123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463 |
- package nebula
- import (
- "context"
- "errors"
- "fmt"
- "io"
- "net/netip"
- "os"
- "runtime"
- "sync/atomic"
- "time"
- "github.com/gaissmai/bart"
- "github.com/rcrowley/go-metrics"
- "github.com/sirupsen/logrus"
- "github.com/slackhq/nebula/config"
- "github.com/slackhq/nebula/firewall"
- "github.com/slackhq/nebula/header"
- "github.com/slackhq/nebula/overlay"
- "github.com/slackhq/nebula/udp"
- )
- const mtu = 9001
- type InterfaceConfig struct {
- HostMap *HostMap
- Outside udp.Conn
- Inside overlay.Device
- pki *PKI
- Cipher string
- Firewall *Firewall
- ServeDns bool
- HandshakeManager *HandshakeManager
- lightHouse *LightHouse
- connectionManager *connectionManager
- DropLocalBroadcast bool
- DropMulticast bool
- routines int
- MessageMetrics *MessageMetrics
- version string
- relayManager *relayManager
- punchy *Punchy
- tryPromoteEvery uint32
- reQueryEvery uint32
- reQueryWait time.Duration
- ConntrackCacheTimeout time.Duration
- l *logrus.Logger
- }
- type Interface struct {
- hostMap *HostMap
- outside udp.Conn
- inside overlay.Device
- pki *PKI
- firewall *Firewall
- connectionManager *connectionManager
- handshakeManager *HandshakeManager
- serveDns bool
- createTime time.Time
- lightHouse *LightHouse
- myBroadcastAddrsTable *bart.Lite
- myVpnAddrs []netip.Addr // A list of addresses assigned to us via our certificate
- myVpnAddrsTable *bart.Lite
- myVpnNetworks []netip.Prefix // A list of networks assigned to us via our certificate
- myVpnNetworksTable *bart.Lite
- dropLocalBroadcast bool
- dropMulticast bool
- routines int
- disconnectInvalid atomic.Bool
- closed atomic.Bool
- relayManager *relayManager
- tryPromoteEvery atomic.Uint32
- reQueryEvery atomic.Uint32
- reQueryWait atomic.Int64
- sendRecvErrorConfig sendRecvErrorConfig
- // rebindCount is used to decide if an active tunnel should trigger a punch notification through a lighthouse
- rebindCount int8
- version string
- conntrackCacheTimeout time.Duration
- writers []udp.Conn
- readers []io.ReadWriteCloser
- metricHandshakes metrics.Histogram
- messageMetrics *MessageMetrics
- cachedPacketMetrics *cachedPacketMetrics
- l *logrus.Logger
- }
- type EncWriter interface {
- SendVia(via *HostInfo,
- relay *Relay,
- ad,
- nb,
- out []byte,
- nocopy bool,
- )
- SendMessageToVpnAddr(t header.MessageType, st header.MessageSubType, vpnAddr netip.Addr, p, nb, out []byte)
- SendMessageToHostInfo(t header.MessageType, st header.MessageSubType, hostinfo *HostInfo, p, nb, out []byte)
- Handshake(vpnAddr netip.Addr)
- GetHostInfo(vpnAddr netip.Addr) *HostInfo
- GetCertState() *CertState
- }
- type sendRecvErrorConfig uint8
- const (
- sendRecvErrorAlways sendRecvErrorConfig = iota
- sendRecvErrorNever
- sendRecvErrorPrivate
- )
- func (s sendRecvErrorConfig) ShouldSendRecvError(endpoint netip.AddrPort) bool {
- switch s {
- case sendRecvErrorPrivate:
- return endpoint.Addr().IsPrivate()
- case sendRecvErrorAlways:
- return true
- case sendRecvErrorNever:
- return false
- default:
- panic(fmt.Errorf("invalid sendRecvErrorConfig value: %d", s))
- }
- }
- func (s sendRecvErrorConfig) String() string {
- switch s {
- case sendRecvErrorAlways:
- return "always"
- case sendRecvErrorNever:
- return "never"
- case sendRecvErrorPrivate:
- return "private"
- default:
- return fmt.Sprintf("invalid(%d)", s)
- }
- }
- func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
- if c.Outside == nil {
- return nil, errors.New("no outside connection")
- }
- if c.Inside == nil {
- return nil, errors.New("no inside interface (tun)")
- }
- if c.pki == nil {
- return nil, errors.New("no certificate state")
- }
- if c.Firewall == nil {
- return nil, errors.New("no firewall rules")
- }
- if c.connectionManager == nil {
- return nil, errors.New("no connection manager")
- }
- cs := c.pki.getCertState()
- ifce := &Interface{
- pki: c.pki,
- hostMap: c.HostMap,
- outside: c.Outside,
- inside: c.Inside,
- firewall: c.Firewall,
- serveDns: c.ServeDns,
- handshakeManager: c.HandshakeManager,
- createTime: time.Now(),
- lightHouse: c.lightHouse,
- dropLocalBroadcast: c.DropLocalBroadcast,
- dropMulticast: c.DropMulticast,
- routines: c.routines,
- version: c.version,
- writers: make([]udp.Conn, c.routines),
- readers: make([]io.ReadWriteCloser, c.routines),
- myVpnNetworks: cs.myVpnNetworks,
- myVpnNetworksTable: cs.myVpnNetworksTable,
- myVpnAddrs: cs.myVpnAddrs,
- myVpnAddrsTable: cs.myVpnAddrsTable,
- myBroadcastAddrsTable: cs.myVpnBroadcastAddrsTable,
