1 //===-- MachODump.cpp - Object file dumping utility for llvm --------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements the MachO-specific dumper for llvm-objdump.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "llvm-objdump.h"
15 #include "llvm-c/Disassembler.h"
16 #include "llvm/ADT/Optional.h"
17 #include "llvm/ADT/STLExtras.h"
18 #include "llvm/ADT/StringExtras.h"
19 #include "llvm/ADT/Triple.h"
20 #include "llvm/Config/config.h"
21 #include "llvm/DebugInfo/DIContext.h"
22 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
23 #include "llvm/MC/MCAsmInfo.h"
24 #include "llvm/MC/MCContext.h"
25 #include "llvm/MC/MCDisassembler.h"
26 #include "llvm/MC/MCInst.h"
27 #include "llvm/MC/MCInstPrinter.h"
28 #include "llvm/MC/MCInstrDesc.h"
29 #include "llvm/MC/MCInstrInfo.h"
30 #include "llvm/MC/MCRegisterInfo.h"
31 #include "llvm/MC/MCSubtargetInfo.h"
32 #include "llvm/Object/MachO.h"
33 #include "llvm/Object/MachOUniversal.h"
34 #include "llvm/Support/Casting.h"
35 #include "llvm/Support/CommandLine.h"
36 #include "llvm/Support/Debug.h"
37 #include "llvm/Support/Endian.h"
38 #include "llvm/Support/Format.h"
39 #include "llvm/Support/FormattedStream.h"
40 #include "llvm/Support/GraphWriter.h"
41 #include "llvm/Support/LEB128.h"
42 #include "llvm/Support/MachO.h"
43 #include "llvm/Support/MemoryBuffer.h"
44 #include "llvm/Support/TargetRegistry.h"
45 #include "llvm/Support/TargetSelect.h"
46 #include "llvm/Support/raw_ostream.h"
47 #include <algorithm>
48 #include <cstring>
49 #include <system_error>
50 
51 #if HAVE_CXXABI_H
52 #include <cxxabi.h>
53 #endif
54 
55 using namespace llvm;
56 using namespace object;
57 
58 static cl::opt<bool>
59     UseDbg("g",
60            cl::desc("Print line information from debug info if available"));
61 
62 static cl::opt<std::string> DSYMFile("dsym",
63                                      cl::desc("Use .dSYM file for debug info"));
64 
65 static cl::opt<bool> FullLeadingAddr("full-leading-addr",
66                                      cl::desc("Print full leading address"));
67 
68 static cl::opt<bool> NoLeadingAddr("no-leading-addr",
69                                    cl::desc("Print no leading address"));
70 
71 cl::opt<bool> llvm::UniversalHeaders("universal-headers",
72                                      cl::desc("Print Mach-O universal headers "
73                                               "(requires -macho)"));
74 
75 cl::opt<bool>
76     llvm::ArchiveHeaders("archive-headers",
77                          cl::desc("Print archive headers for Mach-O archives "
78                                   "(requires -macho)"));
79 
80 cl::opt<bool>
81     ArchiveMemberOffsets("archive-member-offsets",
82                          cl::desc("Print the offset to each archive member for "
83                                   "Mach-O archives (requires -macho and "
84                                   "-archive-headers)"));
85 
86 cl::opt<bool>
87     llvm::IndirectSymbols("indirect-symbols",
88                           cl::desc("Print indirect symbol table for Mach-O "
89                                    "objects (requires -macho)"));
90 
91 cl::opt<bool>
92     llvm::DataInCode("data-in-code",
93                      cl::desc("Print the data in code table for Mach-O objects "
94                               "(requires -macho)"));
95 
96 cl::opt<bool>
97     llvm::LinkOptHints("link-opt-hints",
98                        cl::desc("Print the linker optimization hints for "
99                                 "Mach-O objects (requires -macho)"));
100 
101 cl::opt<bool>
102     llvm::InfoPlist("info-plist",
103                     cl::desc("Print the info plist section as strings for "
104                              "Mach-O objects (requires -macho)"));
105 
106 cl::opt<bool>
107     llvm::DylibsUsed("dylibs-used",
108                      cl::desc("Print the shared libraries used for linked "
109                               "Mach-O files (requires -macho)"));
110 
111 cl::opt<bool>
112     llvm::DylibId("dylib-id",
113                   cl::desc("Print the shared library's id for the dylib Mach-O "
114                            "file (requires -macho)"));
115 
116 cl::opt<bool>
117     llvm::NonVerbose("non-verbose",
118                      cl::desc("Print the info for Mach-O objects in "
119                               "non-verbose or numeric form (requires -macho)"));
120 
121 cl::opt<bool>
122     llvm::ObjcMetaData("objc-meta-data",
123                        cl::desc("Print the Objective-C runtime meta data for "
124                                 "Mach-O files (requires -macho)"));
125 
126 cl::opt<std::string> llvm::DisSymName(
127     "dis-symname",
128     cl::desc("disassemble just this symbol's instructions (requires -macho"));
129 
130 static cl::opt<bool> NoSymbolicOperands(
131     "no-symbolic-operands",
132     cl::desc("do not symbolic operands when disassembling (requires -macho)"));
133 
134 static cl::list<std::string>
135     ArchFlags("arch", cl::desc("architecture(s) from a Mach-O file to dump"),
136               cl::ZeroOrMore);
137 
138 bool ArchAll = false;
139 
140 static std::string ThumbTripleName;
141 
GetTarget(const MachOObjectFile * MachOObj,const char ** McpuDefault,const Target ** ThumbTarget)142 static const Target *GetTarget(const MachOObjectFile *MachOObj,
143                                const char **McpuDefault,
144                                const Target **ThumbTarget) {
145   // Figure out the target triple.
146   if (TripleName.empty()) {
147     llvm::Triple TT("unknown-unknown-unknown");
148     llvm::Triple ThumbTriple = Triple();
149     TT = MachOObj->getArch(McpuDefault, &ThumbTriple);
150     TripleName = TT.str();
151     ThumbTripleName = ThumbTriple.str();
152   }
153 
154   // Get the target specific parser.
155   std::string Error;
156   const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
157   if (TheTarget && ThumbTripleName.empty())
158     return TheTarget;
159 
160   *ThumbTarget = TargetRegistry::lookupTarget(ThumbTripleName, Error);
161   if (*ThumbTarget)
162     return TheTarget;
163 
164   errs() << "llvm-objdump: error: unable to get target for '";
165   if (!TheTarget)
166     errs() << TripleName;
167   else
168     errs() << ThumbTripleName;
169   errs() << "', see --version and --triple.\n";
170   return nullptr;
171 }
172 
173 struct SymbolSorter {
operator ()SymbolSorter174   bool operator()(const SymbolRef &A, const SymbolRef &B) {
175     uint64_t AAddr = (A.getType() != SymbolRef::ST_Function) ? 0 : A.getValue();
176     uint64_t BAddr = (B.getType() != SymbolRef::ST_Function) ? 0 : B.getValue();
177     return AAddr < BAddr;
178   }
179 };
180 
181 // Types for the storted data in code table that is built before disassembly
182 // and the predicate function to sort them.
183 typedef std::pair<uint64_t, DiceRef> DiceTableEntry;
184 typedef std::vector<DiceTableEntry> DiceTable;
185 typedef DiceTable::iterator dice_table_iterator;
186 
187 // This is used to search for a data in code table entry for the PC being
188 // disassembled.  The j parameter has the PC in j.first.  A single data in code
189 // table entry can cover many bytes for each of its Kind's.  So if the offset,
190 // aka the i.first value, of the data in code table entry plus its Length
191 // covers the PC being searched for this will return true.  If not it will
192 // return false.
compareDiceTableEntries(const DiceTableEntry & i,const DiceTableEntry & j)193 static bool compareDiceTableEntries(const DiceTableEntry &i,
194                                     const DiceTableEntry &j) {
195   uint16_t Length;
196   i.second.getLength(Length);
197 
198   return j.first >= i.first && j.first < i.first + Length;
199 }
200 
DumpDataInCode(const uint8_t * bytes,uint64_t Length,unsigned short Kind)201 static uint64_t DumpDataInCode(const uint8_t *bytes, uint64_t Length,
202                                unsigned short Kind) {
203   uint32_t Value, Size = 1;
204 
205   switch (Kind) {
206   default:
207   case MachO::DICE_KIND_DATA:
208     if (Length >= 4) {
209       if (!NoShowRawInsn)
210         dumpBytes(makeArrayRef(bytes, 4), outs());
211       Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
212       outs() << "\t.long " << Value;
213       Size = 4;
214     } else if (Length >= 2) {
215       if (!NoShowRawInsn)
216         dumpBytes(makeArrayRef(bytes, 2), outs());
217       Value = bytes[1] << 8 | bytes[0];
218       outs() << "\t.short " << Value;
219       Size = 2;
220     } else {
221       if (!NoShowRawInsn)
222         dumpBytes(makeArrayRef(bytes, 2), outs());
223       Value = bytes[0];
224       outs() << "\t.byte " << Value;
225       Size = 1;
226     }
227     if (Kind == MachO::DICE_KIND_DATA)
228       outs() << "\t@ KIND_DATA\n";
229     else
230       outs() << "\t@ data in code kind = " << Kind << "\n";
231     break;
232   case MachO::DICE_KIND_JUMP_TABLE8:
233     if (!NoShowRawInsn)
234       dumpBytes(makeArrayRef(bytes, 1), outs());
235     Value = bytes[0];
236     outs() << "\t.byte " << format("%3u", Value) << "\t@ KIND_JUMP_TABLE8\n";
237     Size = 1;
238     break;
239   case MachO::DICE_KIND_JUMP_TABLE16:
240     if (!NoShowRawInsn)
241       dumpBytes(makeArrayRef(bytes, 2), outs());
242     Value = bytes[1] << 8 | bytes[0];
243     outs() << "\t.short " << format("%5u", Value & 0xffff)
244            << "\t@ KIND_JUMP_TABLE16\n";
245     Size = 2;
246     break;
247   case MachO::DICE_KIND_JUMP_TABLE32:
248   case MachO::DICE_KIND_ABS_JUMP_TABLE32:
249     if (!NoShowRawInsn)
250       dumpBytes(makeArrayRef(bytes, 4), outs());
251     Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
252     outs() << "\t.long " << Value;
253     if (Kind == MachO::DICE_KIND_JUMP_TABLE32)
254       outs() << "\t@ KIND_JUMP_TABLE32\n";
255     else
256       outs() << "\t@ KIND_ABS_JUMP_TABLE32\n";
257     Size = 4;
258     break;
259   }
260   return Size;
261 }
262 
getSectionsAndSymbols(MachOObjectFile * MachOObj,std::vector<SectionRef> & Sections,std::vector<SymbolRef> & Symbols,SmallVectorImpl<uint64_t> & FoundFns,uint64_t & BaseSegmentAddress)263 static void getSectionsAndSymbols(MachOObjectFile *MachOObj,
264                                   std::vector<SectionRef> &Sections,
265                                   std::vector<SymbolRef> &Symbols,
266                                   SmallVectorImpl<uint64_t> &FoundFns,
267                                   uint64_t &BaseSegmentAddress) {
268   for (const SymbolRef &Symbol : MachOObj->symbols()) {
269     ErrorOr<StringRef> SymName = Symbol.getName();
270     if (std::error_code EC = SymName.getError())
271       report_fatal_error(EC.message());
272     if (!SymName->startswith("ltmp"))
273       Symbols.push_back(Symbol);
274   }
275 
276   for (const SectionRef &Section : MachOObj->sections()) {
277     StringRef SectName;
278     Section.getName(SectName);
279     Sections.push_back(Section);
280   }
281 
282   bool BaseSegmentAddressSet = false;
283   for (const auto &Command : MachOObj->load_commands()) {
284     if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) {
285       // We found a function starts segment, parse the addresses for later
286       // consumption.
287       MachO::linkedit_data_command LLC =
288           MachOObj->getLinkeditDataLoadCommand(Command);
289 
290       MachOObj->ReadULEB128s(LLC.dataoff, FoundFns);
291     } else if (Command.C.cmd == MachO::LC_SEGMENT) {
292       MachO::segment_command SLC = MachOObj->getSegmentLoadCommand(Command);
293       StringRef SegName = SLC.segname;
294       if (!BaseSegmentAddressSet && SegName != "__PAGEZERO") {
295         BaseSegmentAddressSet = true;
296         BaseSegmentAddress = SLC.vmaddr;
297       }
298     }
299   }
300 }
301 
PrintIndirectSymbolTable(MachOObjectFile * O,bool verbose,uint32_t n,uint32_t count,uint32_t stride,uint64_t addr)302 static void PrintIndirectSymbolTable(MachOObjectFile *O, bool verbose,
303                                      uint32_t n, uint32_t count,
304                                      uint32_t stride, uint64_t addr) {
305   MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
306   uint32_t nindirectsyms = Dysymtab.nindirectsyms;
307   if (n > nindirectsyms)
308     outs() << " (entries start past the end of the indirect symbol "
309               "table) (reserved1 field greater than the table size)";
310   else if (n + count > nindirectsyms)
311     outs() << " (entries extends past the end of the indirect symbol "
312               "table)";
313   outs() << "\n";
314   uint32_t cputype = O->getHeader().cputype;
315   if (cputype & MachO::CPU_ARCH_ABI64)
316     outs() << "address            index";
317   else
318     outs() << "address    index";
319   if (verbose)
320     outs() << " name\n";
321   else
322     outs() << "\n";
323   for (uint32_t j = 0; j < count && n + j < nindirectsyms; j++) {
324     if (cputype & MachO::CPU_ARCH_ABI64)
325       outs() << format("0x%016" PRIx64, addr + j * stride) << " ";
326     else
327       outs() << format("0x%08" PRIx32, addr + j * stride) << " ";
328     MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
329     uint32_t indirect_symbol = O->getIndirectSymbolTableEntry(Dysymtab, n + j);
330     if (indirect_symbol == MachO::INDIRECT_SYMBOL_LOCAL) {
331       outs() << "LOCAL\n";
332       continue;
333     }
334     if (indirect_symbol ==
335         (MachO::INDIRECT_SYMBOL_LOCAL | MachO::INDIRECT_SYMBOL_ABS)) {
336       outs() << "LOCAL ABSOLUTE\n";
337       continue;
338     }
339     if (indirect_symbol == MachO::INDIRECT_SYMBOL_ABS) {
340       outs() << "ABSOLUTE\n";
341       continue;
342     }
343     outs() << format("%5u ", indirect_symbol);
344     if (verbose) {
345       MachO::symtab_command Symtab = O->getSymtabLoadCommand();
346       if (indirect_symbol < Symtab.nsyms) {
347         symbol_iterator Sym = O->getSymbolByIndex(indirect_symbol);
348         SymbolRef Symbol = *Sym;
349         ErrorOr<StringRef> SymName = Symbol.getName();
350         if (std::error_code EC = SymName.getError())
351           report_fatal_error(EC.message());
352         outs() << *SymName;
353       } else {
354         outs() << "?";
355       }
356     }
357     outs() << "\n";
358   }
359 }
360 
PrintIndirectSymbols(MachOObjectFile * O,bool verbose)361 static void PrintIndirectSymbols(MachOObjectFile *O, bool verbose) {
362   for (const auto &Load : O->load_commands()) {
363     if (Load.C.cmd == MachO::LC_SEGMENT_64) {
364       MachO::segment_command_64 Seg = O->getSegment64LoadCommand(Load);
365       for (unsigned J = 0; J < Seg.nsects; ++J) {
366         MachO::section_64 Sec = O->getSection64(Load, J);
367         uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
368         if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
369             section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
370             section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
371             section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
372             section_type == MachO::S_SYMBOL_STUBS) {
373           uint32_t stride;
374           if (section_type == MachO::S_SYMBOL_STUBS)
375             stride = Sec.reserved2;
376           else
377             stride = 8;
378           if (stride == 0) {
379             outs() << "Can't print indirect symbols for (" << Sec.segname << ","
380                    << Sec.sectname << ") "
381                    << "(size of stubs in reserved2 field is zero)\n";
382             continue;
383           }
384           uint32_t count = Sec.size / stride;
385           outs() << "Indirect symbols for (" << Sec.segname << ","
386                  << Sec.sectname << ") " << count << " entries";
387           uint32_t n = Sec.reserved1;
388           PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
389         }
390       }
391     } else if (Load.C.cmd == MachO::LC_SEGMENT) {
392       MachO::segment_command Seg = O->getSegmentLoadCommand(Load);
393       for (unsigned J = 0; J < Seg.nsects; ++J) {
394         MachO::section Sec = O->getSection(Load, J);
395         uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
396         if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
397             section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
398             section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
399             section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
400             section_type == MachO::S_SYMBOL_STUBS) {
401           uint32_t stride;
402           if (section_type == MachO::S_SYMBOL_STUBS)
403             stride = Sec.reserved2;
404           else
405             stride = 4;
406           if (stride == 0) {
407             outs() << "Can't print indirect symbols for (" << Sec.segname << ","
408                    << Sec.sectname << ") "
409                    << "(size of stubs in reserved2 field is zero)\n";
410             continue;
411           }
412           uint32_t count = Sec.size / stride;
413           outs() << "Indirect symbols for (" << Sec.segname << ","
414                  << Sec.sectname << ") " << count << " entries";
415           uint32_t n = Sec.reserved1;
416           PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
417         }
418       }
419     }
420   }
421 }
422 
PrintDataInCodeTable(MachOObjectFile * O,bool verbose)423 static void PrintDataInCodeTable(MachOObjectFile *O, bool verbose) {
424   MachO::linkedit_data_command DIC = O->getDataInCodeLoadCommand();
425   uint32_t nentries = DIC.datasize / sizeof(struct MachO::data_in_code_entry);
426   outs() << "Data in code table (" << nentries << " entries)\n";
427   outs() << "offset     length kind\n";
428   for (dice_iterator DI = O->begin_dices(), DE = O->end_dices(); DI != DE;
429        ++DI) {
430     uint32_t Offset;
431     DI->getOffset(Offset);
432     outs() << format("0x%08" PRIx32, Offset) << " ";
433     uint16_t Length;
434     DI->getLength(Length);
435     outs() << format("%6u", Length) << " ";
436     uint16_t Kind;
437     DI->getKind(Kind);
438     if (verbose) {
439       switch (Kind) {
440       case MachO::DICE_KIND_DATA:
441         outs() << "DATA";
442         break;
443       case MachO::DICE_KIND_JUMP_TABLE8:
444         outs() << "JUMP_TABLE8";
445         break;
446       case MachO::DICE_KIND_JUMP_TABLE16:
447         outs() << "JUMP_TABLE16";
448         break;
449       case MachO::DICE_KIND_JUMP_TABLE32:
450         outs() << "JUMP_TABLE32";
451         break;
452       case MachO::DICE_KIND_ABS_JUMP_TABLE32:
453         outs() << "ABS_JUMP_TABLE32";
454         break;
455       default:
456         outs() << format("0x%04" PRIx32, Kind);
457         break;
458       }
459     } else
460       outs() << format("0x%04" PRIx32, Kind);
461     outs() << "\n";
462   }
463 }
464 
PrintLinkOptHints(MachOObjectFile * O)465 static void PrintLinkOptHints(MachOObjectFile *O) {
466   MachO::linkedit_data_command LohLC = O->getLinkOptHintsLoadCommand();
467   const char *loh = O->getData().substr(LohLC.dataoff, 1).data();
468   uint32_t nloh = LohLC.datasize;
469   outs() << "Linker optimiztion hints (" << nloh << " total bytes)\n";
470   for (uint32_t i = 0; i < nloh;) {
471     unsigned n;
472     uint64_t identifier = decodeULEB128((const uint8_t *)(loh + i), &n);
473     i += n;
474     outs() << "    identifier " << identifier << " ";
475     if (i >= nloh)
476       return;
477     switch (identifier) {
478     case 1:
479       outs() << "AdrpAdrp\n";
480       break;
481     case 2:
482       outs() << "AdrpLdr\n";
483       break;
484     case 3:
485       outs() << "AdrpAddLdr\n";
486       break;
487     case 4:
488       outs() << "AdrpLdrGotLdr\n";
489       break;
490     case 5:
491       outs() << "AdrpAddStr\n";
492       break;
493     case 6:
494       outs() << "AdrpLdrGotStr\n";
495       break;
496     case 7:
497       outs() << "AdrpAdd\n";
498       break;
499     case 8:
500       outs() << "AdrpLdrGot\n";
501       break;
502     default:
503       outs() << "Unknown identifier value\n";
504       break;
505     }
506     uint64_t narguments = decodeULEB128((const uint8_t *)(loh + i), &n);
507     i += n;
508     outs() << "    narguments " << narguments << "\n";
509     if (i >= nloh)
510       return;
511 
512     for (uint32_t j = 0; j < narguments; j++) {
513       uint64_t value = decodeULEB128((const uint8_t *)(loh + i), &n);
514       i += n;
515       outs() << "\tvalue " << format("0x%" PRIx64, value) << "\n";
516       if (i >= nloh)
517         return;
518     }
519   }
520 }
521 
PrintDylibs(MachOObjectFile * O,bool JustId)522 static void PrintDylibs(MachOObjectFile *O, bool JustId) {
523   unsigned Index = 0;
524   for (const auto &Load : O->load_commands()) {
525     if ((JustId && Load.C.cmd == MachO::LC_ID_DYLIB) ||
526         (!JustId && (Load.C.cmd == MachO::LC_ID_DYLIB ||
527                      Load.C.cmd == MachO::LC_LOAD_DYLIB ||
528                      Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
529                      Load.C.cmd == MachO::LC_REEXPORT_DYLIB ||
530                      Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
531                      Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB))) {
532       MachO::dylib_command dl = O->getDylibIDLoadCommand(Load);
533       if (dl.dylib.name < dl.cmdsize) {
534         const char *p = (const char *)(Load.Ptr) + dl.dylib.name;
535         if (JustId)
536           outs() << p << "\n";
537         else {
538           outs() << "\t" << p;
539           outs() << " (compatibility version "
540                  << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
541                  << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
542                  << (dl.dylib.compatibility_version & 0xff) << ",";
543           outs() << " current version "
544                  << ((dl.dylib.current_version >> 16) & 0xffff) << "."
545                  << ((dl.dylib.current_version >> 8) & 0xff) << "."
546                  << (dl.dylib.current_version & 0xff) << ")\n";
547         }
548       } else {
549         outs() << "\tBad offset (" << dl.dylib.name << ") for name of ";
550         if (Load.C.cmd == MachO::LC_ID_DYLIB)
551           outs() << "LC_ID_DYLIB ";
552         else if (Load.C.cmd == MachO::LC_LOAD_DYLIB)
553           outs() << "LC_LOAD_DYLIB ";
554         else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB)
555           outs() << "LC_LOAD_WEAK_DYLIB ";
556         else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB)
557           outs() << "LC_LAZY_LOAD_DYLIB ";
558         else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB)
559           outs() << "LC_REEXPORT_DYLIB ";
560         else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
561           outs() << "LC_LOAD_UPWARD_DYLIB ";
562         else
563           outs() << "LC_??? ";
564         outs() << "command " << Index++ << "\n";
565       }
566     }
567   }
568 }
569 
570 typedef DenseMap<uint64_t, StringRef> SymbolAddressMap;
571 
CreateSymbolAddressMap(MachOObjectFile * O,SymbolAddressMap * AddrMap)572 static void CreateSymbolAddressMap(MachOObjectFile *O,
573                                    SymbolAddressMap *AddrMap) {
574   // Create a map of symbol addresses to symbol names.
575   for (const SymbolRef &Symbol : O->symbols()) {
576     SymbolRef::Type ST = Symbol.getType();
577     if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
578         ST == SymbolRef::ST_Other) {
579       uint64_t Address = Symbol.getValue();
580       ErrorOr<StringRef> SymNameOrErr = Symbol.getName();
581       if (std::error_code EC = SymNameOrErr.getError())
582         report_fatal_error(EC.message());
583       StringRef SymName = *SymNameOrErr;
584       if (!SymName.startswith(".objc"))
585         (*AddrMap)[Address] = SymName;
586     }
587   }
588 }
589 
590 // GuessSymbolName is passed the address of what might be a symbol and a
591 // pointer to the SymbolAddressMap.  It returns the name of a symbol
592 // with that address or nullptr if no symbol is found with that address.
GuessSymbolName(uint64_t value,SymbolAddressMap * AddrMap)593 static const char *GuessSymbolName(uint64_t value, SymbolAddressMap *AddrMap) {
594   const char *SymbolName = nullptr;
595   // A DenseMap can't lookup up some values.
596   if (value != 0xffffffffffffffffULL && value != 0xfffffffffffffffeULL) {
597     StringRef name = AddrMap->lookup(value);
598     if (!name.empty())
599       SymbolName = name.data();
600   }
601   return SymbolName;
602 }
603 
DumpCstringChar(const char c)604 static void DumpCstringChar(const char c) {
605   char p[2];
606   p[0] = c;
607   p[1] = '\0';
608   outs().write_escaped(p);
609 }
610 
DumpCstringSection(MachOObjectFile * O,const char * sect,uint32_t sect_size,uint64_t sect_addr,bool print_addresses)611 static void DumpCstringSection(MachOObjectFile *O, const char *sect,
612                                uint32_t sect_size, uint64_t sect_addr,
613                                bool print_addresses) {
614   for (uint32_t i = 0; i < sect_size; i++) {
615     if (print_addresses) {
616       if (O->is64Bit())
617         outs() << format("%016" PRIx64, sect_addr + i) << "  ";
618       else
619         outs() << format("%08" PRIx64, sect_addr + i) << "  ";
620     }
621     for (; i < sect_size && sect[i] != '\0'; i++)
622       DumpCstringChar(sect[i]);
623     if (i < sect_size && sect[i] == '\0')
624       outs() << "\n";
625   }
626 }
627 
DumpLiteral4(uint32_t l,float f)628 static void DumpLiteral4(uint32_t l, float f) {
629   outs() << format("0x%08" PRIx32, l);
630   if ((l & 0x7f800000) != 0x7f800000)
631     outs() << format(" (%.16e)\n", f);
632   else {
633     if (l == 0x7f800000)
634       outs() << " (+Infinity)\n";
635     else if (l == 0xff800000)
636       outs() << " (-Infinity)\n";
637     else if ((l & 0x00400000) == 0x00400000)
638       outs() << " (non-signaling Not-a-Number)\n";
639     else
640       outs() << " (signaling Not-a-Number)\n";
641   }
642 }
643 
DumpLiteral4Section(MachOObjectFile * O,const char * sect,uint32_t sect_size,uint64_t sect_addr,bool print_addresses)644 static void DumpLiteral4Section(MachOObjectFile *O, const char *sect,
645                                 uint32_t sect_size, uint64_t sect_addr,
646                                 bool print_addresses) {
647   for (uint32_t i = 0; i < sect_size; i += sizeof(float)) {
648     if (print_addresses) {
649       if (O->is64Bit())
650         outs() << format("%016" PRIx64, sect_addr + i) << "  ";
651       else
652         outs() << format("%08" PRIx64, sect_addr + i) << "  ";
653     }
654     float f;
655     memcpy(&f, sect + i, sizeof(float));
656     if (O->isLittleEndian() != sys::IsLittleEndianHost)
657       sys::swapByteOrder(f);
658     uint32_t l;
659     memcpy(&l, sect + i, sizeof(uint32_t));
660     if (O->isLittleEndian() != sys::IsLittleEndianHost)
661       sys::swapByteOrder(l);
662     DumpLiteral4(l, f);
663   }
664 }
665 
DumpLiteral8(MachOObjectFile * O,uint32_t l0,uint32_t l1,double d)666 static void DumpLiteral8(MachOObjectFile *O, uint32_t l0, uint32_t l1,
667                          double d) {
668   outs() << format("0x%08" PRIx32, l0) << " " << format("0x%08" PRIx32, l1);
669   uint32_t Hi, Lo;
670   Hi = (O->isLittleEndian()) ? l1 : l0;
671   Lo = (O->isLittleEndian()) ? l0 : l1;
672 
673   // Hi is the high word, so this is equivalent to if(isfinite(d))
674   if ((Hi & 0x7ff00000) != 0x7ff00000)
675     outs() << format(" (%.16e)\n", d);
676   else {
677     if (Hi == 0x7ff00000 && Lo == 0)
678       outs() << " (+Infinity)\n";
679     else if (Hi == 0xfff00000 && Lo == 0)
680       outs() << " (-Infinity)\n";
681     else if ((Hi & 0x00080000) == 0x00080000)
682       outs() << " (non-signaling Not-a-Number)\n";
683     else
684       outs() << " (signaling Not-a-Number)\n";
685   }
686 }
687 
DumpLiteral8Section(MachOObjectFile * O,const char * sect,uint32_t sect_size,uint64_t sect_addr,bool print_addresses)688 static void DumpLiteral8Section(MachOObjectFile *O, const char *sect,
689                                 uint32_t sect_size, uint64_t sect_addr,
690                                 bool print_addresses) {
691   for (uint32_t i = 0; i < sect_size; i += sizeof(double)) {
692     if (print_addresses) {
693       if (O->is64Bit())
694         outs() << format("%016" PRIx64, sect_addr + i) << "  ";
695       else
696         outs() << format("%08" PRIx64, sect_addr + i) << "  ";
697     }
698     double d;
699     memcpy(&d, sect + i, sizeof(double));
700     if (O->isLittleEndian() != sys::IsLittleEndianHost)
701       sys::swapByteOrder(d);
702     uint32_t l0, l1;
703     memcpy(&l0, sect + i, sizeof(uint32_t));
704     memcpy(&l1, sect + i + sizeof(uint32_t), sizeof(uint32_t));
705     if (O->isLittleEndian() != sys::IsLittleEndianHost) {
706       sys::swapByteOrder(l0);
707       sys::swapByteOrder(l1);
708     }
709     DumpLiteral8(O, l0, l1, d);
710   }
711 }
712 
DumpLiteral16(uint32_t l0,uint32_t l1,uint32_t l2,uint32_t l3)713 static void DumpLiteral16(uint32_t l0, uint32_t l1, uint32_t l2, uint32_t l3) {
714   outs() << format("0x%08" PRIx32, l0) << " ";
715   outs() << format("0x%08" PRIx32, l1) << " ";
716   outs() << format("0x%08" PRIx32, l2) << " ";
717   outs() << format("0x%08" PRIx32, l3) << "\n";
718 }
719 
DumpLiteral16Section(MachOObjectFile * O,const char * sect,uint32_t sect_size,uint64_t sect_addr,bool print_addresses)720 static void DumpLiteral16Section(MachOObjectFile *O, const char *sect,
721                                  uint32_t sect_size, uint64_t sect_addr,
722                                  bool print_addresses) {
723   for (uint32_t i = 0; i < sect_size; i += 16) {
724     if (print_addresses) {
725       if (O->is64Bit())
726         outs() << format("%016" PRIx64, sect_addr + i) << "  ";
727       else
728         outs() << format("%08" PRIx64, sect_addr + i) << "  ";
729     }
730     uint32_t l0, l1, l2, l3;
731     memcpy(&l0, sect + i, sizeof(uint32_t));
732     memcpy(&l1, sect + i + sizeof(uint32_t), sizeof(uint32_t));
733     memcpy(&l2, sect + i + 2 * sizeof(uint32_t), sizeof(uint32_t));
734     memcpy(&l3, sect + i + 3 * sizeof(uint32_t), sizeof(uint32_t));
735     if (O->isLittleEndian() != sys::IsLittleEndianHost) {
736       sys::swapByteOrder(l0);
737       sys::swapByteOrder(l1);
738       sys::swapByteOrder(l2);
739       sys::swapByteOrder(l3);
740     }
741     DumpLiteral16(l0, l1, l2, l3);
742   }
743 }
744 
DumpLiteralPointerSection(MachOObjectFile * O,const SectionRef & Section,const char * sect,uint32_t sect_size,uint64_t sect_addr,bool print_addresses)745 static void DumpLiteralPointerSection(MachOObjectFile *O,
746                                       const SectionRef &Section,
747                                       const char *sect, uint32_t sect_size,
748                                       uint64_t sect_addr,
749                                       bool print_addresses) {
750   // Collect the literal sections in this Mach-O file.
751   std::vector<SectionRef> LiteralSections;
752   for (const SectionRef &Section : O->sections()) {
753     DataRefImpl Ref = Section.getRawDataRefImpl();
754     uint32_t section_type;
755     if (O->is64Bit()) {
756       const MachO::section_64 Sec = O->getSection64(Ref);
757       section_type = Sec.flags & MachO::SECTION_TYPE;
758     } else {
759       const MachO::section Sec = O->getSection(Ref);
760       section_type = Sec.flags & MachO::SECTION_TYPE;
761     }
762     if (section_type == MachO::S_CSTRING_LITERALS ||
763         section_type == MachO::S_4BYTE_LITERALS ||
764         section_type == MachO::S_8BYTE_LITERALS ||
765         section_type == MachO::S_16BYTE_LITERALS)
766       LiteralSections.push_back(Section);
767   }
768 
769   // Set the size of the literal pointer.
770   uint32_t lp_size = O->is64Bit() ? 8 : 4;
771 
772   // Collect the external relocation symbols for the literal pointers.
773   std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
774   for (const RelocationRef &Reloc : Section.relocations()) {
775     DataRefImpl Rel;
776     MachO::any_relocation_info RE;
777     bool isExtern = false;
778     Rel = Reloc.getRawDataRefImpl();
779     RE = O->getRelocation(Rel);
780     isExtern = O->getPlainRelocationExternal(RE);
781     if (isExtern) {
782       uint64_t RelocOffset = Reloc.getOffset();
783       symbol_iterator RelocSym = Reloc.getSymbol();
784       Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
785     }
786   }
787   array_pod_sort(Relocs.begin(), Relocs.end());
788 
789   // Dump each literal pointer.
790   for (uint32_t i = 0; i < sect_size; i += lp_size) {
791     if (print_addresses) {
792       if (O->is64Bit())
793         outs() << format("%016" PRIx64, sect_addr + i) << "  ";
794       else
795         outs() << format("%08" PRIx64, sect_addr + i) << "  ";
796     }
797     uint64_t lp;
798     if (O->is64Bit()) {
799       memcpy(&lp, sect + i, sizeof(uint64_t));
800       if (O->isLittleEndian() != sys::IsLittleEndianHost)
801         sys::swapByteOrder(lp);
802     } else {
803       uint32_t li;
804       memcpy(&li, sect + i, sizeof(uint32_t));
805       if (O->isLittleEndian() != sys::IsLittleEndianHost)
806         sys::swapByteOrder(li);
807       lp = li;
808     }
809 
810     // First look for an external relocation entry for this literal pointer.
811     auto Reloc = std::find_if(
812         Relocs.begin(), Relocs.end(),
813         [&](const std::pair<uint64_t, SymbolRef> &P) { return P.first == i; });
814     if (Reloc != Relocs.end()) {
815       symbol_iterator RelocSym = Reloc->second;
816       ErrorOr<StringRef> SymName = RelocSym->getName();
817       if (std::error_code EC = SymName.getError())
818         report_fatal_error(EC.message());
819       outs() << "external relocation entry for symbol:" << *SymName << "\n";
820       continue;
821     }
822 
823     // For local references see what the section the literal pointer points to.
824     auto Sect = std::find_if(LiteralSections.begin(), LiteralSections.end(),
825                              [&](const SectionRef &R) {
826                                return lp >= R.getAddress() &&
827                                       lp < R.getAddress() + R.getSize();
828                              });
829     if (Sect == LiteralSections.end()) {
830       outs() << format("0x%" PRIx64, lp) << " (not in a literal section)\n";
831       continue;
832     }
833 
834     uint64_t SectAddress = Sect->getAddress();
835     uint64_t SectSize = Sect->getSize();
836 
837     StringRef SectName;
838     Sect->getName(SectName);
839     DataRefImpl Ref = Sect->getRawDataRefImpl();
840     StringRef SegmentName = O->getSectionFinalSegmentName(Ref);
841     outs() << SegmentName << ":" << SectName << ":";
842 
843     uint32_t section_type;
844     if (O->is64Bit()) {
845       const MachO::section_64 Sec = O->getSection64(Ref);
846       section_type = Sec.flags & MachO::SECTION_TYPE;
847     } else {
848       const MachO::section Sec = O->getSection(Ref);
849       section_type = Sec.flags & MachO::SECTION_TYPE;
850     }
851 
852     StringRef BytesStr;
853     Sect->getContents(BytesStr);
854     const char *Contents = reinterpret_cast<const char *>(BytesStr.data());
855 
856     switch (section_type) {
857     case MachO::S_CSTRING_LITERALS:
858       for (uint64_t i = lp - SectAddress; i < SectSize && Contents[i] != '\0';
859            i++) {
860         DumpCstringChar(Contents[i]);
861       }
862       outs() << "\n";
863       break;
864     case MachO::S_4BYTE_LITERALS:
865       float f;
866       memcpy(&f, Contents + (lp - SectAddress), sizeof(float));
867       uint32_t l;
868       memcpy(&l, Contents + (lp - SectAddress), sizeof(uint32_t));
869       if (O->isLittleEndian() != sys::IsLittleEndianHost) {
870         sys::swapByteOrder(f);
871         sys::swapByteOrder(l);
872       }
873       DumpLiteral4(l, f);
874       break;
875     case MachO::S_8BYTE_LITERALS: {
876       double d;
877       memcpy(&d, Contents + (lp - SectAddress), sizeof(double));
878       uint32_t l0, l1;
879       memcpy(&l0, Contents + (lp - SectAddress), sizeof(uint32_t));
880       memcpy(&l1, Contents + (lp - SectAddress) + sizeof(uint32_t),
881              sizeof(uint32_t));
882       if (O->isLittleEndian() != sys::IsLittleEndianHost) {
883         sys::swapByteOrder(f);
884         sys::swapByteOrder(l0);
885         sys::swapByteOrder(l1);
886       }
887       DumpLiteral8(O, l0, l1, d);
888       break;
889     }
890     case MachO::S_16BYTE_LITERALS: {
891       uint32_t l0, l1, l2, l3;
892       memcpy(&l0, Contents + (lp - SectAddress), sizeof(uint32_t));
893       memcpy(&l1, Contents + (lp - SectAddress) + sizeof(uint32_t),
894              sizeof(uint32_t));
895       memcpy(&l2, Contents + (lp - SectAddress) + 2 * sizeof(uint32_t),
896              sizeof(uint32_t));
897       memcpy(&l3, Contents + (lp - SectAddress) + 3 * sizeof(uint32_t),
898              sizeof(uint32_t));
899       if (O->isLittleEndian() != sys::IsLittleEndianHost) {
900         sys::swapByteOrder(l0);
901         sys::swapByteOrder(l1);
902         sys::swapByteOrder(l2);
903         sys::swapByteOrder(l3);
904       }
905       DumpLiteral16(l0, l1, l2, l3);
906       break;
907     }
908     }
909   }
910 }
911 
DumpInitTermPointerSection(MachOObjectFile * O,const char * sect,uint32_t sect_size,uint64_t sect_addr,SymbolAddressMap * AddrMap,bool verbose)912 static void DumpInitTermPointerSection(MachOObjectFile *O, const char *sect,
913                                        uint32_t sect_size, uint64_t sect_addr,
914                                        SymbolAddressMap *AddrMap,
915                                        bool verbose) {
916   uint32_t stride;
917   stride = (O->is64Bit()) ? sizeof(uint64_t) : sizeof(uint32_t);
918   for (uint32_t i = 0; i < sect_size; i += stride) {
919     const char *SymbolName = nullptr;
920     if (O->is64Bit()) {
921       outs() << format("0x%016" PRIx64, sect_addr + i * stride) << " ";
922       uint64_t pointer_value;
923       memcpy(&pointer_value, sect + i, stride);
924       if (O->isLittleEndian() != sys::IsLittleEndianHost)
925         sys::swapByteOrder(pointer_value);
926       outs() << format("0x%016" PRIx64, pointer_value);
927       if (verbose)
928         SymbolName = GuessSymbolName(pointer_value, AddrMap);
929     } else {
930       outs() << format("0x%08" PRIx64, sect_addr + i * stride) << " ";
931       uint32_t pointer_value;
932       memcpy(&pointer_value, sect + i, stride);
933       if (O->isLittleEndian() != sys::IsLittleEndianHost)
934         sys::swapByteOrder(pointer_value);
935       outs() << format("0x%08" PRIx32, pointer_value);
936       if (verbose)
937         SymbolName = GuessSymbolName(pointer_value, AddrMap);
938     }
939     if (SymbolName)
940       outs() << " " << SymbolName;
941     outs() << "\n";
942   }
943 }
944 
DumpRawSectionContents(MachOObjectFile * O,const char * sect,uint32_t size,uint64_t addr)945 static void DumpRawSectionContents(MachOObjectFile *O, const char *sect,
946                                    uint32_t size, uint64_t addr) {
947   uint32_t cputype = O->getHeader().cputype;
948   if (cputype == MachO::CPU_TYPE_I386 || cputype == MachO::CPU_TYPE_X86_64) {
949     uint32_t j;
950     for (uint32_t i = 0; i < size; i += j, addr += j) {
951       if (O->is64Bit())
952         outs() << format("%016" PRIx64, addr) << "\t";
953       else
954         outs() << format("%08" PRIx64, addr) << "\t";
955       for (j = 0; j < 16 && i + j < size; j++) {
956         uint8_t byte_word = *(sect + i + j);
957         outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
958       }
959       outs() << "\n";
960     }
961   } else {
962     uint32_t j;
963     for (uint32_t i = 0; i < size; i += j, addr += j) {
964       if (O->is64Bit())
965         outs() << format("%016" PRIx64, addr) << "\t";
966       else
967         outs() << format("%08" PRIx64, sect) << "\t";
968       for (j = 0; j < 4 * sizeof(int32_t) && i + j < size;
969            j += sizeof(int32_t)) {
970         if (i + j + sizeof(int32_t) < size) {
971           uint32_t long_word;
972           memcpy(&long_word, sect + i + j, sizeof(int32_t));
973           if (O->isLittleEndian() != sys::IsLittleEndianHost)
974             sys::swapByteOrder(long_word);
975           outs() << format("%08" PRIx32, long_word) << " ";
976         } else {
977           for (uint32_t k = 0; i + j + k < size; k++) {
978             uint8_t byte_word = *(sect + i + j);
979             outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
980           }
981         }
982       }
983       outs() << "\n";
984     }
985   }
986 }
987 
988 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
989                              StringRef DisSegName, StringRef DisSectName);
990 static void DumpProtocolSection(MachOObjectFile *O, const char *sect,
991                                 uint32_t size, uint32_t addr);
992 
DumpSectionContents(StringRef Filename,MachOObjectFile * O,bool verbose)993 static void DumpSectionContents(StringRef Filename, MachOObjectFile *O,
994                                 bool verbose) {
995   SymbolAddressMap AddrMap;
996   if (verbose)
997     CreateSymbolAddressMap(O, &AddrMap);
998 
999   for (unsigned i = 0; i < FilterSections.size(); ++i) {
1000     StringRef DumpSection = FilterSections[i];
1001     std::pair<StringRef, StringRef> DumpSegSectName;
1002     DumpSegSectName = DumpSection.split(',');
1003     StringRef DumpSegName, DumpSectName;
1004     if (DumpSegSectName.second.size()) {
1005       DumpSegName = DumpSegSectName.first;
1006       DumpSectName = DumpSegSectName.second;
1007     } else {
1008       DumpSegName = "";
1009       DumpSectName = DumpSegSectName.first;
1010     }
1011     for (const SectionRef &Section : O->sections()) {
1012       StringRef SectName;
1013       Section.getName(SectName);
1014       DataRefImpl Ref = Section.getRawDataRefImpl();
1015       StringRef SegName = O->getSectionFinalSegmentName(Ref);
1016       if ((DumpSegName.empty() || SegName == DumpSegName) &&
1017           (SectName == DumpSectName)) {
1018 
1019         uint32_t section_flags;
1020         if (O->is64Bit()) {
1021           const MachO::section_64 Sec = O->getSection64(Ref);
1022           section_flags = Sec.flags;
1023 
1024         } else {
1025           const MachO::section Sec = O->getSection(Ref);
1026           section_flags = Sec.flags;
1027         }
1028         uint32_t section_type = section_flags & MachO::SECTION_TYPE;
1029 
1030         StringRef BytesStr;
1031         Section.getContents(BytesStr);
1032         const char *sect = reinterpret_cast<const char *>(BytesStr.data());
1033         uint32_t sect_size = BytesStr.size();
1034         uint64_t sect_addr = Section.getAddress();
1035 
1036         outs() << "Contents of (" << SegName << "," << SectName
1037                << ") section\n";
1038 
1039         if (verbose) {
1040           if ((section_flags & MachO::S_ATTR_PURE_INSTRUCTIONS) ||
1041               (section_flags & MachO::S_ATTR_SOME_INSTRUCTIONS)) {
1042             DisassembleMachO(Filename, O, SegName, SectName);
1043             continue;
1044           }
1045           if (SegName == "__TEXT" && SectName == "__info_plist") {
1046             outs() << sect;
1047             continue;
1048           }
1049           if (SegName == "__OBJC" && SectName == "__protocol") {
1050             DumpProtocolSection(O, sect, sect_size, sect_addr);
1051             continue;
1052           }
1053           switch (section_type) {
1054           case MachO::S_REGULAR:
1055             DumpRawSectionContents(O, sect, sect_size, sect_addr);
1056             break;
1057           case MachO::S_ZEROFILL:
1058             outs() << "zerofill section and has no contents in the file\n";
1059             break;
1060           case MachO::S_CSTRING_LITERALS:
1061             DumpCstringSection(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1062             break;
1063           case MachO::S_4BYTE_LITERALS:
1064             DumpLiteral4Section(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1065             break;
1066           case MachO::S_8BYTE_LITERALS:
1067             DumpLiteral8Section(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1068             break;
1069           case MachO::S_16BYTE_LITERALS:
1070             DumpLiteral16Section(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1071             break;
1072           case MachO::S_LITERAL_POINTERS:
1073             DumpLiteralPointerSection(O, Section, sect, sect_size, sect_addr,
1074                                       !NoLeadingAddr);
1075             break;
1076           case MachO::S_MOD_INIT_FUNC_POINTERS:
1077           case MachO::S_MOD_TERM_FUNC_POINTERS:
1078             DumpInitTermPointerSection(O, sect, sect_size, sect_addr, &AddrMap,
1079                                        verbose);
1080             break;
1081           default:
1082             outs() << "Unknown section type ("
1083                    << format("0x%08" PRIx32, section_type) << ")\n";
1084             DumpRawSectionContents(O, sect, sect_size, sect_addr);
1085             break;
1086           }
1087         } else {
1088           if (section_type == MachO::S_ZEROFILL)
1089             outs() << "zerofill section and has no contents in the file\n";
1090           else
1091             DumpRawSectionContents(O, sect, sect_size, sect_addr);
1092         }
1093       }
1094     }
1095   }
1096 }
1097 
DumpInfoPlistSectionContents(StringRef Filename,MachOObjectFile * O)1098 static void DumpInfoPlistSectionContents(StringRef Filename,
1099                                          MachOObjectFile *O) {
1100   for (const SectionRef &Section : O->sections()) {
1101     StringRef SectName;
1102     Section.getName(SectName);
1103     DataRefImpl Ref = Section.getRawDataRefImpl();
1104     StringRef SegName = O->getSectionFinalSegmentName(Ref);
1105     if (SegName == "__TEXT" && SectName == "__info_plist") {
1106       outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
1107       StringRef BytesStr;
1108       Section.getContents(BytesStr);
1109       const char *sect = reinterpret_cast<const char *>(BytesStr.data());
1110       outs() << sect;
1111       return;
1112     }
1113   }
1114 }
1115 
1116 // checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file
1117 // and if it is and there is a list of architecture flags is specified then
1118 // check to make sure this Mach-O file is one of those architectures or all
1119 // architectures were specified.  If not then an error is generated and this
1120 // routine returns false.  Else it returns true.
checkMachOAndArchFlags(ObjectFile * O,StringRef Filename)1121 static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) {
1122   if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) {
1123     MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O);
1124     bool ArchFound = false;
1125     MachO::mach_header H;
1126     MachO::mach_header_64 H_64;
1127     Triple T;
1128     if (MachO->is64Bit()) {
1129       H_64 = MachO->MachOObjectFile::getHeader64();
1130       T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype);
1131     } else {
1132       H = MachO->MachOObjectFile::getHeader();
1133       T = MachOObjectFile::getArch(H.cputype, H.cpusubtype);
1134     }
1135     unsigned i;
1136     for (i = 0; i < ArchFlags.size(); ++i) {
1137       if (ArchFlags[i] == T.getArchName())
1138         ArchFound = true;
1139       break;
1140     }
1141     if (!ArchFound) {
1142       errs() << "llvm-objdump: file: " + Filename + " does not contain "
1143              << "architecture: " + ArchFlags[i] + "\n";
1144       return false;
1145     }
1146   }
1147   return true;
1148 }
1149 
1150 static void printObjcMetaData(MachOObjectFile *O, bool verbose);
1151 
1152 // ProcessMachO() is passed a single opened Mach-O file, which may be an
1153 // archive member and or in a slice of a universal file.  It prints the
1154 // the file name and header info and then processes it according to the
1155 // command line options.
