Source file src/time/zoneinfo_read.go

     1  // Copyright 2009 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  // Parse "zoneinfo" time zone file.
     6  // This is a fairly standard file format used on OS X, Linux, BSD, Sun, and others.
     7  // See tzfile(5), https://en.wikipedia.org/wiki/Zoneinfo,
     8  // and ftp://munnari.oz.au/pub/oldtz/
     9  
    10  package time
    11  
    12  import (
    13  	"errors"
    14  	"internal/bytealg"
    15  	"runtime"
    16  	"syscall"
    17  )
    18  
    19  // registerLoadFromEmbeddedTZData is called by the time/tzdata package,
    20  // if it is imported.
    21  func registerLoadFromEmbeddedTZData(f func(string) (string, error)) {
    22  	loadFromEmbeddedTZData = f
    23  }
    24  
    25  // loadFromEmbeddedTZData is used to load a specific tzdata file
    26  // from tzdata information embedded in the binary itself.
    27  // This is set when the time/tzdata package is imported,
    28  // via registerLoadFromEmbeddedTzdata.
    29  var loadFromEmbeddedTZData func(zipname string) (string, error)
    30  
    31  // maxFileSize is the max permitted size of files read by readFile.
    32  // As reference, the zoneinfo.zip distributed by Go is ~350 KB,
    33  // so 10MB is overkill.
    34  const maxFileSize = 10 << 20
    35  
    36  type fileSizeError string
    37  
    38  func (f fileSizeError) Error() string {
    39  	return "time: file " + string(f) + " is too large"
    40  }
    41  
    42  // Copies of io.Seek* constants to avoid importing "io":
    43  const (
    44  	seekStart   = 0
    45  	seekCurrent = 1
    46  	seekEnd     = 2
    47  )
    48  
    49  // Simple I/O interface to binary blob of data.
    50  type dataIO struct {
    51  	p     []byte
    52  	error bool
    53  }
    54  
    55  func (d *dataIO) read(n int) []byte {
    56  	if len(d.p) < n {
    57  		d.p = nil
    58  		d.error = true
    59  		return nil
    60  	}
    61  	p := d.p[0:n]
    62  	d.p = d.p[n:]
    63  	return p
    64  }
    65  
    66  func (d *dataIO) big4() (n uint32, ok bool) {
    67  	p := d.read(4)
    68  	if len(p) < 4 {
    69  		d.error = true
    70  		return 0, false
    71  	}
    72  	return uint32(p[3]) | uint32(p[2])<<8 | uint32(p[1])<<16 | uint32(p[0])<<24, true
    73  }
    74  
    75  func (d *dataIO) big8() (n uint64, ok bool) {
    76  	n1, ok1 := d.big4()
    77  	n2, ok2 := d.big4()
    78  	if !ok1 || !ok2 {
    79  		d.error = true
    80  		return 0, false
    81  	}
    82  	return (uint64(n1) << 32) | uint64(n2), true
    83  }
    84  
    85  func (d *dataIO) byte() (n byte, ok bool) {
    86  	p := d.read(1)
    87  	if len(p) < 1 {
    88  		d.error = true
    89  		return 0, false
    90  	}
    91  	return p[0], true
    92  }
    93  
    94  // rest returns the rest of the data in the buffer.
    95  func (d *dataIO) rest() []byte {
    96  	r := d.p
    97  	d.p = nil
    98  	return r
    99  }
   100  
   101  // Make a string by stopping at the first NUL
   102  func byteString(p []byte) string {
   103  	if i := bytealg.IndexByte(p, 0); i != -1 {
   104  		p = p[:i]
   105  	}
   106  	return string(p)
   107  }
   108  
   109  var errBadData = errors.New("malformed time zone information")
   110  
   111  // LoadLocationFromTZData returns a Location with the given name
   112  // initialized from the IANA Time Zone database-formatted data.
   113  // The data should be in the format of a standard IANA time zone file
   114  // (for example, the content of /etc/localtime on Unix systems).
