17. L2 Forwarding Eventdev Sample Application

The L2 Forwarding eventdev sample application is a simple example of packet processing using the Data Plane Development Kit (DPDK) to demonstrate usage of poll and event mode packet I/O mechanism.

17.1. Overview

The L2 Forwarding eventdev sample application, performs L2 forwarding for each packet that is received on an RX_PORT. The destination port is the adjacent port from the enabled portmask, that is, if the first four ports are enabled (portmask=0x0f), ports 1 and 2 forward into each other, and ports 3 and 4 forward into each other. Also, if MAC addresses updating is enabled, the MAC addresses are affected as follows:

  • The source MAC address is replaced by the TX_PORT MAC address

  • The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID

Application receives packets from RX_PORT using below mentioned methods:

  • Poll mode

  • Eventdev mode (default)

This application can be used to benchmark performance using a traffic-generator, as shown in the Fig. 17.1.

../_images/l2_fwd_benchmark_setup.svg

Fig. 17.1 Performance Benchmark Setup (Basic Environment)

17.2. Compiling the Application

To compile the sample application see Compiling the Sample Applications.

The application is located in the l2fwd-event sub-directory.

17.3. Running the Application

The application requires a number of command line options:

./<build_dir>/examples/dpdk-l2fwd-event [EAL options] -- -p PORTMASK
                                                    [-q NQ]
                                                    [--[no-]mac-updating]
                                                    [--mode=MODE]
                                                    [--eventq-sched=SCHED_MODE]
                                                    [--event-vector [--event-vector-size SIZE] [--event-vector-tmo NS]]

where,

  • p PORTMASK: A hexadecimal bitmask of the ports to configure

  • q NQ: Maximum number of queues per lcore (default is 1)

  • –[no-]mac-updating: Enable or disable MAC addresses updating (enabled by default).

  • –mode=MODE: Packet transfer mode for I/O, poll or eventdev. Eventdev by default.

  • –eventq-sched=SCHED_MODE: Event queue schedule mode, Ordered, Atomic or Parallel. Atomic by default.

  • –config: Configure forwarding port pair mapping. Alternate port pairs by default.

  • –event-vector: Enable event vectorization. Only valid if –mode=eventdev.

  • –event-vector-size: Max vector size if event vectorization is enabled.

  • –event-vector-tmo: Max timeout to form vector in nanoseconds if event vectorization is enabled.

Sample usage commands are given below to run the application into different mode:

Poll mode with 4 lcores, 16 ports and 8 RX queues per lcore and MAC address updating enabled, issue the command:

./<build_dir>/examples/dpdk-l2fwd-event -l 0-3 -n 4 -- -q 8 -p ffff --mode=poll

Eventdev mode with 4 lcores, 16 ports , sched method ordered and MAC address updating enabled, issue the command:

./<build_dir>/examples/dpdk-l2fwd-event -l 0-3 -n 4 -- -p ffff --eventq-sched=ordered

or

./<build_dir>/examples/dpdk-l2fwd-event -l 0-3 -n 4 -- -q 8 -p ffff --mode=eventdev --eventq-sched=ordered

Refer to the DPDK Getting Started Guide for general information on running applications and the Environment Abstraction Layer (EAL) options.

To run application with S/W scheduler, it uses following DPDK services:

  • Software scheduler

  • Rx adapter service function

  • Tx adapter service function

Application needs service cores to run above mentioned services. Service cores must be provided as EAL parameters along with the –vdev=event_sw0 to enable S/W scheduler. Following is the sample command:

./<build_dir>/examples/dpdk-l2fwd-event -l 0-7 -s 0-3 -n 4 --vdev event_sw0 -- -q 8 -p ffff --mode=eventdev --eventq-sched=ordered

17.4. Explanation

The following sections provide some explanation of the code.

17.4.1. Command Line Arguments

The L2 Forwarding eventdev sample application takes specific parameters, in addition to Environment Abstraction Layer (EAL) arguments. The preferred way to parse parameters is to use the getopt() function, since it is part of a well-defined and portable library.

