[clang] Add test for CWG190 "Layout-compatible POD-struct types" (#121668)
[llvm-project.git] / llvm / lib / Object / MachOObjectFile.cpp
blob69d36e6a77db790508a726af93f814326930c98c
1 //===- MachOObjectFile.cpp - Mach-O object file binding -------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file defines the MachOObjectFile class, which binds the MachOObject
10 // class to the generic ObjectFile wrapper.
12 //===----------------------------------------------------------------------===//
14 #include "llvm/ADT/ArrayRef.h"
15 #include "llvm/ADT/STLExtras.h"
16 #include "llvm/ADT/SmallVector.h"
17 #include "llvm/ADT/StringRef.h"
18 #include "llvm/ADT/StringSwitch.h"
19 #include "llvm/ADT/Twine.h"
20 #include "llvm/ADT/bit.h"
21 #include "llvm/BinaryFormat/MachO.h"
22 #include "llvm/BinaryFormat/Swift.h"
23 #include "llvm/Object/Error.h"
24 #include "llvm/Object/MachO.h"
25 #include "llvm/Object/ObjectFile.h"
26 #include "llvm/Object/SymbolicFile.h"
27 #include "llvm/Support/DataExtractor.h"
28 #include "llvm/Support/Debug.h"
29 #include "llvm/Support/Errc.h"
30 #include "llvm/Support/Error.h"
31 #include "llvm/Support/ErrorHandling.h"
32 #include "llvm/Support/FileSystem.h"
33 #include "llvm/Support/Format.h"
34 #include "llvm/Support/LEB128.h"
35 #include "llvm/Support/MemoryBufferRef.h"
36 #include "llvm/Support/Path.h"
37 #include "llvm/Support/SwapByteOrder.h"
38 #include "llvm/Support/raw_ostream.h"
39 #include "llvm/TargetParser/Host.h"
40 #include "llvm/TargetParser/Triple.h"
41 #include <algorithm>
42 #include <cassert>
43 #include <cstddef>
44 #include <cstdint>
45 #include <cstring>
46 #include <limits>
47 #include <list>
48 #include <memory>
49 #include <system_error>
51 using namespace llvm;
52 using namespace object;
54 namespace {
56 struct section_base {
57 char sectname[16];
58 char segname[16];
61 } // end anonymous namespace
63 static Error malformedError(const Twine &Msg) {
64 return make_error<GenericBinaryError>("truncated or malformed object (" +
65 Msg + ")",
66 object_error::parse_failed);
69 // FIXME: Replace all uses of this function with getStructOrErr.
70 template <typename T>
71 static T getStruct(const MachOObjectFile &O, const char *P) {
72 // Don't read before the beginning or past the end of the file
73 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end())
74 report_fatal_error("Malformed MachO file.");
76 T Cmd;
77 memcpy(&Cmd, P, sizeof(T));
78 if (O.isLittleEndian() != sys::IsLittleEndianHost)
79 MachO::swapStruct(Cmd);
80 return Cmd;
83 template <typename T>
84 static Expected<T> getStructOrErr(const MachOObjectFile &O, const char *P) {
85 // Don't read before the beginning or past the end of the file
86 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end())
87 return malformedError("Structure read out-of-range");
89 T Cmd;
90 memcpy(&Cmd, P, sizeof(T));
91 if (O.isLittleEndian() != sys::IsLittleEndianHost)
92 MachO::swapStruct(Cmd);
93 return Cmd;
96 static const char *
97 getSectionPtr(const MachOObjectFile &O, MachOObjectFile::LoadCommandInfo L,
98 unsigned Sec) {
99 uintptr_t CommandAddr = reinterpret_cast<uintptr_t>(L.Ptr);
101 bool Is64 = O.is64Bit();
102 unsigned SegmentLoadSize = Is64 ? sizeof(MachO::segment_command_64) :
103 sizeof(MachO::segment_command);
104 unsigned SectionSize = Is64 ? sizeof(MachO::section_64) :
105 sizeof(MachO::section);
107 uintptr_t SectionAddr = CommandAddr + SegmentLoadSize + Sec * SectionSize;
108 return reinterpret_cast<const char*>(SectionAddr);
111 static const char *getPtr(const MachOObjectFile &O, size_t Offset,
112 size_t MachOFilesetEntryOffset = 0) {
113 assert(Offset <= O.getData().size() &&
114 MachOFilesetEntryOffset <= O.getData().size());
115 return O.getData().data() + Offset + MachOFilesetEntryOffset;
118 static MachO::nlist_base
119 getSymbolTableEntryBase(const MachOObjectFile &O, DataRefImpl DRI) {
120 const char *P = reinterpret_cast<const char *>(DRI.p);
121 return getStruct<MachO::nlist_base>(O, P);
124 static StringRef parseSegmentOrSectionName(const char *P) {
125 if (P[15] == 0)
126 // Null terminated.
127 return P;
128 // Not null terminated, so this is a 16 char string.
129 return StringRef(P, 16);
132 static unsigned getCPUType(const MachOObjectFile &O) {
133 return O.getHeader().cputype;
136 static unsigned getCPUSubType(const MachOObjectFile &O) {
137 return O.getHeader().cpusubtype & ~MachO::CPU_SUBTYPE_MASK;
140 static uint32_t
141 getPlainRelocationAddress(const MachO::any_relocation_info &RE) {
142 return RE.r_word0;
145 static unsigned
146 getScatteredRelocationAddress(const MachO::any_relocation_info &RE) {
147 return RE.r_word0 & 0xffffff;
150 static bool getPlainRelocationPCRel(const MachOObjectFile &O,
151 const MachO::any_relocation_info &RE) {
152 if (O.isLittleEndian())
153 return (RE.r_word1 >> 24) & 1;
154 return (RE.r_word1 >> 7) & 1;
157 static bool
158 getScatteredRelocationPCRel(const MachO::any_relocation_info &RE) {
159 return (RE.r_word0 >> 30) & 1;
162 static unsigned getPlainRelocationLength(const MachOObjectFile &O,
163 const MachO::any_relocation_info &RE) {
164 if (O.isLittleEndian())
165 return (RE.r_word1 >> 25) & 3;
166 return (RE.r_word1 >> 5) & 3;
169 static unsigned
170 getScatteredRelocationLength(const MachO::any_relocation_info &RE) {
171 return (RE.r_word0 >> 28) & 3;
174 static unsigned getPlainRelocationType(const MachOObjectFile &O,
175 const MachO::any_relocation_info &RE) {
176 if (O.isLittleEndian())
177 return RE.r_word1 >> 28;
178 return RE.r_word1 & 0xf;
181 static uint32_t getSectionFlags(const MachOObjectFile &O,
182 DataRefImpl Sec) {
183 if (O.is64Bit()) {
184 MachO::section_64 Sect = O.getSection64(Sec);
185 return Sect.flags;
187 MachO::section Sect = O.getSection(Sec);
188 return Sect.flags;
191 static Expected<MachOObjectFile::LoadCommandInfo>
192 getLoadCommandInfo(const MachOObjectFile &Obj, const char *Ptr,
193 uint32_t LoadCommandIndex) {
194 if (auto CmdOrErr = getStructOrErr<MachO::load_command>(Obj, Ptr)) {
195 if (CmdOrErr->cmdsize + Ptr > Obj.getData().end())
196 return malformedError("load command " + Twine(LoadCommandIndex) +
197 " extends past end of file");
198 if (CmdOrErr->cmdsize < 8)
199 return malformedError("load command " + Twine(LoadCommandIndex) +
200 " with size less than 8 bytes");
201 return MachOObjectFile::LoadCommandInfo({Ptr, *CmdOrErr});
202 } else
203 return CmdOrErr.takeError();
206 static Expected<MachOObjectFile::LoadCommandInfo>
207 getFirstLoadCommandInfo(const MachOObjectFile &Obj) {
208 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64)
209 : sizeof(MachO::mach_header);
210 if (sizeof(MachO::load_command) > Obj.getHeader().sizeofcmds)
211 return malformedError("load command 0 extends past the end all load "
212 "commands in the file");
213 return getLoadCommandInfo(
214 Obj, getPtr(Obj, HeaderSize, Obj.getMachOFilesetEntryOffset()), 0);
217 static Expected<MachOObjectFile::LoadCommandInfo>
218 getNextLoadCommandInfo(const MachOObjectFile &Obj, uint32_t LoadCommandIndex,
219 const MachOObjectFile::LoadCommandInfo &L) {
220 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64)
221 : sizeof(MachO::mach_header);
222 if (L.Ptr + L.C.cmdsize + sizeof(MachO::load_command) >
223 Obj.getData().data() + Obj.getMachOFilesetEntryOffset() + HeaderSize +
224 Obj.getHeader().sizeofcmds)
225 return malformedError("load command " + Twine(LoadCommandIndex + 1) +
226 " extends past the end all load commands in the file");
227 return getLoadCommandInfo(Obj, L.Ptr + L.C.cmdsize, LoadCommandIndex + 1);
230 template <typename T>
231 static void parseHeader(const MachOObjectFile &Obj, T &Header,
232 Error &Err) {
233 if (sizeof(T) > Obj.getData().size()) {
234 Err = malformedError("the mach header extends past the end of the "
235 "file");
236 return;
238 if (auto HeaderOrErr = getStructOrErr<T>(
239 Obj, getPtr(Obj, 0, Obj.getMachOFilesetEntryOffset())))
240 Header = *HeaderOrErr;
241 else
242 Err = HeaderOrErr.takeError();
245 // This is used to check for overlapping of Mach-O elements.
246 struct MachOElement {
247 uint64_t Offset;
248 uint64_t Size;
249 const char *Name;
252 static Error checkOverlappingElement(std::list<MachOElement> &Elements,
253 uint64_t Offset, uint64_t Size,
254 const char *Name) {
255 if (Size == 0)
256 return Error::success();
258 for (auto it = Elements.begin(); it != Elements.end(); ++it) {
259 const auto &E = *it;
260 if ((Offset >= E.Offset && Offset < E.Offset + E.Size) ||
261 (Offset + Size > E.Offset && Offset + Size < E.Offset + E.Size) ||
262 (Offset <= E.Offset && Offset + Size >= E.Offset + E.Size))
263 return malformedError(Twine(Name) + " at offset " + Twine(Offset) +
264 " with a size of " + Twine(Size) + ", overlaps " +
265 E.Name + " at offset " + Twine(E.Offset) + " with "
266 "a size of " + Twine(E.Size));
267 auto nt = it;
268 nt++;
269 if (nt != Elements.end()) {
270 const auto &N = *nt;
271 if (Offset + Size <= N.Offset) {
272 Elements.insert(nt, {Offset, Size, Name});
273 return Error::success();
277 Elements.push_back({Offset, Size, Name});
278 return Error::success();
281 // Parses LC_SEGMENT or LC_SEGMENT_64 load command, adds addresses of all
282 // sections to \param Sections, and optionally sets
283 // \param IsPageZeroSegment to true.
284 template <typename Segment, typename Section>
285 static Error parseSegmentLoadCommand(
286 const MachOObjectFile &Obj, const MachOObjectFile::LoadCommandInfo &Load,
287 SmallVectorImpl<const char *> &Sections, bool &IsPageZeroSegment,
288 uint32_t LoadCommandIndex, const char *CmdName, uint64_t SizeOfHeaders,
289 std::list<MachOElement> &Elements) {
290 const unsigned SegmentLoadSize = sizeof(Segment);
291 if (Load.C.cmdsize < SegmentLoadSize)
292 return malformedError("load command " + Twine(LoadCommandIndex) +
293 " " + CmdName + " cmdsize too small");
294 if (auto SegOrErr = getStructOrErr<Segment>(Obj, Load.Ptr)) {
295 Segment S = SegOrErr.get();
296 const unsigned SectionSize = sizeof(Section);
297 uint64_t FileSize = Obj.getData().size();
298 if (S.nsects > std::numeric_limits<uint32_t>::max() / SectionSize ||
299 S.nsects * SectionSize > Load.C.cmdsize - SegmentLoadSize)
300 return malformedError("load command " + Twine(LoadCommandIndex) +
301 " inconsistent cmdsize in " + CmdName +
302 " for the number of sections");
303 for (unsigned J = 0; J < S.nsects; ++J) {
304 const char *Sec = getSectionPtr(Obj, Load, J);
305 Sections.push_back(Sec);
306 auto SectionOrErr = getStructOrErr<Section>(Obj, Sec);
307 if (!SectionOrErr)
308 return SectionOrErr.takeError();
309 Section s = SectionOrErr.get();
310 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB &&
311 Obj.getHeader().filetype != MachO::MH_DSYM &&
312 s.flags != MachO::S_ZEROFILL &&
313 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL &&
314 s.offset > FileSize)
315 return malformedError("offset field of section " + Twine(J) + " in " +
316 CmdName + " command " + Twine(LoadCommandIndex) +
317 " extends past the end of the file");
318 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB &&
319 Obj.getHeader().filetype != MachO::MH_DSYM &&
320 s.flags != MachO::S_ZEROFILL &&
321 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && S.fileoff == 0 &&
322 s.offset < SizeOfHeaders && s.size != 0)
323 return malformedError("offset field of section " + Twine(J) + " in " +
324 CmdName + " command " + Twine(LoadCommandIndex) +
325 " not past the headers of the file");
326 uint64_t BigSize = s.offset;
327 BigSize += s.size;
328 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB &&
329 Obj.getHeader().filetype != MachO::MH_DSYM &&
330 s.flags != MachO::S_ZEROFILL &&
331 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL &&
332 BigSize > FileSize)
333 return malformedError("offset field plus size field of section " +
334 Twine(J) + " in " + CmdName + " command " +
335 Twine(LoadCommandIndex) +
336 " extends past the end of the file");
337 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB &&
338 Obj.getHeader().filetype != MachO::MH_DSYM &&
339 s.flags != MachO::S_ZEROFILL &&
340 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL &&
341 s.size > S.filesize)
342 return malformedError("size field of section " +
343 Twine(J) + " in " + CmdName + " command " +
344 Twine(LoadCommandIndex) +
345 " greater than the segment");
346 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB &&
347 Obj.getHeader().filetype != MachO::MH_DSYM && s.size != 0 &&
348 s.addr < S.vmaddr)
349 return malformedError("addr field of section " + Twine(J) + " in " +
350 CmdName + " command " + Twine(LoadCommandIndex) +
351 " less than the segment's vmaddr");
352 BigSize = s.addr;
353 BigSize += s.size;
354 uint64_t BigEnd = S.vmaddr;
355 BigEnd += S.vmsize;
356 if (S.vmsize != 0 && s.size != 0 && BigSize > BigEnd)
357 return malformedError("addr field plus size of section " + Twine(J) +
358 " in " + CmdName + " command " +
359 Twine(LoadCommandIndex) +
360 " greater than than "
361 "the segment's vmaddr plus vmsize");
362 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB &&
363 Obj.getHeader().filetype != MachO::MH_DSYM &&
364 s.flags != MachO::S_ZEROFILL &&
365 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL)
366 if (Error Err = checkOverlappingElement(Elements, s.offset, s.size,
367 "section contents"))
368 return Err;
369 if (s.reloff > FileSize)
370 return malformedError("reloff field of section " + Twine(J) + " in " +
371 CmdName + " command " + Twine(LoadCommandIndex) +
372 " extends past the end of the file");
373 BigSize = s.nreloc;
374 BigSize *= sizeof(struct MachO::relocation_info);
375 BigSize += s.reloff;
376 if (BigSize > FileSize)
377 return malformedError("reloff field plus nreloc field times sizeof("
378 "struct relocation_info) of section " +
379 Twine(J) + " in " + CmdName + " command " +
380 Twine(LoadCommandIndex) +
381 " extends past the end of the file");
382 if (Error Err = checkOverlappingElement(Elements, s.reloff, s.nreloc *
383 sizeof(struct
384 MachO::relocation_info),
385 "section relocation entries"))
386 return Err;
388 if (S.fileoff > FileSize)
389 return malformedError("load command " + Twine(LoadCommandIndex) +
390 " fileoff field in " + CmdName +
391 " extends past the end of the file");
392 uint64_t BigSize = S.fileoff;
393 BigSize += S.filesize;
394 if (BigSize > FileSize)
395 return malformedError("load command " + Twine(LoadCommandIndex) +
396 " fileoff field plus filesize field in " +
397 CmdName + " extends past the end of the file");
398 if (S.vmsize != 0 && S.filesize > S.vmsize)
399 return malformedError("load command " + Twine(LoadCommandIndex) +
400 " filesize field in " + CmdName +
401 " greater than vmsize field");
402 IsPageZeroSegment |= StringRef("__PAGEZERO") == S.segname;
403 } else
404 return SegOrErr.takeError();
406 return Error::success();
409 static Error checkSymtabCommand(const MachOObjectFile &Obj,
410 const MachOObjectFile::LoadCommandInfo &Load,
411 uint32_t LoadCommandIndex,
412 const char **SymtabLoadCmd,
413 std::list<MachOElement> &Elements) {
414 if (Load.C.cmdsize < sizeof(MachO::symtab_command))
415 return malformedError("load command " + Twine(LoadCommandIndex) +
416 " LC_SYMTAB cmdsize too small");
417 if (*SymtabLoadCmd != nullptr)
418 return malformedError("more than one LC_SYMTAB command");
419 auto SymtabOrErr = getStructOrErr<MachO::symtab_command>(Obj, Load.Ptr);
420 if (!SymtabOrErr)
421 return SymtabOrErr.takeError();
422 MachO::symtab_command Symtab = SymtabOrErr.get();
423 if (Symtab.cmdsize != sizeof(MachO::symtab_command))
424 return malformedError("LC_SYMTAB command " + Twine(LoadCommandIndex) +
425 " has incorrect cmdsize");
426 uint64_t FileSize = Obj.getData().size();
427 if (Symtab.symoff > FileSize)
428 return malformedError("symoff field of LC_SYMTAB command " +
429 Twine(LoadCommandIndex) + " extends past the end "
430 "of the file");
431 uint64_t SymtabSize = Symtab.nsyms;
432 const char *struct_nlist_name;
433 if (Obj.is64Bit()) {
434 SymtabSize *= sizeof(MachO::nlist_64);
435 struct_nlist_name = "struct nlist_64";
436 } else {
437 SymtabSize *= sizeof(MachO::nlist);
438 struct_nlist_name = "struct nlist";
440 uint64_t BigSize = SymtabSize;
441 BigSize += Symtab.symoff;
442 if (BigSize > FileSize)
443 return malformedError("symoff field plus nsyms field times sizeof(" +
444 Twine(struct_nlist_name) + ") of LC_SYMTAB command " +
445 Twine(LoadCommandIndex) + " extends past the end "
446 "of the file");
447 if (Error Err = checkOverlappingElement(Elements, Symtab.symoff, SymtabSize,
448 "symbol table"))
449 return Err;
450 if (Symtab.stroff > FileSize)
451 return malformedError("stroff field of LC_SYMTAB command " +
452 Twine(LoadCommandIndex) + " extends past the end "
453 "of the file");
454 BigSize = Symtab.stroff;
455 BigSize += Symtab.strsize;
456 if (BigSize > FileSize)
457 return malformedError("stroff field plus strsize field of LC_SYMTAB "
458 "command " + Twine(LoadCommandIndex) + " extends "
459 "past the end of the file");
460 if (Error Err = checkOverlappingElement(Elements, Symtab.stroff,
461 Symtab.strsize, "string table"))
462 return Err;
463 *SymtabLoadCmd = Load.Ptr;
464 return Error::success();
467 static Error checkDysymtabCommand(const MachOObjectFile &Obj,
468 const MachOObjectFile::LoadCommandInfo &Load,
469 uint32_t LoadCommandIndex,
470 const char **DysymtabLoadCmd,
471 std::list<MachOElement> &Elements) {
472 if (Load.C.cmdsize < sizeof(MachO::dysymtab_command))
473 return malformedError("load command " + Twine(LoadCommandIndex) +
474 " LC_DYSYMTAB cmdsize too small");
475 if (*DysymtabLoadCmd != nullptr)
476 return malformedError("more than one LC_DYSYMTAB command");
477 auto DysymtabOrErr =
478 getStructOrErr<MachO::dysymtab_command>(Obj, Load.Ptr);
479 if (!DysymtabOrErr)
480 return DysymtabOrErr.takeError();
481 MachO::dysymtab_command Dysymtab = DysymtabOrErr.get();
482 if (Dysymtab.cmdsize != sizeof(MachO::dysymtab_command))
483 return malformedError("LC_DYSYMTAB command " + Twine(LoadCommandIndex) +
484 " has incorrect cmdsize");
485 uint64_t FileSize = Obj.getData().size();
486 if (Dysymtab.tocoff > FileSize)
487 return malformedError("tocoff field of LC_DYSYMTAB command " +
488 Twine(LoadCommandIndex) + " extends past the end of "
489 "the file");
490 uint64_t BigSize = Dysymtab.ntoc;
491 BigSize *= sizeof(MachO::dylib_table_of_contents);
492 BigSize += Dysymtab.tocoff;
493 if (BigSize > FileSize)
494 return malformedError("tocoff field plus ntoc field times sizeof(struct "
495 "dylib_table_of_contents) of LC_DYSYMTAB command " +
496 Twine(LoadCommandIndex) + " extends past the end of "
497 "the file");
498 if (Error Err = checkOverlappingElement(Elements, Dysymtab.tocoff,
499 Dysymtab.ntoc * sizeof(struct
500 MachO::dylib_table_of_contents),
501 "table of contents"))
502 return Err;
503 if (Dysymtab.modtaboff > FileSize)
504 return malformedError("modtaboff field of LC_DYSYMTAB command " +
505 Twine(LoadCommandIndex) + " extends past the end of "
506 "the file");
507 BigSize = Dysymtab.nmodtab;
508 const char *struct_dylib_module_name;
509 uint64_t sizeof_modtab;
510 if (Obj.is64Bit()) {
511 sizeof_modtab = sizeof(MachO::dylib_module_64);
512 struct_dylib_module_name = "struct dylib_module_64";
513 } else {
514 sizeof_modtab = sizeof(MachO::dylib_module);
515 struct_dylib_module_name = "struct dylib_module";
517 BigSize *= sizeof_modtab;
518 BigSize += Dysymtab.modtaboff;
519 if (BigSize > FileSize)
520 return malformedError("modtaboff field plus nmodtab field times sizeof(" +
521 Twine(struct_dylib_module_name) + ") of LC_DYSYMTAB "
522 "command " + Twine(LoadCommandIndex) + " extends "
523 "past the end of the file");
524 if (Error Err = checkOverlappingElement(Elements, Dysymtab.modtaboff,
525 Dysymtab.nmodtab * sizeof_modtab,
526 "module table"))
527 return Err;
528 if (Dysymtab.extrefsymoff > FileSize)
529 return malformedError("extrefsymoff field of LC_DYSYMTAB command " +
530 Twine(LoadCommandIndex) + " extends past the end of "
531 "the file");
532 BigSize = Dysymtab.nextrefsyms;
533 BigSize *= sizeof(MachO::dylib_reference);
534 BigSize += Dysymtab.extrefsymoff;
535 if (BigSize > FileSize)
536 return malformedError("extrefsymoff field plus nextrefsyms field times "
537 "sizeof(struct dylib_reference) of LC_DYSYMTAB "
538 "command " + Twine(LoadCommandIndex) + " extends "
539 "past the end of the file");
540 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extrefsymoff,
541 Dysymtab.nextrefsyms *
542 sizeof(MachO::dylib_reference),
543 "reference table"))
544 return Err;
545 if (Dysymtab.indirectsymoff > FileSize)
546 return malformedError("indirectsymoff field of LC_DYSYMTAB command " +
547 Twine(LoadCommandIndex) + " extends past the end of "
548 "the file");
549 BigSize = Dysymtab.nindirectsyms;
550 BigSize *= sizeof(uint32_t);
551 BigSize += Dysymtab.indirectsymoff;
552 if (BigSize > FileSize)
553 return malformedError("indirectsymoff field plus nindirectsyms field times "
554 "sizeof(uint32_t) of LC_DYSYMTAB command " +
555 Twine(LoadCommandIndex) + " extends past the end of "
556 "the file");
557 if (Error Err = checkOverlappingElement(Elements, Dysymtab.indirectsymoff,
558 Dysymtab.nindirectsyms *
559 sizeof(uint32_t),
560 "indirect table"))
561 return Err;
562 if (Dysymtab.extreloff > FileSize)
563 return malformedError("extreloff field of LC_DYSYMTAB command " +
564 Twine(LoadCommandIndex) + " extends past the end of "
565 "the file");
566 BigSize = Dysymtab.nextrel;
567 BigSize *= sizeof(MachO::relocation_info);
568 BigSize += Dysymtab.extreloff;
569 if (BigSize > FileSize)
570 return malformedError("extreloff field plus nextrel field times sizeof"
571 "(struct relocation_info) of LC_DYSYMTAB command " +
572 Twine(LoadCommandIndex) + " extends past the end of "
573 "the file");
574 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extreloff,
575 Dysymtab.nextrel *
576 sizeof(MachO::relocation_info),
577 "external relocation table"))
578 return Err;
579 if (Dysymtab.locreloff > FileSize)
580 return malformedError("locreloff field of LC_DYSYMTAB command " +
581 Twine(LoadCommandIndex) + " extends past the end of "
582 "the file");
583 BigSize = Dysymtab.nlocrel;
584 BigSize *= sizeof(MachO::relocation_info);
585 BigSize += Dysymtab.locreloff;
586 if (BigSize > FileSize)
587 return malformedError("locreloff field plus nlocrel field times sizeof"
588 "(struct relocation_info) of LC_DYSYMTAB command " +
589 Twine(LoadCommandIndex) + " extends past the end of "
590 "the file");
591 if (Error Err = checkOverlappingElement(Elements, Dysymtab.locreloff,
592 Dysymtab.nlocrel *
593 sizeof(MachO::relocation_info),
594 "local relocation table"))
595 return Err;
596 *DysymtabLoadCmd = Load.Ptr;
597 return Error::success();
600 static Error checkLinkeditDataCommand(const MachOObjectFile &Obj,
601 const MachOObjectFile::LoadCommandInfo &Load,
602 uint32_t LoadCommandIndex,
603 const char **LoadCmd, const char *CmdName,
604 std::list<MachOElement> &Elements,
605 const char *ElementName) {
606 if (Load.C.cmdsize < sizeof(MachO::linkedit_data_command))
607 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
608 CmdName + " cmdsize too small");
609 if (*LoadCmd != nullptr)
610 return malformedError("more than one " + Twine(CmdName) + " command");
611 auto LinkDataOrError =
612 getStructOrErr<MachO::linkedit_data_command>(Obj, Load.Ptr);
613 if (!LinkDataOrError)
614 return LinkDataOrError.takeError();
615 MachO::linkedit_data_command LinkData = LinkDataOrError.get();
616 if (LinkData.cmdsize != sizeof(MachO::linkedit_data_command))
617 return malformedError(Twine(CmdName) + " command " +
618 Twine(LoadCommandIndex) + " has incorrect cmdsize");
619 uint64_t FileSize = Obj.getData().size();
620 if (LinkData.dataoff > FileSize)
621 return malformedError("dataoff field of " + Twine(CmdName) + " command " +
622 Twine(LoadCommandIndex) + " extends past the end of "
623 "the file");
624 uint64_t BigSize = LinkData.dataoff;
625 BigSize += LinkData.datasize;
626 if (BigSize > FileSize)
627 return malformedError("dataoff field plus datasize field of " +
628 Twine(CmdName) + " command " +
629 Twine(LoadCommandIndex) + " extends past the end of "
630 "the file");
631 if (Error Err = checkOverlappingElement(Elements, LinkData.dataoff,
632 LinkData.datasize, ElementName))
633 return Err;
634 *LoadCmd = Load.Ptr;
635 return Error::success();
638 static Error checkDyldInfoCommand(const MachOObjectFile &Obj,
639 const MachOObjectFile::LoadCommandInfo &Load,
640 uint32_t LoadCommandIndex,
641 const char **LoadCmd, const char *CmdName,
642 std::list<MachOElement> &Elements) {
643 if (Load.C.cmdsize < sizeof(MachO::dyld_info_command))
644 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
645 CmdName + " cmdsize too small");
646 if (*LoadCmd != nullptr)
647 return malformedError("more than one LC_DYLD_INFO and or LC_DYLD_INFO_ONLY "
648 "command");
649 auto DyldInfoOrErr =
650 getStructOrErr<MachO::dyld_info_command>(Obj, Load.Ptr);
651 if (!DyldInfoOrErr)
652 return DyldInfoOrErr.takeError();
653 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get();
654 if (DyldInfo.cmdsize != sizeof(MachO::dyld_info_command))
655 return malformedError(Twine(CmdName) + " command " +
656 Twine(LoadCommandIndex) + " has incorrect cmdsize");
657 uint64_t FileSize = Obj.getData().size();
658 if (DyldInfo.rebase_off > FileSize)
659 return malformedError("rebase_off field of " + Twine(CmdName) +
660 " command " + Twine(LoadCommandIndex) + " extends "
661 "past the end of the file");
662 uint64_t BigSize = DyldInfo.rebase_off;
663 BigSize += DyldInfo.rebase_size;
664 if (BigSize > FileSize)
665 return malformedError("rebase_off field plus rebase_size field of " +
666 Twine(CmdName) + " command " +
667 Twine(LoadCommandIndex) + " extends past the end of "
668 "the file");
669 if (Error Err = checkOverlappingElement(Elements, DyldInfo.rebase_off,
670 DyldInfo.rebase_size,
671 "dyld rebase info"))
672 return Err;
673 if (DyldInfo.bind_off > FileSize)
674 return malformedError("bind_off field of " + Twine(CmdName) +
675 " command " + Twine(LoadCommandIndex) + " extends "
676 "past the end of the file");
677 BigSize = DyldInfo.bind_off;
678 BigSize += DyldInfo.bind_size;
679 if (BigSize > FileSize)
680 return malformedError("bind_off field plus bind_size field of " +
681 Twine(CmdName) + " command " +
682 Twine(LoadCommandIndex) + " extends past the end of "
683 "the file");
684 if (Error Err = checkOverlappingElement(Elements, DyldInfo.bind_off,
685 DyldInfo.bind_size,
686 "dyld bind info"))
687 return Err;
688 if (DyldInfo.weak_bind_off > FileSize)
689 return malformedError("weak_bind_off field of " + Twine(CmdName) +
690 " command " + Twine(LoadCommandIndex) + " extends "
691 "past the end of the file");
692 BigSize = DyldInfo.weak_bind_off;
693 BigSize += DyldInfo.weak_bind_size;
694 if (BigSize > FileSize)
695 return malformedError("weak_bind_off field plus weak_bind_size field of " +
696 Twine(CmdName) + " command " +
697 Twine(LoadCommandIndex) + " extends past the end of "
698 "the file");
699 if (Error Err = checkOverlappingElement(Elements, DyldInfo.weak_bind_off,
700 DyldInfo.weak_bind_size,
701 "dyld weak bind info"))
702 return Err;
703 if (DyldInfo.lazy_bind_off > FileSize)
704 return malformedError("lazy_bind_off field of " + Twine(CmdName) +
705 " command " + Twine(LoadCommandIndex) + " extends "
706 "past the end of the file");
707 BigSize = DyldInfo.lazy_bind_off;
708 BigSize += DyldInfo.lazy_bind_size;
709 if (BigSize > FileSize)
710 return malformedError("lazy_bind_off field plus lazy_bind_size field of " +
711 Twine(CmdName) + " command " +
712 Twine(LoadCommandIndex) + " extends past the end of "
713 "the file");
714 if (Error Err = checkOverlappingElement(Elements, DyldInfo.lazy_bind_off,
715 DyldInfo.lazy_bind_size,
716 "dyld lazy bind info"))
717 return Err;
718 if (DyldInfo.export_off > FileSize)
719 return malformedError("export_off field of " + Twine(CmdName) +
720 " command " + Twine(LoadCommandIndex) + " extends "
721 "past the end of the file");
722 BigSize = DyldInfo.export_off;
723 BigSize += DyldInfo.export_size;
724 if (BigSize > FileSize)
725 return malformedError("export_off field plus export_size field of " +
726 Twine(CmdName) + " command " +
727 Twine(LoadCommandIndex) + " extends past the end of "
728 "the file");
729 if (Error Err = checkOverlappingElement(Elements, DyldInfo.export_off,
730 DyldInfo.export_size,
731 "dyld export info"))
732 return Err;
733 *LoadCmd = Load.Ptr;
734 return Error::success();
737 static Error checkDylibCommand(const MachOObjectFile &Obj,
738 const MachOObjectFile::LoadCommandInfo &Load,
739 uint32_t LoadCommandIndex, const char *CmdName) {
740 if (Load.C.cmdsize < sizeof(MachO::dylib_command))
741 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
742 CmdName + " cmdsize too small");
743 auto CommandOrErr = getStructOrErr<MachO::dylib_command>(Obj, Load.Ptr);
744 if (!CommandOrErr)
745 return CommandOrErr.takeError();
746 MachO::dylib_command D = CommandOrErr.get();
747 if (D.dylib.name < sizeof(MachO::dylib_command))
748 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
749 CmdName + " name.offset field too small, not past "
750 "the end of the dylib_command struct");
751 if (D.dylib.name >= D.cmdsize)
752 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
753 CmdName + " name.offset field extends past the end "
754 "of the load command");
755 // Make sure there is a null between the starting offset of the name and
756 // the end of the load command.
