p2p: tighten up and test PeerManager (#6034)

This tightens up the `PeerManager` and related code, adds a ton of tests, and fixes a bunch of inconsistencies and bugs.

evidence/reactor.go

@@ -55,7 +55,7 @@ type Reactor struct {
 	evpool      *Pool
 	eventBus    *types.EventBus
 	evidenceCh  *p2p.Channel
-	peerUpdates *p2p.PeerUpdatesCh
+	peerUpdates *p2p.PeerUpdates
 	closeCh     chan struct{}
 
 	peerWG sync.WaitGroup
@@ -70,7 +70,7 @@ type Reactor struct {
 func NewReactor(
 	logger log.Logger,
 	evidenceCh *p2p.Channel,
-	peerUpdates *p2p.PeerUpdatesCh,
+	peerUpdates *p2p.PeerUpdates,
 	evpool *Pool,
 ) *Reactor {
 	r := &Reactor{
@@ -196,9 +196,8 @@ func (r *Reactor) processEvidenceCh() {
 			if err := r.handleMessage(r.evidenceCh.ID(), envelope); err != nil {
 				r.Logger.Error("failed to process message", "ch_id", r.evidenceCh.ID(), "envelope", envelope, "err", err)
 				r.evidenceCh.Error() <- p2p.PeerError{
-					PeerID:   envelope.From,
-					Err:      err,
-					Severity: p2p.PeerErrorSeverityLow,
+					NodeID: envelope.From,
+					Err:    err,
 				}
 			}
 
@@ -221,7 +220,7 @@ func (r *Reactor) processEvidenceCh() {
 //
 // REF: https://github.com/tendermint/tendermint/issues/4727
 func (r *Reactor) processPeerUpdate(peerUpdate p2p.PeerUpdate) {
-	r.Logger.Debug("received peer update", "peer", peerUpdate.PeerID, "status", peerUpdate.Status)
+	r.Logger.Debug("received peer update", "peer", peerUpdate.NodeID, "status", peerUpdate.Status)
 
 	r.mtx.Lock()
 	defer r.mtx.Unlock()
@@ -240,21 +239,21 @@ func (r *Reactor) processPeerUpdate(peerUpdate p2p.PeerUpdate) {
 		// a new done channel so we can explicitly close the goroutine if the peer
 		// is later removed, we increment the waitgroup so the reactor can stop
 		// safely, and finally start the goroutine to broadcast evidence to that peer.
-		_, ok := r.peerRoutines[peerUpdate.PeerID]
+		_, ok := r.peerRoutines[peerUpdate.NodeID]
 		if !ok {
 			closer := tmsync.NewCloser()
 
-			r.peerRoutines[peerUpdate.PeerID] = closer
+			r.peerRoutines[peerUpdate.NodeID] = closer
 			r.peerWG.Add(1)
-			go r.broadcastEvidenceLoop(peerUpdate.PeerID, closer)
+			go r.broadcastEvidenceLoop(peerUpdate.NodeID, closer)
 		}
 
-	case p2p.PeerStatusDown, p2p.PeerStatusRemoved, p2p.PeerStatusBanned:
+	case p2p.PeerStatusDown:
 		// Check if we've started an evidence broadcasting goroutine for this peer.
 		// If we have, we signal to terminate the goroutine via the channel's closure.
 		// This will internally decrement the peer waitgroup and remove the peer
 		// from the map of peer evidence broadcasting goroutines.
-		closer, ok := r.peerRoutines[peerUpdate.PeerID]
+		closer, ok := r.peerRoutines[peerUpdate.NodeID]
 		if ok {
 			closer.Close()
 		}

evidence/reactor_test.go

@@ -42,7 +42,7 @@ type reactorTestSuite struct {
 	evidencePeerErrCh chan p2p.PeerError
 
 	peerUpdatesCh chan p2p.PeerUpdate
-	peerUpdates   *p2p.PeerUpdatesCh
+	peerUpdates   *p2p.PeerUpdates
 }
 
 func setup(t *testing.T, logger log.Logger, pool *evidence.Pool, chBuf uint) *reactorTestSuite {
@@ -224,18 +224,18 @@ func TestReactorMultiDisconnect(t *testing.T) {
 
 	primary.peerUpdatesCh <- p2p.PeerUpdate{
 		Status: p2p.PeerStatusUp,
-		PeerID: secondary.peerID,
+		NodeID: secondary.peerID,
 	}
 
