moby/pkg/archive/archive.go

package archive // import "github.com/docker/docker/pkg/archive"

import (
	"archive/tar"
	"bufio"
	"bytes"
	"compress/bzip2"
	"compress/gzip"
	"context"
	"encoding/binary"
	"fmt"
	"io"
	"os"
	"path/filepath"
	"runtime"
	"strconv"
	"strings"
	"syscall"
	"time"

	"github.com/containerd/containerd/pkg/userns"
	"github.com/docker/docker/pkg/idtools"
	"github.com/docker/docker/pkg/ioutils"
	"github.com/docker/docker/pkg/pools"
	"github.com/docker/docker/pkg/system"
	"github.com/klauspost/compress/zstd"
	"github.com/moby/patternmatcher"
	"github.com/moby/sys/sequential"
	"github.com/pkg/errors"
	"github.com/sirupsen/logrus"
	exec "golang.org/x/sys/execabs"
)

// ImpliedDirectoryMode represents the mode (Unix permissions) applied to directories that are implied by files in a
// tar, but that do not have their own header entry.
//
// The permissions mask is stored in a constant instead of locally to ensure that magic numbers do not
// proliferate in the codebase. The default value 0755 has been selected based on the default umask of 0022, and
// a convention of mkdir(1) calling mkdir(2) with permissions of 0777, resulting in a final value of 0755.
//
// This value is currently implementation-defined, and not captured in any cross-runtime specification. Thus, it is
// subject to change in Moby at any time -- image authors who require consistent or known directory permissions
// should explicitly control them by ensuring that header entries exist for any applicable path.
const ImpliedDirectoryMode = 0755

type (
	// Compression is the state represents if compressed or not.
	Compression int
	// WhiteoutFormat is the format of whiteouts unpacked
	WhiteoutFormat int

	// TarOptions wraps the tar options.
	TarOptions struct {
		IncludeFiles     []string
		ExcludePatterns  []string
		Compression      Compression
		NoLchown         bool
		IDMap            idtools.IdentityMapping
		ChownOpts        *idtools.Identity
		IncludeSourceDir bool
		// WhiteoutFormat is the expected on disk format for whiteout files.
		// This format will be converted to the standard format on pack
		// and from the standard format on unpack.
		WhiteoutFormat WhiteoutFormat
		// When unpacking, specifies whether overwriting a directory with a
		// non-directory is allowed and vice versa.
		NoOverwriteDirNonDir bool
		// For each include when creating an archive, the included name will be
		// replaced with the matching name from this map.
		RebaseNames map[string]string
		InUserNS    bool
	}
)

// Archiver implements the Archiver interface and allows the reuse of most utility functions of
// this package with a pluggable Untar function. Also, to facilitate the passing of specific id
// mappings for untar, an Archiver can be created with maps which will then be passed to Untar operations.
type Archiver struct {
	Untar     func(io.Reader, string, *TarOptions) error
	IDMapping idtools.IdentityMapping
}

// NewDefaultArchiver returns a new Archiver without any IdentityMapping
func NewDefaultArchiver() *Archiver {
	return &Archiver{Untar: Untar}
}

// breakoutError is used to differentiate errors related to breaking out
// When testing archive breakout in the unit tests, this error is expected
// in order for the test to pass.
type breakoutError error

const (
	// Uncompressed represents the uncompressed.
	Uncompressed Compression = iota
	// Bzip2 is bzip2 compression algorithm.
	Bzip2
	// Gzip is gzip compression algorithm.
	Gzip
	// Xz is xz compression algorithm.
	Xz
	// Zstd is zstd compression algorithm.
	Zstd
)

const (
	// AUFSWhiteoutFormat is the default format for whiteouts
	AUFSWhiteoutFormat WhiteoutFormat = iota
	// OverlayWhiteoutFormat formats whiteout according to the overlay
	// standard.
	OverlayWhiteoutFormat
)

// IsArchivePath checks if the (possibly compressed) file at the given path
// starts with a tar file header.
func IsArchivePath(path string) bool {
	file, err := os.Open(path)
	if err != nil {
		return false
	}
	defer file.Close()
	rdr, err := DecompressStream(file)
	if err != nil {
		return false
	}
	defer rdr.Close()
	r := tar.NewReader(rdr)
	_, err = r.Next()
	return err == nil
}

const (
	zstdMagicSkippableStart = 0x184D2A50
	zstdMagicSkippableMask  = 0xFFFFFFF0
)

var (
	bzip2Magic = []byte{0x42, 0x5A, 0x68}
	gzipMagic  = []byte{0x1F, 0x8B, 0x08}
	xzMagic    = []byte{0xFD, 0x37, 0x7A, 0x58, 0x5A, 0x00}
	zstdMagic  = []byte{0x28, 0xb5, 0x2f, 0xfd}
)

type matcher = func([]byte) bool

func magicNumberMatcher(m []byte) matcher {
	return func(source []byte) bool {
		return bytes.HasPrefix(source, m)
	}
}

// zstdMatcher detects zstd compression algorithm.
// Zstandard compressed data is made of one or more frames.
// There are two frame formats defined by Zstandard: Zstandard frames and Skippable frames.
// See https://tools.ietf.org/id/draft-kucherawy-dispatch-zstd-00.html#rfc.section.2 for more details.
func zstdMatcher() matcher {
	return func(source []byte) bool {
		if bytes.HasPrefix(source, zstdMagic) {
			// Zstandard frame
			return true
		}
		// skippable frame
		if len(source) < 8 {
			return false
		}
		// magic number from 0x184D2A50 to 0x184D2A5F.
		if binary.LittleEndian.Uint32(source[:4])&zstdMagicSkippableMask == zstdMagicSkippableStart {
			return true
		}
		return false
	}
}

