feature: support selinux

cmd/nerdctl/container/container_run_security_linux_test.go

@@ -26,9 +26,15 @@ import (
 
 	"gotest.tools/v3/assert"
 
+	"github.com/containerd/nerdctl/mod/tigron/expect"
+	"github.com/containerd/nerdctl/mod/tigron/require"
+	"github.com/containerd/nerdctl/mod/tigron/test"
+	"github.com/containerd/nerdctl/mod/tigron/tig"
+
 	"github.com/containerd/nerdctl/v2/pkg/apparmorutil"
 	"github.com/containerd/nerdctl/v2/pkg/rootlessutil"
 	"github.com/containerd/nerdctl/v2/pkg/testutil"
+	"github.com/containerd/nerdctl/v2/pkg/testutil/nerdtest"
 )
 
 func getCapEff(base *testutil.Base, args ...string) uint64 {
@@ -186,6 +192,105 @@ func TestRunApparmor(t *testing.T) {
 	base.Cmd("run", "--rm", "--privileged", testutil.AlpineImage, "cat", attrCurrentPath).AssertOutContains("unconfined")
 }
 
+func TestRunSelinuxWithSecurityOpt(t *testing.T) {
+	testCase := nerdtest.Setup()
+	testCase.Require = require.Not(nerdtest.NoSelinux)
+	testContainer := testutil.Identifier(t)
+
+	testCase.SubTests = []*test.Case{
+		{
+			Description: "test run with selinux-enabled",
+			Command: func(data test.Data, helpers test.Helpers) test.TestableCommand {
+				return helpers.Command("--selinux-enabled", "run", "-d", "--security-opt", "label=type:container_t", "--name", testContainer, "sleep", "infinity")
+			},
+			Cleanup: func(data test.Data, helpers test.Helpers) {
+				helpers.Anyhow("rm", "-f", testContainer)
+			},
+			Expected: func(data test.Data, helpers test.Helpers) *test.Expected {
+				return &test.Expected{
+					ExitCode: 0,
+					Output: expect.All(
+						func(stdout string, t tig.T) {
+							inspectOut := helpers.Capture("container", "inspect", "--format", "{{.State.Pid}}", testContainer)
+							pid := strings.TrimSpace(inspectOut)
+							fileName := fmt.Sprintf("/proc/%s/attr/current", pid)
+							data, err := os.ReadFile(fileName)
+							assert.NilError(t, err)
+							assert.Equal(t, strings.Contains(string(data), "container_t"), true)
+						},
+					),
+				}
+			},
+		},
+	}
+	testCase.Run(t)
+}
+func TestRunSelinux(t *testing.T) {
+	testCase := nerdtest.Setup()
+	testCase.Require = require.Not(nerdtest.NoSelinux)
+	testContainer := testutil.Identifier(t)
+
+	testCase.SubTests = []*test.Case{
+		{
+			Description: "test run with selinux-enabled",
+			Command: func(data test.Data, helpers test.Helpers) test.TestableCommand {
+				return helpers.Command("--selinux-enabled", "run", "-d", "--name", testContainer, "sleep", "infinity")
+			},
+			Cleanup: func(data test.Data, helpers test.Helpers) {
+				helpers.Anyhow("rm", "-f", testContainer)
+			},
+			Expected: func(data test.Data, helpers test.Helpers) *test.Expected {
+				return &test.Expected{
+					ExitCode: 0,
+					Output: expect.All(
+						func(stdout string, t tig.T) {
+							inspectOut := helpers.Capture("container", "inspect", "--format", "{{.State.Pid}}", testContainer)
+							pid := strings.TrimSpace(inspectOut)
+							fileName := fmt.Sprintf("/proc/%s/attr/current", pid)
+							data, err := os.ReadFile(fileName)
+							assert.NilError(t, err)
+							assert.Equal(t, strings.Contains(string(data), "container_t"), true)
+						},
+					),
+				}
+			},
+		},
+	}
+	testCase.Run(t)
+}
+
+func TestRunSelinuxWithVolumeLabel(t *testing.T) {
+	testCase := nerdtest.Setup()
+	testCase.Require = require.Not(nerdtest.NoSelinux)
+	testContainer := testutil.Identifier(t)
+
+	testCase.SubTests = []*test.Case{
+		{
+			Description: "test run with selinux-enabled",
+			Command: func(data test.Data, helpers test.Helpers) test.TestableCommand {
+				return helpers.Command("--selinux-enabled", "run", "-d", "-v", fmt.Sprintf("/%s:/%s:Z", testContainer, testContainer), "--name", testContainer, "sleep", "infinity")
+			},
+			Cleanup: func(data test.Data, helpers test.Helpers) {
+				helpers.Anyhow("rm", "-f", testContainer)
+			},
+			Expected: func(data test.Data, helpers test.Helpers) *test.Expected {
+				return &test.Expected{
+					ExitCode: 0,
+					Output: expect.All(
+						func(stdout string, t tig.T) {
+							cmd := exec.Command("ls", "-Z", fmt.Sprintf("/%s", testContainer))
+							lsStdout, err := cmd.CombinedOutput()
+							assert.NilError(t, err)
+							assert.Equal(t, strings.Contains(string(lsStdout), "container_t"), true)
+						},
+					),
+				}
+			},
+		},
+	}
+	testCase.Run(t)
+}
+
 // TestRunSeccompCapSysPtrace tests https://github.com/containerd/nerdctl/issues/976
 func TestRunSeccompCapSysPtrace(t *testing.T) {
 	base := testutil.NewBase(t)

