containerd/metadata/content.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 metadata

import (
	"context"
	"encoding/binary"
	"strings"
	"sync"
	"sync/atomic"
	"time"

	"github.com/containerd/containerd/content"
	"github.com/containerd/containerd/errdefs"
	"github.com/containerd/containerd/filters"
	"github.com/containerd/containerd/labels"
	"github.com/containerd/containerd/log"
	"github.com/containerd/containerd/metadata/boltutil"
	"github.com/containerd/containerd/namespaces"
	digest "github.com/opencontainers/go-digest"
	ocispec "github.com/opencontainers/image-spec/specs-go/v1"
	"github.com/pkg/errors"
	bolt "go.etcd.io/bbolt"
)

type contentStore struct {
	content.Store
	db     *DB
	shared bool
	l      sync.RWMutex
}

// newContentStore returns a namespaced content store using an existing
// content store interface.
// policy defines the sharing behavior for content between namespaces. Both
// modes will result in shared storage in the backend for committed. Choose
// "shared" to prevent separate namespaces from having to pull the same content
// twice.  Choose "isolated" if the content must not be shared between
// namespaces.
//
// If the policy is "shared", writes will try to resolve the "expected" digest
// against the backend, allowing imports of content from other namespaces. In
// "isolated" mode, the client must prove they have the content by providing
// the entire blob before the content can be added to another namespace.
//
// Since we have only two policies right now, it's simpler using bool to
// represent it internally.
func newContentStore(db *DB, shared bool, cs content.Store) *contentStore {
	return &contentStore{
		Store:  cs,
		db:     db,
		shared: shared,
	}
}

func (cs *contentStore) Info(ctx context.Context, dgst digest.Digest) (content.Info, error) {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return content.Info{}, err
	}

	var info content.Info
	if err := view(ctx, cs.db, func(tx *bolt.Tx) error {
		bkt := getBlobBucket(tx, ns, dgst)
		if bkt == nil {
			return errors.Wrapf(errdefs.ErrNotFound, "content digest %v", dgst)
		}

		info.Digest = dgst
		return readInfo(&info, bkt)
	}); err != nil {
		return content.Info{}, err
	}

	return info, nil
}

func (cs *contentStore) Update(ctx context.Context, info content.Info, fieldpaths ...string) (content.Info, error) {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return content.Info{}, err
	}

	cs.l.RLock()
	defer cs.l.RUnlock()

	updated := content.Info{
		Digest: info.Digest,
	}
	if err := update(ctx, cs.db, func(tx *bolt.Tx) error {
		bkt := getBlobBucket(tx, ns, info.Digest)
		if bkt == nil {
			return errors.Wrapf(errdefs.ErrNotFound, "content digest %v", info.Digest)
		}

		if err := readInfo(&updated, bkt); err != nil {
			return errors.Wrapf(err, "info %q", info.Digest)
		}

		if len(fieldpaths) > 0 {
			for _, path := range fieldpaths {
				if strings.HasPrefix(path, "labels.") {
					if updated.Labels == nil {
						updated.Labels = map[string]string{}
					}

					key := strings.TrimPrefix(path, "labels.")
					updated.Labels[key] = info.Labels[key]
					continue
				}

				switch path {
				case "labels":
					updated.Labels = info.Labels
				default:
					return errors.Wrapf(errdefs.ErrInvalidArgument, "cannot update %q field on content info %q", path, info.Digest)
				}
			}
		} else {
			// Set mutable fields
			updated.Labels = info.Labels
		}
		if err := validateInfo(&updated); err != nil {
			return err
		}

		updated.UpdatedAt = time.Now().UTC()
		return writeInfo(&updated, bkt)
	}); err != nil {
		return content.Info{}, err
	}
	return updated, nil
}

func (cs *contentStore) Walk(ctx context.Context, fn content.WalkFunc, fs ...string) error {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return err
	}

	filter, err := filters.ParseAll(fs...)
	if err != nil {
		return err
	}

	// TODO: Batch results to keep from reading all info into memory
	var infos []content.Info
	if err := view(ctx, cs.db, func(tx *bolt.Tx) error {
		bkt := getBlobsBucket(tx, ns)
		if bkt == nil {
			return nil
		}

