1 //===-- MachODump.cpp - Object file dumping utility for llvm --------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the MachO-specific dumper for llvm-objdump.
12 //===----------------------------------------------------------------------===//
14 #include "llvm-objdump.h"
15 #include "llvm-c/Disassembler.h"
16 #include "llvm/ADT/STLExtras.h"
17 #include "llvm/ADT/StringExtras.h"
18 #include "llvm/ADT/Triple.h"
19 #include "llvm/Config/config.h"
20 #include "llvm/DebugInfo/DIContext.h"
21 #include "llvm/MC/MCAsmInfo.h"
22 #include "llvm/MC/MCContext.h"
23 #include "llvm/MC/MCDisassembler.h"
24 #include "llvm/MC/MCInst.h"
25 #include "llvm/MC/MCInstPrinter.h"
26 #include "llvm/MC/MCInstrDesc.h"
27 #include "llvm/MC/MCInstrInfo.h"
28 #include "llvm/MC/MCRegisterInfo.h"
29 #include "llvm/MC/MCSubtargetInfo.h"
30 #include "llvm/Object/MachO.h"
31 #include "llvm/Object/MachOUniversal.h"
32 #include "llvm/Support/Casting.h"
33 #include "llvm/Support/CommandLine.h"
34 #include "llvm/Support/Debug.h"
35 #include "llvm/Support/Endian.h"
36 #include "llvm/Support/Format.h"
37 #include "llvm/Support/FormattedStream.h"
38 #include "llvm/Support/GraphWriter.h"
39 #include "llvm/Support/MachO.h"
40 #include "llvm/Support/MemoryBuffer.h"
41 #include "llvm/Support/TargetRegistry.h"
42 #include "llvm/Support/TargetSelect.h"
43 #include "llvm/Support/raw_ostream.h"
46 #include <system_error>
53 using namespace object;
57 cl::desc("Print line information from debug info if available"));
59 static cl::opt<std::string> DSYMFile("dsym",
60 cl::desc("Use .dSYM file for debug info"));
62 static cl::opt<bool> FullLeadingAddr("full-leading-addr",
63 cl::desc("Print full leading address"));
66 PrintImmHex("print-imm-hex",
67 cl::desc("Use hex format for immediate values"));
69 cl::opt<bool> llvm::UniversalHeaders("universal-headers",
70 cl::desc("Print Mach-O universal headers "
71 "(requires -macho)"));
74 llvm::ArchiveHeaders("archive-headers",
75 cl::desc("Print archive headers for Mach-O archives "
76 "(requires -macho)"));
78 static cl::list<std::string>
79 ArchFlags("arch", cl::desc("architecture(s) from a Mach-O file to dump"),
83 static std::string ThumbTripleName;
85 static const Target *GetTarget(const MachOObjectFile *MachOObj,
86 const char **McpuDefault,
87 const Target **ThumbTarget) {
88 // Figure out the target triple.
89 if (TripleName.empty()) {
90 llvm::Triple TT("unknown-unknown-unknown");
91 llvm::Triple ThumbTriple = Triple();
92 TT = MachOObj->getArch(McpuDefault, &ThumbTriple);
93 TripleName = TT.str();
94 ThumbTripleName = ThumbTriple.str();
97 // Get the target specific parser.
99 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
100 if (TheTarget && ThumbTripleName.empty())
103 *ThumbTarget = TargetRegistry::lookupTarget(ThumbTripleName, Error);
107 errs() << "llvm-objdump: error: unable to get target for '";
109 errs() << TripleName;
111 errs() << ThumbTripleName;
112 errs() << "', see --version and --triple.\n";
116 struct SymbolSorter {
117 bool operator()(const SymbolRef &A, const SymbolRef &B) {
118 SymbolRef::Type AType, BType;
122 uint64_t AAddr, BAddr;
123 if (AType != SymbolRef::ST_Function)
127 if (BType != SymbolRef::ST_Function)
131 return AAddr < BAddr;
135 // Types for the storted data in code table that is built before disassembly
136 // and the predicate function to sort them.
137 typedef std::pair<uint64_t, DiceRef> DiceTableEntry;
138 typedef std::vector<DiceTableEntry> DiceTable;
139 typedef DiceTable::iterator dice_table_iterator;
141 // This is used to search for a data in code table entry for the PC being
142 // disassembled. The j parameter has the PC in j.first. A single data in code
143 // table entry can cover many bytes for each of its Kind's. So if the offset,
144 // aka the i.first value, of the data in code table entry plus its Length
145 // covers the PC being searched for this will return true. If not it will
147 static bool compareDiceTableEntries(const DiceTableEntry &i,
148 const DiceTableEntry &j) {
150 i.second.getLength(Length);
152 return j.first >= i.first && j.first < i.first + Length;
155 static uint64_t DumpDataInCode(const char *bytes, uint64_t Length,
156 unsigned short Kind) {
157 uint32_t Value, Size = 1;
161 case MachO::DICE_KIND_DATA:
164 DumpBytes(StringRef(bytes, 4));
165 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
166 outs() << "\t.long " << Value;
168 } else if (Length >= 2) {
170 DumpBytes(StringRef(bytes, 2));
171 Value = bytes[1] << 8 | bytes[0];
172 outs() << "\t.short " << Value;
176 DumpBytes(StringRef(bytes, 2));
178 outs() << "\t.byte " << Value;
181 if (Kind == MachO::DICE_KIND_DATA)
182 outs() << "\t@ KIND_DATA\n";
184 outs() << "\t@ data in code kind = " << Kind << "\n";
186 case MachO::DICE_KIND_JUMP_TABLE8:
188 DumpBytes(StringRef(bytes, 1));
190 outs() << "\t.byte " << format("%3u", Value) << "\t@ KIND_JUMP_TABLE8\n";
193 case MachO::DICE_KIND_JUMP_TABLE16:
195 DumpBytes(StringRef(bytes, 2));
196 Value = bytes[1] << 8 | bytes[0];
197 outs() << "\t.short " << format("%5u", Value & 0xffff)
198 << "\t@ KIND_JUMP_TABLE16\n";
201 case MachO::DICE_KIND_JUMP_TABLE32:
202 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
204 DumpBytes(StringRef(bytes, 4));
205 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
206 outs() << "\t.long " << Value;
207 if (Kind == MachO::DICE_KIND_JUMP_TABLE32)
208 outs() << "\t@ KIND_JUMP_TABLE32\n";
210 outs() << "\t@ KIND_ABS_JUMP_TABLE32\n";
217 static void getSectionsAndSymbols(const MachO::mach_header Header,
218 MachOObjectFile *MachOObj,
219 std::vector<SectionRef> &Sections,
220 std::vector<SymbolRef> &Symbols,
221 SmallVectorImpl<uint64_t> &FoundFns,
222 uint64_t &BaseSegmentAddress) {
223 for (const SymbolRef &Symbol : MachOObj->symbols()) {
225 Symbol.getName(SymName);
226 if (!SymName.startswith("ltmp"))
227 Symbols.push_back(Symbol);
230 for (const SectionRef &Section : MachOObj->sections()) {
232 Section.getName(SectName);
233 Sections.push_back(Section);
236 MachOObjectFile::LoadCommandInfo Command =
237 MachOObj->getFirstLoadCommandInfo();
238 bool BaseSegmentAddressSet = false;
239 for (unsigned i = 0;; ++i) {
240 if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) {
241 // We found a function starts segment, parse the addresses for later
243 MachO::linkedit_data_command LLC =
244 MachOObj->getLinkeditDataLoadCommand(Command);
246 MachOObj->ReadULEB128s(LLC.dataoff, FoundFns);
247 } else if (Command.C.cmd == MachO::LC_SEGMENT) {
248 MachO::segment_command SLC = MachOObj->getSegmentLoadCommand(Command);
249 StringRef SegName = SLC.segname;
250 if (!BaseSegmentAddressSet && SegName != "__PAGEZERO") {
251 BaseSegmentAddressSet = true;
252 BaseSegmentAddress = SLC.vmaddr;
256 if (i == Header.ncmds - 1)
259 Command = MachOObj->getNextLoadCommandInfo(Command);
263 // checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file
264 // and if it is and there is a list of architecture flags is specified then
265 // check to make sure this Mach-O file is one of those architectures or all
266 // architectures were specified. If not then an error is generated and this
267 // routine returns false. Else it returns true.
268 static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) {
269 if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) {
270 MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O);
271 bool ArchFound = false;
272 MachO::mach_header H;
273 MachO::mach_header_64 H_64;
275 if (MachO->is64Bit()) {
276 H_64 = MachO->MachOObjectFile::getHeader64();
277 T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype);
279 H = MachO->MachOObjectFile::getHeader();
280 T = MachOObjectFile::getArch(H.cputype, H.cpusubtype);
283 for (i = 0; i < ArchFlags.size(); ++i) {
284 if (ArchFlags[i] == T.getArchName())
289 errs() << "llvm-objdump: file: " + Filename + " does not contain "
290 << "architecture: " + ArchFlags[i] + "\n";
297 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF);
299 // ProcessMachO() is passed a single opened Mach-O file, which may be an
300 // archive member and or in a slice of a universal file. It prints the
301 // the file name and header info and then processes it according to the
302 // command line options.
303 static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF,
304 StringRef ArchiveMemberName = StringRef(),
305 StringRef ArchitectureName = StringRef()) {
306 // If we are doing some processing here on the Mach-O file print the header
307 // info. And don't print it otherwise like in the case of printing the
309 if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind ||
310 LazyBind || WeakBind) {
312 if (!ArchiveMemberName.empty())
313 outs() << '(' << ArchiveMemberName << ')';
314 if (!ArchitectureName.empty())
315 outs() << " (architecture " << ArchitectureName << ")";
320 DisassembleMachO(Filename, MachOOF);
322 PrintRelocations(MachOOF);
324 PrintSectionHeaders(MachOOF);
326 PrintSectionContents(MachOOF);
328 PrintSymbolTable(MachOOF);
330 printMachOUnwindInfo(MachOOF);
332 printMachOFileHeader(MachOOF);
334 printExportsTrie(MachOOF);
336 printRebaseTable(MachOOF);
338 printBindTable(MachOOF);
340 printLazyBindTable(MachOOF);
342 printWeakBindTable(MachOOF);
345 // printUnknownCPUType() helps print_fat_headers for unknown CPU's.
346 static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) {
347 outs() << " cputype (" << cputype << ")\n";
348 outs() << " cpusubtype (" << cpusubtype << ")\n";
351 // printCPUType() helps print_fat_headers by printing the cputype and
352 // pusubtype (symbolically for the one's it knows about).
353 static void printCPUType(uint32_t cputype, uint32_t cpusubtype) {
355 case MachO::CPU_TYPE_I386:
356 switch (cpusubtype) {
357 case MachO::CPU_SUBTYPE_I386_ALL:
358 outs() << " cputype CPU_TYPE_I386\n";
359 outs() << " cpusubtype CPU_SUBTYPE_I386_ALL\n";
362 printUnknownCPUType(cputype, cpusubtype);
366 case MachO::CPU_TYPE_X86_64:
367 switch (cpusubtype) {
368 case MachO::CPU_SUBTYPE_X86_64_ALL:
369 outs() << " cputype CPU_TYPE_X86_64\n";
370 outs() << " cpusubtype CPU_SUBTYPE_X86_64_ALL\n";
372 case MachO::CPU_SUBTYPE_X86_64_H:
373 outs() << " cputype CPU_TYPE_X86_64\n";
374 outs() << " cpusubtype CPU_SUBTYPE_X86_64_H\n";
377 printUnknownCPUType(cputype, cpusubtype);
381 case MachO::CPU_TYPE_ARM:
382 switch (cpusubtype) {
383 case MachO::CPU_SUBTYPE_ARM_ALL:
384 outs() << " cputype CPU_TYPE_ARM\n";
385 outs() << " cpusubtype CPU_SUBTYPE_ARM_ALL\n";
387 case MachO::CPU_SUBTYPE_ARM_V4T:
388 outs() << " cputype CPU_TYPE_ARM\n";
389 outs() << " cpusubtype CPU_SUBTYPE_ARM_V4T\n";
391 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
392 outs() << " cputype CPU_TYPE_ARM\n";
393 outs() << " cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n";
395 case MachO::CPU_SUBTYPE_ARM_XSCALE:
396 outs() << " cputype CPU_TYPE_ARM\n";
397 outs() << " cpusubtype CPU_SUBTYPE_ARM_XSCALE\n";
399 case MachO::CPU_SUBTYPE_ARM_V6:
400 outs() << " cputype CPU_TYPE_ARM\n";
401 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6\n";
403 case MachO::CPU_SUBTYPE_ARM_V6M:
404 outs() << " cputype CPU_TYPE_ARM\n";
405 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6M\n";
407 case MachO::CPU_SUBTYPE_ARM_V7:
408 outs() << " cputype CPU_TYPE_ARM\n";
409 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7\n";
411 case MachO::CPU_SUBTYPE_ARM_V7EM:
412 outs() << " cputype CPU_TYPE_ARM\n";
413 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7EM\n";
415 case MachO::CPU_SUBTYPE_ARM_V7K:
416 outs() << " cputype CPU_TYPE_ARM\n";
417 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7K\n";
419 case MachO::CPU_SUBTYPE_ARM_V7M:
420 outs() << " cputype CPU_TYPE_ARM\n";
421 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7M\n";
423 case MachO::CPU_SUBTYPE_ARM_V7S:
424 outs() << " cputype CPU_TYPE_ARM\n";
425 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7S\n";
428 printUnknownCPUType(cputype, cpusubtype);
432 case MachO::CPU_TYPE_ARM64:
433 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
434 case MachO::CPU_SUBTYPE_ARM64_ALL:
435 outs() << " cputype CPU_TYPE_ARM64\n";
436 outs() << " cpusubtype CPU_SUBTYPE_ARM64_ALL\n";
439 printUnknownCPUType(cputype, cpusubtype);
444 printUnknownCPUType(cputype, cpusubtype);
449 static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB,
451 outs() << "Fat headers\n";
453 outs() << "fat_magic FAT_MAGIC\n";
455 outs() << "fat_magic " << format("0x%" PRIx32, MachO::FAT_MAGIC) << "\n";
457 uint32_t nfat_arch = UB->getNumberOfObjects();
458 StringRef Buf = UB->getData();
459 uint64_t size = Buf.size();
460 uint64_t big_size = sizeof(struct MachO::fat_header) +
461 nfat_arch * sizeof(struct MachO::fat_arch);
462 outs() << "nfat_arch " << UB->getNumberOfObjects();
464 outs() << " (malformed, contains zero architecture types)\n";
465 else if (big_size > size)
466 outs() << " (malformed, architectures past end of file)\n";
470 for (uint32_t i = 0; i < nfat_arch; ++i) {
471 MachOUniversalBinary::ObjectForArch OFA(UB, i);
472 uint32_t cputype = OFA.getCPUType();
473 uint32_t cpusubtype = OFA.getCPUSubType();
474 outs() << "architecture ";
475 for (uint32_t j = 0; i != 0 && j <= i - 1; j++) {
476 MachOUniversalBinary::ObjectForArch other_OFA(UB, j);
477 uint32_t other_cputype = other_OFA.getCPUType();
478 uint32_t other_cpusubtype = other_OFA.getCPUSubType();
479 if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype &&
480 (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) ==
481 (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) {
482 outs() << "(illegal duplicate architecture) ";
487 outs() << OFA.getArchTypeName() << "\n";
488 printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
491 outs() << " cputype " << cputype << "\n";
492 outs() << " cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK)
496 (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64)
497 outs() << " capabilities CPU_SUBTYPE_LIB64\n";
499 outs() << " capabilities "
500 << format("0x%" PRIx32,
501 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n";
502 outs() << " offset " << OFA.getOffset();
503 if (OFA.getOffset() > size)
504 outs() << " (past end of file)";
505 if (OFA.getOffset() % (1 << OFA.getAlign()) != 0)
506 outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")";
508 outs() << " size " << OFA.getSize();
509 big_size = OFA.getOffset() + OFA.getSize();
511 outs() << " (past end of file)";
513 outs() << " align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign())
518 static void printArchiveChild(Archive::Child &C, bool verbose,
521 outs() << C.getChildOffset() << "\t";
522 sys::fs::perms Mode = C.getAccessMode();
524 // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG.
525 // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG.
527 if (Mode & sys::fs::owner_read)
531 if (Mode & sys::fs::owner_write)
535 if (Mode & sys::fs::owner_exe)
539 if (Mode & sys::fs::group_read)
543 if (Mode & sys::fs::group_write)
547 if (Mode & sys::fs::group_exe)
551 if (Mode & sys::fs::others_read)
555 if (Mode & sys::fs::others_write)
559 if (Mode & sys::fs::others_exe)
564 outs() << format("0%o ", Mode);
567 unsigned UID = C.getUID();
568 outs() << format("%3d/", UID);
569 unsigned GID = C.getGID();
570 outs() << format("%-3d ", GID);
571 uint64_t Size = C.getRawSize();
572 outs() << format("%5d ", Size);
574 StringRef RawLastModified = C.getRawLastModified();
577 if (RawLastModified.getAsInteger(10, Seconds))
578 outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified;
580 // Since cime(3) returns a 26 character string of the form:
581 // "Sun Sep 16 01:03:52 1973\n\0"
582 // just print 24 characters.
584 outs() << format("%.24s ", ctime(&t));
587 outs() << RawLastModified << " ";
591 ErrorOr<StringRef> NameOrErr = C.getName();
592 if (NameOrErr.getError()) {
593 StringRef RawName = C.getRawName();
594 outs() << RawName << "\n";
596 StringRef Name = NameOrErr.get();
597 outs() << Name << "\n";
600 StringRef RawName = C.getRawName();
601 outs() << RawName << "\n";
605 static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) {
606 if (A->hasSymbolTable()) {
607 Archive::child_iterator S = A->getSymbolTableChild();
608 Archive::Child C = *S;
609 printArchiveChild(C, verbose, print_offset);
611 for (Archive::child_iterator I = A->child_begin(), E = A->child_end(); I != E;
613 Archive::Child C = *I;
614 printArchiveChild(C, verbose, print_offset);
618 // ParseInputMachO() parses the named Mach-O file in Filename and handles the
619 // -arch flags selecting just those slices as specified by them and also parses
620 // archive files. Then for each individual Mach-O file ProcessMachO() is
621 // called to process the file based on the command line options.
622 void llvm::ParseInputMachO(StringRef Filename) {
623 // Check for -arch all and verifiy the -arch flags are valid.
624 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
625 if (ArchFlags[i] == "all") {
628 if (!MachOObjectFile::isValidArch(ArchFlags[i])) {
629 errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] +
630 "'for the -arch option\n";
636 // Attempt to open the binary.
637 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename);
638 if (std::error_code EC = BinaryOrErr.getError()) {
639 errs() << "llvm-objdump: '" << Filename << "': " << EC.message() << ".\n";
642 Binary &Bin = *BinaryOrErr.get().getBinary();
644 if (Archive *A = dyn_cast<Archive>(&Bin)) {
645 outs() << "Archive : " << Filename << "\n";
647 printArchiveHeaders(A, true, false);
648 for (Archive::child_iterator I = A->child_begin(), E = A->child_end();
650 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = I->getAsBinary();
651 if (ChildOrErr.getError())
653 if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
654 if (!checkMachOAndArchFlags(O, Filename))
656 ProcessMachO(Filename, O, O->getFileName());
661 if (UniversalHeaders) {
662 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin))
663 printMachOUniversalHeaders(UB, true);
665 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) {
666 // If we have a list of architecture flags specified dump only those.
667 if (!ArchAll && ArchFlags.size() != 0) {
668 // Look for a slice in the universal binary that matches each ArchFlag.
