archive/tar: replace with one from go-1.11

The RawAccounting changes are to be ported on top.

Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>

archive/tar/common.go

@@ -3,70 +3,528 @@
 // license that can be found in the LICENSE file.
 
 // Package tar implements access to tar archives.
-// It aims to cover most of the variations, including those produced
-// by GNU and BSD tars.
 //
-// References:
-//   http://www.freebsd.org/cgi/man.cgi?query=tar&sektion=5
-//   http://www.gnu.org/software/tar/manual/html_node/Standard.html
-//   http://pubs.opengroup.org/onlinepubs/9699919799/utilities/pax.html
+// Tape archives (tar) are a file format for storing a sequence of files that
+// can be read and written in a streaming manner.
+// This package aims to cover most variations of the format,
+// including those produced by GNU and BSD tar tools.
 package tar
 
 import (
-	"bytes"
 	"errors"
 	"fmt"
+	"math"
 	"os"
 	"path"
+	"reflect"
+	"strconv"
+	"strings"
 	"time"
 )
 
-const (
-	blockSize = 512
+// BUG: Use of the Uid and Gid fields in Header could overflow on 32-bit
+// architectures. If a large value is encountered when decoding, the result
+// stored in Header will be the truncated version.
 
-	// Types
-	TypeReg           = '0'    // regular file
-	TypeRegA          = '\x00' // regular file
-	TypeLink          = '1'    // hard link
-	TypeSymlink       = '2'    // symbolic link
-	TypeChar          = '3'    // character device node
-	TypeBlock         = '4'    // block device node
-	TypeDir           = '5'    // directory
-	TypeFifo          = '6'    // fifo node
-	TypeCont          = '7'    // reserved
-	TypeXHeader       = 'x'    // extended header
-	TypeXGlobalHeader = 'g'    // global extended header
-	TypeGNULongName   = 'L'    // Next file has a long name
-	TypeGNULongLink   = 'K'    // Next file symlinks to a file w/ a long name
-	TypeGNUSparse     = 'S'    // sparse file
+var (
+	ErrHeader          = errors.New("archive/tar: invalid tar header")
+	ErrWriteTooLong    = errors.New("archive/tar: write too long")
+	ErrFieldTooLong    = errors.New("archive/tar: header field too long")
+	ErrWriteAfterClose = errors.New("archive/tar: write after close")
+	errMissData        = errors.New("archive/tar: sparse file references non-existent data")
+	errUnrefData       = errors.New("archive/tar: sparse file contains unreferenced data")
+	errWriteHole       = errors.New("archive/tar: write non-NUL byte in sparse hole")
 )
 
