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 #. Go to the example directory:
102 .. code-block:: console
104 export RTE_SDK=/path/to/rte_sdk
105 cd ${RTE_SDK}/examples/l2fwd-jobstats
107 #. Set the target (a default target is used if not specified). For example:
109 .. code-block:: console
111 export RTE_TARGET=x86_64-native-linuxapp-gcc
113 *See the DPDK Getting Started Guide* for possible RTE_TARGET values.
115 #. Build the application:
117 .. code-block:: console
121 Running the Application
122 -----------------------
124 The application requires a number of command line options:
126 .. code-block:: console
128 ./build/l2fwd-jobstats [EAL options] -- -p PORTMASK [-q NQ] [-l]
132 * p PORTMASK: A hexadecimal bitmask of the ports to configure
134 * q NQ: A number of queues (=ports) per lcore (default is 1)
136 * l: Use locale thousands separator when formatting big numbers.
138 To run the application in linuxapp environment with 4 lcores, 16 ports, 8 RX queues per lcore and
139 thousands separator printing, issue the command:
141 .. code-block:: console
143 $ ./build/l2fwd-jobstats -c f -n 4 -- -q 8 -p ffff -l
145 Refer to the *DPDK Getting Started Guide* for general information on running applications
146 and the Environment Abstraction Layer (EAL) options.
151 The following sections provide some explanation of the code.
153 Command Line Arguments
154 ~~~~~~~~~~~~~~~~~~~~~~
156 The L2 Forwarding sample application takes specific parameters,
157 in addition to Environment Abstraction Layer (EAL) arguments (see Section 9.3).
158 The preferred way to parse parameters is to use the getopt() function,
159 since it is part of a well-defined and portable library.
161 The parsing of arguments is done in the l2fwd_parse_args() function.
162 The method of argument parsing is not described here.
163 Refer to the *glibc getopt(3)* man page for details.
165 EAL arguments are parsed first, then application-specific arguments.
166 This is done at the beginning of the main() function:
172 ret = rte_eal_init(argc, argv);
174 rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
179 /* parse application arguments (after the EAL ones) */
181 ret = l2fwd_parse_args(argc, argv);
183 rte_exit(EXIT_FAILURE, "Invalid L2FWD arguments\n");
185 Mbuf Pool Initialization
186 ~~~~~~~~~~~~~~~~~~~~~~~~
188 Once the arguments are parsed, the mbuf pool is created.
189 The mbuf pool contains a set of mbuf objects that will be used by the driver
190 and the application to store network packet data:
194 /* create the mbuf pool */
196 rte_mempool_create("mbuf_pool", NB_MBUF,
198 sizeof(struct rte_pktmbuf_pool_private),
199 rte_pktmbuf_pool_init, NULL,
200 rte_pktmbuf_init, NULL,
203 if (l2fwd_pktmbuf_pool == NULL)
204 rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
206 The rte_mempool is a generic structure used to handle pools of objects.
207 In this case, it is necessary to create a pool that will be used by the driver,
208 which expects to have some reserved space in the mempool structure,
209 sizeof(struct rte_pktmbuf_pool_private) bytes.
210 The number of allocated pkt mbufs is NB_MBUF, with a size of MBUF_SIZE each.
211 A per-lcore cache of 32 mbufs is kept.
212 The memory is allocated in rte_socket_id() socket,
213 but it is possible to extend this code to allocate one mbuf pool per socket.
215 Two callback pointers are also given to the rte_mempool_create() function:
217 * The first callback pointer is to rte_pktmbuf_pool_init() and is used
218 to initialize the private data of the mempool, which is needed by the driver.
219 This function is provided by the mbuf API, but can be copied and extended by the developer.
221 * The second callback pointer given to rte_mempool_create() is the mbuf initializer.
222 The default is used, that is, rte_pktmbuf_init(), which is provided in the rte_mbuf library.
223 If a more complex application wants to extend the rte_pktmbuf structure for its own needs,
224 a new function derived from rte_pktmbuf_init( ) can be created.
