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 Sample usage commands are given below to run the application into different mode:
71 Poll mode with 4 lcores, 16 ports and 8 RX queues per lcore and MAC address updating enabled,
74 .. code-block:: console
76 ./build/l2fwd-event -l 0-3 -n 4 -- -q 8 -p ffff --mode=poll
78 Eventdev mode with 4 lcores, 16 ports , sched method ordered and MAC address updating enabled,
81 .. code-block:: console
83 ./build/l2fwd-event -l 0-3 -n 4 -- -p ffff --eventq-sched=ordered
87 .. code-block:: console
89 ./build/l2fwd-event -l 0-3 -n 4 -- -q 8 -p ffff --mode=eventdev --eventq-sched=ordered
91 Refer to the *DPDK Getting Started Guide* for general information on running
92 applications and the Environment Abstraction Layer (EAL) options.
94 To run application with S/W scheduler, it uses following DPDK services:
97 * Rx adapter service function
98 * Tx adapter service function
100 Application needs service cores to run above mentioned services. Service cores
101 must be provided as EAL parameters along with the --vdev=event_sw0 to enable S/W
102 scheduler. Following is the sample command:
104 .. code-block:: console
106 ./build/l2fwd-event -l 0-7 -s 0-3 -n 4 --vdev event_sw0 -- -q 8 -p ffff --mode=eventdev --eventq-sched=ordered
111 The following sections provide some explanation of the code.
113 .. _l2_fwd_event_app_cmd_arguments:
115 Command Line Arguments
116 ~~~~~~~~~~~~~~~~~~~~~~
118 The L2 Forwarding eventdev sample application takes specific parameters,
119 in addition to Environment Abstraction Layer (EAL) arguments.
120 The preferred way to parse parameters is to use the getopt() function,
121 since it is part of a well-defined and portable library.
123 The parsing of arguments is done in the **l2fwd_parse_args()** function for non
124 eventdev parameters and in **parse_eventdev_args()** for eventdev parameters.
125 The method of argument parsing is not described here. Refer to the
126 *glibc getopt(3)* man page for details.
128 EAL arguments are parsed first, then application-specific arguments.
129 This is done at the beginning of the main() function and eventdev parameters
130 are parsed in eventdev_resource_setup() function during eventdev setup:
136 ret = rte_eal_init(argc, argv);
138 rte_panic("Invalid EAL arguments\n");
143 /* parse application arguments (after the EAL ones) */
145 ret = l2fwd_parse_args(argc, argv);
147 rte_panic("Invalid L2FWD arguments\n");
152 /* Parse eventdev command line options */
153 ret = parse_eventdev_args(argc, argv);
160 .. _l2_fwd_event_app_mbuf_init:
162 Mbuf Pool Initialization
163 ~~~~~~~~~~~~~~~~~~~~~~~~
165 Once the arguments are parsed, the mbuf pool is created.
166 The mbuf pool contains a set of mbuf objects that will be used by the driver
167 and the application to store network packet data:
171 /* create the mbuf pool */
173 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
174 MEMPOOL_CACHE_SIZE, 0,
175 RTE_MBUF_DEFAULT_BUF_SIZE,
177 if (l2fwd_pktmbuf_pool == NULL)
178 rte_panic("Cannot init mbuf pool\n");
180 The rte_mempool is a generic structure used to handle pools of objects.
181 In this case, it is necessary to create a pool that will be used by the driver.
182 The number of allocated pkt mbufs is NB_MBUF, with a data room size of
183 RTE_MBUF_DEFAULT_BUF_SIZE each.
184 A per-lcore cache of 32 mbufs is kept.
185 The memory is allocated in NUMA socket 0,
186 but it is possible to extend this code to allocate one mbuf pool per socket.
188 The rte_pktmbuf_pool_create() function uses the default mbuf pool and mbuf
189 initializers, respectively rte_pktmbuf_pool_init() and rte_pktmbuf_init().
190 An advanced application may want to use the mempool API to create the
191 mbuf pool with more control.
