1// Copyright 2016 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 5package tar 6 7// Constants to identify various tar formats. 8const ( 9 // The format is unknown. 10 formatUnknown = (1 << iota) / 2 // Sequence of 0, 1, 2, 4, 8, etc... 11 12 // The format of the original Unix V7 tar tool prior to standardization. 13 formatV7 14 15 // The old and new GNU formats, which are incompatible with USTAR. 16 // This does cover the old GNU sparse extension. 17 // This does not cover the GNU sparse extensions using PAX headers, 18 // versions 0.0, 0.1, and 1.0; these fall under the PAX format. 19 formatGNU 20 21 // Schily's tar format, which is incompatible with USTAR. 22 // This does not cover STAR extensions to the PAX format; these fall under 23 // the PAX format. 24 formatSTAR 25 26 // USTAR is the former standardization of tar defined in POSIX.1-1988. 27 // This is incompatible with the GNU and STAR formats. 28 formatUSTAR 29 30 // PAX is the latest standardization of tar defined in POSIX.1-2001. 31 // This is an extension of USTAR and is "backwards compatible" with it. 32 // 33 // Some newer formats add their own extensions to PAX, such as GNU sparse 34 // files and SCHILY extended attributes. Since they are backwards compatible 35 // with PAX, they will be labelled as "PAX". 36 formatPAX 37) 38 39// Magics used to identify various formats. 40const ( 41 magicGNU, versionGNU = "ustar ", " \x00" 42 magicUSTAR, versionUSTAR = "ustar\x00", "00" 43 trailerSTAR = "tar\x00" 44) 45 46// Size constants from various tar specifications. 47const ( 48 blockSize = 512 // Size of each block in a tar stream 49 nameSize = 100 // Max length of the name field in USTAR format 50 prefixSize = 155 // Max length of the prefix field in USTAR format 51) 52 53var zeroBlock block 54 55type block [blockSize]byte 56 57// Convert block to any number of formats. 58func (b *block) V7() *headerV7 { return (*headerV7)(b) } 59func (b *block) GNU() *headerGNU { return (*headerGNU)(b) } 60func (b *block) STAR() *headerSTAR { return (*headerSTAR)(b) } 61func (b *block) USTAR() *headerUSTAR { return (*headerUSTAR)(b) } 62func (b *block) Sparse() sparseArray { return (sparseArray)(b[:]) } 63 64// GetFormat checks that the block is a valid tar header based on the checksum. 65// It then attempts to guess the specific format based on magic values. 66// If the checksum fails, then formatUnknown is returned. 67func (b *block) GetFormat() (format int) { 68 // Verify checksum. 69 var p parser 70 value := p.parseOctal(b.V7().Chksum()) 71 chksum1, chksum2 := b.ComputeChecksum() 72 if p.err != nil || (value != chksum1 && value != chksum2) { 73 return formatUnknown 74 } 75 76 // Guess the magic values. 77 magic := string(b.USTAR().Magic()) 78 version := string(b.USTAR().Version()) 79 trailer := string(b.STAR().Trailer()) 80 switch { 81 case magic == magicUSTAR && trailer == trailerSTAR: 82 return formatSTAR 83 case magic == magicUSTAR: 84 return formatUSTAR 85 case magic == magicGNU && version == versionGNU: 86 return formatGNU 87 default: 88 return formatV7 89 } 90} 91 92// SetFormat writes the magic values necessary for specified format 93// and then updates the checksum accordingly. 94func (b *block) SetFormat(format int) { 95 // Set the magic values. 96 switch format { 97 case formatV7: 98 // Do nothing. 99 case formatGNU: 100 copy(b.GNU().Magic(), magicGNU) 101 copy(b.GNU().Version(), versionGNU) 102 case formatSTAR: 103 copy(b.STAR().Magic(), magicUSTAR) 104 copy(b.STAR().Version(), versionUSTAR) 105 copy(b.STAR().Trailer(), trailerSTAR) 106 case formatUSTAR, formatPAX: 107 copy(b.USTAR().Magic(), magicUSTAR) 108 copy(b.USTAR().Version(), versionUSTAR) 109 default: 110 panic("invalid format") 111 } 112 113 // Update checksum. 114 // This field is special in that it is terminated by a NULL then space. 