   115  func LoadLocationFromTZData(name string, data []byte) (*Location, error) {
   116  	d := dataIO{data, false}
   117  
   118  	// 4-byte magic "TZif"
   119  	if magic := d.read(4); string(magic) != "TZif" {
   120  		return nil, errBadData
   121  	}
   122  
   123  	// 1-byte version, then 15 bytes of padding
   124  	var version int
   125  	var p []byte
   126  	if p = d.read(16); len(p) != 16 {
   127  		return nil, errBadData
   128  	} else {
   129  		switch p[0] {
   130  		case 0:
   131  			version = 1
   132  		case '2':
   133  			version = 2
   134  		case '3':
   135  			version = 3
   136  		default:
   137  			return nil, errBadData
   138  		}
   139  	}
   140  
   141  	// six big-endian 32-bit integers:
   142  	//	number of UTC/local indicators
   143  	//	number of standard/wall indicators
   144  	//	number of leap seconds
   145  	//	number of transition times
   146  	//	number of local time zones
   147  	//	number of characters of time zone abbrev strings
   148  	const (
   149  		NUTCLocal = iota
   150  		NStdWall
   151  		NLeap
   152  		NTime
   153  		NZone
   154  		NChar
   155  	)
   156  	var n [6]int
   157  	for i := 0; i < 6; i++ {
   158  		nn, ok := d.big4()
   159  		if !ok {
   160  			return nil, errBadData
   161  		}
   162  		if uint32(int(nn)) != nn {
   163  			return nil, errBadData
   164  		}
   165  		n[i] = int(nn)
   166  	}
   167  
   168  	// If we have version 2 or 3, then the data is first written out
   169  	// in a 32-bit format, then written out again in a 64-bit format.
   170  	// Skip the 32-bit format and read the 64-bit one, as it can
   171  	// describe a broader range of dates.
   172  
   173  	is64 := false
   174  	if version > 1 {
   175  		// Skip the 32-bit data.
   176  		skip := n[NTime]*4 +
   177  			n[NTime] +
   178  			n[NZone]*6 +
   179  			n[NChar] +
   180  			n[NLeap]*8 +
   181  			n[NStdWall] +
   182  			n[NUTCLocal]
   183  		// Skip the version 2 header that we just read.
   184  		skip += 4 + 16
   185  		d.read(skip)
   186  
   187  		is64 = true
   188  
   189  		// Read the counts again, they can differ.
   190  		for i := 0; i < 6; i++ {
   191  			nn, ok := d.big4()
   192  			if !ok {
   193  				return nil, errBadData
   194  			}
   195  			if uint32(int(nn)) != nn {
   196  				return nil, errBadData
   197  			}
   198  			n[i] = int(nn)
   199  		}
   200  	}
   201  
   202  	size := 4
   203  	if is64 {
   204  		size = 8
   205  	}
   206  
   207  	// Transition times.
   208  	txtimes := dataIO{d.read(n[NTime] * size), false}
   209  
   210  	// Time zone indices for transition times.
   211  	txzones := d.read(n[NTime])
   212  
   213  	// Zone info structures
   214  	zonedata := dataIO{d.read(n[NZone] * 6), false}
   215  
   216  	// Time zone abbreviations.
   217  	abbrev := d.read(n[NChar])
   218  
   219  	// Leap-second time pairs
   220  	d.read(n[NLeap] * (size + 4))
   221  
   222  	// Whether tx times associated with local time types
   223  	// are specified as standard time or wall time.
   224  	isstd := d.read(n[NStdWall])
   225  
   226  	// Whether tx times associated with local time types
   227  	// are specified as UTC or local time.
   228  	isutc := d.read(n[NUTCLocal])
   229  
   230  	if d.error { // ran out of data
   231  		return nil, errBadData
   232  	}
   233  
   234  	var extend string
   235  	rest := d.rest()
   236  	if len(rest) > 2 && rest[0] == '\n' && rest[len(rest)-1] == '\n' {
   237  		extend = string(rest[1 : len(rest)-1])
   238  	}
   239  
   240  	// Now we can build up a useful data structure.
   241  	// First the zone information.
   242  	//	utcoff[4] isdst[1] nameindex[1]
   243  	nzone := n[NZone]
   244  	if nzone == 0 {
   245  		// Reject tzdata files with no zones. There's nothing useful in them.
   246  		// This also avoids a panic later when we add and then use a fake transition (golang.org/issue/29437).
