diff --git a/archive/tar/common.go b/archive/tar/common.go new file mode 100644 index 0000000..e363aa7 --- /dev/null +++ b/archive/tar/common.go @@ -0,0 +1,305 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package tar implements access to tar archives. +// It aims to cover most of the variations, including those produced +// by GNU and BSD tars. +// +// References: +// http://www.freebsd.org/cgi/man.cgi?query=tar&sektion=5 +// http://www.gnu.org/software/tar/manual/html_node/Standard.html +// http://pubs.opengroup.org/onlinepubs/9699919799/utilities/pax.html +package tar + +import ( + "bytes" + "errors" + "fmt" + "os" + "path" + "time" +) + +const ( + blockSize = 512 + + // Types + TypeReg = '0' // regular file + TypeRegA = '\x00' // regular file + TypeLink = '1' // hard link + TypeSymlink = '2' // symbolic link + TypeChar = '3' // character device node + TypeBlock = '4' // block device node + TypeDir = '5' // directory + TypeFifo = '6' // fifo node + TypeCont = '7' // reserved + TypeXHeader = 'x' // extended header + TypeXGlobalHeader = 'g' // global extended header + TypeGNULongName = 'L' // Next file has a long name + TypeGNULongLink = 'K' // Next file symlinks to a file w/ a long name + TypeGNUSparse = 'S' // sparse file +) + +// A Header represents a single header in a tar archive. +// Some fields may not be populated. +type Header struct { + Name string // name of header file entry + Mode int64 // permission and mode bits + Uid int // user id of owner + Gid int // group id of owner + Size int64 // length in bytes + ModTime time.Time // modified time + Typeflag byte // type of header entry + Linkname string // target name of link + Uname string // user name of owner + Gname string // group name of owner + Devmajor int64 // major number of character or block device + Devminor int64 // minor number of character or block device + AccessTime time.Time // access time + ChangeTime time.Time // status change time + Xattrs map[string]string +} + +// File name constants from the tar spec. +const ( + fileNameSize = 100 // Maximum number of bytes in a standard tar name. + fileNamePrefixSize = 155 // Maximum number of ustar extension bytes. +) + +// FileInfo returns an os.FileInfo for the Header. +func (h *Header) FileInfo() os.FileInfo { + return headerFileInfo{h} +} + +// headerFileInfo implements os.FileInfo. +type headerFileInfo struct { + h *Header +} + +func (fi headerFileInfo) Size() int64 { return fi.h.Size } +func (fi headerFileInfo) IsDir() bool { return fi.Mode().IsDir() } +func (fi headerFileInfo) ModTime() time.Time { return fi.h.ModTime } +func (fi headerFileInfo) Sys() interface{} { return fi.h } + +// Name returns the base name of the file. +func (fi headerFileInfo) Name() string { + if fi.IsDir() { + return path.Base(path.Clean(fi.h.Name)) + } + return path.Base(fi.h.Name) +} + +// Mode returns the permission and mode bits for the headerFileInfo. +func (fi headerFileInfo) Mode() (mode os.FileMode) { + // Set file permission bits. + mode = os.FileMode(fi.h.Mode).Perm() + + // Set setuid, setgid and sticky bits. + if fi.h.Mode&c_ISUID != 0 { + // setuid + mode |= os.ModeSetuid + } + if fi.h.Mode&c_ISGID != 0 { + // setgid + mode |= os.ModeSetgid + } + if fi.h.Mode&c_ISVTX != 0 { + // sticky + mode |= os.ModeSticky + } + + // Set file mode bits. + // clear perm, setuid, setgid and sticky bits. + m := os.FileMode(fi.h.Mode) &^ 07777 + if m == c_ISDIR { + // directory + mode |= os.ModeDir + } + if m == c_ISFIFO { + // named pipe (FIFO) + mode |= os.ModeNamedPipe + } + if m == c_ISLNK { + // symbolic link + mode |= os.ModeSymlink + } + if m == c_ISBLK { + // device file + mode |= os.ModeDevice + } + if m == c_ISCHR { + // Unix character device + mode |= os.ModeDevice + mode |= os.ModeCharDevice + } + if m == c_ISSOCK { + // Unix domain socket + mode |= os.ModeSocket + } + + switch fi.h.Typeflag { + case TypeLink, TypeSymlink: + // hard link, symbolic link + mode |= os.ModeSymlink + case TypeChar: + // character device node + mode |= os.ModeDevice + mode |= os.ModeCharDevice + case TypeBlock: + // block device node + mode |= os.ModeDevice + case TypeDir: + // directory + mode |= os.ModeDir + case TypeFifo: + // fifo node + mode |= os.ModeNamedPipe + } + + return mode +} + +// sysStat, if non-nil, populates h from system-dependent fields of fi. +var sysStat func(fi os.FileInfo, h *Header) error + +// Mode constants from the tar spec. +const ( + c_ISUID = 04000 // Set uid + c_ISGID = 02000 // Set gid + c_ISVTX = 01000 // Save text (sticky bit) + c_ISDIR = 040000 // Directory + c_ISFIFO = 010000 // FIFO + c_ISREG = 0100000 // Regular file + c_ISLNK = 0120000 // Symbolic link + c_ISBLK = 060000 // Block special file + c_ISCHR = 020000 // Character special file + c_ISSOCK = 0140000 // Socket +) + +// Keywords for the PAX Extended Header +const ( + paxAtime = "atime" + paxCharset = "charset" + paxComment = "comment" + paxCtime = "ctime" // please note that ctime is not a valid pax header. + paxGid = "gid" + paxGname = "gname" + paxLinkpath = "linkpath" + paxMtime = "mtime" + paxPath = "path" + paxSize = "size" + paxUid = "uid" + paxUname = "uname" + paxXattr = "SCHILY.xattr." + paxNone = "" +) + +// FileInfoHeader creates a partially-populated Header from fi. +// If fi describes a symlink, FileInfoHeader records link as the link target. +// If fi describes a directory, a slash is appended to the name. +// Because os.FileInfo's Name method returns only the base name of +// the file it describes, it may be necessary to modify the Name field +// of the returned header to provide the full path name of the file. +func FileInfoHeader(fi os.FileInfo, link string) (*Header, error) { + if fi == nil { + return nil, errors.New("tar: FileInfo is nil") + } + fm := fi.Mode() + h := &Header{ + Name: fi.Name(), + ModTime: fi.ModTime(), + Mode: int64(fm.Perm()), // or'd with c_IS* constants later + } + switch { + case fm.IsRegular(): + h.Mode |= c_ISREG + h.Typeflag = TypeReg + h.Size = fi.Size() + case fi.IsDir(): + h.Typeflag = TypeDir + h.Mode |= c_ISDIR + h.Name += "/" + case fm&os.ModeSymlink != 0: + h.Typeflag = TypeSymlink + h.Mode |= c_ISLNK + h.Linkname = link + case fm&os.ModeDevice != 0: + if fm&os.ModeCharDevice != 0 { + h.Mode |= c_ISCHR + h.Typeflag = TypeChar + } else { + h.Mode |= c_ISBLK + h.Typeflag = TypeBlock + } + case fm&os.ModeNamedPipe != 0: + h.Typeflag = TypeFifo + h.Mode |= c_ISFIFO + case fm&os.ModeSocket != 0: + h.Mode |= c_ISSOCK + default: + return nil, fmt.Errorf("archive/tar: unknown file mode %v", fm) + } + if fm&os.ModeSetuid != 0 { + h.Mode |= c_ISUID + } + if fm&os.ModeSetgid != 0 { + h.Mode |= c_ISGID + } + if fm&os.ModeSticky != 0 { + h.Mode |= c_ISVTX + } + if sysStat != nil { + return h, sysStat(fi, h) + } + return h, nil +} + +var zeroBlock = make([]byte, blockSize) + +// POSIX specifies a sum of the unsigned byte values, but the Sun tar uses signed byte values. +// We compute and return both. +func checksum(header []byte) (unsigned int64, signed int64) { + for i := 0; i < len(header); i++ { + if i == 148 { + // The chksum field (header[148:156]) is special: it should be treated as space bytes. + unsigned += ' ' * 8 + signed += ' ' * 8 + i += 7 + continue + } + unsigned += int64(header[i]) + signed += int64(int8(header[i])) + } + return +} + +type slicer []byte + +func (sp *slicer) next(n int) (b []byte) { + s := *sp + b, *sp = s[0:n], s[n:] + return +} + +func isASCII(s string) bool { + for _, c := range s { + if c >= 0x80 { + return false + } + } + return true +} + +func toASCII(s string) string { + if isASCII(s) { + return s + } + var buf bytes.Buffer + for _, c := range s { + if c < 0x80 { + buf.WriteByte(byte(c)) + } + } + return buf.String() +} diff --git a/archive/tar/example_test.go b/archive/tar/example_test.go new file mode 100644 index 0000000..2317f44 --- /dev/null +++ b/archive/tar/example_test.go @@ -0,0 +1,80 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tar_test + +import ( + "archive/tar" + "bytes" + "fmt" + "io" + "log" + "os" +) + +func Example() { + // Create a buffer to write our archive to. + buf := new(bytes.Buffer) + + // Create a new tar archive. + tw := tar.NewWriter(buf) + + // Add some files to the archive. + var files = []struct { + Name, Body string + }{ + {"readme.txt", "This archive contains some text files."}, + {"gopher.txt", "Gopher names:\nGeorge\nGeoffrey\nGonzo"}, + {"todo.txt", "Get animal handling licence."}, + } + for _, file := range files { + hdr := &tar.Header{ + Name: file.Name, + Mode: 0600, + Size: int64(len(file.Body)), + } + if err := tw.WriteHeader(hdr); err != nil { + log.Fatalln(err) + } + if _, err := tw.Write([]byte(file.Body)); err != nil { + log.Fatalln(err) + } + } + // Make sure to check the error on Close. + if err := tw.Close(); err != nil { + log.Fatalln(err) + } + + // Open the tar archive for reading. + r := bytes.NewReader(buf.Bytes()) + tr := tar.NewReader(r) + + // Iterate through the files in the archive. + for { + hdr, err := tr.Next() + if err == io.EOF { + // end of tar archive + break + } + if err != nil { + log.Fatalln(err) + } + fmt.Printf("Contents of %s:\n", hdr.Name) + if _, err := io.Copy(os.Stdout, tr); err != nil { + log.Fatalln(err) + } + fmt.Println() + } + + // Output: + // Contents of readme.txt: + // This archive contains some text files. + // Contents of gopher.txt: + // Gopher names: + // George + // Geoffrey + // Gonzo + // Contents of todo.txt: + // Get animal handling licence. +} diff --git a/archive/tar/reader.go b/archive/tar/reader.go new file mode 100644 index 0000000..abd8f14 --- /dev/null +++ b/archive/tar/reader.go @@ -0,0 +1,822 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tar + +// TODO(dsymonds): +// - pax extensions + +import ( + "bytes" + "errors" + "io" + "io/ioutil" + "os" + "strconv" + "strings" + "time" +) + +var ( + ErrHeader = errors.New("archive/tar: invalid tar header") +) + +const maxNanoSecondIntSize = 9 + +// A Reader provides sequential access to the contents of a tar archive. +// A tar archive consists of a sequence of files. +// The Next method advances to the next file in the archive (including the first), +// and then it can be treated as an io.Reader to access the file's data. +type Reader struct { + r io.Reader + err error + pad int64 // amount of padding (ignored) after current file entry + curr numBytesReader // reader for current file entry + hdrBuff [blockSize]byte // buffer to use in readHeader +} + +// A