func (m *Mount) bindMount(rootfs, mountLabel string) error {
var (
flags = syscall.MS_BIND | syscall.MS_REC
dest = filepath.Join(rootfs, m.Destination)
)
if !m.Writable {
flags = flags | syscall.MS_RDONLY
}
if m.Slave {
flags = flags | syscall.MS_SLAVE
}
stat, err := os.Stat(m.Source)
if err != nil {
return err
}
// FIXME: (crosbymichael) This does not belong here and should be done a layer above
dest, err = symlink.FollowSymlinkInScope(dest, rootfs)
if err != nil {
return err
}
if err := createIfNotExists(dest, stat.IsDir()); err != nil {
return fmt.Errorf("creating new bind mount target %s", err)
}
if err := syscall.Mount(m.Source, dest, "bind", uintptr(flags), ""); err != nil {
return fmt.Errorf("mounting %s into %s %s", m.Source, dest, err)
}
if !m.Writable {
if err := syscall.Mount(m.Source, dest, "bind", uintptr(flags|syscall.MS_REMOUNT), ""); err != nil {
return fmt.Errorf("remounting %s into %s %s", m.Source, dest, err)
}
}
if m.Relabel != "" {
if err := label.Relabel(m.Source, mountLabel, m.Relabel); err != nil {
return fmt.Errorf("relabeling %s to %s %s", m.Source, mountLabel, err)
}
}
if m.Private {
if err := syscall.Mount("", dest, "none", uintptr(syscall.MS_PRIVATE), ""); err != nil {
return fmt.Errorf("mounting %s private %s", dest, err)
}
}
return nil
}
func setupBindmounts(rootfs string, mountConfig *MountConfig) error {
bindMounts := mountConfig.Mounts
for _, m := range bindMounts.OfType("bind") {
var (
flags = syscall.MS_BIND | syscall.MS_REC
dest = filepath.Join(rootfs, m.Destination)
)
if !m.Writable {
flags = flags | syscall.MS_RDONLY
}
stat, err := os.Stat(m.Source)
if err != nil {
return err
}
dest, err = symlink.FollowSymlinkInScope(dest, rootfs)
if err != nil {
return err
}
if err := createIfNotExists(dest, stat.IsDir()); err != nil {
return fmt.Errorf("Creating new bind-mount target, %s", err)
}
if err := syscall.Mount(m.Source, dest, "bind", uintptr(flags), ""); err != nil {
return fmt.Errorf("mounting %s into %s %s", m.Source, dest, err)
}
if !m.Writable {
if err := syscall.Mount(m.Source, dest, "bind", uintptr(flags|syscall.MS_REMOUNT), ""); err != nil {
return fmt.Errorf("remounting %s into %s %s", m.Source, dest, err)
}
}
if m.Relabel != "" {
if err := label.Relabel(m.Source, mountConfig.MountLabel, m.Relabel); err != nil {
return fmt.Errorf("relabeling %s to %s %s", m.Source, mountConfig.MountLabel, err)
}
}
if m.Private {
if err := syscall.Mount("", dest, "none", uintptr(syscall.MS_PRIVATE), ""); err != nil {
return fmt.Errorf("mounting %s private %s", dest, err)
}
}
}
return nil
}
func (d *Driver) Create(id, parent string) error {
dir := d.dir(id)
if err := os.MkdirAll(path.Dir(dir), 0700); err != nil {
return err
}
if err := os.Mkdir(dir, 0755); err != nil {
return err
}
opts := []string{"level:s0"}
if _, mountLabel, err := label.InitLabels(opts); err == nil {
label.Relabel(dir, mountLabel, "")
}
if parent == "" {
return nil
}
parentDir, err := d.Get(parent, "")
if err != nil {
return fmt.Errorf("%s: %s", parent, err)
}
if err := chrootarchive.CopyWithTar(parentDir, dir); err != nil {
return err
}
return nil
}
func mountToRootfs(m *configs.Mount, rootfs, mountLabel string) error {
var (
dest = m.Destination
data = label.FormatMountLabel(m.Data, mountLabel)
)
if !strings.HasPrefix(dest, rootfs) {
dest = filepath.Join(rootfs, dest)
}
switch m.Device {
case "proc", "sysfs":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), "")
case "mqueue":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), ""); err != nil {
return err
}
return label.SetFileLabel(dest, mountLabel)
case "tmpfs":
stat, err := os.Stat(dest)
if err != nil {
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if stat != nil {
if err = os.Chmod(dest, stat.Mode()); err != nil {
return err
}
}
return nil
case "devpts":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data)
case "bind":
stat, err := os.Stat(m.Source)
if err != nil {
// error out if the source of a bind mount does not exist as we will be
// unable to bind anything to it.
return err
}
// ensure that the destination of the bind mount is resolved of symlinks at mount time because
// any previous mounts can invalidate the next mount's destination.
