1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2010-2014 Intel Corporation.
4 .. _l2_fwd_app_real_and_virtual:
6 L2 Forwarding Sample Application (in Real and Virtualized Environments)
7 =======================================================================
9 The L2 Forwarding sample application is a simple example of packet processing using
10 the Data Plane Development Kit (DPDK) which
11 also takes advantage of Single Root I/O Virtualization (SR-IOV) features in a virtualized environment.
15 Please note that previously a separate L2 Forwarding in Virtualized Environments sample application was used,
16 however, in later DPDK versions these sample applications have been merged.
21 The L2 Forwarding sample application, which can operate in real and virtualized environments,
22 performs L2 forwarding for each packet that is received on an RX_PORT.
23 The destination port is the adjacent port from the enabled portmask, that is,
24 if the first four ports are enabled (portmask 0xf),
25 ports 1 and 2 forward into each other, and ports 3 and 4 forward into each other.
26 Also, if MAC addresses updating is enabled, the MAC addresses are affected as follows:
28 * The source MAC address is replaced by the TX_PORT MAC address
30 * The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID
32 This application can be used to benchmark performance using a traffic-generator, as shown in the :numref:`figure_l2_fwd_benchmark_setup`,
33 or in a virtualized environment as shown in :numref:`figure_l2_fwd_virtenv_benchmark_setup`.
35 .. _figure_l2_fwd_benchmark_setup:
37 .. figure:: img/l2_fwd_benchmark_setup.*
39 Performance Benchmark Setup (Basic Environment)
41 .. _figure_l2_fwd_virtenv_benchmark_setup:
43 .. figure:: img/l2_fwd_virtenv_benchmark_setup.*
45 Performance Benchmark Setup (Virtualized Environment)
47 This application may be used for basic VM to VM communication as shown in :numref:`figure_l2_fwd_vm2vm`,
48 when MAC addresses updating is disabled.
50 .. _figure_l2_fwd_vm2vm:
52 .. figure:: img/l2_fwd_vm2vm.*
54 Virtual Machine to Virtual Machine communication.
56 The L2 Forwarding application can also be used as a starting point for developing a new application based on the DPDK.
60 Virtual Function Setup Instructions
61 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
63 This application can use the virtual function available in the system and
64 therefore can be used in a virtual machine without passing through
65 the whole Network Device into a guest machine in a virtualized scenario.
66 The virtual functions can be enabled in the host machine or the hypervisor with the respective physical function driver.
68 For example, in a Linux* host machine, it is possible to enable a virtual function using the following command:
70 .. code-block:: console
72 modprobe ixgbe max_vfs=2,2
74 This command enables two Virtual Functions on each of Physical Function of the NIC,
75 with two physical ports in the PCI configuration space.
76 It is important to note that enabled Virtual Function 0 and 2 would belong to Physical Function 0
77 and Virtual Function 1 and 3 would belong to Physical Function 1,
78 in this case enabling a total of four Virtual Functions.
80 Compiling the Application
81 -------------------------
83 To compile the sample application see :doc:`compiling`.
85 The application is located in the ``l2fwd`` sub-directory.
87 Running the Application
88 -----------------------
90 The application requires a number of command line options:
92 .. code-block:: console
94 ./build/l2fwd [EAL options] -- -p PORTMASK [-q NQ] --[no-]mac-updating
98 * p PORTMASK: A hexadecimal bitmask of the ports to configure
100 * q NQ: A number of queues (=ports) per lcore (default is 1)
102 * --[no-]mac-updating: Enable or disable MAC addresses updating (enabled by default).
104 To run the application in linux environment with 4 lcores, 16 ports and 8 RX queues per lcore and MAC address
105 updating enabled, issue the command:
107 .. code-block:: console
109 $ ./build/l2fwd -l 0-3 -n 4 -- -q 8 -p ffff
111 Refer to the *DPDK Getting Started Guide* for general information on running applications
112 and the Environment Abstraction Layer (EAL) options.
117 The following sections provide some explanation of the code.
119 .. _l2_fwd_app_cmd_arguments:
121 Command Line Arguments
122 ~~~~~~~~~~~~~~~~~~~~~~
124 The L2 Forwarding sample application takes specific parameters,
125 in addition to Environment Abstraction Layer (EAL) arguments.
126 The preferred way to parse parameters is to use the getopt() function,
127 since it is part of a well-defined and portable library.
