/usr/share/gocode/src/github.com/opencontainers/runc/libcontainer/cgroups/systemd/apply_systemd.go is in golang-github-opencontainers-runc-dev 0.0.8+dfsg-2.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 | // +build linux
package systemd
import (
"errors"
"fmt"
"io/ioutil"
"os"
"path/filepath"
"strconv"
"strings"
"sync"
"time"
systemdDbus "github.com/coreos/go-systemd/dbus"
systemdUtil "github.com/coreos/go-systemd/util"
"github.com/godbus/dbus"
"github.com/opencontainers/runc/libcontainer/cgroups"
"github.com/opencontainers/runc/libcontainer/cgroups/fs"
"github.com/opencontainers/runc/libcontainer/configs"
)
type Manager struct {
mu sync.Mutex
Cgroups *configs.Cgroup
Paths map[string]string
}
type subsystem interface {
// Name returns the name of the subsystem.
Name() string
// Returns the stats, as 'stats', corresponding to the cgroup under 'path'.
GetStats(path string, stats *cgroups.Stats) error
// Set the cgroup represented by cgroup.
Set(path string, cgroup *configs.Cgroup) error
}
var errSubsystemDoesNotExist = errors.New("cgroup: subsystem does not exist")
type subsystemSet []subsystem
func (s subsystemSet) Get(name string) (subsystem, error) {
for _, ss := range s {
if ss.Name() == name {
return ss, nil
}
}
return nil, errSubsystemDoesNotExist
}
var subsystems = subsystemSet{
&fs.CpusetGroup{},
&fs.DevicesGroup{},
&fs.MemoryGroup{},
&fs.CpuGroup{},
&fs.CpuacctGroup{},
&fs.PidsGroup{},
&fs.BlkioGroup{},
&fs.HugetlbGroup{},
&fs.PerfEventGroup{},
&fs.FreezerGroup{},
&fs.NetPrioGroup{},
&fs.NetClsGroup{},
&fs.NameGroup{GroupName: "name=systemd"},
}
const (
testScopeWait = 4
)
var (
connLock sync.Mutex
theConn *systemdDbus.Conn
hasStartTransientUnit bool
hasTransientDefaultDependencies bool
)
func newProp(name string, units interface{}) systemdDbus.Property {
return systemdDbus.Property{
Name: name,
Value: dbus.MakeVariant(units),
}
}
func UseSystemd() bool {
if !systemdUtil.IsRunningSystemd() {
return false
}
connLock.Lock()
defer connLock.Unlock()
if theConn == nil {
var err error
theConn, err = systemdDbus.New()
if err != nil {
return false
}
// Assume we have StartTransientUnit
hasStartTransientUnit = true
// But if we get UnknownMethod error we don't
if _, err := theConn.StartTransientUnit("test.scope", "invalid", nil, nil); err != nil {
if dbusError, ok := err.(dbus.Error); ok {
if dbusError.Name == "org.freedesktop.DBus.Error.UnknownMethod" {
hasStartTransientUnit = false
return hasStartTransientUnit
}
}
}
// Ensure the scope name we use doesn't exist. Use the Pid to
// avoid collisions between multiple libcontainer users on a
// single host.
scope := fmt.Sprintf("libcontainer-%d-systemd-test-default-dependencies.scope", os.Getpid())
testScopeExists := true
for i := 0; i <= testScopeWait; i++ {
if _, err := theConn.StopUnit(scope, "replace", nil); err != nil {
if dbusError, ok := err.(dbus.Error); ok {
if strings.Contains(dbusError.Name, "org.freedesktop.systemd1.NoSuchUnit") {
testScopeExists = false
break
}
}
}
time.Sleep(time.Millisecond)
}
// Bail out if we can't kill this scope without testing for DefaultDependencies
if testScopeExists {
return hasStartTransientUnit
}
// Assume StartTransientUnit on a scope allows DefaultDependencies
hasTransientDefaultDependencies = true
ddf := newProp("DefaultDependencies", false)
if _, err := theConn.StartTransientUnit(scope, "replace", []systemdDbus.Property{ddf}, nil); err != nil {
if dbusError, ok := err.(dbus.Error); ok {
if strings.Contains(dbusError.Name, "org.freedesktop.DBus.Error.PropertyReadOnly") {
hasTransientDefaultDependencies = false
}
}
}
// Not critical because of the stop unit logic above.
