1 //===- Archive.cpp - ar File Format implementation ------------------------===//
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 ArchiveObjectFile class.
10 //
11 //===----------------------------------------------------------------------===//
12
13 #include "llvm/Object/Archive.h"
14 #include "llvm/ADT/Optional.h"
15 #include "llvm/ADT/SmallString.h"
16 #include "llvm/ADT/StringRef.h"
17 #include "llvm/ADT/Twine.h"
18 #include "llvm/Object/Binary.h"
19 #include "llvm/Object/Error.h"
20 #include "llvm/Support/Chrono.h"
21 #include "llvm/Support/Endian.h"
22 #include "llvm/Support/Error.h"
23 #include "llvm/Support/ErrorOr.h"
24 #include "llvm/Support/FileSystem.h"
25 #include "llvm/Support/MemoryBuffer.h"
26 #include "llvm/Support/Path.h"
27 #include "llvm/Support/raw_ostream.h"
28 #include <algorithm>
29 #include <cassert>
30 #include <cstddef>
31 #include <cstdint>
32 #include <cstring>
33 #include <memory>
34 #include <string>
35 #include <system_error>
36
37 using namespace llvm;
38 using namespace object;
39 using namespace llvm::support::endian;
40
41 const char Magic[] = "!<arch>\n";
42 const char ThinMagic[] = "!<thin>\n";
43
anchor()44 void Archive::anchor() {}
45
46 static Error
malformedError(Twine Msg)47 malformedError(Twine Msg) {
48 std::string StringMsg = "truncated or malformed archive (" + Msg.str() + ")";
49 return make_error<GenericBinaryError>(std::move(StringMsg),
50 object_error::parse_failed);
51 }
52
ArchiveMemberHeader(const Archive * Parent,const char * RawHeaderPtr,uint64_t Size,Error * Err)53 ArchiveMemberHeader::ArchiveMemberHeader(const Archive *Parent,
54 const char *RawHeaderPtr,
55 uint64_t Size, Error *Err)
56 : Parent(Parent),
57 ArMemHdr(reinterpret_cast<const ArMemHdrType *>(RawHeaderPtr)) {
58 if (RawHeaderPtr == nullptr)
59 return;
60 ErrorAsOutParameter ErrAsOutParam(Err);
61
62 if (Size < sizeof(ArMemHdrType)) {
63 if (Err) {
64 std::string Msg("remaining size of archive too small for next archive "
65 "member header ");
66 Expected<StringRef> NameOrErr = getName(Size);
67 if (!NameOrErr) {
68 consumeError(NameOrErr.takeError());
69 uint64_t Offset = RawHeaderPtr - Parent->getData().data();
70 *Err = malformedError(Msg + "at offset " + Twine(Offset));
71 } else
72 *Err = malformedError(Msg + "for " + NameOrErr.get());
73 }
74 return;
75 }
76 if (ArMemHdr->Terminator[0] != '`' || ArMemHdr->Terminator[1] != '\n') {
77 if (Err) {
78 std::string Buf;
79 raw_string_ostream OS(Buf);
80 OS.write_escaped(StringRef(ArMemHdr->Terminator,
81 sizeof(ArMemHdr->Terminator)));
82 OS.flush();
83 std::string Msg("terminator characters in archive member \"" + Buf +
84 "\" not the correct \"`\\n\" values for the archive "
85 "member header ");
86 Expected<StringRef> NameOrErr = getName(Size);
87 if (!NameOrErr) {
88 consumeError(NameOrErr.takeError());
89 uint64_t Offset = RawHeaderPtr - Parent->getData().data();
90 *Err = malformedError(Msg + "at offset " + Twine(Offset));
91 } else
92 *Err = malformedError(Msg + "for " + NameOrErr.get());
93 }
94 return;
95 }
96 }
97
98 // This gets the raw name from the ArMemHdr->Name field and checks that it is
99 // valid for the kind of archive. If it is not valid it returns an Error.
getRawName() const100 Expected<StringRef> ArchiveMemberHeader::getRawName() const {
101 char EndCond;
102 auto Kind = Parent->kind();
103 if (Kind == Archive::K_BSD || Kind == Archive::K_DARWIN64) {
104 if (ArMemHdr->Name[0] == ' ') {
105 uint64_t Offset = reinterpret_cast<const char *>(ArMemHdr) -
106 Parent->getData().data();
107 return malformedError("name contains a leading space for archive member "
108 "header at offset " + Twine(Offset));
109 }
110 EndCond = ' ';
111 }
112 else if (ArMemHdr->Name[0] == '/' || ArMemHdr->Name[0] == '#')
113 EndCond = ' ';
114 else
115 EndCond = '/';
116 StringRef::size_type end =
117 StringRef(ArMemHdr->Name, sizeof(ArMemHdr->Name)).find(EndCond);
118 if (end == StringRef::npos)
119 end = sizeof(ArMemHdr->Name);
120 assert(end <= sizeof(ArMemHdr->Name) && end > 0);
121 // Don't include the EndCond if there is one.
