VMware vSphere Integrated Containers provider (#206)

* Add Virtual Kubelet provider for VIC

Initial virtual kubelet provider for VMware VIC.  This provider currently
handles creating and starting of a pod VM via the VIC portlayer and persona
server.  Image store handling via the VIC persona server.  This provider
currently requires the feature/wolfpack branch of VIC.

* Added pod stop and delete.  Also added node capacity.

Added the ability to stop and delete pod VMs via VIC.  Also retrieve
node capacity information from the VCH.

* Cleanup and readme file

Some file clean up and added a Readme.md markdown file for the VIC
provider.

* Cleaned up errors, added function comments, moved operation code

1. Cleaned up error handling.  Set standard for creating errors.
2. Added method prototype comments for all interface functions.
3. Moved PodCreator, PodStarter, PodStopper, and PodDeleter to a new folder.

* Add mocking code and unit tests for podcache, podcreator, and podstarter

Used the unit test framework used in VIC to handle assertions in the provider's
unit test.  Mocking code generated using OSS project mockery, which is compatible
with the testify assertion framework.

* Vendored packages for the VIC provider

Requires feature/wolfpack branch of VIC and a few specific commit sha of
projects used within VIC.

* Implementation of POD Stopper and Deleter unit tests (#4)

* Updated files for initial PR

vendor/github.com/docker/libnetwork/drivers/overlay/filter.go

@@ -0,0 +1,123 @@
+package overlay
+
+import (
+	"fmt"
+	"sync"
+
+	"github.com/Sirupsen/logrus"
+	"github.com/docker/libnetwork/iptables"
+)
+
+const globalChain = "DOCKER-OVERLAY"
+
+var filterOnce sync.Once
+
+func chainExists(cname string) bool {
+	if _, err := iptables.Raw("-L", cname); err != nil {
+		return false
+	}
+
+	return true
+}
+
+func setupGlobalChain() {
+	if err := iptables.RawCombinedOutput("-N", globalChain); err != nil {
+		logrus.Errorf("could not create global overlay chain: %v", err)
+		return
+	}
+
+	if err := iptables.RawCombinedOutput("-A", globalChain, "-j", "RETURN"); err != nil {
+		logrus.Errorf("could not install default return chain in the overlay global chain: %v", err)
+		return
+	}
+}
+
+func setNetworkChain(cname string, remove bool) error {
+	// Initialize the onetime global overlay chain
+	filterOnce.Do(setupGlobalChain)
+
+	exists := chainExists(cname)
+
+	opt := "-N"
+	// In case of remove, make sure to flush the rules in the chain
+	if remove && exists {
+		if err := iptables.RawCombinedOutput("-F", cname); err != nil {
+			return fmt.Errorf("failed to flush overlay network chain %s rules: %v", cname, err)
+		}
+		opt = "-X"
+	}
+
+	if (!remove && !exists) || (remove && exists) {
+		if err := iptables.RawCombinedOutput(opt, cname); err != nil {
+			return fmt.Errorf("failed network chain operation %q for chain %s: %v", opt, cname, err)
+		}
+	}
+
+	if !remove {
+		if !iptables.Exists(iptables.Filter, cname, "-j", "DROP") {
+			if err := iptables.RawCombinedOutput("-A", cname, "-j", "DROP"); err != nil {
+				return fmt.Errorf("failed adding default drop rule to overlay network chain %s: %v", cname, err)
+			}
+		}
+	}
+
+	return nil
+}
+
+func addNetworkChain(cname string) error {
+	return setNetworkChain(cname, false)
+}
+
+func removeNetworkChain(cname string) error {
+	return setNetworkChain(cname, true)
+}
+
+func setFilters(cname, brName string, remove bool) error {
+	opt := "-I"
+	if remove {
+		opt = "-D"
+	}
+
+	// Everytime we set filters for a new subnet make sure to move the global overlay hook to the top of the both the OUTPUT and forward chains
+	if !remove {
+		for _, chain := range []string{"OUTPUT", "FORWARD"} {
+			exists := iptables.Exists(iptables.Filter, chain, "-j", globalChain)
+			if exists {
+				if err := iptables.RawCombinedOutput("-D", chain, "-j", globalChain); err != nil {
+					return fmt.Errorf("failed to delete overlay hook in chain %s while moving the hook: %v", chain, err)
+				}
+			}
+
+			if err := iptables.RawCombinedOutput("-I", chain, "-j", globalChain); err != nil {
+				return fmt.Errorf("failed to insert overlay hook in chain %s: %v", chain, err)
+			}
+		}
+	}
+
+	// Insert/Delete the rule to jump to per-bridge chain
+	exists := iptables.Exists(iptables.Filter, globalChain, "-o", brName, "-j", cname)
+	if (!remove && !exists) || (remove && exists) {
+		if err := iptables.RawCombinedOutput(opt, globalChain, "-o", brName, "-j", cname); err != nil {
+			return fmt.Errorf("failed to add per-bridge filter rule for bridge %s, network chain %s: %v", brName, cname, err)
+		}
+	}
+
+	exists = iptables.Exists(iptables.Filter, cname, "-i", brName, "-j", "ACCEPT")
+	if (!remove && exists) || (remove && !exists) {
+		return nil
+	}
+
+	if err := iptables.RawCombinedOutput(opt, cname, "-i", brName, "-j", "ACCEPT"); err != nil {
+		return fmt.Errorf("failed to add overlay filter rile for network chain %s, bridge %s: %v", cname, brName, err)
+	}
+
+	return nil
+}
+
+func addFilters(cname, brName string) error {
+	return setFilters(cname, brName, false)
+}
+
+func removeFilters(cname, brName string) error {
+	return setFilters(cname, brName, true)
+}