numBytesReader is an io.Reader with a numBytes method, returning the number +// of bytes remaining in the underlying encoded data. +type numBytesReader interface { + io.Reader + numBytes() int64 +} + +// A regFileReader is a numBytesReader for reading file data from a tar archive. +type regFileReader struct { + r io.Reader // underlying reader + nb int64 // number of unread bytes for current file entry +} + +// A sparseFileReader is a numBytesReader for reading sparse file data from a tar archive. +type sparseFileReader struct { + rfr *regFileReader // reads the sparse-encoded file data + sp []sparseEntry // the sparse map for the file + pos int64 // keeps track of file position + tot int64 // total size of the file +} + +// Keywords for GNU sparse files in a PAX extended header +const ( + paxGNUSparseNumBlocks = "GNU.sparse.numblocks" + paxGNUSparseOffset = "GNU.sparse.offset" + paxGNUSparseNumBytes = "GNU.sparse.numbytes" + paxGNUSparseMap = "GNU.sparse.map" + paxGNUSparseName = "GNU.sparse.name" + paxGNUSparseMajor = "GNU.sparse.major" + paxGNUSparseMinor = "GNU.sparse.minor" + paxGNUSparseSize = "GNU.sparse.size" + paxGNUSparseRealSize = "GNU.sparse.realsize" +) + +// Keywords for old GNU sparse headers +const ( + oldGNUSparseMainHeaderOffset = 386 + oldGNUSparseMainHeaderIsExtendedOffset = 482 + oldGNUSparseMainHeaderNumEntries = 4 + oldGNUSparseExtendedHeaderIsExtendedOffset = 504 + oldGNUSparseExtendedHeaderNumEntries = 21 + oldGNUSparseOffsetSize = 12 + oldGNUSparseNumBytesSize = 12 +) + +// NewReader creates a new Reader reading from r. +func NewReader(r io.Reader) *Reader { return &Reader{r: r} } + +// Next advances to the next entry in the tar archive. +// +// io.EOF is returned at the end of the input. +func (tr *Reader) Next() (*Header, error) { + var hdr *Header + if tr.err == nil { + tr.skipUnread() + } + if tr.err != nil { + return hdr, tr.err + } + hdr = tr.readHeader() + if hdr == nil { + return hdr, tr.err + } + // Check for PAX/GNU header. + switch hdr.Typeflag { + case TypeXHeader: + // PAX extended header + headers, err := parsePAX(tr) + if err != nil { + return nil, err + } + // We actually read the whole file, + // but this skips alignment padding + tr.skipUnread() + hdr = tr.readHeader() + mergePAX(hdr, headers) + + // Check for a PAX format sparse file + sp, err := tr.checkForGNUSparsePAXHeaders(hdr, headers) + if err != nil { + tr.err = err + return nil, err + } + if sp != nil { + // Current file is a PAX format GNU sparse file. + // Set the current file reader to a sparse file reader. + tr.curr = &sparseFileReader{rfr: tr.curr.(*regFileReader), sp: sp, tot: hdr.Size} + } + return hdr, nil + case TypeGNULongName: + // We have a GNU long name header. Its contents are the real file name. + realname, err := ioutil.ReadAll(tr) + if err != nil { + return nil, err + } + hdr, err := tr.Next() + hdr.Name = cString(realname) + return hdr, err + case TypeGNULongLink: + // We have a GNU long link header. + realname, err := ioutil.ReadAll(tr) + if err != nil { + return nil, err + } + hdr, err := tr.Next() + hdr.Linkname = cString(realname) + return hdr, err + } + return hdr, tr.err +} + +// checkForGNUSparsePAXHeaders checks the PAX headers for GNU sparse headers. If they are found, then +// this function reads the sparse map and returns it. Unknown sparse formats are ignored, causing the file to +// be treated as a regular file. +func (tr *Reader) checkForGNUSparsePAXHeaders(hdr *Header, headers map[string]string) ([]sparseEntry, error) { + var sparseFormat string + + // Check for sparse format indicators + major, majorOk := headers[paxGNUSparseMajor] + minor, minorOk := headers[paxGNUSparseMinor] + sparseName, sparseNameOk := headers[paxGNUSparseName] + _, sparseMapOk := headers[paxGNUSparseMap] + sparseSize, sparseSizeOk := headers[paxGNUSparseSize] + sparseRealSize, sparseRealSizeOk := headers[paxGNUSparseRealSize] + + // Identify which, if any, sparse format applies from which PAX headers are set + if majorOk && minorOk { + sparseFormat = major + "." + minor + } else if sparseNameOk && sparseMapOk { + sparseFormat = "0.1" + } else if sparseSizeOk { + sparseFormat = "0.0" + } else { + // Not a PAX format GNU sparse file. + return nil, nil + } + + // Check for unknown sparse format + if sparseFormat != "0.0" && sparseFormat != "0.1" && sparseFormat != "1.0" { + return nil, nil + } + + // Update hdr from GNU sparse PAX headers + if sparseNameOk { + hdr.Name = sparseName + } + if sparseSizeOk { + realSize, err := strconv.ParseInt(sparseSize, 10, 0) + if err != nil { + return nil, ErrHeader + } + hdr.Size = realSize + } else if sparseRealSizeOk { + realSize, err := strconv.ParseInt(sparseRealSize, 10, 0) + if err != nil { + return nil, ErrHeader + } + hdr.Size = realSize + } + + // Set up the sparse map, according to the particular sparse format in use + var sp []sparseEntry + var err error + switch sparseFormat { + case "0.0", "0.1": + sp, err = readGNUSparseMap0x1(headers) + case "1.0": + sp, err = readGNUSparseMap1x0(tr.curr) + } + return sp, err +} + +// mergePAX merges well known headers according to PAX standard. +// In general headers with the same name as those found +// in the header struct overwrite those found in the header +// struct with higher precision or longer values. Esp. useful +// for name and linkname fields. +func mergePAX(hdr *Header, headers map[string]string) error { + for k, v := range headers { + switch k { + case paxPath: + hdr.Name = v + case paxLinkpath: + hdr.Linkname = v + case paxGname: + hdr.Gname = v + case paxUname: + hdr.Uname = v + case paxUid: + uid, err := strconv.ParseInt(v, 10, 0) + if err != nil { + return err + } + hdr.Uid = int(uid) + case paxGid: + gid, err := strconv.ParseInt(v, 10, 0) + if err != nil { + return err + } + hdr.Gid = int(gid) + case paxAtime: + t, err := parsePAXTime(v) + if err != nil { + return err + } + hdr.AccessTime = t + case paxMtime: + t, err := parsePAXTime(v) + if err != nil { + return err + } + hdr.ModTime = t + case paxCtime: + t, err := parsePAXTime(v) + if err != nil { + return err + } + hdr.ChangeTime = t + case paxSize: + size, err := strconv.ParseInt(v, 10, 0) + if err != nil { + return err + } + hdr.Size = int64(size) + default: + if strings.HasPrefix(k, paxXattr) { + if hdr.Xattrs == nil { + hdr.Xattrs = make(map[string]string) + } + hdr.Xattrs[k[len(paxXattr):]] = v + } + } + } + return nil +} + +// parsePAXTime takes a string of the form %d.%d as described in +// the PAX specification. +func parsePAXTime(t string) (time.Time, error) { + buf := []byte(t) + pos := bytes.IndexByte(buf, '.') + var seconds, nanoseconds int64 + var err error + if pos == -1 { + seconds, err = strconv.ParseInt(t, 10, 0) + if err != nil { + return time.Time{}, err + } + } else { + seconds, err = strconv.ParseInt(string(buf[:pos]), 10, 0) + if err != nil { + return time.Time{}, err + } + nano_buf := string(buf[pos+1:]) + // Pad as needed before converting to a decimal. + // For example .030 -> .030000000 -> 30000000 nanoseconds + if len(nano_buf) < maxNanoSecondIntSize { + // Right pad + nano_buf += strings.Repeat("0", maxNanoSecondIntSize-len(nano_buf)) + } else if len(nano_buf) > maxNanoSecondIntSize { + // Right truncate + nano_buf = nano_buf[:maxNanoSecondIntSize] + } + nanoseconds, err = strconv.ParseInt(string(nano_buf), 10, 0) + if err != nil { + return time.Time{}, err + } + } + ts := time.Unix(seconds, nanoseconds) + return ts, nil +} + +// parsePAX parses PAX headers. +// If an extended header (type 'x') is invalid, ErrHeader is returned +func parsePAX(r io.Reader) (map[string]string, error) { + buf, err := ioutil.ReadAll(r) + if err != nil { + return nil, err + } + + // For GNU PAX sparse format 0.0 support. + // This function transforms the sparse format 0.0 headers into sparse format 0.1 headers. + var sparseMap bytes.Buffer + + headers := make(map[string]string) + // Each record is constructed as + // "%d %s=%s\n", length, keyword, value + for len(buf) > 0 { + // or the header was empty to start with. + var sp int + // The size field ends at the first space. + sp = bytes.IndexByte(buf, ' ') + if sp == -1 { + return nil, ErrHeader + } + // Parse the first token as a decimal integer. + n, err := strconv.ParseInt(string(buf[:sp]), 10, 0) + if err != nil { + return nil, ErrHeader + } + // Extract everything between the decimal and the n -1 on the + // beginning to eat the ' ', -1 on the end to skip the newline. + var record []byte + record, buf = buf[sp+1:n-1], buf[n:] + // The first equals is guaranteed to mark the end of the key. + // Everything else is value. + eq := bytes.IndexByte(record, '=') + if eq == -1 { + return nil, ErrHeader + } + key, value := record[:eq], record[eq+1:] + + keyStr := string(key) + if keyStr == paxGNUSparseOffset || keyStr == paxGNUSparseNumBytes { + // GNU sparse format 0.0 special key. Write to sparseMap instead of using the headers map. + sparseMap.Write(value) + sparseMap.Write([]byte{','}) + } else { + // Normal key. Set the value in the headers map. + headers[keyStr] = string(value) + } + } + if sparseMap.Len() != 0 { + // Add sparse info to headers, chopping off the extra comma + sparseMap.Truncate(sparseMap.Len() - 1) + headers[paxGNUSparseMap] = sparseMap.String() + } + return headers, nil +} + +// cString parses bytes as a NUL-terminated C-style string. +// If a NUL byte is not found then the whole slice is returned as a string. +func cString(b []byte) string { + n := 0 + for n < len(b) && b[n] != 0 { + n++ + } + return string(b[0:n]) +} + +func (tr *Reader) octal(b []byte) int64 { + // Check for binary format first. + if len(b) > 0 && b[0]&0x80 != 0 { + var x int64 + for i, c := range b { + if i == 0 { + c &= 0x7f // ignore signal bit in first byte + } + x = x<<8 | int64(c) + } + return x + } + + // Because unused fields are filled with NULs, we need + // to skip leading NULs. Fields may also be padded with + // spaces or NULs. + // So we remove leading and trailing NULs and spaces to + // be sure. + b = bytes.Trim(b, " \x00") + + if len(b) == 0 { + return 0 + } + x, err := strconv.ParseUint(cString(b), 8, 64) + if err != nil { + tr.err = err + } + return int64(x) +} + +// skipUnread skips any unread bytes in the existing file entry, as well as any alignment padding. +func (tr *Reader) skipUnread() { + nr := tr.numBytes() + tr.pad // number of bytes to skip + tr.curr, tr.pad = nil, 0 + if sr, ok := tr.r.