// this can happen when a user specifies mounts within other mounts to cause breakouts or other
// evil stuff to try to escape the container's rootfs.
if dest, err = symlink.FollowSymlinkInScope(filepath.Join(rootfs, m.Destination), rootfs); err != nil {
return err
}
if err := checkMountDestination(rootfs, dest); err != nil {
return err
}
if err := createIfNotExists(dest, stat.IsDir()); err != nil {
return err
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if m.Flags&syscall.MS_RDONLY != 0 {
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags|syscall.MS_REMOUNT), ""); err != nil {
return err
}
}
if m.Relabel != "" {
if err := label.Relabel(m.Source, mountLabel, m.Relabel); err != nil {
return err
}
}
if m.Flags&syscall.MS_PRIVATE != 0 {
if err := syscall.Mount("", dest, "none", uintptr(syscall.MS_PRIVATE), ""); err != nil {
return err
}
}
case "cgroup":
mounts, err := cgroups.GetCgroupMounts()
if err != nil {
return err
}
var binds []*configs.Mount
for _, mm := range mounts {
dir, err := mm.GetThisCgroupDir()
if err != nil {
return err
}
binds = append(binds, &configs.Mount{
Device: "bind",
Source: filepath.Join(mm.Mountpoint, dir),
Destination: filepath.Join(m.Destination, strings.Join(mm.Subsystems, ",")),
Flags: syscall.MS_BIND | syscall.MS_REC | syscall.MS_RDONLY,
})
}
tmpfs := &configs.Mount{
Device: "tmpfs",
Destination: m.Destination,
Flags: syscall.MS_NOEXEC | syscall.MS_NOSUID | syscall.MS_NODEV,
}
if err := mountToRootfs(tmpfs, rootfs, mountLabel); err != nil {
return err
}
for _, b := range binds {
if err := mountToRootfs(b, rootfs, mountLabel); err != nil {
return err
}
}
default:
return fmt.Errorf("unknown mount device %q to %q", m.Device, m.Destination)
}
return nil
}
// registerMountPoints initializes the container mount points with the configured volumes and bind mounts.
// It follows the next sequence to decide what to mount in each final destination:
//
// 1. Select the previously configured mount points for the containers, if any.
// 2. Select the volumes mounted from another containers. Overrides previously configured mount point destination.
// 3. Select the bind mounts set by the client. Overrides previously configured mount point destinations.
func (daemon *Daemon) registerMountPoints(container *Container, hostConfig *runconfig.HostConfig) error {
binds := map[string]bool{}
mountPoints := map[string]*mountPoint{}
// 1. Read already configured mount points.
for name, point := range container.MountPoints {
mountPoints[name] = point
}
// 2. Read volumes from other containers.
for _, v := range hostConfig.VolumesFrom {
containerID, mode, err := parseVolumesFrom(v)
if err != nil {
return err
}
c, err := daemon.Get(containerID)
if err != nil {
return err
}
for _, m := range c.MountPoints {
cp := m
cp.RW = m.RW && mode != "ro"
if len(m.Source) == 0 {
v, err := createVolume(m.Name, m.Driver)
if err != nil {
return err
}
cp.Volume = v
}
mountPoints[cp.Destination] = cp
}
}
// 3. Read bind mounts
for _, b := range hostConfig.Binds {
// #10618
bind, err := parseBindMount(b, container.MountLabel, container.Config)
if err != nil {
return err
}
if binds[bind.Destination] {
return fmt.Errorf("Duplicate bind mount %s", bind.Destination)
}
if len(bind.Name) > 0 && len(bind.Driver) > 0 {
// create the volume
v, err := createVolume(bind.Name, bind.Driver)
if err != nil {
return err
}
bind.Volume = v
bind.Source = v.Path()
// Since this is just a named volume and not a typical bind, set to shared mode `z`
if bind.Relabel == "" {
bind.Relabel = "z"
}
}
if err := label.Relabel(bind.Source, container.MountLabel, bind.Relabel); err != nil {
return err
}
binds[bind.Destination] = true
mountPoints[bind.Destination] = bind
}
// Keep backwards compatible structures
bcVolumes := map[string]string{}
bcVolumesRW := map[string]bool{}
for _, m := range mountPoints {
if m.BackwardsCompatible() {
bcVolumes[m.Destination] = m.Path()
bcVolumesRW[m.Destination] = m.RW
}
}
container.Lock()
container.MountPoints = mountPoints
container.Volumes = bcVolumes
container.VolumesRW = bcVolumesRW
container.Unlock()
return nil
}
// Create creates a new container from the given configuration with a given name.