vendor/github.com/docker/libnetwork/drivers/overlay/joinleave.go

@@ -0,0 +1,149 @@
+package overlay
+
+import (
+	"fmt"
+	"net"
+
+	log "github.com/Sirupsen/logrus"
+	"github.com/docker/libnetwork/driverapi"
+	"github.com/docker/libnetwork/types"
+	"github.com/vishvananda/netlink"
+)
+
+// Join method is invoked when a Sandbox is attached to an endpoint.
+func (d *driver) Join(nid, eid string, sboxKey string, jinfo driverapi.JoinInfo, options map[string]interface{}) error {
+	if err := validateID(nid, eid); err != nil {
+		return err
+	}
+
+	n := d.network(nid)
+	if n == nil {
+		return fmt.Errorf("could not find network with id %s", nid)
+	}
+
+	ep := n.endpoint(eid)
+	if ep == nil {
+		return fmt.Errorf("could not find endpoint with id %s", eid)
+	}
+
+	s := n.getSubnetforIP(ep.addr)
+	if s == nil {
+		return fmt.Errorf("could not find subnet for endpoint %s", eid)
+	}
+
+	if err := n.obtainVxlanID(s); err != nil {
+		return fmt.Errorf("couldn't get vxlan id for %q: %v", s.subnetIP.String(), err)
+	}
+
+	if err := n.joinSandbox(); err != nil {
+		return fmt.Errorf("network sandbox join failed: %v", err)
+	}
+
+	if err := n.joinSubnetSandbox(s); err != nil {
+		return fmt.Errorf("subnet sandbox join failed for %q: %v", s.subnetIP.String(), err)
+	}
+
+	// joinSubnetSandbox gets called when an endpoint comes up on a new subnet in the
+	// overlay network. Hence the Endpoint count should be updated outside joinSubnetSandbox
+	n.incEndpointCount()
+
+	sbox := n.sandbox()
+
+	name1, name2, err := createVethPair()
+	if err != nil {
+		return err
+	}
+
+	ep.ifName = name2
+
+	// Set the container interface and its peer MTU to 1450 to allow
+	// for 50 bytes vxlan encap (inner eth header(14) + outer IP(20) +
+	// outer UDP(8) + vxlan header(8))
+	veth, err := netlink.LinkByName(name1)
+	if err != nil {
+		return fmt.Errorf("cound not find link by name %s: %v", name1, err)
+	}
+	err = netlink.LinkSetMTU(veth, vxlanVethMTU)
+	if err != nil {
+		return err
+	}
+
+	if err := sbox.AddInterface(name1, "veth",
+		sbox.InterfaceOptions().Master(s.brName)); err != nil {
+		return fmt.Errorf("could not add veth pair inside the network sandbox: %v", err)
+	}
+
+	veth, err = netlink.LinkByName(name2)
+	if err != nil {
+		return fmt.Errorf("could not find link by name %s: %v", name2, err)
+	}
+	err = netlink.LinkSetMTU(veth, vxlanVethMTU)
+	if err != nil {
+		return err
+	}
+
+	if err := netlink.LinkSetHardwareAddr(veth, ep.mac); err != nil {
+		return fmt.Errorf("could not set mac address (%v) to the container interface: %v", ep.mac, err)
+	}
+
+	for _, sub := range n.subnets {
+		if sub == s {
+			continue
+		}
+		if err := jinfo.AddStaticRoute(sub.subnetIP, types.NEXTHOP, s.gwIP.IP); err != nil {
+			log.Errorf("Adding subnet %s static route in network %q failed\n", s.subnetIP, n.id)
+		}
+	}
+
+	if iNames := jinfo.InterfaceName(); iNames != nil {
+		err = iNames.SetNames(name2, "eth")
+		if err != nil {
+			return err
+		}
+	}
+
+	d.peerDbAdd(nid, eid, ep.addr.IP, ep.addr.Mask, ep.mac,
+		net.ParseIP(d.bindAddress), true)
+	d.pushLocalEndpointEvent("join", nid, eid)
+
+	return nil
+}
+
+// Leave method is invoked when a Sandbox detaches from an endpoint.
+func (d *driver) Leave(nid, eid string) error {
+	if err := validateID(nid, eid); err != nil {
+		return err
+	}
+
+	n := d.network(nid)
+	if n == nil {
+		return fmt.Errorf("could not find network with id %s", nid)
+	}
+
+	ep := n.endpoint(eid)
+
+	if ep == nil {
+		return types.InternalMaskableErrorf("could not find endpoint with id %s", eid)
+	}
+
+	if d.notifyCh != nil {
+		d.notifyCh <- ovNotify{
+			action: "leave",
+			nid:    nid,
+			eid:    eid,
+		}
+	}
+
+	n.leaveSandbox()
+
+	link, err := netlink.LinkByName(ep.ifName)
+	if err != nil {
+		log.Warnf("Failed to retrieve interface link for interface removal on endpoint leave: %v", err)
+		return nil
+	}
+	if err := netlink.LinkDel(link); err != nil {
+		log.Warnf("Failed to delete interface link on endpoint leave: %v", err)
+	}
+
+	return nil
+}

vendor/github.com/docker/libnetwork/drivers/overlay/ov_endpoint.go

@@ -0,0 +1,105 @@
+package overlay
+
+import (
+	"fmt"
+	"net"
+
+	"github.com/docker/libnetwork/driverapi"
+	"github.com/docker/libnetwork/netutils"
+)
+
+type endpointTable map[string]*endpoint
+
+type endpoint struct {
+	id     string
+	ifName string
+	mac    net.HardwareAddr
+	addr   *net.IPNet
+}
+
+func (n *network) endpoint(eid string) *endpoint {
+	n.Lock()
+	defer n.Unlock()
+
+	return n.endpoints[eid]
+}
+
+func (n *network) addEndpoint(ep *endpoint) {
+	n.Lock()
+	n.endpoints[ep.id] = ep
+	n.Unlock()
+}
+
+func (n *network) deleteEndpoint(eid string) {
+	n.Lock()
+	delete(n.endpoints, eid)
+	n.Unlock()
+}
+
+func (d *driver) CreateEndpoint(nid, eid string, ifInfo driverapi.InterfaceInfo,
+	epOptions map[string]interface{}) error {
+	var err error
+
+	if err = validateID(nid, eid); err != nil {
+		return err
+	}
+
+	// Since we perform lazy configuration make sure we try
+	// configuring the driver when we enter CreateEndpoint since
+	// CreateNetwork may not be called in every node.
+	if err := d.configure(); err != nil {
+		return err
+	}
+
+	n := d.network(nid)
+	if n == nil {
+		return fmt.Errorf("network id %q not found", nid)
+	}
+
+	ep := &endpoint{
+		id:   eid,
+		addr: ifInfo.Address(),
+		mac:  ifInfo.MacAddress(),
+	}
+	if ep.addr == nil {
+		return fmt.Errorf("create endpoint was not passed interface IP address")
+	}
+
+	if s := n.getSubnetforIP(ep.addr); s == nil {
+		return fmt.Errorf("no matching subnet for IP %q in network %q\n", ep.addr, nid)
+	}
+
+	if ep.mac == nil {
+		ep.mac = netutils.GenerateMACFromIP(ep.addr.IP)
+		if err := ifInfo.SetMacAddress(ep.mac); err != nil {
+			return err
+		}
+	}
+
+	n.addEndpoint(ep)
+
+	return nil
+}
+
+func (d *driver) DeleteEndpoint(nid, eid string) error {
+	if err := validateID(nid, eid); err != nil {
+		return err
+	}
+
+	n := d.network(nid)
+	if n == nil {
+		return fmt.Errorf("network id %q not found", nid)
+	}
+
+	ep := n.endpoint(eid)
+	if ep == nil {
+		return fmt.Errorf("endpoint id %q not found", eid)
+	}
+
+	n.deleteEndpoint(eid)
+	return nil
+}
+
+func (d *driver) EndpointOperInfo(nid, eid string) (map[string]interface{}, error) {
+	return make(map[string]interface{}, 0), nil
+}

