TCP主动打开 之 第三次握手-发送ACK

Posted 2020-12-02 wanpengcoder

假定客户端执行主动打开，并且已经收到服务器发送的第二次握手包SYN+ACK，在经过一系列处理之后，客户端发送第三次握手包ACK到服务器；其流程比较简单，主要是分配skb，初始化ack包并发送；需要注意的一点是，标记纯ACK包竟然用了skb->truesize=2，好吧；

 1 void tcp_send_ack(struct sock *sk)
 2 {
 3     struct sk_buff *buff;
 4 
 5     /* If we have been reset, we may not send again. */
 6 
 7     /* 已经是CLOSE状态 */
 8     if (sk->sk_state == TCP_CLOSE)
 9         return;
10 
11     /* 拥塞控制处理 */
12     tcp_ca_event(sk, CA_EVENT_NON_DELAYED_ACK);
13 
14     /* We are not putting this on the write queue, so
15      * tcp_transmit_skb() will set the ownership to this
16      * sock.
17      */
18 
19     /* 分配skb，失败需要启用延迟ack定时器 */
20     buff = alloc_skb(MAX_TCP_HEADER,
21              sk_gfp_mask(sk, GFP_ATOMIC | __GFP_NOWARN));
22     if (unlikely(!buff)) {
23         inet_csk_schedule_ack(sk);
24         inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
25         inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
26                       TCP_DELACK_MAX, TCP_RTO_MAX);
27         return;
28     }
29 
30     /* Reserve space for headers and prepare control bits. */
31 
32     /* 预留头部空间，准备控制信息 */
33     skb_reserve(buff, MAX_TCP_HEADER);
34     tcp_init_nondata_skb(buff, tcp_acceptable_seq(sk), TCPHDR_ACK);
35 
36     /* We do not want pure acks influencing TCP Small Queues or fq/pacing
37      * too much.
38      * SKB_TRUESIZE(max(1 .. 66, MAX_TCP_HEADER)) is unfortunately ~784
39      */
40     /* 标记纯ack，做法是将skb->truesize设置为2 */
41     skb_set_tcp_pure_ack(buff);
42 
43     /* Send it off, this clears delayed acks for us. */
44     /* 记录时间戳，发送skb */
45     skb_mstamp_get(&buff->skb_mstamp);
46     tcp_transmit_skb(sk, buff, 0, (__force gfp_t)0);
47 }