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>
46 #include <rte_common.h>
47 #include <rte_byteorder.h>
49 #include <rte_memory.h>
50 #include <rte_memcpy.h>
51 #include <rte_memzone.h>
52 #include <rte_tailq.h>
54 #include <rte_per_lcore.h>
55 #include <rte_launch.h>
56 #include <rte_atomic.h>
57 #include <rte_cycles.h>
58 #include <rte_prefetch.h>
59 #include <rte_lcore.h>
60 #include <rte_per_lcore.h>
61 #include <rte_branch_prediction.h>
62 #include <rte_interrupts.h>
64 #include <rte_random.h>
65 #include <rte_debug.h>
66 #include <rte_ether.h>
67 #include <rte_ethdev.h>
69 #include <rte_mempool.h>
71 #include <rte_malloc.h>
75 #include <rte_string_fns.h>
78 #define APP_LOOKUP_EXACT_MATCH 0
79 #define APP_LOOKUP_LPM 1
80 #define DO_RFC_1812_CHECKS
82 #ifndef APP_LOOKUP_METHOD
83 #define APP_LOOKUP_METHOD APP_LOOKUP_LPM
86 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
88 #elif (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
92 #error "APP_LOOKUP_METHOD set to incorrect value"
95 #define MAX_PKT_BURST 32
97 #include "ipv4_rsmbl.h"
100 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\
101 "%02x%02x:%02x%02x:%02x%02x:%02x%02x"
102 #define IPv6_BYTES(addr) \
103 addr[0], addr[1], addr[2], addr[3], \
104 addr[4], addr[5], addr[6], addr[7], \
105 addr[8], addr[9], addr[10], addr[11],\
106 addr[12], addr[13],addr[14], addr[15]
110 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
112 #define MAX_PORTS RTE_MAX_ETHPORTS
114 #define MAX_JUMBO_PKT_LEN 9600
116 #define IPV6_ADDR_LEN 16
118 #define MEMPOOL_CACHE_SIZE 256
120 #define BUF_SIZE 2048
122 (BUF_SIZE + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
124 #define MAX_FLOW_NUM UINT16_MAX
125 #define MIN_FLOW_NUM 1
126 #define DEF_FLOW_NUM 0x1000
128 /* TTL numbers are in ms. */
129 #define MAX_FLOW_TTL (3600 * MS_PER_S)
130 #define MIN_FLOW_TTL 1
131 #define DEF_FLOW_TTL MS_PER_S
133 #define DEF_MBUF_NUM 0x400
135 /* Should be power of two. */
136 #define IPV4_FRAG_TBL_BUCKET_ENTRIES 2
138 static uint32_t max_flow_num = DEF_FLOW_NUM;
139 static uint32_t max_flow_ttl = DEF_FLOW_TTL;
142 * RX and TX Prefetch, Host, and Write-back threshold values should be
143 * carefully set for optimal performance. Consult the network
144 * controller's datasheet and supporting DPDK documentation for guidance
145 * on how these parameters should be set.
147 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */
148 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */
149 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */
152 * These default values are optimized for use with the Intel(R) 82599 10 GbE
153 * Controller and the DPDK ixgbe PMD. Consider using other values for other
154 * network controllers and/or network drivers.
156 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */
157 #define TX_HTHRESH 0 /**< Default values of TX host threshold reg. */
158 #define TX_WTHRESH 0 /**< Default values of TX write-back threshold reg. */
160 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
164 /* Configure how many packets ahead to prefetch, when reading packets */
165 #define PREFETCH_OFFSET 3
168 * Configurable number of RX/TX ring descriptors
170 #define RTE_TEST_RX_DESC_DEFAULT 128
171 #define RTE_TEST_TX_DESC_DEFAULT 512
173 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
174 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
176 /* ethernet addresses of ports */
177 static struct ether_addr ports_eth_addr[MAX_PORTS];
179 /* mask of enabled ports */
180 static uint32_t enabled_port_mask = 0;
181 static int promiscuous_on = 0; /**< Ports set in promiscuous mode off by default. */
182 static int numa_on = 1; /**< NUMA is enabled by default. */
188 struct rte_mbuf *m_table[0];
191 struct lcore_rx_queue {
194 } __rte_cache_aligned;
196 #define MAX_RX_QUEUE_PER_LCORE 16
197 #define MAX_TX_QUEUE_PER_PORT MAX_PORTS
198 #define MAX_RX_QUEUE_PER_PORT 128
200 #define MAX_LCORE_PARAMS 1024
201 struct lcore_params {
205 } __rte_cache_aligned;
207 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
208 static struct lcore_params lcore_params_array_default[] = {
220 static struct lcore_params * lcore_params = lcore_params_array_default;
221 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
222 sizeof(lcore_params_array_default[0]);
224 static struct rte_eth_conf port_conf = {
226 .max_rx_pkt_len = ETHER_MAX_LEN,
228 .header_split = 0, /**< Header Split disabled */
229 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
230 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
231 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
232 .hw_strip_crc = 0, /**< CRC stripped by hardware */
