...
Run Format

Source file src/archive/tar/reader.go

Documentation: archive/tar

  // 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"
  	"io"
  	"io/ioutil"
  	"strconv"
  	"strings"
  	"time"
  )
  
  // Reader provides sequential access to the contents of a tar archive.
  // Reader.Next advances to the next file in the archive (including the first),
  // and then Reader can be treated as an io.Reader to access the file's data.
  type Reader struct {
  	r    io.Reader
  	pad  int64      // Amount of padding (ignored) after current file entry
  	curr fileReader // Reader for current file entry
  	blk  block      // Buffer to use as temporary local storage
  
  	// err is a persistent error.
  	// It is only the responsibility of every exported method of Reader to
  	// ensure that this error is sticky.
  	err error
  }
  
  type fileReader interface {
  	io.Reader
  	fileState
  
  	WriteTo(io.Writer) (int64, error)
  }
  
  // NewReader creates a new Reader reading from r.
  func NewReader(r io.Reader) *Reader {
  	return &Reader{r: r, curr: &regFileReader{r, 0}}
  }
  
  // Next advances to the next entry in the tar archive.
  // The Header.Size determines how many bytes can be read for the next file.
  // Any remaining data in the current file is automatically discarded.
  //
  // io.EOF is returned at the end of the input.
  func (tr *Reader) Next() (*Header, error) {
  	if tr.err != nil {
  		return nil, tr.err
  	}
  	hdr, err := tr.next()
  	tr.err = err
  	return hdr, err
  }
  
  func (tr *Reader) next() (*Header, error) {
  	var paxHdrs map[string]string
  	var gnuLongName, gnuLongLink string
  
  	// Externally, Next iterates through the tar archive as if it is a series of
  	// files. Internally, the tar format often uses fake "files" to add meta
  	// data that describes the next file. These meta data "files" should not
  	// normally be visible to the outside. As such, this loop iterates through
  	// one or more "header files" until it finds a "normal file".
  	format := FormatUSTAR | FormatPAX | FormatGNU
  	for {
  		// Discard the remainder of the file and any padding.
  		if err := discard(tr.r, tr.curr.PhysicalRemaining()); err != nil {
  			return nil, err
  		}
  		if _, err := tryReadFull(tr.r, tr.blk[:tr.pad]); err != nil {
  			return nil, err
  		}
  		tr.pad = 0
  
  		hdr, rawHdr, err := tr.readHeader()
  		if err != nil {
  			return nil, err
  		}
  		if err := tr.handleRegularFile(hdr); err != nil {
  			return nil, err
  		}
  		format.mayOnlyBe(hdr.Format)
  
  		// Check for PAX/GNU special headers and files.
  		switch hdr.Typeflag {
  		case TypeXHeader, TypeXGlobalHeader:
  			format.mayOnlyBe(FormatPAX)
  			paxHdrs, err = parsePAX(tr)
  			if err != nil {
  				return nil, err
  			}
  			if hdr.Typeflag == TypeXGlobalHeader {
  				mergePAX(hdr, paxHdrs)
  				return &Header{
  					Name:       hdr.Name,
  					Typeflag:   hdr.Typeflag,
  					Xattrs:     hdr.Xattrs,
  					PAXRecords: hdr.PAXRecords,
  					Format:     format,
  				}, nil
  			}
  			continue // This is a meta header affecting the next header
  		case TypeGNULongName, TypeGNULongLink:
  			format.mayOnlyBe(FormatGNU)
  			realname, err := ioutil.ReadAll(tr)
  			if err != nil {
  				return nil, err
  			}
  
  			var p parser
  			switch hdr.Typeflag {
  			case TypeGNULongName:
  				gnuLongName = p.parseString(realname)
  			case TypeGNULongLink:
  				gnuLongLink = p.parseString(realname)
  			}
  			continue // This is a meta header affecting the next header
  		default:
  			// The old GNU sparse format is handled here since it is technically
  			// just a regular file with additional attributes.
  
