converter.go

/*
   Copyright The containerd Authors.

   Licensed under the Apache License, Version 2.0 (the "License");
   you may not use this file except in compliance with the License.
   You may obtain a copy of the License at

       http://www.apache.org/licenses/LICENSE-2.0

   Unless required by applicable law or agreed to in writing, software
   distributed under the License is distributed on an "AS IS" BASIS,
   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
   See the License for the specific language governing permissions and
   limitations under the License.
*/

package schema1

import (
	"bytes"
	"context"
	"encoding/base64"
	"encoding/json"
	"fmt"
	"io"
	"io/ioutil"
	"strconv"
	"strings"
	"sync"
	"time"

	"golang.org/x/sync/errgroup"

	"github.com/containerd/containerd/archive/compression"
	"github.com/containerd/containerd/content"
	"github.com/containerd/containerd/errdefs"
	"github.com/containerd/containerd/images"
	"github.com/containerd/containerd/log"
	"github.com/containerd/containerd/remotes"
	digest "github.com/opencontainers/go-digest"
	specs "github.com/opencontainers/image-spec/specs-go"
	ocispec "github.com/opencontainers/image-spec/specs-go/v1"
	"github.com/pkg/errors"
)

const (
	manifestSizeLimit            = 8e6 // 8MB
	labelDockerSchema1EmptyLayer = "containerd.io/docker.schema1.empty-layer"
)

type blobState struct {
	diffID digest.Digest
	empty  bool
}

// Converter converts schema1 manifests to schema2 on fetch
type Converter struct {
	contentStore content.Store
	fetcher      remotes.Fetcher

	pulledManifest *manifest

	mu         sync.Mutex
	blobMap    map[digest.Digest]blobState
	layerBlobs map[digest.Digest]ocispec.Descriptor
}

// NewConverter returns a new converter
func NewConverter(contentStore content.Store, fetcher remotes.Fetcher) *Converter {
	return &Converter{
		contentStore: contentStore,
		fetcher:      fetcher,
		blobMap:      map[digest.Digest]blobState{},
		layerBlobs:   map[digest.Digest]ocispec.Descriptor{},
	}
}

// Handle fetching descriptors for a docker media type
func (c *Converter) Handle(ctx context.Context, desc ocispec.Descriptor) ([]ocispec.Descriptor, error) {
	switch desc.MediaType {
	case images.MediaTypeDockerSchema1Manifest:
		if err := c.fetchManifest(ctx, desc); err != nil {
			return nil, err
		}

		m := c.pulledManifest
		if len(m.FSLayers) != len(m.History) {
			return nil, errors.New("invalid schema 1 manifest, history and layer mismatch")
		}
		descs := make([]ocispec.Descriptor, 0, len(c.pulledManifest.FSLayers))

		for i := range m.FSLayers {
			if _, ok := c.blobMap[c.pulledManifest.FSLayers[i].BlobSum]; !ok {
				empty, err := isEmptyLayer([]byte(m.History[i].V1Compatibility))
				if err != nil {
					return nil, err
				}

				// Do no attempt to download a known empty blob
				if !empty {
					descs = append([]ocispec.Descriptor{
						{
							MediaType: images.MediaTypeDockerSchema2LayerGzip,
							Digest:    c.pulledManifest.FSLayers[i].BlobSum,
							Size:      -1,
						},
					}, descs...)
				}
				c.blobMap[c.pulledManifest.FSLayers[i].BlobSum] = blobState{
					empty: empty,
				}
			}
		}
		return descs, nil
	case images.MediaTypeDockerSchema2LayerGzip:
		if c.pulledManifest == nil {
			return nil, errors.New("manifest required for schema 1 blob pull")
		}
		return nil, c.fetchBlob(ctx, desc)
	default:
		return nil, fmt.Errorf("%v not support for schema 1 manifests", desc.MediaType)
	}
}

// ConvertOptions provides options on converting a docker schema1 manifest.
type ConvertOptions struct {
	// ManifestMediaType specifies the media type of the manifest OCI descriptor.
	ManifestMediaType string

	// ConfigMediaType specifies the media type of the manifest config OCI
	// descriptor.
	ConfigMediaType string
}

// ConvertOpt allows configuring a convert operation.
type ConvertOpt func(context.Context, *ConvertOptions) error

// UseDockerSchema2 is used to indicate that a schema1 manifest should be
// converted into the media types for a docker schema2 manifest.
func UseDockerSchema2() ConvertOpt {
	return func(ctx context.Context, o *ConvertOptions) error {
		o.ManifestMediaType = images.MediaTypeDockerSchema2Manifest
		o.ConfigMediaType = images.MediaTypeDockerSchema2Config
		return nil
	}
}

