Initial checkin from CRI-O repo

Signed-off-by: Matthew Heon <matthew.heon@gmail.com>

4 files changed, 1133 insertions, 0 deletions

diff --git a/pkg/storage/doc.go b/pkg/storage/doc.go
new file mode 100644
index 000000000..6366b22a4
--- /dev/null
+++ b/pkg/storage/doc.go
@@ -0,0 +1,5 @@
+// Package storage provides helper functions for creating and managing CRI pod
+// sandboxes and containers and metadata associated with them in the format
+// that crio understands.  The API it provides should be considered to be
+// unstable.
+package storage
diff --git a/pkg/storage/image.go b/pkg/storage/image.go
new file mode 100644
index 000000000..ddff82c1c
--- /dev/null
+++ b/pkg/storage/image.go
@@ -0,0 +1,551 @@
+package storage
+
+import (
+	"errors"
+	"fmt"
+	"net"
+	"path"
+	"regexp"
+	"strings"
+
+	"github.com/containers/image/copy"
+	"github.com/containers/image/docker/reference"
+	"github.com/containers/image/image"
+	"github.com/containers/image/signature"
+	istorage "github.com/containers/image/storage"
+	"github.com/containers/image/transports/alltransports"
+	"github.com/containers/image/types"
+	"github.com/containers/storage"
+	distreference "github.com/docker/distribution/reference"
+)
+
+// ImageResult wraps a subset of information about an image: its ID, its names,
+// and the size, if known, or nil if it isn't.
+type ImageResult struct {
+	ID       string
+	Names    []string
+	Digests  []string
+	Size     *uint64
+	ImageRef string
+}
+
+type indexInfo struct {
+	name   string
+	secure bool
+}
+
+type imageService struct {
+	store                 storage.Store
+	defaultTransport      string
+	insecureRegistryCIDRs []*net.IPNet
+	indexConfigs          map[string]*indexInfo
+	registries            []string
+}
+
+// ImageServer wraps up various CRI-related activities into a reusable
+// implementation.
+type ImageServer interface {
+	// ListImages returns list of all images which match the filter.
+	ListImages(systemContext *types.SystemContext, filter string) ([]ImageResult, error)
+	// ImageStatus returns status of an image which matches the filter.
+	ImageStatus(systemContext *types.SystemContext, filter string) (*ImageResult, error)
+	// PullImage imports an image from the specified location.
+	PullImage(systemContext *types.SystemContext, imageName string, options *copy.Options) (types.ImageReference, error)
+	// UntagImage removes a name from the specified image, and if it was
+	// the only name the image had, removes the image.
+	UntagImage(systemContext *types.SystemContext, imageName string) error
+	// RemoveImage deletes the specified image.
+	RemoveImage(systemContext *types.SystemContext, imageName string) error
+	// GetStore returns the reference to the storage library Store which
+	// the image server uses to hold images, and is the destination used
+	// when it's asked to pull an image.
+	GetStore() storage.Store
+	// CanPull preliminary checks whether we're allowed to pull an image
+	CanPull(imageName string, options *copy.Options) (bool, error)
+	// ResolveNames takes an image reference and if it's unqualified (w/o hostname),
+	// it uses crio's default registries to qualify it.
+	ResolveNames(imageName string) ([]string, error)
+}
+
+func (svc *imageService) getRef(name string) (types.ImageReference, error) {
+	ref, err := alltransports.ParseImageName(name)
+	if err != nil {
+		ref2, err2 := istorage.Transport.ParseStoreReference(svc.store, "@"+name)
+		if err2 != nil {
+			ref3, err3 := istorage.Transport.ParseStoreReference(svc.store, name)
+			if err3 != nil {
+				return nil, err
+			}
+			ref2 = ref3
+		}
+		ref = ref2
+	}
+	return ref, nil
+}
+
+func (svc *imageService) ListImages(systemContext *types.SystemContext, filter string) ([]ImageResult, error) {
+	results := []ImageResult{}
+	if filter != "" {
+		ref, err := svc.getRef(filter)
+		if err != nil {
+			return nil, err
+		}
+		if image, err := istorage.Transport.GetStoreImage(svc.store, ref); err == nil {
+			img, err := ref.NewImage(systemContext)
+			if err != nil {
+				return nil, err
+			}
+			size := imageSize(img)
+			img.Close()
+			results = append(results, ImageResult{
+				ID:    image.ID,
+				Names: image.Names,
+				Size:  size,
+			})
+		}
+	} else {
+		images, err := svc.store.Images()
+		if err != nil {
+			return nil, err
+		}
+		for _, image := range images {
+			ref, err := istorage.Transport.ParseStoreReference(svc.store, "@"+image.ID)
+			if err != nil {
+				return nil, err
+			}
+			img, err := ref.NewImage(systemContext)
+			if err != nil {
+				return nil, err
+			}
+			size := imageSize(img)
+			img.Close()
+			results = append(results, ImageResult{
+				ID:    image.ID,
+				Names: image.Names,
+				Size:  size,
+			})
+		}
+	}
+	return results, nil
+}
+
+func (svc *imageService) ImageStatus(systemContext *types.SystemContext, nameOrID string) (*ImageResult, error) {
+	ref, err := alltransports.ParseImageName(nameOrID)
+	if err != nil {
+		ref2, err2 := istorage.Transport.ParseStoreReference(svc.store, "@"+nameOrID)
+		if err2 != nil {
+			ref3, err3 := istorage.Transport.ParseStoreReference(svc.store, nameOrID)
+			if err3 != nil {
+				return nil, err
+			}
+			ref2 = ref3
+		}
+		ref = ref2
+	}
+	image, err := istorage.Transport.GetStoreImage(svc.store, ref)
+	if err != nil {
