1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2010-2014 Intel Corporation.
6 L2 Forwarding Eventdev Sample Application
7 =========================================
9 The L2 Forwarding eventdev sample application is a simple example of packet
10 processing using the Data Plane Development Kit (DPDK) to demonstrate usage of
11 poll and event mode packet I/O mechanism.
16 The L2 Forwarding eventdev sample application, performs L2 forwarding for each
17 packet that is received on an RX_PORT. The destination port is the adjacent port
18 from the enabled portmask, that is, if the first four ports are enabled (portmask=0x0f),
19 ports 1 and 2 forward into each other, and ports 3 and 4 forward into each other.
20 Also, if MAC addresses updating is enabled, the MAC addresses are affected as follows:
22 * The source MAC address is replaced by the TX_PORT MAC address
24 * The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID
26 Application receives packets from RX_PORT using below mentioned methods:
30 * Eventdev mode (default)
32 This application can be used to benchmark performance using a traffic-generator,
33 as shown in the :numref:`figure_l2fwd_event_benchmark_setup`.
35 .. _figure_l2fwd_event_benchmark_setup:
37 .. figure:: img/l2_fwd_benchmark_setup.*
39 Performance Benchmark Setup (Basic Environment)
41 Compiling the Application
42 -------------------------
44 To compile the sample application see :doc:`compiling`.
46 The application is located in the ``l2fwd-event`` sub-directory.
48 Running the Application
49 -----------------------
51 The application requires a number of command line options:
53 .. code-block:: console
55 ./build/l2fwd-event [EAL options] -- -p PORTMASK [-q NQ] --[no-]mac-updating --mode=MODE --eventq-sched=SCHED_MODE
59 * p PORTMASK: A hexadecimal bitmask of the ports to configure
61 * q NQ: A number of queues (=ports) per lcore (default is 1)
63 * --[no-]mac-updating: Enable or disable MAC addresses updating (enabled by default).
65 * --mode=MODE: Packet transfer mode for I/O, poll or eventdev. Eventdev by default.
67 * --eventq-sched=SCHED_MODE: Event queue schedule mode, Ordered, Atomic or Parallel. Atomic by default.
69 * --config: Configure forwarding port pair mapping. Alternate port pairs by default.
71 Sample usage commands are given below to run the application into different mode:
73 Poll mode with 4 lcores, 16 ports and 8 RX queues per lcore and MAC address updating enabled,
76 .. code-block:: console
78 ./build/l2fwd-event -l 0-3 -n 4 -- -q 8 -p ffff --mode=poll
80 Eventdev mode with 4 lcores, 16 ports , sched method ordered and MAC address updating enabled,
83 .. code-block:: console
85 ./build/l2fwd-event -l 0-3 -n 4 -- -p ffff --eventq-sched=ordered
89 .. code-block:: console
91 ./build/l2fwd-event -l 0-3 -n 4 -- -q 8 -p ffff --mode=eventdev --eventq-sched=ordered
93 Refer to the *DPDK Getting Started Guide* for general information on running
94 applications and the Environment Abstraction Layer (EAL) options.
96 To run application with S/W scheduler, it uses following DPDK services:
99 * Rx adapter service function
100 * Tx adapter service function
102 Application needs service cores to run above mentioned services. Service cores
103 must be provided as EAL parameters along with the --vdev=event_sw0 to enable S/W
104 scheduler. Following is the sample command:
106 .. code-block:: console
108 ./build/l2fwd-event -l 0-7 -s 0-3 -n 4 --vdev event_sw0 -- -q 8 -p ffff --mode=eventdev --eventq-sched=ordered
113 The following sections provide some explanation of the code.
115 .. _l2_fwd_event_app_cmd_arguments:
117 Command Line Arguments
118 ~~~~~~~~~~~~~~~~~~~~~~
120 The L2 Forwarding eventdev sample application takes specific parameters,
121 in addition to Environment Abstraction Layer (EAL) arguments.
122 The preferred way to parse parameters is to use the getopt() function,
123 since it is part of a well-defined and portable library.
125 The parsing of arguments is done in the **l2fwd_parse_args()** function for non
126 eventdev parameters and in **parse_eventdev_args()** for eventdev parameters.
127 The method of argument parsing is not described here. Refer to the
128 *glibc getopt(3)* man page for details.