122 return StringRef(ArMemHdr->Name, end);
123 }
124
125 // This gets the name looking up long names. Size is the size of the archive
126 // member including the header, so the size of any name following the header
127 // is checked to make sure it does not overflow.
getName(uint64_t Size) const128 Expected<StringRef> ArchiveMemberHeader::getName(uint64_t Size) const {
129
130 // This can be called from the ArchiveMemberHeader constructor when the
131 // archive header is truncated to produce an error message with the name.
132 // Make sure the name field is not truncated.
133 if (Size < offsetof(ArMemHdrType, Name) + sizeof(ArMemHdr->Name)) {
134 uint64_t ArchiveOffset = reinterpret_cast<const char *>(ArMemHdr) -
135 Parent->getData().data();
136 return malformedError("archive header truncated before the name field "
137 "for archive member header at offset " +
138 Twine(ArchiveOffset));
139 }
140
141 // The raw name itself can be invalid.
142 Expected<StringRef> NameOrErr = getRawName();
143 if (!NameOrErr)
144 return NameOrErr.takeError();
145 StringRef Name = NameOrErr.get();
146
147 // Check if it's a special name.
148 if (Name[0] == '/') {
149 if (Name.size() == 1) // Linker member.
150 return Name;
151 if (Name.size() == 2 && Name[1] == '/') // String table.
152 return Name;
153 // It's a long name.
154 // Get the string table offset.
155 std::size_t StringOffset;
156 if (Name.substr(1).rtrim(' ').getAsInteger(10, StringOffset)) {
157 std::string Buf;
158 raw_string_ostream OS(Buf);
159 OS.write_escaped(Name.substr(1).rtrim(' '));
160 OS.flush();
161 uint64_t ArchiveOffset = reinterpret_cast<const char *>(ArMemHdr) -
162 Parent->getData().data();
163 return malformedError("long name offset characters after the '/' are "
164 "not all decimal numbers: '" + Buf + "' for "
165 "archive member header at offset " +
166 Twine(ArchiveOffset));
167 }
168
169 // Verify it.
170 if (StringOffset >= Parent->getStringTable().size()) {
171 uint64_t ArchiveOffset = reinterpret_cast<const char *>(ArMemHdr) -
172 Parent->getData().data();
173 return malformedError("long name offset " + Twine(StringOffset) + " past "
174 "the end of the string table for archive member "
175 "header at offset " + Twine(ArchiveOffset));
176 }
177
178 // GNU long file names end with a "/\n".
179 if (Parent->kind() == Archive::K_GNU ||
180 Parent->kind() == Archive::K_GNU64) {
181 size_t End = Parent->getStringTable().find('\n', /*From=*/StringOffset);
182 if (End == StringRef::npos || End < 1 ||
183 Parent->getStringTable()[End - 1] != '/') {
184 return malformedError("string table at long name offset " +
185 Twine(StringOffset) + "not terminated");
186 }
187 return Parent->getStringTable().slice(StringOffset, End - 1);
188 }
189 return Parent->getStringTable().begin() + StringOffset;
190 }
191
192 if (Name.startswith("#1/")) {
193 uint64_t NameLength;
194 if (Name.substr(3).rtrim(' ').getAsInteger(10, NameLength)) {
195 std::string Buf;
196 raw_string_ostream OS(Buf);
197 OS.write_escaped(Name.substr(3).rtrim(' '));
198 OS.flush();
199 uint64_t ArchiveOffset = reinterpret_cast<const char *>(ArMemHdr) -
200 Parent->getData().data();
201 return malformedError("long name length characters after the #1/ are "
202 "not all decimal numbers: '" + Buf + "' for "
203 "archive member header at offset " +
204 Twine(ArchiveOffset));
205 }
206 if (getSizeOf() + NameLength > Size) {
207 uint64_t ArchiveOffset = reinterpret_cast<const char *>(ArMemHdr) -
208 Parent->getData().data();
209 return malformedError("long name length: " + Twine(NameLength) +
210 " extends past the end of the member or archive "
211 "for archive member header at offset " +
212 Twine(ArchiveOffset));
213 }
214 return StringRef(reinterpret_cast<const char *>(ArMemHdr) + getSizeOf(),
215 NameLength).rtrim('\0');
216 }
217
218 // It is not a long name so trim the blanks at the end of the name.
219 if (Name[Name.size() - 1] != '/')
220 return Name.rtrim(' ');
221
222 // It's a simple name.
