daemon: use containerd/sys to detect UserNamespaces

The implementation in libcontainer/system is quite complicated,
and we only use it to detect if user-namespaces are enabled.

In addition, the implementation in containerd uses a sync.Once,
so that detection (and reading/parsing `/proc/self/uid_map`) is
only performed once.

Signed-off-by: Sebastiaan van Stijn <github@gone.nl>

daemon/daemon.go

@@ -27,6 +27,7 @@ import (
 	"github.com/containerd/containerd/defaults"
 	"github.com/containerd/containerd/pkg/dialer"
 	"github.com/containerd/containerd/remotes/docker"
+	"github.com/containerd/containerd/sys"
 	"github.com/docker/docker/api/types"
 	containertypes "github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/swarm"
@@ -42,7 +43,6 @@ import (
 	"github.com/docker/docker/errdefs"
 	bkconfig "github.com/moby/buildkit/cmd/buildkitd/config"
 	"github.com/moby/buildkit/util/resolver"
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
 	"github.com/sirupsen/logrus"
 
 	// register graph drivers
@@ -1040,7 +1040,7 @@ func NewDaemon(ctx context.Context, config *config.Config, pluginStore *plugin.S
 	sysInfo := d.RawSysInfo(false)
 	// Check if Devices cgroup is mounted, it is hard requirement for container security,
 	// on Linux.
-	if runtime.GOOS == "linux" && !sysInfo.CgroupDevicesEnabled && !rsystem.RunningInUserNS() {
+	if runtime.GOOS == "linux" && !sysInfo.CgroupDevicesEnabled && !sys.RunningInUserNS() {
 		return nil, errors.New("Devices cgroup isn't mounted")
 	}
 

daemon/daemon_unix.go

@@ -18,6 +18,7 @@ import (
 
 	statsV1 "github.com/containerd/cgroups/stats/v1"
 	statsV2 "github.com/containerd/cgroups/v2/stats"
+	"github.com/containerd/containerd/sys"
 	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/blkiodev"
 	pblkiodev "github.com/docker/docker/api/types/blkiodev"
@@ -44,7 +45,6 @@ import (
 	lntypes "github.com/docker/libnetwork/types"
 	"github.com/moby/sys/mount"
 	"github.com/opencontainers/runc/libcontainer/cgroups"
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
 	specs "github.com/opencontainers/runtime-spec/specs-go"
 	"github.com/opencontainers/selinux/go-selinux/label"
 	"github.com/pkg/errors"
@@ -1668,7 +1668,7 @@ func setMayDetachMounts() error {
 		// Setting may_detach_mounts does not work in an
 		// unprivileged container. Ignore the error, but log
 		// it if we appear not to be in that situation.
-		if !rsystem.RunningInUserNS() {
+		if !sys.RunningInUserNS() {
 			logrus.Debugf("Permission denied writing %q to /proc/sys/fs/may_detach_mounts", "1")
 		}
 		return nil
@@ -1688,7 +1688,7 @@ func setupOOMScoreAdj(score int) error {
 		// Setting oom_score_adj does not work in an
 		// unprivileged container. Ignore the error, but log
 		// it if we appear not to be in that situation.
-		if !rsystem.RunningInUserNS() {
+		if !sys.RunningInUserNS() {
 			logrus.Debugf("Permission denied writing %q to /proc/self/oom_score_adj", stringScore)
 		}
 		return nil

daemon/graphdriver/aufs/aufs.go

@@ -35,6 +35,7 @@ import (
 	"strings"
 	"sync"
 
+	"github.com/containerd/containerd/sys"
 	"github.com/docker/docker/daemon/graphdriver"
 	"github.com/docker/docker/pkg/archive"
 	"github.com/docker/docker/pkg/chrootarchive"
@@ -44,7 +45,6 @@ import (
 	"github.com/docker/docker/pkg/locker"
 	"github.com/docker/docker/pkg/system"
 	"github.com/moby/sys/mount"
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
 	"github.com/opencontainers/selinux/go-selinux/label"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
@@ -177,7 +177,7 @@ func supportsAufs() error {
 	// proc/filesystems for when aufs is supported
 	exec.Command("modprobe", "aufs").Run()
 
-	if rsystem.RunningInUserNS() {
+	if sys.RunningInUserNS() {
 		return ErrAufsNested
 	}
 

daemon/graphdriver/copy/copy.go

@@ -11,9 +11,9 @@ import (
 	"syscall"
 	"time"
 
+	"github.com/containerd/containerd/sys"
 	"github.com/docker/docker/pkg/pools"
 	"github.com/docker/docker/pkg/system"
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
 	"golang.org/x/sys/unix"
 )
 