226 Driver Initialization
227 ~~~~~~~~~~~~~~~~~~~~~
229 The main part of the code in the main() function relates to the initialization of the driver.
230 To fully understand this code, it is recommended to study the chapters that related to the Poll Mode Driver
231 in the *DPDK Programmer's Guide* and the *DPDK API Reference*.
235 nb_ports = rte_eth_dev_count();
238 rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
240 if (nb_ports > RTE_MAX_ETHPORTS)
241 nb_ports = RTE_MAX_ETHPORTS;
243 /* reset l2fwd_dst_ports */
245 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++)
246 l2fwd_dst_ports[portid] = 0;
251 * Each logical core is assigned a dedicated TX queue on each port.
253 for (portid = 0; portid < nb_ports; portid++) {
254 /* skip ports that are not enabled */
255 if ((l2fwd_enabled_port_mask & (1 << portid)) == 0)
258 if (nb_ports_in_mask % 2) {
259 l2fwd_dst_ports[portid] = last_port;
260 l2fwd_dst_ports[last_port] = portid;
267 rte_eth_dev_info_get((uint8_t) portid, &dev_info);
270 The next step is to configure the RX and TX queues.
271 For each port, there is only one RX queue (only one lcore is able to poll a given port).
272 The number of TX queues depends on the number of available lcores.
273 The rte_eth_dev_configure() function is used to configure the number of queues for a port:
277 ret = rte_eth_dev_configure((uint8_t)portid, 1, 1, &port_conf);
279 rte_exit(EXIT_FAILURE, "Cannot configure device: "
283 The global configuration is stored in a static structure:
287 static const struct rte_eth_conf port_conf = {
290 .header_split = 0, /**< Header Split disabled */
291 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
292 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
293 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
294 .hw_strip_crc= 0, /**< CRC stripped by hardware */
298 .mq_mode = ETH_DCB_NONE
302 RX Queue Initialization
303 ~~~~~~~~~~~~~~~~~~~~~~~
305 The application uses one lcore to poll one or several ports, depending on the -q option,
306 which specifies the number of queues per lcore.
308 For example, if the user specifies -q 4, the application is able to poll four ports with one lcore.
309 If there are 16 ports on the target (and if the portmask argument is -p ffff ),
310 the application will need four lcores to poll all the ports.
314 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
315 rte_eth_dev_socket_id(portid),
320 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
321 ret, (unsigned) portid);
323 The list of queues that must be polled for a given lcore is stored in a private structure called struct lcore_queue_conf.
327 struct lcore_queue_conf {
329 unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
330 truct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
332 struct rte_timer rx_timers[MAX_RX_QUEUE_PER_LCORE];
333 struct rte_jobstats port_fwd_jobs[MAX_RX_QUEUE_PER_LCORE];
335 struct rte_timer flush_timer;
336 struct rte_jobstats flush_job;
337 struct rte_jobstats idle_job;
338 struct rte_jobstats_context jobs_context;
340 rte_atomic16_t stats_read_pending;
342 } __rte_cache_aligned;
344 Values of struct lcore_queue_conf:
346 * n_rx_port and rx_port_list[] are used in the main packet processing loop
347 (see Section 9.4.6 "Receive, Process and Transmit Packets" later in this chapter).
349 * rx_timers and flush_timer are used to ensure forced TX on low packet rate.
351 * flush_job, idle_job and jobs_context are librte_jobstats objects used for managing l2fwd jobs.
353 * stats_read_pending and lock are used during job stats read phase.
