Source file src/archive/zip/reader.go

     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  	"io/fs"
    15  	"os"
    16  	"path"
    17  	"sort"
    18  	"strings"
    19  	"sync"
    20  	"time"
    21  )
    22  
    23  var (
    24  	ErrFormat    = errors.New("zip: not a valid zip file")
    25  	ErrAlgorithm = errors.New("zip: unsupported compression algorithm")
    26  	ErrChecksum  = errors.New("zip: checksum error")
    27  )
    28  
    29  // A Reader serves content from a ZIP archive.
    30  type Reader struct {
    31  	r             io.ReaderAt
    32  	File          []*File
    33  	Comment       string
    34  	decompressors map[uint16]Decompressor
    35  
    36  	// fileList is a list of files sorted by ename,
    37  	// for use by the Open method.
    38  	fileListOnce sync.Once
    39  	fileList     []fileListEntry
    40  }
    41  
    42  // A ReadCloser is a Reader that must be closed when no longer needed.
    43  type ReadCloser struct {
    44  	f *os.File
    45  	Reader
    46  }
    47  
    48  // A File is a single file in a ZIP archive.
    49  // The file information is in the embedded FileHeader.
    50  // The file content can be accessed by calling Open.
    51  type File struct {
    52  	FileHeader
    53  	zip          *Reader
    54  	zipr         io.ReaderAt
    55  	headerOffset int64
    56  	zip64        bool  // zip64 extended information extra field presence
    57  	descErr      error // error reading the data descriptor during init
    58  }
    59  
    60  // OpenReader will open the Zip file specified by name and return a ReadCloser.
    61  func OpenReader(name string) (*ReadCloser, error) {
    62  	f, err := os.Open(name)
    63  	if err != nil {
    64  		return nil, err
    65  	}
    66  	fi, err := f.Stat()
    67  	if err != nil {
    68  		f.Close()
    69  		return nil, err
    70  	}
    71  	r := new(ReadCloser)
    72  	if err := r.init(f, fi.Size()); err != nil {
    73  		f.Close()
    74  		return nil, err
    75  	}
    76  	r.f = f
    77  	return r, nil
    78  }
    79  
    80  // NewReader returns a new Reader reading from r, which is assumed to
    81  // have the given size in bytes.
    82  func NewReader(r io.ReaderAt, size int64) (*Reader, error) {
    83  	if size < 0 {
    84  		return nil, errors.New("zip: size cannot be negative")
    85  	}
    86  	zr := new(Reader)
    87  	if err := zr.init(r, size); err != nil {
    88  		return nil, err
    89  	}
    90  	return zr, nil
    91  }
    92  
    93  func (z *Reader) init(r io.ReaderAt, size int64) error {
    94  	end, err := readDirectoryEnd(r, size)
    95  	if err != nil {
    96  		return err
    97  	}
    98  	z.r = r
    99  	// Since the number of directory records is not validated, it is not
   100  	// safe to preallocate z.File without first checking that the specified
   101  	// number of files is reasonable, since a malformed archive may
   102  	// indicate it contains up to 1 << 128 - 1 files. Since each file has a
   103  	// header which will be _at least_ 30 bytes we can safely preallocate
   104  	// if (data size / 30) >= end.directoryRecords.
   105  	if end.directorySize < uint64(size) && (uint64(size)-end.directorySize)/30 >= end.directoryRecords {
   106  		z.File = make([]*File, 0, end.directoryRecords)
   107  	}
   108  	z.Comment = end.comment
   109  	rs := io.NewSectionReader(r, 0, size)
   110  	if _, err = rs.Seek(int64(end.directoryOffset), io.SeekStart); err != nil {
   111  		return err
   112  	}
   113  	buf := bufio.NewReader(rs)
   114  
   115  	// The count of files inside a zip is truncated to fit in a uint16.
   116  	// Gloss over this by reading headers until we encounter
   117  	// a bad one, and then only report an ErrFormat or UnexpectedEOF if
   118  	// the file count modulo 65536 is incorrect.
