use Fx to start and stop the host, swarm, autorelay and quicreuse (#2118

)

* config: refactor AutoNAT construction into separate method

* config: use a lifecycle hook to start listening on swarm addresses

* use Fx to construct the host

* add a test for constructing a routed host

* use Fx hooks to start the host

* config: use Fx lifecycle hooks to start AutoRelay and for PeerRouting

* basichost: don't close the swarm

The swarm is not constructed by the basic host, thus is shouldn't be
closed by it.

* config: use Fx hook to close the quicreuse connection manager

* test for goroutine leaks when starting/stopping fx

To do this, I've had to move a few leaky tests into a separate package.
I've filed a bug for the AutoNAT issue (#2743) but the "error on
startup" issue is going to require some pretty invasive changes (we need
to construct _then_ start).

* go fmt

* Ignore one more top function

* Typo

* Ignore any not top

---------

Co-authored-by: Sukun <sukunrt@gmail.com>
Co-authored-by: Steven Allen <steven@stebalien.com>
Co-authored-by: Marco Munizaga <git@marcopolo.io>

config/config.go

@@ -1,6 +1,7 @@
 package config
 
 import (
+	"context"
 	"crypto/rand"
 	"errors"
 	"fmt"
@@ -38,6 +39,7 @@ import (
 
 	ma "github.com/multiformats/go-multiaddr"
 	madns "github.com/multiformats/go-multiaddr-dns"
+	"github.com/quic-go/quic-go"
 	"go.uber.org/fx"
 	"go.uber.org/fx/fxevent"
 )
@@ -190,20 +192,11 @@ func (cfg *Config) makeSwarm(eventBus event.Bus, enableMetrics bool) (*swarm.Swa
 	return swarm.NewSwarm(pid, cfg.Peerstore, eventBus, opts...)
 }
 
-func (cfg *Config) addTransports(h host.Host) error {
-	swrm, ok := h.Network().(transport.TransportNetwork)
-	if !ok {
-		// Should probably skip this if no transports.
-		return fmt.Errorf("swarm does not support transports")
-	}
-
+func (cfg *Config) addTransports() ([]fx.Option, error) {
 	fxopts := []fx.Option{
 		fx.WithLogger(func() fxevent.Logger { return getFXLogger() }),
 		fx.Provide(fx.Annotate(tptu.New, fx.ParamTags(`name:"security"`))),
 		fx.Supply(cfg.Muxers),
-		fx.Supply(h.ID()),
-		fx.Provide(func() host.Host { return h }),
-		fx.Provide(func() crypto.PrivKey { return h.Peerstore().PrivKey(h.ID()) }),
 		fx.Provide(func() connmgr.ConnectionGater { return cfg.ConnectionGater }),
 		fx.Provide(func() pnet.PSK { return cfg.PSK }),
 		fx.Provide(func() network.ResourceManager { return cfg.ResourceManager }),
@@ -265,56 +258,38 @@ func (cfg *Config) addTransports(h host.Host) error {
 	if cfg.QUICReuse != nil {
 		fxopts = append(fxopts, cfg.QUICReuse...)
 	} else {
-		fxopts = append(fxopts, fx.Provide(quicreuse.NewConnManager)) // TODO: close the ConnManager when shutting down the node
+		fxopts = append(fxopts,
+			fx.Provide(func(key quic.StatelessResetKey, tokenGenerator quic.TokenGeneratorKey, _ *swarm.Swarm, lifecycle fx.Lifecycle) (*quicreuse.ConnManager, error) {
+				cm, err := quicreuse.NewConnManager(key, tokenGenerator)
+				if err != nil {
+					return nil, err
+				}
+				lifecycle.Append(fx.StopHook(cm.Close))
+				return cm, nil
+			}),
+		)
 	}
 
 	fxopts = append(fxopts, fx.Invoke(
 		fx.Annotate(
-			func(tpts []transport.Transport) error {
+			func(swrm *swarm.Swarm, tpts []transport.Transport) error {
 				for _, t := range tpts {
 					if err := swrm.AddTransport(t); err != nil {
 						return err
 					}
 				}
 				return nil
 			},
-			fx.ParamTags(`group:"transport"`),
+			fx.ParamTags("", `group:"transport"`),
 		)),
 	)
 	if cfg.Relay {
 		fxopts = append(fxopts, fx.Invoke(circuitv2.AddTransport))
 	}
-	app := fx.New(fxopts...)
-	if err := app.Err(); err != nil {
-		h.Close()
-		return err
-	}
-	return nil
+	return fxopts, nil
 }
 