The parsing of arguments is done in the l2fwd_parse_args() function for non eventdev parameters and in parse_eventdev_args() for eventdev parameters. The method of argument parsing is not described here. Refer to the glibc getopt(3) man page for details.

EAL arguments are parsed first, then application-specific arguments. This is done at the beginning of the main() function and eventdev parameters are parsed in eventdev_resource_setup() function during eventdev setup:

ret = rte_eal_init(argc, argv);
if (ret < 0)
	rte_panic("Invalid EAL arguments\n");
argc -= ret;
argv += ret;

rsrc = l2fwd_get_rsrc();

signal(SIGINT, signal_handler);
signal(SIGTERM, signal_handler);

/* parse application arguments (after the EAL ones) */
ret = l2fwd_event_parse_args(argc, argv, rsrc);
if (ret < 0)
	rte_panic("Invalid L2FWD arguments\n");

17.4.2. Mbuf Pool Initialization

Once the arguments are parsed, the mbuf pool is created. The mbuf pool contains a set of mbuf objects that will be used by the driver and the application to store network packet data:

rsrc->pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool",
		nb_mbufs, MEMPOOL_CACHE_SIZE, 0,
		RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
if (rsrc->pktmbuf_pool == NULL)
	rte_panic("Cannot init mbuf pool\n");

The rte_mempool is a generic structure used to handle pools of objects. In this case, it is necessary to create a pool that will be used by the driver. The number of allocated pkt mbufs is NB_MBUF, with a data room size of RTE_MBUF_DEFAULT_BUF_SIZE each. A per-lcore cache of 32 mbufs is kept. The memory is allocated in NUMA socket 0, but it is possible to extend this code to allocate one mbuf pool per socket.

The rte_pktmbuf_pool_create() function uses the default mbuf pool and mbuf initializers, respectively rte_pktmbuf_pool_init() and rte_pktmbuf_init(). An advanced application may want to use the mempool API to create the mbuf pool with more control.

17.4.3. Driver Initialization

The main part of the code in the main() function relates to the initialization of the driver. To fully understand this code, it is recommended to study the chapters that related to the Poll Mode and Event mode Driver in the DPDK Programmer’s Guide - Rel 1.4 EAR and the DPDK API Reference.

for (port_id = 0; port_id < RTE_MAX_ETHPORTS; port_id++)
	rsrc->dst_ports[port_id] = UINT32_MAX;

argvopt = argv;
while ((opt = getopt_long(argc, argvopt, short_options,
			  lgopts, &option_index)) != EOF) {

	switch (opt) {
	/* portmask */
	case 'p':
		rsrc->enabled_port_mask =
				l2fwd_event_parse_portmask(optarg);
		if (rsrc->enabled_port_mask == 0) {
			printf("invalid portmask\n");
			l2fwd_event_usage(prgname);
			return -1;
		}
		break;

	/* nqueue */
	case 'q':
		rsrc->rx_queue_per_lcore =
				l2fwd_event_parse_nqueue(optarg);
		if (rsrc->rx_queue_per_lcore == 0) {
			printf("invalid queue number\n");
			l2fwd_event_usage(prgname);
			return -1;
		}
		break;

	/* timer period */
	case 'T':
		timer_secs = l2fwd_event_parse_timer_period(optarg);
		if (timer_secs < 0) {
			printf("invalid timer period\n");
			l2fwd_event_usage(prgname);
			return -1;
		}
		rsrc->timer_period = timer_secs;
		/* convert to number of cycles */
		rsrc->timer_period *= rte_get_timer_hz();
		break;

	case CMD_LINE_OPT_MODE_NUM:
		l2fwd_event_parse_mode(optarg, rsrc);
		break;

	case CMD_LINE_OPT_EVENTQ_SCHED_NUM:
		l2fwd_event_parse_eventq_sched(optarg, rsrc);
		break;