cmd/nerdctl/helpers/flagutil.go

@@ -154,6 +154,10 @@ func ProcessRootCmdFlags(cmd *cobra.Command) (types.GlobalCommandOptions, error)
 		return types.GlobalCommandOptions{}, err
 	}
 
+	selinuxEnabled, err := cmd.Flags().GetBool("selinux-enabled")
+	if err != nil {
+		return types.GlobalCommandOptions{}, err
+	}
 	// Point to dataRoot for filesystem-helpers implementing rollback / backups.
 	err = fs.InitFS(dataRoot)
 	if err != nil {
@@ -180,6 +184,7 @@ func ProcessRootCmdFlags(cmd *cobra.Command) (types.GlobalCommandOptions, error)
 		DNS:              dns,
 		DNSOpts:          dnsOpts,
 		DNSSearch:        dnsSearch,
+		SelinuxEnabled:   selinuxEnabled,
 	}, nil
 }
 

cmd/nerdctl/main.go

@@ -191,6 +191,7 @@ func initRootCmdFlags(rootCmd *cobra.Command, tomlPath string) (*pflag.FlagSet,
 	helpers.AddPersistentStringFlag(rootCmd, "host-gateway-ip", nil, nil, nil, aliasToBeInherited, cfg.HostGatewayIP, "NERDCTL_HOST_GATEWAY_IP", "IP address that the special 'host-gateway' string in --add-host resolves to. Defaults to the IP address of the host. It has no effect without setting --add-host")
 	helpers.AddPersistentStringFlag(rootCmd, "bridge-ip", nil, nil, nil, aliasToBeInherited, cfg.BridgeIP, "NERDCTL_BRIDGE_IP", "IP address for the default nerdctl bridge network")
 	rootCmd.PersistentFlags().Bool("kube-hide-dupe", cfg.KubeHideDupe, "Deduplicate images for Kubernetes with namespace k8s.io")
+	rootCmd.PersistentFlags().Bool("selinux-enabled", cfg.SelinuxEnabled, "Enable selinux support")
 	rootCmd.PersistentFlags().StringSlice("cdi-spec-dirs", cfg.CDISpecDirs, "The directories to search for CDI spec files. Defaults to /etc/cdi,/var/run/cdi")
 	rootCmd.PersistentFlags().String("userns-remap", cfg.UsernsRemap, "Support idmapping for creating and running containers. This options is only supported on linux. If `host` is passed, no idmapping is done. if a user name is passed, it does idmapping based on the uidmap and gidmap ranges specified in /etc/subuid and /etc/subgid respectively")
 	helpers.HiddenPersistentStringArrayFlag(rootCmd, "global-dns", cfg.DNS, "Global DNS servers for containers")

