// +build linux

package mount

import (
	"fmt"
	"github.com/dotcloud/docker/pkg/label"
	"github.com/dotcloud/docker/pkg/libcontainer"
	"github.com/dotcloud/docker/pkg/libcontainer/mount/nodes"
	"github.com/dotcloud/docker/pkg/libcontainer/security/restrict"
	"github.com/dotcloud/docker/pkg/system"
	"os"
	"path/filepath"
	"syscall"
)

// default mount point flags
const defaultMountFlags = syscall.MS_NOEXEC | syscall.MS_NOSUID | syscall.MS_NODEV

type mount struct {
	source string
	path   string
	device string
	flags  int
	data   string
}

// 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
}

// mountSystem sets up linux specific system mounts like sys, proc, shm, and devpts
// inside the mount namespace
func mountSystem(rootfs string, container *libcontainer.Container) error {
	for _, m := range newSystemMounts(rootfs, container.Context["mount_label"], container.Mounts) {
		if err := os.MkdirAll(m.path, 0755); err != nil && !os.IsExist(err) {
			return fmt.Errorf("mkdirall %s %s", m.path, err)
		}
		if err := system.Mount(m.source, m.path, m.device, uintptr(m.flags), m.data); err != nil {
			return fmt.Errorf("mounting %s into %s %s", m.source, m.path, err)
		}
	}
	return nil
}

func setupBindmounts(rootfs string, bindMounts libcontainer.Mounts) error {
	for _, m := range bindMounts.OfType("bind") {
		var (
			flags = syscall.MS_BIND | syscall.MS_REC
			dest  = filepath.Join(rootfs, m.Destination)
		)
		if !m.Writable {
			flags = flags | syscall.MS_RDONLY
		}
		if err := system.Mount(m.Source, dest, "bind", uintptr(flags), ""); err != nil {
			return fmt.Errorf("mounting %s into %s %s", m.Source, dest, err)
		}
		if !m.Writable {
			if err := system.Mount(m.Source, dest, "bind", uintptr(flags|syscall.MS_REMOUNT), ""); err != nil {
				return fmt.Errorf("remounting %s into %s %s", m.Source, dest, err)
			}
		}
		if m.Private {
			if err := system.Mount("", dest, "none", uintptr(syscall.MS_PRIVATE), ""); err != nil {
				return fmt.Errorf("mounting %s private %s", dest, err)
			}
		}
	}
	return nil
}

func newSystemMounts(rootfs, mountLabel string, mounts libcontainer.Mounts) []mount {
	devMounts := []mount{
		{source: "shm", path: filepath.Join(rootfs, "dev", "shm"), device: "tmpfs", flags: defaultMountFlags, data: label.FormatMountLabel("mode=1777,size=65536k", mountLabel)},
		{source: "devpts", path: filepath.Join(rootfs, "dev", "pts"), device: "devpts", flags: syscall.MS_NOSUID | syscall.MS_NOEXEC, data: label.FormatMountLabel("newinstance,ptmxmode=0666,mode=620,gid=5", mountLabel)},
	}

	systemMounts := []mount{
		{source: "proc", path: filepath.Join(rootfs, "proc"), device: "proc", flags: defaultMountFlags},
	}

	if len(mounts.OfType("devtmpfs")) == 1 {
		systemMounts = append(systemMounts, mount{source: "tmpfs", path: filepath.Join(rootfs, "dev"), device: "tmpfs", flags: syscall.MS_NOSUID | syscall.MS_STRICTATIME, data: "mode=755"})
	}
	systemMounts = append(systemMounts, devMounts...)

	if len(mounts.OfType("sysfs")) == 1 {
		systemMounts = append(systemMounts, mount{source: "sysfs", path: filepath.Join(rootfs, "sys"), device: "sysfs", flags: defaultMountFlags})
	}
	return systemMounts
}