// DetectCompression detects the compression algorithm of the source.
func DetectCompression(source []byte) Compression {
	compressionMap := map[Compression]matcher{
		Bzip2: magicNumberMatcher(bzip2Magic),
		Gzip:  magicNumberMatcher(gzipMagic),
		Xz:    magicNumberMatcher(xzMagic),
		Zstd:  zstdMatcher(),
	}
	for _, compression := range []Compression{Bzip2, Gzip, Xz, Zstd} {
		fn := compressionMap[compression]
		if fn(source) {
			return compression
		}
	}
	return Uncompressed
}

func xzDecompress(ctx context.Context, archive io.Reader) (io.ReadCloser, error) {
	args := []string{"xz", "-d", "-c", "-q"}

	return cmdStream(exec.CommandContext(ctx, args[0], args[1:]...), archive)
}

func gzDecompress(ctx context.Context, buf io.Reader) (io.ReadCloser, error) {
	if noPigzEnv := os.Getenv("MOBY_DISABLE_PIGZ"); noPigzEnv != "" {
		noPigz, err := strconv.ParseBool(noPigzEnv)
		if err != nil {
			logrus.WithError(err).Warn("invalid value in MOBY_DISABLE_PIGZ env var")
		}
		if noPigz {
			logrus.Debugf("Use of pigz is disabled due to MOBY_DISABLE_PIGZ=%s", noPigzEnv)
			return gzip.NewReader(buf)
		}
	}

	unpigzPath, err := exec.LookPath("unpigz")
	if err != nil {
		logrus.Debugf("unpigz binary not found, falling back to go gzip library")
		return gzip.NewReader(buf)
	}

	logrus.Debugf("Using %s to decompress", unpigzPath)

	return cmdStream(exec.CommandContext(ctx, unpigzPath, "-d", "-c"), buf)
}

func wrapReadCloser(readBuf io.ReadCloser, cancel context.CancelFunc) io.ReadCloser {
	return ioutils.NewReadCloserWrapper(readBuf, func() error {
		cancel()
		return readBuf.Close()
	})
}

// DecompressStream decompresses the archive and returns a ReaderCloser with the decompressed archive.
func DecompressStream(archive io.Reader) (io.ReadCloser, error) {
	p := pools.BufioReader32KPool
	buf := p.Get(archive)
	bs, err := buf.Peek(10)
	if err != nil && err != io.EOF {
		// Note: we'll ignore any io.EOF error because there are some odd
		// cases where the layer.tar file will be empty (zero bytes) and
		// that results in an io.EOF from the Peek() call. So, in those
		// cases we'll just treat it as a non-compressed stream and
		// that means just create an empty layer.
		// See Issue 18170
		return nil, err
	}

	compression := DetectCompression(bs)
	switch compression {
	case Uncompressed:
		readBufWrapper := p.NewReadCloserWrapper(buf, buf)
		return readBufWrapper, nil
	case Gzip:
		ctx, cancel := context.WithCancel(context.Background())

		gzReader, err := gzDecompress(ctx, buf)
		if err != nil {
			cancel()
			return nil, err
		}
		readBufWrapper := p.NewReadCloserWrapper(buf, gzReader)
		return wrapReadCloser(readBufWrapper, cancel), nil
	case Bzip2:
		bz2Reader := bzip2.NewReader(buf)
		readBufWrapper := p.NewReadCloserWrapper(buf, bz2Reader)
		return readBufWrapper, nil
	case Xz:
		ctx, cancel := context.WithCancel(context.Background())

		xzReader, err := xzDecompress(ctx, buf)
		if err != nil {
			cancel()
			return nil, err
		}
		readBufWrapper := p.NewReadCloserWrapper(buf, xzReader)
		return wrapReadCloser(readBufWrapper, cancel), nil
	case Zstd:
		zstdReader, err := zstd.NewReader(buf)
		if err != nil {
			return nil, err
		}
		readBufWrapper := p.NewReadCloserWrapper(buf, zstdReader)
		return readBufWrapper, nil
	default:
		return nil, fmt.Errorf("Unsupported compression format %s", (&compression).Extension())
	}
}

// CompressStream compresses the dest with specified compression algorithm.
func CompressStream(dest io.Writer, compression Compression) (io.WriteCloser, error) {
	p := pools.BufioWriter32KPool
	buf := p.Get(dest)
	switch compression {
	case Uncompressed:
		writeBufWrapper := p.NewWriteCloserWrapper(buf, buf)
		return writeBufWrapper, nil
	case Gzip:
		gzWriter := gzip.NewWriter(dest)
		writeBufWrapper := p.NewWriteCloserWrapper(buf, gzWriter)
		return writeBufWrapper, nil
	case Bzip2, Xz:
		// archive/bzip2 does not support writing, and there is no xz support at all
		// However, this is not a problem as docker only currently generates gzipped tars
		return nil, fmt.Errorf("Unsupported compression format %s", (&compression).Extension())
	default:
		return nil, fmt.Errorf("Unsupported compression format %s", (&compression).Extension())
	}
}

// TarModifierFunc is a function that can be passed to ReplaceFileTarWrapper to
// modify the contents or header of an entry in the archive. If the file already
// exists in the archive the TarModifierFunc will be called with the Header and
// a reader which will return the files content. If the file does not exist both
// header and content will be nil.
type TarModifierFunc func(path string, header *tar.Header, content io.Reader) (*tar.Header, []byte, error)

// ReplaceFileTarWrapper converts inputTarStream to a new tar stream. Files in the
// tar stream are modified if they match any of the keys in mods.
func ReplaceFileTarWrapper(inputTarStream io.ReadCloser, mods map[string]TarModifierFunc) io.ReadCloser {
	pipeReader, pipeWriter := io.Pipe()

	go func() {
		tarReader := tar.NewReader(inputTarStream)
		tarWriter := tar.NewWriter(pipeWriter)
		defer inputTarStream.Close()
		defer tarWriter.Close()

		modify := func(name string, original *tar.Header, modifier TarModifierFunc, tarReader io.Reader) error {
			header, data, err := modifier(name, original, tarReader)
			switch {
			case err != nil:
				return err
			case header == nil:
				return nil
			}

			if header.Name == "" {
				header.Name = name
			}
			header.Size = int64(len(data))
			if err := tarWriter.WriteHeader(header); err != nil {
				return err
			}
			if len(data) != 0 {
				if _, err := tarWriter.Write(data); err != nil {
					return err
				}
			}
			return nil
		}

		var err error
		var originalHeader *tar.Header
		for {
			originalHeader, err = tarReader.Next()
			if err == io.EOF {
				break
			}
			if err != nil {
				pipeWriter.CloseWithError(err)
				return
			}

			modifier, ok := mods[originalHeader.Name]
			if !ok {
				// No modifiers for this file, copy the header and data
				if err := tarWriter.WriteHeader(originalHeader); err != nil {
					pipeWriter.CloseWithError(err)
					return
				}
				if _, err := pools.Copy(tarWriter, tarReader); err != nil {
					pipeWriter.CloseWithError(err)
					return
				}
				continue
			}
			delete(mods, originalHeader.Name)

			if err := modify(originalHeader.Name, originalHeader, modifier, tarReader); err != nil {
				pipeWriter.CloseWithError(err)
				return
			}
		}

