common.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  // Package tar implements access to tar archives.
     6  //
     7  // Tape archives (tar) are a file format for storing a sequence of files that
     8  // can be read and written in a streaming manner.
     9  // This package aims to cover most variations of the format,
    10  // including those produced by GNU and BSD tar tools.
    11  package tar
    12  
    13  import (
    14  	"errors"
    15  	"fmt"
    16  	"io/fs"
    17  	"math"
    18  	"path"
    19  	"reflect"
    20  	"strconv"
    21  	"strings"
    22  	"time"
    23  )
    24  
    25  // BUG: Use of the Uid and Gid fields in Header could overflow on 32-bit
    26  // architectures. If a large value is encountered when decoding, the result
    27  // stored in Header will be the truncated version.
    28  
    29  var (
    30  	ErrHeader          = errors.New("archive/tar: invalid tar header")
    31  	ErrWriteTooLong    = errors.New("archive/tar: write too long")
    32  	ErrFieldTooLong    = errors.New("archive/tar: header field too long")
    33  	ErrWriteAfterClose = errors.New("archive/tar: write after close")
    34  	errMissData        = errors.New("archive/tar: sparse file references non-existent data")
    35  	errUnrefData       = errors.New("archive/tar: sparse file contains unreferenced data")
    36  	errWriteHole       = errors.New("archive/tar: write non-NUL byte in sparse hole")
    37  )
    38  
    39  type headerError []string
    40  
    41  func (he headerError) Error() string {
    42  	const prefix = "archive/tar: cannot encode header"
    43  	var ss []string
    44  	for _, s := range he {
    45  		if s != "" {
    46  			ss = append(ss, s)
    47  		}
    48  	}
    49  	if len(ss) == 0 {
    50  		return prefix
    51  	}
    52  	return fmt.Sprintf("%s: %v", prefix, strings.Join(ss, "; and "))
    53  }
    54  
    55  // Type flags for Header.Typeflag.
    56  const (
    57  	// Type '0' indicates a regular file.
    58  	TypeReg  = '0'
    59  	TypeRegA = '\x00' // Deprecated: Use TypeReg instead.
    60  
    61  	// Type '1' to '6' are header-only flags and may not have a data body.
    62  	TypeLink    = '1' // Hard link
    63  	TypeSymlink = '2' // Symbolic link
    64  	TypeChar    = '3' // Character device node
    65  	TypeBlock   = '4' // Block device node
    66  	TypeDir     = '5' // Directory
    67  	TypeFifo    = '6' // FIFO node
    68  
    69  	// Type '7' is reserved.
    70  	TypeCont = '7'
    71  
    72  	// Type 'x' is used by the PAX format to store key-value records that
    73  	// are only relevant to the next file.
    74  	// This package transparently handles these types.
    75  	TypeXHeader = 'x'
    76  
    77  	// Type 'g' is used by the PAX format to store key-value records that
    78  	// are relevant to all subsequent files.
    79  	// This package only supports parsing and composing such headers,
    80  	// but does not currently support persisting the global state across files.
    81  	TypeXGlobalHeader = 'g'
    82  
    83  	// Type 'S' indicates a sparse file in the GNU format.
    84  	TypeGNUSparse = 'S'
    85  
    86  	// Types 'L' and 'K' are used by the GNU format for a meta file
    87  	// used to store the path or link name for the next file.
    88  	// This package transparently handles these types.
    89  	TypeGNULongName = 'L'
    90  	TypeGNULongLink = 'K'
    91  )
    92  
    93  // Keywords for PAX extended header records.
    94  const (
    95  	paxNone     = "" // Indicates that no PAX key is suitable
    96  	paxPath     = "path"
    97  	paxLinkpath = "linkpath"
    98  	paxSize     = "size"
    99  	paxUid      = "uid"
   100  	paxGid      = "gid"
   101  	paxUname    = "uname"
   102  	paxGname    = "gname"
   103  	paxMtime    = "mtime"
   104  	paxAtime    = "atime"
   105  	paxCtime    = "ctime"   // Removed from later revision of PAX spec, but was valid
   106  	paxCharset  = "charset" // Currently unused
   107  	paxComment  = "comment" // Currently unused
   108  
   109  	paxSchilyXattr = "SCHILY.xattr."