193 .. _l2_fwd_event_app_drv_init:
195 Driver Initialization
196 ~~~~~~~~~~~~~~~~~~~~~
198 The main part of the code in the main() function relates to the initialization
199 of the driver. To fully understand this code, it is recommended to study the
200 chapters that related to the Poll Mode and Event mode Driver in the
201 *DPDK Programmer's Guide* - Rel 1.4 EAR and the *DPDK API Reference*.
205 if (rte_pci_probe() < 0)
206 rte_panic("Cannot probe PCI\n");
208 /* reset l2fwd_dst_ports */
210 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
211 l2fwd_dst_ports[portid] = 0;
216 * Each logical core is assigned a dedicated TX queue on each port.
219 RTE_ETH_FOREACH_DEV(portid) {
220 /* skip ports that are not enabled */
222 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
225 if (nb_ports_in_mask % 2) {
226 l2fwd_dst_ports[portid] = last_port;
227 l2fwd_dst_ports[last_port] = portid;
234 rte_eth_dev_info_get((uint8_t) portid, &dev_info);
239 * rte_pci_probe() parses the devices on the PCI bus and initializes recognized
242 The next step is to configure the RX and TX queues. For each port, there is only
243 one RX queue (only one lcore is able to poll a given port). The number of TX
244 queues depends on the number of available lcores. The rte_eth_dev_configure()
245 function is used to configure the number of queues for a port:
249 ret = rte_eth_dev_configure((uint8_t)portid, 1, 1, &port_conf);
251 rte_panic("Cannot configure device: err=%d, port=%u\n",
254 .. _l2_fwd_event_app_rx_init:
256 RX Queue Initialization
257 ~~~~~~~~~~~~~~~~~~~~~~~
259 The application uses one lcore to poll one or several ports, depending on the -q
260 option, which specifies the number of queues per lcore.
262 For example, if the user specifies -q 4, the application is able to poll four
263 ports with one lcore. If there are 16 ports on the target (and if the portmask
264 argument is -p ffff ), the application will need four lcores to poll all the
269 ret = rte_eth_rx_queue_setup((uint8_t) portid, 0, nb_rxd, SOCKET0,
270 &rx_conf, l2fwd_pktmbuf_pool);
273 rte_panic("rte_eth_rx_queue_setup: err=%d, port=%u\n",
276 The list of queues that must be polled for a given lcore is stored in a private
277 structure called struct lcore_queue_conf.
281 struct lcore_queue_conf {
283 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
284 struct mbuf_table tx_mbufs[L2FWD_MAX_PORTS];
287 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
289 The values n_rx_port and rx_port_list[] are used in the main packet processing
290 loop (see :ref:`l2_fwd_event_app_rx_tx_packets`).
292 .. _l2_fwd_event_app_tx_init:
294 TX Queue Initialization
295 ~~~~~~~~~~~~~~~~~~~~~~~
297 Each lcore should be able to transmit on any port. For every port, a single TX
298 queue is initialized.