// Convert a docker manifest to an OCI descriptor
func (c *Converter) Convert(ctx context.Context, opts ...ConvertOpt) (ocispec.Descriptor, error) {
	co := ConvertOptions{
		ManifestMediaType: ocispec.MediaTypeImageManifest,
		ConfigMediaType:   ocispec.MediaTypeImageConfig,
	}
	for _, opt := range opts {
		if err := opt(ctx, &co); err != nil {
			return ocispec.Descriptor{}, err
		}
	}

	history, diffIDs, err := c.schema1ManifestHistory()
	if err != nil {
		return ocispec.Descriptor{}, errors.Wrap(err, "schema 1 conversion failed")
	}

	var img ocispec.Image
	if err := json.Unmarshal([]byte(c.pulledManifest.History[0].V1Compatibility), &img); err != nil {
		return ocispec.Descriptor{}, errors.Wrap(err, "failed to unmarshal image from schema 1 history")
	}

	img.History = history
	img.RootFS = ocispec.RootFS{
		Type:    "layers",
		DiffIDs: diffIDs,
	}

	b, err := json.MarshalIndent(img, "", "   ")
	if err != nil {
		return ocispec.Descriptor{}, errors.Wrap(err, "failed to marshal image")
	}

	config := ocispec.Descriptor{
		MediaType: co.ConfigMediaType,
		Digest:    digest.Canonical.FromBytes(b),
		Size:      int64(len(b)),
	}

	layers := make([]ocispec.Descriptor, len(diffIDs))
	for i, diffID := range diffIDs {
		layers[i] = c.layerBlobs[diffID]
	}

	manifest := ocispec.Manifest{
		Versioned: specs.Versioned{
			SchemaVersion: 2,
		},
		Config: config,
		Layers: layers,
	}

	mb, err := json.MarshalIndent(manifest, "", "   ")
	if err != nil {
		return ocispec.Descriptor{}, errors.Wrap(err, "failed to marshal image")
	}

	desc := ocispec.Descriptor{
		MediaType: co.ManifestMediaType,
		Digest:    digest.Canonical.FromBytes(mb),
		Size:      int64(len(mb)),
	}

	labels := map[string]string{}
	labels["containerd.io/gc.ref.content.0"] = manifest.Config.Digest.String()
	for i, ch := range manifest.Layers {
		labels[fmt.Sprintf("containerd.io/gc.ref.content.%d", i+1)] = ch.Digest.String()
	}

	ref := remotes.MakeRefKey(ctx, desc)
	if err := content.WriteBlob(ctx, c.contentStore, ref, bytes.NewReader(mb), desc, content.WithLabels(labels)); err != nil {
		return ocispec.Descriptor{}, errors.Wrap(err, "failed to write image manifest")
	}

	ref = remotes.MakeRefKey(ctx, config)
	if err := content.WriteBlob(ctx, c.contentStore, ref, bytes.NewReader(b), config); err != nil {
		return ocispec.Descriptor{}, errors.Wrap(err, "failed to write image config")
	}

	return desc, nil
}

// ReadStripSignature reads in a schema1 manifest and returns a byte array
// with the "signatures" field stripped
func ReadStripSignature(schema1Blob io.Reader) ([]byte, error) {
	b, err := ioutil.ReadAll(io.LimitReader(schema1Blob, manifestSizeLimit)) // limit to 8MB
	if err != nil {
		return nil, err
	}

	return stripSignature(b)
}

func (c *Converter) fetchManifest(ctx context.Context, desc ocispec.Descriptor) error {
	log.G(ctx).Debug("fetch schema 1")

	rc, err := c.fetcher.Fetch(ctx, desc)
	if err != nil {
		return err
	}

	b, err := ReadStripSignature(rc)
	rc.Close()
	if err != nil {
		return err
	}

	var m manifest
	if err := json.Unmarshal(b, &m); err != nil {
		return err
	}
	c.pulledManifest = &m

	return nil
}

func (c *Converter) fetchBlob(ctx context.Context, desc ocispec.Descriptor) error {
	log.G(ctx).Debug("fetch blob")

	var (
		ref            = remotes.MakeRefKey(ctx, desc)
		calc           = newBlobStateCalculator()
		compressMethod = compression.Gzip
	)