   110  
   111  	// Keywords for GNU sparse files in a PAX extended header.
   112  	paxGNUSparse          = "GNU.sparse."
   113  	paxGNUSparseNumBlocks = "GNU.sparse.numblocks"
   114  	paxGNUSparseOffset    = "GNU.sparse.offset"
   115  	paxGNUSparseNumBytes  = "GNU.sparse.numbytes"
   116  	paxGNUSparseMap       = "GNU.sparse.map"
   117  	paxGNUSparseName      = "GNU.sparse.name"
   118  	paxGNUSparseMajor     = "GNU.sparse.major"
   119  	paxGNUSparseMinor     = "GNU.sparse.minor"
   120  	paxGNUSparseSize      = "GNU.sparse.size"
   121  	paxGNUSparseRealSize  = "GNU.sparse.realsize"
   122  )
   123  
   124  // basicKeys is a set of the PAX keys for which we have built-in support.
   125  // This does not contain "charset" or "comment", which are both PAX-specific,
   126  // so adding them as first-class features of Header is unlikely.
   127  // Users can use the PAXRecords field to set it themselves.
   128  var basicKeys = map[string]bool{
   129  	paxPath: true, paxLinkpath: true, paxSize: true, paxUid: true, paxGid: true,
   130  	paxUname: true, paxGname: true, paxMtime: true, paxAtime: true, paxCtime: true,
   131  }
   132  
   133  // A Header represents a single header in a tar archive.
   134  // Some fields may not be populated.
   135  //
   136  // For forward compatibility, users that retrieve a Header from Reader.Next,
   137  // mutate it in some ways, and then pass it back to Writer.WriteHeader
   138  // should do so by creating a new Header and copying the fields
   139  // that they are interested in preserving.
   140  type Header struct {
   141  	// Typeflag is the type of header entry.
   142  	// The zero value is automatically promoted to either TypeReg or TypeDir
   143  	// depending on the presence of a trailing slash in Name.
   144  	Typeflag byte
   145  
   146  	Name     string // Name of file entry
   147  	Linkname string // Target name of link (valid for TypeLink or TypeSymlink)
   148  
   149  	Size  int64  // Logical file size in bytes
   150  	Mode  int64  // Permission and mode bits
   151  	Uid   int    // User ID of owner
   152  	Gid   int    // Group ID of owner
   153  	Uname string // User name of owner
   154  	Gname string // Group name of owner
   155  
   156  	// If the Format is unspecified, then Writer.WriteHeader rounds ModTime
   157  	// to the nearest second and ignores the AccessTime and ChangeTime fields.
   158  	//
   159  	// To use AccessTime or ChangeTime, specify the Format as PAX or GNU.
   160  	// To use sub-second resolution, specify the Format as PAX.
   161  	ModTime    time.Time // Modification time
   162  	AccessTime time.Time // Access time (requires either PAX or GNU support)
   163  	ChangeTime time.Time // Change time (requires either PAX or GNU support)
   164  
   165  	Devmajor int64 // Major device number (valid for TypeChar or TypeBlock)
   166  	Devminor int64 // Minor device number (valid for TypeChar or TypeBlock)
   167  
   168  	// Xattrs stores extended attributes as PAX records under the
   169  	// "SCHILY.xattr." namespace.
   170  	//
   171  	// The following are semantically equivalent:
   172  	//  h.Xattrs[key] = value
   173  	//  h.PAXRecords["SCHILY.xattr."+key] = value
   174  	//
   175  	// When Writer.WriteHeader is called, the contents of Xattrs will take
   176  	// precedence over those in PAXRecords.
   177  	//
   178  	// Deprecated: Use PAXRecords instead.
   179  	Xattrs map[string]string
   180  
   181  	// PAXRecords is a map of PAX extended header records.
   182  	//
   183  	// User-defined records should have keys of the following form:
   184  	//	VENDOR.keyword
   185  	// Where VENDOR is some namespace in all uppercase, and keyword may
   186  	// not contain the '=' character (e.g., "GOLANG.pkg.version").
   187  	// The key and value should be non-empty UTF-8 strings.
   188  	//
   189  	// When Writer.WriteHeader is called, PAX records derived from the
   190  	// other fields in Header take precedence over PAXRecords.
   191  	PAXRecords map[string]string
   192  
   193  	// Format specifies the format of the tar header.
