17 Mar
2021
17 Mar
'21
8:45 a.m.
On 2021/3/17 5:45, David Miller wrote:
From: Yunsheng Lin linyunsheng@huawei.com Date: Tue, 16 Mar 2021 10:40:56 +0800
On 2021/3/16 7:41, David Miller wrote:
From: Yunsheng Lin linyunsheng@huawei.com
At least for the fast path, taking two locks for lockless qdisc hurts performance when handling requeued skb, especially if the lockless qdisc supports TCQ_F_CAN_BYPASS.
The bad txq and gro skb cases are not "fast path", sorry
You are right, it is more of exceptional data path, but it is still ok to clean that up without obvious risk, right? For I am going to replace qdisc->seqlock with qdisc_lock(q) for lockless qdisc too and remove qdisc->seqlock, which makes more sense.
.