// InitSpec creates an OCI spec from the plugin's config.
func (p *Plugin) InitSpec(s specs.Spec, libRoot string) (*specs.Spec, error) {
rootfs := filepath.Join(libRoot, p.PluginObj.ID, "rootfs")
s.Root = specs.Root{
Path: rootfs,
Readonly: false, // TODO: all plugins should be readonly? settable in manifest?
}
mounts := append(p.PluginObj.Config.Mounts, types.PluginMount{
Source: &p.RuntimeSourcePath,
Destination: defaultPluginRuntimeDestination,
Type: "bind",
Options: []string{"rbind", "rshared"},
})
for _, mount := range mounts {
m := specs.Mount{
Destination: mount.Destination,
Type: mount.Type,
Options: mount.Options,
}
// TODO: if nil, then it's required and user didn't set it
if mount.Source != nil {
m.Source = *mount.Source
}
if m.Source != "" && m.Type == "bind" {
fi, err := os.Lstat(filepath.Join(rootfs, m.Destination)) // TODO: followsymlinks
if err != nil {
return nil, err
}
if fi.IsDir() {
if err := os.MkdirAll(m.Source, 0700); err != nil {
return nil, err
}
}
}
s.Mounts = append(s.Mounts, m)
}
envs := make([]string, 1, len(p.PluginObj.Config.Env)+1)
envs[0] = "PATH=" + system.DefaultPathEnv
envs = append(envs, p.PluginObj.Config.Env...)
args := append(p.PluginObj.Manifest.Entrypoint, p.PluginObj.Config.Args...)
cwd := p.PluginObj.Manifest.Workdir
if len(cwd) == 0 {
cwd = "/"
}
s.Process = specs.Process{
Terminal: false,
Args: args,
Cwd: cwd,
Env: envs,
}
return &s, nil
}
func (daemon *Daemon) populateCommonSpec(s *specs.Spec, c *container.Container) error {
linkedEnv, err := daemon.setupLinkedContainers(c)
if err != nil {
return err
}
s.Root = specs.Root{
Path: c.BaseFS,
Readonly: c.HostConfig.ReadonlyRootfs,
}
rootUID, rootGID := daemon.GetRemappedUIDGID()
if err := c.SetupWorkingDirectory(rootUID, rootGID); err != nil {
return err
}
cwd := c.Config.WorkingDir
if len(cwd) == 0 {
cwd = "/"
}
s.Process.Args = append([]string{c.Path}, c.Args...)
// only add the custom init if it is specified and the container is running in its
// own private pid namespace. It does not make sense to add if it is running in the
// host namespace or another container's pid namespace where we already have an init
if c.HostConfig.PidMode.IsPrivate() {
if (c.HostConfig.Init != nil && *c.HostConfig.Init) ||
(c.HostConfig.Init == nil && daemon.configStore.Init) {
s.Process.Args = append([]string{"/dev/init", c.Path}, c.Args...)
var path string
if daemon.configStore.InitPath == "" && c.HostConfig.InitPath == "" {
path, err = exec.LookPath(DefaultInitBinary)
if err != nil {
return err
}
}
if daemon.configStore.InitPath != "" {
path = daemon.configStore.InitPath
}
if c.HostConfig.InitPath != "" {
path = c.HostConfig.InitPath
}
s.Mounts = append(s.Mounts, specs.Mount{
Destination: "/dev/init",
Type: "bind",
Source: path,
Options: []string{"bind", "ro"},
})
}
}
s.Process.Cwd = cwd
s.Process.Env = c.CreateDaemonEnvironment(c.Config.Tty, linkedEnv)
s.Process.Terminal = c.Config.Tty
s.Hostname = c.FullHostname()
return nil
}
func setNamespaces(daemon *Daemon, s *specs.Spec, c *container.Container) error {
userNS := false
// user
if c.HostConfig.UsernsMode.IsPrivate() {
uidMap, gidMap := daemon.GetUIDGIDMaps()
if uidMap != nil {
userNS = true
ns := specs.Namespace{Type: "user"}
setNamespace(s, ns)
s.Linux.UIDMappings = specMapping(uidMap)
s.Linux.GIDMappings = specMapping(gidMap)
}
}
// network
if !c.Config.NetworkDisabled {
ns := specs.Namespace{Type: "network"}