ProcessMachO(StringRef Filename,MachOObjectFile * MachOOF,StringRef ArchiveMemberName=StringRef (),StringRef ArchitectureName=StringRef ())1156 static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF,
1157                          StringRef ArchiveMemberName = StringRef(),
1158                          StringRef ArchitectureName = StringRef()) {
1159   // If we are doing some processing here on the Mach-O file print the header
1160   // info.  And don't print it otherwise like in the case of printing the
1161   // UniversalHeaders or ArchiveHeaders.
1162   if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind ||
1163       LazyBind || WeakBind || IndirectSymbols || DataInCode || LinkOptHints ||
1164       DylibsUsed || DylibId || ObjcMetaData || (FilterSections.size() != 0)) {
1165     outs() << Filename;
1166     if (!ArchiveMemberName.empty())
1167       outs() << '(' << ArchiveMemberName << ')';
1168     if (!ArchitectureName.empty())
1169       outs() << " (architecture " << ArchitectureName << ")";
1170     outs() << ":\n";
1171   }
1172 
1173   if (Disassemble)
1174     DisassembleMachO(Filename, MachOOF, "__TEXT", "__text");
1175   if (IndirectSymbols)
1176     PrintIndirectSymbols(MachOOF, !NonVerbose);
1177   if (DataInCode)
1178     PrintDataInCodeTable(MachOOF, !NonVerbose);
1179   if (LinkOptHints)
1180     PrintLinkOptHints(MachOOF);
1181   if (Relocations)
1182     PrintRelocations(MachOOF);
1183   if (SectionHeaders)
1184     PrintSectionHeaders(MachOOF);
1185   if (SectionContents)
1186     PrintSectionContents(MachOOF);
1187   if (FilterSections.size() != 0)
1188     DumpSectionContents(Filename, MachOOF, !NonVerbose);
1189   if (InfoPlist)
1190     DumpInfoPlistSectionContents(Filename, MachOOF);
1191   if (DylibsUsed)
1192     PrintDylibs(MachOOF, false);
1193   if (DylibId)
1194     PrintDylibs(MachOOF, true);
1195   if (SymbolTable)
1196     PrintSymbolTable(MachOOF);
1197   if (UnwindInfo)
1198     printMachOUnwindInfo(MachOOF);
1199   if (PrivateHeaders)
1200     printMachOFileHeader(MachOOF);
1201   if (ObjcMetaData)
1202     printObjcMetaData(MachOOF, !NonVerbose);
1203   if (ExportsTrie)
1204     printExportsTrie(MachOOF);
1205   if (Rebase)
1206     printRebaseTable(MachOOF);
1207   if (Bind)
1208     printBindTable(MachOOF);
1209   if (LazyBind)
1210     printLazyBindTable(MachOOF);
1211   if (WeakBind)
1212     printWeakBindTable(MachOOF);
1213 }
1214 
1215 // printUnknownCPUType() helps print_fat_headers for unknown CPU's.
printUnknownCPUType(uint32_t cputype,uint32_t cpusubtype)1216 static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) {
1217   outs() << "    cputype (" << cputype << ")\n";
1218   outs() << "    cpusubtype (" << cpusubtype << ")\n";
1219 }
1220 
1221 // printCPUType() helps print_fat_headers by printing the cputype and
1222 // pusubtype (symbolically for the one's it knows about).
printCPUType(uint32_t cputype,uint32_t cpusubtype)1223 static void printCPUType(uint32_t cputype, uint32_t cpusubtype) {
1224   switch (cputype) {
1225   case MachO::CPU_TYPE_I386:
1226     switch (cpusubtype) {
1227     case MachO::CPU_SUBTYPE_I386_ALL:
1228       outs() << "    cputype CPU_TYPE_I386\n";
1229       outs() << "    cpusubtype CPU_SUBTYPE_I386_ALL\n";
1230       break;
1231     default:
1232       printUnknownCPUType(cputype, cpusubtype);
1233       break;
1234     }
1235     break;
1236   case MachO::CPU_TYPE_X86_64:
1237     switch (cpusubtype) {
1238     case MachO::CPU_SUBTYPE_X86_64_ALL:
1239       outs() << "    cputype CPU_TYPE_X86_64\n";
1240       outs() << "    cpusubtype CPU_SUBTYPE_X86_64_ALL\n";
1241       break;
1242     case MachO::CPU_SUBTYPE_X86_64_H:
1243       outs() << "    cputype CPU_TYPE_X86_64\n";
1244       outs() << "    cpusubtype CPU_SUBTYPE_X86_64_H\n";
1245       break;
1246     default:
1247       printUnknownCPUType(cputype, cpusubtype);
1248       break;
1249     }
1250     break;
1251   case MachO::CPU_TYPE_ARM:
1252     switch (cpusubtype) {
1253     case MachO::CPU_SUBTYPE_ARM_ALL:
1254       outs() << "    cputype CPU_TYPE_ARM\n";
1255       outs() << "    cpusubtype CPU_SUBTYPE_ARM_ALL\n";
1256       break;
1257     case MachO::CPU_SUBTYPE_ARM_V4T:
1258       outs() << "    cputype CPU_TYPE_ARM\n";
1259       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V4T\n";
1260       break;
1261     case MachO::CPU_SUBTYPE_ARM_V5TEJ:
1262       outs() << "    cputype CPU_TYPE_ARM\n";
1263       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n";
1264       break;
1265     case MachO::CPU_SUBTYPE_ARM_XSCALE:
1266       outs() << "    cputype CPU_TYPE_ARM\n";
1267       outs() << "    cpusubtype CPU_SUBTYPE_ARM_XSCALE\n";
1268       break;
1269     case MachO::CPU_SUBTYPE_ARM_V6:
1270       outs() << "    cputype CPU_TYPE_ARM\n";
1271       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V6\n";
1272       break;
1273     case MachO::CPU_SUBTYPE_ARM_V6M:
1274       outs() << "    cputype CPU_TYPE_ARM\n";
1275       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V6M\n";
1276       break;
1277     case MachO::CPU_SUBTYPE_ARM_V7:
1278       outs() << "    cputype CPU_TYPE_ARM\n";
1279       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V7\n";
1280       break;
1281     case MachO::CPU_SUBTYPE_ARM_V7EM:
1282       outs() << "    cputype CPU_TYPE_ARM\n";
1283       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V7EM\n";
1284       break;
1285     case MachO::CPU_SUBTYPE_ARM_V7K:
1286       outs() << "    cputype CPU_TYPE_ARM\n";
1287       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V7K\n";
1288       break;
1289     case MachO::CPU_SUBTYPE_ARM_V7M:
1290       outs() << "    cputype CPU_TYPE_ARM\n";
1291       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V7M\n";
1292       break;
1293     case MachO::CPU_SUBTYPE_ARM_V7S:
1294       outs() << "    cputype CPU_TYPE_ARM\n";
1295       outs() << "    cpusubtype CPU_SUBTYPE_ARM_V7S\n";
1296       break;
1297     default:
1298       printUnknownCPUType(cputype, cpusubtype);
1299       break;
1300     }
1301     break;
1302   case MachO::CPU_TYPE_ARM64:
1303     switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
1304     case MachO::CPU_SUBTYPE_ARM64_ALL:
1305       outs() << "    cputype CPU_TYPE_ARM64\n";
1306       outs() << "    cpusubtype CPU_SUBTYPE_ARM64_ALL\n";
1307       break;
1308     default:
1309       printUnknownCPUType(cputype, cpusubtype);
1310       break;
1311     }
1312     break;
1313   default:
1314     printUnknownCPUType(cputype, cpusubtype);
1315     break;
1316   }
1317 }
1318 
printMachOUniversalHeaders(const object::MachOUniversalBinary * UB,bool verbose)1319 static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB,
1320                                        bool verbose) {
1321   outs() << "Fat headers\n";
1322   if (verbose)
1323     outs() << "fat_magic FAT_MAGIC\n";
1324   else
1325     outs() << "fat_magic " << format("0x%" PRIx32, MachO::FAT_MAGIC) << "\n";
1326 
1327   uint32_t nfat_arch = UB->getNumberOfObjects();
1328   StringRef Buf = UB->getData();
1329   uint64_t size = Buf.size();
1330   uint64_t big_size = sizeof(struct MachO::fat_header) +
1331                       nfat_arch * sizeof(struct MachO::fat_arch);
1332   outs() << "nfat_arch " << UB->getNumberOfObjects();
1333   if (nfat_arch == 0)
1334     outs() << " (malformed, contains zero architecture types)\n";
1335   else if (big_size > size)
1336     outs() << " (malformed, architectures past end of file)\n";
1337   else
1338     outs() << "\n";
1339 
1340   for (uint32_t i = 0; i < nfat_arch; ++i) {
1341     MachOUniversalBinary::ObjectForArch OFA(UB, i);
1342     uint32_t cputype = OFA.getCPUType();
1343     uint32_t cpusubtype = OFA.getCPUSubType();
1344     outs() << "architecture ";
1345     for (uint32_t j = 0; i != 0 && j <= i - 1; j++) {
1346       MachOUniversalBinary::ObjectForArch other_OFA(UB, j);
1347       uint32_t other_cputype = other_OFA.getCPUType();
1348       uint32_t other_cpusubtype = other_OFA.getCPUSubType();
1349       if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype &&
1350           (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) ==
1351               (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) {
1352         outs() << "(illegal duplicate architecture) ";
1353         break;
1354       }
1355     }
1356     if (verbose) {
1357       outs() << OFA.getArchTypeName() << "\n";
1358       printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
1359     } else {
1360       outs() << i << "\n";
1361       outs() << "    cputype " << cputype << "\n";
1362       outs() << "    cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK)
1363              << "\n";
1364     }
1365     if (verbose &&
1366         (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64)
1367       outs() << "    capabilities CPU_SUBTYPE_LIB64\n";
1368     else
1369       outs() << "    capabilities "
1370              << format("0x%" PRIx32,
1371                        (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n";
1372     outs() << "    offset " << OFA.getOffset();
1373     if (OFA.getOffset() > size)
1374       outs() << " (past end of file)";
1375     if (OFA.getOffset() % (1 << OFA.getAlign()) != 0)
1376       outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")";
1377     outs() << "\n";
1378     outs() << "    size " << OFA.getSize();
1379     big_size = OFA.getOffset() + OFA.getSize();
1380     if (big_size > size)
1381       outs() << " (past end of file)";
1382     outs() << "\n";
1383     outs() << "    align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign())
1384            << ")\n";
1385   }
1386 }
1387 
printArchiveChild(const Archive::Child & C,bool verbose,bool print_offset)1388 static void printArchiveChild(const Archive::Child &C, bool verbose,
1389                               bool print_offset) {
1390   if (print_offset)
1391     outs() << C.getChildOffset() << "\t";
1392   sys::fs::perms Mode = C.getAccessMode();
1393   if (verbose) {
1394     // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG.
1395     // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG.
1396     outs() << "-";
1397     outs() << ((Mode & sys::fs::owner_read) ? "r" : "-");
1398     outs() << ((Mode & sys::fs::owner_write) ? "w" : "-");
1399     outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-");
1400     outs() << ((Mode & sys::fs::group_read) ? "r" : "-");
1401     outs() << ((Mode & sys::fs::group_write) ? "w" : "-");
1402     outs() << ((Mode & sys::fs::group_exe) ? "x" : "-");
1403     outs() << ((Mode & sys::fs::others_read) ? "r" : "-");
1404     outs() << ((Mode & sys::fs::others_write) ? "w" : "-");
1405     outs() << ((Mode & sys::fs::others_exe) ? "x" : "-");
1406   } else {
1407     outs() << format("0%o ", Mode);
1408   }
1409 
1410   unsigned UID = C.getUID();
1411   outs() << format("%3d/", UID);
1412   unsigned GID = C.getGID();
1413   outs() << format("%-3d ", GID);
1414   ErrorOr<uint64_t> Size = C.getRawSize();
1415   if (std::error_code EC = Size.getError())
1416     report_fatal_error(EC.message());
1417   outs() << format("%5" PRId64, Size.get()) << " ";
1418 
1419   StringRef RawLastModified = C.getRawLastModified();
1420   if (verbose) {
1421     unsigned Seconds;
1422     if (RawLastModified.getAsInteger(10, Seconds))
1423       outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified;
1424     else {
1425       // Since cime(3) returns a 26 character string of the form:
1426       // "Sun Sep 16 01:03:52 1973\n\0"
1427       // just print 24 characters.
1428       time_t t = Seconds;
1429       outs() << format("%.24s ", ctime(&t));
1430     }
1431   } else {
1432     outs() << RawLastModified << " ";
1433   }
1434 
1435   if (verbose) {
1436     ErrorOr<StringRef> NameOrErr = C.getName();
1437     if (NameOrErr.getError()) {
1438       StringRef RawName = C.getRawName();
1439       outs() << RawName << "\n";
1440     } else {
1441       StringRef Name = NameOrErr.get();
1442       outs() << Name << "\n";
1443     }
1444   } else {
1445     StringRef RawName = C.getRawName();
1446     outs() << RawName << "\n";
1447   }
1448 }
1449 
printArchiveHeaders(Archive * A,bool verbose,bool print_offset)1450 static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) {
1451   for (Archive::child_iterator I = A->child_begin(false), E = A->child_end();
1452        I != E; ++I) {
1453     if (std::error_code EC = I->getError())
1454       report_fatal_error(EC.message());
1455     const Archive::Child &C = **I;
1456     printArchiveChild(C, verbose, print_offset);
1457   }
1458 }
1459 
1460 // ParseInputMachO() parses the named Mach-O file in Filename and handles the
1461 // -arch flags selecting just those slices as specified by them and also parses
1462 // archive files.  Then for each individual Mach-O file ProcessMachO() is
1463 // called to process the file based on the command line options.
ParseInputMachO(StringRef Filename)1464 void llvm::ParseInputMachO(StringRef Filename) {
1465   // Check for -arch all and verifiy the -arch flags are valid.
1466   for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1467     if (ArchFlags[i] == "all") {
1468       ArchAll = true;
1469     } else {
1470       if (!MachOObjectFile::isValidArch(ArchFlags[i])) {
1471         errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] +
1472                       "'for the -arch option\n";
1473         return;
1474       }
1475     }
1476   }
1477 
1478   // Attempt to open the binary.
1479   ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename);
1480   if (std::error_code EC = BinaryOrErr.getError()) {
1481     errs() << "llvm-objdump: '" << Filename << "': " << EC.message() << ".\n";
1482     return;
1483   }
1484   Binary &Bin = *BinaryOrErr.get().getBinary();
1485 
1486   if (Archive *A = dyn_cast<Archive>(&Bin)) {
1487     outs() << "Archive : " << Filename << "\n";
1488     if (ArchiveHeaders)
1489       printArchiveHeaders(A, !NonVerbose, ArchiveMemberOffsets);
1490     for (Archive::child_iterator I = A->child_begin(), E = A->child_end();
1491          I != E; ++I) {
1492       if (std::error_code EC = I->getError())
1493         report_error(Filename, EC);
1494       auto &C = I->get();
1495       ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
1496       if (ChildOrErr.getError())
1497         continue;
1498       if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1499         if (!checkMachOAndArchFlags(O, Filename))
1500           return;
1501         ProcessMachO(Filename, O, O->getFileName());
1502       }
1503     }
1504     return;
1505   }
1506   if (UniversalHeaders) {
1507     if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin))
1508       printMachOUniversalHeaders(UB, !NonVerbose);
1509   }
1510   if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) {
1511     // If we have a list of architecture flags specified dump only those.
1512     if (!ArchAll && ArchFlags.size() != 0) {
1513       // Look for a slice in the universal binary that matches each ArchFlag.
1514       bool ArchFound;
1515       for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1516         ArchFound = false;
1517         for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1518                                                    E = UB->end_objects();
1519              I != E; ++I) {
1520           if (ArchFlags[i] == I->getArchTypeName()) {
1521             ArchFound = true;
1522             ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
1523                 I->getAsObjectFile();
1524             std::string ArchitectureName = "";
1525             if (ArchFlags.size() > 1)
1526               ArchitectureName = I->getArchTypeName();
1527             if (ObjOrErr) {
1528               ObjectFile &O = *ObjOrErr.get();
1529               if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1530                 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1531             } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1532                            I->getAsArchive()) {
1533               std::unique_ptr<Archive> &A = *AOrErr;
1534               outs() << "Archive : " << Filename;
1535               if (!ArchitectureName.empty())
1536                 outs() << " (architecture " << ArchitectureName << ")";
1537               outs() << "\n";
1538               if (ArchiveHeaders)
1539                 printArchiveHeaders(A.get(), !NonVerbose, ArchiveMemberOffsets);
1540               for (Archive::child_iterator AI = A->child_begin(),
1541                                            AE = A->child_end();
1542                    AI != AE; ++AI) {
1543                 if (std::error_code EC = AI->getError())
1544                   report_error(Filename, EC);
1545                 auto &C = AI->get();
1546                 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
1547                 if (ChildOrErr.getError())
1548                   continue;
1549                 if (MachOObjectFile *O =
1550                         dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1551                   ProcessMachO(Filename, O, O->getFileName(), ArchitectureName);
1552               }
1553             }
1554           }
1555         }
1556         if (!ArchFound) {
1557           errs() << "llvm-objdump: file: " + Filename + " does not contain "
1558                  << "architecture: " + ArchFlags[i] + "\n";
1559           return;
1560         }
1561       }
1562       return;
1563     }
1564     // No architecture flags were specified so if this contains a slice that
1565     // matches the host architecture dump only that.
1566     if (!ArchAll) {
1567       for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1568                                                  E = UB->end_objects();
1569            I != E; ++I) {
1570         if (MachOObjectFile::getHostArch().getArchName() ==
1571             I->getArchTypeName()) {
1572           ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1573           std::string ArchiveName;
1574           ArchiveName.clear();
1575           if (ObjOrErr) {
1576             ObjectFile &O = *ObjOrErr.get();
1577             if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1578               ProcessMachO(Filename, MachOOF);
1579           } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1580                          I->getAsArchive()) {
1581             std::unique_ptr<Archive> &A = *AOrErr;
1582             outs() << "Archive : " << Filename << "\n";
1583             if (ArchiveHeaders)
1584               printArchiveHeaders(A.get(), !NonVerbose, ArchiveMemberOffsets);
1585             for (Archive::child_iterator AI = A->child_begin(),
1586                                          AE = A->child_end();
1587                  AI != AE; ++AI) {
1588               if (std::error_code EC = AI->getError())
1589                 report_error(Filename, EC);
1590               auto &C = AI->get();
1591               ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
1592               if (ChildOrErr.getError())
1593                 continue;
1594               if (MachOObjectFile *O =
1595                       dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1596                 ProcessMachO(Filename, O, O->getFileName());
1597             }
1598           }
1599           return;
1600         }
1601       }
1602     }
1603     // Either all architectures have been specified or none have been specified
1604     // and this does not contain the host architecture so dump all the slices.
1605     bool moreThanOneArch = UB->getNumberOfObjects() > 1;
1606     for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1607                                                E = UB->end_objects();
1608          I != E; ++I) {
1609       ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1610       std::string ArchitectureName = "";
1611       if (moreThanOneArch)
1612         ArchitectureName = I->getArchTypeName();
1613       if (ObjOrErr) {
1614         ObjectFile &Obj = *ObjOrErr.get();
1615         if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj))
1616           ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1617       } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) {
1618         std::unique_ptr<Archive> &A = *AOrErr;
1619         outs() << "Archive : " << Filename;
1620         if (!ArchitectureName.empty())
1621           outs() << " (architecture " << ArchitectureName << ")";
1622         outs() << "\n";
1623         if (ArchiveHeaders)
1624           printArchiveHeaders(A.get(), !NonVerbose, ArchiveMemberOffsets);
1625         for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end();
1626              AI != AE; ++AI) {
1627           if (std::error_code EC = AI->getError())
1628             report_error(Filename, EC);
1629           auto &C = AI->get();
1630           ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary();
1631           if (ChildOrErr.getError())
1632             continue;
1633           if (MachOObjectFile *O =
1634                   dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1635             if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O))
1636               ProcessMachO(Filename, MachOOF, MachOOF->getFileName(),
1637                            ArchitectureName);
1638           }
1639         }
1640       }
1641     }
1642     return;
1643   }
1644   if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) {
1645     if (!checkMachOAndArchFlags(O, Filename))
1646       return;
1647     if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) {
1648       ProcessMachO(Filename, MachOOF);
1649     } else
1650       errs() << "llvm-objdump: '" << Filename << "': "
1651              << "Object is not a Mach-O file type.\n";
1652   } else
1653     report_error(Filename, object_error::invalid_file_type);
1654 }
1655 
1656 typedef std::pair<uint64_t, const char *> BindInfoEntry;
1657 typedef std::vector<BindInfoEntry> BindTable;
1658 typedef BindTable::iterator bind_table_iterator;
1659 
1660 // The block of info used by the Symbolizer call backs.
1661 struct DisassembleInfo {
1662   bool verbose;
1663   MachOObjectFile *O;
1664   SectionRef S;
1665   SymbolAddressMap *AddrMap;
1666   std::vector<SectionRef> *Sections;
1667   const char *class_name;
1668   const char *selector_name;
1669   char *method;
1670   char *demangled_name;
1671   uint64_t adrp_addr;
1672   uint32_t adrp_inst;
1673   BindTable *bindtable;
1674   uint32_t depth;
1675 };
1676 
1677 // SymbolizerGetOpInfo() is the operand information call back function.
1678 // This is called to get the symbolic information for operand(s) of an
1679 // instruction when it is being done.  This routine does this from
1680 // the relocation information, symbol table, etc. That block of information
1681 // is a pointer to the struct DisassembleInfo that was passed when the
1682 // disassembler context was created and passed to back to here when
1683 // called back by the disassembler for instruction operands that could have
1684 // relocation information. The address of the instruction containing operand is
1685 // at the Pc parameter.  The immediate value the operand has is passed in
1686 // op_info->Value and is at Offset past the start of the instruction and has a
1687 // byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
1688 // LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
1689 // names and addends of the symbolic expression to add for the operand.  The
1690 // value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
1691 // information is returned then this function returns 1 else it returns 0.
SymbolizerGetOpInfo(void * DisInfo,uint64_t Pc,uint64_t Offset,uint64_t Size,int TagType,void * TagBuf)1692 static int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
1693                                uint64_t Size, int TagType, void *TagBuf) {
1694   struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1695   struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
1696   uint64_t value = op_info->Value;
1697 
1698   // Make sure all fields returned are zero if we don't set them.
1699   memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
1700   op_info->Value = value;
1701 
1702   // If the TagType is not the value 1 which it code knows about or if no
1703   // verbose symbolic information is wanted then just return 0, indicating no
1704   // information is being returned.
1705   if (TagType != 1 || !info->verbose)
1706     return 0;
1707 
1708   unsigned int Arch = info->O->getArch();
1709   if (Arch == Triple::x86) {
1710     if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1711       return 0;
1712     if (info->O->getHeader().filetype != MachO::MH_OBJECT) {
1713       // TODO:
1714       // Search the external relocation entries of a fully linked image
1715       // (if any) for an entry that matches this segment offset.
1716       // uint32_t seg_offset = (Pc + Offset);
1717       return 0;
1718     }
1719     // In MH_OBJECT filetypes search the section's relocation entries (if any)
1720     // for an entry for this section offset.
1721     uint32_t sect_addr = info->S.getAddress();
1722     uint32_t sect_offset = (Pc + Offset) - sect_addr;
1723     bool reloc_found = false;
1724     DataRefImpl Rel;
1725     MachO::any_relocation_info RE;
1726     bool isExtern = false;
1727     SymbolRef Symbol;
1728     bool r_scattered = false;
1729     uint32_t r_value, pair_r_value, r_type;
1730     for (const RelocationRef &Reloc : info->S.relocations()) {
1731       uint64_t RelocOffset = Reloc.getOffset();
1732       if (RelocOffset == sect_offset) {
1733         Rel = Reloc.getRawDataRefImpl();
1734         RE = info->O->getRelocation(Rel);
1735         r_type = info->O->getAnyRelocationType(RE);
1736         r_scattered = info->O->isRelocationScattered(RE);
1737         if (r_scattered) {
1738           r_value = info->O->getScatteredRelocationValue(RE);
1739           if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1740               r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
1741             DataRefImpl RelNext = Rel;
1742             info->O->moveRelocationNext(RelNext);
1743             MachO::any_relocation_info RENext;
1744             RENext = info->O->getRelocation(RelNext);
1745             if (info->O->isRelocationScattered(RENext))
1746               pair_r_value = info->O->getScatteredRelocationValue(RENext);
1747             else
1748               return 0;
1749           }
1750         } else {
1751           isExtern = info->O->getPlainRelocationExternal(RE);
1752           if (isExtern) {
1753             symbol_iterator RelocSym = Reloc.getSymbol();
1754             Symbol = *RelocSym;
1755           }
1756         }
1757         reloc_found = true;
1758         break;
1759       }
1760     }
1761     if (reloc_found && isExtern) {
1762       ErrorOr<StringRef> SymName = Symbol.getName();
1763       if (std::error_code EC = SymName.getError())
1764         report_fatal_error(EC.message());
1765       const char *name = SymName->data();
1766       op_info->AddSymbol.Present = 1;
1767       op_info->AddSymbol.Name = name;
1768       // For i386 extern relocation entries the value in the instruction is
1769       // the offset from the symbol, and value is already set in op_info->Value.
1770       return 1;
1771     }
1772     if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1773                         r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
1774       const char *add = GuessSymbolName(r_value, info->AddrMap);
1775       const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1776       uint32_t offset = value - (r_value - pair_r_value);
1777       op_info->AddSymbol.Present = 1;
1778       if (add != nullptr)
1779         op_info->AddSymbol.Name = add;
1780       else
1781         op_info->AddSymbol.Value = r_value;
1782       op_info->SubtractSymbol.Present = 1;
1783       if (sub != nullptr)
1784         op_info->SubtractSymbol.Name = sub;
1785       else
1786         op_info->SubtractSymbol.Value = pair_r_value;
1787       op_info->Value = offset;
1788       return 1;
1789     }
1790     return 0;
1791   }
1792   if (Arch == Triple::x86_64) {
1793     if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1794       return 0;
1795     if (info->O->getHeader().filetype != MachO::MH_OBJECT) {
1796       // TODO:
1797       // Search the external relocation entries of a fully linked image
1798       // (if any) for an entry that matches this segment offset.
1799       // uint64_t seg_offset = (Pc + Offset);
1800       return 0;
1801     }
1802     // In MH_OBJECT filetypes search the section's relocation entries (if any)
1803     // for an entry for this section offset.
1804     uint64_t sect_addr = info->S.getAddress();
1805     uint64_t sect_offset = (Pc + Offset) - sect_addr;
1806     bool reloc_found = false;
1807     DataRefImpl Rel;
1808     MachO::any_relocation_info RE;
1809     bool isExtern = false;
1810     SymbolRef Symbol;
1811     for (const RelocationRef &Reloc : info->S.relocations()) {
1812       uint64_t RelocOffset = Reloc.getOffset();
1813       if (RelocOffset == sect_offset) {
1814         Rel = Reloc.getRawDataRefImpl();
1815         RE = info->O->getRelocation(Rel);
1816         // NOTE: Scattered relocations don't exist on x86_64.
1817         isExtern = info->O->getPlainRelocationExternal(RE);
1818         if (isExtern) {
1819           symbol_iterator RelocSym = Reloc.getSymbol();
1820           Symbol = *RelocSym;
1821         }
1822         reloc_found = true;
1823         break;
1824       }
1825     }
1826     if (reloc_found && isExtern) {
1827       // The Value passed in will be adjusted by the Pc if the instruction
1828       // adds the Pc.  But for x86_64 external relocation entries the Value
1829       // is the offset from the external symbol.
1830       if (info->O->getAnyRelocationPCRel(RE))
1831         op_info->Value -= Pc + Offset + Size;
1832       ErrorOr<StringRef> SymName = Symbol.getName();
1833       if (std::error_code EC = SymName.getError())
1834         report_fatal_error(EC.message());
1835       const char *name = SymName->data();
1836       unsigned Type = info->O->getAnyRelocationType(RE);
1837       if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
1838         DataRefImpl RelNext = Rel;
1839         info->O->moveRelocationNext(RelNext);
1840         MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1841         unsigned TypeNext = info->O->getAnyRelocationType(RENext);
1842         bool isExternNext = info->O->getPlainRelocationExternal(RENext);
1843         unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
1844         if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
1845           op_info->SubtractSymbol.Present = 1;
1846           op_info->SubtractSymbol.Name = name;
1847           symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
1848           Symbol = *RelocSymNext;
1849           ErrorOr<StringRef> SymNameNext = Symbol.getName();
1850           if (std::error_code EC = SymNameNext.getError())
1851             report_fatal_error(EC.message());
1852           name = SymNameNext->data();
1853         }
1854       }
1855       // TODO: add the VariantKinds to op_info->VariantKind for relocation types
1856       // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
1857       op_info->AddSymbol.Present = 1;
1858       op_info->AddSymbol.Name = name;
1859       return 1;
1860     }
1861     return 0;
1862   }
1863   if (Arch == Triple::arm) {
1864     if (Offset != 0 || (Size != 4 && Size != 2))
1865       return 0;
1866     if (info->O->getHeader().filetype != MachO::MH_OBJECT) {
1867       // TODO:
1868       // Search the external relocation entries of a fully linked image
1869       // (if any) for an entry that matches this segment offset.
1870       // uint32_t seg_offset = (Pc + Offset);
1871       return 0;
1872     }
1873     // In MH_OBJECT filetypes search the section's relocation entries (if any)
1874     // for an entry for this section offset.
1875     uint32_t sect_addr = info->S.getAddress();
1876     uint32_t sect_offset = (Pc + Offset) - sect_addr;
1877     DataRefImpl Rel;
1878     MachO::any_relocation_info RE;
1879     bool isExtern = false;
1880     SymbolRef Symbol;
1881     bool r_scattered = false;
1882     uint32_t r_value, pair_r_value, r_type, r_length, other_half;
1883     auto Reloc =
1884         std::find_if(info->S.relocations().begin(), info->S.relocations().end(),
1885                      [&](const RelocationRef &Reloc) {
1886                        uint64_t RelocOffset = Reloc.getOffset();
1887                        return RelocOffset == sect_offset;
1888                      });
1889 
1890     if (Reloc == info->S.relocations().end())
1891       return 0;
1892 
1893     Rel = Reloc->getRawDataRefImpl();
1894     RE = info->O->getRelocation(Rel);
1895     r_length = info->O->getAnyRelocationLength(RE);
1896     r_scattered = info->O->isRelocationScattered(RE);
1897     if (r_scattered) {
1898       r_value = info->O->getScatteredRelocationValue(RE);
1899       r_type = info->O->getScatteredRelocationType(RE);
1900     } else {
1901       r_type = info->O->getAnyRelocationType(RE);
1902       isExtern = info->O->getPlainRelocationExternal(RE);
1903       if (isExtern) {
1904         symbol_iterator RelocSym = Reloc->getSymbol();
1905         Symbol = *RelocSym;
1906       }
1907     }
1908     if (r_type == MachO::ARM_RELOC_HALF ||
1909         r_type == MachO::ARM_RELOC_SECTDIFF ||
1910         r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
1911         r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1912       DataRefImpl RelNext = Rel;
1913       info->O->moveRelocationNext(RelNext);
1914       MachO::any_relocation_info RENext;
1915       RENext = info->O->getRelocation(RelNext);
1916       other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
1917       if (info->O->isRelocationScattered(RENext))
1918         pair_r_value = info->O->getScatteredRelocationValue(RENext);
1919     }
1920 
1921     if (isExtern) {
1922       ErrorOr<StringRef> SymName = Symbol.getName();
1923       if (std::error_code EC = SymName.getError())
1924         report_fatal_error(EC.message());
1925       const char *name = SymName->data();
1926       op_info->AddSymbol.Present = 1;
1927       op_info->AddSymbol.Name = name;
1928       switch (r_type) {
1929       case MachO::ARM_RELOC_HALF:
1930         if ((r_length & 0x1) == 1) {
1931           op_info->Value = value << 16 | other_half;
1932           op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1933         } else {
1934           op_info->Value = other_half << 16 | value;
1935           op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1936         }
1937         break;
1938       default:
1939         break;
1940       }
1941       return 1;
1942     }
1943     // If we have a branch that is not an external relocation entry then
1944     // return 0 so the code in tryAddingSymbolicOperand() can use the
1945     // SymbolLookUp call back with the branch target address to look up the
1946     // symbol and possiblity add an annotation for a symbol stub.
1947     if (isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
1948                           r_type == MachO::ARM_THUMB_RELOC_BR22))
1949       return 0;
1950 
1951     uint32_t offset = 0;
1952     if (r_type == MachO::ARM_RELOC_HALF ||
1953         r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1954       if ((r_length & 0x1) == 1)
1955         value = value << 16 | other_half;
1956       else
1957         value = other_half << 16 | value;
1958     }
1959     if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
1960                         r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
1961       offset = value - r_value;
1962       value = r_value;
1963     }
1964 
1965     if (r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1966       if ((r_length & 0x1) == 1)
1967         op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1968       else
1969         op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1970       const char *add = GuessSymbolName(r_value, info->AddrMap);
1971       const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1972       int32_t offset = value - (r_value - pair_r_value);
1973       op_info->AddSymbol.Present = 1;
1974       if (add != nullptr)
1975         op_info->AddSymbol.Name = add;
1976       else
1977         op_info->AddSymbol.Value = r_value;
1978       op_info->SubtractSymbol.Present = 1;
1979       if (sub != nullptr)
1980         op_info->SubtractSymbol.Name = sub;
1981       else
1982         op_info->SubtractSymbol.Value = pair_r_value;
1983       op_info->Value = offset;
1984       return 1;
1985     }
1986 
1987     op_info->AddSymbol.Present = 1;
1988     op_info->Value = offset;
1989     if (r_type == MachO::ARM_RELOC_HALF) {
1990       if ((r_length & 0x1) == 1)
1991         op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1992       else
1993         op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1994     }
1995     const char *add = GuessSymbolName(value, info->AddrMap);
1996     if (add != nullptr) {
1997       op_info->AddSymbol.Name = add;
1998       return 1;
1999     }
2000     op_info->AddSymbol.Value = value;
2001     return 1;
2002   }
2003   if (Arch == Triple::aarch64) {
2004     if (Offset != 0 || Size != 4)
2005       return 0;
2006     if (info->O->getHeader().filetype != MachO::MH_OBJECT) {
2007       // TODO:
2008       // Search the external relocation entries of a fully linked image
2009       // (if any) for an entry that matches this segment offset.
2010       // uint64_t seg_offset = (Pc + Offset);
2011       return 0;
2012     }
2013     // In MH_OBJECT filetypes search the section's relocation entries (if any)
2014     // for an entry for this section offset.
2015     uint64_t sect_addr = info->S.getAddress();
2016     uint64_t sect_offset = (Pc + Offset) - sect_addr;
2017     auto Reloc =
2018         std::find_if(info->S.relocations().begin(), info->S.relocations().end(),
2019                      [&](const RelocationRef &Reloc) {
2020                        uint64_t RelocOffset = Reloc.getOffset();
2021                        return RelocOffset == sect_offset;
2022                      });
2023 
2024     if (Reloc == info->S.relocations().end())
2025       return 0;
2026 
2027     DataRefImpl Rel = Reloc->getRawDataRefImpl();
2028     MachO::any_relocation_info RE = info->O->getRelocation(Rel);
2029     uint32_t r_type = info->O->getAnyRelocationType(RE);
2030     if (r_type == MachO::ARM64_RELOC_ADDEND) {
2031       DataRefImpl RelNext = Rel;
2032       info->O->moveRelocationNext(RelNext);
2033       MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
2034       if (value == 0) {
2035         value = info->O->getPlainRelocationSymbolNum(RENext);
2036         op_info->Value = value;
2037       }
2038     }
2039     // NOTE: Scattered relocations don't exist on arm64.
2040     if (!info->O->getPlainRelocationExternal(RE))
2041       return 0;
2042     ErrorOr<StringRef> SymName = Reloc->getSymbol()->getName();
2043     if (std::error_code EC = SymName.getError())
2044       report_fatal_error(EC.message());
2045     const char *name = SymName->data();
2046     op_info->AddSymbol.Present = 1;
2047     op_info->AddSymbol.Name = name;
2048 
2049     switch (r_type) {
2050     case MachO::ARM64_RELOC_PAGE21:
2051       /* @page */
2052       op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE;
2053       break;
2054     case MachO::ARM64_RELOC_PAGEOFF12:
2055       /* @pageoff */
2056       op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF;
2057       break;
2058     case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
2059       /* @gotpage */
2060       op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE;
2061       break;
2062     case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
2063       /* @gotpageoff */
2064       op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF;
2065       break;
2066     case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
2067       /* @tvlppage is not implemented in llvm-mc */
2068       op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP;
2069       break;
2070     case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
2071       /* @tvlppageoff is not implemented in llvm-mc */
2072       op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF;
2073       break;
2074     default:
2075     case MachO::ARM64_RELOC_BRANCH26:
2076       op_info->VariantKind = LLVMDisassembler_VariantKind_None;
2077       break;
2078     }
2079     return 1;
2080   }
2081   return 0;
2082 }
2083 
2084 // GuessCstringPointer is passed the address of what might be a pointer to a
2085 // literal string in a cstring section.  If that address is in a cstring section
2086 // it returns a pointer to that string.  Else it returns nullptr.
GuessCstringPointer(uint64_t ReferenceValue,struct DisassembleInfo * info)2087 static const char *GuessCstringPointer(uint64_t ReferenceValue,
2088                                        struct DisassembleInfo *info) {
2089   for (const auto &Load : info->O->load_commands()) {
2090     if (Load.C.cmd == MachO::LC_SEGMENT_64) {
2091       MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
2092       for (unsigned J = 0; J < Seg.nsects; ++J) {
2093         MachO::section_64 Sec = info->O->getSection64(Load, J);
2094         uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2095         if (section_type == MachO::S_CSTRING_LITERALS &&
2096             ReferenceValue >= Sec.addr &&
2097             ReferenceValue < Sec.addr + Sec.size) {
2098           uint64_t sect_offset = ReferenceValue - Sec.addr;
2099           uint64_t object_offset = Sec.offset + sect_offset;
2100           StringRef MachOContents = info->O->getData();
2101           uint64_t object_size = MachOContents.size();
2102           const char *object_addr = (const char *)MachOContents.data();
2103           if (object_offset < object_size) {
2104             const char *name = object_addr + object_offset;
2105             return name;
2106           } else {
2107             return nullptr;
2108           }
2109         }
2110       }
2111     } else if (Load.C.cmd == MachO::LC_SEGMENT) {
2112       MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
2113       for (unsigned J = 0; J < Seg.nsects; ++J) {
2114         MachO::section Sec = info->O->getSection(Load, J);
2115         uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2116         if (section_type == MachO::S_CSTRING_LITERALS &&
2117             ReferenceValue >= Sec.addr &&
2118             ReferenceValue < Sec.addr + Sec.size) {
2119           uint64_t sect_offset = ReferenceValue - Sec.addr;
2120           uint64_t object_offset = Sec.offset + sect_offset;
2121           StringRef MachOContents = info->O->getData();
2122           uint64_t object_size = MachOContents.size();
2123           const char *object_addr = (const char *)MachOContents.data();
2124           if (object_offset < object_size) {
2125             const char *name = object_addr + object_offset;
2126             return name;
2127           } else {
2128             return nullptr;
2129           }
2130         }
2131       }
2132     }
2133   }
2134   return nullptr;
2135 }
2136 
2137 // GuessIndirectSymbol returns the name of the indirect symbol for the
2138 // ReferenceValue passed in or nullptr.  This is used when ReferenceValue maybe
2139 // an address of a symbol stub or a lazy or non-lazy pointer to associate the
2140 // symbol name being referenced by the stub or pointer.
GuessIndirectSymbol(uint64_t ReferenceValue,struct DisassembleInfo * info)2141 static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
2142                                        struct DisassembleInfo *info) {
2143   MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
2144   MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
2145   for (const auto &Load : info->O->load_commands()) {
2146     if (Load.C.cmd == MachO::LC_SEGMENT_64) {
2147       MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
2148       for (unsigned J = 0; J < Seg.nsects; ++J) {
2149         MachO::section_64 Sec = info->O->getSection64(Load, J);
2150         uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2151         if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
2152              section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
2153              section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
2154              section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
2155              section_type == MachO::S_SYMBOL_STUBS) &&
2156             ReferenceValue >= Sec.addr &&
2157             ReferenceValue < Sec.addr + Sec.size) {
2158           uint32_t stride;
2159           if (section_type == MachO::S_SYMBOL_STUBS)
2160             stride = Sec.reserved2;
2161           else
2162             stride = 8;
2163           if (stride == 0)
2164             return nullptr;
2165           uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
2166           if (index < Dysymtab.nindirectsyms) {
2167             uint32_t indirect_symbol =
2168                 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
2169             if (indirect_symbol < Symtab.nsyms) {
2170               symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
2171               SymbolRef Symbol = *Sym;
2172               ErrorOr<StringRef> SymName = Symbol.getName();
2173               if (std::error_code EC = SymName.getError())
2174                 report_fatal_error(EC.message());
2175               const char *name = SymName->data();
2176               return name;
2177             }
2178           }
2179         }
2180       }
2181     } else if (Load.C.cmd == MachO::LC_SEGMENT) {
2182       MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
2183       for (unsigned J = 0; J < Seg.nsects; ++J) {
2184         MachO::section Sec = info->O->getSection(Load, J);
2185         uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2186         if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
2187              section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
2188              section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
2189              section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
2190              section_type == MachO::S_SYMBOL_STUBS) &&
2191             ReferenceValue >= Sec.addr &&
2192             ReferenceValue < Sec.addr + Sec.size) {
2193           uint32_t stride;
2194           if (section_type == MachO::S_SYMBOL_STUBS)
2195             stride = Sec.reserved2;
2196           else
2197             stride = 4;
2198           if (stride == 0)
2199             return nullptr;
2200           uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
2201           if (index < Dysymtab.nindirectsyms) {
2202             uint32_t indirect_symbol =
2203                 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
2204             if (indirect_symbol < Symtab.nsyms) {
2205               symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
2206               SymbolRef Symbol = *Sym;
2207               ErrorOr<StringRef> SymName = Symbol.getName();
2208               if (std::error_code EC = SymName.getError())
2209                 report_fatal_error(EC.message());
2210               const char *name = SymName->data();
2211               return name;
2212             }
2213           }
2214         }
2215       }
2216     }
2217   }
2218   return nullptr;
2219 }
2220 
2221 // method_reference() is called passing it the ReferenceName that might be
2222 // a reference it to an Objective-C method call.  If so then it allocates and
2223 // assembles a method call string with the values last seen and saved in
2224 // the DisassembleInfo's class_name and selector_name fields.  This is saved
2225 // into the method field of the info and any previous string is free'ed.
2226 // Then the class_name field in the info is set to nullptr.  The method call
2227 // string is set into ReferenceName and ReferenceType is set to
2228 // LLVMDisassembler_ReferenceType_Out_Objc_Message.  If this not a method call
2229 // then both ReferenceType and ReferenceName are left unchanged.
method_reference(struct DisassembleInfo * info,uint64_t * ReferenceType,const char ** ReferenceName)2230 static void method_reference(struct DisassembleInfo *info,
2231                              uint64_t *ReferenceType,
2232                              const char **ReferenceName) {
2233   unsigned int Arch = info->O->getArch();
2234   if (*ReferenceName != nullptr) {
2235     if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
2236       if (info->selector_name != nullptr) {
2237         if (info->method != nullptr)
2238           free(info->method);
2239         if (info->class_name != nullptr) {
2240           info->method = (char *)malloc(5 + strlen(info->class_name) +
2241                                         strlen(info->selector_name));
2242           if (info->method != nullptr) {
2243             strcpy(info->method, "+[");
2244             strcat(info->method, info->class_name);
2245             strcat(info->method, " ");
2246             strcat(info->method, info->selector_name);
2247             strcat(info->method, "]");
2248             *ReferenceName = info->method;
2249             *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
2250           }
2251         } else {
2252           info->method = (char *)malloc(9 + strlen(info->selector_name));
2253           if (info->method != nullptr) {
2254             if (Arch == Triple::x86_64)
2255               strcpy(info->method, "-[%rdi ");
2256             else if (Arch == Triple::aarch64)
2257               strcpy(info->method, "-[x0 ");
2258             else
2259               strcpy(info->method, "-[r? ");
2260             strcat(info->method, info->selector_name);
2261             strcat(info->method, "]");
2262             *ReferenceName = info->method;
2263             *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
2264           }
2265         }
2266         info->class_name = nullptr;
2267       }
2268     } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
2269       if (info->selector_name != nullptr) {
2270         if (info->method != nullptr)
2271           free(info->method);
2272         info->method = (char *)malloc(17 + strlen(info->selector_name));
2273         if (info->method != nullptr) {
2274           if (Arch == Triple::x86_64)
2275             strcpy(info->method, "-[[%rdi super] ");
2276           else if (Arch == Triple::aarch64)
2277             strcpy(info->method, "-[[x0 super] ");
2278           else
2279             strcpy(info->method, "-[[r? super] ");
2280           strcat(info->method, info->selector_name);
2281           strcat(info->method, "]");
2282           *ReferenceName = info->method;
2283           *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
2284         }
2285         info->class_name = nullptr;
2286       }
2287     }
2288   }
2289 }
2290 
2291 // GuessPointerPointer() is passed the address of what might be a pointer to
2292 // a reference to an Objective-C class, selector, message ref or cfstring.
2293 // If so the value of the pointer is returned and one of the booleans are set
2294 // to true.  If not zero is returned and all the booleans are set to false.