- relayManager: c.relayManager,
- connectionManager: c.connectionManager,
- conntrackCacheTimeout: c.ConntrackCacheTimeout,
- metricHandshakes: metrics.GetOrRegisterHistogram("handshakes", nil, metrics.NewExpDecaySample(1028, 0.015)),
- messageMetrics: c.MessageMetrics,
- cachedPacketMetrics: &cachedPacketMetrics{
- sent: metrics.GetOrRegisterCounter("hostinfo.cached_packets.sent", nil),
- dropped: metrics.GetOrRegisterCounter("hostinfo.cached_packets.dropped", nil),
- },
- l: c.l,
- }
- ifce.tryPromoteEvery.Store(c.tryPromoteEvery)
- ifce.reQueryEvery.Store(c.reQueryEvery)
- ifce.reQueryWait.Store(int64(c.reQueryWait))
- ifce.connectionManager.intf = ifce
- return ifce, nil
- }
- // activate creates the interface on the host. After the interface is created, any
- // other services that want to bind listeners to its IP may do so successfully. However,
- // the interface isn't going to process anything until run() is called.
- func (f *Interface) activate() {
- // actually turn on tun dev
- addr, err := f.outside.LocalAddr()
- if err != nil {
- f.l.WithError(err).Error("Failed to get udp listen address")
- }
- f.l.WithField("interface", f.inside.Name()).WithField("networks", f.myVpnNetworks).
- WithField("build", f.version).WithField("udpAddr", addr).
- WithField("boringcrypto", boringEnabled()).
- Info("Nebula interface is active")
- metrics.GetOrRegisterGauge("routines", nil).Update(int64(f.routines))
- // Prepare n tun queues
- var reader io.ReadWriteCloser = f.inside
- for i := 0; i < f.routines; i++ {
- if i > 0 {
- reader, err = f.inside.NewMultiQueueReader()
- if err != nil {
- f.l.Fatal(err)
- }
- }
- f.readers[i] = reader
- }
- if err := f.inside.Activate(); err != nil {
- f.inside.Close()
- f.l.Fatal(err)
- }
- }
- func (f *Interface) run() {
- // Launch n queues to read packets from udp
- for i := 0; i < f.routines; i++ {
- go f.listenOut(i)
- }
- // Launch n queues to read packets from tun dev
- for i := 0; i < f.routines; i++ {
- go f.listenIn(f.readers[i], i)
- }
- }
- func (f *Interface) listenOut(i int) {
- runtime.LockOSThread()
- var li udp.Conn
- if i > 0 {
- li = f.writers[i]
- } else {
- li = f.outside
- }
- ctCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
- lhh := f.lightHouse.NewRequestHandler()
- plaintext := make([]byte, udp.MTU)
- h := &header.H{}
- fwPacket := &firewall.Packet{}
- nb := make([]byte, 12, 12)
- li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
- f.readOutsidePackets(fromUdpAddr, nil, plaintext[:0], payload, h, fwPacket, lhh, nb, i, ctCache.Get(f.l))
- })
- }
- func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
- runtime.LockOSThread()
- packet := make([]byte, mtu)
- out := make([]byte, mtu)
- fwPacket := &firewall.Packet{}
- nb := make([]byte, 12, 12)
- conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
- for {
- n, err := reader.Read(packet)
- if err != nil {
- if errors.Is(err, os.ErrClosed) && f.closed.Load() {
- return
- }
- f.l.WithError(err).Error("Error while reading outbound packet")
- // This only seems to happen when something fatal happens to the fd, so exit.
- os.Exit(2)
- }
- f.consumeInsidePacket(packet[:n], fwPacket, nb, out, i, conntrackCache.Get(f.l))
- }
- }
- func (f *Interface) RegisterConfigChangeCallbacks(c *config.C) {
- c.RegisterReloadCallback(f.reloadFirewall)
- c.RegisterReloadCallback(f.reloadSendRecvError)
- c.RegisterReloadCallback(f.reloadDisconnectInvalid)
- c.RegisterReloadCallback(f.reloadMisc)
- for _, udpConn := range f.writers {
- c.RegisterReloadCallback(udpConn.ReloadConfig)
- }
- }
- func (f *Interface) reloadDisconnectInvalid(c *config.C) {
- initial := c.InitialLoad()
- if initial || c.HasChanged("pki.disconnect_invalid") {
- f.disconnectInvalid.Store(c.GetBool("pki.disconnect_invalid", true))
- if !initial {
- f.l.Infof("pki.disconnect_invalid changed to %v", f.disconnectInvalid.Load())
- }
- }
- }
- func (f *Interface) reloadFirewall(c *config.C) {
- //TODO: need to trigger/detect if the certificate changed too
- if c.HasChanged("firewall") == false {
- f.l.Debug("No firewall config change detected")
- return
- }
- fw, err := NewFirewallFromConfig(f.l, f.pki.getCertState(), c)
- if err != nil {
- f.l.WithError(err).Error("Error while creating firewall during reload")
- return
- }
- oldFw := f.firewall
- conntrack := oldFw.Conntrack
- conntrack.Lock()
- defer conntrack.Unlock()
- fw.rulesVersion = oldFw.rulesVersion + 1
- // If rulesVersion is back to zero, we have wrapped all the way around. Be
- // safe and just reset conntrack in this case.