129 The parsing of arguments is done in the l2fwd_parse_args() function.
130 The method of argument parsing is not described here.
131 Refer to the *glibc getopt(3)* man page for details.
133 EAL arguments are parsed first, then application-specific arguments.
134 This is done at the beginning of the main() function:
140 ret = rte_eal_init(argc, argv);
142 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
147 /* parse application arguments (after the EAL ones) */
149 ret = l2fwd_parse_args(argc, argv);
151 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
153 .. _l2_fwd_app_mbuf_init:
155 Mbuf Pool Initialization
156 ~~~~~~~~~~~~~~~~~~~~~~~~
158 Once the arguments are parsed, the mbuf pool is created.
159 The mbuf pool contains a set of mbuf objects that will be used by the driver
160 and the application to store network packet data:
164 /* create the mbuf pool */
166 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
167 MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
170 if (l2fwd_pktmbuf_pool == NULL)
171 rte_panic("Cannot init mbuf pool\n");
173 The rte_mempool is a generic structure used to handle pools of objects.
174 In this case, it is necessary to create a pool that will be used by the driver.
175 The number of allocated pkt mbufs is NB_MBUF, with a data room size of
176 RTE_MBUF_DEFAULT_BUF_SIZE each.
177 A per-lcore cache of 32 mbufs is kept.
178 The memory is allocated in NUMA socket 0,
179 but it is possible to extend this code to allocate one mbuf pool per socket.
181 The rte_pktmbuf_pool_create() function uses the default mbuf pool and mbuf
182 initializers, respectively rte_pktmbuf_pool_init() and rte_pktmbuf_init().
183 An advanced application may want to use the mempool API to create the
184 mbuf pool with more control.
186 .. _l2_fwd_app_dvr_init:
188 Driver Initialization
189 ~~~~~~~~~~~~~~~~~~~~~
191 The main part of the code in the main() function relates to the initialization of the driver.
192 To fully understand this code, it is recommended to study the chapters that related to the Poll Mode Driver
193 in the *DPDK Programmer's Guide* - Rel 1.4 EAR and the *DPDK API Reference*.
197 if (rte_pci_probe() < 0)
198 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
200 /* reset l2fwd_dst_ports */
202 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
203 l2fwd_dst_ports[portid] = 0;
208 * Each logical core is assigned a dedicated TX queue on each port.
211 RTE_ETH_FOREACH_DEV(portid) {
212 /* skip ports that are not enabled */
214 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
217 if (nb_ports_in_mask % 2) {
218 l2fwd_dst_ports[portid] = last_port;
219 l2fwd_dst_ports[last_port] = portid;
226 rte_eth_dev_info_get((uint8_t) portid, &dev_info);
231 * rte_igb_pmd_init_all() simultaneously registers the driver as a PCI driver and as an Ethernet* Poll Mode Driver.
233 * rte_pci_probe() parses the devices on the PCI bus and initializes recognized devices.
235 The next step is to configure the RX and TX queues.
236 For each port, there is only one RX queue (only one lcore is able to poll a given port).
237 The number of TX queues depends on the number of available lcores.
238 The rte_eth_dev_configure() function is used to configure the number of queues for a port:
242 ret = rte_eth_dev_configure((uint8_t)portid, 1, 1, &port_conf);
244 rte_exit(EXIT_FAILURE, "Cannot configure device: "
248 .. _l2_fwd_app_rx_init:
250 RX Queue Initialization
251 ~~~~~~~~~~~~~~~~~~~~~~~
253 The application uses one lcore to poll one or several ports, depending on the -q option,
254 which specifies the number of queues per lcore.
256 For example, if the user specifies -q 4, the application is able to poll four ports with one lcore.
257 If there are 16 ports on the target (and if the portmask argument is -p ffff ),
258 the application will need four lcores to poll all the ports.
262 ret = rte_eth_rx_queue_setup((uint8_t) portid, 0, nb_rxd, SOCKET0, &rx_conf, l2fwd_pktmbuf_pool);
265 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: "
269 The list of queues that must be polled for a given lcore is stored in a private structure called struct lcore_queue_conf.
273 struct lcore_queue_conf {
275 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
276 struct mbuf_table tx_mbufs[L2FWD_MAX_PORTS];
279 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
281 The values n_rx_port and rx_port_list[] are used in the main packet processing loop
282 (see :ref:`l2_fwd_app_rx_tx_packets`).
