Bug Summary

File:tools/llvm-objdump/MachODump.cpp
Location:line 1987, column 25
Description:Function call argument is an uninitialized value

Annotated Source Code

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/STLExtras.h"
17#include "llvm/ADT/StringExtras.h"
18#include "llvm/ADT/Triple.h"
19#include "llvm/Config/config.h"
20#include "llvm/DebugInfo/DWARF/DIContext.h"
21#include "llvm/MC/MCAsmInfo.h"
22#include "llvm/MC/MCContext.h"
23#include "llvm/MC/MCDisassembler.h"
24#include "llvm/MC/MCInst.h"
25#include "llvm/MC/MCInstPrinter.h"
26#include "llvm/MC/MCInstrDesc.h"
27#include "llvm/MC/MCInstrInfo.h"
28#include "llvm/MC/MCRegisterInfo.h"
29#include "llvm/MC/MCSubtargetInfo.h"
30#include "llvm/Object/MachO.h"
31#include "llvm/Object/MachOUniversal.h"
32#include "llvm/Support/Casting.h"
33#include "llvm/Support/CommandLine.h"
34#include "llvm/Support/Debug.h"
35#include "llvm/Support/Endian.h"
36#include "llvm/Support/Format.h"
37#include "llvm/Support/FormattedStream.h"
38#include "llvm/Support/GraphWriter.h"
39#include "llvm/Support/LEB128.h"
40#include "llvm/Support/MachO.h"
41#include "llvm/Support/MemoryBuffer.h"
42#include "llvm/Support/TargetRegistry.h"
43#include "llvm/Support/TargetSelect.h"
44#include "llvm/Support/raw_ostream.h"
45#include <algorithm>
46#include <cstring>
47#include <system_error>
48
49#if HAVE_CXXABI_H1
50#include <cxxabi.h>
51#endif
52
53using namespace llvm;
54using namespace object;
55
56static cl::opt<bool>
57 UseDbg("g",
58 cl::desc("Print line information from debug info if available"));
59
60static cl::opt<std::string> DSYMFile("dsym",
61 cl::desc("Use .dSYM file for debug info"));
62
63static cl::opt<bool> FullLeadingAddr("full-leading-addr",
64 cl::desc("Print full leading address"));
65
66static cl::opt<bool> NoLeadingAddr("no-leading-addr",
67 cl::desc("Print no leading address"));
68
69static cl::opt<bool>
70 PrintImmHex("print-imm-hex",
71 cl::desc("Use hex format for immediate values"));
72
73cl::opt<bool> llvm::UniversalHeaders("universal-headers",
74 cl::desc("Print Mach-O universal headers "
75 "(requires -macho)"));
76
77cl::opt<bool>
78 llvm::ArchiveHeaders("archive-headers",
79 cl::desc("Print archive headers for Mach-O archives "
80 "(requires -macho)"));
81
82cl::opt<bool>
83 llvm::IndirectSymbols("indirect-symbols",
84 cl::desc("Print indirect symbol table for Mach-O "
85 "objects (requires -macho)"));
86
87cl::opt<bool>
88 llvm::DataInCode("data-in-code",
89 cl::desc("Print the data in code table for Mach-O objects "
90 "(requires -macho)"));
91
92cl::opt<bool>
93 llvm::LinkOptHints("link-opt-hints",
94 cl::desc("Print the linker optimization hints for "
95 "Mach-O objects (requires -macho)"));
96
97cl::list<std::string>
98 llvm::DumpSections("section",
99 cl::desc("Prints the specified segment,section for "
100 "Mach-O objects (requires -macho)"));
101
102cl::opt<bool>
103 llvm::InfoPlist("info-plist",
104 cl::desc("Print the info plist section as strings for "
105 "Mach-O objects (requires -macho)"));
106
107cl::opt<bool>
108 llvm::DylibsUsed("dylibs-used",
109 cl::desc("Print the shared libraries used for linked "
110 "Mach-O files (requires -macho)"));
111
112cl::opt<bool>
113 llvm::DylibId("dylib-id",
114 cl::desc("Print the shared library's id for the dylib Mach-O "
115 "file (requires -macho)"));
116
117cl::opt<bool>
118 llvm::NonVerbose("non-verbose",
119 cl::desc("Print the info for Mach-O objects in "
120 "non-verbose or numeric form (requires -macho)"));
121
122cl::opt<std::string> llvm::DisSymName(
123 "dis-symname",
124 cl::desc("disassemble just this symbol's instructions (requires -macho"));
125
126static cl::opt<bool> NoSymbolicOperands(
127 "no-symbolic-operands",
128 cl::desc("do not symbolic operands when disassembling (requires -macho)"));
129
130
131static cl::list<std::string>
132 ArchFlags("arch", cl::desc("architecture(s) from a Mach-O file to dump"),
133 cl::ZeroOrMore);
134bool ArchAll = false;
135
136static std::string ThumbTripleName;
137
138static const Target *GetTarget(const MachOObjectFile *MachOObj,
139 const char **McpuDefault,
140 const Target **ThumbTarget) {
141 // Figure out the target triple.
142 if (TripleName.empty()) {
143 llvm::Triple TT("unknown-unknown-unknown");
144 llvm::Triple ThumbTriple = Triple();
145 TT = MachOObj->getArch(McpuDefault, &ThumbTriple);
146 TripleName = TT.str();
147 ThumbTripleName = ThumbTriple.str();
148 }
149
150 // Get the target specific parser.
151 std::string Error;
152 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
153 if (TheTarget && ThumbTripleName.empty())
154 return TheTarget;
155
156 *ThumbTarget = TargetRegistry::lookupTarget(ThumbTripleName, Error);
157 if (*ThumbTarget)
158 return TheTarget;
159
160 errs() << "llvm-objdump: error: unable to get target for '";
161 if (!TheTarget)
162 errs() << TripleName;
163 else
164 errs() << ThumbTripleName;
165 errs() << "', see --version and --triple.\n";
166 return nullptr;
167}
168
169struct SymbolSorter {
170 bool operator()(const SymbolRef &A, const SymbolRef &B) {
171 SymbolRef::Type AType, BType;
172 A.getType(AType);
173 B.getType(BType);
174
175 uint64_t AAddr, BAddr;
176 if (AType != SymbolRef::ST_Function)
177 AAddr = 0;
178 else
179 A.getAddress(AAddr);
180 if (BType != SymbolRef::ST_Function)
181 BAddr = 0;
182 else
183 B.getAddress(BAddr);
184 return AAddr < BAddr;
185 }
186};
187
188// Types for the storted data in code table that is built before disassembly
189// and the predicate function to sort them.
190typedef std::pair<uint64_t, DiceRef> DiceTableEntry;
191typedef std::vector<DiceTableEntry> DiceTable;
192typedef DiceTable::iterator dice_table_iterator;
193
194// This is used to search for a data in code table entry for the PC being
195// disassembled. The j parameter has the PC in j.first. A single data in code
196// table entry can cover many bytes for each of its Kind's. So if the offset,
197// aka the i.first value, of the data in code table entry plus its Length
198// covers the PC being searched for this will return true. If not it will
199// return false.
200static bool compareDiceTableEntries(const DiceTableEntry &i,
201 const DiceTableEntry &j) {
202 uint16_t Length;
203 i.second.getLength(Length);
204
205 return j.first >= i.first && j.first < i.first + Length;
206}
207
208static uint64_t DumpDataInCode(const uint8_t *bytes, uint64_t Length,
209 unsigned short Kind) {
210 uint32_t Value, Size = 1;
211
212 switch (Kind) {
213 default:
214 case MachO::DICE_KIND_DATA:
215 if (Length >= 4) {
216 if (!NoShowRawInsn)
217 DumpBytes(ArrayRef<uint8_t>(bytes, 4));
218 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
219 outs() << "\t.long " << Value;
220 Size = 4;
221 } else if (Length >= 2) {
222 if (!NoShowRawInsn)
223 DumpBytes(ArrayRef<uint8_t>(bytes, 2));
224 Value = bytes[1] << 8 | bytes[0];
225 outs() << "\t.short " << Value;
226 Size = 2;
227 } else {
228 if (!NoShowRawInsn)
229 DumpBytes(ArrayRef<uint8_t>(bytes, 2));
230 Value = bytes[0];
231 outs() << "\t.byte " << Value;
232 Size = 1;
233 }
234 if (Kind == MachO::DICE_KIND_DATA)
235 outs() << "\t@ KIND_DATA\n";
236 else
237 outs() << "\t@ data in code kind = " << Kind << "\n";
238 break;
239 case MachO::DICE_KIND_JUMP_TABLE8:
240 if (!NoShowRawInsn)
241 DumpBytes(ArrayRef<uint8_t>(bytes, 1));
242 Value = bytes[0];
243 outs() << "\t.byte " << format("%3u", Value) << "\t@ KIND_JUMP_TABLE8\n";
244 Size = 1;
245 break;
246 case MachO::DICE_KIND_JUMP_TABLE16:
247 if (!NoShowRawInsn)
248 DumpBytes(ArrayRef<uint8_t>(bytes, 2));
249 Value = bytes[1] << 8 | bytes[0];
250 outs() << "\t.short " << format("%5u", Value & 0xffff)
251 << "\t@ KIND_JUMP_TABLE16\n";
252 Size = 2;
253 break;
254 case MachO::DICE_KIND_JUMP_TABLE32:
255 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
256 if (!NoShowRawInsn)
257 DumpBytes(ArrayRef<uint8_t>(bytes, 4));
258 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
259 outs() << "\t.long " << Value;
260 if (Kind == MachO::DICE_KIND_JUMP_TABLE32)
261 outs() << "\t@ KIND_JUMP_TABLE32\n";
262 else
263 outs() << "\t@ KIND_ABS_JUMP_TABLE32\n";
264 Size = 4;
265 break;
266 }
267 return Size;
268}
269
270static void getSectionsAndSymbols(const MachO::mach_header Header,
271 MachOObjectFile *MachOObj,
272 std::vector<SectionRef> &Sections,
273 std::vector<SymbolRef> &Symbols,
274 SmallVectorImpl<uint64_t> &FoundFns,
275 uint64_t &BaseSegmentAddress) {
276 for (const SymbolRef &Symbol : MachOObj->symbols()) {
277 StringRef SymName;
278 Symbol.getName(SymName);
279 if (!SymName.startswith("ltmp"))
280 Symbols.push_back(Symbol);
281 }
282
283 for (const SectionRef &Section : MachOObj->sections()) {
284 StringRef SectName;
285 Section.getName(SectName);
286 Sections.push_back(Section);
287 }
288
289 MachOObjectFile::LoadCommandInfo Command =
290 MachOObj->getFirstLoadCommandInfo();
291 bool BaseSegmentAddressSet = false;
292 for (unsigned i = 0;; ++i) {
293 if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) {
294 // We found a function starts segment, parse the addresses for later
295 // consumption.
296 MachO::linkedit_data_command LLC =
297 MachOObj->getLinkeditDataLoadCommand(Command);
298
299 MachOObj->ReadULEB128s(LLC.dataoff, FoundFns);
300 } else if (Command.C.cmd == MachO::LC_SEGMENT) {
301 MachO::segment_command SLC = MachOObj->getSegmentLoadCommand(Command);
302 StringRef SegName = SLC.segname;
303 if (!BaseSegmentAddressSet && SegName != "__PAGEZERO") {
304 BaseSegmentAddressSet = true;
305 BaseSegmentAddress = SLC.vmaddr;
306 }
307 }
308
309 if (i == Header.ncmds - 1)
310 break;
311 else
312 Command = MachOObj->getNextLoadCommandInfo(Command);
313 }
314}
315
316static void PrintIndirectSymbolTable(MachOObjectFile *O, bool verbose,
317 uint32_t n, uint32_t count,
318 uint32_t stride, uint64_t addr) {
319 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
320 uint32_t nindirectsyms = Dysymtab.nindirectsyms;
321 if (n > nindirectsyms)
322 outs() << " (entries start past the end of the indirect symbol "
323 "table) (reserved1 field greater than the table size)";
324 else if (n + count > nindirectsyms)
325 outs() << " (entries extends past the end of the indirect symbol "
326 "table)";
327 outs() << "\n";
328 uint32_t cputype = O->getHeader().cputype;
329 if (cputype & MachO::CPU_ARCH_ABI64)
330 outs() << "address index";
331 else
332 outs() << "address index";
333 if (verbose)
334 outs() << " name\n";
335 else
336 outs() << "\n";
337 for (uint32_t j = 0; j < count && n + j < nindirectsyms; j++) {
338 if (cputype & MachO::CPU_ARCH_ABI64)
339 outs() << format("0x%016" PRIx64"l" "x", addr + j * stride) << " ";
340 else
341 outs() << format("0x%08" PRIx32"x", addr + j * stride) << " ";
342 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
343 uint32_t indirect_symbol = O->getIndirectSymbolTableEntry(Dysymtab, n + j);
344 if (indirect_symbol == MachO::INDIRECT_SYMBOL_LOCAL) {
345 outs() << "LOCAL\n";
346 continue;
347 }
348 if (indirect_symbol ==
349 (MachO::INDIRECT_SYMBOL_LOCAL | MachO::INDIRECT_SYMBOL_ABS)) {
350 outs() << "LOCAL ABSOLUTE\n";
351 continue;
352 }
353 if (indirect_symbol == MachO::INDIRECT_SYMBOL_ABS) {
354 outs() << "ABSOLUTE\n";
355 continue;
356 }
357 outs() << format("%5u ", indirect_symbol);
358 if (verbose) {
359 MachO::symtab_command Symtab = O->getSymtabLoadCommand();
360 if (indirect_symbol < Symtab.nsyms) {
361 symbol_iterator Sym = O->getSymbolByIndex(indirect_symbol);
362 SymbolRef Symbol = *Sym;
363 StringRef SymName;
364 Symbol.getName(SymName);
365 outs() << SymName;
366 } else {
367 outs() << "?";
368 }
369 }
370 outs() << "\n";
371 }
372}
373
374static void PrintIndirectSymbols(MachOObjectFile *O, bool verbose) {
375 uint32_t LoadCommandCount = O->getHeader().ncmds;
376 MachOObjectFile::LoadCommandInfo Load = O->getFirstLoadCommandInfo();
377 for (unsigned I = 0;; ++I) {
378 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
379 MachO::segment_command_64 Seg = O->getSegment64LoadCommand(Load);
380 for (unsigned J = 0; J < Seg.nsects; ++J) {
381 MachO::section_64 Sec = O->getSection64(Load, J);
382 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
383 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
384 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
385 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
386 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
387 section_type == MachO::S_SYMBOL_STUBS) {
388 uint32_t stride;
389 if (section_type == MachO::S_SYMBOL_STUBS)
390 stride = Sec.reserved2;
391 else
392 stride = 8;
393 if (stride == 0) {
394 outs() << "Can't print indirect symbols for (" << Sec.segname << ","
395 << Sec.sectname << ") "
396 << "(size of stubs in reserved2 field is zero)\n";
397 continue;
398 }
399 uint32_t count = Sec.size / stride;
400 outs() << "Indirect symbols for (" << Sec.segname << ","
401 << Sec.sectname << ") " << count << " entries";
402 uint32_t n = Sec.reserved1;
403 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
404 }
405 }
406 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
407 MachO::segment_command Seg = O->getSegmentLoadCommand(Load);
408 for (unsigned J = 0; J < Seg.nsects; ++J) {
409 MachO::section Sec = O->getSection(Load, J);
410 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
411 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
412 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
413 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
414 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
415 section_type == MachO::S_SYMBOL_STUBS) {
416 uint32_t stride;
417 if (section_type == MachO::S_SYMBOL_STUBS)
418 stride = Sec.reserved2;
419 else
420 stride = 4;
421 if (stride == 0) {
422 outs() << "Can't print indirect symbols for (" << Sec.segname << ","
423 << Sec.sectname << ") "
424 << "(size of stubs in reserved2 field is zero)\n";
425 continue;
426 }
427 uint32_t count = Sec.size / stride;
428 outs() << "Indirect symbols for (" << Sec.segname << ","
429 << Sec.sectname << ") " << count << " entries";
430 uint32_t n = Sec.reserved1;
431 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
432 }
433 }
434 }
435 if (I == LoadCommandCount - 1)
436 break;
437 else
438 Load = O->getNextLoadCommandInfo(Load);
439 }
440}
441
442static void PrintDataInCodeTable(MachOObjectFile *O, bool verbose) {
443 MachO::linkedit_data_command DIC = O->getDataInCodeLoadCommand();
444 uint32_t nentries = DIC.datasize / sizeof(struct MachO::data_in_code_entry);
445 outs() << "Data in code table (" << nentries << " entries)\n";
446 outs() << "offset length kind\n";
447 for (dice_iterator DI = O->begin_dices(), DE = O->end_dices(); DI != DE;
448 ++DI) {
449 uint32_t Offset;
450 DI->getOffset(Offset);
451 outs() << format("0x%08" PRIx32"x", Offset) << " ";
452 uint16_t Length;
453 DI->getLength(Length);
454 outs() << format("%6u", Length) << " ";
455 uint16_t Kind;
456 DI->getKind(Kind);
457 if (verbose) {
458 switch (Kind) {
459 case MachO::DICE_KIND_DATA:
460 outs() << "DATA";
461 break;
462 case MachO::DICE_KIND_JUMP_TABLE8:
463 outs() << "JUMP_TABLE8";
464 break;
465 case MachO::DICE_KIND_JUMP_TABLE16:
466 outs() << "JUMP_TABLE16";
467 break;
468 case MachO::DICE_KIND_JUMP_TABLE32:
469 outs() << "JUMP_TABLE32";
470 break;
471 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
472 outs() << "ABS_JUMP_TABLE32";
473 break;
474 default:
475 outs() << format("0x%04" PRIx32"x", Kind);
476 break;
477 }
478 } else
479 outs() << format("0x%04" PRIx32"x", Kind);
480 outs() << "\n";
481 }
482}
483
484static void PrintLinkOptHints(MachOObjectFile *O) {
485 MachO::linkedit_data_command LohLC = O->getLinkOptHintsLoadCommand();
486 const char *loh = O->getData().substr(LohLC.dataoff, 1).data();
487 uint32_t nloh = LohLC.datasize;
488 outs() << "Linker optimiztion hints (" << nloh << " total bytes)\n";
489 for (uint32_t i = 0; i < nloh;) {
490 unsigned n;
491 uint64_t identifier = decodeULEB128((const uint8_t *)(loh + i), &n);
492 i += n;
493 outs() << " identifier " << identifier << " ";
494 if (i >= nloh)
495 return;
496 switch (identifier) {
497 case 1:
498 outs() << "AdrpAdrp\n";
499 break;
500 case 2:
501 outs() << "AdrpLdr\n";
502 break;
503 case 3:
504 outs() << "AdrpAddLdr\n";
505 break;
506 case 4:
507 outs() << "AdrpLdrGotLdr\n";
508 break;
509 case 5:
510 outs() << "AdrpAddStr\n";
511 break;
512 case 6:
513 outs() << "AdrpLdrGotStr\n";
514 break;
515 case 7:
516 outs() << "AdrpAdd\n";
517 break;
518 case 8:
519 outs() << "AdrpLdrGot\n";
520 break;
521 default:
522 outs() << "Unknown identifier value\n";
523 break;
524 }
525 uint64_t narguments = decodeULEB128((const uint8_t *)(loh + i), &n);
526 i += n;
527 outs() << " narguments " << narguments << "\n";
528 if (i >= nloh)
529 return;
530
531 for (uint32_t j = 0; j < narguments; j++) {
532 uint64_t value = decodeULEB128((const uint8_t *)(loh + i), &n);
533 i += n;
534 outs() << "\tvalue " << format("0x%" PRIx64"l" "x", value) << "\n";
535 if (i >= nloh)
536 return;
537 }
538 }
539}
540
541static void PrintDylibs(MachOObjectFile *O, bool JustId) {
542 uint32_t LoadCommandCount = O->getHeader().ncmds;
543 MachOObjectFile::LoadCommandInfo Load = O->getFirstLoadCommandInfo();
544 for (unsigned I = 0;; ++I) {
545 if ((JustId && Load.C.cmd == MachO::LC_ID_DYLIB) ||
546 (!JustId && (Load.C.cmd == MachO::LC_ID_DYLIB ||
547 Load.C.cmd == MachO::LC_LOAD_DYLIB ||
548 Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
549 Load.C.cmd == MachO::LC_REEXPORT_DYLIB ||
550 Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
551 Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB))) {
552 MachO::dylib_command dl = O->getDylibIDLoadCommand(Load);
553 if (dl.dylib.name < dl.cmdsize) {
554 const char *p = (const char *)(Load.Ptr) + dl.dylib.name;
555 if (JustId)
556 outs() << p << "\n";
557 else {
558 outs() << "\t" << p;
559 outs() << " (compatibility version "
560 << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
561 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
562 << (dl.dylib.compatibility_version & 0xff) << ",";
563 outs() << " current version "
564 << ((dl.dylib.current_version >> 16) & 0xffff) << "."
565 << ((dl.dylib.current_version >> 8) & 0xff) << "."
566 << (dl.dylib.current_version & 0xff) << ")\n";
567 }
568 } else {
569 outs() << "\tBad offset (" << dl.dylib.name << ") for name of ";
570 if (Load.C.cmd == MachO::LC_ID_DYLIB)
571 outs() << "LC_ID_DYLIB ";
572 else if (Load.C.cmd == MachO::LC_LOAD_DYLIB)
573 outs() << "LC_LOAD_DYLIB ";
574 else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB)
575 outs() << "LC_LOAD_WEAK_DYLIB ";
576 else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB)
577 outs() << "LC_LAZY_LOAD_DYLIB ";
578 else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB)
579 outs() << "LC_REEXPORT_DYLIB ";
580 else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
581 outs() << "LC_LOAD_UPWARD_DYLIB ";
582 else
583 outs() << "LC_??? ";
584 outs() << "command " << I << "\n";
585 }
586 }
587 if (I == LoadCommandCount - 1)
588 break;
589 else
590 Load = O->getNextLoadCommandInfo(Load);
591 }
592}
593
594typedef DenseMap<uint64_t, StringRef> SymbolAddressMap;
595
596static void CreateSymbolAddressMap(MachOObjectFile *O,
597 SymbolAddressMap *AddrMap) {
598 // Create a map of symbol addresses to symbol names.
599 for (const SymbolRef &Symbol : O->symbols()) {
600 SymbolRef::Type ST;
601 Symbol.getType(ST);
602 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
603 ST == SymbolRef::ST_Other) {
604 uint64_t Address;
605 Symbol.getAddress(Address);
606 StringRef SymName;
607 Symbol.getName(SymName);
608 (*AddrMap)[Address] = SymName;
609 }
610 }
611}
612
613// GuessSymbolName is passed the address of what might be a symbol and a
614// pointer to the SymbolAddressMap. It returns the name of a symbol
615// with that address or nullptr if no symbol is found with that address.
616static const char *GuessSymbolName(uint64_t value, SymbolAddressMap *AddrMap) {
617 const char *SymbolName = nullptr;
618 // A DenseMap can't lookup up some values.
619 if (value != 0xffffffffffffffffULL && value != 0xfffffffffffffffeULL) {
620 StringRef name = AddrMap->lookup(value);
621 if (!name.empty())
622 SymbolName = name.data();
623 }
624 return SymbolName;
625}
626
627static void DumpCstringChar(const char c) {
628 char p[2];
629 p[0] = c;
630 p[1] = '\0';
631 outs().write_escaped(p);
632}
633
634static void DumpCstringSection(MachOObjectFile *O, const char *sect,
635 uint32_t sect_size, uint64_t sect_addr,
636 bool print_addresses) {
637 for (uint32_t i = 0; i < sect_size; i++) {
638 if (print_addresses) {
639 if (O->is64Bit())
640 outs() << format("%016" PRIx64"l" "x", sect_addr + i) << " ";
641 else
642 outs() << format("%08" PRIx64"l" "x", sect_addr + i) << " ";
643 }
644 for (; i < sect_size && sect[i] != '\0'; i++)
645 DumpCstringChar(sect[i]);
646 if (i < sect_size && sect[i] == '\0')
647 outs() << "\n";
648 }
649}
650
651static void DumpLiteral4(uint32_t l, float f) {
652 outs() << format("0x%08" PRIx32"x", l);
653 if ((l & 0x7f800000) != 0x7f800000)
654 outs() << format(" (%.16e)\n", f);
655 else {
656 if (l == 0x7f800000)
657 outs() << " (+Infinity)\n";
658 else if (l == 0xff800000)
659 outs() << " (-Infinity)\n";
660 else if ((l & 0x00400000) == 0x00400000)
661 outs() << " (non-signaling Not-a-Number)\n";
662 else
663 outs() << " (signaling Not-a-Number)\n";
664 }
665}
666
667static void DumpLiteral4Section(MachOObjectFile *O, const char *sect,
668 uint32_t sect_size, uint64_t sect_addr,
669 bool print_addresses) {
670 for (uint32_t i = 0; i < sect_size; i += sizeof(float)) {
671 if (print_addresses) {
672 if (O->is64Bit())
673 outs() << format("%016" PRIx64"l" "x", sect_addr + i) << " ";
674 else
675 outs() << format("%08" PRIx64"l" "x", sect_addr + i) << " ";
676 }
677 float f;
678 memcpy(&f, sect + i, sizeof(float));
679 if (O->isLittleEndian() != sys::IsLittleEndianHost)
680 sys::swapByteOrder(f);
681 uint32_t l;
682 memcpy(&l, sect + i, sizeof(uint32_t));
683 if (O->isLittleEndian() != sys::IsLittleEndianHost)
684 sys::swapByteOrder(l);
685 DumpLiteral4(l, f);
686 }
687}
688
689static void DumpLiteral8(MachOObjectFile *O, uint32_t l0, uint32_t l1,
690 double d) {
691 outs() << format("0x%08" PRIx32"x", l0) << " " << format("0x%08" PRIx32"x", l1);
692 uint32_t Hi, Lo;
693 if (O->isLittleEndian()) {
694 Hi = l1;
695 Lo = l0;
696 } else {
697 Hi = l0;
698 Lo = l1;
699 }
700 // Hi is the high word, so this is equivalent to if(isfinite(d))
701 if ((Hi & 0x7ff00000) != 0x7ff00000)
702 outs() << format(" (%.16e)\n", d);
703 else {
704 if (Hi == 0x7ff00000 && Lo == 0)
705 outs() << " (+Infinity)\n";
706 else if (Hi == 0xfff00000 && Lo == 0)
707 outs() << " (-Infinity)\n";
708 else if ((Hi & 0x00080000) == 0x00080000)
709 outs() << " (non-signaling Not-a-Number)\n";
710 else
711 outs() << " (signaling Not-a-Number)\n";
712 }
713}
714
715static void DumpLiteral8Section(MachOObjectFile *O, const char *sect,
716 uint32_t sect_size, uint64_t sect_addr,
717 bool print_addresses) {
718 for (uint32_t i = 0; i < sect_size; i += sizeof(double)) {
719 if (print_addresses) {
720 if (O->is64Bit())
721 outs() << format("%016" PRIx64"l" "x", sect_addr + i) << " ";
722 else
723 outs() << format("%08" PRIx64"l" "x", sect_addr + i) << " ";
724 }
725 double d;
726 memcpy(&d, sect + i, sizeof(double));
727 if (O->isLittleEndian() != sys::IsLittleEndianHost)
728 sys::swapByteOrder(d);
729 uint32_t l0, l1;
730 memcpy(&l0, sect + i, sizeof(uint32_t));
731 memcpy(&l1, sect + i + sizeof(uint32_t), sizeof(uint32_t));
732 if (O->isLittleEndian() != sys::IsLittleEndianHost) {
733 sys::swapByteOrder(l0);
734 sys::swapByteOrder(l1);
735 }
736 DumpLiteral8(O, l0, l1, d);
737 }
738}
739
740static void DumpLiteral16(uint32_t l0, uint32_t l1, uint32_t l2, uint32_t l3) {
741 outs() << format("0x%08" PRIx32"x", l0) << " ";
742 outs() << format("0x%08" PRIx32"x", l1) << " ";
743 outs() << format("0x%08" PRIx32"x", l2) << " ";
744 outs() << format("0x%08" PRIx32"x", l3) << "\n";
745}
746
747static void DumpLiteral16Section(MachOObjectFile *O, const char *sect,
748 uint32_t sect_size, uint64_t sect_addr,
749 bool print_addresses) {
750 for (uint32_t i = 0; i < sect_size; i += 16) {
751 if (print_addresses) {
752 if (O->is64Bit())
753 outs() << format("%016" PRIx64"l" "x", sect_addr + i) << " ";
754 else
755 outs() << format("%08" PRIx64"l" "x", sect_addr + i) << " ";
756 }
757 uint32_t l0, l1, l2, l3;
758 memcpy(&l0, sect + i, sizeof(uint32_t));
759 memcpy(&l1, sect + i + sizeof(uint32_t), sizeof(uint32_t));
760 memcpy(&l2, sect + i + 2 * sizeof(uint32_t), sizeof(uint32_t));
761 memcpy(&l3, sect + i + 3 * sizeof(uint32_t), sizeof(uint32_t));
762 if (O->isLittleEndian() != sys::IsLittleEndianHost) {
763 sys::swapByteOrder(l0);
764 sys::swapByteOrder(l1);
765 sys::swapByteOrder(l2);
766 sys::swapByteOrder(l3);
767 }
768 DumpLiteral16(l0, l1, l2, l3);
769 }
770}
771
772static void DumpLiteralPointerSection(MachOObjectFile *O,
773 const SectionRef &Section,
774 const char *sect, uint32_t sect_size,
775 uint64_t sect_addr,
776 bool print_addresses) {
777 // Collect the literal sections in this Mach-O file.