670 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
672 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
673 E = UB->end_objects();
675 if (ArchFlags[i] == I->getArchTypeName()) {
677 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
678 I->getAsObjectFile();
679 std::string ArchitectureName = "";
680 if (ArchFlags.size() > 1)
681 ArchitectureName = I->getArchTypeName();
683 ObjectFile &O = *ObjOrErr.get();
684 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
685 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
686 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
688 std::unique_ptr<Archive> &A = *AOrErr;
689 outs() << "Archive : " << Filename;
690 if (!ArchitectureName.empty())
691 outs() << " (architecture " << ArchitectureName << ")";
694 printArchiveHeaders(A.get(), true, false);
695 for (Archive::child_iterator AI = A->child_begin(),
698 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
699 if (ChildOrErr.getError())
701 if (MachOObjectFile *O =
702 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
703 ProcessMachO(Filename, O, O->getFileName(), ArchitectureName);
709 errs() << "llvm-objdump: file: " + Filename + " does not contain "
710 << "architecture: " + ArchFlags[i] + "\n";
716 // No architecture flags were specified so if this contains a slice that
717 // matches the host architecture dump only that.
719 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
720 E = UB->end_objects();
722 if (MachOObjectFile::getHostArch().getArchName() ==
723 I->getArchTypeName()) {
724 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
725 std::string ArchiveName;
728 ObjectFile &O = *ObjOrErr.get();
729 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
730 ProcessMachO(Filename, MachOOF);
731 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
733 std::unique_ptr<Archive> &A = *AOrErr;
734 outs() << "Archive : " << Filename << "\n";
736 printArchiveHeaders(A.get(), true, false);
737 for (Archive::child_iterator AI = A->child_begin(),
740 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
741 if (ChildOrErr.getError())
743 if (MachOObjectFile *O =
744 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
745 ProcessMachO(Filename, O, O->getFileName());
752 // Either all architectures have been specified or none have been specified
753 // and this does not contain the host architecture so dump all the slices.
754 bool moreThanOneArch = UB->getNumberOfObjects() > 1;
755 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
756 E = UB->end_objects();
758 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
759 std::string ArchitectureName = "";
761 ArchitectureName = I->getArchTypeName();
763 ObjectFile &Obj = *ObjOrErr.get();
764 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj))
765 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
766 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) {
767 std::unique_ptr<Archive> &A = *AOrErr;
768 outs() << "Archive : " << Filename;
769 if (!ArchitectureName.empty())
770 outs() << " (architecture " << ArchitectureName << ")";
773 printArchiveHeaders(A.get(), true, false);
774 for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end();
776 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
777 if (ChildOrErr.getError())
779 if (MachOObjectFile *O =
780 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
781 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O))
782 ProcessMachO(Filename, MachOOF, MachOOF->getFileName(),
790 if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) {
791 if (!checkMachOAndArchFlags(O, Filename))
793 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) {
794 ProcessMachO(Filename, MachOOF);
796 errs() << "llvm-objdump: '" << Filename << "': "
797 << "Object is not a Mach-O file type.\n";
799 errs() << "llvm-objdump: '" << Filename << "': "
800 << "Unrecognized file type.\n";
803 typedef DenseMap<uint64_t, StringRef> SymbolAddressMap;
804 typedef std::pair<uint64_t, const char *> BindInfoEntry;
805 typedef std::vector<BindInfoEntry> BindTable;
806 typedef BindTable::iterator bind_table_iterator;
808 // The block of info used by the Symbolizer call backs.
809 struct DisassembleInfo {
813 SymbolAddressMap *AddrMap;
814 std::vector<SectionRef> *Sections;
815 const char *class_name;
816 const char *selector_name;
818 char *demangled_name;
821 BindTable *bindtable;
824 // GuessSymbolName is passed the address of what might be a symbol and a
825 // pointer to the DisassembleInfo struct. It returns the name of a symbol
826 // with that address or nullptr if no symbol is found with that address.
827 static const char *GuessSymbolName(uint64_t value,
828 struct DisassembleInfo *info) {
829 const char *SymbolName = nullptr;
830 // A DenseMap can't lookup up some values.
831 if (value != 0xffffffffffffffffULL && value != 0xfffffffffffffffeULL) {
832 StringRef name = info->AddrMap->lookup(value);
834 SymbolName = name.data();
839 // SymbolizerGetOpInfo() is the operand information call back function.
840 // This is called to get the symbolic information for operand(s) of an
841 // instruction when it is being done. This routine does this from
842 // the relocation information, symbol table, etc. That block of information
843 // is a pointer to the struct DisassembleInfo that was passed when the
844 // disassembler context was created and passed to back to here when
845 // called back by the disassembler for instruction operands that could have
846 // relocation information. The address of the instruction containing operand is
847 // at the Pc parameter. The immediate value the operand has is passed in
848 // op_info->Value and is at Offset past the start of the instruction and has a
849 // byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
850 // LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
851 // names and addends of the symbolic expression to add for the operand. The
852 // value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
853 // information is returned then this function returns 1 else it returns 0.
854 int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
855 uint64_t Size, int TagType, void *TagBuf) {
856 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
857 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
858 uint64_t value = op_info->Value;
860 // Make sure all fields returned are zero if we don't set them.
861 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
862 op_info->Value = value;
864 // If the TagType is not the value 1 which it code knows about or if no
865 // verbose symbolic information is wanted then just return 0, indicating no
866 // information is being returned.
867 if (TagType != 1 || info->verbose == false)
870 unsigned int Arch = info->O->getArch();
871 if (Arch == Triple::x86) {
872 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
874 // First search the section's relocation entries (if any) for an entry
875 // for this section offset.
876 uint32_t sect_addr = info->S.getAddress();
877 uint32_t sect_offset = (Pc + Offset) - sect_addr;
878 bool reloc_found = false;
880 MachO::any_relocation_info RE;
881 bool isExtern = false;
883 bool r_scattered = false;
884 uint32_t r_value, pair_r_value, r_type;
885 for (const RelocationRef &Reloc : info->S.relocations()) {
886 uint64_t RelocOffset;
887 Reloc.getOffset(RelocOffset);
888 if (RelocOffset == sect_offset) {
889 Rel = Reloc.getRawDataRefImpl();
890 RE = info->O->getRelocation(Rel);
891 r_type = info->O->getAnyRelocationType(RE);
892 r_scattered = info->O->isRelocationScattered(RE);
894 r_value = info->O->getScatteredRelocationValue(RE);
895 if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
896 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
897 DataRefImpl RelNext = Rel;
898 info->O->moveRelocationNext(RelNext);
899 MachO::any_relocation_info RENext;
900 RENext = info->O->getRelocation(RelNext);
901 if (info->O->isRelocationScattered(RENext))
902 pair_r_value = info->O->getScatteredRelocationValue(RENext);
907 isExtern = info->O->getPlainRelocationExternal(RE);
909 symbol_iterator RelocSym = Reloc.getSymbol();
917 if (reloc_found && isExtern) {
919 Symbol.getName(SymName);
920 const char *name = SymName.data();
921 op_info->AddSymbol.Present = 1;
922 op_info->AddSymbol.Name = name;
923 // For i386 extern relocation entries the value in the instruction is
924 // the offset from the symbol, and value is already set in op_info->Value.
927 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
928 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
929 const char *add = GuessSymbolName(r_value, info);
930 const char *sub = GuessSymbolName(pair_r_value, info);
931 uint32_t offset = value - (r_value - pair_r_value);
932 op_info->AddSymbol.Present = 1;
934 op_info->AddSymbol.Name = add;
936 op_info->AddSymbol.Value = r_value;
937 op_info->SubtractSymbol.Present = 1;
939 op_info->SubtractSymbol.Name = sub;
941 op_info->SubtractSymbol.Value = pair_r_value;
942 op_info->Value = offset;
946 // Second search the external relocation entries of a fully linked image
947 // (if any) for an entry that matches this segment offset.
948 // uint32_t seg_offset = (Pc + Offset);
950 } else if (Arch == Triple::x86_64) {
951 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
953 // First search the section's relocation entries (if any) for an entry
954 // for this section offset.
955 uint64_t sect_addr = info->S.getAddress();
956 uint64_t sect_offset = (Pc + Offset) - sect_addr;
957 bool reloc_found = false;
959 MachO::any_relocation_info RE;
960 bool isExtern = false;
962 for (const RelocationRef &Reloc : info->S.relocations()) {
963 uint64_t RelocOffset;
964 Reloc.getOffset(RelocOffset);
965 if (RelocOffset == sect_offset) {
966 Rel = Reloc.getRawDataRefImpl();
967 RE = info->O->getRelocation(Rel);
968 // NOTE: Scattered relocations don't exist on x86_64.
969 isExtern = info->O->getPlainRelocationExternal(RE);
971 symbol_iterator RelocSym = Reloc.getSymbol();
978 if (reloc_found && isExtern) {
979 // The Value passed in will be adjusted by the Pc if the instruction
980 // adds the Pc. But for x86_64 external relocation entries the Value
981 // is the offset from the external symbol.
982 if (info->O->getAnyRelocationPCRel(RE))
983 op_info->Value -= Pc + Offset + Size;
985 Symbol.getName(SymName);
986 const char *name = SymName.data();
987 unsigned Type = info->O->getAnyRelocationType(RE);
988 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
989 DataRefImpl RelNext = Rel;
990 info->O->moveRelocationNext(RelNext);
991 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
992 unsigned TypeNext = info->O->getAnyRelocationType(RENext);
993 bool isExternNext = info->O->getPlainRelocationExternal(RENext);
994 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
995 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
996 op_info->SubtractSymbol.Present = 1;
997 op_info->SubtractSymbol.Name = name;
998 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
999 Symbol = *RelocSymNext;
1000 StringRef SymNameNext;
1001 Symbol.getName(SymNameNext);
1002 name = SymNameNext.data();
1005 // TODO: add the VariantKinds to op_info->VariantKind for relocation types
1006 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
1007 op_info->AddSymbol.Present = 1;
1008 op_info->AddSymbol.Name = name;
1012 // Second search the external relocation entries of a fully linked image
1013 // (if any) for an entry that matches this segment offset.
1014 // uint64_t seg_offset = (Pc + Offset);
1016 } else if (Arch == Triple::arm) {
1017 if (Offset != 0 || (Size != 4 && Size != 2))
1019 // First search the section's relocation entries (if any) for an entry
1020 // for this section offset.
1021 uint32_t sect_addr = info->S.getAddress();
1022 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1023 bool reloc_found = false;
1025 MachO::any_relocation_info RE;
1026 bool isExtern = false;
1028 bool r_scattered = false;
1029 uint32_t r_value, pair_r_value, r_type, r_length, other_half;
1030 for (const RelocationRef &Reloc : info->S.relocations()) {
1031 uint64_t RelocOffset;
1032 Reloc.getOffset(RelocOffset);
1033 if (RelocOffset == sect_offset) {
1034 Rel = Reloc.getRawDataRefImpl();
1035 RE = info->O->getRelocation(Rel);
1036 r_length = info->O->getAnyRelocationLength(RE);
1037 r_scattered = info->O->isRelocationScattered(RE);
1039 r_value = info->O->getScatteredRelocationValue(RE);
1040 r_type = info->O->getScatteredRelocationType(RE);
1042 r_type = info->O->getAnyRelocationType(RE);
1043 isExtern = info->O->getPlainRelocationExternal(RE);
1045 symbol_iterator RelocSym = Reloc.getSymbol();
1049 if (r_type == MachO::ARM_RELOC_HALF ||
1050 r_type == MachO::ARM_RELOC_SECTDIFF ||
1051 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
1052 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1053 DataRefImpl RelNext = Rel;
1054 info->O->moveRelocationNext(RelNext);
1055 MachO::any_relocation_info RENext;
1056 RENext = info->O->getRelocation(RelNext);
1057 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
1058 if (info->O->isRelocationScattered(RENext))
1059 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1065 if (reloc_found && isExtern) {
1067 Symbol.getName(SymName);
1068 const char *name = SymName.data();
1069 op_info->AddSymbol.Present = 1;
1070 op_info->AddSymbol.Name = name;
1073 case MachO::ARM_RELOC_HALF:
1074 if ((r_length & 0x1) == 1) {
1075 op_info->Value = value << 16 | other_half;
1076 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1078 op_info->Value = other_half << 16 | value;
1079 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1087 case MachO::ARM_RELOC_HALF:
1088 if ((r_length & 0x1) == 1) {
1089 op_info->Value = value << 16 | other_half;
1090 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1092 op_info->Value = other_half << 16 | value;
1093 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1102 // If we have a branch that is not an external relocation entry then
1103 // return 0 so the code in tryAddingSymbolicOperand() can use the
1104 // SymbolLookUp call back with the branch target address to look up the
1105 // symbol and possiblity add an annotation for a symbol stub.
1106 if (reloc_found && isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
1107 r_type == MachO::ARM_THUMB_RELOC_BR22))
1110 uint32_t offset = 0;
1112 if (r_type == MachO::ARM_RELOC_HALF ||
1113 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1114 if ((r_length & 0x1) == 1)
1115 value = value << 16 | other_half;
1117 value = other_half << 16 | value;
1119 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
1120 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
1121 offset = value - r_value;
1126 if (reloc_found && r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1127 if ((r_length & 0x1) == 1)
1128 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1130 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1131 const char *add = GuessSymbolName(r_value, info);
1132 const char *sub = GuessSymbolName(pair_r_value, info);
1133 int32_t offset = value - (r_value - pair_r_value);
1134 op_info->AddSymbol.Present = 1;
1136 op_info->AddSymbol.Name = add;
1138 op_info->AddSymbol.Value = r_value;
1139 op_info->SubtractSymbol.Present = 1;
1141 op_info->SubtractSymbol.Name = sub;
1143 op_info->SubtractSymbol.Value = pair_r_value;
1144 op_info->Value = offset;
1148 if (reloc_found == false)
1151 op_info->AddSymbol.Present = 1;
1152 op_info->Value = offset;
1154 if (r_type == MachO::ARM_RELOC_HALF) {
1155 if ((r_length & 0x1) == 1)
1156 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1158 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1161 const char *add = GuessSymbolName(value, info);
1162 if (add != nullptr) {
1163 op_info->AddSymbol.Name = add;
1166 op_info->AddSymbol.Value = value;
1168 } else if (Arch == Triple::aarch64) {
1169 if (Offset != 0 || Size != 4)
1171 // First search the section's relocation entries (if any) for an entry
1172 // for this section offset.
1173 uint64_t sect_addr = info->S.getAddress();
1174 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1175 bool reloc_found = false;
1177 MachO::any_relocation_info RE;
1178 bool isExtern = false;
1180 uint32_t r_type = 0;
1181 for (const RelocationRef &Reloc : info->S.relocations()) {
1182 uint64_t RelocOffset;
1183 Reloc.getOffset(RelocOffset);
1184 if (RelocOffset == sect_offset) {
1185 Rel = Reloc.getRawDataRefImpl();
1186 RE = info->O->getRelocation(Rel);
1187 r_type = info->O->getAnyRelocationType(RE);
1188 if (r_type == MachO::ARM64_RELOC_ADDEND) {
1189 DataRefImpl RelNext = Rel;
1190 info->O->moveRelocationNext(RelNext);
1191 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1193 value = info->O->getPlainRelocationSymbolNum(RENext);
1194 op_info->Value = value;
1197 // NOTE: Scattered relocations don't exist on arm64.
1198 isExtern = info->O->getPlainRelocationExternal(RE);
1200 symbol_iterator RelocSym = Reloc.getSymbol();
1207 if (reloc_found && isExtern) {
1209 Symbol.getName(SymName);
1210 const char *name = SymName.data();
1211 op_info->AddSymbol.Present = 1;
1212 op_info->AddSymbol.Name = name;
1215 case MachO::ARM64_RELOC_PAGE21:
1217 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE;
1219 case MachO::ARM64_RELOC_PAGEOFF12:
1221 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF;
1223 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
1225 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE;
1227 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
1229 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF;
1231 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
1232 /* @tvlppage is not implemented in llvm-mc */
1233 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP;
1235 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
1236 /* @tvlppageoff is not implemented in llvm-mc */
1237 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF;
1240 case MachO::ARM64_RELOC_BRANCH26:
1241 op_info->VariantKind = LLVMDisassembler_VariantKind_None;
1252 // GuessCstringPointer is passed the address of what might be a pointer to a
1253 // literal string in a cstring section. If that address is in a cstring section
1254 // it returns a pointer to that string. Else it returns nullptr.
1255 const char *GuessCstringPointer(uint64_t ReferenceValue,
1256 struct DisassembleInfo *info) {
1257 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1258 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1259 for (unsigned I = 0;; ++I) {
1260 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1261 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1262 for (unsigned J = 0; J < Seg.nsects; ++J) {
1263 MachO::section_64 Sec = info->O->getSection64(Load, J);
1264 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1265 if (section_type == MachO::S_CSTRING_LITERALS &&
1266 ReferenceValue >= Sec.addr &&
1267 ReferenceValue < Sec.addr + Sec.size) {
1268 uint64_t sect_offset = ReferenceValue - Sec.addr;
1269 uint64_t object_offset = Sec.offset + sect_offset;
1270 StringRef MachOContents = info->O->getData();
1271 uint64_t object_size = MachOContents.size();
1272 const char *object_addr = (const char *)MachOContents.data();
1273 if (object_offset < object_size) {
1274 const char *name = object_addr + object_offset;
1281 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1282 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1283 for (unsigned J = 0; J < Seg.nsects; ++J) {
1284 MachO::section Sec = info->O->getSection(Load, J);
1285 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1286 if (section_type == MachO::S_CSTRING_LITERALS &&
1287 ReferenceValue >= Sec.addr &&
1288 ReferenceValue < Sec.addr + Sec.size) {
1289 uint64_t sect_offset = ReferenceValue - Sec.addr;
1290 uint64_t object_offset = Sec.offset + sect_offset;
1291 StringRef MachOContents = info->O->getData();
1292 uint64_t object_size = MachOContents.size();
1293 const char *object_addr = (const char *)MachOContents.data();
1294 if (object_offset < object_size) {
1295 const char *name = object_addr + object_offset;
1303 if (I == LoadCommandCount - 1)
1306 Load = info->O->getNextLoadCommandInfo(Load);
1311 // GuessIndirectSymbol returns the name of the indirect symbol for the
1312 // ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe
1313 // an address of a symbol stub or a lazy or non-lazy pointer to associate the
1314 // symbol name being referenced by the stub or pointer.
1315 static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
1316 struct DisassembleInfo *info) {
1317 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1318 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1319 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
1320 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
1321 for (unsigned I = 0;; ++I) {
1322 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1323 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1324 for (unsigned J = 0; J < Seg.nsects; ++J) {
1325 MachO::section_64 Sec = info->O->getSection64(Load, J);
1326 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1327 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1328 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1329 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1330 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1331 section_type == MachO::S_SYMBOL_STUBS) &&
1332 ReferenceValue >= Sec.addr &&
1333 ReferenceValue < Sec.addr + Sec.size) {
1335 if (section_type == MachO::S_SYMBOL_STUBS)
1336 stride = Sec.reserved2;
1341 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1342 if (index < Dysymtab.nindirectsyms) {
1343 uint32_t indirect_symbol =
1344 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1345 if (indirect_symbol < Symtab.nsyms) {
1346 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1347 SymbolRef Symbol = *Sym;
1349 Symbol.getName(SymName);
1350 const char *name = SymName.data();
1356 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1357 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1358 for (unsigned J = 0; J < Seg.nsects; ++J) {
1359 MachO::section Sec = info->O->getSection(Load, J);
1360 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1361 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1362 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1363 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1364 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1365 section_type == MachO::S_SYMBOL_STUBS) &&
1366 ReferenceValue >= Sec.addr &&
1367 ReferenceValue < Sec.addr + Sec.size) {
1369 if (section_type == MachO::S_SYMBOL_STUBS)
1370 stride = Sec.reserved2;
1375 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1376 if (index < Dysymtab.nindirectsyms) {
1377 uint32_t indirect_symbol =
1378 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
1379 if (indirect_symbol < Symtab.nsyms) {
1380 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1381 SymbolRef Symbol = *Sym;
1383 Symbol.getName(SymName);
1384 const char *name = SymName.data();
1391 if (I == LoadCommandCount - 1)
1394 Load = info->O->getNextLoadCommandInfo(Load);
1399 // method_reference() is called passing it the ReferenceName that might be
1400 // a reference it to an Objective-C method call. If so then it allocates and
1401 // assembles a method call string with the values last seen and saved in
1402 // the DisassembleInfo's class_name and selector_name fields. This is saved
1403 // into the method field of the info and any previous string is free'ed.
