vendor: bump runc to 1.0.0 pre

This is to check if runc 1.0.0 (to be released shortly) works with k8s.

The commands used were (roughly):

	hack/pin-dependency.sh github.com/opencontainers/runc v1.0.0
	hack/lint-dependencies.sh
	# Follow its recommendations.
	hack/pin-dependency.sh github.com/cilium/ebpf v0.6.1
	hack/pin-dependency.sh github.com/opencontainers/selinux v1.8.2
	hack/pin-dependency.sh github.com/sirupsen/logrus v1.8.1
	# Recheck.
	hack/lint-dependencies.sh
	GO111MODULE=on go mod edit -dropreplace github.com/willf/bitset
	hack/update-vendor.sh
	# Recheck.
	hack/lint-dependencies.sh
	hack/update-internal-modules.sh
	# Recheck.
	hack/lint-dependencies.sh

[v2: rebased, updated runc 3a0234e1fe2e82 -> 2f8e8e9d977500]
[v3: testing master + runc pr 3019]
[v4: updated to 93a01cd4d0b7a0f08a]
[v5: updated to f093cca13d3cf8a484]
[v6: rebased]
[v7: updated to runc v1.0.0]
[v8: rebased]

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

vendor/github.com/opencontainers/runc/libcontainer/cgroups/systemd/common.go

@@ -158,14 +158,27 @@ func findDeviceGroup(ruleType devices.Type, ruleMajor int64) (string, error) {
 	return "", nil
 }
 
+// DeviceAllow is the dbus type "a(ss)" which means we need a struct
+// to represent it in Go.
+type deviceAllowEntry struct {
+	Path  string
+	Perms string
+}
+
+func allowAllDevices() []systemdDbus.Property {
+	// Setting mode to auto and removing all DeviceAllow rules
+	// results in allowing access to all devices.
+	return []systemdDbus.Property{
+		newProp("DevicePolicy", "auto"),
+		newProp("DeviceAllow", []deviceAllowEntry{}),
+	}
+}
+
 // generateDeviceProperties takes the configured device rules and generates a
 // corresponding set of systemd properties to configure the devices correctly.
-func generateDeviceProperties(rules []*devices.Rule) ([]systemdDbus.Property, error) {
-	// DeviceAllow is the type "a(ss)" which means we need a temporary struct
-	// to represent it in Go.
-	type deviceAllowEntry struct {
-		Path  string
-		Perms string
+func generateDeviceProperties(r *configs.Resources) ([]systemdDbus.Property, error) {
+	if r.SkipDevices {
+		return nil, nil
 	}
 
 	properties := []systemdDbus.Property{
@@ -177,7 +190,7 @@ func generateDeviceProperties(rules []*devices.Rule) ([]systemdDbus.Property, er
 
 	// Figure out the set of rules.
 	configEmu := &cgroupdevices.Emulator{}
-	for _, rule := range rules {
+	for _, rule := range r.Devices {
 		if err := configEmu.Apply(*rule); err != nil {
 			return nil, errors.Wrap(err, "apply rule for systemd")
 		}
@@ -189,12 +202,7 @@ func generateDeviceProperties(rules []*devices.Rule) ([]systemdDbus.Property, er
 	if configEmu.IsBlacklist() {
 		// However, if we're dealing with an allow-all rule then we can do it.
 		if configEmu.IsAllowAll() {
-			return []systemdDbus.Property{
-				// Run in white-list mode by setting to "auto" and removing all
-				// DeviceAllow rules.
-				newProp("DevicePolicy", "auto"),
-				newProp("DeviceAllow", []deviceAllowEntry{}),
-			}, nil
+			return allowAllDevices(), nil
 		}
 		logrus.Warn("systemd doesn't support blacklist device rules -- applying temporary deny-all rule")
 		return properties, nil
@@ -203,8 +211,7 @@ func generateDeviceProperties(rules []*devices.Rule) ([]systemdDbus.Property, er
 	// Now generate the set of rules we actually need to apply. Unlike the
 	// normal devices cgroup, in "strict" mode systemd defaults to a deny-all
 	// whitelist which is the default for devices.Emulator.
-	baseEmu := &cgroupdevices.Emulator{}
-	finalRules, err := baseEmu.Transition(configEmu)
+	finalRules, err := configEmu.Rules()
 	if err != nil {
 		return nil, errors.Wrap(err, "get simplified rules for systemd")
 	}
@@ -306,7 +313,7 @@ func getUnitName(c *configs.Cgroup) string {
 // isDbusError returns true if the error is a specific dbus error.
 func isDbusError(err error, name string) bool {
 	if err != nil {
-		var derr *dbus.Error
+		var derr dbus.Error
 		if errors.As(err, &derr) {
 			return strings.Contains(derr.Name, name)
 		}
@@ -355,6 +362,9 @@ func stopUnit(cm *dbusConnManager, unitName string) error {
 		return err
 	})
 	if err == nil {
+		timeout := time.NewTimer(30 * time.Second)
+		defer timeout.Stop()
+
 		select {
 		case s := <-statusChan:
 			close(statusChan)
@@ -362,8 +372,8 @@ func stopUnit(cm *dbusConnManager, unitName string) error {
 			if s != "done" {
 				logrus.Warnf("error removing unit `%s`: got `%s`. Continuing...", unitName, s)
 			}
-		case <-time.After(time.Second):
-			logrus.Warnf("Timed out while waiting for StopUnit(%s) completion signal from dbus. Continuing...", unitName)
+		case <-timeout.C:
+			return errors.New("Timed out while waiting for systemd to remove " + unitName)
 		}
 	}
 	return nil
@@ -476,7 +486,7 @@ func addCpuset(cm *dbusConnManager, props *[]systemdDbus.Property, cpus, mems st
 	}
 
