General cleanup of libcontainer

Docker-DCO-1.1-Signed-off-by: Michael Crosby <michael@crosbymichael.com> (github: crosbymichael)

libcontainer/errors.go

@@ -1,9 +0,0 @@
-package libcontainer
-
-import (
-	"errors"
-)
-
-var (
-	ErrInvalidPid = errors.New("no ns pid found")
-)

libcontainer/network/network.go

@@ -1,15 +1,10 @@
 package network
 
 import (
-	"errors"
 	"github.com/dotcloud/docker/pkg/netlink"
 	"net"
 )
 
-var (
-	ErrNoDefaultRoute = errors.New("no default network route found")
-)
-
 func InterfaceUp(name string) error {
 	iface, err := net.InterfaceByName(name)
 	if err != nil {
@@ -46,14 +41,6 @@ func SetInterfaceInNamespacePid(name string, nsPid int) error {
 	return netlink.NetworkSetNsPid(iface, nsPid)
 }
 
-func SetInterfaceInNamespaceFd(name string, fd int) error {
-	iface, err := net.InterfaceByName(name)
-	if err != nil {
-		return err
-	}
-	return netlink.NetworkSetNsFd(iface, fd)
-}
-
 func SetInterfaceMaster(name, master string) error {
 	iface, err := net.InterfaceByName(name)
 	if err != nil {
@@ -89,16 +76,3 @@ func SetMtu(name string, mtu int) error {
 	}
 	return netlink.NetworkSetMTU(iface, mtu)
 }
-
-func GetDefaultMtu() (int, error) {
-	routes, err := netlink.NetworkGetRoutes()
-	if err != nil {
-		return -1, err
-	}
-	for _, r := range routes {
-		if r.Default {
-			return r.Iface.MTU, nil
-		}
-	}
-	return -1, ErrNoDefaultRoute
-}

libcontainer/nsinit/exec.go

@@ -8,65 +8,54 @@ import (
 	"github.com/dotcloud/docker/pkg/system"
 	"github.com/dotcloud/docker/pkg/term"
 	"io"
-	"log"
+	"io/ioutil"
 	"os"
 	"os/exec"
 	"syscall"
 )
 
-func execCommand(container *libcontainer.Container) (pid int, err error) {
+func execCommand(container *libcontainer.Container) (int, error) {
 	master, console, err := createMasterAndConsole()
 	if err != nil {
 		return -1, err
 	}
 
-	// we need CLONE_VFORK so we can wait on the child
+	command := exec.Command("nsinit", "init", console)
-	flag := uintptr(getNamespaceFlags(container.Namespaces) | CLONE_VFORK)
-
-	command := exec.Command("nsinit", console)
 	command.SysProcAttr = &syscall.SysProcAttr{
-		Cloneflags: flag,
+		Cloneflags: uintptr(getNamespaceFlags(container.Namespaces) | syscall.CLONE_VFORK), // we need CLONE_VFORK so we can wait on the child
 	}
 
 	inPipe, err := command.StdinPipe()
 	if err != nil {
 		return -1, err
 	}
-
 	if err := command.Start(); err != nil {
 		return -1, err
 	}
-	pid = command.Process.Pid
+	if err := writePidFile(command); err != nil {
+		return -1, err
+	}
 
 	if container.Network != nil {
 		name1, name2, err := createVethPair()
 		if err != nil {
-			log.Fatal(err)
+			return -1, err
 		}
 		if err := network.SetInterfaceMaster(name1, container.Network.Bridge); err != nil {
-			log.Fatal(err)
+			return -1, err
 		}
 		if err := network.InterfaceUp(name1); err != nil {
-			log.Fatal(err)
+			return -1, err
 		}
-		if err := network.SetInterfaceInNamespacePid(name2, pid); err != nil {
+		if err := network.SetInterfaceInNamespacePid(name2, command.Process.Pid); err != nil {
-			log.Fatal(err)
+			return -1, err
 		}
 		fmt.Fprint(inPipe, name2)
 		inPipe.Close()
 	}
 
-	go func() {
+	go io.Copy(os.Stdout, master)
-		if _, err := io.Copy(os.Stdout, master); err != nil {
+	go io.Copy(master, os.Stdin)
-			log.Println(err)
-		}
-	}()
-
-	go func() {
-		if _, err := io.Copy(master, os.Stdin); err != nil {
-			log.Println(err)
-		}
-	}()
 
 	ws, err := term.GetWinsize(os.Stdin.Fd())
 	if err != nil {
@@ -83,9 +72,11 @@ func execCommand(container *libcontainer.Container) (pid int, err error) {
 	defer term.RestoreTerminal(os.Stdin.Fd(), state)
 