302 /* init one TX queue on each port */
306 ret = rte_eth_tx_queue_setup((uint8_t) portid, 0, nb_txd,
307 rte_eth_dev_socket_id(portid), &tx_conf);
309 rte_panic("rte_eth_tx_queue_setup:err=%d, port=%u\n",
310 ret, (unsigned) portid);
312 To configure eventdev support, application setups following components:
320 .. _l2_fwd_event_app_event_dev_init:
322 Event device Initialization
323 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
324 Application can use either H/W or S/W based event device scheduler
325 implementation and supports single instance of event device. It configures event
326 device as per below configuration
330 struct rte_event_dev_config event_d_conf = {
331 .nb_event_queues = ethdev_count, /* Dedicated to each Ethernet port */
332 .nb_event_ports = num_workers, /* Dedicated to each lcore */
333 .nb_events_limit = 4096,
334 .nb_event_queue_flows = 1024,
335 .nb_event_port_dequeue_depth = 128,
336 .nb_event_port_enqueue_depth = 128
339 ret = rte_event_dev_configure(event_d_id, &event_d_conf);
341 rte_panic("Error in configuring event device\n");
343 In case of S/W scheduler, application runs eventdev scheduler service on service
344 core. Application retrieves service id and finds the best possible service core to
349 rte_event_dev_info_get(evt_rsrc->event_d_id, &evdev_info);
350 if (evdev_info.event_dev_cap & RTE_EVENT_DEV_CAP_DISTRIBUTED_SCHED) {
351 ret = rte_event_dev_service_id_get(evt_rsrc->event_d_id,
353 if (ret != -ESRCH && ret != 0)
354 rte_panic("Error in starting eventdev service\n");
355 l2fwd_event_service_enable(service_id);
358 .. _l2_fwd_app_event_queue_init:
360 Event queue Initialization
361 ~~~~~~~~~~~~~~~~~~~~~~~~~~
362 Each Ethernet device is assigned a dedicated event queue which will be linked
363 to all available event ports i.e. each lcore can dequeue packets from any of the
368 struct rte_event_queue_conf event_q_conf = {
369 .nb_atomic_flows = 1024,
370 .nb_atomic_order_sequences = 1024,
371 .event_queue_cfg = 0,
372 .schedule_type = RTE_SCHED_TYPE_ATOMIC,
373 .priority = RTE_EVENT_DEV_PRIORITY_HIGHEST
376 /* User requested sched mode */
377 event_q_conf.schedule_type = eventq_sched_mode;
378 for (event_q_id = 0; event_q_id < ethdev_count; event_q_id++) {
379 ret = rte_event_queue_setup(event_d_id, event_q_id,
382 rte_panic("Error in configuring event queue\n");
385 In case of S/W scheduler, an extra event queue is created which will be used for
386 Tx adapter service function for enqueue operation.
388 .. _l2_fwd_app_event_port_init:
390 Event port Initialization
391 ~~~~~~~~~~~~~~~~~~~~~~~~~
392 Each worker thread is assigned a dedicated event port for enq/deq operations
393 to/from an event device. All event ports are linked with all available event
398 struct rte_event_port_conf event_p_conf = {
401 .new_event_threshold = 4096
404 for (event_p_id = 0; event_p_id < num_workers; event_p_id++) {
405 ret = rte_event_port_setup(event_d_id, event_p_id,
408 rte_panic("Error in configuring event port %d\n", event_p_id);
410 ret = rte_event_port_link(event_d_id, event_p_id, NULL,
413 rte_panic("Error in linking event port %d to queue\n",
417 In case of S/W scheduler, an extra event port is created by DPDK library which
418 is retrieved by the application and same will be used by Tx adapter service.
422 ret = rte_event_eth_tx_adapter_event_port_get(tx_adptr_id, &tx_port_id);
424 rte_panic("Failed to get Tx adapter port id: %d\n", ret);
426 ret = rte_event_port_link(event_d_id, tx_port_id,
427 &evt_rsrc.evq.event_q_id[
428 evt_rsrc.evq.nb_queues - 1],
431 rte_panic("Unable to link Tx adapter port to Tx queue:err=%d\n",
434 .. _l2_fwd_event_app_adapter_init:
436 Rx/Tx adapter Initialization
437 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
438 Each Ethernet port is assigned a dedicated Rx/Tx adapter for H/W scheduler. Each
439 Ethernet port's Rx queues are connected to its respective event queue at
440 priority 0 via Rx adapter configuration and Ethernet port's tx queues are
441 connected via Tx adapter.