237 .rss_hf = ETH_RSS_IPV4 | ETH_RSS_IPV6,
241 .mq_mode = ETH_MQ_TX_NONE,
245 static const struct rte_eth_rxconf rx_conf = {
247 .pthresh = RX_PTHRESH,
248 .hthresh = RX_HTHRESH,
249 .wthresh = RX_WTHRESH,
251 .rx_free_thresh = 32,
254 static const struct rte_eth_txconf tx_conf = {
256 .pthresh = TX_PTHRESH,
257 .hthresh = TX_HTHRESH,
258 .wthresh = TX_WTHRESH,
260 .tx_free_thresh = 0, /* Use PMD default values */
261 .tx_rs_thresh = 0, /* Use PMD default values */
265 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
267 #ifdef RTE_MACHINE_CPUFLAG_SSE4_2
268 #include <rte_hash_crc.h>
269 #define DEFAULT_HASH_FUNC rte_hash_crc
271 #include <rte_jhash.h>
272 #define DEFAULT_HASH_FUNC rte_jhash
281 } __attribute__((__packed__));
284 uint8_t ip_dst[IPV6_ADDR_LEN];
285 uint8_t ip_src[IPV6_ADDR_LEN];
289 } __attribute__((__packed__));
291 struct ipv4_l3fwd_route {
292 struct ipv4_5tuple key;
296 struct ipv6_l3fwd_route {
297 struct ipv6_5tuple key;
301 static struct ipv4_l3fwd_route ipv4_l3fwd_route_array[] = {
302 {{IPv4(100,10,0,1), IPv4(200,10,0,1), 101, 11, IPPROTO_TCP}, 0},
303 {{IPv4(100,20,0,2), IPv4(200,20,0,2), 102, 12, IPPROTO_TCP}, 1},
304 {{IPv4(100,30,0,3), IPv4(200,30,0,3), 103, 13, IPPROTO_TCP}, 2},
305 {{IPv4(100,40,0,4), IPv4(200,40,0,4), 104, 14, IPPROTO_TCP}, 3},
308 static struct ipv6_l3fwd_route ipv6_l3fwd_route_array[] = {
311 {0xfe, 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
312 0x02, 0x1b, 0x21, 0xff, 0xfe, 0x91, 0x38, 0x05},
313 {0xfe, 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
314 0x02, 0x1e, 0x67, 0xff, 0xfe, 0x0d, 0xb6, 0x0a},
320 typedef struct rte_hash lookup_struct_t;
321 static lookup_struct_t *ipv4_l3fwd_lookup_struct[NB_SOCKETS];
322 static lookup_struct_t *ipv6_l3fwd_lookup_struct[NB_SOCKETS];
324 #define L3FWD_HASH_ENTRIES 1024
326 #define IPV4_L3FWD_NUM_ROUTES \
327 (sizeof(ipv4_l3fwd_route_array) / sizeof(ipv4_l3fwd_route_array[0]))
329 #define IPV6_L3FWD_NUM_ROUTES \
330 (sizeof(ipv6_l3fwd_route_array) / sizeof(ipv6_l3fwd_route_array[0]))
332 static uint8_t ipv4_l3fwd_out_if[L3FWD_HASH_ENTRIES] __rte_cache_aligned;
333 static uint8_t ipv6_l3fwd_out_if[L3FWD_HASH_ENTRIES] __rte_cache_aligned;
336 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
337 struct ipv4_l3fwd_route {
343 struct ipv6_l3fwd_route {
349 static struct ipv4_l3fwd_route ipv4_l3fwd_route_array[] = {
350 {IPv4(1,1,1,0), 24, 0},
351 {IPv4(2,1,1,0), 24, 1},
352 {IPv4(3,1,1,0), 24, 2},
353 {IPv4(4,1,1,0), 24, 3},
354 {IPv4(5,1,1,0), 24, 4},
355 {IPv4(6,1,1,0), 24, 5},
356 {IPv4(7,1,1,0), 24, 6},
357 {IPv4(8,1,1,0), 24, 7},
360 static struct ipv6_l3fwd_route ipv6_l3fwd_route_array[] = {
361 {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0},
362 {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1},
363 {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2},
364 {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3},
365 {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4},
366 {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5},
367 {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6},
368 {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7},
371 #define IPV4_L3FWD_NUM_ROUTES \
372 (sizeof(ipv4_l3fwd_route_array) / sizeof(ipv4_l3fwd_route_array[0]))
373 #define IPV6_L3FWD_NUM_ROUTES \
374 (sizeof(ipv6_l3fwd_route_array) / sizeof(ipv6_l3fwd_route_array[0]))
376 #define IPV4_L3FWD_LPM_MAX_RULES 1024
377 #define IPV6_L3FWD_LPM_MAX_RULES 1024
378 #define IPV6_L3FWD_LPM_NUMBER_TBL8S (1 << 16)
380 typedef struct rte_lpm lookup_struct_t;
381 typedef struct rte_lpm6 lookup6_struct_t;
382 static lookup_struct_t *ipv4_l3fwd_lookup_struct[NB_SOCKETS];
383 static lookup6_struct_t *ipv6_l3fwd_lookup_struct[NB_SOCKETS];
386 struct tx_lcore_stat {
393 #ifdef IPV4_FRAG_TBL_STAT
394 #define TX_LCORE_STAT_UPDATE(s, f, v) ((s)->f += (v))
396 #define TX_LCORE_STAT_UPDATE(s, f, v) do {} while (0)
397 #endif /* IPV4_FRAG_TBL_STAT */
401 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
402 uint16_t tx_queue_id[MAX_PORTS];
403 lookup_struct_t * ipv4_lookup_struct;
404 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
405 lookup6_struct_t * ipv6_lookup_struct;
407 lookup_struct_t * ipv6_lookup_struct;
409 struct ipv4_frag_tbl *frag_tbl[MAX_RX_QUEUE_PER_LCORE];
410 struct rte_mempool *pool[MAX_RX_QUEUE_PER_LCORE];
411 struct ipv4_frag_death_row death_row;
412 struct mbuf_table *tx_mbufs[MAX_PORTS];
413 struct tx_lcore_stat tx_stat;
414 } __rte_cache_aligned;
416 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
419 * If number of queued packets reached given threahold, then
420 * send burst of packets on an output interface.
