moby/pkg/tarsum/tarsum.go
Aaron Lehmann 8691a77e44 Fix build cache false positives when build context tar contains unnormalized paths
If a build context tar has path names of the form 'x/./y', they will be
stored in this unnormalized form internally by tarsum. When the builder
walks the untarred directory tree and queries hashes for each relative
path, it will query paths of the form 'x/y', and they will not be found.

To correct this, have tarsum normalize path names by calling Clean.

Add a test to detect this caching false positive.

Fixes #21715

Signed-off-by: Aaron Lehmann <aaron.lehmann@docker.com>
2016-04-14 17:57:45 -07:00

295 lines
7.8 KiB
Go

// Package tarsum provides algorithms to perform checksum calculation on
// filesystem layers.
//
// The transportation of filesystems, regarding Docker, is done with tar(1)
// archives. There are a variety of tar serialization formats [2], and a key
// concern here is ensuring a repeatable checksum given a set of inputs from a
// generic tar archive. Types of transportation include distribution to and from a
// registry endpoint, saving and loading through commands or Docker daemon APIs,
// transferring the build context from client to Docker daemon, and committing the
// filesystem of a container to become an image.
//
// As tar archives are used for transit, but not preserved in many situations, the
// focus of the algorithm is to ensure the integrity of the preserved filesystem,
// while maintaining a deterministic accountability. This includes neither
// constraining the ordering or manipulation of the files during the creation or
// unpacking of the archive, nor include additional metadata state about the file
// system attributes.
package tarsum
import (
"archive/tar"
"bytes"
"compress/gzip"
"crypto"
"crypto/sha256"
"encoding/hex"
"errors"
"fmt"
"hash"
"io"
"path"
"strings"
)
const (
buf8K = 8 * 1024
buf16K = 16 * 1024
buf32K = 32 * 1024
)
// NewTarSum creates a new interface for calculating a fixed time checksum of a
// tar archive.
//
// This is used for calculating checksums of layers of an image, in some cases
// including the byte payload of the image's json metadata as well, and for
// calculating the checksums for buildcache.
func NewTarSum(r io.Reader, dc bool, v Version) (TarSum, error) {
return NewTarSumHash(r, dc, v, DefaultTHash)
}
// NewTarSumHash creates a new TarSum, providing a THash to use rather than
// the DefaultTHash.
func NewTarSumHash(r io.Reader, dc bool, v Version, tHash THash) (TarSum, error) {
headerSelector, err := getTarHeaderSelector(v)
if err != nil {
return nil, err
}
ts := &tarSum{Reader: r, DisableCompression: dc, tarSumVersion: v, headerSelector: headerSelector, tHash: tHash}
err = ts.initTarSum()
return ts, err
}
// NewTarSumForLabel creates a new TarSum using the provided TarSum version+hash label.
func NewTarSumForLabel(r io.Reader, disableCompression bool, label string) (TarSum, error) {
parts := strings.SplitN(label, "+", 2)
if len(parts) != 2 {
return nil, errors.New("tarsum label string should be of the form: {tarsum_version}+{hash_name}")
}
versionName, hashName := parts[0], parts[1]
version, ok := tarSumVersionsByName[versionName]
if !ok {
return nil, fmt.Errorf("unknown TarSum version name: %q", versionName)
}
hashConfig, ok := standardHashConfigs[hashName]
if !ok {
return nil, fmt.Errorf("unknown TarSum hash name: %q", hashName)
}
tHash := NewTHash(hashConfig.name, hashConfig.hash.New)
return NewTarSumHash(r, disableCompression, version, tHash)
}
// TarSum is the generic interface for calculating fixed time
// checksums of a tar archive.
type TarSum interface {
io.Reader
GetSums() FileInfoSums
Sum([]byte) string
Version() Version
Hash() THash
}
// tarSum struct is the structure for a Version0 checksum calculation.
type tarSum struct {
io.Reader
tarR *tar.Reader
tarW *tar.Writer
writer writeCloseFlusher
bufTar *bytes.Buffer
bufWriter *bytes.Buffer
bufData []byte
h hash.Hash
tHash THash
sums FileInfoSums
fileCounter int64
currentFile string
finished bool
