Re: [PATCH net-next 3/4] net_sched: sch_fq: add 3 bands and WRR scheduling

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Eric Dumazet <edumazet@google.com> writes:

> Before Google adopted FQ for its production servers,
> we had to ensure AF4 packets would get a higher share
> than BE1 ones.
>
> As discussed this week in Netconf 2023 in Paris, it is time
> to upstream this for public use.

IIRC, when you mentioned this at Netconf you said the new behaviour
would probably need to be behind a flag, but I don't see that in this
series. What was the reason you decided to drop that?

[..]
> +static int fq_load_priomap(struct fq_sched_data *q,
> +			   const struct nlattr *attr,
> +			   struct netlink_ext_ack *extack)
> +{
> +	const struct tc_prio_qopt *map = nla_data(attr);
> +	int i;
> +
> +	if (map->bands != FQ_BANDS) {
> +		NL_SET_ERR_MSG_MOD(extack, "FQ only supports 3 bands");
> +		return -EINVAL;
> +	}
> +	for (i = 0; i < TC_PRIO_MAX + 1; i++) {
> +		if (map->priomap[i] >= FQ_BANDS) {
> +			NL_SET_ERR_MSG_MOD(extack, "Incorrect field in FQ priomap");

Can we be a bit more specific than just "incorrect" here? Something like
"FQ priomap field %d maps to a too high band %d"?

-Toke