	case CMD_LINE_OPT_PORT_PAIR_CONF_NUM:
		ret = l2fwd_parse_port_pair_config(optarg, rsrc);
		if (ret) {
			printf("Invalid port pair config\n");
			l2fwd_event_usage(prgname);
			return -1;
		}
		break;
	case CMD_LINE_OPT_ENABLE_VECTOR_NUM:
		printf("event vectorization is enabled\n");
		rsrc->evt_vec.enabled = 1;
		break;
	case CMD_LINE_OPT_VECTOR_SIZE_NUM:
		rsrc->evt_vec.size = strtol(optarg, NULL, 10);
		break;
	case CMD_LINE_OPT_VECTOR_TMO_NS_NUM:
		rsrc->evt_vec.timeout_ns = strtoull(optarg, NULL, 10);
		break;

	/* long options */
	case 0:
		break;

	default:
		l2fwd_event_usage(prgname);
		return -1;
	}
}

rsrc->mac_updating = mac_updating;

if (rsrc->evt_vec.enabled && !rsrc->evt_vec.size) {
	rsrc->evt_vec.size = VECTOR_SIZE_DEFAULT;
	printf("vector size set to default (%" PRIu16 ")\n",
	       rsrc->evt_vec.size);
}

if (rsrc->evt_vec.enabled && !rsrc->evt_vec.timeout_ns) {
	rsrc->evt_vec.timeout_ns = VECTOR_TMO_NS_DEFAULT;
	printf("vector timeout set to default (%" PRIu64 " ns)\n",
	       rsrc->evt_vec.timeout_ns);
}

if (optind >= 0)
	argv[optind-1] = prgname;

ret = optind-1;
optind = 1; /* reset getopt lib */
return ret;

The next step is to configure the RX and TX queues. For each port, there is only one RX queue (only one lcore is able to poll a given port). The number of TX queues depends on the number of available lcores. The rte_eth_dev_configure() function is used to configure the number of queues for a port:

ret = rte_eth_dev_configure(port_id, 1, 1, &local_port_conf);
if (ret < 0)
	rte_panic("Cannot configure device: err=%d, port=%u\n",
		  ret, port_id);

17.4.4. RX Queue Initialization

The application uses one lcore to poll one or several ports, depending on the -q option, which specifies the number of queues per lcore.

For example, if the user specifies -q 4, the application is able to poll four ports with one lcore. If there are 16 ports on the target (and if the portmask argument is -p ffff ), the application will need four lcores to poll all the ports.

ret = rte_eth_rx_queue_setup(port_id, 0, nb_rxd,
			     rte_eth_dev_socket_id(port_id),
			     &rxq_conf,
			     rsrc->pktmbuf_pool);
if (ret < 0)
	rte_panic("rte_eth_rx_queue_setup:err=%d, port=%u\n",
		  ret, port_id);

The list of queues that must be polled for a given lcore is stored in a private structure called struct lcore_queue_conf.

struct __rte_cache_aligned lcore_queue_conf {
	unsigned n_rx_port;
	unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
};
struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];

The values n_rx_port and rx_port_list[] are used in the main packet processing loop (see Receive, Process and Transmit Packets).

17.4.5. TX Queue Initialization

Each lcore should be able to transmit on any port. For every port, a single TX queue is initialized.

fflush(stdout);
txq_conf = dev_info.default_txconf;
txq_conf.offloads = local_port_conf.txmode.offloads;
ret = rte_eth_tx_queue_setup(port_id, 0, nb_txd,
		rte_eth_dev_socket_id(port_id),
		&txq_conf);
if (ret < 0)
	rte_panic("rte_eth_tx_queue_setup:err=%d, port=%u\n",
		  ret, port_id);

To configure eventdev support, application setups following components:

