Merge pull request #45687 from vvoland/volume-mount-subpath

volumes: Implement subpath mount

api/server/router/container/container_routes.go

@@ -628,6 +628,14 @@ func (s *containerRouter) postContainersCreate(ctx context.Context, w http.Respo
 		}
 	}
 
+	if versions.LessThan(version, "1.45") {
+		for _, m := range hostConfig.Mounts {
+			if m.VolumeOptions != nil && m.VolumeOptions.Subpath != "" {
+				return errdefs.InvalidParameter(errors.New("VolumeOptions.Subpath needs API v1.45 or newer"))
+			}
+		}
+	}
+
 	var warnings []string
 	if warn, err := handleMACAddressBC(config, hostConfig, networkingConfig, version); err != nil {
 		return err

api/swagger.yaml

@@ -423,6 +423,10 @@ definitions:
                 type: "object"
                 additionalProperties:
                   type: "string"
+          Subpath:
+            description: "Source path inside the volume. Must be relative without any back traversals."
+            type: "string"
+            example: "dir-inside-volume/subdirectory"
       TmpfsOptions:
         description: "Optional configuration for the `tmpfs` type."
         type: "object"

api/types/mount/mount.go

@@ -96,6 +96,7 @@ type BindOptions struct {
 type VolumeOptions struct {
 	NoCopy       bool              `json:",omitempty"`
 	Labels       map[string]string `json:",omitempty"`
+	Subpath      string            `json:",omitempty"`
 	DriverConfig *Driver           `json:",omitempty"`
 }
 

container/container.go

@@ -514,14 +514,14 @@ func (container *Container) AddMountPointWithVolume(destination string, vol volu
 }
 
 // UnmountVolumes unmounts all volumes
-func (container *Container) UnmountVolumes(volumeEventLog func(name string, action events.Action, attributes map[string]string)) error {
+func (container *Container) UnmountVolumes(ctx context.Context, volumeEventLog func(name string, action events.Action, attributes map[string]string)) error {
 	var errs []string
 	for _, volumeMount := range container.MountPoints {
 		if volumeMount.Volume == nil {
 			continue
 		}
 
-		if err := volumeMount.Cleanup(); err != nil {
+		if err := volumeMount.Cleanup(ctx); err != nil {
 			errs = append(errs, err.Error())
 			continue
 		}

container/container_unix.go

@@ -15,8 +15,6 @@ import (
 	"github.com/docker/docker/api/types/events"
 	mounttypes "github.com/docker/docker/api/types/mount"
 	swarmtypes "github.com/docker/docker/api/types/swarm"
-	"github.com/docker/docker/pkg/stringid"
-	"github.com/docker/docker/volume"
 	volumemounts "github.com/docker/docker/volume/mounts"
 	"github.com/moby/sys/mount"
 	"github.com/opencontainers/selinux/go-selinux/label"
@@ -129,34 +127,11 @@ func (container *Container) NetworkMounts() []Mount {
 }
 
 // CopyImagePathContent copies files in destination to the volume.
-func (container *Container) CopyImagePathContent(v volume.Volume, destination string) error {
-	rootfs, err := container.GetResourcePath(destination)
-	if err != nil {
-		return err
-	}
-
-	if _, err := os.Stat(rootfs); err != nil {
-		if os.IsNotExist(err) {
-			return nil
-		}
-		return err
-	}
-
-	id := stringid.GenerateRandomID()
-	path, err := v.Mount(id)
-	if err != nil {
-		return err
-	}
-
-	defer func() {
-		if err := v.Unmount(id); err != nil {
-			log.G(context.TODO()).Warnf("error while unmounting volume %s: %v", v.Name(), err)
-		}
-	}()
-	if err := label.Relabel(path, container.MountLabel, true); err != nil && !errors.Is(err, syscall.ENOTSUP) {
+func (container *Container) CopyImagePathContent(volumePath, destination string) error {
+	if err := label.Relabel(volumePath, container.MountLabel, true); err != nil && !errors.Is(err, syscall.ENOTSUP) {
 		return err
 	}
-	return copyExistingContents(rootfs, path)
+	return copyExistingContents(destination, volumePath)
 }
 
