// mountContainerCgroups mounts the cgroup controllers hierarchy in the container's
// namespace read-only, leaving the needed knobs in the subcgroup for each-app
// read-write so systemd inside stage1 can apply isolators to them
func mountContainerCgroups(s1Root string, enabledCgroups map[int][]string, subcgroup string, serviceNames []string) error {
if err := cgroup.CreateCgroups(s1Root, enabledCgroups); err != nil {
return fmt.Errorf("error creating container cgroups: %v\n", err)
}
if err := cgroup.RemountCgroupsRO(s1Root, enabledCgroups, subcgroup, serviceNames); err != nil {
return fmt.Errorf("error restricting container cgroups: %v\n", err)
}
return nil
}
// mountContainerCgroups mounts the cgroup controllers hierarchy in the container's
// namespace read-only, leaving the needed knobs in the subcgroup for each-app
// read-write so systemd inside stage1 can apply isolators to them
func mountContainerCgroups(s1Root string, enabledCgroups map[int][]string, subcgroup string, serviceNames []string) error {
if err := cgroup.CreateCgroups(s1Root, enabledCgroups); err != nil {
return errwrap.Wrap(errors.New("error creating container cgroups"), err)
}
if err := cgroup.RemountCgroupsRO(s1Root, enabledCgroups, subcgroup, serviceNames); err != nil {
return errwrap.Wrap(errors.New("error restricting container cgroups"), err)
}
return nil
}
// mountContainerCgroups mounts the cgroup controllers hierarchy in the container's
// namespace read-only, leaving the needed knobs in the subcgroup for each-app
// read-write so systemd inside stage1 can apply isolators to them
func mountContainerCgroups(s1Root string, enabledCgroups map[int][]string, subcgroup string, serviceNames []string) error {
mountContext := os.Getenv(common.EnvSELinuxMountContext)
if err := cgroup.CreateCgroups(s1Root, enabledCgroups, mountContext); err != nil {
return errwrap.Wrap(errors.New("error creating container cgroups"), err)
}
if err := cgroup.RemountCgroupsRO(s1Root, enabledCgroups, subcgroup, serviceNames); err != nil {
return errwrap.Wrap(errors.New("error restricting container cgroups"), err)
}
return nil
}
// mountHostCgroups mounts the host cgroup hierarchy as required by
// systemd-nspawn. We need this because some distributions don't have the
// "name=systemd" cgroup or don't mount the cgroup controllers in
// "/sys/fs/cgroup", and systemd-nspawn needs this. Since this is mounted
// inside the rkt mount namespace, it doesn't affect the host.
func mountHostCgroups(enabledCgroups map[int][]string) error {
systemdControllerPath := "/sys/fs/cgroup/systemd"
if !areHostCgroupsMounted(enabledCgroups) {
if err := cgroup.CreateCgroups("/", enabledCgroups); err != nil {
return fmt.Errorf("error creating host cgroups: %v\n", err)
}
}
if !cgroup.IsControllerMounted("systemd") {
if err := os.MkdirAll(systemdControllerPath, 0700); err != nil {
return err
}
if err := syscall.Mount("cgroup", systemdControllerPath, "cgroup", 0, "none,name=systemd"); err != nil {
return fmt.Errorf("error mounting name=systemd hierarchy on %q: %v", systemdControllerPath, err)
}
}
return nil
}
// mountHostCgroups mounts the host cgroup hierarchy as required by
// systemd-nspawn. We need this because some distributions don't have the
// "name=systemd" cgroup or don't mount the cgroup controllers in
// "/sys/fs/cgroup", and systemd-nspawn needs this. Since this is mounted
// inside the rkt mount namespace, it doesn't affect the host.
