X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FObject%2FArchive.cpp;h=99b0650c8b7e26efc0d7bd832f1cdc4f42cb0250;hb=10a5589d08c1de3fcd715ce23697d4e591519595;hp=8b362d66db084b79938453afe9e47d56da56e280;hpb=4e2b922131ae617cb8738d1871e9d918c44bdb69;p=oota-llvm.git diff --git a/lib/Object/Archive.cpp b/lib/Object/Archive.cpp index 8b362d66db0..99b0650c8b7 100644 --- a/lib/Object/Archive.cpp +++ b/lib/Object/Archive.cpp @@ -17,11 +17,14 @@ #include "llvm/ADT/Twine.h" #include "llvm/Support/Endian.h" #include "llvm/Support/MemoryBuffer.h" +#include "llvm/Support/Path.h" using namespace llvm; using namespace object; +using namespace llvm::support::endian; static const char *const Magic = "!\n"; +static const char *const ThinMagic = "!\n"; void Archive::anchor() { } @@ -40,10 +43,10 @@ StringRef ArchiveMemberHeader::getName() const { return llvm::StringRef(Name, end); } -uint32_t ArchiveMemberHeader::getSize() const { +ErrorOr ArchiveMemberHeader::getSize() const { uint32_t Ret; if (llvm::StringRef(Size, sizeof(Size)).rtrim(" ").getAsInteger(10, Ret)) - llvm_unreachable("Size is not a decimal number."); + return object_error::parse_failed; // Size is not a decimal number. return Ret; } @@ -79,19 +82,30 @@ unsigned ArchiveMemberHeader::getGID() const { return Ret; } -Archive::Child::Child(const Archive *Parent, const char *Start) +Archive::Child::Child(const Archive *Parent, StringRef Data, + uint16_t StartOfFile) + : Parent(Parent), Data(Data), StartOfFile(StartOfFile) {} + +Archive::Child::Child(const Archive *Parent, const char *Start, + std::error_code *EC) : Parent(Parent) { if (!Start) return; - const ArchiveMemberHeader *Header = - reinterpret_cast(Start); - Data = StringRef(Start, sizeof(ArchiveMemberHeader) + Header->getSize()); + uint64_t Size = sizeof(ArchiveMemberHeader); + Data = StringRef(Start, Size); + if (!isThinMember()) { + ErrorOr MemberSize = getRawSize(); + if ((*EC = MemberSize.getError())) + return; + Size += MemberSize.get(); + Data = StringRef(Start, Size); + } // Setup StartOfFile and PaddingBytes. StartOfFile = sizeof(ArchiveMemberHeader); // Don't include attached name. - StringRef Name = Header->getName(); + StringRef Name = getRawName(); if (Name.startswith("#1/")) { uint64_t NameSize; if (Name.substr(3).rtrim(" ").getAsInteger(10, NameSize)) @@ -100,7 +114,49 @@ Archive::Child::Child(const Archive *Parent, const char *Start) } } -Archive::Child Archive::Child::getNext() const { +ErrorOr Archive::Child::getSize() const { + if (Parent->IsThin) { + ErrorOr Size = getHeader()->getSize(); + if (std::error_code EC = Size.getError()) + return EC; + return Size.get(); + } + return Data.size() - StartOfFile; +} + +ErrorOr Archive::Child::getRawSize() const { + ErrorOr Size = getHeader()->getSize(); + if (std::error_code EC = Size.getError()) + return EC; + return Size.get(); +} + +bool Archive::Child::isThinMember() const { + StringRef Name = getHeader()->getName(); + return Parent->IsThin && Name != "/" && Name != "//"; +} + +ErrorOr Archive::Child::getBuffer() const { + if (!isThinMember()) { + ErrorOr Size = getSize(); + if (std::error_code EC = Size.getError()) + return EC; + return StringRef(Data.data() + StartOfFile, Size.get()); + } + ErrorOr Name = getName(); + if (std::error_code EC = Name.getError()) + return EC; + SmallString<128> FullName = sys::path::parent_path( + Parent->getMemoryBufferRef().getBufferIdentifier()); + sys::path::append(FullName, *Name); + ErrorOr> Buf = MemoryBuffer::getFile(FullName); + if (std::error_code EC = Buf.getError()) + return EC; + Parent->ThinBuffers.push_back(std::move(*Buf)); + return