moby/daemon/containerd/service.go
Djordje Lukic aec7a80c6f
c8d: Use reference counting while mounting a snapshot
Some snapshotters (like overlayfs or zfs) can't mount the same
directories twice. For example if the same directroy is used as an upper
directory in two mounts the kernel will output this warning:

    overlayfs: upperdir is in-use as upperdir/workdir of another mount, accessing files from both mounts will result in undefined behavior.

And indeed accessing the files from both mounts will result in an "No
such file or directory" error.

This change introduces reference counts for the mounts, if a directory
is already mounted the mount interface will only increment the mount
counter and return the mount target effectively making sure that the
filesystem doesn't end up in an undefined behavior.

Signed-off-by: Djordje Lukic <djordje.lukic@docker.com>
(cherry picked from commit 32d58144fd)
Signed-off-by: Bjorn Neergaard <bjorn.neergaard@docker.com>
2023-06-20 12:54:03 -06:00

245 lines
7.8 KiB
Go

package containerd
import (
"context"
"encoding/json"
"sync/atomic"
"github.com/containerd/containerd"
"github.com/containerd/containerd/content"
"github.com/containerd/containerd/plugin"
"github.com/containerd/containerd/remotes/docker"
"github.com/containerd/containerd/snapshots"
"github.com/docker/distribution/reference"
imagetypes "github.com/docker/docker/api/types/image"
"github.com/docker/docker/container"
daemonevents "github.com/docker/docker/daemon/events"
"github.com/docker/docker/daemon/images"
"github.com/docker/docker/daemon/snapshotter"
"github.com/docker/docker/errdefs"
"github.com/docker/docker/image"
"github.com/docker/docker/layer"
"github.com/docker/docker/registry"
"github.com/opencontainers/go-digest"
"github.com/opencontainers/image-spec/identity"
ocispec "github.com/opencontainers/image-spec/specs-go/v1"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
)
// ImageService implements daemon.ImageService
type ImageService struct {
client *containerd.Client
containers container.Store
snapshotter string
registryHosts RegistryHostsProvider
registryService RegistryConfigProvider
eventsService *daemonevents.Events
pruneRunning atomic.Bool
refCountMounter snapshotter.Mounter
}
type RegistryHostsProvider interface {
RegistryHosts() docker.RegistryHosts
}
type RegistryConfigProvider interface {
IsInsecureRegistry(host string) bool
ResolveRepository(name reference.Named) (*registry.RepositoryInfo, error)
}
type ImageServiceConfig struct {
Client *containerd.Client
Containers container.Store
Snapshotter string
HostsProvider RegistryHostsProvider
Registry RegistryConfigProvider
EventsService *daemonevents.Events
RefCountMounter snapshotter.Mounter
}
// NewService creates a new ImageService.
func NewService(config ImageServiceConfig) *ImageService {
return &ImageService{
client: config.Client,
containers: config.Containers,
snapshotter: config.Snapshotter,
registryHosts: config.HostsProvider,
registryService: config.Registry,
eventsService: config.EventsService,
refCountMounter: config.RefCountMounter,
}
}
// DistributionServices return services controlling daemon image storage.
func (i *ImageService) DistributionServices() images.DistributionServices {
return images.DistributionServices{}
}
// CountImages returns the number of images stored by ImageService
// called from info.go
func (i *ImageService) CountImages() int {
imgs, err := i.client.ListImages(context.TODO())
if err != nil {
return 0
}
return len(imgs)
}
// CreateLayer creates a filesystem layer for a container.
// called from create.go
// TODO: accept an opt struct instead of container?
func (i *ImageService) CreateLayer(container *container.Container, initFunc layer.MountInit) (layer.RWLayer, error) {
return nil, errdefs.NotImplemented(errdefs.NotImplemented(errors.New("not implemented")))
}
// LayerStoreStatus returns the status for each layer store
// called from info.go
func (i *ImageService) LayerStoreStatus() [][2]string {
// TODO(thaJeztah) do we want to add more details about the driver here?
return [][2]string{
{"driver-type", string(plugin.SnapshotPlugin)},
}
}
// GetLayerMountID returns the mount ID for a layer
// called from daemon.go Daemon.Shutdown(), and Daemon.Cleanup() (cleanup is actually continerCleanup)
// TODO: needs to be refactored to Unmount (see callers), or removed and replaced with GetLayerByID