355 TX Queue Initialization
356 ~~~~~~~~~~~~~~~~~~~~~~~
358 Each lcore should be able to transmit on any port. For every port, a single TX queue is initialized.
362 /* init one TX queue on each port */
365 ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
366 rte_eth_dev_socket_id(portid),
369 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
370 ret, (unsigned) portid);
372 Jobs statistics initialization
373 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
374 There are several statistics objects available:
376 * Flush job statistics
380 rte_jobstats_init(&qconf->flush_job, "flush", drain_tsc, drain_tsc,
383 rte_timer_init(&qconf->flush_timer);
384 ret = rte_timer_reset(&qconf->flush_timer, drain_tsc, PERIODICAL,
385 lcore_id, &l2fwd_flush_job, NULL);
388 rte_exit(1, "Failed to reset flush job timer for lcore %u: %s",
389 lcore_id, rte_strerror(-ret));
392 * Statistics per RX port
396 rte_jobstats_init(job, name, 0, drain_tsc, 0, MAX_PKT_BURST);
397 rte_jobstats_set_update_period_function(job, l2fwd_job_update_cb);
399 rte_timer_init(&qconf->rx_timers[i]);
400 ret = rte_timer_reset(&qconf->rx_timers[i], 0, PERIODICAL, lcore_id,
401 l2fwd_fwd_job, (void *)(uintptr_t)i);
404 rte_exit(1, "Failed to reset lcore %u port %u job timer: %s",
405 lcore_id, qconf->rx_port_list[i], rte_strerror(-ret));
408 Following parameters are passed to rte_jobstats_init():
410 * 0 as minimal poll period
412 * drain_tsc as maximum poll period
414 * MAX_PKT_BURST as desired target value (RX burst size)
419 The forwarding path is reworked comparing to original L2 Forwarding application.
420 In the l2fwd_main_loop() function three loops are placed.
425 rte_spinlock_lock(&qconf->lock);
428 rte_jobstats_context_start(&qconf->jobs_context);
431 * - Read stats_read_pending flag
432 * - check if some real job need to be executed
434 rte_jobstats_start(&qconf->jobs_context, &qconf->idle_job);
438 uint64_t now = rte_get_timer_cycles();
440 need_manage = qconf->flush_timer.expire < now;
441 /* Check if we was esked to give a stats. */
443 rte_atomic16_read(&qconf->stats_read_pending);
444 need_manage |= stats_read_pending;
446 for (i = 0; i < qconf->n_rx_port && !need_manage; i++)
447 need_manage = qconf->rx_timers[i].expire < now;
449 } while (!need_manage);
450 rte_jobstats_finish(&qconf->idle_job, qconf->idle_job.target);
453 rte_jobstats_context_finish(&qconf->jobs_context);
454 } while (likely(stats_read_pending == 0));
456 rte_spinlock_unlock(&qconf->lock);
460 First infinite for loop is to minimize impact of stats reading. Lock is only locked/unlocked when asked.
462 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.
463 In this place functions l2fwd_fwd_job() and l2fwd_flush_job() are called when needed.
464 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
465 and if stats_read_pending is set, loop breaks allowing stats to be read.
467 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
468 for this lcore. Statistics from this part of code is considered as the headroom available for additional processing.
470 Receive, Process and Transmit Packets
471 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
473 The main task of l2fwd_fwd_job() function is to read ingress packets from the RX queue of particular port and forward it.
474 This is done using the following code:
478 total_nb_rx = rte_eth_rx_burst((uint8_t) portid, 0, pkts_burst,
481 for (j = 0; j < total_nb_rx; j++) {
483 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
484 l2fwd_simple_forward(m, portid);
487 Packets are read in a burst of size MAX_PKT_BURST.
488 Then, each mbuf in the table is processed by the l2fwd_simple_forward() function.
489 The processing is very simple: process the TX port from the RX port, then replace the source and destination MAC addresses.
491 The rte_eth_rx_burst() function writes the mbuf pointers in a local table and returns the number of available mbufs in the table.
