Fix cgroup2 mount for rootless case

In case of rootless, cgroup2 mount is not possible (see [1] for more
details), so since commit 9c81440 runc bind-mounts the whole
/sys/fs/cgroup into container.

Problem is, if cgroupns is enabled, /sys/fs/cgroup inside the container
is supposed to show the cgroup files for this cgroup, not the root one.

The fix is to pass through and use the cgroup path in case cgroup2
mount failed, cgroupns is enabled, and the path is non-empty.

Surely this requires the /sys/fs/cgroup mount in the spec, so modify
runc spec --rootless to keep it.

Before:

	$ ./runc run aaa
	# find /sys/fs/cgroup/ -type d
	/sys/fs/cgroup
	/sys/fs/cgroup/user.slice
	/sys/fs/cgroup/user.slice/user-1000.slice
	/sys/fs/cgroup/user.slice/user-1000.slice/user@1000.service
	...
	# ls -l /sys/fs/cgroup/cgroup.controllers
	-r--r--r--    1 nobody   nogroup          0 Feb 24 02:22 /sys/fs/cgroup/cgroup.controllers
	# wc -w /sys/fs/cgroup/cgroup.procs
	142 /sys/fs/cgroup/cgroup.procs
	# cat /sys/fs/cgroup/memory.current
	cat: can't open '/sys/fs/cgroup/memory.current': No such file or directory

After:

	# find /sys/fs/cgroup/ -type d
	/sys/fs/cgroup/
	# ls -l /sys/fs/cgroup/cgroup.controllers
	-r--r--r--    1 root     root             0 Feb 24 02:43 /sys/fs/cgroup/cgroup.controllers
	# wc -w /sys/fs/cgroup/cgroup.procs
	2 /sys/fs/cgroup/cgroup.procs
	# cat /sys/fs/cgroup/memory.current
	577536

[1] opencontainers#2158

Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>

libcontainer/container_linux.go

@@ -626,6 +626,9 @@ func (c *linuxContainer) newInitConfig(process *Process) *initConfig {
 	if len(process.Rlimits) > 0 {
 		cfg.Rlimits = process.Rlimits
 	}
+	if cgroups.IsCgroup2UnifiedMode() {
+		cfg.Cgroup2Path = c.cgroupManager.Path("")
+	}
 
 	return cfg
 }

libcontainer/init_linux.go

@@ -70,6 +70,7 @@ type initConfig struct {
 	RootlessEUID     bool                  `json:"rootless_euid,omitempty"`
 	RootlessCgroups  bool                  `json:"rootless_cgroups,omitempty"`
 	SpecState        *specs.State          `json:"spec_state,omitempty"`
+	Cgroup2Path      string                `json:"cgroup2_path,omitempty"`
 }
 
 type initer interface {

libcontainer/rootfs_linux.go

@@ -31,9 +31,11 @@ import (
 const defaultMountFlags = unix.MS_NOEXEC | unix.MS_NOSUID | unix.MS_NODEV
 
 type mountConfig struct {
-	root     string
-	label    string
-	cgroupns bool
+	root            string
+	label           string
+	cgroup2Path     string
+	rootlessCgroups bool
+	cgroupns        bool
 }
 
 // needsSetupDev returns true if /dev needs to be set up.
@@ -56,9 +58,11 @@ func prepareRootfs(pipe io.ReadWriter, iConfig *initConfig) (err error) {
 	}
 
 	mountConfig := &mountConfig{
-		root:     config.Rootfs,
-		label:    config.MountLabel,
-		cgroupns: config.Namespaces.Contains(configs.NEWCGROUP),
+		root:            config.Rootfs,
+		label:           config.MountLabel,
+		cgroup2Path:     iConfig.Cgroup2Path,
+		rootlessCgroups: iConfig.RootlessCgroups,
+		cgroupns:        config.Namespaces.Contains(configs.NEWCGROUP),
 	}
 	setupDev := needsSetupDev(config)
 	for _, m := range config.Mounts {
@@ -307,7 +311,17 @@ func mountCgroupV2(m *configs.Mount, c *mountConfig) error {
 		// when we are in UserNS but CgroupNS is not unshared, we cannot mount cgroup2 (#2158)
 		if err == unix.EPERM || err == unix.EBUSY {
 			src := fs2.UnifiedMountpoint
-			return unix.Mount(src, dest, "", uintptr(m.Flags)|unix.MS_BIND, "")
+			if c.cgroupns && c.cgroup2Path != "" {
+				// Emulate cgroupns by bind-mounting
+				// the container cgroup path rather than
+				// the whole /sys/fs/cgroup.
+				src = c.cgroup2Path
+			}
+			err = unix.Mount(src, dest, "", uintptr(m.Flags)|unix.MS_BIND, "")
+			if err == unix.ENOENT && c.rootlessCgroups {
+				err = nil
+			}
+			return err
 		}
 		return err
 	}

libcontainer/specconv/example.go

@@ -2,6 +2,7 @@ package specconv
 
 import (
 	"os"
+	"path/filepath"
 	"strings"
 
 	"github.com/opencontainers/runc/libcontainer/cgroups"
@@ -200,8 +201,14 @@ func ToRootless(spec *specs.Spec) {
 	// Fix up mounts.
 	var mounts []specs.Mount
 	for _, mount := range spec.Mounts {
-		// Ignore all mounts that are under /sys.
-		if strings.HasPrefix(mount.Destination, "/sys") {
+		// Replace the /sys mount with an rbind.
+		if filepath.Clean(mount.Destination) == "/sys" {
+			mounts = append(mounts, specs.Mount{
+				Source:      "/sys",
+				Destination: "/sys",
+				Type:        "none",
+				Options:     []string{"rbind", "nosuid", "noexec", "nodev", "ro"},
+			})
 			continue
 		}
 
@@ -216,13 +223,6 @@ func ToRootless(spec *specs.Spec) {
 		mount.Options = options
 		mounts = append(mounts, mount)
 	}
-	// Add the sysfs mount as an rbind.
-	mounts = append(mounts, specs.Mount{
-		Source:      "/sys",
-		Destination: "/sys",
-		Type:        "none",
-		Options:     []string{"rbind", "nosuid", "noexec", "nodev", "ro"},
-	})
 	spec.Mounts = mounts
 
 	// Remove cgroup settings.