778 std::vector<SectionRef> LiteralSections;
779 for (const SectionRef &Section : O->sections()) {
780 DataRefImpl Ref = Section.getRawDataRefImpl();
781 uint32_t section_type;
782 if (O->is64Bit()) {
783 const MachO::section_64 Sec = O->getSection64(Ref);
784 section_type = Sec.flags & MachO::SECTION_TYPE;
785 } else {
786 const MachO::section Sec = O->getSection(Ref);
787 section_type = Sec.flags & MachO::SECTION_TYPE;
788 }
789 if (section_type == MachO::S_CSTRING_LITERALS ||
790 section_type == MachO::S_4BYTE_LITERALS ||
791 section_type == MachO::S_8BYTE_LITERALS ||
792 section_type == MachO::S_16BYTE_LITERALS)
793 LiteralSections.push_back(Section);
794 }
795
796 // Set the size of the literal pointer.
797 uint32_t lp_size = O->is64Bit() ? 8 : 4;
798
799 // Collect the external relocation symbols for the the literal pointers.
800 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
801 for (const RelocationRef &Reloc : Section.relocations()) {
802 DataRefImpl Rel;
803 MachO::any_relocation_info RE;
804 bool isExtern = false;
805 Rel = Reloc.getRawDataRefImpl();
806 RE = O->getRelocation(Rel);
807 isExtern = O->getPlainRelocationExternal(RE);
808 if (isExtern) {
809 uint64_t RelocOffset;
810 Reloc.getOffset(RelocOffset);
811 symbol_iterator RelocSym = Reloc.getSymbol();
812 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
813 }
814 }
815 array_pod_sort(Relocs.begin(), Relocs.end());
816
817 // Dump each literal pointer.
818 for (uint32_t i = 0; i < sect_size; i += lp_size) {
819 if (print_addresses) {
820 if (O->is64Bit())
821 outs() << format("%016" PRIx64"l" "x", sect_addr + i) << " ";
822 else
823 outs() << format("%08" PRIx64"l" "x", sect_addr + i) << " ";
824 }
825 uint64_t lp;
826 if (O->is64Bit()) {
827 memcpy(&lp, sect + i, sizeof(uint64_t));
828 if (O->isLittleEndian() != sys::IsLittleEndianHost)
829 sys::swapByteOrder(lp);
830 } else {
831 uint32_t li;
832 memcpy(&li, sect + i, sizeof(uint32_t));
833 if (O->isLittleEndian() != sys::IsLittleEndianHost)
834 sys::swapByteOrder(li);
835 lp = li;
836 }
837
838 // First look for an external relocation entry for this literal pointer.
839 bool reloc_found = false;
840 for (unsigned j = 0, e = Relocs.size(); j != e; ++j) {
841 if (Relocs[i].first == i) {
842 symbol_iterator RelocSym = Relocs[j].second;
843 StringRef SymName;
844 RelocSym->getName(SymName);
845 outs() << "external relocation entry for symbol:" << SymName << "\n";
846 reloc_found = true;
847 }
848 }
849 if (reloc_found == true)
850 continue;
851
852 // For local references see what the section the literal pointer points to.
853 bool found = false;
854 for (unsigned SectIdx = 0; SectIdx != LiteralSections.size(); SectIdx++) {
855 uint64_t SectAddress = LiteralSections[SectIdx].getAddress();
856 uint64_t SectSize = LiteralSections[SectIdx].getSize();
857 if (lp >= SectAddress && lp < SectAddress + SectSize) {
858 found = true;
859
860 StringRef SectName;
861 LiteralSections[SectIdx].getName(SectName);
862 DataRefImpl Ref = LiteralSections[SectIdx].getRawDataRefImpl();
863 StringRef SegmentName = O->getSectionFinalSegmentName(Ref);
864 outs() << SegmentName << ":" << SectName << ":";
865
866 uint32_t section_type;
867 if (O->is64Bit()) {
868 const MachO::section_64 Sec = O->getSection64(Ref);
869 section_type = Sec.flags & MachO::SECTION_TYPE;
870 } else {
871 const MachO::section Sec = O->getSection(Ref);
872 section_type = Sec.flags & MachO::SECTION_TYPE;
873 }
874
875 StringRef BytesStr;
876 LiteralSections[SectIdx].getContents(BytesStr);
877 const char *Contents = reinterpret_cast<const char *>(BytesStr.data());
878
879 switch (section_type) {
880 case MachO::S_CSTRING_LITERALS:
881 for (uint64_t i = lp - SectAddress;
882 i < SectSize && Contents[i] != '\0'; i++) {
883 DumpCstringChar(Contents[i]);
884 }
885 outs() << "\n";
886 break;
887 case MachO::S_4BYTE_LITERALS:
888 float f;
889 memcpy(&f, Contents + (lp - SectAddress), sizeof(float));
890 uint32_t l;
891 memcpy(&l, Contents + (lp - SectAddress), sizeof(uint32_t));
892 if (O->isLittleEndian() != sys::IsLittleEndianHost) {
893 sys::swapByteOrder(f);
894 sys::swapByteOrder(l);
895 }
896 DumpLiteral4(l, f);
897 break;
898 case MachO::S_8BYTE_LITERALS: {
899 double d;
900 memcpy(&d, Contents + (lp - SectAddress), sizeof(double));
901 uint32_t l0, l1;
902 memcpy(&l0, Contents + (lp - SectAddress), sizeof(uint32_t));
903 memcpy(&l1, Contents + (lp - SectAddress) + sizeof(uint32_t),
904 sizeof(uint32_t));
905 if (O->isLittleEndian() != sys::IsLittleEndianHost) {
906 sys::swapByteOrder(f);
907 sys::swapByteOrder(l0);
908 sys::swapByteOrder(l1);
909 }
910 DumpLiteral8(O, l0, l1, d);
911 break;
912 }
913 case MachO::S_16BYTE_LITERALS: {
914 uint32_t l0, l1, l2, l3;
915 memcpy(&l0, Contents + (lp - SectAddress), sizeof(uint32_t));
916 memcpy(&l1, Contents + (lp - SectAddress) + sizeof(uint32_t),
917 sizeof(uint32_t));
918 memcpy(&l2, Contents + (lp - SectAddress) + 2 * sizeof(uint32_t),
919 sizeof(uint32_t));
920 memcpy(&l3, Contents + (lp - SectAddress) + 3 * sizeof(uint32_t),
921 sizeof(uint32_t));
922 if (O->isLittleEndian() != sys::IsLittleEndianHost) {
923 sys::swapByteOrder(l0);
924 sys::swapByteOrder(l1);
925 sys::swapByteOrder(l2);
926 sys::swapByteOrder(l3);
927 }
928 DumpLiteral16(l0, l1, l2, l3);
929 break;
930 }
931 }
932 }
933 }
934 if (found == false)
935 outs() << format("0x%" PRIx64"l" "x", lp) << " (not in a literal section)\n";
936 }
937}
938
939static void DumpInitTermPointerSection(MachOObjectFile *O, const char *sect,
940 uint32_t sect_size, uint64_t sect_addr,
941 SymbolAddressMap *AddrMap,
942 bool verbose) {
943 uint32_t stride;
944 if (O->is64Bit())
945 stride = sizeof(uint64_t);
946 else
947 stride = sizeof(uint32_t);
948 for (uint32_t i = 0; i < sect_size; i += stride) {
949 const char *SymbolName = nullptr;
950 if (O->is64Bit()) {
951 outs() << format("0x%016" PRIx64"l" "x", sect_addr + i * stride) << " ";
952 uint64_t pointer_value;
953 memcpy(&pointer_value, sect + i, stride);
954 if (O->isLittleEndian() != sys::IsLittleEndianHost)
955 sys::swapByteOrder(pointer_value);
956 outs() << format("0x%016" PRIx64"l" "x", pointer_value);
957 if (verbose)
958 SymbolName = GuessSymbolName(pointer_value, AddrMap);
959 } else {
960 outs() << format("0x%08" PRIx64"l" "x", sect_addr + i * stride) << " ";
961 uint32_t pointer_value;
962 memcpy(&pointer_value, sect + i, stride);
963 if (O->isLittleEndian() != sys::IsLittleEndianHost)
964 sys::swapByteOrder(pointer_value);
965 outs() << format("0x%08" PRIx32"x", pointer_value);
966 if (verbose)
967 SymbolName = GuessSymbolName(pointer_value, AddrMap);
968 }
969 if (SymbolName)
970 outs() << " " << SymbolName;
971 outs() << "\n";
972 }
973}
974
975static void DumpRawSectionContents(MachOObjectFile *O, const char *sect,
976 uint32_t size, uint64_t addr) {
977 uint32_t cputype = O->getHeader().cputype;
978 if (cputype == MachO::CPU_TYPE_I386 || cputype == MachO::CPU_TYPE_X86_64) {
979 uint32_t j;
980 for (uint32_t i = 0; i < size; i += j, addr += j) {
981 if (O->is64Bit())
982 outs() << format("%016" PRIx64"l" "x", addr) << "\t";
983 else
984 outs() << format("%08" PRIx64"l" "x", addr) << "\t";
985 for (j = 0; j < 16 && i + j < size; j++) {
986 uint8_t byte_word = *(sect + i + j);
987 outs() << format("%02" PRIx32"x", (uint32_t)byte_word) << " ";
988 }
989 outs() << "\n";
990 }
991 } else {
992 uint32_t j;
993 for (uint32_t i = 0; i < size; i += j, addr += j) {
994 if (O->is64Bit())
995 outs() << format("%016" PRIx64"l" "x", addr) << "\t";
996 else
997 outs() << format("%08" PRIx64"l" "x", sect) << "\t";
998 for (j = 0; j < 4 * sizeof(int32_t) && i + j < size;
999 j += sizeof(int32_t)) {
1000 if (i + j + sizeof(int32_t) < size) {
1001 uint32_t long_word;
1002 memcpy(&long_word, sect + i + j, sizeof(int32_t));
1003 if (O->isLittleEndian() != sys::IsLittleEndianHost)
1004 sys::swapByteOrder(long_word);
1005 outs() << format("%08" PRIx32"x", long_word) << " ";
1006 } else {
1007 for (uint32_t k = 0; i + j + k < size; k++) {
1008 uint8_t byte_word = *(sect + i + j);
1009 outs() << format("%02" PRIx32"x", (uint32_t)byte_word) << " ";
1010 }
1011 }
1012 }
1013 outs() << "\n";
1014 }
1015 }
1016}
1017
1018static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
1019 StringRef DisSegName, StringRef DisSectName);
1020
1021static void DumpSectionContents(StringRef Filename, MachOObjectFile *O,
1022 bool verbose) {
1023 SymbolAddressMap AddrMap;
1024 if (verbose)
1025 CreateSymbolAddressMap(O, &AddrMap);
1026
1027 for (unsigned i = 0; i < DumpSections.size(); ++i) {
1028 StringRef DumpSection = DumpSections[i];
1029 std::pair<StringRef, StringRef> DumpSegSectName;
1030 DumpSegSectName = DumpSection.split(',');
1031 StringRef DumpSegName, DumpSectName;
1032 if (DumpSegSectName.second.size()) {
1033 DumpSegName = DumpSegSectName.first;
1034 DumpSectName = DumpSegSectName.second;
1035 } else {
1036 DumpSegName = "";
1037 DumpSectName = DumpSegSectName.first;
1038 }
1039 for (const SectionRef &Section : O->sections()) {
1040 StringRef SectName;
1041 Section.getName(SectName);
1042 DataRefImpl Ref = Section.getRawDataRefImpl();
1043 StringRef SegName = O->getSectionFinalSegmentName(Ref);
1044 if ((DumpSegName.empty() || SegName == DumpSegName) &&
1045 (SectName == DumpSectName)) {
1046 outs() << "Contents of (" << SegName << "," << SectName
1047 << ") section\n";
1048 uint32_t section_flags;
1049 if (O->is64Bit()) {
1050 const MachO::section_64 Sec = O->getSection64(Ref);
1051 section_flags = Sec.flags;
1052
1053 } else {
1054 const MachO::section Sec = O->getSection(Ref);
1055 section_flags = Sec.flags;
1056 }
1057 uint32_t section_type = section_flags & MachO::SECTION_TYPE;
1058
1059 StringRef BytesStr;
1060 Section.getContents(BytesStr);
1061 const char *sect = reinterpret_cast<const char *>(BytesStr.data());
1062 uint32_t sect_size = BytesStr.size();
1063 uint64_t sect_addr = Section.getAddress();
1064
1065 if (verbose) {
1066 if ((section_flags & MachO::S_ATTR_PURE_INSTRUCTIONS) ||
1067 (section_flags & MachO::S_ATTR_SOME_INSTRUCTIONS)) {
1068 DisassembleMachO(Filename, O, SegName, SectName);
1069 continue;
1070 }
1071 if (SegName == "__TEXT" && SectName == "__info_plist") {
1072 outs() << sect;
1073 continue;
1074 }
1075 switch (section_type) {
1076 case MachO::S_REGULAR:
1077 DumpRawSectionContents(O, sect, sect_size, sect_addr);
1078 break;
1079 case MachO::S_ZEROFILL:
1080 outs() << "zerofill section and has no contents in the file\n";
1081 break;
1082 case MachO::S_CSTRING_LITERALS:
1083 DumpCstringSection(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1084 break;
1085 case MachO::S_4BYTE_LITERALS:
1086 DumpLiteral4Section(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1087 break;
1088 case MachO::S_8BYTE_LITERALS:
1089 DumpLiteral8Section(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1090 break;
1091 case MachO::S_16BYTE_LITERALS:
1092 DumpLiteral16Section(O, sect, sect_size, sect_addr, !NoLeadingAddr);
1093 break;
1094 case MachO::S_LITERAL_POINTERS:
1095 DumpLiteralPointerSection(O, Section, sect, sect_size, sect_addr,
1096 !NoLeadingAddr);
1097 break;
1098 case MachO::S_MOD_INIT_FUNC_POINTERS:
1099 case MachO::S_MOD_TERM_FUNC_POINTERS:
1100 DumpInitTermPointerSection(O, sect, sect_size, sect_addr, &AddrMap,
1101 verbose);
1102 break;
1103 default:
1104 outs() << "Unknown section type ("
1105 << format("0x%08" PRIx32"x", section_type) << ")\n";
1106 DumpRawSectionContents(O, sect, sect_size, sect_addr);
1107 break;
1108 }
1109 } else {
1110 if (section_type == MachO::S_ZEROFILL)
1111 outs() << "zerofill section and has no contents in the file\n";
1112 else
1113 DumpRawSectionContents(O, sect, sect_size, sect_addr);
1114 }
1115 }
1116 }
1117 }
1118}
1119
1120static void DumpInfoPlistSectionContents(StringRef Filename,
1121 MachOObjectFile *O) {
1122 for (const SectionRef &Section : O->sections()) {
1123 StringRef SectName;
1124 Section.getName(SectName);
1125 DataRefImpl Ref = Section.getRawDataRefImpl();
1126 StringRef SegName = O->getSectionFinalSegmentName(Ref);
1127 if (SegName == "__TEXT" && SectName == "__info_plist") {
1128 outs() << "Contents of (" << SegName << "," << SectName << ") section\n";
1129 StringRef BytesStr;
1130 Section.getContents(BytesStr);
1131 const char *sect = reinterpret_cast<const char *>(BytesStr.data());
1132 outs() << sect;
1133 return;
1134 }
1135 }
1136}
1137
1138// checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file
1139// and if it is and there is a list of architecture flags is specified then
1140// check to make sure this Mach-O file is one of those architectures or all
1141// architectures were specified. If not then an error is generated and this
1142// routine returns false. Else it returns true.
1143static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) {
1144 if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) {
1145 MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O);
1146 bool ArchFound = false;
1147 MachO::mach_header H;
1148 MachO::mach_header_64 H_64;
1149 Triple T;
1150 if (MachO->is64Bit()) {
1151 H_64 = MachO->MachOObjectFile::getHeader64();
1152 T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype);
1153 } else {
1154 H = MachO->MachOObjectFile::getHeader();
1155 T = MachOObjectFile::getArch(H.cputype, H.cpusubtype);
1156 }
1157 unsigned i;
1158 for (i = 0; i < ArchFlags.size(); ++i) {
1159 if (ArchFlags[i] == T.getArchName())
1160 ArchFound = true;
1161 break;
1162 }
1163 if (!ArchFound) {
1164 errs() << "llvm-objdump: file: " + Filename + " does not contain "
1165 << "architecture: " + ArchFlags[i] + "\n";
1166 return false;
1167 }
1168 }
1169 return true;
1170}
1171
1172// ProcessMachO() is passed a single opened Mach-O file, which may be an
1173// archive member and or in a slice of a universal file. It prints the
1174// the file name and header info and then processes it according to the
1175// command line options.
1176static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF,
1177 StringRef ArchiveMemberName = StringRef(),
1178 StringRef ArchitectureName = StringRef()) {
1179 // If we are doing some processing here on the Mach-O file print the header
1180 // info. And don't print it otherwise like in the case of printing the
1181 // UniversalHeaders or ArchiveHeaders.
1182 if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind ||
1183 LazyBind || WeakBind || IndirectSymbols || DataInCode || LinkOptHints ||
1184 DylibsUsed || DylibId || DumpSections.size() != 0) {
1185 outs() << Filename;
1186 if (!ArchiveMemberName.empty())
1187 outs() << '(' << ArchiveMemberName << ')';
1188 if (!ArchitectureName.empty())
1189 outs() << " (architecture " << ArchitectureName << ")";
1190 outs() << ":\n";
1191 }
1192
1193 if (Disassemble)
1194 DisassembleMachO(Filename, MachOOF, "__TEXT", "__text");
1195 if (IndirectSymbols)
1196 PrintIndirectSymbols(MachOOF, !NonVerbose);
1197 if (DataInCode)
1198 PrintDataInCodeTable(MachOOF, !NonVerbose);
1199 if (LinkOptHints)
1200 PrintLinkOptHints(MachOOF);
1201 if (Relocations)
1202 PrintRelocations(MachOOF);
1203 if (SectionHeaders)
1204 PrintSectionHeaders(MachOOF);
1205 if (SectionContents)
1206 PrintSectionContents(MachOOF);
1207 if (DumpSections.size() != 0)
1208 DumpSectionContents(Filename, MachOOF, !NonVerbose);
1209 if (InfoPlist)
1210 DumpInfoPlistSectionContents(Filename, MachOOF);
1211 if (DylibsUsed)
1212 PrintDylibs(MachOOF, false);
1213 if (DylibId)
1214 PrintDylibs(MachOOF, true);
1215 if (SymbolTable)
1216 PrintSymbolTable(MachOOF);
1217 if (UnwindInfo)
1218 printMachOUnwindInfo(MachOOF);
1219 if (PrivateHeaders)
1220 printMachOFileHeader(MachOOF);
1221 if (ExportsTrie)
1222 printExportsTrie(MachOOF);
1223 if (Rebase)
1224 printRebaseTable(MachOOF);
1225 if (Bind)
1226 printBindTable(MachOOF);
1227 if (LazyBind)
1228 printLazyBindTable(MachOOF);
1229 if (WeakBind)
1230 printWeakBindTable(MachOOF);
1231}
1232
1233// printUnknownCPUType() helps print_fat_headers for unknown CPU's.
1234static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) {
1235 outs() << " cputype (" << cputype << ")\n";
1236 outs() << " cpusubtype (" << cpusubtype << ")\n";
1237}
1238
1239// printCPUType() helps print_fat_headers by printing the cputype and
1240// pusubtype (symbolically for the one's it knows about).
1241static void printCPUType(uint32_t cputype, uint32_t cpusubtype) {
1242 switch (cputype) {
1243 case MachO::CPU_TYPE_I386:
1244 switch (cpusubtype) {
1245 case MachO::CPU_SUBTYPE_I386_ALL:
1246 outs() << " cputype CPU_TYPE_I386\n";
1247 outs() << " cpusubtype CPU_SUBTYPE_I386_ALL\n";
1248 break;
1249 default:
1250 printUnknownCPUType(cputype, cpusubtype);
1251 break;
1252 }
1253 break;
1254 case MachO::CPU_TYPE_X86_64:
1255 switch (cpusubtype) {
1256 case MachO::CPU_SUBTYPE_X86_64_ALL:
1257 outs() << " cputype CPU_TYPE_X86_64\n";
1258 outs() << " cpusubtype CPU_SUBTYPE_X86_64_ALL\n";
1259 break;
1260 case MachO::CPU_SUBTYPE_X86_64_H:
1261 outs() << " cputype CPU_TYPE_X86_64\n";
1262 outs() << " cpusubtype CPU_SUBTYPE_X86_64_H\n";
1263 break;
1264 default:
1265 printUnknownCPUType(cputype, cpusubtype);
1266 break;
1267 }
1268 break;
1269 case MachO::CPU_TYPE_ARM:
1270 switch (cpusubtype) {
1271 case MachO::CPU_SUBTYPE_ARM_ALL:
1272 outs() << " cputype CPU_TYPE_ARM\n";
1273 outs() << " cpusubtype CPU_SUBTYPE_ARM_ALL\n";
1274 break;
1275 case MachO::CPU_SUBTYPE_ARM_V4T:
1276 outs() << " cputype CPU_TYPE_ARM\n";
1277 outs() << " cpusubtype CPU_SUBTYPE_ARM_V4T\n";
1278 break;
1279 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
1280 outs() << " cputype CPU_TYPE_ARM\n";
1281 outs() << " cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n";
1282 break;
1283 case MachO::CPU_SUBTYPE_ARM_XSCALE:
1284 outs() << " cputype CPU_TYPE_ARM\n";
1285 outs() << " cpusubtype CPU_SUBTYPE_ARM_XSCALE\n";
1286 break;
1287 case MachO::CPU_SUBTYPE_ARM_V6:
1288 outs() << " cputype CPU_TYPE_ARM\n";
1289 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6\n";
1290 break;
1291 case MachO::CPU_SUBTYPE_ARM_V6M:
1292 outs() << " cputype CPU_TYPE_ARM\n";
1293 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6M\n";
1294 break;
1295 case MachO::CPU_SUBTYPE_ARM_V7:
1296 outs() << " cputype CPU_TYPE_ARM\n";
1297 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7\n";
1298 break;
1299 case MachO::CPU_SUBTYPE_ARM_V7EM:
1300 outs() << " cputype CPU_TYPE_ARM\n";
1301 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7EM\n";
1302 break;
1303 case MachO::CPU_SUBTYPE_ARM_V7K:
1304 outs() << " cputype CPU_TYPE_ARM\n";
1305 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7K\n";
1306 break;
1307 case MachO::CPU_SUBTYPE_ARM_V7M:
1308 outs() << " cputype CPU_TYPE_ARM\n";
1309 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7M\n";
1310 break;
1311 case MachO::CPU_SUBTYPE_ARM_V7S:
1312 outs() << " cputype CPU_TYPE_ARM\n";
1313 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7S\n";
1314 break;
1315 default:
1316 printUnknownCPUType(cputype, cpusubtype);
1317 break;
1318 }
1319 break;
1320 case MachO::CPU_TYPE_ARM64:
1321 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
1322 case MachO::CPU_SUBTYPE_ARM64_ALL:
1323 outs() << " cputype CPU_TYPE_ARM64\n";
1324 outs() << " cpusubtype CPU_SUBTYPE_ARM64_ALL\n";
1325 break;
1326 default:
1327 printUnknownCPUType(cputype, cpusubtype);
1328 break;
1329 }
1330 break;
1331 default:
1332 printUnknownCPUType(cputype, cpusubtype);
1333 break;
1334 }
1335}
1336
1337static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB,
1338 bool verbose) {
1339 outs() << "Fat headers\n";
1340 if (verbose)
1341 outs() << "fat_magic FAT_MAGIC\n";
1342 else
1343 outs() << "fat_magic " << format("0x%" PRIx32"x", MachO::FAT_MAGIC) << "\n";
1344
1345 uint32_t nfat_arch = UB->getNumberOfObjects();
1346 StringRef Buf = UB->getData();
1347 uint64_t size = Buf.size();
1348 uint64_t big_size = sizeof(struct MachO::fat_header) +
1349 nfat_arch * sizeof(struct MachO::fat_arch);
1350 outs() << "nfat_arch " << UB->getNumberOfObjects();
1351 if (nfat_arch == 0)
1352 outs() << " (malformed, contains zero architecture types)\n";
1353 else if (big_size > size)
1354 outs() << " (malformed, architectures past end of file)\n";
1355 else
1356 outs() << "\n";
1357
1358 for (uint32_t i = 0; i < nfat_arch; ++i) {
1359 MachOUniversalBinary::ObjectForArch OFA(UB, i);
1360 uint32_t cputype = OFA.getCPUType();
1361 uint32_t cpusubtype = OFA.getCPUSubType();
1362 outs() << "architecture ";
1363 for (uint32_t j = 0; i != 0 && j <= i - 1; j++) {
1364 MachOUniversalBinary::ObjectForArch other_OFA(UB, j);
1365 uint32_t other_cputype = other_OFA.getCPUType();
1366 uint32_t other_cpusubtype = other_OFA.getCPUSubType();
1367 if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype &&
1368 (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) ==
1369 (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) {
1370 outs() << "(illegal duplicate architecture) ";
1371 break;
1372 }
1373 }
1374 if (verbose) {
1375 outs() << OFA.getArchTypeName() << "\n";
1376 printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
1377 } else {
1378 outs() << i << "\n";
1379 outs() << " cputype " << cputype << "\n";
1380 outs() << " cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK)
1381 << "\n";
1382 }
1383 if (verbose &&
1384 (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64)
1385 outs() << " capabilities CPU_SUBTYPE_LIB64\n";
1386 else
1387 outs() << " capabilities "
1388 << format("0x%" PRIx32"x",
1389 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n";
1390 outs() << " offset " << OFA.getOffset();
1391 if (OFA.getOffset() > size)
1392 outs() << " (past end of file)";
1393 if (OFA.getOffset() % (1 << OFA.getAlign()) != 0)
1394 outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")";
1395 outs() << "\n";
1396 outs() << " size " << OFA.getSize();
1397 big_size = OFA.getOffset() + OFA.getSize();
1398 if (big_size > size)
1399 outs() << " (past end of file)";
1400 outs() << "\n";
1401 outs() << " align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign())
1402 << ")\n";
1403 }
1404}
1405
1406static void printArchiveChild(Archive::Child &C, bool verbose,
1407 bool print_offset) {
1408 if (print_offset)
1409 outs() << C.getChildOffset() << "\t";
1410 sys::fs::perms Mode = C.getAccessMode();
1411 if (verbose) {
1412 // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG.
1413 // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG.
1414 outs() << "-";
1415 if (Mode & sys::fs::owner_read)
1416 outs() << "r";
1417 else
1418 outs() << "-";
1419 if (Mode & sys::fs::owner_write)
1420 outs() << "w";
1421 else
1422 outs() << "-";
1423 if (Mode & sys::fs::owner_exe)
1424 outs() << "x";
1425 else
1426 outs() << "-";
1427 if (Mode & sys::fs::group_read)
1428 outs() << "r";
1429 else
1430 outs() << "-";
1431 if (Mode & sys::fs::group_write)
1432 outs() << "w";
1433 else
1434 outs() << "-";
1435 if (Mode & sys::fs::group_exe)
1436 outs() << "x";
1437 else
1438 outs() << "-";
1439 if (Mode & sys::fs::others_read)
1440 outs() << "r";
1441 else
1442 outs() << "-";
1443 if (Mode & sys::fs::others_write)
1444 outs() << "w";
1445 else
1446 outs() << "-";
1447 if (Mode & sys::fs::others_exe)
1448 outs() << "x";
1449 else
1450 outs() << "-";
1451 } else {
1452 outs() << format("0%o ", Mode);
1453 }
1454
1455 unsigned UID = C.getUID();
1456 outs() << format("%3d/", UID);
1457 unsigned GID = C.getGID();
1458 outs() << format("%-3d ", GID);
1459 uint64_t Size = C.getRawSize();
1460 outs() << format("%5" PRId64"l" "d", Size) << " ";
1461
1462 StringRef RawLastModified = C.getRawLastModified();
1463 if (verbose) {
1464 unsigned Seconds;
1465 if (RawLastModified.getAsInteger(10, Seconds))
1466 outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified;
1467 else {
1468 // Since cime(3) returns a 26 character string of the form:
1469 // "Sun Sep 16 01:03:52 1973\n\0"
1470 // just print 24 characters.
