[dpdk-dev] [PATCH 3/4] eventdev: Add eventdev ethernet Rx adapter

[Nipun Gupta]

> -----Original Message-----
> From: Nikhil Rao [mailto:nikhil.rao at intel.com]
> Sent: Thursday, September 14, 2017 0:23
> To: jerin.jacob at caviumnetworks.com; bruce.richardson at intel.com
> Cc: gage.eads at intel.com; dev at dpdk.org; thomas at monjalon.net;
> harry.van.haaren at intel.com; Hemant Agrawal <hemant.agrawal at nxp.com>;
> Nipun Gupta <nipun.gupta at nxp.com>; narender.vangati at intel.com;
> erik.g.carrillo at intel.com; abhinandan.gujjar at intel.com
> Subject: [PATCH 3/4] eventdev: Add eventdev ethernet Rx adapter
> 
> Add common APIs for configuring packet transfer from ethernet Rx
> queues to event devices across HW & SW packet transfer mechanisms.
> A detailed description of the adapter is contained in the header's
> comments.
> 
> The adapter implementation uses eventdev PMDs to configure the packet
> transfer if HW support is available and if not, it uses an EAL service
> function that reads packets from ethernet Rx queues and injects these
> as events into the event device.
> 
> Signed-off-by: Nikhil Rao <nikhil.rao at intel.com>
> Signed-off-by: Gage Eads <gage.eads at intel.com>
> Signed-off-by: Abhinandan Gujjar <abhinandan.gujjar at intel.com>
> ---
>  lib/librte_eventdev/rte_event_eth_rx_adapter.h |  386 ++++++++
>  lib/librte_eventdev/rte_event_eth_rx_adapter.c | 1239
> ++++++++++++++++++++++++
>  lib/Makefile                                   |    2 +-
>  lib/librte_eventdev/Makefile                   |    2 +
>  lib/librte_eventdev/rte_eventdev_version.map   |   11 +-
>  5 files changed, 1638 insertions(+), 2 deletions(-)
>  create mode 100644 lib/librte_eventdev/rte_event_eth_rx_adapter.h
>  create mode 100644 lib/librte_eventdev/rte_event_eth_rx_adapter.c
> 
> +
> +static inline void
> +fill_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter,
> +	uint8_t dev_id,
> +	uint16_t rx_queue_id,
> +	struct rte_mbuf **mbufs,
> +	uint16_t num)
> +{
> +	uint32_t i;
> +	struct eth_device_info *eth_device_info =
> +					&rx_adapter->eth_devices[dev_id];
> +	struct eth_rx_queue_info *eth_rx_queue_info =
> +					&eth_device_info-
> >rx_queue[rx_queue_id];
> +
> +	int32_t qid = eth_rx_queue_info->event_queue_id;
> +	uint8_t sched_type = eth_rx_queue_info->sched_type;
> +	uint8_t priority = eth_rx_queue_info->priority;
> +	uint32_t flow_id;
> +	struct rte_event events[BATCH_SIZE];
> +	struct rte_mbuf *m = mbufs[0];
> +	uint32_t rss_mask;
> +	uint32_t rss;
> +	int do_rss;
> +
> +	/* 0xffff ffff if PKT_RX_RSS_HASH is set, otherwise 0 */
> +	rss_mask = ~(((m->ol_flags & PKT_RX_RSS_HASH) != 0) - 1);
> +	do_rss = !rss_mask && !eth_rx_queue_info->flow_id_mask;
> +
> +	for (i = 0; i < num; i++) {
> +		m = mbufs[i];
> +		struct rte_event *ev = &events[i];
> +
> +		rss = do_rss ? do_softrss(m) : m->hash.rss;
> +		flow_id =
> +		    eth_rx_queue_info->flow_id &
> +				eth_rx_queue_info->flow_id_mask;
> +		flow_id |= rss & ~eth_rx_queue_info->flow_id_mask;
> +
> +		ev->flow_id = flow_id;
> +		ev->op = RTE_EVENT_OP_NEW;
> +		ev->sched_type = sched_type;
> +		ev->queue_id = qid;
> +		ev->event_type = RTE_EVENT_TYPE_ETHDEV;
> +		ev->sub_event_type = 0;
> +		ev->priority = priority;
> +		ev->mbuf = m;

How will the application get to know about the queue and eth_port from where the
packet is received by the eventdev? Is it the user responsibility to have that information
in the flow_id?

