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>
73 #include <rte_string_fns.h>
76 #define DO_RFC_1812_CHECKS
78 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
80 #define MAX_JUMBO_PKT_LEN 9600
82 #define MEMPOOL_CACHE_SIZE 256
84 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
87 * This expression is used to calculate the number of mbufs needed
88 * depending on user input, taking into account memory for rx and tx hardware
89 * rings, cache per lcore and mtable per port per lcore.
90 * RTE_MAX is used to ensure that NB_MBUF never goes below a
91 * minimum value of 8192
94 #define NB_MBUF RTE_MAX(\
95 (nb_ports * nb_rx_queue*RTE_TEST_RX_DESC_DEFAULT + \
96 nb_ports * nb_lcores * MAX_PKT_BURST + \
97 nb_ports * n_tx_queue * RTE_TEST_TX_DESC_DEFAULT + \
98 nb_lcores * MEMPOOL_CACHE_SIZE), \
101 #define MAX_PKT_BURST 32
102 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
106 /* Configure how many packets ahead to prefetch, when reading packets */
107 #define PREFETCH_OFFSET 3
110 * Configurable number of RX/TX ring descriptors
112 #define RTE_TEST_RX_DESC_DEFAULT 128
113 #define RTE_TEST_TX_DESC_DEFAULT 512
114 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
115 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
117 /* ethernet addresses of ports */
118 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
120 /* mask of enabled ports */
121 static uint32_t enabled_port_mask;
122 static int promiscuous_on; /**< Ports set in promiscuous mode off by default. */
123 static int numa_on = 1; /**< NUMA is enabled by default. */
127 struct rte_mbuf *m_table[MAX_PKT_BURST];
130 struct lcore_rx_queue {
133 } __rte_cache_aligned;
135 #define MAX_RX_QUEUE_PER_LCORE 16
136 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
137 #define MAX_RX_QUEUE_PER_PORT 128
139 #define MAX_LCORE_PARAMS 1024
140 struct lcore_params {
144 } __rte_cache_aligned;
146 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
147 static struct lcore_params lcore_params_array_default[] = {
159 static struct lcore_params *lcore_params = lcore_params_array_default;
160 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
161 sizeof(lcore_params_array_default[0]);
163 static struct rte_eth_conf port_conf = {
165 .mq_mode = ETH_MQ_RX_RSS,
166 .max_rx_pkt_len = ETHER_MAX_LEN,
168 .header_split = 0, /**< Header Split disabled */
169 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
170 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
171 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
172 .hw_strip_crc = 0, /**< CRC stripped by hardware */
177 .rss_hf = ETH_RSS_IPV4 | ETH_RSS_IPV4_TCP
179 | ETH_RSS_IPV6 | ETH_RSS_IPV6_EX
180 | ETH_RSS_IPV6_TCP | ETH_RSS_IPV6_TCP_EX
181 | ETH_RSS_IPV6_UDP | ETH_RSS_IPV6_UDP_EX,
185 .mq_mode = ETH_MQ_TX_NONE,
189 static struct rte_mempool *pktmbuf_pool[NB_SOCKETS];
191 /***********************start of ACL part******************************/
192 #ifdef DO_RFC_1812_CHECKS
194 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len);
197 send_single_packet(struct rte_mbuf *m, uint8_t port);
199 #define MAX_ACL_RULE_NUM 100000
200 #define DEFAULT_MAX_CATEGORIES 1
201 #define L3FWD_ACL_IPV4_NAME "l3fwd-acl-ipv4"
202 #define L3FWD_ACL_IPV6_NAME "l3fwd-acl-ipv6"
203 #define ACL_LEAD_CHAR ('@')
204 #define ROUTE_LEAD_CHAR ('R')
205 #define COMMENT_LEAD_CHAR ('#')
206 #define OPTION_CONFIG "config"
207 #define OPTION_NONUMA "no-numa"
208 #define OPTION_ENBJMO "enable-jumbo"
209 #define OPTION_RULE_IPV4 "rule_ipv4"
210 #define OPTION_RULE_IPV6 "rule_ipv6"
211 #define OPTION_SCALAR "scalar"
212 #define ACL_DENY_SIGNATURE 0xf0000000
213 #define RTE_LOGTYPE_L3FWDACL RTE_LOGTYPE_USER3
214 #define acl_log(format, ...) RTE_LOG(ERR, L3FWDACL, format, ##__VA_ARGS__)
215 #define uint32_t_to_char(ip, a, b, c, d) do {\
216 *a = (unsigned char)(ip >> 24 & 0xff);\
217 *b = (unsigned char)(ip >> 16 & 0xff);\
218 *c = (unsigned char)(ip >> 8 & 0xff);\
219 *d = (unsigned char)(ip & 0xff);\
221 #define OFF_ETHHEAD (sizeof(struct ether_hdr))
222 #define OFF_IPV42PROTO (offsetof(struct ipv4_hdr, next_proto_id))
223 #define OFF_IPV62PROTO (offsetof(struct ipv6_hdr, proto))
224 #define MBUF_IPV4_2PROTO(m) \
225 (rte_pktmbuf_mtod((m), uint8_t *) + OFF_ETHHEAD + OFF_IPV42PROTO)
226 #define MBUF_IPV6_2PROTO(m) \
227 (rte_pktmbuf_mtod((m), uint8_t *) + OFF_ETHHEAD + OFF_IPV62PROTO)
229 #define GET_CB_FIELD(in, fd, base, lim, dlm) do { \
233 val = strtoul((in), &end, (base)); \
234 if (errno != 0 || end[0] != (dlm) || val > (lim)) \
236 (fd) = (typeof(fd))val; \
241 * ACL rules should have higher priorities than route ones to ensure ACL rule
242 * always be found when input packets have multi-matches in the database.
243 * A exception case is performance measure, which can define route rules with
244 * higher priority and route rules will always be returned in each lookup.
245 * Reserve range from ACL_RULE_PRIORITY_MAX + 1 to
246 * RTE_ACL_MAX_PRIORITY for route entries in performance measure
248 #define ACL_RULE_PRIORITY_MAX 0x10000000
251 * Forward port info save in ACL lib starts from 1
252 * since ACL assume 0 is invalid.
253 * So, need add 1 when saving and minus 1 when forwarding packets.
255 #define FWD_PORT_SHIFT 1
258 * Rule and trace formats definitions.
