func (a *Driver) aufsMount(ro []string, rw, target, mountLabel string) (err error) {
defer func() {
if err != nil {
Unmount(target)
}
}()
// Mount options are clipped to page size(4096 bytes). If there are more
// layers then these are remounted individually using append.
offset := 54
if useDirperm() {
offset += len("dirperm1")
}
b := make([]byte, syscall.Getpagesize()-len(mountLabel)-offset) // room for xino & mountLabel
bp := copy(b, fmt.Sprintf("br:%s=rw", rw))
firstMount := true
i := 0
for {
for ; i < len(ro); i++ {
layer := fmt.Sprintf(":%s=ro+wh", ro[i])
if firstMount {
if bp+len(layer) > len(b) {
break
}
bp += copy(b[bp:], layer)
} else {
data := label.FormatMountLabel(fmt.Sprintf("append%s", layer), mountLabel)
if err = mount("none", target, "aufs", syscall.MS_REMOUNT, data); err != nil {
return
}
}
}
if firstMount {
opts := "dio,xino=/dev/shm/aufs.xino"
if useDirperm() {
opts += ",dirperm1"
}
data := label.FormatMountLabel(fmt.Sprintf("%s,%s", string(b[:bp]), opts), mountLabel)
if err = mount("none", target, "aufs", 0, data); err != nil {
return
}
firstMount = false
}
if i == len(ro) {
break
}
}
return
}
// Get returns the mountpoint for the given id after creating the target directories if necessary.
func (d *Driver) Get(id, mountLabel string) (string, error) {
mountpoint := d.mountPath(id)
filesystem := d.zfsPath(id)
options := label.FormatMountLabel("", mountLabel)
logrus.Debugf(`[zfs] mount("%s", "%s", "%s")`, filesystem, mountpoint, options)
rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
if err != nil {
return "", err
}
// Create the target directories if they don't exist
if err := idtools.MkdirAllAs(mountpoint, 0755, rootUID, rootGID); err != nil {
return "", err
}
if err := mount.Mount(filesystem, mountpoint, "zfs", options); err != nil {
return "", fmt.Errorf("error creating zfs mount of %s to %s: %v", filesystem, mountpoint, err)
}
// this could be our first mount after creation of the filesystem, and the root dir may still have root
// permissions instead of the remapped root uid:gid (if user namespaces are enabled):
if err := os.Chown(mountpoint, rootUID, rootGID); err != nil {
return "", fmt.Errorf("error modifying zfs mountpoint (%s) directory ownership: %v", mountpoint, err)
}
return mountpoint, nil
}
func (d *driver) Mount(
device, target, mountOptions, mountLabel string) error {
if d.isNfsDevice(device) {
if err := d.nfsMount(device, target); err != nil {
return err
}
os.MkdirAll(d.volumeMountPath(target), d.fileModeMountPath())
os.Chmod(d.volumeMountPath(target), d.fileModeMountPath())
return nil
}
fsType, err := probeFsType(device)
if err != nil {
return err
}
options := label.FormatMountLabel("", mountLabel)
options = fmt.Sprintf("%s,%s", mountOptions, mountLabel)
if fsType == "xfs" {
options = fmt.Sprintf("%s,nouuid", mountOptions)
}
if err := mount.Mount(device, target, fsType, options); err != nil {
return fmt.Errorf("Couldn't mount directory %s at %s: %s", device, target, err)
}
os.MkdirAll(d.volumeMountPath(target), d.fileModeMountPath())
os.Chmod(d.volumeMountPath(target), d.fileModeMountPath())
return nil
}
// Do the mount operation followed by additional mounts required to take care
// of propagation flags.
