5cc0765b63
This commit renames routing processing method in the funding mangers config from ProcessRoutingMessage to SendToRouter and also modifies the signature to only require the message itself and not the server’s identity public key.
889 lines
24 KiB
Go
889 lines
24 KiB
Go
package main
|
|
|
|
import (
|
|
"encoding/hex"
|
|
"errors"
|
|
"fmt"
|
|
"net"
|
|
"sync"
|
|
"sync/atomic"
|
|
"time"
|
|
|
|
"github.com/btcsuite/fastsha256"
|
|
"github.com/lightningnetwork/lightning-onion"
|
|
"github.com/lightningnetwork/lnd/brontide"
|
|
"github.com/lightningnetwork/lnd/chainntnfs"
|
|
"github.com/lightningnetwork/lnd/channeldb"
|
|
"github.com/lightningnetwork/lnd/lnrpc"
|
|
"github.com/lightningnetwork/lnd/lnwallet"
|
|
"github.com/lightningnetwork/lnd/lnwire"
|
|
"github.com/lightningnetwork/lnd/routing"
|
|
"github.com/roasbeef/btcd/btcec"
|
|
"github.com/roasbeef/btcd/chaincfg/chainhash"
|
|
"github.com/roasbeef/btcd/connmgr"
|
|
"github.com/roasbeef/btcutil"
|
|
)
|
|
|
|
// server is the main server of the Lightning Network Daemon. The server houses
|
|
// global state pertaining to the wallet, database, and the rpcserver.
|
|
// Additionally, the server is also used as a central messaging bus to interact
|
|
// with any of its companion objects.
|
|
type server struct {
|
|
started int32 // atomic
|
|
shutdown int32 // atomic
|
|
|
|
// identityPriv is the private key used to authenticate any incoming
|
|
// connections.
|
|
identityPriv *btcec.PrivateKey
|
|
|
|
// lightningID is the sha256 of the public key corresponding to our
|
|
// long-term identity private key.
|
|
lightningID [32]byte
|
|
|
|
peersMtx sync.RWMutex
|
|
peersByID map[int32]*peer
|
|
peersByPub map[string]*peer
|
|
|
|
rpcServer *rpcServer
|
|
|
|
chainNotifier chainntnfs.ChainNotifier
|
|
|
|
bio lnwallet.BlockChainIO
|
|
lnwallet *lnwallet.LightningWallet
|
|
|
|
fundingMgr *fundingManager
|
|
chanDB *channeldb.DB
|
|
|
|
htlcSwitch *htlcSwitch
|
|
invoices *invoiceRegistry
|
|
breachArbiter *breachArbiter
|
|
|
|
chanRouter *routing.ChannelRouter
|
|
|
|
utxoNursery *utxoNursery
|
|
|
|
sphinx *sphinx.Router
|
|
|
|
connMgr *connmgr.ConnManager
|
|
|
|
pendingConnMtx sync.RWMutex
|
|
persistentConnReqs map[string]*connmgr.ConnReq
|
|
|
|
broadcastRequests chan *broadcastReq
|
|
sendRequests chan *sendReq
|
|
|
|
newPeers chan *peer
|
|
donePeers chan *peer
|
|
queries chan interface{}
|
|
|
|
// globalFeatures feature vector which affects HTLCs and thus are also
|
|
// advertised to other nodes.
|
|
globalFeatures *lnwire.FeatureVector
|
|
|
|
// localFeatures is an feature vector which represent the features which
|
|
// only affect the protocol between these two nodes.
|
|
localFeatures *lnwire.FeatureVector
|
|
|
|
wg sync.WaitGroup
|
|
quit chan struct{}
|
|
}
|
|
|
|
// newServer creates a new instance of the server which is to listen using the
|
|
// passed listener address.
