diff options
Diffstat (limited to 'pkg/server')
-rw-r--r-- | pkg/server/bmp.go | 358 | ||||
-rw-r--r-- | pkg/server/collector.go | 222 | ||||
-rw-r--r-- | pkg/server/fsm.go | 1752 | ||||
-rw-r--r-- | pkg/server/fsm_test.go | 345 | ||||
-rw-r--r-- | pkg/server/grpc_server.go | 3016 | ||||
-rw-r--r-- | pkg/server/mrt.go | 409 | ||||
-rw-r--r-- | pkg/server/peer.go | 522 | ||||
-rw-r--r-- | pkg/server/rpki.go | 712 | ||||
-rw-r--r-- | pkg/server/rpki_test.go | 257 | ||||
-rw-r--r-- | pkg/server/server.go | 3048 | ||||
-rw-r--r-- | pkg/server/server_test.go | 706 | ||||
-rw-r--r-- | pkg/server/sockopt.go | 90 | ||||
-rw-r--r-- | pkg/server/sockopt_bsd.go | 89 | ||||
-rw-r--r-- | pkg/server/sockopt_darwin.go | 64 | ||||
-rw-r--r-- | pkg/server/sockopt_linux.go | 282 | ||||
-rw-r--r-- | pkg/server/sockopt_linux_test.go | 106 | ||||
-rw-r--r-- | pkg/server/sockopt_openbsd.go | 469 | ||||
-rw-r--r-- | pkg/server/sockopt_stub.go | 38 | ||||
-rw-r--r-- | pkg/server/util.go | 117 | ||||
-rw-r--r-- | pkg/server/zclient.go | 450 | ||||
-rw-r--r-- | pkg/server/zclient_test.go | 113 |
21 files changed, 13165 insertions, 0 deletions
diff --git a/pkg/server/bmp.go b/pkg/server/bmp.go new file mode 100644 index 00000000..7904fae1 --- /dev/null +++ b/pkg/server/bmp.go @@ -0,0 +1,358 @@ +// Copyright (C) 2015-2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "fmt" + "net" + "strconv" + "time" + + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + "github.com/osrg/gobgp/pkg/packet/bmp" + log "github.com/sirupsen/logrus" +) + +type ribout map[string][]*table.Path + +func newribout() ribout { + return make(map[string][]*table.Path) +} + +// return true if we need to send the path to the BMP server +func (r ribout) update(p *table.Path) bool { + key := p.GetNlri().String() // TODO expose (*Path).getPrefix() + l := r[key] + if p.IsWithdraw { + if len(l) == 0 { + return false + } + n := make([]*table.Path, 0, len(l)) + for _, q := range l { + if p.GetSource() == q.GetSource() { + continue + } + n = append(n, q) + } + if len(n) == 0 { + delete(r, key) + } else { + r[key] = n + } + return true + } + + if len(l) == 0 { + r[key] = []*table.Path{p} + return true + } + + doAppend := true + for idx, q := range l { + if p.GetSource() == q.GetSource() { + // if we have sent the same path, don't send it again + if p.Equal(q) { + return false + } + l[idx] = p + doAppend = false + } + } + if doAppend { + r[key] = append(r[key], p) + } + return true +} + +func (b *bmpClient) tryConnect() *net.TCPConn { + interval := 1 + for { + log.WithFields(log.Fields{"Topic": "bmp"}).Debugf("Connecting BMP server:%s", b.host) + conn, err := net.Dial("tcp", b.host) + if err != nil { + select { + case <-b.dead: + return nil + default: + } + time.Sleep(time.Duration(interval) * time.Second) + if interval < 30 { + interval *= 2 + } + } else { + log.WithFields(log.Fields{"Topic": "bmp"}).Infof("BMP server is connected:%s", b.host) + return conn.(*net.TCPConn) + } + } +} + +func (b *bmpClient) Stop() { + close(b.dead) +} + +func (b *bmpClient) loop() { + for { + conn := b.tryConnect() + if conn == nil { + break + } + + if func() bool { + ops := []WatchOption{WatchPeerState(true)} + if b.c.RouteMonitoringPolicy == config.BMP_ROUTE_MONITORING_POLICY_TYPE_BOTH { + log.WithFields( + log.Fields{"Topic": "bmp"}, + ).Warn("both option for route-monitoring-policy is obsoleted") + } + if b.c.RouteMonitoringPolicy == config.BMP_ROUTE_MONITORING_POLICY_TYPE_PRE_POLICY || b.c.RouteMonitoringPolicy == config.BMP_ROUTE_MONITORING_POLICY_TYPE_ALL { + ops = append(ops, WatchUpdate(true)) + } + if b.c.RouteMonitoringPolicy == config.BMP_ROUTE_MONITORING_POLICY_TYPE_POST_POLICY || b.c.RouteMonitoringPolicy == config.BMP_ROUTE_MONITORING_POLICY_TYPE_ALL { + ops = append(ops, WatchPostUpdate(true)) + } + if b.c.RouteMonitoringPolicy == config.BMP_ROUTE_MONITORING_POLICY_TYPE_LOCAL_RIB || b.c.RouteMonitoringPolicy == config.BMP_ROUTE_MONITORING_POLICY_TYPE_ALL { + ops = append(ops, WatchBestPath(true)) + } + if b.c.RouteMirroringEnabled { + ops = append(ops, WatchMessage(false)) + } + w := b.s.Watch(ops...) + defer w.Stop() + + var tickerCh <-chan time.Time + if b.c.StatisticsTimeout == 0 { + log.WithFields(log.Fields{"Topic": "bmp"}).Debug("statistics reports disabled") + } else { + t := time.NewTicker(time.Duration(b.c.StatisticsTimeout) * time.Second) + defer t.Stop() + tickerCh = t.C + } + + write := func(msg *bmp.BMPMessage) error { + buf, _ := msg.Serialize() + _, err := conn.Write(buf) + if err != nil { + log.Warnf("failed to write to bmp server %s", b.host) + } + return err + } + + if err := write(bmp.NewBMPInitiation([]bmp.BMPInfoTLVInterface{})); err != nil { + return false + } + + for { + select { + case ev := <-w.Event(): + switch msg := ev.(type) { + case *WatchEventUpdate: + info := &table.PeerInfo{ + Address: msg.PeerAddress, + AS: msg.PeerAS, + ID: msg.PeerID, + } + if msg.Payload == nil { + var pathList []*table.Path + if msg.Init { + pathList = msg.PathList + } else { + for _, p := range msg.PathList { + if b.ribout.update(p) { + pathList = append(pathList, p) + } + } + } + for _, path := range pathList { + for _, u := range table.CreateUpdateMsgFromPaths([]*table.Path{path}) { + payload, _ := u.Serialize() + if err := write(bmpPeerRoute(bmp.BMP_PEER_TYPE_GLOBAL, msg.PostPolicy, 0, true, info, msg.Timestamp.Unix(), payload)); err != nil { + return false + } + } + } + } else { + if err := write(bmpPeerRoute(bmp.BMP_PEER_TYPE_GLOBAL, msg.PostPolicy, 0, msg.FourBytesAs, info, msg.Timestamp.Unix(), msg.Payload)); err != nil { + return false + } + } + case *WatchEventBestPath: + info := &table.PeerInfo{ + Address: net.ParseIP("0.0.0.0").To4(), + AS: b.s.bgpConfig.Global.Config.As, + ID: net.ParseIP(b.s.bgpConfig.Global.Config.RouterId).To4(), + } + for _, p := range msg.PathList { + u := table.CreateUpdateMsgFromPaths([]*table.Path{p})[0] + if payload, err := u.Serialize(); err != nil { + return false + } else if err = write(bmpPeerRoute(bmp.BMP_PEER_TYPE_LOCAL_RIB, false, 0, true, info, p.GetTimestamp().Unix(), payload)); err != nil { + return false + } + } + case *WatchEventPeerState: + if msg.State == bgp.BGP_FSM_ESTABLISHED { + if err := write(bmpPeerUp(msg, bmp.BMP_PEER_TYPE_GLOBAL, false, 0)); err != nil { + return false + } + } else { + if err := write(bmpPeerDown(msg, bmp.BMP_PEER_TYPE_GLOBAL, false, 0)); err != nil { + return false + } + } + case *WatchEventMessage: + info := &table.PeerInfo{ + Address: msg.PeerAddress, + AS: msg.PeerAS, + ID: msg.PeerID, + } + if err := write(bmpPeerRouteMirroring(bmp.BMP_PEER_TYPE_GLOBAL, 0, info, msg.Timestamp.Unix(), msg.Message)); err != nil { + return false + } + } + case <-tickerCh: + neighborList := b.s.GetNeighbor("", true) + for _, n := range neighborList { + if n.State.SessionState != config.SESSION_STATE_ESTABLISHED { + continue + } + if err := write(bmpPeerStats(bmp.BMP_PEER_TYPE_GLOBAL, 0, 0, n)); err != nil { + return false + } + } + case <-b.dead: + term := bmp.NewBMPTermination([]bmp.BMPTermTLVInterface{ + bmp.NewBMPTermTLV16(bmp.BMP_TERM_TLV_TYPE_REASON, bmp.BMP_TERM_REASON_PERMANENTLY_ADMIN), + }) + if err := write(term); err != nil { + return false + } + conn.Close() + return true + } + } + }() { + return + } + } +} + +type bmpClient struct { + s *BgpServer + dead chan struct{} + host string + c *config.BmpServerConfig + ribout ribout +} + +func bmpPeerUp(ev *WatchEventPeerState, t uint8, policy bool, pd uint64) *bmp.BMPMessage { + var flags uint8 = 0 + if policy { + flags |= bmp.BMP_PEER_FLAG_POST_POLICY + } + ph := bmp.NewBMPPeerHeader(t, flags, pd, ev.PeerAddress.String(), ev.PeerAS, ev.PeerID.String(), float64(ev.Timestamp.Unix())) + return bmp.NewBMPPeerUpNotification(*ph, ev.LocalAddress.String(), ev.LocalPort, ev.PeerPort, ev.SentOpen, ev.RecvOpen) +} + +func bmpPeerDown(ev *WatchEventPeerState, t uint8, policy bool, pd uint64) *bmp.BMPMessage { + var flags uint8 = 0 + if policy { + flags |= bmp.BMP_PEER_FLAG_POST_POLICY + } + ph := bmp.NewBMPPeerHeader(t, flags, pd, ev.PeerAddress.String(), ev.PeerAS, ev.PeerID.String(), float64(ev.Timestamp.Unix())) + return bmp.NewBMPPeerDownNotification(*ph, uint8(ev.StateReason.PeerDownReason), ev.StateReason.BGPNotification, ev.StateReason.Data) +} + +func bmpPeerRoute(t uint8, policy bool, pd uint64, fourBytesAs bool, peeri *table.PeerInfo, timestamp int64, payload []byte) *bmp.BMPMessage { + var flags uint8 = 0 + if policy { + flags |= bmp.BMP_PEER_FLAG_POST_POLICY + } + if !fourBytesAs { + flags |= bmp.BMP_PEER_FLAG_TWO_AS + } + ph := bmp.NewBMPPeerHeader(t, flags, pd, peeri.Address.String(), peeri.AS, peeri.ID.String(), float64(timestamp)) + m := bmp.NewBMPRouteMonitoring(*ph, nil) + body := m.Body.(*bmp.BMPRouteMonitoring) + body.BGPUpdatePayload = payload + return m +} + +func bmpPeerStats(peerType uint8, peerDist uint64, timestamp int64, neighConf *config.Neighbor) *bmp.BMPMessage { + var peerFlags uint8 = 0 + ph := bmp.NewBMPPeerHeader(peerType, peerFlags, peerDist, neighConf.State.NeighborAddress, neighConf.State.PeerAs, neighConf.State.RemoteRouterId, float64(timestamp)) + return bmp.NewBMPStatisticsReport( + *ph, + []bmp.BMPStatsTLVInterface{ + bmp.NewBMPStatsTLV64(bmp.BMP_STAT_TYPE_ADJ_RIB_IN, uint64(neighConf.State.AdjTable.Accepted)), + bmp.NewBMPStatsTLV64(bmp.BMP_STAT_TYPE_LOC_RIB, uint64(neighConf.State.AdjTable.Advertised+neighConf.State.AdjTable.Filtered)), + bmp.NewBMPStatsTLV32(bmp.BMP_STAT_TYPE_WITHDRAW_UPDATE, neighConf.State.Messages.Received.WithdrawUpdate), + bmp.NewBMPStatsTLV32(bmp.BMP_STAT_TYPE_WITHDRAW_PREFIX, neighConf.State.Messages.Received.WithdrawPrefix), + }, + ) +} + +func bmpPeerRouteMirroring(peerType uint8, peerDist uint64, peerInfo *table.PeerInfo, timestamp int64, msg *bgp.BGPMessage) *bmp.BMPMessage { + var peerFlags uint8 = 0 + ph := bmp.NewBMPPeerHeader(peerType, peerFlags, peerDist, peerInfo.Address.String(), peerInfo.AS, peerInfo.ID.String(), float64(timestamp)) + return bmp.NewBMPRouteMirroring( + *ph, + []bmp.BMPRouteMirrTLVInterface{ + // RFC7854: BGP Message TLV MUST occur last in the list of TLVs + bmp.NewBMPRouteMirrTLVBGPMsg(bmp.BMP_ROUTE_MIRRORING_TLV_TYPE_BGP_MSG, msg), + }, + ) +} + +func (b *bmpClientManager) addServer(c *config.BmpServerConfig) error { + host := net.JoinHostPort(c.Address, strconv.Itoa(int(c.Port))) + if _, y := b.clientMap[host]; y { + return fmt.Errorf("bmp client %s is already configured", host) + } + b.clientMap[host] = &bmpClient{ + s: b.s, + dead: make(chan struct{}), + host: host, + c: c, + ribout: newribout(), + } + go b.clientMap[host].loop() + return nil +} + +func (b *bmpClientManager) deleteServer(c *config.BmpServerConfig) error { + host := net.JoinHostPort(c.Address, strconv.Itoa(int(c.Port))) + if c, y := b.clientMap[host]; !y { + return fmt.Errorf("bmp client %s isn't found", host) + } else { + c.Stop() + delete(b.clientMap, host) + } + return nil +} + +type bmpClientManager struct { + s *BgpServer + clientMap map[string]*bmpClient +} + +func newBmpClientManager(s *BgpServer) *bmpClientManager { + return &bmpClientManager{ + s: s, + clientMap: make(map[string]*bmpClient), + } +} diff --git a/pkg/server/collector.go b/pkg/server/collector.go new file mode 100644 index 00000000..c219b215 --- /dev/null +++ b/pkg/server/collector.go @@ -0,0 +1,222 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "fmt" + "time" + + "github.com/influxdata/influxdb/client/v2" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + log "github.com/sirupsen/logrus" +) + +type Collector struct { + s *BgpServer + url string + dbName string + interval uint64 + client client.Client +} + +const ( + MEATUREMENT_UPDATE = "update" + MEATUREMENT_PEER = "peer" + MEATUREMENT_TABLE = "table" +) + +func (c *Collector) writePoints(points []*client.Point) error { + bp, _ := client.NewBatchPoints(client.BatchPointsConfig{ + Database: c.dbName, + Precision: "ms", + }) + bp.AddPoints(points) + return c.client.Write(bp) +} + +func (c *Collector) writePeer(msg *WatchEventPeerState) error { + var state string + switch msg.State { + case bgp.BGP_FSM_ESTABLISHED: + state = "Established" + case bgp.BGP_FSM_IDLE: + state = "Idle" + default: + return fmt.Errorf("unexpected fsm state %v", msg.State) + } + + tags := map[string]string{ + "PeerAddress": msg.PeerAddress.String(), + "PeerAS": fmt.Sprintf("%v", msg.PeerAS), + "State": state, + } + + fields := map[string]interface{}{ + "PeerID": msg.PeerID.String(), + } + + pt, err := client.NewPoint(MEATUREMENT_PEER, tags, fields, msg.Timestamp) + if err != nil { + return err + } + return c.writePoints([]*client.Point{pt}) +} + +func path2data(path *table.Path) (map[string]interface{}, map[string]string) { + fields := map[string]interface{}{ + "RouterID": path.GetSource().ID, + } + if asPath := path.GetAsPath(); asPath != nil { + fields["ASPath"] = asPath.String() + } + if origin, err := path.GetOrigin(); err == nil { + typ := "-" + switch origin { + case bgp.BGP_ORIGIN_ATTR_TYPE_IGP: + typ = "i" + case bgp.BGP_ORIGIN_ATTR_TYPE_EGP: + typ = "e" + case bgp.BGP_ORIGIN_ATTR_TYPE_INCOMPLETE: + typ = "?" + } + fields["Origin"] = typ + } + if med, err := path.GetMed(); err == nil { + fields["Med"] = med + } + + tags := map[string]string{ + "PeerAddress": path.GetSource().Address.String(), + "PeerAS": fmt.Sprintf("%v", path.GetSource().AS), + "Timestamp": path.GetTimestamp().String(), + } + if nexthop := path.GetNexthop(); len(nexthop) > 0 { + fields["NextHop"] = nexthop.String() + } + if originAS := path.GetSourceAs(); originAS != 0 { + fields["OriginAS"] = fmt.Sprintf("%v", originAS) + } + + if err := bgp.FlatUpdate(tags, path.GetNlri().Flat()); err != nil { + log.WithFields(log.Fields{"Type": "collector", "Error": err}).Error("NLRI FlatUpdate failed") + } + for _, p := range path.GetPathAttrs() { + if err := bgp.FlatUpdate(tags, p.Flat()); err != nil { + log.WithFields(log.Fields{"Type": "collector", "Error": err}).Error("PathAttr FlatUpdate failed") + } + } + return fields, tags +} + +func (c *Collector) writeUpdate(msg *WatchEventUpdate) error { + if len(msg.PathList) == 0 { + // EOR + return nil + } + now := time.Now() + points := make([]*client.Point, 0, len(msg.PathList)) + for _, path := range msg.PathList { + fields, tags := path2data(path) + tags["Withdraw"] = fmt.Sprintf("%v", path.IsWithdraw) + pt, err := client.NewPoint(MEATUREMENT_UPDATE, tags, fields, now) + if err != nil { + return fmt.Errorf("failed to write update, %v", err) + } + points = append(points, pt) + } + return c.writePoints(points) +} + +func (c *Collector) writeTable(msg *WatchEventAdjIn) error { + now := time.Now() + points := make([]*client.Point, 0, len(msg.PathList)) + for _, path := range msg.PathList { + fields, tags := path2data(path) + pt, err := client.NewPoint(MEATUREMENT_TABLE, tags, fields, now) + if err != nil { + return fmt.Errorf("failed to write table, %v", err) + } + points = append(points, pt) + } + return c.writePoints(points) +} + +func (c *Collector) loop() { + w := c.s.Watch(WatchPeerState(true), WatchUpdate(false)) + defer w.Stop() + + ticker := func() *time.Ticker { + if c.interval == 0 { + return &time.Ticker{} + } + return time.NewTicker(time.Second * time.Duration(c.interval)) + }() + + for { + select { + case <-ticker.C: + w.Generate(WATCH_EVENT_TYPE_PRE_UPDATE) + case ev := <-w.Event(): + switch msg := ev.(type) { + case *WatchEventUpdate: + if err := c.writeUpdate(msg); err != nil { + log.WithFields(log.Fields{"Type": "collector", "Error": err}).Error("Failed to write update event message") + } + case *WatchEventPeerState: + if err := c.writePeer(msg); err != nil { + log.WithFields(log.Fields{"Type": "collector", "Error": err}).Error("Failed to write state changed event message") + } + case *WatchEventAdjIn: + if err := c.writeTable(msg); err != nil { + log.WithFields(log.Fields{"Type": "collector", "Error": err}).Error("Failed to write Adj-In event message") + } + } + } + } +} + +func NewCollector(s *BgpServer, url, dbName string, interval uint64) (*Collector, error) { + c, err := client.NewHTTPClient(client.HTTPConfig{ + Addr: url, + }) + if err != nil { + return nil, err + } + + _, _, err = c.Ping(0) + if err != nil { + log.Error("can not connect to InfluxDB") + log.WithFields(log.Fields{"Type": "collector", "Error": err}).Error("Failed to connect to InfluxDB") + return nil, err + } + + q := client.NewQuery("CREATE DATABASE "+dbName, "", "") + if response, err := c.Query(q); err != nil || response.Error() != nil { + log.WithFields(log.Fields{"Type": "collector", "Error": err}).Errorf("Failed to create database:%s", dbName) + return nil, err + } + + collector := &Collector{ + s: s, + url: url, + dbName: dbName, + interval: interval, + client: c, + } + go collector.loop() + return collector, nil +} diff --git a/pkg/server/fsm.go b/pkg/server/fsm.go new file mode 100644 index 00000000..855066f4 --- /dev/null +++ b/pkg/server/fsm.go @@ -0,0 +1,1752 @@ +// Copyright (C) 2014 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "fmt" + "io" + "math/rand" + "net" + "strconv" + "time" + + "github.com/eapache/channels" + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + "github.com/osrg/gobgp/pkg/packet/bmp" + + log "github.com/sirupsen/logrus" + "gopkg.in/tomb.v2" +) + +type PeerDownReason int + +const ( + PEER_DOWN_REASON_UNKNOWN PeerDownReason = iota + PEER_DOWN_BY_LOCAL + PEER_DOWN_BY_LOCAL_WITHOUT_NOTIFICATION + PEER_DOWN_BY_REMOTE + PEER_DOWN_BY_REMOTE_WITHOUT_NOTIFICATION + PEER_DOWN_BY_BMP_CONFIGURATION +) + +type FsmStateReasonType uint8 + +const ( + FSM_DYING FsmStateReasonType = iota + FSM_ADMIN_DOWN + FSM_READ_FAILED + FSM_WRITE_FAILED + FSM_NOTIFICATION_SENT + FSM_NOTIFICATION_RECV + FSM_HOLD_TIMER_EXPIRED + FSM_IDLE_HOLD_TIMER_EXPIRED + FSM_RESTART_TIMER_EXPIRED + FSM_GRACEFUL_RESTART + FSM_INVALID_MSG + FSM_NEW_CONNECTION + FSM_OPEN_MSG_RECEIVED + FSM_OPEN_MSG_NEGOTIATED + FSM_HARD_RESET +) + +type FsmStateReason struct { + Type FsmStateReasonType + PeerDownReason PeerDownReason + BGPNotification *bgp.BGPMessage + Data []byte +} + +func NewFsmStateReason(typ FsmStateReasonType, notif *bgp.BGPMessage, data []byte) *FsmStateReason { + var reasonCode PeerDownReason + switch typ { + case FSM_DYING, FSM_INVALID_MSG, FSM_NOTIFICATION_SENT, FSM_HOLD_TIMER_EXPIRED, FSM_IDLE_HOLD_TIMER_EXPIRED, FSM_RESTART_TIMER_EXPIRED: + reasonCode = PEER_DOWN_BY_LOCAL + case FSM_ADMIN_DOWN: + reasonCode = PEER_DOWN_BY_LOCAL_WITHOUT_NOTIFICATION + case FSM_NOTIFICATION_RECV, FSM_GRACEFUL_RESTART, FSM_HARD_RESET: + reasonCode = PEER_DOWN_BY_REMOTE + case FSM_READ_FAILED, FSM_WRITE_FAILED: + reasonCode = PEER_DOWN_BY_REMOTE_WITHOUT_NOTIFICATION + } + return &FsmStateReason{ + Type: typ, + PeerDownReason: reasonCode, + BGPNotification: notif, + Data: data, + } +} + +func (r FsmStateReason) String() string { + switch r.Type { + case FSM_DYING: + return "dying" + case FSM_ADMIN_DOWN: + return "admin-down" + case FSM_READ_FAILED: + return "read-failed" + case FSM_WRITE_FAILED: + return "write-failed" + case FSM_NOTIFICATION_SENT: + body := r.BGPNotification.Body.(*bgp.BGPNotification) + return fmt.Sprintf("notification-sent %s", bgp.NewNotificationErrorCode(body.ErrorCode, body.ErrorSubcode).String()) + case FSM_NOTIFICATION_RECV: + body := r.BGPNotification.Body.(*bgp.BGPNotification) + return fmt.Sprintf("notification-received %s", bgp.NewNotificationErrorCode(body.ErrorCode, body.ErrorSubcode).String()) + case FSM_HOLD_TIMER_EXPIRED: + return "hold-timer-expired" + case FSM_IDLE_HOLD_TIMER_EXPIRED: + return "idle-hold-timer-expired" + case FSM_RESTART_TIMER_EXPIRED: + return "restart-timer-expired" + case FSM_GRACEFUL_RESTART: + return "graceful-restart" + case FSM_INVALID_MSG: + return "invalid-msg" + case FSM_NEW_CONNECTION: + return "new-connection" + case FSM_OPEN_MSG_RECEIVED: + return "open-msg-received" + case FSM_OPEN_MSG_NEGOTIATED: + return "open-msg-negotiated" + case FSM_HARD_RESET: + return "hard-reset" + default: + return "unknown" + } +} + +type FsmMsgType int + +const ( + _ FsmMsgType = iota + FSM_MSG_STATE_CHANGE + FSM_MSG_BGP_MESSAGE + FSM_MSG_ROUTE_REFRESH +) + +type FsmMsg struct { + MsgType FsmMsgType + MsgSrc string + MsgData interface{} + StateReason *FsmStateReason + PathList []*table.Path + timestamp time.Time + payload []byte + Version uint +} + +type FsmOutgoingMsg struct { + Paths []*table.Path + Notification *bgp.BGPMessage + StayIdle bool +} + +const ( + HOLDTIME_OPENSENT = 240 + HOLDTIME_IDLE = 5 +) + +type AdminState int + +const ( + ADMIN_STATE_UP AdminState = iota + ADMIN_STATE_DOWN + ADMIN_STATE_PFX_CT +) + +func (s AdminState) String() string { + switch s { + case ADMIN_STATE_UP: + return "ADMIN_STATE_UP" + case ADMIN_STATE_DOWN: + return "ADMIN_STATE_DOWN" + case ADMIN_STATE_PFX_CT: + return "ADMIN_STATE_PFX_CT" + default: + return "Unknown" + } +} + +type AdminStateOperation struct { + State AdminState + Communication []byte +} + +var fsmVersion uint + +type FSM struct { + t tomb.Tomb + gConf *config.Global + pConf *config.Neighbor + state bgp.FSMState + reason *FsmStateReason + conn net.Conn + connCh chan net.Conn + idleHoldTime float64 + opensentHoldTime float64 + adminState AdminState + adminStateCh chan AdminStateOperation + getActiveCh chan struct{} + h *FSMHandler + rfMap map[bgp.RouteFamily]bgp.BGPAddPathMode + capMap map[bgp.BGPCapabilityCode][]bgp.ParameterCapabilityInterface + recvOpen *bgp.BGPMessage + peerInfo *table.PeerInfo + policy *table.RoutingPolicy + gracefulRestartTimer *time.Timer + twoByteAsTrans bool + version uint + marshallingOptions *bgp.MarshallingOption +} + +func (fsm *FSM) bgpMessageStateUpdate(MessageType uint8, isIn bool) { + state := &fsm.pConf.State.Messages + timer := &fsm.pConf.Timers + if isIn { + state.Received.Total++ + } else { + state.Sent.Total++ + } + switch MessageType { + case bgp.BGP_MSG_OPEN: + if isIn { + state.Received.Open++ + } else { + state.Sent.Open++ + } + case bgp.BGP_MSG_UPDATE: + if isIn { + state.Received.Update++ + timer.State.UpdateRecvTime = time.Now().Unix() + } else { + state.Sent.Update++ + } + case bgp.BGP_MSG_NOTIFICATION: + if isIn { + state.Received.Notification++ + } else { + state.Sent.Notification++ + } + case bgp.BGP_MSG_KEEPALIVE: + if isIn { + state.Received.Keepalive++ + } else { + state.Sent.Keepalive++ + } + case bgp.BGP_MSG_ROUTE_REFRESH: + if isIn { + state.Received.Refresh++ + } else { + state.Sent.Refresh++ + } + default: + if isIn { + state.Received.Discarded++ + } else { + state.Sent.Discarded++ + } + } +} + +func (fsm *FSM) bmpStatsUpdate(statType uint16, increment int) { + stats := &fsm.pConf.State.Messages.Received + switch statType { + // TODO + // Support other stat types. + case bmp.BMP_STAT_TYPE_WITHDRAW_UPDATE: + stats.WithdrawUpdate += uint32(increment) + case bmp.BMP_STAT_TYPE_WITHDRAW_PREFIX: + stats.WithdrawPrefix += uint32(increment) + } +} + +func NewFSM(gConf *config.Global, pConf *config.Neighbor, policy *table.RoutingPolicy) *FSM { + adminState := ADMIN_STATE_UP + if pConf.Config.AdminDown { + adminState = ADMIN_STATE_DOWN + } + pConf.State.SessionState = config.IntToSessionStateMap[int(bgp.BGP_FSM_IDLE)] + pConf.Timers.State.Downtime = time.Now().Unix() + fsmVersion++ + fsm := &FSM{ + gConf: gConf, + pConf: pConf, + state: bgp.BGP_FSM_IDLE, + connCh: make(chan net.Conn, 1), + opensentHoldTime: float64(HOLDTIME_OPENSENT), + adminState: adminState, + adminStateCh: make(chan AdminStateOperation, 1), + getActiveCh: make(chan struct{}), + rfMap: make(map[bgp.RouteFamily]bgp.BGPAddPathMode), + capMap: make(map[bgp.BGPCapabilityCode][]bgp.ParameterCapabilityInterface), + peerInfo: table.NewPeerInfo(gConf, pConf), + policy: policy, + gracefulRestartTimer: time.NewTimer(time.Hour), + version: fsmVersion, + } + fsm.gracefulRestartTimer.Stop() + fsm.t.Go(fsm.connectLoop) + return fsm +} + +func (fsm *FSM) StateChange(nextState bgp.FSMState) { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "old": fsm.state.String(), + "new": nextState.String(), + "reason": fsm.reason, + }).Debug("state changed") + fsm.state = nextState + switch nextState { + case bgp.BGP_FSM_ESTABLISHED: + fsm.pConf.Timers.State.Uptime = time.Now().Unix() + fsm.pConf.State.EstablishedCount++ + // reset the state set by the previous session + fsm.twoByteAsTrans = false + if _, y := fsm.capMap[bgp.BGP_CAP_FOUR_OCTET_AS_NUMBER]; !y { + fsm.twoByteAsTrans = true + break + } + y := func() bool { + for _, c := range capabilitiesFromConfig(fsm.pConf) { + switch c.(type) { + case *bgp.CapFourOctetASNumber: + return true + } + } + return false + }() + if !y { + fsm.twoByteAsTrans = true + } + case bgp.BGP_FSM_ACTIVE: + if !fsm.pConf.Transport.Config.PassiveMode { + fsm.getActiveCh <- struct{}{} + } + fallthrough + default: + fsm.pConf.Timers.State.Downtime = time.Now().Unix() + } +} + +func hostport(addr net.Addr) (string, uint16) { + if addr != nil { + host, port, err := net.SplitHostPort(addr.String()) + if err != nil { + return "", 0 + } + p, _ := strconv.ParseUint(port, 10, 16) + return host, uint16(p) + } + return "", 0 +} + +func (fsm *FSM) RemoteHostPort() (string, uint16) { + return hostport(fsm.conn.RemoteAddr()) + +} + +func (fsm *FSM) LocalHostPort() (string, uint16) { + return hostport(fsm.conn.LocalAddr()) +} + +func (fsm *FSM) sendNotificationFromErrorMsg(e *bgp.MessageError) (*bgp.BGPMessage, error) { + if fsm.h != nil && fsm.h.conn != nil { + m := bgp.NewBGPNotificationMessage(e.TypeCode, e.SubTypeCode, e.Data) + b, _ := m.Serialize() + _, err := fsm.h.conn.Write(b) + if err == nil { + fsm.bgpMessageStateUpdate(m.Header.Type, false) + fsm.h.sentNotification = m + } + fsm.h.conn.Close() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "Data": e, + }).Warn("sent notification") + return m, nil + } + return nil, fmt.Errorf("can't send notification to %s since TCP connection is not established", fsm.pConf.State.NeighborAddress) +} + +func (fsm *FSM) sendNotification(code, subType uint8, data []byte, msg string) (*bgp.BGPMessage, error) { + e := bgp.NewMessageError(code, subType, data, msg) + return fsm.sendNotificationFromErrorMsg(e.(*bgp.MessageError)) +} + +func (fsm *FSM) connectLoop() error { + tick := int(fsm.pConf.Timers.Config.ConnectRetry) + if tick < MIN_CONNECT_RETRY { + tick = MIN_CONNECT_RETRY + } + + r := rand.New(rand.NewSource(time.Now().UnixNano())) + + timer := time.NewTimer(time.Duration(tick) * time.Second) + timer.Stop() + + connect := func() { + addr := fsm.pConf.State.NeighborAddress + port := int(bgp.BGP_PORT) + if fsm.pConf.Transport.Config.RemotePort != 0 { + port = int(fsm.pConf.Transport.Config.RemotePort) + } + laddr, err := net.ResolveTCPAddr("tcp", net.JoinHostPort(fsm.pConf.Transport.Config.LocalAddress, "0")) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + }).Warn("failed to resolve local address: %s", err) + return + } + var conn net.Conn + d := TCPDialer{ + Dialer: net.Dialer{ + LocalAddr: laddr, + Timeout: time.Duration(MIN_CONNECT_RETRY-1) * time.Second, + }, + AuthPassword: fsm.pConf.Config.AuthPassword, + } + if fsm.pConf.TtlSecurity.Config.Enabled { + d.Ttl = 255 + d.TtlMin = fsm.pConf.TtlSecurity.Config.TtlMin + } else if fsm.pConf.Config.PeerAs != 0 && fsm.pConf.Config.PeerType == config.PEER_TYPE_EXTERNAL { + d.Ttl = 1 + if fsm.pConf.EbgpMultihop.Config.Enabled { + d.Ttl = fsm.pConf.EbgpMultihop.Config.MultihopTtl + } + } + conn, err = d.DialTCP(addr, port) + if err == nil { + select { + case fsm.connCh <- conn: + return + default: + conn.Close() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + }).Warn("active conn is closed to avoid being blocked") + } + } else { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + }).Debugf("failed to connect: %s", err) + } + + if fsm.state == bgp.BGP_FSM_ACTIVE && !fsm.pConf.GracefulRestart.State.PeerRestarting { + timer.Reset(time.Duration(tick) * time.Second) + } + } + + for { + select { + case <-fsm.t.Dying(): + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + }).Debug("stop connect loop") + return nil + case <-timer.C: + if fsm.state == bgp.BGP_FSM_ACTIVE && !fsm.pConf.GracefulRestart.State.PeerRestarting { + go connect() + } + case <-fsm.getActiveCh: + timer.Reset(time.Duration(r.Intn(MIN_CONNECT_RETRY)+MIN_CONNECT_RETRY) * time.Second) + } + } +} + +type FSMHandler struct { + t tomb.Tomb + fsm *FSM + conn net.Conn + msgCh *channels.InfiniteChannel + stateReasonCh chan FsmStateReason + incoming *channels.InfiniteChannel + stateCh chan *FsmMsg + outgoing *channels.InfiniteChannel + holdTimerResetCh chan bool + sentNotification *bgp.BGPMessage +} + +func NewFSMHandler(fsm *FSM, incoming *channels.InfiniteChannel, stateCh chan *FsmMsg, outgoing *channels.InfiniteChannel) *FSMHandler { + h := &FSMHandler{ + fsm: fsm, + stateReasonCh: make(chan FsmStateReason, 2), + incoming: incoming, + stateCh: stateCh, + outgoing: outgoing, + holdTimerResetCh: make(chan bool, 2), + } + fsm.t.Go(h.loop) + return h +} + +func (h *FSMHandler) idle() (bgp.FSMState, *FsmStateReason) { + fsm := h.fsm + + idleHoldTimer := time.NewTimer(time.Second * time.Duration(fsm.idleHoldTime)) + for { + select { + case <-h.t.Dying(): + return -1, NewFsmStateReason(FSM_DYING, nil, nil) + case <-fsm.gracefulRestartTimer.C: + if fsm.pConf.GracefulRestart.State.PeerRestarting { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("graceful restart timer expired") + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_RESTART_TIMER_EXPIRED, nil, nil) + } + case conn, ok := <-fsm.connCh: + if !ok { + break + } + conn.Close() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("Closed an accepted connection") + case <-idleHoldTimer.C: + + if fsm.adminState == ADMIN_STATE_UP { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "Duration": fsm.idleHoldTime, + }).Debug("IdleHoldTimer expired") + fsm.idleHoldTime = HOLDTIME_IDLE + return bgp.BGP_FSM_ACTIVE, NewFsmStateReason(FSM_IDLE_HOLD_TIMER_EXPIRED, nil, nil) + + } else { + log.WithFields(log.Fields{"Topic": "Peer"}).Debug("IdleHoldTimer expired, but stay at idle because the admin state is DOWN") + } + + case stateOp := <-fsm.adminStateCh: + err := h.changeAdminState(stateOp.State) + if err == nil { + switch stateOp.State { + case ADMIN_STATE_DOWN: + // stop idle hold timer + idleHoldTimer.Stop() + + case ADMIN_STATE_UP: + // restart idle hold timer + idleHoldTimer.Reset(time.Second * time.Duration(fsm.idleHoldTime)) + } + } + } + } +} + +func (h *FSMHandler) active() (bgp.FSMState, *FsmStateReason) { + fsm := h.fsm + for { + select { + case <-h.t.Dying(): + return -1, NewFsmStateReason(FSM_DYING, nil, nil) + case conn, ok := <-fsm.connCh: + if !ok { + break + } + fsm.conn = conn + ttl := 0 + ttlMin := 0 + if fsm.pConf.TtlSecurity.Config.Enabled { + ttl = 255 + ttlMin = int(fsm.pConf.TtlSecurity.Config.TtlMin) + } else if fsm.pConf.Config.PeerAs != 0 && fsm.pConf.Config.PeerType == config.PEER_TYPE_EXTERNAL { + if fsm.pConf.EbgpMultihop.Config.Enabled { + ttl = int(fsm.pConf.EbgpMultihop.Config.MultihopTtl) + } else if fsm.pConf.Transport.Config.Ttl != 0 { + ttl = int(fsm.pConf.Transport.Config.Ttl) + } else { + ttl = 1 + } + } else if fsm.pConf.Transport.Config.Ttl != 0 { + ttl = int(fsm.pConf.Transport.Config.Ttl) + } + if ttl != 0 { + if err := SetTcpTTLSockopt(conn.(*net.TCPConn), ttl); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.Config.NeighborAddress, + "State": fsm.state.String(), + }).Warnf("cannot set TTL(=%d) for peer: %s", ttl, err) + } + } + if ttlMin != 0 { + if err := SetTcpMinTTLSockopt(conn.(*net.TCPConn), ttlMin); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.Config.NeighborAddress, + "State": fsm.state.String(), + }).Warnf("cannot set minimal TTL(=%d) for peer: %s", ttl, err) + } + } + // we don't implement delayed open timer so move to opensent right + // away. + return bgp.BGP_FSM_OPENSENT, NewFsmStateReason(FSM_NEW_CONNECTION, nil, nil) + case <-fsm.gracefulRestartTimer.C: + if fsm.pConf.GracefulRestart.State.PeerRestarting { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("graceful restart timer expired") + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_RESTART_TIMER_EXPIRED, nil, nil) + } + case err := <-h.stateReasonCh: + return bgp.BGP_FSM_IDLE, &err + case stateOp := <-fsm.adminStateCh: + err := h.changeAdminState(stateOp.State) + if err == nil { + switch stateOp.State { + case ADMIN_STATE_DOWN: + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_ADMIN_DOWN, nil, nil) + case ADMIN_STATE_UP: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "AdminState": stateOp.State.String(), + }).Panic("code logic bug") + } + } + } + } +} + +func capAddPathFromConfig(pConf *config.Neighbor) bgp.ParameterCapabilityInterface { + tuples := make([]*bgp.CapAddPathTuple, 0, len(pConf.AfiSafis)) + for _, af := range pConf.AfiSafis { + var mode bgp.BGPAddPathMode + if af.AddPaths.State.Receive { + mode |= bgp.BGP_ADD_PATH_RECEIVE + } + if af.AddPaths.State.SendMax > 0 { + mode |= bgp.BGP_ADD_PATH_SEND + } + if mode > 0 { + tuples = append(tuples, bgp.NewCapAddPathTuple(af.State.Family, mode)) + } + } + if len(tuples) == 0 { + return nil + } + return bgp.NewCapAddPath(tuples) +} + +func capabilitiesFromConfig(pConf *config.Neighbor) []bgp.ParameterCapabilityInterface { + caps := make([]bgp.ParameterCapabilityInterface, 0, 4) + caps = append(caps, bgp.NewCapRouteRefresh()) + for _, af := range pConf.AfiSafis { + caps = append(caps, bgp.NewCapMultiProtocol(af.State.Family)) + } + caps = append(caps, bgp.NewCapFourOctetASNumber(pConf.Config.LocalAs)) + + if c := pConf.GracefulRestart.Config; c.Enabled { + tuples := []*bgp.CapGracefulRestartTuple{} + ltuples := []*bgp.CapLongLivedGracefulRestartTuple{} + + // RFC 4724 4.1 + // To re-establish the session with its peer, the Restarting Speaker + // MUST set the "Restart State" bit in the Graceful Restart Capability + // of the OPEN message. + restarting := pConf.GracefulRestart.State.LocalRestarting + + if !c.HelperOnly { + for i, rf := range pConf.AfiSafis { + if m := rf.MpGracefulRestart.Config; m.Enabled { + // When restarting, always flag forwaring bit. + // This can be a lie, depending on how gobgpd is used. + // For a route-server use-case, since a route-server + // itself doesn't forward packets, and the dataplane + // is a l2 switch which continues to work with no + // relation to bgpd, this behavior is ok. + // TODO consideration of other use-cases + tuples = append(tuples, bgp.NewCapGracefulRestartTuple(rf.State.Family, restarting)) + pConf.AfiSafis[i].MpGracefulRestart.State.Advertised = true + } + if m := rf.LongLivedGracefulRestart.Config; m.Enabled { + ltuples = append(ltuples, bgp.NewCapLongLivedGracefulRestartTuple(rf.State.Family, restarting, m.RestartTime)) + } + } + } + restartTime := c.RestartTime + notification := c.NotificationEnabled + caps = append(caps, bgp.NewCapGracefulRestart(restarting, notification, restartTime, tuples)) + if c.LongLivedEnabled { + caps = append(caps, bgp.NewCapLongLivedGracefulRestart(ltuples)) + } + } + + // unnumbered BGP + if pConf.Config.NeighborInterface != "" { + tuples := []*bgp.CapExtendedNexthopTuple{} + families, _ := config.AfiSafis(pConf.AfiSafis).ToRfList() + for _, family := range families { + if family == bgp.RF_IPv6_UC { + continue + } + tuple := bgp.NewCapExtendedNexthopTuple(family, bgp.AFI_IP6) + tuples = append(tuples, tuple) + } + caps = append(caps, bgp.NewCapExtendedNexthop(tuples)) + } + + // ADD-PATH Capability + if c := capAddPathFromConfig(pConf); c != nil { + caps = append(caps, capAddPathFromConfig(pConf)) + } + + return caps +} + +func buildopen(gConf *config.Global, pConf *config.Neighbor) *bgp.BGPMessage { + caps := capabilitiesFromConfig(pConf) + opt := bgp.NewOptionParameterCapability(caps) + holdTime := uint16(pConf.Timers.Config.HoldTime) + as := pConf.Config.LocalAs + if as > (1<<16)-1 { + as = bgp.AS_TRANS + } + return bgp.NewBGPOpenMessage(uint16(as), holdTime, gConf.Config.RouterId, + []bgp.OptionParameterInterface{opt}) +} + +func readAll(conn net.Conn, length int) ([]byte, error) { + buf := make([]byte, length) + _, err := io.ReadFull(conn, buf) + if err != nil { + return nil, err + } + return buf, nil +} + +func getPathAttrFromBGPUpdate(m *bgp.BGPUpdate, typ bgp.BGPAttrType) bgp.PathAttributeInterface { + for _, a := range m.PathAttributes { + if a.GetType() == typ { + return a + } + } + return nil +} + +func hasOwnASLoop(ownAS uint32, limit int, asPath *bgp.PathAttributeAsPath) bool { + cnt := 0 + for _, param := range asPath.Value { + for _, as := range param.GetAS() { + if as == ownAS { + cnt++ + if cnt > limit { + return true + } + } + } + } + return false +} + +func extractRouteFamily(p *bgp.PathAttributeInterface) *bgp.RouteFamily { + attr := *p + + var afi uint16 + var safi uint8 + + switch a := attr.(type) { + case *bgp.PathAttributeMpReachNLRI: + afi = a.AFI + safi = a.SAFI + case *bgp.PathAttributeMpUnreachNLRI: + afi = a.AFI + safi = a.SAFI + default: + return nil + } + + rf := bgp.AfiSafiToRouteFamily(afi, safi) + return &rf +} + +func (h *FSMHandler) afiSafiDisable(rf bgp.RouteFamily) string { + n := bgp.AddressFamilyNameMap[rf] + + for i, a := range h.fsm.pConf.AfiSafis { + if string(a.Config.AfiSafiName) == n { + h.fsm.pConf.AfiSafis[i].State.Enabled = false + break + } + } + newList := make([]bgp.ParameterCapabilityInterface, 0) + for _, c := range h.fsm.capMap[bgp.BGP_CAP_MULTIPROTOCOL] { + if c.(*bgp.CapMultiProtocol).CapValue == rf { + continue + } + newList = append(newList, c) + } + h.fsm.capMap[bgp.BGP_CAP_MULTIPROTOCOL] = newList + return n +} + +func (h *FSMHandler) handlingError(m *bgp.BGPMessage, e error, useRevisedError bool) bgp.ErrorHandling { + handling := bgp.ERROR_HANDLING_NONE + if m.Header.Type == bgp.BGP_MSG_UPDATE && useRevisedError { + factor := e.(*bgp.MessageError) + handling = factor.ErrorHandling + switch handling { + case bgp.ERROR_HANDLING_ATTRIBUTE_DISCARD: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "State": h.fsm.state.String(), + "error": e, + }).Warn("Some attributes were discarded") + case bgp.ERROR_HANDLING_TREAT_AS_WITHDRAW: + m.Body = bgp.TreatAsWithdraw(m.Body.(*bgp.BGPUpdate)) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "State": h.fsm.state.String(), + "error": e, + }).Warn("the received Update message was treated as withdraw") + case bgp.ERROR_HANDLING_AFISAFI_DISABLE: + rf := extractRouteFamily(factor.ErrorAttribute) + if rf == nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "State": h.fsm.state.String(), + }).Warn("Error occurred during AFI/SAFI disabling") + } else { + n := h.afiSafiDisable(*rf) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "State": h.fsm.state.String(), + "error": e, + }).Warnf("Capability %s was disabled", n) + } + } + } else { + handling = bgp.ERROR_HANDLING_SESSION_RESET + } + return handling +} + +func (h *FSMHandler) recvMessageWithError() (*FsmMsg, error) { + sendToStateReasonCh := func(typ FsmStateReasonType, notif *bgp.BGPMessage) { + // probably doesn't happen but be cautious + select { + case h.stateReasonCh <- *NewFsmStateReason(typ, notif, nil): + default: + } + } + + headerBuf, err := readAll(h.conn, bgp.BGP_HEADER_LENGTH) + if err != nil { + sendToStateReasonCh(FSM_READ_FAILED, nil) + return nil, err + } + + hd := &bgp.BGPHeader{} + err = hd.DecodeFromBytes(headerBuf) + if err != nil { + h.fsm.bgpMessageStateUpdate(0, true) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "State": h.fsm.state.String(), + "error": err, + }).Warn("Session will be reset due to malformed BGP Header") + fmsg := &FsmMsg{ + MsgType: FSM_MSG_BGP_MESSAGE, + MsgSrc: h.fsm.pConf.State.NeighborAddress, + MsgData: err, + Version: h.fsm.version, + } + return fmsg, err + } + + bodyBuf, err := readAll(h.conn, int(hd.Len)-bgp.BGP_HEADER_LENGTH) + if err != nil { + sendToStateReasonCh(FSM_READ_FAILED, nil) + return nil, err + } + + now := time.Now() + useRevisedError := h.fsm.pConf.ErrorHandling.Config.TreatAsWithdraw + handling := bgp.ERROR_HANDLING_NONE + + m, err := bgp.ParseBGPBody(hd, bodyBuf, h.fsm.marshallingOptions) + if err != nil { + handling = h.handlingError(m, err, useRevisedError) + h.fsm.bgpMessageStateUpdate(0, true) + } else { + h.fsm.bgpMessageStateUpdate(m.Header.Type, true) + err = bgp.ValidateBGPMessage(m) + } + fmsg := &FsmMsg{ + MsgType: FSM_MSG_BGP_MESSAGE, + MsgSrc: h.fsm.pConf.State.NeighborAddress, + timestamp: now, + Version: h.fsm.version, + } + + switch handling { + case bgp.ERROR_HANDLING_AFISAFI_DISABLE: + fmsg.MsgData = m + return fmsg, nil + case bgp.ERROR_HANDLING_SESSION_RESET: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "State": h.fsm.state.String(), + "error": err, + }).Warn("Session will be reset due to malformed BGP message") + fmsg.MsgData = err + return fmsg, err + default: + fmsg.MsgData = m + if h.fsm.state == bgp.BGP_FSM_ESTABLISHED { + switch m.Header.Type { + case bgp.BGP_MSG_ROUTE_REFRESH: + fmsg.MsgType = FSM_MSG_ROUTE_REFRESH + case bgp.BGP_MSG_UPDATE: + body := m.Body.(*bgp.BGPUpdate) + isEBGP := h.fsm.pConf.IsEBGPPeer(h.fsm.gConf) + isConfed := h.fsm.pConf.IsConfederationMember(h.fsm.gConf) + + fmsg.payload = make([]byte, len(headerBuf)+len(bodyBuf)) + copy(fmsg.payload, headerBuf) + copy(fmsg.payload[len(headerBuf):], bodyBuf) + + ok, err := bgp.ValidateUpdateMsg(body, h.fsm.rfMap, isEBGP, isConfed) + if !ok { + handling = h.handlingError(m, err, useRevisedError) + } + if handling == bgp.ERROR_HANDLING_SESSION_RESET { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "State": h.fsm.state.String(), + "error": err, + }).Warn("Session will be reset due to malformed BGP update message") + fmsg.MsgData = err + return fmsg, err + } + + if routes := len(body.WithdrawnRoutes); routes > 0 { + h.fsm.bmpStatsUpdate(bmp.BMP_STAT_TYPE_WITHDRAW_UPDATE, 1) + h.fsm.bmpStatsUpdate(bmp.BMP_STAT_TYPE_WITHDRAW_PREFIX, routes) + } else if attr := getPathAttrFromBGPUpdate(body, bgp.BGP_ATTR_TYPE_MP_UNREACH_NLRI); attr != nil { + mpUnreach := attr.(*bgp.PathAttributeMpUnreachNLRI) + if routes = len(mpUnreach.Value); routes > 0 { + h.fsm.bmpStatsUpdate(bmp.BMP_STAT_TYPE_WITHDRAW_UPDATE, 1) + h.fsm.bmpStatsUpdate(bmp.BMP_STAT_TYPE_WITHDRAW_PREFIX, routes) + } + } + + table.UpdatePathAttrs4ByteAs(body) + if err = table.UpdatePathAggregator4ByteAs(body); err != nil { + fmsg.MsgData = err + return fmsg, err + } + + fmsg.PathList = table.ProcessMessage(m, h.fsm.peerInfo, fmsg.timestamp) + fallthrough + case bgp.BGP_MSG_KEEPALIVE: + // if the length of h.holdTimerResetCh + // isn't zero, the timer will be reset + // soon anyway. + select { + case h.holdTimerResetCh <- true: + default: + } + if m.Header.Type == bgp.BGP_MSG_KEEPALIVE { + return nil, nil + } + case bgp.BGP_MSG_NOTIFICATION: + body := m.Body.(*bgp.BGPNotification) + if body.ErrorCode == bgp.BGP_ERROR_CEASE && (body.ErrorSubcode == bgp.BGP_ERROR_SUB_ADMINISTRATIVE_SHUTDOWN || body.ErrorSubcode == bgp.BGP_ERROR_SUB_ADMINISTRATIVE_RESET) { + communication, rest := decodeAdministrativeCommunication(body.Data) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "Code": body.ErrorCode, + "Subcode": body.ErrorSubcode, + "Communicated-Reason": communication, + "Data": rest, + }).Warn("received notification") + } else { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": h.fsm.pConf.State.NeighborAddress, + "Code": body.ErrorCode, + "Subcode": body.ErrorSubcode, + "Data": body.Data, + }).Warn("received notification") + } + + if s := h.fsm.pConf.GracefulRestart.State; s.Enabled && s.NotificationEnabled && body.ErrorCode == bgp.BGP_ERROR_CEASE && body.ErrorSubcode == bgp.BGP_ERROR_SUB_HARD_RESET { + sendToStateReasonCh(FSM_HARD_RESET, m) + } else { + sendToStateReasonCh(FSM_NOTIFICATION_RECV, m) + } + return nil, nil + } + } + } + return fmsg, nil +} + +func (h *FSMHandler) recvMessage() error { + defer h.msgCh.Close() + fmsg, _ := h.recvMessageWithError() + if fmsg != nil { + h.msgCh.In() <- fmsg + } + return nil +} + +func open2Cap(open *bgp.BGPOpen, n *config.Neighbor) (map[bgp.BGPCapabilityCode][]bgp.ParameterCapabilityInterface, map[bgp.RouteFamily]bgp.BGPAddPathMode) { + capMap := make(map[bgp.BGPCapabilityCode][]bgp.ParameterCapabilityInterface) + for _, p := range open.OptParams { + if paramCap, y := p.(*bgp.OptionParameterCapability); y { + for _, c := range paramCap.Capability { + m, ok := capMap[c.Code()] + if !ok { + m = make([]bgp.ParameterCapabilityInterface, 0, 1) + } + capMap[c.Code()] = append(m, c) + } + } + } + + // squash add path cap + if caps, y := capMap[bgp.BGP_CAP_ADD_PATH]; y { + items := make([]*bgp.CapAddPathTuple, 0, len(caps)) + for _, c := range caps { + items = append(items, c.(*bgp.CapAddPath).Tuples...) + } + capMap[bgp.BGP_CAP_ADD_PATH] = []bgp.ParameterCapabilityInterface{bgp.NewCapAddPath(items)} + } + + // remote open message may not include multi-protocol capability + if _, y := capMap[bgp.BGP_CAP_MULTIPROTOCOL]; !y { + capMap[bgp.BGP_CAP_MULTIPROTOCOL] = []bgp.ParameterCapabilityInterface{bgp.NewCapMultiProtocol(bgp.RF_IPv4_UC)} + } + + local := n.CreateRfMap() + remote := make(map[bgp.RouteFamily]bgp.BGPAddPathMode) + for _, c := range capMap[bgp.BGP_CAP_MULTIPROTOCOL] { + family := c.(*bgp.CapMultiProtocol).CapValue + remote[family] = bgp.BGP_ADD_PATH_NONE + for _, a := range capMap[bgp.BGP_CAP_ADD_PATH] { + for _, i := range a.