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>
39 #include <sys/param.h>
41 #include <sys/queue.h>
46 #include <rte_common.h>
47 #include <rte_byteorder.h>
49 #include <rte_memory.h>
50 #include <rte_memcpy.h>
52 #include <rte_launch.h>
53 #include <rte_atomic.h>
54 #include <rte_cycles.h>
55 #include <rte_prefetch.h>
56 #include <rte_lcore.h>
57 #include <rte_per_lcore.h>
58 #include <rte_branch_prediction.h>
59 #include <rte_interrupts.h>
60 #include <rte_random.h>
61 #include <rte_debug.h>
62 #include <rte_ether.h>
63 #include <rte_ethdev.h>
64 #include <rte_mempool.h>
69 #include <rte_string_fns.h>
71 #include <rte_ip_frag.h>
73 #define RTE_LOGTYPE_IP_FRAG RTE_LOGTYPE_USER1
75 /* allow max jumbo frame 9.5 KB */
76 #define JUMBO_FRAME_MAX_SIZE 0x2600
78 #define ROUNDUP_DIV(a, b) (((a) + (b) - 1) / (b))
81 * Default byte size for the IPv6 Maximum Transfer Unit (MTU).
82 * This value includes the size of IPv6 header.
84 #define IPV4_MTU_DEFAULT ETHER_MTU
85 #define IPV6_MTU_DEFAULT ETHER_MTU
88 * Default payload in bytes for the IPv6 packet.
90 #define IPV4_DEFAULT_PAYLOAD (IPV4_MTU_DEFAULT - sizeof(struct ipv4_hdr))
91 #define IPV6_DEFAULT_PAYLOAD (IPV6_MTU_DEFAULT - sizeof(struct ipv6_hdr))
94 * Max number of fragments per packet expected - defined by config file.
96 #define MAX_PACKET_FRAG RTE_LIBRTE_IP_FRAG_MAX_FRAG
100 #define MAX_PKT_BURST 32
101 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
103 /* Configure how many packets ahead to prefetch, when reading packets */
104 #define PREFETCH_OFFSET 3
107 * Configurable number of RX/TX ring descriptors
109 #define RTE_TEST_RX_DESC_DEFAULT 128
110 #define RTE_TEST_TX_DESC_DEFAULT 512
111 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
112 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
114 /* ethernet addresses of ports */
115 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
118 #define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8
119 #define IPv4_BYTES(addr) \
120 (uint8_t) (((addr) >> 24) & 0xFF),\
121 (uint8_t) (((addr) >> 16) & 0xFF),\
122 (uint8_t) (((addr) >> 8) & 0xFF),\
123 (uint8_t) ((addr) & 0xFF)
127 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\
128 "%02x%02x:%02x%02x:%02x%02x:%02x%02x"
129 #define IPv6_BYTES(addr) \
130 addr[0], addr[1], addr[2], addr[3], \
131 addr[4], addr[5], addr[6], addr[7], \
132 addr[8], addr[9], addr[10], addr[11],\
133 addr[12], addr[13],addr[14], addr[15]
136 #define IPV6_ADDR_LEN 16
138 /* mask of enabled ports */
139 static int enabled_port_mask = 0;
141 static int rx_queue_per_lcore = 1;
143 #define MBUF_TABLE_SIZE (2 * MAX(MAX_PKT_BURST, MAX_PACKET_FRAG))
147 struct rte_mbuf *m_table[MBUF_TABLE_SIZE];
151 struct rte_mempool *direct_pool;
152 struct rte_mempool *indirect_pool;
154 struct rte_lpm6 *lpm6;
158 #define MAX_RX_QUEUE_PER_LCORE 16
159 #define MAX_TX_QUEUE_PER_PORT 16
160 struct lcore_queue_conf {
162 uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
163 struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
164 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
165 } __rte_cache_aligned;
166 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
168 static struct rte_eth_conf port_conf = {
170 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
172 .header_split = 0, /**< Header Split disabled */
173 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
174 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
175 .jumbo_frame = 1, /**< Jumbo Frame Support enabled */
176 .hw_strip_crc = 1, /**< CRC stripped by hardware */
179 .mq_mode = ETH_MQ_TX_NONE,
184 * IPv4 forwarding table
186 struct l3fwd_ipv4_route {
192 struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = {
193 {IPv4(100,10,0,0), 16, 0},
194 {IPv4(100,20,0,0), 16, 1},
195 {IPv4(100,30,0,0), 16, 2},
196 {IPv4(100,40,0,0), 16, 3},
197 {IPv4(100,50,0,0), 16, 4},
198 {IPv4(100,60,0,0), 16, 5},
199 {IPv4(100,70,0,0), 16, 6},
200 {IPv4(100,80,0,0), 16, 7},
204 * IPv6 forwarding table
207 struct l3fwd_ipv6_route {
208 uint8_t ip[IPV6_ADDR_LEN];
213 static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = {
214 {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0},
215 {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1},
216 {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2},
217 {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3},
218 {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4},
219 {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5},
220 {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6},
221 {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7},
224 #define LPM_MAX_RULES 1024
225 #define LPM6_MAX_RULES 1024
226 #define LPM6_NUMBER_TBL8S (1 << 16)
228 struct rte_lpm6_config lpm6_config = {
229 .max_rules = LPM6_MAX_RULES,
230 .number_tbl8s = LPM6_NUMBER_TBL8S,
234 static struct rte_mempool *socket_direct_pool[RTE_MAX_NUMA_NODES];
235 static struct rte_mempool *socket_indirect_pool[RTE_MAX_NUMA_NODES];
236 static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES];
237 static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES];
239 /* Send burst of packets on an output interface */
241 send_burst(struct lcore_queue_conf *qconf, uint16_t n, uint16_t port)
243 struct rte_mbuf **m_table;
247 queueid = qconf->tx_queue_id[port];
248 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
250 ret = rte_eth_tx_burst(port, queueid, m_table, n);
251 if (unlikely(ret < n)) {
253 rte_pktmbuf_free(m_table[ret]);
261 l3fwd_simple_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf,
262 uint8_t queueid, uint16_t port_in)
264 struct rx_queue *rxq;
265 uint32_t i, len, next_hop;
271 rxq = &qconf->rx_queue_list[queueid];
273 /* by default, send everything back to the source port */
276 /* Remove the Ethernet header and trailer from the input packet */
277 rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
279 /* Build transmission burst */
280 len = qconf->tx_mbufs[port_out].len;
282 /* if this is an IPv4 packet */
283 if (RTE_ETH_IS_IPV4_HDR(m->packet_type)) {
284 struct ipv4_hdr *ip_hdr;
286 /* Read the lookup key (i.e. ip_dst) from the input packet */
287 ip_hdr = rte_pktmbuf_mtod(m, struct ipv4_hdr *);
288 ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr);
290 /* Find destination port */
291 if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop) == 0 &&
292 (enabled_port_mask & 1 << next_hop) != 0) {
295 /* Build transmission burst for new port */
296 len = qconf->tx_mbufs[port_out].len;
299 /* if we don't need to do any fragmentation */
300 if (likely (IPV4_MTU_DEFAULT >= m->pkt_len)) {
301 qconf->tx_mbufs[port_out].m_table[len] = m;
304 len2 = rte_ipv4_fragment_packet(m,
305 &qconf->tx_mbufs[port_out].m_table[len],
306 (uint16_t)(MBUF_TABLE_SIZE - len),
308 rxq->direct_pool, rxq->indirect_pool);
310 /* Free input packet */
313 /* If we fail to fragment the packet */
314 if (unlikely (len2 < 0))
317 } else if (RTE_ETH_IS_IPV6_HDR(m->packet_type)) {
318 /* if this is an IPv6 packet */
319 struct ipv6_hdr *ip_hdr;
323 /* Read the lookup key (i.e. ip_dst) from the input packet */
324 ip_hdr = rte_pktmbuf_mtod(m, struct ipv6_hdr *);
326 /* Find destination port */
327 if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr,
329 (enabled_port_mask & 1 << next_hop) != 0) {
332 /* Build transmission burst for new port */
333 len = qconf->tx_mbufs[port_out].len;