130 EAL arguments are parsed first, then application-specific arguments.
131 This is done at the beginning of the main() function and eventdev parameters
132 are parsed in eventdev_resource_setup() function during eventdev setup:
138 ret = rte_eal_init(argc, argv);
140 rte_panic("Invalid EAL arguments\n");
145 /* parse application arguments (after the EAL ones) */
147 ret = l2fwd_parse_args(argc, argv);
149 rte_panic("Invalid L2FWD arguments\n");
154 /* Parse eventdev command line options */
155 ret = parse_eventdev_args(argc, argv);
162 .. _l2_fwd_event_app_mbuf_init:
164 Mbuf Pool Initialization
165 ~~~~~~~~~~~~~~~~~~~~~~~~
167 Once the arguments are parsed, the mbuf pool is created.
168 The mbuf pool contains a set of mbuf objects that will be used by the driver
169 and the application to store network packet data:
173 /* create the mbuf pool */
175 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
176 MEMPOOL_CACHE_SIZE, 0,
177 RTE_MBUF_DEFAULT_BUF_SIZE,
179 if (l2fwd_pktmbuf_pool == NULL)
180 rte_panic("Cannot init mbuf pool\n");
182 The rte_mempool is a generic structure used to handle pools of objects.
183 In this case, it is necessary to create a pool that will be used by the driver.
184 The number of allocated pkt mbufs is NB_MBUF, with a data room size of
185 RTE_MBUF_DEFAULT_BUF_SIZE each.
186 A per-lcore cache of 32 mbufs is kept.
187 The memory is allocated in NUMA socket 0,
188 but it is possible to extend this code to allocate one mbuf pool per socket.
190 The rte_pktmbuf_pool_create() function uses the default mbuf pool and mbuf
191 initializers, respectively rte_pktmbuf_pool_init() and rte_pktmbuf_init().
192 An advanced application may want to use the mempool API to create the
193 mbuf pool with more control.
195 .. _l2_fwd_event_app_drv_init:
197 Driver Initialization
198 ~~~~~~~~~~~~~~~~~~~~~
200 The main part of the code in the main() function relates to the initialization
201 of the driver. To fully understand this code, it is recommended to study the
202 chapters that related to the Poll Mode and Event mode Driver in the
203 *DPDK Programmer's Guide* - Rel 1.4 EAR and the *DPDK API Reference*.
207 if (rte_pci_probe() < 0)
208 rte_panic("Cannot probe PCI\n");
210 /* reset l2fwd_dst_ports */
212 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
213 l2fwd_dst_ports[portid] = 0;
218 * Each logical core is assigned a dedicated TX queue on each port.
221 RTE_ETH_FOREACH_DEV(portid) {
222 /* skip ports that are not enabled */
224 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
227 if (nb_ports_in_mask % 2) {
228 l2fwd_dst_ports[portid] = last_port;
229 l2fwd_dst_ports[last_port] = portid;
236 rte_eth_dev_info_get((uint8_t) portid, &dev_info);
241 * rte_pci_probe() parses the devices on the PCI bus and initializes recognized
244 The next step is to configure the RX and TX queues. For each port, there is only
245 one RX queue (only one lcore is able to poll a given port). The number of TX
246 queues depends on the number of available lcores. The rte_eth_dev_configure()
247 function is used to configure the number of queues for a port:
251 ret = rte_eth_dev_configure((uint8_t)portid, 1, 1, &port_conf);
253 rte_panic("Cannot configure device: err=%d, port=%u\n",
256 .. _l2_fwd_event_app_rx_init:
258 RX Queue Initialization
259 ~~~~~~~~~~~~~~~~~~~~~~~
261 The application uses one lcore to poll one or several ports, depending on the -q
262 option, which specifies the number of queues per lcore.
264 For example, if the user specifies -q 4, the application is able to poll four
265 ports with one lcore. If there are 16 ports on the target (and if the portmask
266 argument is -p ffff ), the application will need four lcores to poll all the
271 ret = rte_eth_rx_queue_setup((uint8_t) portid, 0, nb_rxd, SOCKET0,
272 &rx_conf, l2fwd_pktmbuf_pool);
275 rte_panic("rte_eth_rx_queue_setup: err=%d, port=%u\n",
278 The list of queues that must be polled for a given lcore is stored in a private
279 structure called struct lcore_queue_conf.
283 struct lcore_queue_conf {
285 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
286 struct mbuf_table tx_mbufs[L2FWD_MAX_PORTS];
289 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
291 The values n_rx_port and rx_port_list[] are used in the main packet processing
292 loop (see :ref:`l2_fwd_event_app_rx_tx_packets`).