		// Apply the modifiers that haven't matched any files in the archive
		for name, modifier := range mods {
			if err := modify(name, nil, modifier, nil); err != nil {
				pipeWriter.CloseWithError(err)
				return
			}
		}

		pipeWriter.Close()
	}()
	return pipeReader
}

// Extension returns the extension of a file that uses the specified compression algorithm.
func (compression *Compression) Extension() string {
	switch *compression {
	case Uncompressed:
		return "tar"
	case Bzip2:
		return "tar.bz2"
	case Gzip:
		return "tar.gz"
	case Xz:
		return "tar.xz"
	case Zstd:
		return "tar.zst"
	}
	return ""
}

// nosysFileInfo hides the system-dependent info of the wrapped FileInfo to
// prevent tar.FileInfoHeader from introspecting it and potentially calling into
// glibc.
type nosysFileInfo struct {
	os.FileInfo
}

func (fi nosysFileInfo) Sys() interface{} {
	// A Sys value of type *tar.Header is safe as it is system-independent.
	// The tar.FileInfoHeader function copies the fields into the returned
	// header without performing any OS lookups.
	if sys, ok := fi.FileInfo.Sys().(*tar.Header); ok {
		return sys
	}
	return nil
}

// sysStat, if non-nil, populates hdr from system-dependent fields of fi.
var sysStat func(fi os.FileInfo, hdr *tar.Header) error

// FileInfoHeaderNoLookups creates a partially-populated tar.Header from fi.
//
// Compared to the archive/tar.FileInfoHeader function, this function is safe to
// call from a chrooted process as it does not populate fields which would
// require operating system lookups. It behaves identically to
// tar.FileInfoHeader when fi is a FileInfo value returned from
// tar.Header.FileInfo().
//
// When fi is a FileInfo for a native file, such as returned from os.Stat() and
// os.Lstat(), the returned Header value differs from one returned from
// tar.FileInfoHeader in the following ways. The Uname and Gname fields are not
// set as OS lookups would be required to populate them. The AccessTime and
// ChangeTime fields are not currently set (not yet implemented) although that
// is subject to change. Callers which require the AccessTime or ChangeTime
// fields to be zeroed should explicitly zero them out in the returned Header
// value to avoid any compatibility issues in the future.
func FileInfoHeaderNoLookups(fi os.FileInfo, link string) (*tar.Header, error) {
	hdr, err := tar.FileInfoHeader(nosysFileInfo{fi}, link)
	if err != nil {
		return nil, err
	}
	if sysStat != nil {
		return hdr, sysStat(fi, hdr)
	}
	return hdr, nil
}

// FileInfoHeader creates a populated Header from fi.
//
// Compared to the archive/tar package, this function fills in less information
// but is safe to call from a chrooted process. The AccessTime and ChangeTime
// fields are not set in the returned header, ModTime is truncated to one-second
// precision, and the Uname and Gname fields are only set when fi is a FileInfo
// value returned from tar.Header.FileInfo().
func FileInfoHeader(name string, fi os.FileInfo, link string) (*tar.Header, error) {
	hdr, err := FileInfoHeaderNoLookups(fi, link)
	if err != nil {
		return nil, err
	}
	hdr.Format = tar.FormatPAX
	hdr.ModTime = hdr.ModTime.Truncate(time.Second)
	hdr.AccessTime = time.Time{}
	hdr.ChangeTime = time.Time{}
	hdr.Mode = int64(chmodTarEntry(os.FileMode(hdr.Mode)))
	hdr.Name = canonicalTarName(name, fi.IsDir())
	return hdr, nil
}

// ReadSecurityXattrToTarHeader reads security.capability xattr from filesystem
// to a tar header
func ReadSecurityXattrToTarHeader(path string, hdr *tar.Header) error {
	const (
		// Values based on linux/include/uapi/linux/capability.h
		xattrCapsSz2    = 20
		versionOffset   = 3
		vfsCapRevision2 = 2
		vfsCapRevision3 = 3
	)
	capability, _ := system.Lgetxattr(path, "security.capability")
	if capability != nil {
		length := len(capability)
		if capability[versionOffset] == vfsCapRevision3 {
			// Convert VFS_CAP_REVISION_3 to VFS_CAP_REVISION_2 as root UID makes no
			// sense outside the user namespace the archive is built in.
			capability[versionOffset] = vfsCapRevision2
			length = xattrCapsSz2
		}
		hdr.Xattrs = make(map[string]string)
		hdr.Xattrs["security.capability"] = string(capability[:length])
	}
	return nil
}

type tarWhiteoutConverter interface {
	ConvertWrite(*tar.Header, string, os.FileInfo) (*tar.Header, error)
	ConvertRead(*tar.Header, string) (bool, error)
}

type tarAppender struct {
	TarWriter *tar.Writer
	Buffer    *bufio.Writer

	// for hardlink mapping
	SeenFiles       map[uint64]string
	IdentityMapping idtools.IdentityMapping
	ChownOpts       *idtools.Identity

	// For packing and unpacking whiteout files in the
	// non standard format. The whiteout files defined
	// by the AUFS standard are used as the tar whiteout
	// standard.
	WhiteoutConverter tarWhiteoutConverter
}

func newTarAppender(idMapping idtools.IdentityMapping, writer io.Writer, chownOpts *idtools.Identity) *tarAppender {
	return &tarAppender{
		SeenFiles:       make(map[uint64]string),
		TarWriter:       tar.NewWriter(writer),
		Buffer:          pools.BufioWriter32KPool.Get(nil),
		IdentityMapping: idMapping,
		ChownOpts:       chownOpts,
	}
}