336 /* if we don't need to do any fragmentation */
337 if (likely (IPV6_MTU_DEFAULT >= m->pkt_len)) {
338 qconf->tx_mbufs[port_out].m_table[len] = m;
341 len2 = rte_ipv6_fragment_packet(m,
342 &qconf->tx_mbufs[port_out].m_table[len],
343 (uint16_t)(MBUF_TABLE_SIZE - len),
345 rxq->direct_pool, rxq->indirect_pool);
347 /* Free input packet */
350 /* If we fail to fragment the packet */
351 if (unlikely (len2 < 0))
355 /* else, just forward the packet */
357 qconf->tx_mbufs[port_out].m_table[len] = m;
361 for (i = len; i < len + len2; i ++) {
364 m = qconf->tx_mbufs[port_out].m_table[i];
365 struct ether_hdr *eth_hdr = (struct ether_hdr *)
366 rte_pktmbuf_prepend(m, (uint16_t)sizeof(struct ether_hdr));
367 if (eth_hdr == NULL) {
368 rte_panic("No headroom in mbuf.\n");
371 m->l2_len = sizeof(struct ether_hdr);
373 /* 02:00:00:00:00:xx */
374 d_addr_bytes = ð_hdr->d_addr.addr_bytes[0];
375 *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)port_out << 40);
378 ether_addr_copy(&ports_eth_addr[port_out], ð_hdr->s_addr);
380 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv6);
382 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
387 if (likely(len < MAX_PKT_BURST)) {
388 qconf->tx_mbufs[port_out].len = (uint16_t)len;
392 /* Transmit packets */
393 send_burst(qconf, (uint16_t)len, port_out);
394 qconf->tx_mbufs[port_out].len = 0;
397 /* main processing loop */
399 main_loop(__attribute__((unused)) void *dummy)
401 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
403 uint64_t prev_tsc, diff_tsc, cur_tsc;
406 struct lcore_queue_conf *qconf;
407 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
411 lcore_id = rte_lcore_id();
412 qconf = &lcore_queue_conf[lcore_id];
414 if (qconf->n_rx_queue == 0) {
415 RTE_LOG(INFO, IP_FRAG, "lcore %u has nothing to do\n", lcore_id);
419 RTE_LOG(INFO, IP_FRAG, "entering main loop on lcore %u\n", lcore_id);
421 for (i = 0; i < qconf->n_rx_queue; i++) {
423 portid = qconf->rx_queue_list[i].portid;
424 RTE_LOG(INFO, IP_FRAG, " -- lcoreid=%u portid=%d\n", lcore_id,
430 cur_tsc = rte_rdtsc();
433 * TX burst queue drain
435 diff_tsc = cur_tsc - prev_tsc;
436 if (unlikely(diff_tsc > drain_tsc)) {
439 * This could be optimized (use queueid instead of
440 * portid), but it is not called so often
442 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
443 if (qconf->tx_mbufs[portid].len == 0)
445 send_burst(&lcore_queue_conf[lcore_id],
446 qconf->tx_mbufs[portid].len,
448 qconf->tx_mbufs[portid].len = 0;
455 * Read packet from RX queues
457 for (i = 0; i < qconf->n_rx_queue; i++) {
459 portid = qconf->rx_queue_list[i].portid;
460 nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
463 /* Prefetch first packets */
464 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
465 rte_prefetch0(rte_pktmbuf_mtod(
466 pkts_burst[j], void *));
469 /* Prefetch and forward already prefetched packets */
470 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
471 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
472 j + PREFETCH_OFFSET], void *));
473 l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
476 /* Forward remaining prefetched packets */
477 for (; j < nb_rx; j++) {
478 l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
486 print_usage(const char *prgname)
488 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
489 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
490 " -q NQ: number of queue (=ports) per lcore (default is 1)\n",
495 parse_portmask(const char *portmask)
500 /* parse hexadecimal string */
501 pm = strtoul(portmask, &end, 16);
502 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
512 parse_nqueue(const char *q_arg)
517 /* parse hexadecimal string */
