github.com/docker/engine@v22.0.0-20211208180946-d456264580cf+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  	"github.com/docker/docker/api/types"
    27  	containertypes "github.com/docker/docker/api/types/container"
    28  	"github.com/docker/docker/api/types/swarm"
    29  	"github.com/docker/docker/builder"
    30  	"github.com/docker/docker/container"
    31  	"github.com/docker/docker/daemon/config"
    32  	"github.com/docker/docker/daemon/discovery"
    33  	"github.com/docker/docker/daemon/events"
    34  	"github.com/docker/docker/daemon/exec"
    35  	_ "github.com/docker/docker/daemon/graphdriver/register" // register graph drivers
    36  	"github.com/docker/docker/daemon/images"
    37  	"github.com/docker/docker/daemon/logger"
    38  	"github.com/docker/docker/daemon/network"
    39  	"github.com/docker/docker/daemon/stats"
    40  	dmetadata "github.com/docker/docker/distribution/metadata"
    41  	"github.com/docker/docker/dockerversion"
    42  	"github.com/docker/docker/errdefs"
    43  	"github.com/docker/docker/image"
    44  	"github.com/docker/docker/layer"
    45  	libcontainerdtypes "github.com/docker/docker/libcontainerd/types"
    46  	"github.com/docker/docker/libnetwork"
    47  	"github.com/docker/docker/libnetwork/cluster"
    48  	nwconfig "github.com/docker/docker/libnetwork/config"
    49  	"github.com/docker/docker/pkg/fileutils"
    50  	"github.com/docker/docker/pkg/idtools"
    51  	"github.com/docker/docker/pkg/plugingetter"
    52  	"github.com/docker/docker/pkg/system"
    53  	"github.com/docker/docker/pkg/truncindex"
    54  	"github.com/docker/docker/plugin"
    55  	pluginexec "github.com/docker/docker/plugin/executor/containerd"
    56  	refstore "github.com/docker/docker/reference"
    57  	"github.com/docker/docker/registry"
    58  	"github.com/docker/docker/runconfig"
    59  	volumesservice "github.com/docker/docker/volume/service"
    60  	"github.com/moby/buildkit/util/resolver"
    61  	"github.com/moby/locker"
    62  	"github.com/pkg/errors"
    63  	"github.com/sirupsen/logrus"
    64  	"go.etcd.io/bbolt"
    65  	"golang.org/x/sync/semaphore"
    66  	"golang.org/x/sync/singleflight"
    67  	"google.golang.org/grpc"
    68  	"google.golang.org/grpc/backoff"
    69  )
    70  
    71  // ContainersNamespace is the name of the namespace used for users containers
    72  const (
    73  	ContainersNamespace = "moby"
    74  )
    75  
    76  var (
    77  	errSystemNotSupported = errors.New("the Docker daemon is not supported on this platform")
    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          *exec.Store
    87  	imageService          *images.ImageService
    88  	idIndex               *truncindex.TruncIndex
    89  	configStore           *config.Config
    90  	statsCollector        *stats.Collector
    91  	defaultLogConfig      containertypes.LogConfig
    92  	RegistryService       registry.Service
    93  	EventsService         *events.Events
    94  	netController         libnetwork.NetworkController
    95  	volumes               *volumesservice.VolumesService
    96  	discoveryWatcher      discovery.Reloader
    97  	root                  string
    98  	seccompEnabled        bool
    99  	apparmorEnabled       bool
   100  	shutdown              bool
   101  	idMapping             *idtools.IdentityMapping
   102  	graphDriver           string        // TODO: move graphDriver field to an InfoService
   103  	PluginStore           *plugin.Store // TODO: remove
   104  	pluginManager         *plugin.Manager
   105  	linkIndex             *linkIndex
   106  	containerdCli         *containerd.Client
   107  	containerd            libcontainerdtypes.Client
   108  	defaultIsolation      containertypes.Isolation // Default isolation mode on Windows
   109  	clusterProvider       cluster.Provider
   110  	cluster               Cluster
   111  	genericResources      []swarm.GenericResource
   112  	metricsPluginListener net.Listener
   113  
   114  	machineMemory uint64
   115  
   116  	seccompProfile     []byte
   117  	seccompProfilePath string
   118  
   119  	usage singleflight.Group
   120  
   121  	pruneRunning int32
   122  	hosts        map[string]bool // hosts stores the addresses the daemon is listening on
   123  	startupDone  chan struct{}
   124  
   125  	attachmentStore       network.AttachmentStore
   126  	attachableNetworkLock *locker.Locker
   127  
   128  	// This is used for Windows which doesn't currently support running on containerd
   129  	// It stores metadata for the content store (used for manifest caching)
   130  	// This needs to be closed on daemon exit
   131  	mdDB *bbolt.DB
   132  }
   133  
   134  // StoreHosts stores the addresses the daemon is listening on
   135  func (daemon *Daemon) StoreHosts(hosts []string) {
   136  	if daemon.hosts == nil {
   137  		daemon.hosts = make(map[string]bool)
   138  	}
   139  	for _, h := range hosts {
   140  		daemon.hosts[h] = true
   141  	}
   142  }
   143  
   144  // HasExperimental returns whether the experimental features of the daemon are enabled or not
   145  func (daemon *Daemon) HasExperimental() bool {
   146  	return daemon.configStore != nil && daemon.configStore.Experimental
   147  }
   148  
   149  // Features returns the features map from configStore
   150  func (daemon *Daemon) Features() *map[string]bool {
   151  	return &daemon.configStore.Features
   152  }
   153  
   154  // RegistryHosts returns registry configuration in containerd resolvers format
   155  func (daemon *Daemon) RegistryHosts() docker.RegistryHosts {
   156  	var (
   157  		registryKey = "docker.io"
   158  		mirrors     = make([]string, len(daemon.configStore.Mirrors))
   159  		m           = map[string]resolver.RegistryConfig{}
   160  	)
   161  	// must trim "https://" or "http://" prefix
   162  	for i, v := range daemon.configStore.Mirrors {
   163  		if uri, err := url.Parse(v); err == nil {
   164  			v = uri.Host
   165  		}
   166  		mirrors[i] = v
   167  	}
   168  	// set mirrors for default registry
   169  	m[registryKey] = resolver.RegistryConfig{Mirrors: mirrors}
   170  
   171  	for _, v := range daemon.configStore.InsecureRegistries {
   172  		u, err := url.Parse(v)
   173  		c := resolver.RegistryConfig{}
   174  		if err == nil {
   175  			v = u.Host
   176  			t := true
   177  			if u.Scheme == "http" {
   178  				c.PlainHTTP = &t
   179  			} else {
   180  				c.Insecure = &t
   181  			}
   182  		}
   183  		m[v] = c
   184  	}
   185  
   186  	for k, v := range m {
   187  		if d, err := registry.HostCertsDir(k); err == nil {
   188  			v.TLSConfigDir = []string{d}
   189  			m[k] = v
   190  		}
   191  	}
   192  
   193  	certsDir := registry.CertsDir()
   194  	if fis, err := os.ReadDir(certsDir); err == nil {
   195  		for _, fi := range fis {
   196  			if _, ok := m[fi.Name()]; !ok {
   197  				m[fi.Name()] = resolver.RegistryConfig{
   198  					TLSConfigDir: []string{filepath.Join(certsDir, fi.Name())},
   199  				}
   200  			}
   201  		}
   202  	}
   203  
   204  	return resolver.NewRegistryConfig(m)
   205  }
   206  
   207  func (daemon *Daemon) restore() error {
   208  	var mapLock sync.Mutex
   209  	containers := make(map[string]*container.Container)
   210  
   211  	logrus.Info("Loading containers: start.")
   212  
   213  	dir, err := os.ReadDir(daemon.repository)
   214  	if err != nil {
   215  		return err
   216  	}
   217  
   218  	// parallelLimit is the maximum number of parallel startup jobs that we
   219  	// allow (this is the limited used for all startup semaphores). The multipler
   220  	// (128) was chosen after some fairly significant benchmarking -- don't change
   221  	// it unless you've tested it significantly (this value is adjusted if
   222  	// RLIMIT_NOFILE is small to avoid EMFILE).
   223  	parallelLimit := adjustParallelLimit(len(dir), 128*runtime.NumCPU())
   224  
   225  	// Re-used for all parallel startup jobs.