vendor/github.com/docker/libnetwork/drivers/overlay/ov_network.go

@@ -0,0 +1,633 @@
+package overlay
+
+import (
+	"encoding/json"
+	"fmt"
+	"net"
+	"os"
+	"sync"
+	"syscall"
+
+	"github.com/Sirupsen/logrus"
+	"github.com/docker/libnetwork/datastore"
+	"github.com/docker/libnetwork/driverapi"
+	"github.com/docker/libnetwork/netutils"
+	"github.com/docker/libnetwork/osl"
+	"github.com/docker/libnetwork/resolvconf"
+	"github.com/docker/libnetwork/types"
+	"github.com/vishvananda/netlink"
+	"github.com/vishvananda/netlink/nl"
+)
+
+var (
+	hostMode     bool
+	hostModeOnce sync.Once
+)
+
+type networkTable map[string]*network
+
+type subnet struct {
+	once      *sync.Once
+	vxlanName string
+	brName    string
+	vni       uint32
+	initErr   error
+	subnetIP  *net.IPNet
+	gwIP      *net.IPNet
+}
+
+type subnetJSON struct {
+	SubnetIP string
+	GwIP     string
+	Vni      uint32
+}
+
+type network struct {
+	id        string
+	dbIndex   uint64
+	dbExists  bool
+	sbox      osl.Sandbox
+	endpoints endpointTable
+	driver    *driver
+	joinCnt   int
+	once      *sync.Once
+	initEpoch int
+	initErr   error
+	subnets   []*subnet
+	sync.Mutex
+}
+
+func (d *driver) CreateNetwork(id string, option map[string]interface{}, ipV4Data, ipV6Data []driverapi.IPAMData) error {
+	if id == "" {
+		return fmt.Errorf("invalid network id")
+	}
+
+	// Since we perform lazy configuration make sure we try
+	// configuring the driver when we enter CreateNetwork
+	if err := d.configure(); err != nil {
+		return err
+	}
+
+	n := &network{
+		id:        id,
+		driver:    d,
+		endpoints: endpointTable{},
+		once:      &sync.Once{},
+		subnets:   []*subnet{},
+	}
+
+	for _, ipd := range ipV4Data {
+		s := &subnet{
+			subnetIP: ipd.Pool,
+			gwIP:     ipd.Gateway,
+			once:     &sync.Once{},
+		}
+		n.subnets = append(n.subnets, s)
+	}
+
+	if err := n.writeToStore(); err != nil {
+		return fmt.Errorf("failed to update data store for network %v: %v", n.id, err)
+	}
+
+	d.addNetwork(n)
+
+	return nil
+}
+
+func (d *driver) DeleteNetwork(nid string) error {
+	if nid == "" {
+		return fmt.Errorf("invalid network id")
+	}
+
+	n := d.network(nid)
+	if n == nil {
+		return fmt.Errorf("could not find network with id %s", nid)
+	}
+
+	d.deleteNetwork(nid)
+
+	return n.releaseVxlanID()
+}
+
+func (n *network) incEndpointCount() {
+	n.Lock()
+	defer n.Unlock()
+	n.joinCnt++
+}
+
+func (n *network) joinSandbox() error {
+	// If there is a race between two go routines here only one will win
+	// the other will wait.
+	n.once.Do(func() {
+		// save the error status of initSandbox in n.initErr so that
+		// all the racing go routines are able to know the status.
+		n.initErr = n.initSandbox()
+	})
+
+	return n.initErr
+}
+
+func (n *network) joinSubnetSandbox(s *subnet) error {
+	s.once.Do(func() {
+		s.initErr = n.initSubnetSandbox(s)
+	})
+	return s.initErr
+}
+
+func (n *network) leaveSandbox() {
+	n.Lock()
+	n.joinCnt--
+	if n.joinCnt != 0 {
+		n.Unlock()
+		return
+	}
+
+	// We are about to destroy sandbox since the container is leaving the network
+	// Reinitialize the once variable so that we will be able to trigger one time
+	// sandbox initialization(again) when another container joins subsequently.
+	n.once = &sync.Once{}
+	for _, s := range n.subnets {
+		s.once = &sync.Once{}
+	}
+	n.Unlock()
+
+	n.destroySandbox()
+}
+
+func (n *network) destroySandbox() {
+	sbox := n.sandbox()
+	if sbox != nil {
+		for _, iface := range sbox.Info().Interfaces() {
+			iface.Remove()
+		}
+
+		for _, s := range n.subnets {
+			if hostMode {
+				if err := removeFilters(n.id[:12], s.brName); err != nil {
+					logrus.Warnf("Could not remove overlay filters: %v", err)
+				}
+			}
+
+			if s.vxlanName != "" {
+				err := deleteInterface(s.vxlanName)
+				if err != nil {
+					logrus.Warnf("could not cleanup sandbox properly: %v", err)
+				}
+			}
+		}
+
+		if hostMode {
+			if err := removeNetworkChain(n.id[:12]); err != nil {
+				logrus.Warnf("could not remove network chain: %v", err)
+			}
+		}
+
+		sbox.Destroy()
+		n.setSandbox(nil)
+	}
+}
+
+func setHostMode() {
+	if os.Getenv("_OVERLAY_HOST_MODE") != "" {
+		hostMode = true
+		return
+	}
+
+	err := createVxlan("testvxlan", 1)
+	if err != nil {
+		logrus.Errorf("Failed to create testvxlan interface: %v", err)
+		return
+	}
+
+	defer deleteInterface("testvxlan")
+
+	path := "/proc/self/ns/net"
+	f, err := os.OpenFile(path, os.O_RDONLY, 0)
+	if err != nil {
+		logrus.Errorf("Failed to open path %s for network namespace for setting host mode: %v", path, err)
+		return
+	}
+	defer f.Close()
+
+	nsFD := f.Fd()
+
+	iface, err := netlink.LinkByName("testvxlan")
+	if err != nil {
+		logrus.Errorf("Failed to get link testvxlan: %v", err)
+		return
+	}
+
+	// If we are not able to move the vxlan interface to a namespace
+	// then fallback to host mode
+	if err := netlink.LinkSetNsFd(iface, int(nsFD)); err != nil {
+		hostMode = true
+	}
+}
+
+func (n *network) generateVxlanName(s *subnet) string {
+	return "vx-" + fmt.Sprintf("%06x", n.vxlanID(s)) + "-" + n.id[:5]
+}
+
+func (n *network) generateBridgeName(s *subnet) string {
+	return "ov-" + fmt.Sprintf("%06x", n.vxlanID(s)) + "-" + n.id[:5]
+}
+
+func isOverlap(nw *net.IPNet) bool {
+	var nameservers []string
+
+	if rc, err := resolvconf.Get(); err == nil {
+		nameservers = resolvconf.GetNameserversAsCIDR(rc.Content)
+	}
+
+	if err := netutils.CheckNameserverOverlaps(nameservers, nw); err != nil {
+		return true
+	}
+
+	if err := netutils.CheckRouteOverlaps(nw); err != nil {
+		return true
+	}
+
+	return false
+}
+
+func (n *network) initSubnetSandbox(s *subnet) error {
+	brName := n.generateBridgeName(s)
+	vxlanName := n.generateVxlanName(s)
+
+	if hostMode {
+		// Try to delete stale bridge interface if it exists
+		deleteInterface(brName)
+		// Try to delete the vxlan interface by vni if already present
+		deleteVxlanByVNI(n.vxlanID(s))
+
+		if isOverlap(s.subnetIP) {
+			return fmt.Errorf("overlay subnet %s has conflicts in the host while running in host mode", s.subnetIP.String())
+		}
+	}
+
+	// create a bridge and vxlan device for this subnet and move it to the sandbox
+	sbox := n.sandbox()
+
+	if err := sbox.AddInterface(brName, "br",
+		sbox.InterfaceOptions().Address(s.gwIP),
+		sbox.InterfaceOptions().Bridge(true)); err != nil {
+		return fmt.Errorf("bridge creation in sandbox failed for subnet %q: %v", s.subnetIP.String(), err)
+	}
+
+	err := createVxlan(vxlanName, n.vxlanID(s))
+	if err != nil {
+		return err
+	}
+
+	if err := sbox.AddInterface(vxlanName, "vxlan",
+		sbox.InterfaceOptions().Master(brName)); err != nil {
+		return fmt.Errorf("vxlan interface creation failed for subnet %q: %v", s.subnetIP.String(), err)
+	}
+
+	if hostMode {
+		if err := addFilters(n.id[:12], brName); err != nil {
+			return err
+		}
+	}
+
+	n.Lock()
+	s.vxlanName = vxlanName
+	s.brName = brName
+	n.Unlock()
+
+	return nil
+}
+
+func (n *network) initSandbox() error {
+	n.Lock()
+	n.initEpoch++