1404 // Then the class_name field in the info is set to nullptr. The method call
1405 // string is set into ReferenceName and ReferenceType is set to
1406 // LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call
1407 // then both ReferenceType and ReferenceName are left unchanged.
1408 static void method_reference(struct DisassembleInfo *info,
1409 uint64_t *ReferenceType,
1410 const char **ReferenceName) {
1411 unsigned int Arch = info->O->getArch();
1412 if (*ReferenceName != nullptr) {
1413 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
1414 if (info->selector_name != nullptr) {
1415 if (info->method != nullptr)
1417 if (info->class_name != nullptr) {
1418 info->method = (char *)malloc(5 + strlen(info->class_name) +
1419 strlen(info->selector_name));
1420 if (info->method != nullptr) {
1421 strcpy(info->method, "+[");
1422 strcat(info->method, info->class_name);
1423 strcat(info->method, " ");
1424 strcat(info->method, info->selector_name);
1425 strcat(info->method, "]");
1426 *ReferenceName = info->method;
1427 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1430 info->method = (char *)malloc(9 + strlen(info->selector_name));
1431 if (info->method != nullptr) {
1432 if (Arch == Triple::x86_64)
1433 strcpy(info->method, "-[%rdi ");
1434 else if (Arch == Triple::aarch64)
1435 strcpy(info->method, "-[x0 ");
1437 strcpy(info->method, "-[r? ");
1438 strcat(info->method, info->selector_name);
1439 strcat(info->method, "]");
1440 *ReferenceName = info->method;
1441 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1444 info->class_name = nullptr;
1446 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
1447 if (info->selector_name != nullptr) {
1448 if (info->method != nullptr)
1450 info->method = (char *)malloc(17 + strlen(info->selector_name));
1451 if (info->method != nullptr) {
1452 if (Arch == Triple::x86_64)
1453 strcpy(info->method, "-[[%rdi super] ");
1454 else if (Arch == Triple::aarch64)
1455 strcpy(info->method, "-[[x0 super] ");
1457 strcpy(info->method, "-[[r? super] ");
1458 strcat(info->method, info->selector_name);
1459 strcat(info->method, "]");
1460 *ReferenceName = info->method;
1461 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1463 info->class_name = nullptr;
1469 // GuessPointerPointer() is passed the address of what might be a pointer to
1470 // a reference to an Objective-C class, selector, message ref or cfstring.
1471 // If so the value of the pointer is returned and one of the booleans are set
1472 // to true. If not zero is returned and all the booleans are set to false.
1473 static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
1474 struct DisassembleInfo *info,
1475 bool &classref, bool &selref, bool &msgref,
1481 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1482 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1483 for (unsigned I = 0;; ++I) {
1484 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1485 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1486 for (unsigned J = 0; J < Seg.nsects; ++J) {
1487 MachO::section_64 Sec = info->O->getSection64(Load, J);
1488 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
1489 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1490 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
1491 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
1492 strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
1493 ReferenceValue >= Sec.addr &&
1494 ReferenceValue < Sec.addr + Sec.size) {
1495 uint64_t sect_offset = ReferenceValue - Sec.addr;
1496 uint64_t object_offset = Sec.offset + sect_offset;
1497 StringRef MachOContents = info->O->getData();
1498 uint64_t object_size = MachOContents.size();
1499 const char *object_addr = (const char *)MachOContents.data();
1500 if (object_offset < object_size) {
1501 uint64_t pointer_value;
1502 memcpy(&pointer_value, object_addr + object_offset,
1504 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1505 sys::swapByteOrder(pointer_value);
1506 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
1508 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1509 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
1511 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
1512 ReferenceValue + 8 < Sec.addr + Sec.size) {
1514 memcpy(&pointer_value, object_addr + object_offset + 8,
1516 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1517 sys::swapByteOrder(pointer_value);
1518 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
1520 return pointer_value;
1527 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
1528 if (I == LoadCommandCount - 1)
1531 Load = info->O->getNextLoadCommandInfo(Load);
1536 // get_pointer_64 returns a pointer to the bytes in the object file at the
1537 // Address from a section in the Mach-O file. And indirectly returns the
1538 // offset into the section, number of bytes left in the section past the offset
1539 // and which section is was being referenced. If the Address is not in a
1540 // section nullptr is returned.
1541 const char *get_pointer_64(uint64_t Address, uint32_t &offset, uint32_t &left,
1542 SectionRef &S, DisassembleInfo *info) {
1546 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
1547 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
1548 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
1549 if (Address >= SectAddress && Address < SectAddress + SectSize) {
1550 S = (*(info->Sections))[SectIdx];
1551 offset = Address - SectAddress;
1552 left = SectSize - offset;
1553 StringRef SectContents;
1554 ((*(info->Sections))[SectIdx]).getContents(SectContents);
1555 return SectContents.data() + offset;
1561 // get_symbol_64() returns the name of a symbol (or nullptr) and the address of
1562 // the symbol indirectly through n_value. Based on the relocation information
1563 // for the specified section offset in the specified section reference.
1564 const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
1565 DisassembleInfo *info, uint64_t &n_value) {
1567 if (info->verbose == false)
1570 // See if there is an external relocation entry at the sect_offset.
1571 bool reloc_found = false;
1573 MachO::any_relocation_info RE;
1574 bool isExtern = false;
1576 for (const RelocationRef &Reloc : S.relocations()) {
1577 uint64_t RelocOffset;
1578 Reloc.getOffset(RelocOffset);
1579 if (RelocOffset == sect_offset) {
1580 Rel = Reloc.getRawDataRefImpl();
1581 RE = info->O->getRelocation(Rel);
1582 if (info->O->isRelocationScattered(RE))
1584 isExtern = info->O->getPlainRelocationExternal(RE);
1586 symbol_iterator RelocSym = Reloc.getSymbol();
1593 // If there is an external relocation entry for a symbol in this section
1594 // at this section_offset then use that symbol's value for the n_value
1595 // and return its name.
1596 const char *SymbolName = nullptr;
1597 if (reloc_found && isExtern) {
1598 Symbol.getAddress(n_value);
1600 Symbol.getName(name);
1601 if (!name.empty()) {
1602 SymbolName = name.data();
1607 // TODO: For fully linked images, look through the external relocation
1608 // entries off the dynamic symtab command. For these the r_offset is from the
1609 // start of the first writeable segment in the Mach-O file. So the offset
1610 // to this section from that segment is passed to this routine by the caller,
1611 // as the database_offset. Which is the difference of the section's starting
1612 // address and the first writable segment.
1614 // NOTE: need add passing the database_offset to this routine.
1616 // TODO: We did not find an external relocation entry so look up the
1617 // ReferenceValue as an address of a symbol and if found return that symbol's
1620 // NOTE: need add passing the ReferenceValue to this routine. Then that code
1621 // would simply be this:
1622 // SymbolName = GuessSymbolName(ReferenceValue, info);
1627 // These are structs in the Objective-C meta data and read to produce the
1628 // comments for disassembly. While these are part of the ABI they are no
1629 // public defintions. So the are here not in include/llvm/Support/MachO.h .
1631 // The cfstring object in a 64-bit Mach-O file.
1632 struct cfstring64_t {
1633 uint64_t isa; // class64_t * (64-bit pointer)
1634 uint64_t flags; // flag bits
1635 uint64_t characters; // char * (64-bit pointer)
1636 uint64_t length; // number of non-NULL characters in above
1639 // The class object in a 64-bit Mach-O file.
1641 uint64_t isa; // class64_t * (64-bit pointer)
1642 uint64_t superclass; // class64_t * (64-bit pointer)
1643 uint64_t cache; // Cache (64-bit pointer)
1644 uint64_t vtable; // IMP * (64-bit pointer)
1645 uint64_t data; // class_ro64_t * (64-bit pointer)
1648 struct class_ro64_t {
1650 uint32_t instanceStart;
1651 uint32_t instanceSize;
1653 uint64_t ivarLayout; // const uint8_t * (64-bit pointer)
1654 uint64_t name; // const char * (64-bit pointer)
1655 uint64_t baseMethods; // const method_list_t * (64-bit pointer)
1656 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer)
1657 uint64_t ivars; // const ivar_list_t * (64-bit pointer)
1658 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
1659 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
1662 inline void swapStruct(struct cfstring64_t &cfs) {
1663 sys::swapByteOrder(cfs.isa);
1664 sys::swapByteOrder(cfs.flags);
1665 sys::swapByteOrder(cfs.characters);
1666 sys::swapByteOrder(cfs.length);
1669 inline void swapStruct(struct class64_t &c) {
1670 sys::swapByteOrder(c.isa);
1671 sys::swapByteOrder(c.superclass);
1672 sys::swapByteOrder(c.cache);
1673 sys::swapByteOrder(c.vtable);
1674 sys::swapByteOrder(c.data);
1677 inline void swapStruct(struct class_ro64_t &cro) {
1678 sys::swapByteOrder(cro.flags);
1679 sys::swapByteOrder(cro.instanceStart);
1680 sys::swapByteOrder(cro.instanceSize);
1681 sys::swapByteOrder(cro.reserved);
1682 sys::swapByteOrder(cro.ivarLayout);
1683 sys::swapByteOrder(cro.name);
1684 sys::swapByteOrder(cro.baseMethods);
1685 sys::swapByteOrder(cro.baseProtocols);
1686 sys::swapByteOrder(cro.ivars);
1687 sys::swapByteOrder(cro.weakIvarLayout);
1688 sys::swapByteOrder(cro.baseProperties);
1691 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
1692 struct DisassembleInfo *info);
1694 // get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
1695 // to an Objective-C class and returns the class name. It is also passed the
1696 // address of the pointer, so when the pointer is zero as it can be in an .o
1697 // file, that is used to look for an external relocation entry with a symbol
1699 const char *get_objc2_64bit_class_name(uint64_t pointer_value,
1700 uint64_t ReferenceValue,
1701 struct DisassembleInfo *info) {
1703 uint32_t offset, left;
1706 // The pointer_value can be 0 in an object file and have a relocation
1707 // entry for the class symbol at the ReferenceValue (the address of the
1709 if (pointer_value == 0) {
1710 r = get_pointer_64(ReferenceValue, offset, left, S, info);
1711 if (r == nullptr || left < sizeof(uint64_t))
1714 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
1715 if (symbol_name == nullptr)
1717 const char *class_name = strrchr(symbol_name, '$');
1718 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
1719 return class_name + 2;
1724 // The case were the pointer_value is non-zero and points to a class defined
1725 // in this Mach-O file.
1726 r = get_pointer_64(pointer_value, offset, left, S, info);
1727 if (r == nullptr || left < sizeof(struct class64_t))
1730 memcpy(&c, r, sizeof(struct class64_t));
1731 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1735 r = get_pointer_64(c.data, offset, left, S, info);
1736 if (r == nullptr || left < sizeof(struct class_ro64_t))
1738 struct class_ro64_t cro;
1739 memcpy(&cro, r, sizeof(struct class_ro64_t));
1740 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1744 const char *name = get_pointer_64(cro.name, offset, left, S, info);
1748 // get_objc2_64bit_cfstring_name is used for disassembly and is passed a
1749 // pointer to a cfstring and returns its name or nullptr.
1750 const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
1751 struct DisassembleInfo *info) {
1752 const char *r, *name;
1753 uint32_t offset, left;
1755 struct cfstring64_t cfs;
1756 uint64_t cfs_characters;
1758 r = get_pointer_64(ReferenceValue, offset, left, S, info);
1759 if (r == nullptr || left < sizeof(struct cfstring64_t))
1761 memcpy(&cfs, r, sizeof(struct cfstring64_t));
1762 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1764 if (cfs.characters == 0) {
1766 const char *symbol_name = get_symbol_64(
1767 offset + offsetof(struct cfstring64_t, characters), S, info, n_value);
1768 if (symbol_name == nullptr)
1770 cfs_characters = n_value;
1772 cfs_characters = cfs.characters;
1773 name = get_pointer_64(cfs_characters, offset, left, S, info);
1778 // get_objc2_64bit_selref() is used for disassembly and is passed a the address
1779 // of a pointer to an Objective-C selector reference when the pointer value is
1780 // zero as in a .o file and is likely to have a external relocation entry with
1781 // who's symbol's n_value is the real pointer to the selector name. If that is
1782 // the case the real pointer to the selector name is returned else 0 is
1784 uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
1785 struct DisassembleInfo *info) {
1786 uint32_t offset, left;
1789 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
1790 if (r == nullptr || left < sizeof(uint64_t))
1793 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
1794 if (symbol_name == nullptr)
1799 // GuessLiteralPointer returns a string which for the item in the Mach-O file
1800 // for the address passed in as ReferenceValue for printing as a comment with
1801 // the instruction and also returns the corresponding type of that item
1802 // indirectly through ReferenceType.
1804 // If ReferenceValue is an address of literal cstring then a pointer to the
1805 // cstring is returned and ReferenceType is set to
1806 // LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
1808 // If ReferenceValue is an address of an Objective-C CFString, Selector ref or
1809 // Class ref that name is returned and the ReferenceType is set accordingly.
1811 // Lastly, literals which are Symbol address in a literal pool are looked for
1812 // and if found the symbol name is returned and ReferenceType is set to
1813 // LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
1815 // If there is no item in the Mach-O file for the address passed in as
1816 // ReferenceValue nullptr is returned and ReferenceType is unchanged.
1817 const char *GuessLiteralPointer(uint64_t ReferenceValue, uint64_t ReferencePC,
1818 uint64_t *ReferenceType,
1819 struct DisassembleInfo *info) {
1820 // First see if there is an external relocation entry at the ReferencePC.
1821 uint64_t sect_addr = info->S.getAddress();
1822 uint64_t sect_offset = ReferencePC - sect_addr;
1823 bool reloc_found = false;
1825 MachO::any_relocation_info RE;
1826 bool isExtern = false;
1828 for (const RelocationRef &Reloc : info->S.relocations()) {
1829 uint64_t RelocOffset;
1830 Reloc.getOffset(RelocOffset);
1831 if (RelocOffset == sect_offset) {
1832 Rel = Reloc.getRawDataRefImpl();
1833 RE = info->O->getRelocation(Rel);
1834 if (info->O->isRelocationScattered(RE))
1836 isExtern = info->O->getPlainRelocationExternal(RE);
1838 symbol_iterator RelocSym = Reloc.getSymbol();
1845 // If there is an external relocation entry for a symbol in a section
1846 // then used that symbol's value for the value of the reference.
1847 if (reloc_found && isExtern) {
1848 if (info->O->getAnyRelocationPCRel(RE)) {
1849 unsigned Type = info->O->getAnyRelocationType(RE);
1850 if (Type == MachO::X86_64_RELOC_SIGNED) {
1851 Symbol.getAddress(ReferenceValue);
1856 // Look for literals such as Objective-C CFStrings refs, Selector refs,
1857 // Message refs and Class refs.
1858 bool classref, selref, msgref, cfstring;
1859 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
1860 selref, msgref, cfstring);
1861 if (classref == true && pointer_value == 0) {
1862 // Note the ReferenceValue is a pointer into the __objc_classrefs section.
1863 // And the pointer_value in that section is typically zero as it will be
1864 // set by dyld as part of the "bind information".
1865 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
1866 if (name != nullptr) {
1867 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
1868 const char *class_name = strrchr(name, '$');
1869 if (class_name != nullptr && class_name[1] == '_' &&
1870 class_name[2] != '\0') {
1871 info->class_name = class_name + 2;
1877 if (classref == true) {
1878 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
1880 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
1881 if (name != nullptr)
1882 info->class_name = name;
1884 name = "bad class ref";
1888 if (cfstring == true) {
1889 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref;
1890 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
1894 if (selref == true && pointer_value == 0)
1895 pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
1897 if (pointer_value != 0)
1898 ReferenceValue = pointer_value;
1900 const char *name = GuessCstringPointer(ReferenceValue, info);
1902 if (pointer_value != 0 && selref == true) {
1903 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref;
1904 info->selector_name = name;
1905 } else if (pointer_value != 0 && msgref == true) {
1906 info->class_name = nullptr;
1907 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref;
1908 info->selector_name = name;
1910 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr;
1914 // Lastly look for an indirect symbol with this ReferenceValue which is in
1915 // a literal pool. If found return that symbol name.
1916 name = GuessIndirectSymbol(ReferenceValue, info);
1918 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr;
1925 // SymbolizerSymbolLookUp is the symbol lookup function passed when creating
1926 // the Symbolizer. It looks up the ReferenceValue using the info passed via the
1927 // pointer to the struct DisassembleInfo that was passed when MCSymbolizer
1928 // is created and returns the symbol name that matches the ReferenceValue or
1929 // nullptr if none. The ReferenceType is passed in for the IN type of
1930 // reference the instruction is making from the values in defined in the header
1931 // "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific
1932 // Out type and the ReferenceName will also be set which is added as a comment
1933 // to the disassembled instruction.
1936 // If the symbol name is a C++ mangled name then the demangled name is
1937 // returned through ReferenceName and ReferenceType is set to
1938 // LLVMDisassembler_ReferenceType_DeMangled_Name .
1941 // When this is called to get a symbol name for a branch target then the
1942 // ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
1943 // SymbolValue will be looked for in the indirect symbol table to determine if
1944 // it is an address for a symbol stub. If so then the symbol name for that
1945 // stub is returned indirectly through ReferenceName and then ReferenceType is
1946 // set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
1948 // When this is called with an value loaded via a PC relative load then
1949 // ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
1950 // SymbolValue is checked to be an address of literal pointer, symbol pointer,
1951 // or an Objective-C meta data reference. If so the output ReferenceType is
1952 // set to correspond to that as well as setting the ReferenceName.
1953 const char *SymbolizerSymbolLookUp(void *DisInfo, uint64_t ReferenceValue,
1954 uint64_t *ReferenceType,
1955 uint64_t ReferencePC,
1956 const char **ReferenceName) {
1957 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1958 // If no verbose symbolic information is wanted then just return nullptr.
1959 if (info->verbose == false) {
1960 *ReferenceName = nullptr;
1961 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1965 const char *SymbolName = GuessSymbolName(ReferenceValue, info);
1967 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) {
1968 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
1969 if (*ReferenceName != nullptr) {
1970 method_reference(info, ReferenceType, ReferenceName);
1971 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message)
1972 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub;
1975 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
1976 if (info->demangled_name != nullptr)
1977 free(info->demangled_name);
1979 info->demangled_name =
1980 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
1981 if (info->demangled_name != nullptr) {
1982 *ReferenceName = info->demangled_name;
1983 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
1985 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1988 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1989 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) {
1991 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
1993 method_reference(info, ReferenceType, ReferenceName);
1995 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
1996 // If this is arm64 and the reference is an adrp instruction save the
1997 // instruction, passed in ReferenceValue and the address of the instruction
1998 // for use later if we see and add immediate instruction.
1999 } else if (info->O->getArch() == Triple::aarch64 &&
2000 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) {
2001 info->adrp_inst = ReferenceValue;
2002 info->adrp_addr = ReferencePC;
2003 SymbolName = nullptr;
2004 *ReferenceName = nullptr;
2005 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2006 // If this is arm64 and reference is an add immediate instruction and we
2008 // seen an adrp instruction just before it and the adrp's Xd register
2010 // this add's Xn register reconstruct the value being referenced and look to
2011 // see if it is a literal pointer. Note the add immediate instruction is
2012 // passed in ReferenceValue.
