2 Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions
9 * Redistributions of source code must retain the above copyright
10 notice, this list of conditions and the following disclaimer.
11 * Redistributions in binary form must reproduce the above copyright
12 notice, this list of conditions and the following disclaimer in
13 the documentation and/or other materials provided with the
15 * Neither the name of Intel Corporation nor the names of its
16 contributors may be used to endorse or promote products derived
17 from this software without specific prior written permission.
19 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31 L2 Forwarding Sample Application (in Real and Virtualized Environments) with core load statistics.
32 ==================================================================================================
34 The L2 Forwarding sample application is a simple example of packet processing using
35 the Data Plane Development Kit (DPDK) which
36 also takes advantage of Single Root I/O Virtualization (SR-IOV) features in a virtualized environment.
40 This application is a variation of L2 Forwarding sample application. It demonstrate possible
41 scheme of job stats library usage therefore some parts of this document is identical with original
42 L2 forwarding application.
47 The L2 Forwarding sample application, which can operate in real and virtualized environments,
48 performs L2 forwarding for each packet that is received.
49 The destination port is the adjacent port from the enabled portmask, that is,
50 if the first four ports are enabled (portmask 0xf),
51 ports 1 and 2 forward into each other, and ports 3 and 4 forward into each other.
52 Also, the MAC addresses are affected as follows:
54 * The source MAC address is replaced by the TX port MAC address
56 * The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID
58 This application can be used to benchmark performance using a traffic-generator, as shown in the :numref:`figure_l2_fwd_benchmark_setup_jobstats`.
60 The application can also be used in a virtualized environment as shown in :numref:`figure_l2_fwd_virtenv_benchmark_setup_jobstats`.
62 The L2 Forwarding application can also be used as a starting point for developing a new application based on the DPDK.
64 .. _figure_l2_fwd_benchmark_setup_jobstats:
66 .. figure:: img/l2_fwd_benchmark_setup.*
68 Performance Benchmark Setup (Basic Environment)
70 .. _figure_l2_fwd_virtenv_benchmark_setup_jobstats:
72 .. figure:: img/l2_fwd_virtenv_benchmark_setup.*
74 Performance Benchmark Setup (Virtualized Environment)
77 Virtual Function Setup Instructions
78 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
80 This application can use the virtual function available in the system and
81 therefore can be used in a virtual machine without passing through
82 the whole Network Device into a guest machine in a virtualized scenario.
83 The virtual functions can be enabled in the host machine or the hypervisor with the respective physical function driver.
85 For example, in a Linux* host machine, it is possible to enable a virtual function using the following command:
87 .. code-block:: console
89 modprobe ixgbe max_vfs=2,2
91 This command enables two Virtual Functions on each of Physical Function of the NIC,
92 with two physical ports in the PCI configuration space.
93 It is important to note that enabled Virtual Function 0 and 2 would belong to Physical Function 0
94 and Virtual Function 1 and 3 would belong to Physical Function 1,
95 in this case enabling a total of four Virtual Functions.
97 Compiling the Application
98 -------------------------
100 To compile the sample application see :doc:`compiling`.
102 The application is located in the ``l2fwd-jobstats`` sub-directory.
104 Running the Application
105 -----------------------
107 The application requires a number of command line options:
109 .. code-block:: console
111 ./build/l2fwd-jobstats [EAL options] -- -p PORTMASK [-q NQ] [-l]
115 * p PORTMASK: A hexadecimal bitmask of the ports to configure
117 * q NQ: A number of queues (=ports) per lcore (default is 1)
119 * l: Use locale thousands separator when formatting big numbers.
121 To run the application in linuxapp environment with 4 lcores, 16 ports, 8 RX queues per lcore and
122 thousands separator printing, issue the command:
124 .. code-block:: console
126 $ ./build/l2fwd-jobstats -l 0-3 -n 4 -- -q 8 -p ffff -l
128 Refer to the *DPDK Getting Started Guide* for general information on running applications
129 and the Environment Abstraction Layer (EAL) options.
134 The following sections provide some explanation of the code.
136 Command Line Arguments
137 ~~~~~~~~~~~~~~~~~~~~~~
139 The L2 Forwarding sample application takes specific parameters,
140 in addition to Environment Abstraction Layer (EAL) arguments
141 (see `Running the Application`_).
142 The preferred way to parse parameters is to use the getopt() function,
143 since it is part of a well-defined and portable library.