   226  	var group sync.WaitGroup
   227  	sem := semaphore.NewWeighted(int64(parallelLimit))
   228  
   229  	for _, v := range dir {
   230  		group.Add(1)
   231  		go func(id string) {
   232  			defer group.Done()
   233  			_ = sem.Acquire(context.Background(), 1)
   234  			defer sem.Release(1)
   235  
   236  			log := logrus.WithField("container", id)
   237  
   238  			c, err := daemon.load(id)
   239  			if err != nil {
   240  				log.WithError(err).Error("failed to load container")
   241  				return
   242  			}
   243  			if !system.IsOSSupported(c.OS) {
   244  				log.Errorf("failed to load container: %s (%q)", system.ErrNotSupportedOperatingSystem, c.OS)
   245  				return
   246  			}
   247  			// Ignore the container if it does not support the current driver being used by the graph
   248  			if (c.Driver == "" && daemon.graphDriver == "aufs") || c.Driver == daemon.graphDriver {
   249  				rwlayer, err := daemon.imageService.GetLayerByID(c.ID)
   250  				if err != nil {
   251  					log.WithError(err).Error("failed to load container mount")
   252  					return
   253  				}
   254  				c.RWLayer = rwlayer
   255  				log.WithFields(logrus.Fields{
   256  					"running": c.IsRunning(),
   257  					"paused":  c.IsPaused(),
   258  				}).Debug("loaded container")
   259  
   260  				mapLock.Lock()
   261  				containers[c.ID] = c
   262  				mapLock.Unlock()
   263  			} else {
   264  				log.Debugf("cannot load container because it was created with another storage driver")
   265  			}
   266  		}(v.Name())
   267  	}
   268  	group.Wait()
   269  
   270  	removeContainers := make(map[string]*container.Container)
   271  	restartContainers := make(map[*container.Container]chan struct{})
   272  	activeSandboxes := make(map[string]interface{})
   273  
   274  	for _, c := range containers {
   275  		group.Add(1)
   276  		go func(c *container.Container) {
   277  			defer group.Done()
   278  			_ = sem.Acquire(context.Background(), 1)
   279  			defer sem.Release(1)
   280  
   281  			log := logrus.WithField("container", c.ID)
   282  
   283  			if err := daemon.registerName(c); err != nil {
   284  				log.WithError(err).Errorf("failed to register container name: %s", c.Name)
   285  				mapLock.Lock()
   286  				delete(containers, c.ID)
   287  				mapLock.Unlock()
   288  				return
   289  			}
   290  			if err := daemon.Register(c); err != nil {
   291  				log.WithError(err).Error("failed to register container")
   292  				mapLock.Lock()
   293  				delete(containers, c.ID)
   294  				mapLock.Unlock()
   295  				return
   296  			}
   297  		}(c)
   298  	}
   299  	group.Wait()
   300  
   301  	for _, c := range containers {
   302  		group.Add(1)
   303  		go func(c *container.Container) {
   304  			defer group.Done()
   305  			_ = sem.Acquire(context.Background(), 1)
   306  			defer sem.Release(1)
   307  
   308  			log := logrus.WithField("container", c.ID)
   309  
   310  			daemon.backportMountSpec(c)
   311  			if err := daemon.checkpointAndSave(c); err != nil {
   312  				log.WithError(err).Error("error saving backported mountspec to disk")
   313  			}
   314  
   315  			daemon.setStateCounter(c)
   316  
   317  			logger := func(c *container.Container) *logrus.Entry {
   318  				return log.WithFields(logrus.Fields{
   319  					"running":    c.IsRunning(),
   320  					"paused":     c.IsPaused(),
   321  					"restarting": c.IsRestarting(),
   322  				})
   323  			}
   324  
   325  			logger(c).Debug("restoring container")
   326  
   327  			var (
   328  				err      error
   329  				alive    bool
   330  				ec       uint32
   331  				exitedAt time.Time
   332  				process  libcontainerdtypes.Process
   333  			)
   334  
   335  			alive, _, process, err = daemon.containerd.Restore(context.Background(), c.ID, c.InitializeStdio)
   336  			if err != nil && !errdefs.IsNotFound(err) {
   337  				logger(c).WithError(err).Error("failed to restore container with containerd")
   338  				return
   339  			}
   340  			logger(c).Debugf("alive: %v", alive)
   341  			if !alive {
   342  				// If process is not nil, cleanup dead container from containerd.
   343  				// If process is nil then the above `containerd.Restore` returned an errdefs.NotFoundError,
   344  				// and docker's view of the container state will be updated accorrdingly via SetStopped further down.
   345  				if process != nil {
   346  					logger(c).Debug("cleaning up dead container process")
   347  					ec, exitedAt, err = process.Delete(context.Background())
   348  					if err != nil && !errdefs.IsNotFound(err) {
   349  						logger(c).WithError(err).Error("failed to delete container from containerd")
   350  						return
   351  					}
   352  				}
   353  			} else if !daemon.configStore.LiveRestoreEnabled {
   354  				logger(c).Debug("shutting down container considered alive by containerd")
   355  				if err := daemon.shutdownContainer(c); err != nil && !errdefs.IsNotFound(err) {
   356  					log.WithError(err).Error("error shutting down container")
   357  					return
   358  				}
   359  				c.ResetRestartManager(false)
   360  			}
   361  
   362  			if c.IsRunning() || c.IsPaused() {
   363  				logger(c).Debug("syncing container on disk state with real state")
   364  
   365  				c.RestartManager().Cancel() // manually start containers because some need to wait for swarm networking
   366  
   367  				switch {
   368  				case c.IsPaused() && alive:
   369  					s, err := daemon.containerd.Status(context.Background(), c.ID)
   370  					if err != nil {
   371  						logger(c).WithError(err).Error("failed to get container status")
   372  					} else {
   373  						logger(c).WithField("state", s).Info("restored container paused")
   374  						switch s {
   375  						case containerd.Paused, containerd.Pausing:
   376  							// nothing to do
   377  						case containerd.Stopped:
   378  							alive = false
   379  						case containerd.Unknown:
   380  							log.Error("unknown status for paused container during restore")
   381  						default:
   382  							// running
   383  							c.Lock()
   384  							c.Paused = false
   385  							daemon.setStateCounter(c)
   386  							daemon.updateHealthMonitor(c)
   387  							if err := c.CheckpointTo(daemon.containersReplica); err != nil {
   388  								log.WithError(err).Error("failed to update paused container state")
   389  							}
   390  							c.Unlock()
   391  						}
   392  					}
   393  				case !c.IsPaused() && alive:
   394  					logger(c).Debug("restoring healthcheck")
   395  					c.Lock()
   396  					daemon.updateHealthMonitor(c)
   397  					c.Unlock()
   398  				}
   399  
   400  				if !alive {
   401  					logger(c).Debug("setting stopped state")
   402  					c.Lock()
   403  					c.SetStopped(&container.ExitStatus{ExitCode: int(ec), ExitedAt: exitedAt})
   404  					daemon.Cleanup(c)
   405  					if err := c.CheckpointTo(daemon.containersReplica); err != nil {
   406  						log.WithError(err).Error("failed to update stopped container state")
   407  					}
   408  					c.Unlock()
   409  					logger(c).Debug("set stopped state")
   410  				}
   411  
   412  				// we call Mount and then Unmount to get BaseFs of the container
   413  				if err := daemon.Mount(c); err != nil {
   414  					// The mount is unlikely to fail. However, in case mount fails
   415  					// the container should be allowed to restore here. Some functionalities
   416  					// (like docker exec -u user) might be missing but container is able to be
   417  					// stopped/restarted/removed.
   418  					// See #29365 for related information.
   419  					// The error is only logged here.
   420  					logger(c).WithError(err).Warn("failed to mount container to get BaseFs path")
   421  				} else {
   422  					if err := daemon.Unmount(c); err != nil {
   423  						logger(c).WithError(err).Warn("failed to umount container to get BaseFs path")
   424  					}
   425  				}
   426  
   427  				c.ResetRestartManager(false)
   428  				if !c.HostConfig.NetworkMode.IsContainer() && c.IsRunning() {
   429  					options, err := daemon.buildSandboxOptions(c)
   430  					if err != nil {
   431  						logger(c).WithError(err).Warn("failed to build sandbox option to restore container")
   432  					}
   433  					mapLock.Lock()
   434  					activeSandboxes[c.NetworkSettings.SandboxID] = options
   435  					mapLock.Unlock()
   436  				}
   437  			}
   438  
   439  			// get list of containers we need to restart
   440  
   441  			// Do not autostart containers which
   442  			// has endpoints in a swarm scope
   443  			// network yet since the cluster is
   444  			// not initialized yet. We will start
   445  			// it after the cluster is
   446  			// initialized.