422 static inline uint32_t
423 send_burst(struct lcore_conf *qconf, uint32_t thresh, uint8_t port)
425 uint32_t fill, len, k, n;
426 struct mbuf_table *txmb;
428 txmb = qconf->tx_mbufs[port];
431 if ((int32_t)(fill = txmb->head - txmb->tail) < 0)
434 if (fill >= thresh) {
435 n = RTE_MIN(len - txmb->tail, fill);
437 k = rte_eth_tx_burst(port, qconf->tx_queue_id[port],
438 txmb->m_table + txmb->tail, (uint16_t)n);
440 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, call, 1);
441 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, send, k);
444 if ((txmb->tail += k) == len)
451 /* Enqueue a single packet, and send burst if queue is filled */
453 send_single_packet(struct rte_mbuf *m, uint8_t port)
455 uint32_t fill, lcore_id, len;
456 struct lcore_conf *qconf;
457 struct mbuf_table *txmb;
459 lcore_id = rte_lcore_id();
460 qconf = &lcore_conf[lcore_id];
462 txmb = qconf->tx_mbufs[port];
465 fill = send_burst(qconf, MAX_PKT_BURST, port);
467 if (fill == len - 1) {
468 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, drop, 1);
469 rte_pktmbuf_free(txmb->m_table[txmb->tail]);
470 if (++txmb->tail == len)
474 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, queue, 1);
475 txmb->m_table[txmb->head] = m;
476 if(++txmb->head == len)
482 #ifdef DO_RFC_1812_CHECKS
484 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
486 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
488 * 1. The packet length reported by the Link Layer must be large
489 * enough to hold the minimum length legal IP datagram (20 bytes).
491 if (link_len < sizeof(struct ipv4_hdr))
494 /* 2. The IP checksum must be correct. */
495 /* this is checked in H/W */
498 * 3. The IP version number must be 4. If the version number is not 4
499 * then the packet may be another version of IP, such as IPng or
502 if (((pkt->version_ihl) >> 4) != 4)
505 * 4. The IP header length field must be large enough to hold the
506 * minimum length legal IP datagram (20 bytes = 5 words).
508 if ((pkt->version_ihl & 0xf) < 5)
512 * 5. The IP total length field must be large enough to hold the IP
513 * datagram header, whose length is specified in the IP header length
516 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
523 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
525 print_ipv4_key(struct ipv4_5tuple key)
527 printf("IP dst = %08x, IP src = %08x, port dst = %d, port src = %d, proto = %d\n",
528 (unsigned)key.ip_dst, (unsigned)key.ip_src, key.port_dst, key.port_src, key.proto);
531 print_ipv6_key(struct ipv6_5tuple key)
533 printf( "IP dst = " IPv6_BYTES_FMT ", IP src = " IPv6_BYTES_FMT ", "
534 "port dst = %d, port src = %d, proto = %d\n",
535 IPv6_BYTES(key.ip_dst), IPv6_BYTES(key.ip_src),
536 key.port_dst, key.port_src, key.proto);
539 static inline uint8_t
540 get_ipv4_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * ipv4_l3fwd_lookup_struct)
542 struct ipv4_5tuple key;
547 key.ip_dst = rte_be_to_cpu_32(ipv4_hdr->dst_addr);
548 key.ip_src = rte_be_to_cpu_32(ipv4_hdr->src_addr);
549 key.proto = ipv4_hdr->next_proto_id;
551 switch (ipv4_hdr->next_proto_id) {
553 tcp = (struct tcp_hdr *)((unsigned char *) ipv4_hdr +
554 sizeof(struct ipv4_hdr));
555 key.port_dst = rte_be_to_cpu_16(tcp->dst_port);
556 key.port_src = rte_be_to_cpu_16(tcp->src_port);
560 udp = (struct udp_hdr *)((unsigned char *) ipv4_hdr +
561 sizeof(struct ipv4_hdr));
562 key.port_dst = rte_be_to_cpu_16(udp->dst_port);
563 key.port_src = rte_be_to_cpu_16(udp->src_port);
572 /* Find destination port */
573 ret = rte_hash_lookup(ipv4_l3fwd_lookup_struct, (const void *)&key);
574 return (uint8_t)((ret < 0)? portid : ipv4_l3fwd_out_if[ret]);
577 static inline uint8_t
578 get_ipv6_dst_port(struct ipv6_hdr *ipv6_hdr, uint8_t portid, lookup_struct_t * ipv6_l3fwd_lookup_struct)
580 struct ipv6_5tuple key;
585 memcpy(key.ip_dst, ipv6_hdr->dst_addr, IPV6_ADDR_LEN);
586 memcpy(key.ip_src, ipv6_hdr->src_addr, IPV6_ADDR_LEN);
588 key.proto = ipv6_hdr->proto;
590 switch (ipv6_hdr->proto) {
592 tcp = (struct tcp_hdr *)((unsigned char *) ipv6_hdr +
593 sizeof(struct ipv6_hdr));
594 key.port_dst = rte_be_to_cpu_16(tcp->dst_port);
595 key.port_src = rte_be_to_cpu_16(tcp->src_port);
599 udp = (struct udp_hdr *)((unsigned char *) ipv6_hdr +
600 sizeof(struct ipv6_hdr));
601 key.port_dst = rte_be_to_cpu_16(udp->dst_port);
602 key.port_src = rte_be_to_cpu_16(udp->src_port);
611 /* Find destination port */
612 ret = rte_hash_lookup(ipv6_l3fwd_lookup_struct, (const void *)&key);
613 return (uint8_t)((ret < 0)? portid : ipv6_l3fwd_out_if[ret]);
617 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
618 static inline uint8_t
619 get_ipv4_dst_port(struct ipv4_hdr *ipv4_hdr, uint8_t portid, lookup_struct_t * ipv4_l3fwd_lookup_struct)
623 return (uint8_t) ((rte_lpm_lookup(ipv4_l3fwd_lookup_struct,
624 rte_be_to_cpu_32(ipv4_hdr->dst_addr), &next_hop) == 0)?
628 static inline uint8_t
629 get_ipv6_dst_port(struct ipv6_hdr *ipv6_hdr, uint8_t portid, lookup6_struct_t * ipv6_l3fwd_lookup_struct)
633 return (uint8_t) ((rte_lpm6_lookup(ipv6_l3fwd_lookup_struct,
634 ipv6_hdr->dst_addr, &next_hop) == 0)?
