moby--moby/libnetwork/drivers/overlay/joinleave.go

233 lines
5.9 KiB
Go
Raw Normal View History

package overlay
import (
"fmt"
"net"
"syscall"
"github.com/docker/libnetwork/driverapi"
"github.com/docker/libnetwork/ns"
"github.com/docker/libnetwork/types"
"github.com/gogo/protobuf/proto"
"github.com/sirupsen/logrus"
)
// 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)
}
if n.secure && len(d.keys) == 0 {
return fmt.Errorf("cannot join secure network: encryption keys not present")
}
nlh := ns.NlHandle()
if n.secure && !nlh.SupportsNetlinkFamily(syscall.NETLINK_XFRM) {
return fmt.Errorf("cannot join secure network: required modules to install IPSEC rules are missing on host")
}
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)
}
Refactor locking for join/leave to avoid race Instead of using "sync.Once" to determine whether to initialize a network sandbox or subnet sandbox, we use a traditional mutex + initialization boolean. This is because the initialization state isn't truly a once-and-done condition. Rather, libnetwork destroys network and subnet sandboxes when the last endpoint leaves them. The use of sync.Once in this kind of scenario requires, therefore, re-initializing the Once which is impoissible. So the approach that libnetwork currently takes is to use a pointer to a Once and redirect that pointer to a new Once on reset. This leads to nasty race conditions. In addition to refactoring the locking, this patch merges the functions joinSandbox(), and joinSubnetSandbox(). This makes the code both cleaner and it also holds the network and subnet locks through the series of read-modify-writes avoiding further potential races. This does reduce the potential parallelism which could be applied should there be many joins coming in on many different subnets in the same overlay network. However, this should be an extremely minor performance hit for a very obscure case. One important pattern in this commit is that it is crucial to avoid sending peerDB messages while holding a driver or network lock. The changes herein defer such (asynchronous) notifications until after release of such locks. This prevents deadlocks where the peerDB blocks acquiring said locks while the network method blocks trying to send to the peerDB's channel. Signed-off-by: Chris Telfer <ctelfer@docker.com>
2018-05-04 18:33:00 +00:00
if err := n.joinSandbox(s, false, true); err != nil {
return fmt.Errorf("network sandbox join failed: %v", err)
}
sbox := n.sandbox()
overlayIfName, containerIfName, err := createVethPair()
if err != nil {
return err
}
ep.ifName = containerIfName
if err = d.writeEndpointToStore(ep); err != nil {
return fmt.Errorf("failed to update overlay endpoint %.7s to local data store: %v", ep.id, err)
}
// 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))
mtu := n.maxMTU()
veth, err := nlh.LinkByName(overlayIfName)
if err != nil {
return fmt.Errorf("cound not find link by name %s: %v", overlayIfName, err)
}
err = nlh.LinkSetMTU(veth, mtu)
if err != nil {
return err
}
if err = sbox.AddInterface(overlayIfName, "veth",
sbox.InterfaceOptions().Master(s.brName)); err != nil {
return fmt.Errorf("could not add veth pair inside the network sandbox: %v", err)
}
veth, err = nlh.LinkByName(containerIfName)
if err != nil {
return fmt.Errorf("could not find link by name %s: %v", containerIfName, err)
}
err = nlh.LinkSetMTU(veth, mtu)
if err != nil {
return err
}
if err = nlh.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 {
logrus.Errorf("Adding subnet %s static route in network %q failed\n", s.subnetIP, n.id)
}
}
if iNames := jinfo.InterfaceName(); iNames != nil {
err = iNames.SetNames(containerIfName, "eth")
if err != nil {
return err
}
}
d.peerAdd(nid, eid, ep.addr.IP, ep.addr.Mask, ep.mac, net.ParseIP(d.advertiseAddress), false, false, true)
if err = d.checkEncryption(nid, nil, n.vxlanID(s), true, true); err != nil {
logrus.Warn(err)
}
buf, err := proto.Marshal(&PeerRecord{
EndpointIP: ep.addr.String(),
EndpointMAC: ep.mac.String(),
TunnelEndpointIP: d.advertiseAddress,
})
if err != nil {
return err
}
if err := jinfo.AddTableEntry(ovPeerTable, eid, buf); err != nil {
logrus.Errorf("overlay: Failed adding table entry to joininfo: %v", err)
}
d.pushLocalEndpointEvent("join", nid, eid)
return nil
}
func (d *driver) DecodeTableEntry(tablename string, key string, value []byte) (string, map[string]string) {
if tablename != ovPeerTable {
logrus.Errorf("DecodeTableEntry: unexpected table name %s", tablename)
return "", nil
}
var peer PeerRecord
if err := proto.Unmarshal(value, &peer); err != nil {
logrus.Errorf("DecodeTableEntry: failed to unmarshal peer record for key %s: %v", key, err)
return "", nil
}
return key, map[string]string{
"Host IP": peer.TunnelEndpointIP,
}
}
func (d *driver) EventNotify(etype driverapi.EventType, nid, tableName, key string, value []byte) {
if tableName != ovPeerTable {
logrus.Errorf("Unexpected table notification for table %s received", tableName)
return
}
eid := key
var peer PeerRecord
if err := proto.Unmarshal(value, &peer); err != nil {
logrus.Errorf("Failed to unmarshal peer record: %v", err)
return
}
// Ignore local peers. We already know about them and they
// should not be added to vxlan fdb.
if peer.TunnelEndpointIP == d.advertiseAddress {
return
}
addr, err := types.ParseCIDR(peer.EndpointIP)
if err != nil {
logrus.Errorf("Invalid peer IP %s received in event notify", peer.EndpointIP)
return
}
mac, err := net.ParseMAC(peer.EndpointMAC)
if err != nil {
logrus.Errorf("Invalid mac %s received in event notify", peer.EndpointMAC)
return
}
vtep := net.ParseIP(peer.TunnelEndpointIP)
if vtep == nil {
logrus.Errorf("Invalid VTEP %s received in event notify", peer.TunnelEndpointIP)
return
}
if etype == driverapi.Delete {
d.peerDelete(nid, eid, addr.IP, addr.Mask, mac, vtep, false)
return
}
d.peerAdd(nid, eid, addr.IP, addr.Mask, mac, vtep, false, false, false)
}
// 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",
nw: n,
ep: ep,
}
}
d.peerDelete(nid, eid, ep.addr.IP, ep.addr.Mask, ep.mac, net.ParseIP(d.advertiseAddress), true)
n.leaveSandbox()
return nil
}