// InitializeMountNamespace setups up the devices, mount points, and filesystems for use inside a
// new mount namepsace
func InitializeMountNamespace(rootfs, console string, container *libcontainer.Container) error {
var (
err error
flag = syscall.MS_PRIVATE
)
if container.NoPivotRoot {
flag = syscall.MS_SLAVE
}
if err := system.Mount("", "/", "", uintptr(flag|syscall.MS_REC), ""); err != nil {
return fmt.Errorf("mounting / as slave %s", err)
}
if err := system.Mount(rootfs, rootfs, "bind", syscall.MS_BIND|syscall.MS_REC, ""); err != nil {
return fmt.Errorf("mouting %s as bind %s", rootfs, err)
}
if err := mountSystem(rootfs, container); err != nil {
return fmt.Errorf("mount system %s", err)
}
if err := setupBindmounts(rootfs, container.Mounts); err != nil {
return fmt.Errorf("bind mounts %s", err)
}
if err := nodes.CopyN(rootfs, nodes.DefaultNodes); err != nil {
return fmt.Errorf("copy dev nodes %s", err)
}
if restrictionPath := container.Context["restriction_path"]; restrictionPath != "" {
if err := restrict.Restrict(rootfs, restrictionPath); err != nil {
return fmt.Errorf("restrict %s", err)
}
}
if err := SetupPtmx(rootfs, console, container.Context["mount_label"]); err != nil {
return err
}
if err := system.Chdir(rootfs); err != nil {
return fmt.Errorf("chdir into %s %s", rootfs, err)
}
if container.NoPivotRoot {
err = MsMoveRoot(rootfs)
} else {
err = PivotRoot(rootfs)
}
if err != nil {
return err
}
if container.ReadonlyFs {
if err := SetReadonly(); err != nil {
return fmt.Errorf("set readonly %s", err)
}
}
system.Umask(0022)
return nil
}
func init() {
execdriver.RegisterInitFunc(DriverName, func(args *execdriver.InitArgs) error {
if err := setupEnv(args); err != nil {
return err
}
if err := setupHostname(args); err != nil {
return err
}
if err := setupNetworking(args); err != nil {
return err
}
if !args.Privileged {
if err := restrict.Restrict(); err != nil {
return err
}
}
if err := setupCapabilities(args); err != nil {
return err
}
if err := setupWorkingDirectory(args); err != nil {
return err
}
if err := system.CloseFdsFrom(3); err != nil {
return err
}
if err := changeUser(args); err != nil {
return err
}
path, err := exec.LookPath(args.Args[0])
if err != nil {
log.Printf("Unable to locate %v", args.Args[0])
os.Exit(127)
}
if err := syscall.Exec(path, args.Args, os.Environ()); err != nil {
return fmt.Errorf("dockerinit unable to execute %s - %s", path, err)
}
panic("Unreachable")
})
}
// Init is the init process that first runs inside a new namespace to setup mounts, users, networking,
// and other options required for the new container.
func Init(container *libcontainer.Container, uncleanRootfs, consolePath string, syncPipe *SyncPipe, args []string) error {
rootfs, err := utils.ResolveRootfs(uncleanRootfs)
if err != nil {
return err
}
// clear the current processes env and replace it with the environment
// defined on the container
if err := LoadContainerEnvironment(container); err != nil {
return err
}
// We always read this as it is a way to sync with the parent as well
context, err := syncPipe.ReadFromParent()
if err != nil {
syncPipe.Close()
return err
}
syncPipe.Close()
if consolePath != "" {
if err := console.OpenAndDup(consolePath); err != nil {
return err
}
}
if _, err := system.Setsid(); err != nil {
return fmt.Errorf("setsid %s", err)
}
if consolePath != "" {
if err := system.Setctty(); err != nil {
return fmt.Errorf("setctty %s", err)
}
}
if err := setupNetwork(container, context); err != nil {
return fmt.Errorf("setup networking %s", err)
}
label.Init()
if err := mount.InitializeMountNamespace(rootfs, consolePath, container); err != nil {
return fmt.Errorf("setup mount namespace %s", err)
}
if container.Hostname != "" {
if err := system.Sethostname(container.Hostname); err != nil {
return fmt.Errorf("sethostname %s", err)
}
}
runtime.LockOSThread()
if err := apparmor.ApplyProfile(container.Context["apparmor_profile"]); err != nil {
return fmt.Errorf("set apparmor profile %s: %s", container.Context["apparmor_profile"], err)
}
if err := label.SetProcessLabel(container.Context["process_label"]); err != nil {
return fmt.Errorf("set process label %s", err)
}
if container.Context["restrictions"] != "" {
if err := restrict.Restrict("proc", "sys"); err != nil {
return err
}
}
if err := FinalizeNamespace(container); err != nil {
return fmt.Errorf("finalize namespace %s", err)
}
return system.Execv(args[0], args[0:], container.Env)
}