GuessPointerPointer(uint64_t ReferenceValue,struct DisassembleInfo * info,bool & classref,bool & selref,bool & msgref,bool & cfstring)2295 static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
2296                                     struct DisassembleInfo *info,
2297                                     bool &classref, bool &selref, bool &msgref,
2298                                     bool &cfstring) {
2299   classref = false;
2300   selref = false;
2301   msgref = false;
2302   cfstring = false;
2303   for (const auto &Load : info->O->load_commands()) {
2304     if (Load.C.cmd == MachO::LC_SEGMENT_64) {
2305       MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
2306       for (unsigned J = 0; J < Seg.nsects; ++J) {
2307         MachO::section_64 Sec = info->O->getSection64(Load, J);
2308         if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
2309              strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
2310              strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
2311              strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
2312              strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
2313             ReferenceValue >= Sec.addr &&
2314             ReferenceValue < Sec.addr + Sec.size) {
2315           uint64_t sect_offset = ReferenceValue - Sec.addr;
2316           uint64_t object_offset = Sec.offset + sect_offset;
2317           StringRef MachOContents = info->O->getData();
2318           uint64_t object_size = MachOContents.size();
2319           const char *object_addr = (const char *)MachOContents.data();
2320           if (object_offset < object_size) {
2321             uint64_t pointer_value;
2322             memcpy(&pointer_value, object_addr + object_offset,
2323                    sizeof(uint64_t));
2324             if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2325               sys::swapByteOrder(pointer_value);
2326             if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
2327               selref = true;
2328             else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
2329                      strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
2330               classref = true;
2331             else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
2332                      ReferenceValue + 8 < Sec.addr + Sec.size) {
2333               msgref = true;
2334               memcpy(&pointer_value, object_addr + object_offset + 8,
2335                      sizeof(uint64_t));
2336               if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2337                 sys::swapByteOrder(pointer_value);
2338             } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
2339               cfstring = true;
2340             return pointer_value;
2341           } else {
2342             return 0;
2343           }
2344         }
2345       }
2346     }
2347     // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
2348   }
2349   return 0;
2350 }
2351 
2352 // get_pointer_64 returns a pointer to the bytes in the object file at the
2353 // Address from a section in the Mach-O file.  And indirectly returns the
2354 // offset into the section, number of bytes left in the section past the offset
2355 // and which section is was being referenced.  If the Address is not in a
2356 // section nullptr is returned.
get_pointer_64(uint64_t Address,uint32_t & offset,uint32_t & left,SectionRef & S,DisassembleInfo * info,bool objc_only=false)2357 static const char *get_pointer_64(uint64_t Address, uint32_t &offset,
2358                                   uint32_t &left, SectionRef &S,
2359                                   DisassembleInfo *info,
2360                                   bool objc_only = false) {
2361   offset = 0;
2362   left = 0;
2363   S = SectionRef();
2364   for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
2365     uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
2366     uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
2367     if (SectSize == 0)
2368       continue;
2369     if (objc_only) {
2370       StringRef SectName;
2371       ((*(info->Sections))[SectIdx]).getName(SectName);
2372       DataRefImpl Ref = ((*(info->Sections))[SectIdx]).getRawDataRefImpl();
2373       StringRef SegName = info->O->getSectionFinalSegmentName(Ref);
2374       if (SegName != "__OBJC" && SectName != "__cstring")
2375         continue;
2376     }
2377     if (Address >= SectAddress && Address < SectAddress + SectSize) {
2378       S = (*(info->Sections))[SectIdx];
2379       offset = Address - SectAddress;
2380       left = SectSize - offset;
2381       StringRef SectContents;
2382       ((*(info->Sections))[SectIdx]).getContents(SectContents);
2383       return SectContents.data() + offset;
2384     }
2385   }
2386   return nullptr;
2387 }
2388 
get_pointer_32(uint32_t Address,uint32_t & offset,uint32_t & left,SectionRef & S,DisassembleInfo * info,bool objc_only=false)2389 static const char *get_pointer_32(uint32_t Address, uint32_t &offset,
2390                                   uint32_t &left, SectionRef &S,
2391                                   DisassembleInfo *info,
2392                                   bool objc_only = false) {
2393   return get_pointer_64(Address, offset, left, S, info, objc_only);
2394 }
2395 
2396 // get_symbol_64() returns the name of a symbol (or nullptr) and the address of
2397 // the symbol indirectly through n_value. Based on the relocation information
2398 // for the specified section offset in the specified section reference.
2399 // If no relocation information is found and a non-zero ReferenceValue for the
2400 // symbol is passed, look up that address in the info's AddrMap.
get_symbol_64(uint32_t sect_offset,SectionRef S,DisassembleInfo * info,uint64_t & n_value,uint64_t ReferenceValue=0)2401 static const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
2402                                  DisassembleInfo *info, uint64_t &n_value,
2403                                  uint64_t ReferenceValue = 0) {
2404   n_value = 0;
2405   if (!info->verbose)
2406     return nullptr;
2407 
2408   // See if there is an external relocation entry at the sect_offset.
2409   bool reloc_found = false;
2410   DataRefImpl Rel;
2411   MachO::any_relocation_info RE;
2412   bool isExtern = false;
2413   SymbolRef Symbol;
2414   for (const RelocationRef &Reloc : S.relocations()) {
2415     uint64_t RelocOffset = Reloc.getOffset();
2416     if (RelocOffset == sect_offset) {
2417       Rel = Reloc.getRawDataRefImpl();
2418       RE = info->O->getRelocation(Rel);
2419       if (info->O->isRelocationScattered(RE))
2420         continue;
2421       isExtern = info->O->getPlainRelocationExternal(RE);
2422       if (isExtern) {
2423         symbol_iterator RelocSym = Reloc.getSymbol();
2424         Symbol = *RelocSym;
2425       }
2426       reloc_found = true;
2427       break;
2428     }
2429   }
2430   // If there is an external relocation entry for a symbol in this section
2431   // at this section_offset then use that symbol's value for the n_value
2432   // and return its name.
2433   const char *SymbolName = nullptr;
2434   if (reloc_found && isExtern) {
2435     n_value = Symbol.getValue();
2436     ErrorOr<StringRef> NameOrError = Symbol.getName();
2437     if (std::error_code EC = NameOrError.getError())
2438       report_fatal_error(EC.message());
2439     StringRef Name = *NameOrError;
2440     if (!Name.empty()) {
2441       SymbolName = Name.data();
2442       return SymbolName;
2443     }
2444   }
2445 
2446   // TODO: For fully linked images, look through the external relocation
2447   // entries off the dynamic symtab command. For these the r_offset is from the
2448   // start of the first writeable segment in the Mach-O file.  So the offset
2449   // to this section from that segment is passed to this routine by the caller,
2450   // as the database_offset. Which is the difference of the section's starting
2451   // address and the first writable segment.
2452   //
2453   // NOTE: need add passing the database_offset to this routine.
2454 
2455   // We did not find an external relocation entry so look up the ReferenceValue
2456   // as an address of a symbol and if found return that symbol's name.
2457   SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2458 
2459   return SymbolName;
2460 }
2461 
get_symbol_32(uint32_t sect_offset,SectionRef S,DisassembleInfo * info,uint32_t ReferenceValue)2462 static const char *get_symbol_32(uint32_t sect_offset, SectionRef S,
2463                                  DisassembleInfo *info,
2464                                  uint32_t ReferenceValue) {
2465   uint64_t n_value64;
2466   return get_symbol_64(sect_offset, S, info, n_value64, ReferenceValue);
2467 }
2468 
2469 // These are structs in the Objective-C meta data and read to produce the
2470 // comments for disassembly.  While these are part of the ABI they are no
2471 // public defintions.  So the are here not in include/llvm/Support/MachO.h .
2472 
2473 // The cfstring object in a 64-bit Mach-O file.
2474 struct cfstring64_t {
2475   uint64_t isa;        // class64_t * (64-bit pointer)
2476   uint64_t flags;      // flag bits
2477   uint64_t characters; // char * (64-bit pointer)
2478   uint64_t length;     // number of non-NULL characters in above
2479 };
2480 
2481 // The class object in a 64-bit Mach-O file.
2482 struct class64_t {
2483   uint64_t isa;        // class64_t * (64-bit pointer)
2484   uint64_t superclass; // class64_t * (64-bit pointer)
2485   uint64_t cache;      // Cache (64-bit pointer)
2486   uint64_t vtable;     // IMP * (64-bit pointer)
2487   uint64_t data;       // class_ro64_t * (64-bit pointer)
2488 };
2489 
2490 struct class32_t {
2491   uint32_t isa;        /* class32_t * (32-bit pointer) */
2492   uint32_t superclass; /* class32_t * (32-bit pointer) */
2493   uint32_t cache;      /* Cache (32-bit pointer) */
2494   uint32_t vtable;     /* IMP * (32-bit pointer) */
2495   uint32_t data;       /* class_ro32_t * (32-bit pointer) */
2496 };
2497 
2498 struct class_ro64_t {
2499   uint32_t flags;
2500   uint32_t instanceStart;
2501   uint32_t instanceSize;
2502   uint32_t reserved;
2503   uint64_t ivarLayout;     // const uint8_t * (64-bit pointer)
2504   uint64_t name;           // const char * (64-bit pointer)
2505   uint64_t baseMethods;    // const method_list_t * (64-bit pointer)
2506   uint64_t baseProtocols;  // const protocol_list_t * (64-bit pointer)
2507   uint64_t ivars;          // const ivar_list_t * (64-bit pointer)
2508   uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
2509   uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
2510 };
2511 
2512 struct class_ro32_t {
2513   uint32_t flags;
2514   uint32_t instanceStart;
2515   uint32_t instanceSize;
2516   uint32_t ivarLayout;     /* const uint8_t * (32-bit pointer) */
2517   uint32_t name;           /* const char * (32-bit pointer) */
2518   uint32_t baseMethods;    /* const method_list_t * (32-bit pointer) */
2519   uint32_t baseProtocols;  /* const protocol_list_t * (32-bit pointer) */
2520   uint32_t ivars;          /* const ivar_list_t * (32-bit pointer) */
2521   uint32_t weakIvarLayout; /* const uint8_t * (32-bit pointer) */
2522   uint32_t baseProperties; /* const struct objc_property_list *
2523                                                    (32-bit pointer) */
2524 };
2525 
2526 /* Values for class_ro{64,32}_t->flags */
2527 #define RO_META (1 << 0)
2528 #define RO_ROOT (1 << 1)
2529 #define RO_HAS_CXX_STRUCTORS (1 << 2)
2530 
2531 struct method_list64_t {
2532   uint32_t entsize;
2533   uint32_t count;
2534   /* struct method64_t first;  These structures follow inline */
2535 };
2536 
2537 struct method_list32_t {
2538   uint32_t entsize;
2539   uint32_t count;
2540   /* struct method32_t first;  These structures follow inline */
2541 };
2542 
2543 struct method64_t {
2544   uint64_t name;  /* SEL (64-bit pointer) */
2545   uint64_t types; /* const char * (64-bit pointer) */
2546   uint64_t imp;   /* IMP (64-bit pointer) */
2547 };
2548 
2549 struct method32_t {
2550   uint32_t name;  /* SEL (32-bit pointer) */
2551   uint32_t types; /* const char * (32-bit pointer) */
2552   uint32_t imp;   /* IMP (32-bit pointer) */
2553 };
2554 
2555 struct protocol_list64_t {
2556   uint64_t count; /* uintptr_t (a 64-bit value) */
2557   /* struct protocol64_t * list[0];  These pointers follow inline */
2558 };
2559 
2560 struct protocol_list32_t {
2561   uint32_t count; /* uintptr_t (a 32-bit value) */
2562   /* struct protocol32_t * list[0];  These pointers follow inline */
2563 };
2564 
2565 struct protocol64_t {
2566   uint64_t isa;                     /* id * (64-bit pointer) */
2567   uint64_t name;                    /* const char * (64-bit pointer) */
2568   uint64_t protocols;               /* struct protocol_list64_t *
2569                                                     (64-bit pointer) */
2570   uint64_t instanceMethods;         /* method_list_t * (64-bit pointer) */
2571   uint64_t classMethods;            /* method_list_t * (64-bit pointer) */
2572   uint64_t optionalInstanceMethods; /* method_list_t * (64-bit pointer) */
2573   uint64_t optionalClassMethods;    /* method_list_t * (64-bit pointer) */
2574   uint64_t instanceProperties;      /* struct objc_property_list *
2575                                                        (64-bit pointer) */
2576 };
2577 
2578 struct protocol32_t {
2579   uint32_t isa;                     /* id * (32-bit pointer) */
2580   uint32_t name;                    /* const char * (32-bit pointer) */
2581   uint32_t protocols;               /* struct protocol_list_t *
2582                                                     (32-bit pointer) */
2583   uint32_t instanceMethods;         /* method_list_t * (32-bit pointer) */
2584   uint32_t classMethods;            /* method_list_t * (32-bit pointer) */
2585   uint32_t optionalInstanceMethods; /* method_list_t * (32-bit pointer) */
2586   uint32_t optionalClassMethods;    /* method_list_t * (32-bit pointer) */
2587   uint32_t instanceProperties;      /* struct objc_property_list *
2588                                                        (32-bit pointer) */
2589 };
2590 
2591 struct ivar_list64_t {
2592   uint32_t entsize;
2593   uint32_t count;
2594   /* struct ivar64_t first;  These structures follow inline */
2595 };
2596 
2597 struct ivar_list32_t {
2598   uint32_t entsize;
2599   uint32_t count;
2600   /* struct ivar32_t first;  These structures follow inline */
2601 };
2602 
2603 struct ivar64_t {
2604   uint64_t offset; /* uintptr_t * (64-bit pointer) */
2605   uint64_t name;   /* const char * (64-bit pointer) */
2606   uint64_t type;   /* const char * (64-bit pointer) */
2607   uint32_t alignment;
2608   uint32_t size;
2609 };
2610 
2611 struct ivar32_t {
2612   uint32_t offset; /* uintptr_t * (32-bit pointer) */
2613   uint32_t name;   /* const char * (32-bit pointer) */
2614   uint32_t type;   /* const char * (32-bit pointer) */
2615   uint32_t alignment;
2616   uint32_t size;
2617 };
2618 
2619 struct objc_property_list64 {
2620   uint32_t entsize;
2621   uint32_t count;
2622   /* struct objc_property64 first;  These structures follow inline */
2623 };
2624 
2625 struct objc_property_list32 {
2626   uint32_t entsize;
2627   uint32_t count;
2628   /* struct objc_property32 first;  These structures follow inline */
2629 };
2630 
2631 struct objc_property64 {
2632   uint64_t name;       /* const char * (64-bit pointer) */
2633   uint64_t attributes; /* const char * (64-bit pointer) */
2634 };
2635 
2636 struct objc_property32 {
2637   uint32_t name;       /* const char * (32-bit pointer) */
2638   uint32_t attributes; /* const char * (32-bit pointer) */
2639 };
2640 
2641 struct category64_t {
2642   uint64_t name;               /* const char * (64-bit pointer) */
2643   uint64_t cls;                /* struct class_t * (64-bit pointer) */
2644   uint64_t instanceMethods;    /* struct method_list_t * (64-bit pointer) */
2645   uint64_t classMethods;       /* struct method_list_t * (64-bit pointer) */
2646   uint64_t protocols;          /* struct protocol_list_t * (64-bit pointer) */
2647   uint64_t instanceProperties; /* struct objc_property_list *
2648                                   (64-bit pointer) */
2649 };
2650 
2651 struct category32_t {
2652   uint32_t name;               /* const char * (32-bit pointer) */
2653   uint32_t cls;                /* struct class_t * (32-bit pointer) */
2654   uint32_t instanceMethods;    /* struct method_list_t * (32-bit pointer) */
2655   uint32_t classMethods;       /* struct method_list_t * (32-bit pointer) */
2656   uint32_t protocols;          /* struct protocol_list_t * (32-bit pointer) */
2657   uint32_t instanceProperties; /* struct objc_property_list *
2658                                   (32-bit pointer) */
2659 };
2660 
2661 struct objc_image_info64 {
2662   uint32_t version;
2663   uint32_t flags;
2664 };
2665 struct objc_image_info32 {
2666   uint32_t version;
2667   uint32_t flags;
2668 };
2669 struct imageInfo_t {
2670   uint32_t version;
2671   uint32_t flags;
2672 };
2673 /* masks for objc_image_info.flags */
2674 #define OBJC_IMAGE_IS_REPLACEMENT (1 << 0)
2675 #define OBJC_IMAGE_SUPPORTS_GC (1 << 1)
2676 
2677 struct message_ref64 {
2678   uint64_t imp; /* IMP (64-bit pointer) */
2679   uint64_t sel; /* SEL (64-bit pointer) */
2680 };
2681 
2682 struct message_ref32 {
2683   uint32_t imp; /* IMP (32-bit pointer) */
2684   uint32_t sel; /* SEL (32-bit pointer) */
2685 };
2686 
2687 // Objective-C 1 (32-bit only) meta data structs.
2688 
2689 struct objc_module_t {
2690   uint32_t version;
2691   uint32_t size;
2692   uint32_t name;   /* char * (32-bit pointer) */
2693   uint32_t symtab; /* struct objc_symtab * (32-bit pointer) */
2694 };
2695 
2696 struct objc_symtab_t {
2697   uint32_t sel_ref_cnt;
2698   uint32_t refs; /* SEL * (32-bit pointer) */
2699   uint16_t cls_def_cnt;
2700   uint16_t cat_def_cnt;
2701   // uint32_t defs[1];        /* void * (32-bit pointer) variable size */
2702 };
2703 
2704 struct objc_class_t {
2705   uint32_t isa;         /* struct objc_class * (32-bit pointer) */
2706   uint32_t super_class; /* struct objc_class * (32-bit pointer) */
2707   uint32_t name;        /* const char * (32-bit pointer) */
2708   int32_t version;
2709   int32_t info;
2710   int32_t instance_size;
2711   uint32_t ivars;       /* struct objc_ivar_list * (32-bit pointer) */
2712   uint32_t methodLists; /* struct objc_method_list ** (32-bit pointer) */
2713   uint32_t cache;       /* struct objc_cache * (32-bit pointer) */
2714   uint32_t protocols;   /* struct objc_protocol_list * (32-bit pointer) */
2715 };
2716 
2717 #define CLS_GETINFO(cls, infomask) ((cls)->info & (infomask))
2718 // class is not a metaclass
2719 #define CLS_CLASS 0x1
2720 // class is a metaclass
2721 #define CLS_META 0x2
2722 
2723 struct objc_category_t {
2724   uint32_t category_name;    /* char * (32-bit pointer) */
2725   uint32_t class_name;       /* char * (32-bit pointer) */
2726   uint32_t instance_methods; /* struct objc_method_list * (32-bit pointer) */
2727   uint32_t class_methods;    /* struct objc_method_list * (32-bit pointer) */
2728   uint32_t protocols;        /* struct objc_protocol_list * (32-bit ptr) */
2729 };
2730 
2731 struct objc_ivar_t {
2732   uint32_t ivar_name; /* char * (32-bit pointer) */
2733   uint32_t ivar_type; /* char * (32-bit pointer) */
2734   int32_t ivar_offset;
2735 };
2736 
2737 struct objc_ivar_list_t {
2738   int32_t ivar_count;
2739   // struct objc_ivar_t ivar_list[1];          /* variable length structure */
2740 };
2741 
2742 struct objc_method_list_t {
2743   uint32_t obsolete; /* struct objc_method_list * (32-bit pointer) */
2744   int32_t method_count;
2745   // struct objc_method_t method_list[1];      /* variable length structure */
2746 };
2747 
2748 struct objc_method_t {
2749   uint32_t method_name;  /* SEL, aka struct objc_selector * (32-bit pointer) */
2750   uint32_t method_types; /* char * (32-bit pointer) */
2751   uint32_t method_imp;   /* IMP, aka function pointer, (*IMP)(id, SEL, ...)
2752                             (32-bit pointer) */
2753 };
2754 
2755 struct objc_protocol_list_t {
2756   uint32_t next; /* struct objc_protocol_list * (32-bit pointer) */
2757   int32_t count;
2758   // uint32_t list[1];   /* Protocol *, aka struct objc_protocol_t *
2759   //                        (32-bit pointer) */
2760 };
2761 
2762 struct objc_protocol_t {
2763   uint32_t isa;              /* struct objc_class * (32-bit pointer) */
2764   uint32_t protocol_name;    /* char * (32-bit pointer) */
2765   uint32_t protocol_list;    /* struct objc_protocol_list * (32-bit pointer) */
2766   uint32_t instance_methods; /* struct objc_method_description_list *
2767                                 (32-bit pointer) */
2768   uint32_t class_methods;    /* struct objc_method_description_list *
2769                                 (32-bit pointer) */
2770 };
2771 
2772 struct objc_method_description_list_t {
2773   int32_t count;
2774   // struct objc_method_description_t list[1];
2775 };
2776 
2777 struct objc_method_description_t {
2778   uint32_t name;  /* SEL, aka struct objc_selector * (32-bit pointer) */
2779   uint32_t types; /* char * (32-bit pointer) */
2780 };
2781 
swapStruct(struct cfstring64_t & cfs)2782 inline void swapStruct(struct cfstring64_t &cfs) {
2783   sys::swapByteOrder(cfs.isa);
2784   sys::swapByteOrder(cfs.flags);
2785   sys::swapByteOrder(cfs.characters);
2786   sys::swapByteOrder(cfs.length);
2787 }
2788 
swapStruct(struct class64_t & c)2789 inline void swapStruct(struct class64_t &c) {
2790   sys::swapByteOrder(c.isa);
2791   sys::swapByteOrder(c.superclass);
2792   sys::swapByteOrder(c.cache);
2793   sys::swapByteOrder(c.vtable);
2794   sys::swapByteOrder(c.data);
2795 }
2796 
swapStruct(struct class32_t & c)2797 inline void swapStruct(struct class32_t &c) {
2798   sys::swapByteOrder(c.isa);
2799   sys::swapByteOrder(c.superclass);
2800   sys::swapByteOrder(c.cache);
2801   sys::swapByteOrder(c.vtable);
2802   sys::swapByteOrder(c.data);
2803 }
2804 
swapStruct(struct class_ro64_t & cro)2805 inline void swapStruct(struct class_ro64_t &cro) {
2806   sys::swapByteOrder(cro.flags);
2807   sys::swapByteOrder(cro.instanceStart);
2808   sys::swapByteOrder(cro.instanceSize);
2809   sys::swapByteOrder(cro.reserved);
2810   sys::swapByteOrder(cro.ivarLayout);
2811   sys::swapByteOrder(cro.name);
2812   sys::swapByteOrder(cro.baseMethods);
2813   sys::swapByteOrder(cro.baseProtocols);
2814   sys::swapByteOrder(cro.ivars);
2815   sys::swapByteOrder(cro.weakIvarLayout);
2816   sys::swapByteOrder(cro.baseProperties);
2817 }
2818 
swapStruct(struct class_ro32_t & cro)2819 inline void swapStruct(struct class_ro32_t &cro) {
2820   sys::swapByteOrder(cro.flags);
2821   sys::swapByteOrder(cro.instanceStart);
2822   sys::swapByteOrder(cro.instanceSize);
2823   sys::swapByteOrder(cro.ivarLayout);
2824   sys::swapByteOrder(cro.name);
2825   sys::swapByteOrder(cro.baseMethods);
2826   sys::swapByteOrder(cro.baseProtocols);
2827   sys::swapByteOrder(cro.ivars);
2828   sys::swapByteOrder(cro.weakIvarLayout);
2829   sys::swapByteOrder(cro.baseProperties);
2830 }
2831 
swapStruct(struct method_list64_t & ml)2832 inline void swapStruct(struct method_list64_t &ml) {
2833   sys::swapByteOrder(ml.entsize);
2834   sys::swapByteOrder(ml.count);
2835 }
2836 
swapStruct(struct method_list32_t & ml)2837 inline void swapStruct(struct method_list32_t &ml) {
2838   sys::swapByteOrder(ml.entsize);
2839   sys::swapByteOrder(ml.count);
2840 }
2841 
swapStruct(struct method64_t & m)2842 inline void swapStruct(struct method64_t &m) {
2843   sys::swapByteOrder(m.name);
2844   sys::swapByteOrder(m.types);
2845   sys::swapByteOrder(m.imp);
2846 }
2847 
swapStruct(struct method32_t & m)2848 inline void swapStruct(struct method32_t &m) {
2849   sys::swapByteOrder(m.name);
2850   sys::swapByteOrder(m.types);
2851   sys::swapByteOrder(m.imp);
2852 }
2853 
swapStruct(struct protocol_list64_t & pl)2854 inline void swapStruct(struct protocol_list64_t &pl) {
2855   sys::swapByteOrder(pl.count);
2856 }
2857 
swapStruct(struct protocol_list32_t & pl)2858 inline void swapStruct(struct protocol_list32_t &pl) {
2859   sys::swapByteOrder(pl.count);
2860 }
2861 
swapStruct(struct protocol64_t & p)2862 inline void swapStruct(struct protocol64_t &p) {
2863   sys::swapByteOrder(p.isa);
2864   sys::swapByteOrder(p.name);
2865   sys::swapByteOrder(p.protocols);
2866   sys::swapByteOrder(p.instanceMethods);
2867   sys::swapByteOrder(p.classMethods);
2868   sys::swapByteOrder(p.optionalInstanceMethods);
2869   sys::swapByteOrder(p.optionalClassMethods);
2870   sys::swapByteOrder(p.instanceProperties);
2871 }
2872 
swapStruct(struct protocol32_t & p)2873 inline void swapStruct(struct protocol32_t &p) {
2874   sys::swapByteOrder(p.isa);
2875   sys::swapByteOrder(p.name);
2876   sys::swapByteOrder(p.protocols);
2877   sys::swapByteOrder(p.instanceMethods);
2878   sys::swapByteOrder(p.classMethods);
2879   sys::swapByteOrder(p.optionalInstanceMethods);
2880   sys::swapByteOrder(p.optionalClassMethods);
2881   sys::swapByteOrder(p.instanceProperties);
2882 }
2883 
swapStruct(struct ivar_list64_t & il)2884 inline void swapStruct(struct ivar_list64_t &il) {
2885   sys::swapByteOrder(il.entsize);
2886   sys::swapByteOrder(il.count);
2887 }
2888 
swapStruct(struct ivar_list32_t & il)2889 inline void swapStruct(struct ivar_list32_t &il) {
2890   sys::swapByteOrder(il.entsize);
2891   sys::swapByteOrder(il.count);
2892 }
2893 
swapStruct(struct ivar64_t & i)2894 inline void swapStruct(struct ivar64_t &i) {
2895   sys::swapByteOrder(i.offset);
2896   sys::swapByteOrder(i.name);
2897   sys::swapByteOrder(i.type);
2898   sys::swapByteOrder(i.alignment);
2899   sys::swapByteOrder(i.size);
2900 }
2901 
swapStruct(struct ivar32_t & i)2902 inline void swapStruct(struct ivar32_t &i) {
2903   sys::swapByteOrder(i.offset);
2904   sys::swapByteOrder(i.name);
2905   sys::swapByteOrder(i.type);
2906   sys::swapByteOrder(i.alignment);
2907   sys::swapByteOrder(i.size);
2908 }
2909 
swapStruct(struct objc_property_list64 & pl)2910 inline void swapStruct(struct objc_property_list64 &pl) {
2911   sys::swapByteOrder(pl.entsize);
2912   sys::swapByteOrder(pl.count);
2913 }
2914 
swapStruct(struct objc_property_list32 & pl)2915 inline void swapStruct(struct objc_property_list32 &pl) {
2916   sys::swapByteOrder(pl.entsize);
2917   sys::swapByteOrder(pl.count);
2918 }
2919 
swapStruct(struct objc_property64 & op)2920 inline void swapStruct(struct objc_property64 &op) {
2921   sys::swapByteOrder(op.name);
2922   sys::swapByteOrder(op.attributes);
2923 }
2924 
swapStruct(struct objc_property32 & op)2925 inline void swapStruct(struct objc_property32 &op) {
2926   sys::swapByteOrder(op.name);
2927   sys::swapByteOrder(op.attributes);
2928 }
2929 
swapStruct(struct category64_t & c)2930 inline void swapStruct(struct category64_t &c) {
2931   sys::swapByteOrder(c.name);
2932   sys::swapByteOrder(c.cls);
2933   sys::swapByteOrder(c.instanceMethods);
2934   sys::swapByteOrder(c.classMethods);
2935   sys::swapByteOrder(c.protocols);
2936   sys::swapByteOrder(c.instanceProperties);
2937 }
2938 
swapStruct(struct category32_t & c)2939 inline void swapStruct(struct category32_t &c) {
2940   sys::swapByteOrder(c.name);
2941   sys::swapByteOrder(c.cls);
2942   sys::swapByteOrder(c.instanceMethods);
2943   sys::swapByteOrder(c.classMethods);
2944   sys::swapByteOrder(c.protocols);
2945   sys::swapByteOrder(c.instanceProperties);
2946 }
2947 
swapStruct(struct objc_image_info64 & o)2948 inline void swapStruct(struct objc_image_info64 &o) {
2949   sys::swapByteOrder(o.version);
2950   sys::swapByteOrder(o.flags);
2951 }
2952 
swapStruct(struct objc_image_info32 & o)2953 inline void swapStruct(struct objc_image_info32 &o) {
2954   sys::swapByteOrder(o.version);
2955   sys::swapByteOrder(o.flags);
2956 }
2957 
swapStruct(struct imageInfo_t & o)2958 inline void swapStruct(struct imageInfo_t &o) {
2959   sys::swapByteOrder(o.version);
2960   sys::swapByteOrder(o.flags);
2961 }
2962 
swapStruct(struct message_ref64 & mr)2963 inline void swapStruct(struct message_ref64 &mr) {
2964   sys::swapByteOrder(mr.imp);
2965   sys::swapByteOrder(mr.sel);
2966 }
2967 
swapStruct(struct message_ref32 & mr)2968 inline void swapStruct(struct message_ref32 &mr) {
2969   sys::swapByteOrder(mr.imp);
2970   sys::swapByteOrder(mr.sel);
2971 }
2972 
swapStruct(struct objc_module_t & module)2973 inline void swapStruct(struct objc_module_t &module) {
2974   sys::swapByteOrder(module.version);
2975   sys::swapByteOrder(module.size);
2976   sys::swapByteOrder(module.name);
2977   sys::swapByteOrder(module.symtab);
2978 }
2979 
swapStruct(struct objc_symtab_t & symtab)2980 inline void swapStruct(struct objc_symtab_t &symtab) {
2981   sys::swapByteOrder(symtab.sel_ref_cnt);
2982   sys::swapByteOrder(symtab.refs);
2983   sys::swapByteOrder(symtab.cls_def_cnt);
2984   sys::swapByteOrder(symtab.cat_def_cnt);
2985 }
2986 
swapStruct(struct objc_class_t & objc_class)2987 inline void swapStruct(struct objc_class_t &objc_class) {
2988   sys::swapByteOrder(objc_class.isa);
2989   sys::swapByteOrder(objc_class.super_class);
2990   sys::swapByteOrder(objc_class.name);
2991   sys::swapByteOrder(objc_class.version);
2992   sys::swapByteOrder(objc_class.info);
2993   sys::swapByteOrder(objc_class.instance_size);
2994   sys::swapByteOrder(objc_class.ivars);
2995   sys::swapByteOrder(objc_class.methodLists);
2996   sys::swapByteOrder(objc_class.cache);
2997   sys::swapByteOrder(objc_class.protocols);
2998 }
2999 
swapStruct(struct objc_category_t & objc_category)3000 inline void swapStruct(struct objc_category_t &objc_category) {
3001   sys::swapByteOrder(objc_category.category_name);
3002   sys::swapByteOrder(objc_category.class_name);
3003   sys::swapByteOrder(objc_category.instance_methods);
3004   sys::swapByteOrder(objc_category.class_methods);
3005   sys::swapByteOrder(objc_category.protocols);
3006 }
3007 
swapStruct(struct objc_ivar_list_t & objc_ivar_list)3008 inline void swapStruct(struct objc_ivar_list_t &objc_ivar_list) {
3009   sys::swapByteOrder(objc_ivar_list.ivar_count);
3010 }
3011 
swapStruct(struct objc_ivar_t & objc_ivar)3012 inline void swapStruct(struct objc_ivar_t &objc_ivar) {
3013   sys::swapByteOrder(objc_ivar.ivar_name);
3014   sys::swapByteOrder(objc_ivar.ivar_type);
3015   sys::swapByteOrder(objc_ivar.ivar_offset);
3016 }
3017 
swapStruct(struct objc_method_list_t & method_list)3018 inline void swapStruct(struct objc_method_list_t &method_list) {
3019   sys::swapByteOrder(method_list.obsolete);
3020   sys::swapByteOrder(method_list.method_count);
3021 }
3022 
swapStruct(struct objc_method_t & method)3023 inline void swapStruct(struct objc_method_t &method) {
3024   sys::swapByteOrder(method.method_name);
3025   sys::swapByteOrder(method.method_types);
3026   sys::swapByteOrder(method.method_imp);
3027 }
3028 
swapStruct(struct objc_protocol_list_t & protocol_list)3029 inline void swapStruct(struct objc_protocol_list_t &protocol_list) {
3030   sys::swapByteOrder(protocol_list.next);
3031   sys::swapByteOrder(protocol_list.count);
3032 }
3033 
swapStruct(struct objc_protocol_t & protocol)3034 inline void swapStruct(struct objc_protocol_t &protocol) {
3035   sys::swapByteOrder(protocol.isa);
3036   sys::swapByteOrder(protocol.protocol_name);
3037   sys::swapByteOrder(protocol.protocol_list);
3038   sys::swapByteOrder(protocol.instance_methods);
3039   sys::swapByteOrder(protocol.class_methods);
3040 }
3041 
swapStruct(struct objc_method_description_list_t & mdl)3042 inline void swapStruct(struct objc_method_description_list_t &mdl) {
3043   sys::swapByteOrder(mdl.count);
3044 }
3045 
swapStruct(struct objc_method_description_t & md)3046 inline void swapStruct(struct objc_method_description_t &md) {
3047   sys::swapByteOrder(md.name);
3048   sys::swapByteOrder(md.types);
3049 }
3050 
3051 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
3052                                                  struct DisassembleInfo *info);
3053 
3054 // get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
3055 // to an Objective-C class and returns the class name.  It is also passed the
3056 // address of the pointer, so when the pointer is zero as it can be in an .o
3057 // file, that is used to look for an external relocation entry with a symbol
3058 // name.
get_objc2_64bit_class_name(uint64_t pointer_value,uint64_t ReferenceValue,struct DisassembleInfo * info)3059 static const char *get_objc2_64bit_class_name(uint64_t pointer_value,
3060                                               uint64_t ReferenceValue,
3061                                               struct DisassembleInfo *info) {
3062   const char *r;
3063   uint32_t offset, left;
3064   SectionRef S;
3065 
3066   // The pointer_value can be 0 in an object file and have a relocation
3067   // entry for the class symbol at the ReferenceValue (the address of the
3068   // pointer).
3069   if (pointer_value == 0) {
3070     r = get_pointer_64(ReferenceValue, offset, left, S, info);
3071     if (r == nullptr || left < sizeof(uint64_t))
3072       return nullptr;
3073     uint64_t n_value;
3074     const char *symbol_name = get_symbol_64(offset, S, info, n_value);
3075     if (symbol_name == nullptr)
3076       return nullptr;
3077     const char *class_name = strrchr(symbol_name, '$');
3078     if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
3079       return class_name + 2;
3080     else
3081       return nullptr;
3082   }
3083 
3084   // The case were the pointer_value is non-zero and points to a class defined
3085   // in this Mach-O file.
3086   r = get_pointer_64(pointer_value, offset, left, S, info);
3087   if (r == nullptr || left < sizeof(struct class64_t))
3088     return nullptr;
3089   struct class64_t c;
3090   memcpy(&c, r, sizeof(struct class64_t));
3091   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3092     swapStruct(c);
3093   if (c.data == 0)
3094     return nullptr;
3095   r = get_pointer_64(c.data, offset, left, S, info);
3096   if (r == nullptr || left < sizeof(struct class_ro64_t))
3097     return nullptr;
3098   struct class_ro64_t cro;
3099   memcpy(&cro, r, sizeof(struct class_ro64_t));
3100   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3101     swapStruct(cro);
3102   if (cro.name == 0)
3103     return nullptr;
3104   const char *name = get_pointer_64(cro.name, offset, left, S, info);
3105   return name;
3106 }
3107 
3108 // get_objc2_64bit_cfstring_name is used for disassembly and is passed a
3109 // pointer to a cfstring and returns its name or nullptr.