+type headerError []string
+
+func (he headerError) Error() string {
+	const prefix = "archive/tar: cannot encode header"
+	var ss []string
+	for _, s := range he {
+		if s != "" {
+			ss = append(ss, s)
+		}
+	}
+	if len(ss) == 0 {
+		return prefix
+	}
+	return fmt.Sprintf("%s: %v", prefix, strings.Join(ss, "; and "))
+}
+
+// Type flags for Header.Typeflag.
+const (
+	// Type '0' indicates a regular file.
+	TypeReg  = '0'
+	TypeRegA = '\x00' // Deprecated: Use TypeReg instead.
+
+	// Type '1' to '6' are header-only flags and may not have a data body.
+	TypeLink    = '1' // Hard link
+	TypeSymlink = '2' // Symbolic link
+	TypeChar    = '3' // Character device node
+	TypeBlock   = '4' // Block device node
+	TypeDir     = '5' // Directory
+	TypeFifo    = '6' // FIFO node
+
+	// Type '7' is reserved.
+	TypeCont = '7'
+
+	// Type 'x' is used by the PAX format to store key-value records that
+	// are only relevant to the next file.
+	// This package transparently handles these types.
+	TypeXHeader = 'x'
+
+	// Type 'g' is used by the PAX format to store key-value records that
+	// are relevant to all subsequent files.
+	// This package only supports parsing and composing such headers,
+	// but does not currently support persisting the global state across files.
+	TypeXGlobalHeader = 'g'
+
+	// Type 'S' indicates a sparse file in the GNU format.
+	TypeGNUSparse = 'S'
+
+	// Types 'L' and 'K' are used by the GNU format for a meta file
+	// used to store the path or link name for the next file.
+	// This package transparently handles these types.
+	TypeGNULongName = 'L'
+	TypeGNULongLink = 'K'
+)
+
+// Keywords for PAX extended header records.
+const (
+	paxNone     = "" // Indicates that no PAX key is suitable
+	paxPath     = "path"
+	paxLinkpath = "linkpath"
+	paxSize     = "size"
+	paxUid      = "uid"
+	paxGid      = "gid"
+	paxUname    = "uname"
+	paxGname    = "gname"
+	paxMtime    = "mtime"
+	paxAtime    = "atime"
+	paxCtime    = "ctime"   // Removed from later revision of PAX spec, but was valid
+	paxCharset  = "charset" // Currently unused
+	paxComment  = "comment" // Currently unused
+
+	paxSchilyXattr = "SCHILY.xattr."
+
+	// Keywords for GNU sparse files in a PAX extended header.
+	paxGNUSparse          = "GNU.sparse."
+	paxGNUSparseNumBlocks = "GNU.sparse.numblocks"
+	paxGNUSparseOffset    = "GNU.sparse.offset"
+	paxGNUSparseNumBytes  = "GNU.sparse.numbytes"
+	paxGNUSparseMap       = "GNU.sparse.map"
+	paxGNUSparseName      = "GNU.sparse.name"
+	paxGNUSparseMajor     = "GNU.sparse.major"
+	paxGNUSparseMinor     = "GNU.sparse.minor"
+	paxGNUSparseSize      = "GNU.sparse.size"
+	paxGNUSparseRealSize  = "GNU.sparse.realsize"
+)
+
+// basicKeys is a set of the PAX keys for which we have built-in support.
+// This does not contain "charset" or "comment", which are both PAX-specific,
+// so adding them as first-class features of Header is unlikely.
+// Users can use the PAXRecords field to set it themselves.
+var basicKeys = map[string]bool{
+	paxPath: true, paxLinkpath: true, paxSize: true, paxUid: true, paxGid: true,
+	paxUname: true, paxGname: true, paxMtime: true, paxAtime: true, paxCtime: true,
+}
+
 // A Header represents a single header in a tar archive.
 // Some fields may not be populated.
+//
+// For forward compatibility, users that retrieve a Header from Reader.Next,
+// mutate it in some ways, and then pass it back to Writer.WriteHeader
+// should do so by creating a new Header and copying the fields
+// that they are interested in preserving.
 type Header struct {
-	Name       string    // name of header file entry
-	Mode       int64     // permission and mode bits
-	Uid        int       // user id of owner
-	Gid        int       // group id of owner
-	Size       int64     // length in bytes
-	ModTime    time.Time // modified time
-	Typeflag   byte      // type of header entry
-	Linkname   string    // target name of link
-	Uname      string    // user name of owner
-	Gname      string    // group name of owner
-	Devmajor   int64     // major number of character or block device
-	Devminor   int64     // minor number of character or block device
-	AccessTime time.Time // access time
-	ChangeTime time.Time // status change time
-	Xattrs     map[string]string
+	// Typeflag is the type of header entry.
+	// The zero value is automatically promoted to either TypeReg or TypeDir
+	// depending on the presence of a trailing slash in Name.
+	Typeflag byte
+
+	Name     string // Name of file entry
+	Linkname string // Target name of link (valid for TypeLink or TypeSymlink)
+
+	Size  int64  // Logical file size in bytes
+	Mode  int64  // Permission and mode bits
+	Uid   int    // User ID of owner
+	Gid   int    // Group ID of owner
+	Uname string // User name of owner
+	Gname string // Group name of owner
+
+	// If the Format is unspecified, then Writer.WriteHeader rounds ModTime
+	// to the nearest second and ignores the AccessTime and ChangeTime fields.
+	//
+	// To use AccessTime or ChangeTime, specify the Format as PAX or GNU.
+	// To use sub-second resolution, specify the Format as PAX.
+	ModTime    time.Time // Modification time
+	AccessTime time.Time // Access time (requires either PAX or GNU support)
+	ChangeTime time.Time // Change time (requires either PAX or GNU support)
+
+	Devmajor int64 // Major device number (valid for TypeChar or TypeBlock)
+	Devminor int64 // Minor device number (valid for TypeChar or TypeBlock)
+
+	// Xattrs stores extended attributes as PAX records under the
+	// "SCHILY.xattr." namespace.
+	//
+	// The following are semantically equivalent:
+	//  h.Xattrs[key] = value
+	//  h.PAXRecords["SCHILY.xattr."+key] = value
+	//
+	// When Writer.WriteHeader is called, the contents of Xattrs will take
+	// precedence over those in PAXRecords.
+	//
+	// Deprecated: Use PAXRecords instead.
+	Xattrs map[string]string
+
+	// PAXRecords is a map of PAX extended header records.
+	//
+	// User-defined records should have keys of the following form:
+	//	VENDOR.keyword
+	// Where VENDOR is some namespace in all uppercase, and keyword may
+	// not contain the '=' character (e.g., "GOLANG.pkg.version").
+	// The key and value should be non-empty UTF-8 strings.
+	//
+	// When Writer.WriteHeader is called, PAX records derived from the
+	// other fields in Header take precedence over PAXRecords.
+	PAXRecords map[string]string
+
+	// Format specifies the format of the tar header.
+	//
+	// This is set by Reader.Next as a best-effort guess at the format.
+	// Since the Reader liberally reads some non-compliant files,
+	// it is possible for this to be FormatUnknown.
+	//
+	// If the format is unspecified when Writer.WriteHeader is called,
+	// then it uses the first format (in the order of USTAR, PAX, GNU)
+	// capable of encoding this Header (see Format).
+	Format Format
 }
 