640 l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid, uint32_t queue,
641 struct lcore_conf *qconf, uint64_t tms)
643 struct ether_hdr *eth_hdr;
644 struct ipv4_hdr *ipv4_hdr;
647 uint16_t flag_offset, ip_flag, ip_ofs;
649 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
651 if (m->ol_flags & PKT_RX_IPV4_HDR) {
652 /* Handle IPv4 headers.*/
653 ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
655 #ifdef DO_RFC_1812_CHECKS
656 /* Check to make sure the packet is valid (RFC1812) */
657 if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt.pkt_len) < 0) {
662 /* Update time to live and header checksum */
663 --(ipv4_hdr->time_to_live);
664 ++(ipv4_hdr->hdr_checksum);
667 flag_offset = rte_be_to_cpu_16(ipv4_hdr->fragment_offset);
668 ip_ofs = (uint16_t)(flag_offset & IPV4_HDR_OFFSET_MASK);
669 ip_flag = (uint16_t)(flag_offset & IPV4_HDR_MF_FLAG);
671 /* if it is a fragmented packet, then try to reassemble. */
672 if (ip_flag != 0 || ip_ofs != 0) {
675 struct ipv4_frag_tbl *tbl;
676 struct ipv4_frag_death_row *dr;
678 tbl = qconf->frag_tbl[queue];
679 dr = &qconf->death_row;
681 /* prepare mbuf: setup l2_len/l3_len. */
682 m->pkt.vlan_macip.f.l2_len = sizeof(*eth_hdr);
683 m->pkt.vlan_macip.f.l3_len = sizeof(*ipv4_hdr);
685 /* process this fragment. */
686 if ((mo = ipv4_frag_mbuf(tbl, dr, m, tms, ipv4_hdr,
687 ip_ofs, ip_flag)) == NULL)
688 /* no packet to send out. */
691 /* we have our packet reassembled. */
694 eth_hdr = rte_pktmbuf_mtod(m,
696 ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
700 dst_port = get_ipv4_dst_port(ipv4_hdr, portid,
701 qconf->ipv4_lookup_struct);
702 if (dst_port >= MAX_PORTS ||
703 (enabled_port_mask & 1 << dst_port) == 0)
706 /* 02:00:00:00:00:xx */
707 d_addr_bytes = ð_hdr->d_addr.addr_bytes[0];
708 *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)dst_port << 40);
711 ether_addr_copy(&ports_eth_addr[dst_port], ð_hdr->s_addr);
713 send_single_packet(m, dst_port);
716 /* Handle IPv6 headers.*/
717 struct ipv6_hdr *ipv6_hdr;
719 ipv6_hdr = (struct ipv6_hdr *)(rte_pktmbuf_mtod(m, unsigned char *) +
720 sizeof(struct ether_hdr));
722 dst_port = get_ipv6_dst_port(ipv6_hdr, portid, qconf->ipv6_lookup_struct);
724 if (dst_port >= MAX_PORTS || (enabled_port_mask & 1 << dst_port) == 0)
727 /* 02:00:00:00:00:xx */
728 d_addr_bytes = ð_hdr->d_addr.addr_bytes[0];
729 *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)dst_port << 40);
732 ether_addr_copy(&ports_eth_addr[dst_port], ð_hdr->s_addr);
734 send_single_packet(m, dst_port);
739 /* main processing loop */
741 main_loop(__attribute__((unused)) void *dummy)
743 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
745 uint64_t diff_tsc, cur_tsc, prev_tsc;
747 uint8_t portid, queueid;
748 struct lcore_conf *qconf;
749 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
753 lcore_id = rte_lcore_id();
754 qconf = &lcore_conf[lcore_id];
756 if (qconf->n_rx_queue == 0) {
757 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
761 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
763 for (i = 0; i < qconf->n_rx_queue; i++) {
765 portid = qconf->rx_queue_list[i].port_id;
766 queueid = qconf->rx_queue_list[i].queue_id;
767 RTE_LOG(INFO, L3FWD, " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n", lcore_id,
773 cur_tsc = rte_rdtsc();
776 * TX burst queue drain
778 diff_tsc = cur_tsc - prev_tsc;
779 if (unlikely(diff_tsc > drain_tsc)) {
782 * This could be optimized (use queueid instead of
783 * portid), but it is not called so often
785 for (portid = 0; portid < MAX_PORTS; portid++) {
786 if ((enabled_port_mask & (1 << portid)) != 0)
787 send_burst(qconf, 1, portid);
794 * Read packet from RX queues
796 for (i = 0; i < qconf->n_rx_queue; ++i) {
798 portid = qconf->rx_queue_list[i].port_id;
799 queueid = qconf->rx_queue_list[i].queue_id;
801 nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst,
804 /* Prefetch first packets */
805 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
806 rte_prefetch0(rte_pktmbuf_mtod(
807 pkts_burst[j], void *));
810 /* Prefetch and forward already prefetched packets */
811 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
812 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
813 j + PREFETCH_OFFSET], void *));
814 l3fwd_simple_forward(pkts_burst[j], portid,
818 /* Forward remaining prefetched packets */
819 for (; j < nb_rx; j++) {
820 l3fwd_simple_forward(pkts_burst[j], portid,
824 ipv4_frag_free_death_row(&qconf->death_row,
831 check_lcore_params(void)
833 uint8_t queue, lcore;
837 for (i = 0; i < nb_lcore_params; ++i) {
838 queue = lcore_params[i].queue_id;
