4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
38 #include <sys/types.h>
40 #include <sys/queue.h>
45 #include <rte_common.h>
46 #include <rte_byteorder.h>
48 #include <rte_memory.h>
49 #include <rte_memcpy.h>
50 #include <rte_memzone.h>
51 #include <rte_tailq.h>
53 #include <rte_per_lcore.h>
54 #include <rte_launch.h>
55 #include <rte_atomic.h>
56 #include <rte_cycles.h>
57 #include <rte_prefetch.h>
58 #include <rte_lcore.h>
59 #include <rte_per_lcore.h>
60 #include <rte_branch_prediction.h>
61 #include <rte_interrupts.h>
63 #include <rte_random.h>
64 #include <rte_debug.h>
65 #include <rte_ether.h>
66 #include <rte_ethdev.h>
68 #include <rte_mempool.h>
70 #include <rte_string_fns.h>
77 static struct rte_eth_conf port_conf = {
79 .mq_mode = ETH_MQ_RX_RSS,
81 .header_split = 0, /**< Header Split disabled */
82 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
83 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
84 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
85 .hw_strip_crc = 0, /**< CRC stripped by hardware */
94 .mq_mode = ETH_MQ_TX_NONE,
98 static struct rte_eth_rxconf rx_conf = {
100 .pthresh = APP_DEFAULT_NIC_RX_PTHRESH,
101 .hthresh = APP_DEFAULT_NIC_RX_HTHRESH,
102 .wthresh = APP_DEFAULT_NIC_RX_WTHRESH,
104 .rx_free_thresh = APP_DEFAULT_NIC_RX_FREE_THRESH,
105 .rx_drop_en = APP_DEFAULT_NIC_RX_DROP_EN,
108 static struct rte_eth_txconf tx_conf = {
110 .pthresh = APP_DEFAULT_NIC_TX_PTHRESH,
111 .hthresh = APP_DEFAULT_NIC_TX_HTHRESH,
112 .wthresh = APP_DEFAULT_NIC_TX_WTHRESH,
114 .tx_free_thresh = APP_DEFAULT_NIC_TX_FREE_THRESH,
115 .tx_rs_thresh = APP_DEFAULT_NIC_TX_RS_THRESH,
119 app_assign_worker_ids(void)
121 uint32_t lcore, worker_id;
123 /* Assign ID for each worker */
125 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
126 struct app_lcore_params_worker *lp_worker = &app.lcore_params[lcore].worker;
128 if (app.lcore_params[lcore].type != e_APP_LCORE_WORKER) {
132 lp_worker->worker_id = worker_id;
138 app_init_mbuf_pools(void)
140 unsigned socket, lcore;
142 /* Init the buffer pools */
143 for (socket = 0; socket < APP_MAX_SOCKETS; socket ++) {
145 if (app_is_socket_used(socket) == 0) {
149 snprintf(name, sizeof(name), "mbuf_pool_%u", socket);
150 printf("Creating the mbuf pool for socket %u ...\n", socket);
151 app.pools[socket] = rte_mempool_create(
153 APP_DEFAULT_MEMPOOL_BUFFERS,
154 APP_DEFAULT_MBUF_SIZE,
155 APP_DEFAULT_MEMPOOL_CACHE_SIZE,
156 sizeof(struct rte_pktmbuf_pool_private),
157 rte_pktmbuf_pool_init, NULL,
158 rte_pktmbuf_init, NULL,
161 if (app.pools[socket] == NULL) {
162 rte_panic("Cannot create mbuf pool on socket %u\n", socket);
166 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
167 if (app.lcore_params[lcore].type == e_APP_LCORE_DISABLED) {
171 socket = rte_lcore_to_socket_id(lcore);
172 app.lcore_params[lcore].pool = app.pools[socket];
177 app_init_lpm_tables(void)
179 unsigned socket, lcore;
181 /* Init the LPM tables */
182 for (socket = 0; socket < APP_MAX_SOCKETS; socket ++) {
186 if (app_is_socket_used(socket) == 0) {
190 snprintf(name, sizeof(name), "lpm_table_%u", socket);
191 printf("Creating the LPM table for socket %u ...\n", socket);
192 app.lpm_tables[socket] = rte_lpm_create(
197 if (app.lpm_tables[socket] == NULL) {
198 rte_panic("Unable to create LPM table on socket %u\n", socket);
201 for (rule = 0; rule < app.n_lpm_rules; rule ++) {
204 ret = rte_lpm_add(app.lpm_tables[socket],
205 app.lpm_rules[rule].ip,
206 app.lpm_rules[rule].depth,
207 app.lpm_rules[rule].if_out);