|
|
func newServer(listenAddrs []string, notifier chainntnfs.ChainNotifier,
|
|
bio lnwallet.BlockChainIO, wallet *lnwallet.LightningWallet,
|
|
chanDB *channeldb.DB) (*server, error) {
|
|
|
|
privKey, err := wallet.GetIdentitykey()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
privKey.Curve = btcec.S256()
|
|
|
|
listeners := make([]net.Listener, len(listenAddrs))
|
|
for i, addr := range listenAddrs {
|
|
listeners[i], err = brontide.NewListener(privKey, addr)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
}
|
|
|
|
serializedPubKey := privKey.PubKey().SerializeCompressed()
|
|
s := &server{
|
|
lnwallet: wallet,
|
|
bio: bio,
|
|
chainNotifier: notifier,
|
|
chanDB: chanDB,
|
|
|
|
invoices: newInvoiceRegistry(chanDB),
|
|
utxoNursery: newUtxoNursery(chanDB, notifier, wallet),
|
|
htlcSwitch: newHtlcSwitch(),
|
|
|
|
identityPriv: privKey,
|
|
|
|
// TODO(roasbeef): derive proper onion key based on rotation
|
|
// schedule
|
|
sphinx: sphinx.NewRouter(privKey, activeNetParams.Params),
|
|
lightningID: fastsha256.Sum256(serializedPubKey),
|
|
|
|
persistentConnReqs: make(map[string]*connmgr.ConnReq),
|
|
|
|
peersByID: make(map[int32]*peer),
|
|
peersByPub: make(map[string]*peer),
|
|
|
|
newPeers: make(chan *peer, 10),
|
|
donePeers: make(chan *peer, 10),
|
|
|
|
broadcastRequests: make(chan *broadcastReq),
|
|
sendRequests: make(chan *sendReq),
|
|
|
|
globalFeatures: globalFeatures,
|
|
localFeatures: localFeatures,
|
|
|
|
queries: make(chan interface{}),
|
|
quit: make(chan struct{}),
|
|
}
|
|
|
|
// If the debug HTLC flag is on, then we invoice a "master debug"
|
|
// invoice which all outgoing payments will be sent and all incoming
|
|
// HTLCs with the debug R-Hash immediately settled.
|
|
if cfg.DebugHTLC {
|
|
kiloCoin := btcutil.Amount(btcutil.SatoshiPerBitcoin * 1000)
|
|
s.invoices.AddDebugInvoice(kiloCoin, *debugPre)
|
|
srvrLog.Debugf("Debug HTLC invoice inserted, preimage=%x, hash=%x",
|
|
debugPre[:], debugHash[:])
|
|
}
|
|
|
|
// TODO(roasbeef): add --externalip flag?
|
|
selfAddr, ok := listeners[0].Addr().(*net.TCPAddr)
|
|
if !ok {
|
|
return nil, fmt.Errorf("default listener must be TCP")
|
|
}
|
|
|
|
chanGraph := chanDB.ChannelGraph()
|
|
self := &channeldb.LightningNode{
|
|
LastUpdate: time.Now(),
|
|
Address: selfAddr,
|
|
PubKey: privKey.PubKey(),
|
|
// TODO(roasbeef): make alias configurable
|
|
Alias: hex.EncodeToString(serializedPubKey[:10]),
|
|
}
|
|
if err := chanGraph.SetSourceNode(self); err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
s.chanRouter, err = routing.New(routing.Config{
|
|
Graph: chanGraph,
|
|
Chain: bio,
|
|
Notifier: notifier,
|
|
Broadcast: s.broadcastMessage,
|
|
SendMessages: s.sendToPeer,
|
|
SendToSwitch: func(firstHop *btcec.PublicKey,
|
|
htlcAdd *lnwire.UpdateAddHTLC) ([32]byte, error) {
|
|
|
|
firstHopPub := firstHop.SerializeCompressed()
|
|
destInterface := chainhash.Hash(fastsha256.Sum256(firstHopPub))
|
|
|
|
return s.htlcSwitch.SendHTLC(&htlcPacket{
|
|
dest: destInterface,
|
|
msg: htlcAdd,
|
|
})
|
|
},
|
|
})
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
s.rpcServer = newRPCServer(s)
|
|
s.breachArbiter = newBreachArbiter(wallet, chanDB, notifier, s.htlcSwitch)
|
|
|
|
s.fundingMgr, err = newFundingManager(fundingConfig{
|
|
IDKey: s.identityPriv.PubKey(),
|
|
Wallet: wallet,
|
|
Notifier: s.chainNotifier,
|
|
SendToRouter: func(msg lnwire.Message) {
|
|
s.chanRouter.ProcessRoutingMessage(msg,
|
|
s.identityPriv.PubKey())
|
|
},
|
|
ArbiterChan: s.breachArbiter.newContracts,
|
|
SendToPeer: s.sendToPeer,
|
|
FindPeer: s.findPeer,
|
|
FindChannel: s.rpcServer.fetchActiveChannel,
|
|
})
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// TODO(roasbeef): introduce closure and config system to decouple the
|
|
// initialization above ^
|
|
|
|
// Create the connection manager which will be responsible for
|
|
// maintaining persistent outbound connections and also accepting new
|
|
// incoming connections
|
|
cmgr, err := connmgr.New(&connmgr.Config{
|
|
Listeners: listeners,
|
|
OnAccept: s.inboundPeerConnected,
|
|
RetryDuration: time.Second * 5,
|
|
TargetOutbound: 100,
|
|
GetNewAddress: nil,
|
|
Dial: noiseDial(s.identityPriv),
|
|
OnConnection: s.outboundPeerConnected,
|
|
})
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
s.connMgr = cmgr
|
|
|
|
// In order to promote liveness of our active channels, instruct the
|
|
// connection manager to attempt to establish and maintain persistent
|
|
// connections to all our direct channel counterparties.