 // ShmResourcePath returns path to shm
@@ -396,7 +371,7 @@ func (container *Container) DetachAndUnmount(volumeEventLog func(name string, ac
 				Warn("Unable to unmount")
 		}
 	}
-	return container.UnmountVolumes(volumeEventLog)
+	return container.UnmountVolumes(ctx, volumeEventLog)
 }
 
 // ignoreUnsupportedXAttrs ignores errors when extended attributes
@@ -419,9 +394,13 @@ func copyExistingContents(source, destination string) error {
 		return err
 	}
 	if len(dstList) != 0 {
-		// destination is not empty, do not copy
+		log.G(context.TODO()).WithFields(log.Fields{
+			"source":      source,
+			"destination": destination,
+		}).Debug("destination is not empty, do not copy")
 		return nil
 	}
+
 	return fs.CopyDir(destination, source, ignoreUnsupportedXAttrs())
 }
 

container/container_windows.go

@@ -1,6 +1,7 @@
 package container // import "github.com/docker/docker/container"
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -128,7 +129,7 @@ func (container *Container) ConfigMounts() []Mount {
 // On Windows it only delegates to `UnmountVolumes` since there is nothing to
 // force unmount.
 func (container *Container) DetachAndUnmount(volumeEventLog func(name string, action events.Action, attributes map[string]string)) error {
-	return container.UnmountVolumes(volumeEventLog)
+	return container.UnmountVolumes(context.TODO(), volumeEventLog)
 }
 
 // TmpfsMounts returns the list of tmpfs mounts

daemon/container_operations_unix.go

@@ -99,6 +99,45 @@ func (daemon *Daemon) getPIDContainer(id string) (*container.Container, error) {
 	return ctr, nil
 }
 
+// setupContainerDirs sets up base container directories (root, ipc, tmpfs and secrets).
+func (daemon *Daemon) setupContainerDirs(c *container.Container) (_ []container.Mount, err error) {
+	if err := daemon.setupContainerMountsRoot(c); err != nil {
+		return nil, err
+	}
+
+	if err := daemon.setupIPCDirs(c); err != nil {
+		return nil, err
+	}
+
+	if err := daemon.setupSecretDir(c); err != nil {
+		return nil, err
+	}
+	defer func() {
+		if err != nil {
+			daemon.cleanupSecretDir(c)
+		}
+	}()
+
+	var ms []container.Mount
+	if !c.HostConfig.IpcMode.IsPrivate() && !c.HostConfig.IpcMode.IsEmpty() {
+		ms = append(ms, c.IpcMounts()...)
+	}
+
+	tmpfsMounts, err := c.TmpfsMounts()
+	if err != nil {
+		return nil, err
+	}
+	ms = append(ms, tmpfsMounts...)
+
+	secretMounts, err := c.SecretMounts()
+	if err != nil {
+		return nil, err
+	}
+	ms = append(ms, secretMounts...)
+
+	return ms, nil
+}
+
 func (daemon *Daemon) setupIPCDirs(c *container.Container) error {
 	ipcMode := c.HostConfig.IpcMode
 

daemon/containerfs_linux.go

@@ -17,6 +17,7 @@ import (
 
 	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/container"
+	"github.com/docker/docker/internal/compatcontext"
 	"github.com/docker/docker/internal/mounttree"
 	"github.com/docker/docker/internal/unshare"
 	"github.com/docker/docker/pkg/fileutils"
@@ -54,6 +55,8 @@ type containerFSView struct {
 
 // openContainerFS opens a new view of the container's filesystem.
 func (daemon *Daemon) openContainerFS(container *container.Container) (_ *containerFSView, err error) {
+	ctx := context.TODO()
+
 	if err := daemon.Mount(container); err != nil {
 		return nil, err
 	}
@@ -63,13 +66,15 @@ func (daemon *Daemon) openContainerFS(container *container.Container) (_ *contai
 		}
 	}()
 
