Re: Why do we prefer skb->priority to tc filters?

[Eric Dumazet <eric.dumazet@gmail.com>]

On Wed, 2015-03-11 at 11:08 -0700, Cong Wang wrote:
> On Wed, Mar 11, 2015 at 10:34 AM, Cong Wang <cwang@twopensource.com> wrote:
> >
> > Given that skb->priority can be specified in user-space, doesn't this
> > mean some application can always override our rules specified by tc
> > filters? I think we should always respect tc filters over any
> > application setting.
> >
> 
> Hmm, on the other hand, skb->priority can be changed in cgroup
> too, in this case the current behavior seems correct. :-/ Interesting.

Note that even HTB has the following code :

static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch,
                                      int *qerr)
{
        struct htb_sched *q = qdisc_priv(sch);
        struct htb_class *cl;
        struct tcf_result res;
        struct tcf_proto *tcf;
        int result;

        /* allow to select class by setting skb->priority to valid classid;
         * note that nfmark can be used too by attaching filter fw with no
         * rules in it
         */
        if (skb->priority == sch->handle)
                return HTB_DIRECT;      /* X:0 (direct flow) selected */
        cl = htb_find(skb->priority, sch);
        if (cl) {
                if (cl->level == 0)
                        return cl;
                /* Start with inner filter chain if a non-leaf class is selected */
                tcf = rcu_dereference_bh(cl->filter_list);
        } else {
                tcf = rcu_dereference_bh(q->filter_list);
        }


So if skb->priority happens to match sch->handle or any class handle,
we queue packet into a queue, without calling tc_classify()