tcp: avoid setting cwnd to invalid ssthresh after cwnd reduction states
authorYuchung Cheng <ycheng@google.com>
Tue, 1 Aug 2017 20:22:32 +0000 (13:22 -0700)
committerDavid S. Miller <davem@davemloft.net>
Wed, 2 Aug 2017 17:51:07 +0000 (10:51 -0700)
If the sender switches the congestion control during ECN-triggered
cwnd-reduction state (CA_CWR), upon exiting recovery cwnd is set to
the ssthresh value calculated by the previous congestion control. If
the previous congestion control is BBR that always keep ssthresh
to TCP_INIFINITE_SSTHRESH, cwnd ends up being infinite. The safe
step is to avoid assigning invalid ssthresh value when recovery ends.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_input.c

index 2920e0cb09f8d3e743eb4f49c16060ba1af48ed4..dad026fcfd098a78aab373bae78dbf88a0c50df1 100644 (file)
@@ -2520,8 +2520,8 @@ static inline void tcp_end_cwnd_reduction(struct sock *sk)
                return;
 
        /* Reset cwnd to ssthresh in CWR or Recovery (unless it's undone) */
-       if (inet_csk(sk)->icsk_ca_state == TCP_CA_CWR ||
-           (tp->undo_marker && tp->snd_ssthresh < TCP_INFINITE_SSTHRESH)) {
+       if (tp->snd_ssthresh < TCP_INFINITE_SSTHRESH &&
+           (inet_csk(sk)->icsk_ca_state == TCP_CA_CWR || tp->undo_marker)) {
                tp->snd_cwnd = tp->snd_ssthresh;
                tp->snd_cwnd_stamp = tcp_jiffies32;
        }