@@ -184,7 +184,7 @@ func DirCopy(srcDir, dstDir string, copyMode Mode, copyXattrs bool) error {
 			}
 
 		case mode&os.ModeDevice != 0:
-			if rsystem.RunningInUserNS() {
+			if sys.RunningInUserNS() {
 				// cannot create a device if running in user namespace
 				return nil
 			}

daemon/graphdriver/fuse-overlayfs/fuseoverlayfs.go

@@ -14,6 +14,7 @@ import (
 	"path/filepath"
 	"strings"
 
+	"github.com/containerd/containerd/sys"
 	"github.com/docker/docker/daemon/graphdriver"
 	"github.com/docker/docker/daemon/graphdriver/overlayutils"
 	"github.com/docker/docker/pkg/archive"
@@ -25,7 +26,6 @@ import (
 	"github.com/docker/docker/pkg/parsers/kernel"
 	"github.com/docker/docker/pkg/system"
 	"github.com/moby/sys/mount"
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
 	"github.com/opencontainers/selinux/go-selinux/label"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
@@ -475,7 +475,7 @@ func (d *Driver) ApplyDiff(id string, parent string, diff io.Reader) (size int64
 		GIDMaps: d.gidMaps,
 		// Use AUFS whiteout format: https://github.com/containers/storage/blob/39a8d5ed9843844eafb5d2ba6e6a7510e0126f40/drivers/overlay/overlay.go#L1084-L1089
 		WhiteoutFormat: archive.AUFSWhiteoutFormat,
-		InUserNS:       rsystem.RunningInUserNS(),
+		InUserNS:       sys.RunningInUserNS(),
 	}); err != nil {
 		return 0, err
 	}

daemon/graphdriver/overlay2/overlay.go

@@ -15,6 +15,7 @@ import (
 	"strings"
 	"sync"
 
+	"github.com/containerd/containerd/sys"
 	"github.com/docker/docker/daemon/graphdriver"
 	"github.com/docker/docker/daemon/graphdriver/overlayutils"
 	"github.com/docker/docker/daemon/graphdriver/quota"
@@ -29,7 +30,6 @@ import (
 	"github.com/docker/docker/pkg/system"
 	units "github.com/docker/go-units"
 	"github.com/moby/sys/mount"
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
 	"github.com/opencontainers/selinux/go-selinux/label"
 	"github.com/sirupsen/logrus"
 	"golang.org/x/sys/unix"
@@ -683,7 +683,7 @@ func (d *Driver) ApplyDiff(id string, parent string, diff io.Reader) (size int64
 		UIDMaps:        d.uidMaps,
 		GIDMaps:        d.gidMaps,
 		WhiteoutFormat: archive.OverlayWhiteoutFormat,
-		InUserNS:       rsystem.RunningInUserNS(),
+		InUserNS:       sys.RunningInUserNS(),
 	}); err != nil {
 		return 0, err
 	}

daemon/graphdriver/quota/projectquota.go

@@ -57,7 +57,7 @@ import (
 	"path/filepath"
 	"unsafe"
 
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
+	"github.com/containerd/containerd/sys"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
 	"golang.org/x/sys/unix"
@@ -90,7 +90,7 @@ func NewControl(basePath string) (*Control, error) {
 	// If we are running in a user namespace quota won't be supported for
 	// now since makeBackingFsDev() will try to mknod().
 	//
-	if rsystem.RunningInUserNS() {
+	if sys.RunningInUserNS() {
 		return nil, ErrQuotaNotSupported
 	}
 

daemon/oci_linux.go

@@ -14,6 +14,7 @@ import (
 
 	"github.com/containerd/containerd/containers"
 	coci "github.com/containerd/containerd/oci"
+	"github.com/containerd/containerd/sys"
 	containertypes "github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/container"
 	daemonconfig "github.com/docker/docker/daemon/config"
@@ -28,7 +29,6 @@ import (
 	"github.com/opencontainers/runc/libcontainer/apparmor"
 	"github.com/opencontainers/runc/libcontainer/cgroups"
 	"github.com/opencontainers/runc/libcontainer/devices"
-	rsystem "github.com/opencontainers/runc/libcontainer/system"
 	"github.com/opencontainers/runc/libcontainer/user"
 	specs "github.com/opencontainers/runtime-spec/specs-go"
 	"github.com/pkg/errors"
@@ -857,7 +857,7 @@ func WithDevices(daemon *Daemon, c *container.Container) coci.SpecOpts {
 		var devs []specs.LinuxDevice
 		devPermissions := s.Linux.Resources.Devices
 
-		if c.HostConfig.Privileged && !rsystem.RunningInUserNS() {
+		if c.HostConfig.Privileged && !sys.RunningInUserNS() {
 			hostDevices, err := devices.HostDevices()
 			if err != nil {
 				return err