Add comments to many functions
Docker-DCO-1.1-Signed-off-by: Michael Crosby <michael@crosbymichael.com> (github: crosbymichael)
This commit is contained in:
parent
e0ff0f4dd6
commit
70593be139
10 changed files with 97 additions and 45 deletions
|
@ -41,6 +41,7 @@ func DropCapabilities(container *libcontainer.Container) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
// getCapabilities returns the specific cap values for the libcontainer types
|
||||
func getCapabilities(container *libcontainer.Container) []capability.Cap {
|
||||
drop := []capability.Cap{}
|
||||
for _, c := range container.Capabilities {
|
||||
|
|
|
@ -1,16 +1,22 @@
|
|||
package libcontainer
|
||||
|
||||
// Container defines configuration options for how a
|
||||
// container is setup inside a directory and how a process should be executed
|
||||
type Container struct {
|
||||
Hostname string `json:"hostname,omitempty"`
|
||||
ReadonlyFs bool `json:"readonly_fs,omitempty"`
|
||||
User string `json:"user,omitempty"`
|
||||
WorkingDir string `json:"working_dir,omitempty"`
|
||||
Env []string `json:"environment,omitempty"`
|
||||
Namespaces Namespaces `json:"namespaces,omitempty"`
|
||||
Capabilities Capabilities `json:"capabilities,omitempty"`
|
||||
Network *Network `json:"network,omitempty"`
|
||||
Hostname string `json:"hostname,omitempty"` // hostname
|
||||
ReadonlyFs bool `json:"readonly_fs,omitempty"` // set the containers rootfs as readonly
|
||||
User string `json:"user,omitempty"` // user to execute the process as
|
||||
WorkingDir string `json:"working_dir,omitempty"` // current working directory
|
||||
Env []string `json:"environment,omitempty"` // environment to set
|
||||
Namespaces Namespaces `json:"namespaces,omitempty"` // namespaces to apply
|
||||
Capabilities Capabilities `json:"capabilities,omitempty"` // capabilities to drop
|
||||
Network *Network `json:"network,omitempty"` // nil for host's network stack
|
||||
}
|
||||
|
||||
// Network defines configuration for a container's networking stack
|
||||
//
|
||||
// The network configuration can be omited from a container causing the
|
||||
// container to be setup with the host's networking stack
|
||||
type Network struct {
|
||||
IP string `json:"ip,omitempty"`
|
||||
Gateway string `json:"gateway,omitempty"`
|
||||
|
|
|
@ -38,7 +38,7 @@ func execCommand(container *libcontainer.Container, args []string) (int, error)
|
|||
defer deletePidFile()
|
||||
|
||||
if container.Network != nil {
|
||||
vethPair, err := setupVeth(container.Network.Bridge, command.Process.Pid)
|
||||
vethPair, err := initializeContainerVeth(container.Network.Bridge, command.Process.Pid)
|
||||
if err != nil {
|
||||
return -1, err
|
||||
}
|
||||
|
@ -63,14 +63,21 @@ func execCommand(container *libcontainer.Container, args []string) (int, error)
|
|||
return command.ProcessState.Sys().(syscall.WaitStatus).ExitStatus(), nil
|
||||
}
|
||||
|
||||
// sendVethName writes the veth pair name to the child's stdin then closes the
|
||||
// pipe so that the child stops waiting for more data
|
||||
func sendVethName(name string, pipe io.WriteCloser) {
|
||||
// write the veth pair name to the child's stdin then close the
|
||||
// pipe so that the child stops waiting
|
||||
fmt.Fprint(pipe, name)
|
||||
pipe.Close()
|
||||
}
|
||||
|
||||
func setupVeth(bridge string, nspid int) (string, error) {
|
||||
// initializeContainerVeth will create a veth pair and setup the host's
|
||||
// side of the pair by setting the specified bridge as the master and bringing
|
||||
// up the interface.