// CanonicalTarNameForPath canonicalizes relativePath to a POSIX-style path using
// forward slashes. It is an alias for filepath.ToSlash, which is a no-op on
// Linux and Unix.
func CanonicalTarNameForPath(relativePath string) string {
	return filepath.ToSlash(relativePath)
}

// canonicalTarName provides a platform-independent and consistent POSIX-style
// path for files and directories to be archived regardless of the platform.
func canonicalTarName(name string, isDir bool) string {
	name = filepath.ToSlash(name)

	// suffix with '/' for directories
	if isDir && !strings.HasSuffix(name, "/") {
		name += "/"
	}
	return name
}

// addTarFile adds to the tar archive a file from `path` as `name`
func (ta *tarAppender) addTarFile(path, name string) error {
	fi, err := os.Lstat(path)
	if err != nil {
		return err
	}

	var link string
	if fi.Mode()&os.ModeSymlink != 0 {
		var err error
		link, err = os.Readlink(path)
		if err != nil {
			return err
		}
	}

	hdr, err := FileInfoHeader(name, fi, link)
	if err != nil {
		return err
	}
	if err := ReadSecurityXattrToTarHeader(path, hdr); err != nil {
		return err
	}

	// if it's not a directory and has more than 1 link,
	// it's hard linked, so set the type flag accordingly
	if !fi.IsDir() && hasHardlinks(fi) {
		inode, err := getInodeFromStat(fi.Sys())
		if err != nil {
			return err
		}
		// a link should have a name that it links too
		// and that linked name should be first in the tar archive
		if oldpath, ok := ta.SeenFiles[inode]; ok {
			hdr.Typeflag = tar.TypeLink
			hdr.Linkname = oldpath
			hdr.Size = 0 // This Must be here for the writer math to add up!
		} else {
			ta.SeenFiles[inode] = name
		}
	}

	// check whether the file is overlayfs whiteout
	// if yes, skip re-mapping container ID mappings.
	isOverlayWhiteout := fi.Mode()&os.ModeCharDevice != 0 && hdr.Devmajor == 0 && hdr.Devminor == 0

	// handle re-mapping container ID mappings back to host ID mappings before
	// writing tar headers/files. We skip whiteout files because they were written
	// by the kernel and already have proper ownership relative to the host
	if !isOverlayWhiteout && !strings.HasPrefix(filepath.Base(hdr.Name), WhiteoutPrefix) && !ta.IdentityMapping.Empty() {
		fileIDPair, err := getFileUIDGID(fi.Sys())
		if err != nil {
			return err
		}
		hdr.Uid, hdr.Gid, err = ta.IdentityMapping.ToContainer(fileIDPair)
		if err != nil {
			return err
		}
	}

	// explicitly override with ChownOpts
	if ta.ChownOpts != nil {
		hdr.Uid = ta.ChownOpts.UID
		hdr.Gid = ta.ChownOpts.GID
	}

	if ta.WhiteoutConverter != nil {
		wo, err := ta.WhiteoutConverter.ConvertWrite(hdr, path, fi)
		if err != nil {
			return err
		}

		// If a new whiteout file exists, write original hdr, then
		// replace hdr with wo to be written after. Whiteouts should
		// always be written after the original. Note the original
		// hdr may have been updated to be a whiteout with returning
		// a whiteout header
		if wo != nil {
			if err := ta.TarWriter.WriteHeader(hdr); err != nil {
				return err
			}
			if hdr.Typeflag == tar.TypeReg && hdr.Size > 0 {
				return fmt.Errorf("tar: cannot use whiteout for non-empty file")
			}
			hdr = wo
		}
	}

	if err := ta.TarWriter.WriteHeader(hdr); err != nil {
		return err
	}

	if hdr.Typeflag == tar.TypeReg && hdr.Size > 0 {
		// We use sequential file access to avoid depleting the standby list on
		// Windows. On Linux, this equates to a regular os.Open.
		file, err := sequential.Open(path)
		if err != nil {
			return err
		}

		ta.Buffer.Reset(ta.TarWriter)
		defer ta.Buffer.Reset(nil)
		_, err = io.Copy(ta.Buffer, file)
		file.Close()
		if err != nil {
			return err
		}
		err = ta.Buffer.Flush()
		if err != nil {
			return err
		}
	}

	return nil
}

func createTarFile(path, extractDir string, hdr *tar.Header, reader io.Reader, Lchown bool, chownOpts *idtools.Identity, inUserns bool) error {
	// hdr.Mode is in linux format, which we can use for sycalls,
	// but for os.Foo() calls we need the mode converted to os.FileMode,
	// so use hdrInfo.Mode() (they differ for e.g. setuid bits)
	hdrInfo := hdr.FileInfo()

	switch hdr.Typeflag {
	case tar.TypeDir:
		// Create directory unless it exists as a directory already.
		// In that case we just want to merge the two
		if fi, err := os.Lstat(path); !(err == nil && fi.IsDir()) {
			if err := os.Mkdir(path, hdrInfo.Mode()); err != nil {
				return err
			}
		}

	case tar.TypeReg, tar.TypeRegA:
		// Source is regular file. We use sequential file access to avoid depleting
		// the standby list on Windows. On Linux, this equates to a regular os.OpenFile.
		file, err := sequential.OpenFile(path, os.O_CREATE|os.O_WRONLY, hdrInfo.Mode())
		if err != nil {
			return err
		}
		if _, err := io.Copy(file, reader); err != nil {
			file.Close()
			return err
		}
		file.Close()

	case tar.TypeBlock, tar.TypeChar:
		if inUserns { // cannot create devices in a userns
			return nil
		}
		// Handle this is an OS-specific way
		if err := handleTarTypeBlockCharFifo(hdr, path); err != nil {
			return err
		}

	case tar.TypeFifo:
		// Handle this is an OS-specific way
		if err := handleTarTypeBlockCharFifo(hdr, path); err != nil {
			return err
		}

	case tar.TypeLink:
		// #nosec G305 -- The target path is checked for path traversal.
		targetPath := filepath.Join(extractDir, hdr.Linkname)
		// check for hardlink breakout
		if !strings.HasPrefix(targetPath, extractDir) {
			return breakoutError(fmt.Errorf("invalid hardlink %q -> %q", targetPath, hdr.Linkname))
		}
		if err := os.Link(targetPath, path); err != nil {
			return err
		}

	case tar.TypeSymlink:
		// 	path 				-> hdr.Linkname = targetPath
		// e.g. /extractDir/path/to/symlink 	-> ../2/file	= /extractDir/path/2/file
		targetPath := filepath.Join(filepath.Dir(path), hdr.Linkname) // #nosec G305 -- The target path is checked for path traversal.

