@@ -358,7 +358,7 @@ struct tcp_sock {
*/
int lost_cnt_hint;
- int retransmit_cnt_hint;
+ u32 retransmit_high; /* L-bits may be on up to this seqno */
u32 lost_retrans_low; /* Sent seq after any rxmit (lowest) */
@@ -1044,6 +1044,9 @@ static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
tp->forward_skb_hint = NULL;
}
+extern void tcp_verify_retransmit_hint(struct tcp_sock *tp,
+ struct sk_buff *skb);
+
/* MD5 Signature */
struct crypto_hash;
@@ -984,12 +984,16 @@ static void tcp_update_reordering(struct sock *sk, const int metric,
* retransmitted past LOST markings in the first place? I'm not fully sure
* about undo and end of connection cases, which can cause R without L?
*/
-static void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb)
+void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb)
{
- if ((tp->retransmit_skb_hint != NULL) &&
+ if ((tp->retransmit_skb_hint == NULL) ||
before(TCP_SKB_CB(skb)->seq,
TCP_SKB_CB(tp->retransmit_skb_hint)->seq))
- tp->retransmit_skb_hint = NULL;
+ tp->retransmit_skb_hint = skb;
+
+ if (!tp->lost_out ||
+ after(TCP_SKB_CB(skb)->end_seq, tp->retransmit_high))
+ tp->retransmit_high = TCP_SKB_CB(skb)->end_seq;
}
static void __tcp_skb_mark_lost(struct tcp_sock *tp, struct sk_buff *skb)
@@ -1888,6 +1892,7 @@ static void tcp_enter_frto_loss(struct sock *sk, int allowed_segments, int flag)
if (!(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)) {
TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
tp->lost_out += tcp_skb_pcount(skb);
+ tp->retransmit_high = TCP_SKB_CB(skb)->end_seq;
}
}
tcp_verify_left_out(tp);
@@ -1972,6 +1977,7 @@ void tcp_enter_loss(struct sock *sk, int how)
TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_ACKED;
TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
tp->lost_out += tcp_skb_pcount(skb);
+ tp->retransmit_high = TCP_SKB_CB(skb)->end_seq;
}
}
tcp_verify_left_out(tp);
@@ -1851,6 +1851,7 @@ void tcp_simple_retransmit(struct sock *sk)
tcp_skb_mark_lost(tp, skb);
lost = 1;
}
+ tcp_verify_retransmit_hint(tp, skb);
}
}
@@ -2006,15 +2007,11 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
const struct inet_connection_sock *icsk = inet_csk(sk);
struct tcp_sock *tp = tcp_sk(sk);
struct sk_buff *skb;
- int packet_cnt;
- if (tp->retransmit_skb_hint) {
+ if (tp->retransmit_skb_hint)
skb = tp->retransmit_skb_hint;
- packet_cnt = tp->retransmit_cnt_hint;
- } else {
+ else
skb = tcp_write_queue_head(sk);
- packet_cnt = 0;
- }
/* First pass: retransmit lost packets. */
if (tp->lost_out) {
@@ -2025,7 +2022,6 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
break;
/* we could do better than to assign each time */
tp->retransmit_skb_hint = skb;
- tp->retransmit_cnt_hint = packet_cnt;
/* Assume this retransmit will generate
* only one packet for congestion window
@@ -2036,6 +2032,8 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
*/
if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
return;
+ if (!before(TCP_SKB_CB(skb)->seq, tp->retransmit_high))
+ break;
if (sacked & TCPCB_LOST) {
if (!(sacked & (TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))) {
@@ -2056,10 +2054,6 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
inet_csk(sk)->icsk_rto,
TCP_RTO_MAX);
}
-
- packet_cnt += tcp_skb_pcount(skb);
- if (packet_cnt >= tp->lost_out)
- break;
}
}
}
Main benefit in this is that we can then freely point the retransmit_skb_hint to anywhere we want to because there's no longer need to know what would be the count changes involve, and since this is really used only as a terminator, unnecessary work is one time walk at most, and if some retransmissions are necessary after that point later on, the walk is not full waste of time anyway. Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> --- include/linux/tcp.h | 2 +- include/net/tcp.h | 3 +++ net/ipv4/tcp_input.c | 12 +++++++++--- net/ipv4/tcp_output.c | 16 +++++----------- 4 files changed, 18 insertions(+), 15 deletions(-)