-// NewNode constructs a new libp2p Host from the Config.
-//
-// This function consumes the config. Do not reuse it (really!).
-func (cfg *Config) NewNode() (host.Host, error) {
-	// If possible check that the resource manager conn limit is higher than the
-	// limit set in the conn manager.
-	if l, ok := cfg.ResourceManager.(connmgr.GetConnLimiter); ok {
-		err := cfg.ConnManager.CheckLimit(l)
-		if err != nil {
-			log.Warn(fmt.Sprintf("rcmgr limit conflicts with connmgr limit: %v", err))
-		}
-	}
-
-	eventBus := eventbus.NewBus(eventbus.WithMetricsTracer(eventbus.NewMetricsTracer(eventbus.WithRegisterer(cfg.PrometheusRegisterer))))
-	swrm, err := cfg.makeSwarm(eventBus, !cfg.DisableMetrics)
-	if err != nil {
-		return nil, err
-	}
-
-	if !cfg.DisableMetrics {
-		rcmgr.MustRegisterWith(cfg.PrometheusRegisterer)
-	}
-
+func (cfg *Config) newBasicHost(swrm *swarm.Swarm, eventBus event.Bus) (*bhost.BasicHost, error) {
 	h, err := bhost.NewHost(swrm, &bhost.HostOpts{
 		EventBus:             eventBus,
 		ConnManager:          cfg.ConnManager,
@@ -331,10 +306,8 @@ func (cfg *Config) NewNode() (host.Host, error) {
 		PrometheusRegisterer: cfg.PrometheusRegisterer,
 	})
 	if err != nil {
-		swrm.Close()
 		return nil, err
 	}
-
 	if cfg.Relay {
 		// If we've enabled the relay, we should filter out relay
 		// addresses by default.
@@ -345,60 +318,137 @@ func (cfg *Config) NewNode() (host.Host, error) {
 			return oldFactory(autorelay.Filter(addrs))
 		}
 	}
+	return h, nil
+}
 
-	if err := cfg.addTransports(h); err != nil {
-		h.Close()
-		return nil, err
+// NewNode constructs a new libp2p Host from the Config.
+//
+// This function consumes the config. Do not reuse it (really!).
+func (cfg *Config) NewNode() (host.Host, error) {
+	if cfg.EnableAutoRelay && !cfg.Relay {
+		return nil, fmt.Errorf("cannot enable autorelay; relay is not enabled")
+	}
+	// If possible check that the resource manager conn limit is higher than the
+	// limit set in the conn manager.
+	if l, ok := cfg.ResourceManager.(connmgr.GetConnLimiter); ok {
+		err := cfg.ConnManager.CheckLimit(l)
+		if err != nil {
+			log.Warn(fmt.Sprintf("rcmgr limit conflicts with connmgr limit: %v", err))
+		}
 	}
 