223 return Name.drop_back(1);
224 }
225
getSize() const226 Expected<uint64_t> ArchiveMemberHeader::getSize() const {
227 uint64_t Ret;
228 if (StringRef(ArMemHdr->Size,
229 sizeof(ArMemHdr->Size)).rtrim(" ").getAsInteger(10, Ret)) {
230 std::string Buf;
231 raw_string_ostream OS(Buf);
232 OS.write_escaped(StringRef(ArMemHdr->Size,
233 sizeof(ArMemHdr->Size)).rtrim(" "));
234 OS.flush();
235 uint64_t Offset = reinterpret_cast<const char *>(ArMemHdr) -
236 Parent->getData().data();
237 return malformedError("characters in size field in archive header are not "
238 "all decimal numbers: '" + Buf + "' for archive "
239 "member header at offset " + Twine(Offset));
240 }
241 return Ret;
242 }
243
getAccessMode() const244 Expected<sys::fs::perms> ArchiveMemberHeader::getAccessMode() const {
245 unsigned Ret;
246 if (StringRef(ArMemHdr->AccessMode,
247 sizeof(ArMemHdr->AccessMode)).rtrim(' ').getAsInteger(8, Ret)) {
248 std::string Buf;
249 raw_string_ostream OS(Buf);
250 OS.write_escaped(StringRef(ArMemHdr->AccessMode,
251 sizeof(ArMemHdr->AccessMode)).rtrim(" "));
252 OS.flush();
253 uint64_t Offset = reinterpret_cast<const char *>(ArMemHdr) -
254 Parent->getData().data();
255 return malformedError("characters in AccessMode field in archive header "
256 "are not all decimal numbers: '" + Buf + "' for the "
257 "archive member header at offset " + Twine(Offset));
258 }
259 return static_cast<sys::fs::perms>(Ret);
260 }
261
262 Expected<sys::TimePoint<std::chrono::seconds>>
getLastModified() const263 ArchiveMemberHeader::getLastModified() const {
264 unsigned Seconds;
265 if (StringRef(ArMemHdr->LastModified,
266 sizeof(ArMemHdr->LastModified)).rtrim(' ')
267 .getAsInteger(10, Seconds)) {
268 std::string Buf;
269 raw_string_ostream OS(Buf);
270 OS.write_escaped(StringRef(ArMemHdr->LastModified,
271 sizeof(ArMemHdr->LastModified)).rtrim(" "));
272 OS.flush();
273 uint64_t Offset = reinterpret_cast<const char *>(ArMemHdr) -
274 Parent->getData().data();
275 return malformedError("characters in LastModified field in archive header "
276 "are not all decimal numbers: '" + Buf + "' for the "
277 "archive member header at offset " + Twine(Offset));
278 }
279
280 return sys::toTimePoint(Seconds);
281 }
282
getUID() const283 Expected<unsigned> ArchiveMemberHeader::getUID() const {
284 unsigned Ret;
285 StringRef User = StringRef(ArMemHdr->UID, sizeof(ArMemHdr->UID)).rtrim(' ');
286 if (User.empty())
287 return 0;
288 if (User.getAsInteger(10, Ret)) {
289 std::string Buf;
290 raw_string_ostream OS(Buf);
291 OS.write_escaped(User);
292 OS.flush();
293 uint64_t Offset = reinterpret_cast<const char *>(ArMemHdr) -
294 Parent->getData().data();
295 return malformedError("characters in UID field in archive header "
296 "are not all decimal numbers: '" + Buf + "' for the "
297 "archive member header at offset " + Twine(Offset));
298 }
299 return Ret;
300 }
301
getGID() const302 Expected<unsigned> ArchiveMemberHeader::getGID() const {
303 unsigned Ret;
304 StringRef Group = StringRef(ArMemHdr->GID, sizeof(ArMemHdr->GID)).rtrim(' ');
305 if (Group.empty())
306 return 0;
307 if (Group.getAsInteger(10, Ret)) {
308 std::string Buf;
309 raw_string_ostream OS(Buf);
310 OS.write_escaped(Group);
311 OS.flush();
312 uint64_t Offset = reinterpret_cast<const char *>(ArMemHdr) -
313 Parent->getData().data();
314 return malformedError("characters in GID field in archive header "
315 "are not all decimal numbers: '" + Buf + "' for the "
316 "archive member header at offset " + Twine(Offset));
317 }
318 return Ret;
319 }
320
Child(const Archive * Parent,StringRef Data,uint16_t StartOfFile)321 Archive::Child::Child(const Archive *Parent, StringRef Data,
322 uint16_t StartOfFile)
323 : Parent(Parent), Header(Parent, Data.data(), Data.size(), nullptr),
324 Data(Data), StartOfFile(StartOfFile) {
325 }
326
Child(const Archive * Parent,const char * Start,Error * Err)327 Archive::Child::Child(const Archive *Parent, const char *Start, Error *Err)
328 : Parent(Parent),
329 Header(Parent, Start,
330 Parent
331 ? Parent->getData().size() - (Start - Parent->getData().data())
332 : 0, Err) {
333 if (!Start)
334 return;
335
336 // If we are pointed to real data, Start is not a nullptr, then there must be
337 // a non-null Err pointer available to report malformed data on. Only in
338 // the case sentinel value is being constructed is Err is permitted to be a
339 // nullptr.
340 assert(Err && "Err can't be nullptr if Start is not a nullptr");
341
342 ErrorAsOutParameter ErrAsOutParam(Err);
343
344 // If there was an error in the construction of the Header
345 // then just return with the error now set.