1471 time_t t = Seconds;
1472 outs() << format("%.24s ", ctime(&t));
1473 }
1474 } else {
1475 outs() << RawLastModified << " ";
1476 }
1477
1478 if (verbose) {
1479 ErrorOr<StringRef> NameOrErr = C.getName();
1480 if (NameOrErr.getError()) {
1481 StringRef RawName = C.getRawName();
1482 outs() << RawName << "\n";
1483 } else {
1484 StringRef Name = NameOrErr.get();
1485 outs() << Name << "\n";
1486 }
1487 } else {
1488 StringRef RawName = C.getRawName();
1489 outs() << RawName << "\n";
1490 }
1491}
1492
1493static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) {
1494 if (A->hasSymbolTable()) {
1495 Archive::child_iterator S = A->getSymbolTableChild();
1496 Archive::Child C = *S;
1497 printArchiveChild(C, verbose, print_offset);
1498 }
1499 for (Archive::child_iterator I = A->child_begin(), E = A->child_end(); I != E;
1500 ++I) {
1501 Archive::Child C = *I;
1502 printArchiveChild(C, verbose, print_offset);
1503 }
1504}
1505
1506// ParseInputMachO() parses the named Mach-O file in Filename and handles the
1507// -arch flags selecting just those slices as specified by them and also parses
1508// archive files. Then for each individual Mach-O file ProcessMachO() is
1509// called to process the file based on the command line options.
1510void llvm::ParseInputMachO(StringRef Filename) {
1511 // Check for -arch all and verifiy the -arch flags are valid.
1512 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1513 if (ArchFlags[i] == "all") {
1514 ArchAll = true;
1515 } else {
1516 if (!MachOObjectFile::isValidArch(ArchFlags[i])) {
1517 errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] +
1518 "'for the -arch option\n";
1519 return;
1520 }
1521 }
1522 }
1523
1524 // Attempt to open the binary.
1525 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename);
1526 if (std::error_code EC = BinaryOrErr.getError()) {
1527 errs() << "llvm-objdump: '" << Filename << "': " << EC.message() << ".\n";
1528 return;
1529 }
1530 Binary &Bin = *BinaryOrErr.get().getBinary();
1531
1532 if (Archive *A = dyn_cast<Archive>(&Bin)) {
1533 outs() << "Archive : " << Filename << "\n";
1534 if (ArchiveHeaders)
1535 printArchiveHeaders(A, true, false);
1536 for (Archive::child_iterator I = A->child_begin(), E = A->child_end();
1537 I != E; ++I) {
1538 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = I->getAsBinary();
1539 if (ChildOrErr.getError())
1540 continue;
1541 if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1542 if (!checkMachOAndArchFlags(O, Filename))
1543 return;
1544 ProcessMachO(Filename, O, O->getFileName());
1545 }
1546 }
1547 return;
1548 }
1549 if (UniversalHeaders) {
1550 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin))
1551 printMachOUniversalHeaders(UB, !NonVerbose);
1552 }
1553 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) {
1554 // If we have a list of architecture flags specified dump only those.
1555 if (!ArchAll && ArchFlags.size() != 0) {
1556 // Look for a slice in the universal binary that matches each ArchFlag.
1557 bool ArchFound;
1558 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1559 ArchFound = false;
1560 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1561 E = UB->end_objects();
1562 I != E; ++I) {
1563 if (ArchFlags[i] == I->getArchTypeName()) {
1564 ArchFound = true;
1565 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
1566 I->getAsObjectFile();
1567 std::string ArchitectureName = "";
1568 if (ArchFlags.size() > 1)
1569 ArchitectureName = I->getArchTypeName();
1570 if (ObjOrErr) {
1571 ObjectFile &O = *ObjOrErr.get();
1572 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1573 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1574 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1575 I->getAsArchive()) {
1576 std::unique_ptr<Archive> &A = *AOrErr;
1577 outs() << "Archive : " << Filename;
1578 if (!ArchitectureName.empty())
1579 outs() << " (architecture " << ArchitectureName << ")";
1580 outs() << "\n";
1581 if (ArchiveHeaders)
1582 printArchiveHeaders(A.get(), true, false);
1583 for (Archive::child_iterator AI = A->child_begin(),
1584 AE = A->child_end();
1585 AI != AE; ++AI) {
1586 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1587 if (ChildOrErr.getError())
1588 continue;
1589 if (MachOObjectFile *O =
1590 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1591 ProcessMachO(Filename, O, O->getFileName(), ArchitectureName);
1592 }
1593 }
1594 }
1595 }
1596 if (!ArchFound) {
1597 errs() << "llvm-objdump: file: " + Filename + " does not contain "
1598 << "architecture: " + ArchFlags[i] + "\n";
1599 return;
1600 }
1601 }
1602 return;
1603 }
1604 // No architecture flags were specified so if this contains a slice that
1605 // matches the host architecture dump only that.
1606 if (!ArchAll) {
1607 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1608 E = UB->end_objects();
1609 I != E; ++I) {
1610 if (MachOObjectFile::getHostArch().getArchName() ==
1611 I->getArchTypeName()) {
1612 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1613 std::string ArchiveName;
1614 ArchiveName.clear();
1615 if (ObjOrErr) {
1616 ObjectFile &O = *ObjOrErr.get();
1617 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
1618 ProcessMachO(Filename, MachOOF);
1619 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1620 I->getAsArchive()) {
1621 std::unique_ptr<Archive> &A = *AOrErr;
1622 outs() << "Archive : " << Filename << "\n";
1623 if (ArchiveHeaders)
1624 printArchiveHeaders(A.get(), true, false);
1625 for (Archive::child_iterator AI = A->child_begin(),
1626 AE = A->child_end();
1627 AI != AE; ++AI) {
1628 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1629 if (ChildOrErr.getError())
1630 continue;
1631 if (MachOObjectFile *O =
1632 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
1633 ProcessMachO(Filename, O, O->getFileName());
1634 }
1635 }
1636 return;
1637 }
1638 }
1639 }
1640 // Either all architectures have been specified or none have been specified
1641 // and this does not contain the host architecture so dump all the slices.
1642 bool moreThanOneArch = UB->getNumberOfObjects() > 1;
1643 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1644 E = UB->end_objects();
1645 I != E; ++I) {
1646 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
1647 std::string ArchitectureName = "";
1648 if (moreThanOneArch)
1649 ArchitectureName = I->getArchTypeName();
1650 if (ObjOrErr) {
1651 ObjectFile &Obj = *ObjOrErr.get();
1652 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj))
1653 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
1654 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) {
1655 std::unique_ptr<Archive> &A = *AOrErr;
1656 outs() << "Archive : " << Filename;
1657 if (!ArchitectureName.empty())
1658 outs() << " (architecture " << ArchitectureName << ")";
1659 outs() << "\n";
1660 if (ArchiveHeaders)
1661 printArchiveHeaders(A.get(), true, false);
1662 for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end();
1663 AI != AE; ++AI) {
1664 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1665 if (ChildOrErr.getError())
1666 continue;
1667 if (MachOObjectFile *O =
1668 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1669 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O))
1670 ProcessMachO(Filename, MachOOF, MachOOF->getFileName(),
1671 ArchitectureName);
1672 }
1673 }
1674 }
1675 }
1676 return;
1677 }
1678 if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) {
1679 if (!checkMachOAndArchFlags(O, Filename))
1680 return;
1681 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) {
1682 ProcessMachO(Filename, MachOOF);
1683 } else
1684 errs() << "llvm-objdump: '" << Filename << "': "
1685 << "Object is not a Mach-O file type.\n";
1686 } else
1687 errs() << "llvm-objdump: '" << Filename << "': "
1688 << "Unrecognized file type.\n";
1689}
1690
1691typedef std::pair<uint64_t, const char *> BindInfoEntry;
1692typedef std::vector<BindInfoEntry> BindTable;
1693typedef BindTable::iterator bind_table_iterator;
1694
1695// The block of info used by the Symbolizer call backs.
1696struct DisassembleInfo {
1697 bool verbose;
1698 MachOObjectFile *O;
1699 SectionRef S;
1700 SymbolAddressMap *AddrMap;
1701 std::vector<SectionRef> *Sections;
1702 const char *class_name;
1703 const char *selector_name;
1704 char *method;
1705 char *demangled_name;
1706 uint64_t adrp_addr;
1707 uint32_t adrp_inst;
1708 BindTable *bindtable;
1709};
1710
1711// SymbolizerGetOpInfo() is the operand information call back function.
1712// This is called to get the symbolic information for operand(s) of an
1713// instruction when it is being done. This routine does this from
1714// the relocation information, symbol table, etc. That block of information
1715// is a pointer to the struct DisassembleInfo that was passed when the
1716// disassembler context was created and passed to back to here when
1717// called back by the disassembler for instruction operands that could have
1718// relocation information. The address of the instruction containing operand is
1719// at the Pc parameter. The immediate value the operand has is passed in
1720// op_info->Value and is at Offset past the start of the instruction and has a
1721// byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
1722// LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
1723// names and addends of the symbolic expression to add for the operand. The
1724// value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
1725// information is returned then this function returns 1 else it returns 0.
1726static int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
1727 uint64_t Size, int TagType, void *TagBuf) {
1728 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1729 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
1730 uint64_t value = op_info->Value;
1731
1732 // Make sure all fields returned are zero if we don't set them.
1733 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
1734 op_info->Value = value;
1735
1736 // If the TagType is not the value 1 which it code knows about or if no
1737 // verbose symbolic information is wanted then just return 0, indicating no
1738 // information is being returned.
1739 if (TagType != 1 || info->verbose == false)
1
Assuming 'TagType' is equal to 1
2
Taking false branch
1740 return 0;
1741
1742 unsigned int Arch = info->O->getArch();
1743 if (Arch == Triple::x86) {
3
Assuming 'Arch' is not equal to x86
4
Taking false branch
1744 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1745 return 0;
1746 // First search the section's relocation entries (if any) for an entry
1747 // for this section offset.
1748 uint32_t sect_addr = info->S.getAddress();
1749 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1750 bool reloc_found = false;
1751 DataRefImpl Rel;
1752 MachO::any_relocation_info RE;
1753 bool isExtern = false;
1754 SymbolRef Symbol;
1755 bool r_scattered = false;
1756 uint32_t r_value, pair_r_value, r_type;
1757 for (const RelocationRef &Reloc : info->S.relocations()) {
1758 uint64_t RelocOffset;
1759 Reloc.getOffset(RelocOffset);
1760 if (RelocOffset == sect_offset) {
1761 Rel = Reloc.getRawDataRefImpl();
1762 RE = info->O->getRelocation(Rel);
1763 r_type = info->O->getAnyRelocationType(RE);
1764 r_scattered = info->O->isRelocationScattered(RE);
1765 if (r_scattered) {
1766 r_value = info->O->getScatteredRelocationValue(RE);
1767 if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1768 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
1769 DataRefImpl RelNext = Rel;
1770 info->O->moveRelocationNext(RelNext);
1771 MachO::any_relocation_info RENext;
1772 RENext = info->O->getRelocation(RelNext);
1773 if (info->O->isRelocationScattered(RENext))
1774 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1775 else
1776 return 0;
1777 }
1778 } else {
1779 isExtern = info->O->getPlainRelocationExternal(RE);
1780 if (isExtern) {
1781 symbol_iterator RelocSym = Reloc.getSymbol();
1782 Symbol = *RelocSym;
1783 }
1784 }
1785 reloc_found = true;
1786 break;
1787 }
1788 }
1789 if (reloc_found && isExtern) {
1790 StringRef SymName;
1791 Symbol.getName(SymName);
1792 const char *name = SymName.data();
1793 op_info->AddSymbol.Present = 1;
1794 op_info->AddSymbol.Name = name;
1795 // For i386 extern relocation entries the value in the instruction is
1796 // the offset from the symbol, and value is already set in op_info->Value.
1797 return 1;
1798 }
1799 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1800 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
1801 const char *add = GuessSymbolName(r_value, info->AddrMap);
1802 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1803 uint32_t offset = value - (r_value - pair_r_value);
1804 op_info->AddSymbol.Present = 1;
1805 if (add != nullptr)
1806 op_info->AddSymbol.Name = add;
1807 else
1808 op_info->AddSymbol.Value = r_value;
1809 op_info->SubtractSymbol.Present = 1;
1810 if (sub != nullptr)
1811 op_info->SubtractSymbol.Name = sub;
1812 else
1813 op_info->SubtractSymbol.Value = pair_r_value;
1814 op_info->Value = offset;
1815 return 1;
1816 }
1817 // TODO:
1818 // Second search the external relocation entries of a fully linked image
1819 // (if any) for an entry that matches this segment offset.
1820 // uint32_t seg_offset = (Pc + Offset);
1821 return 0;
1822 } else if (Arch == Triple::x86_64) {
5
Assuming 'Arch' is not equal to x86_64
6
Taking false branch
1823 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1824 return 0;
1825 // First search the section's relocation entries (if any) for an entry
1826 // for this section offset.
1827 uint64_t sect_addr = info->S.getAddress();
1828 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1829 bool reloc_found = false;
1830 DataRefImpl Rel;
1831 MachO::any_relocation_info RE;
1832 bool isExtern = false;
1833 SymbolRef Symbol;
1834 for (const RelocationRef &Reloc : info->S.relocations()) {
1835 uint64_t RelocOffset;
1836 Reloc.getOffset(RelocOffset);
1837 if (RelocOffset == sect_offset) {
1838 Rel = Reloc.getRawDataRefImpl();
1839 RE = info->O->getRelocation(Rel);
1840 // NOTE: Scattered relocations don't exist on x86_64.
1841 isExtern = info->O->getPlainRelocationExternal(RE);
1842 if (isExtern) {
1843 symbol_iterator RelocSym = Reloc.getSymbol();
1844 Symbol = *RelocSym;
1845 }
1846 reloc_found = true;
1847 break;
1848 }
1849 }
1850 if (reloc_found && isExtern) {
1851 // The Value passed in will be adjusted by the Pc if the instruction
1852 // adds the Pc. But for x86_64 external relocation entries the Value
1853 // is the offset from the external symbol.
1854 if (info->O->getAnyRelocationPCRel(RE))
1855 op_info->Value -= Pc + Offset + Size;
1856 StringRef SymName;
1857 Symbol.getName(SymName);
1858 const char *name = SymName.data();
1859 unsigned Type = info->O->getAnyRelocationType(RE);
1860 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
1861 DataRefImpl RelNext = Rel;
1862 info->O->moveRelocationNext(RelNext);
1863 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1864 unsigned TypeNext = info->O->getAnyRelocationType(RENext);
1865 bool isExternNext = info->O->getPlainRelocationExternal(RENext);
1866 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
1867 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
1868 op_info->SubtractSymbol.Present = 1;
1869 op_info->SubtractSymbol.Name = name;
1870 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
1871 Symbol = *RelocSymNext;
1872 StringRef SymNameNext;
1873 Symbol.getName(SymNameNext);
1874 name = SymNameNext.data();
1875 }
1876 }
1877 // TODO: add the VariantKinds to op_info->VariantKind for relocation types
1878 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
1879 op_info->AddSymbol.Present = 1;
1880 op_info->AddSymbol.Name = name;
1881 return 1;
1882 }
1883 // TODO:
1884 // Second search the external relocation entries of a fully linked image
1885 // (if any) for an entry that matches this segment offset.
1886 // uint64_t seg_offset = (Pc + Offset);
1887 return 0;
1888 } else if (Arch == Triple::arm) {
7
Assuming 'Arch' is equal to arm
8
Taking true branch
1889 if (Offset != 0 || (Size != 4 && Size != 2))
9
Assuming 'Offset' is equal to 0
10
Assuming 'Size' is equal to 4
1890 return 0;
1891 // First search the section's relocation entries (if any) for an entry
1892 // for this section offset.
1893 uint32_t sect_addr = info->S.getAddress();
1894 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1895 bool reloc_found = false;
1896 DataRefImpl Rel;
1897 MachO::any_relocation_info RE;
1898 bool isExtern = false;
1899 SymbolRef Symbol;
1900 bool r_scattered = false;
1901 uint32_t r_value, pair_r_value, r_type, r_length, other_half;
11
'r_value' declared without an initial value
1902 for (const RelocationRef &Reloc : info->S.relocations()) {
1903 uint64_t RelocOffset;
1904 Reloc.getOffset(RelocOffset);
1905 if (RelocOffset == sect_offset) {
12
Taking true branch
1906 Rel = Reloc.getRawDataRefImpl();
1907 RE = info->O->getRelocation(Rel);
1908 r_length = info->O->getAnyRelocationLength(RE);
1909 r_scattered = info->O->isRelocationScattered(RE);
1910 if (r_scattered) {
13
Assuming 'r_scattered' is 0
14
Taking false branch
1911 r_value = info->O->getScatteredRelocationValue(RE);
1912 r_type = info->O->getScatteredRelocationType(RE);
1913 } else {
1914 r_type = info->O->getAnyRelocationType(RE);
1915 isExtern = info->O->getPlainRelocationExternal(RE);
1916 if (isExtern) {
15
Assuming 'isExtern' is 0
16
Taking false branch
1917 symbol_iterator RelocSym = Reloc.getSymbol();
1918 Symbol = *RelocSym;
1919 }
1920 }
1921 if (r_type == MachO::ARM_RELOC_HALF ||
17
Assuming 'r_type' is not equal to ARM_RELOC_HALF
21
Taking true branch
1922 r_type == MachO::ARM_RELOC_SECTDIFF ||
18
Assuming 'r_type' is not equal to ARM_RELOC_SECTDIFF
1923 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
19
Assuming 'r_type' is not equal to ARM_RELOC_LOCAL_SECTDIFF
1924 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
20
Assuming 'r_type' is equal to ARM_RELOC_HALF_SECTDIFF
1925 DataRefImpl RelNext = Rel;
1926 info->O->moveRelocationNext(RelNext);
1927 MachO::any_relocation_info RENext;
1928 RENext = info->O->getRelocation(RelNext);
1929 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
1930 if (info->O->isRelocationScattered(RENext))
22
Taking false branch
1931 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1932 }
1933 reloc_found = true;
1934 break;
23
Execution continues on line 1937
1935 }
1936 }
1937 if (reloc_found && isExtern) {
24
Taking false branch
1938 StringRef SymName;
1939 Symbol.getName(SymName);
1940 const char *name = SymName.data();
1941 op_info->AddSymbol.Present = 1;
1942 op_info->AddSymbol.Name = name;
1943 switch (r_type) {
1944 case MachO::ARM_RELOC_HALF:
1945 if ((r_length & 0x1) == 1) {
1946 op_info->Value = value << 16 | other_half;
1947 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI161;
1948 } else {
1949 op_info->Value = other_half << 16 | value;
1950 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO162;
1951 }
1952 break;
1953 default:
1954 break;
1955 }
1956 return 1;
1957 }
1958 // If we have a branch that is not an external relocation entry then
1959 // return 0 so the code in tryAddingSymbolicOperand() can use the
1960 // SymbolLookUp call back with the branch target address to look up the
1961 // symbol and possiblity add an annotation for a symbol stub.
1962 if (reloc_found && isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
25
Taking false branch
1963 r_type == MachO::ARM_THUMB_RELOC_BR22))
1964 return 0;
1965
1966 uint32_t offset = 0;
1967 if (reloc_found) {
26
Taking true branch
1968 if (r_type == MachO::ARM_RELOC_HALF ||
27
Taking true branch
1969 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1970 if ((r_length & 0x1) == 1)
28
Taking false branch
1971 value = value << 16 | other_half;
1972 else
1973 value = other_half << 16 | value;
1974 }
1975 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
1976 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
1977 offset = value - r_value;
1978 value = r_value;
1979 }
1980 }
1981
1982 if (reloc_found && r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
29
Taking true branch
1983 if ((r_length & 0x1) == 1)
30
Taking false branch
1984 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI161;
1985 else
1986 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO162;
1987 const char *add = GuessSymbolName(r_value, info->AddrMap);
31
Function call argument is an uninitialized value
1988 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
1989 int32_t offset = value - (r_value - pair_r_value);
1990 op_info->AddSymbol.Present = 1;
1991 if (add != nullptr)
1992 op_info->AddSymbol.Name = add;
1993 else
1994 op_info->AddSymbol.Value = r_value;
1995 op_info->SubtractSymbol.Present = 1;
1996 if (sub != nullptr)
1997 op_info->SubtractSymbol.Name = sub;
1998 else
1999 op_info->SubtractSymbol.Value = pair_r_value;
2000 op_info->Value = offset;
2001 return 1;
2002 }
2003
2004 if (reloc_found == false)
2005 return 0;
2006
2007 op_info->AddSymbol.Present = 1;
2008 op_info->Value = offset;
2009 if (reloc_found) {
2010 if (r_type == MachO::ARM_RELOC_HALF) {
2011 if ((r_length & 0x1) == 1)
2012 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI161;
2013 else
2014 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO162;
2015 }
2016 }
2017 const char *add = GuessSymbolName(value, info->AddrMap);
2018 if (add != nullptr) {
2019 op_info->AddSymbol.Name = add;
2020 return 1;
2021 }
2022 op_info->AddSymbol.Value = value;
2023 return 1;
2024 } else if (Arch == Triple::aarch64) {
2025 if (Offset != 0 || Size != 4)
2026 return 0;
2027 // First search the section's relocation entries (if any) for an entry
2028 // for this section offset.
2029 uint64_t sect_addr = info->S.getAddress();
2030 uint64_t sect_offset = (Pc + Offset) - sect_addr;
2031 bool reloc_found = false;
2032 DataRefImpl Rel;
2033 MachO::any_relocation_info RE;
2034 bool isExtern = false;
2035 SymbolRef Symbol;
2036 uint32_t r_type = 0;
2037 for (const RelocationRef &Reloc : info->S.relocations()) {
2038 uint64_t RelocOffset;
2039 Reloc.getOffset(RelocOffset);
2040 if (RelocOffset == sect_offset) {
2041 Rel = Reloc.getRawDataRefImpl();
2042 RE = info->O->getRelocation(Rel);
2043 r_type = info->O->getAnyRelocationType(RE);
2044 if (r_type == MachO::ARM64_RELOC_ADDEND) {
2045 DataRefImpl RelNext = Rel;
2046 info->O->moveRelocationNext(RelNext);
2047 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
2048 if (value == 0) {
2049 value = info->O->getPlainRelocationSymbolNum(RENext);
2050 op_info->Value = value;
2051 }
2052 }
2053 // NOTE: Scattered relocations don't exist on arm64.
2054 isExtern = info->O->getPlainRelocationExternal(RE);
2055 if (isExtern) {
2056 symbol_iterator RelocSym = Reloc.getSymbol();
2057 Symbol = *RelocSym;
2058 }
2059 reloc_found = true;
2060 break;
2061 }
2062 }
2063 if (reloc_found && isExtern) {
2064 StringRef SymName;
2065 Symbol.getName(SymName);
2066 const char *name = SymName.data();
2067 op_info->AddSymbol.Present = 1;
2068 op_info->AddSymbol.Name = name;
2069
2070 switch (r_type) {
2071 case MachO::ARM64_RELOC_PAGE21:
2072 /* @page */
2073 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE1;
2074 break;
2075 case MachO::ARM64_RELOC_PAGEOFF12:
2076 /* @pageoff */
2077 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF2;
2078 break;
2079 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
2080 /* @gotpage */
2081 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE3;
2082 break;
2083 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
2084 /* @gotpageoff */
2085 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF4;
2086 break;
2087 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
2088 /* @tvlppage is not implemented in llvm-mc */
2089 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP5;
2090 break;
2091 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
2092 /* @tvlppageoff is not implemented in llvm-mc */
2093 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF6;
2094 break;
2095 default:
2096 case MachO::ARM64_RELOC_BRANCH26:
2097 op_info->VariantKind = LLVMDisassembler_VariantKind_None0;
2098 break;
2099 }
2100 return 1;
2101 }
2102 return 0;
2103 } else {
2104 return 0;
2105 }
2106}
2107
2108// GuessCstringPointer is passed the address of what might be a pointer to a
2109// literal string in a cstring section. If that address is in a cstring section
2110// it returns a pointer to that string. Else it returns nullptr.
2111static const char *GuessCstringPointer(uint64_t ReferenceValue,
2112 struct DisassembleInfo *info) {
2113 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
2114 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
2115 for (unsigned I = 0;; ++I) {
2116 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
2117 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
2118 for (unsigned J = 0; J < Seg.nsects; ++J) {
2119 MachO::section_64 Sec = info->O->getSection64(Load, J);
2120 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2121 if (section_type == MachO::S_CSTRING_LITERALS &&
2122 ReferenceValue >= Sec.addr &&
2123 ReferenceValue < Sec.addr + Sec.size) {
2124 uint64_t sect_offset = ReferenceValue - Sec.addr;
2125 uint64_t object_offset = Sec.offset + sect_offset;
2126 StringRef MachOContents = info->O->getData();
2127 uint64_t object_size = MachOContents.size();
2128 const char *object_addr = (const char *)MachOContents.data();
2129 if (object_offset < object_size) {
2130 const char *name = object_addr + object_offset;
2131 return name;
2132 } else {
2133 return nullptr;
2134 }
2135 }
2136 }
2137 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
2138 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
2139 for (unsigned J = 0; J < Seg.nsects; ++J) {
2140 MachO::section Sec = info->O->getSection(Load, J);
2141 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2142 if (section_type == MachO::S_CSTRING_LITERALS &&
2143 ReferenceValue >= Sec.addr &&
2144 ReferenceValue < Sec.addr + Sec.size) {
2145 uint64_t sect_offset = ReferenceValue - Sec.addr;
2146 uint64_t object_offset = Sec.offset + sect_offset;
2147 StringRef MachOContents = info->O->getData();
2148 uint64_t object_size = MachOContents.size();
2149 const char *object_addr = (const char *)MachOContents.data();
2150 if (object_offset < object_size) {
2151 const char *name = object_addr + object_offset;
2152 return name;
2153 } else {
2154 return nullptr;
2155 }
2156 }
2157 }
2158 }
2159 if (I == LoadCommandCount - 1)
2160 break;
2161 else
2162 Load = info->O->getNextLoadCommandInfo(Load);
2163 }
2164 return nullptr;
2165}
2166
2167// GuessIndirectSymbol returns the name of the indirect symbol for the
2168// ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe
2169// an address of a symbol stub or a lazy or non-lazy pointer to associate the
2170// symbol name being referenced by the stub or pointer.
2171static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
2172 struct DisassembleInfo *info) {
2173 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
2174 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
2175 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
2176 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
2177 for (unsigned I = 0;; ++I) {
2178 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
2179 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
2180 for (unsigned J = 0; J < Seg.nsects; ++J) {
2181 MachO::section_64 Sec = info->O->getSection64(Load, J);
2182 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2183 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
2184 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
2185 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
2186 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
2187 section_type == MachO::S_SYMBOL_STUBS) &&
2188 ReferenceValue >= Sec.addr &&
2189 ReferenceValue < Sec.addr + Sec.size) {
2190 uint32_t stride;
2191 if (section_type == MachO::S_SYMBOL_STUBS)
2192 stride = Sec.reserved2;
2193 else
2194 stride = 8;
2195 if (stride == 0)
2196 return nullptr;
2197 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
2198 if (index < Dysymtab.nindirectsyms) {
2199 uint32_t indirect_symbol =
2200 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
2201 if (indirect_symbol < Symtab.nsyms) {
2202 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
2203 SymbolRef Symbol = *Sym;
2204 StringRef SymName;
2205 Symbol.getName(SymName);
2206 const char *name = SymName.data();
2207 return name;
2208 }
2209 }
2210 }
2211 }
2212 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
2213 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
2214 for (unsigned J = 0; J < Seg.nsects; ++J) {
2215 MachO::section Sec = info->O->getSection(Load, J);
2216 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
2217 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
2218 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
2219 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
2220 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
2221 section_type == MachO::S_SYMBOL_STUBS) &&
2222 ReferenceValue >= Sec.addr &&
2223 ReferenceValue < Sec.addr + Sec.size) {
2224 uint32_t stride;
2225 if (section_type == MachO::S_SYMBOL_STUBS)
2226 stride = Sec.reserved2;
2227 else
2228 stride = 4;
2229 if (stride == 0)
2230 return nullptr;
2231 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
2232 if (index < Dysymtab.nindirectsyms) {
2233 uint32_t indirect_symbol =
2234 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
2235 if (indirect_symbol < Symtab.nsyms) {
2236 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
2237 SymbolRef Symbol = *Sym;
2238 StringRef SymName;
2239 Symbol.getName(SymName);
2240 const char *name = SymName.data();
2241 return name;
2242 }
2243 }
2244 }
2245 }
2246 }
2247 if (I == LoadCommandCount - 1)
2248 break;
2249 else
2250 Load = info->O->getNextLoadCommandInfo(Load);
2251 }
2252 return nullptr;
2253}
2254
2255// method_reference() is called passing it the ReferenceName that might be
2256// a reference it to an Objective-C method call. If so then it allocates and
2257// assembles a method call string with the values last seen and saved in
2258// the DisassembleInfo's class_name and selector_name fields. This is saved
2259// into the method field of the info and any previous string is free'ed.
