github.com/gnolang/gno@v0.0.0-20240520182011-228e9d0192ce/tm2/pkg/p2p/switch.go

package p2p

import (
	"context"
	"fmt"
	"math"
	"sync"
	"time"

	"github.com/gnolang/gno/tm2/pkg/cmap"
	"github.com/gnolang/gno/tm2/pkg/errors"
	"github.com/gnolang/gno/tm2/pkg/p2p/config"
	"github.com/gnolang/gno/tm2/pkg/p2p/conn"
	"github.com/gnolang/gno/tm2/pkg/random"
	"github.com/gnolang/gno/tm2/pkg/service"
	"github.com/gnolang/gno/tm2/pkg/telemetry"
	"github.com/gnolang/gno/tm2/pkg/telemetry/metrics"
)

const (
	// wait a random amount of time from this interval
	// before dialing peers or reconnecting to help prevent DoS
	dialRandomizerIntervalMilliseconds = 3000

	// repeatedly try to reconnect for a few minutes
	// ie. 5 * 20 = 100s
	reconnectAttempts = 20
	reconnectInterval = 5 * time.Second

	// then move into exponential backoff mode for ~1day
	// ie. 3**10 = 16hrs
	reconnectBackOffAttempts    = 10
	reconnectBackOffBaseSeconds = 3
)

// MConnConfig returns an MConnConfig with fields updated
// from the P2PConfig.
func MConnConfig(cfg *config.P2PConfig) conn.MConnConfig {
	mConfig := conn.DefaultMConnConfig()
	mConfig.FlushThrottle = cfg.FlushThrottleTimeout
	mConfig.SendRate = cfg.SendRate
	mConfig.RecvRate = cfg.RecvRate
	mConfig.MaxPacketMsgPayloadSize = cfg.MaxPacketMsgPayloadSize
	return mConfig
}

// PeerFilterFunc to be implemented by filter hooks after a new Peer has been
// fully setup.
type PeerFilterFunc func(IPeerSet, Peer) error

// -----------------------------------------------------------------------------

// Switch handles peer connections and exposes an API to receive incoming messages
// on `Reactors`.  Each `Reactor` is responsible for handling incoming messages of one
// or more `Channels`.  So while sending outgoing messages is typically performed on the peer,
// incoming messages are received on the reactor.
type Switch struct {
	service.BaseService

	config       *config.P2PConfig
	reactors     map[string]Reactor
	chDescs      []*conn.ChannelDescriptor
	reactorsByCh map[byte]Reactor
	peers        *PeerSet
	dialing      *cmap.CMap
	reconnecting *cmap.CMap
	nodeInfo     NodeInfo // our node info
	nodeKey      *NodeKey // our node privkey
	// peers addresses with whom we'll maintain constant connection
	persistentPeersAddrs []*NetAddress

	transport Transport

	filterTimeout time.Duration
	peerFilters   []PeerFilterFunc

	rng *random.Rand // seed for randomizing dial times and orders
}

// NetAddress returns the address the switch is listening on.
func (sw *Switch) NetAddress() *NetAddress {
	addr := sw.transport.NetAddress()
	return &addr
}

// SwitchOption sets an optional parameter on the Switch.
type SwitchOption func(*Switch)

// NewSwitch creates a new Switch with the given config.
func NewSwitch(
	cfg *config.P2PConfig,
	transport Transport,
	options ...SwitchOption,
) *Switch {
	sw := &Switch{
		config:               cfg,
		reactors:             make(map[string]Reactor),
		chDescs:              make([]*conn.ChannelDescriptor, 0),
		reactorsByCh:         make(map[byte]Reactor),
		peers:                NewPeerSet(),
		dialing:              cmap.NewCMap(),
		reconnecting:         cmap.NewCMap(),
		transport:            transport,
		filterTimeout:        defaultFilterTimeout,
		persistentPeersAddrs: make([]*NetAddress, 0),
	}