2013 } else if (info->O->getArch() == Triple::aarch64 &&
2014 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri &&
2015 ReferencePC - 4 == info->adrp_addr &&
2016 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2017 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2018 uint32_t addxri_inst;
2019 uint64_t adrp_imm, addxri_imm;
2022 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2023 if (info->adrp_inst & 0x0200000)
2024 adrp_imm |= 0xfffffffffc000000LL;
2026 addxri_inst = ReferenceValue;
2027 addxri_imm = (addxri_inst >> 10) & 0xfff;
2028 if (((addxri_inst >> 22) & 0x3) == 1)
2031 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2032 (adrp_imm << 12) + addxri_imm;
2035 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2036 if (*ReferenceName == nullptr)
2037 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2038 // If this is arm64 and the reference is a load register instruction and we
2039 // have seen an adrp instruction just before it and the adrp's Xd register
2040 // matches this add's Xn register reconstruct the value being referenced and
2041 // look to see if it is a literal pointer. Note the load register
2042 // instruction is passed in ReferenceValue.
2043 } else if (info->O->getArch() == Triple::aarch64 &&
2044 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui &&
2045 ReferencePC - 4 == info->adrp_addr &&
2046 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2047 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2048 uint32_t ldrxui_inst;
2049 uint64_t adrp_imm, ldrxui_imm;
2052 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2053 if (info->adrp_inst & 0x0200000)
2054 adrp_imm |= 0xfffffffffc000000LL;
2056 ldrxui_inst = ReferenceValue;
2057 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
2059 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2060 (adrp_imm << 12) + (ldrxui_imm << 3);
2063 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2064 if (*ReferenceName == nullptr)
2065 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2067 // If this arm64 and is an load register (PC-relative) instruction the
2068 // ReferenceValue is the PC plus the immediate value.
2069 else if (info->O->getArch() == Triple::aarch64 &&
2070 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl ||
2071 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) {
2073 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2074 if (*ReferenceName == nullptr)
2075 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2078 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2079 if (info->demangled_name != nullptr)
2080 free(info->demangled_name);
2082 info->demangled_name =
2083 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2084 if (info->demangled_name != nullptr) {
2085 *ReferenceName = info->demangled_name;
2086 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2091 *ReferenceName = nullptr;
2092 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2098 /// \brief Emits the comments that are stored in the CommentStream.
2099 /// Each comment in the CommentStream must end with a newline.
2100 static void emitComments(raw_svector_ostream &CommentStream,
2101 SmallString<128> &CommentsToEmit,
2102 formatted_raw_ostream &FormattedOS,
2103 const MCAsmInfo &MAI) {
2104 // Flush the stream before taking its content.
2105 CommentStream.flush();
2106 StringRef Comments = CommentsToEmit.str();
2107 // Get the default information for printing a comment.
2108 const char *CommentBegin = MAI.getCommentString();
2109 unsigned CommentColumn = MAI.getCommentColumn();
2110 bool IsFirst = true;
2111 while (!Comments.empty()) {
2113 FormattedOS << '\n';
2114 // Emit a line of comments.
2115 FormattedOS.PadToColumn(CommentColumn);
2116 size_t Position = Comments.find('\n');
2117 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
2118 // Move after the newline character.
2119 Comments = Comments.substr(Position + 1);
2122 FormattedOS.flush();
2124 // Tell the comment stream that the vector changed underneath it.
2125 CommentsToEmit.clear();
2126 CommentStream.resync();
2129 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF) {
2130 const char *McpuDefault = nullptr;
2131 const Target *ThumbTarget = nullptr;
2132 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
2134 // GetTarget prints out stuff.
2137 if (MCPU.empty() && McpuDefault)
2140 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
2141 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
2143 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
2145 // Package up features to be passed to target/subtarget
2146 std::string FeaturesStr;
2147 if (MAttrs.size()) {
2148 SubtargetFeatures Features;
2149 for (unsigned i = 0; i != MAttrs.size(); ++i)
2150 Features.AddFeature(MAttrs[i]);
2151 FeaturesStr = Features.getString();
2154 // Set up disassembler.
2155 std::unique_ptr<const MCRegisterInfo> MRI(
2156 TheTarget->createMCRegInfo(TripleName));
2157 std::unique_ptr<const MCAsmInfo> AsmInfo(
2158 TheTarget->createMCAsmInfo(*MRI, TripleName));
2159 std::unique_ptr<const MCSubtargetInfo> STI(
2160 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
2161 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
2162 std::unique_ptr<MCDisassembler> DisAsm(
2163 TheTarget->createMCDisassembler(*STI, Ctx));
2164 std::unique_ptr<MCSymbolizer> Symbolizer;
2165 struct DisassembleInfo SymbolizerInfo;
2166 std::unique_ptr<MCRelocationInfo> RelInfo(
2167 TheTarget->createMCRelocationInfo(TripleName, Ctx));
2169 Symbolizer.reset(TheTarget->createMCSymbolizer(
2170 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2171 &SymbolizerInfo, &Ctx, std::move(RelInfo)));
2172 DisAsm->setSymbolizer(std::move(Symbolizer));
2174 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
2175 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
2176 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
2177 // Set the display preference for hex vs. decimal immediates.
2178 IP->setPrintImmHex(PrintImmHex);
2179 // Comment stream and backing vector.
2180 SmallString<128> CommentsToEmit;
2181 raw_svector_ostream CommentStream(CommentsToEmit);
2182 // FIXME: Setting the CommentStream in the InstPrinter is problematic in that
2183 // if it is done then arm64 comments for string literals don't get printed
2184 // and some constant get printed instead and not setting it causes intel
2185 // (32-bit and 64-bit) comments printed with different spacing before the
2186 // comment causing different diffs with the 'C' disassembler library API.
2187 // IP->setCommentStream(CommentStream);
2189 if (!AsmInfo || !STI || !DisAsm || !IP) {
2190 errs() << "error: couldn't initialize disassembler for target "
2191 << TripleName << '\n';
2195 // Set up thumb disassembler.
2196 std::unique_ptr<const MCRegisterInfo> ThumbMRI;
2197 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
2198 std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
2199 std::unique_ptr<MCDisassembler> ThumbDisAsm;
2200 std::unique_ptr<MCInstPrinter> ThumbIP;
2201 std::unique_ptr<MCContext> ThumbCtx;
2202 std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
2203 struct DisassembleInfo ThumbSymbolizerInfo;
2204 std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
2206 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
2208 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
2210 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
2211 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
2212 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
2213 MCContext *PtrThumbCtx = ThumbCtx.get();
2215 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
2217 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
2218 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
2219 &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo)));
2220 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
2222 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
2223 ThumbIP.reset(ThumbTarget->createMCInstPrinter(
2224 ThumbAsmPrinterVariant, *ThumbAsmInfo, *ThumbInstrInfo, *ThumbMRI,
2226 // Set the display preference for hex vs. decimal immediates.
2227 ThumbIP->setPrintImmHex(PrintImmHex);
2230 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
2231 errs() << "error: couldn't initialize disassembler for target "
2232 << ThumbTripleName << '\n';
2236 MachO::mach_header Header = MachOOF->getHeader();
2238 // FIXME: Using the -cfg command line option, this code used to be able to
2239 // annotate relocations with the referenced symbol's name, and if this was
2240 // inside a __[cf]string section, the data it points to. This is now replaced
2241 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
2242 std::vector<SectionRef> Sections;
2243 std::vector<SymbolRef> Symbols;
2244 SmallVector<uint64_t, 8> FoundFns;
2245 uint64_t BaseSegmentAddress;
2247 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
2248 BaseSegmentAddress);
2250 // Sort the symbols by address, just in case they didn't come in that way.
2251 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
2253 // Build a data in code table that is sorted on by the address of each entry.
2254 uint64_t BaseAddress = 0;
2255 if (Header.filetype == MachO::MH_OBJECT)
2256 BaseAddress = Sections[0].getAddress();
2258 BaseAddress = BaseSegmentAddress;
2260 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
2263 DI->getOffset(Offset);
2264 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
2266 array_pod_sort(Dices.begin(), Dices.end());
2269 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
2271 raw_ostream &DebugOut = nulls();
2274 std::unique_ptr<DIContext> diContext;
2275 ObjectFile *DbgObj = MachOOF;
2276 // Try to find debug info and set up the DIContext for it.
2278 // A separate DSym file path was specified, parse it as a macho file,
2279 // get the sections and supply it to the section name parsing machinery.
2280 if (!DSYMFile.empty()) {
2281 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
2282 MemoryBuffer::getFileOrSTDIN(DSYMFile);
2283 if (std::error_code EC = BufOrErr.getError()) {
2284 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
2288 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
2293 // Setup the DIContext
2294 diContext.reset(DIContext::getDWARFContext(*DbgObj));
2297 // TODO: For now this only disassembles the (__TEXT,__text) section (see the
2298 // checks in the code below at the top of this loop). It should allow a
2299 // darwin otool(1) like -s option to disassemble any named segment & section
2300 // that is marked as containing instructions with the attributes
2301 // S_ATTR_PURE_INSTRUCTIONS or S_ATTR_SOME_INSTRUCTIONS in the flags field of
2302 // the section structure.
2303 outs() << "(__TEXT,__text) section\n";
2305 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
2307 bool SectIsText = Sections[SectIdx].isText();
2308 if (SectIsText == false)
2312 if (Sections[SectIdx].getName(SectName) || SectName != "__text")
2313 continue; // Skip non-text sections
2315 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
2317 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
2318 if (SegmentName != "__TEXT")
2322 Sections[SectIdx].getContents(BytesStr);
2323 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
2325 uint64_t SectAddress = Sections[SectIdx].getAddress();
2327 bool symbolTableWorked = false;
2329 // Parse relocations.
2330 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
2331 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
2332 uint64_t RelocOffset;
2333 Reloc.getOffset(RelocOffset);
2334 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2335 RelocOffset -= SectionAddress;
2337 symbol_iterator RelocSym = Reloc.getSymbol();
2339 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
2341 array_pod_sort(Relocs.begin(), Relocs.end());
2343 // Create a map of symbol addresses to symbol names for use by
2344 // the SymbolizerSymbolLookUp() routine.
2345 SymbolAddressMap AddrMap;
2346 for (const SymbolRef &Symbol : MachOOF->symbols()) {
2349 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
2350 ST == SymbolRef::ST_Other) {
2352 Symbol.getAddress(Address);
2354 Symbol.getName(SymName);
2355 AddrMap[Address] = SymName;
2358 // Set up the block of info used by the Symbolizer call backs.
2359 SymbolizerInfo.verbose = true;
2360 SymbolizerInfo.O = MachOOF;
2361 SymbolizerInfo.S = Sections[SectIdx];
2362 SymbolizerInfo.AddrMap = &AddrMap;
2363 SymbolizerInfo.Sections = &Sections;
2364 SymbolizerInfo.class_name = nullptr;
2365 SymbolizerInfo.selector_name = nullptr;
2366 SymbolizerInfo.method = nullptr;
2367 SymbolizerInfo.demangled_name = nullptr;
2368 SymbolizerInfo.bindtable = nullptr;
2369 SymbolizerInfo.adrp_addr = 0;
2370 SymbolizerInfo.adrp_inst = 0;
2371 // Same for the ThumbSymbolizer
2372 ThumbSymbolizerInfo.verbose = true;
2373 ThumbSymbolizerInfo.O = MachOOF;
2374 ThumbSymbolizerInfo.S = Sections[SectIdx];
2375 ThumbSymbolizerInfo.AddrMap = &AddrMap;
2376 ThumbSymbolizerInfo.Sections = &Sections;
2377 ThumbSymbolizerInfo.class_name = nullptr;
2378 ThumbSymbolizerInfo.selector_name = nullptr;
2379 ThumbSymbolizerInfo.method = nullptr;
2380 ThumbSymbolizerInfo.demangled_name = nullptr;
2381 ThumbSymbolizerInfo.bindtable = nullptr;
2382 ThumbSymbolizerInfo.adrp_addr = 0;
2383 ThumbSymbolizerInfo.adrp_inst = 0;
2385 // Disassemble symbol by symbol.
2386 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
2388 Symbols[SymIdx].getName(SymName);
2391 Symbols[SymIdx].getType(ST);
2392 if (ST != SymbolRef::ST_Function)
2395 // Make sure the symbol is defined in this section.
2396 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
2400 // Start at the address of the symbol relative to the section's address.
2402 uint64_t SectionAddress = Sections[SectIdx].getAddress();
2403 Symbols[SymIdx].getAddress(Start);
2404 Start -= SectionAddress;
2406 // Stop disassembling either at the beginning of the next symbol or at
2407 // the end of the section.
2408 bool containsNextSym = false;
2409 uint64_t NextSym = 0;
2410 uint64_t NextSymIdx = SymIdx + 1;
2411 while (Symbols.size() > NextSymIdx) {
2412 SymbolRef::Type NextSymType;
2413 Symbols[NextSymIdx].getType(NextSymType);
2414 if (NextSymType == SymbolRef::ST_Function) {
2416 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
2417 Symbols[NextSymIdx].getAddress(NextSym);
2418 NextSym -= SectionAddress;
2424 uint64_t SectSize = Sections[SectIdx].getSize();
2425 uint64_t End = containsNextSym ? NextSym : SectSize;
2428 symbolTableWorked = true;
2430 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
2432 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
2434 outs() << SymName << ":\n";
2435 DILineInfo lastLine;
2436 for (uint64_t Index = Start; Index < End; Index += Size) {
2439 uint64_t PC = SectAddress + Index;
2440 if (FullLeadingAddr) {
2441 if (MachOOF->is64Bit())
2442 outs() << format("%016" PRIx64, PC);
2444 outs() << format("%08" PRIx64, PC);
2446 outs() << format("%8" PRIx64 ":", PC);
2451 // Check the data in code table here to see if this is data not an
2452 // instruction to be disassembled.
2454 Dice.push_back(std::make_pair(PC, DiceRef()));
2455 dice_table_iterator DTI =
2456 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
2457 compareDiceTableEntries);
2458 if (DTI != Dices.end()) {
2460 DTI->second.getLength(Length);
2462 DTI->second.getKind(Kind);
2463 Size = DumpDataInCode(reinterpret_cast<const char *>(Bytes.data()) +
2466 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
2467 (PC == (DTI->first + Length - 1)) && (Length & 1))
2472 SmallVector<char, 64> AnnotationsBytes;
2473 raw_svector_ostream Annotations(AnnotationsBytes);
2477 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
2478 PC, DebugOut, Annotations);
2480 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
2481 DebugOut, Annotations);
2483 if (!NoShowRawInsn) {
2484 DumpBytes(StringRef(
2485 reinterpret_cast<const char *>(Bytes.data()) + Index, Size));
2487 formatted_raw_ostream FormattedOS(outs());
2488 Annotations.flush();
2489 StringRef AnnotationsStr = Annotations.str();
2491 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr);
2493 IP->printInst(&Inst, FormattedOS, AnnotationsStr);
2494 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
2496 // Print debug info.
2498 DILineInfo dli = diContext->getLineInfoForAddress(PC);
2499 // Print valid line info if it changed.
2500 if (dli != lastLine && dli.Line != 0)
2501 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
2507 unsigned int Arch = MachOOF->getArch();
2508 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2509 outs() << format("\t.byte 0x%02x #bad opcode\n",
2510 *(Bytes.data() + Index) & 0xff);
2511 Size = 1; // skip exactly one illegible byte and move on.
2512 } else if (Arch == Triple::aarch64) {
2513 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
2514 (*(Bytes.data() + Index + 1) & 0xff) << 8 |
2515 (*(Bytes.data() + Index + 2) & 0xff) << 16 |
2516 (*(Bytes.data() + Index + 3) & 0xff) << 24;
2517 outs() << format("\t.long\t0x%08x\n", opcode);
2520 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2522 Size = 1; // skip illegible bytes
2527 if (!symbolTableWorked) {
2528 // Reading the symbol table didn't work, disassemble the whole section.
2529 uint64_t SectAddress = Sections[SectIdx].getAddress();
2530 uint64_t SectSize = Sections[SectIdx].getSize();
2532 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
2535 uint64_t PC = SectAddress + Index;
2536 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
2537 DebugOut, nulls())) {
2538 if (FullLeadingAddr) {
2539 if (MachOOF->is64Bit())
2540 outs() << format("%016" PRIx64, PC);
2542 outs() << format("%08" PRIx64, PC);
2544 outs() << format("%8" PRIx64 ":", PC);
2546 if (!NoShowRawInsn) {
2549 StringRef(reinterpret_cast<const char *>(Bytes.data()) + Index,
2552 IP->printInst(&Inst, outs(), "");
2555 unsigned int Arch = MachOOF->getArch();
2556 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
2557 outs() << format("\t.byte 0x%02x #bad opcode\n",
2558 *(Bytes.data() + Index) & 0xff);
2559 InstSize = 1; // skip exactly one illegible byte and move on.
2561 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2563 InstSize = 1; // skip illegible bytes
2568 // The TripleName's need to be reset if we are called again for a different
2571 ThumbTripleName = "";
2573 if (SymbolizerInfo.method != nullptr)
2574 free(SymbolizerInfo.method);
2575 if (SymbolizerInfo.demangled_name != nullptr)
2576 free(SymbolizerInfo.demangled_name);
2577 if (SymbolizerInfo.bindtable != nullptr)
2578 delete SymbolizerInfo.bindtable;
2579 if (ThumbSymbolizerInfo.method != nullptr)
2580 free(ThumbSymbolizerInfo.method);
2581 if (ThumbSymbolizerInfo.demangled_name != nullptr)
2582 free(ThumbSymbolizerInfo.demangled_name);
2583 if (ThumbSymbolizerInfo.bindtable != nullptr)
2584 delete ThumbSymbolizerInfo.bindtable;
2588 //===----------------------------------------------------------------------===//
2589 // __compact_unwind section dumping
2590 //===----------------------------------------------------------------------===//
2594 template <typename T> static uint64_t readNext(const char *&Buf) {
2595 using llvm::support::little;
2596 using llvm::support::unaligned;
2598 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
2603 struct CompactUnwindEntry {
2604 uint32_t OffsetInSection;
2606 uint64_t FunctionAddr;
2608 uint32_t CompactEncoding;
2609 uint64_t PersonalityAddr;
2612 RelocationRef FunctionReloc;
2613 RelocationRef PersonalityReloc;
2614 RelocationRef LSDAReloc;
2616 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
2617 : OffsetInSection(Offset) {
2619 read<uint64_t>(Contents.data() + Offset);
2621 read<uint32_t>(Contents.data() + Offset);
2625 template <typename UIntPtr> void read(const char *Buf) {
2626 FunctionAddr = readNext<UIntPtr>(Buf);
2627 Length = readNext<uint32_t>(Buf);
2628 CompactEncoding = readNext<uint32_t>(Buf);
2629 PersonalityAddr = readNext<UIntPtr>(Buf);
2630 LSDAAddr = readNext<UIntPtr>(Buf);
2635 /// Given a relocation from __compact_unwind, consisting of the RelocationRef
2636 /// and data being relocated, determine the best base Name and Addend to use for
2637 /// display purposes.
2639 /// 1. An Extern relocation will directly reference a symbol (and the data is
2640 /// then already an addend), so use that.
2641 /// 2. Otherwise the data is an offset in the object file's layout; try to find
2642 // a symbol before it in the same section, and use the offset from there.
2643 /// 3. Finally, if all that fails, fall back to an offset from the start of the
2644 /// referenced section.
