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

527 lines
14 KiB
Go
Raw Normal View History

package overlay
import (
"context"
"fmt"
"net"
"sync"
"syscall"
"github.com/docker/libnetwork/internal/caller"
"github.com/docker/libnetwork/internal/setmatrix"
"github.com/docker/libnetwork/osl"
"github.com/sirupsen/logrus"
)
const ovPeerTable = "overlay_peer_table"
type peerKey struct {
peerIP net.IP
peerMac net.HardwareAddr
}
type peerEntry struct {
eid string
vtep net.IP
peerIPMask net.IPMask
isLocal bool
}
func (p *peerEntry) MarshalDB() peerEntryDB {
ones, bits := p.peerIPMask.Size()
return peerEntryDB{
eid: p.eid,
vtep: p.vtep.String(),
peerIPMaskOnes: ones,
peerIPMaskBits: bits,
isLocal: p.isLocal,
}
}
// This the structure saved into the set (SetMatrix), due to the implementation of it
// the value inserted in the set has to be Hashable so the []byte had to be converted into
// strings
type peerEntryDB struct {
eid string
vtep string
peerIPMaskOnes int
peerIPMaskBits int
isLocal bool
}
func (p *peerEntryDB) UnMarshalDB() peerEntry {
return peerEntry{
eid: p.eid,
vtep: net.ParseIP(p.vtep),
peerIPMask: net.CIDRMask(p.peerIPMaskOnes, p.peerIPMaskBits),
isLocal: p.isLocal,
}
}
type peerMap struct {
// set of peerEntry, note they have to be objects and not pointers to maintain the proper equality checks
mp setmatrix.SetMatrix
sync.Mutex
}
type peerNetworkMap struct {
// map with key peerKey
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))
return err
}
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]
d.peerDb.Unlock()
if !ok {
return nil
}
mp := map[string]peerEntry{}
pMap.Lock()
for _, pKeyStr := range pMap.mp.Keys() {
entryDBList, ok := pMap.mp.Get(pKeyStr)
if ok {
peerEntryDB := entryDBList[0].(peerEntryDB)
mp[pKeyStr] = peerEntryDB.UnMarshalDB()
}
}
pMap.Unlock()
for pKeyStr, pEntry := range mp {
var pKey peerKey
if _, err := fmt.Sscan(pKeyStr, &pKey); err != nil {
logrus.Warnf("Peer key scan on network %s failed: %v", nid, err)
}
if f(&pKey, &pEntry) {
return nil
}
}
return nil
}
func (d *driver) peerDbSearch(nid string, peerIP net.IP) (*peerKey, *peerEntry, error) {
var pKeyMatched *peerKey
var pEntryMatched *peerEntry
err := d.peerDbNetworkWalk(nid, func(pKey *peerKey, pEntry *peerEntry) bool {
if pKey.peerIP.Equal(peerIP) {
pKeyMatched = pKey
pEntryMatched = pEntry
return true
}
return false
})
if err != nil {
return nil, nil, fmt.Errorf("peerdb search for peer ip %q failed: %v", peerIP, err)
}
if pKeyMatched == nil || pEntryMatched == nil {
return nil, nil, fmt.Errorf("peer ip %q not found in peerdb", peerIP)
}
return pKeyMatched, pEntryMatched, nil
}
func (d *driver) peerDbAdd(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
peerMac net.HardwareAddr, vtep net.IP, isLocal bool) (bool, int) {
d.peerDb.Lock()
pMap, ok := d.peerDb.mp[nid]
if !ok {
d.peerDb.mp[nid] = &peerMap{
mp: setmatrix.NewSetMatrix(),
}
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()
defer pMap.Unlock()
b, i := pMap.mp.Insert(pKey.String(), pEntry.MarshalDB())
if i != 1 {
// Transient case, there is more than one endpoint that is using the same IP,MAC pair
s, _ := pMap.mp.String(pKey.String())
logrus.Warnf("peerDbAdd transient condition - Key:%s cardinality:%d db state:%s", pKey.String(), i, s)
}
return b, i
}
func (d *driver) peerDbDelete(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
peerMac net.HardwareAddr, vtep net.IP, isLocal bool) (bool, int) {
d.peerDb.Lock()
pMap, ok := d.peerDb.mp[nid]
if !ok {
d.peerDb.Unlock()
return false, 0
}
d.peerDb.Unlock()
pKey := peerKey{
peerIP: peerIP,
peerMac: peerMac,
}
pEntry := peerEntry{
eid: eid,
vtep: vtep,
peerIPMask: peerIPMask,
isLocal: isLocal,
}
pMap.Lock()
defer pMap.Unlock()
b, i := pMap.mp.Remove(pKey.String(), pEntry.MarshalDB())
if i != 0 {
// Transient case, there is more than one endpoint that is using the same IP,MAC pair
s, _ := pMap.mp.String(pKey.String())
logrus.Warnf("peerDbDelete transient condition - Key:%s cardinality:%d db state:%s", pKey.String(), i, s)
}
return b, i
}
// The overlay uses a lazy initialization approach, this means that when a network is created
// and the driver registered the overlay does not allocate resources till the moment that a
// sandbox is actually created.