  • Event dev

  • Event queue

  • Event Port

  • Rx/Tx adapters

  • Ethernet ports

17.4.6. Event device Initialization

Application can use either H/W or S/W based event device scheduler implementation and supports single instance of event device. It configures event device as per below configuration

struct rte_event_dev_config event_d_conf = {
	.nb_events_limit  = 4096,
	.nb_event_queue_flows = 1024,
	.nb_event_port_dequeue_depth = 128,
	.nb_event_port_enqueue_depth = 128
};

In case of S/W scheduler, application runs eventdev scheduler service on service core. Application retrieves service id and finds the best possible service core to run S/W scheduler.

rte_event_dev_info_get(evt_rsrc->event_d_id, &evdev_info);
if (!(evdev_info.event_dev_cap & RTE_EVENT_DEV_CAP_DISTRIBUTED_SCHED)) {
	ret = rte_event_dev_service_id_get(evt_rsrc->event_d_id,
			&service_id);
	if (ret != -ESRCH && ret != 0)
		rte_panic("Error in starting eventdev service\n");
	l2fwd_event_service_enable(service_id);
}

17.4.7. Event queue Initialization

Each Ethernet device is assigned a dedicated event queue which will be linked to all available event ports i.e. each lcore can dequeue packets from any of the Ethernet ports.

struct rte_event_queue_conf event_q_conf = {
	.nb_atomic_flows = 1024,
	.nb_atomic_order_sequences = 1024,
	.event_queue_cfg = event_queue_cfg,
	.priority = RTE_EVENT_DEV_PRIORITY_NORMAL
};
struct rte_event_queue_conf def_q_conf;
uint8_t event_q_id;
int32_t ret;

event_q_conf.schedule_type = rsrc->sched_type;
evt_rsrc->evq.event_q_id = (uint8_t *)malloc(sizeof(uint8_t) *
				evt_rsrc->evq.nb_queues);
if (!evt_rsrc->evq.event_q_id)
	rte_panic("Memory allocation failure\n");

ret = rte_event_queue_default_conf_get(event_d_id, 0, &def_q_conf);
if (ret < 0)
	rte_panic("Error to get default config of event queue\n");

In case of S/W scheduler, an extra event queue is created which will be used for Tx adapter service function for enqueue operation.

17.4.8. Event port Initialization

Each worker thread is assigned a dedicated event port for enq/deq operations to/from an event device. All event ports are linked with all available event queues.

struct rte_event_port_conf event_p_conf = {
	.dequeue_depth = 32,
	.enqueue_depth = 32,
	.new_event_threshold = 4096
};
struct rte_event_port_conf def_p_conf;
uint8_t event_p_id;
int32_t ret;

evt_rsrc->evp.event_p_id = (uint8_t *)malloc(sizeof(uint8_t) *
				evt_rsrc->evp.nb_ports);
if (!evt_rsrc->evp.event_p_id)
	rte_panic("No space is available\n");

memset(&def_p_conf, 0, sizeof(struct rte_event_port_conf));
ret = rte_event_port_default_conf_get(event_d_id, 0, &def_p_conf);
if (ret < 0)
	rte_panic("Error to get default configuration of event port\n");

if (def_p_conf.new_event_threshold < event_p_conf.new_event_threshold)
	event_p_conf.new_event_threshold =
		def_p_conf.new_event_threshold;

if (def_p_conf.dequeue_depth < event_p_conf.dequeue_depth)
	event_p_conf.dequeue_depth = def_p_conf.dequeue_depth;

if (def_p_conf.enqueue_depth < event_p_conf.enqueue_depth)
	event_p_conf.enqueue_depth = def_p_conf.enqueue_depth;

event_p_conf.event_port_cfg = 0;
if (evt_rsrc->disable_implicit_release)
	event_p_conf.event_port_cfg |=
		RTE_EVENT_PORT_CFG_DISABLE_IMPL_REL;

evt_rsrc->deq_depth = def_p_conf.dequeue_depth;