2260// Then the class_name field in the info is set to nullptr. The method call
2261// string is set into ReferenceName and ReferenceType is set to
2262// LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call
2263// then both ReferenceType and ReferenceName are left unchanged.
2264static void method_reference(struct DisassembleInfo *info,
2265 uint64_t *ReferenceType,
2266 const char **ReferenceName) {
2267 unsigned int Arch = info->O->getArch();
2268 if (*ReferenceName != nullptr) {
2269 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
2270 if (info->selector_name != nullptr) {
2271 if (info->method != nullptr)
2272 free(info->method);
2273 if (info->class_name != nullptr) {
2274 info->method = (char *)malloc(5 + strlen(info->class_name) +
2275 strlen(info->selector_name));
2276 if (info->method != nullptr) {
2277 strcpy(info->method, "+[");
2278 strcat(info->method, info->class_name);
2279 strcat(info->method, " ");
2280 strcat(info->method, info->selector_name);
2281 strcat(info->method, "]");
2282 *ReferenceName = info->method;
2283 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message5;
2284 }
2285 } else {
2286 info->method = (char *)malloc(9 + strlen(info->selector_name));
2287 if (info->method != nullptr) {
2288 if (Arch == Triple::x86_64)
2289 strcpy(info->method, "-[%rdi ");
2290 else if (Arch == Triple::aarch64)
2291 strcpy(info->method, "-[x0 ");
2292 else
2293 strcpy(info->method, "-[r? ");
2294 strcat(info->method, info->selector_name);
2295 strcat(info->method, "]");
2296 *ReferenceName = info->method;
2297 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message5;
2298 }
2299 }
2300 info->class_name = nullptr;
2301 }
2302 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
2303 if (info->selector_name != nullptr) {
2304 if (info->method != nullptr)
2305 free(info->method);
2306 info->method = (char *)malloc(17 + strlen(info->selector_name));
2307 if (info->method != nullptr) {
2308 if (Arch == Triple::x86_64)
2309 strcpy(info->method, "-[[%rdi super] ");
2310 else if (Arch == Triple::aarch64)
2311 strcpy(info->method, "-[[x0 super] ");
2312 else
2313 strcpy(info->method, "-[[r? super] ");
2314 strcat(info->method, info->selector_name);
2315 strcat(info->method, "]");
2316 *ReferenceName = info->method;
2317 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message5;
2318 }
2319 info->class_name = nullptr;
2320 }
2321 }
2322 }
2323}
2324
2325// GuessPointerPointer() is passed the address of what might be a pointer to
2326// a reference to an Objective-C class, selector, message ref or cfstring.
2327// If so the value of the pointer is returned and one of the booleans are set
2328// to true. If not zero is returned and all the booleans are set to false.
2329static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
2330 struct DisassembleInfo *info,
2331 bool &classref, bool &selref, bool &msgref,
2332 bool &cfstring) {
2333 classref = false;
2334 selref = false;
2335 msgref = false;
2336 cfstring = false;
2337 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
2338 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
2339 for (unsigned I = 0;; ++I) {
2340 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
2341 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
2342 for (unsigned J = 0; J < Seg.nsects; ++J) {
2343 MachO::section_64 Sec = info->O->getSection64(Load, J);
2344 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
2345 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
2346 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
2347 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
2348 strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
2349 ReferenceValue >= Sec.addr &&
2350 ReferenceValue < Sec.addr + Sec.size) {
2351 uint64_t sect_offset = ReferenceValue - Sec.addr;
2352 uint64_t object_offset = Sec.offset + sect_offset;
2353 StringRef MachOContents = info->O->getData();
2354 uint64_t object_size = MachOContents.size();
2355 const char *object_addr = (const char *)MachOContents.data();
2356 if (object_offset < object_size) {
2357 uint64_t pointer_value;
2358 memcpy(&pointer_value, object_addr + object_offset,
2359 sizeof(uint64_t));
2360 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2361 sys::swapByteOrder(pointer_value);
2362 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
2363 selref = true;
2364 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
2365 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
2366 classref = true;
2367 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
2368 ReferenceValue + 8 < Sec.addr + Sec.size) {
2369 msgref = true;
2370 memcpy(&pointer_value, object_addr + object_offset + 8,
2371 sizeof(uint64_t));
2372 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2373 sys::swapByteOrder(pointer_value);
2374 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
2375 cfstring = true;
2376 return pointer_value;
2377 } else {
2378 return 0;
2379 }
2380 }
2381 }
2382 }
2383 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
2384 if (I == LoadCommandCount - 1)
2385 break;
2386 else
2387 Load = info->O->getNextLoadCommandInfo(Load);
2388 }
2389 return 0;
2390}
2391
2392// get_pointer_64 returns a pointer to the bytes in the object file at the
2393// Address from a section in the Mach-O file. And indirectly returns the
2394// offset into the section, number of bytes left in the section past the offset
2395// and which section is was being referenced. If the Address is not in a
2396// section nullptr is returned.
2397static const char *get_pointer_64(uint64_t Address, uint32_t &offset,
2398 uint32_t &left, SectionRef &S,
2399 DisassembleInfo *info) {
2400 offset = 0;
2401 left = 0;
2402 S = SectionRef();
2403 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
2404 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
2405 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
2406 if (Address >= SectAddress && Address < SectAddress + SectSize) {
2407 S = (*(info->Sections))[SectIdx];
2408 offset = Address - SectAddress;
2409 left = SectSize - offset;
2410 StringRef SectContents;
2411 ((*(info->Sections))[SectIdx]).getContents(SectContents);
2412 return SectContents.data() + offset;
2413 }
2414 }
2415 return nullptr;
2416}
2417
2418// get_symbol_64() returns the name of a symbol (or nullptr) and the address of
2419// the symbol indirectly through n_value. Based on the relocation information
2420// for the specified section offset in the specified section reference.
2421static const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
2422 DisassembleInfo *info, uint64_t &n_value) {
2423 n_value = 0;
2424 if (info->verbose == false)
2425 return nullptr;
2426
2427 // See if there is an external relocation entry at the sect_offset.
2428 bool reloc_found = false;
2429 DataRefImpl Rel;
2430 MachO::any_relocation_info RE;
2431 bool isExtern = false;
2432 SymbolRef Symbol;
2433 for (const RelocationRef &Reloc : S.relocations()) {
2434 uint64_t RelocOffset;
2435 Reloc.getOffset(RelocOffset);
2436 if (RelocOffset == sect_offset) {
2437 Rel = Reloc.getRawDataRefImpl();
2438 RE = info->O->getRelocation(Rel);
2439 if (info->O->isRelocationScattered(RE))
2440 continue;
2441 isExtern = info->O->getPlainRelocationExternal(RE);
2442 if (isExtern) {
2443 symbol_iterator RelocSym = Reloc.getSymbol();
2444 Symbol = *RelocSym;
2445 }
2446 reloc_found = true;
2447 break;
2448 }
2449 }
2450 // If there is an external relocation entry for a symbol in this section
2451 // at this section_offset then use that symbol's value for the n_value
2452 // and return its name.
2453 const char *SymbolName = nullptr;
2454 if (reloc_found && isExtern) {
2455 Symbol.getAddress(n_value);
2456 StringRef name;
2457 Symbol.getName(name);
2458 if (!name.empty()) {
2459 SymbolName = name.data();
2460 return SymbolName;
2461 }
2462 }
2463
2464 // TODO: For fully linked images, look through the external relocation
2465 // entries off the dynamic symtab command. For these the r_offset is from the
2466 // start of the first writeable segment in the Mach-O file. So the offset
2467 // to this section from that segment is passed to this routine by the caller,
2468 // as the database_offset. Which is the difference of the section's starting
2469 // address and the first writable segment.
2470 //
2471 // NOTE: need add passing the database_offset to this routine.
2472
2473 // TODO: We did not find an external relocation entry so look up the
2474 // ReferenceValue as an address of a symbol and if found return that symbol's
2475 // name.
2476 //
2477 // NOTE: need add passing the ReferenceValue to this routine. Then that code
2478 // would simply be this:
2479 // SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2480
2481 return SymbolName;
2482}
2483
2484// These are structs in the Objective-C meta data and read to produce the
2485// comments for disassembly. While these are part of the ABI they are no
2486// public defintions. So the are here not in include/llvm/Support/MachO.h .
2487
2488// The cfstring object in a 64-bit Mach-O file.
2489struct cfstring64_t {
2490 uint64_t isa; // class64_t * (64-bit pointer)
2491 uint64_t flags; // flag bits
2492 uint64_t characters; // char * (64-bit pointer)
2493 uint64_t length; // number of non-NULL characters in above
2494};
2495
2496// The class object in a 64-bit Mach-O file.
2497struct class64_t {
2498 uint64_t isa; // class64_t * (64-bit pointer)
2499 uint64_t superclass; // class64_t * (64-bit pointer)
2500 uint64_t cache; // Cache (64-bit pointer)
2501 uint64_t vtable; // IMP * (64-bit pointer)
2502 uint64_t data; // class_ro64_t * (64-bit pointer)
2503};
2504
2505struct class_ro64_t {
2506 uint32_t flags;
2507 uint32_t instanceStart;
2508 uint32_t instanceSize;
2509 uint32_t reserved;
2510 uint64_t ivarLayout; // const uint8_t * (64-bit pointer)
2511 uint64_t name; // const char * (64-bit pointer)
2512 uint64_t baseMethods; // const method_list_t * (64-bit pointer)
2513 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer)
2514 uint64_t ivars; // const ivar_list_t * (64-bit pointer)
2515 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
2516 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
2517};
2518
2519inline void swapStruct(struct cfstring64_t &cfs) {
2520 sys::swapByteOrder(cfs.isa);
2521 sys::swapByteOrder(cfs.flags);
2522 sys::swapByteOrder(cfs.characters);
2523 sys::swapByteOrder(cfs.length);
2524}
2525
2526inline void swapStruct(struct class64_t &c) {
2527 sys::swapByteOrder(c.isa);
2528 sys::swapByteOrder(c.superclass);
2529 sys::swapByteOrder(c.cache);
2530 sys::swapByteOrder(c.vtable);
2531 sys::swapByteOrder(c.data);
2532}
2533
2534inline void swapStruct(struct class_ro64_t &cro) {
2535 sys::swapByteOrder(cro.flags);
2536 sys::swapByteOrder(cro.instanceStart);
2537 sys::swapByteOrder(cro.instanceSize);
2538 sys::swapByteOrder(cro.reserved);
2539 sys::swapByteOrder(cro.ivarLayout);
2540 sys::swapByteOrder(cro.name);
2541 sys::swapByteOrder(cro.baseMethods);
2542 sys::swapByteOrder(cro.baseProtocols);
2543 sys::swapByteOrder(cro.ivars);
2544 sys::swapByteOrder(cro.weakIvarLayout);
2545 sys::swapByteOrder(cro.baseProperties);
2546}
2547
2548static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
2549 struct DisassembleInfo *info);
2550
2551// get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
2552// to an Objective-C class and returns the class name. It is also passed the
2553// address of the pointer, so when the pointer is zero as it can be in an .o
2554// file, that is used to look for an external relocation entry with a symbol
2555// name.
2556static const char *get_objc2_64bit_class_name(uint64_t pointer_value,
2557 uint64_t ReferenceValue,
2558 struct DisassembleInfo *info) {
2559 const char *r;
2560 uint32_t offset, left;
2561 SectionRef S;
2562
2563 // The pointer_value can be 0 in an object file and have a relocation
2564 // entry for the class symbol at the ReferenceValue (the address of the
2565 // pointer).
2566 if (pointer_value == 0) {
2567 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2568 if (r == nullptr || left < sizeof(uint64_t))
2569 return nullptr;
2570 uint64_t n_value;
2571 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2572 if (symbol_name == nullptr)
2573 return nullptr;
2574 const char *class_name = strrchr(symbol_name, '$');
2575 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
2576 return class_name + 2;
2577 else
2578 return nullptr;
2579 }
2580
2581 // The case were the pointer_value is non-zero and points to a class defined
2582 // in this Mach-O file.
2583 r = get_pointer_64(pointer_value, offset, left, S, info);
2584 if (r == nullptr || left < sizeof(struct class64_t))
2585 return nullptr;
2586 struct class64_t c;
2587 memcpy(&c, r, sizeof(struct class64_t));
2588 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2589 swapStruct(c);
2590 if (c.data == 0)
2591 return nullptr;
2592 r = get_pointer_64(c.data, offset, left, S, info);
2593 if (r == nullptr || left < sizeof(struct class_ro64_t))
2594 return nullptr;
2595 struct class_ro64_t cro;
2596 memcpy(&cro, r, sizeof(struct class_ro64_t));
2597 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2598 swapStruct(cro);
2599 if (cro.name == 0)
2600 return nullptr;
2601 const char *name = get_pointer_64(cro.name, offset, left, S, info);
2602 return name;
2603}
2604
2605// get_objc2_64bit_cfstring_name is used for disassembly and is passed a
2606// pointer to a cfstring and returns its name or nullptr.
2607static const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
2608 struct DisassembleInfo *info) {
2609 const char *r, *name;
2610 uint32_t offset, left;
2611 SectionRef S;
2612 struct cfstring64_t cfs;
2613 uint64_t cfs_characters;
2614
2615 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2616 if (r == nullptr || left < sizeof(struct cfstring64_t))
2617 return nullptr;
2618 memcpy(&cfs, r, sizeof(struct cfstring64_t));
2619 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2620 swapStruct(cfs);
2621 if (cfs.characters == 0) {
2622 uint64_t n_value;
2623 const char *symbol_name = get_symbol_64(
2624 offset + offsetof(struct cfstring64_t, characters)__builtin_offsetof(struct cfstring64_t, characters), S, info, n_value);
2625 if (symbol_name == nullptr)
2626 return nullptr;
2627 cfs_characters = n_value;
2628 } else
2629 cfs_characters = cfs.characters;
2630 name = get_pointer_64(cfs_characters, offset, left, S, info);
2631
2632 return name;
2633}
2634
2635// get_objc2_64bit_selref() is used for disassembly and is passed a the address
2636// of a pointer to an Objective-C selector reference when the pointer value is
2637// zero as in a .o file and is likely to have a external relocation entry with
2638// who's symbol's n_value is the real pointer to the selector name. If that is
2639// the case the real pointer to the selector name is returned else 0 is
2640// returned
2641static uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
2642 struct DisassembleInfo *info) {
2643 uint32_t offset, left;
2644 SectionRef S;
2645
2646 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
2647 if (r == nullptr || left < sizeof(uint64_t))
2648 return 0;
2649 uint64_t n_value;
2650 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2651 if (symbol_name == nullptr)
2652 return 0;
2653 return n_value;
2654}
2655
2656// GuessLiteralPointer returns a string which for the item in the Mach-O file
2657// for the address passed in as ReferenceValue for printing as a comment with
2658// the instruction and also returns the corresponding type of that item
2659// indirectly through ReferenceType.
2660//
2661// If ReferenceValue is an address of literal cstring then a pointer to the
2662// cstring is returned and ReferenceType is set to
2663// LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
2664//
2665// If ReferenceValue is an address of an Objective-C CFString, Selector ref or
2666// Class ref that name is returned and the ReferenceType is set accordingly.
2667//
2668// Lastly, literals which are Symbol address in a literal pool are looked for
2669// and if found the symbol name is returned and ReferenceType is set to
2670// LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
2671//
2672// If there is no item in the Mach-O file for the address passed in as
2673// ReferenceValue nullptr is returned and ReferenceType is unchanged.
2674static const char *GuessLiteralPointer(uint64_t ReferenceValue,
2675 uint64_t ReferencePC,
2676 uint64_t *ReferenceType,
2677 struct DisassembleInfo *info) {
2678 // First see if there is an external relocation entry at the ReferencePC.
2679 uint64_t sect_addr = info->S.getAddress();
2680 uint64_t sect_offset = ReferencePC - sect_addr;
2681 bool reloc_found = false;
2682 DataRefImpl Rel;
2683 MachO::any_relocation_info RE;
2684 bool isExtern = false;
2685 SymbolRef Symbol;
2686 for (const RelocationRef &Reloc : info->S.relocations()) {
2687 uint64_t RelocOffset;
2688 Reloc.getOffset(RelocOffset);
2689 if (RelocOffset == sect_offset) {
2690 Rel = Reloc.getRawDataRefImpl();
2691 RE = info->O->getRelocation(Rel);
2692 if (info->O->isRelocationScattered(RE))
2693 continue;
2694 isExtern = info->O->getPlainRelocationExternal(RE);
2695 if (isExtern) {
2696 symbol_iterator RelocSym = Reloc.getSymbol();
2697 Symbol = *RelocSym;
2698 }
2699 reloc_found = true;
2700 break;
2701 }
2702 }
2703 // If there is an external relocation entry for a symbol in a section
2704 // then used that symbol's value for the value of the reference.
2705 if (reloc_found && isExtern) {
2706 if (info->O->getAnyRelocationPCRel(RE)) {
2707 unsigned Type = info->O->getAnyRelocationType(RE);
2708 if (Type == MachO::X86_64_RELOC_SIGNED) {
2709 Symbol.getAddress(ReferenceValue);
2710 }
2711 }
2712 }
2713
2714 // Look for literals such as Objective-C CFStrings refs, Selector refs,
2715 // Message refs and Class refs.
2716 bool classref, selref, msgref, cfstring;
2717 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
2718 selref, msgref, cfstring);
2719 if (classref == true && pointer_value == 0) {
2720 // Note the ReferenceValue is a pointer into the __objc_classrefs section.
2721 // And the pointer_value in that section is typically zero as it will be
2722 // set by dyld as part of the "bind information".
2723 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
2724 if (name != nullptr) {
2725 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref8;
2726 const char *class_name = strrchr(name, '$');
2727 if (class_name != nullptr && class_name[1] == '_' &&
2728 class_name[2] != '\0') {
2729 info->class_name = class_name + 2;
2730 return name;
2731 }
2732 }
2733 }
2734
2735 if (classref == true) {
2736 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref8;
2737 const char *name =
2738 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
2739 if (name != nullptr)
2740 info->class_name = name;
2741 else
2742 name = "bad class ref";
2743 return name;
2744 }
2745
2746 if (cfstring == true) {
2747 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref4;
2748 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
2749 return name;
2750 }
2751
2752 if (selref == true && pointer_value == 0)
2753 pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
2754
2755 if (pointer_value != 0)
2756 ReferenceValue = pointer_value;
2757
2758 const char *name = GuessCstringPointer(ReferenceValue, info);
2759 if (name) {
2760 if (pointer_value != 0 && selref == true) {
2761 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref7;
2762 info->selector_name = name;
2763 } else if (pointer_value != 0 && msgref == true) {
2764 info->class_name = nullptr;
2765 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref6;
2766 info->selector_name = name;
2767 } else
2768 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr3;
2769 return name;
2770 }
2771
2772 // Lastly look for an indirect symbol with this ReferenceValue which is in
2773 // a literal pool. If found return that symbol name.
2774 name = GuessIndirectSymbol(ReferenceValue, info);
2775 if (name) {
2776 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr2;
2777 return name;
2778 }
2779
2780 return nullptr;
2781}
2782
2783// SymbolizerSymbolLookUp is the symbol lookup function passed when creating
2784// the Symbolizer. It looks up the ReferenceValue using the info passed via the
2785// pointer to the struct DisassembleInfo that was passed when MCSymbolizer
2786// is created and returns the symbol name that matches the ReferenceValue or
2787// nullptr if none. The ReferenceType is passed in for the IN type of
2788// reference the instruction is making from the values in defined in the header
2789// "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific
2790// Out type and the ReferenceName will also be set which is added as a comment
2791// to the disassembled instruction.
2792//
2793#if HAVE_CXXABI_H1
2794// If the symbol name is a C++ mangled name then the demangled name is
2795// returned through ReferenceName and ReferenceType is set to
2796// LLVMDisassembler_ReferenceType_DeMangled_Name .
2797#endif
2798//
2799// When this is called to get a symbol name for a branch target then the
2800// ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
2801// SymbolValue will be looked for in the indirect symbol table to determine if
2802// it is an address for a symbol stub. If so then the symbol name for that
2803// stub is returned indirectly through ReferenceName and then ReferenceType is
2804// set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
2805//
2806// When this is called with an value loaded via a PC relative load then
2807// ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
2808// SymbolValue is checked to be an address of literal pointer, symbol pointer,
2809// or an Objective-C meta data reference. If so the output ReferenceType is
2810// set to correspond to that as well as setting the ReferenceName.
2811static const char *SymbolizerSymbolLookUp(void *DisInfo,
2812 uint64_t ReferenceValue,
2813 uint64_t *ReferenceType,
2814 uint64_t ReferencePC,
2815 const char **ReferenceName) {
2816 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
2817 // If no verbose symbolic information is wanted then just return nullptr.
2818 if (info->verbose == false) {
2819 *ReferenceName = nullptr;
2820 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2821 return nullptr;
2822 }
2823
2824 const char *SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
2825
2826 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch1) {
2827 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
2828 if (*ReferenceName != nullptr) {
2829 method_reference(info, ReferenceType, ReferenceName);
2830 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message5)
2831 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub1;
2832 } else
2833#if HAVE_CXXABI_H1
2834 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2835 if (info->demangled_name != nullptr)
2836 free(info->demangled_name);
2837 int status;
2838 info->demangled_name =
2839 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2840 if (info->demangled_name != nullptr) {
2841 *ReferenceName = info->demangled_name;
2842 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name9;
2843 } else
2844 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2845 } else
2846#endif
2847 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2848 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load2) {
2849 *ReferenceName =
2850 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2851 if (*ReferenceName)
2852 method_reference(info, ReferenceType, ReferenceName);
2853 else
2854 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2855 // If this is arm64 and the reference is an adrp instruction save the
2856 // instruction, passed in ReferenceValue and the address of the instruction
2857 // for use later if we see and add immediate instruction.
2858 } else if (info->O->getArch() == Triple::aarch64 &&
2859 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP0x100000001) {
2860 info->adrp_inst = ReferenceValue;
2861 info->adrp_addr = ReferencePC;
2862 SymbolName = nullptr;
2863 *ReferenceName = nullptr;
2864 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2865 // If this is arm64 and reference is an add immediate instruction and we
2866 // have
2867 // seen an adrp instruction just before it and the adrp's Xd register
2868 // matches
2869 // this add's Xn register reconstruct the value being referenced and look to
2870 // see if it is a literal pointer. Note the add immediate instruction is
2871 // passed in ReferenceValue.
2872 } else if (info->O->getArch() == Triple::aarch64 &&
2873 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri0x100000002 &&
2874 ReferencePC - 4 == info->adrp_addr &&
2875 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2876 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2877 uint32_t addxri_inst;
2878 uint64_t adrp_imm, addxri_imm;
2879
2880 adrp_imm =
2881 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2882 if (info->adrp_inst & 0x0200000)
2883 adrp_imm |= 0xfffffffffc000000LL;
2884
2885 addxri_inst = ReferenceValue;
2886 addxri_imm = (addxri_inst >> 10) & 0xfff;
2887 if (((addxri_inst >> 22) & 0x3) == 1)
2888 addxri_imm <<= 12;
2889
2890 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2891 (adrp_imm << 12) + addxri_imm;
2892
2893 *ReferenceName =
2894 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2895 if (*ReferenceName == nullptr)
2896 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2897 // If this is arm64 and the reference is a load register instruction and we
2898 // have seen an adrp instruction just before it and the adrp's Xd register
2899 // matches this add's Xn register reconstruct the value being referenced and
2900 // look to see if it is a literal pointer. Note the load register
2901 // instruction is passed in ReferenceValue.
2902 } else if (info->O->getArch() == Triple::aarch64 &&
2903 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui0x100000003 &&
2904 ReferencePC - 4 == info->adrp_addr &&
2905 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2906 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2907 uint32_t ldrxui_inst;
2908 uint64_t adrp_imm, ldrxui_imm;
2909
2910 adrp_imm =
2911 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2912 if (info->adrp_inst & 0x0200000)
2913 adrp_imm |= 0xfffffffffc000000LL;
2914
2915 ldrxui_inst = ReferenceValue;
2916 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
2917
2918 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2919 (adrp_imm << 12) + (ldrxui_imm << 3);
2920
2921 *ReferenceName =
2922 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2923 if (*ReferenceName == nullptr)
2924 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2925 }
2926 // If this arm64 and is an load register (PC-relative) instruction the
2927 // ReferenceValue is the PC plus the immediate value.
2928 else if (info->O->getArch() == Triple::aarch64 &&
2929 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl0x100000004 ||
2930 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR0x100000005)) {
2931 *ReferenceName =
2932 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2933 if (*ReferenceName == nullptr)
2934 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2935 }
2936#if HAVE_CXXABI_H1
2937 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2938 if (info->demangled_name != nullptr)
2939 free(info->demangled_name);
2940 int status;
2941 info->demangled_name =
2942 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
2943 if (info->demangled_name != nullptr) {
2944 *ReferenceName = info->demangled_name;
2945 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name9;
2946 }
2947 }
2948#endif
2949 else {
2950 *ReferenceName = nullptr;
2951 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None0;
2952 }
2953
2954 return SymbolName;
2955}
2956
2957/// \brief Emits the comments that are stored in the CommentStream.
2958/// Each comment in the CommentStream must end with a newline.
2959static void emitComments(raw_svector_ostream &CommentStream,
2960 SmallString<128> &CommentsToEmit,
2961 formatted_raw_ostream &FormattedOS,
2962 const MCAsmInfo &MAI) {
2963 // Flush the stream before taking its content.
2964 CommentStream.flush();
2965 StringRef Comments = CommentsToEmit.str();
2966 // Get the default information for printing a comment.
2967 const char *CommentBegin = MAI.getCommentString();
2968 unsigned CommentColumn = MAI.getCommentColumn();
2969 bool IsFirst = true;
2970 while (!Comments.empty()) {
2971 if (!IsFirst)
2972 FormattedOS << '\n';
2973 // Emit a line of comments.
2974 FormattedOS.PadToColumn(CommentColumn);
2975 size_t Position = Comments.find('\n');
2976 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
2977 // Move after the newline character.
2978 Comments = Comments.substr(Position + 1);
2979 IsFirst = false;
2980 }
2981 FormattedOS.flush();
2982
2983 // Tell the comment stream that the vector changed underneath it.
2984 CommentsToEmit.clear();
2985 CommentStream.resync();
2986}
2987
2988static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF,
2989 StringRef DisSegName, StringRef DisSectName) {
2990 const char *McpuDefault = nullptr;
2991 const Target *ThumbTarget = nullptr;
2992 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
2993 if (!TheTarget) {
2994 // GetTarget prints out stuff.
2995 return;
2996 }
2997 if (MCPU.empty() && McpuDefault)
2998 MCPU = McpuDefault;
2999
3000 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
3001 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
3002 if (ThumbTarget)
3003 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
3004
3005 // Package up features to be passed to target/subtarget
3006 std::string FeaturesStr;
3007 if (MAttrs.size()) {
3008 SubtargetFeatures Features;
3009 for (unsigned i = 0; i != MAttrs.size(); ++i)
3010 Features.AddFeature(MAttrs[i]);
3011 FeaturesStr = Features.getString();
3012 }
3013
3014 // Set up disassembler.