270 struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
272 .type = RTE_ACL_FIELD_TYPE_BITMASK,
273 .size = sizeof(uint8_t),
274 .field_index = PROTO_FIELD_IPV4,
275 .input_index = RTE_ACL_IPV4VLAN_PROTO,
279 .type = RTE_ACL_FIELD_TYPE_MASK,
280 .size = sizeof(uint32_t),
281 .field_index = SRC_FIELD_IPV4,
282 .input_index = RTE_ACL_IPV4VLAN_SRC,
283 .offset = offsetof(struct ipv4_hdr, src_addr) -
284 offsetof(struct ipv4_hdr, next_proto_id),
287 .type = RTE_ACL_FIELD_TYPE_MASK,
288 .size = sizeof(uint32_t),
289 .field_index = DST_FIELD_IPV4,
290 .input_index = RTE_ACL_IPV4VLAN_DST,
291 .offset = offsetof(struct ipv4_hdr, dst_addr) -
292 offsetof(struct ipv4_hdr, next_proto_id),
295 .type = RTE_ACL_FIELD_TYPE_RANGE,
296 .size = sizeof(uint16_t),
297 .field_index = SRCP_FIELD_IPV4,
298 .input_index = RTE_ACL_IPV4VLAN_PORTS,
299 .offset = sizeof(struct ipv4_hdr) -
300 offsetof(struct ipv4_hdr, next_proto_id),
303 .type = RTE_ACL_FIELD_TYPE_RANGE,
304 .size = sizeof(uint16_t),
305 .field_index = DSTP_FIELD_IPV4,
306 .input_index = RTE_ACL_IPV4VLAN_PORTS,
307 .offset = sizeof(struct ipv4_hdr) -
308 offsetof(struct ipv4_hdr, next_proto_id) +
313 #define IPV6_ADDR_LEN 16
314 #define IPV6_ADDR_U16 (IPV6_ADDR_LEN / sizeof(uint16_t))
315 #define IPV6_ADDR_U32 (IPV6_ADDR_LEN / sizeof(uint32_t))
332 struct rte_acl_field_def ipv6_defs[NUM_FIELDS_IPV6] = {
334 .type = RTE_ACL_FIELD_TYPE_BITMASK,
335 .size = sizeof(uint8_t),
336 .field_index = PROTO_FIELD_IPV6,
337 .input_index = PROTO_FIELD_IPV6,
341 .type = RTE_ACL_FIELD_TYPE_MASK,
342 .size = sizeof(uint32_t),
343 .field_index = SRC1_FIELD_IPV6,
344 .input_index = SRC1_FIELD_IPV6,
345 .offset = offsetof(struct ipv6_hdr, src_addr) -
346 offsetof(struct ipv6_hdr, proto),
349 .type = RTE_ACL_FIELD_TYPE_MASK,
350 .size = sizeof(uint32_t),
351 .field_index = SRC2_FIELD_IPV6,
352 .input_index = SRC2_FIELD_IPV6,
353 .offset = offsetof(struct ipv6_hdr, src_addr) -
354 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
357 .type = RTE_ACL_FIELD_TYPE_MASK,
358 .size = sizeof(uint32_t),
359 .field_index = SRC3_FIELD_IPV6,
360 .input_index = SRC3_FIELD_IPV6,
361 .offset = offsetof(struct ipv6_hdr, src_addr) -
362 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
365 .type = RTE_ACL_FIELD_TYPE_MASK,
366 .size = sizeof(uint32_t),
367 .field_index = SRC4_FIELD_IPV6,
368 .input_index = SRC4_FIELD_IPV6,
369 .offset = offsetof(struct ipv6_hdr, src_addr) -
370 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
373 .type = RTE_ACL_FIELD_TYPE_MASK,
374 .size = sizeof(uint32_t),
375 .field_index = DST1_FIELD_IPV6,
376 .input_index = DST1_FIELD_IPV6,
377 .offset = offsetof(struct ipv6_hdr, dst_addr)
378 - offsetof(struct ipv6_hdr, proto),
381 .type = RTE_ACL_FIELD_TYPE_MASK,
382 .size = sizeof(uint32_t),
383 .field_index = DST2_FIELD_IPV6,
384 .input_index = DST2_FIELD_IPV6,
385 .offset = offsetof(struct ipv6_hdr, dst_addr) -
386 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
389 .type = RTE_ACL_FIELD_TYPE_MASK,
390 .size = sizeof(uint32_t),
391 .field_index = DST3_FIELD_IPV6,
392 .input_index = DST3_FIELD_IPV6,
393 .offset = offsetof(struct ipv6_hdr, dst_addr) -
394 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
397 .type = RTE_ACL_FIELD_TYPE_MASK,
398 .size = sizeof(uint32_t),
399 .field_index = DST4_FIELD_IPV6,
400 .input_index = DST4_FIELD_IPV6,
401 .offset = offsetof(struct ipv6_hdr, dst_addr) -
402 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
405 .type = RTE_ACL_FIELD_TYPE_RANGE,
406 .size = sizeof(uint16_t),
407 .field_index = SRCP_FIELD_IPV6,
408 .input_index = SRCP_FIELD_IPV6,
409 .offset = sizeof(struct ipv6_hdr) -
410 offsetof(struct ipv6_hdr, proto),
413 .type = RTE_ACL_FIELD_TYPE_RANGE,
414 .size = sizeof(uint16_t),
415 .field_index = DSTP_FIELD_IPV6,
416 .input_index = SRCP_FIELD_IPV6,
417 .offset = sizeof(struct ipv6_hdr) -
418 offsetof(struct ipv6_hdr, proto) + sizeof(uint16_t),
427 CB_FLD_SRC_PORT_HIGH,
430 CB_FLD_DST_PORT_HIGH,
436 RTE_ACL_RULE_DEF(acl4_rule, RTE_DIM(ipv4_defs));
437 RTE_ACL_RULE_DEF(acl6_rule, RTE_DIM(ipv6_defs));
439 struct acl_search_t {
440 const uint8_t *data_ipv4[MAX_PKT_BURST];
441 struct rte_mbuf *m_ipv4[MAX_PKT_BURST];
442 uint32_t res_ipv4[MAX_PKT_BURST];
445 const uint8_t *data_ipv6[MAX_PKT_BURST];
446 struct rte_mbuf *m_ipv6[MAX_PKT_BURST];
447 uint32_t res_ipv6[MAX_PKT_BURST];
452 char mapped[NB_SOCKETS];
453 struct rte_acl_ctx *acx_ipv4[NB_SOCKETS];
454 struct rte_acl_ctx *acx_ipv6[NB_SOCKETS];
455 #ifdef L3FWDACL_DEBUG
456 struct acl4_rule *rule_ipv4;
457 struct acl6_rule *rule_ipv6;
462 const char *rule_ipv4_name;
463 const char *rule_ipv6_name;
467 const char cb_port_delim[] = ":";
470 print_one_ipv4_rule(struct acl4_rule *rule, int extra)
472 unsigned char a, b, c, d;
474 uint32_t_to_char(rule->field[SRC_FIELD_IPV4].value.u32,
476 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
477 rule->field[SRC_FIELD_IPV4].mask_range.u32);
478 uint32_t_to_char(rule->field[DST_FIELD_IPV4].value.u32,
480 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
481 rule->field[DST_FIELD_IPV4].mask_range.u32);
482 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
