Re: [dpdk-dev] [PATCH v12] ethdev: new Rx/Tx offloads API

[Thomas Monjalon <thomas@monjalon.net>]

Hi,

A first general comment: a lot of spaces are still inside parens.
You can grep '( )'.

10/05/2018 13:56, Wei Dai:
> --- a/doc/guides/prog_guide/poll_mode_drv.rst
> +++ b/doc/guides/prog_guide/poll_mode_drv.rst
> +A per-queue offloading can be enabled on a queue and disabled on another queue at the same time.
> +A pure per-port offload is the one supported by device but not per-queue type.

Another way to say it: pure per-port offloads are not directly advertised but
are the port offloads capabilities minus the queue capabilities.
port capabilities = pure per-port capabilities + queue capabilities

> +A pure per-port offloading can't be enabled on a queue and disabled on another queue at the same time.
> +A pure per-port offloading must be enabled or disabled on all queues at the same time.
> +Any offloading is per-queue or pure per-port type, but can't be both types at same devices.
> +A per-port offloading can be enabled or disabled on all queues at the same time.

This sentence is useless: it says any offload can be setup for the whole port.

> +It is certain that both per-queue and pure per-port offloading are per-port type.

This sentence is confusing. I cannot understand it.


>  The different offloads capabilities can be queried using ``rte_eth_dev_info_get()``.
> +The dev_info->[rt]x_queue_offload_capa returned from ``rte_eth_dev_info_get()`` includes all per-queue offloading capabilities.
> +The dev_info->[rt]x_offload_capa returned from ``rte_eth_dev_info_get()`` includes all per-port and per-queue offloading capabilities.

If you want to stick with pure per-port wording, you should say
[rt]x_offload_capa is the port capabilities (including pure per-port and per-queue).


> --- a/lib/librte_ethdev/rte_ethdev.c
> +++ b/lib/librte_ethdev/rte_ethdev.c
> +	/* Any requested offloading must be within its device capabilities */
> +	if ((local_conf.rxmode.offloads & dev_info.rx_offload_capa) !=
> +	     local_conf.rxmode.offloads) {
> +		ethdev_log(ERR, "ethdev port_id=%d requested Rx offloads "
> +				"0x%" PRIx64 " doesn't match Rx offloads "
> +				"capabilities 0x%" PRIx64 " in %s( )\n",
> +				port_id,
> +				local_conf.rxmode.offloads,
> +				dev_info.rx_offload_capa,
> +				__func__);

We could have a comment saying that an error will be returned in next version.

> +	}
> +	if ((local_conf.txmode.offloads & dev_info.tx_offload_capa) !=
> +	     local_conf.txmode.offloads) {
> +		ethdev_log(ERR, "ethdev port_id=%d requested Tx offloads "
> +				"0x%" PRIx64 " doesn't match Tx offloads "
> +				"capabilities 0x%" PRIx64 " in %s( )\n",
> +				port_id,
> +				local_conf.txmode.offloads,
> +				dev_info.tx_offload_capa,
> +				__func__);

idem

> +	}


> +	/*
> +	 * If an offloading has already been enabled in
> +	 * rte_eth_dev_configure(), it has been enabled on all queues,
> +	 * so there is no need to enable it in this queue again.
> +	 * The local_conf.offloads input to underlying PMD only carries
> +	 * those offloadings which are only enabled on this queue and
> +	 * not enabled on all queues.
> +	 * The underlying PMD must be aware of this point.

I think the last sentence is useless.

> +	 */
> +	local_conf.offloads &= ~dev->data->dev_conf.rxmode.offloads;
> +
> +	/*
> +	 * New added offloadings for this queue are those not enabled in
> +	 * rte_eth_dev_configure( ) and they must be per-queue type.
> +	 * A pure per-port offloading can't be enabled on a queue while
> +	 * disabled on another queue. A pure per-port offloading can't
> +	 * be enabled for any queue as new added one if it hasn't been
> +	 * enabled in rte_eth_dev_configure( ).
> +	 */
> +	if ((local_conf.offloads & dev_info.rx_queue_offload_capa) !=
> +	     local_conf.offloads) {
> +		ethdev_log(ERR, "Ethdev port_id=%d rx_queue_id=%d, new "
> +				"added offloads 0x%" PRIx64 " must be "
> +				"within pre-queue offload capabilities 0x%"
> +				PRIx64 " in %s( )\n",
> +				port_id,
> +				rx_queue_id,
> +				local_conf.offloads,
> +				dev_info.rx_queue_offload_capa,
> +				__func__);

idem, we can have a comment about error in next version

> +	}


> --- a/lib/librte_ethdev/rte_ethdev.h
> +++ b/lib/librte_ethdev/rte_ethdev.h
>  	uint64_t rx_offload_capa;
> -	/**< Device per port RX offload capabilities. */
> +	/**< All RX offload capabilities including all per queue ones */

OK
per queue -> per-queue

>  	uint64_t tx_offload_capa;
> -	/**< Device per port TX offload capabilities. */
> +	/**< All TX offload capabilities.including all per-queue ones */

Typo: there is a dot instead of space.

>  	uint64_t rx_queue_offload_capa;
>  	/**< Device per queue RX offload capabilities. */

Here you should add more comments:
	No need to repeat flags already enabled at port level.
	A flag enabled at port level, cannot be disabled at queue level.


> + *     -  Any offloading set in eth_conf->[rt]xmode.offloads must be within
> + *        the [rt]x_offload_capa returned from rte_eth_dev_infos_get().

OK

> + *        Any type of device supported offloading set in the input argument
> + *        eth_conf->[rt]xmode.offloads to rte_eth_dev_configure() is enabled
> + *        on all [RT]x queues and it can't be disabled no matter whether
> + *        it is cleared or set in the input argument [rt]x_conf->offloads
> + *        to rte_eth_[rt]x_queue_setup().

last part can be simpler: cannot be disabled in queue setup.
"[RT]x queues" can be simply "queues".


> + *   If an offloading set in rx_conf->offloads
> + *   hasn't been set in the input argument eth_conf->rxmode.offloads
> + *   to rte_eth_dev_configure(), it is a new added offloading, it must be
> + *   per-queue type and it is enabled for the queue.

OK
Another wording:
The offloads not advertised in queue capabilities, and not already enabled
at port level, are rejected.