(*bgp.CapAddPath).Tuples { + if i.RouteFamily == family { + remote[family] = i.Mode + } + } + } + } + negotiated := make(map[bgp.RouteFamily]bgp.BGPAddPathMode) + for family, mode := range local { + if m, y := remote[family]; y { + n := bgp.BGP_ADD_PATH_NONE + if mode&bgp.BGP_ADD_PATH_SEND > 0 && m&bgp.BGP_ADD_PATH_RECEIVE > 0 { + n |= bgp.BGP_ADD_PATH_SEND + } + if mode&bgp.BGP_ADD_PATH_RECEIVE > 0 && m&bgp.BGP_ADD_PATH_SEND > 0 { + n |= bgp.BGP_ADD_PATH_RECEIVE + } + negotiated[family] = n + } + } + return capMap, negotiated +} + +func (h *FSMHandler) opensent() (bgp.FSMState, *FsmStateReason) { + fsm := h.fsm + m := buildopen(fsm.gConf, fsm.pConf) + b, _ := m.Serialize() + fsm.conn.Write(b) + fsm.bgpMessageStateUpdate(m.Header.Type, false) + + h.msgCh = channels.NewInfiniteChannel() + h.conn = fsm.conn + + h.t.Go(h.recvMessage) + + // RFC 4271 P.60 + // sets its HoldTimer to a large value + // A HoldTimer value of 4 minutes is suggested as a "large value" + // for the HoldTimer + holdTimer := time.NewTimer(time.Second * time.Duration(fsm.opensentHoldTime)) + + for { + select { + case <-h.t.Dying(): + h.conn.Close() + return -1, NewFsmStateReason(FSM_DYING, nil, nil) + case conn, ok := <-fsm.connCh: + if !ok { + break + } + conn.Close() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("Closed an accepted connection") + case <-fsm.gracefulRestartTimer.C: + if fsm.pConf.GracefulRestart.State.PeerRestarting { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("graceful restart timer expired") + h.conn.Close() + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_RESTART_TIMER_EXPIRED, nil, nil) + } + case i, ok := <-h.msgCh.Out(): + if !ok { + continue + } + e := i.(*FsmMsg) + switch e.MsgData.(type) { + case *bgp.BGPMessage: + m := e.MsgData.(*bgp.BGPMessage) + if m.Header.Type == bgp.BGP_MSG_OPEN { + fsm.recvOpen = m + body := m.Body.(*bgp.BGPOpen) + peerAs, err := bgp.ValidateOpenMsg(body, fsm.pConf.Config.PeerAs) + if err != nil { + m, _ := fsm.sendNotificationFromErrorMsg(err.(*bgp.MessageError)) + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_INVALID_MSG, m, nil) + } + + // ASN negotiation was skipped + if fsm.pConf.Config.PeerAs == 0 { + typ := config.PEER_TYPE_EXTERNAL + if fsm.peerInfo.LocalAS == peerAs { + typ = config.PEER_TYPE_INTERNAL + } + fsm.pConf.State.PeerType = typ + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Infof("skipped asn negotiation: peer-as: %d, peer-type: %s", peerAs, typ) + } else { + fsm.pConf.State.PeerType = fsm.pConf.Config.PeerType + } + fsm.pConf.State.PeerAs = peerAs + fsm.peerInfo.AS = peerAs + fsm.peerInfo.ID = body.ID + fsm.capMap, fsm.rfMap = open2Cap(body, fsm.pConf) + + if _, y := fsm.capMap[bgp.BGP_CAP_ADD_PATH]; y { + fsm.marshallingOptions = &bgp.MarshallingOption{ + AddPath: fsm.rfMap, + } + } else { + fsm.marshallingOptions = nil + } + + // calculate HoldTime + // RFC 4271 P.13 + // a BGP speaker MUST calculate the value of the Hold Timer + // by using the smaller of its configured Hold Time and the Hold Time + // received in the OPEN message. + holdTime := float64(body.HoldTime) + myHoldTime := fsm.pConf.Timers.Config.HoldTime + if holdTime > myHoldTime { + fsm.pConf.Timers.State.NegotiatedHoldTime = myHoldTime + } else { + fsm.pConf.Timers.State.NegotiatedHoldTime = holdTime + } + + keepalive := fsm.pConf.Timers.Config.KeepaliveInterval + if n := fsm.pConf.Timers.State.NegotiatedHoldTime; n < myHoldTime { + keepalive = n / 3 + } + fsm.pConf.Timers.State.KeepaliveInterval = keepalive + + gr, ok := fsm.capMap[bgp.BGP_CAP_GRACEFUL_RESTART] + if fsm.pConf.GracefulRestart.Config.Enabled && ok { + state := &fsm.pConf.GracefulRestart.State + state.Enabled = true + cap := gr[len(gr)-1].(*bgp.CapGracefulRestart) + state.PeerRestartTime = uint16(cap.Time) + + for _, t := range cap.Tuples { + n := bgp.AddressFamilyNameMap[bgp.AfiSafiToRouteFamily(t.AFI, t.SAFI)] + for i, a := range fsm.pConf.AfiSafis { + if string(a.Config.AfiSafiName) == n { + fsm.pConf.AfiSafis[i].MpGracefulRestart.State.Enabled = true + fsm.pConf.AfiSafis[i].MpGracefulRestart.State.Received = true + break + } + } + } + + // RFC 4724 4.1 + // To re-establish the session with its peer, the Restarting Speaker + // MUST set the "Restart State" bit in the Graceful Restart Capability + // of the OPEN message. + if fsm.pConf.GracefulRestart.State.PeerRestarting && cap.Flags&0x08 == 0 { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("restart flag is not set") + // send notification? + h.conn.Close() + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_INVALID_MSG, nil, nil) + } + + // RFC 4724 3 + // The most significant bit is defined as the Restart State (R) + // bit, ...(snip)... When set (value 1), this bit + // indicates that the BGP speaker has restarted, and its peer MUST + // NOT wait for the End-of-RIB marker from the speaker before + // advertising routing information to the speaker. + if fsm.pConf.GracefulRestart.State.LocalRestarting && cap.Flags&0x08 != 0 { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Debug("peer has restarted, skipping wait for EOR") + for i := range fsm.pConf.AfiSafis { + fsm.pConf.AfiSafis[i].MpGracefulRestart.State.EndOfRibReceived = true + } + } + if fsm.pConf.GracefulRestart.Config.NotificationEnabled && cap.Flags&0x04 > 0 { + fsm.pConf.GracefulRestart.State.NotificationEnabled = true + } + } + llgr, ok2 := fsm.capMap[bgp.BGP_CAP_LONG_LIVED_GRACEFUL_RESTART] + if fsm.pConf.GracefulRestart.Config.LongLivedEnabled && ok && ok2 { + fsm.pConf.GracefulRestart.State.LongLivedEnabled = true + cap := llgr[len(llgr)-1].(*bgp.CapLongLivedGracefulRestart) + for _, t := range cap.Tuples { + n := bgp.AddressFamilyNameMap[bgp.AfiSafiToRouteFamily(t.AFI, t.SAFI)] + for i, a := range fsm.pConf.AfiSafis { + if string(a.Config.AfiSafiName) == n { + fsm.pConf.AfiSafis[i].LongLivedGracefulRestart.State.Enabled = true + fsm.pConf.AfiSafis[i].LongLivedGracefulRestart.State.Received = true + fsm.pConf.AfiSafis[i].LongLivedGracefulRestart.State.PeerRestartTime = t.RestartTime + break + } + } + } + } + + msg := bgp.NewBGPKeepAliveMessage() + b, _ := msg.Serialize() + fsm.conn.Write(b) + fsm.bgpMessageStateUpdate(msg.Header.Type, false) + return bgp.BGP_FSM_OPENCONFIRM, NewFsmStateReason(FSM_OPEN_MSG_RECEIVED, nil, nil) + } else { + // send notification? + h.conn.Close() + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_INVALID_MSG, nil, nil) + } + case *bgp.MessageError: + m, _ := fsm.sendNotificationFromErrorMsg(e.MsgData.(*bgp.MessageError)) + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_INVALID_MSG, m, nil) + default: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Data": e.MsgData, + }).Panic("unknown msg type") + } + case err := <-h.stateReasonCh: + h.conn.Close() + return bgp.BGP_FSM_IDLE, &err + case <-holdTimer.C: + m, _ := fsm.sendNotification(bgp.BGP_ERROR_HOLD_TIMER_EXPIRED, 0, nil, "hold timer expired") + h.t.Kill(nil) + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_HOLD_TIMER_EXPIRED, m, nil) + case stateOp := <-fsm.adminStateCh: + err := h.changeAdminState(stateOp.State) + if err == nil { + switch stateOp.State { + case ADMIN_STATE_DOWN: + h.conn.Close() + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_ADMIN_DOWN, m, nil) + case ADMIN_STATE_UP: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "AdminState": stateOp.State.String(), + }).Panic("code logic bug") + } + } + } + } +} + +func keepaliveTicker(fsm *FSM) *time.Ticker { + negotiatedTime := fsm.pConf.Timers.State.NegotiatedHoldTime + if negotiatedTime == 0 { + return &time.Ticker{} + } + sec := time.Second * time.Duration(fsm.pConf.Timers.State.KeepaliveInterval) + if sec == 0 { + sec = time.Second + } + return time.NewTicker(sec) +} + +func (h *FSMHandler) openconfirm() (bgp.FSMState, *FsmStateReason) { + fsm := h.fsm + ticker := keepaliveTicker(fsm) + h.msgCh = channels.NewInfiniteChannel() + h.conn = fsm.conn + + h.t.Go(h.recvMessage) + + var holdTimer *time.Timer + if fsm.pConf.Timers.State.NegotiatedHoldTime == 0 { + holdTimer = &time.Timer{} + } else { + // RFC 4271 P.65 + // sets the HoldTimer according to the negotiated value + holdTimer = time.NewTimer(time.Second * time.Duration(fsm.pConf.Timers.State.NegotiatedHoldTime)) + } + + for { + select { + case <-h.t.Dying(): + h.conn.Close() + return -1, NewFsmStateReason(FSM_DYING, nil, nil) + case conn, ok := <-fsm.connCh: + if !ok { + break + } + conn.Close() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("Closed an accepted connection") + case <-fsm.gracefulRestartTimer.C: + if fsm.pConf.GracefulRestart.State.PeerRestarting { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("graceful restart timer expired") + h.conn.Close() + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_RESTART_TIMER_EXPIRED, nil, nil) + } + case <-ticker.C: + m := bgp.NewBGPKeepAliveMessage() + b, _ := m.Serialize() + // TODO: check error + fsm.conn.Write(b) + fsm.bgpMessageStateUpdate(m.Header.Type, false) + case i, ok := <-h.msgCh.Out(): + if !ok { + continue + } + e := i.(*FsmMsg) + switch e.MsgData.(type) { + case *bgp.BGPMessage: + m := e.MsgData.(*bgp.BGPMessage) + if m.Header.Type == bgp.BGP_MSG_KEEPALIVE { + return bgp.BGP_FSM_ESTABLISHED, NewFsmStateReason(FSM_OPEN_MSG_NEGOTIATED, nil, nil) + } + // send notification ? + h.conn.Close() + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_INVALID_MSG, nil, nil) + case *bgp.MessageError: + m, _ := fsm.sendNotificationFromErrorMsg(e.MsgData.(*bgp.MessageError)) + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_INVALID_MSG, m, nil) + default: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Data": e.MsgData, + }).Panic("unknown msg type") + } + case err := <-h.stateReasonCh: + h.conn.Close() + return bgp.BGP_FSM_IDLE, &err + case <-holdTimer.C: + m, _ := fsm.sendNotification(bgp.BGP_ERROR_HOLD_TIMER_EXPIRED, 0, nil, "hold timer expired") + h.t.Kill(nil) + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_HOLD_TIMER_EXPIRED, m, nil) + case stateOp := <-fsm.adminStateCh: + err := h.changeAdminState(stateOp.State) + if err == nil { + switch stateOp.State { + case ADMIN_STATE_DOWN: + h.conn.Close() + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_ADMIN_DOWN, nil, nil) + case ADMIN_STATE_UP: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "AdminState": stateOp.State.String(), + }).Panic("code logic bug") + } + } + } + } +} + +func (h *FSMHandler) sendMessageloop() error { + conn := h.conn + fsm := h.fsm + ticker := keepaliveTicker(fsm) + send := func(m *bgp.BGPMessage) error { + if fsm.twoByteAsTrans && m.Header.Type == bgp.BGP_MSG_UPDATE { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Data": m, + }).Debug("update for 2byte AS peer") + table.UpdatePathAttrs2ByteAs(m.Body.(*bgp.BGPUpdate)) + table.UpdatePathAggregator2ByteAs(m.Body.(*bgp.BGPUpdate)) + } + b, err := m.Serialize(h.fsm.marshallingOptions) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Data": err, + }).Warn("failed to serialize") + fsm.bgpMessageStateUpdate(0, false) + return nil + } + if err := conn.SetWriteDeadline(time.Now().Add(time.Second * time.Duration(fsm.pConf.Timers.State.NegotiatedHoldTime))); err != nil { + h.stateReasonCh <- *NewFsmStateReason(FSM_WRITE_FAILED, nil, nil) + conn.Close() + return fmt.Errorf("failed to set write deadline") + } + _, err = conn.Write(b) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Data": err, + }).Warn("failed to send") + h.stateReasonCh <- *NewFsmStateReason(FSM_WRITE_FAILED, nil, nil) + conn.Close() + return fmt.Errorf("closed") + } + fsm.bgpMessageStateUpdate(m.Header.Type, false) + + switch m.Header.Type { + case bgp.BGP_MSG_NOTIFICATION: + body := m.Body.(*bgp.BGPNotification) + if body.ErrorCode == bgp.BGP_ERROR_CEASE && (body.ErrorSubcode == bgp.BGP_ERROR_SUB_ADMINISTRATIVE_SHUTDOWN || body.ErrorSubcode == bgp.BGP_ERROR_SUB_ADMINISTRATIVE_RESET) { + communication, rest := decodeAdministrativeCommunication(body.Data) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Code": body.ErrorCode, + "Subcode": body.ErrorSubcode, + "Communicated-Reason": communication, + "Data": rest, + }).Warn("sent notification") + } else { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Code": body.ErrorCode, + "Subcode": body.ErrorSubcode, + "Data": body.Data, + }).Warn("sent notification") + } + h.stateReasonCh <- *NewFsmStateReason(FSM_NOTIFICATION_SENT, m, nil) + conn.Close() + return fmt.Errorf("closed") + case bgp.BGP_MSG_UPDATE: + update := m.Body.(*bgp.BGPUpdate) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "nlri": update.NLRI, + "withdrawals": update.WithdrawnRoutes, + "attributes": update.PathAttributes, + }).Debug("sent update") + default: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "data": m, + }).Debug("sent") + } + return nil + } + + for { + select { + case <-h.t.Dying(): + return nil + case o := <-h.outgoing.Out(): + m := o.(*FsmOutgoingMsg) + for _, msg := range table.CreateUpdateMsgFromPaths(m.Paths, h.fsm.marshallingOptions) { + if err := send(msg); err != nil { + return nil + } + } + if m.Notification != nil { + if m.StayIdle { + // current user is only prefix-limit + // fix me if this is not the case + h.changeAdminState(ADMIN_STATE_PFX_CT) + } + if err := send(m.Notification); err != nil { + return nil + } + } + case <-ticker.C: + if err := send(bgp.NewBGPKeepAliveMessage()); err != nil { + return nil + } + } + } +} + +func (h *FSMHandler) recvMessageloop() error { + for { + fmsg, err := h.recvMessageWithError() + if fmsg != nil { + h.msgCh.In() <- fmsg + } + if err != nil { + return nil + } + } +} + +func (h *FSMHandler) established() (bgp.FSMState, *FsmStateReason) { + fsm := h.fsm + h.conn = fsm.conn + h.t.Go(h.sendMessageloop) + h.msgCh = h.incoming + h.t.Go(h.recvMessageloop) + + var holdTimer *time.Timer + if fsm.pConf.Timers.State.NegotiatedHoldTime == 0 { + holdTimer = &time.Timer{} + } else { + holdTimer = time.NewTimer(time.Second * time.Duration(fsm.pConf.Timers.State.NegotiatedHoldTime)) + } + + fsm.gracefulRestartTimer.Stop() + + for { + select { + case <-h.t.Dying(): + return -1, NewFsmStateReason(FSM_DYING, nil, nil) + case conn, ok := <-fsm.connCh: + if !ok { + break + } + conn.Close() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("Closed an accepted connection") + case err := <-h.stateReasonCh: + h.conn.Close() + h.t.Kill(nil) + if s := fsm.pConf.GracefulRestart.State; s.Enabled && + (s.NotificationEnabled && err.Type == FSM_NOTIFICATION_RECV || + err.Type == FSM_READ_FAILED || + err.Type == FSM_WRITE_FAILED) { + err = *NewFsmStateReason(FSM_GRACEFUL_RESTART, nil, nil) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Info("peer graceful restart") + fsm.gracefulRestartTimer.Reset(time.Duration(fsm.pConf.GracefulRestart.State.PeerRestartTime) * time.Second) + } + return bgp.BGP_FSM_IDLE, &err + case <-holdTimer.C: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("hold timer expired") + m := bgp.NewBGPNotificationMessage(bgp.BGP_ERROR_HOLD_TIMER_EXPIRED, 0, nil) + h.outgoing.In() <- &FsmOutgoingMsg{Notification: m} + return bgp.BGP_FSM_IDLE, NewFsmStateReason(FSM_HOLD_TIMER_EXPIRED, m, nil) + case <-h.holdTimerResetCh: + if fsm.pConf.Timers.State.NegotiatedHoldTime != 0 { + holdTimer.Reset(time.Second * time.Duration(fsm.pConf.Timers.State.NegotiatedHoldTime)) + } + case stateOp := <-fsm.adminStateCh: + err := h.changeAdminState(stateOp.State) + if err == nil { + switch stateOp.State { + case ADMIN_STATE_DOWN: + m := bgp.NewBGPNotificationMessage(bgp.BGP_ERROR_CEASE, bgp.BGP_ERROR_SUB_ADMINISTRATIVE_SHUTDOWN, stateOp.Communication) + h.outgoing.In() <- &FsmOutgoingMsg{Notification: m} + } + } + } + } +} + +func (h *FSMHandler) loop() error { + fsm := h.fsm + ch := make(chan bgp.FSMState) + oldState := fsm.state + + var reason *FsmStateReason + f := func() error { + nextState := bgp.FSMState(-1) + switch fsm.state { + case bgp.BGP_FSM_IDLE: + nextState, reason = h.idle() + // case bgp.BGP_FSM_CONNECT: + // nextState = h.connect() + case bgp.BGP_FSM_ACTIVE: + nextState, reason = h.active() + case bgp.BGP_FSM_OPENSENT: + nextState, reason = h.opensent() + case bgp.BGP_FSM_OPENCONFIRM: + nextState, reason = h.openconfirm() + case bgp.BGP_FSM_ESTABLISHED: + nextState, reason = h.established() + } + fsm.reason = reason + ch <- nextState + return nil + } + + h.t.Go(f) + + nextState := <-ch + + if nextState == bgp.BGP_FSM_ESTABLISHED && oldState == bgp.BGP_FSM_OPENCONFIRM { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Info("Peer Up") + } + + if oldState == bgp.BGP_FSM_ESTABLISHED { + // The main goroutine sent the notificaiton due to + // deconfiguration or something. + reason := fsm.reason + if fsm.h.sentNotification != nil { + reason.Type = FSM_NOTIFICATION_SENT + reason.PeerDownReason = PEER_DOWN_BY_LOCAL + reason.BGPNotification = fsm.h.sentNotification + } + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "Reason": reason.String(), + }).Info("Peer Down") + } + + e := time.AfterFunc(time.Second*120, func() { + log.WithFields(log.Fields{"Topic": "Peer"}).Fatalf("failed to free the fsm.h.t for %s %s %s", fsm.pConf.State.NeighborAddress, oldState, nextState) + }) + h.t.Wait() + e.Stop() + + // under zero means that tomb.Dying() + if nextState >= bgp.BGP_FSM_IDLE { + h.stateCh <- &FsmMsg{ + MsgType: FSM_MSG_STATE_CHANGE, + MsgSrc: fsm.pConf.State.NeighborAddress, + MsgData: nextState, + StateReason: reason, + Version: h.fsm.version, + } + } + return nil +} + +func (h *FSMHandler) changeAdminState(s AdminState) error { + fsm := h.fsm + if fsm.adminState != s { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + "AdminState": s.String(), + }).Debug("admin state changed") + + fsm.adminState = s + fsm.pConf.State.AdminDown = !fsm.pConf.State.AdminDown + + switch s { + case ADMIN_STATE_UP: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Info("Administrative start") + case ADMIN_STATE_DOWN: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Info("Administrative shutdown") + case ADMIN_STATE_PFX_CT: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Info("Administrative shutdown(Prefix limit reached)") + } + + } else { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": fsm.pConf.State.NeighborAddress, + "State": fsm.state.String(), + }).Warn("cannot change to the same state") + + return fmt.Errorf("cannot change to the same state.") + } + return nil +} diff --git a/pkg/server/fsm_test.go b/pkg/server/fsm_test.go new file mode 100644 index 00000000..dad6563a --- /dev/null +++ b/pkg/server/fsm_test.go @@ -0,0 +1,345 @@ +// Copyright (C) 2014 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "errors" + "net" + "strconv" + "sync" + "testing" + "time" + + "github.com/eapache/channels" + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + + log "github.com/sirupsen/logrus" + "github.com/stretchr/testify/assert" +) + +type MockConnection struct { + *testing.T + net.Conn + recvCh chan chan byte + sendBuf [][]byte + currentCh chan byte + isClosed bool + wait int + mtx sync.Mutex +} + +func NewMockConnection(t *testing.T) *MockConnection { + m := &MockConnection{ + T: t, + recvCh: make(chan chan byte, 128), + sendBuf: make([][]byte, 0), + isClosed: false, + } + return m +} + +func (m *MockConnection) SetWriteDeadline(t time.Time) error { + return nil +} + +func (m *MockConnection) setData(data []byte) int { + dataChan := make(chan byte, 4096) + for _, b := range data { + dataChan <- b + } + m.recvCh <- dataChan + return len(dataChan) +} + +func (m *MockConnection) Read(buf []byte) (int, error) { + m.mtx.Lock() + closed := m.isClosed + m.mtx.Unlock() + if closed { + return 0, errors.New("already closed") + } + + if m.currentCh == nil { + m.currentCh = <-m.recvCh + } + + length := 0 + rest := len(buf) + for i := 0; i < rest; i++ { + if len(m.currentCh) > 0 { + val := <-m.currentCh + buf[i] = val + length++ + } else { + m.currentCh = nil + break + } + } + + m.Logf("%d bytes read from peer", length) + return length, nil +} + +func (m *MockConnection) Write(buf []byte) (int, error) { + time.Sleep(time.Duration(m.wait) * time.Millisecond) + m.sendBuf = append(m.sendBuf, buf) + msg, _ := bgp.ParseBGPMessage(buf) + m.Logf("%d bytes written by gobgp message type : %s", + len(buf), showMessageType(msg.Header.Type)) + return len(buf), nil +} + +func showMessageType(t uint8) string { + switch t { + case bgp.BGP_MSG_KEEPALIVE: + return "BGP_MSG_KEEPALIVE" + case bgp.BGP_MSG_NOTIFICATION: + return "BGP_MSG_NOTIFICATION" + case bgp.BGP_MSG_OPEN: + return "BGP_MSG_OPEN" + case bgp.BGP_MSG_UPDATE: + return "BGP_MSG_UPDATE" + case bgp.BGP_MSG_ROUTE_REFRESH: + return "BGP_MSG_ROUTE_REFRESH" + } + return strconv.Itoa(int(t)) +} + +func (m *MockConnection) Close() error { + m.mtx.Lock() + defer m.mtx.Unlock() + if !m.isClosed { + close(m.recvCh) + m.isClosed = true + } + return nil +} + +func (m *MockConnection) LocalAddr() net.Addr { + return &net.TCPAddr{ + IP: net.ParseIP("10.10.10.10"), + Port: bgp.BGP_PORT} +} + +func TestReadAll(t *testing.T) { + assert := assert.New(t) + m := NewMockConnection(t) + msg := open() + expected1, _ := msg.Header.Serialize() + expected2, _ := msg.Body.Serialize() + + pushBytes := func() { + m.Log("push 5 bytes") + m.setData(expected1[0:5]) + m.Log("push rest") + m.setData(expected1[5:]) + m.Log("push bytes at once") + m.setData(expected2) + } + + go pushBytes() + + var actual1 []byte + actual1, _ = readAll(m, bgp.BGP_HEADER_LENGTH) + m.Log(actual1) + assert.Equal(expected1, actual1) + + var actual2 []byte + actual2, _ = readAll(m, len(expected2)) + m.Log(actual2) + assert.Equal(expected2, actual2) +} + +func TestFSMHandlerOpensent_HoldTimerExpired(t *testing.T) { + assert := assert.New(t) + m := NewMockConnection(t) + + p, h := makePeerAndHandler() + + // push mock connection + p.fsm.conn = m + p.fsm.h = h + + // set keepalive ticker + p.fsm.pConf.Timers.State.NegotiatedHoldTime = 3 + + // set holdtime + p.fsm.opensentHoldTime = 2 + + state, _ := h.opensent() + + assert.Equal(bgp.BGP_FSM_IDLE, state) + lastMsg := m.sendBuf[len(m.sendBuf)-1] + sent, _ := bgp.ParseBGPMessage(lastMsg) + assert.Equal(uint8(bgp.BGP_MSG_NOTIFICATION), sent.Header.Type) + assert.Equal(uint8(bgp.BGP_ERROR_HOLD_TIMER_EXPIRED), sent.Body.(*bgp.BGPNotification).ErrorCode) + +} + +func TestFSMHandlerOpenconfirm_HoldTimerExpired(t *testing.T) { + assert := assert.New(t) + m := NewMockConnection(t) + + p, h := makePeerAndHandler() + + // push mock connection + p.fsm.conn = m + p.fsm.h = h + + // set up keepalive ticker + p.fsm.pConf.Timers.Config.KeepaliveInterval = 1 + + // set holdtime + p.fsm.pConf.Timers.State.NegotiatedHoldTime = 2 + state, _ := h.openconfirm() + + assert.Equal(bgp.BGP_FSM_IDLE, state) + lastMsg := m.sendBuf[len(m.sendBuf)-1] + sent, _ := bgp.ParseBGPMessage(lastMsg) + assert.Equal(uint8(bgp.BGP_MSG_NOTIFICATION), sent.Header.Type) + assert.Equal(uint8(bgp.BGP_ERROR_HOLD_TIMER_EXPIRED), sent.Body.(*bgp.BGPNotification).ErrorCode) + +} + +func TestFSMHandlerEstablish_HoldTimerExpired(t *testing.T) { + assert := assert.New(t) + m := NewMockConnection(t) + + p, h := makePeerAndHandler() + + // push mock connection + p.fsm.conn = m + p.fsm.h = h + + // set keepalive ticker + p.fsm.pConf.Timers.State.NegotiatedHoldTime = 3 + + msg := keepalive() + header, _ := msg.Header.Serialize() + body, _ := msg.Body.Serialize() + + pushPackets := func() { + // first keepalive from peer + m.setData(header) + m.setData(body) + } + + // set holdtime + p.fsm.pConf.Timers.Config.HoldTime = 2 + p.fsm.pConf.Timers.State.NegotiatedHoldTime = 2 + + go pushPackets() + state, _ := h.established() + time.Sleep(time.Second * 1) + assert.Equal(bgp.BGP_FSM_IDLE, state) + m.mtx.Lock() + lastMsg := m.sendBuf[len(m.sendBuf)-1] + m.mtx.Unlock() + sent, _ := bgp.ParseBGPMessage(lastMsg) + assert.Equal(uint8(bgp.BGP_MSG_NOTIFICATION), sent.Header.Type) + assert.Equal(uint8(bgp.BGP_ERROR_HOLD_TIMER_EXPIRED), sent.Body.(*bgp.BGPNotification).ErrorCode) +} + +func TestFSMHandlerOpenconfirm_HoldtimeZero(t *testing.T) { + log.SetLevel(log.DebugLevel) + assert := assert.New(t) + m := NewMockConnection(t) + + p, h := makePeerAndHandler() + + // push mock connection + p.fsm.conn = m + p.fsm.h = h + + // set up keepalive ticker + p.fsm.pConf.Timers.Config.KeepaliveInterval = 1 + // set holdtime + p.fsm.pConf.Timers.State.NegotiatedHoldTime = 0 + go h.openconfirm() + + time.Sleep(100 * time.Millisecond) + + assert.Equal(0, len(m.sendBuf)) + +} + +func TestFSMHandlerEstablished_HoldtimeZero(t *testing.T) { + log.SetLevel(log.DebugLevel) + assert := assert.New(t) + m := NewMockConnection(t) + + p, h := makePeerAndHandler() + + // push mock connection + p.fsm.conn = m + p.fsm.h = h + + // set holdtime + p.fsm.pConf.Timers.State.NegotiatedHoldTime = 0 + + go h.established() + + time.Sleep(100 * time.Millisecond) + + assert.Equal(0, len(m.sendBuf)) +} + +func TestCheckOwnASLoop(t *testing.T) { + assert := assert.New(t) + aspathParam := []bgp.AsPathParamInterface{bgp.NewAs4PathParam(2, []uint32{65100})} + aspath := bgp.NewPathAttributeAsPath(aspathParam) + assert.False(hasOwnASLoop(65100, 10, aspath)) + assert.True(hasOwnASLoop(65100, 0, aspath)) + assert.False(hasOwnASLoop(65200, 0, aspath)) +} + +func makePeerAndHandler() (*Peer, *FSMHandler) { + p := &Peer{ + fsm: NewFSM(&config.Global{}, &config.Neighbor{}, table.NewRoutingPolicy()), + outgoing: channels.NewInfiniteChannel(), + } + + h := &FSMHandler{ + fsm: p.fsm, + stateReasonCh: make(chan FsmStateReason, 2), + incoming: channels.NewInfiniteChannel(), + outgoing: p.outgoing, + } + + return p, h + +} + +func open() *bgp.BGPMessage { + p1 := bgp.NewOptionParameterCapability( + []bgp.ParameterCapabilityInterface{bgp.NewCapRouteRefresh()}) + p2 := bgp.NewOptionParameterCapability( + []bgp.ParameterCapabilityInterface{bgp.NewCapMultiProtocol(bgp.RF_IPv4_UC)}) + g := &bgp.CapGracefulRestartTuple{AFI: 4, SAFI: 2, Flags: 3} + p3 := bgp.NewOptionParameterCapability( + []bgp.ParameterCapabilityInterface{bgp.NewCapGracefulRestart(true, true, 100, + []*bgp.CapGracefulRestartTuple{g})}) + p4 := bgp.NewOptionParameterCapability( + []bgp.ParameterCapabilityInterface{bgp.NewCapFourOctetASNumber(100000)}) + return bgp.NewBGPOpenMessage(11033, 303, "100.4.10.3", + []bgp.OptionParameterInterface{p1, p2, p3, p4}) +} + +func keepalive() *bgp.BGPMessage { + return bgp.NewBGPKeepAliveMessage() +} diff --git a/pkg/server/grpc_server.go b/pkg/server/grpc_server.go new file mode 100644 index 00000000..37b7c57b --- /dev/null +++ b/pkg/server/grpc_server.go @@ -0,0 +1,3016 @@ +// Copyright (C) 2014-2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "bytes" + "fmt" + "io" + "net" + "reflect" + "regexp" + "strconv" + "strings" + "sync" + "time" + + farm "github.com/dgryski/go-farm" + "github.com/golang/protobuf/ptypes/any" + log "github.com/sirupsen/logrus" + "golang.org/x/net/context" + "google.golang.org/grpc" + + api "github.com/osrg/gobgp/api" + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/internal/pkg/zebra" + "github.com/osrg/gobgp/pkg/packet/bgp" +) + +type Server struct { + bgpServer *BgpServer + grpcServer *grpc.Server + hosts string +} + +func NewGrpcServer(b *BgpServer, hosts string) *Server { + size := 256 << 20 + return NewServer(b, grpc.NewServer(grpc.MaxRecvMsgSize(size), grpc.MaxSendMsgSize(size)), hosts) +} + +func NewServer(b *BgpServer, g *grpc.Server, hosts string) *Server { + grpc.EnableTracing = false + s := &Server{ + bgpServer: b, + grpcServer: g, + hosts: hosts, + } + api.RegisterGobgpApiServer(g, s) + return s +} + +func (s *Server) Serve() error { + var wg sync.WaitGroup + l := strings.Split(s.hosts, ",") + wg.Add(len(l)) + + serve := func(host string) { + defer wg.Done() + lis, err := net.Listen("tcp", host) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "grpc", + "Key": host, + "Error": err, + }).Warn("listen failed") + return + } + err = s.grpcServer.Serve(lis) + log.WithFields(log.Fields{ + "Topic": "grpc", + "Key": host, + "Error": err, + }).Warn("accept failed") + } + + for _, host := range l { + go serve(host) + } + wg.Wait() + return nil +} + +func NewMpGracefulRestartFromConfigStruct(c *config.MpGracefulRestart) *api.MpGracefulRestart { + return &api.MpGracefulRestart{ + Config: &api.MpGracefulRestartConfig{ + Enabled: c.Config.Enabled, + }, + } +} + +func extractFamilyFromConfigAfiSafi(c *config.AfiSafi) uint32 { + if c == nil { + return 0 + } + // If address family value is already stored in AfiSafiState structure, + // we prefer to use this value. + if c.State.Family != 0 { + return uint32(c.State.Family) + } + // In case that Neighbor structure came from CLI or gRPC, address family + // value in AfiSafiState structure can be omitted. + // Here extracts value from AfiSafiName field in AfiSafiConfig structure. + if rf, err := bgp.GetRouteFamily(string(c.Config.AfiSafiName)); err == nil { + return uint32(rf) + } + // Ignores invalid address family name + return 0 +} + +func NewAfiSafiConfigFromConfigStruct(c *config.AfiSafi) *api.AfiSafiConfig { + return &api.AfiSafiConfig{ + Family: extractFamilyFromConfigAfiSafi(c), + Enabled: c.Config.Enabled, + } +} + +func NewApplyPolicyFromConfigStruct(c *config.ApplyPolicy) *api.ApplyPolicy { + applyPolicy := &api.ApplyPolicy{ + ImportPolicy: &api.PolicyAssignment{ + Type: api.PolicyType_IMPORT, + Default: api.RouteAction(c.Config.DefaultImportPolicy.ToInt()), + }, + ExportPolicy: &api.PolicyAssignment{ + Type: api.PolicyType_EXPORT, + Default: api.RouteAction(c.Config.DefaultExportPolicy.ToInt()), + }, + InPolicy: &api.PolicyAssignment{ + Type: api.PolicyType_IN, + Default: api.RouteAction(c.Config.DefaultInPolicy.ToInt()), + }, + } + + for _, pname := range c.Config.ImportPolicyList { + applyPolicy.ImportPolicy.Policies = append(applyPolicy.ImportPolicy.Policies, &api.Policy{Name: pname}) + } + for _, pname := range c.Config.ExportPolicyList { + applyPolicy.ExportPolicy.Policies = append(applyPolicy.ExportPolicy.Policies, &api.Policy{Name: pname}) + } + for _, pname := range c.Config.InPolicyList { + applyPolicy.InPolicy.Policies = append(applyPolicy.InPolicy.Policies, &api.Policy{Name: pname}) + } + + return applyPolicy +} + +func NewRouteSelectionOptionsFromConfigStruct(c *config.RouteSelectionOptions) *api.RouteSelectionOptions { + return &api.RouteSelectionOptions{ + Config: &api.RouteSelectionOptionsConfig{ + AlwaysCompareMed: c.Config.AlwaysCompareMed, + IgnoreAsPathLength: c.Config.IgnoreAsPathLength, + ExternalCompareRouterId: c.Config.ExternalCompareRouterId, + AdvertiseInactiveRoutes: c.Config.AdvertiseInactiveRoutes, + EnableAigp: c.Config.EnableAigp, + IgnoreNextHopIgpMetric: c.Config.IgnoreNextHopIgpMetric, + }, + } +} + +func NewUseMultiplePathsFromConfigStruct(c *config.UseMultiplePaths) *api.UseMultiplePaths { + return &api.UseMultiplePaths{ + Config: &api.UseMultiplePathsConfig{ + Enabled: c.Config.Enabled, + }, + Ebgp: &api.Ebgp{ + Config: &api.EbgpConfig{ + AllowMultipleAs: c.Ebgp.Config.AllowMultipleAs, + MaximumPaths: c.Ebgp.Config.MaximumPaths, + }, + }, + Ibgp: &api.Ibgp{ + Config: &api.IbgpConfig{ + MaximumPaths: c.Ibgp.Config.MaximumPaths, + }, + }, + } +} + +func NewPrefixLimitFromConfigStruct(c *config.AfiSafi) *api.PrefixLimit { + if c.PrefixLimit.Config.MaxPrefixes == 0 { + return nil + } + + return &api.PrefixLimit{ + Family: uint32(c.State.Family), + MaxPrefixes: c.PrefixLimit.Config.MaxPrefixes, + ShutdownThresholdPct: uint32(c.PrefixLimit.Config.ShutdownThresholdPct), + } +} + +func NewRouteTargetMembershipFromConfigStruct(c *config.RouteTargetMembership) *api.RouteTargetMembership { + return &api.RouteTargetMembership{ + Config: &api.RouteTargetMembershipConfig{ + DeferralTime: uint32(c.Config.DeferralTime), + }, + } +} + +func NewLongLivedGracefulRestartFromConfigStruct(c *config.LongLivedGracefulRestart) *api.LongLivedGracefulRestart { + return &api.LongLivedGracefulRestart{ + Config: &api.LongLivedGracefulRestartConfig{ + Enabled: c.Config.Enabled, + RestartTime: c.Config.RestartTime, + }, + } +} + +func NewAddPathsFromConfigStruct(c *config.AddPaths) *api.AddPaths { + return &api.AddPaths{ + Config: &api.AddPathsConfig{ + Receive: c.Config.Receive, + SendMax: uint32(c.Config.SendMax), + }, + } +} + +func NewAfiSafiFromConfigStruct(c *config.AfiSafi) *api.AfiSafi { + return &api.AfiSafi{ + MpGracefulRestart: NewMpGracefulRestartFromConfigStruct(&c.MpGracefulRestart), + Config: NewAfiSafiConfigFromConfigStruct(c), + ApplyPolicy: NewApplyPolicyFromConfigStruct(&c.ApplyPolicy), + RouteSelectionOptions: NewRouteSelectionOptionsFromConfigStruct(&c.RouteSelectionOptions), + UseMultiplePaths: NewUseMultiplePathsFromConfigStruct(&c.UseMultiplePaths), + PrefixLimits: NewPrefixLimitFromConfigStruct(c), + RouteTargetMembership: NewRouteTargetMembershipFromConfigStruct(&c.RouteTargetMembership), + LongLivedGracefulRestart: NewLongLivedGracefulRestartFromConfigStruct(&c.LongLivedGracefulRestart), + AddPaths: NewAddPathsFromConfigStruct(&c.AddPaths), + } +} + +func NewPeerFromConfigStruct(pconf *config.Neighbor) *api.Peer { + families := make([]uint32, 0, len(pconf.AfiSafis)) + prefixLimits := make([]*api.PrefixLimit, 0, len(pconf.AfiSafis)) + afiSafis := make([]*api.AfiSafi, 0, len(pconf.AfiSafis)) + for _, f := range pconf.AfiSafis { + families = append(families, extractFamilyFromConfigAfiSafi(&f)) + if prefixLimit := NewPrefixLimitFromConfigStruct(&f); prefixLimit != nil { + prefixLimits = append(prefixLimits, prefixLimit) + } + if afiSafi := NewAfiSafiFromConfigStruct(&f); afiSafi != nil { + afiSafis = append(afiSafis, afiSafi) + } + } + + timer := pconf.Timers + s := pconf.State + localAddress := pconf.Transport.Config.LocalAddress + if pconf.Transport.State.LocalAddress != "" { + localAddress = pconf.Transport.State.LocalAddress + } + remoteCap, err := api.MarshalCapabilities(pconf.State.RemoteCapabilityList) + if err != nil { + return nil + } + localCap, err := api.MarshalCapabilities(pconf.State.LocalCapabilityList) + if err != nil { + return nil + } + var removePrivateAs api.PeerConf_RemovePrivateAs + switch pconf.Config.RemovePrivateAs { + case config.REMOVE_PRIVATE_AS_OPTION_ALL: + removePrivateAs = api.PeerConf_ALL + case config.REMOVE_PRIVATE_AS_OPTION_REPLACE: + removePrivateAs = api.PeerConf_REPLACE + } + return &api.Peer{ + Families: families, + ApplyPolicy: NewApplyPolicyFromConfigStruct(&pconf.ApplyPolicy), + Conf: &api.PeerConf{ + NeighborAddress: pconf.Config.NeighborAddress, + Id: s.RemoteRouterId, + PeerAs: pconf.Config.PeerAs, + LocalAs: pconf.Config.LocalAs, + PeerType: uint32(pconf.Config.PeerType.ToInt()), + AuthPassword: pconf.Config.AuthPassword, + RouteFlapDamping: pconf.Config.RouteFlapDamping, + Description: pconf.Config.Description, + PeerGroup: pconf.Config.PeerGroup, + RemoteCap: remoteCap, + LocalCap: localCap, + PrefixLimits: prefixLimits, + LocalAddress: localAddress, + NeighborInterface: pconf.Config.NeighborInterface, + Vrf: pconf.Config.Vrf, + AllowOwnAs: uint32(pconf.AsPathOptions.Config.AllowOwnAs), + RemovePrivateAs: removePrivateAs, + ReplacePeerAs: pconf.AsPathOptions.Config.ReplacePeerAs, + }, + Info: &api.PeerState{ + BgpState: string(s.SessionState), + AdminState: api.PeerState_AdminState(s.AdminState.ToInt()), + Messages: &api.Messages{ + Received: &api.Message{ + NOTIFICATION: s.Messages.Received.Notification, + UPDATE: s.Messages.Received.Update, + OPEN: s.Messages.Received.Open, + KEEPALIVE: s.Messages.Received.Keepalive, + REFRESH: s.Messages.Received.Refresh, + DISCARDED: s.Messages.Received.Discarded, + TOTAL: s.Messages.Received.Total, + }, + Sent: &api.Message{ + NOTIFICATION: s.Messages.Sent.Notification, + UPDATE: s.Messages.Sent.Update, + OPEN: s.Messages.Sent.Open, + KEEPALIVE: s.Messages.Sent.Keepalive, + REFRESH: s.Messages.Sent.Refresh, + DISCARDED: s.Messages.Sent.Discarded, + TOTAL: s.Messages.Sent.Total, + }, + }, + Received: s.AdjTable.Received, + Accepted: s.AdjTable.Accepted, + Advertised: s.AdjTable.Advertised, + PeerAs: s.PeerAs, + PeerType: uint32(s.PeerType.ToInt()), + NeighborAddress: pconf.State.NeighborAddress, + }, + Timers: &api.Timers{ + Config: &api.TimersConfig{ + ConnectRetry: uint64(timer.Config.ConnectRetry), + HoldTime: uint64(timer.Config.HoldTime), + KeepaliveInterval: uint64(timer.Config.KeepaliveInterval), + }, + State: &api.TimersState{ + KeepaliveInterval: uint64(timer.State.KeepaliveInterval), + NegotiatedHoldTime: uint64(timer.State.NegotiatedHoldTime), + Uptime: uint64(timer.State.Uptime), + Downtime: uint64(timer.State.Downtime), + }, + }, + RouteReflector: &api.RouteReflector{ + RouteReflectorClient: pconf.RouteReflector.Config.RouteReflectorClient, + RouteReflectorClusterId: string(pconf.RouteReflector.State.RouteReflectorClusterId), + }, + RouteServer: &api.RouteServer{ + RouteServerClient: pconf.RouteServer.Config.RouteServerClient, + }, + GracefulRestart: &api.GracefulRestart{ + Enabled: pconf.GracefulRestart.Config.Enabled, + RestartTime: uint32(pconf.GracefulRestart.Config.RestartTime), + HelperOnly: pconf.GracefulRestart.Config.HelperOnly, + DeferralTime: uint32(pconf.GracefulRestart.Config.DeferralTime), + NotificationEnabled: pconf.GracefulRestart.Config.NotificationEnabled, + LonglivedEnabled: pconf.GracefulRestart.Config.LongLivedEnabled, + LocalRestarting: pconf.GracefulRestart.State.LocalRestarting, + }, + Transport: &api.Transport{ + RemotePort: uint32(pconf.Transport.Config.RemotePort), + LocalAddress: pconf.Transport.Config.LocalAddress, + PassiveMode: pconf.Transport.Config.PassiveMode, + }, + AfiSafis: afiSafis, + AddPaths: NewAddPathsFromConfigStruct(&pconf.AddPaths), + } +} + +func NewPeerGroupFromConfigStruct(pconf *config.PeerGroup) *api.PeerGroup { + families := make([]uint32, 0, len(pconf.AfiSafis)) + afiSafis := make([]*api.AfiSafi, 0, len(pconf.AfiSafis)) + for _, f := range pconf.AfiSafis { + families = append(families, extractFamilyFromConfigAfiSafi(&f)) + if afiSafi := NewAfiSafiFromConfigStruct(&f); afiSafi != nil { + afiSafis = append(afiSafis, afiSafi) + } + } + + timer := pconf.Timers + s := pconf.State + return &api.PeerGroup{ + Families: families, + ApplyPolicy: NewApplyPolicyFromConfigStruct(&pconf.ApplyPolicy), + Conf: &api.PeerGroupConf{ + PeerAs: pconf.Config.PeerAs, + LocalAs: pconf.Config.LocalAs, + PeerType: uint32(pconf.Config.PeerType.ToInt()), + AuthPassword: pconf.Config.AuthPassword, + RouteFlapDamping: pconf.Config.RouteFlapDamping, + Description: pconf.Config.Description, + PeerGroupName: pconf.Config.PeerGroupName, + }, + Info: &api.PeerGroupState{ + PeerAs: s.PeerAs, + PeerType: uint32(s.PeerType.ToInt()), + TotalPaths: s.TotalPaths, + TotalPrefixes: s.TotalPrefixes, + }, + Timers: &api.Timers{ + Config: &api.TimersConfig{ + ConnectRetry: uint64(timer.Config.ConnectRetry), + HoldTime: uint64(timer.Config.HoldTime), + KeepaliveInterval: uint64(timer.Config.KeepaliveInterval), + }, + State: &api.TimersState{ + KeepaliveInterval: uint64(timer.State.KeepaliveInterval), + NegotiatedHoldTime: uint64(timer.State.NegotiatedHoldTime), + Uptime: uint64(timer.State.Uptime), + Downtime: uint64(timer.State.Downtime), + }, + }, + RouteReflector: &api.RouteReflector{ + RouteReflectorClient: pconf.RouteReflector.Config.RouteReflectorClient, + RouteReflectorClusterId: string(pconf.RouteReflector.Config.RouteReflectorClusterId), + }, + RouteServer: &api.RouteServer{ + RouteServerClient: pconf.RouteServer.Config.RouteServerClient, + }, + GracefulRestart: &api.GracefulRestart{ + Enabled: pconf.GracefulRestart.Config.Enabled, + RestartTime: uint32(pconf.GracefulRestart.Config.RestartTime), + HelperOnly: pconf.GracefulRestart.Config.HelperOnly, + DeferralTime: uint32(pconf.GracefulRestart.Config.DeferralTime), + NotificationEnabled: pconf.GracefulRestart.Config.NotificationEnabled, + LonglivedEnabled: pconf.GracefulRestart.Config.LongLivedEnabled, + LocalRestarting: pconf.GracefulRestart.State.LocalRestarting, + }, + Transport: &api.Transport{ + RemotePort: uint32(pconf.Transport.Config.RemotePort), + LocalAddress: pconf.Transport.Config.LocalAddress, + PassiveMode: pconf.Transport.Config.PassiveMode, + }, + AfiSafis: afiSafis, + AddPaths: NewAddPathsFromConfigStruct(&pconf.AddPaths), + } +} + +func (s *Server) GetNeighbor(ctx context.Context, arg *api.GetNeighborRequest) (*api.GetNeighborResponse, error) { + if arg == nil { + return nil, fmt.Errorf("invalid request") + } + neighbors := s.bgpServer.GetNeighbor(arg.Address, arg.EnableAdvertised) + peers := make([]*api.Peer, 0, len(neighbors)) + for _, e := range neighbors { + peers = append(peers, NewPeerFromConfigStruct(e)) + } + return &api.GetNeighborResponse{Peers: peers}, nil +} + +func NewValidationFromTableStruct(v *table.Validation) *api.RPKIValidation { + if v == nil { + return &api.RPKIValidation{} + } + return &api.RPKIValidation{ + Reason: api.RPKIValidation_Reason(v.Reason.ToInt()), + Matched: NewRoaListFromTableStructList(v.Matched), + UnmatchedAs: NewRoaListFromTableStructList(v.UnmatchedAs), + UnmatchedLength: NewRoaListFromTableStructList(v.UnmatchedLength), + } +} + +func toPathAPI(binNlri []byte, binPattrs [][]byte, anyNlri *any.Any, anyPattrs []*any.Any, path *table.Path, v *table.Validation) *api.Path { + nlri := path.GetNlri() + family := uint32(path.GetRouteFamily()) + vv := config.RPKI_VALIDATION_RESULT_TYPE_NONE.ToInt() + if v != nil { + vv = v.Status.ToInt() + } + p := &api.Path{ + Nlri: binNlri, + Pattrs: binPattrs, + Age: path.GetTimestamp().Unix(), + IsWithdraw: path.IsWithdraw, + Validation: int32(vv), + ValidationDetail: NewValidationFromTableStruct(v), + Family: family, + Stale: path.IsStale(), + IsFromExternal: path.IsFromExternal(), + NoImplicitWithdraw: path.NoImplicitWithdraw(), + IsNexthopInvalid: path.IsNexthopInvalid, + Identifier: nlri.PathIdentifier(), + LocalIdentifier: nlri.PathLocalIdentifier(), + AnyNlri: anyNlri, + AnyPattrs: anyPattrs, + } + if s := path.GetSource(); s != nil { + p.SourceAsn = s.AS + p.SourceId = s.ID.String() + p.NeighborIp = s.Address.String() + } + return p +} + +func ToPathApi(path *table.Path, v *table.Validation) *api.Path { + nlri := path.GetNlri() + anyNlri := api.MarshalNLRI(nlri) + if path.IsWithdraw { + return toPathAPI(nil, nil, anyNlri, nil, path, v) + } + anyPattrs := api.MarshalPathAttributes(path.GetPathAttrs()) + return toPathAPI(nil, nil, anyNlri, anyPattrs, path, v) +} + +func getValidation(v []*table.Validation, i int) *table.Validation { + if v == nil { + return nil + } else { + return v[i] + } +} + +func (s *Server) GetRib(ctx context.Context, arg *api.GetRibRequest) (*api.GetRibResponse, error) { + if arg == nil || arg.Table == nil { + return nil, fmt.Errorf("invalid request") + } + f := func() []*table.LookupPrefix { + l := make([]*table.LookupPrefix, 0, len(arg.Table.Destinations)) + for _, p := range arg.Table.Destinations { + l = append(l, &table.LookupPrefix{ + Prefix: p.Prefix, + LookupOption: func() table.LookupOption { + if p.LongerPrefixes { + return table.LOOKUP_LONGER + } else if p.ShorterPrefixes { + return table.LOOKUP_SHORTER + } + return table.LOOKUP_EXACT + }(), + }) + } + return l + } + + var in bool + var err error + var tbl *table.Table + var v []*table.Validation + + family := bgp.RouteFamily(arg.Table.Family) + switch arg.Table.Type { + case api.Resource_LOCAL, api.Resource_GLOBAL: + tbl, v, err = s.bgpServer.GetRib(arg.Table.Name, family, f()) + case api.Resource_ADJ_IN: + in = true + fallthrough + case api.Resource_ADJ_OUT: + tbl, v, err = s.bgpServer.GetAdjRib(arg.Table.Name, family, in, f()) + case api.Resource_VRF: + tbl, err = s.bgpServer.GetVrfRib(arg.Table.Name, family, []*table.LookupPrefix{}) + default: + return nil, fmt.Errorf("unsupported resource type: %v", arg.Table.Type) + } + + if err != nil { + return nil, err + } + + tblDsts := tbl.GetDestinations() + dsts := make([]*api.Destination, 0, len(tblDsts)) + idx := 0 + for _, dst := range tblDsts { + dsts = append(dsts, &api.Destination{ + Prefix: dst.GetNlri().String(), + Paths: func(paths []*table.Path) []*api.Path { + l := make([]*api.Path, 0, len(paths)) + for i, p := range paths { + pp := ToPathApi(p, getValidation(v, idx)) + idx++ + switch arg.Table.Type { + case api.Resource_LOCAL, api.Resource_GLOBAL: + if i == 0 && !table.SelectionOptions.DisableBestPathSelection { + pp.Best = true + } + } + l = append(l, pp) + } + return l + }(dst.GetAllKnownPathList()), + }) + } + + return &api.GetRibResponse{Table: &api.Table{ + Type: arg.Table.Type, + Family: uint32(tbl.GetRoutefamily()), + Destinations: dsts}, + }, err +} + +func (s *Server) GetPath(arg *api.GetPathRequest, stream api.GobgpApi_GetPathServer) error { + f := func() []*table.LookupPrefix { + l := make([]*table.LookupPrefix, 0, len(arg.Prefixes)) + for _, p := range arg.Prefixes { + l = append(l, &table.LookupPrefix{ + Prefix: p.Prefix, + LookupOption: table.LookupOption(p.LookupOption), + }) + } + return l + } + + in := false + family := bgp.RouteFamily(arg.Family) + var tbl *table.Table + var err error + var v []*table.Validation + switch arg.Type { + case api.Resource_LOCAL, api.Resource_GLOBAL: + tbl, v, err = s.bgpServer.GetRib(arg.Name, family, f()) + case api.Resource_ADJ_IN: + in = true + fallthrough + case api.Resource_ADJ_OUT: + tbl, v, err = s.bgpServer.GetAdjRib(arg.Name, family, in, f()) + case api.Resource_VRF: + tbl, err = s.bgpServer.GetVrfRib(arg.Name, family, []*table.LookupPrefix{}) + default: + return fmt.Errorf("unsupported resource type: %v", arg.Type) + } + if err != nil { + return err + } + + idx := 0 + return func() error { + for _, dst := range tbl.GetDestinations() { + for i, path := range dst.GetAllKnownPathList() { + p := ToPathApi(path, getValidation(v, idx)) + idx++ + if i == 0 && !table.SelectionOptions.DisableBestPathSelection { + switch arg.Type { + case api.Resource_LOCAL, api.Resource_GLOBAL: + p.Best = true + } + } + if err := stream.Send(p); err != nil { + return err + } + } + } + return nil + }() +} + +func (s *Server) MonitorRib(arg *api.MonitorRibRequest, stream api.GobgpApi_MonitorRibServer) error { + if arg == nil || arg.Table == nil { + return fmt.Errorf("invalid request") + } + t := arg.Table + w, err := func() (*Watcher, error) { + switch t.Type { + case api.Resource_GLOBAL: + return s.bgpServer.Watch(WatchBestPath(arg.Current)), nil + case api.Resource_ADJ_IN: + if t.PostPolicy { + return s.bgpServer.Watch(WatchPostUpdate(arg.Current)), nil + } + return s.bgpServer.Watch(WatchUpdate(arg.Current)), nil + default: + return nil, fmt.Errorf("unsupported resource type: %v", t.Type) + } + }() + if err != nil { + return nil + } + + return func() error { + defer func() { w.Stop() }() + + sendPath := func(pathList []*table.Path) error { + dsts := make(map[string]*api.Destination) + for _, path := range pathList { + if path == nil || (t.Family != 0 && bgp.RouteFamily(t.Family) != path.GetRouteFamily()) { + continue + } + if dst, y := dsts[path.GetNlri().String()]; y { + dst.Paths = append(dst.Paths, ToPathApi(path, nil)) + } else { + dsts[path.GetNlri().String()] = &api.Destination{ + Prefix: path.GetNlri().String(), + Paths: []*api.Path{ToPathApi(path, nil)}, + } + } + } + for _, dst := range dsts { + if err := stream.Send(dst); err != nil { + return err + } + } + return nil + } + + for ev := range w.Event() { + switch msg := ev.