493 After first read second try is issued.
497 if (total_nb_rx == MAX_PKT_BURST) {
498 const uint16_t nb_rx = rte_eth_rx_burst((uint8_t) portid, 0, pkts_burst,
501 total_nb_rx += nb_rx;
502 for (j = 0; j < nb_rx; j++) {
504 rte_prefetch0(rte_pktmbuf_mtod(m, void *));
505 l2fwd_simple_forward(m, portid);
509 This second read is important to give job stats library a feedback how many packets was processed.
513 /* Adjust period time in which we are running here. */
514 if (rte_jobstats_finish(job, total_nb_rx) != 0) {
515 rte_timer_reset(&qconf->rx_timers[port_idx], job->period, PERIODICAL,
516 lcore_id, l2fwd_fwd_job, arg);
519 To maximize performance exactly MAX_PKT_BURST is expected (the target value) to be read for each l2fwd_fwd_job() call.
520 If total_nb_rx is smaller than target value job->period will be increased. If it is greater the period will be decreased.
524 In the following code, one line for getting the output port requires some explanation.
526 During the initialization process, a static array of destination ports (l2fwd_dst_ports[]) is filled such that for each source port,
527 a destination port is assigned that is either the next or previous enabled port from the portmask.
528 Naturally, the number of ports in the portmask must be even, otherwise, the application exits.
533 l2fwd_simple_forward(struct rte_mbuf *m, unsigned portid)
535 struct ether_hdr *eth;
539 dst_port = l2fwd_dst_ports[portid];
541 eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
543 /* 02:00:00:00:00:xx */
545 tmp = ð->d_addr.addr_bytes[0];
547 *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t) dst_port << 40);
551 ether_addr_copy(&l2fwd_ports_eth_addr[dst_port], ð->s_addr);
553 l2fwd_send_packet(m, (uint8_t) dst_port);
556 Then, the packet is sent using the l2fwd_send_packet (m, dst_port) function.
557 For this test application, the processing is exactly the same for all packets arriving on the same RX port.
558 Therefore, it would have been possible to call the l2fwd_send_burst() function directly from the main loop
559 to send all the received packets on the same TX port,
560 using the burst-oriented send function, which is more efficient.
562 However, in real-life applications (such as, L3 routing),
563 packet N is not necessarily forwarded on the same port as packet N-1.
564 The application is implemented to illustrate that, so the same approach can be reused in a more complex application.
566 The l2fwd_send_packet() function stores the packet in a per-lcore and per-txport table.
567 If the table is full, the whole packets table is transmitted using the l2fwd_send_burst() function:
571 /* Send the packet on an output interface */
574 l2fwd_send_packet(struct rte_mbuf *m, uint8_t port)
576 unsigned lcore_id, len;
577 struct lcore_queue_conf *qconf;
579 lcore_id = rte_lcore_id();
580 qconf = &lcore_queue_conf[lcore_id];
581 len = qconf->tx_mbufs[port].len;
582 qconf->tx_mbufs[port].m_table[len] = m;
585 /* enough pkts to be sent */
587 if (unlikely(len == MAX_PKT_BURST)) {
588 l2fwd_send_burst(qconf, MAX_PKT_BURST, port);
592 qconf->tx_mbufs[port].len = len; return 0;
595 To ensure that no packets remain in the tables, the flush job exists. The l2fwd_flush_job()
596 is called periodically to for each lcore draining TX queue of each port.
597 This technique introduces some latency when there are not many packets to send,
598 however it improves performance:
603 l2fwd_flush_job(__rte_unused struct rte_timer *timer, __rte_unused void *arg)
607 struct lcore_queue_conf *qconf;
608 struct mbuf_table *m_table;
611 lcore_id = rte_lcore_id();
612 qconf = &lcore_queue_conf[lcore_id];
614 rte_jobstats_start(&qconf->jobs_context, &qconf->flush_job);
616 now = rte_get_timer_cycles();
617 lcore_id = rte_lcore_id();
618 qconf = &lcore_queue_conf[lcore_id];
619 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
620 m_table = &qconf->tx_mbufs[portid];
621 if (m_table->len == 0 || m_table->next_flush_time <= now)
624 l2fwd_send_burst(qconf, portid);
628 /* Pass target to indicate that this job is happy of time interval
629 * in which it was called. */
630 rte_jobstats_finish(&qconf->flush_job, qconf->flush_job.target);