839 if (queue >= MAX_RX_QUEUE_PER_PORT) {
840 printf("invalid queue number: %hhu\n", queue);
843 lcore = lcore_params[i].lcore_id;
844 if (!rte_lcore_is_enabled(lcore)) {
845 printf("error: lcore %hhu is not enabled in lcore mask\n", lcore);
848 if ((socketid = rte_lcore_to_socket_id(lcore) != 0) &&
850 printf("warning: lcore %hhu is on socket %d with numa off \n",
858 check_port_config(const unsigned nb_ports)
863 for (i = 0; i < nb_lcore_params; ++i) {
864 portid = lcore_params[i].port_id;
865 if ((enabled_port_mask & (1 << portid)) == 0) {
866 printf("port %u is not enabled in port mask\n", portid);
869 if (portid >= nb_ports) {
870 printf("port %u is not present on the board\n", portid);
878 get_port_n_rx_queues(const uint8_t port)
883 for (i = 0; i < nb_lcore_params; ++i) {
884 if (lcore_params[i].port_id == port && lcore_params[i].queue_id > queue)
885 queue = lcore_params[i].queue_id;
887 return (uint8_t)(++queue);
891 init_lcore_rx_queues(void)
893 uint16_t i, nb_rx_queue;
896 for (i = 0; i < nb_lcore_params; ++i) {
897 lcore = lcore_params[i].lcore_id;
898 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
899 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
900 printf("error: too many queues (%u) for lcore: %u\n",
901 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
904 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
905 lcore_params[i].port_id;
906 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
907 lcore_params[i].queue_id;
908 lcore_conf[lcore].n_rx_queue++;
916 print_usage(const char *prgname)
918 printf ("%s [EAL options] -- -p PORTMASK -P"
919 " [--config (port,queue,lcore)[,(port,queue,lcore]]"
920 " [--enable-jumbo [--max-pkt-len PKTLEN]]"
921 " [--maxflows=<flows>] [--flowttl=<ttl>[(s|ms)]]\n"
922 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
923 " -P : enable promiscuous mode\n"
924 " --config (port,queue,lcore): rx queues configuration\n"
925 " --no-numa: optional, disable numa awareness\n"
926 " --enable-jumbo: enable jumbo frame"
927 " which max packet len is PKTLEN in decimal (64-9600)\n"
928 " --maxflows=<flows>: optional, maximum number of flows "
930 " --flowttl=<ttl>[(s|ms)]: optional, maximum TTL for each "
936 parse_flow_num(const char *str, uint32_t min, uint32_t max, uint32_t *val)
941 /* parse decimal string */
943 v = strtoul(str, &end, 10);
944 if (errno != 0 || *end != '\0')
947 if (v < min || v > max)
955 parse_flow_ttl(const char *str, uint32_t min, uint32_t max, uint32_t *val)
960 static const char frmt_sec[] = "s";
961 static const char frmt_msec[] = "ms";
963 /* parse decimal string */
965 v = strtoul(str, &end, 10);
970 if (strncmp(frmt_sec, end, sizeof(frmt_sec)) == 0)
972 else if (strncmp(frmt_msec, end, sizeof (frmt_msec)) != 0)
976 if (v < min || v > max)
984 static int parse_max_pkt_len(const char *pktlen)
989 /* parse decimal string */
990 len = strtoul(pktlen, &end, 10);
991 if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1001 parse_portmask(const char *portmask)
1006 /* parse hexadecimal string */
1007 pm = strtoul(portmask, &end, 16);
1008 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1018 parse_config(const char *q_arg)
1021 const char *p, *p0 = q_arg;
1029 unsigned long int_fld[_NUM_FLD];
1030 char *str_fld[_NUM_FLD];
1034 nb_lcore_params = 0;
1036 while ((p = strchr(p0,'(')) != NULL) {
1038 if((p0 = strchr(p,')')) == NULL)
1042 if(size >= sizeof(s))
1045 rte_snprintf(s, sizeof(s), "%.*s", size, p);
1046 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') != _NUM_FLD)
1048 for (i = 0; i < _NUM_FLD; i++){
1050 int_fld[i] = strtoul(str_fld[i], &end, 0);
1051 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1054 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1055 printf("exceeded max number of lcore params: %hu\n",
1059 lcore_params_array[nb_lcore_params].port_id = (uint8_t)int_fld[FLD_PORT];
1060 lcore_params_array[nb_lcore_params].queue_id = (uint8_t)int_fld[FLD_QUEUE];
1061 lcore_params_array[nb_lcore_params].lcore_id = (uint8_t)int_fld[FLD_LCORE];
1064 lcore_params = lcore_params_array;
1068 /* Parse the argument given in the command line of the application */
1070 parse_args(int argc, char **argv)
1075 char *prgname = argv[0];
1076 static struct option lgopts[] = {
1077 {"config", 1, 0, 0},
1078 {"no-numa", 0, 0, 0},
1079 {"enable-jumbo", 0, 0, 0},
1080 {"maxflows", 1, 0, 0},
1081 {"flowttl", 1, 0, 0},
1087 while ((opt = getopt_long(argc, argvopt, "p:P",
1088 lgopts, &option_index)) != EOF) {
1093 enabled_port_mask = parse_portmask(optarg);
1094 if (enabled_port_mask == 0) {