115 var f formatter 116 field := b.V7().Chksum() 117 chksum, _ := b.ComputeChecksum() // Possible values are 256..128776 118 f.formatOctal(field[:7], chksum) // Never fails since 128776 < 262143 119 field[7] = ' ' 120} 121 122// ComputeChecksum computes the checksum for the header block. 123// POSIX specifies a sum of the unsigned byte values, but the Sun tar used 124// signed byte values. 125// We compute and return both. 126func (b *block) ComputeChecksum() (unsigned, signed int64) { 127 for i, c := range b { 128 if 148 <= i && i < 156 { 129 c = ' ' // Treat the checksum field itself as all spaces. 130 } 131 unsigned += int64(uint8(c)) 132 signed += int64(int8(c)) 133 } 134 return unsigned, signed 135} 136 137type headerV7 [blockSize]byte 138 139func (h *headerV7) Name() []byte { return h[000:][:100] } 140func (h *headerV7) Mode() []byte { return h[100:][:8] } 141func (h *headerV7) UID() []byte { return h[108:][:8] } 142func (h *headerV7) GID() []byte { return h[116:][:8] } 143func (h *headerV7) Size() []byte { return h[124:][:12] } 144func (h *headerV7) ModTime() []byte { return h[136:][:12] } 145func (h *headerV7) Chksum() []byte { return h[148:][:8] } 146func (h *headerV7) TypeFlag() []byte { return h[156:][:1] } 147func (h *headerV7) LinkName() []byte { return h[157:][:100] } 148 149type headerGNU [blockSize]byte 150 151func (h *headerGNU) V7() *headerV7 { return (*headerV7)(h) } 152func (h *headerGNU) Magic() []byte { return h[257:][:6] } 153func (h *headerGNU) Version() []byte { return h[263:][:2] } 154func (h *headerGNU) UserName() []byte { return h[265:][:32] } 155func (h *headerGNU) GroupName() []byte { return h[297:][:32] } 156func (h *headerGNU) DevMajor() []byte { return h[329:][:8] } 157func (h *headerGNU) DevMinor() []byte { return h[337:][:8] } 158func (h *headerGNU) AccessTime() []byte { return h[345:][:12] } 159func (h *headerGNU) ChangeTime() []byte { return h[357:][:12] } 160func (h *headerGNU) Sparse() sparseArray { return (sparseArray)(h[386:][:24*4+1]) } 161func (h *headerGNU) RealSize() []byte { return h[483:][:12] } 162 163type headerSTAR [blockSize]byte 164 165func (h *headerSTAR) V7() *headerV7 { return (*headerV7)(h) } 166func (h *headerSTAR) Magic() []byte { return h[257:][:6] } 167func (h *headerSTAR) Version() []byte { return h[263:][:2] } 168func (h *headerSTAR) UserName() []byte { return h[265:][:32] } 169func (h *headerSTAR) GroupName() []byte { return h[297:][:32] } 170func (h *headerSTAR) DevMajor() []byte { return h[329:][:8] } 171func (h *headerSTAR) DevMinor() []byte { return h[337:][:8] } 172func (h *headerSTAR) Prefix() []byte { return h[345:][:131] } 173func (h *headerSTAR) AccessTime() []byte { return h[476:][:12] } 174func (h *headerSTAR) ChangeTime() []byte { return h[488:][:12] } 175func (h *headerSTAR) Trailer() []byte { return h[508:][:4] } 176 177type headerUSTAR [blockSize]byte 178 179func (h *headerUSTAR) V7() *headerV7 { return (*headerV7)(h) } 180func (h *headerUSTAR) Magic() []byte { return h[257:][:6] } 181func (h *headerUSTAR) Version() []byte { return h[263:][:2] } 182func (h *headerUSTAR) UserName() []byte { return h[265:][:32] } 183func (h *headerUSTAR) GroupName() []byte { return h[297:][:32] } 184func (h *headerUSTAR) DevMajor() []byte { return h[329:][:8] } 185func (h *headerUSTAR) DevMinor() []byte { return h[337:][:8] } 186func (h *headerUSTAR) Prefix() []byte { return h[345:][:155] } 187 188type sparseArray []byte 189 190func (s sparseArray) Entry(i int) sparseNode { return (sparseNode)(s[i*24:]) } 191func (s sparseArray) IsExtended() []byte { return s[24*s.MaxEntries():][:1] } 192func (s sparseArray) MaxEntries() int { return len(s) / 24 } 193 194type sparseNode []byte 195 196func (s sparseNode) Offset() []byte { return s[00:][:12] } 197func (s sparseNode) NumBytes() []byte { return s[12:][:12] } 198