   194  	//
   195  	// This is set by Reader.Next as a best-effort guess at the format.
   196  	// Since the Reader liberally reads some non-compliant files,
   197  	// it is possible for this to be FormatUnknown.
   198  	//
   199  	// If the format is unspecified when Writer.WriteHeader is called,
   200  	// then it uses the first format (in the order of USTAR, PAX, GNU)
   201  	// capable of encoding this Header (see Format).
   202  	Format Format
   203  }
   204  
   205  // sparseEntry represents a Length-sized fragment at Offset in the file.
   206  type sparseEntry struct{ Offset, Length int64 }
   207  
   208  func (s sparseEntry) endOffset() int64 { return s.Offset + s.Length }
   209  
   210  // A sparse file can be represented as either a sparseDatas or a sparseHoles.
   211  // As long as the total size is known, they are equivalent and one can be
   212  // converted to the other form and back. The various tar formats with sparse
   213  // file support represent sparse files in the sparseDatas form. That is, they
   214  // specify the fragments in the file that has data, and treat everything else as
   215  // having zero bytes. As such, the encoding and decoding logic in this package
   216  // deals with sparseDatas.
   217  //
   218  // However, the external API uses sparseHoles instead of sparseDatas because the
   219  // zero value of sparseHoles logically represents a normal file (i.e., there are
   220  // no holes in it). On the other hand, the zero value of sparseDatas implies
   221  // that the file has no data in it, which is rather odd.
   222  //
   223  // As an example, if the underlying raw file contains the 10-byte data:
   224  //	var compactFile = "abcdefgh"
   225  //
   226  // And the sparse map has the following entries:
   227  //	var spd sparseDatas = []sparseEntry{
   228  //		{Offset: 2,  Length: 5},  // Data fragment for 2..6
   229  //		{Offset: 18, Length: 3},  // Data fragment for 18..20
   230  //	}
   231  //	var sph sparseHoles = []sparseEntry{
   232  //		{Offset: 0,  Length: 2},  // Hole fragment for 0..1
   233  //		{Offset: 7,  Length: 11}, // Hole fragment for 7..17
   234  //		{Offset: 21, Length: 4},  // Hole fragment for 21..24
   235  //	}
   236  //
   237  // Then the content of the resulting sparse file with a Header.Size of 25 is:
   238  //	var sparseFile = "\x00"*2 + "abcde" + "\x00"*11 + "fgh" + "\x00"*4
   239  type (
   240  	sparseDatas []sparseEntry
   241  	sparseHoles []sparseEntry
   242  )
   243  
   244  // validateSparseEntries reports whether sp is a valid sparse map.
   245  // It does not matter whether sp represents data fragments or hole fragments.
   246  func validateSparseEntries(sp []sparseEntry, size int64) bool {
   247  	// Validate all sparse entries. These are the same checks as performed by
   248  	// the BSD tar utility.
   249  	if size < 0 {
   250  		return false
   251  	}
   252  	var pre sparseEntry
   253  	for _, cur := range sp {
   254  		switch {
   255  		case cur.Offset < 0 || cur.Length < 0:
   256  			return false // Negative values are never okay
   257  		case cur.Offset > math.MaxInt64-cur.Length:
   258  			return false // Integer overflow with large length
   259  		case cur.endOffset() > size:
   260  			return false // Region extends beyond the actual size
   261  		case pre.endOffset() > cur.Offset:
   262  			return false // Regions cannot overlap and must be in order
   263  		}
   264  		pre = cur
   265  	}
   266  	return true
   267  }
   268  
   269  // alignSparseEntries mutates src and returns dst where each fragment's
   270  // starting offset is aligned up to the nearest block edge, and each
   271  // ending offset is aligned down to the nearest block edge.
   272  //
   273  // Even though the Go tar Reader and the BSD tar utility can handle entries
   274  // with arbitrary offsets and lengths, the GNU tar utility can only handle
   275  // offsets and lengths that are multiples of blockSize.