   119  	for {
   120  		f := &File{zip: z, zipr: r}
   121  		err = readDirectoryHeader(f, buf)
   122  		if err == ErrFormat || err == io.ErrUnexpectedEOF {
   123  			break
   124  		}
   125  		if err != nil {
   126  			return err
   127  		}
   128  		z.File = append(z.File, f)
   129  	}
   130  	if uint16(len(z.File)) != uint16(end.directoryRecords) { // only compare 16 bits here
   131  		// Return the readDirectoryHeader error if we read
   132  		// the wrong number of directory entries.
   133  		return err
   134  	}
   135  	return nil
   136  }
   137  
   138  // RegisterDecompressor registers or overrides a custom decompressor for a
   139  // specific method ID. If a decompressor for a given method is not found,
   140  // Reader will default to looking up the decompressor at the package level.
   141  func (z *Reader) RegisterDecompressor(method uint16, dcomp Decompressor) {
   142  	if z.decompressors == nil {
   143  		z.decompressors = make(map[uint16]Decompressor)
   144  	}
   145  	z.decompressors[method] = dcomp
   146  }
   147  
   148  func (z *Reader) decompressor(method uint16) Decompressor {
   149  	dcomp := z.decompressors[method]
   150  	if dcomp == nil {
   151  		dcomp = decompressor(method)
   152  	}
   153  	return dcomp
   154  }
   155  
   156  // Close closes the Zip file, rendering it unusable for I/O.
   157  func (rc *ReadCloser) Close() error {
   158  	return rc.f.Close()
   159  }
   160  
   161  // DataOffset returns the offset of the file's possibly-compressed
   162  // data, relative to the beginning of the zip file.
   163  //
   164  // Most callers should instead use Open, which transparently
   165  // decompresses data and verifies checksums.
   166  func (f *File) DataOffset() (offset int64, err error) {
   167  	bodyOffset, err := f.findBodyOffset()
   168  	if err != nil {
   169  		return
   170  	}
   171  	return f.headerOffset + bodyOffset, nil
   172  }
   173  
   174  // Open returns a ReadCloser that provides access to the File's contents.
   175  // Multiple files may be read concurrently.
   176  func (f *File) Open() (io.ReadCloser, error) {
   177  	bodyOffset, err := f.findBodyOffset()
   178  	if err != nil {
   179  		return nil, err
   180  	}
   181  	size := int64(f.CompressedSize64)
   182  	r := io.NewSectionReader(f.zipr, f.headerOffset+bodyOffset, size)
   183  	dcomp := f.zip.decompressor(f.Method)
   184  	if dcomp == nil {
   185  		return nil, ErrAlgorithm
   186  	}
   187  	var rc io.ReadCloser = dcomp(r)
   188  	var desr io.Reader
   189  	if f.hasDataDescriptor() {
   190  		desr = io.NewSectionReader(f.zipr, f.headerOffset+bodyOffset+size, dataDescriptorLen)
   191  	}
   192  	rc = &checksumReader{
   193  		rc:   rc,
   194  		hash: crc32.NewIEEE(),
   195  		f:    f,
   196  		desr: desr,
   197  	}
   198  	return rc, nil
   199  }
   200  
   201  // OpenRaw returns a Reader that provides access to the File's contents without
   202  // decompression.
   203  func (f *File) OpenRaw() (io.Reader, error) {
   204  	bodyOffset, err := f.findBodyOffset()
   205  	if err != nil {
   206  		return nil, err
   207  	}
   208  	r := io.NewSectionReader(f.zipr, f.headerOffset+bodyOffset, int64(f.CompressedSize64))
   209  	return r, nil
   210  }
   211  
   212  type checksumReader struct {
   213  	rc    io.ReadCloser
   214  	hash  hash.Hash32
   215  	nread uint64 // number of bytes read so far
   216  	f     *File
   217  	desr  io.Reader // if non-nil, where to read the data descriptor
   218  	err   error     // sticky error
   219  }
   220  
   221  func (r *checksumReader) Stat() (fs.FileInfo, error) {
   222  	return headerFileInfo{&r.f.FileHeader}, nil
   223  }
   224  
   225  func (r *checksumReader) Read(b []byte) (n int, err error) {
   226  	if r.err != nil {
   227  		return 0, r.err
   228  	}
   229  	n, err = r.rc.Read(b)
   230  	r.hash.Write(b[:n])
   231  	r.nread += uint64(n)
   232  	if err == nil {
   233  		return
   234  	}
   235  	if err == io.EOF {
   236  		if r.nread != r.f.UncompressedSize64 {
   237  			return 0, io.ErrUnexpectedEOF
   238  		}
   239  		if r.desr != nil {
   240  			if err1 := readDataDescriptor(r.desr, r.f); err1 != nil {
   241  				if err1 == io.EOF {
   242  					err = io.ErrUnexpectedEOF
   243  				} else {
   244  					err = err1
   245  				}
   246  			} else if r.hash.Sum32() != r.f.CRC32 {
   247  				err = ErrChecksum
   248  			}
   249  		} else {
   250  			// If there's not a data descriptor, we still compare
   251  			// the CRC32 of what we've read against the file header
   252  			// or TOC's CRC32, if it seems like it was set.