(type) { + case *WatchEventBestPath: + if err := sendPath(func() []*table.Path { + if len(msg.MultiPathList) > 0 { + l := make([]*table.Path, 0) + for _, p := range msg.MultiPathList { + l = append(l, p...) + } + return l + } else { + return msg.PathList + } + }()); err != nil { + return err + } + case *WatchEventUpdate: + if err := sendPath(msg.PathList); err != nil { + return err + } + } + } + return nil + }() +} + +func (s *Server) MonitorPeerState(arg *api.Arguments, stream api.GobgpApi_MonitorPeerStateServer) error { + if arg == nil { + return fmt.Errorf("invalid request") + } + return func() error { + w := s.bgpServer.Watch(WatchPeerState(arg.Current)) + defer func() { w.Stop() }() + + for ev := range w.Event() { + switch msg := ev.(type) { + case *WatchEventPeerState: + if len(arg.Name) > 0 && arg.Name != msg.PeerAddress.String() && arg.Name != msg.PeerInterface { + continue + } + if err := stream.Send(&api.Peer{ + Conf: &api.PeerConf{ + PeerAs: msg.PeerAS, + LocalAs: msg.LocalAS, + NeighborAddress: msg.PeerAddress.String(), + Id: msg.PeerID.String(), + NeighborInterface: msg.PeerInterface, + }, + Info: &api.PeerState{ + PeerAs: msg.PeerAS, + LocalAs: msg.LocalAS, + NeighborAddress: msg.PeerAddress.String(), + BgpState: msg.State.String(), + AdminState: api.PeerState_AdminState(msg.AdminState), + }, + Transport: &api.Transport{ + LocalAddress: msg.LocalAddress.String(), + LocalPort: uint32(msg.LocalPort), + RemotePort: uint32(msg.PeerPort), + }, + }); err != nil { + return err + } + } + } + return nil + }() +} + +func (s *Server) ResetNeighbor(ctx context.Context, arg *api.ResetNeighborRequest) (*api.ResetNeighborResponse, error) { + return &api.ResetNeighborResponse{}, s.bgpServer.ResetNeighbor(arg.Address, arg.Communication) +} + +func (s *Server) SoftResetNeighbor(ctx context.Context, arg *api.SoftResetNeighborRequest) (*api.SoftResetNeighborResponse, error) { + var err error + addr := arg.Address + if addr == "all" { + addr = "" + } + family := bgp.RouteFamily(0) + switch arg.Direction { + case api.SoftResetNeighborRequest_IN: + err = s.bgpServer.SoftResetIn(addr, family) + case api.SoftResetNeighborRequest_OUT: + err = s.bgpServer.SoftResetOut(addr, family) + default: + err = s.bgpServer.SoftReset(addr, family) + } + return &api.SoftResetNeighborResponse{}, err +} + +func (s *Server) ShutdownNeighbor(ctx context.Context, arg *api.ShutdownNeighborRequest) (*api.ShutdownNeighborResponse, error) { + return &api.ShutdownNeighborResponse{}, s.bgpServer.ShutdownNeighbor(arg.Address, arg.Communication) +} + +func (s *Server) EnableNeighbor(ctx context.Context, arg *api.EnableNeighborRequest) (*api.EnableNeighborResponse, error) { + return &api.EnableNeighborResponse{}, s.bgpServer.EnableNeighbor(arg.Address) +} + +func (s *Server) DisableNeighbor(ctx context.Context, arg *api.DisableNeighborRequest) (*api.DisableNeighborResponse, error) { + return &api.DisableNeighborResponse{}, s.bgpServer.DisableNeighbor(arg.Address, arg.Communication) +} + +func (s *Server) UpdatePolicy(ctx context.Context, arg *api.UpdatePolicyRequest) (*api.UpdatePolicyResponse, error) { + rp, err := NewRoutingPolicyFromApiStruct(arg) + if err != nil { + return nil, err + } + return &api.UpdatePolicyResponse{}, s.bgpServer.UpdatePolicy(*rp) +} + +func NewAPIRoutingPolicyFromConfigStruct(c *config.RoutingPolicy) (*api.RoutingPolicy, error) { + definedSets, err := NewAPIDefinedSetsFromConfigStruct(&c.DefinedSets) + if err != nil { + return nil, err + } + policies := make([]*api.Policy, 0, len(c.PolicyDefinitions)) + for _, policy := range c.PolicyDefinitions { + policies = append(policies, toPolicyApi(&policy)) + } + + return &api.RoutingPolicy{ + DefinedSet: definedSets, + PolicyDefinition: policies, + }, nil +} + +func NewRoutingPolicyFromApiStruct(arg *api.UpdatePolicyRequest) (*config.RoutingPolicy, error) { + policyDefinitions := make([]config.PolicyDefinition, 0, len(arg.Policies)) + for _, p := range arg.Policies { + pd, err := NewConfigPolicyFromApiStruct(p) + if err != nil { + return nil, err + } + policyDefinitions = append(policyDefinitions, *pd) + } + + definedSets, err := NewConfigDefinedSetsFromApiStruct(arg.Sets) + if err != nil { + return nil, err + } + + return &config.RoutingPolicy{ + DefinedSets: *definedSets, + PolicyDefinitions: policyDefinitions, + }, nil +} + +func (s *Server) api2PathList(resource api.Resource, ApiPathList []*api.Path) ([]*table.Path, error) { + var pi *table.PeerInfo + + pathList := make([]*table.Path, 0, len(ApiPathList)) + for _, path := range ApiPathList { + var nlri bgp.AddrPrefixInterface + var nexthop string + + if path.SourceAsn != 0 { + pi = &table.PeerInfo{ + AS: path.SourceAsn, + LocalID: net.ParseIP(path.SourceId), + } + } + + nlri, err := path.GetNativeNlri() + if err != nil { + return nil, err + } + nlri.SetPathIdentifier(path.Identifier) + + attrList, err := path.GetNativePathAttributes() + if err != nil { + return nil, err + } + + pattrs := make([]bgp.PathAttributeInterface, 0) + seen := make(map[bgp.BGPAttrType]struct{}) + for _, attr := range attrList { + attrType := attr.GetType() + if _, ok := seen[attrType]; !ok { + seen[attrType] = struct{}{} + } else { + return nil, fmt.Errorf("duplicated path attribute type: %d", attrType) + } + + switch a := attr.(type) { + case *bgp.PathAttributeNextHop: + nexthop = a.Value.String() + case *bgp.PathAttributeMpReachNLRI: + nlri = a.Value[0] + nexthop = a.Nexthop.String() + default: + pattrs = append(pattrs, attr) + } + } + + if nlri == nil { + return nil, fmt.Errorf("nlri not found") + } else if !path.IsWithdraw && nexthop == "" { + return nil, fmt.Errorf("nexthop not found") + } + + if resource != api.Resource_VRF && bgp.RouteFamily(path.Family) == bgp.RF_IPv4_UC && net.ParseIP(nexthop).To4() != nil { + pattrs = append(pattrs, bgp.NewPathAttributeNextHop(nexthop)) + } else { + pattrs = append(pattrs, bgp.NewPathAttributeMpReachNLRI(nexthop, []bgp.AddrPrefixInterface{nlri})) + } + + newPath := table.NewPath(pi, nlri, path.IsWithdraw, pattrs, time.Now(), path.NoImplicitWithdraw) + if !path.IsWithdraw { + total := bytes.NewBuffer(make([]byte, 0)) + for _, a := range newPath.GetPathAttrs() { + if a.GetType() == bgp.BGP_ATTR_TYPE_MP_REACH_NLRI { + continue + } + b, _ := a.Serialize() + total.Write(b) + } + newPath.SetHash(farm.Hash32(total.Bytes())) + } + newPath.SetIsFromExternal(path.IsFromExternal) + pathList = append(pathList, newPath) + } + return pathList, nil +} + +func (s *Server) AddPath(ctx context.Context, arg *api.AddPathRequest) (*api.AddPathResponse, error) { + pathList, err := s.api2PathList(arg.Resource, []*api.Path{arg.Path}) + var uuid []byte + if err == nil { + uuid, err = s.bgpServer.AddPath(arg.VrfId, pathList) + } + return &api.AddPathResponse{Uuid: uuid}, err +} + +func (s *Server) DeletePath(ctx context.Context, arg *api.DeletePathRequest) (*api.DeletePathResponse, error) { + pathList, err := func() ([]*table.Path, error) { + if arg.Path != nil { + arg.Path.IsWithdraw = true + return s.api2PathList(arg.Resource, []*api.Path{arg.Path}) + } + return []*table.Path{}, nil + }() + if err != nil { + return nil, err + } + return &api.DeletePathResponse{}, s.bgpServer.DeletePath(arg.Uuid, bgp.RouteFamily(arg.Family), arg.VrfId, pathList) +} + +func (s *Server) EnableMrt(ctx context.Context, arg *api.EnableMrtRequest) (*api.EnableMrtResponse, error) { + return &api.EnableMrtResponse{}, s.bgpServer.EnableMrt(&config.MrtConfig{ + RotationInterval: arg.Interval, + DumpType: config.IntToMrtTypeMap[int(arg.DumpType)], + FileName: arg.Filename, + }) +} + +func (s *Server) DisableMrt(ctx context.Context, arg *api.DisableMrtRequest) (*api.DisableMrtResponse, error) { + return &api.DisableMrtResponse{}, s.bgpServer.DisableMrt(&config.MrtConfig{}) +} + +func (s *Server) InjectMrt(stream api.GobgpApi_InjectMrtServer) error { + for { + arg, err := stream.Recv() + + if err == io.EOF { + break + } else if err != nil { + return err + } + + if arg.Resource != api.Resource_GLOBAL && arg.Resource != api.Resource_VRF { + return fmt.Errorf("unsupported resource: %s", arg.Resource) + } + + if pathList, err := s.api2PathList(arg.Resource, arg.Paths); err != nil { + return err + } else { + if _, err = s.bgpServer.AddPath("", pathList); err != nil { + return err + } + } + } + return stream.SendAndClose(&api.InjectMrtResponse{}) +} + +func (s *Server) AddBmp(ctx context.Context, arg *api.AddBmpRequest) (*api.AddBmpResponse, error) { + t, ok := config.IntToBmpRouteMonitoringPolicyTypeMap[int(arg.Type)] + if !ok { + return nil, fmt.Errorf("invalid bmp route monitoring policy: %d", arg.Type) + } + return &api.AddBmpResponse{}, s.bgpServer.AddBmp(&config.BmpServerConfig{ + Address: arg.Address, + Port: arg.Port, + RouteMonitoringPolicy: t, + }) +} + +func (s *Server) DeleteBmp(ctx context.Context, arg *api.DeleteBmpRequest) (*api.DeleteBmpResponse, error) { + return &api.DeleteBmpResponse{}, s.bgpServer.DeleteBmp(&config.BmpServerConfig{ + Address: arg.Address, + Port: arg.Port, + }) +} + +func (s *Server) ValidateRib(ctx context.Context, arg *api.ValidateRibRequest) (*api.ValidateRibResponse, error) { + return &api.ValidateRibResponse{}, nil +} + +func (s *Server) AddRpki(ctx context.Context, arg *api.AddRpkiRequest) (*api.AddRpkiResponse, error) { + return &api.AddRpkiResponse{}, s.bgpServer.AddRpki(&config.RpkiServerConfig{ + Address: arg.Address, + Port: arg.Port, + RecordLifetime: arg.Lifetime, + }) +} + +func (s *Server) DeleteRpki(ctx context.Context, arg *api.DeleteRpkiRequest) (*api.DeleteRpkiResponse, error) { + return &api.DeleteRpkiResponse{}, s.bgpServer.DeleteRpki(&config.RpkiServerConfig{ + Address: arg.Address, + Port: arg.Port, + }) +} + +func (s *Server) EnableRpki(ctx context.Context, arg *api.EnableRpkiRequest) (*api.EnableRpkiResponse, error) { + return &api.EnableRpkiResponse{}, s.bgpServer.EnableRpki(&config.RpkiServerConfig{ + Address: arg.Address, + }) +} + +func (s *Server) DisableRpki(ctx context.Context, arg *api.DisableRpkiRequest) (*api.DisableRpkiResponse, error) { + return &api.DisableRpkiResponse{}, s.bgpServer.DisableRpki(&config.RpkiServerConfig{ + Address: arg.Address, + }) +} + +func (s *Server) ResetRpki(ctx context.Context, arg *api.ResetRpkiRequest) (*api.ResetRpkiResponse, error) { + return &api.ResetRpkiResponse{}, s.bgpServer.ResetRpki(&config.RpkiServerConfig{ + Address: arg.Address, + }) +} + +func (s *Server) SoftResetRpki(ctx context.Context, arg *api.SoftResetRpkiRequest) (*api.SoftResetRpkiResponse, error) { + return &api.SoftResetRpkiResponse{}, s.bgpServer.SoftResetRpki(&config.RpkiServerConfig{ + Address: arg.Address, + }) +} + +func (s *Server) GetRpki(ctx context.Context, arg *api.GetRpkiRequest) (*api.GetRpkiResponse, error) { + servers, err := s.bgpServer.GetRpki() + if err != nil { + return nil, err + } + l := make([]*api.Rpki, 0, len(servers)) + for _, s := range servers { + received := &s.State.RpkiMessages.RpkiReceived + sent := &s.State.RpkiMessages.RpkiSent + rpki := &api.Rpki{ + Conf: &api.RPKIConf{ + Address: s.Config.Address, + RemotePort: strconv.Itoa(int(s.Config.Port)), + }, + State: &api.RPKIState{ + Uptime: s.State.Uptime, + Downtime: s.State.Downtime, + Up: s.State.Up, + RecordIpv4: s.State.RecordsV4, + RecordIpv6: s.State.RecordsV6, + PrefixIpv4: s.State.PrefixesV4, + PrefixIpv6: s.State.PrefixesV6, + Serial: s.State.SerialNumber, + ReceivedIpv4: received.Ipv4Prefix, + ReceivedIpv6: received.Ipv6Prefix, + SerialNotify: received.SerialNotify, + CacheReset: received.CacheReset, + CacheResponse: received.CacheResponse, + EndOfData: received.EndOfData, + Error: received.Error, + SerialQuery: sent.SerialQuery, + ResetQuery: sent.ResetQuery, + }, + } + l = append(l, rpki) + } + return &api.GetRpkiResponse{Servers: l}, nil +} + +func (s *Server) GetRoa(ctx context.Context, arg *api.GetRoaRequest) (*api.GetRoaResponse, error) { + roas, err := s.bgpServer.GetRoa(bgp.RouteFamily(arg.Family)) + if err != nil { + return nil, err + } + return &api.GetRoaResponse{Roas: NewRoaListFromTableStructList(roas)}, nil +} + +func (s *Server) EnableZebra(ctx context.Context, arg *api.EnableZebraRequest) (*api.EnableZebraResponse, error) { + for _, p := range arg.RouteTypes { + if _, err := zebra.RouteTypeFromString(p); err != nil { + return &api.EnableZebraResponse{}, err + } + } + return &api.EnableZebraResponse{}, s.bgpServer.StartZebraClient(&config.ZebraConfig{ + Url: arg.Url, + RedistributeRouteTypeList: arg.RouteTypes, + Version: uint8(arg.Version), + NexthopTriggerEnable: arg.NexthopTriggerEnable, + NexthopTriggerDelay: uint8(arg.NexthopTriggerDelay), + }) +} + +func (s *Server) GetVrf(ctx context.Context, arg *api.GetVrfRequest) (*api.GetVrfResponse, error) { + toApi := func(v *table.Vrf) *api.Vrf { + return &api.Vrf{ + Name: v.Name, + Rd: api.MarshalRD(v.Rd), + Id: v.Id, + ImportRt: api.MarshalRTs(v.ImportRt), + ExportRt: api.MarshalRTs(v.ExportRt), + } + } + vrfs := s.bgpServer.GetVrf() + l := make([]*api.Vrf, 0, len(vrfs)) + for _, v := range vrfs { + l = append(l, toApi(v)) + } + return &api.GetVrfResponse{Vrfs: l}, nil +} + +func (s *Server) AddVrf(ctx context.Context, arg *api.AddVrfRequest) (r *api.AddVrfResponse, err error) { + if arg == nil || arg.Vrf == nil { + return nil, fmt.Errorf("invalid request") + } + rd, err := api.UnmarshalRD(arg.Vrf.Rd) + if err != nil { + return nil, err + } + im, err := api.UnmarshalRTs(arg.Vrf.ImportRt) + if err != nil { + return nil, err + } + ex, err := api.UnmarshalRTs(arg.Vrf.ExportRt) + if err != nil { + return nil, err + } + return &api.AddVrfResponse{}, s.bgpServer.AddVrf(arg.Vrf.Name, arg.Vrf.Id, rd, im, ex) +} + +func (s *Server) DeleteVrf(ctx context.Context, arg *api.DeleteVrfRequest) (*api.DeleteVrfResponse, error) { + if arg == nil || arg.Vrf == nil { + return nil, fmt.Errorf("invalid request") + } + return &api.DeleteVrfResponse{}, s.bgpServer.DeleteVrf(arg.Vrf.Name) +} + +func ReadMpGracefulRestartFromAPIStruct(c *config.MpGracefulRestart, a *api.MpGracefulRestart) { + if c == nil || a == nil { + return + } + if a.Config != nil { + c.Config.Enabled = a.Config.Enabled + } +} + +func ReadAfiSafiConfigFromAPIStruct(c *config.AfiSafiConfig, a *api.AfiSafiConfig) { + if c == nil || a == nil { + return + } + c.AfiSafiName = config.AfiSafiType(bgp.RouteFamily(a.Family).String()) + c.Enabled = a.Enabled +} + +func ReadAfiSafiStateFromAPIStruct(s *config.AfiSafiState, a *api.AfiSafiConfig) { + if s == nil || a == nil { + return + } + // Store only address family value for the convenience + s.Family = bgp.RouteFamily(a.Family) +} + +func ReadPrefixLimitFromAPIStruct(c *config.PrefixLimit, a *api.PrefixLimit) { + if c == nil || a == nil { + return + } + c.Config.MaxPrefixes = a.MaxPrefixes + c.Config.ShutdownThresholdPct = config.Percentage(a.ShutdownThresholdPct) +} + +func ReadApplyPolicyFromAPIStruct(c *config.ApplyPolicy, a *api.ApplyPolicy) { + if c == nil || a == nil { + return + } + if a.ImportPolicy != nil { + c.Config.DefaultImportPolicy = config.IntToDefaultPolicyTypeMap[int(a.ImportPolicy.Default)] + for _, p := range a.ImportPolicy.Policies { + c.Config.ImportPolicyList = append(c.Config.ImportPolicyList, p.Name) + } + } + if a.ExportPolicy != nil { + c.Config.DefaultExportPolicy = config.IntToDefaultPolicyTypeMap[int(a.ExportPolicy.Default)] + for _, p := range a.ExportPolicy.Policies { + c.Config.ExportPolicyList = append(c.Config.ExportPolicyList, p.Name) + } + } + if a.InPolicy != nil { + c.Config.DefaultInPolicy = config.IntToDefaultPolicyTypeMap[int(a.InPolicy.Default)] + for _, p := range a.InPolicy.Policies { + c.Config.InPolicyList = append(c.Config.InPolicyList, p.Name) + } + } +} + +func ReadRouteSelectionOptionsFromAPIStruct(c *config.RouteSelectionOptions, a *api.RouteSelectionOptions) { + if c == nil || a == nil { + return + } + if a.Config != nil { + c.Config.AlwaysCompareMed = a.Config.AlwaysCompareMed + c.Config.IgnoreAsPathLength = a.Config.IgnoreAsPathLength + c.Config.ExternalCompareRouterId = a.Config.ExternalCompareRouterId + c.Config.AdvertiseInactiveRoutes = a.Config.AdvertiseInactiveRoutes + c.Config.EnableAigp = a.Config.EnableAigp + c.Config.IgnoreNextHopIgpMetric = a.Config.IgnoreNextHopIgpMetric + } +} + +func ReadUseMultiplePathsFromAPIStruct(c *config.UseMultiplePaths, a *api.UseMultiplePaths) { + if c == nil || a == nil { + return + } + if a.Config != nil { + c.Config.Enabled = a.Config.Enabled + } + if a.Ebgp != nil && a.Ebgp.Config != nil { + c.Ebgp = config.Ebgp{ + Config: config.EbgpConfig{ + AllowMultipleAs: a.Ebgp.Config.AllowMultipleAs, + MaximumPaths: a.Ebgp.Config.MaximumPaths, + }, + } + } + if a.Ibgp != nil && a.Ibgp.Config != nil { + c.Ibgp = config.Ibgp{ + Config: config.IbgpConfig{ + MaximumPaths: a.Ibgp.Config.MaximumPaths, + }, + } + } +} + +func ReadRouteTargetMembershipFromAPIStruct(c *config.RouteTargetMembership, a *api.RouteTargetMembership) { + if c == nil || a == nil { + return + } + if a.Config != nil { + c.Config.DeferralTime = uint16(a.Config.DeferralTime) + } +} + +func ReadLongLivedGracefulRestartFromAPIStruct(c *config.LongLivedGracefulRestart, a *api.LongLivedGracefulRestart) { + if c == nil || a == nil { + return + } + if a.Config != nil { + c.Config.Enabled = a.Config.Enabled + c.Config.RestartTime = a.Config.RestartTime + } +} + +func ReadAddPathsFromAPIStruct(c *config.AddPaths, a *api.AddPaths) { + if c == nil || a == nil { + return + } + if a.Config != nil { + c.Config.Receive = a.Config.Receive + c.Config.SendMax = uint8(a.Config.SendMax) + } +} + +func NewNeighborFromAPIStruct(a *api.Peer) (*config.Neighbor, error) { + pconf := &config.Neighbor{} + if a.Conf != nil { + pconf.Config.PeerAs = a.Conf.PeerAs + pconf.Config.LocalAs = a.Conf.LocalAs + pconf.Config.AuthPassword = a.Conf.AuthPassword + pconf.Config.RouteFlapDamping = a.Conf.RouteFlapDamping + pconf.Config.Description = a.Conf.Description + pconf.Config.PeerGroup = a.Conf.PeerGroup + pconf.Config.PeerType = config.IntToPeerTypeMap[int(a.Conf.PeerType)] + pconf.Config.NeighborAddress = a.Conf.NeighborAddress + pconf.Config.NeighborInterface = a.Conf.NeighborInterface + pconf.Config.Vrf = a.Conf.Vrf + pconf.AsPathOptions.Config.AllowOwnAs = uint8(a.Conf.AllowOwnAs) + pconf.AsPathOptions.Config.ReplacePeerAs = a.Conf.ReplacePeerAs + + switch a.Conf.RemovePrivateAs { + case api.PeerConf_ALL: + pconf.Config.RemovePrivateAs = config.REMOVE_PRIVATE_AS_OPTION_ALL + case api.PeerConf_REPLACE: + pconf.Config.RemovePrivateAs = config.REMOVE_PRIVATE_AS_OPTION_REPLACE + } + + localCaps, err := api.UnmarshalCapabilities(a.Conf.LocalCap) + if err != nil { + return nil, err + } + remoteCaps, err := api.UnmarshalCapabilities(a.Conf.RemoteCap) + if err != nil { + return nil, err + } + pconf.State.LocalCapabilityList = localCaps + pconf.State.RemoteCapabilityList = remoteCaps + + pconf.State.RemoteRouterId = a.Conf.Id + + for _, af := range a.AfiSafis { + afiSafi := config.AfiSafi{} + ReadMpGracefulRestartFromAPIStruct(&afiSafi.MpGracefulRestart, af.MpGracefulRestart) + ReadAfiSafiConfigFromAPIStruct(&afiSafi.Config, af.Config) + ReadAfiSafiStateFromAPIStruct(&afiSafi.State, af.Config) + ReadApplyPolicyFromAPIStruct(&afiSafi.ApplyPolicy, af.ApplyPolicy) + ReadRouteSelectionOptionsFromAPIStruct(&afiSafi.RouteSelectionOptions, af.RouteSelectionOptions) + ReadUseMultiplePathsFromAPIStruct(&afiSafi.UseMultiplePaths, af.UseMultiplePaths) + ReadPrefixLimitFromAPIStruct(&afiSafi.PrefixLimit, af.PrefixLimits) + ReadRouteTargetMembershipFromAPIStruct(&afiSafi.RouteTargetMembership, af.RouteTargetMembership) + ReadLongLivedGracefulRestartFromAPIStruct(&afiSafi.LongLivedGracefulRestart, af.LongLivedGracefulRestart) + ReadAddPathsFromAPIStruct(&afiSafi.AddPaths, af.AddPaths) + pconf.AfiSafis = append(pconf.AfiSafis, afiSafi) + } + // For the backward compatibility, we override AfiSafi configurations + // with Peer.Families. + for _, family := range a.Families { + found := false + for _, afiSafi := range pconf.AfiSafis { + if uint32(afiSafi.State.Family) == family { + // If Peer.Families contains the same address family, + // we enable this address family. + afiSafi.Config.Enabled = true + found = true + } + } + if !found { + // If Peer.Families does not contain the same address family, + // we append AfiSafi structure with the default value. + pconf.AfiSafis = append(pconf.AfiSafis, config.AfiSafi{ + Config: config.AfiSafiConfig{ + AfiSafiName: config.AfiSafiType(bgp.RouteFamily(family).String()), + Enabled: true, + }, + }) + } + } + // For the backward compatibility, we override AfiSafi configurations + // with Peer.Conf.PrefixLimits. + for _, prefixLimit := range a.Conf.PrefixLimits { + for _, afiSafi := range pconf.AfiSafis { + // If Peer.Conf.PrefixLimits contains the configuration for + // the same address family, we override AfiSafi.PrefixLimit. + if uint32(afiSafi.State.Family) == prefixLimit.Family { + ReadPrefixLimitFromAPIStruct(&afiSafi.PrefixLimit, prefixLimit) + } + } + } + } + + if a.Timers != nil { + if a.Timers.Config != nil { + pconf.Timers.Config.ConnectRetry = float64(a.Timers.Config.ConnectRetry) + pconf.Timers.Config.HoldTime = float64(a.Timers.Config.HoldTime) + pconf.Timers.Config.KeepaliveInterval = float64(a.Timers.Config.KeepaliveInterval) + pconf.Timers.Config.MinimumAdvertisementInterval = float64(a.Timers.Config.MinimumAdvertisementInterval) + } + if a.Timers.State != nil { + pconf.Timers.State.KeepaliveInterval = float64(a.Timers.State.KeepaliveInterval) + pconf.Timers.State.NegotiatedHoldTime = float64(a.Timers.State.NegotiatedHoldTime) + pconf.Timers.State.Uptime = int64(a.Timers.State.Uptime) + pconf.Timers.State.Downtime = int64(a.Timers.State.Downtime) + } + } + if a.RouteReflector != nil { + pconf.RouteReflector.Config.RouteReflectorClusterId = config.RrClusterIdType(a.RouteReflector.RouteReflectorClusterId) + pconf.RouteReflector.Config.RouteReflectorClient = a.RouteReflector.RouteReflectorClient + } + if a.RouteServer != nil { + pconf.RouteServer.Config.RouteServerClient = a.RouteServer.RouteServerClient + } + if a.GracefulRestart != nil { + pconf.GracefulRestart.Config.Enabled = a.GracefulRestart.Enabled + pconf.GracefulRestart.Config.RestartTime = uint16(a.GracefulRestart.RestartTime) + pconf.GracefulRestart.Config.HelperOnly = a.GracefulRestart.HelperOnly + pconf.GracefulRestart.Config.DeferralTime = uint16(a.GracefulRestart.DeferralTime) + pconf.GracefulRestart.Config.NotificationEnabled = a.GracefulRestart.NotificationEnabled + pconf.GracefulRestart.Config.LongLivedEnabled = a.GracefulRestart.LonglivedEnabled + pconf.GracefulRestart.State.LocalRestarting = a.GracefulRestart.LocalRestarting + } + ReadApplyPolicyFromAPIStruct(&pconf.ApplyPolicy, a.ApplyPolicy) + if a.Transport != nil { + pconf.Transport.Config.LocalAddress = a.Transport.LocalAddress + pconf.Transport.Config.PassiveMode = a.Transport.PassiveMode + pconf.Transport.Config.RemotePort = uint16(a.Transport.RemotePort) + } + if a.EbgpMultihop != nil { + pconf.EbgpMultihop.Config.Enabled = a.EbgpMultihop.Enabled + pconf.EbgpMultihop.Config.MultihopTtl = uint8(a.EbgpMultihop.MultihopTtl) + } + if a.Info != nil { + pconf.State.SessionState = config.SessionState(a.Info.BgpState) + pconf.State.AdminState = config.IntToAdminStateMap[int(a.Info.AdminState)] + + pconf.State.AdjTable.Received = a.Info.Received + pconf.State.AdjTable.Accepted = a.Info.Accepted + pconf.State.AdjTable.Advertised = a.Info.Advertised + pconf.State.PeerAs = a.Info.PeerAs + pconf.State.PeerType = config.IntToPeerTypeMap[int(a.Info.PeerType)] + pconf.State.NeighborAddress = a.Info.NeighborAddress + + if a.Info.Messages != nil { + if a.Info.Messages.Sent != nil { + pconf.State.Messages.Sent.Update = a.Info.Messages.Sent.UPDATE + pconf.State.Messages.Sent.Notification = a.Info.Messages.Sent.NOTIFICATION + pconf.State.Messages.Sent.Open = a.Info.Messages.Sent.OPEN + pconf.State.Messages.Sent.Refresh = a.Info.Messages.Sent.REFRESH + pconf.State.Messages.Sent.Keepalive = a.Info.Messages.Sent.KEEPALIVE + pconf.State.Messages.Sent.Discarded = a.Info.Messages.Sent.DISCARDED + pconf.State.Messages.Sent.Total = a.Info.Messages.Sent.TOTAL + } + if a.Info.Messages.Received != nil { + pconf.State.Messages.Received.Update = a.Info.Messages.Received.UPDATE + pconf.State.Messages.Received.Open = a.Info.Messages.Received.OPEN + pconf.State.Messages.Received.Refresh = a.Info.Messages.Received.REFRESH + pconf.State.Messages.Received.Keepalive = a.Info.Messages.Received.KEEPALIVE + pconf.State.Messages.Received.Discarded = a.Info.Messages.Received.DISCARDED + pconf.State.Messages.Received.Total = a.Info.Messages.Received.TOTAL + } + } + } + ReadAddPathsFromAPIStruct(&pconf.AddPaths, a.AddPaths) + return pconf, nil +} + +func NewPeerGroupFromAPIStruct(a *api.PeerGroup) (*config.PeerGroup, error) { + pconf := &config.PeerGroup{} + if a.Conf != nil { + pconf.Config.PeerAs = a.Conf.PeerAs + pconf.Config.LocalAs = a.Conf.LocalAs + pconf.Config.AuthPassword = a.Conf.AuthPassword + pconf.Config.RouteFlapDamping = a.Conf.RouteFlapDamping + pconf.Config.Description = a.Conf.Description + pconf.Config.PeerGroupName = a.Conf.PeerGroupName + + switch a.Conf.RemovePrivateAs { + case api.PeerGroupConf_ALL: + pconf.Config.RemovePrivateAs = config.REMOVE_PRIVATE_AS_OPTION_ALL + case api.PeerGroupConf_REPLACE: + pconf.Config.RemovePrivateAs = config.REMOVE_PRIVATE_AS_OPTION_REPLACE + } + + for _, af := range a.AfiSafis { + afiSafi := config.AfiSafi{} + ReadMpGracefulRestartFromAPIStruct(&afiSafi.MpGracefulRestart, af.MpGracefulRestart) + ReadAfiSafiConfigFromAPIStruct(&afiSafi.Config, af.Config) + ReadAfiSafiStateFromAPIStruct(&afiSafi.State, af.Config) + ReadApplyPolicyFromAPIStruct(&afiSafi.ApplyPolicy, af.ApplyPolicy) + ReadRouteSelectionOptionsFromAPIStruct(&afiSafi.RouteSelectionOptions, af.RouteSelectionOptions) + ReadUseMultiplePathsFromAPIStruct(&afiSafi.UseMultiplePaths, af.UseMultiplePaths) + ReadPrefixLimitFromAPIStruct(&afiSafi.PrefixLimit, af.PrefixLimits) + ReadRouteTargetMembershipFromAPIStruct(&afiSafi.RouteTargetMembership, af.RouteTargetMembership) + ReadLongLivedGracefulRestartFromAPIStruct(&afiSafi.LongLivedGracefulRestart, af.LongLivedGracefulRestart) + ReadAddPathsFromAPIStruct(&afiSafi.AddPaths, af.AddPaths) + pconf.AfiSafis = append(pconf.AfiSafis, afiSafi) + } + // For the backward compatibility, we override AfiSafi configurations + // with Peer.Families. + for _, family := range a.Families { + found := false + for _, afiSafi := range pconf.AfiSafis { + if uint32(afiSafi.State.Family) == family { + // If Peer.Families contains the same address family, + // we enable this address family. + afiSafi.Config.Enabled = true + found = true + } + } + if !found { + // If Peer.Families does not contain the same address family, + // we append AfiSafi structure with the default value. + pconf.AfiSafis = append(pconf.AfiSafis, config.AfiSafi{ + Config: config.AfiSafiConfig{ + AfiSafiName: config.AfiSafiType(bgp.RouteFamily(family).String()), + Enabled: true, + }, + }) + } + } + } + + if a.Timers != nil { + if a.Timers.Config != nil { + pconf.Timers.Config.ConnectRetry = float64(a.Timers.Config.ConnectRetry) + pconf.Timers.Config.HoldTime = float64(a.Timers.Config.HoldTime) + pconf.Timers.Config.KeepaliveInterval = float64(a.Timers.Config.KeepaliveInterval) + pconf.Timers.Config.MinimumAdvertisementInterval = float64(a.Timers.Config.MinimumAdvertisementInterval) + } + if a.Timers.State != nil { + pconf.Timers.State.KeepaliveInterval = float64(a.Timers.State.KeepaliveInterval) + pconf.Timers.State.NegotiatedHoldTime = float64(a.Timers.State.NegotiatedHoldTime) + pconf.Timers.State.Uptime = int64(a.Timers.State.Uptime) + pconf.Timers.State.Downtime = int64(a.Timers.State.Downtime) + } + } + if a.RouteReflector != nil { + pconf.RouteReflector.Config.RouteReflectorClusterId = config.RrClusterIdType(a.RouteReflector.RouteReflectorClusterId) + pconf.RouteReflector.Config.RouteReflectorClient = a.RouteReflector.RouteReflectorClient + } + if a.RouteServer != nil { + pconf.RouteServer.Config.RouteServerClient = a.RouteServer.RouteServerClient + } + if a.GracefulRestart != nil { + pconf.GracefulRestart.Config.Enabled = a.GracefulRestart.Enabled + pconf.GracefulRestart.Config.RestartTime = uint16(a.GracefulRestart.RestartTime) + pconf.GracefulRestart.Config.HelperOnly = a.GracefulRestart.HelperOnly + pconf.GracefulRestart.Config.DeferralTime = uint16(a.GracefulRestart.DeferralTime) + pconf.GracefulRestart.Config.NotificationEnabled = a.GracefulRestart.NotificationEnabled + pconf.GracefulRestart.Config.LongLivedEnabled = a.GracefulRestart.LonglivedEnabled + pconf.GracefulRestart.State.LocalRestarting = a.GracefulRestart.LocalRestarting + } + ReadApplyPolicyFromAPIStruct(&pconf.ApplyPolicy, a.ApplyPolicy) + if a.Transport != nil { + pconf.Transport.Config.LocalAddress = a.Transport.LocalAddress + pconf.Transport.Config.PassiveMode = a.Transport.PassiveMode + pconf.Transport.Config.RemotePort = uint16(a.Transport.RemotePort) + } + if a.EbgpMultihop != nil { + pconf.EbgpMultihop.Config.Enabled = a.EbgpMultihop.Enabled + pconf.EbgpMultihop.Config.MultihopTtl = uint8(a.EbgpMultihop.MultihopTtl) + } + if a.Info != nil { + pconf.State.TotalPaths = a.Info.TotalPaths + pconf.State.TotalPrefixes = a.Info.TotalPrefixes + pconf.State.PeerAs = a.Info.PeerAs + pconf.State.PeerType = config.IntToPeerTypeMap[int(a.Info.PeerType)] + } + ReadAddPathsFromAPIStruct(&pconf.AddPaths, a.AddPaths) + return pconf, nil +} + +func (s *Server) AddNeighbor(ctx context.Context, arg *api.AddNeighborRequest) (*api.AddNeighborResponse, error) { + c, err := NewNeighborFromAPIStruct(arg.Peer) + if err != nil { + return nil, err + } + return &api.AddNeighborResponse{}, s.bgpServer.AddNeighbor(c) +} + +func (s *Server) DeleteNeighbor(ctx context.Context, arg *api.DeleteNeighborRequest) (*api.DeleteNeighborResponse, error) { + return &api.DeleteNeighborResponse{}, s.bgpServer.DeleteNeighbor(&config.Neighbor{Config: config.NeighborConfig{ + NeighborAddress: arg.Peer.Conf.NeighborAddress, + NeighborInterface: arg.Peer.Conf.NeighborInterface, + }}) +} + +func (s *Server) UpdateNeighbor(ctx context.Context, arg *api.UpdateNeighborRequest) (*api.UpdateNeighborResponse, error) { + c, err := NewNeighborFromAPIStruct(arg.Peer) + if err != nil { + return nil, err + } + needsSoftResetIn, err := s.bgpServer.UpdateNeighbor(c) + if err != nil { + return nil, err + } + if arg.DoSoftResetIn && needsSoftResetIn { + return &api.UpdateNeighborResponse{NeedsSoftResetIn: false}, s.bgpServer.SoftResetIn("", bgp.RouteFamily(0)) + } + return &api.UpdateNeighborResponse{NeedsSoftResetIn: needsSoftResetIn}, nil +} + +func (s *Server) AddPeerGroup(ctx context.Context, arg *api.AddPeerGroupRequest) (*api.AddPeerGroupResponse, error) { + c, err := NewPeerGroupFromAPIStruct(arg.PeerGroup) + if err != nil { + return nil, err + } + return &api.AddPeerGroupResponse{}, s.bgpServer.AddPeerGroup(c) +} + +func (s *Server) DeletePeerGroup(ctx context.Context, arg *api.DeletePeerGroupRequest) (*api.DeletePeerGroupResponse, error) { + return &api.DeletePeerGroupResponse{}, s.bgpServer.DeletePeerGroup(&config.PeerGroup{Config: config.PeerGroupConfig{ + PeerGroupName: arg.PeerGroup.Conf.PeerGroupName, + }}) +} + +func (s *Server) UpdatePeerGroup(ctx context.Context, arg *api.UpdatePeerGroupRequest) (*api.UpdatePeerGroupResponse, error) { + c, err := NewPeerGroupFromAPIStruct(arg.PeerGroup) + if err != nil { + return nil, err + } + needsSoftResetIn, err := s.bgpServer.UpdatePeerGroup(c) + if err != nil { + return nil, err + } + if arg.DoSoftResetIn && needsSoftResetIn { + return &api.UpdatePeerGroupResponse{NeedsSoftResetIn: false}, s.bgpServer.SoftResetIn("", bgp.RouteFamily(0)) + } + return &api.UpdatePeerGroupResponse{NeedsSoftResetIn: needsSoftResetIn}, nil +} + +func (s *Server) AddDynamicNeighbor(ctx context.Context, arg *api.AddDynamicNeighborRequest) (*api.AddDynamicNeighborResponse, error) { + return &api.AddDynamicNeighborResponse{}, s.bgpServer.AddDynamicNeighbor(&config.DynamicNeighbor{Config: config.DynamicNeighborConfig{ + Prefix: arg.DynamicNeighbor.Prefix, + PeerGroup: arg.DynamicNeighbor.PeerGroup, + }}) +} + +func NewPrefixFromApiStruct(a *api.Prefix) (*table.Prefix, error) { + _, prefix, err := net.ParseCIDR(a.IpPrefix) + if err != nil { + return nil, err + } + rf := bgp.RF_IPv4_UC + if strings.Contains(a.IpPrefix, ":") { + rf = bgp.RF_IPv6_UC + } + return &table.Prefix{ + Prefix: prefix, + AddressFamily: rf, + MasklengthRangeMin: uint8(a.MaskLengthMin), + MasklengthRangeMax: uint8(a.MaskLengthMax), + }, nil +} + +func NewConfigPrefixFromAPIStruct(a *api.Prefix) (*config.Prefix, error) { + _, prefix, err := net.ParseCIDR(a.IpPrefix) + if err != nil { + return nil, err + } + return &config.Prefix{ + IpPrefix: prefix.String(), + MasklengthRange: fmt.Sprintf("%d..%d", a.MaskLengthMin, a.MaskLengthMax), + }, nil +} + +func NewAPIPrefixFromConfigStruct(c config.Prefix) (*api.Prefix, error) { + min, max, err := config.ParseMaskLength(c.IpPrefix, c.MasklengthRange) + if err != nil { + return nil, err + } + return &api.Prefix{ + IpPrefix: c.IpPrefix, + MaskLengthMin: uint32(min), + MaskLengthMax: uint32(max), + }, nil +} + +func NewAPIDefinedSetFromTableStruct(t table.DefinedSet) (*api.DefinedSet, error) { + a := &api.DefinedSet{ + Type: api.DefinedType(t.Type()), + Name: t.Name(), + } + switch t.Type() { + case table.DEFINED_TYPE_PREFIX: + s := t.(*table.PrefixSet) + c := s.ToConfig() + for _, p := range c.PrefixList { + ap, err := NewAPIPrefixFromConfigStruct(p) + if err != nil { + return nil, err + } + a.Prefixes = append(a.Prefixes, ap) + } + case table.DEFINED_TYPE_NEIGHBOR: + s := t.(*table.NeighborSet) + c := s.ToConfig() + a.List = append(a.List, c.NeighborInfoList...) + case table.DEFINED_TYPE_AS_PATH: + s := t.(*table.AsPathSet) + c := s.ToConfig() + a.List = append(a.List, c.AsPathList...) + case table.DEFINED_TYPE_COMMUNITY: + s := t.(*table.CommunitySet) + c := s.ToConfig() + a.List = append(a.List, c.CommunityList...) + case table.DEFINED_TYPE_EXT_COMMUNITY: + s := t.(*table.ExtCommunitySet) + c := s.ToConfig() + a.List = append(a.List, c.ExtCommunityList...) + case table.DEFINED_TYPE_LARGE_COMMUNITY: + s := t.(*table.LargeCommunitySet) + c := s.ToConfig() + a.List = append(a.List, c.LargeCommunityList...) + default: + return nil, fmt.Errorf("invalid defined type") + } + return a, nil +} + +func NewAPIDefinedSetsFromConfigStruct(t *config.DefinedSets) ([]*api.DefinedSet, error) { + definedSets := make([]*api.DefinedSet, 0) + + for _, ps := range t.PrefixSets { + prefixes := make([]*api.Prefix, 0) + for _, p := range ps.PrefixList { + ap, err := NewAPIPrefixFromConfigStruct(p) + if err != nil { + return nil, err + } + prefixes = append(prefixes, ap) + } + definedSets = append(definedSets, &api.DefinedSet{ + Type: api.DefinedType_PREFIX, + Name: ps.PrefixSetName, + Prefixes: prefixes, + }) + } + + for _, ns := range t.NeighborSets { + definedSets = append(definedSets, &api.DefinedSet{ + Type: api.DefinedType_NEIGHBOR, + Name: ns.NeighborSetName, + List: ns.NeighborInfoList, + }) + } + + bs := t.BgpDefinedSets + for _, cs := range bs.CommunitySets { + definedSets = append(definedSets, &api.DefinedSet{ + Type: api.DefinedType_COMMUNITY, + Name: cs.CommunitySetName, + List: cs.CommunityList, + }) + } + + for _, es := range bs.ExtCommunitySets { + definedSets = append(definedSets, &api.DefinedSet{ + Type: api.DefinedType_EXT_COMMUNITY, + Name: es.ExtCommunitySetName, + List: es.ExtCommunityList, + }) + } + + for _, ls := range bs.LargeCommunitySets { + definedSets = append(definedSets, &api.DefinedSet{ + Type: api.DefinedType_LARGE_COMMUNITY, + Name: ls.LargeCommunitySetName, + List: ls.LargeCommunityList, + }) + } + + for _, as := range bs.AsPathSets { + definedSets = append(definedSets, &api.DefinedSet{ + Type: api.DefinedType_AS_PATH, + Name: as.AsPathSetName, + List: as.AsPathList, + }) + } + + return definedSets, nil +} + +func NewConfigDefinedSetsFromApiStruct(a []*api.DefinedSet) (*config.DefinedSets, error) { + ps := make([]config.PrefixSet, 0) + ns := make([]config.NeighborSet, 0) + as := make([]config.AsPathSet, 0) + cs := make([]config.CommunitySet, 0) + es := make([]config.ExtCommunitySet, 0) + ls := make([]config.LargeCommunitySet, 0) + + for _, ds := range a { + if ds.Name == "" { + return nil, fmt.Errorf("empty neighbor set name") + } + switch table.DefinedType(ds.Type) { + case table.DEFINED_TYPE_PREFIX: + prefixes := make([]config.Prefix, 0, len(ds.Prefixes)) + for _, p := range ds.Prefixes { + prefix, err := NewConfigPrefixFromAPIStruct(p) + if err != nil { + return nil, err + } + prefixes = append(prefixes, *prefix) + } + ps = append(ps, config.PrefixSet{ + PrefixSetName: ds.Name, + PrefixList: prefixes, + }) + case table.DEFINED_TYPE_NEIGHBOR: + ns = append(ns, config.NeighborSet{ + NeighborSetName: ds.Name, + NeighborInfoList: ds.List, + }) + case table.DEFINED_TYPE_AS_PATH: + as = append(as, config.AsPathSet{ + AsPathSetName: ds.Name, + AsPathList: ds.List, + }) + case table.DEFINED_TYPE_COMMUNITY: + cs = append(cs, config.CommunitySet{ + CommunitySetName: ds.Name, + CommunityList: ds.List, + }) + case table.DEFINED_TYPE_EXT_COMMUNITY: + es = append(es, config.ExtCommunitySet{ + ExtCommunitySetName: ds.Name, + ExtCommunityList: ds.List, + }) + case table.DEFINED_TYPE_LARGE_COMMUNITY: + ls = append(ls, config.LargeCommunitySet{ + LargeCommunitySetName: ds.Name, + LargeCommunityList: ds.List, + }) + default: + return nil, fmt.Errorf("invalid defined type") + } + } + + return &config.DefinedSets{ + PrefixSets: ps, + NeighborSets: ns, + BgpDefinedSets: config.BgpDefinedSets{ + AsPathSets: as, + CommunitySets: cs, + ExtCommunitySets: es, + LargeCommunitySets: ls, + }, + }, nil +} + +func NewDefinedSetFromApiStruct(a *api.DefinedSet) (table.DefinedSet, error) { + if a.Name == "" { + return nil, fmt.Errorf("empty neighbor set name") + } + switch table.DefinedType(a.Type) { + case table.DEFINED_TYPE_PREFIX: + prefixes := make([]*table.Prefix, 0, len(a.Prefixes)) + for _, p := range a.Prefixes { + prefix, err := NewPrefixFromApiStruct(p) + if err != nil { + return nil, err + } + prefixes = append(prefixes, prefix) + } + return table.NewPrefixSetFromApiStruct(a.Name, prefixes) + case table.DEFINED_TYPE_NEIGHBOR: + list := make([]net.IPNet, 0, len(a.List)) + for _, x := range a.List { + _, addr, err := net.ParseCIDR(x) + if err != nil { + return nil, fmt.Errorf("invalid address or prefix: %s", x) + } + list = append(list, *addr) + } + return table.NewNeighborSetFromApiStruct(a.Name, list) + case table.DEFINED_TYPE_AS_PATH: + return table.NewAsPathSet(config.AsPathSet{ + AsPathSetName: a.Name, + AsPathList: a.List, + }) + case table.DEFINED_TYPE_COMMUNITY: + return table.NewCommunitySet(config.CommunitySet{ + CommunitySetName: a.Name, + CommunityList: a.List, + }) + case table.DEFINED_TYPE_EXT_COMMUNITY: + return table.NewExtCommunitySet(config.ExtCommunitySet{ + ExtCommunitySetName: a.Name, + ExtCommunityList: a.List, + }) + case table.DEFINED_TYPE_LARGE_COMMUNITY: + return table.NewLargeCommunitySet(config.LargeCommunitySet{ + LargeCommunitySetName: a.Name, + LargeCommunityList: a.List, + }) + default: + return nil, fmt.Errorf("invalid defined type") + } +} + +var _regexpPrefixMaskLengthRange = regexp.MustCompile(`(\d+)\.\.(\d+)`) + +func (s *Server) GetDefinedSet(ctx context.Context, arg *api.GetDefinedSetRequest) (*api.GetDefinedSetResponse, error) { + cd, err := s.bgpServer.GetDefinedSet(table.DefinedType(arg.Type), arg.Name) + if err != nil { + return nil, err + } + sets := make([]*api.DefinedSet, 0) + for _, cs := range cd.PrefixSets { + ad := &api.DefinedSet{ + Type: api.DefinedType_PREFIX, + Name: cs.PrefixSetName, + Prefixes: func() []*api.Prefix { + l := make([]*api.Prefix, 0, len(cs.PrefixList)) + for _, p := range cs.PrefixList { + elems := _regexpPrefixMaskLengthRange.FindStringSubmatch(p.MasklengthRange) + min, _ := strconv.ParseUint(elems[1], 10, 32) + max, _ := strconv.ParseUint(elems[2], 10, 32) + + l = append(l, &api.Prefix{IpPrefix: p.IpPrefix, MaskLengthMin: uint32(min), MaskLengthMax: uint32(max)}) + } + return l + }(), + } + sets = append(sets, ad) + + } + for _, cs := range cd.NeighborSets { + ad := &api.DefinedSet{ + Type: api.DefinedType_NEIGHBOR, + Name: cs.NeighborSetName, + List: cs.NeighborInfoList, + } + sets = append(sets, ad) + } + for _, cs := range cd.BgpDefinedSets.CommunitySets { + ad := &api.DefinedSet{ + Type: api.DefinedType_COMMUNITY, + Name: cs.CommunitySetName, + List: cs.CommunityList, + } + sets = append(sets, ad) + } + for _, cs := range cd.BgpDefinedSets.ExtCommunitySets { + ad := &api.DefinedSet{ + Type: api.DefinedType_EXT_COMMUNITY, + Name: cs.ExtCommunitySetName, + List: cs.ExtCommunityList, + } + sets = append(sets, ad) + } + for _, cs := range cd.BgpDefinedSets.LargeCommunitySets { + ad := &api.DefinedSet{ + Type: api.DefinedType_LARGE_COMMUNITY, + Name: cs.LargeCommunitySetName, + List: cs.LargeCommunityList, + } + sets = append(sets, ad) + } + for _, cs := range cd.BgpDefinedSets.AsPathSets { + ad := &api.DefinedSet{ + Type: api.DefinedType_AS_PATH, + Name: cs.AsPathSetName, + List: cs.AsPathList, + } + sets = append(sets, ad) + } + + return &api.GetDefinedSetResponse{Sets: sets}, nil +} + +func (s *Server) AddDefinedSet(ctx context.Context, arg *api.AddDefinedSetRequest) (*api.AddDefinedSetResponse, error) { + if arg == nil || arg.Set == nil { + return nil, fmt.Errorf("invalid request") + } + set, err := NewDefinedSetFromApiStruct(arg.Set) + if err != nil { + return nil, err + } + return &api.AddDefinedSetResponse{}, s.bgpServer.AddDefinedSet(set) +} + +func (s *Server) DeleteDefinedSet(ctx context.Context, arg *api.DeleteDefinedSetRequest) (*api.DeleteDefinedSetResponse, error) { + if arg == nil || arg.Set == nil { + return nil, fmt.Errorf("invalid request") + } + set, err := NewDefinedSetFromApiStruct(arg.Set) + if err != nil { + return nil, err + } + return &api.DeleteDefinedSetResponse{}, s.bgpServer.DeleteDefinedSet(set, arg.All) +} + +func (s *Server) ReplaceDefinedSet(ctx context.Context, arg *api.ReplaceDefinedSetRequest) (*api.ReplaceDefinedSetResponse, error) { + if arg == nil || arg.Set == nil { + return nil, fmt.Errorf("invalid request") + } + set, err := NewDefinedSetFromApiStruct(arg.Set) + if err != nil { + return nil, err + } + return &api.ReplaceDefinedSetResponse{}, s.bgpServer.ReplaceDefinedSet(set) +} + +func NewAPIStatementFromTableStruct(t *table.Statement) *api.Statement { + return toStatementApi(t.ToConfig()) +} + +var _regexpMedActionType = regexp.MustCompile(`([+-]?)