   247  		return nil, errBadData
   248  	}
   249  	zones := make([]zone, nzone)
   250  	for i := range zones {
   251  		var ok bool
   252  		var n uint32
   253  		if n, ok = zonedata.big4(); !ok {
   254  			return nil, errBadData
   255  		}
   256  		if uint32(int(n)) != n {
   257  			return nil, errBadData
   258  		}
   259  		zones[i].offset = int(int32(n))
   260  		var b byte
   261  		if b, ok = zonedata.byte(); !ok {
   262  			return nil, errBadData
   263  		}
   264  		zones[i].isDST = b != 0
   265  		if b, ok = zonedata.byte(); !ok || int(b) >= len(abbrev) {
   266  			return nil, errBadData
   267  		}
   268  		zones[i].name = byteString(abbrev[b:])
   269  		if runtime.GOOS == "aix" && len(name) > 8 && (name[:8] == "Etc/GMT+" || name[:8] == "Etc/GMT-") {
   270  			// There is a bug with AIX 7.2 TL 0 with files in Etc,
   271  			// GMT+1 will return GMT-1 instead of GMT+1 or -01.
   272  			if name != "Etc/GMT+0" {
   273  				// GMT+0 is OK
   274  				zones[i].name = name[4:]
   275  			}
   276  		}
   277  	}
   278  
   279  	// Now the transition time info.
   280  	tx := make([]zoneTrans, n[NTime])
   281  	for i := range tx {
   282  		var n int64
   283  		if !is64 {
   284  			if n4, ok := txtimes.big4(); !ok {
   285  				return nil, errBadData
   286  			} else {
   287  				n = int64(int32(n4))
   288  			}
   289  		} else {
   290  			if n8, ok := txtimes.big8(); !ok {
   291  				return nil, errBadData
   292  			} else {
   293  				n = int64(n8)
   294  			}
   295  		}
   296  		tx[i].when = n
   297  		if int(txzones[i]) >= len(zones) {
   298  			return nil, errBadData
   299  		}
   300  		tx[i].index = txzones[i]
   301  		if i < len(isstd) {
   302  			tx[i].isstd = isstd[i] != 0
   303  		}
   304  		if i < len(isutc) {
   305  			tx[i].isutc = isutc[i] != 0
   306  		}
   307  	}
   308  
   309  	if len(tx) == 0 {
   310  		// Build fake transition to cover all time.
   311  		// This happens in fixed locations like "Etc/GMT0".
   312  		tx = append(tx, zoneTrans{when: alpha, index: 0})
   313  	}
   314  
   315  	// Committed to succeed.
   316  	l := &Location{zone: zones, tx: tx, name: name, extend: extend}
   317  
   318  	// Fill in the cache with information about right now,
   319  	// since that will be the most common lookup.
   320  	sec, _, _ := now()
   321  	for i := range tx {
   322  		if tx[i].when <= sec && (i+1 == len(tx) || sec < tx[i+1].when) {
   323  			l.cacheStart = tx[i].when
   324  			l.cacheEnd = omega
   325  			l.cacheZone = &l.zone[tx[i].index]
   326  			if i+1 < len(tx) {
   327  				l.cacheEnd = tx[i+1].when
   328  			} else if l.extend != "" {
   329  				// If we're at the end of the known zone transitions,
   330  				// try the extend string.
   331  				if name, offset, estart, eend, isDST, ok := tzset(l.extend, l.cacheStart, sec); ok {
   332  					l.cacheStart = estart
   333  					l.cacheEnd = eend
   334  					// Find the zone that is returned by tzset to avoid allocation if possible.
   335  					if zoneIdx := findZone(l.zone, name, offset, isDST); zoneIdx != -1 {
   336  						l.cacheZone = &l.zone[zoneIdx]
   337  					} else {
   338  						l.cacheZone = &zone{
   339  							name:   name,
   340  							offset: offset,
   341  							isDST:  isDST,
   342  						}
   343  					}
   344  				}
   345  			}
   346  			break
   347  		}
   348  	}
   349  
   350  	return l, nil
   351  }
   352  
   353  func findZone(zones []zone, name string, offset int, isDST bool) int {
   354  	for i, z := range zones {
   355  		if z.name == name && z.offset == offset && z.isDST == isDST {
   356  			return i
   357  		}
   358  	}
   359  	return -1
   360  }
   361  
   362  // loadTzinfoFromDirOrZip returns the contents of the file with the given name
   363  // in dir. dir can either be an uncompressed zip file, or a directory.