-	// TODO: This method succeeds if listening on one address succeeds. We
-	// should probably fail if listening on *any* addr fails.
-	if err := h.Network().Listen(cfg.ListenAddrs...); err != nil {
-		h.Close()
+	if !cfg.DisableMetrics {
+		rcmgr.MustRegisterWith(cfg.PrometheusRegisterer)
+	}
+
+	fxopts := []fx.Option{
+		fx.Provide(func() event.Bus {
+			return eventbus.NewBus(eventbus.WithMetricsTracer(eventbus.NewMetricsTracer(eventbus.WithRegisterer(cfg.PrometheusRegisterer))))
+		}),
+		fx.Provide(func(eventBus event.Bus, lifecycle fx.Lifecycle) (*swarm.Swarm, error) {
+			sw, err := cfg.makeSwarm(eventBus, !cfg.DisableMetrics)
+			if err != nil {
+				return nil, err
+			}
+			lifecycle.Append(fx.StopHook(sw.Close))
+			return sw, nil
+		}),
+		// Make sure the swarm constructor depends on the quicreuse.ConnManager.
+		// That way, the ConnManager will be started before the swarm, and more importantly,
+		// the swarm will be stopped before the ConnManager.
+		fx.Decorate(func(sw *swarm.Swarm, _ *quicreuse.ConnManager, lifecycle fx.Lifecycle) *swarm.Swarm {
+			lifecycle.Append(fx.Hook{
+				OnStart: func(context.Context) error {
+					// TODO: This method succeeds if listening on one address succeeds. We
+					// should probably fail if listening on *any* addr fails.
+					return sw.Listen(cfg.ListenAddrs...)
+				},
+				OnStop: func(context.Context) error {
+					return sw.Close()
+				},
+			})
+			return sw
+		}),
+		fx.Provide(cfg.newBasicHost),
+		fx.Provide(func(h *bhost.BasicHost, lifecycle fx.Lifecycle) host.Host {
+			lifecycle.Append(fx.StartHook(h.Start))
+			return h
+		}),
+		fx.Provide(func(h host.Host) peer.ID { return h.ID() }),
+		fx.Provide(func(h host.Host) crypto.PrivKey { return h.Peerstore().PrivKey(h.ID()) }),
+	}
+	transportOpts, err := cfg.addTransports()
+	if err != nil {
 		return nil, err
 	}
+	fxopts = append(fxopts, transportOpts...)
 
 	// Configure routing and autorelay
-	var router routing.PeerRouting
 	if cfg.Routing != nil {
-		router, err = cfg.Routing(h)
-		if err != nil {
-			h.Close()
-			return nil, err
-		}
+		fxopts = append(fxopts,
+			fx.Provide(cfg.Routing),
+			fx.Provide(func(h host.Host, router routing.PeerRouting) *routed.RoutedHost {
+				return routed.Wrap(h, router)
+			}),
+		)
 	}
 
 	// Note: h.AddrsFactory may be changed by relayFinder, but non-relay version is
 	// used by AutoNAT below.
-	var ar *autorelay.AutoRelay
-	addrF := h.AddrsFactory
 	if cfg.EnableAutoRelay {
-		if !cfg.Relay {
-			h.Close()
-			return nil, fmt.Errorf("cannot enable autorelay; relay is not enabled")
-		}
 		if !cfg.DisableMetrics {
 			mt := autorelay.WithMetricsTracer(
 				autorelay.NewMetricsTracer(autorelay.WithRegisterer(cfg.PrometheusRegisterer)))
 			mtOpts := []autorelay.Option{mt}
 			cfg.AutoRelayOpts = append(mtOpts, cfg.AutoRelayOpts...)
 		}
+		fxopts = append(fxopts,
+			fx.Invoke(func(h *bhost.BasicHost, lifecycle fx.Lifecycle) (*autorelay.AutoRelay, error) {
+				ar, err := autorelay.NewAutoRelay(h, cfg.AutoRelayOpts...)
+				if err != nil {
+					return nil, err
+				}
+				lifecycle.Append(fx.StartStopHook(ar.Start, ar.Close))
+				return ar, nil
+			}),
+		)
+	}
 
-		ar, err = autorelay.NewAutoRelay(h, cfg.AutoRelayOpts...)
-		if err != nil {
-			return nil, err
-		}
+	var bh *bhost.BasicHost
+	fxopts = append(fxopts, fx.Invoke(func(bho *bhost.BasicHost) { bh = bho }))
+
+	var rh *routed.RoutedHost
+	if cfg.Routing != nil {
+		fxopts = append(fxopts, fx.Invoke(func(bho *routed.RoutedHost) { rh = bho }))
+	}
+
+	app := fx.New(fxopts...)
+	if err := app.Start(context.Background()); err != nil {
+		return nil, err
 	}
 