		// the reason we don't need to check symlinks in the path (with FollowSymlinkInScope) is because
		// that symlink would first have to be created, which would be caught earlier, at this very check:
		if !strings.HasPrefix(targetPath, extractDir) {
			return breakoutError(fmt.Errorf("invalid symlink %q -> %q", path, hdr.Linkname))
		}
		if err := os.Symlink(hdr.Linkname, path); err != nil {
			return err
		}

	case tar.TypeXGlobalHeader:
		logrus.Debug("PAX Global Extended Headers found and ignored")
		return nil

	default:
		return fmt.Errorf("unhandled tar header type %d", hdr.Typeflag)
	}

	// Lchown is not supported on Windows.
	if Lchown && runtime.GOOS != "windows" {
		if chownOpts == nil {
			chownOpts = &idtools.Identity{UID: hdr.Uid, GID: hdr.Gid}
		}
		if err := os.Lchown(path, chownOpts.UID, chownOpts.GID); err != nil {
			msg := "failed to Lchown %q for UID %d, GID %d"
			if errors.Is(err, syscall.EINVAL) && userns.RunningInUserNS() {
				msg += " (try increasing the number of subordinate IDs in /etc/subuid and /etc/subgid)"
			}
			return errors.Wrapf(err, msg, path, hdr.Uid, hdr.Gid)
		}
	}

	var errors []string
	for key, value := range hdr.Xattrs {
		if err := system.Lsetxattr(path, key, []byte(value), 0); err != nil {
			if err == syscall.ENOTSUP || err == syscall.EPERM {
				// We ignore errors here because not all graphdrivers support
				// xattrs *cough* old versions of AUFS *cough*. However only
				// ENOTSUP should be emitted in that case, otherwise we still
				// bail.
				// EPERM occurs if modifying xattrs is not allowed. This can
				// happen when running in userns with restrictions (ChromeOS).
				errors = append(errors, err.Error())
				continue
			}
			return err
		}
	}

	if len(errors) > 0 {
		logrus.WithFields(logrus.Fields{
			"errors": errors,
		}).Warn("ignored xattrs in archive: underlying filesystem doesn't support them")
	}

	// There is no LChmod, so ignore mode for symlink. Also, this
	// must happen after chown, as that can modify the file mode
	if err := handleLChmod(hdr, path, hdrInfo); err != nil {
		return err
	}

	aTime := hdr.AccessTime
	if aTime.Before(hdr.ModTime) {
		// Last access time should never be before last modified time.
		aTime = hdr.ModTime
	}

	// system.Chtimes doesn't support a NOFOLLOW flag atm
	if hdr.Typeflag == tar.TypeLink {
		if fi, err := os.Lstat(hdr.Linkname); err == nil && (fi.Mode()&os.ModeSymlink == 0) {
			if err := system.Chtimes(path, aTime, hdr.ModTime); err != nil {
				return err
			}
		}
	} else if hdr.Typeflag != tar.TypeSymlink {
		if err := system.Chtimes(path, aTime, hdr.ModTime); err != nil {
			return err
		}
	} else {
		ts := []syscall.Timespec{timeToTimespec(aTime), timeToTimespec(hdr.ModTime)}
		if err := system.LUtimesNano(path, ts); err != nil && err != system.ErrNotSupportedPlatform {
			return err
		}
	}
	return nil
}

// Tar creates an archive from the directory at `path`, and returns it as a
// stream of bytes.
func Tar(path string, compression Compression) (io.ReadCloser, error) {
	return TarWithOptions(path, &TarOptions{Compression: compression})
}

// TarWithOptions creates an archive from the directory at `path`, only including files whose relative
// paths are included in `options.IncludeFiles` (if non-nil) or not in `options.ExcludePatterns`.
func TarWithOptions(srcPath string, options *TarOptions) (io.ReadCloser, error) {
	// Fix the source path to work with long path names. This is a no-op
	// on platforms other than Windows.
	srcPath = fixVolumePathPrefix(srcPath)

	pm, err := patternmatcher.New(options.ExcludePatterns)
	if err != nil {
		return nil, err
	}

	pipeReader, pipeWriter := io.Pipe()

	compressWriter, err := CompressStream(pipeWriter, options.Compression)
	if err != nil {
		return nil, err
	}

	whiteoutConverter, err := getWhiteoutConverter(options.WhiteoutFormat, options.InUserNS)
	if err != nil {
		return nil, err
	}

	go func() {
		ta := newTarAppender(
			options.IDMap,
			compressWriter,
			options.ChownOpts,
		)
		ta.WhiteoutConverter = whiteoutConverter

		defer func() {
			// Make sure to check the error on Close.
			if err := ta.TarWriter.Close(); err != nil {
				logrus.Errorf("Can't close tar writer: %s", err)
			}
			if err := compressWriter.Close(); err != nil {
				logrus.Errorf("Can't close compress writer: %s", err)
			}
			if err := pipeWriter.Close(); err != nil {
				logrus.Errorf("Can't close pipe writer: %s", err)
			}
		}()

		// this buffer is needed for the duration of this piped stream
		defer pools.BufioWriter32KPool.Put(ta.Buffer)