 	if cpus != "" {
-		bits, err := rangeToBits(cpus)
+		bits, err := RangeToBits(cpus)
 		if err != nil {
 			return fmt.Errorf("resources.CPU.Cpus=%q conversion error: %w",
 				cpus, err)
@@ -485,7 +495,7 @@ func addCpuset(cm *dbusConnManager, props *[]systemdDbus.Property, cpus, mems st
 			newProp("AllowedCPUs", bits))
 	}
 	if mems != "" {
-		bits, err := rangeToBits(mems)
+		bits, err := RangeToBits(mems)
 		if err != nil {
 			return fmt.Errorf("resources.CPU.Mems=%q conversion error: %w",
 				mems, err)

vendor/github.com/opencontainers/runc/libcontainer/cgroups/systemd/cpuset.go

@@ -5,15 +5,15 @@ import (
 	"strconv"
 	"strings"
 
+	"github.com/bits-and-blooms/bitset"
 	"github.com/pkg/errors"
-	"github.com/willf/bitset"
 )
 
-// rangeToBits converts a text representation of a CPU mask (as written to
+// RangeToBits converts a text representation of a CPU mask (as written to
 // or read from cgroups' cpuset.* files, e.g. "1,3-5") to a slice of bytes
 // with the corresponding bits set (as consumed by systemd over dbus as
 // AllowedCPUs/AllowedMemoryNodes unit property value).
-func rangeToBits(str string) ([]byte, error) {
+func RangeToBits(str string) ([]byte, error) {
 	bits := &bitset.BitSet{}
 
 	for _, r := range strings.Split(str, ",") {

vendor/github.com/opencontainers/runc/libcontainer/cgroups/systemd/dbus.go

@@ -17,14 +17,16 @@ var (
 	dbusRootless bool
 )
 
-type dbusConnManager struct {
-}
+type dbusConnManager struct{}
 
 // newDbusConnManager initializes systemd dbus connection manager.
 func newDbusConnManager(rootless bool) *dbusConnManager {
+	dbusMu.Lock()
+	defer dbusMu.Unlock()
 	if dbusInited && rootless != dbusRootless {
 		panic("can't have both root and rootless dbus")
 	}
+	dbusInited = true
 	dbusRootless = rootless
 	return &dbusConnManager{}
 }

vendor/github.com/opencontainers/runc/libcontainer/cgroups/systemd/v1.go

@@ -61,7 +61,7 @@ var legacySubsystems = []subsystem{
 func genV1ResourcesProperties(r *configs.Resources, cm *dbusConnManager) ([]systemdDbus.Property, error) {
 	var properties []systemdDbus.Property
 
-	deviceProperties, err := generateDeviceProperties(r.Devices)
+	deviceProperties, err := generateDeviceProperties(r)
 	if err != nil {
 		return nil, err
 	}
@@ -207,9 +207,10 @@ func (m *legacyManager) Destroy() error {
 
 	stopErr := stopUnit(m.dbus, getUnitName(m.cgroups))
 
-	// Both on success and on error, cleanup all the cgroups we are aware of.
-	// Some of them were created directly by Apply() and are not managed by systemd.
-	if err := cgroups.RemovePaths(m.paths); err != nil {
+	// Both on success and on error, cleanup all the cgroups
+	// we are aware of, as some of them were created directly
+	// by Apply() and are not managed by systemd.
+	if err := cgroups.RemovePaths(m.paths); err != nil && stopErr == nil {
 		return err
 	}
 