for (event_p_id = 0; event_p_id < evt_rsrc->evp.nb_ports;
							event_p_id++) {
	ret = rte_event_port_setup(event_d_id, event_p_id,
				   &event_p_conf);
	if (ret < 0)
		rte_panic("Error in configuring event port %d\n",
			  event_p_id);

	ret = rte_event_port_link(event_d_id, event_p_id,
				  evt_rsrc->evq.event_q_id,
				  NULL,
				  evt_rsrc->evq.nb_queues - 1);
	if (ret != (evt_rsrc->evq.nb_queues - 1))
		rte_panic("Error in linking event port %d to queues\n",
			  event_p_id);
	evt_rsrc->evp.event_p_id[event_p_id] = event_p_id;

In case of S/W scheduler, an extra event port is created by DPDK library which is retrieved by the application and same will be used by Tx adapter service.

ret = rte_event_eth_tx_adapter_event_port_get(tx_adptr_id, &tx_port_id);
if (ret)
	rte_panic("Failed to get Tx adapter port id: %d\n", ret);

ret = rte_event_port_link(event_d_id, tx_port_id,
			  &evt_rsrc->evq.event_q_id[
				evt_rsrc->evq.nb_queues - 1],
			  NULL, 1);
if (ret != 1)
	rte_panic("Unable to link Tx adapter port to Tx queue:err=%d\n",
		 ret);

17.4.9. Rx/Tx adapter Initialization

Each Ethernet port is assigned a dedicated Rx/Tx adapter for H/W scheduler. Each Ethernet port’s Rx queues are connected to its respective event queue at priority 0 via Rx adapter configuration and Ethernet port’s tx queues are connected via Tx adapter.

RTE_ETH_FOREACH_DEV(port_id) {
	if ((rsrc->enabled_port_mask & (1 << port_id)) == 0)
		continue;

	if (rsrc->evt_vec.enabled) {
		uint32_t cap;

		if (rte_event_eth_rx_adapter_caps_get(event_d_id,
						      port_id, &cap))
			rte_panic(
				"Failed to get event rx adapter capability");

		if (cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_EVENT_VECTOR) {
			eth_q_conf.vector_sz = rsrc->evt_vec.size;
			eth_q_conf.vector_timeout_ns =
				rsrc->evt_vec.timeout_ns;
			eth_q_conf.vector_mp = rsrc->evt_vec_pool;
			eth_q_conf.rx_queue_flags |=
			RTE_EVENT_ETH_RX_ADAPTER_QUEUE_EVENT_VECTOR;
		} else {
			rte_panic(
				"Rx adapter doesn't support event vector");
		}
	}

	ret = rte_event_eth_rx_adapter_create(adapter_id, event_d_id,
					&evt_rsrc->def_p_conf);
	if (ret)
		rte_panic("Failed to create rx adapter[%d]\n",
			  adapter_id);

	/* Configure user requested sched type*/
	eth_q_conf.ev.sched_type = rsrc->sched_type;
	eth_q_conf.ev.queue_id = evt_rsrc->evq.event_q_id[q_id];
	ret = rte_event_eth_rx_adapter_queue_add(adapter_id, port_id,
						 -1, &eth_q_conf);
	if (ret)
		rte_panic("Failed to add queues to Rx adapter\n");

	ret = rte_event_eth_rx_adapter_start(adapter_id);
	if (ret)
		rte_panic("Rx adapter[%d] start Failed\n", adapter_id);

	evt_rsrc->rx_adptr.rx_adptr[adapter_id] = adapter_id;
	adapter_id++;
	if (q_id < evt_rsrc->evq.nb_queues)
		q_id++;
}

evt_rsrc->tx_adptr.nb_tx_adptr = nb_adapter;
evt_rsrc->tx_adptr.tx_adptr = (uint8_t *)malloc(sizeof(uint8_t) *
				evt_rsrc->tx_adptr.nb_tx_adptr);
if (!evt_rsrc->tx_adptr.tx_adptr) {
	free(evt_rsrc->rx_adptr.rx_adptr);
	free(evt_rsrc->evp.event_p_id);
	free(evt_rsrc->evq.event_q_id);
	rte_panic("Failed to allocate memery for Rx adapter\n");
}