+		return nil, err
+	}
+
+	img, err := ref.NewImage(systemContext)
+	if err != nil {
+		return nil, err
+	}
+	size := imageSize(img)
+	img.Close()
+
+	result := ImageResult{
+		ID:    image.ID,
+		Names: image.Names,
+		Size:  size,
+	}
+	if len(image.Names) > 0 {
+		result.ImageRef = image.Names[0]
+		if ref2, err2 := istorage.Transport.ParseStoreReference(svc.store, image.Names[0]); err2 == nil {
+			if dref := ref2.DockerReference(); dref != nil {
+				result.ImageRef = reference.FamiliarString(dref)
+			}
+		}
+	}
+
+	return &result, nil
+}
+
+func imageSize(img types.Image) *uint64 {
+	if sum, err := img.Size(); err == nil {
+		usum := uint64(sum)
+		return &usum
+	}
+	return nil
+}
+
+func (svc *imageService) CanPull(imageName string, options *copy.Options) (bool, error) {
+	srcRef, err := svc.prepareImage(imageName, options)
+	if err != nil {
+		return false, err
+	}
+	rawSource, err := srcRef.NewImageSource(options.SourceCtx)
+	if err != nil {
+		return false, err
+	}
+	src, err := image.FromSource(rawSource)
+	if err != nil {
+		rawSource.Close()
+		return false, err
+	}
+	src.Close()
+	return true, nil
+}
+
+// prepareImage creates an image reference from an image string and set options
+// for the source context
+func (svc *imageService) prepareImage(imageName string, options *copy.Options) (types.ImageReference, error) {
+	if imageName == "" {
+		return nil, storage.ErrNotAnImage
+	}
+
+	srcRef, err := alltransports.ParseImageName(imageName)
+	if err != nil {
+		if svc.defaultTransport == "" {
+			return nil, err
+		}
+		srcRef2, err2 := alltransports.ParseImageName(svc.defaultTransport + imageName)
+		if err2 != nil {
+			return nil, err
+		}
+		srcRef = srcRef2
+	}
+
+	if options.SourceCtx == nil {
+		options.SourceCtx = &types.SystemContext{}
+	}
+
+	hostname := reference.Domain(srcRef.DockerReference())
+	if secure := svc.isSecureIndex(hostname); !secure {
+		options.SourceCtx.DockerInsecureSkipTLSVerify = !secure
+	}
+	return srcRef, nil
+}
+
+func (svc *imageService) PullImage(systemContext *types.SystemContext, imageName string, options *copy.Options) (types.ImageReference, error) {
+	policy, err := signature.DefaultPolicy(systemContext)
+	if err != nil {
+		return nil, err
+	}
+	policyContext, err := signature.NewPolicyContext(policy)
+	if err != nil {
+		return nil, err
+	}
+	if options == nil {
+		options = &copy.Options{}
+	}
+
+	srcRef, err := svc.prepareImage(imageName, options)
+	if err != nil {
+		return nil, err
+	}
+
+	dest := imageName
+	if srcRef.DockerReference() != nil {
+		dest = srcRef.DockerReference().Name()
+		if tagged, ok := srcRef.DockerReference().(reference.NamedTagged); ok {
+			dest = dest + ":" + tagged.Tag()
+		}
+		if canonical, ok := srcRef.DockerReference().(reference.Canonical); ok {
+			dest = dest + "@" + canonical.Digest().String()
+		}
+	}
+	destRef, err := istorage.Transport.ParseStoreReference(svc.store, dest)
+	if err != nil {
+		return nil, err
+	}
+	err = copy.Image(policyContext, destRef, srcRef, options)
+	if err != nil {
+		return nil, err
+	}
+	return destRef, nil
+}
+
+func (svc *imageService) UntagImage(systemContext *types.SystemContext, nameOrID string) error {
+	ref, err := alltransports.ParseImageName(nameOrID)
+	if err != nil {
+		ref2, err2 := istorage.Transport.ParseStoreReference(svc.store, "@"+nameOrID)
+		if err2 != nil {
+			ref3, err3 := istorage.Transport.ParseStoreReference(svc.store, nameOrID)
+			if err3 != nil {
+				return err
+			}
+			ref2 = ref3
+		}
+		ref = ref2
+	}
+
+	img, err := istorage.Transport.GetStoreImage(svc.store, ref)
+	if err != nil {
+		return err
+	}
+
+	if nameOrID != img.ID {
+		namedRef, err := svc.prepareImage(nameOrID, &copy.Options{})
+		if err != nil {
+			return err
+		}
+
+		name := nameOrID
+		if namedRef.DockerReference() != nil {
+			name = namedRef.DockerReference().Name()
+			if tagged, ok := namedRef.DockerReference().(reference.NamedTagged); ok {
+				name = name + ":" + tagged.Tag()
+			}
+			if canonical, ok := namedRef.DockerReference().(reference.Canonical); ok {
+				name = name + "@" + canonical.Digest().String()
+			}
+		}
+
+		prunedNames := make([]string, 0, len(img.Names))
+		for _, imgName := range img.Names {
+			if imgName != name && imgName != nameOrID {
+				prunedNames = append(prunedNames, imgName)
+			}
+		}
+
+		if len(prunedNames) > 0 {
+			return svc.store.SetNames(img.ID, prunedNames)
+		}
+	}
+
+	return ref.DeleteImage(systemContext)
+}
+
+func (svc *imageService) RemoveImage(systemContext *types.SystemContext, nameOrID string) error {
+	ref, err := alltransports.ParseImageName(nameOrID)
+	if err != nil {
+		ref2, err2 := istorage.Transport.ParseStoreReference(svc.store, "@"+nameOrID)
+		if err2 != nil {
+			ref3, err3 := istorage.Transport.ParseStoreReference(svc.store, nameOrID)
+			if err3 != nil {
+				return err
+			}
+			ref2 = ref3
+		}
+		ref = ref2
+	}
+	return ref.DeleteImage(systemContext)
+}
+
+func (svc *imageService) GetStore() storage.Store {
+	return svc.store
+}
+
+func (svc *imageService) isSecureIndex(indexName string) bool {
+	if index, ok := svc.indexConfigs[indexName]; ok {
+		return index.secure
+	}
+
+	host, _, err := net.SplitHostPort(indexName)
+	if err != nil {
+		// assume indexName is of the form `host` without the port and go on.