3015 std::unique_ptr<const MCRegisterInfo> MRI(
3016 TheTarget->createMCRegInfo(TripleName));
3017 std::unique_ptr<const MCAsmInfo> AsmInfo(
3018 TheTarget->createMCAsmInfo(*MRI, TripleName));
3019 std::unique_ptr<const MCSubtargetInfo> STI(
3020 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
3021 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
3022 std::unique_ptr<MCDisassembler> DisAsm(
3023 TheTarget->createMCDisassembler(*STI, Ctx));
3024 std::unique_ptr<MCSymbolizer> Symbolizer;
3025 struct DisassembleInfo SymbolizerInfo;
3026 std::unique_ptr<MCRelocationInfo> RelInfo(
3027 TheTarget->createMCRelocationInfo(TripleName, Ctx));
3028 if (RelInfo) {
3029 Symbolizer.reset(TheTarget->createMCSymbolizer(
3030 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
3031 &SymbolizerInfo, &Ctx, std::move(RelInfo)));
3032 DisAsm->setSymbolizer(std::move(Symbolizer));
3033 }
3034 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
3035 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
3036 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
3037 // Set the display preference for hex vs. decimal immediates.
3038 IP->setPrintImmHex(PrintImmHex);
3039 // Comment stream and backing vector.
3040 SmallString<128> CommentsToEmit;
3041 raw_svector_ostream CommentStream(CommentsToEmit);
3042 // FIXME: Setting the CommentStream in the InstPrinter is problematic in that
3043 // if it is done then arm64 comments for string literals don't get printed
3044 // and some constant get printed instead and not setting it causes intel
3045 // (32-bit and 64-bit) comments printed with different spacing before the
3046 // comment causing different diffs with the 'C' disassembler library API.
3047 // IP->setCommentStream(CommentStream);
3048
3049 if (!AsmInfo || !STI || !DisAsm || !IP) {
3050 errs() << "error: couldn't initialize disassembler for target "
3051 << TripleName << '\n';
3052 return;
3053 }
3054
3055 // Set up thumb disassembler.
3056 std::unique_ptr<const MCRegisterInfo> ThumbMRI;
3057 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
3058 std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
3059 std::unique_ptr<MCDisassembler> ThumbDisAsm;
3060 std::unique_ptr<MCInstPrinter> ThumbIP;
3061 std::unique_ptr<MCContext> ThumbCtx;
3062 std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
3063 struct DisassembleInfo ThumbSymbolizerInfo;
3064 std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
3065 if (ThumbTarget) {
3066 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
3067 ThumbAsmInfo.reset(
3068 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
3069 ThumbSTI.reset(
3070 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
3071 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
3072 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
3073 MCContext *PtrThumbCtx = ThumbCtx.get();
3074 ThumbRelInfo.reset(
3075 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
3076 if (ThumbRelInfo) {
3077 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
3078 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
3079 &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo)));
3080 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
3081 }
3082 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
3083 ThumbIP.reset(ThumbTarget->createMCInstPrinter(
3084 ThumbAsmPrinterVariant, *ThumbAsmInfo, *ThumbInstrInfo, *ThumbMRI,
3085 *ThumbSTI));
3086 // Set the display preference for hex vs. decimal immediates.
3087 ThumbIP->setPrintImmHex(PrintImmHex);
3088 }
3089
3090 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
3091 errs() << "error: couldn't initialize disassembler for target "
3092 << ThumbTripleName << '\n';
3093 return;
3094 }
3095
3096 MachO::mach_header Header = MachOOF->getHeader();
3097
3098 // FIXME: Using the -cfg command line option, this code used to be able to
3099 // annotate relocations with the referenced symbol's name, and if this was
3100 // inside a __[cf]string section, the data it points to. This is now replaced
3101 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
3102 std::vector<SectionRef> Sections;
3103 std::vector<SymbolRef> Symbols;
3104 SmallVector<uint64_t, 8> FoundFns;
3105 uint64_t BaseSegmentAddress;
3106
3107 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
3108 BaseSegmentAddress);
3109
3110 // Sort the symbols by address, just in case they didn't come in that way.
3111 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
3112
3113 // Build a data in code table that is sorted on by the address of each entry.
3114 uint64_t BaseAddress = 0;
3115 if (Header.filetype == MachO::MH_OBJECT)
3116 BaseAddress = Sections[0].getAddress();
3117 else
3118 BaseAddress = BaseSegmentAddress;
3119 DiceTable Dices;
3120 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
3121 DI != DE; ++DI) {
3122 uint32_t Offset;
3123 DI->getOffset(Offset);
3124 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
3125 }
3126 array_pod_sort(Dices.begin(), Dices.end());
3127
3128#ifndef NDEBUG
3129 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
3130#else
3131 raw_ostream &DebugOut = nulls();
3132#endif
3133
3134 std::unique_ptr<DIContext> diContext;
3135 ObjectFile *DbgObj = MachOOF;
3136 // Try to find debug info and set up the DIContext for it.
3137 if (UseDbg) {
3138 // A separate DSym file path was specified, parse it as a macho file,
3139 // get the sections and supply it to the section name parsing machinery.
3140 if (!DSYMFile.empty()) {
3141 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
3142 MemoryBuffer::getFileOrSTDIN(DSYMFile);
3143 if (std::error_code EC = BufOrErr.getError()) {
3144 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
3145 return;
3146 }
3147 DbgObj =
3148 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
3149 .get()
3150 .release();
3151 }
3152
3153 // Setup the DIContext
3154 diContext.reset(DIContext::getDWARFContext(*DbgObj));
3155 }
3156
3157 if (DumpSections.size() == 0)
3158 outs() << "(" << DisSegName << "," << DisSectName << ") section\n";
3159
3160 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
3161 StringRef SectName;
3162 if (Sections[SectIdx].getName(SectName) || SectName != DisSectName)
3163 continue;
3164
3165 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
3166
3167 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
3168 if (SegmentName != DisSegName)
3169 continue;
3170
3171 StringRef BytesStr;
3172 Sections[SectIdx].getContents(BytesStr);
3173 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
3174 BytesStr.size());
3175 uint64_t SectAddress = Sections[SectIdx].getAddress();
3176
3177 bool symbolTableWorked = false;
3178
3179 // Parse relocations.
3180 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
3181 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
3182 uint64_t RelocOffset;
3183 Reloc.getOffset(RelocOffset);
3184 uint64_t SectionAddress = Sections[SectIdx].getAddress();
3185 RelocOffset -= SectionAddress;
3186
3187 symbol_iterator RelocSym = Reloc.getSymbol();
3188
3189 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
3190 }
3191 array_pod_sort(Relocs.begin(), Relocs.end());
3192
3193 // Create a map of symbol addresses to symbol names for use by
3194 // the SymbolizerSymbolLookUp() routine.
3195 SymbolAddressMap AddrMap;
3196 bool DisSymNameFound = false;
3197 for (const SymbolRef &Symbol : MachOOF->symbols()) {
3198 SymbolRef::Type ST;
3199 Symbol.getType(ST);
3200 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
3201 ST == SymbolRef::ST_Other) {
3202 uint64_t Address;
3203 Symbol.getAddress(Address);
3204 StringRef SymName;
3205 Symbol.getName(SymName);
3206 AddrMap[Address] = SymName;
3207 if (!DisSymName.empty() && DisSymName == SymName)
3208 DisSymNameFound = true;
3209 }
3210 }
3211 if (!DisSymName.empty() && DisSymNameFound == false) {
3212 outs() << "Can't find -dis-symname: " << DisSymName << "\n";
3213 return;
3214 }
3215 // Set up the block of info used by the Symbolizer call backs.
3216 SymbolizerInfo.verbose = !NoSymbolicOperands;
3217 SymbolizerInfo.O = MachOOF;
3218 SymbolizerInfo.S = Sections[SectIdx];
3219 SymbolizerInfo.AddrMap = &AddrMap;
3220 SymbolizerInfo.Sections = &Sections;
3221 SymbolizerInfo.class_name = nullptr;
3222 SymbolizerInfo.selector_name = nullptr;
3223 SymbolizerInfo.method = nullptr;
3224 SymbolizerInfo.demangled_name = nullptr;
3225 SymbolizerInfo.bindtable = nullptr;
3226 SymbolizerInfo.adrp_addr = 0;
3227 SymbolizerInfo.adrp_inst = 0;
3228 // Same for the ThumbSymbolizer
3229 ThumbSymbolizerInfo.verbose = !NoSymbolicOperands;
3230 ThumbSymbolizerInfo.O = MachOOF;
3231 ThumbSymbolizerInfo.S = Sections[SectIdx];
3232 ThumbSymbolizerInfo.AddrMap = &AddrMap;
3233 ThumbSymbolizerInfo.Sections = &Sections;
3234 ThumbSymbolizerInfo.class_name = nullptr;
3235 ThumbSymbolizerInfo.selector_name = nullptr;
3236 ThumbSymbolizerInfo.method = nullptr;
3237 ThumbSymbolizerInfo.demangled_name = nullptr;
3238 ThumbSymbolizerInfo.bindtable = nullptr;
3239 ThumbSymbolizerInfo.adrp_addr = 0;
3240 ThumbSymbolizerInfo.adrp_inst = 0;
3241
3242 // Disassemble symbol by symbol.
3243 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
3244 StringRef SymName;
3245 Symbols[SymIdx].getName(SymName);
3246
3247 SymbolRef::Type ST;
3248 Symbols[SymIdx].getType(ST);
3249 if (ST != SymbolRef::ST_Function)
3250 continue;
3251
3252 // Make sure the symbol is defined in this section.
3253 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
3254 if (!containsSym)
3255 continue;
3256
3257 // If we are only disassembling one symbol see if this is that symbol.
3258 if (!DisSymName.empty() && DisSymName != SymName)
3259 continue;
3260
3261 // Start at the address of the symbol relative to the section's address.
3262 uint64_t Start = 0;
3263 uint64_t SectionAddress = Sections[SectIdx].getAddress();
3264 Symbols[SymIdx].getAddress(Start);
3265 Start -= SectionAddress;
3266
3267 // Stop disassembling either at the beginning of the next symbol or at
3268 // the end of the section.
3269 bool containsNextSym = false;
3270 uint64_t NextSym = 0;
3271 uint64_t NextSymIdx = SymIdx + 1;
3272 while (Symbols.size() > NextSymIdx) {
3273 SymbolRef::Type NextSymType;
3274 Symbols[NextSymIdx].getType(NextSymType);
3275 if (NextSymType == SymbolRef::ST_Function) {
3276 containsNextSym =
3277 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
3278 Symbols[NextSymIdx].getAddress(NextSym);
3279 NextSym -= SectionAddress;
3280 break;
3281 }
3282 ++NextSymIdx;
3283 }
3284
3285 uint64_t SectSize = Sections[SectIdx].getSize();
3286 uint64_t End = containsNextSym ? NextSym : SectSize;
3287 uint64_t Size;
3288
3289 symbolTableWorked = true;
3290
3291 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
3292 bool isThumb =
3293 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
3294
3295 outs() << SymName << ":\n";
3296 DILineInfo lastLine;
3297 for (uint64_t Index = Start; Index < End; Index += Size) {
3298 MCInst Inst;
3299
3300 uint64_t PC = SectAddress + Index;
3301 if (!NoLeadingAddr) {
3302 if (FullLeadingAddr) {
3303 if (MachOOF->is64Bit())
3304 outs() << format("%016" PRIx64"l" "x", PC);
3305 else
3306 outs() << format("%08" PRIx64"l" "x", PC);
3307 } else {
3308 outs() << format("%8" PRIx64"l" "x" ":", PC);
3309 }
3310 }
3311 if (!NoShowRawInsn)
3312 outs() << "\t";
3313
3314 // Check the data in code table here to see if this is data not an
3315 // instruction to be disassembled.
3316 DiceTable Dice;
3317 Dice.push_back(std::make_pair(PC, DiceRef()));
3318 dice_table_iterator DTI =
3319 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
3320 compareDiceTableEntries);
3321 if (DTI != Dices.end()) {
3322 uint16_t Length;
3323 DTI->second.getLength(Length);
3324 uint16_t Kind;
3325 DTI->second.getKind(Kind);
3326 Size = DumpDataInCode(Bytes.data() + Index, Length, Kind);
3327 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
3328 (PC == (DTI->first + Length - 1)) && (Length & 1))
3329 Size++;
3330 continue;
3331 }
3332
3333 SmallVector<char, 64> AnnotationsBytes;
3334 raw_svector_ostream Annotations(AnnotationsBytes);
3335
3336 bool gotInst;
3337 if (isThumb)
3338 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
3339 PC, DebugOut, Annotations);
3340 else
3341 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
3342 DebugOut, Annotations);
3343 if (gotInst) {
3344 if (!NoShowRawInsn) {
3345 DumpBytes(ArrayRef<uint8_t>(Bytes.data() + Index, Size));
3346 }
3347 formatted_raw_ostream FormattedOS(outs());
3348 Annotations.flush();
3349 StringRef AnnotationsStr = Annotations.str();
3350 if (isThumb)
3351 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr);
3352 else
3353 IP->printInst(&Inst, FormattedOS, AnnotationsStr);
3354 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
3355
3356 // Print debug info.
3357 if (diContext) {
3358 DILineInfo dli = diContext->getLineInfoForAddress(PC);
3359 // Print valid line info if it changed.
3360 if (dli != lastLine && dli.Line != 0)
3361 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
3362 << dli.Column;
3363 lastLine = dli;
3364 }
3365 outs() << "\n";
3366 } else {
3367 unsigned int Arch = MachOOF->getArch();
3368 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
3369 outs() << format("\t.byte 0x%02x #bad opcode\n",
3370 *(Bytes.data() + Index) & 0xff);
3371 Size = 1; // skip exactly one illegible byte and move on.
3372 } else if (Arch == Triple::aarch64) {
3373 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
3374 (*(Bytes.data() + Index + 1) & 0xff) << 8 |
3375 (*(Bytes.data() + Index + 2) & 0xff) << 16 |
3376 (*(Bytes.data() + Index + 3) & 0xff) << 24;
3377 outs() << format("\t.long\t0x%08x\n", opcode);
3378 Size = 4;
3379 } else {
3380 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
3381 if (Size == 0)
3382 Size = 1; // skip illegible bytes
3383 }
3384 }
3385 }
3386 }
3387 if (!symbolTableWorked) {
3388 // Reading the symbol table didn't work, disassemble the whole section.
3389 uint64_t SectAddress = Sections[SectIdx].getAddress();
3390 uint64_t SectSize = Sections[SectIdx].getSize();
3391 uint64_t InstSize;
3392 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
3393 MCInst Inst;
3394
3395 uint64_t PC = SectAddress + Index;
3396 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
3397 DebugOut, nulls())) {
3398 if (!NoLeadingAddr) {
3399 if (FullLeadingAddr) {
3400 if (MachOOF->is64Bit())
3401 outs() << format("%016" PRIx64"l" "x", PC);
3402 else
3403 outs() << format("%08" PRIx64"l" "x", PC);
3404 } else {
3405 outs() << format("%8" PRIx64"l" "x" ":", PC);
3406 }
3407 }
3408 if (!NoShowRawInsn) {
3409 outs() << "\t";
3410 DumpBytes(ArrayRef<uint8_t>(Bytes.data() + Index, InstSize));
3411 }
3412 IP->printInst(&Inst, outs(), "");
3413 outs() << "\n";
3414 } else {
3415 unsigned int Arch = MachOOF->getArch();
3416 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
3417 outs() << format("\t.byte 0x%02x #bad opcode\n",
3418 *(Bytes.data() + Index) & 0xff);
3419 InstSize = 1; // skip exactly one illegible byte and move on.
3420 } else {
3421 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
3422 if (InstSize == 0)
3423 InstSize = 1; // skip illegible bytes
3424 }
3425 }
3426 }
3427 }
3428 // The TripleName's need to be reset if we are called again for a different
3429 // archtecture.
3430 TripleName = "";
3431 ThumbTripleName = "";
3432
3433 if (SymbolizerInfo.method != nullptr)
3434 free(SymbolizerInfo.method);
3435 if (SymbolizerInfo.demangled_name != nullptr)
3436 free(SymbolizerInfo.demangled_name);
3437 if (SymbolizerInfo.bindtable != nullptr)
3438 delete SymbolizerInfo.bindtable;
3439 if (ThumbSymbolizerInfo.method != nullptr)
3440 free(ThumbSymbolizerInfo.method);
3441 if (ThumbSymbolizerInfo.demangled_name != nullptr)
3442 free(ThumbSymbolizerInfo.demangled_name);
3443 if (ThumbSymbolizerInfo.bindtable != nullptr)
3444 delete ThumbSymbolizerInfo.bindtable;
3445 }
3446}
3447
3448//===----------------------------------------------------------------------===//
3449// __compact_unwind section dumping
3450//===----------------------------------------------------------------------===//
3451
3452namespace {
3453
3454template <typename T> static uint64_t readNext(const char *&Buf) {
3455 using llvm::support::little;
3456 using llvm::support::unaligned;
3457
3458 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
3459 Buf += sizeof(T);
3460 return Val;
3461}
3462
3463struct CompactUnwindEntry {
3464 uint32_t OffsetInSection;
3465
3466 uint64_t FunctionAddr;
3467 uint32_t Length;
3468 uint32_t CompactEncoding;
3469 uint64_t PersonalityAddr;
3470 uint64_t LSDAAddr;
3471
3472 RelocationRef FunctionReloc;
3473 RelocationRef PersonalityReloc;
3474 RelocationRef LSDAReloc;
3475
3476 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
3477 : OffsetInSection(Offset) {
3478 if (Is64)
3479 read<uint64_t>(Contents.data() + Offset);
3480 else
3481 read<uint32_t>(Contents.data() + Offset);
3482 }
3483
3484private:
3485 template <typename UIntPtr> void read(const char *Buf) {
3486 FunctionAddr = readNext<UIntPtr>(Buf);
3487 Length = readNext<uint32_t>(Buf);
3488 CompactEncoding = readNext<uint32_t>(Buf);
3489 PersonalityAddr = readNext<UIntPtr>(Buf);
3490 LSDAAddr = readNext<UIntPtr>(Buf);
3491 }
3492};
3493}
3494
3495/// Given a relocation from __compact_unwind, consisting of the RelocationRef
3496/// and data being relocated, determine the best base Name and Addend to use for
3497/// display purposes.
3498///
3499/// 1. An Extern relocation will directly reference a symbol (and the data is
3500/// then already an addend), so use that.
3501/// 2. Otherwise the data is an offset in the object file's layout; try to find
3502// a symbol before it in the same section, and use the offset from there.
3503/// 3. Finally, if all that fails, fall back to an offset from the start of the
3504/// referenced section.
3505static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
3506 std::map<uint64_t, SymbolRef> &Symbols,
3507 const RelocationRef &Reloc, uint64_t Addr,
3508 StringRef &Name, uint64_t &Addend) {
3509 if (Reloc.getSymbol() != Obj->symbol_end()) {
3510 Reloc.getSymbol()->getName(Name);
3511 Addend = Addr;
3512 return;
3513 }
3514
3515 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
3516 SectionRef RelocSection = Obj->getRelocationSection(RE);
3517
3518 uint64_t SectionAddr = RelocSection.getAddress();
3519
3520 auto Sym = Symbols.upper_bound(Addr);
3521 if (Sym == Symbols.begin()) {
3522 // The first symbol in the object is after this reference, the best we can
3523 // do is section-relative notation.
3524 RelocSection.getName(Name);
3525 Addend = Addr - SectionAddr;
3526 return;
3527 }
3528
3529 // Go back one so that SymbolAddress <= Addr.
3530 --Sym;
3531
3532 section_iterator SymSection = Obj->section_end();
3533 Sym->second.getSection(SymSection);
3534 if (RelocSection == *SymSection) {
3535 // There's a valid symbol in the same section before this reference.
3536 Sym->second.getName(Name);
3537 Addend = Addr - Sym->first;
3538 return;
3539 }
3540
3541 // There is a symbol before this reference, but it's in a different
3542 // section. Probably not helpful to mention it, so use the section name.
3543 RelocSection.getName(Name);
3544 Addend = Addr - SectionAddr;
3545}
3546
3547static void printUnwindRelocDest(const MachOObjectFile *Obj,
3548 std::map<uint64_t, SymbolRef> &Symbols,
3549 const RelocationRef &Reloc, uint64_t Addr) {
3550 StringRef Name;
3551 uint64_t Addend;
3552
3553 if (!Reloc.getObjectFile())
3554 return;
3555
3556 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
3557
3558 outs() << Name;
3559 if (Addend)
3560 outs() << " + " << format("0x%" PRIx64"l" "x", Addend);
3561}
3562
3563static void
3564printMachOCompactUnwindSection(const MachOObjectFile *Obj,
3565 std::map<uint64_t, SymbolRef> &Symbols,
3566 const SectionRef &CompactUnwind) {
3567
3568 assert(Obj->isLittleEndian() &&((Obj->isLittleEndian() && "There should not be a big-endian .o with __compact_unwind"
) ? static_cast<void> (0) : __assert_fail ("Obj->isLittleEndian() && \"There should not be a big-endian .o with __compact_unwind\""
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3569, __PRETTY_FUNCTION__))
3569 "There should not be a big-endian .o with __compact_unwind")((Obj->isLittleEndian() && "There should not be a big-endian .o with __compact_unwind"
) ? static_cast<void> (0) : __assert_fail ("Obj->isLittleEndian() && \"There should not be a big-endian .o with __compact_unwind\""
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3569, __PRETTY_FUNCTION__))
;
3570
3571 bool Is64 = Obj->is64Bit();
3572 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
3573 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
3574
3575 StringRef Contents;
3576 CompactUnwind.getContents(Contents);
3577
3578 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
3579
3580 // First populate the initial raw offsets, encodings and so on from the entry.
3581 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
3582 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
3583 CompactUnwinds.push_back(Entry);
3584 }
3585
3586 // Next we need to look at the relocations to find out what objects are
3587 // actually being referred to.
3588 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
3589 uint64_t RelocAddress;
3590 Reloc.getOffset(RelocAddress);
3591
3592 uint32_t EntryIdx = RelocAddress / EntrySize;
3593 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
3594 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
3595
3596 if (OffsetInEntry == 0)
3597 Entry.FunctionReloc = Reloc;
3598 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
3599 Entry.PersonalityReloc = Reloc;
3600 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
3601 Entry.LSDAReloc = Reloc;
3602 else
3603 llvm_unreachable("Unexpected relocation in __compact_unwind section")::llvm::llvm_unreachable_internal("Unexpected relocation in __compact_unwind section"
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3603)
;
3604 }
3605
3606 // Finally, we're ready to print the data we've gathered.
3607 outs() << "Contents of __compact_unwind section:\n";
3608 for (auto &Entry : CompactUnwinds) {
3609 outs() << " Entry at offset "
3610 << format("0x%" PRIx32"x", Entry.OffsetInSection) << ":\n";
3611
3612 // 1. Start of the region this entry applies to.
3613 outs() << " start: " << format("0x%" PRIx64"l" "x",
3614 Entry.FunctionAddr) << ' ';
3615 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
3616 outs() << '\n';
3617
3618 // 2. Length of the region this entry applies to.
3619 outs() << " length: " << format("0x%" PRIx32"x", Entry.Length)
3620 << '\n';
3621 // 3. The 32-bit compact encoding.
3622 outs() << " compact encoding: "
3623 << format("0x%08" PRIx32"x", Entry.CompactEncoding) << '\n';
3624
3625 // 4. The personality function, if present.
3626 if (Entry.PersonalityReloc.getObjectFile()) {
3627 outs() << " personality function: "
3628 << format("0x%" PRIx64"l" "x", Entry.PersonalityAddr) << ' ';
3629 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
3630 Entry.PersonalityAddr);
3631 outs() << '\n';
3632 }
3633
3634 // 5. This entry's language-specific data area.
3635 if (Entry.LSDAReloc.getObjectFile()) {
3636 outs() << " LSDA: " << format("0x%" PRIx64"l" "x",
3637 Entry.LSDAAddr) << ' ';
3638 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
3639 outs() << '\n';
3640 }
3641 }
3642}
3643
3644//===----------------------------------------------------------------------===//
3645// __unwind_info section dumping
3646//===----------------------------------------------------------------------===//
3647
3648static void printRegularSecondLevelUnwindPage(const char *PageStart) {
3649 const char *Pos = PageStart;
3650 uint32_t Kind = readNext<uint32_t>(Pos);
3651 (void)Kind;
3652 assert(Kind == 2 && "kind for a regular 2nd level index should be 2")((Kind == 2 && "kind for a regular 2nd level index should be 2"
) ? static_cast<void> (0) : __assert_fail ("Kind == 2 && \"kind for a regular 2nd level index should be 2\""
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3652, __PRETTY_FUNCTION__))
;
3653
3654 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3655 uint16_t NumEntries = readNext<uint16_t>(Pos);
3656
3657 Pos = PageStart + EntriesStart;
3658 for (unsigned i = 0; i < NumEntries; ++i) {
3659 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3660 uint32_t Encoding = readNext<uint32_t>(Pos);
3661
3662 outs() << " [" << i << "]: "
3663 << "function offset=" << format("0x%08" PRIx32"x", FunctionOffset)
3664 << ", "
3665 << "encoding=" << format("0x%08" PRIx32"x", Encoding) << '\n';
3666 }
3667}
3668
3669static void printCompressedSecondLevelUnwindPage(
3670 const char *PageStart, uint32_t FunctionBase,
3671 const SmallVectorImpl<uint32_t> &CommonEncodings) {
3672 const char *Pos = PageStart;
3673 uint32_t Kind = readNext<uint32_t>(Pos);
3674 (void)Kind;
3675 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3")((Kind == 3 && "kind for a compressed 2nd level index should be 3"
) ? static_cast<void> (0) : __assert_fail ("Kind == 3 && \"kind for a compressed 2nd level index should be 3\""
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3675, __PRETTY_FUNCTION__))
;
3676
3677 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3678 uint16_t NumEntries = readNext<uint16_t>(Pos);
3679
3680 uint16_t EncodingsStart = readNext<uint16_t>(Pos);
3681 readNext<uint16_t>(Pos);
3682 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
3683 PageStart + EncodingsStart);
3684
3685 Pos = PageStart + EntriesStart;
3686 for (unsigned i = 0; i < NumEntries; ++i) {
3687 uint32_t Entry = readNext<uint32_t>(Pos);
3688 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
3689 uint32_t EncodingIdx = Entry >> 24;
3690
3691 uint32_t Encoding;
3692 if (EncodingIdx < CommonEncodings.size())
3693 Encoding = CommonEncodings[EncodingIdx];
3694 else
3695 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
3696
3697 outs() << " [" << i << "]: "
3698 << "function offset=" << format("0x%08" PRIx32"x", FunctionOffset)
3699 << ", "
3700 << "encoding[" << EncodingIdx
3701 << "]=" << format("0x%08" PRIx32"x", Encoding) << '\n';
3702 }
3703}
3704
3705static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
3706 std::map<uint64_t, SymbolRef> &Symbols,
3707 const SectionRef &UnwindInfo) {
3708
3709 assert(Obj->isLittleEndian() &&((Obj->isLittleEndian() && "There should not be a big-endian .o with __unwind_info"
) ? static_cast<void> (0) : __assert_fail ("Obj->isLittleEndian() && \"There should not be a big-endian .o with __unwind_info\""
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3710, __PRETTY_FUNCTION__))
3710 "There should not be a big-endian .o with __unwind_info")((Obj->isLittleEndian() && "There should not be a big-endian .o with __unwind_info"
) ? static_cast<void> (0) : __assert_fail ("Obj->isLittleEndian() && \"There should not be a big-endian .o with __unwind_info\""
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3710, __PRETTY_FUNCTION__))
;
3711
3712 outs() << "Contents of __unwind_info section:\n";
3713
3714 StringRef Contents;
3715 UnwindInfo.getContents(Contents);
3716 const char *Pos = Contents.data();
3717
3718 //===----------------------------------
3719 // Section header
3720 //===----------------------------------
3721
3722 uint32_t Version = readNext<uint32_t>(Pos);
3723 outs() << " Version: "
3724 << format("0x%" PRIx32"x", Version) << '\n';
3725 assert(Version == 1 && "only understand version 1")((Version == 1 && "only understand version 1") ? static_cast
<void> (0) : __assert_fail ("Version == 1 && \"only understand version 1\""
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3725, __PRETTY_FUNCTION__))
;
3726
3727 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
3728 outs() << " Common encodings array section offset: "
3729 << format("0x%" PRIx32"x", CommonEncodingsStart) << '\n';
3730 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
3731 outs() << " Number of common encodings in array: "
3732 << format("0x%" PRIx32"x", NumCommonEncodings) << '\n';
3733
3734 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
3735 outs() << " Personality function array section offset: "
3736 << format("0x%" PRIx32"x", PersonalitiesStart) << '\n';
3737 uint32_t NumPersonalities = readNext<uint32_t>(Pos);
3738 outs() << " Number of personality functions in array: "
3739 << format("0x%" PRIx32"x", NumPersonalities) << '\n';
3740
3741 uint32_t IndicesStart = readNext<uint32_t>(Pos);
3742 outs() << " Index array section offset: "
3743 << format("0x%" PRIx32"x", IndicesStart) << '\n';
3744 uint32_t NumIndices = readNext<uint32_t>(Pos);
3745 outs() << " Number of indices in array: "
3746 << format("0x%" PRIx32"x", NumIndices) << '\n';
3747
3748 //===----------------------------------
3749 // A shared list of common encodings
3750 //===----------------------------------
3751
3752 // These occupy indices in the range [0, N] whenever an encoding is referenced
3753 // from a compressed 2nd level index table. In practice the linker only
3754 // creates ~128 of these, so that indices are available to embed encodings in
3755 // the 2nd level index.