adapter_id = 0;
RTE_ETH_FOREACH_DEV(port_id) {
	if ((rsrc->enabled_port_mask & (1 << port_id)) == 0)
		continue;
	ret = rte_event_eth_tx_adapter_create(adapter_id, event_d_id,
					&evt_rsrc->def_p_conf);
	if (ret)
		rte_panic("Failed to create tx adapter[%d]\n",
			  adapter_id);

	ret = rte_event_eth_tx_adapter_queue_add(adapter_id, port_id,
						 -1);
	if (ret)
		rte_panic("Failed to add queues to Tx adapter\n");

	ret = rte_event_eth_tx_adapter_start(adapter_id);
	if (ret)
		rte_panic("Tx adapter[%d] start Failed\n", adapter_id);

	evt_rsrc->tx_adptr.tx_adptr[adapter_id] = adapter_id;
	adapter_id++;
}

For S/W scheduler instead of dedicated adapters, common Rx/Tx adapters are configured which will be shared among all the Ethernet ports. Also DPDK library need service cores to run internal services for Rx/Tx adapters. Application gets service id for Rx/Tx adapters and after successful setup it runs the services on dedicated service cores.

for (i = 0; i < evt_rsrc->rx_adptr.nb_rx_adptr; i++) {
	ret = rte_event_eth_rx_adapter_caps_get(evt_rsrc->event_d_id,
			evt_rsrc->rx_adptr.rx_adptr[i], &caps);
	if (ret < 0)
		rte_panic("Failed to get Rx adapter[%d] caps\n",
			  evt_rsrc->rx_adptr.rx_adptr[i]);
	ret = rte_event_eth_rx_adapter_service_id_get(
			evt_rsrc->event_d_id,
			&service_id);
	if (ret != -ESRCH && ret != 0)
		rte_panic("Error in starting Rx adapter[%d] service\n",
			  evt_rsrc->rx_adptr.rx_adptr[i]);
	l2fwd_event_service_enable(service_id);
}

for (i = 0; i < evt_rsrc->tx_adptr.nb_tx_adptr; i++) {
	ret = rte_event_eth_tx_adapter_caps_get(evt_rsrc->event_d_id,
			evt_rsrc->tx_adptr.tx_adptr[i], &caps);
	if (ret < 0)
		rte_panic("Failed to get Rx adapter[%d] caps\n",
			  evt_rsrc->tx_adptr.tx_adptr[i]);
	ret = rte_event_eth_tx_adapter_service_id_get(
			evt_rsrc->event_d_id,
			&service_id);
	if (ret != -ESRCH && ret != 0)
		rte_panic("Error in starting Rx adapter[%d] service\n",
			  evt_rsrc->tx_adptr.tx_adptr[i]);
	l2fwd_event_service_enable(service_id);
}

17.4.10. Receive, Process and Transmit Packets

In the l2fwd_main_loop() function, the main task is to read ingress packets from the RX queues. This is done using the following code:


/* Read packet from RX queues */
for (i = 0; i < qconf->n_rx_port; i++) {

	port_id = qconf->rx_port_list[i];
	nb_rx = rte_eth_rx_burst(port_id, 0, pkts_burst,
				 MAX_PKT_BURST);

	rsrc->port_stats[port_id].rx += nb_rx;

	for (j = 0; j < nb_rx; j++) {
		m = pkts_burst[j];
		rte_prefetch0(rte_pktmbuf_mtod(m, void *));
		l2fwd_poll_simple_forward(rsrc, m, port_id);
	}
}

Packets are read in a burst of size MAX_PKT_BURST. The rte_eth_rx_burst() function writes the mbuf pointers in a local table and returns the number of available mbufs in the table.