   276  func alignSparseEntries(src []sparseEntry, size int64) []sparseEntry {
   277  	dst := src[:0]
   278  	for _, s := range src {
   279  		pos, end := s.Offset, s.endOffset()
   280  		pos += blockPadding(+pos) // Round-up to nearest blockSize
   281  		if end != size {
   282  			end -= blockPadding(-end) // Round-down to nearest blockSize
   283  		}
   284  		if pos < end {
   285  			dst = append(dst, sparseEntry{Offset: pos, Length: end - pos})
   286  		}
   287  	}
   288  	return dst
   289  }
   290  
   291  // invertSparseEntries converts a sparse map from one form to the other.
   292  // If the input is sparseHoles, then it will output sparseDatas and vice-versa.
   293  // The input must have been already validated.
   294  //
   295  // This function mutates src and returns a normalized map where:
   296  //	* adjacent fragments are coalesced together
   297  //	* only the last fragment may be empty
   298  //	* the endOffset of the last fragment is the total size
   299  func invertSparseEntries(src []sparseEntry, size int64) []sparseEntry {
   300  	dst := src[:0]
   301  	var pre sparseEntry
   302  	for _, cur := range src {
   303  		if cur.Length == 0 {
   304  			continue // Skip empty fragments
   305  		}
   306  		pre.Length = cur.Offset - pre.Offset
   307  		if pre.Length > 0 {
   308  			dst = append(dst, pre) // Only add non-empty fragments
   309  		}
   310  		pre.Offset = cur.endOffset()
   311  	}
   312  	pre.Length = size - pre.Offset // Possibly the only empty fragment
   313  	return append(dst, pre)
   314  }
   315  
   316  // fileState tracks the number of logical (includes sparse holes) and physical
   317  // (actual in tar archive) bytes remaining for the current file.
   318  //
   319  // Invariant: logicalRemaining >= physicalRemaining
   320  type fileState interface {
   321  	logicalRemaining() int64
   322  	physicalRemaining() int64
   323  }
   324  
   325  // allowedFormats determines which formats can be used.
   326  // The value returned is the logical OR of multiple possible formats.
   327  // If the value is FormatUnknown, then the input Header cannot be encoded
   328  // and an error is returned explaining why.
   329  //
   330  // As a by-product of checking the fields, this function returns paxHdrs, which
   331  // contain all fields that could not be directly encoded.
   332  // A value receiver ensures that this method does not mutate the source Header.
   333  func (h Header) allowedFormats() (format Format, paxHdrs map[string]string, err error) {
   334  	format = FormatUSTAR | FormatPAX | FormatGNU
   335  	paxHdrs = make(map[string]string)
   336  
   337  	var whyNoUSTAR, whyNoPAX, whyNoGNU string
   338  	var preferPAX bool // Prefer PAX over USTAR
   339  	verifyString := func(s string, size int, name, paxKey string) {
   340  		// NUL-terminator is optional for path and linkpath.
   341  		// Technically, it is required for uname and gname,
   342  		// but neither GNU nor BSD tar checks for it.
   343  		tooLong := len(s) > size
   344  		allowLongGNU := paxKey == paxPath || paxKey == paxLinkpath
   345  		if hasNUL(s) || (tooLong && !allowLongGNU) {
   346  			whyNoGNU = fmt.Sprintf("GNU cannot encode %s=%q", name, s)
   347  			format.mustNotBe(FormatGNU)
   348  		}
   349  		if !isASCII(s) || tooLong {
   350  			canSplitUSTAR := paxKey == paxPath
   351  			if _, _, ok := splitUSTARPath(s); !canSplitUSTAR || !ok {
   352  				whyNoUSTAR = fmt.Sprintf("USTAR cannot encode %s=%q", name, s)
   353  				format.mustNotBe(FormatUSTAR)
   354  			}