  			if err := mergePAX(hdr, paxHdrs); err != nil {
  				return nil, err
  			}
  			if gnuLongName != "" {
  				hdr.Name = gnuLongName
  			}
  			if gnuLongLink != "" {
  				hdr.Linkname = gnuLongLink
  			}
  			if hdr.Typeflag == TypeRegA {
  				if strings.HasSuffix(hdr.Name, "/") {
  					hdr.Typeflag = TypeDir // Legacy archives use trailing slash for directories
  				} else {
  					hdr.Typeflag = TypeReg
  				}
  			}
  
  			// The extended headers may have updated the size.
  			// Thus, setup the regFileReader again after merging PAX headers.
  			if err := tr.handleRegularFile(hdr); err != nil {
  				return nil, err
  			}
  
  			// Sparse formats rely on being able to read from the logical data
  			// section; there must be a preceding call to handleRegularFile.
  			if err := tr.handleSparseFile(hdr, rawHdr); err != nil {
  				return nil, err
  			}
  
  			// Set the final guess at the format.
  			if format.has(FormatUSTAR) && format.has(FormatPAX) {
  				format.mayOnlyBe(FormatUSTAR)
  			}
  			hdr.Format = format
  			return hdr, nil // This is a file, so stop
  		}
  	}
  }
  
  // handleRegularFile sets up the current file reader and padding such that it
  // can only read the following logical data section. It will properly handle
  // special headers that contain no data section.
  func (tr *Reader) handleRegularFile(hdr *Header) error {
  	nb := hdr.Size
  	if isHeaderOnlyType(hdr.Typeflag) {
  		nb = 0
  	}
  	if nb < 0 {
  		return ErrHeader
  	}
  
  	tr.pad = blockPadding(nb)
  	tr.curr = &regFileReader{r: tr.r, nb: nb}
  	return nil
  }
  
  // handleSparseFile checks if the current file is a sparse format of any type
  // and sets the curr reader appropriately.
  func (tr *Reader) handleSparseFile(hdr *Header, rawHdr *block) error {
  	var spd sparseDatas
  	var err error
  	if hdr.Typeflag == TypeGNUSparse {
  		spd, err = tr.readOldGNUSparseMap(hdr, rawHdr)
  	} else {
  		spd, err = tr.readGNUSparsePAXHeaders(hdr)
  	}
  
  	// If sp is non-nil, then this is a sparse file.
  	// Note that it is possible for len(sp) == 0.
  	if err == nil && spd != nil {
  		if isHeaderOnlyType(hdr.Typeflag) || !validateSparseEntries(spd, hdr.Size) {
  			return ErrHeader
  		}
  		sph := invertSparseEntries(spd, hdr.Size)
  		tr.curr = &sparseFileReader{tr.curr, sph, 0}
  	}
  	return err
  }
  
  // readGNUSparsePAXHeaders checks the PAX headers for GNU sparse headers.
  // If they are found, then this function reads the sparse map and returns it.
  // This assumes that 0.0 headers have already been converted to 0.1 headers
  // by the PAX header parsing logic.
  func (tr *Reader) readGNUSparsePAXHeaders(hdr *Header) (sparseDatas, error) {
  	// Identify the version of GNU headers.
  	var is1x0 bool
  	major, minor := hdr.PAXRecords[paxGNUSparseMajor], hdr.PAXRecords[paxGNUSparseMinor]
  	switch {
  	case major == "0" && (minor == "0" || minor == "1"):
  		is1x0 = false
  	case major == "1" && minor == "0":
  		is1x0 = true
  	case major != "" || minor != "":
  		return nil, nil // Unknown GNU sparse PAX version
  	case hdr.PAXRecords[paxGNUSparseMap] != "":
  		is1x0 = false // 0.0 and 0.1 did not have explicit version records, so guess
  	default:
  		return nil, nil // Not a PAX format GNU sparse file.
  	}
  	hdr.Format.mayOnlyBe(FormatPAX)
  