		// In general we log errors here but ignore them because
		// during e.g. a diff operation the container can continue
		// mutating the filesystem and we can see transient errors
		// from this

		stat, err := os.Lstat(srcPath)
		if err != nil {
			return
		}

		if !stat.IsDir() {
			// We can't later join a non-dir with any includes because the
			// 'walk' will error if "file/." is stat-ed and "file" is not a
			// directory. So, we must split the source path and use the
			// basename as the include.
			if len(options.IncludeFiles) > 0 {
				logrus.Warn("Tar: Can't archive a file with includes")
			}

			dir, base := SplitPathDirEntry(srcPath)
			srcPath = dir
			options.IncludeFiles = []string{base}
		}

		if len(options.IncludeFiles) == 0 {
			options.IncludeFiles = []string{"."}
		}

		seen := make(map[string]bool)

		for _, include := range options.IncludeFiles {
			rebaseName := options.RebaseNames[include]

			var (
				parentMatchInfo []patternmatcher.MatchInfo
				parentDirs      []string
			)

			walkRoot := getWalkRoot(srcPath, include)
			filepath.Walk(walkRoot, func(filePath string, f os.FileInfo, err error) error {
				if err != nil {
					logrus.Errorf("Tar: Can't stat file %s to tar: %s", srcPath, err)
					return nil
				}

				relFilePath, err := filepath.Rel(srcPath, filePath)
				if err != nil || (!options.IncludeSourceDir && relFilePath == "." && f.IsDir()) {
					// Error getting relative path OR we are looking
					// at the source directory path. Skip in both situations.
					return nil
				}

				if options.IncludeSourceDir && include == "." && relFilePath != "." {
					relFilePath = strings.Join([]string{".", relFilePath}, string(filepath.Separator))
				}

				skip := false

				// If "include" is an exact match for the current file
				// then even if there's an "excludePatterns" pattern that
				// matches it, don't skip it. IOW, assume an explicit 'include'
				// is asking for that file no matter what - which is true
				// for some files, like .dockerignore and Dockerfile (sometimes)
				if include != relFilePath {
					for len(parentDirs) != 0 {
						lastParentDir := parentDirs[len(parentDirs)-1]
						if strings.HasPrefix(relFilePath, lastParentDir+string(os.PathSeparator)) {
							break
						}
						parentDirs = parentDirs[:len(parentDirs)-1]
						parentMatchInfo = parentMatchInfo[:len(parentMatchInfo)-1]
					}

					var matchInfo patternmatcher.MatchInfo
					if len(parentMatchInfo) != 0 {
						skip, matchInfo, err = pm.MatchesUsingParentResults(relFilePath, parentMatchInfo[len(parentMatchInfo)-1])
					} else {
						skip, matchInfo, err = pm.MatchesUsingParentResults(relFilePath, patternmatcher.MatchInfo{})
					}
					if err != nil {
						logrus.Errorf("Error matching %s: %v", relFilePath, err)
						return err
					}

					if f.IsDir() {
						parentDirs = append(parentDirs, relFilePath)
						parentMatchInfo = append(parentMatchInfo, matchInfo)
					}
				}

				if skip {
					// If we want to skip this file and its a directory
					// then we should first check to see if there's an
					// excludes pattern (e.g. !dir/file) that starts with this
					// dir. If so then we can't skip this dir.

					// Its not a dir then so we can just return/skip.
					if !f.IsDir() {
						return nil
					}

					// No exceptions (!...) in patterns so just skip dir
					if !pm.Exclusions() {
						return filepath.SkipDir
					}

					dirSlash := relFilePath + string(filepath.Separator)

					for _, pat := range pm.Patterns() {
						if !pat.Exclusion() {
							continue
						}
						if strings.HasPrefix(pat.String()+string(filepath.Separator), dirSlash) {
							// found a match - so can't skip this dir
							return nil
						}
					}

					// No matching exclusion dir so just skip dir
					return filepath.SkipDir
				}

				if seen[relFilePath] {
					return nil
				}
				seen[relFilePath] = true

				// Rename the base resource.
				if rebaseName != "" {
					var replacement string
					if rebaseName != string(filepath.Separator) {
						// Special case the root directory to replace with an
						// empty string instead so that we don't end up with
						// double slashes in the paths.
						replacement = rebaseName
					}

					relFilePath = strings.Replace(relFilePath, include, replacement, 1)
				}

				if err := ta.addTarFile(filePath, relFilePath); err != nil {
					logrus.Errorf("Can't add file %s to tar: %s", filePath, err)
					// if pipe is broken, stop writing tar stream to it
					if err == io.ErrClosedPipe {
						return err
					}
				}
				return nil
			})
		}
	}()

	return pipeReader, nil
}

// Unpack unpacks the decompressedArchive to dest with options.
func Unpack(decompressedArchive io.Reader, dest string, options *TarOptions) error {
	tr := tar.NewReader(decompressedArchive)
	trBuf := pools.BufioReader32KPool.Get(nil)
	defer pools.BufioReader32KPool.Put(trBuf)

	var dirs []*tar.Header
	whiteoutConverter, err := getWhiteoutConverter(options.WhiteoutFormat, options.InUserNS)
	if err != nil {
		return err
	}

	// Iterate through the files in the archive.
loop:
	for {
		hdr, err := tr.Next()
		if err == io.EOF {
			// end of tar archive
			break
		}
		if err != nil {
			return err
		}

		// ignore XGlobalHeader early to avoid creating parent directories for them
		if hdr.Typeflag == tar.TypeXGlobalHeader {
			logrus.Debugf("PAX Global Extended Headers found for %s and ignored", hdr.Name)
			continue
		}

		// Normalize name, for safety and for a simple is-root check
		// This keeps "../" as-is, but normalizes "/../" to "/". Or Windows:
		// This keeps "..\" as-is, but normalizes "\..\" to "\".
		hdr.Name = filepath.Clean(hdr.Name)

		for _, exclude := range options.ExcludePatterns {
			if strings.HasPrefix(hdr.Name, exclude) {
				continue loop
			}
		}

		// Ensure that the parent directory exists.
		err = createImpliedDirectories(dest, hdr, options)
		if err != nil {
			return err
		}