346 if (*Err)
347 return;
348
349 uint64_t Size = Header.getSizeOf();
350 Data = StringRef(Start, Size);
351 Expected<bool> isThinOrErr = isThinMember();
352 if (!isThinOrErr) {
353 *Err = isThinOrErr.takeError();
354 return;
355 }
356 bool isThin = isThinOrErr.get();
357 if (!isThin) {
358 Expected<uint64_t> MemberSize = getRawSize();
359 if (!MemberSize) {
360 *Err = MemberSize.takeError();
361 return;
362 }
363 Size += MemberSize.get();
364 Data = StringRef(Start, Size);
365 }
366
367 // Setup StartOfFile and PaddingBytes.
368 StartOfFile = Header.getSizeOf();
369 // Don't include attached name.
370 Expected<StringRef> NameOrErr = getRawName();
371 if (!NameOrErr){
372 *Err = NameOrErr.takeError();
373 return;
374 }
375 StringRef Name = NameOrErr.get();
376 if (Name.startswith("#1/")) {
377 uint64_t NameSize;
378 if (Name.substr(3).rtrim(' ').getAsInteger(10, NameSize)) {
379 std::string Buf;
380 raw_string_ostream OS(Buf);
381 OS.write_escaped(Name.substr(3).rtrim(' '));
382 OS.flush();
383 uint64_t Offset = Start - Parent->getData().data();
384 *Err = malformedError("long name length characters after the #1/ are "
385 "not all decimal numbers: '" + Buf + "' for "
386 "archive member header at offset " +
387 Twine(Offset));
388 return;
389 }
390 StartOfFile += NameSize;
391 }
392 }
393
getSize() const394 Expected<uint64_t> Archive::Child::getSize() const {
395 if (Parent->IsThin)
396 return Header.getSize();
397 return Data.size() - StartOfFile;
398 }
399
getRawSize() const400 Expected<uint64_t> Archive::Child::getRawSize() const {
401 return Header.getSize();
402 }
403
isThinMember() const404 Expected<bool> Archive::Child::isThinMember() const {
405 Expected<StringRef> NameOrErr = Header.getRawName();
406 if (!NameOrErr)
407 return NameOrErr.takeError();
408 StringRef Name = NameOrErr.get();
409 return Parent->IsThin && Name != "/" && Name != "//";
410 }
411
getFullName() const412 Expected<std::string> Archive::Child::getFullName() const {
413 Expected<bool> isThin = isThinMember();
414 if (!isThin)
415 return isThin.takeError();
416 assert(isThin.get());
417 Expected<StringRef> NameOrErr = getName();
418 if (!NameOrErr)
419 return NameOrErr.takeError();
420 StringRef Name = *NameOrErr;
421 if (sys::path::is_absolute(Name))
422 return std::string(Name);
423
424 SmallString<128> FullName = sys::path::parent_path(
425 Parent->getMemoryBufferRef().getBufferIdentifier());
426 sys::path::append(FullName, Name);
427 return std::string(FullName.str());
428 }
429
getBuffer() const430 Expected<StringRef> Archive::Child::getBuffer() const {
431 Expected<bool> isThinOrErr = isThinMember();
432 if (!isThinOrErr)
433 return isThinOrErr.takeError();
434 bool isThin = isThinOrErr.get();
435 if (!isThin) {
436 Expected<uint64_t> Size = getSize();
437 if (!Size)
438 return Size.takeError();
439 return StringRef(Data.data() + StartOfFile, Size.get());
440 }
441 Expected<std::string> FullNameOrErr = getFullName();
442 if (!FullNameOrErr)
443 return FullNameOrErr.takeError();
444 const std::string &FullName = *FullNameOrErr;
445 ErrorOr<std::unique_ptr<MemoryBuffer>> Buf = MemoryBuffer::getFile(FullName);
446 if (std::error_code EC = Buf.getError())
447 return errorCodeToError(EC);
448 Parent->ThinBuffers.push_back(std::move(*Buf));
449 return Parent->ThinBuffers.back()->getBuffer();
450 }
451
getNext() const452 Expected<Archive::Child> Archive::Child::getNext() const {
453 size_t SpaceToSkip = Data.size();
454 // If it's odd, add 1 to make it even.
455 if (SpaceToSkip & 1)
456 ++SpaceToSkip;
457
458 const char *NextLoc = Data.data() + SpaceToSkip;
459
460 // Check to see if this is at the end of the archive.
461 if (NextLoc == Parent->Data.getBufferEnd())
462 return Child(nullptr, nullptr, nullptr);
463
464 // Check to see if this is past the end of the archive.