first bool
DisableCompression bool // false by default. When false, the output gzip compressed.
tarSumVersion Version // this field is not exported so it can not be mutated during use
headerSelector tarHeaderSelector // handles selecting and ordering headers for files in the archive
}
func (ts tarSum) Hash() THash {
return ts.tHash
}
func (ts tarSum) Version() Version {
return ts.tarSumVersion
}
// THash provides a hash.Hash type generator and its name.
type THash interface {
Hash() hash.Hash
Name() string
}
// NewTHash is a convenience method for creating a THash.
func NewTHash(name string, h func() hash.Hash) THash {
return simpleTHash{n: name, h: h}
}
type tHashConfig struct {
name string
hash crypto.Hash
}
var (
// NOTE: DO NOT include MD5 or SHA1, which are considered insecure.
standardHashConfigs = map[string]tHashConfig{
"sha256": {name: "sha256", hash: crypto.SHA256},
"sha512": {name: "sha512", hash: crypto.SHA512},
}
)
// DefaultTHash is default TarSum hashing algorithm - "sha256".
var DefaultTHash = NewTHash("sha256", sha256.New)
type simpleTHash struct {
n string
h func() hash.Hash
}
func (sth simpleTHash) Name() string { return sth.n }
func (sth simpleTHash) Hash() hash.Hash { return sth.h() }
func (ts *tarSum) encodeHeader(h *tar.Header) error {
for _, elem := range ts.headerSelector.selectHeaders(h) {
if _, err := ts.h.Write([]byte(elem[0] + elem[1])); err != nil {
return err
}
}
return nil
}
func (ts *tarSum) initTarSum() error {
ts.bufTar = bytes.NewBuffer([]byte{})
ts.bufWriter = bytes.NewBuffer([]byte{})
ts.tarR = tar.NewReader(ts.Reader)
ts.tarW = tar.NewWriter(ts.bufTar)
if !ts.DisableCompression {
ts.writer = gzip.NewWriter(ts.bufWriter)
} else {
ts.writer = &nopCloseFlusher{Writer: ts.bufWriter}
}
if ts.tHash == nil {
ts.tHash = DefaultTHash
}
ts.h = ts.tHash.Hash()
ts.h.Reset()
ts.first = true
ts.sums = FileInfoSums{}
return nil
}
func (ts *tarSum) Read(buf []byte) (int, error) {
if ts.finished {
return ts.bufWriter.Read(buf)
}
if len(ts.bufData) < len(buf) {
switch {
case len(buf) <= buf8K:
ts.bufData = make([]byte, buf8K)
case len(buf) <= buf16K:
ts.bufData = make([]byte, buf16K)
case len(buf) <= buf32K:
ts.bufData = make([]byte, buf32K)
default:
ts.bufData = make([]byte, len(buf))
}
}
buf2 := ts.bufData[:len(buf)]
n, err := ts.tarR.Read(buf2)
if err != nil {
if err == io.EOF {
if _, err := ts.h.Write(buf2[:n]); err != nil {
return 0, err
}
if !ts.first {
ts.sums = append(ts.sums, fileInfoSum{name: ts.currentFile, sum: hex.EncodeToString(ts.h.Sum(nil)), pos: ts.fileCounter})
ts.fileCounter++
ts.h.Reset()
} else {
ts.first = false
}
currentHeader, err := ts.tarR.Next()
if err != nil {
if err == io.EOF {
if err := ts.tarW.Close(); err != nil {
return 0, err
}
if _, err := io.Copy(ts.writer, ts.bufTar); err != nil {
return 0, err
}
if err := ts.writer.Close(); err != nil {
return 0, err
}
ts.finished = true
return n, nil
}
return n, err
}
ts.currentFile = path.Clean(currentHeader.Name)
if err := ts.encodeHeader(currentHeader); err != nil {
return 0, err
}
if err := ts.tarW.WriteHeader(currentHeader); err != nil {
return 0, err
}
if _, err := ts.tarW.Write(buf2[:n]); err != nil {
return 0, err
}
ts.tarW.Flush()
if _, err := io.Copy(ts.writer, ts.bufTar); err != nil {
return 0, err
}
ts.writer.Flush()
return ts.bufWriter.Read(buf)
}
return n, err
}
// Filling the hash buffer
if _, err = ts.h.Write(buf2[:n]); err != nil {
return 0, err
}
// Filling the tar writer
if _, err = ts.tarW.Write(buf2[:n]); err != nil {
return 0, err
}
ts.tarW.Flush()
// Filling the output writer
if _, err = io.Copy(ts.writer, ts.bufTar); err != nil {
return 0, err
}
ts.writer.Flush()
return ts.bufWriter.Read(buf)
}
func (ts *tarSum) Sum(extra []byte) string {
ts.sums.SortBySums()
h := ts.tHash.Hash()
if extra != nil {
h.Write(extra)
}
for _, fis := range ts.sums {
h.Write([]byte(fis.Sum()))
}
checksum := ts.Version().String() + "+" + ts.tHash.Name() + ":" + hex.EncodeToString(h.Sum(nil))
return checksum
}
func (ts *tarSum) GetSums() FileInfoSums {
return ts.sums
}