  	// Update hdr from GNU sparse PAX headers.
  	if name := hdr.PAXRecords[paxGNUSparseName]; name != "" {
  		hdr.Name = name
  	}
  	size := hdr.PAXRecords[paxGNUSparseSize]
  	if size == "" {
  		size = hdr.PAXRecords[paxGNUSparseRealSize]
  	}
  	if size != "" {
  		n, err := strconv.ParseInt(size, 10, 64)
  		if err != nil {
  			return nil, ErrHeader
  		}
  		hdr.Size = n
  	}
  
  	// Read the sparse map according to the appropriate format.
  	if is1x0 {
  		return readGNUSparseMap1x0(tr.curr)
  	}
  	return readGNUSparseMap0x1(hdr.PAXRecords)
  }
  
  // mergePAX merges paxHdrs into hdr for all relevant fields of Header.
  func mergePAX(hdr *Header, paxHdrs map[string]string) (err error) {
  	for k, v := range paxHdrs {
  		if v == "" {
  			continue // Keep the original USTAR value
  		}
  		var id64 int64
  		switch k {
  		case paxPath:
  			hdr.Name = v
  		case paxLinkpath:
  			hdr.Linkname = v
  		case paxUname:
  			hdr.Uname = v
  		case paxGname:
  			hdr.Gname = v
  		case paxUid:
  			id64, err = strconv.ParseInt(v, 10, 64)
  			hdr.Uid = int(id64) // Integer overflow possible
  		case paxGid:
  			id64, err = strconv.ParseInt(v, 10, 64)
  			hdr.Gid = int(id64) // Integer overflow possible
  		case paxAtime:
  			hdr.AccessTime, err = parsePAXTime(v)
  		case paxMtime:
  			hdr.ModTime, err = parsePAXTime(v)
  		case paxCtime:
  			hdr.ChangeTime, err = parsePAXTime(v)
  		case paxSize:
  			hdr.Size, err = strconv.ParseInt(v, 10, 64)
  		default:
  			if strings.HasPrefix(k, paxSchilyXattr) {
  				if hdr.Xattrs == nil {
  					hdr.Xattrs = make(map[string]string)
  				}
  				hdr.Xattrs[k[len(paxSchilyXattr):]] = v
  			}
  		}
  		if err != nil {
  			return ErrHeader
  		}
  	}
  	hdr.PAXRecords = paxHdrs
  	return 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
  	}
  	sbuf := string(buf)
  
  	// For GNU PAX sparse format 0.0 support.
  	// This function transforms the sparse format 0.0 headers into format 0.1
  	// headers since 0.0 headers were not PAX compliant.
  	var sparseMap []string
  
  	paxHdrs := make(map[string]string)
  	for len(sbuf) > 0 {
  		key, value, residual, err := parsePAXRecord(sbuf)
  		if err != nil {
  			return nil, ErrHeader
  		}
  		sbuf = residual
  
  		switch key {
  		case paxGNUSparseOffset, paxGNUSparseNumBytes:
  			// Validate sparse header order and value.
  			if (len(sparseMap)%2 == 0 && key != paxGNUSparseOffset) ||
  				(len(sparseMap)%2 == 1 && key != paxGNUSparseNumBytes) ||
  				strings.Contains(value, ",") {
  				return nil, ErrHeader
  			}
  			sparseMap = append(sparseMap, value)
  		default:
  			paxHdrs[key] = value
  		}
  	}
  	if len(sparseMap) > 0 {
  		paxHdrs[paxGNUSparseMap] = strings.Join(sparseMap, ",")
  	}
  	return paxHdrs, nil
  }
  
  // readHeader reads the next block header and assumes that the underlying reader
  // is already aligned to a block boundary. It returns the raw block of the
  // header in case further processing is required.
  //
  // The err will be set to io.EOF only when one of the following occurs:
  //	* Exactly 0 bytes are read and EOF is hit.
  //	* Exactly 1 block of zeros is read and EOF is hit.
  //	* At least 2 blocks of zeros are read.
  func (tr *Reader) readHeader() (*Header, *block, error) {
  	// Two blocks of zero bytes marks the end of the archive.
  	if _, err := io.ReadFull(tr.r, tr.blk[:]); err != nil {
  		return nil, nil, err // EOF is okay here; exactly 0 bytes read
  	}
  	if bytes.Equal(tr.blk[:], zeroBlock[:]) {
  		if _, err := io.ReadFull(tr.r, tr.blk[:]); err != nil {
  			return nil, nil, err // EOF is okay here; exactly 1 block of zeros read
  		}
  		if bytes.Equal(tr.blk[:], zeroBlock[:]) {
  			return nil, nil, io.EOF // normal EOF; exactly 2 block of zeros read
  		}
  		return nil, nil, ErrHeader // Zero block and then non-zero block
  	}
  