theConn.StopUnit(scope, "replace", nil)
}
return hasStartTransientUnit
}
func getIfaceForUnit(unitName string) string {
if strings.HasSuffix(unitName, ".scope") {
return "Scope"
}
if strings.HasSuffix(unitName, ".service") {
return "Service"
}
return "Unit"
}
func (m *Manager) Apply(pid int) error {
var (
c = m.Cgroups
unitName = getUnitName(c)
slice = "system.slice"
properties []systemdDbus.Property
)
if c.Paths != nil {
paths := make(map[string]string)
for name, path := range c.Paths {
_, err := getSubsystemPath(m.Cgroups, name)
if err != nil {
// Don't fail if a cgroup hierarchy was not found, just skip this subsystem
if cgroups.IsNotFound(err) {
continue
}
return err
}
paths[name] = path
}
m.Paths = paths
return cgroups.EnterPid(m.Paths, pid)
}
if c.Parent != "" {
slice = c.Parent
}
properties = append(properties,
systemdDbus.PropSlice(slice),
systemdDbus.PropDescription("docker container "+c.Name),
newProp("PIDs", []uint32{uint32(pid)}),
)
// Always enable accounting, this gets us the same behaviour as the fs implementation,
// plus the kernel has some problems with joining the memory cgroup at a later time.
properties = append(properties,
newProp("MemoryAccounting", true),
newProp("CPUAccounting", true),
newProp("BlockIOAccounting", true))
if hasTransientDefaultDependencies {
properties = append(properties,
newProp("DefaultDependencies", false))
}
if c.Resources.Memory != 0 {
properties = append(properties,
newProp("MemoryLimit", uint64(c.Resources.Memory)))
}
if c.Resources.CpuShares != 0 {
properties = append(properties,
newProp("CPUShares", uint64(c.Resources.CpuShares)))
}
if c.Resources.BlkioWeight != 0 {
properties = append(properties,
newProp("BlockIOWeight", uint64(c.Resources.BlkioWeight)))
}
// We need to set kernel memory before processes join cgroup because
// kmem.limit_in_bytes can only be set when the cgroup is empty.
// And swap memory limit needs to be set after memory limit, only
// memory limit is handled by systemd, so it's kind of ugly here.
if c.Resources.KernelMemory > 0 {
if err := setKernelMemory(c); err != nil {
return err
}
}
if _, err := theConn.StartTransientUnit(unitName, "replace", properties, nil); err != nil {
return err
}
if err := joinDevices(c, pid); err != nil {
return err
}
// TODO: CpuQuota and CpuPeriod not available in systemd
// we need to manually join the cpu.cfs_quota_us and cpu.cfs_period_us
if err := joinCpu(c, pid); err != nil {
return err
}
// TODO: MemoryReservation and MemorySwap not available in systemd
if err := joinMemory(c, pid); err != nil {
return err
}
// we need to manually join the freezer, net_cls, net_prio, pids and cpuset cgroup in systemd
// because it does not currently support it via the dbus api.
if err := joinFreezer(c, pid); err != nil {
return err
}
if err := joinNetPrio(c, pid); err != nil {
return err
}
if err := joinNetCls(c, pid); err != nil {
return err
}
if err := joinPids(c, pid); err != nil {
return err
}
if err := joinCpuset(c, pid); err != nil {
return err
}
if err := joinHugetlb(c, pid); err != nil {
return err
}
if err := joinPerfEvent(c, pid); err != nil {
return err
}
// FIXME: Systemd does have `BlockIODeviceWeight` property, but we got problem
// using that (at least on systemd 208, see https://github.com/opencontainers/runc/libcontainer/pull/354),
// so use fs work around for now.