(io.Seeker); ok { + if _, err := sr.Seek(nr, os.SEEK_CUR); err == nil { + return + } + } + _, tr.err = io.CopyN(ioutil.Discard, tr.r, nr) +} + +func (tr *Reader) verifyChecksum(header []byte) bool { + if tr.err != nil { + return false + } + + given := tr.octal(header[148:156]) + unsigned, signed := checksum(header) + return given == unsigned || given == signed +} + +func (tr *Reader) readHeader() *Header { + header := tr.hdrBuff[:] + copy(header, zeroBlock) + + if _, tr.err = io.ReadFull(tr.r, header); tr.err != nil { + return nil + } + + // Two blocks of zero bytes marks the end of the archive. + if bytes.Equal(header, zeroBlock[0:blockSize]) { + if _, tr.err = io.ReadFull(tr.r, header); tr.err != nil { + return nil + } + if bytes.Equal(header, zeroBlock[0:blockSize]) { + tr.err = io.EOF + } else { + tr.err = ErrHeader // zero block and then non-zero block + } + return nil + } + + if !tr.verifyChecksum(header) { + tr.err = ErrHeader + return nil + } + + // Unpack + hdr := new(Header) + s := slicer(header) + + hdr.Name = cString(s.next(100)) + hdr.Mode = tr.octal(s.next(8)) + hdr.Uid = int(tr.octal(s.next(8))) + hdr.Gid = int(tr.octal(s.next(8))) + hdr.Size = tr.octal(s.next(12)) + hdr.ModTime = time.Unix(tr.octal(s.next(12)), 0) + s.next(8) // chksum + hdr.Typeflag = s.next(1)[0] + hdr.Linkname = cString(s.next(100)) + + // The remainder of the header depends on the value of magic. + // The original (v7) version of tar had no explicit magic field, + // so its magic bytes, like the rest of the block, are NULs. + magic := string(s.next(8)) // contains version field as well. + var format string + switch { + case magic[:6] == "ustar\x00": // POSIX tar (1003.1-1988) + if string(header[508:512]) == "tar\x00" { + format = "star" + } else { + format = "posix" + } + case magic == "ustar \x00": // old GNU tar + format = "gnu" + } + + switch format { + case "posix", "gnu", "star": + hdr.Uname = cString(s.next(32)) + hdr.Gname = cString(s.next(32)) + devmajor := s.next(8) + devminor := s.next(8) + if hdr.Typeflag == TypeChar || hdr.Typeflag == TypeBlock { + hdr.Devmajor = tr.octal(devmajor) + hdr.Devminor = tr.octal(devminor) + } + var prefix string + switch format { + case "posix", "gnu": + prefix = cString(s.next(155)) + case "star": + prefix = cString(s.next(131)) + hdr.AccessTime = time.Unix(tr.octal(s.next(12)), 0) + hdr.ChangeTime = time.Unix(tr.octal(s.next(12)), 0) + } + if len(prefix) > 0 { + hdr.Name = prefix + "/" + hdr.Name + } + } + + if tr.err != nil { + tr.err = ErrHeader + return nil + } + + // Maximum value of hdr.Size is 64 GB (12 octal digits), + // so there's no risk of int64 overflowing. + nb := int64(hdr.Size) + tr.pad = -nb & (blockSize - 1) // blockSize is a power of two + + // Set the current file reader. + tr.curr = ®FileReader{r: tr.r, nb: nb} + + // Check for old GNU sparse format entry. + if hdr.Typeflag == TypeGNUSparse { + // Get the real size of the file. + hdr.Size = tr.octal(header[483:495]) + + // Read the sparse map. + sp := tr.readOldGNUSparseMap(header) + if tr.err != nil { + return nil + } + // Current file is a GNU sparse file. Update the current file reader. + tr.curr = &sparseFileReader{rfr: tr.curr.(*regFileReader), sp: sp, tot: hdr.Size} + } + + return hdr +} + +// A sparseEntry holds a single entry in a sparse file's sparse map. +// A sparse entry indicates the offset and size in a sparse file of a +// block of data. +type sparseEntry struct { + offset int64 + numBytes int64 +} + +// readOldGNUSparseMap reads the sparse map as stored in the old GNU sparse format. +// The sparse map is stored in the tar header if it's small enough. If it's larger than four entries, +// then one or more extension headers are used to store the rest of the sparse map. +func (tr *Reader) readOldGNUSparseMap(header []byte) []sparseEntry { + isExtended := header[oldGNUSparseMainHeaderIsExtendedOffset] != 0 + spCap := oldGNUSparseMainHeaderNumEntries + if isExtended { + spCap += oldGNUSparseExtendedHeaderNumEntries + } + sp := make([]sparseEntry, 0, spCap) + s := slicer(header[oldGNUSparseMainHeaderOffset:]) + + // Read the four entries from the main tar header + for i := 0; i < oldGNUSparseMainHeaderNumEntries; i++ { + offset := tr.octal(s.next(oldGNUSparseOffsetSize)) + numBytes := tr.octal(s.next(oldGNUSparseNumBytesSize)) + if tr.err != nil { + tr.err = ErrHeader + return nil + } + if offset == 0 && numBytes == 0 { + break + } + sp = append(sp, sparseEntry{offset: offset, numBytes: numBytes}) + } + + for isExtended { + // There are more entries. Read an extension header and parse its entries. + sparseHeader := make([]byte, blockSize) + if _, tr.err = io.ReadFull(tr.r, sparseHeader); tr.err != nil { + return nil + } + isExtended = sparseHeader[oldGNUSparseExtendedHeaderIsExtendedOffset] != 0 + s = slicer(sparseHeader) + for i := 0; i < oldGNUSparseExtendedHeaderNumEntries; i++ { + offset := tr.octal(s.next(oldGNUSparseOffsetSize)) + numBytes := tr.octal(s.next(oldGNUSparseNumBytesSize)) + if tr.err != nil { + tr.err = ErrHeader + return nil + } + if offset == 0 && numBytes == 0 { + break + } + sp = append(sp, sparseEntry{offset: offset, numBytes: numBytes}) + } + } + return sp +} + +// readGNUSparseMap1x0 reads the sparse map as stored in GNU's PAX sparse format version 1.0. +// The sparse map is stored just before the file data and padded out to the nearest block boundary. +func readGNUSparseMap1x0(r io.Reader) ([]sparseEntry, error) { + buf := make([]byte, 2*blockSize) + sparseHeader := buf[:blockSize] + + // readDecimal is a helper function to read a decimal integer from the sparse map + // while making sure to read from the file in blocks of size blockSize + readDecimal := func() (int64, error) { + // Look for newline + nl := bytes.IndexByte(sparseHeader, '\n') + if nl == -1 { + if len(sparseHeader) >= blockSize { + // This is an error + return 0, ErrHeader + } + oldLen := len(sparseHeader) + newLen := oldLen + blockSize + if cap(sparseHeader) < newLen { + // There's more header, but we need to make room for the next block + copy(buf, sparseHeader) + sparseHeader = buf[:newLen] + } else { + // There's more header, and we can just reslice + sparseHeader = sparseHeader[:newLen] + } + + // Now that sparseHeader is large enough, read next block + if _, err := io.ReadFull(r, sparseHeader[oldLen:newLen]); err != nil { + return 0, err + } + + // Look for a newline in the new data + nl = bytes.IndexByte(sparseHeader[oldLen:newLen], '\n') + if nl == -1 { + // This is an error + return 0, ErrHeader + } + nl += oldLen // We want the position from the beginning + } + // Now that we've found a newline, read a number + n, err := strconv.ParseInt(string(sparseHeader[:nl]), 10, 0) + if err != nil { + return 0, ErrHeader + } + + // Update sparseHeader to consume this number + sparseHeader = sparseHeader[nl+1:] + return n, nil + } + + // Read the first block + if _, err := io.ReadFull(r, sparseHeader); err != nil { + return nil, err + } + + // The first line contains the number of entries + numEntries, err := readDecimal() + if err != nil { + return nil, err + } + + // Read all the entries + sp := make([]sparseEntry, 0, numEntries) + for i := int64(0); i < numEntries; i++ { + // Read the offset + offset, err := readDecimal() + if err != nil { + return nil, err + } + // Read numBytes + numBytes, err := readDecimal() + if err != nil { + return nil, err + } + + sp = append(sp, sparseEntry{offset: offset, numBytes: numBytes}) + } + + return sp, nil +} + +// readGNUSparseMap0x1 reads the sparse map as stored in GNU's PAX sparse format version 0.1. +// The sparse map is stored in the PAX headers. +func readGNUSparseMap0x1(headers map[string]string) ([]sparseEntry, error) { + // Get number of entries + numEntriesStr, ok := headers[paxGNUSparseNumBlocks] + if !ok { + return nil, ErrHeader + } + numEntries, err := strconv.ParseInt(numEntriesStr, 10, 0) + if err != nil { + return nil, ErrHeader + } + + sparseMap := strings.Split(headers[paxGNUSparseMap], ",") + + // There should be two numbers in sparseMap for each entry + if int64(len(sparseMap)) != 2*numEntries { + return nil, ErrHeader + } + + // Loop through the entries in the sparse map + sp := make([]sparseEntry, 0, numEntries) + for i := int64(0); i < numEntries; i++ { + offset, err := strconv.ParseInt(sparseMap[2*i], 10, 0) + if err != nil { + return nil, ErrHeader + } + numBytes, err := strconv.ParseInt(sparseMap[2*i+1], 10, 0) + if err != nil { + return nil, ErrHeader + } + sp = append(sp, sparseEntry{offset: offset, numBytes: numBytes}) + } + + return sp, nil +} + +// numBytes returns the number of bytes left to read in the current file's entry +// in the tar archive, or 0 if there is no current file. +func (tr *Reader) numBytes() int64 { + if tr.curr == nil { + // No current file, so no bytes + return 0 + } + return tr.curr.numBytes() +} + +// Read reads from the current entry in the tar archive. +// It returns 0, io.EOF when it reaches the end of that entry, +// until Next is called to advance to the next entry. +func (tr *Reader) Read(b []byte) (n int, err error) { + if tr.curr == nil { + return 0, io.EOF + } + n, err = tr.curr.Read(b) + if err != nil && err != io.EOF { + tr.err = err + } + return +} + +func (rfr *regFileReader) Read(b []byte) (n int, err error) { + if rfr.nb == 0 { + // file consumed + return 0, io.EOF + } + if int64(len(b)) > rfr.nb { + b = b[0:rfr.nb] + } + n, err = rfr.r.Read(b) + rfr.nb -= int64(n) + + if err == io.EOF && rfr.nb > 0 { + err = io.ErrUnexpectedEOF + } + return +} + +// numBytes returns the number of bytes left to read in the file's data in the tar archive. +func (rfr *regFileReader) numBytes() int64 { + return rfr.nb +} + +// readHole reads a sparse file hole ending at offset toOffset +func (sfr *sparseFileReader) readHole(b []byte, toOffset int64) int { + n64 := toOffset - sfr.pos + if n64 > int64(len(b)) { + n64 = int64(len(b)) + } + n := int(n64) + for i := 0; i < n; i++ { + b[i] = 0 + } + sfr.pos += n64 + return n +} + +// Read reads the sparse file data in expanded form. +func (sfr *sparseFileReader) Read(b []byte) (n int, err error) { + if len(sfr.sp) == 0 { + // No more data fragments to read from. + if sfr.pos < sfr.tot { + // We're in the last hole + n = sfr.readHole(b, sfr.tot) + return + } + // Otherwise, we're at the end of the file + return 0, io.EOF + } + if sfr.pos < sfr.sp[0].offset { + // We're in a hole + n = sfr.readHole(b, sfr.sp[0].offset) + return + } + + // We're not in a hole, so we'll read from the next data