netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH net] tcp: fix RTO calculated from cached RTT
@ 2013-09-17  1:44 Neal Cardwell
  2013-09-17 12:21 ` Eric Dumazet
  2013-09-17 14:32 ` Yuchung Cheng
  0 siblings, 2 replies; 4+ messages in thread
From: Neal Cardwell @ 2013-09-17  1:44 UTC (permalink / raw)
  To: David Miller; +Cc: netdev, Neal Cardwell, Eric Dumazet, Yuchung Cheng

Commit 1b7fdd2ab5852 ("tcp: do not use cached RTT for RTT estimation")
did not correctly account for the fact that crtt is the RTT shifted
left 3 bits. Fix the calculation to consistently reflect this fact.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
---
 net/ipv4/tcp_metrics.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/tcp_metrics.c b/net/ipv4/tcp_metrics.c
index 4a22f3e..52f3c6b 100644
--- a/net/ipv4/tcp_metrics.c
+++ b/net/ipv4/tcp_metrics.c
@@ -502,7 +502,9 @@ reset:
 	 * ACKs, wait for troubles.
 	 */
 	if (crtt > tp->srtt) {
-		inet_csk(sk)->icsk_rto = crtt + max(crtt >> 2, tcp_rto_min(sk));
+		/* Set RTO like tcp_rtt_estimator(), but from cached RTT. */
+		crtt >>= 3;
+		inet_csk(sk)->icsk_rto = crtt + max(2 * crtt, tcp_rto_min(sk));
 	} else if (tp->srtt == 0) {
 		/* RFC6298: 5.7 We've failed to get a valid RTT sample from
 		 * 3WHS. This is most likely due to retransmission,
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH net] tcp: fix RTO calculated from cached RTT
  2013-09-17  1:44 [PATCH net] tcp: fix RTO calculated from cached RTT Neal Cardwell
@ 2013-09-17 12:21 ` Eric Dumazet
  2013-09-17 14:32 ` Yuchung Cheng
  1 sibling, 0 replies; 4+ messages in thread
From: Eric Dumazet @ 2013-09-17 12:21 UTC (permalink / raw)
  To: Neal Cardwell; +Cc: David Miller, netdev, Eric Dumazet, Yuchung Cheng

On Mon, 2013-09-16 at 21:44 -0400, Neal Cardwell wrote:
> Commit 1b7fdd2ab5852 ("tcp: do not use cached RTT for RTT estimation")
> did not correctly account for the fact that crtt is the RTT shifted
> left 3 bits. Fix the calculation to consistently reflect this fact.
> 
> Signed-off-by: Neal Cardwell <ncardwell@google.com>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> ---
>  net/ipv4/tcp_metrics.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)

Acked-by: Eric Dumazet <edumazet@google.com>

Instead of rto = 3*crtt, we had rto = 10*crtt, which was indeed
suboptimal ;)

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH net] tcp: fix RTO calculated from cached RTT
  2013-09-17  1:44 [PATCH net] tcp: fix RTO calculated from cached RTT Neal Cardwell
  2013-09-17 12:21 ` Eric Dumazet
@ 2013-09-17 14:32 ` Yuchung Cheng
  2013-09-17 23:08   ` David Miller
  1 sibling, 1 reply; 4+ messages in thread
From: Yuchung Cheng @ 2013-09-17 14:32 UTC (permalink / raw)
  To: Neal Cardwell; +Cc: David Miller, netdev, Eric Dumazet

On Mon, Sep 16, 2013 at 6:44 PM, Neal Cardwell <ncardwell@google.com> wrote:
> Commit 1b7fdd2ab5852 ("tcp: do not use cached RTT for RTT estimation")
> did not correctly account for the fact that crtt is the RTT shifted
> left 3 bits. Fix the calculation to consistently reflect this fact.
>
> Signed-off-by: Neal Cardwell <ncardwell@google.com>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Yuchung Cheng <ycheng@google.com>
> ---
>  net/ipv4/tcp_metrics.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
Acked-By: Yuchung Cheng <ycheng@google.com>

Thanks for discovering and fixing it!

>
> diff --git a/net/ipv4/tcp_metrics.c b/net/ipv4/tcp_metrics.c
> index 4a22f3e..52f3c6b 100644
> --- a/net/ipv4/tcp_metrics.c
> +++ b/net/ipv4/tcp_metrics.c
> @@ -502,7 +502,9 @@ reset:
>          * ACKs, wait for troubles.
>          */
>         if (crtt > tp->srtt) {
> -               inet_csk(sk)->icsk_rto = crtt + max(crtt >> 2, tcp_rto_min(sk));
> +               /* Set RTO like tcp_rtt_estimator(), but from cached RTT. */
> +               crtt >>= 3;
> +               inet_csk(sk)->icsk_rto = crtt + max(2 * crtt, tcp_rto_min(sk));
>         } else if (tp->srtt == 0) {
>                 /* RFC6298: 5.7 We've failed to get a valid RTT sample from
>                  * 3WHS. This is most likely due to retransmission,
> --
> 1.8.4
>

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH net] tcp: fix RTO calculated from cached RTT
  2013-09-17 14:32 ` Yuchung Cheng
@ 2013-09-17 23:08   ` David Miller
  0 siblings, 0 replies; 4+ messages in thread
From: David Miller @ 2013-09-17 23:08 UTC (permalink / raw)
  To: ycheng; +Cc: ncardwell, netdev, edumazet

From: Yuchung Cheng <ycheng@google.com>
Date: Tue, 17 Sep 2013 07:32:38 -0700

> On Mon, Sep 16, 2013 at 6:44 PM, Neal Cardwell <ncardwell@google.com> wrote:
>> Commit 1b7fdd2ab5852 ("tcp: do not use cached RTT for RTT estimation")
>> did not correctly account for the fact that crtt is the RTT shifted
>> left 3 bits. Fix the calculation to consistently reflect this fact.
>>
>> Signed-off-by: Neal Cardwell <ncardwell@google.com>
>> Cc: Eric Dumazet <edumazet@google.com>
>> Cc: Yuchung Cheng <ycheng@google.com>
>> ---
>>  net/ipv4/tcp_metrics.c | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
> Acked-By: Yuchung Cheng <ycheng@google.com>
> 
> Thanks for discovering and fixing it!

Applied, thanks everyone.

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2013-09-17 23:09 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-09-17  1:44 [PATCH net] tcp: fix RTO calculated from cached RTT Neal Cardwell
2013-09-17 12:21 ` Eric Dumazet
2013-09-17 14:32 ` Yuchung Cheng
2013-09-17 23:08   ` David Miller

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).