483 rule->field[SRCP_FIELD_IPV4].value.u16,
484 rule->field[SRCP_FIELD_IPV4].mask_range.u16,
485 rule->field[DSTP_FIELD_IPV4].value.u16,
486 rule->field[DSTP_FIELD_IPV4].mask_range.u16,
487 rule->field[PROTO_FIELD_IPV4].value.u8,
488 rule->field[PROTO_FIELD_IPV4].mask_range.u8);
490 printf("0x%x-0x%x-0x%x ",
491 rule->data.category_mask,
493 rule->data.userdata);
497 print_one_ipv6_rule(struct acl6_rule *rule, int extra)
499 unsigned char a, b, c, d;
501 uint32_t_to_char(rule->field[SRC1_FIELD_IPV6].value.u32,
503 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
504 uint32_t_to_char(rule->field[SRC2_FIELD_IPV6].value.u32,
506 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
507 uint32_t_to_char(rule->field[SRC3_FIELD_IPV6].value.u32,
509 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
510 uint32_t_to_char(rule->field[SRC4_FIELD_IPV6].value.u32,
512 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
513 rule->field[SRC1_FIELD_IPV6].mask_range.u32
514 + rule->field[SRC2_FIELD_IPV6].mask_range.u32
515 + rule->field[SRC3_FIELD_IPV6].mask_range.u32
516 + rule->field[SRC4_FIELD_IPV6].mask_range.u32);
518 uint32_t_to_char(rule->field[DST1_FIELD_IPV6].value.u32,
520 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
521 uint32_t_to_char(rule->field[DST2_FIELD_IPV6].value.u32,
523 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
524 uint32_t_to_char(rule->field[DST3_FIELD_IPV6].value.u32,
526 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
527 uint32_t_to_char(rule->field[DST4_FIELD_IPV6].value.u32,
529 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
530 rule->field[DST1_FIELD_IPV6].mask_range.u32
531 + rule->field[DST2_FIELD_IPV6].mask_range.u32
532 + rule->field[DST3_FIELD_IPV6].mask_range.u32
533 + rule->field[DST4_FIELD_IPV6].mask_range.u32);
535 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
536 rule->field[SRCP_FIELD_IPV6].value.u16,
537 rule->field[SRCP_FIELD_IPV6].mask_range.u16,
538 rule->field[DSTP_FIELD_IPV6].value.u16,
539 rule->field[DSTP_FIELD_IPV6].mask_range.u16,
540 rule->field[PROTO_FIELD_IPV6].value.u8,
541 rule->field[PROTO_FIELD_IPV6].mask_range.u8);
543 printf("0x%x-0x%x-0x%x ",
544 rule->data.category_mask,
546 rule->data.userdata);
549 /* Bypass comment and empty lines */
551 is_bypass_line(char *buff)
556 if (buff[0] == COMMENT_LEAD_CHAR)
559 while (buff[i] != '\0') {
560 if (!isspace(buff[i]))
567 #ifdef L3FWDACL_DEBUG
569 dump_acl4_rule(struct rte_mbuf *m, uint32_t sig)
571 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
572 unsigned char a, b, c, d;
573 struct ipv4_hdr *ipv4_hdr = (struct ipv4_hdr *)
574 (rte_pktmbuf_mtod(m, unsigned char *) +
575 sizeof(struct ether_hdr));
577 uint32_t_to_char(rte_bswap32(ipv4_hdr->src_addr), &a, &b, &c, &d);
578 printf("Packet Src:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
579 uint32_t_to_char(rte_bswap32(ipv4_hdr->dst_addr), &a, &b, &c, &d);
580 printf("Dst:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
582 printf("Src port:%hu,Dst port:%hu ",
583 rte_bswap16(*(uint16_t *)(ipv4_hdr + 1)),
584 rte_bswap16(*((uint16_t *)(ipv4_hdr + 1) + 1)));
585 printf("hit ACL %d - ", offset);
587 print_one_ipv4_rule(acl_config.rule_ipv4 + offset, 1);
593 dump_acl6_rule(struct rte_mbuf *m, uint32_t sig)
596 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
597 struct ipv6_hdr *ipv6_hdr = (struct ipv6_hdr *)
598 (rte_pktmbuf_mtod(m, unsigned char *) +
599 sizeof(struct ether_hdr));
601 printf("Packet Src");
602 for (i = 0; i < RTE_DIM(ipv6_hdr->src_addr); i += sizeof(uint16_t))
604 ipv6_hdr->src_addr[i], ipv6_hdr->src_addr[i + 1]);
607 for (i = 0; i < RTE_DIM(ipv6_hdr->dst_addr); i += sizeof(uint16_t))
609 ipv6_hdr->dst_addr[i], ipv6_hdr->dst_addr[i + 1]);
611 printf("\nSrc port:%hu,Dst port:%hu ",
612 rte_bswap16(*(uint16_t *)(ipv6_hdr + 1)),
613 rte_bswap16(*((uint16_t *)(ipv6_hdr + 1) + 1)));
614 printf("hit ACL %d - ", offset);
616 print_one_ipv6_rule(acl_config.rule_ipv6 + offset, 1);
620 #endif /* L3FWDACL_DEBUG */
623 dump_ipv4_rules(struct acl4_rule *rule, int num, int extra)
627 for (i = 0; i < num; i++, rule++) {
628 printf("\t%d:", i + 1);
629 print_one_ipv4_rule(rule, extra);
635 dump_ipv6_rules(struct acl6_rule *rule, int num, int extra)
639 for (i = 0; i < num; i++, rule++) {
640 printf("\t%d:", i + 1);
641 print_one_ipv6_rule(rule, extra);
646 #ifdef DO_RFC_1812_CHECKS
648 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
651 struct ipv4_hdr *ipv4_hdr;
652 struct rte_mbuf *pkt = pkts_in[index];
654 int type = pkt->ol_flags & (PKT_RX_IPV4_HDR | PKT_RX_IPV6_HDR);
656 if (type == PKT_RX_IPV4_HDR) {
658 ipv4_hdr = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt,
659 unsigned char *) + sizeof(struct ether_hdr));
661 /* Check to make sure the packet is valid (RFC1812) */
662 if (is_valid_ipv4_pkt(ipv4_hdr, pkt->pkt_len) >= 0) {
664 /* Update time to live and header checksum */
665 --(ipv4_hdr->time_to_live);
666 ++(ipv4_hdr->hdr_checksum);
668 /* Fill acl structure */
669 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
670 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
673 /* Not a valid IPv4 packet */