		return bkt.ForEach(func(k, v []byte) error {
			dgst, err := digest.Parse(string(k))
			if err != nil {
				// Not a digest, skip
				return nil
			}
			bbkt := bkt.Bucket(k)
			if bbkt == nil {
				return nil
			}
			info := content.Info{
				Digest: dgst,
			}
			if err := readInfo(&info, bkt.Bucket(k)); err != nil {
				return err
			}
			if filter.Match(content.AdaptInfo(info)) {
				infos = append(infos, info)
			}
			return nil
		})
	}); err != nil {
		return err
	}

	for _, info := range infos {
		if err := fn(info); err != nil {
			return err
		}
	}

	return nil
}

func (cs *contentStore) Delete(ctx context.Context, dgst digest.Digest) error {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return err
	}

	cs.l.RLock()
	defer cs.l.RUnlock()

	return update(ctx, cs.db, func(tx *bolt.Tx) error {
		bkt := getBlobBucket(tx, ns, dgst)
		if bkt == nil {
			return errors.Wrapf(errdefs.ErrNotFound, "content digest %v", dgst)
		}

		if err := getBlobsBucket(tx, ns).DeleteBucket([]byte(dgst.String())); err != nil {
			return err
		}
		if err := removeContentLease(ctx, tx, dgst); err != nil {
			return err
		}

		// Mark content store as dirty for triggering garbage collection
		atomic.AddUint32(&cs.db.dirty, 1)
		cs.db.dirtyCS = true

		return nil
	})
}

func (cs *contentStore) ListStatuses(ctx context.Context, fs ...string) ([]content.Status, error) {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return nil, err
	}

	filter, err := filters.ParseAll(fs...)
	if err != nil {
		return nil, err
	}

	brefs := map[string]string{}
	if err := view(ctx, cs.db, func(tx *bolt.Tx) error {
		bkt := getIngestsBucket(tx, ns)
		if bkt == nil {
			return nil
		}

		return bkt.ForEach(func(k, v []byte) error {
			if v == nil {
				// TODO(dmcgowan): match name and potentially labels here
				brefs[string(k)] = string(bkt.Bucket(k).Get(bucketKeyRef))
			}
			return nil
		})
	}); err != nil {
		return nil, err
	}

	statuses := make([]content.Status, 0, len(brefs))
	for k, bref := range brefs {
		status, err := cs.Store.Status(ctx, bref)
		if err != nil {
			if errdefs.IsNotFound(err) {
				continue
			}
			return nil, err
		}
		status.Ref = k

		if filter.Match(adaptContentStatus(status)) {
			statuses = append(statuses, status)
		}
	}

	return statuses, nil

}

func getRef(tx *bolt.Tx, ns, ref string) string {
	bkt := getIngestBucket(tx, ns, ref)
	if bkt == nil {
		return ""
	}
	v := bkt.Get(bucketKeyRef)
	if len(v) == 0 {
		return ""
	}
	return string(v)
}

func (cs *contentStore) Status(ctx context.Context, ref string) (content.Status, error) {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return content.Status{}, err
	}

	var bref string
	if err := view(ctx, cs.db, func(tx *bolt.Tx) error {
		bref = getRef(tx, ns, ref)
		if bref == "" {
			return errors.Wrapf(errdefs.ErrNotFound, "reference %v", ref)
		}

		return nil
	}); err != nil {
		return content.Status{}, err
	}

	st, err := cs.Store.Status(ctx, bref)
	if err != nil {
		return content.Status{}, err
	}
	st.Ref = ref
	return st, nil
}

func (cs *contentStore) Abort(ctx context.Context, ref string) error {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return err
	}

	cs.l.RLock()
	defer cs.l.RUnlock()