145 The parsing of arguments is done in the l2fwd_parse_args() function.
146 The method of argument parsing is not described here.
147 Refer to the *glibc getopt(3)* man page for details.
149 EAL arguments are parsed first, then application-specific arguments.
150 This is done at the beginning of the main() function:
156 ret = rte_eal_init(argc, argv);
158 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
163 /* parse application arguments (after the EAL ones) */
165 ret = l2fwd_parse_args(argc, argv);
167 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
169 Mbuf Pool Initialization
170 ~~~~~~~~~~~~~~~~~~~~~~~~
172 Once the arguments are parsed, the mbuf pool is created.
173 The mbuf pool contains a set of mbuf objects that will be used by the driver
174 and the application to store network packet data:
178 /* create the mbuf pool */
179 l2fwd_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NB_MBUF,
180 MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
183 if (l2fwd_pktmbuf_pool == NULL)
184 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
186 The rte_mempool is a generic structure used to handle pools of objects.
187 In this case, it is necessary to create a pool that will be used by the driver.
188 The number of allocated pkt mbufs is NB_MBUF, with a data room size of
189 RTE_MBUF_DEFAULT_BUF_SIZE each.
190 A per-lcore cache of MEMPOOL_CACHE_SIZE mbufs is kept.
191 The memory is allocated in rte_socket_id() socket,
192 but it is possible to extend this code to allocate one mbuf pool per socket.
194 The rte_pktmbuf_pool_create() function uses the default mbuf pool and mbuf
195 initializers, respectively rte_pktmbuf_pool_init() and rte_pktmbuf_init().
196 An advanced application may want to use the mempool API to create the
197 mbuf pool with more control.
199 Driver Initialization
200 ~~~~~~~~~~~~~~~~~~~~~
202 The main part of the code in the main() function relates to the initialization of the driver.
203 To fully understand this code, it is recommended to study the chapters that related to the Poll Mode Driver
204 in the *DPDK Programmer's Guide* and the *DPDK API Reference*.
208 nb_ports = rte_eth_dev_count();
211 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
213 /* reset l2fwd_dst_ports */
215 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
216 l2fwd_dst_ports[portid] = 0;
221 * Each logical core is assigned a dedicated TX queue on each port.
223 for (portid = 0; portid < nb_ports; portid++) {
224 /* skip ports that are not enabled */
225 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
228 if (nb_ports_in_mask % 2) {
229 l2fwd_dst_ports[portid] = last_port;
230 l2fwd_dst_ports[last_port] = portid;
237 rte_eth_dev_info_get((uint8_t) portid, &dev_info);
240 The next step is to configure the RX and TX queues.
241 For each port, there is only one RX queue (only one lcore is able to poll a given port).
242 The number of TX queues depends on the number of available lcores.
243 The rte_eth_dev_configure() function is used to configure the number of queues for a port:
247 ret = rte_eth_dev_configure((uint8_t)portid, 1, 1, &port_conf);
249 rte_exit(EXIT_FAILURE, "Cannot configure device: "
253 The global configuration is stored in a static structure:
257 static const struct rte_eth_conf port_conf = {
260 .header_split = 0, /**< Header Split disabled */
261 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
262 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
263 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
264 .hw_strip_crc= 0, /**< CRC stripped by hardware */
268 .mq_mode = ETH_DCB_NONE
272 RX Queue Initialization
273 ~~~~~~~~~~~~~~~~~~~~~~~
275 The application uses one lcore to poll one or several ports, depending on the -q option,
276 which specifies the number of queues per lcore.
278 For example, if the user specifies -q 4, the application is able to poll four ports with one lcore.
279 If there are 16 ports on the target (and if the portmask argument is -p ffff ),
280 the application will need four lcores to poll all the ports.
284 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
285 rte_eth_dev_socket_id(portid),
290 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
291 ret, (unsigned) portid);
293 The list of queues that must be polled for a given lcore is stored in a private structure called struct lcore_queue_conf.
297 struct lcore_queue_conf {
299 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
300 truct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
302 struct rte_timer rx_timers[MAX_RX_QUEUE_PER_LCORE];
303 struct rte_jobstats port_fwd_jobs[MAX_RX_QUEUE_PER_LCORE];
305 struct rte_timer flush_timer;
306 struct rte_jobstats flush_job;
307 struct rte_jobstats idle_job;
308 struct rte_jobstats_context jobs_context;
310 rte_atomic16_t stats_read_pending;
312 } __rte_cache_aligned;
314 Values of struct lcore_queue_conf:
316 * n_rx_port and rx_port_list[] are used in the main packet processing loop
317 (see Section `Receive, Process and Transmit Packets`_ later in this chapter).