Parent->ThinBuffers.back()->getBuffer(); +} + +ErrorOr Archive::Child::getNext() const { size_t SpaceToSkip = Data.size(); // If it's odd, add 1 to make it even. if (SpaceToSkip & 1) @@ -108,122 +164,133 @@ Archive::Child Archive::Child::getNext() const { const char *NextLoc = Data.data() + SpaceToSkip; + // Check to see if this is at the end of the archive. + if (NextLoc == Parent->Data.getBufferEnd()) + return Child(Parent, nullptr, nullptr); + // Check to see if this is past the end of the archive. - if (NextLoc >= Parent->Data->getBufferEnd()) - return Child(Parent, nullptr); + if (NextLoc > Parent->Data.getBufferEnd()) + return object_error::parse_failed; - return Child(Parent, NextLoc); + std::error_code EC; + Child Ret(Parent, NextLoc, &EC); + if (EC) + return EC; + return Ret; } -std::error_code Archive::Child::getName(StringRef &Result) const { +uint64_t Archive::Child::getChildOffset() const { + const char *a = Parent->Data.getBuffer().data(); + const char *c = Data.data(); + uint64_t offset = c - a; + return offset; +} + +ErrorOr Archive::Child::getName() const { StringRef name = getRawName(); // Check if it's a special name. if (name[0] == '/') { - if (name.size() == 1) { // Linker member. - Result = name; - return object_error::success; - } - if (name.size() == 2 && name[1] == '/') { // String table. - Result = name; - return object_error::success; - } + if (name.size() == 1) // Linker member. + return name; + if (name.size() == 2 && name[1] == '/') // String table. + return name; // It's a long name. // Get the offset. std::size_t offset; if (name.substr(1).rtrim(" ").getAsInteger(10, offset)) llvm_unreachable("Long name offset is not an integer"); - const char *addr = Parent->StringTable->Data.begin() - + sizeof(ArchiveMemberHeader) - + offset; + // Verify it. - if (Parent->StringTable == Parent->child_end() - || addr < (Parent->StringTable->Data.begin() - + sizeof(ArchiveMemberHeader)) - || addr > (Parent->StringTable->Data.begin() - + sizeof(ArchiveMemberHeader) - + Parent->StringTable->getSize())) + if (offset >= Parent->StringTable.size()) return object_error::parse_failed; + const char *addr = Parent->StringTable.begin() + offset; - // GNU long file names end with a /. - if (Parent->kind() == K_GNU) { - StringRef::size_type End = StringRef(addr).find('/'); - Result = StringRef(addr, End); - } else { - Result = addr; + // GNU long file names end with a "/\n". + if (Parent->kind() == K_GNU || Parent->kind() == K_MIPS64) { + StringRef::size_type End = StringRef(addr).find('\n'); + return StringRef(addr, End - 1); } - return object_error::success; + return StringRef(addr); } else if (name.startswith("#1/")) { uint64_t name_size; if (name.substr(3).rtrim(" ").getAsInteger(10, name_size)) llvm_unreachable("Long name length is not an ingeter"); - Result = Data.substr(sizeof(ArchiveMemberHeader), name_size) + return Data.substr(sizeof(ArchiveMemberHeader), name_size) .rtrim(StringRef("\0", 1)); - return object_error::success; } // It's a simple name. if (name[name.size() - 1] == '/') - Result = name.substr(0, name.size() - 1); - else - Result = name; - return object_error::success; + return name.substr(0, name.size() - 1); + return name; } -std::error_code -Archive::Child::getMemoryBuffer(std::unique_ptr &Result, - bool FullPath) const { - StringRef Name; - if (std::error_code ec = getName(Name)) - return ec; - SmallString<128> Path; - Result.reset(MemoryBuffer::getMemBuffer( - getBuffer(), FullPath ? (Twine(Parent->getFileName()) + "(" + Name + ")") - .toStringRef(Path) - : Name, - false)); - return std::error_code(); +ErrorOr Archive::Child::getMemoryBufferRef() const { + ErrorOr NameOrErr = getName(); + if (std::error_code EC = NameOrErr.getError()) + return EC; + StringRef Name = NameOrErr.get(); + ErrorOr Buf = getBuffer(); + if (std::error_code EC = Buf.getError()) + return EC; + return MemoryBufferRef(*Buf, Name); } -std::error_code Archive::Child::getAsBinary(std::unique_ptr &Result, - LLVMContext *Context) const { - std::unique_ptr ret; - std::unique_ptr Buff; - if (std::error_code ec = getMemoryBuffer(Buff)) - return ec; - ErrorOr BinaryOrErr = createBinary(Buff.release(), Context); - if (std::error_code EC = BinaryOrErr.getError()) +ErrorOr> +Archive::Child::getAsBinary(LLVMContext *Context) const { + ErrorOr BuffOrErr = getMemoryBufferRef(); + if (std::error_code EC = BuffOrErr.getError()) return EC; - Result.reset(BinaryOrErr.get()); - return object_error::success; + + return createBinary(BuffOrErr.get(), Context); } -ErrorOr Archive::create(MemoryBuffer *Source) { +ErrorOr> Archive::create(MemoryBufferRef Source) { std::error_code EC; std::unique_ptr Ret(new Archive(Source, EC)); if (EC) return EC; - return Ret.release(); + return std::move(Ret); } -Archive::Archive(MemoryBuffer *source, std::error_code &ec) - : Binary(Binary::ID_Archive, source), SymbolTable(child_end()) { +void Archive::setFirstRegular(const Child &C) { + FirstRegularData = C.Data; + FirstRegularStartOfFile = C.StartOfFile; +} + +Archive::Archive(MemoryBufferRef Source, std::error_code &ec) + : Binary(Binary::ID_Archive, Source) { + StringRef Buffer = Data.getBuffer(); // Check for sufficient magic. - assert(source); - if (source->getBufferSize() < 8 || - StringRef(source->getBufferStart(), 8) != Magic) { + if (Buffer.startswith(ThinMagic)) { + IsThin = true; + } else if (Buffer.startswith(Magic)) { + IsThin = false; + } else { ec = object_error::invalid_file_type; return; } // Get the special members. - child_iterator i = child_begin(false); - child_iterator e = child_end(); + child_iterator I = child_begin(false); + if ((ec = I->getError())) + return; + child_iterator E = child_end(); - if (i == e) { - ec = object_error::success; + if (I == E) { + ec = std::error_code(); return; } + const Child *C = &**I; - StringRef Name = i->getRawName(); + auto Increment = [&]() { + ++I; + if ((ec = I->getError())) + return true; + C = &**I; + return false; + }; + + StringRef Name = C->getRawName(); // Below is the pattern that is used to figure out the archive format // GNU archive format @@ -246,51 +313,74 @@ Archive::Archive(MemoryBuffer *source, std::error_code &ec) if (Name == "__.SYMDEF") { Format = K_BSD; - SymbolTable = i; - ++i; - FirstRegular = i; - ec = object_error::success; + // We know that the symbol table is not an external file, so we just assert + // there is no error. + SymbolTable = *C->getBuffer(); + if (Increment()) + return; + setFirstRegular(*C); + + ec = std::error_code(); return; } if (Name.startswith("#1/")) { Format = K_BSD; // We know this is BSD, so getName will work since there is no string table. - ec = i->getName(Name); + ErrorOr NameOrErr = C->getName(); + ec = NameOrErr.getError(); if (ec) return; - if (Name == "__.SYMDEF SORTED") { - SymbolTable = i; - ++i; + Name = NameOrErr.get(); + if (Name == "__.SYMDEF SORTED" || Name == "__.SYMDEF") { + // We know that the symbol table is not an external file, so we just + // assert there is no error. + SymbolTable = *C->getBuffer(); + if (Increment()) + return; } - FirstRegular = i; + setFirstRegular(*C); return; } - if (Name == "/") { - SymbolTable = i; + // MIPS 64-bit ELF archives use a special format of a symbol table. + // This format is marked by `ar_name` field