|
||||
//
|
||||
// Then will with set the other side of the veth pair into the container's namespaced
|
||||
// using the pid and returns the veth's interface name to provide to the container to
|
||||
// finish setting up the interface inside the namespace
|
||||
func initializeContainerVeth(bridge string, nspid int) (string, error) {
|
||||
name1, name2, err := createVethPair()
|
||||
if err != nil {
|
||||
return "", err
|
||||
|
@ -98,6 +105,8 @@ func setupWindow(master *os.File) (*term.State, error) {
|
|||
return term.SetRawTerminal(os.Stdin.Fd())
|
||||
}
|
||||
|
||||
// createMasterAndConsole will open /dev/ptmx on the host and retreive the
|
||||
// pts name for use as the pty slave inside the container
|
||||
func createMasterAndConsole() (*os.File, string, error) {
|
||||
master, err := os.OpenFile("/dev/ptmx", syscall.O_RDWR|syscall.O_NOCTTY|syscall.O_CLOEXEC, 0)
|
||||
if err != nil {
|
||||
|
@ -113,6 +122,8 @@ func createMasterAndConsole() (*os.File, string, error) {
|
|||
return master, console, nil
|
||||
}
|
||||
|
||||
// createVethPair will automatically generage two random names for
|
||||
// the veth pair and ensure that they have been created
|
||||
func createVethPair() (name1 string, name2 string, err error) {
|
||||
name1, err = utils.GenerateRandomName("dock", 4)
|
||||
if err != nil {
|
||||
|
@ -128,6 +139,7 @@ func createVethPair() (name1 string, name2 string, err error) {
|
|||
return
|
||||
}
|
||||
|
||||
// writePidFile writes the namespaced processes pid to .nspid in the rootfs for the container
|
||||
func writePidFile(command *exec.Cmd) error {
|
||||
return ioutil.WriteFile(".nspid", []byte(fmt.Sprint(command.Process.Pid)), 0655)
|
||||
}
|
||||
|
@ -136,6 +148,9 @@ func deletePidFile() error {
|
|||
return os.Remove(".nspid")
|
||||
}
|
||||
|
||||
// createCommand will return an exec.Cmd with the Cloneflags set to the proper namespaces
|
||||
// defined on the container's configuration and use the current binary as the init with the
|
||||
// args provided
|
||||
func createCommand(container *libcontainer.Container, console string, args []string) *exec.Cmd {
|
||||
command := exec.Command("nsinit", append([]string{"init", console}, args...)...)
|
||||
command.SysProcAttr = &syscall.SysProcAttr{
|
||||
|
|
|
@ -28,6 +28,7 @@ func execinCommand(container *libcontainer.Container, nspid int, args []string)
|
|||
return -1, err
|
||||
}
|
||||
|
||||
// foreach namespace fd, use setns to join an existing container's namespaces
|
||||
for _, fd := range fds {
|
||||
if fd > 0 {
|
||||
if err := system.Setns(fd, 0); err != nil {
|
||||
|
@ -42,7 +43,6 @@ func execinCommand(container *libcontainer.Container, nspid int, args []string)
|
|||
// remount proc and sys to pick up the changes
|
||||
if container.Namespaces.Contains(libcontainer.CLONE_NEWNS) &&
|
||||
container.Namespaces.Contains(libcontainer.CLONE_NEWPID) {
|
||||
|
||||
pid, err := system.Fork()
|
||||
if err != nil {
|
||||
return -1, err
|
||||
|
@ -58,12 +58,7 @@ func execinCommand(container *libcontainer.Container, nspid int, args []string)
|
|||
if err := remountSys(); err != nil {
|
||||
return -1, fmt.Errorf("remount sys %s", err)
|
||||
}
|
||||
if err := capabilities.DropCapabilities(container); err != nil {
|
||||
return -1, fmt.Errorf("drop capabilities %s", err)
|
||||
}
|
||||
if err := system.Exec(args[0], args[0:], container.Env); err != nil {
|
||||
return -1, err
|
||||
}
|
||||
goto dropAndExec
|
||||
}
|
||||
proc, err := os.FindProcess(pid)
|
||||
if err != nil {
|
||||
|
@ -75,6 +70,7 @@ func execinCommand(container *libcontainer.Container, nspid int, args []string)
|
|||
}
|
||||
os.Exit(state.Sys().(syscall.WaitStatus).ExitStatus())
|
||||
}
|
||||
dropAndExec:
|
||||
if err := capabilities.DropCapabilities(container); err != nil {
|
||||
return -1, fmt.Errorf("drop capabilities %s", err)
|
||||
}
|
||||
|
|
|
@ -37,9 +37,6 @@ func initCommand(container *libcontainer.Container, console string, args []strin
|
|||
if err != nil {
|
||||
return fmt.Errorf("open terminal %s", err)
|
||||
}
|
||||
if slave.Fd() != 0 {
|
||||
return fmt.Errorf("slave fd should be 0")
|
||||
}
|
||||
if err := dupSlave(slave); err != nil {
|
||||
return fmt.Errorf("dup2 slave %s", err)
|
||||
}
|
||||
|
@ -55,7 +52,7 @@ func initCommand(container *libcontainer.Container, console string, args []strin
|
|||
if err := setupNewMountNamespace(rootfs, console, container.ReadonlyFs); err != nil {
|
||||
return fmt.Errorf("setup mount namespace %s", err)
|
||||
}
|
||||
if err := setupNetworking(container.Network, tempVethName); err != nil {
|
||||
if err := setupVethNetwork(container.Network, tempVethName); err != nil {
|
||||
return fmt.Errorf("setup networking %s", err)
|
||||
}
|
||||
if err := system.Sethostname(container.Hostname); err != nil {
|
||||
|
@ -78,6 +75,8 @@ func initCommand(container *libcontainer.Container, console string, args []strin
|
|||
panic("unreachable")
|
||||
}
|
||||
|
||||
// resolveRootfs ensures that the current working directory is
|
||||
// not a symlink and returns the absolute path to the rootfs
|
||||
func resolveRootfs() (string, error) {
|
||||
cwd, err := os.Getwd()
|
||||
if err != nil {
|
||||
|
@ -104,8 +103,9 @@ func setupUser(container *libcontainer.Container) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
// dupSlave dup2 the pty slave's fd into stdout and stdin and ensures that
|
||||
// the slave's fd is 0, or stdin
|
||||
func dupSlave(slave *os.File) error {
|
||||
// we close Stdin,etc so our pty slave should have fd 0
|
||||
if slave.Fd() != 0 {
|
||||
return fmt.Errorf("slave fd not 0 %d", slave.Fd())
|
||||
}
|
||||
|
@ -118,7 +118,8 @@ func dupSlave(slave *os.File) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
// openTerminal is a clone of os.OpenFile without the O_CLOEXEC addition.
