(2) linux 3.x

http://blog.csdn.net/zhangskd/article/details/17576911

本文主要分析：服务器端如何构造和发送SYNACK段。

内核版本：3.6

Author：zhangskd @ csdn blog

发送入口

tcp_v4_send_synack()用于发送SYNACK段，在tcp_v4_conn_request()中被调用。

首先调用tcp_make_synack()构造SYNACK段，主要是构造TCP报头和初始化skb中的一些字段。

然后调用ip_build_and_send_pkt()添加IP报头后发送出去。

/* Send a SYN-ACK after having received a SYN.
 * This still operates on a request_sock only, not on a big socket.
 */

static int tcp_v4_send_synack(struct sock *sk, struct dst_entry *dst, struct request_sock *req,
     struct request_values *rvp, u16 queue_mapping, bool nocache)
{
    const struct inet_request_sock *ireq = inet_rsk(req);
    struct flowi4 fl4;
    int err = -1;
    struct sk_buff *skb;

    /* First, grab a route.
     * 获取路由缓存。
     */
    if (! dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL)
        return -1;

    /* 构造一个SYNACK段，初始化TCP首部和skb中的一些字段。*/
    skb = tcp_make_synack(sk, dst, req, rvp);

    if (skb) {
        /* 计算TCP报文的校验和 */
        __tcp_v4_send_check(skb, ireq->loc_addr, ireq->rmt_addr);
        skb_set_queue_mapping(skb, queue_mapping);

        /* 添加IP报头，并把此SYNACK段发送出去 */
        err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr, ireq->rmt_addr, ireq->opt);
        err = net_xmit_eval(err);
    }

    return err;
}

构造SYNACK段

构造一个SYNACK段，初始化TCP报头和skb中的一些字段。

/**
 * tcp_make_synack - Prepare a SYN-ACK.
 * sk: listener socket
 * dst: dst entry attached to the SYNACK
 * req: request_sock pointer
 * rvp: request_values pointer
 *
 * Allocate one skb and build a SYNACK packet.
 * @dst is consumed: Caller should not use it again.
 */

struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst, struct request_sock *req,
     struct request_values *rvp)
{
    struct tcp_out_options opts;
    struct tcp_extend_values *xvp = tcp_xv(rvp);
    struct inet_request_sock *ireq = inet_rsk(req);
    struct tcp_sock *tp = tcp_sk(sk);
    const struct tcp_cookie_values *cvp = tp->cookie_values;
    struct tcphdr *th;
    struct sk_buff *skb;
    struct tcp_md5sig_key *md5;
    int tcp_header_size;
    int mss;
    int s_data_desired = 0;

    if (cvp != NULL && cvp->s_data_constant && cvp->s_data_desired)
        s_data_desired = cvp->s_data_desired;

    /* 申请一个skb，用于发送SYNACK */
    skb = alloc_skb(MAX_TCP_HEADER + 15 + s_data_desired, sk_gfp_atomic(sk, GFP_ATOMIC));
    if (unlikely(! skb)) {
        dst_release(dst);
        return NULL;
    }

    /* Reserve space for headers.
     * 拓展headroom，为MAC、IP、TCP协议头预留空间。
     */
    skb_reserve(skb, MAX_TCP_HEADER);
    skb_dst_set(skb, dst); /* 保存路由缓存的地址 */

    /* 从路由缓存中获取本端的通告MSS */
    mss = dst_metric_advmss(dst);
    if (tp->rx_opt.user_mss && tp->rx_opt.user_mss < mss)
        mss = tp->rx_opt.user_mss; /* 如果用户使用了TCP_MAXSEG选项 */

    /* ignored for retransmitted syns.
     * 设置接收窗口的初始值、窗口扩大因子、通告窗口的上限。
     */
    if (req->rcv_wnd == 0) {
        __u8 rcv_wscale;
        /* 最大的通告窗口 */
        req->window_clamp = tp->window_clamp ? : dst_metric(dst, RTAX_WINDOW);

