fix: resolve query lock bottleneck, relay finder untracked worker, and metrics pool memory leak

.github/workflows/go-test-template.yml

@@ -146,7 +146,7 @@ jobs:
         id: coverages
         run: echo "files=$(find . -type f -name 'module-coverage.txt' | tr -s '\n' ',' | sed 's/,$//')" >> $GITHUB_OUTPUT
       - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@54bcd8715eee62d40e33596ef5e8f0f48dbbccab # v4.1.0
+        uses: codecov/codecov-action@v4
         with:
           files: ${{ steps.coverages.outputs.files }}
           env_vars: OS=${{ matrix.os }}, GO=${{ steps.go.outputs.version }}

core/routing/query.go

@@ -49,7 +49,7 @@ type QueryEvent struct {
 
 type routingQueryKey struct{}
 type eventChannel struct {
-	mu  sync.Mutex
+	mu  sync.RWMutex
 	ctx context.Context
 	ch  chan<- *QueryEvent
 }
@@ -59,20 +59,22 @@ type eventChannel struct {
 func (e *eventChannel) waitThenClose() {
 	<-e.ctx.Done()
 	e.mu.Lock()
-	close(e.ch)
-	// 1. Signals that we're done.
-	// 2. Frees memory (in case we end up hanging on to this for a while).
-	e.ch = nil
+	if e.ch != nil {
+		close(e.ch)
+		// 1. Signals that we're done.
+		// 2. Frees memory (in case we end up hanging on to this for a while).
+		e.ch = nil
+	}
 	e.mu.Unlock()
 }
 
 // send sends an event on the event channel, aborting if either the passed or
 // the internal context expire.
 func (e *eventChannel) send(ctx context.Context, ev *QueryEvent) {
-	e.mu.Lock()
+	e.mu.RLock()
 	// Closed.
 	if e.ch == nil {
-		e.mu.Unlock()
+		e.mu.RUnlock()
 		return
 	}
 	// in case the passed context is unrelated, wait on both.
@@ -81,7 +83,7 @@ func (e *eventChannel) send(ctx context.Context, ev *QueryEvent) {
 	case <-e.ctx.Done():
 	case <-ctx.Done():
 	}
-	e.mu.Unlock()
+	e.mu.RUnlock()
 }
 
 // RegisterForQueryEvents registers a query event channel with the given

p2p/host/autorelay/relay_finder.go

@@ -207,7 +207,9 @@ func (rf *relayFinder) background(ctx context.Context) {
 	workTimer := rf.conf.clock.InstantTimer(rf.runScheduledWork(ctx, now, scheduledWork, peerSourceRateLimiter))
 	defer workTimer.Stop()
 
-	go rf.cleanupDisconnectedPeers(ctx)
+	rf.refCount.Go(func() {
+		rf.cleanupDisconnectedPeers(ctx)
+	})
 
 	// update addrs on starting the relay finder.
 	rf.updateAddrs()

p2p/metricshelper/pool.go

@@ -22,5 +22,6 @@ func PutStringSlice(s *[]string) {
 	if c := cap(*s); c < capacity {
 		panic(fmt.Sprintf("expected a string slice with capacity 8 or greater, got %d", c))
 	}
+	clear(*s)
 	stringPool.Put(s)
 }