// At the moment of this call, that happens when a sandbox is initialized, is possible that
// networkDB has already delivered some events of peers already available on remote nodes,
// these peers are saved into the peerDB and this function is used to properly configure
// the network sandbox with all those peers that got previously notified.
// Note also that this method sends a single message on the channel and the go routine on the
// other side, will atomically loop on the whole table of peers and will program their state
// in one single atomic operation. This is fundamental to guarantee consistency, and avoid that
// new peerAdd or peerDelete gets reordered during the sandbox init.
func (d *driver) initSandboxPeerDB(nid string) {
d.peerInit(nid)
}
type peerOperationType int32
const (
peerOperationINIT peerOperationType = iota
peerOperationADD
peerOperationDELETE
peerOperationFLUSH
)
type peerOperation struct {
opType peerOperationType
networkID string
endpointID string
peerIP net.IP
peerIPMask net.IPMask
peerMac net.HardwareAddr
vtepIP net.IP
l2Miss bool
l3Miss bool
localPeer bool
callerName string
}
func (d *driver) peerOpRoutine(ctx context.Context, ch chan *peerOperation) {
var err error
for {
select {
case <-ctx.Done():
return
case op := <-ch:
switch op.opType {
case peerOperationINIT:
err = d.peerInitOp(op.networkID)
case peerOperationADD:
err = d.peerAddOp(op.networkID, op.endpointID, op.peerIP, op.peerIPMask, op.peerMac, op.vtepIP, op.l2Miss, op.l3Miss, true, op.localPeer)
case peerOperationDELETE:
err = d.peerDeleteOp(op.networkID, op.endpointID, op.peerIP, op.peerIPMask, op.peerMac, op.vtepIP, op.localPeer)
case peerOperationFLUSH:
err = d.peerFlushOp(op.networkID)
}
if err != nil {
logrus.Warnf("Peer operation failed:%s op:%v", err, op)
}
}
}
}
func (d *driver) peerInit(nid string) {
callerName := caller.Name(1)
d.peerOpCh <- &peerOperation{
opType: peerOperationINIT,
networkID: nid,
callerName: callerName,
}
}
func (d *driver) peerInitOp(nid string) error {
return d.peerDbNetworkWalk(nid, func(pKey *peerKey, pEntry *peerEntry) bool {
// Local entries do not need to be added
if pEntry.isLocal {
return false
}
d.peerAddOp(nid, pEntry.eid, pKey.peerIP, pEntry.peerIPMask, pKey.peerMac, pEntry.vtep, false, false, false, pEntry.isLocal)
// return false to loop on all entries
return false
})
}
func (d *driver) peerAdd(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
peerMac net.HardwareAddr, vtep net.IP, l2Miss, l3Miss, localPeer bool) {
d.peerOpCh <- &peerOperation{
opType: peerOperationADD,
networkID: nid,
endpointID: eid,
peerIP: peerIP,
peerIPMask: peerIPMask,
peerMac: peerMac,
vtepIP: vtep,
l2Miss: l2Miss,
l3Miss: l3Miss,
localPeer: localPeer,
callerName: caller.Name(1),
}
}
func (d *driver) peerAddOp(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
peerMac net.HardwareAddr, vtep net.IP, l2Miss, l3Miss, updateDB, localPeer bool) error {
if err := validateID(nid, eid); err != nil {
return err
}
var dbEntries int
var inserted bool
if updateDB {
inserted, dbEntries = d.peerDbAdd(nid, eid, peerIP, peerIPMask, peerMac, vtep, localPeer)
if !inserted {
logrus.Warnf("Entry already present in db: nid:%s eid:%s peerIP:%v peerMac:%v isLocal:%t vtep:%v",
nid, eid, peerIP, peerMac, localPeer, vtep)
}
}
// Local peers do not need any further configuration
if localPeer {
return nil
}
n := d.network(nid)
if n == nil {
return nil
}
sbox := n.sandbox()
if sbox == nil {
// We are hitting this case for all the events that are arriving before that the sandbox
// is being created. The peer got already added into the database and the sanbox init will
// call the peerDbUpdateSandbox that will configure all these peers from the database
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", 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)
}
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, false); err != nil {
return fmt.Errorf("subnet sandbox join failed for %q: %v", s.subnetIP.String(), err)
}
if err := d.checkEncryption(nid, vtep, n.vxlanID(s), false, true); err != nil {
