[dpdk-dev] [PATCH v2 09/11] net/octeontx_ep: Added dev start and stop

[Ferruh Yigit]

On 1/18/2021 9:36 AM, Nalla Pradeep wrote:
> Dev start and stop operations are added. To accomplish this internal
> functions to enable or disable io queues are incorporated.
> 
> Signed-off-by: Nalla Pradeep <pnalla at marvell.com>

<...>

> +static int
> +otx_ep_dev_start(struct rte_eth_dev *eth_dev)
> +{
> +	struct otx_ep_device *otx_epvf;
> +	unsigned int q;
> +	int ret;
> +
> +	otx_epvf = (struct otx_ep_device *)OTX_EP_DEV(eth_dev);
> +	/* Enable IQ/OQ for this device */
> +	ret = otx_epvf->fn_list.enable_io_queues(otx_epvf);
> +	if (ret) {
> +		otx_ep_err("IOQ enable failed\n");
> +		return ret;
> +	}
> +
> +	for (q = 0; q < otx_epvf->nb_rx_queues; q++) {
> +		rte_write32(otx_epvf->droq[q]->nb_desc,
> +			    otx_epvf->droq[q]->pkts_credit_reg);
> +
> +		rte_wmb();
> +		otx_ep_info("OQ[%d] dbells [%d]\n", q,
> +		rte_read32(otx_epvf->droq[q]->pkts_credit_reg));
> +	}
> +
> +	otx_epvf->started = 1;
> +
> +	rte_wmb();

is this memory barrier required, isn't the one in the loop should be enough? And 
no ones seems reading the 'started'.

> +	otx_ep_info("dev started\n");
> +
> +	return 0;
> +}
> +

<...>