[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9d01e66a-5855-ea9a-ff63-c029ceb46c98@yandex-team.ru>
Date: Tue, 15 Aug 2017 17:33:06 +0300
From: Konstantin Khlebnikov <khlebnikov@...dex-team.ru>
To: Eric Dumazet <eric.dumazet@...il.com>
Cc: netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
Cong Wang <xiyou.wangcong@...il.com>,
Jamal Hadi Salim <jhs@...atatu.com>
Subject: Re: [PATCH] net_sched/sfq: update hierarchical backlog when drop
packet
On 15.08.2017 17:09, Eric Dumazet wrote:
> On Tue, 2017-08-15 at 16:37 +0300, Konstantin Khlebnikov wrote:
>> When sfq_enqueue() drops head packet or packet from another queue it
>> have to update backlog at upper qdiscs too.
>>
>> Signed-off-by: Konstantin Khlebnikov <khlebnikov@...dex-team.ru>
>> Fixes: 2ccccf5fb43f ("net_sched: update hierarchical backlog too")
>> ---
>> net/sched/sch_sfq.c | 5 ++++-
>> 1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/sched/sch_sfq.c b/net/sched/sch_sfq.c
>> index f80ea2cc5f1f..82469ef9655e 100644
>> --- a/net/sched/sch_sfq.c
>> +++ b/net/sched/sch_sfq.c
>> @@ -437,6 +437,7 @@ sfq_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free)
>> qdisc_drop(head, sch, to_free);
>>
>> slot_queue_add(slot, skb);
>> + qdisc_tree_reduce_backlog(sch, 0, delta);
>> return NET_XMIT_CN;
>> }
>>
>> @@ -468,8 +469,10 @@ sfq_enqueue(struct sk_buff *skb, struct Qdisc *sch, struct sk_buff **to_free)
>> /* Return Congestion Notification only if we dropped a packet
>> * from this flow.
>> */
>> - if (qlen != slot->qlen)
>> + if (qlen != slot->qlen) {
>> + qdisc_tree_reduce_backlog(sch, 0, dropped - qdisc_pkt_len(skb));
>> return NET_XMIT_CN;
>> + }
>>
>> /* As we dropped a packet, better let upper stack know this */
>> qdisc_tree_reduce_backlog(sch, 1, dropped);
>>
>
> Are you sure you have tested this patch ?
>
Nope. I'm not sure. But we have something similar in our 4.4 kernel for a while.
Also fq_codel and pfifo_head_drop do something similar tho this.
Probably this might crash without "[PATCH 1/2] net_sched: call qlen_notify only if child qdisc is empty".
I hadn't tested them separately.
Powered by blists - more mailing lists