在 2020/7/18 20:30, Florian Westphal 写道:
> we...@ucloud.cn <we...@ucloud.cn> wrote:
>> From: wenxu <we...@ucloud.cn>
>>
>> The fragment packets do defrag in tcf_ct_handle_fragments
>> will clear the skb->cb which make the qdisc_skb_cb clear
>> too. So the qdsic_skb_cb should be store before defrag and
>> restore after that.
>> It also update the pkt_len after all the
>> fragments finish the defrag to one packet and make the
>> following actions counter correct.
>>
>> Fixes: b57dc7c13ea9 ("net/sched: Introduce action ct")
>> Signed-off-by: wenxu <we...@ucloud.cn>
> Looks ok to me. One question:
>
>> @@ -1014,6 +1017,7 @@ static int tcf_ct_act(struct sk_buff *skb, const
>> struct tc_action *a,
>>
>> out:
>> tcf_action_update_bstats(&c->common, skb);
>> + qdisc_skb_cb(skb)->pkt_len = skb->len;
>> return retval;
> This appears to be unconditional, I would have expected that
> this only done for reassembled skbs?
Yes.
>
> Otherwise we will lose the value calculated by core via
> qdisc_calculate_pkt_len().
qdisc_calculate_pkt_len only be cablled in dev_xmit_skb and qdisc_enqueue. If
all the fragment will
pass those before defrag, it will caculate correctly. If the reassembled
packets pass those, it also caculate correctly after we recaculate the pkt_len
of qdisc_skb_cb