+	n.Unlock()
+
+	hostModeOnce.Do(setHostMode)
+
+	if hostMode {
+		if err := addNetworkChain(n.id[:12]); err != nil {
+			return err
+		}
+	}
+
+	sbox, err := osl.NewSandbox(
+		osl.GenerateKey(fmt.Sprintf("%d-", n.initEpoch)+n.id), !hostMode)
+	if err != nil {
+		return fmt.Errorf("could not create network sandbox: %v", err)
+	}
+
+	n.setSandbox(sbox)
+
+	n.driver.peerDbUpdateSandbox(n.id)
+
+	var nlSock *nl.NetlinkSocket
+	sbox.InvokeFunc(func() {
+		nlSock, err = nl.Subscribe(syscall.NETLINK_ROUTE, syscall.RTNLGRP_NEIGH)
+		if err != nil {
+			err = fmt.Errorf("failed to subscribe to neighbor group netlink messages")
+		}
+	})
+
+	go n.watchMiss(nlSock)
+	return nil
+}
+
+func (n *network) watchMiss(nlSock *nl.NetlinkSocket) {
+	for {
+		msgs, err := nlSock.Receive()
+		if err != nil {
+			logrus.Errorf("Failed to receive from netlink: %v ", err)
+			continue
+		}
+
+		for _, msg := range msgs {
+			if msg.Header.Type != syscall.RTM_GETNEIGH && msg.Header.Type != syscall.RTM_NEWNEIGH {
+				continue
+			}
+
+			neigh, err := netlink.NeighDeserialize(msg.Data)
+			if err != nil {
+				logrus.Errorf("Failed to deserialize netlink ndmsg: %v", err)
+				continue
+			}
+
+			if neigh.IP.To16() != nil {
+				continue
+			}
+
+			if neigh.State&(netlink.NUD_STALE|netlink.NUD_INCOMPLETE) == 0 {
+				continue
+			}
+
+			mac, IPmask, vtep, err := n.driver.resolvePeer(n.id, neigh.IP)
+			if err != nil {
+				logrus.Errorf("could not resolve peer %q: %v", neigh.IP, err)
+				continue
+			}
+
+			if err := n.driver.peerAdd(n.id, "dummy", neigh.IP, IPmask, mac, vtep, true); err != nil {
+				logrus.Errorf("could not add neighbor entry for missed peer %q: %v", neigh.IP, err)
+			}
+		}
+	}
+}
+
+func (d *driver) addNetwork(n *network) {
+	d.Lock()
+	d.networks[n.id] = n
+	d.Unlock()
+}
+
+func (d *driver) deleteNetwork(nid string) {
+	d.Lock()
+	delete(d.networks, nid)
+	d.Unlock()
+}
+
+func (d *driver) network(nid string) *network {
+	d.Lock()
+	networks := d.networks
+	d.Unlock()
+
+	n, ok := networks[nid]
+	if !ok {
+		n = d.getNetworkFromStore(nid)
+		if n != nil {
+			n.driver = d
+			n.endpoints = endpointTable{}
+			n.once = &sync.Once{}
+			networks[nid] = n
+		}
+	}
+
+	return n
+}
+
+func (d *driver) getNetworkFromStore(nid string) *network {
+	if d.store == nil {
+		return nil
+	}
+
+	n := &network{id: nid}
+	if err := d.store.GetObject(datastore.Key(n.Key()...), n); err != nil {
+		return nil
+	}
+
+	return n
+}
+
+func (n *network) sandbox() osl.Sandbox {
+	n.Lock()
+	defer n.Unlock()
+
+	return n.sbox
+}
+
+func (n *network) setSandbox(sbox osl.Sandbox) {
+	n.Lock()
+	n.sbox = sbox
+	n.Unlock()
+}
+
+func (n *network) vxlanID(s *subnet) uint32 {
+	n.Lock()
+	defer n.Unlock()
+
+	return s.vni
+}
+
+func (n *network) setVxlanID(s *subnet, vni uint32) {
+	n.Lock()
+	s.vni = vni
+	n.Unlock()
+}
+
+func (n *network) Key() []string {
+	return []string{"overlay", "network", n.id}
+}
+
+func (n *network) KeyPrefix() []string {
+	return []string{"overlay", "network"}
+}
+
+func (n *network) Value() []byte {
+	netJSON := []*subnetJSON{}
+
+	for _, s := range n.subnets {
+		sj := &subnetJSON{
+			SubnetIP: s.subnetIP.String(),
+			GwIP:     s.gwIP.String(),
+			Vni:      s.vni,
+		}
+		netJSON = append(netJSON, sj)
+	}
+
+	b, err := json.Marshal(netJSON)
+
+	if err != nil {
+		return []byte{}
+	}
+	return b
+}
+
+func (n *network) Index() uint64 {
+	return n.dbIndex
+}
+
+func (n *network) SetIndex(index uint64) {
+	n.dbIndex = index
+	n.dbExists = true
+}
+
+func (n *network) Exists() bool {
+	return n.dbExists
+}
+
+func (n *network) Skip() bool {
+	return false
+}
+
+func (n *network) SetValue(value []byte) error {
+	var newNet bool
+	netJSON := []*subnetJSON{}
+
+	err := json.Unmarshal(value, &netJSON)
+	if err != nil {
+		return err
+	}
+
+	if len(n.subnets) == 0 {
+		newNet = true
+	}
+
+	for _, sj := range netJSON {
+		subnetIPstr := sj.SubnetIP
+		gwIPstr := sj.GwIP
+		vni := sj.Vni
+
+		subnetIP, _ := types.ParseCIDR(subnetIPstr)
+		gwIP, _ := types.ParseCIDR(gwIPstr)
+
+		if newNet {
+			s := &subnet{
+				subnetIP: subnetIP,
+				gwIP:     gwIP,
+				vni:      vni,
+				once:     &sync.Once{},
+			}
+			n.subnets = append(n.subnets, s)
+		} else {
+			sNet := n.getMatchingSubnet(subnetIP)
+			if sNet != nil {
+				sNet.vni = vni
+			}
+		}
+	}
+	return nil
+}
+
+func (n *network) DataScope() string {
+	return datastore.GlobalScope
+}
+
+func (n *network) writeToStore() error {
+	return n.driver.store.PutObjectAtomic(n)
+}
+
+func (n *network) releaseVxlanID() error {
+	if n.driver.store == nil {
+		return fmt.Errorf("no datastore configured. cannot release vxlan id")
+	}
+
+	if len(n.subnets) == 0 {
+		return nil
+	}
+
+	if err := n.driver.store.DeleteObjectAtomic(n); err != nil {
+		if err == datastore.ErrKeyModified || err == datastore.ErrKeyNotFound {
+			// In both the above cases we can safely assume that the key has been removed by some other
+			// instance and so simply get out of here
+			return nil
+		}
+
+		return fmt.Errorf("failed to delete network to vxlan id map: %v", err)
+	}
+
+	for _, s := range n.subnets {
+		n.driver.vxlanIdm.Release(uint64(n.vxlanID(s)))
+		n.setVxlanID(s, 0)
+	}
+	return nil
+}
+
+func (n *network) obtainVxlanID(s *subnet) error {
+	//return if the subnet already has a vxlan id assigned
+	if s.vni != 0 {
+		return nil
+	}
+
+	if n.driver.store == nil {
+		return fmt.Errorf("no datastore configured. cannot obtain vxlan id")
+	}
+
+	for {
+		if err := n.driver.store.GetObject(datastore.Key(n.Key()...), n); err != nil {
+			return fmt.Errorf("getting network %q from datastore failed %v", n.id, err)
+		}
+
+		if s.vni == 0 {
+			vxlanID, err := n.driver.vxlanIdm.GetID()
+			if err != nil {
+				return fmt.Errorf("failed to allocate vxlan id: %v", err)
+			}
+
+			n.setVxlanID(s, uint32(vxlanID))
+			if err := n.writeToStore(); err != nil {
+				n.driver.vxlanIdm.Release(uint64(n.vxlanID(s)))
+				n.setVxlanID(s, 0)
+				if err == datastore.ErrKeyModified {
+					continue
+				}
+				return fmt.Errorf("network %q failed to update data store: %v", n.id, err)
+			}
+			return nil
+		}
+		return nil
+	}
+}
+
+// getSubnetforIP returns the subnet to which the given IP belongs
+func (n *network) getSubnetforIP(ip *net.IPNet) *subnet {
+	for _, s := range n.subnets {
+		// first check if the mask lengths are the same
+		i, _ := s.subnetIP.Mask.Size()
+		j, _ := ip.Mask.Size()
+		if i != j {
+			continue
+		}
+		if s.subnetIP.Contains(ip.IP) {
+			return s
+		}
+	}
+	return nil
+}
+
+// getMatchingSubnet return the network's subnet that matches the input
+func (n *network) getMatchingSubnet(ip *net.IPNet) *subnet {
+	if ip == nil {
+		return nil
+	}
+	for _, s := range n.subnets {
+		// first check if the mask lengths are the same
+		i, _ := s.subnetIP.Mask.Size()
+		j, _ := ip.Mask.Size()
+		if i != j {
+			continue
+		}
+		if s.subnetIP.IP.Equal(ip.IP) {
+			return s
+		}
+	}
+	return nil
+}