		// #nosec G305 -- The joined path is checked for path traversal.
		path := filepath.Join(dest, hdr.Name)
		rel, err := filepath.Rel(dest, path)
		if err != nil {
			return err
		}
		if strings.HasPrefix(rel, ".."+string(os.PathSeparator)) {
			return breakoutError(fmt.Errorf("%q is outside of %q", hdr.Name, dest))
		}

		// If path exits we almost always just want to remove and replace it
		// The only exception is when it is a directory *and* the file from
		// the layer is also a directory. Then we want to merge them (i.e.
		// just apply the metadata from the layer).
		if fi, err := os.Lstat(path); err == nil {
			if options.NoOverwriteDirNonDir && fi.IsDir() && hdr.Typeflag != tar.TypeDir {
				// If NoOverwriteDirNonDir is true then we cannot replace
				// an existing directory with a non-directory from the archive.
				return fmt.Errorf("cannot overwrite directory %q with non-directory %q", path, dest)
			}

			if options.NoOverwriteDirNonDir && !fi.IsDir() && hdr.Typeflag == tar.TypeDir {
				// If NoOverwriteDirNonDir is true then we cannot replace
				// an existing non-directory with a directory from the archive.
				return fmt.Errorf("cannot overwrite non-directory %q with directory %q", path, dest)
			}

			if fi.IsDir() && hdr.Name == "." {
				continue
			}

			if !(fi.IsDir() && hdr.Typeflag == tar.TypeDir) {
				if err := os.RemoveAll(path); err != nil {
					return err
				}
			}
		}
		trBuf.Reset(tr)

		if err := remapIDs(options.IDMap, hdr); err != nil {
			return err
		}

		if whiteoutConverter != nil {
			writeFile, err := whiteoutConverter.ConvertRead(hdr, path)
			if err != nil {
				return err
			}
			if !writeFile {
				continue
			}
		}

		if err := createTarFile(path, dest, hdr, trBuf, !options.NoLchown, options.ChownOpts, options.InUserNS); err != nil {
			return err
		}

		// Directory mtimes must be handled at the end to avoid further
		// file creation in them to modify the directory mtime
		if hdr.Typeflag == tar.TypeDir {
			dirs = append(dirs, hdr)
		}
	}

	for _, hdr := range dirs {
		// #nosec G305 -- The header was checked for path traversal before it was appended to the dirs slice.
		path := filepath.Join(dest, hdr.Name)

		if err := system.Chtimes(path, hdr.AccessTime, hdr.ModTime); err != nil {
			return err
		}
	}
	return nil
}

// createImpliedDirectories will create all parent directories of the current path with default permissions, if they do
// not already exist. This is possible as the tar format supports 'implicit' directories, where their existence is
// defined by the paths of files in the tar, but there are no header entries for the directories themselves, and thus
// we most both create them and choose metadata like permissions.
//
// The caller should have performed filepath.Clean(hdr.Name), so hdr.Name will now be in the filepath format for the OS
// on which the daemon is running. This precondition is required because this function assumes a OS-specific path
// separator when checking that a path is not the root.
func createImpliedDirectories(dest string, hdr *tar.Header, options *TarOptions) error {
	// Not the root directory, ensure that the parent directory exists
	if !strings.HasSuffix(hdr.Name, string(os.PathSeparator)) {
		parent := filepath.Dir(hdr.Name)
		parentPath := filepath.Join(dest, parent)
		if _, err := os.Lstat(parentPath); err != nil && os.IsNotExist(err) {
			// RootPair() is confined inside this loop as most cases will not require a call, so we can spend some
			// unneeded function calls in the uncommon case to encapsulate logic -- implied directories are a niche
			// usage that reduces the portability of an image.
			rootIDs := options.IDMap.RootPair()

			err = idtools.MkdirAllAndChownNew(parentPath, ImpliedDirectoryMode, rootIDs)
			if err != nil {
				return err
			}
		}
	}

	return nil
}

// Untar reads a stream of bytes from `archive`, parses it as a tar archive,
// and unpacks it into the directory at `dest`.
// The archive may be compressed with one of the following algorithms:
// identity (uncompressed), gzip, bzip2, xz.
//
// FIXME: specify behavior when target path exists vs. doesn't exist.
func Untar(tarArchive io.Reader, dest string, options *TarOptions) error {
	return untarHandler(tarArchive, dest, options, true)
}

// UntarUncompressed reads a stream of bytes from `archive`, parses it as a tar archive,
// and unpacks it into the directory at `dest`.
// The archive must be an uncompressed stream.
func UntarUncompressed(tarArchive io.Reader, dest string, options *TarOptions) error {
	return untarHandler(tarArchive, dest, options, false)
}

// Handler for teasing out the automatic decompression
func untarHandler(tarArchive io.Reader, dest string, options *TarOptions, decompress bool) error {
	if tarArchive == nil {
		return fmt.Errorf("Empty archive")
	}
	dest = filepath.Clean(dest)
	if options == nil {
		options = &TarOptions{}
	}
	if options.ExcludePatterns == nil {
		options.ExcludePatterns = []string{}
	}

	r := tarArchive
	if decompress {
		decompressedArchive, err := DecompressStream(tarArchive)
		if err != nil {
			return err
		}
		defer decompressedArchive.Close()
		r = decompressedArchive
	}

	return Unpack(r, dest, options)
}

// TarUntar is a convenience function which calls Tar and Untar, with the output of one piped into the other.
// If either Tar or Untar fails, TarUntar aborts and returns the error.
func (archiver *Archiver) TarUntar(src, dst string) error {
	archive, err := TarWithOptions(src, &TarOptions{Compression: Uncompressed})
	if err != nil {
		return err
	}
	defer archive.Close()
	options := &TarOptions{
		IDMap: archiver.IDMapping,
	}
	return archiver.Untar(archive, dst, options)
}

// UntarPath untar a file from path to a destination, src is the source tar file path.
func (archiver *Archiver) UntarPath(src, dst string) error {
	archive, err := os.Open(src)
	if err != nil {
		return err
	}
	defer archive.Close()
	options := &TarOptions{
		IDMap: archiver.IDMapping,
	}
	return archiver.Untar(archive, dst, options)
}