	return update(ctx, cs.db, func(tx *bolt.Tx) error {
		ibkt := getIngestsBucket(tx, ns)
		if ibkt == nil {
			return errors.Wrapf(errdefs.ErrNotFound, "reference %v", ref)
		}
		bkt := ibkt.Bucket([]byte(ref))
		if bkt == nil {
			return errors.Wrapf(errdefs.ErrNotFound, "reference %v", ref)
		}
		bref := string(bkt.Get(bucketKeyRef))
		if bref == "" {
			return errors.Wrapf(errdefs.ErrNotFound, "reference %v", ref)
		}
		expected := string(bkt.Get(bucketKeyExpected))
		if err := ibkt.DeleteBucket([]byte(ref)); err != nil {
			return err
		}

		if err := removeIngestLease(ctx, tx, ref); err != nil {
			return err
		}

		// if not shared content, delete active ingest on backend
		if expected == "" {
			return cs.Store.Abort(ctx, bref)
		}

		return nil
	})

}

func (cs *contentStore) Writer(ctx context.Context, opts ...content.WriterOpt) (content.Writer, error) {
	var wOpts content.WriterOpts
	for _, opt := range opts {
		if err := opt(&wOpts); err != nil {
			return nil, err
		}
	}
	// TODO(AkihiroSuda): we could create a random string or one calculated based on the context
	// https://github.com/containerd/containerd/issues/2129#issuecomment-380255019
	if wOpts.Ref == "" {
		return nil, errors.Wrap(errdefs.ErrInvalidArgument, "ref must not be empty")
	}
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return nil, err
	}

	cs.l.RLock()
	defer cs.l.RUnlock()

	var (
		w      content.Writer
		exists bool
		bref   string
	)
	if err := update(ctx, cs.db, func(tx *bolt.Tx) error {
		var shared bool
		if wOpts.Desc.Digest != "" {
			cbkt := getBlobBucket(tx, ns, wOpts.Desc.Digest)
			if cbkt != nil {
				// Add content to lease to prevent other reference removals
				// from effecting this object during a provided lease
				if err := addContentLease(ctx, tx, wOpts.Desc.Digest); err != nil {
					return errors.Wrap(err, "unable to lease content")
				}
				// Return error outside of transaction to ensure
				// commit succeeds with the lease.
				exists = true
				return nil
			}

			if cs.shared {
				if st, err := cs.Store.Info(ctx, wOpts.Desc.Digest); err == nil {
					// Ensure the expected size is the same, it is likely
					// an error if the size is mismatched but the caller
					// must resolve this on commit
					if wOpts.Desc.Size == 0 || wOpts.Desc.Size == st.Size {
						shared = true
						wOpts.Desc.Size = st.Size
					}
				}
			}
		}

		bkt, err := createIngestBucket(tx, ns, wOpts.Ref)
		if err != nil {
			return err
		}

		leased, err := addIngestLease(ctx, tx, wOpts.Ref)
		if err != nil {
			return err
		}

		brefb := bkt.Get(bucketKeyRef)
		if brefb == nil {
			sid, err := bkt.NextSequence()
			if err != nil {
				return err
			}

			bref = createKey(sid, ns, wOpts.Ref)
			if err := bkt.Put(bucketKeyRef, []byte(bref)); err != nil {
				return err
			}
		} else {
			bref = string(brefb)
		}
		if !leased {
			// Add timestamp to allow aborting once stale
			// When lease is set the ingest should be aborted
			// after lease it belonged to is deleted.
			// Expiration can be configurable in the future to
			// give more control to the daemon, however leases
			// already give users more control of expiration.
			expireAt := time.Now().UTC().Add(24 * time.Hour)
			if err := writeExpireAt(expireAt, bkt); err != nil {
				return err
			}
		}

		if shared {
			if err := bkt.Put(bucketKeyExpected, []byte(wOpts.Desc.Digest)); err != nil {
				return err
			}
		} else {
			// Do not use the passed in expected value here since it was
			// already checked against the user metadata. The content must
			// be committed in the namespace before it will be seen as
			// available in the current namespace.
			desc := wOpts.Desc
			desc.Digest = ""
			w, err = cs.Store.Writer(ctx, content.WithRef(bref), content.WithDescriptor(desc))
		}
		return err
	}); err != nil {
		return nil, err
	}
	if exists {
		return nil, errors.Wrapf(errdefs.ErrAlreadyExists, "content %v", wOpts.Desc.Digest)
	}

