[TCP]: No need to re-count fackets_out/sacked_out at RTO
authorIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Fri, 12 Oct 2007 00:34:57 +0000 (17:34 -0700)
committerDavid S. Miller <davem@davemloft.net>
Fri, 12 Oct 2007 00:34:57 +0000 (17:34 -0700)
Both sacked_out and fackets_out are directly known from how
parameter. Since fackets_out is accurate, there's no need for
recounting (sacked_out was previously unnecessarily counted
in the loop anyway).

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_input.c

index bd18c25..a066039 100644 (file)
@@ -1711,18 +1711,23 @@ static void tcp_enter_frto_loss(struct sock *sk, int allowed_segments, int flag)
        tcp_clear_retrans_hints_partial(tp);
 }
 
-void tcp_clear_retrans(struct tcp_sock *tp)
+static void tcp_clear_retrans_partial(struct tcp_sock *tp)
 {
        tp->retrans_out = 0;
-
-       tp->fackets_out = 0;
-       tp->sacked_out = 0;
        tp->lost_out = 0;
 
        tp->undo_marker = 0;
        tp->undo_retrans = 0;
 }
 
+void tcp_clear_retrans(struct tcp_sock *tp)
+{
+       tcp_clear_retrans_partial(tp);
+
+       tp->fackets_out = 0;
+       tp->sacked_out = 0;
+}
+
 /* Enter Loss state. If "how" is not zero, forget all SACK information
  * and reset tags completely, otherwise preserve SACKs. If receiver
  * dropped its ofo queue, we will know this due to reneging detection.
@@ -1732,7 +1737,6 @@ void tcp_enter_loss(struct sock *sk, int how)
        const struct inet_connection_sock *icsk = inet_csk(sk);
        struct tcp_sock *tp = tcp_sk(sk);
        struct sk_buff *skb;
-       int cnt = 0;
 
        /* Reduce ssthresh if it has not yet been made inside this window. */
        if (icsk->icsk_ca_state <= TCP_CA_Disorder || tp->snd_una == tp->high_seq ||
@@ -1746,7 +1750,10 @@ void tcp_enter_loss(struct sock *sk, int how)
        tp->snd_cwnd_stamp = tcp_time_stamp;
 
        tp->bytes_acked = 0;
-       tcp_clear_retrans(tp);
+       tcp_clear_retrans_partial(tp);
+
+       if (tcp_is_reno(tp))
+               tcp_reset_reno_sack(tp);
 
        if (!how) {
                /* Push undo marker, if it was plain RTO and nothing
@@ -1754,13 +1761,15 @@ void tcp_enter_loss(struct sock *sk, int how)
                tp->undo_marker = tp->snd_una;
                tcp_clear_retrans_hints_partial(tp);
        } else {
+               tp->sacked_out = 0;
+               tp->fackets_out = 0;
                tcp_clear_all_retrans_hints(tp);
        }
 
        tcp_for_write_queue(skb, sk) {
                if (skb == tcp_send_head(sk))
                        break;
-               cnt += tcp_skb_pcount(skb);
+
                if (TCP_SKB_CB(skb)->sacked&TCPCB_RETRANS)
                        tp->undo_marker = 0;
                TCP_SKB_CB(skb)->sacked &= (~TCPCB_TAGBITS)|TCPCB_SACKED_ACKED;
@@ -1768,9 +1777,6 @@ 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);
-               } else {
-                       tp->sacked_out += tcp_skb_pcount(skb);
-                       tp->fackets_out = cnt;
                }
        }
        tcp_verify_left_out(tp);