if err := joinBlkio(c, pid); err != nil {
return err
}
paths := make(map[string]string)
for _, s := range subsystems {
subsystemPath, err := getSubsystemPath(m.Cgroups, s.Name())
if err != nil {
// Don't fail if a cgroup hierarchy was not found, just skip this subsystem
if cgroups.IsNotFound(err) {
continue
}
return err
}
paths[s.Name()] = subsystemPath
}
m.Paths = paths
return nil
}
func (m *Manager) Destroy() error {
if m.Cgroups.Paths != nil {
return nil
}
m.mu.Lock()
defer m.mu.Unlock()
theConn.StopUnit(getUnitName(m.Cgroups), "replace", nil)
if err := cgroups.RemovePaths(m.Paths); err != nil {
return err
}
m.Paths = make(map[string]string)
return nil
}
func (m *Manager) GetPaths() map[string]string {
m.mu.Lock()
paths := m.Paths
m.mu.Unlock()
return paths
}
func writeFile(dir, file, data string) error {
// Normally dir should not be empty, one case is that cgroup subsystem
// is not mounted, we will get empty dir, and we want it fail here.
if dir == "" {
return fmt.Errorf("no such directory for %s.", file)
}
return ioutil.WriteFile(filepath.Join(dir, file), []byte(data), 0700)
}
func join(c *configs.Cgroup, subsystem string, pid int) (string, error) {
path, err := getSubsystemPath(c, subsystem)
if err != nil {
return "", err
}
if err := os.MkdirAll(path, 0755); err != nil {
return "", err
}
if err := writeFile(path, "cgroup.procs", strconv.Itoa(pid)); err != nil {
return "", err
}
return path, nil
}
func joinCpu(c *configs.Cgroup, pid int) error {
_, err := join(c, "cpu", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
func joinFreezer(c *configs.Cgroup, pid int) error {
_, err := join(c, "freezer", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
func joinNetPrio(c *configs.Cgroup, pid int) error {
_, err := join(c, "net_prio", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
func joinNetCls(c *configs.Cgroup, pid int) error {
_, err := join(c, "net_cls", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
func joinPids(c *configs.Cgroup, pid int) error {
_, err := join(c, "pids", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
// systemd represents slice heirarchy using `-`, so we need to follow suit when
// generating the path of slice. Essentially, test-a-b.slice becomes
// test.slice/test-a.slice/test-a-b.slice.
func expandSlice(slice string) (string, error) {
suffix := ".slice"
// Name has to end with ".slice", but can't be just ".slice".
if len(slice) < len(suffix) || !strings.HasSuffix(slice, suffix) {
return "", fmt.Errorf("invalid slice name: %s", slice)
}
// Path-separators are not allowed.
if strings.Contains(slice, "/") {
return "", fmt.Errorf("invalid slice name: %s", slice)
}
var path, prefix string
sliceName := strings.TrimSuffix(slice, suffix)
for _, component := range strings.Split(sliceName, "-") {
// test--a.slice isn't permitted, nor is -test.slice.
if component == "" {
return "", fmt.Errorf("invalid slice name: %s", slice)
}
// Append the component to the path and to the prefix.