|
||||
// openTerminal is a clone of os.OpenFile without the O_CLOEXEC
|
||||
// used to open the pty slave inside the container namespace
|
||||
func openTerminal(name string, flag int) (*os.File, error) {
|
||||
r, e := syscall.Open(name, flag, 0)
|
||||
if e != nil {
|
||||
|
@ -127,7 +128,10 @@ func openTerminal(name string, flag int) (*os.File, error) {
|
|||
return os.NewFile(uintptr(r), name), nil
|
||||
}
|
||||
|
||||
func setupNetworking(config *libcontainer.Network, tempVethName string) error {
|
||||
// setupVethNetwork uses the Network config if it is not nil to initialize
|
||||
// the new veth interface inside the container for use by changing the name to eth0
|
||||
// setting the MTU and IP address along with the default gateway
|
||||
func setupVethNetwork(config *libcontainer.Network, tempVethName string) error {
|
||||
if config != nil {
|
||||
if err := network.InterfaceDown(tempVethName); err != nil {
|
||||
return fmt.Errorf("interface down %s %s", tempVethName, err)
|
||||
|
|
|
@ -26,7 +26,7 @@ func main() {
|
|||
log.Fatal(ErrWrongArguments)
|
||||
}
|
||||
switch os.Args[1] {
|
||||
case "exec":
|
||||
case "exec": // this is executed outside of the namespace in the cwd
|
||||
var exitCode int
|
||||
nspid, err := readPid()
|
||||
if err != nil {
|
||||
|
@ -43,7 +43,7 @@ func main() {
|
|||
log.Fatal(err)
|
||||
}
|
||||
os.Exit(exitCode)
|
||||
case "init":
|
||||
case "init": // this is executed inside of the namespace to setup the container
|
||||
if argc < 3 {
|
||||
log.Fatal(ErrWrongArguments)
|
||||
}
|
||||
|
|
|
@ -10,10 +10,16 @@ import (
|
|||
"syscall"
|
||||
)
|
||||
|
||||
// default mount point options
|
||||
// default mount point flags
|
||||
const defaultMountFlags = syscall.MS_NOEXEC | syscall.MS_NOSUID | syscall.MS_NODEV
|
||||
|
||||
// setupNewMountNamespace is used to initialize a new mount namespace for an new
|
||||
// container in the rootfs that is specified.