319 * rx_timers and flush_timer are used to ensure forced TX on low packet rate.
321 * flush_job, idle_job and jobs_context are librte_jobstats objects used for managing l2fwd jobs.
323 * stats_read_pending and lock are used during job stats read phase.
325 TX Queue Initialization
326 ~~~~~~~~~~~~~~~~~~~~~~~
328 Each lcore should be able to transmit on any port. For every port, a single TX queue is initialized.
332 /* init one TX queue on each port */
335 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
336 rte_eth_dev_socket_id(portid),
339 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
340 ret, (unsigned) portid);
342 Jobs statistics initialization
343 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
344 There are several statistics objects available:
346 * Flush job statistics
350 rte_jobstats_init(&qconf->flush_job, "flush", drain_tsc, drain_tsc,
353 rte_timer_init(&qconf->flush_timer);
354 ret = rte_timer_reset(&qconf->flush_timer, drain_tsc, PERIODICAL,
355 lcore_id, &l2fwd_flush_job, NULL);
358 rte_exit(1, "Failed to reset flush job timer for lcore %u: %s",
359 lcore_id, rte_strerror(-ret));
362 * Statistics per RX port
366 rte_jobstats_init(job, name, 0, drain_tsc, 0, MAX_PKT_BURST);
367 rte_jobstats_set_update_period_function(job, l2fwd_job_update_cb);
369 rte_timer_init(&qconf->rx_timers[i]);
370 ret = rte_timer_reset(&qconf->rx_timers[i], 0, PERIODICAL, lcore_id,
371 l2fwd_fwd_job, (void *)(uintptr_t)i);
374 rte_exit(1, "Failed to reset lcore %u port %u job timer: %s",
375 lcore_id, qconf->rx_port_list[i], rte_strerror(-ret));
378 Following parameters are passed to rte_jobstats_init():
380 * 0 as minimal poll period
382 * drain_tsc as maximum poll period
384 * MAX_PKT_BURST as desired target value (RX burst size)
389 The forwarding path is reworked comparing to original L2 Forwarding application.
390 In the l2fwd_main_loop() function three loops are placed.
395 rte_spinlock_lock(&qconf->lock);
398 rte_jobstats_context_start(&qconf->jobs_context);
401 * - Read stats_read_pending flag
402 * - check if some real job need to be executed
404 rte_jobstats_start(&qconf->jobs_context, &qconf->idle_job);
408 uint64_t now = rte_get_timer_cycles();
410 need_manage = qconf->flush_timer.expire < now;
411 /* Check if we was esked to give a stats. */
413 rte_atomic16_read(&qconf->stats_read_pending);
414 need_manage |= stats_read_pending;
416 for (i = 0; i < qconf->n_rx_port && !need_manage; i++)
417 need_manage = qconf->rx_timers[i].expire < now;
419 } while (!need_manage);
420 rte_jobstats_finish(&qconf->idle_job, qconf->idle_job.target);
423 rte_jobstats_context_finish(&qconf->jobs_context);
424 } while (likely(stats_read_pending == 0));
426 rte_spinlock_unlock(&qconf->lock);
430 First infinite for loop is to minimize impact of stats reading. Lock is only locked/unlocked when asked.
432 Second inner while loop do the whole jobs management. When any job is ready, the use rte_timer_manage() is used to call the job handler.
433 In this place functions l2fwd_fwd_job() and l2fwd_flush_job() are called when needed.
434 Then rte_jobstats_context_finish() is called to mark loop end - no other jobs are ready to execute. By this time stats are ready to be read
435 and if stats_read_pending is set, loop breaks allowing stats to be read.
437 Third do-while loop is the idle job (idle stats counter). Its only purpose is monitoring if any job is ready or stats job read is pending
438 for this lcore. Statistics from this part of code is considered as the headroom available for additional processing.
440 Receive, Process and Transmit Packets
441 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
443 The main task of l2fwd_fwd_job() function is to read ingress packets from the RX queue of particular port and forward it.
444 This is done using the following code:
448 total_nb_rx = rte_eth_rx_burst((uint8_t) portid, 0, pkts_burst,
451 for (j = 0; j < total_nb_rx; j++) {
453 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
454 l2fwd_simple_forward(m, portid);
457 Packets are read in a burst of size MAX_PKT_BURST.
458 Then, each mbuf in the table is processed by the l2fwd_simple_forward() function.
459 The processing is very simple: process the TX port from the RX port, then replace the source and destination MAC addresses.