   253  			if r.f.CRC32 != 0 && r.hash.Sum32() != r.f.CRC32 {
   254  				err = ErrChecksum
   255  			}
   256  		}
   257  	}
   258  	r.err = err
   259  	return
   260  }
   261  
   262  func (r *checksumReader) Close() error { return r.rc.Close() }
   263  
   264  // findBodyOffset does the minimum work to verify the file has a header
   265  // and returns the file body offset.
   266  func (f *File) findBodyOffset() (int64, error) {
   267  	var buf [fileHeaderLen]byte
   268  	if _, err := f.zipr.ReadAt(buf[:], f.headerOffset); err != nil {
   269  		return 0, err
   270  	}
   271  	b := readBuf(buf[:])
   272  	if sig := b.uint32(); sig != fileHeaderSignature {
   273  		return 0, ErrFormat
   274  	}
   275  	b = b[22:] // skip over most of the header
   276  	filenameLen := int(b.uint16())
   277  	extraLen := int(b.uint16())
   278  	return int64(fileHeaderLen + filenameLen + extraLen), nil
   279  }
   280  
   281  // readDirectoryHeader attempts to read a directory header from r.
   282  // It returns io.ErrUnexpectedEOF if it cannot read a complete header,
   283  // and ErrFormat if it doesn't find a valid header signature.
   284  func readDirectoryHeader(f *File, r io.Reader) error {
   285  	var buf [directoryHeaderLen]byte
   286  	if _, err := io.ReadFull(r, buf[:]); err != nil {
   287  		return err
   288  	}
   289  	b := readBuf(buf[:])
   290  	if sig := b.uint32(); sig != directoryHeaderSignature {
   291  		return ErrFormat
   292  	}
   293  	f.CreatorVersion = b.uint16()
   294  	f.ReaderVersion = b.uint16()
   295  	f.Flags = b.uint16()
   296  	f.Method = b.uint16()
   297  	f.ModifiedTime = b.uint16()
   298  	f.ModifiedDate = b.uint16()
   299  	f.CRC32 = b.uint32()
   300  	f.CompressedSize = b.uint32()
   301  	f.UncompressedSize = b.uint32()
   302  	f.CompressedSize64 = uint64(f.CompressedSize)
   303  	f.UncompressedSize64 = uint64(f.UncompressedSize)
   304  	filenameLen := int(b.uint16())
   305  	extraLen := int(b.uint16())
   306  	commentLen := int(b.uint16())
   307  	b = b[4:] // skipped start disk number and internal attributes (2x uint16)
   308  	f.ExternalAttrs = b.uint32()
   309  	f.headerOffset = int64(b.uint32())
   310  	d := make([]byte, filenameLen+extraLen+commentLen)
   311  	if _, err := io.ReadFull(r, d); err != nil {
   312  		return err
   313  	}
   314  	f.Name = string(d[:filenameLen])
   315  	f.Extra = d[filenameLen : filenameLen+extraLen]
   316  	f.Comment = string(d[filenameLen+extraLen:])
   317  
   318  	// Determine the character encoding.
   319  	utf8Valid1, utf8Require1 := detectUTF8(f.Name)
   320  	utf8Valid2, utf8Require2 := detectUTF8(f.Comment)
   321  	switch {
   322  	case !utf8Valid1 || !utf8Valid2:
   323  		// Name and Comment definitely not UTF-8.
   324  		f.NonUTF8 = true
   325  	case !utf8Require1 && !utf8Require2:
   326  		// Name and Comment use only single-byte runes that overlap with UTF-8.