vendor/github.com/docker/libnetwork/drivers/overlay/ov_serf.go

@@ -0,0 +1,234 @@
+package overlay
+
+import (
+	"fmt"
+	"net"
+	"strings"
+	"time"
+
+	"github.com/Sirupsen/logrus"
+	"github.com/hashicorp/serf/serf"
+)
+
+type ovNotify struct {
+	action string
+	eid    string
+	nid    string
+}
+
+type logWriter struct{}
+
+func (l *logWriter) Write(p []byte) (int, error) {
+	str := string(p)
+
+	switch {
+	case strings.Contains(str, "[WARN]"):
+		logrus.Warn(str)
+	case strings.Contains(str, "[DEBUG]"):
+		logrus.Debug(str)
+	case strings.Contains(str, "[INFO]"):
+		logrus.Info(str)
+	case strings.Contains(str, "[ERR]"):
+		logrus.Error(str)
+	}
+
+	return len(p), nil
+}
+
+func (d *driver) serfInit() error {
+	var err error
+
+	config := serf.DefaultConfig()
+	config.Init()
+	config.MemberlistConfig.BindAddr = d.bindAddress
+
+	d.eventCh = make(chan serf.Event, 4)
+	config.EventCh = d.eventCh
+	config.UserCoalescePeriod = 1 * time.Second
+	config.UserQuiescentPeriod = 50 * time.Millisecond
+
+	config.LogOutput = &logWriter{}
+	config.MemberlistConfig.LogOutput = config.LogOutput
+
+	s, err := serf.Create(config)
+	if err != nil {
+		return fmt.Errorf("failed to create cluster node: %v", err)
+	}
+	defer func() {
+		if err != nil {
+			s.Shutdown()
+		}
+	}()
+
+	d.serfInstance = s
+
+	d.notifyCh = make(chan ovNotify)
+	d.exitCh = make(chan chan struct{})
+
+	go d.startSerfLoop(d.eventCh, d.notifyCh, d.exitCh)
+	return nil
+}
+
+func (d *driver) serfJoin(neighIP string) error {
+	if neighIP == "" {
+		return fmt.Errorf("no neighbor to join")
+	}
+	if _, err := d.serfInstance.Join([]string{neighIP}, false); err != nil {
+		return fmt.Errorf("Failed to join the cluster at neigh IP %s: %v",
+			neighIP, err)
+	}
+	return nil
+}
+
+func (d *driver) notifyEvent(event ovNotify) {
+	n := d.network(event.nid)
+	ep := n.endpoint(event.eid)
+
+	ePayload := fmt.Sprintf("%s %s %s %s", event.action, ep.addr.IP.String(),
+		net.IP(ep.addr.Mask).String(), ep.mac.String())
+	eName := fmt.Sprintf("jl %s %s %s", d.serfInstance.LocalMember().Addr.String(),
+		event.nid, event.eid)
+
+	if err := d.serfInstance.UserEvent(eName, []byte(ePayload), true); err != nil {
+		logrus.Errorf("Sending user event failed: %v\n", err)
+	}
+}
+
+func (d *driver) processEvent(u serf.UserEvent) {
+	logrus.Debugf("Received user event name:%s, payload:%s\n", u.Name,
+		string(u.Payload))
+
+	var dummy, action, vtepStr, nid, eid, ipStr, maskStr, macStr string
+	if _, err := fmt.Sscan(u.Name, &dummy, &vtepStr, &nid, &eid); err != nil {
+		fmt.Printf("Failed to scan name string: %v\n", err)
+	}
+
+	if _, err := fmt.Sscan(string(u.Payload), &action,
+		&ipStr, &maskStr, &macStr); err != nil {
+		fmt.Printf("Failed to scan value string: %v\n", err)
+	}
+
+	logrus.Debugf("Parsed data = %s/%s/%s/%s/%s/%s\n", nid, eid, vtepStr, ipStr, maskStr, macStr)
+
+	mac, err := net.ParseMAC(macStr)
+	if err != nil {
+		logrus.Errorf("Failed to parse mac: %v\n", err)
+	}
+
+	if d.serfInstance.LocalMember().Addr.String() == vtepStr {
+		return
+	}
+
+	switch action {
+	case "join":
+		if err := d.peerAdd(nid, eid, net.ParseIP(ipStr), net.IPMask(net.ParseIP(maskStr).To4()), mac,
+			net.ParseIP(vtepStr), true); err != nil {
+			logrus.Errorf("Peer add failed in the driver: %v\n", err)
+		}
+	case "leave":
+		if err := d.peerDelete(nid, eid, net.ParseIP(ipStr), net.IPMask(net.ParseIP(maskStr).To4()), mac,
+			net.ParseIP(vtepStr), true); err != nil {
+			logrus.Errorf("Peer delete failed in the driver: %v\n", err)
+		}
+	}
+}
+
+func (d *driver) processQuery(q *serf.Query) {
+	logrus.Debugf("Received query name:%s, payload:%s\n", q.Name,
+		string(q.Payload))
+
+	var nid, ipStr string
+	if _, err := fmt.Sscan(string(q.Payload), &nid, &ipStr); err != nil {
+		fmt.Printf("Failed to scan query payload string: %v\n", err)
+	}
+
+	peerMac, peerIPMask, vtep, err := d.peerDbSearch(nid, net.ParseIP(ipStr))
+	if err != nil {
+		return
+	}
+
+	q.Respond([]byte(fmt.Sprintf("%s %s %s", peerMac.String(), net.IP(peerIPMask).String(), vtep.String())))
+}
+
+func (d *driver) resolvePeer(nid string, peerIP net.IP) (net.HardwareAddr, net.IPMask, net.IP, error) {
+	if d.serfInstance == nil {
+		return nil, nil, nil, fmt.Errorf("could not resolve peer: serf instance not initialized")
+	}
+
+	qPayload := fmt.Sprintf("%s %s", string(nid), peerIP.String())
+	resp, err := d.serfInstance.Query("peerlookup", []byte(qPayload), nil)
+	if err != nil {
+		return nil, nil, nil, fmt.Errorf("resolving peer by querying the cluster failed: %v", err)
+	}
+
+	respCh := resp.ResponseCh()
+	select {
+	case r := <-respCh:
+		var macStr, maskStr, vtepStr string
+		if _, err := fmt.Sscan(string(r.Payload), &macStr, &maskStr, &vtepStr); err != nil {
+			return nil, nil, nil, fmt.Errorf("bad response %q for the resolve query: %v", string(r.Payload), err)
+		}
+
+		mac, err := net.ParseMAC(macStr)
+		if err != nil {
+			return nil, nil, nil, fmt.Errorf("failed to parse mac: %v", err)
+		}
+
+		return mac, net.IPMask(net.ParseIP(maskStr).To4()), net.ParseIP(vtepStr), nil
+
+	case <-time.After(time.Second):
+		return nil, nil, nil, fmt.Errorf("timed out resolving peer by querying the cluster")
+	}
+}
+
+func (d *driver) startSerfLoop(eventCh chan serf.Event, notifyCh chan ovNotify,
+	exitCh chan chan struct{}) {
+
+	for {
+		select {
+		case notify, ok := <-notifyCh:
+			if !ok {
+				break
+			}
+
+			d.notifyEvent(notify)
+		case ch, ok := <-exitCh:
+			if !ok {
+				break
+			}
+
+			if err := d.serfInstance.Leave(); err != nil {
+				logrus.Errorf("failed leaving the cluster: %v\n", err)
+			}
+
+			d.serfInstance.Shutdown()
+			close(ch)
+			return
+		case e, ok := <-eventCh:
+			if !ok {
+				break
+			}
+
+			if e.EventType() == serf.EventQuery {
+				d.processQuery(e.(*serf.Query))
+				break
+			}
+
+			u, ok := e.(serf.UserEvent)
+			if !ok {
+				break
+			}
+			d.processEvent(u)
+		}
+	}
+}
+
+func (d *driver) isSerfAlive() bool {
+	d.Lock()
+	serfInstance := d.serfInstance
+	d.Unlock()
+	if serfInstance == nil || serfInstance.State() != serf.SerfAlive {
+		return false
+	}
+	return true
+}