parts := strings.SplitN(string(c.HostConfig.NetworkMode), ":", 2)
if parts[0] == "container" {
nc, err := daemon.getNetworkedContainer(c.ID, c.HostConfig.NetworkMode.ConnectedContainer())
if err != nil {
return err
}
ns.Path = fmt.Sprintf("/proc/%d/ns/net", nc.State.GetPID())
if userNS {
// to share a net namespace, they must also share a user namespace
nsUser := specs.Namespace{Type: "user"}
nsUser.Path = fmt.Sprintf("/proc/%d/ns/user", nc.State.GetPID())
setNamespace(s, nsUser)
}
} else if c.HostConfig.NetworkMode.IsHost() {
ns.Path = c.NetworkSettings.SandboxKey
}
setNamespace(s, ns)
}
// ipc
if c.HostConfig.IpcMode.IsContainer() {
ns := specs.Namespace{Type: "ipc"}
ic, err := daemon.getIpcContainer(c)
if err != nil {
return err
}
ns.Path = fmt.Sprintf("/proc/%d/ns/ipc", ic.State.GetPID())
setNamespace(s, ns)
if userNS {
// to share an IPC namespace, they must also share a user namespace
nsUser := specs.Namespace{Type: "user"}
nsUser.Path = fmt.Sprintf("/proc/%d/ns/user", ic.State.GetPID())
setNamespace(s, nsUser)
}
} else if c.HostConfig.IpcMode.IsHost() {
oci.RemoveNamespace(s, specs.NamespaceType("ipc"))
} else {
ns := specs.Namespace{Type: "ipc"}
setNamespace(s, ns)
}
// pid
if c.HostConfig.PidMode.IsContainer() {
ns := specs.Namespace{Type: "pid"}
pc, err := daemon.getPidContainer(c)
if err != nil {
return err
}
ns.Path = fmt.Sprintf("/proc/%d/ns/pid", pc.State.GetPID())
setNamespace(s, ns)
if userNS {
// to share a PID namespace, they must also share a user namespace
nsUser := specs.Namespace{Type: "user"}
nsUser.Path = fmt.Sprintf("/proc/%d/ns/user", pc.State.GetPID())
setNamespace(s, nsUser)
}
} else if c.HostConfig.PidMode.IsHost() {
oci.RemoveNamespace(s, specs.NamespaceType("pid"))
} else {
ns := specs.Namespace{Type: "pid"}
setNamespace(s, ns)
}
// uts
if c.HostConfig.UTSMode.IsHost() {
oci.RemoveNamespace(s, specs.NamespaceType("uts"))
s.Hostname = ""
}
return nil
}
func setMounts(daemon *Daemon, s *specs.Spec, c *container.Container, mounts []container.Mount) error {
userMounts := make(map[string]struct{})
for _, m := range mounts {
userMounts[m.Destination] = struct{}{}
}
// Filter out mounts that are overridden by user supplied mounts
var defaultMounts []specs.Mount
_, mountDev := userMounts["/dev"]
for _, m := range s.Mounts {
if _, ok := userMounts[m.Destination]; !ok {
if mountDev && strings.HasPrefix(m.Destination, "/dev/") {
continue
}
defaultMounts = append(defaultMounts, m)
}
}
s.Mounts = defaultMounts
for _, m := range mounts {
for _, cm := range s.Mounts {
if cm.Destination == m.Destination {
return fmt.Errorf("Duplicate mount point '%s'", m.Destination)
}
}
if m.Source == "tmpfs" {
data := m.Data
options := []string{"noexec", "nosuid", "nodev", string(volume.DefaultPropagationMode)}
if data != "" {
options = append(options, strings.Split(data, ",")...)
}
merged, err := mount.MergeTmpfsOptions(options)
if err != nil {
return err
}
s.Mounts = append(s.Mounts, specs.Mount{Destination: m.Destination, Source: m.Source, Type: "tmpfs", Options: merged})
continue
}
mt := specs.Mount{Destination: m.Destination, Source: m.Source, Type: "bind"}
// Determine property of RootPropagation based on volume
// properties. If a volume is shared, then keep root propagation
// shared. This should work for slave and private volumes too.
//
// For slave volumes, it can be either [r]shared/[r]slave.
//
// For private volumes any root propagation value should work.