   364  func loadTzinfoFromDirOrZip(dir, name string) ([]byte, error) {
   365  	if len(dir) > 4 && dir[len(dir)-4:] == ".zip" {
   366  		return loadTzinfoFromZip(dir, name)
   367  	}
   368  	if dir != "" {
   369  		name = dir + "/" + name
   370  	}
   371  	return readFile(name)
   372  }
   373  
   374  // There are 500+ zoneinfo files. Rather than distribute them all
   375  // individually, we ship them in an uncompressed zip file.
   376  // Used this way, the zip file format serves as a commonly readable
   377  // container for the individual small files. We choose zip over tar
   378  // because zip files have a contiguous table of contents, making
   379  // individual file lookups faster, and because the per-file overhead
   380  // in a zip file is considerably less than tar's 512 bytes.
   381  
   382  // get4 returns the little-endian 32-bit value in b.
   383  func get4(b []byte) int {
   384  	if len(b) < 4 {
   385  		return 0
   386  	}
   387  	return int(b[0]) | int(b[1])<<8 | int(b[2])<<16 | int(b[3])<<24
   388  }
   389  
   390  // get2 returns the little-endian 16-bit value in b.
   391  func get2(b []byte) int {
   392  	if len(b) < 2 {
   393  		return 0
   394  	}
   395  	return int(b[0]) | int(b[1])<<8
   396  }
   397  
   398  // loadTzinfoFromZip returns the contents of the file with the given name
   399  // in the given uncompressed zip file.
   400  func loadTzinfoFromZip(zipfile, name string) ([]byte, error) {
   401  	fd, err := open(zipfile)
   402  	if err != nil {
   403  		return nil, err
   404  	}
   405  	defer closefd(fd)
   406  
   407  	const (
   408  		zecheader = 0x06054b50
   409  		zcheader  = 0x02014b50
   410  		ztailsize = 22
   411  
   412  		zheadersize = 30
   413  		zheader     = 0x04034b50
   414  	)
   415  
   416  	buf := make([]byte, ztailsize)
   417  	if err := preadn(fd, buf, -ztailsize); err != nil || get4(buf) != zecheader {
   418  		return nil, errors.New("corrupt zip file " + zipfile)
   419  	}
   420  	n := get2(buf[10:])
   421  	size := get4(buf[12:])
   422  	off := get4(buf[16:])
   423  
   424  	buf = make([]byte, size)
   425  	if err := preadn(fd, buf, off); err != nil {
   426  		return nil, errors.New("corrupt zip file " + zipfile)
   427  	}
   428  
   429  	for i := 0; i < n; i++ {
   430  		// zip entry layout:
   431  		//	0	magic[4]
   432  		//	4	madevers[1]
   433  		//	5	madeos[1]
   434  		//	6	extvers[1]
   435  		//	7	extos[1]
   436  		//	8	flags[2]
   437  		//	10	meth[2]
   438  		//	12	modtime[2]
   439  		//	14	moddate[2]
   440  		//	16	crc[4]
   441  		//	20	csize[4]
   442  		//	24	uncsize[4]
   443  		//	28	namelen[2]
   444  		//	30	xlen[2]
   445  		//	32	fclen[2]
   446  		//	34	disknum[2]
   447  		//	36	iattr[2]
   448  		//	38	eattr[4]
   449  		//	42	off[4]
   450  		//	46	name[namelen]
   451  		//	46+namelen+xlen+fclen - next header
   452  		//
   453  		if get4(buf) != zcheader {
   454  			break
   455  		}
   456  		meth := get2(buf[10:])
   457  		size := get4(buf[24:])
   458  		namelen := get2(buf[28:])
   459  		xlen := get2(buf[30:])
   460  		fclen := get2(buf[32:])
   461  		off := get4(buf[42:])
   462  		zname := buf[46 : 46+namelen]
   463  		buf = buf[46+namelen+xlen+fclen:]
   464  		if string(zname) != name {
   465  			continue
   466  		}
   467  		if meth != 0 {
   468  			return nil, errors.New("unsupported compression for " + name + " in " + zipfile)
   469  		}
   470  
   471  		// zip per-file header layout:
   472  		//	0	magic[4]
   473  		//	4	extvers[1]
   474  		//	5	extos[1]
   475  		//	6	flags[2]
   476  		//	8	meth[2]
   477  		//	10	modtime[2]
   478  		//	12	moddate[2]
   479  		//	14	crc[4]
   480  		//	18	csize[4]
   481  		//	22	uncsize[4]
   482  		//	26	namelen[2]
   483  		//	28	xlen[2]
   484  		//	30	name[namelen]
   485  		//	30+namelen+xlen - file data
   486  		//
   487  		buf = make([]byte, zheadersize+namelen)
   488  		if err := preadn(fd, buf, off); err != nil ||
   489  			get4(buf) != zheader ||
   490  			get2(buf[8:]) != meth ||
   491  			get2(buf[26:]) != namelen ||
   492  			string(buf[30:30+namelen]) != name {
   493  			return nil, errors.New("corrupt zip file " + zipfile)
   494  		}
   495  		xlen = get2(buf[28:])
   496  
   497  		buf = make([]byte, size)
   498  		if err := preadn(fd, buf, off+30+namelen+xlen); err != nil {
   499  			return nil, errors.New("corrupt zip file " + zipfile)
   500  		}
   501  
   502  		return buf, nil
   503  	}
   504  
   505  	return nil, syscall.ENOENT
   506  }
   507  
   508  // loadTzinfoFromTzdata returns the time zone information of the time zone
   509  // with the given name, from a tzdata database file as they are typically
   510  // found on android.