	return &namespacedWriter{
		ctx:       ctx,
		ref:       wOpts.Ref,
		namespace: ns,
		db:        cs.db,
		provider:  cs.Store,
		l:         &cs.l,
		w:         w,
		bref:      bref,
		started:   time.Now(),
		desc:      wOpts.Desc,
	}, nil
}

type namespacedWriter struct {
	ctx       context.Context
	ref       string
	namespace string
	db        transactor
	provider  interface {
		content.Provider
		content.Ingester
	}
	l *sync.RWMutex

	w content.Writer

	bref    string
	started time.Time
	desc    ocispec.Descriptor
}

func (nw *namespacedWriter) Close() error {
	if nw.w != nil {
		return nw.w.Close()
	}
	return nil
}

func (nw *namespacedWriter) Write(p []byte) (int, error) {
	// if no writer, first copy and unshare before performing write
	if nw.w == nil {
		if len(p) == 0 {
			return 0, nil
		}

		if err := nw.createAndCopy(nw.ctx, nw.desc); err != nil {
			return 0, err
		}
	}

	return nw.w.Write(p)
}

func (nw *namespacedWriter) Digest() digest.Digest {
	if nw.w != nil {
		return nw.w.Digest()
	}
	return nw.desc.Digest
}

func (nw *namespacedWriter) Truncate(size int64) error {
	if nw.w != nil {
		return nw.w.Truncate(size)
	}
	desc := nw.desc
	desc.Size = size
	desc.Digest = ""
	return nw.createAndCopy(nw.ctx, desc)
}

func (nw *namespacedWriter) createAndCopy(ctx context.Context, desc ocispec.Descriptor) error {
	nwDescWithoutDigest := desc
	nwDescWithoutDigest.Digest = ""
	w, err := nw.provider.Writer(ctx, content.WithRef(nw.bref), content.WithDescriptor(nwDescWithoutDigest))
	if err != nil {
		return err
	}

	if desc.Size > 0 {
		ra, err := nw.provider.ReaderAt(ctx, nw.desc)
		if err != nil {
			return err
		}
		defer ra.Close()

		if err := content.CopyReaderAt(w, ra, desc.Size); err != nil {
			nw.w.Close()
			nw.w = nil
			return err
		}
	}
	nw.w = w

	return nil
}

func (nw *namespacedWriter) Commit(ctx context.Context, size int64, expected digest.Digest, opts ...content.Opt) error {
	ctx = namespaces.WithNamespace(ctx, nw.namespace)

	nw.l.RLock()
	defer nw.l.RUnlock()

	var innerErr error

	if err := update(ctx, nw.db, func(tx *bolt.Tx) error {
		dgst, err := nw.commit(ctx, tx, size, expected, opts...)
		if err != nil {
			if !errdefs.IsAlreadyExists(err) {
				return err
			}
			innerErr = err
		}
		bkt := getIngestsBucket(tx, nw.namespace)
		if bkt != nil {
			if err := bkt.DeleteBucket([]byte(nw.ref)); err != nil && err != bolt.ErrBucketNotFound {
				return err
			}
		}
		if err := removeIngestLease(ctx, tx, nw.ref); err != nil {
			return err
		}
		return addContentLease(ctx, tx, dgst)
	}); err != nil {
		return err
	}

	return innerErr
}

func (nw *namespacedWriter) commit(ctx context.Context, tx *bolt.Tx, size int64, expected digest.Digest, opts ...content.Opt) (digest.Digest, error) {
	var base content.Info
	for _, opt := range opts {
		if err := opt(&base); err != nil {
			if nw.w != nil {
				nw.w.Close()
			}
			return "", err
		}
	}
	if err := validateInfo(&base); err != nil {
		if nw.w != nil {
			nw.w.Close()
		}
		return "", err
	}

