github.com/rawahars/moby@v24.0.4+incompatible/daemon/daemon.go (about)

     1  // Package daemon exposes the functions that occur on the host server
     2  // that the Docker daemon is running.
     3  //
     4  // In implementing the various functions of the daemon, there is often
     5  // a method-specific struct for configuring the runtime behavior.
     6  package daemon // import "github.com/docker/docker/daemon"
     7  
     8  import (
     9  	"context"
    10  	"fmt"
    11  	"net"
    12  	"net/url"
    13  	"os"
    14  	"path"
    15  	"path/filepath"
    16  	"runtime"
    17  	"strings"
    18  	"sync"
    19  	"time"
    20  
    21  	"github.com/containerd/containerd"
    22  	"github.com/containerd/containerd/defaults"
    23  	"github.com/containerd/containerd/pkg/dialer"
    24  	"github.com/containerd/containerd/pkg/userns"
    25  	"github.com/containerd/containerd/remotes/docker"
    26  	dist "github.com/docker/distribution"
    27  	"github.com/docker/distribution/reference"
    28  	"github.com/docker/docker/api/types"
    29  	containertypes "github.com/docker/docker/api/types/container"
    30  	registrytypes "github.com/docker/docker/api/types/registry"
    31  	"github.com/docker/docker/api/types/swarm"
    32  	"github.com/docker/docker/api/types/volume"
    33  	"github.com/docker/docker/builder"
    34  	"github.com/docker/docker/container"
    35  	executorpkg "github.com/docker/docker/daemon/cluster/executor"
    36  	"github.com/docker/docker/daemon/config"
    37  	ctrd "github.com/docker/docker/daemon/containerd"
    38  	"github.com/docker/docker/daemon/events"
    39  	_ "github.com/docker/docker/daemon/graphdriver/register" // register graph drivers
    40  	"github.com/docker/docker/daemon/images"
    41  	dlogger "github.com/docker/docker/daemon/logger"
    42  	"github.com/docker/docker/daemon/network"
    43  	"github.com/docker/docker/daemon/snapshotter"
    44  	"github.com/docker/docker/daemon/stats"
    45  	"github.com/docker/docker/distribution"
    46  	dmetadata "github.com/docker/docker/distribution/metadata"
    47  	"github.com/docker/docker/dockerversion"
    48  	"github.com/docker/docker/errdefs"
    49  	"github.com/docker/docker/image"
    50  	"github.com/docker/docker/layer"
    51  	libcontainerdtypes "github.com/docker/docker/libcontainerd/types"
    52  	"github.com/docker/docker/libnetwork"
    53  	"github.com/docker/docker/libnetwork/cluster"
    54  	nwconfig "github.com/docker/docker/libnetwork/config"
    55  	"github.com/docker/docker/pkg/authorization"
    56  	"github.com/docker/docker/pkg/fileutils"
    57  	"github.com/docker/docker/pkg/idtools"
    58  	"github.com/docker/docker/pkg/plugingetter"
    59  	"github.com/docker/docker/pkg/sysinfo"
    60  	"github.com/docker/docker/pkg/system"
    61  	"github.com/docker/docker/plugin"
    62  	pluginexec "github.com/docker/docker/plugin/executor/containerd"
    63  	refstore "github.com/docker/docker/reference"
    64  	"github.com/docker/docker/registry"
    65  	"github.com/docker/docker/runconfig"
    66  	volumesservice "github.com/docker/docker/volume/service"
    67  	"github.com/moby/buildkit/util/resolver"
    68  	resolverconfig "github.com/moby/buildkit/util/resolver/config"
    69  	"github.com/moby/locker"
    70  	"github.com/pkg/errors"
    71  	"github.com/sirupsen/logrus"
    72  	"go.etcd.io/bbolt"
    73  	"golang.org/x/sync/semaphore"
    74  	"google.golang.org/grpc"
    75  	"google.golang.org/grpc/backoff"
    76  	"google.golang.org/grpc/credentials/insecure"
    77  	"resenje.org/singleflight"
    78  )
    79  
    80  // Daemon holds information about the Docker daemon.
    81  type Daemon struct {
    82  	id                    string
    83  	repository            string
    84  	containers            container.Store
    85  	containersReplica     *container.ViewDB
    86  	execCommands          *container.ExecStore
    87  	imageService          ImageService
    88  	configStore           *config.Config
    89  	statsCollector        *stats.Collector
    90  	defaultLogConfig      containertypes.LogConfig
    91  	registryService       *registry.Service
    92  	EventsService         *events.Events
    93  	netController         *libnetwork.Controller
    94  	volumes               *volumesservice.VolumesService
    95  	root                  string
    96  	sysInfoOnce           sync.Once
    97  	sysInfo               *sysinfo.SysInfo
    98  	shutdown              bool
    99  	idMapping             idtools.IdentityMapping
   100  	PluginStore           *plugin.Store // TODO: remove
   101  	pluginManager         *plugin.Manager
   102  	linkIndex             *linkIndex
   103  	containerdCli         *containerd.Client
   104  	containerd            libcontainerdtypes.Client
   105  	defaultIsolation      containertypes.Isolation // Default isolation mode on Windows
   106  	clusterProvider       cluster.Provider
   107  	cluster               Cluster
   108  	genericResources      []swarm.GenericResource
   109  	metricsPluginListener net.Listener
   110  	ReferenceStore        refstore.Store
   111  
   112  	machineMemory uint64
   113  
   114  	seccompProfile     []byte
   115  	seccompProfilePath string
   116  
   117  	usageContainers singleflight.Group[struct{}, []*types.Container]
   118  	usageImages     singleflight.Group[struct{}, []*types.ImageSummary]
   119  	usageVolumes    singleflight.Group[struct{}, []*volume.Volume]
   120  	usageLayer      singleflight.Group[struct{}, int64]
   121  
   122  	pruneRunning int32
   123  	hosts        map[string]bool // hosts stores the addresses the daemon is listening on
   124  	startupDone  chan struct{}
   125  
   126  	attachmentStore       network.AttachmentStore
   127  	attachableNetworkLock *locker.Locker
   128  
   129  	// This is used for Windows which doesn't currently support running on containerd
   130  	// It stores metadata for the content store (used for manifest caching)
   131  	// This needs to be closed on daemon exit
   132  	mdDB *bbolt.DB
   133  
   134  	usesSnapshotter bool
   135  }
   136  
   137  // ID returns the daemon id
   138  func (daemon *Daemon) ID() string {
   139  	return daemon.id
   140  }
   141  
   142  // StoreHosts stores the addresses the daemon is listening on
   143  func (daemon *Daemon) StoreHosts(hosts []string) {
   144  	if daemon.hosts == nil {
   145  		daemon.hosts = make(map[string]bool)
   146  	}
   147  	for _, h := range hosts {
   148  		daemon.hosts[h] = true
   149  	}
   150  }
   151  
   152  // HasExperimental returns whether the experimental features of the daemon are enabled or not
   153  func (daemon *Daemon) HasExperimental() bool {
   154  	return daemon.configStore != nil && daemon.configStore.Experimental
   155  }
   156  
   157  // Features returns the features map from configStore
   158  func (daemon *Daemon) Features() *map[string]bool {
   159  	return &daemon.configStore.Features
   160  }
   161  
   162  // UsesSnapshotter returns true if feature flag to use containerd snapshotter is enabled
   163  func (daemon *Daemon) UsesSnapshotter() bool {
   164  	return daemon.usesSnapshotter
   165  }
   166  
   167  // RegistryHosts returns registry configuration in containerd resolvers format
   168  func (daemon *Daemon) RegistryHosts() docker.RegistryHosts {
   169  	var (
   170  		registryKey = "docker.io"
   171  		mirrors     = make([]string, len(daemon.configStore.Mirrors))
   172  		m           = map[string]resolverconfig.RegistryConfig{}
   173  	)
   174  	// must trim "https://" or "http://" prefix
   175  	for i, v := range daemon.configStore.Mirrors {
   176  		if uri, err := url.Parse(v); err == nil {
   177  			v = uri.Host
   178  		}
   179  		mirrors[i] = v
   180  	}
   181  	// set mirrors for default registry
   182  	m[registryKey] = resolverconfig.RegistryConfig{Mirrors: mirrors}
   183  
   184  	for _, v := range daemon.configStore.InsecureRegistries {
   185  		u, err := url.Parse(v)
   186  		if err != nil && !strings.HasPrefix(v, "http://") && !strings.HasPrefix(v, "https://") {
   187  			originalErr := err
   188  			u, err = url.Parse("http://" + v)
   189  			if err != nil {
   190  				err = originalErr
   191  			}
   192  		}
   193  		c := resolverconfig.RegistryConfig{}
   194  		if err == nil {
   195  			v = u.Host
   196  			t := true
   197  			if u.Scheme == "http" {
   198  				c.PlainHTTP = &t
   199  			} else {
   200  				c.Insecure = &t
   201  			}
   202  		}
   203  		m[v] = c
   204  	}
   205  
   206  	for k, v := range m {
   207  		v.TLSConfigDir = []string{registry.HostCertsDir(k)}
   208  		m[k] = v
   209  	}
   210  
   211  	certsDir := registry.CertsDir()
   212  	if fis, err := os.ReadDir(certsDir); err == nil {
   213  		for _, fi := range fis {
   214  			if _, ok := m[fi.Name()]; !ok {
   215  				m[fi.Name()] = resolverconfig.RegistryConfig{
   216  					TLSConfigDir: []string{filepath.Join(certsDir, fi.Name())},
   217  				}
   218  			}
   219  		}
   220  	}
   221  
   222  	return resolver.NewRegistryConfig(m)
   223  }
   224  
   225  // layerAccessor may be implemented by ImageService
   226  type layerAccessor interface {
   227  	GetLayerByID(cid string) (layer.RWLayer, error)
   228  }
   229  
   230  func (daemon *Daemon) restore() error {
   231  	var mapLock sync.Mutex
   232  	containers := make(map[string]*container.Container)
   233  
   234  	logrus.Info("Loading containers: start.")