518 n = strtoul(q_arg, &end, 10);
519 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
523 if (n >= MAX_RX_QUEUE_PER_LCORE)
529 /* Parse the argument given in the command line of the application */
531 parse_args(int argc, char **argv)
536 char *prgname = argv[0];
537 static struct option lgopts[] = {
543 while ((opt = getopt_long(argc, argvopt, "p:q:",
544 lgopts, &option_index)) != EOF) {
549 enabled_port_mask = parse_portmask(optarg);
550 if (enabled_port_mask < 0) {
551 printf("invalid portmask\n");
552 print_usage(prgname);
559 rx_queue_per_lcore = parse_nqueue(optarg);
560 if (rx_queue_per_lcore < 0) {
561 printf("invalid queue number\n");
562 print_usage(prgname);
569 print_usage(prgname);
573 print_usage(prgname);
578 if (enabled_port_mask == 0) {
579 printf("portmask not specified\n");
580 print_usage(prgname);
585 argv[optind-1] = prgname;
588 optind = 1; /* reset getopt lib */
593 print_ethaddr(const char *name, struct ether_addr *eth_addr)
595 char buf[ETHER_ADDR_FMT_SIZE];
596 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
597 printf("%s%s", name, buf);
600 /* Check the link status of all ports in up to 9s, and print them finally */
602 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask)
604 #define CHECK_INTERVAL 100 /* 100ms */
605 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
607 uint8_t count, all_ports_up, print_flag = 0;
608 struct rte_eth_link link;
610 printf("\nChecking link status");
612 for (count = 0; count <= MAX_CHECK_TIME; count++) {
614 for (portid = 0; portid < port_num; portid++) {
615 if ((port_mask & (1 << portid)) == 0)
617 memset(&link, 0, sizeof(link));
618 rte_eth_link_get_nowait(portid, &link);
619 /* print link status if flag set */
620 if (print_flag == 1) {
621 if (link.link_status)
623 "Port%d Link Up .Speed %u Mbps - %s\n",
624 portid, link.link_speed,
625 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
626 ("full-duplex") : ("half-duplex\n"));
628 printf("Port %d Link Down\n", portid);
631 /* clear all_ports_up flag if any link down */
632 if (link.link_status == ETH_LINK_DOWN) {
637 /* after finally printing all link status, get out */
641 if (all_ports_up == 0) {
644 rte_delay_ms(CHECK_INTERVAL);
647 /* set the print_flag if all ports up or timeout */
648 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
655 /* Check L3 packet type detection capablity of the NIC port */
657 check_ptype(int portid)
660 int ptype_l3_ipv4 = 0, ptype_l3_ipv6 = 0;
661 uint32_t ptype_mask = RTE_PTYPE_L3_MASK;
663 ret = rte_eth_dev_get_supported_ptypes(portid, ptype_mask, NULL, 0);
667 uint32_t ptypes[ret];
669 ret = rte_eth_dev_get_supported_ptypes(portid, ptype_mask, ptypes, ret);
670 for (i = 0; i < ret; ++i) {
671 if (ptypes[i] & RTE_PTYPE_L3_IPV4)
673 if (ptypes[i] & RTE_PTYPE_L3_IPV6)
677 if (ptype_l3_ipv4 == 0)
678 printf("port %d cannot parse RTE_PTYPE_L3_IPV4\n", portid);
680 if (ptype_l3_ipv6 == 0)
681 printf("port %d cannot parse RTE_PTYPE_L3_IPV6\n", portid);
683 if (ptype_l3_ipv4 && ptype_l3_ipv6)
690 /* Parse packet type of a packet by SW */
692 parse_ptype(struct rte_mbuf *m)
694 struct ether_hdr *eth_hdr;
695 uint32_t packet_type = RTE_PTYPE_UNKNOWN;
698 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
699 ether_type = eth_hdr->ether_type;
700 if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv4))
701 packet_type |= RTE_PTYPE_L3_IPV4_EXT_UNKNOWN;
702 else if (ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv6))
703 packet_type |= RTE_PTYPE_L3_IPV6_EXT_UNKNOWN;
705 m->packet_type = packet_type;
708 /* callback function to detect packet type for a queue of a port */
710 cb_parse_ptype(uint16_t port __rte_unused, uint16_t queue __rte_unused,