> +
> +		buf_event_enqueue(rx_adapter, ev);
> +	}
> +}
> +

<snip>

> +int
> +rte_event_eth_rx_adapter_queue_add(uint8_t id,
> +		uint8_t eth_dev_id,
> +		int32_t rx_queue_id,
> +		const struct rte_event_eth_rx_adapter_queue_conf
> *queue_conf)
> +{
> +	int ret;
> +	uint32_t rx_adapter_cap;
> +	struct rte_event_eth_rx_adapter *rx_adapter;
> +	struct rte_eventdev *dev;
> +	struct eth_device_info *dev_info;
> +	int start_service = 0;
> +
> +	RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL);
> +	RTE_ETH_VALID_PORTID_OR_ERR_RET(eth_dev_id, -EINVAL);
> +
> +	rx_adapter = id_to_rx_adapter(id);
> +	if (!rx_adapter || !queue_conf)
> +		return -EINVAL;
> +
> +	dev = &rte_eventdevs[rx_adapter->eventdev_id];
> +	ret = (*dev->dev_ops->eth_rx_adapter_caps_get)(dev, eth_dev_id,
> +						&rx_adapter_cap);
> +	if (ret) {
> +		RTE_EDEV_LOG_ERR("Failed to get adapter caps edev %" PRIu8
> +			"eth port %" PRIu8, id, eth_dev_id);
> +		return ret;
> +	}
> +
> +	if (!(rx_adapter_cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_FLOW_ID)
> &&
> +		!(queue_conf->rx_queue_flags &
> +
> 	RTE_EVENT_ETH_RX_ADAPTER_QUEUE_FLOW_ID_VALID)) {
> +		RTE_EDEV_LOG_ERR("Flow ID required for configuration,"
> +				" eth port: %" PRIu8 " adapter id: %" PRIu8,
> +				eth_dev_id, id);
> +		return -EINVAL;
> +	}
> +
> +	if ((rx_adapter_cap &
> RTE_EVENT_ETH_RX_ADAPTER_CAP_SINGLE_EVENTQ) &&
> +		(rx_queue_id != -1)) {
> +		RTE_EDEV_LOG_ERR("Rx queues can only be connected to
> single "
> +			"event queue id %u eth port %u", id, eth_dev_id);
> +		return -EINVAL;
> +	}
> +
> +	if (rx_queue_id != -1 && (uint16_t)rx_queue_id >=
> +			rte_eth_devices[eth_dev_id].data->nb_rx_queues) {
> +		RTE_EDEV_LOG_ERR("Invalid rx queue_id %" PRIu16,
> +			 (uint16_t)rx_queue_id);
> +		return -EINVAL;
> +	}
> +
> +	start_service = 0;
> +	dev_info = &rx_adapter->eth_devices[eth_dev_id];
> +
> +	if (rx_adapter_cap &
> RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT) {
> +		RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> >eth_rx_adapter_queue_add,
> +					-ENOTSUP);
> +		if (!dev_info->rx_queue) {
> +			dev_info->rx_queue =
> +			    rte_zmalloc_socket(rx_adapter->mem_name,
> +					dev_info->dev->data->nb_rx_queues *
> +					sizeof(struct eth_rx_queue_info), 0,
> +					rx_adapter->socket_id);
> +			if (!dev_info->rx_queue)
> +				return -ENOMEM;
> +		}
> +
> +		ret = (*dev->dev_ops->eth_rx_adapter_queue_add)(dev,
> eth_dev_id,
> +				rx_queue_id, queue_conf);

IMO, Instead of eth_dev_id rte_eth_dev parameter should be passed to the underlying driver.
Otherwise the underlying driver needs to use rte_eth_devices[] global variable to fetch the device.
This does not seems very right to me.