210 rte_panic("Unable to add entry %u (%x/%u => %u) to the LPM table on socket %u (%d)\n",
212 (unsigned) app.lpm_rules[rule].ip,
213 (unsigned) app.lpm_rules[rule].depth,
214 (unsigned) app.lpm_rules[rule].if_out,
222 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
223 if (app.lcore_params[lcore].type != e_APP_LCORE_WORKER) {
227 socket = rte_lcore_to_socket_id(lcore);
228 app.lcore_params[lcore].worker.lpm_table = app.lpm_tables[socket];
233 app_init_rings_rx(void)
237 /* Initialize the rings for the RX side */
238 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
239 struct app_lcore_params_io *lp_io = &app.lcore_params[lcore].io;
240 unsigned socket_io, lcore_worker;
242 if ((app.lcore_params[lcore].type != e_APP_LCORE_IO) ||
243 (lp_io->rx.n_nic_queues == 0)) {
247 socket_io = rte_lcore_to_socket_id(lcore);
249 for (lcore_worker = 0; lcore_worker < APP_MAX_LCORES; lcore_worker ++) {
251 struct app_lcore_params_worker *lp_worker = &app.lcore_params[lcore_worker].worker;
252 struct rte_ring *ring = NULL;
254 if (app.lcore_params[lcore_worker].type != e_APP_LCORE_WORKER) {
258 printf("Creating ring to connect I/O lcore %u (socket %u) with worker lcore %u ...\n",
262 snprintf(name, sizeof(name), "app_ring_rx_s%u_io%u_w%u",
266 ring = rte_ring_create(
270 RING_F_SP_ENQ | RING_F_SC_DEQ);
272 rte_panic("Cannot create ring to connect I/O core %u with worker core %u\n",
277 lp_io->rx.rings[lp_io->rx.n_rings] = ring;
278 lp_io->rx.n_rings ++;
280 lp_worker->rings_in[lp_worker->n_rings_in] = ring;
281 lp_worker->n_rings_in ++;
285 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
286 struct app_lcore_params_io *lp_io = &app.lcore_params[lcore].io;
288 if ((app.lcore_params[lcore].type != e_APP_LCORE_IO) ||
289 (lp_io->rx.n_nic_queues == 0)) {
293 if (lp_io->rx.n_rings != app_get_lcores_worker()) {
294 rte_panic("Algorithmic error (I/O RX rings)\n");
298 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
299 struct app_lcore_params_worker *lp_worker = &app.lcore_params[lcore].worker;
301 if (app.lcore_params[lcore].type != e_APP_LCORE_WORKER) {
305 if (lp_worker->n_rings_in != app_get_lcores_io_rx()) {
306 rte_panic("Algorithmic error (worker input rings)\n");
312 app_init_rings_tx(void)
316 /* Initialize the rings for the TX side */
317 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
318 struct app_lcore_params_worker *lp_worker = &app.lcore_params[lcore].worker;
321 if (app.lcore_params[lcore].type != e_APP_LCORE_WORKER) {
325 for (port = 0; port < APP_MAX_NIC_PORTS; port ++) {
327 struct app_lcore_params_io *lp_io = NULL;
328 struct rte_ring *ring;
329 uint32_t socket_io, lcore_io;
331 if (app.nic_tx_port_mask[port] == 0) {
335 if (app_get_lcore_for_nic_tx((uint8_t) port, &lcore_io) < 0) {
336 rte_panic("Algorithmic error (no I/O core to handle TX of port %u)\n",
340 lp_io = &app.lcore_params[lcore_io].io;
341 socket_io = rte_lcore_to_socket_id(lcore_io);
343 printf("Creating ring to connect worker lcore %u with TX port %u (through I/O lcore %u) (socket %u) ...\n",
344 lcore, port, (unsigned)lcore_io, (unsigned)socket_io);
345 snprintf(name, sizeof(name), "app_ring_tx_s%u_w%u_p%u", socket_io, lcore, port);
346 ring = rte_ring_create(
350 RING_F_SP_ENQ | RING_F_SC_DEQ);
352 rte_panic("Cannot create ring to connect worker core %u with TX port %u\n",
357 lp_worker->rings_out[port] = ring;
358 lp_io->tx.rings[port][lp_worker->worker_id] = ring;
362 for (lcore = 0; lcore < APP_MAX_LCORES; lcore ++) {
363 struct app_lcore_params_io *lp_io = &app.lcore_params[lcore].io;