	// Ensure we have a completely undeterministic PRNG.
	sw.rng = random.NewRand()

	sw.BaseService = *service.NewBaseService(nil, "P2P Switch", sw)

	for _, option := range options {
		option(sw)
	}

	return sw
}

// SwitchFilterTimeout sets the timeout used for peer filters.
func SwitchFilterTimeout(timeout time.Duration) SwitchOption {
	return func(sw *Switch) { sw.filterTimeout = timeout }
}

// SwitchPeerFilters sets the filters for rejection of new peers.
func SwitchPeerFilters(filters ...PeerFilterFunc) SwitchOption {
	return func(sw *Switch) { sw.peerFilters = filters }
}

// ---------------------------------------------------------------------
// Switch setup

// AddReactor adds the given reactor to the switch.
// NOTE: Not goroutine safe.
func (sw *Switch) AddReactor(name string, reactor Reactor) Reactor {
	for _, chDesc := range reactor.GetChannels() {
		chID := chDesc.ID
		// No two reactors can share the same channel.
		if sw.reactorsByCh[chID] != nil {
			panic(fmt.Sprintf("Channel %X has multiple reactors %v & %v", chID, sw.reactorsByCh[chID], reactor))
		}
		sw.chDescs = append(sw.chDescs, chDesc)
		sw.reactorsByCh[chID] = reactor
	}
	sw.reactors[name] = reactor
	reactor.SetSwitch(sw)
	return reactor
}

// RemoveReactor removes the given Reactor from the Switch.
// NOTE: Not goroutine safe.
func (sw *Switch) RemoveReactor(name string, reactor Reactor) {
	for _, chDesc := range reactor.GetChannels() {
		// remove channel description
		for i := 0; i < len(sw.chDescs); i++ {
			if chDesc.ID == sw.chDescs[i].ID {
				sw.chDescs = append(sw.chDescs[:i], sw.chDescs[i+1:]...)
				break
			}
		}
		delete(sw.reactorsByCh, chDesc.ID)
	}
	delete(sw.reactors, name)
	reactor.SetSwitch(nil)
}

// Reactors returns a map of reactors registered on the switch.
// NOTE: Not goroutine safe.
func (sw *Switch) Reactors() map[string]Reactor {
	return sw.reactors
}

// Reactor returns the reactor with the given name.
// NOTE: Not goroutine safe.
func (sw *Switch) Reactor(name string) Reactor {
	return sw.reactors[name]
}

// SetNodeInfo sets the switch's NodeInfo for checking compatibility and handshaking with other nodes.
// NOTE: Not goroutine safe.
func (sw *Switch) SetNodeInfo(nodeInfo NodeInfo) {
	sw.nodeInfo = nodeInfo
}

// NodeInfo returns the switch's NodeInfo.
// NOTE: Not goroutine safe.
func (sw *Switch) NodeInfo() NodeInfo {
	return sw.nodeInfo
}

// SetNodeKey sets the switch's private key for authenticated encryption.
// NOTE: Not goroutine safe.
func (sw *Switch) SetNodeKey(nodeKey *NodeKey) {
	sw.nodeKey = nodeKey
}

// ---------------------------------------------------------------------
// Service start/stop

// OnStart implements BaseService. It starts all the reactors and peers.
func (sw *Switch) OnStart() error {
	// Start reactors
	for _, reactor := range sw.reactors {
		err := reactor.Start()
		if err != nil {
			return errors.Wrap(err, "failed to start %v", reactor)
		}
	}

	// Start accepting Peers.
	go sw.acceptRoutine()

	return nil
}

// OnStop implements BaseService. It stops all peers and reactors.
func (sw *Switch) OnStop() {
	// Stop transport
	if t, ok := sw.transport.(TransportLifecycle); ok {
		err := t.Close()
		if err != nil {
			sw.Logger.Error("Error stopping transport on stop: ", err)
		}
	}

