Source file src/archive/zip/reader.go

Documentation: archive/zip

     1  // Copyright 2010 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package zip
     6  
     7  import (
     8  	"bufio"
     9  	"encoding/binary"
    10  	"errors"
    11  	"hash"
    12  	"hash/crc32"
    13  	"io"
    14  	"os"
    15  	"time"
    16  )
    17  
    18  var (
    19  	ErrFormat    = errors.New("zip: not a valid zip file")
    20  	ErrAlgorithm = errors.New("zip: unsupported compression algorithm")
    21  	ErrChecksum  = errors.New("zip: checksum error")
    22  )
    23  
    24  type Reader struct {
    25  	r             io.ReaderAt
    26  	File          []*File
    27  	Comment       string
    28  	decompressors map[uint16]Decompressor
    29  }
    30  
    31  type ReadCloser struct {
    32  	f *os.File
    33  	Reader
    34  }
    35  
    36  type File struct {
    37  	FileHeader
    38  	zip          *Reader
    39  	zipr         io.ReaderAt
    40  	zipsize      int64
    41  	headerOffset int64
    42  }
    43  
    44  func (f *File) hasDataDescriptor() bool {
    45  	return f.Flags&0x8 != 0
    46  }
    47  
    48  // OpenReader will open the Zip file specified by name and return a ReadCloser.
    49  func OpenReader(name string) (*ReadCloser, error) {
    50  	f, err := os.Open(name)
    51  	if err != nil {
    52  		return nil, err
    53  	}
    54  	fi, err := f.Stat()
    55  	if err != nil {
    56  		f.Close()
    57  		return nil, err
    58  	}
    59  	r := new(ReadCloser)
    60  	if err := r.init(f, fi.Size()); err != nil {
    61  		f.Close()
    62  		return nil, err
    63  	}
    64  	r.f = f
    65  	return r, nil
    66  }
    67  
    68  // NewReader returns a new Reader reading from r, which is assumed to
    69  // have the given size in bytes.
    70  func NewReader(r io.ReaderAt, size int64) (*Reader, error) {
    71  	if size < 0 {
    72  		return nil, errors.New("zip: size cannot be negative")
    73  	}
    74  	zr := new(Reader)
    75  	if err := zr.init(r, size); err != nil {
    76  		return nil, err
    77  	}
    78  	return zr, nil
    79  }
    80  
    81  func (z *Reader) init(r io.ReaderAt, size int64) error {
    82  	end, err := readDirectoryEnd(r, size)
    83  	if err != nil {
    84  		return err
    85  	}
    86  	z.r = r
    87  	z.File = make([]*File, 0, end.directoryRecords)
    88  	z.Comment = end.comment
    89  	rs := io.NewSectionReader(r, 0, size)
    90  	if _, err = rs.Seek(int64(end.directoryOffset), io.SeekStart); err != nil {
    91  		return err
    92  	}
    93  	buf := bufio.NewReader(rs)
    94  
    95  	// The count of files inside a zip is truncated to fit in a uint16.
    96  	// Gloss over this by reading headers until we encounter
    97  	// a bad one, and then only report an ErrFormat or UnexpectedEOF if
    98  	// the file count modulo 65536 is incorrect.
    99  	for {
   100  		f := &File{zip: z, zipr: r, zipsize: size}
   101  		err = readDirectoryHeader(f, buf)
   102  		if err == ErrFormat || err == io.ErrUnexpectedEOF {
   103  			break
   104  		}
   105  		if err != nil {
   106  			return err
   107  		}
   108  		z.File = append(z.File, f)
   109  	}
   110  	if uint16(len(z.File)) != uint16(end.directoryRecords) { // only compare 16 bits here
   111  		// Return the readDirectoryHeader error if we read
   112  		// the wrong number of directory entries.
   113  		return err
   114  	}
   115  	return nil
   116  }
   117  
   118  // RegisterDecompressor registers or overrides a custom decompressor for a
   119  // specific method ID. If a decompressor for a given method is not found,
   120  // Reader will default to looking up the decompressor at the package level.