   327  		f.NonUTF8 = false
   328  	default:
   329  		// Might be UTF-8, might be some other encoding; preserve existing flag.
   330  		// Some ZIP writers use UTF-8 encoding without setting the UTF-8 flag.
   331  		// Since it is impossible to always distinguish valid UTF-8 from some
   332  		// other encoding (e.g., GBK or Shift-JIS), we trust the flag.
   333  		f.NonUTF8 = f.Flags&0x800 == 0
   334  	}
   335  
   336  	needUSize := f.UncompressedSize == ^uint32(0)
   337  	needCSize := f.CompressedSize == ^uint32(0)
   338  	needHeaderOffset := f.headerOffset == int64(^uint32(0))
   339  
   340  	// Best effort to find what we need.
   341  	// Other zip authors might not even follow the basic format,
   342  	// and we'll just ignore the Extra content in that case.
   343  	var modified time.Time
   344  parseExtras:
   345  	for extra := readBuf(f.Extra); len(extra) >= 4; { // need at least tag and size
   346  		fieldTag := extra.uint16()
   347  		fieldSize := int(extra.uint16())
   348  		if len(extra) < fieldSize {
   349  			break
   350  		}
   351  		fieldBuf := extra.sub(fieldSize)
   352  
   353  		switch fieldTag {
   354  		case zip64ExtraID:
   355  			f.zip64 = true
   356  
   357  			// update directory values from the zip64 extra block.
   358  			// They should only be consulted if the sizes read earlier
   359  			// are maxed out.
   360  			// See golang.org/issue/13367.
   361  			if needUSize {
   362  				needUSize = false
   363  				if len(fieldBuf) < 8 {
   364  					return ErrFormat
   365  				}
   366  				f.UncompressedSize64 = fieldBuf.uint64()
   367  			}
   368  			if needCSize {
   369  				needCSize = false
   370  				if len(fieldBuf) < 8 {
   371  					return ErrFormat
   372  				}
   373  				f.CompressedSize64 = fieldBuf.uint64()
   374  			}
   375  			if needHeaderOffset {
   376  				needHeaderOffset = false
   377  				if len(fieldBuf) < 8 {
   378  					return ErrFormat
   379  				}
   380  				f.headerOffset = int64(fieldBuf.uint64())
   381  			}
   382  		case ntfsExtraID:
   383  			if len(fieldBuf) < 4 {
   384  				continue parseExtras
   385  			}
   386  			fieldBuf.uint32()        // reserved (ignored)
   387  			for len(fieldBuf) >= 4 { // need at least tag and size
   388  				attrTag := fieldBuf.uint16()
   389  				attrSize := int(fieldBuf.uint16())
   390  				if len(fieldBuf) < attrSize {
   391  					continue parseExtras
   392  				}
   393  				attrBuf := fieldBuf.sub(attrSize)
   394  				if attrTag != 1 || attrSize != 24 {
   395  					continue // Ignore irrelevant attributes
   396  				}
   397  
   398  				const ticksPerSecond = 1e7    // Windows timestamp resolution
   399  				ts := int64(attrBuf.uint64()) // ModTime since Windows epoch
   400  				secs := int64(ts / ticksPerSecond)
   401  				nsecs := (1e9 / ticksPerSecond) * int64(ts%ticksPerSecond)
   402  				epoch := time.Date(1601, time.January, 1, 0, 0, 0, 0, time.UTC)
   403  				modified = time.Unix(epoch.Unix()+secs, nsecs)
   404  			}
   405  		case unixExtraID, infoZipUnixExtraID:
   406  			if len(fieldBuf) < 8 {
   407  				continue parseExtras
   408  			}
   409  			fieldBuf.uint32()              // AcTime (ignored)
   410  			ts := int64(fieldBuf.uint32()) // ModTime since Unix epoch
   411  			modified = time.Unix(ts, 0)
   412  		case extTimeExtraID:
   413  			if len(fieldBuf) < 5 || fieldBuf.uint8()&1 == 0 {
   414  				continue parseExtras
   415  			}
   416  			ts := int64(fieldBuf.uint32()) // ModTime since Unix epoch
   417  			modified = time.Unix(ts, 0)
   418  		}
   419  	}
   420  
   421  	msdosModified := msDosTimeToTime(f.ModifiedDate, f.ModifiedTime)
   422  	f.Modified = msdosModified
   423  	if !modified.IsZero() {
   424  		f.Modified = modified.UTC()
   425  
   426  		// If legacy MS-DOS timestamps are set, we can use the delta between
   427  		// the legacy and extended versions to estimate timezone offset.