	// size may be unknown, set to zero for content ingest
	ingestDesc := desc
	if ingestDesc.Size == -1 {
		ingestDesc.Size = 0
	}

	cw, err := content.OpenWriter(ctx, c.contentStore, content.WithRef(ref), content.WithDescriptor(ingestDesc))
	if err != nil {
		if !errdefs.IsAlreadyExists(err) {
			return err
		}

		reuse, err := c.reuseLabelBlobState(ctx, desc)
		if err != nil {
			return err
		}

		if reuse {
			return nil
		}

		ra, err := c.contentStore.ReaderAt(ctx, desc)
		if err != nil {
			return err
		}
		defer ra.Close()

		r, err := compression.DecompressStream(content.NewReader(ra))
		if err != nil {
			return err
		}

		compressMethod = r.GetCompression()
		_, err = io.Copy(calc, r)
		r.Close()
		if err != nil {
			return err
		}
	} else {
		defer cw.Close()

		rc, err := c.fetcher.Fetch(ctx, desc)
		if err != nil {
			return err
		}
		defer rc.Close()

		eg, _ := errgroup.WithContext(ctx)
		pr, pw := io.Pipe()

		eg.Go(func() error {
			r, err := compression.DecompressStream(pr)
			if err != nil {
				return err
			}

			compressMethod = r.GetCompression()
			_, err = io.Copy(calc, r)
			r.Close()
			pr.CloseWithError(err)
			return err
		})

		eg.Go(func() error {
			defer pw.Close()

			return content.Copy(ctx, cw, io.TeeReader(rc, pw), ingestDesc.Size, ingestDesc.Digest)
		})

		if err := eg.Wait(); err != nil {
			return err
		}
	}

	if desc.Size == -1 {
		info, err := c.contentStore.Info(ctx, desc.Digest)
		if err != nil {
			return errors.Wrap(err, "failed to get blob info")
		}
		desc.Size = info.Size
	}

	if compressMethod == compression.Uncompressed {
		log.G(ctx).WithField("id", desc.Digest).Debugf("changed media type for uncompressed schema1 layer blob")
		desc.MediaType = images.MediaTypeDockerSchema2Layer
	}

	state := calc.State()

	cinfo := content.Info{
		Digest: desc.Digest,
		Labels: map[string]string{
			"containerd.io/uncompressed": state.diffID.String(),
			labelDockerSchema1EmptyLayer: strconv.FormatBool(state.empty),
		},
	}

	if _, err := c.contentStore.Update(ctx, cinfo, "labels.containerd.io/uncompressed", fmt.Sprintf("labels.%s", labelDockerSchema1EmptyLayer)); err != nil {
		return errors.Wrap(err, "failed to update uncompressed label")
	}

	c.mu.Lock()
	c.blobMap[desc.Digest] = state
	c.layerBlobs[state.diffID] = desc
	c.mu.Unlock()

	return nil
}

func (c *Converter) reuseLabelBlobState(ctx context.Context, desc ocispec.Descriptor) (bool, error) {
	cinfo, err := c.contentStore.Info(ctx, desc.Digest)
	if err != nil {
		return false, errors.Wrap(err, "failed to get blob info")
	}
	desc.Size = cinfo.Size

	diffID, ok := cinfo.Labels["containerd.io/uncompressed"]
	if !ok {
		return false, nil
	}

	emptyVal, ok := cinfo.Labels[labelDockerSchema1EmptyLayer]
	if !ok {
		return false, nil
	}

	isEmpty, err := strconv.ParseBool(emptyVal)
	if err != nil {
		log.G(ctx).WithField("id", desc.Digest).Warnf("failed to parse bool from label %s: %v", labelDockerSchema1EmptyLayer, isEmpty)
		return false, nil
	}

	bState := blobState{empty: isEmpty}

	if bState.diffID, err = digest.Parse(diffID); err != nil {
		log.G(ctx).WithField("id", desc.Digest).Warnf("failed to parse digest from label containerd.io/uncompressed: %v", diffID)
		return false, nil
	}

	// NOTE: there is no need to read header to get compression method
	// because there are only two kinds of methods.
	if bState.diffID == desc.Digest {
		desc.MediaType = images.MediaTypeDockerSchema2Layer
	} else {
		desc.MediaType = images.MediaTypeDockerSchema2LayerGzip
	}

	c.mu.Lock()
	c.blobMap[desc.Digest] = bState
	c.layerBlobs[bState.diffID] = desc
	c.mu.Unlock()
	return true, nil
}

func (c *Converter) schema1ManifestHistory() ([]ocispec.History, []digest.Digest, error) {
	if c.pulledManifest == nil {
		return nil, nil, errors.New("missing schema 1 manifest for conversion")
	}
	m := *c.pulledManifest

