Re: [PATCH net-next v6 2/6] net/sched: sch_cake: Factor out config variables into separate struct

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Toke Høiland-Jørgensen wrote:
> Willem de Bruijn <willemdebruijn.kernel@gmail.com> writes:
> 
> >>  static int cake_init(struct Qdisc *sch, struct nlattr *opt,
> >>  		     struct netlink_ext_ack *extack)
> >>  {
> >> -	struct cake_sched_data *q = qdisc_priv(sch);
> >> +	struct cake_sched_data *qd = qdisc_priv(sch);
> >> +	struct cake_sched_config *q;
> >>  	int i, j, err;
> >>  
> >> +	q = kvcalloc(1, sizeof(struct cake_sched_config), GFP_KERNEL);
> >> +	if (!q)
> >> +		return -ENOMEM;
> >> +
> >
> > Can this just be a regular kzalloc?
> 
> Yeah, I guess so. I'll change this if there's a need to respin for other
> reasons, but probably not worth respinning for this on its own? Seeing
> as it'll all end up in the same kmalloc call anyway :)

Sounds good.

> 
> > More importantly, where is q assigned to qd->config after init?
> 
> Just below:
> 
> >>  	sch->limit = 10240;
> >>  	sch->flags |= TCQ_F_DEQUEUE_DROPS;
> >>  
> >> @@ -2742,33 +2755,36 @@ static int cake_init(struct Qdisc *sch, struct nlattr *opt,
> >>  			       * for 5 to 10% of interval
> >>  			       */
> >>  	q->rate_flags |= CAKE_FLAG_SPLIT_GSO;
> >> -	q->cur_tin = 0;
> >> -	q->cur_flow  = 0;
> >> +	qd->cur_tin = 0;
> >> +	qd->cur_flow  = 0;
> >> +	qd->config = q;
> 
> Here:   ^^^^^^^

I'm blind. Thanks!

Reviewed-by: Willem de Bruijn <willemb@google.com>