Merge pull request #5570 from kolyshkin/cgroup-manager

Add --cgroup-manager, deprecate --systemd-cgroup

CHANGELOG.md

@@ -12,11 +12,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   says, the destination is a path inside the container, and it is now used
   as such. (#5556)
 
+### Added ###
+- New `--cgroup-manager` global option, to select the cgroup manager
+  (`cgroupfs` or `systemd`). This is compatible with crun. (#5570)
+
 ### Deprecated ###
 - `libcontainer.Stats.Interfaces` and `types.Stats.NetworkInterfaces` (shown
   as `network_interfaces` in `runc events --stats` output) are deprecated, as
   they are always empty since the removal of veth network strategy in runc
   1.0.0-rc6. (#5523)
+- The `--systemd-cgroup` global option is deprecated in favor of
+  `--cgroup-manager=systemd`. It still works, and there are no plans to
+  remove it. (#5570)
 
 ### Changed ###
 - A container configuration which uses `--no-pivot` or sets

contrib/completions/bash/runc

@@ -228,6 +228,7 @@ _runc_runc() {
 		--systemd-cgroup
 	"
 	local options_with_args="
+		--cgroup-manager
 		--log
 		--log-format
 		--root
@@ -255,6 +256,11 @@ _runc_runc() {
 		return
 		;;
 
+	--cgroup-manager)
+		COMPREPLY=($(compgen -W 'cgroupfs systemd' -- "$cur"))
+		return
+		;;
+
 	$(__runc_to_extglob "$options_with_args"))
 		return
 		;;

docs/cgroup-v2.md

@@ -28,7 +28,7 @@ Kernel older than 5.2 is not recommended due to lack of freezer.
 Notably, kernel older than 4.15 MUST NOT be used (unless you are running containers with user namespaces), as it lacks support for controlling permissions of devices.
 
 ### Systemd
-On cgroup v2 hosts, it is highly recommended to run runc with the systemd cgroup driver (`runc --systemd-cgroup`), though not mandatory.
+On cgroup v2 hosts, it is highly recommended to run runc with the systemd cgroup driver (`runc --cgroup-manager=systemd`), though not mandatory.
 
 The recommended systemd version is 244 or later. Older systemd does not support delegation of `cpuset` controller.
 
@@ -45,7 +45,7 @@ On cgroup v2 hosts, rootless runc can talk to systemd to get cgroup permissions
 ```bash
 runc spec --rootless
 jq '.linux.cgroupsPath="user.slice:runc:foo"' config.json | sponge config.json
-runc --systemd-cgroup run foo
+runc --cgroup-manager=systemd run foo
 ```
 
 The container processes are executed in a cgroup like `/user.slice/user-$(id -u).slice/user@$(id -u).service/user.slice/runc-foo.scope`.

docs/systemd.md

@@ -1,9 +1,9 @@
 ## systemd cgroup driver
 
 By default, runc creates cgroups and sets cgroup limits on its own (this mode
-is known as fs cgroup driver). When `--systemd-cgroup` global option is given
-(as in e.g. `runc --systemd-cgroup run ...`), runc switches to systemd cgroup
-driver. This document describes its features and peculiarities.
+is known as fs cgroup driver). When `--cgroup-manager=systemd` global option is
+given (as in e.g. `runc --cgroup-manager=systemd run ...`), runc switches to
+systemd cgroup driver. This document describes its features and peculiarities.
 
 ### systemd unit name and placement
 

main.go

@@ -123,9 +123,14 @@ func main() {
 			Value: root,
 			Usage: "root directory for storage of container state (this should be located in tmpfs)",
 		},
+		&cli.StringFlag{
+			Name:  "cgroup-manager",
+			Value: "cgroupfs",
+			Usage: "cgroup manager to use ('cgroupfs' or 'systemd'); the latter expects cgroupsPath to be of form \"slice:prefix:name\" for e.g. \"system.slice:runc:434234\"",
+		},
 		&cli.BoolFlag{
 			Name:  "systemd-cgroup",
-			Usage: "enable systemd cgroup support, expects cgroupsPath to be of form \"slice:prefix:name\" for e.g. \"system.slice:runc:434234\"",
+			Usage: "same as --cgroup-manager=systemd (deprecated)",
 		},
 		&cli.StringFlag{
 			Name:  "rootless",
@@ -169,6 +174,9 @@ func main() {
 		if err := reviseRootDir(cmd); err != nil {
 			return ctx, err
 		}
+		if err := reviseCgroupManager(cmd); err != nil {
+			return ctx, err
+		}
 
 		if err := configLogrus(cmd); err != nil {
 			return ctx, err

man/runc.8.md

@@ -106,10 +106,13 @@ These options can be used with any command, and must precede the **command**.
 located on tmpfs. Default is */run/runc*, or *$XDG_RUNTIME_DIR/runc* for
 rootless containers.
 
+**--cgroup-manager** **cgroupfs**|**systemd**
+: Set the cgroup manager to use (default is **cgroupfs**). With **systemd**,
+the container spec (_config.json_) is expected to have **cgroupsPath** value
+in the *slice:prefix:name* form (e.g. **system.slice:runc:434234**).
+
 **--systemd-cgroup**
-: Enable systemd cgroup support. If this is set, the container spec
-(_config.json_) is expected to have **cgroupsPath** value in the
-*slice:prefix:name* form (e.g. **system.slice:runc:434234**).
+: Same as **--cgroup-manager=systemd**. Deprecated.
 