 	if err := command.Wait(); err != nil {
-		return pid, err
+		if _, ok := err.(*exec.ExitError); !ok {
+			return -1, err
 		}
-	return pid, nil
+	}
+	return command.ProcessState.Sys().(syscall.WaitStatus).ExitStatus(), nil
 }
 
 func createMasterAndConsole() (*os.File, string, error) {
@@ -93,12 +84,10 @@ func createMasterAndConsole() (*os.File, string, error) {
 	if err != nil {
 		return nil, "", err
 	}
-
 	console, err := system.Ptsname(master)
 	if err != nil {
 		return nil, "", err
 	}
-
 	if err := system.Unlockpt(master); err != nil {
 		return nil, "", err
 	}
@@ -119,3 +108,7 @@ func createVethPair() (name1 string, name2 string, err error) {
 	}
 	return
 }
+
+func writePidFile(command *exec.Cmd) error {
+	return ioutil.WriteFile(".nspid", []byte(fmt.Sprint(command.Process.Pid)), 0655)
+}

libcontainer/nsinit/init.go

@@ -1,7 +1,6 @@
 package main
 
 import (
-	"encoding/json"
 	"fmt"
 	"github.com/dotcloud/docker/pkg/libcontainer"
 	"github.com/dotcloud/docker/pkg/libcontainer/capabilities"
@@ -14,49 +13,21 @@ import (
 	"syscall"
 )
 
-func loadContainer() (*libcontainer.Container, error) {
+func initCommand(container *libcontainer.Container, console string) error {
-	f, err := os.Open("container.json")
-	if err != nil {
-		return nil, err
-	}
-	defer f.Close()
-
-	var container *libcontainer.Container
-	if err := json.NewDecoder(f).Decode(&container); err != nil {
-		return nil, err
-	}
-	return container, nil
-}
-
-func main() {
-	container, err := loadContainer()
-	if err != nil {
-		log.Fatal(err)
-	}
-
-	if os.Args[1] == "exec" {
-		_, err := execCommand(container)
-		if err != nil {
-			log.Fatal(err)
-		}
-		os.Exit(0)
-	}
-	console := os.Args[1]
-
 	if err := setLogFile(container); err != nil {
-		log.Fatal(err)
+		return err
 	}
 
 	rootfs, err := resolveRootfs()
 	if err != nil {
-		log.Fatal(err)
+		return err
 	}
 
 	var tempVethName string
 	if container.Network != nil {
 		data, err := ioutil.ReadAll(os.Stdin)
 		if err != nil {
-			log.Fatalf("error reading from stdin %s", err)
+			return fmt.Errorf("error reading from stdin %s", err)
 		}
 		tempVethName = string(data)
 	}
@@ -68,48 +39,48 @@ func main() {
 
 	slave, err := openTerminal(console, syscall.O_RDWR)
 	if err != nil {
-		log.Fatalf("open terminal %s", err)
+		return fmt.Errorf("open terminal %s", err)
 	}
 	if slave.Fd() != 0 {
-		log.Fatalf("slave fd should be 0")
+		return fmt.Errorf("slave fd should be 0")
 	}
 	if err := dupSlave(slave); err != nil {
-		log.Fatalf("dup2 slave %s", err)
+		return fmt.Errorf("dup2 slave %s", err)
 	}
 	if _, err := system.Setsid(); err != nil {
-		log.Fatalf("setsid %s", err)
+		return fmt.Errorf("setsid %s", err)
 	}
 	if err := system.Setctty(); err != nil {
-		log.Fatalf("setctty %s", err)
+		return fmt.Errorf("setctty %s", err)
 	}
 	if err := system.ParentDeathSignal(); err != nil {
-		log.Fatalf("parent deth signal %s", err)
+		return fmt.Errorf("parent deth signal %s", err)
 	}
 	if err := setupNewMountNamespace(rootfs, console, container.ReadonlyFs); err != nil {
-		log.Fatalf("setup mount namespace %s", err)
+		return fmt.Errorf("setup mount namespace %s", err)
 	}
 	if container.Network != nil {
 		if err := setupNetworking(container.Network, tempVethName); err != nil {
-			log.Fatalf("setup networking %s", err)
+			return fmt.Errorf("setup networking %s", err)
 		}
 	}
 
 	if err := system.Sethostname(container.ID); err != nil {
-		log.Fatalf("sethostname %s", err)
+		return fmt.Errorf("sethostname %s", err)
 	}
 	if err := capabilities.DropCapabilities(container); err != nil {
-		log.Fatalf("drop capabilities %s", err)
+		return fmt.Errorf("drop capabilities %s", err)
 	}
 	if err := setupUser(container); err != nil {
-		log.Fatalf("setup user %s", err)
+		return fmt.Errorf("setup user %s", err)
 	}
 	if container.WorkingDir != "" {
 		if err := system.Chdir(container.WorkingDir); err != nil {
-			log.Fatalf("chdir to %s %s", container.WorkingDir, err)
+			return fmt.Errorf("chdir to %s %s", container.WorkingDir, err)
 		}
 	}
 	if err := system.Exec(container.Command.Args[0], container.Command.Args[0:], container.Command.Env); err != nil {
-		log.Fatalf("exec %s", err)
+		return fmt.Errorf("exec %s", err)
 	}
 	panic("unreachable")
 }