   447  			if daemon.configStore.AutoRestart && c.ShouldRestart() && !c.NetworkSettings.HasSwarmEndpoint && c.HasBeenStartedBefore {
   448  				mapLock.Lock()
   449  				restartContainers[c] = make(chan struct{})
   450  				mapLock.Unlock()
   451  			} else if c.HostConfig != nil && c.HostConfig.AutoRemove {
   452  				mapLock.Lock()
   453  				removeContainers[c.ID] = c
   454  				mapLock.Unlock()
   455  			}
   456  
   457  			c.Lock()
   458  			if c.RemovalInProgress {
   459  				// We probably crashed in the middle of a removal, reset
   460  				// the flag.
   461  				//
   462  				// We DO NOT remove the container here as we do not
   463  				// know if the user had requested for either the
   464  				// associated volumes, network links or both to also
   465  				// be removed. So we put the container in the "dead"
   466  				// state and leave further processing up to them.
   467  				c.RemovalInProgress = false
   468  				c.Dead = true
   469  				if err := c.CheckpointTo(daemon.containersReplica); err != nil {
   470  					log.WithError(err).Error("failed to update RemovalInProgress container state")
   471  				} else {
   472  					log.Debugf("reset RemovalInProgress state for container")
   473  				}
   474  			}
   475  			c.Unlock()
   476  			logger(c).Debug("done restoring container")
   477  		}(c)
   478  	}
   479  	group.Wait()
   480  
   481  	daemon.netController, err = daemon.initNetworkController(daemon.configStore, activeSandboxes)
   482  	if err != nil {
   483  		return fmt.Errorf("Error initializing network controller: %v", err)
   484  	}
   485  
   486  	// Now that all the containers are registered, register the links
   487  	for _, c := range containers {
   488  		group.Add(1)
   489  		go func(c *container.Container) {
   490  			_ = sem.Acquire(context.Background(), 1)
   491  
   492  			if err := daemon.registerLinks(c, c.HostConfig); err != nil {
   493  				logrus.WithField("container", c.ID).WithError(err).Error("failed to register link for container")
   494  			}
   495  
   496  			sem.Release(1)
   497  			group.Done()
   498  		}(c)
   499  	}
   500  	group.Wait()
   501  
   502  	for c, notifier := range restartContainers {
   503  		group.Add(1)
   504  		go func(c *container.Container, chNotify chan struct{}) {
   505  			_ = sem.Acquire(context.Background(), 1)
   506  
   507  			log := logrus.WithField("container", c.ID)
   508  
   509  			log.Debug("starting container")
   510  
   511  			// ignore errors here as this is a best effort to wait for children to be
   512  			//   running before we try to start the container
   513  			children := daemon.children(c)
   514  			timeout := time.NewTimer(5 * time.Second)
   515  			defer timeout.Stop()
   516  
   517  			for _, child := range children {
   518  				if notifier, exists := restartContainers[child]; exists {
   519  					select {
   520  					case <-notifier:
   521  					case <-timeout.C:
   522  					}
   523  				}
   524  			}
   525  
   526  			// Make sure networks are available before starting
   527  			daemon.waitForNetworks(c)
   528  			if err := daemon.containerStart(c, "", "", true); err != nil {
   529  				log.WithError(err).Error("failed to start container")
   530  			}
   531  			close(chNotify)
   532  
   533  			sem.Release(1)
   534  			group.Done()
   535  		}(c, notifier)
   536  	}
   537  	group.Wait()
   538  
   539  	for id := range removeContainers {
   540  		group.Add(1)
   541  		go func(cid string) {
   542  			_ = sem.Acquire(context.Background(), 1)
   543  
   544  			if err := daemon.ContainerRm(cid, &types.ContainerRmConfig{ForceRemove: true, RemoveVolume: true}); err != nil {
   545  				logrus.WithField("container", cid).WithError(err).Error("failed to remove container")
   546  			}
   547  
   548  			sem.Release(1)
   549  			group.Done()
   550  		}(id)
   551  	}
   552  	group.Wait()
   553  
   554  	// any containers that were started above would already have had this done,
   555  	// however we need to now prepare the mountpoints for the rest of the containers as well.
   556  	// This shouldn't cause any issue running on the containers that already had this run.
   557  	// This must be run after any containers with a restart policy so that containerized plugins
   558  	// can have a chance to be running before we try to initialize them.
   559  	for _, c := range containers {
   560  		// if the container has restart policy, do not
   561  		// prepare the mountpoints since it has been done on restarting.
   562  		// This is to speed up the daemon start when a restart container
   563  		// has a volume and the volume driver is not available.
   564  		if _, ok := restartContainers[c]; ok {
   565  			continue
   566  		} else if _, ok := removeContainers[c.ID]; ok {
   567  			// container is automatically removed, skip it.
   568  			continue
   569  		}
   570  
   571  		group.Add(1)
   572  		go func(c *container.Container) {
   573  			_ = sem.Acquire(context.Background(), 1)
   574  
   575  			if err := daemon.prepareMountPoints(c); err != nil {
   576  				logrus.WithField("container", c.ID).WithError(err).Error("failed to prepare mountpoints for container")
   577  			}
   578  
   579  			sem.Release(1)
   580  			group.Done()
   581  		}(c)
   582  	}
   583  	group.Wait()
   584  
   585  	logrus.Info("Loading containers: done.")
   586  
   587  	return nil
   588  }
   589  
   590  // RestartSwarmContainers restarts any autostart container which has a
   591  // swarm endpoint.
   592  func (daemon *Daemon) RestartSwarmContainers() {
   593  	ctx := context.Background()
   594  
   595  	// parallelLimit is the maximum number of parallel startup jobs that we
   596  	// allow (this is the limited used for all startup semaphores). The multipler
   597  	// (128) was chosen after some fairly significant benchmarking -- don't change
   598  	// it unless you've tested it significantly (this value is adjusted if
   599  	// RLIMIT_NOFILE is small to avoid EMFILE).
   600  	parallelLimit := adjustParallelLimit(len(daemon.List()), 128*runtime.NumCPU())
   601  
   602  	var group sync.WaitGroup
   603  	sem := semaphore.NewWeighted(int64(parallelLimit))
   604  
   605  	for _, c := range daemon.List() {
   606  		if !c.IsRunning() && !c.IsPaused() {
   607  			// Autostart all the containers which has a
   608  			// swarm endpoint now that the cluster is
   609  			// initialized.
   610  			if daemon.configStore.AutoRestart && c.ShouldRestart() && c.NetworkSettings.HasSwarmEndpoint && c.HasBeenStartedBefore {
   611  				group.Add(1)
   612  				go func(c *container.Container) {
   613  					if err := sem.Acquire(ctx, 1); err != nil {
   614  						// ctx is done.
   615  						group.Done()
   616  						return
   617  					}
   618  
   619  					if err := daemon.containerStart(c, "", "", true); err != nil {
   620  						logrus.WithField("container", c.ID).WithError(err).Error("failed to start swarm container")
   621  					}
   622  
   623  					sem.Release(1)
   624  					group.Done()
   625  				}(c)
   626  			}
   627  		}
   628  	}
   629  	group.Wait()
   630  }
   631  
   632  // waitForNetworks is used during daemon initialization when starting up containers
   633  // It ensures that all of a container's networks are available before the daemon tries to start the container.
   634  // In practice it just makes sure the discovery service is available for containers which use a network that require discovery.