757 uint32_t i;
758 const char *P = (const char *)Load.Ptr;
759 for (i = D.dylib.name; i < D.cmdsize; i++)
760 if (P[i] == '\0')
761 break;
762 if (i >= D.cmdsize)
763 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
764 CmdName + " library name extends past the end of the "
765 "load command");
766 return Error::success();
769 static Error checkDylibIdCommand(const MachOObjectFile &Obj,
770 const MachOObjectFile::LoadCommandInfo &Load,
771 uint32_t LoadCommandIndex,
772 const char **LoadCmd) {
773 if (Error Err = checkDylibCommand(Obj, Load, LoadCommandIndex,
774 "LC_ID_DYLIB"))
775 return Err;
776 if (*LoadCmd != nullptr)
777 return malformedError("more than one LC_ID_DYLIB command");
778 if (Obj.getHeader().filetype != MachO::MH_DYLIB &&
779 Obj.getHeader().filetype != MachO::MH_DYLIB_STUB)
780 return malformedError("LC_ID_DYLIB load command in non-dynamic library "
781 "file type");
782 *LoadCmd = Load.Ptr;
783 return Error::success();
786 static Error checkDyldCommand(const MachOObjectFile &Obj,
787 const MachOObjectFile::LoadCommandInfo &Load,
788 uint32_t LoadCommandIndex, const char *CmdName) {
789 if (Load.C.cmdsize < sizeof(MachO::dylinker_command))
790 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
791 CmdName + " cmdsize too small");
792 auto CommandOrErr = getStructOrErr<MachO::dylinker_command>(Obj, Load.Ptr);
793 if (!CommandOrErr)
794 return CommandOrErr.takeError();
795 MachO::dylinker_command D = CommandOrErr.get();
796 if (D.name < sizeof(MachO::dylinker_command))
797 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
798 CmdName + " name.offset field too small, not past "
799 "the end of the dylinker_command struct");
800 if (D.name >= D.cmdsize)
801 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
802 CmdName + " name.offset field extends past the end "
803 "of the load command");
804 // Make sure there is a null between the starting offset of the name and
805 // the end of the load command.
806 uint32_t i;
807 const char *P = (const char *)Load.Ptr;
808 for (i = D.name; i < D.cmdsize; i++)
809 if (P[i] == '\0')
810 break;
811 if (i >= D.cmdsize)
812 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
813 CmdName + " dyld name extends past the end of the "
814 "load command");
815 return Error::success();
818 static Error checkVersCommand(const MachOObjectFile &Obj,
819 const MachOObjectFile::LoadCommandInfo &Load,
820 uint32_t LoadCommandIndex,
821 const char **LoadCmd, const char *CmdName) {
822 if (Load.C.cmdsize != sizeof(MachO::version_min_command))
823 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
824 CmdName + " has incorrect cmdsize");
825 if (*LoadCmd != nullptr)
826 return malformedError("more than one LC_VERSION_MIN_MACOSX, "
827 "LC_VERSION_MIN_IPHONEOS, LC_VERSION_MIN_TVOS or "
828 "LC_VERSION_MIN_WATCHOS command");
829 *LoadCmd = Load.Ptr;
830 return Error::success();
833 static Error checkNoteCommand(const MachOObjectFile &Obj,
834 const MachOObjectFile::LoadCommandInfo &Load,
835 uint32_t LoadCommandIndex,
836 std::list<MachOElement> &Elements) {
837 if (Load.C.cmdsize != sizeof(MachO::note_command))
838 return malformedError("load command " + Twine(LoadCommandIndex) +
839 " LC_NOTE has incorrect cmdsize");
840 auto NoteCmdOrErr = getStructOrErr<MachO::note_command>(Obj, Load.Ptr);
841 if (!NoteCmdOrErr)
842 return NoteCmdOrErr.takeError();
843 MachO::note_command Nt = NoteCmdOrErr.get();
844 uint64_t FileSize = Obj.getData().size();
845 if (Nt.offset > FileSize)
846 return malformedError("offset field of LC_NOTE command " +
847 Twine(LoadCommandIndex) + " extends "
848 "past the end of the file");
849 uint64_t BigSize = Nt.offset;
850 BigSize += Nt.size;
851 if (BigSize > FileSize)
852 return malformedError("size field plus offset field of LC_NOTE command " +
853 Twine(LoadCommandIndex) + " extends past the end of "
854 "the file");
855 if (Error Err = checkOverlappingElement(Elements, Nt.offset, Nt.size,
856 "LC_NOTE data"))
857 return Err;
858 return Error::success();
861 static Error
862 parseBuildVersionCommand(const MachOObjectFile &Obj,
863 const MachOObjectFile::LoadCommandInfo &Load,
864 SmallVectorImpl<const char*> &BuildTools,
865 uint32_t LoadCommandIndex) {
866 auto BVCOrErr =
867 getStructOrErr<MachO::build_version_command>(Obj, Load.Ptr);
868 if (!BVCOrErr)
869 return BVCOrErr.takeError();
870 MachO::build_version_command BVC = BVCOrErr.get();
871 if (Load.C.cmdsize !=
872 sizeof(MachO::build_version_command) +
873 BVC.ntools * sizeof(MachO::build_tool_version))
874 return malformedError("load command " + Twine(LoadCommandIndex) +
875 " LC_BUILD_VERSION_COMMAND has incorrect cmdsize");
877 auto Start = Load.Ptr + sizeof(MachO::build_version_command);
878 BuildTools.resize(BVC.ntools);
879 for (unsigned i = 0; i < BVC.ntools; ++i)
880 BuildTools[i] = Start + i * sizeof(MachO::build_tool_version);
882 return Error::success();
885 static Error checkRpathCommand(const MachOObjectFile &Obj,
886 const MachOObjectFile::LoadCommandInfo &Load,
887 uint32_t LoadCommandIndex) {
888 if (Load.C.cmdsize < sizeof(MachO::rpath_command))
889 return malformedError("load command " + Twine(LoadCommandIndex) +
890 " LC_RPATH cmdsize too small");
891 auto ROrErr = getStructOrErr<MachO::rpath_command>(Obj, Load.Ptr);
892 if (!ROrErr)
893 return ROrErr.takeError();
894 MachO::rpath_command R = ROrErr.get();
895 if (R.path < sizeof(MachO::rpath_command))
896 return malformedError("load command " + Twine(LoadCommandIndex) +
897 " LC_RPATH path.offset field too small, not past "
898 "the end of the rpath_command struct");
899 if (R.path >= R.cmdsize)
900 return malformedError("load command " + Twine(LoadCommandIndex) +
901 " LC_RPATH path.offset field extends past the end "
902 "of the load command");
903 // Make sure there is a null between the starting offset of the path and
904 // the end of the load command.
905 uint32_t i;
906 const char *P = (const char *)Load.Ptr;
907 for (i = R.path; i < R.cmdsize; i++)
908 if (P[i] == '\0')
909 break;
910 if (i >= R.cmdsize)
911 return malformedError("load command " + Twine(LoadCommandIndex) +
912 " LC_RPATH library name extends past the end of the "
913 "load command");
914 return Error::success();
917 static Error checkEncryptCommand(const MachOObjectFile &Obj,
918 const MachOObjectFile::LoadCommandInfo &Load,
919 uint32_t LoadCommandIndex,
920 uint64_t cryptoff, uint64_t cryptsize,
921 const char **LoadCmd, const char *CmdName) {
922 if (*LoadCmd != nullptr)
923 return malformedError("more than one LC_ENCRYPTION_INFO and or "
924 "LC_ENCRYPTION_INFO_64 command");
925 uint64_t FileSize = Obj.getData().size();
926 if (cryptoff > FileSize)
927 return malformedError("cryptoff field of " + Twine(CmdName) +
928 " command " + Twine(LoadCommandIndex) + " extends "
929 "past the end of the file");
930 uint64_t BigSize = cryptoff;
931 BigSize += cryptsize;
932 if (BigSize > FileSize)
933 return malformedError("cryptoff field plus cryptsize field of " +
934 Twine(CmdName) + " command " +
935 Twine(LoadCommandIndex) + " extends past the end of "
936 "the file");
937 *LoadCmd = Load.Ptr;
938 return Error::success();
941 static Error checkLinkerOptCommand(const MachOObjectFile &Obj,
942 const MachOObjectFile::LoadCommandInfo &Load,
943 uint32_t LoadCommandIndex) {
944 if (Load.C.cmdsize < sizeof(MachO::linker_option_command))
945 return malformedError("load command " + Twine(LoadCommandIndex) +
946 " LC_LINKER_OPTION cmdsize too small");
947 auto LinkOptionOrErr =
948 getStructOrErr<MachO::linker_option_command>(Obj, Load.Ptr);
949 if (!LinkOptionOrErr)
950 return LinkOptionOrErr.takeError();
951 MachO::linker_option_command L = LinkOptionOrErr.get();
952 // Make sure the count of strings is correct.
953 const char *string = (const char *)Load.Ptr +
954 sizeof(struct MachO::linker_option_command);
955 uint32_t left = L.cmdsize - sizeof(struct MachO::linker_option_command);
956 uint32_t i = 0;
957 while (left > 0) {
958 while (*string == '\0' && left > 0) {
959 string++;
960 left--;
962 if (left > 0) {
963 i++;
964 uint32_t NullPos = StringRef(string, left).find('\0');
965 if (0xffffffff == NullPos)
966 return malformedError("load command " + Twine(LoadCommandIndex) +
967 " LC_LINKER_OPTION string #" + Twine(i) +
968 " is not NULL terminated");
969 uint32_t len = std::min(NullPos, left) + 1;
970 string += len;
971 left -= len;
974 if (L.count != i)
975 return malformedError("load command " + Twine(LoadCommandIndex) +
976 " LC_LINKER_OPTION string count " + Twine(L.count) +
977 " does not match number of strings");
978 return Error::success();
981 static Error checkSubCommand(const MachOObjectFile &Obj,
982 const MachOObjectFile::LoadCommandInfo &Load,
983 uint32_t LoadCommandIndex, const char *CmdName,
984 size_t SizeOfCmd, const char *CmdStructName,
985 uint32_t PathOffset, const char *PathFieldName) {
986 if (PathOffset < SizeOfCmd)
987 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
988 CmdName + " " + PathFieldName + ".offset field too "
989 "small, not past the end of the " + CmdStructName);
990 if (PathOffset >= Load.C.cmdsize)
991 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
992 CmdName + " " + PathFieldName + ".offset field "
993 "extends past the end of the load command");
994 // Make sure there is a null between the starting offset of the path and
995 // the end of the load command.
996 uint32_t i;
997 const char *P = (const char *)Load.Ptr;
998 for (i = PathOffset; i < Load.C.cmdsize; i++)
999 if (P[i] == '\0')
1000 break;
1001 if (i >= Load.C.cmdsize)
1002 return malformedError("load command " + Twine(LoadCommandIndex) + " " +
1003 CmdName + " " + PathFieldName + " name extends past "
1004 "the end of the load command");
1005 return Error::success();
1008 static Error checkThreadCommand(const MachOObjectFile &Obj,
1009 const MachOObjectFile::LoadCommandInfo &Load,
1010 uint32_t LoadCommandIndex,
1011 const char *CmdName) {
1012 if (Load.C.cmdsize < sizeof(MachO::thread_command))
1013 return malformedError("load command " + Twine(LoadCommandIndex) +
1014 CmdName + " cmdsize too small");
1015 auto ThreadCommandOrErr =
1016 getStructOrErr<MachO::thread_command>(Obj, Load.Ptr);
1017 if (!ThreadCommandOrErr)
1018 return ThreadCommandOrErr.takeError();
1019 MachO::thread_command T = ThreadCommandOrErr.get();
1020 const char *state = Load.Ptr + sizeof(MachO::thread_command);
1021 const char *end = Load.Ptr + T.cmdsize;
1022 uint32_t nflavor = 0;
1023 uint32_t cputype = getCPUType(Obj);
1024 while (state < end) {
1025 if(state + sizeof(uint32_t) > end)
1026 return malformedError("load command " + Twine(LoadCommandIndex) +
1027 "flavor in " + CmdName + " extends past end of "
1028 "command");
1029 uint32_t flavor;
1030 memcpy(&flavor, state, sizeof(uint32_t));
1031 if (Obj.isLittleEndian() != sys::IsLittleEndianHost)
1032 sys::swapByteOrder(flavor);
1033 state += sizeof(uint32_t);
1035 if(state + sizeof(uint32_t) > end)
1036 return malformedError("load command " + Twine(LoadCommandIndex) +
1037 " count in " + CmdName + " extends past end of "
1038 "command");
1039 uint32_t count;
1040 memcpy(&count, state, sizeof(uint32_t));
1041 if (Obj.isLittleEndian() != sys::IsLittleEndianHost)
1042 sys::swapByteOrder(count);
1043 state += sizeof(uint32_t);
1045 if (cputype == MachO::CPU_TYPE_I386) {
1046 if (flavor == MachO::x86_THREAD_STATE32) {
1047 if (count != MachO::x86_THREAD_STATE32_COUNT)
1048 return malformedError("load command " + Twine(LoadCommandIndex) +
1049 " count not x86_THREAD_STATE32_COUNT for "
1050 "flavor number " + Twine(nflavor) + " which is "
1051 "a x86_THREAD_STATE32 flavor in " + CmdName +
1052 " command");
1053 if (state + sizeof(MachO::x86_thread_state32_t) > end)
1054 return malformedError("load command " + Twine(LoadCommandIndex) +
1055 " x86_THREAD_STATE32 extends past end of "
1056 "command in " + CmdName + " command");
1057 state += sizeof(MachO::x86_thread_state32_t);
1058 } else {
1059 return malformedError("load command " + Twine(LoadCommandIndex) +
1060 " unknown flavor (" + Twine(flavor) + ") for "
1061 "flavor number " + Twine(nflavor) + " in " +
1062 CmdName + " command");
1064 } else if (cputype == MachO::CPU_TYPE_X86_64) {
1065 if (flavor == MachO::x86_THREAD_STATE) {
1066 if (count != MachO::x86_THREAD_STATE_COUNT)
1067 return malformedError("load command " + Twine(LoadCommandIndex) +
1068 " count not x86_THREAD_STATE_COUNT for "
1069 "flavor number " + Twine(nflavor) + " which is "
1070 "a x86_THREAD_STATE flavor in " + CmdName +
1071 " command");
1072 if (state + sizeof(MachO::x86_thread_state_t) > end)
1073 return malformedError("load command " + Twine(LoadCommandIndex) +
1074 " x86_THREAD_STATE extends past end of "
1075 "command in " + CmdName + " command");
1076 state += sizeof(MachO::x86_thread_state_t);
1077 } else if (flavor == MachO::x86_FLOAT_STATE) {
1078 if (count != MachO::x86_FLOAT_STATE_COUNT)
1079 return malformedError("load command " + Twine(LoadCommandIndex) +
1080 " count not x86_FLOAT_STATE_COUNT for "
1081 "flavor number " + Twine(nflavor) + " which is "
1082 "a x86_FLOAT_STATE flavor in " + CmdName +
1083 " command");
1084 if (state + sizeof(MachO::x86_float_state_t) > end)
1085 return malformedError("load command " + Twine(LoadCommandIndex) +
1086 " x86_FLOAT_STATE extends past end of "
1087 "command in " + CmdName + " command");
1088 state += sizeof(MachO::x86_float_state_t);
1089 } else if (flavor == MachO::x86_EXCEPTION_STATE) {
1090 if (count != MachO::x86_EXCEPTION_STATE_COUNT)
1091 return malformedError("load command " + Twine(LoadCommandIndex) +
1092 " count not x86_EXCEPTION_STATE_COUNT for "
1093 "flavor number " + Twine(nflavor) + " which is "
1094 "a x86_EXCEPTION_STATE flavor in " + CmdName +
1095 " command");
1096 if (state + sizeof(MachO::x86_exception_state_t) > end)
1097 return malformedError("load command " + Twine(LoadCommandIndex) +
1098 " x86_EXCEPTION_STATE extends past end of "
1099 "command in " + CmdName + " command");
1100 state += sizeof(MachO::x86_exception_state_t);
1101 } else if (flavor == MachO::x86_THREAD_STATE64) {
1102 if (count != MachO::x86_THREAD_STATE64_COUNT)
1103 return malformedError("load command " + Twine(LoadCommandIndex) +
1104 " count not x86_THREAD_STATE64_COUNT for "
1105 "flavor number " + Twine(nflavor) + " which is "
1106 "a x86_THREAD_STATE64 flavor in " + CmdName +
1107 " command");
1108 if (state + sizeof(MachO::x86_thread_state64_t) > end)
1109 return malformedError("load command " + Twine(LoadCommandIndex) +
1110 " x86_THREAD_STATE64 extends past end of "
1111 "command in " + CmdName + " command");
1112 state += sizeof(MachO::x86_thread_state64_t);
1113 } else if (flavor == MachO::x86_EXCEPTION_STATE64) {
1114 if (count != MachO::x86_EXCEPTION_STATE64_COUNT)
1115 return malformedError("load command " + Twine(LoadCommandIndex) +
1116 " count not x86_EXCEPTION_STATE64_COUNT for "
1117 "flavor number " + Twine(nflavor) + " which is "
1118 "a x86_EXCEPTION_STATE64 flavor in " + CmdName +
1119 " command");
1120 if (state + sizeof(MachO::x86_exception_state64_t) > end)
1121 return malformedError("load command " + Twine(LoadCommandIndex) +
1122 " x86_EXCEPTION_STATE64 extends past end of "
1123 "command in " + CmdName + " command");
1124 state += sizeof(MachO::x86_exception_state64_t);
1125 } else {
1126 return malformedError("load command " + Twine(LoadCommandIndex) +
1127 " unknown flavor (" + Twine(flavor) + ") for "
1128 "flavor number " + Twine(nflavor) + " in " +
1129 CmdName + " command");
1131 } else if (cputype == MachO::CPU_TYPE_ARM) {
1132 if (flavor == MachO::ARM_THREAD_STATE) {
1133 if (count != MachO::ARM_THREAD_STATE_COUNT)
1134 return malformedError("load command " + Twine(LoadCommandIndex) +
1135 " count not ARM_THREAD_STATE_COUNT for "
1136 "flavor number " + Twine(nflavor) + " which is "
1137 "a ARM_THREAD_STATE flavor in " + CmdName +
1138 " command");
1139 if (state + sizeof(MachO::arm_thread_state32_t) > end)
1140 return malformedError("load command " + Twine(LoadCommandIndex) +
1141 " ARM_THREAD_STATE extends past end of "
1142 "command in " + CmdName + " command");
1143 state += sizeof(MachO::arm_thread_state32_t);
1144 } else {
1145 return malformedError("load command " + Twine(LoadCommandIndex) +
1146 " unknown flavor (" + Twine(flavor) + ") for "
1147 "flavor number " + Twine(nflavor) + " in " +
1148 CmdName + " command");
1150 } else if (cputype == MachO::CPU_TYPE_ARM64 ||
1151 cputype == MachO::CPU_TYPE_ARM64_32) {
1152 if (flavor == MachO::ARM_THREAD_STATE64) {
1153 if (count != MachO::ARM_THREAD_STATE64_COUNT)
1154 return malformedError("load command " + Twine(LoadCommandIndex) +
1155 " count not ARM_THREAD_STATE64_COUNT for "
1156 "flavor number " + Twine(nflavor) + " which is "
1157 "a ARM_THREAD_STATE64 flavor in " + CmdName +
1158 " command");
1159 if (state + sizeof(MachO::arm_thread_state64_t) > end)
1160 return malformedError("load command " + Twine(LoadCommandIndex) +
1161 " ARM_THREAD_STATE64 extends past end of "
1162 "command in " + CmdName + " command");
1163 state += sizeof(MachO::arm_thread_state64_t);
1164 } else {
1165 return malformedError("load command " + Twine(LoadCommandIndex) +
1166 " unknown flavor (" + Twine(flavor) + ") for "
1167 "flavor number " + Twine(nflavor) + " in " +
1168 CmdName + " command");
1170 } else if (cputype == MachO::CPU_TYPE_POWERPC) {
1171 if (flavor == MachO::PPC_THREAD_STATE) {
1172 if (count != MachO::PPC_THREAD_STATE_COUNT)
1173 return malformedError("load command " + Twine(LoadCommandIndex) +
1174 " count not PPC_THREAD_STATE_COUNT for "
1175 "flavor number " + Twine(nflavor) + " which is "
1176 "a PPC_THREAD_STATE flavor in " + CmdName +
1177 " command");
1178 if (state + sizeof(MachO::ppc_thread_state32_t) > end)
1179 return malformedError("load command " + Twine(LoadCommandIndex) +
1180 " PPC_THREAD_STATE extends past end of "
1181 "command in " + CmdName + " command");
1182 state += sizeof(MachO::ppc_thread_state32_t);
1183 } else {
1184 return malformedError("load command " + Twine(LoadCommandIndex) +
1185 " unknown flavor (" + Twine(flavor) + ") for "
1186 "flavor number " + Twine(nflavor) + " in " +
1187 CmdName + " command");
1189 } else {
1190 return malformedError("unknown cputype (" + Twine(cputype) + ") load "
1191 "command " + Twine(LoadCommandIndex) + " for " +
1192 CmdName + " command can't be checked");
1194 nflavor++;
1196 return Error::success();
1199 static Error checkTwoLevelHintsCommand(const MachOObjectFile &Obj,
1200 const MachOObjectFile::LoadCommandInfo
1201 &Load,
1202 uint32_t LoadCommandIndex,
1203 const char **LoadCmd,
1204 std::list<MachOElement> &Elements) {
1205 if (Load.C.cmdsize != sizeof(MachO::twolevel_hints_command))
1206 return malformedError("load command " + Twine(LoadCommandIndex) +
1207 " LC_TWOLEVEL_HINTS has incorrect cmdsize");
1208 if (*LoadCmd != nullptr)
1209 return malformedError("more than one LC_TWOLEVEL_HINTS command");
1210 auto HintsOrErr = getStructOrErr<MachO::twolevel_hints_command>(Obj, Load.Ptr);
1211 if(!HintsOrErr)
1212 return HintsOrErr.takeError();
1213 MachO::twolevel_hints_command Hints = HintsOrErr.get();
1214 uint64_t FileSize = Obj.getData().size();
1215 if (Hints.offset > FileSize)
1216 return malformedError("offset field of LC_TWOLEVEL_HINTS command " +
1217 Twine(LoadCommandIndex) + " extends past the end of "
1218 "the file");
1219 uint64_t BigSize = Hints.nhints;
1220 BigSize *= sizeof(MachO::twolevel_hint);
1221 BigSize += Hints.offset;
1222 if (BigSize > FileSize)
1223 return malformedError("offset field plus nhints times sizeof(struct "
1224 "twolevel_hint) field of LC_TWOLEVEL_HINTS command " +
1225 Twine(LoadCommandIndex) + " extends past the end of "
1226 "the file");
1227 if (Error Err = checkOverlappingElement(Elements, Hints.offset, Hints.nhints *
1228 sizeof(MachO::twolevel_hint),
1229 "two level hints"))
1230 return Err;
1231 *LoadCmd = Load.Ptr;
1232 return Error::success();
1235 // Returns true if the libObject code does not support the load command and its
1236 // contents. The cmd value it is treated as an unknown load command but with
1237 // an error message that says the cmd value is obsolete.