|
|
linkNodes, err := s.chanDB.FetchAllLinkNodes()
|
|
if err != nil && err != channeldb.ErrLinkNodesNotFound {
|
|
return nil, err
|
|
}
|
|
for _, node := range linkNodes {
|
|
// Create a wrapper address which couples the IP and the pubkey
|
|
// so the brontide authenticated connection can be established.
|
|
lnAddr := &lnwire.NetAddress{
|
|
IdentityKey: node.IdentityPub,
|
|
Address: node.Addresses[0],
|
|
}
|
|
pubStr := string(node.IdentityPub.SerializeCompressed())
|
|
srvrLog.Debugf("Attempting persistent connection to channel "+
|
|
"peer %v", lnAddr)
|
|
|
|
// Send the persistent connection request to the connection
|
|
// manager, saving the request itself so we can cancel/restart
|
|
// the process as needed.
|
|
// TODO(roasbeef): use default addr
|
|
connReq := &connmgr.ConnReq{
|
|
Addr: lnAddr,
|
|
Permanent: true,
|
|
}
|
|
s.persistentConnReqs[pubStr] = connReq
|
|
go s.connMgr.Connect(connReq)
|
|
}
|
|
|
|
return s, nil
|
|
}
|
|
|
|
// Start starts the main daemon server, all requested listeners, and any helper
|
|
// goroutines.
|
|
func (s *server) Start() error {
|
|
// Already running?
|
|
if atomic.AddInt32(&s.started, 1) != 1 {
|
|
return nil
|
|
}
|
|
|
|
// Start the notification server. This is used so channel management
|
|
// goroutines can be notified when a funding transaction reaches a
|
|
// sufficient number of confirmations, or when the input for the
|
|
// funding transaction is spent in an attempt at an uncooperative close
|
|
// by the counterparty.
|
|
if err := s.chainNotifier.Start(); err != nil {
|
|
return err
|
|
}
|
|
|
|
if err := s.rpcServer.Start(); err != nil {
|
|
return err
|
|
}
|
|
if err := s.fundingMgr.Start(); err != nil {
|
|
return err
|
|
}
|
|
if err := s.htlcSwitch.Start(); err != nil {
|
|
return err
|
|
}
|
|
if err := s.utxoNursery.Start(); err != nil {
|
|
return err
|
|
}
|
|
if err := s.breachArbiter.Start(); err != nil {
|
|
return err
|
|
}
|
|
if err := s.chanRouter.Start(); err != nil {
|
|
return err
|
|
}
|
|
|
|
s.wg.Add(1)
|
|
go s.queryHandler()
|
|
|
|
return nil
|
|
}
|
|
|
|
// Stop gracefully shutsdown the main daemon server. This function will signal
|
|
// any active goroutines, or helper objects to exit, then blocks until they've
|
|
// all successfully exited. Additionally, any/all listeners are closed.
|
|
func (s *server) Stop() error {
|
|
// Bail if we're already shutting down.
|
|
if atomic.AddInt32(&s.shutdown, 1) != 1 {
|
|
return nil
|
|
}
|
|
|
|
// Shutdown the wallet, funding manager, and the rpc server.
|
|
s.chainNotifier.Stop()
|
|
s.rpcServer.Stop()
|
|
s.fundingMgr.Stop()
|
|
s.chanRouter.Stop()
|
|
s.htlcSwitch.Stop()
|
|
s.utxoNursery.Stop()
|
|
s.breachArbiter.Stop()
|
|
|
|
s.lnwallet.Shutdown()
|
|
|
|
// Signal all the lingering goroutines to quit.
