Merge branch 'main' into fix/pkg-loss

This commit is contained in:
Zoltán Papp
2025-09-03 16:27:17 +02:00
872 changed files with 75624 additions and 22190 deletions
+285 -264
View File
@@ -5,61 +5,61 @@ import (
"fmt"
"math/rand"
"net"
"net/netip"
"os"
"runtime"
"strings"
"sync"
"time"
"github.com/pion/ice/v3"
"github.com/pion/ice/v4"
log "github.com/sirupsen/logrus"
"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
"github.com/netbirdio/netbird/client/iface/configurer"
"github.com/netbirdio/netbird/client/iface/wgproxy"
"github.com/netbirdio/netbird/client/internal/peer/conntype"
"github.com/netbirdio/netbird/client/internal/peer/dispatcher"
"github.com/netbirdio/netbird/client/internal/peer/guard"
icemaker "github.com/netbirdio/netbird/client/internal/peer/ice"
"github.com/netbirdio/netbird/client/internal/peer/id"
"github.com/netbirdio/netbird/client/internal/peer/worker"
"github.com/netbirdio/netbird/client/internal/stdnet"
relayClient "github.com/netbirdio/netbird/relay/client"
"github.com/netbirdio/netbird/route"
nbnet "github.com/netbirdio/netbird/util/net"
relayClient "github.com/netbirdio/netbird/shared/relay/client"
semaphoregroup "github.com/netbirdio/netbird/util/semaphore-group"
)
type ConnPriority int
func (cp ConnPriority) String() string {
switch cp {
case connPriorityNone:
return "None"
case connPriorityRelay:
return "PriorityRelay"
case connPriorityICETurn:
return "PriorityICETurn"
case connPriorityICEP2P:
return "PriorityICEP2P"
default:
return fmt.Sprintf("ConnPriority(%d)", cp)
}
}
const (
defaultWgKeepAlive = 25 * time.Second
connPriorityNone ConnPriority = 0
connPriorityRelay ConnPriority = 1
connPriorityICETurn ConnPriority = 2
connPriorityICEP2P ConnPriority = 3
)
type ServiceDependencies struct {
StatusRecorder *Status
Signaler *Signaler
IFaceDiscover stdnet.ExternalIFaceDiscover
RelayManager *relayClient.Manager
SrWatcher *guard.SRWatcher
Semaphore *semaphoregroup.SemaphoreGroup
PeerConnDispatcher *dispatcher.ConnectionDispatcher
}
type WgConfig struct {
WgListenPort int
RemoteKey string
WgInterface WGIface
AllowedIps string
AllowedIps []netip.Prefix
PreSharedKey *wgtypes.Key
}
type RosenpassConfig struct {
// RosenpassPubKey is this peer's Rosenpass public key
PubKey []byte
// RosenpassPubKey is this peer's RosenpassAddr server address (IP:port)
Addr string
PermissiveMode bool
}
// ConnConfig is a peer Connection configuration
type ConnConfig struct {
// Key is a public key of a remote peer
@@ -67,157 +67,169 @@ type ConnConfig struct {
// LocalKey is a public key of a local peer
LocalKey string
AgentVersion string
Timeout time.Duration
WgConfig WgConfig
LocalWgPort int
// RosenpassPubKey is this peer's Rosenpass public key
RosenpassPubKey []byte
// RosenpassPubKey is this peer's RosenpassAddr server address (IP:port)
RosenpassAddr string
RosenpassConfig RosenpassConfig
// ICEConfig ICE protocol configuration
ICEConfig icemaker.Config
}
type Conn struct {
log *log.Entry
Log *log.Entry
mu sync.Mutex
ctx context.Context
ctxCancel context.CancelFunc
config ConnConfig
statusRecorder *Status
signaler *Signaler
iFaceDiscover stdnet.ExternalIFaceDiscover
relayManager *relayClient.Manager
allowedIP net.IP
handshaker *Handshaker
srWatcher *guard.SRWatcher
onConnected func(remoteWireGuardKey string, remoteRosenpassPubKey []byte, wireGuardIP string, remoteRosenpassAddr string)
onDisconnected func(remotePeer string, wgIP string)
onDisconnected func(remotePeer string)
statusRelay *AtomicConnStatus
statusICE *AtomicConnStatus
currentConnPriority ConnPriority
statusRelay *worker.AtomicWorkerStatus
statusICE *worker.AtomicWorkerStatus
currentConnPriority conntype.ConnPriority
opened bool // this flag is used to prevent close in case of not opened connection
workerICE *WorkerICE
workerRelay *WorkerRelay
wgWatcherWg sync.WaitGroup
connIDRelay nbnet.ConnectionID
connIDICE nbnet.ConnectionID
beforeAddPeerHooks []nbnet.AddHookFunc
afterRemovePeerHooks []nbnet.RemoveHookFunc
// used to store the remote Rosenpass key for Relayed connection in case of connection update from ice
rosenpassRemoteKey []byte
wgProxyICE wgproxy.Proxy
wgProxyRelay wgproxy.Proxy
handshaker *Handshaker
guard *guard.Guard
semaphore *semaphoregroup.SemaphoreGroup
wg sync.WaitGroup
// debug purpose
dumpState *stateDump
}
// NewConn creates a new not opened Conn to the remote peer.
// To establish a connection run Conn.Open
func NewConn(engineCtx context.Context, config ConnConfig, statusRecorder *Status, signaler *Signaler, iFaceDiscover stdnet.ExternalIFaceDiscover, relayManager *relayClient.Manager, srWatcher *guard.SRWatcher, semaphore *semaphoregroup.SemaphoreGroup) (*Conn, error) {
allowedIP, _, err := net.ParseCIDR(config.WgConfig.AllowedIps)
if err != nil {
log.Errorf("failed to parse allowedIPS: %v", err)
return nil, err
func NewConn(config ConnConfig, services ServiceDependencies) (*Conn, error) {
if len(config.WgConfig.AllowedIps) == 0 {
return nil, fmt.Errorf("allowed IPs is empty")
}
ctx, ctxCancel := context.WithCancel(engineCtx)
connLog := log.WithField("peer", config.Key)
var conn = &Conn{
log: connLog,
ctx: ctx,
ctxCancel: ctxCancel,
Log: connLog,
config: config,
statusRecorder: statusRecorder,
signaler: signaler,
relayManager: relayManager,
allowedIP: allowedIP,
statusRelay: NewAtomicConnStatus(),
statusICE: NewAtomicConnStatus(),
semaphore: semaphore,
statusRecorder: services.StatusRecorder,
signaler: services.Signaler,
iFaceDiscover: services.IFaceDiscover,
relayManager: services.RelayManager,
srWatcher: services.SrWatcher,
semaphore: services.Semaphore,
statusRelay: worker.NewAtomicStatus(),
statusICE: worker.NewAtomicStatus(),
dumpState: newStateDump(config.Key, connLog, services.StatusRecorder),
}
ctrl := isController(config)
conn.workerRelay = NewWorkerRelay(connLog, ctrl, config, conn, relayManager)
relayIsSupportedLocally := conn.workerRelay.RelayIsSupportedLocally()
conn.workerICE, err = NewWorkerICE(ctx, connLog, config, conn, signaler, iFaceDiscover, statusRecorder, relayIsSupportedLocally)
if err != nil {
return nil, err
}
conn.handshaker = NewHandshaker(ctx, connLog, config, signaler, conn.workerICE, conn.workerRelay)
conn.handshaker.AddOnNewOfferListener(conn.workerRelay.OnNewOffer)
if os.Getenv("NB_FORCE_RELAY") != "true" {
conn.handshaker.AddOnNewOfferListener(conn.workerICE.OnNewOffer)
}
conn.guard = guard.NewGuard(connLog, ctrl, conn.isConnectedOnAllWay, config.Timeout, srWatcher)
go conn.handshaker.Listen()
return conn, nil
}
// Open opens connection to the remote peer
// It will try to establish a connection using ICE and in parallel with relay. The higher priority connection type will
// be used.
func (conn *Conn) Open() {
conn.semaphore.Add(conn.ctx)
conn.log.Debugf("open connection to peer")
func (conn *Conn) Open(engineCtx context.Context) error {
conn.semaphore.Add(engineCtx)
conn.mu.Lock()
defer conn.mu.Unlock()
conn.opened = true
if conn.opened {
conn.semaphore.Done(engineCtx)
return nil
}
conn.ctx, conn.ctxCancel = context.WithCancel(engineCtx)
conn.workerRelay = NewWorkerRelay(conn.ctx, conn.Log, isController(conn.config), conn.config, conn, conn.relayManager, conn.dumpState)
relayIsSupportedLocally := conn.workerRelay.RelayIsSupportedLocally()
workerICE, err := NewWorkerICE(conn.ctx, conn.Log, conn.config, conn, conn.signaler, conn.iFaceDiscover, conn.statusRecorder, relayIsSupportedLocally)
if err != nil {
return err
}
conn.workerICE = workerICE
conn.handshaker = NewHandshaker(conn.Log, conn.config, conn.signaler, conn.workerICE, conn.workerRelay)
conn.handshaker.AddOnNewOfferListener(conn.workerRelay.OnNewOffer)
if os.Getenv("NB_FORCE_RELAY") != "true" {
conn.handshaker.AddOnNewOfferListener(conn.workerICE.OnNewOffer)
}
conn.guard = guard.NewGuard(conn.Log, conn.isConnectedOnAllWay, conn.config.Timeout, conn.srWatcher)
conn.wg.Add(1)
go func() {
defer conn.wg.Done()
conn.handshaker.Listen(conn.ctx)
}()
go conn.dumpState.Start(conn.ctx)
peerState := State{
PubKey: conn.config.Key,
IP: strings.Split(conn.config.WgConfig.AllowedIps, "/")[0],
ConnStatusUpdate: time.Now(),
ConnStatus: StatusDisconnected,
ConnStatus: StatusConnecting,
Mux: new(sync.RWMutex),
}
err := conn.statusRecorder.UpdatePeerState(peerState)
if err != nil {
conn.log.Warnf("error while updating the state err: %v", err)
if err := conn.statusRecorder.UpdatePeerState(peerState); err != nil {
conn.Log.Warnf("error while updating the state err: %v", err)
}
go conn.startHandshakeAndReconnect(conn.ctx)
}
conn.wg.Add(1)
go func() {
defer conn.wg.Done()
func (conn *Conn) startHandshakeAndReconnect(ctx context.Context) {
defer conn.semaphore.Done(conn.ctx)
conn.waitInitialRandomSleepTime(ctx)
conn.waitInitialRandomSleepTime(conn.ctx)
conn.semaphore.Done(conn.ctx)
err := conn.handshaker.sendOffer()
if err != nil {
conn.log.Errorf("failed to send initial offer: %v", err)
}
go conn.guard.Start(ctx)
go conn.listenGuardEvent(ctx)
conn.guard.Start(conn.ctx, conn.onGuardEvent)
}()
conn.opened = true
return nil
}
// Close closes this peer Conn issuing a close event to the Conn closeCh
func (conn *Conn) Close() {
func (conn *Conn) Close(signalToRemote bool) {
conn.mu.Lock()
defer conn.wgWatcherWg.Wait()
defer conn.mu.Unlock()
conn.log.Infof("close peer connection")
conn.ctxCancel()
if !conn.opened {
conn.log.Debugf("ignore close connection to peer")
conn.Log.Debugf("ignore close connection to peer")
return
}
if signalToRemote {
if err := conn.signaler.SignalIdle(conn.config.Key); err != nil {
conn.Log.Errorf("failed to signal idle state to peer: %v", err)
}
}
conn.Log.Infof("close peer connection")
conn.ctxCancel()
conn.workerRelay.DisableWgWatcher()
conn.workerRelay.CloseConn()
conn.workerICE.Close()
@@ -225,7 +237,7 @@ func (conn *Conn) Close() {
if conn.wgProxyRelay != nil {
err := conn.wgProxyRelay.CloseConn()
if err != nil {
conn.log.Errorf("failed to close wg proxy for relay: %v", err)
conn.Log.Errorf("failed to close wg proxy for relay: %v", err)
}
conn.wgProxyRelay = nil
}
@@ -233,56 +245,53 @@ func (conn *Conn) Close() {
if conn.wgProxyICE != nil {
err := conn.wgProxyICE.CloseConn()
if err != nil {
conn.log.Errorf("failed to close wg proxy for ice: %v", err)
conn.Log.Errorf("failed to close wg proxy for ice: %v", err)
}
conn.wgProxyICE = nil
}
if err := conn.removeWgPeer(); err != nil {
conn.log.Errorf("failed to remove wg endpoint: %v", err)
conn.Log.Errorf("failed to remove wg endpoint: %v", err)
}
conn.freeUpConnID()
if conn.evalStatus() == StatusConnected && conn.onDisconnected != nil {
conn.onDisconnected(conn.config.WgConfig.RemoteKey, conn.config.WgConfig.AllowedIps)
conn.onDisconnected(conn.config.WgConfig.RemoteKey)
}
conn.setStatusToDisconnected()
conn.opened = false
conn.wg.Wait()
conn.Log.Infof("peer connection closed")
}
// OnRemoteAnswer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
// doesn't block, discards the message if connection wasn't ready
func (conn *Conn) OnRemoteAnswer(answer OfferAnswer) bool {
conn.log.Debugf("OnRemoteAnswer, status ICE: %s, status relay: %s", conn.statusICE, conn.statusRelay)
return conn.handshaker.OnRemoteAnswer(answer)
func (conn *Conn) OnRemoteAnswer(answer OfferAnswer) {
conn.dumpState.RemoteAnswer()
conn.Log.Infof("OnRemoteAnswer, priority: %s, status ICE: %s, status relay: %s", conn.currentConnPriority, conn.statusICE, conn.statusRelay)
conn.handshaker.OnRemoteAnswer(answer)
}
// OnRemoteCandidate Handles ICE connection Candidate provided by the remote peer.