366 if ((app.lcore_params[lcore].type != e_APP_LCORE_IO) ||
367 (lp_io->tx.n_nic_ports == 0)) {
371 for (i = 0; i < lp_io->tx.n_nic_ports; i ++){
374 port = lp_io->tx.nic_ports[i];
375 for (j = 0; j < app_get_lcores_worker(); j ++) {
376 if (lp_io->tx.rings[port][j] == NULL) {
377 rte_panic("Algorithmic error (I/O TX rings)\n");
384 /* Check the link status of all ports in up to 9s, and print them finally */
386 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
388 #define CHECK_INTERVAL 100 /* 100ms */
389 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
390 uint8_t portid, count, all_ports_up, print_flag = 0;
391 struct rte_eth_link link;
392 uint32_t n_rx_queues, n_tx_queues;
394 printf("\nChecking link status");
396 for (count = 0; count <= MAX_CHECK_TIME; count++) {
398 for (portid = 0; portid < port_num; portid++) {
399 if ((port_mask & (1 << portid)) == 0)
401 n_rx_queues = app_get_nic_rx_queues_per_port(portid);
402 n_tx_queues = app.nic_tx_port_mask[portid];
403 if ((n_rx_queues == 0) && (n_tx_queues == 0))
405 memset(&link, 0, sizeof(link));
406 rte_eth_link_get_nowait(portid, &link);
407 /* print link status if flag set */
408 if (print_flag == 1) {
409 if (link.link_status)
410 printf("Port %d Link Up - speed %u "
411 "Mbps - %s\n", (uint8_t)portid,
412 (unsigned)link.link_speed,
413 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
414 ("full-duplex") : ("half-duplex\n"));
416 printf("Port %d Link Down\n",
420 /* clear all_ports_up flag if any link down */
421 if (link.link_status == 0) {
426 /* after finally printing all link status, get out */
430 if (all_ports_up == 0) {
433 rte_delay_ms(CHECK_INTERVAL);
436 /* set the print_flag if all ports up or timeout */
437 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
451 uint32_t n_rx_queues, n_tx_queues;
453 if (rte_eal_pci_probe() < 0) {
454 rte_panic("Cannot probe PCI\n");
457 /* Init NIC ports and queues, then start the ports */
458 for (port = 0; port < APP_MAX_NIC_PORTS; port ++) {
459 struct rte_mempool *pool;
461 n_rx_queues = app_get_nic_rx_queues_per_port(port);
462 n_tx_queues = app.nic_tx_port_mask[port];
464 if ((n_rx_queues == 0) && (n_tx_queues == 0)) {
469 printf("Initializing NIC port %u ...\n", (unsigned) port);
470 ret = rte_eth_dev_configure(
472 (uint8_t) n_rx_queues,
473 (uint8_t) n_tx_queues,
476 rte_panic("Cannot init NIC port %u (%d)\n", (unsigned) port, ret);
478 rte_eth_promiscuous_enable(port);
481 for (queue = 0; queue < APP_MAX_RX_QUEUES_PER_NIC_PORT; queue ++) {
482 if (app.nic_rx_queue_mask[port][queue] == 0) {
486 app_get_lcore_for_nic_rx(port, queue, &lcore);
487 socket = rte_lcore_to_socket_id(lcore);
488 pool = app.lcore_params[lcore].pool;
490 printf("Initializing NIC port %u RX queue %u ...\n",
493 ret = rte_eth_rx_queue_setup(
496 (uint16_t) app.nic_rx_ring_size,
501 rte_panic("Cannot init RX queue %u for port %u (%d)\n",
509 if (app.nic_tx_port_mask[port] == 1) {
510 app_get_lcore_for_nic_tx(port, &lcore);
511 socket = rte_lcore_to_socket_id(lcore);
512 printf("Initializing NIC port %u TX queue 0 ...\n",
514 ret = rte_eth_tx_queue_setup(
517 (uint16_t) app.nic_tx_ring_size,
521 rte_panic("Cannot init TX queue 0 for port %d (%d)\n",
528 ret = rte_eth_dev_start(port);
530 rte_panic("Cannot start port %d (%d)\n", port, ret);
534 check_all_ports_link_status(APP_MAX_NIC_PORTS, (~0x0));
540 app_assign_worker_ids();
541 app_init_mbuf_pools();
542 app_init_lpm_tables();
547 printf("Initialization completed.\n");