-// File name constants from the tar spec.
-const (
-	fileNameSize       = 100 // Maximum number of bytes in a standard tar name.
-	fileNamePrefixSize = 155 // Maximum number of ustar extension bytes.
+// sparseEntry represents a Length-sized fragment at Offset in the file.
+type sparseEntry struct{ Offset, Length int64 }
+
+func (s sparseEntry) endOffset() int64 { return s.Offset + s.Length }
+
+// A sparse file can be represented as either a sparseDatas or a sparseHoles.
+// As long as the total size is known, they are equivalent and one can be
+// converted to the other form and back. The various tar formats with sparse
+// file support represent sparse files in the sparseDatas form. That is, they
+// specify the fragments in the file that has data, and treat everything else as
+// having zero bytes. As such, the encoding and decoding logic in this package
+// deals with sparseDatas.
+//
+// However, the external API uses sparseHoles instead of sparseDatas because the
+// zero value of sparseHoles logically represents a normal file (i.e., there are
+// no holes in it). On the other hand, the zero value of sparseDatas implies
+// that the file has no data in it, which is rather odd.
+//
+// As an example, if the underlying raw file contains the 10-byte data:
+//	var compactFile = "abcdefgh"
+//
+// And the sparse map has the following entries:
+//	var spd sparseDatas = []sparseEntry{
+//		{Offset: 2,  Length: 5},  // Data fragment for 2..6
+//		{Offset: 18, Length: 3},  // Data fragment for 18..20
+//	}
+//	var sph sparseHoles = []sparseEntry{
+//		{Offset: 0,  Length: 2},  // Hole fragment for 0..1
+//		{Offset: 7,  Length: 11}, // Hole fragment for 7..17
+//		{Offset: 21, Length: 4},  // Hole fragment for 21..24
+//	}
+//
+// Then the content of the resulting sparse file with a Header.Size of 25 is:
+//	var sparseFile = "\x00"*2 + "abcde" + "\x00"*11 + "fgh" + "\x00"*4
+type (
+	sparseDatas []sparseEntry
+	sparseHoles []sparseEntry
 )
 