(\d+)`) + +func toStatementApi(s *config.Statement) *api.Statement { + cs := &api.Conditions{} + if s.Conditions.MatchPrefixSet.PrefixSet != "" { + o, _ := table.NewMatchOption(s.Conditions.MatchPrefixSet.MatchSetOptions) + cs.PrefixSet = &api.MatchSet{ + Type: api.MatchType(o), + Name: s.Conditions.MatchPrefixSet.PrefixSet, + } + } + if s.Conditions.MatchNeighborSet.NeighborSet != "" { + o, _ := table.NewMatchOption(s.Conditions.MatchNeighborSet.MatchSetOptions) + cs.NeighborSet = &api.MatchSet{ + Type: api.MatchType(o), + Name: s.Conditions.MatchNeighborSet.NeighborSet, + } + } + if s.Conditions.BgpConditions.AsPathLength.Operator != "" { + cs.AsPathLength = &api.AsPathLength{ + Length: s.Conditions.BgpConditions.AsPathLength.Value, + Type: api.AsPathLengthType(s.Conditions.BgpConditions.AsPathLength.Operator.ToInt()), + } + } + if s.Conditions.BgpConditions.MatchAsPathSet.AsPathSet != "" { + cs.AsPathSet = &api.MatchSet{ + Type: api.MatchType(s.Conditions.BgpConditions.MatchAsPathSet.MatchSetOptions.ToInt()), + Name: s.Conditions.BgpConditions.MatchAsPathSet.AsPathSet, + } + } + if s.Conditions.BgpConditions.MatchCommunitySet.CommunitySet != "" { + cs.CommunitySet = &api.MatchSet{ + Type: api.MatchType(s.Conditions.BgpConditions.MatchCommunitySet.MatchSetOptions.ToInt()), + Name: s.Conditions.BgpConditions.MatchCommunitySet.CommunitySet, + } + } + if s.Conditions.BgpConditions.MatchExtCommunitySet.ExtCommunitySet != "" { + cs.ExtCommunitySet = &api.MatchSet{ + Type: api.MatchType(s.Conditions.BgpConditions.MatchExtCommunitySet.MatchSetOptions.ToInt()), + Name: s.Conditions.BgpConditions.MatchExtCommunitySet.ExtCommunitySet, + } + } + if s.Conditions.BgpConditions.MatchLargeCommunitySet.LargeCommunitySet != "" { + cs.LargeCommunitySet = &api.MatchSet{ + Type: api.MatchType(s.Conditions.BgpConditions.MatchLargeCommunitySet.MatchSetOptions.ToInt()), + Name: s.Conditions.BgpConditions.MatchLargeCommunitySet.LargeCommunitySet, + } + } + if s.Conditions.BgpConditions.RouteType != "" { + cs.RouteType = api.Conditions_RouteType(s.Conditions.BgpConditions.RouteType.ToInt()) + } + if len(s.Conditions.BgpConditions.NextHopInList) > 0 { + cs.NextHopInList = s.Conditions.BgpConditions.NextHopInList + } + if s.Conditions.BgpConditions.AfiSafiInList != nil { + afiSafiIn := make([]api.Family, 0) + for _, afiSafiType := range s.Conditions.BgpConditions.AfiSafiInList { + if mapped, ok := bgp.AddressFamilyValueMap[string(afiSafiType)]; ok { + afiSafiIn = append(afiSafiIn, api.Family(mapped)) + } + } + cs.AfiSafiIn = afiSafiIn + } + cs.RpkiResult = int32(s.Conditions.BgpConditions.RpkiValidationResult.ToInt()) + as := &api.Actions{ + RouteAction: func() api.RouteAction { + switch s.Actions.RouteDisposition { + case config.ROUTE_DISPOSITION_ACCEPT_ROUTE: + return api.RouteAction_ACCEPT + case config.ROUTE_DISPOSITION_REJECT_ROUTE: + return api.RouteAction_REJECT + } + return api.RouteAction_NONE + }(), + Community: func() *api.CommunityAction { + if len(s.Actions.BgpActions.SetCommunity.SetCommunityMethod.CommunitiesList) == 0 { + return nil + } + return &api.CommunityAction{ + Type: api.CommunityActionType(config.BgpSetCommunityOptionTypeToIntMap[config.BgpSetCommunityOptionType(s.Actions.BgpActions.SetCommunity.Options)]), + Communities: s.Actions.BgpActions.SetCommunity.SetCommunityMethod.CommunitiesList} + }(), + Med: func() *api.MedAction { + medStr := strings.TrimSpace(string(s.Actions.BgpActions.SetMed)) + if len(medStr) == 0 { + return nil + } + matches := _regexpMedActionType.FindStringSubmatch(medStr) + if len(matches) == 0 { + return nil + } + action := api.MedActionType_MED_REPLACE + switch matches[1] { + case "+", "-": + action = api.MedActionType_MED_MOD + } + value, err := strconv.ParseInt(matches[1]+matches[2], 10, 64) + if err != nil { + return nil + } + return &api.MedAction{ + Value: value, + Type: action, + } + }(), + AsPrepend: func() *api.AsPrependAction { + if len(s.Actions.BgpActions.SetAsPathPrepend.As) == 0 { + return nil + } + var asn uint64 + useleft := false + if s.Actions.BgpActions.SetAsPathPrepend.As != "last-as" { + asn, _ = strconv.ParseUint(s.Actions.BgpActions.SetAsPathPrepend.As, 10, 32) + } else { + useleft = true + } + return &api.AsPrependAction{ + Asn: uint32(asn), + Repeat: uint32(s.Actions.BgpActions.SetAsPathPrepend.RepeatN), + UseLeftMost: useleft, + } + }(), + ExtCommunity: func() *api.CommunityAction { + if len(s.Actions.BgpActions.SetExtCommunity.SetExtCommunityMethod.CommunitiesList) == 0 { + return nil + } + return &api.CommunityAction{ + Type: api.CommunityActionType(config.BgpSetCommunityOptionTypeToIntMap[config.BgpSetCommunityOptionType(s.Actions.BgpActions.SetExtCommunity.Options)]), + Communities: s.Actions.BgpActions.SetExtCommunity.SetExtCommunityMethod.CommunitiesList, + } + }(), + LargeCommunity: func() *api.CommunityAction { + if len(s.Actions.BgpActions.SetLargeCommunity.SetLargeCommunityMethod.CommunitiesList) == 0 { + return nil + } + return &api.CommunityAction{ + Type: api.CommunityActionType(config.BgpSetCommunityOptionTypeToIntMap[config.BgpSetCommunityOptionType(s.Actions.BgpActions.SetLargeCommunity.Options)]), + Communities: s.Actions.BgpActions.SetLargeCommunity.SetLargeCommunityMethod.CommunitiesList, + } + }(), + Nexthop: func() *api.NexthopAction { + if len(string(s.Actions.BgpActions.SetNextHop)) == 0 { + return nil + } + + if string(s.Actions.BgpActions.SetNextHop) == "self" { + return &api.NexthopAction{ + Self: true, + } + } + return &api.NexthopAction{ + Address: string(s.Actions.BgpActions.SetNextHop), + } + }(), + LocalPref: func() *api.LocalPrefAction { + if s.Actions.BgpActions.SetLocalPref == 0 { + return nil + } + return &api.LocalPrefAction{Value: s.Actions.BgpActions.SetLocalPref} + }(), + } + return &api.Statement{ + Name: s.Name, + Conditions: cs, + Actions: as, + } +} + +func toConfigMatchSetOption(a api.MatchType) (config.MatchSetOptionsType, error) { + var typ config.MatchSetOptionsType + switch a { + case api.MatchType_ANY: + typ = config.MATCH_SET_OPTIONS_TYPE_ANY + case api.MatchType_ALL: + typ = config.MATCH_SET_OPTIONS_TYPE_ALL + case api.MatchType_INVERT: + typ = config.MATCH_SET_OPTIONS_TYPE_INVERT + default: + return typ, fmt.Errorf("invalid match type") + } + return typ, nil +} + +func toConfigMatchSetOptionRestricted(a api.MatchType) (config.MatchSetOptionsRestrictedType, error) { + var typ config.MatchSetOptionsRestrictedType + switch a { + case api.MatchType_ANY: + typ = config.MATCH_SET_OPTIONS_RESTRICTED_TYPE_ANY + case api.MatchType_INVERT: + typ = config.MATCH_SET_OPTIONS_RESTRICTED_TYPE_INVERT + default: + return typ, fmt.Errorf("invalid match type") + } + return typ, nil +} + +func NewPrefixConditionFromApiStruct(a *api.MatchSet) (*table.PrefixCondition, error) { + if a == nil { + return nil, nil + } + typ, err := toConfigMatchSetOptionRestricted(a.Type) + if err != nil { + return nil, err + } + c := config.MatchPrefixSet{ + PrefixSet: a.Name, + MatchSetOptions: typ, + } + return table.NewPrefixCondition(c) +} + +func NewNeighborConditionFromApiStruct(a *api.MatchSet) (*table.NeighborCondition, error) { + if a == nil { + return nil, nil + } + typ, err := toConfigMatchSetOptionRestricted(a.Type) + if err != nil { + return nil, err + } + c := config.MatchNeighborSet{ + NeighborSet: a.Name, + MatchSetOptions: typ, + } + return table.NewNeighborCondition(c) +} + +func NewAsPathLengthConditionFromApiStruct(a *api.AsPathLength) (*table.AsPathLengthCondition, error) { + if a == nil { + return nil, nil + } + return table.NewAsPathLengthCondition(config.AsPathLength{ + Operator: config.IntToAttributeComparisonMap[int(a.Type)], + Value: a.Length, + }) +} + +func NewAsPathConditionFromApiStruct(a *api.MatchSet) (*table.AsPathCondition, error) { + if a == nil { + return nil, nil + } + typ, err := toConfigMatchSetOption(a.Type) + if err != nil { + return nil, err + } + c := config.MatchAsPathSet{ + AsPathSet: a.Name, + MatchSetOptions: typ, + } + return table.NewAsPathCondition(c) +} + +func NewRpkiValidationConditionFromApiStruct(a int32) (*table.RpkiValidationCondition, error) { + if a < 1 { + return nil, nil + } + return table.NewRpkiValidationCondition(config.IntToRpkiValidationResultTypeMap[int(a)]) +} + +func NewRouteTypeConditionFromApiStruct(a api.Conditions_RouteType) (*table.RouteTypeCondition, error) { + if a == 0 { + return nil, nil + } + typ, ok := config.IntToRouteTypeMap[int(a)] + if !ok { + return nil, fmt.Errorf("invalid route type: %d", a) + } + return table.NewRouteTypeCondition(typ) +} + +func NewCommunityConditionFromApiStruct(a *api.MatchSet) (*table.CommunityCondition, error) { + if a == nil { + return nil, nil + } + typ, err := toConfigMatchSetOption(a.Type) + if err != nil { + return nil, err + } + c := config.MatchCommunitySet{ + CommunitySet: a.Name, + MatchSetOptions: typ, + } + return table.NewCommunityCondition(c) +} + +func NewExtCommunityConditionFromApiStruct(a *api.MatchSet) (*table.ExtCommunityCondition, error) { + if a == nil { + return nil, nil + } + typ, err := toConfigMatchSetOption(a.Type) + if err != nil { + return nil, err + } + c := config.MatchExtCommunitySet{ + ExtCommunitySet: a.Name, + MatchSetOptions: typ, + } + return table.NewExtCommunityCondition(c) +} + +func NewLargeCommunityConditionFromApiStruct(a *api.MatchSet) (*table.LargeCommunityCondition, error) { + if a == nil { + return nil, nil + } + typ, err := toConfigMatchSetOption(a.Type) + if err != nil { + return nil, err + } + c := config.MatchLargeCommunitySet{ + LargeCommunitySet: a.Name, + MatchSetOptions: typ, + } + return table.NewLargeCommunityCondition(c) +} + +func NewNextHopConditionFromApiStruct(a []string) (*table.NextHopCondition, error) { + if a == nil { + return nil, nil + } + + return table.NewNextHopCondition(a) +} + +func NewAfiSafiInConditionFromApiStruct(a []api.Family) (*table.AfiSafiInCondition, error) { + if a == nil { + return nil, nil + } + afiSafiTypes := make([]config.AfiSafiType, 0, len(a)) + for _, aType := range a { + if configType, ok := bgp.AddressFamilyNameMap[bgp.RouteFamily(aType)]; ok { + afiSafiTypes = append(afiSafiTypes, config.AfiSafiType(configType)) + } else { + return nil, fmt.Errorf("unknown afi-safi-in type value: %d", aType) + } + } + return table.NewAfiSafiInCondition(afiSafiTypes) +} + +func NewRoutingActionFromApiStruct(a api.RouteAction) (*table.RoutingAction, error) { + if a == api.RouteAction_NONE { + return nil, nil + } + accept := false + if a == api.RouteAction_ACCEPT { + accept = true + } + return &table.RoutingAction{ + AcceptRoute: accept, + }, nil +} + +func NewCommunityActionFromApiStruct(a *api.CommunityAction) (*table.CommunityAction, error) { + if a == nil { + return nil, nil + } + return table.NewCommunityAction(config.SetCommunity{ + Options: string(config.IntToBgpSetCommunityOptionTypeMap[int(a.Type)]), + SetCommunityMethod: config.SetCommunityMethod{ + CommunitiesList: a.Communities, + }, + }) +} + +func NewExtCommunityActionFromApiStruct(a *api.CommunityAction) (*table.ExtCommunityAction, error) { + if a == nil { + return nil, nil + } + return table.NewExtCommunityAction(config.SetExtCommunity{ + Options: string(config.IntToBgpSetCommunityOptionTypeMap[int(a.Type)]), + SetExtCommunityMethod: config.SetExtCommunityMethod{ + CommunitiesList: a.Communities, + }, + }) +} + +func NewLargeCommunityActionFromApiStruct(a *api.CommunityAction) (*table.LargeCommunityAction, error) { + if a == nil { + return nil, nil + } + return table.NewLargeCommunityAction(config.SetLargeCommunity{ + Options: config.IntToBgpSetCommunityOptionTypeMap[int(a.Type)], + SetLargeCommunityMethod: config.SetLargeCommunityMethod{ + CommunitiesList: a.Communities, + }, + }) +} + +func NewMedActionFromApiStruct(a *api.MedAction) (*table.MedAction, error) { + if a == nil { + return nil, nil + } + return table.NewMedActionFromApiStruct(table.MedActionType(a.Type), a.Value), nil +} + +func NewLocalPrefActionFromApiStruct(a *api.LocalPrefAction) (*table.LocalPrefAction, error) { + if a == nil || a.Value == 0 { + return nil, nil + } + return table.NewLocalPrefAction(a.Value) +} + +func NewAsPathPrependActionFromApiStruct(a *api.AsPrependAction) (*table.AsPathPrependAction, error) { + if a == nil { + return nil, nil + } + return table.NewAsPathPrependAction(config.SetAsPathPrepend{ + RepeatN: uint8(a.Repeat), + As: func() string { + if a.UseLeftMost { + return "last-as" + } + return fmt.Sprintf("%d", a.Asn) + }(), + }) +} + +func NewNexthopActionFromApiStruct(a *api.NexthopAction) (*table.NexthopAction, error) { + if a == nil { + return nil, nil + } + return table.NewNexthopAction(config.BgpNextHopType( + func() string { + if a.Self { + return "self" + } + return a.Address + }(), + )) +} + +func NewStatementFromApiStruct(a *api.Statement) (*table.Statement, error) { + if a.Name == "" { + return nil, fmt.Errorf("empty statement name") + } + var ra table.Action + var as []table.Action + var cs []table.Condition + var err error + if a.Conditions != nil { + cfs := []func() (table.Condition, error){ + func() (table.Condition, error) { + return NewPrefixConditionFromApiStruct(a.Conditions.PrefixSet) + }, + func() (table.Condition, error) { + return NewNeighborConditionFromApiStruct(a.Conditions.NeighborSet) + }, + func() (table.Condition, error) { + return NewAsPathLengthConditionFromApiStruct(a.Conditions.AsPathLength) + }, + func() (table.Condition, error) { + return NewRpkiValidationConditionFromApiStruct(a.Conditions.RpkiResult) + }, + func() (table.Condition, error) { + return NewRouteTypeConditionFromApiStruct(a.Conditions.RouteType) + }, + func() (table.Condition, error) { + return NewAsPathConditionFromApiStruct(a.Conditions.AsPathSet) + }, + func() (table.Condition, error) { + return NewCommunityConditionFromApiStruct(a.Conditions.CommunitySet) + }, + func() (table.Condition, error) { + return NewExtCommunityConditionFromApiStruct(a.Conditions.ExtCommunitySet) + }, + func() (table.Condition, error) { + return NewLargeCommunityConditionFromApiStruct(a.Conditions.LargeCommunitySet) + }, + func() (table.Condition, error) { + return NewNextHopConditionFromApiStruct(a.Conditions.NextHopInList) + }, + func() (table.Condition, error) { + return NewAfiSafiInConditionFromApiStruct(a.Conditions.AfiSafiIn) + }, + } + cs = make([]table.Condition, 0, len(cfs)) + for _, f := range cfs { + c, err := f() + if err != nil { + return nil, err + } + if !reflect.ValueOf(c).IsNil() { + cs = append(cs, c) + } + } + } + if a.Actions != nil { + ra, err = NewRoutingActionFromApiStruct(a.Actions.RouteAction) + if err != nil { + return nil, err + } + afs := []func() (table.Action, error){ + func() (table.Action, error) { + return NewCommunityActionFromApiStruct(a.Actions.Community) + }, + func() (table.Action, error) { + return NewExtCommunityActionFromApiStruct(a.Actions.ExtCommunity) + }, + func() (table.Action, error) { + return NewLargeCommunityActionFromApiStruct(a.Actions.LargeCommunity) + }, + func() (table.Action, error) { + return NewMedActionFromApiStruct(a.Actions.Med) + }, + func() (table.Action, error) { + return NewLocalPrefActionFromApiStruct(a.Actions.LocalPref) + }, + func() (table.Action, error) { + return NewAsPathPrependActionFromApiStruct(a.Actions.AsPrepend) + }, + func() (table.Action, error) { + return NewNexthopActionFromApiStruct(a.Actions.Nexthop) + }, + } + as = make([]table.Action, 0, len(afs)) + for _, f := range afs { + a, err := f() + if err != nil { + return nil, err + } + if !reflect.ValueOf(a).IsNil() { + as = append(as, a) + } + } + } + return &table.Statement{ + Name: a.Name, + Conditions: cs, + RouteAction: ra, + ModActions: as, + }, nil +} + +func (s *Server) GetStatement(ctx context.Context, arg *api.GetStatementRequest) (*api.GetStatementResponse, error) { + l := make([]*api.Statement, 0) + for _, s := range s.bgpServer.GetStatement() { + l = append(l, toStatementApi(s)) + } + return &api.GetStatementResponse{Statements: l}, nil +} + +func (s *Server) AddStatement(ctx context.Context, arg *api.AddStatementRequest) (*api.AddStatementResponse, error) { + if arg == nil || arg.Statement == nil { + return nil, fmt.Errorf("invalid request") + } + st, err := NewStatementFromApiStruct(arg.Statement) + if err == nil { + err = s.bgpServer.AddStatement(st) + } + return &api.AddStatementResponse{}, err +} + +func (s *Server) DeleteStatement(ctx context.Context, arg *api.DeleteStatementRequest) (*api.DeleteStatementResponse, error) { + if arg == nil || arg.Statement == nil { + return nil, fmt.Errorf("invalid request") + } + st, err := NewStatementFromApiStruct(arg.Statement) + if err == nil { + err = s.bgpServer.DeleteStatement(st, arg.All) + } + return &api.DeleteStatementResponse{}, err +} + +func (s *Server) ReplaceStatement(ctx context.Context, arg *api.ReplaceStatementRequest) (*api.ReplaceStatementResponse, error) { + if arg == nil || arg.Statement == nil { + return nil, fmt.Errorf("invalid request") + } + st, err := NewStatementFromApiStruct(arg.Statement) + if err == nil { + err = s.bgpServer.ReplaceStatement(st) + } + return &api.ReplaceStatementResponse{}, err +} + +func NewAPIPolicyFromTableStruct(p *table.Policy) *api.Policy { + return toPolicyApi(p.ToConfig()) +} + +func toPolicyApi(p *config.PolicyDefinition) *api.Policy { + return &api.Policy{ + Name: p.Name, + Statements: func() []*api.Statement { + l := make([]*api.Statement, 0) + for _, s := range p.Statements { + l = append(l, toStatementApi(&s)) + } + return l + }(), + } +} + +func NewAPIPolicyAssignmentFromTableStruct(t *table.PolicyAssignment) *api.PolicyAssignment { + return &api.PolicyAssignment{ + Type: func() api.PolicyType { + switch t.Type { + case table.POLICY_DIRECTION_IMPORT: + return api.PolicyType_IMPORT + case table.POLICY_DIRECTION_EXPORT: + return api.PolicyType_EXPORT + } + log.Errorf("invalid policy-type: %s", t.Type) + return api.PolicyType(-1) + }(), + Default: func() api.RouteAction { + switch t.Default { + case table.ROUTE_TYPE_ACCEPT: + return api.RouteAction_ACCEPT + case table.ROUTE_TYPE_REJECT: + return api.RouteAction_REJECT + } + return api.RouteAction_NONE + }(), + Name: t.Name, + Resource: func() api.Resource { + if t.Name != "" { + return api.Resource_LOCAL + } + return api.Resource_GLOBAL + }(), + Policies: func() []*api.Policy { + l := make([]*api.Policy, 0) + for _, p := range t.Policies { + l = append(l, NewAPIPolicyFromTableStruct(p)) + } + return l + }(), + } +} + +func NewConfigPolicyFromApiStruct(a *api.Policy) (*config.PolicyDefinition, error) { + if a.Name == "" { + return nil, fmt.Errorf("empty policy name") + } + stmts := make([]config.Statement, 0, len(a.Statements)) + for idx, x := range a.Statements { + if x.Name == "" { + x.Name = fmt.Sprintf("%s_stmt%d", a.Name, idx) + } + y, err := NewStatementFromApiStruct(x) + if err != nil { + return nil, err + } + stmt := y.ToConfig() + stmts = append(stmts, *stmt) + } + return &config.PolicyDefinition{ + Name: a.Name, + Statements: stmts, + }, nil +} + +func NewPolicyFromApiStruct(a *api.Policy) (*table.Policy, error) { + if a.Name == "" { + return nil, fmt.Errorf("empty policy name") + } + stmts := make([]*table.Statement, 0, len(a.Statements)) + for idx, x := range a.Statements { + if x.Name == "" { + x.Name = fmt.Sprintf("%s_stmt%d", a.Name, idx) + } + y, err := NewStatementFromApiStruct(x) + if err != nil { + return nil, err + } + stmts = append(stmts, y) + } + return &table.Policy{ + Name: a.Name, + Statements: stmts, + }, nil +} + +func NewRoaListFromTableStructList(origin []*table.ROA) []*api.Roa { + l := make([]*api.Roa, 0) + for _, r := range origin { + host, port, _ := net.SplitHostPort(r.Src) + l = append(l, &api.Roa{ + As: r.AS, + Maxlen: uint32(r.MaxLen), + Prefixlen: uint32(r.Prefix.Length), + Prefix: r.Prefix.Prefix.String(), + Conf: &api.RPKIConf{ + Address: host, + RemotePort: port, + }, + }) + } + return l +} + +func (s *Server) GetPolicy(ctx context.Context, arg *api.GetPolicyRequest) (*api.GetPolicyResponse, error) { + l := make([]*api.Policy, 0) + for _, p := range s.bgpServer.GetPolicy() { + l = append(l, toPolicyApi(p)) + } + return &api.GetPolicyResponse{Policies: l}, nil +} + +func (s *Server) AddPolicy(ctx context.Context, arg *api.AddPolicyRequest) (*api.AddPolicyResponse, error) { + if arg == nil || arg.Policy == nil { + return nil, fmt.Errorf("invalid request") + } + x, err := NewPolicyFromApiStruct(arg.Policy) + if err != nil { + return nil, err + } + return &api.AddPolicyResponse{}, s.bgpServer.AddPolicy(x, arg.ReferExistingStatements) +} + +func (s *Server) DeletePolicy(ctx context.Context, arg *api.DeletePolicyRequest) (*api.DeletePolicyResponse, error) { + if arg == nil || arg.Policy == nil { + return nil, fmt.Errorf("invalid request") + } + x, err := NewPolicyFromApiStruct(arg.Policy) + if err != nil { + return nil, err + } + return &api.DeletePolicyResponse{}, s.bgpServer.DeletePolicy(x, arg.All, arg.PreserveStatements) +} + +func (s *Server) ReplacePolicy(ctx context.Context, arg *api.ReplacePolicyRequest) (*api.ReplacePolicyResponse, error) { + if arg == nil || arg.Policy == nil { + return nil, fmt.Errorf("invalid request") + } + x, err := NewPolicyFromApiStruct(arg.Policy) + if err != nil { + return nil, err + } + return &api.ReplacePolicyResponse{}, s.bgpServer.ReplacePolicy(x, arg.ReferExistingStatements, arg.PreserveStatements) +} + +func toPolicyAssignmentName(a *api.PolicyAssignment) (string, table.PolicyDirection, error) { + switch a.Resource { + case api.Resource_GLOBAL: + switch a.Type { + case api.PolicyType_IMPORT: + return "", table.POLICY_DIRECTION_IMPORT, nil + case api.PolicyType_EXPORT: + return "", table.POLICY_DIRECTION_EXPORT, nil + default: + return "", table.POLICY_DIRECTION_NONE, fmt.Errorf("invalid policy type") + } + case api.Resource_LOCAL: + switch a.Type { + case api.PolicyType_IMPORT: + return a.Name, table.POLICY_DIRECTION_IMPORT, nil + case api.PolicyType_EXPORT: + return a.Name, table.POLICY_DIRECTION_EXPORT, nil + default: + return "", table.POLICY_DIRECTION_NONE, fmt.Errorf("invalid policy type") + } + default: + return "", table.POLICY_DIRECTION_NONE, fmt.Errorf("invalid resource type") + } + +} + +func (s *Server) GetPolicyAssignment(ctx context.Context, arg *api.GetPolicyAssignmentRequest) (*api.GetPolicyAssignmentResponse, error) { + if arg == nil || arg.Assignment == nil { + return nil, fmt.Errorf("invalid request") + } + name, dir, err := toPolicyAssignmentName(arg.Assignment) + if err != nil { + return nil, err + } + def, pols, err := s.bgpServer.GetPolicyAssignment(name, dir) + if err != nil { + return nil, err + } + policies := make([]*table.Policy, 0, len(pols)) + for _, p := range pols { + t, err := table.NewPolicy(*p) + if err != nil { + return nil, err + } + policies = append(policies, t) + } + t := &table.PolicyAssignment{ + Name: name, + Type: dir, + Default: def, + Policies: policies, + } + return &api.GetPolicyAssignmentResponse{NewAPIPolicyAssignmentFromTableStruct(t)}, err +} + +func defaultRouteType(d api.RouteAction) table.RouteType { + switch d { + case api.RouteAction_ACCEPT: + return table.ROUTE_TYPE_ACCEPT + case api.RouteAction_REJECT: + return table.ROUTE_TYPE_REJECT + default: + return table.ROUTE_TYPE_NONE + } +} + +func toPolicyDefinition(policies []*api.Policy) []*config.PolicyDefinition { + l := make([]*config.PolicyDefinition, 0, len(policies)) + for _, p := range policies { + l = append(l, &config.PolicyDefinition{Name: p.Name}) + } + return l +} + +func (s *Server) AddPolicyAssignment(ctx context.Context, arg *api.AddPolicyAssignmentRequest) (*api.AddPolicyAssignmentResponse, error) { + if arg == nil || arg.Assignment == nil { + return nil, fmt.Errorf("invalid request") + } + name, dir, err := toPolicyAssignmentName(arg.Assignment) + if err != nil { + return nil, err + } + return &api.AddPolicyAssignmentResponse{}, s.bgpServer.AddPolicyAssignment(name, dir, toPolicyDefinition(arg.Assignment.Policies), defaultRouteType(arg.Assignment.Default)) +} + +func (s *Server) DeletePolicyAssignment(ctx context.Context, arg *api.DeletePolicyAssignmentRequest) (*api.DeletePolicyAssignmentResponse, error) { + if arg == nil || arg.Assignment == nil { + return nil, fmt.Errorf("invalid request") + } + name, dir, err := toPolicyAssignmentName(arg.Assignment) + if err != nil { + return nil, err + } + return &api.DeletePolicyAssignmentResponse{}, s.bgpServer.DeletePolicyAssignment(name, dir, toPolicyDefinition(arg.Assignment.Policies), arg.All) +} + +func (s *Server) ReplacePolicyAssignment(ctx context.Context, arg *api.ReplacePolicyAssignmentRequest) (*api.ReplacePolicyAssignmentResponse, error) { + if arg == nil || arg.Assignment == nil { + return nil, fmt.Errorf("invalid request") + } + name, dir, err := toPolicyAssignmentName(arg.Assignment) + if err != nil { + return nil, err + } + return &api.ReplacePolicyAssignmentResponse{}, s.bgpServer.ReplacePolicyAssignment(name, dir, toPolicyDefinition(arg.Assignment.Policies), defaultRouteType(arg.Assignment.Default)) +} + +func (s *Server) GetServer(ctx context.Context, arg *api.GetServerRequest) (*api.GetServerResponse, error) { + g := s.bgpServer.GetServer() + return &api.GetServerResponse{ + Global: &api.Global{ + As: g.Config.As, + RouterId: g.Config.RouterId, + ListenPort: g.Config.Port, + ListenAddresses: g.Config.LocalAddressList, + UseMultiplePaths: g.UseMultiplePaths.Config.Enabled, + }, + }, nil +} + +func NewGlobalFromAPIStruct(a *api.Global) *config.Global { + families := make([]config.AfiSafi, 0, len(a.Families)) + for _, f := range a.Families { + name := config.IntToAfiSafiTypeMap[int(f)] + rf, _ := bgp.GetRouteFamily(string(name)) + families = append(families, config.AfiSafi{ + Config: config.AfiSafiConfig{ + AfiSafiName: name, + Enabled: true, + }, + State: config.AfiSafiState{ + AfiSafiName: name, + Enabled: true, + Family: rf, + }, + }) + } + + applyPolicy := &config.ApplyPolicy{} + ReadApplyPolicyFromAPIStruct(applyPolicy, a.ApplyPolicy) + + global := &config.Global{ + Config: config.GlobalConfig{ + As: a.As, + RouterId: a.RouterId, + Port: a.ListenPort, + LocalAddressList: a.ListenAddresses, + }, + ApplyPolicy: *applyPolicy, + AfiSafis: families, + UseMultiplePaths: config.UseMultiplePaths{ + Config: config.UseMultiplePathsConfig{ + Enabled: a.UseMultiplePaths, + }, + }, + } + if a.RouteSelectionOptions != nil { + global.RouteSelectionOptions = config.RouteSelectionOptions{ + Config: config.RouteSelectionOptionsConfig{ + AlwaysCompareMed: a.RouteSelectionOptions.AlwaysCompareMed, + IgnoreAsPathLength: a.RouteSelectionOptions.IgnoreAsPathLength, + ExternalCompareRouterId: a.RouteSelectionOptions.ExternalCompareRouterId, + AdvertiseInactiveRoutes: a.RouteSelectionOptions.AdvertiseInactiveRoutes, + EnableAigp: a.RouteSelectionOptions.EnableAigp, + IgnoreNextHopIgpMetric: a.RouteSelectionOptions.IgnoreNextHopIgpMetric, + DisableBestPathSelection: a.RouteSelectionOptions.DisableBestPathSelection, + }, + } + } + if a.DefaultRouteDistance != nil { + global.DefaultRouteDistance = config.DefaultRouteDistance{ + Config: config.DefaultRouteDistanceConfig{ + ExternalRouteDistance: uint8(a.DefaultRouteDistance.ExternalRouteDistance), + InternalRouteDistance: uint8(a.DefaultRouteDistance.InternalRouteDistance), + }, + } + } + if a.Confederation != nil { + global.Confederation = config.Confederation{ + Config: config.ConfederationConfig{ + Enabled: a.Confederation.Enabled, + Identifier: a.Confederation.Identifier, + MemberAsList: a.Confederation.MemberAsList, + }, + } + } + if a.GracefulRestart != nil { + global.GracefulRestart = config.GracefulRestart{ + Config: config.GracefulRestartConfig{ + Enabled: a.GracefulRestart.Enabled, + RestartTime: uint16(a.GracefulRestart.RestartTime), + StaleRoutesTime: float64(a.GracefulRestart.StaleRoutesTime), + HelperOnly: a.GracefulRestart.HelperOnly, + DeferralTime: uint16(a.GracefulRestart.DeferralTime), + NotificationEnabled: a.GracefulRestart.NotificationEnabled, + LongLivedEnabled: a.GracefulRestart.LonglivedEnabled, + }, + } + } + return global +} + +func NewGlobalFromConfigStruct(c *config.Global) *api.Global { + families := make([]uint32, 0, len(c.AfiSafis)) + for _, f := range c.AfiSafis { + families = append(families, uint32(config.AfiSafiTypeToIntMap[f.Config.AfiSafiName])) + } + + applyPolicy := NewApplyPolicyFromConfigStruct(&c.ApplyPolicy) + + return &api.Global{ + As: c.Config.As, + RouterId: c.Config.RouterId, + ListenPort: c.Config.Port, + ListenAddresses: c.Config.LocalAddressList, + Families: families, + UseMultiplePaths: c.UseMultiplePaths.Config.Enabled, + RouteSelectionOptions: &api.RouteSelectionOptionsConfig{ + AlwaysCompareMed: c.RouteSelectionOptions.Config.AlwaysCompareMed, + IgnoreAsPathLength: c.RouteSelectionOptions.Config.IgnoreAsPathLength, + ExternalCompareRouterId: c.RouteSelectionOptions.Config.ExternalCompareRouterId, + AdvertiseInactiveRoutes: c.RouteSelectionOptions.Config.AdvertiseInactiveRoutes, + EnableAigp: c.RouteSelectionOptions.Config.EnableAigp, + IgnoreNextHopIgpMetric: c.RouteSelectionOptions.Config.IgnoreNextHopIgpMetric, + DisableBestPathSelection: c.RouteSelectionOptions.Config.DisableBestPathSelection, + }, + DefaultRouteDistance: &api.DefaultRouteDistance{ + ExternalRouteDistance: uint32(c.DefaultRouteDistance.Config.ExternalRouteDistance), + InternalRouteDistance: uint32(c.DefaultRouteDistance.Config.InternalRouteDistance), + }, + Confederation: &api.Confederation{ + Enabled: c.Confederation.Config.Enabled, + Identifier: c.Confederation.Config.Identifier, + MemberAsList: c.Confederation.Config.MemberAsList, + }, + GracefulRestart: &api.GracefulRestart{ + Enabled: c.GracefulRestart.Config.Enabled, + RestartTime: uint32(c.GracefulRestart.Config.RestartTime), + StaleRoutesTime: uint32(c.GracefulRestart.Config.StaleRoutesTime), + HelperOnly: c.GracefulRestart.Config.HelperOnly, + DeferralTime: uint32(c.GracefulRestart.Config.DeferralTime), + NotificationEnabled: c.GracefulRestart.Config.NotificationEnabled, + LonglivedEnabled: c.GracefulRestart.Config.LongLivedEnabled, + }, + ApplyPolicy: applyPolicy, + } +} + +func (s *Server) StartServer(ctx context.Context, arg *api.StartServerRequest) (*api.StartServerResponse, error) { + if arg == nil || arg.Global == nil { + return nil, fmt.Errorf("invalid request") + } + g := arg.Global + if net.ParseIP(g.RouterId) == nil { + return nil, fmt.Errorf("invalid router-id format: %s", g.RouterId) + } + + global := NewGlobalFromAPIStruct(arg.Global) + + return &api.StartServerResponse{}, s.bgpServer.Start(global) +} + +func (s *Server) StopServer(ctx context.Context, arg *api.StopServerRequest) (*api.StopServerResponse, error) { + return &api.StopServerResponse{}, s.bgpServer.Stop() +} + +func (s *Server) GetRibInfo(ctx context.Context, arg *api.GetRibInfoRequest) (*api.GetRibInfoResponse, error) { + if arg == nil || arg.Info == nil { + return nil, fmt.Errorf("invalid request") + } + family := bgp.RouteFamily(arg.Info.Family) + var in bool + var err error + var info *table.TableInfo + switch arg.Info.Type { + case api.Resource_GLOBAL, api.Resource_LOCAL: + info, err = s.bgpServer.GetRibInfo(arg.Info.Name, family) + case api.Resource_ADJ_IN: + in = true + fallthrough + case api.Resource_ADJ_OUT: + info, err = s.bgpServer.GetAdjRibInfo(arg.Info.Name, family, in) + default: + return nil, fmt.Errorf("unsupported resource type: %s", arg.Info.Type) + } + + if err != nil { + return nil, err + } + + return &api.GetRibInfoResponse{ + Info: &api.TableInfo{ + Type: arg.Info.Type, + Family: arg.Info.Family, + Name: arg.Info.Name, + NumDestination: uint64(info.NumDestination), + NumPath: uint64(info.NumPath), + NumAccepted: uint64(info.NumAccepted), + }, + }, nil +} + +func (s *Server) AddCollector(ctx context.Context, arg *api.AddCollectorRequest) (*api.AddCollectorResponse, error) { + return &api.AddCollectorResponse{}, s.bgpServer.AddCollector(&config.CollectorConfig{ + Url: arg.Url, + DbName: arg.DbName, + TableDumpInterval: arg.TableDumpInterval, + }) +} + +func (s *Server) Shutdown(ctx context.Context, arg *api.ShutdownRequest) (*api.ShutdownResponse, error) { + s.bgpServer.Shutdown() + return &api.ShutdownResponse{}, nil +} diff --git a/pkg/server/mrt.go b/pkg/server/mrt.go new file mode 100644 index 00000000..c654cb78 --- /dev/null +++ b/pkg/server/mrt.go @@ -0,0 +1,409 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "bytes" + "fmt" + "os" + "time" + + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + "github.com/osrg/gobgp/pkg/packet/mrt" + + log "github.com/sirupsen/logrus" +) + +const ( + MIN_ROTATION_INTERVAL = 60 + MIN_DUMP_INTERVAL = 60 +) + +type mrtWriter struct { + dead chan struct{} + s *BgpServer + c *config.MrtConfig + file *os.File + rotationInterval uint64 + dumpInterval uint64 +} + +func (m *mrtWriter) Stop() { + close(m.dead) +} + +func (m *mrtWriter) loop() error { + ops := []WatchOption{} + switch m.c.DumpType { + case config.MRT_TYPE_UPDATES: + ops = append(ops, WatchUpdate(false)) + case config.MRT_TYPE_TABLE: + if len(m.c.TableName) > 0 { + ops = append(ops, WatchTableName(m.c.TableName)) + } + } + w := m.s.Watch(ops...) + rotator := func() *time.Ticker { + if m.rotationInterval == 0 { + return &time.Ticker{} + } + return time.NewTicker(time.Second * time.Duration(m.rotationInterval)) + }() + dump := func() *time.Ticker { + if m.dumpInterval == 0 { + return &time.Ticker{} + } + return time.NewTicker(time.Second * time.Duration(m.dumpInterval)) + }() + + defer func() { + if m.file != nil { + m.file.Close() + } + if m.rotationInterval != 0 { + rotator.Stop() + } + if m.dumpInterval != 0 { + dump.Stop() + } + w.Stop() + }() + + for { + serialize := func(ev WatchEvent) []*mrt.MRTMessage { + msg := make([]*mrt.MRTMessage, 0, 1) + switch e := ev.(type) { + case *WatchEventUpdate: + if e.Init { + return nil + } + mp := mrt.NewBGP4MPMessage(e.PeerAS, e.LocalAS, 0, e.PeerAddress.String(), e.LocalAddress.String(), e.FourBytesAs, nil) + mp.BGPMessagePayload = e.Payload + isAddPath := e.Neighbor.IsAddPathReceiveEnabled(e.PathList[0].GetRouteFamily()) + subtype := mrt.MESSAGE + switch { + case isAddPath && e.FourBytesAs: + subtype = mrt.MESSAGE_AS4_ADDPATH + case isAddPath: + subtype = mrt.MESSAGE_ADDPATH + case e.FourBytesAs: + subtype = mrt.MESSAGE_AS4 + } + if bm, err := mrt.NewMRTMessage(uint32(e.Timestamp.Unix()), mrt.BGP4MP, subtype, mp); err != nil { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Data": e, + "Error": err, + }).Warnf("Failed to create MRT BGP4MP message (subtype %d)", subtype) + } else { + msg = append(msg, bm) + } + case *WatchEventTable: + t := uint32(time.Now().Unix()) + + peers := make([]*mrt.Peer, 1, len(e.Neighbor)+1) + // Adding dummy Peer record for locally generated routes + peers[0] = mrt.NewPeer("0.0.0.0", "0.0.0.0", 0, true) + neighborMap := make(map[string]*config.Neighbor) + for _, pconf := range e.Neighbor { + peers = append(peers, mrt.NewPeer(pconf.State.RemoteRouterId, pconf.State.NeighborAddress, pconf.Config.PeerAs, true)) + neighborMap[pconf.State.NeighborAddress] = pconf + } + + if bm, err := mrt.NewMRTMessage(t, mrt.TABLE_DUMPv2, mrt.PEER_INDEX_TABLE, mrt.NewPeerIndexTable(e.RouterId, "", peers)); err != nil { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Data": e, + "Error": err, + }).Warnf("Failed to create MRT TABLE_DUMPv2 message (subtype %d)", mrt.PEER_INDEX_TABLE) + break + } else { + msg = append(msg, bm) + } + + idx := func(p *table.Path) uint16 { + for i, pconf := range e.Neighbor { + if p.GetSource().Address.String() == pconf.State.NeighborAddress { + return uint16(i) + } + } + return uint16(len(e.Neighbor)) + } + + subtype := func(p *table.Path, isAddPath bool) mrt.MRTSubTypeTableDumpv2 { + t := mrt.RIB_GENERIC + switch p.GetRouteFamily() { + case bgp.RF_IPv4_UC: + t = mrt.RIB_IPV4_UNICAST + case bgp.RF_IPv4_MC: + t = mrt.RIB_IPV4_MULTICAST + case bgp.RF_IPv6_UC: + t = mrt.RIB_IPV6_UNICAST + case bgp.RF_IPv6_MC: + t = mrt.RIB_IPV6_MULTICAST + } + if isAddPath { + // Shift non-additional-path version to *_ADDPATH + t += 6 + } + return t + } + + seq := uint32(0) + appendTableDumpMsg := func(path *table.Path, entries []*mrt.RibEntry, isAddPath bool) { + st := subtype(path, isAddPath) + if bm, err := mrt.NewMRTMessage(t, mrt.TABLE_DUMPv2, st, mrt.NewRib(seq, path.GetNlri(), entries)); err != nil { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Data": e, + "Error": err, + }).Warnf("Failed to create MRT TABLE_DUMPv2 message (subtype %d)", st) + } else { + msg = append(msg, bm) + seq++ + } + } + for _, pathList := range e.PathList { + entries := make([]*mrt.RibEntry, 0, len(pathList)) + entriesAddPath := make([]*mrt.RibEntry, 0, len(pathList)) + for _, path := range pathList { + isAddPath := false + if path.IsLocal() { + isAddPath = true + } else if neighbor, ok := neighborMap[path.GetSource().Address.String()]; ok { + isAddPath = neighbor.IsAddPathReceiveEnabled(path.GetRouteFamily()) + } + if !isAddPath { + entries = append(entries, mrt.NewRibEntry(idx(path), uint32(path.GetTimestamp().Unix()), 0, path.GetPathAttrs(), false)) + } else { + entriesAddPath = append(entriesAddPath, mrt.NewRibEntry(idx(path), uint32(path.GetTimestamp().Unix()), path.GetNlri().PathIdentifier(), path.GetPathAttrs(), true)) + } + } + if len(entries) > 0 { + appendTableDumpMsg(pathList[0], entries, false) + } + if len(entriesAddPath) > 0 { + appendTableDumpMsg(pathList[0], entriesAddPath, true) + } + } + } + return msg + } + + drain := func(ev WatchEvent) { + events := make([]WatchEvent, 0, 1+len(w.Event())) + if ev != nil { + events = append(events, ev) + } + + for len(w.Event()) > 0 { + events = append(events, <-w.Event()) + } + + w := func(buf []byte) { + if _, err := m.file.Write(buf); err == nil { + m.file.Sync() + } else { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Error": err, + }).Warn("Can't write to destination MRT file") + } + } + + var b bytes.Buffer + for _, e := range events { + for _, m := range serialize(e) { + if buf, err := m.Serialize(); err != nil { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Data": e, + "Error": err, + }).Warn("Failed to serialize event") + } else { + b.Write(buf) + if b.Len() > 1*1000*1000 { + w(b.Bytes()) + b.Reset() + } + } + } + } + if b.Len() > 0 { + w(b.Bytes()) + } + } + rotate := func() { + m.file.Close() + file, err := mrtFileOpen(m.c.FileName, m.rotationInterval) + if err == nil { + m.file = file + } else { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Error": err, + }).Warn("can't rotate MRT file") + } + } + + select { + case <-m.dead: + drain(nil) + return nil + case e := <-w.Event(): + drain(e) + if m.c.DumpType == config.MRT_TYPE_TABLE && m.rotationInterval != 0 { + rotate() + } + case <-rotator.C: + if m.c.DumpType == config.MRT_TYPE_UPDATES { + rotate() + } else { + w.Generate(WATCH_EVENT_TYPE_TABLE) + } + case <-dump.C: + w.Generate(WATCH_EVENT_TYPE_TABLE) + } + } +} + +func mrtFileOpen(filename string, interval uint64) (*os.File, error) { + realname := filename + if interval != 0 { + realname = time.Now().Format(filename) + } + log.WithFields(log.Fields{ + "Topic": "mrt", + "Filename": realname, + "Dump Interval": interval, + }).Debug("Setting new MRT destination file") + + i := len(realname) + for i > 0 && os.IsPathSeparator(realname[i-1]) { + // skip trailing path separators + i-- + } + j := i + + for j > 0 && !os.IsPathSeparator(realname[j-1]) { + j-- + } + + if j > 0 { + if err := os.MkdirAll(realname[0:j-1], 0755); err != nil { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Error": err, + }).Warn("can't create MRT destination directory") + return nil, err + } + } + + file, err := os.OpenFile(realname, os.O_CREATE|os.O_RDWR|os.O_APPEND, 0644) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "mrt", + "Error": err, + }).Warn("can't create MRT destination file") + } + return file, err +} + +func newMrtWriter(s *BgpServer, c *config.MrtConfig, rInterval, dInterval uint64) (*mrtWriter, error) { + file, err := mrtFileOpen(c.FileName, rInterval) + if err != nil { + return nil, err + } + m := mrtWriter{ + s: s, + c: c, + file: file, + rotationInterval: rInterval, + dumpInterval: dInterval, + } + go m.loop() + return &m, nil +} + +type mrtManager struct { + bgpServer *BgpServer + writer map[string]*mrtWriter +} + +func (m *mrtManager) enable(c *config.MrtConfig) error { + if _, ok := m.writer[c.FileName]; ok { + return fmt.Errorf("%s already exists", c.FileName) + } + + rInterval := c.RotationInterval + dInterval := c.DumpInterval + + setRotationMin := func() { + if rInterval < MIN_ROTATION_INTERVAL { + log.WithFields(log.Fields{ + "Topic": "MRT", + }).Infof("minimum mrt rotation interval is %d seconds", MIN_ROTATION_INTERVAL) + rInterval = MIN_ROTATION_INTERVAL + } + } + + if c.DumpType == config.MRT_TYPE_TABLE { + if rInterval == 0 { + if dInterval < MIN_DUMP_INTERVAL { + log.WithFields(log.Fields{ + "Topic": "MRT", + }).Infof("minimum mrt dump interval is %d seconds", MIN_DUMP_INTERVAL) + dInterval = MIN_DUMP_INTERVAL + } + } else if dInterval == 0 { + setRotationMin() + } else { + return fmt.Errorf("can't specify both intervals in the table dump type") + } + } else if c.DumpType == config.MRT_TYPE_UPDATES { + // ignore the dump interval + dInterval = 0 + if len(c.TableName) > 0 { + return fmt.Errorf("can't specify the table name with the update dump type") + } + setRotationMin() + } + + w, err := newMrtWriter(m.bgpServer, c, rInterval, dInterval) + if err == nil { + m.writer[c.FileName] = w + } + return err +} + +func (m *mrtManager) disable(c *config.MrtConfig) error { + w, ok := m.writer[c.FileName] + if !ok { + return fmt.Errorf("%s doesn't exists", c.FileName) + } + w.Stop() + delete(m.writer, c.FileName) + return nil +} + +func newMrtManager(s *BgpServer) *mrtManager { + return &mrtManager{ + bgpServer: s, + writer: make(map[string]*mrtWriter), + } +} diff --git a/pkg/server/peer.go b/pkg/server/peer.go new file mode 100644 index 00000000..1f3dd8d8 --- /dev/null +++ b/pkg/server/peer.go @@ -0,0 +1,522 @@ +// Copyright (C) 2014-2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "fmt" + "net" + "time" + + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + + "github.com/eapache/channels" + log "github.com/sirupsen/logrus" +) + +const ( + FLOP_THRESHOLD = time.Second * 30 + MIN_CONNECT_RETRY = 10 +) + +type PeerGroup struct { + Conf *config.PeerGroup + members map[string]config.Neighbor + dynamicNeighbors map[string]*config.DynamicNeighbor +} + +func NewPeerGroup(c *config.PeerGroup) *PeerGroup { + return &PeerGroup{ + Conf: c, + members: make(map[string]config.Neighbor), + dynamicNeighbors: make(map[string]*config.DynamicNeighbor), + } +} + +func (pg *PeerGroup) AddMember(c config.Neighbor) { + pg.members[c.State.NeighborAddress] = c +} + +func (pg *PeerGroup) DeleteMember(c config.Neighbor) { + delete(pg.members, c.State.NeighborAddress) +} + +func (pg *PeerGroup) AddDynamicNeighbor(c *config.DynamicNeighbor) { + pg.dynamicNeighbors[c.Config.Prefix] = c +} + +func newDynamicPeer(g *config.Global, neighborAddress string, pg *config.PeerGroup, loc *table.TableManager, policy *table.RoutingPolicy) *Peer { + conf := config.Neighbor{ + Config: config.NeighborConfig{ + PeerGroup: pg.Config.PeerGroupName, + }, + State: config.NeighborState{ + NeighborAddress: neighborAddress, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + PassiveMode: true, + }, + }, + } + if err := config.OverwriteNeighborConfigWithPeerGroup(&conf, pg); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": neighborAddress, + }).Debugf("Can't overwrite neighbor config: %s", err) + return nil + } + if err := config.SetDefaultNeighborConfigValues(&conf, pg, g); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": neighborAddress, + }).Debugf("Can't set default config: %s", err) + return nil + } + peer := NewPeer(g, &conf, loc, policy) + peer.fsm.state = bgp.BGP_FSM_ACTIVE + return peer +} + +type Peer struct { + tableId string + fsm *FSM + adjRibIn *table.AdjRib + outgoing *channels.InfiniteChannel + policy *table.RoutingPolicy + localRib *table.TableManager + prefixLimitWarned map[bgp.RouteFamily]bool + llgrEndChs []chan struct{} +} + +func NewPeer(g *config.Global, conf *config.Neighbor, loc *table.TableManager, policy *table.RoutingPolicy) *Peer { + peer := &Peer{ + outgoing: channels.NewInfiniteChannel(), + localRib: loc, + policy: policy, + fsm: NewFSM(g, conf, policy), + prefixLimitWarned: make(map[bgp.RouteFamily]bool), + } + if peer.isRouteServerClient() { + peer.tableId = conf.State.NeighborAddress + } else { + peer.tableId = table.GLOBAL_RIB_NAME + } + rfs, _ := config.AfiSafis(conf.AfiSafis).ToRfList() + peer.adjRibIn = table.NewAdjRib(rfs) + return peer +} + +func (peer *Peer) AS() uint32 { + return peer.fsm.pConf.State.PeerAs +} + +func (peer *Peer) ID() string { + return peer.fsm.pConf.State.NeighborAddress +} + +func (peer *Peer) TableID() string { + return peer.tableId +} + +func (peer *Peer) isIBGPPeer() bool { + return peer.fsm.pConf.State.PeerAs == peer.fsm.gConf.Config.As +} + +func (peer *Peer) isRouteServerClient() bool { + return peer.fsm.pConf.RouteServer.Config.RouteServerClient +} + +func (peer *Peer) isRouteReflectorClient() bool { + return peer.fsm.pConf.RouteReflector.Config.RouteReflectorClient +} + +func (peer *Peer) isGracefulRestartEnabled() bool { + return peer.fsm.pConf.GracefulRestart.State.Enabled +} + +func (peer *Peer) getAddPathMode(family bgp.RouteFamily) bgp.BGPAddPathMode { + if mode, y := peer.fsm.rfMap[family]; y { + return mode + } + return bgp.BGP_ADD_PATH_NONE +} + +func (peer *Peer) isAddPathReceiveEnabled(family bgp.RouteFamily) bool { + return (peer.getAddPathMode(family) & bgp.BGP_ADD_PATH_RECEIVE) > 0 +} + +func (peer *Peer) isAddPathSendEnabled(family bgp.RouteFamily) bool { + return (peer.getAddPathMode(family) & bgp.BGP_ADD_PATH_SEND) > 0 +} + +func (peer *Peer) isDynamicNeighbor() bool { + return peer.fsm.pConf.Config.NeighborAddress == "" && peer.fsm.pConf.Config.NeighborInterface == "" +} + +func (peer *Peer) recvedAllEOR() bool { + for _, a := range peer.fsm.pConf.AfiSafis { + if s := a.MpGracefulRestart.State; s.Enabled && !s.EndOfRibReceived { + return false + } + } + return true +} + +func (peer *Peer) configuredRFlist() []bgp.RouteFamily { + rfs, _ := config.AfiSafis(peer.fsm.pConf.AfiSafis).ToRfList() + return rfs +} + +func (peer *Peer) negotiatedRFList() []bgp.RouteFamily { + l := make([]bgp.RouteFamily, 0, len(peer.fsm.rfMap)) + for family, _ := range peer.fsm.rfMap { + l = append(l, family) + } + return l +} + +func (peer *Peer) toGlobalFamilies(families []bgp.RouteFamily) []bgp.RouteFamily { + if peer.fsm.pConf.Config.Vrf != "" { + fs := make([]bgp.RouteFamily, 0, len(families)) + for _, f := range families { + switch f { + case bgp.RF_IPv4_UC: + fs = append(fs, bgp.RF_IPv4_VPN) + case bgp.RF_IPv6_UC: + fs = append(fs, bgp.RF_IPv6_VPN) + default: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Family": f, + "VRF": peer.fsm.pConf.Config.Vrf, + }).Warn("invalid family configured for neighbor with vrf") + } + } + families = fs + } + return families +} + +func classifyFamilies(all, part []bgp.RouteFamily) ([]bgp.RouteFamily, []bgp.RouteFamily) { + a := []bgp.RouteFamily{} + b := []bgp.RouteFamily{} + for _, f := range all { + p := true + for _, g := range part { + if f == g { + p = false + a = append(a, f) + break + } + } + if p { + b = append(b, f) + } + } + return a, b +} + +func (peer *Peer) forwardingPreservedFamilies() ([]bgp.RouteFamily, []bgp.RouteFamily) { + list := []bgp.RouteFamily{} + for _, a := range peer.fsm.pConf.AfiSafis { + if s := a.MpGracefulRestart.State; s.Enabled && s.Received { + list = append(list, a.State.Family) + } + } + return classifyFamilies(peer.configuredRFlist(), list) +} + +func (peer *Peer) llgrFamilies() ([]bgp.RouteFamily, []bgp.RouteFamily) { + list := []bgp.RouteFamily{} + for _, a := range peer.fsm.pConf.AfiSafis { + if a.LongLivedGracefulRestart.State.Enabled { + list = append(list, a.State.Family) + } + } + return classifyFamilies(peer.configuredRFlist(), list) +} + +func (peer *Peer) isLLGREnabledFamily(family bgp.RouteFamily) bool { + if !peer.fsm.pConf.GracefulRestart.Config.LongLivedEnabled { + return false + } + fs, _ := peer.llgrFamilies() + for _, f := range fs { + if f == family { + return true + } + } + return false +} + +func (peer *Peer) llgrRestartTime(family bgp.RouteFamily) uint32 { + for _, a := range peer.fsm.pConf.AfiSafis { + if a.State.Family == family { + return a.LongLivedGracefulRestart.State.PeerRestartTime + } + } + return 0 +} + +func (peer *Peer) llgrRestartTimerExpired(family bgp.RouteFamily) bool { + all := true + for _, a := range peer.fsm.pConf.AfiSafis { + if a.State.Family == family { + a.LongLivedGracefulRestart.State.PeerRestartTimerExpired = true + } + s := a.LongLivedGracefulRestart.State + if s.Received && !s.PeerRestartTimerExpired { + all = false + } + } + return all +} + +func (peer *Peer) markLLGRStale(fs []bgp.RouteFamily) []*table.Path { + paths := peer.adjRibIn.PathList(fs, true) + for i, p := range paths { + doStale := true + for _, c := range p.GetCommunities() { + if c == uint32(bgp.COMMUNITY_NO_LLGR) { + doStale = false + p = p.Clone(true) + break + } + } + if doStale { + p = p.Clone(false) + p.SetCommunities([]uint32{uint32(bgp.COMMUNITY_LLGR_STALE)}, false) + } + paths[i] = p + } + return paths +} + +func (peer *Peer) stopPeerRestarting() { + peer.fsm.pConf.GracefulRestart.State.PeerRestarting = false + for _, ch := range peer.llgrEndChs { + close(ch) + } + peer.llgrEndChs = make([]chan struct{}, 0) + +} + +func (peer *Peer) filterPathFromSourcePeer(path, old *table.Path) *table.Path { + if peer.ID() != path.GetSource().Address.String() { + return path + } + + // Note: Multiple paths having the same prefix could exist the withdrawals + // list in the case of Route Server setup with import policies modifying + // paths. In such case, gobgp sends duplicated update messages; withdraw + // messages for the same prefix. + if !peer.isRouteServerClient() { + if peer.isRouteReflectorClient() && path.GetRouteFamily() == bgp.RF_RTC_UC { + // When the peer is a Route Reflector client and the given path + // contains the Route Tartget Membership NLRI, the path should not + // be withdrawn in order to signal the client to distribute routes + // with the specific RT to Route Reflector. + return path + } else if !path.IsWithdraw && old != nil && old.GetSource().Address.String() != peer.ID() { + // Say, peer A and B advertized same prefix P, and best path + // calculation chose a path from B as best. When B withdraws prefix + // P, best path calculation chooses the path from A as best. For + // peers other than A, this path should be advertised (as implicit + // withdrawal). However for A, we should advertise the withdrawal + // path. Thing is same when peer A and we advertized prefix P (as + // local route), then, we withdraws the prefix. + return old.Clone(true) + } + } + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Data": path, + }).Debug("From me, ignore.") + return nil +} + +func (peer *Peer) doPrefixLimit(k bgp.RouteFamily, c *config.PrefixLimitConfig) *bgp.BGPMessage { + if maxPrefixes := int(c.MaxPrefixes); maxPrefixes > 0 { + count := peer.adjRibIn.Count([]bgp.RouteFamily{k}) + pct := int(c.ShutdownThresholdPct) + if pct > 0 && !peer.prefixLimitWarned[k] && count > (maxPrefixes*pct/100) { + peer.prefixLimitWarned[k] = true + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "AddressFamily": k.String(), + }).Warnf("prefix limit %d%% reached", pct) + } + if count > maxPrefixes { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "AddressFamily": k.String(), + }).Warnf("prefix limit reached") + return bgp.NewBGPNotificationMessage(bgp.BGP_ERROR_CEASE, bgp.BGP_ERROR_SUB_MAXIMUM_NUMBER_OF_PREFIXES_REACHED, nil) + } + } + return nil + +} + +func (peer *Peer) updatePrefixLimitConfig(c []config.AfiSafi) error { + x := peer.fsm.pConf.AfiSafis + y := c + if len(x) != len(y) { + return fmt.Errorf("changing supported afi-safi is not allowed") + } + m := make(map[bgp.RouteFamily]config.PrefixLimitConfig) + for _, e := range x { + m[e.State.Family] = e.PrefixLimit.Config + } + for _, e := range y { + if p, ok := m[e.State.Family]; !ok { + return fmt.Errorf("changing supported afi-safi is not allowed") + } else if !p.Equal(&e.PrefixLimit.Config) { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "AddressFamily": e.Config.AfiSafiName, + "OldMaxPrefixes": p.MaxPrefixes, + "NewMaxPrefixes": e.PrefixLimit.Config.MaxPrefixes, + "OldShutdownThresholdPct": p.ShutdownThresholdPct, + "NewShutdownThresholdPct": e.PrefixLimit.Config.ShutdownThresholdPct, + }).Warnf("update prefix limit configuration") + peer.prefixLimitWarned[e.State.Family] = false + if msg := peer.doPrefixLimit(e.State.Family, &e.PrefixLimit.Config); msg != nil { + sendFsmOutgoingMsg(peer, nil, msg, true) + } + } + } + peer.fsm.pConf.AfiSafis = c + return nil +} + +func (peer *Peer) handleUpdate(e *FsmMsg) ([]*table.Path, []bgp.RouteFamily, *bgp.BGPMessage) { + m := e.MsgData.