2645 static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
2646 std::map<uint64_t, SymbolRef> &Symbols,
2647 const RelocationRef &Reloc, uint64_t Addr,
2648 StringRef &Name, uint64_t &Addend) {
2649 if (Reloc.getSymbol() != Obj->symbol_end()) {
2650 Reloc.getSymbol()->getName(Name);
2655 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
2656 SectionRef RelocSection = Obj->getRelocationSection(RE);
2658 uint64_t SectionAddr = RelocSection.getAddress();
2660 auto Sym = Symbols.upper_bound(Addr);
2661 if (Sym == Symbols.begin()) {
2662 // The first symbol in the object is after this reference, the best we can
2663 // do is section-relative notation.
2664 RelocSection.getName(Name);
2665 Addend = Addr - SectionAddr;
2669 // Go back one so that SymbolAddress <= Addr.
2672 section_iterator SymSection = Obj->section_end();
2673 Sym->second.getSection(SymSection);
2674 if (RelocSection == *SymSection) {
2675 // There's a valid symbol in the same section before this reference.
2676 Sym->second.getName(Name);
2677 Addend = Addr - Sym->first;
2681 // There is a symbol before this reference, but it's in a different
2682 // section. Probably not helpful to mention it, so use the section name.
2683 RelocSection.getName(Name);
2684 Addend = Addr - SectionAddr;
2687 static void printUnwindRelocDest(const MachOObjectFile *Obj,
2688 std::map<uint64_t, SymbolRef> &Symbols,
2689 const RelocationRef &Reloc, uint64_t Addr) {
2693 if (!Reloc.getObjectFile())
2696 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
2700 outs() << " + " << format("0x%" PRIx64, Addend);
2704 printMachOCompactUnwindSection(const MachOObjectFile *Obj,
2705 std::map<uint64_t, SymbolRef> &Symbols,
2706 const SectionRef &CompactUnwind) {
2708 assert(Obj->isLittleEndian() &&
2709 "There should not be a big-endian .o with __compact_unwind");
2711 bool Is64 = Obj->is64Bit();
2712 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
2713 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
2716 CompactUnwind.getContents(Contents);
2718 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
2720 // First populate the initial raw offsets, encodings and so on from the entry.
2721 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
2722 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
2723 CompactUnwinds.push_back(Entry);
2726 // Next we need to look at the relocations to find out what objects are
2727 // actually being referred to.
2728 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
2729 uint64_t RelocAddress;
2730 Reloc.getOffset(RelocAddress);
2732 uint32_t EntryIdx = RelocAddress / EntrySize;
2733 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
2734 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
2736 if (OffsetInEntry == 0)
2737 Entry.FunctionReloc = Reloc;
2738 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
2739 Entry.PersonalityReloc = Reloc;
2740 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
2741 Entry.LSDAReloc = Reloc;
2743 llvm_unreachable("Unexpected relocation in __compact_unwind section");
2746 // Finally, we're ready to print the data we've gathered.
2747 outs() << "Contents of __compact_unwind section:\n";
2748 for (auto &Entry : CompactUnwinds) {
2749 outs() << " Entry at offset "
2750 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
2752 // 1. Start of the region this entry applies to.
2753 outs() << " start: " << format("0x%" PRIx64,
2754 Entry.FunctionAddr) << ' ';
2755 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
2758 // 2. Length of the region this entry applies to.
2759 outs() << " length: " << format("0x%" PRIx32, Entry.Length)
2761 // 3. The 32-bit compact encoding.
2762 outs() << " compact encoding: "
2763 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
2765 // 4. The personality function, if present.
2766 if (Entry.PersonalityReloc.getObjectFile()) {
2767 outs() << " personality function: "
2768 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
2769 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
2770 Entry.PersonalityAddr);
2774 // 5. This entry's language-specific data area.
2775 if (Entry.LSDAReloc.getObjectFile()) {
2776 outs() << " LSDA: " << format("0x%" PRIx64,
2777 Entry.LSDAAddr) << ' ';
2778 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
2784 //===----------------------------------------------------------------------===//
2785 // __unwind_info section dumping
2786 //===----------------------------------------------------------------------===//
2788 static void printRegularSecondLevelUnwindPage(const char *PageStart) {
2789 const char *Pos = PageStart;
2790 uint32_t Kind = readNext<uint32_t>(Pos);
2792 assert(Kind == 2 && "kind for a regular 2nd level index should be 2");
2794 uint16_t EntriesStart = readNext<uint16_t>(Pos);
2795 uint16_t NumEntries = readNext<uint16_t>(Pos);
2797 Pos = PageStart + EntriesStart;
2798 for (unsigned i = 0; i < NumEntries; ++i) {
2799 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
2800 uint32_t Encoding = readNext<uint32_t>(Pos);
2802 outs() << " [" << i << "]: "
2803 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
2805 << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n';
2809 static void printCompressedSecondLevelUnwindPage(
2810 const char *PageStart, uint32_t FunctionBase,
2811 const SmallVectorImpl<uint32_t> &CommonEncodings) {
2812 const char *Pos = PageStart;
2813 uint32_t Kind = readNext<uint32_t>(Pos);
2815 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3");
2817 uint16_t EntriesStart = readNext<uint16_t>(Pos);
2818 uint16_t NumEntries = readNext<uint16_t>(Pos);
2820 uint16_t EncodingsStart = readNext<uint16_t>(Pos);
2821 readNext<uint16_t>(Pos);
2822 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
2823 PageStart + EncodingsStart);
2825 Pos = PageStart + EntriesStart;
2826 for (unsigned i = 0; i < NumEntries; ++i) {
2827 uint32_t Entry = readNext<uint32_t>(Pos);
2828 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
2829 uint32_t EncodingIdx = Entry >> 24;
2832 if (EncodingIdx < CommonEncodings.size())
2833 Encoding = CommonEncodings[EncodingIdx];
2835 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
2837 outs() << " [" << i << "]: "
2838 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
2840 << "encoding[" << EncodingIdx
2841 << "]=" << format("0x%08" PRIx32, Encoding) << '\n';
2845 static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
2846 std::map<uint64_t, SymbolRef> &Symbols,
2847 const SectionRef &UnwindInfo) {
2849 assert(Obj->isLittleEndian() &&
2850 "There should not be a big-endian .o with __unwind_info");
2852 outs() << "Contents of __unwind_info section:\n";
2855 UnwindInfo.getContents(Contents);
2856 const char *Pos = Contents.data();
2858 //===----------------------------------
2860 //===----------------------------------
2862 uint32_t Version = readNext<uint32_t>(Pos);
2863 outs() << " Version: "
2864 << format("0x%" PRIx32, Version) << '\n';
2865 assert(Version == 1 && "only understand version 1");
2867 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
2868 outs() << " Common encodings array section offset: "
2869 << format("0x%" PRIx32, CommonEncodingsStart) << '\n';
2870 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
2871 outs() << " Number of common encodings in array: "
2872 << format("0x%" PRIx32, NumCommonEncodings) << '\n';
2874 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
2875 outs() << " Personality function array section offset: "
2876 << format("0x%" PRIx32, PersonalitiesStart) << '\n';
2877 uint32_t NumPersonalities = readNext<uint32_t>(Pos);
2878 outs() << " Number of personality functions in array: "
2879 << format("0x%" PRIx32, NumPersonalities) << '\n';
2881 uint32_t IndicesStart = readNext<uint32_t>(Pos);
2882 outs() << " Index array section offset: "
2883 << format("0x%" PRIx32, IndicesStart) << '\n';
2884 uint32_t NumIndices = readNext<uint32_t>(Pos);
2885 outs() << " Number of indices in array: "
2886 << format("0x%" PRIx32, NumIndices) << '\n';
2888 //===----------------------------------
2889 // A shared list of common encodings
2890 //===----------------------------------
2892 // These occupy indices in the range [0, N] whenever an encoding is referenced
2893 // from a compressed 2nd level index table. In practice the linker only
2894 // creates ~128 of these, so that indices are available to embed encodings in
2895 // the 2nd level index.
2897 SmallVector<uint32_t, 64> CommonEncodings;
2898 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n";
2899 Pos = Contents.data() + CommonEncodingsStart;
2900 for (unsigned i = 0; i < NumCommonEncodings; ++i) {
2901 uint32_t Encoding = readNext<uint32_t>(Pos);
2902 CommonEncodings.push_back(Encoding);
2904 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding)
2908 //===----------------------------------
2909 // Personality functions used in this executable
2910 //===----------------------------------
2912 // There should be only a handful of these (one per source language,
2913 // roughly). Particularly since they only get 2 bits in the compact encoding.
2915 outs() << " Personality functions: (count = " << NumPersonalities << ")\n";
2916 Pos = Contents.data() + PersonalitiesStart;
2917 for (unsigned i = 0; i < NumPersonalities; ++i) {
2918 uint32_t PersonalityFn = readNext<uint32_t>(Pos);
2919 outs() << " personality[" << i + 1
2920 << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n';
2923 //===----------------------------------
2924 // The level 1 index entries
2925 //===----------------------------------
2927 // These specify an approximate place to start searching for the more detailed
2928 // information, sorted by PC.
2931 uint32_t FunctionOffset;
2932 uint32_t SecondLevelPageStart;
2936 SmallVector<IndexEntry, 4> IndexEntries;
2938 outs() << " Top level indices: (count = " << NumIndices << ")\n";
2939 Pos = Contents.data() + IndicesStart;
2940 for (unsigned i = 0; i < NumIndices; ++i) {
2943 Entry.FunctionOffset = readNext<uint32_t>(Pos);
2944 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
2945 Entry.LSDAStart = readNext<uint32_t>(Pos);
2946 IndexEntries.push_back(Entry);
2948 outs() << " [" << i << "]: "
2949 << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset)
2951 << "2nd level page offset="
2952 << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", "
2953 << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n';
2956 //===----------------------------------
2957 // Next come the LSDA tables
2958 //===----------------------------------
2960 // The LSDA layout is rather implicit: it's a contiguous array of entries from
2961 // the first top-level index's LSDAOffset to the last (sentinel).
2963 outs() << " LSDA descriptors:\n";
2964 Pos = Contents.data() + IndexEntries[0].LSDAStart;
2965 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
2966 (2 * sizeof(uint32_t));
2967 for (int i = 0; i < NumLSDAs; ++i) {
2968 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
2969 uint32_t LSDAOffset = readNext<uint32_t>(Pos);
2970 outs() << " [" << i << "]: "
2971 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
2973 << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n';
2976 //===----------------------------------
2977 // Finally, the 2nd level indices
2978 //===----------------------------------
2980 // Generally these are 4K in size, and have 2 possible forms:
2981 // + Regular stores up to 511 entries with disparate encodings
2982 // + Compressed stores up to 1021 entries if few enough compact encoding
2984 outs() << " Second level indices:\n";
2985 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
2986 // The final sentinel top-level index has no associated 2nd level page
2987 if (IndexEntries[i].SecondLevelPageStart == 0)
2990 outs() << " Second level index[" << i << "]: "
2991 << "offset in section="
2992 << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart)
2994 << "base function offset="
2995 << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n';
2997 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
2998 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
3000 printRegularSecondLevelUnwindPage(Pos);
3002 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
3005 llvm_unreachable("Do not know how to print this kind of 2nd level page");
3009 void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
3010 std::map<uint64_t, SymbolRef> Symbols;
3011 for (const SymbolRef &SymRef : Obj->symbols()) {
3012 // Discard any undefined or absolute symbols. They're not going to take part
3013 // in the convenience lookup for unwind info and just take up resources.
3014 section_iterator Section = Obj->section_end();
3015 SymRef.getSection(Section);
3016 if (Section == Obj->section_end())
3020 SymRef.getAddress(Addr);
3021 Symbols.insert(std::make_pair(Addr, SymRef));
3024 for (const SectionRef &Section : Obj->sections()) {
3026 Section.getName(SectName);
3027 if (SectName == "__compact_unwind")
3028 printMachOCompactUnwindSection(Obj, Symbols, Section);
3029 else if (SectName == "__unwind_info")
3030 printMachOUnwindInfoSection(Obj, Symbols, Section);
3031 else if (SectName == "__eh_frame")
3032 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
3036 static void PrintMachHeader(uint32_t magic, uint32_t cputype,
3037 uint32_t cpusubtype, uint32_t filetype,
3038 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
3040 outs() << "Mach header\n";
3041 outs() << " magic cputype cpusubtype caps filetype ncmds "
3042 "sizeofcmds flags\n";
3044 if (magic == MachO::MH_MAGIC)
3045 outs() << " MH_MAGIC";
3046 else if (magic == MachO::MH_MAGIC_64)
3047 outs() << "MH_MAGIC_64";
3049 outs() << format(" 0x%08" PRIx32, magic);
3051 case MachO::CPU_TYPE_I386:
3053 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3054 case MachO::CPU_SUBTYPE_I386_ALL:
3058 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3062 case MachO::CPU_TYPE_X86_64:
3063 outs() << " X86_64";
3064 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3065 case MachO::CPU_SUBTYPE_X86_64_ALL:
3068 case MachO::CPU_SUBTYPE_X86_64_H:
3069 outs() << " Haswell";
3072 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3076 case MachO::CPU_TYPE_ARM:
3078 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3079 case MachO::CPU_SUBTYPE_ARM_ALL:
3082 case MachO::CPU_SUBTYPE_ARM_V4T:
3085 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
3088 case MachO::CPU_SUBTYPE_ARM_XSCALE:
3089 outs() << " XSCALE";
3091 case MachO::CPU_SUBTYPE_ARM_V6:
3094 case MachO::CPU_SUBTYPE_ARM_V6M:
3097 case MachO::CPU_SUBTYPE_ARM_V7:
3100 case MachO::CPU_SUBTYPE_ARM_V7EM:
3103 case MachO::CPU_SUBTYPE_ARM_V7K:
3106 case MachO::CPU_SUBTYPE_ARM_V7M:
3109 case MachO::CPU_SUBTYPE_ARM_V7S:
3113 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3117 case MachO::CPU_TYPE_ARM64:
3119 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3120 case MachO::CPU_SUBTYPE_ARM64_ALL:
3124 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3128 case MachO::CPU_TYPE_POWERPC:
3130 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3131 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3135 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3139 case MachO::CPU_TYPE_POWERPC64:
3141 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3142 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3146 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3151 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
3154 outs() << format(" 0x%02" PRIx32,
3155 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3158 case MachO::MH_OBJECT:
3159 outs() << " OBJECT";
3161 case MachO::MH_EXECUTE:
3162 outs() << " EXECUTE";
3164 case MachO::MH_FVMLIB:
3165 outs() << " FVMLIB";
3167 case MachO::MH_CORE:
3170 case MachO::MH_PRELOAD:
3171 outs() << " PRELOAD";
3173 case MachO::MH_DYLIB:
3176 case MachO::MH_DYLIB_STUB:
3177 outs() << " DYLIB_STUB";
3179 case MachO::MH_DYLINKER:
3180 outs() << " DYLINKER";
3182 case MachO::MH_BUNDLE:
3183 outs() << " BUNDLE";
3185 case MachO::MH_DSYM:
3188 case MachO::MH_KEXT_BUNDLE:
3189 outs() << " KEXTBUNDLE";
3192 outs() << format(" %10u", filetype);
3195 outs() << format(" %5u", ncmds);
3196 outs() << format(" %10u", sizeofcmds);
3198 if (f & MachO::MH_NOUNDEFS) {
3199 outs() << " NOUNDEFS";
3200 f &= ~MachO::MH_NOUNDEFS;
3202 if (f & MachO::MH_INCRLINK) {
3203 outs() << " INCRLINK";
3204 f &= ~MachO::MH_INCRLINK;
3206 if (f & MachO::MH_DYLDLINK) {
3207 outs() << " DYLDLINK";
3208 f &= ~MachO::MH_DYLDLINK;
3210 if (f & MachO::MH_BINDATLOAD) {
3211 outs() << " BINDATLOAD";
3212 f &= ~MachO::MH_BINDATLOAD;
3214 if (f & MachO::MH_PREBOUND) {
3215 outs() << " PREBOUND";
3216 f &= ~MachO::MH_PREBOUND;
3218 if (f & MachO::MH_SPLIT_SEGS) {
3219 outs() << " SPLIT_SEGS";
3220 f &= ~MachO::MH_SPLIT_SEGS;
3222 if (f & MachO::MH_LAZY_INIT) {
3223 outs() << " LAZY_INIT";
3224 f &= ~MachO::MH_LAZY_INIT;
3226 if (f & MachO::MH_TWOLEVEL) {
3227 outs() << " TWOLEVEL";
3228 f &= ~MachO::MH_TWOLEVEL;
3230 if (f & MachO::MH_FORCE_FLAT) {
3231 outs() << " FORCE_FLAT";
3232 f &= ~MachO::MH_FORCE_FLAT;
3234 if (f & MachO::MH_NOMULTIDEFS) {
3235 outs() << " NOMULTIDEFS";
3236 f &= ~MachO::MH_NOMULTIDEFS;
3238 if (f & MachO::MH_NOFIXPREBINDING) {
3239 outs() << " NOFIXPREBINDING";
3240 f &= ~MachO::MH_NOFIXPREBINDING;
3242 if (f & MachO::MH_PREBINDABLE) {
3243 outs() << " PREBINDABLE";
3244 f &= ~MachO::MH_PREBINDABLE;
3246 if (f & MachO::MH_ALLMODSBOUND) {
3247 outs() << " ALLMODSBOUND";
3248 f &= ~MachO::MH_ALLMODSBOUND;
3250 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
3251 outs() << " SUBSECTIONS_VIA_SYMBOLS";
3252 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
3254 if (f & MachO::MH_CANONICAL) {
3255 outs() << " CANONICAL";
3256 f &= ~MachO::MH_CANONICAL;
3258 if (f & MachO::MH_WEAK_DEFINES) {
3259 outs() << " WEAK_DEFINES";
3260 f &= ~MachO::MH_WEAK_DEFINES;
3262 if (f & MachO::MH_BINDS_TO_WEAK) {
3263 outs() << " BINDS_TO_WEAK";
3264 f &= ~MachO::MH_BINDS_TO_WEAK;
3266 if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
3267 outs() << " ALLOW_STACK_EXECUTION";
3268 f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
3270 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
3271 outs() << " DEAD_STRIPPABLE_DYLIB";
3272 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
3274 if (f & MachO::MH_PIE) {
3276 f &= ~MachO::MH_PIE;
3278 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
3279 outs() << " NO_REEXPORTED_DYLIBS";
3280 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
3282 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