3756
3757 SmallVector<uint32_t, 64> CommonEncodings;
3758 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n";
3759 Pos = Contents.data() + CommonEncodingsStart;
3760 for (unsigned i = 0; i < NumCommonEncodings; ++i) {
3761 uint32_t Encoding = readNext<uint32_t>(Pos);
3762 CommonEncodings.push_back(Encoding);
3763
3764 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32"x", Encoding)
3765 << '\n';
3766 }
3767
3768 //===----------------------------------
3769 // Personality functions used in this executable
3770 //===----------------------------------
3771
3772 // There should be only a handful of these (one per source language,
3773 // roughly). Particularly since they only get 2 bits in the compact encoding.
3774
3775 outs() << " Personality functions: (count = " << NumPersonalities << ")\n";
3776 Pos = Contents.data() + PersonalitiesStart;
3777 for (unsigned i = 0; i < NumPersonalities; ++i) {
3778 uint32_t PersonalityFn = readNext<uint32_t>(Pos);
3779 outs() << " personality[" << i + 1
3780 << "]: " << format("0x%08" PRIx32"x", PersonalityFn) << '\n';
3781 }
3782
3783 //===----------------------------------
3784 // The level 1 index entries
3785 //===----------------------------------
3786
3787 // These specify an approximate place to start searching for the more detailed
3788 // information, sorted by PC.
3789
3790 struct IndexEntry {
3791 uint32_t FunctionOffset;
3792 uint32_t SecondLevelPageStart;
3793 uint32_t LSDAStart;
3794 };
3795
3796 SmallVector<IndexEntry, 4> IndexEntries;
3797
3798 outs() << " Top level indices: (count = " << NumIndices << ")\n";
3799 Pos = Contents.data() + IndicesStart;
3800 for (unsigned i = 0; i < NumIndices; ++i) {
3801 IndexEntry Entry;
3802
3803 Entry.FunctionOffset = readNext<uint32_t>(Pos);
3804 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
3805 Entry.LSDAStart = readNext<uint32_t>(Pos);
3806 IndexEntries.push_back(Entry);
3807
3808 outs() << " [" << i << "]: "
3809 << "function offset=" << format("0x%08" PRIx32"x", Entry.FunctionOffset)
3810 << ", "
3811 << "2nd level page offset="
3812 << format("0x%08" PRIx32"x", Entry.SecondLevelPageStart) << ", "
3813 << "LSDA offset=" << format("0x%08" PRIx32"x", Entry.LSDAStart) << '\n';
3814 }
3815
3816 //===----------------------------------
3817 // Next come the LSDA tables
3818 //===----------------------------------
3819
3820 // The LSDA layout is rather implicit: it's a contiguous array of entries from
3821 // the first top-level index's LSDAOffset to the last (sentinel).
3822
3823 outs() << " LSDA descriptors:\n";
3824 Pos = Contents.data() + IndexEntries[0].LSDAStart;
3825 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
3826 (2 * sizeof(uint32_t));
3827 for (int i = 0; i < NumLSDAs; ++i) {
3828 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3829 uint32_t LSDAOffset = readNext<uint32_t>(Pos);
3830 outs() << " [" << i << "]: "
3831 << "function offset=" << format("0x%08" PRIx32"x", FunctionOffset)
3832 << ", "
3833 << "LSDA offset=" << format("0x%08" PRIx32"x", LSDAOffset) << '\n';
3834 }
3835
3836 //===----------------------------------
3837 // Finally, the 2nd level indices
3838 //===----------------------------------
3839
3840 // Generally these are 4K in size, and have 2 possible forms:
3841 // + Regular stores up to 511 entries with disparate encodings
3842 // + Compressed stores up to 1021 entries if few enough compact encoding
3843 // values are used.
3844 outs() << " Second level indices:\n";
3845 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
3846 // The final sentinel top-level index has no associated 2nd level page
3847 if (IndexEntries[i].SecondLevelPageStart == 0)
3848 break;
3849
3850 outs() << " Second level index[" << i << "]: "
3851 << "offset in section="
3852 << format("0x%08" PRIx32"x", IndexEntries[i].SecondLevelPageStart)
3853 << ", "
3854 << "base function offset="
3855 << format("0x%08" PRIx32"x", IndexEntries[i].FunctionOffset) << '\n';
3856
3857 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
3858 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
3859 if (Kind == 2)
3860 printRegularSecondLevelUnwindPage(Pos);
3861 else if (Kind == 3)
3862 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
3863 CommonEncodings);
3864 else
3865 llvm_unreachable("Do not know how to print this kind of 2nd level page")::llvm::llvm_unreachable_internal("Do not know how to print this kind of 2nd level page"
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 3865)
;
3866 }
3867}
3868
3869void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
3870 std::map<uint64_t, SymbolRef> Symbols;
3871 for (const SymbolRef &SymRef : Obj->symbols()) {
3872 // Discard any undefined or absolute symbols. They're not going to take part
3873 // in the convenience lookup for unwind info and just take up resources.
3874 section_iterator Section = Obj->section_end();
3875 SymRef.getSection(Section);
3876 if (Section == Obj->section_end())
3877 continue;
3878
3879 uint64_t Addr;
3880 SymRef.getAddress(Addr);
3881 Symbols.insert(std::make_pair(Addr, SymRef));
3882 }
3883
3884 for (const SectionRef &Section : Obj->sections()) {
3885 StringRef SectName;
3886 Section.getName(SectName);
3887 if (SectName == "__compact_unwind")
3888 printMachOCompactUnwindSection(Obj, Symbols, Section);
3889 else if (SectName == "__unwind_info")
3890 printMachOUnwindInfoSection(Obj, Symbols, Section);
3891 else if (SectName == "__eh_frame")
3892 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
3893 }
3894}
3895
3896static void PrintMachHeader(uint32_t magic, uint32_t cputype,
3897 uint32_t cpusubtype, uint32_t filetype,
3898 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
3899 bool verbose) {
3900 outs() << "Mach header\n";
3901 outs() << " magic cputype cpusubtype caps filetype ncmds "
3902 "sizeofcmds flags\n";
3903 if (verbose) {
3904 if (magic == MachO::MH_MAGIC)
3905 outs() << " MH_MAGIC";
3906 else if (magic == MachO::MH_MAGIC_64)
3907 outs() << "MH_MAGIC_64";
3908 else
3909 outs() << format(" 0x%08" PRIx32"x", magic);
3910 switch (cputype) {
3911 case MachO::CPU_TYPE_I386:
3912 outs() << " I386";
3913 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3914 case MachO::CPU_SUBTYPE_I386_ALL:
3915 outs() << " ALL";
3916 break;
3917 default:
3918 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3919 break;
3920 }
3921 break;
3922 case MachO::CPU_TYPE_X86_64:
3923 outs() << " X86_64";
3924 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3925 case MachO::CPU_SUBTYPE_X86_64_ALL:
3926 outs() << " ALL";
3927 break;
3928 case MachO::CPU_SUBTYPE_X86_64_H:
3929 outs() << " Haswell";
3930 break;
3931 default:
3932 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3933 break;
3934 }
3935 break;
3936 case MachO::CPU_TYPE_ARM:
3937 outs() << " ARM";
3938 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3939 case MachO::CPU_SUBTYPE_ARM_ALL:
3940 outs() << " ALL";
3941 break;
3942 case MachO::CPU_SUBTYPE_ARM_V4T:
3943 outs() << " V4T";
3944 break;
3945 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
3946 outs() << " V5TEJ";
3947 break;
3948 case MachO::CPU_SUBTYPE_ARM_XSCALE:
3949 outs() << " XSCALE";
3950 break;
3951 case MachO::CPU_SUBTYPE_ARM_V6:
3952 outs() << " V6";
3953 break;
3954 case MachO::CPU_SUBTYPE_ARM_V6M:
3955 outs() << " V6M";
3956 break;
3957 case MachO::CPU_SUBTYPE_ARM_V7:
3958 outs() << " V7";
3959 break;
3960 case MachO::CPU_SUBTYPE_ARM_V7EM:
3961 outs() << " V7EM";
3962 break;
3963 case MachO::CPU_SUBTYPE_ARM_V7K:
3964 outs() << " V7K";
3965 break;
3966 case MachO::CPU_SUBTYPE_ARM_V7M:
3967 outs() << " V7M";
3968 break;
3969 case MachO::CPU_SUBTYPE_ARM_V7S:
3970 outs() << " V7S";
3971 break;
3972 default:
3973 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3974 break;
3975 }
3976 break;
3977 case MachO::CPU_TYPE_ARM64:
3978 outs() << " ARM64";
3979 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3980 case MachO::CPU_SUBTYPE_ARM64_ALL:
3981 outs() << " ALL";
3982 break;
3983 default:
3984 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3985 break;
3986 }
3987 break;
3988 case MachO::CPU_TYPE_POWERPC:
3989 outs() << " PPC";
3990 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3991 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3992 outs() << " ALL";
3993 break;
3994 default:
3995 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3996 break;
3997 }
3998 break;
3999 case MachO::CPU_TYPE_POWERPC64:
4000 outs() << " PPC64";
4001 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
4002 case MachO::CPU_SUBTYPE_POWERPC_ALL:
4003 outs() << " ALL";
4004 break;
4005 default:
4006 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
4007 break;
4008 }
4009 break;
4010 }
4011 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
4012 outs() << " LIB64";
4013 } else {
4014 outs() << format(" 0x%02" PRIx32"x",
4015 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
4016 }
4017 switch (filetype) {
4018 case MachO::MH_OBJECT:
4019 outs() << " OBJECT";
4020 break;
4021 case MachO::MH_EXECUTE:
4022 outs() << " EXECUTE";
4023 break;
4024 case MachO::MH_FVMLIB:
4025 outs() << " FVMLIB";
4026 break;
4027 case MachO::MH_CORE:
4028 outs() << " CORE";
4029 break;
4030 case MachO::MH_PRELOAD:
4031 outs() << " PRELOAD";
4032 break;
4033 case MachO::MH_DYLIB:
4034 outs() << " DYLIB";
4035 break;
4036 case MachO::MH_DYLIB_STUB:
4037 outs() << " DYLIB_STUB";
4038 break;
4039 case MachO::MH_DYLINKER:
4040 outs() << " DYLINKER";
4041 break;
4042 case MachO::MH_BUNDLE:
4043 outs() << " BUNDLE";
4044 break;
4045 case MachO::MH_DSYM:
4046 outs() << " DSYM";
4047 break;
4048 case MachO::MH_KEXT_BUNDLE:
4049 outs() << " KEXTBUNDLE";
4050 break;
4051 default:
4052 outs() << format(" %10u", filetype);
4053 break;
4054 }
4055 outs() << format(" %5u", ncmds);
4056 outs() << format(" %10u", sizeofcmds);
4057 uint32_t f = flags;
4058 if (f & MachO::MH_NOUNDEFS) {
4059 outs() << " NOUNDEFS";
4060 f &= ~MachO::MH_NOUNDEFS;
4061 }
4062 if (f & MachO::MH_INCRLINK) {
4063 outs() << " INCRLINK";
4064 f &= ~MachO::MH_INCRLINK;
4065 }
4066 if (f & MachO::MH_DYLDLINK) {
4067 outs() << " DYLDLINK";
4068 f &= ~MachO::MH_DYLDLINK;
4069 }
4070 if (f & MachO::MH_BINDATLOAD) {
4071 outs() << " BINDATLOAD";
4072 f &= ~MachO::MH_BINDATLOAD;
4073 }
4074 if (f & MachO::MH_PREBOUND) {
4075 outs() << " PREBOUND";
4076 f &= ~MachO::MH_PREBOUND;
4077 }
4078 if (f & MachO::MH_SPLIT_SEGS) {
4079 outs() << " SPLIT_SEGS";
4080 f &= ~MachO::MH_SPLIT_SEGS;
4081 }
4082 if (f & MachO::MH_LAZY_INIT) {
4083 outs() << " LAZY_INIT";
4084 f &= ~MachO::MH_LAZY_INIT;
4085 }
4086 if (f & MachO::MH_TWOLEVEL) {
4087 outs() << " TWOLEVEL";
4088 f &= ~MachO::MH_TWOLEVEL;
4089 }
4090 if (f & MachO::MH_FORCE_FLAT) {
4091 outs() << " FORCE_FLAT";
4092 f &= ~MachO::MH_FORCE_FLAT;
4093 }
4094 if (f & MachO::MH_NOMULTIDEFS) {
4095 outs() << " NOMULTIDEFS";
4096 f &= ~MachO::MH_NOMULTIDEFS;
4097 }
4098 if (f & MachO::MH_NOFIXPREBINDING) {
4099 outs() << " NOFIXPREBINDING";
4100 f &= ~MachO::MH_NOFIXPREBINDING;
4101 }
4102 if (f & MachO::MH_PREBINDABLE) {
4103 outs() << " PREBINDABLE";
4104 f &= ~MachO::MH_PREBINDABLE;
4105 }
4106 if (f & MachO::MH_ALLMODSBOUND) {
4107 outs() << " ALLMODSBOUND";
4108 f &= ~MachO::MH_ALLMODSBOUND;
4109 }
4110 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
4111 outs() << " SUBSECTIONS_VIA_SYMBOLS";
4112 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
4113 }
4114 if (f & MachO::MH_CANONICAL) {
4115 outs() << " CANONICAL";
4116 f &= ~MachO::MH_CANONICAL;
4117 }
4118 if (f & MachO::MH_WEAK_DEFINES) {
4119 outs() << " WEAK_DEFINES";
4120 f &= ~MachO::MH_WEAK_DEFINES;
4121 }
4122 if (f & MachO::MH_BINDS_TO_WEAK) {
4123 outs() << " BINDS_TO_WEAK";
4124 f &= ~MachO::MH_BINDS_TO_WEAK;
4125 }
4126 if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
4127 outs() << " ALLOW_STACK_EXECUTION";
4128 f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
4129 }
4130 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
4131 outs() << " DEAD_STRIPPABLE_DYLIB";
4132 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
4133 }
4134 if (f & MachO::MH_PIE) {
4135 outs() << " PIE";
4136 f &= ~MachO::MH_PIE;
4137 }
4138 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
4139 outs() << " NO_REEXPORTED_DYLIBS";
4140 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
4141 }
4142 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
4143 outs() << " MH_HAS_TLV_DESCRIPTORS";
4144 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
4145 }
4146 if (f & MachO::MH_NO_HEAP_EXECUTION) {
4147 outs() << " MH_NO_HEAP_EXECUTION";
4148 f &= ~MachO::MH_NO_HEAP_EXECUTION;
4149 }
4150 if (f & MachO::MH_APP_EXTENSION_SAFE) {
4151 outs() << " APP_EXTENSION_SAFE";
4152 f &= ~MachO::MH_APP_EXTENSION_SAFE;
4153 }
4154 if (f != 0 || flags == 0)
4155 outs() << format(" 0x%08" PRIx32"x", f);
4156 } else {
4157 outs() << format(" 0x%08" PRIx32"x", magic);
4158 outs() << format(" %7d", cputype);
4159 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
4160 outs() << format(" 0x%02" PRIx32"x",
4161 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
4162 outs() << format(" %10u", filetype);
4163 outs() << format(" %5u", ncmds);
4164 outs() << format(" %10u", sizeofcmds);
4165 outs() << format(" 0x%08" PRIx32"x", flags);
4166 }
4167 outs() << "\n";
4168}
4169
4170static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
4171 StringRef SegName, uint64_t vmaddr,
4172 uint64_t vmsize, uint64_t fileoff,
4173 uint64_t filesize, uint32_t maxprot,
4174 uint32_t initprot, uint32_t nsects,
4175 uint32_t flags, uint32_t object_size,
4176 bool verbose) {
4177 uint64_t expected_cmdsize;
4178 if (cmd == MachO::LC_SEGMENT) {
4179 outs() << " cmd LC_SEGMENT\n";
4180 expected_cmdsize = nsects;
4181 expected_cmdsize *= sizeof(struct MachO::section);
4182 expected_cmdsize += sizeof(struct MachO::segment_command);
4183 } else {
4184 outs() << " cmd LC_SEGMENT_64\n";
4185 expected_cmdsize = nsects;
4186 expected_cmdsize *= sizeof(struct MachO::section_64);
4187 expected_cmdsize += sizeof(struct MachO::segment_command_64);
4188 }
4189 outs() << " cmdsize " << cmdsize;
4190 if (cmdsize != expected_cmdsize)
4191 outs() << " Inconsistent size\n";
4192 else
4193 outs() << "\n";
4194 outs() << " segname " << SegName << "\n";
4195 if (cmd == MachO::LC_SEGMENT_64) {
4196 outs() << " vmaddr " << format("0x%016" PRIx64"l" "x", vmaddr) << "\n";
4197 outs() << " vmsize " << format("0x%016" PRIx64"l" "x", vmsize) << "\n";
4198 } else {
4199 outs() << " vmaddr " << format("0x%08" PRIx64"l" "x", vmaddr) << "\n";
4200 outs() << " vmsize " << format("0x%08" PRIx64"l" "x", vmsize) << "\n";
4201 }
4202 outs() << " fileoff " << fileoff;
4203 if (fileoff > object_size)
4204 outs() << " (past end of file)\n";
4205 else
4206 outs() << "\n";
4207 outs() << " filesize " << filesize;
4208 if (fileoff + filesize > object_size)
4209 outs() << " (past end of file)\n";
4210 else
4211 outs() << "\n";
4212 if (verbose) {
4213 if ((maxprot &
4214 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
4215 MachO::VM_PROT_EXECUTE)) != 0)
4216 outs() << " maxprot ?" << format("0x%08" PRIx32"x", maxprot) << "\n";
4217 else {
4218 if (maxprot & MachO::VM_PROT_READ)
4219 outs() << " maxprot r";
4220 else
4221 outs() << " maxprot -";
4222 if (maxprot & MachO::VM_PROT_WRITE)
4223 outs() << "w";
4224 else
4225 outs() << "-";
4226 if (maxprot & MachO::VM_PROT_EXECUTE)
4227 outs() << "x\n";
4228 else
4229 outs() << "-\n";
4230 }
4231 if ((initprot &
4232 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
4233 MachO::VM_PROT_EXECUTE)) != 0)
4234 outs() << " initprot ?" << format("0x%08" PRIx32"x", initprot) << "\n";
4235 else {
4236 if (initprot & MachO::VM_PROT_READ)
4237 outs() << " initprot r";
4238 else
4239 outs() << " initprot -";
4240 if (initprot & MachO::VM_PROT_WRITE)
4241 outs() << "w";
4242 else
4243 outs() << "-";
4244 if (initprot & MachO::VM_PROT_EXECUTE)
4245 outs() << "x\n";
4246 else
4247 outs() << "-\n";
4248 }
4249 } else {
4250 outs() << " maxprot " << format("0x%08" PRIx32"x", maxprot) << "\n";
4251 outs() << " initprot " << format("0x%08" PRIx32"x", initprot) << "\n";
4252 }
4253 outs() << " nsects " << nsects << "\n";
4254 if (verbose) {
4255 outs() << " flags";
4256 if (flags == 0)
4257 outs() << " (none)\n";
4258 else {
4259 if (flags & MachO::SG_HIGHVM) {
4260 outs() << " HIGHVM";
4261 flags &= ~MachO::SG_HIGHVM;
4262 }
4263 if (flags & MachO::SG_FVMLIB) {
4264 outs() << " FVMLIB";
4265 flags &= ~MachO::SG_FVMLIB;
4266 }
4267 if (flags & MachO::SG_NORELOC) {
4268 outs() << " NORELOC";
4269 flags &= ~MachO::SG_NORELOC;
4270 }
4271 if (flags & MachO::SG_PROTECTED_VERSION_1) {
4272 outs() << " PROTECTED_VERSION_1";
4273 flags &= ~MachO::SG_PROTECTED_VERSION_1;
4274 }
4275 if (flags)
4276 outs() << format(" 0x%08" PRIx32"x", flags) << " (unknown flags)\n";
4277 else
4278 outs() << "\n";
4279 }
4280 } else {
4281 outs() << " flags " << format("0x%" PRIx32"x", flags) << "\n";
4282 }
4283}
4284
4285static void PrintSection(const char *sectname, const char *segname,
4286 uint64_t addr, uint64_t size, uint32_t offset,
4287 uint32_t align, uint32_t reloff, uint32_t nreloc,
4288 uint32_t flags, uint32_t reserved1, uint32_t reserved2,
4289 uint32_t cmd, const char *sg_segname,
4290 uint32_t filetype, uint32_t object_size,
4291 bool verbose) {
4292 outs() << "Section\n";
4293 outs() << " sectname " << format("%.16s\n", sectname);
4294 outs() << " segname " << format("%.16s", segname);
4295 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
4296 outs() << " (does not match segment)\n";
4297 else
4298 outs() << "\n";
4299 if (cmd == MachO::LC_SEGMENT_64) {
4300 outs() << " addr " << format("0x%016" PRIx64"l" "x", addr) << "\n";
4301 outs() << " size " << format("0x%016" PRIx64"l" "x", size);
4302 } else {
4303 outs() << " addr " << format("0x%08" PRIx64"l" "x", addr) << "\n";
4304 outs() << " size " << format("0x%08" PRIx64"l" "x", size);
4305 }
4306 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
4307 outs() << " (past end of file)\n";
4308 else
4309 outs() << "\n";
4310 outs() << " offset " << offset;
4311 if (offset > object_size)
4312 outs() << " (past end of file)\n";
4313 else
4314 outs() << "\n";
4315 uint32_t align_shifted = 1 << align;
4316 outs() << " align 2^" << align << " (" << align_shifted << ")\n";
4317 outs() << " reloff " << reloff;
4318 if (reloff > object_size)
4319 outs() << " (past end of file)\n";
4320 else
4321 outs() << "\n";
4322 outs() << " nreloc " << nreloc;
4323 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
4324 outs() << " (past end of file)\n";
4325 else
4326 outs() << "\n";
4327 uint32_t section_type = flags & MachO::SECTION_TYPE;
4328 if (verbose) {
4329 outs() << " type";
4330 if (section_type == MachO::S_REGULAR)
4331 outs() << " S_REGULAR\n";
4332 else if (section_type == MachO::S_ZEROFILL)
4333 outs() << " S_ZEROFILL\n";
4334 else if (section_type == MachO::S_CSTRING_LITERALS)
4335 outs() << " S_CSTRING_LITERALS\n";
4336 else if (section_type == MachO::S_4BYTE_LITERALS)
4337 outs() << " S_4BYTE_LITERALS\n";
4338 else if (section_type == MachO::S_8BYTE_LITERALS)
4339 outs() << " S_8BYTE_LITERALS\n";
4340 else if (section_type == MachO::S_16BYTE_LITERALS)
4341 outs() << " S_16BYTE_LITERALS\n";
4342 else if (section_type == MachO::S_LITERAL_POINTERS)
4343 outs() << " S_LITERAL_POINTERS\n";
4344 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
4345 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
4346 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
4347 outs() << " S_LAZY_SYMBOL_POINTERS\n";
4348 else if (section_type == MachO::S_SYMBOL_STUBS)
4349 outs() << " S_SYMBOL_STUBS\n";
4350 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
4351 outs() << " S_MOD_INIT_FUNC_POINTERS\n";
4352 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
4353 outs() << " S_MOD_TERM_FUNC_POINTERS\n";
4354 else if (section_type == MachO::S_COALESCED)
4355 outs() << " S_COALESCED\n";
4356 else if (section_type == MachO::S_INTERPOSING)
4357 outs() << " S_INTERPOSING\n";
4358 else if (section_type == MachO::S_DTRACE_DOF)
4359 outs() << " S_DTRACE_DOF\n";
4360 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
4361 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
4362 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
4363 outs() << " S_THREAD_LOCAL_REGULAR\n";
4364 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
4365 outs() << " S_THREAD_LOCAL_ZEROFILL\n";
4366 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
4367 outs() << " S_THREAD_LOCAL_VARIABLES\n";
4368 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
4369 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
4370 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
4371 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
4372 else
4373 outs() << format("0x%08" PRIx32"x", section_type) << "\n";
4374 outs() << "attributes";
4375 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
4376 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
4377 outs() << " PURE_INSTRUCTIONS";
4378 if (section_attributes & MachO::S_ATTR_NO_TOC)
4379 outs() << " NO_TOC";
4380 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
4381 outs() << " STRIP_STATIC_SYMS";
4382 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
4383 outs() << " NO_DEAD_STRIP";
4384 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
4385 outs() << " LIVE_SUPPORT";
4386 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
4387 outs() << " SELF_MODIFYING_CODE";
4388 if (section_attributes & MachO::S_ATTR_DEBUG)
4389 outs() << " DEBUG";
4390 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
4391 outs() << " SOME_INSTRUCTIONS";
4392 if (section_attributes & MachO::S_ATTR_EXT_RELOC)
4393 outs() << " EXT_RELOC";
4394 if (section_attributes & MachO::S_ATTR_LOC_RELOC)
4395 outs() << " LOC_RELOC";
4396 if (section_attributes == 0)
4397 outs() << " (none)";
4398 outs() << "\n";
4399 } else
4400 outs() << " flags " << format("0x%08" PRIx32"x", flags) << "\n";
4401 outs() << " reserved1 " << reserved1;
4402 if (section_type == MachO::S_SYMBOL_STUBS ||
4403 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
4404 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
4405 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
4406 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
4407 outs() << " (index into indirect symbol table)\n";
4408 else
4409 outs() << "\n";
4410 outs() << " reserved2 " << reserved2;
4411 if (section_type == MachO::S_SYMBOL_STUBS)
4412 outs() << " (size of stubs)\n";
4413 else
4414 outs() << "\n";
4415}
4416
4417static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
4418 uint32_t object_size) {
4419 outs() << " cmd LC_SYMTAB\n";
4420 outs() << " cmdsize " << st.cmdsize;
4421 if (st.cmdsize != sizeof(struct MachO::symtab_command))
4422 outs() << " Incorrect size\n";
4423 else
4424 outs() << "\n";
4425 outs() << " symoff " << st.symoff;
4426 if (st.symoff > object_size)
4427 outs() << " (past end of file)\n";
4428 else
4429 outs() << "\n";
4430 outs() << " nsyms " << st.nsyms;
4431 uint64_t big_size;
4432 if (Is64Bit) {
4433 big_size = st.nsyms;
4434 big_size *= sizeof(struct MachO::nlist_64);
4435 big_size += st.symoff;
4436 if (big_size > object_size)
4437 outs() << " (past end of file)\n";
4438 else
4439 outs() << "\n";
4440 } else {
4441 big_size = st.nsyms;
4442 big_size *= sizeof(struct MachO::nlist);
4443 big_size += st.symoff;
4444 if (big_size > object_size)
4445 outs() << " (past end of file)\n";
4446 else
4447 outs() << "\n";
4448 }
4449 outs() << " stroff " << st.stroff;
4450 if (st.stroff > object_size)
4451 outs() << " (past end of file)\n";
4452 else
4453 outs() << "\n";
4454 outs() << " strsize " << st.strsize;
4455 big_size = st.stroff;
4456 big_size += st.strsize;
4457 if (big_size > object_size)
4458 outs() << " (past end of file)\n";
4459 else
4460 outs() << "\n";
4461}
4462
4463static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
4464 uint32_t nsyms, uint32_t object_size,
4465 bool Is64Bit) {
4466 outs() << " cmd LC_DYSYMTAB\n";
4467 outs() << " cmdsize " << dyst.cmdsize;
4468 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
4469 outs() << " Incorrect size\n";
4470 else
4471 outs() << "\n";
4472 outs() << " ilocalsym " << dyst.ilocalsym;
4473 if (dyst.ilocalsym > nsyms)
4474 outs() << " (greater than the number of symbols)\n";
4475 else
4476 outs() << "\n";
4477 outs() << " nlocalsym " << dyst.nlocalsym;
4478 uint64_t big_size;
4479 big_size = dyst.ilocalsym;
4480 big_size += dyst.nlocalsym;
4481 if (big_size > nsyms)
4482 outs() << " (past the end of the symbol table)\n";
4483 else
4484 outs() << "\n";
4485 outs() << " iextdefsym " << dyst.iextdefsym;
4486 if (dyst.iextdefsym > nsyms)
4487 outs() << " (greater than the number of symbols)\n";
4488 else
4489 outs() << "\n";
4490 outs() << " nextdefsym " << dyst.nextdefsym;
4491 big_size = dyst.iextdefsym;
4492 big_size += dyst.nextdefsym;
4493 if (big_size > nsyms)
4494 outs() << " (past the end of the symbol table)\n";
4495 else
4496 outs() << "\n";
4497 outs() << " iundefsym " << dyst.iundefsym;
4498 if (dyst.iundefsym > nsyms)
4499 outs() << " (greater than the number of symbols)\n";
4500 else
4501 outs() << "\n";
4502 outs() << " nundefsym " << dyst.nundefsym;
4503 big_size = dyst.iundefsym;
4504 big_size += dyst.nundefsym;
4505 if (big_size > nsyms)
4506 outs() << " (past the end of the symbol table)\n";
4507 else
4508 outs() << "\n";
4509 outs() << " tocoff " << dyst.tocoff;
4510 if (dyst.tocoff > object_size)
4511 outs() << " (past end of file)\n";
4512 else
4513 outs() << "\n";
4514 outs() << " ntoc " << dyst.ntoc;
4515 big_size = dyst.ntoc;
4516 big_size *= sizeof(struct MachO::dylib_table_of_contents);
4517 big_size += dyst.tocoff;
4518 if (big_size > object_size)
4519 outs() << " (past end of file)\n";
4520 else
4521 outs() << "\n";
4522 outs() << " modtaboff " << dyst.modtaboff;
4523 if (dyst.modtaboff > object_size)