vendor/github.com/docker/libnetwork/drivers/overlay/ov_utils.go

@@ -0,0 +1,105 @@
+package overlay
+
+import (
+	"fmt"
+
+	"github.com/docker/libnetwork/netutils"
+	"github.com/docker/libnetwork/osl"
+	"github.com/vishvananda/netlink"
+	"github.com/vishvananda/netlink/nl"
+)
+
+func validateID(nid, eid string) error {
+	if nid == "" {
+		return fmt.Errorf("invalid network id")
+	}
+
+	if eid == "" {
+		return fmt.Errorf("invalid endpoint id")
+	}
+
+	return nil
+}
+
+func createVethPair() (string, string, error) {
+	defer osl.InitOSContext()()
+
+	// Generate a name for what will be the host side pipe interface
+	name1, err := netutils.GenerateIfaceName(vethPrefix, vethLen)
+	if err != nil {
+		return "", "", fmt.Errorf("error generating veth name1: %v", err)
+	}
+
+	// Generate a name for what will be the sandbox side pipe interface
+	name2, err := netutils.GenerateIfaceName(vethPrefix, vethLen)
+	if err != nil {
+		return "", "", fmt.Errorf("error generating veth name2: %v", err)
+	}
+
+	// Generate and add the interface pipe host <-> sandbox
+	veth := &netlink.Veth{
+		LinkAttrs: netlink.LinkAttrs{Name: name1, TxQLen: 0},
+		PeerName:  name2}
+	if err := netlink.LinkAdd(veth); err != nil {
+		return "", "", fmt.Errorf("error creating veth pair: %v", err)
+	}
+
+	return name1, name2, nil
+}
+
+func createVxlan(name string, vni uint32) error {
+	defer osl.InitOSContext()()
+
+	vxlan := &netlink.Vxlan{
+		LinkAttrs: netlink.LinkAttrs{Name: name},
+		VxlanId:   int(vni),
+		Learning:  true,
+		Port:      int(nl.Swap16(vxlanPort)), //network endian order
+		Proxy:     true,
+		L3miss:    true,
+		L2miss:    true,
+	}
+
+	if err := netlink.LinkAdd(vxlan); err != nil {
+		return fmt.Errorf("error creating vxlan interface: %v", err)
+	}
+
+	return nil
+}
+
+func deleteInterface(name string) error {
+	defer osl.InitOSContext()()
+
+	link, err := netlink.LinkByName(name)
+	if err != nil {
+		return fmt.Errorf("failed to find interface with name %s: %v", name, err)
+	}
+
+	if err := netlink.LinkDel(link); err != nil {
+		return fmt.Errorf("error deleting interface with name %s: %v", name, err)
+	}
+
+	return nil
+}
+
+func deleteVxlanByVNI(vni uint32) error {
+	defer osl.InitOSContext()()
+
+	links, err := netlink.LinkList()
+	if err != nil {
+		return fmt.Errorf("failed to list interfaces while deleting vxlan interface by vni: %v", err)
+	}
+
+	for _, l := range links {
+		if l.Type() == "vxlan" && l.(*netlink.Vxlan).VxlanId == int(vni) {
+			err = netlink.LinkDel(l)
+			if err != nil {
+				return fmt.Errorf("error deleting vxlan interface with id %d: %v", vni, err)
+			}
+
+			return nil
+		}
+	}
+
+	return fmt.Errorf("could not find a vxlan interface to delete with id %d", vni)
+}

vendor/github.com/docker/libnetwork/drivers/overlay/overlay.go

@@ -0,0 +1,209 @@
+package overlay
+
+import (
+	"fmt"
+	"net"
+	"sync"
+
+	"github.com/Sirupsen/logrus"
+	"github.com/docker/libkv/store"
+	"github.com/docker/libnetwork/datastore"
+	"github.com/docker/libnetwork/driverapi"
+	"github.com/docker/libnetwork/idm"
+	"github.com/docker/libnetwork/netlabel"
+	"github.com/hashicorp/serf/serf"
+)
+
+const (
+	networkType  = "overlay"
+	vethPrefix   = "veth"
+	vethLen      = 7
+	vxlanIDStart = 256
+	vxlanIDEnd   = 1000
+	vxlanPort    = 4789
+	vxlanVethMTU = 1450
+)
+
+type driver struct {
+	eventCh      chan serf.Event
+	notifyCh     chan ovNotify
+	exitCh       chan chan struct{}
+	bindAddress  string
+	neighIP      string
+	config       map[string]interface{}
+	peerDb       peerNetworkMap
+	serfInstance *serf.Serf
+	networks     networkTable
+	store        datastore.DataStore
+	ipAllocator  *idm.Idm
+	vxlanIdm     *idm.Idm
+	once         sync.Once
+	joinOnce     sync.Once
+	sync.Mutex
+}
+
+// Init registers a new instance of overlay driver
+func Init(dc driverapi.DriverCallback, config map[string]interface{}) error {
+	c := driverapi.Capability{
+		DataScope: datastore.GlobalScope,
+	}
+
+	d := &driver{
+		networks: networkTable{},
+		peerDb: peerNetworkMap{
+			mp: map[string]*peerMap{},
+		},
+		config: config,
+	}
+
+	return dc.RegisterDriver(networkType, d, c)
+}
+
+// Fini cleans up the driver resources
+func Fini(drv driverapi.Driver) {
+	d := drv.(*driver)
+
+	if d.exitCh != nil {
+		waitCh := make(chan struct{})
+
+		d.exitCh <- waitCh
+
+		<-waitCh
+	}
+}
+
+func (d *driver) configure() error {
+	var err error
+
+	if len(d.config) == 0 {
+		return nil
+	}
+
+	d.once.Do(func() {
+		provider, provOk := d.config[netlabel.GlobalKVProvider]
+		provURL, urlOk := d.config[netlabel.GlobalKVProviderURL]
+
+		if provOk && urlOk {
+			cfg := &datastore.ScopeCfg{
+				Client: datastore.ScopeClientCfg{
+					Provider: provider.(string),
+					Address:  provURL.(string),
+				},
+			}
+			provConfig, confOk := d.config[netlabel.GlobalKVProviderConfig]
+			if confOk {
+				cfg.Client.Config = provConfig.(*store.Config)
+			}
+			d.store, err = datastore.NewDataStore(datastore.GlobalScope, cfg)
+			if err != nil {
+				err = fmt.Errorf("failed to initialize data store: %v", err)
+				return
+			}
+		}
+
+		d.vxlanIdm, err = idm.New(d.store, "vxlan-id", vxlanIDStart, vxlanIDEnd)
+		if err != nil {
+			err = fmt.Errorf("failed to initialize vxlan id manager: %v", err)
+			return
+		}
+
+		d.ipAllocator, err = idm.New(d.store, "ipam-id", 1, 0xFFFF-2)
+		if err != nil {
+			err = fmt.Errorf("failed to initalize ipam id manager: %v", err)
+			return
+		}
+	})
+
+	return err
+}
+
+func (d *driver) Type() string {
+	return networkType
+}
+
+func validateSelf(node string) error {
+	advIP := net.ParseIP(node)
+	if advIP == nil {
+		return fmt.Errorf("invalid self address (%s)", node)
+	}
+
+	addrs, err := net.InterfaceAddrs()
+	if err != nil {
+		return fmt.Errorf("Unable to get interface addresses %v", err)
+	}
+	for _, addr := range addrs {
+		ip, _, err := net.ParseCIDR(addr.String())
+		if err == nil && ip.Equal(advIP) {
+			return nil
+		}
+	}
+	return fmt.Errorf("Multi-Host overlay networking requires cluster-advertise(%s) to be configured with a local ip-address that is reachable within the cluster", advIP.String())
+}
+
+func (d *driver) nodeJoin(node string, self bool) {
+	if self && !d.isSerfAlive() {
+		if err := validateSelf(node); err != nil {
+			logrus.Errorf("%s", err.Error())
+		}
+		d.Lock()
+		d.bindAddress = node
+		d.Unlock()
+		err := d.serfInit()
+		if err != nil {
+			logrus.Errorf("initializing serf instance failed: %v", err)
+			return
+		}
+	}
+
+	d.Lock()
+	if !self {
+		d.neighIP = node
+	}
+	neighIP := d.neighIP
+	d.Unlock()
+
+	if d.serfInstance != nil && neighIP != "" {
+		var err error
+		d.joinOnce.Do(func() {
+			err = d.serfJoin(neighIP)
+			if err == nil {
+				d.pushLocalDb()
+			}
+		})
+		if err != nil {
+			logrus.Errorf("joining serf neighbor %s failed: %v", node, err)
+			d.Lock()
+			d.joinOnce = sync.Once{}
+			d.Unlock()
+			return
+		}
+	}
+}
+
+func (d *driver) pushLocalEndpointEvent(action, nid, eid string) {
+	if !d.isSerfAlive() {
+		return
+	}
+	d.notifyCh <- ovNotify{
+		action: "join",
+		nid:    nid,
+		eid:    eid,
+	}
+}
+
+// DiscoverNew is a notification for a new discovery event, such as a new node joining a cluster
+func (d *driver) DiscoverNew(dType driverapi.DiscoveryType, data interface{}) error {
+	if dType == driverapi.NodeDiscovery {
+		nodeData, ok := data.(driverapi.NodeDiscoveryData)
+		if !ok || nodeData.Address == "" {
+			return fmt.Errorf("invalid discovery data")
+		}
+		d.nodeJoin(nodeData.Address, nodeData.Self)
+	}
+	return nil
+}
+
+// DiscoverDelete is a notification for a discovery delete event, such as a node leaving a cluster
+func (d *driver) DiscoverDelete(dType driverapi.DiscoveryType, data interface{}) error {
+	return nil
+}