711 struct rte_mbuf *pkts[], uint16_t nb_pkts,
712 uint16_t max_pkts __rte_unused,
713 void *user_param __rte_unused)
717 for (i = 0; i < nb_pkts; ++i)
718 parse_ptype(pkts[i]);
724 init_routing_table(void)
727 struct rte_lpm6 *lpm6;
731 for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) {
732 if (socket_lpm[socket]) {
733 lpm = socket_lpm[socket];
734 /* populate the LPM table */
735 for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) {
736 ret = rte_lpm_add(lpm,
737 l3fwd_ipv4_route_array[i].ip,
738 l3fwd_ipv4_route_array[i].depth,
739 l3fwd_ipv4_route_array[i].if_out);
742 RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
747 RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv4_BYTES_FMT
750 IPv4_BYTES(l3fwd_ipv4_route_array[i].ip),
751 l3fwd_ipv4_route_array[i].depth,
752 l3fwd_ipv4_route_array[i].if_out);
756 if (socket_lpm6[socket]) {
757 lpm6 = socket_lpm6[socket];
758 /* populate the LPM6 table */
759 for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) {
760 ret = rte_lpm6_add(lpm6,
761 l3fwd_ipv6_route_array[i].ip,
762 l3fwd_ipv6_route_array[i].depth,
763 l3fwd_ipv6_route_array[i].if_out);
766 RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
771 RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv6_BYTES_FMT
774 IPv6_BYTES(l3fwd_ipv6_route_array[i].ip),
775 l3fwd_ipv6_route_array[i].depth,
776 l3fwd_ipv6_route_array[i].if_out);
787 struct rte_mempool *mp;
789 struct rte_lpm6 *lpm6;
790 struct rte_lpm_config lpm_config;
794 /* traverse through lcores and initialize structures on each socket */
796 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
798 if (rte_lcore_is_enabled(lcore_id) == 0)
801 socket = rte_lcore_to_socket_id(lcore_id);
803 if (socket == SOCKET_ID_ANY)
806 if (socket_direct_pool[socket] == NULL) {
807 RTE_LOG(INFO, IP_FRAG, "Creating direct mempool on socket %i\n",
809 snprintf(buf, sizeof(buf), "pool_direct_%i", socket);
811 mp = rte_pktmbuf_pool_create(buf, NB_MBUF, 32,
812 0, RTE_MBUF_DEFAULT_BUF_SIZE, socket);
814 RTE_LOG(ERR, IP_FRAG, "Cannot create direct mempool\n");
817 socket_direct_pool[socket] = mp;
820 if (socket_indirect_pool[socket] == NULL) {
821 RTE_LOG(INFO, IP_FRAG, "Creating indirect mempool on socket %i\n",
823 snprintf(buf, sizeof(buf), "pool_indirect_%i", socket);
825 mp = rte_pktmbuf_pool_create(buf, NB_MBUF, 32, 0, 0,
828 RTE_LOG(ERR, IP_FRAG, "Cannot create indirect mempool\n");
831 socket_indirect_pool[socket] = mp;
834 if (socket_lpm[socket] == NULL) {
835 RTE_LOG(INFO, IP_FRAG, "Creating LPM table on socket %i\n", socket);
836 snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
838 lpm_config.max_rules = LPM_MAX_RULES;
839 lpm_config.number_tbl8s = 256;
840 lpm_config.flags = 0;
842 lpm = rte_lpm_create(buf, socket, &lpm_config);
844 RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
847 socket_lpm[socket] = lpm;
850 if (socket_lpm6[socket] == NULL) {
851 RTE_LOG(INFO, IP_FRAG, "Creating LPM6 table on socket %i\n", socket);
852 snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
854 lpm6 = rte_lpm6_create(buf, socket, &lpm6_config);
856 RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
859 socket_lpm6[socket] = lpm6;
867 main(int argc, char **argv)
869 struct lcore_queue_conf *qconf;
870 struct rte_eth_dev_info dev_info;
871 struct rte_eth_txconf *txconf;
872 struct rx_queue *rxq;
875 uint16_t queueid = 0;
876 unsigned lcore_id = 0, rx_lcore_id = 0;
877 uint32_t n_tx_queue, nb_lcores;
881 ret = rte_eal_init(argc, argv);
883 rte_exit(EXIT_FAILURE, "rte_eal_init failed");
887 /* parse application arguments (after the EAL ones) */
888 ret = parse_args(argc, argv);
890 rte_exit(EXIT_FAILURE, "Invalid arguments");