1238 static bool isLoadCommandObsolete(uint32_t cmd) {
1239 if (cmd == MachO::LC_SYMSEG ||
1240 cmd == MachO::LC_LOADFVMLIB ||
1241 cmd == MachO::LC_IDFVMLIB ||
1242 cmd == MachO::LC_IDENT ||
1243 cmd == MachO::LC_FVMFILE ||
1244 cmd == MachO::LC_PREPAGE ||
1245 cmd == MachO::LC_PREBOUND_DYLIB ||
1246 cmd == MachO::LC_TWOLEVEL_HINTS ||
1247 cmd == MachO::LC_PREBIND_CKSUM)
1248 return true;
1249 return false;
1252 Expected<std::unique_ptr<MachOObjectFile>>
1253 MachOObjectFile::create(MemoryBufferRef Object, bool IsLittleEndian,
1254 bool Is64Bits, uint32_t UniversalCputype,
1255 uint32_t UniversalIndex,
1256 size_t MachOFilesetEntryOffset) {
1257 Error Err = Error::success();
1258 std::unique_ptr<MachOObjectFile> Obj(new MachOObjectFile(
1259 std::move(Object), IsLittleEndian, Is64Bits, Err, UniversalCputype,
1260 UniversalIndex, MachOFilesetEntryOffset));
1261 if (Err)
1262 return std::move(Err);
1263 return std::move(Obj);
1266 MachOObjectFile::MachOObjectFile(MemoryBufferRef Object, bool IsLittleEndian,
1267 bool Is64bits, Error &Err,
1268 uint32_t UniversalCputype,
1269 uint32_t UniversalIndex,
1270 size_t MachOFilesetEntryOffset)
1271 : ObjectFile(getMachOType(IsLittleEndian, Is64bits), Object),
1272 MachOFilesetEntryOffset(MachOFilesetEntryOffset) {
1273 ErrorAsOutParameter ErrAsOutParam(Err);
1274 uint64_t SizeOfHeaders;
1275 uint32_t cputype;
1276 if (is64Bit()) {
1277 parseHeader(*this, Header64, Err);
1278 SizeOfHeaders = sizeof(MachO::mach_header_64);
1279 cputype = Header64.cputype;
1280 } else {
1281 parseHeader(*this, Header, Err);
1282 SizeOfHeaders = sizeof(MachO::mach_header);
1283 cputype = Header.cputype;
1285 if (Err)
1286 return;
1287 SizeOfHeaders += getHeader().sizeofcmds;
1288 if (getData().data() + SizeOfHeaders > getData().end()) {
1289 Err = malformedError("load commands extend past the end of the file");
1290 return;
1292 if (UniversalCputype != 0 && cputype != UniversalCputype) {
1293 Err = malformedError("universal header architecture: " +
1294 Twine(UniversalIndex) + "'s cputype does not match "
1295 "object file's mach header");
1296 return;
1298 std::list<MachOElement> Elements;
1299 Elements.push_back({0, SizeOfHeaders, "Mach-O headers"});
1301 uint32_t LoadCommandCount = getHeader().ncmds;
1302 LoadCommandInfo Load;
1303 if (LoadCommandCount != 0) {
1304 if (auto LoadOrErr = getFirstLoadCommandInfo(*this))
1305 Load = *LoadOrErr;
1306 else {
1307 Err = LoadOrErr.takeError();
1308 return;
1312 const char *DyldIdLoadCmd = nullptr;
1313 const char *SplitInfoLoadCmd = nullptr;
1314 const char *CodeSignDrsLoadCmd = nullptr;
1315 const char *CodeSignLoadCmd = nullptr;
1316 const char *VersLoadCmd = nullptr;
1317 const char *SourceLoadCmd = nullptr;
1318 const char *EntryPointLoadCmd = nullptr;
1319 const char *EncryptLoadCmd = nullptr;
1320 const char *RoutinesLoadCmd = nullptr;
1321 const char *UnixThreadLoadCmd = nullptr;
1322 const char *TwoLevelHintsLoadCmd = nullptr;
1323 for (unsigned I = 0; I < LoadCommandCount; ++I) {
1324 if (is64Bit()) {
1325 if (Load.C.cmdsize % 8 != 0) {
1326 // We have a hack here to allow 64-bit Mach-O core files to have
1327 // LC_THREAD commands that are only a multiple of 4 and not 8 to be
1328 // allowed since the macOS kernel produces them.
1329 if (getHeader().filetype != MachO::MH_CORE ||
1330 Load.C.cmd != MachO::LC_THREAD || Load.C.cmdsize % 4) {
1331 Err = malformedError("load command " + Twine(I) + " cmdsize not a "
1332 "multiple of 8");
1333 return;
1336 } else {
1337 if (Load.C.cmdsize % 4 != 0) {
1338 Err = malformedError("load command " + Twine(I) + " cmdsize not a "
1339 "multiple of 4");
1340 return;
1343 LoadCommands.push_back(Load);
1344 if (Load.C.cmd == MachO::LC_SYMTAB) {
1345 if ((Err = checkSymtabCommand(*this, Load, I, &SymtabLoadCmd, Elements)))
1346 return;
1347 } else if (Load.C.cmd == MachO::LC_DYSYMTAB) {
1348 if ((Err = checkDysymtabCommand(*this, Load, I, &DysymtabLoadCmd,
1349 Elements)))
1350 return;
1351 } else if (Load.C.cmd == MachO::LC_DATA_IN_CODE) {
1352 if ((Err = checkLinkeditDataCommand(*this, Load, I, &DataInCodeLoadCmd,
1353 "LC_DATA_IN_CODE", Elements,
1354 "data in code info")))
1355 return;
1356 } else if (Load.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
1357 if ((Err = checkLinkeditDataCommand(*this, Load, I, &LinkOptHintsLoadCmd,
1358 "LC_LINKER_OPTIMIZATION_HINT",
1359 Elements, "linker optimization "
1360 "hints")))
1361 return;
1362 } else if (Load.C.cmd == MachO::LC_FUNCTION_STARTS) {
1363 if ((Err = checkLinkeditDataCommand(*this, Load, I, &FuncStartsLoadCmd,
1364 "LC_FUNCTION_STARTS", Elements,
1365 "function starts data")))
1366 return;
1367 } else if (Load.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO) {
1368 if ((Err = checkLinkeditDataCommand(*this, Load, I, &SplitInfoLoadCmd,
1369 "LC_SEGMENT_SPLIT_INFO", Elements,
1370 "split info data")))
1371 return;
1372 } else if (Load.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS) {
1373 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignDrsLoadCmd,
1374 "LC_DYLIB_CODE_SIGN_DRS", Elements,
1375 "code signing RDs data")))
1376 return;
1377 } else if (Load.C.cmd == MachO::LC_CODE_SIGNATURE) {
1378 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignLoadCmd,
1379 "LC_CODE_SIGNATURE", Elements,
1380 "code signature data")))
1381 return;
1382 } else if (Load.C.cmd == MachO::LC_DYLD_INFO) {
1383 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd,
1384 "LC_DYLD_INFO", Elements)))
1385 return;
1386 } else if (Load.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
1387 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd,
1388 "LC_DYLD_INFO_ONLY", Elements)))
1389 return;
1390 } else if (Load.C.cmd == MachO::LC_DYLD_CHAINED_FIXUPS) {
1391 if ((Err = checkLinkeditDataCommand(
1392 *this, Load, I, &DyldChainedFixupsLoadCmd,
1393 "LC_DYLD_CHAINED_FIXUPS", Elements, "chained fixups")))
1394 return;
1395 } else if (Load.C.cmd == MachO::LC_DYLD_EXPORTS_TRIE) {
1396 if ((Err = checkLinkeditDataCommand(
1397 *this, Load, I, &DyldExportsTrieLoadCmd, "LC_DYLD_EXPORTS_TRIE",
1398 Elements, "exports trie")))
1399 return;
1400 } else if (Load.C.cmd == MachO::LC_UUID) {
1401 if (Load.C.cmdsize != sizeof(MachO::uuid_command)) {
1402 Err = malformedError("LC_UUID command " + Twine(I) + " has incorrect "
1403 "cmdsize");
1404 return;
1406 if (UuidLoadCmd) {
1407 Err = malformedError("more than one LC_UUID command");
1408 return;
1410 UuidLoadCmd = Load.Ptr;
1411 } else if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1412 if ((Err = parseSegmentLoadCommand<MachO::segment_command_64,
1413 MachO::section_64>(
1414 *this, Load, Sections, HasPageZeroSegment, I,
1415 "LC_SEGMENT_64", SizeOfHeaders, Elements)))
1416 return;
1417 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1418 if ((Err = parseSegmentLoadCommand<MachO::segment_command,
1419 MachO::section>(
1420 *this, Load, Sections, HasPageZeroSegment, I,
1421 "LC_SEGMENT", SizeOfHeaders, Elements)))
1422 return;
1423 } else if (Load.C.cmd == MachO::LC_ID_DYLIB) {
1424 if ((Err = checkDylibIdCommand(*this, Load, I, &DyldIdLoadCmd)))
1425 return;
1426 } else if (Load.C.cmd == MachO::LC_LOAD_DYLIB) {
1427 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_DYLIB")))
1428 return;
1429 Libraries.push_back(Load.Ptr);
1430 } else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB) {
1431 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_WEAK_DYLIB")))
1432 return;
1433 Libraries.push_back(Load.Ptr);
1434 } else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB) {
1435 if ((Err = checkDylibCommand(*this, Load, I, "LC_LAZY_LOAD_DYLIB")))
1436 return;
1437 Libraries.push_back(Load.Ptr);
1438 } else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB) {
1439 if ((Err = checkDylibCommand(*this, Load, I, "LC_REEXPORT_DYLIB")))
1440 return;
1441 Libraries.push_back(Load.Ptr);
1442 } else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
1443 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_UPWARD_DYLIB")))
1444 return;
1445 Libraries.push_back(Load.Ptr);
1446 } else if (Load.C.cmd == MachO::LC_ID_DYLINKER) {
1447 if ((Err = checkDyldCommand(*this, Load, I, "LC_ID_DYLINKER")))
1448 return;
1449 } else if (Load.C.cmd == MachO::LC_LOAD_DYLINKER) {
1450 if ((Err = checkDyldCommand(*this, Load, I, "LC_LOAD_DYLINKER")))
1451 return;
1452 } else if (Load.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
1453 if ((Err = checkDyldCommand(*this, Load, I, "LC_DYLD_ENVIRONMENT")))
1454 return;
1455 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_MACOSX) {
1456 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd,
1457 "LC_VERSION_MIN_MACOSX")))
1458 return;
1459 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) {
1460 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd,
1461 "LC_VERSION_MIN_IPHONEOS")))
1462 return;
1463 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_TVOS) {
1464 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd,
1465 "LC_VERSION_MIN_TVOS")))
1466 return;
1467 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_WATCHOS) {
1468 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd,
1469 "LC_VERSION_MIN_WATCHOS")))
1470 return;
1471 } else if (Load.C.cmd == MachO::LC_NOTE) {
1472 if ((Err = checkNoteCommand(*this, Load, I, Elements)))
1473 return;
1474 } else if (Load.C.cmd == MachO::LC_BUILD_VERSION) {
1475 if ((Err = parseBuildVersionCommand(*this, Load, BuildTools, I)))
1476 return;
1477 } else if (Load.C.cmd == MachO::LC_RPATH) {
1478 if ((Err = checkRpathCommand(*this, Load, I)))
1479 return;
1480 } else if (Load.C.cmd == MachO::LC_SOURCE_VERSION) {
1481 if (Load.C.cmdsize != sizeof(MachO::source_version_command)) {
1482 Err = malformedError("LC_SOURCE_VERSION command " + Twine(I) +
1483 " has incorrect cmdsize");
1484 return;
1486 if (SourceLoadCmd) {
1487 Err = malformedError("more than one LC_SOURCE_VERSION command");
1488 return;
1490 SourceLoadCmd = Load.Ptr;
1491 } else if (Load.C.cmd == MachO::LC_MAIN) {
1492 if (Load.C.cmdsize != sizeof(MachO::entry_point_command)) {
1493 Err = malformedError("LC_MAIN command " + Twine(I) +
1494 " has incorrect cmdsize");
1495 return;
1497 if (EntryPointLoadCmd) {
1498 Err = malformedError("more than one LC_MAIN command");
1499 return;
1501 EntryPointLoadCmd = Load.Ptr;
1502 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO) {
1503 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command)) {
1504 Err = malformedError("LC_ENCRYPTION_INFO command " + Twine(I) +
1505 " has incorrect cmdsize");
1506 return;
1508 MachO::encryption_info_command E =
1509 getStruct<MachO::encryption_info_command>(*this, Load.Ptr);
1510 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize,
1511 &EncryptLoadCmd, "LC_ENCRYPTION_INFO")))
1512 return;
1513 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
1514 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command_64)) {
1515 Err = malformedError("LC_ENCRYPTION_INFO_64 command " + Twine(I) +
1516 " has incorrect cmdsize");
1517 return;
1519 MachO::encryption_info_command_64 E =
1520 getStruct<MachO::encryption_info_command_64>(*this, Load.Ptr);
1521 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize,
1522 &EncryptLoadCmd, "LC_ENCRYPTION_INFO_64")))
1523 return;
1524 } else if (Load.C.cmd == MachO::LC_LINKER_OPTION) {
1525 if ((Err = checkLinkerOptCommand(*this, Load, I)))
1526 return;
1527 } else if (Load.C.cmd == MachO::LC_SUB_FRAMEWORK) {
1528 if (Load.C.cmdsize < sizeof(MachO::sub_framework_command)) {
1529 Err = malformedError("load command " + Twine(I) +
1530 " LC_SUB_FRAMEWORK cmdsize too small");
1531 return;
1533 MachO::sub_framework_command S =
1534 getStruct<MachO::sub_framework_command>(*this, Load.Ptr);
1535 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_FRAMEWORK",
1536 sizeof(MachO::sub_framework_command),
1537 "sub_framework_command", S.umbrella,
1538 "umbrella")))
1539 return;
1540 } else if (Load.C.cmd == MachO::LC_SUB_UMBRELLA) {
1541 if (Load.C.cmdsize < sizeof(MachO::sub_umbrella_command)) {
1542 Err = malformedError("load command " + Twine(I) +
1543 " LC_SUB_UMBRELLA cmdsize too small");
1544 return;
1546 MachO::sub_umbrella_command S =
1547 getStruct<MachO::sub_umbrella_command>(*this, Load.Ptr);
1548 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_UMBRELLA",
1549 sizeof(MachO::sub_umbrella_command),
1550 "sub_umbrella_command", S.sub_umbrella,
1551 "sub_umbrella")))
1552 return;
1553 } else if (Load.C.cmd == MachO::LC_SUB_LIBRARY) {
1554 if (Load.C.cmdsize < sizeof(MachO::sub_library_command)) {
1555 Err = malformedError("load command " + Twine(I) +
1556 " LC_SUB_LIBRARY cmdsize too small");
1557 return;
1559 MachO::sub_library_command S =
1560 getStruct<MachO::sub_library_command>(*this, Load.Ptr);
1561 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_LIBRARY",
1562 sizeof(MachO::sub_library_command),
1563 "sub_library_command", S.sub_library,
1564 "sub_library")))
1565 return;
1566 } else if (Load.C.cmd == MachO::LC_SUB_CLIENT) {
1567 if (Load.C.cmdsize < sizeof(MachO::sub_client_command)) {
1568 Err = malformedError("load command " + Twine(I) +
1569 " LC_SUB_CLIENT cmdsize too small");
1570 return;
1572 MachO::sub_client_command S =
1573 getStruct<MachO::sub_client_command>(*this, Load.Ptr);
1574 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_CLIENT",
1575 sizeof(MachO::sub_client_command),
1576 "sub_client_command", S.client, "client")))
1577 return;
1578 } else if (Load.C.cmd == MachO::LC_ROUTINES) {
1579 if (Load.C.cmdsize != sizeof(MachO::routines_command)) {
1580 Err = malformedError("LC_ROUTINES command " + Twine(I) +
1581 " has incorrect cmdsize");
1582 return;
1584 if (RoutinesLoadCmd) {
1585 Err = malformedError("more than one LC_ROUTINES and or LC_ROUTINES_64 "
1586 "command");
1587 return;
1589 RoutinesLoadCmd = Load.Ptr;
1590 } else if (Load.C.cmd == MachO::LC_ROUTINES_64) {
1591 if (Load.C.cmdsize != sizeof(MachO::routines_command_64)) {
1592 Err = malformedError("LC_ROUTINES_64 command " + Twine(I) +
1593 " has incorrect cmdsize");
1594 return;
1596 if (RoutinesLoadCmd) {
1597 Err = malformedError("more than one LC_ROUTINES_64 and or LC_ROUTINES "
1598 "command");
1599 return;
1601 RoutinesLoadCmd = Load.Ptr;
1602 } else if (Load.C.cmd == MachO::LC_UNIXTHREAD) {
1603 if ((Err = checkThreadCommand(*this, Load, I, "LC_UNIXTHREAD")))
1604 return;
1605 if (UnixThreadLoadCmd) {
1606 Err = malformedError("more than one LC_UNIXTHREAD command");
1607 return;
1609 UnixThreadLoadCmd = Load.Ptr;
1610 } else if (Load.C.cmd == MachO::LC_THREAD) {
1611 if ((Err = checkThreadCommand(*this, Load, I, "LC_THREAD")))
1612 return;
1613 // Note: LC_TWOLEVEL_HINTS is really obsolete and is not supported.
1614 } else if (Load.C.cmd == MachO::LC_TWOLEVEL_HINTS) {
1615 if ((Err = checkTwoLevelHintsCommand(*this, Load, I,
1616 &TwoLevelHintsLoadCmd, Elements)))
1617 return;
1618 } else if (Load.C.cmd == MachO::LC_IDENT) {
1619 // Note: LC_IDENT is ignored.
1620 continue;
1621 } else if (isLoadCommandObsolete(Load.C.cmd)) {
1622 Err = malformedError("load command " + Twine(I) + " for cmd value of: " +
1623 Twine(Load.C.cmd) + " is obsolete and not "
1624 "supported");
1625 return;
1627 // TODO: generate a error for unknown load commands by default. But still
1628 // need work out an approach to allow or not allow unknown values like this
1629 // as an option for some uses like lldb.
1630 if (I < LoadCommandCount - 1) {
1631 if (auto LoadOrErr = getNextLoadCommandInfo(*this, I, Load))
1632 Load = *LoadOrErr;
1633 else {
1634 Err = LoadOrErr.takeError();
1635 return;
1639 if (!SymtabLoadCmd) {
1640 if (DysymtabLoadCmd) {
1641 Err = malformedError("contains LC_DYSYMTAB load command without a "
1642 "LC_SYMTAB load command");
1643 return;
1645 } else if (DysymtabLoadCmd) {
1646 MachO::symtab_command Symtab =
1647 getStruct<MachO::symtab_command>(*this, SymtabLoadCmd);
1648 MachO::dysymtab_command Dysymtab =
1649 getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd);
1650 if (Dysymtab.nlocalsym != 0 && Dysymtab.ilocalsym > Symtab.nsyms) {
1651 Err = malformedError("ilocalsym in LC_DYSYMTAB load command "
1652 "extends past the end of the symbol table");
1653 return;
1655 uint64_t BigSize = Dysymtab.ilocalsym;
1656 BigSize += Dysymtab.nlocalsym;
1657 if (Dysymtab.nlocalsym != 0 && BigSize > Symtab.nsyms) {
1658 Err = malformedError("ilocalsym plus nlocalsym in LC_DYSYMTAB load "
1659 "command extends past the end of the symbol table");
1660 return;
1662 if (Dysymtab.nextdefsym != 0 && Dysymtab.iextdefsym > Symtab.nsyms) {
1663 Err = malformedError("iextdefsym in LC_DYSYMTAB load command "
1664 "extends past the end of the symbol table");
1665 return;
1667 BigSize = Dysymtab.iextdefsym;
1668 BigSize += Dysymtab.nextdefsym;
1669 if (Dysymtab.nextdefsym != 0 && BigSize > Symtab.nsyms) {
1670 Err = malformedError("iextdefsym plus nextdefsym in LC_DYSYMTAB "
1671 "load command extends past the end of the symbol "
1672 "table");
1673 return;
1675 if (Dysymtab.nundefsym != 0 && Dysymtab.iundefsym > Symtab.nsyms) {
1676 Err = malformedError("iundefsym in LC_DYSYMTAB load command "
1677 "extends past the end of the symbol table");
1678 return;
1680 BigSize = Dysymtab.iundefsym;
1681 BigSize += Dysymtab.nundefsym;
1682 if (Dysymtab.nundefsym != 0 && BigSize > Symtab.nsyms) {
1683 Err = malformedError("iundefsym plus nundefsym in LC_DYSYMTAB load "
1684 " command extends past the end of the symbol table");
1685 return;
1688 if ((getHeader().filetype == MachO::MH_DYLIB ||
1689 getHeader().filetype == MachO::MH_DYLIB_STUB) &&
1690 DyldIdLoadCmd == nullptr) {
1691 Err = malformedError("no LC_ID_DYLIB load command in dynamic library "
1692 "filetype");
1693 return;
1695 assert(LoadCommands.size() == LoadCommandCount);
1697 Err = Error::success();
1700 Error MachOObjectFile::checkSymbolTable() const {
1701 uint32_t Flags = 0;
1702 if (is64Bit()) {
1703 MachO::mach_header_64 H_64 = MachOObjectFile::getHeader64();
1704 Flags = H_64.flags;
1705 } else {
1706 MachO::mach_header H = MachOObjectFile::getHeader();
1707 Flags = H.flags;
1709 uint8_t NType = 0;
1710 uint8_t NSect = 0;
1711 uint16_t NDesc = 0;
1712 uint32_t NStrx = 0;
1713 uint64_t NValue = 0;
1714 uint32_t SymbolIndex = 0;
1715 MachO::symtab_command S = getSymtabLoadCommand();
1716 for (const SymbolRef &Symbol : symbols()) {
1717 DataRefImpl SymDRI = Symbol.getRawDataRefImpl();
1718 if (is64Bit()) {
1719 MachO::nlist_64 STE_64 = getSymbol64TableEntry(SymDRI);
1720 NType = STE_64.n_type;
1721 NSect = STE_64.n_sect;
1722 NDesc = STE_64.n_desc;
1723 NStrx = STE_64.n_strx;
1724 NValue = STE_64.n_value;
1725 } else {
1726 MachO::nlist STE = getSymbolTableEntry(SymDRI);
1727 NType = STE.n_type;
1728 NSect = STE.n_sect;
1729 NDesc = STE.n_desc;
1730 NStrx = STE.n_strx;
1731 NValue = STE.n_value;
1733 if ((NType & MachO::N_STAB) == 0) {
1734 if ((NType & MachO::N_TYPE) == MachO::N_SECT) {
1735 if (NSect == 0 || NSect > Sections.size())
1736 return malformedError("bad section index: " + Twine((int)NSect) +
1737 " for symbol at index " + Twine(SymbolIndex));
1739 if ((NType & MachO::N_TYPE) == MachO::N_INDR) {
1740 if (NValue >= S.strsize)
1741 return malformedError("bad n_value: " + Twine((int)NValue) + " past "
1742 "the end of string table, for N_INDR symbol at "
1743 "index " + Twine(SymbolIndex));
1745 if ((Flags & MachO::MH_TWOLEVEL) == MachO::MH_TWOLEVEL &&
1746 (((NType & MachO::N_TYPE) == MachO::N_UNDF && NValue == 0) ||
1747 (NType & MachO::N_TYPE) == MachO::N_PBUD)) {
1748 uint32_t LibraryOrdinal = MachO::GET_LIBRARY_ORDINAL(NDesc);
1749 if (LibraryOrdinal != 0 &&
1750 LibraryOrdinal != MachO::EXECUTABLE_ORDINAL &&
1751 LibraryOrdinal != MachO::DYNAMIC_LOOKUP_ORDINAL &&
1752 LibraryOrdinal - 1 >= Libraries.size() ) {
1753 return malformedError("bad library ordinal: " + Twine(LibraryOrdinal) +
1754 " for symbol at index " + Twine(SymbolIndex));
1758 if (NStrx >= S.strsize)
1759 return malformedError("bad string table index: " + Twine((int)NStrx) +
1760 " past the end of string table, for symbol at "
1761 "index " + Twine(SymbolIndex));
1762 SymbolIndex++;
1764 return Error::success();
1767 void MachOObjectFile::moveSymbolNext(DataRefImpl &Symb) const {
1768 unsigned SymbolTableEntrySize = is64Bit() ?