(*bgp.BGPMessage) + update := m.Body.(*bgp.BGPUpdate) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.fsm.pConf.State.NeighborAddress, + "nlri": update.NLRI, + "withdrawals": update.WithdrawnRoutes, + "attributes": update.PathAttributes, + }).Debug("received update") + peer.fsm.pConf.Timers.State.UpdateRecvTime = time.Now().Unix() + if len(e.PathList) > 0 { + paths := make([]*table.Path, 0, len(e.PathList)) + eor := []bgp.RouteFamily{} + for _, path := range e.PathList { + if path.IsEOR() { + family := path.GetRouteFamily() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "AddressFamily": family, + }).Debug("EOR received") + eor = append(eor, family) + continue + } + // RFC4271 9.1.2 Phase 2: Route Selection + // + // If the AS_PATH attribute of a BGP route contains an AS loop, the BGP + // route should be excluded from the Phase 2 decision function. + if aspath := path.GetAsPath(); aspath != nil { + if hasOwnASLoop(peer.fsm.peerInfo.LocalAS, int(peer.fsm.pConf.AsPathOptions.Config.AllowOwnAs), aspath) { + path.SetAsLooped(true) + continue + } + } + paths = append(paths, path) + } + peer.adjRibIn.Update(e.PathList) + for _, af := range peer.fsm.pConf.AfiSafis { + if msg := peer.doPrefixLimit(af.State.Family, &af.PrefixLimit.Config); msg != nil { + return nil, nil, msg + } + } + return paths, eor, nil + } + return nil, nil, nil +} + +func (peer *Peer) startFSMHandler(incoming *channels.InfiniteChannel, stateCh chan *FsmMsg) { + peer.fsm.h = NewFSMHandler(peer.fsm, incoming, stateCh, peer.outgoing) +} + +func (peer *Peer) StaleAll(rfList []bgp.RouteFamily) []*table.Path { + return peer.adjRibIn.StaleAll(rfList) +} + +func (peer *Peer) PassConn(conn *net.TCPConn) { + select { + case peer.fsm.connCh <- conn: + default: + conn.Close() + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + }).Warn("accepted conn is closed to avoid be blocked") + } +} + +func (peer *Peer) DropAll(rfList []bgp.RouteFamily) { + peer.adjRibIn.Drop(rfList) +} + +func (peer *Peer) stopFSM() error { + failed := false + addr := peer.fsm.pConf.State.NeighborAddress + t1 := time.AfterFunc(time.Minute*5, func() { + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Warnf("Failed to free the fsm.h.t for %s", addr) + failed = true + }) + peer.fsm.h.t.Kill(nil) + peer.fsm.h.t.Wait() + t1.Stop() + if !failed { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Debug("freed fsm.h.t") + cleanInfiniteChannel(peer.outgoing) + } + failed = false + t2 := time.AfterFunc(time.Minute*5, func() { + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Warnf("Failed to free the fsm.t for %s", addr) + failed = true + }) + peer.fsm.t.Kill(nil) + peer.fsm.t.Wait() + t2.Stop() + if !failed { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Debug("freed fsm.t") + return nil + } + return fmt.Errorf("Failed to free FSM for %s", addr) +} diff --git a/pkg/server/rpki.go b/pkg/server/rpki.go new file mode 100644 index 00000000..606b18ab --- /dev/null +++ b/pkg/server/rpki.go @@ -0,0 +1,712 @@ +// Copyright (C) 2015,2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "encoding/binary" + "fmt" + "io" + "net" + "sort" + "strconv" + "time" + + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + "github.com/osrg/gobgp/pkg/packet/rtr" + + "github.com/armon/go-radix" + log "github.com/sirupsen/logrus" + "golang.org/x/net/context" +) + +const ( + CONNECT_RETRY_INTERVAL = 30 +) + +func before(a, b uint32) bool { + return int32(a-b) < 0 +} + +type RoaBucket struct { + Prefix *table.IPPrefix + entries []*table.ROA +} + +func (r *RoaBucket) GetEntries() []*table.ROA { + return r.entries +} + +type roas []*table.ROA + +func (r roas) Len() int { + return len(r) +} + +func (r roas) Swap(i, j int) { + r[i], r[j] = r[j], r[i] +} + +func (r roas) Less(i, j int) bool { + r1 := r[i] + r2 := r[j] + + if r1.MaxLen < r2.MaxLen { + return true + } else if r1.MaxLen > r2.MaxLen { + return false + } + + if r1.AS < r2.AS { + return true + } + return false +} + +type ROAEventType uint8 + +const ( + CONNECTED ROAEventType = iota + DISCONNECTED + RTR + LIFETIMEOUT +) + +type ROAEvent struct { + EventType ROAEventType + Src string + Data []byte + conn *net.TCPConn +} + +type roaManager struct { + AS uint32 + Roas map[bgp.RouteFamily]*radix.Tree + eventCh chan *ROAEvent + clientMap map[string]*roaClient +} + +func NewROAManager(as uint32) (*roaManager, error) { + m := &roaManager{ + AS: as, + Roas: make(map[bgp.RouteFamily]*radix.Tree), + } + m.Roas[bgp.RF_IPv4_UC] = radix.New() + m.Roas[bgp.RF_IPv6_UC] = radix.New() + m.eventCh = make(chan *ROAEvent) + m.clientMap = make(map[string]*roaClient) + return m, nil +} + +func (c *roaManager) enabled() bool { + return len(c.clientMap) != 0 +} + +func (m *roaManager) SetAS(as uint32) error { + if m.AS != 0 { + return fmt.Errorf("AS was already configured") + } + m.AS = as + return nil +} + +func (m *roaManager) AddServer(host string, lifetime int64) error { + address, port, err := net.SplitHostPort(host) + if err != nil { + return err + } + if lifetime == 0 { + lifetime = 3600 + } + if _, ok := m.clientMap[host]; ok { + return fmt.Errorf("ROA server exists %s", host) + } + m.clientMap[host] = NewRoaClient(address, port, m.eventCh, lifetime) + return nil +} + +func (m *roaManager) DeleteServer(host string) error { + client, ok := m.clientMap[host] + if !ok { + return fmt.Errorf("ROA server doesn't exists %s", host) + } + client.stop() + m.deleteAllROA(host) + delete(m.clientMap, host) + return nil +} + +func (m *roaManager) deleteAllROA(network string) { + for _, tree := range m.Roas { + deleteKeys := make([]string, 0, tree.Len()) + tree.Walk(func(s string, v interface{}) bool { + b, _ := v.(*RoaBucket) + newEntries := make([]*table.ROA, 0, len(b.entries)) + for _, r := range b.entries { + if r.Src != network { + newEntries = append(newEntries, r) + } + } + if len(newEntries) > 0 { + b.entries = newEntries + } else { + deleteKeys = append(deleteKeys, s) + } + return false + }) + for _, key := range deleteKeys { + tree.Delete(key) + } + } +} + +func (m *roaManager) Enable(address string) error { + for network, client := range m.clientMap { + add, _, _ := net.SplitHostPort(network) + if add == address { + client.enable(client.serialNumber) + return nil + } + } + return fmt.Errorf("ROA server not found %s", address) +} + +func (m *roaManager) Disable(address string) error { + for network, client := range m.clientMap { + add, _, _ := net.SplitHostPort(network) + if add == address { + client.reset() + m.deleteAllROA(add) + return nil + } + } + return fmt.Errorf("ROA server not found %s", address) +} + +func (m *roaManager) Reset(address string) error { + return m.Disable(address) +} + +func (m *roaManager) SoftReset(address string) error { + for network, client := range m.clientMap { + add, _, _ := net.SplitHostPort(network) + if add == address { + client.softReset() + m.deleteAllROA(network) + return nil + } + } + return fmt.Errorf("ROA server not found %s", address) +} + +func (c *roaManager) ReceiveROA() chan *ROAEvent { + return c.eventCh +} + +func (c *roaClient) lifetimeout() { + c.eventCh <- &ROAEvent{ + EventType: LIFETIMEOUT, + Src: c.host, + } +} + +func (m *roaManager) HandleROAEvent(ev *ROAEvent) { + client, y := m.clientMap[ev.Src] + if !y { + if ev.EventType == CONNECTED { + ev.conn.Close() + } + log.WithFields(log.Fields{"Topic": "rpki"}).Errorf("Can't find %s ROA server configuration", ev.Src) + return + } + switch ev.EventType { + case DISCONNECTED: + log.WithFields(log.Fields{"Topic": "rpki"}).Infof("ROA server %s is disconnected", ev.Src) + client.state.Downtime = time.Now().Unix() + // clear state + client.endOfData = false + client.pendingROAs = make([]*table.ROA, 0) + client.state.RpkiMessages = config.RpkiMessages{} + client.conn = nil + go client.tryConnect() + client.timer = time.AfterFunc(time.Duration(client.lifetime)*time.Second, client.lifetimeout) + client.oldSessionID = client.sessionID + case CONNECTED: + log.WithFields(log.Fields{"Topic": "rpki"}).Infof("ROA server %s is connected", ev.Src) + client.conn = ev.conn + client.state.Uptime = time.Now().Unix() + go client.established() + case RTR: + m.handleRTRMsg(client, &client.state, ev.Data) + case LIFETIMEOUT: + // a) already reconnected but hasn't received + // EndOfData -> needs to delete stale ROAs + // b) not reconnected -> needs to delete stale ROAs + // + // c) already reconnected and received EndOfData so + // all stale ROAs were deleted -> timer was cancelled + // so should not be here. + if client.oldSessionID != client.sessionID { + log.WithFields(log.Fields{"Topic": "rpki"}).Infof("Reconnected to %s. Ignore timeout", client.host) + } else { + log.WithFields(log.Fields{"Topic": "rpki"}).Infof("Deleting all ROAs due to timeout with:%s", client.host) + m.deleteAllROA(client.host) + } + } +} + +func (m *roaManager) roa2tree(roa *table.ROA) (*radix.Tree, string) { + tree := m.Roas[bgp.RF_IPv4_UC] + if roa.Family == bgp.AFI_IP6 { + tree = m.Roas[bgp.RF_IPv6_UC] + } + return tree, table.IpToRadixkey(roa.Prefix.Prefix, roa.Prefix.Length) +} + +func (m *roaManager) deleteROA(roa *table.ROA) { + tree, key := m.roa2tree(roa) + b, _ := tree.Get(key) + if b != nil { + bucket := b.(*RoaBucket) + newEntries := make([]*table.ROA, 0, len(bucket.entries)) + for _, r := range bucket.entries { + if !r.Equal(roa) { + newEntries = append(newEntries, r) + } + } + if len(newEntries) != len(bucket.entries) { + bucket.entries = newEntries + if len(newEntries) == 0 { + tree.Delete(key) + } + return + } + } + log.WithFields(log.Fields{ + "Topic": "rpki", + "Prefix": roa.Prefix.Prefix.String(), + "Prefix Length": roa.Prefix.Length, + "AS": roa.AS, + "Max Length": roa.MaxLen, + }).Info("Can't withdraw a ROA") +} + +func (m *roaManager) DeleteROA(roa *table.ROA) { + m.deleteROA(roa) +} + +func (m *roaManager) addROA(roa *table.ROA) { + tree, key := m.roa2tree(roa) + b, _ := tree.Get(key) + var bucket *RoaBucket + if b == nil { + bucket = &RoaBucket{ + Prefix: roa.Prefix, + entries: make([]*table.ROA, 0), + } + tree.Insert(key, bucket) + } else { + bucket = b.(*RoaBucket) + for _, r := range bucket.entries { + if r.Equal(roa) { + // we already have the same one + return + } + } + } + bucket.entries = append(bucket.entries, roa) +} + +func (m *roaManager) AddROA(roa *table.ROA) { + m.addROA(roa) +} + +func (c *roaManager) handleRTRMsg(client *roaClient, state *config.RpkiServerState, buf []byte) { + received := &state.RpkiMessages.RpkiReceived + + m, err := rtr.ParseRTR(buf) + if err == nil { + switch msg := m.(type) { + case *rtr.RTRSerialNotify: + if before(client.serialNumber, msg.RTRCommon.SerialNumber) { + client.enable(client.serialNumber) + } else if client.serialNumber == msg.RTRCommon.SerialNumber { + // nothing + } else { + // should not happen. try to get the whole ROAs. + client.softReset() + } + received.SerialNotify++ + case *rtr.RTRSerialQuery: + case *rtr.RTRResetQuery: + case *rtr.RTRCacheResponse: + received.CacheResponse++ + client.endOfData = false + case *rtr.RTRIPPrefix: + family := bgp.AFI_IP + if msg.Type == rtr.RTR_IPV4_PREFIX { + received.Ipv4Prefix++ + } else { + family = bgp.AFI_IP6 + received.Ipv6Prefix++ + } + roa := table.NewROA(family, msg.Prefix, msg.PrefixLen, msg.MaxLen, msg.AS, client.host) + if (msg.Flags & 1) == 1 { + if client.endOfData { + c.addROA(roa) + } else { + client.pendingROAs = append(client.pendingROAs, roa) + } + } else { + c.deleteROA(roa) + } + case *rtr.RTREndOfData: + received.EndOfData++ + if client.sessionID != msg.RTRCommon.SessionID { + // remove all ROAs related with the + // previous session + c.deleteAllROA(client.host) + } + client.sessionID = msg.RTRCommon.SessionID + client.serialNumber = msg.RTRCommon.SerialNumber + client.endOfData = true + if client.timer != nil { + client.timer.Stop() + client.timer = nil + } + for _, roa := range client.pendingROAs { + c.addROA(roa) + } + client.pendingROAs = make([]*table.ROA, 0) + case *rtr.RTRCacheReset: + client.softReset() + received.CacheReset++ + case *rtr.RTRErrorReport: + received.Error++ + } + } else { + log.WithFields(log.Fields{ + "Topic": "rpki", + "Host": client.host, + "Error": err, + }).Info("Failed to parse an RTR message") + } +} + +func (c *roaManager) GetServers() []*config.RpkiServer { + f := func(tree *radix.Tree) (map[string]uint32, map[string]uint32) { + records := make(map[string]uint32) + prefixes := make(map[string]uint32) + + tree.Walk(func(s string, v interface{}) bool { + b, _ := v.(*RoaBucket) + tmpRecords := make(map[string]uint32) + for _, roa := range b.entries { + tmpRecords[roa.Src]++ + } + + for src, r := range tmpRecords { + if r > 0 { + records[src] += r + prefixes[src]++ + } + } + return false + }) + return records, prefixes + } + + recordsV4, prefixesV4 := f(c.Roas[bgp.RF_IPv4_UC]) + recordsV6, prefixesV6 := f(c.Roas[bgp.RF_IPv6_UC]) + + l := make([]*config.RpkiServer, 0, len(c.clientMap)) + for _, client := range c.clientMap { + state := &client.state + + if client.conn == nil { + state.Up = false + } else { + state.Up = true + } + f := func(m map[string]uint32, key string) uint32 { + if r, ok := m[key]; ok { + return r + } + return 0 + } + state.RecordsV4 = f(recordsV4, client.host) + state.RecordsV6 = f(recordsV6, client.host) + state.PrefixesV4 = f(prefixesV4, client.host) + state.PrefixesV6 = f(prefixesV6, client.host) + state.SerialNumber = client.serialNumber + + addr, port, _ := net.SplitHostPort(client.host) + l = append(l, &config.RpkiServer{ + Config: config.RpkiServerConfig{ + Address: addr, + // Note: RpkiServerConfig.Port is uint32 type, but the TCP/UDP + // port is 16-bit length. + Port: func() uint32 { p, _ := strconv.ParseUint(port, 10, 16); return uint32(p) }(), + }, + State: client.state, + }) + } + return l +} + +func (c *roaManager) GetRoa(family bgp.RouteFamily) ([]*table.ROA, error) { + if len(c.clientMap) == 0 { + return []*table.ROA{}, fmt.Errorf("RPKI server isn't configured.") + } + var rfList []bgp.RouteFamily + switch family { + case bgp.RF_IPv4_UC: + rfList = []bgp.RouteFamily{bgp.RF_IPv4_UC} + case bgp.RF_IPv6_UC: + rfList = []bgp.RouteFamily{bgp.RF_IPv6_UC} + default: + rfList = []bgp.RouteFamily{bgp.RF_IPv4_UC, bgp.RF_IPv6_UC} + } + l := make([]*table.ROA, 0) + for _, rf := range rfList { + if tree, ok := c.Roas[rf]; ok { + tree.Walk(func(s string, v interface{}) bool { + b, _ := v.(*RoaBucket) + var roaList roas + for _, r := range b.entries { + roaList = append(roaList, r) + } + sort.Sort(roaList) + for _, roa := range roaList { + l = append(l, roa) + } + return false + }) + } + } + return l, nil +} + +func ValidatePath(ownAs uint32, tree *radix.Tree, cidr string, asPath *bgp.PathAttributeAsPath) *table.Validation { + var as uint32 + + validation := &table.Validation{ + Status: config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND, + Reason: table.RPKI_VALIDATION_REASON_TYPE_NONE, + Matched: make([]*table.ROA, 0), + UnmatchedLength: make([]*table.ROA, 0), + UnmatchedAs: make([]*table.ROA, 0), + } + + if asPath == nil || len(asPath.Value) == 0 { + as = ownAs + } else { + param := asPath.Value[len(asPath.Value)-1] + switch param.GetType() { + case bgp.BGP_ASPATH_ATTR_TYPE_SEQ: + asList := param.GetAS() + if len(asList) == 0 { + as = ownAs + } else { + as = asList[len(asList)-1] + } + case bgp.BGP_ASPATH_ATTR_TYPE_CONFED_SET, bgp.BGP_ASPATH_ATTR_TYPE_CONFED_SEQ: + as = ownAs + default: + return validation + } + } + _, n, _ := net.ParseCIDR(cidr) + ones, _ := n.Mask.Size() + prefixLen := uint8(ones) + key := table.IpToRadixkey(n.IP, prefixLen) + _, b, _ := tree.LongestPrefix(key) + if b == nil { + return validation + } + + var bucket *RoaBucket + fn := radix.WalkFn(func(k string, v interface{}) bool { + bucket, _ = v.(*RoaBucket) + for _, r := range bucket.entries { + if prefixLen <= r.MaxLen { + if r.AS != 0 && r.AS == as { + validation.Matched = append(validation.Matched, r) + } else { + validation.UnmatchedAs = append(validation.UnmatchedAs, r) + } + } else { + validation.UnmatchedLength = append(validation.UnmatchedLength, r) + } + } + return false + }) + tree.WalkPath(key, fn) + + if len(validation.Matched) != 0 { + validation.Status = config.RPKI_VALIDATION_RESULT_TYPE_VALID + validation.Reason = table.RPKI_VALIDATION_REASON_TYPE_NONE + } else if len(validation.UnmatchedAs) != 0 { + validation.Status = config.RPKI_VALIDATION_RESULT_TYPE_INVALID + validation.Reason = table.RPKI_VALIDATION_REASON_TYPE_AS + } else if len(validation.UnmatchedLength) != 0 { + validation.Status = config.RPKI_VALIDATION_RESULT_TYPE_INVALID + validation.Reason = table.RPKI_VALIDATION_REASON_TYPE_LENGTH + } else { + validation.Status = config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND + validation.Reason = table.RPKI_VALIDATION_REASON_TYPE_NONE + } + + return validation +} + +func (c *roaManager) validate(path *table.Path) *table.Validation { + if len(c.clientMap) == 0 || path.IsWithdraw || path.IsEOR() { + // RPKI isn't enabled or invalid path + return nil + } + if tree, ok := c.Roas[path.GetRouteFamily()]; ok { + return ValidatePath(c.AS, tree, path.GetNlri().String(), path.GetAsPath()) + } + return nil +} + +type roaClient struct { + host string + conn *net.TCPConn + state config.RpkiServerState + eventCh chan *ROAEvent + sessionID uint16 + oldSessionID uint16 + serialNumber uint32 + timer *time.Timer + lifetime int64 + endOfData bool + pendingROAs []*table.ROA + cancelfnc context.CancelFunc + ctx context.Context +} + +func NewRoaClient(address, port string, ch chan *ROAEvent, lifetime int64) *roaClient { + ctx, cancel := context.WithCancel(context.Background()) + c := &roaClient{ + host: net.JoinHostPort(address, port), + eventCh: ch, + lifetime: lifetime, + pendingROAs: make([]*table.ROA, 0), + ctx: ctx, + cancelfnc: cancel, + } + go c.tryConnect() + return c +} + +func (c *roaClient) enable(serial uint32) error { + if c.conn != nil { + r := rtr.NewRTRSerialQuery(c.sessionID, serial) + data, _ := r.Serialize() + _, err := c.conn.Write(data) + if err != nil { + return err + } + c.state.RpkiMessages.RpkiSent.SerialQuery++ + } + return nil +} + +func (c *roaClient) softReset() error { + if c.conn != nil { + r := rtr.NewRTRResetQuery() + data, _ := r.Serialize() + _, err := c.conn.Write(data) + if err != nil { + return err + } + c.state.RpkiMessages.RpkiSent.ResetQuery++ + c.endOfData = false + c.pendingROAs = make([]*table.ROA, 0) + } + return nil +} + +func (c *roaClient) reset() { + if c.conn != nil { + c.conn.Close() + } +} + +func (c *roaClient) stop() { + c.cancelfnc() + c.reset() +} + +func (c *roaClient) tryConnect() { + for { + select { + case <-c.ctx.Done(): + return + default: + } + if conn, err := net.Dial("tcp", c.host); err != nil { + // better to use context with timeout + time.Sleep(CONNECT_RETRY_INTERVAL * time.Second) + } else { + c.eventCh <- &ROAEvent{ + EventType: CONNECTED, + Src: c.host, + conn: conn.(*net.TCPConn), + } + return + } + } +} + +func (c *roaClient) established() (err error) { + defer func() { + c.conn.Close() + c.eventCh <- &ROAEvent{ + EventType: DISCONNECTED, + Src: c.host, + } + }() + + if err := c.softReset(); err != nil { + return err + } + + for { + header := make([]byte, rtr.RTR_MIN_LEN) + if _, err = io.ReadFull(c.conn, header); err != nil { + return err + } + totalLen := binary.BigEndian.Uint32(header[4:8]) + if totalLen < rtr.RTR_MIN_LEN { + return fmt.Errorf("too short header length %v", totalLen) + } + + body := make([]byte, totalLen-rtr.RTR_MIN_LEN) + if _, err = io.ReadFull(c.conn, body); err != nil { + return + } + + c.eventCh <- &ROAEvent{ + EventType: RTR, + Src: c.host, + Data: append(header, body...), + } + } +} diff --git a/pkg/server/rpki_test.go b/pkg/server/rpki_test.go new file mode 100644 index 00000000..0d77dbb2 --- /dev/null +++ b/pkg/server/rpki_test.go @@ -0,0 +1,257 @@ +// Copyright (C) 2015 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "net" + "strconv" + "strings" + "testing" + + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + + radix "github.com/armon/go-radix" + "github.com/stretchr/testify/assert" +) + +func strToASParam(str string) *bgp.PathAttributeAsPath { + toList := func(asstr, sep string) []uint32 { + as := make([]uint32, 0) + l := strings.Split(asstr, sep) + for _, s := range l { + v, _ := strconv.ParseUint(s, 10, 32) + as = append(as, uint32(v)) + } + return as + } + var atype uint8 + var as []uint32 + if strings.HasPrefix(str, "{") { + atype = bgp.BGP_ASPATH_ATTR_TYPE_SET + as = toList(str[1:len(str)-1], ",") + } else if strings.HasPrefix(str, "(") { + atype = bgp.BGP_ASPATH_ATTR_TYPE_CONFED_SET + as = toList(str[1:len(str)-1], " ") + } else { + atype = bgp.BGP_ASPATH_ATTR_TYPE_SEQ + as = toList(str, " ") + } + + return bgp.NewPathAttributeAsPath([]bgp.AsPathParamInterface{bgp.NewAs4PathParam(atype, as)}) +} + +func validateOne(tree *radix.Tree, cidr, aspathStr string) config.RpkiValidationResultType { + r := ValidatePath(65500, tree, cidr, strToASParam(aspathStr)) + return r.Status +} + +func TestValidate0(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("192.168.0.0").To4(), 24, 32, 100, "")) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("192.168.0.0").To4(), 24, 24, 200, "")) + + var r config.RpkiValidationResultType + + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "192.168.0.0/24", "100") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) + + r = validateOne(tree, "192.168.0.0/24", "100 200") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) + + r = validateOne(tree, "192.168.0.0/24", "300") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) + + r = validateOne(tree, "192.168.0.0/25", "100") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) + + r = validateOne(tree, "192.168.0.0/25", "200") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) + + r = validateOne(tree, "192.168.0.0/25", "300") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) +} + +func TestValidate1(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 16, 65000, "")) + + var r config.RpkiValidationResultType + + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/16", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) + + r = validateOne(tree, "10.0.0.0/16", "65001") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) +} + +func TestValidate2(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + + var r config.RpkiValidationResultType + + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/16", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) + + r = validateOne(tree, "10.0.0.0/16", "65001") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) +} + +func TestValidate3(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 16, 65000, "")) + + var r config.RpkiValidationResultType + + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/8", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) + + r = validateOne(tree, "10.0.0.0/17", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) + + manager, _ = NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 24, 65000, "")) + + tree = manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/17", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) +} + +func TestValidate4(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 16, 65000, "")) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 16, 65001, "")) + + var r config.RpkiValidationResultType + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/16", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) + + r = validateOne(tree, "10.0.0.0/16", "65001") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) +} + +func TestValidate5(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 17, 17, 65000, "")) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.128.0").To4(), 17, 17, 65000, "")) + + var r config.RpkiValidationResultType + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/16", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) +} + +func TestValidate6(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 8, 32, 0, "")) + + var r config.RpkiValidationResultType + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/7", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) + + r = validateOne(tree, "10.0.0.0/8", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) + + r = validateOne(tree, "10.0.0.0/24", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) +} + +func TestValidate7(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 24, 65000, "")) + + var r config.RpkiValidationResultType + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/24", "{65000}") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) + + r = validateOne(tree, "10.0.0.0/24", "{65001}") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) + + r = validateOne(tree, "10.0.0.0/24", "{65000,65001}") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_NOT_FOUND) +} + +func TestValidate8(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 24, 0, "")) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 24, 65000, "")) + + var r config.RpkiValidationResultType + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/24", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) + + r = validateOne(tree, "10.0.0.0/24", "65001") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) +} + +func TestValidate9(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 24, 24, 65000, "")) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 24, 65001, "")) + + var r config.RpkiValidationResultType + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/24", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) + + r = validateOne(tree, "10.0.0.0/24", "65001") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) +} + +func TestValidate10(t *testing.T) { + assert := assert.New(t) + + manager, _ := NewROAManager(0) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 24, 24, 0, "")) + manager.addROA(table.NewROA(bgp.AFI_IP, net.ParseIP("10.0.0.0").To4(), 16, 24, 65001, "")) + + var r config.RpkiValidationResultType + tree := manager.Roas[bgp.RF_IPv4_UC] + r = validateOne(tree, "10.0.0.0/24", "65000") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_INVALID) + + r = validateOne(tree, "10.0.0.0/24", "65001") + assert.Equal(r, config.RPKI_VALIDATION_RESULT_TYPE_VALID) +} diff --git a/pkg/server/server.go b/pkg/server/server.go new file mode 100644 index 00000000..b00bb55d --- /dev/null +++ b/pkg/server/server.go @@ -0,0 +1,3048 @@ +// Copyright (C) 2014-2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "bytes" + "fmt" + "net" + "strconv" + "sync" + "time" + + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + + "github.com/eapache/channels" + uuid "github.com/satori/go.uuid" + log "github.com/sirupsen/logrus" +) + +type TCPListener struct { + l *net.TCPListener + ch chan struct{} +} + +func (l *TCPListener) Close() error { + if err := l.l.Close(); err != nil { + return err + } + t := time.NewTicker(time.Second) + select { + case <-l.ch: + case <-t.C: + return fmt.Errorf("close timeout") + } + return nil +} + +// avoid mapped IPv6 address +func NewTCPListener(address string, port uint32, ch chan *net.TCPConn) (*TCPListener, error) { + proto := "tcp4" + if ip := net.ParseIP(address); ip == nil { + return nil, fmt.Errorf("can't listen on %s", address) + } else if ip.To4() == nil { + proto = "tcp6" + } + addr, err := net.ResolveTCPAddr(proto, net.JoinHostPort(address, strconv.Itoa(int(port)))) + if err != nil { + return nil, err + } + + l, err := net.ListenTCP(proto, addr) + if err != nil { + return nil, err + } + // Note: Set TTL=255 for incoming connection listener in order to accept + // connection in case for the neighbor has TTL Security settings. + if err := SetListenTcpTTLSockopt(l, 255); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warnf("cannot set TTL(=%d) for TCPListener: %s", 255, err) + } + + closeCh := make(chan struct{}) + go func() error { + for { + conn, err := l.AcceptTCP() + if err != nil { + close(closeCh) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Error": err, + }).Warn("Failed to AcceptTCP") + return err + } + ch <- conn + } + }() + return &TCPListener{ + l: l, + ch: closeCh, + }, nil +} + +type BgpServer struct { + bgpConfig config.Bgp + fsmincomingCh *channels.InfiniteChannel + fsmStateCh chan *FsmMsg + acceptCh chan *net.TCPConn + + mgmtCh chan *mgmtOp + policy *table.RoutingPolicy + listeners []*TCPListener + neighborMap map[string]*Peer + peerGroupMap map[string]*PeerGroup + globalRib *table.TableManager + rsRib *table.TableManager + roaManager *roaManager + shutdownWG *sync.WaitGroup + watcherMap map[WatchEventType][]*Watcher + zclient *zebraClient + bmpManager *bmpClientManager + mrtManager *mrtManager + uuidMap map[uuid.UUID]string +} + +func NewBgpServer() *BgpServer { + roaManager, _ := NewROAManager(0) + s := &BgpServer{ + neighborMap: make(map[string]*Peer), + peerGroupMap: make(map[string]*PeerGroup), + policy: table.NewRoutingPolicy(), + roaManager: roaManager, + mgmtCh: make(chan *mgmtOp, 1), + watcherMap: make(map[WatchEventType][]*Watcher), + uuidMap: make(map[uuid.UUID]string), + } + s.bmpManager = newBmpClientManager(s) + s.mrtManager = newMrtManager(s) + return s +} + +func (server *BgpServer) Listeners(addr string) []*net.TCPListener { + list := make([]*net.TCPListener, 0, len(server.listeners)) + rhs := net.ParseIP(addr).To4() != nil + for _, l := range server.listeners { + host, _, _ := net.SplitHostPort(l.l.Addr().String()) + lhs := net.ParseIP(host).To4() != nil + if lhs == rhs { + list = append(list, l.l) + } + } + return list +} + +func (s *BgpServer) active() error { + if s.bgpConfig.Global.Config.As == 0 { + return fmt.Errorf("bgp server hasn't started yet") + } + return nil +} + +type mgmtOp struct { + f func() error + errCh chan error + checkActive bool // check BGP global setting is configured before calling f() +} + +func (server *BgpServer) handleMGMTOp(op *mgmtOp) { + if op.checkActive { + if err := server.active(); err != nil { + op.errCh <- err + return + } + } + op.errCh <- op.f() +} + +func (s *BgpServer) mgmtOperation(f func() error, checkActive bool) (err error) { + ch := make(chan error) + defer func() { err = <-ch }() + s.mgmtCh <- &mgmtOp{ + f: f, + errCh: ch, + checkActive: checkActive, + } + return +} + +func (server *BgpServer) Serve() { + server.listeners = make([]*TCPListener, 0, 2) + server.fsmincomingCh = channels.NewInfiniteChannel() + server.fsmStateCh = make(chan *FsmMsg, 4096) + + handleFsmMsg := func(e *FsmMsg) { + peer, found := server.neighborMap[e.MsgSrc] + if !found { + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Warnf("Can't find the neighbor %s", e.MsgSrc) + return + } + if e.Version != peer.fsm.version { + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Debug("FSM version inconsistent") + return + } + server.handleFSMMessage(peer, e) + } + + for { + passConn := func(conn *net.TCPConn) { + host, _, _ := net.SplitHostPort(conn.RemoteAddr().String()) + ipaddr, _ := net.ResolveIPAddr("ip", host) + remoteAddr := ipaddr.String() + peer, found := server.neighborMap[remoteAddr] + if found { + if peer.fsm.adminState != ADMIN_STATE_UP { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Remote Addr": remoteAddr, + "Admin State": peer.fsm.adminState, + }).Debug("New connection for non admin-state-up peer") + conn.Close() + return + } + localAddrValid := func(laddr string) bool { + if laddr == "0.0.0.0" || laddr == "::" { + return true + } + l := conn.LocalAddr() + if l == nil { + // already closed + return false + } + + host, _, _ := net.SplitHostPort(l.String()) + if host != laddr { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": remoteAddr, + "Configured addr": laddr, + "Addr": host, + }).Info("Mismatched local address") + return false + } + return true + }(peer.fsm.pConf.Transport.Config.LocalAddress) + + if !localAddrValid { + conn.Close() + return + } + + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Debugf("Accepted a new passive connection from:%s", remoteAddr) + peer.PassConn(conn) + } else if pg := server.matchLongestDynamicNeighborPrefix(remoteAddr); pg != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Debugf("Accepted a new dynamic neighbor from:%s", remoteAddr) + rib := server.globalRib + if pg.Conf.RouteServer.Config.RouteServerClient { + rib = server.rsRib + } + peer := newDynamicPeer(&server.bgpConfig.Global, remoteAddr, pg.Conf, rib, server.policy) + if peer == nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": remoteAddr, + }).Infof("Can't create new Dynamic Peer") + conn.Close() + return + } + server.policy.Reset(nil, map[string]config.ApplyPolicy{peer.ID(): peer.fsm.pConf.ApplyPolicy}) + server.neighborMap[remoteAddr] = peer + peer.startFSMHandler(server.fsmincomingCh, server.fsmStateCh) + server.broadcastPeerState(peer, bgp.BGP_FSM_ACTIVE, nil) + peer.PassConn(conn) + } else { + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Infof("Can't find configuration for a new passive connection from:%s", remoteAddr) + conn.Close() + } + } + + select { + case op := <-server.mgmtCh: + server.handleMGMTOp(op) + case conn := <-server.acceptCh: + passConn(conn) + default: + } + + for { + select { + case e := <-server.fsmStateCh: + handleFsmMsg(e) + default: + goto CONT + } + } + CONT: + + select { + case op := <-server.mgmtCh: + server.handleMGMTOp(op) + case rmsg := <-server.roaManager.ReceiveROA(): + server.roaManager.HandleROAEvent(rmsg) + case conn := <-server.acceptCh: + passConn(conn) + case e, ok := <-server.fsmincomingCh.Out(): + if !ok { + continue + } + handleFsmMsg(e.(*FsmMsg)) + case e := <-server.fsmStateCh: + handleFsmMsg(e) + } + } +} + +func (server *BgpServer) matchLongestDynamicNeighborPrefix(a string) *PeerGroup { + ipAddr := net.ParseIP(a) + longestMask := net.CIDRMask(0, 32).String() + var longestPG *PeerGroup + for _, pg := range server.peerGroupMap { + for _, d := range pg.dynamicNeighbors { + _, netAddr, _ := net.ParseCIDR(d.Config.Prefix) + if netAddr.Contains(ipAddr) { + if netAddr.Mask.String() > longestMask { + longestMask = netAddr.Mask.String() + longestPG = pg + } + } + } + } + return longestPG +} + +func sendFsmOutgoingMsg(peer *Peer, paths []*table.Path, notification *bgp.BGPMessage, stayIdle bool) { + peer.outgoing.In() <- &FsmOutgoingMsg{ + Paths: paths, + Notification: notification, + StayIdle: stayIdle, + } +} + +func isASLoop(peer *Peer, path *table.Path) bool { + for _, as := range path.GetAsList() { + if as == peer.AS() { + return true + } + } + return false +} + +func filterpath(peer *Peer, path, old *table.Path) *table.Path { + if path == nil { + return nil + } + if _, ok := peer.fsm.rfMap[path.GetRouteFamily()]; !ok { + return nil + } + + //RFC4684 Constrained Route Distribution + if _, y := peer.fsm.rfMap[bgp.RF_RTC_UC]; y && path.GetRouteFamily() != bgp.RF_RTC_UC { + ignore := true + for _, ext := range path.GetExtCommunities() { + for _, p := range peer.adjRibIn.PathList([]bgp.RouteFamily{bgp.RF_RTC_UC}, true) { + rt := p.GetNlri().