-	mounts, err := daemon.setupMounts(container)
+	mounts, cleanup, err := daemon.setupMounts(ctx, container)
 	if err != nil {
 		return nil, err
 	}
 	defer func() {
+		ctx := compatcontext.WithoutCancel(ctx)
+		cleanup(ctx)
 		if err != nil {
-			_ = container.UnmountVolumes(daemon.LogVolumeEvent)
+			_ = container.UnmountVolumes(ctx, daemon.LogVolumeEvent)
 		}
 	}()
 
@@ -207,7 +212,7 @@ func (vw *containerFSView) Close() error {
 	runtime.SetFinalizer(vw, nil)
 	close(vw.todo)
 	err := multierror.Append(nil, <-vw.done)
-	err = multierror.Append(err, vw.ctr.UnmountVolumes(vw.d.LogVolumeEvent))
+	err = multierror.Append(err, vw.ctr.UnmountVolumes(context.TODO(), vw.d.LogVolumeEvent))
 	err = multierror.Append(err, vw.d.Unmount(vw.ctr))
 	return err.ErrorOrNil()
 }

daemon/create.go

@@ -222,7 +222,7 @@ func (daemon *Daemon) create(ctx context.Context, daemonCfg *config.Config, opts
 		return nil, err
 	}
 
-	if err := daemon.createContainerOSSpecificSettings(ctr, opts.params.Config, opts.params.HostConfig); err != nil {
+	if err := daemon.createContainerOSSpecificSettings(ctx, ctr, opts.params.Config, opts.params.HostConfig); err != nil {
 		return nil, err
 	}
 

daemon/create_unix.go

@@ -12,13 +12,17 @@ import (
 	containertypes "github.com/docker/docker/api/types/container"
 	mounttypes "github.com/docker/docker/api/types/mount"
 	"github.com/docker/docker/container"
+	"github.com/docker/docker/errdefs"
+	"github.com/docker/docker/internal/compatcontext"
 	"github.com/docker/docker/oci"
+	volumemounts "github.com/docker/docker/volume/mounts"
 	volumeopts "github.com/docker/docker/volume/service/opts"
 	"github.com/opencontainers/selinux/go-selinux/label"
+	"github.com/pkg/errors"
 )
 
 // createContainerOSSpecificSettings performs host-OS specific container create functionality
-func (daemon *Daemon) createContainerOSSpecificSettings(container *container.Container, config *containertypes.Config, hostConfig *containertypes.HostConfig) error {
+func (daemon *Daemon) createContainerOSSpecificSettings(ctx context.Context, container *container.Container, config *containertypes.Config, hostConfig *containertypes.HostConfig) error {
 	if err := daemon.Mount(container); err != nil {
 		return err
 	}
@@ -45,7 +49,7 @@ func (daemon *Daemon) createContainerOSSpecificSettings(container *container.Con
 		// Skip volumes for which we already have something mounted on that
 		// destination because of a --volume-from.
 		if container.HasMountFor(destination) {
-			log.G(context.TODO()).WithField("container", container.ID).WithField("destination", spec).Debug("mountpoint already exists, skipping anonymous volume")
+			log.G(ctx).WithField("container", container.ID).WithField("destination", spec).Debug("mountpoint already exists, skipping anonymous volume")
 			// Not an error, this could easily have come from the image config.
 			continue
 		}
@@ -70,12 +74,12 @@ func (daemon *Daemon) createContainerOSSpecificSettings(container *container.Con
 
 		container.AddMountPointWithVolume(destination, &volumeWrapper{v: v, s: daemon.volumes}, true)
 	}
-	return daemon.populateVolumes(container)
+	return daemon.populateVolumes(ctx, container)
 }
 
 // populateVolumes copies data from the container's rootfs into the volume for non-binds.
 // this is only called when the container is created.
-func (daemon *Daemon) populateVolumes(c *container.Container) error {
+func (daemon *Daemon) populateVolumes(ctx context.Context, c *container.Container) error {
 	for _, mnt := range c.MountPoints {
 		if mnt.Volume == nil {
 			continue
@@ -85,10 +89,41 @@ func (daemon *Daemon) populateVolumes(c *container.Container) error {
 			continue
 		}
 