+		host = indexName
+	}
+
+	addrs, err := net.LookupIP(host)
+	if err != nil {
+		ip := net.ParseIP(host)
+		if ip != nil {
+			addrs = []net.IP{ip}
+		}
+
+		// if ip == nil, then `host` is neither an IP nor it could be looked up,
+		// either because the index is unreachable, or because the index is behind an HTTP proxy.
+		// So, len(addrs) == 0 and we're not aborting.
+	}
+
+	// Try CIDR notation only if addrs has any elements, i.e. if `host`'s IP could be determined.
+	for _, addr := range addrs {
+		for _, ipnet := range svc.insecureRegistryCIDRs {
+			// check if the addr falls in the subnet
+			if ipnet.Contains(addr) {
+				return false
+			}
+		}
+	}
+
+	return true
+}
+
+func isValidHostname(hostname string) bool {
+	return hostname != "" && !strings.Contains(hostname, "/") &&
+		(strings.Contains(hostname, ".") ||
+			strings.Contains(hostname, ":") || hostname == "localhost")
+}
+
+func isReferenceFullyQualified(reposName reference.Named) bool {
+	indexName, _, _ := splitReposName(reposName)
+	return indexName != ""
+}
+
+const (
+	// defaultHostname is the default built-in hostname
+	defaultHostname = "docker.io"
+	// legacyDefaultHostname is automatically converted to DefaultHostname
+	legacyDefaultHostname = "index.docker.io"
+	// defaultRepoPrefix is the prefix used for default repositories in default host
+	defaultRepoPrefix = "library/"
+)
+
+// splitReposName breaks a reposName into an index name and remote name
+func splitReposName(reposName reference.Named) (indexName string, remoteName reference.Named, err error) {
+	var remoteNameStr string
+	indexName, remoteNameStr = distreference.SplitHostname(reposName)
+	if !isValidHostname(indexName) {
+		// This is a Docker Index repos (ex: samalba/hipache or ubuntu)
+		// 'docker.io'
+		indexName = ""
+		remoteName = reposName
+	} else {
+		remoteName, err = withName(remoteNameStr)
+	}
+	return
+}
+
+func validateName(name string) error {
+	if err := validateID(strings.TrimPrefix(name, defaultHostname+"/")); err == nil {
+		return fmt.Errorf("Invalid repository name (%s), cannot specify 64-byte hexadecimal strings", name)
+	}
+	return nil
+}
+
+var validHex = regexp.MustCompile(`^([a-f0-9]{64})$`)
+
+// validateID checks whether an ID string is a valid image ID.
+func validateID(id string) error {
+	if ok := validHex.MatchString(id); !ok {
+		return fmt.Errorf("image ID %q is invalid", id)
+	}
+	return nil
+}
+
+// withName returns a named object representing the given string. If the input
+// is invalid ErrReferenceInvalidFormat will be returned.
+func withName(name string) (reference.Named, error) {
+	name, err := normalize(name)
+	if err != nil {
+		return nil, err
+	}
+	if err := validateName(name); err != nil {
+		return nil, err
+	}
+	r, err := distreference.WithName(name)
+	return r, err
+}
+
+// splitHostname splits a repository name to hostname and remotename string.
+// If no valid hostname is found, empty string will be returned as a resulting
+// hostname. Repository name needs to be already validated before.
+func splitHostname(name string) (hostname, remoteName string) {
+	i := strings.IndexRune(name, '/')
+	if i == -1 || (!strings.ContainsAny(name[:i], ".:") && name[:i] != "localhost") {
+		hostname, remoteName = "", name
+	} else {
+		hostname, remoteName = name[:i], name[i+1:]
+	}
+	if hostname == legacyDefaultHostname {
+		hostname = defaultHostname
+	}
+	if hostname == defaultHostname && !strings.ContainsRune(remoteName, '/') {
+		remoteName = defaultRepoPrefix + remoteName
+	}
+	return
+}
+
+// normalize returns a repository name in its normalized form, meaning it
+// will contain library/ prefix for official images.