	if len(m.History) == 0 {
		return nil, nil, errors.New("no history")
	}

	history := make([]ocispec.History, len(m.History))
	diffIDs := []digest.Digest{}
	for i := range m.History {
		var h v1History
		if err := json.Unmarshal([]byte(m.History[i].V1Compatibility), &h); err != nil {
			return nil, nil, errors.Wrap(err, "failed to unmarshal history")
		}

		blobSum := m.FSLayers[i].BlobSum

		state := c.blobMap[blobSum]

		history[len(history)-i-1] = ocispec.History{
			Author:     h.Author,
			Comment:    h.Comment,
			Created:    &h.Created,
			CreatedBy:  strings.Join(h.ContainerConfig.Cmd, " "),
			EmptyLayer: state.empty,
		}

		if !state.empty {
			diffIDs = append([]digest.Digest{state.diffID}, diffIDs...)

		}
	}

	return history, diffIDs, nil
}

type fsLayer struct {
	BlobSum digest.Digest `json:"blobSum"`
}

type history struct {
	V1Compatibility string `json:"v1Compatibility"`
}

type manifest struct {
	FSLayers []fsLayer `json:"fsLayers"`
	History  []history `json:"history"`
}

type v1History struct {
	Author          string    `json:"author,omitempty"`
	Created         time.Time `json:"created"`
	Comment         string    `json:"comment,omitempty"`
	ThrowAway       *bool     `json:"throwaway,omitempty"`
	Size            *int      `json:"Size,omitempty"` // used before ThrowAway field
	ContainerConfig struct {
		Cmd []string `json:"Cmd,omitempty"`
	} `json:"container_config,omitempty"`
}

// isEmptyLayer returns whether the v1 compatibility history describes an
// empty layer. A return value of true indicates the layer is empty,
// however false does not indicate non-empty.
func isEmptyLayer(compatHistory []byte) (bool, error) {
	var h v1History
	if err := json.Unmarshal(compatHistory, &h); err != nil {
		return false, err
	}

	if h.ThrowAway != nil {
		return *h.ThrowAway, nil
	}
	if h.Size != nil {
		return *h.Size == 0, nil
	}

	// If no `Size` or `throwaway` field is given, then
	// it cannot be determined whether the layer is empty
	// from the history, return false
	return false, nil
}

type signature struct {
	Signatures []jsParsedSignature `json:"signatures"`
}

type jsParsedSignature struct {
	Protected string `json:"protected"`
}

type protectedBlock struct {
	Length int    `json:"formatLength"`
	Tail   string `json:"formatTail"`
}

// joseBase64UrlDecode decodes the given string using the standard base64 url
// decoder but first adds the appropriate number of trailing '=' characters in
// accordance with the jose specification.
// http://tools.ietf.org/html/draft-ietf-jose-json-web-signature-31#section-2
func joseBase64UrlDecode(s string) ([]byte, error) {
	switch len(s) % 4 {
	case 0:
	case 2:
		s += "=="
	case 3:
		s += "="
	default:
		return nil, errors.New("illegal base64url string")
	}
	return base64.URLEncoding.DecodeString(s)
}

func stripSignature(b []byte) ([]byte, error) {
	var sig signature
	if err := json.Unmarshal(b, &sig); err != nil {
		return nil, err
	}
	if len(sig.Signatures) == 0 {
		return nil, errors.New("no signatures")
	}
	pb, err := joseBase64UrlDecode(sig.Signatures[0].Protected)
	if err != nil {
		return nil, errors.Wrapf(err, "could not decode %s", sig.Signatures[0].Protected)
	}

	var protected protectedBlock
	if err := json.Unmarshal(pb, &protected); err != nil {
		return nil, err
	}

	if protected.Length > len(b) {
		return nil, errors.New("invalid protected length block")
	}

	tail, err := joseBase64UrlDecode(protected.Tail)
	if err != nil {
		return nil, errors.Wrap(err, "invalid tail base 64 value")
	}

	return append(b[:protected.Length], tail...), nil
}

type blobStateCalculator struct {
	empty    bool
	digester digest.Digester
}

func newBlobStateCalculator() *blobStateCalculator {
	return &blobStateCalculator{
		empty:    true,
		digester: digest.Canonical.Digester(),
	}
}

func (c *blobStateCalculator) Write(p []byte) (int, error) {
	if c.empty {
		for _, b := range p {
			if b != 0x00 {
				c.empty = false
				break
			}
		}
	}
	return c.digester.Hash().Write(p)
}

func (c *blobStateCalculator) State() blobState {
	return blobState{
		empty:  c.empty,
		diffID: c.digester.Digest(),
	}
}