get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,struct DisassembleInfo * info)3110 static const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
3111                                                  struct DisassembleInfo *info) {
3112   const char *r, *name;
3113   uint32_t offset, left;
3114   SectionRef S;
3115   struct cfstring64_t cfs;
3116   uint64_t cfs_characters;
3117 
3118   r = get_pointer_64(ReferenceValue, offset, left, S, info);
3119   if (r == nullptr || left < sizeof(struct cfstring64_t))
3120     return nullptr;
3121   memcpy(&cfs, r, sizeof(struct cfstring64_t));
3122   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3123     swapStruct(cfs);
3124   if (cfs.characters == 0) {
3125     uint64_t n_value;
3126     const char *symbol_name = get_symbol_64(
3127         offset + offsetof(struct cfstring64_t, characters), S, info, n_value);
3128     if (symbol_name == nullptr)
3129       return nullptr;
3130     cfs_characters = n_value;
3131   } else
3132     cfs_characters = cfs.characters;
3133   name = get_pointer_64(cfs_characters, offset, left, S, info);
3134 
3135   return name;
3136 }
3137 
3138 // get_objc2_64bit_selref() is used for disassembly and is passed a the address
3139 // of a pointer to an Objective-C selector reference when the pointer value is
3140 // zero as in a .o file and is likely to have a external relocation entry with
3141 // who's symbol's n_value is the real pointer to the selector name.  If that is
3142 // the case the real pointer to the selector name is returned else 0 is
3143 // returned
get_objc2_64bit_selref(uint64_t ReferenceValue,struct DisassembleInfo * info)3144 static uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
3145                                        struct DisassembleInfo *info) {
3146   uint32_t offset, left;
3147   SectionRef S;
3148 
3149   const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
3150   if (r == nullptr || left < sizeof(uint64_t))
3151     return 0;
3152   uint64_t n_value;
3153   const char *symbol_name = get_symbol_64(offset, S, info, n_value);
3154   if (symbol_name == nullptr)
3155     return 0;
3156   return n_value;
3157 }
3158 
get_section(MachOObjectFile * O,const char * segname,const char * sectname)3159 static const SectionRef get_section(MachOObjectFile *O, const char *segname,
3160                                     const char *sectname) {
3161   for (const SectionRef &Section : O->sections()) {
3162     StringRef SectName;
3163     Section.getName(SectName);
3164     DataRefImpl Ref = Section.getRawDataRefImpl();
3165     StringRef SegName = O->getSectionFinalSegmentName(Ref);
3166     if (SegName == segname && SectName == sectname)
3167       return Section;
3168   }
3169   return SectionRef();
3170 }
3171 
3172 static void
walk_pointer_list_64(const char * listname,const SectionRef S,MachOObjectFile * O,struct DisassembleInfo * info,void (* func)(uint64_t,struct DisassembleInfo * info))3173 walk_pointer_list_64(const char *listname, const SectionRef S,
3174                      MachOObjectFile *O, struct DisassembleInfo *info,
3175                      void (*func)(uint64_t, struct DisassembleInfo *info)) {
3176   if (S == SectionRef())
3177     return;
3178 
3179   StringRef SectName;
3180   S.getName(SectName);
3181   DataRefImpl Ref = S.getRawDataRefImpl();
3182   StringRef SegName = O->getSectionFinalSegmentName(Ref);
3183   outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
3184 
3185   StringRef BytesStr;
3186   S.getContents(BytesStr);
3187   const char *Contents = reinterpret_cast<const char *>(BytesStr.data());
3188 
3189   for (uint32_t i = 0; i < S.getSize(); i += sizeof(uint64_t)) {
3190     uint32_t left = S.getSize() - i;
3191     uint32_t size = left < sizeof(uint64_t) ? left : sizeof(uint64_t);
3192     uint64_t p = 0;
3193     memcpy(&p, Contents + i, size);
3194     if (i + sizeof(uint64_t) > S.getSize())
3195       outs() << listname << " list pointer extends past end of (" << SegName
3196              << "," << SectName << ") section\n";
3197     outs() << format("%016" PRIx64, S.getAddress() + i) << " ";
3198 
3199     if (O->isLittleEndian() != sys::IsLittleEndianHost)
3200       sys::swapByteOrder(p);
3201 
3202     uint64_t n_value = 0;
3203     const char *name = get_symbol_64(i, S, info, n_value, p);
3204     if (name == nullptr)
3205       name = get_dyld_bind_info_symbolname(S.getAddress() + i, info);
3206 
3207     if (n_value != 0) {
3208       outs() << format("0x%" PRIx64, n_value);
3209       if (p != 0)
3210         outs() << " + " << format("0x%" PRIx64, p);
3211     } else
3212       outs() << format("0x%" PRIx64, p);
3213     if (name != nullptr)
3214       outs() << " " << name;
3215     outs() << "\n";
3216 
3217     p += n_value;
3218     if (func)
3219       func(p, info);
3220   }
3221 }
3222 
3223 static void
walk_pointer_list_32(const char * listname,const SectionRef S,MachOObjectFile * O,struct DisassembleInfo * info,void (* func)(uint32_t,struct DisassembleInfo * info))3224 walk_pointer_list_32(const char *listname, const SectionRef S,
3225                      MachOObjectFile *O, struct DisassembleInfo *info,
3226                      void (*func)(uint32_t, struct DisassembleInfo *info)) {
3227   if (S == SectionRef())
3228     return;
3229 
3230   StringRef SectName;
3231   S.getName(SectName);
3232   DataRefImpl Ref = S.getRawDataRefImpl();
3233   StringRef SegName = O->getSectionFinalSegmentName(Ref);
3234   outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
3235 
3236   StringRef BytesStr;
3237   S.getContents(BytesStr);
3238   const char *Contents = reinterpret_cast<const char *>(BytesStr.data());
3239 
3240   for (uint32_t i = 0; i < S.getSize(); i += sizeof(uint32_t)) {
3241     uint32_t left = S.getSize() - i;
3242     uint32_t size = left < sizeof(uint32_t) ? left : sizeof(uint32_t);
3243     uint32_t p = 0;
3244     memcpy(&p, Contents + i, size);
3245     if (i + sizeof(uint32_t) > S.getSize())
3246       outs() << listname << " list pointer extends past end of (" << SegName
3247              << "," << SectName << ") section\n";
3248     uint32_t Address = S.getAddress() + i;
3249     outs() << format("%08" PRIx32, Address) << " ";
3250 
3251     if (O->isLittleEndian() != sys::IsLittleEndianHost)
3252       sys::swapByteOrder(p);
3253     outs() << format("0x%" PRIx32, p);
3254 
3255     const char *name = get_symbol_32(i, S, info, p);
3256     if (name != nullptr)
3257       outs() << " " << name;
3258     outs() << "\n";
3259 
3260     if (func)
3261       func(p, info);
3262   }
3263 }
3264 
print_layout_map(const char * layout_map,uint32_t left)3265 static void print_layout_map(const char *layout_map, uint32_t left) {
3266   if (layout_map == nullptr)
3267     return;
3268   outs() << "                layout map: ";
3269   do {
3270     outs() << format("0x%02" PRIx32, (*layout_map) & 0xff) << " ";
3271     left--;
3272     layout_map++;
3273   } while (*layout_map != '\0' && left != 0);
3274   outs() << "\n";
3275 }
3276 
print_layout_map64(uint64_t p,struct DisassembleInfo * info)3277 static void print_layout_map64(uint64_t p, struct DisassembleInfo *info) {
3278   uint32_t offset, left;
3279   SectionRef S;
3280   const char *layout_map;
3281 
3282   if (p == 0)
3283     return;
3284   layout_map = get_pointer_64(p, offset, left, S, info);
3285   print_layout_map(layout_map, left);
3286 }
3287 
print_layout_map32(uint32_t p,struct DisassembleInfo * info)3288 static void print_layout_map32(uint32_t p, struct DisassembleInfo *info) {
3289   uint32_t offset, left;
3290   SectionRef S;
3291   const char *layout_map;
3292 
3293   if (p == 0)
3294     return;
3295   layout_map = get_pointer_32(p, offset, left, S, info);
3296   print_layout_map(layout_map, left);
3297 }
3298 
print_method_list64_t(uint64_t p,struct DisassembleInfo * info,const char * indent)3299 static void print_method_list64_t(uint64_t p, struct DisassembleInfo *info,
3300                                   const char *indent) {
3301   struct method_list64_t ml;
3302   struct method64_t m;
3303   const char *r;
3304   uint32_t offset, xoffset, left, i;
3305   SectionRef S, xS;
3306   const char *name, *sym_name;
3307   uint64_t n_value;
3308 
3309   r = get_pointer_64(p, offset, left, S, info);
3310   if (r == nullptr)
3311     return;
3312   memset(&ml, '\0', sizeof(struct method_list64_t));
3313   if (left < sizeof(struct method_list64_t)) {
3314     memcpy(&ml, r, left);
3315     outs() << "   (method_list_t entends past the end of the section)\n";
3316   } else
3317     memcpy(&ml, r, sizeof(struct method_list64_t));
3318   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3319     swapStruct(ml);
3320   outs() << indent << "\t\t   entsize " << ml.entsize << "\n";
3321   outs() << indent << "\t\t     count " << ml.count << "\n";
3322 
3323   p += sizeof(struct method_list64_t);
3324   offset += sizeof(struct method_list64_t);
3325   for (i = 0; i < ml.count; i++) {
3326     r = get_pointer_64(p, offset, left, S, info);
3327     if (r == nullptr)
3328       return;
3329     memset(&m, '\0', sizeof(struct method64_t));
3330     if (left < sizeof(struct method64_t)) {
3331       memcpy(&m, r, left);
3332       outs() << indent << "   (method_t extends past the end of the section)\n";
3333     } else
3334       memcpy(&m, r, sizeof(struct method64_t));
3335     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3336       swapStruct(m);
3337 
3338     outs() << indent << "\t\t      name ";
3339     sym_name = get_symbol_64(offset + offsetof(struct method64_t, name), S,
3340                              info, n_value, m.name);
3341     if (n_value != 0) {
3342       if (info->verbose && sym_name != nullptr)
3343         outs() << sym_name;
3344       else
3345         outs() << format("0x%" PRIx64, n_value);
3346       if (m.name != 0)
3347         outs() << " + " << format("0x%" PRIx64, m.name);
3348     } else
3349       outs() << format("0x%" PRIx64, m.name);
3350     name = get_pointer_64(m.name + n_value, xoffset, left, xS, info);
3351     if (name != nullptr)
3352       outs() << format(" %.*s", left, name);
3353     outs() << "\n";
3354 
3355     outs() << indent << "\t\t     types ";
3356     sym_name = get_symbol_64(offset + offsetof(struct method64_t, types), S,
3357                              info, n_value, m.types);
3358     if (n_value != 0) {
3359       if (info->verbose && sym_name != nullptr)
3360         outs() << sym_name;
3361       else
3362         outs() << format("0x%" PRIx64, n_value);
3363       if (m.types != 0)
3364         outs() << " + " << format("0x%" PRIx64, m.types);
3365     } else
3366       outs() << format("0x%" PRIx64, m.types);
3367     name = get_pointer_64(m.types + n_value, xoffset, left, xS, info);
3368     if (name != nullptr)
3369       outs() << format(" %.*s", left, name);
3370     outs() << "\n";
3371 
3372     outs() << indent << "\t\t       imp ";
3373     name = get_symbol_64(offset + offsetof(struct method64_t, imp), S, info,
3374                          n_value, m.imp);
3375     if (info->verbose && name == nullptr) {
3376       if (n_value != 0) {
3377         outs() << format("0x%" PRIx64, n_value) << " ";
3378         if (m.imp != 0)
3379           outs() << "+ " << format("0x%" PRIx64, m.imp) << " ";
3380       } else
3381         outs() << format("0x%" PRIx64, m.imp) << " ";
3382     }
3383     if (name != nullptr)
3384       outs() << name;
3385     outs() << "\n";
3386 
3387     p += sizeof(struct method64_t);
3388     offset += sizeof(struct method64_t);
3389   }
3390 }
3391 
print_method_list32_t(uint64_t p,struct DisassembleInfo * info,const char * indent)3392 static void print_method_list32_t(uint64_t p, struct DisassembleInfo *info,
3393                                   const char *indent) {
3394   struct method_list32_t ml;
3395   struct method32_t m;
3396   const char *r, *name;
3397   uint32_t offset, xoffset, left, i;
3398   SectionRef S, xS;
3399 
3400   r = get_pointer_32(p, offset, left, S, info);
3401   if (r == nullptr)
3402     return;
3403   memset(&ml, '\0', sizeof(struct method_list32_t));
3404   if (left < sizeof(struct method_list32_t)) {
3405     memcpy(&ml, r, left);
3406     outs() << "   (method_list_t entends past the end of the section)\n";
3407   } else
3408     memcpy(&ml, r, sizeof(struct method_list32_t));
3409   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3410     swapStruct(ml);
3411   outs() << indent << "\t\t   entsize " << ml.entsize << "\n";
3412   outs() << indent << "\t\t     count " << ml.count << "\n";
3413 
3414   p += sizeof(struct method_list32_t);
3415   offset += sizeof(struct method_list32_t);
3416   for (i = 0; i < ml.count; i++) {
3417     r = get_pointer_32(p, offset, left, S, info);
3418     if (r == nullptr)
3419       return;
3420     memset(&m, '\0', sizeof(struct method32_t));
3421     if (left < sizeof(struct method32_t)) {
3422       memcpy(&ml, r, left);
3423       outs() << indent << "   (method_t entends past the end of the section)\n";
3424     } else
3425       memcpy(&m, r, sizeof(struct method32_t));
3426     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3427       swapStruct(m);
3428 
3429     outs() << indent << "\t\t      name " << format("0x%" PRIx32, m.name);
3430     name = get_pointer_32(m.name, xoffset, left, xS, info);
3431     if (name != nullptr)
3432       outs() << format(" %.*s", left, name);
3433     outs() << "\n";
3434 
3435     outs() << indent << "\t\t     types " << format("0x%" PRIx32, m.types);
3436     name = get_pointer_32(m.types, xoffset, left, xS, info);
3437     if (name != nullptr)
3438       outs() << format(" %.*s", left, name);
3439     outs() << "\n";
3440 
3441     outs() << indent << "\t\t       imp " << format("0x%" PRIx32, m.imp);
3442     name = get_symbol_32(offset + offsetof(struct method32_t, imp), S, info,
3443                          m.imp);
3444     if (name != nullptr)
3445       outs() << " " << name;
3446     outs() << "\n";
3447 
3448     p += sizeof(struct method32_t);
3449     offset += sizeof(struct method32_t);
3450   }
3451 }
3452 
print_method_list(uint32_t p,struct DisassembleInfo * info)3453 static bool print_method_list(uint32_t p, struct DisassembleInfo *info) {
3454   uint32_t offset, left, xleft;
3455   SectionRef S;
3456   struct objc_method_list_t method_list;
3457   struct objc_method_t method;
3458   const char *r, *methods, *name, *SymbolName;
3459   int32_t i;
3460 
3461   r = get_pointer_32(p, offset, left, S, info, true);
3462   if (r == nullptr)
3463     return true;
3464 
3465   outs() << "\n";
3466   if (left > sizeof(struct objc_method_list_t)) {
3467     memcpy(&method_list, r, sizeof(struct objc_method_list_t));
3468   } else {
3469     outs() << "\t\t objc_method_list extends past end of the section\n";
3470     memset(&method_list, '\0', sizeof(struct objc_method_list_t));
3471     memcpy(&method_list, r, left);
3472   }
3473   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3474     swapStruct(method_list);
3475 
3476   outs() << "\t\t         obsolete "
3477          << format("0x%08" PRIx32, method_list.obsolete) << "\n";
3478   outs() << "\t\t     method_count " << method_list.method_count << "\n";
3479 
3480   methods = r + sizeof(struct objc_method_list_t);
3481   for (i = 0; i < method_list.method_count; i++) {
3482     if ((i + 1) * sizeof(struct objc_method_t) > left) {
3483       outs() << "\t\t remaining method's extend past the of the section\n";
3484       break;
3485     }
3486     memcpy(&method, methods + i * sizeof(struct objc_method_t),
3487            sizeof(struct objc_method_t));
3488     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3489       swapStruct(method);
3490 
3491     outs() << "\t\t      method_name "
3492            << format("0x%08" PRIx32, method.method_name);
3493     if (info->verbose) {
3494       name = get_pointer_32(method.method_name, offset, xleft, S, info, true);
3495       if (name != nullptr)
3496         outs() << format(" %.*s", xleft, name);
3497       else
3498         outs() << " (not in an __OBJC section)";
3499     }
3500     outs() << "\n";
3501 
3502     outs() << "\t\t     method_types "
3503            << format("0x%08" PRIx32, method.method_types);
3504     if (info->verbose) {
3505       name = get_pointer_32(method.method_types, offset, xleft, S, info, true);
3506       if (name != nullptr)
3507         outs() << format(" %.*s", xleft, name);
3508       else
3509         outs() << " (not in an __OBJC section)";
3510     }
3511     outs() << "\n";
3512 
3513     outs() << "\t\t       method_imp "
3514            << format("0x%08" PRIx32, method.method_imp) << " ";
3515     if (info->verbose) {
3516       SymbolName = GuessSymbolName(method.method_imp, info->AddrMap);
3517       if (SymbolName != nullptr)
3518         outs() << SymbolName;
3519     }
3520     outs() << "\n";
3521   }
3522   return false;
3523 }
3524 
print_protocol_list64_t(uint64_t p,struct DisassembleInfo * info)3525 static void print_protocol_list64_t(uint64_t p, struct DisassembleInfo *info) {
3526   struct protocol_list64_t pl;
3527   uint64_t q, n_value;
3528   struct protocol64_t pc;
3529   const char *r;
3530   uint32_t offset, xoffset, left, i;
3531   SectionRef S, xS;
3532   const char *name, *sym_name;
3533 
3534   r = get_pointer_64(p, offset, left, S, info);
3535   if (r == nullptr)
3536     return;
3537   memset(&pl, '\0', sizeof(struct protocol_list64_t));
3538   if (left < sizeof(struct protocol_list64_t)) {
3539     memcpy(&pl, r, left);
3540     outs() << "   (protocol_list_t entends past the end of the section)\n";
3541   } else
3542     memcpy(&pl, r, sizeof(struct protocol_list64_t));
3543   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3544     swapStruct(pl);
3545   outs() << "                      count " << pl.count << "\n";
3546 
3547   p += sizeof(struct protocol_list64_t);
3548   offset += sizeof(struct protocol_list64_t);
3549   for (i = 0; i < pl.count; i++) {
3550     r = get_pointer_64(p, offset, left, S, info);
3551     if (r == nullptr)
3552       return;
3553     q = 0;
3554     if (left < sizeof(uint64_t)) {
3555       memcpy(&q, r, left);
3556       outs() << "   (protocol_t * entends past the end of the section)\n";
3557     } else
3558       memcpy(&q, r, sizeof(uint64_t));
3559     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3560       sys::swapByteOrder(q);
3561 
3562     outs() << "\t\t      list[" << i << "] ";
3563     sym_name = get_symbol_64(offset, S, info, n_value, q);
3564     if (n_value != 0) {
3565       if (info->verbose && sym_name != nullptr)
3566         outs() << sym_name;
3567       else
3568         outs() << format("0x%" PRIx64, n_value);
3569       if (q != 0)
3570         outs() << " + " << format("0x%" PRIx64, q);
3571     } else
3572       outs() << format("0x%" PRIx64, q);
3573     outs() << " (struct protocol_t *)\n";
3574 
3575     r = get_pointer_64(q + n_value, offset, left, S, info);
3576     if (r == nullptr)
3577       return;
3578     memset(&pc, '\0', sizeof(struct protocol64_t));
3579     if (left < sizeof(struct protocol64_t)) {
3580       memcpy(&pc, r, left);
3581       outs() << "   (protocol_t entends past the end of the section)\n";
3582     } else
3583       memcpy(&pc, r, sizeof(struct protocol64_t));
3584     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3585       swapStruct(pc);
3586 
3587     outs() << "\t\t\t      isa " << format("0x%" PRIx64, pc.isa) << "\n";
3588 
3589     outs() << "\t\t\t     name ";
3590     sym_name = get_symbol_64(offset + offsetof(struct protocol64_t, name), S,
3591                              info, n_value, pc.name);
3592     if (n_value != 0) {
3593       if (info->verbose && sym_name != nullptr)
3594         outs() << sym_name;
3595       else
3596         outs() << format("0x%" PRIx64, n_value);
3597       if (pc.name != 0)
3598         outs() << " + " << format("0x%" PRIx64, pc.name);
3599     } else
3600       outs() << format("0x%" PRIx64, pc.name);
3601     name = get_pointer_64(pc.name + n_value, xoffset, left, xS, info);
3602     if (name != nullptr)
3603       outs() << format(" %.*s", left, name);
3604     outs() << "\n";
3605 
3606     outs() << "\t\t\tprotocols " << format("0x%" PRIx64, pc.protocols) << "\n";
3607 
3608     outs() << "\t\t  instanceMethods ";
3609     sym_name =
3610         get_symbol_64(offset + offsetof(struct protocol64_t, instanceMethods),
3611                       S, info, n_value, pc.instanceMethods);
3612     if (n_value != 0) {
3613       if (info->verbose && sym_name != nullptr)
3614         outs() << sym_name;
3615       else
3616         outs() << format("0x%" PRIx64, n_value);
3617       if (pc.instanceMethods != 0)
3618         outs() << " + " << format("0x%" PRIx64, pc.instanceMethods);
3619     } else
3620       outs() << format("0x%" PRIx64, pc.instanceMethods);
3621     outs() << " (struct method_list_t *)\n";
3622     if (pc.instanceMethods + n_value != 0)
3623       print_method_list64_t(pc.instanceMethods + n_value, info, "\t");
3624 
3625     outs() << "\t\t     classMethods ";
3626     sym_name =
3627         get_symbol_64(offset + offsetof(struct protocol64_t, classMethods), S,
3628                       info, n_value, pc.classMethods);
3629     if (n_value != 0) {
3630       if (info->verbose && sym_name != nullptr)
3631         outs() << sym_name;
3632       else
3633         outs() << format("0x%" PRIx64, n_value);
3634       if (pc.classMethods != 0)
3635         outs() << " + " << format("0x%" PRIx64, pc.classMethods);
3636     } else
3637       outs() << format("0x%" PRIx64, pc.classMethods);
3638     outs() << " (struct method_list_t *)\n";
3639     if (pc.classMethods + n_value != 0)
3640       print_method_list64_t(pc.classMethods + n_value, info, "\t");
3641 
3642     outs() << "\t  optionalInstanceMethods "
3643            << format("0x%" PRIx64, pc.optionalInstanceMethods) << "\n";
3644     outs() << "\t     optionalClassMethods "
3645            << format("0x%" PRIx64, pc.optionalClassMethods) << "\n";
3646     outs() << "\t       instanceProperties "
3647            << format("0x%" PRIx64, pc.instanceProperties) << "\n";
3648 
3649     p += sizeof(uint64_t);
3650     offset += sizeof(uint64_t);
3651   }
3652 }
3653 
print_protocol_list32_t(uint32_t p,struct DisassembleInfo * info)3654 static void print_protocol_list32_t(uint32_t p, struct DisassembleInfo *info) {
3655   struct protocol_list32_t pl;
3656   uint32_t q;
3657   struct protocol32_t pc;
3658   const char *r;
3659   uint32_t offset, xoffset, left, i;
3660   SectionRef S, xS;
3661   const char *name;
3662 
3663   r = get_pointer_32(p, offset, left, S, info);
3664   if (r == nullptr)
3665     return;
3666   memset(&pl, '\0', sizeof(struct protocol_list32_t));
3667   if (left < sizeof(struct protocol_list32_t)) {
3668     memcpy(&pl, r, left);
3669     outs() << "   (protocol_list_t entends past the end of the section)\n";
3670   } else
3671     memcpy(&pl, r, sizeof(struct protocol_list32_t));
3672   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3673     swapStruct(pl);
3674   outs() << "                      count " << pl.count << "\n";
3675 
3676   p += sizeof(struct protocol_list32_t);
3677   offset += sizeof(struct protocol_list32_t);
3678   for (i = 0; i < pl.count; i++) {
3679     r = get_pointer_32(p, offset, left, S, info);
3680     if (r == nullptr)
3681       return;
3682     q = 0;
3683     if (left < sizeof(uint32_t)) {
3684       memcpy(&q, r, left);
3685       outs() << "   (protocol_t * entends past the end of the section)\n";
3686     } else
3687       memcpy(&q, r, sizeof(uint32_t));
3688     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3689       sys::swapByteOrder(q);
3690     outs() << "\t\t      list[" << i << "] " << format("0x%" PRIx32, q)
3691            << " (struct protocol_t *)\n";
3692     r = get_pointer_32(q, offset, left, S, info);
3693     if (r == nullptr)
3694       return;
3695     memset(&pc, '\0', sizeof(struct protocol32_t));
3696     if (left < sizeof(struct protocol32_t)) {
3697       memcpy(&pc, r, left);
3698       outs() << "   (protocol_t entends past the end of the section)\n";
3699     } else
3700       memcpy(&pc, r, sizeof(struct protocol32_t));
3701     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3702       swapStruct(pc);
3703     outs() << "\t\t\t      isa " << format("0x%" PRIx32, pc.isa) << "\n";
3704     outs() << "\t\t\t     name " << format("0x%" PRIx32, pc.name);
3705     name = get_pointer_32(pc.name, xoffset, left, xS, info);
3706     if (name != nullptr)
3707       outs() << format(" %.*s", left, name);
3708     outs() << "\n";
3709     outs() << "\t\t\tprotocols " << format("0x%" PRIx32, pc.protocols) << "\n";
3710     outs() << "\t\t  instanceMethods "
3711            << format("0x%" PRIx32, pc.instanceMethods)
3712            << " (struct method_list_t *)\n";
3713     if (pc.instanceMethods != 0)
3714       print_method_list32_t(pc.instanceMethods, info, "\t");
3715     outs() << "\t\t     classMethods " << format("0x%" PRIx32, pc.classMethods)
3716            << " (struct method_list_t *)\n";
3717     if (pc.classMethods != 0)
3718       print_method_list32_t(pc.classMethods, info, "\t");
3719     outs() << "\t  optionalInstanceMethods "
3720            << format("0x%" PRIx32, pc.optionalInstanceMethods) << "\n";
3721     outs() << "\t     optionalClassMethods "
3722            << format("0x%" PRIx32, pc.optionalClassMethods) << "\n";
3723     outs() << "\t       instanceProperties "
3724            << format("0x%" PRIx32, pc.instanceProperties) << "\n";
3725     p += sizeof(uint32_t);
3726     offset += sizeof(uint32_t);
3727   }
3728 }
3729 
print_indent(uint32_t indent)3730 static void print_indent(uint32_t indent) {
3731   for (uint32_t i = 0; i < indent;) {
3732     if (indent - i >= 8) {
3733       outs() << "\t";
3734       i += 8;
3735     } else {
3736       for (uint32_t j = i; j < indent; j++)
3737         outs() << " ";
3738       return;
3739     }
3740   }
3741 }
3742 
print_method_description_list(uint32_t p,uint32_t indent,struct DisassembleInfo * info)3743 static bool print_method_description_list(uint32_t p, uint32_t indent,
3744                                           struct DisassembleInfo *info) {
3745   uint32_t offset, left, xleft;
3746   SectionRef S;
3747   struct objc_method_description_list_t mdl;
3748   struct objc_method_description_t md;
3749   const char *r, *list, *name;
3750   int32_t i;
3751 
3752   r = get_pointer_32(p, offset, left, S, info, true);
3753   if (r == nullptr)
3754     return true;
3755 
3756   outs() << "\n";
3757   if (left > sizeof(struct objc_method_description_list_t)) {
3758     memcpy(&mdl, r, sizeof(struct objc_method_description_list_t));
3759   } else {
3760     print_indent(indent);
3761     outs() << " objc_method_description_list extends past end of the section\n";
3762     memset(&mdl, '\0', sizeof(struct objc_method_description_list_t));
3763     memcpy(&mdl, r, left);
3764   }
3765   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3766     swapStruct(mdl);
3767 
3768   print_indent(indent);
3769   outs() << "        count " << mdl.count << "\n";
3770 
3771   list = r + sizeof(struct objc_method_description_list_t);
3772   for (i = 0; i < mdl.count; i++) {
3773     if ((i + 1) * sizeof(struct objc_method_description_t) > left) {
3774       print_indent(indent);
3775       outs() << " remaining list entries extend past the of the section\n";
3776       break;
3777     }
3778     print_indent(indent);
3779     outs() << "        list[" << i << "]\n";
3780     memcpy(&md, list + i * sizeof(struct objc_method_description_t),
3781            sizeof(struct objc_method_description_t));
3782     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3783       swapStruct(md);
3784 
3785     print_indent(indent);
3786     outs() << "             name " << format("0x%08" PRIx32, md.name);
3787     if (info->verbose) {
3788       name = get_pointer_32(md.name, offset, xleft, S, info, true);
3789       if (name != nullptr)
3790         outs() << format(" %.*s", xleft, name);
3791       else
3792         outs() << " (not in an __OBJC section)";
3793     }
3794     outs() << "\n";
3795 
3796     print_indent(indent);
3797     outs() << "            types " << format("0x%08" PRIx32, md.types);
3798     if (info->verbose) {
3799       name = get_pointer_32(md.types, offset, xleft, S, info, true);
3800       if (name != nullptr)
3801         outs() << format(" %.*s", xleft, name);
3802       else
3803         outs() << " (not in an __OBJC section)";
3804     }
3805     outs() << "\n";
3806   }
3807   return false;
3808 }
3809 
3810 static bool print_protocol_list(uint32_t p, uint32_t indent,
3811                                 struct DisassembleInfo *info);
3812 
print_protocol(uint32_t p,uint32_t indent,struct DisassembleInfo * info)3813 static bool print_protocol(uint32_t p, uint32_t indent,
3814                            struct DisassembleInfo *info) {
3815   uint32_t offset, left;
3816   SectionRef S;
3817   struct objc_protocol_t protocol;
3818   const char *r, *name;
3819 
3820   r = get_pointer_32(p, offset, left, S, info, true);
3821   if (r == nullptr)
3822     return true;
3823 
3824   outs() << "\n";
3825   if (left >= sizeof(struct objc_protocol_t)) {
3826     memcpy(&protocol, r, sizeof(struct objc_protocol_t));
3827   } else {
3828     print_indent(indent);
3829     outs() << "            Protocol extends past end of the section\n";
3830     memset(&protocol, '\0', sizeof(struct objc_protocol_t));
3831     memcpy(&protocol, r, left);
3832   }
3833   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3834     swapStruct(protocol);
3835 
3836   print_indent(indent);
3837   outs() << "              isa " << format("0x%08" PRIx32, protocol.isa)
3838          << "\n";
3839 
3840   print_indent(indent);
3841   outs() << "    protocol_name "
3842          << format("0x%08" PRIx32, protocol.protocol_name);
3843   if (info->verbose) {
3844     name = get_pointer_32(protocol.protocol_name, offset, left, S, info, true);
3845     if (name != nullptr)
3846       outs() << format(" %.*s", left, name);
3847     else
3848       outs() << " (not in an __OBJC section)";
3849   }
3850   outs() << "\n";
3851 
3852   print_indent(indent);
3853   outs() << "    protocol_list "
3854          << format("0x%08" PRIx32, protocol.protocol_list);
3855   if (print_protocol_list(protocol.protocol_list, indent + 4, info))
3856     outs() << " (not in an __OBJC section)\n";
3857 
3858   print_indent(indent);
3859   outs() << " instance_methods "
3860          << format("0x%08" PRIx32, protocol.instance_methods);
3861   if (print_method_description_list(protocol.instance_methods, indent, info))
3862     outs() << " (not in an __OBJC section)\n";
3863 
3864   print_indent(indent);
3865   outs() << "    class_methods "
3866          << format("0x%08" PRIx32, protocol.class_methods);
3867   if (print_method_description_list(protocol.class_methods, indent, info))
3868     outs() << " (not in an __OBJC section)\n";
3869 
3870   return false;
3871 }
3872 
print_protocol_list(uint32_t p,uint32_t indent,struct DisassembleInfo * info)3873 static bool print_protocol_list(uint32_t p, uint32_t indent,
3874                                 struct DisassembleInfo *info) {
3875   uint32_t offset, left, l;
3876   SectionRef S;
3877   struct objc_protocol_list_t protocol_list;
3878   const char *r, *list;
3879   int32_t i;
3880 
3881   r = get_pointer_32(p, offset, left, S, info, true);
3882   if (r == nullptr)
3883     return true;
3884 
3885   outs() << "\n";
3886   if (left > sizeof(struct objc_protocol_list_t)) {
3887     memcpy(&protocol_list, r, sizeof(struct objc_protocol_list_t));
3888   } else {
3889     outs() << "\t\t objc_protocol_list_t extends past end of the section\n";
3890     memset(&protocol_list, '\0', sizeof(struct objc_protocol_list_t));
3891     memcpy(&protocol_list, r, left);
3892   }
3893   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3894     swapStruct(protocol_list);
3895 
3896   print_indent(indent);
3897   outs() << "         next " << format("0x%08" PRIx32, protocol_list.next)
3898          << "\n";
3899   print_indent(indent);
3900   outs() << "        count " << protocol_list.count << "\n";
3901 
3902   list = r + sizeof(struct objc_protocol_list_t);
3903   for (i = 0; i < protocol_list.count; i++) {
3904     if ((i + 1) * sizeof(uint32_t) > left) {
3905       outs() << "\t\t remaining list entries extend past the of the section\n";
3906       break;
3907     }
3908     memcpy(&l, list + i * sizeof(uint32_t), sizeof(uint32_t));
3909     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3910       sys::swapByteOrder(l);
3911 
3912     print_indent(indent);
3913     outs() << "      list[" << i << "] " << format("0x%08" PRIx32, l);
3914     if (print_protocol(l, indent, info))
3915       outs() << "(not in an __OBJC section)\n";
3916   }
3917   return false;
3918 }
3919 
print_ivar_list64_t(uint64_t p,struct DisassembleInfo * info)3920 static void print_ivar_list64_t(uint64_t p, struct DisassembleInfo *info) {
3921   struct ivar_list64_t il;
3922   struct ivar64_t i;
3923   const char *r;
3924   uint32_t offset, xoffset, left, j;
3925   SectionRef S, xS;
3926   const char *name, *sym_name, *ivar_offset_p;
3927   uint64_t ivar_offset, n_value;
3928 
3929   r = get_pointer_64(p, offset, left, S, info);
3930   if (r == nullptr)
3931     return;
3932   memset(&il, '\0', sizeof(struct ivar_list64_t));
3933   if (left < sizeof(struct ivar_list64_t)) {
3934     memcpy(&il, r, left);
3935     outs() << "   (ivar_list_t entends past the end of the section)\n";
3936   } else
3937     memcpy(&il, r, sizeof(struct ivar_list64_t));
3938   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3939     swapStruct(il);
3940   outs() << "                    entsize " << il.entsize << "\n";
3941   outs() << "                      count " << il.count << "\n";
3942 
3943   p += sizeof(struct ivar_list64_t);
3944   offset += sizeof(struct ivar_list64_t);
3945   for (j = 0; j < il.count; j++) {
3946     r = get_pointer_64(p, offset, left, S, info);
3947     if (r == nullptr)
3948       return;
3949     memset(&i, '\0', sizeof(struct ivar64_t));
3950     if (left < sizeof(struct ivar64_t)) {
3951       memcpy(&i, r, left);
3952       outs() << "   (ivar_t entends past the end of the section)\n";
3953     } else
3954       memcpy(&i, r, sizeof(struct ivar64_t));
3955     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3956       swapStruct(i);
3957 
3958     outs() << "\t\t\t   offset ";
3959     sym_name = get_symbol_64(offset + offsetof(struct ivar64_t, offset), S,
3960                              info, n_value, i.offset);
3961     if (n_value != 0) {
3962       if (info->verbose && sym_name != nullptr)
3963         outs() << sym_name;
3964       else
3965         outs() << format("0x%" PRIx64, n_value);
3966       if (i.offset != 0)
3967         outs() << " + " << format("0x%" PRIx64, i.offset);
3968     } else
3969       outs() << format("0x%" PRIx64, i.offset);
3970     ivar_offset_p = get_pointer_64(i.offset + n_value, xoffset, left, xS, info);
3971     if (ivar_offset_p != nullptr && left >= sizeof(*ivar_offset_p)) {
3972       memcpy(&ivar_offset, ivar_offset_p, sizeof(ivar_offset));
3973       if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
3974         sys::swapByteOrder(ivar_offset);
3975       outs() << " " << ivar_offset << "\n";
3976     } else
3977       outs() << "\n";
3978 
3979     outs() << "\t\t\t     name ";
3980     sym_name = get_symbol_64(offset + offsetof(struct ivar64_t, name), S, info,
3981                              n_value, i.name);
3982     if (n_value != 0) {
3983       if (info->verbose && sym_name != nullptr)
3984         outs() << sym_name;
3985       else
3986         outs() << format("0x%" PRIx64, n_value);
3987       if (i.name != 0)
3988         outs() << " + " << format("0x%" PRIx64, i.name);
3989     } else
3990       outs() << format("0x%" PRIx64, i.name);
3991     name = get_pointer_64(i.name + n_value, xoffset, left, xS, info);
3992     if (name != nullptr)
3993       outs() << format(" %.*s", left, name);
3994     outs() << "\n";
3995 
3996     outs() << "\t\t\t     type ";
3997     sym_name = get_symbol_64(offset + offsetof(struct ivar64_t, type), S, info,
3998                              n_value, i.name);
3999     name = get_pointer_64(i.type + n_value, xoffset, left, xS, info);
4000     if (n_value != 0) {
4001       if (info->verbose && sym_name != nullptr)
4002         outs() << sym_name;
4003       else
4004         outs() << format("0x%" PRIx64, n_value);
4005       if (i.type != 0)
4006         outs() << " + " << format("0x%" PRIx64, i.type);
4007     } else
4008       outs() << format("0x%" PRIx64, i.type);
4009     if (name != nullptr)
4010       outs() << format(" %.*s", left, name);
4011     outs() << "\n";
4012 
4013     outs() << "\t\t\talignment " << i.alignment << "\n";
4014     outs() << "\t\t\t     size " << i.size << "\n";
4015 
4016     p += sizeof(struct ivar64_t);
4017     offset += sizeof(struct ivar64_t);
4018   }
4019 }
4020 
print_ivar_list32_t(uint32_t p,struct DisassembleInfo * info)4021 static void print_ivar_list32_t(uint32_t p, struct DisassembleInfo *info) {
4022   struct ivar_list32_t il;
4023   struct ivar32_t i;
4024   const char *r;
4025   uint32_t offset, xoffset, left, j;
4026   SectionRef S, xS;
4027   const char *name, *ivar_offset_p;
4028   uint32_t ivar_offset;
4029 
4030   r = get_pointer_32(p, offset, left, S, info);
4031   if (r == nullptr)
4032     return;
4033   memset(&il, '\0', sizeof(struct ivar_list32_t));
4034   if (left < sizeof(struct ivar_list32_t)) {
4035     memcpy(&il, r, left);
4036     outs() << "   (ivar_list_t entends past the end of the section)\n";
4037   } else
4038     memcpy(&il, r, sizeof(struct ivar_list32_t));
4039   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4040     swapStruct(il);
4041   outs() << "                    entsize " << il.entsize << "\n";
4042   outs() << "                      count " << il.count << "\n";
4043 
4044   p += sizeof(struct ivar_list32_t);
4045   offset += sizeof(struct ivar_list32_t);
4046   for (j = 0; j < il.count; j++) {
4047     r = get_pointer_32(p, offset, left, S, info);
4048     if (r == nullptr)
4049       return;
4050     memset(&i, '\0', sizeof(struct ivar32_t));
4051     if (left < sizeof(struct ivar32_t)) {
4052       memcpy(&i, r, left);
4053       outs() << "   (ivar_t entends past the end of the section)\n";
4054     } else
4055       memcpy(&i, r, sizeof(struct ivar32_t));
4056     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4057       swapStruct(i);
4058 
4059     outs() << "\t\t\t   offset " << format("0x%" PRIx32, i.offset);
4060     ivar_offset_p = get_pointer_32(i.offset, xoffset, left, xS, info);
4061     if (ivar_offset_p != nullptr && left >= sizeof(*ivar_offset_p)) {
4062       memcpy(&ivar_offset, ivar_offset_p, sizeof(ivar_offset));
4063       if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4064         sys::swapByteOrder(ivar_offset);
4065       outs() << " " << ivar_offset << "\n";
4066     } else
4067       outs() << "\n";
4068 
4069     outs() << "\t\t\t     name " << format("0x%" PRIx32, i.name);
4070     name = get_pointer_32(i.name, xoffset, left, xS, info);
4071     if (name != nullptr)
4072       outs() << format(" %.*s", left, name);
4073     outs() << "\n";
4074 
4075     outs() << "\t\t\t     type " << format("0x%" PRIx32, i.type);
4076     name = get_pointer_32(i.type, xoffset, left, xS, info);
4077     if (name != nullptr)
4078       outs() << format(" %.*s", left, name);
4079     outs() << "\n";
4080 
4081     outs() << "\t\t\talignment " << i.alignment << "\n";
4082     outs() << "\t\t\t     size " << i.size << "\n";
4083 
4084     p += sizeof(struct ivar32_t);
4085     offset += sizeof(struct ivar32_t);
4086   }
4087 }
4088 
print_objc_property_list64(uint64_t p,struct DisassembleInfo * info)4089 static void print_objc_property_list64(uint64_t p,
4090                                        struct DisassembleInfo *info) {
4091   struct objc_property_list64 opl;
4092   struct objc_property64 op;
4093   const char *r;
4094   uint32_t offset, xoffset, left, j;
4095   SectionRef S, xS;
4096   const char *name, *sym_name;
4097   uint64_t n_value;
4098 
4099   r = get_pointer_64(p, offset, left, S, info);
4100   if (r == nullptr)
4101     return;
4102   memset(&opl, '\0', sizeof(struct objc_property_list64));
4103   if (left < sizeof(struct objc_property_list64)) {
4104     memcpy(&opl, r, left);
4105     outs() << "   (objc_property_list entends past the end of the section)\n";
4106   } else
4107     memcpy(&opl, r, sizeof(struct objc_property_list64));
4108   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4109     swapStruct(opl);
4110   outs() << "                    entsize " << opl.entsize << "\n";
4111   outs() << "                      count " << opl.count << "\n";
4112 
4113   p += sizeof(struct objc_property_list64);
4114   offset += sizeof(struct objc_property_list64);
4115   for (j = 0; j < opl.count; j++) {
4116     r = get_pointer_64(p, offset, left, S, info);
4117     if (r == nullptr)
4118       return;
4119     memset(&op, '\0', sizeof(struct objc_property64));
4120     if (left < sizeof(struct objc_property64)) {
4121       memcpy(&op, r, left);
4122       outs() << "   (objc_property entends past the end of the section)\n";
4123     } else
4124       memcpy(&op, r, sizeof(struct objc_property64));
4125     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4126       swapStruct(op);
4127 
4128     outs() << "\t\t\t     name ";
4129     sym_name = get_symbol_64(offset + offsetof(struct objc_property64, name), S,
4130                              info, n_value, op.name);
4131     if (n_value != 0) {
4132       if (info->verbose && sym_name != nullptr)
4133         outs() << sym_name;
4134       else
4135         outs() << format("0x%" PRIx64, n_value);
4136       if (op.name != 0)
4137         outs() << " + " << format("0x%" PRIx64, op.name);
4138     } else
4139       outs() << format("0x%" PRIx64, op.name);
4140     name = get_pointer_64(op.name + n_value, xoffset, left, xS, info);
4141     if (name != nullptr)
4142       outs() << format(" %.*s", left, name);
4143     outs() << "\n";
4144 
4145     outs() << "\t\t\tattributes ";
4146     sym_name =
4147         get_symbol_64(offset + offsetof(struct objc_property64, attributes), S,
4148                       info, n_value, op.attributes);
4149     if (n_value != 0) {
4150       if (info->verbose && sym_name != nullptr)
4151         outs() << sym_name;
4152       else
4153         outs() << format("0x%" PRIx64, n_value);
4154       if (op.attributes != 0)
4155         outs() << " + " << format("0x%" PRIx64, op.attributes);
4156     } else
4157       outs() << format("0x%" PRIx64, op.attributes);
4158     name = get_pointer_64(op.attributes + n_value, xoffset, left, xS, info);
4159     if (name != nullptr)
4160       outs() << format(" %.*s", left, name);
4161     outs() << "\n";
4162 
4163     p += sizeof(struct objc_property64);
4164     offset += sizeof(struct objc_property64);
4165   }
4166 }
4167 
print_objc_property_list32(uint32_t p,struct DisassembleInfo * info)4168 static void print_objc_property_list32(uint32_t p,
4169                                        struct DisassembleInfo *info) {
4170   struct objc_property_list32 opl;
4171   struct objc_property32 op;
4172   const char *r;
4173   uint32_t offset, xoffset, left, j;
4174   SectionRef S, xS;
4175   const char *name;
4176 
4177   r = get_pointer_32(p, offset, left, S, info);
4178   if (r == nullptr)
4179     return;
4180   memset(&opl, '\0', sizeof(struct objc_property_list32));
4181   if (left < sizeof(struct objc_property_list32)) {
4182     memcpy(&opl, r, left);
4183     outs() << "   (objc_property_list entends past the end of the section)\n";
4184   } else
4185     memcpy(&opl, r, sizeof(struct objc_property_list32));
4186   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4187     swapStruct(opl);
4188   outs() << "                    entsize " << opl.entsize << "\n";
4189   outs() << "                      count " << opl.count << "\n";
4190 
4191   p += sizeof(struct objc_property_list32);
4192   offset += sizeof(struct objc_property_list32);
4193   for (j = 0; j < opl.count; j++) {
4194     r = get_pointer_32(p, offset, left, S, info);
4195     if (r == nullptr)
4196       return;
4197     memset(&op, '\0', sizeof(struct objc_property32));
4198     if (left < sizeof(struct objc_property32)) {
4199       memcpy(&op, r, left);
4200       outs() << "   (objc_property entends past the end of the section)\n";
4201     } else
4202       memcpy(&op, r, sizeof(struct objc_property32));
4203     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4204       swapStruct(op);
4205 
4206     outs() << "\t\t\t     name " << format("0x%" PRIx32, op.name);
4207     name = get_pointer_32(op.name, xoffset, left, xS, info);
4208     if (name != nullptr)
4209       outs() << format(" %.*s", left, name);
4210     outs() << "\n";
4211 
4212     outs() << "\t\t\tattributes " << format("0x%" PRIx32, op.attributes);
4213     name = get_pointer_32(op.attributes, xoffset, left, xS, info);
4214     if (name != nullptr)
4215       outs() << format(" %.*s", left, name);
4216     outs() << "\n";
4217 
4218     p += sizeof(struct objc_property32);
4219     offset += sizeof(struct objc_property32);
4220   }
4221 }
4222 
print_class_ro64_t(uint64_t p,struct DisassembleInfo * info,bool & is_meta_class)4223 static bool print_class_ro64_t(uint64_t p, struct DisassembleInfo *info,
4224                                bool &is_meta_class) {
4225   struct class_ro64_t cro;
4226   const char *r;
4227   uint32_t offset, xoffset, left;
4228   SectionRef S, xS;
4229   const char *name, *sym_name;
4230   uint64_t n_value;
4231 
4232   r = get_pointer_64(p, offset, left, S, info);
4233   if (r == nullptr || left < sizeof(struct class_ro64_t))
4234     return false;
4235   memset(&cro, '\0', sizeof(struct class_ro64_t));
4236   if (left < sizeof(struct class_ro64_t)) {
4237     memcpy(&cro, r, left);
4238     outs() << "   (class_ro_t entends past the end of the section)\n";
4239   } else
4240     memcpy(&cro, r, sizeof(struct class_ro64_t));
4241   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4242     swapStruct(cro);
4243   outs() << "                    flags " << format("0x%" PRIx32, cro.flags);
4244   if (cro.flags & RO_META)
4245     outs() << " RO_META";
4246   if (cro.flags & RO_ROOT)
4247     outs() << " RO_ROOT";
4248   if (cro.flags & RO_HAS_CXX_STRUCTORS)
4249     outs() << " RO_HAS_CXX_STRUCTORS";
4250   outs() << "\n";
4251   outs() << "            instanceStart " << cro.instanceStart << "\n";
4252   outs() << "             instanceSize " << cro.instanceSize << "\n";
4253   outs() << "                 reserved " << format("0x%" PRIx32, cro.reserved)
4254          << "\n";
4255   outs() << "               ivarLayout " << format("0x%" PRIx64, cro.ivarLayout)
4256          << "\n";
4257   print_layout_map64(cro.ivarLayout, info);
4258 
4259   outs() << "                     name ";
4260   sym_name = get_symbol_64(offset + offsetof(struct class_ro64_t, name), S,
4261                            info, n_value, cro.name);
4262   if (n_value != 0) {
4263     if (info->verbose && sym_name != nullptr)
4264       outs() << sym_name;
4265     else
4266       outs() << format("0x%" PRIx64, n_value);
4267     if (cro.name != 0)
4268       outs() << " + " << format("0x%" PRIx64, cro.name);
4269   } else
4270     outs() << format("0x%" PRIx64, cro.name);
4271   name = get_pointer_64(cro.name + n_value, xoffset, left, xS, info);
4272   if (name != nullptr)
4273     outs() << format(" %.*s", left, name);
4274   outs() << "\n";
4275 
4276   outs() << "              baseMethods ";
4277   sym_name = get_symbol_64(offset + offsetof(struct class_ro64_t, baseMethods),
4278                            S, info, n_value, cro.baseMethods);
4279   if (n_value != 0) {
4280     if (info->verbose && sym_name != nullptr)
4281       outs() << sym_name;
4282     else
4283       outs() << format("0x%" PRIx64, n_value);
4284     if (cro.baseMethods != 0)
4285       outs() << " + " << format("0x%" PRIx64, cro.baseMethods);
4286   } else
4287     outs() << format("0x%" PRIx64, cro.baseMethods);
4288   outs() << " (struct method_list_t *)\n";
4289   if (cro.baseMethods + n_value != 0)
4290     print_method_list64_t(cro.baseMethods + n_value, info, "");
4291 
4292   outs() << "            baseProtocols ";
4293   sym_name =
4294       get_symbol_64(offset + offsetof(struct class_ro64_t, baseProtocols), S,
4295                     info, n_value, cro.baseProtocols);
4296   if (n_value != 0) {
4297     if (info->verbose && sym_name != nullptr)
4298       outs() << sym_name;
4299     else
4300       outs() << format("0x%" PRIx64, n_value);
4301     if (cro.baseProtocols != 0)
4302       outs() << " + " << format("0x%" PRIx64, cro.baseProtocols);
4303   } else
4304     outs() << format("0x%" PRIx64, cro.baseProtocols);
4305   outs() << "\n";
4306   if (cro.baseProtocols + n_value != 0)
4307     print_protocol_list64_t(cro.baseProtocols + n_value, info);
4308 
4309   outs() << "                    ivars ";
4310   sym_name = get_symbol_64(offset + offsetof(struct class_ro64_t, ivars), S,
4311                            info, n_value, cro.ivars);
4312   if (n_value != 0) {
4313     if (info->verbose && sym_name != nullptr)
4314       outs() << sym_name;
4315     else
4316       outs() << format("0x%" PRIx64, n_value);
4317     if (cro.ivars != 0)
4318       outs() << " + " << format("0x%" PRIx64, cro.ivars);
4319   } else
4320     outs() << format("0x%" PRIx64, cro.ivars);
4321   outs() << "\n";
4322   if (cro.ivars + n_value != 0)
4323     print_ivar_list64_t(cro.ivars + n_value, info);
4324 
4325   outs() << "           weakIvarLayout ";
4326   sym_name =
4327       get_symbol_64(offset + offsetof(struct class_ro64_t, weakIvarLayout), S,
4328                     info, n_value, cro.weakIvarLayout);
4329   if (n_value != 0) {
4330     if (info->verbose && sym_name != nullptr)
4331       outs() << sym_name;
4332     else
4333       outs() << format("0x%" PRIx64, n_value);
4334     if (cro.weakIvarLayout != 0)
4335       outs() << " + " << format("0x%" PRIx64, cro.weakIvarLayout);
4336   } else
4337     outs() << format("0x%" PRIx64, cro.weakIvarLayout);
4338   outs() << "\n";
4339   print_layout_map64(cro.weakIvarLayout + n_value, info);
4340 
4341   outs() << "           baseProperties ";
4342   sym_name =
4343       get_symbol_64(offset + offsetof(struct class_ro64_t, baseProperties), S,
4344                     info, n_value, cro.baseProperties);
4345   if (n_value != 0) {
4346     if (info->verbose && sym_name != nullptr)
4347       outs() << sym_name;
4348     else
4349       outs() << format("0x%" PRIx64, n_value);
4350     if (cro.baseProperties != 0)
4351       outs() << " + " << format("0x%" PRIx64, cro.baseProperties);
4352   } else
4353     outs() << format("0x%" PRIx64, cro.baseProperties);
4354   outs() << "\n";
4355   if (cro.baseProperties + n_value != 0)
4356     print_objc_property_list64(cro.baseProperties + n_value, info);
4357 
4358   is_meta_class = (cro.flags & RO_META) != 0;
4359   return true;
4360 }
4361 
print_class_ro32_t(uint32_t p,struct DisassembleInfo * info,bool & is_meta_class)4362 static bool print_class_ro32_t(uint32_t p, struct DisassembleInfo *info,
4363                                bool &is_meta_class) {
4364   struct class_ro32_t cro;
4365   const char *r;
4366   uint32_t offset, xoffset, left;
4367   SectionRef S, xS;
4368   const char *name;
4369 
4370   r = get_pointer_32(p, offset, left, S, info);
4371   if (r == nullptr)
4372     return false;
4373   memset(&cro, '\0', sizeof(struct class_ro32_t));
4374   if (left < sizeof(struct class_ro32_t)) {
4375     memcpy(&cro, r, left);
4376     outs() << "   (class_ro_t entends past the end of the section)\n";
4377   } else
4378     memcpy(&cro, r, sizeof(struct class_ro32_t));
4379   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4380     swapStruct(cro);
4381   outs() << "                    flags " << format("0x%" PRIx32, cro.flags);
4382   if (cro.flags & RO_META)
4383     outs() << " RO_META";
4384   if (cro.flags & RO_ROOT)
4385     outs() << " RO_ROOT";
4386   if (cro.flags & RO_HAS_CXX_STRUCTORS)
4387     outs() << " RO_HAS_CXX_STRUCTORS";
4388   outs() << "\n";
4389   outs() << "            instanceStart " << cro.instanceStart << "\n";
4390   outs() << "             instanceSize " << cro.instanceSize << "\n";
4391   outs() << "               ivarLayout " << format("0x%" PRIx32, cro.ivarLayout)
4392          << "\n";
4393   print_layout_map32(cro.ivarLayout, info);
4394 
4395   outs() << "                     name " << format("0x%" PRIx32, cro.name);
4396   name = get_pointer_32(cro.name, xoffset, left, xS, info);
4397   if (name != nullptr)
4398     outs() << format(" %.*s", left, name);
4399   outs() << "\n";
4400 
4401   outs() << "              baseMethods "
4402          << format("0x%" PRIx32, cro.baseMethods)
4403          << " (struct method_list_t *)\n";
4404   if (cro.baseMethods != 0)
4405     print_method_list32_t(cro.baseMethods, info, "");
4406 
4407   outs() << "            baseProtocols "
4408          << format("0x%" PRIx32, cro.baseProtocols) << "\n";
4409   if (cro.baseProtocols != 0)
4410     print_protocol_list32_t(cro.baseProtocols, info);
4411   outs() << "                    ivars " << format("0x%" PRIx32, cro.ivars)
4412          << "\n";
4413   if (cro.ivars != 0)
4414     print_ivar_list32_t(cro.ivars, info);
4415   outs() << "           weakIvarLayout "
4416          << format("0x%" PRIx32, cro.weakIvarLayout) << "\n";
4417   print_layout_map32(cro.weakIvarLayout, info);
4418   outs() << "           baseProperties "
4419          << format("0x%" PRIx32, cro.baseProperties) << "\n";
4420   if (cro.baseProperties != 0)
4421     print_objc_property_list32(cro.baseProperties, info);
4422   is_meta_class = (cro.flags & RO_META) != 0;
4423   return true;
4424 }
4425 
print_class64_t(uint64_t p,struct DisassembleInfo * info)4426 static void print_class64_t(uint64_t p, struct DisassembleInfo *info) {
4427   struct class64_t c;
4428   const char *r;
4429   uint32_t offset, left;
4430   SectionRef S;
4431   const char *name;
4432   uint64_t isa_n_value, n_value;
4433 
4434   r = get_pointer_64(p, offset, left, S, info);
4435   if (r == nullptr || left < sizeof(struct class64_t))
4436     return;
4437   memset(&c, '\0', sizeof(struct class64_t));
4438   if (left < sizeof(struct class64_t)) {
4439     memcpy(&c, r, left);
4440     outs() << "   (class_t entends past the end of the section)\n";
4441   } else
4442     memcpy(&c, r, sizeof(struct class64_t));
4443   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4444     swapStruct(c);
4445 
4446   outs() << "           isa " << format("0x%" PRIx64, c.isa);
4447   name = get_symbol_64(offset + offsetof(struct class64_t, isa), S, info,
4448                        isa_n_value, c.isa);
4449   if (name != nullptr)
4450     outs() << " " << name;
4451   outs() << "\n";
4452 
4453   outs() << "    superclass " << format("0x%" PRIx64, c.superclass);
4454   name = get_symbol_64(offset + offsetof(struct class64_t, superclass), S, info,
4455                        n_value, c.superclass);
4456   if (name != nullptr)
4457     outs() << " " << name;
4458   outs() << "\n";
4459 
4460   outs() << "         cache " << format("0x%" PRIx64, c.cache);
4461   name = get_symbol_64(offset + offsetof(struct class64_t, cache), S, info,
4462                        n_value, c.cache);
4463   if (name != nullptr)
4464     outs() << " " << name;
4465   outs() << "\n";
4466 
4467   outs() << "        vtable " << format("0x%" PRIx64, c.vtable);
4468   name = get_symbol_64(offset + offsetof(struct class64_t, vtable), S, info,
4469                        n_value, c.vtable);
4470   if (name != nullptr)
4471     outs() << " " << name;
4472   outs() << "\n";
4473 
4474   name = get_symbol_64(offset + offsetof(struct class64_t, data), S, info,
4475                        n_value, c.data);
4476   outs() << "          data ";
4477   if (n_value != 0) {
4478     if (info->verbose && name != nullptr)
4479       outs() << name;
4480     else
4481       outs() << format("0x%" PRIx64, n_value);
4482     if (c.data != 0)
4483       outs() << " + " << format("0x%" PRIx64, c.data);
4484   } else
4485     outs() << format("0x%" PRIx64, c.data);
4486   outs() << " (struct class_ro_t *)";
4487 
4488   // This is a Swift class if some of the low bits of the pointer are set.