Then, each mbuf in the table is processed by the l2fwd_simple_forward() function. The processing is very simple: process the TX port from the RX port, then replace the source and destination MAC addresses if MAC addresses updating is enabled.

During the initialization process, a static array of destination ports (l2fwd_dst_ports[]) is filled such that for each source port, a destination port is assigned that is either the next or previous enabled port from the portmask. If number of ports are odd in portmask then packet from last port will be forwarded to first port i.e. if portmask=0x07, then forwarding will take place like p0—>p1, p1—>p2, p2—>p0.

Also to optimize enqueue operation, l2fwd_simple_forward() stores incoming mbufs up to MAX_PKT_BURST. Once it reaches up to limit, all packets are transmitted to destination ports.

static void
l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
{
	unsigned dst_port;
	int sent;
	struct rte_eth_dev_tx_buffer *buffer;

	dst_port = l2fwd_dst_ports[portid];

	if (mac_updating)
		l2fwd_mac_updating(m, dst_port);

	buffer = tx_buffer[dst_port];
	sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
	if (sent)
		port_statistics[dst_port].tx += sent;
}

For this test application, the processing is exactly the same for all packets arriving on the same RX port. Therefore, it would have been possible to call the rte_eth_tx_buffer() function directly from the main loop to send all the received packets on the same TX port, using the burst-oriented send function, which is more efficient.

However, in real-life applications (such as, L3 routing), packet N is not necessarily forwarded on the same port as packet N-1. The application is implemented to illustrate that, so the same approach can be reused in a more complex application.

To ensure that no packets remain in the tables, each lcore does a draining of TX queue in its main loop. This technique introduces some latency when there are not many packets to send, however it improves performance:

cur_tsc = rte_rdtsc();

/*
 * TX burst queue drain
 */
diff_tsc = cur_tsc - prev_tsc;
if (unlikely(diff_tsc > drain_tsc)) {
	for (i = 0; i < qconf->n_rx_port; i++) {
		port_id =
			rsrc->dst_ports[qconf->rx_port_list[i]];
		buf = poll_rsrc->tx_buffer[port_id];
		sent = rte_eth_tx_buffer_flush(port_id, 0, buf);
		if (sent)
			rsrc->port_stats[port_id].tx += sent;
	}

	prev_tsc = cur_tsc;
}

In the l2fwd_event_loop() function, the main task is to read ingress packets from the event ports. This is done using the following code:

nb_rx = rte_event_dequeue_burst(event_d_id, port_id, ev,
				deq_len, 0);
if (nb_rx == 0)
	continue;

for (i = 0; i < nb_rx; i++) {
	l2fwd_event_fwd(rsrc, &ev[i], tx_q_id, timer_period,
			flags);
}

Before reading packets, deq_len is fetched to ensure correct allowed deq length by the eventdev. The rte_event_dequeue_burst() function writes the mbuf pointers in a local table and returns the number of available mbufs in the table.

Then, each mbuf in the table is processed by the l2fwd_eventdev_forward() function. The processing is very simple: process the TX port from the RX port, then replace the source and destination MAC addresses if MAC addresses updating is enabled.

During the initialization process, a static array of destination ports (l2fwd_dst_ports[]) is filled such that for each source port, a destination port is assigned that is either the next or previous enabled port from the portmask. If number of ports are odd in portmask then packet from last port will be forwarded to first port i.e. if portmask=0x07, then forwarding will take place like p0—>p1, p1—>p2, p2—>p0.

l2fwd_eventdev_forward() does not stores incoming mbufs. Packet will forwarded be to destination ports via Tx adapter or generic event dev enqueue API depending H/W or S/W scheduler is used.

nb_rx = rte_event_dequeue_burst(event_d_id, port_id, ev,
				deq_len, 0);
if (nb_rx == 0)
	continue;

for (i = 0; i < nb_rx; i++) {
	l2fwd_event_fwd(rsrc, &ev[i], tx_q_id, timer_period,
			flags);
}