   235  
   236  	dir, err := os.ReadDir(daemon.repository)
   237  	if err != nil {
   238  		return err
   239  	}
   240  
   241  	// parallelLimit is the maximum number of parallel startup jobs that we
   242  	// allow (this is the limited used for all startup semaphores). The multipler
   243  	// (128) was chosen after some fairly significant benchmarking -- don't change
   244  	// it unless you've tested it significantly (this value is adjusted if
   245  	// RLIMIT_NOFILE is small to avoid EMFILE).
   246  	parallelLimit := adjustParallelLimit(len(dir), 128*runtime.NumCPU())
   247  
   248  	// Re-used for all parallel startup jobs.
   249  	var group sync.WaitGroup
   250  	sem := semaphore.NewWeighted(int64(parallelLimit))
   251  
   252  	for _, v := range dir {
   253  		group.Add(1)
   254  		go func(id string) {
   255  			defer group.Done()
   256  			_ = sem.Acquire(context.Background(), 1)
   257  			defer sem.Release(1)
   258  
   259  			log := logrus.WithField("container", id)
   260  
   261  			c, err := daemon.load(id)
   262  			if err != nil {
   263  				log.WithError(err).Error("failed to load container")
   264  				return
   265  			}
   266  			if c.Driver != daemon.imageService.StorageDriver() {
   267  				// Ignore the container if it wasn't created with the current storage-driver
   268  				log.Debugf("not restoring container because it was created with another storage driver (%s)", c.Driver)
   269  				return
   270  			}
   271  			if accessor, ok := daemon.imageService.(layerAccessor); ok {
   272  				rwlayer, err := accessor.GetLayerByID(c.ID)
   273  				if err != nil {
   274  					log.WithError(err).Error("failed to load container mount")
   275  					return
   276  				}
   277  				c.RWLayer = rwlayer
   278  			}
   279  			log.WithFields(logrus.Fields{
   280  				"running": c.IsRunning(),
   281  				"paused":  c.IsPaused(),
   282  			}).Debug("loaded container")
   283  
   284  			mapLock.Lock()
   285  			containers[c.ID] = c
   286  			mapLock.Unlock()
   287  		}(v.Name())
   288  	}
   289  	group.Wait()
   290  
   291  	removeContainers := make(map[string]*container.Container)
   292  	restartContainers := make(map[*container.Container]chan struct{})
   293  	activeSandboxes := make(map[string]interface{})
   294  
   295  	for _, c := range containers {
   296  		group.Add(1)
   297  		go func(c *container.Container) {
   298  			defer group.Done()
   299  			_ = sem.Acquire(context.Background(), 1)
   300  			defer sem.Release(1)
   301  
   302  			log := logrus.WithField("container", c.ID)
   303  
   304  			if err := daemon.registerName(c); err != nil {
   305  				log.WithError(err).Errorf("failed to register container name: %s", c.Name)
   306  				mapLock.Lock()
   307  				delete(containers, c.ID)
   308  				mapLock.Unlock()
   309  				return
   310  			}
   311  			if err := daemon.Register(c); err != nil {
   312  				log.WithError(err).Error("failed to register container")
   313  				mapLock.Lock()
   314  				delete(containers, c.ID)
   315  				mapLock.Unlock()
   316  				return
   317  			}
   318  		}(c)
   319  	}
   320  	group.Wait()
   321  
   322  	for _, c := range containers {
   323  		group.Add(1)
   324  		go func(c *container.Container) {
   325  			defer group.Done()
   326  			_ = sem.Acquire(context.Background(), 1)
   327  			defer sem.Release(1)
   328  
   329  			log := logrus.WithField("container", c.ID)
   330  
   331  			if err := daemon.checkpointAndSave(c); err != nil {
   332  				log.WithError(err).Error("error saving backported mountspec to disk")
   333  			}
   334  
   335  			daemon.setStateCounter(c)
   336  
   337  			logger := func(c *container.Container) *logrus.Entry {
   338  				return log.WithFields(logrus.Fields{
   339  					"running":    c.IsRunning(),
   340  					"paused":     c.IsPaused(),
   341  					"restarting": c.IsRestarting(),
   342  				})
   343  			}
   344  
   345  			logger(c).Debug("restoring container")
   346  
   347  			var es *containerd.ExitStatus
   348  
   349  			if err := c.RestoreTask(context.Background(), daemon.containerd); err != nil && !errdefs.IsNotFound(err) {
   350  				logger(c).WithError(err).Error("failed to restore container with containerd")
   351  				return
   352  			}
   353  
   354  			alive := false
   355  			status := containerd.Unknown
   356  			if tsk, ok := c.Task(); ok {
   357  				s, err := tsk.Status(context.Background())
   358  				if err != nil {
   359  					logger(c).WithError(err).Error("failed to get task status")
   360  				} else {
   361  					status = s.Status
   362  					alive = status != containerd.Stopped
   363  					if !alive {
   364  						logger(c).Debug("cleaning up dead container process")
   365  						es, err = tsk.Delete(context.Background())
   366  						if err != nil && !errdefs.IsNotFound(err) {
   367  							logger(c).WithError(err).Error("failed to delete task from containerd")
   368  							return
   369  						}
   370  					} else if !daemon.configStore.LiveRestoreEnabled {
   371  						logger(c).Debug("shutting down container considered alive by containerd")
   372  						if err := daemon.shutdownContainer(c); err != nil && !errdefs.IsNotFound(err) {
   373  							log.WithError(err).Error("error shutting down container")
   374  							return
   375  						}
   376  						status = containerd.Stopped
   377  						alive = false
   378  						c.ResetRestartManager(false)
   379  					}
   380  				}
   381  			}
   382  			// If the containerd task for the container was not found, docker's view of the
   383  			// container state will be updated accordingly via SetStopped further down.