+func normalize(name string) (string, error) {
+	host, remoteName := splitHostname(name)
+	if strings.ToLower(remoteName) != remoteName {
+		return "", errors.New("invalid reference format: repository name must be lowercase")
+	}
+	if host == defaultHostname {
+		if strings.HasPrefix(remoteName, defaultRepoPrefix) {
+			remoteName = strings.TrimPrefix(remoteName, defaultRepoPrefix)
+		}
+		return host + "/" + remoteName, nil
+	}
+	return name, nil
+}
+
+func (svc *imageService) ResolveNames(imageName string) ([]string, error) {
+	r, err := reference.ParseNormalizedNamed(imageName)
+	if err != nil {
+		return nil, err
+	}
+	if isReferenceFullyQualified(r) {
+		// this means the image is already fully qualified
+		return []string{imageName}, nil
+	}
+	// we got an unqualified image here, we can't go ahead w/o registries configured
+	// properly.
+	if len(svc.registries) == 0 {
+		return nil, errors.New("no registries configured while trying to pull an unqualified image")
+	}
+	// this means we got an image in the form of "busybox"
+	// we need to use additional registries...
+	// normalize the unqualified image to be domain/repo/image...
+	_, rest := splitDomain(r.Name())
+	images := []string{}
+	for _, r := range svc.registries {
+		images = append(images, path.Join(r, rest))
+	}
+	return images, nil
+}
+
+// GetImageService returns an ImageServer that uses the passed-in store, and
+// which will prepend the passed-in defaultTransport value to an image name if
+// a name that's passed to its PullImage() method can't be resolved to an image
+// in the store and can't be resolved to a source on its own.
+func GetImageService(store storage.Store, defaultTransport string, insecureRegistries []string, registries []string) (ImageServer, error) {
+	if store == nil {
+		var err error
+		store, err = storage.GetStore(storage.DefaultStoreOptions)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	seenRegistries := make(map[string]bool, len(registries))
+	cleanRegistries := []string{}
+	for _, r := range registries {
+		if seenRegistries[r] {
+			continue
+		}
+		cleanRegistries = append(cleanRegistries, r)
+		seenRegistries[r] = true
+	}
+
+	is := &imageService{
+		store:                 store,
+		defaultTransport:      defaultTransport,
+		indexConfigs:          make(map[string]*indexInfo, 0),
+		insecureRegistryCIDRs: make([]*net.IPNet, 0),
+		registries:            cleanRegistries,
+	}
+
+	insecureRegistries = append(insecureRegistries, "127.0.0.0/8")
+	// Split --insecure-registry into CIDR and registry-specific settings.
+	for _, r := range insecureRegistries {
+		// Check if CIDR was passed to --insecure-registry
+		_, ipnet, err := net.ParseCIDR(r)
+		if err == nil {
+			// Valid CIDR.
+			is.insecureRegistryCIDRs = append(is.insecureRegistryCIDRs, ipnet)
+		} else {
+			// Assume `host:port` if not CIDR.
+			is.indexConfigs[r] = &indexInfo{
+				name:   r,
+				secure: false,
+			}
+		}
+	}
+
+	return is, nil
+}
diff --git a/pkg/storage/image_regexp.go b/pkg/storage/image_regexp.go
new file mode 100644
index 000000000..96de64884
--- /dev/null
+++ b/pkg/storage/image_regexp.go
@@ -0,0 +1,125 @@
+package storage
+
+// This is a fork of docker/distribution code to be used when manipulating image
+// references.
+// DO NOT EDIT THIS FILE.
+
+import "regexp"
+
+var (
+	// alphaNumericRegexp defines the alpha numeric atom, typically a
+	// component of names. This only allows lower case characters and digits.
+	alphaNumericRegexp = match(`[a-z0-9]+`)
+
+	// separatorRegexp defines the separators allowed to be embedded in name
+	// components. This allow one period, one or two underscore and multiple
+	// dashes.
+	separatorRegexp = match(`(?:[._]|__|[-]*)`)
+
+	// nameComponentRegexp restricts registry path component names to start
+	// with at least one letter or number, with following parts able to be
+	// separated by one period, one or two underscore and multiple dashes.
+	nameComponentRegexp = expression(
+		alphaNumericRegexp,
+		optional(repeated(separatorRegexp, alphaNumericRegexp)))
+
+	// domainComponentRegexp restricts the registry domain component of a
+	// repository name to start with a component as defined by domainRegexp
+	// and followed by an optional port.
+	domainComponentRegexp = match(`(?:[a-zA-Z0-9]|[a-zA-Z0-9][a-zA-Z0-9-]*[a-zA-Z0-9])`)
+
+	// domainRegexp defines the structure of potential domain components
+	// that may be part of image names. This is purposely a subset of what is
+	// allowed by DNS to ensure backwards compatibility with Docker image
+	// names.
+	domainRegexp = expression(
+		domainComponentRegexp,
+		optional(repeated(literal(`.`), domainComponentRegexp)),
+		optional(literal(`:`), match(`[0-9]+`)))
+
+	// NameRegexp is the format for the name component of references. The
+	// regexp has capturing groups for the domain and name part omitting
+	// the separating forward slash from either.
+	NameRegexp = expression(
+		optional(domainRegexp, literal(`/`)),
+		nameComponentRegexp,
+		optional(repeated(literal(`/`), nameComponentRegexp)))
+
+	// anchoredNameRegexp is used to parse a name value, capturing the
+	// domain and trailing components.
+	anchoredNameRegexp = anchored(
+		optional(capture(domainRegexp), literal(`/`)),
+		capture(nameComponentRegexp,
+			optional(repeated(literal(`/`), nameComponentRegexp))))
+
+	// IdentifierRegexp is the format for string identifier used as a
+	// content addressable identifier using sha256. These identifiers
+	// are like digests without the algorithm, since sha256 is used.