+	if err := cfg.addAutoNAT(bh); err != nil {
+		rh.Close()
+		return nil, err
+	}
+
+	if cfg.Routing != nil {
+		return &closableRoutedHost{App: app, RoutedHost: rh}, nil
+	}
+	return &closableBasicHost{App: app, BasicHost: bh}, nil
+}
+
+func (cfg *Config) addAutoNAT(h *bhost.BasicHost) error {
+	addrF := h.AddrsFactory
 	autonatOpts := []autonat.Option{
 		autonat.UsingAddresses(func() []ma.Multiaddr {
 			return addrF(h.AllAddrs())
 		}),
 	}
 	if !cfg.DisableMetrics {
-		autonatOpts = append(autonatOpts,
-			autonat.WithMetricsTracer(
-				autonat.NewMetricsTracer(autonat.WithRegisterer(cfg.PrometheusRegisterer))))
+		autonatOpts = append(autonatOpts, autonat.WithMetricsTracer(
+			autonat.NewMetricsTracer(autonat.WithRegisterer(cfg.PrometheusRegisterer)),
+		))
 	}
 	if cfg.AutoNATConfig.ThrottleInterval != 0 {
 		autonatOpts = append(autonatOpts,
@@ -408,11 +458,11 @@ func (cfg *Config) NewNode() (host.Host, error) {
 	if cfg.AutoNATConfig.EnableService {
 		autonatPrivKey, _, err := crypto.GenerateEd25519Key(rand.Reader)
 		if err != nil {
-			return nil, err
+			return err
 		}
 		ps, err := pstoremem.NewPeerstore()
 		if err != nil {
-			return nil, err
+			return err
 		}
 
 		// Pull out the pieces of the config that we _actually_ care about.
@@ -438,14 +488,23 @@ func (cfg *Config) NewNode() (host.Host, error) {
 
 		dialer, err := autoNatCfg.makeSwarm(eventbus.NewBus(), false)
 		if err != nil {
-			h.Close()
-			return nil, err
+			return err
 		}
 		dialerHost := blankhost.NewBlankHost(dialer)
-		if err := autoNatCfg.addTransports(dialerHost); err != nil {
+		fxopts, err := autoNatCfg.addTransports()
+		if err != nil {
+			dialerHost.Close()
+			return err
+		}
+		fxopts = append(fxopts,
+			fx.Supply(dialerHost.ID()),
+			fx.Supply(dialer),
+			fx.Provide(func() crypto.PrivKey { return autonatPrivKey }),
+		)
+		app := fx.New(fxopts...)
+		if err := app.Err(); err != nil {
 			dialerHost.Close()
-			h.Close()
-			return nil, err
+			return err
 		}
 		// NOTE: We're dropping the blank host here but that's fine. It
 		// doesn't really _do_ anything and doesn't even need to be
@@ -458,25 +517,10 @@ func (cfg *Config) NewNode() (host.Host, error) {
 
 	autonat, err := autonat.New(h, autonatOpts...)
 	if err != nil {
-		h.Close()
-		return nil, fmt.Errorf("cannot enable autorelay; autonat failed to start: %v", err)
+		return fmt.Errorf("cannot enable autorelay; autonat failed to start: %v", err)
 	}
 	h.SetAutoNat(autonat)
-
-	// start the host background tasks
-	h.Start()
-
-	var ho host.Host
-	ho = h
-	if router != nil {
-		ho = routed.Wrap(h, router)
-	}
-	if ar != nil {
-		arh := autorelay.NewAutoRelayHost(ho, ar)
-		arh.Start()
-		ho = arh
-	}
-	return ho, nil
+	return nil
 }
 
 // Option is a libp2p config option that can be given to the libp2p constructor

config/host.go

@@ -0,0 +1,30 @@
+package config
+
+import (
+	"context"
+
+	basichost "github.com/libp2p/go-libp2p/p2p/host/basic"
+	routed "github.com/libp2p/go-libp2p/p2p/host/routed"
+
+	"go.uber.org/fx"
+)
+
+type closableBasicHost struct {
+	*fx.App
+	*basichost.BasicHost
+}
+
+func (h *closableBasicHost) Close() error {
+	_ = h.App.Stop(context.Background())
+	return h.BasicHost.Close()
+}
+
+type closableRoutedHost struct {
+	*fx.App
+	*routed.RoutedHost
+}
+
+func (h *closableRoutedHost) Close() error {
+	_ = h.App.Stop(context.Background())
+	return h.RoutedHost.Close()
+}

leaky_tests/README.md

@@ -0,0 +1 @@
+Tests that leak goroutines for various reasons. Mostly because libp2p node shutdown logic doesn't run if we fail to construct the node.