   384  
   385  			if c.IsRunning() || c.IsPaused() {
   386  				logger(c).Debug("syncing container on disk state with real state")
   387  
   388  				c.RestartManager().Cancel() // manually start containers because some need to wait for swarm networking
   389  
   390  				switch {
   391  				case c.IsPaused() && alive:
   392  					logger(c).WithField("state", status).Info("restored container paused")
   393  					switch status {
   394  					case containerd.Paused, containerd.Pausing:
   395  						// nothing to do
   396  					case containerd.Unknown, containerd.Stopped, "":
   397  						log.WithField("status", status).Error("unexpected status for paused container during restore")
   398  					default:
   399  						// running
   400  						c.Lock()
   401  						c.Paused = false
   402  						daemon.setStateCounter(c)
   403  						daemon.updateHealthMonitor(c)
   404  						if err := c.CheckpointTo(daemon.containersReplica); err != nil {
   405  							log.WithError(err).Error("failed to update paused container state")
   406  						}
   407  						c.Unlock()
   408  					}
   409  				case !c.IsPaused() && alive:
   410  					logger(c).Debug("restoring healthcheck")
   411  					c.Lock()
   412  					daemon.updateHealthMonitor(c)
   413  					c.Unlock()
   414  				}
   415  
   416  				if !alive {
   417  					logger(c).Debug("setting stopped state")
   418  					c.Lock()
   419  					var ces container.ExitStatus
   420  					if es != nil {
   421  						ces.ExitCode = int(es.ExitCode())
   422  						ces.ExitedAt = es.ExitTime()
   423  					} else {
   424  						ces.ExitCode = 255
   425  					}
   426  					c.SetStopped(&ces)
   427  					daemon.Cleanup(c)
   428  					if err := c.CheckpointTo(daemon.containersReplica); err != nil {
   429  						log.WithError(err).Error("failed to update stopped container state")
   430  					}
   431  					c.Unlock()
   432  					logger(c).Debug("set stopped state")
   433  				}
   434  
   435  				// we call Mount and then Unmount to get BaseFs of the container
   436  				if err := daemon.Mount(c); err != nil {
   437  					// The mount is unlikely to fail. However, in case mount fails
   438  					// the container should be allowed to restore here. Some functionalities
   439  					// (like docker exec -u user) might be missing but container is able to be
   440  					// stopped/restarted/removed.
   441  					// See #29365 for related information.
   442  					// The error is only logged here.
   443  					logger(c).WithError(err).Warn("failed to mount container to get BaseFs path")
   444  				} else {
   445  					if err := daemon.Unmount(c); err != nil {
   446  						logger(c).WithError(err).Warn("failed to umount container to get BaseFs path")
   447  					}
   448  				}
   449  
   450  				c.ResetRestartManager(false)
   451  				if !c.HostConfig.NetworkMode.IsContainer() && c.IsRunning() {
   452  					options, err := daemon.buildSandboxOptions(c)
   453  					if err != nil {
   454  						logger(c).WithError(err).Warn("failed to build sandbox option to restore container")
   455  					}
   456  					mapLock.Lock()
   457  					activeSandboxes[c.NetworkSettings.SandboxID] = options
   458  					mapLock.Unlock()
   459  				}
   460  			}
   461  
   462  			// get list of containers we need to restart
   463  
   464  			// Do not autostart containers which
   465  			// has endpoints in a swarm scope
   466  			// network yet since the cluster is
   467  			// not initialized yet. We will start
   468  			// it after the cluster is
   469  			// initialized.
   470  			if daemon.configStore.AutoRestart && c.ShouldRestart() && !c.NetworkSettings.HasSwarmEndpoint && c.HasBeenStartedBefore {
   471  				mapLock.Lock()
   472  				restartContainers[c] = make(chan struct{})
   473  				mapLock.Unlock()
   474  			} else if c.HostConfig != nil && c.HostConfig.AutoRemove {
   475  				mapLock.Lock()
   476  				removeContainers[c.ID] = c
   477  				mapLock.Unlock()
   478  			}
   479  
   480  			c.Lock()
   481  			if c.RemovalInProgress {
   482  				// We probably crashed in the middle of a removal, reset
   483  				// the flag.
   484  				//
   485  				// We DO NOT remove the container here as we do not
   486  				// know if the user had requested for either the
   487  				// associated volumes, network links or both to also
   488  				// be removed. So we put the container in the "dead"
   489  				// state and leave further processing up to them.
   490  				c.RemovalInProgress = false
   491  				c.Dead = true
   492  				if err := c.CheckpointTo(daemon.containersReplica); err != nil {
   493  					log.WithError(err).Error("failed to update RemovalInProgress container state")
   494  				} else {
   495  					log.Debugf("reset RemovalInProgress state for container")
   496  				}
   497  			}
   498  			c.Unlock()
   499  			logger(c).Debug("done restoring container")
   500  		}(c)
   501  	}
   502  	group.Wait()
   503  
   504  	// Initialize the network controller and configure network settings.
   505  	//
   506  	// Note that we cannot initialize the network controller earlier, as it
   507  	// needs to know if there's active sandboxes (running containers).
   508  	if err = daemon.initNetworkController(activeSandboxes); err != nil {
   509  		return fmt.Errorf("Error initializing network controller: %v", err)
   510  	}
   511  
   512  	// Now that all the containers are registered, register the links
   513  	for _, c := range containers {
   514  		group.Add(1)
   515  		go func(c *container.Container) {
   516  			_ = sem.Acquire(context.Background(), 1)
   517  
   518  			if err := daemon.registerLinks(c, c.HostConfig); err != nil {
   519  				logrus.WithField("container", c.ID).WithError(err).Error("failed to register link for container")
   520  			}
   521  
   522  			sem.Release(1)
   523  			group.Done()
   524  		}(c)
   525  	}
   526  	group.Wait()
   527  
   528  	for c, notifier := range restartContainers {
   529  		group.Add(1)
   530  		go func(c *container.Container, chNotify chan struct{}) {
   531  			_ = sem.Acquire(context.Background(), 1)
   532  
   533  			log := logrus.WithField("container", c.ID)
   534  
   535  			log.Debug("starting container")
   536  
   537  			// ignore errors here as this is a best effort to wait for children to be
   538  			//   running before we try to start the container
   539  			children := daemon.children(c)
   540  			timeout := time.NewTimer(5 * time.Second)
   541  			defer timeout.Stop()
   542  
   543  			for _, child := range children {
   544  				if notifier, exists := restartContainers[child]; exists {
   545  					select {
   546  					case <-notifier:
   547  					case <-timeout.C:
   548  					}
   549  				}
   550  			}
   551  
   552  			if err := daemon.prepareMountPoints(c); err != nil {
   553  				log.WithError(err).Error("failed to prepare mount points for container")
   554  			}
   555  			if err := daemon.containerStart(context.Background(), c, "", "", true); err != nil {
   556  				log.WithError(err).Error("failed to start container")
   557  			}
   558  			close(chNotify)
   559  
   560  			sem.Release(1)
   561  			group.Done()
   562  		}(c, notifier)
   563  	}
   564  	group.Wait()
   565  
   566  	for id := range removeContainers {
   567  		group.Add(1)
   568  		go func(cid string) {
   569  			_ = sem.Acquire(context.Background(), 1)
   570  
   571  			if err := daemon.ContainerRm(cid, &types.ContainerRmConfig{ForceRemove: true, RemoveVolume: true}); err != nil {
   572  				logrus.WithField("container", cid).WithError(err).Error("failed to remove container")
   573  			}
   574  
   575  			sem.Release(1)
   576  			group.Done()
   577  		}(id)
   578  	}
   579  	group.Wait()
   580  
   581  	// any containers that were started above would already have had this done,
   582  	// however we need to now prepare the mountpoints for the rest of the containers as well.
   583  	// This shouldn't cause any issue running on the containers that already had this run.
   584  	// This must be run after any containers with a restart policy so that containerized plugins
   585  	// can have a chance to be running before we try to initialize them.
   586  	for _, c := range containers {
   587  		// if the container has restart policy, do not
   588  		// prepare the mountpoints since it has been done on restarting.
   589  		// This is to speed up the daemon start when a restart container
   590  		// has a volume and the volume driver is not available.
   591  		if _, ok := restartContainers[c]; ok {
   592  			continue
   593  		} else if _, ok := removeContainers[c.ID]; ok {
   594  			// container is automatically removed, skip it.
   595  			continue
   596  		}
   597  
   598  		group.Add(1)
   599  		go func(c *container.Container) {
   600  			_ = sem.Acquire(context.Background(), 1)
   601  
   602  			if err := daemon.prepareMountPoints(c); err != nil {
   603  				logrus.WithField("container", c.ID).WithError(err).Error("failed to prepare mountpoints for container")
   604  			}
   605  
   606  			sem.Release(1)
   607  			group.Done()
   608  		}(c)
   609  	}
   610  	group.Wait()
   611  
   612  	logrus.Info("Loading containers: done.")
   613  
   614  	return nil
   615  }
   616  
   617  // RestartSwarmContainers restarts any autostart container which has a
   618  // swarm endpoint.