   121  func (z *Reader) RegisterDecompressor(method uint16, dcomp Decompressor) {
   122  	if z.decompressors == nil {
   123  		z.decompressors = make(map[uint16]Decompressor)
   124  	}
   125  	z.decompressors[method] = dcomp
   126  }
   127  
   128  func (z *Reader) decompressor(method uint16) Decompressor {
   129  	dcomp := z.decompressors[method]
   130  	if dcomp == nil {
   131  		dcomp = decompressor(method)
   132  	}
   133  	return dcomp
   134  }
   135  
   136  // Close closes the Zip file, rendering it unusable for I/O.
   137  func (rc *ReadCloser) Close() error {
   138  	return rc.f.Close()
   139  }
   140  
   141  // DataOffset returns the offset of the file's possibly-compressed
   142  // data, relative to the beginning of the zip file.
   143  //
   144  // Most callers should instead use Open, which transparently
   145  // decompresses data and verifies checksums.
   146  func (f *File) DataOffset() (offset int64, err error) {
   147  	bodyOffset, err := f.findBodyOffset()
   148  	if err != nil {
   149  		return
   150  	}
   151  	return f.headerOffset + bodyOffset, nil
   152  }
   153  
   154  // Open returns a ReadCloser that provides access to the File's contents.
   155  // Multiple files may be read concurrently.
   156  func (f *File) Open() (io.ReadCloser, error) {
   157  	bodyOffset, err := f.findBodyOffset()
   158  	if err != nil {
   159  		return nil, err
   160  	}
   161  	size := int64(f.CompressedSize64)
   162  	r := io.NewSectionReader(f.zipr, f.headerOffset+bodyOffset, size)
   163  	dcomp := f.zip.decompressor(f.Method)
   164  	if dcomp == nil {
   165  		return nil, ErrAlgorithm
   166  	}
   167  	var rc io.ReadCloser = dcomp(r)
   168  	var desr io.Reader
   169  	if f.hasDataDescriptor() {
   170  		desr = io.NewSectionReader(f.zipr, f.headerOffset+bodyOffset+size, dataDescriptorLen)
   171  	}
   172  	rc = &checksumReader{
   173  		rc:   rc,
   174  		hash: crc32.NewIEEE(),
   175  		f:    f,
   176  		desr: desr,
   177  	}
   178  	return rc, nil
   179  }
   180  
   181  type checksumReader struct {
   182  	rc    io.ReadCloser
   183  	hash  hash.Hash32
   184  	nread uint64 // number of bytes read so far
   185  	f     *File
   186  	desr  io.Reader // if non-nil, where to read the data descriptor
   187  	err   error     // sticky error
   188  }
   189  
   190  func (r *checksumReader) Read(b []byte) (n int, err error) {
   191  	if r.err != nil {
   192  		return 0, r.err
   193  	}
   194  	n, err = r.rc.Read(b)
   195  	r.hash.Write(b[:n])
   196  	r.nread += uint64(n)
   197  	if err == nil {
   198  		return
   199  	}
   200  	if err == io.EOF {
   201  		if r.nread != r.f.UncompressedSize64 {
   202  			return 0, io.ErrUnexpectedEOF
   203  		}
   204  		if r.desr != nil {
   205  			if err1 := readDataDescriptor(r.desr, r.f); err1 != nil {
   206  				if err1 == io.EOF {
   207  					err = io.ErrUnexpectedEOF
   208  				} else {
   209  					err = err1
   210  				}
   211  			} else if r.hash.Sum32() != r.f.CRC32 {
   212  				err = ErrChecksum
   213  			}
   214  		} else {
   215  			// If there's not a data descriptor, we still compare
   216  			// the CRC32 of what we've read against the file header
   217  			// or TOC's CRC32, if it seems like it was set.
   218  			if r.f.CRC32 != 0 && r.hash.Sum32() != r.f.CRC32 {
   219  				err = ErrChecksum
   220  			}
   221  		}
   222  	}
   223  	r.err = err
   224  	return
   225  }
   226  
   227  func (r *checksumReader) Close() error { return r.rc.Close() }
   228  
   229  // findBodyOffset does the minimum work to verify the file has a header
   230  // and returns the file body offset.
