tcp: apply Kern's check on RTTs used for congestion control
authorYuchung Cheng <ycheng@google.com>
Sat, 17 Oct 2015 04:57:41 +0000 (21:57 -0700)
committerDavid S. Miller <davem@davemloft.net>
Wed, 21 Oct 2015 14:00:41 +0000 (07:00 -0700)
Currently ca_seq_rtt_us does not use Kern's check. Fix that by
checking if any packet acked is a retransmit, for both RTT used
for RTT estimation and congestion control.

Fixes: 5b08e47ca ("tcp: prefer packet timing to TS-ECR for RTT")
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_input.c

index 944eaca6911596676af66a934a17685c27fc11cc..62ee71efd1ce2e65128ba5d929f70f144c1c1d02 100644 (file)
@@ -2925,9 +2925,6 @@ static inline bool tcp_ack_update_rtt(struct sock *sk, const int flag,
         * Karn's algorithm forbids taking RTT if some retransmitted data
         * is acked (RFC6298).
         */
-       if (flag & FLAG_RETRANS_DATA_ACKED)
-               seq_rtt_us = -1L;
-
        if (seq_rtt_us < 0)
                seq_rtt_us = sack_rtt_us;
 
@@ -3169,7 +3166,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets,
                flag |= FLAG_SACK_RENEGING;
 
        skb_mstamp_get(&now);
-       if (likely(first_ackt.v64)) {
+       if (likely(first_ackt.v64) && !(flag & FLAG_RETRANS_DATA_ACKED)) {
                seq_rtt_us = skb_mstamp_us_delta(&now, &first_ackt);
                ca_rtt_us = skb_mstamp_us_delta(&now, &last_ackt);
        }