func (conn *Conn) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMap) {
conn.dumpState.RemoteCandidate()
conn.workerICE.OnRemoteCandidate(candidate, haRoutes)
}
func (conn *Conn) AddBeforeAddPeerHook(hook nbnet.AddHookFunc) {
conn.beforeAddPeerHooks = append(conn.beforeAddPeerHooks, hook)
}
func (conn *Conn) AddAfterRemovePeerHook(hook nbnet.RemoveHookFunc) {
conn.afterRemovePeerHooks = append(conn.afterRemovePeerHooks, hook)
}
// SetOnConnected sets a handler function to be triggered by Conn when a new connection to a remote peer established
func (conn *Conn) SetOnConnected(handler func(remoteWireGuardKey string, remoteRosenpassPubKey []byte, wireGuardIP string, remoteRosenpassAddr string)) {
conn.onConnected = handler
}
// SetOnDisconnected sets a handler function to be triggered by Conn when a connection to a remote disconnected
func (conn *Conn) SetOnDisconnected(handler func(remotePeer string, wgIP string)) {
func (conn *Conn) SetOnDisconnected(handler func(remotePeer string)) {
conn.onDisconnected = handler
}
func (conn *Conn) OnRemoteOffer(offer OfferAnswer) bool {
conn.log.Debugf("OnRemoteOffer, on status ICE: %s, status Relay: %s", conn.statusICE, conn.statusRelay)
return conn.handshaker.OnRemoteOffer(offer)
func (conn *Conn) OnRemoteOffer(offer OfferAnswer) {
conn.dumpState.RemoteOffer()
conn.Log.Infof("OnRemoteOffer, on status ICE: %s, status Relay: %s", conn.statusICE, conn.statusRelay)
conn.handshaker.OnRemoteOffer(offer)
}
// WgConfig returns the WireGuard config
@@ -290,19 +299,24 @@ func (conn *Conn) WgConfig() WgConfig {
return conn.config.WgConfig
}
// Status returns current status of the Conn
func (conn *Conn) Status() ConnStatus {
// IsConnected returns true if the peer is connected
func (conn *Conn) IsConnected() bool {
conn.mu.Lock()
defer conn.mu.Unlock()
return conn.evalStatus()
return conn.evalStatus() == StatusConnected
}
func (conn *Conn) GetKey() string {
return conn.config.Key
}
func (conn *Conn) ConnID() id.ConnID {
return id.ConnID(conn)
}
// configureConnection starts proxying traffic from/to local Wireguard and sets connection status to StatusConnected
func (conn *Conn) onICEConnectionIsReady(priority ConnPriority, iceConnInfo ICEConnInfo) {
func (conn *Conn) onICEConnectionIsReady(priority conntype.ConnPriority, iceConnInfo ICEConnInfo) {
conn.mu.Lock()
defer conn.mu.Unlock()
@@ -310,21 +324,22 @@ func (conn *Conn) onICEConnectionIsReady(priority ConnPriority, iceConnInfo ICEC
return
}
if remoteConnNil(conn.log, iceConnInfo.RemoteConn) {
conn.log.Errorf("remote ICE connection is nil")
if remoteConnNil(conn.Log, iceConnInfo.RemoteConn) {
conn.Log.Errorf("remote ICE connection is nil")
return
}
// this never should happen, because Relay is the lower priority and ICE always close the deprecated connection before upgrade
// todo consider to remove this check
if conn.currentConnPriority > priority {
conn.log.Infof("current connection priority (%s) is higher than the new one (%s), do not upgrade connection", conn.currentConnPriority, priority)
conn.statusICE.Set(StatusConnected)
conn.Log.Infof("current connection priority (%s) is higher than the new one (%s), do not upgrade connection", conn.currentConnPriority, priority)
conn.statusICE.SetConnected()
conn.updateIceState(iceConnInfo)
return
}
conn.log.Infof("set ICE to active connection")
conn.Log.Infof("set ICE to active connection")
conn.dumpState.P2PConnected()
var (
ep *net.UDPAddr
@@ -332,9 +347,10 @@ func (conn *Conn) onICEConnectionIsReady(priority ConnPriority, iceConnInfo ICEC
err error
)
if iceConnInfo.RelayedOnLocal {
conn.dumpState.NewLocalProxy()
wgProxy, err = conn.newProxy(iceConnInfo.RemoteConn)
if err != nil {
conn.log.Errorf("failed to add turn net.Conn to local proxy: %v", err)
conn.Log.Errorf("failed to add turn net.Conn to local proxy: %v", err)
return
}
ep = wgProxy.EndpointAddr()
@@ -349,36 +365,32 @@ func (conn *Conn) onICEConnectionIsReady(priority ConnPriority, iceConnInfo ICEC
ep = directEp
}
if err := conn.runBeforeAddPeerHooks(ep.IP); err != nil {
conn.log.Errorf("Before add peer hook failed: %v", err)
}
conn.workerRelay.DisableWgWatcher()
// todo consider to run conn.wgWatcherWg.Wait() here
if conn.wgProxyRelay != nil {
conn.log.Debugf("pause Relayed proxy")
conn.wgProxyRelay.Pause()
}
if wgProxy != nil {
conn.log.Debugf("run ICE proxy")
wgProxy.Work()
}
conn.log.Infof("configure WireGuard endpoint to: %s", ep.String())
if err = conn.configureWGEndpoint(ep); err != nil {
conn.Log.Infof("configure WireGuard endpoint to: %s", ep.String())
if err = conn.configureWGEndpoint(ep, iceConnInfo.RosenpassPubKey); err != nil {
conn.handleConfigurationFailure(err, wgProxy)
return
}
wgConfigWorkaround()
if conn.wgProxyRelay != nil {
conn.log.Debugf("redirect packages from relayed conn to WireGuard")
conn.Log.Debugf("redirect packages from relayed conn to WireGuard")
conn.wgProxyRelay.RedirectAs(ep)
}
conn.currentConnPriority = priority
conn.statusICE.Set(StatusConnected)
conn.statusICE.SetConnected()
conn.updateIceState(iceConnInfo)
conn.doOnConnected(iceConnInfo.RosenpassPubKey, iceConnInfo.RosenpassAddr)
}
@@ -391,35 +403,41 @@ func (conn *Conn) onICEStateDisconnected() {
return
}
conn.log.Tracef("ICE connection state changed to disconnected")
conn.Log.Tracef("ICE connection state changed to disconnected")
if conn.wgProxyICE != nil {
if err := conn.wgProxyICE.CloseConn(); err != nil {
conn.log.Warnf("failed to close deprecated wg proxy conn: %v", err)
conn.Log.Warnf("failed to close deprecated wg proxy conn: %v", err)
}
}
// switch back to relay connection
if conn.isReadyToUpgrade() {
conn.log.Infof("ICE disconnected, set Relay to active connection")
conn.Log.Infof("ICE disconnected, set Relay to active connection")
conn.dumpState.SwitchToRelay()
conn.wgProxyRelay.Work()
if err := conn.configureWGEndpoint(conn.wgProxyRelay.EndpointAddr()); err != nil {
conn.log.Errorf("failed to switch to relay conn: %v", err)
if err := conn.configureWGEndpoint(conn.wgProxyRelay.EndpointAddr(), conn.rosenpassRemoteKey); err != nil {
conn.Log.Errorf("failed to switch to relay conn: %v", err)
}
conn.wgWatcherWg.Add(1)
go func() {
defer conn.wgWatcherWg.Done()
conn.workerRelay.EnableWgWatcher(conn.ctx)
}()
conn.wgProxyRelay.Work()
conn.workerRelay.EnableWgWatcher(conn.ctx)
conn.currentConnPriority = connPriorityRelay
conn.currentConnPriority = conntype.Relay
} else {
conn.log.Infof("ICE disconnected, do not switch to Relay. Reset priority to: %s", connPriorityNone.String())
conn.currentConnPriority = connPriorityNone
conn.Log.Infof("ICE disconnected, do not switch to Relay. Reset priority to: %s", conntype.None.String())
conn.currentConnPriority = conntype.None
}
changed := conn.statusICE.Get() != StatusDisconnected
changed := conn.statusICE.Get() != worker.StatusDisconnected
if changed {
conn.guard.SetICEConnDisconnected()
}
conn.statusICE.Set(StatusDisconnected)
conn.statusICE.SetDisconnected()
peerState := State{
PubKey: conn.config.Key,
@@ -430,7 +448,7 @@ func (conn *Conn) onICEStateDisconnected() {
err := conn.statusRecorder.UpdatePeerICEStateToDisconnected(peerState)
if err != nil {
conn.log.Warnf("unable to set peer's state to disconnected ice, got error: %v", err)
conn.Log.Warnf("unable to set peer's state to disconnected ice, got error: %v", err)
}
}
@@ -440,49 +458,55 @@ func (conn *Conn) onRelayConnectionIsReady(rci RelayConnInfo) {
if conn.ctx.Err() != nil {
if err := rci.relayedConn.Close(); err != nil {
conn.log.Warnf("failed to close unnecessary relayed connection: %v", err)
conn.Log.Warnf("failed to close unnecessary relayed connection: %v", err)
}
return
}
conn.log.Debugf("Relay connection has been established, setup the WireGuard")
conn.dumpState.RelayConnected()
conn.Log.Debugf("Relay connection has been established, setup the WireGuard")
wgProxy, err := conn.newProxy(rci.relayedConn)
if err != nil {
conn.log.Errorf("failed to add relayed net.Conn to local proxy: %v", err)
conn.Log.Errorf("failed to add relayed net.Conn to local proxy: %v", err)
return
}
wgProxy.SetDisconnectListener(conn.onRelayDisconnected)
conn.log.Infof("created new wgProxy for relay connection: %s", wgProxy.EndpointAddr().String())
conn.dumpState.NewLocalProxy()
if conn.iceP2PIsActive() {
conn.log.Debugf("do not switch to relay because current priority is: %s", conn.currentConnPriority.String())
conn.Log.Infof("created new wgProxy for relay connection: %s", wgProxy.EndpointAddr().String())
if conn.isICEActive() {
conn.Log.Debugf("do not switch to relay because current priority is: %s", conn.currentConnPriority.String())
conn.setRelayedProxy(wgProxy)
conn.statusRelay.Set(StatusConnected)
conn.statusRelay.SetConnected()
conn.updateRelayStatus(rci.relayedConn.RemoteAddr().String(), rci.rosenpassPubKey)
return
}
if err := conn.runBeforeAddPeerHooks(wgProxy.EndpointAddr().IP); err != nil {
conn.log.Errorf("Before add peer hook failed: %v", err)
}
wgProxy.Work()
if err := conn.configureWGEndpoint(wgProxy.EndpointAddr()); err != nil {
if err := conn.configureWGEndpoint(wgProxy.EndpointAddr(), rci.rosenpassPubKey); err != nil {
if err := wgProxy.CloseConn(); err != nil {
conn.log.Warnf("Failed to close relay connection: %v", err)
conn.Log.Warnf("Failed to close relay connection: %v", err)
}
conn.log.Errorf("Failed to update WireGuard peer configuration: %v", err)
conn.Log.Errorf("Failed to update WireGuard peer configuration: %v", err)
return
}
conn.workerRelay.EnableWgWatcher(conn.ctx)
conn.wgWatcherWg.Add(1)
go func() {
defer conn.wgWatcherWg.Done()
conn.workerRelay.EnableWgWatcher(conn.ctx)
}()
wgConfigWorkaround()
conn.currentConnPriority = connPriorityRelay
conn.statusRelay.Set(StatusConnected)
conn.rosenpassRemoteKey = rci.rosenpassPubKey
conn.currentConnPriority = conntype.Relay
conn.statusRelay.SetConnected()
conn.setRelayedProxy(wgProxy)
conn.updateRelayStatus(rci.relayedConn.RemoteAddr().String(), rci.rosenpassPubKey)
conn.log.Infof("start to communicate with peer via relay")
conn.Log.Infof("start to communicate with peer via relay")
conn.doOnConnected(rci.rosenpassPubKey, rci.rosenpassAddr)
}
@@ -494,13 +518,11 @@ func (conn *Conn) onRelayDisconnected() {
return
}
conn.log.Debugf("relay connection is disconnected")
conn.Log.Debugf("relay connection is disconnected")
if conn.currentConnPriority == connPriorityRelay {
conn.log.Debugf("clean up WireGuard config")
if err := conn.removeWgPeer(); err != nil {
conn.log.Errorf("failed to remove wg endpoint: %v", err)
}
if conn.currentConnPriority == conntype.Relay {
conn.Log.Debugf("clean up WireGuard config")
conn.currentConnPriority = conntype.None
}
if conn.wgProxyRelay != nil {
@@ -508,11 +530,11 @@ func (conn *Conn) onRelayDisconnected() {
conn.wgProxyRelay = nil
}
changed := conn.statusRelay.Get() != StatusDisconnected
changed := conn.statusRelay.Get() != worker.StatusDisconnected
if changed {
conn.guard.SetRelayedConnDisconnected()
}
conn.statusRelay.Set(StatusDisconnected)
conn.statusRelay.SetDisconnected()
peerState := State{
PubKey: conn.config.Key,
@@ -521,31 +543,25 @@ func (conn *Conn) onRelayDisconnected() {
ConnStatusUpdate: time.Now(),
}
if err := conn.statusRecorder.UpdatePeerRelayedStateToDisconnected(peerState); err != nil {
conn.log.Warnf("unable to save peer's state to Relay disconnected, got error: %v", err)
conn.Log.Warnf("unable to save peer's state to Relay disconnected, got error: %v", err)
}
}
func (conn *Conn) listenGuardEvent(ctx context.Context) {
for {
select {
case <-conn.guard.Reconnect:
conn.log.Debugf("send offer to peer")
if err := conn.handshaker.SendOffer(); err != nil {
conn.log.Errorf("failed to send offer: %v", err)
}
case <-ctx.Done():
return
}
func (conn *Conn) onGuardEvent() {
conn.dumpState.SendOffer()
if err := conn.handshaker.SendOffer(); err != nil {
conn.Log.Errorf("failed to send offer: %v", err)
}
}
func (conn *Conn) configureWGEndpoint(addr *net.UDPAddr) error {
func (conn *Conn) configureWGEndpoint(addr *net.UDPAddr, remoteRPKey []byte) error {
presharedKey := conn.presharedKey(remoteRPKey)
return conn.config.WgConfig.WgInterface.UpdatePeer(
conn.config.WgConfig.RemoteKey,
conn.config.WgConfig.AllowedIps,
defaultWgKeepAlive,
addr,
conn.config.WgConfig.PreSharedKey,
presharedKey,
)
}
@@ -561,7 +577,7 @@ func (conn *Conn) updateRelayStatus(relayServerAddr string, rosenpassPubKey []by
err := conn.statusRecorder.UpdatePeerRelayedState(peerState)
if err != nil {
conn.log.Warnf("unable to save peer's Relay state, got error: %v", err)
conn.Log.Warnf("unable to save peer's Relay state, got error: %v", err)
}
}
@@ -580,17 +596,18 @@ func (conn *Conn) updateIceState(iceConnInfo ICEConnInfo) {
err := conn.statusRecorder.UpdatePeerICEState(peerState)
if err != nil {
conn.log.Warnf("unable to save peer's ICE state, got error: %v", err)
conn.Log.Warnf("unable to save peer's ICE state, got error: %v", err)
}
}
func (conn *Conn) setStatusToDisconnected() {
conn.statusRelay.Set(StatusDisconnected)
conn.statusICE.Set(StatusDisconnected)
conn.statusRelay.SetDisconnected()
conn.statusICE.SetDisconnected()
conn.currentConnPriority = conntype.None
peerState := State{
PubKey: conn.config.Key,
ConnStatus: StatusDisconnected,
ConnStatus: StatusIdle,
ConnStatusUpdate: time.Now(),
Mux: new(sync.RWMutex),
}
@@ -598,10 +615,10 @@ func (conn *Conn) setStatusToDisconnected() {
if err != nil {
// pretty common error because by that time Engine can already remove the peer and status won't be available.
// todo rethink status updates
conn.log.Debugf("error while updating peer's state, err: %v", err)
conn.Log.Debugf("error while updating peer's state, err: %v", err)
}
if err := conn.statusRecorder.UpdateWireGuardPeerState(conn.config.Key, configurer.WGStats{}); err != nil {
conn.log.Debugf("failed to reset wireguard stats for peer: %s", err)
conn.Log.Debugf("failed to reset wireguard stats for peer: %s", err)
}
}
@@ -611,7 +628,7 @@ func (conn *Conn) doOnConnected(remoteRosenpassPubKey []byte, remoteRosenpassAdd
}
if conn.onConnected != nil {
conn.onConnected(conn.config.Key, remoteRosenpassPubKey, conn.allowedIP.String(), remoteRosenpassAddr)
conn.onConnected(conn.config.Key, remoteRosenpassPubKey, conn.config.WgConfig.AllowedIps[0].Addr().String(), remoteRosenpassAddr)
}
}
@@ -629,32 +646,24 @@ func (conn *Conn) waitInitialRandomSleepTime(ctx context.Context) {
}
func (conn *Conn) isRelayed() bool {
if conn.statusRelay.Get() == StatusDisconnected && (conn.statusICE.Get() == StatusDisconnected || conn.statusICE.Get() == StatusConnecting) {
switch conn.currentConnPriority {
case conntype.Relay, conntype.ICETurn:
return true
default:
return false
}
if conn.currentConnPriority == connPriorityICEP2P {
return false
}
return true
}
func (conn *Conn) evalStatus() ConnStatus {
if conn.statusRelay.Get() == StatusConnected || conn.statusICE.Get() == StatusConnected {
if conn.statusRelay.Get() == worker.StatusConnected || conn.statusICE.Get() == worker.StatusConnected {
return StatusConnected
}
if conn.statusRelay.Get() == StatusConnecting || conn.statusICE.Get() == StatusConnecting {
return StatusConnecting
}
return StatusDisconnected
return StatusConnecting
}
func (conn *Conn) isConnectedOnAllWay() (connected bool) {
conn.mu.Lock()
defer conn.mu.Unlock()
// would be better to protect this with a mutex, but it could cause deadlock with Close function
defer func() {
if !connected {
@@ -662,12 +671,12 @@ func (conn *Conn) isConnectedOnAllWay() (connected bool) {
}
}()
if conn.statusICE.Get() == StatusDisconnected {
if conn.statusICE.Get() == worker.StatusDisconnected && !conn.workerICE.InProgress() {
return false
}
if conn.workerRelay.IsRelayConnectionSupportedWithPeer() {
if conn.statusRelay.Get() != StatusConnected {
if conn.statusRelay.Get() == worker.StatusDisconnected {
return false
}
}
@@ -675,57 +684,27 @@ func (conn *Conn) isConnectedOnAllWay() (connected bool) {
return true
}
func (conn *Conn) runBeforeAddPeerHooks(ip net.IP) error {
conn.connIDICE = nbnet.GenerateConnID()
for _, hook := range conn.beforeAddPeerHooks {
if err := hook(conn.connIDICE, ip); err != nil {
return err
}
}
return nil
}
func (conn *Conn) freeUpConnID() {
if conn.connIDRelay != "" {
for _, hook := range conn.afterRemovePeerHooks {
if err := hook(conn.connIDRelay); err != nil {
conn.log.Errorf("After remove peer hook failed: %v", err)
}
}
conn.connIDRelay = ""
}
if conn.connIDICE != "" {
for _, hook := range conn.afterRemovePeerHooks {
if err := hook(conn.connIDICE); err != nil {
conn.log.Errorf("After remove peer hook failed: %v", err)
}
}
conn.connIDICE = ""
}
}
func (conn *Conn) newProxy(remoteConn net.Conn) (wgproxy.Proxy, error) {
conn.log.Debugf("setup proxied WireGuard connection")
conn.Log.Debugf("setup proxied WireGuard connection")
udpAddr := &net.UDPAddr{
IP: conn.allowedIP,
IP: conn.config.WgConfig.AllowedIps[0].Addr().AsSlice(),
Port: conn.config.WgConfig.WgListenPort,
}
wgProxy := conn.config.WgConfig.WgInterface.GetProxy()
if err := wgProxy.AddTurnConn(conn.ctx, udpAddr, remoteConn); err != nil {
conn.log.Errorf("failed to add turn net.Conn to local proxy: %v", err)
conn.Log.Errorf("failed to add turn net.Conn to local proxy: %v", err)
return nil, err
}
return wgProxy, nil
}
func (conn *Conn) isReadyToUpgrade() bool {
return conn.wgProxyRelay != nil && conn.currentConnPriority != connPriorityRelay
return conn.wgProxyRelay != nil && conn.currentConnPriority != conntype.Relay
}
func (conn *Conn) iceP2PIsActive() bool {
return conn.currentConnPriority == connPriorityICEP2P && conn.statusICE.Get() == StatusConnected
func (conn *Conn) isICEActive() bool {
return (conn.currentConnPriority == conntype.ICEP2P || conn.currentConnPriority == conntype.ICETurn) && conn.statusICE.Get() == worker.StatusConnected
}
func (conn *Conn) removeWgPeer() error {
@@ -733,10 +712,10 @@ func (conn *Conn) removeWgPeer() error {
}
func (conn *Conn) handleConfigurationFailure(err error, wgProxy wgproxy.Proxy) {
conn.log.Warnf("Failed to update wg peer configuration: %v", err)
conn.Log.Warnf("Failed to update wg peer configuration: %v", err)
if wgProxy != nil {
if ierr := wgProxy.CloseConn(); ierr != nil {
conn.log.Warnf("Failed to close wg proxy: %v", ierr)
conn.Log.Warnf("Failed to close wg proxy: %v", ierr)
}
}
if conn.wgProxyRelay != nil {
@@ -746,24 +725,66 @@ func (conn *Conn) handleConfigurationFailure(err error, wgProxy wgproxy.Proxy) {
func (conn *Conn) logTraceConnState() {
if conn.workerRelay.IsRelayConnectionSupportedWithPeer() {
conn.log.Tracef("connectivity guard check, relay state: %s, ice state: %s", conn.statusRelay, conn.statusICE)
conn.Log.Tracef("connectivity guard check, relay state: %s, ice state: %s", conn.statusRelay, conn.statusICE)
} else {
conn.log.Tracef("connectivity guard check, ice state: %s", conn.statusICE)
conn.Log.Tracef("connectivity guard check, ice state: %s", conn.statusICE)
}
}
func (conn *Conn) setRelayedProxy(proxy wgproxy.Proxy) {
if conn.wgProxyRelay != nil {
if err := conn.wgProxyRelay.CloseConn(); err != nil {
conn.log.Warnf("failed to close deprecated wg proxy conn: %v", err)
conn.Log.Warnf("failed to close deprecated wg proxy conn: %v", err)
}
}
conn.wgProxyRelay = proxy
}
// AllowedIP returns the allowed IP of the remote peer
func (conn *Conn) AllowedIP() net.IP {
return conn.allowedIP
func (conn *Conn) AllowedIP() netip.Addr {
return conn.config.WgConfig.AllowedIps[0].Addr()
}
func (conn *Conn) AgentVersionString() string {
return conn.config.AgentVersion
}
func (conn *Conn) presharedKey(remoteRosenpassKey []byte) *wgtypes.Key {
if conn.config.RosenpassConfig.PubKey == nil {
return conn.config.WgConfig.PreSharedKey
}
if remoteRosenpassKey == nil && conn.config.RosenpassConfig.PermissiveMode {
return conn.config.WgConfig.PreSharedKey
}
determKey, err := conn.rosenpassDetermKey()
if err != nil {
conn.Log.Errorf("failed to generate Rosenpass initial key: %v", err)
return conn.config.WgConfig.PreSharedKey
}
return determKey
}
// todo: move this logic into Rosenpass package
func (conn *Conn) rosenpassDetermKey() (*wgtypes.Key, error) {
lk := []byte(conn.config.LocalKey)
rk := []byte(conn.config.Key) // remote key
var keyInput []byte
if string(lk) > string(rk) {
//nolint:gocritic
keyInput = append(lk[:16], rk[:16]...)