 	// Ensure "disconnecting" the secondary peer from the primary more than once
 	// is handled gracefully.
 	primary.peerUpdatesCh <- p2p.PeerUpdate{
 		Status: p2p.PeerStatusDown,
-		PeerID: secondary.peerID,
+		NodeID: secondary.peerID,
 	}
 	primary.peerUpdatesCh <- p2p.PeerUpdate{
 		Status: p2p.PeerStatusDown,
-		PeerID: secondary.peerID,
+		NodeID: secondary.peerID,
 	}
 }
 
@@ -276,7 +276,7 @@ func TestReactorBroadcastEvidence(t *testing.T) {
 	for _, suite := range secondaries {
 		primary.peerUpdatesCh <- p2p.PeerUpdate{
 			Status: p2p.PeerStatusUp,
-			PeerID: suite.peerID,
+			NodeID: suite.peerID,
 		}
 	}
 
@@ -327,7 +327,7 @@ func TestReactorBroadcastEvidence_Lagging(t *testing.T) {
 	for _, suite := range secondaries {
 		primary.peerUpdatesCh <- p2p.PeerUpdate{
 			Status: p2p.PeerStatusUp,
-			PeerID: suite.peerID,
+			NodeID: suite.peerID,
 		}
 	}
 
@@ -378,7 +378,7 @@ func TestReactorBroadcastEvidence_Pending(t *testing.T) {
 	// add the secondary reactor as a peer to the primary reactor
 	primary.peerUpdatesCh <- p2p.PeerUpdate{
 		Status: p2p.PeerStatusUp,
-		PeerID: secondary.peerID,
+		NodeID: secondary.peerID,
 	}
 
 	// The secondary reactor should have received all the evidence ignoring the
@@ -438,7 +438,7 @@ func TestReactorBroadcastEvidence_Committed(t *testing.T) {
 	// add the secondary reactor as a peer to the primary reactor
 	primary.peerUpdatesCh <- p2p.PeerUpdate{
 		Status: p2p.PeerStatusUp,
-		PeerID: secondary.peerID,
+		NodeID: secondary.peerID,
 	}
 
 	// The secondary reactor should have received all the evidence ignoring the
@@ -487,7 +487,7 @@ func TestReactorBroadcastEvidence_FullyConnected(t *testing.T) {
 			if suiteI.peerID != suiteJ.peerID {
 				suiteI.peerUpdatesCh <- p2p.PeerUpdate{
 					Status: p2p.PeerStatusUp,
-					PeerID: suiteJ.peerID,
+					NodeID: suiteJ.peerID,
 				}
 			}
 		}
@@ -530,7 +530,7 @@ func TestReactorBroadcastEvidence_RemovePeer(t *testing.T) {
 	// add the secondary reactor as a peer to the primary reactor
 	primary.peerUpdatesCh <- p2p.PeerUpdate{
 		Status: p2p.PeerStatusUp,
-		PeerID: secondary.peerID,
+		NodeID: secondary.peerID,
 	}
 
 	// have the secondary reactor receive only half the evidence
@@ -539,7 +539,7 @@ func TestReactorBroadcastEvidence_RemovePeer(t *testing.T) {
 	// disconnect the peer
 	primary.peerUpdatesCh <- p2p.PeerUpdate{
 		Status: p2p.PeerStatusDown,
-		PeerID: secondary.peerID,
+		NodeID: secondary.peerID,
 	}
 
 	// Ensure the secondary only received half of the evidence before being