docs/command-reference.md

@@ -255,6 +255,7 @@ Security flags:
 
 - :whale: `--security-opt seccomp=<PROFILE_JSON_FILE>`: specify custom seccomp profile
 - :whale: `--security-opt apparmor=<PROFILE>`: specify custom AppArmor profile
+  :whale: `--security-opt label=<selinuxbel>`: specify custom selinux label
 - :whale: `--security-opt no-new-privileges`: disallow privilege escalation, e.g., setuid and file capabilities
 - :whale: `--security-opt systempaths=unconfined`: Turn off confinement for system paths (masked paths, read-only paths) for the container
 - :whale: `--security-opt writable-cgroups`: making the cgroups writeable
@@ -1977,6 +1978,7 @@ Flags:
 - :nerd_face: `--host-gateway-ip`: IP address that the special 'host-gateway' string in --add-host resolves to. It has no effect without setting --add-host
   - Default: the IP address of the host
 - :nerd_face: `--userns-remap=<username>:<groupname>`: Support idmapping of containers. This options is only supported on rootful linux for container create and run if a user name and optionally group name is passed, it does idmapping based on the uidmap and gidmap ranges specified in /etc/subuid and /etc/subgid respectively. Note: `--userns-remap` is not supported for building containers. Nerdctl Build doesn't support userns-remap feature. (format: <name|uid>[:<group|gid>])
+- :nerd_face: `--selinux-enabled`: Enable selinux support
 
 The global flags can be also specified in `/etc/nerdctl/nerdctl.toml` (rootful) and `~/.config/nerdctl/nerdctl.toml` (rootless).
 See [`./config.md`](./config.md).