        /* 如果用户使用了SO_RCVBUF选项做限制 */
        if (sk->sk_userlocks & SOCK_RCVBUF_LOCK &&
             (req->window_clamp > tcp_full_space(sk) || req->window_clamp == 0))
            req->window_clamp = tcp_full_space(sk); /* 3/4 * sk->sk_rcvbuf */

        /* 获取接收窗口的初始值、窗口扩大因子和接收窗口的上限 */
        tcp_select_initial_window(tcp_full_space(sk), mss - (ireq->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0),
            &req->rcv_wnd, &req->window_clamp, ireq->wscale_ok, &rcv_wscale, dst_metric(dst, RTAX_INITRWND));
        ireq->rcv_wscale = rcv_wscale;
    }
    memset(&opts, 0, sizeof(opts));

#ifdef CONFIG_SYN_COOKIES
    /* 如果SYNACK段使用了SYN COOKIE，并且使用时间戳选项，
     * 则把TCP选项信息保存在SYNACK段中tsval的低6位。
     */
    if (unlikely(req->cookie_ts))
        TCP_SKB_CB(skb)->when = cookie_init_timestamp(req);
    else
#endif
    TCP_SKB_CB(skb)->when = tcp_time_stamp;

    /* TCP首部和选项的长度，赋值TCP选项实例tcp_out_options */
    tcp_header_size = tcp_synack_options(sk, req, mss, skb, &opts, &md5, xvp) + sizeof(*th);

    /* 向headroom扩展，使data room包含TCP首部和选项 */
    skb_push(skb, tcp_header_size);
    skb_reset_transport_header(skb);

    th = tcp_hdr(skb);
    memset(th, 0, sizeof(struct tcphdr));
    th->syn = 1;
    th->ack = 1;
    /* 如果连接支持ECN，把th->ece置1 */
    TCP_ECN_make_synack(req, th);
    th->source = ireq->loc_port; /* 源端口 */
    th->dest = ireq->rmt_port; /* 目的端口 */
    /* 初始化skb中的一些控制字段 */
    tcp_init_nondata_skb(skb, tcp_rsk(req)->snt_isn, TCPHDR_SYN | TCPHDR_ACK);

    /* TCPCT，新内核已废弃此选项，不做分析 */
    if (OPTION_COOKIE_EXTENSION & opts.options) {
        ...
    }

    th->seq = htonl(TCP_SKB_CB(skb)->seq); /* 序号 */
    th->ack_seq = htonl(tcp_rsk(seq)->rcv_isn + 1); /* 确认序号 */
    /* RFC1323: The window in SYN & SYN/ACK segments is never scaled. */
    th->window = htons(min(req->rcv_wnd, 65535U)); /* 初始接收窗口 */

    /* 把TCP选项实例tcp_out_options写到skb中 */
    tcp_options_write((__u32 *) (th + 1), tp, &opts);
    th->doff = (tcp_header_size >> 2); /* TCP首部长度 */
    TCP_ADD_STATS(sock_net(sk), TCP_MIB_OUTSEGS, tcp_skb_pcount(skb));

#ifdef CONFIG_TCP_MD5SIG
    /* Okay, we have all we need - do the md5 hash if needed */
    if (md5)
        tcp_rsk(req)->af_specific->calc_md5_hash(opts.hash_location, md5, NULL, req, skb);
#endif

    return skb;
}

/* MAC层、IP层、TCP层，首部和选项的最大长度 */
#define MAX_TCP_HEADER (128 + MAX_HEADER)
#define MAX_TCP_OPTION_SPACE 40

static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
{
    /* Allow access to emergency reserves */
    return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
}

