libct/cg/sd: support setting cpu.idle via systemd

Systemd v252 (available in CentOS Stream 9 in our CI) added support
for setting cpu.idle (see [1]). The way it works is:
 - if CPUWeight == 0, cpu.idle is set to 1;
 - if CPUWeight != 0, cpu.idle is set to 0.

This commit implements setting cpu.idle in systemd cgroup driver via a
unit property. In case CPUIdle is set to non-zero value, the driver sets
adds CPUWeight=0 property, which will result in systemd setting cpu.idle
to 1.

Unfortunately, there's no way to set cpu.idle to 0 without also changing
the CPUWeight value, so the driver doesn't do anything if CPUIdle is
explicitly set to 0. This case is handled by the fs driver which is
always used as a followup to setting systemd unit properties.

Also, handle cpu.idle set via unified map. In case it is set to non-zero
value, add CPUWeight=0 property, and ignore cpu.weight (otherwise we'll
get two different CPUWeight properties set).

Add a unit test for new values in unified map, and an integration test case.

[1] systemd/systemd#23299
[2] opencontainers#3786

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

docs/systemd.md

@@ -91,6 +91,7 @@ The following tables summarize which properties are translated.
 | cpu.mems                | AllowedMemoryNodes    | v244                |
 | unified.cpu.max         | CPUQuota, CPUQuotaPeriodSec | v242          |
 | unified.cpu.weight      | CPUWeight             |                     |
+| unified.cpu.idle        | CPUWeight             | v252                |
 | unified.cpuset.cpus     | AllowedCPUs           | v244                |
 | unified.cpuset.mems     | AllowedMemoryNodes    | v244                |
 | unified.memory.high     | MemoryHigh            |                     |

libcontainer/cgroups/systemd/systemd_test.go

@@ -2,8 +2,10 @@ package systemd
 
 import (
 	"os"
+	"reflect"
 	"testing"
 
+	systemdDbus "github.com/coreos/go-systemd/v22/dbus"
 	"github.com/opencontainers/runc/libcontainer/cgroups"
 	"github.com/opencontainers/runc/libcontainer/configs"
 )
@@ -95,3 +97,85 @@ func TestUnitExistsIgnored(t *testing.T) {
 		}
 	}
 }
+
+func TestUnifiedResToSystemdProps(t *testing.T) {
+	if !IsRunningSystemd() {
+		t.Skip("Test requires systemd.")
+	}
+	if !cgroups.IsCgroup2UnifiedMode() {
+		t.Skip("cgroup v2 is required")
+	}
+
+	cm := newDbusConnManager(os.Geteuid() != 0)
+
+	testCases := []struct {
+		name     string
+		minVer   int
+		res      map[string]string
+		expError bool
+		expProps []systemdDbus.Property
+	}{
+		{
+			name: "empty map",
+			res:  map[string]string{},
+		},
+		{
+			name:   "only cpu.idle=1",
+			minVer: cpuIdleSupportedVersion,
+			res: map[string]string{
+				"cpu.idle": "1",
+			},
+			expProps: []systemdDbus.Property{
+				newProp("CPUWeight", uint64(0)),
+			},
+		},
+		{
+			name:   "only cpu.idle=0",
+			minVer: cpuIdleSupportedVersion,
+			res: map[string]string{
+				"cpu.idle": "0",
+			},
+		},
+		{
+			name:   "cpu.idle=1 and cpu.weight=1000",
+			minVer: cpuIdleSupportedVersion,
+			res: map[string]string{
+				"cpu.idle":   "1",
+				"cpu.weight": "1000",
+			},
+			expProps: []systemdDbus.Property{
+				newProp("CPUWeight", uint64(0)),
+			},
+		},
+		{
+			name:   "cpu.idle=0 and cpu.weight=1000",
+			minVer: cpuIdleSupportedVersion,
+			res: map[string]string{
+				"cpu.idle":   "0",
+				"cpu.weight": "1000",
+			},
+			expProps: []systemdDbus.Property{
+				newProp("CPUWeight", uint64(1000)),
+			},
+		},
+	}
+
+	for _, tc := range testCases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			if tc.minVer != 0 && systemdVersion(cm) < tc.minVer {
+				t.Skipf("requires systemd >= %d", tc.minVer)
+			}
+			props, err := unifiedResToSystemdProps(cm, tc.res)
+			if err != nil && !tc.expError {
+				t.Fatalf("expected no error, got: %v", err)
+			}
+			if err == nil && tc.expError {
+				t.Fatal("expected error, got nil")
+			}
+			if !reflect.DeepEqual(tc.expProps, props) {
+				t.Errorf("wrong properties (exp %+v, got %+v)", tc.expProps, props)
+			}
+		})
+	}
+}

libcontainer/cgroups/systemd/v2.go

@@ -20,6 +20,10 @@ import (
 	"github.com/opencontainers/runc/libcontainer/configs"
 )
 