465 if (NextLoc > Parent->Data.getBufferEnd()) {
466 std::string Msg("offset to next archive member past the end of the archive "
467 "after member ");
468 Expected<StringRef> NameOrErr = getName();
469 if (!NameOrErr) {
470 consumeError(NameOrErr.takeError());
471 uint64_t Offset = Data.data() - Parent->getData().data();
472 return malformedError(Msg + "at offset " + Twine(Offset));
473 } else
474 return malformedError(Msg + NameOrErr.get());
475 }
476
477 Error Err = Error::success();
478 Child Ret(Parent, NextLoc, &Err);
479 if (Err)
480 return std::move(Err);
481 return Ret;
482 }
483
getChildOffset() const484 uint64_t Archive::Child::getChildOffset() const {
485 const char *a = Parent->Data.getBuffer().data();
486 const char *c = Data.data();
487 uint64_t offset = c - a;
488 return offset;
489 }
490
getName() const491 Expected<StringRef> Archive::Child::getName() const {
492 Expected<uint64_t> RawSizeOrErr = getRawSize();
493 if (!RawSizeOrErr)
494 return RawSizeOrErr.takeError();
495 uint64_t RawSize = RawSizeOrErr.get();
496 Expected<StringRef> NameOrErr = Header.getName(Header.getSizeOf() + RawSize);
497 if (!NameOrErr)
498 return NameOrErr.takeError();
499 StringRef Name = NameOrErr.get();
500 return Name;
501 }
502
getMemoryBufferRef() const503 Expected<MemoryBufferRef> Archive::Child::getMemoryBufferRef() const {
504 Expected<StringRef> NameOrErr = getName();
505 if (!NameOrErr)
506 return NameOrErr.takeError();
507 StringRef Name = NameOrErr.get();
508 Expected<StringRef> Buf = getBuffer();
509 if (!Buf)
510 return createFileError(Name, Buf.takeError());
511 return MemoryBufferRef(*Buf, Name);
512 }
513
514 Expected<std::unique_ptr<Binary>>
getAsBinary(LLVMContext * Context) const515 Archive::Child::getAsBinary(LLVMContext *Context) const {
516 Expected<MemoryBufferRef> BuffOrErr = getMemoryBufferRef();
517 if (!BuffOrErr)
518 return BuffOrErr.takeError();
519
520 auto BinaryOrErr = createBinary(BuffOrErr.get(), Context);
521 if (BinaryOrErr)
522 return std::move(*BinaryOrErr);
523 return BinaryOrErr.takeError();
524 }
525
create(MemoryBufferRef Source)526 Expected<std::unique_ptr<Archive>> Archive::create(MemoryBufferRef Source) {
527 Error Err = Error::success();
528 std::unique_ptr<Archive> Ret(new Archive(Source, Err));
529 if (Err)
530 return std::move(Err);
531 return std::move(Ret);
532 }
533
setFirstRegular(const Child & C)534 void Archive::setFirstRegular(const Child &C) {
535 FirstRegularData = C.Data;
536 FirstRegularStartOfFile = C.StartOfFile;
537 }
538
Archive(MemoryBufferRef Source,Error & Err)539 Archive::Archive(MemoryBufferRef Source, Error &Err)
540 : Binary(Binary::ID_Archive, Source) {
541 ErrorAsOutParameter ErrAsOutParam(&Err);
542 StringRef Buffer = Data.getBuffer();
543 // Check for sufficient magic.
544 if (Buffer.startswith(ThinMagic)) {
545 IsThin = true;
546 } else if (Buffer.startswith(Magic)) {
547 IsThin = false;
548 } else {
549 Err = make_error<GenericBinaryError>("file too small to be an archive",
550 object_error::invalid_file_type);
551 return;
552 }
553
554 // Make sure Format is initialized before any call to
555 // ArchiveMemberHeader::getName() is made. This could be a valid empty
556 // archive which is the same in all formats. So claiming it to be gnu to is
557 // fine if not totally correct before we look for a string table or table of
558 // contents.
559 Format = K_GNU;
560
561 // Get the special members.
562 child_iterator I = child_begin(Err, false);
563 if (Err)
564 return;
565 child_iterator E = child_end();
566
567 // See if this is a valid empty archive and if so return.
568 if (I == E) {
569 Err = Error::success();
570 return;
571 }
572 const Child *C = &*I;
573
574 auto Increment = [&]() {
575 ++I;
576 if (Err)
577 return true;
578 C = &*I;
579 return false;
580 };
581
582 Expected<StringRef> NameOrErr = C->getRawName();
583 if (!NameOrErr) {
584 Err = NameOrErr.takeError();
585 return;
586 }
587 StringRef Name = NameOrErr.get();
588
589 // Below is the pattern that is used to figure out the archive format
590 // GNU archive format
591 // First member : / (may exist, if it exists, points to the symbol table )
592 // Second member : // (may exist, if it exists, points to the string table)
593 // Note : The string table is used if the filename exceeds 15 characters
594 // BSD archive format
595 // First member : __.SYMDEF or "__.SYMDEF SORTED" (the symbol table)
596 // There is no string table, if the filename exceeds 15 characters or has a
597 // embedded space, the filename has #1/<size>, The size represents the size
598 // of the filename that needs to be read after the archive header
599 // COFF archive format
600 // First member : /
601 // Second member : / (provides a directory of symbols)
602 // Third member : // (may exist, if it exists, contains the string table)
603 // Note: Microsoft PE/COFF Spec 8.3 says that the third member is present
604 // even if the string table is empty. However, lib.exe does not in fact
605 // seem to create the third member if there's no member whose filename
606 // exceeds 15 characters. So the third member is optional.