	// Stop peers
	for _, p := range sw.peers.List() {
		sw.stopAndRemovePeer(p, nil)
	}

	// Stop reactors
	sw.Logger.Debug("Switch: Stopping reactors")
	for _, reactor := range sw.reactors {
		reactor.Stop()
	}
}

// ---------------------------------------------------------------------
// Peers

// Broadcast runs a go routine for each attempted send, which will block trying
// to send for defaultSendTimeoutSeconds. Returns a channel which receives
// success values for each attempted send (false if times out). Channel will be
// closed once msg bytes are sent to all peers (or time out).
//
// NOTE: Broadcast uses goroutines, so order of broadcast may not be preserved.
func (sw *Switch) Broadcast(chID byte, msgBytes []byte) chan bool {
	startTime := time.Now()

	sw.Logger.Debug(
		"Broadcast",
		"channel", chID,
		"value", fmt.Sprintf("%X", msgBytes),
	)

	peers := sw.peers.List()
	var wg sync.WaitGroup
	wg.Add(len(peers))
	successChan := make(chan bool, len(peers))

	for _, peer := range peers {
		go func(p Peer) {
			defer wg.Done()
			success := p.Send(chID, msgBytes)
			successChan <- success
		}(peer)
	}

	go func() {
		wg.Wait()
		close(successChan)
		if telemetry.MetricsEnabled() {
			metrics.BroadcastTxTimer.Record(context.Background(), time.Since(startTime).Milliseconds())
		}
	}()

	return successChan
}

// NumPeers returns the count of outbound/inbound and outbound-dialing peers.
func (sw *Switch) NumPeers() (outbound, inbound, dialing int) {
	peers := sw.peers.List()
	for _, peer := range peers {
		if peer.IsOutbound() {
			outbound++
		} else {
			inbound++
		}
	}
	dialing = sw.dialing.Size()
	return
}

// MaxNumOutboundPeers returns a maximum number of outbound peers.
func (sw *Switch) MaxNumOutboundPeers() int {
	return sw.config.MaxNumOutboundPeers
}

// Peers returns the set of peers that are connected to the switch.
func (sw *Switch) Peers() IPeerSet {
	return sw.peers
}

// StopPeerForError disconnects from a peer due to external error.
// If the peer is persistent, it will attempt to reconnect.
// TODO: make record depending on reason.
func (sw *Switch) StopPeerForError(peer Peer, reason interface{}) {
	sw.Logger.Error("Stopping peer for error", "peer", peer, "err", reason)
	sw.stopAndRemovePeer(peer, reason)

	if peer.IsPersistent() {
		var addr *NetAddress
		if peer.IsOutbound() { // socket address for outbound peers
			addr = peer.SocketAddr()
		} else { // self-reported address for inbound peers
			addr = peer.NodeInfo().NetAddress
		}
		go sw.reconnectToPeer(addr)
	}
}

// StopPeerGracefully disconnects from a peer gracefully.
// TODO: handle graceful disconnects.
func (sw *Switch) StopPeerGracefully(peer Peer) {
	sw.Logger.Info("Stopping peer gracefully")
	sw.stopAndRemovePeer(peer, nil)
}

func (sw *Switch) stopAndRemovePeer(peer Peer, reason interface{}) {
	sw.transport.Cleanup(peer)
	peer.Stop()

	for _, reactor := range sw.reactors {
		reactor.RemovePeer(peer, reason)
	}

	// Removing a peer should go last to avoid a situation where a peer
	// reconnect to our node and the switch calls InitPeer before
	// RemovePeer is finished.
	// https://github.com/tendermint/classic/issues/3338
	sw.peers.Remove(peer)
}