|
|
close(s.quit)
|
|
s.wg.Wait()
|
|
|
|
return nil
|
|
}
|
|
|
|
// WaitForShutdown blocks all goroutines have been stopped.
|
|
func (s *server) WaitForShutdown() {
|
|
s.wg.Wait()
|
|
}
|
|
|
|
// broadcastReq is a message sent to the server by a related subsystem when it
|
|
// wishes to broadcast one or more messages to all connected peers. Thi
|
|
type broadcastReq struct {
|
|
ignore *btcec.PublicKey
|
|
msgs []lnwire.Message
|
|
|
|
errChan chan error // MUST be buffered.
|
|
}
|
|
|
|
// broadcastMessage sends a request to the server to broadcast a set of
|
|
// messages to all peers other than the one specified by the `skip` parameter.
|
|
func (s *server) broadcastMessage(skip *btcec.PublicKey, msgs ...lnwire.Message) error {
|
|
errChan := make(chan error, 1)
|
|
|
|
msgsToSend := make([]lnwire.Message, 0, len(msgs))
|
|
msgsToSend = append(msgsToSend, msgs...)
|
|
broadcastReq := &broadcastReq{
|
|
ignore: skip,
|
|
msgs: msgsToSend,
|
|
errChan: errChan,
|
|
}
|
|
|
|
select {
|
|
case s.broadcastRequests <- broadcastReq:
|
|
case <-s.quit:
|
|
return errors.New("server shutting down")
|
|
}
|
|
|
|
select {
|
|
case err := <-errChan:
|
|
return err
|
|
case <-s.quit:
|
|
return errors.New("server shutting down")
|
|
}
|
|
}
|
|
|
|
// sendReq is message sent to the server by a related subsystem which it
|
|
// wishes to send a set of messages to a specified peer.
|
|
type sendReq struct {
|
|
target *btcec.PublicKey
|
|
msgs []lnwire.Message
|
|
|
|
errChan chan error
|
|
}
|
|
|
|
// sendToPeer send a message to the server telling it to send the specific set
|
|
// of message to a particular peer. If the peer connect be found, then this
|
|
// method will return a non-nil error.
|
|
func (s *server) sendToPeer(target *btcec.PublicKey, msgs ...lnwire.Message) error {
|
|
errChan := make(chan error, 1)
|
|
|
|
msgsToSend := make([]lnwire.Message, 0, len(msgs))
|
|
msgsToSend = append(msgsToSend, msgs...)
|
|
sMsg := &sendReq{
|
|
target: target,
|
|
msgs: msgsToSend,
|
|
errChan: errChan,
|
|
}
|
|
|
|
select {
|
|
case s.sendRequests <- sMsg:
|
|
case <-s.quit:
|
|
return errors.New("server shutting down")
|
|
}
|
|
|
|
select {
|
|
case err := <-errChan:
|
|
return err
|
|
case <-s.quit:
|
|
return errors.New("server shutting down")
|
|
}
|
|
}
|
|
|
|
// findPeer will return the peer that corresponds to the passed in public key.
|
|
// This function is used by the funding manager, allowing it to update the
|
|
// daemon's local representation of the remote peer.
|
|
func (s *server) findPeer(peerKey *btcec.PublicKey) (*peer, error) {
|
|
serializedIDKey := string(peerKey.SerializeCompressed())
|
|
|
|
s.peersMtx.RLock()
|
|
peer := s.peersByPub[serializedIDKey]
|
|
s.peersMtx.RUnlock()
|
|
|
|
if peer == nil {
|
|
return nil, errors.New("Peer not found. Pubkey: " +
|
|
string(peerKey.SerializeCompressed()))
|
|
}
|
|
|
|
return peer, nil
|
|
}
|
|
|
|
// peerConnected is a function that handles initialization a newly connected
|
|
// peer by adding it to the server's global list of all active peers, and
|
|
// starting all the goroutines the peer needs to function properly.