// CopyWithTar creates a tar archive of filesystem path `src`, and
// unpacks it at filesystem path `dst`.
// The archive is streamed directly with fixed buffering and no
// intermediary disk IO.
func (archiver *Archiver) CopyWithTar(src, dst string) error {
	srcSt, err := os.Stat(src)
	if err != nil {
		return err
	}
	if !srcSt.IsDir() {
		return archiver.CopyFileWithTar(src, dst)
	}

	// if this Archiver is set up with ID mapping we need to create
	// the new destination directory with the remapped root UID/GID pair
	// as owner
	rootIDs := archiver.IDMapping.RootPair()
	// Create dst, copy src's content into it
	if err := idtools.MkdirAllAndChownNew(dst, 0755, rootIDs); err != nil {
		return err
	}
	return archiver.TarUntar(src, dst)
}

// CopyFileWithTar emulates the behavior of the 'cp' command-line
// for a single file. It copies a regular file from path `src` to
// path `dst`, and preserves all its metadata.
func (archiver *Archiver) CopyFileWithTar(src, dst string) (err error) {
	srcSt, err := os.Stat(src)
	if err != nil {
		return err
	}

	if srcSt.IsDir() {
		return fmt.Errorf("Can't copy a directory")
	}

	// Clean up the trailing slash. This must be done in an operating
	// system specific manner.
	if dst[len(dst)-1] == os.PathSeparator {
		dst = filepath.Join(dst, filepath.Base(src))
	}
	// Create the holding directory if necessary
	if err := system.MkdirAll(filepath.Dir(dst), 0700); err != nil {
		return err
	}

	r, w := io.Pipe()
	errC := make(chan error, 1)

	go func() {
		defer close(errC)

		errC <- func() error {
			defer w.Close()

			srcF, err := os.Open(src)
			if err != nil {
				return err
			}
			defer srcF.Close()

			hdr, err := FileInfoHeaderNoLookups(srcSt, "")
			if err != nil {
				return err
			}
			hdr.Format = tar.FormatPAX
			hdr.ModTime = hdr.ModTime.Truncate(time.Second)
			hdr.AccessTime = time.Time{}
			hdr.ChangeTime = time.Time{}
			hdr.Name = filepath.Base(dst)
			hdr.Mode = int64(chmodTarEntry(os.FileMode(hdr.Mode)))

			if err := remapIDs(archiver.IDMapping, hdr); err != nil {
				return err
			}

			tw := tar.NewWriter(w)
			defer tw.Close()
			if err := tw.WriteHeader(hdr); err != nil {
				return err
			}
			if _, err := io.Copy(tw, srcF); err != nil {
				return err
			}
			return nil
		}()
	}()
	defer func() {
		if er := <-errC; err == nil && er != nil {
			err = er
		}
	}()

	err = archiver.Untar(r, filepath.Dir(dst), nil)
	if err != nil {
		r.CloseWithError(err)
	}
	return err
}

// IdentityMapping returns the IdentityMapping of the archiver.
func (archiver *Archiver) IdentityMapping() idtools.IdentityMapping {
	return archiver.IDMapping
}

func remapIDs(idMapping idtools.IdentityMapping, hdr *tar.Header) error {
	ids, err := idMapping.ToHost(idtools.Identity{UID: hdr.Uid, GID: hdr.Gid})
	hdr.Uid, hdr.Gid = ids.UID, ids.GID
	return err
}

// cmdStream executes a command, and returns its stdout as a stream.
// If the command fails to run or doesn't complete successfully, an error
// will be returned, including anything written on stderr.
func cmdStream(cmd *exec.Cmd, input io.Reader) (io.ReadCloser, error) {
	cmd.Stdin = input
	pipeR, pipeW := io.Pipe()
	cmd.Stdout = pipeW
	var errBuf bytes.Buffer
	cmd.Stderr = &errBuf

	// Run the command and return the pipe
	if err := cmd.Start(); err != nil {
		return nil, err
	}

	// Ensure the command has exited before we clean anything up
	done := make(chan struct{})

	// Copy stdout to the returned pipe
	go func() {
		if err := cmd.Wait(); err != nil {
			pipeW.CloseWithError(fmt.Errorf("%s: %s", err, errBuf.String()))
		} else {
			pipeW.Close()
		}
		close(done)
	}()

	return ioutils.NewReadCloserWrapper(pipeR, func() error {
		// Close pipeR, and then wait for the command to complete before returning. We have to close pipeR first, as
		// cmd.Wait waits for any non-file stdout/stderr/stdin to close.
		err := pipeR.Close()
		<-done
		return err
	}), nil
}

// NewTempArchive reads the content of src into a temporary file, and returns the contents
// of that file as an archive. The archive can only be read once - as soon as reading completes,
// the file will be deleted.
func NewTempArchive(src io.Reader, dir string) (*TempArchive, error) {
	f, err := os.CreateTemp(dir, "")
	if err != nil {
		return nil, err
	}
	if _, err := io.Copy(f, src); err != nil {
		return nil, err
	}
	if _, err := f.Seek(0, 0); err != nil {
		return nil, err
	}
	st, err := f.Stat()
	if err != nil {
		return nil, err
	}
	size := st.Size()
	return &TempArchive{File: f, Size: size}, nil
}

// TempArchive is a temporary archive. The archive can only be read once - as soon as reading completes,
// the file will be deleted.
type TempArchive struct {
	*os.File
	Size   int64 // Pre-computed from Stat().Size() as a convenience
	read   int64
	closed bool
}

// Close closes the underlying file if it's still open, or does a no-op
// to allow callers to try to close the TempArchive multiple times safely.
func (archive *TempArchive) Close() error {
	if archive.closed {
		return nil
	}

	archive.closed = true

	return archive.File.Close()
}

func (archive *TempArchive) Read(data []byte) (int, error) {
	n, err := archive.File.Read(data)
	archive.read += int64(n)
	if err != nil || archive.read == archive.Size {
		archive.Close()
		os.Remove(archive.File.Name())
	}
	return n, err
}