- if fw.rulesVersion == 0 {
- f.l.WithField("firewallHashes", fw.GetRuleHashes()).
- WithField("oldFirewallHashes", oldFw.GetRuleHashes()).
- WithField("rulesVersion", fw.rulesVersion).
- Warn("firewall rulesVersion has overflowed, resetting conntrack")
- } else {
- fw.Conntrack = conntrack
- }
- f.firewall = fw
- oldFw.Destroy()
- f.l.WithField("firewallHashes", fw.GetRuleHashes()).
- WithField("oldFirewallHashes", oldFw.GetRuleHashes()).
- WithField("rulesVersion", fw.rulesVersion).
- Info("New firewall has been installed")
- }
- func (f *Interface) reloadSendRecvError(c *config.C) {
- if c.InitialLoad() || c.HasChanged("listen.send_recv_error") {
- stringValue := c.GetString("listen.send_recv_error", "always")
- switch stringValue {
- case "always":
- f.sendRecvErrorConfig = sendRecvErrorAlways
- case "never":
- f.sendRecvErrorConfig = sendRecvErrorNever
- case "private":
- f.sendRecvErrorConfig = sendRecvErrorPrivate
- default:
- if c.GetBool("listen.send_recv_error", true) {
- f.sendRecvErrorConfig = sendRecvErrorAlways
- } else {
- f.sendRecvErrorConfig = sendRecvErrorNever
- }
- }
- f.l.WithField("sendRecvError", f.sendRecvErrorConfig.String()).
- Info("Loaded send_recv_error config")
- }
- }
- func (f *Interface) reloadMisc(c *config.C) {
- if c.HasChanged("counters.try_promote") {
- n := c.GetUint32("counters.try_promote", defaultPromoteEvery)
- f.tryPromoteEvery.Store(n)
- f.l.Info("counters.try_promote has changed")
- }
- if c.HasChanged("counters.requery_every_packets") {
- n := c.GetUint32("counters.requery_every_packets", defaultReQueryEvery)
- f.reQueryEvery.Store(n)
- f.l.Info("counters.requery_every_packets has changed")
- }
- if c.HasChanged("timers.requery_wait_duration") {
- n := c.GetDuration("timers.requery_wait_duration", defaultReQueryWait)
- f.reQueryWait.Store(int64(n))
- f.l.Info("timers.requery_wait_duration has changed")
- }
- }
- func (f *Interface) emitStats(ctx context.Context, i time.Duration) {
- ticker := time.NewTicker(i)
- defer ticker.Stop()
- udpStats := udp.NewUDPStatsEmitter(f.writers)
- certExpirationGauge := metrics.GetOrRegisterGauge("certificate.ttl_seconds", nil)
- certInitiatingVersion := metrics.GetOrRegisterGauge("certificate.initiating_version", nil)
- certMaxVersion := metrics.GetOrRegisterGauge("certificate.max_version", nil)
- for {
- select {
- case <-ctx.Done():
- return
- case <-ticker.C:
- f.firewall.EmitStats()
- f.handshakeManager.EmitStats()
- udpStats()
- certState := f.pki.getCertState()
- defaultCrt := certState.GetDefaultCertificate()
- certExpirationGauge.Update(int64(defaultCrt.NotAfter().Sub(time.Now()) / time.Second))
- certInitiatingVersion.Update(int64(defaultCrt.Version()))
- // Report the max certificate version we are capable of using
- if certState.v2Cert != nil {
- certMaxVersion.Update(int64(certState.v2Cert.Version()))
- } else {
- certMaxVersion.Update(int64(certState.v1Cert.Version()))
- }
- }
- }
- }
- func (f *Interface) GetHostInfo(vpnIp netip.Addr) *HostInfo {
- return f.hostMap.QueryVpnAddr(vpnIp)
- }
- func (f *Interface) GetCertState() *CertState {
- return f.pki.getCertState()
- }
- func (f *Interface) Close() error {
- f.closed.Store(true)
- for _, u := range f.writers {
- err := u.Close()
- if err != nil {
- f.l.WithError(err).Error("Error while closing udp socket")
- }
- }
- // Release the tun device
- return f.inside.Close()
- }
|