   231  func (f *File) findBodyOffset() (int64, error) {
   232  	var buf [fileHeaderLen]byte
   233  	if _, err := f.zipr.ReadAt(buf[:], f.headerOffset); err != nil {
   234  		return 0, err
   235  	}
   236  	b := readBuf(buf[:])
   237  	if sig := b.uint32(); sig != fileHeaderSignature {
   238  		return 0, ErrFormat
   239  	}
   240  	b = b[22:] // skip over most of the header
   241  	filenameLen := int(b.uint16())
   242  	extraLen := int(b.uint16())
   243  	return int64(fileHeaderLen + filenameLen + extraLen), nil
   244  }
   245  
   246  // readDirectoryHeader attempts to read a directory header from r.
   247  // It returns io.ErrUnexpectedEOF if it cannot read a complete header,
   248  // and ErrFormat if it doesn't find a valid header signature.
   249  func readDirectoryHeader(f *File, r io.Reader) error {
   250  	var buf [directoryHeaderLen]byte
   251  	if _, err := io.ReadFull(r, buf[:]); err != nil {
   252  		return err
   253  	}
   254  	b := readBuf(buf[:])
   255  	if sig := b.uint32(); sig != directoryHeaderSignature {
   256  		return ErrFormat
   257  	}
   258  	f.CreatorVersion = b.uint16()
   259  	f.ReaderVersion = b.uint16()
   260  	f.Flags = b.uint16()
   261  	f.Method = b.uint16()
   262  	f.ModifiedTime = b.uint16()
   263  	f.ModifiedDate = b.uint16()
   264  	f.CRC32 = b.uint32()
   265  	f.CompressedSize = b.uint32()
   266  	f.UncompressedSize = b.uint32()
   267  	f.CompressedSize64 = uint64(f.CompressedSize)
   268  	f.UncompressedSize64 = uint64(f.UncompressedSize)
   269  	filenameLen := int(b.uint16())
   270  	extraLen := int(b.uint16())
   271  	commentLen := int(b.uint16())
   272  	b = b[4:] // skipped start disk number and internal attributes (2x uint16)
   273  	f.ExternalAttrs = b.uint32()
   274  	f.headerOffset = int64(b.uint32())
   275  	d := make([]byte, filenameLen+extraLen+commentLen)
   276  	if _, err := io.ReadFull(r, d); err != nil {
   277  		return err
   278  	}
   279  	f.Name = string(d[:filenameLen])
   280  	f.Extra = d[filenameLen : filenameLen+extraLen]
   281  	f.Comment = string(d[filenameLen+extraLen:])
   282  
   283  	// Determine the character encoding.
   284  	utf8Valid1, utf8Require1 := detectUTF8(f.Name)
   285  	utf8Valid2, utf8Require2 := detectUTF8(f.Comment)
   286  	switch {
   287  	case !utf8Valid1 || !utf8Valid2:
   288  		// Name and Comment definitely not UTF-8.
   289  		f.NonUTF8 = true
   290  	case !utf8Require1 && !utf8Require2:
   291  		// Name and Comment use only single-byte runes that overlap with UTF-8.
   292  		f.NonUTF8 = false
   293  	default:
   294  		// Might be UTF-8, might be some other encoding; preserve existing flag.
   295  		// Some ZIP writers use UTF-8 encoding without setting the UTF-8 flag.
   296  		// Since it is impossible to always distinguish valid UTF-8 from some
   297  		// other encoding (e.g., GBK or Shift-JIS), we trust the flag.
   298  		f.NonUTF8 = f.Flags&0x800 == 0
   299  	}
   300  
   301  	needUSize := f.UncompressedSize == ^uint32(0)
   302  	needCSize := f.CompressedSize == ^uint32(0)
   303  	needHeaderOffset := f.headerOffset == int64(^uint32(0))
   304  
   305  	// Best effort to find what we need.
   306  	// Other zip authors might not even follow the basic format,
   307  	// and we'll just ignore the Extra content in that case.
   308  	var modified time.Time
   309  parseExtras:
   310  	for extra := readBuf(f.Extra); len(extra) >= 4; { // need at least tag and size
   311  		fieldTag := extra.uint16()
   312  		fieldSize := int(extra.uint16())
   313  		if len(extra) < fieldSize {
   314  			break
   315  		}
   316  		fieldBuf := extra.sub(fieldSize)
   317  
   318  		switch fieldTag {
   319  		case zip64ExtraID:
   320  			// update directory values from the zip64 extra block.