+// validateSparseEntries reports whether sp is a valid sparse map.
+// It does not matter whether sp represents data fragments or hole fragments.
+func validateSparseEntries(sp []sparseEntry, size int64) bool {
+	// Validate all sparse entries. These are the same checks as performed by
+	// the BSD tar utility.
+	if size < 0 {
+		return false
+	}
+	var pre sparseEntry
+	for _, cur := range sp {
+		switch {
+		case cur.Offset < 0 || cur.Length < 0:
+			return false // Negative values are never okay
+		case cur.Offset > math.MaxInt64-cur.Length:
+			return false // Integer overflow with large length
+		case cur.endOffset() > size:
+			return false // Region extends beyond the actual size
+		case pre.endOffset() > cur.Offset:
+			return false // Regions cannot overlap and must be in order
+		}
+		pre = cur
+	}
+	return true
+}
+
+// alignSparseEntries mutates src and returns dst where each fragment's
+// starting offset is aligned up to the nearest block edge, and each
+// ending offset is aligned down to the nearest block edge.
+//
+// Even though the Go tar Reader and the BSD tar utility can handle entries
+// with arbitrary offsets and lengths, the GNU tar utility can only handle
+// offsets and lengths that are multiples of blockSize.
+func alignSparseEntries(src []sparseEntry, size int64) []sparseEntry {
+	dst := src[:0]
+	for _, s := range src {
+		pos, end := s.Offset, s.endOffset()
+		pos += blockPadding(+pos) // Round-up to nearest blockSize
+		if end != size {
+			end -= blockPadding(-end) // Round-down to nearest blockSize
+		}
+		if pos < end {
+			dst = append(dst, sparseEntry{Offset: pos, Length: end - pos})
+		}
+	}
+	return dst
+}
+
+// invertSparseEntries converts a sparse map from one form to the other.
+// If the input is sparseHoles, then it will output sparseDatas and vice-versa.
+// The input must have been already validated.
+//
+// This function mutates src and returns a normalized map where:
+//	* adjacent fragments are coalesced together
+//	* only the last fragment may be empty
+//	* the endOffset of the last fragment is the total size
+func invertSparseEntries(src []sparseEntry, size int64) []sparseEntry {
+	dst := src[:0]
+	var pre sparseEntry
+	for _, cur := range src {
+		if cur.Length == 0 {
+			continue // Skip empty fragments
+		}
+		pre.Length = cur.Offset - pre.Offset
+		if pre.Length > 0 {
+			dst = append(dst, pre) // Only add non-empty fragments
+		}
+		pre.Offset = cur.endOffset()
+	}
+	pre.Length = size - pre.Offset // Possibly the only empty fragment
+	return append(dst, pre)
+}
+
+// fileState tracks the number of logical (includes sparse holes) and physical
+// (actual in tar archive) bytes remaining for the current file.
+//
+// Invariant: LogicalRemaining >= PhysicalRemaining
+type fileState interface {
+	LogicalRemaining() int64
+	PhysicalRemaining() int64
+}
+
+// allowedFormats determines which formats can be used.
+// The value returned is the logical OR of multiple possible formats.
+// If the value is FormatUnknown, then the input Header cannot be encoded
+// and an error is returned explaining why.
+//
+// As a by-product of checking the fields, this function returns paxHdrs, which
+// contain all fields that could not be directly encoded.
+// A value receiver ensures that this method does not mutate the source Header.
+func (h Header) allowedFormats() (format Format, paxHdrs map[string]string, err error) {
+	format = FormatUSTAR | FormatPAX | FormatGNU
+	paxHdrs = make(map[string]string)
+
+	var whyNoUSTAR, whyNoPAX, whyNoGNU string
+	var preferPAX bool // Prefer PAX over USTAR
+	verifyString := func(s string, size int, name, paxKey string) {
+		// NUL-terminator is optional for path and linkpath.
+		// Technically, it is required for uname and gname,
+		// but neither GNU nor BSD tar checks for it.
+		tooLong := len(s) > size
+		allowLongGNU := paxKey == paxPath || paxKey == paxLinkpath
+		if hasNUL(s) || (tooLong && !allowLongGNU) {
+			whyNoGNU = fmt.Sprintf("GNU cannot encode %s=%q", name, s)
+			format.mustNotBe(FormatGNU)
+		}
+		if !isASCII(s) || tooLong {
+			canSplitUSTAR := paxKey == paxPath
+			if _, _, ok := splitUSTARPath(s); !canSplitUSTAR || !ok {
+				whyNoUSTAR = fmt.Sprintf("USTAR cannot encode %s=%q", name, s)
+				format.mustNotBe(FormatUSTAR)
+			}
+			if paxKey == paxNone {
+				whyNoPAX = fmt.Sprintf("PAX cannot encode %s=%q", name, s)
+				format.mustNotBe(FormatPAX)
+			} else {
+				paxHdrs[paxKey] = s
+			}
+		}
+		if v, ok := h.PAXRecords[paxKey]; ok && v == s {
+			paxHdrs[paxKey] = v
+		}
+	}
+	verifyNumeric := func(n int64, size int, name, paxKey string) {
+		if !fitsInBase256(size, n) {
+			whyNoGNU = fmt.Sprintf("GNU cannot encode %s=%d", name, n)
+			format.mustNotBe(FormatGNU)
+		}
+		if !fitsInOctal(size, n) {
+			whyNoUSTAR = fmt.Sprintf("USTAR cannot encode %s=%d", name, n)
+			format.mustNotBe(FormatUSTAR)
+			if paxKey == paxNone {
+				whyNoPAX = fmt.Sprintf("PAX cannot encode %s=%d", name, n)
+				format.mustNotBe(FormatPAX)
+			} else {
+				paxHdrs[paxKey] = strconv.FormatInt(n, 10)
+			}
+		}
+		if v, ok := h.PAXRecords[paxKey]; ok && v == strconv.FormatInt(n, 10) {
+			paxHdrs[paxKey] = v
+		}
+	}
+	verifyTime := func(ts time.Time, size int, name, paxKey string) {
+		if ts.IsZero() {
+			return // Always okay
+		}
+		if !fitsInBase256(size, ts.Unix()) {
