Use blocking lock in worker_pool_mark — workers never hold mutex with GVL

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: samuel-williams-shopify

ext/io/event/worker_pool.c

@@ -109,11 +109,11 @@ static void worker_pool_mark(void *ptr)
 	// queue we need to mark from here to keep blocking_operation_value alive
 	// across any GC that might run during that window.
 	//
-	// We use trylock rather than lock: if a worker thread is currently
-	// dequeuing (which it does while holding the mutex but without the GVL),
-	// we skip this cycle.  The work will still be on the blocked fiber's
-	// C stack and the conservative scan will keep the VALUEs alive.
-	if (pthread_mutex_trylock(&pool->mutex) == 0) {
+	// Mark all queued work items. Workers only hold this mutex during brief
+	// non-Ruby queue operations (no GVL acquired while locked), so a blocking
+	// lock here is safe — no deadlock risk.
+	pthread_mutex_lock(&pool->mutex);
+	{
 		struct IO_Event_WorkerPool_Work *work = pool->work_queue;
 		while (work) {
 			rb_gc_mark(work->blocking_operation_value);
@@ -122,7 +122,8 @@ static void worker_pool_mark(void *ptr)
 			rb_gc_mark(work->fiber);
 			work = work->next;
 		}
-		pthread_mutex_unlock(&pool->mutex);
+	}
+	pthread_mutex_unlock(&pool->mutex);
 	}
 }