892 nb_ports = rte_eth_dev_count();
894 rte_exit(EXIT_FAILURE, "No ports found!\n");
896 nb_lcores = rte_lcore_count();
898 /* initialize structures (mempools, lpm etc.) */
900 rte_panic("Cannot initialize memory structures!\n");
902 /* check if portmask has non-existent ports */
903 if (enabled_port_mask & ~(RTE_LEN2MASK(nb_ports, unsigned)))
904 rte_exit(EXIT_FAILURE, "Non-existent ports in portmask!\n");
906 /* initialize all ports */
907 for (portid = 0; portid < nb_ports; portid++) {
908 /* skip ports that are not enabled */
909 if ((enabled_port_mask & (1 << portid)) == 0) {
910 printf("Skipping disabled port %d\n", portid);
914 qconf = &lcore_queue_conf[rx_lcore_id];
916 /* limit the frame size to the maximum supported by NIC */
917 rte_eth_dev_info_get(portid, &dev_info);
918 port_conf.rxmode.max_rx_pkt_len = RTE_MIN(
919 dev_info.max_rx_pktlen, port_conf.rxmode.max_rx_pkt_len);
921 /* get the lcore_id for this port */
922 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
923 qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
926 if (rx_lcore_id >= RTE_MAX_LCORE)
927 rte_exit(EXIT_FAILURE, "Not enough cores\n");
929 qconf = &lcore_queue_conf[rx_lcore_id];
932 socket = (int) rte_lcore_to_socket_id(rx_lcore_id);
933 if (socket == SOCKET_ID_ANY)
936 rxq = &qconf->rx_queue_list[qconf->n_rx_queue];
937 rxq->portid = portid;
938 rxq->direct_pool = socket_direct_pool[socket];
939 rxq->indirect_pool = socket_indirect_pool[socket];
940 rxq->lpm = socket_lpm[socket];
941 rxq->lpm6 = socket_lpm6[socket];
945 printf("Initializing port %d on lcore %u...", portid,
949 n_tx_queue = nb_lcores;
950 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
951 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
952 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
956 rte_exit(EXIT_FAILURE, "Cannot configure device: "
961 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
965 rte_exit(EXIT_FAILURE, "Cannot adjust number of "
966 "descriptors: err=%d, port=%d\n", ret, portid);
969 /* init one RX queue */
970 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
972 socket_direct_pool[socket]);
975 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: "
980 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
981 print_ethaddr(" Address:", &ports_eth_addr[portid]);
984 /* init one TX queue per couple (lcore,port) */
986 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
987 if (rte_lcore_is_enabled(lcore_id) == 0)
990 socket = (int) rte_lcore_to_socket_id(lcore_id);
991 printf("txq=%u,%d ", lcore_id, queueid);
994 txconf = &dev_info.default_txconf;
995 txconf->txq_flags = 0;
996 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
1000 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: "
1001 "err=%d, port=%d\n", ret, portid);
1004 qconf = &lcore_queue_conf[lcore_id];
1005 qconf->tx_queue_id[portid] = queueid;
1015 for (portid = 0; portid < nb_ports; portid++) {
1016 if ((enabled_port_mask & (1 << portid)) == 0) {
1020 ret = rte_eth_dev_start(portid);
1022 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
1025 rte_eth_promiscuous_enable(portid);
1027 if (check_ptype(portid) == 0) {
1028 rte_eth_add_rx_callback(portid, 0, cb_parse_ptype, NULL);
1029 printf("Add Rx callback function to detect L3 packet type by SW :"
1030 " port = %d\n", portid);
1034 if (init_routing_table() < 0)
1035 rte_exit(EXIT_FAILURE, "Cannot init routing table\n");
1037 check_all_ports_link_status(nb_ports, enabled_port_mask);
1039 /* launch per-lcore init on every lcore */
1040 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1041 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1042 if (rte_eal_wait_lcore(lcore_id) < 0)