+	IdentifierRegexp = match(`([a-f0-9]{64})`)
+
+	// ShortIdentifierRegexp is the format used to represent a prefix
+	// of an identifier. A prefix may be used to match a sha256 identifier
+	// within a list of trusted identifiers.
+	ShortIdentifierRegexp = match(`([a-f0-9]{6,64})`)
+)
+
+// match compiles the string to a regular expression.
+var match = regexp.MustCompile
+
+// literal compiles s into a literal regular expression, escaping any regexp
+// reserved characters.
+func literal(s string) *regexp.Regexp {
+	re := match(regexp.QuoteMeta(s))
+
+	if _, complete := re.LiteralPrefix(); !complete {
+		panic("must be a literal")
+	}
+
+	return re
+}
+
+func splitDomain(name string) (string, string) {
+	match := anchoredNameRegexp.FindStringSubmatch(name)
+	if len(match) != 3 {
+		return "", name
+	}
+	return match[1], match[2]
+}
+
+// expression defines a full expression, where each regular expression must
+// follow the previous.
+func expression(res ...*regexp.Regexp) *regexp.Regexp {
+	var s string
+	for _, re := range res {
+		s += re.String()
+	}
+
+	return match(s)
+}
+
+// optional wraps the expression in a non-capturing group and makes the
+// production optional.
+func optional(res ...*regexp.Regexp) *regexp.Regexp {
+	return match(group(expression(res...)).String() + `?`)
+}
+
+// repeated wraps the regexp in a non-capturing group to get one or more
+// matches.
+func repeated(res ...*regexp.Regexp) *regexp.Regexp {
+	return match(group(expression(res...)).String() + `+`)
+}
+
+// group wraps the regexp in a non-capturing group.
+func group(res ...*regexp.Regexp) *regexp.Regexp {
+	return match(`(?:` + expression(res...).String() + `)`)
+}
+
+// capture wraps the expression in a capturing group.
+func capture(res ...*regexp.Regexp) *regexp.Regexp {
+	return match(`(` + expression(res...).String() + `)`)
+}
+
+// anchored anchors the regular expression by adding start and end delimiters.
+func anchored(res ...*regexp.Regexp) *regexp.Regexp {
+	return match(`^` + expression(res...).String() + `$`)
+}
diff --git a/pkg/storage/runtime.go b/pkg/storage/runtime.go
new file mode 100644
index 000000000..bc402f413
--- /dev/null
+++ b/pkg/storage/runtime.go
@@ -0,0 +1,452 @@
+package storage
+
+import (
+	"encoding/json"
+	"fmt"
+	"time"
+
+	"github.com/containers/image/copy"
+	istorage "github.com/containers/image/storage"
+	"github.com/containers/image/transports/alltransports"
+	"github.com/containers/image/types"
+	"github.com/containers/storage"
+	"github.com/opencontainers/image-spec/specs-go/v1"
+	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
+)
+
+var (
+	// ErrInvalidPodName is returned when a pod name specified to a
+	// function call is found to be invalid (most often, because it's
+	// empty).
+	ErrInvalidPodName = errors.New("invalid pod name")
+	// ErrInvalidImageName is returned when an image name specified to a
+	// function call is found to be invalid (most often, because it's
+	// empty).
+	ErrInvalidImageName = errors.New("invalid image name")
+	// ErrInvalidContainerName is returned when a container name specified
+	// to a function call is found to be invalid (most often, because it's
+	// empty).
+	ErrInvalidContainerName = errors.New("invalid container name")
+	// ErrInvalidSandboxID is returned when a sandbox ID specified to a
+	// function call is found to be invalid (because it's either
+	// empty or doesn't match a valid sandbox).
+	ErrInvalidSandboxID = errors.New("invalid sandbox ID")
+	// ErrInvalidContainerID is returned when a container ID specified to a
+	// function call is found to be invalid (because it's either
+	// empty or doesn't match a valid container).
+	ErrInvalidContainerID = errors.New("invalid container ID")
+)
+
+type runtimeService struct {
+	storageImageServer ImageServer
+	pauseImage         string
+}
+
+// ContainerInfo wraps a subset of information about a container: its ID and
+// the locations of its nonvolatile and volatile per-container directories,
+// along with a copy of the configuration blob from the image that was used to
+// create the container, if the image had a configuration.
+type ContainerInfo struct {
+	ID     string
+	Dir    string
+	RunDir string
+	Config *v1.Image
+}
+
+// RuntimeServer wraps up various CRI-related activities into a reusable
+// implementation.
+type RuntimeServer interface {
+	// CreatePodSandbox creates a pod infrastructure container, using the
+	// specified PodID for the infrastructure container's ID.  In the CRI
+	// view of things, a sandbox is distinct from its containers, including
+	// its infrastructure container, but at this level the sandbox is
+	// essentially the same as its infrastructure container, with a
+	// container's membership in a pod being signified by it listing the
+	// same pod ID in its metadata that the pod's other members do, and
+	// with the pod's infrastructure container having the same value for
+	// both its pod's ID and its container ID.
+	// Pointer arguments can be nil.  Either the image name or ID can be
+	// omitted, but not both.  All other arguments are required.
+	CreatePodSandbox(systemContext *types.SystemContext, podName, podID, imageName, imageID, containerName, metadataName, uid, namespace string, attempt uint32, copyOptions *copy.Options) (ContainerInfo, error)
+	// RemovePodSandbox deletes a pod sandbox's infrastructure container.
