Revert "Set temporary single CPU affinity..."

There's too much logic here figuring out which CPUs to use. Runc is a
low level tool and is not supposed to be that "smart". What's worse,
this logic is executed on every exec, making it slower. Some of the
logic in (*setnsProcess).start is executed even if no annotation is set,
thus making ALL execs slow.

Also, this should be a property of a process, rather than annotation.

The plan is to rework this.

This reverts commit afc23e3.

Signed-off-by: Kir Kolyshkin <[email protected]>

Author: kolyshkin

docs/isolated-cpu-affinity-transition.md

@@ -68,7 +68,6 @@ var featuresCommand = cli.Command{
 				"bundle",
 				"org.systemd.property.", // prefix form
 				"org.criu.config",
-				"org.opencontainers.runc.exec.isolated-cpu-affinity-transition",
 			},
 		}
 

features.go

@@ -72,8 +72,4 @@ type Manager interface {
 
 	// OOMKillCount reports OOM kill count for the cgroup.
 	OOMKillCount() (uint64, error)
-
-	// GetEffectiveCPUs returns the effective CPUs of the cgroup, an empty
-	// value means that the cgroups cpuset subsystem/controller is not enabled.
-	GetEffectiveCPUs() string
 }

libcontainer/cgroups/cgroups.go

@@ -4,8 +4,6 @@ import (
 	"errors"
 	"fmt"
 	"os"
-	"path/filepath"
-	"strings"
 	"sync"
 
 	"golang.org/x/sys/unix"
@@ -265,28 +263,3 @@ func (m *Manager) OOMKillCount() (uint64, error) {
 
 	return c, err
 }
-
-func (m *Manager) GetEffectiveCPUs() string {
-	return GetEffectiveCPUs(m.Path("cpuset"), m.cgroups)
-}
-
-func GetEffectiveCPUs(cpusetPath string, cgroups *configs.Cgroup) string {
-	// Fast path.
-	if cgroups.CpusetCpus != "" {
-		return cgroups.CpusetCpus
-	} else if !strings.HasPrefix(cpusetPath, defaultCgroupRoot) {
-		return ""
-	}
-
-	// Iterates until it goes to the cgroup root path.
-	// It's required for containers in which cpuset controller
-	// is not enabled, in this case a parent cgroup is used.
-	for path := cpusetPath; path != defaultCgroupRoot; path = filepath.Dir(path) {
-		cpus, err := fscommon.GetCgroupParamString(path, "cpuset.effective_cpus")
-		if err == nil {
-			return cpus
-		}
-	}
-
-	return ""
-}

libcontainer/cgroups/fs/fs.go

@@ -4,13 +4,11 @@ import (
 	"errors"
 	"fmt"
 	"os"
-	"path/filepath"
 	"strings"
 
 	"github.com/opencontainers/runc/libcontainer/cgroups"
 	"github.com/opencontainers/runc/libcontainer/cgroups/fscommon"
 	"github.com/opencontainers/runc/libcontainer/configs"
-	"github.com/opencontainers/runc/libcontainer/utils"
 )
 
 type parseError = fscommon.ParseError
@@ -34,9 +32,6 @@ func NewManager(config *configs.Cgroup, dirPath string) (*Manager, error) {
 		if err != nil {
 			return nil, err
 		}
-	} else {
-		// Clean path for safety.
-		dirPath = utils.CleanPath(dirPath)
 	}
 
 	m := &Manager{
@@ -321,26 +316,3 @@ func CheckMemoryUsage(dirPath string, r *configs.Resources) error {
 
 	return nil
 }
-
-func (m *Manager) GetEffectiveCPUs() string {
-	// Fast path.
-	if m.config.CpusetCpus != "" {
-		return m.config.CpusetCpus
-	} else if !strings.HasPrefix(m.dirPath, UnifiedMountpoint) {
-		return ""
-	}
-
-	// Iterates until it goes outside of the cgroup root path.
-	// It's required for containers in which cpuset controller
-	// is not enabled, in this case a parent cgroup is used.
-	outsidePath := filepath.Dir(UnifiedMountpoint)
-
-	for path := m.dirPath; path != outsidePath; path = filepath.Dir(path) {
-		cpus, err := fscommon.GetCgroupParamString(path, "cpuset.cpus.effective")
-		if err == nil {
-			return cpus
-		}
-	}
-
-	return ""
-}

libcontainer/cgroups/fs2/fs2.go

@@ -411,7 +411,3 @@ func (m *LegacyManager) Exists() bool {
 func (m *LegacyManager) OOMKillCount() (uint64, error) {
 	return fs.OOMKillCount(m.Path("memory"))
 }
-
-func (m *LegacyManager) GetEffectiveCPUs() string {
-	return fs.GetEffectiveCPUs(m.Path("cpuset"), m.cgroups)
-}

libcontainer/cgroups/systemd/v1.go

@@ -514,7 +514,3 @@ func (m *UnifiedManager) Exists() bool {
 func (m *UnifiedManager) OOMKillCount() (uint64, error) {
 	return m.fsMgr.OOMKillCount()
 }
-
-func (m *UnifiedManager) GetEffectiveCPUs() string {
-	return m.fsMgr.GetEffectiveCPUs()
-}

libcontainer/cgroups/systemd/v2.go

@@ -136,18 +136,18 @@ func GetAllSubsystems() ([]string, error) {
 	return subsystems, nil
 }
 
-func readProcsFile(dir string) ([]int, error) {
-	f, err := OpenFile(dir, CgroupProcesses, os.O_RDONLY)
+func readProcsFile(dir string) (out []int, _ error) {
+	file := CgroupProcesses
+	retry := true
+
+again:
+	f, err := OpenFile(dir, file, os.O_RDONLY)
 	if err != nil {
 		return nil, err
 	}
 	defer f.Close()
 
-	var (
-		s   = bufio.NewScanner(f)
-		out = []int{}
-	)
-
+	s := bufio.NewScanner(f)
 	for s.Scan() {
 		if t := s.Text(); t != "" {
 			pid, err := strconv.Atoi(t)
@@ -157,6 +157,13 @@ func readProcsFile(dir string) ([]int, error) {
 			out = append(out, pid)
 		}
 	}
+	if errors.Is(s.Err(), unix.ENOTSUP) && retry {
+		// For a threaded cgroup, read returns ENOTSUP, and we should
+		// read from cgroup.threads instead.
+		file = "cgroup.threads"
+		retry = false
+		goto again
+	}
 	return out, s.Err()
 }
 

libcontainer/cgroups/utils.go

@@ -69,10 +69,6 @@ func (m *mockCgroupManager) GetFreezerState() (configs.FreezerState, error) {
 	return configs.Thawed, nil
 }
 
-func (m *mockCgroupManager) GetEffectiveCPUs() string {
-	return ""
-}
-
 type mockProcess struct {
 	_pid    int
 	started uint64