   428  		//
   429  		// A non-UTC timezone is always used (even if offset is zero).
   430  		// Thus, FileHeader.Modified.Location() == time.UTC is useful for
   431  		// determining whether extended timestamps are present.
   432  		// This is necessary for users that need to do additional time
   433  		// calculations when dealing with legacy ZIP formats.
   434  		if f.ModifiedTime != 0 || f.ModifiedDate != 0 {
   435  			f.Modified = modified.In(timeZone(msdosModified.Sub(modified)))
   436  		}
   437  	}
   438  
   439  	// Assume that uncompressed size 2³²-1 could plausibly happen in
   440  	// an old zip32 file that was sharding inputs into the largest chunks
   441  	// possible (or is just malicious; search the web for 42.zip).
   442  	// If needUSize is true still, it means we didn't see a zip64 extension.
   443  	// As long as the compressed size is not also 2³²-1 (implausible)
   444  	// and the header is not also 2³²-1 (equally implausible),
   445  	// accept the uncompressed size 2³²-1 as valid.
   446  	// If nothing else, this keeps archive/zip working with 42.zip.
   447  	_ = needUSize
   448  
   449  	if needCSize || needHeaderOffset {
   450  		return ErrFormat
   451  	}
   452  
   453  	return nil
   454  }
   455  
   456  func readDataDescriptor(r io.Reader, f *File) error {
   457  	var buf [dataDescriptorLen]byte
   458  	// The spec says: "Although not originally assigned a
   459  	// signature, the value 0x08074b50 has commonly been adopted
   460  	// as a signature value for the data descriptor record.
   461  	// Implementers should be aware that ZIP files may be
   462  	// encountered with or without this signature marking data
   463  	// descriptors and should account for either case when reading
   464  	// ZIP files to ensure compatibility."
   465  	//
   466  	// dataDescriptorLen includes the size of the signature but
   467  	// first read just those 4 bytes to see if it exists.
   468  	if _, err := io.ReadFull(r, buf[:4]); err != nil {
   469  		return err
   470  	}
   471  	off := 0
   472  	maybeSig := readBuf(buf[:4])
   473  	if maybeSig.uint32() != dataDescriptorSignature {
   474  		// No data descriptor signature. Keep these four
   475  		// bytes.
   476  		off += 4
   477  	}
   478  	if _, err := io.ReadFull(r, buf[off:12]); err != nil {
   479  		return err
   480  	}
   481  	b := readBuf(buf[:12])
   482  	if b.uint32() != f.CRC32 {
   483  		return ErrChecksum
   484  	}
   485  
   486  	// The two sizes that follow here can be either 32 bits or 64 bits
   487  	// but the spec is not very clear on this and different
   488  	// interpretations has been made causing incompatibilities. We
   489  	// already have the sizes from the central directory so we can
   490  	// just ignore these.