+			whyNoGNU = fmt.Sprintf("GNU cannot encode %s=%v", name, ts)
+			format.mustNotBe(FormatGNU)
+		}
+		isMtime := paxKey == paxMtime
+		fitsOctal := fitsInOctal(size, ts.Unix())
+		if (isMtime && !fitsOctal) || !isMtime {
+			whyNoUSTAR = fmt.Sprintf("USTAR cannot encode %s=%v", name, ts)
+			format.mustNotBe(FormatUSTAR)
+		}
+		needsNano := ts.Nanosecond() != 0
+		if !isMtime || !fitsOctal || needsNano {
+			preferPAX = true // USTAR may truncate sub-second measurements
+			if paxKey == paxNone {
+				whyNoPAX = fmt.Sprintf("PAX cannot encode %s=%v", name, ts)
+				format.mustNotBe(FormatPAX)
+			} else {
+				paxHdrs[paxKey] = formatPAXTime(ts)
+			}
+		}
+		if v, ok := h.PAXRecords[paxKey]; ok && v == formatPAXTime(ts) {
+			paxHdrs[paxKey] = v
+		}
+	}
+
+	// Check basic fields.
+	var blk block
+	v7 := blk.V7()
+	ustar := blk.USTAR()
+	gnu := blk.GNU()
+	verifyString(h.Name, len(v7.Name()), "Name", paxPath)
+	verifyString(h.Linkname, len(v7.LinkName()), "Linkname", paxLinkpath)
+	verifyString(h.Uname, len(ustar.UserName()), "Uname", paxUname)
+	verifyString(h.Gname, len(ustar.GroupName()), "Gname", paxGname)
+	verifyNumeric(h.Mode, len(v7.Mode()), "Mode", paxNone)
+	verifyNumeric(int64(h.Uid), len(v7.UID()), "Uid", paxUid)
+	verifyNumeric(int64(h.Gid), len(v7.GID()), "Gid", paxGid)
+	verifyNumeric(h.Size, len(v7.Size()), "Size", paxSize)
+	verifyNumeric(h.Devmajor, len(ustar.DevMajor()), "Devmajor", paxNone)
+	verifyNumeric(h.Devminor, len(ustar.DevMinor()), "Devminor", paxNone)
+	verifyTime(h.ModTime, len(v7.ModTime()), "ModTime", paxMtime)
+	verifyTime(h.AccessTime, len(gnu.AccessTime()), "AccessTime", paxAtime)
+	verifyTime(h.ChangeTime, len(gnu.ChangeTime()), "ChangeTime", paxCtime)
+
+	// Check for header-only types.
+	var whyOnlyPAX, whyOnlyGNU string
+	switch h.Typeflag {
+	case TypeReg, TypeChar, TypeBlock, TypeFifo, TypeGNUSparse:
+		// Exclude TypeLink and TypeSymlink, since they may reference directories.
+		if strings.HasSuffix(h.Name, "/") {
+			return FormatUnknown, nil, headerError{"filename may not have trailing slash"}
+		}
+	case TypeXHeader, TypeGNULongName, TypeGNULongLink:
+		return FormatUnknown, nil, headerError{"cannot manually encode TypeXHeader, TypeGNULongName, or TypeGNULongLink headers"}
+	case TypeXGlobalHeader:
+		h2 := Header{Name: h.Name, Typeflag: h.Typeflag, Xattrs: h.Xattrs, PAXRecords: h.PAXRecords, Format: h.Format}
+		if !reflect.DeepEqual(h, h2) {
+			return FormatUnknown, nil, headerError{"only PAXRecords should be set for TypeXGlobalHeader"}
+		}
+		whyOnlyPAX = "only PAX supports TypeXGlobalHeader"
+		format.mayOnlyBe(FormatPAX)
+	}
+	if !isHeaderOnlyType(h.Typeflag) && h.Size < 0 {
+		return FormatUnknown, nil, headerError{"negative size on header-only type"}
+	}
+
+	// Check PAX records.
+	if len(h.Xattrs) > 0 {
+		for k, v := range h.Xattrs {
+			paxHdrs[paxSchilyXattr+k] = v
+		}
+		whyOnlyPAX = "only PAX supports Xattrs"
+		format.mayOnlyBe(FormatPAX)
+	}
+	if len(h.PAXRecords) > 0 {
+		for k, v := range h.PAXRecords {
+			switch _, exists := paxHdrs[k]; {
+			case exists:
+				continue // Do not overwrite existing records
+			case h.Typeflag == TypeXGlobalHeader:
+				paxHdrs[k] = v // Copy all records
+			case !basicKeys[k] && !strings.HasPrefix(k, paxGNUSparse):
+				paxHdrs[k] = v // Ignore local records that may conflict
+			}
+		}
+		whyOnlyPAX = "only PAX supports PAXRecords"
+		format.mayOnlyBe(FormatPAX)
+	}
+	for k, v := range paxHdrs {
+		if !validPAXRecord(k, v) {
+			return FormatUnknown, nil, headerError{fmt.Sprintf("invalid PAX record: %q", k+" = "+v)}
+		}
+	}
+
+	// TODO(dsnet): Re-enable this when adding sparse support.
+	// See https://golang.org/issue/22735
+	/*
+		// Check sparse files.
+		if len(h.SparseHoles) > 0 || h.Typeflag == TypeGNUSparse {
+			if isHeaderOnlyType(h.Typeflag) {
+				return FormatUnknown, nil, headerError{"header-only type cannot be sparse"}
+			}
+			if !validateSparseEntries(h.SparseHoles, h.Size) {
+				return FormatUnknown, nil, headerError{"invalid sparse holes"}
+			}
+			if h.Typeflag == TypeGNUSparse {
+				whyOnlyGNU = "only GNU supports TypeGNUSparse"
+				format.mayOnlyBe(FormatGNU)
+			} else {
+				whyNoGNU = "GNU supports sparse files only with TypeGNUSparse"
+				format.mustNotBe(FormatGNU)
+			}
+			whyNoUSTAR = "USTAR does not support sparse files"
+			format.mustNotBe(FormatUSTAR)
+		}
+	*/
+
+	// Check desired format.
+	if wantFormat := h.Format; wantFormat != FormatUnknown {
+		if wantFormat.has(FormatPAX) && !preferPAX {
+			wantFormat.mayBe(FormatUSTAR) // PAX implies USTAR allowed too
+		}
+		format.mayOnlyBe(wantFormat) // Set union of formats allowed and format wanted
+	}
+	if format == FormatUnknown {
+		switch h.Format {
+		case FormatUSTAR:
+			err = headerError{"Format specifies USTAR", whyNoUSTAR, whyOnlyPAX, whyOnlyGNU}
+		case FormatPAX:
+			err = headerError{"Format specifies PAX", whyNoPAX, whyOnlyGNU}
+		case FormatGNU:
+			err = headerError{"Format specifies GNU", whyNoGNU, whyOnlyPAX}
+		default:
+			err = headerError{whyNoUSTAR, whyNoPAX, whyNoGNU, whyOnlyPAX, whyOnlyGNU}
+		}
+	}
+	return format, paxHdrs, err
+}
+
 // FileInfo returns an os.FileInfo for the Header.
 func (h *Header) FileInfo() os.FileInfo {
 	return headerFileInfo{h}
@@ -97,63 +555,43 @@ func (fi headerFileInfo) Mode() (mode os.FileMode) {
 
