> 10 dec. 2015 kl. 16:37 skrev Neal Cardwell <ncardw...@google.com>:
>
>> On Thu, Dec 10, 2015 at 1:51 AM, Per Hurtig <per.hur...@kau.se> wrote:
>>
>>> On 08 Dec 2015, at 14:47, Eric Dumazet <eric.duma...@gmail.com> wrote:
>>>
>>> On Tue, 2015-12-08 at 10:19 +0100, Per Hurtig wrote:
>>>
>>>> +static u32 tcp_unsent_pkts(const struct sock *sk, u32 ulimit)
>>>> +{
>>>> + struct sk_buff *skb = tcp_send_head(sk);
>>>> + u32 pkts = 0;
>>>> +
>>>> + if (skb)
>>>> + tcp_for_write_queue_from(skb, sk) {
>>>> + pkts += tcp_skb_pcount(skb);
>>>> +
>>>> + if (ulimit && pkts >= ulimit)
>>>> + return ulimit;
>>>> + }
>>>> +
>>>> + return pkts;
>>>> +}
>>>
>>>
>>> Considering Yuchung feedback, have you looked at using an approximation
>>> instead ?
>>>
>>> (ie using tp->write_seq - tp->snd_nxt)
>>
>> Well, an approximation is rather “dangerous” as missing a single packet
>> could inhibit the desired behaviour. If looping is undesired, I think a
>> better solution is to actually *not* do this check at all and instead rely
>> solely on the
>>
>> tp->packets_out < TCP_RTORESTART_THRESH
>
> Yes, this simpler version seems very much preferable, IMHO. I agree
> that it does not seem worth the complexity to try to cover the kind of
> corner cases you outline.
>
> I would also suggest a TCP_RTORESTART_THRESH value higher than 4.
>
> In the ID at https://tools.ietf.org/html/draft-ietf-tcpm-rtorestart-10 it
> says:
>
> The RECOMMENDED value of rrthresh is four, as this value will ensure
> that RTOR is only used when fast retransmit cannot be triggered.
>
> But my sense is that fast retransmit is often not triggered at
> in-flight counts of much higher than 4, due to drop-tail queues, TSO
> bursts, the initial IW10 being unpaced, etc. It would be interesting
> to see A/B experiments for a few TCP_RTORESTART_THRESH values, say, 4
> vs 10.
>
> neal
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majord...@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
Sure. One idea could also be to use the "reordering" value as a dynamic
threshhold?
-- Per
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