(*bgp.RouteTargetMembershipNLRI).RouteTarget + // Note: nil RT means the default route target + if rt == nil || ext.String() == rt.String() { + ignore = false + break + } + } + if !ignore { + break + } + } + if ignore { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Data": path, + }).Debug("Filtered by Route Target Constraint, ignore") + return nil + } + } + + //iBGP handling + if peer.isIBGPPeer() { + ignore := false + if !path.IsLocal() { + ignore = true + info := path.GetSource() + //if the path comes from eBGP peer + if info.AS != peer.AS() { + ignore = false + } + // RFC4456 8. Avoiding Routing Information Loops + // A router that recognizes the ORIGINATOR_ID attribute SHOULD + // ignore a route received with its BGP Identifier as the ORIGINATOR_ID. + if id := path.GetOriginatorID(); peer.fsm.gConf.Config.RouterId == id.String() { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "OriginatorID": id, + "Data": path, + }).Debug("Originator ID is mine, ignore") + return nil + } + if info.RouteReflectorClient { + ignore = false + } + if peer.isRouteReflectorClient() { + // RFC4456 8. Avoiding Routing Information Loops + // If the local CLUSTER_ID is found in the CLUSTER_LIST, + // the advertisement received SHOULD be ignored. + for _, clusterID := range path.GetClusterList() { + if clusterID.Equal(peer.fsm.peerInfo.RouteReflectorClusterID) { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "ClusterID": clusterID, + "Data": path, + }).Debug("cluster list path attribute has local cluster id, ignore") + return nil + } + } + ignore = false + } + } + + if ignore { + if !path.IsWithdraw && old != nil { + oldSource := old.GetSource() + if old.IsLocal() || oldSource.Address.String() != peer.ID() && oldSource.AS != peer.AS() { + // In this case, we suppose this peer has the same prefix + // received from another iBGP peer. + // So we withdraw the old best which was injected locally + // (from CLI or gRPC for example) in order to avoid the + // old best left on peers. + // Also, we withdraw the eBGP route which is the old best. + // When we got the new best from iBGP, we don't advertise + // the new best and need to withdraw the old best. + return old.Clone(true) + } + } + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Data": path, + }).Debug("From same AS, ignore.") + return nil + } + } + + if path = peer.filterPathFromSourcePeer(path, old); path == nil { + return nil + } + + if !peer.isRouteServerClient() && isASLoop(peer, path) { + return nil + } + return path +} + +func (s *BgpServer) filterpath(peer *Peer, path, old *table.Path) *table.Path { + // Special handling for RTM NLRI. + if path != nil && path.GetRouteFamily() == bgp.RF_RTC_UC && !path.IsWithdraw { + // If the given "path" is locally generated and the same with "old", we + // assumes "path" was already sent before. This assumption avoids the + // infinite UPDATE loop between Route Reflector and its clients. + if path.IsLocal() && path == old { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.fsm.pConf.State.NeighborAddress, + "Path": path, + }).Debug("given rtm nlri is already sent, skipping to advertise") + return nil + } + + if old != nil && old.IsLocal() { + // We assumes VRF with the specific RT is deleted. + path = old.Clone(true) + } else if peer.isRouteReflectorClient() { + // We need to send the path even if the peer is originator of the + // path in order to signal that the client should distribute route + // with the given RT. + } else { + // We send a path even if it is not the best path. See comments in + // (*Destination) GetChanges(). + dst := peer.localRib.GetDestination(path) + path = nil + for _, p := range dst.GetKnownPathList(peer.TableID(), peer.AS()) { + srcPeer := p.GetSource() + if peer.ID() != srcPeer.Address.String() { + if srcPeer.RouteReflectorClient { + // The path from a RR client is preferred than others + // for the case that RR and non RR client peering + // (e.g., peering of different RR clusters). + path = p + break + } else if path == nil { + path = p + } + } + } + } + } + + // only allow vpnv4 and vpnv6 paths to be advertised to VRFed neighbors. + // also check we can import this path using table.CanImportToVrf() + // if we can, make it local path by calling (*Path).ToLocal() + if path != nil && peer.fsm.pConf.Config.Vrf != "" { + if f := path.GetRouteFamily(); f != bgp.RF_IPv4_VPN && f != bgp.RF_IPv6_VPN { + return nil + } + vrf := peer.localRib.Vrfs[peer.fsm.pConf.Config.Vrf] + if table.CanImportToVrf(vrf, path) { + path = path.ToLocal() + } else { + return nil + } + } + + // replace-peer-as handling + if path != nil && !path.IsWithdraw && peer.fsm.pConf.AsPathOptions.State.ReplacePeerAs { + path = path.ReplaceAS(peer.fsm.pConf.Config.LocalAs, peer.fsm.pConf.Config.PeerAs) + } + + if path = filterpath(peer, path, old); path == nil { + return nil + } + + options := &table.PolicyOptions{ + Info: peer.fsm.peerInfo, + OldNextHop: path.GetNexthop(), + } + path = table.UpdatePathAttrs(peer.fsm.gConf, peer.fsm.pConf, peer.fsm.peerInfo, path) + + if v := s.roaManager.validate(path); v != nil { + options.ValidationResult = v + } + + path = peer.policy.ApplyPolicy(peer.TableID(), table.POLICY_DIRECTION_EXPORT, path, options) + // When 'path' is filtered (path == nil), check 'old' has been sent to this peer. + // If it has, send withdrawal to the peer. + if path == nil && old != nil { + o := peer.policy.ApplyPolicy(peer.TableID(), table.POLICY_DIRECTION_EXPORT, old, options) + if o != nil { + path = old.Clone(true) + } + } + + // draft-uttaro-idr-bgp-persistence-02 + // 4.3. Processing LLGR_STALE Routes + // + // The route SHOULD NOT be advertised to any neighbor from which the + // Long-lived Graceful Restart Capability has not been received. The + // exception is described in the Optional Partial Deployment + // Procedure section (Section 4.7). Note that this requirement + // implies that such routes should be withdrawn from any such neighbor. + if path != nil && !path.IsWithdraw && !peer.isLLGREnabledFamily(path.GetRouteFamily()) && path.IsLLGRStale() { + // we send unnecessary withdrawn even if we didn't + // sent the route. + path = path.Clone(true) + } + + // remove local-pref attribute + // we should do this after applying export policy since policy may + // set local-preference + if path != nil && !peer.isIBGPPeer() && !peer.isRouteServerClient() { + path.RemoveLocalPref() + } + return path +} + +func clonePathList(pathList []*table.Path) []*table.Path { + l := make([]*table.Path, 0, len(pathList)) + for _, p := range pathList { + if p != nil { + l = append(l, p.Clone(p.IsWithdraw)) + } + } + return l +} + +func (server *BgpServer) notifyBestWatcher(best []*table.Path, multipath [][]*table.Path) { + if table.SelectionOptions.DisableBestPathSelection { + // Note: If best path selection disabled, no best path to notify. + return + } + clonedM := make([][]*table.Path, len(multipath)) + for i, pathList := range multipath { + clonedM[i] = clonePathList(pathList) + } + clonedB := clonePathList(best) + m := make(map[string]uint16) + for _, p := range clonedB { + switch p.GetRouteFamily() { + case bgp.RF_IPv4_VPN, bgp.RF_IPv6_VPN: + for _, vrf := range server.globalRib.Vrfs { + if vrf.Id != 0 && table.CanImportToVrf(vrf, p) { + m[p.GetNlri().String()] = uint16(vrf.Id) + } + } + } + } + w := &WatchEventBestPath{PathList: clonedB, MultiPathList: clonedM} + if len(m) > 0 { + w.Vrf = m + } + server.notifyWatcher(WATCH_EVENT_TYPE_BEST_PATH, w) +} + +func (s *BgpServer) ToConfig(peer *Peer, getAdvertised bool) *config.Neighbor { + // create copy which can be access to without mutex + conf := *peer.fsm.pConf + + conf.AfiSafis = make([]config.AfiSafi, len(peer.fsm.pConf.AfiSafis)) + for i, af := range peer.fsm.pConf.AfiSafis { + conf.AfiSafis[i] = af + conf.AfiSafis[i].AddPaths.State.Receive = peer.isAddPathReceiveEnabled(af.State.Family) + if peer.isAddPathSendEnabled(af.State.Family) { + conf.AfiSafis[i].AddPaths.State.SendMax = af.AddPaths.State.SendMax + } else { + conf.AfiSafis[i].AddPaths.State.SendMax = 0 + } + } + + remoteCap := make([]bgp.ParameterCapabilityInterface, 0, len(peer.fsm.capMap)) + for _, caps := range peer.fsm.capMap { + for _, m := range caps { + // need to copy all values here + buf, _ := m.Serialize() + c, _ := bgp.DecodeCapability(buf) + remoteCap = append(remoteCap, c) + } + } + conf.State.RemoteCapabilityList = remoteCap + conf.State.LocalCapabilityList = capabilitiesFromConfig(peer.fsm.pConf) + + conf.State.SessionState = config.IntToSessionStateMap[int(peer.fsm.state)] + conf.State.AdminState = config.IntToAdminStateMap[int(peer.fsm.adminState)] + + if peer.fsm.state == bgp.BGP_FSM_ESTABLISHED { + rfList := peer.configuredRFlist() + if getAdvertised { + pathList, filtered := s.getBestFromLocal(peer, rfList) + conf.State.AdjTable.Advertised = uint32(len(pathList)) + conf.State.AdjTable.Filtered = uint32(len(filtered)) + } else { + conf.State.AdjTable.Advertised = 0 + } + conf.State.AdjTable.Received = uint32(peer.adjRibIn.Count(rfList)) + conf.State.AdjTable.Accepted = uint32(peer.adjRibIn.Accepted(rfList)) + + conf.Transport.State.LocalAddress, conf.Transport.State.LocalPort = peer.fsm.LocalHostPort() + _, conf.Transport.State.RemotePort = peer.fsm.RemoteHostPort() + buf, _ := peer.fsm.recvOpen.Serialize() + // need to copy all values here + conf.State.ReceivedOpenMessage, _ = bgp.ParseBGPMessage(buf) + conf.State.RemoteRouterId = peer.fsm.peerInfo.ID.To4().String() + } + return &conf +} + +func (server *BgpServer) notifyPrePolicyUpdateWatcher(peer *Peer, pathList []*table.Path, msg *bgp.BGPMessage, timestamp time.Time, payload []byte) { + if !server.isWatched(WATCH_EVENT_TYPE_PRE_UPDATE) || peer == nil { + return + } + cloned := clonePathList(pathList) + if len(cloned) == 0 { + return + } + _, y := peer.fsm.capMap[bgp.BGP_CAP_FOUR_OCTET_AS_NUMBER] + l, _ := peer.fsm.LocalHostPort() + ev := &WatchEventUpdate{ + Message: msg, + PeerAS: peer.fsm.peerInfo.AS, + LocalAS: peer.fsm.peerInfo.LocalAS, + PeerAddress: peer.fsm.peerInfo.Address, + LocalAddress: net.ParseIP(l), + PeerID: peer.fsm.peerInfo.ID, + FourBytesAs: y, + Timestamp: timestamp, + Payload: payload, + PostPolicy: false, + PathList: cloned, + Neighbor: server.ToConfig(peer, false), + } + server.notifyWatcher(WATCH_EVENT_TYPE_PRE_UPDATE, ev) +} + +func (server *BgpServer) notifyPostPolicyUpdateWatcher(peer *Peer, pathList []*table.Path) { + if !server.isWatched(WATCH_EVENT_TYPE_POST_UPDATE) || peer == nil { + return + } + cloned := clonePathList(pathList) + if len(cloned) == 0 { + return + } + _, y := peer.fsm.capMap[bgp.BGP_CAP_FOUR_OCTET_AS_NUMBER] + l, _ := peer.fsm.LocalHostPort() + ev := &WatchEventUpdate{ + PeerAS: peer.fsm.peerInfo.AS, + LocalAS: peer.fsm.peerInfo.LocalAS, + PeerAddress: peer.fsm.peerInfo.Address, + LocalAddress: net.ParseIP(l), + PeerID: peer.fsm.peerInfo.ID, + FourBytesAs: y, + Timestamp: cloned[0].GetTimestamp(), + PostPolicy: true, + PathList: cloned, + Neighbor: server.ToConfig(peer, false), + } + server.notifyWatcher(WATCH_EVENT_TYPE_POST_UPDATE, ev) +} + +func newWatchEventPeerState(peer *Peer, m *FsmMsg) *WatchEventPeerState { + _, rport := peer.fsm.RemoteHostPort() + laddr, lport := peer.fsm.LocalHostPort() + sentOpen := buildopen(peer.fsm.gConf, peer.fsm.pConf) + recvOpen := peer.fsm.recvOpen + e := &WatchEventPeerState{ + PeerAS: peer.fsm.peerInfo.AS, + LocalAS: peer.fsm.peerInfo.LocalAS, + PeerAddress: peer.fsm.peerInfo.Address, + LocalAddress: net.ParseIP(laddr), + PeerPort: rport, + LocalPort: lport, + PeerID: peer.fsm.peerInfo.ID, + SentOpen: sentOpen, + RecvOpen: recvOpen, + State: peer.fsm.state, + AdminState: peer.fsm.adminState, + Timestamp: time.Now(), + PeerInterface: peer.fsm.pConf.Config.NeighborInterface, + } + + if m != nil { + e.StateReason = m.StateReason + } + return e +} + +func (server *BgpServer) broadcastPeerState(peer *Peer, oldState bgp.FSMState, e *FsmMsg) { + newState := peer.fsm.state + if oldState == bgp.BGP_FSM_ESTABLISHED || newState == bgp.BGP_FSM_ESTABLISHED { + server.notifyWatcher(WATCH_EVENT_TYPE_PEER_STATE, newWatchEventPeerState(peer, e)) + } +} + +func (server *BgpServer) notifyMessageWatcher(peer *Peer, timestamp time.Time, msg *bgp.BGPMessage, isSent bool) { + // validation should be done in the caller of this function + _, y := peer.fsm.capMap[bgp.BGP_CAP_FOUR_OCTET_AS_NUMBER] + l, _ := peer.fsm.LocalHostPort() + ev := &WatchEventMessage{ + Message: msg, + PeerAS: peer.fsm.peerInfo.AS, + LocalAS: peer.fsm.peerInfo.LocalAS, + PeerAddress: peer.fsm.peerInfo.Address, + LocalAddress: net.ParseIP(l), + PeerID: peer.fsm.peerInfo.ID, + FourBytesAs: y, + Timestamp: timestamp, + IsSent: isSent, + } + if !isSent { + server.notifyWatcher(WATCH_EVENT_TYPE_RECV_MSG, ev) + } +} + +func (server *BgpServer) notifyRecvMessageWatcher(peer *Peer, timestamp time.Time, msg *bgp.BGPMessage) { + if peer == nil || !server.isWatched(WATCH_EVENT_TYPE_RECV_MSG) { + return + } + server.notifyMessageWatcher(peer, timestamp, msg, false) +} + +func (s *BgpServer) getBestFromLocal(peer *Peer, rfList []bgp.RouteFamily) ([]*table.Path, []*table.Path) { + pathList := []*table.Path{} + filtered := []*table.Path{} + for _, family := range peer.toGlobalFamilies(rfList) { + pl := func() []*table.Path { + if peer.isAddPathSendEnabled(family) { + return peer.localRib.GetPathList(peer.TableID(), peer.AS(), []bgp.RouteFamily{family}) + } + return peer.localRib.GetBestPathList(peer.TableID(), peer.AS(), []bgp.RouteFamily{family}) + }() + for _, path := range pl { + if p := s.filterpath(peer, path, nil); p != nil { + pathList = append(pathList, p) + } else { + filtered = append(filtered, path) + } + } + } + if peer.isGracefulRestartEnabled() { + for _, family := range rfList { + pathList = append(pathList, table.NewEOR(family)) + } + } + return pathList, filtered +} + +func (s *BgpServer) processOutgoingPaths(peer *Peer, paths, olds []*table.Path) []*table.Path { + if peer.fsm.state != bgp.BGP_FSM_ESTABLISHED { + return nil + } + if peer.fsm.pConf.GracefulRestart.State.LocalRestarting { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.fsm.pConf.State.NeighborAddress, + }).Debug("now syncing, suppress sending updates") + return nil + } + + outgoing := make([]*table.Path, 0, len(paths)) + + for idx, path := range paths { + var old *table.Path + if olds != nil { + old = olds[idx] + } + if p := s.filterpath(peer, path, old); p != nil { + outgoing = append(outgoing, p) + } + } + return outgoing +} + +func (s *BgpServer) handleRouteRefresh(peer *Peer, e *FsmMsg) []*table.Path { + m := e.MsgData.(*bgp.BGPMessage) + rr := m.Body.(*bgp.BGPRouteRefresh) + rf := bgp.AfiSafiToRouteFamily(rr.AFI, rr.SAFI) + if _, ok := peer.fsm.rfMap[rf]; !ok { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Data": rf, + }).Warn("Route family isn't supported") + return nil + } + if _, ok := peer.fsm.capMap[bgp.BGP_CAP_ROUTE_REFRESH]; !ok { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + }).Warn("ROUTE_REFRESH received but the capability wasn't advertised") + return nil + } + rfList := []bgp.RouteFamily{rf} + accepted, filtered := s.getBestFromLocal(peer, rfList) + for _, path := range filtered { + path.IsWithdraw = true + accepted = append(accepted, path) + } + return accepted +} + +func (server *BgpServer) propagateUpdate(peer *Peer, pathList []*table.Path) { + rs := peer != nil && peer.isRouteServerClient() + vrf := !rs && peer != nil && peer.fsm.pConf.Config.Vrf != "" + + tableId := table.GLOBAL_RIB_NAME + rib := server.globalRib + if rs { + tableId = peer.TableID() + rib = server.rsRib + } + + for _, path := range pathList { + if vrf { + path = path.ToGlobal(rib.Vrfs[peer.fsm.pConf.Config.Vrf]) + } + + policyOptions := &table.PolicyOptions{} + + if !rs && peer != nil { + policyOptions.Info = peer.fsm.peerInfo + } + if v := server.roaManager.validate(path); v != nil { + policyOptions.ValidationResult = v + } + + if p := server.policy.ApplyPolicy(tableId, table.POLICY_DIRECTION_IMPORT, path, policyOptions); p != nil { + path = p + } else { + path = path.Clone(true) + } + + if !rs { + server.notifyPostPolicyUpdateWatcher(peer, []*table.Path{path}) + + // RFC4684 Constrained Route Distribution 6. Operation + // + // When a BGP speaker receives a BGP UPDATE that advertises or withdraws + // a given Route Target membership NLRI, it should examine the RIB-OUTs + // of VPN NLRIs and re-evaluate the advertisement status of routes that + // match the Route Target in question. + // + // A BGP speaker should generate the minimum set of BGP VPN route + // updates (advertisements and/or withdraws) necessary to transition + // between the previous and current state of the route distribution + // graph that is derived from Route Target membership information. + if peer != nil && path != nil && path.GetRouteFamily() == bgp.RF_RTC_UC { + rt := path.GetNlri().(*bgp.RouteTargetMembershipNLRI).RouteTarget + fs := make([]bgp.RouteFamily, 0, len(peer.negotiatedRFList())) + for _, f := range peer.negotiatedRFList() { + if f != bgp.RF_RTC_UC { + fs = append(fs, f) + } + } + var candidates []*table.Path + if path.IsWithdraw { + // Note: The paths to be withdrawn are filtered because the + // given RT on RTM NLRI is already removed from adj-RIB-in. + _, candidates = server.getBestFromLocal(peer, fs) + } else { + candidates = server.globalRib.GetBestPathList(peer.TableID(), 0, fs) + } + paths := make([]*table.Path, 0, len(candidates)) + for _, p := range candidates { + for _, ext := range p.GetExtCommunities() { + if rt == nil || ext.String() == rt.String() { + if path.IsWithdraw { + p = p.Clone(true) + } + paths = append(paths, p) + break + } + } + } + if path.IsWithdraw { + // Skips filtering because the paths are already filtered + // and the withdrawal does not need the path attributes. + } else { + paths = server.processOutgoingPaths(peer, paths, nil) + } + sendFsmOutgoingMsg(peer, paths, nil, false) + } + } + + if dsts := rib.Update(path); len(dsts) > 0 { + server.propagateUpdateToNeighbors(peer, path, dsts, true) + } + } +} + +func (server *BgpServer) dropPeerAllRoutes(peer *Peer, families []bgp.RouteFamily) { + rib := server.globalRib + if peer.isRouteServerClient() { + rib = server.rsRib + } + for _, family := range peer.toGlobalFamilies(families) { + for _, path := range rib.GetPathListByPeer(peer.fsm.peerInfo, family) { + p := path.Clone(true) + if dsts := rib.Update(p); len(dsts) > 0 { + server.propagateUpdateToNeighbors(peer, p, dsts, false) + } + } + } +} + +func dstsToPaths(id string, as uint32, dsts []*table.Update) ([]*table.Path, []*table.Path, [][]*table.Path) { + bestList := make([]*table.Path, 0, len(dsts)) + oldList := make([]*table.Path, 0, len(dsts)) + mpathList := make([][]*table.Path, 0, len(dsts)) + + for _, dst := range dsts { + best, old, mpath := dst.GetChanges(id, as, false) + bestList = append(bestList, best) + oldList = append(oldList, old) + if mpath != nil { + mpathList = append(mpathList, mpath) + } + } + return bestList, oldList, mpathList +} + +func (server *BgpServer) propagateUpdateToNeighbors(source *Peer, newPath *table.Path, dsts []*table.Update, needOld bool) { + if table.SelectionOptions.DisableBestPathSelection { + return + } + var gBestList, gOldList, bestList, oldList []*table.Path + var mpathList [][]*table.Path + if source == nil || !source.isRouteServerClient() { + gBestList, gOldList, mpathList = dstsToPaths(table.GLOBAL_RIB_NAME, 0, dsts) + server.notifyBestWatcher(gBestList, mpathList) + } + family := newPath.GetRouteFamily() + for _, targetPeer := range server.neighborMap { + if (source == nil && targetPeer.isRouteServerClient()) || (source != nil && source.isRouteServerClient() != targetPeer.isRouteServerClient()) { + continue + } + f := func() bgp.RouteFamily { + if targetPeer.fsm.pConf.Config.Vrf != "" { + switch family { + case bgp.RF_IPv4_VPN: + return bgp.RF_IPv4_UC + case bgp.RF_IPv6_VPN: + return bgp.RF_IPv6_UC + } + } + return family + }() + if targetPeer.isAddPathSendEnabled(f) { + if newPath.IsWithdraw { + bestList = func() []*table.Path { + l := make([]*table.Path, 0, len(dsts)) + for _, d := range dsts { + l = append(l, d.GetWithdrawnPath()...) + } + return l + }() + } else { + bestList = []*table.Path{newPath} + if newPath.GetRouteFamily() == bgp.RF_RTC_UC { + // we assumes that new "path" nlri was already sent before. This assumption avoids the + // infinite UPDATE loop between Route Reflector and its clients. + for _, old := range dsts[0].OldKnownPathList { + if old.IsLocal() { + bestList = []*table.Path{} + break + } + } + } + } + oldList = nil + } else if targetPeer.isRouteServerClient() { + bestList, oldList, _ = dstsToPaths(targetPeer.TableID(), targetPeer.AS(), dsts) + } else { + bestList = gBestList + oldList = gOldList + } + if !needOld { + oldList = nil + } + if paths := server.processOutgoingPaths(targetPeer, bestList, oldList); len(paths) > 0 { + sendFsmOutgoingMsg(targetPeer, paths, nil, false) + } + } +} + +func (server *BgpServer) handleFSMMessage(peer *Peer, e *FsmMsg) { + switch e.MsgType { + case FSM_MSG_STATE_CHANGE: + nextState := e.MsgData.(bgp.FSMState) + oldState := bgp.FSMState(peer.fsm.pConf.State.SessionState.ToInt()) + peer.fsm.pConf.State.SessionState = config.IntToSessionStateMap[int(nextState)] + peer.fsm.StateChange(nextState) + + // PeerDown + if oldState == bgp.BGP_FSM_ESTABLISHED { + t := time.Now() + if t.Sub(time.Unix(peer.fsm.pConf.Timers.State.Uptime, 0)) < FLOP_THRESHOLD { + peer.fsm.pConf.State.Flops++ + } + var drop []bgp.RouteFamily + if peer.fsm.reason.Type == FSM_GRACEFUL_RESTART { + peer.fsm.pConf.GracefulRestart.State.PeerRestarting = true + var p []bgp.RouteFamily + p, drop = peer.forwardingPreservedFamilies() + server.propagateUpdate(peer, peer.StaleAll(p)) + } else { + drop = peer.configuredRFlist() + } + peer.prefixLimitWarned = make(map[bgp.RouteFamily]bool) + peer.DropAll(drop) + server.dropPeerAllRoutes(peer, drop) + if peer.fsm.pConf.Config.PeerAs == 0 { + peer.fsm.pConf.State.PeerAs = 0 + peer.fsm.peerInfo.AS = 0 + } + if peer.isDynamicNeighbor() { + peer.stopPeerRestarting() + go peer.stopFSM() + delete(server.neighborMap, peer.fsm.pConf.State.NeighborAddress) + server.broadcastPeerState(peer, oldState, e) + return + } + } else if peer.fsm.pConf.GracefulRestart.State.PeerRestarting && nextState == bgp.BGP_FSM_IDLE { + if peer.fsm.pConf.GracefulRestart.State.LongLivedEnabled { + llgr, no_llgr := peer.llgrFamilies() + + peer.DropAll(no_llgr) + server.dropPeerAllRoutes(peer, no_llgr) + + // attach LLGR_STALE community to paths in peer's adj-rib-in + // paths with NO_LLGR are deleted + pathList := peer.markLLGRStale(llgr) + + // calculate again + // wheh path with LLGR_STALE chosen as best, + // peer which doesn't support LLGR will drop the path + // if it is in adj-rib-out, do withdrawal + server.propagateUpdate(peer, pathList) + + for _, f := range llgr { + endCh := make(chan struct{}) + peer.llgrEndChs = append(peer.llgrEndChs, endCh) + go func(family bgp.RouteFamily, endCh chan struct{}) { + t := peer.llgrRestartTime(family) + timer := time.NewTimer(time.Second * time.Duration(t)) + + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Family": family, + }).Debugf("start LLGR restart timer (%d sec) for %s", t, family) + + select { + case <-timer.C: + server.mgmtOperation(func() error { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Family": family, + }).Debugf("LLGR restart timer (%d sec) for %s expired", t, family) + peer.DropAll([]bgp.RouteFamily{family}) + server.dropPeerAllRoutes(peer, []bgp.RouteFamily{family}) + + // when all llgr restart timer expired, stop PeerRestarting + if peer.llgrRestartTimerExpired(family) { + peer.stopPeerRestarting() + } + return nil + }, false) + case <-endCh: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Family": family, + }).Debugf("stop LLGR restart timer (%d sec) for %s", t, family) + } + }(f, endCh) + } + } else { + // RFC 4724 4.2 + // If the session does not get re-established within the "Restart Time" + // that the peer advertised previously, the Receiving Speaker MUST + // delete all the stale routes from the peer that it is retaining. + peer.fsm.pConf.GracefulRestart.State.PeerRestarting = false + peer.DropAll(peer.configuredRFlist()) + server.dropPeerAllRoutes(peer, peer.configuredRFlist()) + } + } + + cleanInfiniteChannel(peer.outgoing) + peer.outgoing = channels.NewInfiniteChannel() + if nextState == bgp.BGP_FSM_ESTABLISHED { + // update for export policy + laddr, _ := peer.fsm.LocalHostPort() + // may include zone info + peer.fsm.pConf.Transport.State.LocalAddress = laddr + // exclude zone info + ipaddr, _ := net.ResolveIPAddr("ip", laddr) + peer.fsm.peerInfo.LocalAddress = ipaddr.IP + deferralExpiredFunc := func(family bgp.RouteFamily) func() { + return func() { + server.mgmtOperation(func() error { + server.softResetOut(peer.fsm.pConf.State.NeighborAddress, family, true) + return nil + }, false) + } + } + if !peer.fsm.pConf.GracefulRestart.State.LocalRestarting { + // When graceful-restart cap (which means intention + // of sending EOR) and route-target address family are negotiated, + // send route-target NLRIs first, and wait to send others + // till receiving EOR of route-target address family. + // This prevents sending uninterested routes to peers. + // + // However, when the peer is graceful restarting, give up + // waiting sending non-route-target NLRIs since the peer won't send + // any routes (and EORs) before we send ours (or deferral-timer expires). + var pathList []*table.Path + _, y := peer.fsm.rfMap[bgp.RF_RTC_UC] + if c := peer.fsm.pConf.GetAfiSafi(bgp.RF_RTC_UC); y && !peer.fsm.pConf.GracefulRestart.State.PeerRestarting && c.RouteTargetMembership.Config.DeferralTime > 0 { + pathList, _ = server.getBestFromLocal(peer, []bgp.RouteFamily{bgp.RF_RTC_UC}) + t := c.RouteTargetMembership.Config.DeferralTime + for _, f := range peer.negotiatedRFList() { + if f != bgp.RF_RTC_UC { + time.AfterFunc(time.Second*time.Duration(t), deferralExpiredFunc(f)) + } + } + } else { + pathList, _ = server.getBestFromLocal(peer, peer.negotiatedRFList()) + } + + if len(pathList) > 0 { + sendFsmOutgoingMsg(peer, pathList, nil, false) + } + } else { + // RFC 4724 4.1 + // Once the session between the Restarting Speaker and the Receiving + // Speaker is re-established, ...snip... it MUST defer route + // selection for an address family until it either (a) receives the + // End-of-RIB marker from all its peers (excluding the ones with the + // "Restart State" bit set in the received capability and excluding the + // ones that do not advertise the graceful restart capability) or (b) + // the Selection_Deferral_Timer referred to below has expired. + allEnd := func() bool { + for _, p := range server.neighborMap { + if !p.recvedAllEOR() { + return false + } + } + return true + }() + if allEnd { + for _, p := range server.neighborMap { + p.fsm.pConf.GracefulRestart.State.LocalRestarting = false + if !p.isGracefulRestartEnabled() { + continue + } + paths, _ := server.getBestFromLocal(p, p.configuredRFlist()) + if len(paths) > 0 { + sendFsmOutgoingMsg(p, paths, nil, false) + } + } + log.WithFields(log.Fields{ + "Topic": "Server", + }).Info("sync finished") + } else { + deferral := peer.fsm.pConf.GracefulRestart.Config.DeferralTime + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + }).Debugf("Now syncing, suppress sending updates. start deferral timer(%d)", deferral) + time.AfterFunc(time.Second*time.Duration(deferral), deferralExpiredFunc(bgp.RouteFamily(0))) + } + } + } else { + if server.shutdownWG != nil && nextState == bgp.BGP_FSM_IDLE { + die := true + for _, p := range server.neighborMap { + if p.fsm.state != bgp.BGP_FSM_IDLE { + die = false + break + } + } + if die { + server.shutdownWG.Done() + } + } + peer.fsm.pConf.Timers.State.Downtime = time.Now().Unix() + } + // clear counter + if peer.fsm.adminState == ADMIN_STATE_DOWN { + peer.fsm.pConf.State = config.NeighborState{} + peer.fsm.pConf.State.NeighborAddress = peer.fsm.pConf.Config.NeighborAddress + peer.fsm.pConf.State.PeerAs = peer.fsm.pConf.Config.PeerAs + peer.fsm.pConf.Timers.State = config.TimersState{} + } + peer.startFSMHandler(server.fsmincomingCh, server.fsmStateCh) + server.broadcastPeerState(peer, oldState, e) + case FSM_MSG_ROUTE_REFRESH: + if peer.fsm.state != bgp.BGP_FSM_ESTABLISHED || e.timestamp.Unix() < peer.fsm.pConf.Timers.State.Uptime { + return + } + if paths := server.handleRouteRefresh(peer, e); len(paths) > 0 { + sendFsmOutgoingMsg(peer, paths, nil, false) + return + } + case FSM_MSG_BGP_MESSAGE: + switch m := e.MsgData.(type) { + case *bgp.MessageError: + sendFsmOutgoingMsg(peer, nil, bgp.NewBGPNotificationMessage(m.TypeCode, m.SubTypeCode, m.Data), false) + return + case *bgp.BGPMessage: + server.notifyRecvMessageWatcher(peer, e.timestamp, m) + if peer.fsm.state != bgp.BGP_FSM_ESTABLISHED || e.timestamp.Unix() < peer.fsm.pConf.Timers.State.Uptime { + return + } + pathList, eor, notification := peer.handleUpdate(e) + if notification != nil { + sendFsmOutgoingMsg(peer, nil, notification, true) + return + } + if m.Header.Type == bgp.BGP_MSG_UPDATE { + server.notifyPrePolicyUpdateWatcher(peer, pathList, m, e.timestamp, e.payload) + } + + if len(pathList) > 0 { + server.propagateUpdate(peer, pathList) + } + + if len(eor) > 0 { + rtc := false + for _, f := range eor { + if f == bgp.RF_RTC_UC { + rtc = true + } + for i, a := range peer.fsm.pConf.AfiSafis { + if a.State.Family == f { + peer.fsm.pConf.AfiSafis[i].MpGracefulRestart.State.EndOfRibReceived = true + } + } + } + + // RFC 4724 4.1 + // Once the session between the Restarting Speaker and the Receiving + // Speaker is re-established, ...snip... it MUST defer route + // selection for an address family until it either (a) receives the + // End-of-RIB marker from all its peers (excluding the ones with the + // "Restart State" bit set in the received capability and excluding the + // ones that do not advertise the graceful restart capability) or ...snip... + if peer.fsm.pConf.GracefulRestart.State.LocalRestarting { + allEnd := func() bool { + for _, p := range server.neighborMap { + if !p.recvedAllEOR() { + return false + } + } + return true + }() + if allEnd { + for _, p := range server.neighborMap { + p.fsm.pConf.GracefulRestart.State.LocalRestarting = false + if !p.isGracefulRestartEnabled() { + continue + } + paths, _ := server.getBestFromLocal(p, p.negotiatedRFList()) + if len(paths) > 0 { + sendFsmOutgoingMsg(p, paths, nil, false) + } + } + log.WithFields(log.Fields{ + "Topic": "Server", + }).Info("sync finished") + + } + + // we don't delay non-route-target NLRIs when local-restarting + rtc = false + } + if peer.fsm.pConf.GracefulRestart.State.PeerRestarting { + if peer.recvedAllEOR() { + peer.stopPeerRestarting() + pathList := peer.adjRibIn.DropStale(peer.configuredRFlist()) + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.fsm.pConf.State.NeighborAddress, + }).Debugf("withdraw %d stale routes", len(pathList)) + server.propagateUpdate(peer, pathList) + } + + // we don't delay non-route-target NLRIs when peer is restarting + rtc = false + } + + // received EOR of route-target address family + // outbound filter is now ready, let's flash non-route-target NLRIs + if c := peer.fsm.pConf.GetAfiSafi(bgp.RF_RTC_UC); rtc && c != nil && c.RouteTargetMembership.Config.DeferralTime > 0 { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + }).Debug("received route-target eor. flash non-route-target NLRIs") + families := make([]bgp.RouteFamily, 0, len(peer.negotiatedRFList())) + for _, f := range peer.negotiatedRFList() { + if f != bgp.RF_RTC_UC { + families = append(families, f) + } + } + if paths, _ := server.getBestFromLocal(peer, families); len(paths) > 0 { + sendFsmOutgoingMsg(peer, paths, nil, false) + } + } + } + default: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.fsm.pConf.State.NeighborAddress, + "Data": e.MsgData, + }).Panic("unknown msg type") + } + } +} + +func (s *BgpServer) AddCollector(c *config.CollectorConfig) error { + return s.mgmtOperation(func() error { + _, err := NewCollector(s, c.Url, c.DbName, c.TableDumpInterval) + return err + }, false) +} + +func (s *BgpServer) StartZebraClient(c *config.ZebraConfig) error { + return s.mgmtOperation(func() error { + if s.zclient != nil { + return fmt.Errorf("already connected to Zebra") + } + protos := make([]string, 0, len(c.RedistributeRouteTypeList)) + for _, p := range c.RedistributeRouteTypeList { + protos = append(protos, string(p)) + } + var err error + s.zclient, err = newZebraClient(s, c.Url, protos, c.Version, c.NexthopTriggerEnable, c.NexthopTriggerDelay) + return err + }, false) +} + +func (s *BgpServer) AddBmp(c *config.BmpServerConfig) error { + return s.mgmtOperation(func() error { + return s.bmpManager.addServer(c) + }, true) +} + +func (s *BgpServer) DeleteBmp(c *config.BmpServerConfig) error { + return s.mgmtOperation(func() error { + return s.bmpManager.deleteServer(c) + }, true) +} + +func (s *BgpServer) Shutdown() { + s.mgmtOperation(func() error { + s.shutdownWG = new(sync.WaitGroup) + s.shutdownWG.Add(1) + stateOp := AdminStateOperation{ + State: ADMIN_STATE_DOWN, + Communication: nil, + } + for _, p := range s.neighborMap { + p.fsm.adminStateCh <- stateOp + } + // TODO: call fsmincomingCh.Close() + return nil + }, false) + + // Waits for all goroutines per peer to stop. + // Note: This should not be wrapped with s.mgmtOperation() in order to + // avoid the deadlock in the main goroutine of BgpServer. + if s.shutdownWG != nil { + s.shutdownWG.Wait() + s.shutdownWG = nil + } +} + +func (s *BgpServer) UpdatePolicy(policy config.RoutingPolicy) error { + return s.mgmtOperation(func() error { + ap := make(map[string]config.ApplyPolicy, len(s.neighborMap)+1) + ap[table.GLOBAL_RIB_NAME] = s.bgpConfig.Global.ApplyPolicy + for _, peer := range s.neighborMap { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.fsm.pConf.State.NeighborAddress, + }).Info("call set policy") + ap[peer.ID()] = peer.fsm.pConf.ApplyPolicy + } + return s.policy.Reset(&policy, ap) + }, false) +} + +// EVPN MAC MOBILITY HANDLING +// +// We don't have multihoming function now, so ignore +// ESI comparison. +// +// RFC7432 15. MAC Mobility +// +// A PE detecting a locally attached MAC address for which it had +// previously received a MAC/IP Advertisement route with the same zero +// Ethernet segment identifier (single-homed scenarios) advertises it +// with a MAC Mobility extended community attribute with the sequence +// number set properly. In the case of single-homed scenarios, there +// is no need for ESI comparison. + +func getMacMobilityExtendedCommunity(etag uint32, mac net.HardwareAddr, evpnPaths []*table.Path) *bgp.MacMobilityExtended { + seqs := make([]struct { + seq int + isLocal bool + }, 0) + + for _, path := range evpnPaths { + nlri := path.GetNlri().(*bgp.EVPNNLRI) + target, ok := nlri.RouteTypeData.(*bgp.EVPNMacIPAdvertisementRoute) + if !ok { + continue + } + if target.ETag == etag && bytes.Equal(target.MacAddress, mac) { + found := false + for _, ec := range path.GetExtCommunities() { + if t, st := ec.GetTypes(); t == bgp.EC_TYPE_EVPN && st == bgp.EC_SUBTYPE_MAC_MOBILITY { + seqs = append(seqs, struct { + seq int + isLocal bool + }{int(ec.(*bgp.MacMobilityExtended).Sequence), path.IsLocal()}) + found = true + break + } + } + + if !found { + seqs = append(seqs, struct { + seq int + isLocal bool + }{-1, path.IsLocal()}) + } + } + } + + if len(seqs) > 0 { + newSeq := -2 + var isLocal bool + for _, seq := range seqs { + if seq.seq > newSeq { + newSeq = seq.seq + isLocal = seq.isLocal + } + } + + if !isLocal { + newSeq += 1 + } + + if newSeq != -1 { + return &bgp.MacMobilityExtended{ + Sequence: uint32(newSeq), + } + } + } + return nil +} + +func (server *BgpServer) fixupApiPath(vrfId string, pathList []*table.Path) error { + pi := &table.PeerInfo{ + AS: server.bgpConfig.Global.Config.As, + LocalID: net.ParseIP(server.bgpConfig.Global.Config.RouterId).To4(), + } + + for _, path := range pathList { + if !path.IsWithdraw { + if _, err := path.GetOrigin(); err != nil { + return err + } + } + + if path.GetSource() == nil { + path.SetSource(pi) + } + + if vrfId != "" { + vrf := server.globalRib.Vrfs[vrfId] + if vrf == nil { + return fmt.Errorf("vrf %s not found", vrfId) + } + if err := vrf.ToGlobalPath(path); err != nil { + return err + } + } + + // Address Family specific Handling + switch nlri := path.GetNlri().(type) { + case *bgp.EVPNNLRI: + switch r := nlri.RouteTypeData.(type) { + case *bgp.EVPNMacIPAdvertisementRoute: + // MAC Mobility Extended Community + paths := server.globalRib.GetBestPathList(table.GLOBAL_RIB_NAME, 0, []bgp.RouteFamily{bgp.RF_EVPN}) + if m := getMacMobilityExtendedCommunity(r.ETag, r.MacAddress, paths); m != nil { + path.SetExtCommunities([]bgp.ExtendedCommunityInterface{m}, false) + } + case *bgp.EVPNEthernetSegmentRoute: + // RFC7432: BGP MPLS-Based Ethernet VPN + // 7.6. ES-Import Route Target + // The value is derived automatically for the ESI Types 1, 2, + // and 3, by encoding the high-order 6-octet portion of the 9-octet ESI + // Value, which corresponds to a MAC address, in the ES-Import Route + // Target. + // Note: If the given path already has the ES-Import Route Target, + // skips deriving a new one. + found := false + for _, extComm := range path.GetExtCommunities() { + if _, found = extComm.(*bgp.ESImportRouteTarget); found { + break + } + } + if !found { + switch r.ESI.Type { + case bgp.ESI_LACP, bgp.ESI_MSTP, bgp.ESI_MAC: + mac := net.HardwareAddr(r.ESI.Value[0:6]) + rt := &bgp.ESImportRouteTarget{ESImport: mac} + path.SetExtCommunities([]bgp.ExtendedCommunityInterface{rt}, false) + } + } + } + } + } + return nil +} + +func pathTokey(path *table.Path) string { + return fmt.Sprintf("%d:%s", path.GetNlri().PathIdentifier(), path.GetNlri().String()) +} + +func (s *BgpServer) AddPath(vrfId string, pathList []*table.Path) (uuidBytes []byte, err error) { + err = s.mgmtOperation(func() error { + if err := s.fixupApiPath(vrfId, pathList); err != nil { + return err + } + if len(pathList) == 1 { + path := pathList[0] + id, _ := uuid.NewV4() + s.uuidMap[id] = pathTokey(path) + uuidBytes = id.Bytes() + } + s.propagateUpdate(nil, pathList) + return nil + }, true) + return +} + +func (s *BgpServer) DeletePath(uuidBytes []byte, f bgp.RouteFamily, vrfId string, pathList []*table.Path) error { + return s.mgmtOperation(func() error { + deletePathList := make([]*table.Path, 0) + if len(uuidBytes) > 0 { + // Delete locally generated path which has the given UUID + path := func() *table.Path { + id, _ := uuid.FromBytes(uuidBytes) + if key, ok := s.uuidMap[id]; !ok { + return nil + } else { + for _, path := range s.globalRib.GetPathList(table.GLOBAL_RIB_NAME, 0, s.globalRib.GetRFlist()) { + if path.IsLocal() && key == pathTokey(path) { + delete(s.uuidMap, id) + return path + } + } + } + return nil + }() + if path == nil { + return fmt.Errorf("Can't find a specified path") + } + deletePathList = append(deletePathList, path.Clone(true)) + } else if len(pathList) == 0 { + // Delete all locally generated paths + families := s.globalRib.GetRFlist() + if f != 0 { + families = []bgp.RouteFamily{f} + } + for _, path := range s.globalRib.GetPathList(table.GLOBAL_RIB_NAME, 0, families) { + if path.IsLocal() { + deletePathList = append(deletePathList, path.Clone(true)) + } + } + s.uuidMap = make(map[uuid.UUID]string) + } else { + if err := s.fixupApiPath(vrfId, pathList); err != nil { + return err + } + deletePathList = pathList + } + s.propagateUpdate(nil, deletePathList) + return nil + }, true) +} + +func (s *BgpServer) UpdatePath(vrfId string, pathList []*table.Path) error { + err := s.mgmtOperation(func() error { + if err := s.fixupApiPath(vrfId, pathList); err != nil { + return err + } + s.propagateUpdate(nil, pathList) + return nil + }, true) + return err +} + +func (s *BgpServer) Start(c *config.Global) error { + return s.mgmtOperation(func() error { + if err := config.SetDefaultGlobalConfigValues(c); err != nil { + return err + } + + if c.Config.Port > 0 { + acceptCh := make(chan *net.TCPConn, 4096) + for _, addr := range c.Config.LocalAddressList { + l, err := NewTCPListener(addr, uint32(c.Config.Port), acceptCh) + if err != nil { + return err + } + s.listeners = append(s.listeners, l) + } + s.acceptCh = acceptCh + } + + rfs, _ := config.AfiSafis(c.AfiSafis).ToRfList() + s.globalRib = table.NewTableManager(rfs) + s.rsRib = table.NewTableManager(rfs) + + if err := s.policy.Reset(&config.RoutingPolicy{}, map[string]config.ApplyPolicy{}); err != nil { + return err + } + s.bgpConfig.Global = *c + // update route selection options + table.SelectionOptions = c.RouteSelectionOptions.Config + table.UseMultiplePaths = c.UseMultiplePaths.Config + + s.roaManager.SetAS(s.bgpConfig.Global.Config.As) + return nil + }, false) +} + +func (s *BgpServer) GetVrf() (l []*table.Vrf) { + s.mgmtOperation(func() error { + l = make([]*table.Vrf, 0, len(s.globalRib.Vrfs)) + for _, vrf := range s.globalRib.Vrfs { + l = append(l, vrf.Clone()) + } + return nil + }, true) + return l +} + +func (s *BgpServer) AddVrf(name string, id uint32, rd bgp.RouteDistinguisherInterface, im, ex []bgp.ExtendedCommunityInterface) error { + return s.mgmtOperation(func() error { + pi := &table.PeerInfo{ + AS: s.bgpConfig.Global.Config.As, + LocalID: net.ParseIP(s.bgpConfig.Global.Config.RouterId).To4(), + } + if pathList, err := s.globalRib.AddVrf(name, id, rd, im, ex, pi); err != nil { + return err + } else if len(pathList) > 0 { + s.propagateUpdate(nil, pathList) + } + return nil + }, true) +} + +func (s *BgpServer) DeleteVrf(name string) error { + return s.mgmtOperation(func() error { + for _, n := range s.neighborMap { + if n.fsm.pConf.Config.Vrf == name { + return fmt.Errorf("failed to delete VRF %s: neighbor %s is in use", name, n.ID()) + } + } + pathList, err := s.globalRib.DeleteVrf(name) + if err != nil { + return err + } + if len(pathList) > 0 { + s.propagateUpdate(nil, pathList) + } + return nil + }, true) +} + +func (s *BgpServer) Stop() error { + return s.mgmtOperation(func() error { + for k, _ := range s.neighborMap { + if err := s.deleteNeighbor(&config.Neighbor{Config: config.NeighborConfig{ + NeighborAddress: k}}, bgp.BGP_ERROR_CEASE, bgp.BGP_ERROR_SUB_PEER_DECONFIGURED); err != nil { + return err + } + } + for _, l := range s.listeners { + l.Close() + } + s.bgpConfig.Global = config.Global{} + return nil + }, true) +} + +func familiesForSoftreset(peer *Peer, family bgp.RouteFamily) []bgp.RouteFamily { + if family == bgp.RouteFamily(0) { + configured := peer.configuredRFlist() + families := make([]bgp.RouteFamily, 0, len(configured)) + for _, f := range configured { + if f != bgp.RF_RTC_UC { + families = append(families, f) + } + } + return families + } + return []bgp.RouteFamily{family} +} + +func (s *BgpServer) softResetIn(addr string, family bgp.RouteFamily) error { + peers, err := s.addrToPeers(addr) + if err != nil { + return err + } + for _, peer := range peers { + families := familiesForSoftreset(peer, family) + + pathList := make([]*table.Path, 0, peer.adjRibIn.Count(families)) + for _, path := range peer.adjRibIn.PathList(families, false) { + // RFC4271 9.1.2 Phase 2: Route Selection + // + // If the AS_PATH attribute of a BGP route contains an AS loop, the BGP + // route should be excluded from the Phase 2 decision function. + isLooped := false + if aspath := path.GetAsPath(); aspath != nil { + isLooped = hasOwnASLoop(peer.fsm.peerInfo.LocalAS, int(peer.fsm.pConf.AsPathOptions.Config.AllowOwnAs), aspath) + } + if path.IsAsLooped() != isLooped { + // can't modify the existing one. needs to create one + path = path.Clone(false) + path.SetAsLooped(isLooped) + // update accepted counter + peer.adjRibIn.Update([]*table.Path{path}) + } + if !path.IsAsLooped() { + pathList = append(pathList, path) + } + } + s.propagateUpdate(peer, pathList) + } + return err +} + +func (s *BgpServer) softResetOut(addr string, family bgp.RouteFamily, deferral bool) error { + peers, err := s.addrToPeers(addr) + if err != nil { + return err + } + for _, peer := range peers { + if peer.fsm.state != bgp.BGP_FSM_ESTABLISHED { + continue + } + families := familiesForSoftreset(peer, family) + + if deferral { + _, y := peer.fsm.rfMap[bgp.RF_RTC_UC] + if peer.fsm.pConf.GracefulRestart.State.LocalRestarting { + peer.fsm.pConf.GracefulRestart.State.LocalRestarting = false + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Families": families, + }).Debug("deferral timer expired") + } else if c := peer.fsm.pConf.GetAfiSafi(bgp.RF_RTC_UC); y && !c.MpGracefulRestart.State.EndOfRibReceived { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "Families": families, + }).Debug("route-target deferral timer expired") + } else { + continue + } + } + + pathList, filtered := s.getBestFromLocal(peer, families) + if len(pathList) > 0 { + sendFsmOutgoingMsg(peer, pathList, nil, false) + } + if !deferral && len(filtered) > 0 { + withdrawnList := make([]*table.Path, 0, len(filtered)) + for _, p := range filtered { + withdrawnList = append(withdrawnList, p.Clone(true)) + } + sendFsmOutgoingMsg(peer, withdrawnList, nil, false) + } + } + return nil +} + +func (s *BgpServer) SoftResetIn(addr string, family bgp.RouteFamily) error { + return s.mgmtOperation(func() error { + log.WithFields(log.Fields{ + "Topic": "Operation", + "Key": addr, + }).Info("Neighbor soft reset in") + return s.softResetIn(addr, family) + }, true) +} + +func (s *BgpServer) SoftResetOut(addr string, family bgp.RouteFamily) error { + return s.mgmtOperation(func() error { + log.WithFields(log.Fields{ + "Topic": "Operation", + "Key": addr, + }).Info("Neighbor soft reset out") + return s.softResetOut(addr, family, false) + }, true) +} + +func (s *BgpServer) SoftReset(addr string, family bgp.RouteFamily) error { + return s.mgmtOperation(func() error { + log.WithFields(log.Fields{ + "Topic": "Operation", + "Key": addr, + }).Info("Neighbor soft reset") + err := s.softResetIn(addr, family) + if err != nil { + return err + } + return s.softResetOut(addr, family, false) + }, true) +} + +func (s *BgpServer) validateTable(r *table.Table) (v []*table.Validation) { + if s.roaManager.enabled() { + v = make([]*table.Validation, 0, len(r.GetDestinations())) + for _, d := range r.GetDestinations() { + for _, p := range d.GetAllKnownPathList() { + v = append(v, s.roaManager.validate(p)) + } + } + } + return +} + +func (s *BgpServer) GetRib(addr string, family bgp.RouteFamily, prefixes []*table.LookupPrefix) (rib *table.Table, v []*table.Validation, err error) { + err = s.mgmtOperation(func() error { + m := s.globalRib + id := table.GLOBAL_RIB_NAME + as := uint32(0) + if len(addr) > 0 { + peer, ok := s.neighborMap[addr] + if !ok { + return fmt.Errorf("Neighbor that has %v doesn't exist.", addr) + } + if !peer.isRouteServerClient() { + return fmt.Errorf("Neighbor %v doesn't have local rib", addr) + } + id = peer.ID() + as = peer.AS() + m = s.rsRib + } + af := bgp.RouteFamily(family) + tbl, ok := m.Tables[af] + if !ok { + return fmt.Errorf("address family: %s not supported", af) + } + rib, err = tbl.Select(table.TableSelectOption{ID: id, AS: as, LookupPrefixes: prefixes}) + v = s.validateTable(rib) + return err + }, true) + return +} + +func (s *BgpServer) GetVrfRib(name string, family bgp.RouteFamily, prefixes []*table.LookupPrefix) (rib *table.Table, err error) { + err = s.mgmtOperation(func() error { + m := s.globalRib + vrfs := m.Vrfs + if _, ok := vrfs[name]; !ok { + return fmt.Errorf("vrf %s not found", name) + } + var af bgp.RouteFamily + switch family { + case bgp.RF_IPv4_UC: + af = bgp.RF_IPv4_VPN + case bgp.RF_IPv6_UC: + af = bgp.RF_IPv6_VPN + case bgp.RF_EVPN: + af = bgp.RF_EVPN + } + tbl, ok := m.Tables[af] + if !ok { + return fmt.Errorf("address family: %s not supported", af) + } + rib, err = tbl.Select(table.TableSelectOption{VRF: vrfs[name], LookupPrefixes: prefixes}) + return err + }, true) + return +} + +func (s *BgpServer) GetAdjRib(addr string, family bgp.RouteFamily, in bool, prefixes []*table.LookupPrefix) (rib *table.Table, v []*table.Validation, err error) { + err = s.mgmtOperation(func() error { + peer, ok := s.neighborMap[addr] + if !ok { + return fmt.Errorf("Neighbor that has %v doesn't exist.", addr) + } + id := peer.ID() + as := peer.AS() + + var adjRib *table.AdjRib + if in { + adjRib = peer.adjRibIn + } else { + adjRib = table.NewAdjRib(peer.configuredRFlist()) + accepted, _ := s.getBestFromLocal(peer, peer.configuredRFlist()) + adjRib.Update(accepted) + } + rib, err = adjRib.Select(family, false, table.TableSelectOption{ID: id, AS: as, LookupPrefixes: prefixes}) + v = s.validateTable(rib) + return err + }, true) + return +} + +func (s *BgpServer) GetRibInfo(addr string, family bgp.RouteFamily) (info *table.TableInfo, err error) { + err = s.mgmtOperation(func() error { + m := s.globalRib + id := table.GLOBAL_RIB_NAME + as := uint32(0) + if len(addr) > 0 { + peer, ok := s.neighborMap[addr] + if !ok { + return fmt.Errorf("Neighbor that has %v doesn't exist.", addr) + } + if !peer.isRouteServerClient() { + return fmt.Errorf("Neighbor %v doesn't have local rib", addr) + } + id = peer.ID() + as = peer.AS() + m = s.rsRib + } + info, err = m.TableInfo(id, as, family) + return err + }, true) + return +} + +func (s *BgpServer) GetAdjRibInfo(addr string, family bgp.RouteFamily, in bool) (info *table.TableInfo, err error) { + err = s.mgmtOperation(func() error { + peer, ok := s.neighborMap[addr] + if !ok { + return fmt.Errorf("Neighbor that has %v doesn't exist.", addr) + } + + var adjRib *table.AdjRib + if in { + adjRib = peer.adjRibIn + } else { + adjRib = table.NewAdjRib(peer.configuredRFlist()) + accepted, _ := s.getBestFromLocal(peer, peer.configuredRFlist()) + adjRib.Update(accepted) + } + info, err = adjRib.TableInfo(family) + return err + }, true) + return +} + +func (s *BgpServer) GetServer() (c *config.Global) { + s.mgmtOperation(func() error { + g := s.bgpConfig.Global + c = &g + return nil + }, false) + return c +} + +func (s *BgpServer) GetNeighbor(address string, getAdvertised bool) (l []*config.Neighbor) { + s.mgmtOperation(func() error { + l = make([]*config.Neighbor, 0, len(s.neighborMap)) + for k, peer := range s.neighborMap { + if address != "" && address != k && address != peer.fsm.pConf.Config.NeighborInterface { + continue + } + l = append(l, s.ToConfig(peer, getAdvertised)) + } + return nil + }, false) + return l +} + +func (server *BgpServer) addPeerGroup(c *config.PeerGroup) error { + name := c.Config.PeerGroupName + if _, y := server.peerGroupMap[name]; y { + return fmt.Errorf("Can't overwrite the existing peer-group: %s", name) + } + + log.WithFields(log.Fields{ + "Topic": "Peer", + "Name": name, + }).Info("Add a peer group configuration") + + server.peerGroupMap[c.Config.PeerGroupName] = NewPeerGroup(c) + + return nil +} + +func (server *BgpServer) addNeighbor(c *config.Neighbor) error { + addr, err := c.ExtractNeighborAddress() + if err != nil { + return err + } + + if _, y := server.neighborMap[addr]; y { + return fmt.Errorf("Can't overwrite the existing peer: %s", addr) + } + + var pgConf *config.PeerGroup + if c.Config.PeerGroup != "" { + pg, ok := server.peerGroupMap[c.Config.PeerGroup] + if !ok { + return fmt.Errorf("no such peer-group: %s", c.Config.PeerGroup) + } + pgConf = pg.Conf + } + + if err := config.SetDefaultNeighborConfigValues(c, pgConf, &server.bgpConfig.Global); err != nil { + return err + } + + if vrf := c.Config.Vrf; vrf != "" { + if c.RouteServer.Config.RouteServerClient { + return fmt.Errorf("route server client can't be enslaved to VRF") + } + families, _ := config.AfiSafis(c.AfiSafis).ToRfList() + for _, f := range families { + if f != bgp.RF_IPv4_UC && f != bgp.RF_IPv6_UC { + return fmt.Errorf("%s is not supported for VRF enslaved neighbor", f) + } + } + _, y := server.globalRib.Vrfs[vrf] + if !y { + return fmt.Errorf("VRF not found: %s", vrf) + } + } + + if c.RouteServer.Config.RouteServerClient && c.RouteReflector.Config.RouteReflectorClient { + return fmt.Errorf("can't be both route-server-client and route-reflector-client") + } + + if server.bgpConfig.Global.Config.Port > 0 { + for _, l := range server.Listeners(addr) { + if c.Config.AuthPassword != "" { + if err := SetTcpMD5SigSockopt(l, addr, c.Config.AuthPassword); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warnf("failed to set md5: %s", err) + } + } + } + } + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Infof("Add a peer configuration for:%s", addr) + + rib := server.globalRib + if c.RouteServer.Config.RouteServerClient { + rib = server.rsRib + } + peer := NewPeer(&server.bgpConfig.Global, c, rib, server.policy) + server.policy.Reset(nil, map[string]config.ApplyPolicy{peer.ID(): c.ApplyPolicy}) + server.neighborMap[addr] = peer + if name := c.Config.PeerGroup; name != "" { + server.peerGroupMap[name].AddMember(*c) + } + peer.startFSMHandler(server.fsmincomingCh, server.fsmStateCh) + server.broadcastPeerState(peer, bgp.BGP_FSM_IDLE, nil) + return nil +} + +func (s *BgpServer) AddPeerGroup(c *config.PeerGroup) error { + return s.mgmtOperation(func() error { + return s.addPeerGroup(c) + }, true) +} + +func (s *BgpServer) AddNeighbor(c *config.Neighbor) error { + return s.mgmtOperation(func() error { + return s.addNeighbor(c) + }, true) +} + +func (s *BgpServer) AddDynamicNeighbor(c *config.DynamicNeighbor) error { + return s.mgmtOperation(func() error { + s.peerGroupMap[c.Config.PeerGroup].AddDynamicNeighbor(c) + return nil + }, true) +} + +func (server *BgpServer) deletePeerGroup(pg *config.PeerGroup) error { + name := pg.Config.PeerGroupName + + if _, y := server.peerGroupMap[name]; !y { + return fmt.Errorf("Can't delete a peer-group %s which does not exist", name) + } + + log.WithFields(log.Fields{ + "Topic": "Peer", + "Name": name, + }).Info("Delete a peer group configuration") + + delete(server.peerGroupMap, name) + return nil +} + +func (server *BgpServer) deleteNeighbor(c *config.Neighbor, code, subcode uint8) error { + if c.Config.PeerGroup != "" { + _, y := server.peerGroupMap[c.Config.PeerGroup] + if y { + server.peerGroupMap[c.Config.PeerGroup].DeleteMember(*c) + } + } + + addr, err := c.ExtractNeighborAddress() + if err != nil { + return err + } + + if intf := c.Config.NeighborInterface; intf != "" { + var err error + addr, err = config.GetIPv6LinkLocalNeighborAddress(intf) + if err != nil { + return err + } + } + n, y := server.neighborMap[addr] + if !y { + return fmt.Errorf("Can't delete a peer configuration for %s", addr) + } + for _, l := range server.Listeners(addr) { + if err := SetTcpMD5SigSockopt(l, addr, ""); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warnf("failed to unset md5: %s", err) + } + } + log.WithFields(log.Fields{ + "Topic": "Peer", + }).Infof("Delete a peer configuration for:%s", addr) + + n.fsm.sendNotification(code, subcode, nil, "") + n.stopPeerRestarting() + + go n.stopFSM() + delete(server.neighborMap, addr) + server.dropPeerAllRoutes(n, n.configuredRFlist()) + return nil +} + +func (s *BgpServer) DeletePeerGroup(c *config.PeerGroup) error { + return s.mgmtOperation(func() error { + name := c.Config.PeerGroupName + for _, n := range s.neighborMap { + if n.fsm.pConf.Config.PeerGroup == name { + return fmt.Errorf("failed to delete peer-group %s: neighbor %s is in use", name, n.ID()) + } + } + return s.deletePeerGroup(c) + }, true) +} + +func (s *BgpServer) DeleteNeighbor(c *config.Neighbor) error { + return s.mgmtOperation(func() error { + return s.deleteNeighbor(c, bgp.BGP_ERROR_CEASE, bgp.BGP_ERROR_SUB_PEER_DECONFIGURED) + }, true) +} + +func (s *BgpServer) updatePeerGroup(pg *config.PeerGroup) (needsSoftResetIn bool, err error) { + name := pg.Config.PeerGroupName + + _, ok := s.peerGroupMap[name] + if !ok { + return false, fmt.Errorf("Peer-group %s doesn't exist.", name) + } + s.peerGroupMap[name].Conf = pg + + for _, n := range s.peerGroupMap[name].members { + c := n + u, err := s.updateNeighbor(&c) + if err != nil { + return needsSoftResetIn, err + } + needsSoftResetIn = needsSoftResetIn || u + } + return needsSoftResetIn, nil +} + +func (s *BgpServer) UpdatePeerGroup(pg *config.PeerGroup) (needsSoftResetIn bool, err error) { + err = s.mgmtOperation(func() error { + needsSoftResetIn, err = s.updatePeerGroup(pg) + return err + }, true) + return needsSoftResetIn, err +} + +func (s *BgpServer) updateNeighbor(c *config.Neighbor) (needsSoftResetIn bool, err error) { + if c.Config.PeerGroup != "" { + if pg, ok := s.peerGroupMap[c.Config.PeerGroup]; ok { + if err := config.SetDefaultNeighborConfigValues(c, pg.Conf, &s.bgpConfig.Global); err != nil { + return needsSoftResetIn, err + } + } else { + return needsSoftResetIn, fmt.Errorf("no such peer-group: %s", c.Config.PeerGroup) + } + } + + addr, err := c.ExtractNeighborAddress() + if err != nil { + return needsSoftResetIn, err + } + + peer, ok := s.neighborMap[addr] + if !ok { + return needsSoftResetIn, fmt.Errorf("Neighbor that has %v doesn't exist.", addr) + } + + if !peer.fsm.pConf.ApplyPolicy.Equal(&c.ApplyPolicy) { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Info("Update ApplyPolicy") + s.policy.Reset(nil, map[string]config.ApplyPolicy{peer.ID(): c.ApplyPolicy}) + peer.fsm.pConf.ApplyPolicy = c.ApplyPolicy + needsSoftResetIn = true + } + original := peer.fsm.pConf + + if !original.AsPathOptions.Config.Equal(&c.AsPathOptions.Config) { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + }).Info("Update aspath options") + peer.fsm.pConf.AsPathOptions = c.AsPathOptions + needsSoftResetIn = true + } + + if original.NeedsResendOpenMessage(c) { + sub := uint8(bgp.BGP_ERROR_SUB_OTHER_CONFIGURATION_CHANGE) + if original.Config.AdminDown != c.Config.AdminDown { + sub = bgp.BGP_ERROR_SUB_ADMINISTRATIVE_SHUTDOWN + state := "Admin Down" + + if !c.Config.AdminDown { + state = "Admin Up" + } + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + "State": state, + }).Info("Update admin-state configuration") + } else if original.Config.PeerAs != c.Config.PeerAs { + sub = bgp.BGP_ERROR_SUB_PEER_DECONFIGURED + } + if err = s.deleteNeighbor(peer.fsm.pConf, bgp.BGP_ERROR_CEASE, sub); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Error(err) + return needsSoftResetIn, err + } + err = s.addNeighbor(c) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Error(err) + } + return needsSoftResetIn, err + } + + if !original.Timers.Config.Equal(&c.Timers.Config) { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.ID(), + }).Info("Update timer configuration") + peer.fsm.pConf.Timers.Config = c.Timers.Config + } + + err = peer.updatePrefixLimitConfig(c.AfiSafis) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Error(err) + // rollback to original state + peer.fsm.pConf = original + } + return needsSoftResetIn, err +} + +func (s *BgpServer) UpdateNeighbor(c *config.Neighbor) (needsSoftResetIn bool, err error) { + err = s.mgmtOperation(func() error { + needsSoftResetIn, err = s.updateNeighbor(c) + return err + }, true) + return needsSoftResetIn, err +} + +func (s *BgpServer) addrToPeers(addr string) (l []*Peer, err error) { + if len(addr) == 0 { + for _, p := range s.neighborMap { + l = append(l, p) + } + return l, nil + } + peer, found := s.neighborMap[addr] + if !found { + return l, fmt.Errorf("Neighbor that has %v doesn't exist.", addr) + } + return []*Peer{peer}, nil +} + +func (s *BgpServer) resetNeighbor(op, addr string, subcode uint8, data []byte) error { + log.WithFields(log.Fields{ + "Topic": "Operation", + "Key": addr, + }).Info(op) + + peers, err := s.addrToPeers(addr) + if err == nil { + m := bgp.NewBGPNotificationMessage(bgp.BGP_ERROR_CEASE, subcode, data) + for _, peer := range peers { + sendFsmOutgoingMsg(peer, nil, m, false) + } + } + return err +} + +func (s *BgpServer) ShutdownNeighbor(addr, communication string) error { + return s.mgmtOperation(func() error { + return s.resetNeighbor("Neighbor shutdown", addr, bgp.BGP_ERROR_SUB_ADMINISTRATIVE_SHUTDOWN, newAdministrativeCommunication(communication)) + }, true) +} + +func (s *BgpServer) ResetNeighbor(addr, communication string) error { + return s.mgmtOperation(func() error { + err := s.resetNeighbor("Neighbor reset", addr, bgp.BGP_ERROR_SUB_ADMINISTRATIVE_RESET, newAdministrativeCommunication(communication)) + if err != nil { + return err + } + peers, _ := s.addrToPeers(addr) + for _, peer := range peers { + peer.fsm.idleHoldTime = peer.fsm.pConf.Timers.Config.IdleHoldTimeAfterReset + } + return nil + }, true) +} + +func (s *BgpServer) setAdminState(addr, communication string, enable bool) error { + peers, err := s.addrToPeers(addr) + if err != nil { + return err + } + for _, peer := range peers { + f := func(stateOp *AdminStateOperation, message string) { + select { + case peer.fsm.adminStateCh <- *stateOp: + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": peer.fsm.pConf.State.NeighborAddress, + }).Debug(message) + default: + log.Warning("previous request is still remaining. : ", peer.fsm.pConf.State.NeighborAddress) + } + } + if enable { + f(&AdminStateOperation{ADMIN_STATE_UP, nil}, "ADMIN_STATE_UP requested") + } else { + f(&AdminStateOperation{ADMIN_STATE_DOWN, newAdministrativeCommunication(communication)}, "ADMIN_STATE_DOWN requested") + } + } + return nil +} + +func (s *BgpServer) EnableNeighbor(addr string) error { + return s.mgmtOperation(func() error { + return s.setAdminState(addr, "", true) + }, true) +} + +func (s *BgpServer) DisableNeighbor(addr, communication string) error { + return s.mgmtOperation(func() error { + return s.setAdminState(addr, communication, false) + }, true) +} + +func (s *BgpServer) GetDefinedSet(typ table.DefinedType, name string) (sets *config.DefinedSets, err error) { + err = s.mgmtOperation(func() error { + sets, err = s.policy.GetDefinedSet(typ, name) + return nil + }, false) + return sets, err +} + +func (s *BgpServer) AddDefinedSet(a table.DefinedSet) error { + return s.mgmtOperation(func() error { + return s.policy.AddDefinedSet(a) + }, false) +} + +func (s *BgpServer) DeleteDefinedSet(a table.DefinedSet, all bool) error { + return s.mgmtOperation(func() error { + return s.policy.DeleteDefinedSet(a, all) + }, false) +} + +func (s *BgpServer) ReplaceDefinedSet(a table.DefinedSet) error { + return s.mgmtOperation(func() error { + return s.policy.ReplaceDefinedSet(a) + }, false) +} + +func (s *BgpServer) GetStatement() (l []*config.Statement) { + s.mgmtOperation(func() error { + l = s.policy.GetStatement() + return nil + }, false) + return l +} + +func (s *BgpServer) AddStatement(st *table.Statement) error { + return s.mgmtOperation(func() error { + return s.policy.AddStatement(st) + }, false) +} + +func (s *BgpServer) DeleteStatement(st *table.Statement, all bool) error { + return s.mgmtOperation(func() error { + return s.policy.DeleteStatement(st, all) + }, false) +} + +func (s *BgpServer) ReplaceStatement(st *table.Statement) error { + return s.mgmtOperation(func() error { + return s.policy.ReplaceStatement(st) + }, false) +} + +func (s *BgpServer) GetPolicy() (l []*config.PolicyDefinition) { + s.mgmtOperation(func() error { + l = s.policy.GetAllPolicy() + return nil + }, false) + return l +} + +func (s *BgpServer) AddPolicy(x *table.Policy, refer bool) error { + return s.mgmtOperation(func() error { + return s.policy.AddPolicy(x, refer) + }, false) +} + +func (s *BgpServer) DeletePolicy(x *table.Policy, all, preserve bool) error { + return s.mgmtOperation(func() error { + l := make([]string, 0, len(s.neighborMap)+1) + for _, peer := range s.neighborMap { + l = append(l, peer.ID()) + } + l = append(l, table.GLOBAL_RIB_NAME) + + return s.policy.DeletePolicy(x, all, preserve, l) + }, false) +} + +func (s *BgpServer) ReplacePolicy(x *table.Policy, refer, preserve bool) error { + return s.mgmtOperation(func() error { + return s.policy.ReplacePolicy(x, refer, preserve) + }, false) +} + +func (server *BgpServer) toPolicyInfo(name string, dir table.PolicyDirection) (string, error) { + if name == "" { + switch dir { + case table.POLICY_DIRECTION_IMPORT, table.POLICY_DIRECTION_EXPORT: + return table.GLOBAL_RIB_NAME, nil + } + return "", fmt.Errorf("invalid policy type") + } else { + peer, ok := server.neighborMap[name] + if !ok { + return "", fmt.Errorf("not found peer %s", name) + } + if !peer.isRouteServerClient() { + return "", fmt.Errorf("non-rs-client peer %s doesn't have per peer policy", name) + } + return peer.ID(), nil + } +} + +func (s *BgpServer) GetPolicyAssignment(name string, dir table.PolicyDirection) (rt table.RouteType, l []*config.PolicyDefinition, err error) { + err = s.mgmtOperation(func() error { + var id string + id, err = s.toPolicyInfo(name, dir) + if err != nil { + rt = table.ROUTE_TYPE_NONE + return err + } + rt, l, err = s.policy.GetPolicyAssignment(id, dir) + return nil + }, false) + return rt, l, err +} + +func (s *BgpServer) AddPolicyAssignment(name string, dir table.PolicyDirection, policies []*config.PolicyDefinition, def table.RouteType) error { + return s.mgmtOperation(func() error { + id, err := s.toPolicyInfo(name, dir) + if err != nil { + return err + } + return s.policy.AddPolicyAssignment(id, dir, policies, def) + }, false) +} + +func (s *BgpServer) DeletePolicyAssignment(name string, dir table.PolicyDirection, policies []*config.PolicyDefinition, all bool) error { + return s.mgmtOperation(func() error { + id, err := s.toPolicyInfo(name, dir) + if err != nil { + return err + } + return s.policy.DeletePolicyAssignment(id, dir, policies, all) + }, false) +} + +func (s *BgpServer) ReplacePolicyAssignment(name string, dir table.PolicyDirection, policies []*config.PolicyDefinition, def table.RouteType) error { + return s.mgmtOperation(func() error { + id, err := s.toPolicyInfo(name, dir) + if err != nil { + return err + } + return s.policy.ReplacePolicyAssignment(id, dir, policies, def) + }, false) +} + +func (s *BgpServer) EnableMrt(c *config.MrtConfig) error { + return s.mgmtOperation(func() error { + return s.mrtManager.enable(c) + }, false) +} + +func (s *BgpServer) DisableMrt(c *config.MrtConfig) error { + return s.mgmtOperation(func() error { + return s.mrtManager.disable(c) + }, false) +} + +func (s *BgpServer) GetRpki() (l []*config.RpkiServer, err error) { + err = s.mgmtOperation(func() error { + l = s.roaManager.GetServers() + return nil + }, false) + return l, err +} + +func (s *BgpServer) GetRoa(family bgp.RouteFamily) (l []*table.ROA, err error) { + s.mgmtOperation(func() error { + l, err = s.roaManager.GetRoa(family) + return nil + }, false) + return l, err +} + +func (s *BgpServer) AddRpki(c *config.RpkiServerConfig) error { + return s.mgmtOperation(func() error { + return s.roaManager.AddServer(net.JoinHostPort(c.Address, strconv.Itoa(int(c.Port))), c.RecordLifetime) + }, false) +} + +func (s *BgpServer) DeleteRpki(c *config.RpkiServerConfig) error { + return s.mgmtOperation(func() error { + return s.roaManager.DeleteServer(c.Address) + }, false) +} + +func (s *BgpServer) EnableRpki(c *config.RpkiServerConfig) error { + return s.mgmtOperation(func() error { + return s.roaManager.Enable(c.Address) + }, false) +} + +func (s *BgpServer) DisableRpki(c *config.RpkiServerConfig) error { + return s.mgmtOperation(func() error { + return s.roaManager.Disable(c.Address) + }, false) +} + +func (s *BgpServer) ResetRpki(c *config.RpkiServerConfig) error { + return s.mgmtOperation(func() error { + return s.roaManager.Reset(c.Address) + }, false) +} + +func (s *BgpServer) SoftResetRpki(c *config.RpkiServerConfig) error { + return s.mgmtOperation(func() error { + return s.roaManager.SoftReset(c.Address) + }, false) +} + +type WatchEventType string + +const ( + WATCH_EVENT_TYPE_BEST_PATH WatchEventType = "bestpath" + WATCH_EVENT_TYPE_PRE_UPDATE WatchEventType = "preupdate" + WATCH_EVENT_TYPE_POST_UPDATE WatchEventType = "postupdate" + WATCH_EVENT_TYPE_PEER_STATE WatchEventType = "peerstate" + WATCH_EVENT_TYPE_TABLE WatchEventType = "table" + WATCH_EVENT_TYPE_RECV_MSG WatchEventType = "receivedmessage" +) + +type WatchEvent interface { +} + +type WatchEventUpdate struct { + Message *bgp.BGPMessage + PeerAS uint32 + LocalAS uint32 + PeerAddress net.IP + LocalAddress net.IP + PeerID net.IP + FourBytesAs bool + Timestamp time.Time + Payload []byte + PostPolicy bool + Init bool + PathList []*table.Path + Neighbor *config.Neighbor +} + +type WatchEventPeerState struct { + PeerAS uint32 + LocalAS uint32 + PeerAddress net.IP + LocalAddress net.IP + PeerPort uint16 + LocalPort uint16 + PeerID net.IP + SentOpen *bgp.BGPMessage + RecvOpen *bgp.BGPMessage + State bgp.FSMState + StateReason *FsmStateReason + AdminState AdminState + Timestamp time.Time + PeerInterface string +} + +type WatchEventAdjIn struct { + PathList []*table.Path +} + +type WatchEventTable struct { + RouterId string + PathList map[string][]*table.Path + Neighbor []*config.Neighbor +} + +type WatchEventBestPath struct { + PathList []*table.Path + MultiPathList [][]*table.Path + Vrf map[string]uint16 +} + +type WatchEventMessage struct { + Message *bgp.BGPMessage + PeerAS uint32 + LocalAS uint32 + PeerAddress net.IP + LocalAddress net.IP + PeerID net.IP + FourBytesAs bool + Timestamp time.Time + IsSent bool +} + +type watchOptions struct { + bestpath bool + preUpdate bool + postUpdate bool + peerState bool + initBest bool + initUpdate bool + initPostUpdate bool + initPeerState bool + tableName string + recvMessage bool +} + +type WatchOption func(*watchOptions) + +func WatchBestPath(current bool) WatchOption { + return func(o *watchOptions) { + o.bestpath = true + if current { + o.initBest = true + } + } +} + +func WatchUpdate(current bool) WatchOption { + return func(o *watchOptions) { + o.preUpdate = true + if current { + o.initUpdate = true + } + } +} + +func WatchPostUpdate(current bool) WatchOption { + return func(o *watchOptions) { + o.postUpdate = true + if current { + o.initPostUpdate = true + } + } +} + +func WatchPeerState(current bool) WatchOption { + return func(o *watchOptions) { + o.peerState = true + if current { + o.initPeerState = true + } + } +} + +func WatchTableName(name string) WatchOption { + return func(o *watchOptions) { + o.tableName = name + } +} + +func WatchMessage(isSent bool) WatchOption { + return func(o *watchOptions) { + if isSent { + log.WithFields(log.Fields{ + "Topic": "Server", + }).Warn("watch event for sent messages is not implemented yet") + // o.sentMessage = true + } else { + o.recvMessage = true + } + } +} + +type Watcher struct { + opts watchOptions + realCh chan WatchEvent + ch *channels.InfiniteChannel + s *BgpServer +} + +func (w *Watcher) Event() <-chan WatchEvent { + return w.realCh +} + +func (w *Watcher) Generate(t WatchEventType) error { + return w.s.mgmtOperation(func() error { + switch t { + case WATCH_EVENT_TYPE_PRE_UPDATE: + pathList := make([]*table.Path, 0) + for _, peer := range w.s.neighborMap { + pathList = append(pathList, peer.adjRibIn.PathList(peer.configuredRFlist(), false)...) + } + w.notify(&WatchEventAdjIn{PathList: clonePathList(pathList)}) + case WATCH_EVENT_TYPE_TABLE: + rib := w.s.globalRib + as := uint32(0) + id := table.GLOBAL_RIB_NAME + if len(w.opts.tableName) > 0 { + peer, ok := w.s.neighborMap[w.opts.tableName] + if !ok { + return fmt.Errorf("Neighbor that has %v doesn't exist.", w.opts.tableName) + } + if !peer.isRouteServerClient() { + return fmt.Errorf("Neighbor %v doesn't have local rib", w.opts.tableName) + } + id = peer.ID() + as = peer.AS() + rib = w.s.rsRib + } + + pathList := func() map[string][]*table.Path { + pathList := make(map[string][]*table.Path) + for _, t := range rib.Tables { + for _, dst := range t.GetDestinations() { + if paths := dst.GetKnownPathList(id, as); len(paths) > 0 { + pathList[dst.GetNlri().String()] = clonePathList(paths) + } + } + } + return pathList + }() + l := make([]*config.Neighbor, 0, len(w.s.neighborMap)) + for _, peer := range w.s.neighborMap { + l = append(l, w.s.ToConfig(peer, false)) + } + w.notify(&WatchEventTable{PathList: pathList, Neighbor: l}) + default: + return fmt.Errorf("unsupported type %v", t) + } + return nil + }, false) +} + +func (w *Watcher) notify(v WatchEvent) { + w.ch.In() <- v +} + +func (w *Watcher) loop() { + for ev := range w.ch.Out() { + w.realCh <- ev.(WatchEvent) + } + close(w.realCh) +} + +func (w *Watcher) Stop() { + w.s.mgmtOperation(func() error { + for k, l := range w.s.watcherMap { + for i, v := range l { + if w == v { + w.s.watcherMap[k] = append(l[:i], l[i+1:]...) + break + } + } + } + + cleanInfiniteChannel(w.ch) + // the loop function goroutine might be blocked for + // writing to realCh. make sure it finishes. + for range w.realCh { + } + return nil + }, false) +} + +func (s *BgpServer) isWatched(typ WatchEventType) bool { + return len(s.watcherMap[typ]) != 0 +} + +func (s *BgpServer) notifyWatcher(typ WatchEventType, ev WatchEvent) { + for _, w := range s.watcherMap[typ] { + w.notify(ev) + } +} + +func (s *BgpServer) Watch(opts ...WatchOption) (w *Watcher) { + s.mgmtOperation(func() error { + w = &Watcher{ + s: s, + realCh: make(chan WatchEvent, 8), + ch: channels.NewInfiniteChannel(), + } + + for _, opt := range opts { + opt(&w.opts) + } + + register := func(t WatchEventType, w *Watcher) { + s.watcherMap[t] = append(s.watcherMap[t], w) + } + + if w.opts.bestpath { + register(WATCH_EVENT_TYPE_BEST_PATH, w) + } + if w.opts.preUpdate { + register(WATCH_EVENT_TYPE_PRE_UPDATE, w) + } + if w.opts.postUpdate { + register(WATCH_EVENT_TYPE_POST_UPDATE, w) + } + if w.opts.peerState { + register(WATCH_EVENT_TYPE_PEER_STATE, w) + } + if w.opts.initPeerState { + for _, peer := range s.neighborMap { + if peer.fsm.state != bgp.BGP_FSM_ESTABLISHED { + continue + } + w.notify(newWatchEventPeerState(peer, nil)) + } + } + if w.opts.initBest && s.active() == nil { + w.notify(&WatchEventBestPath{ + PathList: s.globalRib.GetBestPathList(table.GLOBAL_RIB_NAME, 0, nil), + MultiPathList: s.globalRib.GetBestMultiPathList(table.GLOBAL_RIB_NAME, nil), + }) + } + if w.opts.initUpdate { + for _, peer := range s.neighborMap { + if peer.fsm.state != bgp.BGP_FSM_ESTABLISHED { + continue + } + configNeighbor := w.s.ToConfig(peer, false) + for _, rf := range peer.configuredRFlist() { + _, y := peer.fsm.capMap[bgp.BGP_CAP_FOUR_OCTET_AS_NUMBER] + l, _ := peer.fsm.LocalHostPort() + w.notify(&WatchEventUpdate{ + PeerAS: peer.fsm.peerInfo.AS, + LocalAS: peer.fsm.peerInfo.LocalAS, + PeerAddress: peer.fsm.peerInfo.Address, + LocalAddress: net.ParseIP(l), + PeerID: peer.fsm.peerInfo.ID, + FourBytesAs: y, + Init: true, + PostPolicy: false, + Neighbor: configNeighbor, + PathList: peer.adjRibIn.PathList([]bgp.RouteFamily{rf}, false), + }) + + eor := bgp.NewEndOfRib(rf) + eorBuf, _ := eor.Serialize() + w.notify(&WatchEventUpdate{ + Message: eor, + PeerAS: peer.fsm.peerInfo.AS, + LocalAS: peer.fsm.peerInfo.LocalAS, + PeerAddress: peer.fsm.peerInfo.Address, + LocalAddress: net.ParseIP(l), + PeerID: peer.fsm.peerInfo.ID, + FourBytesAs: y, + Timestamp: time.Now(), + Init: true, + Payload: eorBuf, + PostPolicy: false, + Neighbor: configNeighbor, + }) + } + } + } + if w.opts.initPostUpdate && s.active() == nil { + for _, rf := range s.globalRib.GetRFlist() { + if len(s.globalRib.Tables[rf].GetDestinations()) == 0 { + continue + } + pathsByPeer := make(map[*table.PeerInfo][]*table.Path) + for _, path := range s.globalRib.GetPathList(table.GLOBAL_RIB_NAME, 0, []bgp.RouteFamily{rf}) { + pathsByPeer[path.GetSource()] = append(pathsByPeer[path.GetSource()], path) + } + for peerInfo, paths := range pathsByPeer { + // create copy which can be access to without mutex + var configNeighbor *config.Neighbor + if peer, ok := s.neighborMap[peerInfo.Address.String()]; ok { + configNeighbor = w.s.ToConfig(peer, false) + } + + w.notify(&WatchEventUpdate{ + PeerAS: peerInfo.AS, + PeerAddress: peerInfo.Address, + PeerID: peerInfo.ID, + PostPolicy: true, + Neighbor: configNeighbor, + PathList: paths, + Init: true, + }) + + eor := bgp.NewEndOfRib(rf) + eorBuf, _ := eor.Serialize() + w.notify(&WatchEventUpdate{ + Message: eor, + PeerAS: peerInfo.AS, + PeerAddress: peerInfo.Address, + PeerID: peerInfo.ID, + Timestamp: time.Now(), + Payload: eorBuf, + PostPolicy: true, + Neighbor: configNeighbor, + Init: true, + }) + } + } + } + if w.opts.recvMessage { + register(WATCH_EVENT_TYPE_RECV_MSG, w) + } + + go w.loop() + return nil + }, false) + return w +} diff --git a/pkg/server/server_test.go b/pkg/server/server_test.go new file mode 100644 index 00000000..e4a5e677 --- /dev/null +++ b/pkg/server/server_test.go @@ -0,0 +1,706 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "context" + "net" + "runtime" + "testing" + "time" + + "github.com/osrg/gobgp/internal/pkg/config" + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/pkg/packet/bgp" + + log "github.com/sirupsen/logrus" + "github.com/stretchr/testify/assert" + "github.com/stretchr/testify/require" +) + +func TestModPolicyAssign(t *testing.T) { + assert := assert.New(t) + s := NewBgpServer() + go s.Serve() + err := s.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 1, + RouterId: "1.1.1.1", + Port: -1, + }, + }) + assert.Nil(err) + defer s.Stop() + + err = s.AddPolicy(&table.Policy{Name: "p1"}, false) + assert.Nil(err) + + err = s.AddPolicy(&table.Policy{Name: "p2"}, false) + assert.Nil(err) + + err = s.AddPolicy(&table.Policy{Name: "p3"}, false) + assert.Nil(err) + + err = s.AddPolicyAssignment("", table.POLICY_DIRECTION_IMPORT, + []*config.PolicyDefinition{&config.PolicyDefinition{Name: "p1"}, &config.PolicyDefinition{Name: "p2"}, &config.PolicyDefinition{Name: "p3"}}, table.ROUTE_TYPE_ACCEPT) + assert.Nil(err) + + err = s.DeletePolicyAssignment("", table.POLICY_DIRECTION_IMPORT, + []*config.PolicyDefinition{&config.PolicyDefinition{Name: "p1"}}, false) + assert.Nil(err) + + _, ps, _ := s.GetPolicyAssignment("", table.POLICY_DIRECTION_IMPORT) + assert.Equal(len(ps), 2) +} + +func TestMonitor(test *testing.T) { + assert := assert.New(test) + s := NewBgpServer() + go s.Serve() + err := s.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 1, + RouterId: "1.1.1.1", + Port: 10179, + }, + }) + assert.Nil(err) + defer s.Stop() + + n := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerAs: 2, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + PassiveMode: true, + }, + }, + } + err = s.AddNeighbor(n) + assert.Nil(err) + + t := NewBgpServer() + go t.Serve() + err = t.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 2, + RouterId: "2.2.2.2", + Port: -1, + }, + }) + assert.Nil(err) + defer t.Stop() + + m := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerAs: 1, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + RemotePort: 10179, + }, + }, + } + err = t.AddNeighbor(m) + assert.Nil(err) + + for { + time.Sleep(time.Second) + if t.GetNeighbor("", false)[0].State.SessionState == config.SESSION_STATE_ESTABLISHED { + break + } + } + + // Test WatchBestPath. + w := s.Watch(WatchBestPath(false)) + + // Advertises a route. + attrs := []bgp.PathAttributeInterface{ + bgp.NewPathAttributeOrigin(0), + bgp.NewPathAttributeNextHop("10.0.0.1"), + } + if _, err := t.AddPath("", []*table.Path{table.NewPath(nil, bgp.NewIPAddrPrefix(24, "10.0.0.0"), false, attrs, time.Now(), false)}); err != nil { + log.Fatal(err) + } + ev := <-w.Event() + b := ev.(*WatchEventBestPath) + assert.Equal(1, len(b.PathList)) + assert.Equal("10.0.0.0/24", b.PathList[0].GetNlri().String()) + assert.False(b.PathList[0].IsWithdraw) + + // Withdraws the previous route. + // NOTE: Withdow should not require any path attribute. + if _, err := t.AddPath("", []*table.Path{table.NewPath(nil, bgp.NewIPAddrPrefix(24, "10.0.0.0"), true, nil, time.Now(), false)}); err != nil { + log.Fatal(err) + } + ev = <-w.Event() + b = ev.(*WatchEventBestPath) + assert.Equal(1, len(b.PathList)) + assert.Equal("10.0.0.0/24", b.PathList[0].GetNlri().String()) + assert.True(b.PathList[0].IsWithdraw) + + // Stops the watcher still having an item. + w.Stop() + + // Prepares an initial route to test WatchUpdate with "current" flag. + if _, err := t.AddPath("", []*table.Path{table.NewPath(nil, bgp.NewIPAddrPrefix(24, "10.1.0.0"), false, attrs, time.Now(), false)}); err != nil { + log.Fatal(err) + } + for { + // Waits for the initial route will be advertised. + rib, _, err := s.GetRib("", bgp.RF_IPv4_UC, nil) + if err != nil { + log.Fatal(err) + } + if len(rib.GetKnownPathList("", 0)) > 0 { + break + } + time.Sleep(100 * time.Millisecond) + } + + // Test WatchUpdate with "current" flag. + w = s.Watch(WatchUpdate(true)) + + // Test the initial route. + ev = <-w.Event() + u := ev.(*WatchEventUpdate) + assert.Equal(1, len(u.PathList)) + assert.Equal("10.1.0.0/24", u.PathList[0].GetNlri().String()) + assert.False(u.PathList[0].IsWithdraw) + ev = <-w.Event() + u = ev.(*WatchEventUpdate) + assert.Equal(len(u.PathList), 0) // End of RIB + + // Advertises an additional route. + if _, err := t.AddPath("", []*table.Path{table.NewPath(nil, bgp.NewIPAddrPrefix(24, "10.2.0.0"), false, attrs, time.Now(), false)}); err != nil { + log.Fatal(err) + } + ev = <-w.Event() + u = ev.(*WatchEventUpdate) + assert.Equal(1, len(u.PathList)) + assert.Equal("10.2.0.0/24", u.PathList[0].GetNlri().String()) + assert.False(u.PathList[0].IsWithdraw) + + // Withdraws the previous route. + // NOTE: Withdow should not require any path attribute. + if _, err := t.AddPath("", []*table.Path{table.NewPath(nil, bgp.NewIPAddrPrefix(24, "10.2.0.0"), true, nil, time.Now(), false)}); err != nil { + log.Fatal(err) + } + ev = <-w.Event() + u = ev.(*WatchEventUpdate) + assert.Equal(1, len(u.PathList)) + assert.Equal("10.2.0.0/24", u.PathList[0].GetNlri().String()) + assert.True(u.PathList[0].IsWithdraw) + + // Stops the watcher still having an item. + w.Stop() +} + +func TestNumGoroutineWithAddDeleteNeighbor(t *testing.T) { + assert := assert.New(t) + s := NewBgpServer() + go s.Serve() + err := s.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 1, + RouterId: "1.1.1.1", + Port: -1, + }, + }) + assert.Nil(err) + defer s.Stop() + + // wait a few seconds to avoid taking effect from other test cases. + time.Sleep(time.Second * 5) + + num := runtime.NumGoroutine() + + n := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerAs: 2, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + PassiveMode: true, + }, + }, + } + err = s.AddNeighbor(n) + assert.Nil(err) + + err = s.DeleteNeighbor(n) + assert.Nil(err) + // wait goroutines to finish (e.g. internal goroutine for + // InfiniteChannel) + time.Sleep(time.Second * 5) + assert.Equal(num, runtime.NumGoroutine()) +} + +func newPeerandInfo(myAs, as uint32, address string, rib *table.TableManager) (*Peer, *table.PeerInfo) { + nConf := &config.Neighbor{Config: config.NeighborConfig{PeerAs: as, NeighborAddress: address}} + gConf := &config.Global{Config: config.GlobalConfig{As: myAs}} + config.SetDefaultNeighborConfigValues(nConf, nil, gConf) + policy := table.NewRoutingPolicy() + policy.Reset(&config.RoutingPolicy{}, nil) + p := NewPeer( + &config.Global{Config: config.GlobalConfig{As: myAs}}, + nConf, + rib, + policy) + for _, f := range rib.GetRFlist() { + p.fsm.rfMap[f] = bgp.BGP_ADD_PATH_NONE + } + return p, &table.PeerInfo{AS: as, Address: net.ParseIP(address)} +} + +func process(rib *table.TableManager, l []*table.Path) (*table.Path, *table.Path) { + dsts := make([]*table.Update, 0) + for _, path := range l { + dsts = append(dsts, rib.Update(path)...) + } + news, olds, _ := dstsToPaths(table.GLOBAL_RIB_NAME, 0, dsts) + if len(news) != 1 { + panic("can't handle multiple paths") + } + + return news[0], olds[0] +} + +func TestFilterpathWitheBGP(t *testing.T) { + as := uint32(65000) + p1As := uint32(65001) + p2As := uint32(65002) + rib := table.NewTableManager([]bgp.RouteFamily{bgp.RF_IPv4_UC}) + p1, pi1 := newPeerandInfo(as, p1As, "192.168.0.1", rib) + p2, pi2 := newPeerandInfo(as, p2As, "192.168.0.2", rib) + + nlri := bgp.NewIPAddrPrefix(24, "10.10.10.0") + pa1 := []bgp.PathAttributeInterface{bgp.NewPathAttributeAsPath([]bgp.AsPathParamInterface{bgp.NewAs4PathParam(2, []uint32{p1As})}), bgp.NewPathAttributeLocalPref(200)} + pa2 := []bgp.PathAttributeInterface{bgp.NewPathAttributeAsPath([]bgp.AsPathParamInterface{bgp.NewAs4PathParam(2, []uint32{p2As})})} + + path1 := table.NewPath(pi1, nlri, false, pa1, time.Now(), false) + path2 := table.NewPath(pi2, nlri, false, pa2, time.Now(), false) + rib.Update(path2) + d := rib.Update(path1) + new, old, _ := d[0].GetChanges(table.GLOBAL_RIB_NAME, 0, false) + assert.Equal(t, new, path1) + filterpath(p1, new, old) + filterpath(p2, new, old) + + new, old = process(rib, []*table.Path{path1.Clone(true)}) + assert.Equal(t, new, path2) + // p1 and p2 advertized the same prefix and p1's was best. Then p1 withdraw it, so p2 must get withdawal. + path := filterpath(p2, new, old) + assert.NotNil(t, path) + assert.True(t, path.IsWithdraw) + + // p1 should get the new best (from p2) + assert.Equal(t, filterpath(p1, new, old), path2) + + new, old = process(rib, []*table.Path{path2.Clone(true)}) + assert.True(t, new.IsWithdraw) + // p2 withdraw so p1 should get withdrawal. + path = filterpath(p1, new, old) + assert.True(t, path.IsWithdraw) + + // p2 withdraw so p2 should get nothing. + path = filterpath(p2, new, old) + assert.Nil(t, path) +} + +func TestFilterpathWithiBGP(t *testing.T) { + as := uint32(65000) + + rib := table.NewTableManager([]bgp.RouteFamily{bgp.RF_IPv4_UC}) + p1, pi1 := newPeerandInfo(as, as, "192.168.0.1", rib) + //p2, pi2 := newPeerandInfo(as, as, "192.168.0.2", rib) + p2, _ := newPeerandInfo(as, as, "192.168.0.2", rib) + + nlri := bgp.NewIPAddrPrefix(24, "10.10.10.0") + pa1 := []bgp.PathAttributeInterface{bgp.NewPathAttributeAsPath([]bgp.AsPathParamInterface{bgp.NewAs4PathParam(2, []uint32{as})}), bgp.NewPathAttributeLocalPref(200)} + //pa2 := []bgp.PathAttributeInterface{bgp.NewPathAttributeAsPath([]bgp.AsPathParamInterface{bgp.NewAs4PathParam(2, []uint32{as})})} + + path1 := table.NewPath(pi1, nlri, false, pa1, time.Now(), false) + //path2 := table.NewPath(pi2, nlri, false, pa2, time.Now(), false) + + new, old := process(rib, []*table.Path{path1}) + assert.Equal(t, new, path1) + path := filterpath(p1, new, old) + assert.Nil(t, path) + path = filterpath(p2, new, old) + assert.Nil(t, path) + + new, old = process(rib, []*table.Path{path1.Clone(true)}) + path = filterpath(p1, new, old) + assert.Nil(t, path) + path = filterpath(p2, new, old) + assert.Nil(t, path) + +} + +func TestFilterpathWithRejectPolicy(t *testing.T) { + rib1 := table.NewTableManager([]bgp.RouteFamily{bgp.RF_IPv4_UC}) + _, pi1 := newPeerandInfo(1, 2, "192.168.0.1", rib1) + rib2 := table.NewTableManager([]bgp.RouteFamily{bgp.RF_IPv4_UC}) + p2, _ := newPeerandInfo(1, 3, "192.168.0.2", rib2) + + comSet1 := config.CommunitySet{ + CommunitySetName: "comset1", + CommunityList: []string{"100:100"}, + } + s, _ := table.NewCommunitySet(comSet1) + p2.policy.AddDefinedSet(s) + + statement := config.Statement{ + Name: "stmt1", + Conditions: config.Conditions{ + BgpConditions: config.BgpConditions{ + MatchCommunitySet: config.MatchCommunitySet{ + CommunitySet: "comset1", + }, + }, + }, + Actions: config.Actions{ + RouteDisposition: config.ROUTE_DISPOSITION_REJECT_ROUTE, + }, + } + policy := config.PolicyDefinition{ + Name: "policy1", + Statements: []config.Statement{statement}, + } + p, _ := table.NewPolicy(policy) + p2.policy.AddPolicy(p, false) + policies := []*config.PolicyDefinition{ + &config.PolicyDefinition{ + Name: "policy1", + }, + } + p2.policy.AddPolicyAssignment(p2.TableID(), table.POLICY_DIRECTION_EXPORT, policies, table.ROUTE_TYPE_ACCEPT) + + for _, addCommunity := range []bool{false, true, false, true} { + nlri := bgp.NewIPAddrPrefix(24, "10.10.10.0") + pa1 := []bgp.PathAttributeInterface{bgp.NewPathAttributeAsPath([]bgp.AsPathParamInterface{bgp.NewAs4PathParam(2, []uint32{1})}), bgp.NewPathAttributeLocalPref(200)} + if addCommunity { + pa1 = append(pa1, bgp.NewPathAttributeCommunities([]uint32{100<<16 | 100})) + } + path1 := table.NewPath(pi1, nlri, false, pa1, time.Now(), false) + new, old := process(rib2, []*table.Path{path1}) + assert.Equal(t, new, path1) + s := NewBgpServer() + path2 := s.filterpath(p2, new, old) + if addCommunity { + assert.True(t, path2.IsWithdraw) + } else { + assert.False(t, path2.IsWithdraw) + } + } + +} + +func TestPeerGroup(test *testing.T) { + assert := assert.New(test) + log.SetLevel(log.DebugLevel) + s := NewBgpServer() + go s.Serve() + err := s.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 1, + RouterId: "1.1.1.1", + Port: 10179, + }, + }) + assert.Nil(err) + defer s.Stop() + + g := &config.PeerGroup{ + Config: config.PeerGroupConfig{ + PeerAs: 2, + PeerGroupName: "g", + }, + } + err = s.AddPeerGroup(g) + assert.Nil(err) + + n := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerGroup: "g", + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + PassiveMode: true, + }, + }, + } + configured := map[string]interface{}{ + "config": map[string]interface{}{ + "neigbor-address": "127.0.0.1", + "peer-group": "g", + }, + "transport": map[string]interface{}{ + "config": map[string]interface{}{ + "passive-mode": true, + }, + }, + } + config.RegisterConfiguredFields("127.0.0.1", configured) + err = s.AddNeighbor(n) + assert.Nil(err) + + t := NewBgpServer() + go t.Serve() + err = t.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 2, + RouterId: "2.2.2.2", + Port: -1, + }, + }) + assert.Nil(err) + defer t.Stop() + + m := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerAs: 1, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + RemotePort: 10179, + }, + }, + } + err = t.AddNeighbor(m) + assert.Nil(err) + + for { + time.Sleep(time.Second) + if t.GetNeighbor("", false)[0].State.SessionState == config.SESSION_STATE_ESTABLISHED { + break + } + } +} + +func TestDynamicNeighbor(t *testing.T) { + assert := assert.New(t) + log.SetLevel(log.DebugLevel) + s1 := NewBgpServer() + go s1.Serve() + err := s1.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 1, + RouterId: "1.1.1.1", + Port: 10179, + }, + }) + assert.Nil(err) + defer s1.Stop() + + g := &config.PeerGroup{ + Config: config.PeerGroupConfig{ + PeerAs: 2, + PeerGroupName: "g", + }, + } + err = s1.AddPeerGroup(g) + assert.Nil(err) + + d := &config.DynamicNeighbor{ + Config: config.DynamicNeighborConfig{ + Prefix: "127.0.0.0/24", + PeerGroup: "g", + }, + } + err = s1.AddDynamicNeighbor(d) + assert.Nil(err) + + s2 := NewBgpServer() + go s2.Serve() + err = s2.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 2, + RouterId: "2.2.2.2", + Port: -1, + }, + }) + assert.Nil(err) + defer s2.Stop() + + m := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerAs: 1, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + RemotePort: 10179, + }, + }, + } + err = s2.AddNeighbor(m) + + assert.Nil(err) + + for { + time.Sleep(time.Second) + if s2.GetNeighbor("", false)[0].State.SessionState == config.SESSION_STATE_ESTABLISHED { + break + } + } +} + +func TestGracefulRestartTimerExpired(t *testing.T) { + assert := assert.New(t) + s1 := NewBgpServer() + go s1.Serve() + err := s1.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 1, + RouterId: "1.1.1.1", + Port: 10179, + }, + }) + assert.Nil(err) + defer s1.Stop() + + n := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerAs: 2, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + PassiveMode: true, + }, + }, + GracefulRestart: config.GracefulRestart{ + Config: config.GracefulRestartConfig{ + Enabled: true, + RestartTime: 1, + }, + }, + } + err = s1.AddNeighbor(n) + assert.Nil(err) + + s2 := NewBgpServer() + go s2.Serve() + err = s2.Start(&config.Global{ + Config: config.GlobalConfig{ + As: 2, + RouterId: "2.2.2.2", + Port: -1, + }, + }) + require.NoError(t, err) + defer s2.Stop() + + m := &config.Neighbor{ + Config: config.NeighborConfig{ + NeighborAddress: "127.0.0.1", + PeerAs: 1, + }, + Transport: config.Transport{ + Config: config.TransportConfig{ + RemotePort: 10179, + }, + }, + GracefulRestart: config.GracefulRestart{ + Config: config.GracefulRestartConfig{ + Enabled: true, + RestartTime: 1, + }, + }, + } + err = s2.AddNeighbor(m) + assert.Nil(err) + + // Waiting for BGP session established. + for { + time.Sleep(time.Second) + if s2.GetNeighbor("", false)[0].State.SessionState == config.SESSION_STATE_ESTABLISHED { + break + } + } + + // Force TCP session disconnected in order to cause Graceful Restart at s1 + // side. + for _, n := range s2.neighborMap { + n.fsm.conn.Close() + } + s2.Stop() + + time.Sleep(5 * time.Second) + + // Create dummy session which does NOT send BGP OPEN message in order to + // cause Graceful Restart timer expired. + var conn net.Conn + + conn, err = net.Dial("tcp", "127.0.0.1:10179") + require.NoError(t, err) + defer conn.Close() + + // this seems to take around 22 seconds... need to address this whole thing + ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second) + defer cancel() + + // Waiting for Graceful Restart timer expired and moving on to IDLE state. + for { + if s1.GetNeighbor("", false)[0].State.SessionState == config.SESSION_STATE_IDLE { + break + } + + select { + case <-time.After(time.Second): + case <-ctx.Done(): + t.Fatalf("failed to enter IDLE state in the deadline") + return + } + } +} + +func TestFamiliesForSoftreset(t *testing.T) { + f := func(f bgp.RouteFamily) config.AfiSafi { + return config.AfiSafi{ + State: config.AfiSafiState{ + Family: f, + }, + } + } + peer := &Peer{ + fsm: &FSM{ + pConf: &config.Neighbor{ + AfiSafis: []config.AfiSafi{f(bgp.RF_RTC_UC), f(bgp.RF_IPv4_UC), f(bgp.RF_IPv6_UC)}, + }, + }, + } + + families := familiesForSoftreset(peer, bgp.RF_IPv4_UC) + assert.Equal(t, len(families), 1) + assert.Equal(t, families[0], bgp.RF_IPv4_UC) + + families = familiesForSoftreset(peer, bgp.RF_RTC_UC) + assert.Equal(t, len(families), 1) + assert.Equal(t, families[0], bgp.RF_RTC_UC) + + families = familiesForSoftreset(peer, bgp.RouteFamily(0)) + assert.Equal(t, len(families), 2) + assert.NotContains(t, families, bgp.RF_RTC_UC) +} diff --git a/pkg/server/sockopt.go b/pkg/server/sockopt.go new file mode 100644 index 00000000..e1c9c467 --- /dev/null +++ b/pkg/server/sockopt.go @@ -0,0 +1,90 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +build !linux,!openbsd + +package server + +import ( + "fmt" + "net" + + log "github.com/sirupsen/logrus" +) + +func SetTcpMD5SigSockopt(l *net.TCPListener, address string, key string) error { + return setTcpMD5SigSockopt(l, address, key) +} + +func SetListenTcpTTLSockopt(l *net.TCPListener, ttl int) error { + return setListenTcpTTLSockopt(l, ttl) +} + +func SetTcpTTLSockopt(conn *net.TCPConn, ttl int) error { + return setTcpTTLSockopt(conn, ttl) +} + +func SetTcpMinTTLSockopt(conn *net.TCPConn, ttl int) error { + return setTcpMinTTLSockopt(conn, ttl) +} + +type TCPDialer struct { + net.Dialer + + // MD5 authentication password. + AuthPassword string + + // The TTL value to set outgoing connection. + Ttl uint8 + + // The minimum TTL value for incoming packets. + TtlMin uint8 +} + +func (d *TCPDialer) DialTCP(addr string, port int) (*net.TCPConn, error) { + if d.AuthPassword != "" { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warn("setting md5 for active connection is not supported") + } + if d.Ttl != 0 { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warn("setting ttl for active connection is not supported") + } + if d.TtlMin != 0 { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warn("setting min ttl for active connection is not supported") + } + + raddr, err := net.ResolveTCPAddr("tcp", net.JoinHostPort(addr, fmt.Sprintf("%d", port))) + if err != nil { + return nil, fmt.Errorf("invalid remote address: %s", err) + } + laddr, err := net.ResolveTCPAddr("tcp", d.LocalAddr.String()) + if err != nil { + return nil, fmt.Errorf("invalid local address: %s", err) + } + + dialer := net.Dialer{LocalAddr: laddr, Timeout: d.Timeout} + conn, err := dialer.Dial("tcp", raddr.String()) + if err != nil { + return nil, err + } + return conn.