   635  func (daemon *Daemon) waitForNetworks(c *container.Container) {
   636  	if daemon.discoveryWatcher == nil {
   637  		return
   638  	}
   639  
   640  	// Make sure if the container has a network that requires discovery that the discovery service is available before starting
   641  	for netName := range c.NetworkSettings.Networks {
   642  		// If we get `ErrNoSuchNetwork` here, we can assume that it is due to discovery not being ready
   643  		// Most likely this is because the K/V store used for discovery is in a container and needs to be started
   644  		if _, err := daemon.netController.NetworkByName(netName); err != nil {
   645  			if _, ok := err.(libnetwork.ErrNoSuchNetwork); !ok {
   646  				continue
   647  			}
   648  
   649  			// use a longish timeout here due to some slowdowns in libnetwork if the k/v store is on anything other than --net=host
   650  			// FIXME: why is this slow???
   651  			dur := 60 * time.Second
   652  			timer := time.NewTimer(dur)
   653  
   654  			logrus.WithField("container", c.ID).Debugf("Container %s waiting for network to be ready", c.Name)
   655  			select {
   656  			case <-daemon.discoveryWatcher.ReadyCh():
   657  			case <-timer.C:
   658  			}
   659  			timer.Stop()
   660  
   661  			return
   662  		}
   663  	}
   664  }
   665  
   666  func (daemon *Daemon) children(c *container.Container) map[string]*container.Container {
   667  	return daemon.linkIndex.children(c)
   668  }
   669  
   670  // parents returns the names of the parent containers of the container
   671  // with the given name.
   672  func (daemon *Daemon) parents(c *container.Container) map[string]*container.Container {
   673  	return daemon.linkIndex.parents(c)
   674  }
   675  
   676  func (daemon *Daemon) registerLink(parent, child *container.Container, alias string) error {
   677  	fullName := path.Join(parent.Name, alias)
   678  	if err := daemon.containersReplica.ReserveName(fullName, child.ID); err != nil {
   679  		if err == container.ErrNameReserved {
   680  			logrus.Warnf("error registering link for %s, to %s, as alias %s, ignoring: %v", parent.ID, child.ID, alias, err)
   681  			return nil
   682  		}
   683  		return err
   684  	}
   685  	daemon.linkIndex.link(parent, child, fullName)
   686  	return nil
   687  }
   688  
   689  // DaemonJoinsCluster informs the daemon has joined the cluster and provides
   690  // the handler to query the cluster component
   691  func (daemon *Daemon) DaemonJoinsCluster(clusterProvider cluster.Provider) {
   692  	daemon.setClusterProvider(clusterProvider)
   693  }
   694  
   695  // DaemonLeavesCluster informs the daemon has left the cluster
   696  func (daemon *Daemon) DaemonLeavesCluster() {
   697  	// Daemon is in charge of removing the attachable networks with
   698  	// connected containers when the node leaves the swarm
   699  	daemon.clearAttachableNetworks()
   700  	// We no longer need the cluster provider, stop it now so that
   701  	// the network agent will stop listening to cluster events.
   702  	daemon.setClusterProvider(nil)
   703  	// Wait for the networking cluster agent to stop
   704  	daemon.netController.AgentStopWait()
   705  	// Daemon is in charge of removing the ingress network when the
   706  	// node leaves the swarm. Wait for job to be done or timeout.
   707  	// This is called also on graceful daemon shutdown. We need to
   708  	// wait, because the ingress release has to happen before the
   709  	// network controller is stopped.
   710  
   711  	if done, err := daemon.ReleaseIngress(); err == nil {
   712  		timeout := time.NewTimer(5 * time.Second)
   713  		defer timeout.Stop()
   714  
   715  		select {
   716  		case <-done:
   717  		case <-timeout.C:
   718  			logrus.Warn("timeout while waiting for ingress network removal")
   719  		}
   720  	} else {
   721  		logrus.Warnf("failed to initiate ingress network removal: %v", err)
   722  	}
   723  
   724  	daemon.attachmentStore.ClearAttachments()
   725  }
   726  
   727  // setClusterProvider sets a component for querying the current cluster state.
   728  func (daemon *Daemon) setClusterProvider(clusterProvider cluster.Provider) {
   729  	daemon.clusterProvider = clusterProvider
   730  	daemon.netController.SetClusterProvider(clusterProvider)
   731  	daemon.attachableNetworkLock = locker.New()
   732  }
   733  
   734  // IsSwarmCompatible verifies if the current daemon
   735  // configuration is compatible with the swarm mode
   736  func (daemon *Daemon) IsSwarmCompatible() error {
   737  	if daemon.configStore == nil {
   738  		return nil
   739  	}
   740  	return daemon.configStore.IsSwarmCompatible()
   741  }
   742  
   743  // NewDaemon sets up everything for the daemon to be able to service
   744  // requests from the webserver.
   745  func NewDaemon(ctx context.Context, config *config.Config, pluginStore *plugin.Store) (daemon *Daemon, err error) {
   746  	setDefaultMtu(config)
   747  
   748  	registryService, err := registry.NewService(config.ServiceOptions)
   749  	if err != nil {
   750  		return nil, err
   751  	}
   752  
   753  	// Ensure that we have a correct root key limit for launching containers.
   754  	if err := modifyRootKeyLimit(); err != nil {
   755  		logrus.Warnf("unable to modify root key limit, number of containers could be limited by this quota: %v", err)
   756  	}
   757  
   758  	// Ensure we have compatible and valid configuration options
   759  	if err := verifyDaemonSettings(config); err != nil {
   760  		return nil, err
   761  	}
   762  
   763  	// Do we have a disabled network?
   764  	config.DisableBridge = isBridgeNetworkDisabled(config)
   765  
   766  	// Setup the resolv.conf
   767  	setupResolvConf(config)
   768  
   769  	// Verify the platform is supported as a daemon
   770  	if !platformSupported {
   771  		return nil, errSystemNotSupported
   772  	}
   773  
   774  	// Validate platform-specific requirements
   775  	if err := checkSystem(); err != nil {
   776  		return nil, err
   777  	}
   778  
   779  	idMapping, err := setupRemappedRoot(config)
   780  	if err != nil {
   781  		return nil, err
   782  	}
   783  	rootIDs := idMapping.RootPair()
   784  	if err := setupDaemonProcess(config); err != nil {
   785  		return nil, err
   786  	}
   787  
   788  	// set up the tmpDir to use a canonical path
   789  	tmp, err := prepareTempDir(config.Root)
   790  	if err != nil {
   791  		return nil, fmt.Errorf("Unable to get the TempDir under %s: %s", config.Root, err)
   792  	}
   793  	realTmp, err := fileutils.ReadSymlinkedDirectory(tmp)
   794  	if err != nil {
   795  		return nil, fmt.Errorf("Unable to get the full path to the TempDir (%s): %s", tmp, err)
   796  	}
   797  	if isWindows {
   798  		if _, err := os.Stat(realTmp); err != nil && os.IsNotExist(err) {
   799  			if err := system.MkdirAll(realTmp, 0700); err != nil {
   800  				return nil, fmt.Errorf("Unable to create the TempDir (%s): %s", realTmp, err)
   801  			}
   802  		}
   803  		os.Setenv("TEMP", realTmp)
   804  		os.Setenv("TMP", realTmp)
   805  	} else {
   806  		os.Setenv("TMPDIR", realTmp)
   807  	}
   808  
   809  	d := &Daemon{
   810  		configStore: config,
   811  		PluginStore: pluginStore,
   812  		startupDone: make(chan struct{}),
   813  	}
   814  
   815  	// Ensure the daemon is properly shutdown if there is a failure during
   816  	// initialization
   817  	defer func() {
   818  		if err != nil {
   819  			if err := d.Shutdown(); err != nil {
   820  				logrus.Error(err)
   821  			}
   822  		}
   823  	}()
   824  
   825  	if err := d.setGenericResources(config); err != nil {
   826  		return nil, err
   827  	}
   828  	// set up SIGUSR1 handler on Unix-like systems, or a Win32 global event
   829  	// on Windows to dump Go routine stacks
   830  	stackDumpDir := config.Root
   831  	if execRoot := config.GetExecRoot(); execRoot != "" {
   832  		stackDumpDir = execRoot
   833  	}
   834  	d.setupDumpStackTrap(stackDumpDir)
   835  
   836  	if err := d.setupSeccompProfile(); err != nil {
   837  		return nil, err
   838  	}
   839  
   840  	// Set the default isolation mode (only applicable on Windows)
   841  	if err := d.setDefaultIsolation(); err != nil {
   842  		return nil, fmt.Errorf("error setting default isolation mode: %v", err)
   843  	}
   844  
   845  	if err := configureMaxThreads(config); err != nil {
   846  		logrus.Warnf("Failed to configure golang's threads limit: %v", err)
   847  	}
   848  
   849  	// ensureDefaultAppArmorProfile does nothing if apparmor is disabled
   850  	if err := ensureDefaultAppArmorProfile(); err != nil {
   851  		logrus.Errorf(err.Error())
   852  	}
   853  
   854  	daemonRepo := filepath.Join(config.Root, "containers")
   855  	if err := idtools.MkdirAllAndChown(daemonRepo, 0710, idtools.Identity{
   856  		UID: idtools.CurrentIdentity().UID,
   857  		GID: rootIDs.GID,
   858  	}); err != nil {
   859  		return nil, err
   860  	}
   861  
   862  	// Create the directory where we'll store the runtime scripts (i.e. in
   863  	// order to support runtimeArgs)
   864  	daemonRuntimes := filepath.Join(config.Root, "runtimes")
   865  	if err := system.MkdirAll(daemonRuntimes, 0700); err != nil {
   866  		return nil, err
   867  	}
   868  	if err := d.loadRuntimes(); err != nil {
   869  		return nil, err
   870  	}
   871  
   872  	if isWindows {
   873  		if err := system.MkdirAll(filepath.Join(config.Root, "credentialspecs"), 0); err != nil {
   874  			return nil, err
   875  		}
   876  	}
   877  
   878  	if isWindows {
   879  		// On Windows we don't support the environment variable, or a user supplied graphdriver
   880  		d.graphDriver = "windowsfilter"
   881  	} else {
   882  		// Unix platforms however run a single graphdriver for all containers, and it can
   883  		// be set through an environment variable, a daemon start parameter, or chosen through
   884  		// initialization of the layerstore through driver priority order for example.