|
|
func (s *server) peerConnected(conn net.Conn, connReq *connmgr.ConnReq, inbound bool) {
|
|
brontideConn := conn.(*brontide.Conn)
|
|
peerAddr := &lnwire.NetAddress{
|
|
IdentityKey: brontideConn.RemotePub(),
|
|
Address: conn.RemoteAddr().(*net.TCPAddr),
|
|
ChainNet: activeNetParams.Net,
|
|
}
|
|
|
|
// Now that we've established a connection, create a peer, and
|
|
// it to the set of currently active peers.
|
|
p, err := newPeer(conn, connReq, s, peerAddr, inbound)
|
|
if err != nil {
|
|
srvrLog.Errorf("unable to create peer %v", err)
|
|
if p.connReq != nil {
|
|
s.connMgr.Remove(p.connReq.ID())
|
|
}
|
|
p.Disconnect()
|
|
return
|
|
}
|
|
|
|
// TODO(roasbeef): update IP address for link-node
|
|
// * also mark last-seen, do it one single transaction?
|
|
|
|
if err := p.Start(); err != nil {
|
|
srvrLog.Errorf("unable to start peer: %v", err)
|
|
if p.connReq != nil {
|
|
s.connMgr.Remove(p.connReq.ID())
|
|
}
|
|
p.Disconnect()
|
|
return
|
|
}
|
|
|
|
s.newPeers <- p
|
|
}
|
|
|
|
// inboundPeerConnected initializes a new peer in response to a new inbound
|
|
// connection.
|
|
func (s *server) inboundPeerConnected(conn net.Conn) {
|
|
s.peersMtx.Lock()
|
|
defer s.peersMtx.Unlock()
|
|
|
|
srvrLog.Tracef("New inbound connection from %v", conn.RemoteAddr())
|
|
|
|
nodePub := conn.(*brontide.Conn).RemotePub()
|
|
|
|
// If we already have an outbound connection to this peer, simply drop
|
|
// the connection.
|
|
pubStr := string(nodePub.SerializeCompressed())
|
|
if _, ok := s.peersByPub[pubStr]; ok {
|
|
srvrLog.Errorf("Received inbound connection from peer %x, but "+
|
|
"already connected, dropping conn",
|
|
nodePub.SerializeCompressed())
|
|
conn.Close()
|
|
return
|
|
}
|
|
|
|
// However, if we receive an incoming connection from a peer we're
|
|
// attempting to maintain a persistent connection with then we need to
|
|
// cancel the ongoing connection attempts to ensure that we don't end
|
|
// up with a duplicate connecting to the same peer.
|
|
s.pendingConnMtx.RLock()
|
|
if connReq, ok := s.persistentConnReqs[pubStr]; ok {
|
|
s.connMgr.Remove(connReq.ID())
|
|
}
|
|
s.pendingConnMtx.RUnlock()
|
|
|
|
s.peerConnected(conn, nil, false)
|
|
}
|
|
|
|
// outboundPeerConnected initializes a new peer in response to a new outbound
|
|
// connection.
|
|
func (s *server) outboundPeerConnected(connReq *connmgr.ConnReq, conn net.Conn) {
|
|
s.peersMtx.Lock()
|
|
defer s.peersMtx.Unlock()
|
|
|
|
srvrLog.Tracef("Established connection to: %v", conn.RemoteAddr())
|
|
|
|
nodePub := conn.(*brontide.Conn).RemotePub()
|
|
|
|
// If we already have an inbound connection from this peer, simply drop
|
|
// the connection.
|
|
pubStr := string(nodePub.SerializeCompressed())
|
|
if _, ok := s.peersByPub[pubStr]; ok {
|
|
srvrLog.Errorf("Established outbound connection to peer %x, but "+
|
|
"already connected, dropping conn",
|
|
nodePub.SerializeCompressed())
|
|
s.connMgr.Remove(connReq.ID())
|
|
conn.Close()
|
|
return
|
|
}
|
|
|
|
s.peerConnected(conn, connReq, true)
|
|
}
|
|
|
|
// addPeer adds the passed peer to the server's global state of all active
|
|
// peers.
|
|
func (s *server) addPeer(p *peer) {
|
|
if p == nil {
|
|
return
|
|
}
|
|
|
|
// Ignore new peers if we're shutting down.
|
|
if atomic.LoadInt32(&s.shutdown) != 0 {
|
|
p.Stop()
|
|
return
|
|
}
|
|
|
|
// Track the new peer in our indexes so we can quickly look it up either
|
|
// according to its public key, or it's peer ID.