@@ -237,7 +238,7 @@ func (m *legacyManager) joinCgroups(pid int) error {
 			}
 		default:
 			if path, ok := m.paths[name]; ok {
-				if err := os.MkdirAll(path, 0755); err != nil {
+				if err := os.MkdirAll(path, 0o755); err != nil {
 					return err
 				}
 				if err := cgroups.WriteCgroupProc(path, pid); err != nil {
@@ -338,27 +339,24 @@ func (m *legacyManager) Set(r *configs.Resources) error {
 		return err
 	}
 
+	// Figure out the current freezer state, so we can revert to it after we
+	// temporarily freeze the container.
+	targetFreezerState, err := m.GetFreezerState()
+	if err != nil {
+		return err
+	}
+	if targetFreezerState == configs.Undefined {
+		targetFreezerState = configs.Thawed
+	}
+
 	// We have to freeze the container while systemd sets the cgroup settings.
 	// The reason for this is that systemd's application of DeviceAllow rules
 	// is done disruptively, resulting in spurrious errors to common devices
 	// (unlike our fs driver, they will happily write deny-all rules to running
 	// containers). So we freeze the container to avoid them hitting the cgroup
 	// error. But if the freezer cgroup isn't supported, we just warn about it.
-	targetFreezerState := configs.Undefined
-	if !m.cgroups.SkipDevices {
-		// Figure out the current freezer state, so we can revert to it after we
-		// temporarily freeze the container.
-		targetFreezerState, err = m.GetFreezerState()
-		if err != nil {
-			return err
-		}
-		if targetFreezerState == configs.Undefined {
-			targetFreezerState = configs.Thawed
-		}
-
-		if err := m.Freeze(configs.Frozen); err != nil {
-			logrus.Infof("freeze container before SetUnitProperties failed: %v", err)
-		}
+	if err := m.Freeze(configs.Frozen); err != nil {
+		logrus.Infof("freeze container before SetUnitProperties failed: %v", err)
 	}
 
 	if err := setUnitProperties(m.dbus, getUnitName(m.cgroups), properties...); err != nil {

vendor/github.com/opencontainers/runc/libcontainer/cgroups/systemd/v2.go

@@ -96,7 +96,7 @@ func unifiedResToSystemdProps(cm *dbusConnManager, res map[string]string) (props
 				newProp("CPUWeight", num))
 
 		case "cpuset.cpus", "cpuset.mems":
-			bits, err := rangeToBits(v)
+			bits, err := RangeToBits(v)
 			if err != nil {
 				return nil, fmt.Errorf("unified resource %q=%q conversion error: %w", k, v, err)
 			}
@@ -172,7 +172,7 @@ func genV2ResourcesProperties(r *configs.Resources, cm *dbusConnManager) ([]syst
 	//       aren't the end of the world, but it is a bit concerning. However
 	//       it's unclear if systemd removes all eBPF programs attached when
 	//       doing SetUnitProperties...
-	deviceProperties, err := generateDeviceProperties(r.Devices)
+	deviceProperties, err := generateDeviceProperties(r)
 	if err != nil {
 		return nil, err
 	}
@@ -418,27 +418,24 @@ func (m *unifiedManager) Set(r *configs.Resources) error {
 		return err
 	}
 
+	// Figure out the current freezer state, so we can revert to it after we
+	// temporarily freeze the container.
+	targetFreezerState, err := m.GetFreezerState()
+	if err != nil {
+		return err
+	}
+	if targetFreezerState == configs.Undefined {
+		targetFreezerState = configs.Thawed
+	}
+
 	// We have to freeze the container while systemd sets the cgroup settings.
 	// The reason for this is that systemd's application of DeviceAllow rules
 	// is done disruptively, resulting in spurrious errors to common devices
 	// (unlike our fs driver, they will happily write deny-all rules to running
 	// containers). So we freeze the container to avoid them hitting the cgroup
 	// error. But if the freezer cgroup isn't supported, we just warn about it.
-	targetFreezerState := configs.Undefined
-	if !m.cgroups.SkipDevices {
-		// Figure out the current freezer state, so we can revert to it after we
-		// temporarily freeze the container.
-		targetFreezerState, err = m.GetFreezerState()
-		if err != nil {
-			return err
-		}
-		if targetFreezerState == configs.Undefined {
-			targetFreezerState = configs.Thawed
-		}
-
-		if err := m.Freeze(configs.Frozen); err != nil {
-			logrus.Infof("freeze container before SetUnitProperties failed: %v", err)
-		}
+	if err := m.Freeze(configs.Frozen); err != nil {
+		logrus.Infof("freeze container before SetUnitProperties failed: %v", err)
 	}
 
 	if err := setUnitProperties(m.dbus, getUnitName(m.cgroups), properties...); err != nil {