3283 outs() << " MH_HAS_TLV_DESCRIPTORS";
3284 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
3286 if (f & MachO::MH_NO_HEAP_EXECUTION) {
3287 outs() << " MH_NO_HEAP_EXECUTION";
3288 f &= ~MachO::MH_NO_HEAP_EXECUTION;
3290 if (f & MachO::MH_APP_EXTENSION_SAFE) {
3291 outs() << " APP_EXTENSION_SAFE";
3292 f &= ~MachO::MH_APP_EXTENSION_SAFE;
3294 if (f != 0 || flags == 0)
3295 outs() << format(" 0x%08" PRIx32, f);
3297 outs() << format(" 0x%08" PRIx32, magic);
3298 outs() << format(" %7d", cputype);
3299 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3300 outs() << format(" 0x%02" PRIx32,
3301 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3302 outs() << format(" %10u", filetype);
3303 outs() << format(" %5u", ncmds);
3304 outs() << format(" %10u", sizeofcmds);
3305 outs() << format(" 0x%08" PRIx32, flags);
3310 static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
3311 StringRef SegName, uint64_t vmaddr,
3312 uint64_t vmsize, uint64_t fileoff,
3313 uint64_t filesize, uint32_t maxprot,
3314 uint32_t initprot, uint32_t nsects,
3315 uint32_t flags, uint32_t object_size,
3317 uint64_t expected_cmdsize;
3318 if (cmd == MachO::LC_SEGMENT) {
3319 outs() << " cmd LC_SEGMENT\n";
3320 expected_cmdsize = nsects;
3321 expected_cmdsize *= sizeof(struct MachO::section);
3322 expected_cmdsize += sizeof(struct MachO::segment_command);
3324 outs() << " cmd LC_SEGMENT_64\n";
3325 expected_cmdsize = nsects;
3326 expected_cmdsize *= sizeof(struct MachO::section_64);
3327 expected_cmdsize += sizeof(struct MachO::segment_command_64);
3329 outs() << " cmdsize " << cmdsize;
3330 if (cmdsize != expected_cmdsize)
3331 outs() << " Inconsistent size\n";
3334 outs() << " segname " << SegName << "\n";
3335 if (cmd == MachO::LC_SEGMENT_64) {
3336 outs() << " vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n";
3337 outs() << " vmsize " << format("0x%016" PRIx64, vmsize) << "\n";
3339 outs() << " vmaddr " << format("0x%08" PRIx64, vmaddr) << "\n";
3340 outs() << " vmsize " << format("0x%08" PRIx64, vmsize) << "\n";
3342 outs() << " fileoff " << fileoff;
3343 if (fileoff > object_size)
3344 outs() << " (past end of file)\n";
3347 outs() << " filesize " << filesize;
3348 if (fileoff + filesize > object_size)
3349 outs() << " (past end of file)\n";
3354 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3355 MachO::VM_PROT_EXECUTE)) != 0)
3356 outs() << " maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n";
3358 if (maxprot & MachO::VM_PROT_READ)
3359 outs() << " maxprot r";
3361 outs() << " maxprot -";
3362 if (maxprot & MachO::VM_PROT_WRITE)
3366 if (maxprot & MachO::VM_PROT_EXECUTE)
3372 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3373 MachO::VM_PROT_EXECUTE)) != 0)
3374 outs() << " initprot ?" << format("0x%08" PRIx32, initprot) << "\n";
3376 if (initprot & MachO::VM_PROT_READ)
3377 outs() << " initprot r";
3379 outs() << " initprot -";
3380 if (initprot & MachO::VM_PROT_WRITE)
3384 if (initprot & MachO::VM_PROT_EXECUTE)
3390 outs() << " maxprot " << format("0x%08" PRIx32, maxprot) << "\n";
3391 outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n";
3393 outs() << " nsects " << nsects << "\n";
3397 outs() << " (none)\n";
3399 if (flags & MachO::SG_HIGHVM) {
3400 outs() << " HIGHVM";
3401 flags &= ~MachO::SG_HIGHVM;
3403 if (flags & MachO::SG_FVMLIB) {
3404 outs() << " FVMLIB";
3405 flags &= ~MachO::SG_FVMLIB;
3407 if (flags & MachO::SG_NORELOC) {
3408 outs() << " NORELOC";
3409 flags &= ~MachO::SG_NORELOC;
3411 if (flags & MachO::SG_PROTECTED_VERSION_1) {
3412 outs() << " PROTECTED_VERSION_1";
3413 flags &= ~MachO::SG_PROTECTED_VERSION_1;
3416 outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n";
3421 outs() << " flags " << format("0x%" PRIx32, flags) << "\n";
3425 static void PrintSection(const char *sectname, const char *segname,
3426 uint64_t addr, uint64_t size, uint32_t offset,
3427 uint32_t align, uint32_t reloff, uint32_t nreloc,
3428 uint32_t flags, uint32_t reserved1, uint32_t reserved2,
3429 uint32_t cmd, const char *sg_segname,
3430 uint32_t filetype, uint32_t object_size,
3432 outs() << "Section\n";
3433 outs() << " sectname " << format("%.16s\n", sectname);
3434 outs() << " segname " << format("%.16s", segname);
3435 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
3436 outs() << " (does not match segment)\n";
3439 if (cmd == MachO::LC_SEGMENT_64) {
3440 outs() << " addr " << format("0x%016" PRIx64, addr) << "\n";
3441 outs() << " size " << format("0x%016" PRIx64, size);
3443 outs() << " addr " << format("0x%08" PRIx64, addr) << "\n";
3444 outs() << " size " << format("0x%08" PRIx64, size);
3446 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
3447 outs() << " (past end of file)\n";
3450 outs() << " offset " << offset;
3451 if (offset > object_size)
3452 outs() << " (past end of file)\n";
3455 uint32_t align_shifted = 1 << align;
3456 outs() << " align 2^" << align << " (" << align_shifted << ")\n";
3457 outs() << " reloff " << reloff;
3458 if (reloff > object_size)
3459 outs() << " (past end of file)\n";
3462 outs() << " nreloc " << nreloc;
3463 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
3464 outs() << " (past end of file)\n";
3467 uint32_t section_type = flags & MachO::SECTION_TYPE;
3470 if (section_type == MachO::S_REGULAR)
3471 outs() << " S_REGULAR\n";
3472 else if (section_type == MachO::S_ZEROFILL)
3473 outs() << " S_ZEROFILL\n";
3474 else if (section_type == MachO::S_CSTRING_LITERALS)
3475 outs() << " S_CSTRING_LITERALS\n";
3476 else if (section_type == MachO::S_4BYTE_LITERALS)
3477 outs() << " S_4BYTE_LITERALS\n";
3478 else if (section_type == MachO::S_8BYTE_LITERALS)
3479 outs() << " S_8BYTE_LITERALS\n";
3480 else if (section_type == MachO::S_16BYTE_LITERALS)
3481 outs() << " S_16BYTE_LITERALS\n";
3482 else if (section_type == MachO::S_LITERAL_POINTERS)
3483 outs() << " S_LITERAL_POINTERS\n";
3484 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
3485 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
3486 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
3487 outs() << " S_LAZY_SYMBOL_POINTERS\n";
3488 else if (section_type == MachO::S_SYMBOL_STUBS)
3489 outs() << " S_SYMBOL_STUBS\n";
3490 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
3491 outs() << " S_MOD_INIT_FUNC_POINTERS\n";
3492 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
3493 outs() << " S_MOD_TERM_FUNC_POINTERS\n";
3494 else if (section_type == MachO::S_COALESCED)
3495 outs() << " S_COALESCED\n";
3496 else if (section_type == MachO::S_INTERPOSING)
3497 outs() << " S_INTERPOSING\n";
3498 else if (section_type == MachO::S_DTRACE_DOF)
3499 outs() << " S_DTRACE_DOF\n";
3500 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
3501 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
3502 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
3503 outs() << " S_THREAD_LOCAL_REGULAR\n";
3504 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
3505 outs() << " S_THREAD_LOCAL_ZEROFILL\n";
3506 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
3507 outs() << " S_THREAD_LOCAL_VARIABLES\n";
3508 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3509 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
3510 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
3511 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
3513 outs() << format("0x%08" PRIx32, section_type) << "\n";
3514 outs() << "attributes";
3515 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
3516 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
3517 outs() << " PURE_INSTRUCTIONS";
3518 if (section_attributes & MachO::S_ATTR_NO_TOC)
3519 outs() << " NO_TOC";
3520 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
3521 outs() << " STRIP_STATIC_SYMS";
3522 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
3523 outs() << " NO_DEAD_STRIP";
3524 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
3525 outs() << " LIVE_SUPPORT";
3526 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
3527 outs() << " SELF_MODIFYING_CODE";
3528 if (section_attributes & MachO::S_ATTR_DEBUG)
3530 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
3531 outs() << " SOME_INSTRUCTIONS";
3532 if (section_attributes & MachO::S_ATTR_EXT_RELOC)
3533 outs() << " EXT_RELOC";
3534 if (section_attributes & MachO::S_ATTR_LOC_RELOC)
3535 outs() << " LOC_RELOC";
3536 if (section_attributes == 0)
3537 outs() << " (none)";
3540 outs() << " flags " << format("0x%08" PRIx32, flags) << "\n";
3541 outs() << " reserved1 " << reserved1;
3542 if (section_type == MachO::S_SYMBOL_STUBS ||
3543 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
3544 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
3545 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
3546 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3547 outs() << " (index into indirect symbol table)\n";
3550 outs() << " reserved2 " << reserved2;
3551 if (section_type == MachO::S_SYMBOL_STUBS)
3552 outs() << " (size of stubs)\n";
3557 static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
3558 uint32_t object_size) {
3559 outs() << " cmd LC_SYMTAB\n";
3560 outs() << " cmdsize " << st.cmdsize;
3561 if (st.cmdsize != sizeof(struct MachO::symtab_command))
3562 outs() << " Incorrect size\n";
3565 outs() << " symoff " << st.symoff;
3566 if (st.symoff > object_size)
3567 outs() << " (past end of file)\n";
3570 outs() << " nsyms " << st.nsyms;
3573 big_size = st.nsyms;
3574 big_size *= sizeof(struct MachO::nlist_64);
3575 big_size += st.symoff;
3576 if (big_size > object_size)
3577 outs() << " (past end of file)\n";
3581 big_size = st.nsyms;
3582 big_size *= sizeof(struct MachO::nlist);
3583 big_size += st.symoff;
3584 if (big_size > object_size)
3585 outs() << " (past end of file)\n";
3589 outs() << " stroff " << st.stroff;
3590 if (st.stroff > object_size)
3591 outs() << " (past end of file)\n";
3594 outs() << " strsize " << st.strsize;
3595 big_size = st.stroff;
3596 big_size += st.strsize;
3597 if (big_size > object_size)
3598 outs() << " (past end of file)\n";
3603 static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
3604 uint32_t nsyms, uint32_t object_size,
3606 outs() << " cmd LC_DYSYMTAB\n";
3607 outs() << " cmdsize " << dyst.cmdsize;
3608 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
3609 outs() << " Incorrect size\n";
3612 outs() << " ilocalsym " << dyst.ilocalsym;
3613 if (dyst.ilocalsym > nsyms)
3614 outs() << " (greater than the number of symbols)\n";
3617 outs() << " nlocalsym " << dyst.nlocalsym;
3619 big_size = dyst.ilocalsym;
3620 big_size += dyst.nlocalsym;
3621 if (big_size > nsyms)
3622 outs() << " (past the end of the symbol table)\n";
3625 outs() << " iextdefsym " << dyst.iextdefsym;
3626 if (dyst.iextdefsym > nsyms)
3627 outs() << " (greater than the number of symbols)\n";
3630 outs() << " nextdefsym " << dyst.nextdefsym;
3631 big_size = dyst.iextdefsym;
3632 big_size += dyst.nextdefsym;
3633 if (big_size > nsyms)
3634 outs() << " (past the end of the symbol table)\n";
3637 outs() << " iundefsym " << dyst.iundefsym;
3638 if (dyst.iundefsym > nsyms)
3639 outs() << " (greater than the number of symbols)\n";
3642 outs() << " nundefsym " << dyst.nundefsym;
3643 big_size = dyst.iundefsym;
3644 big_size += dyst.nundefsym;
3645 if (big_size > nsyms)
3646 outs() << " (past the end of the symbol table)\n";
3649 outs() << " tocoff " << dyst.tocoff;
3650 if (dyst.tocoff > object_size)
3651 outs() << " (past end of file)\n";
3654 outs() << " ntoc " << dyst.ntoc;
3655 big_size = dyst.ntoc;
3656 big_size *= sizeof(struct MachO::dylib_table_of_contents);
3657 big_size += dyst.tocoff;
3658 if (big_size > object_size)
3659 outs() << " (past end of file)\n";
3662 outs() << " modtaboff " << dyst.modtaboff;
3663 if (dyst.modtaboff > object_size)
3664 outs() << " (past end of file)\n";
3667 outs() << " nmodtab " << dyst.nmodtab;
3670 modtabend = dyst.nmodtab;
3671 modtabend *= sizeof(struct MachO::dylib_module_64);
3672 modtabend += dyst.modtaboff;
3674 modtabend = dyst.nmodtab;
3675 modtabend *= sizeof(struct MachO::dylib_module);
3676 modtabend += dyst.modtaboff;
3678 if (modtabend > object_size)
3679 outs() << " (past end of file)\n";
3682 outs() << " extrefsymoff " << dyst.extrefsymoff;
3683 if (dyst.extrefsymoff > object_size)
3684 outs() << " (past end of file)\n";
3687 outs() << " nextrefsyms " << dyst.nextrefsyms;
3688 big_size = dyst.nextrefsyms;
3689 big_size *= sizeof(struct MachO::dylib_reference);
3690 big_size += dyst.extrefsymoff;
3691 if (big_size > object_size)
3692 outs() << " (past end of file)\n";
3695 outs() << " indirectsymoff " << dyst.indirectsymoff;
3696 if (dyst.indirectsymoff > object_size)
3697 outs() << " (past end of file)\n";
3700 outs() << " nindirectsyms " << dyst.nindirectsyms;
3701 big_size = dyst.nindirectsyms;
3702 big_size *= sizeof(uint32_t);
3703 big_size += dyst.indirectsymoff;
3704 if (big_size > object_size)
3705 outs() << " (past end of file)\n";
3708 outs() << " extreloff " << dyst.extreloff;
3709 if (dyst.extreloff > object_size)
3710 outs() << " (past end of file)\n";
3713 outs() << " nextrel " << dyst.nextrel;
3714 big_size = dyst.nextrel;
3715 big_size *= sizeof(struct MachO::relocation_info);
3716 big_size += dyst.extreloff;
3717 if (big_size > object_size)
3718 outs() << " (past end of file)\n";
3721 outs() << " locreloff " << dyst.locreloff;
3722 if (dyst.locreloff > object_size)
3723 outs() << " (past end of file)\n";
3726 outs() << " nlocrel " << dyst.nlocrel;
3727 big_size = dyst.nlocrel;
3728 big_size *= sizeof(struct MachO::relocation_info);
3729 big_size += dyst.locreloff;
3730 if (big_size > object_size)
3731 outs() << " (past end of file)\n";
3736 static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
3737 uint32_t object_size) {
3738 if (dc.cmd == MachO::LC_DYLD_INFO)
3739 outs() << " cmd LC_DYLD_INFO\n";
3741 outs() << " cmd LC_DYLD_INFO_ONLY\n";
3742 outs() << " cmdsize " << dc.cmdsize;
3743 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
3744 outs() << " Incorrect size\n";
3747 outs() << " rebase_off " << dc.rebase_off;
3748 if (dc.rebase_off > object_size)
3749 outs() << " (past end of file)\n";
3752 outs() << " rebase_size " << dc.rebase_size;
3754 big_size = dc.rebase_off;
3755 big_size += dc.rebase_size;
3756 if (big_size > object_size)
3757 outs() << " (past end of file)\n";
3760 outs() << " bind_off " << dc.bind_off;
3761 if (dc.bind_off > object_size)
3762 outs() << " (past end of file)\n";
3765 outs() << " bind_size " << dc.bind_size;
3766 big_size = dc.bind_off;
3767 big_size += dc.bind_size;
3768 if (big_size > object_size)
3769 outs() << " (past end of file)\n";
3772 outs() << " weak_bind_off " << dc.weak_bind_off;
3773 if (dc.weak_bind_off > object_size)
3774 outs() << " (past end of file)\n";
3777 outs() << " weak_bind_size " << dc.weak_bind_size;
3778 big_size = dc.weak_bind_off;
3779 big_size += dc.weak_bind_size;
3780 if (big_size > object_size)
3781 outs() << " (past end of file)\n";
3784 outs() << " lazy_bind_off " << dc.lazy_bind_off;
3785 if (dc.lazy_bind_off > object_size)
3786 outs() << " (past end of file)\n";
3789 outs() << " lazy_bind_size " << dc.lazy_bind_size;
3790 big_size = dc.lazy_bind_off;
3791 big_size += dc.lazy_bind_size;
3792 if (big_size > object_size)
3793 outs() << " (past end of file)\n";
3796 outs() << " export_off " << dc.export_off;
3797 if (dc.export_off > object_size)
3798 outs() << " (past end of file)\n";
3801 outs() << " export_size " << dc.export_size;
3802 big_size = dc.export_off;
3803 big_size += dc.export_size;
3804 if (big_size > object_size)
3805 outs() << " (past end of file)\n";
3810 static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
3812 if (dyld.cmd == MachO::LC_ID_DYLINKER)
3813 outs() << " cmd LC_ID_DYLINKER\n";
3814 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
3815 outs() << " cmd LC_LOAD_DYLINKER\n";
3816 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
3817 outs() << " cmd LC_DYLD_ENVIRONMENT\n";
3819 outs() << " cmd ?(" << dyld.cmd << ")\n";
3820 outs() << " cmdsize " << dyld.cmdsize;
3821 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
3822 outs() << " Incorrect size\n";
3825 if (dyld.name >= dyld.cmdsize)
3826 outs() << " name ?(bad offset " << dyld.name << ")\n";
3828 const char *P = (const char *)(Ptr) + dyld.name;
3829 outs() << " name " << P << " (offset " << dyld.name << ")\n";
3833 static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
3834 outs() << " cmd LC_UUID\n";
3835 outs() << " cmdsize " << uuid.cmdsize;
3836 if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
3837 outs() << " Incorrect size\n";
3841 outs() << format("%02" PRIX32, uuid.uuid[0]);
3842 outs() << format("%02" PRIX32, uuid.uuid[1]);
3843 outs() << format("%02" PRIX32, uuid.uuid[2]);
3844 outs() << format("%02" PRIX32, uuid.uuid[3]);
3846 outs() << format("%02" PRIX32, uuid.uuid[4]);
3847 outs() << format("%02" PRIX32, uuid.uuid[5]);
3849 outs() << format("%02" PRIX32, uuid.uuid[6]);
3850 outs() << format("%02" PRIX32, uuid.uuid[7]);
3852 outs() << format("%02" PRIX32, uuid.uuid[8]);
3853 outs() << format("%02" PRIX32, uuid.uuid[9]);
3855 outs() << format("%02" PRIX32, uuid.uuid[10]);
3856 outs() << format("%02" PRIX32, uuid.uuid[11]);
3857 outs() << format("%02" PRIX32, uuid.uuid[12]);
3858 outs() << format("%02" PRIX32, uuid.uuid[13]);
3859 outs() << format("%02" PRIX32, uuid.uuid[14]);
3860 outs() << format("%02" PRIX32, uuid.uuid[15]);
3864 static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) {
3865 outs() << " cmd LC_RPATH\n";
3866 outs() << " cmdsize " << rpath.cmdsize;
3867 if (rpath.cmdsize < sizeof(struct MachO::rpath_command))
3868 outs() << " Incorrect size\n";
3871 if (rpath.path >= rpath.cmdsize)
3872 outs() << " path ?(bad offset " << rpath.path << ")\n";
3874 const char *P = (const char *)(Ptr) + rpath.path;
3875 outs() << " path " << P << " (offset " << rpath.path << ")\n";
3879 static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
3880 if (vd.cmd == MachO::LC_VERSION_MIN_MACOSX)
3881 outs() << " cmd LC_VERSION_MIN_MACOSX\n";
3882 else if (vd.cmd == MachO::LC_VERSION_MIN_IPHONEOS)
3883 outs() << " cmd LC_VERSION_MIN_IPHONEOS\n";
3885 outs() << " cmd " << vd.cmd << " (?)\n";
3886 outs() << " cmdsize " << vd.cmdsize;
3887 if (vd.cmdsize != sizeof(struct MachO::version_min_command))
3888 outs() << " Incorrect size\n";
3891 outs() << " version " << ((vd.version >> 16) & 0xffff) << "."
3892 << ((vd.version >> 8) & 0xff);
3893 if ((vd.version & 0xff) != 0)
3894 outs() << "." << (vd.version & 0xff);
3897 outs() << " sdk n/a";
3899 outs() << " sdk " << ((vd.sdk >> 16) & 0xffff) << "."
