Open bind mount sources from the host userns

The source of the bind mount might not be accessible in a different user
namespace because a component of the source path might not be traversed
under the users and groups mapped inside the user namespace. This caused
errors such as the following:

  # time="2020-06-22T13:48:26Z" level=error msg="container_linux.go:367:
  starting container process caused: process_linux.go:459:
  container init caused: rootfs_linux.go:58:
  mounting \"/tmp/busyboxtest/source-inaccessible/dir\"
  to rootfs at \"/tmp/inaccessible\" caused:
  stat /tmp/busyboxtest/source-inaccessible/dir: permission denied"

To solve this problem, this patch performs the following:

1. in nsexec.c, it opens the source path in the host userns (so we have
   the right permissions to open it) but in the container mntns (so the
   kernel cross mntns mount check let us mount it later:
   https://github.com/torvalds/linux/blob/v5.8/fs/namespace.c#L2312).

2. in nsexec.c, it passes the file descriptors of the source to the
   child process with SCM_RIGHTS.

3. In runc-init in Golang, it finishes the mounts while inside the
   userns even without access to the some components of the source
   paths.

Passing the fds with SCM_RIGHTS is necessary because once the child
process is in the container mntns, it is already in the container userns
so it cannot temporarily join the host mntns.

This patch uses the existing mechanism with _LIBCONTAINER_* environment
variables to pass the file descriptors from runc to runc init.

This patch uses the existing mechanism with the Netlink-style bootstrap
to pass information about the list of source mounts to nsexec.c.

Signed-off-by: Alban Crequy <[email protected]>

libcontainer/container_linux.go

@@ -548,6 +548,24 @@ func (c *linuxContainer) newInitProcess(p *Process, cmd *exec.Cmd, messageSockPa
 	if err != nil {
 		return nil, err
 	}
+
+	for i, m := range c.config.Mounts {
+		if m.Device != "bind" {
+			continue
+		}
+		// The fd passed here will not be used: nsexec.c will overwrite it with dup2(). We just need
+		// to allocate a fd so that we know the number to pass in the environment variable. The fd
+		// must not be closed before cmd.Start(), so we reuse messageSockPair.child because the
+		// lifecycle of that fd is already taken care of.
+		cmd.ExtraFiles = append(cmd.ExtraFiles, messageSockPair.child)
+		cmd.Env = append(cmd.Env,
+			fmt.Sprintf("_LIBCONTAINER_MOUNT_FILE_%d=%d", i, stdioFdCount+len(cmd.ExtraFiles)-1),
+		)
+	}
+	cmd.Env = append(cmd.Env,
+		fmt.Sprintf("_LIBCONTAINER_MOUNT_FILE_COUNT=%d", len(c.config.Mounts)),
+	)
+
 	init := &initProcess{
 		cmd:             cmd,
 		messageSockPair: messageSockPair,
@@ -1170,7 +1188,9 @@ func (c *linuxContainer) makeCriuRestoreMountpoints(m *configs.Mount) error {
 	case "bind":
 		// The prepareBindMount() function checks if source
 		// exists. So it cannot be used for other filesystem types.
-		if err := prepareBindMount(m, c.config.Rootfs); err != nil {
+		// TODO: pass something else than nil? Not sure if criu is
+		// impacted by issue #2484
+		if err := prepareBindMount(m, c.config.Rootfs, nil); err != nil {
 			return err
 		}
 	default:
@@ -2053,6 +2073,25 @@ func (c *linuxContainer) bootstrapData(cloneFlags uintptr, nsMaps map[configs.Na
 		Value: c.config.RootlessEUID,
 	})
 