+	// The CRI expects that a sandbox can't be removed unless its only
+	// container is its infrastructure container, but we don't enforce that
+	// here, since we're just keeping track of it for higher level APIs.
+	RemovePodSandbox(idOrName string) error
+
+	// GetContainerMetadata returns the metadata we've stored for a container.
+	GetContainerMetadata(idOrName string) (RuntimeContainerMetadata, error)
+	// SetContainerMetadata updates the metadata we've stored for a container.
+	SetContainerMetadata(idOrName string, metadata RuntimeContainerMetadata) error
+
+	// CreateContainer creates a container with the specified ID.
+	// Pointer arguments can be nil.  Either the image name or ID can be
+	// omitted, but not both.  All other arguments are required.
+	CreateContainer(systemContext *types.SystemContext, podName, podID, imageName, imageID, containerName, containerID, metadataName string, attempt uint32, mountLabel string, copyOptions *copy.Options) (ContainerInfo, error)
+	// DeleteContainer deletes a container, unmounting it first if need be.
+	DeleteContainer(idOrName string) error
+
+	// StartContainer makes sure a container's filesystem is mounted, and
+	// returns the location of its root filesystem, which is not guaranteed
+	// by lower-level drivers to never change.
+	StartContainer(idOrName string) (string, error)
+	// StopContainer attempts to unmount a container's root filesystem,
+	// freeing up any kernel resources which may be limited.
+	StopContainer(idOrName string) error
+
+	// GetWorkDir returns the path of a nonvolatile directory on the
+	// filesystem (somewhere under the Store's Root directory) which can be
+	// used to store arbitrary data that is specific to the container.  It
+	// will be removed automatically when the container is deleted.
+	GetWorkDir(id string) (string, error)
+	// GetRunDir returns the path of a volatile directory (does not survive
+	// the host rebooting, somewhere under the Store's RunRoot directory)
+	// on the filesystem which can be used to store arbitrary data that is
+	// specific to the container.  It will be removed automatically when
+	// the container is deleted.
+	GetRunDir(id string) (string, error)
+}
+
+// RuntimeContainerMetadata is the structure that we encode as JSON and store
+// in the metadata field of storage.Container objects.  It is used for
+// specifying attributes of pod sandboxes and containers when they are being
+// created, and allows a container's MountLabel, and possibly other values, to
+// be modified in one read/write cycle via calls to
+// RuntimeServer.ContainerMetadata, RuntimeContainerMetadata.SetMountLabel,
+// and RuntimeServer.SetContainerMetadata.
+type RuntimeContainerMetadata struct {
+	// Pod is true if this is the pod's infrastructure container.
+	Pod bool `json:"pod,omitempty"` // Applicable to both PodSandboxes and Containers
+	// The pod's name and ID, kept for use by upper layers in determining
+	// which containers belong to which pods.
+	PodName string `json:"pod-name"` // Applicable to both PodSandboxes and Containers, mandatory
+	PodID   string `json:"pod-id"`   // Applicable to both PodSandboxes and Containers, mandatory
+	// The provided name and the ID of the image that was used to
+	// instantiate the container.
+	ImageName string `json:"image-name"` // Applicable to both PodSandboxes and Containers
+	ImageID   string `json:"image-id"`   // Applicable to both PodSandboxes and Containers
+	// The container's name, which for an infrastructure container is usually PodName + "-infra".
+	ContainerName string `json:"name"` // Applicable to both PodSandboxes and Containers, mandatory
+	// The name as originally specified in PodSandbox or Container CRI metadata.
+	MetadataName string `json:"metadata-name"`        // Applicable to both PodSandboxes and Containers, mandatory
+	UID          string `json:"uid,omitempty"`        // Only applicable to pods
+	Namespace    string `json:"namespace,omitempty"`  // Only applicable to pods
+	Attempt      uint32 `json:"attempt,omitempty"`    // Applicable to both PodSandboxes and Containers
+	CreatedAt    int64  `json:"created-at"`           // Applicable to both PodSandboxes and Containers
+	MountLabel   string `json:"mountlabel,omitempty"` // Applicable to both PodSandboxes and Containers
+}
+
+// SetMountLabel updates the mount label held by a RuntimeContainerMetadata
+// object.
+func (metadata *RuntimeContainerMetadata) SetMountLabel(mountLabel string) {
+	metadata.MountLabel = mountLabel
+}
+
+func (r *runtimeService) createContainerOrPodSandbox(systemContext *types.SystemContext, podName, podID, imageName, imageID, containerName, containerID, metadataName, uid, namespace string, attempt uint32, mountLabel string, options *copy.Options) (ContainerInfo, error) {
+	var ref types.ImageReference
+	if podName == "" || podID == "" {
+		return ContainerInfo{}, ErrInvalidPodName
+	}
+	if imageName == "" && imageID == "" {
+		return ContainerInfo{}, ErrInvalidImageName
+	}
+	if containerName == "" {
+		return ContainerInfo{}, ErrInvalidContainerName
+	}
+	if metadataName == "" {
+		metadataName = containerName
+	}
+
+	// Check if we have the specified image.
+	ref, err := istorage.Transport.ParseStoreReference(r.storageImageServer.GetStore(), imageName)
+	if err != nil {
+		// Maybe it's some other transport's copy of the image?
+		otherRef, err2 := alltransports.ParseImageName(imageName)
+		if err2 == nil && otherRef.DockerReference() != nil {
+			ref, err = istorage.Transport.ParseStoreReference(r.storageImageServer.GetStore(), otherRef.DockerReference().Name())
+		}
+		if err != nil {
+			// Maybe the image ID is sufficient?