284 .. _l2_fwd_app_tx_init:
286 TX Queue Initialization
287 ~~~~~~~~~~~~~~~~~~~~~~~
289 Each lcore should be able to transmit on any port. For every port, a single TX queue is initialized.
293 /* init one TX queue on each port */
297 ret = rte_eth_tx_queue_setup((uint8_t) portid, 0, nb_txd, rte_eth_dev_socket_id(portid), &tx_conf);
299 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n", ret, (unsigned) portid);
301 The global configuration for TX queues is stored in a static structure:
305 static const struct rte_eth_txconf tx_conf = {
307 .pthresh = TX_PTHRESH,
308 .hthresh = TX_HTHRESH,
309 .wthresh = TX_WTHRESH,
311 .tx_free_thresh = RTE_TEST_TX_DESC_DEFAULT + 1, /* disable feature */
314 .. _l2_fwd_app_rx_tx_packets:
316 Receive, Process and Transmit Packets
317 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
319 In the l2fwd_main_loop() function, the main task is to read ingress packets from the RX queues.
320 This is done using the following code:
325 * Read packet from RX queues
328 for (i = 0; i < qconf->n_rx_port; i++) {
329 portid = qconf->rx_port_list[i];
330 nb_rx = rte_eth_rx_burst((uint8_t) portid, 0, pkts_burst, MAX_PKT_BURST);
332 for (j = 0; j < nb_rx; j++) {
334 rte_prefetch0[rte_pktmbuf_mtod(m, void *)); l2fwd_simple_forward(m, portid);
338 Packets are read in a burst of size MAX_PKT_BURST.
339 The rte_eth_rx_burst() function writes the mbuf pointers in a local table and returns the number of available mbufs in the table.
341 Then, each mbuf in the table is processed by the l2fwd_simple_forward() function.
342 The processing is very simple: process the TX port from the RX port, then replace the source and destination MAC addresses if MAC
343 addresses updating is enabled.
347 In the following code, one line for getting the output port requires some explanation.
349 During the initialization process, a static array of destination ports (l2fwd_dst_ports[]) is filled such that for each source port,
350 a destination port is assigned that is either the next or previous enabled port from the portmask.
351 Naturally, the number of ports in the portmask must be even, otherwise, the application exits.
356 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
358 struct rte_ether_hdr *eth;
362 dst_port = l2fwd_dst_ports[portid];
364 eth = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
366 /* 02:00:00:00:00:xx */
368 tmp = ð->d_addr.addr_bytes[0];
370 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t) dst_port << 40);
374 rte_ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
376 l2fwd_send_packet(m, (uint8_t) dst_port);
379 Then, the packet is sent using the l2fwd_send_packet (m, dst_port) function.
380 For this test application, the processing is exactly the same for all packets arriving on the same RX port.
381 Therefore, it would have been possible to call the l2fwd_send_burst() function directly from the main loop
382 to send all the received packets on the same TX port,
383 using the burst-oriented send function, which is more efficient.
385 However, in real-life applications (such as, L3 routing),
386 packet N is not necessarily forwarded on the same port as packet N-1.
387 The application is implemented to illustrate that, so the same approach can be reused in a more complex application.
389 The l2fwd_send_packet() function stores the packet in a per-lcore and per-txport table.
390 If the table is full, the whole packets table is transmitted using the l2fwd_send_burst() function:
394 /* Send the packet on an output interface */
397 l2fwd_send_packet(struct rte_mbuf *m, uint16_t port)
399 unsigned lcore_id, len;
400 struct lcore_queue_conf *qconf;
402 lcore_id = rte_lcore_id();
403 qconf = &lcore_queue_conf[lcore_id];
404 len = qconf->tx_mbufs[port].len;
405 qconf->tx_mbufs[port].m_table[len] = m;
408 /* enough pkts to be sent */
410 if (unlikely(len == MAX_PKT_BURST)) {
411 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
415 qconf->tx_mbufs[port].len = len; return 0;
418 To ensure that no packets remain in the tables, each lcore does a draining of TX queue in its main loop.
419 This technique introduces some latency when there are not many packets to send,
420 however it improves performance:
424 cur_tsc = rte_rdtsc();
427 * TX burst queue drain
430 diff_tsc = cur_tsc - prev_tsc;
432 if (unlikely(diff_tsc > drain_tsc)) {
433 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
434 if (qconf->tx_mbufs[portid].len == 0)
437 l2fwd_send_burst(&lcore_queue_conf[lcore_id], qconf->tx_mbufs[portid].len, (uint8_t) portid);
439 qconf->tx_mbufs[portid].len = 0;
442 /* if timer is enabled */
444 if (timer_period > 0) {
445 /* advance the timer */
447 timer_tsc += diff_tsc;
449 /* if timer has reached its timeout */
451 if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
452 /* do this only on master core */
454 if (lcore_id == rte_get_master_lcore()) {
457 /* reset the timer */