} else {
//nolint:gocritic
keyInput = append(rk[:16], lk[:16]...)
}
key, err := wgtypes.NewKey(keyInput)
if err != nil {
return nil, err
}
return &key, nil
}
func isController(config ConnConfig) bool {
+6 -35
View File
@@ -1,58 +1,29 @@
package peer
import (
"sync/atomic"
log "github.com/sirupsen/logrus"
)
const (
// StatusConnected indicate the peer is in connected state
StatusConnected ConnStatus = iota
// StatusIdle indicate the peer is in disconnected state
StatusIdle ConnStatus = iota
// StatusConnecting indicate the peer is in connecting state
StatusConnecting
// StatusDisconnected indicate the peer is in disconnected state
StatusDisconnected
// StatusConnected indicate the peer is in connected state
StatusConnected
)
// ConnStatus describe the status of a peer's connection
type ConnStatus int32
// AtomicConnStatus is a thread-safe wrapper for ConnStatus
type AtomicConnStatus struct {
status atomic.Int32
}
// NewAtomicConnStatus creates a new AtomicConnStatus with the given initial status
func NewAtomicConnStatus() *AtomicConnStatus {
acs := &AtomicConnStatus{}
acs.Set(StatusDisconnected)
return acs
}
// Get returns the current connection status
func (acs *AtomicConnStatus) Get() ConnStatus {
return ConnStatus(acs.status.Load())
}
// Set updates the connection status
func (acs *AtomicConnStatus) Set(status ConnStatus) {
acs.status.Store(int32(status))
}
// String returns the string representation of the current status
func (acs *AtomicConnStatus) String() string {
return acs.Get().String()
}
func (s ConnStatus) String() string {
switch s {
case StatusConnecting:
return "Connecting"
case StatusConnected:
return "Connected"
case StatusDisconnected:
return "Disconnected"
case StatusIdle:
return "Idle"
default:
log.Errorf("unknown status: %d", s)
return "INVALID_PEER_CONNECTION_STATUS"
+1 -2
View File
@@ -14,7 +14,7 @@ func TestConnStatus_String(t *testing.T) {
want string
}{
{"StatusConnected", StatusConnected, "Connected"},
{"StatusDisconnected", StatusDisconnected, "Disconnected"},
{"StatusIdle", StatusIdle, "Idle"},
{"StatusConnecting", StatusConnecting, "Connecting"},
}
@@ -24,5 +24,4 @@ func TestConnStatus_String(t *testing.T) {
assert.Equal(t, got, table.want, "they should be equal")
})
}
}
+201 -78
View File
@@ -2,14 +2,15 @@ package peer
import (
"context"
"fmt"
"os"
"sync"
"testing"
"time"
"github.com/stretchr/testify/assert"
"github.com/netbirdio/netbird/client/iface"
"github.com/netbirdio/netbird/client/internal/peer/dispatcher"
"github.com/netbirdio/netbird/client/internal/peer/guard"
"github.com/netbirdio/netbird/client/internal/peer/ice"
"github.com/netbirdio/netbird/client/internal/stdnet"
@@ -17,6 +18,8 @@ import (
semaphoregroup "github.com/netbirdio/netbird/util/semaphore-group"
)
var testDispatcher = dispatcher.NewConnectionDispatcher()
var connConf = ConnConfig{
Key: "LLHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
LocalKey: "RRHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
@@ -28,7 +31,7 @@ var connConf = ConnConfig{
}
func TestMain(m *testing.M) {
_ = util.InitLog("trace", "console")
_ = util.InitLog("trace", util.LogConsole)
code := m.Run()
os.Exit(code)
}
@@ -47,7 +50,13 @@ func TestNewConn_interfaceFilter(t *testing.T) {
func TestConn_GetKey(t *testing.T) {
swWatcher := guard.NewSRWatcher(nil, nil, nil, connConf.ICEConfig)
conn, err := NewConn(context.Background(), connConf, nil, nil, nil, nil, swWatcher, semaphoregroup.NewSemaphoreGroup(1))
sd := ServiceDependencies{
SrWatcher: swWatcher,
Semaphore: semaphoregroup.NewSemaphoreGroup(1),
PeerConnDispatcher: testDispatcher,
}
conn, err := NewConn(connConf, sd)
if err != nil {
return
}
@@ -59,105 +68,219 @@ func TestConn_GetKey(t *testing.T) {
func TestConn_OnRemoteOffer(t *testing.T) {
swWatcher := guard.NewSRWatcher(nil, nil, nil, connConf.ICEConfig)
conn, err := NewConn(context.Background(), connConf, NewRecorder("https://mgm"), nil, nil, nil, swWatcher, semaphoregroup.NewSemaphoreGroup(1))
sd := ServiceDependencies{
StatusRecorder: NewRecorder("https://mgm"),
SrWatcher: swWatcher,
Semaphore: semaphoregroup.NewSemaphoreGroup(1),
PeerConnDispatcher: testDispatcher,
}
conn, err := NewConn(connConf, sd)
if err != nil {
return
}
wg := sync.WaitGroup{}
wg.Add(2)
go func() {
<-conn.handshaker.remoteOffersCh
wg.Done()
}()
onNewOffeChan := make(chan struct{})
go func() {
for {
accepted := conn.OnRemoteOffer(OfferAnswer{
IceCredentials: IceCredentials{
UFrag: "test",
Pwd: "test",
},
WgListenPort: 0,
Version: "",
})
if accepted {
wg.Done()
return
}
}
}()
conn.handshaker.AddOnNewOfferListener(func(remoteOfferAnswer *OfferAnswer) {
onNewOffeChan <- struct{}{}
})
wg.Wait()
conn.OnRemoteOffer(OfferAnswer{
IceCredentials: IceCredentials{
UFrag: "test",
Pwd: "test",
},
WgListenPort: 0,
Version: "",
})
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
select {
case <-onNewOffeChan:
// success
case <-ctx.Done():
t.Error("expected to receive a new offer notification, but timed out")
}
}
func TestConn_OnRemoteAnswer(t *testing.T) {
swWatcher := guard.NewSRWatcher(nil, nil, nil, connConf.ICEConfig)
conn, err := NewConn(context.Background(), connConf, NewRecorder("https://mgm"), nil, nil, nil, swWatcher, semaphoregroup.NewSemaphoreGroup(1))
sd := ServiceDependencies{
StatusRecorder: NewRecorder("https://mgm"),
SrWatcher: swWatcher,
Semaphore: semaphoregroup.NewSemaphoreGroup(1),
PeerConnDispatcher: testDispatcher,
}
conn, err := NewConn(connConf, sd)
if err != nil {
return
}
wg := sync.WaitGroup{}
wg.Add(2)
go func() {
<-conn.handshaker.remoteAnswerCh
wg.Done()
}()
onNewOffeChan := make(chan struct{})
go func() {
for {
accepted := conn.OnRemoteAnswer(OfferAnswer{
IceCredentials: IceCredentials{
UFrag: "test",
Pwd: "test",
},
WgListenPort: 0,
Version: "",
})
if accepted {
wg.Done()
return
}
}
}()
conn.handshaker.AddOnNewOfferListener(func(remoteOfferAnswer *OfferAnswer) {
onNewOffeChan <- struct{}{}
})
wg.Wait()
conn.OnRemoteAnswer(OfferAnswer{
IceCredentials: IceCredentials{
UFrag: "test",
Pwd: "test",
},
WgListenPort: 0,
Version: "",
})
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
select {
case <-onNewOffeChan:
// success
case <-ctx.Done():
t.Error("expected to receive a new offer notification, but timed out")
}
}
func TestConn_Status(t *testing.T) {
swWatcher := guard.NewSRWatcher(nil, nil, nil, connConf.ICEConfig)
conn, err := NewConn(context.Background(), connConf, NewRecorder("https://mgm"), nil, nil, nil, swWatcher, semaphoregroup.NewSemaphoreGroup(1))
if err != nil {
return
func TestConn_presharedKey(t *testing.T) {
conn1 := Conn{
config: ConnConfig{
Key: "LLHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
LocalKey: "RRHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
RosenpassConfig: RosenpassConfig{},
},
}
conn2 := Conn{
config: ConnConfig{
Key: "RRHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
LocalKey: "LLHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
RosenpassConfig: RosenpassConfig{},
},
}
tables := []struct {
name string
statusIce ConnStatus
statusRelay ConnStatus
want ConnStatus
tests := []struct {
conn1Permissive bool
conn1RosenpassEnabled bool
conn2Permissive bool
conn2RosenpassEnabled bool
conn1ExpectedInitialKey bool
conn2ExpectedInitialKey bool
}{
{"StatusConnected", StatusConnected, StatusConnected, StatusConnected},
{"StatusDisconnected", StatusDisconnected, StatusDisconnected, StatusDisconnected},
{"StatusConnecting", StatusConnecting, StatusConnecting, StatusConnecting},
{"StatusConnectingIce", StatusConnecting, StatusDisconnected, StatusConnecting},
{"StatusConnectingIceAlternative", StatusConnecting, StatusConnected, StatusConnected},
{"StatusConnectingRelay", StatusDisconnected, StatusConnecting, StatusConnecting},
{"StatusConnectingRelayAlternative", StatusConnected, StatusConnecting, StatusConnected},
{
conn1Permissive: false,
conn1RosenpassEnabled: false,
conn2Permissive: false,
conn2RosenpassEnabled: false,
conn1ExpectedInitialKey: false,
conn2ExpectedInitialKey: false,
},
{
conn1Permissive: false,
conn1RosenpassEnabled: true,
conn2Permissive: false,
conn2RosenpassEnabled: true,
conn1ExpectedInitialKey: true,
conn2ExpectedInitialKey: true,
},
{
conn1Permissive: false,
conn1RosenpassEnabled: true,
conn2Permissive: false,
conn2RosenpassEnabled: false,
conn1ExpectedInitialKey: true,
conn2ExpectedInitialKey: false,
},
{
conn1Permissive: false,
conn1RosenpassEnabled: false,
conn2Permissive: false,
conn2RosenpassEnabled: true,
conn1ExpectedInitialKey: false,
conn2ExpectedInitialKey: true,
},
{
conn1Permissive: true,
conn1RosenpassEnabled: true,
conn2Permissive: false,
conn2RosenpassEnabled: false,
conn1ExpectedInitialKey: false,
conn2ExpectedInitialKey: false,
},
{
conn1Permissive: false,
conn1RosenpassEnabled: false,
conn2Permissive: true,
conn2RosenpassEnabled: true,
conn1ExpectedInitialKey: false,
conn2ExpectedInitialKey: false,
},
{
conn1Permissive: true,
conn1RosenpassEnabled: true,
conn2Permissive: true,
conn2RosenpassEnabled: true,
conn1ExpectedInitialKey: true,
conn2ExpectedInitialKey: true,
},
{
conn1Permissive: false,
conn1RosenpassEnabled: false,
conn2Permissive: false,
conn2RosenpassEnabled: true,
conn1ExpectedInitialKey: false,
conn2ExpectedInitialKey: true,
},
{
conn1Permissive: false,
conn1RosenpassEnabled: true,
conn2Permissive: true,
conn2RosenpassEnabled: true,
conn1ExpectedInitialKey: true,
conn2ExpectedInitialKey: true,
},
}
for _, table := range tables {
t.Run(table.name, func(t *testing.T) {
si := NewAtomicConnStatus()
si.Set(table.statusIce)
conn.statusICE = si
conn1.config.RosenpassConfig.PermissiveMode = true
for i, test := range tests {
tcase := i + 1
t.Run(fmt.Sprintf("Rosenpass test case %d", tcase), func(t *testing.T) {
conn1.config.RosenpassConfig = RosenpassConfig{}
conn2.config.RosenpassConfig = RosenpassConfig{}
sr := NewAtomicConnStatus()
sr.Set(table.statusRelay)
conn.statusRelay = sr
if test.conn1RosenpassEnabled {
conn1.config.RosenpassConfig.PubKey = []byte("dummykey")
}
conn1.config.RosenpassConfig.PermissiveMode = test.conn1Permissive
got := conn.Status()
assert.Equal(t, got, table.want, "they should be equal")
if test.conn2RosenpassEnabled {
conn2.config.RosenpassConfig.PubKey = []byte("dummykey")
}
conn2.config.RosenpassConfig.PermissiveMode = test.conn2Permissive
conn1PresharedKey := conn1.presharedKey(conn2.config.RosenpassConfig.PubKey)
conn2PresharedKey := conn2.presharedKey(conn1.config.RosenpassConfig.PubKey)
if test.conn1ExpectedInitialKey {
if conn1PresharedKey == nil {
t.Errorf("Case %d: Expected conn1 to have a non-nil key, but got nil", tcase)
}
} else {
if conn1PresharedKey != nil {
t.Errorf("Case %d: Expected conn1 to have a nil key, but got %v", tcase, conn1PresharedKey)
}
}
// Assert conn2's key expectation
if test.conn2ExpectedInitialKey {
if conn2PresharedKey == nil {
t.Errorf("Case %d: Expected conn2 to have a non-nil key, but got nil", tcase)
}
} else {
if conn2PresharedKey != nil {
t.Errorf("Case %d: Expected conn2 to have a nil key, but got %v", tcase, conn2PresharedKey)
}
}
})
}
}
+29
View File
@@ -0,0 +1,29 @@
package conntype
import (
"fmt"
)
const (
None ConnPriority = 0
Relay ConnPriority = 1
ICETurn ConnPriority = 2
ICEP2P ConnPriority = 3
)
type ConnPriority int
func (cp ConnPriority) String() string {
switch cp {
case None:
return "None"
case Relay:
return "PriorityRelay"
case ICETurn:
return "PriorityICETurn"
case ICEP2P:
return "PriorityICEP2P"
default:
return fmt.Sprintf("ConnPriority(%d)", cp)
}
}
@@ -0,0 +1,52 @@
package dispatcher
import (
"sync"
"github.com/netbirdio/netbird/client/internal/peer/id"
)
type ConnectionListener struct {
OnConnected func(peerID id.ConnID)
OnDisconnected func(peerID id.ConnID)
}
type ConnectionDispatcher struct {
listeners map[*ConnectionListener]struct{}
mu sync.Mutex
}
func NewConnectionDispatcher() *ConnectionDispatcher {
return &ConnectionDispatcher{
listeners: make(map[*ConnectionListener]struct{}),
}
}
func (e *ConnectionDispatcher) AddListener(listener *ConnectionListener) {
e.mu.Lock()
defer e.mu.Unlock()
e.listeners[listener] = struct{}{}
}
func (e *ConnectionDispatcher) RemoveListener(listener *ConnectionListener) {
e.mu.Lock()
defer e.mu.Unlock()
delete(e.listeners, listener)
}
func (e *ConnectionDispatcher) NotifyConnected(peerConnID id.ConnID) {
e.mu.Lock()
defer e.mu.Unlock()
for listener := range e.listeners {
listener.OnConnected(peerConnID)
}
}
func (e *ConnectionDispatcher) NotifyDisconnected(peerConnID id.ConnID) {
e.mu.Lock()
defer e.mu.Unlock()
for listener := range e.listeners {
listener.OnDisconnected(peerConnID)
}
}
+19 -64
View File
@@ -8,10 +8,6 @@ import (
log "github.com/sirupsen/logrus"
)
const (
reconnectMaxElapsedTime = 30 * time.Minute
)
type isConnectedFunc func() bool
// Guard is responsible for the reconnection logic.