docs/config.md

@@ -30,6 +30,7 @@ userns_remap   = ""
 dns            = ["8.8.8.8", "1.1.1.1"]
 dns_opts       = ["ndots:1", "timeout:2"]
 dns_search     = ["example.com", "example.org"]
+selinux_enabled= true
 ```
 
 ## Properties

go.mod

@@ -53,6 +53,7 @@ require (
 	github.com/opencontainers/go-digest v1.0.0
 	github.com/opencontainers/image-spec v1.1.1
 	github.com/opencontainers/runtime-spec v1.3.0
+	github.com/opencontainers/selinux v1.13.1
 	github.com/pelletier/go-toml/v2 v2.3.0
 	github.com/rootless-containers/bypass4netns v0.4.2 //gomodjail:unconfined
 	github.com/rootless-containers/rootlesskit/v2 v2.3.6 //gomodjail:unconfined
@@ -113,7 +114,6 @@ require (
 	github.com/multiformats/go-multihash v0.2.3 // indirect
 	github.com/multiformats/go-varint v0.1.0 // indirect
 	github.com/opencontainers/runtime-tools v0.9.1-0.20251114084447-edf4cb3d2116 // indirect
-	github.com/opencontainers/selinux v1.13.1 // indirect
 	github.com/petermattis/goid v0.0.0-20240813172612-4fcff4a6cae7 // indirect
 	github.com/philhofer/fwd v1.2.0 // indirect
 	github.com/pkg/errors v0.9.1 // indirect

pkg/cmd/container/run_linux.go

@@ -71,7 +71,7 @@ func setPlatformOptions(ctx context.Context, client *containerd.Client, id, uts
 	}
 	opts = append(opts, capOpts...)
 	securityOptsMaps := strutil.ConvertKVStringsToMap(strutil.DedupeStrSlice(options.SecurityOpt))
-	secOpts, err := generateSecurityOpts(options.Privileged, securityOptsMaps)
+	secOpts, err := generateSecurityOpts(options.Privileged, options.GOptions.SelinuxEnabled, securityOptsMaps)
 	if err != nil {
 		return nil, err
 	}

pkg/cmd/container/run_security_linux.go

@@ -17,14 +17,19 @@
 package container
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"strconv"
 	"strings"
 	"sync"
 
+	"github.com/opencontainers/runtime-spec/specs-go"
+	"github.com/opencontainers/selinux/go-selinux/label"
+
 	"github.com/containerd/containerd/v2/contrib/apparmor"
 	"github.com/containerd/containerd/v2/contrib/seccomp"
+	"github.com/containerd/containerd/v2/core/containers"
 	"github.com/containerd/containerd/v2/pkg/cap"
 	"github.com/containerd/containerd/v2/pkg/oci"
 	"github.com/containerd/log"
@@ -51,10 +56,10 @@ const (
 	systemPathsUnconfined = "unconfined"
 )
 
-func generateSecurityOpts(privileged bool, securityOptsMap map[string]string) ([]oci.SpecOpts, error) {
+func generateSecurityOpts(privileged bool, selinuxEnabled bool, securityOptsMap map[string]string) ([]oci.SpecOpts, error) {
 	for k := range securityOptsMap {
 		switch k {
-		case "seccomp", "apparmor", "no-new-privileges", "systempaths", "privileged-without-host-devices", "writable-cgroups":
+		case "seccomp", "apparmor", "no-new-privileges", "systempaths", "privileged-without-host-devices", "writable-cgroups", "label":
 		default:
 			log.L.Warnf("unknown security-opt: %q", k)
 		}
@@ -95,6 +100,27 @@ func generateSecurityOpts(privileged bool, securityOptsMap map[string]string) ([
 			opts = append(opts, apparmor.WithProfile(defaults.AppArmorProfileName))
 		}
 	}
+	var labelOpts []string
+	// TODO: should set unique MCS categorie.
+	if !privileged && selinuxEnabled {
+		if selinuxLabel, ok := securityOptsMap["label"]; ok {
+			labelOpts = append(labelOpts, selinuxLabel)
+			processLabel, mountLabel, err := label.InitLabels(labelOpts)
+			if err != nil {
+				return nil, err
+			}
+			opts = append(opts, WithSelinuxLabel(processLabel, mountLabel))
+		}
+
+		// if selinux-enabled=true and security-opt selinux label is not set.
+		if len(labelOpts) == 0 {
+			processLabel, mountLabel, err := label.InitLabels(labelOpts)
+			if err != nil {
+				return nil, err
+			}
+			opts = append(opts, WithSelinuxLabel(processLabel, mountLabel))
+		}
+	}
 
