tcp: cwnd does not increase in TCP YeAH
authorArtem Germanov <agermanov@anchorfree.com>
Wed, 7 Sep 2016 17:49:36 +0000 (10:49 -0700)
committerDavid S. Miller <davem@davemloft.net>
Fri, 9 Sep 2016 00:16:12 +0000 (17:16 -0700)
Commit 76174004a0f19785a328f40388e87e982bbf69b9
(tcp: do not slow start when cwnd equals ssthresh )
introduced regression in TCP YeAH. Using 100ms delay 1% loss virtual
ethernet link kernel 4.2 shows bandwidth ~500KB/s for single TCP
connection and kernel 4.3 and above (including 4.8-rc4) shows bandwidth
~100KB/s.
   That is caused by stalled cwnd when cwnd equals ssthresh. This patch
fixes it by proper increasing cwnd in this case.

Signed-off-by: Artem Germanov <agermanov@anchorfree.com>
Acked-by: Dmitry Adamushko <d.adamushko@anchorfree.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_yeah.c

index 028eb04..9c5fc97 100644 (file)
@@ -76,7 +76,7 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, u32 acked)
        if (!tcp_is_cwnd_limited(sk))
                return;
 
        if (!tcp_is_cwnd_limited(sk))
                return;
 
-       if (tp->snd_cwnd <= tp->snd_ssthresh)
+       if (tcp_in_slow_start(tp))
                tcp_slow_start(tp, acked);
 
        else if (!yeah->doing_reno_now) {
                tcp_slow_start(tp, acked);
 
        else if (!yeah->doing_reno_now) {