libcontainer/nsinit/main.go

@@ -0,0 +1,42 @@
+package main
+
+import (
+	"encoding/json"
+	"github.com/dotcloud/docker/pkg/libcontainer"
+	"log"
+	"os"
+)
+
+func main() {
+	container, err := loadContainer()
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	switch os.Args[1] {
+	case "exec":
+		exitCode, err := execCommand(container)
+		if err != nil {
+			log.Fatal(err)
+		}
+		os.Exit(exitCode)
+	case "init":
+		if err := initCommand(container, os.Args[2]); err != nil {
+			log.Fatal(err)
+		}
+	}
+}
+
+func loadContainer() (*libcontainer.Container, error) {
+	f, err := os.Open("container.json")
+	if err != nil {
+		return nil, err
+	}
+	defer f.Close()
+
+	var container *libcontainer.Container
+	if err := json.NewDecoder(f).Decode(&container); err != nil {
+		return nil, err
+	}
+	return container, nil
+}

libcontainer/nsinit/mount.go

@@ -3,68 +3,47 @@ package main
 import (
 	"fmt"
 	"github.com/dotcloud/docker/pkg/system"
-	"log"
 	"os"
 	"path/filepath"
 	"syscall"
 )
 
-var (
 // default mount point options
-	defaults = syscall.MS_NOEXEC | syscall.MS_NOSUID | syscall.MS_NODEV
+const defaultMountFlags = syscall.MS_NOEXEC | syscall.MS_NOSUID | syscall.MS_NODEV
-)
 