   355  			if paxKey == paxNone {
   356  				whyNoPAX = fmt.Sprintf("PAX cannot encode %s=%q", name, s)
   357  				format.mustNotBe(FormatPAX)
   358  			} else {
   359  				paxHdrs[paxKey] = s
   360  			}
   361  		}
   362  		if v, ok := h.PAXRecords[paxKey]; ok && v == s {
   363  			paxHdrs[paxKey] = v
   364  		}
   365  	}
   366  	verifyNumeric := func(n int64, size int, name, paxKey string) {
   367  		if !fitsInBase256(size, n) {
   368  			whyNoGNU = fmt.Sprintf("GNU cannot encode %s=%d", name, n)
   369  			format.mustNotBe(FormatGNU)
   370  		}
   371  		if !fitsInOctal(size, n) {
   372  			whyNoUSTAR = fmt.Sprintf("USTAR cannot encode %s=%d", name, n)
   373  			format.mustNotBe(FormatUSTAR)
   374  			if paxKey == paxNone {
   375  				whyNoPAX = fmt.Sprintf("PAX cannot encode %s=%d", name, n)
   376  				format.mustNotBe(FormatPAX)
   377  			} else {
   378  				paxHdrs[paxKey] = strconv.FormatInt(n, 10)
   379  			}
   380  		}
   381  		if v, ok := h.PAXRecords[paxKey]; ok && v == strconv.FormatInt(n, 10) {
   382  			paxHdrs[paxKey] = v
   383  		}
   384  	}
   385  	verifyTime := func(ts time.Time, size int, name, paxKey string) {
   386  		if ts.IsZero() {
   387  			return // Always okay
   388  		}
   389  		if !fitsInBase256(size, ts.Unix()) {
   390  			whyNoGNU = fmt.Sprintf("GNU cannot encode %s=%v", name, ts)
   391  			format.mustNotBe(FormatGNU)
   392  		}
   393  		isMtime := paxKey == paxMtime
   394  		fitsOctal := fitsInOctal(size, ts.Unix())
   395  		if (isMtime && !fitsOctal) || !isMtime {
   396  			whyNoUSTAR = fmt.Sprintf("USTAR cannot encode %s=%v", name, ts)
   397  			format.mustNotBe(FormatUSTAR)
   398  		}
   399  		needsNano := ts.Nanosecond() != 0
   400  		if !isMtime || !fitsOctal || needsNano {
   401  			preferPAX = true // USTAR may truncate sub-second measurements
   402  			if paxKey == paxNone {
   403  				whyNoPAX = fmt.Sprintf("PAX cannot encode %s=%v", name, ts)
   404  				format.mustNotBe(FormatPAX)
   405  			} else {
   406  				paxHdrs[paxKey] = formatPAXTime(ts)
   407  			}
   408  		}
   409  		if v, ok := h.PAXRecords[paxKey]; ok && v == formatPAXTime(ts) {
   410  			paxHdrs[paxKey] = v
   411  		}
   412  	}
   413  
   414  	// Check basic fields.
   415  	var blk block
   416  	v7 := blk.toV7()
   417  	ustar := blk.toUSTAR()
   418  	gnu := blk.toGNU()
   419  	verifyString(h.Name, len(v7.name()), "Name", paxPath)
   420  	verifyString(h.Linkname, len(v7.linkName()), "Linkname", paxLinkpath)
   421  	verifyString(h.Uname, len(ustar.userName()), "Uname", paxUname)
   422  	verifyString(h.Gname, len(ustar.groupName()), "Gname", paxGname)
   423  	verifyNumeric(h.Mode, len(v7.mode()), "Mode", paxNone)
   424  	verifyNumeric(int64(h.Uid), len(v7.uid()), "Uid", paxUid)
   425  	verifyNumeric(int64(h.Gid), len(v7.gid()), "Gid", paxGid)
   426  	verifyNumeric(h.Size, len(v7.size()), "Size", paxSize)
   427  	verifyNumeric(h.Devmajor, len(ustar.devMajor()), "Devmajor", paxNone)
   428  	verifyNumeric(h.Devminor, len(ustar.devMinor()), "Devminor", paxNone)
   429  	verifyTime(h.ModTime, len(v7.modTime()), "ModTime", paxMtime)
   430  	verifyTime(h.AccessTime, len(gnu.accessTime()), "AccessTime", paxAtime)
   431  	verifyTime(h.ChangeTime, len(gnu.changeTime()), "ChangeTime", paxCtime)
   432  
   433  	// Check for header-only types.
   434  	var whyOnlyPAX, whyOnlyGNU string
   435  	switch h.Typeflag {
   436  	case TypeReg, TypeChar, TypeBlock, TypeFifo, TypeGNUSparse:
   437  		// Exclude TypeLink and TypeSymlink, since they may reference directories.