1769 sizeof(MachO::nlist_64) :
1770 sizeof(MachO::nlist);
1771 Symb.p += SymbolTableEntrySize;
1774 Expected<StringRef> MachOObjectFile::getSymbolName(DataRefImpl Symb) const {
1775 StringRef StringTable = getStringTableData();
1776 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb);
1777 if (Entry.n_strx == 0)
1778 // A n_strx value of 0 indicates that no name is associated with a
1779 // particular symbol table entry.
1780 return StringRef();
1781 const char *Start = &StringTable.data()[Entry.n_strx];
1782 if (Start < getData().begin() || Start >= getData().end()) {
1783 return malformedError("bad string index: " + Twine(Entry.n_strx) +
1784 " for symbol at index " + Twine(getSymbolIndex(Symb)));
1786 return StringRef(Start);
1789 unsigned MachOObjectFile::getSectionType(SectionRef Sec) const {
1790 DataRefImpl DRI = Sec.getRawDataRefImpl();
1791 uint32_t Flags = getSectionFlags(*this, DRI);
1792 return Flags & MachO::SECTION_TYPE;
1795 uint64_t MachOObjectFile::getNValue(DataRefImpl Sym) const {
1796 if (is64Bit()) {
1797 MachO::nlist_64 Entry = getSymbol64TableEntry(Sym);
1798 return Entry.n_value;
1800 MachO::nlist Entry = getSymbolTableEntry(Sym);
1801 return Entry.n_value;
1804 // getIndirectName() returns the name of the alias'ed symbol who's string table
1805 // index is in the n_value field.
1806 std::error_code MachOObjectFile::getIndirectName(DataRefImpl Symb,
1807 StringRef &Res) const {
1808 StringRef StringTable = getStringTableData();
1809 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb);
1810 if ((Entry.n_type & MachO::N_TYPE) != MachO::N_INDR)
1811 return object_error::parse_failed;
1812 uint64_t NValue = getNValue(Symb);
1813 if (NValue >= StringTable.size())
1814 return object_error::parse_failed;
1815 const char *Start = &StringTable.data()[NValue];
1816 Res = StringRef(Start);
1817 return std::error_code();
1820 uint64_t MachOObjectFile::getSymbolValueImpl(DataRefImpl Sym) const {
1821 return getNValue(Sym);
1824 Expected<uint64_t> MachOObjectFile::getSymbolAddress(DataRefImpl Sym) const {
1825 return getSymbolValue(Sym);
1828 uint32_t MachOObjectFile::getSymbolAlignment(DataRefImpl DRI) const {
1829 uint32_t Flags = cantFail(getSymbolFlags(DRI));
1830 if (Flags & SymbolRef::SF_Common) {
1831 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI);
1832 return 1 << MachO::GET_COMM_ALIGN(Entry.n_desc);
1834 return 0;
1837 uint64_t MachOObjectFile::getCommonSymbolSizeImpl(DataRefImpl DRI) const {
1838 return getNValue(DRI);
1841 Expected<SymbolRef::Type>
1842 MachOObjectFile::getSymbolType(DataRefImpl Symb) const {
1843 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb);
1844 uint8_t n_type = Entry.n_type;
1846 // If this is a STAB debugging symbol, we can do nothing more.
1847 if (n_type & MachO::N_STAB)
1848 return SymbolRef::ST_Debug;
1850 switch (n_type & MachO::N_TYPE) {
1851 case MachO::N_UNDF :
1852 return SymbolRef::ST_Unknown;
1853 case MachO::N_SECT :
1854 Expected<section_iterator> SecOrError = getSymbolSection(Symb);
1855 if (!SecOrError)
1856 return SecOrError.takeError();
1857 section_iterator Sec = *SecOrError;
1858 if (Sec == section_end())
1859 return SymbolRef::ST_Other;
1860 if (Sec->isData() || Sec->isBSS())
1861 return SymbolRef::ST_Data;
1862 return SymbolRef::ST_Function;
1864 return SymbolRef::ST_Other;
1867 Expected<uint32_t> MachOObjectFile::getSymbolFlags(DataRefImpl DRI) const {
1868 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI);
1870 uint8_t MachOType = Entry.n_type;
1871 uint16_t MachOFlags = Entry.n_desc;
1873 uint32_t Result = SymbolRef::SF_None;
1875 if ((MachOType & MachO::N_TYPE) == MachO::N_INDR)
1876 Result |= SymbolRef::SF_Indirect;
1878 if (MachOType & MachO::N_STAB)
1879 Result |= SymbolRef::SF_FormatSpecific;
1881 if (MachOType & MachO::N_EXT) {
1882 Result |= SymbolRef::SF_Global;
1883 if ((MachOType & MachO::N_TYPE) == MachO::N_UNDF) {
1884 if (getNValue(DRI))
1885 Result |= SymbolRef::SF_Common;
1886 else
1887 Result |= SymbolRef::SF_Undefined;
1890 if (MachOType & MachO::N_PEXT)
1891 Result |= SymbolRef::SF_Hidden;
1892 else
1893 Result |= SymbolRef::SF_Exported;
1895 } else if (MachOType & MachO::N_PEXT)
1896 Result |= SymbolRef::SF_Hidden;
1898 if (MachOFlags & (MachO::N_WEAK_REF | MachO::N_WEAK_DEF))
1899 Result |= SymbolRef::SF_Weak;
1901 if (MachOFlags & (MachO::N_ARM_THUMB_DEF))
1902 Result |= SymbolRef::SF_Thumb;
1904 if ((MachOType & MachO::N_TYPE) == MachO::N_ABS)
1905 Result |= SymbolRef::SF_Absolute;
1907 return Result;
1910 Expected<section_iterator>
1911 MachOObjectFile::getSymbolSection(DataRefImpl Symb) const {
1912 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb);
1913 uint8_t index = Entry.n_sect;
1915 if (index == 0)
1916 return section_end();
1917 DataRefImpl DRI;
1918 DRI.d.a = index - 1;
1919 if (DRI.d.a >= Sections.size()){
1920 return malformedError("bad section index: " + Twine((int)index) +
1921 " for symbol at index " + Twine(getSymbolIndex(Symb)));
1923 return section_iterator(SectionRef(DRI, this));
1926 unsigned MachOObjectFile::getSymbolSectionID(SymbolRef Sym) const {
1927 MachO::nlist_base Entry =
1928 getSymbolTableEntryBase(*this, Sym.getRawDataRefImpl());
1929 return Entry.n_sect - 1;
1932 void MachOObjectFile::moveSectionNext(DataRefImpl &Sec) const {
1933 Sec.d.a++;
1936 Expected<StringRef> MachOObjectFile::getSectionName(DataRefImpl Sec) const {
1937 ArrayRef<char> Raw = getSectionRawName(Sec);
1938 return parseSegmentOrSectionName(Raw.data());
1941 uint64_t MachOObjectFile::getSectionAddress(DataRefImpl Sec) const {
1942 if (is64Bit())
1943 return getSection64(Sec).addr;
1944 return getSection(Sec).addr;
1947 uint64_t MachOObjectFile::getSectionIndex(DataRefImpl Sec) const {
1948 return Sec.d.a;
1951 uint64_t MachOObjectFile::getSectionSize(DataRefImpl Sec) const {
1952 // In the case if a malformed Mach-O file where the section offset is past
1953 // the end of the file or some part of the section size is past the end of
1954 // the file return a size of zero or a size that covers the rest of the file
1955 // but does not extend past the end of the file.
1956 uint32_t SectOffset, SectType;
1957 uint64_t SectSize;
1959 if (is64Bit()) {
1960 MachO::section_64 Sect = getSection64(Sec);
1961 SectOffset = Sect.offset;
1962 SectSize = Sect.size;
1963 SectType = Sect.flags & MachO::SECTION_TYPE;
1964 } else {
1965 MachO::section Sect = getSection(Sec);
1966 SectOffset = Sect.offset;
1967 SectSize = Sect.size;
1968 SectType = Sect.flags & MachO::SECTION_TYPE;
1970 if (SectType == MachO::S_ZEROFILL || SectType == MachO::S_GB_ZEROFILL)
1971 return SectSize;
1972 uint64_t FileSize = getData().size();
1973 if (SectOffset > FileSize)
1974 return 0;
1975 if (FileSize - SectOffset < SectSize)
1976 return FileSize - SectOffset;
1977 return SectSize;
1980 ArrayRef<uint8_t> MachOObjectFile::getSectionContents(uint32_t Offset,
1981 uint64_t Size) const {
1982 return arrayRefFromStringRef(getData().substr(Offset, Size));
1985 Expected<ArrayRef<uint8_t>>
1986 MachOObjectFile::getSectionContents(DataRefImpl Sec) const {
1987 uint32_t Offset;
1988 uint64_t Size;
1990 if (is64Bit()) {
1991 MachO::section_64 Sect = getSection64(Sec);
1992 Offset = Sect.offset;
1993 Size = Sect.size;
1994 } else {
1995 MachO::section Sect = getSection(Sec);
1996 Offset = Sect.offset;
1997 Size = Sect.size;
2000 return getSectionContents(Offset, Size);
2003 uint64_t MachOObjectFile::getSectionAlignment(DataRefImpl Sec) const {
2004 uint32_t Align;
2005 if (is64Bit()) {
2006 MachO::section_64 Sect = getSection64(Sec);
2007 Align = Sect.align;
2008 } else {
2009 MachO::section Sect = getSection(Sec);
2010 Align = Sect.align;
2013 return uint64_t(1) << Align;
2016 Expected<SectionRef> MachOObjectFile::getSection(unsigned SectionIndex) const {
2017 if (SectionIndex < 1 || SectionIndex > Sections.size())
2018 return malformedError("bad section index: " + Twine((int)SectionIndex));
2020 DataRefImpl DRI;
2021 DRI.d.a = SectionIndex - 1;
2022 return SectionRef(DRI, this);
2025 Expected<SectionRef> MachOObjectFile::getSection(StringRef SectionName) const {
2026 for (const SectionRef &Section : sections()) {
2027 auto NameOrErr = Section.getName();
2028 if (!NameOrErr)
2029 return NameOrErr.takeError();
2030 if (*NameOrErr == SectionName)
2031 return Section;
2033 return errorCodeToError(object_error::parse_failed);
2036 bool MachOObjectFile::isSectionCompressed(DataRefImpl Sec) const {
2037 return false;
2040 bool MachOObjectFile::isSectionText(DataRefImpl Sec) const {
2041 uint32_t Flags = getSectionFlags(*this, Sec);
2042 return Flags & MachO::S_ATTR_PURE_INSTRUCTIONS;
2045 bool MachOObjectFile::isSectionData(DataRefImpl Sec) const {
2046 uint32_t Flags = getSectionFlags(*this, Sec);
2047 unsigned SectionType = Flags & MachO::SECTION_TYPE;
2048 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) &&
2049 !(SectionType == MachO::S_ZEROFILL ||
2050 SectionType == MachO::S_GB_ZEROFILL);
2053 bool MachOObjectFile::isSectionBSS(DataRefImpl Sec) const {
2054 uint32_t Flags = getSectionFlags(*this, Sec);
2055 unsigned SectionType = Flags & MachO::SECTION_TYPE;
2056 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) &&
2057 (SectionType == MachO::S_ZEROFILL ||
2058 SectionType == MachO::S_GB_ZEROFILL);
2061 bool MachOObjectFile::isDebugSection(DataRefImpl Sec) const {
2062 Expected<StringRef> SectionNameOrErr = getSectionName(Sec);
2063 if (!SectionNameOrErr) {
2064 // TODO: Report the error message properly.
2065 consumeError(SectionNameOrErr.takeError());
2066 return false;
2068 StringRef SectionName = SectionNameOrErr.get();
2069 return SectionName.starts_with("__debug") ||
2070 SectionName.starts_with("__zdebug") ||
2071 SectionName.starts_with("__apple") || SectionName == "__gdb_index" ||
2072 SectionName == "__swift_ast";
2075 namespace {
2076 template <typename LoadCommandType>
2077 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj,
2078 MachOObjectFile::LoadCommandInfo LoadCmd,
2079 StringRef SegmentName) {
2080 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr);
2081 if (!SegmentOrErr) {
2082 consumeError(SegmentOrErr.takeError());
2083 return {};
2085 auto &Segment = SegmentOrErr.get();
2086 if (StringRef(Segment.segname, 16).starts_with(SegmentName))
2087 return arrayRefFromStringRef(Obj.getData().slice(
2088 Segment.fileoff, Segment.fileoff + Segment.filesize));
2089 return {};
2092 template <typename LoadCommandType>
2093 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj,
2094 MachOObjectFile::LoadCommandInfo LoadCmd) {
2095 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr);
2096 if (!SegmentOrErr) {
2097 consumeError(SegmentOrErr.takeError());
2098 return {};
2100 auto &Segment = SegmentOrErr.get();
2101 return arrayRefFromStringRef(
2102 Obj.getData().substr(Segment.fileoff, Segment.filesize));
2104 } // namespace
2106 ArrayRef<uint8_t>
2107 MachOObjectFile::getSegmentContents(StringRef SegmentName) const {
2108 for (auto LoadCmd : load_commands()) {
2109 ArrayRef<uint8_t> Contents;
2110 switch (LoadCmd.C.cmd) {
2111 case MachO::LC_SEGMENT:
2112 Contents = ::getSegmentContents<MachO::segment_command>(*this, LoadCmd,
2113 SegmentName);
2114 break;
2115 case MachO::LC_SEGMENT_64:
2116 Contents = ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd,
2117 SegmentName);
2118 break;
2119 default:
2120 continue;
2122 if (!Contents.empty())
2123 return Contents;
2125 return {};
2128 ArrayRef<uint8_t>
2129 MachOObjectFile::getSegmentContents(size_t SegmentIndex) const {
2130 size_t Idx = 0;
2131 for (auto LoadCmd : load_commands()) {
2132 switch (LoadCmd.C.cmd) {
2133 case MachO::LC_SEGMENT:
2134 if (Idx == SegmentIndex)
2135 return ::getSegmentContents<MachO::segment_command>(*this, LoadCmd);
2136 ++Idx;
2137 break;
2138 case MachO::LC_SEGMENT_64:
2139 if (Idx == SegmentIndex)
2140 return ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd);
2141 ++Idx;
2142 break;
2143 default:
2144 continue;
2147 return {};
2150 unsigned MachOObjectFile::getSectionID(SectionRef Sec) const {
2151 return Sec.getRawDataRefImpl().d.a;
2154 bool MachOObjectFile::isSectionVirtual(DataRefImpl Sec) const {
2155 uint32_t Flags = getSectionFlags(*this, Sec);
2156 unsigned SectionType = Flags & MachO::SECTION_TYPE;
2157 return SectionType == MachO::S_ZEROFILL ||
2158 SectionType == MachO::S_GB_ZEROFILL;
2161 bool MachOObjectFile::isSectionBitcode(DataRefImpl Sec) const {
2162 StringRef SegmentName = getSectionFinalSegmentName(Sec);
2163 if (Expected<StringRef> NameOrErr = getSectionName(Sec))
2164 return (SegmentName == "__LLVM" && *NameOrErr == "__bitcode");
2165 return false;
2168 bool MachOObjectFile::isSectionStripped(DataRefImpl Sec) const {
2169 if (is64Bit())
2170 return getSection64(Sec).offset == 0;
2171 return getSection(Sec).offset == 0;
2174 relocation_iterator MachOObjectFile::section_rel_begin(DataRefImpl Sec) const {
2175 DataRefImpl Ret;
2176 Ret.d.a = Sec.d.a;
2177 Ret.d.b = 0;
2178 return relocation_iterator(RelocationRef(Ret, this));
2181 relocation_iterator
2182 MachOObjectFile::section_rel_end(DataRefImpl Sec) const {
2183 uint32_t Num;
2184 if (is64Bit()) {
2185 MachO::section_64 Sect = getSection64(Sec);
2186 Num = Sect.nreloc;
2187 } else {
2188 MachO::section Sect = getSection(Sec);
2189 Num = Sect.nreloc;
2192 DataRefImpl Ret;
2193 Ret.d.a = Sec.d.a;
2194 Ret.d.b = Num;
2195 return relocation_iterator(RelocationRef(Ret, this));
2198 relocation_iterator MachOObjectFile::extrel_begin() const {
2199 DataRefImpl Ret;
2200 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations
2201 Ret.d.a = 0; // Would normally be a section index.
2202 Ret.d.b = 0; // Index into the external relocations
2203 return relocation_iterator(RelocationRef(Ret, this));
2206 relocation_iterator MachOObjectFile::extrel_end() const {
2207 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand();
2208 DataRefImpl Ret;
2209 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations
2210 Ret.d.a = 0; // Would normally be a section index.
2211 Ret.d.b = DysymtabLoadCmd.nextrel; // Index into the external relocations
2212 return relocation_iterator(RelocationRef(Ret, this));
2215 relocation_iterator MachOObjectFile::locrel_begin() const {
2216 DataRefImpl Ret;
2217 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations
2218 Ret.d.a = 1; // Would normally be a section index.
2219 Ret.d.b = 0; // Index into the local relocations
2220 return relocation_iterator(RelocationRef(Ret, this));
2223 relocation_iterator MachOObjectFile::locrel_end() const {
2224 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand();
2225 DataRefImpl Ret;
2226 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations
2227 Ret.d.a = 1; // Would normally be a section index.
2228 Ret.d.b = DysymtabLoadCmd.nlocrel; // Index into the local relocations
2229 return relocation_iterator(RelocationRef(Ret, this));
2232 void MachOObjectFile::moveRelocationNext(DataRefImpl &Rel) const {
2233 ++Rel.d.b;
2236 uint64_t MachOObjectFile::getRelocationOffset(DataRefImpl Rel) const {
2237 assert((getHeader().filetype == MachO::MH_OBJECT ||
2238 getHeader().filetype == MachO::MH_KEXT_BUNDLE) &&
2239 "Only implemented for MH_OBJECT && MH_KEXT_BUNDLE");
2240 MachO::any_relocation_info RE = getRelocation(Rel);
2241 return getAnyRelocationAddress(RE);
2244 symbol_iterator
2245 MachOObjectFile::getRelocationSymbol(DataRefImpl Rel) const {
2246 MachO::any_relocation_info RE = getRelocation(Rel);
2247 if (isRelocationScattered(RE))
2248 return symbol_end();
2250 uint32_t SymbolIdx = getPlainRelocationSymbolNum(RE);
2251 bool isExtern = getPlainRelocationExternal(RE);
2252 if (!isExtern)
2253 return symbol_end();
2255 MachO::symtab_command S = getSymtabLoadCommand();
2256 unsigned SymbolTableEntrySize = is64Bit() ?
2257 sizeof(MachO::nlist_64) :
2258 sizeof(MachO::nlist);
2259 uint64_t Offset = S.symoff + SymbolIdx * SymbolTableEntrySize;
2260 DataRefImpl Sym;
2261 Sym.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset));
2262 return symbol_iterator(SymbolRef(Sym, this));
2265 section_iterator
2266 MachOObjectFile::getRelocationSection(DataRefImpl Rel) const {
2267 return section_iterator(getAnyRelocationSection(getRelocation(Rel)));
2270 uint64_t MachOObjectFile::getRelocationType(DataRefImpl Rel) const {
2271 MachO::any_relocation_info RE = getRelocation(Rel);
2272 return getAnyRelocationType(RE);
2275 void MachOObjectFile::getRelocationTypeName(
2276 DataRefImpl Rel, SmallVectorImpl<char> &Result) const {
2277 StringRef res;
2278 uint64_t RType = getRelocationType(Rel);
2280 unsigned Arch = this->getArch();
2282 switch (Arch) {
2283 case Triple::x86: {
2284 static const char *const Table[] = {
2285 "GENERIC_RELOC_VANILLA",
2286 "GENERIC_RELOC_PAIR",
2287 "GENERIC_RELOC_SECTDIFF",
2288 "GENERIC_RELOC_PB_LA_PTR",
2289 "GENERIC_RELOC_LOCAL_SECTDIFF",
2290 "GENERIC_RELOC_TLV" };
2292 if (RType > 5)
2293 res = "Unknown";
2294 else
2295 res = Table[RType];
2296 break;
2298 case Triple::x86_64: {
2299 static const char *const Table[] = {
2300 "X86_64_RELOC_UNSIGNED",
2301 "X86_64_RELOC_SIGNED",
2302 "X86_64_RELOC_BRANCH",
2303 "X86_64_RELOC_GOT_LOAD",
2304 "X86_64_RELOC_GOT",
2305 "X86_64_RELOC_SUBTRACTOR",
2306 "X86_64_RELOC_SIGNED_1",
2307 "X86_64_RELOC_SIGNED_2",
2308 "X86_64_RELOC_SIGNED_4",
2309 "X86_64_RELOC_TLV" };
2311 if (RType > 9)
2312 res = "Unknown";
2313 else
2314 res = Table[RType];
2315 break;
2317 case Triple::arm: {
2318 static const char *const Table[] = {
2319 "ARM_RELOC_VANILLA",
2320 "ARM_RELOC_PAIR",
2321 "ARM_RELOC_SECTDIFF",
2322 "ARM_RELOC_LOCAL_SECTDIFF",
2323 "ARM_RELOC_PB_LA_PTR",
2324 "ARM_RELOC_BR24",
2325 "ARM_THUMB_RELOC_BR22",
2326 "ARM_THUMB_32BIT_BRANCH",
2327 "ARM_RELOC_HALF",
2328 "ARM_RELOC_HALF_SECTDIFF" };
2330 if (RType > 9)
2331 res = "Unknown";
2332 else
2333 res = Table[RType];
2334 break;
2336 case Triple::aarch64:
2337 case Triple::aarch64_32: {
2338 static const char *const Table[] = {
2339 "ARM64_RELOC_UNSIGNED", "ARM64_RELOC_SUBTRACTOR",
2340 "ARM64_RELOC_BRANCH26", "ARM64_RELOC_PAGE21",
2341 "ARM64_RELOC_PAGEOFF12", "ARM64_RELOC_GOT_LOAD_PAGE21",
2342 "ARM64_RELOC_GOT_LOAD_PAGEOFF12", "ARM64_RELOC_POINTER_TO_GOT",
2343 "ARM64_RELOC_TLVP_LOAD_PAGE21", "ARM64_RELOC_TLVP_LOAD_PAGEOFF12",
2344 "ARM64_RELOC_ADDEND", "ARM64_RELOC_AUTHENTICATED_POINTER"
2347 if (RType >= std::size(Table))
2348 res = "Unknown";
2349 else
2350 res = Table[RType];
2351 break;
2353 case Triple::ppc: {
2354 static const char *const Table[] = {
2355 "PPC_RELOC_VANILLA",
2356 "PPC_RELOC_PAIR",
2357 "PPC_RELOC_BR14",
2358 "PPC_RELOC_BR24",
2359 "PPC_RELOC_HI16",
2360 "PPC_RELOC_LO16",
2361 "PPC_RELOC_HA16",
2362 "PPC_RELOC_LO14",
2363 "PPC_RELOC_SECTDIFF",
2364 "PPC_RELOC_PB_LA_PTR",
2365 "PPC_RELOC_HI16_SECTDIFF",
2366 "PPC_RELOC_LO16_SECTDIFF",
2367 "PPC_RELOC_HA16_SECTDIFF",
2368 "PPC_RELOC_JBSR",
2369 "PPC_RELOC_LO14_SECTDIFF",
2370 "PPC_RELOC_LOCAL_SECTDIFF" };
2372 if (RType > 15)
2373 res = "Unknown";
2374 else
2375 res = Table[RType];
2376 break;
2378 case Triple::UnknownArch:
2379 res = "Unknown";
2380 break;
2382 Result.append(res.begin(), res.end());
2385 uint8_t MachOObjectFile::getRelocationLength(DataRefImpl Rel) const {
2386 MachO::any_relocation_info RE = getRelocation(Rel);
2387 return getAnyRelocationLength(RE);
2391 // guessLibraryShortName() is passed a name of a dynamic library and returns a
2392 // guess on what the short name is. Then name is returned as a substring of the
2393 // StringRef Name passed in. The name of the dynamic library is recognized as
2394 // a framework if it has one of the two following forms:
2395 // Foo.framework/Versions/A/Foo
2396 // Foo.framework/Foo
2397 // Where A and Foo can be any string. And may contain a trailing suffix
2398 // starting with an underbar. If the Name is recognized as a framework then
2399 // isFramework is set to true else it is set to false. If the Name has a
2400 // suffix then Suffix is set to the substring in Name that contains the suffix
2401 // else it is set to a NULL StringRef.
2403 // The Name of the dynamic library is recognized as a library name if it has
2404 // one of the two following forms:
2405 // libFoo.A.dylib
2406 // libFoo.dylib
2408 // The library may have a suffix trailing the name Foo of the form:
2409 // libFoo_profile.A.dylib
2410 // libFoo_profile.dylib
2411 // These dyld image suffixes are separated from the short name by a '_'
2412 // character. Because the '_' character is commonly used to separate words in
2413 // filenames guessLibraryShortName() cannot reliably separate a dylib's short
2414 // name from an arbitrary image suffix; imagine if both the short name and the
2415 // suffix contains an '_' character! To better deal with this ambiguity,
2416 // guessLibraryShortName() will recognize only "_debug" and "_profile" as valid
2417 // Suffix values. Calling code needs to be tolerant of guessLibraryShortName()
2418 // guessing incorrectly.
2420 // The Name of the dynamic library is also recognized as a library name if it
2421 // has the following form:
2422 // Foo.qtx
2424 // If the Name of the dynamic library is none of the forms above then a NULL
2425 // StringRef is returned.
2426 StringRef MachOObjectFile::guessLibraryShortName(StringRef Name,
2427 bool &isFramework,
2428 StringRef &Suffix) {
2429 StringRef Foo, F, DotFramework, V, Dylib, Lib, Dot, Qtx;
2430 size_t a, b, c, d, Idx;
2432 isFramework = false;
2433 Suffix = StringRef();
2435 // Pull off the last component and make Foo point to it
2436 a = Name.rfind('/');
2437 if (a == Name.npos || a == 0)
2438 goto guess_library;
2439 Foo = Name.substr(a + 1);
2441 // Look for a suffix starting with a '_'
2442 Idx = Foo.rfind('_');
2443 if (Idx != Foo.npos && Foo.size() >= 2) {
2444 Suffix = Foo.substr(Idx);
2445 if (Suffix != "_debug" && Suffix != "_profile")
2446 Suffix = StringRef();
2447 else
2448 Foo = Foo.slice(0, Idx);
2451 // First look for the form Foo.framework/Foo
2452 b = Name.rfind('/', a);
2453 if (b == Name.npos)
2454 Idx = 0;
2455 else
2456 Idx = b+1;
2457 F = Name.substr(Idx, Foo.size());
2458 DotFramework = Name.substr(Idx + Foo.size(), sizeof(".framework/") - 1);
2459 if (F == Foo && DotFramework == ".framework/") {
2460 isFramework = true;
2461 return Foo;
2464 // Next look for the form Foo.framework/Versions/A/Foo
2465 if (b == Name.npos)
2466 goto guess_library;
2467 c = Name.rfind('/', b);
2468 if (c == Name.npos || c == 0)
2469 goto guess_library;
2470 V = Name.substr(c + 1);
2471 if (!V.starts_with("Versions/"))
2472 goto guess_library;
2473 d = Name.rfind('/', c);
2474 if (d == Name.npos)
2475 Idx = 0;
2476 else
2477 Idx = d+1;
2478 F = Name.substr(Idx, Foo.size());
2479 DotFramework = Name.substr(Idx + Foo.size(), sizeof(".framework/") - 1);
2480 if (F == Foo && DotFramework == ".framework/") {
2481 isFramework = true;
2482 return Foo;
2485 guess_library:
2486 // pull off the suffix after the "." and make a point to it
2487 a = Name.rfind('.');
2488 if (a == Name.npos || a == 0)
2489 return StringRef();
2490 Dylib = Name.substr(a);
2491 if (Dylib != ".dylib")
2492 goto guess_qtx;
2494 // First pull off the version letter for the form Foo.A.dylib if any.
