p2p: enable discovery after all local protocols were registered (#4546)

related: #4542

libp2p communicates the set of enabled protocols after completing secure handshake.
if some of the protocols were not registered before completing initial identify protocol round, it will result in errors
as in attached github issue.

node/node.go

@@ -905,7 +905,9 @@ func (app *App) initServices(ctx context.Context, poetClients []activation.PoetP
 			peersync.WithConfig(app.Config.TIME.Peersync),
 		)
 	}
-
+	if err := app.host.Start(); err != nil {
+		return err
+	}
 	return nil
 }
 

p2p/peerexchange/discovery.go

@@ -44,6 +44,7 @@ type Discovery struct {
 	host   host.Host
 	cfg    Config
 
+	ctx    context.Context
 	cancel context.CancelFunc
 	eg     errgroup.Group
 
@@ -60,6 +61,7 @@ func New(logger log.Log, h host.Host, config Config) (*Discovery, error) {
 		cfg:    config,
 		logger: logger,
 		host:   h,
+		ctx:    ctx,
 		cancel: cancel,
 		book:   book.New(),
 	}
@@ -103,10 +105,15 @@ func New(logger log.Log, h host.Host, config Config) (*Discovery, error) {
 			return nil, err
 		}
 	}
-	d.scanPeers(ctx)
 	return d, nil
 }
 
+// StartScan starts background goroutine to connect with known peers
+// and scan for new ones.
+func (d *Discovery) StartScan() {
+	d.scanPeers(d.ctx)
+}
+
 func (d *Discovery) recovery(ctx context.Context) error {
 	fpath := filepath.Join(d.cfg.DataDir, peersFile)
 	f, err := os.Open(fpath)
@@ -203,7 +210,6 @@ func (d *Discovery) scanPeers(ctx context.Context) {
 				}
 				return err
 			}
-
 		}
 	})
 }

p2p/peerexchange/discovery_test.go

@@ -38,6 +38,7 @@ func TestDiscovery_CrawlMesh(t *testing.T) {
 		cfg.Bootnodes = append(cfg.Bootnodes, bootnode.String())
 		instance, err := New(logger, h, cfg)
 		require.NoError(t, err)
+		instance.StartScan()
 		t.Cleanup(instance.Stop)
 	}
 

p2p/upgrade.go

@@ -2,7 +2,9 @@ package p2p
 
 import (
 	"context"
+	"errors"
 	"fmt"
+	"sync"
 	"time"
 
 	"github.com/libp2p/go-libp2p/core/host"
@@ -53,6 +55,11 @@ type Host struct {
 	cfg    Config
 	logger log.Log
 
+	closed struct {
+		sync.Mutex
+		closed bool
+	}
+
 	host.Host
 	*pubsub.PubSub
 
@@ -131,8 +138,24 @@ func (fh *Host) PeerCount() uint64 {
 	return uint64(len(fh.Host.Network().Peers()))
 }
 
+func (fh *Host) Start() error {
+	fh.closed.Lock()
+	defer fh.closed.Unlock()
+	if fh.closed.closed {
+		return errors.New("p2p: closed")
+	}
+	fh.discovery.StartScan()
+	return nil
+}
+
 // Stop background workers and release external resources.
 func (fh *Host) Stop() error {
+	fh.closed.Lock()
+	defer fh.closed.Unlock()
+	if fh.closed.closed {
+		return errors.New("p2p: closed")
+	}
+	fh.closed.closed = true
 	fh.discovery.Stop()
 	if err := fh.Host.Close(); err != nil {
 		return fmt.Errorf("failed to close libp2p host: %w", err)