674 rte_pktmbuf_free(pkt);
677 } else if (type == PKT_RX_IPV6_HDR) {
679 /* Fill acl structure */
680 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
681 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
684 /* Unknown type, drop the packet */
685 rte_pktmbuf_free(pkt);
691 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
694 struct rte_mbuf *pkt = pkts_in[index];
696 int type = pkt->ol_flags & (PKT_RX_IPV4_HDR | PKT_RX_IPV6_HDR);
698 if (type == PKT_RX_IPV4_HDR) {
700 /* Fill acl structure */
701 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
702 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
705 } else if (type == PKT_RX_IPV6_HDR) {
707 /* Fill acl structure */
708 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
709 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
711 /* Unknown type, drop the packet */
712 rte_pktmbuf_free(pkt);
715 #endif /* DO_RFC_1812_CHECKS */
718 prepare_acl_parameter(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
726 /* Prefetch first packets */
727 for (i = 0; i < PREFETCH_OFFSET && i < nb_rx; i++) {
728 rte_prefetch0(rte_pktmbuf_mtod(
729 pkts_in[i], void *));
732 for (i = 0; i < (nb_rx - PREFETCH_OFFSET); i++) {
733 rte_prefetch0(rte_pktmbuf_mtod(pkts_in[
734 i + PREFETCH_OFFSET], void *));
735 prepare_one_packet(pkts_in, acl, i);
738 /* Process left packets */
739 for (; i < nb_rx; i++)
740 prepare_one_packet(pkts_in, acl, i);
744 send_one_packet(struct rte_mbuf *m, uint32_t res)
746 if (likely((res & ACL_DENY_SIGNATURE) == 0 && res != 0)) {
747 /* forward packets */
748 send_single_packet(m,
749 (uint8_t)(res - FWD_PORT_SHIFT));
751 /* in the ACL list, drop it */
752 #ifdef L3FWDACL_DEBUG
753 if ((res & ACL_DENY_SIGNATURE) != 0) {
754 if (m->ol_flags & PKT_RX_IPV4_HDR)
755 dump_acl4_rule(m, res);
757 dump_acl6_rule(m, res);
767 send_packets(struct rte_mbuf **m, uint32_t *res, int num)
771 /* Prefetch first packets */
772 for (i = 0; i < PREFETCH_OFFSET && i < num; i++) {
773 rte_prefetch0(rte_pktmbuf_mtod(
777 for (i = 0; i < (num - PREFETCH_OFFSET); i++) {
778 rte_prefetch0(rte_pktmbuf_mtod(m[
779 i + PREFETCH_OFFSET], void *));
780 send_one_packet(m[i], res[i]);
783 /* Process left packets */
785 send_one_packet(m[i], res[i]);
789 * Parses IPV6 address, exepcts the following format:
790 * XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX (where X - is a hexedecimal digit).
793 parse_ipv6_addr(const char *in, const char **end, uint32_t v[IPV6_ADDR_U32],
796 uint32_t addr[IPV6_ADDR_U16];
798 GET_CB_FIELD(in, addr[0], 16, UINT16_MAX, ':');
799 GET_CB_FIELD(in, addr[1], 16, UINT16_MAX, ':');
800 GET_CB_FIELD(in, addr[2], 16, UINT16_MAX, ':');
801 GET_CB_FIELD(in, addr[3], 16, UINT16_MAX, ':');
802 GET_CB_FIELD(in, addr[4], 16, UINT16_MAX, ':');
803 GET_CB_FIELD(in, addr[5], 16, UINT16_MAX, ':');
804 GET_CB_FIELD(in, addr[6], 16, UINT16_MAX, ':');
805 GET_CB_FIELD(in, addr[7], 16, UINT16_MAX, dlm);
809 v[0] = (addr[0] << 16) + addr[1];
810 v[1] = (addr[2] << 16) + addr[3];
811 v[2] = (addr[4] << 16) + addr[5];
812 v[3] = (addr[6] << 16) + addr[7];
818 parse_ipv6_net(const char *in, struct rte_acl_field field[4])
823 const uint32_t nbu32 = sizeof(uint32_t) * CHAR_BIT;
826 rc = parse_ipv6_addr(in, &mp, v, '/');
831 GET_CB_FIELD(mp, m, 0, CHAR_BIT * sizeof(v), 0);
833 /* put all together. */
834 for (i = 0; i != RTE_DIM(v); i++) {
835 if (m >= (i + 1) * nbu32)
836 field[i].mask_range.u32 = nbu32;
838 field[i].mask_range.u32 = m > (i * nbu32) ?
841 field[i].value.u32 = v[i];
848 parse_cb_ipv6_rule(char *str, struct rte_acl_rule *v, int has_userdata)
851 char *s, *sp, *in[CB_FLD_NUM];
852 static const char *dlm = " \t\n";
853 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
856 for (i = 0; i != dim; i++, s = NULL) {
857 in[i] = strtok_r(s, dlm, &sp);
862 rc = parse_ipv6_net(in[CB_FLD_SRC_ADDR], v->field + SRC1_FIELD_IPV6);
864 acl_log("failed to read source address/mask: %s\n",
865 in[CB_FLD_SRC_ADDR]);
869 rc = parse_ipv6_net(in[CB_FLD_DST_ADDR], v->field + DST1_FIELD_IPV6);
871 acl_log("failed to read destination address/mask: %s\n",
872 in[CB_FLD_DST_ADDR]);
877 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
878 v->field[SRCP_FIELD_IPV6].value.u16,
880 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
881 v->field[SRCP_FIELD_IPV6].mask_range.u16,
884 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
885 sizeof(cb_port_delim)) != 0)
888 /* destination port. */
889 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
890 v->field[DSTP_FIELD_IPV6].value.u16,
892 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
893 v->field[DSTP_FIELD_IPV6].mask_range.u16,
896 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
897 sizeof(cb_port_delim)) != 0)
900 if (v->field[SRCP_FIELD_IPV6].mask_range.u16
901 < v->field[SRCP_FIELD_IPV6].value.u16
902 || v->field[DSTP_FIELD_IPV6].mask_range.u16
903 < v->field[DSTP_FIELD_IPV6].value.u16)
906 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].value.u8,
908 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].mask_range.u8,
912 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata,
919 * Parse ClassBench rules file.