@@ -23,9 +19,7 @@ type isConnectedFunc func() bool
// - Relayed connection disconnected
// - ICE candidate changes
type Guard struct {
Reconnect chan struct{}
log *log.Entry
isController bool
isConnectedOnAllWay isConnectedFunc
timeout time.Duration
srWatcher *SRWatcher
@@ -33,11 +27,9 @@ type Guard struct {
iCEConnDisconnected chan struct{}
}
func NewGuard(log *log.Entry, isController bool, isConnectedFn isConnectedFunc, timeout time.Duration, srWatcher *SRWatcher) *Guard {
func NewGuard(log *log.Entry, isConnectedFn isConnectedFunc, timeout time.Duration, srWatcher *SRWatcher) *Guard {
return &Guard{
Reconnect: make(chan struct{}, 1),
log: log,
isController: isController,
isConnectedOnAllWay: isConnectedFn,
timeout: timeout,
srWatcher: srWatcher,
@@ -46,12 +38,9 @@ func NewGuard(log *log.Entry, isController bool, isConnectedFn isConnectedFunc,
}
}
func (g *Guard) Start(ctx context.Context) {
if g.isController {
g.reconnectLoopWithRetry(ctx)
} else {
g.listenForDisconnectEvents(ctx)
}
func (g *Guard) Start(ctx context.Context, eventCallback func()) {
g.log.Infof("starting guard for reconnection with MaxInterval: %s", g.timeout)
g.reconnectLoopWithRetry(ctx, eventCallback)
}
func (g *Guard) SetRelayedConnDisconnected() {
@@ -68,20 +57,17 @@ func (g *Guard) SetICEConnDisconnected() {
}
}
// reconnectLoopWithRetry periodically check (max 30 min) the connection status.
// reconnectLoopWithRetry periodically check the connection status.
// Try to send offer while the P2P is not established or while the Relay is not connected if is it supported
func (g *Guard) reconnectLoopWithRetry(ctx context.Context) {
waitForInitialConnectionTry(ctx)
func (g *Guard) reconnectLoopWithRetry(ctx context.Context, callback func()) {
srReconnectedChan := g.srWatcher.NewListener()
defer g.srWatcher.RemoveListener(srReconnectedChan)
ticker := g.prepareExponentTicker(ctx)
ticker := g.initialTicker(ctx)
defer ticker.Stop()
tickerChannel := ticker.C
g.log.Infof("start reconnect loop...")
for {
select {
case t := <-tickerChannel:
@@ -93,9 +79,8 @@ func (g *Guard) reconnectLoopWithRetry(ctx context.Context) {
}
if !g.isConnectedOnAllWay() {
g.triggerOfferSending()
callback()
}
case <-g.relayedConnDisconnected:
g.log.Debugf("Relay connection changed, reset reconnection ticker")
ticker.Stop()
@@ -121,30 +106,18 @@ func (g *Guard) reconnectLoopWithRetry(ctx context.Context) {
}
}
// listenForDisconnectEvents is used when the peer is not a controller and it should reconnect to the peer
// when the connection is lost. It will try to establish a connection only once time if before the connection was established
// It track separately the ice and relay connection status. Just because a lower priority connection reestablished it does not
// mean that to switch to it. We always force to use the higher priority connection.
func (g *Guard) listenForDisconnectEvents(ctx context.Context) {
srReconnectedChan := g.srWatcher.NewListener()
defer g.srWatcher.RemoveListener(srReconnectedChan)
// initialTicker give chance to the peer to establish the initial connection.
func (g *Guard) initialTicker(ctx context.Context) *backoff.Ticker {
bo := backoff.WithContext(&backoff.ExponentialBackOff{
InitialInterval: 3 * time.Second,
RandomizationFactor: 0.1,
Multiplier: 2,
MaxInterval: g.timeout,
Stop: backoff.Stop,
Clock: backoff.SystemClock,
}, ctx)
g.log.Infof("start listen for reconnect events...")
for {
select {
case <-g.relayedConnDisconnected:
g.log.Debugf("Relay connection changed, triggering reconnect")
g.triggerOfferSending()
case <-g.iCEConnDisconnected:
g.log.Debugf("ICE state changed, try to send new offer")
g.triggerOfferSending()
case <-srReconnectedChan:
g.triggerOfferSending()
case <-ctx.Done():
g.log.Debugf("context is done, stop reconnect loop")
return
}
}
return backoff.NewTicker(bo)
}
func (g *Guard) prepareExponentTicker(ctx context.Context) *backoff.Ticker {
@@ -153,7 +126,6 @@ func (g *Guard) prepareExponentTicker(ctx context.Context) *backoff.Ticker {
RandomizationFactor: 0.1,
Multiplier: 2,
MaxInterval: g.timeout,
MaxElapsedTime: reconnectMaxElapsedTime,
Stop: backoff.Stop,
Clock: backoff.SystemClock,
}, ctx)
@@ -163,20 +135,3 @@ func (g *Guard) prepareExponentTicker(ctx context.Context) *backoff.Ticker {
return ticker
}
func (g *Guard) triggerOfferSending() {
select {
case g.Reconnect <- struct{}{}:
default:
}
}
// Give chance to the peer to establish the initial connection.
// With it, we can decrease to send necessary offer
func waitForInitialConnectionTry(ctx context.Context) {
select {
case <-ctx.Done():
return
case <-time.After(3 * time.Second):
}
}
+1 -1
View File
@@ -6,7 +6,7 @@ import (
"sync"
"time"
"github.com/pion/ice/v3"
"github.com/pion/ice/v4"
log "github.com/sirupsen/logrus"
icemaker "github.com/netbirdio/netbird/client/internal/peer/ice"
+53 -68
View File
@@ -39,11 +39,19 @@ type OfferAnswer struct {
// relay server address
RelaySrvAddress string
// SessionID is the unique identifier of the session, used to discard old messages
SessionID *ICESessionID
}
func (oa *OfferAnswer) SessionIDString() string {
if oa.SessionID == nil {
return "unknown"
}
return oa.SessionID.String()
}
type Handshaker struct {
mu sync.Mutex
ctx context.Context
log *log.Entry
config ConnConfig
signaler *Signaler
@@ -57,9 +65,8 @@ type Handshaker struct {
remoteAnswerCh chan OfferAnswer
}
func NewHandshaker(ctx context.Context, log *log.Entry, config ConnConfig, signaler *Signaler, ice *WorkerICE, relay *WorkerRelay) *Handshaker {
func NewHandshaker(log *log.Entry, config ConnConfig, signaler *Signaler, ice *WorkerICE, relay *WorkerRelay) *Handshaker {
return &Handshaker{
ctx: ctx,
log: log,
config: config,
signaler: signaler,
@@ -74,23 +81,27 @@ func (h *Handshaker) AddOnNewOfferListener(offer func(remoteOfferAnswer *OfferAn
h.onNewOfferListeners = append(h.onNewOfferListeners, offer)
}
func (h *Handshaker) Listen() {
func (h *Handshaker) Listen(ctx context.Context) {
for {
h.log.Debugf("wait for remote offer confirmation")
remoteOfferAnswer, err := h.waitForRemoteOfferConfirmation()
if err != nil {
var connectionClosedError *ConnectionClosedError
if errors.As(err, &connectionClosedError) {
h.log.Tracef("stop handshaker")
return
select {
case remoteOfferAnswer := <-h.remoteOffersCh:
// received confirmation from the remote peer -> ready to proceed
if err := h.sendAnswer(); err != nil {
h.log.Errorf("failed to send remote offer confirmation: %s", err)
continue
}
h.log.Errorf("failed to received remote offer confirmation: %s", err)
continue
}
h.log.Debugf("received connection confirmation, running version %s and with remote WireGuard listen port %d", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort)
for _, listener := range h.onNewOfferListeners {
go listener(remoteOfferAnswer)
for _, listener := range h.onNewOfferListeners {
listener(&remoteOfferAnswer)
}
h.log.Infof("received offer, running version %s, remote WireGuard listen port %d, session id: %s", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort, remoteOfferAnswer.SessionIDString())
case remoteOfferAnswer := <-h.remoteAnswerCh:
h.log.Infof("received answer, running version %s, remote WireGuard listen port %d, session id: %s", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort, remoteOfferAnswer.SessionIDString())
for _, listener := range h.onNewOfferListeners {
listener(&remoteOfferAnswer)
}
case <-ctx.Done():
h.log.Infof("stop listening for remote offers and answers")
return
}
}
}
@@ -103,44 +114,27 @@ func (h *Handshaker) SendOffer() error {
// OnRemoteOffer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
// doesn't block, discards the message if connection wasn't ready
func (h *Handshaker) OnRemoteOffer(offer OfferAnswer) bool {
func (h *Handshaker) OnRemoteOffer(offer OfferAnswer) {
select {
case h.remoteOffersCh <- offer:
return true
return
default:
h.log.Debugf("OnRemoteOffer skipping message because is not ready")
h.log.Warnf("skipping remote offer message because receiver not ready")
// connection might not be ready yet to receive so we ignore the message
return false
return
}
}
// OnRemoteAnswer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
// doesn't block, discards the message if connection wasn't ready
func (h *Handshaker) OnRemoteAnswer(answer OfferAnswer) bool {
func (h *Handshaker) OnRemoteAnswer(answer OfferAnswer) {
select {
case h.remoteAnswerCh <- answer:
return true
return
default:
// connection might not be ready yet to receive so we ignore the message
h.log.Debugf("OnRemoteAnswer skipping message because is not ready")
return false
}
}
func (h *Handshaker) waitForRemoteOfferConfirmation() (*OfferAnswer, error) {
select {
case remoteOfferAnswer := <-h.remoteOffersCh:
// received confirmation from the remote peer -> ready to proceed
err := h.sendAnswer()
if err != nil {
return nil, err
}
return &remoteOfferAnswer, nil
case remoteOfferAnswer := <-h.remoteAnswerCh:
return &remoteOfferAnswer, nil
case <-h.ctx.Done():
// closed externally
return nil, NewConnectionClosedError(h.config.Key)
h.log.Warnf("skipping remote answer message because receiver not ready")
return
}
}
@@ -150,43 +144,34 @@ func (h *Handshaker) sendOffer() error {
return ErrSignalIsNotReady
}
iceUFrag, icePwd := h.ice.GetLocalUserCredentials()
offer := OfferAnswer{
IceCredentials: IceCredentials{iceUFrag, icePwd},
WgListenPort: h.config.LocalWgPort,
Version: version.NetbirdVersion(),
RosenpassPubKey: h.config.RosenpassPubKey,
RosenpassAddr: h.config.RosenpassAddr,
}
addr, err := h.relay.RelayInstanceAddress()
if err == nil {
offer.RelaySrvAddress = addr
}
offer := h.buildOfferAnswer()
h.log.Infof("sending offer with serial: %s", offer.SessionIDString())
return h.signaler.SignalOffer(offer, h.config.Key)
}
func (h *Handshaker) sendAnswer() error {
h.log.Debugf("sending answer")
uFrag, pwd := h.ice.GetLocalUserCredentials()
answer := h.buildOfferAnswer()
h.log.Infof("sending answer with serial: %s", answer.SessionIDString())
return h.signaler.SignalAnswer(answer, h.config.Key)
}
func (h *Handshaker) buildOfferAnswer() OfferAnswer {
uFrag, pwd := h.ice.GetLocalUserCredentials()
sid := h.ice.SessionID()
answer := OfferAnswer{
IceCredentials: IceCredentials{uFrag, pwd},
WgListenPort: h.config.LocalWgPort,
Version: version.NetbirdVersion(),
RosenpassPubKey: h.config.RosenpassPubKey,
RosenpassAddr: h.config.RosenpassAddr,
RosenpassPubKey: h.config.RosenpassConfig.PubKey,
RosenpassAddr: h.config.RosenpassConfig.Addr,
SessionID: &sid,
}
addr, err := h.relay.RelayInstanceAddress()
if err == nil {
if addr, err := h.relay.RelayInstanceAddress(); err == nil {
answer.RelaySrvAddress = addr
}
err = h.signaler.SignalAnswer(answer, h.config.Key)
if err != nil {
return err
}
return nil
return answer
}
+1 -1
View File
@@ -3,7 +3,7 @@ package ice
import (
"sync/atomic"
"github.com/pion/stun/v2"
"github.com/pion/stun/v3"
)
type StunTurn atomic.Value
+30 -7
View File
@@ -1,9 +1,11 @@
package ice
import (
"sync"
"time"
"github.com/pion/ice/v3"
"github.com/pion/ice/v4"
"github.com/pion/logging"
"github.com/pion/randutil"
log "github.com/sirupsen/logrus"
@@ -17,17 +19,28 @@ const (
iceKeepAliveDefault = 4 * time.Second
iceDisconnectedTimeoutDefault = 6 * time.Second
iceFailedTimeoutDefault = 6 * time.Second
// iceRelayAcceptanceMinWaitDefault is the same as in the Pion ICE package
iceRelayAcceptanceMinWaitDefault = 2 * time.Second
)
var (
failedTimeout = 6 * time.Second
)
type ThreadSafeAgent struct {
*ice.Agent
once sync.Once
}
func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candidateTypes []ice.CandidateType, ufrag string, pwd string) (*ice.Agent, error) {
func (a *ThreadSafeAgent) Close() error {
var err error
a.once.Do(func() {
err = a.Agent.Close()
})
return err
}
func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candidateTypes []ice.CandidateType, ufrag string, pwd string) (*ThreadSafeAgent, error) {
iceKeepAlive := iceKeepAlive()
iceDisconnectedTimeout := iceDisconnectedTimeout()
iceFailedTimeout := iceFailedTimeout()
iceRelayAcceptanceMinWait := iceRelayAcceptanceMinWait()
transportNet, err := newStdNet(iFaceDiscover, config.InterfaceBlackList)
@@ -35,6 +48,10 @@ func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candida
log.Errorf("failed to create pion's stdnet: %s", err)
}
fac := logging.NewDefaultLoggerFactory()
//fac.Writer = log.StandardLogger().Writer()
agentConfig := &ice.AgentConfig{
MulticastDNSMode: ice.MulticastDNSModeDisabled,
NetworkTypes: []ice.NetworkType{ice.NetworkTypeUDP4, ice.NetworkTypeUDP6},
@@ -45,19 +62,25 @@ func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candida
UDPMuxSrflx: config.UDPMuxSrflx,
NAT1To1IPs: config.NATExternalIPs,
Net: transportNet,
FailedTimeout: &failedTimeout,
FailedTimeout: &iceFailedTimeout,
DisconnectedTimeout: &iceDisconnectedTimeout,
KeepaliveInterval: &iceKeepAlive,
RelayAcceptanceMinWait: &iceRelayAcceptanceMinWait,
LocalUfrag: ufrag,
LocalPwd: pwd,
LoggerFactory: fac,
}
if config.DisableIPv6Discovery {
agentConfig.NetworkTypes = []ice.NetworkType{ice.NetworkTypeUDP4}
}
return ice.NewAgent(agentConfig)
agent, err := ice.NewAgent(agentConfig)
if err != nil {
return nil, err
}
return &ThreadSafeAgent{Agent: agent}, nil
}
func GenerateICECredentials() (string, string, error) {
+1 -1
View File
@@ -1,7 +1,7 @@
package ice
import (
"github.com/pion/ice/v3"
"github.com/pion/ice/v4"
)
type Config struct {
+17
View File
@@ -13,6 +13,7 @@ const (
envICEForceRelayConn = "NB_ICE_FORCE_RELAY_CONN"
envICEKeepAliveIntervalSec = "NB_ICE_KEEP_ALIVE_INTERVAL_SEC"
envICEDisconnectedTimeoutSec = "NB_ICE_DISCONNECTED_TIMEOUT_SEC"
envICEFailedTimeoutSec = "NB_ICE_FAILED_TIMEOUT_SEC"
envICERelayAcceptanceMinWaitSec = "NB_ICE_RELAY_ACCEPTANCE_MIN_WAIT_SEC"
msgWarnInvalidValue = "invalid value %s set for %s, using default %v"
@@ -55,6 +56,22 @@ func iceDisconnectedTimeout() time.Duration {
return time.Duration(disconnectedTimeoutSec) * time.Second
}
func iceFailedTimeout() time.Duration {
failedTimeoutEnv := os.Getenv(envICEFailedTimeoutSec)
if failedTimeoutEnv == "" {
return iceFailedTimeoutDefault
}
log.Infof("setting ICE failed timeout to %s seconds", failedTimeoutEnv)
failedTimeoutSec, err := strconv.Atoi(failedTimeoutEnv)
if err != nil {
log.Warnf(msgWarnInvalidValue, failedTimeoutEnv, envICEFailedTimeoutSec, iceFailedTimeoutDefault)
return iceFailedTimeoutDefault
}
return time.Duration(failedTimeoutSec) * time.Second
}
func iceRelayAcceptanceMinWait() time.Duration {
