51. Event Ethernet Tx Adapter Library

The DPDK Eventdev API allows the application to use an event driven programming model for packet processing in which the event device distributes events referencing packets to the application cores in a dynamic load balanced fashion while handling atomicity and packet ordering. Event adapters provide the interface between the ethernet, crypto and timer devices and the event device. Event adapter APIs enable common application code by abstracting PMD specific capabilities. The Event ethernet Tx adapter provides configuration and data path APIs for the transmit stage of the application allowing the same application code to use eventdev PMD support or in its absence, a common implementation.

In the common implementation, the application enqueues mbufs to the adapter which runs as a rte_service function. The service function dequeues events from its event port and transmits the mbufs referenced by these events.

51.1. API Walk-through

This section will introduce the reader to the adapter API. The application has to first instantiate an adapter which is associated with a single eventdev, next the adapter instance is configured with Tx queues, finally the adapter is started and the application can start enqueuing mbufs to it.

51.1.1. Creating an Adapter Instance

An adapter instance is created using rte_event_eth_tx_adapter_create(). This function is passed the event device to be associated with the adapter and port configuration for the adapter to setup an event port if the adapter needs to use a service function.

If the application desires to have finer control of eventdev port configuration, it can use the rte_event_eth_tx_adapter_create_ext() function. The rte_event_eth_tx_adapter_create_ext() function is passed a callback function. The callback function is invoked if the adapter needs to use a service function and needs to create an event port for it. The callback is expected to fill the struct rte_event_eth_tx_adapter_conf structure passed to it.

struct rte_event_dev_info dev_info;
struct rte_event_port_conf tx_p_conf = {0};

err = rte_event_dev_info_get(id, &dev_info);

tx_p_conf.new_event_threshold = dev_info.max_num_events;
tx_p_conf.dequeue_depth = dev_info.max_event_port_dequeue_depth;
tx_p_conf.enqueue_depth = dev_info.max_event_port_enqueue_depth;

err = rte_event_eth_tx_adapter_create(id, dev_id, &tx_p_conf);

51.1.2. Event device configuration for service based adapter

When rte_event_eth_tx_adapter_create() is used for creating adapter instance, rte_event_dev_config::nb_event_ports is automatically incremented, and the event device is reconfigured with additional event port during service initialization. This event device reconfigure logic also increments the rte_event_dev_config::nb_single_link_event_port_queues parameter if the adapter event port config is of type RTE_EVENT_PORT_CFG_SINGLE_LINK.

Application no longer needs to configure the event device with rte_event_dev_config::nb_event_ports and rte_event_dev_config::nb_single_link_event_port_queues parameters required for eth Tx adapter when the adapter is created using the above-mentioned API.

51.1.3. Adding Tx Queues to the Adapter Instance

Ethdev Tx queues are added to the instance using the rte_event_eth_tx_adapter_queue_add() function. A queue value of -1 is used to indicate all queues within a device.

int err = rte_event_eth_tx_adapter_queue_add(id,
                                             eth_dev_id,
                                             q);

51.1.4. Querying Adapter Capabilities

The rte_event_eth_tx_adapter_caps_get() function allows the application to query the adapter capabilities for an eventdev and ethdev combination. Currently, the only capability flag defined is RTE_EVENT_ETH_TX_ADAPTER_CAP_INTERNAL_PORT, the application can query this flag to determine if a service function is associated with the adapter and retrieve its service identifier using the rte_event_eth_tx_adapter_service_id_get() API.

int err = rte_event_eth_tx_adapter_caps_get(dev_id, eth_dev_id, &cap);

if (!(cap & RTE_EVENT_ETH_TX_ADAPTER_CAP_INTERNAL_PORT))
        err = rte_event_eth_tx_adapter_service_id_get(id, &service_id);

51.1.5. Linking a Queue to the Adapter’s Event Port

If the adapter uses a service function as described in the previous section, the application is required to link a queue to the adapter’s event port. The adapter’s event port can be obtained using the rte_event_eth_tx_adapter_event_port_get() function. The queue can be configured with the RTE_EVENT_QUEUE_CFG_SINGLE_LINK since it is linked to a single event port.