921 * '@'<src_ipv4_addr>'/'<masklen> <space> \
922 * <dst_ipv4_addr>'/'<masklen> <space> \
923 * <src_port_low> <space> ":" <src_port_high> <space> \
924 * <dst_port_low> <space> ":" <dst_port_high> <space> \
928 parse_ipv4_net(const char *in, uint32_t *addr, uint32_t *mask_len)
930 uint8_t a, b, c, d, m;
932 GET_CB_FIELD(in, a, 0, UINT8_MAX, '.');
933 GET_CB_FIELD(in, b, 0, UINT8_MAX, '.');
934 GET_CB_FIELD(in, c, 0, UINT8_MAX, '.');
935 GET_CB_FIELD(in, d, 0, UINT8_MAX, '/');
936 GET_CB_FIELD(in, m, 0, sizeof(uint32_t) * CHAR_BIT, 0);
938 addr[0] = IPv4(a, b, c, d);
945 parse_cb_ipv4vlan_rule(char *str, struct rte_acl_rule *v, int has_userdata)
948 char *s, *sp, *in[CB_FLD_NUM];
949 static const char *dlm = " \t\n";
950 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
953 for (i = 0; i != dim; i++, s = NULL) {
954 in[i] = strtok_r(s, dlm, &sp);
959 rc = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
960 &v->field[SRC_FIELD_IPV4].value.u32,
961 &v->field[SRC_FIELD_IPV4].mask_range.u32);
963 acl_log("failed to read source address/mask: %s\n",
964 in[CB_FLD_SRC_ADDR]);
968 rc = parse_ipv4_net(in[CB_FLD_DST_ADDR],
969 &v->field[DST_FIELD_IPV4].value.u32,
970 &v->field[DST_FIELD_IPV4].mask_range.u32);
972 acl_log("failed to read destination address/mask: %s\n",
973 in[CB_FLD_DST_ADDR]);
977 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
978 v->field[SRCP_FIELD_IPV4].value.u16,
980 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
981 v->field[SRCP_FIELD_IPV4].mask_range.u16,
984 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
985 sizeof(cb_port_delim)) != 0)
988 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
989 v->field[DSTP_FIELD_IPV4].value.u16,
991 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
992 v->field[DSTP_FIELD_IPV4].mask_range.u16,
995 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
996 sizeof(cb_port_delim)) != 0)
999 if (v->field[SRCP_FIELD_IPV4].mask_range.u16
1000 < v->field[SRCP_FIELD_IPV4].value.u16
1001 || v->field[DSTP_FIELD_IPV4].mask_range.u16
1002 < v->field[DSTP_FIELD_IPV4].value.u16)
1005 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].value.u8,
1007 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].mask_range.u8,
1011 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata, 0,
1018 add_rules(const char *rule_path,
1019 struct rte_acl_rule **proute_base,
1020 unsigned int *proute_num,
1021 struct rte_acl_rule **pacl_base,
1022 unsigned int *pacl_num, uint32_t rule_size,
1023 int (*parser)(char *, struct rte_acl_rule*, int))
1025 uint8_t *acl_rules, *route_rules;
1026 struct rte_acl_rule *next;
1027 unsigned int acl_num = 0, route_num = 0, total_num = 0;
1028 unsigned int acl_cnt = 0, route_cnt = 0;
1029 char buff[LINE_MAX];
1030 FILE *fh = fopen(rule_path, "rb");
1034 rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
1037 while ((fgets(buff, LINE_MAX, fh) != NULL)) {
1038 if (buff[0] == ROUTE_LEAD_CHAR)
1040 else if (buff[0] == ACL_LEAD_CHAR)
1045 rte_exit(EXIT_FAILURE, "Not find any route entries in %s!\n",
1048 fseek(fh, 0, SEEK_SET);
1050 acl_rules = (uint8_t *)calloc(acl_num, rule_size);
1052 if (NULL == acl_rules)
1053 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1056 route_rules = (uint8_t *)calloc(route_num, rule_size);
1058 if (NULL == route_rules)
1059 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1063 while (fgets(buff, LINE_MAX, fh) != NULL) {
1066 if (is_bypass_line(buff))
1072 if (s == ROUTE_LEAD_CHAR)
1073 next = (struct rte_acl_rule *)(route_rules +
1074 route_cnt * rule_size);
1077 else if (s == ACL_LEAD_CHAR)
1078 next = (struct rte_acl_rule *)(acl_rules +
1079 acl_cnt * rule_size);
1083 rte_exit(EXIT_FAILURE,
1084 "%s Line %u: should start with leading "
1086 rule_path, i, ROUTE_LEAD_CHAR, ACL_LEAD_CHAR);
1088 if (parser(buff + 1, next, s == ROUTE_LEAD_CHAR) != 0)
1089 rte_exit(EXIT_FAILURE,
1090 "%s Line %u: parse rules error\n",
1093 if (s == ROUTE_LEAD_CHAR) {
1094 /* Check the forwarding port number */
1095 if ((enabled_port_mask & (1 << next->data.userdata)) ==
1097 rte_exit(EXIT_FAILURE,
1098 "%s Line %u: fwd number illegal:%u\n",
1099 rule_path, i, next->data.userdata);
1100 next->data.userdata += FWD_PORT_SHIFT;
1103 next->data.userdata = ACL_DENY_SIGNATURE + acl_cnt;
1107 next->data.priority = RTE_ACL_MAX_PRIORITY - total_num;
1108 next->data.category_mask = -1;
1114 *pacl_base = (struct rte_acl_rule *)acl_rules;
1115 *pacl_num = acl_num;
1116 *proute_base = (struct rte_acl_rule *)route_rules;
1117 *proute_num = route_cnt;
1123 dump_acl_config(void)
1125 printf("ACL option are:\n");
1126 printf(OPTION_RULE_IPV4": %s\n", parm_config.rule_ipv4_name);
1127 printf(OPTION_RULE_IPV6": %s\n", parm_config.rule_ipv6_name);
1128 printf(OPTION_SCALAR": %d\n", parm_config.scalar);
1132 check_acl_config(void)
1134 if (parm_config.rule_ipv4_name == NULL) {
1135 acl_log("ACL IPv4 rule file not specified\n");
1137 } else if (parm_config.rule_ipv6_name == NULL) {
1138 acl_log("ACL IPv6 rule file not specified\n");
1145 static struct rte_acl_ctx*
1146 setup_acl(struct rte_acl_rule *route_base,
1147 struct rte_acl_rule *acl_base, unsigned int route_num,
1148 unsigned int acl_num, int ipv6, int socketid)
1150 char name[PATH_MAX];
1151 struct rte_acl_param acl_param;