iceRelayAcceptanceMinWaitEnv := os.Getenv(envICERelayAcceptanceMinWaitSec)
if iceRelayAcceptanceMinWaitEnv == "" {
+5
View File
@@ -0,0 +1,5 @@
package id
import "unsafe"
type ConnID unsafe.Pointer
+8 -4
View File
@@ -2,16 +2,20 @@ package peer
import (
"net"
"net/netip"
"time"
"github.com/netbirdio/netbird/client/iface/configurer"
"github.com/netbirdio/netbird/client/iface/wgproxy"
"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
"github.com/netbirdio/netbird/client/iface/configurer"
"github.com/netbirdio/netbird/client/iface/wgaddr"
"github.com/netbirdio/netbird/client/iface/wgproxy"
)
type WGIface interface {
UpdatePeer(peerKey string, allowedIps string, keepAlive time.Duration, endpoint *net.UDPAddr, preSharedKey *wgtypes.Key) error
UpdatePeer(peerKey string, allowedIps []netip.Prefix, keepAlive time.Duration, endpoint *net.UDPAddr, preSharedKey *wgtypes.Key) error
RemovePeer(peerKey string) error
GetStats(peerKey string) (configurer.WGStats, error)
GetStats() (map[string]configurer.WGStats, error)
GetProxy() wgproxy.Proxy
Address() wgaddr.Address
}
+65 -37
View File
@@ -18,6 +18,8 @@ type notifier struct {
currentClientState bool
lastNotification int
lastNumberOfPeers int
lastFqdnAddress string
lastIPAddress string
}
func newNotifier() *notifier {
@@ -25,15 +27,22 @@ func newNotifier() *notifier {
}
func (n *notifier) setListener(listener Listener) {
n.serverStateLock.Lock()
lastNotification := n.lastNotification
numOfPeers := n.lastNumberOfPeers
fqdnAddress := n.lastFqdnAddress
address := n.lastIPAddress
n.serverStateLock.Unlock()
n.listenersLock.Lock()
defer n.listenersLock.Unlock()
n.serverStateLock.Lock()
n.notifyListener(listener, n.lastNotification)
listener.OnPeersListChanged(n.lastNumberOfPeers)
n.serverStateLock.Unlock()
n.listener = listener
listener.OnAddressChanged(fqdnAddress, address)
notifyListener(listener, lastNotification)
// run on go routine to avoid on Java layer to call go functions on same thread
go listener.OnPeersListChanged(numOfPeers)
}
func (n *notifier) removeListener() {
@@ -44,41 +53,44 @@ func (n *notifier) removeListener() {
func (n *notifier) updateServerStates(mgmState bool, signalState bool) {
n.serverStateLock.Lock()
defer n.serverStateLock.Unlock()
calculatedState := n.calculateState(mgmState, signalState)
if !n.isServerStateChanged(calculatedState) {
n.serverStateLock.Unlock()
return
}
n.lastNotification = calculatedState
n.serverStateLock.Unlock()
n.notify(n.lastNotification)
n.notify(calculatedState)
}
func (n *notifier) clientStart() {
n.serverStateLock.Lock()
defer n.serverStateLock.Unlock()
n.currentClientState = true
n.lastNotification = stateConnecting
n.notify(n.lastNotification)
n.serverStateLock.Unlock()
n.notify(stateConnecting)
}
func (n *notifier) clientStop() {
n.serverStateLock.Lock()
defer n.serverStateLock.Unlock()
n.currentClientState = false
n.lastNotification = stateDisconnected
n.notify(n.lastNotification)
n.serverStateLock.Unlock()
n.notify(stateDisconnected)
}
func (n *notifier) clientTearDown() {
n.serverStateLock.Lock()
defer n.serverStateLock.Unlock()
n.currentClientState = false
n.lastNotification = stateDisconnecting
n.notify(n.lastNotification)
n.serverStateLock.Unlock()
n.notify(stateDisconnecting)
}
func (n *notifier) isServerStateChanged(newState int) bool {
@@ -87,26 +99,14 @@ func (n *notifier) isServerStateChanged(newState int) bool {
func (n *notifier) notify(state int) {
n.listenersLock.Lock()
defer n.listenersLock.Unlock()
if n.listener == nil {
listener := n.listener
n.listenersLock.Unlock()
if listener == nil {
return
}
n.notifyListener(n.listener, state)
}
func (n *notifier) notifyListener(l Listener, state int) {
go func() {
switch state {
case stateDisconnected:
l.OnDisconnected()
case stateConnected:
l.OnConnected()
case stateConnecting:
l.OnConnecting()
case stateDisconnecting:
l.OnDisconnecting()
}
}()
notifyListener(listener, state)
}
func (n *notifier) calculateState(managementConn, signalConn bool) int {
@@ -126,20 +126,48 @@ func (n *notifier) calculateState(managementConn, signalConn bool) int {
}
func (n *notifier) peerListChanged(numOfPeers int) {
n.serverStateLock.Lock()
n.lastNumberOfPeers = numOfPeers
n.serverStateLock.Unlock()
n.listenersLock.Lock()
defer n.listenersLock.Unlock()
if n.listener == nil {
listener := n.listener
n.listenersLock.Unlock()
if listener == nil {
return
}
n.listener.OnPeersListChanged(numOfPeers)
// run on go routine to avoid on Java layer to call go functions on same thread
go listener.OnPeersListChanged(numOfPeers)
}
func (n *notifier) localAddressChanged(fqdn, address string) {
n.serverStateLock.Lock()
n.lastFqdnAddress = fqdn
n.lastIPAddress = address
n.serverStateLock.Unlock()
n.listenersLock.Lock()
defer n.listenersLock.Unlock()
if n.listener == nil {
listener := n.listener
n.listenersLock.Unlock()
if listener == nil {
return
}
n.listener.OnAddressChanged(fqdn, address)
listener.OnAddressChanged(fqdn, address)
}
func notifyListener(l Listener, state int) {
switch state {
case stateDisconnected:
l.OnDisconnected()
case stateConnected:
l.OnConnected()
case stateConnecting:
l.OnConnecting()
case stateDisconnecting:
l.OnDisconnecting()
}
}
+133
View File
@@ -0,0 +1,133 @@
package peer
import (
"net/netip"
"sort"
"sync"
log "github.com/sirupsen/logrus"
"github.com/netbirdio/netbird/route"
)
// routeEntry holds the route prefix and the corresponding resource ID.
type routeEntry struct {
prefix netip.Prefix
resourceID route.ResID
}
type routeIDLookup struct {
localRoutes []routeEntry
localLock sync.RWMutex
remoteRoutes []routeEntry
remoteLock sync.RWMutex
resolvedIPs sync.Map
}
func (r *routeIDLookup) AddLocalRouteID(resourceID route.ResID, route netip.Prefix) {
r.localLock.Lock()
defer r.localLock.Unlock()
// update the resource id if the route already exists.
for i, entry := range r.localRoutes {
if entry.prefix == route {
r.localRoutes[i].resourceID = resourceID
log.Tracef("resourceID for route %v updated to %s in local routes", route, resourceID)
return
}
}
// append and sort descending by prefix bits (more specific first)
r.localRoutes = append(r.localRoutes, routeEntry{prefix: route, resourceID: resourceID})
sort.Slice(r.localRoutes, func(i, j int) bool {
return r.localRoutes[i].prefix.Bits() > r.localRoutes[j].prefix.Bits()
})
}
func (r *routeIDLookup) RemoveLocalRouteID(route netip.Prefix) {
r.localLock.Lock()
defer r.localLock.Unlock()
for i, entry := range r.localRoutes {
if entry.prefix == route {
r.localRoutes = append(r.localRoutes[:i], r.localRoutes[i+1:]...)
return
}
}
}
func (r *routeIDLookup) AddRemoteRouteID(resourceID route.ResID, route netip.Prefix) {
r.remoteLock.Lock()
defer r.remoteLock.Unlock()
for i, entry := range r.remoteRoutes {
if entry.prefix == route {
r.remoteRoutes[i].resourceID = resourceID
log.Tracef("resourceID for route %v updated to %s in remote routes", route, resourceID)
return
}
}
// append and sort descending by prefix bits.
r.remoteRoutes = append(r.remoteRoutes, routeEntry{prefix: route, resourceID: resourceID})
sort.Slice(r.remoteRoutes, func(i, j int) bool {
return r.remoteRoutes[i].prefix.Bits() > r.remoteRoutes[j].prefix.Bits()
})
}
func (r *routeIDLookup) RemoveRemoteRouteID(route netip.Prefix) {
r.remoteLock.Lock()
defer r.remoteLock.Unlock()
for i, entry := range r.remoteRoutes {
if entry.prefix == route {
r.remoteRoutes = append(r.remoteRoutes[:i], r.remoteRoutes[i+1:]...)
return
}
}
}
func (r *routeIDLookup) AddResolvedIP(resourceID route.ResID, route netip.Prefix) {
r.resolvedIPs.Store(route.Addr(), resourceID)
}
func (r *routeIDLookup) RemoveResolvedIP(route netip.Prefix) {
r.resolvedIPs.Delete(route.Addr())
}
// Lookup returns the resource ID for the given IP address
// and a bool indicating if the IP is an exit node.
func (r *routeIDLookup) Lookup(ip netip.Addr) (route.ResID, bool) {
if res, ok := r.resolvedIPs.Load(ip); ok {
return res.(route.ResID), false
}
var resourceID route.ResID
var isExitNode bool
r.localLock.RLock()
for _, entry := range r.localRoutes {
if entry.prefix.Contains(ip) {
resourceID = entry.resourceID
isExitNode = entry.prefix.Bits() == 0
break
}
}
r.localLock.RUnlock()
if resourceID == "" {
r.remoteLock.RLock()
for _, entry := range r.remoteRoutes {
if entry.prefix.Contains(ip) {
resourceID = entry.resourceID
isExitNode = entry.prefix.Bits() == 0
break
}
}
r.remoteLock.RUnlock()
}
return resourceID, isExitNode
}
+47
View File
@@ -0,0 +1,47 @@
package peer
import (
"crypto/rand"
"encoding/hex"
"fmt"
"io"
)
const sessionIDSize = 5
type ICESessionID string
// NewICESessionID generates a new session ID for distinguishing sessions
func NewICESessionID() (ICESessionID, error) {
b := make([]byte, sessionIDSize)
if _, err := io.ReadFull(rand.Reader, b); err != nil {
return "", fmt.Errorf("failed to generate session ID: %w", err)
}
return ICESessionID(hex.EncodeToString(b)), nil
}
func ICESessionIDFromBytes(b []byte) (ICESessionID, error) {
if len(b) != sessionIDSize {
return "", fmt.Errorf("invalid session ID length: %d", len(b))
}
return ICESessionID(hex.EncodeToString(b)), nil
}
// Bytes returns the raw bytes of the session ID for protobuf serialization
func (id ICESessionID) Bytes() ([]byte, error) {
if len(id) == 0 {
return nil, fmt.Errorf("ICE session ID is empty")
}
b, err := hex.DecodeString(string(id))
if err != nil {
return nil, fmt.Errorf("invalid ICE session ID encoding: %w", err)
}
if len(b) != sessionIDSize {
return nil, fmt.Errorf("invalid ICE session ID length: expected %d bytes, got %d", sessionIDSize, len(b))
}
return b, nil
}
func (id ICESessionID) String() string {
return string(id)
}
+21 -6
View File
@@ -1,11 +1,12 @@
package peer
import (
"github.com/pion/ice/v3"
"github.com/pion/ice/v4"
log "github.com/sirupsen/logrus"
"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
signal "github.com/netbirdio/netbird/signal/client"
sProto "github.com/netbirdio/netbird/signal/proto"
signal "github.com/netbirdio/netbird/shared/signal/client"
sProto "github.com/netbirdio/netbird/shared/signal/proto"
)
type Signaler struct {
@@ -45,6 +46,10 @@ func (s *Signaler) Ready() bool {
// SignalOfferAnswer signals either an offer or an answer to remote peer
func (s *Signaler) signalOfferAnswer(offerAnswer OfferAnswer, remoteKey string, bodyType sProto.Body_Type) error {
sessionIDBytes, err := offerAnswer.SessionID.Bytes()
if err != nil {
log.Warnf("failed to get session ID bytes: %v", err)
}
msg, err := signal.MarshalCredential(
s.wgPrivateKey,
offerAnswer.WgListenPort,
@@ -56,15 +61,25 @@ func (s *Signaler) signalOfferAnswer(offerAnswer OfferAnswer, remoteKey string,
bodyType,
offerAnswer.RosenpassPubKey,
offerAnswer.RosenpassAddr,
offerAnswer.RelaySrvAddress)
offerAnswer.RelaySrvAddress,
sessionIDBytes)
if err != nil {
return err
}
err = s.signal.Send(msg)
if err != nil {
if err = s.signal.Send(msg); err != nil {
return err
}
return nil
}
func (s *Signaler) SignalIdle(remoteKey string) error {
return s.signal.Send(&sProto.Message{
Key: s.wgPrivateKey.PublicKey().String(),
RemoteKey: remoteKey,
Body: &sProto.Body{
Type: sProto.Body_GO_IDLE,
},
})
}
+122
View File
@@ -0,0 +1,122 @@
package peer
import (
"context"
"sync"
"time"
log "github.com/sirupsen/logrus"
)
type stateDump struct {
log *log.Entry
status *Status
key string
sentOffer int
remoteOffer int
remoteAnswer int
remoteCandidate int
p2pConnected int
switchToRelay int
wgCheckSuccess int
relayConnected int
localProxies int
mu sync.Mutex
}
func newStateDump(key string, log *log.Entry, statusRecorder *Status) *stateDump {
return &stateDump{
log: log,
status: statusRecorder,
key: key,
}
}
func (s *stateDump) Start(ctx context.Context) {
ticker := time.NewTicker(10 * time.Minute)
defer ticker.Stop()
for {
select {
case <-ticker.C:
s.dumpState()
case <-ctx.Done():
return
}
}
}
func (s *stateDump) RemoteOffer() {
s.mu.Lock()
defer s.mu.Unlock()
s.remoteOffer++
}
func (s *stateDump) RemoteCandidate() {
s.mu.Lock()
defer s.mu.Unlock()
s.remoteCandidate++
}
func (s *stateDump) SendOffer() {
s.mu.Lock()
defer s.mu.Unlock()
s.sentOffer++
}
func (s *stateDump) dumpState() {
s.mu.Lock()
defer s.mu.Unlock()
status := "unknown"
state, e := s.status.GetPeer(s.key)
if e == nil {
status = state.ConnStatus.String()
}
s.log.Infof("Dump stat: Status: %s, SentOffer: %d, RemoteOffer: %d, RemoteAnswer: %d, RemoteCandidate: %d, P2PConnected: %d, SwitchToRelay: %d, WGCheckSuccess: %d, RelayConnected: %d, LocalProxies: %d",
status, s.sentOffer, s.remoteOffer, s.remoteAnswer, s.remoteCandidate, s.p2pConnected, s.switchToRelay, s.wgCheckSuccess, s.relayConnected, s.localProxies)
}
func (s *stateDump) RemoteAnswer() {
s.mu.Lock()
defer s.mu.Unlock()
s.remoteAnswer++
}
func (s *stateDump) P2PConnected() {
s.mu.Lock()
defer s.mu.Unlock()
s.p2pConnected++
}
func (s *stateDump) SwitchToRelay() {
s.mu.Lock()
defer s.mu.Unlock()
s.switchToRelay++
}
func (s *stateDump) WGcheckSuccess() {
s.mu.Lock()
defer s.mu.Unlock()
s.wgCheckSuccess++
}
func (s *stateDump) RelayConnected() {
s.mu.Lock()
defer s.mu.Unlock()
s.relayConnected++
}
func (s *stateDump) NewLocalProxy() {
s.mu.Lock()
defer s.mu.Unlock()
s.localProxies++
}
+311 -78
View File
@@ -1,7 +1,9 @@
package peer
import (
"context"
"errors"
"fmt"
"net/netip"
"slices"
"sync"
@@ -14,11 +16,14 @@ import (
gstatus "google.golang.org/grpc/status"
"google.golang.org/protobuf/types/known/timestamppb"
firewall "github.com/netbirdio/netbird/client/firewall/manager"
"github.com/netbirdio/netbird/client/iface/configurer"
"github.com/netbirdio/netbird/client/internal/ingressgw"
"github.com/netbirdio/netbird/client/internal/relay"
"github.com/netbirdio/netbird/client/proto"
"github.com/netbirdio/netbird/management/domain"
relayClient "github.com/netbirdio/netbird/relay/client"
"github.com/netbirdio/netbird/shared/management/domain"
relayClient "github.com/netbirdio/netbird/shared/relay/client"
"github.com/netbirdio/netbird/route"
)
const eventQueueSize = 10
@@ -28,10 +33,21 @@ type ResolvedDomainInfo struct {
ParentDomain domain.Domain
}
type WGIfaceStatus interface {
FullStats() (*configurer.Stats, error)
}
type EventListener interface {
OnEvent(event *proto.SystemEvent)
}
// RouterState status for router peers. This contains relevant fields for route manager
type RouterState struct {
Status ConnStatus
Relayed bool
Latency time.Duration
}
// State contains the latest state of a peer
type State struct {
Mux *sync.RWMutex
@@ -124,7 +140,7 @@ type RosenpassState struct {
// whether it's enabled, and the last error message encountered during probing.