4489   if ((c.data + n_value) & 0x7)
4490     outs() << " Swift class";
4491   outs() << "\n";
4492   bool is_meta_class;
4493   if (!print_class_ro64_t((c.data + n_value) & ~0x7, info, is_meta_class))
4494     return;
4495 
4496   if (!is_meta_class &&
4497       c.isa + isa_n_value != p &&
4498       c.isa + isa_n_value != 0 &&
4499       info->depth < 100) {
4500       info->depth++;
4501       outs() << "Meta Class\n";
4502       print_class64_t(c.isa + isa_n_value, info);
4503   }
4504 }
4505 
print_class32_t(uint32_t p,struct DisassembleInfo * info)4506 static void print_class32_t(uint32_t p, struct DisassembleInfo *info) {
4507   struct class32_t c;
4508   const char *r;
4509   uint32_t offset, left;
4510   SectionRef S;
4511   const char *name;
4512 
4513   r = get_pointer_32(p, offset, left, S, info);
4514   if (r == nullptr)
4515     return;
4516   memset(&c, '\0', sizeof(struct class32_t));
4517   if (left < sizeof(struct class32_t)) {
4518     memcpy(&c, r, left);
4519     outs() << "   (class_t entends past the end of the section)\n";
4520   } else
4521     memcpy(&c, r, sizeof(struct class32_t));
4522   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4523     swapStruct(c);
4524 
4525   outs() << "           isa " << format("0x%" PRIx32, c.isa);
4526   name =
4527       get_symbol_32(offset + offsetof(struct class32_t, isa), S, info, c.isa);
4528   if (name != nullptr)
4529     outs() << " " << name;
4530   outs() << "\n";
4531 
4532   outs() << "    superclass " << format("0x%" PRIx32, c.superclass);
4533   name = get_symbol_32(offset + offsetof(struct class32_t, superclass), S, info,
4534                        c.superclass);
4535   if (name != nullptr)
4536     outs() << " " << name;
4537   outs() << "\n";
4538 
4539   outs() << "         cache " << format("0x%" PRIx32, c.cache);
4540   name = get_symbol_32(offset + offsetof(struct class32_t, cache), S, info,
4541                        c.cache);
4542   if (name != nullptr)
4543     outs() << " " << name;
4544   outs() << "\n";
4545 
4546   outs() << "        vtable " << format("0x%" PRIx32, c.vtable);
4547   name = get_symbol_32(offset + offsetof(struct class32_t, vtable), S, info,
4548                        c.vtable);
4549   if (name != nullptr)
4550     outs() << " " << name;
4551   outs() << "\n";
4552 
4553   name =
4554       get_symbol_32(offset + offsetof(struct class32_t, data), S, info, c.data);
4555   outs() << "          data " << format("0x%" PRIx32, c.data)
4556          << " (struct class_ro_t *)";
4557 
4558   // This is a Swift class if some of the low bits of the pointer are set.
4559   if (c.data & 0x3)
4560     outs() << " Swift class";
4561   outs() << "\n";
4562   bool is_meta_class;
4563   if (!print_class_ro32_t(c.data & ~0x3, info, is_meta_class))
4564     return;
4565 
4566   if (!is_meta_class) {
4567     outs() << "Meta Class\n";
4568     print_class32_t(c.isa, info);
4569   }
4570 }
4571 
print_objc_class_t(struct objc_class_t * objc_class,struct DisassembleInfo * info)4572 static void print_objc_class_t(struct objc_class_t *objc_class,
4573                                struct DisassembleInfo *info) {
4574   uint32_t offset, left, xleft;
4575   const char *name, *p, *ivar_list;
4576   SectionRef S;
4577   int32_t i;
4578   struct objc_ivar_list_t objc_ivar_list;
4579   struct objc_ivar_t ivar;
4580 
4581   outs() << "\t\t      isa " << format("0x%08" PRIx32, objc_class->isa);
4582   if (info->verbose && CLS_GETINFO(objc_class, CLS_META)) {
4583     name = get_pointer_32(objc_class->isa, offset, left, S, info, true);
4584     if (name != nullptr)
4585       outs() << format(" %.*s", left, name);
4586     else
4587       outs() << " (not in an __OBJC section)";
4588   }
4589   outs() << "\n";
4590 
4591   outs() << "\t      super_class "
4592          << format("0x%08" PRIx32, objc_class->super_class);
4593   if (info->verbose) {
4594     name = get_pointer_32(objc_class->super_class, offset, left, S, info, true);
4595     if (name != nullptr)
4596       outs() << format(" %.*s", left, name);
4597     else
4598       outs() << " (not in an __OBJC section)";
4599   }
4600   outs() << "\n";
4601 
4602   outs() << "\t\t     name " << format("0x%08" PRIx32, objc_class->name);
4603   if (info->verbose) {
4604     name = get_pointer_32(objc_class->name, offset, left, S, info, true);
4605     if (name != nullptr)
4606       outs() << format(" %.*s", left, name);
4607     else
4608       outs() << " (not in an __OBJC section)";
4609   }
4610   outs() << "\n";
4611 
4612   outs() << "\t\t  version " << format("0x%08" PRIx32, objc_class->version)
4613          << "\n";
4614 
4615   outs() << "\t\t     info " << format("0x%08" PRIx32, objc_class->info);
4616   if (info->verbose) {
4617     if (CLS_GETINFO(objc_class, CLS_CLASS))
4618       outs() << " CLS_CLASS";
4619     else if (CLS_GETINFO(objc_class, CLS_META))
4620       outs() << " CLS_META";
4621   }
4622   outs() << "\n";
4623 
4624   outs() << "\t    instance_size "
4625          << format("0x%08" PRIx32, objc_class->instance_size) << "\n";
4626 
4627   p = get_pointer_32(objc_class->ivars, offset, left, S, info, true);
4628   outs() << "\t\t    ivars " << format("0x%08" PRIx32, objc_class->ivars);
4629   if (p != nullptr) {
4630     if (left > sizeof(struct objc_ivar_list_t)) {
4631       outs() << "\n";
4632       memcpy(&objc_ivar_list, p, sizeof(struct objc_ivar_list_t));
4633     } else {
4634       outs() << " (entends past the end of the section)\n";
4635       memset(&objc_ivar_list, '\0', sizeof(struct objc_ivar_list_t));
4636       memcpy(&objc_ivar_list, p, left);
4637     }
4638     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4639       swapStruct(objc_ivar_list);
4640     outs() << "\t\t       ivar_count " << objc_ivar_list.ivar_count << "\n";
4641     ivar_list = p + sizeof(struct objc_ivar_list_t);
4642     for (i = 0; i < objc_ivar_list.ivar_count; i++) {
4643       if ((i + 1) * sizeof(struct objc_ivar_t) > left) {
4644         outs() << "\t\t remaining ivar's extend past the of the section\n";
4645         break;
4646       }
4647       memcpy(&ivar, ivar_list + i * sizeof(struct objc_ivar_t),
4648              sizeof(struct objc_ivar_t));
4649       if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4650         swapStruct(ivar);
4651 
4652       outs() << "\t\t\tivar_name " << format("0x%08" PRIx32, ivar.ivar_name);
4653       if (info->verbose) {
4654         name = get_pointer_32(ivar.ivar_name, offset, xleft, S, info, true);
4655         if (name != nullptr)
4656           outs() << format(" %.*s", xleft, name);
4657         else
4658           outs() << " (not in an __OBJC section)";
4659       }
4660       outs() << "\n";
4661 
4662       outs() << "\t\t\tivar_type " << format("0x%08" PRIx32, ivar.ivar_type);
4663       if (info->verbose) {
4664         name = get_pointer_32(ivar.ivar_type, offset, xleft, S, info, true);
4665         if (name != nullptr)
4666           outs() << format(" %.*s", xleft, name);
4667         else
4668           outs() << " (not in an __OBJC section)";
4669       }
4670       outs() << "\n";
4671 
4672       outs() << "\t\t      ivar_offset "
4673              << format("0x%08" PRIx32, ivar.ivar_offset) << "\n";
4674     }
4675   } else {
4676     outs() << " (not in an __OBJC section)\n";
4677   }
4678 
4679   outs() << "\t\t  methods " << format("0x%08" PRIx32, objc_class->methodLists);
4680   if (print_method_list(objc_class->methodLists, info))
4681     outs() << " (not in an __OBJC section)\n";
4682 
4683   outs() << "\t\t    cache " << format("0x%08" PRIx32, objc_class->cache)
4684          << "\n";
4685 
4686   outs() << "\t\tprotocols " << format("0x%08" PRIx32, objc_class->protocols);
4687   if (print_protocol_list(objc_class->protocols, 16, info))
4688     outs() << " (not in an __OBJC section)\n";
4689 }
4690 
print_objc_objc_category_t(struct objc_category_t * objc_category,struct DisassembleInfo * info)4691 static void print_objc_objc_category_t(struct objc_category_t *objc_category,
4692                                        struct DisassembleInfo *info) {
4693   uint32_t offset, left;
4694   const char *name;
4695   SectionRef S;
4696 
4697   outs() << "\t       category name "
4698          << format("0x%08" PRIx32, objc_category->category_name);
4699   if (info->verbose) {
4700     name = get_pointer_32(objc_category->category_name, offset, left, S, info,
4701                           true);
4702     if (name != nullptr)
4703       outs() << format(" %.*s", left, name);
4704     else
4705       outs() << " (not in an __OBJC section)";
4706   }
4707   outs() << "\n";
4708 
4709   outs() << "\t\t  class name "
4710          << format("0x%08" PRIx32, objc_category->class_name);
4711   if (info->verbose) {
4712     name =
4713         get_pointer_32(objc_category->class_name, offset, left, S, info, true);
4714     if (name != nullptr)
4715       outs() << format(" %.*s", left, name);
4716     else
4717       outs() << " (not in an __OBJC section)";
4718   }
4719   outs() << "\n";
4720 
4721   outs() << "\t    instance methods "
4722          << format("0x%08" PRIx32, objc_category->instance_methods);
4723   if (print_method_list(objc_category->instance_methods, info))
4724     outs() << " (not in an __OBJC section)\n";
4725 
4726   outs() << "\t       class methods "
4727          << format("0x%08" PRIx32, objc_category->class_methods);
4728   if (print_method_list(objc_category->class_methods, info))
4729     outs() << " (not in an __OBJC section)\n";
4730 }
4731 
print_category64_t(uint64_t p,struct DisassembleInfo * info)4732 static void print_category64_t(uint64_t p, struct DisassembleInfo *info) {
4733   struct category64_t c;
4734   const char *r;
4735   uint32_t offset, xoffset, left;
4736   SectionRef S, xS;
4737   const char *name, *sym_name;
4738   uint64_t n_value;
4739 
4740   r = get_pointer_64(p, offset, left, S, info);
4741   if (r == nullptr)
4742     return;
4743   memset(&c, '\0', sizeof(struct category64_t));
4744   if (left < sizeof(struct category64_t)) {
4745     memcpy(&c, r, left);
4746     outs() << "   (category_t entends past the end of the section)\n";
4747   } else
4748     memcpy(&c, r, sizeof(struct category64_t));
4749   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4750     swapStruct(c);
4751 
4752   outs() << "              name ";
4753   sym_name = get_symbol_64(offset + offsetof(struct category64_t, name), S,
4754                            info, n_value, c.name);
4755   if (n_value != 0) {
4756     if (info->verbose && sym_name != nullptr)
4757       outs() << sym_name;
4758     else
4759       outs() << format("0x%" PRIx64, n_value);
4760     if (c.name != 0)
4761       outs() << " + " << format("0x%" PRIx64, c.name);
4762   } else
4763     outs() << format("0x%" PRIx64, c.name);
4764   name = get_pointer_64(c.name + n_value, xoffset, left, xS, info);
4765   if (name != nullptr)
4766     outs() << format(" %.*s", left, name);
4767   outs() << "\n";
4768 
4769   outs() << "               cls ";
4770   sym_name = get_symbol_64(offset + offsetof(struct category64_t, cls), S, info,
4771                            n_value, c.cls);
4772   if (n_value != 0) {
4773     if (info->verbose && sym_name != nullptr)
4774       outs() << sym_name;
4775     else
4776       outs() << format("0x%" PRIx64, n_value);
4777     if (c.cls != 0)
4778       outs() << " + " << format("0x%" PRIx64, c.cls);
4779   } else
4780     outs() << format("0x%" PRIx64, c.cls);
4781   outs() << "\n";
4782   if (c.cls + n_value != 0)
4783     print_class64_t(c.cls + n_value, info);
4784 
4785   outs() << "   instanceMethods ";
4786   sym_name =
4787       get_symbol_64(offset + offsetof(struct category64_t, instanceMethods), S,
4788                     info, n_value, c.instanceMethods);
4789   if (n_value != 0) {
4790     if (info->verbose && sym_name != nullptr)
4791       outs() << sym_name;
4792     else
4793       outs() << format("0x%" PRIx64, n_value);
4794     if (c.instanceMethods != 0)
4795       outs() << " + " << format("0x%" PRIx64, c.instanceMethods);
4796   } else
4797     outs() << format("0x%" PRIx64, c.instanceMethods);
4798   outs() << "\n";
4799   if (c.instanceMethods + n_value != 0)
4800     print_method_list64_t(c.instanceMethods + n_value, info, "");
4801 
4802   outs() << "      classMethods ";
4803   sym_name = get_symbol_64(offset + offsetof(struct category64_t, classMethods),
4804                            S, info, n_value, c.classMethods);
4805   if (n_value != 0) {
4806     if (info->verbose && sym_name != nullptr)
4807       outs() << sym_name;
4808     else
4809       outs() << format("0x%" PRIx64, n_value);
4810     if (c.classMethods != 0)
4811       outs() << " + " << format("0x%" PRIx64, c.classMethods);
4812   } else
4813     outs() << format("0x%" PRIx64, c.classMethods);
4814   outs() << "\n";
4815   if (c.classMethods + n_value != 0)
4816     print_method_list64_t(c.classMethods + n_value, info, "");
4817 
4818   outs() << "         protocols ";
4819   sym_name = get_symbol_64(offset + offsetof(struct category64_t, protocols), S,
4820                            info, n_value, c.protocols);
4821   if (n_value != 0) {
4822     if (info->verbose && sym_name != nullptr)
4823       outs() << sym_name;
4824     else
4825       outs() << format("0x%" PRIx64, n_value);
4826     if (c.protocols != 0)
4827       outs() << " + " << format("0x%" PRIx64, c.protocols);
4828   } else
4829     outs() << format("0x%" PRIx64, c.protocols);
4830   outs() << "\n";
4831   if (c.protocols + n_value != 0)
4832     print_protocol_list64_t(c.protocols + n_value, info);
4833 
4834   outs() << "instanceProperties ";
4835   sym_name =
4836       get_symbol_64(offset + offsetof(struct category64_t, instanceProperties),
4837                     S, info, n_value, c.instanceProperties);
4838   if (n_value != 0) {
4839     if (info->verbose && sym_name != nullptr)
4840       outs() << sym_name;
4841     else
4842       outs() << format("0x%" PRIx64, n_value);
4843     if (c.instanceProperties != 0)
4844       outs() << " + " << format("0x%" PRIx64, c.instanceProperties);
4845   } else
4846     outs() << format("0x%" PRIx64, c.instanceProperties);
4847   outs() << "\n";
4848   if (c.instanceProperties + n_value != 0)
4849     print_objc_property_list64(c.instanceProperties + n_value, info);
4850 }
4851 
print_category32_t(uint32_t p,struct DisassembleInfo * info)4852 static void print_category32_t(uint32_t p, struct DisassembleInfo *info) {
4853   struct category32_t c;
4854   const char *r;
4855   uint32_t offset, left;
4856   SectionRef S, xS;
4857   const char *name;
4858 
4859   r = get_pointer_32(p, offset, left, S, info);
4860   if (r == nullptr)
4861     return;
4862   memset(&c, '\0', sizeof(struct category32_t));
4863   if (left < sizeof(struct category32_t)) {
4864     memcpy(&c, r, left);
4865     outs() << "   (category_t entends past the end of the section)\n";
4866   } else
4867     memcpy(&c, r, sizeof(struct category32_t));
4868   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4869     swapStruct(c);
4870 
4871   outs() << "              name " << format("0x%" PRIx32, c.name);
4872   name = get_symbol_32(offset + offsetof(struct category32_t, name), S, info,
4873                        c.name);
4874   if (name)
4875     outs() << " " << name;
4876   outs() << "\n";
4877 
4878   outs() << "               cls " << format("0x%" PRIx32, c.cls) << "\n";
4879   if (c.cls != 0)
4880     print_class32_t(c.cls, info);
4881   outs() << "   instanceMethods " << format("0x%" PRIx32, c.instanceMethods)
4882          << "\n";
4883   if (c.instanceMethods != 0)
4884     print_method_list32_t(c.instanceMethods, info, "");
4885   outs() << "      classMethods " << format("0x%" PRIx32, c.classMethods)
4886          << "\n";
4887   if (c.classMethods != 0)
4888     print_method_list32_t(c.classMethods, info, "");
4889   outs() << "         protocols " << format("0x%" PRIx32, c.protocols) << "\n";
4890   if (c.protocols != 0)
4891     print_protocol_list32_t(c.protocols, info);
4892   outs() << "instanceProperties " << format("0x%" PRIx32, c.instanceProperties)
4893          << "\n";
4894   if (c.instanceProperties != 0)
4895     print_objc_property_list32(c.instanceProperties, info);
4896 }
4897 
print_message_refs64(SectionRef S,struct DisassembleInfo * info)4898 static void print_message_refs64(SectionRef S, struct DisassembleInfo *info) {
4899   uint32_t i, left, offset, xoffset;
4900   uint64_t p, n_value;
4901   struct message_ref64 mr;
4902   const char *name, *sym_name;
4903   const char *r;
4904   SectionRef xS;
4905 
4906   if (S == SectionRef())
4907     return;
4908 
4909   StringRef SectName;
4910   S.getName(SectName);
4911   DataRefImpl Ref = S.getRawDataRefImpl();
4912   StringRef SegName = info->O->getSectionFinalSegmentName(Ref);
4913   outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
4914   offset = 0;
4915   for (i = 0; i < S.getSize(); i += sizeof(struct message_ref64)) {
4916     p = S.getAddress() + i;
4917     r = get_pointer_64(p, offset, left, S, info);
4918     if (r == nullptr)
4919       return;
4920     memset(&mr, '\0', sizeof(struct message_ref64));
4921     if (left < sizeof(struct message_ref64)) {
4922       memcpy(&mr, r, left);
4923       outs() << "   (message_ref entends past the end of the section)\n";
4924     } else
4925       memcpy(&mr, r, sizeof(struct message_ref64));
4926     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4927       swapStruct(mr);
4928 
4929     outs() << "  imp ";
4930     name = get_symbol_64(offset + offsetof(struct message_ref64, imp), S, info,
4931                          n_value, mr.imp);
4932     if (n_value != 0) {
4933       outs() << format("0x%" PRIx64, n_value) << " ";
4934       if (mr.imp != 0)
4935         outs() << "+ " << format("0x%" PRIx64, mr.imp) << " ";
4936     } else
4937       outs() << format("0x%" PRIx64, mr.imp) << " ";
4938     if (name != nullptr)
4939       outs() << " " << name;
4940     outs() << "\n";
4941 
4942     outs() << "  sel ";
4943     sym_name = get_symbol_64(offset + offsetof(struct message_ref64, sel), S,
4944                              info, n_value, mr.sel);
4945     if (n_value != 0) {
4946       if (info->verbose && sym_name != nullptr)
4947         outs() << sym_name;
4948       else
4949         outs() << format("0x%" PRIx64, n_value);
4950       if (mr.sel != 0)
4951         outs() << " + " << format("0x%" PRIx64, mr.sel);
4952     } else
4953       outs() << format("0x%" PRIx64, mr.sel);
4954     name = get_pointer_64(mr.sel + n_value, xoffset, left, xS, info);
4955     if (name != nullptr)
4956       outs() << format(" %.*s", left, name);
4957     outs() << "\n";
4958 
4959     offset += sizeof(struct message_ref64);
4960   }
4961 }
4962 
print_message_refs32(SectionRef S,struct DisassembleInfo * info)4963 static void print_message_refs32(SectionRef S, struct DisassembleInfo *info) {
4964   uint32_t i, left, offset, xoffset, p;
4965   struct message_ref32 mr;
4966   const char *name, *r;
4967   SectionRef xS;
4968 
4969   if (S == SectionRef())
4970     return;
4971 
4972   StringRef SectName;
4973   S.getName(SectName);
4974   DataRefImpl Ref = S.getRawDataRefImpl();
4975   StringRef SegName = info->O->getSectionFinalSegmentName(Ref);
4976   outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
4977   offset = 0;
4978   for (i = 0; i < S.getSize(); i += sizeof(struct message_ref64)) {
4979     p = S.getAddress() + i;
4980     r = get_pointer_32(p, offset, left, S, info);
4981     if (r == nullptr)
4982       return;
4983     memset(&mr, '\0', sizeof(struct message_ref32));
4984     if (left < sizeof(struct message_ref32)) {
4985       memcpy(&mr, r, left);
4986       outs() << "   (message_ref entends past the end of the section)\n";
4987     } else
4988       memcpy(&mr, r, sizeof(struct message_ref32));
4989     if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
4990       swapStruct(mr);
4991 
4992     outs() << "  imp " << format("0x%" PRIx32, mr.imp);
4993     name = get_symbol_32(offset + offsetof(struct message_ref32, imp), S, info,
4994                          mr.imp);
4995     if (name != nullptr)
4996       outs() << " " << name;
4997     outs() << "\n";
4998 
4999     outs() << "  sel " << format("0x%" PRIx32, mr.sel);
5000     name = get_pointer_32(mr.sel, xoffset, left, xS, info);
5001     if (name != nullptr)
5002       outs() << " " << name;
5003     outs() << "\n";
5004 
5005     offset += sizeof(struct message_ref32);
5006   }
5007 }
5008 
print_image_info64(SectionRef S,struct DisassembleInfo * info)5009 static void print_image_info64(SectionRef S, struct DisassembleInfo *info) {
5010   uint32_t left, offset, swift_version;
5011   uint64_t p;
5012   struct objc_image_info64 o;
5013   const char *r;
5014 
5015   if (S == SectionRef())
5016     return;
5017 
5018   StringRef SectName;
5019   S.getName(SectName);
5020   DataRefImpl Ref = S.getRawDataRefImpl();
5021   StringRef SegName = info->O->getSectionFinalSegmentName(Ref);
5022   outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
5023   p = S.getAddress();
5024   r = get_pointer_64(p, offset, left, S, info);
5025   if (r == nullptr)
5026     return;
5027   memset(&o, '\0', sizeof(struct objc_image_info64));
5028   if (left < sizeof(struct objc_image_info64)) {
5029     memcpy(&o, r, left);
5030     outs() << "   (objc_image_info entends past the end of the section)\n";
5031   } else
5032     memcpy(&o, r, sizeof(struct objc_image_info64));
5033   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
5034     swapStruct(o);
5035   outs() << "  version " << o.version << "\n";
5036   outs() << "    flags " << format("0x%" PRIx32, o.flags);
5037   if (o.flags & OBJC_IMAGE_IS_REPLACEMENT)
5038     outs() << " OBJC_IMAGE_IS_REPLACEMENT";
5039   if (o.flags & OBJC_IMAGE_SUPPORTS_GC)
5040     outs() << " OBJC_IMAGE_SUPPORTS_GC";
5041   swift_version = (o.flags >> 8) & 0xff;
5042   if (swift_version != 0) {
5043     if (swift_version == 1)
5044       outs() << " Swift 1.0";
5045     else if (swift_version == 2)
5046       outs() << " Swift 1.1";
5047     else
5048       outs() << " unknown future Swift version (" << swift_version << ")";
5049   }
5050   outs() << "\n";
5051 }
5052 
print_image_info32(SectionRef S,struct DisassembleInfo * info)5053 static void print_image_info32(SectionRef S, struct DisassembleInfo *info) {
5054   uint32_t left, offset, swift_version, p;
5055   struct objc_image_info32 o;
5056   const char *r;
5057 
5058   StringRef SectName;
5059   S.getName(SectName);
5060   DataRefImpl Ref = S.getRawDataRefImpl();
5061   StringRef SegName = info->O->getSectionFinalSegmentName(Ref);
5062   outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
5063   p = S.getAddress();
5064   r = get_pointer_32(p, offset, left, S, info);
5065   if (r == nullptr)
5066     return;
5067   memset(&o, '\0', sizeof(struct objc_image_info32));
5068   if (left < sizeof(struct objc_image_info32)) {
5069     memcpy(&o, r, left);
5070     outs() << "   (objc_image_info entends past the end of the section)\n";
5071   } else
5072     memcpy(&o, r, sizeof(struct objc_image_info32));
5073   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
5074     swapStruct(o);
5075   outs() << "  version " << o.version << "\n";
5076   outs() << "    flags " << format("0x%" PRIx32, o.flags);
5077   if (o.flags & OBJC_IMAGE_IS_REPLACEMENT)
5078     outs() << " OBJC_IMAGE_IS_REPLACEMENT";
5079   if (o.flags & OBJC_IMAGE_SUPPORTS_GC)
5080     outs() << " OBJC_IMAGE_SUPPORTS_GC";
5081   swift_version = (o.flags >> 8) & 0xff;
5082   if (swift_version != 0) {
5083     if (swift_version == 1)
5084       outs() << " Swift 1.0";
5085     else if (swift_version == 2)
5086       outs() << " Swift 1.1";
5087     else
5088       outs() << " unknown future Swift version (" << swift_version << ")";
5089   }
5090   outs() << "\n";
5091 }
5092 
print_image_info(SectionRef S,struct DisassembleInfo * info)5093 static void print_image_info(SectionRef S, struct DisassembleInfo *info) {
5094   uint32_t left, offset, p;
5095   struct imageInfo_t o;
5096   const char *r;
5097 
5098   StringRef SectName;
5099   S.getName(SectName);
5100   DataRefImpl Ref = S.getRawDataRefImpl();
5101   StringRef SegName = info->O->getSectionFinalSegmentName(Ref);
5102   outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
5103   p = S.getAddress();
5104   r = get_pointer_32(p, offset, left, S, info);
5105   if (r == nullptr)
5106     return;
5107   memset(&o, '\0', sizeof(struct imageInfo_t));
5108   if (left < sizeof(struct imageInfo_t)) {
5109     memcpy(&o, r, left);
5110     outs() << " (imageInfo entends past the end of the section)\n";
5111   } else
5112     memcpy(&o, r, sizeof(struct imageInfo_t));
5113   if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
5114     swapStruct(o);
5115   outs() << "  version " << o.version << "\n";
5116   outs() << "    flags " << format("0x%" PRIx32, o.flags);
5117   if (o.flags & 0x1)
5118     outs() << "  F&C";
5119   if (o.flags & 0x2)
5120     outs() << " GC";
5121   if (o.flags & 0x4)
5122     outs() << " GC-only";
5123   else
5124     outs() << " RR";
5125   outs() << "\n";
5126 }
5127 
printObjc2_64bit_MetaData(MachOObjectFile * O,bool verbose)5128 static void printObjc2_64bit_MetaData(MachOObjectFile *O, bool verbose) {
5129   SymbolAddressMap AddrMap;
5130   if (verbose)
5131     CreateSymbolAddressMap(O, &AddrMap);
5132 
5133   std::vector<SectionRef> Sections;
5134   for (const SectionRef &Section : O->sections()) {
5135     StringRef SectName;
5136     Section.getName(SectName);
5137     Sections.push_back(Section);
5138   }
5139 
5140   struct DisassembleInfo info;
5141   // Set up the block of info used by the Symbolizer call backs.
5142   info.verbose = verbose;
5143   info.O = O;
5144   info.AddrMap = &AddrMap;
5145   info.Sections = &Sections;
5146   info.class_name = nullptr;
5147   info.selector_name = nullptr;
5148   info.method = nullptr;
5149   info.demangled_name = nullptr;
5150   info.bindtable = nullptr;
5151   info.adrp_addr = 0;
5152   info.adrp_inst = 0;
5153 
5154   info.depth = 0;
5155   SectionRef CL = get_section(O, "__OBJC2", "__class_list");
5156   if (CL == SectionRef())
5157     CL = get_section(O, "__DATA", "__objc_classlist");
5158   info.S = CL;
5159   walk_pointer_list_64("class", CL, O, &info, print_class64_t);
5160 
5161   SectionRef CR = get_section(O, "__OBJC2", "__class_refs");
5162   if (CR == SectionRef())
5163     CR = get_section(O, "__DATA", "__objc_classrefs");
5164   info.S = CR;
5165   walk_pointer_list_64("class refs", CR, O, &info, nullptr);
5166 
5167   SectionRef SR = get_section(O, "__OBJC2", "__super_refs");
5168   if (SR == SectionRef())
5169     SR = get_section(O, "__DATA", "__objc_superrefs");
5170   info.S = SR;
5171   walk_pointer_list_64("super refs", SR, O, &info, nullptr);
5172 
5173   SectionRef CA = get_section(O, "__OBJC2", "__category_list");
5174   if (CA == SectionRef())
5175     CA = get_section(O, "__DATA", "__objc_catlist");
5176   info.S = CA;
5177   walk_pointer_list_64("category", CA, O, &info, print_category64_t);
5178 
5179   SectionRef PL = get_section(O, "__OBJC2", "__protocol_list");
5180   if (PL == SectionRef())
5181     PL = get_section(O, "__DATA", "__objc_protolist");
5182   info.S = PL;
5183   walk_pointer_list_64("protocol", PL, O, &info, nullptr);
5184 
5185   SectionRef MR = get_section(O, "__OBJC2", "__message_refs");
5186   if (MR == SectionRef())
5187     MR = get_section(O, "__DATA", "__objc_msgrefs");
5188   info.S = MR;
5189   print_message_refs64(MR, &info);
5190 
5191   SectionRef II = get_section(O, "__OBJC2", "__image_info");
5192   if (II == SectionRef())
5193     II = get_section(O, "__DATA", "__objc_imageinfo");
5194   info.S = II;
5195   print_image_info64(II, &info);
5196 
5197   if (info.bindtable != nullptr)
5198     delete info.bindtable;
5199 }
5200 
printObjc2_32bit_MetaData(MachOObjectFile * O,bool verbose)5201 static void printObjc2_32bit_MetaData(MachOObjectFile *O, bool verbose) {
5202   SymbolAddressMap AddrMap;
5203   if (verbose)
5204     CreateSymbolAddressMap(O, &AddrMap);
5205 
5206   std::vector<SectionRef> Sections;
5207   for (const SectionRef &Section : O->sections()) {
5208     StringRef SectName;
5209     Section.getName(SectName);
5210     Sections.push_back(Section);
5211   }
5212 
5213   struct DisassembleInfo info;
5214   // Set up the block of info used by the Symbolizer call backs.
5215   info.verbose = verbose;
5216   info.O = O;
5217   info.AddrMap = &AddrMap;
5218   info.Sections = &Sections;
5219   info.class_name = nullptr;
5220   info.selector_name = nullptr;
5221   info.method = nullptr;
5222   info.demangled_name = nullptr;
5223   info.bindtable = nullptr;
5224   info.adrp_addr = 0;
5225   info.adrp_inst = 0;
5226 
5227   const SectionRef CL = get_section(O, "__OBJC2", "__class_list");
5228   if (CL != SectionRef()) {
5229     info.S = CL;
5230     walk_pointer_list_32("class", CL, O, &info, print_class32_t);
5231   } else {
5232     const SectionRef CL = get_section(O, "__DATA", "__objc_classlist");
5233     info.S = CL;
5234     walk_pointer_list_32("class", CL, O, &info, print_class32_t);
5235   }
5236 
5237   const SectionRef CR = get_section(O, "__OBJC2", "__class_refs");
5238   if (CR != SectionRef()) {
5239     info.S = CR;
5240     walk_pointer_list_32("class refs", CR, O, &info, nullptr);
5241   } else {
5242     const SectionRef CR = get_section(O, "__DATA", "__objc_classrefs");
5243     info.S = CR;
5244     walk_pointer_list_32("class refs", CR, O, &info, nullptr);
5245   }
5246 
5247   const SectionRef SR = get_section(O, "__OBJC2", "__super_refs");
5248   if (SR != SectionRef()) {
5249     info.S = SR;
5250     walk_pointer_list_32("super refs", SR, O, &info, nullptr);
5251   } else {
5252     const SectionRef SR = get_section(O, "__DATA", "__objc_superrefs");
5253     info.S = SR;
5254     walk_pointer_list_32("super refs", SR, O, &info, nullptr);
5255   }
5256 
5257   const SectionRef CA = get_section(O, "__OBJC2", "__category_list");
5258   if (CA != SectionRef()) {
5259     info.S = CA;
5260     walk_pointer_list_32("category", CA, O, &info, print_category32_t);
5261   } else {
5262     const SectionRef CA = get_section(O, "__DATA", "__objc_catlist");
5263     info.S = CA;
5264     walk_pointer_list_32("category", CA, O, &info, print_category32_t);
5265   }
5266 
5267   const SectionRef PL = get_section(O, "__OBJC2", "__protocol_list");
5268   if (PL != SectionRef()) {
5269     info.S = PL;
5270     walk_pointer_list_32("protocol", PL, O, &info, nullptr);
5271   } else {
5272     const SectionRef PL = get_section(O, "__DATA", "__objc_protolist");
5273     info.S = PL;
5274     walk_pointer_list_32("protocol", PL, O, &info, nullptr);
5275   }
5276 
5277   const SectionRef MR = get_section(O, "__OBJC2", "__message_refs");
5278   if (MR != SectionRef()) {
5279     info.S = MR;
5280     print_message_refs32(MR, &info);
5281   } else {
5282     const SectionRef MR = get_section(O, "__DATA", "__objc_msgrefs");
5283     info.S = MR;
5284     print_message_refs32(MR, &info);
5285   }
5286 
5287   const SectionRef II = get_section(O, "__OBJC2", "__image_info");
5288   if (II != SectionRef()) {
5289     info.S = II;
5290     print_image_info32(II, &info);
5291   } else {
5292     const SectionRef II = get_section(O, "__DATA", "__objc_imageinfo");
5293     info.S = II;
5294     print_image_info32(II, &info);
5295   }
5296 }
5297 
printObjc1_32bit_MetaData(MachOObjectFile * O,bool verbose)5298 static bool printObjc1_32bit_MetaData(MachOObjectFile *O, bool verbose) {
5299   uint32_t i, j, p, offset, xoffset, left, defs_left, def;
5300   const char *r, *name, *defs;
5301   struct objc_module_t module;
5302   SectionRef S, xS;
5303   struct objc_symtab_t symtab;
5304   struct objc_class_t objc_class;
5305   struct objc_category_t objc_category;
5306 
5307   outs() << "Objective-C segment\n";
5308   S = get_section(O, "__OBJC", "__module_info");
5309   if (S == SectionRef())
5310     return false;
5311 
5312   SymbolAddressMap AddrMap;
5313   if (verbose)
5314     CreateSymbolAddressMap(O, &AddrMap);
5315 
5316   std::vector<SectionRef> Sections;
5317   for (const SectionRef &Section : O->sections()) {
5318     StringRef SectName;
5319     Section.getName(SectName);
5320     Sections.push_back(Section);
5321   }
5322 
5323   struct DisassembleInfo info;
5324   // Set up the block of info used by the Symbolizer call backs.
5325   info.verbose = verbose;
5326   info.O = O;
5327   info.AddrMap = &AddrMap;
5328   info.Sections = &Sections;
5329   info.class_name = nullptr;
5330   info.selector_name = nullptr;
5331   info.method = nullptr;
5332   info.demangled_name = nullptr;
5333   info.bindtable = nullptr;
5334   info.adrp_addr = 0;
5335   info.adrp_inst = 0;
5336 
5337   for (i = 0; i < S.getSize(); i += sizeof(struct objc_module_t)) {
5338     p = S.getAddress() + i;
5339     r = get_pointer_32(p, offset, left, S, &info, true);
5340     if (r == nullptr)
5341       return true;
5342     memset(&module, '\0', sizeof(struct objc_module_t));
5343     if (left < sizeof(struct objc_module_t)) {
5344       memcpy(&module, r, left);
5345       outs() << "   (module extends past end of __module_info section)\n";
5346     } else
5347       memcpy(&module, r, sizeof(struct objc_module_t));
5348     if (O->isLittleEndian() != sys::IsLittleEndianHost)
5349       swapStruct(module);
5350 
5351     outs() << "Module " << format("0x%" PRIx32, p) << "\n";
5352     outs() << "    version " << module.version << "\n";
5353     outs() << "       size " << module.size << "\n";
5354     outs() << "       name ";
5355     name = get_pointer_32(module.name, xoffset, left, xS, &info, true);
5356     if (name != nullptr)
5357       outs() << format("%.*s", left, name);
5358     else
5359       outs() << format("0x%08" PRIx32, module.name)
5360              << "(not in an __OBJC section)";
5361     outs() << "\n";
5362 
5363     r = get_pointer_32(module.symtab, xoffset, left, xS, &info, true);
5364     if (module.symtab == 0 || r == nullptr) {
5365       outs() << "     symtab " << format("0x%08" PRIx32, module.symtab)
5366              << " (not in an __OBJC section)\n";
5367       continue;
5368     }
5369     outs() << "     symtab " << format("0x%08" PRIx32, module.symtab) << "\n";
5370     memset(&symtab, '\0', sizeof(struct objc_symtab_t));
5371     defs_left = 0;
5372     defs = nullptr;
5373     if (left < sizeof(struct objc_symtab_t)) {
5374       memcpy(&symtab, r, left);
5375       outs() << "\tsymtab extends past end of an __OBJC section)\n";
5376     } else {
5377       memcpy(&symtab, r, sizeof(struct objc_symtab_t));
5378       if (left > sizeof(struct objc_symtab_t)) {
5379         defs_left = left - sizeof(struct objc_symtab_t);
5380         defs = r + sizeof(struct objc_symtab_t);
5381       }
5382     }
5383     if (O->isLittleEndian() != sys::IsLittleEndianHost)
5384       swapStruct(symtab);
5385 
5386     outs() << "\tsel_ref_cnt " << symtab.sel_ref_cnt << "\n";
5387     r = get_pointer_32(symtab.refs, xoffset, left, xS, &info, true);
5388     outs() << "\trefs " << format("0x%08" PRIx32, symtab.refs);
5389     if (r == nullptr)
5390       outs() << " (not in an __OBJC section)";
5391     outs() << "\n";
5392     outs() << "\tcls_def_cnt " << symtab.cls_def_cnt << "\n";
5393     outs() << "\tcat_def_cnt " << symtab.cat_def_cnt << "\n";
5394     if (symtab.cls_def_cnt > 0)
5395       outs() << "\tClass Definitions\n";
5396     for (j = 0; j < symtab.cls_def_cnt; j++) {
5397       if ((j + 1) * sizeof(uint32_t) > defs_left) {
5398         outs() << "\t(remaining class defs entries entends past the end of the "
5399                << "section)\n";
5400         break;
5401       }
5402       memcpy(&def, defs + j * sizeof(uint32_t), sizeof(uint32_t));
5403       if (O->isLittleEndian() != sys::IsLittleEndianHost)
5404         sys::swapByteOrder(def);
5405 
5406       r = get_pointer_32(def, xoffset, left, xS, &info, true);
5407       outs() << "\tdefs[" << j << "] " << format("0x%08" PRIx32, def);
5408       if (r != nullptr) {
5409         if (left > sizeof(struct objc_class_t)) {
5410           outs() << "\n";
5411           memcpy(&objc_class, r, sizeof(struct objc_class_t));
5412         } else {
5413           outs() << " (entends past the end of the section)\n";
5414           memset(&objc_class, '\0', sizeof(struct objc_class_t));
5415           memcpy(&objc_class, r, left);
5416         }
5417         if (O->isLittleEndian() != sys::IsLittleEndianHost)
5418           swapStruct(objc_class);
5419         print_objc_class_t(&objc_class, &info);
5420       } else {
5421         outs() << "(not in an __OBJC section)\n";
5422       }
5423 
5424       if (CLS_GETINFO(&objc_class, CLS_CLASS)) {
5425         outs() << "\tMeta Class";
5426         r = get_pointer_32(objc_class.isa, xoffset, left, xS, &info, true);
5427         if (r != nullptr) {
5428           if (left > sizeof(struct objc_class_t)) {
5429             outs() << "\n";
5430             memcpy(&objc_class, r, sizeof(struct objc_class_t));
5431           } else {
5432             outs() << " (entends past the end of the section)\n";
5433             memset(&objc_class, '\0', sizeof(struct objc_class_t));
5434             memcpy(&objc_class, r, left);
5435           }
5436           if (O->isLittleEndian() != sys::IsLittleEndianHost)
5437             swapStruct(objc_class);
5438           print_objc_class_t(&objc_class, &info);
5439         } else {
5440           outs() << "(not in an __OBJC section)\n";
5441         }
5442       }
5443     }
5444     if (symtab.cat_def_cnt > 0)
5445       outs() << "\tCategory Definitions\n";
5446     for (j = 0; j < symtab.cat_def_cnt; j++) {
5447       if ((j + symtab.cls_def_cnt + 1) * sizeof(uint32_t) > defs_left) {
5448         outs() << "\t(remaining category defs entries entends past the end of "
5449                << "the section)\n";
5450         break;
5451       }
5452       memcpy(&def, defs + (j + symtab.cls_def_cnt) * sizeof(uint32_t),
5453              sizeof(uint32_t));
5454       if (O->isLittleEndian() != sys::IsLittleEndianHost)
5455         sys::swapByteOrder(def);
5456 
5457       r = get_pointer_32(def, xoffset, left, xS, &info, true);
5458       outs() << "\tdefs[" << j + symtab.cls_def_cnt << "] "
5459              << format("0x%08" PRIx32, def);
5460       if (r != nullptr) {
5461         if (left > sizeof(struct objc_category_t)) {
5462           outs() << "\n";
5463           memcpy(&objc_category, r, sizeof(struct objc_category_t));
5464         } else {
5465           outs() << " (entends past the end of the section)\n";
5466           memset(&objc_category, '\0', sizeof(struct objc_category_t));
5467           memcpy(&objc_category, r, left);
5468         }
5469         if (O->isLittleEndian() != sys::IsLittleEndianHost)
5470           swapStruct(objc_category);
5471         print_objc_objc_category_t(&objc_category, &info);
5472       } else {
5473         outs() << "(not in an __OBJC section)\n";
5474       }
5475     }
5476   }
5477   const SectionRef II = get_section(O, "__OBJC", "__image_info");
5478   if (II != SectionRef())
5479     print_image_info(II, &info);
5480 
5481   return true;
5482 }
5483 
DumpProtocolSection(MachOObjectFile * O,const char * sect,uint32_t size,uint32_t addr)5484 static void DumpProtocolSection(MachOObjectFile *O, const char *sect,
5485                                 uint32_t size, uint32_t addr) {
5486   SymbolAddressMap AddrMap;
5487   CreateSymbolAddressMap(O, &AddrMap);
5488 
5489   std::vector<SectionRef> Sections;
5490   for (const SectionRef &Section : O->sections()) {
5491     StringRef SectName;
5492     Section.getName(SectName);
5493     Sections.push_back(Section);
5494   }
5495 
5496   struct DisassembleInfo info;
5497   // Set up the block of info used by the Symbolizer call backs.
5498   info.verbose = true;
5499   info.O = O;
5500   info.AddrMap = &AddrMap;
5501   info.Sections = &Sections;
5502   info.class_name = nullptr;
5503   info.selector_name = nullptr;
5504   info.method = nullptr;
5505   info.demangled_name = nullptr;
5506   info.bindtable = nullptr;
5507   info.adrp_addr = 0;
5508   info.adrp_inst = 0;
5509 
5510   const char *p;
5511   struct objc_protocol_t protocol;
5512   uint32_t left, paddr;
5513   for (p = sect; p < sect + size; p += sizeof(struct objc_protocol_t)) {
5514     memset(&protocol, '\0', sizeof(struct objc_protocol_t));
5515     left = size - (p - sect);
5516     if (left < sizeof(struct objc_protocol_t)) {
5517       outs() << "Protocol extends past end of __protocol section\n";
5518       memcpy(&protocol, p, left);
5519     } else
5520       memcpy(&protocol, p, sizeof(struct objc_protocol_t));
5521     if (O->isLittleEndian() != sys::IsLittleEndianHost)
5522       swapStruct(protocol);
5523     paddr = addr + (p - sect);
5524     outs() << "Protocol " << format("0x%" PRIx32, paddr);
5525     if (print_protocol(paddr, 0, &info))
5526       outs() << "(not in an __OBJC section)\n";
5527   }
5528 }
5529 
printObjcMetaData(MachOObjectFile * O,bool verbose)5530 static void printObjcMetaData(MachOObjectFile *O, bool verbose) {
5531   if (O->is64Bit())
5532     printObjc2_64bit_MetaData(O, verbose);
5533   else {
5534     MachO::mach_header H;
5535     H = O->getHeader();
5536     if (H.cputype == MachO::CPU_TYPE_ARM)
5537       printObjc2_32bit_MetaData(O, verbose);
5538     else {
5539       // This is the 32-bit non-arm cputype case.  Which is normally
5540       // the first Objective-C ABI.  But it may be the case of a
5541       // binary for the iOS simulator which is the second Objective-C
5542       // ABI.  In that case printObjc1_32bit_MetaData() will determine that
5543       // and return false.
5544       if (!printObjc1_32bit_MetaData(O, verbose))
5545         printObjc2_32bit_MetaData(O, verbose);
5546     }
5547   }
5548 }
5549 
5550 // GuessLiteralPointer returns a string which for the item in the Mach-O file
5551 // for the address passed in as ReferenceValue for printing as a comment with
5552 // the instruction and also returns the corresponding type of that item
5553 // indirectly through ReferenceType.
5554 //
5555 // If ReferenceValue is an address of literal cstring then a pointer to the
5556 // cstring is returned and ReferenceType is set to
5557 // LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
5558 //
5559 // If ReferenceValue is an address of an Objective-C CFString, Selector ref or
5560 // Class ref that name is returned and the ReferenceType is set accordingly.
5561 //
5562 // Lastly, literals which are Symbol address in a literal pool are looked for
5563 // and if found the symbol name is returned and ReferenceType is set to
5564 // LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
5565 //
5566 // If there is no item in the Mach-O file for the address passed in as
5567 // ReferenceValue nullptr is returned and ReferenceType is unchanged.
GuessLiteralPointer(uint64_t ReferenceValue,uint64_t ReferencePC,uint64_t * ReferenceType,struct DisassembleInfo * info)5568 static const char *GuessLiteralPointer(uint64_t ReferenceValue,
5569                                        uint64_t ReferencePC,
5570                                        uint64_t *ReferenceType,
5571                                        struct DisassembleInfo *info) {
5572   // First see if there is an external relocation entry at the ReferencePC.
5573   if (info->O->getHeader().filetype == MachO::MH_OBJECT) {
5574     uint64_t sect_addr = info->S.getAddress();
5575     uint64_t sect_offset = ReferencePC - sect_addr;
5576     bool reloc_found = false;
5577     DataRefImpl Rel;
5578     MachO::any_relocation_info RE;
5579     bool isExtern = false;
5580     SymbolRef Symbol;
5581     for (const RelocationRef &Reloc : info->S.relocations()) {
5582       uint64_t RelocOffset = Reloc.getOffset();
5583       if (RelocOffset == sect_offset) {
5584         Rel = Reloc.getRawDataRefImpl();
5585         RE = info->O->getRelocation(Rel);
5586         if (info->O->isRelocationScattered(RE))
5587           continue;
5588         isExtern = info->O->getPlainRelocationExternal(RE);
5589         if (isExtern) {
5590           symbol_iterator RelocSym = Reloc.getSymbol();
5591           Symbol = *RelocSym;
5592         }
5593         reloc_found = true;
5594         break;
5595       }
5596     }
5597     // If there is an external relocation entry for a symbol in a section
5598     // then used that symbol's value for the value of the reference.
