&Binary)) 1601 DumpArchive(a); 1602 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 1603 DumpObject(o); 1604 else 1605 report_error(file, object_error::invalid_file_type); 1606} 1607 1608int main(int argc, char **argv) { 1609 // Print a stack trace if we signal out. 1610 sys::PrintStackTraceOnErrorSignal(); 1611 PrettyStackTraceProgram X(argc, argv); 1612 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 1613 1614 // Initialize targets and assembly printers/parsers. 1615 llvm::InitializeAllTargetInfos(); 1616 llvm::InitializeAllTargetMCs(); 1617 llvm::InitializeAllDisassemblers(); 1618 1619 // Register the target printer for --version. 1620 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 1621 1622 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 1623 TripleName = Triple::normalize(TripleName); 1624 1625 ToolName = argv[0]; 1626 1627 // Defaults to a.out if no filenames specified. 1628 if (InputFilenames.size() == 0) 1629 InputFilenames.push_back("a.out"); 1630 1631 if (DisassembleAll) 1632 Disassemble = true; 1633 if (!Disassemble 1634 && !Relocations 1635 && !SectionHeaders 1636 && !SectionContents 1637 && !SymbolTable 1638 && !UnwindInfo 1639 && !PrivateHeaders 1640 && !FirstPrivateHeader 1641 && !ExportsTrie 1642 && !Rebase 1643 && !Bind 1644 && !LazyBind 1645 && !WeakBind 1646 && !RawClangAST 1647 && !(UniversalHeaders && MachOOpt) 1648 && !(ArchiveHeaders && MachOOpt) 1649 && !(IndirectSymbols && MachOOpt) 1650 && !(DataInCode && MachOOpt) 1651 && !(LinkOptHints && MachOOpt) 1652 && !(InfoPlist && MachOOpt) 1653 && !(DylibsUsed && MachOOpt) 1654 && !(DylibId && MachOOpt) 1655 && !(ObjcMetaData && MachOOpt) 1656 && !(FilterSections.size() != 0 && MachOOpt) 1657 && !PrintFaultMaps) { 1658 cl::PrintHelpMessage(); 1659 return 2; 1660 } 1661 1662 std::for_each(InputFilenames.begin(), InputFilenames.end(), 1663 DumpInput); 1664 1665 return EXIT_SUCCESS; 1666} 1667