445 RTE_ETH_FOREACH_DEV(port_id) {
446 if ((rsrc->enabled_port_mask & (1 << port_id)) == 0)
448 ret = rte_event_eth_rx_adapter_create(adapter_id, event_d_id,
449 &evt_rsrc->def_p_conf);
451 rte_panic("Failed to create rx adapter[%d]\n",
454 /* Configure user requested sched type*/
455 eth_q_conf.ev.sched_type = rsrc->sched_type;
456 eth_q_conf.ev.queue_id = evt_rsrc->evq.event_q_id[q_id];
457 ret = rte_event_eth_rx_adapter_queue_add(adapter_id, port_id,
460 rte_panic("Failed to add queues to Rx adapter\n");
462 ret = rte_event_eth_rx_adapter_start(adapter_id);
464 rte_panic("Rx adapter[%d] start Failed\n", adapter_id);
466 evt_rsrc->rx_adptr.rx_adptr[adapter_id] = adapter_id;
468 if (q_id < evt_rsrc->evq.nb_queues)
473 RTE_ETH_FOREACH_DEV(port_id) {
474 if ((rsrc->enabled_port_mask & (1 << port_id)) == 0)
476 ret = rte_event_eth_tx_adapter_create(adapter_id, event_d_id,
477 &evt_rsrc->def_p_conf);
479 rte_panic("Failed to create tx adapter[%d]\n",
482 ret = rte_event_eth_tx_adapter_queue_add(adapter_id, port_id,
485 rte_panic("Failed to add queues to Tx adapter\n");
487 ret = rte_event_eth_tx_adapter_start(adapter_id);
489 rte_panic("Tx adapter[%d] start Failed\n", adapter_id);
491 evt_rsrc->tx_adptr.tx_adptr[adapter_id] = adapter_id;
495 For S/W scheduler instead of dedicated adapters, common Rx/Tx adapters are
496 configured which will be shared among all the Ethernet ports. Also DPDK library
497 need service cores to run internal services for Rx/Tx adapters. Application gets
498 service id for Rx/Tx adapters and after successful setup it runs the services
499 on dedicated service cores.
503 for (i = 0; i < evt_rsrc->rx_adptr.nb_rx_adptr; i++) {
504 ret = rte_event_eth_rx_adapter_caps_get(evt_rsrc->event_d_id,
505 evt_rsrc->rx_adptr.rx_adptr[i], &caps);
507 rte_panic("Failed to get Rx adapter[%d] caps\n",
508 evt_rsrc->rx_adptr.rx_adptr[i]);
509 ret = rte_event_eth_rx_adapter_service_id_get(
510 evt_rsrc->event_d_id,
512 if (ret != -ESRCH && ret != 0)
513 rte_panic("Error in starting Rx adapter[%d] service\n",
514 evt_rsrc->rx_adptr.rx_adptr[i]);
515 l2fwd_event_service_enable(service_id);
518 for (i = 0; i < evt_rsrc->tx_adptr.nb_tx_adptr; i++) {
519 ret = rte_event_eth_tx_adapter_caps_get(evt_rsrc->event_d_id,
520 evt_rsrc->tx_adptr.tx_adptr[i], &caps);
522 rte_panic("Failed to get Rx adapter[%d] caps\n",
523 evt_rsrc->tx_adptr.tx_adptr[i]);
524 ret = rte_event_eth_tx_adapter_service_id_get(
525 evt_rsrc->event_d_id,
527 if (ret != -ESRCH && ret != 0)
528 rte_panic("Error in starting Rx adapter[%d] service\n",
529 evt_rsrc->tx_adptr.tx_adptr[i]);
530 l2fwd_event_service_enable(service_id);
533 .. _l2_fwd_event_app_rx_tx_packets:
535 Receive, Process and Transmit Packets
536 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
538 In the **l2fwd_main_loop()** function, the main task is to read ingress packets from
539 the RX queues. This is done using the following code:
544 * Read packet from RX queues
547 for (i = 0; i < qconf->n_rx_port; i++) {
548 portid = qconf->rx_port_list[i];
549 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0, pkts_burst,
552 for (j = 0; j < nb_rx; j++) {
554 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
555 l2fwd_simple_forward(m, portid);
559 Packets are read in a burst of size MAX_PKT_BURST. The rte_eth_rx_burst()
560 function writes the mbuf pointers in a local table and returns the number of
561 available mbufs in the table.
563 Then, each mbuf in the table is processed by the l2fwd_simple_forward()
564 function. The processing is very simple: process the TX port from the RX port,
565 then replace the source and destination MAC addresses if MAC addresses updating
568 During the initialization process, a static array of destination ports
569 (l2fwd_dst_ports[]) is filled such that for each source port, a destination port
570 is assigned that is either the next or previous enabled port from the portmask.