vendor/github.com/docker/libnetwork/drivers/overlay/overlay_test.go

@@ -0,0 +1,134 @@
+package overlay
+
+import (
+	"net"
+	"testing"
+	"time"
+
+	"github.com/docker/libnetwork/driverapi"
+	_ "github.com/docker/libnetwork/testutils"
+)
+
+type driverTester struct {
+	t *testing.T
+	d *driver
+}
+
+const testNetworkType = "overlay"
+
+func setupDriver(t *testing.T) *driverTester {
+	dt := &driverTester{t: t}
+	if err := Init(dt, nil); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := dt.d.configure(); err != nil {
+		t.Fatal(err)
+	}
+
+	iface, err := net.InterfaceByName("eth0")
+	if err != nil {
+		t.Fatal(err)
+	}
+	addrs, err := iface.Addrs()
+	if err != nil || len(addrs) == 0 {
+		t.Fatal(err)
+	}
+	data := driverapi.NodeDiscoveryData{
+		Address: addrs[0].String(),
+		Self:    true,
+	}
+	dt.d.DiscoverNew(driverapi.NodeDiscovery, data)
+	return dt
+}
+
+func cleanupDriver(t *testing.T, dt *driverTester) {
+	ch := make(chan struct{})
+	go func() {
+		Fini(dt.d)
+		close(ch)
+	}()
+
+	select {
+	case <-ch:
+	case <-time.After(10 * time.Second):
+		t.Fatal("test timed out because Fini() did not return on time")
+	}
+}
+
+func (dt *driverTester) RegisterDriver(name string, drv driverapi.Driver,
+	cap driverapi.Capability) error {
+	if name != testNetworkType {
+		dt.t.Fatalf("Expected driver register name to be %q. Instead got %q",
+			testNetworkType, name)
+	}
+
+	if _, ok := drv.(*driver); !ok {
+		dt.t.Fatalf("Expected driver type to be %T. Instead got %T",
+			&driver{}, drv)
+	}
+
+	dt.d = drv.(*driver)
+	return nil
+}
+
+func TestOverlayInit(t *testing.T) {
+	if err := Init(&driverTester{t: t}, nil); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestOverlayFiniWithoutConfig(t *testing.T) {
+	dt := &driverTester{t: t}
+	if err := Init(dt, nil); err != nil {
+		t.Fatal(err)
+	}
+
+	cleanupDriver(t, dt)
+}
+
+func TestOverlayNilConfig(t *testing.T) {
+	dt := &driverTester{t: t}
+	if err := Init(dt, nil); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := dt.d.configure(); err != nil {
+		t.Fatal(err)
+	}
+
+	cleanupDriver(t, dt)
+}
+
+func TestOverlayConfig(t *testing.T) {
+	dt := setupDriver(t)
+
+	time.Sleep(1 * time.Second)
+
+	d := dt.d
+	if d.notifyCh == nil {
+		t.Fatal("Driver notify channel wasn't initialzed after Config method")
+	}
+
+	if d.exitCh == nil {
+		t.Fatal("Driver serfloop exit channel wasn't initialzed after Config method")
+	}
+
+	if d.serfInstance == nil {
+		t.Fatal("Driver serfinstance  hasn't been initialized after Config method")
+	}
+
+	cleanupDriver(t, dt)
+}
+
+func TestOverlayType(t *testing.T) {
+	dt := &driverTester{t: t}
+	if err := Init(dt, nil); err != nil {
+		t.Fatal(err)
+	}
+
+	if dt.d.Type() != testNetworkType {
+		t.Fatalf("Expected Type() to return %q. Instead got %q", testNetworkType,
+			dt.d.Type())
+	}
+}

