kernel TCP/IP implementation - about accept queue length - c

Question

I'm recently learning the TCP/IP implementation in linux kernel(version 4.4), and got really confused about the accept queue. I know there is a queue in struct inet_connection_sock which is called a accept queue:
struct inet_connection_sock {
...
/* #icsk_accept_queue: FIFO of established children */
struct request_sock_queue icsk_accept_queue;
...
}
and there is a qlen member in it, I think it is used to indicate the length of the queue.
struct request_sock_queue {
...
/* length of the queue? */
atomic_t qlen;
...
};
Here is what I think I know: when a LISTEN socket receives a SYN packet, in function tcp_conn_request, inet_csk_reqsk_queue_hash_add is called to put the newly created NEW_SYN_RECV sock into ehash table (not icsk_accept_queue)
void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
unsigned long timeout)
{
reqsk_queue_hash_req(req, timeout); // add to ehash table
inet_csk_reqsk_queue_added(sk); // increment icsk_accept_queue.qlen
}
But in this funcfion, inet_csk_reqsk_queue_added is called to increment icsk_accept_queue.qlen. My question is why increment qlen since nothing is inserted to icsk_accept_queue? Is it not the length of icsk_accept_queue?
Also, tcp_conn_request called inet_csk_reqsk_queue_add to add a fast open sock(if enabled) to icsk_accept_queue:
struct sock *inet_csk_reqsk_queue_add(struct sock *sk,
struct request_sock *req,
struct sock *child)
{
struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
spin_lock(&queue->rskq_lock);
if (unlikely(sk->sk_state != TCP_LISTEN)) {
inet_child_forget(sk, req, child);
child = NULL;
} else {
req->sk = child;
req->dl_next = NULL;
if (queue->rskq_accept_head == NULL)
queue->rskq_accept_head = req;
else
queue->rskq_accept_tail->dl_next = req;
queue->rskq_accept_tail = req;
sk_acceptq_added(sk); // increment sk.sk_ack_backlog
}
spin_unlock(&queue->rskq_lock);
return child;
}
But this function ended up incrementing sk.sk_ack_backlog (by calling sk_acceptq_added) instead of icsk_accept_queue.qlen. Why not?