   619  func (daemon *Daemon) RestartSwarmContainers() {
   620  	ctx := context.Background()
   621  
   622  	// parallelLimit is the maximum number of parallel startup jobs that we
   623  	// allow (this is the limited used for all startup semaphores). The multipler
   624  	// (128) was chosen after some fairly significant benchmarking -- don't change
   625  	// it unless you've tested it significantly (this value is adjusted if
   626  	// RLIMIT_NOFILE is small to avoid EMFILE).
   627  	parallelLimit := adjustParallelLimit(len(daemon.List()), 128*runtime.NumCPU())
   628  
   629  	var group sync.WaitGroup
   630  	sem := semaphore.NewWeighted(int64(parallelLimit))
   631  
   632  	for _, c := range daemon.List() {
   633  		if !c.IsRunning() && !c.IsPaused() {
   634  			// Autostart all the containers which has a
   635  			// swarm endpoint now that the cluster is
   636  			// initialized.
   637  			if daemon.configStore.AutoRestart && c.ShouldRestart() && c.NetworkSettings.HasSwarmEndpoint && c.HasBeenStartedBefore {
   638  				group.Add(1)
   639  				go func(c *container.Container) {
   640  					if err := sem.Acquire(ctx, 1); err != nil {
   641  						// ctx is done.
   642  						group.Done()
   643  						return
   644  					}
   645  
   646  					if err := daemon.containerStart(ctx, c, "", "", true); err != nil {
   647  						logrus.WithField("container", c.ID).WithError(err).Error("failed to start swarm container")
   648  					}
   649  
   650  					sem.Release(1)
   651  					group.Done()
   652  				}(c)
   653  			}
   654  		}
   655  	}
   656  	group.Wait()
   657  }
   658  
   659  func (daemon *Daemon) children(c *container.Container) map[string]*container.Container {
   660  	return daemon.linkIndex.children(c)
   661  }
   662  
   663  // parents returns the names of the parent containers of the container
   664  // with the given name.
   665  func (daemon *Daemon) parents(c *container.Container) map[string]*container.Container {
   666  	return daemon.linkIndex.parents(c)
   667  }
   668  
   669  func (daemon *Daemon) registerLink(parent, child *container.Container, alias string) error {
   670  	fullName := path.Join(parent.Name, alias)
   671  	if err := daemon.containersReplica.ReserveName(fullName, child.ID); err != nil {
   672  		if errors.Is(err, container.ErrNameReserved) {
   673  			logrus.Warnf("error registering link for %s, to %s, as alias %s, ignoring: %v", parent.ID, child.ID, alias, err)
   674  			return nil
   675  		}
   676  		return err
   677  	}
   678  	daemon.linkIndex.link(parent, child, fullName)
   679  	return nil
   680  }
   681  
   682  // DaemonJoinsCluster informs the daemon has joined the cluster and provides
   683  // the handler to query the cluster component
   684  func (daemon *Daemon) DaemonJoinsCluster(clusterProvider cluster.Provider) {
   685  	daemon.setClusterProvider(clusterProvider)
   686  }
   687  
   688  // DaemonLeavesCluster informs the daemon has left the cluster
   689  func (daemon *Daemon) DaemonLeavesCluster() {
   690  	// Daemon is in charge of removing the attachable networks with
   691  	// connected containers when the node leaves the swarm
   692  	daemon.clearAttachableNetworks()
   693  	// We no longer need the cluster provider, stop it now so that
   694  	// the network agent will stop listening to cluster events.
   695  	daemon.setClusterProvider(nil)
   696  	// Wait for the networking cluster agent to stop
   697  	daemon.netController.AgentStopWait()
   698  	// Daemon is in charge of removing the ingress network when the
   699  	// node leaves the swarm. Wait for job to be done or timeout.
   700  	// This is called also on graceful daemon shutdown. We need to
   701  	// wait, because the ingress release has to happen before the
   702  	// network controller is stopped.
   703  
   704  	if done, err := daemon.ReleaseIngress(); err == nil {
   705  		timeout := time.NewTimer(5 * time.Second)
   706  		defer timeout.Stop()
   707  
   708  		select {
   709  		case <-done:
   710  		case <-timeout.C:
   711  			logrus.Warn("timeout while waiting for ingress network removal")
   712  		}
   713  	} else {
   714  		logrus.Warnf("failed to initiate ingress network removal: %v", err)
   715  	}
   716  
   717  	daemon.attachmentStore.ClearAttachments()
   718  }
   719  
   720  // setClusterProvider sets a component for querying the current cluster state.
   721  func (daemon *Daemon) setClusterProvider(clusterProvider cluster.Provider) {
   722  	daemon.clusterProvider = clusterProvider
   723  	daemon.netController.SetClusterProvider(clusterProvider)
   724  	daemon.attachableNetworkLock = locker.New()
   725  }
   726  
   727  // IsSwarmCompatible verifies if the current daemon
   728  // configuration is compatible with the swarm mode
   729  func (daemon *Daemon) IsSwarmCompatible() error {
   730  	if daemon.configStore == nil {
   731  		return nil
   732  	}
   733  	return daemon.configStore.IsSwarmCompatible()
   734  }
   735  
   736  // NewDaemon sets up everything for the daemon to be able to service
   737  // requests from the webserver.
   738  func NewDaemon(ctx context.Context, config *config.Config, pluginStore *plugin.Store, authzMiddleware *authorization.Middleware) (daemon *Daemon, err error) {
   739  	// Verify platform-specific requirements.
   740  	// TODO(thaJeztah): this should be called before we try to create the daemon; perhaps together with the config validation.
   741  	if err := checkSystem(); err != nil {
   742  		return nil, err
   743  	}
   744  
   745  	registryService, err := registry.NewService(config.ServiceOptions)
   746  	if err != nil {
   747  		return nil, err
   748  	}
   749  
   750  	// Ensure that we have a correct root key limit for launching containers.
   751  	if err := modifyRootKeyLimit(); err != nil {
   752  		logrus.Warnf("unable to modify root key limit, number of containers could be limited by this quota: %v", err)
   753  	}
   754  
   755  	// Ensure we have compatible and valid configuration options
   756  	if err := verifyDaemonSettings(config); err != nil {
   757  		return nil, err
   758  	}
   759  
   760  	// Do we have a disabled network?
   761  	config.DisableBridge = isBridgeNetworkDisabled(config)
   762  
   763  	// Setup the resolv.conf
   764  	setupResolvConf(config)
   765  
   766  	idMapping, err := setupRemappedRoot(config)
   767  	if err != nil {
   768  		return nil, err
   769  	}
   770  	rootIDs := idMapping.RootPair()
   771  	if err := setupDaemonProcess(config); err != nil {
   772  		return nil, err
   773  	}
   774  
   775  	// set up the tmpDir to use a canonical path
   776  	tmp, err := prepareTempDir(config.Root)
   777  	if err != nil {
   778  		return nil, fmt.Errorf("Unable to get the TempDir under %s: %s", config.Root, err)
   779  	}
   780  	realTmp, err := fileutils.ReadSymlinkedDirectory(tmp)
   781  	if err != nil {
   782  		return nil, fmt.Errorf("Unable to get the full path to the TempDir (%s): %s", tmp, err)
   783  	}
   784  	if isWindows {
   785  		if err := system.MkdirAll(realTmp, 0); err != nil {
   786  			return nil, fmt.Errorf("Unable to create the TempDir (%s): %s", realTmp, err)
   787  		}
   788  		os.Setenv("TEMP", realTmp)
   789  		os.Setenv("TMP", realTmp)
   790  	} else {
   791  		os.Setenv("TMPDIR", realTmp)
   792  	}
   793  
   794  	d := &Daemon{
   795  		configStore: config,
   796  		PluginStore: pluginStore,
   797  		startupDone: make(chan struct{}),
   798  	}
   799  
   800  	// TEST_INTEGRATION_USE_SNAPSHOTTER is used for integration tests only.
   801  	if os.Getenv("TEST_INTEGRATION_USE_SNAPSHOTTER") != "" {
   802  		d.usesSnapshotter = true
   803  	} else {
   804  		d.usesSnapshotter = config.Features["containerd-snapshotter"]
   805  	}
   806  
   807  	// Ensure the daemon is properly shutdown if there is a failure during
   808  	// initialization
   809  	defer func() {
   810  		if err != nil {
   811  			// Use a fresh context here. Passed context could be cancelled.