   321  			// They should only be consulted if the sizes read earlier
   322  			// are maxed out.
   323  			// See golang.org/issue/13367.
   324  			if needUSize {
   325  				needUSize = false
   326  				if len(fieldBuf) < 8 {
   327  					return ErrFormat
   328  				}
   329  				f.UncompressedSize64 = fieldBuf.uint64()
   330  			}
   331  			if needCSize {
   332  				needCSize = false
   333  				if len(fieldBuf) < 8 {
   334  					return ErrFormat
   335  				}
   336  				f.CompressedSize64 = fieldBuf.uint64()
   337  			}
   338  			if needHeaderOffset {
   339  				needHeaderOffset = false
   340  				if len(fieldBuf) < 8 {
   341  					return ErrFormat
   342  				}
   343  				f.headerOffset = int64(fieldBuf.uint64())
   344  			}
   345  		case ntfsExtraID:
   346  			if len(fieldBuf) < 4 {
   347  				continue parseExtras
   348  			}
   349  			fieldBuf.uint32()        // reserved (ignored)
   350  			for len(fieldBuf) >= 4 { // need at least tag and size
   351  				attrTag := fieldBuf.uint16()
   352  				attrSize := int(fieldBuf.uint16())
   353  				if len(fieldBuf) < attrSize {
   354  					continue parseExtras
   355  				}
   356  				attrBuf := fieldBuf.sub(attrSize)
   357  				if attrTag != 1 || attrSize != 24 {
   358  					continue // Ignore irrelevant attributes
   359  				}
   360  
   361  				const ticksPerSecond = 1e7    // Windows timestamp resolution
   362  				ts := int64(attrBuf.uint64()) // ModTime since Windows epoch
   363  				secs := int64(ts / ticksPerSecond)
   364  				nsecs := (1e9 / ticksPerSecond) * int64(ts%ticksPerSecond)
   365  				epoch := time.Date(1601, time.January, 1, 0, 0, 0, 0, time.UTC)
   366  				modified = time.Unix(epoch.Unix()+secs, nsecs)
   367  			}
   368  		case unixExtraID, infoZipUnixExtraID:
   369  			if len(fieldBuf) < 8 {
   370  				continue parseExtras
   371  			}
   372  			fieldBuf.uint32()              // AcTime (ignored)
   373  			ts := int64(fieldBuf.uint32()) // ModTime since Unix epoch
   374  			modified = time.Unix(ts, 0)
   375  		case extTimeExtraID:
   376  			if len(fieldBuf) < 5 || fieldBuf.uint8()&1 == 0 {
   377  				continue parseExtras
   378  			}
   379  			ts := int64(fieldBuf.uint32()) // ModTime since Unix epoch
   380  			modified = time.Unix(ts, 0)
   381  		}
   382  	}
   383  
   384  	msdosModified := msDosTimeToTime(f.ModifiedDate, f.ModifiedTime)
   385  	f.Modified = msdosModified
   386  	if !modified.IsZero() {
   387  		f.Modified = modified.UTC()
   388  
   389  		// If legacy MS-DOS timestamps are set, we can use the delta between
   390  		// the legacy and extended versions to estimate timezone offset.
   391  		//
   392  		// A non-UTC timezone is always used (even if offset is zero).
   393  		// Thus, FileHeader.Modified.Location() == time.UTC is useful for
   394  		// determining whether extended timestamps are present.
   395  		// This is necessary for users that need to do additional time
   396  		// calculations when dealing with legacy ZIP formats.
   397  		if f.ModifiedTime != 0 || f.ModifiedDate != 0 {
   398  			f.Modified = modified.In(timeZone(msdosModified.Sub(modified)))
   399  		}
   400  	}
   401  
   402  	// Assume that uncompressed size 2³²-1 could plausibly happen in
   403  	// an old zip32 file that was sharding inputs into the largest chunks
   404  	// possible (or is just malicious; search the web for 42.zip).
   405  	// If needUSize is true still, it means we didn't see a zip64 extension.