如果SYNACK段使用SYN Cookie，并且使用时间戳选项，则把TCP选项信息保存在SYNACK段

中tsval的低6位。

/* When syncookies are in effect and tcp timestamps are enabled we encode tcp options
 * in the lower bits of the timestamp value that will be sent in the syn-ack.
 * Since subsequent timestamps use the normal tcp_time_stamp value, we must make
 * sure that the resulting initial timestamp is <= tcp_time_stamp.
 */
__u32 cookie_init_timestamp(struct request_sock *req)
{
    struct inet_request_sock *ireq;
    u32 ts, ts_now = tcp_time_stamp;
    u32 options = 0;
    ireq = inet_rsk(req);

    options = ireq->wscale_ok ? ireq->snd_wscale : 0xf;
    options |= ireq->sack_ok << 4;
    options |= ireq->ecn_ok << 5;

    ts = ts_now & ~TSMASK;
    ts |= options;

    if (ts > ts_now) {
        ts >>= TSBITS;
        ts--;
        ts <<= TSBITS;
        ts |= options;
    }
    return ts;
}

#define TSBITS 6
#define TSMASK (((__u32) 1 << TSBITS) - 1)

TCP Cookie Transaction (TCPCT) 选项功能类似与SYN Cookie，是2009年加入内核，

2013/3从内核中移除，本文不对其进行分析。

#define OPTION_SACK_ADVERTISE (1 << 0)
#define OPTION_TS (1 << 1)
#define OPTION_MD5 (1 << 2)
#define OPTION_WSCALE (1 << 3)
#define OPTION_COOKIE_EXTENSION (1 << 4)
#define OPTION_FAST_OPEN_COOKIE (1 << 8)

/* TCP选项实例，用于发送。*/
struct tcp_out_options {
    u16 options; /* bit field of OPTION_* */
    u16 mss; /* 0 to disable，Max Segment Size选项 */
    u8 ws; /* window scale, 0 to disable，Window Scaling选项 */
    u8 num_sack_blocks; /* number of SACK blocks to include */
    u8 hash_size; /* bytes in hash_location */
    __u8 *hash_location; /* temporary pointer, overloaded */
    __u32 tsval, tsecr; /* need to include OPTION_TS，时间戳 */
    struct tcp_fastopen_cookie *fastopen_cookie; /* Fast open cookie，Fast Open选项 */
};

赋值TCP选项实例tcp_out_options，用于构造SYNACK段。

/* Set up TCP options for SYN-ACKs. */
static unsigned int tcp_synack_options(struct sock *sk, struct request_sock *req,
    unsigned int mss, struct sk_buff *skb, struct tcp_out_options *opts,
    struct tcp_md5sig_key **md5, struct tcp_extend_values *xvp)
{
    struct inet_request_sock *ireq = inet_rsk(req);
    unsigned int remaining = MAX_TCP_OPTION_SPACE;
    u8 cookie_plus = (xvp != NULL && ! xvp->cookie_out_never) ? xvp->cookie_plus : 0;

#define CONFIG_TCP_MD5SIG
    *md5 = tcp_rsk(req)->af_specific->md5_lookup(sk, req);
    if (*md5) {
        opts->options |= OPTION_MD5;
        remaining -= TCPOLEN_MD5SIG_ALIGNED;

        /* We can't fit any SACK blocks in a packet with MD5 + TS options.
         * There was discussion about disabling SACK rather than TS in order to
         * fit in better with old, buggy kernels, but that was deemed to be unnecessary.
         */
        ireq->tstamp_ok &= ! ireq->sack_ok;
    }
#else
    *md5 = NULL;
#endif

    /* We always send an MSS option. */
    opt->mss = mss; /* Max Segment Size选项 */
    remaining -= TCPOLEN_MSS_ALIGNED;

    if (likely(ireq->wscale_ok)) { /* Window Scaling选项 */
        opts->ws = ireq->rcv_wscale;
        opts->options |= OPTION_WSCALE;
        remaining -= TCPOLEN_WSCALE_ALIGNED;
    }

    if (likely(ireq->tstamp_ok)) { /* 时间戳选项 */
        opts->options |= OPTION_TS;
        opts->tsval = TCP_SKB_CB(skb)->when;
        opts->tsecr = req->ts_recent;
        remaining -= TCPOLEN_TSTAMP_ALIGNED;
    }

    if (likely(ireq->sack_ok)) { /* SACK Permit选项 */
        opts->options |= OPTION_SACK_ADVERTISE;
        if (unlikely(! ireq->tstamp_ok))
            remaining -= TCPOLEN_SACKPERM_ALIGNED;
    }