   812  			if err := d.Shutdown(context.Background()); err != nil {
   813  				logrus.Error(err)
   814  			}
   815  		}
   816  	}()
   817  
   818  	if err := d.setGenericResources(config); err != nil {
   819  		return nil, err
   820  	}
   821  	// set up SIGUSR1 handler on Unix-like systems, or a Win32 global event
   822  	// on Windows to dump Go routine stacks
   823  	stackDumpDir := config.Root
   824  	if execRoot := config.GetExecRoot(); execRoot != "" {
   825  		stackDumpDir = execRoot
   826  	}
   827  	d.setupDumpStackTrap(stackDumpDir)
   828  
   829  	if err := d.setupSeccompProfile(); err != nil {
   830  		return nil, err
   831  	}
   832  
   833  	// Set the default isolation mode (only applicable on Windows)
   834  	if err := d.setDefaultIsolation(); err != nil {
   835  		return nil, fmt.Errorf("error setting default isolation mode: %v", err)
   836  	}
   837  
   838  	if err := configureMaxThreads(config); err != nil {
   839  		logrus.Warnf("Failed to configure golang's threads limit: %v", err)
   840  	}
   841  
   842  	// ensureDefaultAppArmorProfile does nothing if apparmor is disabled
   843  	if err := ensureDefaultAppArmorProfile(); err != nil {
   844  		logrus.Errorf(err.Error())
   845  	}
   846  
   847  	daemonRepo := filepath.Join(config.Root, "containers")
   848  	if err := idtools.MkdirAllAndChown(daemonRepo, 0o710, idtools.Identity{
   849  		UID: idtools.CurrentIdentity().UID,
   850  		GID: rootIDs.GID,
   851  	}); err != nil {
   852  		return nil, err
   853  	}
   854  
   855  	// Create the directory where we'll store the runtime scripts (i.e. in
   856  	// order to support runtimeArgs)
   857  	if err = os.Mkdir(filepath.Join(config.Root, "runtimes"), 0o700); err != nil && !errors.Is(err, os.ErrExist) {
   858  		return nil, err
   859  	}
   860  	if err := d.loadRuntimes(); err != nil {
   861  		return nil, err
   862  	}
   863  
   864  	if isWindows {
   865  		// Note that permissions (0o700) are ignored on Windows; passing them to
   866  		// show intent only. We could consider using idtools.MkdirAndChown here
   867  		// to apply an ACL.
   868  		if err = os.Mkdir(filepath.Join(config.Root, "credentialspecs"), 0o700); err != nil && !errors.Is(err, os.ErrExist) {
   869  			return nil, err
   870  		}
   871  	}
   872  
   873  	d.registryService = registryService
   874  	dlogger.RegisterPluginGetter(d.PluginStore)
   875  
   876  	metricsSockPath, err := d.listenMetricsSock()
   877  	if err != nil {
   878  		return nil, err
   879  	}
   880  	registerMetricsPluginCallback(d.PluginStore, metricsSockPath)
   881  
   882  	backoffConfig := backoff.DefaultConfig
   883  	backoffConfig.MaxDelay = 3 * time.Second
   884  	connParams := grpc.ConnectParams{
   885  		Backoff: backoffConfig,
   886  	}
   887  	gopts := []grpc.DialOption{
   888  		// WithBlock makes sure that the following containerd request
   889  		// is reliable.
   890  		//
   891  		// NOTE: In one edge case with high load pressure, kernel kills
   892  		// dockerd, containerd and containerd-shims caused by OOM.
   893  		// When both dockerd and containerd restart, but containerd
   894  		// will take time to recover all the existing containers. Before
   895  		// containerd serving, dockerd will failed with gRPC error.
   896  		// That bad thing is that restore action will still ignore the
   897  		// any non-NotFound errors and returns running state for
   898  		// already stopped container. It is unexpected behavior. And
   899  		// we need to restart dockerd to make sure that anything is OK.
   900  		//
   901  		// It is painful. Add WithBlock can prevent the edge case. And
   902  		// n common case, the containerd will be serving in shortly.
   903  		// It is not harm to add WithBlock for containerd connection.
   904  		grpc.WithBlock(),
   905  
   906  		grpc.WithTransportCredentials(insecure.NewCredentials()),
   907  		grpc.WithConnectParams(connParams),
   908  		grpc.WithContextDialer(dialer.ContextDialer),
   909  
   910  		// TODO(stevvooe): We may need to allow configuration of this on the client.
   911  		grpc.WithDefaultCallOptions(grpc.MaxCallRecvMsgSize(defaults.DefaultMaxRecvMsgSize)),
   912  		grpc.WithDefaultCallOptions(grpc.MaxCallSendMsgSize(defaults.DefaultMaxSendMsgSize)),
   913  	}
   914  
   915  	if config.ContainerdAddr != "" {
   916  		d.containerdCli, err = containerd.New(config.ContainerdAddr, containerd.WithDefaultNamespace(config.ContainerdNamespace), containerd.WithDialOpts(gopts), containerd.WithTimeout(60*time.Second))
   917  		if err != nil {
   918  			return nil, errors.Wrapf(err, "failed to dial %q", config.ContainerdAddr)
   919  		}
   920  	}
   921  
   922  	createPluginExec := func(m *plugin.Manager) (plugin.Executor, error) {
   923  		var pluginCli *containerd.Client
   924  
   925  		if config.ContainerdAddr != "" {
   926  			pluginCli, err = containerd.New(config.ContainerdAddr, containerd.WithDefaultNamespace(config.ContainerdPluginNamespace), containerd.WithDialOpts(gopts), containerd.WithTimeout(60*time.Second))
   927  			if err != nil {
   928  				return nil, errors.Wrapf(err, "failed to dial %q", config.ContainerdAddr)
   929  			}
   930  		}
   931  
   932  		var (
   933  			shim     string
   934  			shimOpts interface{}
   935  		)
   936  		if runtime.GOOS != "windows" {
   937  			shim, shimOpts, err = d.getRuntime(config.GetDefaultRuntimeName())
   938  			if err != nil {
   939  				return nil, err
   940  			}
   941  		}
   942  		return pluginexec.New(ctx, getPluginExecRoot(config), pluginCli, config.ContainerdPluginNamespace, m, shim, shimOpts)
   943  	}
   944  
   945  	// Plugin system initialization should happen before restore. Do not change order.
   946  	d.pluginManager, err = plugin.NewManager(plugin.ManagerConfig{
   947  		Root:               filepath.Join(config.Root, "plugins"),
   948  		ExecRoot:           getPluginExecRoot(config),
   949  		Store:              d.PluginStore,
   950  		CreateExecutor:     createPluginExec,
   951  		RegistryService:    registryService,
   952  		LiveRestoreEnabled: config.LiveRestoreEnabled,
   953  		LogPluginEvent:     d.LogPluginEvent, // todo: make private
   954  		AuthzMiddleware:    authzMiddleware,
   955  	})
   956  	if err != nil {
   957  		return nil, errors.Wrap(err, "couldn't create plugin manager")
   958  	}
   959  
   960  	if err := d.setupDefaultLogConfig(); err != nil {
   961  		return nil, err
   962  	}
   963  
   964  	d.volumes, err = volumesservice.NewVolumeService(config.Root, d.PluginStore, rootIDs, d)
   965  	if err != nil {
   966  		return nil, err
   967  	}
   968  
   969  	// Check if Devices cgroup is mounted, it is hard requirement for container security,
   970  	// on Linux.
   971  	//
   972  	// Important: we call getSysInfo() directly here, without storing the results,
   973  	// as networking has not yet been set up, so we only have partial system info
   974  	// at this point.
   975  	//
   976  	// TODO(thaJeztah) add a utility to only collect the CgroupDevicesEnabled information
   977  	if runtime.GOOS == "linux" && !userns.RunningInUserNS() && !getSysInfo(d).CgroupDevicesEnabled {
   978  		return nil, errors.New("Devices cgroup isn't mounted")
   979  	}
   980  
   981  	d.id, err = loadOrCreateID(filepath.Join(config.Root, "engine-id"))
   982  	if err != nil {
   983  		return nil, err
   984  	}
   985  	d.repository = daemonRepo
   986  	d.containers = container.NewMemoryStore()
   987  	if d.containersReplica, err = container.NewViewDB(); err != nil {
   988  		return nil, err
   989  	}
   990  	d.execCommands = container.NewExecStore()
   991  	d.statsCollector = d.newStatsCollector(1 * time.Second)
   992  
   993  	d.EventsService = events.New()
   994  	d.root = config.Root
   995  	d.idMapping = idMapping
   996  
   997  	d.linkIndex = newLinkIndex()
   998  
   999  	// On Windows we don't support the environment variable, or a user supplied graphdriver
  1000  	// Unix platforms however run a single graphdriver for all containers, and it can
  1001  	// be set through an environment variable, a daemon start parameter, or chosen through
  1002  	// initialization of the layerstore through driver priority order for example.