 	nnp, err := maputil.MapBoolValueAsOpt(securityOptsMap, "no-new-privileges")
 	if err != nil {
@@ -141,6 +167,21 @@ func generateSecurityOpts(privileged bool, securityOptsMap map[string]string) ([
 	return opts, nil
 }
 
+// WithSelinuxLabels sets the mount and process labels
+func WithSelinuxLabel(process, mount string) oci.SpecOpts {
+	return func(_ context.Context, _ oci.Client, _ *containers.Container, s *oci.Spec) error {
+		if s.Linux == nil {
+			s.Linux = &specs.Linux{}
+		}
+		if s.Process == nil {
+			s.Process = &specs.Process{}
+		}
+		s.Linux.MountLabel = mount
+		s.Process.SelinuxLabel = process
+		return nil
+	}
+}
+
 func canonicalizeCapName(s string) string {
 	if s == "" {
 		return ""

pkg/config/config.go

@@ -47,6 +47,7 @@ type Config struct {
 	DNSOpts          []string `toml:"dns_opts,omitempty"`
 	DNSSearch        []string `toml:"dns_search,omitempty"`
 	DisableHCSystemd bool     `toml:"disable_hc_systemd"`
+	SelinuxEnabled   bool     `toml:"selinux_enabled"`
 }
 
 // New creates a default Config object statically,
@@ -63,6 +64,7 @@ func New() *Config {
 		DataRoot:         ncdefaults.DataRoot(),
 		CgroupManager:    ncdefaults.CgroupManager(),
 		InsecureRegistry: false,
+		SelinuxEnabled:   false,
 		HostsDir:         ncdefaults.HostsDirs(),
 		Experimental:     true,
 		HostGatewayIP:    ncdefaults.HostGatewayIP(),

pkg/mountutil/mountutil_linux.go

@@ -28,6 +28,7 @@ import (
 	"github.com/docker/go-units"
 	mobymount "github.com/moby/sys/mount"
 	"github.com/opencontainers/runtime-spec/specs-go"
+	"github.com/opencontainers/selinux/go-selinux/label"
 	"golang.org/x/sys/unix"
 
 	"github.com/containerd/containerd/v2/core/containers"
@@ -112,6 +113,7 @@ func parseVolumeOptionsWithMountInfo(vType, src, optsRaw string, getMountInfoFun
 		propagationRawOpts []string
 		bindOpts           []string
 	)
+	var specOpts []oci.SpecOpts
 	for _, opt := range strings.Split(optsRaw, ",") {
 		switch opt {
 		case "rw", "ro", "rro":
@@ -121,6 +123,15 @@ func parseVolumeOptionsWithMountInfo(vType, src, optsRaw string, getMountInfoFun
 		case "bind", "rbind":
 			// bind means not recursively bind-mounted, rbind is the opposite
 			bindOpts = append(bindOpts, opt)
+		case "Z", "z":
+			specOpts = append(specOpts, func(ctx context.Context, cli oci.Client, c *containers.Container, s *oci.Spec) error {
+				if s.Linux != nil && s.Linux.MountLabel != "" {
+					if err := label.Relabel(src, s.Linux.MountLabel, opt == "z"); err != nil {
+						return err
+					}
+				}
+				return nil
+			})
 		case "":
 			// NOP
 		default:
@@ -129,7 +140,6 @@ func parseVolumeOptionsWithMountInfo(vType, src, optsRaw string, getMountInfoFun
 	}
 
 	var opts []string
-	var specOpts []oci.SpecOpts
 
 	if len(bindOpts) > 0 && vType != Bind {
 		return nil, nil, fmt.Errorf("volume bind/rbind option is only supported for bind mount: %+v", bindOpts)

pkg/testutil/nerdtest/requirements.go

@@ -25,6 +25,7 @@ import (
 	"strings"
 
 	"github.com/Masterminds/semver/v3"
+	"github.com/opencontainers/selinux/go-selinux"
 	"gotest.tools/v3/assert"
 
 	"github.com/containerd/containerd/v2/defaults"
@@ -161,6 +162,19 @@ var Rootless = &test.Requirement{
 	},
 }
 
+// NoSelinux marks a test as suitable only for the noselinux enable environment
+var NoSelinux = &test.Requirement{
+	Check: func(data test.Data, helpers test.Helpers) (ret bool, mess string) {
+		ret = !selinux.GetEnabled()
+		if ret {
+			mess = "selinux is disabled"
+		} else {
+			mess = "selinux is enabled"
+		}
+		return ret, mess
+	},
+}
+
 // RootlessWithDetachNetNS marks a test as suitable only for rootless environment with detached netns support.
 var RootlessWithDetachNetNS = &test.Requirement{
 	Check: func(data test.Data, helpers test.Helpers) (ret bool, mess string) {