+const (
+	cpuIdleSupportedVersion = 252
+)
+
 type UnifiedManager struct {
 	mu      sync.Mutex
 	cgroups *configs.Cgroup
@@ -60,6 +64,13 @@ func NewUnifiedManager(config *configs.Cgroup, path string) (*UnifiedManager, er
 func unifiedResToSystemdProps(cm *dbusConnManager, res map[string]string) (props []systemdDbus.Property, _ error) {
 	var err error
 
+	shouldSetIdle := func(v string) bool {
+		// The only valid values for cpu.idle is 0 or 1. Since it is
+		// not possible to directly set cpu.idle to 0 via systemd,
+		// ignore 0.
+		return v == "1" && systemdVersion(cm) >= cpuIdleSupportedVersion
+	}
+
 	for k, v := range res {
 		if strings.Contains(k, "/") {
 			return nil, fmt.Errorf("unified resource %q must be a file name (no slashes)", k)
@@ -97,13 +108,26 @@ func unifiedResToSystemdProps(cm *dbusConnManager, res map[string]string) (props
 			addCpuQuota(cm, &props, quota, period)
 
 		case "cpu.weight":
+			if shouldSetIdle(strings.TrimSpace(res["cpu.idle"])) {
+				// Do not add duplicate CPUWeight property
+				// (see case "cpu.idle" below).
+				continue
+			}
 			num, err := strconv.ParseUint(v, 10, 64)
 			if err != nil {
 				return nil, fmt.Errorf("unified resource %q value conversion error: %w", k, err)
 			}
 			props = append(props,
 				newProp("CPUWeight", num))
 
+		case "cpu.idle":
+			if shouldSetIdle(v) {
+				// Setting CPUWeight to 0 tells systemd
+				// to set cpu.idle to 1.
+				props = append(props,
+					newProp("CPUWeight", uint64(0)))
+			}
+
 		case "cpuset.cpus", "cpuset.mems":
 			bits, err := RangeToBits(v)
 			if err != nil {
@@ -212,7 +236,17 @@ func genV2ResourcesProperties(dirPath string, r *configs.Resources, cm *dbusConn
 			newProp("MemorySwapMax", uint64(swap)))
 	}
 
-	if r.CpuWeight != 0 {
+	CPUWeightSet := false
+	if r.CPUIdle != nil {
+		sdVer := systemdVersion(cm)
+		if sdVer >= cpuIdleSupportedVersion && *r.CPUIdle != 0 {
+			properties = append(properties,
+				newProp("CPUWeight", uint64(0)))
+			CPUWeightSet = true
+		}
+	}
+	// Ignore CpuWeight if CPUIdle is set.
+	if !CPUWeightSet && r.CpuWeight != 0 {
 		properties = append(properties,
 			newProp("CPUWeight", r.CpuWeight))
 	}

tests/integration/update.bats

@@ -462,6 +462,55 @@ EOF
 	check_cgroup_value "cpu.idle" "1"
 }
 
+@test "update cgroup cpu.idle via systemd v252+" {
+	requires systemd cgroups_cpu_idle
+	[ $EUID -ne 0 ] && requires rootless_cgroup
+	if [ "$(systemd_version)" -lt 252 ]; then
+		skip "requires systemd >= v252"
+	fi
+
+	runc run -d --console-socket "$CONSOLE_SOCKET" test_update
+	[ "$status" -eq 0 ]
+	check_cgroup_value "cpu.idle" "0"
+
+	runc update --cpu-idle 1 test_update
+	check_cgroup_value "cpu.idle" "1"
+
+	runc update --cpu-idle 0 test_update
+	check_cgroup_value "cpu.idle" "0"
+
+	# If cpu-idle is set, cpu-period is ignored.
+	runc update --cpu-period 10000 --cpu-idle 1 test_update
+	check_cgroup_value "cpu.idle" "1"
+
+	# Setting any cpu-period should reset cpu.idle to 0.
+	runc update --cpu-period 20000 test_update
+	check_cgroup_value "cpu.idle" "0"
+
+	# Setting values via unified map.
+
+	# If cpu.idle is set, cpu.weight is ignored.
+	runc update -r - test_update <<EOF
+{
+  "unified": {
+    "cpu.idle": "1",
+    "cpu.weight": "8"
+  }
+}
+EOF
+	check_cgroup_value "cpu.idle" "1"
+
+	# Setting any cpu.weight should reset cpu.idle to 0.
+	runc update -r - test_update <<EOF
+{
+  "unified": {
+    "cpu.weight": "8"
+  }
+}
+EOF
+	check_cgroup_value "cpu.idle" "0"
+}
+
 @test "update cgroup v2 resources via unified map" {
 	[ $EUID -ne 0 ] && requires rootless_cgroup
 	requires cgroups_v2