  1003  	driverName := os.Getenv("DOCKER_DRIVER")
  1004  	if isWindows {
  1005  		driverName = "windowsfilter"
  1006  	} else if driverName != "" {
  1007  		logrus.Infof("Setting the storage driver from the $DOCKER_DRIVER environment variable (%s)", driverName)
  1008  	} else {
  1009  		driverName = config.GraphDriver
  1010  	}
  1011  
  1012  	if d.UsesSnapshotter() {
  1013  		if os.Getenv("TEST_INTEGRATION_USE_SNAPSHOTTER") != "" {
  1014  			logrus.Warn("Enabling containerd snapshotter through the $TEST_INTEGRATION_USE_SNAPSHOTTER environment variable. This should only be used for testing.")
  1015  		}
  1016  		logrus.Info("Starting daemon with containerd snapshotter integration enabled")
  1017  
  1018  		// FIXME(thaJeztah): implement automatic snapshotter-selection similar to graph-driver selection; see https://github.com/moby/moby/issues/44076
  1019  		if driverName == "" {
  1020  			driverName = containerd.DefaultSnapshotter
  1021  		}
  1022  
  1023  		// Configure and validate the kernels security support. Note this is a Linux/FreeBSD
  1024  		// operation only, so it is safe to pass *just* the runtime OS graphdriver.
  1025  		if err := configureKernelSecuritySupport(config, driverName); err != nil {
  1026  			return nil, err
  1027  		}
  1028  		d.imageService = ctrd.NewService(ctrd.ImageServiceConfig{
  1029  			Client:          d.containerdCli,
  1030  			Containers:      d.containers,
  1031  			Snapshotter:     driverName,
  1032  			HostsProvider:   d,
  1033  			Registry:        d.registryService,
  1034  			EventsService:   d.EventsService,
  1035  			RefCountMounter: snapshotter.NewMounter(config.Root, driverName, idMapping),
  1036  		})
  1037  	} else {
  1038  		layerStore, err := layer.NewStoreFromOptions(layer.StoreOptions{
  1039  			Root:                      config.Root,
  1040  			MetadataStorePathTemplate: filepath.Join(config.Root, "image", "%s", "layerdb"),
  1041  			GraphDriver:               driverName,
  1042  			GraphDriverOptions:        config.GraphOptions,
  1043  			IDMapping:                 idMapping,
  1044  			PluginGetter:              d.PluginStore,
  1045  			ExperimentalEnabled:       config.Experimental,
  1046  		})
  1047  		if err != nil {
  1048  			return nil, err
  1049  		}
  1050  
  1051  		// Configure and validate the kernels security support. Note this is a Linux/FreeBSD
  1052  		// operation only, so it is safe to pass *just* the runtime OS graphdriver.
  1053  		if err := configureKernelSecuritySupport(config, layerStore.DriverName()); err != nil {
  1054  			return nil, err
  1055  		}
  1056  
  1057  		imageRoot := filepath.Join(config.Root, "image", layerStore.DriverName())
  1058  		ifs, err := image.NewFSStoreBackend(filepath.Join(imageRoot, "imagedb"))
  1059  		if err != nil {
  1060  			return nil, err
  1061  		}
  1062  
  1063  		// We have a single tag/reference store for the daemon globally. However, it's
  1064  		// stored under the graphdriver. On host platforms which only support a single
  1065  		// container OS, but multiple selectable graphdrivers, this means depending on which
  1066  		// graphdriver is chosen, the global reference store is under there. For
  1067  		// platforms which support multiple container operating systems, this is slightly
  1068  		// more problematic as where does the global ref store get located? Fortunately,
  1069  		// for Windows, which is currently the only daemon supporting multiple container
  1070  		// operating systems, the list of graphdrivers available isn't user configurable.
  1071  		// For backwards compatibility, we just put it under the windowsfilter
  1072  		// directory regardless.
  1073  		refStoreLocation := filepath.Join(imageRoot, `repositories.json`)
  1074  		rs, err := refstore.NewReferenceStore(refStoreLocation)
  1075  		if err != nil {
  1076  			return nil, fmt.Errorf("Couldn't create reference store repository: %s", err)
  1077  		}
  1078  		d.ReferenceStore = rs
  1079  
  1080  		imageStore, err := image.NewImageStore(ifs, layerStore)
  1081  		if err != nil {
  1082  			return nil, err
  1083  		}
  1084  
  1085  		distributionMetadataStore, err := dmetadata.NewFSMetadataStore(filepath.Join(imageRoot, "distribution"))
  1086  		if err != nil {
  1087  			return nil, err
  1088  		}
  1089  
  1090  		imgSvcConfig := images.ImageServiceConfig{
  1091  			ContainerStore:            d.containers,
  1092  			DistributionMetadataStore: distributionMetadataStore,
  1093  			EventsService:             d.EventsService,
  1094  			ImageStore:                imageStore,
  1095  			LayerStore:                layerStore,
  1096  			MaxConcurrentDownloads:    config.MaxConcurrentDownloads,
  1097  			MaxConcurrentUploads:      config.MaxConcurrentUploads,
  1098  			MaxDownloadAttempts:       config.MaxDownloadAttempts,
  1099  			ReferenceStore:            rs,
  1100  			RegistryService:           registryService,
  1101  			ContentNamespace:          config.ContainerdNamespace,
  1102  		}
  1103  
  1104  		// containerd is not currently supported with Windows.
  1105  		// So sometimes d.containerdCli will be nil
  1106  		// In that case we'll create a local content store... but otherwise we'll use containerd
  1107  		if d.containerdCli != nil {
  1108  			imgSvcConfig.Leases = d.containerdCli.LeasesService()
  1109  			imgSvcConfig.ContentStore = d.containerdCli.ContentStore()
  1110  		} else {
  1111  			cs, lm, err := d.configureLocalContentStore(config.ContainerdNamespace)
  1112  			if err != nil {
  1113  				return nil, err
  1114  			}
  1115  			imgSvcConfig.ContentStore = cs
  1116  			imgSvcConfig.Leases = lm
  1117  		}
  1118  
  1119  		// TODO: imageStore, distributionMetadataStore, and ReferenceStore are only
  1120  		// used above to run migration. They could be initialized in ImageService
  1121  		// if migration is called from daemon/images. layerStore might move as well.
  1122  		d.imageService = images.NewImageService(imgSvcConfig)
  1123  
  1124  		logrus.Debugf("Max Concurrent Downloads: %d", imgSvcConfig.MaxConcurrentDownloads)
  1125  		logrus.Debugf("Max Concurrent Uploads: %d", imgSvcConfig.MaxConcurrentUploads)
  1126  		logrus.Debugf("Max Download Attempts: %d", imgSvcConfig.MaxDownloadAttempts)
  1127  	}
  1128  
  1129  	go d.execCommandGC()
  1130  
  1131  	if err := d.initLibcontainerd(ctx); err != nil {
  1132  		return nil, err
  1133  	}
  1134  
  1135  	if err := d.restore(); err != nil {
  1136  		return nil, err
  1137  	}
  1138  	close(d.startupDone)
  1139  
  1140  	info := d.SystemInfo()
  1141  	for _, w := range info.Warnings {
  1142  		logrus.Warn(w)
  1143  	}
  1144  
  1145  	engineInfo.WithValues(
  1146  		dockerversion.Version,
  1147  		dockerversion.GitCommit,
  1148  		info.Architecture,
  1149  		info.Driver,
  1150  		info.KernelVersion,
  1151  		info.OperatingSystem,
  1152  		info.OSType,
  1153  		info.OSVersion,
  1154  		info.ID,
  1155  	).Set(1)
  1156  	engineCpus.Set(float64(info.NCPU))
  1157  	engineMemory.Set(float64(info.MemTotal))
  1158  
  1159  	logrus.WithFields(logrus.Fields{
  1160  		"version":     dockerversion.Version,
  1161  		"commit":      dockerversion.GitCommit,
  1162  		"graphdriver": d.ImageService().StorageDriver(),
  1163  	}).Info("Docker daemon")
  1164  
  1165  	return d, nil
  1166  }
  1167  
  1168  // DistributionServices returns services controlling daemon storage
  1169  func (daemon *Daemon) DistributionServices() images.DistributionServices {
  1170  	return daemon.imageService.DistributionServices()
  1171  }
  1172  
  1173  func (daemon *Daemon) waitForStartupDone() {
  1174  	<-daemon.startupDone
  1175  }
  1176  
  1177  func (daemon *Daemon) shutdownContainer(c *container.Container) error {
  1178  	// If container failed to exit in stopTimeout seconds of SIGTERM, then using the force
  1179  	if err := daemon.containerStop(context.TODO(), c, containertypes.StopOptions{}); err != nil {
  1180  		return fmt.Errorf("Failed to stop container %s with error: %v", c.ID, err)
  1181  	}
  1182  
  1183  	// Wait without timeout for the container to exit.