2495 if (a >= 3) {
2496 Dot = Name.substr(a - 2, 1);
2497 if (Dot == ".")
2498 a = a - 2;
2501 b = Name.rfind('/', a);
2502 if (b == Name.npos)
2503 b = 0;
2504 else
2505 b = b+1;
2506 // ignore any suffix after an underbar like Foo_profile.A.dylib
2507 Idx = Name.rfind('_');
2508 if (Idx != Name.npos && Idx != b) {
2509 Lib = Name.slice(b, Idx);
2510 Suffix = Name.slice(Idx, a);
2511 if (Suffix != "_debug" && Suffix != "_profile") {
2512 Suffix = StringRef();
2513 Lib = Name.slice(b, a);
2516 else
2517 Lib = Name.slice(b, a);
2518 // There are incorrect library names of the form:
2519 // libATS.A_profile.dylib so check for these.
2520 if (Lib.size() >= 3) {
2521 Dot = Lib.substr(Lib.size() - 2, 1);
2522 if (Dot == ".")
2523 Lib = Lib.slice(0, Lib.size()-2);
2525 return Lib;
2527 guess_qtx:
2528 Qtx = Name.substr(a);
2529 if (Qtx != ".qtx")
2530 return StringRef();
2531 b = Name.rfind('/', a);
2532 if (b == Name.npos)
2533 Lib = Name.slice(0, a);
2534 else
2535 Lib = Name.slice(b+1, a);
2536 // There are library names of the form: QT.A.qtx so check for these.
2537 if (Lib.size() >= 3) {
2538 Dot = Lib.substr(Lib.size() - 2, 1);
2539 if (Dot == ".")
2540 Lib = Lib.slice(0, Lib.size()-2);
2542 return Lib;
2545 // getLibraryShortNameByIndex() is used to get the short name of the library
2546 // for an undefined symbol in a linked Mach-O binary that was linked with the
2547 // normal two-level namespace default (that is MH_TWOLEVEL in the header).
2548 // It is passed the index (0 - based) of the library as translated from
2549 // GET_LIBRARY_ORDINAL (1 - based).
2550 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index,
2551 StringRef &Res) const {
2552 if (Index >= Libraries.size())
2553 return object_error::parse_failed;
2555 // If the cache of LibrariesShortNames is not built up do that first for
2556 // all the Libraries.
2557 if (LibrariesShortNames.size() == 0) {
2558 for (unsigned i = 0; i < Libraries.size(); i++) {
2559 auto CommandOrErr =
2560 getStructOrErr<MachO::dylib_command>(*this, Libraries[i]);
2561 if (!CommandOrErr)
2562 return object_error::parse_failed;
2563 MachO::dylib_command D = CommandOrErr.get();
2564 if (D.dylib.name >= D.cmdsize)
2565 return object_error::parse_failed;
2566 const char *P = (const char *)(Libraries[i]) + D.dylib.name;
2567 StringRef Name = StringRef(P);
2568 if (D.dylib.name+Name.size() >= D.cmdsize)
2569 return object_error::parse_failed;
2570 StringRef Suffix;
2571 bool isFramework;
2572 StringRef shortName = guessLibraryShortName(Name, isFramework, Suffix);
2573 if (shortName.empty())
2574 LibrariesShortNames.push_back(Name);
2575 else
2576 LibrariesShortNames.push_back(shortName);
2580 Res = LibrariesShortNames[Index];
2581 return std::error_code();
2584 uint32_t MachOObjectFile::getLibraryCount() const {
2585 return Libraries.size();
2588 section_iterator
2589 MachOObjectFile::getRelocationRelocatedSection(relocation_iterator Rel) const {
2590 DataRefImpl Sec;
2591 Sec.d.a = Rel->getRawDataRefImpl().d.a;
2592 return section_iterator(SectionRef(Sec, this));
2595 basic_symbol_iterator MachOObjectFile::symbol_begin() const {
2596 DataRefImpl DRI;
2597 MachO::symtab_command Symtab = getSymtabLoadCommand();
2598 if (!SymtabLoadCmd || Symtab.nsyms == 0)
2599 return basic_symbol_iterator(SymbolRef(DRI, this));
2601 return getSymbolByIndex(0);
2604 basic_symbol_iterator MachOObjectFile::symbol_end() const {
2605 DataRefImpl DRI;
2606 MachO::symtab_command Symtab = getSymtabLoadCommand();
2607 if (!SymtabLoadCmd || Symtab.nsyms == 0)
2608 return basic_symbol_iterator(SymbolRef(DRI, this));
2610 unsigned SymbolTableEntrySize = is64Bit() ?
2611 sizeof(MachO::nlist_64) :
2612 sizeof(MachO::nlist);
2613 unsigned Offset = Symtab.symoff +
2614 Symtab.nsyms * SymbolTableEntrySize;
2615 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset));
2616 return basic_symbol_iterator(SymbolRef(DRI, this));
2619 symbol_iterator MachOObjectFile::getSymbolByIndex(unsigned Index) const {
2620 MachO::symtab_command Symtab = getSymtabLoadCommand();
2621 if (!SymtabLoadCmd || Index >= Symtab.nsyms)
2622 report_fatal_error("Requested symbol index is out of range.");
2623 unsigned SymbolTableEntrySize =
2624 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist);
2625 DataRefImpl DRI;
2626 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff));
2627 DRI.p += Index * SymbolTableEntrySize;
2628 return basic_symbol_iterator(SymbolRef(DRI, this));
2631 uint64_t MachOObjectFile::getSymbolIndex(DataRefImpl Symb) const {
2632 MachO::symtab_command Symtab = getSymtabLoadCommand();
2633 if (!SymtabLoadCmd)
2634 report_fatal_error("getSymbolIndex() called with no symbol table symbol");
2635 unsigned SymbolTableEntrySize =
2636 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist);
2637 DataRefImpl DRIstart;
2638 DRIstart.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff));
2639 uint64_t Index = (Symb.p - DRIstart.p) / SymbolTableEntrySize;
2640 return Index;
2643 section_iterator MachOObjectFile::section_begin() const {
2644 DataRefImpl DRI;
2645 return section_iterator(SectionRef(DRI, this));
2648 section_iterator MachOObjectFile::section_end() const {
2649 DataRefImpl DRI;
2650 DRI.d.a = Sections.size();
2651 return section_iterator(SectionRef(DRI, this));
2654 uint8_t MachOObjectFile::getBytesInAddress() const {
2655 return is64Bit() ? 8 : 4;
2658 StringRef MachOObjectFile::getFileFormatName() const {
2659 unsigned CPUType = getCPUType(*this);
2660 if (!is64Bit()) {
2661 switch (CPUType) {
2662 case MachO::CPU_TYPE_I386:
2663 return "Mach-O 32-bit i386";
2664 case MachO::CPU_TYPE_ARM:
2665 return "Mach-O arm";
2666 case MachO::CPU_TYPE_ARM64_32:
2667 return "Mach-O arm64 (ILP32)";
2668 case MachO::CPU_TYPE_POWERPC:
2669 return "Mach-O 32-bit ppc";
2670 default:
2671 return "Mach-O 32-bit unknown";
2675 switch (CPUType) {
2676 case MachO::CPU_TYPE_X86_64:
2677 return "Mach-O 64-bit x86-64";
2678 case MachO::CPU_TYPE_ARM64:
2679 return "Mach-O arm64";
2680 case MachO::CPU_TYPE_POWERPC64:
2681 return "Mach-O 64-bit ppc64";
2682 default:
2683 return "Mach-O 64-bit unknown";
2687 Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) {
2688 switch (CPUType) {
2689 case MachO::CPU_TYPE_I386:
2690 return Triple::x86;
2691 case MachO::CPU_TYPE_X86_64:
2692 return Triple::x86_64;
2693 case MachO::CPU_TYPE_ARM:
2694 return Triple::arm;
2695 case MachO::CPU_TYPE_ARM64:
2696 return Triple::aarch64;
2697 case MachO::CPU_TYPE_ARM64_32:
2698 return Triple::aarch64_32;
2699 case MachO::CPU_TYPE_POWERPC:
2700 return Triple::ppc;
2701 case MachO::CPU_TYPE_POWERPC64:
2702 return Triple::ppc64;
2703 default:
2704 return Triple::UnknownArch;
2708 Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType,
2709 const char **McpuDefault,
2710 const char **ArchFlag) {
2711 if (McpuDefault)
2712 *McpuDefault = nullptr;
2713 if (ArchFlag)
2714 *ArchFlag = nullptr;
2716 switch (CPUType) {
2717 case MachO::CPU_TYPE_I386:
2718 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
2719 case MachO::CPU_SUBTYPE_I386_ALL:
2720 if (ArchFlag)
2721 *ArchFlag = "i386";
2722 return Triple("i386-apple-darwin");
2723 default:
2724 return Triple();
2726 case MachO::CPU_TYPE_X86_64:
2727 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
2728 case MachO::CPU_SUBTYPE_X86_64_ALL:
2729 if (ArchFlag)
2730 *ArchFlag = "x86_64";
2731 return Triple("x86_64-apple-darwin");
2732 case MachO::CPU_SUBTYPE_X86_64_H:
2733 if (ArchFlag)
2734 *ArchFlag = "x86_64h";
2735 return Triple("x86_64h-apple-darwin");
2736 default:
2737 return Triple();
2739 case MachO::CPU_TYPE_ARM:
2740 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
2741 case MachO::CPU_SUBTYPE_ARM_V4T:
2742 if (ArchFlag)
2743 *ArchFlag = "armv4t";
2744 return Triple("armv4t-apple-darwin");
2745 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
2746 if (ArchFlag)
2747 *ArchFlag = "armv5e";
2748 return Triple("armv5e-apple-darwin");
2749 case MachO::CPU_SUBTYPE_ARM_XSCALE:
2750 if (ArchFlag)
2751 *ArchFlag = "xscale";
2752 return Triple("xscale-apple-darwin");
2753 case MachO::CPU_SUBTYPE_ARM_V6:
2754 if (ArchFlag)
2755 *ArchFlag = "armv6";
2756 return Triple("armv6-apple-darwin");
2757 case MachO::CPU_SUBTYPE_ARM_V6M:
2758 if (McpuDefault)
2759 *McpuDefault = "cortex-m0";
2760 if (ArchFlag)
2761 *ArchFlag = "armv6m";
2762 return Triple("armv6m-apple-darwin");
2763 case MachO::CPU_SUBTYPE_ARM_V7:
2764 if (ArchFlag)
2765 *ArchFlag = "armv7";
2766 return Triple("armv7-apple-darwin");
2767 case MachO::CPU_SUBTYPE_ARM_V7EM:
2768 if (McpuDefault)
2769 *McpuDefault = "cortex-m4";
2770 if (ArchFlag)
2771 *ArchFlag = "armv7em";
2772 return Triple("thumbv7em-apple-darwin");
2773 case MachO::CPU_SUBTYPE_ARM_V7K:
2774 if (McpuDefault)
2775 *McpuDefault = "cortex-a7";
2776 if (ArchFlag)
2777 *ArchFlag = "armv7k";
2778 return Triple("armv7k-apple-darwin");
2779 case MachO::CPU_SUBTYPE_ARM_V7M:
2780 if (McpuDefault)
2781 *McpuDefault = "cortex-m3";
2782 if (ArchFlag)
2783 *ArchFlag = "armv7m";
2784 return Triple("thumbv7m-apple-darwin");
2785 case MachO::CPU_SUBTYPE_ARM_V7S:
2786 if (McpuDefault)
2787 *McpuDefault = "cortex-a7";
2788 if (ArchFlag)
2789 *ArchFlag = "armv7s";
2790 return Triple("armv7s-apple-darwin");
2791 default:
2792 return Triple();
2794 case MachO::CPU_TYPE_ARM64:
2795 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
2796 case MachO::CPU_SUBTYPE_ARM64_ALL:
2797 if (McpuDefault)
2798 *McpuDefault = "cyclone";
2799 if (ArchFlag)
2800 *ArchFlag = "arm64";
2801 return Triple("arm64-apple-darwin");
2802 case MachO::CPU_SUBTYPE_ARM64E:
2803 if (McpuDefault)
2804 *McpuDefault = "apple-a12";
2805 if (ArchFlag)
2806 *ArchFlag = "arm64e";
2807 return Triple("arm64e-apple-darwin");
2808 default:
2809 return Triple();
2811 case MachO::CPU_TYPE_ARM64_32:
2812 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
2813 case MachO::CPU_SUBTYPE_ARM64_32_V8:
2814 if (McpuDefault)
2815 *McpuDefault = "cyclone";
2816 if (ArchFlag)
2817 *ArchFlag = "arm64_32";
2818 return Triple("arm64_32-apple-darwin");
2819 default:
2820 return Triple();
2822 case MachO::CPU_TYPE_POWERPC:
2823 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
2824 case MachO::CPU_SUBTYPE_POWERPC_ALL:
2825 if (ArchFlag)
2826 *ArchFlag = "ppc";
2827 return Triple("ppc-apple-darwin");
2828 default:
2829 return Triple();
2831 case MachO::CPU_TYPE_POWERPC64:
2832 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) {
2833 case MachO::CPU_SUBTYPE_POWERPC_ALL:
2834 if (ArchFlag)
2835 *ArchFlag = "ppc64";
2836 return Triple("ppc64-apple-darwin");
2837 default:
2838 return Triple();
2840 default:
2841 return Triple();
2845 Triple MachOObjectFile::getHostArch() {
2846 return Triple(sys::getDefaultTargetTriple());
2849 bool MachOObjectFile::isValidArch(StringRef ArchFlag) {
2850 auto validArchs = getValidArchs();
2851 return llvm::is_contained(validArchs, ArchFlag);
2854 ArrayRef<StringRef> MachOObjectFile::getValidArchs() {
2855 static const std::array<StringRef, 18> ValidArchs = {{
2856 "i386",
2857 "x86_64",
2858 "x86_64h",
2859 "armv4t",
2860 "arm",
2861 "armv5e",
2862 "armv6",
2863 "armv6m",
2864 "armv7",
2865 "armv7em",
2866 "armv7k",
2867 "armv7m",
2868 "armv7s",
2869 "arm64",
2870 "arm64e",
2871 "arm64_32",
2872 "ppc",
2873 "ppc64",
2876 return ValidArchs;
2879 Triple::ArchType MachOObjectFile::getArch() const {
2880 return getArch(getCPUType(*this), getCPUSubType(*this));
2883 Triple MachOObjectFile::getArchTriple(const char **McpuDefault) const {
2884 return getArchTriple(Header.cputype, Header.cpusubtype, McpuDefault);
2887 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const {
2888 DataRefImpl DRI;
2889 DRI.d.a = Index;
2890 return section_rel_begin(DRI);
2893 relocation_iterator MachOObjectFile::section_rel_end(unsigned Index) const {
2894 DataRefImpl DRI;
2895 DRI.d.a = Index;
2896 return section_rel_end(DRI);
2899 dice_iterator MachOObjectFile::begin_dices() const {
2900 DataRefImpl DRI;
2901 if (!DataInCodeLoadCmd)
2902 return dice_iterator(DiceRef(DRI, this));
2904 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand();
2905 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, DicLC.dataoff));
2906 return dice_iterator(DiceRef(DRI, this));
2909 dice_iterator MachOObjectFile::end_dices() const {
2910 DataRefImpl DRI;
2911 if (!DataInCodeLoadCmd)
2912 return dice_iterator(DiceRef(DRI, this));
2914 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand();
2915 unsigned Offset = DicLC.dataoff + DicLC.datasize;
2916 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset));
2917 return dice_iterator(DiceRef(DRI, this));
2920 ExportEntry::ExportEntry(Error *E, const MachOObjectFile *O,
2921 ArrayRef<uint8_t> T) : E(E), O(O), Trie(T) {}
2923 void ExportEntry::moveToFirst() {
2924 ErrorAsOutParameter ErrAsOutParam(E);
2925 pushNode(0);
2926 if (*E)
2927 return;
2928 pushDownUntilBottom();
2931 void ExportEntry::moveToEnd() {
2932 Stack.clear();
2933 Done = true;
2936 bool ExportEntry::operator==(const ExportEntry &Other) const {
2937 // Common case, one at end, other iterating from begin.
2938 if (Done || Other.Done)
2939 return (Done == Other.Done);
2940 // Not equal if different stack sizes.
2941 if (Stack.size() != Other.Stack.size())
2942 return false;
2943 // Not equal if different cumulative strings.
2944 if (!CumulativeString.equals(Other.CumulativeString))
2945 return false;
2946 // Equal if all nodes in both stacks match.
2947 for (unsigned i=0; i < Stack.size(); ++i) {
2948 if (Stack[i].Start != Other.Stack[i].Start)
2949 return false;
2951 return true;
2954 uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr, const char **error) {
2955 unsigned Count;
2956 uint64_t Result = decodeULEB128(Ptr, &Count, Trie.end(), error);
2957 Ptr += Count;
2958 if (Ptr > Trie.end())
2959 Ptr = Trie.end();
2960 return Result;
2963 StringRef ExportEntry::name() const {
2964 return CumulativeString;
2967 uint64_t ExportEntry::flags() const {
2968 return Stack.back().Flags;
2971 uint64_t ExportEntry::address() const {
2972 return Stack.back().Address;
2975 uint64_t ExportEntry::other() const {
2976 return Stack.back().Other;
2979 StringRef ExportEntry::otherName() const {
2980 const char* ImportName = Stack.back().ImportName;
2981 if (ImportName)
2982 return StringRef(ImportName);
2983 return StringRef();
2986 uint32_t ExportEntry::nodeOffset() const {
2987 return Stack.back().Start - Trie.begin();
2990 ExportEntry::NodeState::NodeState(const uint8_t *Ptr)
2991 : Start(Ptr), Current(Ptr) {}
2993 void ExportEntry::pushNode(uint64_t offset) {
2994 ErrorAsOutParameter ErrAsOutParam(E);
2995 const uint8_t *Ptr = Trie.begin() + offset;
2996 NodeState State(Ptr);
2997 const char *error = nullptr;
2998 uint64_t ExportInfoSize = readULEB128(State.Current, &error);
2999 if (error) {
3000 *E = malformedError("export info size " + Twine(error) +
3001 " in export trie data at node: 0x" +
3002 Twine::utohexstr(offset));
3003 moveToEnd();
3004 return;
3006 State.IsExportNode = (ExportInfoSize != 0);
3007 const uint8_t* Children = State.Current + ExportInfoSize;
3008 if (Children > Trie.end()) {
3009 *E = malformedError(
3010 "export info size: 0x" + Twine::utohexstr(ExportInfoSize) +
3011 " in export trie data at node: 0x" + Twine::utohexstr(offset) +
3012 " too big and extends past end of trie data");
3013 moveToEnd();
3014 return;
3016 if (State.IsExportNode) {
3017 const uint8_t *ExportStart = State.Current;
3018 State.Flags = readULEB128(State.Current, &error);
3019 if (error) {
3020 *E = malformedError("flags " + Twine(error) +
3021 " in export trie data at node: 0x" +
3022 Twine::utohexstr(offset));
3023 moveToEnd();
3024 return;
3026 uint64_t Kind = State.Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK;
3027 if (State.Flags != 0 &&
3028 (Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_REGULAR &&
3029 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE &&
3030 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL)) {
3031 *E = malformedError(
3032 "unsupported exported symbol kind: " + Twine((int)Kind) +
3033 " in flags: 0x" + Twine::utohexstr(State.Flags) +
3034 " in export trie data at node: 0x" + Twine::utohexstr(offset));
3035 moveToEnd();
3036 return;
3038 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) {
3039 State.Address = 0;
3040 State.Other = readULEB128(State.Current, &error); // dylib ordinal
3041 if (error) {
3042 *E = malformedError("dylib ordinal of re-export " + Twine(error) +
3043 " in export trie data at node: 0x" +
3044 Twine::utohexstr(offset));
3045 moveToEnd();
3046 return;
3048 if (O != nullptr) {
3049 // Only positive numbers represent library ordinals. Zero and negative
3050 // numbers have special meaning (see BindSpecialDylib).
3051 if ((int64_t)State.Other > 0 && State.Other > O->getLibraryCount()) {
3052 *E = malformedError(
3053 "bad library ordinal: " + Twine((int)State.Other) + " (max " +
3054 Twine((int)O->getLibraryCount()) +
3055 ") in export trie data at node: 0x" + Twine::utohexstr(offset));
3056 moveToEnd();
3057 return;
3060 State.ImportName = reinterpret_cast<const char*>(State.Current);
3061 if (*State.ImportName == '\0') {
3062 State.Current++;
3063 } else {
3064 const uint8_t *End = State.Current + 1;
3065 if (End >= Trie.end()) {
3066 *E = malformedError("import name of re-export in export trie data at "
3067 "node: 0x" +
3068 Twine::utohexstr(offset) +
3069 " starts past end of trie data");
3070 moveToEnd();
3071 return;
3073 while(*End != '\0' && End < Trie.end())
3074 End++;
3075 if (*End != '\0') {
3076 *E = malformedError("import name of re-export in export trie data at "
3077 "node: 0x" +
3078 Twine::utohexstr(offset) +
3079 " extends past end of trie data");
3080 moveToEnd();
3081 return;
3083 State.Current = End + 1;
3085 } else {
3086 State.Address = readULEB128(State.Current, &error);
3087 if (error) {
3088 *E = malformedError("address " + Twine(error) +
3089 " in export trie data at node: 0x" +
3090 Twine::utohexstr(offset));
3091 moveToEnd();
3092 return;
3094 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) {
3095 State.Other = readULEB128(State.Current, &error);
3096 if (error) {
3097 *E = malformedError("resolver of stub and resolver " + Twine(error) +
3098 " in export trie data at node: 0x" +
3099 Twine::utohexstr(offset));
3100 moveToEnd();
3101 return;
3105 if (ExportStart + ExportInfoSize < State.Current) {
3106 *E = malformedError(
3107 "inconsistent export info size: 0x" +
3108 Twine::utohexstr(ExportInfoSize) + " where actual size was: 0x" +
3109 Twine::utohexstr(State.Current - ExportStart) +
3110 " in export trie data at node: 0x" + Twine::utohexstr(offset));
3111 moveToEnd();
3112 return;
3115 State.ChildCount = *Children;
3116 if (State.ChildCount != 0 && Children + 1 >= Trie.end()) {
3117 *E = malformedError("byte for count of childern in export trie data at "
3118 "node: 0x" +
3119 Twine::utohexstr(offset) +
3120 " extends past end of trie data");
3121 moveToEnd();
3122 return;
3124 State.Current = Children + 1;
3125 State.NextChildIndex = 0;
3126 State.ParentStringLength = CumulativeString.size();
3127 Stack.push_back(State);
3130 void ExportEntry::pushDownUntilBottom() {
3131 ErrorAsOutParameter ErrAsOutParam(E);
3132 const char *error = nullptr;
3133 while (Stack.back().NextChildIndex < Stack.back().ChildCount) {
3134 NodeState &Top = Stack.back();
3135 CumulativeString.resize(Top.ParentStringLength);
3136 for (;*Top.Current != 0 && Top.Current < Trie.end(); Top.Current++) {
3137 char C = *Top.Current;
3138 CumulativeString.push_back(C);
3140 if (Top.Current >= Trie.end()) {
3141 *E = malformedError("edge sub-string in export trie data at node: 0x" +
3142 Twine::utohexstr(Top.Start - Trie.begin()) +
3143 " for child #" + Twine((int)Top.NextChildIndex) +
3144 " extends past end of trie data");
3145 moveToEnd();
3146 return;
3148 Top.Current += 1;
3149 uint64_t childNodeIndex = readULEB128(Top.Current, &error);
3150 if (error) {
3151 *E = malformedError("child node offset " + Twine(error) +
3152 " in export trie data at node: 0x" +
3153 Twine::utohexstr(Top.Start - Trie.begin()));
3154 moveToEnd();
3155 return;
3157 for (const NodeState &node : nodes()) {
3158 if (node.Start == Trie.begin() + childNodeIndex){
3159 *E = malformedError("loop in childern in export trie data at node: 0x" +
3160 Twine::utohexstr(Top.Start - Trie.begin()) +
3161 " back to node: 0x" +
3162 Twine::utohexstr(childNodeIndex));
3163 moveToEnd();
3164 return;
3167 Top.NextChildIndex += 1;
3168 pushNode(childNodeIndex);
3169 if (*E)
3170 return;
3172 if (!Stack.back().IsExportNode) {
3173 *E = malformedError("node is not an export node in export trie data at "
3174 "node: 0x" +
3175 Twine::utohexstr(Stack.back().Start - Trie.begin()));
3176 moveToEnd();
3177 return;
3181 // We have a trie data structure and need a way to walk it that is compatible
3182 // with the C++ iterator model. The solution is a non-recursive depth first
3183 // traversal where the iterator contains a stack of parent nodes along with a
3184 // string that is the accumulation of all edge strings along the parent chain
3185 // to this point.
3187 // There is one "export" node for each exported symbol. But because some
3188 // symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export
3189 // node may have child nodes too.
3191 // The algorithm for moveNext() is to keep moving down the leftmost unvisited
3192 // child until hitting a node with no children (which is an export node or
3193 // else the trie is malformed). On the way down, each node is pushed on the
3194 // stack ivar. If there is no more ways down, it pops up one and tries to go
3195 // down a sibling path until a childless node is reached.
3196 void ExportEntry::moveNext() {
3197 assert(!Stack.empty() && "ExportEntry::moveNext() with empty node stack");
3198 if (!Stack.back().IsExportNode) {
3199 *E = malformedError("node is not an export node in export trie data at "
3200 "node: 0x" +
3201 Twine::utohexstr(Stack.back().Start - Trie.begin()));
3202 moveToEnd();
3203 return;
3206 Stack.pop_back();
3207 while (!Stack.empty()) {
3208 NodeState &Top = Stack.back();
3209 if (Top.NextChildIndex < Top.ChildCount) {
3210 pushDownUntilBottom();
3211 // Now at the next export node.
3212 return;
3213 } else {
3214 if (Top.IsExportNode) {
3215 // This node has no children but is itself an export node.