294 .. _l2_fwd_event_app_tx_init:
296 TX Queue Initialization
297 ~~~~~~~~~~~~~~~~~~~~~~~
299 Each lcore should be able to transmit on any port. For every port, a single TX
300 queue is initialized.
304 /* init one TX queue on each port */
308 ret = rte_eth_tx_queue_setup((uint8_t) portid, 0, nb_txd,
309 rte_eth_dev_socket_id(portid), &tx_conf);
311 rte_panic("rte_eth_tx_queue_setup:err=%d, port=%u\n",
312 ret, (unsigned) portid);
314 To configure eventdev support, application setups following components:
322 .. _l2_fwd_event_app_event_dev_init:
324 Event device Initialization
325 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
326 Application can use either H/W or S/W based event device scheduler
327 implementation and supports single instance of event device. It configures event
328 device as per below configuration
332 struct rte_event_dev_config event_d_conf = {
333 .nb_event_queues = ethdev_count, /* Dedicated to each Ethernet port */
334 .nb_event_ports = num_workers, /* Dedicated to each lcore */
335 .nb_events_limit = 4096,
336 .nb_event_queue_flows = 1024,
337 .nb_event_port_dequeue_depth = 128,
338 .nb_event_port_enqueue_depth = 128
341 ret = rte_event_dev_configure(event_d_id, &event_d_conf);
343 rte_panic("Error in configuring event device\n");
345 In case of S/W scheduler, application runs eventdev scheduler service on service
346 core. Application retrieves service id and finds the best possible service core to
351 rte_event_dev_info_get(evt_rsrc->event_d_id, &evdev_info);
352 if (evdev_info.event_dev_cap & RTE_EVENT_DEV_CAP_DISTRIBUTED_SCHED) {
353 ret = rte_event_dev_service_id_get(evt_rsrc->event_d_id,
355 if (ret != -ESRCH && ret != 0)
356 rte_panic("Error in starting eventdev service\n");
357 l2fwd_event_service_enable(service_id);
360 .. _l2_fwd_app_event_queue_init:
362 Event queue Initialization
363 ~~~~~~~~~~~~~~~~~~~~~~~~~~
364 Each Ethernet device is assigned a dedicated event queue which will be linked
365 to all available event ports i.e. each lcore can dequeue packets from any of the
370 struct rte_event_queue_conf event_q_conf = {
371 .nb_atomic_flows = 1024,
372 .nb_atomic_order_sequences = 1024,
373 .event_queue_cfg = 0,
374 .schedule_type = RTE_SCHED_TYPE_ATOMIC,
375 .priority = RTE_EVENT_DEV_PRIORITY_HIGHEST
378 /* User requested sched mode */
379 event_q_conf.schedule_type = eventq_sched_mode;
380 for (event_q_id = 0; event_q_id < ethdev_count; event_q_id++) {
381 ret = rte_event_queue_setup(event_d_id, event_q_id,
384 rte_panic("Error in configuring event queue\n");
387 In case of S/W scheduler, an extra event queue is created which will be used for
388 Tx adapter service function for enqueue operation.
390 .. _l2_fwd_app_event_port_init:
392 Event port Initialization
393 ~~~~~~~~~~~~~~~~~~~~~~~~~
394 Each worker thread is assigned a dedicated event port for enq/deq operations
395 to/from an event device. All event ports are linked with all available event
400 struct rte_event_port_conf event_p_conf = {
403 .new_event_threshold = 4096
406 for (event_p_id = 0; event_p_id < num_workers; event_p_id++) {
407 ret = rte_event_port_setup(event_d_id, event_p_id,
410 rte_panic("Error in configuring event port %d\n", event_p_id);
412 ret = rte_event_port_link(event_d_id, event_p_id, NULL,
415 rte_panic("Error in linking event port %d to queue\n",
419 In case of S/W scheduler, an extra event port is created by DPDK library which
420 is retrieved by the application and same will be used by Tx adapter service.
424 ret = rte_event_eth_tx_adapter_event_port_get(tx_adptr_id, &tx_port_id);
426 rte_panic("Failed to get Tx adapter port id: %d\n", ret);
428 ret = rte_event_port_link(event_d_id, tx_port_id,
429 &evt_rsrc.evq.event_q_id[
430 evt_rsrc.evq.nb_queues - 1],
433 rte_panic("Unable to link Tx adapter port to Tx queue:err=%d\n",
436 .. _l2_fwd_event_app_adapter_init:
438 Rx/Tx adapter Initialization
439 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
440 Each Ethernet port is assigned a dedicated Rx/Tx adapter for H/W scheduler. Each
441 Ethernet port's Rx queues are connected to its respective event queue at
442 priority 0 via Rx adapter configuration and Ethernet port's tx queues are
443 connected via Tx adapter.