pFlag := mountPropagationMap[m.Propagation]
if pFlag == mount.SHARED || pFlag == mount.RSHARED {
if err := ensureShared(m.Source); err != nil {
return err
}
rootpg := mountPropagationMap[s.Linux.RootfsPropagation]
if rootpg != mount.SHARED && rootpg != mount.RSHARED {
s.Linux.RootfsPropagation = mountPropagationReverseMap[mount.SHARED]
}
} else if pFlag == mount.SLAVE || pFlag == mount.RSLAVE {
if err := ensureSharedOrSlave(m.Source); err != nil {
return err
}
rootpg := mountPropagationMap[s.Linux.RootfsPropagation]
if rootpg != mount.SHARED && rootpg != mount.RSHARED && rootpg != mount.SLAVE && rootpg != mount.RSLAVE {
s.Linux.RootfsPropagation = mountPropagationReverseMap[mount.RSLAVE]
}
}
opts := []string{"rbind"}
if !m.Writable {
opts = append(opts, "ro")
}
if pFlag != 0 {
opts = append(opts, mountPropagationReverseMap[pFlag])
}
mt.Options = opts
s.Mounts = append(s.Mounts, mt)
}
if s.Root.Readonly {
for i, m := range s.Mounts {
switch m.Destination {
case "/proc", "/dev/pts", "/dev/mqueue": // /dev is remounted by runc
continue
}
if _, ok := userMounts[m.Destination]; !ok {
if !stringutils.InSlice(m.Options, "ro") {
s.Mounts[i].Options = append(s.Mounts[i].Options, "ro")
}
}
}
}
if c.HostConfig.Privileged {
if !s.Root.Readonly {
// clear readonly for /sys
for i := range s.Mounts {
if s.Mounts[i].Destination == "/sys" {
clearReadOnly(&s.Mounts[i])
}
}
}
s.Linux.ReadonlyPaths = nil
s.Linux.MaskedPaths = nil
}
// TODO: until a kernel/mount solution exists for handling remount in a user namespace,
// we must clear the readonly flag for the cgroups mount (@mrunalp concurs)
if uidMap, _ := daemon.GetUIDGIDMaps(); uidMap != nil || c.HostConfig.Privileged {
for i, m := range s.Mounts {
if m.Type == "cgroup" {
clearReadOnly(&s.Mounts[i])
}
}
}
return nil
}
// InitSpec creates an OCI spec from the plugin's config.
func (p *Plugin) InitSpec(s specs.Spec) (*specs.Spec, error) {
s.Root = specs.Root{
Path: p.Rootfs,
Readonly: false, // TODO: all plugins should be readonly? settable in config?
}
userMounts := make(map[string]struct{}, len(p.PluginObj.Settings.Mounts))
for _, m := range p.PluginObj.Settings.Mounts {
userMounts[m.Destination] = struct{}{}
}
if err := os.MkdirAll(p.runtimeSourcePath, 0755); err != nil {
return nil, err
}
mounts := append(p.PluginObj.Config.Mounts, types.PluginMount{
Source: &p.runtimeSourcePath,
Destination: defaultPluginRuntimeDestination,
Type: "bind",
Options: []string{"rbind", "rshared"},
})
if p.PluginObj.Config.Network.Type != "" {
// TODO: if net == bridge, use libnetwork controller to create a new plugin-specific bridge, bind mount /etc/hosts and /etc/resolv.conf look at the docker code (allocateNetwork, initialize)
if p.PluginObj.Config.Network.Type == "host" {
oci.RemoveNamespace(&s, specs.NamespaceType("network"))
}
etcHosts := "/etc/hosts"
resolvConf := "/etc/resolv.conf"
mounts = append(mounts,
types.PluginMount{
Source: &etcHosts,
Destination: etcHosts,
Type: "bind",
Options: []string{"rbind", "ro"},
},
types.PluginMount{
Source: &resolvConf,
Destination: resolvConf,
Type: "bind",
Options: []string{"rbind", "ro"},
})
}
for _, mnt := range mounts {
m := specs.Mount{
Destination: mnt.Destination,
Type: mnt.Type,
Options: mnt.Options,
}
if mnt.Source == nil {
return nil, errors.New("mount source is not specified")
}
m.Source = *mnt.Source
s.Mounts = append(s.Mounts, m)
}
for i, m := range s.Mounts {
if strings.HasPrefix(m.Destination, "/dev/") {
if _, ok := userMounts[m.Destination]; ok {
s.Mounts = append(s.Mounts[:i], s.Mounts[i+1:]...)
}
}
}
if p.PluginObj.Config.PropagatedMount != "" {
p.PropagatedMount = filepath.Join(p.Rootfs, p.PluginObj.Config.PropagatedMount)
s.Linux.RootfsPropagation = "rshared"
}
if p.PluginObj.Config.Linux.DeviceCreation {
rwm := "rwm"
s.Linux.Resources.Devices = []specs.DeviceCgroup{{Allow: true, Access: &rwm}}
}
for _, dev := range p.PluginObj.Settings.Devices {
path := *dev.Path
d, dPermissions, err := oci.DevicesFromPath(path, path, "rwm")
if err != nil {
return nil, err
}
s.Linux.Devices = append(s.Linux.Devices, d...)
s.Linux.Resources.Devices = append(s.Linux.Resources.Devices, dPermissions...)
}
envs := make([]string, 1, len(p.PluginObj.Settings.Env)+1)
envs[0] = "PATH=" + system.DefaultPathEnv
envs = append(envs, p.PluginObj.Settings.Env...)
args := append(p.PluginObj.Config.Entrypoint, p.PluginObj.Settings.Args...)
cwd := p.PluginObj.Config.Workdir
if len(cwd) == 0 {
cwd = "/"
}
s.Process.Terminal = false
s.Process.Args = args
s.Process.Cwd = cwd
s.Process.Env = envs
s.Process.Capabilities = append(s.Process.Capabilities, p.PluginObj.Config.Linux.Capabilities...)
return &s, nil
}