4524 outs() << " (past end of file)\n";
4525 else
4526 outs() << "\n";
4527 outs() << " nmodtab " << dyst.nmodtab;
4528 uint64_t modtabend;
4529 if (Is64Bit) {
4530 modtabend = dyst.nmodtab;
4531 modtabend *= sizeof(struct MachO::dylib_module_64);
4532 modtabend += dyst.modtaboff;
4533 } else {
4534 modtabend = dyst.nmodtab;
4535 modtabend *= sizeof(struct MachO::dylib_module);
4536 modtabend += dyst.modtaboff;
4537 }
4538 if (modtabend > object_size)
4539 outs() << " (past end of file)\n";
4540 else
4541 outs() << "\n";
4542 outs() << " extrefsymoff " << dyst.extrefsymoff;
4543 if (dyst.extrefsymoff > object_size)
4544 outs() << " (past end of file)\n";
4545 else
4546 outs() << "\n";
4547 outs() << " nextrefsyms " << dyst.nextrefsyms;
4548 big_size = dyst.nextrefsyms;
4549 big_size *= sizeof(struct MachO::dylib_reference);
4550 big_size += dyst.extrefsymoff;
4551 if (big_size > object_size)
4552 outs() << " (past end of file)\n";
4553 else
4554 outs() << "\n";
4555 outs() << " indirectsymoff " << dyst.indirectsymoff;
4556 if (dyst.indirectsymoff > object_size)
4557 outs() << " (past end of file)\n";
4558 else
4559 outs() << "\n";
4560 outs() << " nindirectsyms " << dyst.nindirectsyms;
4561 big_size = dyst.nindirectsyms;
4562 big_size *= sizeof(uint32_t);
4563 big_size += dyst.indirectsymoff;
4564 if (big_size > object_size)
4565 outs() << " (past end of file)\n";
4566 else
4567 outs() << "\n";
4568 outs() << " extreloff " << dyst.extreloff;
4569 if (dyst.extreloff > object_size)
4570 outs() << " (past end of file)\n";
4571 else
4572 outs() << "\n";
4573 outs() << " nextrel " << dyst.nextrel;
4574 big_size = dyst.nextrel;
4575 big_size *= sizeof(struct MachO::relocation_info);
4576 big_size += dyst.extreloff;
4577 if (big_size > object_size)
4578 outs() << " (past end of file)\n";
4579 else
4580 outs() << "\n";
4581 outs() << " locreloff " << dyst.locreloff;
4582 if (dyst.locreloff > object_size)
4583 outs() << " (past end of file)\n";
4584 else
4585 outs() << "\n";
4586 outs() << " nlocrel " << dyst.nlocrel;
4587 big_size = dyst.nlocrel;
4588 big_size *= sizeof(struct MachO::relocation_info);
4589 big_size += dyst.locreloff;
4590 if (big_size > object_size)
4591 outs() << " (past end of file)\n";
4592 else
4593 outs() << "\n";
4594}
4595
4596static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
4597 uint32_t object_size) {
4598 if (dc.cmd == MachO::LC_DYLD_INFO)
4599 outs() << " cmd LC_DYLD_INFO\n";
4600 else
4601 outs() << " cmd LC_DYLD_INFO_ONLY\n";
4602 outs() << " cmdsize " << dc.cmdsize;
4603 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
4604 outs() << " Incorrect size\n";
4605 else
4606 outs() << "\n";
4607 outs() << " rebase_off " << dc.rebase_off;
4608 if (dc.rebase_off > object_size)
4609 outs() << " (past end of file)\n";
4610 else
4611 outs() << "\n";
4612 outs() << " rebase_size " << dc.rebase_size;
4613 uint64_t big_size;
4614 big_size = dc.rebase_off;
4615 big_size += dc.rebase_size;
4616 if (big_size > object_size)
4617 outs() << " (past end of file)\n";
4618 else
4619 outs() << "\n";
4620 outs() << " bind_off " << dc.bind_off;
4621 if (dc.bind_off > object_size)
4622 outs() << " (past end of file)\n";
4623 else
4624 outs() << "\n";
4625 outs() << " bind_size " << dc.bind_size;
4626 big_size = dc.bind_off;
4627 big_size += dc.bind_size;
4628 if (big_size > object_size)
4629 outs() << " (past end of file)\n";
4630 else
4631 outs() << "\n";
4632 outs() << " weak_bind_off " << dc.weak_bind_off;
4633 if (dc.weak_bind_off > object_size)
4634 outs() << " (past end of file)\n";
4635 else
4636 outs() << "\n";
4637 outs() << " weak_bind_size " << dc.weak_bind_size;
4638 big_size = dc.weak_bind_off;
4639 big_size += dc.weak_bind_size;
4640 if (big_size > object_size)
4641 outs() << " (past end of file)\n";
4642 else
4643 outs() << "\n";
4644 outs() << " lazy_bind_off " << dc.lazy_bind_off;
4645 if (dc.lazy_bind_off > object_size)
4646 outs() << " (past end of file)\n";
4647 else
4648 outs() << "\n";
4649 outs() << " lazy_bind_size " << dc.lazy_bind_size;
4650 big_size = dc.lazy_bind_off;
4651 big_size += dc.lazy_bind_size;
4652 if (big_size > object_size)
4653 outs() << " (past end of file)\n";
4654 else
4655 outs() << "\n";
4656 outs() << " export_off " << dc.export_off;
4657 if (dc.export_off > object_size)
4658 outs() << " (past end of file)\n";
4659 else
4660 outs() << "\n";
4661 outs() << " export_size " << dc.export_size;
4662 big_size = dc.export_off;
4663 big_size += dc.export_size;
4664 if (big_size > object_size)
4665 outs() << " (past end of file)\n";
4666 else
4667 outs() << "\n";
4668}
4669
4670static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
4671 const char *Ptr) {
4672 if (dyld.cmd == MachO::LC_ID_DYLINKER)
4673 outs() << " cmd LC_ID_DYLINKER\n";
4674 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
4675 outs() << " cmd LC_LOAD_DYLINKER\n";
4676 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
4677 outs() << " cmd LC_DYLD_ENVIRONMENT\n";
4678 else
4679 outs() << " cmd ?(" << dyld.cmd << ")\n";
4680 outs() << " cmdsize " << dyld.cmdsize;
4681 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
4682 outs() << " Incorrect size\n";
4683 else
4684 outs() << "\n";
4685 if (dyld.name >= dyld.cmdsize)
4686 outs() << " name ?(bad offset " << dyld.name << ")\n";
4687 else {
4688 const char *P = (const char *)(Ptr) + dyld.name;
4689 outs() << " name " << P << " (offset " << dyld.name << ")\n";
4690 }
4691}
4692
4693static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
4694 outs() << " cmd LC_UUID\n";
4695 outs() << " cmdsize " << uuid.cmdsize;
4696 if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
4697 outs() << " Incorrect size\n";
4698 else
4699 outs() << "\n";
4700 outs() << " uuid ";
4701 outs() << format("%02" PRIX32"X", uuid.uuid[0]);
4702 outs() << format("%02" PRIX32"X", uuid.uuid[1]);
4703 outs() << format("%02" PRIX32"X", uuid.uuid[2]);
4704 outs() << format("%02" PRIX32"X", uuid.uuid[3]);
4705 outs() << "-";
4706 outs() << format("%02" PRIX32"X", uuid.uuid[4]);
4707 outs() << format("%02" PRIX32"X", uuid.uuid[5]);
4708 outs() << "-";
4709 outs() << format("%02" PRIX32"X", uuid.uuid[6]);
4710 outs() << format("%02" PRIX32"X", uuid.uuid[7]);
4711 outs() << "-";
4712 outs() << format("%02" PRIX32"X", uuid.uuid[8]);
4713 outs() << format("%02" PRIX32"X", uuid.uuid[9]);
4714 outs() << "-";
4715 outs() << format("%02" PRIX32"X", uuid.uuid[10]);
4716 outs() << format("%02" PRIX32"X", uuid.uuid[11]);
4717 outs() << format("%02" PRIX32"X", uuid.uuid[12]);
4718 outs() << format("%02" PRIX32"X", uuid.uuid[13]);
4719 outs() << format("%02" PRIX32"X", uuid.uuid[14]);
4720 outs() << format("%02" PRIX32"X", uuid.uuid[15]);
4721 outs() << "\n";
4722}
4723
4724static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) {
4725 outs() << " cmd LC_RPATH\n";
4726 outs() << " cmdsize " << rpath.cmdsize;
4727 if (rpath.cmdsize < sizeof(struct MachO::rpath_command))
4728 outs() << " Incorrect size\n";
4729 else
4730 outs() << "\n";
4731 if (rpath.path >= rpath.cmdsize)
4732 outs() << " path ?(bad offset " << rpath.path << ")\n";
4733 else {
4734 const char *P = (const char *)(Ptr) + rpath.path;
4735 outs() << " path " << P << " (offset " << rpath.path << ")\n";
4736 }
4737}
4738
4739static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
4740 if (vd.cmd == MachO::LC_VERSION_MIN_MACOSX)
4741 outs() << " cmd LC_VERSION_MIN_MACOSX\n";
4742 else if (vd.cmd == MachO::LC_VERSION_MIN_IPHONEOS)
4743 outs() << " cmd LC_VERSION_MIN_IPHONEOS\n";
4744 else
4745 outs() << " cmd " << vd.cmd << " (?)\n";
4746 outs() << " cmdsize " << vd.cmdsize;
4747 if (vd.cmdsize != sizeof(struct MachO::version_min_command))
4748 outs() << " Incorrect size\n";
4749 else
4750 outs() << "\n";
4751 outs() << " version " << ((vd.version >> 16) & 0xffff) << "."
4752 << ((vd.version >> 8) & 0xff);
4753 if ((vd.version & 0xff) != 0)
4754 outs() << "." << (vd.version & 0xff);
4755 outs() << "\n";
4756 if (vd.sdk == 0)
4757 outs() << " sdk n/a";
4758 else {
4759 outs() << " sdk " << ((vd.sdk >> 16) & 0xffff) << "."
4760 << ((vd.sdk >> 8) & 0xff);
4761 }
4762 if ((vd.sdk & 0xff) != 0)
4763 outs() << "." << (vd.sdk & 0xff);
4764 outs() << "\n";
4765}
4766
4767static void PrintSourceVersionCommand(MachO::source_version_command sd) {
4768 outs() << " cmd LC_SOURCE_VERSION\n";
4769 outs() << " cmdsize " << sd.cmdsize;
4770 if (sd.cmdsize != sizeof(struct MachO::source_version_command))
4771 outs() << " Incorrect size\n";
4772 else
4773 outs() << "\n";
4774 uint64_t a = (sd.version >> 40) & 0xffffff;
4775 uint64_t b = (sd.version >> 30) & 0x3ff;
4776 uint64_t c = (sd.version >> 20) & 0x3ff;
4777 uint64_t d = (sd.version >> 10) & 0x3ff;
4778 uint64_t e = sd.version & 0x3ff;
4779 outs() << " version " << a << "." << b;
4780 if (e != 0)
4781 outs() << "." << c << "." << d << "." << e;
4782 else if (d != 0)
4783 outs() << "." << c << "." << d;
4784 else if (c != 0)
4785 outs() << "." << c;
4786 outs() << "\n";
4787}
4788
4789static void PrintEntryPointCommand(MachO::entry_point_command ep) {
4790 outs() << " cmd LC_MAIN\n";
4791 outs() << " cmdsize " << ep.cmdsize;
4792 if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
4793 outs() << " Incorrect size\n";
4794 else
4795 outs() << "\n";
4796 outs() << " entryoff " << ep.entryoff << "\n";
4797 outs() << " stacksize " << ep.stacksize << "\n";
4798}
4799
4800static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec,
4801 uint32_t object_size) {
4802 outs() << " cmd LC_ENCRYPTION_INFO\n";
4803 outs() << " cmdsize " << ec.cmdsize;
4804 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command))
4805 outs() << " Incorrect size\n";
4806 else
4807 outs() << "\n";
4808 outs() << " cryptoff " << ec.cryptoff;
4809 if (ec.cryptoff > object_size)
4810 outs() << " (past end of file)\n";
4811 else
4812 outs() << "\n";
4813 outs() << " cryptsize " << ec.cryptsize;
4814 if (ec.cryptsize > object_size)
4815 outs() << " (past end of file)\n";
4816 else
4817 outs() << "\n";
4818 outs() << " cryptid " << ec.cryptid << "\n";
4819}
4820
4821static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,
4822 uint32_t object_size) {
4823 outs() << " cmd LC_ENCRYPTION_INFO_64\n";
4824 outs() << " cmdsize " << ec.cmdsize;
4825 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64))
4826 outs() << " Incorrect size\n";
4827 else
4828 outs() << "\n";
4829 outs() << " cryptoff " << ec.cryptoff;
4830 if (ec.cryptoff > object_size)
4831 outs() << " (past end of file)\n";
4832 else
4833 outs() << "\n";
4834 outs() << " cryptsize " << ec.cryptsize;
4835 if (ec.cryptsize > object_size)
4836 outs() << " (past end of file)\n";
4837 else
4838 outs() << "\n";
4839 outs() << " cryptid " << ec.cryptid << "\n";
4840 outs() << " pad " << ec.pad << "\n";
4841}
4842
4843static void PrintLinkerOptionCommand(MachO::linker_option_command lo,
4844 const char *Ptr) {
4845 outs() << " cmd LC_LINKER_OPTION\n";
4846 outs() << " cmdsize " << lo.cmdsize;
4847 if (lo.cmdsize < sizeof(struct MachO::linker_option_command))
4848 outs() << " Incorrect size\n";
4849 else
4850 outs() << "\n";
4851 outs() << " count " << lo.count << "\n";
4852 const char *string = Ptr + sizeof(struct MachO::linker_option_command);
4853 uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command);
4854 uint32_t i = 0;
4855 while (left > 0) {
4856 while (*string == '\0' && left > 0) {
4857 string++;
4858 left--;
4859 }
4860 if (left > 0) {
4861 i++;
4862 outs() << " string #" << i << " " << format("%.*s\n", left, string);
4863 uint32_t NullPos = StringRef(string, left).find('\0');
4864 uint32_t len = std::min(NullPos, left) + 1;
4865 string += len;
4866 left -= len;
4867 }
4868 }
4869 if (lo.count != i)
4870 outs() << " count " << lo.count << " does not match number of strings "
4871 << i << "\n";
4872}
4873
4874static void PrintSubFrameworkCommand(MachO::sub_framework_command sub,
4875 const char *Ptr) {
4876 outs() << " cmd LC_SUB_FRAMEWORK\n";
4877 outs() << " cmdsize " << sub.cmdsize;
4878 if (sub.cmdsize < sizeof(struct MachO::sub_framework_command))
4879 outs() << " Incorrect size\n";
4880 else
4881 outs() << "\n";
4882 if (sub.umbrella < sub.cmdsize) {
4883 const char *P = Ptr + sub.umbrella;
4884 outs() << " umbrella " << P << " (offset " << sub.umbrella << ")\n";
4885 } else {
4886 outs() << " umbrella ?(bad offset " << sub.umbrella << ")\n";
4887 }
4888}
4889
4890static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,
4891 const char *Ptr) {
4892 outs() << " cmd LC_SUB_UMBRELLA\n";
4893 outs() << " cmdsize " << sub.cmdsize;
4894 if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command))
4895 outs() << " Incorrect size\n";
4896 else
4897 outs() << "\n";
4898 if (sub.sub_umbrella < sub.cmdsize) {
4899 const char *P = Ptr + sub.sub_umbrella;
4900 outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n";
4901 } else {
4902 outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n";
4903 }
4904}
4905
4906static void PrintSubLibraryCommand(MachO::sub_library_command sub,
4907 const char *Ptr) {
4908 outs() << " cmd LC_SUB_LIBRARY\n";
4909 outs() << " cmdsize " << sub.cmdsize;
4910 if (sub.cmdsize < sizeof(struct MachO::sub_library_command))
4911 outs() << " Incorrect size\n";
4912 else
4913 outs() << "\n";
4914 if (sub.sub_library < sub.cmdsize) {
4915 const char *P = Ptr + sub.sub_library;
4916 outs() << " sub_library " << P << " (offset " << sub.sub_library << ")\n";
4917 } else {
4918 outs() << " sub_library ?(bad offset " << sub.sub_library << ")\n";
4919 }
4920}
4921
4922static void PrintSubClientCommand(MachO::sub_client_command sub,
4923 const char *Ptr) {
4924 outs() << " cmd LC_SUB_CLIENT\n";
4925 outs() << " cmdsize " << sub.cmdsize;
4926 if (sub.cmdsize < sizeof(struct MachO::sub_client_command))
4927 outs() << " Incorrect size\n";
4928 else
4929 outs() << "\n";
4930 if (sub.client < sub.cmdsize) {
4931 const char *P = Ptr + sub.client;
4932 outs() << " client " << P << " (offset " << sub.client << ")\n";
4933 } else {
4934 outs() << " client ?(bad offset " << sub.client << ")\n";
4935 }
4936}
4937
4938static void PrintRoutinesCommand(MachO::routines_command r) {
4939 outs() << " cmd LC_ROUTINES\n";
4940 outs() << " cmdsize " << r.cmdsize;
4941 if (r.cmdsize != sizeof(struct MachO::routines_command))
4942 outs() << " Incorrect size\n";
4943 else
4944 outs() << "\n";
4945 outs() << " init_address " << format("0x%08" PRIx32"x", r.init_address) << "\n";
4946 outs() << " init_module " << r.init_module << "\n";
4947 outs() << " reserved1 " << r.reserved1 << "\n";
4948 outs() << " reserved2 " << r.reserved2 << "\n";
4949 outs() << " reserved3 " << r.reserved3 << "\n";
4950 outs() << " reserved4 " << r.reserved4 << "\n";
4951 outs() << " reserved5 " << r.reserved5 << "\n";
4952 outs() << " reserved6 " << r.reserved6 << "\n";
4953}
4954
4955static void PrintRoutinesCommand64(MachO::routines_command_64 r) {
4956 outs() << " cmd LC_ROUTINES_64\n";
4957 outs() << " cmdsize " << r.cmdsize;
4958 if (r.cmdsize != sizeof(struct MachO::routines_command_64))
4959 outs() << " Incorrect size\n";
4960 else
4961 outs() << "\n";
4962 outs() << " init_address " << format("0x%016" PRIx64"l" "x", r.init_address) << "\n";
4963 outs() << " init_module " << r.init_module << "\n";
4964 outs() << " reserved1 " << r.reserved1 << "\n";
4965 outs() << " reserved2 " << r.reserved2 << "\n";
4966 outs() << " reserved3 " << r.reserved3 << "\n";
4967 outs() << " reserved4 " << r.reserved4 << "\n";
4968 outs() << " reserved5 " << r.reserved5 << "\n";
4969 outs() << " reserved6 " << r.reserved6 << "\n";
4970}
4971
4972static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) {
4973 outs() << " rax " << format("0x%016" PRIx64"l" "x", cpu64.rax);
4974 outs() << " rbx " << format("0x%016" PRIx64"l" "x", cpu64.rbx);
4975 outs() << " rcx " << format("0x%016" PRIx64"l" "x", cpu64.rcx) << "\n";
4976 outs() << " rdx " << format("0x%016" PRIx64"l" "x", cpu64.rdx);
4977 outs() << " rdi " << format("0x%016" PRIx64"l" "x", cpu64.rdi);
4978 outs() << " rsi " << format("0x%016" PRIx64"l" "x", cpu64.rsi) << "\n";
4979 outs() << " rbp " << format("0x%016" PRIx64"l" "x", cpu64.rbp);
4980 outs() << " rsp " << format("0x%016" PRIx64"l" "x", cpu64.rsp);
4981 outs() << " r8 " << format("0x%016" PRIx64"l" "x", cpu64.r8) << "\n";
4982 outs() << " r9 " << format("0x%016" PRIx64"l" "x", cpu64.r9);
4983 outs() << " r10 " << format("0x%016" PRIx64"l" "x", cpu64.r10);
4984 outs() << " r11 " << format("0x%016" PRIx64"l" "x", cpu64.r11) << "\n";
4985 outs() << " r12 " << format("0x%016" PRIx64"l" "x", cpu64.r12);
4986 outs() << " r13 " << format("0x%016" PRIx64"l" "x", cpu64.r13);
4987 outs() << " r14 " << format("0x%016" PRIx64"l" "x", cpu64.r14) << "\n";
4988 outs() << " r15 " << format("0x%016" PRIx64"l" "x", cpu64.r15);
4989 outs() << " rip " << format("0x%016" PRIx64"l" "x", cpu64.rip) << "\n";
4990 outs() << "rflags " << format("0x%016" PRIx64"l" "x", cpu64.rflags);
4991 outs() << " cs " << format("0x%016" PRIx64"l" "x", cpu64.cs);
4992 outs() << " fs " << format("0x%016" PRIx64"l" "x", cpu64.fs) << "\n";
4993 outs() << " gs " << format("0x%016" PRIx64"l" "x", cpu64.gs) << "\n";
4994}
4995
4996static void Print_mmst_reg(MachO::mmst_reg_t &r) {
4997 uint32_t f;
4998 outs() << "\t mmst_reg ";
4999 for (f = 0; f < 10; f++)
5000 outs() << format("%02" PRIx32"x", (r.mmst_reg[f] & 0xff)) << " ";
5001 outs() << "\n";
5002 outs() << "\t mmst_rsrv ";
5003 for (f = 0; f < 6; f++)
5004 outs() << format("%02" PRIx32"x", (r.mmst_rsrv[f] & 0xff)) << " ";
5005 outs() << "\n";
5006}
5007
5008static void Print_xmm_reg(MachO::xmm_reg_t &r) {
5009 uint32_t f;
5010 outs() << "\t xmm_reg ";
5011 for (f = 0; f < 16; f++)
5012 outs() << format("%02" PRIx32"x", (r.xmm_reg[f] & 0xff)) << " ";
5013 outs() << "\n";
5014}
5015
5016static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) {
5017 outs() << "\t fpu_reserved[0] " << fpu.fpu_reserved[0];
5018 outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n";
5019 outs() << "\t control: invalid " << fpu.fpu_fcw.invalid;
5020 outs() << " denorm " << fpu.fpu_fcw.denorm;
5021 outs() << " zdiv " << fpu.fpu_fcw.zdiv;
5022 outs() << " ovrfl " << fpu.fpu_fcw.ovrfl;
5023 outs() << " undfl " << fpu.fpu_fcw.undfl;
5024 outs() << " precis " << fpu.fpu_fcw.precis << "\n";
5025 outs() << "\t\t pc ";
5026 if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B)
5027 outs() << "FP_PREC_24B ";
5028 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B)
5029 outs() << "FP_PREC_53B ";
5030 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B)
5031 outs() << "FP_PREC_64B ";
5032 else
5033 outs() << fpu.fpu_fcw.pc << " ";
5034 outs() << "rc ";
5035 if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR)
5036 outs() << "FP_RND_NEAR ";
5037 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN)
5038 outs() << "FP_RND_DOWN ";
5039 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP)
5040 outs() << "FP_RND_UP ";
5041 else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP)
5042 outs() << "FP_CHOP ";
5043 outs() << "\n";
5044 outs() << "\t status: invalid " << fpu.fpu_fsw.invalid;
5045 outs() << " denorm " << fpu.fpu_fsw.denorm;
5046 outs() << " zdiv " << fpu.fpu_fsw.zdiv;
5047 outs() << " ovrfl " << fpu.fpu_fsw.ovrfl;
5048 outs() << " undfl " << fpu.fpu_fsw.undfl;
5049 outs() << " precis " << fpu.fpu_fsw.precis;
5050 outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n";
5051 outs() << "\t errsumm " << fpu.fpu_fsw.errsumm;
5052 outs() << " c0 " << fpu.fpu_fsw.c0;
5053 outs() << " c1 " << fpu.fpu_fsw.c1;
5054 outs() << " c2 " << fpu.fpu_fsw.c2;
5055 outs() << " tos " << fpu.fpu_fsw.tos;
5056 outs() << " c3 " << fpu.fpu_fsw.c3;
5057 outs() << " busy " << fpu.fpu_fsw.busy << "\n";
5058 outs() << "\t fpu_ftw " << format("0x%02" PRIx32"x", fpu.fpu_ftw);
5059 outs() << " fpu_rsrv1 " << format("0x%02" PRIx32"x", fpu.fpu_rsrv1);
5060 outs() << " fpu_fop " << format("0x%04" PRIx32"x", fpu.fpu_fop);
5061 outs() << " fpu_ip " << format("0x%08" PRIx32"x", fpu.fpu_ip) << "\n";
5062 outs() << "\t fpu_cs " << format("0x%04" PRIx32"x", fpu.fpu_cs);
5063 outs() << " fpu_rsrv2 " << format("0x%04" PRIx32"x", fpu.fpu_rsrv2);
5064 outs() << " fpu_dp " << format("0x%08" PRIx32"x", fpu.fpu_dp);
5065 outs() << " fpu_ds " << format("0x%04" PRIx32"x", fpu.fpu_ds) << "\n";
5066 outs() << "\t fpu_rsrv3 " << format("0x%04" PRIx32"x", fpu.fpu_rsrv3);
5067 outs() << " fpu_mxcsr " << format("0x%08" PRIx32"x", fpu.fpu_mxcsr);
5068 outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32"x", fpu.fpu_mxcsrmask);
5069 outs() << "\n";
5070 outs() << "\t fpu_stmm0:\n";
5071 Print_mmst_reg(fpu.fpu_stmm0);
5072 outs() << "\t fpu_stmm1:\n";
5073 Print_mmst_reg(fpu.fpu_stmm1);
5074 outs() << "\t fpu_stmm2:\n";
5075 Print_mmst_reg(fpu.fpu_stmm2);
5076 outs() << "\t fpu_stmm3:\n";
5077 Print_mmst_reg(fpu.fpu_stmm3);
5078 outs() << "\t fpu_stmm4:\n";
5079 Print_mmst_reg(fpu.fpu_stmm4);
5080 outs() << "\t fpu_stmm5:\n";
5081 Print_mmst_reg(fpu.fpu_stmm5);
5082 outs() << "\t fpu_stmm6:\n";
5083 Print_mmst_reg(fpu.fpu_stmm6);
5084 outs() << "\t fpu_stmm7:\n";
5085 Print_mmst_reg(fpu.fpu_stmm7);
5086 outs() << "\t fpu_xmm0:\n";
5087 Print_xmm_reg(fpu.fpu_xmm0);
5088 outs() << "\t fpu_xmm1:\n";
5089 Print_xmm_reg(fpu.fpu_xmm1);
5090 outs() << "\t fpu_xmm2:\n";
5091 Print_xmm_reg(fpu.fpu_xmm2);
5092 outs() << "\t fpu_xmm3:\n";
5093 Print_xmm_reg(fpu.fpu_xmm3);
5094 outs() << "\t fpu_xmm4:\n";
5095 Print_xmm_reg(fpu.fpu_xmm4);
5096 outs() << "\t fpu_xmm5:\n";
5097 Print_xmm_reg(fpu.fpu_xmm5);
5098 outs() << "\t fpu_xmm6:\n";
5099 Print_xmm_reg(fpu.fpu_xmm6);
5100 outs() << "\t fpu_xmm7:\n";
5101 Print_xmm_reg(fpu.fpu_xmm7);
5102 outs() << "\t fpu_xmm8:\n";
5103 Print_xmm_reg(fpu.fpu_xmm8);
5104 outs() << "\t fpu_xmm9:\n";
5105 Print_xmm_reg(fpu.fpu_xmm9);
5106 outs() << "\t fpu_xmm10:\n";
5107 Print_xmm_reg(fpu.fpu_xmm10);
5108 outs() << "\t fpu_xmm11:\n";
5109 Print_xmm_reg(fpu.fpu_xmm11);
5110 outs() << "\t fpu_xmm12:\n";
5111 Print_xmm_reg(fpu.fpu_xmm12);
5112 outs() << "\t fpu_xmm13:\n";
5113 Print_xmm_reg(fpu.fpu_xmm13);
5114 outs() << "\t fpu_xmm14:\n";
5115 Print_xmm_reg(fpu.fpu_xmm14);
5116 outs() << "\t fpu_xmm15:\n";
5117 Print_xmm_reg(fpu.fpu_xmm15);
5118 outs() << "\t fpu_rsrv4:\n";
5119 for (uint32_t f = 0; f < 6; f++) {
5120 outs() << "\t ";
5121 for (uint32_t g = 0; g < 16; g++)
5122 outs() << format("%02" PRIx32"x", fpu.fpu_rsrv4[f * g]) << " ";
5123 outs() << "\n";
5124 }
5125 outs() << "\t fpu_reserved1 " << format("0x%08" PRIx32"x", fpu.fpu_reserved1);
5126 outs() << "\n";
5127}
5128
5129static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) {
5130 outs() << "\t trapno " << format("0x%08" PRIx32"x", exc64.trapno);
5131 outs() << " err " << format("0x%08" PRIx32"x", exc64.err);
5132 outs() << " faultvaddr " << format("0x%016" PRIx64"l" "x", exc64.faultvaddr) << "\n";
5133}
5134
5135static void PrintThreadCommand(MachO::thread_command t, const char *Ptr,
5136 bool isLittleEndian, uint32_t cputype) {
5137 if (t.cmd == MachO::LC_THREAD)
5138 outs() << " cmd LC_THREAD\n";
5139 else if (t.cmd == MachO::LC_UNIXTHREAD)
5140 outs() << " cmd LC_UNIXTHREAD\n";
5141 else
5142 outs() << " cmd " << t.cmd << " (unknown)\n";
5143 outs() << " cmdsize " << t.cmdsize;
5144 if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t))
5145 outs() << " Incorrect size\n";
5146 else
5147 outs() << "\n";
5148
5149 const char *begin = Ptr + sizeof(struct MachO::thread_command);
5150 const char *end = Ptr + t.cmdsize;
5151 uint32_t flavor, count, left;
5152 if (cputype == MachO::CPU_TYPE_X86_64) {
5153 while (begin < end) {
5154 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
5155 memcpy((char *)&flavor, begin, sizeof(uint32_t));
5156 begin += sizeof(uint32_t);
5157 } else {
5158 flavor = 0;
5159 begin = end;
5160 }
5161 if (isLittleEndian != sys::IsLittleEndianHost)
5162 sys::swapByteOrder(flavor);
5163 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
5164 memcpy((char *)&count, begin, sizeof(uint32_t));
5165 begin += sizeof(uint32_t);
5166 } else {
5167 count = 0;
5168 begin = end;
5169 }
5170 if (isLittleEndian != sys::IsLittleEndianHost)
5171 sys::swapByteOrder(count);
5172 if (flavor == MachO::x86_THREAD_STATE64) {
5173 outs() << " flavor x86_THREAD_STATE64\n";
5174 if (count == MachO::x86_THREAD_STATE64_COUNT)
5175 outs() << " count x86_THREAD_STATE64_COUNT\n";
5176 else
5177 outs() << " count " << count
5178 << " (not x86_THREAD_STATE64_COUNT)\n";
5179 MachO::x86_thread_state64_t cpu64;
5180 left = end - begin;
5181 if (left >= sizeof(MachO::x86_thread_state64_t)) {
5182 memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t));
5183 begin += sizeof(MachO::x86_thread_state64_t);
5184 } else {
5185 memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t));
5186 memcpy(&cpu64, begin, left);
5187 begin += left;
5188 }
5189 if (isLittleEndian != sys::IsLittleEndianHost)
5190 swapStruct(cpu64);
5191 Print_x86_thread_state64_t(cpu64);
5192 } else if (flavor == MachO::x86_THREAD_STATE) {
5193 outs() << " flavor x86_THREAD_STATE\n";
5194 if (count == MachO::x86_THREAD_STATE_COUNT)
5195 outs() << " count x86_THREAD_STATE_COUNT\n";
5196 else
5197 outs() << " count " << count
5198 << " (not x86_THREAD_STATE_COUNT)\n";
5199 struct MachO::x86_thread_state_t ts;
5200 left = end - begin;
5201 if (left >= sizeof(MachO::x86_thread_state_t)) {
5202 memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t));
5203 begin += sizeof(MachO::x86_thread_state_t);
5204 } else {
5205 memset(&ts, '\0', sizeof(MachO::x86_thread_state_t));
5206 memcpy(&ts, begin, left);
5207 begin += left;
5208 }
5209 if (isLittleEndian != sys::IsLittleEndianHost)
5210 swapStruct(ts);
5211 if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) {
5212 outs() << "\t tsh.flavor x86_THREAD_STATE64 ";
5213 if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT)
5214 outs() << "tsh.count x86_THREAD_STATE64_COUNT\n";
5215 else
5216 outs() << "tsh.count " << ts.tsh.count