type NSGroupState struct {
ID string
Servers []string
Servers []netip.AddrPort
Domains []string
Enabled bool
Error error
@@ -132,20 +148,43 @@ type NSGroupState struct {
// FullStatus contains the full state held by the Status instance
type FullStatus struct {
Peers []State
ManagementState ManagementState
SignalState SignalState
LocalPeerState LocalPeerState
RosenpassState RosenpassState
Relays []relay.ProbeResult
NSGroupStates []NSGroupState
Peers []State
ManagementState ManagementState
SignalState SignalState
LocalPeerState LocalPeerState
RosenpassState RosenpassState
Relays []relay.ProbeResult
NSGroupStates []NSGroupState
NumOfForwardingRules int
LazyConnectionEnabled bool
}
type StatusChangeSubscription struct {
peerID string
id string
eventsChan chan map[string]RouterState
ctx context.Context
}
func newStatusChangeSubscription(ctx context.Context, peerID string) *StatusChangeSubscription {
return &StatusChangeSubscription{
ctx: ctx,
peerID: peerID,
id: uuid.New().String(),
// it is a buffer for notifications to block less the status recorded
eventsChan: make(chan map[string]RouterState, 8),
}
}
func (s *StatusChangeSubscription) Events() chan map[string]RouterState {
return s.eventsChan
}
// Status holds a state of peers, signal, management connections and relays
type Status struct {
mux sync.Mutex
peers map[string]State
changeNotify map[string]chan struct{}
changeNotify map[string]map[string]*StatusChangeSubscription // map[peerID]map[subscriptionID]*StatusChangeSubscription
signalState bool
signalError error
managementState bool
@@ -160,6 +199,7 @@ type Status struct {
rosenpassPermissive bool
nsGroupStates []NSGroupState
resolvedDomainsStates map[domain.Domain]ResolvedDomainInfo
lazyConnectionEnabled bool
// To reduce the number of notification invocation this bool will be true when need to call the notification
// Some Peer actions mostly used by in a batch when the network map has been synchronized. In these type of events
@@ -171,13 +211,18 @@ type Status struct {
eventMux sync.RWMutex
eventStreams map[string]chan *proto.SystemEvent
eventQueue *EventQueue
ingressGwMgr *ingressgw.Manager
routeIDLookup routeIDLookup
wgIface WGIfaceStatus
}
// NewRecorder returns a new Status instance
func NewRecorder(mgmAddress string) *Status {
return &Status{
peers: make(map[string]State),
changeNotify: make(map[string]chan struct{}),
changeNotify: make(map[string]map[string]*StatusChangeSubscription),
eventStreams: make(map[string]chan *proto.SystemEvent),
eventQueue: NewEventQueue(eventQueueSize),
offlinePeers: make([]State, 0),
@@ -193,6 +238,12 @@ func (d *Status) SetRelayMgr(manager *relayClient.Manager) {
d.relayMgr = manager
}
func (d *Status) SetIngressGwMgr(ingressGwMgr *ingressgw.Manager) {
d.mux.Lock()
defer d.mux.Unlock()
d.ingressGwMgr = ingressGwMgr
}
// ReplaceOfflinePeers replaces
func (d *Status) ReplaceOfflinePeers(replacement []State) {
d.mux.Lock()
@@ -205,7 +256,7 @@ func (d *Status) ReplaceOfflinePeers(replacement []State) {
}
// AddPeer adds peer to Daemon status map
func (d *Status) AddPeer(peerPubKey string, fqdn string) error {
func (d *Status) AddPeer(peerPubKey string, fqdn string, ip string) error {
d.mux.Lock()
defer d.mux.Unlock()
@@ -215,7 +266,8 @@ func (d *Status) AddPeer(peerPubKey string, fqdn string) error {
}
d.peers[peerPubKey] = State{
PubKey: peerPubKey,
ConnStatus: StatusDisconnected,
IP: ip,
ConnStatus: StatusIdle,
FQDN: fqdn,
Mux: new(sync.RWMutex),
}
@@ -235,6 +287,18 @@ func (d *Status) GetPeer(peerPubKey string) (State, error) {
return state, nil
}
func (d *Status) PeerByIP(ip string) (string, bool) {
d.mux.Lock()
defer d.mux.Unlock()
for _, state := range d.peers {
if state.IP == ip {
return state.FQDN, true
}
}
return "", false
}
// RemovePeer removes peer from Daemon status map
func (d *Status) RemovePeer(peerPubKey string) error {
d.mux.Lock()
@@ -260,11 +324,7 @@ func (d *Status) UpdatePeerState(receivedState State) error {
return errors.New("peer doesn't exist")
}
if receivedState.IP != "" {
peerState.IP = receivedState.IP
}
skipNotification := shouldSkipNotify(receivedState.ConnStatus, peerState)
oldState := peerState.ConnStatus
if receivedState.ConnStatus != peerState.ConnStatus {
peerState.ConnStatus = receivedState.ConnStatus
@@ -280,15 +340,18 @@ func (d *Status) UpdatePeerState(receivedState State) error {
d.peers[receivedState.PubKey] = peerState
if skipNotification {
return nil
if hasConnStatusChanged(oldState, receivedState.ConnStatus) {
d.notifyPeerListChanged()
}
d.notifyPeerListChanged()
// when we close the connection we will not notify the router manager
if receivedState.ConnStatus == StatusIdle {
d.notifyPeerStateChangeListeners(receivedState.PubKey)
}
return nil
}
func (d *Status) AddPeerStateRoute(peer string, route string) error {
func (d *Status) AddPeerStateRoute(peer string, route string, resourceId route.ResID) error {
d.mux.Lock()
defer d.mux.Unlock()
@@ -300,6 +363,11 @@ func (d *Status) AddPeerStateRoute(peer string, route string) error {
peerState.AddRoute(route)
d.peers[peer] = peerState
pref, err := netip.ParsePrefix(route)
if err == nil {
d.routeIDLookup.AddRemoteRouteID(resourceId, pref)
}
// todo: consider to make sense of this notification or not
d.notifyPeerListChanged()
return nil
@@ -317,11 +385,26 @@ func (d *Status) RemovePeerStateRoute(peer string, route string) error {
peerState.DeleteRoute(route)
d.peers[peer] = peerState
pref, err := netip.ParsePrefix(route)
if err == nil {
d.routeIDLookup.RemoveRemoteRouteID(pref)
}
// todo: consider to make sense of this notification or not
d.notifyPeerListChanged()
return nil
}
// CheckRoutes checks if the source and destination addresses are within the same route
// and returns the resource ID of the route that contains the addresses
func (d *Status) CheckRoutes(ip netip.Addr) ([]byte, bool) {
if d == nil {
return nil, false
}
resId, isExitNode := d.routeIDLookup.Lookup(ip)
return []byte(resId), isExitNode
}
func (d *Status) UpdatePeerICEState(receivedState State) error {
d.mux.Lock()
defer d.mux.Unlock()
@@ -331,11 +414,8 @@ func (d *Status) UpdatePeerICEState(receivedState State) error {
return errors.New("peer doesn't exist")
}
if receivedState.IP != "" {
peerState.IP = receivedState.IP
}
skipNotification := shouldSkipNotify(receivedState.ConnStatus, peerState)
oldState := peerState.ConnStatus
oldIsRelayed := peerState.Relayed
peerState.ConnStatus = receivedState.ConnStatus
peerState.ConnStatusUpdate = receivedState.ConnStatusUpdate
@@ -348,12 +428,13 @@ func (d *Status) UpdatePeerICEState(receivedState State) error {
d.peers[receivedState.PubKey] = peerState
if skipNotification {
return nil
if hasConnStatusChanged(oldState, receivedState.ConnStatus) {
d.notifyPeerListChanged()
}
d.notifyPeerStateChangeListeners(receivedState.PubKey)
d.notifyPeerListChanged()
if hasStatusOrRelayedChange(oldState, receivedState.ConnStatus, oldIsRelayed, receivedState.Relayed) {
d.notifyPeerStateChangeListeners(receivedState.PubKey)
}
return nil
}
@@ -366,7 +447,8 @@ func (d *Status) UpdatePeerRelayedState(receivedState State) error {
return errors.New("peer doesn't exist")
}
skipNotification := shouldSkipNotify(receivedState.ConnStatus, peerState)
oldState := peerState.ConnStatus
oldIsRelayed := peerState.Relayed
peerState.ConnStatus = receivedState.ConnStatus
peerState.ConnStatusUpdate = receivedState.ConnStatusUpdate
@@ -376,12 +458,13 @@ func (d *Status) UpdatePeerRelayedState(receivedState State) error {
d.peers[receivedState.PubKey] = peerState
if skipNotification {
return nil
if hasConnStatusChanged(oldState, receivedState.ConnStatus) {
d.notifyPeerListChanged()
}
d.notifyPeerStateChangeListeners(receivedState.PubKey)
d.notifyPeerListChanged()
if hasStatusOrRelayedChange(oldState, receivedState.ConnStatus, oldIsRelayed, receivedState.Relayed) {
d.notifyPeerStateChangeListeners(receivedState.PubKey)
}
return nil
}
@@ -394,7 +477,8 @@ func (d *Status) UpdatePeerRelayedStateToDisconnected(receivedState State) error
return errors.New("peer doesn't exist")
}
skipNotification := shouldSkipNotify(receivedState.ConnStatus, peerState)
oldState := peerState.ConnStatus
oldIsRelayed := peerState.Relayed
peerState.ConnStatus = receivedState.ConnStatus
peerState.Relayed = receivedState.Relayed
@@ -403,12 +487,13 @@ func (d *Status) UpdatePeerRelayedStateToDisconnected(receivedState State) error
d.peers[receivedState.PubKey] = peerState
if skipNotification {
return nil
if hasConnStatusChanged(oldState, receivedState.ConnStatus) {
d.notifyPeerListChanged()
}
d.notifyPeerStateChangeListeners(receivedState.PubKey)
d.notifyPeerListChanged()
if hasStatusOrRelayedChange(oldState, receivedState.ConnStatus, oldIsRelayed, receivedState.Relayed) {
d.notifyPeerStateChangeListeners(receivedState.PubKey)
}
return nil
}
@@ -421,7 +506,8 @@ func (d *Status) UpdatePeerICEStateToDisconnected(receivedState State) error {
return errors.New("peer doesn't exist")
}
skipNotification := shouldSkipNotify(receivedState.ConnStatus, peerState)
oldState := peerState.ConnStatus
oldIsRelayed := peerState.Relayed
peerState.ConnStatus = receivedState.ConnStatus
peerState.Relayed = receivedState.Relayed
@@ -433,12 +519,13 @@ func (d *Status) UpdatePeerICEStateToDisconnected(receivedState State) error {
d.peers[receivedState.PubKey] = peerState
if skipNotification {
return nil
if hasConnStatusChanged(oldState, receivedState.ConnStatus) {
d.notifyPeerListChanged()
}
d.notifyPeerStateChangeListeners(receivedState.PubKey)
d.notifyPeerListChanged()
if hasStatusOrRelayedChange(oldState, receivedState.ConnStatus, oldIsRelayed, receivedState.Relayed) {
d.notifyPeerStateChangeListeners(receivedState.PubKey)
}
return nil
}
@@ -461,17 +548,12 @@ func (d *Status) UpdateWireGuardPeerState(pubKey string, wgStats configurer.WGSt
return nil
}
func shouldSkipNotify(receivedConnStatus ConnStatus, curr State) bool {
switch {
case receivedConnStatus == StatusConnecting:
return true
case receivedConnStatus == StatusDisconnected && curr.ConnStatus == StatusConnecting:
return true
case receivedConnStatus == StatusDisconnected && curr.ConnStatus == StatusDisconnected:
return curr.IP != ""
default:
return false
}
func hasStatusOrRelayedChange(oldConnStatus, newConnStatus ConnStatus, oldRelayed, newRelayed bool) bool {
return oldRelayed != newRelayed || hasConnStatusChanged(newConnStatus, oldConnStatus)
}
func hasConnStatusChanged(oldStatus, newStatus ConnStatus) bool {
return newStatus != oldStatus
}
// UpdatePeerFQDN update peer's state fqdn only
@@ -493,30 +575,55 @@ func (d *Status) UpdatePeerFQDN(peerPubKey, fqdn string) error {
// FinishPeerListModifications this event invoke the notification
func (d *Status) FinishPeerListModifications() {
d.mux.Lock()
defer d.mux.Unlock()
if !d.peerListChangedForNotification {
d.mux.Unlock()
return
}
d.peerListChangedForNotification = false
d.mux.Unlock()
d.notifyPeerListChanged()
for key := range d.peers {
d.notifyPeerStateChangeListeners(key)
}
}
// GetPeerStateChangeNotifier returns a change notifier channel for a peer
func (d *Status) GetPeerStateChangeNotifier(peer string) <-chan struct{} {
func (d *Status) SubscribeToPeerStateChanges(ctx context.Context, peerID string) *StatusChangeSubscription {
d.mux.Lock()
defer d.mux.Unlock()
ch, found := d.changeNotify[peer]
if found {
return ch
sub := newStatusChangeSubscription(ctx, peerID)
if _, ok := d.changeNotify[peerID]; !ok {
d.changeNotify[peerID] = make(map[string]*StatusChangeSubscription)
}
d.changeNotify[peerID][sub.id] = sub
return sub
}
func (d *Status) UnsubscribePeerStateChanges(subscription *StatusChangeSubscription) {
d.mux.Lock()
defer d.mux.Unlock()
if subscription == nil {
return
}
ch = make(chan struct{})
d.changeNotify[peer] = ch
return ch
channels, ok := d.changeNotify[subscription.peerID]
if !ok {
return
}
sub, exists := channels[subscription.id]
if !exists {
return
}
delete(channels, subscription.id)
if len(channels) == 0 {
delete(d.changeNotify, sub.peerID)
}
}
// GetLocalPeerState returns the local peer state
@@ -535,6 +642,63 @@ func (d *Status) UpdateLocalPeerState(localPeerState LocalPeerState) {
d.notifyAddressChanged()
}
// AddLocalPeerStateRoute adds a route to the local peer state
func (d *Status) AddLocalPeerStateRoute(route string, resourceId route.ResID) {
d.mux.Lock()
defer d.mux.Unlock()
pref, err := netip.ParsePrefix(route)
if err == nil {
d.routeIDLookup.AddLocalRouteID(resourceId, pref)
}
if d.localPeer.Routes == nil {
d.localPeer.Routes = map[string]struct{}{}
}
d.localPeer.Routes[route] = struct{}{}
}
// RemoveLocalPeerStateRoute removes a route from the local peer state
func (d *Status) RemoveLocalPeerStateRoute(route string) {
d.mux.Lock()
defer d.mux.Unlock()
pref, err := netip.ParsePrefix(route)
if err == nil {
d.routeIDLookup.RemoveLocalRouteID(pref)
}
delete(d.localPeer.Routes, route)
}
// AddResolvedIPLookupEntry adds a resolved IP lookup entry
func (d *Status) AddResolvedIPLookupEntry(prefix netip.Prefix, resourceId route.ResID) {
d.mux.Lock()
defer d.mux.Unlock()
d.routeIDLookup.AddResolvedIP(resourceId, prefix)
}
// RemoveResolvedIPLookupEntry removes a resolved IP lookup entry
func (d *Status) RemoveResolvedIPLookupEntry(route string) {
d.mux.Lock()
defer d.mux.Unlock()
pref, err := netip.ParsePrefix(route)
if err == nil {