 	// Set setuid, setgid and sticky bits.
 	if fi.h.Mode&c_ISUID != 0 {
-		// setuid
 		mode |= os.ModeSetuid
 	}
 	if fi.h.Mode&c_ISGID != 0 {
-		// setgid
 		mode |= os.ModeSetgid
 	}
 	if fi.h.Mode&c_ISVTX != 0 {
-		// sticky
 		mode |= os.ModeSticky
 	}
 
-	// Set file mode bits.
-	// clear perm, setuid, setgid and sticky bits.
-	m := os.FileMode(fi.h.Mode) &^ 07777
-	if m == c_ISDIR {
-		// directory
+	// Set file mode bits; clear perm, setuid, setgid, and sticky bits.
+	switch m := os.FileMode(fi.h.Mode) &^ 07777; m {
+	case c_ISDIR:
 		mode |= os.ModeDir
-	}
-	if m == c_ISFIFO {
-		// named pipe (FIFO)
+	case c_ISFIFO:
 		mode |= os.ModeNamedPipe
-	}
-	if m == c_ISLNK {
-		// symbolic link
+	case c_ISLNK:
 		mode |= os.ModeSymlink
-	}
-	if m == c_ISBLK {
-		// device file
+	case c_ISBLK:
 		mode |= os.ModeDevice
-	}
-	if m == c_ISCHR {
-		// Unix character device
+	case c_ISCHR:
 		mode |= os.ModeDevice
 		mode |= os.ModeCharDevice
-	}
-	if m == c_ISSOCK {
-		// Unix domain socket
+	case c_ISSOCK:
 		mode |= os.ModeSocket
 	}
 