  	// Verify the header matches a known format.
  	format := tr.blk.GetFormat()
  	if format == FormatUnknown {
  		return nil, nil, ErrHeader
  	}
  
  	var p parser
  	hdr := new(Header)
  
  	// Unpack the V7 header.
  	v7 := tr.blk.V7()
  	hdr.Typeflag = v7.TypeFlag()[0]
  	hdr.Name = p.parseString(v7.Name())
  	hdr.Linkname = p.parseString(v7.LinkName())
  	hdr.Size = p.parseNumeric(v7.Size())
  	hdr.Mode = p.parseNumeric(v7.Mode())
  	hdr.Uid = int(p.parseNumeric(v7.UID()))
  	hdr.Gid = int(p.parseNumeric(v7.GID()))
  	hdr.ModTime = time.Unix(p.parseNumeric(v7.ModTime()), 0)
  
  	// Unpack format specific fields.
  	if format > formatV7 {
  		ustar := tr.blk.USTAR()
  		hdr.Uname = p.parseString(ustar.UserName())
  		hdr.Gname = p.parseString(ustar.GroupName())
  		hdr.Devmajor = p.parseNumeric(ustar.DevMajor())
  		hdr.Devminor = p.parseNumeric(ustar.DevMinor())
  
  		var prefix string
  		switch {
  		case format.has(FormatUSTAR | FormatPAX):
  			hdr.Format = format
  			ustar := tr.blk.USTAR()
  			prefix = p.parseString(ustar.Prefix())
  
  			// For Format detection, check if block is properly formatted since
  			// the parser is more liberal than what USTAR actually permits.
  			notASCII := func(r rune) bool { return r >= 0x80 }
  			if bytes.IndexFunc(tr.blk[:], notASCII) >= 0 {
  				hdr.Format = FormatUnknown // Non-ASCII characters in block.
  			}
  			nul := func(b []byte) bool { return int(b[len(b)-1]) == 0 }
  			if !(nul(v7.Size()) && nul(v7.Mode()) && nul(v7.UID()) && nul(v7.GID()) &&
  				nul(v7.ModTime()) && nul(ustar.DevMajor()) && nul(ustar.DevMinor())) {
  				hdr.Format = FormatUnknown // Numeric fields must end in NUL
  			}
  		case format.has(formatSTAR):
  			star := tr.blk.STAR()
  			prefix = p.parseString(star.Prefix())
  			hdr.AccessTime = time.Unix(p.parseNumeric(star.AccessTime()), 0)
  			hdr.ChangeTime = time.Unix(p.parseNumeric(star.ChangeTime()), 0)
  		case format.has(FormatGNU):
  			hdr.Format = format
  			var p2 parser
  			gnu := tr.blk.GNU()
  			if b := gnu.AccessTime(); b[0] != 0 {
  				hdr.AccessTime = time.Unix(p2.parseNumeric(b), 0)
  			}
  			if b := gnu.ChangeTime(); b[0] != 0 {
  				hdr.ChangeTime = time.Unix(p2.parseNumeric(b), 0)
  			}
  