   406  	// As long as the compressed size is not also 2³²-1 (implausible)
   407  	// and the header is not also 2³²-1 (equally implausible),
   408  	// accept the uncompressed size 2³²-1 as valid.
   409  	// If nothing else, this keeps archive/zip working with 42.zip.
   410  	_ = needUSize
   411  
   412  	if needCSize || needHeaderOffset {
   413  		return ErrFormat
   414  	}
   415  
   416  	return nil
   417  }
   418  
   419  func readDataDescriptor(r io.Reader, f *File) error {
   420  	var buf [dataDescriptorLen]byte
   421  
   422  	// The spec says: "Although not originally assigned a
   423  	// signature, the value 0x08074b50 has commonly been adopted
   424  	// as a signature value for the data descriptor record.
   425  	// Implementers should be aware that ZIP files may be
   426  	// encountered with or without this signature marking data
   427  	// descriptors and should account for either case when reading
   428  	// ZIP files to ensure compatibility."
   429  	//
   430  	// dataDescriptorLen includes the size of the signature but
   431  	// first read just those 4 bytes to see if it exists.
   432  	if _, err := io.ReadFull(r, buf[:4]); err != nil {
   433  		return err
   434  	}
   435  	off := 0
   436  	maybeSig := readBuf(buf[:4])
   437  	if maybeSig.uint32() != dataDescriptorSignature {
   438  		// No data descriptor signature. Keep these four
   439  		// bytes.
   440  		off += 4
   441  	}
   442  	if _, err := io.ReadFull(r, buf[off:12]); err != nil {
   443  		return err
   444  	}
   445  	b := readBuf(buf[:12])
   446  	if b.uint32() != f.CRC32 {
   447  		return ErrChecksum
   448  	}
   449  
   450  	// The two sizes that follow here can be either 32 bits or 64 bits
   451  	// but the spec is not very clear on this and different
   452  	// interpretations has been made causing incompatibilities. We
   453  	// already have the sizes from the central directory so we can
   454  	// just ignore these.
   455  
   456  	return nil
   457  }
   458  
   459  func readDirectoryEnd(r io.ReaderAt, size int64) (dir *directoryEnd, err error) {
   460  	// look for directoryEndSignature in the last 1k, then in the last 65k
   461  	var buf []byte
   462  	var directoryEndOffset int64
   463  	for i, bLen := range []int64{1024, 65 * 1024} {
   464  		if bLen > size {
   465  			bLen = size
   466  		}
   467  		buf = make([]byte, int(bLen))
   468  		if _, err := r.ReadAt(buf, size-bLen); err != nil && err != io.EOF {
   469  			return nil, err
   470  		}
   471  		if p := findSignatureInBlock(buf); p >= 0 {
   472  			buf = buf[p:]
   473  			directoryEndOffset = size - bLen + int64(p)
   474  			break
   475  		}
   476  		if i == 1 || bLen == size {
   477  			return nil, ErrFormat
   478  		}
   479  	}
   480  
   481  	// read header into struct
   482  	b := readBuf(buf[4:]) // skip signature
   483  	d := &directoryEnd{
   484  		diskNbr:            uint32(b.uint16()),
   485  		dirDiskNbr:         uint32(b.uint16()),
   486  		dirRecordsThisDisk: uint64(b.uint16()),
   487  		directoryRecords:   uint64(b.uint16()),
   488  		directorySize:      uint64(b.uint32()),
   489  		directoryOffset:    uint64(b.uint32()),
   490  		commentLen:         b.uint16(),
   491  	}
   492  	l := int(d.commentLen)
   493  	if l > len(b) {
   494  		return nil, errors.New("zip: invalid comment length")
   495  	}
   496  	d.comment = string(b[:l])
   497  
   498  	// These values mean that the file can be a zip64 file
   499  	if d.directoryRecords == 0xffff || d.directorySize == 0xffff || d.directoryOffset == 0xffffffff {
   500  		p, err := findDirectory64End(r, directoryEndOffset)
   501  		if err == nil && p >= 0 {
   502  			err = readDirectory64End(r, p, d)
   503  		}
   504  		if err != nil {
   505  			return nil, err
   506  		}
   507  	}
   508  	// Make sure directoryOffset points to somewhere in our file.