447 RTE_ETH_FOREACH_DEV(port_id) {
448 if ((rsrc->enabled_port_mask & (1 << port_id)) == 0)
450 ret = rte_event_eth_rx_adapter_create(adapter_id, event_d_id,
451 &evt_rsrc->def_p_conf);
453 rte_panic("Failed to create rx adapter[%d]\n",
456 /* Configure user requested sched type*/
457 eth_q_conf.ev.sched_type = rsrc->sched_type;
458 eth_q_conf.ev.queue_id = evt_rsrc->evq.event_q_id[q_id];
459 ret = rte_event_eth_rx_adapter_queue_add(adapter_id, port_id,
462 rte_panic("Failed to add queues to Rx adapter\n");
464 ret = rte_event_eth_rx_adapter_start(adapter_id);
466 rte_panic("Rx adapter[%d] start Failed\n", adapter_id);
468 evt_rsrc->rx_adptr.rx_adptr[adapter_id] = adapter_id;
470 if (q_id < evt_rsrc->evq.nb_queues)
475 RTE_ETH_FOREACH_DEV(port_id) {
476 if ((rsrc->enabled_port_mask & (1 << port_id)) == 0)
478 ret = rte_event_eth_tx_adapter_create(adapter_id, event_d_id,
479 &evt_rsrc->def_p_conf);
481 rte_panic("Failed to create tx adapter[%d]\n",
484 ret = rte_event_eth_tx_adapter_queue_add(adapter_id, port_id,
487 rte_panic("Failed to add queues to Tx adapter\n");
489 ret = rte_event_eth_tx_adapter_start(adapter_id);
491 rte_panic("Tx adapter[%d] start Failed\n", adapter_id);
493 evt_rsrc->tx_adptr.tx_adptr[adapter_id] = adapter_id;
497 For S/W scheduler instead of dedicated adapters, common Rx/Tx adapters are
498 configured which will be shared among all the Ethernet ports. Also DPDK library
499 need service cores to run internal services for Rx/Tx adapters. Application gets
500 service id for Rx/Tx adapters and after successful setup it runs the services
501 on dedicated service cores.
505 for (i = 0; i < evt_rsrc->rx_adptr.nb_rx_adptr; i++) {
506 ret = rte_event_eth_rx_adapter_caps_get(evt_rsrc->event_d_id,
507 evt_rsrc->rx_adptr.rx_adptr[i], &caps);
509 rte_panic("Failed to get Rx adapter[%d] caps\n",
510 evt_rsrc->rx_adptr.rx_adptr[i]);
511 ret = rte_event_eth_rx_adapter_service_id_get(
512 evt_rsrc->event_d_id,
514 if (ret != -ESRCH && ret != 0)
515 rte_panic("Error in starting Rx adapter[%d] service\n",
516 evt_rsrc->rx_adptr.rx_adptr[i]);
517 l2fwd_event_service_enable(service_id);
520 for (i = 0; i < evt_rsrc->tx_adptr.nb_tx_adptr; i++) {
521 ret = rte_event_eth_tx_adapter_caps_get(evt_rsrc->event_d_id,
522 evt_rsrc->tx_adptr.tx_adptr[i], &caps);
524 rte_panic("Failed to get Rx adapter[%d] caps\n",
525 evt_rsrc->tx_adptr.tx_adptr[i]);
526 ret = rte_event_eth_tx_adapter_service_id_get(
527 evt_rsrc->event_d_id,
529 if (ret != -ESRCH && ret != 0)
530 rte_panic("Error in starting Rx adapter[%d] service\n",
531 evt_rsrc->tx_adptr.tx_adptr[i]);
532 l2fwd_event_service_enable(service_id);
535 .. _l2_fwd_event_app_rx_tx_packets:
537 Receive, Process and Transmit Packets
538 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
540 In the **l2fwd_main_loop()** function, the main task is to read ingress packets from
541 the RX queues. This is done using the following code:
546 * Read packet from RX queues
549 for (i = 0; i < qconf->n_rx_port; i++) {
550 portid = qconf->rx_port_list[i];
551 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0, pkts_burst,
554 for (j = 0; j < nb_rx; j++) {
556 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
557 l2fwd_simple_forward(m, portid);
561 Packets are read in a burst of size MAX_PKT_BURST. The rte_eth_rx_burst()
562 function writes the mbuf pointers in a local table and returns the number of
563 available mbufs in the table.