   491  
   492  	return nil
   493  }
   494  
   495  func readDirectoryEnd(r io.ReaderAt, size int64) (dir *directoryEnd, err error) {
   496  	// look for directoryEndSignature in the last 1k, then in the last 65k
   497  	var buf []byte
   498  	var directoryEndOffset int64
   499  	for i, bLen := range []int64{1024, 65 * 1024} {
   500  		if bLen > size {
   501  			bLen = size
   502  		}
   503  		buf = make([]byte, int(bLen))
   504  		if _, err := r.ReadAt(buf, size-bLen); err != nil && err != io.EOF {
   505  			return nil, err
   506  		}
   507  		if p := findSignatureInBlock(buf); p >= 0 {
   508  			buf = buf[p:]
   509  			directoryEndOffset = size - bLen + int64(p)
   510  			break
   511  		}
   512  		if i == 1 || bLen == size {
   513  			return nil, ErrFormat
   514  		}
   515  	}
   516  
   517  	// read header into struct
   518  	b := readBuf(buf[4:]) // skip signature
   519  	d := &directoryEnd{
   520  		diskNbr:            uint32(b.uint16()),
   521  		dirDiskNbr:         uint32(b.uint16()),
   522  		dirRecordsThisDisk: uint64(b.uint16()),
   523  		directoryRecords:   uint64(b.uint16()),
   524  		directorySize:      uint64(b.uint32()),
   525  		directoryOffset:    uint64(b.uint32()),
   526  		commentLen:         b.uint16(),
   527  	}
   528  	l := int(d.commentLen)
   529  	if l > len(b) {
   530  		return nil, errors.New("zip: invalid comment length")
   531  	}
   532  	d.comment = string(b[:l])
   533  
   534  	// These values mean that the file can be a zip64 file
   535  	if d.directoryRecords == 0xffff || d.directorySize == 0xffff || d.directoryOffset == 0xffffffff {
   536  		p, err := findDirectory64End(r, directoryEndOffset)
   537  		if err == nil && p >= 0 {
   538  			err = readDirectory64End(r, p, d)
   539  		}
   540  		if err != nil {
   541  			return nil, err
   542  		}
   543  	}
   544  	// Make sure directoryOffset points to somewhere in our file.
   545  	if o := int64(d.directoryOffset); o < 0 || o >= size {
   546  		return nil, ErrFormat
   547  	}
   548  	return d, nil
   549  }
   550  
   551  // findDirectory64End tries to read the zip64 locator just before the
   552  // directory end and returns the offset of the zip64 directory end if
   553  // found.
   554  func findDirectory64End(r io.ReaderAt, directoryEndOffset int64) (int64, error) {
   555  	locOffset := directoryEndOffset - directory64LocLen
   556  	if locOffset < 0 {
   557  		return -1, nil // no need to look for a header outside the file
   558  	}
   559  	buf := make([]byte, directory64LocLen)
   560  	if _, err := r.ReadAt(buf, locOffset); err != nil {
   561  		return -1, err
   562  	}
   563  	b := readBuf(buf)
   564  	if sig := b.uint32(); sig != directory64LocSignature {
   565  		return -1, nil
   566  	}
   567  	if b.uint32() != 0 { // number of the disk with the start of the zip64 end of central directory
   568  		return -1, nil // the file is not a valid zip64-file
   569  	}
   570  	p := b.uint64()      // relative offset of the zip64 end of central directory record
   571  	if b.uint32() != 1 { // total number of disks
   572  		return -1, nil // the file is not a valid zip64-file
   573  	}
   574  	return int64(p), nil
   575  }
   576  
   577  // readDirectory64End reads the zip64 directory end and updates the
   578  // directory end with the zip64 directory end values.
   579  func readDirectory64End(r io.ReaderAt, offset int64, d *directoryEnd) (err error) {
   580  	buf := make([]byte, directory64EndLen)
   581  	if _, err := r.ReadAt(buf, offset); err != nil {
   582  		return err
   583  	}
   584  
   585  	b := readBuf(buf)
   586  	if sig := b.uint32(); sig != directory64EndSignature {
   587  		return ErrFormat
   588  	}
   589  
   590  	b = b[12:]                        // skip dir size, version and version needed (uint64 + 2x uint16)
   591  	d.diskNbr = b.uint32()            // number of this disk
   592  	d.dirDiskNbr = b.uint32()         // number of the disk with the start of the central directory
   593  	d.dirRecordsThisDisk = b.uint64() // total number of entries in the central directory on this disk
   594  	d.directoryRecords = b.uint64()   // total number of entries in the central directory
   595  	d.directorySize = b.uint64()      // size of the central directory
   596  	d.directoryOffset = b.uint64()    // offset of start of central directory with respect to the starting disk number
   597  
   598  	return nil
   599  }
   600  
   601  func findSignatureInBlock(b []byte) int {
   602  	for i := len(b) - directoryEndLen; i >= 0; i-- {
   603  		// defined from directoryEndSignature in struct.go
   604  		if b[i] == 'P' && b[i+1] == 'K' && b[i+2] == 0x05 && b[i+3] == 0x06 {
   605  			// n is length of comment
   606  			n := int(b[i+directoryEndLen-2]) | int(b[i+directoryEndLen-1])<<8
   607  			if n+directoryEndLen+i <= len(b) {
   608  				return i
   609  			}
   610  		}
   611  	}
   612  	return -1
   613  }
   614  
   615  type readBuf []byte
   616  
   617  func (b *readBuf) uint8() uint8 {
   618  	v := (*b)[0]
   619  	*b = (*b)[1:]
   620  	return v
   621  }
   622  
   623  func (b *readBuf) uint16() uint16 {
   624  	v := binary.LittleEndian.Uint16(*b)
   625  	*b = (*b)[2:]
   626  	return v
   627  }
   628  
   629  func (b *readBuf) uint32() uint32 {
   630  	v := binary.LittleEndian.Uint32(*b)
   631  	*b = (*b)[4:]
   632  	return v
   633  }
   634  
   635  func (b *readBuf) uint64() uint64 {
   636  	v := binary.LittleEndian.Uint64(*b)
   637  	*b = (*b)[8:]
   638  	return v
   639  }
   640  
   641  func (b *readBuf) sub(n int) readBuf {
   642  	b2 := (*b)[:n]
   643  	*b = (*b)[n:]
   644  	return b2
   645  }
   646  
   647  // A fileListEntry is a File and its ename.