logrus.Warn(err)
}
// Add neighbor entry for the peer IP
if err := sbox.AddNeighbor(peerIP, peerMac, l3Miss, sbox.NeighborOptions().LinkName(s.vxlanName)); err != nil {
if _, ok := err.(osl.NeighborSearchError); ok && dbEntries > 1 {
// We are in the transient case so only the first configuration is programmed into the kernel
// Upon deletion if the active configuration is deleted the next one from the database will be restored
// Note we are skipping also the next configuration
return nil
}
return fmt.Errorf("could not add neighbor entry for nid:%s eid:%s into the sandbox:%v", nid, eid, err)
}
// Add fdb entry to the bridge for the peer mac
if err := sbox.AddNeighbor(vtep, peerMac, l2Miss, sbox.NeighborOptions().LinkName(s.vxlanName),
sbox.NeighborOptions().Family(syscall.AF_BRIDGE)); err != nil {
return fmt.Errorf("could not add fdb entry for nid:%s eid:%s into the sandbox:%v", nid, eid, err)
}
return nil
}
func (d *driver) peerDelete(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
peerMac net.HardwareAddr, vtep net.IP, localPeer bool) {
d.peerOpCh <- &peerOperation{
opType: peerOperationDELETE,
networkID: nid,
endpointID: eid,
peerIP: peerIP,
peerIPMask: peerIPMask,
peerMac: peerMac,
vtepIP: vtep,
callerName: caller.Name(1),
localPeer: localPeer,
}
}
func (d *driver) peerDeleteOp(nid, eid string, peerIP net.IP, peerIPMask net.IPMask,
peerMac net.HardwareAddr, vtep net.IP, localPeer bool) error {
if err := validateID(nid, eid); err != nil {
return err
}
deleted, dbEntries := d.peerDbDelete(nid, eid, peerIP, peerIPMask, peerMac, vtep, localPeer)
if !deleted {
logrus.Warnf("Entry was not in db: nid:%s eid:%s peerIP:%v peerMac:%v isLocal:%t vtep:%v",
nid, eid, peerIP, peerMac, localPeer, vtep)
}
n := d.network(nid)
if n == nil {
return nil
}
sbox := n.sandbox()
if sbox == nil {
return nil
}
if err := d.checkEncryption(nid, vtep, 0, localPeer, false); err != nil {
logrus.Warn(err)
}
// Local peers do not have any local configuration to delete
if !localPeer {
// Remove fdb entry to the bridge for the peer mac
if err := sbox.DeleteNeighbor(vtep, peerMac, true); err != nil {
if _, ok := err.(osl.NeighborSearchError); ok && dbEntries > 0 {
// We fall in here if there is a transient state and if the neighbor that is being deleted
// was never been configured into the kernel (we allow only 1 configuration at the time per <ip,mac> mapping)
return nil
}
return fmt.Errorf("could not delete fdb entry for nid:%s eid:%s into the sandbox:%v", nid, eid, err)
}
// Delete neighbor entry for the peer IP
if err := sbox.DeleteNeighbor(peerIP, peerMac, true); err != nil {
return fmt.Errorf("could not delete neighbor entry for nid:%s eid:%s into the sandbox:%v", nid, eid, err)
}
}
if dbEntries == 0 {
return nil
}
// If there is still an entry into the database and the deletion went through without errors means that there is now no
// configuration active in the kernel.
// Restore one configuration for the <ip,mac> directly from the database, note that is guaranteed that there is one
peerKey, peerEntry, err := d.peerDbSearch(nid, peerIP)
if err != nil {
logrus.Errorf("peerDeleteOp unable to restore a configuration for nid:%s ip:%v mac:%v err:%s", nid, peerIP, peerMac, err)
return err
}
return d.peerAddOp(nid, peerEntry.eid, peerIP, peerEntry.peerIPMask, peerKey.peerMac, peerEntry.vtep, false, false, false, peerEntry.isLocal)
}
func (d *driver) peerFlush(nid string) {
d.peerOpCh <- &peerOperation{
opType: peerOperationFLUSH,
networkID: nid,
callerName: caller.Name(1),
}
}
func (d *driver) peerFlushOp(nid string) error {
d.peerDb.Lock()
defer d.peerDb.Unlock()
_, ok := d.peerDb.mp[nid]
if !ok {
return fmt.Errorf("Unable to find the peerDB for nid:%s", nid)
}
delete(d.peerDb.mp, nid)
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
})
}
func (d *driver) peerDBUpdateSelf() {
d.peerDbWalk(func(nid string, pkey *peerKey, pEntry *peerEntry) bool {
if pEntry.isLocal {
pEntry.vtep = net.ParseIP(d.advertiseAddress)
}
return false
})
}