   885  		if drv := os.Getenv("DOCKER_DRIVER"); drv != "" {
   886  			d.graphDriver = drv
   887  			logrus.Infof("Setting the storage driver from the $DOCKER_DRIVER environment variable (%s)", drv)
   888  		} else {
   889  			d.graphDriver = config.GraphDriver // May still be empty. Layerstore init determines instead.
   890  		}
   891  	}
   892  
   893  	d.RegistryService = registryService
   894  	logger.RegisterPluginGetter(d.PluginStore)
   895  
   896  	metricsSockPath, err := d.listenMetricsSock()
   897  	if err != nil {
   898  		return nil, err
   899  	}
   900  	registerMetricsPluginCallback(d.PluginStore, metricsSockPath)
   901  
   902  	backoffConfig := backoff.DefaultConfig
   903  	backoffConfig.MaxDelay = 3 * time.Second
   904  	connParams := grpc.ConnectParams{
   905  		Backoff: backoffConfig,
   906  	}
   907  	gopts := []grpc.DialOption{
   908  		// WithBlock makes sure that the following containerd request
   909  		// is reliable.
   910  		//
   911  		// NOTE: In one edge case with high load pressure, kernel kills
   912  		// dockerd, containerd and containerd-shims caused by OOM.
   913  		// When both dockerd and containerd restart, but containerd
   914  		// will take time to recover all the existing containers. Before
   915  		// containerd serving, dockerd will failed with gRPC error.
   916  		// That bad thing is that restore action will still ignore the
   917  		// any non-NotFound errors and returns running state for
   918  		// already stopped container. It is unexpected behavior. And
   919  		// we need to restart dockerd to make sure that anything is OK.
   920  		//
   921  		// It is painful. Add WithBlock can prevent the edge case. And
   922  		// n common case, the containerd will be serving in shortly.
   923  		// It is not harm to add WithBlock for containerd connection.
   924  		grpc.WithBlock(),
   925  
   926  		grpc.WithInsecure(),
   927  		grpc.WithConnectParams(connParams),
   928  		grpc.WithContextDialer(dialer.ContextDialer),
   929  
   930  		// TODO(stevvooe): We may need to allow configuration of this on the client.
   931  		grpc.WithDefaultCallOptions(grpc.MaxCallRecvMsgSize(defaults.DefaultMaxRecvMsgSize)),
   932  		grpc.WithDefaultCallOptions(grpc.MaxCallSendMsgSize(defaults.DefaultMaxSendMsgSize)),
   933  	}
   934  
   935  	if config.ContainerdAddr != "" {
   936  		d.containerdCli, err = containerd.New(config.ContainerdAddr, containerd.WithDefaultNamespace(config.ContainerdNamespace), containerd.WithDialOpts(gopts), containerd.WithTimeout(60*time.Second))
   937  		if err != nil {
   938  			return nil, errors.Wrapf(err, "failed to dial %q", config.ContainerdAddr)
   939  		}
   940  	}
   941  
   942  	createPluginExec := func(m *plugin.Manager) (plugin.Executor, error) {
   943  		var pluginCli *containerd.Client
   944  
   945  		if config.ContainerdAddr != "" {
   946  			pluginCli, err = containerd.New(config.ContainerdAddr, containerd.WithDefaultNamespace(config.ContainerdPluginNamespace), containerd.WithDialOpts(gopts), containerd.WithTimeout(60*time.Second))
   947  			if err != nil {
   948  				return nil, errors.Wrapf(err, "failed to dial %q", config.ContainerdAddr)
   949  			}
   950  		}
   951  
   952  		var rt types.Runtime
   953  		if runtime.GOOS != "windows" {
   954  			rtPtr, err := d.getRuntime(config.GetDefaultRuntimeName())
   955  			if err != nil {
   956  				return nil, err
   957  			}
   958  			rt = *rtPtr
   959  		}
   960  		return pluginexec.New(ctx, getPluginExecRoot(config.Root), pluginCli, config.ContainerdPluginNamespace, m, rt)
   961  	}
   962  
   963  	// Plugin system initialization should happen before restore. Do not change order.
   964  	d.pluginManager, err = plugin.NewManager(plugin.ManagerConfig{
   965  		Root:               filepath.Join(config.Root, "plugins"),
   966  		ExecRoot:           getPluginExecRoot(config.Root),
   967  		Store:              d.PluginStore,
   968  		CreateExecutor:     createPluginExec,
   969  		RegistryService:    registryService,
   970  		LiveRestoreEnabled: config.LiveRestoreEnabled,
   971  		LogPluginEvent:     d.LogPluginEvent, // todo: make private
   972  		AuthzMiddleware:    config.AuthzMiddleware,
   973  	})
   974  	if err != nil {
   975  		return nil, errors.Wrap(err, "couldn't create plugin manager")
   976  	}
   977  
   978  	if err := d.setupDefaultLogConfig(); err != nil {
   979  		return nil, err
   980  	}
   981  
   982  	layerStore, err := layer.NewStoreFromOptions(layer.StoreOptions{
   983  		Root:                      config.Root,
   984  		MetadataStorePathTemplate: filepath.Join(config.Root, "image", "%s", "layerdb"),
   985  		GraphDriver:               d.graphDriver,
   986  		GraphDriverOptions:        config.GraphOptions,
   987  		IDMapping:                 idMapping,
   988  		PluginGetter:              d.PluginStore,
   989  		ExperimentalEnabled:       config.Experimental,
   990  		OS:                        runtime.GOOS,
   991  	})
   992  	if err != nil {
   993  		return nil, err
   994  	}
   995  
   996  	// As layerstore initialization may set the driver
   997  	d.graphDriver = layerStore.DriverName()
   998  
   999  	// Configure and validate the kernels security support. Note this is a Linux/FreeBSD
  1000  	// operation only, so it is safe to pass *just* the runtime OS graphdriver.