func mountPropagate(m *configs.Mount, rootfs string, mountLabel string) error {
var (
dest = m.Destination
data = label.FormatMountLabel(m.Data, mountLabel)
flags = m.Flags
)
if libcontainerUtils.CleanPath(dest) == "/dev" {
flags &= ^syscall.MS_RDONLY
}
copyUp := m.Extensions&configs.EXT_COPYUP == configs.EXT_COPYUP
if !(copyUp || strings.HasPrefix(dest, rootfs)) {
dest = filepath.Join(rootfs, dest)
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(flags), data); err != nil {
return err
}
for _, pflag := range m.PropagationFlags {
if err := syscall.Mount("", dest, "", uintptr(pflag), ""); err != nil {
return err
}
}
return nil
}
// Get creates and mounts the required file system for the given id and returns the mount path.
func (d *Driver) Get(id string, mountLabel string) (string, error) {
// Protect the d.active from concurrent access
d.Lock()
defer d.Unlock()
mount := d.active[id]
if mount != nil {
mount.count++
return mount.path, nil
}
mount = &ActiveMount{count: 1}
dir := d.dir(id)
if _, err := os.Stat(dir); err != nil {
return "", err
}
// If id has a root, just return it
rootDir := path.Join(dir, "root")
if _, err := os.Stat(rootDir); err == nil {
mount.path = rootDir
d.active[id] = mount
return mount.path, nil
}
lowerID, err := ioutil.ReadFile(path.Join(dir, "lower-id"))
if err != nil {
return "", err
}
lowerDir := path.Join(d.dir(string(lowerID)), "root")
upperDir := path.Join(dir, "upper")
workDir := path.Join(dir, "work")
mergedDir := path.Join(dir, "merged")
opts := fmt.Sprintf("lowerdir=%s,upperdir=%s,workdir=%s", lowerDir, upperDir, workDir)
if err := syscall.Mount("overlay", mergedDir, "overlay", 0, label.FormatMountLabel(opts, mountLabel)); err != nil {
return "", fmt.Errorf("error creating overlay mount to %s: %v", mergedDir, err)
}
// chown "workdir/work" to the remapped root UID/GID. Overlay fs inside a
// user namespace requires this to move a directory from lower to upper.
rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
if err != nil {
return "", err
}
if err := os.Chown(path.Join(workDir, "work"), rootUID, rootGID); err != nil {
return "", err
}
mount.path = mergedDir
mount.mounted = true
d.active[id] = mount
return mount.path, nil
}
// Get creates and mounts the required file system for the given id and returns the mount path.
func (d *Driver) Get(id string, mountLabel string) (string, error) {
dir := d.dir(id)
if _, err := os.Stat(dir); err != nil {
return "", err
}
// If id has a root, just return it
rootDir := path.Join(dir, "root")
if _, err := os.Stat(rootDir); err == nil {
d.pathCacheLock.Lock()
d.pathCache[id] = rootDir
d.pathCacheLock.Unlock()
return rootDir, nil
}
lowerID, err := ioutil.ReadFile(path.Join(dir, "lower-id"))
if err != nil {
return "", err
}
lowerDir := path.Join(d.dir(string(lowerID)), "root")
upperDir := path.Join(dir, "upper")
workDir := path.Join(dir, "work")
mergedDir := path.Join(dir, "merged")
opts := fmt.Sprintf("lowerdir=%s,upperdir=%s,workdir=%s", lowerDir, upperDir, workDir)
// if it's mounted already, just return
mounted, err := d.mounted(mergedDir)
if err != nil {
return "", err
}
if mounted {
return mergedDir, nil
}
if err := syscall.Mount("overlay", mergedDir, "overlay", 0, label.FormatMountLabel(opts, mountLabel)); err != nil {
return "", fmt.Errorf("error creating overlay mount to %s: %v", mergedDir, err)
}
// chown "workdir/work" to the remapped root UID/GID. Overlay fs inside a
// user namespace requires this to move a directory from lower to upper.
rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
if err != nil {
return "", err
}
if err := os.Chown(path.Join(workDir, "work"), rootUID, rootGID); err != nil {
return "", err
}
d.pathCacheLock.Lock()
d.pathCache[id] = mergedDir
d.pathCacheLock.Unlock()
return mergedDir, nil
}
func (devices *RbdSet) MountDevice(hash, mountPoint, mountLabel string) error {
info, err := devices.lookupDevice(hash)
if err != nil {
return err
}
info.lock.Lock()
defer info.lock.Unlock()
if info.mountCount > 0 {
if mountPoint != info.mountPath {
return fmt.Errorf("Trying to mount rbd device in multple places (%s, %s)", info.mountPath, info.Device)
}
info.mountCount++
return nil
}
log.Debugf("[rbdset] Mount image %s with device %s to %s", info.Hash, info.Device, info.mountPath)
if err := devices.mapImageToRbdDevice(info); err != nil {
return err
}
var flags uintptr = syscall.MS_MGC_VAL
fstype, err := ProbeFsType(info.Device)
if err != nil {
return err
}
options := ""
if fstype == "xfs" {
// XFS needs nouuid or it can't mount filesystems with the same fs
options = joinMountOptions(options, "nouuid")
}
options = joinMountOptions(options, devices.mountOptions)
options = joinMountOptions(options, label.FormatMountLabel("", mountLabel))
err = syscall.Mount(info.Device, mountPoint, fstype, flags, joinMountOptions("discard", options))
if err != nil && err == syscall.EINVAL {
err = syscall.Mount(info.Device, mountPoint, fstype, flags, options)
}
if err != nil {
return fmt.Errorf("Error mounting '%s' on '%s': %s", info.Device, mountPoint, err)
}
info.mountCount = 1
info.mountPath = mountPoint
return nil
}
// Get creates and mounts the required file system for the given id and returns the mount path.
func (d *Driver) Get(id string, mountLabel string) (s string, err error) {
dir := d.dir(id)
if _, err := os.Stat(dir); err != nil {
return "", err
}
diffDir := path.Join(dir, "diff")
lowers, err := ioutil.ReadFile(path.Join(dir, lowerFile))
if err != nil {
// If no lower, just return diff directory
if os.IsNotExist(err) {
return diffDir, nil
}
return "", err
}
mergedDir := path.Join(dir, "merged")
if count := d.ctr.Increment(mergedDir); count > 1 {
return mergedDir, nil
}
defer func() {
if err != nil {
if c := d.ctr.Decrement(mergedDir); c <= 0 {
syscall.Unmount(mergedDir, 0)
}
}
}()
workDir := path.Join(dir, "work")
opts := fmt.Sprintf("lowerdir=%s,upperdir=%s,workdir=%s", string(lowers), path.Join(id, "diff"), path.Join(id, "work"))
mountLabel = label.FormatMountLabel(opts, mountLabel)
if len(mountLabel) > syscall.Getpagesize() {
return "", fmt.Errorf("cannot mount layer, mount label too large %d", len(mountLabel))
}
if err := mountFrom(d.home, "overlay", path.Join(id, "merged"), "overlay", mountLabel); err != nil {
return "", fmt.Errorf("error creating overlay mount to %s: %v", mergedDir, err)
}
// chown "workdir/work" to the remapped root UID/GID. Overlay fs inside a
// user namespace requires this to move a directory from lower to upper.
rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
if err != nil {
return "", err
}
if err := os.Chown(path.Join(workDir, "work"), rootUID, rootGID); err != nil {
return "", err
}
return mergedDir, nil
}
// Get creates and mounts the required file system for the given id and returns the mount path.
