[TCP]: Add bytes_acked (ABC) clearing to FRTO too
Ilpo Järvinen [Fri, 12 Oct 2007 00:32:31 +0000 (17:32 -0700)]
I was reading tcp_enter_loss while looking for Cedric's bug and
noticed bytes_acked adjustment is missing from FRTO side.

Since bytes_acked will only be used in tcp_cong_avoid, I think
it's safe to assume RTO would be spurious. During FRTO cwnd
will be not controlled by tcp_cong_avoid and if FRTO calls for
conventional recovery, cwnd is adjusted and the result of wrong
assumption is cleared from bytes_acked. If RTO was in fact
spurious, we did normal ABC already and can continue without
any additional adjustments.

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

net/ipv4/tcp_input.c

index 101054c..e40857e 100644 (file)
@@ -1687,6 +1687,7 @@ static void tcp_enter_frto_loss(struct sock *sk, int allowed_segments, int flag)
        tp->snd_cwnd_cnt = 0;
        tp->snd_cwnd_stamp = tcp_time_stamp;
        tp->frto_counter = 0;
+       tp->bytes_acked = 0;
 
        tp->reordering = min_t(unsigned int, tp->reordering,
                                             sysctl_tcp_reordering);
@@ -2824,6 +2825,7 @@ static void tcp_conservative_spur_to_response(struct tcp_sock *tp)
 {
        tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh);
        tp->snd_cwnd_cnt = 0;
+       tp->bytes_acked = 0;
        TCP_ECN_queue_cwr(tp);
        tcp_moderate_cwnd(tp);
 }