|
|
// TODO(roasbeef): pipe all requests through to the
|
|
// queryHandler/peerManager
|
|
s.peersMtx.Lock()
|
|
s.peersByID[p.id] = p
|
|
s.peersByPub[string(p.addr.IdentityKey.SerializeCompressed())] = p
|
|
s.peersMtx.Unlock()
|
|
|
|
// Once the peer has been added to our indexes, send a message to the
|
|
// channel router so we can synchronize our view of the channel graph
|
|
// with this new peer.
|
|
go s.chanRouter.SynchronizeNode(p.addr.IdentityKey)
|
|
}
|
|
|
|
// removePeer removes the passed peer from the server's state of all active
|
|
// peers.
|
|
func (s *server) removePeer(p *peer) {
|
|
s.peersMtx.Lock()
|
|
defer s.peersMtx.Unlock()
|
|
|
|
srvrLog.Debugf("removing peer %v", p)
|
|
|
|
if p == nil {
|
|
return
|
|
}
|
|
|
|
// As the peer is now finished, ensure that the TCP connection is
|
|
// closed and all of its related goroutines have exited.
|
|
if err := p.Stop(); err != nil {
|
|
peerLog.Errorf("unable to stop peer: %v", err)
|
|
}
|
|
|
|
// Ignore deleting peers if we're shutting down.
|
|
if atomic.LoadInt32(&s.shutdown) != 0 {
|
|
return
|
|
}
|
|
|
|
delete(s.peersByID, p.id)
|
|
delete(s.peersByPub, string(p.addr.IdentityKey.SerializeCompressed()))
|
|
}
|
|
|
|
// connectPeerMsg is a message requesting the server to open a connection to a
|
|
// particular peer. This message also houses an error channel which will be
|
|
// used to report success/failure.
|
|
type connectPeerMsg struct {
|
|
addr *lnwire.NetAddress
|
|
persistent bool
|
|
|
|
err chan error
|
|
}
|
|
|
|
// listPeersMsg is a message sent to the server in order to obtain a listing
|
|
// of all currently active channels.
|
|
type listPeersMsg struct {
|
|
resp chan []*peer
|
|
}
|
|
|
|
// openChanReq is a message sent to the server in order to request the
|
|
// initiation of a channel funding workflow to the peer with either the specified
|
|
// relative peer ID, or a global lightning ID.
|
|
type openChanReq struct {
|
|
targetPeerID int32
|
|
targetPubkey *btcec.PublicKey
|
|
|
|
// TODO(roasbeef): make enums in lnwire
|
|
channelType uint8
|
|
coinType uint64
|
|
|
|
localFundingAmt btcutil.Amount
|
|
remoteFundingAmt btcutil.Amount
|
|
|
|
pushAmt btcutil.Amount
|
|
|
|
numConfs uint32
|
|
|
|
updates chan *lnrpc.OpenStatusUpdate
|
|
err chan error
|
|
}
|
|
|
|
// queryHandler handles any requests to modify the server's internal state of
|
|
// all active peers, or query/mutate the server's global state. Additionally,
|
|
// any queries directed at peers will be handled by this goroutine.
|
|
//
|
|
// NOTE: This MUST be run as a goroutine.
|
|
func (s *server) queryHandler() {
|
|
go s.connMgr.Start()
|
|
|
|
out:
|
|
for {
|
|
select {
|
|
// New peers.
|
|
case p := <-s.newPeers:
|
|
s.addPeer(p)
|
|
|
|
// Finished peers.
|
|
case p := <-s.donePeers:
|
|
s.removePeer(p)
|
|
|
|
case bMsg := <-s.broadcastRequests:
|
|
ignore := bMsg.ignore
|
|
|
|
srvrLog.Debugf("Broadcasting %v messages", len(bMsg.msgs))
|
|
|
|
// Launch a new goroutine to handle the broadcast
|
|
// request, this allows us process this request
|
|
// asynchronously without blocking subsequent broadcast
|
|
// requests.