d.routeIDLookup.RemoveResolvedIP(pref)
}
}
// CleanLocalPeerStateRoutes cleans all routes from the local peer state
func (d *Status) CleanLocalPeerStateRoutes() {
d.mux.Lock()
defer d.mux.Unlock()
d.localPeer.Routes = map[string]struct{}{}
}
// CleanLocalPeerState cleans local peer status
func (d *Status) CleanLocalPeerState() {
d.mux.Lock()
@@ -586,6 +750,12 @@ func (d *Status) UpdateRosenpass(rosenpassEnabled, rosenpassPermissive bool) {
d.rosenpassEnabled = rosenpassEnabled
}
func (d *Status) UpdateLazyConnection(enabled bool) {
d.mux.Lock()
defer d.mux.Unlock()
d.lazyConnectionEnabled = enabled
}
// MarkSignalDisconnected sets SignalState to disconnected
func (d *Status) MarkSignalDisconnected(err error) {
d.mux.Lock()
@@ -618,7 +788,7 @@ func (d *Status) UpdateDNSStates(dnsStates []NSGroupState) {
d.nsGroupStates = dnsStates
}
func (d *Status) UpdateResolvedDomainsStates(originalDomain domain.Domain, resolvedDomain domain.Domain, prefixes []netip.Prefix) {
func (d *Status) UpdateResolvedDomainsStates(originalDomain domain.Domain, resolvedDomain domain.Domain, prefixes []netip.Prefix, resourceId route.ResID) {
d.mux.Lock()
defer d.mux.Unlock()
@@ -627,6 +797,10 @@ func (d *Status) UpdateResolvedDomainsStates(originalDomain domain.Domain, resol
Prefixes: prefixes,
ParentDomain: originalDomain,
}
for _, prefix := range prefixes {
d.routeIDLookup.AddResolvedIP(resourceId, prefix)
}
}
func (d *Status) DeleteResolvedDomainsStates(domain domain.Domain) {
@@ -637,6 +811,10 @@ func (d *Status) DeleteResolvedDomainsStates(domain domain.Domain) {
for k, v := range d.resolvedDomainsStates {
if v.ParentDomain == domain {
delete(d.resolvedDomainsStates, k)
for _, prefix := range v.Prefixes {
d.routeIDLookup.RemoveResolvedIP(prefix)
}
}
}
}
@@ -650,6 +828,12 @@ func (d *Status) GetRosenpassState() RosenpassState {
}
}
func (d *Status) GetLazyConnection() bool {
d.mux.Lock()
defer d.mux.Unlock()
return d.lazyConnectionEnabled
}
func (d *Status) GetManagementState() ManagementState {
d.mux.Lock()
defer d.mux.Unlock()
@@ -734,6 +918,16 @@ func (d *Status) GetRelayStates() []relay.ProbeResult {
return append(relayStates, relayState)
}
func (d *Status) ForwardingRules() []firewall.ForwardRule {
d.mux.Lock()
defer d.mux.Unlock()
if d.ingressGwMgr == nil {
return nil
}
return d.ingressGwMgr.Rules()
}
func (d *Status) GetDNSStates() []NSGroupState {
d.mux.Lock()
defer d.mux.Unlock()
@@ -751,11 +945,13 @@ func (d *Status) GetResolvedDomainsStates() map[domain.Domain]ResolvedDomainInfo
// GetFullStatus gets full status
func (d *Status) GetFullStatus() FullStatus {
fullStatus := FullStatus{
ManagementState: d.GetManagementState(),
SignalState: d.GetSignalState(),
Relays: d.GetRelayStates(),
RosenpassState: d.GetRosenpassState(),
NSGroupStates: d.GetDNSStates(),
ManagementState: d.GetManagementState(),
SignalState: d.GetSignalState(),
Relays: d.GetRelayStates(),
RosenpassState: d.GetRosenpassState(),
NSGroupStates: d.GetDNSStates(),
NumOfForwardingRules: len(d.ForwardingRules()),
LazyConnectionEnabled: d.GetLazyConnection(),
}
d.mux.Lock()
@@ -802,13 +998,33 @@ func (d *Status) onConnectionChanged() {
// notifyPeerStateChangeListeners notifies route manager about the change in peer state
func (d *Status) notifyPeerStateChangeListeners(peerID string) {
ch, found := d.changeNotify[peerID]
if !found {
subs, ok := d.changeNotify[peerID]
if !ok {
return
}
close(ch)
delete(d.changeNotify, peerID)
// collect the relevant data for router peers
routerPeers := make(map[string]RouterState, len(d.changeNotify))
for pid := range d.changeNotify {
s, ok := d.peers[pid]
if !ok {
log.Warnf("router peer not found in peers list: %s", pid)
continue
}
routerPeers[pid] = RouterState{
Status: s.ConnStatus,
Relayed: s.Relayed,
Latency: s.Latency,
}
}
for _, sub := range subs {
select {
case sub.eventsChan <- routerPeers:
case <-sub.ctx.Done():
}
}
}
func (d *Status) notifyPeerListChanged() {
@@ -892,6 +1108,23 @@ func (d *Status) GetEventHistory() []*proto.SystemEvent {
return d.eventQueue.GetAll()
}
func (d *Status) SetWgIface(wgInterface WGIfaceStatus) {
d.mux.Lock()
defer d.mux.Unlock()
d.wgIface = wgInterface
}
func (d *Status) PeersStatus() (*configurer.Stats, error) {
d.mux.Lock()
defer d.mux.Unlock()
if d.wgIface == nil {
return nil, fmt.Errorf("wgInterface is nil, cannot retrieve peers status")
}
return d.wgIface.FullStats()
}
type EventQueue struct {
maxSize int
events []*proto.SystemEvent
+27 -21
View File
@@ -1,31 +1,35 @@
package peer
import (
"context"
"errors"
"sync"
"testing"
"time"
"github.com/stretchr/testify/assert"
)
func TestAddPeer(t *testing.T) {
key := "abc"
ip := "100.108.254.1"
status := NewRecorder("https://mgm")
err := status.AddPeer(key, "abc.netbird")
err := status.AddPeer(key, "abc.netbird", ip)
assert.NoError(t, err, "shouldn't return error")
_, exists := status.peers[key]
assert.True(t, exists, "value was found")
err = status.AddPeer(key, "abc.netbird")
err = status.AddPeer(key, "abc.netbird", ip)
assert.Error(t, err, "should return error on duplicate")
}
func TestGetPeer(t *testing.T) {
key := "abc"
ip := "100.108.254.1"
status := NewRecorder("https://mgm")
err := status.AddPeer(key, "abc.netbird")
err := status.AddPeer(key, "abc.netbird", ip)
assert.NoError(t, err, "shouldn't return error")
peerStatus, err := status.GetPeer(key)
@@ -40,16 +44,16 @@ func TestGetPeer(t *testing.T) {
func TestUpdatePeerState(t *testing.T) {
key := "abc"
ip := "10.10.10.10"
fqdn := "peer-a.netbird.local"
status := NewRecorder("https://mgm")
_ = status.AddPeer(key, fqdn, ip)
peerState := State{
PubKey: key,
Mux: new(sync.RWMutex),
PubKey: key,
ConnStatusUpdate: time.Now(),
ConnStatus: StatusConnecting,
}
status.peers[key] = peerState
peerState.IP = ip
err := status.UpdatePeerState(peerState)
assert.NoError(t, err, "shouldn't return error")
@@ -81,25 +85,27 @@ func TestGetPeerStateChangeNotifierLogic(t *testing.T) {
key := "abc"
ip := "10.10.10.10"
status := NewRecorder("https://mgm")
_ = status.AddPeer(key, "abc.netbird", ip)
sub := status.SubscribeToPeerStateChanges(context.Background(), key)
assert.NotNil(t, sub, "channel shouldn't be nil")
peerState := State{
PubKey: key,
Mux: new(sync.RWMutex),
PubKey: key,
ConnStatus: StatusConnecting,
Relayed: false,
ConnStatusUpdate: time.Now(),
}
status.peers[key] = peerState
ch := status.GetPeerStateChangeNotifier(key)
assert.NotNil(t, ch, "channel shouldn't be nil")
peerState.IP = ip
err := status.UpdatePeerRelayedStateToDisconnected(peerState)
assert.NoError(t, err, "shouldn't return error")
timeoutCtx, cancel := context.WithTimeout(context.Background(), 1*time.Second)
defer cancel()
select {
case <-ch:
default:
t.Errorf("channel wasn't closed after update")
case <-sub.eventsChan:
case <-timeoutCtx.Done():
t.Errorf("timed out waiting for event")
}
}
+14 -9
View File
@@ -2,6 +2,7 @@ package peer
import (
"context"
"fmt"
"sync"
"time"
@@ -20,25 +21,26 @@ var (
)
type WGInterfaceStater interface {
GetStats(key string) (configurer.WGStats, error)
GetStats() (map[string]configurer.WGStats, error)
}
type WGWatcher struct {
log *log.Entry
wgIfaceStater WGInterfaceStater
peerKey string
stateDump *stateDump
ctx context.Context
ctxCancel context.CancelFunc
ctxLock sync.Mutex
waitGroup sync.WaitGroup
}
func NewWGWatcher(log *log.Entry, wgIfaceStater WGInterfaceStater, peerKey string) *WGWatcher {
func NewWGWatcher(log *log.Entry, wgIfaceStater WGInterfaceStater, peerKey string, stateDump *stateDump) *WGWatcher {
return &WGWatcher{
log: log,
wgIfaceStater: wgIfaceStater,
peerKey: peerKey,
stateDump: stateDump,
}
}
@@ -46,24 +48,24 @@ func NewWGWatcher(log *log.Entry, wgIfaceStater WGInterfaceStater, peerKey strin
func (w *WGWatcher) EnableWgWatcher(parentCtx context.Context, onDisconnectedFn func()) {
w.log.Debugf("enable WireGuard watcher")
w.ctxLock.Lock()
defer w.ctxLock.Unlock()
if w.ctx != nil && w.ctx.Err() == nil {
w.log.Errorf("WireGuard watcher already enabled")
w.ctxLock.Unlock()
return
}
ctx, ctxCancel := context.WithCancel(parentCtx)
w.ctx = ctx
w.ctxCancel = ctxCancel
w.ctxLock.Unlock()
initialHandshake, err := w.wgState()
if err != nil {
w.log.Warnf("failed to read initial wg stats: %v", err)
}
w.waitGroup.Add(1)
go w.periodicHandshakeCheck(ctx, ctxCancel, onDisconnectedFn, initialHandshake)
w.periodicHandshakeCheck(ctx, ctxCancel, onDisconnectedFn, initialHandshake)
}
// DisableWgWatcher stops the WireGuard watcher and wait for the watcher to exit
@@ -79,13 +81,11 @@ func (w *WGWatcher) DisableWgWatcher() {
w.ctxCancel()
w.ctxCancel = nil
w.waitGroup.Wait()
}
// wgStateCheck help to check the state of the WireGuard handshake and relay connection
func (w *WGWatcher) periodicHandshakeCheck(ctx context.Context, ctxCancel context.CancelFunc, onDisconnectedFn func(), initialHandshake time.Time) {
w.log.Infof("WireGuard watcher started")
defer w.waitGroup.Done()
timer := time.NewTimer(wgHandshakeOvertime)
defer timer.Stop()
@@ -105,6 +105,7 @@ func (w *WGWatcher) periodicHandshakeCheck(ctx context.Context, ctxCancel contex
resetTime := time.Until(handshake.Add(checkPeriod))
timer.Reset(resetTime)
w.stateDump.WGcheckSuccess()
w.log.Debugf("WireGuard watcher reset timer: %v", resetTime)
case <-ctx.Done():
@@ -146,9 +147,13 @@ func (w *WGWatcher) handshakeCheck(lastHandshake time.Time) (*time.Time, bool) {
}
func (w *WGWatcher) wgState() (time.Time, error) {
wgState, err := w.wgIfaceStater.GetStats(w.peerKey)
wgStates, err := w.wgIfaceStater.GetStats()
if err != nil {
return time.Time{}, err
}
wgState, ok := wgStates[w.peerKey]
if !ok {
return time.Time{}, fmt.Errorf("peer %s not found in WireGuard endpoints", w.peerKey)
}
return wgState.LastHandshake, nil
}
+9 -23
View File
@@ -11,26 +11,11 @@ import (
)
type MocWgIface struct {
initial bool
lastHandshake time.Time
stop bool
stop bool
}
func (m *MocWgIface) GetStats(key string) (configurer.WGStats, error) {
if !m.initial {
m.initial = true
return configurer.WGStats{}, nil
}
if !m.stop {
m.lastHandshake = time.Now()
}
stats := configurer.WGStats{
LastHandshake: m.lastHandshake,
}
return stats, nil
func (m *MocWgIface) GetStats() (map[string]configurer.WGStats, error) {
return map[string]configurer.WGStats{}, nil
}
func (m *MocWgIface) disconnect() {
@@ -43,13 +28,13 @@ func TestWGWatcher_EnableWgWatcher(t *testing.T) {
mlog := log.WithField("peer", "tet")
mocWgIface := &MocWgIface{}
watcher := NewWGWatcher(mlog, mocWgIface, "")
watcher := NewWGWatcher(mlog, mocWgIface, "", newStateDump("peer", mlog, &Status{}))
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
onDisconnected := make(chan struct{}, 1)
watcher.EnableWgWatcher(ctx, func() {
go watcher.EnableWgWatcher(ctx, func() {
mlog.Infof("onDisconnectedFn")
onDisconnected <- struct{}{}
})
@@ -72,17 +57,18 @@ func TestWGWatcher_ReEnable(t *testing.T) {
mlog := log.WithField("peer", "tet")
mocWgIface := &MocWgIface{}
watcher := NewWGWatcher(mlog, mocWgIface, "")
watcher := NewWGWatcher(mlog, mocWgIface, "", newStateDump("peer", mlog, &Status{}))
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
onDisconnected := make(chan struct{}, 1)
watcher.EnableWgWatcher(ctx, func() {})
go watcher.EnableWgWatcher(ctx, func() {})
time.Sleep(1 * time.Second)
watcher.DisableWgWatcher()
watcher.EnableWgWatcher(ctx, func() {
go watcher.EnableWgWatcher(ctx, func() {
onDisconnected <- struct{}{}
})
+55
View File
@@ -0,0 +1,55 @@
package worker
import (
"sync/atomic"
log "github.com/sirupsen/logrus"
)
const (
StatusDisconnected Status = iota
StatusConnected
)
type Status int32
func (s Status) String() string {
switch s {
case StatusDisconnected:
return "Disconnected"
case StatusConnected:
return "Connected"
default:
log.Errorf("unknown status: %d", s)
return "unknown"
}
}
// AtomicWorkerStatus is a thread-safe wrapper for worker status
type AtomicWorkerStatus struct {
status atomic.Int32
}
func NewAtomicStatus() *AtomicWorkerStatus {
acs := &AtomicWorkerStatus{}
acs.SetDisconnected()
return acs
}
// Get returns the current connection status
func (acs *AtomicWorkerStatus) Get() Status {
return Status(acs.status.Load())
}
func (acs *AtomicWorkerStatus) SetConnected() {
acs.status.Store(int32(StatusConnected))
}
func (acs *AtomicWorkerStatus) SetDisconnected() {
acs.status.Store(int32(StatusDisconnected))
}
// String returns the string representation of the current status
func (acs *AtomicWorkerStatus) String() string {
return acs.Get().String()
}
+233 -129
View File
@@ -8,12 +8,13 @@ import (
"sync"
"time"
"github.com/pion/ice/v3"
"github.com/pion/ice/v4"
"github.com/pion/stun/v2"
log "github.com/sirupsen/logrus"
"github.com/netbirdio/netbird/client/iface"
"github.com/netbirdio/netbird/client/iface/bind"
"github.com/netbirdio/netbird/client/internal/peer/conntype"
icemaker "github.com/netbirdio/netbird/client/internal/peer/ice"
"github.com/netbirdio/netbird/client/internal/stdnet"
"github.com/netbirdio/netbird/route"
@@ -41,8 +42,18 @@ type WorkerICE struct {
statusRecorder *Status
hasRelayOnLocally bool
agent *ice.Agent
muxAgent sync.Mutex
agent *icemaker.ThreadSafeAgent
agentDialerCancel context.CancelFunc
agentConnecting bool // while it is true, drop all incoming offers
lastSuccess time.Time // with this avoid the too frequent ICE agent recreation
// remoteSessionID represents the peer's session identifier from the latest remote offer.