fragment + posInFragment := sfr.pos - sfr.sp[0].offset + bytesLeft := sfr.sp[0].numBytes - posInFragment + if int64(len(b)) > bytesLeft { + b = b[0:bytesLeft] + } + + n, err = sfr.rfr.Read(b) + sfr.pos += int64(n) + + if int64(n) == bytesLeft { + // We're done with this fragment + sfr.sp = sfr.sp[1:] + } + + if err == io.EOF && sfr.pos < sfr.tot { + // We reached the end of the last fragment's data, but there's a final hole + err = nil + } + return +} + +// numBytes returns the number of bytes left to read in the sparse file's +// sparse-encoded data in the tar archive. +func (sfr *sparseFileReader) numBytes() int64 { + return sfr.rfr.nb +} diff --git a/archive/tar/reader_test.go b/archive/tar/reader_test.go new file mode 100644 index 0000000..9601ffe --- /dev/null +++ b/archive/tar/reader_test.go @@ -0,0 +1,743 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tar + +import ( + "bytes" + "crypto/md5" + "fmt" + "io" + "io/ioutil" + "os" + "reflect" + "strings" + "testing" + "time" +) + +type untarTest struct { + file string + headers []*Header + cksums []string +} + +var gnuTarTest = &untarTest{ + file: "testdata/gnu.tar", + headers: []*Header{ + { + Name: "small.txt", + Mode: 0640, + Uid: 73025, + Gid: 5000, + Size: 5, + ModTime: time.Unix(1244428340, 0), + Typeflag: '0', + Uname: "dsymonds", + Gname: "eng", + }, + { + Name: "small2.txt", + Mode: 0640, + Uid: 73025, + Gid: 5000, + Size: 11, + ModTime: time.Unix(1244436044, 0), + Typeflag: '0', + Uname: "dsymonds", + Gname: "eng", + }, + }, + cksums: []string{ + "e38b27eaccb4391bdec553a7f3ae6b2f", + "c65bd2e50a56a2138bf1716f2fd56fe9", + }, +} + +var sparseTarTest = &untarTest{ + file: "testdata/sparse-formats.tar", + headers: []*Header{ + { + Name: "sparse-gnu", + Mode: 420, + Uid: 1000, + Gid: 1000, + Size: 200, + ModTime: time.Unix(1392395740, 0), + Typeflag: 0x53, + Linkname: "", + Uname: "david", + Gname: "david", + Devmajor: 0, + Devminor: 0, + }, + { + Name: "sparse-posix-0.0", + Mode: 420, + Uid: 1000, + Gid: 1000, + Size: 200, + ModTime: time.Unix(1392342187, 0), + Typeflag: 0x30, + Linkname: "", + Uname: "david", + Gname: "david", + Devmajor: 0, + Devminor: 0, + }, + { + Name: "sparse-posix-0.1", + Mode: 420, + Uid: 1000, + Gid: 1000, + Size: 200, + ModTime: time.Unix(1392340456, 0), + Typeflag: 0x30, + Linkname: "", + Uname: "david", + Gname: "david", + Devmajor: 0, + Devminor: 0, + }, + { + Name: "sparse-posix-1.0", + Mode: 420, + Uid: 1000, + Gid: 1000, + Size: 200, + ModTime: time.Unix(1392337404, 0), + Typeflag: 0x30, + Linkname: "", + Uname: "david", + Gname: "david", + Devmajor: 0, + Devminor: 0, + }, + { + Name: "end", + Mode: 420, + Uid: 1000, + Gid: 1000, + Size: 4, + ModTime: time.Unix(1392398319, 0), + Typeflag: 0x30, + Linkname: "", + Uname: "david", + Gname: "david", + Devmajor: 0, + Devminor: 0, + }, + }, + cksums: []string{ + "6f53234398c2449fe67c1812d993012f", + "6f53234398c2449fe67c1812d993012f", + "6f53234398c2449fe67c1812d993012f", + "6f53234398c2449fe67c1812d993012f", + "b0061974914468de549a2af8ced10316", + }, +} + +var untarTests = []*untarTest{ + gnuTarTest, + sparseTarTest, + { + file: "testdata/star.tar", + headers: []*Header{ + { + Name: "small.txt", + Mode: 0640, + Uid: 73025, + Gid: 5000, + Size: 5, + ModTime: time.Unix(1244592783, 0), + Typeflag: '0', + Uname: "dsymonds", + Gname: "eng", + AccessTime: time.Unix(1244592783, 0), + ChangeTime: time.Unix(1244592783, 0), + }, + { + Name: "small2.txt", + Mode: 0640, + Uid: 73025, + Gid: 5000, + Size: 11, + ModTime: time.Unix(1244592783, 0), + Typeflag: '0', + Uname: "dsymonds", + Gname: "eng", + AccessTime: time.Unix(1244592783, 0), + ChangeTime: time.Unix(1244592783, 0), + }, + }, + }, + { + file: "testdata/v7.tar", + headers: []*Header{ + { + Name: "small.txt", + Mode: 0444, + Uid: 73025, + Gid: 5000, + Size: 5, + ModTime: time.Unix(1244593104, 0), + Typeflag: '\x00', + }, + { + Name: "small2.txt", + Mode: 0444, + Uid: 73025, + Gid: 5000, + Size: 11, + ModTime: time.Unix(1244593104, 0), + Typeflag: '\x00', + }, + }, + }, + { + file: "testdata/pax.tar", + headers: []*Header{ + { + Name: "a/123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100", + Mode: 0664, + Uid: 1000, + Gid: 1000, + Uname: "shane", + Gname: "shane", + Size: 7, + ModTime: time.Unix(1350244992, 23960108), + ChangeTime: time.Unix(1350244992, 23960108), + AccessTime: time.Unix(1350244992, 23960108), + Typeflag: TypeReg, + }, + { + Name: "a/b", + Mode: 0777, + Uid: 1000, + Gid: 1000, + Uname: "shane", + Gname: "shane", + Size: 0, + ModTime: time.Unix(1350266320, 910238425), + ChangeTime: time.Unix(1350266320, 910238425), + AccessTime: time.Unix(1350266320, 910238425), + Typeflag: TypeSymlink, + Linkname: "123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100", + }, + }, + }, + { + file: "testdata/nil-uid.tar", // golang.org/issue/5290 + headers: []*Header{ + { + Name: "P1050238.JPG.log", + Mode: 0664, + Uid: 0, + Gid: 0, + Size: 14, + ModTime: time.Unix(1365454838, 0), + Typeflag: TypeReg, + Linkname: "", + Uname: "eyefi", + Gname: "eyefi", + Devmajor: 0, + Devminor: 0, + }, + }, + }, + { + file: "testdata/xattrs.tar", + headers: []*Header{ + { + Name: "small.txt", + Mode: 0644, + Uid: 1000, + Gid: 10, + Size: 5, + ModTime: time.Unix(1386065770, 448252320), + Typeflag: '0', + Uname: "alex", + Gname: "wheel", + AccessTime: time.Unix(1389782991, 419875220), + ChangeTime: time.Unix(1389782956, 794414986), + Xattrs: map[string]string{ + "user.key": "value", + "user.key2": "value2", + // Interestingly, selinux encodes the terminating null inside the xattr + "security.selinux": "unconfined_u:object_r:default_t:s0\x00", + }, + }, + { + Name: "small2.txt", + Mode: 0644, + Uid: 1000, + Gid: 10, + Size: 11, + ModTime: time.Unix(1386065770, 449252304), + Typeflag: '0', + Uname: "alex", + Gname: "wheel", + AccessTime: time.Unix(1389782991, 419875220), + ChangeTime: time.Unix(1386065770, 449252304), + Xattrs: map[string]string{ + "security.selinux": "unconfined_u:object_r:default_t:s0\x00", + }, + }, + }, + }, +} + +func TestReader(t *testing.T) { +testLoop: + for i, test := range untarTests { + f, err := os.Open(test.file) + if err != nil { + t.Errorf("test %d: Unexpected error: %v", i, err) + continue + } + defer f.Close() + tr := NewReader(f) + for j, header := range test.headers { + hdr, err := tr.Next() + if err != nil || hdr == nil { + t.Errorf("test %d, entry %d: Didn't get entry: %v", i, j, err) + f.Close() + continue testLoop + } + if !reflect.DeepEqual(*hdr, *header) { + t.Errorf("test %d, entry %d: Incorrect header:\nhave %+v\nwant %+v", + i, j, *hdr, *header) + } + } + hdr, err := tr.Next() + if err == io.EOF { + continue testLoop + } + if hdr != nil || err != nil { + t.Errorf("test %d: Unexpected entry or error: hdr=%v err=%v", i, hdr, err) + } + } +} + +func TestPartialRead(t *testing.T) { + f, err := os.Open("testdata/gnu.tar") + if err != nil { + t.Fatalf("Unexpected error: %v", err) + } + defer f.Close() + + tr := NewReader(f) + + // Read the first four bytes; Next() should skip the last byte. + hdr, err := tr.Next() + if err != nil || hdr == nil { + t.Fatalf("Didn't get first file: %v", err) + } + buf := make([]byte, 4) + if _, err := io.ReadFull(tr, buf); err != nil { + t.Fatalf("Unexpected error: %v", err) + } + if expected := []byte("Kilt"); !bytes.Equal(buf, expected) { + t.Errorf("Contents = %v, want %v", buf, expected) + } + + // Second file + hdr, err = tr.Next() + if err != nil || hdr == nil { + t.Fatalf("Didn't get second file: %v", err) + } + buf = make([]byte, 6) + if _, err := io.ReadFull(tr, buf); err != nil { + t.Fatalf("Unexpected error: %v", err) + } + if expected := []byte("Google"); !bytes.Equal(buf, expected) { + t.Errorf("Contents = %v, want %v", buf, expected) + } +} + +func TestIncrementalRead(t *testing.T) { + test := gnuTarTest + f, err := os.Open(test.file) + if err != nil { + t.Fatalf("Unexpected error: %v", err) + } + defer f.Close() + + tr := NewReader(f) + + headers := test.headers + cksums := test.cksums + nread := 0 + + // loop over all files + for ; ; nread++ { + hdr, err := tr.Next() + if hdr == nil || err == io.EOF { + break + } + + // check the header + if !reflect.DeepEqual(*hdr, *headers[nread]) { + t.Errorf("Incorrect header:\nhave %+v\nwant %+v", + *hdr, headers[nread]) + } + + // read file contents in little chunks EOF, + // checksumming all the way + h := md5.New() + rdbuf := make([]uint8, 8) + for { + nr, err := tr.Read(rdbuf) + if err == io.EOF { + break + } + if err != nil { + t.Errorf("Read: unexpected error %v\n", err) + break + } + h.Write(rdbuf[0:nr]) + } + // verify checksum + have := fmt.Sprintf("%x", h.Sum(nil)) + want := cksums[nread] + if want != have { + t.Errorf("Bad checksum on file %s:\nhave %+v\nwant %+v", hdr.Name, have, want) + } + } + if nread != len(headers) { + t.Errorf("Didn't process all files\nexpected: %d\nprocessed %d\n", len(headers), nread) + } +} + +func TestNonSeekable(t *testing.T) { + test := gnuTarTest + f, err := os.Open(test.file) + if err != nil { + t.Fatalf("Unexpected error: %v", err) + } + defer f.Close() + + type readerOnly struct { + io.Reader + } + tr := NewReader(readerOnly{f}) + nread := 0 + + for ; ; nread++ { + _, err := tr.Next() + if err == io.EOF { + break + } + if err != nil { + t.Fatalf("Unexpected error: %v", err) + } + } + + if nread != len(test.headers) { + t.Errorf("Didn't process all files\nexpected: %d\nprocessed %d\n", len(test.headers), nread) + } +} + +func TestParsePAXHeader(t *testing.T) { + paxTests := [][3]string{ + {"a", "a=name", "10 a=name\n"}, // Test case involving multiple acceptable lengths + {"a", "a=name", "9 a=name\n"}, // Test case involving multiple acceptable length + {"mtime", "mtime=1350244992.023960108", "30 mtime=1350244992.023960108\n"}} + for _, test := range paxTests { + key, expected, raw := test[0], test[1], test[2] + reader := bytes.NewReader([]byte(raw)) + headers, err := parsePAX(reader) + if err != nil { + t.Errorf("Couldn't parse correctly formatted headers: %v", err) + continue + } + if strings.EqualFold(headers[key], expected) { + t.Errorf("mtime header incorrectly parsed: got %s, wanted %s", headers[key], expected) + continue + } + trailer := make([]byte, 100) + n, err := reader.Read(trailer) + if err != io.EOF || n != 0 { + t.Error("Buffer wasn't consumed") + } + } + badHeader := bytes.NewReader([]byte("3 somelongkey=")) + if _, err := parsePAX(badHeader); err != ErrHeader { + t.Fatal("Unexpected success when parsing