func (d *Driver) Get(id string, mountLabel string) (s string, err error) {
dir := d.dir(id)
if _, err := os.Stat(dir); err != nil {
return "", err
}
// If id has a root, just return it
rootDir := path.Join(dir, "root")
if _, err := os.Stat(rootDir); err == nil {
return rootDir, nil
}
mergedDir := path.Join(dir, "merged")
if count := d.ctr.Increment(mergedDir); count > 1 {
return mergedDir, nil
}
defer func() {
if err != nil {
if c := d.ctr.Decrement(mergedDir); c <= 0 {
syscall.Unmount(mergedDir, 0)
}
}
}()
lowerID, err := ioutil.ReadFile(path.Join(dir, "lower-id"))
if err != nil {
return "", err
}
var (
lowerDir = path.Join(d.dir(string(lowerID)), "root")
upperDir = path.Join(dir, "upper")
workDir = path.Join(dir, "work")
opts = fmt.Sprintf("lowerdir=%s,upperdir=%s,workdir=%s", lowerDir, upperDir, workDir)
)
if err := syscall.Mount("overlay", mergedDir, "overlay", 0, label.FormatMountLabel(opts, mountLabel)); err != nil {
return "", fmt.Errorf("error creating overlay mount to %s: %v", mergedDir, err)
}
// chown "workdir/work" to the remapped root UID/GID. Overlay fs inside a
// user namespace requires this to move a directory from lower to upper.
rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
if err != nil {
return "", err
}
if err := os.Chown(path.Join(workDir, "work"), rootUID, rootGID); err != nil {
return "", err
}
return mergedDir, nil
}
func (d *Driver) Get(id string, mountLabel string) (string, error) {
// Protect the d.active from concurrent access
d.Lock()
defer d.Unlock()
mount := d.active[id]
if mount != nil {
mount.count++
return mount.path, nil
}
mount = &ActiveMount{count: 1}
dir := d.dir(id)
if _, err := os.Stat(dir); err != nil {
return "", err
}
// If id has a root, just return it
rootDir := path.Join(dir, "root")
if _, err := os.Stat(rootDir); err == nil {
mount.path = rootDir
d.active[id] = mount
return mount.path, nil
}
lowerId, err := ioutil.ReadFile(path.Join(dir, "lower-id"))
if err != nil {
return "", err
}
lowerDir := path.Join(d.dir(string(lowerId)), "root")
upperDir := path.Join(dir, "upper")
workDir := path.Join(dir, "work")
mergedDir := path.Join(dir, "merged")
opts := fmt.Sprintf("lowerdir=%s,upperdir=%s,workdir=%s", lowerDir, upperDir, workDir)
if err := syscall.Mount("overlay", mergedDir, "overlay", 0, label.FormatMountLabel(opts, mountLabel)); err != nil {
return "", fmt.Errorf("error creating overlay mount to %s: %v", mergedDir, err)
}
mount.path = mergedDir
mount.mounted = true
d.active[id] = mount
return mount.path, nil
}
// Get returns the mountpoint for the given id after creating the target directories if necessary.
func (d *Driver) Get(id, mountLabel string) (string, error) {
mountpoint := d.MountPath(id)
filesystem := d.ZfsPath(id)
options := label.FormatMountLabel("", mountLabel)
logrus.Debugf(`[zfs] mount("%s", "%s", "%s")`, filesystem, mountpoint, options)
// Create the target directories if they don't exist
if err := os.MkdirAll(mountpoint, 0755); err != nil {
return "", err
}
err := mount.Mount(filesystem, mountpoint, "zfs", options)
if err != nil {
return "", fmt.Errorf("error creating zfs mount of %s to %s: %v", filesystem, mountpoint, err)
}
return mountpoint, nil
}
func (driver *Driver) Mount(device, target, mountOptions, mountLabel string) error {
fsType, err := probeFsType(device)
if err != nil {
return err
}
options := label.FormatMountLabel("", mountLabel)
options = fmt.Sprintf("%s,%s", mountOptions, mountLabel)
if fsType == "xfs" {
options = fmt.Sprintf("%s,nouuid", mountOptions)
}
if err := mount.Mount(device, target, fsType, options); err != nil {
return fmt.Errorf("Couldn't mount directory %s at %s: %s", device, target, err)
}
return nil
}
// Do the mount operation followed by additional mounts required to take care
// of propagation flags.