565 Then, each mbuf in the table is processed by the l2fwd_simple_forward()
566 function. The processing is very simple: process the TX port from the RX port,
567 then replace the source and destination MAC addresses if MAC addresses updating
570 During the initialization process, a static array of destination ports
571 (l2fwd_dst_ports[]) is filled such that for each source port, a destination port
572 is assigned that is either the next or previous enabled port from the portmask.
573 If number of ports are odd in portmask then packet from last port will be
574 forwarded to first port i.e. if portmask=0x07, then forwarding will take place
575 like p0--->p1, p1--->p2, p2--->p0.
577 Also to optimize enqueue operation, l2fwd_simple_forward() stores incoming mbufs
578 up to MAX_PKT_BURST. Once it reaches up to limit, all packets are transmitted to
584 l2fwd_simple_forward(struct rte_mbuf *m, uint32_t portid)
588 struct rte_eth_dev_tx_buffer *buffer;
590 dst_port = l2fwd_dst_ports[portid];
593 l2fwd_mac_updating(m, dst_port);
595 buffer = tx_buffer[dst_port];
596 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
598 port_statistics[dst_port].tx += sent;
601 For this test application, the processing is exactly the same for all packets
602 arriving on the same RX port. Therefore, it would have been possible to call
603 the rte_eth_tx_buffer() function directly from the main loop to send all the
604 received packets on the same TX port, using the burst-oriented send function,
605 which is more efficient.
607 However, in real-life applications (such as, L3 routing),
608 packet N is not necessarily forwarded on the same port as packet N-1.
609 The application is implemented to illustrate that, so the same approach can be
610 reused in a more complex application.
612 To ensure that no packets remain in the tables, each lcore does a draining of TX
613 queue in its main loop. This technique introduces some latency when there are
614 not many packets to send, however it improves performance:
618 cur_tsc = rte_rdtsc();
621 * TX burst queue drain
623 diff_tsc = cur_tsc - prev_tsc;
624 if (unlikely(diff_tsc > drain_tsc)) {
625 for (i = 0; i < qconf->n_rx_port; i++) {
626 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
627 buffer = tx_buffer[portid];
628 sent = rte_eth_tx_buffer_flush(portid, 0,
631 port_statistics[portid].tx += sent;
634 /* if timer is enabled */
635 if (timer_period > 0) {
636 /* advance the timer */
637 timer_tsc += diff_tsc;
639 /* if timer has reached its timeout */
640 if (unlikely(timer_tsc >= timer_period)) {
641 /* do this only on master core */
642 if (lcore_id == rte_get_master_lcore()) {
644 /* reset the timer */
653 In the **l2fwd_event_loop()** function, the main task is to read ingress
654 packets from the event ports. This is done using the following code:
658 /* Read packet from eventdev */
659 nb_rx = rte_event_dequeue_burst(event_d_id, event_p_id,
666 for (i = 0; i < nb_rx; i++) {
667 mbuf[i] = events[i].mbuf;
668 rte_prefetch0(rte_pktmbuf_mtod(mbuf[i], void *));
672 Before reading packets, deq_len is fetched to ensure correct allowed deq length
674 The rte_event_dequeue_burst() function writes the mbuf pointers in a local table
675 and returns the number of available mbufs in the table.
677 Then, each mbuf in the table is processed by the l2fwd_eventdev_forward()
678 function. The processing is very simple: process the TX port from the RX port,
679 then replace the source and destination MAC addresses if MAC addresses updating
682 During the initialization process, a static array of destination ports
683 (l2fwd_dst_ports[]) is filled such that for each source port, a destination port
684 is assigned that is either the next or previous enabled port from the portmask.
685 If number of ports are odd in portmask then packet from last port will be
686 forwarded to first port i.e. if portmask=0x07, then forwarding will take place
687 like p0--->p1, p1--->p2, p2--->p0.
689 l2fwd_eventdev_forward() does not stores incoming mbufs. Packet will forwarded
690 be to destination ports via Tx adapter or generic event dev enqueue API
691 depending H/W or S/W scheduler is used.
695 nb_tx = rte_event_eth_tx_adapter_enqueue(event_d_id, port_id, ev,
697 while (nb_tx < nb_rx && !rsrc->force_quit)
698 nb_tx += rte_event_eth_tx_adapter_enqueue(
700 ev + nb_tx, nb_rx - nb_tx);