+			ref, err = istorage.Transport.ParseStoreReference(r.storageImageServer.GetStore(), "@"+imageID)
+			if err != nil {
+				return ContainerInfo{}, err
+			}
+		}
+	}
+	img, err := istorage.Transport.GetStoreImage(r.storageImageServer.GetStore(), ref)
+	if img == nil && errors.Cause(err) == storage.ErrImageUnknown && imageName == r.pauseImage {
+		image := imageID
+		if imageName != "" {
+			image = imageName
+		}
+		if image == "" {
+			return ContainerInfo{}, ErrInvalidImageName
+		}
+		logrus.Debugf("couldn't find image %q, retrieving it", image)
+		ref, err = r.storageImageServer.PullImage(systemContext, image, options)
+		if err != nil {
+			return ContainerInfo{}, err
+		}
+		img, err = istorage.Transport.GetStoreImage(r.storageImageServer.GetStore(), ref)
+		if err != nil {
+			return ContainerInfo{}, err
+		}
+		logrus.Debugf("successfully pulled image %q", image)
+	}
+	if img == nil && errors.Cause(err) == storage.ErrImageUnknown {
+		if imageID == "" {
+			return ContainerInfo{}, fmt.Errorf("image %q not present in image store", imageName)
+		}
+		if imageName == "" {
+			return ContainerInfo{}, fmt.Errorf("image with ID %q not present in image store", imageID)
+		}
+		return ContainerInfo{}, fmt.Errorf("image %q with ID %q not present in image store", imageName, imageID)
+	}
+
+	// Pull out a copy of the image's configuration.
+	image, err := ref.NewImage(systemContext)
+	if err != nil {
+		return ContainerInfo{}, err
+	}
+	defer image.Close()
+
+	imageConfig, err := image.OCIConfig()
+	if err != nil {
+		return ContainerInfo{}, err
+	}
+
+	// Update the image name and ID.
+	if imageName == "" && len(img.Names) > 0 {
+		imageName = img.Names[0]
+	}
+	imageID = img.ID
+
+	// Build metadata to store with the container.
+	metadata := RuntimeContainerMetadata{
+		Pod:           containerID == podID,
+		PodName:       podName,
+		PodID:         podID,
+		ImageName:     imageName,
+		ImageID:       imageID,
+		ContainerName: containerName,
+		MetadataName:  metadataName,
+		UID:           uid,
+		Namespace:     namespace,
+		Attempt:       attempt,
+		CreatedAt:     time.Now().Unix(),
+		MountLabel:    mountLabel,
+	}
+	mdata, err := json.Marshal(&metadata)
+	if err != nil {
+		return ContainerInfo{}, err
+	}
+
+	// Build the container.
+	names := []string{metadata.ContainerName}
+	if metadata.Pod {
+		names = append(names, metadata.PodName)
+	}
+	container, err := r.storageImageServer.GetStore().CreateContainer(containerID, names, img.ID, "", string(mdata), nil)
+	if err != nil {
+		if metadata.Pod {
+			logrus.Debugf("failed to create pod sandbox %s(%s): %v", metadata.PodName, metadata.PodID, err)
+		} else {
+			logrus.Debugf("failed to create container %s(%s): %v", metadata.ContainerName, containerID, err)
+		}
+		return ContainerInfo{}, err
+	}
+	if metadata.Pod {
+		logrus.Debugf("created pod sandbox %q", container.ID)
+	} else {
+		logrus.Debugf("created container %q", container.ID)
+	}
+
+	// If anything fails after this point, we need to delete the incomplete
+	// container before returning.
+	defer func() {
+		if err != nil {
+			if err2 := r.storageImageServer.GetStore().DeleteContainer(container.ID); err2 != nil {
+				if metadata.Pod {
+					logrus.Infof("%v deleting partially-created pod sandbox %q", err2, container.ID)
+				} else {
+					logrus.Infof("%v deleting partially-created container %q", err2, container.ID)
+				}
+				return
+			}
+			logrus.Infof("deleted partially-created container %q", container.ID)
+		}
+	}()
+
+	// Add a name to the container's layer so that it's easier to follow
+	// what's going on if we're just looking at the storage-eye view of things.
+	layerName := metadata.ContainerName + "-layer"
+	names, err = r.storageImageServer.GetStore().Names(container.LayerID)
+	if err != nil {
+		return ContainerInfo{}, err
+	}
+	names = append(names, layerName)
+	err = r.storageImageServer.GetStore().SetNames(container.LayerID, names)
+	if err != nil {
+		return ContainerInfo{}, err
+	}
+
+	// Find out where the container work directories are, so that we can return them.