3900 << ((vd.sdk >> 8) & 0xff);
3902 if ((vd.sdk & 0xff) != 0)
3903 outs() << "." << (vd.sdk & 0xff);
3907 static void PrintSourceVersionCommand(MachO::source_version_command sd) {
3908 outs() << " cmd LC_SOURCE_VERSION\n";
3909 outs() << " cmdsize " << sd.cmdsize;
3910 if (sd.cmdsize != sizeof(struct MachO::source_version_command))
3911 outs() << " Incorrect size\n";
3914 uint64_t a = (sd.version >> 40) & 0xffffff;
3915 uint64_t b = (sd.version >> 30) & 0x3ff;
3916 uint64_t c = (sd.version >> 20) & 0x3ff;
3917 uint64_t d = (sd.version >> 10) & 0x3ff;
3918 uint64_t e = sd.version & 0x3ff;
3919 outs() << " version " << a << "." << b;
3921 outs() << "." << c << "." << d << "." << e;
3923 outs() << "." << c << "." << d;
3929 static void PrintEntryPointCommand(MachO::entry_point_command ep) {
3930 outs() << " cmd LC_MAIN\n";
3931 outs() << " cmdsize " << ep.cmdsize;
3932 if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
3933 outs() << " Incorrect size\n";
3936 outs() << " entryoff " << ep.entryoff << "\n";
3937 outs() << " stacksize " << ep.stacksize << "\n";
3940 static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec,
3941 uint32_t object_size) {
3942 outs() << " cmd LC_ENCRYPTION_INFO\n";
3943 outs() << " cmdsize " << ec.cmdsize;
3944 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command))
3945 outs() << " Incorrect size\n";
3948 outs() << " cryptoff " << ec.cryptoff;
3949 if (ec.cryptoff > object_size)
3950 outs() << " (past end of file)\n";
3953 outs() << " cryptsize " << ec.cryptsize;
3954 if (ec.cryptsize > object_size)
3955 outs() << " (past end of file)\n";
3958 outs() << " cryptid " << ec.cryptid << "\n";
3961 static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,
3962 uint32_t object_size) {
3963 outs() << " cmd LC_ENCRYPTION_INFO_64\n";
3964 outs() << " cmdsize " << ec.cmdsize;
3965 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64))
3966 outs() << " Incorrect size\n";
3969 outs() << " cryptoff " << ec.cryptoff;
3970 if (ec.cryptoff > object_size)
3971 outs() << " (past end of file)\n";
3974 outs() << " cryptsize " << ec.cryptsize;
3975 if (ec.cryptsize > object_size)
3976 outs() << " (past end of file)\n";
3979 outs() << " cryptid " << ec.cryptid << "\n";
3980 outs() << " pad " << ec.pad << "\n";
3983 static void PrintLinkerOptionCommand(MachO::linker_option_command lo,
3985 outs() << " cmd LC_LINKER_OPTION\n";
3986 outs() << " cmdsize " << lo.cmdsize;
3987 if (lo.cmdsize < sizeof(struct MachO::linker_option_command))
3988 outs() << " Incorrect size\n";
3991 outs() << " count " << lo.count << "\n";
3992 const char *string = Ptr + sizeof(struct MachO::linker_option_command);
3993 uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command);
3996 while (*string == '\0' && left > 0) {
4002 outs() << " string #" << i << " " << format("%.*s\n", left, string);
4003 uint32_t NullPos = StringRef(string, left).find('\0');
4004 uint32_t len = std::min(NullPos, left) + 1;
4010 outs() << " count " << lo.count << " does not match number of strings "
4014 static void PrintSubFrameworkCommand(MachO::sub_framework_command sub,
4016 outs() << " cmd LC_SUB_FRAMEWORK\n";
4017 outs() << " cmdsize " << sub.cmdsize;
4018 if (sub.cmdsize < sizeof(struct MachO::sub_framework_command))
4019 outs() << " Incorrect size\n";
4022 if (sub.umbrella < sub.cmdsize) {
4023 const char *P = Ptr + sub.umbrella;
4024 outs() << " umbrella " << P << " (offset " << sub.umbrella << ")\n";
4026 outs() << " umbrella ?(bad offset " << sub.umbrella << ")\n";
4030 static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,
4032 outs() << " cmd LC_SUB_UMBRELLA\n";
4033 outs() << " cmdsize " << sub.cmdsize;
4034 if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command))
4035 outs() << " Incorrect size\n";
4038 if (sub.sub_umbrella < sub.cmdsize) {
4039 const char *P = Ptr + sub.sub_umbrella;
4040 outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n";
4042 outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n";
4046 static void PrintSubLibraryCommand(MachO::sub_library_command sub,
4048 outs() << " cmd LC_SUB_LIBRARY\n";
4049 outs() << " cmdsize " << sub.cmdsize;
4050 if (sub.cmdsize < sizeof(struct MachO::sub_library_command))
4051 outs() << " Incorrect size\n";
4054 if (sub.sub_library < sub.cmdsize) {
4055 const char *P = Ptr + sub.sub_library;
4056 outs() << " sub_library " << P << " (offset " << sub.sub_library << ")\n";
4058 outs() << " sub_library ?(bad offset " << sub.sub_library << ")\n";
4062 static void PrintSubClientCommand(MachO::sub_client_command sub,
4064 outs() << " cmd LC_SUB_CLIENT\n";
4065 outs() << " cmdsize " << sub.cmdsize;
4066 if (sub.cmdsize < sizeof(struct MachO::sub_client_command))
4067 outs() << " Incorrect size\n";
4070 if (sub.client < sub.cmdsize) {
4071 const char *P = Ptr + sub.client;
4072 outs() << " client " << P << " (offset " << sub.client << ")\n";
4074 outs() << " client ?(bad offset " << sub.client << ")\n";
4078 static void PrintRoutinesCommand(MachO::routines_command r) {
4079 outs() << " cmd LC_ROUTINES\n";
4080 outs() << " cmdsize " << r.cmdsize;
4081 if (r.cmdsize != sizeof(struct MachO::routines_command))
4082 outs() << " Incorrect size\n";
4085 outs() << " init_address " << format("0x%08" PRIx32, r.init_address) << "\n";
4086 outs() << " init_module " << r.init_module << "\n";
4087 outs() << " reserved1 " << r.reserved1 << "\n";
4088 outs() << " reserved2 " << r.reserved2 << "\n";
4089 outs() << " reserved3 " << r.reserved3 << "\n";
4090 outs() << " reserved4 " << r.reserved4 << "\n";
4091 outs() << " reserved5 " << r.reserved5 << "\n";
4092 outs() << " reserved6 " << r.reserved6 << "\n";
4095 static void PrintRoutinesCommand64(MachO::routines_command_64 r) {
4096 outs() << " cmd LC_ROUTINES_64\n";
4097 outs() << " cmdsize " << r.cmdsize;
4098 if (r.cmdsize != sizeof(struct MachO::routines_command_64))
4099 outs() << " Incorrect size\n";
4102 outs() << " init_address " << format("0x%016" PRIx64, r.init_address) << "\n";
4103 outs() << " init_module " << r.init_module << "\n";
4104 outs() << " reserved1 " << r.reserved1 << "\n";
4105 outs() << " reserved2 " << r.reserved2 << "\n";
4106 outs() << " reserved3 " << r.reserved3 << "\n";
4107 outs() << " reserved4 " << r.reserved4 << "\n";
4108 outs() << " reserved5 " << r.reserved5 << "\n";
4109 outs() << " reserved6 " << r.reserved6 << "\n";
4112 static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) {
4113 outs() << " rax " << format("0x%016" PRIx64, cpu64.rax);
4114 outs() << " rbx " << format("0x%016" PRIx64, cpu64.rbx);
4115 outs() << " rcx " << format("0x%016" PRIx64, cpu64.rcx) << "\n";
4116 outs() << " rdx " << format("0x%016" PRIx64, cpu64.rdx);
4117 outs() << " rdi " << format("0x%016" PRIx64, cpu64.rdi);
4118 outs() << " rsi " << format("0x%016" PRIx64, cpu64.rsi) << "\n";
4119 outs() << " rbp " << format("0x%016" PRIx64, cpu64.rbp);
4120 outs() << " rsp " << format("0x%016" PRIx64, cpu64.rsp);
4121 outs() << " r8 " << format("0x%016" PRIx64, cpu64.r8) << "\n";
4122 outs() << " r9 " << format("0x%016" PRIx64, cpu64.r9);
4123 outs() << " r10 " << format("0x%016" PRIx64, cpu64.r10);
4124 outs() << " r11 " << format("0x%016" PRIx64, cpu64.r11) << "\n";
4125 outs() << " r12 " << format("0x%016" PRIx64, cpu64.r12);
4126 outs() << " r13 " << format("0x%016" PRIx64, cpu64.r13);
4127 outs() << " r14 " << format("0x%016" PRIx64, cpu64.r14) << "\n";
4128 outs() << " r15 " << format("0x%016" PRIx64, cpu64.r15);
4129 outs() << " rip " << format("0x%016" PRIx64, cpu64.rip) << "\n";
4130 outs() << "rflags " << format("0x%016" PRIx64, cpu64.rflags);
4131 outs() << " cs " << format("0x%016" PRIx64, cpu64.cs);
4132 outs() << " fs " << format("0x%016" PRIx64, cpu64.fs) << "\n";
4133 outs() << " gs " << format("0x%016" PRIx64, cpu64.gs) << "\n";
4136 static void Print_mmst_reg(MachO::mmst_reg_t &r) {
4138 outs() << "\t mmst_reg ";
4139 for (f = 0; f < 10; f++)
4140 outs() << format("%02" PRIx32, (r.mmst_reg[f] & 0xff)) << " ";
4142 outs() << "\t mmst_rsrv ";
4143 for (f = 0; f < 6; f++)
4144 outs() << format("%02" PRIx32, (r.mmst_rsrv[f] & 0xff)) << " ";
4148 static void Print_xmm_reg(MachO::xmm_reg_t &r) {
4150 outs() << "\t xmm_reg ";
4151 for (f = 0; f < 16; f++)
4152 outs() << format("%02" PRIx32, (r.xmm_reg[f] & 0xff)) << " ";
4156 static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) {
4157 outs() << "\t fpu_reserved[0] " << fpu.fpu_reserved[0];
4158 outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n";
4159 outs() << "\t control: invalid " << fpu.fpu_fcw.invalid;
4160 outs() << " denorm " << fpu.fpu_fcw.denorm;
4161 outs() << " zdiv " << fpu.fpu_fcw.zdiv;
4162 outs() << " ovrfl " << fpu.fpu_fcw.ovrfl;
4163 outs() << " undfl " << fpu.fpu_fcw.undfl;
4164 outs() << " precis " << fpu.fpu_fcw.precis << "\n";
4165 outs() << "\t\t pc ";
4166 if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B)
4167 outs() << "FP_PREC_24B ";
4168 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B)
4169 outs() << "FP_PREC_53B ";
4170 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B)
4171 outs() << "FP_PREC_64B ";
4173 outs() << fpu.fpu_fcw.pc << " ";
4175 if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR)
4176 outs() << "FP_RND_NEAR ";
4177 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN)
4178 outs() << "FP_RND_DOWN ";
4179 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP)
4180 outs() << "FP_RND_UP ";
4181 else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP)
4182 outs() << "FP_CHOP ";
4184 outs() << "\t status: invalid " << fpu.fpu_fsw.invalid;
4185 outs() << " denorm " << fpu.fpu_fsw.denorm;
4186 outs() << " zdiv " << fpu.fpu_fsw.zdiv;
4187 outs() << " ovrfl " << fpu.fpu_fsw.ovrfl;
4188 outs() << " undfl " << fpu.fpu_fsw.undfl;
4189 outs() << " precis " << fpu.fpu_fsw.precis;
4190 outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n";
4191 outs() << "\t errsumm " << fpu.fpu_fsw.errsumm;
4192 outs() << " c0 " << fpu.fpu_fsw.c0;
4193 outs() << " c1 " << fpu.fpu_fsw.c1;
4194 outs() << " c2 " << fpu.fpu_fsw.c2;
4195 outs() << " tos " << fpu.fpu_fsw.tos;
4196 outs() << " c3 " << fpu.fpu_fsw.c3;
4197 outs() << " busy " << fpu.fpu_fsw.busy << "\n";
4198 outs() << "\t fpu_ftw " << format("0x%02" PRIx32, fpu.fpu_ftw);
4199 outs() << " fpu_rsrv1 " << format("0x%02" PRIx32, fpu.fpu_rsrv1);
4200 outs() << " fpu_fop " << format("0x%04" PRIx32, fpu.fpu_fop);
4201 outs() << " fpu_ip " << format("0x%08" PRIx32, fpu.fpu_ip) << "\n";
4202 outs() << "\t fpu_cs " << format("0x%04" PRIx32, fpu.fpu_cs);
4203 outs() << " fpu_rsrv2 " << format("0x%04" PRIx32, fpu.fpu_rsrv2);
4204 outs() << " fpu_dp " << format("0x%08" PRIx32, fpu.fpu_dp);
4205 outs() << " fpu_ds " << format("0x%04" PRIx32, fpu.fpu_ds) << "\n";
4206 outs() << "\t fpu_rsrv3 " << format("0x%04" PRIx32, fpu.fpu_rsrv3);
4207 outs() << " fpu_mxcsr " << format("0x%08" PRIx32, fpu.fpu_mxcsr);
4208 outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32, fpu.fpu_mxcsrmask);
4210 outs() << "\t fpu_stmm0:\n";
4211 Print_mmst_reg(fpu.fpu_stmm0);
4212 outs() << "\t fpu_stmm1:\n";
4213 Print_mmst_reg(fpu.fpu_stmm1);
4214 outs() << "\t fpu_stmm2:\n";
4215 Print_mmst_reg(fpu.fpu_stmm2);
4216 outs() << "\t fpu_stmm3:\n";
4217 Print_mmst_reg(fpu.fpu_stmm3);
4218 outs() << "\t fpu_stmm4:\n";
4219 Print_mmst_reg(fpu.fpu_stmm4);
4220 outs() << "\t fpu_stmm5:\n";
4221 Print_mmst_reg(fpu.fpu_stmm5);
4222 outs() << "\t fpu_stmm6:\n";
4223 Print_mmst_reg(fpu.fpu_stmm6);
4224 outs() << "\t fpu_stmm7:\n";
4225 Print_mmst_reg(fpu.fpu_stmm7);
4226 outs() << "\t fpu_xmm0:\n";
4227 Print_xmm_reg(fpu.fpu_xmm0);
4228 outs() << "\t fpu_xmm1:\n";
4229 Print_xmm_reg(fpu.fpu_xmm1);
4230 outs() << "\t fpu_xmm2:\n";
4231 Print_xmm_reg(fpu.fpu_xmm2);
4232 outs() << "\t fpu_xmm3:\n";
4233 Print_xmm_reg(fpu.fpu_xmm3);
4234 outs() << "\t fpu_xmm4:\n";
4235 Print_xmm_reg(fpu.fpu_xmm4);
4236 outs() << "\t fpu_xmm5:\n";
4237 Print_xmm_reg(fpu.fpu_xmm5);
4238 outs() << "\t fpu_xmm6:\n";
4239 Print_xmm_reg(fpu.fpu_xmm6);
4240 outs() << "\t fpu_xmm7:\n";
4241 Print_xmm_reg(fpu.fpu_xmm7);
4242 outs() << "\t fpu_xmm8:\n";
4243 Print_xmm_reg(fpu.fpu_xmm8);
4244 outs() << "\t fpu_xmm9:\n";
4245 Print_xmm_reg(fpu.fpu_xmm9);
4246 outs() << "\t fpu_xmm10:\n";
4247 Print_xmm_reg(fpu.fpu_xmm10);
4248 outs() << "\t fpu_xmm11:\n";
4249 Print_xmm_reg(fpu.fpu_xmm11);
4250 outs() << "\t fpu_xmm12:\n";
4251 Print_xmm_reg(fpu.fpu_xmm12);
4252 outs() << "\t fpu_xmm13:\n";
4253 Print_xmm_reg(fpu.fpu_xmm13);
4254 outs() << "\t fpu_xmm14:\n";
4255 Print_xmm_reg(fpu.fpu_xmm14);
4256 outs() << "\t fpu_xmm15:\n";
4257 Print_xmm_reg(fpu.fpu_xmm15);
4258 outs() << "\t fpu_rsrv4:\n";
4259 for (uint32_t f = 0; f < 6; f++) {
4261 for (uint32_t g = 0; g < 16; g++)
4262 outs() << format("%02" PRIx32, fpu.fpu_rsrv4[f * g]) << " ";
4265 outs() << "\t fpu_reserved1 " << format("0x%08" PRIx32, fpu.fpu_reserved1);
4269 static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) {
4270 outs() << "\t trapno " << format("0x%08" PRIx32, exc64.trapno);
4271 outs() << " err " << format("0x%08" PRIx32, exc64.err);
4272 outs() << " faultvaddr " << format("0x%016" PRIx64, exc64.faultvaddr) << "\n";
4275 static void PrintThreadCommand(MachO::thread_command t, const char *Ptr,
4276 bool isLittleEndian, uint32_t cputype) {
4277 if (t.cmd == MachO::LC_THREAD)
4278 outs() << " cmd LC_THREAD\n";
4279 else if (t.cmd == MachO::LC_UNIXTHREAD)
4280 outs() << " cmd LC_UNIXTHREAD\n";
4282 outs() << " cmd " << t.cmd << " (unknown)\n";
4283 outs() << " cmdsize " << t.cmdsize;
4284 if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t))
4285 outs() << " Incorrect size\n";
4289 const char *begin = Ptr + sizeof(struct MachO::thread_command);
4290 const char *end = Ptr + t.cmdsize;
4291 uint32_t flavor, count, left;
4292 if (cputype == MachO::CPU_TYPE_X86_64) {
4293 while (begin < end) {
4294 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4295 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4296 begin += sizeof(uint32_t);
4301 if (isLittleEndian != sys::IsLittleEndianHost)
4302 sys::swapByteOrder(flavor);
4303 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4304 memcpy((char *)&count, begin, sizeof(uint32_t));
4305 begin += sizeof(uint32_t);
4310 if (isLittleEndian != sys::IsLittleEndianHost)
4311 sys::swapByteOrder(count);
4312 if (flavor == MachO::x86_THREAD_STATE64) {
4313 outs() << " flavor x86_THREAD_STATE64\n";
4314 if (count == MachO::x86_THREAD_STATE64_COUNT)
4315 outs() << " count x86_THREAD_STATE64_COUNT\n";
4317 outs() << " count " << count
4318 << " (not x86_THREAD_STATE64_COUNT)\n";
4319 MachO::x86_thread_state64_t cpu64;
4321 if (left >= sizeof(MachO::x86_thread_state64_t)) {
4322 memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t));
4323 begin += sizeof(MachO::x86_thread_state64_t);
4325 memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t));
4326 memcpy(&cpu64, begin, left);
4329 if (isLittleEndian != sys::IsLittleEndianHost)
4331 Print_x86_thread_state64_t(cpu64);
4332 } else if (flavor == MachO::x86_THREAD_STATE) {
4333 outs() << " flavor x86_THREAD_STATE\n";
4334 if (count == MachO::x86_THREAD_STATE_COUNT)
4335 outs() << " count x86_THREAD_STATE_COUNT\n";
4337 outs() << " count " << count
4338 << " (not x86_THREAD_STATE_COUNT)\n";
4339 struct MachO::x86_thread_state_t ts;
4341 if (left >= sizeof(MachO::x86_thread_state_t)) {
4342 memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t));
4343 begin += sizeof(MachO::x86_thread_state_t);
4345 memset(&ts, '\0', sizeof(MachO::x86_thread_state_t));
4346 memcpy(&ts, begin, left);
4349 if (isLittleEndian != sys::IsLittleEndianHost)
4351 if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) {
4352 outs() << "\t tsh.flavor x86_THREAD_STATE64 ";
4353 if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT)
4354 outs() << "tsh.count x86_THREAD_STATE64_COUNT\n";
4356 outs() << "tsh.count " << ts.tsh.count
4357 << " (not x86_THREAD_STATE64_COUNT\n";
4358 Print_x86_thread_state64_t(ts.uts.ts64);
4360 outs() << "\t tsh.flavor " << ts.tsh.flavor << " tsh.count "
4361 << ts.tsh.count << "\n";
4363 } else if (flavor == MachO::x86_FLOAT_STATE) {
4364 outs() << " flavor x86_FLOAT_STATE\n";
4365 if (count == MachO::x86_FLOAT_STATE_COUNT)
4366 outs() << " count x86_FLOAT_STATE_COUNT\n";
4368 outs() << " count " << count << " (not x86_FLOAT_STATE_COUNT)\n";
4369 struct MachO::x86_float_state_t fs;
4371 if (left >= sizeof(MachO::x86_float_state_t)) {
4372 memcpy(&fs, begin, sizeof(MachO::x86_float_state_t));
4373 begin += sizeof(MachO::x86_float_state_t);
4375 memset(&fs, '\0', sizeof(MachO::x86_float_state_t));
4376 memcpy(&fs, begin, left);
4379 if (isLittleEndian != sys::IsLittleEndianHost)
4381 if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) {
4382 outs() << "\t fsh.flavor x86_FLOAT_STATE64 ";
4383 if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT)
4384 outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n";
4386 outs() << "fsh.count " << fs.fsh.count
4387 << " (not x86_FLOAT_STATE64_COUNT\n";
4388 Print_x86_float_state_t(fs.ufs.fs64);
4390 outs() << "\t fsh.flavor " << fs.fsh.flavor << " fsh.count "
4391 << fs.fsh.count << "\n";
4393 } else if (flavor == MachO::x86_EXCEPTION_STATE) {
4394 outs() << " flavor x86_EXCEPTION_STATE\n";
4395 if (count == MachO::x86_EXCEPTION_STATE_COUNT)
4396 outs() << " count x86_EXCEPTION_STATE_COUNT\n";
4398 outs() << " count " << count
4399 << " (not x86_EXCEPTION_STATE_COUNT)\n";
4400 struct MachO::x86_exception_state_t es;
4402 if (left >= sizeof(MachO::x86_exception_state_t)) {
4403 memcpy(&es, begin, sizeof(MachO::x86_exception_state_t));
4404 begin += sizeof(MachO::x86_exception_state_t);
4406 memset(&es, '\0', sizeof(MachO::x86_exception_state_t));
4407 memcpy(&es, begin, left);
4410 if (isLittleEndian != sys::IsLittleEndianHost)
4412 if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) {
4413 outs() << "\t esh.flavor x86_EXCEPTION_STATE64\n";
4414 if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT)
4415 outs() << "\t esh.count x86_EXCEPTION_STATE64_COUNT\n";
4417 outs() << "\t esh.count " << es.esh.count
4418 << " (not x86_EXCEPTION_STATE64_COUNT\n";
4419 Print_x86_exception_state_t(es.ues.es64);
4421 outs() << "\t esh.flavor " << es.esh.flavor << " esh.count "
4422 << es.esh.count << "\n";
4425 outs() << " flavor " << flavor << " (unknown)\n";
4426 outs() << " count " << count << "\n";
4427 outs() << " state (unknown)\n";
4428 begin += count * sizeof(uint32_t);
4432 while (begin < end) {
4433 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4434 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4435 begin += sizeof(uint32_t);
4440 if (isLittleEndian != sys::IsLittleEndianHost)
4441 sys::swapByteOrder(flavor);
4442 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4443 memcpy((char *)&count, begin, sizeof(uint32_t));
4444 begin += sizeof(uint32_t);
4449 if (isLittleEndian != sys::IsLittleEndianHost)
4450 sys::swapByteOrder(count);
4451 outs() << " flavor " << flavor << "\n";
4452 outs() << " count " << count << "\n";
4453 outs() << " state (Unknown cputype/cpusubtype)\n";
4454 begin += count * sizeof(uint32_t);
4459 static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
4460 if (dl.cmd == MachO::LC_ID_DYLIB)
4461 outs() << " cmd LC_ID_DYLIB\n";
4462 else if (dl.cmd == MachO::LC_LOAD_DYLIB)
4463 outs() << " cmd LC_LOAD_DYLIB\n";
4464 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
4465 outs() << " cmd LC_LOAD_WEAK_DYLIB\n";
4466 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
4467 outs() << " cmd LC_REEXPORT_DYLIB\n";
4468 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
4469 outs() << " cmd LC_LAZY_LOAD_DYLIB\n";
4470 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
4471 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n";
4473 outs() << " cmd " << dl.cmd << " (unknown)\n";
4474 outs() << " cmdsize " << dl.cmdsize;
4475 if (dl.cmdsize < sizeof(struct MachO::dylib_command))
4476 outs() << " Incorrect size\n";
4479 if (dl.dylib.name < dl.cmdsize) {
4480 const char *P = (const char *)(Ptr) + dl.dylib.name;
4481 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n";
4483 outs() << " name ?(bad offset " << dl.dylib.name << ")\n";
4485 outs() << " time stamp " << dl.dylib.timestamp << " ";
4486 time_t t = dl.dylib.timestamp;
4487 outs() << ctime(&t);
4488 outs() << " current version ";
4489 if (dl.dylib.current_version == 0xffffffff)
4492 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
4493 << ((dl.dylib.current_version >> 8) & 0xff) << "."