 	switch fi.h.Typeflag {
 	case TypeSymlink:
-		// symbolic link
 		mode |= os.ModeSymlink
 	case TypeChar:
-		// character device node
 		mode |= os.ModeDevice
 		mode |= os.ModeCharDevice
 	case TypeBlock:
-		// block device node
 		mode |= os.ModeDevice
 	case TypeDir:
-		// directory
 		mode |= os.ModeDir
 	case TypeFifo:
-		// fifo node
 		mode |= os.ModeNamedPipe
 	}
 
@@ -163,11 +601,15 @@ func (fi headerFileInfo) Mode() (mode os.FileMode) {
 // sysStat, if non-nil, populates h from system-dependent fields of fi.
 var sysStat func(fi os.FileInfo, h *Header) error
 
-// Mode constants from the tar spec.
 const (
-	c_ISUID  = 04000   // Set uid
-	c_ISGID  = 02000   // Set gid
-	c_ISVTX  = 01000   // Save text (sticky bit)
+	// Mode constants from the USTAR spec:
+	// See http://pubs.opengroup.org/onlinepubs/9699919799/utilities/pax.html#tag_20_92_13_06
+	c_ISUID = 04000 // Set uid
+	c_ISGID = 02000 // Set gid
+	c_ISVTX = 01000 // Save text (sticky bit)
+
+	// Common Unix mode constants; these are not defined in any common tar standard.
+	// Header.FileInfo understands these, but FileInfoHeader will never produce these.
 	c_ISDIR  = 040000  // Directory
 	c_ISFIFO = 010000  // FIFO
 	c_ISREG  = 0100000 // Regular file
@@ -177,33 +619,16 @@ const (
 	c_ISSOCK = 0140000 // Socket
 )
 