  			// Prior to Go1.8, the Writer had a bug where it would output
  			// an invalid tar file in certain rare situations because the logic
  			// incorrectly believed that the old GNU format had a prefix field.
  			// This is wrong and leads to an output file that mangles the
  			// atime and ctime fields, which are often left unused.
  			//
  			// In order to continue reading tar files created by former, buggy
  			// versions of Go, we skeptically parse the atime and ctime fields.
  			// If we are unable to parse them and the prefix field looks like
  			// an ASCII string, then we fallback on the pre-Go1.8 behavior
  			// of treating these fields as the USTAR prefix field.
  			//
  			// Note that this will not use the fallback logic for all possible
  			// files generated by a pre-Go1.8 toolchain. If the generated file
  			// happened to have a prefix field that parses as valid
  			// atime and ctime fields (e.g., when they are valid octal strings),
  			// then it is impossible to distinguish between an valid GNU file
  			// and an invalid pre-Go1.8 file.
  			//
  			// See https://golang.org/issues/12594
  			// See https://golang.org/issues/21005
  			if p2.err != nil {
  				hdr.AccessTime, hdr.ChangeTime = time.Time{}, time.Time{}
  				ustar := tr.blk.USTAR()
  				if s := p.parseString(ustar.Prefix()); isASCII(s) {
  					prefix = s
  				}
  				hdr.Format = FormatUnknown // Buggy file is not GNU
  			}
  		}
  		if len(prefix) > 0 {
  			hdr.Name = prefix + "/" + hdr.Name
  		}
  	}
  	return hdr, &tr.blk, p.err
  }
  
  // readOldGNUSparseMap reads the sparse map from 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.
  //
  // The Header.Size does not reflect the size of any extended headers used.
  // Thus, this function will read from the raw io.Reader to fetch extra headers.
  // This method mutates blk in the process.
  func (tr *Reader) readOldGNUSparseMap(hdr *Header, blk *block) (sparseDatas, error) {
  	// Make sure that the input format is GNU.
  	// Unfortunately, the STAR format also has a sparse header format that uses
  	// the same type flag but has a completely different layout.
  	if blk.GetFormat() != FormatGNU {
  		return nil, ErrHeader
  	}
  	hdr.Format.mayOnlyBe(FormatGNU)
  
  	var p parser
  	hdr.Size = p.parseNumeric(blk.GNU().RealSize())
  	if p.err != nil {
  		return nil, p.err
  	}
  	s := blk.GNU().Sparse()
  	spd := make(sparseDatas, 0, s.MaxEntries())
  	for {
  		for i := 0; i < s.MaxEntries(); i++ {
  			// This termination condition is identical to GNU and BSD tar.
  			if s.Entry(i).Offset()[0] == 0x00 {
  				break // Don't return, need to process extended headers (even if empty)
  			}
  			offset := p.parseNumeric(s.Entry(i).Offset())
  			length := p.parseNumeric(s.Entry(i).Length())
  			if p.err != nil {
  				return nil, p.err
  			}
  			spd = append(spd, sparseEntry{Offset: offset, Length: length})
  		}
  
  		if s.IsExtended()[0] > 0 {
  			// There are more entries. Read an extension header and parse its entries.
  			if _, err := mustReadFull(tr.r, blk[:]); err != nil {
  				return nil, err
  			}
  			s = blk.Sparse()
  			continue
  		}
  		return spd, nil // Done
  	}
  }
  
  // readGNUSparseMap1x0 reads the sparse map as stored in GNU's PAX sparse format
  // version 1.0. The format of the sparse map consists of a series of
  // newline-terminated numeric fields. The first field is the number of entries
  // and is always present. Following this are the entries, consisting of two
  // fields (offset, length). This function must stop reading at the end
  // boundary of the block containing the last newline.
  //
  // Note that the GNU manual says that numeric values should be encoded in octal
  // format. However, the GNU tar utility itself outputs these values in decimal.
  // As such, this library treats values as being encoded in decimal.
  func readGNUSparseMap1x0(r io.Reader) (sparseDatas, error) {
  	var (
  		cntNewline int64
  		buf        bytes.Buffer
  		blk        block
  	)
  
  	// feedTokens copies data in blocks from r into buf until there are
  	// at least cnt newlines in buf. It will not read more blocks than needed.
  	feedTokens := func(n int64) error {
  		for cntNewline < n {
  			if _, err := mustReadFull(r, blk[:]); err != nil {
  				return err
  			}
  			buf.Write(blk[:])
  			for _, c := range blk {
  				if c == '\n' {
  					cntNewline++
  				}
  			}
  		}
  		return nil
  	}
  