4494 << (dl.dylib.current_version & 0xff) << "\n";
4495 outs() << "compatibility version ";
4496 if (dl.dylib.compatibility_version == 0xffffffff)
4499 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
4500 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
4501 << (dl.dylib.compatibility_version & 0xff) << "\n";
4504 static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
4505 uint32_t object_size) {
4506 if (ld.cmd == MachO::LC_CODE_SIGNATURE)
4507 outs() << " cmd LC_FUNCTION_STARTS\n";
4508 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
4509 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n";
4510 else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
4511 outs() << " cmd LC_FUNCTION_STARTS\n";
4512 else if (ld.cmd == MachO::LC_DATA_IN_CODE)
4513 outs() << " cmd LC_DATA_IN_CODE\n";
4514 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
4515 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n";
4516 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
4517 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n";
4519 outs() << " cmd " << ld.cmd << " (?)\n";
4520 outs() << " cmdsize " << ld.cmdsize;
4521 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
4522 outs() << " Incorrect size\n";
4525 outs() << " dataoff " << ld.dataoff;
4526 if (ld.dataoff > object_size)
4527 outs() << " (past end of file)\n";
4530 outs() << " datasize " << ld.datasize;
4531 uint64_t big_size = ld.dataoff;
4532 big_size += ld.datasize;
4533 if (big_size > object_size)
4534 outs() << " (past end of file)\n";
4539 static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t ncmds,
4540 uint32_t filetype, uint32_t cputype,
4544 StringRef Buf = Obj->getData();
4545 MachOObjectFile::LoadCommandInfo Command = Obj->getFirstLoadCommandInfo();
4546 for (unsigned i = 0;; ++i) {
4547 outs() << "Load command " << i << "\n";
4548 if (Command.C.cmd == MachO::LC_SEGMENT) {
4549 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
4550 const char *sg_segname = SLC.segname;
4551 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
4552 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
4553 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
4555 for (unsigned j = 0; j < SLC.nsects; j++) {
4556 MachO::section S = Obj->getSection(Command, j);
4557 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
4558 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
4559 SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
4561 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
4562 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
4563 const char *sg_segname = SLC_64.segname;
4564 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
4565 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
4566 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
4567 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
4568 for (unsigned j = 0; j < SLC_64.nsects; j++) {
4569 MachO::section_64 S_64 = Obj->getSection64(Command, j);
4570 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
4571 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
4572 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
4573 sg_segname, filetype, Buf.size(), verbose);
4575 } else if (Command.C.cmd == MachO::LC_SYMTAB) {
4576 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
4577 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
4578 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
4579 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
4580 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
4581 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
4583 } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
4584 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
4585 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
4586 PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
4587 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
4588 Command.C.cmd == MachO::LC_ID_DYLINKER ||
4589 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
4590 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
4591 PrintDyldLoadCommand(Dyld, Command.Ptr);
4592 } else if (Command.C.cmd == MachO::LC_UUID) {
4593 MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
4594 PrintUuidLoadCommand(Uuid);
4595 } else if (Command.C.cmd == MachO::LC_RPATH) {
4596 MachO::rpath_command Rpath = Obj->getRpathCommand(Command);
4597 PrintRpathLoadCommand(Rpath, Command.Ptr);
4598 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX ||
4599 Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) {
4600 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
4601 PrintVersionMinLoadCommand(Vd);
4602 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
4603 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
4604 PrintSourceVersionCommand(Sd);
4605 } else if (Command.C.cmd == MachO::LC_MAIN) {
4606 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
4607 PrintEntryPointCommand(Ep);
4608 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) {
4609 MachO::encryption_info_command Ei =
4610 Obj->getEncryptionInfoCommand(Command);
4611 PrintEncryptionInfoCommand(Ei, Buf.size());
4612 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
4613 MachO::encryption_info_command_64 Ei =
4614 Obj->getEncryptionInfoCommand64(Command);
4615 PrintEncryptionInfoCommand64(Ei, Buf.size());
4616 } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) {
4617 MachO::linker_option_command Lo =
4618 Obj->getLinkerOptionLoadCommand(Command);
4619 PrintLinkerOptionCommand(Lo, Command.Ptr);
4620 } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) {
4621 MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command);
4622 PrintSubFrameworkCommand(Sf, Command.Ptr);
4623 } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) {
4624 MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command);
4625 PrintSubUmbrellaCommand(Sf, Command.Ptr);
4626 } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) {
4627 MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command);
4628 PrintSubLibraryCommand(Sl, Command.Ptr);
4629 } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) {
4630 MachO::sub_client_command Sc = Obj->getSubClientCommand(Command);
4631 PrintSubClientCommand(Sc, Command.Ptr);
4632 } else if (Command.C.cmd == MachO::LC_ROUTINES) {
4633 MachO::routines_command Rc = Obj->getRoutinesCommand(Command);
4634 PrintRoutinesCommand(Rc);
4635 } else if (Command.C.cmd == MachO::LC_ROUTINES_64) {
4636 MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command);
4637 PrintRoutinesCommand64(Rc);
4638 } else if (Command.C.cmd == MachO::LC_THREAD ||
4639 Command.C.cmd == MachO::LC_UNIXTHREAD) {
4640 MachO::thread_command Tc = Obj->getThreadCommand(Command);
4641 PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype);
4642 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
4643 Command.C.cmd == MachO::LC_ID_DYLIB ||
4644 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
4645 Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
4646 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
4647 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
4648 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
4649 PrintDylibCommand(Dl, Command.Ptr);
4650 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
4651 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
4652 Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
4653 Command.C.cmd == MachO::LC_DATA_IN_CODE ||
4654 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
4655 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
4656 MachO::linkedit_data_command Ld =
4657 Obj->getLinkeditDataLoadCommand(Command);
4658 PrintLinkEditDataCommand(Ld, Buf.size());
4660 outs() << " cmd ?(" << format("0x%08" PRIx32, Command.C.cmd)
4662 outs() << " cmdsize " << Command.C.cmdsize << "\n";
4663 // TODO: get and print the raw bytes of the load command.
4665 // TODO: print all the other kinds of load commands.
4669 Command = Obj->getNextLoadCommandInfo(Command);
4673 static void getAndPrintMachHeader(const MachOObjectFile *Obj, uint32_t &ncmds,
4674 uint32_t &filetype, uint32_t &cputype,
4676 if (Obj->is64Bit()) {
4677 MachO::mach_header_64 H_64;
4678 H_64 = Obj->getHeader64();
4679 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
4680 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
4682 filetype = H_64.filetype;
4683 cputype = H_64.cputype;
4685 MachO::mach_header H;
4686 H = Obj->getHeader();
4687 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
4688 H.sizeofcmds, H.flags, verbose);
4690 filetype = H.filetype;
4691 cputype = H.cputype;
4695 void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
4696 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
4698 uint32_t filetype = 0;
4699 uint32_t cputype = 0;
4700 getAndPrintMachHeader(file, ncmds, filetype, cputype, true);
4701 PrintLoadCommands(file, ncmds, filetype, cputype, true);
4704 //===----------------------------------------------------------------------===//
4705 // export trie dumping
4706 //===----------------------------------------------------------------------===//
4708 void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
4709 for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
4710 uint64_t Flags = Entry.flags();
4711 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
4712 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
4713 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
4714 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
4715 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
4716 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
4717 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
4719 outs() << "[re-export] ";
4721 outs() << format("0x%08llX ",
4722 Entry.address()); // FIXME:add in base address
4723 outs() << Entry.name();
4724 if (WeakDef || ThreadLocal || Resolver || Abs) {
4725 bool NeedsComma = false;
4728 outs() << "weak_def";
4734 outs() << "per-thread";
4740 outs() << "absolute";
4746 outs() << format("resolver=0x%08llX", Entry.other());
4752 StringRef DylibName = "unknown";
4753 int Ordinal = Entry.other() - 1;
4754 Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
4755 if (Entry.otherName().empty())
4756 outs() << " (from " << DylibName << ")";
4758 outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
4764 //===----------------------------------------------------------------------===//
4765 // rebase table dumping
4766 //===----------------------------------------------------------------------===//
4771 SegInfo(const object::MachOObjectFile *Obj);
4773 StringRef segmentName(uint32_t SegIndex);
4774 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
4775 uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
4778 struct SectionInfo {
4781 StringRef SectionName;
4782 StringRef SegmentName;
4783 uint64_t OffsetInSegment;
4784 uint64_t SegmentStartAddress;
4785 uint32_t SegmentIndex;
4787 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
4788 SmallVector<SectionInfo, 32> Sections;
4792 SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
4793 // Build table of sections so segIndex/offset pairs can be translated.
4794 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
4795 StringRef CurSegName;
4796 uint64_t CurSegAddress;
4797 for (const SectionRef &Section : Obj->sections()) {
4799 if (error(Section.getName(Info.SectionName)))
4801 Info.Address = Section.getAddress();
4802 Info.Size = Section.getSize();
4804 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
4805 if (!Info.SegmentName.equals(CurSegName)) {
4807 CurSegName = Info.SegmentName;
4808 CurSegAddress = Info.Address;
4810 Info.SegmentIndex = CurSegIndex - 1;
4811 Info.OffsetInSegment = Info.Address - CurSegAddress;
4812 Info.SegmentStartAddress = CurSegAddress;
4813 Sections.push_back(Info);
4817 StringRef SegInfo::segmentName(uint32_t SegIndex) {
4818 for (const SectionInfo &SI : Sections) {
4819 if (SI.SegmentIndex == SegIndex)
4820 return SI.SegmentName;
4822 llvm_unreachable("invalid segIndex");
4825 const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
4826 uint64_t OffsetInSeg) {
4827 for (const SectionInfo &SI : Sections) {
4828 if (SI.SegmentIndex != SegIndex)
4830 if (SI.OffsetInSegment > OffsetInSeg)
4832 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
4836 llvm_unreachable("segIndex and offset not in any section");
4839 StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
4840 return findSection(SegIndex, OffsetInSeg).SectionName;
4843 uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
4844 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
4845 return SI.SegmentStartAddress + OffsetInSeg;
4848 void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
4849 // Build table of sections so names can used in final output.
4850 SegInfo sectionTable(Obj);
4852 outs() << "segment section address type\n";
4853 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
4854 uint32_t SegIndex = Entry.segmentIndex();
4855 uint64_t OffsetInSeg = Entry.segmentOffset();
4856 StringRef SegmentName = sectionTable.segmentName(SegIndex);
4857 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
4858 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
4860 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer
4861 outs() << format("%-8s %-18s 0x%08" PRIX64 " %s\n",
4862 SegmentName.str().c_str(), SectionName.str().c_str(),
4863 Address, Entry.typeName().str().c_str());
4867 static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
4868 StringRef DylibName;
4870 case MachO::BIND_SPECIAL_DYLIB_SELF:
4871 return "this-image";
4872 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
4873 return "main-executable";
4874 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
4875 return "flat-namespace";
4878 std::error_code EC =
4879 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
4881 return "<<bad library ordinal>>";
4885 return "<<unknown special ordinal>>";
4888 //===----------------------------------------------------------------------===//
4889 // bind table dumping
4890 //===----------------------------------------------------------------------===//
4892 void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
4893 // Build table of sections so names can used in final output.
4894 SegInfo sectionTable(Obj);
4896 outs() << "segment section address type "
4897 "addend dylib symbol\n";
4898 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
4899 uint32_t SegIndex = Entry.segmentIndex();
4900 uint64_t OffsetInSeg = Entry.segmentOffset();
4901 StringRef SegmentName = sectionTable.segmentName(SegIndex);
4902 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
4903 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
4905 // Table lines look like:
4906 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard
4908 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
4909 Attr = " (weak_import)";
4910 outs() << left_justify(SegmentName, 8) << " "
4911 << left_justify(SectionName, 18) << " "
4912 << format_hex(Address, 10, true) << " "
4913 << left_justify(Entry.typeName(), 8) << " "
4914 << format_decimal(Entry.addend(), 8) << " "
4915 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
4916 << Entry.symbolName() << Attr << "\n";
4920 //===----------------------------------------------------------------------===//
4921 // lazy bind table dumping
4922 //===----------------------------------------------------------------------===//
4924 void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
4925 // Build table of sections so names can used in final output.
4926 SegInfo sectionTable(Obj);
4928 outs() << "segment section address "
4930 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
4931 uint32_t SegIndex = Entry.segmentIndex();
4932 uint64_t OffsetInSeg = Entry.segmentOffset();
4933 StringRef SegmentName = sectionTable.segmentName(SegIndex);
4934 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
4935 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
4937 // Table lines look like:
4938 // __DATA __got 0x00012010 libSystem ___stack_chk_guard
4939 outs() << left_justify(SegmentName, 8) << " "
4940 << left_justify(SectionName, 18) << " "
4941 << format_hex(Address, 10, true) << " "
4942 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
4943 << Entry.symbolName() << "\n";
4947 //===----------------------------------------------------------------------===//
4948 // weak bind table dumping
4949 //===----------------------------------------------------------------------===//
4951 void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
4952 // Build table of sections so names can used in final output.
4953 SegInfo sectionTable(Obj);
4955 outs() << "segment section address "
4956 "type addend symbol\n";
4957 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
4958 // Strong symbols don't have a location to update.
4959 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
4960 outs() << " strong "
4961 << Entry.symbolName() << "\n";
4964 uint32_t SegIndex = Entry.segmentIndex();
4965 uint64_t OffsetInSeg = Entry.segmentOffset();
4966 StringRef SegmentName = sectionTable.segmentName(SegIndex);
4967 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
4968 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
4970 // Table lines look like:
4971 // __DATA __data 0x00001000 pointer 0 _foo
4972 outs() << left_justify(SegmentName, 8) << " "
4973 << left_justify(SectionName, 18) << " "
4974 << format_hex(Address, 10, true) << " "
4975 << left_justify(Entry.typeName(), 8) << " "
4976 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName()
4981 // get_dyld_bind_info_symbolname() is used for disassembly and passed an
4982 // address, ReferenceValue, in the Mach-O file and looks in the dyld bind
4983 // information for that address. If the address is found its binding symbol
4984 // name is returned. If not nullptr is returned.
4985 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
4986 struct DisassembleInfo *info) {
4987 if (info->bindtable == nullptr) {
4988 info->bindtable = new (BindTable);
4989 SegInfo sectionTable(info->O);
4990 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
4991 uint32_t SegIndex = Entry.segmentIndex();
4992 uint64_t OffsetInSeg = Entry.segmentOffset();
4993 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
4994 const char *SymbolName = nullptr;
4995 StringRef name = Entry.symbolName();
4997 SymbolName = name.data();
4998 info->bindtable->push_back(std::make_pair(Address, SymbolName));
5001 for (bind_table_iterator BI = info->bindtable->begin(),
5002 BE = info->bindtable->end();
5004 uint64_t Address = BI->first;
5005 if (ReferenceValue == Address) {
5006 const char *SymbolName = BI->second;