  1001  	if err := configureKernelSecuritySupport(config, d.graphDriver); err != nil {
  1002  		return nil, err
  1003  	}
  1004  
  1005  	imageRoot := filepath.Join(config.Root, "image", d.graphDriver)
  1006  	ifs, err := image.NewFSStoreBackend(filepath.Join(imageRoot, "imagedb"))
  1007  	if err != nil {
  1008  		return nil, err
  1009  	}
  1010  
  1011  	imageStore, err := image.NewImageStore(ifs, layerStore)
  1012  	if err != nil {
  1013  		return nil, err
  1014  	}
  1015  
  1016  	d.volumes, err = volumesservice.NewVolumeService(config.Root, d.PluginStore, rootIDs, d)
  1017  	if err != nil {
  1018  		return nil, err
  1019  	}
  1020  
  1021  	trustKey, err := loadOrCreateTrustKey(config.TrustKeyPath)
  1022  	if err != nil {
  1023  		return nil, err
  1024  	}
  1025  
  1026  	trustDir := filepath.Join(config.Root, "trust")
  1027  
  1028  	if err := system.MkdirAll(trustDir, 0700); err != nil {
  1029  		return nil, err
  1030  	}
  1031  
  1032  	// We have a single tag/reference store for the daemon globally. However, it's
  1033  	// stored under the graphdriver. On host platforms which only support a single
  1034  	// container OS, but multiple selectable graphdrivers, this means depending on which
  1035  	// graphdriver is chosen, the global reference store is under there. For
  1036  	// platforms which support multiple container operating systems, this is slightly
  1037  	// more problematic as where does the global ref store get located? Fortunately,
  1038  	// for Windows, which is currently the only daemon supporting multiple container
  1039  	// operating systems, the list of graphdrivers available isn't user configurable.
  1040  	// For backwards compatibility, we just put it under the windowsfilter
  1041  	// directory regardless.
  1042  	refStoreLocation := filepath.Join(imageRoot, `repositories.json`)
  1043  	rs, err := refstore.NewReferenceStore(refStoreLocation)
  1044  	if err != nil {
  1045  		return nil, fmt.Errorf("Couldn't create reference store repository: %s", err)
  1046  	}
  1047  
  1048  	distributionMetadataStore, err := dmetadata.NewFSMetadataStore(filepath.Join(imageRoot, "distribution"))
  1049  	if err != nil {
  1050  		return nil, err
  1051  	}
  1052  
  1053  	// Discovery is only enabled when the daemon is launched with an address to advertise.  When
  1054  	// initialized, the daemon is registered and we can store the discovery backend as it's read-only
  1055  	if err := d.initDiscovery(config); err != nil {
  1056  		return nil, err
  1057  	}
  1058  
  1059  	sysInfo := d.RawSysInfo()
  1060  	for _, w := range sysInfo.Warnings {
  1061  		logrus.Warn(w)
  1062  	}
  1063  	// Check if Devices cgroup is mounted, it is hard requirement for container security,
  1064  	// on Linux.
  1065  	if runtime.GOOS == "linux" && !sysInfo.CgroupDevicesEnabled && !userns.RunningInUserNS() {
  1066  		return nil, errors.New("Devices cgroup isn't mounted")
  1067  	}
  1068  
  1069  	d.ID = trustKey.PublicKey().KeyID()
  1070  	d.repository = daemonRepo
  1071  	d.containers = container.NewMemoryStore()
  1072  	if d.containersReplica, err = container.NewViewDB(); err != nil {
  1073  		return nil, err
  1074  	}
  1075  	d.execCommands = exec.NewStore()
  1076  	d.idIndex = truncindex.NewTruncIndex([]string{})
  1077  	d.statsCollector = d.newStatsCollector(1 * time.Second)
  1078  
  1079  	d.EventsService = events.New()
  1080  	d.root = config.Root
  1081  	d.idMapping = idMapping
  1082  	d.seccompEnabled = sysInfo.Seccomp
  1083  	d.apparmorEnabled = sysInfo.AppArmor
  1084  
  1085  	d.linkIndex = newLinkIndex()
  1086  
  1087  	imgSvcConfig := images.ImageServiceConfig{
  1088  		ContainerStore:            d.containers,
  1089  		DistributionMetadataStore: distributionMetadataStore,
  1090  		EventsService:             d.EventsService,
  1091  		ImageStore:                imageStore,
  1092  		LayerStore:                layerStore,
  1093  		MaxConcurrentDownloads:    *config.MaxConcurrentDownloads,
  1094  		MaxConcurrentUploads:      *config.MaxConcurrentUploads,
  1095  		MaxDownloadAttempts:       *config.MaxDownloadAttempts,
  1096  		ReferenceStore:            rs,
  1097  		RegistryService:           registryService,
  1098  		TrustKey:                  trustKey,
  1099  		ContentNamespace:          config.ContainerdNamespace,
  1100  	}
  1101  
  1102  	// containerd is not currently supported with Windows.
  1103  	// So sometimes d.containerdCli will be nil
  1104  	// In that case we'll create a local content store... but otherwise we'll use containerd
  1105  	if d.containerdCli != nil {
  1106  		imgSvcConfig.Leases = d.containerdCli.LeasesService()
  1107  		imgSvcConfig.ContentStore = d.containerdCli.ContentStore()
  1108  	} else {
  1109  		cs, lm, err := d.configureLocalContentStore()
  1110  		if err != nil {
  1111  			return nil, err
  1112  		}
  1113  		imgSvcConfig.ContentStore = cs
  1114  		imgSvcConfig.Leases = lm
  1115  	}
  1116  
  1117  	// TODO: imageStore, distributionMetadataStore, and ReferenceStore are only
  1118  	// used above to run migration. They could be initialized in ImageService
  1119  	// if migration is called from daemon/images. layerStore might move as well.
  1120  	d.imageService = images.NewImageService(imgSvcConfig)
  1121  
  1122  	go d.execCommandGC()
  1123  
  1124  	if err := d.initLibcontainerd(ctx); err != nil {
  1125  		return nil, err
  1126  	}
  1127  
  1128  	if err := d.restore(); err != nil {
  1129  		return nil, err
  1130  	}
  1131  	close(d.startupDone)
  1132  
  1133  	info := d.SystemInfo()
  1134  
  1135  	engineInfo.WithValues(
  1136  		dockerversion.Version,
  1137  		dockerversion.GitCommit,
  1138  		info.Architecture,
  1139  		info.Driver,
  1140  		info.KernelVersion,
  1141  		info.OperatingSystem,
  1142  		info.OSType,
  1143  		info.OSVersion,
  1144  		info.ID,
  1145  	).Set(1)
  1146  	engineCpus.Set(float64(info.NCPU))
  1147  	engineMemory.Set(float64(info.MemTotal))
  1148  
  1149  	logrus.WithFields(logrus.Fields{
  1150  		"version":     dockerversion.Version,
  1151  		"commit":      dockerversion.GitCommit,
  1152  		"graphdriver": d.graphDriver,
  1153  	}).Info("Docker daemon")
  1154  
  1155  	return d, nil
  1156  }
  1157  
  1158  // DistributionServices returns services controlling daemon storage
  1159  func (daemon *Daemon) DistributionServices() images.DistributionServices {
  1160  	return daemon.imageService.DistributionServices()
  1161  }
  1162  
  1163  func (daemon *Daemon) waitForStartupDone() {
  1164  	<-daemon.startupDone
  1165  }
  1166  
  1167  func (daemon *Daemon) shutdownContainer(c *container.Container) error {
  1168  	stopTimeout := c.StopTimeout()
  1169  
  1170  	// If container failed to exit in stopTimeout seconds of SIGTERM, then using the force
  1171  	if err := daemon.containerStop(c, stopTimeout); err != nil {
  1172  		return fmt.Errorf("Failed to stop container %s with error: %v", c.ID, err)
  1173  	}
  1174  
  1175  	// Wait without timeout for the container to exit.
  1176  	// Ignore the result.
  1177  	<-c.Wait(context.Background(), container.WaitConditionNotRunning)
  1178  	return nil
  1179  }
  1180  
  1181  // ShutdownTimeout returns the timeout (in seconds) before containers are forcibly
  1182  // killed during shutdown. The default timeout can be configured both on the daemon
  1183  // and per container, and the longest timeout will be used. A grace-period of
  1184  // 5 seconds is added to the configured timeout.
  1185  //
  1186  // A negative (-1) timeout means "indefinitely", which means that containers
  1187  // are not forcibly killed, and the daemon shuts down after all containers exit.
  1188  func (daemon *Daemon) ShutdownTimeout() int {
  1189  	shutdownTimeout := daemon.configStore.ShutdownTimeout
  1190  	if shutdownTimeout < 0 {
  1191  		return -1
  1192  	}
  1193  	if daemon.containers == nil {
  1194  		return shutdownTimeout
  1195  	}
  1196  
  1197  	graceTimeout := 5
  1198  	for _, c := range daemon.containers.List() {
  1199  		stopTimeout := c.StopTimeout()
  1200  		if stopTimeout < 0 {
  1201  			return -1
  1202  		}
  1203  		if stopTimeout+graceTimeout > shutdownTimeout {
  1204  			shutdownTimeout = stopTimeout + graceTimeout
  1205  		}
  1206  	}
  1207  	return shutdownTimeout
  1208  }
  1209  
  1210  // Shutdown stops the daemon.