  	// nextToken gets the next token delimited by a newline. This assumes that
  	// at least one newline exists in the buffer.
  	nextToken := func() string {
  		cntNewline--
  		tok, _ := buf.ReadString('\n')
  		return strings.TrimRight(tok, "\n")
  	}
  
  	// Parse for the number of entries.
  	// Use integer overflow resistant math to check this.
  	if err := feedTokens(1); err != nil {
  		return nil, err
  	}
  	numEntries, err := strconv.ParseInt(nextToken(), 10, 0) // Intentionally parse as native int
  	if err != nil || numEntries < 0 || int(2*numEntries) < int(numEntries) {
  		return nil, ErrHeader
  	}
  
  	// Parse for all member entries.
  	// numEntries is trusted after this since a potential attacker must have
  	// committed resources proportional to what this library used.
  	if err := feedTokens(2 * numEntries); err != nil {
  		return nil, err
  	}
  	spd := make(sparseDatas, 0, numEntries)
  	for i := int64(0); i < numEntries; i++ {
  		offset, err1 := strconv.ParseInt(nextToken(), 10, 64)
  		length, err2 := strconv.ParseInt(nextToken(), 10, 64)
  		if err1 != nil || err2 != nil {
  			return nil, ErrHeader
  		}
  		spd = append(spd, sparseEntry{Offset: offset, Length: length})
  	}
  	return spd, 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(paxHdrs map[string]string) (sparseDatas, error) {
  	// Get number of entries.
  	// Use integer overflow resistant math to check this.
  	numEntriesStr := paxHdrs[paxGNUSparseNumBlocks]
  	numEntries, err := strconv.ParseInt(numEntriesStr, 10, 0) // Intentionally parse as native int
  	if err != nil || numEntries < 0 || int(2*numEntries) < int(numEntries) {
  		return nil, ErrHeader
  	}
  
  	// There should be two numbers in sparseMap for each entry.
  	sparseMap := strings.Split(paxHdrs[paxGNUSparseMap], ",")
  	if len(sparseMap) == 1 && sparseMap[0] == "" {
  		sparseMap = sparseMap[:0]
  	}
  	if int64(len(sparseMap)) != 2*numEntries {
  		return nil, ErrHeader
  	}
  
  	// Loop through the entries in the sparse map.
  	// numEntries is trusted now.
  	spd := make(sparseDatas, 0, numEntries)
  	for len(sparseMap) >= 2 {
  		offset, err1 := strconv.ParseInt(sparseMap[0], 10, 64)
  		length, err2 := strconv.ParseInt(sparseMap[1], 10, 64)
  		if err1 != nil || err2 != nil {
  			return nil, ErrHeader
  		}
  		spd = append(spd, sparseEntry{Offset: offset, Length: length})
  		sparseMap = sparseMap[2:]
  	}
  	return spd, nil
  }
  
  // Read reads from the current file in the tar archive.
  // It returns (0, io.EOF) when it reaches the end of that file,
  // until Next is called to advance to the next file.
  //
  // If the current file is sparse, then the regions marked as a hole
  // are read back as NUL-bytes.
  //
  // Calling Read on special types like TypeLink, TypeSymlink, TypeChar,
  // TypeBlock, TypeDir, and TypeFifo returns (0, io.EOF) regardless of what
  // the Header.Size claims.
  func (tr *Reader) Read(b []byte) (int, error) {
  	if tr.err != nil {
  		return 0, tr.err
  	}
  	n, err := tr.curr.Read(b)
  	if err != nil && err != io.EOF {
  		tr.err = err
  	}
  	return n, err
  }
  