607
608 if (Name == "__.SYMDEF" || Name == "__.SYMDEF_64") {
609 if (Name == "__.SYMDEF")
610 Format = K_BSD;
611 else // Name == "__.SYMDEF_64"
612 Format = K_DARWIN64;
613 // We know that the symbol table is not an external file, but we still must
614 // check any Expected<> return value.
615 Expected<StringRef> BufOrErr = C->getBuffer();
616 if (!BufOrErr) {
617 Err = BufOrErr.takeError();
618 return;
619 }
620 SymbolTable = BufOrErr.get();
621 if (Increment())
622 return;
623 setFirstRegular(*C);
624
625 Err = Error::success();
626 return;
627 }
628
629 if (Name.startswith("#1/")) {
630 Format = K_BSD;
631 // We know this is BSD, so getName will work since there is no string table.
632 Expected<StringRef> NameOrErr = C->getName();
633 if (!NameOrErr) {
634 Err = NameOrErr.takeError();
635 return;
636 }
637 Name = NameOrErr.get();
638 if (Name == "__.SYMDEF SORTED" || Name == "__.SYMDEF") {
639 // We know that the symbol table is not an external file, but we still
640 // must check any Expected<> return value.
641 Expected<StringRef> BufOrErr = C->getBuffer();
642 if (!BufOrErr) {
643 Err = BufOrErr.takeError();
644 return;
645 }
646 SymbolTable = BufOrErr.get();
647 if (Increment())
648 return;
649 }
650 else if (Name == "__.SYMDEF_64 SORTED" || Name == "__.SYMDEF_64") {
651 Format = K_DARWIN64;
652 // We know that the symbol table is not an external file, but we still
653 // must check any Expected<> return value.
654 Expected<StringRef> BufOrErr = C->getBuffer();
655 if (!BufOrErr) {
656 Err = BufOrErr.takeError();
657 return;
658 }
659 SymbolTable = BufOrErr.get();
660 if (Increment())
661 return;
662 }
663 setFirstRegular(*C);
664 return;
665 }
666
667 // MIPS 64-bit ELF archives use a special format of a symbol table.
668 // This format is marked by `ar_name` field equals to "/SYM64/".
669 // For detailed description see page 96 in the following document:
670 // http://techpubs.sgi.com/library/manuals/4000/007-4658-001/pdf/007-4658-001.pdf
671
672 bool has64SymTable = false;
673 if (Name == "/" || Name == "/SYM64/") {
674 // We know that the symbol table is not an external file, but we still
675 // must check any Expected<> return value.
676 Expected<StringRef> BufOrErr = C->getBuffer();
677 if (!BufOrErr) {
678 Err = BufOrErr.takeError();
679 return;
680 }
681 SymbolTable = BufOrErr.get();
682 if (Name == "/SYM64/")
683 has64SymTable = true;
684
685 if (Increment())
686 return;
687 if (I == E) {
688 Err = Error::success();
689 return;
690 }
691 Expected<StringRef> NameOrErr = C->getRawName();
692 if (!NameOrErr) {
693 Err = NameOrErr.takeError();
694 return;
695 }
696 Name = NameOrErr.get();
697 }
698
699 if (Name == "//") {
700 Format = has64SymTable ? K_GNU64 : K_GNU;
701 // The string table is never an external member, but we still
702 // must check any Expected<> return value.
703 Expected<StringRef> BufOrErr = C->getBuffer();
704 if (!BufOrErr) {
705 Err = BufOrErr.takeError();
706 return;
707 }
708 StringTable = BufOrErr.get();
709 if (Increment())
710 return;
711 setFirstRegular(*C);
712 Err = Error::success();
713 return;
714 }
715
716 if (Name[0] != '/') {
717 Format = has64SymTable ? K_GNU64 : K_GNU;
718 setFirstRegular(*C);
719 Err = Error::success();
720 return;
721 }
722
723 if (Name != "/") {
724 Err = errorCodeToError(object_error::parse_failed);
725 return;
726 }
727
728 Format = K_COFF;
729 // We know that the symbol table is not an external file, but we still
730 // must check any Expected<> return value.
731 Expected<StringRef> BufOrErr = C->getBuffer();
732 if (!BufOrErr) {
733 Err = BufOrErr.takeError();
734 return;
735 }
736 SymbolTable = BufOrErr.get();
737
738 if (Increment())
739 return;
740
741 if (I == E) {
742 setFirstRegular(*C);
743 Err = Error::success();
744 return;
745 }
746
747 NameOrErr = C->getRawName();
748 if (!NameOrErr) {
749 Err = NameOrErr.takeError();
750 return;
751 }
752 Name = NameOrErr.get();
753
754 if (Name == "//") {
755 // The string table is never an external member, but we still
756 // must check any Expected<> return value.