  1211  func (daemon *Daemon) Shutdown() error {
  1212  	daemon.shutdown = true
  1213  	// Keep mounts and networking running on daemon shutdown if
  1214  	// we are to keep containers running and restore them.
  1215  
  1216  	if daemon.configStore.LiveRestoreEnabled && daemon.containers != nil {
  1217  		// check if there are any running containers, if none we should do some cleanup
  1218  		if ls, err := daemon.Containers(&types.ContainerListOptions{}); len(ls) != 0 || err != nil {
  1219  			// metrics plugins still need some cleanup
  1220  			daemon.cleanupMetricsPlugins()
  1221  			return nil
  1222  		}
  1223  	}
  1224  
  1225  	if daemon.containers != nil {
  1226  		logrus.Debugf("daemon configured with a %d seconds minimum shutdown timeout", daemon.configStore.ShutdownTimeout)
  1227  		logrus.Debugf("start clean shutdown of all containers with a %d seconds timeout...", daemon.ShutdownTimeout())
  1228  		daemon.containers.ApplyAll(func(c *container.Container) {
  1229  			if !c.IsRunning() {
  1230  				return
  1231  			}
  1232  			log := logrus.WithField("container", c.ID)
  1233  			log.Debug("shutting down container")
  1234  			if err := daemon.shutdownContainer(c); err != nil {
  1235  				log.WithError(err).Error("failed to shut down container")
  1236  				return
  1237  			}
  1238  			if mountid, err := daemon.imageService.GetLayerMountID(c.ID); err == nil {
  1239  				daemon.cleanupMountsByID(mountid)
  1240  			}
  1241  			log.Debugf("shut down container")
  1242  		})
  1243  	}
  1244  
  1245  	if daemon.volumes != nil {
  1246  		if err := daemon.volumes.Shutdown(); err != nil {
  1247  			logrus.Errorf("Error shutting down volume store: %v", err)
  1248  		}
  1249  	}
  1250  
  1251  	if daemon.imageService != nil {
  1252  		daemon.imageService.Cleanup()
  1253  	}
  1254  
  1255  	// If we are part of a cluster, clean up cluster's stuff
  1256  	if daemon.clusterProvider != nil {
  1257  		logrus.Debugf("start clean shutdown of cluster resources...")
  1258  		daemon.DaemonLeavesCluster()
  1259  	}
  1260  
  1261  	daemon.cleanupMetricsPlugins()
  1262  
  1263  	// Shutdown plugins after containers and layerstore. Don't change the order.
  1264  	daemon.pluginShutdown()
  1265  
  1266  	// trigger libnetwork Stop only if it's initialized
  1267  	if daemon.netController != nil {
  1268  		daemon.netController.Stop()
  1269  	}
  1270  
  1271  	if daemon.containerdCli != nil {
  1272  		daemon.containerdCli.Close()
  1273  	}
  1274  
  1275  	if daemon.mdDB != nil {
  1276  		daemon.mdDB.Close()
  1277  	}
  1278  
  1279  	return daemon.cleanupMounts()
  1280  }
  1281  
  1282  // Mount sets container.BaseFS
  1283  // (is it not set coming in? why is it unset?)
  1284  func (daemon *Daemon) Mount(container *container.Container) error {
  1285  	if container.RWLayer == nil {
  1286  		return errors.New("RWLayer of container " + container.ID + " is unexpectedly nil")
  1287  	}
  1288  	dir, err := container.RWLayer.Mount(container.GetMountLabel())
  1289  	if err != nil {
  1290  		return err
  1291  	}
  1292  	logrus.WithField("container", container.ID).Debugf("container mounted via layerStore: %v", dir)
  1293  
  1294  	if container.BaseFS != nil && container.BaseFS.Path() != dir.Path() {
  1295  		// The mount path reported by the graph driver should always be trusted on Windows, since the
  1296  		// volume path for a given mounted layer may change over time.  This should only be an error
  1297  		// on non-Windows operating systems.
  1298  		if runtime.GOOS != "windows" {
  1299  			daemon.Unmount(container)
  1300  			return fmt.Errorf("Error: driver %s is returning inconsistent paths for container %s ('%s' then '%s')",
  1301  				daemon.imageService.GraphDriverName(), container.ID, container.BaseFS, dir)
  1302  		}
  1303  	}
  1304  	container.BaseFS = dir // TODO: combine these fields
  1305  	return nil
  1306  }
  1307  
  1308  // Unmount unsets the container base filesystem
  1309  func (daemon *Daemon) Unmount(container *container.Container) error {
  1310  	if container.RWLayer == nil {
  1311  		return errors.New("RWLayer of container " + container.ID + " is unexpectedly nil")
  1312  	}
  1313  	if err := container.RWLayer.Unmount(); err != nil {
  1314  		logrus.WithField("container", container.ID).WithError(err).Error("error unmounting container")
  1315  		return err
  1316  	}
  1317  
  1318  	return nil
  1319  }
  1320  
  1321  // Subnets return the IPv4 and IPv6 subnets of networks that are manager by Docker.
  1322  func (daemon *Daemon) Subnets() ([]net.IPNet, []net.IPNet) {
  1323  	var v4Subnets []net.IPNet
  1324  	var v6Subnets []net.IPNet
  1325  
  1326  	managedNetworks := daemon.netController.Networks()
  1327  
  1328  	for _, managedNetwork := range managedNetworks {
  1329  		v4infos, v6infos := managedNetwork.Info().IpamInfo()
  1330  		for _, info := range v4infos {
  1331  			if info.IPAMData.Pool != nil {
  1332  				v4Subnets = append(v4Subnets, *info.IPAMData.Pool)
  1333  			}
  1334  		}
  1335  		for _, info := range v6infos {
  1336  			if info.IPAMData.Pool != nil {
  1337  				v6Subnets = append(v6Subnets, *info.IPAMData.Pool)
  1338  			}
  1339  		}
  1340  	}
  1341  
  1342  	return v4Subnets, v6Subnets
  1343  }
  1344  
  1345  // prepareTempDir prepares and returns the default directory to use
  1346  // for temporary files.
  1347  // If it doesn't exist, it is created. If it exists, its content is removed.
  1348  func prepareTempDir(rootDir string) (string, error) {
  1349  	var tmpDir string
  1350  	if tmpDir = os.Getenv("DOCKER_TMPDIR"); tmpDir == "" {
  1351  		tmpDir = filepath.Join(rootDir, "tmp")
  1352  		newName := tmpDir + "-old"
  1353  		if err := os.Rename(tmpDir, newName); err == nil {
  1354  			go func() {
  1355  				if err := os.RemoveAll(newName); err != nil {
  1356  					logrus.Warnf("failed to delete old tmp directory: %s", newName)
  1357  				}
  1358  			}()
  1359  		} else if !os.IsNotExist(err) {
  1360  			logrus.Warnf("failed to rename %s for background deletion: %s. Deleting synchronously", tmpDir, err)
  1361  			if err := os.RemoveAll(tmpDir); err != nil {
  1362  				logrus.Warnf("failed to delete old tmp directory: %s", tmpDir)
  1363  			}
  1364  		}
  1365  	}
  1366  	return tmpDir, idtools.MkdirAllAndChown(tmpDir, 0700, idtools.CurrentIdentity())
  1367  }
  1368  
  1369  func (daemon *Daemon) setGenericResources(conf *config.Config) error {
  1370  	genericResources, err := config.ParseGenericResources(conf.NodeGenericResources)
  1371  	if err != nil {
  1372  		return err
  1373  	}
  1374  
  1375  	daemon.genericResources = genericResources
  1376  
  1377  	return nil
  1378  }
  1379  
  1380  func setDefaultMtu(conf *config.Config) {
  1381  	// do nothing if the config does not have the default 0 value.