  1184  	// Ignore the result.
  1185  	<-c.Wait(context.Background(), container.WaitConditionNotRunning)
  1186  	return nil
  1187  }
  1188  
  1189  // ShutdownTimeout returns the timeout (in seconds) before containers are forcibly
  1190  // killed during shutdown. The default timeout can be configured both on the daemon
  1191  // and per container, and the longest timeout will be used. A grace-period of
  1192  // 5 seconds is added to the configured timeout.
  1193  //
  1194  // A negative (-1) timeout means "indefinitely", which means that containers
  1195  // are not forcibly killed, and the daemon shuts down after all containers exit.
  1196  func (daemon *Daemon) ShutdownTimeout() int {
  1197  	shutdownTimeout := daemon.configStore.ShutdownTimeout
  1198  	if shutdownTimeout < 0 {
  1199  		return -1
  1200  	}
  1201  	if daemon.containers == nil {
  1202  		return shutdownTimeout
  1203  	}
  1204  
  1205  	graceTimeout := 5
  1206  	for _, c := range daemon.containers.List() {
  1207  		stopTimeout := c.StopTimeout()
  1208  		if stopTimeout < 0 {
  1209  			return -1
  1210  		}
  1211  		if stopTimeout+graceTimeout > shutdownTimeout {
  1212  			shutdownTimeout = stopTimeout + graceTimeout
  1213  		}
  1214  	}
  1215  	return shutdownTimeout
  1216  }
  1217  
  1218  // Shutdown stops the daemon.
  1219  func (daemon *Daemon) Shutdown(ctx context.Context) error {
  1220  	daemon.shutdown = true
  1221  	// Keep mounts and networking running on daemon shutdown if
  1222  	// we are to keep containers running and restore them.
  1223  
  1224  	if daemon.configStore.LiveRestoreEnabled && daemon.containers != nil {
  1225  		// check if there are any running containers, if none we should do some cleanup
  1226  		if ls, err := daemon.Containers(ctx, &types.ContainerListOptions{}); len(ls) != 0 || err != nil {
  1227  			// metrics plugins still need some cleanup
  1228  			daemon.cleanupMetricsPlugins()
  1229  			return err
  1230  		}
  1231  	}
  1232  
  1233  	if daemon.containers != nil {
  1234  		logrus.Debugf("daemon configured with a %d seconds minimum shutdown timeout", daemon.configStore.ShutdownTimeout)
  1235  		logrus.Debugf("start clean shutdown of all containers with a %d seconds timeout...", daemon.ShutdownTimeout())
  1236  		daemon.containers.ApplyAll(func(c *container.Container) {
  1237  			if !c.IsRunning() {
  1238  				return
  1239  			}
  1240  			log := logrus.WithField("container", c.ID)
  1241  			log.Debug("shutting down container")
  1242  			if err := daemon.shutdownContainer(c); err != nil {
  1243  				log.WithError(err).Error("failed to shut down container")
  1244  				return
  1245  			}
  1246  			if mountid, err := daemon.imageService.GetLayerMountID(c.ID); err == nil {
  1247  				daemon.cleanupMountsByID(mountid)
  1248  			}
  1249  			log.Debugf("shut down container")
  1250  		})
  1251  	}
  1252  
  1253  	if daemon.volumes != nil {
  1254  		if err := daemon.volumes.Shutdown(); err != nil {
  1255  			logrus.Errorf("Error shutting down volume store: %v", err)
  1256  		}
  1257  	}
  1258  
  1259  	if daemon.imageService != nil {
  1260  		if err := daemon.imageService.Cleanup(); err != nil {
  1261  			logrus.Error(err)
  1262  		}
  1263  	}
  1264  
  1265  	// If we are part of a cluster, clean up cluster's stuff
  1266  	if daemon.clusterProvider != nil {
  1267  		logrus.Debugf("start clean shutdown of cluster resources...")
  1268  		daemon.DaemonLeavesCluster()
  1269  	}
  1270  
  1271  	daemon.cleanupMetricsPlugins()
  1272  
  1273  	// Shutdown plugins after containers and layerstore. Don't change the order.
  1274  	daemon.pluginShutdown()
  1275  
  1276  	// trigger libnetwork Stop only if it's initialized
  1277  	if daemon.netController != nil {
  1278  		daemon.netController.Stop()
  1279  	}
  1280  
  1281  	if daemon.containerdCli != nil {
  1282  		daemon.containerdCli.Close()
  1283  	}
  1284  
  1285  	if daemon.mdDB != nil {
  1286  		daemon.mdDB.Close()
  1287  	}
  1288  
  1289  	return daemon.cleanupMounts()
  1290  }
  1291  
  1292  // Mount sets container.BaseFS
  1293  func (daemon *Daemon) Mount(container *container.Container) error {
  1294  	return daemon.imageService.Mount(context.Background(), container)
  1295  }
  1296  
  1297  // Unmount unsets the container base filesystem
  1298  func (daemon *Daemon) Unmount(container *container.Container) error {
  1299  	return daemon.imageService.Unmount(context.Background(), container)
  1300  }
  1301  
  1302  // Subnets return the IPv4 and IPv6 subnets of networks that are manager by Docker.
  1303  func (daemon *Daemon) Subnets() ([]net.IPNet, []net.IPNet) {
  1304  	var v4Subnets []net.IPNet
  1305  	var v6Subnets []net.IPNet
  1306  
  1307  	managedNetworks := daemon.netController.Networks()
  1308  
  1309  	for _, managedNetwork := range managedNetworks {
  1310  		v4infos, v6infos := managedNetwork.Info().IpamInfo()
  1311  		for _, info := range v4infos {
  1312  			if info.IPAMData.Pool != nil {
  1313  				v4Subnets = append(v4Subnets, *info.IPAMData.Pool)
  1314  			}
  1315  		}
  1316  		for _, info := range v6infos {
  1317  			if info.IPAMData.Pool != nil {
  1318  				v6Subnets = append(v6Subnets, *info.IPAMData.Pool)
  1319  			}
  1320  		}
  1321  	}
  1322  
  1323  	return v4Subnets, v6Subnets
  1324  }
  1325  
  1326  // prepareTempDir prepares and returns the default directory to use
  1327  // for temporary files.
  1328  // If it doesn't exist, it is created. If it exists, its content is removed.