1095 printf("invalid portmask\n");
1096 print_usage(prgname);
1101 printf("Promiscuous mode selected\n");
1107 if (!strncmp(lgopts[option_index].name, "config", 6)) {
1108 ret = parse_config(optarg);
1110 printf("invalid config\n");
1111 print_usage(prgname);
1116 if (!strncmp(lgopts[option_index].name, "no-numa", 7)) {
1117 printf("numa is disabled \n");
1121 if (!strncmp(lgopts[option_index].name,
1123 if ((ret = parse_flow_num(optarg, MIN_FLOW_NUM,
1125 &max_flow_num)) != 0) {
1126 printf("invalid value: \"%s\" for "
1129 lgopts[option_index].name);
1130 print_usage(prgname);
1135 if (!strncmp(lgopts[option_index].name, "flowttl", 7)) {
1136 if ((ret = parse_flow_ttl(optarg, MIN_FLOW_TTL,
1138 &max_flow_ttl)) != 0) {
1139 printf("invalid value: \"%s\" for "
1142 lgopts[option_index].name);
1143 print_usage(prgname);
1148 if (!strncmp(lgopts[option_index].name, "enable-jumbo", 12)) {
1149 struct option lenopts = {"max-pkt-len", required_argument, 0, 0};
1151 printf("jumbo frame is enabled \n");
1152 port_conf.rxmode.jumbo_frame = 1;
1154 /* if no max-pkt-len set, use the default value ETHER_MAX_LEN */
1155 if (0 == getopt_long(argc, argvopt, "", &lenopts, &option_index)) {
1156 ret = parse_max_pkt_len(optarg);
1157 if ((ret < 64) || (ret > MAX_JUMBO_PKT_LEN)){
1158 printf("invalid packet length\n");
1159 print_usage(prgname);
1162 port_conf.rxmode.max_rx_pkt_len = ret;
1164 printf("set jumbo frame max packet length to %u\n",
1165 (unsigned int)port_conf.rxmode.max_rx_pkt_len);
1171 print_usage(prgname);
1177 argv[optind-1] = prgname;
1180 optind = 0; /* reset getopt lib */
1185 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
1187 printf ("%s%02X:%02X:%02X:%02X:%02X:%02X", name,
1188 eth_addr->addr_bytes[0],
1189 eth_addr->addr_bytes[1],
1190 eth_addr->addr_bytes[2],
1191 eth_addr->addr_bytes[3],
1192 eth_addr->addr_bytes[4],
1193 eth_addr->addr_bytes[5]);
1196 #if (APP_LOOKUP_METHOD == APP_LOOKUP_EXACT_MATCH)
1198 setup_hash(int socketid)
1200 struct rte_hash_parameters ipv4_l3fwd_hash_params = {
1202 .entries = L3FWD_HASH_ENTRIES,
1203 .bucket_entries = 4,
1204 .key_len = sizeof(struct ipv4_5tuple),
1205 .hash_func = DEFAULT_HASH_FUNC,
1206 .hash_func_init_val = 0,
1209 struct rte_hash_parameters ipv6_l3fwd_hash_params = {
1211 .entries = L3FWD_HASH_ENTRIES,
1212 .bucket_entries = 4,
1213 .key_len = sizeof(struct ipv6_5tuple),
1214 .hash_func = DEFAULT_HASH_FUNC,
1215 .hash_func_init_val = 0,
1222 /* create ipv4 hash */
1223 rte_snprintf(s, sizeof(s), "ipv4_l3fwd_hash_%d", socketid);
1224 ipv4_l3fwd_hash_params.name = s;
1225 ipv4_l3fwd_hash_params.socket_id = socketid;
1226 ipv4_l3fwd_lookup_struct[socketid] = rte_hash_create(&ipv4_l3fwd_hash_params);
1227 if (ipv4_l3fwd_lookup_struct[socketid] == NULL)
1228 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd hash on "
1229 "socket %d\n", socketid);
1231 /* create ipv6 hash */
1232 rte_snprintf(s, sizeof(s), "ipv6_l3fwd_hash_%d", socketid);
1233 ipv6_l3fwd_hash_params.name = s;
1234 ipv6_l3fwd_hash_params.socket_id = socketid;
1235 ipv6_l3fwd_lookup_struct[socketid] = rte_hash_create(&ipv6_l3fwd_hash_params);
1236 if (ipv6_l3fwd_lookup_struct[socketid] == NULL)
1237 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd hash on "
1238 "socket %d\n", socketid);
1241 /* populate the ipv4 hash */
1242 for (i = 0; i < IPV4_L3FWD_NUM_ROUTES; i++) {
1243 ret = rte_hash_add_key (ipv4_l3fwd_lookup_struct[socketid],
1244 (void *) &ipv4_l3fwd_route_array[i].key);
1246 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the"
1247 "l3fwd hash on socket %d\n", i, socketid);
1249 ipv4_l3fwd_out_if[ret] = ipv4_l3fwd_route_array[i].if_out;
1250 printf("Hash: Adding key\n");
1251 print_ipv4_key(ipv4_l3fwd_route_array[i].key);
1254 /* populate the ipv6 hash */
1255 for (i = 0; i < IPV6_L3FWD_NUM_ROUTES; i++) {
1256 ret = rte_hash_add_key (ipv6_l3fwd_lookup_struct[socketid],
1257 (void *) &ipv6_l3fwd_route_array[i].key);
1259 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the"
1260 "l3fwd hash on socket %d\n", i, socketid);
1262 ipv6_l3fwd_out_if[ret] = ipv6_l3fwd_route_array[i].if_out;
1263 printf("Hash: Adding key\n");
1264 print_ipv6_key(ipv6_l3fwd_route_array[i].key);
1269 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
1271 setup_lpm(int socketid)
1273 struct rte_lpm6_config config;
1278 /* create the LPM table */
1279 rte_snprintf(s, sizeof(s), "IPV4_L3FWD_LPM_%d", socketid);
1280 ipv4_l3fwd_lookup_struct[socketid] = rte_lpm_create(s, socketid,
1281 IPV4_L3FWD_LPM_MAX_RULES, 0);
1282 if (ipv4_l3fwd_lookup_struct[socketid] == NULL)