bad header") + } +} + +func TestParsePAXTime(t *testing.T) { + // Some valid PAX time values + timestamps := map[string]time.Time{ + "1350244992.023960108": time.Unix(1350244992, 23960108), // The common case + "1350244992.02396010": time.Unix(1350244992, 23960100), // Lower precision value + "1350244992.0239601089": time.Unix(1350244992, 23960108), // Higher precision value + "1350244992": time.Unix(1350244992, 0), // Low precision value + } + for input, expected := range timestamps { + ts, err := parsePAXTime(input) + if err != nil { + t.Fatal(err) + } + if !ts.Equal(expected) { + t.Fatalf("Time parsing failure %s %s", ts, expected) + } + } +} + +func TestMergePAX(t *testing.T) { + hdr := new(Header) + // Test a string, integer, and time based value. + headers := map[string]string{ + "path": "a/b/c", + "uid": "1000", + "mtime": "1350244992.023960108", + } + err := mergePAX(hdr, headers) + if err != nil { + t.Fatal(err) + } + want := &Header{ + Name: "a/b/c", + Uid: 1000, + ModTime: time.Unix(1350244992, 23960108), + } + if !reflect.DeepEqual(hdr, want) { + t.Errorf("incorrect merge: got %+v, want %+v", hdr, want) + } +} + +func TestSparseEndToEnd(t *testing.T) { + test := sparseTarTest + f, err := os.Open(test.file) + if err != nil { + t.Fatalf("Unexpected error: %v", err) + } + defer f.Close() + + tr := NewReader(f) + + headers := test.headers + cksums := test.cksums + nread := 0 + + // loop over all files + for ; ; nread++ { + hdr, err := tr.Next() + if hdr == nil || err == io.EOF { + break + } + + // check the header + if !reflect.DeepEqual(*hdr, *headers[nread]) { + t.Errorf("Incorrect header:\nhave %+v\nwant %+v", + *hdr, headers[nread]) + } + + // read and checksum the file data + h := md5.New() + _, err = io.Copy(h, tr) + if err != nil { + t.Fatalf("Unexpected error: %v", err) + } + + // verify checksum + have := fmt.Sprintf("%x", h.Sum(nil)) + want := cksums[nread] + if want != have { + t.Errorf("Bad checksum on file %s:\nhave %+v\nwant %+v", hdr.Name, have, want) + } + } + if nread != len(headers) { + t.Errorf("Didn't process all files\nexpected: %d\nprocessed %d\n", len(headers), nread) + } +} + +type sparseFileReadTest struct { + sparseData []byte + sparseMap []sparseEntry + realSize int64 + expected []byte +} + +var sparseFileReadTests = []sparseFileReadTest{ + { + sparseData: []byte("abcde"), + sparseMap: []sparseEntry{ + {offset: 0, numBytes: 2}, + {offset: 5, numBytes: 3}, + }, + realSize: 8, + expected: []byte("ab\x00\x00\x00cde"), + }, + { + sparseData: []byte("abcde"), + sparseMap: []sparseEntry{ + {offset: 0, numBytes: 2}, + {offset: 5, numBytes: 3}, + }, + realSize: 10, + expected: []byte("ab\x00\x00\x00cde\x00\x00"), + }, + { + sparseData: []byte("abcde"), + sparseMap: []sparseEntry{ + {offset: 1, numBytes: 3}, + {offset: 6, numBytes: 2}, + }, + realSize: 8, + expected: []byte("\x00abc\x00\x00de"), + }, + { + sparseData: []byte("abcde"), + sparseMap: []sparseEntry{ + {offset: 1, numBytes: 3}, + {offset: 6, numBytes: 2}, + }, + realSize: 10, + expected: []byte("\x00abc\x00\x00de\x00\x00"), + }, + { + sparseData: []byte(""), + sparseMap: nil, + realSize: 2, + expected: []byte("\x00\x00"), + }, +} + +func TestSparseFileReader(t *testing.T) { + for i, test := range sparseFileReadTests { + r := bytes.NewReader(test.sparseData) + nb := int64(r.Len()) + sfr := &sparseFileReader{ + rfr: ®FileReader{r: r, nb: nb}, + sp: test.sparseMap, + pos: 0, + tot: test.realSize, + } + if sfr.numBytes() != nb { + t.Errorf("test %d: Before reading, sfr.numBytes() = %d, want %d", i, sfr.numBytes(), nb) + } + buf, err := ioutil.ReadAll(sfr) + if err != nil { + t.Errorf("test %d: Unexpected error: %v", i, err) + } + if e := test.expected; !bytes.Equal(buf, e) { + t.Errorf("test %d: Contents = %v, want %v", i, buf, e) + } + if sfr.numBytes() != 0 { + t.Errorf("test %d: After draining the reader, numBytes() was nonzero", i) + } + } +} + +func TestSparseIncrementalRead(t *testing.T) { + sparseMap := []sparseEntry{{10, 2}} + sparseData := []byte("Go") + expected := "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00Go\x00\x00\x00\x00\x00\x00\x00\x00" + + r := bytes.NewReader(sparseData) + nb := int64(r.Len()) + sfr := &sparseFileReader{ + rfr: ®FileReader{r: r, nb: nb}, + sp: sparseMap, + pos: 0, + tot: int64(len(expected)), + } + + // We'll read the data 6 bytes at a time, with a hole of size 10 at + // the beginning and one of size 8 at the end. + var outputBuf bytes.Buffer + buf := make([]byte, 6) + for { + n, err := sfr.Read(buf) + if err == io.EOF { + break + } + if err != nil { + t.Errorf("Read: unexpected error %v\n", err) + } + if n > 0 { + _, err := outputBuf.Write(buf[:n]) + if err != nil { + t.Errorf("Write: unexpected error %v\n", err) + } + } + } + got := outputBuf.String() + if got != expected { + t.Errorf("Contents = %v, want %v", got, expected) + } +} + +func TestReadGNUSparseMap0x1(t *testing.T) { + headers := map[string]string{ + paxGNUSparseNumBlocks: "4", + paxGNUSparseMap: "0,5,10,5,20,5,30,5", + } + expected := []sparseEntry{ + {offset: 0, numBytes: 5}, + {offset: 10, numBytes: 5}, + {offset: 20, numBytes: 5}, + {offset: 30, numBytes: 5}, + } + + sp, err := readGNUSparseMap0x1(headers) + if err != nil { + t.Errorf("Unexpected error: %v", err) + } + if !reflect.DeepEqual(sp, expected) { + t.Errorf("Incorrect sparse map: got %v, wanted %v", sp, expected) + } +} + +func TestReadGNUSparseMap1x0(t *testing.T) { + // This test uses lots of holes so the sparse header takes up more than two blocks + numEntries := 100 + expected := make([]sparseEntry, 0, numEntries) + sparseMap := new(bytes.Buffer) + + fmt.Fprintf(sparseMap, "%d\n", numEntries) + for i := 0; i < numEntries; i++ { + offset := int64(2048 * i) + numBytes := int64(1024) + expected = append(expected, sparseEntry{offset: offset, numBytes: numBytes}) + fmt.Fprintf(sparseMap, "%d\n%d\n", offset, numBytes) + } + + // Make the header the smallest multiple of blockSize that fits the sparseMap + headerBlocks := (sparseMap.Len() + blockSize - 1) / blockSize + bufLen := blockSize * headerBlocks + buf := make([]byte, bufLen) + copy(buf, sparseMap.Bytes()) + + // Get an reader to read the sparse map + r := bytes.NewReader(buf) + + // Read the sparse map + sp, err := readGNUSparseMap1x0(r) + if err != nil { + t.Errorf("Unexpected error: %v", err) + } + if !reflect.DeepEqual(sp, expected) { + t.Errorf("Incorrect sparse map: got %v, wanted %v", sp, expected) + } +} + +func TestUninitializedRead(t *testing.T) { + test := gnuTarTest + f, err := os.Open(test.file) + if err != nil { + t.Fatalf("Unexpected error: %v", err) + } + defer f.Close() + + tr := NewReader(f) + _, err = tr.Read([]byte{}) + if err == nil || err != io.EOF { + t.Errorf("Unexpected error: %v, wanted %v", err, io.EOF) + } + +} diff --git a/archive/tar/stat_atim.go b/archive/tar/stat_atim.go new file mode 100644 index 0000000..cf9cc79 --- /dev/null +++ b/archive/tar/stat_atim.go @@ -0,0 +1,20 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// +build linux dragonfly openbsd solaris + +package tar + +import ( + "syscall" + "time" +) + +func statAtime(st *syscall.Stat_t) time.Time { + return time.Unix(st.Atim.Unix()) +} + +func statCtime(st *syscall.Stat_t) time.Time { + return time.Unix(st.Ctim.Unix()) +} diff --git a/archive/tar/stat_atimespec.go b/archive/tar/stat_atimespec.go new file mode 100644 index 0000000..6f17dbe --- /dev/null +++ b/archive/tar/stat_atimespec.go @@ -0,0 +1,20 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// +build darwin freebsd netbsd + +package tar + +import ( + "syscall" + "time" +) + +func statAtime(st *syscall.Stat_t) time.Time { + return time.Unix(st.Atimespec.Unix()) +} + +func statCtime(st *syscall.Stat_t) time.Time { + return time.Unix(st.Ctimespec.Unix()) +} diff --git a/archive/tar/stat_unix.go b/archive/tar/stat_unix.go new file mode 100644 index 0000000..cb843db --- /dev/null +++ b/archive/tar/stat_unix.go @@ -0,0 +1,32 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// +build linux darwin dragonfly freebsd openbsd netbsd solaris + +package tar + +import ( + "os" + "syscall" +) + +func init() { + sysStat = statUnix +} + +func statUnix(fi os.FileInfo, h *Header) error { + sys, ok := fi.Sys().(*syscall.Stat_t) + if !ok { + return nil + } + h.Uid = int(sys.Uid) + h.Gid = int(sys.Gid) + // TODO(bradfitz): populate username & group. os/user + // doesn't cache LookupId lookups, and lacks group + // lookup functions. + h.AccessTime = statAtime(sys) + h.ChangeTime = statCtime(sys) + // TODO(bradfitz): major/minor device numbers? + return nil +} diff --git a/archive/tar/tar_test.go b/archive/tar/tar_test.go new file mode 100644 index 0000000..ed333f3 --- /dev/null +++ b/archive/tar/tar_test.go @@ -0,0 +1,284 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tar + +import ( + "bytes" + "io/ioutil" + "os" + "path" + "reflect" + "strings" + "testing" + "time" +) + +func TestFileInfoHeader(t *testing.T) { + fi, err := os.Stat("testdata/small.txt") + if err != nil { + t.Fatal(err) + } + h, err := FileInfoHeader(fi, "") + if err != nil { + t.Fatalf("FileInfoHeader: %v", err) + } + if g, e := h.Name, "small.txt"; g != e { + t.Errorf("Name = %q; want %q", g, e) + } + if g, e := h.Mode, int64(fi.Mode().Perm())|c_ISREG; g != e { + t.Errorf("Mode = %#o; want %#o", g, e) + } + if g, e := h.Size, int64(5); g != e { + t.Errorf("Size = %v; want %v", g, e) + } + if g, e := h.ModTime, fi.ModTime(); !g.Equal(e) { + t.Errorf("ModTime = %v; want %v", g, e) + } + // FileInfoHeader should error when passing nil FileInfo + if _, err := FileInfoHeader(nil, ""); err == nil { + t.Fatalf("Expected error when passing nil to FileInfoHeader") + } +} + +func TestFileInfoHeaderDir(t *testing.T) { + fi, err := os.Stat("testdata") + if err != nil { + t.Fatal(err) + } + h, err := FileInfoHeader(fi, "") + if err != nil { + t.Fatalf("FileInfoHeader: %v", err) + } + if g, e := h.Name, "testdata/"; g != e { + t.Errorf("Name = %q; want %q", g, e) + } + // Ignoring c_ISGID for golang.org/issue/4867 + if g, e := h.Mode&^c_ISGID, int64(fi.Mode().Perm())|c_ISDIR; g != e { + t.Errorf("Mode = %#o; want %#o", g, e) + } + if g, e := h.Size, int64(0); g != e { + t.Errorf("Size = %v; want %v", g, e) + } + if g, e := h.ModTime, fi.ModTime(); !g.Equal(e) { + t.Errorf("ModTime = %v; want %v", g, e) + } +} + +func TestFileInfoHeaderSymlink(t *testing.T) { + h, err := FileInfoHeader(symlink{}, "some-target") + if err != nil { + t.Fatal(err) + } + if g, e := h.Name, "some-symlink"; g != e { + t.Errorf("Name = %q; want %q", g, e) + } + if g, e := h.Linkname, "some-target"; g != e { + t.Errorf("Linkname = %q; want %q", g, e) + } +} + +type symlink struct{} + +func (symlink) Name() string { return "some-symlink" } +func (symlink) Size() int64 { return 0 } +func (symlink) Mode() os.FileMode { return os.ModeSymlink } +func (symlink) ModTime() time.Time { return time.Time{} } +func (symlink) IsDir() bool { return false } +func (symlink) Sys() interface{} { return nil } + +func TestRoundTrip(t *testing.T) { + data := []byte("some file contents") + + var b bytes.Buffer + tw := NewWriter(&b) + hdr := &Header{ + Name: "file.txt", + Uid: 1 << 21, // too big for 8 octal digits + Size: int64(len(data)), + ModTime: time.Now(), + } + // tar only supports second precision. + hdr.ModTime = hdr.ModTime.Add(-time.Duration(hdr.ModTime.Nanosecond()) * time.Nanosecond) + if err := tw.WriteHeader(hdr); err != nil { + t.Fatalf("tw.WriteHeader: %v", err) + } + if _, err := tw.Write(data); err != nil { + t.Fatalf("tw.Write: %v", err) + } + if err := tw.Close(); err != nil { + t.Fatalf("tw.Close: %v", err) + } + + // Read it back. + tr := NewReader(&b) + rHdr, err := tr.Next() + if err != nil { + t.Fatalf("tr.Next: %v", err) + } + if !reflect.DeepEqual(rHdr, hdr) { + t.Errorf("Header mismatch.