3216 CumulativeString.resize(Top.ParentStringLength);
3217 return;
3219 Stack.pop_back();
3222 Done = true;
3225 iterator_range<export_iterator>
3226 MachOObjectFile::exports(Error &E, ArrayRef<uint8_t> Trie,
3227 const MachOObjectFile *O) {
3228 ExportEntry Start(&E, O, Trie);
3229 if (Trie.empty())
3230 Start.moveToEnd();
3231 else
3232 Start.moveToFirst();
3234 ExportEntry Finish(&E, O, Trie);
3235 Finish.moveToEnd();
3237 return make_range(export_iterator(Start), export_iterator(Finish));
3240 iterator_range<export_iterator> MachOObjectFile::exports(Error &Err) const {
3241 ArrayRef<uint8_t> Trie;
3242 if (DyldInfoLoadCmd)
3243 Trie = getDyldInfoExportsTrie();
3244 else if (DyldExportsTrieLoadCmd)
3245 Trie = getDyldExportsTrie();
3247 return exports(Err, Trie, this);
3250 MachOAbstractFixupEntry::MachOAbstractFixupEntry(Error *E,
3251 const MachOObjectFile *O)
3252 : E(E), O(O) {
3253 // Cache the vmaddress of __TEXT
3254 for (const auto &Command : O->load_commands()) {
3255 if (Command.C.cmd == MachO::LC_SEGMENT) {
3256 MachO::segment_command SLC = O->getSegmentLoadCommand(Command);
3257 if (StringRef(SLC.segname) == "__TEXT") {
3258 TextAddress = SLC.vmaddr;
3259 break;
3261 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
3262 MachO::segment_command_64 SLC_64 = O->getSegment64LoadCommand(Command);
3263 if (StringRef(SLC_64.segname) == "__TEXT") {
3264 TextAddress = SLC_64.vmaddr;
3265 break;
3271 int32_t MachOAbstractFixupEntry::segmentIndex() const { return SegmentIndex; }
3273 uint64_t MachOAbstractFixupEntry::segmentOffset() const {
3274 return SegmentOffset;
3277 uint64_t MachOAbstractFixupEntry::segmentAddress() const {
3278 return O->BindRebaseAddress(SegmentIndex, 0);
3281 StringRef MachOAbstractFixupEntry::segmentName() const {
3282 return O->BindRebaseSegmentName(SegmentIndex);
3285 StringRef MachOAbstractFixupEntry::sectionName() const {
3286 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset);
3289 uint64_t MachOAbstractFixupEntry::address() const {
3290 return O->BindRebaseAddress(SegmentIndex, SegmentOffset);
3293 StringRef MachOAbstractFixupEntry::symbolName() const { return SymbolName; }
3295 int64_t MachOAbstractFixupEntry::addend() const { return Addend; }
3297 uint32_t MachOAbstractFixupEntry::flags() const { return Flags; }
3299 int MachOAbstractFixupEntry::ordinal() const { return Ordinal; }
3301 StringRef MachOAbstractFixupEntry::typeName() const { return "unknown"; }
3303 void MachOAbstractFixupEntry::moveToFirst() {
3304 SegmentOffset = 0;
3305 SegmentIndex = -1;
3306 Ordinal = 0;
3307 Flags = 0;
3308 Addend = 0;
3309 Done = false;
3312 void MachOAbstractFixupEntry::moveToEnd() { Done = true; }
3314 void MachOAbstractFixupEntry::moveNext() {}
3316 MachOChainedFixupEntry::MachOChainedFixupEntry(Error *E,
3317 const MachOObjectFile *O,
3318 bool Parse)
3319 : MachOAbstractFixupEntry(E, O) {
3320 ErrorAsOutParameter e(E);
3321 if (!Parse)
3322 return;
3324 if (auto FixupTargetsOrErr = O->getDyldChainedFixupTargets()) {
3325 FixupTargets = *FixupTargetsOrErr;
3326 } else {
3327 *E = FixupTargetsOrErr.takeError();
3328 return;
3331 if (auto SegmentsOrErr = O->getChainedFixupsSegments()) {
3332 Segments = std::move(SegmentsOrErr->second);
3333 } else {
3334 *E = SegmentsOrErr.takeError();
3335 return;
3339 void MachOChainedFixupEntry::findNextPageWithFixups() {
3340 auto FindInSegment = [this]() {
3341 const ChainedFixupsSegment &SegInfo = Segments[InfoSegIndex];
3342 while (PageIndex < SegInfo.PageStarts.size() &&
3343 SegInfo.PageStarts[PageIndex] == MachO::DYLD_CHAINED_PTR_START_NONE)
3344 ++PageIndex;
3345 return PageIndex < SegInfo.PageStarts.size();
3348 while (InfoSegIndex < Segments.size()) {
3349 if (FindInSegment()) {
3350 PageOffset = Segments[InfoSegIndex].PageStarts[PageIndex];
3351 SegmentData = O->getSegmentContents(Segments[InfoSegIndex].SegIdx);
3352 return;
3355 InfoSegIndex++;
3356 PageIndex = 0;
3360 void MachOChainedFixupEntry::moveToFirst() {
3361 MachOAbstractFixupEntry::moveToFirst();
3362 if (Segments.empty()) {
3363 Done = true;
3364 return;
3367 InfoSegIndex = 0;
3368 PageIndex = 0;
3370 findNextPageWithFixups();
3371 moveNext();
3374 void MachOChainedFixupEntry::moveToEnd() {
3375 MachOAbstractFixupEntry::moveToEnd();
3378 void MachOChainedFixupEntry::moveNext() {
3379 ErrorAsOutParameter ErrAsOutParam(E);
3381 if (InfoSegIndex == Segments.size()) {
3382 Done = true;
3383 return;
3386 const ChainedFixupsSegment &SegInfo = Segments[InfoSegIndex];
3387 SegmentIndex = SegInfo.SegIdx;
3388 SegmentOffset = SegInfo.Header.page_size * PageIndex + PageOffset;
3390 // FIXME: Handle other pointer formats.
3391 uint16_t PointerFormat = SegInfo.Header.pointer_format;
3392 if (PointerFormat != MachO::DYLD_CHAINED_PTR_64 &&
3393 PointerFormat != MachO::DYLD_CHAINED_PTR_64_OFFSET) {
3394 *E = createError("segment " + Twine(SegmentIndex) +
3395 " has unsupported chained fixup pointer_format " +
3396 Twine(PointerFormat));
3397 moveToEnd();
3398 return;
3401 Ordinal = 0;
3402 Flags = 0;
3403 Addend = 0;
3404 PointerValue = 0;
3405 SymbolName = {};
3407 if (SegmentOffset + sizeof(RawValue) > SegmentData.size()) {
3408 *E = malformedError("fixup in segment " + Twine(SegmentIndex) +
3409 " at offset " + Twine(SegmentOffset) +
3410 " extends past segment's end");
3411 moveToEnd();
3412 return;
3415 static_assert(sizeof(RawValue) == sizeof(MachO::dyld_chained_import_addend));
3416 memcpy(&RawValue, SegmentData.data() + SegmentOffset, sizeof(RawValue));
3417 if (O->isLittleEndian() != sys::IsLittleEndianHost)
3418 sys::swapByteOrder(RawValue);
3420 // The bit extraction below assumes little-endian fixup entries.
3421 assert(O->isLittleEndian() && "big-endian object should have been rejected "
3422 "by getDyldChainedFixupTargets()");
3423 auto Field = [this](uint8_t Right, uint8_t Count) {
3424 return (RawValue >> Right) & ((1ULL << Count) - 1);
3427 // The `bind` field (most significant bit) of the encoded fixup determines
3428 // whether it is dyld_chained_ptr_64_bind or dyld_chained_ptr_64_rebase.
3429 bool IsBind = Field(63, 1);
3430 Kind = IsBind ? FixupKind::Bind : FixupKind::Rebase;
3431 uint32_t Next = Field(51, 12);
3432 if (IsBind) {
3433 uint32_t ImportOrdinal = Field(0, 24);
3434 uint8_t InlineAddend = Field(24, 8);
3436 if (ImportOrdinal >= FixupTargets.size()) {
3437 *E = malformedError("fixup in segment " + Twine(SegmentIndex) +
3438 " at offset " + Twine(SegmentOffset) +
3439 " has out-of range import ordinal " +
3440 Twine(ImportOrdinal));
3441 moveToEnd();
3442 return;
3445 ChainedFixupTarget &Target = FixupTargets[ImportOrdinal];
3446 Ordinal = Target.libOrdinal();
3447 Addend = InlineAddend ? InlineAddend : Target.addend();
3448 Flags = Target.weakImport() ? MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT : 0;
3449 SymbolName = Target.symbolName();
3450 } else {
3451 uint64_t Target = Field(0, 36);
3452 uint64_t High8 = Field(36, 8);
3454 PointerValue = Target | (High8 << 56);
3455 if (PointerFormat == MachO::DYLD_CHAINED_PTR_64_OFFSET)
3456 PointerValue += textAddress();
3459 // The stride is 4 bytes for DYLD_CHAINED_PTR_64(_OFFSET).
3460 if (Next != 0) {
3461 PageOffset += 4 * Next;
3462 } else {
3463 ++PageIndex;
3464 findNextPageWithFixups();
3468 bool MachOChainedFixupEntry::operator==(
3469 const MachOChainedFixupEntry &Other) const {
3470 if (Done && Other.Done)
3471 return true;
3472 if (Done != Other.Done)
3473 return false;
3474 return InfoSegIndex == Other.InfoSegIndex && PageIndex == Other.PageIndex &&
3475 PageOffset == Other.PageOffset;
3478 MachORebaseEntry::MachORebaseEntry(Error *E, const MachOObjectFile *O,
3479 ArrayRef<uint8_t> Bytes, bool is64Bit)
3480 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()),
3481 PointerSize(is64Bit ? 8 : 4) {}
3483 void MachORebaseEntry::moveToFirst() {
3484 Ptr = Opcodes.begin();
3485 moveNext();
3488 void MachORebaseEntry::moveToEnd() {
3489 Ptr = Opcodes.end();
3490 RemainingLoopCount = 0;
3491 Done = true;
3494 void MachORebaseEntry::moveNext() {
3495 ErrorAsOutParameter ErrAsOutParam(E);
3496 // If in the middle of some loop, move to next rebasing in loop.
3497 SegmentOffset += AdvanceAmount;
3498 if (RemainingLoopCount) {
3499 --RemainingLoopCount;
3500 return;
3503 bool More = true;
3504 while (More) {
3505 // REBASE_OPCODE_DONE is only used for padding if we are not aligned to
3506 // pointer size. Therefore it is possible to reach the end without ever
3507 // having seen REBASE_OPCODE_DONE.
3508 if (Ptr == Opcodes.end()) {
3509 Done = true;
3510 return;
3513 // Parse next opcode and set up next loop.
3514 const uint8_t *OpcodeStart = Ptr;
3515 uint8_t Byte = *Ptr++;
3516 uint8_t ImmValue = Byte & MachO::REBASE_IMMEDIATE_MASK;
3517 uint8_t Opcode = Byte & MachO::REBASE_OPCODE_MASK;
3518 uint64_t Count, Skip;
3519 const char *error = nullptr;
3520 switch (Opcode) {
3521 case MachO::REBASE_OPCODE_DONE:
3522 More = false;
3523 Done = true;
3524 moveToEnd();
3525 DEBUG_WITH_TYPE("mach-o-rebase", dbgs() << "REBASE_OPCODE_DONE\n");
3526 break;
3527 case MachO::REBASE_OPCODE_SET_TYPE_IMM:
3528 RebaseType = ImmValue;
3529 if (RebaseType > MachO::REBASE_TYPE_TEXT_PCREL32) {
3530 *E = malformedError("for REBASE_OPCODE_SET_TYPE_IMM bad bind type: " +
3531 Twine((int)RebaseType) + " for opcode at: 0x" +
3532 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3533 moveToEnd();
3534 return;
3536 DEBUG_WITH_TYPE(
3537 "mach-o-rebase",
3538 dbgs() << "REBASE_OPCODE_SET_TYPE_IMM: "
3539 << "RebaseType=" << (int) RebaseType << "\n");
3540 break;
3541 case MachO::REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB:
3542 SegmentIndex = ImmValue;
3543 SegmentOffset = readULEB128(&error);
3544 if (error) {
3545 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " +
3546 Twine(error) + " for opcode at: 0x" +
3547 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3548 moveToEnd();
3549 return;
3551 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
3552 PointerSize);
3553 if (error) {
3554 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " +
3555 Twine(error) + " for opcode at: 0x" +
3556 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3557 moveToEnd();
3558 return;
3560 DEBUG_WITH_TYPE(
3561 "mach-o-rebase",
3562 dbgs() << "REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: "
3563 << "SegmentIndex=" << SegmentIndex << ", "
3564 << format("SegmentOffset=0x%06X", SegmentOffset)
3565 << "\n");
3566 break;
3567 case MachO::REBASE_OPCODE_ADD_ADDR_ULEB:
3568 SegmentOffset += readULEB128(&error);
3569 if (error) {
3570 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) +
3571 " for opcode at: 0x" +
3572 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3573 moveToEnd();
3574 return;
3576 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
3577 PointerSize);
3578 if (error) {
3579 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) +
3580 " for opcode at: 0x" +
3581 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3582 moveToEnd();
3583 return;
3585 DEBUG_WITH_TYPE("mach-o-rebase",
3586 dbgs() << "REBASE_OPCODE_ADD_ADDR_ULEB: "
3587 << format("SegmentOffset=0x%06X",
3588 SegmentOffset) << "\n");
3589 break;
3590 case MachO::REBASE_OPCODE_ADD_ADDR_IMM_SCALED:
3591 SegmentOffset += ImmValue * PointerSize;
3592 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
3593 PointerSize);
3594 if (error) {
3595 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_IMM_SCALED " +
3596 Twine(error) + " for opcode at: 0x" +
3597 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3598 moveToEnd();
3599 return;
3601 DEBUG_WITH_TYPE("mach-o-rebase",
3602 dbgs() << "REBASE_OPCODE_ADD_ADDR_IMM_SCALED: "
3603 << format("SegmentOffset=0x%06X",
3604 SegmentOffset) << "\n");
3605 break;
3606 case MachO::REBASE_OPCODE_DO_REBASE_IMM_TIMES:
3607 AdvanceAmount = PointerSize;
3608 Skip = 0;
3609 Count = ImmValue;
3610 if (ImmValue != 0)
3611 RemainingLoopCount = ImmValue - 1;
3612 else
3613 RemainingLoopCount = 0;
3614 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
3615 PointerSize, Count, Skip);
3616 if (error) {
3617 *E = malformedError("for REBASE_OPCODE_DO_REBASE_IMM_TIMES " +
3618 Twine(error) + " for opcode at: 0x" +
3619 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3620 moveToEnd();
3621 return;
3623 DEBUG_WITH_TYPE(
3624 "mach-o-rebase",
3625 dbgs() << "REBASE_OPCODE_DO_REBASE_IMM_TIMES: "
3626 << format("SegmentOffset=0x%06X", SegmentOffset)
3627 << ", AdvanceAmount=" << AdvanceAmount
3628 << ", RemainingLoopCount=" << RemainingLoopCount
3629 << "\n");
3630 return;
3631 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES:
3632 AdvanceAmount = PointerSize;
3633 Skip = 0;
3634 Count = readULEB128(&error);
3635 if (error) {
3636 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " +
3637 Twine(error) + " for opcode at: 0x" +
3638 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3639 moveToEnd();
3640 return;
3642 if (Count != 0)
3643 RemainingLoopCount = Count - 1;
3644 else
3645 RemainingLoopCount = 0;
3646 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
3647 PointerSize, Count, Skip);
3648 if (error) {
3649 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " +
3650 Twine(error) + " for opcode at: 0x" +
3651 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3652 moveToEnd();
3653 return;
3655 DEBUG_WITH_TYPE(
3656 "mach-o-rebase",
3657 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES: "
3658 << format("SegmentOffset=0x%06X", SegmentOffset)
3659 << ", AdvanceAmount=" << AdvanceAmount
3660 << ", RemainingLoopCount=" << RemainingLoopCount
3661 << "\n");
3662 return;
3663 case MachO::REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB:
3664 Skip = readULEB128(&error);
3665 if (error) {
3666 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " +
3667 Twine(error) + " for opcode at: 0x" +
3668 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3669 moveToEnd();
3670 return;
3672 AdvanceAmount = Skip + PointerSize;
3673 Count = 1;
3674 RemainingLoopCount = 0;
3675 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
3676 PointerSize, Count, Skip);
3677 if (error) {
3678 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " +
3679 Twine(error) + " for opcode at: 0x" +
3680 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3681 moveToEnd();
3682 return;
3684 DEBUG_WITH_TYPE(
3685 "mach-o-rebase",
3686 dbgs() << "REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: "
3687 << format("SegmentOffset=0x%06X", SegmentOffset)
3688 << ", AdvanceAmount=" << AdvanceAmount
3689 << ", RemainingLoopCount=" << RemainingLoopCount
3690 << "\n");
3691 return;
3692 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB:
3693 Count = readULEB128(&error);
3694 if (error) {
3695 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_"
3696 "ULEB " +
3697 Twine(error) + " for opcode at: 0x" +
3698 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3699 moveToEnd();
3700 return;
3702 if (Count != 0)
3703 RemainingLoopCount = Count - 1;
3704 else
3705 RemainingLoopCount = 0;
3706 Skip = readULEB128(&error);
3707 if (error) {
3708 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_"
3709 "ULEB " +
3710 Twine(error) + " for opcode at: 0x" +
3711 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3712 moveToEnd();
3713 return;
3715 AdvanceAmount = Skip + PointerSize;
3717 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
3718 PointerSize, Count, Skip);
3719 if (error) {
3720 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_"
3721 "ULEB " +
3722 Twine(error) + " for opcode at: 0x" +
3723 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3724 moveToEnd();
3725 return;
3727 DEBUG_WITH_TYPE(
3728 "mach-o-rebase",
3729 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: "
3730 << format("SegmentOffset=0x%06X", SegmentOffset)
3731 << ", AdvanceAmount=" << AdvanceAmount
3732 << ", RemainingLoopCount=" << RemainingLoopCount
3733 << "\n");
3734 return;
3735 default:
3736 *E = malformedError("bad rebase info (bad opcode value 0x" +
3737 Twine::utohexstr(Opcode) + " for opcode at: 0x" +
3738 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3739 moveToEnd();
3740 return;
3745 uint64_t MachORebaseEntry::readULEB128(const char **error) {
3746 unsigned Count;
3747 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error);
3748 Ptr += Count;
3749 if (Ptr > Opcodes.end())
3750 Ptr = Opcodes.end();
3751 return Result;
3754 int32_t MachORebaseEntry::segmentIndex() const { return SegmentIndex; }
3756 uint64_t MachORebaseEntry::segmentOffset() const { return SegmentOffset; }
3758 StringRef MachORebaseEntry::typeName() const {
3759 switch (RebaseType) {
3760 case MachO::REBASE_TYPE_POINTER:
3761 return "pointer";
3762 case MachO::REBASE_TYPE_TEXT_ABSOLUTE32:
3763 return "text abs32";
3764 case MachO::REBASE_TYPE_TEXT_PCREL32:
3765 return "text rel32";
3767 return "unknown";
3770 // For use with the SegIndex of a checked Mach-O Rebase entry
3771 // to get the segment name.
3772 StringRef MachORebaseEntry::segmentName() const {
3773 return O->BindRebaseSegmentName(SegmentIndex);
3776 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry
3777 // to get the section name.
3778 StringRef MachORebaseEntry::sectionName() const {
3779 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset);
3782 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry
3783 // to get the address.
3784 uint64_t MachORebaseEntry::address() const {
3785 return O->BindRebaseAddress(SegmentIndex, SegmentOffset);
3788 bool MachORebaseEntry::operator==(const MachORebaseEntry &Other) const {
3789 #ifdef EXPENSIVE_CHECKS
3790 assert(Opcodes == Other.Opcodes && "compare iterators of different files");
3791 #else
3792 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files");
3793 #endif
3794 return (Ptr == Other.Ptr) &&
3795 (RemainingLoopCount == Other.RemainingLoopCount) &&
3796 (Done == Other.Done);
3799 iterator_range<rebase_iterator>
3800 MachOObjectFile::rebaseTable(Error &Err, MachOObjectFile *O,
3801 ArrayRef<uint8_t> Opcodes, bool is64) {
3802 if (O->BindRebaseSectionTable == nullptr)
3803 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O);
3804 MachORebaseEntry Start(&Err, O, Opcodes, is64);
3805 Start.moveToFirst();
3807 MachORebaseEntry Finish(&Err, O, Opcodes, is64);
3808 Finish.moveToEnd();
3810 return make_range(rebase_iterator(Start), rebase_iterator(Finish));
3813 iterator_range<rebase_iterator> MachOObjectFile::rebaseTable(Error &Err) {
3814 return rebaseTable(Err, this, getDyldInfoRebaseOpcodes(), is64Bit());
3817 MachOBindEntry::MachOBindEntry(Error *E, const MachOObjectFile *O,
3818 ArrayRef<uint8_t> Bytes, bool is64Bit, Kind BK)
3819 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()),
3820 PointerSize(is64Bit ? 8 : 4), TableKind(BK) {}
3822 void MachOBindEntry::moveToFirst() {
3823 Ptr = Opcodes.begin();
3824 moveNext();
3827 void MachOBindEntry::moveToEnd() {
3828 Ptr = Opcodes.end();
3829 RemainingLoopCount = 0;
3830 Done = true;
3833 void MachOBindEntry::moveNext() {
3834 ErrorAsOutParameter ErrAsOutParam(E);
3835 // If in the middle of some loop, move to next binding in loop.
3836 SegmentOffset += AdvanceAmount;
3837 if (RemainingLoopCount) {
3838 --RemainingLoopCount;
3839 return;
3842 bool More = true;
3843 while (More) {
3844 // BIND_OPCODE_DONE is only used for padding if we are not aligned to
3845 // pointer size. Therefore it is possible to reach the end without ever
3846 // having seen BIND_OPCODE_DONE.
3847 if (Ptr == Opcodes.end()) {
3848 Done = true;
3849 return;
3852 // Parse next opcode and set up next loop.
3853 const uint8_t *OpcodeStart = Ptr;
3854 uint8_t Byte = *Ptr++;
3855 uint8_t ImmValue = Byte & MachO::BIND_IMMEDIATE_MASK;
3856 uint8_t Opcode = Byte & MachO::BIND_OPCODE_MASK;
3857 int8_t SignExtended;
3858 const uint8_t *SymStart;
3859 uint64_t Count, Skip;
3860 const char *error = nullptr;
3861 switch (Opcode) {
3862 case MachO::BIND_OPCODE_DONE:
3863 if (TableKind == Kind::Lazy) {
3864 // Lazying bindings have a DONE opcode between entries. Need to ignore
3865 // it to advance to next entry. But need not if this is last entry.
3866 bool NotLastEntry = false;
3867 for (const uint8_t *P = Ptr; P < Opcodes.end(); ++P) {
3868 if (*P) {
3869 NotLastEntry = true;
3872 if (NotLastEntry)
3873 break;
3875 More = false;
3876 moveToEnd();
3877 DEBUG_WITH_TYPE("mach-o-bind", dbgs() << "BIND_OPCODE_DONE\n");
3878 break;
3879 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_IMM:
3880 if (TableKind == Kind::Weak) {
3881 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_IMM not allowed in "
3882 "weak bind table for opcode at: 0x" +
3883 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3884 moveToEnd();
3885 return;
3887 Ordinal = ImmValue;
3888 LibraryOrdinalSet = true;
3889 if (ImmValue > O->getLibraryCount()) {
3890 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad "
3891 "library ordinal: " +
3892 Twine((int)ImmValue) + " (max " +
3893 Twine((int)O->getLibraryCount()) +
3894 ") for opcode at: 0x" +
3895 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3896 moveToEnd();
3897 return;
3899 DEBUG_WITH_TYPE(
3900 "mach-o-bind",
3901 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: "
3902 << "Ordinal=" << Ordinal << "\n");
3903 break;
3904 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB:
3905 if (TableKind == Kind::Weak) {
3906 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB not allowed in "
3907 "weak bind table for opcode at: 0x" +
3908 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3909 moveToEnd();
3910 return;
3912 Ordinal = readULEB128(&error);
3913 LibraryOrdinalSet = true;
3914 if (error) {
3915 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB " +
3916 Twine(error) + " for opcode at: 0x" +
3917 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3918 moveToEnd();
3919 return;
3921 if (Ordinal > (int)O->getLibraryCount()) {
3922 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad "
3923 "library ordinal: " +
3924 Twine((int)Ordinal) + " (max " +
3925 Twine((int)O->getLibraryCount()) +
3926 ") for opcode at: 0x" +
3927 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3928 moveToEnd();
3929 return;
3931 DEBUG_WITH_TYPE(
3932 "mach-o-bind",
3933 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: "
3934 << "Ordinal=" << Ordinal << "\n");
3935 break;
3936 case MachO::BIND_OPCODE_SET_DYLIB_SPECIAL_IMM:
3937 if (TableKind == Kind::Weak) {
3938 *E = malformedError("BIND_OPCODE_SET_DYLIB_SPECIAL_IMM not allowed in "
3939 "weak bind table for opcode at: 0x" +
3940 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3941 moveToEnd();
3942 return;
3944 if (ImmValue) {
3945 SignExtended = MachO::BIND_OPCODE_MASK | ImmValue;
3946 Ordinal = SignExtended;
3947 if (Ordinal < MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) {
3948 *E = malformedError("for BIND_OPCODE_SET_DYLIB_SPECIAL_IMM unknown "
3949 "special ordinal: " +
3950 Twine((int)Ordinal) + " for opcode at: 0x" +
3951 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3952 moveToEnd();
3953 return;
3955 } else
3956 Ordinal = 0;
3957 LibraryOrdinalSet = true;
3958 DEBUG_WITH_TYPE(
3959 "mach-o-bind",
3960 dbgs() << "BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: "
3961 << "Ordinal=" << Ordinal << "\n");
3962 break;
3963 case MachO::BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM:
3964 Flags = ImmValue;
3965 SymStart = Ptr;
3966 while (*Ptr && (Ptr < Opcodes.end())) {
3967 ++Ptr;
3969 if (Ptr == Opcodes.end()) {
3970 *E = malformedError(
3971 "for BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM "
3972 "symbol name extends past opcodes for opcode at: 0x" +
3973 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3974 moveToEnd();
3975 return;
3977 SymbolName = StringRef(reinterpret_cast<const char*>(SymStart),
3978 Ptr-SymStart);
3979 ++Ptr;
3980 DEBUG_WITH_TYPE(
3981 "mach-o-bind",
3982 dbgs() << "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: "
3983 << "SymbolName=" << SymbolName << "\n");
3984 if (TableKind == Kind::Weak) {
3985 if (ImmValue & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION)
3986 return;
3988 break;
3989 case MachO::BIND_OPCODE_SET_TYPE_IMM:
3990 BindType = ImmValue;
3991 if (ImmValue > MachO::BIND_TYPE_TEXT_PCREL32) {
3992 *E = malformedError("for BIND_OPCODE_SET_TYPE_IMM bad bind type: " +
3993 Twine((int)ImmValue) + " for opcode at: 0x" +
3994 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
3995 moveToEnd();
3996 return;
3998 DEBUG_WITH_TYPE(
3999 "mach-o-bind",
4000 dbgs() << "BIND_OPCODE_SET_TYPE_IMM: "
4001 << "BindType=" << (int)BindType << "\n");
4002 break;
4003 case MachO::BIND_OPCODE_SET_ADDEND_SLEB:
4004 Addend = readSLEB128(&error);
4005 if (error) {
4006 *E = malformedError("for BIND_OPCODE_SET_ADDEND_SLEB " + Twine(error) +
4007 " for opcode at: 0x" +
4008 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4009 moveToEnd();
4010 return;
4012 DEBUG_WITH_TYPE(
4013 "mach-o-bind",
4014 dbgs() << "BIND_OPCODE_SET_ADDEND_SLEB: "
4015 << "Addend=" << Addend << "\n");
4016 break;
4017 case MachO::BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB:
4018 SegmentIndex = ImmValue;
4019 SegmentOffset = readULEB128(&error);
4020 if (error) {
4021 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " +
4022 Twine(error) + " for opcode at: 0x" +
4023 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4024 moveToEnd();
4025 return;
4027 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
4028 PointerSize);
4029 if (error) {
4030 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " +
4031 Twine(error) + " for opcode at: 0x" +
4032 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4033 moveToEnd();
4034 return;
4036 DEBUG_WITH_TYPE(
4037 "mach-o-bind",
4038 dbgs() << "BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: "
4039 << "SegmentIndex=" << SegmentIndex << ", "
4040 << format("SegmentOffset=0x%06X", SegmentOffset)
4041 << "\n");
4042 break;
4043 case MachO::BIND_OPCODE_ADD_ADDR_ULEB:
4044 SegmentOffset += readULEB128(&error);
4045 if (error) {
4046 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) +
4047 " for opcode at: 0x" +
4048 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4049 moveToEnd();
4050 return;
4052 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
4053 PointerSize);
4054 if (error) {
4055 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) +
4056 " for opcode at: 0x" +
4057 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4058 moveToEnd();
4059 return;
4061 DEBUG_WITH_TYPE("mach-o-bind",
4062 dbgs() << "BIND_OPCODE_ADD_ADDR_ULEB: "
4063 << format("SegmentOffset=0x%06X",
4064 SegmentOffset) << "\n");
4065 break;
4066 case MachO::BIND_OPCODE_DO_BIND:
4067 AdvanceAmount = PointerSize;
4068 RemainingLoopCount = 0;
4069 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
4070 PointerSize);
4071 if (error) {
4072 *E = malformedError("for BIND_OPCODE_DO_BIND " + Twine(error) +
4073 " for opcode at: 0x" +
4074 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4075 moveToEnd();
4076 return;
4078 if (SymbolName == StringRef()) {
4079 *E = malformedError(
4080 "for BIND_OPCODE_DO_BIND missing preceding "
4081 "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode at: 0x" +
4082 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4083 moveToEnd();
4084 return;
4086 if (!LibraryOrdinalSet && TableKind != Kind::Weak) {
4087 *E =
4088 malformedError("for BIND_OPCODE_DO_BIND missing preceding "
4089 "BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" +
4090 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4091 moveToEnd();
4092 return;
4094 DEBUG_WITH_TYPE("mach-o-bind",
4095 dbgs() << "BIND_OPCODE_DO_BIND: "
4096 << format("SegmentOffset=0x%06X",
4097 SegmentOffset) << "\n");
4098 return;
4099 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB:
4100 if (TableKind == Kind::Lazy) {
4101 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB not allowed in "
4102 "lazy bind table for opcode at: 0x" +
4103 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4104 moveToEnd();
4105 return;
4107 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
4108 PointerSize);
4109 if (error) {
4110 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " +
4111 Twine(error) + " for opcode at: 0x" +
4112 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4113 moveToEnd();
4114 return;
4116 if (SymbolName == StringRef()) {
4117 *E = malformedError(
4118 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing "
4119 "preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode "
4120 "at: 0x" +
4121 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4122 moveToEnd();
4123 return;
4125 if (!LibraryOrdinalSet && TableKind != Kind::Weak) {
4126 *E = malformedError(
4127 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing "
4128 "preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" +
4129 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4130 moveToEnd();
4131 return;
4133 AdvanceAmount = readULEB128(&error) + PointerSize;
4134 if (error) {
4135 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " +
4136 Twine(error) + " for opcode at: 0x" +
4137 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4138 moveToEnd();
4139 return;
4141 // Note, this is not really an error until the next bind but make no sense
4142 // for a BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB to not be followed by another
4143 // bind operation.