1283 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd LPM table"
1284 " on socket %d\n", socketid);
1286 /* populate the LPM table */
1287 for (i = 0; i < IPV4_L3FWD_NUM_ROUTES; i++) {
1288 ret = rte_lpm_add(ipv4_l3fwd_lookup_struct[socketid],
1289 ipv4_l3fwd_route_array[i].ip,
1290 ipv4_l3fwd_route_array[i].depth,
1291 ipv4_l3fwd_route_array[i].if_out);
1294 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the "
1295 "l3fwd LPM table on socket %d\n",
1299 printf("LPM: Adding route 0x%08x / %d (%d)\n",
1300 (unsigned)ipv4_l3fwd_route_array[i].ip,
1301 ipv4_l3fwd_route_array[i].depth,
1302 ipv4_l3fwd_route_array[i].if_out);
1305 /* create the LPM6 table */
1306 rte_snprintf(s, sizeof(s), "IPV6_L3FWD_LPM_%d", socketid);
1308 config.max_rules = IPV6_L3FWD_LPM_MAX_RULES;
1309 config.number_tbl8s = IPV6_L3FWD_LPM_NUMBER_TBL8S;
1311 ipv6_l3fwd_lookup_struct[socketid] = rte_lpm6_create(s, socketid,
1313 if (ipv6_l3fwd_lookup_struct[socketid] == NULL)
1314 rte_exit(EXIT_FAILURE, "Unable to create the l3fwd LPM table"
1315 " on socket %d\n", socketid);
1317 /* populate the LPM table */
1318 for (i = 0; i < IPV6_L3FWD_NUM_ROUTES; i++) {
1319 ret = rte_lpm6_add(ipv6_l3fwd_lookup_struct[socketid],
1320 ipv6_l3fwd_route_array[i].ip,
1321 ipv6_l3fwd_route_array[i].depth,
1322 ipv6_l3fwd_route_array[i].if_out);
1325 rte_exit(EXIT_FAILURE, "Unable to add entry %u to the "
1326 "l3fwd LPM table on socket %d\n",
1330 printf("LPM: Adding route %s / %d (%d)\n",
1332 ipv6_l3fwd_route_array[i].depth,
1333 ipv6_l3fwd_route_array[i].if_out);
1341 struct lcore_conf *qconf;
1345 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1346 if (rte_lcore_is_enabled(lcore_id) == 0)
1350 socketid = rte_lcore_to_socket_id(lcore_id);
1354 if (socketid >= NB_SOCKETS) {
1355 rte_exit(EXIT_FAILURE,
1356 "Socket %d of lcore %u is out of range %d\n",
1357 socketid, lcore_id, NB_SOCKETS);
1360 #if (APP_LOOKUP_METHOD == APP_LOOKUP_LPM)
1361 setup_lpm(socketid);
1363 setup_hash(socketid);
1365 qconf = &lcore_conf[lcore_id];
1366 qconf->ipv4_lookup_struct = ipv4_l3fwd_lookup_struct[socketid];
1367 qconf->ipv6_lookup_struct = ipv6_l3fwd_lookup_struct[socketid];
1372 /* Check the link status of all ports in up to 9s, and print them finally */
1374 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
1376 #define CHECK_INTERVAL 100 /* 100ms */
1377 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1378 uint8_t portid, count, all_ports_up, print_flag = 0;
1379 struct rte_eth_link link;
1381 printf("\nChecking link status");
1383 for (count = 0; count <= MAX_CHECK_TIME; count++) {
1385 for (portid = 0; portid < port_num; portid++) {
1386 if ((port_mask & (1 << portid)) == 0)
1388 memset(&link, 0, sizeof(link));
1389 rte_eth_link_get_nowait(portid, &link);
1390 /* print link status if flag set */
1391 if (print_flag == 1) {
1392 if (link.link_status)
1393 printf("Port %d Link Up - speed %u "
1394 "Mbps - %s\n", (uint8_t)portid,
1395 (unsigned)link.link_speed,
1396 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1397 ("full-duplex") : ("half-duplex\n"));
1399 printf("Port %d Link Down\n",
1403 /* clear all_ports_up flag if any link down */
1404 if (link.link_status == 0) {
1409 /* after finally printing all link status, get out */
1410 if (print_flag == 1)
1413 if (all_ports_up == 0) {
1416 rte_delay_ms(CHECK_INTERVAL);
1419 /* set the print_flag if all ports up or timeout */
1420 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1427 setup_port_tbl(struct lcore_conf *qconf, uint32_t lcore, int socket,
1430 struct mbuf_table *mtb;
1434 n = RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST);
1435 sz = sizeof (*mtb) + sizeof (mtb->m_table[0]) * n;
1437 if ((mtb = rte_zmalloc_socket(__func__, sz, CACHE_LINE_SIZE,
1439 rte_exit(EXIT_FAILURE, "%s() for lcore: %u, port: %u "
1440 "failed to allocate %zu bytes\n",
1441 __func__, lcore, port, sz);
1444 qconf->tx_mbufs[port] = mtb;
1448 setup_queue_tbl(struct lcore_conf *qconf, uint32_t lcore, int socket,
1452 uint64_t frag_cycles;
1453 char buf[RTE_MEMPOOL_NAMESIZE];
1455 frag_cycles = (rte_get_tsc_hz() + MS_PER_S - 1) / MS_PER_S *
1458 if ((qconf->frag_tbl[queue] = ipv4_frag_tbl_create(max_flow_num,
1459 IPV4_FRAG_TBL_BUCKET_ENTRIES, max_flow_num, frag_cycles,
1461 rte_exit(EXIT_FAILURE, "ipv4_frag_tbl_create(%u) on "
1462 "lcore: %u for queue: %u failed\n",
1463 max_flow_num, lcore, queue);
1466 * At any given moment up to <max_flow_num * (MAX_FRAG_NUM - 1)>
1467 * mbufs could be stored int the fragment table.
1468 * Plus, each TX queue can hold up to <max_flow_num> packets.