1152 struct rte_acl_config acl_build_param;
1153 struct rte_acl_ctx *context;
1154 int dim = ipv6 ? RTE_DIM(ipv6_defs) : RTE_DIM(ipv4_defs);
1156 /* Create ACL contexts */
1157 snprintf(name, sizeof(name), "%s%d",
1158 ipv6 ? L3FWD_ACL_IPV6_NAME : L3FWD_ACL_IPV4_NAME,
1161 acl_param.name = name;
1162 acl_param.socket_id = socketid;
1163 acl_param.rule_size = RTE_ACL_RULE_SZ(dim);
1164 acl_param.max_rule_num = MAX_ACL_RULE_NUM;
1166 if ((context = rte_acl_create(&acl_param)) == NULL)
1167 rte_exit(EXIT_FAILURE, "Failed to create ACL context\n");
1169 if (parm_config.scalar && rte_acl_set_ctx_classify(context,
1170 RTE_ACL_CLASSIFY_SCALAR) != 0)
1171 rte_exit(EXIT_FAILURE,
1172 "Failed to setup classify method for ACL context\n");
1174 if (rte_acl_add_rules(context, route_base, route_num) < 0)
1175 rte_exit(EXIT_FAILURE, "add rules failed\n");
1177 if (rte_acl_add_rules(context, acl_base, acl_num) < 0)
1178 rte_exit(EXIT_FAILURE, "add rules failed\n");
1180 /* Perform builds */
1181 acl_build_param.num_categories = DEFAULT_MAX_CATEGORIES;
1183 acl_build_param.num_fields = dim;
1184 memcpy(&acl_build_param.defs, ipv6 ? ipv6_defs : ipv4_defs,
1185 ipv6 ? sizeof(ipv6_defs) : sizeof(ipv4_defs));
1187 if (rte_acl_build(context, &acl_build_param) != 0)
1188 rte_exit(EXIT_FAILURE, "Failed to build ACL trie\n");
1190 rte_acl_dump(context);
1201 struct rte_acl_rule *acl_base_ipv4, *route_base_ipv4,
1202 *acl_base_ipv6, *route_base_ipv6;
1203 unsigned int acl_num_ipv4 = 0, route_num_ipv4 = 0,
1204 acl_num_ipv6 = 0, route_num_ipv6 = 0;
1206 if (check_acl_config() != 0)
1207 rte_exit(EXIT_FAILURE, "Failed to get valid ACL options\n");
1211 /* Load rules from the input file */
1212 if (add_rules(parm_config.rule_ipv4_name, &route_base_ipv4,
1213 &route_num_ipv4, &acl_base_ipv4, &acl_num_ipv4,
1214 sizeof(struct acl4_rule), &parse_cb_ipv4vlan_rule) < 0)
1215 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1217 acl_log("IPv4 Route entries %u:\n", route_num_ipv4);
1218 dump_ipv4_rules((struct acl4_rule *)route_base_ipv4, route_num_ipv4, 1);
1220 acl_log("IPv4 ACL entries %u:\n", acl_num_ipv4);
1221 dump_ipv4_rules((struct acl4_rule *)acl_base_ipv4, acl_num_ipv4, 1);
1223 if (add_rules(parm_config.rule_ipv6_name, &route_base_ipv6,
1225 &acl_base_ipv6, &acl_num_ipv6,
1226 sizeof(struct acl6_rule), &parse_cb_ipv6_rule) < 0)
1227 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1229 acl_log("IPv6 Route entries %u:\n", route_num_ipv6);
1230 dump_ipv6_rules((struct acl6_rule *)route_base_ipv6, route_num_ipv6, 1);
1232 acl_log("IPv6 ACL entries %u:\n", acl_num_ipv6);
1233 dump_ipv6_rules((struct acl6_rule *)acl_base_ipv6, acl_num_ipv6, 1);
1235 memset(&acl_config, 0, sizeof(acl_config));
1237 /* Check sockets a context should be created on */
1239 acl_config.mapped[0] = 1;
1241 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1242 if (rte_lcore_is_enabled(lcore_id) == 0)
1245 socketid = rte_lcore_to_socket_id(lcore_id);
1246 if (socketid >= NB_SOCKETS) {
1247 acl_log("Socket %d of lcore %u is out "
1249 socketid, lcore_id, NB_SOCKETS);
1250 free(route_base_ipv4);
1251 free(route_base_ipv6);
1252 free(acl_base_ipv4);
1253 free(acl_base_ipv6);
1257 acl_config.mapped[socketid] = 1;
1261 for (i = 0; i < NB_SOCKETS; i++) {
1262 if (acl_config.mapped[i]) {
1263 acl_config.acx_ipv4[i] = setup_acl(route_base_ipv4,
1264 acl_base_ipv4, route_num_ipv4, acl_num_ipv4,
1267 acl_config.acx_ipv6[i] = setup_acl(route_base_ipv6,
1268 acl_base_ipv6, route_num_ipv6, acl_num_ipv6,
1273 free(route_base_ipv4);
1274 free(route_base_ipv6);
1276 #ifdef L3FWDACL_DEBUG
1277 acl_config.rule_ipv4 = (struct acl4_rule *)acl_base_ipv4;
1278 acl_config.rule_ipv6 = (struct acl6_rule *)acl_base_ipv6;
1280 free(acl_base_ipv4);
1281 free(acl_base_ipv6);
1287 /***********************end of ACL part******************************/
1290 uint16_t n_rx_queue;
1291 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
1292 uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
1293 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
1294 } __rte_cache_aligned;
1296 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
1298 /* Send burst of packets on an output interface */
1300 send_burst(struct lcore_conf *qconf, uint16_t n, uint8_t port)
1302 struct rte_mbuf **m_table;
1306 queueid = qconf->tx_queue_id[port];
1307 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
1309 ret = rte_eth_tx_burst(port, queueid, m_table, n);
1310 if (unlikely(ret < n)) {
1312 rte_pktmbuf_free(m_table[ret]);
1313 } while (++ret < n);
1319 /* Enqueue a single packet, and send burst if queue is filled */
1321 send_single_packet(struct rte_mbuf *m, uint8_t port)
1325 struct lcore_conf *qconf;
1327 lcore_id = rte_lcore_id();
1329 qconf = &lcore_conf[lcore_id];
1330 len = qconf->tx_mbufs[port].len;
1331 qconf->tx_mbufs[port].m_table[len] = m;
1334 /* enough pkts to be sent */
1335 if (unlikely(len == MAX_PKT_BURST)) {
1336 send_burst(qconf, MAX_PKT_BURST, port);
1340 qconf->tx_mbufs[port].len = len;
1344 #ifdef DO_RFC_1812_CHECKS
1346 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
1348 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
1350 * 1. The packet length reported by the Link Layer must be large
1351 * enough to hold the minimum length legal IP datagram (20 bytes).