func (daemon *Daemon) Create(config *runconfig.Config, hostConfig *runconfig.HostConfig, name string) (*Container, []string, error) {
var (
container *Container
warnings []string
img *image.Image
imgID string
err error
)
if config.Image != "" {
img, err = daemon.repositories.LookupImage(config.Image)
if err != nil {
return nil, nil, err
}
if err = img.CheckDepth(); err != nil {
return nil, nil, err
}
imgID = img.ID
}
if err := daemon.mergeAndVerifyConfig(config, img); err != nil {
return nil, nil, err
}
if !config.NetworkDisabled && daemon.SystemConfig().IPv4ForwardingDisabled {
warnings = append(warnings, "IPv4 forwarding is disabled.")
}
if hostConfig == nil {
hostConfig = &runconfig.HostConfig{}
}
if hostConfig.SecurityOpt == nil {
hostConfig.SecurityOpt, err = daemon.GenerateSecurityOpt(hostConfig.IpcMode, hostConfig.PidMode)
if err != nil {
return nil, nil, err
}
}
if container, err = daemon.newContainer(name, config, imgID); err != nil {
return nil, nil, err
}
if err := daemon.Register(container); err != nil {
return nil, nil, err
}
if err := daemon.createRootfs(container); err != nil {
return nil, nil, err
}
if err := daemon.setHostConfig(container, hostConfig); err != nil {
return nil, nil, err
}
if err := container.Mount(); err != nil {
return nil, nil, err
}
defer container.Unmount()
for spec := range config.Volumes {
var (
name, destination string
parts = strings.Split(spec, ":")
)
switch len(parts) {
case 2:
name, destination = parts[0], filepath.Clean(parts[1])
default:
name = stringid.GenerateRandomID()
destination = filepath.Clean(parts[0])
}
// Skip volumes for which we already have something mounted on that
// destination because of a --volume-from.
if container.isDestinationMounted(destination) {
continue
}
path, err := container.GetResourcePath(destination)
if err != nil {
return nil, nil, err
}
stat, err := os.Stat(path)
if err == nil && !stat.IsDir() {
return nil, nil, fmt.Errorf("cannot mount volume over existing file, file exists %s", path)
}
v, err := createVolume(name, config.VolumeDriver)
if err != nil {
return nil, nil, err
}
if err := label.Relabel(v.Path(), container.MountLabel, "z"); err != nil {
return nil, nil, err
}
if err := container.copyImagePathContent(v, destination); err != nil {
return nil, nil, err
}
container.addMountPointWithVolume(destination, v, true)
}
if err := container.ToDisk(); err != nil {
return nil, nil, err
}
return container, warnings, nil
}
func mountToRootfs(m *configs.Mount, rootfs, mountLabel string) error {
var (
dest = m.Destination
data = label.FormatMountLabel(m.Data, mountLabel)
)
if !strings.HasPrefix(dest, rootfs) {
dest = filepath.Join(rootfs, dest)
}
switch m.Device {
case "proc", "mqueue", "sysfs":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), "")
case "tmpfs":
stat, err := os.Stat(dest)
if err != nil {
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if stat != nil {
if err = os.Chmod(dest, stat.Mode()); err != nil {
return err
}
}
return nil
case "devpts":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data)
case "bind":
stat, err := os.Stat(m.Source)
if err != nil {
// error out if the source of a bind mount does not exist as we will be
// unable to bind anything to it.