func mountPropagate(m *configs.Mount, rootfs string, mountLabel string) error {
var (
dest = m.Destination
data = label.FormatMountLabel(m.Data, mountLabel)
)
if !strings.HasPrefix(dest, rootfs) {
dest = filepath.Join(rootfs, dest)
}
if err := syscall.Mount(m.Source, dest, m.Device, uintptr(m.Flags), data); err != nil {
return err
}
for _, pflag := range m.PropagationFlags {
if err := syscall.Mount("", dest, "", uintptr(pflag), ""); err != nil {
return err
}
}
return nil
}
// Get returns the mountpoint for the given id after creating the target directories if necessary.
func (d *Driver) Get(id, mountLabel string) (string, error) {
mountpoint := d.mountPath(id)
filesystem := d.zfsPath(id)
options := label.FormatMountLabel("", mountLabel)
logrus.Debugf(`[zfs] mount("%s", "%s", "%s")`, filesystem, mountpoint, options)
rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
if err != nil {
return "", err
}
// Create the target directories if they don't exist
if err := idtools.MkdirAllAs(mountpoint, 0755, rootUID, rootGID); err != nil {
return "", err
}
err = mount.Mount(filesystem, mountpoint, "zfs", options)
if err != nil {
return "", fmt.Errorf("error creating zfs mount of %s to %s: %v", filesystem, mountpoint, err)
}
return mountpoint, nil
}
func (daemon *Daemon) setupIpcDirs(container *Container) error {
rootUID, rootGID := daemon.GetRemappedUIDGID()
if !container.hasMountFor("/dev/shm") {
shmPath, err := container.shmPath()
if err != nil {
return err
}
if err := idtools.MkdirAllAs(shmPath, 0700, rootUID, rootGID); err != nil {
return err
}
// When ShmSize is 0 or less, the SHM size is set to 64MB.
if container.hostConfig.ShmSize <= 0 {
container.hostConfig.ShmSize = 67108864
}
shmproperty := "mode=1777,size=" + strconv.FormatInt(container.hostConfig.ShmSize, 10)
if err := syscall.Mount("shm", shmPath, "tmpfs", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), label.FormatMountLabel(shmproperty, container.getMountLabel())); err != nil {
return fmt.Errorf("mounting shm tmpfs: %s", err)
}
if err := os.Chown(shmPath, rootUID, rootGID); err != nil {
return err
}
}
if !container.hasMountFor("/dev/mqueue") {
mqueuePath, err := container.mqueuePath()
if err != nil {
return err
}
if err := idtools.MkdirAllAs(mqueuePath, 0700, rootUID, rootGID); err != nil {
return err
}
if err := syscall.Mount("mqueue", mqueuePath, "mqueue", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), ""); err != nil {
return fmt.Errorf("mounting mqueue mqueue : %s", err)
}
if err := os.Chown(mqueuePath, rootUID, rootGID); err != nil {
return err
}
}
return nil
}
func (daemon *Daemon) setupIpcDirs(c *container.Container) error {
rootUID, rootGID := daemon.GetRemappedUIDGID()
if !c.HasMountFor("/dev/shm") {
shmPath, err := c.ShmResourcePath()
if err != nil {
return err
}
if err := idtools.MkdirAllAs(shmPath, 0700, rootUID, rootGID); err != nil {
return err
}
shmSize := container.DefaultSHMSize
if c.HostConfig.ShmSize != 0 {
shmSize = c.HostConfig.ShmSize
}
shmproperty := "mode=1777,size=" + strconv.FormatInt(shmSize, 10)
if err := syscall.Mount("shm", shmPath, "tmpfs", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), label.FormatMountLabel(shmproperty, c.GetMountLabel())); err != nil {
return fmt.Errorf("mounting shm tmpfs: %s", err)
}
if err := os.Chown(shmPath, rootUID, rootGID); err != nil {
return err
}
}
if !c.HasMountFor("/dev/mqueue") {
mqueuePath, err := c.MqueueResourcePath()
if err != nil {
return err
}
if err := idtools.MkdirAllAs(mqueuePath, 0700, rootUID, rootGID); err != nil {
return err
}
if err := syscall.Mount("mqueue", mqueuePath, "mqueue", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), ""); err != nil {