   509  	if o := int64(d.directoryOffset); o < 0 || o >= size {
   510  		return nil, ErrFormat
   511  	}
   512  	return d, nil
   513  }
   514  
   515  // findDirectory64End tries to read the zip64 locator just before the
   516  // directory end and returns the offset of the zip64 directory end if
   517  // found.
   518  func findDirectory64End(r io.ReaderAt, directoryEndOffset int64) (int64, error) {
   519  	locOffset := directoryEndOffset - directory64LocLen
   520  	if locOffset < 0 {
   521  		return -1, nil // no need to look for a header outside the file
   522  	}
   523  	buf := make([]byte, directory64LocLen)
   524  	if _, err := r.ReadAt(buf, locOffset); err != nil {
   525  		return -1, err
   526  	}
   527  	b := readBuf(buf)
   528  	if sig := b.uint32(); sig != directory64LocSignature {
   529  		return -1, nil
   530  	}
   531  	if b.uint32() != 0 { // number of the disk with the start of the zip64 end of central directory
   532  		return -1, nil // the file is not a valid zip64-file
   533  	}
   534  	p := b.uint64()      // relative offset of the zip64 end of central directory record
   535  	if b.uint32() != 1 { // total number of disks
   536  		return -1, nil // the file is not a valid zip64-file
   537  	}
   538  	return int64(p), nil
   539  }
   540  
   541  // readDirectory64End reads the zip64 directory end and updates the
   542  // directory end with the zip64 directory end values.
   543  func readDirectory64End(r io.ReaderAt, offset int64, d *directoryEnd) (err error) {
   544  	buf := make([]byte, directory64EndLen)
   545  	if _, err := r.ReadAt(buf, offset); err != nil {
   546  		return err
   547  	}
   548  
   549  	b := readBuf(buf)
   550  	if sig := b.uint32(); sig != directory64EndSignature {
   551  		return ErrFormat
   552  	}
   553  
   554  	b = b[12:]                        // skip dir size, version and version needed (uint64 + 2x uint16)
   555  	d.diskNbr = b.uint32()            // number of this disk
   556  	d.dirDiskNbr = b.uint32()         // number of the disk with the start of the central directory
   557  	d.dirRecordsThisDisk = b.uint64() // total number of entries in the central directory on this disk
   558  	d.directoryRecords = b.uint64()   // total number of entries in the central directory
   559  	d.directorySize = b.uint64()      // size of the central directory
   560  	d.directoryOffset = b.uint64()    // offset of start of central directory with respect to the starting disk number
   561  
   562  	return nil
   563  }
   564  
   565  func findSignatureInBlock(b []byte) int {
   566  	for i := len(b) - directoryEndLen; i >= 0; i-- {
   567  		// defined from directoryEndSignature in struct.go
   568  		if b[i] == 'P' && b[i+1] == 'K' && b[i+2] == 0x05 && b[i+3] == 0x06 {
   569  			// n is length of comment
   570  			n := int(b[i+directoryEndLen-2]) | int(b[i+directoryEndLen-1])<<8
   571  			if n+directoryEndLen+i <= len(b) {
   572  				return i
   573  			}
   574  		}
   575  	}
   576  	return -1
   577  }
   578  
   579  type readBuf []byte
   580  
   581  func (b *readBuf) uint8() uint8 {
   582  	v := (*b)[0]
   583  	*b = (*b)[1:]
   584  	return v
   585  }
   586  
   587  func (b *readBuf) uint16() uint16 {
   588  	v := binary.LittleEndian.Uint16(*b)
   589  	*b = (*b)[2:]
   590  	return v
   591  }
   592  
   593  func (b *readBuf) uint32() uint32 {
   594  	v := binary.LittleEndian.Uint32(*b)
   595  	*b = (*b)[4:]
   596  	return v
   597  }
   598  
   599  func (b *readBuf) uint64() uint64 {
   600  	v := binary.LittleEndian.Uint64(*b)
   601  	*b = (*b)[8:]
   602  	return v
   603  }
   604  
   605  func (b *readBuf) sub(n int) readBuf {
   606  	b2 := (*b)[:n]
   607  	*b = (*b)[n:]
   608  	return b2
   609  }
   610  

View as plain text