757 Expected<StringRef> BufOrErr = C->getBuffer();
758 if (!BufOrErr) {
759 Err = BufOrErr.takeError();
760 return;
761 }
762 StringTable = BufOrErr.get();
763 if (Increment())
764 return;
765 }
766
767 setFirstRegular(*C);
768 Err = Error::success();
769 }
770
child_begin(Error & Err,bool SkipInternal) const771 Archive::child_iterator Archive::child_begin(Error &Err,
772 bool SkipInternal) const {
773 if (isEmpty())
774 return child_end();
775
776 if (SkipInternal)
777 return child_iterator::itr(
778 Child(this, FirstRegularData, FirstRegularStartOfFile), Err);
779
780 const char *Loc = Data.getBufferStart() + strlen(Magic);
781 Child C(this, Loc, &Err);
782 if (Err)
783 return child_end();
784 return child_iterator::itr(C, Err);
785 }
786
child_end() const787 Archive::child_iterator Archive::child_end() const {
788 return child_iterator::end(Child(nullptr, nullptr, nullptr));
789 }
790
getName() const791 StringRef Archive::Symbol::getName() const {
792 return Parent->getSymbolTable().begin() + StringIndex;
793 }
794
getMember() const795 Expected<Archive::Child> Archive::Symbol::getMember() const {
796 const char *Buf = Parent->getSymbolTable().begin();
797 const char *Offsets = Buf;
798 if (Parent->kind() == K_GNU64 || Parent->kind() == K_DARWIN64)
799 Offsets += sizeof(uint64_t);
800 else
801 Offsets += sizeof(uint32_t);
802 uint64_t Offset = 0;
803 if (Parent->kind() == K_GNU) {
804 Offset = read32be(Offsets + SymbolIndex * 4);
805 } else if (Parent->kind() == K_GNU64) {
806 Offset = read64be(Offsets + SymbolIndex * 8);
807 } else if (Parent->kind() == K_BSD) {
808 // The SymbolIndex is an index into the ranlib structs that start at
809 // Offsets (the first uint32_t is the number of bytes of the ranlib
810 // structs). The ranlib structs are a pair of uint32_t's the first
811 // being a string table offset and the second being the offset into
812 // the archive of the member that defines the symbol. Which is what
813 // is needed here.
814 Offset = read32le(Offsets + SymbolIndex * 8 + 4);
815 } else if (Parent->kind() == K_DARWIN64) {
816 // The SymbolIndex is an index into the ranlib_64 structs that start at
817 // Offsets (the first uint64_t is the number of bytes of the ranlib_64
818 // structs). The ranlib_64 structs are a pair of uint64_t's the first
819 // being a string table offset and the second being the offset into
820 // the archive of the member that defines the symbol. Which is what
821 // is needed here.
822 Offset = read64le(Offsets + SymbolIndex * 16 + 8);
823 } else {
824 // Skip offsets.
825 uint32_t MemberCount = read32le(Buf);
826 Buf += MemberCount * 4 + 4;
827
828 uint32_t SymbolCount = read32le(Buf);
829 if (SymbolIndex >= SymbolCount)
830 return errorCodeToError(object_error::parse_failed);
831
832 // Skip SymbolCount to get to the indices table.
833 const char *Indices = Buf + 4;
834
835 // Get the index of the offset in the file member offset table for this
836 // symbol.
837 uint16_t OffsetIndex = read16le(Indices + SymbolIndex * 2);
838 // Subtract 1 since OffsetIndex is 1 based.
839 --OffsetIndex;
840
841 if (OffsetIndex >= MemberCount)
842 return errorCodeToError(object_error::parse_failed);
843
844 Offset = read32le(Offsets + OffsetIndex * 4);
845 }
846
847 const char *Loc = Parent->getData().begin() + Offset;
848 Error Err = Error::success();
849 Child C(Parent, Loc, &Err);
850 if (Err)
851 return std::move(Err);
852 return C;
853 }
854
getNext() const855 Archive::Symbol Archive::Symbol::getNext() const {
856 Symbol t(*this);
857 if (Parent->kind() == K_BSD) {
858 // t.StringIndex is an offset from the start of the __.SYMDEF or
859 // "__.SYMDEF SORTED" member into the string table for the ranlib
860 // struct indexed by t.SymbolIndex . To change t.StringIndex to the
861 // offset in the string table for t.SymbolIndex+1 we subtract the
862 // its offset from the start of the string table for t.SymbolIndex
863 // and add the offset of the string table for t.SymbolIndex+1.
864
865 // The __.SYMDEF or "__.SYMDEF SORTED" member starts with a uint32_t
866 // which is the number of bytes of ranlib structs that follow. The ranlib
867 // structs are a pair of uint32_t's the first being a string table offset
868 // and the second being the offset into the archive of the member that
869 // define the symbol. After that the next uint32_t is the byte count of
870 // the string table followed by the string table.
871 const char *Buf = Parent->getSymbolTable().begin();
872 uint32_t RanlibCount = 0;
873 RanlibCount = read32le(Buf) / 8;
874 // If t.SymbolIndex + 1 will be past the count of symbols (the RanlibCount)
875 // don't change the t.StringIndex as we don't want to reference a ranlib
876 // past RanlibCount.