5599     if (reloc_found && isExtern) {
5600       if (info->O->getAnyRelocationPCRel(RE)) {
5601         unsigned Type = info->O->getAnyRelocationType(RE);
5602         if (Type == MachO::X86_64_RELOC_SIGNED) {
5603           ReferenceValue = Symbol.getValue();
5604         }
5605       }
5606     }
5607   }
5608 
5609   // Look for literals such as Objective-C CFStrings refs, Selector refs,
5610   // Message refs and Class refs.
5611   bool classref, selref, msgref, cfstring;
5612   uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
5613                                                selref, msgref, cfstring);
5614   if (classref && pointer_value == 0) {
5615     // Note the ReferenceValue is a pointer into the __objc_classrefs section.
5616     // And the pointer_value in that section is typically zero as it will be
5617     // set by dyld as part of the "bind information".
5618     const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
5619     if (name != nullptr) {
5620       *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
5621       const char *class_name = strrchr(name, '$');
5622       if (class_name != nullptr && class_name[1] == '_' &&
5623           class_name[2] != '\0') {
5624         info->class_name = class_name + 2;
5625         return name;
5626       }
5627     }
5628   }
5629 
5630   if (classref) {
5631     *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
5632     const char *name =
5633         get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
5634     if (name != nullptr)
5635       info->class_name = name;
5636     else
5637       name = "bad class ref";
5638     return name;
5639   }
5640 
5641   if (cfstring) {
5642     *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref;
5643     const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
5644     return name;
5645   }
5646 
5647   if (selref && pointer_value == 0)
5648     pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
5649 
5650   if (pointer_value != 0)
5651     ReferenceValue = pointer_value;
5652 
5653   const char *name = GuessCstringPointer(ReferenceValue, info);
5654   if (name) {
5655     if (pointer_value != 0 && selref) {
5656       *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref;
5657       info->selector_name = name;
5658     } else if (pointer_value != 0 && msgref) {
5659       info->class_name = nullptr;
5660       *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref;
5661       info->selector_name = name;
5662     } else
5663       *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr;
5664     return name;
5665   }
5666 
5667   // Lastly look for an indirect symbol with this ReferenceValue which is in
5668   // a literal pool.  If found return that symbol name.
5669   name = GuessIndirectSymbol(ReferenceValue, info);
5670   if (name) {
5671     *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr;
5672     return name;
5673   }
5674 
5675   return nullptr;
5676 }
5677 
5678 // SymbolizerSymbolLookUp is the symbol lookup function passed when creating
5679 // the Symbolizer.  It looks up the ReferenceValue using the info passed via the
5680 // pointer to the struct DisassembleInfo that was passed when MCSymbolizer
5681 // is created and returns the symbol name that matches the ReferenceValue or
5682 // nullptr if none.  The ReferenceType is passed in for the IN type of
5683 // reference the instruction is making from the values in defined in the header
5684 // "llvm-c/Disassembler.h".  On return the ReferenceType can set to a specific
5685 // Out type and the ReferenceName will also be set which is added as a comment
5686 // to the disassembled instruction.
5687 //
5688 #if HAVE_CXXABI_H
5689 // If the symbol name is a C++ mangled name then the demangled name is
5690 // returned through ReferenceName and ReferenceType is set to
5691 // LLVMDisassembler_ReferenceType_DeMangled_Name .
5692 #endif
5693 //
5694 // When this is called to get a symbol name for a branch target then the
5695 // ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
5696 // SymbolValue will be looked for in the indirect symbol table to determine if
5697 // it is an address for a symbol stub.  If so then the symbol name for that
5698 // stub is returned indirectly through ReferenceName and then ReferenceType is
5699 // set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
5700 //
5701 // When this is called with an value loaded via a PC relative load then
5702 // ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
5703 // SymbolValue is checked to be an address of literal pointer, symbol pointer,
5704 // or an Objective-C meta data reference.  If so the output ReferenceType is
5705 // set to correspond to that as well as setting the ReferenceName.
SymbolizerSymbolLookUp(void * DisInfo,uint64_t ReferenceValue,uint64_t * ReferenceType,uint64_t ReferencePC,const char ** ReferenceName)5706 static const char *SymbolizerSymbolLookUp(void *DisInfo,
5707                                           uint64_t ReferenceValue,
5708                                           uint64_t *ReferenceType,
5709                                           uint64_t ReferencePC,
5710                                           const char **ReferenceName) {
5711   struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
5712   // If no verbose symbolic information is wanted then just return nullptr.
5713   if (!info->verbose) {
5714     *ReferenceName = nullptr;
5715     *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5716     return nullptr;
5717   }
5718 
5719   const char *SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
5720 
5721   if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) {
5722     *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
5723     if (*ReferenceName != nullptr) {
5724       method_reference(info, ReferenceType, ReferenceName);
5725       if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message)
5726         *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub;
5727     } else
5728 #if HAVE_CXXABI_H
5729         if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
5730       if (info->demangled_name != nullptr)
5731         free(info->demangled_name);
5732       int status;
5733       info->demangled_name =
5734           abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
5735       if (info->demangled_name != nullptr) {
5736         *ReferenceName = info->demangled_name;
5737         *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
5738       } else
5739         *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5740     } else
5741 #endif
5742       *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5743   } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) {
5744     *ReferenceName =
5745         GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
5746     if (*ReferenceName)
5747       method_reference(info, ReferenceType, ReferenceName);
5748     else
5749       *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5750     // If this is arm64 and the reference is an adrp instruction save the
5751     // instruction, passed in ReferenceValue and the address of the instruction
5752     // for use later if we see and add immediate instruction.
5753   } else if (info->O->getArch() == Triple::aarch64 &&
5754              *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) {
5755     info->adrp_inst = ReferenceValue;
5756     info->adrp_addr = ReferencePC;
5757     SymbolName = nullptr;
5758     *ReferenceName = nullptr;
5759     *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5760     // If this is arm64 and reference is an add immediate instruction and we
5761     // have
5762     // seen an adrp instruction just before it and the adrp's Xd register
5763     // matches
5764     // this add's Xn register reconstruct the value being referenced and look to
5765     // see if it is a literal pointer.  Note the add immediate instruction is
5766     // passed in ReferenceValue.
5767   } else if (info->O->getArch() == Triple::aarch64 &&
5768              *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri &&
5769              ReferencePC - 4 == info->adrp_addr &&
5770              (info->adrp_inst & 0x9f000000) == 0x90000000 &&
5771              (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
5772     uint32_t addxri_inst;
5773     uint64_t adrp_imm, addxri_imm;
5774 
5775     adrp_imm =
5776         ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
5777     if (info->adrp_inst & 0x0200000)
5778       adrp_imm |= 0xfffffffffc000000LL;
5779 
5780     addxri_inst = ReferenceValue;
5781     addxri_imm = (addxri_inst >> 10) & 0xfff;
5782     if (((addxri_inst >> 22) & 0x3) == 1)
5783       addxri_imm <<= 12;
5784 
5785     ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
5786                      (adrp_imm << 12) + addxri_imm;
5787 
5788     *ReferenceName =
5789         GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
5790     if (*ReferenceName == nullptr)
5791       *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5792     // If this is arm64 and the reference is a load register instruction and we
5793     // have seen an adrp instruction just before it and the adrp's Xd register
5794     // matches this add's Xn register reconstruct the value being referenced and
5795     // look to see if it is a literal pointer.  Note the load register
5796     // instruction is passed in ReferenceValue.
5797   } else if (info->O->getArch() == Triple::aarch64 &&
5798              *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui &&
5799              ReferencePC - 4 == info->adrp_addr &&
5800              (info->adrp_inst & 0x9f000000) == 0x90000000 &&
5801              (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
5802     uint32_t ldrxui_inst;
5803     uint64_t adrp_imm, ldrxui_imm;
5804 
5805     adrp_imm =
5806         ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
5807     if (info->adrp_inst & 0x0200000)
5808       adrp_imm |= 0xfffffffffc000000LL;
5809 
5810     ldrxui_inst = ReferenceValue;
5811     ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
5812 
5813     ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
5814                      (adrp_imm << 12) + (ldrxui_imm << 3);
5815 
5816     *ReferenceName =
5817         GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
5818     if (*ReferenceName == nullptr)
5819       *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5820   }
5821   // If this arm64 and is an load register (PC-relative) instruction the
5822   // ReferenceValue is the PC plus the immediate value.
5823   else if (info->O->getArch() == Triple::aarch64 &&
5824            (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl ||
5825             *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) {
5826     *ReferenceName =
5827         GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
5828     if (*ReferenceName == nullptr)
5829       *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5830   }
5831 #if HAVE_CXXABI_H
5832   else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
5833     if (info->demangled_name != nullptr)
5834       free(info->demangled_name);
5835     int status;
5836     info->demangled_name =
5837         abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
5838     if (info->demangled_name != nullptr) {
5839       *ReferenceName = info->demangled_name;
5840       *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
5841     }
5842   }
5843 #endif
5844   else {
5845     *ReferenceName = nullptr;
5846     *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
5847   }
5848 
5849   return SymbolName;
5850 }
5851 
5852 /// \brief Emits the comments that are stored in the CommentStream.
5853 /// Each comment in the CommentStream must end with a newline.
emitComments(raw_svector_ostream & CommentStream,SmallString<128> & CommentsToEmit,formatted_raw_ostream & FormattedOS,const MCAsmInfo & MAI)5854 static void emitComments(raw_svector_ostream &CommentStream,
5855                          SmallString<128> &CommentsToEmit,
5856                          formatted_raw_ostream &FormattedOS,
5857                          const MCAsmInfo &MAI) {
5858   // Flush the stream before taking its content.
5859   StringRef Comments = CommentsToEmit.str();
5860   // Get the default information for printing a comment.
5861   const char *CommentBegin = MAI.getCommentString();
5862   unsigned CommentColumn = MAI.getCommentColumn();
5863   bool IsFirst = true;
5864   while (!Comments.empty()) {
5865     if (!IsFirst)
5866       FormattedOS << '\n';
5867     // Emit a line of comments.
5868     FormattedOS.PadToColumn(CommentColumn);
5869     size_t Position = Comments.find('\n');
5870     FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
5871     // Move after the newline character.
5872     Comments = Comments.substr(Position + 1);
5873     IsFirst = false;
5874   }
5875   FormattedOS.flush();
5876 
5877   // Tell the comment stream that the vector changed underneath it.
5878   CommentsToEmit.clear();
5879 }
5880 
DisassembleMachO(StringRef Filename,MachOObjectFile * MachOOF,StringRef DisSegName,StringRef DisSectName)5881 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
5882                              StringRef DisSegName, StringRef DisSectName) {
5883   const char *McpuDefault = nullptr;
5884   const Target *ThumbTarget = nullptr;
5885   const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
5886   if (!TheTarget) {
5887     // GetTarget prints out stuff.
5888     return;
5889   }
5890   if (MCPU.empty() && McpuDefault)
5891     MCPU = McpuDefault;
5892 
5893   std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
5894   std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
5895   if (ThumbTarget)
5896     ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
5897 
5898   // Package up features to be passed to target/subtarget
5899   std::string FeaturesStr;
5900   if (MAttrs.size()) {
5901     SubtargetFeatures Features;
5902     for (unsigned i = 0; i != MAttrs.size(); ++i)
5903       Features.AddFeature(MAttrs[i]);
5904     FeaturesStr = Features.getString();
5905   }
5906 
5907   // Set up disassembler.
5908   std::unique_ptr<const MCRegisterInfo> MRI(
5909       TheTarget->createMCRegInfo(TripleName));
5910   std::unique_ptr<const MCAsmInfo> AsmInfo(
5911       TheTarget->createMCAsmInfo(*MRI, TripleName));
5912   std::unique_ptr<const MCSubtargetInfo> STI(
5913       TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
5914   MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
5915   std::unique_ptr<MCDisassembler> DisAsm(
5916       TheTarget->createMCDisassembler(*STI, Ctx));
5917   std::unique_ptr<MCSymbolizer> Symbolizer;
5918   struct DisassembleInfo SymbolizerInfo;
5919   std::unique_ptr<MCRelocationInfo> RelInfo(
5920       TheTarget->createMCRelocationInfo(TripleName, Ctx));
5921   if (RelInfo) {
5922     Symbolizer.reset(TheTarget->createMCSymbolizer(
5923         TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
5924         &SymbolizerInfo, &Ctx, std::move(RelInfo)));
5925     DisAsm->setSymbolizer(std::move(Symbolizer));
5926   }
5927   int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
5928   std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
5929       Triple(TripleName), AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI));
5930   // Set the display preference for hex vs. decimal immediates.
5931   IP->setPrintImmHex(PrintImmHex);
5932   // Comment stream and backing vector.
5933   SmallString<128> CommentsToEmit;
5934   raw_svector_ostream CommentStream(CommentsToEmit);
5935   // FIXME: Setting the CommentStream in the InstPrinter is problematic in that
5936   // if it is done then arm64 comments for string literals don't get printed
5937   // and some constant get printed instead and not setting it causes intel
5938   // (32-bit and 64-bit) comments printed with different spacing before the
5939   // comment causing different diffs with the 'C' disassembler library API.
5940   // IP->setCommentStream(CommentStream);
5941 
5942   if (!AsmInfo || !STI || !DisAsm || !IP) {
5943     errs() << "error: couldn't initialize disassembler for target "
5944            << TripleName << '\n';
5945     return;
5946   }
5947 
5948   // Set up thumb disassembler.
5949   std::unique_ptr<const MCRegisterInfo> ThumbMRI;
5950   std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
5951   std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
5952   std::unique_ptr<MCDisassembler> ThumbDisAsm;
5953   std::unique_ptr<MCInstPrinter> ThumbIP;
5954   std::unique_ptr<MCContext> ThumbCtx;
5955   std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
5956   struct DisassembleInfo ThumbSymbolizerInfo;
5957   std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
5958   if (ThumbTarget) {
5959     ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
5960     ThumbAsmInfo.reset(
5961         ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
5962     ThumbSTI.reset(
5963         ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
5964     ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
5965     ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
5966     MCContext *PtrThumbCtx = ThumbCtx.get();
5967     ThumbRelInfo.reset(
5968         ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
5969     if (ThumbRelInfo) {
5970       ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
5971           ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
5972           &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo)));
5973       ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
5974     }
5975     int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
5976     ThumbIP.reset(ThumbTarget->createMCInstPrinter(
5977         Triple(ThumbTripleName), ThumbAsmPrinterVariant, *ThumbAsmInfo,
5978         *ThumbInstrInfo, *ThumbMRI));
5979     // Set the display preference for hex vs. decimal immediates.
5980     ThumbIP->setPrintImmHex(PrintImmHex);
5981   }
5982 
5983   if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
5984     errs() << "error: couldn't initialize disassembler for target "
5985            << ThumbTripleName << '\n';
5986     return;
5987   }
5988 
5989   MachO::mach_header Header = MachOOF->getHeader();
5990 
5991   // FIXME: Using the -cfg command line option, this code used to be able to
5992   // annotate relocations with the referenced symbol's name, and if this was
5993   // inside a __[cf]string section, the data it points to. This is now replaced
5994   // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
5995   std::vector<SectionRef> Sections;
5996   std::vector<SymbolRef> Symbols;
5997   SmallVector<uint64_t, 8> FoundFns;
5998   uint64_t BaseSegmentAddress;
5999 
6000   getSectionsAndSymbols(MachOOF, Sections, Symbols, FoundFns,
6001                         BaseSegmentAddress);
6002 
6003   // Sort the symbols by address, just in case they didn't come in that way.
6004   std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
6005 
6006   // Build a data in code table that is sorted on by the address of each entry.
6007   uint64_t BaseAddress = 0;
6008   if (Header.filetype == MachO::MH_OBJECT)
6009     BaseAddress = Sections[0].getAddress();
6010   else
6011     BaseAddress = BaseSegmentAddress;
6012   DiceTable Dices;
6013   for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
6014        DI != DE; ++DI) {
6015     uint32_t Offset;
6016     DI->getOffset(Offset);
6017     Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
6018   }
6019   array_pod_sort(Dices.begin(), Dices.end());
6020 
6021 #ifndef NDEBUG
6022   raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
6023 #else
6024   raw_ostream &DebugOut = nulls();
6025 #endif
6026 
6027   std::unique_ptr<DIContext> diContext;
6028   ObjectFile *DbgObj = MachOOF;
6029   // Try to find debug info and set up the DIContext for it.
6030   if (UseDbg) {
6031     // A separate DSym file path was specified, parse it as a macho file,
6032     // get the sections and supply it to the section name parsing machinery.
6033     if (!DSYMFile.empty()) {
6034       ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
6035           MemoryBuffer::getFileOrSTDIN(DSYMFile);
6036       if (std::error_code EC = BufOrErr.getError()) {
6037         errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
6038         return;
6039       }
6040       DbgObj =
6041           ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
6042               .get()
6043               .release();
6044     }
6045 
6046     // Setup the DIContext
6047     diContext.reset(new DWARFContextInMemory(*DbgObj));
6048   }
6049 
6050   if (FilterSections.size() == 0)
6051     outs() << "(" << DisSegName << "," << DisSectName << ") section\n";
6052 
6053   for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
6054     StringRef SectName;
6055     if (Sections[SectIdx].getName(SectName) || SectName != DisSectName)
6056       continue;
6057 
6058     DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
6059 
6060     StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
6061     if (SegmentName != DisSegName)
6062       continue;
6063 
6064     StringRef BytesStr;
6065     Sections[SectIdx].getContents(BytesStr);
6066     ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
6067                             BytesStr.size());
6068     uint64_t SectAddress = Sections[SectIdx].getAddress();
6069 
6070     bool symbolTableWorked = false;
6071 
6072     // Create a map of symbol addresses to symbol names for use by
6073     // the SymbolizerSymbolLookUp() routine.
6074     SymbolAddressMap AddrMap;
6075     bool DisSymNameFound = false;
6076     for (const SymbolRef &Symbol : MachOOF->symbols()) {
6077       SymbolRef::Type ST = Symbol.getType();
6078       if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
6079           ST == SymbolRef::ST_Other) {
6080         uint64_t Address = Symbol.getValue();
6081         ErrorOr<StringRef> SymNameOrErr = Symbol.getName();
6082         if (std::error_code EC = SymNameOrErr.getError())
6083           report_fatal_error(EC.message());
6084         StringRef SymName = *SymNameOrErr;
6085         AddrMap[Address] = SymName;
6086         if (!DisSymName.empty() && DisSymName == SymName)
6087           DisSymNameFound = true;
6088       }
6089     }
6090     if (!DisSymName.empty() && !DisSymNameFound) {
6091       outs() << "Can't find -dis-symname: " << DisSymName << "\n";
6092       return;
6093     }
6094     // Set up the block of info used by the Symbolizer call backs.
6095     SymbolizerInfo.verbose = !NoSymbolicOperands;
6096     SymbolizerInfo.O = MachOOF;
6097     SymbolizerInfo.S = Sections[SectIdx];
6098     SymbolizerInfo.AddrMap = &AddrMap;
6099     SymbolizerInfo.Sections = &Sections;
6100     SymbolizerInfo.class_name = nullptr;
6101     SymbolizerInfo.selector_name = nullptr;
6102     SymbolizerInfo.method = nullptr;
6103     SymbolizerInfo.demangled_name = nullptr;
6104     SymbolizerInfo.bindtable = nullptr;
6105     SymbolizerInfo.adrp_addr = 0;
6106     SymbolizerInfo.adrp_inst = 0;
6107     // Same for the ThumbSymbolizer
6108     ThumbSymbolizerInfo.verbose = !NoSymbolicOperands;
6109     ThumbSymbolizerInfo.O = MachOOF;
6110     ThumbSymbolizerInfo.S = Sections[SectIdx];
6111     ThumbSymbolizerInfo.AddrMap = &AddrMap;
6112     ThumbSymbolizerInfo.Sections = &Sections;
6113     ThumbSymbolizerInfo.class_name = nullptr;
6114     ThumbSymbolizerInfo.selector_name = nullptr;
6115     ThumbSymbolizerInfo.method = nullptr;
6116     ThumbSymbolizerInfo.demangled_name = nullptr;
6117     ThumbSymbolizerInfo.bindtable = nullptr;
6118     ThumbSymbolizerInfo.adrp_addr = 0;
6119     ThumbSymbolizerInfo.adrp_inst = 0;
6120 
6121     // Disassemble symbol by symbol.
6122     for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
6123       ErrorOr<StringRef> SymNameOrErr = Symbols[SymIdx].getName();
6124       if (std::error_code EC = SymNameOrErr.getError())
6125         report_fatal_error(EC.message());
6126       StringRef SymName = *SymNameOrErr;
6127 
6128       SymbolRef::Type ST = Symbols[SymIdx].getType();
6129       if (ST != SymbolRef::ST_Function && ST != SymbolRef::ST_Data)
6130         continue;
6131 
6132       // Make sure the symbol is defined in this section.
6133       bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
6134       if (!containsSym)
6135         continue;
6136 
6137       // If we are only disassembling one symbol see if this is that symbol.
6138       if (!DisSymName.empty() && DisSymName != SymName)
6139         continue;
6140 
6141       // Start at the address of the symbol relative to the section's address.
6142       uint64_t Start = Symbols[SymIdx].getValue();
6143       uint64_t SectionAddress = Sections[SectIdx].getAddress();
6144       Start -= SectionAddress;
6145 
6146       // Stop disassembling either at the beginning of the next symbol or at
6147       // the end of the section.
6148       bool containsNextSym = false;
6149       uint64_t NextSym = 0;
6150       uint64_t NextSymIdx = SymIdx + 1;
6151       while (Symbols.size() > NextSymIdx) {
6152         SymbolRef::Type NextSymType = Symbols[NextSymIdx].getType();
6153         if (NextSymType == SymbolRef::ST_Function) {
6154           containsNextSym =
6155               Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
6156           NextSym = Symbols[NextSymIdx].getValue();
6157           NextSym -= SectionAddress;
6158           break;
6159         }
6160         ++NextSymIdx;
6161       }
6162 
6163       uint64_t SectSize = Sections[SectIdx].getSize();
6164       uint64_t End = containsNextSym ? NextSym : SectSize;
6165       uint64_t Size;
6166 
6167       symbolTableWorked = true;
6168 
6169       DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
6170       bool isThumb =
6171           (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
6172 
6173       outs() << SymName << ":\n";
6174       DILineInfo lastLine;
6175       for (uint64_t Index = Start; Index < End; Index += Size) {
6176         MCInst Inst;
6177 
6178         uint64_t PC = SectAddress + Index;
6179         if (!NoLeadingAddr) {
6180           if (FullLeadingAddr) {
6181             if (MachOOF->is64Bit())
6182               outs() << format("%016" PRIx64, PC);
6183             else
6184               outs() << format("%08" PRIx64, PC);
6185           } else {
6186             outs() << format("%8" PRIx64 ":", PC);
6187           }
6188         }
6189         if (!NoShowRawInsn)
6190           outs() << "\t";
6191 
6192         // Check the data in code table here to see if this is data not an
6193         // instruction to be disassembled.
6194         DiceTable Dice;
6195         Dice.push_back(std::make_pair(PC, DiceRef()));
6196         dice_table_iterator DTI =
6197             std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
6198                         compareDiceTableEntries);
6199         if (DTI != Dices.end()) {
6200           uint16_t Length;
6201           DTI->second.getLength(Length);
6202           uint16_t Kind;
6203           DTI->second.getKind(Kind);
6204           Size = DumpDataInCode(Bytes.data() + Index, Length, Kind);
6205           if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
6206               (PC == (DTI->first + Length - 1)) && (Length & 1))
6207             Size++;
6208           continue;
6209         }
6210 
6211         SmallVector<char, 64> AnnotationsBytes;
6212         raw_svector_ostream Annotations(AnnotationsBytes);
6213 
6214         bool gotInst;
6215         if (isThumb)
6216           gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
6217                                                 PC, DebugOut, Annotations);
6218         else
6219           gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
6220                                            DebugOut, Annotations);
6221         if (gotInst) {
6222           if (!NoShowRawInsn) {
6223             dumpBytes(makeArrayRef(Bytes.data() + Index, Size), outs());
6224           }
6225           formatted_raw_ostream FormattedOS(outs());
6226           StringRef AnnotationsStr = Annotations.str();
6227           if (isThumb)
6228             ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr, *ThumbSTI);
6229           else
6230             IP->printInst(&Inst, FormattedOS, AnnotationsStr, *STI);
6231           emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
6232 
6233           // Print debug info.
6234           if (diContext) {
6235             DILineInfo dli = diContext->getLineInfoForAddress(PC);
6236             // Print valid line info if it changed.
6237             if (dli != lastLine && dli.Line != 0)
6238               outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
6239                      << dli.Column;
6240             lastLine = dli;
6241           }
6242           outs() << "\n";
6243         } else {
6244           unsigned int Arch = MachOOF->getArch();
6245           if (Arch == Triple::x86_64 || Arch == Triple::x86) {
6246             outs() << format("\t.byte 0x%02x #bad opcode\n",
6247                              *(Bytes.data() + Index) & 0xff);
6248             Size = 1; // skip exactly one illegible byte and move on.
6249           } else if (Arch == Triple::aarch64) {
6250             uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
6251                               (*(Bytes.data() + Index + 1) & 0xff) << 8 |
6252                               (*(Bytes.data() + Index + 2) & 0xff) << 16 |
6253                               (*(Bytes.data() + Index + 3) & 0xff) << 24;
6254             outs() << format("\t.long\t0x%08x\n", opcode);
6255             Size = 4;
6256           } else {
6257             errs() << "llvm-objdump: warning: invalid instruction encoding\n";
6258             if (Size == 0)
6259               Size = 1; // skip illegible bytes
6260           }
6261         }
6262       }
6263     }
6264     if (!symbolTableWorked) {
6265       // Reading the symbol table didn't work, disassemble the whole section.
6266       uint64_t SectAddress = Sections[SectIdx].getAddress();
6267       uint64_t SectSize = Sections[SectIdx].getSize();
6268       uint64_t InstSize;
6269       for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
6270         MCInst Inst;
6271 
6272         uint64_t PC = SectAddress + Index;
6273         if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
6274                                    DebugOut, nulls())) {
6275           if (!NoLeadingAddr) {
6276             if (FullLeadingAddr) {
6277               if (MachOOF->is64Bit())
6278                 outs() << format("%016" PRIx64, PC);
6279               else
6280                 outs() << format("%08" PRIx64, PC);
6281             } else {
6282               outs() << format("%8" PRIx64 ":", PC);
6283             }
6284           }
6285           if (!NoShowRawInsn) {
6286             outs() << "\t";
6287             dumpBytes(makeArrayRef(Bytes.data() + Index, InstSize), outs());
6288           }
6289           IP->printInst(&Inst, outs(), "", *STI);
6290           outs() << "\n";
6291         } else {
6292           unsigned int Arch = MachOOF->getArch();
6293           if (Arch == Triple::x86_64 || Arch == Triple::x86) {
6294             outs() << format("\t.byte 0x%02x #bad opcode\n",
6295                              *(Bytes.data() + Index) & 0xff);
6296             InstSize = 1; // skip exactly one illegible byte and move on.
6297           } else {
6298             errs() << "llvm-objdump: warning: invalid instruction encoding\n";
6299             if (InstSize == 0)
6300               InstSize = 1; // skip illegible bytes
6301           }
6302         }
6303       }
6304     }
6305     // The TripleName's need to be reset if we are called again for a different
6306     // archtecture.
6307     TripleName = "";
6308     ThumbTripleName = "";
6309 
6310     if (SymbolizerInfo.method != nullptr)
6311       free(SymbolizerInfo.method);
6312     if (SymbolizerInfo.demangled_name != nullptr)
6313       free(SymbolizerInfo.demangled_name);
6314     if (SymbolizerInfo.bindtable != nullptr)
6315       delete SymbolizerInfo.bindtable;
6316     if (ThumbSymbolizerInfo.method != nullptr)
6317       free(ThumbSymbolizerInfo.method);
6318     if (ThumbSymbolizerInfo.demangled_name != nullptr)
6319       free(ThumbSymbolizerInfo.demangled_name);
6320     if (ThumbSymbolizerInfo.bindtable != nullptr)
6321       delete ThumbSymbolizerInfo.bindtable;
6322   }
6323 }
6324 
6325 //===----------------------------------------------------------------------===//
6326 // __compact_unwind section dumping
6327 //===----------------------------------------------------------------------===//
6328 
6329 namespace {
6330 
readNext(const char * & Buf)6331 template <typename T> static uint64_t readNext(const char *&Buf) {
6332   using llvm::support::little;
6333   using llvm::support::unaligned;
6334 
6335   uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
6336   Buf += sizeof(T);
6337   return Val;
6338 }
6339 
6340 struct CompactUnwindEntry {
6341   uint32_t OffsetInSection;
6342 
6343   uint64_t FunctionAddr;
6344   uint32_t Length;
6345   uint32_t CompactEncoding;
6346   uint64_t PersonalityAddr;
6347   uint64_t LSDAAddr;
6348 
6349   RelocationRef FunctionReloc;
6350   RelocationRef PersonalityReloc;
6351   RelocationRef LSDAReloc;
6352 
CompactUnwindEntry__anon514500d20511::CompactUnwindEntry6353   CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
6354       : OffsetInSection(Offset) {
6355     if (Is64)
6356       read<uint64_t>(Contents.data() + Offset);
6357     else
6358       read<uint32_t>(Contents.data() + Offset);
6359   }
6360 
6361 private:
read__anon514500d20511::CompactUnwindEntry6362   template <typename UIntPtr> void read(const char *Buf) {
6363     FunctionAddr = readNext<UIntPtr>(Buf);
6364     Length = readNext<uint32_t>(Buf);
6365     CompactEncoding = readNext<uint32_t>(Buf);
6366     PersonalityAddr = readNext<UIntPtr>(Buf);
6367     LSDAAddr = readNext<UIntPtr>(Buf);
6368   }
6369 };
6370 }
6371 
6372 /// Given a relocation from __compact_unwind, consisting of the RelocationRef
6373 /// and data being relocated, determine the best base Name and Addend to use for
6374 /// display purposes.
6375 ///
6376 /// 1. An Extern relocation will directly reference a symbol (and the data is
6377 ///    then already an addend), so use that.
6378 /// 2. Otherwise the data is an offset in the object file's layout; try to find
6379 //     a symbol before it in the same section, and use the offset from there.
6380 /// 3. Finally, if all that fails, fall back to an offset from the start of the
6381 ///    referenced section.
findUnwindRelocNameAddend(const MachOObjectFile * Obj,std::map<uint64_t,SymbolRef> & Symbols,const RelocationRef & Reloc,uint64_t Addr,StringRef & Name,uint64_t & Addend)6382 static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
6383                                       std::map<uint64_t, SymbolRef> &Symbols,
6384                                       const RelocationRef &Reloc, uint64_t Addr,
6385                                       StringRef &Name, uint64_t &Addend) {
6386   if (Reloc.getSymbol() != Obj->symbol_end()) {
6387     ErrorOr<StringRef> NameOrErr = Reloc.getSymbol()->getName();
6388     if (std::error_code EC = NameOrErr.getError())
6389       report_fatal_error(EC.message());
6390     Name = *NameOrErr;
6391     Addend = Addr;
6392     return;
6393   }
6394 
6395   auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
6396   SectionRef RelocSection = Obj->getAnyRelocationSection(RE);
6397 
6398   uint64_t SectionAddr = RelocSection.getAddress();
6399 
6400   auto Sym = Symbols.upper_bound(Addr);
6401   if (Sym == Symbols.begin()) {
6402     // The first symbol in the object is after this reference, the best we can
6403     // do is section-relative notation.
6404     RelocSection.getName(Name);
6405     Addend = Addr - SectionAddr;
6406     return;
6407   }
6408 
6409   // Go back one so that SymbolAddress <= Addr.
6410   --Sym;
6411 
6412   section_iterator SymSection = *Sym->second.getSection();
6413   if (RelocSection == *SymSection) {
6414     // There's a valid symbol in the same section before this reference.
6415     ErrorOr<StringRef> NameOrErr = Sym->second.getName();
6416     if (std::error_code EC = NameOrErr.getError())
6417       report_fatal_error(EC.message());
6418     Name = *NameOrErr;
6419     Addend = Addr - Sym->first;
6420     return;
6421   }
6422 
6423   // There is a symbol before this reference, but it's in a different
6424   // section. Probably not helpful to mention it, so use the section name.
6425   RelocSection.getName(Name);
6426   Addend = Addr - SectionAddr;
6427 }
6428 
printUnwindRelocDest(const MachOObjectFile * Obj,std::map<uint64_t,SymbolRef> & Symbols,const RelocationRef & Reloc,uint64_t Addr)6429 static void printUnwindRelocDest(const MachOObjectFile *Obj,
6430                                  std::map<uint64_t, SymbolRef> &Symbols,
6431                                  const RelocationRef &Reloc, uint64_t Addr) {
6432   StringRef Name;
6433   uint64_t Addend;
6434 
6435   if (!Reloc.getObject())
6436     return;
6437 
6438   findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
6439 
6440   outs() << Name;
6441   if (Addend)
6442     outs() << " + " << format("0x%" PRIx64, Addend);
6443 }
6444 
6445 static void
printMachOCompactUnwindSection(const MachOObjectFile * Obj,std::map<uint64_t,SymbolRef> & Symbols,const SectionRef & CompactUnwind)6446 printMachOCompactUnwindSection(const MachOObjectFile *Obj,
6447                                std::map<uint64_t, SymbolRef> &Symbols,
6448                                const SectionRef &CompactUnwind) {
6449 
6450   assert(Obj->isLittleEndian() &&
6451          "There should not be a big-endian .o with __compact_unwind");
6452 
6453   bool Is64 = Obj->is64Bit();
6454   uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
6455   uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
6456 
6457   StringRef Contents;
6458   CompactUnwind.getContents(Contents);
6459 
6460   SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
6461 
6462   // First populate the initial raw offsets, encodings and so on from the entry.
6463   for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
6464     CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
6465     CompactUnwinds.push_back(Entry);
6466   }
6467 
6468   // Next we need to look at the relocations to find out what objects are
6469   // actually being referred to.
6470   for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
6471     uint64_t RelocAddress = Reloc.getOffset();
6472 
6473     uint32_t EntryIdx = RelocAddress / EntrySize;
6474     uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
6475     CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
6476 
6477     if (OffsetInEntry == 0)
6478       Entry.FunctionReloc = Reloc;
6479     else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
6480       Entry.PersonalityReloc = Reloc;
6481     else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
6482       Entry.LSDAReloc = Reloc;
6483     else
6484       llvm_unreachable("Unexpected relocation in __compact_unwind section");
6485   }
6486 
6487   // Finally, we're ready to print the data we've gathered.
6488   outs() << "Contents of __compact_unwind section:\n";
6489   for (auto &Entry : CompactUnwinds) {
6490     outs() << "  Entry at offset "
6491            << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
6492 
6493     // 1. Start of the region this entry applies to.
6494     outs() << "    start:                " << format("0x%" PRIx64,
6495                                                      Entry.FunctionAddr) << ' ';
6496     printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
6497     outs() << '\n';
6498 
6499     // 2. Length of the region this entry applies to.
6500     outs() << "    length:               " << format("0x%" PRIx32, Entry.Length)
6501            << '\n';
6502     // 3. The 32-bit compact encoding.
6503     outs() << "    compact encoding:     "
6504            << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
6505 
6506     // 4. The personality function, if present.
6507     if (Entry.PersonalityReloc.getObject()) {
6508       outs() << "    personality function: "
6509              << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
6510       printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
6511                            Entry.PersonalityAddr);
6512       outs() << '\n';
6513     }
6514 
6515     // 5. This entry's language-specific data area.
6516     if (Entry.LSDAReloc.getObject()) {
6517       outs() << "    LSDA:                 " << format("0x%" PRIx64,
6518                                                        Entry.LSDAAddr) << ' ';
6519       printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
6520       outs() << '\n';
6521     }
6522   }
6523 }
6524 
6525 //===----------------------------------------------------------------------===//
6526 // __unwind_info section dumping
6527 //===----------------------------------------------------------------------===//
6528 
printRegularSecondLevelUnwindPage(const char * PageStart)6529 static void printRegularSecondLevelUnwindPage(const char *PageStart) {
6530   const char *Pos = PageStart;
6531   uint32_t Kind = readNext<uint32_t>(Pos);
6532   (void)Kind;
6533   assert(Kind == 2 && "kind for a regular 2nd level index should be 2");
6534 
6535   uint16_t EntriesStart = readNext<uint16_t>(Pos);
6536   uint16_t NumEntries = readNext<uint16_t>(Pos);
6537 
6538   Pos = PageStart + EntriesStart;
6539   for (unsigned i = 0; i < NumEntries; ++i) {
6540     uint32_t FunctionOffset = readNext<uint32_t>(Pos);
6541     uint32_t Encoding = readNext<uint32_t>(Pos);
6542 
6543     outs() << "      [" << i << "]: "
6544            << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
6545            << ", "
6546            << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n';
6547   }
6548 }
6549 
printCompressedSecondLevelUnwindPage(const char * PageStart,uint32_t FunctionBase,const SmallVectorImpl<uint32_t> & CommonEncodings)6550 static void printCompressedSecondLevelUnwindPage(
6551     const char *PageStart, uint32_t FunctionBase,
6552     const SmallVectorImpl<uint32_t> &CommonEncodings) {
6553   const char *Pos = PageStart;
6554   uint32_t Kind = readNext<uint32_t>(Pos);
6555   (void)Kind;
6556   assert(Kind == 3 && "kind for a compressed 2nd level index should be 3");
6557 
6558   uint16_t EntriesStart = readNext<uint16_t>(Pos);
6559   uint16_t NumEntries = readNext<uint16_t>(Pos);
6560 
6561   uint16_t EncodingsStart = readNext<uint16_t>(Pos);
6562   readNext<uint16_t>(Pos);
6563   const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
6564       PageStart + EncodingsStart);
6565 
6566   Pos = PageStart + EntriesStart;
6567   for (unsigned i = 0; i < NumEntries; ++i) {
6568     uint32_t Entry = readNext<uint32_t>(Pos);
6569     uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
6570     uint32_t EncodingIdx = Entry >> 24;
6571 
6572     uint32_t Encoding;
6573     if (EncodingIdx < CommonEncodings.size())
6574       Encoding = CommonEncodings[EncodingIdx];
6575     else
6576       Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
6577 
6578     outs() << "      [" << i << "]: "
6579            << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
6580            << ", "
6581            << "encoding[" << EncodingIdx
6582            << "]=" << format("0x%08" PRIx32, Encoding) << '\n';
6583   }
6584 }
6585 
printMachOUnwindInfoSection(const MachOObjectFile * Obj,std::map<uint64_t,SymbolRef> & Symbols,const SectionRef & UnwindInfo)6586 static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
6587                                         std::map<uint64_t, SymbolRef> &Symbols,
6588                                         const SectionRef &UnwindInfo) {
6589 
6590   assert(Obj->isLittleEndian() &&
6591          "There should not be a big-endian .o with __unwind_info");
6592 
6593   outs() << "Contents of __unwind_info section:\n";
6594 
6595   StringRef Contents;
6596   UnwindInfo.getContents(Contents);
6597   const char *Pos = Contents.data();
6598 
6599   //===----------------------------------
6600   // Section header
6601   //===----------------------------------
6602 
6603   uint32_t Version = readNext<uint32_t>(Pos);
6604   outs() << "  Version:                                   "
6605          << format("0x%" PRIx32, Version) << '\n';
6606   assert(Version == 1 && "only understand version 1");
6607 
6608   uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
6609   outs() << "  Common encodings array section offset:     "
6610          << format("0x%" PRIx32, CommonEncodingsStart) << '\n';
6611   uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
6612   outs() << "  Number of common encodings in array:       "
6613          << format("0x%" PRIx32, NumCommonEncodings) << '\n';
6614 
6615   uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
6616   outs() << "  Personality function array section offset: "
6617          << format("0x%" PRIx32, PersonalitiesStart) << '\n';
6618   uint32_t NumPersonalities = readNext<uint32_t>(Pos);
6619   outs() << "  Number of personality functions in array:  "
6620          << format("0x%" PRIx32, NumPersonalities) << '\n';
6621 
6622   uint32_t IndicesStart = readNext<uint32_t>(Pos);
6623   outs() << "  Index array section offset:                "
6624          << format("0x%" PRIx32, IndicesStart) << '\n';
6625   uint32_t NumIndices = readNext<uint32_t>(Pos);
6626   outs() << "  Number of indices in array:                "
6627          << format("0x%" PRIx32, NumIndices) << '\n';
6628 
6629   //===----------------------------------
6630   // A shared list of common encodings
6631   //===----------------------------------
6632 
6633   // These occupy indices in the range [0, N] whenever an encoding is referenced
6634   // from a compressed 2nd level index table. In practice the linker only
6635   // creates ~128 of these, so that indices are available to embed encodings in
6636   // the 2nd level index.
6637 
6638   SmallVector<uint32_t, 64> CommonEncodings;
6639   outs() << "  Common encodings: (count = " << NumCommonEncodings << ")\n";
6640   Pos = Contents.data() + CommonEncodingsStart;
6641   for (unsigned i = 0; i < NumCommonEncodings; ++i) {
6642     uint32_t Encoding = readNext<uint32_t>(Pos);
6643     CommonEncodings.push_back(Encoding);
6644 
6645     outs() << "    encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding)
6646            << '\n';
6647   }
6648 
6649   //===----------------------------------
6650   // Personality functions used in this executable
6651   //===----------------------------------
6652 
6653   // There should be only a handful of these (one per source language,
6654   // roughly). Particularly since they only get 2 bits in the compact encoding.
6655 
6656   outs() << "  Personality functions: (count = " << NumPersonalities << ")\n";
6657   Pos = Contents.data() + PersonalitiesStart;
6658   for (unsigned i = 0; i < NumPersonalities; ++i) {
6659     uint32_t PersonalityFn = readNext<uint32_t>(Pos);
6660     outs() << "    personality[" << i + 1
6661            << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n';
6662   }
6663 
6664   //===----------------------------------
6665   // The level 1 index entries
6666   //===----------------------------------
6667 
6668   // These specify an approximate place to start searching for the more detailed
6669   // information, sorted by PC.
6670 
6671   struct IndexEntry {
6672     uint32_t FunctionOffset;
6673     uint32_t SecondLevelPageStart;
6674     uint32_t LSDAStart;
6675   };
6676 
6677   SmallVector<IndexEntry, 4> IndexEntries;
6678 
6679   outs() << "  Top level indices: (count = " << NumIndices << ")\n";
6680   Pos = Contents.data() + IndicesStart;
6681   for (unsigned i = 0; i < NumIndices; ++i) {
6682     IndexEntry Entry;
6683 
6684     Entry.FunctionOffset = readNext<uint32_t>(Pos);
6685     Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
6686     Entry.LSDAStart = readNext<uint32_t>(Pos);
6687     IndexEntries.push_back(Entry);
6688 
6689     outs() << "    [" << i << "]: "
6690            << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset)
6691            << ", "
6692            << "2nd level page offset="
6693            << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", "
6694            << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n';
6695   }
6696 
6697   //===----------------------------------
6698   // Next come the LSDA tables
6699   //===----------------------------------
6700 
6701   // The LSDA layout is rather implicit: it's a contiguous array of entries from
6702   // the first top-level index's LSDAOffset to the last (sentinel).
6703 
6704   outs() << "  LSDA descriptors:\n";
6705   Pos = Contents.data() + IndexEntries[0].LSDAStart;
6706   int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
6707                  (2 * sizeof(uint32_t));
6708   for (int i = 0; i < NumLSDAs; ++i) {
6709     uint32_t FunctionOffset = readNext<uint32_t>(Pos);
6710     uint32_t LSDAOffset = readNext<uint32_t>(Pos);
6711     outs() << "    [" << i << "]: "
6712            << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
6713            << ", "
6714            << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n';
6715   }
6716 
6717   //===----------------------------------
6718   // Finally, the 2nd level indices
6719   //===----------------------------------
6720 
6721   // Generally these are 4K in size, and have 2 possible forms:
6722   //   + Regular stores up to 511 entries with disparate encodings
6723   //   + Compressed stores up to 1021 entries if few enough compact encoding
6724   //     values are used.
6725   outs() << "  Second level indices:\n";
6726   for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
6727     // The final sentinel top-level index has no associated 2nd level page
6728     if (IndexEntries[i].SecondLevelPageStart == 0)
6729       break;
6730 
6731     outs() << "    Second level index[" << i << "]: "
6732            << "offset in section="
6733            << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart)
6734            << ", "
6735            << "base function offset="
6736            << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n';
6737 
6738     Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
6739     uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
6740     if (Kind == 2)
6741       printRegularSecondLevelUnwindPage(Pos);
6742     else if (Kind == 3)
6743       printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
6744                                            CommonEncodings);
6745     else
6746       llvm_unreachable("Do not know how to print this kind of 2nd level page");
6747   }
6748 }
6749 
getSizeForEncoding(bool is64Bit,unsigned symbolEncoding)6750 static unsigned getSizeForEncoding(bool is64Bit,
6751                                    unsigned symbolEncoding) {
6752   unsigned format = symbolEncoding & 0x0f;
6753   switch (format) {
6754     default: llvm_unreachable("Unknown Encoding");
6755     case dwarf::DW_EH_PE_absptr:
6756     case dwarf::DW_EH_PE_signed:
6757       return is64Bit ? 8 : 4;
6758     case dwarf::DW_EH_PE_udata2:
6759     case dwarf::DW_EH_PE_sdata2:
6760       return 2;
6761     case dwarf::DW_EH_PE_udata4:
6762     case dwarf::DW_EH_PE_sdata4:
6763       return 4;
6764     case dwarf::DW_EH_PE_udata8:
6765     case dwarf::DW_EH_PE_sdata8:
6766       return 8;
6767   }
6768 }
6769 
readPointer(const char * & Pos,bool is64Bit,unsigned Encoding)6770 static uint64_t readPointer(const char *&Pos, bool is64Bit, unsigned Encoding) {
6771   switch (getSizeForEncoding(is64Bit, Encoding)) {
6772     case 2:
6773       return readNext<uint16_t>(Pos);
6774       break;
6775     case 4:
6776       return readNext<uint32_t>(Pos);
6777       break;
6778     case 8:
6779       return readNext<uint64_t>(Pos);
6780       break;
6781     default:
6782       llvm_unreachable("Illegal data size");
6783   }
6784 }
6785 
printMachOEHFrameSection(const MachOObjectFile * Obj,std::map<uint64_t,SymbolRef> & Symbols,const SectionRef & EHFrame)6786 static void printMachOEHFrameSection(const MachOObjectFile *Obj,
6787                                      std::map<uint64_t, SymbolRef> &Symbols,
6788                                      const SectionRef &EHFrame) {
6789   if (!Obj->isLittleEndian()) {
6790     outs() << "warning: cannot handle big endian __eh_frame section\n";
6791     return;
6792   }
6793 
6794   bool is64Bit = Obj->is64Bit();
6795 
6796   outs() << "Contents of __eh_frame section:\n";
6797 
6798   StringRef Contents;
6799   EHFrame.getContents(Contents);
6800 
6801   /// A few fields of the CIE are used when decoding the FDE's.  This struct
6802   /// will cache those fields we need so that we don't have to decode it
6803   /// repeatedly for each FDE that references it.