461 The rte_eth_rx_burst() function writes the mbuf pointers in a local table and returns the number of available mbufs in the table.
463 After first read second try is issued.
467 if (total_nb_rx == MAX_PKT_BURST) {
468 const uint16_t nb_rx = rte_eth_rx_burst((uint8_t) portid, 0, pkts_burst,
471 total_nb_rx += nb_rx;
472 for (j = 0; j < nb_rx; j++) {
474 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
475 l2fwd_simple_forward(m, portid);
479 This second read is important to give job stats library a feedback how many packets was processed.
483 /* Adjust period time in which we are running here. */
484 if (rte_jobstats_finish(job, total_nb_rx) != 0) {
485 rte_timer_reset(&qconf->rx_timers[port_idx], job->period, PERIODICAL,
486 lcore_id, l2fwd_fwd_job, arg);
489 To maximize performance exactly MAX_PKT_BURST is expected (the target value) to be read for each l2fwd_fwd_job() call.
490 If total_nb_rx is smaller than target value job->period will be increased. If it is greater the period will be decreased.
494 In the following code, one line for getting the output port requires some explanation.
496 During the initialization process, a static array of destination ports (l2fwd_dst_ports[]) is filled such that for each source port,
497 a destination port is assigned that is either the next or previous enabled port from the portmask.
498 Naturally, the number of ports in the portmask must be even, otherwise, the application exits.
503 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
505 struct ether_hdr *eth;
509 dst_port = l2fwd_dst_ports[portid];
511 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
513 /* 02:00:00:00:00:xx */
515 tmp = ð->d_addr.addr_bytes[0];
517 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t) dst_port << 40);
521 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
523 l2fwd_send_packet(m, (uint8_t) dst_port);
526 Then, the packet is sent using the l2fwd_send_packet (m, dst_port) function.
527 For this test application, the processing is exactly the same for all packets arriving on the same RX port.
528 Therefore, it would have been possible to call the l2fwd_send_burst() function directly from the main loop
529 to send all the received packets on the same TX port,
530 using the burst-oriented send function, which is more efficient.
532 However, in real-life applications (such as, L3 routing),
533 packet N is not necessarily forwarded on the same port as packet N-1.
534 The application is implemented to illustrate that, so the same approach can be reused in a more complex application.
536 The l2fwd_send_packet() function stores the packet in a per-lcore and per-txport table.
537 If the table is full, the whole packets table is transmitted using the l2fwd_send_burst() function:
541 /* Send the packet on an output interface */
544 l2fwd_send_packet(struct rte_mbuf *m, uint16_t port)
546 unsigned lcore_id, len;
547 struct lcore_queue_conf *qconf;
549 lcore_id = rte_lcore_id();
550 qconf = &lcore_queue_conf[lcore_id];
551 len = qconf->tx_mbufs[port].len;
552 qconf->tx_mbufs[port].m_table[len] = m;
555 /* enough pkts to be sent */
557 if (unlikely(len == MAX_PKT_BURST)) {
558 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
562 qconf->tx_mbufs[port].len = len; return 0;
565 To ensure that no packets remain in the tables, the flush job exists. The l2fwd_flush_job()
566 is called periodically to for each lcore draining TX queue of each port.
567 This technique introduces some latency when there are not many packets to send,
568 however it improves performance:
573 l2fwd_flush_job(__rte_unused struct rte_timer *timer, __rte_unused void *arg)
577 struct lcore_queue_conf *qconf;
578 struct mbuf_table *m_table;
581 lcore_id = rte_lcore_id();
582 qconf = &lcore_queue_conf[lcore_id];
584 rte_jobstats_start(&qconf->jobs_context, &qconf->flush_job);
586 now = rte_get_timer_cycles();
587 lcore_id = rte_lcore_id();
588 qconf = &lcore_queue_conf[lcore_id];
589 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
590 m_table = &qconf->tx_mbufs[portid];
591 if (m_table->len == 0 || m_table->next_flush_time <= now)
594 l2fwd_send_burst(qconf, portid);
598 /* Pass target to indicate that this job is happy of time interval
599 * in which it was called. */
600 rte_jobstats_finish(&qconf->flush_job, qconf->flush_job.target);