1353 if (link_len < sizeof(struct ipv4_hdr))
1356 /* 2. The IP checksum must be correct. */
1357 /* this is checked in H/W */
1360 * 3. The IP version number must be 4. If the version number is not 4
1361 * then the packet may be another version of IP, such as IPng or
1364 if (((pkt->version_ihl) >> 4) != 4)
1367 * 4. The IP header length field must be large enough to hold the
1368 * minimum length legal IP datagram (20 bytes = 5 words).
1370 if ((pkt->version_ihl & 0xf) < 5)
1374 * 5. The IP total length field must be large enough to hold the IP
1375 * datagram header, whose length is specified in the IP header length
1378 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
1385 /* main processing loop */
1387 main_loop(__attribute__((unused)) void *dummy)
1389 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1391 uint64_t prev_tsc, diff_tsc, cur_tsc;
1393 uint8_t portid, queueid;
1394 struct lcore_conf *qconf;
1396 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
1397 / US_PER_S * BURST_TX_DRAIN_US;
1400 lcore_id = rte_lcore_id();
1401 qconf = &lcore_conf[lcore_id];
1402 socketid = rte_lcore_to_socket_id(lcore_id);
1404 if (qconf->n_rx_queue == 0) {
1405 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
1409 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
1411 for (i = 0; i < qconf->n_rx_queue; i++) {
1413 portid = qconf->rx_queue_list[i].port_id;
1414 queueid = qconf->rx_queue_list[i].queue_id;
1415 RTE_LOG(INFO, L3FWD,
1416 " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n",
1417 lcore_id, portid, queueid);
1422 cur_tsc = rte_rdtsc();
1425 * TX burst queue drain
1427 diff_tsc = cur_tsc - prev_tsc;
1428 if (unlikely(diff_tsc > drain_tsc)) {
1431 * This could be optimized (use queueid instead of
1432 * portid), but it is not called so often
1434 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
1435 if (qconf->tx_mbufs[portid].len == 0)
1437 send_burst(&lcore_conf[lcore_id],
1438 qconf->tx_mbufs[portid].len,
1440 qconf->tx_mbufs[portid].len = 0;
1447 * Read packet from RX queues
1449 for (i = 0; i < qconf->n_rx_queue; ++i) {
1451 portid = qconf->rx_queue_list[i].port_id;
1452 queueid = qconf->rx_queue_list[i].queue_id;
1453 nb_rx = rte_eth_rx_burst(portid, queueid,
1454 pkts_burst, MAX_PKT_BURST);
1457 struct acl_search_t acl_search;
1459 prepare_acl_parameter(pkts_burst, &acl_search,
1462 if (acl_search.num_ipv4) {
1464 acl_config.acx_ipv4[socketid],
1465 acl_search.data_ipv4,
1466 acl_search.res_ipv4,
1467 acl_search.num_ipv4,
1468 DEFAULT_MAX_CATEGORIES);
1470 send_packets(acl_search.m_ipv4,
1471 acl_search.res_ipv4,
1472 acl_search.num_ipv4);
1475 if (acl_search.num_ipv6) {
1477 acl_config.acx_ipv6[socketid],
1478 acl_search.data_ipv6,
1479 acl_search.res_ipv6,
1480 acl_search.num_ipv6,
1481 DEFAULT_MAX_CATEGORIES);
1483 send_packets(acl_search.m_ipv6,
1484 acl_search.res_ipv6,
1485 acl_search.num_ipv6);
1493 check_lcore_params(void)
1495 uint8_t queue, lcore;
1499 for (i = 0; i < nb_lcore_params; ++i) {
1500 queue = lcore_params[i].queue_id;
1501 if (queue >= MAX_RX_QUEUE_PER_PORT) {
1502 printf("invalid queue number: %hhu\n", queue);
1505 lcore = lcore_params[i].lcore_id;
1506 if (!rte_lcore_is_enabled(lcore)) {
1507 printf("error: lcore %hhu is not enabled in "
1508 "lcore mask\n", lcore);
1511 socketid = rte_lcore_to_socket_id(lcore);
1512 if (socketid != 0 && numa_on == 0) {
1513 printf("warning: lcore %hhu is on socket %d "
1522 check_port_config(const unsigned nb_ports)
1527 for (i = 0; i < nb_lcore_params; ++i) {
1528 portid = lcore_params[i].port_id;
1530 if ((enabled_port_mask & (1 << portid)) == 0) {
1531 printf("port %u is not enabled in port mask\n", portid);
1534 if (portid >= nb_ports) {
1535 printf("port %u is not present on the board\n", portid);
1543 get_port_n_rx_queues(const uint8_t port)
1548 for (i = 0; i < nb_lcore_params; ++i) {
1549 if (lcore_params[i].port_id == port &&
1550 lcore_params[i].queue_id > queue)
1551 queue = lcore_params[i].queue_id;
1553 return (uint8_t)(++queue);
1557 init_lcore_rx_queues(void)
1559 uint16_t i, nb_rx_queue;
1562 for (i = 0; i < nb_lcore_params; ++i) {
1563 lcore = lcore_params[i].lcore_id;
1564 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
1565 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
1566 printf("error: too many queues (%u) for lcore: %u\n",
1567 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
1570 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
1571 lcore_params[i].port_id;
1572 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
1573 lcore_params[i].queue_id;
1574 lcore_conf[lcore].n_rx_queue++;
1582 print_usage(const char *prgname)
1584 printf("%s [EAL options] -- -p PORTMASK -P"
1585 "--"OPTION_RULE_IPV4"=FILE"
1586 "--"OPTION_RULE_IPV6"=FILE"
1587 " [--"OPTION_CONFIG" (port,queue,lcore)[,(port,queue,lcore]]"
1588 " [--"OPTION_ENBJMO" [--max-pkt-len PKTLEN]]\n"
1589 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
1590 " -P : enable promiscuous mode\n"
1591 " --"OPTION_CONFIG": (port,queue,lcore): "
1592 "rx queues configuration\n"
1593 " --"OPTION_NONUMA": optional, disable numa awareness\n"
1594 " --"OPTION_ENBJMO": enable jumbo frame"
1595 " which max packet len is PKTLEN in decimal (64-9600)\n"
1596 " --"OPTION_RULE_IPV4"=FILE: specify the ipv4 rules entries "