   438  		if strings.HasSuffix(h.Name, "/") {
   439  			return FormatUnknown, nil, headerError{"filename may not have trailing slash"}
   440  		}
   441  	case TypeXHeader, TypeGNULongName, TypeGNULongLink:
   442  		return FormatUnknown, nil, headerError{"cannot manually encode TypeXHeader, TypeGNULongName, or TypeGNULongLink headers"}
   443  	case TypeXGlobalHeader:
   444  		h2 := Header{Name: h.Name, Typeflag: h.Typeflag, Xattrs: h.Xattrs, PAXRecords: h.PAXRecords, Format: h.Format}
   445  		if !reflect.DeepEqual(h, h2) {
   446  			return FormatUnknown, nil, headerError{"only PAXRecords should be set for TypeXGlobalHeader"}
   447  		}
   448  		whyOnlyPAX = "only PAX supports TypeXGlobalHeader"
   449  		format.mayOnlyBe(FormatPAX)
   450  	}
   451  	if !isHeaderOnlyType(h.Typeflag) && h.Size < 0 {
   452  		return FormatUnknown, nil, headerError{"negative size on header-only type"}
   453  	}
   454  
   455  	// Check PAX records.
   456  	if len(h.Xattrs) > 0 {
   457  		for k, v := range h.Xattrs {
   458  			paxHdrs[paxSchilyXattr+k] = v
   459  		}
   460  		whyOnlyPAX = "only PAX supports Xattrs"
   461  		format.mayOnlyBe(FormatPAX)
   462  	}
   463  	if len(h.PAXRecords) > 0 {
   464  		for k, v := range h.PAXRecords {
   465  			switch _, exists := paxHdrs[k]; {
   466  			case exists:
   467  				continue // Do not overwrite existing records
   468  			case h.Typeflag == TypeXGlobalHeader:
   469  				paxHdrs[k] = v // Copy all records
   470  			case !basicKeys[k] && !strings.HasPrefix(k, paxGNUSparse):
   471  				paxHdrs[k] = v // Ignore local records that may conflict
   472  			}
   473  		}
   474  		whyOnlyPAX = "only PAX supports PAXRecords"
   475  		format.mayOnlyBe(FormatPAX)
   476  	}
   477  	for k, v := range paxHdrs {
   478  		if !validPAXRecord(k, v) {
   479  			return FormatUnknown, nil, headerError{fmt.Sprintf("invalid PAX record: %q", k+" = "+v)}
   480  		}
   481  	}
   482  
   483  	// TODO(dsnet): Re-enable this when adding sparse support.
   484  	// See https://golang.org/issue/22735
   485  	/*
   486  		// Check sparse files.
   487  		if len(h.SparseHoles) > 0 || h.Typeflag == TypeGNUSparse {
   488  			if isHeaderOnlyType(h.Typeflag) {
   489  				return FormatUnknown, nil, headerError{"header-only type cannot be sparse"}
   490  			}
   491  			if !validateSparseEntries(h.SparseHoles, h.Size) {
   492  				return FormatUnknown, nil, headerError{"invalid sparse holes"}
   493  			}
   494  			if h.Typeflag == TypeGNUSparse {
   495  				whyOnlyGNU = "only GNU supports TypeGNUSparse"
   496  				format.mayOnlyBe(FormatGNU)
   497  			} else {
   498  				whyNoGNU = "GNU supports sparse files only with TypeGNUSparse"
   499  				format.mustNotBe(FormatGNU)
   500  			}
   501  			whyNoUSTAR = "USTAR does not support sparse files"
   502  			format.mustNotBe(FormatUSTAR)
   503  		}
   504  	*/
   505  
   506  	// Check desired format.
   507  	if wantFormat := h.Format; wantFormat != FormatUnknown {
   508  		if wantFormat.has(FormatPAX) && !preferPAX {
   509  			wantFormat.mayBe(FormatUSTAR) // PAX implies USTAR allowed too
   510  		}
   511  		format.mayOnlyBe(wantFormat) // Set union of formats allowed and format wanted
   512  	}
   513  	if format == FormatUnknown {
   514  		switch h.Format {
   515  		case FormatUSTAR:
   516  			err = headerError{"Format specifies USTAR", whyNoUSTAR, whyOnlyPAX, whyOnlyGNU}
   517  		case FormatPAX:
   518  			err = headerError{"Format specifies PAX", whyNoPAX, whyOnlyGNU}
   519  		case FormatGNU:
   520  			err = headerError{"Format specifies GNU", whyNoGNU, whyOnlyPAX}
   521  		default:
   522  			err = headerError{whyNoUSTAR, whyNoPAX, whyNoGNU, whyOnlyPAX, whyOnlyGNU}
   523  		}
   524  	}
   525  	return format, paxHdrs, err
   526  }
   527  
   528  // FileInfo returns an fs.FileInfo for the Header.