  // writeTo writes the content of the current file to w.
  // The bytes written matches the number of remaining bytes in the current file.
  //
  // If the current file is sparse and w is an io.WriteSeeker,
  // then writeTo uses Seek to skip past holes defined in Header.SparseHoles,
  // assuming that skipped regions are filled with NULs.
  // This always writes the last byte to ensure w is the right size.
  //
  // TODO(dsnet): Re-export this when adding sparse file support.
  // See https://golang.org/issue/22735
  func (tr *Reader) writeTo(w io.Writer) (int64, error) {
  	if tr.err != nil {
  		return 0, tr.err
  	}
  	n, err := tr.curr.WriteTo(w)
  	if err != nil {
  		tr.err = err
  	}
  	return n, err
  }
  
  // regFileReader is a fileReader for reading data from a regular file entry.
  type regFileReader struct {
  	r  io.Reader // Underlying Reader
  	nb int64     // Number of remaining bytes to read
  }
  
  func (fr *regFileReader) Read(b []byte) (n int, err error) {
  	if int64(len(b)) > fr.nb {
  		b = b[:fr.nb]
  	}
  	if len(b) > 0 {
  		n, err = fr.r.Read(b)
  		fr.nb -= int64(n)
  	}
  	switch {
  	case err == io.EOF && fr.nb > 0:
  		return n, io.ErrUnexpectedEOF
  	case err == nil && fr.nb == 0:
  		return n, io.EOF
  	default:
  		return n, err
  	}
  }
  
  func (fr *regFileReader) WriteTo(w io.Writer) (int64, error) {
  	return io.Copy(w, struct{ io.Reader }{fr})
  }
  
  func (fr regFileReader) LogicalRemaining() int64 {
  	return fr.nb
  }
  
  func (fr regFileReader) PhysicalRemaining() int64 {
  	return fr.nb
  }
  
  // sparseFileReader is a fileReader for reading data from a sparse file entry.
  type sparseFileReader struct {
  	fr  fileReader  // Underlying fileReader
  	sp  sparseHoles // Normalized list of sparse holes
  	pos int64       // Current position in sparse file
  }
  
  func (sr *sparseFileReader) Read(b []byte) (n int, err error) {
  	finished := int64(len(b)) >= sr.LogicalRemaining()
  	if finished {
  		b = b[:sr.LogicalRemaining()]
  	}
  
  	b0 := b
  	endPos := sr.pos + int64(len(b))
  	for endPos > sr.pos && err == nil {
  		var nf int // Bytes read in fragment
  		holeStart, holeEnd := sr.sp[0].Offset, sr.sp[0].endOffset()
  		if sr.pos < holeStart { // In a data fragment
  			bf := b[:min(int64(len(b)), holeStart-sr.pos)]
  			nf, err = tryReadFull(sr.fr, bf)
  		} else { // In a hole fragment
  			bf := b[:min(int64(len(b)), holeEnd-sr.pos)]
  			nf, err = tryReadFull(zeroReader{}, bf)
  		}
  		b = b[nf:]
  		sr.pos += int64(nf)
  		if sr.pos >= holeEnd && len(sr.sp) > 1 {
  			sr.sp = sr.sp[1:] // Ensure last fragment always remains
  		}
  	}
  
  	n = len(b0) - len(b)
  	switch {
  	case err == io.EOF:
  		return n, errMissData // Less data in dense file than sparse file
  	case err != nil:
  		return n, err
  	case sr.LogicalRemaining() == 0 && sr.PhysicalRemaining() > 0:
  		return n, errUnrefData // More data in dense file than sparse file
  	case finished:
  		return n, io.EOF
  	default:
  		return n, nil
  	}
  }
  
  func (sr *sparseFileReader) WriteTo(w io.Writer) (n int64, err error) {
  	ws, ok := w.(io.WriteSeeker)
  	if ok {
  		if _, err := ws.Seek(0, io.SeekCurrent); err != nil {
  			ok = false // Not all io.Seeker can really seek
  		}
  	}
  	if !ok {
  		return io.Copy(w, struct{ io.Reader }{sr})
  	}
  