   511  var loadTzinfoFromTzdata func(file, name string) ([]byte, error)
   512  
   513  // loadTzinfo returns the time zone information of the time zone
   514  // with the given name, from a given source. A source may be a
   515  // timezone database directory, tzdata database file or an uncompressed
   516  // zip file, containing the contents of such a directory.
   517  func loadTzinfo(name string, source string) ([]byte, error) {
   518  	if len(source) >= 6 && source[len(source)-6:] == "tzdata" {
   519  		return loadTzinfoFromTzdata(source, name)
   520  	}
   521  	return loadTzinfoFromDirOrZip(source, name)
   522  }
   523  
   524  // loadLocation returns the Location with the given name from one of
   525  // the specified sources. See loadTzinfo for a list of supported sources.
   526  // The first timezone data matching the given name that is successfully loaded
   527  // and parsed is returned as a Location.
   528  func loadLocation(name string, sources []string) (z *Location, firstErr error) {
   529  	for _, source := range sources {
   530  		zoneData, err := loadTzinfo(name, source)
   531  		if err == nil {
   532  			if z, err = LoadLocationFromTZData(name, zoneData); err == nil {
   533  				return z, nil
   534  			}
   535  		}
   536  		if firstErr == nil && err != syscall.ENOENT {
   537  			firstErr = err
   538  		}
   539  	}
   540  	if loadFromEmbeddedTZData != nil {
   541  		zoneData, err := loadFromEmbeddedTZData(name)
   542  		if err == nil {
   543  			if z, err = LoadLocationFromTZData(name, []byte(zoneData)); err == nil {
   544  				return z, nil
   545  			}
   546  		}
   547  		if firstErr == nil && err != syscall.ENOENT {
   548  			firstErr = err
   549  		}
   550  	}
   551  	if source, ok := gorootZoneSource(runtime.GOROOT()); ok {
   552  		zoneData, err := loadTzinfo(name, source)
   553  		if err == nil {
   554  			if z, err = LoadLocationFromTZData(name, zoneData); err == nil {
   555  				return z, nil
   556  			}
   557  		}
   558  		if firstErr == nil && err != syscall.ENOENT {
   559  			firstErr = err
   560  		}
   561  	}
   562  	if firstErr != nil {
   563  		return nil, firstErr
   564  	}
   565  	return nil, errors.New("unknown time zone " + name)
   566  }
   567  
   568  // readFile reads and returns the content of the named file.
   569  // It is a trivial implementation of os.ReadFile, reimplemented
   570  // here to avoid depending on io/ioutil or os.
   571  // It returns an error if name exceeds maxFileSize bytes.
   572  func readFile(name string) ([]byte, error) {
   573  	f, err := open(name)
   574  	if err != nil {
   575  		return nil, err
   576  	}
   577  	defer closefd(f)
   578  	var (
   579  		buf [4096]byte
   580  		ret []byte
   581  		n   int
   582  	)
   583  	for {
   584  		n, err = read(f, buf[:])
   585  		if n > 0 {
   586  			ret = append(ret, buf[:n]...)
   587  		}
   588  		if n == 0 || err != nil {
   589  			break
   590  		}
   591  		if len(ret) > maxFileSize {
   592  			return nil, fileSizeError(name)
   593  		}
   594  	}
   595  	return ret, err
   596  }
   597  

View as plain text