return fmt.Errorf("mounting mqueue mqueue : %s", 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":
binds, err := getCgroupMounts(m)
if err != nil {
return err
}
var merged []string
for _, b := range binds {
ss := filepath.Base(b.Destination)
if strings.Contains(ss, ",") {
merged = append(merged, ss)
}
}
tmpfs := &configs.Mount{
Source: "tmpfs",
Device: "tmpfs",
Destination: m.Destination,
Flags: defaultMountFlags,
}
if err := mountToRootfs(tmpfs, rootfs, mountLabel); err != nil {
return err
}
for _, b := range binds {
if err := mountToRootfs(b, rootfs, mountLabel); err != nil {
return err
}
}
// create symlinks for merged cgroups
cwd, err := os.Getwd()
if err != nil {
return err
}
if err := os.Chdir(filepath.Join(rootfs, m.Destination)); err != nil {
return err
}
for _, mc := range merged {
for _, ss := range strings.Split(mc, ",") {
if err := os.Symlink(mc, ss); err != nil {
// if cgroup already exists, then okay(it could have been created before)
if os.IsExist(err) {
continue
}
os.Chdir(cwd)
return err
}
}
}
if err := os.Chdir(cwd); err != nil {
return err
}
default:
return fmt.Errorf("unknown mount device %q to %q", m.Device, m.Destination)
}
return nil
}
// Get creates and mounts the required file system for the given id and returns the mount path.
func (d *Driver) Get(id string, mountLabel string) (s string, err error) {
dir := d.dir(id)
if _, err := os.Stat(dir); err != nil {
return "", err
}
diffDir := path.Join(dir, "diff")
lowers, err := ioutil.ReadFile(path.Join(dir, lowerFile))
if err != nil {
// If no lower, just return diff directory
if os.IsNotExist(err) {
return diffDir, nil
}
return "", err
}
mergedDir := path.Join(dir, "merged")
if count := d.ctr.Increment(mergedDir); count > 1 {
return mergedDir, nil
}
defer func() {
if err != nil {
if c := d.ctr.Decrement(mergedDir); c <= 0 {
syscall.Unmount(mergedDir, 0)
}
}
}()
workDir := path.Join(dir, "work")
splitLowers := strings.Split(string(lowers), ":")
absLowers := make([]string, len(splitLowers))
for i, s := range splitLowers {
absLowers[i] = path.Join(d.home, s)
}
opts := fmt.Sprintf("lowerdir=%s,upperdir=%s,workdir=%s", strings.Join(absLowers, ":"), path.Join(dir, "diff"), path.Join(dir, "work"))
mountData := label.FormatMountLabel(opts, mountLabel)
mount := syscall.Mount
mountTarget := mergedDir
pageSize := syscall.Getpagesize()
// Go can return a larger page size than supported by the system
// as of go 1.7. This will be fixed in 1.8 and this block can be
// removed when building with 1.8.
// See https://github.com/golang/go/commit/1b9499b06989d2831e5b156161d6c07642926ee1
// See https://github.com/docker/docker/issues/27384
if pageSize > 4096 {
pageSize = 4096
}
// Use relative paths and mountFrom when the mount data has exceeded
// the page size. The mount syscall fails if the mount data cannot
// fit within a page and relative links make the mount data much
// smaller at the expense of requiring a fork exec to chroot.
if len(mountData) > pageSize {
opts = fmt.Sprintf("lowerdir=%s,upperdir=%s,workdir=%s", string(lowers), path.Join(id, "diff"), path.Join(id, "work"))
mountData = label.FormatMountLabel(opts, mountLabel)
if len(mountData) > pageSize {
return "", fmt.Errorf("cannot mount layer, mount label too large %d", len(mountData))
}
mount = func(source string, target string, mType string, flags uintptr, label string) error {
return mountFrom(d.home, source, target, mType, flags, label)
}
mountTarget = path.Join(id, "merged")
}
if err := mount("overlay", mountTarget, "overlay", 0, mountData); err != nil {
return "", fmt.Errorf("error creating overlay mount to %s: %v", mergedDir, err)
}
// chown "workdir/work" to the remapped root UID/GID. Overlay fs inside a
// user namespace requires this to move a directory from lower to upper.
rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
if err != nil {
return "", err
}
if err := os.Chown(path.Join(workDir, "work"), rootUID, rootGID); err != nil {
return "", err
}
return mergedDir, nil
}
func (container *Container) setupIpcDirs() error {
rootUID, rootGID := container.daemon.GetRemappedUIDGID()
if !container.hasMountFor("/dev/shm") {
shmPath, err := container.shmPath()
if err != nil {
return err
}
if err := idtools.MkdirAllAs(shmPath, 0700, rootUID, rootGID); err != nil {
return err
}
if err := syscall.Mount("shm", shmPath, "tmpfs", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), label.FormatMountLabel("mode=1777,size=65536k", container.getMountLabel())); err != nil {
return fmt.Errorf("mounting shm tmpfs: %s", err)
}
if err := os.Chown(shmPath, rootUID, rootGID); err != nil {
return err
}
}
if !container.hasMountFor("/dev/mqueue") {
mqueuePath, err := container.mqueuePath()
if err != nil {
return err
}
if err := idtools.MkdirAllAs(mqueuePath, 0700, rootUID, rootGID); err != nil {
return err
}
if err := syscall.Mount("mqueue", mqueuePath, "mqueue", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), ""); err != nil {
return fmt.Errorf("mounting mqueue mqueue : %s", err)
}
if err := os.Chown(mqueuePath, rootUID, rootGID); err != nil {
return err
}
}
return nil
}
func (daemon *Daemon) setupIpcDirs(c *container.Container) error {
var err error
c.ShmPath, err = c.ShmResourcePath()
if err != nil {
return err
}
if c.HostConfig.IpcMode.IsContainer() {
ic, err := daemon.getIpcContainer(c)
if err != nil {
return err
}
c.ShmPath = ic.ShmPath
} else if c.HostConfig.IpcMode.IsHost() {
if _, err := os.Stat("/dev/shm"); err != nil {
return fmt.Errorf("/dev/shm is not mounted, but must be for --ipc=host")
}
c.ShmPath = "/dev/shm"
} else {
rootUID, rootGID := daemon.GetRemappedUIDGID()
if !c.HasMountFor("/dev/shm") {
shmPath, err := c.ShmResourcePath()
if err != nil {
return err
}
if err := idtools.MkdirAllAs(shmPath, 0700, rootUID, rootGID); err != nil {
return err
}
shmSize := container.DefaultSHMSize
if c.HostConfig.ShmSize != 0 {
shmSize = c.HostConfig.ShmSize
}
shmproperty := "mode=1777,size=" + strconv.FormatInt(shmSize, 10)
if err := syscall.Mount("shm", shmPath, "tmpfs", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), label.FormatMountLabel(shmproperty, c.GetMountLabel())); err != nil {
return fmt.Errorf("mounting shm tmpfs: %s", err)
}
if err := os.Chown(shmPath, rootUID, rootGID); err != nil {
return err
}
}
}
return nil
}
func (container *Container) setupIpcDirs() error {
shmPath, err := container.shmPath()
if err != nil {
return err
}
if err := os.MkdirAll(shmPath, 0700); err != nil {
return err
}
if err := syscall.Mount("shm", shmPath, "tmpfs", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), label.FormatMountLabel("mode=1777,size=65536k", container.getMountLabel())); err != nil {
return fmt.Errorf("mounting shm tmpfs: %s", err)
}
mqueuePath, err := container.mqueuePath()
if err != nil {
return err
}
if err := os.MkdirAll(mqueuePath, 0700); err != nil {
return err
}
if err := syscall.Mount("mqueue", mqueuePath, "mqueue", uintptr(syscall.MS_NOEXEC|syscall.MS_NOSUID|syscall.MS_NODEV), ""); err != nil {
return fmt.Errorf("mounting mqueue mqueue : %s", 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",
Source: "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
}