equals to "/SYM64/". + // For detailed description see page 96 in the following document: + // http://techpubs.sgi.com/library/manuals/4000/007-4658-001/pdf/007-4658-001.pdf - ++i; - if (i == e) { - ec = object_error::parse_failed; + bool has64SymTable = false; + if (Name == "/" || Name == "/SYM64/") { + // We know that the symbol table is not an external file, so we just assert + // there is no error. + SymbolTable = *C->getBuffer(); + if (Name == "/SYM64/") + has64SymTable = true; + + if (Increment()) + return; + if (I == E) { + ec = std::error_code(); return; } - Name = i->getRawName(); + Name = C->getRawName(); } if (Name == "//") { - Format = K_GNU; - StringTable = i; - ++i; - FirstRegular = i; - ec = object_error::success; + Format = has64SymTable ? K_MIPS64 : K_GNU; + // The string table is never an external member, so we just assert on the + // ErrorOr. + StringTable = *C->getBuffer(); + if (Increment()) + return; + setFirstRegular(*C); + ec = std::error_code(); return; } if (Name[0] != '/') { - Format = K_GNU; - FirstRegular = i; - ec = object_error::success; + Format = has64SymTable ? K_MIPS64 : K_GNU; + setFirstRegular(*C); + ec = std::error_code(); return; } @@ -300,97 +390,143 @@ Archive::Archive(MemoryBuffer *source, std::error_code &ec) } Format = K_COFF; - SymbolTable = i; + // We know that the symbol table is not an external file, so we just assert + // there is no error. + SymbolTable = *C->getBuffer(); + + if (Increment()) + return; - ++i; - if (i == e) { - FirstRegular = i; - ec = object_error::success; + if (I == E) { + setFirstRegular(*C); + ec = std::error_code(); return; } - Name = i->getRawName(); + Name = C->getRawName(); if (Name == "//") { - StringTable = i; - ++i; + // The string table is never an external member, so we just assert on the + // ErrorOr. + StringTable = *C->getBuffer(); + if (Increment()) + return; } - FirstRegular = i; - ec = object_error::success; + setFirstRegular(*C); + ec = std::error_code(); } Archive::child_iterator Archive::child_begin(bool SkipInternal) const { - if (Data->getBufferSize() == 8) // empty archive. + if (Data.getBufferSize() == 8) // empty archive. return child_end(); if (SkipInternal) - return FirstRegular; + return Child(this, FirstRegularData, FirstRegularStartOfFile); - const char *Loc = Data->getBufferStart() + strlen(Magic); - Child c(this, Loc); - return c; + const char *Loc = Data.getBufferStart() + strlen(Magic); + std::error_code EC; + Child c(this, Loc, &EC); + if (EC) + return child_iterator(EC); + return child_iterator(c); } Archive::child_iterator Archive::child_end() const { - return Child(this, nullptr); + return Child(this, nullptr, nullptr); } -std::error_code Archive::Symbol::getName(StringRef &Result) const { - Result = StringRef(Parent->SymbolTable->getBuffer().begin() + StringIndex); - return object_error::success; +StringRef Archive::Symbol::getName() const { + return Parent->getSymbolTable().begin() + StringIndex; } -std::error_code Archive::Symbol::getMember(child_iterator &Result) const { - const char *Buf = Parent->SymbolTable->getBuffer().begin(); - const char *Offsets = Buf + 4; +ErrorOr Archive::Symbol::getMember() const { + const char *Buf = Parent->getSymbolTable().begin(); + const char *Offsets = Buf; + if (Parent->kind() == K_MIPS64) + Offsets += sizeof(uint64_t); + else + Offsets += sizeof(uint32_t); uint32_t Offset = 0; if (Parent->kind() == K_GNU) { - Offset = *(reinterpret_cast(Offsets) - + SymbolIndex); + Offset = read32be(Offsets + SymbolIndex * 4); + } else if (Parent->kind() == K_MIPS64) { + Offset = read64be(Offsets + SymbolIndex * 8); } else if (Parent->kind() == K_BSD) { - llvm_unreachable("BSD format is not supported"); + // The SymbolIndex is an index into the ranlib structs that start at + // Offsets (the first uint32_t is the number of bytes of the ranlib + // structs). The ranlib structs are a pair of uint32_t's the first + // being a string table offset and the second being the offset into + // the archive of the member that defines the symbol. Which is what + // is needed here. + Offset = read32le(Offsets + SymbolIndex * 8 + 4); } else { - uint32_t MemberCount = *reinterpret_cast(Buf); - // Skip offsets. - Buf += sizeof(support::ulittle32_t) - + (MemberCount * sizeof(support::ulittle32_t)); - - uint32_t SymbolCount = *reinterpret_cast(Buf); + uint32_t MemberCount = read32le(Buf); + Buf += MemberCount * 4 + 4; + uint32_t SymbolCount = read32le(Buf); if (SymbolIndex >= SymbolCount) return object_error::parse_failed; // Skip SymbolCount to get to the indices table. - const char *Indices = Buf + sizeof(support::ulittle32_t); + const char *Indices = Buf + 4; // Get the index of the offset in the file member offset table for this // symbol. - uint16_t OffsetIndex = - *(reinterpret_cast(Indices) - + SymbolIndex); + uint16_t OffsetIndex = read16le(Indices + SymbolIndex * 2); // Subtract 1 since OffsetIndex is 1 based. --OffsetIndex; if (OffsetIndex >= MemberCount) return object_error::parse_failed; - Offset = *(reinterpret_cast(Offsets) - + OffsetIndex); + Offset = read32le(Offsets + OffsetIndex * 4); } const char *Loc = Parent->getData().begin() + Offset; - Result = Child(Parent, Loc); - - return object_error::success; + std::error_code EC; + Child C(Parent, Loc, &EC); + if (EC) + return EC; + return C; } Archive::Symbol Archive::Symbol::getNext() const { Symbol t(*this); - // Go to one past next null. - t.StringIndex = - Parent->SymbolTable->getBuffer().find('\0', t.StringIndex) + 1; + if (Parent->kind() == K_BSD) { + // t.StringIndex is an offset from the start of the __.SYMDEF or + // "__.SYMDEF SORTED" member into the string table for the ranlib + // struct indexed by t.SymbolIndex . To change t.StringIndex to the + // offset in the string table for t.SymbolIndex+1 we subtract the + // its offset from the start of the string table for t.SymbolIndex + // and add the offset of the string table for t.SymbolIndex+1. + + // The __.SYMDEF or "__.SYMDEF SORTED" member starts with a uint32_t + // which is the number of bytes of ranlib structs that follow. The ranlib + // structs are a pair of uint32_t's the first being a string table offset + // and the second being the offset into the archive of the member that + // define the symbol. After that the next uint32_t is the byte count of + // the string table followed by the string table. + const char *Buf = Parent->getSymbolTable().begin(); + uint32_t RanlibCount = 0; + RanlibCount = read32le(Buf) / 8; + // If t.SymbolIndex + 1 will be past the count of symbols (the RanlibCount) + // don't change the t.StringIndex as we don't want to reference a ranlib + // past RanlibCount. + if (t.SymbolIndex + 1 < RanlibCount) { + const char *Ranlibs = Buf + 4; + uint32_t CurRanStrx = 0; + uint32_t NextRanStrx = 0; + CurRanStrx = read32le(Ranlibs + t.SymbolIndex * 8); + NextRanStrx = read32le(Ranlibs + (t.SymbolIndex + 1) * 8); + t.StringIndex -= CurRanStrx; + t.StringIndex += NextRanStrx; + } + } else { + // Go to one past next null. + t.StringIndex = Parent->getSymbolTable().find('\0', t.StringIndex) + 1; + } ++t.SymbolIndex; return t; } @@ -399,63 +535,77 @@ Archive::symbol_iterator Archive::symbol_begin() const { if (!hasSymbolTable()) return symbol_iterator(Symbol(this, 