// reconnectToPeer tries to reconnect to the addr, first repeatedly
// with a fixed interval, then with exponential backoff.
// If no success after all that, it stops trying.
// NOTE: this will keep trying even if the handshake or auth fails.
// TODO: be more explicit with error types so we only retry on certain failures
//   - ie. if we're getting ErrDuplicatePeer we can stop
func (sw *Switch) reconnectToPeer(addr *NetAddress) {
	if sw.reconnecting.Has(addr.ID.String()) {
		return
	}
	sw.reconnecting.Set(addr.ID.String(), addr)
	defer sw.reconnecting.Delete(addr.ID.String())

	start := time.Now()
	sw.Logger.Info("Reconnecting to peer", "addr", addr)
	for i := 0; i < reconnectAttempts; i++ {
		if !sw.IsRunning() {
			return
		}

		err := sw.DialPeerWithAddress(addr)
		if err == nil {
			return // success
		} else if _, ok := err.(CurrentlyDialingOrExistingAddressError); ok {
			return
		}

		sw.Logger.Info("Error reconnecting to peer. Trying again", "tries", i, "err", err, "addr", addr)
		// sleep a set amount
		sw.randomSleep(reconnectInterval)
		continue
	}

	sw.Logger.Error("Failed to reconnect to peer. Beginning exponential backoff",
		"addr", addr, "elapsed", time.Since(start))
	for i := 0; i < reconnectBackOffAttempts; i++ {
		if !sw.IsRunning() {
			return
		}

		// sleep an exponentially increasing amount
		sleepIntervalSeconds := math.Pow(reconnectBackOffBaseSeconds, float64(i))
		sw.randomSleep(time.Duration(sleepIntervalSeconds) * time.Second)

		err := sw.DialPeerWithAddress(addr)
		if err == nil {
			return // success
		} else if _, ok := err.(CurrentlyDialingOrExistingAddressError); ok {
			return
		}
		sw.Logger.Info("Error reconnecting to peer. Trying again", "tries", i, "err", err, "addr", addr)
	}
	sw.Logger.Error("Failed to reconnect to peer. Giving up", "addr", addr, "elapsed", time.Since(start))
}

// ---------------------------------------------------------------------
// Dialing

// DialPeersAsync dials a list of peers asynchronously in random order.
// Used to dial peers from config on startup or from unsafe-RPC (trusted sources).
// It ignores NetAddressLookupError. However, if there are other errors, first
// encounter is returned.
// Nop if there are no peers.
func (sw *Switch) DialPeersAsync(peers []string) error {
	netAddrs, errs := NewNetAddressFromStrings(peers)
	// report all the errors
	for _, err := range errs {
		sw.Logger.Error("Error in peer's address", "err", err)
	}
	// return first non-NetAddressLookupError error
	for _, err := range errs {
		if _, ok := err.(NetAddressLookupError); ok {
			continue
		}
		return err
	}
	sw.dialPeersAsync(netAddrs)
	return nil
}

func (sw *Switch) dialPeersAsync(netAddrs []*NetAddress) {
	ourAddr := sw.NetAddress()

	// permute the list, dial them in random order.
	perm := sw.rng.Perm(len(netAddrs))
	for i := 0; i < len(perm); i++ {
		go func(i int) {
			j := perm[i]
			addr := netAddrs[j]

			if addr.Same(ourAddr) {
				sw.Logger.Debug("Ignore attempt to connect to ourselves", "addr", addr, "ourAddr", ourAddr)
				return
			}

			sw.randomSleep(0)

			err := sw.DialPeerWithAddress(addr)
			if err != nil {
				switch err.(type) {
				case SwitchConnectToSelfError, SwitchDuplicatePeerIDError, CurrentlyDialingOrExistingAddressError:
					sw.Logger.Debug("Error dialing peer", "err", err)
				default:
					sw.Logger.Error("Error dialing peer", "err", err)
				}
			}
		}(i)
	}
}