  	var writeLastByte bool
  	pos0 := sr.pos
  	for sr.LogicalRemaining() > 0 && !writeLastByte && err == nil {
  		var nf int64 // Size of fragment
  		holeStart, holeEnd := sr.sp[0].Offset, sr.sp[0].endOffset()
  		if sr.pos < holeStart { // In a data fragment
  			nf = holeStart - sr.pos
  			nf, err = io.CopyN(ws, sr.fr, nf)
  		} else { // In a hole fragment
  			nf = holeEnd - sr.pos
  			if sr.PhysicalRemaining() == 0 {
  				writeLastByte = true
  				nf--
  			}
  			_, err = ws.Seek(nf, io.SeekCurrent)
  		}
  		sr.pos += nf
  		if sr.pos >= holeEnd && len(sr.sp) > 1 {
  			sr.sp = sr.sp[1:] // Ensure last fragment always remains
  		}
  	}
  
  	// If the last fragment is a hole, then seek to 1-byte before EOF, and
  	// write a single byte to ensure the file is the right size.
  	if writeLastByte && err == nil {
  		_, err = ws.Write([]byte{0})
  		sr.pos++
  	}
  
  	n = sr.pos - pos0
  	switch {
  	case err == io.EOF:
  		return n, errMissData // Less data in dense file than sparse file
  	case err != nil:
  		return n, err
  	case sr.LogicalRemaining() == 0 && sr.PhysicalRemaining() > 0:
  		return n, errUnrefData // More data in dense file than sparse file
  	default:
  		return n, nil
  	}
  }
  
  func (sr sparseFileReader) LogicalRemaining() int64 {
  	return sr.sp[len(sr.sp)-1].endOffset() - sr.pos
  }
  func (sr sparseFileReader) PhysicalRemaining() int64 {
  	return sr.fr.PhysicalRemaining()
  }
  
  type zeroReader struct{}
  
  func (zeroReader) Read(b []byte) (int, error) {
  	for i := range b {
  		b[i] = 0
  	}
  	return len(b), nil
  }
  
  // mustReadFull is like io.ReadFull except it returns
  // io.ErrUnexpectedEOF when io.EOF is hit before len(b) bytes are read.
  func mustReadFull(r io.Reader, b []byte) (int, error) {
  	n, err := tryReadFull(r, b)
  	if err == io.EOF {
  		err = io.ErrUnexpectedEOF
  	}
  	return n, err
  }
  
  // tryReadFull is like io.ReadFull except it returns
  // io.EOF when it is hit before len(b) bytes are read.
  func tryReadFull(r io.Reader, b []byte) (n int, err error) {
  	for len(b) > n && err == nil {
  		var nn int
  		nn, err = r.Read(b[n:])
  		n += nn
  	}
  	if len(b) == n && err == io.EOF {
  		err = nil
  	}
  	return n, err
  }
  
  // discard skips n bytes in r, reporting an error if unable to do so.
  func discard(r io.Reader, n int64) error {
  	// If possible, Seek to the last byte before the end of the data section.
  	// Do this because Seek is often lazy about reporting errors; this will mask
  	// the fact that the stream may be truncated. We can rely on the
  	// io.CopyN done shortly afterwards to trigger any IO errors.
  	var seekSkipped int64 // Number of bytes skipped via Seek
  	if sr, ok := r.(io.Seeker); ok && n > 1 {
  		// Not all io.Seeker can actually Seek. For example, os.Stdin implements
  		// io.Seeker, but calling Seek always returns an error and performs
  		// no action. Thus, we try an innocent seek to the current position
  		// to see if Seek is really supported.
  		pos1, err := sr.Seek(0, io.SeekCurrent)
  		if pos1 >= 0 && err == nil {
  			// Seek seems supported, so perform the real Seek.
  			pos2, err := sr.Seek(n-1, io.SeekCurrent)
  			if pos2 < 0 || err != nil {
  				return err
  			}
  			seekSkipped = pos2 - pos1
  		}
  	}
  
  	copySkipped, err := io.CopyN(ioutil.Discard, r, n-seekSkipped)
  	if err == io.EOF && seekSkipped+copySkipped < n {
  		err = io.ErrUnexpectedEOF
  	}
  	return err
  }
  

View as plain text