return err
}
if err := createIfNotExists(dest, stat.IsDir()); err != nil {
return err
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if m.Flags&syscall.MS_RDONLY != 0 {
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags|syscall.MS_REMOUNT), ""); err != nil {
return err
}
}
if m.Relabel != "" {
if err := label.Relabel(m.Source, mountLabel, m.Relabel); err != nil {
return err
}
}
if m.Flags&syscall.MS_PRIVATE != 0 {
if err := syscall.Mount("", dest, "none", uintptr(syscall.MS_PRIVATE), ""); err != nil {
return err
}
}
default:
return fmt.Errorf("unknown mount device %q to %q", m.Device, m.Destination)
}
return nil
}
func mountToRootfs(m *configs.Mount, rootfs, mountLabel string) error {
var (
dest = m.Destination
data = label.FormatMountLabel(m.Data, mountLabel)
)
if !strings.HasPrefix(dest, rootfs) {
dest = filepath.Join(rootfs, dest)
}
switch m.Device {
case "proc", "mqueue", "sysfs":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), "")
case "tmpfs":
stat, err := os.Stat(dest)
if err != nil {
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if stat != nil {
if err = os.Chmod(dest, stat.Mode()); err != nil {
return err
}
}
return nil
case "devpts":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data)
case "bind":
stat, err := os.Stat(m.Source)
if err != nil {
// error out if the source of a bind mount does not exist as we will be
// unable to bind anything to it.
return err
}
// ensure that the destination of the bind mount is resolved of symlinks at mount time because
// any previous mounts can invalidate the next mount's destination.
// this can happen when a user specifies mounts within other mounts to cause breakouts or other
// evil stuff to try to escape the container's rootfs.
if dest, err = symlink.FollowSymlinkInScope(filepath.Join(rootfs, m.Destination), rootfs); err != nil {
return err
}
if err := checkMountDestination(rootfs, dest); err != nil {
return err
}
if err := createIfNotExists(dest, stat.IsDir()); err != nil {
return err
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if m.Flags&syscall.MS_RDONLY != 0 {
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags|syscall.MS_REMOUNT), ""); err != nil {
return err
}
}
if m.Relabel != "" {
if err := label.Relabel(m.Source, mountLabel, m.Relabel); err != nil {
return err
}
}
if m.Flags&syscall.MS_PRIVATE != 0 {
if err := syscall.Mount("", dest, "none", uintptr(syscall.MS_PRIVATE), ""); err != nil {
return err
}
}
default:
return fmt.Errorf("unknown mount device %q to %q", m.Device, m.Destination)
}
return nil
}
func mountToRootfs(m *configs.Mount, rootfs, mountLabel string) error {
var (
dest = m.Destination
data = label.FormatMountLabel(m.Data, mountLabel)
)
if !strings.HasPrefix(dest, rootfs) {
dest = filepath.Join(rootfs, dest)
}
switch m.Device {
case "proc", "sysfs":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), "")
case "mqueue":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), ""); err != nil {
return err
}
return label.SetFileLabel(dest, mountLabel)
case "tmpfs":
stat, err := os.Stat(dest)
if err != nil {
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if stat != nil {
if err = os.Chmod(dest, stat.Mode()); err != nil {
return err
}
}
return nil
case "devpts":
if err := os.MkdirAll(dest, 0755); err != nil && !os.IsExist(err) {
return err
}
return syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data)
case "bind":
stat, err := os.Stat(m.Source)
if err != nil {
// error out if the source of a bind mount does not exist as we will be
// unable to bind anything to it.
return err
}
if err := createIfNotExists(dest, stat.IsDir()); err != nil {
return err
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
if m.Flags&syscall.MS_RDONLY != 0 {
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags|syscall.MS_REMOUNT), ""); err != nil {
return err
}
}
if m.Relabel != "" {
if err := label.Relabel(m.Source, mountLabel, m.Relabel); err != nil {
return err
}
}
if m.Flags&syscall.MS_PRIVATE != 0 {
if err := syscall.Mount("", dest, "none", uintptr(syscall.MS_PRIVATE), ""); err != nil {
return err
}
}
case "cgroup":
mounts, err := cgroups.GetCgroupMounts()
if err != nil {
return err
}
var binds []*configs.Mount
for _, mm := range mounts {
dir, err := mm.GetThisCgroupDir()
if err != nil {
return err
}
binds = append(binds, &configs.Mount{
Device: "bind",
Source: filepath.Join(mm.Mountpoint, dir),
Destination: filepath.Join(m.Destination, strings.Join(mm.Subsystems, ",")),
Flags: syscall.MS_BIND | syscall.MS_REC | syscall.MS_RDONLY,
})
}
tmpfs := &configs.Mount{
Device: "tmpfs",
Destination: m.Destination,
Flags: syscall.MS_NOEXEC | syscall.MS_NOSUID | syscall.MS_NODEV,
}
if err := mountToRootfs(tmpfs, rootfs, mountLabel); err != nil {
return err
}
for _, b := range binds {
if err := mountToRootfs(b, rootfs, mountLabel); err != nil {
return err
}
}
default:
return fmt.Errorf("unknown mount device %q to %q", m.Device, m.Destination)
}
return nil
}