> +		if (!ret) {
> +			update_queue_info(rx_adapter,
> +					&rx_adapter-
> >eth_devices[eth_dev_id],
> +					rx_queue_id,
> +					1);
> +		}
> +	} else {
> +		rte_spinlock_lock(&rx_adapter->rx_lock);
> +		ret = init_service(rx_adapter, id);
> +		if (!ret)
> +			ret = add_rx_queue(rx_adapter, eth_dev_id,
> rx_queue_id,
> +					queue_conf);
> +		rte_spinlock_unlock(&rx_adapter->rx_lock);
> +		if (!ret)
> +			start_service =
> !!sw_rx_adapter_queue_count(rx_adapter);
> +	}
> +
> +	if (ret)
> +		return ret;
> +
> +	if (start_service)
> +		rte_service_component_runstate_set(rx_adapter->service_id,
> 1);
> +
> +	return 0;
> +}
> +
> +int
> +rte_event_eth_rx_adapter_queue_del(uint8_t id, uint8_t eth_dev_id,
> +				int32_t rx_queue_id)
> +{
> +	int ret = 0;
> +	struct rte_eventdev *dev;
> +	struct rte_event_eth_rx_adapter *rx_adapter;
> +	struct eth_device_info *dev_info;
> +	uint32_t rx_adapter_cap;
> +	uint16_t i;
> +
> +	RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL);
> +	RTE_ETH_VALID_PORTID_OR_ERR_RET(eth_dev_id, -EINVAL);
> +
> +	rx_adapter = id_to_rx_adapter(id);
> +	if (!rx_adapter)
> +		return -EINVAL;
> +
> +	dev = &rte_eventdevs[rx_adapter->eventdev_id];
> +	ret = dev->dev_ops->eth_rx_adapter_caps_get(dev, eth_dev_id,
> +						&rx_adapter_cap);
> +	if (ret)
> +		return ret;
> +
> +	if (rx_queue_id != -1 && (uint16_t)rx_queue_id >=
> +		rte_eth_devices[eth_dev_id].data->nb_rx_queues) {
> +		RTE_EDEV_LOG_ERR("Invalid rx queue_id %" PRIu16,
> +			 (uint16_t)rx_queue_id);
> +		return -EINVAL;
> +	}
> +
> +	dev_info = &rx_adapter->eth_devices[eth_dev_id];
> +
> +	if (rx_adapter_cap &
> RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT) {
> +		RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops-
> >eth_rx_adapter_queue_del,
> +				 -ENOTSUP);
> +		ret = (*dev->dev_ops->eth_rx_adapter_queue_del)(dev,
> eth_dev_id,
> +							rx_queue_id);

Here too rte_eth_device shall be passed.

> +		if (!ret) {
> +			update_queue_info(rx_adapter,
> +					&rx_adapter-
> >eth_devices[eth_dev_id],
> +					rx_queue_id,
> +					0);
> +			if (!dev_info->nb_dev_queues) {
> +				rte_free(dev_info->rx_queue);
> +				dev_info->rx_queue = NULL;
> +			}
> +		}
> +	} else {
> +		int rc;
> +		rte_spinlock_lock(&rx_adapter->rx_lock);
> +		if (rx_queue_id == -1) {
> +			for (i = 0; i < dev_info->dev->data->nb_rx_queues; i++)
> +
> 	_rte_event_eth_rx_adapter_queue_del(rx_adapter,
> +								dev_info,
> +								i);
> +		} else {
> +			_rte_event_eth_rx_adapter_queue_del(rx_adapter,
> +							dev_info,
> +
> 	(uint16_t)rx_queue_id);
> +		}
> +
> +		rc = eth_poll_wrr_calc(rx_adapter);
> +		if (rc)
> +			RTE_EDEV_LOG_ERR("WRR recalculation failed %"
> PRId32,
> +					rc);
> +
> +		if (!dev_info->nb_dev_queues) {
> +			rte_free(dev_info->rx_queue);
> +			dev_info->rx_queue = NULL;
> +		}
> +
> +		rte_spinlock_unlock(&rx_adapter->rx_lock);
> +		rte_service_component_runstate_set(rx_adapter->service_id,
> +				sw_rx_adapter_queue_count(rx_adapter));
> +	}
> +
> +	return ret;
> +}
> +
> +
> +int
> +rte_event_eth_rx_adapter_start(uint8_t id)
> +{
> +	return rx_adapter_ctrl(id, 1);
> +}
> +
> +int
> +rte_event_eth_rx_adapter_stop(uint8_t id)
> +{
> +	return rx_adapter_ctrl(id, 0);
> +}
> +
> +int
> +rte_event_eth_rx_adapter_stats_get(uint8_t id,
> +			       struct rte_event_eth_rx_adapter_stats *stats)
> +{
> +	struct rte_event_eth_rx_adapter *rx_adapter;
> +	struct rte_event_eth_rx_adapter_stats dev_stats_sum = { 0 };
> +	struct rte_event_eth_rx_adapter_stats dev_stats;
> +	struct rte_eventdev *dev;
> +	struct eth_device_info *dev_info;
> +	uint32_t i;
> +	int ret;
> +
> +	RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL);
> +
> +	rx_adapter = id_to_rx_adapter(id);
> +	if (!rx_adapter || !stats)
> +		return -EINVAL;
> +
> +	dev = &rte_eventdevs[rx_adapter->eventdev_id];
> +	memset(stats, 0, sizeof(*stats));
> +	for (i = 0; i < rte_eth_dev_count(); i++) {
> +		dev_info = &rx_adapter->eth_devices[i];
> +		if (!dev_info->internal_event_port)
> +			continue;
> +		ret = (*dev->dev_ops->eth_rx_adapter_stats_get)(dev, i,
> +						&dev_stats);

Preferable to check if the adapter stats get and stats set API are registered from the driver.

Regards,
Nipun