6804   struct DecodedCIE {
6805     Optional<uint32_t> FDEPointerEncoding;
6806     Optional<uint32_t> LSDAPointerEncoding;
6807     bool hasAugmentationLength;
6808   };
6809 
6810   // Map from the start offset of the CIE to the cached data for that CIE.
6811   DenseMap<uint64_t, DecodedCIE> CachedCIEs;
6812 
6813   for (const char *Pos = Contents.data(), *End = Contents.end(); Pos != End; ) {
6814 
6815     const char *EntryStartPos = Pos;
6816 
6817     uint64_t Length = readNext<uint32_t>(Pos);
6818     if (Length == 0xffffffff)
6819       Length = readNext<uint64_t>(Pos);
6820 
6821     // Save the Pos so that we can check the length we encoded against what we
6822     // end up decoding.
6823     const char *PosAfterLength = Pos;
6824     const char *EntryEndPos = PosAfterLength + Length;
6825 
6826     assert(EntryEndPos <= End &&
6827            "__eh_frame entry length exceeds section size");
6828 
6829     uint32_t ID = readNext<uint32_t>(Pos);
6830     if (ID == 0) {
6831       // This is a CIE.
6832 
6833       uint32_t Version = readNext<uint8_t>(Pos);
6834 
6835       // Parse a null terminated augmentation string
6836       SmallString<8> AugmentationString;
6837       for (uint8_t Char = readNext<uint8_t>(Pos); Char;
6838            Char = readNext<uint8_t>(Pos))
6839         AugmentationString.push_back(Char);
6840 
6841       // Optionally parse the EH data if the augmentation string says it's there.
6842       Optional<uint64_t> EHData;
6843       if (StringRef(AugmentationString).count("eh"))
6844         EHData = is64Bit ? readNext<uint64_t>(Pos) : readNext<uint32_t>(Pos);
6845 
6846       unsigned ULEBByteCount;
6847       uint64_t CodeAlignmentFactor = decodeULEB128((const uint8_t *)Pos,
6848                                                    &ULEBByteCount);
6849       Pos += ULEBByteCount;
6850 
6851       int64_t DataAlignmentFactor = decodeSLEB128((const uint8_t *)Pos,
6852                                                    &ULEBByteCount);
6853       Pos += ULEBByteCount;
6854 
6855       uint32_t ReturnAddressRegister = readNext<uint8_t>(Pos);
6856 
6857       Optional<uint64_t> AugmentationLength;
6858       Optional<uint32_t> LSDAPointerEncoding;
6859       Optional<uint32_t> PersonalityEncoding;
6860       Optional<uint64_t> Personality;
6861       Optional<uint32_t> FDEPointerEncoding;
6862       if (!AugmentationString.empty() && AugmentationString.front() == 'z') {
6863         AugmentationLength = decodeULEB128((const uint8_t *)Pos,
6864                                            &ULEBByteCount);
6865         Pos += ULEBByteCount;
6866 
6867         // Walk the augmentation string to get all the augmentation data.
6868         for (unsigned i = 1, e = AugmentationString.size(); i != e; ++i) {
6869           char Char = AugmentationString[i];
6870           switch (Char) {
6871             case 'e':
6872               assert((i + 1) != e && AugmentationString[i + 1] == 'h' &&
6873                      "Expected 'eh' in augmentation string");
6874               break;
6875             case 'L':
6876               assert(!LSDAPointerEncoding && "Duplicate LSDA encoding");
6877               LSDAPointerEncoding = readNext<uint8_t>(Pos);
6878               break;
6879             case 'P': {
6880               assert(!Personality && "Duplicate personality");
6881               PersonalityEncoding = readNext<uint8_t>(Pos);
6882               Personality = readPointer(Pos, is64Bit, *PersonalityEncoding);
6883               break;
6884             }
6885             case 'R':
6886               assert(!FDEPointerEncoding && "Duplicate FDE encoding");
6887               FDEPointerEncoding = readNext<uint8_t>(Pos);
6888               break;
6889             case 'z':
6890               llvm_unreachable("'z' must be first in the augmentation string");
6891           }
6892         }
6893       }
6894 
6895       outs() << "CIE:\n";
6896       outs() << "  Length: " << Length << "\n";
6897       outs() << "  CIE ID: " << ID << "\n";
6898       outs() << "  Version: " << Version << "\n";
6899       outs() << "  Augmentation String: " << AugmentationString << "\n";
6900       if (EHData)
6901         outs() << "  EHData: " << *EHData << "\n";
6902       outs() << "  Code Alignment Factor: " << CodeAlignmentFactor << "\n";
6903       outs() << "  Data Alignment Factor: " << DataAlignmentFactor << "\n";
6904       outs() << "  Return Address Register: " << ReturnAddressRegister << "\n";
6905       if (AugmentationLength) {
6906         outs() << "  Augmentation Data Length: " << *AugmentationLength << "\n";
6907         if (LSDAPointerEncoding) {
6908           outs() << "  FDE LSDA Pointer Encoding: "
6909                  << *LSDAPointerEncoding << "\n";
6910         }
6911         if (Personality) {
6912           outs() << "  Personality Encoding: " << *PersonalityEncoding << "\n";
6913           outs() << "  Personality: " << *Personality << "\n";
6914         }
6915         if (FDEPointerEncoding) {
6916           outs() << "  FDE Address Pointer Encoding: "
6917                  << *FDEPointerEncoding << "\n";
6918         }
6919       }
6920       // FIXME: Handle instructions.
6921       // For now just emit some bytes
6922       outs() << "  Instructions:\n  ";
6923       dumpBytes(makeArrayRef((const uint8_t*)Pos, (const uint8_t*)EntryEndPos),
6924                 outs());
6925       outs() << "\n";
6926       Pos = EntryEndPos;
6927 
6928       // Cache this entry.
6929       uint64_t Offset = EntryStartPos - Contents.data();
6930       CachedCIEs[Offset] = { FDEPointerEncoding, LSDAPointerEncoding,
6931                              AugmentationLength.hasValue() };
6932       continue;
6933     }
6934 
6935     // This is an FDE.
6936     // The CIE pointer for an FDE is the same location as the ID which we
6937     // already read.
6938     uint32_t CIEPointer = ID;
6939 
6940     const char *CIEStart = PosAfterLength - CIEPointer;
6941     assert(CIEStart >= Contents.data() &&
6942            "FDE points to CIE before the __eh_frame start");
6943 
6944     uint64_t CIEOffset = CIEStart - Contents.data();
6945     auto CIEIt = CachedCIEs.find(CIEOffset);
6946     if (CIEIt == CachedCIEs.end())
6947       llvm_unreachable("Couldn't find CIE at offset in to __eh_frame section");
6948 
6949     const DecodedCIE &CIE = CIEIt->getSecond();
6950     assert(CIE.FDEPointerEncoding &&
6951            "FDE references CIE which did not set pointer encoding");
6952 
6953     uint64_t PCPointerSize = getSizeForEncoding(is64Bit,
6954                                                 *CIE.FDEPointerEncoding);
6955 
6956     uint64_t PCBegin = readPointer(Pos, is64Bit, *CIE.FDEPointerEncoding);
6957     uint64_t PCRange = readPointer(Pos, is64Bit, *CIE.FDEPointerEncoding);
6958 
6959     Optional<uint64_t> AugmentationLength;
6960     uint32_t LSDAPointerSize;
6961     Optional<uint64_t> LSDAPointer;
6962     if (CIE.hasAugmentationLength) {
6963       unsigned ULEBByteCount;
6964       AugmentationLength = decodeULEB128((const uint8_t *)Pos,
6965                                          &ULEBByteCount);
6966       Pos += ULEBByteCount;
6967 
6968       // Decode the LSDA if the CIE augmentation string said we should.
6969       if (CIE.LSDAPointerEncoding) {
6970         LSDAPointerSize = getSizeForEncoding(is64Bit, *CIE.LSDAPointerEncoding);
6971         LSDAPointer = readPointer(Pos, is64Bit, *CIE.LSDAPointerEncoding);
6972       }
6973     }
6974 
6975     outs() << "FDE:\n";
6976     outs() << "  Length: " << Length << "\n";
6977     outs() << "  CIE Offset: " << CIEOffset << "\n";
6978 
6979     if (PCPointerSize == 8) {
6980       outs() << format("  PC Begin: %016" PRIx64, PCBegin) << "\n";
6981       outs() << format("  PC Range: %016" PRIx64, PCRange) << "\n";
6982     } else {
6983       outs() << format("  PC Begin: %08" PRIx64, PCBegin) << "\n";
6984       outs() << format("  PC Range: %08" PRIx64, PCRange) << "\n";
6985     }
6986     if (AugmentationLength) {
6987       outs() << "  Augmentation Data Length: " << *AugmentationLength << "\n";
6988       if (LSDAPointer) {
6989         if (LSDAPointerSize == 8)
6990           outs() << format("  LSDA Pointer: %016\n" PRIx64, *LSDAPointer);
6991         else
6992           outs() << format("  LSDA Pointer: %08\n" PRIx64, *LSDAPointer);
6993       }
6994     }
6995 
6996     // FIXME: Handle instructions.
6997     // For now just emit some bytes
6998     outs() << "  Instructions:\n  ";
6999     dumpBytes(makeArrayRef((const uint8_t*)Pos, (const uint8_t*)EntryEndPos),
7000               outs());
7001     outs() << "\n";
7002     Pos = EntryEndPos;
7003   }
7004 }
7005 
printMachOUnwindInfo(const MachOObjectFile * Obj)7006 void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
7007   std::map<uint64_t, SymbolRef> Symbols;
7008   for (const SymbolRef &SymRef : Obj->symbols()) {
7009     // Discard any undefined or absolute symbols. They're not going to take part
7010     // in the convenience lookup for unwind info and just take up resources.
7011     section_iterator Section = *SymRef.getSection();
7012     if (Section == Obj->section_end())
7013       continue;
7014 
7015     uint64_t Addr = SymRef.getValue();
7016     Symbols.insert(std::make_pair(Addr, SymRef));
7017   }
7018 
7019   for (const SectionRef &Section : Obj->sections()) {
7020     StringRef SectName;
7021     Section.getName(SectName);
7022     if (SectName == "__compact_unwind")
7023       printMachOCompactUnwindSection(Obj, Symbols, Section);
7024     else if (SectName == "__unwind_info")
7025       printMachOUnwindInfoSection(Obj, Symbols, Section);
7026     else if (SectName == "__eh_frame")
7027       printMachOEHFrameSection(Obj, Symbols, Section);
7028   }
7029 }
7030 
PrintMachHeader(uint32_t magic,uint32_t cputype,uint32_t cpusubtype,uint32_t filetype,uint32_t ncmds,uint32_t sizeofcmds,uint32_t flags,bool verbose)7031 static void PrintMachHeader(uint32_t magic, uint32_t cputype,
7032                             uint32_t cpusubtype, uint32_t filetype,
7033                             uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
7034                             bool verbose) {
7035   outs() << "Mach header\n";
7036   outs() << "      magic cputype cpusubtype  caps    filetype ncmds "
7037             "sizeofcmds      flags\n";
7038   if (verbose) {
7039     if (magic == MachO::MH_MAGIC)
7040       outs() << "   MH_MAGIC";
7041     else if (magic == MachO::MH_MAGIC_64)
7042       outs() << "MH_MAGIC_64";
7043     else
7044       outs() << format(" 0x%08" PRIx32, magic);
7045     switch (cputype) {
7046     case MachO::CPU_TYPE_I386:
7047       outs() << "    I386";
7048       switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
7049       case MachO::CPU_SUBTYPE_I386_ALL:
7050         outs() << "        ALL";
7051         break;
7052       default:
7053         outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
7054         break;
7055       }
7056       break;
7057     case MachO::CPU_TYPE_X86_64:
7058       outs() << "  X86_64";
7059       switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
7060       case MachO::CPU_SUBTYPE_X86_64_ALL:
7061         outs() << "        ALL";
7062         break;
7063       case MachO::CPU_SUBTYPE_X86_64_H:
7064         outs() << "    Haswell";
7065         break;
7066       default:
7067         outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
7068         break;
7069       }
7070       break;
7071     case MachO::CPU_TYPE_ARM:
7072       outs() << "     ARM";
7073       switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
7074       case MachO::CPU_SUBTYPE_ARM_ALL:
7075         outs() << "        ALL";
7076         break;
7077       case MachO::CPU_SUBTYPE_ARM_V4T:
7078         outs() << "        V4T";
7079         break;
7080       case MachO::CPU_SUBTYPE_ARM_V5TEJ:
7081         outs() << "      V5TEJ";
7082         break;
7083       case MachO::CPU_SUBTYPE_ARM_XSCALE:
7084         outs() << "     XSCALE";
7085         break;
7086       case MachO::CPU_SUBTYPE_ARM_V6:
7087         outs() << "         V6";
7088         break;
7089       case MachO::CPU_SUBTYPE_ARM_V6M:
7090         outs() << "        V6M";
7091         break;
7092       case MachO::CPU_SUBTYPE_ARM_V7:
7093         outs() << "         V7";
7094         break;
7095       case MachO::CPU_SUBTYPE_ARM_V7EM:
7096         outs() << "       V7EM";
7097         break;
7098       case MachO::CPU_SUBTYPE_ARM_V7K:
7099         outs() << "        V7K";
7100         break;
7101       case MachO::CPU_SUBTYPE_ARM_V7M:
7102         outs() << "        V7M";
7103         break;
7104       case MachO::CPU_SUBTYPE_ARM_V7S:
7105         outs() << "        V7S";
7106         break;
7107       default:
7108         outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
7109         break;
7110       }
7111       break;
7112     case MachO::CPU_TYPE_ARM64:
7113       outs() << "   ARM64";
7114       switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
7115       case MachO::CPU_SUBTYPE_ARM64_ALL:
7116         outs() << "        ALL";
7117         break;
7118       default:
7119         outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
7120         break;
7121       }
7122       break;
7123     case MachO::CPU_TYPE_POWERPC:
7124       outs() << "     PPC";
7125       switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
7126       case MachO::CPU_SUBTYPE_POWERPC_ALL:
7127         outs() << "        ALL";
7128         break;
7129       default:
7130         outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
7131         break;
7132       }
7133       break;
7134     case MachO::CPU_TYPE_POWERPC64:
7135       outs() << "   PPC64";
7136       switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
7137       case MachO::CPU_SUBTYPE_POWERPC_ALL:
7138         outs() << "        ALL";
7139         break;
7140       default:
7141         outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
7142         break;
7143       }
7144       break;
7145     }
7146     if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
7147       outs() << " LIB64";
7148     } else {
7149       outs() << format("  0x%02" PRIx32,
7150                        (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
7151     }
7152     switch (filetype) {
7153     case MachO::MH_OBJECT:
7154       outs() << "      OBJECT";
7155       break;
7156     case MachO::MH_EXECUTE:
7157       outs() << "     EXECUTE";
7158       break;
7159     case MachO::MH_FVMLIB:
7160       outs() << "      FVMLIB";
7161       break;
7162     case MachO::MH_CORE:
7163       outs() << "        CORE";
7164       break;
7165     case MachO::MH_PRELOAD:
7166       outs() << "     PRELOAD";
7167       break;
7168     case MachO::MH_DYLIB:
7169       outs() << "       DYLIB";
7170       break;
7171     case MachO::MH_DYLIB_STUB:
7172       outs() << "  DYLIB_STUB";
7173       break;
7174     case MachO::MH_DYLINKER:
7175       outs() << "    DYLINKER";
7176       break;
7177     case MachO::MH_BUNDLE:
7178       outs() << "      BUNDLE";
7179       break;
7180     case MachO::MH_DSYM:
7181       outs() << "        DSYM";
7182       break;
7183     case MachO::MH_KEXT_BUNDLE:
7184       outs() << "  KEXTBUNDLE";
7185       break;
7186     default:
7187       outs() << format("  %10u", filetype);
7188       break;
7189     }
7190     outs() << format(" %5u", ncmds);
7191     outs() << format(" %10u", sizeofcmds);
7192     uint32_t f = flags;
7193     if (f & MachO::MH_NOUNDEFS) {
7194       outs() << "   NOUNDEFS";
7195       f &= ~MachO::MH_NOUNDEFS;
7196     }
7197     if (f & MachO::MH_INCRLINK) {
7198       outs() << " INCRLINK";
7199       f &= ~MachO::MH_INCRLINK;
7200     }
7201     if (f & MachO::MH_DYLDLINK) {
7202       outs() << " DYLDLINK";
7203       f &= ~MachO::MH_DYLDLINK;
7204     }
7205     if (f & MachO::MH_BINDATLOAD) {
7206       outs() << " BINDATLOAD";
7207       f &= ~MachO::MH_BINDATLOAD;
7208     }
7209     if (f & MachO::MH_PREBOUND) {
7210       outs() << " PREBOUND";
7211       f &= ~MachO::MH_PREBOUND;
7212     }
7213     if (f & MachO::MH_SPLIT_SEGS) {
7214       outs() << " SPLIT_SEGS";
7215       f &= ~MachO::MH_SPLIT_SEGS;
7216     }
7217     if (f & MachO::MH_LAZY_INIT) {
7218       outs() << " LAZY_INIT";
7219       f &= ~MachO::MH_LAZY_INIT;
7220     }
7221     if (f & MachO::MH_TWOLEVEL) {
7222       outs() << " TWOLEVEL";
7223       f &= ~MachO::MH_TWOLEVEL;
7224     }
7225     if (f & MachO::MH_FORCE_FLAT) {
7226       outs() << " FORCE_FLAT";
7227       f &= ~MachO::MH_FORCE_FLAT;
7228     }
7229     if (f & MachO::MH_NOMULTIDEFS) {
7230       outs() << " NOMULTIDEFS";
7231       f &= ~MachO::MH_NOMULTIDEFS;
7232     }
7233     if (f & MachO::MH_NOFIXPREBINDING) {
7234       outs() << " NOFIXPREBINDING";
7235       f &= ~MachO::MH_NOFIXPREBINDING;
7236     }
7237     if (f & MachO::MH_PREBINDABLE) {
7238       outs() << " PREBINDABLE";
7239       f &= ~MachO::MH_PREBINDABLE;
7240     }
7241     if (f & MachO::MH_ALLMODSBOUND) {
7242       outs() << " ALLMODSBOUND";
7243       f &= ~MachO::MH_ALLMODSBOUND;
7244     }
7245     if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
7246       outs() << " SUBSECTIONS_VIA_SYMBOLS";
7247       f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
7248     }
7249     if (f & MachO::MH_CANONICAL) {
7250       outs() << " CANONICAL";
7251       f &= ~MachO::MH_CANONICAL;
7252     }
7253     if (f & MachO::MH_WEAK_DEFINES) {
7254       outs() << " WEAK_DEFINES";
7255       f &= ~MachO::MH_WEAK_DEFINES;
7256     }
7257     if (f & MachO::MH_BINDS_TO_WEAK) {
7258       outs() << " BINDS_TO_WEAK";
7259       f &= ~MachO::MH_BINDS_TO_WEAK;
7260     }
7261     if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
7262       outs() << " ALLOW_STACK_EXECUTION";
7263       f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
7264     }
7265     if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
7266       outs() << " DEAD_STRIPPABLE_DYLIB";
7267       f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
7268     }
7269     if (f & MachO::MH_PIE) {
7270       outs() << " PIE";
7271       f &= ~MachO::MH_PIE;
7272     }
7273     if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
7274       outs() << " NO_REEXPORTED_DYLIBS";
7275       f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
7276     }
7277     if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
7278       outs() << " MH_HAS_TLV_DESCRIPTORS";
7279       f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
7280     }
7281     if (f & MachO::MH_NO_HEAP_EXECUTION) {
7282       outs() << " MH_NO_HEAP_EXECUTION";
7283       f &= ~MachO::MH_NO_HEAP_EXECUTION;
7284     }
7285     if (f & MachO::MH_APP_EXTENSION_SAFE) {
7286       outs() << " APP_EXTENSION_SAFE";
7287       f &= ~MachO::MH_APP_EXTENSION_SAFE;
7288     }
7289     if (f != 0 || flags == 0)
7290       outs() << format(" 0x%08" PRIx32, f);
7291   } else {
7292     outs() << format(" 0x%08" PRIx32, magic);
7293     outs() << format(" %7d", cputype);
7294     outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
7295     outs() << format("  0x%02" PRIx32,
7296                      (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
7297     outs() << format("  %10u", filetype);
7298     outs() << format(" %5u", ncmds);
7299     outs() << format(" %10u", sizeofcmds);
7300     outs() << format(" 0x%08" PRIx32, flags);
7301   }
7302   outs() << "\n";
7303 }
7304 
PrintSegmentCommand(uint32_t cmd,uint32_t cmdsize,StringRef SegName,uint64_t vmaddr,uint64_t vmsize,uint64_t fileoff,uint64_t filesize,uint32_t maxprot,uint32_t initprot,uint32_t nsects,uint32_t flags,uint32_t object_size,bool verbose)7305 static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
7306                                 StringRef SegName, uint64_t vmaddr,
7307                                 uint64_t vmsize, uint64_t fileoff,
7308                                 uint64_t filesize, uint32_t maxprot,
7309                                 uint32_t initprot, uint32_t nsects,
7310                                 uint32_t flags, uint32_t object_size,
7311                                 bool verbose) {
7312   uint64_t expected_cmdsize;
7313   if (cmd == MachO::LC_SEGMENT) {
7314     outs() << "      cmd LC_SEGMENT\n";
7315     expected_cmdsize = nsects;
7316     expected_cmdsize *= sizeof(struct MachO::section);
7317     expected_cmdsize += sizeof(struct MachO::segment_command);
7318   } else {
7319     outs() << "      cmd LC_SEGMENT_64\n";
7320     expected_cmdsize = nsects;
7321     expected_cmdsize *= sizeof(struct MachO::section_64);
7322     expected_cmdsize += sizeof(struct MachO::segment_command_64);
7323   }
7324   outs() << "  cmdsize " << cmdsize;
7325   if (cmdsize != expected_cmdsize)
7326     outs() << " Inconsistent size\n";
7327   else
7328     outs() << "\n";
7329   outs() << "  segname " << SegName << "\n";
7330   if (cmd == MachO::LC_SEGMENT_64) {
7331     outs() << "   vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n";
7332     outs() << "   vmsize " << format("0x%016" PRIx64, vmsize) << "\n";
7333   } else {
7334     outs() << "   vmaddr " << format("0x%08" PRIx64, vmaddr) << "\n";
7335     outs() << "   vmsize " << format("0x%08" PRIx64, vmsize) << "\n";
7336   }
7337   outs() << "  fileoff " << fileoff;
7338   if (fileoff > object_size)
7339     outs() << " (past end of file)\n";
7340   else
7341     outs() << "\n";
7342   outs() << " filesize " << filesize;
7343   if (fileoff + filesize > object_size)
7344     outs() << " (past end of file)\n";
7345   else
7346     outs() << "\n";
7347   if (verbose) {
7348     if ((maxprot &
7349          ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
7350            MachO::VM_PROT_EXECUTE)) != 0)
7351       outs() << "  maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n";
7352     else {
7353       outs() << "  maxprot ";
7354       outs() << ((maxprot & MachO::VM_PROT_READ) ? "r" : "-");
7355       outs() << ((maxprot & MachO::VM_PROT_WRITE) ? "w" : "-");
7356       outs() << ((maxprot & MachO::VM_PROT_EXECUTE) ? "x\n" : "-\n");
7357     }
7358     if ((initprot &
7359          ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
7360            MachO::VM_PROT_EXECUTE)) != 0)
7361       outs() << "  initprot ?" << format("0x%08" PRIx32, initprot) << "\n";
7362     else {
7363       outs() << "  initprot ";
7364       outs() << ((initprot & MachO::VM_PROT_READ) ? "r" : "-");
7365       outs() << ((initprot & MachO::VM_PROT_WRITE) ? "w" : "-");
7366       outs() << ((initprot & MachO::VM_PROT_EXECUTE) ? "x\n" : "-\n");
7367     }
7368   } else {
7369     outs() << "  maxprot " << format("0x%08" PRIx32, maxprot) << "\n";
7370     outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n";
7371   }
7372   outs() << "   nsects " << nsects << "\n";
7373   if (verbose) {
7374     outs() << "    flags";
7375     if (flags == 0)
7376       outs() << " (none)\n";
7377     else {
7378       if (flags & MachO::SG_HIGHVM) {
7379         outs() << " HIGHVM";
7380         flags &= ~MachO::SG_HIGHVM;
7381       }
7382       if (flags & MachO::SG_FVMLIB) {
7383         outs() << " FVMLIB";
7384         flags &= ~MachO::SG_FVMLIB;
7385       }
7386       if (flags & MachO::SG_NORELOC) {
7387         outs() << " NORELOC";
7388         flags &= ~MachO::SG_NORELOC;
7389       }
7390       if (flags & MachO::SG_PROTECTED_VERSION_1) {
7391         outs() << " PROTECTED_VERSION_1";
7392         flags &= ~MachO::SG_PROTECTED_VERSION_1;
7393       }
7394       if (flags)
7395         outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n";
7396       else
7397         outs() << "\n";
7398     }
7399   } else {
7400     outs() << "    flags " << format("0x%" PRIx32, flags) << "\n";
7401   }
7402 }
7403 
PrintSection(const char * sectname,const char * segname,uint64_t addr,uint64_t size,uint32_t offset,uint32_t align,uint32_t reloff,uint32_t nreloc,uint32_t flags,uint32_t reserved1,uint32_t reserved2,uint32_t cmd,const char * sg_segname,uint32_t filetype,uint32_t object_size,bool verbose)7404 static void PrintSection(const char *sectname, const char *segname,
7405                          uint64_t addr, uint64_t size, uint32_t offset,
7406                          uint32_t align, uint32_t reloff, uint32_t nreloc,
7407                          uint32_t flags, uint32_t reserved1, uint32_t reserved2,
7408                          uint32_t cmd, const char *sg_segname,
7409                          uint32_t filetype, uint32_t object_size,
7410                          bool verbose) {
7411   outs() << "Section\n";
7412   outs() << "  sectname " << format("%.16s\n", sectname);
7413   outs() << "   segname " << format("%.16s", segname);
7414   if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
7415     outs() << " (does not match segment)\n";
7416   else
7417     outs() << "\n";
7418   if (cmd == MachO::LC_SEGMENT_64) {
7419     outs() << "      addr " << format("0x%016" PRIx64, addr) << "\n";
7420     outs() << "      size " << format("0x%016" PRIx64, size);
7421   } else {
7422     outs() << "      addr " << format("0x%08" PRIx64, addr) << "\n";
7423     outs() << "      size " << format("0x%08" PRIx64, size);
7424   }
7425   if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
7426     outs() << " (past end of file)\n";
7427   else
7428     outs() << "\n";
7429   outs() << "    offset " << offset;
7430   if (offset > object_size)
7431     outs() << " (past end of file)\n";
7432   else
7433     outs() << "\n";
7434   uint32_t align_shifted = 1 << align;
7435   outs() << "     align 2^" << align << " (" << align_shifted << ")\n";
7436   outs() << "    reloff " << reloff;
7437   if (reloff > object_size)
7438     outs() << " (past end of file)\n";
7439   else
7440     outs() << "\n";
7441   outs() << "    nreloc " << nreloc;
7442   if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
7443     outs() << " (past end of file)\n";
7444   else
7445     outs() << "\n";
7446   uint32_t section_type = flags & MachO::SECTION_TYPE;
7447   if (verbose) {
7448     outs() << "      type";
7449     if (section_type == MachO::S_REGULAR)
7450       outs() << " S_REGULAR\n";
7451     else if (section_type == MachO::S_ZEROFILL)
7452       outs() << " S_ZEROFILL\n";
7453     else if (section_type == MachO::S_CSTRING_LITERALS)
7454       outs() << " S_CSTRING_LITERALS\n";
7455     else if (section_type == MachO::S_4BYTE_LITERALS)
7456       outs() << " S_4BYTE_LITERALS\n";
7457     else if (section_type == MachO::S_8BYTE_LITERALS)
7458       outs() << " S_8BYTE_LITERALS\n";
7459     else if (section_type == MachO::S_16BYTE_LITERALS)
7460       outs() << " S_16BYTE_LITERALS\n";
7461     else if (section_type == MachO::S_LITERAL_POINTERS)
7462       outs() << " S_LITERAL_POINTERS\n";
7463     else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
7464       outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
7465     else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
7466       outs() << " S_LAZY_SYMBOL_POINTERS\n";
7467     else if (section_type == MachO::S_SYMBOL_STUBS)
7468       outs() << " S_SYMBOL_STUBS\n";
7469     else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
7470       outs() << " S_MOD_INIT_FUNC_POINTERS\n";
7471     else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
7472       outs() << " S_MOD_TERM_FUNC_POINTERS\n";
7473     else if (section_type == MachO::S_COALESCED)
7474       outs() << " S_COALESCED\n";
7475     else if (section_type == MachO::S_INTERPOSING)
7476       outs() << " S_INTERPOSING\n";
7477     else if (section_type == MachO::S_DTRACE_DOF)
7478       outs() << " S_DTRACE_DOF\n";
7479     else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
7480       outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
7481     else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
7482       outs() << " S_THREAD_LOCAL_REGULAR\n";
7483     else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
7484       outs() << " S_THREAD_LOCAL_ZEROFILL\n";
7485     else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
7486       outs() << " S_THREAD_LOCAL_VARIABLES\n";
7487     else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
7488       outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
7489     else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
7490       outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
7491     else
7492       outs() << format("0x%08" PRIx32, section_type) << "\n";
7493     outs() << "attributes";
7494     uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
7495     if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
7496       outs() << " PURE_INSTRUCTIONS";
7497     if (section_attributes & MachO::S_ATTR_NO_TOC)
7498       outs() << " NO_TOC";
7499     if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
7500       outs() << " STRIP_STATIC_SYMS";
7501     if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
7502       outs() << " NO_DEAD_STRIP";
7503     if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
7504       outs() << " LIVE_SUPPORT";
7505     if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
7506       outs() << " SELF_MODIFYING_CODE";
7507     if (section_attributes & MachO::S_ATTR_DEBUG)
7508       outs() << " DEBUG";
7509     if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
7510       outs() << " SOME_INSTRUCTIONS";
7511     if (section_attributes & MachO::S_ATTR_EXT_RELOC)
7512       outs() << " EXT_RELOC";
7513     if (section_attributes & MachO::S_ATTR_LOC_RELOC)
7514       outs() << " LOC_RELOC";
7515     if (section_attributes == 0)
7516       outs() << " (none)";
7517     outs() << "\n";
7518   } else
7519     outs() << "     flags " << format("0x%08" PRIx32, flags) << "\n";
7520   outs() << " reserved1 " << reserved1;
7521   if (section_type == MachO::S_SYMBOL_STUBS ||
7522       section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
7523       section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
7524       section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
7525       section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
7526     outs() << " (index into indirect symbol table)\n";
7527   else
7528     outs() << "\n";
7529   outs() << " reserved2 " << reserved2;
7530   if (section_type == MachO::S_SYMBOL_STUBS)
7531     outs() << " (size of stubs)\n";
7532   else
7533     outs() << "\n";
7534 }
7535 
PrintSymtabLoadCommand(MachO::symtab_command st,bool Is64Bit,uint32_t object_size)7536 static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
7537                                    uint32_t object_size) {
7538   outs() << "     cmd LC_SYMTAB\n";
7539   outs() << " cmdsize " << st.cmdsize;
7540   if (st.cmdsize != sizeof(struct MachO::symtab_command))
7541     outs() << " Incorrect size\n";
7542   else
7543     outs() << "\n";
7544   outs() << "  symoff " << st.symoff;
7545   if (st.symoff > object_size)
7546     outs() << " (past end of file)\n";
7547   else
7548     outs() << "\n";
7549   outs() << "   nsyms " << st.nsyms;
7550   uint64_t big_size;
7551   if (Is64Bit) {
7552     big_size = st.nsyms;
7553     big_size *= sizeof(struct MachO::nlist_64);
7554     big_size += st.symoff;
7555     if (big_size > object_size)
7556       outs() << " (past end of file)\n";
7557     else
7558       outs() << "\n";
7559   } else {
7560     big_size = st.nsyms;
7561     big_size *= sizeof(struct MachO::nlist);
7562     big_size += st.symoff;
7563     if (big_size > object_size)
7564       outs() << " (past end of file)\n";
7565     else
7566       outs() << "\n";
7567   }
7568   outs() << "  stroff " << st.stroff;
7569   if (st.stroff > object_size)
7570     outs() << " (past end of file)\n";
7571   else
7572     outs() << "\n";
7573   outs() << " strsize " << st.strsize;
7574   big_size = st.stroff;
7575   big_size += st.strsize;
7576   if (big_size > object_size)
7577     outs() << " (past end of file)\n";
7578   else
7579     outs() << "\n";
7580 }
7581 
PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,uint32_t nsyms,uint32_t object_size,bool Is64Bit)7582 static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
7583                                      uint32_t nsyms, uint32_t object_size,
7584                                      bool Is64Bit) {
7585   outs() << "            cmd LC_DYSYMTAB\n";
7586   outs() << "        cmdsize " << dyst.cmdsize;
7587   if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
7588     outs() << " Incorrect size\n";
7589   else
7590     outs() << "\n";
7591   outs() << "      ilocalsym " << dyst.ilocalsym;
7592   if (dyst.ilocalsym > nsyms)
7593     outs() << " (greater than the number of symbols)\n";
7594   else
7595     outs() << "\n";
7596   outs() << "      nlocalsym " << dyst.nlocalsym;
7597   uint64_t big_size;
7598   big_size = dyst.ilocalsym;
7599   big_size += dyst.nlocalsym;
7600   if (big_size > nsyms)
7601     outs() << " (past the end of the symbol table)\n";
7602   else
7603     outs() << "\n";
7604   outs() << "     iextdefsym " << dyst.iextdefsym;
7605   if (dyst.iextdefsym > nsyms)
7606     outs() << " (greater than the number of symbols)\n";
7607   else
7608     outs() << "\n";
7609   outs() << "     nextdefsym " << dyst.nextdefsym;
7610   big_size = dyst.iextdefsym;
7611   big_size += dyst.nextdefsym;
7612   if (big_size > nsyms)
7613     outs() << " (past the end of the symbol table)\n";
7614   else
7615     outs() << "\n";
7616   outs() << "      iundefsym " << dyst.iundefsym;
7617   if (dyst.iundefsym > nsyms)
7618     outs() << " (greater than the number of symbols)\n";
7619   else
7620     outs() << "\n";
7621   outs() << "      nundefsym " << dyst.nundefsym;
7622   big_size = dyst.iundefsym;
7623   big_size += dyst.nundefsym;
7624   if (big_size > nsyms)
7625     outs() << " (past the end of the symbol table)\n";
7626   else
7627     outs() << "\n";
7628   outs() << "         tocoff " << dyst.tocoff;
7629   if (dyst.tocoff > object_size)
7630     outs() << " (past end of file)\n";
7631   else
7632     outs() << "\n";
7633   outs() << "           ntoc " << dyst.ntoc;
7634   big_size = dyst.ntoc;
7635   big_size *= sizeof(struct MachO::dylib_table_of_contents);
7636   big_size += dyst.tocoff;
7637   if (big_size > object_size)
7638     outs() << " (past end of file)\n";
7639   else
7640     outs() << "\n";
7641   outs() << "      modtaboff " << dyst.modtaboff;
7642   if (dyst.modtaboff > object_size)
7643     outs() << " (past end of file)\n";
7644   else
7645     outs() << "\n";
7646   outs() << "        nmodtab " << dyst.nmodtab;
7647   uint64_t modtabend;
7648   if (Is64Bit) {
7649     modtabend = dyst.nmodtab;
7650     modtabend *= sizeof(struct MachO::dylib_module_64);
7651     modtabend += dyst.modtaboff;
7652   } else {
7653     modtabend = dyst.nmodtab;
7654     modtabend *= sizeof(struct MachO::dylib_module);
7655     modtabend += dyst.modtaboff;
7656   }
7657   if (modtabend > object_size)
7658     outs() << " (past end of file)\n";
7659   else
7660     outs() << "\n";
7661   outs() << "   extrefsymoff " << dyst.extrefsymoff;
7662   if (dyst.extrefsymoff > object_size)
7663     outs() << " (past end of file)\n";
7664   else
7665     outs() << "\n";
7666   outs() << "    nextrefsyms " << dyst.nextrefsyms;
7667   big_size = dyst.nextrefsyms;
7668   big_size *= sizeof(struct MachO::dylib_reference);
7669   big_size += dyst.extrefsymoff;
7670   if (big_size > object_size)
7671     outs() << " (past end of file)\n";
7672   else
7673     outs() << "\n";
7674   outs() << " indirectsymoff " << dyst.indirectsymoff;
7675   if (dyst.indirectsymoff > object_size)
7676     outs() << " (past end of file)\n";
7677   else
7678     outs() << "\n";
7679   outs() << "  nindirectsyms " << dyst.nindirectsyms;
7680   big_size = dyst.nindirectsyms;
7681   big_size *= sizeof(uint32_t);
7682   big_size += dyst.indirectsymoff;
7683   if (big_size > object_size)
7684     outs() << " (past end of file)\n";
7685   else
7686     outs() << "\n";
7687   outs() << "      extreloff " << dyst.extreloff;
7688   if (dyst.extreloff > object_size)
7689     outs() << " (past end of file)\n";
7690   else
7691     outs() << "\n";
7692   outs() << "        nextrel " << dyst.nextrel;
7693   big_size = dyst.nextrel;
7694   big_size *= sizeof(struct MachO::relocation_info);
7695   big_size += dyst.extreloff;
7696   if (big_size > object_size)
7697     outs() << " (past end of file)\n";
7698   else
7699     outs() << "\n";
7700   outs() << "      locreloff " << dyst.locreloff;
7701   if (dyst.locreloff > object_size)
7702     outs() << " (past end of file)\n";
7703   else
7704     outs() << "\n";
7705   outs() << "        nlocrel " << dyst.nlocrel;
7706   big_size = dyst.nlocrel;
7707   big_size *= sizeof(struct MachO::relocation_info);
7708   big_size += dyst.locreloff;
7709   if (big_size > object_size)
7710     outs() << " (past end of file)\n";
7711   else
7712     outs() << "\n";
7713 }
7714 
PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,uint32_t object_size)7715 static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
7716                                      uint32_t object_size) {
7717   if (dc.cmd == MachO::LC_DYLD_INFO)
7718     outs() << "            cmd LC_DYLD_INFO\n";
7719   else
7720     outs() << "            cmd LC_DYLD_INFO_ONLY\n";
7721   outs() << "        cmdsize " << dc.cmdsize;
7722   if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
7723     outs() << " Incorrect size\n";
7724   else
7725     outs() << "\n";
7726   outs() << "     rebase_off " << dc.rebase_off;
7727   if (dc.rebase_off > object_size)
7728     outs() << " (past end of file)\n";
7729   else
7730     outs() << "\n";
7731   outs() << "    rebase_size " << dc.rebase_size;
7732   uint64_t big_size;
7733   big_size = dc.rebase_off;
7734   big_size += dc.rebase_size;
7735   if (big_size > object_size)
7736     outs() << " (past end of file)\n";
7737   else
7738     outs() << "\n";
7739   outs() << "       bind_off " << dc.bind_off;
7740   if (dc.bind_off > object_size)
7741     outs() << " (past end of file)\n";
7742   else
7743     outs() << "\n";
7744   outs() << "      bind_size " << dc.bind_size;
7745   big_size = dc.bind_off;
7746   big_size += dc.bind_size;
7747   if (big_size > object_size)
7748     outs() << " (past end of file)\n";
7749   else
7750     outs() << "\n";
7751   outs() << "  weak_bind_off " << dc.weak_bind_off;
7752   if (dc.weak_bind_off > object_size)
7753     outs() << " (past end of file)\n";
7754   else
7755     outs() << "\n";
7756   outs() << " weak_bind_size " << dc.weak_bind_size;
7757   big_size = dc.weak_bind_off;
7758   big_size += dc.weak_bind_size;
7759   if (big_size > object_size)
7760     outs() << " (past end of file)\n";
7761   else
7762     outs() << "\n";
7763   outs() << "  lazy_bind_off " << dc.lazy_bind_off;
7764   if (dc.lazy_bind_off > object_size)
7765     outs() << " (past end of file)\n";
7766   else
7767     outs() << "\n";
7768   outs() << " lazy_bind_size " << dc.lazy_bind_size;
7769   big_size = dc.lazy_bind_off;
7770   big_size += dc.lazy_bind_size;
7771   if (big_size > object_size)
7772     outs() << " (past end of file)\n";
7773   else
7774     outs() << "\n";
7775   outs() << "     export_off " << dc.export_off;
7776   if (dc.export_off > object_size)
7777     outs() << " (past end of file)\n";
7778   else
7779     outs() << "\n";
7780   outs() << "    export_size " << dc.export_size;
7781   big_size = dc.export_off;
7782   big_size += dc.export_size;
7783   if (big_size > object_size)
7784     outs() << " (past end of file)\n";
7785   else
7786     outs() << "\n";
7787 }
7788 
PrintDyldLoadCommand(MachO::dylinker_command dyld,const char * Ptr)7789 static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
7790                                  const char *Ptr) {
7791   if (dyld.cmd == MachO::LC_ID_DYLINKER)
7792     outs() << "          cmd LC_ID_DYLINKER\n";
7793   else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
7794     outs() << "          cmd LC_LOAD_DYLINKER\n";
7795   else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
7796     outs() << "          cmd LC_DYLD_ENVIRONMENT\n";
7797   else
7798     outs() << "          cmd ?(" << dyld.cmd << ")\n";
7799   outs() << "      cmdsize " << dyld.cmdsize;
7800   if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
7801     outs() << " Incorrect size\n";
7802   else
7803     outs() << "\n";
7804   if (dyld.name >= dyld.cmdsize)
7805     outs() << "         name ?(bad offset " << dyld.name << ")\n";
7806   else {
7807     const char *P = (const char *)(Ptr) + dyld.name;
7808     outs() << "         name " << P << " (offset " << dyld.name << ")\n";
7809   }
7810 }
7811 
PrintUuidLoadCommand(MachO::uuid_command uuid)7812 static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
7813   outs() << "     cmd LC_UUID\n";
7814   outs() << " cmdsize " << uuid.cmdsize;
7815   if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
7816     outs() << " Incorrect size\n";
7817   else
7818     outs() << "\n";
7819   outs() << "    uuid ";
7820   for (int i = 0; i < 16; ++i) {
7821     outs() << format("%02" PRIX32, uuid.uuid[i]);
7822     if (i == 3 || i == 5 || i == 7 || i == 9)
7823       outs() << "-";
7824   }
7825   outs() << "\n";
7826 }
7827 
PrintRpathLoadCommand(MachO::rpath_command rpath,const char * Ptr)7828 static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) {
7829   outs() << "          cmd LC_RPATH\n";
7830   outs() << "      cmdsize " << rpath.cmdsize;
7831   if (rpath.cmdsize < sizeof(struct MachO::rpath_command))
7832     outs() << " Incorrect size\n";
7833   else
7834     outs() << "\n";
7835   if (rpath.path >= rpath.cmdsize)
7836     outs() << "         path ?(bad offset " << rpath.path << ")\n";
7837   else {
7838     const char *P = (const char *)(Ptr) + rpath.path;
7839     outs() << "         path " << P << " (offset " << rpath.path << ")\n";
7840   }
7841 }
7842 
PrintVersionMinLoadCommand(MachO::version_min_command vd)7843 static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
7844   StringRef LoadCmdName;
7845   switch (vd.cmd) {
7846   case MachO::LC_VERSION_MIN_MACOSX:
7847     LoadCmdName = "LC_VERSION_MIN_MACOSX";
7848     break;
7849   case MachO::LC_VERSION_MIN_IPHONEOS:
7850     LoadCmdName = "LC_VERSION_MIN_IPHONEOS";
7851     break;
7852   case MachO::LC_VERSION_MIN_TVOS:
7853     LoadCmdName = "LC_VERSION_MIN_TVOS";
7854     break;
7855   case MachO::LC_VERSION_MIN_WATCHOS:
7856     LoadCmdName = "LC_VERSION_MIN_WATCHOS";
7857     break;
7858   default:
7859     llvm_unreachable("Unknown version min load command");
7860   }
7861 
7862   outs() << "      cmd " << LoadCmdName << '\n';
7863   outs() << "  cmdsize " << vd.cmdsize;
7864   if (vd.cmdsize != sizeof(struct MachO::version_min_command))
7865     outs() << " Incorrect size\n";
7866   else
7867     outs() << "\n";
7868   outs() << "  version "
7869          << MachOObjectFile::getVersionMinMajor(vd, false) << "."
7870          << MachOObjectFile::getVersionMinMinor(vd, false);
7871   uint32_t Update = MachOObjectFile::getVersionMinUpdate(vd, false);
7872   if (Update != 0)
7873     outs() << "." << Update;
7874   outs() << "\n";
7875   if (vd.sdk == 0)
7876     outs() << "      sdk n/a";
7877   else {
7878     outs() << "      sdk "
7879            << MachOObjectFile::getVersionMinMajor(vd, true) << "."