   529  func (h *Header) FileInfo() fs.FileInfo {
   530  	return headerFileInfo{h}
   531  }
   532  
   533  // headerFileInfo implements fs.FileInfo.
   534  type headerFileInfo struct {
   535  	h *Header
   536  }
   537  
   538  func (fi headerFileInfo) Size() int64        { return fi.h.Size }
   539  func (fi headerFileInfo) IsDir() bool        { return fi.Mode().IsDir() }
   540  func (fi headerFileInfo) ModTime() time.Time { return fi.h.ModTime }
   541  func (fi headerFileInfo) Sys() any           { return fi.h }
   542  
   543  // Name returns the base name of the file.
   544  func (fi headerFileInfo) Name() string {
   545  	if fi.IsDir() {
   546  		return path.Base(path.Clean(fi.h.Name))
   547  	}
   548  	return path.Base(fi.h.Name)
   549  }
   550  
   551  // Mode returns the permission and mode bits for the headerFileInfo.
   552  func (fi headerFileInfo) Mode() (mode fs.FileMode) {
   553  	// Set file permission bits.
   554  	mode = fs.FileMode(fi.h.Mode).Perm()
   555  
   556  	// Set setuid, setgid and sticky bits.
   557  	if fi.h.Mode&c_ISUID != 0 {
   558  		mode |= fs.ModeSetuid
   559  	}
   560  	if fi.h.Mode&c_ISGID != 0 {
   561  		mode |= fs.ModeSetgid
   562  	}
   563  	if fi.h.Mode&c_ISVTX != 0 {
   564  		mode |= fs.ModeSticky
   565  	}
   566  
   567  	// Set file mode bits; clear perm, setuid, setgid, and sticky bits.
   568  	switch m := fs.FileMode(fi.h.Mode) &^ 07777; m {
   569  	case c_ISDIR:
   570  		mode |= fs.ModeDir
   571  	case c_ISFIFO:
   572  		mode |= fs.ModeNamedPipe
   573  	case c_ISLNK:
   574  		mode |= fs.ModeSymlink
   575  	case c_ISBLK:
   576  		mode |= fs.ModeDevice
   577  	case c_ISCHR:
   578  		mode |= fs.ModeDevice
   579  		mode |= fs.ModeCharDevice
   580  	case c_ISSOCK:
   581  		mode |= fs.ModeSocket
   582  	}
   583  
   584  	switch fi.h.Typeflag {
   585  	case TypeSymlink:
   586  		mode |= fs.ModeSymlink
   587  	case TypeChar:
   588  		mode |= fs.ModeDevice
   589  		mode |= fs.ModeCharDevice
   590  	case TypeBlock:
   591  		mode |= fs.ModeDevice
   592  	case TypeDir:
   593  		mode |= fs.ModeDir
   594  	case TypeFifo:
   595  		mode |= fs.ModeNamedPipe
   596  	}
   597  
   598  	return mode
   599  }
   600  
   601  // sysStat, if non-nil, populates h from system-dependent fields of fi.
   602  var sysStat func(fi fs.FileInfo, h *Header) error
   603  
   604  const (
   605  	// Mode constants from the USTAR spec:
   606  	// See http://pubs.opengroup.org/onlinepubs/9699919799/utilities/pax.html#tag_20_92_13_06
   607  	c_ISUID = 04000 // Set uid
   608  	c_ISGID = 02000 // Set gid
   609  	c_ISVTX = 01000 // Save text (sticky bit)
   610  
   611  	// Common Unix mode constants; these are not defined in any common tar standard.
   612  	// Header.FileInfo understands these, but FileInfoHeader will never produce these.
   613  	c_ISDIR  = 040000  // Directory
   614  	c_ISFIFO = 010000  // FIFO
   615  	c_ISREG  = 0100000 // Regular file
   616  	c_ISLNK  = 0120000 // Symbolic link
   617  	c_ISBLK  = 060000  // Block special file
   618  	c_ISCHR  = 020000  // Character special file
   619  	c_ISSOCK = 0140000 // Socket
   620  )
   621  
   622  // FileInfoHeader creates a partially-populated Header from fi.