+	// bind mount source to open
+	if len(c.config.Mounts) > 0 &&
+		c.config.Namespaces.Contains(configs.NEWUSER) &&
+		c.config.Namespaces.Contains(configs.NEWNS) {
+
+		var mounts []byte
+		for _, m := range c.config.Mounts {
+			if m.Device == "bind" {
+				mounts = append(mounts, []byte(m.Source)...)
+			}
+			mounts = append(mounts, byte(0))
+		}
+
+		r.AddData(&Bytemsg{
+			Type:  MountSourcesAttr,
+			Value: mounts,
+		})
+	}
+
 	return bytes.NewReader(r.Serialize()), nil
 }
 

libcontainer/factory_linux.go

@@ -336,11 +336,13 @@ func (l *LinuxFactory) Type() string {
 // This is a low level implementation detail of the reexec and should not be consumed externally
 func (l *LinuxFactory) StartInitialization() (err error) {
 	var (
-		pipefd, fifofd int
-		consoleSocket  *os.File
-		envInitPipe    = os.Getenv("_LIBCONTAINER_INITPIPE")
-		envFifoFd      = os.Getenv("_LIBCONTAINER_FIFOFD")
-		envConsole     = os.Getenv("_LIBCONTAINER_CONSOLE")
+		pipefd, fifofd    int
+		mountFileCount    int
+		consoleSocket     *os.File
+		envInitPipe       = os.Getenv("_LIBCONTAINER_INITPIPE")
+		envFifoFd         = os.Getenv("_LIBCONTAINER_FIFOFD")
+		envConsole        = os.Getenv("_LIBCONTAINER_CONSOLE")
+		envMountFileCount = os.Getenv("_LIBCONTAINER_MOUNT_FILE_COUNT")
 	)
 
 	// Get the INITPIPE.
@@ -372,6 +374,28 @@ func (l *LinuxFactory) StartInitialization() (err error) {
 		defer consoleSocket.Close()
 	}
 
+	// Get the mount files (O_PATH)
+	mountFileCount, err = strconv.Atoi(envMountFileCount)
+	if err != nil {
+		return fmt.Errorf("unable to convert _LIBCONTAINER_MOUNT_FILE_COUNT=%s to int: %s",
+			envMountFileCount, err)
+	}
+	mountFiles := make([]*os.File, mountFileCount)
+	for i := 0; i < mountFileCount; i++ {
+		envMountFile := os.Getenv(fmt.Sprintf("_LIBCONTAINER_MOUNT_FILE_%d", i))
+		if envMountFile == "" {
+			continue
+		}
+		mountFileFd, err := strconv.Atoi(envMountFile)
+		if err != nil {
+			return fmt.Errorf("unable to convert _LIBCONTAINER_MOUNT_FILE_%d=%s to int: %s",
+				i, envMountFile, err)
+		}
+		mountFile := os.NewFile(uintptr(mountFileFd), "mount-file")
+		defer mountFile.Close()
+		mountFiles[i] = mountFile
+	}
+
 	// clear the current process's environment to clean any libcontainer
 	// specific env vars.
 	os.Clearenv()
@@ -394,7 +418,7 @@ func (l *LinuxFactory) StartInitialization() (err error) {
 		}
 	}()
 
-	i, err := newContainerInit(it, pipe, consoleSocket, fifofd)
+	i, err := newContainerInit(it, pipe, consoleSocket, fifofd, mountFiles)
 	if err != nil {
 		return err
 	}

libcontainer/init_linux.go

@@ -75,7 +75,7 @@ type initer interface {
 	Init() error
 }
 
-func newContainerInit(t initType, pipe *os.File, consoleSocket *os.File, fifoFd int) (initer, error) {
+func newContainerInit(t initType, pipe *os.File, consoleSocket *os.File, fifoFd int, mountFiles []*os.File) (initer, error) {
 	var config *initConfig
 	if err := json.NewDecoder(pipe).Decode(&config); err != nil {
 		return nil, err
@@ -97,6 +97,7 @@ func newContainerInit(t initType, pipe *os.File, consoleSocket *os.File, fifoFd
 			parentPid:     unix.Getppid(),
 			config:        config,
 			fifoFd:        fifoFd,
+			mountFiles:    mountFiles,
 		}, nil
 	}
 	return nil, fmt.Errorf("unknown init type %q", t)

libcontainer/message_linux.go

@@ -20,6 +20,7 @@ const (
 	RootlessEUIDAttr uint16 = 27287
 	UidmapPathAttr   uint16 = 27288
 	GidmapPathAttr   uint16 = 27289
+	MountSourcesAttr uint16 = 27290
 )
 
 type Int32msg struct {