vendor/github.com/docker/libnetwork/drivers/overlay/peerdb.go

@@ -0,0 +1,329 @@
+package overlay
+
+import (
+	"fmt"
+	"net"
+	"sync"
+	"syscall"
+)
+
+type peerKey struct {
+	peerIP  net.IP
+	peerMac net.HardwareAddr
+}
+
+type peerEntry struct {
+	eid        string
+	vtep       net.IP
+	peerIPMask net.IPMask
+	inSandbox  bool
+	isLocal    bool
+}
+
+type peerMap struct {
+	mp map[string]peerEntry
+	sync.Mutex
+}
+
+type peerNetworkMap struct {
+	mp map[string]*peerMap
+	sync.Mutex
+}
+
+func (pKey peerKey) String() string {
+	return fmt.Sprintf("%s %s", pKey.peerIP, pKey.peerMac)
+}
+
+func (pKey *peerKey) Scan(state fmt.ScanState, verb rune) error {
+	ipB, err := state.Token(true, nil)
+	if err != nil {
+		return err
+	}
+
+	pKey.peerIP = net.ParseIP(string(ipB))
+
+	macB, err := state.Token(true, nil)
+	if err != nil {
+		return err
+	}
+
+	pKey.peerMac, err = net.ParseMAC(string(macB))
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+var peerDbWg sync.WaitGroup
+
+func (d *driver) peerDbWalk(f func(string, *peerKey, *peerEntry) bool) error {
+	d.peerDb.Lock()
+	nids := []string{}
+	for nid := range d.peerDb.mp {
+		nids = append(nids, nid)
+	}
+	d.peerDb.Unlock()
+
+	for _, nid := range nids {
+		d.peerDbNetworkWalk(nid, func(pKey *peerKey, pEntry *peerEntry) bool {
+			return f(nid, pKey, pEntry)
+		})
+	}
+	return nil
+}
+
+func (d *driver) peerDbNetworkWalk(nid string, f func(*peerKey, *peerEntry) bool) error {
+	d.peerDb.Lock()
+	pMap, ok := d.peerDb.mp[nid]
+	if !ok {
+		d.peerDb.Unlock()
+		return nil
+	}
+	d.peerDb.Unlock()
+
+	pMap.Lock()
+	for pKeyStr, pEntry := range pMap.mp {
+		var pKey peerKey
+		if _, err := fmt.Sscan(pKeyStr, &pKey); err != nil {
+			fmt.Printf("peer key scan failed: %v", err)
+		}
+
+		if f(&pKey, &pEntry) {
+			pMap.Unlock()
+			return nil
+		}
+	}
+	pMap.Unlock()
+
+	return nil
+}
+
+func (d *driver) peerDbSearch(nid string, peerIP net.IP) (net.HardwareAddr, net.IPMask, net.IP, error) {
+	var (
+		peerMac    net.HardwareAddr
+		vtep       net.IP
+		peerIPMask net.IPMask
+		found      bool
+	)
+
+	err := d.peerDbNetworkWalk(nid, func(pKey *peerKey, pEntry *peerEntry) bool {
+		if pKey.peerIP.Equal(peerIP) {
+			peerMac = pKey.peerMac
+			peerIPMask = pEntry.peerIPMask
+			vtep = pEntry.vtep
+			found = true
+			return found
+		}
+
+		return found
+	})
+
+	if err != nil {
+		return nil, nil, nil, fmt.Errorf("peerdb search for peer ip %q failed: %v", peerIP, err)
+	}
+
+	if !found {
+		return nil, nil, nil, fmt.Errorf("peer ip %q not found in peerdb", peerIP)
+	}
+
+	return peerMac, peerIPMask, vtep, nil
+}
+
+func (d *driver) peerDbAdd(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
+	peerMac net.HardwareAddr, vtep net.IP, isLocal bool) {
+
+	peerDbWg.Wait()
+
+	d.peerDb.Lock()
+	pMap, ok := d.peerDb.mp[nid]
+	if !ok {
+		d.peerDb.mp[nid] = &peerMap{
+			mp: make(map[string]peerEntry),
+		}
+
+		pMap = d.peerDb.mp[nid]
+	}
+	d.peerDb.Unlock()
+
+	pKey := peerKey{
+		peerIP:  peerIP,
+		peerMac: peerMac,
+	}
+
+	pEntry := peerEntry{
+		eid:        eid,
+		vtep:       vtep,
+		peerIPMask: peerIPMask,
+		isLocal:    isLocal,
+	}
+
+	pMap.Lock()
+	pMap.mp[pKey.String()] = pEntry
+	pMap.Unlock()
+}
+
+func (d *driver) peerDbDelete(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
+	peerMac net.HardwareAddr, vtep net.IP) {
+	peerDbWg.Wait()
+
+	d.peerDb.Lock()
+	pMap, ok := d.peerDb.mp[nid]
+	if !ok {
+		d.peerDb.Unlock()
+		return
+	}
+	d.peerDb.Unlock()
+
+	pKey := peerKey{
+		peerIP:  peerIP,
+		peerMac: peerMac,
+	}
+
+	pMap.Lock()
+	delete(pMap.mp, pKey.String())
+	pMap.Unlock()
+}
+
+func (d *driver) peerDbUpdateSandbox(nid string) {
+	d.peerDb.Lock()
+	pMap, ok := d.peerDb.mp[nid]
+	if !ok {
+		d.peerDb.Unlock()
+		return
+	}
+	d.peerDb.Unlock()
+
+	peerDbWg.Add(1)
+
+	var peerOps []func()
+	pMap.Lock()
+	for pKeyStr, pEntry := range pMap.mp {
+		var pKey peerKey
+		if _, err := fmt.Sscan(pKeyStr, &pKey); err != nil {
+			fmt.Printf("peer key scan failed: %v", err)
+		}
+
+		if pEntry.isLocal {
+			continue
+		}
+
+		// Go captures variables by reference. The pEntry could be
+		// pointing to the same memory location for every iteration. Make
+		// a copy of pEntry before capturing it in the following closure.
+		entry := pEntry
+		op := func() {
+			if err := d.peerAdd(nid, entry.eid, pKey.peerIP, entry.peerIPMask,
+				pKey.peerMac, entry.vtep,
+				false); err != nil {
+				fmt.Printf("peerdbupdate in sandbox failed for ip %s and mac %s: %v",
+					pKey.peerIP, pKey.peerMac, err)
+			}
+		}
+
+		peerOps = append(peerOps, op)
+	}
+	pMap.Unlock()
+
+	for _, op := range peerOps {
+		op()
+	}
+
+	peerDbWg.Done()
+}
+
+func (d *driver) peerAdd(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
+	peerMac net.HardwareAddr, vtep net.IP, updateDb bool) error {
+
+	if err := validateID(nid, eid); err != nil {
+		return err
+	}
+
+	if updateDb {
+		d.peerDbAdd(nid, eid, peerIP, peerIPMask, peerMac, vtep, false)
+	}
+
+	n := d.network(nid)
+	if n == nil {
+		return nil
+	}
+
+	sbox := n.sandbox()
+	if sbox == nil {
+		return nil
+	}
+
+	IP := &net.IPNet{
+		IP:   peerIP,
+		Mask: peerIPMask,
+	}
+
+	s := n.getSubnetforIP(IP)
+	if s == nil {
+		return fmt.Errorf("couldn't find the subnet %q in network %q\n", IP.String(), n.id)
+	}
+
+	if err := n.obtainVxlanID(s); err != nil {
+		return fmt.Errorf("couldn't get vxlan id for %q: %v", s.subnetIP.String(), err)
+	}
+
+	if err := n.joinSubnetSandbox(s); err != nil {
+		return fmt.Errorf("subnet sandbox join failed for %q: %v", s.subnetIP.String(), err)
+	}
+
+	// Add neighbor entry for the peer IP
+	if err := sbox.AddNeighbor(peerIP, peerMac, sbox.NeighborOptions().LinkName(s.vxlanName)); err != nil {
+		return fmt.Errorf("could not add neigbor entry into the sandbox: %v", err)
+	}
+
+	// Add fdb entry to the bridge for the peer mac
+	if err := sbox.AddNeighbor(vtep, peerMac, sbox.NeighborOptions().LinkName(s.vxlanName),
+		sbox.NeighborOptions().Family(syscall.AF_BRIDGE)); err != nil {
+		return fmt.Errorf("could not add fdb entry into the sandbox: %v", err)
+	}
+
+	return nil
+}
+
+func (d *driver) peerDelete(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
+	peerMac net.HardwareAddr, vtep net.IP, updateDb bool) error {
+
+	if err := validateID(nid, eid); err != nil {
+		return err
+	}
+
+	if updateDb {
+		d.peerDbDelete(nid, eid, peerIP, peerIPMask, peerMac, vtep)
+	}
+
+	n := d.network(nid)
+	if n == nil {
+		return nil
+	}
+
+	sbox := n.sandbox()
+	if sbox == nil {
+		return nil
+	}
+
+	// Delete fdb entry to the bridge for the peer mac
+	if err := sbox.DeleteNeighbor(vtep, peerMac); err != nil {
+		return fmt.Errorf("could not delete fdb entry into the sandbox: %v", err)
+	}
+
+	// Delete neighbor entry for the peer IP
+	if err := sbox.DeleteNeighbor(peerIP, peerMac); err != nil {
+		return fmt.Errorf("could not delete neigbor entry into the sandbox: %v", err)
+	}
+
+	return nil
+}
+
+func (d *driver) pushLocalDb() {
+	d.peerDbWalk(func(nid string, pKey *peerKey, pEntry *peerEntry) bool {
+		if pEntry.isLocal {
+			d.pushLocalEndpointEvent("join", nid, pEntry.eid)
+		}
+		return false
+	})
+}