func mountHostCgroups(enabledCgroups map[int][]string) error {
systemdControllerPath := "/sys/fs/cgroup/systemd"
if !areHostCgroupsMounted(enabledCgroups) {
mountContext := os.Getenv(common.EnvSELinuxMountContext)
if err := cgroup.CreateCgroups("/", enabledCgroups, mountContext); err != nil {
return errwrap.Wrap(errors.New("error creating host cgroups"), err)
}
}
if !cgroup.IsControllerMounted("systemd") {
if err := os.MkdirAll(systemdControllerPath, 0700); err != nil {
return err
}
if err := syscall.Mount("cgroup", systemdControllerPath, "cgroup", 0, "none,name=systemd"); err != nil {
return errwrap.Wrap(fmt.Errorf("error mounting name=systemd hierarchy on %q", systemdControllerPath), err)
}
}
return nil
}
func stage1() int {
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
fmt.Fprintln(os.Stderr, "UUID is missing or malformed")
return 1
}
root := "."
p, err := LoadPod(root, uuid)
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to load pod: %v\n", err)
return 1
}
// set close-on-exec flag on RKT_LOCK_FD so it gets correctly closed when invoking
// network plugins
lfd, err := common.GetRktLockFD()
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to get rkt lock fd: %v\n", err)
return 1
}
if err := sys.CloseOnExec(lfd, true); err != nil {
fmt.Fprintf(os.Stderr, "Failed to set FD_CLOEXEC on rkt lock: %v\n", err)
return 1
}
mirrorLocalZoneInfo(p.Root)
flavor, _, err := p.getFlavor()
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to get stage1 flavor: %v\n", err)
return 3
}
var n *networking.Networking
if privNet.Any() {
fps, err := forwardedPorts(p)
if err != nil {
fmt.Fprintln(os.Stderr, err.Error())
return 6
}
n, err = networking.Setup(root, p.UUID, fps, privNet, localConfig, flavor)
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to setup network: %v\n", err)
return 6
}
if err = n.Save(); err != nil {
fmt.Fprintf(os.Stderr, "Failed to save networking state %v\n", err)
n.Teardown(flavor)
return 6
}
if len(mdsToken) > 0 {
hostIP, err := n.GetDefaultHostIP()
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to get default Host IP: %v\n", err)
return 6
}
p.MetadataServiceURL = common.MetadataServicePublicURL(hostIP, mdsToken)
}
} else {
if flavor == "kvm" {
fmt.Fprintf(os.Stderr, "Flavor kvm requires private network configuration (try --private-net).\n")
return 6
}
if len(mdsToken) > 0 {
p.MetadataServiceURL = common.MetadataServicePublicURL(localhostIP, mdsToken)
}
}
if err = p.WritePrepareAppTemplate(); err != nil {
fmt.Fprintf(os.Stderr, "Failed to write prepare-app service template: %v\n", err)
return 2
}
if err = p.PodToSystemd(interactive, flavor); err != nil {
fmt.Fprintf(os.Stderr, "Failed to configure systemd: %v\n", err)
return 2
}
args, env, err := getArgsEnv(p, flavor, debug, n)
if err != nil {
fmt.Fprintf(os.Stderr, "%v\n", err)
return 3
}
// create a separate mount namespace so the cgroup filesystems
// are unmounted when exiting the pod
if err := syscall.Unshare(syscall.CLONE_NEWNS); err != nil {
log.Fatalf("error unsharing: %v", err)
}
// we recursively make / a "shared and slave" so mount events from the
// new namespace don't propagate to the host namespace but mount events
// from the host propagate to the new namespace and are forwarded to
// its peer group
// See https://www.kernel.org/doc/Documentation/filesystems/sharedsubtree.txt
if err := syscall.Mount("", "/", "none", syscall.MS_REC|syscall.MS_SLAVE, ""); err != nil {
log.Fatalf("error making / a slave mount: %v", err)
}
if err := syscall.Mount("", "/", "none", syscall.MS_REC|syscall.MS_SHARED, ""); err != nil {
log.Fatalf("error making / a shared and slave mount: %v", err)
}
var serviceNames []string
for _, app := range p.Manifest.Apps {
serviceNames = append(serviceNames, ServiceUnitName(app.Name))
}
s1Root := common.Stage1RootfsPath(p.Root)
machineID := p.GetMachineID()
subcgroup, err := getContainerSubCgroup(machineID)
if err == nil {
if err := cgroup.CreateCgroups(s1Root, subcgroup, serviceNames); err != nil {
fmt.Fprintf(os.Stderr, "Error creating cgroups: %v\n", err)
return 5
}
} else {
fmt.Fprintf(os.Stderr, "Continuing with per-app isolators disabled: %v\n", err)
}
if err = writePpid(os.Getpid()); err != nil {
fmt.Fprintln(os.Stderr, err.Error())
return 4
}
err = withClearedCloExec(lfd, func() error {
return syscall.Exec(args[0], args, env)
})
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to execute %q: %v\n", args[0], err)
return 7
}
return 0
}
func stage1() int {
uuid, err := types.NewUUID(flag.Arg(0))
if err != nil {
fmt.Fprintln(os.Stderr, "UUID is missing or malformed")
return 1
}
root := "."