   623  // If fi describes a symlink, FileInfoHeader records link as the link target.
   624  // If fi describes a directory, a slash is appended to the name.
   625  //
   626  // Since fs.FileInfo's Name method only returns the base name of
   627  // the file it describes, it may be necessary to modify Header.Name
   628  // to provide the full path name of the file.
   629  func FileInfoHeader(fi fs.FileInfo, link string) (*Header, error) {
   630  	if fi == nil {
   631  		return nil, errors.New("archive/tar: FileInfo is nil")
   632  	}
   633  	fm := fi.Mode()
   634  	h := &Header{
   635  		Name:    fi.Name(),
   636  		ModTime: fi.ModTime(),
   637  		Mode:    int64(fm.Perm()), // or'd with c_IS* constants later
   638  	}
   639  	switch {
   640  	case fm.IsRegular():
   641  		h.Typeflag = TypeReg
   642  		h.Size = fi.Size()
   643  	case fi.IsDir():
   644  		h.Typeflag = TypeDir
   645  		h.Name += "/"
   646  	case fm&fs.ModeSymlink != 0:
   647  		h.Typeflag = TypeSymlink
   648  		h.Linkname = link
   649  	case fm&fs.ModeDevice != 0:
   650  		if fm&fs.ModeCharDevice != 0 {
   651  			h.Typeflag = TypeChar
   652  		} else {
   653  			h.Typeflag = TypeBlock
   654  		}
   655  	case fm&fs.ModeNamedPipe != 0:
   656  		h.Typeflag = TypeFifo
   657  	case fm&fs.ModeSocket != 0:
   658  		return nil, fmt.Errorf("archive/tar: sockets not supported")
   659  	default:
   660  		return nil, fmt.Errorf("archive/tar: unknown file mode %v", fm)
   661  	}
   662  	if fm&fs.ModeSetuid != 0 {
   663  		h.Mode |= c_ISUID
   664  	}
   665  	if fm&fs.ModeSetgid != 0 {
   666  		h.Mode |= c_ISGID
   667  	}
   668  	if fm&fs.ModeSticky != 0 {
   669  		h.Mode |= c_ISVTX
   670  	}
   671  	// If possible, populate additional fields from OS-specific
   672  	// FileInfo fields.
   673  	if sys, ok := fi.Sys().(*Header); ok {
   674  		// This FileInfo came from a Header (not the OS). Use the
   675  		// original Header to populate all remaining fields.
   676  		h.Uid = sys.Uid
   677  		h.Gid = sys.Gid
   678  		h.Uname = sys.Uname
   679  		h.Gname = sys.Gname
   680  		h.AccessTime = sys.AccessTime
   681  		h.ChangeTime = sys.ChangeTime
   682  		if sys.Xattrs != nil {
   683  			h.Xattrs = make(map[string]string)
   684  			for k, v := range sys.Xattrs {
   685  				h.Xattrs[k] = v
   686  			}
   687  		}
   688  		if sys.Typeflag == TypeLink {
   689  			// hard link
   690  			h.Typeflag = TypeLink
   691  			h.Size = 0
   692  			h.Linkname = sys.Linkname
   693  		}
   694  		if sys.PAXRecords != nil {
   695  			h.PAXRecords = make(map[string]string)
   696  			for k, v := range sys.PAXRecords {
   697  				h.PAXRecords[k] = v
   698  			}
   699  		}
   700  	}
   701  	if sysStat != nil {
   702  		return h, sysStat(fi, h)
   703  	}
   704  	return h, nil
   705  }
   706  
   707  // isHeaderOnlyType checks if the given type flag is of the type that has no
   708  // data section even if a size is specified.
   709  func isHeaderOnlyType(flag byte) bool {
   710  	switch flag {
   711  	case TypeLink, TypeSymlink, TypeChar, TypeBlock, TypeDir, TypeFifo:
   712  		return true
   713  	default:
   714  		return false
   715  	}
   716  }
   717  
   718  func min(a, b int64) int64 {
   719  	if a < b {
   720  		return a
   721  	}
   722  	return b
   723  }
   724
View as plain text