  1329  func prepareTempDir(rootDir string) (string, error) {
  1330  	var tmpDir string
  1331  	if tmpDir = os.Getenv("DOCKER_TMPDIR"); tmpDir == "" {
  1332  		tmpDir = filepath.Join(rootDir, "tmp")
  1333  		newName := tmpDir + "-old"
  1334  		if err := os.Rename(tmpDir, newName); err == nil {
  1335  			go func() {
  1336  				if err := os.RemoveAll(newName); err != nil {
  1337  					logrus.Warnf("failed to delete old tmp directory: %s", newName)
  1338  				}
  1339  			}()
  1340  		} else if !os.IsNotExist(err) {
  1341  			logrus.Warnf("failed to rename %s for background deletion: %s. Deleting synchronously", tmpDir, err)
  1342  			if err := os.RemoveAll(tmpDir); err != nil {
  1343  				logrus.Warnf("failed to delete old tmp directory: %s", tmpDir)
  1344  			}
  1345  		}
  1346  	}
  1347  	return tmpDir, idtools.MkdirAllAndChown(tmpDir, 0o700, idtools.CurrentIdentity())
  1348  }
  1349  
  1350  func (daemon *Daemon) setGenericResources(conf *config.Config) error {
  1351  	genericResources, err := config.ParseGenericResources(conf.NodeGenericResources)
  1352  	if err != nil {
  1353  		return err
  1354  	}
  1355  
  1356  	daemon.genericResources = genericResources
  1357  
  1358  	return nil
  1359  }
  1360  
  1361  // IsShuttingDown tells whether the daemon is shutting down or not
  1362  func (daemon *Daemon) IsShuttingDown() bool {
  1363  	return daemon.shutdown
  1364  }
  1365  
  1366  func isBridgeNetworkDisabled(conf *config.Config) bool {
  1367  	return conf.BridgeConfig.Iface == config.DisableNetworkBridge
  1368  }
  1369  
  1370  func (daemon *Daemon) networkOptions(pg plugingetter.PluginGetter, activeSandboxes map[string]interface{}) ([]nwconfig.Option, error) {
  1371  	options := []nwconfig.Option{}
  1372  	if daemon.configStore == nil {
  1373  		return options, nil
  1374  	}
  1375  	conf := daemon.configStore
  1376  	dd := runconfig.DefaultDaemonNetworkMode()
  1377  
  1378  	options = []nwconfig.Option{
  1379  		nwconfig.OptionDataDir(conf.Root),
  1380  		nwconfig.OptionExecRoot(conf.GetExecRoot()),
  1381  		nwconfig.OptionDefaultDriver(string(dd)),
  1382  		nwconfig.OptionDefaultNetwork(dd.NetworkName()),
  1383  		nwconfig.OptionLabels(conf.Labels),
  1384  		nwconfig.OptionNetworkControlPlaneMTU(conf.NetworkControlPlaneMTU),
  1385  		driverOptions(conf),
  1386  	}
  1387  
  1388  	if len(conf.NetworkConfig.DefaultAddressPools.Value()) > 0 {
  1389  		options = append(options, nwconfig.OptionDefaultAddressPoolConfig(conf.NetworkConfig.DefaultAddressPools.Value()))
  1390  	}
  1391  	if conf.LiveRestoreEnabled && len(activeSandboxes) != 0 {
  1392  		options = append(options, nwconfig.OptionActiveSandboxes(activeSandboxes))
  1393  	}
  1394  	if pg != nil {
  1395  		options = append(options, nwconfig.OptionPluginGetter(pg))
  1396  	}
  1397  
  1398  	return options, nil
  1399  }
  1400  
  1401  // GetCluster returns the cluster
  1402  func (daemon *Daemon) GetCluster() Cluster {
  1403  	return daemon.cluster
  1404  }
  1405  
  1406  // SetCluster sets the cluster
  1407  func (daemon *Daemon) SetCluster(cluster Cluster) {
  1408  	daemon.cluster = cluster
  1409  }
  1410  
  1411  func (daemon *Daemon) pluginShutdown() {
  1412  	manager := daemon.pluginManager
  1413  	// Check for a valid manager object. In error conditions, daemon init can fail
  1414  	// and shutdown called, before plugin manager is initialized.
  1415  	if manager != nil {
  1416  		manager.Shutdown()
  1417  	}
  1418  }
  1419  
  1420  // PluginManager returns current pluginManager associated with the daemon
  1421  func (daemon *Daemon) PluginManager() *plugin.Manager { // set up before daemon to avoid this method
  1422  	return daemon.pluginManager
  1423  }
  1424  
  1425  // PluginGetter returns current pluginStore associated with the daemon
  1426  func (daemon *Daemon) PluginGetter() *plugin.Store {
  1427  	return daemon.PluginStore
  1428  }
  1429  
  1430  // CreateDaemonRoot creates the root for the daemon
  1431  func CreateDaemonRoot(config *config.Config) error {
  1432  	// get the canonical path to the Docker root directory
  1433  	var realRoot string
  1434  	if _, err := os.Stat(config.Root); err != nil && os.IsNotExist(err) {
  1435  		realRoot = config.Root
  1436  	} else {
  1437  		realRoot, err = fileutils.ReadSymlinkedDirectory(config.Root)
  1438  		if err != nil {
  1439  			return fmt.Errorf("Unable to get the full path to root (%s): %s", config.Root, err)
  1440  		}
  1441  	}
  1442  
  1443  	idMapping, err := setupRemappedRoot(config)
  1444  	if err != nil {
  1445  		return err
  1446  	}
  1447  	return setupDaemonRoot(config, realRoot, idMapping.RootPair())
  1448  }
  1449  
  1450  // checkpointAndSave grabs a container lock to safely call container.CheckpointTo
  1451  func (daemon *Daemon) checkpointAndSave(container *container.Container) error {
  1452  	container.Lock()
  1453  	defer container.Unlock()
  1454  	if err := container.CheckpointTo(daemon.containersReplica); err != nil {
  1455  		return fmt.Errorf("Error saving container state: %v", err)
  1456  	}
  1457  	return nil
  1458  }
  1459  
  1460  // because the CLI sends a -1 when it wants to unset the swappiness value
  1461  // we need to clear it on the server side
  1462  func fixMemorySwappiness(resources *containertypes.Resources) {
  1463  	if resources.MemorySwappiness != nil && *resources.MemorySwappiness == -1 {
  1464  		resources.MemorySwappiness = nil
  1465  	}
  1466  }
  1467  
  1468  // GetAttachmentStore returns current attachment store associated with the daemon
  1469  func (daemon *Daemon) GetAttachmentStore() *network.AttachmentStore {
  1470  	return &daemon.attachmentStore
  1471  }
  1472  
  1473  // IdentityMapping returns uid/gid mapping or a SID (in the case of Windows) for the builder
  1474  func (daemon *Daemon) IdentityMapping() idtools.IdentityMapping {
  1475  	return daemon.idMapping
  1476  }
  1477  
  1478  // ImageService returns the Daemon's ImageService
  1479  func (daemon *Daemon) ImageService() ImageService {
  1480  	return daemon.imageService
  1481  }
  1482  
  1483  // ImageBackend returns an image-backend for Swarm and the distribution router.
  1484  func (daemon *Daemon) ImageBackend() executorpkg.ImageBackend {
  1485  	return &imageBackend{
  1486  		ImageService:    daemon.imageService,
  1487  		registryService: daemon.registryService,
  1488  	}
  1489  }
  1490  
  1491  // RegistryService returns the Daemon's RegistryService
  1492  func (daemon *Daemon) RegistryService() *registry.Service {
  1493  	return daemon.registryService
  1494  }
  1495  
  1496  // BuilderBackend returns the backend used by builder
  1497  func (daemon *Daemon) BuilderBackend() builder.Backend {
  1498  	return struct {
  1499  		*Daemon
  1500  		ImageService
  1501  	}{daemon, daemon.imageService}
  1502  }
  1503  
  1504  // RawSysInfo returns *sysinfo.SysInfo .
  1505  func (daemon *Daemon) RawSysInfo() *sysinfo.SysInfo {
  1506  	daemon.sysInfoOnce.Do(func() {
  1507  		// We check if sysInfo is not set here, to allow some test to
  1508  		// override the actual sysInfo.
  1509  		if daemon.sysInfo == nil {
  1510  			daemon.sysInfo = getSysInfo(daemon)
  1511  		}
  1512  	})
  1513  
  1514  	return daemon.sysInfo
  1515  }
  1516  
  1517  // imageBackend is used to satisfy the [executorpkg.ImageBackend] and
  1518  // [github.com/docker/docker/api/server/router/distribution.Backend]
  1519  // interfaces.
  1520  type imageBackend struct {
  1521  	ImageService
  1522  	registryService *registry.Service
  1523  }
  1524  
  1525  // GetRepository returns a repository from the registry.
  1526  func (i *imageBackend) GetRepository(ctx context.Context, ref reference.Named, authConfig *registrytypes.AuthConfig) (dist.Repository, error) {
  1527  	return distribution.GetRepository(ctx, ref, &distribution.ImagePullConfig{
  1528  		Config: distribution.Config{
  1529  			AuthConfig:      authConfig,
  1530  			RegistryService: i.registryService,
  1531  		},
  1532  	})
  1533  }