0, 0)); - const char *buf = SymbolTable->getBuffer().begin(); + const char *buf = getSymbolTable().begin(); if (kind() == K_GNU) { uint32_t symbol_count = 0; - symbol_count = *reinterpret_cast(buf); + symbol_count = read32be(buf); buf += sizeof(uint32_t) + (symbol_count * (sizeof(uint32_t))); + } else if (kind() == K_MIPS64) { + uint64_t symbol_count = read64be(buf); + buf += sizeof(uint64_t) + (symbol_count * (sizeof(uint64_t))); } else if (kind() == K_BSD) { - llvm_unreachable("BSD archive format is not supported"); + // The __.SYMDEF or "__.SYMDEF SORTED" member starts with a uint32_t + // which is the number of bytes of ranlib structs that follow. The ranlib + // structs are a pair of uint32_t's the first being a string table offset + // and the second being the offset into the archive of the member that + // define the symbol. After that the next uint32_t is the byte count of + // the string table followed by the string table. + uint32_t ranlib_count = 0; + ranlib_count = read32le(buf) / 8; + const char *ranlibs = buf + 4; + uint32_t ran_strx = 0; + ran_strx = read32le(ranlibs); + buf += sizeof(uint32_t) + (ranlib_count * (2 * (sizeof(uint32_t)))); + // Skip the byte count of the string table. + buf += sizeof(uint32_t); + buf += ran_strx; } else { uint32_t member_count = 0; uint32_t symbol_count = 0; - member_count = *reinterpret_cast(buf); + member_count = read32le(buf); buf += 4 + (member_count * 4); // Skip offsets. - symbol_count = *reinterpret_cast(buf); + symbol_count = read32le(buf); buf += 4 + (symbol_count * 2); // Skip indices. } - uint32_t string_start_offset = buf - SymbolTable->getBuffer().begin(); + uint32_t string_start_offset = buf - getSymbolTable().begin(); return symbol_iterator(Symbol(this, 0, string_start_offset)); } Archive::symbol_iterator Archive::symbol_end() const { - if (!hasSymbolTable()) - return symbol_iterator(Symbol(this, 0, 0)); + return symbol_iterator(Symbol(this, getNumberOfSymbols(), 0)); +} - const char *buf = SymbolTable->getBuffer().begin(); - uint32_t symbol_count = 0; - if (kind() == K_GNU) { - symbol_count = *reinterpret_cast(buf); - } else if (kind() == K_BSD) { - llvm_unreachable("BSD archive format is not supported"); - } else { - uint32_t member_count = 0; - member_count = *reinterpret_cast(buf); - buf += 4 + (member_count * 4); // Skip offsets. - symbol_count = *reinterpret_cast(buf); - } - return symbol_iterator( - Symbol(this, symbol_count, 0)); +uint32_t Archive::getNumberOfSymbols() const { + if (!hasSymbolTable()) + return 0; + const char *buf = getSymbolTable().begin(); + if (kind() == K_GNU) + return read32be(buf); + if (kind() == K_MIPS64) + return read64be(buf); + if (kind() == K_BSD) + return read32le(buf) / 8; + uint32_t member_count = 0; + member_count = read32le(buf); + buf += 4 + (member_count * 4); // Skip offsets. + return read32le(buf); } Archive::child_iterator Archive::findSym(StringRef name) const { Archive::symbol_iterator bs = symbol_begin(); Archive::symbol_iterator es = symbol_end(); - Archive::child_iterator result; - - StringRef symname; + for (; bs != es; ++bs) { - if (bs->getName(symname)) + StringRef SymName = bs->getName(); + if (SymName == name) { + ErrorOr ResultOrErr = bs->getMember(); + // FIXME: Should we really eat the error? + if (ResultOrErr.getError()) return child_end(); - if (symname == name) { - if (bs->getMember(result)) - return child_end(); - return result; + return ResultOrErr.get(); } } return child_end(); } -bool Archive::hasSymbolTable() const { - return SymbolTable != child_end(); -} +bool Archive::hasSymbolTable() const { return !SymbolTable.empty(); }