|
|
go func() {
|
|
s.peersMtx.RLock()
|
|
for _, sPeer := range s.peersByPub {
|
|
if ignore != nil &&
|
|
sPeer.addr.IdentityKey.IsEqual(ignore) {
|
|
|
|
srvrLog.Debugf("Skipping %v in broadcast",
|
|
ignore.SerializeCompressed())
|
|
|
|
continue
|
|
}
|
|
|
|
go func(p *peer) {
|
|
for _, msg := range bMsg.msgs {
|
|
p.queueMsg(msg, nil)
|
|
}
|
|
}(sPeer)
|
|
}
|
|
s.peersMtx.RUnlock()
|
|
|
|
bMsg.errChan <- nil
|
|
}()
|
|
case sMsg := <-s.sendRequests:
|
|
// TODO(roasbeef): use [33]byte everywhere instead
|
|
// * eliminate usage of mutexes, funnel all peer
|
|
// mutation to this goroutine
|
|
target := sMsg.target.SerializeCompressed()
|
|
|
|
srvrLog.Debugf("Attempting to send msgs %v to: %x",
|
|
len(sMsg.msgs), target)
|
|
|
|
// Launch a new goroutine to handle this send request,
|
|
// this allows us process this request asynchronously
|
|
// without blocking future send requests.
|
|
go func() {
|
|
s.peersMtx.RLock()
|
|
targetPeer, ok := s.peersByPub[string(target)]
|
|
if !ok {
|
|
s.peersMtx.RUnlock()
|
|
srvrLog.Errorf("unable to send message to %x, "+
|
|
"peer not found", target)
|
|
sMsg.errChan <- errors.New("peer not found")
|
|
return
|
|
}
|
|
s.peersMtx.RUnlock()
|
|
|
|
sMsg.errChan <- nil
|
|
|
|
for _, msg := range sMsg.msgs {
|
|
targetPeer.queueMsg(msg, nil)
|
|
}
|
|
}()
|
|
case query := <-s.queries:
|
|
switch msg := query.(type) {
|
|
case *connectPeerMsg:
|
|
s.handleConnectPeer(msg)
|
|
case *listPeersMsg:
|
|
s.handleListPeers(msg)
|
|
case *openChanReq:
|
|
s.handleOpenChanReq(msg)
|
|
}
|
|
case <-s.quit:
|
|
break out
|
|
}
|
|
}
|
|
|
|
s.connMgr.Stop()
|
|
|
|
s.wg.Done()
|
|
}
|
|
|
|
// handleListPeers sends a lice of all currently active peers to the original
|
|
// caller.
|
|
func (s *server) handleListPeers(msg *listPeersMsg) {
|
|
s.peersMtx.RLock()
|
|
|
|
peers := make([]*peer, 0, len(s.peersByID))
|
|
for _, peer := range s.peersByID {
|
|
peers = append(peers, peer)
|
|
}
|
|
|
|
s.peersMtx.RUnlock()
|
|
|
|
msg.resp <- peers
|
|
}
|
|
|
|
// handleConnectPeer attempts to establish a connection to the address enclosed
|
|
// within the passed connectPeerMsg. This function is *async*, a goroutine will
|
|
// be spawned in order to finish the request, and respond to the caller.
|
|
func (s *server) handleConnectPeer(msg *connectPeerMsg) {
|
|
addr := msg.addr
|
|
|
|
targetPub := string(msg.addr.IdentityKey.SerializeCompressed())
|
|
|
|
// Ensure we're not already connected to this
|
|
// peer.
|
|
s.peersMtx.RLock()
|
|
peer, ok := s.peersByPub[targetPub]
|
|
if ok {
|
|
s.peersMtx.RUnlock()
|
|
msg.err <- fmt.Errorf("already connected to peer: %v", peer)
|
|
return
|
|
}
|
|
s.peersMtx.RUnlock()
|
|
|
|
// If there's already a pending connection request for this pubkey,
|
|
// then we ignore this request to ensure we don't create a redundant
|
|
// connection.
|
|
s.pendingConnMtx.RLock()
|
|
if _, ok := s.persistentConnReqs[targetPub]; ok {
|
|
s.pendingConnMtx.RUnlock()
|
|
msg.err <- fmt.Errorf("connection attempt to %v is pending",
|
|
addr)
|
|
return
|
|
}
|
|
s.pendingConnMtx.RUnlock()
|
|
|
|
// If there's not already a pending or active connection to this node,
|
|
// then instruct the connection manager to attempt to establish a
|
|
// persistent connection to the peer.
|
|
srvrLog.Debugf("Connecting to %v", addr)
|
|
if msg.persistent {
|
|
go s.connMgr.Connect(&connmgr.ConnReq{
|
|
Addr: addr,
|
|
Permanent: true,
|
|
})
|
|
msg.err <- nil
|
|
} else {
|
|
// If we're not making a persistent connection, then we'll
|
|
// attempt to connect o the target peer, returning an error
|
|
// which indicates success of failure.