4144 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset +
4145 AdvanceAmount, PointerSize);
4146 if (error) {
4147 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB (after adding "
4148 "ULEB) " +
4149 Twine(error) + " for opcode at: 0x" +
4150 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4151 moveToEnd();
4152 return;
4154 RemainingLoopCount = 0;
4155 DEBUG_WITH_TYPE(
4156 "mach-o-bind",
4157 dbgs() << "BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: "
4158 << format("SegmentOffset=0x%06X", SegmentOffset)
4159 << ", AdvanceAmount=" << AdvanceAmount
4160 << ", RemainingLoopCount=" << RemainingLoopCount
4161 << "\n");
4162 return;
4163 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED:
4164 if (TableKind == Kind::Lazy) {
4165 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED not "
4166 "allowed in lazy bind table for opcode at: 0x" +
4167 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4168 moveToEnd();
4169 return;
4171 if (SymbolName == StringRef()) {
4172 *E = malformedError(
4173 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED "
4174 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for "
4175 "opcode at: 0x" +
4176 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4177 moveToEnd();
4178 return;
4180 if (!LibraryOrdinalSet && TableKind != Kind::Weak) {
4181 *E = malformedError(
4182 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED "
4183 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode "
4184 "at: 0x" +
4185 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4186 moveToEnd();
4187 return;
4189 AdvanceAmount = ImmValue * PointerSize + PointerSize;
4190 RemainingLoopCount = 0;
4191 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset +
4192 AdvanceAmount, PointerSize);
4193 if (error) {
4194 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " +
4195 Twine(error) + " for opcode at: 0x" +
4196 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4197 moveToEnd();
4198 return;
4200 DEBUG_WITH_TYPE("mach-o-bind",
4201 dbgs()
4202 << "BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: "
4203 << format("SegmentOffset=0x%06X", SegmentOffset) << "\n");
4204 return;
4205 case MachO::BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB:
4206 if (TableKind == Kind::Lazy) {
4207 *E = malformedError("BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB not "
4208 "allowed in lazy bind table for opcode at: 0x" +
4209 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4210 moveToEnd();
4211 return;
4213 Count = readULEB128(&error);
4214 if (Count != 0)
4215 RemainingLoopCount = Count - 1;
4216 else
4217 RemainingLoopCount = 0;
4218 if (error) {
4219 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB "
4220 " (count value) " +
4221 Twine(error) + " for opcode at: 0x" +
4222 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4223 moveToEnd();
4224 return;
4226 Skip = readULEB128(&error);
4227 AdvanceAmount = Skip + PointerSize;
4228 if (error) {
4229 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB "
4230 " (skip value) " +
4231 Twine(error) + " for opcode at: 0x" +
4232 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4233 moveToEnd();
4234 return;
4236 if (SymbolName == StringRef()) {
4237 *E = malformedError(
4238 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB "
4239 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for "
4240 "opcode at: 0x" +
4241 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4242 moveToEnd();
4243 return;
4245 if (!LibraryOrdinalSet && TableKind != Kind::Weak) {
4246 *E = malformedError(
4247 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB "
4248 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode "
4249 "at: 0x" +
4250 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4251 moveToEnd();
4252 return;
4254 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset,
4255 PointerSize, Count, Skip);
4256 if (error) {
4257 *E =
4258 malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " +
4259 Twine(error) + " for opcode at: 0x" +
4260 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4261 moveToEnd();
4262 return;
4264 DEBUG_WITH_TYPE(
4265 "mach-o-bind",
4266 dbgs() << "BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: "
4267 << format("SegmentOffset=0x%06X", SegmentOffset)
4268 << ", AdvanceAmount=" << AdvanceAmount
4269 << ", RemainingLoopCount=" << RemainingLoopCount
4270 << "\n");
4271 return;
4272 default:
4273 *E = malformedError("bad bind info (bad opcode value 0x" +
4274 Twine::utohexstr(Opcode) + " for opcode at: 0x" +
4275 Twine::utohexstr(OpcodeStart - Opcodes.begin()));
4276 moveToEnd();
4277 return;
4282 uint64_t MachOBindEntry::readULEB128(const char **error) {
4283 unsigned Count;
4284 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error);
4285 Ptr += Count;
4286 if (Ptr > Opcodes.end())
4287 Ptr = Opcodes.end();
4288 return Result;
4291 int64_t MachOBindEntry::readSLEB128(const char **error) {
4292 unsigned Count;
4293 int64_t Result = decodeSLEB128(Ptr, &Count, Opcodes.end(), error);
4294 Ptr += Count;
4295 if (Ptr > Opcodes.end())
4296 Ptr = Opcodes.end();
4297 return Result;
4300 int32_t MachOBindEntry::segmentIndex() const { return SegmentIndex; }
4302 uint64_t MachOBindEntry::segmentOffset() const { return SegmentOffset; }
4304 StringRef MachOBindEntry::typeName() const {
4305 switch (BindType) {
4306 case MachO::BIND_TYPE_POINTER:
4307 return "pointer";
4308 case MachO::BIND_TYPE_TEXT_ABSOLUTE32:
4309 return "text abs32";
4310 case MachO::BIND_TYPE_TEXT_PCREL32:
4311 return "text rel32";
4313 return "unknown";
4316 StringRef MachOBindEntry::symbolName() const { return SymbolName; }
4318 int64_t MachOBindEntry::addend() const { return Addend; }
4320 uint32_t MachOBindEntry::flags() const { return Flags; }
4322 int MachOBindEntry::ordinal() const { return Ordinal; }
4324 // For use with the SegIndex of a checked Mach-O Bind entry
4325 // to get the segment name.
4326 StringRef MachOBindEntry::segmentName() const {
4327 return O->BindRebaseSegmentName(SegmentIndex);
4330 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry
4331 // to get the section name.
4332 StringRef MachOBindEntry::sectionName() const {
4333 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset);
4336 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry
4337 // to get the address.
4338 uint64_t MachOBindEntry::address() const {
4339 return O->BindRebaseAddress(SegmentIndex, SegmentOffset);
4342 bool MachOBindEntry::operator==(const MachOBindEntry &Other) const {
4343 #ifdef EXPENSIVE_CHECKS
4344 assert(Opcodes == Other.Opcodes && "compare iterators of different files");
4345 #else
4346 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files");
4347 #endif
4348 return (Ptr == Other.Ptr) &&
4349 (RemainingLoopCount == Other.RemainingLoopCount) &&
4350 (Done == Other.Done);
4353 // Build table of sections so SegIndex/SegOffset pairs can be translated.
4354 BindRebaseSegInfo::BindRebaseSegInfo(const object::MachOObjectFile *Obj) {
4355 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
4356 StringRef CurSegName;
4357 uint64_t CurSegAddress;
4358 for (const SectionRef &Section : Obj->sections()) {
4359 SectionInfo Info;
4360 Expected<StringRef> NameOrErr = Section.getName();
4361 if (!NameOrErr)
4362 consumeError(NameOrErr.takeError());
4363 else
4364 Info.SectionName = *NameOrErr;
4365 Info.Address = Section.getAddress();
4366 Info.Size = Section.getSize();
4367 Info.SegmentName =
4368 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
4369 if (Info.SegmentName != CurSegName) {
4370 ++CurSegIndex;
4371 CurSegName = Info.SegmentName;
4372 CurSegAddress = Info.Address;
4374 Info.SegmentIndex = CurSegIndex - 1;
4375 Info.OffsetInSegment = Info.Address - CurSegAddress;
4376 Info.SegmentStartAddress = CurSegAddress;
4377 Sections.push_back(Info);
4379 MaxSegIndex = CurSegIndex;
4382 // For use with a SegIndex, SegOffset, and PointerSize triple in
4383 // MachOBindEntry::moveNext() to validate a MachOBindEntry or MachORebaseEntry.
4385 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists
4386 // that fully contains a pointer at that location. Multiple fixups in a bind
4387 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can
4388 // be tested via the Count and Skip parameters.
4389 const char *BindRebaseSegInfo::checkSegAndOffsets(int32_t SegIndex,
4390 uint64_t SegOffset,
4391 uint8_t PointerSize,
4392 uint64_t Count,
4393 uint64_t Skip) {
4394 if (SegIndex == -1)
4395 return "missing preceding *_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB";
4396 if (SegIndex >= MaxSegIndex)
4397 return "bad segIndex (too large)";
4398 for (uint64_t i = 0; i < Count; ++i) {
4399 uint64_t Start = SegOffset + i * (PointerSize + Skip);
4400 uint64_t End = Start + PointerSize;
4401 bool Found = false;
4402 for (const SectionInfo &SI : Sections) {
4403 if (SI.SegmentIndex != SegIndex)
4404 continue;
4405 if ((SI.OffsetInSegment<=Start) && (Start<(SI.OffsetInSegment+SI.Size))) {
4406 if (End <= SI.OffsetInSegment + SI.Size) {
4407 Found = true;
4408 break;
4410 else
4411 return "bad offset, extends beyond section boundary";
4414 if (!Found)
4415 return "bad offset, not in section";
4417 return nullptr;
4420 // For use with the SegIndex of a checked Mach-O Bind or Rebase entry
4421 // to get the segment name.
4422 StringRef BindRebaseSegInfo::segmentName(int32_t SegIndex) {
4423 for (const SectionInfo &SI : Sections) {
4424 if (SI.SegmentIndex == SegIndex)
4425 return SI.SegmentName;
4427 llvm_unreachable("invalid SegIndex");
4430 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase
4431 // to get the SectionInfo.
4432 const BindRebaseSegInfo::SectionInfo &BindRebaseSegInfo::findSection(
4433 int32_t SegIndex, uint64_t SegOffset) {
4434 for (const SectionInfo &SI : Sections) {
4435 if (SI.SegmentIndex != SegIndex)
4436 continue;
4437 if (SI.OffsetInSegment > SegOffset)
4438 continue;
4439 if (SegOffset >= (SI.OffsetInSegment + SI.Size))
4440 continue;
4441 return SI;
4443 llvm_unreachable("SegIndex and SegOffset not in any section");
4446 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase
4447 // entry to get the section name.
4448 StringRef BindRebaseSegInfo::sectionName(int32_t SegIndex,
4449 uint64_t SegOffset) {
4450 return findSection(SegIndex, SegOffset).SectionName;
4453 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase
4454 // entry to get the address.
4455 uint64_t BindRebaseSegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
4456 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
4457 return SI.SegmentStartAddress + OffsetInSeg;
4460 iterator_range<bind_iterator>
4461 MachOObjectFile::bindTable(Error &Err, MachOObjectFile *O,
4462 ArrayRef<uint8_t> Opcodes, bool is64,
4463 MachOBindEntry::Kind BKind) {
4464 if (O->BindRebaseSectionTable == nullptr)
4465 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O);
4466 MachOBindEntry Start(&Err, O, Opcodes, is64, BKind);
4467 Start.moveToFirst();
4469 MachOBindEntry Finish(&Err, O, Opcodes, is64, BKind);
4470 Finish.moveToEnd();
4472 return make_range(bind_iterator(Start), bind_iterator(Finish));
4475 iterator_range<bind_iterator> MachOObjectFile::bindTable(Error &Err) {
4476 return bindTable(Err, this, getDyldInfoBindOpcodes(), is64Bit(),
4477 MachOBindEntry::Kind::Regular);
4480 iterator_range<bind_iterator> MachOObjectFile::lazyBindTable(Error &Err) {
4481 return bindTable(Err, this, getDyldInfoLazyBindOpcodes(), is64Bit(),
4482 MachOBindEntry::Kind::Lazy);
4485 iterator_range<bind_iterator> MachOObjectFile::weakBindTable(Error &Err) {
4486 return bindTable(Err, this, getDyldInfoWeakBindOpcodes(), is64Bit(),
4487 MachOBindEntry::Kind::Weak);
4490 iterator_range<fixup_iterator> MachOObjectFile::fixupTable(Error &Err) {
4491 if (BindRebaseSectionTable == nullptr)
4492 BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(this);
4494 MachOChainedFixupEntry Start(&Err, this, true);
4495 Start.moveToFirst();
4497 MachOChainedFixupEntry Finish(&Err, this, false);
4498 Finish.moveToEnd();
4500 return make_range(fixup_iterator(Start), fixup_iterator(Finish));
4503 MachOObjectFile::load_command_iterator
4504 MachOObjectFile::begin_load_commands() const {
4505 return LoadCommands.begin();
4508 MachOObjectFile::load_command_iterator
4509 MachOObjectFile::end_load_commands() const {
4510 return LoadCommands.end();
4513 iterator_range<MachOObjectFile::load_command_iterator>
4514 MachOObjectFile::load_commands() const {
4515 return make_range(begin_load_commands(), end_load_commands());
4518 StringRef
4519 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const {
4520 ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec);
4521 return parseSegmentOrSectionName(Raw.data());
4524 ArrayRef<char>
4525 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const {
4526 assert(Sec.d.a < Sections.size() && "Should have detected this earlier");
4527 const section_base *Base =
4528 reinterpret_cast<const section_base *>(Sections[Sec.d.a]);
4529 return ArrayRef(Base->sectname);
4532 ArrayRef<char>
4533 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const {
4534 assert(Sec.d.a < Sections.size() && "Should have detected this earlier");
4535 const section_base *Base =
4536 reinterpret_cast<const section_base *>(Sections[Sec.d.a]);
4537 return ArrayRef(Base->segname);
4540 bool
4541 MachOObjectFile::isRelocationScattered(const MachO::any_relocation_info &RE)
4542 const {
4543 if (getCPUType(*this) == MachO::CPU_TYPE_X86_64)
4544 return false;
4545 return getPlainRelocationAddress(RE) & MachO::R_SCATTERED;
4548 unsigned MachOObjectFile::getPlainRelocationSymbolNum(
4549 const MachO::any_relocation_info &RE) const {
4550 if (isLittleEndian())
4551 return RE.r_word1 & 0xffffff;
4552 return RE.r_word1 >> 8;
4555 bool MachOObjectFile::getPlainRelocationExternal(
4556 const MachO::any_relocation_info &RE) const {
4557 if (isLittleEndian())
4558 return (RE.r_word1 >> 27) & 1;
4559 return (RE.r_word1 >> 4) & 1;
4562 bool MachOObjectFile::getScatteredRelocationScattered(
4563 const MachO::any_relocation_info &RE) const {
4564 return RE.r_word0 >> 31;
4567 uint32_t MachOObjectFile::getScatteredRelocationValue(
4568 const MachO::any_relocation_info &RE) const {
4569 return RE.r_word1;
4572 uint32_t MachOObjectFile::getScatteredRelocationType(
4573 const MachO::any_relocation_info &RE) const {
4574 return (RE.r_word0 >> 24) & 0xf;
4577 unsigned MachOObjectFile::getAnyRelocationAddress(
4578 const MachO::any_relocation_info &RE) const {
4579 if (isRelocationScattered(RE))
4580 return getScatteredRelocationAddress(RE);
4581 return getPlainRelocationAddress(RE);
4584 unsigned MachOObjectFile::getAnyRelocationPCRel(
4585 const MachO::any_relocation_info &RE) const {
4586 if (isRelocationScattered(RE))
4587 return getScatteredRelocationPCRel(RE);
4588 return getPlainRelocationPCRel(*this, RE);
4591 unsigned MachOObjectFile::getAnyRelocationLength(
4592 const MachO::any_relocation_info &RE) const {
4593 if (isRelocationScattered(RE))
4594 return getScatteredRelocationLength(RE);
4595 return getPlainRelocationLength(*this, RE);
4598 unsigned
4599 MachOObjectFile::getAnyRelocationType(
4600 const MachO::any_relocation_info &RE) const {
4601 if (isRelocationScattered(RE))
4602 return getScatteredRelocationType(RE);
4603 return getPlainRelocationType(*this, RE);
4606 SectionRef
4607 MachOObjectFile::getAnyRelocationSection(
4608 const MachO::any_relocation_info &RE) const {
4609 if (isRelocationScattered(RE) || getPlainRelocationExternal(RE))
4610 return *section_end();
4611 unsigned SecNum = getPlainRelocationSymbolNum(RE);
4612 if (SecNum == MachO::R_ABS || SecNum > Sections.size())
4613 return *section_end();
4614 DataRefImpl DRI;
4615 DRI.d.a = SecNum - 1;
4616 return SectionRef(DRI, this);
4619 MachO::section MachOObjectFile::getSection(DataRefImpl DRI) const {
4620 assert(DRI.d.a < Sections.size() && "Should have detected this earlier");
4621 return getStruct<MachO::section>(*this, Sections[DRI.d.a]);
4624 MachO::section_64 MachOObjectFile::getSection64(DataRefImpl DRI) const {
4625 assert(DRI.d.a < Sections.size() && "Should have detected this earlier");
4626 return getStruct<MachO::section_64>(*this, Sections[DRI.d.a]);
4629 MachO::section MachOObjectFile::getSection(const LoadCommandInfo &L,
4630 unsigned Index) const {
4631 const char *Sec = getSectionPtr(*this, L, Index);
4632 return getStruct<MachO::section>(*this, Sec);
4635 MachO::section_64 MachOObjectFile::getSection64(const LoadCommandInfo &L,
4636 unsigned Index) const {
4637 const char *Sec = getSectionPtr(*this, L, Index);
4638 return getStruct<MachO::section_64>(*this, Sec);
4641 MachO::nlist
4642 MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI) const {
4643 const char *P = reinterpret_cast<const char *>(DRI.p);
4644 return getStruct<MachO::nlist>(*this, P);
4647 MachO::nlist_64
4648 MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI) const {
4649 const char *P = reinterpret_cast<const char *>(DRI.p);
4650 return getStruct<MachO::nlist_64>(*this, P);
4653 MachO::linkedit_data_command
4654 MachOObjectFile::getLinkeditDataLoadCommand(const LoadCommandInfo &L) const {
4655 return getStruct<MachO::linkedit_data_command>(*this, L.Ptr);
4658 MachO::segment_command
4659 MachOObjectFile::getSegmentLoadCommand(const LoadCommandInfo &L) const {
4660 return getStruct<MachO::segment_command>(*this, L.Ptr);
4663 MachO::segment_command_64
4664 MachOObjectFile::getSegment64LoadCommand(const LoadCommandInfo &L) const {
4665 return getStruct<MachO::segment_command_64>(*this, L.Ptr);
4668 MachO::linker_option_command
4669 MachOObjectFile::getLinkerOptionLoadCommand(const LoadCommandInfo &L) const {
4670 return getStruct<MachO::linker_option_command>(*this, L.Ptr);
4673 MachO::version_min_command
4674 MachOObjectFile::getVersionMinLoadCommand(const LoadCommandInfo &L) const {
4675 return getStruct<MachO::version_min_command>(*this, L.Ptr);
4678 MachO::note_command
4679 MachOObjectFile::getNoteLoadCommand(const LoadCommandInfo &L) const {
4680 return getStruct<MachO::note_command>(*this, L.Ptr);
4683 MachO::build_version_command
4684 MachOObjectFile::getBuildVersionLoadCommand(const LoadCommandInfo &L) const {
4685 return getStruct<MachO::build_version_command>(*this, L.Ptr);
4688 MachO::build_tool_version
4689 MachOObjectFile::getBuildToolVersion(unsigned index) const {
4690 return getStruct<MachO::build_tool_version>(*this, BuildTools[index]);
4693 MachO::dylib_command
4694 MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const {
4695 return getStruct<MachO::dylib_command>(*this, L.Ptr);
4698 MachO::dyld_info_command
4699 MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const {
4700 return getStruct<MachO::dyld_info_command>(*this, L.Ptr);
4703 MachO::dylinker_command
4704 MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const {
4705 return getStruct<MachO::dylinker_command>(*this, L.Ptr);
4708 MachO::uuid_command
4709 MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const {
4710 return getStruct<MachO::uuid_command>(*this, L.Ptr);
4713 MachO::rpath_command
4714 MachOObjectFile::getRpathCommand(const LoadCommandInfo &L) const {
4715 return getStruct<MachO::rpath_command>(*this, L.Ptr);
4718 MachO::source_version_command
4719 MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const {
4720 return getStruct<MachO::source_version_command>(*this, L.Ptr);
4723 MachO::entry_point_command
4724 MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const {
4725 return getStruct<MachO::entry_point_command>(*this, L.Ptr);
4728 MachO::encryption_info_command
4729 MachOObjectFile::getEncryptionInfoCommand(const LoadCommandInfo &L) const {
4730 return getStruct<MachO::encryption_info_command>(*this, L.Ptr);
4733 MachO::encryption_info_command_64
4734 MachOObjectFile::getEncryptionInfoCommand64(const LoadCommandInfo &L) const {
4735 return getStruct<MachO::encryption_info_command_64>(*this, L.Ptr);
4738 MachO::sub_framework_command
4739 MachOObjectFile::getSubFrameworkCommand(const LoadCommandInfo &L) const {
4740 return getStruct<MachO::sub_framework_command>(*this, L.Ptr);
4743 MachO::sub_umbrella_command
4744 MachOObjectFile::getSubUmbrellaCommand(const LoadCommandInfo &L) const {
4745 return getStruct<MachO::sub_umbrella_command>(*this, L.Ptr);
4748 MachO::sub_library_command
4749 MachOObjectFile::getSubLibraryCommand(const LoadCommandInfo &L) const {
4750 return getStruct<MachO::sub_library_command>(*this, L.Ptr);
4753 MachO::sub_client_command
4754 MachOObjectFile::getSubClientCommand(const LoadCommandInfo &L) const {
4755 return getStruct<MachO::sub_client_command>(*this, L.Ptr);
4758 MachO::routines_command
4759 MachOObjectFile::getRoutinesCommand(const LoadCommandInfo &L) const {
4760 return getStruct<MachO::routines_command>(*this, L.Ptr);
4763 MachO::routines_command_64
4764 MachOObjectFile::getRoutinesCommand64(const LoadCommandInfo &L) const {
4765 return getStruct<MachO::routines_command_64>(*this, L.Ptr);
4768 MachO::thread_command
4769 MachOObjectFile::getThreadCommand(const LoadCommandInfo &L) const {
4770 return getStruct<MachO::thread_command>(*this, L.Ptr);
4773 MachO::fileset_entry_command
4774 MachOObjectFile::getFilesetEntryLoadCommand(const LoadCommandInfo &L) const {
4775 return getStruct<MachO::fileset_entry_command>(*this, L.Ptr);
4778 MachO::any_relocation_info
4779 MachOObjectFile::getRelocation(DataRefImpl Rel) const {
4780 uint32_t Offset;
4781 if (getHeader().filetype == MachO::MH_OBJECT) {
4782 DataRefImpl Sec;
4783 Sec.d.a = Rel.d.a;
4784 if (is64Bit()) {
4785 MachO::section_64 Sect = getSection64(Sec);
4786 Offset = Sect.reloff;
4787 } else {
4788 MachO::section Sect = getSection(Sec);
4789 Offset = Sect.reloff;
4791 } else {
4792 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand();
4793 if (Rel.d.a == 0)
4794 Offset = DysymtabLoadCmd.extreloff; // Offset to the external relocations
4795 else
4796 Offset = DysymtabLoadCmd.locreloff; // Offset to the local relocations
4799 auto P = reinterpret_cast<const MachO::any_relocation_info *>(
4800 getPtr(*this, Offset)) + Rel.d.b;
4801 return getStruct<MachO::any_relocation_info>(
4802 *this, reinterpret_cast<const char *>(P));
4805 MachO::data_in_code_entry
4806 MachOObjectFile::getDice(DataRefImpl Rel) const {
4807 const char *P = reinterpret_cast<const char *>(Rel.p);
4808 return getStruct<MachO::data_in_code_entry>(*this, P);
4811 const MachO::mach_header &MachOObjectFile::getHeader() const {
4812 return Header;
4815 const MachO::mach_header_64 &MachOObjectFile::getHeader64() const {
4816 assert(is64Bit());
4817 return Header64;
4820 uint32_t MachOObjectFile::getIndirectSymbolTableEntry(
4821 const MachO::dysymtab_command &DLC,
4822 unsigned Index) const {
4823 uint64_t Offset = DLC.indirectsymoff + Index * sizeof(uint32_t);
4824 return getStruct<uint32_t>(*this, getPtr(*this, Offset));
4827 MachO::data_in_code_entry
4828 MachOObjectFile::getDataInCodeTableEntry(uint32_t DataOffset,
4829 unsigned Index) const {
4830 uint64_t Offset = DataOffset + Index * sizeof(MachO::data_in_code_entry);
4831 return getStruct<MachO::data_in_code_entry>(*this, getPtr(*this, Offset));
4834 MachO::symtab_command MachOObjectFile::getSymtabLoadCommand() const {
4835 if (SymtabLoadCmd)
4836 return getStruct<MachO::symtab_command>(*this, SymtabLoadCmd);
4838 // If there is no SymtabLoadCmd return a load command with zero'ed fields.
