commit 0af00f8105bc66d792f73076d770a5b30323a956 Author: Patrick McManus Date: Fri Aug 20 20:20:24 2010 -0400 insturment out_of_order delays diff --git a/include/linux/tcp.h b/include/linux/tcp.h index a778ee0..31a19cd 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -459,6 +459,13 @@ struct tcp_sock { * contains related tcp_cookie_transactions fields. */ struct tcp_cookie_values *cookie_values; + + struct + { + u32 rcv_isn; + u32 pkts_rcv; + u8 fin_reported; + } ofo; }; static inline struct tcp_sock *tcp_sk(const struct sock *sk) @@ -488,6 +495,10 @@ static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk) return (struct tcp_timewait_sock *)sk; } + +void ofo_notify(struct sock *sk, struct sk_buff *skb, int flags); + + #endif /* __KERNEL__ */ #endif /* _LINUX_TCP_H */ diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 65afeae..2eb4a19 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -2087,6 +2087,8 @@ int tcp_disconnect(struct sock *sk, int flags) tp->snd_cwnd_cnt = 0; tp->bytes_acked = 0; tp->window_clamp = 0; + tp->ofo.pkts_rcv = 0; + tp->ofo.fin_reported = 0; tcp_set_ca_state(sk, TCP_CA_Open); tcp_clear_retrans(tp); inet_csk_delack_init(sk); diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index 548d575..70ea32d 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -4321,6 +4321,7 @@ static void tcp_ofo_queue(struct sock *sk) if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) { SOCK_DEBUG(sk, "ofo packet was already received\n"); + ofo_notify(sk, skb, 0); __skb_unlink(skb, &tp->out_of_order_queue); __kfree_skb(skb); continue; @@ -4329,6 +4330,7 @@ static void tcp_ofo_queue(struct sock *sk) tp->rcv_nxt, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq); + ofo_notify(sk, skb, 0); __skb_unlink(skb, &tp->out_of_order_queue); __skb_queue_tail(&sk->sk_receive_queue, skb); tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq; @@ -4367,7 +4369,7 @@ static void tcp_data_queue(struct sock *sk, struct sk_buff *skb) if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) goto drop; - + skb_dst_drop(skb); __skb_pull(skb, th->doff * 4); @@ -4487,7 +4489,7 @@ drop: tp->rcv_nxt, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq); skb_set_owner_r(skb, sk); - + if (!skb_peek(&tp->out_of_order_queue)) { /* Initial out of order segment, build 1 SACK. */ if (tcp_is_sack(tp)) { @@ -4547,6 +4549,7 @@ drop: skb1); } } + if (!skb1) __skb_queue_head(&tp->out_of_order_queue, skb); else @@ -5229,6 +5232,9 @@ int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, struct tcp_sock *tp = tcp_sk(sk); int res; + if (TCP_SKB_CB(skb)->end_seq - TCP_SKB_CB(skb)->seq) + tp->ofo.pkts_rcv++; + /* * Header prediction. * The code loosely follows the one in the famous @@ -5506,7 +5512,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb, */ tp->rcv_nxt = TCP_SKB_CB(skb)->seq + 1; tp->rcv_wup = TCP_SKB_CB(skb)->seq + 1; - + tp->ofo.rcv_isn = TCP_SKB_CB(skb)->seq; + /* RFC1323: The window in SYN & SYN/ACK segments is * never scaled. */ @@ -5942,3 +5949,83 @@ EXPORT_SYMBOL(tcp_parse_md5sig_option); EXPORT_SYMBOL(tcp_rcv_established); EXPORT_SYMBOL(tcp_rcv_state_process); EXPORT_SYMBOL(tcp_initialize_rcv_mss); + +#include + +struct uofo_notifier +{ + struct timeval tstamp1; + struct timeval tstamp2; + __u32 seq; + __u16 orig_len; + __u16 actual_len; + __u32 daddr; + __u32 saddr; + __u16 dport; + __u16 sport; + __u32 pkts_rcv; + __u16 flags; /* 1 - FIN */ + + char data[0]; +}; + +static atomic_t cn_ofo_nextseq; +static struct cb_id cn_ofo_id = { CN_NETLINK_USERS + 100, 1 }; + +// flags - fin/rst 0x01 +void ofo_notify(struct sock *sk, struct sk_buff *skb, int flags) +{ + struct tcp_sock *tp = tcp_sk(sk); + struct uofo_notifier *notifier; + u16 actual = 0; + struct cn_msg *m; + + BUG_ON (flags & ~0x01); + + if (flags & 0x01) + { + if (tp->ofo.fin_reported) return; + tp->ofo.fin_reported = 1; + } + + m = kzalloc(sizeof(struct cn_msg) + sizeof(struct uofo_notifier) + actual, GFP_ATOMIC); + if (!m) return; + notifier = (struct uofo_notifier *) (m + 1); + + notifier->daddr = inet_sk(sk)->inet_daddr; + notifier->saddr = inet_sk(sk)->inet_saddr; + notifier->dport = inet_sk(sk)->inet_dport; + notifier->sport = inet_sk(sk)->inet_sport; + notifier->actual_len = actual; + + if (skb) + { + notifier->orig_len = TCP_SKB_CB(skb)->end_seq - TCP_SKB_CB(skb)->seq; + notifier->seq = TCP_SKB_CB(skb)->seq - tp->ofo.rcv_isn; + + skb_get_timestamp(skb, ¬ifier->tstamp1); + do_gettimeofday(¬ifier->tstamp2); + memcpy (notifier->data, skb->data, actual); + } + else + { + notifier->orig_len = 0; + notifier->seq = 0; + actual = 0; + do_gettimeofday(¬ifier->tstamp1); + memcpy (¬ifier->tstamp2, ¬ifier->tstamp1, sizeof (struct timeval)); + } + + notifier->pkts_rcv = tp->ofo.pkts_rcv; + + if (flags & 0x01) + notifier->flags |= 0x01; + + memcpy(&m->id, &cn_ofo_id, sizeof(m->id)); + m->seq = (flags & 0x01) ? atomic_read (&cn_ofo_nextseq) : atomic_inc_return (&cn_ofo_nextseq); + m->len = sizeof(struct uofo_notifier) + actual; + cn_netlink_send(m, 1, GFP_ATOMIC); + kfree(m); + + return; +} diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c index fe193e5..3da94fc 100644 --- a/net/ipv4/tcp_ipv4.c +++ b/net/ipv4/tcp_ipv4.c @@ -1661,6 +1661,9 @@ process: if (sk->sk_state == TCP_TIME_WAIT) goto do_time_wait; + if ((sk->sk_state == TCP_ESTABLISHED) && (th->fin || th->rst)) + ofo_notify(sk, skb, 1); + if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) { NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); goto discard_and_relse; diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c index 794c2e1..3448f53 100644 --- a/net/ipv4/tcp_minisocks.c +++ b/net/ipv4/tcp_minisocks.c @@ -273,6 +273,8 @@ void tcp_time_wait(struct sock *sk, int state, int timeo) const struct tcp_sock *tp = tcp_sk(sk); int recycle_ok = 0; + ofo_notify(sk, NULL, 1); + if (tcp_death_row.sysctl_tw_recycle && tp->rx_opt.ts_recent_stamp) recycle_ok = icsk->icsk_af_ops->remember_stamp(sk); @@ -418,7 +420,8 @@ struct sock *tcp_create_openreq_child(struct sock *sk, struct request_sock *req, newtp->rcv_wup = newtp->copied_seq = newtp->rcv_nxt = treq->rcv_isn + 1; - + newtp->ofo.rcv_isn = treq->rcv_isn; + newtp->snd_sml = newtp->snd_una = newtp->snd_nxt = newtp->snd_up = treq->snt_isn + 1 + tcp_s_data_size(oldtp);