-// Keywords for the PAX Extended Header
-const (
-	paxAtime    = "atime"
-	paxCharset  = "charset"
-	paxComment  = "comment"
-	paxCtime    = "ctime" // please note that ctime is not a valid pax header.
-	paxGid      = "gid"
-	paxGname    = "gname"
-	paxLinkpath = "linkpath"
-	paxMtime    = "mtime"
-	paxPath     = "path"
-	paxSize     = "size"
-	paxUid      = "uid"
-	paxUname    = "uname"
-	paxXattr    = "SCHILY.xattr."
-	paxNone     = ""
-)
-
 // FileInfoHeader creates a partially-populated Header from fi.
 // If fi describes a symlink, FileInfoHeader records link as the link target.
 // If fi describes a directory, a slash is appended to the name.
-// Because os.FileInfo's Name method returns only the base name of
-// the file it describes, it may be necessary to modify the Name field
-// of the returned header to provide the full path name of the file.
+//
+// Since os.FileInfo's Name method only returns the base name of
+// the file it describes, it may be necessary to modify Header.Name
+// to provide the full path name of the file.
 func FileInfoHeader(fi os.FileInfo, link string) (*Header, error) {
 	if fi == nil {
-		return nil, errors.New("tar: FileInfo is nil")
+		return nil, errors.New("archive/tar: FileInfo is nil")
 	}
 	fm := fi.Mode()
 	h := &Header{
@@ -213,30 +638,24 @@ func FileInfoHeader(fi os.FileInfo, link string) (*Header, error) {
 	}
 	switch {
 	case fm.IsRegular():
-		h.Mode |= c_ISREG
 		h.Typeflag = TypeReg
 		h.Size = fi.Size()
 	case fi.IsDir():
 		h.Typeflag = TypeDir
-		h.Mode |= c_ISDIR
 		h.Name += "/"
 	case fm&os.ModeSymlink != 0:
 		h.Typeflag = TypeSymlink
-		h.Mode |= c_ISLNK
 		h.Linkname = link
 	case fm&os.ModeDevice != 0:
 		if fm&os.ModeCharDevice != 0 {
-			h.Mode |= c_ISCHR
 			h.Typeflag = TypeChar
 		} else {
-			h.Mode |= c_ISBLK
 			h.Typeflag = TypeBlock
 		}
 	case fm&os.ModeNamedPipe != 0:
 		h.Typeflag = TypeFifo
-		h.Mode |= c_ISFIFO
 	case fm&os.ModeSocket != 0:
-		h.Mode |= c_ISSOCK
+		return nil, fmt.Errorf("archive/tar: sockets not supported")
 	default:
 		return nil, fmt.Errorf("archive/tar: unknown file mode %v", fm)
 	}
@@ -272,6 +691,12 @@ func FileInfoHeader(fi os.FileInfo, link string) (*Header, error) {
 			h.Size = 0
 			h.Linkname = sys.Linkname
 		}
+		if sys.PAXRecords != nil {
+			h.PAXRecords = make(map[string]string)
+			for k, v := range sys.PAXRecords {
+				h.PAXRecords[k] = v
+			}
+		}
 	}
 	if sysStat != nil {
 		return h, sysStat(fi, h)
@@ -279,55 +704,6 @@ func FileInfoHeader(fi os.FileInfo, link string) (*Header, error) {
 	return h, nil
 }
 
-var zeroBlock = make([]byte, blockSize)
-
-// POSIX specifies a sum of the unsigned byte values, but the Sun tar uses signed byte values.
-// We compute and return both.
-func checksum(header []byte) (unsigned int64, signed int64) {
-	for i := 0; i < len(header); i++ {
-		if i == 148 {
-			// The chksum field (header[148:156]) is special: it should be treated as space bytes.
-			unsigned += ' ' * 8
-			signed += ' ' * 8
-			i += 7
-			continue
-		}
-		unsigned += int64(header[i])
-		signed += int64(int8(header[i]))
-	}
-	return
-}
-
-type slicer []byte
-
-func (sp *slicer) next(n int) (b []byte) {
-	s := *sp
-	b, *sp = s[0:n], s[n:]
-	return
-}
-
-func isASCII(s string) bool {
-	for _, c := range s {
-		if c >= 0x80 {
-			return false
-		}
-	}
-	return true
-}
-
-func toASCII(s string) string {
-	if isASCII(s) {
-		return s
-	}
-	var buf bytes.Buffer
-	for _, c := range s {
-		if c < 0x80 {
-			buf.WriteByte(byte(c))
-		}
-	}
-	return buf.String()
-}
-
 // isHeaderOnlyType checks if the given type flag is of the type that has no
 // data section even if a size is specified.
 func isHeaderOnlyType(flag byte) bool {
@@ -338,3 +714,10 @@ func isHeaderOnlyType(flag byte) bool {
 		return false
 	}
 }
+
+func min(a, b int64) int64 {
+	if a < b {
+		return a
+	}
+	return b
+}