// DialPeerWithAddress dials the given peer and runs sw.addPeer if it connects
// and authenticates successfully.
// If we're currently dialing this address or it belongs to an existing peer,
// CurrentlyDialingOrExistingAddressError is returned.
func (sw *Switch) DialPeerWithAddress(addr *NetAddress) error {
	if sw.IsDialingOrExistingAddress(addr) {
		return CurrentlyDialingOrExistingAddressError{addr.String()}
	}

	sw.dialing.Set(addr.ID.String(), addr)
	defer sw.dialing.Delete(addr.ID.String())

	return sw.addOutboundPeerWithConfig(addr, sw.config)
}

// sleep for interval plus some random amount of ms on [0, dialRandomizerIntervalMilliseconds]
func (sw *Switch) randomSleep(interval time.Duration) {
	r := time.Duration(sw.rng.Int63n(dialRandomizerIntervalMilliseconds)) * time.Millisecond
	time.Sleep(r + interval)
}

// IsDialingOrExistingAddress returns true if switch has a peer with the given
// address or dialing it at the moment.
func (sw *Switch) IsDialingOrExistingAddress(addr *NetAddress) bool {
	return sw.dialing.Has(addr.ID.String()) ||
		sw.peers.Has(addr.ID) ||
		(!sw.config.AllowDuplicateIP && sw.peers.HasIP(addr.IP))
}

// AddPersistentPeers allows you to set persistent peers. It ignores
// NetAddressLookupError. However, if there are other errors, first encounter is
// returned.
func (sw *Switch) AddPersistentPeers(addrs []string) error {
	sw.Logger.Info("Adding persistent peers", "addrs", addrs)
	netAddrs, errs := NewNetAddressFromStrings(addrs)
	// report all the errors
	for _, err := range errs {
		sw.Logger.Error("Error in peer's address", "err", err)
	}
	// return first non-NetAddressLookupError error
	for _, err := range errs {
		if _, ok := err.(NetAddressLookupError); ok {
			continue
		}
		return err
	}
	sw.persistentPeersAddrs = netAddrs
	return nil
}

func (sw *Switch) isPeerPersistentFn() func(*NetAddress) bool {
	return func(na *NetAddress) bool {
		for _, pa := range sw.persistentPeersAddrs {
			if pa.Equals(na) {
				return true
			}
		}
		return false
	}
}

func (sw *Switch) acceptRoutine() {
	for {
		p, err := sw.transport.Accept(peerConfig{
			chDescs:      sw.chDescs,
			onPeerError:  sw.StopPeerForError,
			reactorsByCh: sw.reactorsByCh,
			isPersistent: sw.isPeerPersistentFn(),
		})
		if err != nil {
			switch err := err.(type) {
			case RejectedError:
				if err.IsSelf() {
					// TODO: warn?
				}

				sw.Logger.Info(
					"Inbound Peer rejected",
					"err", err,
					"numPeers", sw.peers.Size(),
				)

				continue
			case FilterTimeoutError:
				sw.Logger.Error(
					"Peer filter timed out",
					"err", err,
				)

				continue
			case TransportClosedError:
				sw.Logger.Error(
					"Stopped accept routine, as transport is closed",
					"numPeers", sw.peers.Size(),
				)
			default:
				sw.Logger.Error(
					"Accept on transport errored",
					"err", err,
					"numPeers", sw.peers.Size(),
				)
				// We could instead have a retry loop around the acceptRoutine,
				// but that would need to stop and let the node shutdown eventually.
				// So might as well panic and let process managers restart the node.
				// There's no point in letting the node run without the acceptRoutine,
				// since it won't be able to accept new connections.
				panic(fmt.Errorf("accept routine exited: %w", err))
			}

			break
		}

		// Ignore connection if we already have enough peers.
		_, in, _ := sw.NumPeers()
		if in >= sw.config.MaxNumInboundPeers {
			sw.Logger.Info(
				"Ignoring inbound connection: already have enough inbound peers",
				"address", p.SocketAddr(),
				"have", in,
				"max", sw.config.MaxNumInboundPeers,
			)

			sw.transport.Cleanup(p)

			continue
		}

		if err := sw.addPeer(p); err != nil {
			sw.transport.Cleanup(p)
			if p.IsRunning() {
				_ = p.Stop()
			}
			sw.Logger.Info(
				"Ignoring inbound connection: error while adding peer",
				"err", err,
				"id", p.ID(),
			)
		}
	}
}