	var actual digest.Digest
	if nw.w == nil {
		if size != 0 && size != nw.desc.Size {
			return "", errors.Wrapf(errdefs.ErrFailedPrecondition, "%q failed size validation: %v != %v", nw.ref, nw.desc.Size, size)
		}
		if expected != "" && expected != nw.desc.Digest {
			return "", errors.Wrapf(errdefs.ErrFailedPrecondition, "%q unexpected digest", nw.ref)
		}
		size = nw.desc.Size
		actual = nw.desc.Digest
	} else {
		status, err := nw.w.Status()
		if err != nil {
			nw.w.Close()
			return "", err
		}
		if size != 0 && size != status.Offset {
			nw.w.Close()
			return "", errors.Wrapf(errdefs.ErrFailedPrecondition, "%q failed size validation: %v != %v", nw.ref, status.Offset, size)
		}
		size = status.Offset

		if err := nw.w.Commit(ctx, size, expected); err != nil && !errdefs.IsAlreadyExists(err) {
			return "", err
		}
		actual = nw.w.Digest()
	}

	bkt, err := createBlobBucket(tx, nw.namespace, actual)
	if err != nil {
		if err == bolt.ErrBucketExists {
			return actual, errors.Wrapf(errdefs.ErrAlreadyExists, "content %v", actual)
		}
		return "", err
	}

	commitTime := time.Now().UTC()

	sizeEncoded, err := encodeInt(size)
	if err != nil {
		return "", err
	}

	if err := boltutil.WriteTimestamps(bkt, commitTime, commitTime); err != nil {
		return "", err
	}
	if err := boltutil.WriteLabels(bkt, base.Labels); err != nil {
		return "", err
	}
	return actual, bkt.Put(bucketKeySize, sizeEncoded)
}

func (nw *namespacedWriter) Status() (st content.Status, err error) {
	if nw.w != nil {
		st, err = nw.w.Status()
	} else {
		st.Offset = nw.desc.Size
		st.Total = nw.desc.Size
		st.StartedAt = nw.started
		st.UpdatedAt = nw.started
		st.Expected = nw.desc.Digest
	}
	if err == nil {
		st.Ref = nw.ref
	}
	return
}

func (cs *contentStore) ReaderAt(ctx context.Context, desc ocispec.Descriptor) (content.ReaderAt, error) {
	if err := cs.checkAccess(ctx, desc.Digest); err != nil {
		return nil, err
	}
	return cs.Store.ReaderAt(ctx, desc)
}

func (cs *contentStore) checkAccess(ctx context.Context, dgst digest.Digest) error {
	ns, err := namespaces.NamespaceRequired(ctx)
	if err != nil {
		return err
	}

	return view(ctx, cs.db, func(tx *bolt.Tx) error {
		bkt := getBlobBucket(tx, ns, dgst)
		if bkt == nil {
			return errors.Wrapf(errdefs.ErrNotFound, "content digest %v", dgst)
		}
		return nil
	})
}

func validateInfo(info *content.Info) error {
	for k, v := range info.Labels {
		if err := labels.Validate(k, v); err == nil {
			return errors.Wrapf(err, "info.Labels")
		}
	}

	return nil
}

func readInfo(info *content.Info, bkt *bolt.Bucket) error {
	if err := boltutil.ReadTimestamps(bkt, &info.CreatedAt, &info.UpdatedAt); err != nil {
		return err
	}

	labels, err := boltutil.ReadLabels(bkt)
	if err != nil {
		return err
	}
	info.Labels = labels

	if v := bkt.Get(bucketKeySize); len(v) > 0 {
		info.Size, _ = binary.Varint(v)
	}

	return nil
}

func writeInfo(info *content.Info, bkt *bolt.Bucket) error {
	if err := boltutil.WriteTimestamps(bkt, info.CreatedAt, info.UpdatedAt); err != nil {
		return err
	}

	if err := boltutil.WriteLabels(bkt, info.Labels); err != nil {
		return errors.Wrapf(err, "writing labels for info %v", info.Digest)
	}