5217 << " (not x86_THREAD_STATE64_COUNT\n";
5218 Print_x86_thread_state64_t(ts.uts.ts64);
5219 } else {
5220 outs() << "\t tsh.flavor " << ts.tsh.flavor << " tsh.count "
5221 << ts.tsh.count << "\n";
5222 }
5223 } else if (flavor == MachO::x86_FLOAT_STATE) {
5224 outs() << " flavor x86_FLOAT_STATE\n";
5225 if (count == MachO::x86_FLOAT_STATE_COUNT)
5226 outs() << " count x86_FLOAT_STATE_COUNT\n";
5227 else
5228 outs() << " count " << count << " (not x86_FLOAT_STATE_COUNT)\n";
5229 struct MachO::x86_float_state_t fs;
5230 left = end - begin;
5231 if (left >= sizeof(MachO::x86_float_state_t)) {
5232 memcpy(&fs, begin, sizeof(MachO::x86_float_state_t));
5233 begin += sizeof(MachO::x86_float_state_t);
5234 } else {
5235 memset(&fs, '\0', sizeof(MachO::x86_float_state_t));
5236 memcpy(&fs, begin, left);
5237 begin += left;
5238 }
5239 if (isLittleEndian != sys::IsLittleEndianHost)
5240 swapStruct(fs);
5241 if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) {
5242 outs() << "\t fsh.flavor x86_FLOAT_STATE64 ";
5243 if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT)
5244 outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n";
5245 else
5246 outs() << "fsh.count " << fs.fsh.count
5247 << " (not x86_FLOAT_STATE64_COUNT\n";
5248 Print_x86_float_state_t(fs.ufs.fs64);
5249 } else {
5250 outs() << "\t fsh.flavor " << fs.fsh.flavor << " fsh.count "
5251 << fs.fsh.count << "\n";
5252 }
5253 } else if (flavor == MachO::x86_EXCEPTION_STATE) {
5254 outs() << " flavor x86_EXCEPTION_STATE\n";
5255 if (count == MachO::x86_EXCEPTION_STATE_COUNT)
5256 outs() << " count x86_EXCEPTION_STATE_COUNT\n";
5257 else
5258 outs() << " count " << count
5259 << " (not x86_EXCEPTION_STATE_COUNT)\n";
5260 struct MachO::x86_exception_state_t es;
5261 left = end - begin;
5262 if (left >= sizeof(MachO::x86_exception_state_t)) {
5263 memcpy(&es, begin, sizeof(MachO::x86_exception_state_t));
5264 begin += sizeof(MachO::x86_exception_state_t);
5265 } else {
5266 memset(&es, '\0', sizeof(MachO::x86_exception_state_t));
5267 memcpy(&es, begin, left);
5268 begin += left;
5269 }
5270 if (isLittleEndian != sys::IsLittleEndianHost)
5271 swapStruct(es);
5272 if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) {
5273 outs() << "\t esh.flavor x86_EXCEPTION_STATE64\n";
5274 if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT)
5275 outs() << "\t esh.count x86_EXCEPTION_STATE64_COUNT\n";
5276 else
5277 outs() << "\t esh.count " << es.esh.count
5278 << " (not x86_EXCEPTION_STATE64_COUNT\n";
5279 Print_x86_exception_state_t(es.ues.es64);
5280 } else {
5281 outs() << "\t esh.flavor " << es.esh.flavor << " esh.count "
5282 << es.esh.count << "\n";
5283 }
5284 } else {
5285 outs() << " flavor " << flavor << " (unknown)\n";
5286 outs() << " count " << count << "\n";
5287 outs() << " state (unknown)\n";
5288 begin += count * sizeof(uint32_t);
5289 }
5290 }
5291 } else {
5292 while (begin < end) {
5293 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
5294 memcpy((char *)&flavor, begin, sizeof(uint32_t));
5295 begin += sizeof(uint32_t);
5296 } else {
5297 flavor = 0;
5298 begin = end;
5299 }
5300 if (isLittleEndian != sys::IsLittleEndianHost)
5301 sys::swapByteOrder(flavor);
5302 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
5303 memcpy((char *)&count, begin, sizeof(uint32_t));
5304 begin += sizeof(uint32_t);
5305 } else {
5306 count = 0;
5307 begin = end;
5308 }
5309 if (isLittleEndian != sys::IsLittleEndianHost)
5310 sys::swapByteOrder(count);
5311 outs() << " flavor " << flavor << "\n";
5312 outs() << " count " << count << "\n";
5313 outs() << " state (Unknown cputype/cpusubtype)\n";
5314 begin += count * sizeof(uint32_t);
5315 }
5316 }
5317}
5318
5319static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
5320 if (dl.cmd == MachO::LC_ID_DYLIB)
5321 outs() << " cmd LC_ID_DYLIB\n";
5322 else if (dl.cmd == MachO::LC_LOAD_DYLIB)
5323 outs() << " cmd LC_LOAD_DYLIB\n";
5324 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
5325 outs() << " cmd LC_LOAD_WEAK_DYLIB\n";
5326 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
5327 outs() << " cmd LC_REEXPORT_DYLIB\n";
5328 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
5329 outs() << " cmd LC_LAZY_LOAD_DYLIB\n";
5330 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
5331 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n";
5332 else
5333 outs() << " cmd " << dl.cmd << " (unknown)\n";
5334 outs() << " cmdsize " << dl.cmdsize;
5335 if (dl.cmdsize < sizeof(struct MachO::dylib_command))
5336 outs() << " Incorrect size\n";
5337 else
5338 outs() << "\n";
5339 if (dl.dylib.name < dl.cmdsize) {
5340 const char *P = (const char *)(Ptr) + dl.dylib.name;
5341 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n";
5342 } else {
5343 outs() << " name ?(bad offset " << dl.dylib.name << ")\n";
5344 }
5345 outs() << " time stamp " << dl.dylib.timestamp << " ";
5346 time_t t = dl.dylib.timestamp;
5347 outs() << ctime(&t);
5348 outs() << " current version ";
5349 if (dl.dylib.current_version == 0xffffffff)
5350 outs() << "n/a\n";
5351 else
5352 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
5353 << ((dl.dylib.current_version >> 8) & 0xff) << "."
5354 << (dl.dylib.current_version & 0xff) << "\n";
5355 outs() << "compatibility version ";
5356 if (dl.dylib.compatibility_version == 0xffffffff)
5357 outs() << "n/a\n";
5358 else
5359 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
5360 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
5361 << (dl.dylib.compatibility_version & 0xff) << "\n";
5362}
5363
5364static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
5365 uint32_t object_size) {
5366 if (ld.cmd == MachO::LC_CODE_SIGNATURE)
5367 outs() << " cmd LC_FUNCTION_STARTS\n";
5368 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
5369 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n";
5370 else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
5371 outs() << " cmd LC_FUNCTION_STARTS\n";
5372 else if (ld.cmd == MachO::LC_DATA_IN_CODE)
5373 outs() << " cmd LC_DATA_IN_CODE\n";
5374 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
5375 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n";
5376 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
5377 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n";
5378 else
5379 outs() << " cmd " << ld.cmd << " (?)\n";
5380 outs() << " cmdsize " << ld.cmdsize;
5381 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
5382 outs() << " Incorrect size\n";
5383 else
5384 outs() << "\n";
5385 outs() << " dataoff " << ld.dataoff;
5386 if (ld.dataoff > object_size)
5387 outs() << " (past end of file)\n";
5388 else
5389 outs() << "\n";
5390 outs() << " datasize " << ld.datasize;
5391 uint64_t big_size = ld.dataoff;
5392 big_size += ld.datasize;
5393 if (big_size > object_size)
5394 outs() << " (past end of file)\n";
5395 else
5396 outs() << "\n";
5397}
5398
5399static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t ncmds,
5400 uint32_t filetype, uint32_t cputype,
5401 bool verbose) {
5402 if (ncmds == 0)
5403 return;
5404 StringRef Buf = Obj->getData();
5405 MachOObjectFile::LoadCommandInfo Command = Obj->getFirstLoadCommandInfo();
5406 for (unsigned i = 0;; ++i) {
5407 outs() << "Load command " << i << "\n";
5408 if (Command.C.cmd == MachO::LC_SEGMENT) {
5409 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
5410 const char *sg_segname = SLC.segname;
5411 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
5412 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
5413 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
5414 verbose);
5415 for (unsigned j = 0; j < SLC.nsects; j++) {
5416 MachO::section S = Obj->getSection(Command, j);
5417 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
5418 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
5419 SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
5420 }
5421 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
5422 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
5423 const char *sg_segname = SLC_64.segname;
5424 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
5425 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
5426 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
5427 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
5428 for (unsigned j = 0; j < SLC_64.nsects; j++) {
5429 MachO::section_64 S_64 = Obj->getSection64(Command, j);
5430 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
5431 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
5432 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
5433 sg_segname, filetype, Buf.size(), verbose);
5434 }
5435 } else if (Command.C.cmd == MachO::LC_SYMTAB) {
5436 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
5437 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
5438 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
5439 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
5440 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
5441 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
5442 Obj->is64Bit());
5443 } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
5444 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
5445 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
5446 PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
5447 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
5448 Command.C.cmd == MachO::LC_ID_DYLINKER ||
5449 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
5450 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
5451 PrintDyldLoadCommand(Dyld, Command.Ptr);
5452 } else if (Command.C.cmd == MachO::LC_UUID) {
5453 MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
5454 PrintUuidLoadCommand(Uuid);
5455 } else if (Command.C.cmd == MachO::LC_RPATH) {
5456 MachO::rpath_command Rpath = Obj->getRpathCommand(Command);
5457 PrintRpathLoadCommand(Rpath, Command.Ptr);
5458 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX ||
5459 Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) {
5460 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
5461 PrintVersionMinLoadCommand(Vd);
5462 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
5463 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
5464 PrintSourceVersionCommand(Sd);
5465 } else if (Command.C.cmd == MachO::LC_MAIN) {
5466 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
5467 PrintEntryPointCommand(Ep);
5468 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) {
5469 MachO::encryption_info_command Ei =
5470 Obj->getEncryptionInfoCommand(Command);
5471 PrintEncryptionInfoCommand(Ei, Buf.size());
5472 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
5473 MachO::encryption_info_command_64 Ei =
5474 Obj->getEncryptionInfoCommand64(Command);
5475 PrintEncryptionInfoCommand64(Ei, Buf.size());
5476 } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) {
5477 MachO::linker_option_command Lo =
5478 Obj->getLinkerOptionLoadCommand(Command);
5479 PrintLinkerOptionCommand(Lo, Command.Ptr);
5480 } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) {
5481 MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command);
5482 PrintSubFrameworkCommand(Sf, Command.Ptr);
5483 } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) {
5484 MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command);
5485 PrintSubUmbrellaCommand(Sf, Command.Ptr);
5486 } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) {
5487 MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command);
5488 PrintSubLibraryCommand(Sl, Command.Ptr);
5489 } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) {
5490 MachO::sub_client_command Sc = Obj->getSubClientCommand(Command);
5491 PrintSubClientCommand(Sc, Command.Ptr);
5492 } else if (Command.C.cmd == MachO::LC_ROUTINES) {
5493 MachO::routines_command Rc = Obj->getRoutinesCommand(Command);
5494 PrintRoutinesCommand(Rc);
5495 } else if (Command.C.cmd == MachO::LC_ROUTINES_64) {
5496 MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command);
5497 PrintRoutinesCommand64(Rc);
5498 } else if (Command.C.cmd == MachO::LC_THREAD ||
5499 Command.C.cmd == MachO::LC_UNIXTHREAD) {
5500 MachO::thread_command Tc = Obj->getThreadCommand(Command);
5501 PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype);
5502 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
5503 Command.C.cmd == MachO::LC_ID_DYLIB ||
5504 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
5505 Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
5506 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
5507 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
5508 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
5509 PrintDylibCommand(Dl, Command.Ptr);
5510 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
5511 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
5512 Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
5513 Command.C.cmd == MachO::LC_DATA_IN_CODE ||
5514 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
5515 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
5516 MachO::linkedit_data_command Ld =
5517 Obj->getLinkeditDataLoadCommand(Command);
5518 PrintLinkEditDataCommand(Ld, Buf.size());
5519 } else {
5520 outs() << " cmd ?(" << format("0x%08" PRIx32"x", Command.C.cmd)
5521 << ")\n";
5522 outs() << " cmdsize " << Command.C.cmdsize << "\n";
5523 // TODO: get and print the raw bytes of the load command.
5524 }
5525 // TODO: print all the other kinds of load commands.
5526 if (i == ncmds - 1)
5527 break;
5528 else
5529 Command = Obj->getNextLoadCommandInfo(Command);
5530 }
5531}
5532
5533static void getAndPrintMachHeader(const MachOObjectFile *Obj, uint32_t &ncmds,
5534 uint32_t &filetype, uint32_t &cputype,
5535 bool verbose) {
5536 if (Obj->is64Bit()) {
5537 MachO::mach_header_64 H_64;
5538 H_64 = Obj->getHeader64();
5539 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
5540 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
5541 ncmds = H_64.ncmds;
5542 filetype = H_64.filetype;
5543 cputype = H_64.cputype;
5544 } else {
5545 MachO::mach_header H;
5546 H = Obj->getHeader();
5547 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
5548 H.sizeofcmds, H.flags, verbose);
5549 ncmds = H.ncmds;
5550 filetype = H.filetype;
5551 cputype = H.cputype;
5552 }
5553}
5554
5555void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
5556 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
5557 uint32_t ncmds = 0;
5558 uint32_t filetype = 0;
5559 uint32_t cputype = 0;
5560 getAndPrintMachHeader(file, ncmds, filetype, cputype, !NonVerbose);
5561 PrintLoadCommands(file, ncmds, filetype, cputype, !NonVerbose);
5562}
5563
5564//===----------------------------------------------------------------------===//
5565// export trie dumping
5566//===----------------------------------------------------------------------===//
5567
5568void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
5569 for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
5570 uint64_t Flags = Entry.flags();
5571 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
5572 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
5573 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5574 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
5575 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5576 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
5577 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
5578 if (ReExport)
5579 outs() << "[re-export] ";
5580 else
5581 outs() << format("0x%08llX ",
5582 Entry.address()); // FIXME:add in base address
5583 outs() << Entry.name();
5584 if (WeakDef || ThreadLocal || Resolver || Abs) {
5585 bool NeedsComma = false;
5586 outs() << " [";
5587 if (WeakDef) {
5588 outs() << "weak_def";
5589 NeedsComma = true;
5590 }
5591 if (ThreadLocal) {
5592 if (NeedsComma)
5593 outs() << ", ";
5594 outs() << "per-thread";
5595 NeedsComma = true;
5596 }
5597 if (Abs) {
5598 if (NeedsComma)
5599 outs() << ", ";
5600 outs() << "absolute";
5601 NeedsComma = true;
5602 }
5603 if (Resolver) {
5604 if (NeedsComma)
5605 outs() << ", ";
5606 outs() << format("resolver=0x%08llX", Entry.other());
5607 NeedsComma = true;
5608 }
5609 outs() << "]";
5610 }
5611 if (ReExport) {
5612 StringRef DylibName = "unknown";
5613 int Ordinal = Entry.other() - 1;
5614 Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
5615 if (Entry.otherName().empty())
5616 outs() << " (from " << DylibName << ")";
5617 else
5618 outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
5619 }
5620 outs() << "\n";
5621 }
5622}
5623
5624//===----------------------------------------------------------------------===//
5625// rebase table dumping
5626//===----------------------------------------------------------------------===//
5627
5628namespace {
5629class SegInfo {
5630public:
5631 SegInfo(const object::MachOObjectFile *Obj);
5632
5633 StringRef segmentName(uint32_t SegIndex);
5634 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
5635 uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
5636
5637private:
5638 struct SectionInfo {
5639 uint64_t Address;
5640 uint64_t Size;
5641 StringRef SectionName;
5642 StringRef SegmentName;
5643 uint64_t OffsetInSegment;
5644 uint64_t SegmentStartAddress;
5645 uint32_t SegmentIndex;
5646 };
5647 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
5648 SmallVector<SectionInfo, 32> Sections;
5649};
5650}
5651
5652SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
5653 // Build table of sections so segIndex/offset pairs can be translated.
5654 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
5655 StringRef CurSegName;
5656 uint64_t CurSegAddress;
5657 for (const SectionRef &Section : Obj->sections()) {
5658 SectionInfo Info;
5659 if (error(Section.getName(Info.SectionName)))
5660 return;
5661 Info.Address = Section.getAddress();
5662 Info.Size = Section.getSize();
5663 Info.SegmentName =
5664 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
5665 if (!Info.SegmentName.equals(CurSegName)) {
5666 ++CurSegIndex;
5667 CurSegName = Info.SegmentName;
5668 CurSegAddress = Info.Address;
5669 }
5670 Info.SegmentIndex = CurSegIndex - 1;
5671 Info.OffsetInSegment = Info.Address - CurSegAddress;
5672 Info.SegmentStartAddress = CurSegAddress;
5673 Sections.push_back(Info);
5674 }
5675}
5676
5677StringRef SegInfo::segmentName(uint32_t SegIndex) {
5678 for (const SectionInfo &SI : Sections) {
5679 if (SI.SegmentIndex == SegIndex)
5680 return SI.SegmentName;
5681 }
5682 llvm_unreachable("invalid segIndex")::llvm::llvm_unreachable_internal("invalid segIndex", "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 5682)
;
5683}
5684
5685const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
5686 uint64_t OffsetInSeg) {
5687 for (const SectionInfo &SI : Sections) {
5688 if (SI.SegmentIndex != SegIndex)
5689 continue;
5690 if (SI.OffsetInSegment > OffsetInSeg)
5691 continue;
5692 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
5693 continue;
5694 return SI;
5695 }
5696 llvm_unreachable("segIndex and offset not in any section")::llvm::llvm_unreachable_internal("segIndex and offset not in any section"
, "/tmp/buildd/llvm-toolchain-snapshot-3.7~svn232922/tools/llvm-objdump/MachODump.cpp"
, 5696)
;
5697}
5698
5699StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
5700 return findSection(SegIndex, OffsetInSeg).SectionName;
5701}
5702
5703uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
5704 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
5705 return SI.SegmentStartAddress + OffsetInSeg;
5706}
5707
5708void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
5709 // Build table of sections so names can used in final output.
5710 SegInfo sectionTable(Obj);
5711
5712 outs() << "segment section address type\n";
5713 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
5714 uint32_t SegIndex = Entry.segmentIndex();
5715 uint64_t OffsetInSeg = Entry.segmentOffset();
5716 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5717 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5718 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5719
5720 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer
5721 outs() << format("%-8s %-18s 0x%08" PRIX64"l" "X" " %s\n",
5722 SegmentName.str().c_str(), SectionName.str().c_str(),
5723 Address, Entry.typeName().str().c_str());
5724 }
5725}
5726
5727static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
5728 StringRef DylibName;
5729 switch (Ordinal) {
5730 case MachO::BIND_SPECIAL_DYLIB_SELF:
5731 return "this-image";
5732 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
5733 return "main-executable";
5734 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
5735 return "flat-namespace";
5736 default:
5737 if (Ordinal > 0) {
5738 std::error_code EC =
5739 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
5740 if (EC)
5741 return "<<bad library ordinal>>";
5742 return DylibName;
5743 }
5744 }
5745 return "<<unknown special ordinal>>";
5746}
5747
5748//===----------------------------------------------------------------------===//
5749// bind table dumping
5750//===----------------------------------------------------------------------===//
5751
5752void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
5753 // Build table of sections so names can used in final output.
5754 SegInfo sectionTable(Obj);
5755
5756 outs() << "segment section address type "
5757 "addend dylib symbol\n";
5758 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
5759 uint32_t SegIndex = Entry.segmentIndex();
5760 uint64_t OffsetInSeg = Entry.segmentOffset();
5761 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5762 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5763 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5764
5765 // Table lines look like:
5766 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard
5767 StringRef Attr;
5768 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
5769 Attr = " (weak_import)";
5770 outs() << left_justify(SegmentName, 8) << " "
5771 << left_justify(SectionName, 18) << " "
5772 << format_hex(Address, 10, true) << " "
5773 << left_justify(Entry.typeName(), 8) << " "
5774 << format_decimal(Entry.addend(), 8) << " "
5775 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5776 << Entry.symbolName() << Attr << "\n";
5777 }
5778}
5779
5780//===----------------------------------------------------------------------===//
5781// lazy bind table dumping
5782//===----------------------------------------------------------------------===//
5783
5784void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
5785 // Build table of sections so names can used in final output.
5786 SegInfo sectionTable(Obj);
5787
5788 outs() << "segment section address "
5789 "dylib symbol\n";
5790 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
5791 uint32_t SegIndex = Entry.segmentIndex();
5792 uint64_t OffsetInSeg = Entry.segmentOffset();
5793 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5794 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5795 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5796
5797 // Table lines look like:
5798 // __DATA __got 0x00012010 libSystem ___stack_chk_guard
5799 outs() << left_justify(SegmentName, 8) << " "
5800 << left_justify(SectionName, 18) << " "
5801 << format_hex(Address, 10, true) << " "
5802 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
5803 << Entry.symbolName() << "\n";
5804 }
5805}
5806
5807//===----------------------------------------------------------------------===//
5808// weak bind table dumping
5809//===----------------------------------------------------------------------===//
5810
5811void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
5812 // Build table of sections so names can used in final output.
5813 SegInfo sectionTable(Obj);
5814
5815 outs() << "segment section address "
5816 "type addend symbol\n";
5817 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
5818 // Strong symbols don't have a location to update.
5819 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
5820 outs() << " strong "
5821 << Entry.symbolName() << "\n";
5822 continue;
5823 }
5824 uint32_t SegIndex = Entry.segmentIndex();
5825 uint64_t OffsetInSeg = Entry.segmentOffset();
5826 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5827 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5828 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5829
5830 // Table lines look like:
5831 // __DATA __data 0x00001000 pointer 0 _foo
5832 outs() << left_justify(SegmentName, 8) << " "
5833 << left_justify(SectionName, 18) << " "
5834 << format_hex(Address, 10, true) << " "
5835 << left_justify(Entry.typeName(), 8) << " "
5836 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName()
5837 << "\n";
5838 }
5839}
5840
5841// get_dyld_bind_info_symbolname() is used for disassembly and passed an
5842// address, ReferenceValue, in the Mach-O file and looks in the dyld bind
5843// information for that address. If the address is found its binding symbol
5844// name is returned. If not nullptr is returned.
5845static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
5846 struct DisassembleInfo *info) {
5847 if (info->bindtable == nullptr) {
5848 info->bindtable = new (BindTable);
5849 SegInfo sectionTable(info->O);
5850 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
5851 uint32_t SegIndex = Entry.segmentIndex();
5852 uint64_t OffsetInSeg = Entry.segmentOffset();
5853 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5854 const char *SymbolName = nullptr;
5855 StringRef name = Entry.symbolName();
5856 if (!name.empty())
5857 SymbolName = name.data();
5858 info->bindtable->push_back(std::make_pair(Address, SymbolName));
5859 }
5860 }
5861 for (bind_table_iterator BI = info->bindtable->begin(),
5862 BE = info->bindtable->end();
5863 BI != BE; ++BI) {
5864 uint64_t Address = BI->first;
5865 if (ReferenceValue == Address) {
5866 const char *SymbolName = BI->second;
5867 return SymbolName;
5868 }
5869 }
5870 return nullptr;
5871}