 func setupNewMountNamespace(rootfs, console string, readonly bool) error {
 	if err := system.Mount("", "/", "", syscall.MS_SLAVE|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 readonly {
 		if err := system.Mount(rootfs, rootfs, "bind", syscall.MS_BIND|syscall.MS_REMOUNT|syscall.MS_RDONLY|syscall.MS_REC, ""); err != nil {
 			return fmt.Errorf("mounting %s as readonly %s", rootfs, err)
 		}
 	}
-
 	if err := mountSystem(rootfs); err != nil {
 		return fmt.Errorf("mount system %s", err)
 	}
-
 	if err := copyDevNodes(rootfs); err != nil {
 		return fmt.Errorf("copy dev nodes %s", err)
 	}
-
-	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 := setupDev(rootfs); err != nil {
 		return err
 	}
-
+	if err := setupPtmx(rootfs, console); err != nil {
-	if err := setupConsole(rootfs, console); err != nil {
 		return err
 	}
-
 	if err := system.Chdir(rootfs); err != nil {
 		return fmt.Errorf("chdir into %s %s", rootfs, err)
 	}
-
 	if err := system.Mount(rootfs, "/", "", syscall.MS_MOVE, ""); err != nil {
 		return fmt.Errorf("mount move %s into / %s", rootfs, err)
 	}
-
 	if err := system.Chroot("."); err != nil {
 		return fmt.Errorf("chroot . %s", err)
 	}
-
 	if err := system.Chdir("/"); err != nil {
 		return fmt.Errorf("chdir / %s", err)
 	}
@@ -90,13 +69,10 @@ func copyDevNodes(rootfs string) error {
 		if err != nil {
 			return err
 		}
-
 		var (
 			dest = filepath.Join(rootfs, "dev", node)
 			st   = stat.Sys().(*syscall.Stat_t)
 		)
-
-		log.Printf("copy %s to %s %d\n", node, dest, st.Rdev)
 		if err := system.Mknod(dest, st.Mode, int(st.Rdev)); err != nil && !os.IsExist(err) {
 			return fmt.Errorf("copy %s %s", node, err)
 		}
@@ -134,24 +110,22 @@ func setupConsole(rootfs, console string) error {
 	if err != nil {
 		return fmt.Errorf("stat console %s %s", console, err)
 	}
-	st := stat.Sys().(*syscall.Stat_t)
+	var (
-
+		st   = stat.Sys().(*syscall.Stat_t)
-	dest := filepath.Join(rootfs, "dev/console")
+		dest = filepath.Join(rootfs, "dev/console")
+	)
 	if err := os.Remove(dest); err != nil && !os.IsNotExist(err) {
 		return fmt.Errorf("remove %s %s", dest, err)
 	}
-
 	if err := os.Chmod(console, 0600); err != nil {
 		return err
 	}
 	if err := os.Chown(console, 0, 0); err != nil {
 		return err
 	}
-
 	if err := system.Mknod(dest, (st.Mode&^07777)|0600, int(st.Rdev)); err != nil {
 		return fmt.Errorf("mknod %s %s", dest, err)
 	}
-
 	if err := system.Mount(console, dest, "bind", syscall.MS_BIND, ""); err != nil {
 		return fmt.Errorf("bind %s to %s %s", console, dest, err)
 	}
@@ -168,10 +142,10 @@ func mountSystem(rootfs string) error {
 		flags  int
 		data   string
 	}{
-		{source: "proc", path: filepath.Join(rootfs, "proc"), device: "proc", flags: defaults},
+		{source: "proc", path: filepath.Join(rootfs, "proc"), device: "proc", flags: defaultMountFlags},
-		{source: "sysfs", path: filepath.Join(rootfs, "sys"), device: "sysfs", flags: defaults},
+		{source: "sysfs", path: filepath.Join(rootfs, "sys"), device: "sysfs", flags: defaultMountFlags},
 		{source: "tmpfs", path: filepath.Join(rootfs, "dev"), device: "tmpfs", flags: syscall.MS_NOSUID | syscall.MS_STRICTATIME, data: "mode=755"},
-		{source: "shm", path: filepath.Join(rootfs, "dev", "shm"), device: "tmpfs", flags: defaults, data: "mode=1777"},
+		{source: "shm", path: filepath.Join(rootfs, "dev", "shm"), device: "tmpfs", flags: defaultMountFlags, data: "mode=1777"},
 		{source: "devpts", path: filepath.Join(rootfs, "dev", "pts"), device: "devpts", flags: syscall.MS_NOSUID | syscall.MS_NOEXEC, data: "newinstance,ptmxmode=0666,mode=620,gid=5"},
 		{source: "tmpfs", path: filepath.Join(rootfs, "run"), device: "tmpfs", flags: syscall.MS_NOSUID | syscall.MS_NODEV | syscall.MS_STRICTATIME, data: "mode=755"},
 	} {
@@ -189,7 +163,7 @@ func remountProc() error {
 	if err := system.Unmount("/proc", syscall.MNT_DETACH); err != nil {
 		return err
 	}
-	if err := system.Mount("proc", "/proc", "proc", uintptr(defaults), ""); err != nil {
+	if err := system.Mount("proc", "/proc", "proc", uintptr(defaultMountFlags), ""); err != nil {
 		return err
 	}
 	return nil
@@ -201,9 +175,20 @@ func remountSys() error {
 			return err
 		}
 	} else {
-		if err := system.Mount("sysfs", "/sys", "sysfs", uintptr(defaults), ""); err != nil {
+		if err := system.Mount("sysfs", "/sys", "sysfs", uintptr(defaultMountFlags), ""); err != nil {
 			return err
 		}
 	}
 	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)
+}

libcontainer/nsinit/ns_linux.go

@@ -2,27 +2,16 @@ package main
 
 import (
 	"github.com/dotcloud/docker/pkg/libcontainer"
-)
+	"syscall"
-
-const (
-	SIGCHLD       = 0x14
-	CLONE_VFORK   = 0x00004000
-	CLONE_NEWNS   = 0x00020000
-	CLONE_NEWUTS  = 0x04000000
-	CLONE_NEWIPC  = 0x08000000
-	CLONE_NEWUSER = 0x10000000
-	CLONE_NEWPID  = 0x20000000
-	CLONE_NEWNET  = 0x40000000
 )
 
 var namespaceMap = map[libcontainer.Namespace]int{
-	"": 0,
+	libcontainer.CLONE_NEWNS:   syscall.CLONE_NEWNS,
-	libcontainer.CLONE_NEWNS:   CLONE_NEWNS,
+	libcontainer.CLONE_NEWUTS:  syscall.CLONE_NEWUTS,
-	libcontainer.CLONE_NEWUTS:  CLONE_NEWUTS,
+	libcontainer.CLONE_NEWIPC:  syscall.CLONE_NEWIPC,
-	libcontainer.CLONE_NEWIPC:  CLONE_NEWIPC,
+	libcontainer.CLONE_NEWUSER: syscall.CLONE_NEWUSER,
-	libcontainer.CLONE_NEWUSER: CLONE_NEWUSER,
+	libcontainer.CLONE_NEWPID:  syscall.CLONE_NEWPID,
-	libcontainer.CLONE_NEWPID:  CLONE_NEWPID,
+	libcontainer.CLONE_NEWNET:  syscall.CLONE_NEWNET,
-	libcontainer.CLONE_NEWNET:  CLONE_NEWNET,
 }
 
 // getNamespaceFlags parses the container's Namespaces options to set the correct