7880            << MachOObjectFile::getVersionMinMinor(vd, true);
7881   }
7882   Update = MachOObjectFile::getVersionMinUpdate(vd, true);
7883   if (Update != 0)
7884     outs() << "." << Update;
7885   outs() << "\n";
7886 }
7887 
PrintSourceVersionCommand(MachO::source_version_command sd)7888 static void PrintSourceVersionCommand(MachO::source_version_command sd) {
7889   outs() << "      cmd LC_SOURCE_VERSION\n";
7890   outs() << "  cmdsize " << sd.cmdsize;
7891   if (sd.cmdsize != sizeof(struct MachO::source_version_command))
7892     outs() << " Incorrect size\n";
7893   else
7894     outs() << "\n";
7895   uint64_t a = (sd.version >> 40) & 0xffffff;
7896   uint64_t b = (sd.version >> 30) & 0x3ff;
7897   uint64_t c = (sd.version >> 20) & 0x3ff;
7898   uint64_t d = (sd.version >> 10) & 0x3ff;
7899   uint64_t e = sd.version & 0x3ff;
7900   outs() << "  version " << a << "." << b;
7901   if (e != 0)
7902     outs() << "." << c << "." << d << "." << e;
7903   else if (d != 0)
7904     outs() << "." << c << "." << d;
7905   else if (c != 0)
7906     outs() << "." << c;
7907   outs() << "\n";
7908 }
7909 
PrintEntryPointCommand(MachO::entry_point_command ep)7910 static void PrintEntryPointCommand(MachO::entry_point_command ep) {
7911   outs() << "       cmd LC_MAIN\n";
7912   outs() << "   cmdsize " << ep.cmdsize;
7913   if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
7914     outs() << " Incorrect size\n";
7915   else
7916     outs() << "\n";
7917   outs() << "  entryoff " << ep.entryoff << "\n";
7918   outs() << " stacksize " << ep.stacksize << "\n";
7919 }
7920 
PrintEncryptionInfoCommand(MachO::encryption_info_command ec,uint32_t object_size)7921 static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec,
7922                                        uint32_t object_size) {
7923   outs() << "          cmd LC_ENCRYPTION_INFO\n";
7924   outs() << "      cmdsize " << ec.cmdsize;
7925   if (ec.cmdsize != sizeof(struct MachO::encryption_info_command))
7926     outs() << " Incorrect size\n";
7927   else
7928     outs() << "\n";
7929   outs() << "     cryptoff " << ec.cryptoff;
7930   if (ec.cryptoff > object_size)
7931     outs() << " (past end of file)\n";
7932   else
7933     outs() << "\n";
7934   outs() << "    cryptsize " << ec.cryptsize;
7935   if (ec.cryptsize > object_size)
7936     outs() << " (past end of file)\n";
7937   else
7938     outs() << "\n";
7939   outs() << "      cryptid " << ec.cryptid << "\n";
7940 }
7941 
PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,uint32_t object_size)7942 static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,
7943                                          uint32_t object_size) {
7944   outs() << "          cmd LC_ENCRYPTION_INFO_64\n";
7945   outs() << "      cmdsize " << ec.cmdsize;
7946   if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64))
7947     outs() << " Incorrect size\n";
7948   else
7949     outs() << "\n";
7950   outs() << "     cryptoff " << ec.cryptoff;
7951   if (ec.cryptoff > object_size)
7952     outs() << " (past end of file)\n";
7953   else
7954     outs() << "\n";
7955   outs() << "    cryptsize " << ec.cryptsize;
7956   if (ec.cryptsize > object_size)
7957     outs() << " (past end of file)\n";
7958   else
7959     outs() << "\n";
7960   outs() << "      cryptid " << ec.cryptid << "\n";
7961   outs() << "          pad " << ec.pad << "\n";
7962 }
7963 
PrintLinkerOptionCommand(MachO::linker_option_command lo,const char * Ptr)7964 static void PrintLinkerOptionCommand(MachO::linker_option_command lo,
7965                                      const char *Ptr) {
7966   outs() << "     cmd LC_LINKER_OPTION\n";
7967   outs() << " cmdsize " << lo.cmdsize;
7968   if (lo.cmdsize < sizeof(struct MachO::linker_option_command))
7969     outs() << " Incorrect size\n";
7970   else
7971     outs() << "\n";
7972   outs() << "   count " << lo.count << "\n";
7973   const char *string = Ptr + sizeof(struct MachO::linker_option_command);
7974   uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command);
7975   uint32_t i = 0;
7976   while (left > 0) {
7977     while (*string == '\0' && left > 0) {
7978       string++;
7979       left--;
7980     }
7981     if (left > 0) {
7982       i++;
7983       outs() << "  string #" << i << " " << format("%.*s\n", left, string);
7984       uint32_t NullPos = StringRef(string, left).find('\0');
7985       uint32_t len = std::min(NullPos, left) + 1;
7986       string += len;
7987       left -= len;
7988     }
7989   }
7990   if (lo.count != i)
7991     outs() << "   count " << lo.count << " does not match number of strings "
7992            << i << "\n";
7993 }
7994 
PrintSubFrameworkCommand(MachO::sub_framework_command sub,const char * Ptr)7995 static void PrintSubFrameworkCommand(MachO::sub_framework_command sub,
7996                                      const char *Ptr) {
7997   outs() << "          cmd LC_SUB_FRAMEWORK\n";
7998   outs() << "      cmdsize " << sub.cmdsize;
7999   if (sub.cmdsize < sizeof(struct MachO::sub_framework_command))
8000     outs() << " Incorrect size\n";
8001   else
8002     outs() << "\n";
8003   if (sub.umbrella < sub.cmdsize) {
8004     const char *P = Ptr + sub.umbrella;
8005     outs() << "     umbrella " << P << " (offset " << sub.umbrella << ")\n";
8006   } else {
8007     outs() << "     umbrella ?(bad offset " << sub.umbrella << ")\n";
8008   }
8009 }
8010 
PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,const char * Ptr)8011 static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,
8012                                     const char *Ptr) {
8013   outs() << "          cmd LC_SUB_UMBRELLA\n";
8014   outs() << "      cmdsize " << sub.cmdsize;
8015   if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command))
8016     outs() << " Incorrect size\n";
8017   else
8018     outs() << "\n";
8019   if (sub.sub_umbrella < sub.cmdsize) {
8020     const char *P = Ptr + sub.sub_umbrella;
8021     outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n";
8022   } else {
8023     outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n";
8024   }
8025 }
8026 
PrintSubLibraryCommand(MachO::sub_library_command sub,const char * Ptr)8027 static void PrintSubLibraryCommand(MachO::sub_library_command sub,
8028                                    const char *Ptr) {
8029   outs() << "          cmd LC_SUB_LIBRARY\n";
8030   outs() << "      cmdsize " << sub.cmdsize;
8031   if (sub.cmdsize < sizeof(struct MachO::sub_library_command))
8032     outs() << " Incorrect size\n";
8033   else
8034     outs() << "\n";
8035   if (sub.sub_library < sub.cmdsize) {
8036     const char *P = Ptr + sub.sub_library;
8037     outs() << "  sub_library " << P << " (offset " << sub.sub_library << ")\n";
8038   } else {
8039     outs() << "  sub_library ?(bad offset " << sub.sub_library << ")\n";
8040   }
8041 }
8042 
PrintSubClientCommand(MachO::sub_client_command sub,const char * Ptr)8043 static void PrintSubClientCommand(MachO::sub_client_command sub,
8044                                   const char *Ptr) {
8045   outs() << "          cmd LC_SUB_CLIENT\n";
8046   outs() << "      cmdsize " << sub.cmdsize;
8047   if (sub.cmdsize < sizeof(struct MachO::sub_client_command))
8048     outs() << " Incorrect size\n";
8049   else
8050     outs() << "\n";
8051   if (sub.client < sub.cmdsize) {
8052     const char *P = Ptr + sub.client;
8053     outs() << "       client " << P << " (offset " << sub.client << ")\n";
8054   } else {
8055     outs() << "       client ?(bad offset " << sub.client << ")\n";
8056   }
8057 }
8058 
PrintRoutinesCommand(MachO::routines_command r)8059 static void PrintRoutinesCommand(MachO::routines_command r) {
8060   outs() << "          cmd LC_ROUTINES\n";
8061   outs() << "      cmdsize " << r.cmdsize;
8062   if (r.cmdsize != sizeof(struct MachO::routines_command))
8063     outs() << " Incorrect size\n";
8064   else
8065     outs() << "\n";
8066   outs() << " init_address " << format("0x%08" PRIx32, r.init_address) << "\n";
8067   outs() << "  init_module " << r.init_module << "\n";
8068   outs() << "    reserved1 " << r.reserved1 << "\n";
8069   outs() << "    reserved2 " << r.reserved2 << "\n";
8070   outs() << "    reserved3 " << r.reserved3 << "\n";
8071   outs() << "    reserved4 " << r.reserved4 << "\n";
8072   outs() << "    reserved5 " << r.reserved5 << "\n";
8073   outs() << "    reserved6 " << r.reserved6 << "\n";
8074 }
8075 
PrintRoutinesCommand64(MachO::routines_command_64 r)8076 static void PrintRoutinesCommand64(MachO::routines_command_64 r) {
8077   outs() << "          cmd LC_ROUTINES_64\n";
8078   outs() << "      cmdsize " << r.cmdsize;
8079   if (r.cmdsize != sizeof(struct MachO::routines_command_64))
8080     outs() << " Incorrect size\n";
8081   else
8082     outs() << "\n";
8083   outs() << " init_address " << format("0x%016" PRIx64, r.init_address) << "\n";
8084   outs() << "  init_module " << r.init_module << "\n";
8085   outs() << "    reserved1 " << r.reserved1 << "\n";
8086   outs() << "    reserved2 " << r.reserved2 << "\n";
8087   outs() << "    reserved3 " << r.reserved3 << "\n";
8088   outs() << "    reserved4 " << r.reserved4 << "\n";
8089   outs() << "    reserved5 " << r.reserved5 << "\n";
8090   outs() << "    reserved6 " << r.reserved6 << "\n";
8091 }
8092 
Print_x86_thread_state64_t(MachO::x86_thread_state64_t & cpu64)8093 static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) {
8094   outs() << "   rax  " << format("0x%016" PRIx64, cpu64.rax);
8095   outs() << " rbx " << format("0x%016" PRIx64, cpu64.rbx);
8096   outs() << " rcx  " << format("0x%016" PRIx64, cpu64.rcx) << "\n";
8097   outs() << "   rdx  " << format("0x%016" PRIx64, cpu64.rdx);
8098   outs() << " rdi " << format("0x%016" PRIx64, cpu64.rdi);
8099   outs() << " rsi  " << format("0x%016" PRIx64, cpu64.rsi) << "\n";
8100   outs() << "   rbp  " << format("0x%016" PRIx64, cpu64.rbp);
8101   outs() << " rsp " << format("0x%016" PRIx64, cpu64.rsp);
8102   outs() << " r8   " << format("0x%016" PRIx64, cpu64.r8) << "\n";
8103   outs() << "    r9  " << format("0x%016" PRIx64, cpu64.r9);
8104   outs() << " r10 " << format("0x%016" PRIx64, cpu64.r10);
8105   outs() << " r11  " << format("0x%016" PRIx64, cpu64.r11) << "\n";
8106   outs() << "   r12  " << format("0x%016" PRIx64, cpu64.r12);
8107   outs() << " r13 " << format("0x%016" PRIx64, cpu64.r13);
8108   outs() << " r14  " << format("0x%016" PRIx64, cpu64.r14) << "\n";
8109   outs() << "   r15  " << format("0x%016" PRIx64, cpu64.r15);
8110   outs() << " rip " << format("0x%016" PRIx64, cpu64.rip) << "\n";
8111   outs() << "rflags  " << format("0x%016" PRIx64, cpu64.rflags);
8112   outs() << " cs  " << format("0x%016" PRIx64, cpu64.cs);
8113   outs() << " fs   " << format("0x%016" PRIx64, cpu64.fs) << "\n";
8114   outs() << "    gs  " << format("0x%016" PRIx64, cpu64.gs) << "\n";
8115 }
8116 
Print_mmst_reg(MachO::mmst_reg_t & r)8117 static void Print_mmst_reg(MachO::mmst_reg_t &r) {
8118   uint32_t f;
8119   outs() << "\t      mmst_reg  ";
8120   for (f = 0; f < 10; f++)
8121     outs() << format("%02" PRIx32, (r.mmst_reg[f] & 0xff)) << " ";
8122   outs() << "\n";
8123   outs() << "\t      mmst_rsrv ";
8124   for (f = 0; f < 6; f++)
8125     outs() << format("%02" PRIx32, (r.mmst_rsrv[f] & 0xff)) << " ";
8126   outs() << "\n";
8127 }
8128 
Print_xmm_reg(MachO::xmm_reg_t & r)8129 static void Print_xmm_reg(MachO::xmm_reg_t &r) {
8130   uint32_t f;
8131   outs() << "\t      xmm_reg ";
8132   for (f = 0; f < 16; f++)
8133     outs() << format("%02" PRIx32, (r.xmm_reg[f] & 0xff)) << " ";
8134   outs() << "\n";
8135 }
8136 
Print_x86_float_state_t(MachO::x86_float_state64_t & fpu)8137 static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) {
8138   outs() << "\t    fpu_reserved[0] " << fpu.fpu_reserved[0];
8139   outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n";
8140   outs() << "\t    control: invalid " << fpu.fpu_fcw.invalid;
8141   outs() << " denorm " << fpu.fpu_fcw.denorm;
8142   outs() << " zdiv " << fpu.fpu_fcw.zdiv;
8143   outs() << " ovrfl " << fpu.fpu_fcw.ovrfl;
8144   outs() << " undfl " << fpu.fpu_fcw.undfl;
8145   outs() << " precis " << fpu.fpu_fcw.precis << "\n";
8146   outs() << "\t\t     pc ";
8147   if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B)
8148     outs() << "FP_PREC_24B ";
8149   else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B)
8150     outs() << "FP_PREC_53B ";
8151   else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B)
8152     outs() << "FP_PREC_64B ";
8153   else
8154     outs() << fpu.fpu_fcw.pc << " ";
8155   outs() << "rc ";
8156   if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR)
8157     outs() << "FP_RND_NEAR ";
8158   else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN)
8159     outs() << "FP_RND_DOWN ";
8160   else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP)
8161     outs() << "FP_RND_UP ";
8162   else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP)
8163     outs() << "FP_CHOP ";
8164   outs() << "\n";
8165   outs() << "\t    status: invalid " << fpu.fpu_fsw.invalid;
8166   outs() << " denorm " << fpu.fpu_fsw.denorm;
8167   outs() << " zdiv " << fpu.fpu_fsw.zdiv;
8168   outs() << " ovrfl " << fpu.fpu_fsw.ovrfl;
8169   outs() << " undfl " << fpu.fpu_fsw.undfl;
8170   outs() << " precis " << fpu.fpu_fsw.precis;
8171   outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n";
8172   outs() << "\t            errsumm " << fpu.fpu_fsw.errsumm;
8173   outs() << " c0 " << fpu.fpu_fsw.c0;
8174   outs() << " c1 " << fpu.fpu_fsw.c1;
8175   outs() << " c2 " << fpu.fpu_fsw.c2;
8176   outs() << " tos " << fpu.fpu_fsw.tos;
8177   outs() << " c3 " << fpu.fpu_fsw.c3;
8178   outs() << " busy " << fpu.fpu_fsw.busy << "\n";
8179   outs() << "\t    fpu_ftw " << format("0x%02" PRIx32, fpu.fpu_ftw);
8180   outs() << " fpu_rsrv1 " << format("0x%02" PRIx32, fpu.fpu_rsrv1);
8181   outs() << " fpu_fop " << format("0x%04" PRIx32, fpu.fpu_fop);
8182   outs() << " fpu_ip " << format("0x%08" PRIx32, fpu.fpu_ip) << "\n";
8183   outs() << "\t    fpu_cs " << format("0x%04" PRIx32, fpu.fpu_cs);
8184   outs() << " fpu_rsrv2 " << format("0x%04" PRIx32, fpu.fpu_rsrv2);
8185   outs() << " fpu_dp " << format("0x%08" PRIx32, fpu.fpu_dp);
8186   outs() << " fpu_ds " << format("0x%04" PRIx32, fpu.fpu_ds) << "\n";
8187   outs() << "\t    fpu_rsrv3 " << format("0x%04" PRIx32, fpu.fpu_rsrv3);
8188   outs() << " fpu_mxcsr " << format("0x%08" PRIx32, fpu.fpu_mxcsr);
8189   outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32, fpu.fpu_mxcsrmask);
8190   outs() << "\n";
8191   outs() << "\t    fpu_stmm0:\n";
8192   Print_mmst_reg(fpu.fpu_stmm0);
8193   outs() << "\t    fpu_stmm1:\n";
8194   Print_mmst_reg(fpu.fpu_stmm1);
8195   outs() << "\t    fpu_stmm2:\n";
8196   Print_mmst_reg(fpu.fpu_stmm2);
8197   outs() << "\t    fpu_stmm3:\n";
8198   Print_mmst_reg(fpu.fpu_stmm3);
8199   outs() << "\t    fpu_stmm4:\n";
8200   Print_mmst_reg(fpu.fpu_stmm4);
8201   outs() << "\t    fpu_stmm5:\n";
8202   Print_mmst_reg(fpu.fpu_stmm5);
8203   outs() << "\t    fpu_stmm6:\n";
8204   Print_mmst_reg(fpu.fpu_stmm6);
8205   outs() << "\t    fpu_stmm7:\n";
8206   Print_mmst_reg(fpu.fpu_stmm7);
8207   outs() << "\t    fpu_xmm0:\n";
8208   Print_xmm_reg(fpu.fpu_xmm0);
8209   outs() << "\t    fpu_xmm1:\n";
8210   Print_xmm_reg(fpu.fpu_xmm1);
8211   outs() << "\t    fpu_xmm2:\n";
8212   Print_xmm_reg(fpu.fpu_xmm2);
8213   outs() << "\t    fpu_xmm3:\n";
8214   Print_xmm_reg(fpu.fpu_xmm3);
8215   outs() << "\t    fpu_xmm4:\n";
8216   Print_xmm_reg(fpu.fpu_xmm4);
8217   outs() << "\t    fpu_xmm5:\n";
8218   Print_xmm_reg(fpu.fpu_xmm5);
8219   outs() << "\t    fpu_xmm6:\n";
8220   Print_xmm_reg(fpu.fpu_xmm6);
8221   outs() << "\t    fpu_xmm7:\n";
8222   Print_xmm_reg(fpu.fpu_xmm7);
8223   outs() << "\t    fpu_xmm8:\n";
8224   Print_xmm_reg(fpu.fpu_xmm8);
8225   outs() << "\t    fpu_xmm9:\n";
8226   Print_xmm_reg(fpu.fpu_xmm9);
8227   outs() << "\t    fpu_xmm10:\n";
8228   Print_xmm_reg(fpu.fpu_xmm10);
8229   outs() << "\t    fpu_xmm11:\n";
8230   Print_xmm_reg(fpu.fpu_xmm11);
8231   outs() << "\t    fpu_xmm12:\n";
8232   Print_xmm_reg(fpu.fpu_xmm12);
8233   outs() << "\t    fpu_xmm13:\n";
8234   Print_xmm_reg(fpu.fpu_xmm13);
8235   outs() << "\t    fpu_xmm14:\n";
8236   Print_xmm_reg(fpu.fpu_xmm14);
8237   outs() << "\t    fpu_xmm15:\n";
8238   Print_xmm_reg(fpu.fpu_xmm15);
8239   outs() << "\t    fpu_rsrv4:\n";
8240   for (uint32_t f = 0; f < 6; f++) {
8241     outs() << "\t            ";
8242     for (uint32_t g = 0; g < 16; g++)
8243       outs() << format("%02" PRIx32, fpu.fpu_rsrv4[f * g]) << " ";
8244     outs() << "\n";
8245   }
8246   outs() << "\t    fpu_reserved1 " << format("0x%08" PRIx32, fpu.fpu_reserved1);
8247   outs() << "\n";
8248 }
8249 
Print_x86_exception_state_t(MachO::x86_exception_state64_t & exc64)8250 static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) {
8251   outs() << "\t    trapno " << format("0x%08" PRIx32, exc64.trapno);
8252   outs() << " err " << format("0x%08" PRIx32, exc64.err);
8253   outs() << " faultvaddr " << format("0x%016" PRIx64, exc64.faultvaddr) << "\n";
8254 }
8255 
PrintThreadCommand(MachO::thread_command t,const char * Ptr,bool isLittleEndian,uint32_t cputype)8256 static void PrintThreadCommand(MachO::thread_command t, const char *Ptr,
8257                                bool isLittleEndian, uint32_t cputype) {
8258   if (t.cmd == MachO::LC_THREAD)
8259     outs() << "        cmd LC_THREAD\n";
8260   else if (t.cmd == MachO::LC_UNIXTHREAD)
8261     outs() << "        cmd LC_UNIXTHREAD\n";
8262   else
8263     outs() << "        cmd " << t.cmd << " (unknown)\n";
8264   outs() << "    cmdsize " << t.cmdsize;
8265   if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t))
8266     outs() << " Incorrect size\n";
8267   else
8268     outs() << "\n";
8269 
8270   const char *begin = Ptr + sizeof(struct MachO::thread_command);
8271   const char *end = Ptr + t.cmdsize;
8272   uint32_t flavor, count, left;
8273   if (cputype == MachO::CPU_TYPE_X86_64) {
8274     while (begin < end) {
8275       if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
8276         memcpy((char *)&flavor, begin, sizeof(uint32_t));
8277         begin += sizeof(uint32_t);
8278       } else {
8279         flavor = 0;
8280         begin = end;
8281       }
8282       if (isLittleEndian != sys::IsLittleEndianHost)
8283         sys::swapByteOrder(flavor);
8284       if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
8285         memcpy((char *)&count, begin, sizeof(uint32_t));
8286         begin += sizeof(uint32_t);
8287       } else {
8288         count = 0;
8289         begin = end;
8290       }
8291       if (isLittleEndian != sys::IsLittleEndianHost)
8292         sys::swapByteOrder(count);
8293       if (flavor == MachO::x86_THREAD_STATE64) {
8294         outs() << "     flavor x86_THREAD_STATE64\n";
8295         if (count == MachO::x86_THREAD_STATE64_COUNT)
8296           outs() << "      count x86_THREAD_STATE64_COUNT\n";
8297         else
8298           outs() << "      count " << count
8299                  << " (not x86_THREAD_STATE64_COUNT)\n";
8300         MachO::x86_thread_state64_t cpu64;
8301         left = end - begin;
8302         if (left >= sizeof(MachO::x86_thread_state64_t)) {
8303           memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t));
8304           begin += sizeof(MachO::x86_thread_state64_t);
8305         } else {
8306           memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t));
8307           memcpy(&cpu64, begin, left);
8308           begin += left;
8309         }
8310         if (isLittleEndian != sys::IsLittleEndianHost)
8311           swapStruct(cpu64);
8312         Print_x86_thread_state64_t(cpu64);
8313       } else if (flavor == MachO::x86_THREAD_STATE) {
8314         outs() << "     flavor x86_THREAD_STATE\n";
8315         if (count == MachO::x86_THREAD_STATE_COUNT)
8316           outs() << "      count x86_THREAD_STATE_COUNT\n";
8317         else
8318           outs() << "      count " << count
8319                  << " (not x86_THREAD_STATE_COUNT)\n";
8320         struct MachO::x86_thread_state_t ts;
8321         left = end - begin;
8322         if (left >= sizeof(MachO::x86_thread_state_t)) {
8323           memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t));
8324           begin += sizeof(MachO::x86_thread_state_t);
8325         } else {
8326           memset(&ts, '\0', sizeof(MachO::x86_thread_state_t));
8327           memcpy(&ts, begin, left);
8328           begin += left;
8329         }
8330         if (isLittleEndian != sys::IsLittleEndianHost)
8331           swapStruct(ts);
8332         if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) {
8333           outs() << "\t    tsh.flavor x86_THREAD_STATE64 ";
8334           if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT)
8335             outs() << "tsh.count x86_THREAD_STATE64_COUNT\n";
8336           else
8337             outs() << "tsh.count " << ts.tsh.count
8338                    << " (not x86_THREAD_STATE64_COUNT\n";
8339           Print_x86_thread_state64_t(ts.uts.ts64);
8340         } else {
8341           outs() << "\t    tsh.flavor " << ts.tsh.flavor << "  tsh.count "
8342                  << ts.tsh.count << "\n";
8343         }
8344       } else if (flavor == MachO::x86_FLOAT_STATE) {
8345         outs() << "     flavor x86_FLOAT_STATE\n";
8346         if (count == MachO::x86_FLOAT_STATE_COUNT)
8347           outs() << "      count x86_FLOAT_STATE_COUNT\n";
8348         else
8349           outs() << "      count " << count << " (not x86_FLOAT_STATE_COUNT)\n";
8350         struct MachO::x86_float_state_t fs;
8351         left = end - begin;
8352         if (left >= sizeof(MachO::x86_float_state_t)) {
8353           memcpy(&fs, begin, sizeof(MachO::x86_float_state_t));
8354           begin += sizeof(MachO::x86_float_state_t);
8355         } else {
8356           memset(&fs, '\0', sizeof(MachO::x86_float_state_t));
8357           memcpy(&fs, begin, left);
8358           begin += left;
8359         }
8360         if (isLittleEndian != sys::IsLittleEndianHost)
8361           swapStruct(fs);
8362         if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) {
8363           outs() << "\t    fsh.flavor x86_FLOAT_STATE64 ";
8364           if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT)
8365             outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n";
8366           else
8367             outs() << "fsh.count " << fs.fsh.count
8368                    << " (not x86_FLOAT_STATE64_COUNT\n";
8369           Print_x86_float_state_t(fs.ufs.fs64);
8370         } else {
8371           outs() << "\t    fsh.flavor " << fs.fsh.flavor << "  fsh.count "
8372                  << fs.fsh.count << "\n";
8373         }
8374       } else if (flavor == MachO::x86_EXCEPTION_STATE) {
8375         outs() << "     flavor x86_EXCEPTION_STATE\n";
8376         if (count == MachO::x86_EXCEPTION_STATE_COUNT)
8377           outs() << "      count x86_EXCEPTION_STATE_COUNT\n";
8378         else
8379           outs() << "      count " << count
8380                  << " (not x86_EXCEPTION_STATE_COUNT)\n";
8381         struct MachO::x86_exception_state_t es;
8382         left = end - begin;
8383         if (left >= sizeof(MachO::x86_exception_state_t)) {
8384           memcpy(&es, begin, sizeof(MachO::x86_exception_state_t));
8385           begin += sizeof(MachO::x86_exception_state_t);
8386         } else {
8387           memset(&es, '\0', sizeof(MachO::x86_exception_state_t));
8388           memcpy(&es, begin, left);
8389           begin += left;
8390         }
8391         if (isLittleEndian != sys::IsLittleEndianHost)
8392           swapStruct(es);
8393         if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) {
8394           outs() << "\t    esh.flavor x86_EXCEPTION_STATE64\n";
8395           if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT)
8396             outs() << "\t    esh.count x86_EXCEPTION_STATE64_COUNT\n";
8397           else
8398             outs() << "\t    esh.count " << es.esh.count
8399                    << " (not x86_EXCEPTION_STATE64_COUNT\n";
8400           Print_x86_exception_state_t(es.ues.es64);
8401         } else {
8402           outs() << "\t    esh.flavor " << es.esh.flavor << "  esh.count "
8403                  << es.esh.count << "\n";
8404         }
8405       } else {
8406         outs() << "     flavor " << flavor << " (unknown)\n";
8407         outs() << "      count " << count << "\n";
8408         outs() << "      state (unknown)\n";
8409         begin += count * sizeof(uint32_t);
8410       }
8411     }
8412   } else {
8413     while (begin < end) {
8414       if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
8415         memcpy((char *)&flavor, begin, sizeof(uint32_t));
8416         begin += sizeof(uint32_t);
8417       } else {
8418         flavor = 0;
8419         begin = end;
8420       }
8421       if (isLittleEndian != sys::IsLittleEndianHost)
8422         sys::swapByteOrder(flavor);
8423       if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
8424         memcpy((char *)&count, begin, sizeof(uint32_t));
8425         begin += sizeof(uint32_t);
8426       } else {
8427         count = 0;
8428         begin = end;
8429       }
8430       if (isLittleEndian != sys::IsLittleEndianHost)
8431         sys::swapByteOrder(count);
8432       outs() << "     flavor " << flavor << "\n";
8433       outs() << "      count " << count << "\n";
8434       outs() << "      state (Unknown cputype/cpusubtype)\n";
8435       begin += count * sizeof(uint32_t);
8436     }
8437   }
8438 }
8439 
PrintDylibCommand(MachO::dylib_command dl,const char * Ptr)8440 static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
8441   if (dl.cmd == MachO::LC_ID_DYLIB)
8442     outs() << "          cmd LC_ID_DYLIB\n";
8443   else if (dl.cmd == MachO::LC_LOAD_DYLIB)
8444     outs() << "          cmd LC_LOAD_DYLIB\n";
8445   else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
8446     outs() << "          cmd LC_LOAD_WEAK_DYLIB\n";
8447   else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
8448     outs() << "          cmd LC_REEXPORT_DYLIB\n";
8449   else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
8450     outs() << "          cmd LC_LAZY_LOAD_DYLIB\n";
8451   else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
8452     outs() << "          cmd LC_LOAD_UPWARD_DYLIB\n";
8453   else
8454     outs() << "          cmd " << dl.cmd << " (unknown)\n";
8455   outs() << "      cmdsize " << dl.cmdsize;
8456   if (dl.cmdsize < sizeof(struct MachO::dylib_command))
8457     outs() << " Incorrect size\n";
8458   else
8459     outs() << "\n";
8460   if (dl.dylib.name < dl.cmdsize) {
8461     const char *P = (const char *)(Ptr) + dl.dylib.name;
8462     outs() << "         name " << P << " (offset " << dl.dylib.name << ")\n";
8463   } else {
8464     outs() << "         name ?(bad offset " << dl.dylib.name << ")\n";
8465   }
8466   outs() << "   time stamp " << dl.dylib.timestamp << " ";
8467   time_t t = dl.dylib.timestamp;
8468   outs() << ctime(&t);
8469   outs() << "      current version ";
8470   if (dl.dylib.current_version == 0xffffffff)
8471     outs() << "n/a\n";
8472   else
8473     outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
8474            << ((dl.dylib.current_version >> 8) & 0xff) << "."
8475            << (dl.dylib.current_version & 0xff) << "\n";
8476   outs() << "compatibility version ";
8477   if (dl.dylib.compatibility_version == 0xffffffff)
8478     outs() << "n/a\n";
8479   else
8480     outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
8481            << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
8482            << (dl.dylib.compatibility_version & 0xff) << "\n";
8483 }
8484 
PrintLinkEditDataCommand(MachO::linkedit_data_command ld,uint32_t object_size)8485 static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
8486                                      uint32_t object_size) {
8487   if (ld.cmd == MachO::LC_CODE_SIGNATURE)
8488     outs() << "      cmd LC_FUNCTION_STARTS\n";
8489   else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
8490     outs() << "      cmd LC_SEGMENT_SPLIT_INFO\n";
8491   else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
8492     outs() << "      cmd LC_FUNCTION_STARTS\n";
8493   else if (ld.cmd == MachO::LC_DATA_IN_CODE)
8494     outs() << "      cmd LC_DATA_IN_CODE\n";
8495   else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
8496     outs() << "      cmd LC_DYLIB_CODE_SIGN_DRS\n";
8497   else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
8498     outs() << "      cmd LC_LINKER_OPTIMIZATION_HINT\n";
8499   else
8500     outs() << "      cmd " << ld.cmd << " (?)\n";
8501   outs() << "  cmdsize " << ld.cmdsize;
8502   if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
8503     outs() << " Incorrect size\n";
8504   else
8505     outs() << "\n";
8506   outs() << "  dataoff " << ld.dataoff;
8507   if (ld.dataoff > object_size)
8508     outs() << " (past end of file)\n";
8509   else
8510     outs() << "\n";
8511   outs() << " datasize " << ld.datasize;
8512   uint64_t big_size = ld.dataoff;
8513   big_size += ld.datasize;
8514   if (big_size > object_size)
8515     outs() << " (past end of file)\n";
8516   else
8517     outs() << "\n";
8518 }
8519 
PrintLoadCommands(const MachOObjectFile * Obj,uint32_t filetype,uint32_t cputype,bool verbose)8520 static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t filetype,
8521                               uint32_t cputype, bool verbose) {
8522   StringRef Buf = Obj->getData();
8523   unsigned Index = 0;
8524   for (const auto &Command : Obj->load_commands()) {
8525     outs() << "Load command " << Index++ << "\n";
8526     if (Command.C.cmd == MachO::LC_SEGMENT) {
8527       MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
8528       const char *sg_segname = SLC.segname;
8529       PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
8530                           SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
8531                           SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
8532                           verbose);
8533       for (unsigned j = 0; j < SLC.nsects; j++) {
8534         MachO::section S = Obj->getSection(Command, j);
8535         PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
8536                      S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
8537                      SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
8538       }
8539     } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
8540       MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
8541       const char *sg_segname = SLC_64.segname;
8542       PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
8543                           SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
8544                           SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
8545                           SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
8546       for (unsigned j = 0; j < SLC_64.nsects; j++) {
8547         MachO::section_64 S_64 = Obj->getSection64(Command, j);
8548         PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
8549                      S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
8550                      S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
8551                      sg_segname, filetype, Buf.size(), verbose);
8552       }
8553     } else if (Command.C.cmd == MachO::LC_SYMTAB) {
8554       MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
8555       PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
8556     } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
8557       MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
8558       MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
8559       PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
8560                                Obj->is64Bit());
8561     } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
8562                Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
8563       MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
8564       PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
8565     } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
8566                Command.C.cmd == MachO::LC_ID_DYLINKER ||
8567                Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
8568       MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
8569       PrintDyldLoadCommand(Dyld, Command.Ptr);
8570     } else if (Command.C.cmd == MachO::LC_UUID) {
8571       MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
8572       PrintUuidLoadCommand(Uuid);
8573     } else if (Command.C.cmd == MachO::LC_RPATH) {
8574       MachO::rpath_command Rpath = Obj->getRpathCommand(Command);
8575       PrintRpathLoadCommand(Rpath, Command.Ptr);
8576     } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX ||
8577                Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS ||
8578                Command.C.cmd == MachO::LC_VERSION_MIN_TVOS ||
8579                Command.C.cmd == MachO::LC_VERSION_MIN_WATCHOS) {
8580       MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
8581       PrintVersionMinLoadCommand(Vd);
8582     } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
8583       MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
8584       PrintSourceVersionCommand(Sd);
8585     } else if (Command.C.cmd == MachO::LC_MAIN) {
8586       MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
8587       PrintEntryPointCommand(Ep);
8588     } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) {
8589       MachO::encryption_info_command Ei =
8590           Obj->getEncryptionInfoCommand(Command);
8591       PrintEncryptionInfoCommand(Ei, Buf.size());
8592     } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
8593       MachO::encryption_info_command_64 Ei =
8594           Obj->getEncryptionInfoCommand64(Command);
8595       PrintEncryptionInfoCommand64(Ei, Buf.size());
8596     } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) {
8597       MachO::linker_option_command Lo =
8598           Obj->getLinkerOptionLoadCommand(Command);
8599       PrintLinkerOptionCommand(Lo, Command.Ptr);
8600     } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) {
8601       MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command);
8602       PrintSubFrameworkCommand(Sf, Command.Ptr);
8603     } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) {
8604       MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command);
8605       PrintSubUmbrellaCommand(Sf, Command.Ptr);
8606     } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) {
8607       MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command);
8608       PrintSubLibraryCommand(Sl, Command.Ptr);
8609     } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) {
8610       MachO::sub_client_command Sc = Obj->getSubClientCommand(Command);
8611       PrintSubClientCommand(Sc, Command.Ptr);
8612     } else if (Command.C.cmd == MachO::LC_ROUTINES) {
8613       MachO::routines_command Rc = Obj->getRoutinesCommand(Command);
8614       PrintRoutinesCommand(Rc);
8615     } else if (Command.C.cmd == MachO::LC_ROUTINES_64) {
8616       MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command);
8617       PrintRoutinesCommand64(Rc);
8618     } else if (Command.C.cmd == MachO::LC_THREAD ||
8619                Command.C.cmd == MachO::LC_UNIXTHREAD) {
8620       MachO::thread_command Tc = Obj->getThreadCommand(Command);
8621       PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype);
8622     } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
8623                Command.C.cmd == MachO::LC_ID_DYLIB ||
8624                Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
8625                Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
8626                Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
8627                Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
8628       MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
8629       PrintDylibCommand(Dl, Command.Ptr);
8630     } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
8631                Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
8632                Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
8633                Command.C.cmd == MachO::LC_DATA_IN_CODE ||
8634                Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
8635                Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
8636       MachO::linkedit_data_command Ld =
8637           Obj->getLinkeditDataLoadCommand(Command);
8638       PrintLinkEditDataCommand(Ld, Buf.size());
8639     } else {
8640       outs() << "      cmd ?(" << format("0x%08" PRIx32, Command.C.cmd)
8641              << ")\n";
8642       outs() << "  cmdsize " << Command.C.cmdsize << "\n";
8643       // TODO: get and print the raw bytes of the load command.
8644     }
8645     // TODO: print all the other kinds of load commands.
8646   }
8647 }
8648 
getAndPrintMachHeader(const MachOObjectFile * Obj,uint32_t & filetype,uint32_t & cputype,bool verbose)8649 static void getAndPrintMachHeader(const MachOObjectFile *Obj,
8650                                   uint32_t &filetype, uint32_t &cputype,
8651                                   bool verbose) {
8652   if (Obj->is64Bit()) {
8653     MachO::mach_header_64 H_64;
8654     H_64 = Obj->getHeader64();
8655     PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
8656                     H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
8657     filetype = H_64.filetype;
8658     cputype = H_64.cputype;
8659   } else {
8660     MachO::mach_header H;
8661     H = Obj->getHeader();
8662     PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
8663                     H.sizeofcmds, H.flags, verbose);
8664     filetype = H.filetype;
8665     cputype = H.cputype;
8666   }
8667 }
8668 
printMachOFileHeader(const object::ObjectFile * Obj)8669 void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
8670   const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
8671   uint32_t filetype = 0;
8672   uint32_t cputype = 0;
8673   getAndPrintMachHeader(file, filetype, cputype, !NonVerbose);
8674   PrintLoadCommands(file, filetype, cputype, !NonVerbose);
8675 }
8676 
8677 //===----------------------------------------------------------------------===//
8678 // export trie dumping
8679 //===----------------------------------------------------------------------===//
8680 
printMachOExportsTrie(const object::MachOObjectFile * Obj)8681 void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
8682   for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
8683     uint64_t Flags = Entry.flags();
8684     bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
8685     bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
8686     bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
8687                         MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
8688     bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
8689                 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
8690     bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
8691     if (ReExport)
8692       outs() << "[re-export] ";
8693     else
8694       outs() << format("0x%08llX  ",
8695                        Entry.address()); // FIXME:add in base address
8696     outs() << Entry.name();
8697     if (WeakDef || ThreadLocal || Resolver || Abs) {
8698       bool NeedsComma = false;
8699       outs() << " [";
8700       if (WeakDef) {
8701         outs() << "weak_def";
8702         NeedsComma = true;
8703       }
8704       if (ThreadLocal) {
8705         if (NeedsComma)
8706           outs() << ", ";
8707         outs() << "per-thread";
8708         NeedsComma = true;
8709       }
8710       if (Abs) {
8711         if (NeedsComma)
8712           outs() << ", ";
8713         outs() << "absolute";
8714         NeedsComma = true;
8715       }
8716       if (Resolver) {
8717         if (NeedsComma)
8718           outs() << ", ";
8719         outs() << format("resolver=0x%08llX", Entry.other());
8720         NeedsComma = true;
8721       }
8722       outs() << "]";
8723     }
8724     if (ReExport) {
8725       StringRef DylibName = "unknown";
8726       int Ordinal = Entry.other() - 1;
8727       Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
8728       if (Entry.otherName().empty())
8729         outs() << " (from " << DylibName << ")";
8730       else
8731         outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
8732     }
8733     outs() << "\n";
8734   }
8735 }
8736 
8737 //===----------------------------------------------------------------------===//
8738 // rebase table dumping
8739 //===----------------------------------------------------------------------===//
8740 
8741 namespace {
8742 class SegInfo {
8743 public:
8744   SegInfo(const object::MachOObjectFile *Obj);
8745 
8746   StringRef segmentName(uint32_t SegIndex);
8747   StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
8748   uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
8749   bool isValidSegIndexAndOffset(uint32_t SegIndex, uint64_t SegOffset);
8750 
8751 private:
8752   struct SectionInfo {
8753     uint64_t Address;
8754     uint64_t Size;
8755     StringRef SectionName;
8756     StringRef SegmentName;
8757     uint64_t OffsetInSegment;
8758     uint64_t SegmentStartAddress;
8759     uint32_t SegmentIndex;
8760   };
8761   const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
8762   SmallVector<SectionInfo, 32> Sections;
8763 };
8764 }
8765 
SegInfo(const object::MachOObjectFile * Obj)8766 SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
8767   // Build table of sections so segIndex/offset pairs can be translated.
8768   uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
8769   StringRef CurSegName;
8770   uint64_t CurSegAddress;
8771   for (const SectionRef &Section : Obj->sections()) {
8772     SectionInfo Info;
8773     error(Section.getName(Info.SectionName));
8774     Info.Address = Section.getAddress();
8775     Info.Size = Section.getSize();
8776     Info.SegmentName =
8777         Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
8778     if (!Info.SegmentName.equals(CurSegName)) {
8779       ++CurSegIndex;
8780       CurSegName = Info.SegmentName;
8781       CurSegAddress = Info.Address;
8782     }
8783     Info.SegmentIndex = CurSegIndex - 1;
8784     Info.OffsetInSegment = Info.Address - CurSegAddress;
8785     Info.SegmentStartAddress = CurSegAddress;
8786     Sections.push_back(Info);
8787   }
8788 }
8789 
segmentName(uint32_t SegIndex)8790 StringRef SegInfo::segmentName(uint32_t SegIndex) {
8791   for (const SectionInfo &SI : Sections) {
8792     if (SI.SegmentIndex == SegIndex)
8793       return SI.SegmentName;
8794   }
8795   llvm_unreachable("invalid segIndex");
8796 }
8797 
isValidSegIndexAndOffset(uint32_t SegIndex,uint64_t OffsetInSeg)8798 bool SegInfo::isValidSegIndexAndOffset(uint32_t SegIndex,
8799                                        uint64_t OffsetInSeg) {
8800   for (const SectionInfo &SI : Sections) {
8801     if (SI.SegmentIndex != SegIndex)
8802       continue;
8803     if (SI.OffsetInSegment > OffsetInSeg)
8804       continue;
8805     if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
8806       continue;
8807     return true;
8808   }
8809   return false;
8810 }
8811 
findSection(uint32_t SegIndex,uint64_t OffsetInSeg)8812 const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
8813                                                  uint64_t OffsetInSeg) {
8814   for (const SectionInfo &SI : Sections) {
8815     if (SI.SegmentIndex != SegIndex)
8816       continue;
8817     if (SI.OffsetInSegment > OffsetInSeg)
8818       continue;
8819     if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
8820       continue;
8821     return SI;
8822   }
8823   llvm_unreachable("segIndex and offset not in any section");
8824 }
8825 
sectionName(uint32_t SegIndex,uint64_t OffsetInSeg)8826 StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
8827   return findSection(SegIndex, OffsetInSeg).SectionName;
8828 }
8829 
address(uint32_t SegIndex,uint64_t OffsetInSeg)8830 uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
8831   const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
8832   return SI.SegmentStartAddress + OffsetInSeg;
8833 }
8834 
printMachORebaseTable(const object::MachOObjectFile * Obj)8835 void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
8836   // Build table of sections so names can used in final output.
8837   SegInfo sectionTable(Obj);
8838 
8839   outs() << "segment  section            address     type\n";
8840   for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
8841     uint32_t SegIndex = Entry.segmentIndex();
8842     uint64_t OffsetInSeg = Entry.segmentOffset();
8843     StringRef SegmentName = sectionTable.segmentName(SegIndex);
8844     StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
8845     uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
8846 
8847     // Table lines look like: __DATA  __nl_symbol_ptr  0x0000F00C  pointer
8848     outs() << format("%-8s %-18s 0x%08" PRIX64 "  %s\n",
8849                      SegmentName.str().c_str(), SectionName.str().c_str(),
8850                      Address, Entry.typeName().str().c_str());
8851   }
8852 }
8853 
ordinalName(const object::MachOObjectFile * Obj,int Ordinal)8854 static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
8855   StringRef DylibName;
8856   switch (Ordinal) {
8857   case MachO::BIND_SPECIAL_DYLIB_SELF:
8858     return "this-image";
8859   case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
8860     return "main-executable";
8861   case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
8862     return "flat-namespace";
8863   default:
8864     if (Ordinal > 0) {
8865       std::error_code EC =
8866           Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
8867       if (EC)
8868         return "<<bad library ordinal>>";
8869       return DylibName;
8870     }
8871   }
8872   return "<<unknown special ordinal>>";
8873 }
8874 
8875 //===----------------------------------------------------------------------===//
8876 // bind table dumping
8877 //===----------------------------------------------------------------------===//
8878 
printMachOBindTable(const object::MachOObjectFile * Obj)8879 void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
8880   // Build table of sections so names can used in final output.
8881   SegInfo sectionTable(Obj);
8882 
8883   outs() << "segment  section            address    type       "
8884             "addend dylib            symbol\n";
8885   for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
8886     uint32_t SegIndex = Entry.segmentIndex();
8887     uint64_t OffsetInSeg = Entry.segmentOffset();
8888     StringRef SegmentName = sectionTable.segmentName(SegIndex);
8889     StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
8890     uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
8891 
8892     // Table lines look like:
8893     //  __DATA  __got  0x00012010    pointer   0 libSystem ___stack_chk_guard
8894     StringRef Attr;
8895     if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
8896       Attr = " (weak_import)";
8897     outs() << left_justify(SegmentName, 8) << " "
8898            << left_justify(SectionName, 18) << " "
8899            << format_hex(Address, 10, true) << " "
8900            << left_justify(Entry.typeName(), 8) << " "
8901            << format_decimal(Entry.addend(), 8) << " "
8902            << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
8903            << Entry.symbolName() << Attr << "\n";
8904   }
8905 }
8906 
8907 //===----------------------------------------------------------------------===//
8908 // lazy bind table dumping
8909 //===----------------------------------------------------------------------===//
8910 
printMachOLazyBindTable(const object::MachOObjectFile * Obj)8911 void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
8912   // Build table of sections so names can used in final output.
8913   SegInfo sectionTable(Obj);
8914 
8915   outs() << "segment  section            address     "
8916             "dylib            symbol\n";
8917   for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
8918     uint32_t SegIndex = Entry.segmentIndex();
8919     uint64_t OffsetInSeg = Entry.segmentOffset();
8920     StringRef SegmentName = sectionTable.segmentName(SegIndex);
8921     StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
8922     uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
8923 
8924     // Table lines look like:
8925     //  __DATA  __got  0x00012010 libSystem ___stack_chk_guard
8926     outs() << left_justify(SegmentName, 8) << " "
8927            << left_justify(SectionName, 18) << " "
8928            << format_hex(Address, 10, true) << " "
8929            << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
8930            << Entry.symbolName() << "\n";
8931   }
8932 }
8933 
8934 //===----------------------------------------------------------------------===//
8935 // weak bind table dumping
8936 //===----------------------------------------------------------------------===//
8937 
printMachOWeakBindTable(const object::MachOObjectFile * Obj)8938 void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
8939   // Build table of sections so names can used in final output.
8940   SegInfo sectionTable(Obj);
8941 
8942   outs() << "segment  section            address     "
8943             "type       addend   symbol\n";
8944   for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
8945     // Strong symbols don't have a location to update.
8946     if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
8947       outs() << "                                        strong              "
8948              << Entry.symbolName() << "\n";
8949       continue;
8950     }
8951     uint32_t SegIndex = Entry.segmentIndex();
8952     uint64_t OffsetInSeg = Entry.segmentOffset();
8953     StringRef SegmentName = sectionTable.segmentName(SegIndex);
8954     StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
8955     uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
8956 
8957     // Table lines look like:
8958     // __DATA  __data  0x00001000  pointer    0   _foo
8959     outs() << left_justify(SegmentName, 8) << " "
8960            << left_justify(SectionName, 18) << " "
8961            << format_hex(Address, 10, true) << " "
8962            << left_justify(Entry.typeName(), 8) << " "
8963            << format_decimal(Entry.addend(), 8) << "   " << Entry.symbolName()
8964            << "\n";
8965   }
8966 }
8967 
8968 // get_dyld_bind_info_symbolname() is used for disassembly and passed an
8969 // address, ReferenceValue, in the Mach-O file and looks in the dyld bind
8970 // information for that address. If the address is found its binding symbol
8971 // name is returned.  If not nullptr is returned.
get_dyld_bind_info_symbolname(uint64_t ReferenceValue,struct DisassembleInfo * info)8972 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
8973                                                  struct DisassembleInfo *info) {
8974   if (info->bindtable == nullptr) {
8975     info->bindtable = new (BindTable);
8976     SegInfo sectionTable(info->O);
8977     for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
8978       uint32_t SegIndex = Entry.segmentIndex();
8979       uint64_t OffsetInSeg = Entry.segmentOffset();
8980       if (!sectionTable.isValidSegIndexAndOffset(SegIndex, OffsetInSeg))
8981         continue;
8982       uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
8983       const char *SymbolName = nullptr;
8984       StringRef name = Entry.symbolName();
8985       if (!name.empty())
8986         SymbolName = name.data();
8987       info->bindtable->push_back(std::make_pair(Address, SymbolName));
8988     }
8989   }
8990   for (bind_table_iterator BI = info->bindtable->begin(),
8991                            BE = info->bindtable->end();
8992        BI != BE; ++BI) {
8993     uint64_t Address = BI->first;
8994     if (ReferenceValue == Address) {
8995       const char *SymbolName = BI->second;
8996       return SymbolName;
8997     }
8998   }
8999   return nullptr;
9000 }
9001