|
||||
//
|
||||
// There is no need to unmount the new mounts because as soon as the mount namespace
|
||||
// is no longer in use, the mounts will be removed automatically
|
||||
func setupNewMountNamespace(rootfs, console string, readonly bool) error {
|
||||
// mount as slave so that the new mounts do not propagate to the host
|
||||
if err := system.Mount("", "/", "", syscall.MS_SLAVE|syscall.MS_REC, ""); err != nil {
|
||||
return fmt.Errorf("mounting / as slave %s", err)
|
||||
}
|
||||
|
@ -55,6 +61,7 @@ func setupNewMountNamespace(rootfs, console string, readonly bool) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
// copyDevNodes mknods the hosts devices so the new container has access to them
|
||||
func copyDevNodes(rootfs string) error {
|
||||
oldMask := system.Umask(0000)
|
||||
defer system.Umask(oldMask)
|
||||
|
@ -82,6 +89,8 @@ func copyDevNodes(rootfs string) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
// setupDev symlinks the current processes pipes into the
|
||||
// appropriate destination on the containers rootfs
|
||||
func setupDev(rootfs string) error {
|
||||
for _, link := range []struct {
|
||||
from string
|
||||
|
@ -104,6 +113,7 @@ func setupDev(rootfs string) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
// setupConsole ensures that the container has a proper /dev/console setup
|
||||
func setupConsole(rootfs, console string) error {
|
||||
oldMask := system.Umask(0000)
|
||||
defer system.Umask(oldMask)
|
||||
|
@ -161,6 +171,24 @@ func mountSystem(rootfs string) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
// setupPtmx adds a symlink to pts/ptmx for /dev/ptmx and
|
||||
// finishes setting up /dev/console
|
||||
func setupPtmx(rootfs, console string) error {
|
||||
ptmx := filepath.Join(rootfs, "dev/ptmx")
|
||||
if err := os.Remove(ptmx); err != nil && !os.IsNotExist(err) {
|
||||
return err
|
||||
}
|
||||
if err := os.Symlink("pts/ptmx", ptmx); err != nil {
|
||||
return fmt.Errorf("symlink dev ptmx %s", err)
|
||||
}
|
||||
if err := setupConsole(rootfs, console); err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// remountProc is used to detach and remount the proc filesystem
|
||||
// commonly needed with running a new process inside an existing container
|
||||
func remountProc() error {
|
||||
if err := system.Unmount("/proc", syscall.MNT_DETACH); err != nil {
|
||||
return err
|
||||
|
@ -183,14 +211,3 @@ func remountSys() error {
|
|||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func setupPtmx(rootfs, console string) error {
|
||||
ptmx := filepath.Join(rootfs, "dev/ptmx")
|
||||
if err := os.Remove(ptmx); err != nil && !os.IsNotExist(err) {
|
||||
return err
|
||||
}
|
||||
if err := os.Symlink("pts/ptmx", ptmx); err != nil {
|
||||
return fmt.Errorf("symlink dev ptmx %s", err)
|
||||
}
|
||||
return setupConsole(rootfs, console)
|
||||
}
|
||||
|
|
|
@ -14,6 +14,9 @@ var namespaceMap = map[libcontainer.Namespace]int{
|
|||
libcontainer.CLONE_NEWNET: syscall.CLONE_NEWNET,
|
||||
}
|
||||
|
||||
// namespaceFileMap is used to convert the libcontainer types
|
||||
// into the names of the files located in /proc/<pid>/ns/* for
|
||||
// each namespace
|
||||
var namespaceFileMap = map[libcontainer.Namespace]string{
|
||||
libcontainer.CLONE_NEWNS: "mnt",
|
||||
libcontainer.CLONE_NEWUTS: "uts",
|
||||
|
|
|
@ -1,5 +1,8 @@
|
|||
package libcontainer
|
||||
|
||||
// These constants are defined as string types so that
|
||||
// it is clear when adding the configuration in config files
|
||||
// instead of using ints or other types
|
||||
const (
|
||||
CAP_SETPCAP Capability = "SETPCAP"
|
||||
CAP_SYS_MODULE Capability = "SYS_MODULE"
|
||||
|
@ -25,9 +28,15 @@ const (
|
|||
CLONE_NEWNET Namespace = "NEWNET" // network
|
||||
)
|
||||
|
||||
type Namespace string
|
||||
type Namespaces []Namespace
|
||||
type (
|
||||
Namespace string
|
||||
Namespaces []Namespace
|
||||
Capability string
|
||||
Capabilities []Capability
|
||||
)
|
||||
|
||||
// Contains returns true if the specified Namespace is
|
||||
// in the slice
|
||||
func (n Namespaces) Contains(ns Namespace) bool {
|
||||
for _, nns := range n {
|
||||
if nns == ns {
|
||||
|
@ -37,9 +46,8 @@ func (n Namespaces) Contains(ns Namespace) bool {
|
|||
return false
|
||||
}
|
||||
|
||||
type Capability string
|
||||
type Capabilities []Capability
|
||||
|
||||
// Contains returns true if the specified Capability is
|
||||
// in the slice
|
||||
func (c Capabilities) Contains(capp Capability) bool {
|
||||
for _, cc := range c {
|
||||
if cc == capp {
|
||||
|
|
|
@ -6,6 +6,8 @@ import (
|
|||
"io"
|
||||
)
|
||||
|
||||
// GenerateRandomName returns a new name joined with a prefix. This size
|
||||
// specified is used to truncate the randomly generated value
|
||||
func GenerateRandomName(prefix string, size int) (string, error) {
|
||||
id := make([]byte, 32)
|
||||
if _, err := io.ReadFull(rand.Reader, id); err != nil {
|
||||
|
|
Loading…
Reference in a new issue