remoteSessionID ICESessionID
// sessionID is used to track the current session ID of the ICE agent
// increase by one when disconnecting the agent
// with it the remote peer can discard the already deprecated offer/answer
// Without it the remote peer may recreate a workable ICE connection
sessionID ICESessionID
muxAgent sync.Mutex
StunTurn []*stun.URI
@@ -56,6 +67,11 @@ type WorkerICE struct {
}
func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *Conn, signaler *Signaler, ifaceDiscover stdnet.ExternalIFaceDiscover, statusRecorder *Status, hasRelayOnLocally bool) (*WorkerICE, error) {
sessionID, err := NewICESessionID()
if err != nil {
return nil, err
}
w := &WorkerICE{
ctx: ctx,
log: log,
@@ -65,6 +81,8 @@ func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *
iFaceDiscover: ifaceDiscover,
statusRecorder: statusRecorder,
hasRelayOnLocally: hasRelayOnLocally,
lastKnownState: ice.ConnectionStateDisconnected,
sessionID: sessionID,
}
localUfrag, localPwd, err := icemaker.GenerateICECredentials()
@@ -77,15 +95,36 @@ func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *
}
func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
w.log.Debugf("OnNewOffer for ICE")
w.log.Debugf("OnNewOffer for ICE, serial: %s", remoteOfferAnswer.SessionIDString())
w.muxAgent.Lock()
if w.agent != nil {
w.log.Debugf("agent already exists, skipping the offer")
if w.agentConnecting {
w.log.Debugf("agent connection is in progress, skipping the offer")
w.muxAgent.Unlock()
return
}
if w.agent != nil {
// backward compatibility with old clients that do not send session ID
if remoteOfferAnswer.SessionID == nil {
w.log.Debugf("agent already exists, skipping the offer")
w.muxAgent.Unlock()
return
}
if w.remoteSessionID == *remoteOfferAnswer.SessionID {
w.log.Debugf("agent already exists and session ID matches, skipping the offer: %s", remoteOfferAnswer.SessionIDString())
w.muxAgent.Unlock()
return
}
w.log.Debugf("agent already exists, recreate the connection")
w.agentDialerCancel()
if err := w.agent.Close(); err != nil {
w.log.Warnf("failed to close ICE agent: %s", err)
}
w.agent = nil
// todo consider to switch to Relay connection while establishing a new ICE connection
}
var preferredCandidateTypes []ice.CandidateType
if w.hasRelayOnLocally && remoteOfferAnswer.RelaySrvAddress != "" {
preferredCandidateTypes = icemaker.CandidateTypesP2P()
@@ -94,36 +133,125 @@ func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
}
w.log.Debugf("recreate ICE agent")
agentCtx, agentCancel := context.WithCancel(w.ctx)
agent, err := w.reCreateAgent(agentCancel, preferredCandidateTypes)
dialerCtx, dialerCancel := context.WithCancel(w.ctx)
agent, err := w.reCreateAgent(dialerCancel, preferredCandidateTypes)
if err != nil {
w.log.Errorf("failed to recreate ICE Agent: %s", err)
w.muxAgent.Unlock()
return
}
w.sentExtraSrflx = false
w.agent = agent
w.agentDialerCancel = dialerCancel
w.agentConnecting = true
w.muxAgent.Unlock()
go w.connect(dialerCtx, agent, remoteOfferAnswer)
}
// OnRemoteCandidate Handles ICE connection Candidate provided by the remote peer.
func (w *WorkerICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMap) {
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
w.log.Debugf("OnRemoteCandidate from peer %s -> %s", w.config.Key, candidate.String())
if w.agent == nil {
w.log.Warnf("ICE Agent is not initialized yet")
return
}
if candidateViaRoutes(candidate, haRoutes) {
return
}
if err := w.agent.AddRemoteCandidate(candidate); err != nil {
w.log.Errorf("error while handling remote candidate")
return
}
}
func (w *WorkerICE) GetLocalUserCredentials() (frag string, pwd string) {
return w.localUfrag, w.localPwd
}
func (w *WorkerICE) InProgress() bool {
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
return w.agentConnecting
}
func (w *WorkerICE) Close() {
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
if w.agent == nil {
return
}
w.agentDialerCancel()
if err := w.agent.Close(); err != nil {
w.log.Warnf("failed to close ICE agent: %s", err)
}
w.agent = nil
}
func (w *WorkerICE) reCreateAgent(dialerCancel context.CancelFunc, candidates []ice.CandidateType) (*icemaker.ThreadSafeAgent, error) {
agent, err := icemaker.NewAgent(w.iFaceDiscover, w.config.ICEConfig, candidates, w.localUfrag, w.localPwd)
if err != nil {
return nil, fmt.Errorf("create agent: %w", err)
}
if err := agent.OnCandidate(w.onICECandidate); err != nil {
return nil, err
}
if err := agent.OnConnectionStateChange(w.onConnectionStateChange(agent, dialerCancel)); err != nil {
return nil, err
}
if err := agent.OnSelectedCandidatePairChange(w.onICESelectedCandidatePair); err != nil {
return nil, err
}
return agent, nil
}
func (w *WorkerICE) SessionID() ICESessionID {
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
return w.sessionID
}
// will block until connection succeeded
// but it won't release if ICE Agent went into Disconnected or Failed state,
// so we have to cancel it with the provided context once agent detected a broken connection
func (w *WorkerICE) connect(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) {
w.log.Debugf("gather candidates")
err = w.agent.GatherCandidates()
if err != nil {
w.log.Debugf("failed to gather candidates: %s", err)
if err := agent.GatherCandidates(); err != nil {
w.log.Warnf("failed to gather candidates: %s", err)
w.closeAgent(agent, w.agentDialerCancel)
return
}
// will block until connection succeeded
// but it won't release if ICE Agent went into Disconnected or Failed state,
// so we have to cancel it with the provided context once agent detected a broken connection
w.log.Debugf("turn agent dial")
remoteConn, err := w.turnAgentDial(agentCtx, remoteOfferAnswer)
remoteConn, err := w.turnAgentDial(ctx, agent, remoteOfferAnswer)
if err != nil {
w.log.Debugf("failed to dial the remote peer: %s", err)
w.closeAgent(agent, w.agentDialerCancel)
return
}
w.log.Debugf("agent dial succeeded")
pair, err := w.agent.GetSelectedCandidatePair()
pair, err := agent.GetSelectedCandidatePair()
if err != nil {
w.closeAgent(agent, w.agentDialerCancel)
return
}
if pair == nil {
w.log.Warnf("selected candidate pair is nil, cannot proceed")
w.closeAgent(agent, w.agentDialerCancel)
return
}
@@ -150,114 +278,39 @@ func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
RelayedOnLocal: isRelayCandidate(pair.Local),
}
w.log.Debugf("on ICE conn is ready to use")
go w.conn.onICEConnectionIsReady(selectedPriority(pair), ci)
}
// OnRemoteCandidate Handles ICE connection Candidate provided by the remote peer.
func (w *WorkerICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMap) {
w.log.Infof("connection succeeded with offer session: %s", remoteOfferAnswer.SessionIDString())
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
w.log.Debugf("OnRemoteCandidate from peer %s -> %s", w.config.Key, candidate.String())
if w.agent == nil {
w.log.Warnf("ICE Agent is not initialized yet")
return
w.agentConnecting = false
w.lastSuccess = time.Now()
if remoteOfferAnswer.SessionID != nil {
w.remoteSessionID = *remoteOfferAnswer.SessionID
}
w.muxAgent.Unlock()
if candidateViaRoutes(candidate, haRoutes) {
return
}
err := w.agent.AddRemoteCandidate(candidate)
if err != nil {
w.log.Errorf("error while handling remote candidate")
return
}
// todo: the potential problem is a race between the onConnectionStateChange
w.conn.onICEConnectionIsReady(selectedPriority(pair), ci)
}
func (w *WorkerICE) GetLocalUserCredentials() (frag string, pwd string) {
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
return w.localUfrag, w.localPwd
}
func (w *WorkerICE) Close() {
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
if w.agent == nil {
return
}
if err := w.agent.Close(); err != nil {
w.log.Warnf("failed to close ICE agent: %s", err)
}
}
func (w *WorkerICE) reCreateAgent(agentCancel context.CancelFunc, candidates []ice.CandidateType) (*ice.Agent, error) {
w.sentExtraSrflx = false
agent, err := icemaker.NewAgent(w.iFaceDiscover, w.config.ICEConfig, candidates, w.localUfrag, w.localPwd)
if err != nil {
return nil, fmt.Errorf("create agent: %w", err)
}
err = agent.OnCandidate(w.onICECandidate)
if err != nil {
return nil, err
}
err = agent.OnConnectionStateChange(func(state ice.ConnectionState) {
w.log.Debugf("ICE ConnectionState has changed to %s", state.String())
switch state {
case ice.ConnectionStateConnected:
w.lastKnownState = ice.ConnectionStateConnected
return
case ice.ConnectionStateFailed, ice.ConnectionStateDisconnected:
if w.lastKnownState != ice.ConnectionStateDisconnected {
w.lastKnownState = ice.ConnectionStateDisconnected
w.conn.onICEStateDisconnected()
}
w.closeAgent(agentCancel)
default:
return
}
})
if err != nil {
return nil, err
}
err = agent.OnSelectedCandidatePairChange(w.onICESelectedCandidatePair)
if err != nil {
return nil, err
}
err = agent.OnSuccessfulSelectedPairBindingResponse(func(p *ice.CandidatePair) {
err := w.statusRecorder.UpdateLatency(w.config.Key, p.Latency())
if err != nil {
w.log.Debugf("failed to update latency for peer: %s", err)
return
}
})
if err != nil {
return nil, fmt.Errorf("failed setting binding response callback: %w", err)
}
return agent, nil
}
func (w *WorkerICE) closeAgent(cancel context.CancelFunc) {
w.muxAgent.Lock()
defer w.muxAgent.Unlock()
func (w *WorkerICE) closeAgent(agent *icemaker.ThreadSafeAgent, cancel context.CancelFunc) {
cancel()
if w.agent == nil {
return
}
if err := w.agent.Close(); err != nil {
if err := agent.Close(); err != nil {
w.log.Warnf("failed to close ICE agent: %s", err)
}
w.agent = nil
w.muxAgent.Lock()
// todo review does it make sense to generate new session ID all the time when w.agent==agent
sessionID, err := NewICESessionID()
if err != nil {
w.log.Errorf("failed to create new session ID: %s", err)
}
w.sessionID = sessionID
if w.agent == agent {
w.agent = nil
w.agentConnecting = false
}
w.muxAgent.Unlock()
}
func (w *WorkerICE) punchRemoteWGPort(pair *ice.CandidatePair, remoteWgPort int) {
@@ -327,6 +380,46 @@ func (w *WorkerICE) onICECandidate(candidate ice.Candidate) {
func (w *WorkerICE) onICESelectedCandidatePair(c1 ice.Candidate, c2 ice.Candidate) {
w.log.Debugf("selected candidate pair [local <-> remote] -> [%s <-> %s], peer %s", c1.String(), c2.String(),
w.config.Key)
w.muxAgent.Lock()
pair, err := w.agent.GetSelectedCandidatePair()
if err != nil {
w.log.Warnf("failed to get selected candidate pair: %s", err)
w.muxAgent.Unlock()
return
}
if pair == nil {
w.log.Warnf("selected candidate pair is nil, cannot proceed")
w.muxAgent.Unlock()
return
}
w.muxAgent.Unlock()
duration := time.Duration(pair.CurrentRoundTripTime() * float64(time.Second))
if err := w.statusRecorder.UpdateLatency(w.config.Key, duration); err != nil {
w.log.Debugf("failed to update latency for peer: %s", err)
return
}
}
func (w *WorkerICE) onConnectionStateChange(agent *icemaker.ThreadSafeAgent, dialerCancel context.CancelFunc) func(ice.ConnectionState) {
return func(state ice.ConnectionState) {
w.log.Debugf("ICE ConnectionState has changed to %s", state.String())
switch state {
case ice.ConnectionStateConnected:
w.lastKnownState = ice.ConnectionStateConnected
return
case ice.ConnectionStateFailed, ice.ConnectionStateDisconnected:
if w.lastKnownState == ice.ConnectionStateConnected {
w.lastKnownState = ice.ConnectionStateDisconnected
w.conn.onICEStateDisconnected()
}
w.closeAgent(agent, dialerCancel)
default:
return
}
}
}
func (w *WorkerICE) shouldSendExtraSrflxCandidate(candidate ice.Candidate) bool {
@@ -336,18 +429,18 @@ func (w *WorkerICE) shouldSendExtraSrflxCandidate(candidate ice.Candidate) bool
return false
}
func (w *WorkerICE) turnAgentDial(ctx context.Context, remoteOfferAnswer *OfferAnswer) (*ice.Conn, error) {
func (w *WorkerICE) turnAgentDial(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) (*ice.Conn, error) {
isControlling := w.config.LocalKey > w.config.Key
if isControlling {
return w.agent.Dial(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
return agent.Dial(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
} else {
return w.agent.Accept(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
return agent.Accept(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
}
}
func extraSrflxCandidate(candidate ice.Candidate) (*ice.CandidateServerReflexive, error) {
relatedAdd := candidate.RelatedAddress()
return ice.NewCandidateServerReflexive(&ice.CandidateServerReflexiveConfig{
ec, err := ice.NewCandidateServerReflexive(&ice.CandidateServerReflexiveConfig{
Network: candidate.NetworkType().String(),
Address: candidate.Address(),
Port: relatedAdd.Port,
@@ -355,9 +448,26 @@ func extraSrflxCandidate(candidate ice.Candidate) (*ice.CandidateServerReflexive
RelAddr: relatedAdd.Address,
RelPort: relatedAdd.Port,
})
if err != nil {
return nil, err
}
for _, e := range candidate.Extensions() {
if err := ec.AddExtension(e); err != nil {
return nil, err
}
}
return ec, nil
}
func candidateViaRoutes(candidate ice.Candidate, clientRoutes route.HAMap) bool {
addr, err := netip.ParseAddr(candidate.Address())
if err != nil {
log.Errorf("Failed to parse IP address %s: %v", candidate.Address(), err)
return false
}
var routePrefixes []netip.Prefix
for _, routes := range clientRoutes {
if len(routes) > 0 && routes[0] != nil {
@@ -365,14 +475,8 @@ func candidateViaRoutes(candidate ice.Candidate, clientRoutes route.HAMap) bool
}
}
addr, err := netip.ParseAddr(candidate.Address())
if err != nil {
log.Errorf("Failed to parse IP address %s: %v", candidate.Address(), err)
return false
}
for _, prefix := range routePrefixes {
// default route is
// default route is handled by route exclusion / ip rules
if prefix.Bits() == 0 {
continue
}
@@ -396,10 +500,10 @@ func isRelayed(pair *ice.CandidatePair) bool {
return false
}
func selectedPriority(pair *ice.CandidatePair) ConnPriority {
func selectedPriority(pair *ice.CandidatePair) conntype.ConnPriority {
if isRelayed(pair) {
return connPriorityICETurn
return conntype.ICETurn
} else {
return connPriorityICEP2P
return conntype.ICEP2P
}
}
+7 -5
View File
@@ -9,7 +9,7 @@ import (
log "github.com/sirupsen/logrus"
relayClient "github.com/netbirdio/netbird/relay/client"
relayClient "github.com/netbirdio/netbird/shared/relay/client"
)
type RelayConnInfo struct {
@@ -19,11 +19,12 @@ type RelayConnInfo struct {
}
type WorkerRelay struct {
peerCtx context.Context
log *log.Entry
isController bool
config ConnConfig
conn *Conn
relayManager relayClient.ManagerService
relayManager *relayClient.Manager
relayedConn net.Conn
relayLock sync.Mutex
@@ -33,14 +34,15 @@ type WorkerRelay struct {
wgWatcher *WGWatcher
}
func NewWorkerRelay(log *log.Entry, ctrl bool, config ConnConfig, conn *Conn, relayManager relayClient.ManagerService) *WorkerRelay {
func NewWorkerRelay(ctx context.Context, log *log.Entry, ctrl bool, config ConnConfig, conn *Conn, relayManager *relayClient.Manager, stateDump *stateDump) *WorkerRelay {
r := &WorkerRelay{
peerCtx: ctx,
log: log,
isController: ctrl,
config: config,
conn: conn,
relayManager: relayManager,
wgWatcher: NewWGWatcher(log, config.WgConfig.WgInterface, config.Key),
wgWatcher: NewWGWatcher(log, config.WgConfig.WgInterface, config.Key, stateDump),
}
return r
}
@@ -62,7 +64,7 @@ func (w *WorkerRelay) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
srv := w.preferredRelayServer(currentRelayAddress, remoteOfferAnswer.RelaySrvAddress)
relayedConn, err := w.relayManager.OpenConn(srv, w.config.Key)
relayedConn, err := w.relayManager.OpenConn(w.peerCtx, srv, w.config.Key)
if err != nil {
if errors.Is(err, relayClient.ErrConnAlreadyExists) {
w.log.Debugf("handled offer by reusing existing relay connection")