51.1.6. Configuring the Service Function

If the adapter uses a service function, the application can assign a service core to the service function as shown below.

if (rte_event_eth_tx_adapter_service_id_get(id, &service_id) == 0)
        rte_service_map_lcore_set(service_id, TX_CORE_ID);

51.1.7. Starting the Adapter Instance

The application calls rte_event_eth_tx_adapter_start() to start the adapter. This function calls the start callback of the eventdev PMD if supported, and the rte_service_run_state_set() to enable the service function if one exists.

51.1.8. Enqueuing Packets to the Adapter

The application needs to notify the adapter about the transmit port and queue used to send the packet. The transmit port is set in the struct rte mbuf::port field and the transmit queue is set using the rte_event_eth_tx_adapter_txq_set() function.

If the eventdev PMD supports the RTE_EVENT_ETH_TX_ADAPTER_CAP_INTERNAL_PORT capability for a given ethernet device, the application should use the rte_event_eth_tx_adapter_enqueue() function to enqueue packets to the adapter.

If the adapter uses a service function for the ethernet device then the application should use the rte_event_enqueue_burst() function.

struct rte_event event;

if (cap & RTE_EVENT_ETH_TX_ADAPTER_CAP_INTERNAL_PORT) {

        event.mbuf = m;
        eq_flags = 0;

        m->port = tx_port;
        rte_event_eth_tx_adapter_txq_set(m, tx_queue_id);

        rte_event_eth_tx_adapter_enqueue(dev_id, ev_port, &event, 1, eq_flags);
} else {

        event.queue_id = qid; /* event queue linked to adapter port */
        event.op = RTE_EVENT_OP_NEW;
        event.event_type = RTE_EVENT_TYPE_CPU;
        event.sched_type = RTE_SCHED_TYPE_ATOMIC;
        event.mbuf = m;

        m->port = tx_port;
        rte_event_eth_tx_adapter_txq_set(m, tx_queue_id);

        rte_event_enqueue_burst(dev_id, ev_port, &event, 1);
}

51.1.9. Getting Adapter Statistics

The rte_event_eth_tx_adapter_stats_get() function reports counters defined in struct rte_event_eth_tx_adapter_stats. The counter values are the sum of the counts from the eventdev PMD callback if the callback is supported, and the counts maintained by the service function, if one exists.

51.1.10. Getting Adapter Instance ID

The rte_event_eth_tx_adapter_instance_get() function reports Tx adapter instance ID for a specified ethernet device ID and Tx queue index.

51.1.11. Tx event vectorization

The event device, ethernet device pairs which support the capability RTE_EVENT_ETH_TX_ADAPTER_CAP_EVENT_VECTOR can process event vector of mbufs. Additionally, application can provide a hint to the Tx adapter that all the mbufs are destined to the same ethernet port and queue by setting the bit rte_event_vector::attr_valid and filling rte_event_vector::port` and rte_event_vector::queue. If rte_event_vector::attr_valid is not set then the Tx adapter should peek into each mbuf and transmit them to the requested ethernet port and queue pair.

51.1.12. Queue start/stop

The adapter can be configured to start/stop enqueueing of packets to a associated NIC queue using rte_event_eth_tx_adapter_queue_start() or rte_event_eth_tx_adapter_queue_stop() respectively. By default the queue is in start state.

These APIs help avoid some unexpected behavior with application stopping ethdev Tx queues and adapter being unaware of it. With these APIs, the application can call stop API to notify adapter that corresponding ethdev Tx queue is stopped and any in-flight packets are freed by adapter dataplane code. Adapter queue stop API is called before stopping the ethdev Tx queue. When ethdev Tx queue is enabled, application can notify adapter to resume processing of the packets for that queue by calling the start API. The ethdev Tx queue is started before calling adapter start API.

Start function enables the adapter runtime to start enqueueing of packets to the Tx queue.

Stop function stops the adapter runtime function from enqueueing any packets to the associated Tx queue. This API also frees any packets that may have been buffered for this queue. All inflight packets destined to the queue are freed by the adapter runtime until the queue is started again.

51.1.13. Set/Get adapter runtime configuration parameters

The runtime configuration parameters of adapter can be set/get using rte_event_eth_tx_adapter_runtime_params_set() and rte_event_eth_tx_adapter_runtime_params_get() respectively. The parameters that can be set/get are defined in struct rte_event_eth_tx_adapter_runtime_params.