// dial the peer; make secret connection; authenticate against the dialed ID;
// add the peer.
// if dialing fails, start the reconnect loop. If handshake fails, it's over.
// If peer is started successfully, reconnectLoop will start when
// StopPeerForError is called.
func (sw *Switch) addOutboundPeerWithConfig(
	addr *NetAddress,
	cfg *config.P2PConfig,
) error {
	sw.Logger.Info("Dialing peer", "address", addr)

	// XXX(xla): Remove the leakage of test concerns in implementation.
	if cfg.TestDialFail {
		go sw.reconnectToPeer(addr)
		return fmt.Errorf("dial err (peerConfig.DialFail == true)")
	}

	p, err := sw.transport.Dial(*addr, peerConfig{
		chDescs:      sw.chDescs,
		onPeerError:  sw.StopPeerForError,
		isPersistent: sw.isPeerPersistentFn(),
		reactorsByCh: sw.reactorsByCh,
	})
	if err != nil {
		if e, ok := err.(RejectedError); ok {
			if e.IsSelf() {
				// TODO: warn?
				return err
			}
		}

		// retry persistent peers after
		// any dial error besides IsSelf()
		if sw.isPeerPersistentFn()(addr) {
			go sw.reconnectToPeer(addr)
		}

		return err
	}

	if err := sw.addPeer(p); err != nil {
		sw.transport.Cleanup(p)
		if p.IsRunning() {
			_ = p.Stop()
		}
		return err
	}

	return nil
}

func (sw *Switch) filterPeer(p Peer) error {
	// Avoid duplicate
	if sw.peers.Has(p.ID()) {
		return RejectedError{id: p.ID(), isDuplicate: true}
	}

	errc := make(chan error, len(sw.peerFilters))

	for _, f := range sw.peerFilters {
		go func(f PeerFilterFunc, p Peer, errc chan<- error) {
			errc <- f(sw.peers, p)
		}(f, p, errc)
	}

	for i := 0; i < cap(errc); i++ {
		select {
		case err := <-errc:
			if err != nil {
				return RejectedError{id: p.ID(), err: err, isFiltered: true}
			}
		case <-time.After(sw.filterTimeout):
			return FilterTimeoutError{}
		}
	}

	return nil
}

// addPeer starts up the Peer and adds it to the Switch. Error is returned if
// the peer is filtered out or failed to start or can't be added.
func (sw *Switch) addPeer(p Peer) error {
	if err := sw.filterPeer(p); err != nil {
		return err
	}

	p.SetLogger(sw.Logger.With("peer", p.SocketAddr()))

	// Handle the shut down case where the switch has stopped but we're
	// concurrently trying to add a peer.
	if !sw.IsRunning() {
		// XXX should this return an error or just log and terminate?
		sw.Logger.Error("Won't start a peer - switch is not running", "peer", p)
		return nil
	}

	// Add some data to the peer, which is required by reactors.
	for _, reactor := range sw.reactors {
		p = reactor.InitPeer(p)
	}

	// Start the peer's send/recv routines.
	// Must start it before adding it to the peer set
	// to prevent Start and Stop from being called concurrently.
	err := p.Start()
	if err != nil {
		// Should never happen
		sw.Logger.Error("Error starting peer", "err", err, "peer", p)
		return err
	}

	// Add the peer to PeerSet. Do this before starting the reactors
	// so that if Receive errors, we will find the peer and remove it.
	// Add should not err since we already checked peers.Has().
	if err := sw.peers.Add(p); err != nil {
		return err
	}

	// Start all the reactor protocols on the peer.
	for _, reactor := range sw.reactors {
		reactor.AddPeer(p)
	}

	sw.Logger.Info("Added peer", "peer", p)

	return nil
}