	// Write size
	sizeEncoded, err := encodeInt(info.Size)
	if err != nil {
		return err
	}

	return bkt.Put(bucketKeySize, sizeEncoded)
}

func readExpireAt(bkt *bolt.Bucket) (*time.Time, error) {
	v := bkt.Get(bucketKeyExpireAt)
	if v == nil {
		return nil, nil
	}
	t := &time.Time{}
	if err := t.UnmarshalBinary(v); err != nil {
		return nil, err
	}
	return t, nil
}

func writeExpireAt(expire time.Time, bkt *bolt.Bucket) error {
	expireAt, err := expire.MarshalBinary()
	if err != nil {
		return err
	}
	return bkt.Put(bucketKeyExpireAt, expireAt)
}

func (cs *contentStore) garbageCollect(ctx context.Context) (d time.Duration, err error) {
	cs.l.Lock()
	t1 := time.Now()
	defer func() {
		if err == nil {
			d = time.Since(t1)
		}
		cs.l.Unlock()
	}()

	contentSeen := map[string]struct{}{}
	ingestSeen := map[string]struct{}{}
	if err := cs.db.View(func(tx *bolt.Tx) error {
		v1bkt := tx.Bucket(bucketKeyVersion)
		if v1bkt == nil {
			return nil
		}

		// iterate through each namespace
		v1c := v1bkt.Cursor()

		for k, v := v1c.First(); k != nil; k, v = v1c.Next() {
			if v != nil {
				continue
			}

			cbkt := v1bkt.Bucket(k).Bucket(bucketKeyObjectContent)
			if cbkt == nil {
				continue
			}
			bbkt := cbkt.Bucket(bucketKeyObjectBlob)
			if bbkt != nil {
				if err := bbkt.ForEach(func(ck, cv []byte) error {
					if cv == nil {
						contentSeen[string(ck)] = struct{}{}
					}
					return nil
				}); err != nil {
					return err
				}
			}

			ibkt := cbkt.Bucket(bucketKeyObjectIngests)
			if ibkt != nil {
				if err := ibkt.ForEach(func(ref, v []byte) error {
					if v == nil {
						bkt := ibkt.Bucket(ref)
						// expected here may be from a different namespace
						// so much be explicitly retained from the ingest
						// in case it was removed from the other namespace
						expected := bkt.Get(bucketKeyExpected)
						if len(expected) > 0 {
							contentSeen[string(expected)] = struct{}{}
						}
						bref := bkt.Get(bucketKeyRef)
						if len(bref) > 0 {
							ingestSeen[string(bref)] = struct{}{}
						}
					}
					return nil
				}); err != nil {
					return err
				}
			}
		}

		return nil
	}); err != nil {
		return 0, err
	}

	err = cs.Store.Walk(ctx, func(info content.Info) error {
		if _, ok := contentSeen[info.Digest.String()]; !ok {
			if err := cs.Store.Delete(ctx, info.Digest); err != nil {
				return err
			}
			log.G(ctx).WithField("digest", info.Digest).Debug("removed content")
		}
		return nil
	})
	if err != nil {
		return
	}

	// If the content store has implemented a more efficient walk function
	// then use that else fallback to reading all statuses which may
	// cause reading of unneeded metadata.
	type statusWalker interface {
		WalkStatusRefs(context.Context, func(string) error) error
	}
	if w, ok := cs.Store.(statusWalker); ok {
		err = w.WalkStatusRefs(ctx, func(ref string) error {
			if _, ok := ingestSeen[ref]; !ok {
				if err := cs.Store.Abort(ctx, ref); err != nil {
					return err
				}
				log.G(ctx).WithField("ref", ref).Debug("cleanup aborting ingest")
			}
			return nil
		})
	} else {
		var statuses []content.Status
		statuses, err = cs.Store.ListStatuses(ctx)
		if err != nil {
			return 0, err
		}
		for _, status := range statuses {
			if _, ok := ingestSeen[status.Ref]; !ok {
				if err = cs.Store.Abort(ctx, status.Ref); err != nil {
					return
				}
				log.G(ctx).WithField("ref", status.Ref).Debug("cleanup aborting ingest")
			}
		}
	}
	return
}