[PATCH] tcp: undo spurious timeout after SACK reneging

[PATCH] tcp: undo spurious timeout after SACK reneging

[Yuchung Cheng]

On SACK reneging the sender immediately retransmits and forces a
timeout but disables Eifel (undo). If the (buggy) receiver does not
drop any packet this can trigger a false slow-start retransmit storm
driven by the ACKs of the original packets. This can be detected with
undo and TCP timestamps.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
---
 net/ipv4/tcp_input.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index b2b3619..6d9ca35 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -1858,11 +1858,8 @@ void tcp_enter_loss(struct sock *sk, int how)
 	if (tcp_is_reno(tp))
 		tcp_reset_reno_sack(tp);
 
-	if (!how) {
-		/* Push undo marker, if it was plain RTO and nothing
-		 * was retransmitted. */
-		tp->undo_marker = tp->snd_una;
-	} else {
+	tp->undo_marker = tp->snd_una;
+	if (how) {
 		tp->sacked_out = 0;
 		tp->fackets_out = 0;
 	}
-- 
1.8.1.3

Re: [PATCH] tcp: undo spurious timeout after SACK reneging

[Neal Cardwell]

On Sun, Mar 24, 2013 at 4:42 PM, Yuchung Cheng <ycheng@google.com> wrote:
> On SACK reneging the sender immediately retransmits and forces a
> timeout but disables Eifel (undo). If the (buggy) receiver does not
> drop any packet this can trigger a false slow-start retransmit storm
> driven by the ACKs of the original packets. This can be detected with
> undo and TCP timestamps.
>
> Signed-off-by: Yuchung Cheng <ycheng@google.com>
> ---
>  net/ipv4/tcp_input.c | 7 ++-----
>  1 file changed, 2 insertions(+), 5 deletions(-)

Acked-by: Neal Cardwell <ncardwell@google.com>

neal

Re: [PATCH] tcp: undo spurious timeout after SACK reneging

[David Miller]

From: Neal Cardwell <ncardwell@google.com>
Date: Sun, 24 Mar 2013 16:50:41 -0400

> On Sun, Mar 24, 2013 at 4:42 PM, Yuchung Cheng <ycheng@google.com> wrote:
>> On SACK reneging the sender immediately retransmits and forces a
>> timeout but disables Eifel (undo). If the (buggy) receiver does not
>> drop any packet this can trigger a false slow-start retransmit storm
>> driven by the ACKs of the original packets. This can be detected with
>> undo and TCP timestamps.
>>
>> Signed-off-by: Yuchung Cheng <ycheng@google.com>
>> ---
>>  net/ipv4/tcp_input.c | 7 ++-----
>>  1 file changed, 2 insertions(+), 5 deletions(-)
> 
> Acked-by: Neal Cardwell <ncardwell@google.com>

Applied.