   648  // If file == nil, the fileListEntry describes a directory without metadata.
   649  type fileListEntry struct {
   650  	name  string
   651  	file  *File
   652  	isDir bool
   653  }
   654  
   655  type fileInfoDirEntry interface {
   656  	fs.FileInfo
   657  	fs.DirEntry
   658  }
   659  
   660  func (e *fileListEntry) stat() fileInfoDirEntry {
   661  	if !e.isDir {
   662  		return headerFileInfo{&e.file.FileHeader}
   663  	}
   664  	return e
   665  }
   666  
   667  // Only used for directories.
   668  func (f *fileListEntry) Name() string      { _, elem, _ := split(f.name); return elem }
   669  func (f *fileListEntry) Size() int64       { return 0 }
   670  func (f *fileListEntry) Mode() fs.FileMode { return fs.ModeDir | 0555 }
   671  func (f *fileListEntry) Type() fs.FileMode { return fs.ModeDir }
   672  func (f *fileListEntry) IsDir() bool       { return true }
   673  func (f *fileListEntry) Sys() any          { return nil }
   674  
   675  func (f *fileListEntry) ModTime() time.Time {
   676  	if f.file == nil {
   677  		return time.Time{}
   678  	}
   679  	return f.file.FileHeader.Modified.UTC()
   680  }
   681  
   682  func (f *fileListEntry) Info() (fs.FileInfo, error) { return f, nil }
   683  
   684  // toValidName coerces name to be a valid name for fs.FS.Open.
   685  func toValidName(name string) string {
   686  	name = strings.ReplaceAll(name, `\`, `/`)
   687  	p := path.Clean(name)
   688  	if strings.HasPrefix(p, "/") {
   689  		p = p[len("/"):]
   690  	}
   691  	for strings.HasPrefix(p, "../") {
   692  		p = p[len("../"):]
   693  	}
   694  	return p
   695  }
   696  
   697  func (r *Reader) initFileList() {
   698  	r.fileListOnce.Do(func() {
   699  		dirs := make(map[string]bool)
   700  		knownDirs := make(map[string]bool)
   701  		for _, file := range r.File {
   702  			isDir := len(file.Name) > 0 && file.Name[len(file.Name)-1] == '/'
   703  			name := toValidName(file.Name)
   704  			if name == "" {
   705  				continue
   706  			}
   707  			for dir := path.Dir(name); dir != "."; dir = path.Dir(dir) {
   708  				dirs[dir] = true
   709  			}
   710  			entry := fileListEntry{
   711  				name:  name,
   712  				file:  file,
   713  				isDir: isDir,
   714  			}
   715  			r.fileList = append(r.fileList, entry)
   716  			if isDir {
   717  				knownDirs[name] = true
   718  			}
   719  		}
   720  		for dir := range dirs {
   721  			if !knownDirs[dir] {
   722  				entry := fileListEntry{
   723  					name:  dir,
   724  					file:  nil,
   725  					isDir: true,
   726  				}
   727  				r.fileList = append(r.fileList, entry)
   728  			}
   729  		}
   730  
   731  		sort.Slice(r.fileList, func(i, j int) bool { return fileEntryLess(r.fileList[i].name, r.fileList[j].name) })
   732  	})
   733  }
   734  
   735  func fileEntryLess(x, y string) bool {
   736  	xdir, xelem, _ := split(x)
   737  	ydir, yelem, _ := split(y)
   738  	return xdir < ydir || xdir == ydir && xelem < yelem
   739  }
   740  
   741  // Open opens the named file in the ZIP archive,
   742  // using the semantics of fs.FS.Open:
   743  // paths are always slash separated, with no
   744  // leading / or ../ elements.