-		log.G(context.TODO()).Debugf("copying image data from %s:%s, to %s", c.ID, mnt.Destination, mnt.Name)
-		if err := c.CopyImagePathContent(mnt.Volume, mnt.Destination); err != nil {
+		if err := daemon.populateVolume(ctx, c, mnt); err != nil {
 			return err
 		}
 	}
 	return nil
 }
+
+func (daemon *Daemon) populateVolume(ctx context.Context, c *container.Container, mnt *volumemounts.MountPoint) error {
+	ctrDestPath, err := c.GetResourcePath(mnt.Destination)
+	if err != nil {
+		return err
+	}
+
+	if _, err := os.Stat(ctrDestPath); err != nil {
+		if os.IsNotExist(err) {
+			return nil
+		}
+		return err
+	}
+
+	volumePath, cleanup, err := mnt.Setup(ctx, c.MountLabel, daemon.idMapping.RootPair(), nil)
+	if err != nil {
+		if errdefs.IsNotFound(err) {
+			return nil
+		}
+		log.G(ctx).WithError(err).Debugf("can't copy data from %s:%s, to %s", c.ID, mnt.Destination, volumePath)
+		return errors.Wrapf(err, "failed to populate volume")
+	}
+	defer mnt.Cleanup(compatcontext.WithoutCancel(ctx))
+	defer cleanup(compatcontext.WithoutCancel(ctx))
+
+	log.G(ctx).Debugf("copying image data from %s:%s, to %s", c.ID, mnt.Destination, volumePath)
+	if err := c.CopyImagePathContent(volumePath, ctrDestPath); err != nil {
+		return err
+	}
+
+	return nil
+}

daemon/create_windows.go

@@ -11,7 +11,7 @@ import (
 )
 
 // createContainerOSSpecificSettings performs host-OS specific container create functionality
-func (daemon *Daemon) createContainerOSSpecificSettings(container *container.Container, config *containertypes.Config, hostConfig *containertypes.HostConfig) error {
+func (daemon *Daemon) createContainerOSSpecificSettings(ctx context.Context, container *container.Container, config *containertypes.Config, hostConfig *containertypes.HostConfig) error {
 	if containertypes.Isolation.IsDefault(hostConfig.Isolation) {
 		// Make sure the host config has the default daemon isolation if not specified by caller.
 		hostConfig.Isolation = daemon.defaultIsolation
@@ -34,7 +34,7 @@ func (daemon *Daemon) createContainerOSSpecificSettings(container *container.Con
 
 		// Create the volume in the volume driver. If it doesn't exist,
 		// a new one will be created.
-		v, err := daemon.volumes.Create(context.TODO(), "", volumeDriver, volumeopts.WithCreateReference(container.ID))
+		v, err := daemon.volumes.Create(ctx, "", volumeDriver, volumeopts.WithCreateReference(container.ID))
 		if err != nil {
 			return err
 		}

daemon/daemon.go

@@ -466,7 +466,7 @@ func (daemon *Daemon) restore(cfg *configStore) error {
 						ces.ExitCode = 255
 					}
 					c.SetStopped(&ces)
-					daemon.Cleanup(c)
+					daemon.Cleanup(context.TODO(), c)
 					if err := c.CheckpointTo(daemon.containersReplica); err != nil {
 						baseLogger.WithError(err).Error("failed to update stopped container state")
 					}

daemon/monitor.go

@@ -89,7 +89,7 @@ func (daemon *Daemon) handleContainerExit(c *container.Container, e *libcontaine
 		"exitCode":     strconv.Itoa(exitStatus.ExitCode),
 		"execDuration": strconv.Itoa(int(execDuration.Seconds())),
 	}
-	daemon.Cleanup(c)
+	daemon.Cleanup(context.TODO(), c)
 
 	if restart {
 		c.RestartCount++