    /* TCPCT选项，新内核以废弃 */
    if (*md5 == NULL && ireq->tstamp_ok && cookie_plus > TCPOLEN_COOKIE_BASE) {
        ...
    }

    return MAX_TCP_OPTION_SPACE - remaining; /* TCP选项长度 */
}

初始化不携带数据skb的一些控制字段。

/* Constructs common control bits of non-data skb. If SYN/FIN is present,
 * auto increment end seqno.
 */

static void tcp_init_nondata_skb(struct sk_buff *skb, u32 seq, u8 flags)
{
    skb->ip_summed = CHECKSUM_PARTIAL; /* 由硬件计算协议首部和数据的校验和。*/
    skb->csum = 0;

    TCP_SKB_CB(skb)->tcp_flags = flags;
    TCP_SKB_CB(skb)->sacked = 0;

    skb_shinfo(skb)->gso_segs = 1;
    skb_shinfo(skb)->gso_size = 0;
    skb_shinfo(skb)->gso_type = 0;

    TCP_SKB_CB(skb)->seq = seq;
    if (flags & (TCPHDR_SYN | TCPHDR_FIN))
        seq++;
    TCP_SKB_CB(skb)->end_seq = seq;
}

发送到IP层

TCP报头中的校验和字段还没赋值，用__tcp_v4_send_check()来计算。

static void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr)
{
    struct tcphdr *th = tcp_hdr(skb);

    if (skb->ip_summed == CHECKSUM_PARTIAL) {
        /* 只计算伪首部，TCP报头和TCP数据的累加由硬件完成 */
        th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0);
        skb->csum_start = skb_transport_header(skb) - skb->head; /* 下次计算开始位置的偏移 */
        skb->csum_offset = offsetof(struct tcphdr, check); /* 校验和值在TCP首部的偏移 */

    } else {
        /* tcp_v4_check()累加伪首部，获取最终的校验和。
         * csum_partial()累加TCP报头。
         * skb->csum是TCP数据部分的累加，这是在从用户空间复制时顺便累加的。
         */
        th->check = tcp_v4_check(skb->len, saddr, daddr, csum_partial(th, th->doff << 2, skb->csum));
    }
}

给skb添加一个IP报头，然后发送出去。

/* Add an ip header to a skbuff and send it out. */
int ip_build_and_send_pkt(struct sk_buff *skb, struct sock *sk, __be32 saddr, __be32 daddr,
         struct ip_options_rcu *opt)
{
    struct inet_sock *inet = inet_sk(sk);
    struct rtable *rt = skb_rtable(skb);
    struct iphdr *iph;

     /* 向headroom扩展，使data room包含IP首部和选项 */
    skb_push(skb, sizeof(struct iphdr) + (opt ? opt->opt.optlen : 0));
    skb_reset_network_header(skb);

    iph = ip_hdr(skb);
    iph->version = 4;
    iph->ihl = 5;
    ip->tos = inet->tos;

    if (ip_dont_fragment(sk, &rt->dst))
        iph->frag_off = htons(IP_DF); /* 设置不允许分片标志 */
    else
        iph->frag_off = 0;

    iph->ttl = ip_select_ttl(inet, &rt->dst);
    iph->daddr = (opt && opt->opt.srr ? opt->opt.faddr : daddr);
    iph->saddr = saddr;
    iph->protocol = sk->sk_protocol;
    ip_select_ident(iph, &rt->dst, sk);

    if (opt && opt->opt.optlen) {
        iph->ihl += opt->opt.optlen >> 2;
        ip_options_build(skb, &opt->opt, daddr, rt, 0);
    }

    skb->priority = sk->sk_priority;
    skb->mark = sk->sk_mark;

    /* Send it out. */
    return ip_local_out(skb); /* 位于NF_INET_LOCAT_OUT之前 */

}