  1382  	if conf.Mtu != 0 {
  1383  		return
  1384  	}
  1385  	conf.Mtu = config.DefaultNetworkMtu
  1386  }
  1387  
  1388  // IsShuttingDown tells whether the daemon is shutting down or not
  1389  func (daemon *Daemon) IsShuttingDown() bool {
  1390  	return daemon.shutdown
  1391  }
  1392  
  1393  // initDiscovery initializes the discovery watcher for this daemon.
  1394  func (daemon *Daemon) initDiscovery(conf *config.Config) error {
  1395  	advertise, err := config.ParseClusterAdvertiseSettings(conf.ClusterStore, conf.ClusterAdvertise)
  1396  	if err != nil {
  1397  		if err == discovery.ErrDiscoveryDisabled {
  1398  			return nil
  1399  		}
  1400  		return err
  1401  	}
  1402  
  1403  	conf.ClusterAdvertise = advertise
  1404  	discoveryWatcher, err := discovery.Init(conf.ClusterStore, conf.ClusterAdvertise, conf.ClusterOpts)
  1405  	if err != nil {
  1406  		return fmt.Errorf("discovery initialization failed (%v)", err)
  1407  	}
  1408  
  1409  	daemon.discoveryWatcher = discoveryWatcher
  1410  	return nil
  1411  }
  1412  
  1413  func isBridgeNetworkDisabled(conf *config.Config) bool {
  1414  	return conf.BridgeConfig.Iface == config.DisableNetworkBridge
  1415  }
  1416  
  1417  func (daemon *Daemon) networkOptions(dconfig *config.Config, pg plugingetter.PluginGetter, activeSandboxes map[string]interface{}) ([]nwconfig.Option, error) {
  1418  	options := []nwconfig.Option{}
  1419  	if dconfig == nil {
  1420  		return options, nil
  1421  	}
  1422  
  1423  	options = append(options, nwconfig.OptionExperimental(dconfig.Experimental))
  1424  	options = append(options, nwconfig.OptionDataDir(dconfig.Root))
  1425  	options = append(options, nwconfig.OptionExecRoot(dconfig.GetExecRoot()))
  1426  
  1427  	dd := runconfig.DefaultDaemonNetworkMode()
  1428  	dn := runconfig.DefaultDaemonNetworkMode().NetworkName()
  1429  	options = append(options, nwconfig.OptionDefaultDriver(string(dd)))
  1430  	options = append(options, nwconfig.OptionDefaultNetwork(dn))
  1431  
  1432  	if strings.TrimSpace(dconfig.ClusterStore) != "" {
  1433  		kv := strings.Split(dconfig.ClusterStore, "://")
  1434  		if len(kv) != 2 {
  1435  			return nil, errors.New("kv store daemon config must be of the form KV-PROVIDER://KV-URL")
  1436  		}
  1437  		options = append(options, nwconfig.OptionKVProvider(kv[0]))
  1438  		options = append(options, nwconfig.OptionKVProviderURL(kv[1]))
  1439  	}
  1440  	if len(dconfig.ClusterOpts) > 0 {
  1441  		options = append(options, nwconfig.OptionKVOpts(dconfig.ClusterOpts))
  1442  	}
  1443  
  1444  	if daemon.discoveryWatcher != nil {
  1445  		options = append(options, nwconfig.OptionDiscoveryWatcher(daemon.discoveryWatcher))
  1446  	}
  1447  
  1448  	if dconfig.ClusterAdvertise != "" {
  1449  		options = append(options, nwconfig.OptionDiscoveryAddress(dconfig.ClusterAdvertise))
  1450  	}
  1451  
  1452  	options = append(options, nwconfig.OptionLabels(dconfig.Labels))
  1453  	options = append(options, driverOptions(dconfig))
  1454  
  1455  	if len(dconfig.NetworkConfig.DefaultAddressPools.Value()) > 0 {
  1456  		options = append(options, nwconfig.OptionDefaultAddressPoolConfig(dconfig.NetworkConfig.DefaultAddressPools.Value()))
  1457  	}
  1458  
  1459  	if daemon.configStore != nil && daemon.configStore.LiveRestoreEnabled && len(activeSandboxes) != 0 {
  1460  		options = append(options, nwconfig.OptionActiveSandboxes(activeSandboxes))
  1461  	}
  1462  
  1463  	if pg != nil {
  1464  		options = append(options, nwconfig.OptionPluginGetter(pg))
  1465  	}
  1466  
  1467  	options = append(options, nwconfig.OptionNetworkControlPlaneMTU(dconfig.NetworkControlPlaneMTU))
  1468  
  1469  	return options, nil
  1470  }
  1471  
  1472  // GetCluster returns the cluster
  1473  func (daemon *Daemon) GetCluster() Cluster {
  1474  	return daemon.cluster
  1475  }
  1476  
  1477  // SetCluster sets the cluster
  1478  func (daemon *Daemon) SetCluster(cluster Cluster) {
  1479  	daemon.cluster = cluster
  1480  }
  1481  
  1482  func (daemon *Daemon) pluginShutdown() {
  1483  	manager := daemon.pluginManager
  1484  	// Check for a valid manager object. In error conditions, daemon init can fail
  1485  	// and shutdown called, before plugin manager is initialized.
  1486  	if manager != nil {
  1487  		manager.Shutdown()
  1488  	}
  1489  }
  1490  
  1491  // PluginManager returns current pluginManager associated with the daemon
  1492  func (daemon *Daemon) PluginManager() *plugin.Manager { // set up before daemon to avoid this method
  1493  	return daemon.pluginManager
  1494  }
  1495  
  1496  // PluginGetter returns current pluginStore associated with the daemon
  1497  func (daemon *Daemon) PluginGetter() *plugin.Store {
  1498  	return daemon.PluginStore
  1499  }
  1500  
  1501  // CreateDaemonRoot creates the root for the daemon
  1502  func CreateDaemonRoot(config *config.Config) error {
  1503  	// get the canonical path to the Docker root directory
  1504  	var realRoot string
  1505  	if _, err := os.Stat(config.Root); err != nil && os.IsNotExist(err) {
  1506  		realRoot = config.Root
  1507  	} else {
  1508  		realRoot, err = fileutils.ReadSymlinkedDirectory(config.Root)
  1509  		if err != nil {
  1510  			return fmt.Errorf("Unable to get the full path to root (%s): %s", config.Root, err)
  1511  		}
  1512  	}
  1513  
  1514  	idMapping, err := setupRemappedRoot(config)
  1515  	if err != nil {
  1516  		return err
  1517  	}
  1518  	return setupDaemonRoot(config, realRoot, idMapping.RootPair())
  1519  }
  1520  
  1521  // checkpointAndSave grabs a container lock to safely call container.CheckpointTo
  1522  func (daemon *Daemon) checkpointAndSave(container *container.Container) error {
  1523  	container.Lock()
  1524  	defer container.Unlock()
  1525  	if err := container.CheckpointTo(daemon.containersReplica); err != nil {
  1526  		return fmt.Errorf("Error saving container state: %v", err)
  1527  	}
  1528  	return nil
  1529  }
  1530  
  1531  // because the CLI sends a -1 when it wants to unset the swappiness value
  1532  // we need to clear it on the server side
  1533  func fixMemorySwappiness(resources *containertypes.Resources) {
  1534  	if resources.MemorySwappiness != nil && *resources.MemorySwappiness == -1 {
  1535  		resources.MemorySwappiness = nil
  1536  	}
  1537  }
  1538  
  1539  // GetAttachmentStore returns current attachment store associated with the daemon
  1540  func (daemon *Daemon) GetAttachmentStore() *network.AttachmentStore {
  1541  	return &daemon.attachmentStore
  1542  }
  1543  
  1544  // IdentityMapping returns uid/gid mapping or a SID (in the case of Windows) for the builder
  1545  func (daemon *Daemon) IdentityMapping() *idtools.IdentityMapping {
  1546  	return daemon.idMapping
  1547  }
  1548  
  1549  // ImageService returns the Daemon's ImageService
  1550  func (daemon *Daemon) ImageService() *images.ImageService {
  1551  	return daemon.imageService
  1552  }
  1553  
  1554  // BuilderBackend returns the backend used by builder
  1555  func (daemon *Daemon) BuilderBackend() builder.Backend {
  1556  	return struct {
  1557  		*Daemon
  1558  		*images.ImageService
  1559  	}{daemon, daemon.imageService}
  1560  }