+	containerDir, err := r.storageImageServer.GetStore().ContainerDirectory(container.ID)
+	if err != nil {
+		return ContainerInfo{}, err
+	}
+	if metadata.Pod {
+		logrus.Debugf("pod sandbox %q has work directory %q", container.ID, containerDir)
+	} else {
+		logrus.Debugf("container %q has work directory %q", container.ID, containerDir)
+	}
+
+	containerRunDir, err := r.storageImageServer.GetStore().ContainerRunDirectory(container.ID)
+	if err != nil {
+		return ContainerInfo{}, err
+	}
+	if metadata.Pod {
+		logrus.Debugf("pod sandbox %q has run directory %q", container.ID, containerRunDir)
+	} else {
+		logrus.Debugf("container %q has run directory %q", container.ID, containerRunDir)
+	}
+
+	return ContainerInfo{
+		ID:     container.ID,
+		Dir:    containerDir,
+		RunDir: containerRunDir,
+		Config: imageConfig,
+	}, nil
+}
+
+func (r *runtimeService) CreatePodSandbox(systemContext *types.SystemContext, podName, podID, imageName, imageID, containerName, metadataName, uid, namespace string, attempt uint32, copyOptions *copy.Options) (ContainerInfo, error) {
+	return r.createContainerOrPodSandbox(systemContext, podName, podID, imageName, imageID, containerName, podID, metadataName, uid, namespace, attempt, "", copyOptions)
+}
+
+func (r *runtimeService) CreateContainer(systemContext *types.SystemContext, podName, podID, imageName, imageID, containerName, containerID, metadataName string, attempt uint32, mountLabel string, copyOptions *copy.Options) (ContainerInfo, error) {
+	return r.createContainerOrPodSandbox(systemContext, podName, podID, imageName, imageID, containerName, containerID, metadataName, "", "", attempt, mountLabel, copyOptions)
+}
+
+func (r *runtimeService) RemovePodSandbox(idOrName string) error {
+	container, err := r.storageImageServer.GetStore().Container(idOrName)
+	if err != nil {
+		if errors.Cause(err) == storage.ErrContainerUnknown {
+			return ErrInvalidSandboxID
+		}
+		return err
+	}
+	err = r.storageImageServer.GetStore().DeleteContainer(container.ID)
+	if err != nil {
+		logrus.Debugf("failed to delete pod sandbox %q: %v", container.ID, err)
+		return err
+	}
+	return nil
+}
+
+func (r *runtimeService) DeleteContainer(idOrName string) error {
+	if idOrName == "" {
+		return ErrInvalidContainerID
+	}
+	container, err := r.storageImageServer.GetStore().Container(idOrName)
+	if err != nil {
+		return err
+	}
+	err = r.storageImageServer.GetStore().DeleteContainer(container.ID)
+	if err != nil {
+		logrus.Debugf("failed to delete container %q: %v", container.ID, err)
+		return err
+	}
+	return nil
+}
+
+func (r *runtimeService) SetContainerMetadata(idOrName string, metadata RuntimeContainerMetadata) error {
+	mdata, err := json.Marshal(&metadata)
+	if err != nil {
+		logrus.Debugf("failed to encode metadata for %q: %v", idOrName, err)
+		return err
+	}
+	return r.storageImageServer.GetStore().SetMetadata(idOrName, string(mdata))
+}
+
+func (r *runtimeService) GetContainerMetadata(idOrName string) (RuntimeContainerMetadata, error) {
+	metadata := RuntimeContainerMetadata{}
+	mdata, err := r.storageImageServer.GetStore().Metadata(idOrName)
+	if err != nil {
+		return metadata, err
+	}
+	if err = json.Unmarshal([]byte(mdata), &metadata); err != nil {
+		return metadata, err
+	}
+	return metadata, nil
+}
+
+func (r *runtimeService) StartContainer(idOrName string) (string, error) {
+	container, err := r.storageImageServer.GetStore().Container(idOrName)
+	if err != nil {
+		if errors.Cause(err) == storage.ErrContainerUnknown {
+			return "", ErrInvalidContainerID
+		}
+		return "", err
+	}
+	metadata := RuntimeContainerMetadata{}
+	if err = json.Unmarshal([]byte(container.Metadata), &metadata); err != nil {
+		return "", err
+	}
+	mountPoint, err := r.storageImageServer.GetStore().Mount(container.ID, metadata.MountLabel)
+	if err != nil {
+		logrus.Debugf("failed to mount container %q: %v", container.ID, err)
+		return "", err
+	}
+	logrus.Debugf("mounted container %q at %q", container.ID, mountPoint)
+	return mountPoint, nil
+}
+
+func (r *runtimeService) StopContainer(idOrName string) error {
+	if idOrName == "" {
+		return ErrInvalidContainerID
+	}
+	container, err := r.storageImageServer.GetStore().Container(idOrName)
+	if err != nil {
+		return err
+	}
+	err = r.storageImageServer.GetStore().Unmount(container.ID)
+	if err != nil {
+		logrus.Debugf("failed to unmount container %q: %v", container.ID, err)
+		return err
+	}
+	logrus.Debugf("unmounted container %q", container.ID)
+	return nil
+}
+
+func (r *runtimeService) GetWorkDir(id string) (string, error) {
+	container, err := r.storageImageServer.GetStore().Container(id)
+	if err != nil {
+		if errors.Cause(err) == storage.ErrContainerUnknown {
+			return "", ErrInvalidContainerID
+		}
+		return "", err
+	}
+	return r.storageImageServer.GetStore().ContainerDirectory(container.ID)
+}
+
+func (r *runtimeService) GetRunDir(id string) (string, error) {
+	container, err := r.storageImageServer.GetStore().Container(id)
+	if err != nil {
+		if errors.Cause(err) == storage.ErrContainerUnknown {
+			return "", ErrInvalidContainerID
+		}
+		return "", err
+	}
+	return r.storageImageServer.GetStore().ContainerRunDirectory(container.ID)
+}
+
+// GetRuntimeService returns a RuntimeServer that uses the passed-in image
+// service to pull and manage images, and its store to manage containers based
+// on those images.
+func GetRuntimeService(storageImageServer ImageServer, pauseImage string) RuntimeServer {
+	return &runtimeService{
+		storageImageServer: storageImageServer,
+		pauseImage:         pauseImage,
+	}
+}