1471 nb_mbuf = 2 * RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST) * MAX_FRAG_NUM;
1472 nb_mbuf *= (port_conf.rxmode.max_rx_pkt_len + BUF_SIZE - 1) / BUF_SIZE;
1473 nb_mbuf += RTE_TEST_RX_DESC_DEFAULT + RTE_TEST_TX_DESC_DEFAULT;
1475 nb_mbuf = RTE_MAX(nb_mbuf, (uint32_t)DEF_MBUF_NUM);
1477 rte_snprintf(buf, sizeof(buf), "mbuf_pool_%u_%u", lcore, queue);
1479 if ((qconf->pool[queue] = rte_mempool_create(buf, nb_mbuf, MBUF_SIZE, 0,
1480 sizeof(struct rte_pktmbuf_pool_private),
1481 rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
1482 socket, MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET)) == NULL)
1483 rte_exit(EXIT_FAILURE, "mempool_create(%s) failed", buf);
1487 queue_dump_stat(void)
1490 const struct lcore_conf *qconf;
1492 for (lcore = 0; lcore < RTE_MAX_LCORE; lcore++) {
1493 if (rte_lcore_is_enabled(lcore) == 0)
1496 qconf = lcore_conf + lcore;
1497 for (i = 0; i < qconf->n_rx_queue; i++) {
1499 fprintf(stdout, " -- lcoreid=%u portid=%hhu "
1500 "rxqueueid=%hhu frag tbl stat:\n",
1501 lcore, qconf->rx_queue_list[i].port_id,
1502 qconf->rx_queue_list[i].queue_id);
1503 ipv4_frag_tbl_dump_stat(stdout, qconf->frag_tbl[i]);
1504 fprintf(stdout, "TX bursts:\t%" PRIu64 "\n"
1505 "TX packets _queued:\t%" PRIu64 "\n"
1506 "TX packets dropped:\t%" PRIu64 "\n"
1507 "TX packets send:\t%" PRIu64 "\n",
1508 qconf->tx_stat.call,
1509 qconf->tx_stat.queue,
1510 qconf->tx_stat.drop,
1511 qconf->tx_stat.send);
1517 signal_handler(int signum)
1520 if (signum != SIGUSR1)
1521 rte_exit(0, "received signal: %d, exiting\n", signum);
1525 MAIN(int argc, char **argv)
1527 struct lcore_conf *qconf;
1532 uint32_t n_tx_queue, nb_lcores;
1533 uint8_t portid, nb_rx_queue, queue, socketid;
1536 ret = rte_eal_init(argc, argv);
1538 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1542 /* parse application arguments (after the EAL ones) */
1543 ret = parse_args(argc, argv);
1545 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
1547 if (check_lcore_params() < 0)
1548 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1550 ret = init_lcore_rx_queues();
1552 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1555 /* init driver(s) */
1556 if (rte_pmd_init_all() < 0)
1557 rte_exit(EXIT_FAILURE, "Cannot init pmd\n");
1559 if (rte_eal_pci_probe() < 0)
1560 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n");
1562 nb_ports = rte_eth_dev_count();
1563 if (nb_ports > MAX_PORTS)
1564 nb_ports = MAX_PORTS;
1566 if (check_port_config(nb_ports) < 0)
1567 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1569 nb_lcores = rte_lcore_count();
1571 /* initialize all ports */
1572 for (portid = 0; portid < nb_ports; portid++) {
1573 /* skip ports that are not enabled */
1574 if ((enabled_port_mask & (1 << portid)) == 0) {
1575 printf("\nSkipping disabled port %d\n", portid);
1580 printf("Initializing port %d ... ", portid );
1583 nb_rx_queue = get_port_n_rx_queues(portid);
1584 n_tx_queue = nb_lcores;
1585 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1586 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1587 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1588 nb_rx_queue, (unsigned)n_tx_queue );
1589 ret = rte_eth_dev_configure(portid, nb_rx_queue,
1590 (uint16_t)n_tx_queue, &port_conf);
1592 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n",
1595 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1596 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1602 rte_exit(EXIT_FAILURE, "init_mem failed\n");
1604 /* init one TX queue per couple (lcore,port) */
1606 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1607 if (rte_lcore_is_enabled(lcore_id) == 0)
1611 socketid = (uint8_t)rte_lcore_to_socket_id(lcore_id);
1615 printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
1617 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
1618 socketid, &tx_conf);
1620 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
1621 "port=%d\n", ret, portid);
1623 qconf = &lcore_conf[lcore_id];
1624 qconf->tx_queue_id[portid] = queueid;
1625 setup_port_tbl(qconf, lcore_id, socketid, portid);
1631 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1632 if (rte_lcore_is_enabled(lcore_id) == 0)
1634 qconf = &lcore_conf[lcore_id];
1635 printf("\nInitializing rx queues on lcore %u ... ", lcore_id );
1637 /* init RX queues */
1638 for(queue = 0; queue < qconf->n_rx_queue; ++queue) {
1639 portid = qconf->rx_queue_list[queue].port_id;
1640 queueid = qconf->rx_queue_list[queue].queue_id;
1643 socketid = (uint8_t)rte_lcore_to_socket_id(lcore_id);
1647 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
1650 setup_queue_tbl(qconf, lcore_id, socketid, queue);
1652 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
1653 socketid, &rx_conf, qconf->pool[queue]);
1655 rte_exit(EXIT_FAILURE,
1656 "rte_eth_rx_queue_setup: err=%d,"
1657 "port=%d\n", ret, portid);
1664 for (portid = 0; portid < nb_ports; portid++) {
1665 if ((enabled_port_mask & (1 << portid)) == 0) {
1669 ret = rte_eth_dev_start(portid);
1671 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
1675 * If enabled, put device in promiscuous mode.
1676 * This allows IO forwarding mode to forward packets
1677 * to itself through 2 cross-connected ports of the
1681 rte_eth_promiscuous_enable(portid);
1684 check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
1686 signal(SIGUSR1, signal_handler);
1687 signal(SIGTERM, signal_handler);
1688 signal(SIGINT, signal_handler);
1690 /* launch per-lcore init on every lcore */
1691 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1692 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1693 if (rte_eal_wait_lcore(lcore_id) < 0)