\n got %+v\nwant %+v", rHdr, hdr) + } + rData, err := ioutil.ReadAll(tr) + if err != nil { + t.Fatalf("Read: %v", err) + } + if !bytes.Equal(rData, data) { + t.Errorf("Data mismatch.\n got %q\nwant %q", rData, data) + } +} + +type headerRoundTripTest struct { + h *Header + fm os.FileMode +} + +func TestHeaderRoundTrip(t *testing.T) { + golden := []headerRoundTripTest{ + // regular file. + { + h: &Header{ + Name: "test.txt", + Mode: 0644 | c_ISREG, + Size: 12, + ModTime: time.Unix(1360600916, 0), + Typeflag: TypeReg, + }, + fm: 0644, + }, + // hard link. + { + h: &Header{ + Name: "hard.txt", + Mode: 0644 | c_ISLNK, + Size: 0, + ModTime: time.Unix(1360600916, 0), + Typeflag: TypeLink, + }, + fm: 0644 | os.ModeSymlink, + }, + // symbolic link. + { + h: &Header{ + Name: "link.txt", + Mode: 0777 | c_ISLNK, + Size: 0, + ModTime: time.Unix(1360600852, 0), + Typeflag: TypeSymlink, + }, + fm: 0777 | os.ModeSymlink, + }, + // character device node. + { + h: &Header{ + Name: "dev/null", + Mode: 0666 | c_ISCHR, + Size: 0, + ModTime: time.Unix(1360578951, 0), + Typeflag: TypeChar, + }, + fm: 0666 | os.ModeDevice | os.ModeCharDevice, + }, + // block device node. + { + h: &Header{ + Name: "dev/sda", + Mode: 0660 | c_ISBLK, + Size: 0, + ModTime: time.Unix(1360578954, 0), + Typeflag: TypeBlock, + }, + fm: 0660 | os.ModeDevice, + }, + // directory. + { + h: &Header{ + Name: "dir/", + Mode: 0755 | c_ISDIR, + Size: 0, + ModTime: time.Unix(1360601116, 0), + Typeflag: TypeDir, + }, + fm: 0755 | os.ModeDir, + }, + // fifo node. + { + h: &Header{ + Name: "dev/initctl", + Mode: 0600 | c_ISFIFO, + Size: 0, + ModTime: time.Unix(1360578949, 0), + Typeflag: TypeFifo, + }, + fm: 0600 | os.ModeNamedPipe, + }, + // setuid. + { + h: &Header{ + Name: "bin/su", + Mode: 0755 | c_ISREG | c_ISUID, + Size: 23232, + ModTime: time.Unix(1355405093, 0), + Typeflag: TypeReg, + }, + fm: 0755 | os.ModeSetuid, + }, + // setguid. + { + h: &Header{ + Name: "group.txt", + Mode: 0750 | c_ISREG | c_ISGID, + Size: 0, + ModTime: time.Unix(1360602346, 0), + Typeflag: TypeReg, + }, + fm: 0750 | os.ModeSetgid, + }, + // sticky. + { + h: &Header{ + Name: "sticky.txt", + Mode: 0600 | c_ISREG | c_ISVTX, + Size: 7, + ModTime: time.Unix(1360602540, 0), + Typeflag: TypeReg, + }, + fm: 0600 | os.ModeSticky, + }, + } + + for i, g := range golden { + fi := g.h.FileInfo() + h2, err := FileInfoHeader(fi, "") + if err != nil { + t.Error(err) + continue + } + if strings.Contains(fi.Name(), "/") { + t.Errorf("FileInfo of %q contains slash: %q", g.h.Name, fi.Name()) + } + name := path.Base(g.h.Name) + if fi.IsDir() { + name += "/" + } + if got, want := h2.Name, name; got != want { + t.Errorf("i=%d: Name: got %v, want %v", i, got, want) + } + if got, want := h2.Size, g.h.Size; got != want { + t.Errorf("i=%d: Size: got %v, want %v", i, got, want) + } + if got, want := h2.Mode, g.h.Mode; got != want { + t.Errorf("i=%d: Mode: got %o, want %o", i, got, want) + } + if got, want := fi.Mode(), g.fm; got != want { + t.Errorf("i=%d: fi.Mode: got %o, want %o", i, got, want) + } + if got, want := h2.ModTime, g.h.ModTime; got != want { + t.Errorf("i=%d: ModTime: got %v, want %v", i, got, want) + } + if sysh, ok := fi.Sys().(*Header); !ok || sysh != g.h { + t.Errorf("i=%d: Sys didn't return original *Header", i) + } + } +} diff --git a/archive/tar/testdata/gnu.tar b/archive/tar/testdata/gnu.tar new file mode 100644 index 0000000..fc899dc Binary files /dev/null and b/archive/tar/testdata/gnu.tar differ diff --git a/archive/tar/testdata/nil-uid.tar b/archive/tar/testdata/nil-uid.tar new file mode 100644 index 0000000..cc9cfaa Binary files /dev/null and b/archive/tar/testdata/nil-uid.tar differ diff --git a/archive/tar/testdata/pax.tar b/archive/tar/testdata/pax.tar new file mode 100644 index 0000000..9bc24b6 Binary files /dev/null and b/archive/tar/testdata/pax.tar differ diff --git a/archive/tar/testdata/small.txt b/archive/tar/testdata/small.txt new file mode 100644 index 0000000..b249bfc --- /dev/null +++ b/archive/tar/testdata/small.txt @@ -0,0 +1 @@ +Kilts \ No newline at end of file diff --git a/archive/tar/testdata/small2.txt b/archive/tar/testdata/small2.txt new file mode 100644 index 0000000..394ee3e --- /dev/null +++ b/archive/tar/testdata/small2.txt @@ -0,0 +1 @@ +Google.com diff --git a/archive/tar/testdata/sparse-formats.tar b/archive/tar/testdata/sparse-formats.tar new file mode 100644 index 0000000..8bd4e74 Binary files /dev/null and b/archive/tar/testdata/sparse-formats.tar differ diff --git a/archive/tar/testdata/star.tar b/archive/tar/testdata/star.tar new file mode 100644 index 0000000..59e2d4e Binary files /dev/null and b/archive/tar/testdata/star.tar differ diff --git a/archive/tar/testdata/ustar.tar b/archive/tar/testdata/ustar.tar new file mode 100644 index 0000000..29679d9 Binary files /dev/null and b/archive/tar/testdata/ustar.tar differ diff --git a/archive/tar/testdata/v7.tar b/archive/tar/testdata/v7.tar new file mode 100644 index 0000000..eb65fc9 Binary files /dev/null and b/archive/tar/testdata/v7.tar differ diff --git a/archive/tar/testdata/writer-big-long.tar b/archive/tar/testdata/writer-big-long.tar new file mode 100644 index 0000000..5960ee8 Binary files /dev/null and b/archive/tar/testdata/writer-big-long.tar differ diff --git a/archive/tar/testdata/writer-big.tar b/archive/tar/testdata/writer-big.tar new file mode 100644 index 0000000..753e883 Binary files /dev/null and b/archive/tar/testdata/writer-big.tar differ diff --git a/archive/tar/testdata/writer.tar b/archive/tar/testdata/writer.tar new file mode 100644 index 0000000..e6d816a Binary files /dev/null and b/archive/tar/testdata/writer.tar differ diff --git a/archive/tar/testdata/xattrs.tar b/archive/tar/testdata/xattrs.tar new file mode 100644 index 0000000..9701950 Binary files /dev/null and b/archive/tar/testdata/xattrs.tar differ diff --git a/archive/tar/writer.go b/archive/tar/writer.go new file mode 100644 index 0000000..dafb2ca --- /dev/null +++ b/archive/tar/writer.go @@ -0,0 +1,396 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tar + +// TODO(dsymonds): +// - catch more errors (no first header, etc.) + +import ( + "bytes" + "errors" + "fmt" + "io" + "os" + "path" + "strconv" + "strings" + "time" +) + +var ( + ErrWriteTooLong = errors.New("archive/tar: write too long") + ErrFieldTooLong = errors.New("archive/tar: header field too long") + ErrWriteAfterClose = errors.New("archive/tar: write after close") + errNameTooLong = errors.New("archive/tar: name too long") + errInvalidHeader = errors.New("archive/tar: header field too long or contains invalid values") +) + +// A Writer provides sequential writing of a tar archive in POSIX.1 format. +// A tar archive consists of a sequence of files. +// Call WriteHeader to begin a new file, and then call Write to supply that file's data, +// writing at most hdr.Size bytes in total. +type Writer struct { + w io.Writer + err error + nb int64 // number of unwritten bytes for current file entry + pad int64 // amount of padding to write after current file entry + closed bool + usedBinary bool // whether the binary numeric field extension was used + preferPax bool // use pax header instead of binary numeric header + hdrBuff [blockSize]byte // buffer to use in writeHeader when writing a regular header + paxHdrBuff [blockSize]byte // buffer to use in writeHeader when writing a pax header +} + +// NewWriter creates a new Writer writing to w. +func NewWriter(w io.Writer) *Writer { return &Writer{w: w} } + +// Flush finishes writing the current file (optional). +func (tw *Writer) Flush() error { + if tw.nb > 0 { + tw.err = fmt.Errorf("archive/tar: missed writing %d bytes", tw.nb) + return tw.err + } + + n := tw.nb + tw.pad + for n > 0 && tw.err == nil { + nr := n + if nr > blockSize { + nr = blockSize + } + var nw int + nw, tw.err = tw.w.Write(zeroBlock[0:nr]) + n -= int64(nw) + } + tw.nb = 0 + tw.pad = 0 + return tw.err +} + +// Write s into b, terminating it with a NUL if there is room. +// If the value is too long for the field and allowPax is true add a paxheader record instead +func (tw *Writer) cString(b []byte, s string, allowPax bool, paxKeyword string, paxHeaders map[string]string) { + needsPaxHeader := allowPax && len(s) > len(b) || !isASCII(s) + if needsPaxHeader { + paxHeaders[paxKeyword] = s + return + } + if len(s) > len(b) { + if tw.err == nil { + tw.err = ErrFieldTooLong + } + return + } + ascii := toASCII(s) + copy(b, ascii) + if len(ascii) < len(b) { + b[len(ascii)] = 0 + } +} + +// Encode x as an octal ASCII string and write it into b with leading zeros. +func (tw *Writer) octal(b []byte, x int64) { + s := strconv.FormatInt(x, 8) + // leading zeros, but leave room for a NUL. + for len(s)+1 < len(b) { + s = "0" + s + } + tw.cString(b, s, false, paxNone, nil) +} + +// Write x into b, either as octal or as binary (GNUtar/star extension). +// If the value is too long for the field and writingPax is enabled both for the field and the add a paxheader record instead +func (tw *Writer) numeric(b []byte, x int64, allowPax bool, paxKeyword string, paxHeaders map[string]string) { + // Try octal first. + s := strconv.FormatInt(x, 8) + if len(s) < len(b) { + tw.octal(b, x) + return + } + + // If it is too long for octal, and pax is preferred, use a pax header + if allowPax && tw.preferPax { + tw.octal(b, 0) + s := strconv.FormatInt(x, 10) + paxHeaders[paxKeyword] = s + return + } + + // Too big: use binary (big-endian). + tw.usedBinary = true + for i := len(b) - 1; x > 0 && i >= 0; i-- { + b[i] = byte(x) + x >>= 8 + } + b[0] |= 0x80 // highest bit indicates binary format +} + +var ( + minTime = time.Unix(0, 0) + // There is room for 11 octal digits (33 bits) of mtime. + maxTime = minTime.Add((1<<33 - 1) * time.Second) +) + +// WriteHeader writes hdr and prepares to accept the file's contents. +// WriteHeader calls Flush if it is not the first header. +// Calling after a Close will return ErrWriteAfterClose. +func (tw *Writer) WriteHeader(hdr *Header) error { + return tw.writeHeader(hdr, true) +} + +// WriteHeader writes hdr and prepares to accept the file's contents. +// WriteHeader calls Flush if it is not the first header. +// Calling after a Close will return ErrWriteAfterClose. +// As this method is called internally by writePax header to allow it to +// suppress writing the pax header. +func (tw *Writer) writeHeader(hdr *Header, allowPax bool) error { + if tw.closed { + return ErrWriteAfterClose + } + if tw.err == nil { + tw.Flush() + } + if tw.err != nil { + return tw.err + } + + // a map to hold pax header records, if any are needed + paxHeaders := make(map[string]string) + + // TODO(shanemhansen): we might want to use PAX headers for + // subsecond time resolution, but for now let's just capture + // too long fields or non ascii characters + + var header []byte + + // We need to select which scratch buffer to use carefully, + // since this method is called recursively to write PAX headers. + // If allowPax is true, this is the non-recursive call, and we will use hdrBuff. + // If allowPax is false, we are being called by writePAXHeader, and hdrBuff is + // already being used by the non-recursive call, so we must use paxHdrBuff. + header = tw.hdrBuff[:] + if !allowPax { + header = tw.paxHdrBuff[:] + } + copy(header, zeroBlock) + s := slicer(header) + + // keep a reference to the filename to allow to overwrite it later if we detect that we can use ustar longnames instead of pax + pathHeaderBytes := s.next(fileNameSize) + + tw.cString(pathHeaderBytes, hdr.Name, true, paxPath, paxHeaders) + + // Handle out of range ModTime carefully. + var modTime int64 + if !hdr.ModTime.Before(minTime) && !hdr.ModTime.After(maxTime) { + modTime = hdr.ModTime.Unix() + } + + tw.octal(s.next(8), hdr.Mode) // 100:108 + tw.numeric(s.next(8), int64(hdr.Uid), true, paxUid, paxHeaders) // 108:116 + tw.numeric(s.next(8), int64(hdr.Gid), true, paxGid, paxHeaders) // 116:124 + tw.numeric(s.next(12), hdr.Size, true, paxSize, paxHeaders) // 124:136 + tw.numeric(s.next(12), modTime, false, paxNone, nil) // 136:148 --- consider using pax for finer granularity + s.next(8) // chksum (148:156) + s.next(1)[0] = hdr.Typeflag // 156:157 + + tw.cString(s.next(100), hdr.Linkname, true, paxLinkpath, paxHeaders) + + copy(s.next(8), []byte("ustar\x0000")) // 257:265 + tw.cString(s.next(32), hdr.Uname, true, paxUname, paxHeaders) // 265:297 + tw.cString(s.next(32), hdr.Gname, true, paxGname, paxHeaders) // 297:329 + tw.numeric(s.next(8), hdr.Devmajor, false, paxNone, nil) // 329:337 + tw.numeric(s.next(8), hdr.Devminor, false, paxNone, nil) // 337:345 + + // keep a reference to the prefix to allow to overwrite it later if we detect that we can use ustar longnames instead of pax + prefixHeaderBytes := s.next(155) + tw.cString(prefixHeaderBytes, "", false, paxNone, nil) // 345:500 prefix + + // Use the GNU magic instead of POSIX magic if we used any GNU extensions. + if tw.usedBinary { + copy(header[257:265], []byte("ustar \x00")) + } + + _, paxPathUsed := paxHeaders[paxPath] + // try to use a ustar header when only the name is too long + if !tw.preferPax && len(paxHeaders) == 1 && paxPathUsed { + suffix := hdr.Name + prefix := "" + if len(hdr.Name) > fileNameSize && isASCII(hdr.Name) { + var err error + prefix, suffix, err = tw.splitUSTARLongName(hdr.Name) + if err == nil { + // ok we can use a ustar long name instead of pax, now correct the fields + + // remove the path field from the pax header. this will suppress the pax header + delete(paxHeaders, paxPath) + + // update the path fields + tw.cString(pathHeaderBytes, suffix, false, paxNone, nil) + tw.cString(prefixHeaderBytes, prefix, false, paxNone, nil) + + // Use the ustar magic if we used ustar long names. + if len(prefix) > 0 && !tw.usedBinary { + copy(header[257:265], []byte("ustar\x00")) + } + } + } + } + + // The chksum field is terminated by a NUL and a space. + // This is different from the other octal fields. + chksum, _ := checksum(header) + tw.octal(header[148:155], chksum) + header[155] = ' ' + + if tw.err != nil { + // problem with header; probably integer too big for a field. + return tw.err + } + + if allowPax { + for k, v := range hdr.Xattrs { + paxHeaders[paxXattr+k] = v + } + } + + if len(paxHeaders) > 0 { + if !allowPax { + return errInvalidHeader + } + if err := tw.writePAXHeader(hdr, paxHeaders); err != nil { + return err + } + } + tw.nb = int64(hdr.Size) + tw.pad = (blockSize - (tw.nb % blockSize)) % blockSize + + _, tw.err = tw.w.Write(header) + return tw.err +} + +// writeUSTARLongName splits a USTAR long name hdr.Name. +// name must be < 256 characters. errNameTooLong is returned +// if hdr.Name can't be split. The splitting heuristic +// is compatible with gnu tar. +func (tw *Writer) splitUSTARLongName(name string) (prefix, suffix string, err error) { + length := len(name) + if length > fileNamePrefixSize+1 { + length = fileNamePrefixSize + 1 + } else if name[length-1] == '/' { + length-- + } + i := strings.LastIndex(name[:length], "/") + // nlen contains the resulting length in the name field. + // plen contains the resulting length in the prefix field. + nlen := len(name) - i - 1 + plen := i + if i <= 0 || nlen > fileNameSize || nlen == 0 || plen > fileNamePrefixSize { + err = errNameTooLong + return + } + prefix, suffix = name[:i], name[i+1:] + return +} + +// writePaxHeader writes an extended pax header to the +// archive. +func (tw *Writer) writePAXHeader(hdr *Header, paxHeaders map[string]string) error { + // Prepare extended header + ext := new(Header) + ext.Typeflag = TypeXHeader + // Setting ModTime is required for reader parsing to + // succeed, and seems harmless enough. + ext.ModTime = hdr.ModTime + // The spec asks that we namespace our pseudo files + // with the current pid. + pid := os.Getpid() + dir, file := path.Split(hdr.Name) + fullName := path.Join(dir, + fmt.Sprintf("PaxHeaders.%d", pid), file) + + ascii := toASCII(fullName) + if len(ascii) > 100 { + ascii = ascii[:100] + } + ext.Name = ascii + // Construct the body + var buf bytes.Buffer + + for k, v := range paxHeaders { + fmt.Fprint(&buf, paxHeader(k+"="+v)) + } + + ext.Size = int64(len(buf.Bytes())) + if err := tw.writeHeader(ext, false); err != nil { + return err + } + if _, err := tw.Write(buf.Bytes()); err != nil { + return err + } + if err := tw.Flush(); err != nil { + return err + } + return nil +} + +// paxHeader formats a single pax record, prefixing it with the appropriate length +func paxHeader(msg string) string { + const padding = 2 // Extra padding for space and newline + size := len(msg) + padding + size += len(strconv.Itoa(size)) + record := fmt.Sprintf("%d %s\n", size, msg) + if len(record) != size { + // Final adjustment if adding size increased + // the number of digits in size + size = len(record) + record = fmt.Sprintf("%d %s\n", size, msg) + } + return record +} + +// Write writes to the current entry in the tar archive. +// Write returns the error ErrWriteTooLong if more than +// hdr.Size bytes are written after WriteHeader. +func (tw *Writer) Write(b []byte) (n int, err error) { + if tw.closed { + err = ErrWriteTooLong + return + } + overwrite := false + if int64(len(b)) > tw.nb { + b = b[0:tw.nb] + overwrite = true + } + n, err = tw.w.Write(b) + tw.nb -= int64(n) + if err == nil && overwrite { + err = ErrWriteTooLong + return + } + tw.err = err + return +} + +// Close closes the tar archive, flushing any unwritten +// data to the underlying writer. +func (tw *Writer) Close() error { + if tw.err != nil || tw.closed { + return tw.err + } + tw.Flush() + tw.closed = true + if tw.err != nil { + return tw.err + } + + // trailer: two zero blocks + for i := 0; i < 2; i++ { + _, tw.err = tw.w.Write(zeroBlock) + if tw.err != nil { + break + } + } + return tw.err +} diff --git a/archive/tar/writer_test.go b/archive/tar/writer_test.go new file mode 100644 index 0000000..5e42e32 --- /dev/null +++ b/archive/tar/writer_test.go @@ -0,0 +1,491 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tar + +import ( + "bytes" + "fmt" + "io" + "io/ioutil" + "os" + "reflect" + "strings" + "testing" + "testing/iotest" + "time" +) + +type writerTestEntry struct { + header *Header + contents string +} + +type writerTest struct { + file string // filename of expected output + entries []*writerTestEntry +} + +var writerTests = []*writerTest{ + // The writer test file was produced with this command: + // tar (GNU tar) 1.26 + // ln -s small.txt link.txt + // tar -b 1 --format=ustar -c -f writer.tar small.txt small2.txt link.txt + { + file: "testdata/writer.tar", + entries: []*writerTestEntry{ + { + header: &Header{ + Name: "small.txt", + Mode: 0640, + Uid: 73025, + Gid: 5000, + Size: 5, + ModTime: time.Unix(1246508266, 0), + Typeflag: '0', + Uname: "dsymonds", + Gname: "eng", + }, + contents: "Kilts", + }, + { + header: &Header{ + Name: "small2.txt", + Mode: 0640, + Uid: 73025, + Gid: 5000, + Size: 11, + ModTime: time.Unix(1245217492, 0), + Typeflag: '0', + Uname: "dsymonds", + Gname: "eng", + }, + contents: "Google.com\n", + }, + { + header: &Header{ + Name: "link.txt", + Mode: 0777, + Uid: 1000, + Gid: 1000, + Size: 0, + ModTime: time.Unix(1314603082, 0), + Typeflag: '2', + Linkname: "small.txt", + Uname: "strings", + Gname: "strings", + }, + // no contents + }, + }, + }, + // The truncated test file was produced using these commands: + // dd if=/dev/zero bs=1048576 count=16384 > /tmp/16gig.txt + // tar -b 1 -c -f- /tmp/16gig.txt | dd bs=512 count=8 > writer-big.tar + { + file: "testdata/writer-big.tar", + entries: []*writerTestEntry{ + { + header: &Header{ + Name: "tmp/16gig.txt", + Mode: 0640, + Uid: 73025, + Gid: 5000, + Size: 16 << 30, + ModTime: time.Unix(1254699560, 0), + Typeflag: '0', + Uname: "dsymonds", + Gname: "eng", + }, + // fake contents + contents: strings.Repeat("\x00", 4<<10), + }, + }, + }, + // The truncated test file was produced using these commands: + // dd if=/dev/zero bs=1048576 count=16384 > (longname/)*15 /16gig.txt + // tar -b 1 -c -f- (longname/)*15 /16gig.txt | dd bs=512 count=8 > writer-big-long.tar + { + file: "testdata/writer-big-long.tar", + entries: []*writerTestEntry{ + { + header: &Header{ + Name: strings.Repeat("longname/", 15) + "16gig.txt", + Mode: 0644, + Uid: 1000, + Gid: 1000, + Size: 16 << 30, + ModTime: time.Unix(1399583047, 0), + Typeflag: '0', + Uname: "guillaume", + Gname: "guillaume", + }, + // fake contents + contents: strings.Repeat("\x00", 4<<10), + }, + }, + }, + // This file was produced using gnu tar 1.17 + // gnutar -b 4 --format=ustar (longname/)*15 + file.txt + { + file: "testdata/ustar.tar", + entries: []*writerTestEntry{ + { + header: &Header{ + Name: strings.Repeat("longname/", 15) + "file.txt", + Mode: 0644, + Uid: 0765, + Gid: 024, + Size: 06, + ModTime: time.Unix(1360135598, 0), + Typeflag: '0', + Uname: "shane", + Gname: "staff", + }, + contents: "hello\n", + }, + }, + }, +} + +// Render byte array in a two-character hexadecimal string, spaced for easy visual inspection. +func bytestr(offset int, b []byte) string { + const rowLen = 32 + s := fmt.Sprintf("%04x ", offset) + for _, ch := range b { + switch { + case '0' <= ch && ch <= '9', 'A' <= ch && ch <= 'Z', 'a' <= ch && ch <= 'z': + s += fmt.Sprintf(" %c", ch) + default: + s += fmt.Sprintf(" %02x", ch) + } + } + return s +} + +// Render a pseudo-diff between two blocks of bytes. +func bytediff(a []byte, b []byte) string { + const rowLen = 32 + s := fmt.Sprintf("(%d bytes vs. %d bytes)\n", len(a), len(b)) + for offset := 0; len(a)+len(b) > 0; offset += rowLen { + na, nb := rowLen, rowLen + if na > len(a) { + na = len(a) + } + if nb > len(b) { + nb = len(b) + } + sa := bytestr(offset, a[0:na]) + sb := bytestr(offset, b[0:nb]) + if sa != sb { + s += fmt.Sprintf("-%v\n+%v\n", sa, sb) + } + a = a[na:] + b = b[nb:] + } + return s +} + +func TestWriter(t *testing.T) { +testLoop: + for i, test := range writerTests { + expected, err := ioutil.ReadFile(test.file) + if err != nil { + t.Errorf("test %d: Unexpected error: %v", i, err) + continue + } + + buf := new(bytes.Buffer) + tw := NewWriter(iotest.TruncateWriter(buf, 4<<10)) // only catch the first 4 KB + big := false + for j, entry := range test.entries { + big = big || entry.header.Size > 1<<10 + if err := tw.WriteHeader(entry.header); err != nil { + t.Errorf("test %d, entry %d: Failed writing header: %v", i, j, err) + continue testLoop + } + if _, err := io.WriteString(tw, entry.contents); err != nil { + t.Errorf("test %d, entry %d: Failed writing contents: %v", i, j, err) + continue testLoop + } + } + // Only interested in Close failures for the small tests. + if err := tw.Close(); err != nil && !big { + t.Errorf("test %d: Failed closing archive: %v", i, err) + continue testLoop + } + + actual := buf.Bytes() + if !bytes.Equal(expected, actual) { + t.Errorf("test %d: Incorrect result: (-=expected, +=actual)\n%v", + i, bytediff(expected, actual)) + } + if testing.Short() { // The second test is expensive. + break + } + } +} + +func TestPax(t *testing.T) { + // Create an archive with a large name + fileinfo, err := os.Stat("testdata/small.txt") + if err != nil { + t.Fatal(err) + } + hdr, err := FileInfoHeader(fileinfo, "") + if err != nil { + t.Fatalf("os.Stat: %v", err) + } + // Force a PAX long name to be written + longName := strings.Repeat("ab", 100) + contents := strings.Repeat(" ", int(hdr.Size)) + hdr.Name = longName + var buf bytes.Buffer + writer := NewWriter(&buf) + if err := writer.WriteHeader(hdr); err != nil { + t.Fatal(err) + } + if _, err = writer.Write([]byte(contents)); err != nil { + t.Fatal(err) + } + if err := writer.Close(); err != nil { + t.Fatal(err) + } + // Simple test to make sure PAX extensions are in effect + if !bytes.Contains(buf.Bytes(), []byte("PaxHeaders.")) { + t.Fatal("Expected at least one PAX header to be written.") + } + // Test that we can get a long name back out of the archive. + reader := NewReader(&buf) + hdr, err = reader.Next() + if err != nil { + t.Fatal(err) + } + if hdr.Name != longName { + t.Fatal("Couldn't recover long file name") + } +} + +func TestPaxSymlink(t *testing.T) { + // Create an archive with a large linkname + fileinfo, err := os.Stat("testdata/small.txt") + if err != nil { + t.Fatal(err) + } + hdr, err := FileInfoHeader(fileinfo, "") + hdr.Typeflag = TypeSymlink + if err != nil { + t.Fatalf("os.Stat:1 %v", err) + } + // Force a PAX long linkname to be written + longLinkname := strings.Repeat("1234567890/1234567890", 10) + hdr.Linkname = longLinkname + + hdr.Size = 0 + var buf bytes.Buffer + writer := NewWriter(&buf) + if err := writer.WriteHeader(hdr); err != nil { + t.Fatal(err) + } + if err := writer.Close(); err != nil { + t.Fatal(err) + } + // Simple test to make sure PAX extensions are in effect + if !bytes.Contains(buf.Bytes(), []byte("PaxHeaders.")) { + t.Fatal("Expected at least one PAX header to be written.") + } + // Test that we can get a long name back out of the archive. + reader := NewReader(&buf) + hdr, err = reader.Next() + if err != nil { + t.Fatal(err) + } + if hdr.Linkname != longLinkname { + t.Fatal("Couldn't recover long link name") + } +} + +func TestPaxNonAscii(t *testing.T) { + // Create an archive with non ascii. These should trigger a pax header + // because pax headers have a defined utf-8 encoding. + fileinfo, err := os.Stat("testdata/small.txt") + if err != nil { + t.Fatal(err) + } + + hdr, err := FileInfoHeader(fileinfo, "") + if err != nil { + t.Fatalf("os.Stat:1 %v", err) + } + + // some sample data + chineseFilename := "文件名" + chineseGroupname := "組" + chineseUsername := "用戶名" + + hdr.Name = chineseFilename + hdr.Gname = chineseGroupname + hdr.Uname = chineseUsername + + contents := strings.Repeat(" ", int(hdr.Size)) + + var buf bytes.Buffer + writer := NewWriter(&buf) + if err := writer.WriteHeader(hdr); err != nil { + t.Fatal(err) + } + if _, err = writer.Write([]byte(contents)); err != nil { + t.Fatal(err) + } + if err := writer.Close(); err != nil { + t.Fatal(err) + } + // Simple test to make sure PAX extensions are in effect + if !bytes.Contains(buf.Bytes(), []byte("PaxHeaders.")) { + t.Fatal("Expected at least one PAX header to be written.") + } + // Test that we can get a long name back out of the archive. + reader := NewReader(&buf) + hdr, err = reader.Next() + if err != nil { + t.Fatal(err) + } + if hdr.Name != chineseFilename { + t.Fatal("Couldn't recover unicode name") + } + if hdr.Gname != chineseGroupname { + t.Fatal("Couldn't recover unicode group") + } + if hdr.Uname != chineseUsername { + t.Fatal("Couldn't recover unicode user") + } +} + +func TestPaxXattrs(t *testing.T) { + xattrs := map[string]string{ + "user.key": "value", + } + + // Create an archive with an xattr + fileinfo, err := os.Stat("testdata/small.txt") + if err != nil { + t.Fatal(err) + } + hdr, err := FileInfoHeader(fileinfo, "") + if err != nil { + t.Fatalf("os.Stat: %v", err) + } + contents := "Kilts" + hdr.Xattrs = xattrs + var buf bytes.Buffer + writer := NewWriter(&buf) + if err := writer.WriteHeader(hdr); err != nil { + t.Fatal(err) + } + if _, err = writer.Write([]byte(contents)); err != nil { + t.Fatal(err) + } + if err := writer.Close(); err != nil { + t.Fatal(err) + } + // Test that we can get the xattrs back out of the archive. + reader := NewReader(&buf) + hdr, err = reader.Next() + if err != nil { + t.Fatal(err) + } + if !reflect.DeepEqual(hdr.Xattrs, xattrs) { + t.Fatalf("xattrs did not survive round trip: got %+v, want %+v", + hdr.Xattrs, xattrs) + } +} + +func TestPAXHeader(t *testing.T) { + medName := strings.Repeat("CD", 50) + longName := strings.Repeat("AB", 100) + paxTests := [][2]string{ + {paxPath + "=/etc/hosts", "19 path=/etc/hosts\n"}, + {"a=b", "6 a=b\n"}, // Single digit length + {"a=names", "11 a=names\n"}, // Test case involving carries + {paxPath + "=" + longName, fmt.Sprintf("210 path=%s\n", longName)}, + {paxPath + "=" + medName, fmt.Sprintf("110 path=%s\n", medName)}} + + for _, test := range paxTests { + key, expected := test[0], test[1] + if result := paxHeader(key); result != expected { + t.Fatalf("paxHeader: got %s, expected %s", result, expected) + } + } +} + +func TestUSTARLongName(t *testing.T) { + // Create an archive with a path that failed to split with USTAR extension in previous versions. + fileinfo, err := os.Stat("testdata/small.txt") + if err != nil { + t.Fatal(err) + } + hdr, err := FileInfoHeader(fileinfo, "") + hdr.Typeflag = TypeDir + if err != nil { + t.Fatalf("os.Stat:1 %v", err) + } + // Force a PAX long name to be written. The name was taken from a practical example + // that fails and replaced ever char through numbers to anonymize the sample. + longName := "/0000_0000000/00000-000000000/0000_0000000/00000-0000000000000/0000_0000000/00000-0000000-00000000/0000_0000000/00000000/0000_0000000/000/0000_0000000/00000000v00/0000_0000000/000000/0000_0000000/0000000/0000_0000000/00000y-00/0000/0000/00000000/0x000000/" + hdr.Name = longName + + hdr.Size = 0 + var buf bytes.Buffer + writer := NewWriter(&buf) + if err := writer.WriteHeader(hdr); err != nil { + t.Fatal(err) + } + if err := writer.Close(); err != nil { + t.Fatal(err) + } + // Test that we can get a long name back out of the archive. + reader := NewReader(&buf) + hdr, err = reader.Next() + if err != nil { + t.Fatal(err) + } + if hdr.Name != longName { + t.Fatal("Couldn't recover long name") + } +} + +func TestValidTypeflagWithPAXHeader(t *testing.T) { + var buffer bytes.Buffer + tw := NewWriter(&buffer) + + fileName := strings.Repeat("ab", 100) + + hdr := &Header{ + Name: fileName, + Size: 4, + Typeflag: 0, + } + if err := tw.WriteHeader(hdr); err != nil { + t.Fatalf("Failed to write header: %s", err) + } + if _, err := tw.Write([]byte("fooo")); err != nil { + t.Fatalf("Failed to write the file's data: %s", err) + } + tw.Close() + + tr := NewReader(&buffer) + + for { + header, err := tr.Next() + if err == io.EOF { + break + } + if err != nil { + t.Fatalf("Failed to read header: %s", err) + } + if header.Typeflag != 0 { + t.Fatalf("Typeflag should've been 0, found %d", header.Typeflag) + } + } +}