(*net.TCPConn), nil +} diff --git a/pkg/server/sockopt_bsd.go b/pkg/server/sockopt_bsd.go new file mode 100644 index 00000000..651e4e58 --- /dev/null +++ b/pkg/server/sockopt_bsd.go @@ -0,0 +1,89 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +build dragonfly freebsd netbsd + +package server + +import ( + "net" + "os" + "syscall" +) + +const ( + TCP_MD5SIG = 0x10 // TCP MD5 Signature (RFC2385) + IPV6_MINHOPCOUNT = 73 // Generalized TTL Security Mechanism (RFC5082) +) + +func setsockoptTcpMD5Sig(fd int, address string, key string) error { + // always enable and assumes that the configuration is done by setkey() + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, syscall.IPPROTO_TCP, TCP_MD5SIG, 1)) +} + +func setTcpMD5SigSockopt(l *net.TCPListener, address string, key string) error { + fi, _, err := extractFileAndFamilyFromTCPListener(l) + defer fi.Close() + if err != nil { + return err + } + return setsockoptTcpMD5Sig(int(fi.Fd()), address, key) +} + +func setsockoptIpTtl(fd int, family int, value int) error { + level := syscall.IPPROTO_IP + name := syscall.IP_TTL + if family == syscall.AF_INET6 { + level = syscall.IPPROTO_IPV6 + name = syscall.IPV6_UNICAST_HOPS + } + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, level, name, value)) +} + +func setListenTcpTTLSockopt(l *net.TCPListener, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPListener(l) + defer fi.Close() + if err != nil { + return err + } + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func setTcpTTLSockopt(conn *net.TCPConn, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPConn(conn) + defer fi.Close() + if err != nil { + return err + } + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func setsockoptIpMinTtl(fd int, family int, value int) error { + level := syscall.IPPROTO_IP + name := syscall.IP_MINTTL + if family == syscall.AF_INET6 { + level = syscall.IPPROTO_IPV6 + name = IPV6_MINHOPCOUNT + } + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, level, name, value)) +} + +func setTcpMinTTLSockopt(conn *net.TCPConn, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPConn(conn) + defer fi.Close() + if err != nil { + return err + } + return setsockoptIpMinTtl(int(fi.Fd()), family, ttl) +} diff --git a/pkg/server/sockopt_darwin.go b/pkg/server/sockopt_darwin.go new file mode 100644 index 00000000..4bad54ff --- /dev/null +++ b/pkg/server/sockopt_darwin.go @@ -0,0 +1,64 @@ +// Copyright (C) 2016-2017 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +build darwin + +package server + +import ( + "fmt" + "net" + "os" + "syscall" +) + +func setTcpMD5SigSockopt(l *net.TCPListener, address string, key string) error { + return fmt.Errorf("setting md5 is not supported") +} + +func setsockoptIpTtl(fd int, family int, value int) error { + level := syscall.IPPROTO_IP + name := syscall.IP_TTL + if family == syscall.AF_INET6 { + level = syscall.IPPROTO_IPV6 + name = syscall.IPV6_UNICAST_HOPS + } + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, level, name, value)) +} + +func setListenTcpTTLSockopt(l *net.TCPListener, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPListener(l) + if err != nil { + return err + } + + defer fi.Close() + + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func setTcpTTLSockopt(conn *net.TCPConn, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPConn(conn) + if err != nil { + return err + } + + defer fi.Close() + + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func setTcpMinTTLSockopt(conn *net.TCPConn, ttl int) error { + return fmt.Errorf("setting min ttl is not supported") +} diff --git a/pkg/server/sockopt_linux.go b/pkg/server/sockopt_linux.go new file mode 100644 index 00000000..9fe02ba5 --- /dev/null +++ b/pkg/server/sockopt_linux.go @@ -0,0 +1,282 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +build linux + +package server + +import ( + "fmt" + "net" + "os" + "syscall" + "unsafe" +) + +const ( + TCP_MD5SIG = 14 // TCP MD5 Signature (RFC2385) + IPV6_MINHOPCOUNT = 73 // Generalized TTL Security Mechanism (RFC5082) +) + +type tcpmd5sig struct { + ss_family uint16 + ss [126]byte + // padding the struct + _ uint16 + keylen uint16 + // padding the struct + _ uint32 + key [80]byte +} + +func buildTcpMD5Sig(address string, key string) (tcpmd5sig, error) { + t := tcpmd5sig{} + addr := net.ParseIP(address) + if addr.To4() != nil { + t.ss_family = syscall.AF_INET + copy(t.ss[2:], addr.To4()) + } else { + t.ss_family = syscall.AF_INET6 + copy(t.ss[6:], addr.To16()) + } + + t.keylen = uint16(len(key)) + copy(t.key[0:], []byte(key)) + + return t, nil +} + +func setsockoptTcpMD5Sig(fd int, address string, key string) error { + t, err := buildTcpMD5Sig(address, key) + if err != nil { + return err + } + b := *(*[unsafe.Sizeof(t)]byte)(unsafe.Pointer(&t)) + return os.NewSyscallError("setsockopt", syscall.SetsockoptString(fd, syscall.IPPROTO_TCP, TCP_MD5SIG, string(b[:]))) +} + +func SetTcpMD5SigSockopt(l *net.TCPListener, address string, key string) error { + fi, _, err := extractFileAndFamilyFromTCPListener(l) + if err != nil { + return err + } + defer fi.Close() + + return setsockoptTcpMD5Sig(int(fi.Fd()), address, key) +} + +func setsockoptIpTtl(fd int, family int, value int) error { + level := syscall.IPPROTO_IP + name := syscall.IP_TTL + if family == syscall.AF_INET6 { + level = syscall.IPPROTO_IPV6 + name = syscall.IPV6_UNICAST_HOPS + } + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, level, name, value)) +} + +func SetListenTcpTTLSockopt(l *net.TCPListener, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPListener(l) + if err != nil { + return err + } + defer fi.Close() + + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func SetTcpTTLSockopt(conn *net.TCPConn, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPConn(conn) + if err != nil { + return err + } + defer fi.Close() + + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func setsockoptIpMinTtl(fd int, family int, value int) error { + level := syscall.IPPROTO_IP + name := syscall.IP_MINTTL + if family == syscall.AF_INET6 { + level = syscall.IPPROTO_IPV6 + name = IPV6_MINHOPCOUNT + } + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, level, name, value)) +} + +func SetTcpMinTTLSockopt(conn *net.TCPConn, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPConn(conn) + if err != nil { + return err + } + defer fi.Close() + + return setsockoptIpMinTtl(int(fi.Fd()), family, ttl) +} + +type TCPDialer struct { + net.Dialer + + // MD5 authentication password. + AuthPassword string + + // The TTL value to set outgoing connection. + Ttl uint8 + + // The minimum TTL value for incoming packets. + TtlMin uint8 +} + +func (d *TCPDialer) DialTCP(addr string, port int) (*net.TCPConn, error) { + var family int + var ra, la syscall.Sockaddr + + raddr, err := net.ResolveTCPAddr("tcp", net.JoinHostPort(addr, fmt.Sprintf("%d", port))) + if err != nil { + return nil, fmt.Errorf("invalid remote address: %s", err) + } + laddr, err := net.ResolveTCPAddr("tcp", d.LocalAddr.String()) + if err != nil { + return nil, fmt.Errorf("invalid local address: %s", err) + } + if raddr.IP.To4() != nil { + family = syscall.AF_INET + rsockaddr := &syscall.SockaddrInet4{Port: port} + copy(rsockaddr.Addr[:], raddr.IP.To4()) + ra = rsockaddr + lsockaddr := &syscall.SockaddrInet4{} + copy(lsockaddr.Addr[:], laddr.IP.To4()) + la = lsockaddr + } else { + family = syscall.AF_INET6 + rsockaddr := &syscall.SockaddrInet6{Port: port} + copy(rsockaddr.Addr[:], raddr.IP.To16()) + ra = rsockaddr + var zone uint32 + if laddr.Zone != "" { + if intf, err := net.InterfaceByName(laddr.Zone); err != nil { + return nil, err + } else { + zone = uint32(intf.Index) + } + } + lsockaddr := &syscall.SockaddrInet6{ZoneId: zone} + copy(lsockaddr.Addr[:], laddr.IP.To16()) + la = lsockaddr + } + + sockType := syscall.SOCK_STREAM | syscall.SOCK_CLOEXEC | syscall.SOCK_NONBLOCK + proto := 0 + fd, err := syscall.Socket(family, sockType, proto) + if err != nil { + return nil, err + } + fi := os.NewFile(uintptr(fd), "") + defer fi.Close() + // A new socket was created so we must close it before this + // function returns either on failure or success. On success, + // net.FileConn() in newTCPConn() increases the refcount of + // the socket so this fi.Close() doesn't destroy the socket. + // The caller must call Close() with the file later. + // Note that the above os.NewFile() doesn't play with the + // refcount. + + if err = syscall.SetsockoptInt(fd, syscall.SOL_SOCKET, syscall.SO_BROADCAST, 1); err != nil { + return nil, os.NewSyscallError("setsockopt", err) + } + + if err = syscall.SetsockoptInt(fd, syscall.IPPROTO_TCP, syscall.TCP_NODELAY, 1); err != nil { + return nil, os.NewSyscallError("setsockopt", err) + } + + if d.AuthPassword != "" { + if err = setsockoptTcpMD5Sig(fd, addr, d.AuthPassword); err != nil { + return nil, err + } + } + + if d.Ttl != 0 { + if err = setsockoptIpTtl(fd, family, int(d.Ttl)); err != nil { + return nil, err + } + } + + if d.TtlMin != 0 { + if err = setsockoptIpMinTtl(fd, family, int(d.Ttl)); err != nil { + return nil, err + } + } + + if err = syscall.Bind(fd, la); err != nil { + return nil, os.NewSyscallError("bind", err) + } + + newTCPConn := func(fi *os.File) (*net.TCPConn, error) { + if conn, err := net.FileConn(fi); err != nil { + return nil, err + } else { + return conn.(*net.TCPConn), err + } + } + + err = syscall.Connect(fd, ra) + switch err { + case syscall.EINPROGRESS, syscall.EALREADY, syscall.EINTR: + // do timeout handling + case nil, syscall.EISCONN: + return newTCPConn(fi) + default: + return nil, os.NewSyscallError("connect", err) + } + + epfd, e := syscall.EpollCreate1(syscall.EPOLL_CLOEXEC) + if e != nil { + return nil, e + } + defer syscall.Close(epfd) + + var event syscall.EpollEvent + events := make([]syscall.EpollEvent, 1) + + event.Events = syscall.EPOLLIN | syscall.EPOLLOUT | syscall.EPOLLPRI + event.Fd = int32(fd) + if e = syscall.EpollCtl(epfd, syscall.EPOLL_CTL_ADD, fd, &event); e != nil { + return nil, e + } + + for { + nevents, e := syscall.EpollWait(epfd, events, int(d.Timeout/1000000) /*msec*/) + if e != nil { + return nil, e + } + if nevents == 0 { + return nil, fmt.Errorf("timeout") + } else if nevents == 1 && events[0].Fd == int32(fd) { + nerr, err := syscall.GetsockoptInt(fd, syscall.SOL_SOCKET, syscall.SO_ERROR) + if err != nil { + return nil, os.NewSyscallError("getsockopt", err) + } + switch err := syscall.Errno(nerr); err { + case syscall.EINPROGRESS, syscall.EALREADY, syscall.EINTR: + case syscall.Errno(0), syscall.EISCONN: + return newTCPConn(fi) + default: + return nil, os.NewSyscallError("getsockopt", err) + } + } else { + return nil, fmt.Errorf("unexpected epoll behavior") + } + } +} diff --git a/pkg/server/sockopt_linux_test.go b/pkg/server/sockopt_linux_test.go new file mode 100644 index 00000000..a08e7fc7 --- /dev/null +++ b/pkg/server/sockopt_linux_test.go @@ -0,0 +1,106 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +build linux + +package server + +import ( + "bytes" + "fmt" + "net" + "os" + "syscall" + "testing" + "time" + "unsafe" +) + +func Test_buildTcpMD5Sig(t *testing.T) { + s, _ := buildTcpMD5Sig("1.2.3.4", "hello") + + if unsafe.Sizeof(s) != 216 { + t.Error("TCPM5Sig struct size is wrong", unsafe.Sizeof(s)) + } + + buf1 := make([]uint8, 216) + p := unsafe.Pointer(&s) + for i := uintptr(0); i < 216; i++ { + buf1[i] = *(*byte)(unsafe.Pointer(uintptr(p) + i)) + } + + buf2 := []uint8{2, 0, 0, 0, 1, 2, 3, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 0, 104, 101, 108, 108, 111, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0} + + if bytes.Equal(buf1, buf2) { + t.Log("OK") + } else { + t.Error("Something wrong v4") + } +} + +func Test_buildTcpMD5Sigv6(t *testing.T) { + s, _ := buildTcpMD5Sig("fe80::4850:31ff:fe01:fc55", "helloworld") + + buf1 := make([]uint8, 216) + p := unsafe.Pointer(&s) + for i := uintptr(0); i < 216; i++ { + buf1[i] = *(*byte)(unsafe.Pointer(uintptr(p) + i)) + } + + buf2 := []uint8{10, 0, 0, 0, 0, 0, 0, 0, 254, 128, 0, 0, 0, 0, 0, 0, 72, 80, 49, 255, 254, 1, 252, 85, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 10, 0, 0, 0, 0, 0, 104, 101, 108, 108, 111, 119, 111, 114, 108, 100, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0} + + buf2[0] = syscall.AF_INET6 + + if bytes.Equal(buf1, buf2) { + t.Log("OK") + } else { + t.Error("Something wrong v6") + } +} + +func Test_DialTCP_FDleak(t *testing.T) { + openFds := func() int { + pid := os.Getpid() + f, err := os.OpenFile(fmt.Sprintf("/proc/%d/fdinfo", pid), os.O_RDONLY, 0) + if err != nil { + t.Fatal(err) + } + defer f.Close() + names, err := f.Readdirnames(0) + if err != nil { + t.Fatal(err) + } + return len(names) + } + + before := openFds() + + for i := 0; i < 10; i++ { + laddr, _ := net.ResolveTCPAddr("tcp", net.JoinHostPort("127.0.0.1", "0")) + d := TCPDialer{ + Dialer: net.Dialer{ + LocalAddr: laddr, + Timeout: 1 * time.Second, + }, + } + if _, err := d.DialTCP("127.0.0.1", 1); err == nil { + t.Fatalf("should not succeed") + } + + } + + if after := openFds(); before != after { + t.Fatalf("could be fd leak, %d %d", before, after) + } +} diff --git a/pkg/server/sockopt_openbsd.go b/pkg/server/sockopt_openbsd.go new file mode 100644 index 00000000..f52c1447 --- /dev/null +++ b/pkg/server/sockopt_openbsd.go @@ -0,0 +1,469 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +build openbsd + +package server + +import ( + "encoding/binary" + "fmt" + "net" + "os" + "syscall" + "unsafe" + + log "github.com/sirupsen/logrus" +) + +const ( + PF_KEY_V2 = 2 + + SADB_X_SATYPE_TCPSIGNATURE = 8 + + SADB_EXT_SA = 1 + SADB_EXT_ADDRESS_SRC = 5 + SADB_EXT_ADDRESS_DST = 6 + SADB_EXT_KEY_AUTH = 8 + SADB_EXT_SPIRANGE = 16 + + SADB_GETSPI = 1 + SADB_UPDATE = 2 + SADB_DELETE = 4 + + SADB_X_EALG_AES = 12 + + SADB_SASTATE_MATURE = 1 +) + +type sadbMsg struct { + sadbMsgVersion uint8 + sadbMsgType uint8 + sadbMsgErrno uint8 + sadbMsgSatype uint8 + sadbMsgLen uint16 + sadbMsgReserved uint16 + sadbMsgSeq uint32 + sadbMsgPid uint32 +} + +func (s *sadbMsg) DecodeFromBytes(data []byte) error { + if len(data) < SADB_MSG_SIZE { + fmt.Errorf("too short for sadbMsg %d", len(data)) + } + s.sadbMsgVersion = data[0] + s.sadbMsgType = data[1] + s.sadbMsgErrno = data[2] + s.sadbMsgSatype = data[3] + s.sadbMsgLen = binary.LittleEndian.Uint16(data[4:6]) + s.sadbMsgSeq = binary.LittleEndian.Uint32(data[8:12]) + s.sadbMsgPid = binary.LittleEndian.Uint32(data[12:16]) + return nil +} + +type sadbSpirange struct { + sadbSpirangeLen uint16 + sadbSpirangeExttype uint16 + sadbSpirangeMin uint32 + sadbSpirangeMax uint32 + sadbSpirangeReserved uint32 +} + +type sadbAddress struct { + sadbAddressLen uint16 + sadbAddressExttype uint16 + sadbAddressReserved uint32 +} + +type sadbExt struct { + sadbExtLen uint16 + sadbExtType uint16 +} + +type sadbSa struct { + sadbSaLen uint16 + sadbSaExttype uint16 + sadbSaSpi uint32 + sadbSaReplay uint8 + sadbSaState uint8 + sadbSaAuth uint8 + sadbSaEncrypt uint8 + sadbSaFlags uint32 +} + +type sadbKey struct { + sadbKeyLen uint16 + sadbKeyExttype uint16 + sadbKeyBits uint16 + sadbKeyReserved uint16 +} + +const ( + SADB_MSG_SIZE = int(unsafe.Sizeof(sadbMsg{})) + SADB_SPIRANGE_SIZE = int(unsafe.Sizeof(sadbSpirange{})) + SADB_ADDRESS_SIZE = int(unsafe.Sizeof(sadbAddress{})) + SADB_SA_SIZE = int(unsafe.Sizeof(sadbSa{})) + SADB_KEY_SIZE = int(unsafe.Sizeof(sadbKey{})) +) + +type sockaddrIn struct { + ssLen uint8 + ssFamily uint8 + ssPort uint16 + ssAddr uint32 + pad [8]byte +} + +func newSockaddrIn(addr string) sockaddrIn { + if len(addr) == 0 { + return sockaddrIn{ + ssLen: 16, + } + } + v := net.ParseIP(addr).To4() + return sockaddrIn{ + ssAddr: uint32(v[3])<<24 | uint32(v[2])<<16 | uint32(v[1])<<8 | uint32(v[0]), + ssLen: 16, + ssFamily: syscall.AF_INET, + } +} + +func roundUp(v int) int { + if v%8 != 0 { + v += 8 - v%8 + } + return v +} + +func b(p unsafe.Pointer, length int) []byte { + buf := make([]byte, length) + for i := 0; i < length; i++ { + buf[i] = *(*byte)(p) + p = unsafe.Pointer(uintptr(p) + 1) + } + return buf +} + +var seq uint32 +var fd int + +var spiInMap map[string]uint32 = map[string]uint32{} +var spiOutMap map[string]uint32 = map[string]uint32{} + +func pfkeyReply() (spi uint32, err error) { + buf := make([]byte, SADB_MSG_SIZE) + if count, _, _, _, _ := syscall.Recvmsg(fd, buf, nil, syscall.MSG_PEEK); count != len(buf) { + return spi, fmt.Errorf("incomplete sadb msg %d %d", len(buf), count) + } + h := sadbMsg{} + h.DecodeFromBytes(buf) + if h.sadbMsgErrno != 0 { + return spi, fmt.Errorf("sadb msg reply error %d", h.sadbMsgErrno) + } + + if h.sadbMsgSeq != seq { + return spi, fmt.Errorf("sadb msg sequence doesn't match %d %d", h.sadbMsgSeq, seq) + } + + if h.sadbMsgPid != uint32(os.Getpid()) { + return spi, fmt.Errorf("sadb msg pid doesn't match %d %d", h.sadbMsgPid, os.Getpid()) + } + + buf = make([]byte, int(8*h.sadbMsgLen)) + if count, _, _, _, _ := syscall.Recvmsg(fd, buf, nil, 0); count != len(buf) { + return spi, fmt.Errorf("incomplete sadb msg body %d %d", len(buf), count) + } + + buf = buf[SADB_MSG_SIZE:] + + for len(buf) >= 4 { + l := binary.LittleEndian.Uint16(buf[0:2]) * 8 + t := binary.LittleEndian.Uint16(buf[2:4]) + if t == SADB_EXT_SA { + return binary.LittleEndian.Uint32(buf[4:8]), nil + } + + if len(buf) <= int(l) { + break + } + buf = buf[l:] + } + return spi, err +} + +func sendSadbMsg(msg *sadbMsg, body []byte) (err error) { + if fd == 0 { + fd, err = syscall.Socket(syscall.AF_KEY, syscall.SOCK_RAW, PF_KEY_V2) + if err != nil { + return err + } + } + + seq++ + msg.sadbMsgSeq = seq + msg.sadbMsgLen = uint16((len(body) + SADB_MSG_SIZE) / 8) + + buf := append(b(unsafe.Pointer(msg), SADB_MSG_SIZE), body...) + + r, err := syscall.Write(fd, buf) + if r != len(buf) { + return fmt.Errorf("short write %d %d", r, len(buf)) + } + return err +} + +func rfkeyRequest(msgType uint8, src, dst string, spi uint32, key string) error { + h := sadbMsg{ + sadbMsgVersion: PF_KEY_V2, + sadbMsgType: msgType, + sadbMsgSatype: SADB_X_SATYPE_TCPSIGNATURE, + sadbMsgPid: uint32(os.Getpid()), + } + + ssrc := newSockaddrIn(src) + sa_src := sadbAddress{ + sadbAddressExttype: SADB_EXT_ADDRESS_SRC, + sadbAddressLen: uint16(SADB_ADDRESS_SIZE+roundUp(int(ssrc.ssLen))) / 8, + } + + sdst := newSockaddrIn(dst) + sa_dst := sadbAddress{ + sadbAddressExttype: SADB_EXT_ADDRESS_DST, + sadbAddressLen: uint16(SADB_ADDRESS_SIZE+roundUp(int(sdst.ssLen))) / 8, + } + + buf := make([]byte, 0) + switch msgType { + case SADB_UPDATE, SADB_DELETE: + sa := sadbSa{ + sadbSaLen: uint16(SADB_SA_SIZE / 8), + sadbSaExttype: SADB_EXT_SA, + sadbSaSpi: spi, + sadbSaState: SADB_SASTATE_MATURE, + sadbSaEncrypt: SADB_X_EALG_AES, + } + buf = append(buf, b(unsafe.Pointer(&sa), SADB_SA_SIZE)...) + case SADB_GETSPI: + spirange := sadbSpirange{ + sadbSpirangeLen: uint16(SADB_SPIRANGE_SIZE) / 8, + sadbSpirangeExttype: SADB_EXT_SPIRANGE, + sadbSpirangeMin: 0x100, + sadbSpirangeMax: 0xffffffff, + } + buf = append(buf, b(unsafe.Pointer(&spirange), SADB_SPIRANGE_SIZE)...) + } + + buf = append(buf, b(unsafe.Pointer(&sa_dst), SADB_ADDRESS_SIZE)...) + buf = append(buf, b(unsafe.Pointer(&sdst), roundUp(int(sdst.ssLen)))...) + buf = append(buf, b(unsafe.Pointer(&sa_src), SADB_ADDRESS_SIZE)...) + buf = append(buf, b(unsafe.Pointer(&ssrc), roundUp(int(ssrc.ssLen)))...) + + switch msgType { + case SADB_UPDATE: + keylen := roundUp(len(key)) + sa_akey := sadbKey{ + sadbKeyLen: uint16((SADB_KEY_SIZE + keylen) / 8), + sadbKeyExttype: SADB_EXT_KEY_AUTH, + sadbKeyBits: uint16(len(key) * 8), + } + k := []byte(key) + if pad := keylen - len(k); pad != 0 { + k = append(k, make([]byte, pad)...) + } + buf = append(buf, b(unsafe.Pointer(&sa_akey), SADB_KEY_SIZE)...) + buf = append(buf, k...) + } + + return sendSadbMsg(&h, buf) +} + +func saAdd(address, key string) error { + f := func(src, dst string) error { + if err := rfkeyRequest(SADB_GETSPI, src, dst, 0, ""); err != nil { + return err + } + spi, err := pfkeyReply() + if err != nil { + return err + } + if src == "" { + spiOutMap[address] = spi + } else { + spiInMap[address] = spi + } + + if err := rfkeyRequest(SADB_UPDATE, src, dst, spi, key); err != nil { + return err + } + _, err = pfkeyReply() + return err + } + + if err := f(address, ""); err != nil { + return err + } + + return f("", address) +} + +func saDelete(address string) error { + if spi, y := spiInMap[address]; y { + if err := rfkeyRequest(SADB_DELETE, address, "", spi, ""); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": address, + }).Info("failed to delete md5 for incoming") + } else { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": address, + }).Info("can't find spi for md5 for incoming") + } + } + if spi, y := spiOutMap[address]; y { + if err := rfkeyRequest(SADB_DELETE, "", address, spi, ""); err != nil { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": address, + }).Info("failed to delete md5 for outgoing") + } else { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": address, + }).Info("can't find spi for md5 for outgoing") + } + } + return nil +} + +const ( + TCP_MD5SIG = 0x4 // TCP MD5 Signature (RFC2385) + IPV6_MINHOPCOUNT = 73 // Generalized TTL Security Mechanism (RFC5082) +) + +func setsockoptTcpMD5Sig(fd int, address string, key string) error { + if err := syscall.SetsockoptInt(fd, syscall.IPPROTO_TCP, TCP_MD5SIG, 1); err != nil { + return os.NewSyscallError("setsockopt", err) + } + if len(key) > 0 { + return saAdd(address, key) + } + return saDelete(address) +} + +func SetTcpMD5SigSockopt(l *net.TCPListener, address string, key string) error { + fi, _, err := extractFileAndFamilyFromTCPListener(l) + defer fi.Close() + if err != nil { + return err + } + return setsockoptTcpMD5Sig(int(fi.Fd()), address, key) +} + +func setsockoptIpTtl(fd int, family int, value int) error { + level := syscall.IPPROTO_IP + name := syscall.IP_TTL + if family == syscall.AF_INET6 { + level = syscall.IPPROTO_IPV6 + name = syscall.IPV6_UNICAST_HOPS + } + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, level, name, value)) +} + +func SetListenTcpTTLSockopt(l *net.TCPListener, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPListener(l) + defer fi.Close() + if err != nil { + return err + } + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func SetTcpTTLSockopt(conn *net.TCPConn, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPConn(conn) + defer fi.Close() + if err != nil { + return err + } + return setsockoptIpTtl(int(fi.Fd()), family, ttl) +} + +func setsockoptIpMinTtl(fd int, family int, value int) error { + level := syscall.IPPROTO_IP + name := syscall.IP_MINTTL + if family == syscall.AF_INET6 { + level = syscall.IPPROTO_IPV6 + name = IPV6_MINHOPCOUNT + } + return os.NewSyscallError("setsockopt", syscall.SetsockoptInt(fd, level, name, value)) +} + +func SetTcpMinTTLSockopt(conn *net.TCPConn, ttl int) error { + fi, family, err := extractFileAndFamilyFromTCPConn(conn) + defer fi.Close() + if err != nil { + return err + } + return setsockoptIpMinTtl(int(fi.Fd()), family, ttl) +} + +type TCPDialer struct { + net.Dialer + + // MD5 authentication password. + AuthPassword string + + // The TTL value to set outgoing connection. + Ttl uint8 + + // The minimum TTL value for incoming packets. + TtlMin uint8 +} + +func (d *TCPDialer) DialTCP(addr string, port int) (*net.TCPConn, error) { + if d.AuthPassword != "" { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warn("setting md5 for active connection is not supported") + } + if d.Ttl != 0 { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warn("setting ttl for active connection is not supported") + } + if d.TtlMin != 0 { + log.WithFields(log.Fields{ + "Topic": "Peer", + "Key": addr, + }).Warn("setting min ttl for active connection is not supported") + } + + raddr, err := net.ResolveTCPAddr("tcp", net.JoinHostPort(addr, fmt.Sprintf("%d", port))) + if err != nil { + return nil, fmt.Errorf("invalid remote address: %s", err) + } + laddr, err := net.ResolveTCPAddr("tcp", d.LocalAddr.String()) + if err != nil { + return nil, fmt.Errorf("invalid local address: %s", err) + } + + dialer := net.Dialer{LocalAddr: laddr, Timeout: d.Timeout} + conn, err := dialer.Dial("tcp", raddr.String()) + if err != nil { + return nil, err + } + return conn.(*net.TCPConn), nil +} diff --git a/pkg/server/sockopt_stub.go b/pkg/server/sockopt_stub.go new file mode 100644 index 00000000..9e888dc5 --- /dev/null +++ b/pkg/server/sockopt_stub.go @@ -0,0 +1,38 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +build !linux,!dragonfly,!freebsd,!netbsd,!openbsd,!darwin + +package server + +import ( + "fmt" + "net" +) + +func setTcpMD5SigSockopt(l *net.TCPListener, address string, key string) error { + return fmt.Errorf("setting md5 is not supported") +} + +func setListenTcpTTLSockopt(l *net.TCPListener, ttl int) error { + return fmt.Errorf("setting ttl is not supported") +} + +func setTcpTTLSockopt(conn *net.TCPConn, ttl int) error { + return fmt.Errorf("setting ttl is not supported") +} + +func setTcpMinTTLSockopt(conn *net.TCPConn, ttl int) error { + return fmt.Errorf("setting min ttl is not supported") +} diff --git a/pkg/server/util.go b/pkg/server/util.go new file mode 100644 index 00000000..ba6fe70a --- /dev/null +++ b/pkg/server/util.go @@ -0,0 +1,117 @@ +// Copyright (C) 2016 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "net" + "os" + "strings" + "syscall" + + "github.com/eapache/channels" + + "github.com/osrg/gobgp/pkg/packet/bgp" +) + +func cleanInfiniteChannel(ch *channels.InfiniteChannel) { + ch.Close() + // drain all remaining items + for range ch.Out() { + } +} + +// Returns the binary formatted Administrative Shutdown Communication from the +// given string value. +func newAdministrativeCommunication(communication string) (data []byte) { + if communication == "" { + return nil + } + com := []byte(communication) + if len(com) > bgp.BGP_ERROR_ADMINISTRATIVE_COMMUNICATION_MAX { + data = []byte{bgp.BGP_ERROR_ADMINISTRATIVE_COMMUNICATION_MAX} + data = append(data, com[:bgp.BGP_ERROR_ADMINISTRATIVE_COMMUNICATION_MAX]...) + } else { + data = []byte{byte(len(com))} + data = append(data, com...) + } + return data +} + +// Parses the given NOTIFICATION message data as a binary value and returns +// the Administrative Shutdown Communication in string and the rest binary. +func decodeAdministrativeCommunication(data []byte) (string, []byte) { + if len(data) == 0 { + return "", data + } + communicationLen := int(data[0]) + if communicationLen > bgp.BGP_ERROR_ADMINISTRATIVE_COMMUNICATION_MAX { + communicationLen = bgp.BGP_ERROR_ADMINISTRATIVE_COMMUNICATION_MAX + } + if communicationLen > len(data)+1 { + communicationLen = len(data) + 1 + } + return string(data[1 : communicationLen+1]), data[communicationLen+1:] +} + +func extractFileAndFamilyFromTCPListener(l *net.TCPListener) (*os.File, int, error) { + // Note #1: TCPListener.File() has the unexpected side-effect of putting + // the original socket into blocking mode. See Note #2. + fi, err := l.File() + if err != nil { + return nil, 0, err + } + + // Note #2: Call net.FileListener() to put the original socket back into + // non-blocking mode. + fl, err := net.FileListener(fi) + if err != nil { + fi.Close() + return nil, 0, err + } + fl.Close() + + family := syscall.AF_INET + if strings.Contains(l.Addr().String(), "[") { + family = syscall.AF_INET6 + } + + return fi, family, nil +} + +func extractFileAndFamilyFromTCPConn(conn *net.TCPConn) (*os.File, int, error) { + // Note #1: TCPConn.File() has the unexpected side-effect of putting + // the original socket into blocking mode. See Note #2. + fi, err := conn.File() + if err != nil { + return nil, 0, err + } + + // Note #2: Call net.FileConn() to put the original socket back into + // non-blocking mode. + fc, err := net.FileConn(fi) + if err != nil { + fi.Close() + return nil, 0, err + } + fc.Close() + + family := syscall.AF_INET + if strings.Contains(conn.RemoteAddr().String(), "[") { + family = syscall.AF_INET6 + } + + return fi, family, nil +} diff --git a/pkg/server/zclient.go b/pkg/server/zclient.go new file mode 100644 index 00000000..3ef057ce --- /dev/null +++ b/pkg/server/zclient.go @@ -0,0 +1,450 @@ +// Copyright (C) 2015 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "fmt" + "math" + "net" + "strconv" + "strings" + "syscall" + "time" + + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/internal/pkg/zebra" + "github.com/osrg/gobgp/pkg/packet/bgp" + + log "github.com/sirupsen/logrus" +) + +// nexthopStateCache stores a map of nexthop IP to metric value. Especially, +// the metric value of math.MaxUint32 means the nexthop is unreachable. +type nexthopStateCache map[string]uint32 + +func (m nexthopStateCache) applyToPathList(paths []*table.Path) []*table.Path { + updated := make([]*table.Path, 0, len(paths)) + for _, path := range paths { + if path == nil || path.IsWithdraw { + continue + } + metric, ok := m[path.GetNexthop().String()] + if !ok { + continue + } + isNexthopInvalid := metric == math.MaxUint32 + med, err := path.GetMed() + if err == nil && med == metric && path.IsNexthopInvalid == isNexthopInvalid { + // If the nexthop state of the given path is already up to date, + // skips this path. + continue + } + newPath := path.Clone(false) + if isNexthopInvalid { + newPath.IsNexthopInvalid = true + } else { + newPath.IsNexthopInvalid = false + newPath.SetMed(int64(metric), true) + } + updated = append(updated, newPath) + } + return updated +} + +func (m nexthopStateCache) updateByNexthopUpdate(body *zebra.NexthopUpdateBody) (updated bool) { + if len(body.Nexthops) == 0 { + // If NEXTHOP_UPDATE message does not contain any nexthop, the given + // nexthop is unreachable. + if _, ok := m[body.Prefix.String()]; !ok { + // Zebra will send an empty NEXTHOP_UPDATE message as the fist + // response for the NEXTHOP_REGISTER message. Here ignores it. + return false + } + m[body.Prefix.String()] = math.MaxUint32 // means unreachable + } else { + m[body.Prefix.String()] = body.Metric + } + return true +} + +func (m nexthopStateCache) filterPathToRegister(paths []*table.Path) []*table.Path { + filteredPaths := make([]*table.Path, 0, len(paths)) + for _, path := range paths { + // Here filters out: + // - Nil path + // - Withdrawn path + // - External path (advertised from Zebra) in order avoid sending back + // - Unspecified nexthop address + // - Already registered nexthop + if path == nil || path.IsWithdraw || path.IsFromExternal() { + continue + } else if nexthop := path.GetNexthop(); nexthop.IsUnspecified() { + continue + } else if _, ok := m[nexthop.String()]; ok { + continue + } + filteredPaths = append(filteredPaths, path) + } + return filteredPaths +} + +func filterOutExternalPath(paths []*table.Path) []*table.Path { + filteredPaths := make([]*table.Path, 0, len(paths)) + for _, path := range paths { + // Here filters out: + // - Nil path + // - External path (advertised from Zebra) in order avoid sending back + // - Unreachable path because invalidated by Zebra + if path == nil || path.IsFromExternal() || path.IsNexthopInvalid { + continue + } + filteredPaths = append(filteredPaths, path) + } + return filteredPaths +} + +func newIPRouteBody(dst []*table.Path) (body *zebra.IPRouteBody, isWithdraw bool) { + paths := filterOutExternalPath(dst) + if len(paths) == 0 { + return nil, false + } + path := paths[0] + + l := strings.SplitN(path.GetNlri().String(), "/", 2) + var prefix net.IP + nexthops := make([]net.IP, 0, len(paths)) + switch path.GetRouteFamily() { + case bgp.RF_IPv4_UC, bgp.RF_IPv4_VPN: + if path.GetRouteFamily() == bgp.RF_IPv4_UC { + prefix = path.GetNlri().(*bgp.IPAddrPrefix).IPAddrPrefixDefault.Prefix.To4() + } else { + prefix = path.GetNlri().(*bgp.LabeledVPNIPAddrPrefix).IPAddrPrefixDefault.Prefix.To4() + } + for _, p := range paths { + nexthops = append(nexthops, p.GetNexthop().To4()) + } + case bgp.RF_IPv6_UC, bgp.RF_IPv6_VPN: + if path.GetRouteFamily() == bgp.RF_IPv6_UC { + prefix = path.GetNlri().(*bgp.IPv6AddrPrefix).IPAddrPrefixDefault.Prefix.To16() + } else { + prefix = path.GetNlri().(*bgp.LabeledVPNIPv6AddrPrefix).IPAddrPrefixDefault.Prefix.To16() + } + for _, p := range paths { + nexthops = append(nexthops, p.GetNexthop().To16()) + } + default: + return nil, false + } + msgFlags := zebra.MESSAGE_NEXTHOP + plen, _ := strconv.ParseUint(l[1], 10, 8) + med, err := path.GetMed() + if err == nil { + msgFlags |= zebra.MESSAGE_METRIC + } + var flags zebra.FLAG + info := path.GetSource() + if info.AS == info.LocalAS { + flags = zebra.FLAG_IBGP | zebra.FLAG_INTERNAL + } else if info.MultihopTtl > 0 { + flags = zebra.FLAG_INTERNAL + } + return &zebra.IPRouteBody{ + Type: zebra.ROUTE_BGP, + Flags: flags, + SAFI: zebra.SAFI_UNICAST, + Message: msgFlags, + Prefix: prefix, + PrefixLength: uint8(plen), + Nexthops: nexthops, + Metric: med, + }, path.IsWithdraw +} + +func newNexthopRegisterBody(paths []*table.Path, nexthopCache nexthopStateCache) *zebra.NexthopRegisterBody { + paths = nexthopCache.filterPathToRegister(paths) + if len(paths) == 0 { + return nil + } + path := paths[0] + + family := path.GetRouteFamily() + nexthops := make([]*zebra.RegisteredNexthop, 0, len(paths)) + for _, p := range paths { + nexthop := p.GetNexthop() + var nh *zebra.RegisteredNexthop + switch family { + case bgp.RF_IPv4_UC, bgp.RF_IPv4_VPN: + nh = &zebra.RegisteredNexthop{ + Family: syscall.AF_INET, + Prefix: nexthop.To4(), + } + case bgp.RF_IPv6_UC, bgp.RF_IPv6_VPN: + nh = &zebra.RegisteredNexthop{ + Family: syscall.AF_INET6, + Prefix: nexthop.To16(), + } + default: + continue + } + nexthops = append(nexthops, nh) + } + + // If no nexthop needs to be registered or unregistered, skips to send + // message. + if len(nexthops) == 0 { + return nil + } + + return &zebra.NexthopRegisterBody{ + Nexthops: nexthops, + } +} + +func newNexthopUnregisterBody(family uint16, prefix net.IP) *zebra.NexthopRegisterBody { + return &zebra.NexthopRegisterBody{ + Nexthops: []*zebra.RegisteredNexthop{{ + Family: family, + Prefix: prefix, + }}, + } +} + +func newPathFromIPRouteMessage(m *zebra.Message) *table.Path { + header := m.Header + body := m.Body.(*zebra.IPRouteBody) + family := body.RouteFamily() + isWithdraw := body.IsWithdraw() + + var nlri bgp.AddrPrefixInterface + pattr := make([]bgp.PathAttributeInterface, 0) + origin := bgp.NewPathAttributeOrigin(bgp.BGP_ORIGIN_ATTR_TYPE_IGP) + pattr = append(pattr, origin) + + log.WithFields(log.Fields{ + "Topic": "Zebra", + "RouteType": body.Type.String(), + "Flag": body.Flags.String(), + "Message": body.Message, + "Prefix": body.Prefix, + "PrefixLength": body.PrefixLength, + "Nexthop": body.Nexthops, + "IfIndex": body.Ifindexs, + "Metric": body.Metric, + "Distance": body.Distance, + "Mtu": body.Mtu, + "api": header.Command.String(), + }).Debugf("create path from ip route message.") + + switch family { + case bgp.RF_IPv4_UC: + nlri = bgp.NewIPAddrPrefix(body.PrefixLength, body.Prefix.String()) + if len(body.Nexthops) > 0 { + pattr = append(pattr, bgp.NewPathAttributeNextHop(body.Nexthops[0].String())) + } + case bgp.RF_IPv6_UC: + nlri = bgp.NewIPv6AddrPrefix(body.PrefixLength, body.Prefix.String()) + nexthop := "" + if len(body.Nexthops) > 0 { + nexthop = body.Nexthops[0].String() + } + pattr = append(pattr, bgp.NewPathAttributeMpReachNLRI(nexthop, []bgp.AddrPrefixInterface{nlri})) + default: + log.WithFields(log.Fields{ + "Topic": "Zebra", + }).Errorf("unsupport address family: %s", family) + return nil + } + + med := bgp.NewPathAttributeMultiExitDisc(body.Metric) + pattr = append(pattr, med) + + path := table.NewPath(nil, nlri, isWithdraw, pattr, time.Now(), false) + path.SetIsFromExternal(true) + return path +} + +type zebraClient struct { + client *zebra.Client + server *BgpServer + nexthopCache nexthopStateCache + dead chan struct{} +} + +func (z *zebraClient) getPathListWithNexthopUpdate(body *zebra.NexthopUpdateBody) []*table.Path { + rib := &table.TableManager{ + Tables: make(map[bgp.RouteFamily]*table.Table), + } + + var rfList []bgp.RouteFamily + switch body.Family { + case uint16(syscall.AF_INET): + rfList = []bgp.RouteFamily{bgp.RF_IPv4_UC, bgp.RF_IPv4_VPN} + case uint16(syscall.AF_INET6): + rfList = []bgp.RouteFamily{bgp.RF_IPv6_UC, bgp.RF_IPv6_VPN} + } + + for _, rf := range rfList { + tbl, _, err := z.server.GetRib("", rf, nil) + if err != nil { + log.WithFields(log.Fields{ + "Topic": "Zebra", + "Family": rf.String(), + "Error": err, + }).Error("failed to get global rib") + continue + } + rib.Tables[rf] = tbl + } + + return rib.GetPathListWithNexthop(table.GLOBAL_RIB_NAME, rfList, body.Prefix) +} + +func (z *zebraClient) updatePathByNexthopCache(paths []*table.Path) { + paths = z.nexthopCache.applyToPathList(paths) + if len(paths) > 0 { + if err := z.server.UpdatePath("", paths); err != nil { + log.WithFields(log.Fields{ + "Topic": "Zebra", + "PathList": paths, + }).Error("failed to update nexthop reachability") + } + } +} + +func (z *zebraClient) loop() { + w := z.server.Watch([]WatchOption{ + WatchBestPath(true), + WatchPostUpdate(true), + }...) + defer w.Stop() + + for { + select { + case <-z.dead: + return + case msg := <-z.client.Receive(): + switch body := msg.Body.(type) { + case *zebra.IPRouteBody: + if path := newPathFromIPRouteMessage(msg); path != nil { + if _, err := z.server.AddPath("", []*table.Path{path}); err != nil { + log.WithFields(log.Fields{ + "Topic": "Zebra", + "Path": path, + "Error": err, + }).Error("failed to add path from zebra") + } + } + case *zebra.NexthopUpdateBody: + if updated := z.nexthopCache.updateByNexthopUpdate(body); !updated { + continue + } + paths := z.getPathListWithNexthopUpdate(body) + if len(paths) == 0 { + // If there is no path bound for the given nexthop, send + // NEXTHOP_UNREGISTER message. + z.client.SendNexthopRegister(msg.Header.VrfId, newNexthopUnregisterBody(body.Family, body.Prefix), true) + delete(z.nexthopCache, body.Prefix.String()) + } + z.updatePathByNexthopCache(paths) + } + case ev := <-w.Event(): + switch msg := ev.(type) { + case *WatchEventBestPath: + if table.UseMultiplePaths.Enabled { + for _, paths := range msg.MultiPathList { + z.updatePathByNexthopCache(paths) + if body, isWithdraw := newIPRouteBody(paths); body != nil { + z.client.SendIPRoute(0, body, isWithdraw) + } + if body := newNexthopRegisterBody(paths, z.nexthopCache); body != nil { + z.client.SendNexthopRegister(0, body, false) + } + } + } else { + z.updatePathByNexthopCache(msg.PathList) + for _, path := range msg.PathList { + vrfs := []uint16{0} + if msg.Vrf != nil { + if v, ok := msg.Vrf[path.GetNlri().String()]; ok { + vrfs = append(vrfs, v) + } + } + for _, i := range vrfs { + if body, isWithdraw := newIPRouteBody([]*table.Path{path}); body != nil { + z.client.SendIPRoute(i, body, isWithdraw) + } + if body := newNexthopRegisterBody([]*table.Path{path}, z.nexthopCache); body != nil { + z.client.SendNexthopRegister(i, body, false) + } + } + } + } + case *WatchEventUpdate: + if body := newNexthopRegisterBody(msg.PathList, z.nexthopCache); body != nil { + vrfID := uint16(0) + for _, vrf := range z.server.GetVrf() { + if vrf.Name == msg.Neighbor.Config.Vrf { + vrfID = uint16(vrf.Id) + } + } + z.client.SendNexthopRegister(vrfID, body, false) + } + } + } + } +} + +func newZebraClient(s *BgpServer, url string, protos []string, version uint8, nhtEnable bool, nhtDelay uint8) (*zebraClient, error) { + l := strings.SplitN(url, ":", 2) + if len(l) != 2 { + return nil, fmt.Errorf("unsupported url: %s", url) + } + var cli *zebra.Client + var err error + for _, ver := range []uint8{version, 2, 3, 4} { + cli, err = zebra.NewClient(l[0], l[1], zebra.ROUTE_BGP, ver) + if err == nil { + break + } + // Retry with another Zebra message version + log.WithFields(log.Fields{ + "Topic": "Zebra", + }).Warnf("cannot connect to Zebra with message version %d. going to retry another version...", ver) + } + if cli == nil { + return nil, err + } + // Note: HELLO/ROUTER_ID_ADD messages are automatically sent to negotiate + // the Zebra message version in zebra.NewClient(). + // cli.SendHello() + // cli.SendRouterIDAdd() + cli.SendInterfaceAdd() + for _, typ := range protos { + t, err := zebra.RouteTypeFromString(typ) + if err != nil { + return nil, err + } + cli.SendRedistribute(t, zebra.VRF_DEFAULT) + } + w := &zebraClient{ + client: cli, + server: s, + nexthopCache: make(nexthopStateCache), + dead: make(chan struct{}), + } + go w.loop() + return w, nil +} diff --git a/pkg/server/zclient_test.go b/pkg/server/zclient_test.go new file mode 100644 index 00000000..c7868c0b --- /dev/null +++ b/pkg/server/zclient_test.go @@ -0,0 +1,113 @@ +// Copyright (C) 2014 Nippon Telegraph and Telephone Corporation. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or +// implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package server + +import ( + "net" + "testing" + "time" + + "github.com/stretchr/testify/assert" + + "github.com/osrg/gobgp/internal/pkg/table" + "github.com/osrg/gobgp/internal/pkg/zebra" +) + +func Test_newPathFromIPRouteMessage(t *testing.T) { + assert := assert.New(t) + + // IPv4 Route Add + m := &zebra.Message{} + h := &zebra.Header{ + Len: zebra.HeaderSize(2), + Marker: zebra.HEADER_MARKER, + Version: 2, + Command: zebra.IPV4_ROUTE_ADD, + } + b := &zebra.IPRouteBody{ + Type: zebra.ROUTE_TYPE(zebra.ROUTE_STATIC), + Flags: zebra.FLAG(zebra.FLAG_SELECTED), + Message: zebra.MESSAGE_NEXTHOP | zebra.MESSAGE_DISTANCE | zebra.MESSAGE_METRIC | zebra.MESSAGE_MTU, + SAFI: zebra.SAFI(zebra.SAFI_UNICAST), + Prefix: net.ParseIP("192.168.100.0"), + PrefixLength: uint8(24), + Nexthops: []net.IP{net.ParseIP("0.0.0.0")}, + Ifindexs: []uint32{1}, + Distance: uint8(0), + Metric: uint32(100), + Mtu: uint32(0), + Api: zebra.API_TYPE(zebra.IPV4_ROUTE_ADD), + } + m.Header = *h + m.Body = b + + path := newPathFromIPRouteMessage(m) + pp := table.NewPath(nil, path.GetNlri(), path.IsWithdraw, path.GetPathAttrs(), time.Now(), false) + pp.SetIsFromExternal(path.IsFromExternal()) + assert.Equal("0.0.0.0", pp.GetNexthop().String()) + assert.Equal("192.168.100.0/24", pp.GetNlri().String()) + assert.True(pp.IsFromExternal()) + assert.False(pp.IsWithdraw) + + // IPv4 Route Delete + h.Command = zebra.IPV4_ROUTE_DELETE + b.Api = zebra.IPV4_ROUTE_DELETE + m.Header = *h + m.Body = b + + path = newPathFromIPRouteMessage(m) + pp = table.NewPath(nil, path.GetNlri(), path.IsWithdraw, path.GetPathAttrs(), time.Now(), false) + pp.SetIsFromExternal(path.IsFromExternal()) + assert.Equal("0.0.0.0", pp.GetNexthop().String()) + assert.Equal("192.168.100.0/24", pp.GetNlri().String()) + med, _ := pp.GetMed() + assert.Equal(uint32(100), med) + assert.True(pp.IsFromExternal()) + assert.True(pp.IsWithdraw) + + // IPv6 Route Add + h.Command = zebra.IPV6_ROUTE_ADD + b.Api = zebra.IPV6_ROUTE_ADD + b.Prefix = net.ParseIP("2001:db8:0:f101::") + b.PrefixLength = uint8(64) + b.Nexthops = []net.IP{net.ParseIP("::")} + m.Header = *h + m.Body = b + + path = newPathFromIPRouteMessage(m) + pp = table.NewPath(nil, path.GetNlri(), path.IsWithdraw, path.GetPathAttrs(), time.Now(), false) + pp.SetIsFromExternal(path.IsFromExternal()) + assert.Equal("::", pp.GetNexthop().String()) + assert.Equal("2001:db8:0:f101::/64", pp.GetNlri().String()) + med, _ = pp.GetMed() + assert.Equal(uint32(100), med) + assert.True(pp.IsFromExternal()) + assert.False(pp.IsWithdraw) + + // IPv6 Route Delete + h.Command = zebra.IPV6_ROUTE_DELETE + b.Api = zebra.IPV6_ROUTE_DELETE + m.Header = *h + m.Body = b + + path = newPathFromIPRouteMessage(m) + pp = table.NewPath(nil, path.GetNlri(), path.IsWithdraw, path.GetPathAttrs(), time.Now(), false) + pp.SetIsFromExternal(path.IsFromExternal()) + assert.Equal("::", pp.GetNexthop().String()) + assert.Equal("2001:db8:0:f101::/64", pp.GetNlri().String()) + assert.True(pp.IsFromExternal()) + assert.True(pp.IsWithdraw) +} |