   745  func (r *Reader) Open(name string) (fs.File, error) {
   746  	r.initFileList()
   747  
   748  	if !fs.ValidPath(name) {
   749  		return nil, &fs.PathError{Op: "open", Path: name, Err: fs.ErrInvalid}
   750  	}
   751  	e := r.openLookup(name)
   752  	if e == nil {
   753  		return nil, &fs.PathError{Op: "open", Path: name, Err: fs.ErrNotExist}
   754  	}
   755  	if e.isDir {
   756  		return &openDir{e, r.openReadDir(name), 0}, nil
   757  	}
   758  	rc, err := e.file.Open()
   759  	if err != nil {
   760  		return nil, err
   761  	}
   762  	return rc.(fs.File), nil
   763  }
   764  
   765  func split(name string) (dir, elem string, isDir bool) {
   766  	if len(name) > 0 && name[len(name)-1] == '/' {
   767  		isDir = true
   768  		name = name[:len(name)-1]
   769  	}
   770  	i := len(name) - 1
   771  	for i >= 0 && name[i] != '/' {
   772  		i--
   773  	}
   774  	if i < 0 {
   775  		return ".", name, isDir
   776  	}
   777  	return name[:i], name[i+1:], isDir
   778  }
   779  
   780  var dotFile = &fileListEntry{name: "./", isDir: true}
   781  
   782  func (r *Reader) openLookup(name string) *fileListEntry {
   783  	if name == "." {
   784  		return dotFile
   785  	}
   786  
   787  	dir, elem, _ := split(name)
   788  	files := r.fileList
   789  	i := sort.Search(len(files), func(i int) bool {
   790  		idir, ielem, _ := split(files[i].name)
   791  		return idir > dir || idir == dir && ielem >= elem
   792  	})
   793  	if i < len(files) {
   794  		fname := files[i].name
   795  		if fname == name || len(fname) == len(name)+1 && fname[len(name)] == '/' && fname[:len(name)] == name {
   796  			return &files[i]
   797  		}
   798  	}
   799  	return nil
   800  }
   801  
   802  func (r *Reader) openReadDir(dir string) []fileListEntry {
   803  	files := r.fileList
   804  	i := sort.Search(len(files), func(i int) bool {
   805  		idir, _, _ := split(files[i].name)
   806  		return idir >= dir
   807  	})
   808  	j := sort.Search(len(files), func(j int) bool {
   809  		jdir, _, _ := split(files[j].name)
   810  		return jdir > dir
   811  	})
   812  	return files[i:j]
   813  }
   814  
   815  type openDir struct {
   816  	e      *fileListEntry
   817  	files  []fileListEntry
   818  	offset int
   819  }
   820  
   821  func (d *openDir) Close() error               { return nil }
   822  func (d *openDir) Stat() (fs.FileInfo, error) { return d.e.stat(), nil }
   823  
   824  func (d *openDir) Read([]byte) (int, error) {
   825  	return 0, &fs.PathError{Op: "read", Path: d.e.name, Err: errors.New("is a directory")}
   826  }
   827  
   828  func (d *openDir) ReadDir(count int) ([]fs.DirEntry, error) {
   829  	n := len(d.files) - d.offset
   830  	if count > 0 && n > count {
   831  		n = count
   832  	}
   833  	if n == 0 {
   834  		if count <= 0 {
   835  			return nil, nil
   836  		}
   837  		return nil, io.EOF
   838  	}
   839  	list := make([]fs.DirEntry, n)
   840  	for i := range list {
   841  		list[i] = d.files[d.offset+i].stat()
   842  	}
   843  	d.offset += n
   844  	return list, nil
   845  }
   846  

View as plain text