|
|
go func() {
|
|
// Attempt to connect to the remote node. If the we
|
|
// can't make the connection, or the crypto negotiation
|
|
// breaks down, then return an error to the caller.
|
|
conn, err := brontide.Dial(s.identityPriv, addr)
|
|
if err != nil {
|
|
msg.err <- err
|
|
return
|
|
}
|
|
|
|
s.outboundPeerConnected(nil, conn)
|
|
msg.err <- nil
|
|
}()
|
|
}
|
|
}
|
|
|
|
// handleOpenChanReq first locates the target peer, and if found hands off the
|
|
// request to the funding manager allowing it to initiate the channel funding
|
|
// workflow.
|
|
func (s *server) handleOpenChanReq(req *openChanReq) {
|
|
var (
|
|
targetPeer *peer
|
|
pubKeyBytes []byte
|
|
)
|
|
|
|
// If the user is targeting the peer by public key, then we'll need to
|
|
// convert that into a string for our map. Otherwise, we expect them to
|
|
// target by peer ID instead.
|
|
if req.targetPubkey != nil {
|
|
pubKeyBytes = req.targetPubkey.SerializeCompressed()
|
|
}
|
|
|
|
// First attempt to locate the target peer to open a channel with, if
|
|
// we're unable to locate the peer then this request will fail.
|
|
s.peersMtx.RLock()
|
|
if peer, ok := s.peersByID[req.targetPeerID]; ok {
|
|
targetPeer = peer
|
|
} else if peer, ok := s.peersByPub[string(pubKeyBytes)]; ok {
|
|
targetPeer = peer
|
|
}
|
|
s.peersMtx.RUnlock()
|
|
|
|
if targetPeer == nil {
|
|
req.err <- fmt.Errorf("unable to find peer nodeID(%x), "+
|
|
"peerID(%v)", pubKeyBytes, req.targetPeerID)
|
|
return
|
|
}
|
|
|
|
// Spawn a goroutine to send the funding workflow request to the funding
|
|
// manager. This allows the server to continue handling queries instead
|
|
// of blocking on this request which is exported as a synchronous
|
|
// request to the outside world.
|
|
// TODO(roasbeef): server semaphore to restrict num goroutines
|
|
go s.fundingMgr.initFundingWorkflow(targetPeer.addr, req)
|
|
}
|
|
|
|
// ConnectToPeer requests that the server connect to a Lightning Network peer
|
|
// at the specified address. This function will *block* until either a
|
|
// connection is established, or the initial handshake process fails.
|
|
func (s *server) ConnectToPeer(addr *lnwire.NetAddress,
|
|
perm bool) error {
|
|
|
|
errChan := make(chan error, 1)
|
|
|
|
s.queries <- &connectPeerMsg{
|
|
addr: addr,
|
|
persistent: perm,
|
|
err: errChan,
|
|
}
|
|
|
|
return <-errChan
|
|
}
|
|
|
|
// OpenChannel sends a request to the server to open a channel to the specified
|
|
// peer identified by ID with the passed channel funding paramters.
|
|
func (s *server) OpenChannel(peerID int32, nodeKey *btcec.PublicKey,
|
|
localAmt, pushAmt btcutil.Amount,
|
|
numConfs uint32) (chan *lnrpc.OpenStatusUpdate, chan error) {
|
|
|
|
errChan := make(chan error, 1)
|
|
updateChan := make(chan *lnrpc.OpenStatusUpdate, 1)
|
|
|
|
req := &openChanReq{
|
|
targetPeerID: peerID,
|
|
targetPubkey: nodeKey,
|
|
localFundingAmt: localAmt,
|
|
pushAmt: pushAmt,
|
|
numConfs: numConfs,
|
|
updates: updateChan,
|
|
err: errChan,
|
|
}
|
|
|
|
s.queries <- req
|
|
|
|
return updateChan, errChan
|
|
}
|
|
|
|
// Peers returns a slice of all active peers.
|
|
func (s *server) Peers() []*peer {
|
|
resp := make(chan []*peer, 1)
|
|
|
|
s.queries <- &listPeersMsg{resp}
|
|
|
|
return <-resp
|
|
}
|