 **--rootless** **true**|**false**|**auto**
 : Enable or disable rootless mode. Default is **auto**, meaning to auto-detect

rootless_linux.go

@@ -37,7 +37,7 @@ func shouldUseRootlessCgroupManager(cmd *cli.Command) (bool, error) {
 	//
 	// On error, we assume we are root. An error may happen during shelling out to `busctl` CLI,
 	// mostly when $DBUS_SESSION_BUS_ADDRESS is unset.
-	if cmd.Bool("systemd-cgroup") {
+	if cmd.String("cgroup-manager") == "systemd" {
 		ownerUID, err := systemd.DetectUID()
 		if err != nil {
 			logrus.WithError(err).Debug("failed to get the OwnerUID value, assuming the value to be 0")

script/setup_host.sh

@@ -71,7 +71,7 @@ dnf clean all
 # Setup rootless user.
 "$SCRIPTDIR"/setup_rootless.sh
 
-# Delegate all cgroup v2 controllers to rootless user via --systemd-cgroup
+# Delegate all cgroup v2 controllers to rootless user via --cgroup-manager=systemd
 if [ -e /sys/fs/cgroup/cgroup.controllers ]; then
 	mkdir -p /etc/systemd/system/user@.service.d
 	cat >/etc/systemd/system/user@.service.d/delegate.conf <<'EOF'

tests/integration/bin/runc

@@ -11,7 +11,7 @@
 set -u -e
 
 opts=()
-[ -v RUNC_USE_SYSTEMD ] && opts+=("--systemd-cgroup")
+[ -v RUNC_USE_SYSTEMD ] && opts+=("--cgroup-manager=systemd")
 [ -n "${ROOT:-}" ] && opts+=("--root" "$ROOT/state")
 
 exec "$RUNC" "${opts[@]}" "$@"

tests/integration/cgroups.bats

@@ -15,6 +15,27 @@ function setup() {
 	run -0 runc run -d --console-socket "$CONSOLE_SOCKET" test_cgroups_permissions
 }
 
+@test "runc --cgroup-manager with an invalid value fails" {
+	run -1 runc --cgroup-manager=foo list
+	assert_output --partial 'invalid --cgroup-manager value'
+}
+
+@test "runc --systemd-cgroup conflicts with --cgroup-manager=cgroupfs" {
+	run -1 runc --systemd-cgroup --cgroup-manager=cgroupfs list
+	assert_output --partial "--systemd-cgroup conflicts with --cgroup-manager=cgroupfs"
+}
+
+@test "runc --systemd-cgroup is the same as --cgroup-manager=systemd" {
+	requires systemd
+	[ $EUID -ne 0 ] && requires rootless_cgroup
+
+	set_cgroups_path
+
+	# Not using the runc wrapper, as it adds --cgroup-manager=systemd.
+	run -0 "$RUNC" --root "$ROOT/state" --systemd-cgroup run -d --console-socket "$CONSOLE_SOCKET" test_cgroups_systemd
+	check_systemd_value "ActiveState" "active"
+}
+
 @test "runc create (rootless + no limits + cgrouppath + no permission) fails with permission error" {
 	requires rootless rootless_no_cgroup
 

utils.go

@@ -121,6 +121,24 @@ func reviseRootDir(cmd *cli.Command) error {
 	return cmd.Set("root", root)
 }
 
+// reviseCgroupManager checks the --cgroup-manager option argument, and
+// converts the deprecated --systemd-cgroup option to --cgroup-manager=systemd.
+func reviseCgroupManager(cmd *cli.Command) error {
+	manager := cmd.String("cgroup-manager")
+	switch manager {
+	case "cgroupfs", "systemd":
+	default:
+		return fmt.Errorf("invalid --cgroup-manager value %q (must be 'cgroupfs' or 'systemd')", manager)
+	}
+	if !cmd.Bool("systemd-cgroup") {
+		return nil
+	}
+	if cmd.IsSet("cgroup-manager") && manager != "systemd" {
+		return fmt.Errorf("--systemd-cgroup conflicts with --cgroup-manager=%s", manager)
+	}
+	return cmd.Set("cgroup-manager", "systemd")
+}
+
 // parseBoolOrAuto returns (nil, nil) if s is empty or "auto"
 func parseBoolOrAuto(s string) (*bool, error) {
 	if s == "" || strings.ToLower(s) == "auto" {

utils_linux.go

@@ -188,7 +188,7 @@ func createContainer(cmd *cli.Command, id string, spec *specs.Spec) (*libcontain
 	}
 	config, err := specconv.CreateLibcontainerConfig(&specconv.CreateOpts{
 		CgroupName:       id,
-		UseSystemdCgroup: cmd.Bool("systemd-cgroup"),
+		UseSystemdCgroup: cmd.String("cgroup-manager") == "systemd",
 		NoPivotRoot:      cmd.Bool("no-pivot"),
 		NoNewKeyring:     cmd.Bool("no-new-keyring"),
 		Spec:             spec,