Re: [PATCH net-next v4 5/7] net: stmmac: support fp parameter of tc-mqprio

[Vladimir Oltean <olteanv@gmail.com>]

On Tue, Aug 20, 2024 at 05:38:33PM +0800, Furong Xu wrote:
> +static int tc_setup_mqprio(struct stmmac_priv *priv,
> +			   struct tc_mqprio_qopt_offload *mqprio)
> +{
> +	struct netlink_ext_ack *extack = mqprio->extack;
> +	struct tc_mqprio_qopt *qopt = &mqprio->qopt;
> +	struct net_device *ndev = priv->dev;
> +	int num_stack_tx_queues = 0;
> +	int num_tc = qopt->num_tc;
> +	u16 offset, count;
> +	int tc, err;
> +
> +	if (!num_tc) {
> +		stmmac_reset_tc_mqprio(ndev, extack);
> +		return 0;
> +	}
> +
> +	if (mqprio->preemptible_tcs && !ethtool_dev_mm_supported(ndev)) {
> +		NL_SET_ERR_MSG_MOD(extack, "Device does not support preemption");
> +		return -EOPNOTSUPP;
> +	}

When I said that "this condition is dealt with by the core, now"
https://lore.kernel.org/netdev/20240819114242.2m6okk7bq64e437c@skbuf/
I meant that the driver doesn't need to check anything - the check has
already run once, in the Qdisc layer. See taprio_parse_tc_entries() and
mqprio_parse_tc_entries(). I was not asking to insert this test, just to
completely remove, rather than adapt, the entire block:

	if (fpe && !priv->dma_cap.fpesel) {
		mutex_unlock(&priv->est_lock);
		return -EOPNOTSUPP;
	}