p, err := LoadPod(root, uuid)
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to load pod: %v\n", err)
return 1
}
// set close-on-exec flag on RKT_LOCK_FD so it gets correctly closed when invoking
// network plugins
lfd, err := common.GetRktLockFD()
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to get rkt lock fd: %v\n", err)
return 1
}
if err := sys.CloseOnExec(lfd, true); err != nil {
fmt.Fprintf(os.Stderr, "Failed to set FD_CLOEXEC on rkt lock: %v\n", err)
return 1
}
mirrorLocalZoneInfo(p.Root)
if privNet.Any() {
fps, err := forwardedPorts(p)
if err != nil {
fmt.Fprintln(os.Stderr, err.Error())
return 6
}
n, err := networking.Setup(root, p.UUID, fps, privNet)
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to setup network: %v\n", err)
return 6
}
defer n.Teardown()
if err = n.Save(); err != nil {
fmt.Fprintf(os.Stderr, "Failed to save networking state %v\n", err)
return 6
}
hostIP, err := n.GetDefaultHostIP()
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to get default Host IP: %v\n", err)
return 6
}
mdsToken, err := generateMDSToken()
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to generate MDS token: %v", err)
return 8
}
p.MetadataServiceURL = common.MetadataServicePublicURL(hostIP, mdsToken)
if err = registerPod(p, mdsToken); err != nil {
fmt.Fprintf(os.Stderr, "Failed to register pod: %v\n", err)
return 8
}
defer unregisterPod(p)
}
flavor, systemdStage1Version, err := p.getFlavor()
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to get stage1 flavor: %v\n", err)
return 3
}
if err = p.WritePrepareAppTemplate(systemdStage1Version); err != nil {
fmt.Fprintf(os.Stderr, "Failed to write prepare-app service template: %v\n", err)
return 2
}
if err = p.PodToSystemd(interactive); err != nil {
fmt.Fprintf(os.Stderr, "Failed to configure systemd: %v\n", err)
return 2
}
args, env, err := getArgsEnv(p, flavor, systemdStage1Version, debug)
if err != nil {
fmt.Fprintf(os.Stderr, "%v\n", err)
return 3
}
appHashes := p.GetAppHashes()
s1Root := common.Stage1RootfsPath(p.Root)
machineID := p.GetMachineID()
subcgroup, err := getContainerSubCgroup(machineID)
if err == nil {
if err := cgroup.CreateCgroups(s1Root, subcgroup, appHashes); err != nil {
fmt.Fprintf(os.Stderr, "Error creating cgroups: %v\n", err)
return 5
}
} else {
fmt.Fprintf(os.Stderr, "Continuing with per-app isolators disabled: %v\n", err)
}
var execFn func() error
if privNet.Any() {
cmd := exec.Cmd{
Path: args[0],
Args: args,
Stdin: os.Stdin,
Stdout: os.Stdout,
Stderr: os.Stderr,
Env: env,
}
execFn = func() error {
err = cmd.Start()
if err != nil {
return fmt.Errorf("Failed to start nspawn: %v\n", err)
}
if err = writePpid(cmd.Process.Pid); err != nil {
return err
}
return cmd.Wait()
}
} else {
if err = writePpid(os.Getpid()); err != nil {
fmt.Fprintln(os.Stderr, err.Error())
return 4
}
execFn = func() error {
return syscall.Exec(args[0], args, env)
}
}
err = withClearedCloExec(lfd, execFn)
if err != nil {
fmt.Fprintf(os.Stderr, "Failed to execute nspawn: %v\n", err)
return 7
}
return 0
}