877 if (t.SymbolIndex + 1 < RanlibCount) {
878 const char *Ranlibs = Buf + 4;
879 uint32_t CurRanStrx = 0;
880 uint32_t NextRanStrx = 0;
881 CurRanStrx = read32le(Ranlibs + t.SymbolIndex * 8);
882 NextRanStrx = read32le(Ranlibs + (t.SymbolIndex + 1) * 8);
883 t.StringIndex -= CurRanStrx;
884 t.StringIndex += NextRanStrx;
885 }
886 } else {
887 // Go to one past next null.
888 t.StringIndex = Parent->getSymbolTable().find('\0', t.StringIndex) + 1;
889 }
890 ++t.SymbolIndex;
891 return t;
892 }
893
symbol_begin() const894 Archive::symbol_iterator Archive::symbol_begin() const {
895 if (!hasSymbolTable())
896 return symbol_iterator(Symbol(this, 0, 0));
897
898 const char *buf = getSymbolTable().begin();
899 if (kind() == K_GNU) {
900 uint32_t symbol_count = 0;
901 symbol_count = read32be(buf);
902 buf += sizeof(uint32_t) + (symbol_count * (sizeof(uint32_t)));
903 } else if (kind() == K_GNU64) {
904 uint64_t symbol_count = read64be(buf);
905 buf += sizeof(uint64_t) + (symbol_count * (sizeof(uint64_t)));
906 } else if (kind() == K_BSD) {
907 // The __.SYMDEF or "__.SYMDEF SORTED" member starts with a uint32_t
908 // which is the number of bytes of ranlib structs that follow. The ranlib
909 // structs are a pair of uint32_t's the first being a string table offset
910 // and the second being the offset into the archive of the member that
911 // define the symbol. After that the next uint32_t is the byte count of
912 // the string table followed by the string table.
913 uint32_t ranlib_count = 0;
914 ranlib_count = read32le(buf) / 8;
915 const char *ranlibs = buf + 4;
916 uint32_t ran_strx = 0;
917 ran_strx = read32le(ranlibs);
918 buf += sizeof(uint32_t) + (ranlib_count * (2 * (sizeof(uint32_t))));
919 // Skip the byte count of the string table.
920 buf += sizeof(uint32_t);
921 buf += ran_strx;
922 } else if (kind() == K_DARWIN64) {
923 // The __.SYMDEF_64 or "__.SYMDEF_64 SORTED" member starts with a uint64_t
924 // which is the number of bytes of ranlib_64 structs that follow. The
925 // ranlib_64 structs are a pair of uint64_t's the first being a string
926 // table offset and the second being the offset into the archive of the
927 // member that define the symbol. After that the next uint64_t is the byte
928 // count of the string table followed by the string table.
929 uint64_t ranlib_count = 0;
930 ranlib_count = read64le(buf) / 16;
931 const char *ranlibs = buf + 8;
932 uint64_t ran_strx = 0;
933 ran_strx = read64le(ranlibs);
934 buf += sizeof(uint64_t) + (ranlib_count * (2 * (sizeof(uint64_t))));
935 // Skip the byte count of the string table.
936 buf += sizeof(uint64_t);
937 buf += ran_strx;
938 } else {
939 uint32_t member_count = 0;
940 uint32_t symbol_count = 0;
941 member_count = read32le(buf);
942 buf += 4 + (member_count * 4); // Skip offsets.
943 symbol_count = read32le(buf);
944 buf += 4 + (symbol_count * 2); // Skip indices.
945 }
946 uint32_t string_start_offset = buf - getSymbolTable().begin();
947 return symbol_iterator(Symbol(this, 0, string_start_offset));
948 }
949
symbol_end() const950 Archive::symbol_iterator Archive::symbol_end() const {
951 return symbol_iterator(Symbol(this, getNumberOfSymbols(), 0));
952 }
953
getNumberOfSymbols() const954 uint32_t Archive::getNumberOfSymbols() const {
955 if (!hasSymbolTable())
956 return 0;
957 const char *buf = getSymbolTable().begin();
958 if (kind() == K_GNU)
959 return read32be(buf);
960 if (kind() == K_GNU64)
961 return read64be(buf);
962 if (kind() == K_BSD)
963 return read32le(buf) / 8;
964 if (kind() == K_DARWIN64)
965 return read64le(buf) / 16;
966 uint32_t member_count = 0;
967 member_count = read32le(buf);
968 buf += 4 + (member_count * 4); // Skip offsets.
969 return read32le(buf);
970 }
971
findSym(StringRef name) const972 Expected<Optional<Archive::Child>> Archive::findSym(StringRef name) const {
973 Archive::symbol_iterator bs = symbol_begin();
974 Archive::symbol_iterator es = symbol_end();
975
976 for (; bs != es; ++bs) {
977 StringRef SymName = bs->getName();
978 if (SymName == name) {
979 if (auto MemberOrErr = bs->getMember())
980 return Child(*MemberOrErr);
981 else
982 return MemberOrErr.takeError();
983 }
984 }
985 return Optional<Child>();
986 }
987
988 // Returns true if archive file contains no member file.
isEmpty() const989 bool Archive::isEmpty() const { return Data.getBufferSize() == 8; }
990
hasSymbolTable() const991 bool Archive::hasSymbolTable() const { return !SymbolTable.empty(); }
992