[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAM_iQpUwACsXVe9GAkQ1XC1TTU8aT3pqnrkZYT+oVDrP_1pKzw@mail.gmail.com>
Date: Thu, 9 Jul 2020 23:09:49 -0700
From: Cong Wang <xiyou.wangcong@...il.com>
To: "YU, Xiangning" <xiangning.yu@...baba-inc.com>
Cc: Linux Kernel Network Developers <netdev@...r.kernel.org>
Subject: Re: [PATCH net-next 2/2] net: sched: Lockless Token Bucket (LTB) Qdisc
On Thu, Jul 9, 2020 at 10:49 PM YU, Xiangning
<xiangning.yu@...baba-inc.com> wrote:
>
> Well, we do ask packets from a flow to be classified to a single class, not multiple ones. It doesn't have to be socket priority, it could be five tuple hash, or even container classid.
I don't see how it is so in your code, without skb priority your code
simply falls back to default class:
+ /* Allow to select a class by setting skb->priority */
+ if (likely(skb->priority != 0)) {
+ cl = ltb_find_class(sch, skb->priority);
+ if (cl)
+ return cl;
+ }
+ return rcu_dereference_bh(ltb->default_cls);
Mind to be more specific here?
BTW, your qdisc does not even support TC filters, does it?
At least I don't see that tcf_classify() is called.
>
> I think it's ok to have this requirement, even if we use htb, I would suggest the same. Why do you think this is a problem?
Because HTB does not have a per-cpu queue for each class,
yours does, cl->aggr_queues[cpu], if your point here is why we
don't blame HTB.
Thanks.
Powered by blists - more mailing lists