571 If number of ports are odd in portmask then packet from last port will be
572 forwarded to first port i.e. if portmask=0x07, then forwarding will take place
573 like p0--->p1, p1--->p2, p2--->p0.
575 Also to optimize enqueue operation, l2fwd_simple_forward() stores incoming mbufs
576 up to MAX_PKT_BURST. Once it reaches up to limit, all packets are transmitted to
582 l2fwd_simple_forward(struct rte_mbuf *m, uint32_t portid)
586 struct rte_eth_dev_tx_buffer *buffer;
588 dst_port = l2fwd_dst_ports[portid];
591 l2fwd_mac_updating(m, dst_port);
593 buffer = tx_buffer[dst_port];
594 sent = rte_eth_tx_buffer(dst_port, 0, buffer, m);
596 port_statistics[dst_port].tx += sent;
599 For this test application, the processing is exactly the same for all packets
600 arriving on the same RX port. Therefore, it would have been possible to call
601 the rte_eth_tx_buffer() function directly from the main loop to send all the
602 received packets on the same TX port, using the burst-oriented send function,
603 which is more efficient.
605 However, in real-life applications (such as, L3 routing),
606 packet N is not necessarily forwarded on the same port as packet N-1.
607 The application is implemented to illustrate that, so the same approach can be
608 reused in a more complex application.
610 To ensure that no packets remain in the tables, each lcore does a draining of TX
611 queue in its main loop. This technique introduces some latency when there are
612 not many packets to send, however it improves performance:
616 cur_tsc = rte_rdtsc();
619 * TX burst queue drain
621 diff_tsc = cur_tsc - prev_tsc;
622 if (unlikely(diff_tsc > drain_tsc)) {
623 for (i = 0; i < qconf->n_rx_port; i++) {
624 portid = l2fwd_dst_ports[qconf->rx_port_list[i]];
625 buffer = tx_buffer[portid];
626 sent = rte_eth_tx_buffer_flush(portid, 0,
629 port_statistics[portid].tx += sent;
632 /* if timer is enabled */
633 if (timer_period > 0) {
634 /* advance the timer */
635 timer_tsc += diff_tsc;
637 /* if timer has reached its timeout */
638 if (unlikely(timer_tsc >= timer_period)) {
639 /* do this only on master core */
640 if (lcore_id == rte_get_master_lcore()) {
642 /* reset the timer */
651 In the **l2fwd_event_loop()** function, the main task is to read ingress
652 packets from the event ports. This is done using the following code:
656 /* Read packet from eventdev */
657 nb_rx = rte_event_dequeue_burst(event_d_id, event_p_id,
664 for (i = 0; i < nb_rx; i++) {
665 mbuf[i] = events[i].mbuf;
666 rte_prefetch0(rte_pktmbuf_mtod(mbuf[i], void *));
670 Before reading packets, deq_len is fetched to ensure correct allowed deq length
672 The rte_event_dequeue_burst() function writes the mbuf pointers in a local table
673 and returns the number of available mbufs in the table.
675 Then, each mbuf in the table is processed by the l2fwd_eventdev_forward()
676 function. The processing is very simple: process the TX port from the RX port,
677 then replace the source and destination MAC addresses if MAC addresses updating
680 During the initialization process, a static array of destination ports
681 (l2fwd_dst_ports[]) is filled such that for each source port, a destination port
682 is assigned that is either the next or previous enabled port from the portmask.
683 If number of ports are odd in portmask then packet from last port will be
684 forwarded to first port i.e. if portmask=0x07, then forwarding will take place
685 like p0--->p1, p1--->p2, p2--->p0.
687 l2fwd_eventdev_forward() does not stores incoming mbufs. Packet will forwarded
688 be to destination ports via Tx adapter or generic event dev enqueue API
689 depending H/W or S/W scheduler is used.
693 nb_tx = rte_event_eth_tx_adapter_enqueue(event_d_id, port_id, ev,
695 while (nb_tx < nb_rx && !rsrc->force_quit)
696 nb_tx += rte_event_eth_tx_adapter_enqueue(
698 ev + nb_tx, nb_rx - nb_tx);