func (i *ImageService) GetLayerMountID(cid string) (string, error) {
return "", errdefs.NotImplemented(errors.New("not implemented"))
}
// Cleanup resources before the process is shutdown.
// called from daemon.go Daemon.Shutdown()
func (i *ImageService) Cleanup() error {
return nil
}
// StorageDriver returns the name of the default storage-driver (snapshotter)
// used by the ImageService.
func (i *ImageService) StorageDriver() string {
return i.snapshotter
}
// ReleaseLayer releases a layer allowing it to be removed
// called from delete.go Daemon.cleanupContainer(), and Daemon.containerExport()
func (i *ImageService) ReleaseLayer(rwlayer layer.RWLayer) error {
return errdefs.NotImplemented(errors.New("not implemented"))
}
// LayerDiskUsage returns the number of bytes used by layer stores
// called from disk_usage.go
func (i *ImageService) LayerDiskUsage(ctx context.Context) (int64, error) {
var allLayersSize int64
// TODO(thaJeztah): do we need to take multiple snapshotters into account? See https://github.com/moby/moby/issues/45273
snapshotter := i.client.SnapshotService(i.snapshotter)
snapshotter.Walk(ctx, func(ctx context.Context, info snapshots.Info) error {
usage, err := snapshotter.Usage(ctx, info.Name)
if err != nil {
return err
}
allLayersSize += usage.Size
return nil
})
return allLayersSize, nil
}
// UpdateConfig values
//
// called from reload.go
func (i *ImageService) UpdateConfig(maxDownloads, maxUploads int) {
panic("not implemented")
}
// GetLayerFolders returns the layer folders from an image RootFS.
func (i *ImageService) GetLayerFolders(img *image.Image, rwLayer layer.RWLayer) ([]string, error) {
return nil, errdefs.NotImplemented(errors.New("not implemented"))
}
// GetContainerLayerSize returns the real size & virtual size of the container.
func (i *ImageService) GetContainerLayerSize(ctx context.Context, containerID string) (int64, int64, error) {
ctr := i.containers.Get(containerID)
if ctr == nil {
return 0, 0, nil
}
snapshotter := i.client.SnapshotService(ctr.Driver)
usage, err := snapshotter.Usage(ctx, containerID)
if err != nil {
return 0, 0, err
}
imageManifest, err := getContainerImageManifest(ctr)
if err != nil {
// Best efforts attempt to pick an image.
// We don't have platform information at this point, so we can only
// assume that the platform matches host.
// Otherwise this will give a wrong base image size (different
// platform), but should be close enough.
mfst, err := i.GetImageManifest(ctx, ctr.Config.Image, imagetypes.GetImageOpts{})
if err != nil {
// Log error, don't error out whole operation.
logrus.WithFields(logrus.Fields{
logrus.ErrorKey: err,
"container": containerID,
}).Warn("empty ImageManifest, can't calculate base image size")
return usage.Size, 0, nil
}
imageManifest = *mfst
}
cs := i.client.ContentStore()
imageManifestBytes, err := content.ReadBlob(ctx, cs, imageManifest)
if err != nil {
return 0, 0, err
}
var manifest ocispec.Manifest
if err := json.Unmarshal(imageManifestBytes, &manifest); err != nil {
return 0, 0, err
}
imageConfigBytes, err := content.ReadBlob(ctx, cs, manifest.Config)
if err != nil {
return 0, 0, err
}
var img ocispec.Image
if err := json.Unmarshal(imageConfigBytes, &img); err != nil {
return 0, 0, err
}
sizeCache := make(map[digest.Digest]int64)
snapshotSizeFn := func(d digest.Digest) (int64, error) {
if s, ok := sizeCache[d]; ok {
return s, nil
}
u, err := snapshotter.Usage(ctx, d.String())
if err != nil {
return 0, err
}
sizeCache[d] = u.Size
return u.Size, nil
}
chainIDs := identity.ChainIDs(img.RootFS.DiffIDs)
snapShotSize, err := computeSnapshotSize(chainIDs, snapshotSizeFn)
if err != nil {
return 0, 0, err
}
// TODO(thaJeztah): include content-store size for the image (similar to "GET /images/json")
return usage.Size, usage.Size + snapShotSize, nil
}
// getContainerImageManifest safely dereferences ImageManifest.
// ImageManifest can be nil for containers created with Docker Desktop with old
// containerd image store integration enabled which didn't set this field.
func getContainerImageManifest(ctr *container.Container) (ocispec.Descriptor, error) {
if ctr.ImageManifest == nil {
return ocispec.Descriptor{}, errdefs.InvalidParameter(errors.New("container is missing ImageManifest (probably created on old version), please recreate it"))
}
return *ctr.ImageManifest, nil
}