path += prefix + component + suffix + "/"
prefix += component + "-"
}
return path, nil
}
func getSubsystemPath(c *configs.Cgroup, subsystem string) (string, error) {
mountpoint, err := cgroups.FindCgroupMountpoint(subsystem)
if err != nil {
return "", err
}
initPath, err := cgroups.GetInitCgroupDir(subsystem)
if err != nil {
return "", err
}
slice := "system.slice"
if c.Parent != "" {
slice = c.Parent
}
slice, err = expandSlice(slice)
if err != nil {
return "", err
}
return filepath.Join(mountpoint, initPath, slice, getUnitName(c)), nil
}
func (m *Manager) Freeze(state configs.FreezerState) error {
path, err := getSubsystemPath(m.Cgroups, "freezer")
if err != nil {
return err
}
prevState := m.Cgroups.Resources.Freezer
m.Cgroups.Resources.Freezer = state
freezer, err := subsystems.Get("freezer")
if err != nil {
return err
}
err = freezer.Set(path, m.Cgroups)
if err != nil {
m.Cgroups.Resources.Freezer = prevState
return err
}
return nil
}
func (m *Manager) GetPids() ([]int, error) {
path, err := getSubsystemPath(m.Cgroups, "devices")
if err != nil {
return nil, err
}
return cgroups.GetPids(path)
}
func (m *Manager) GetAllPids() ([]int, error) {
path, err := getSubsystemPath(m.Cgroups, "devices")
if err != nil {
return nil, err
}
return cgroups.GetAllPids(path)
}
func (m *Manager) GetStats() (*cgroups.Stats, error) {
m.mu.Lock()
defer m.mu.Unlock()
stats := cgroups.NewStats()
for name, path := range m.Paths {
sys, err := subsystems.Get(name)
if err == errSubsystemDoesNotExist || !cgroups.PathExists(path) {
continue
}
if err := sys.GetStats(path, stats); err != nil {
return nil, err
}
}
return stats, nil
}
func (m *Manager) Set(container *configs.Config) error {
for _, sys := range subsystems {
// Get the subsystem path, but don't error out for not found cgroups.
path, err := getSubsystemPath(container.Cgroups, sys.Name())
if err != nil && !cgroups.IsNotFound(err) {
return err
}
if err := sys.Set(path, container.Cgroups); err != nil {
return err
}
}
if m.Paths["cpu"] != "" {
if err := fs.CheckCpushares(m.Paths["cpu"], container.Cgroups.Resources.CpuShares); err != nil {
return err
}
}
return nil
}
func getUnitName(c *configs.Cgroup) string {
return fmt.Sprintf("%s-%s.scope", c.ScopePrefix, c.Name)
}
// Atm we can't use the systemd device support because of two missing things:
// * Support for wildcards to allow mknod on any device
// * Support for wildcards to allow /dev/pts support
//
// The second is available in more recent systemd as "char-pts", but not in e.g. v208 which is
// in wide use. When both these are available we will be able to switch, but need to keep the old
// implementation for backwards compat.
//
// Note: we can't use systemd to set up the initial limits, and then change the cgroup
// because systemd will re-write the device settings if it needs to re-apply the cgroup context.
// This happens at least for v208 when any sibling unit is started.
func joinDevices(c *configs.Cgroup, pid int) error {
_, err := join(c, "devices", pid)
// Even if it's `not found` error, we'll return err because devices cgroup
// is hard requirement for container security.
if err != nil {
return err
}
return nil
}
func setKernelMemory(c *configs.Cgroup) error {
path, err := getSubsystemPath(c, "memory")
if err != nil && !cgroups.IsNotFound(err) {
return err
}
if err := os.MkdirAll(path, 0755); err != nil {
return err
}
// This doesn't get called by manager.Set, so we need to do it here.
s := &fs.MemoryGroup{}
return s.SetKernelMemory(path, c)
}
func joinMemory(c *configs.Cgroup, pid int) error {
_, err := join(c, "memory", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
// systemd does not atm set up the cpuset controller, so we must manually
// join it. Additionally that is a very finicky controller where each
// level must have a full setup as the default for a new directory is "no cpus"
func joinCpuset(c *configs.Cgroup, pid int) error {
path, err := getSubsystemPath(c, "cpuset")
if err != nil && !cgroups.IsNotFound(err) {
return err
}
s := &fs.CpusetGroup{}
return s.ApplyDir(path, c, pid)
}
// `BlockIODeviceWeight` property of systemd does not work properly, and systemd
// expects device path instead of major minor numbers, which is also confusing
// for users. So we use fs work around for now.
func joinBlkio(c *configs.Cgroup, pid int) error {
_, err := join(c, "blkio", pid)
if err != nil {
return err
}
return nil
}
func joinHugetlb(c *configs.Cgroup, pid int) error {
_, err := join(c, "hugetlb", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
func joinPerfEvent(c *configs.Cgroup, pid int) error {
_, err := join(c, "perf_event", pid)
if err != nil && !cgroups.IsNotFound(err) {
return err
}
return nil
}
|