4839 MachO::symtab_command Cmd;
4840 Cmd.cmd = MachO::LC_SYMTAB;
4841 Cmd.cmdsize = sizeof(MachO::symtab_command);
4842 Cmd.symoff = 0;
4843 Cmd.nsyms = 0;
4844 Cmd.stroff = 0;
4845 Cmd.strsize = 0;
4846 return Cmd;
4849 MachO::dysymtab_command MachOObjectFile::getDysymtabLoadCommand() const {
4850 if (DysymtabLoadCmd)
4851 return getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd);
4853 // If there is no DysymtabLoadCmd return a load command with zero'ed fields.
4854 MachO::dysymtab_command Cmd;
4855 Cmd.cmd = MachO::LC_DYSYMTAB;
4856 Cmd.cmdsize = sizeof(MachO::dysymtab_command);
4857 Cmd.ilocalsym = 0;
4858 Cmd.nlocalsym = 0;
4859 Cmd.iextdefsym = 0;
4860 Cmd.nextdefsym = 0;
4861 Cmd.iundefsym = 0;
4862 Cmd.nundefsym = 0;
4863 Cmd.tocoff = 0;
4864 Cmd.ntoc = 0;
4865 Cmd.modtaboff = 0;
4866 Cmd.nmodtab = 0;
4867 Cmd.extrefsymoff = 0;
4868 Cmd.nextrefsyms = 0;
4869 Cmd.indirectsymoff = 0;
4870 Cmd.nindirectsyms = 0;
4871 Cmd.extreloff = 0;
4872 Cmd.nextrel = 0;
4873 Cmd.locreloff = 0;
4874 Cmd.nlocrel = 0;
4875 return Cmd;
4878 MachO::linkedit_data_command
4879 MachOObjectFile::getDataInCodeLoadCommand() const {
4880 if (DataInCodeLoadCmd)
4881 return getStruct<MachO::linkedit_data_command>(*this, DataInCodeLoadCmd);
4883 // If there is no DataInCodeLoadCmd return a load command with zero'ed fields.
4884 MachO::linkedit_data_command Cmd;
4885 Cmd.cmd = MachO::LC_DATA_IN_CODE;
4886 Cmd.cmdsize = sizeof(MachO::linkedit_data_command);
4887 Cmd.dataoff = 0;
4888 Cmd.datasize = 0;
4889 return Cmd;
4892 MachO::linkedit_data_command
4893 MachOObjectFile::getLinkOptHintsLoadCommand() const {
4894 if (LinkOptHintsLoadCmd)
4895 return getStruct<MachO::linkedit_data_command>(*this, LinkOptHintsLoadCmd);
4897 // If there is no LinkOptHintsLoadCmd return a load command with zero'ed
4898 // fields.
4899 MachO::linkedit_data_command Cmd;
4900 Cmd.cmd = MachO::LC_LINKER_OPTIMIZATION_HINT;
4901 Cmd.cmdsize = sizeof(MachO::linkedit_data_command);
4902 Cmd.dataoff = 0;
4903 Cmd.datasize = 0;
4904 return Cmd;
4907 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const {
4908 if (!DyldInfoLoadCmd)
4909 return {};
4911 auto DyldInfoOrErr =
4912 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd);
4913 if (!DyldInfoOrErr)
4914 return {};
4915 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get();
4916 const uint8_t *Ptr =
4917 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.rebase_off));
4918 return ArrayRef(Ptr, DyldInfo.rebase_size);
4921 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const {
4922 if (!DyldInfoLoadCmd)
4923 return {};
4925 auto DyldInfoOrErr =
4926 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd);
4927 if (!DyldInfoOrErr)
4928 return {};
4929 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get();
4930 const uint8_t *Ptr =
4931 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.bind_off));
4932 return ArrayRef(Ptr, DyldInfo.bind_size);
4935 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const {
4936 if (!DyldInfoLoadCmd)
4937 return {};
4939 auto DyldInfoOrErr =
4940 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd);
4941 if (!DyldInfoOrErr)
4942 return {};
4943 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get();
4944 const uint8_t *Ptr =
4945 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.weak_bind_off));
4946 return ArrayRef(Ptr, DyldInfo.weak_bind_size);
4949 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const {
4950 if (!DyldInfoLoadCmd)
4951 return {};
4953 auto DyldInfoOrErr =
4954 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd);
4955 if (!DyldInfoOrErr)
4956 return {};
4957 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get();
4958 const uint8_t *Ptr =
4959 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.lazy_bind_off));
4960 return ArrayRef(Ptr, DyldInfo.lazy_bind_size);
4963 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const {
4964 if (!DyldInfoLoadCmd)
4965 return {};
4967 auto DyldInfoOrErr =
4968 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd);
4969 if (!DyldInfoOrErr)
4970 return {};
4971 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get();
4972 const uint8_t *Ptr =
4973 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.export_off));
4974 return ArrayRef(Ptr, DyldInfo.export_size);
4977 Expected<std::optional<MachO::linkedit_data_command>>
4978 MachOObjectFile::getChainedFixupsLoadCommand() const {
4979 // Load the dyld chained fixups load command.
4980 if (!DyldChainedFixupsLoadCmd)
4981 return std::nullopt;
4982 auto DyldChainedFixupsOrErr = getStructOrErr<MachO::linkedit_data_command>(
4983 *this, DyldChainedFixupsLoadCmd);
4984 if (!DyldChainedFixupsOrErr)
4985 return DyldChainedFixupsOrErr.takeError();
4986 const MachO::linkedit_data_command &DyldChainedFixups =
4987 *DyldChainedFixupsOrErr;
4989 // If the load command is present but the data offset has been zeroed out,
4990 // as is the case for dylib stubs, return std::nullopt (no error).
4991 if (!DyldChainedFixups.dataoff)
4992 return std::nullopt;
4993 return DyldChainedFixups;
4996 Expected<std::optional<MachO::dyld_chained_fixups_header>>
4997 MachOObjectFile::getChainedFixupsHeader() const {
4998 auto CFOrErr = getChainedFixupsLoadCommand();
4999 if (!CFOrErr)
5000 return CFOrErr.takeError();
5001 if (!CFOrErr->has_value())
5002 return std::nullopt;
5004 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr;
5006 uint64_t CFHeaderOffset = DyldChainedFixups.dataoff;
5007 uint64_t CFSize = DyldChainedFixups.datasize;
5009 // Load the dyld chained fixups header.
5010 const char *CFHeaderPtr = getPtr(*this, CFHeaderOffset);
5011 auto CFHeaderOrErr =
5012 getStructOrErr<MachO::dyld_chained_fixups_header>(*this, CFHeaderPtr);
5013 if (!CFHeaderOrErr)
5014 return CFHeaderOrErr.takeError();
5015 MachO::dyld_chained_fixups_header CFHeader = CFHeaderOrErr.get();
5017 // Reject unknown chained fixup formats.
5018 if (CFHeader.fixups_version != 0)
5019 return malformedError(Twine("bad chained fixups: unknown version: ") +
5020 Twine(CFHeader.fixups_version));
5021 if (CFHeader.imports_format < 1 || CFHeader.imports_format > 3)
5022 return malformedError(
5023 Twine("bad chained fixups: unknown imports format: ") +
5024 Twine(CFHeader.imports_format));
5026 // Validate the image format.
5028 // Load the image starts.
5029 uint64_t CFImageStartsOffset = (CFHeaderOffset + CFHeader.starts_offset);
5030 if (CFHeader.starts_offset < sizeof(MachO::dyld_chained_fixups_header)) {
5031 return malformedError(Twine("bad chained fixups: image starts offset ") +
5032 Twine(CFHeader.starts_offset) +
5033 " overlaps with chained fixups header");
5035 uint32_t EndOffset = CFHeaderOffset + CFSize;
5036 if (CFImageStartsOffset + sizeof(MachO::dyld_chained_starts_in_image) >
5037 EndOffset) {
5038 return malformedError(Twine("bad chained fixups: image starts end ") +
5039 Twine(CFImageStartsOffset +
5040 sizeof(MachO::dyld_chained_starts_in_image)) +
5041 " extends past end " + Twine(EndOffset));
5044 return CFHeader;
5047 Expected<std::pair<size_t, std::vector<ChainedFixupsSegment>>>
5048 MachOObjectFile::getChainedFixupsSegments() const {
5049 auto CFOrErr = getChainedFixupsLoadCommand();
5050 if (!CFOrErr)
5051 return CFOrErr.takeError();
5053 std::vector<ChainedFixupsSegment> Segments;
5054 if (!CFOrErr->has_value())
5055 return std::make_pair(0, Segments);
5057 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr;
5059 auto HeaderOrErr = getChainedFixupsHeader();
5060 if (!HeaderOrErr)
5061 return HeaderOrErr.takeError();
5062 if (!HeaderOrErr->has_value())
5063 return std::make_pair(0, Segments);
5064 const MachO::dyld_chained_fixups_header &Header = **HeaderOrErr;
5066 const char *Contents = getPtr(*this, DyldChainedFixups.dataoff);
5068 auto ImageStartsOrErr = getStructOrErr<MachO::dyld_chained_starts_in_image>(
5069 *this, Contents + Header.starts_offset);
5070 if (!ImageStartsOrErr)
5071 return ImageStartsOrErr.takeError();
5072 const MachO::dyld_chained_starts_in_image &ImageStarts = *ImageStartsOrErr;
5074 const char *SegOffsPtr =
5075 Contents + Header.starts_offset +
5076 offsetof(MachO::dyld_chained_starts_in_image, seg_info_offset);
5077 const char *SegOffsEnd =
5078 SegOffsPtr + ImageStarts.seg_count * sizeof(uint32_t);
5079 if (SegOffsEnd > Contents + DyldChainedFixups.datasize)
5080 return malformedError(
5081 "bad chained fixups: seg_info_offset extends past end");
5083 const char *LastSegEnd = nullptr;
5084 for (size_t I = 0, N = ImageStarts.seg_count; I < N; ++I) {
5085 auto OffOrErr =
5086 getStructOrErr<uint32_t>(*this, SegOffsPtr + I * sizeof(uint32_t));
5087 if (!OffOrErr)
5088 return OffOrErr.takeError();
5089 // seg_info_offset == 0 means there is no associated starts_in_segment
5090 // entry.
5091 if (!*OffOrErr)
5092 continue;
5094 auto Fail = [&](Twine Message) {
5095 return malformedError("bad chained fixups: segment info" + Twine(I) +
5096 " at offset " + Twine(*OffOrErr) + Message);
5099 const char *SegPtr = Contents + Header.starts_offset + *OffOrErr;
5100 if (LastSegEnd && SegPtr < LastSegEnd)
5101 return Fail(" overlaps with previous segment info");
5103 auto SegOrErr =
5104 getStructOrErr<MachO::dyld_chained_starts_in_segment>(*this, SegPtr);
5105 if (!SegOrErr)
5106 return SegOrErr.takeError();
5107 const MachO::dyld_chained_starts_in_segment &Seg = *SegOrErr;
5109 LastSegEnd = SegPtr + Seg.size;
5110 if (Seg.pointer_format < 1 || Seg.pointer_format > 12)
5111 return Fail(" has unknown pointer format: " + Twine(Seg.pointer_format));
5113 const char *PageStart =
5114 SegPtr + offsetof(MachO::dyld_chained_starts_in_segment, page_start);
5115 const char *PageEnd = PageStart + Seg.page_count * sizeof(uint16_t);
5116 if (PageEnd > SegPtr + Seg.size)
5117 return Fail(" : page_starts extend past seg_info size");
5119 // FIXME: This does not account for multiple offsets on a single page
5120 // (DYLD_CHAINED_PTR_START_MULTI; 32-bit only).
5121 std::vector<uint16_t> PageStarts;
5122 for (size_t PageIdx = 0; PageIdx < Seg.page_count; ++PageIdx) {
5123 uint16_t Start;
5124 memcpy(&Start, PageStart + PageIdx * sizeof(uint16_t), sizeof(uint16_t));
5125 if (isLittleEndian() != sys::IsLittleEndianHost)
5126 sys::swapByteOrder(Start);
5127 PageStarts.push_back(Start);
5130 Segments.emplace_back(I, *OffOrErr, Seg, std::move(PageStarts));
5133 return std::make_pair(ImageStarts.seg_count, Segments);
5136 // The special library ordinals have a negative value, but they are encoded in
5137 // an unsigned bitfield, so we need to sign extend the value.
5138 template <typename T> static int getEncodedOrdinal(T Value) {
5139 if (Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE) ||
5140 Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) ||
5141 Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_WEAK_LOOKUP))
5142 return SignExtend32<sizeof(T) * CHAR_BIT>(Value);
5143 return Value;
5146 template <typename T, unsigned N>
5147 static std::array<T, N> getArray(const MachOObjectFile &O, const void *Ptr) {
5148 std::array<T, N> RawValue;
5149 memcpy(RawValue.data(), Ptr, N * sizeof(T));
5150 if (O.isLittleEndian() != sys::IsLittleEndianHost)
5151 for (auto &Element : RawValue)
5152 sys::swapByteOrder(Element);
5153 return RawValue;
5156 Expected<std::vector<ChainedFixupTarget>>
5157 MachOObjectFile::getDyldChainedFixupTargets() const {
5158 auto CFOrErr = getChainedFixupsLoadCommand();
5159 if (!CFOrErr)
5160 return CFOrErr.takeError();
5162 std::vector<ChainedFixupTarget> Targets;
5163 if (!CFOrErr->has_value())
5164 return Targets;
5166 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr;
5168 auto CFHeaderOrErr = getChainedFixupsHeader();
5169 if (!CFHeaderOrErr)
5170 return CFHeaderOrErr.takeError();
5171 if (!(*CFHeaderOrErr))
5172 return Targets;
5173 const MachO::dyld_chained_fixups_header &Header = **CFHeaderOrErr;
5175 size_t ImportSize = 0;
5176 if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT)
5177 ImportSize = sizeof(MachO::dyld_chained_import);
5178 else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND)
5179 ImportSize = sizeof(MachO::dyld_chained_import_addend);
5180 else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND64)
5181 ImportSize = sizeof(MachO::dyld_chained_import_addend64);
5182 else
5183 return malformedError("bad chained fixups: unknown imports format: " +
5184 Twine(Header.imports_format));
5186 const char *Contents = getPtr(*this, DyldChainedFixups.dataoff);
5187 const char *Imports = Contents + Header.imports_offset;
5188 size_t ImportsEndOffset =
5189 Header.imports_offset + ImportSize * Header.imports_count;
5190 const char *ImportsEnd = Contents + ImportsEndOffset;
5191 const char *Symbols = Contents + Header.symbols_offset;
5192 const char *SymbolsEnd = Contents + DyldChainedFixups.datasize;
5194 if (ImportsEnd > Symbols)
5195 return malformedError("bad chained fixups: imports end " +
5196 Twine(ImportsEndOffset) + " overlaps with symbols");
5198 // We use bit manipulation to extract data from the bitfields. This is correct
5199 // for both LE and BE hosts, but we assume that the object is little-endian.
5200 if (!isLittleEndian())
5201 return createError("parsing big-endian chained fixups is not implemented");
5202 for (const char *ImportPtr = Imports; ImportPtr < ImportsEnd;
5203 ImportPtr += ImportSize) {
5204 int LibOrdinal;
5205 bool WeakImport;
5206 uint32_t NameOffset;
5207 uint64_t Addend;
5208 if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT) {
5209 static_assert(sizeof(uint32_t) == sizeof(MachO::dyld_chained_import));
5210 auto RawValue = getArray<uint32_t, 1>(*this, ImportPtr);
5212 LibOrdinal = getEncodedOrdinal<uint8_t>(RawValue[0] & 0xFF);
5213 WeakImport = (RawValue[0] >> 8) & 1;
5214 NameOffset = RawValue[0] >> 9;
5215 Addend = 0;
5216 } else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND) {
5217 static_assert(sizeof(uint64_t) ==
5218 sizeof(MachO::dyld_chained_import_addend));
5219 auto RawValue = getArray<uint32_t, 2>(*this, ImportPtr);
5221 LibOrdinal = getEncodedOrdinal<uint8_t>(RawValue[0] & 0xFF);
5222 WeakImport = (RawValue[0] >> 8) & 1;
5223 NameOffset = RawValue[0] >> 9;
5224 Addend = bit_cast<int32_t>(RawValue[1]);
5225 } else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND64) {
5226 static_assert(2 * sizeof(uint64_t) ==
5227 sizeof(MachO::dyld_chained_import_addend64));
5228 auto RawValue = getArray<uint64_t, 2>(*this, ImportPtr);
5230 LibOrdinal = getEncodedOrdinal<uint16_t>(RawValue[0] & 0xFFFF);
5231 NameOffset = (RawValue[0] >> 16) & 1;
5232 WeakImport = RawValue[0] >> 17;
5233 Addend = RawValue[1];
5234 } else {
5235 llvm_unreachable("Import format should have been checked");
5238 const char *Str = Symbols + NameOffset;
5239 if (Str >= SymbolsEnd)
5240 return malformedError("bad chained fixups: symbol offset " +
5241 Twine(NameOffset) + " extends past end " +
5242 Twine(DyldChainedFixups.datasize));
5243 Targets.emplace_back(LibOrdinal, NameOffset, Str, Addend, WeakImport);
5246 return std::move(Targets);
5249 ArrayRef<uint8_t> MachOObjectFile::getDyldExportsTrie() const {
5250 if (!DyldExportsTrieLoadCmd)
5251 return {};
5253 auto DyldExportsTrieOrError = getStructOrErr<MachO::linkedit_data_command>(
5254 *this, DyldExportsTrieLoadCmd);
5255 if (!DyldExportsTrieOrError)
5256 return {};
5257 MachO::linkedit_data_command DyldExportsTrie = DyldExportsTrieOrError.get();
5258 const uint8_t *Ptr =
5259 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldExportsTrie.dataoff));
5260 return ArrayRef(Ptr, DyldExportsTrie.datasize);
5263 SmallVector<uint64_t> MachOObjectFile::getFunctionStarts() const {
5264 if (!FuncStartsLoadCmd)
5265 return {};
5267 auto InfoOrErr =
5268 getStructOrErr<MachO::linkedit_data_command>(*this, FuncStartsLoadCmd);
5269 if (!InfoOrErr)
5270 return {};
5272 MachO::linkedit_data_command Info = InfoOrErr.get();
5273 SmallVector<uint64_t, 8> FunctionStarts;
5274 this->ReadULEB128s(Info.dataoff, FunctionStarts);
5275 return std::move(FunctionStarts);
5278 ArrayRef<uint8_t> MachOObjectFile::getUuid() const {
5279 if (!UuidLoadCmd)
5280 return {};
5281 // Returning a pointer is fine as uuid doesn't need endian swapping.
5282 const char *Ptr = UuidLoadCmd + offsetof(MachO::uuid_command, uuid);
5283 return ArrayRef(reinterpret_cast<const uint8_t *>(Ptr), 16);
5286 StringRef MachOObjectFile::getStringTableData() const {
5287 MachO::symtab_command S = getSymtabLoadCommand();
5288 return getData().substr(S.stroff, S.strsize);
5291 bool MachOObjectFile::is64Bit() const {
5292 return getType() == getMachOType(false, true) ||
5293 getType() == getMachOType(true, true);
5296 void MachOObjectFile::ReadULEB128s(uint64_t Index,
5297 SmallVectorImpl<uint64_t> &Out) const {
5298 DataExtractor extractor(ObjectFile::getData(), true, 0);
5300 uint64_t offset = Index;
5301 uint64_t data = 0;
5302 while (uint64_t delta = extractor.getULEB128(&offset)) {
5303 data += delta;
5304 Out.push_back(data);
5308 bool MachOObjectFile::isRelocatableObject() const {
5309 return getHeader().filetype == MachO::MH_OBJECT;
5312 /// Create a MachOObjectFile instance from a given buffer.
5314 /// \param Buffer Memory buffer containing the MachO binary data.
5315 /// \param UniversalCputype CPU type when the MachO part of a universal binary.
5316 /// \param UniversalIndex Index of the MachO within a universal binary.
5317 /// \param MachOFilesetEntryOffset Offset of the MachO entry in a fileset MachO.
5318 /// \returns A std::unique_ptr to a MachOObjectFile instance on success.
5319 Expected<std::unique_ptr<MachOObjectFile>> ObjectFile::createMachOObjectFile(
5320 MemoryBufferRef Buffer, uint32_t UniversalCputype, uint32_t UniversalIndex,
5321 size_t MachOFilesetEntryOffset) {
5322 StringRef Magic = Buffer.getBuffer().slice(0, 4);
5323 if (Magic == "\xFE\xED\xFA\xCE")
5324 return MachOObjectFile::create(Buffer, false, false, UniversalCputype,
5325 UniversalIndex, MachOFilesetEntryOffset);
5326 if (Magic == "\xCE\xFA\xED\xFE")
5327 return MachOObjectFile::create(Buffer, true, false, UniversalCputype,
5328 UniversalIndex, MachOFilesetEntryOffset);
5329 if (Magic == "\xFE\xED\xFA\xCF")
5330 return MachOObjectFile::create(Buffer, false, true, UniversalCputype,
5331 UniversalIndex, MachOFilesetEntryOffset);
5332 if (Magic == "\xCF\xFA\xED\xFE")
5333 return MachOObjectFile::create(Buffer, true, true, UniversalCputype,
5334 UniversalIndex, MachOFilesetEntryOffset);
5335 return make_error<GenericBinaryError>("Unrecognized MachO magic number",
5336 object_error::invalid_file_type);
5339 StringRef MachOObjectFile::mapDebugSectionName(StringRef Name) const {
5340 return StringSwitch<StringRef>(Name)
5341 .Case("debug_str_offs", "debug_str_offsets")
5342 .Default(Name);
5345 Expected<std::vector<std::string>>
5346 MachOObjectFile::findDsymObjectMembers(StringRef Path) {
5347 SmallString<256> BundlePath(Path);
5348 // Normalize input path. This is necessary to accept `bundle.dSYM/`.
5349 sys::path::remove_dots(BundlePath);
5350 if (!sys::fs::is_directory(BundlePath) ||
5351 sys::path::extension(BundlePath) != ".dSYM")
5352 return std::vector<std::string>();
5353 sys::path::append(BundlePath, "Contents", "Resources", "DWARF");
5354 bool IsDir;
5355 auto EC = sys::fs::is_directory(BundlePath, IsDir);
5356 if (EC == errc::no_such_file_or_directory || (!EC && !IsDir))
5357 return createStringError(
5358 EC, "%s: expected directory 'Contents/Resources/DWARF' in dSYM bundle",
5359 Path.str().c_str());
5360 if (EC)
5361 return createFileError(BundlePath, errorCodeToError(EC));
5363 std::vector<std::string> ObjectPaths;
5364 for (sys::fs::directory_iterator Dir(BundlePath, EC), DirEnd;
5365 Dir != DirEnd && !EC; Dir.increment(EC)) {
5366 StringRef ObjectPath = Dir->path();
5367 sys::fs::file_status Status;
5368 if (auto EC = sys::fs::status(ObjectPath, Status))
5369 return createFileError(ObjectPath, errorCodeToError(EC));
5370 switch (Status.type()) {
5371 case sys::fs::file_type::regular_file:
5372 case sys::fs::file_type::symlink_file:
5373 case sys::fs::file_type::type_unknown:
5374 ObjectPaths.push_back(ObjectPath.str());
5375 break;
5376 default: /*ignore*/;
5379 if (EC)
5380 return createFileError(BundlePath, errorCodeToError(EC));
5381 if (ObjectPaths.empty())
5382 return createStringError(std::error_code(),
5383 "%s: no objects found in dSYM bundle",
5384 Path.str().c_str());
5385 return ObjectPaths;
5388 llvm::binaryformat::Swift5ReflectionSectionKind
5389 MachOObjectFile::mapReflectionSectionNameToEnumValue(
5390 StringRef SectionName) const {
5391 #define HANDLE_SWIFT_SECTION(KIND, MACHO, ELF, COFF) \
5392 .Case(MACHO, llvm::binaryformat::Swift5ReflectionSectionKind::KIND)
5393 return StringSwitch<llvm::binaryformat::Swift5ReflectionSectionKind>(
5394 SectionName)
5395 #include "llvm/BinaryFormat/Swift.def"
5396 .Default(llvm::binaryformat::Swift5ReflectionSectionKind::unknown);
5397 #undef HANDLE_SWIFT_SECTION
5400 bool MachOObjectFile::isMachOPairedReloc(uint64_t RelocType, uint64_t Arch) {
5401 switch (Arch) {
5402 case Triple::x86:
5403 return RelocType == MachO::GENERIC_RELOC_SECTDIFF ||
5404 RelocType == MachO::GENERIC_RELOC_LOCAL_SECTDIFF;
5405 case Triple::x86_64:
5406 return RelocType == MachO::X86_64_RELOC_SUBTRACTOR;
5407 case Triple::arm:
5408 case Triple::thumb:
5409 return RelocType == MachO::ARM_RELOC_SECTDIFF ||
5410 RelocType == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
5411 RelocType == MachO::ARM_RELOC_HALF ||
5412 RelocType == MachO::ARM_RELOC_HALF_SECTDIFF;
5413 case Triple::aarch64:
5414 return RelocType == MachO::ARM64_RELOC_SUBTRACTOR;
5415 default:
5416 return false;