1598 "Each rule occupy one line. "
1599 "2 kinds of rules are supported. "
1600 "One is ACL entry at while line leads with character '%c', "
1601 "another is route entry at while line leads with "
1603 " --"OPTION_RULE_IPV6"=FILE: specify the ipv6 rules "
1605 " --"OPTION_SCALAR": Use scalar function to do lookup\n",
1606 prgname, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR);
1610 parse_max_pkt_len(const char *pktlen)
1615 /* parse decimal string */
1616 len = strtoul(pktlen, &end, 10);
1617 if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1627 parse_portmask(const char *portmask)
1632 /* parse hexadecimal string */
1633 pm = strtoul(portmask, &end, 16);
1634 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1644 parse_config(const char *q_arg)
1647 const char *p, *p0 = q_arg;
1655 unsigned long int_fld[_NUM_FLD];
1656 char *str_fld[_NUM_FLD];
1660 nb_lcore_params = 0;
1662 while ((p = strchr(p0, '(')) != NULL) {
1664 if ((p0 = strchr(p, ')')) == NULL)
1668 if (size >= sizeof(s))
1671 snprintf(s, sizeof(s), "%.*s", size, p);
1672 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') !=
1675 for (i = 0; i < _NUM_FLD; i++) {
1677 int_fld[i] = strtoul(str_fld[i], &end, 0);
1678 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1681 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1682 printf("exceeded max number of lcore params: %hu\n",
1686 lcore_params_array[nb_lcore_params].port_id =
1687 (uint8_t)int_fld[FLD_PORT];
1688 lcore_params_array[nb_lcore_params].queue_id =
1689 (uint8_t)int_fld[FLD_QUEUE];
1690 lcore_params_array[nb_lcore_params].lcore_id =
1691 (uint8_t)int_fld[FLD_LCORE];
1694 lcore_params = lcore_params_array;
1698 /* Parse the argument given in the command line of the application */
1700 parse_args(int argc, char **argv)
1705 char *prgname = argv[0];
1706 static struct option lgopts[] = {
1707 {OPTION_CONFIG, 1, 0, 0},
1708 {OPTION_NONUMA, 0, 0, 0},
1709 {OPTION_ENBJMO, 0, 0, 0},
1710 {OPTION_RULE_IPV4, 1, 0, 0},
1711 {OPTION_RULE_IPV6, 1, 0, 0},
1712 {OPTION_SCALAR, 0, 0, 0},
1718 while ((opt = getopt_long(argc, argvopt, "p:P",
1719 lgopts, &option_index)) != EOF) {
1724 enabled_port_mask = parse_portmask(optarg);
1725 if (enabled_port_mask == 0) {
1726 printf("invalid portmask\n");
1727 print_usage(prgname);
1732 printf("Promiscuous mode selected\n");
1738 if (!strncmp(lgopts[option_index].name,
1740 sizeof(OPTION_CONFIG))) {
1741 ret = parse_config(optarg);
1743 printf("invalid config\n");
1744 print_usage(prgname);
1749 if (!strncmp(lgopts[option_index].name,
1751 sizeof(OPTION_NONUMA))) {
1752 printf("numa is disabled\n");
1756 if (!strncmp(lgopts[option_index].name,
1757 OPTION_ENBJMO, sizeof(OPTION_ENBJMO))) {
1758 struct option lenopts = {
1765 printf("jumbo frame is enabled\n");
1766 port_conf.rxmode.jumbo_frame = 1;
1769 * if no max-pkt-len set, then use the
1770 * default value ETHER_MAX_LEN
1772 if (0 == getopt_long(argc, argvopt, "",
1773 &lenopts, &option_index)) {
1774 ret = parse_max_pkt_len(optarg);
1776 (ret > MAX_JUMBO_PKT_LEN)) {
1777 printf("invalid packet "
1779 print_usage(prgname);
1782 port_conf.rxmode.max_rx_pkt_len = ret;
1784 printf("set jumbo frame max packet length "
1787 port_conf.rxmode.max_rx_pkt_len);
1790 if (!strncmp(lgopts[option_index].name,
1792 sizeof(OPTION_RULE_IPV4)))
1793 parm_config.rule_ipv4_name = optarg;
1795 if (!strncmp(lgopts[option_index].name,
1797 sizeof(OPTION_RULE_IPV6))) {
1798 parm_config.rule_ipv6_name = optarg;
1801 if (!strncmp(lgopts[option_index].name,
1802 OPTION_SCALAR, sizeof(OPTION_SCALAR)))
1803 parm_config.scalar = 1;
1809 print_usage(prgname);
1815 argv[optind-1] = prgname;
1818 optind = 0; /* reset getopt lib */
1823 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
1825 char buf[ETHER_ADDR_FMT_SIZE];
1826 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
1827 printf("%s%s", name, buf);
1831 init_mem(unsigned nb_mbuf)
1837 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1838 if (rte_lcore_is_enabled(lcore_id) == 0)
1842 socketid = rte_lcore_to_socket_id(lcore_id);
1846 if (socketid >= NB_SOCKETS) {
1847 rte_exit(EXIT_FAILURE,
1848 "Socket %d of lcore %u is out of range %d\n",
1849 socketid, lcore_id, NB_SOCKETS);
1851 if (pktmbuf_pool[socketid] == NULL) {
1852 snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
1853 pktmbuf_pool[socketid] =
1854 rte_mempool_create(s, nb_mbuf, MBUF_SIZE,
1856 sizeof(struct rte_pktmbuf_pool_private),
1857 rte_pktmbuf_pool_init, NULL,
1858 rte_pktmbuf_init, NULL,
1860 if (pktmbuf_pool[socketid] == NULL)
1861 rte_exit(EXIT_FAILURE,
1862 "Cannot init mbuf pool on socket %d\n",
1865 printf("Allocated mbuf pool on socket %d\n",
1872 /* Check the link status of all ports in up to 9s, and print them finally */
1874 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
1876 #define CHECK_INTERVAL 100 /* 100ms */
1877 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1878 uint8_t portid, count, all_ports_up, print_flag = 0;
1879 struct rte_eth_link link;
1881 printf("\nChecking link status");
1883 for (count = 0; count <= MAX_CHECK_TIME; count++) {
1885 for (portid = 0; portid < port_num; portid++) {
1886 if ((port_mask & (1 << portid)) == 0)
1888 memset(&link, 0, sizeof(link));
1889 rte_eth_link_get_nowait(portid, &link);
1890 /* print link status if flag set */
1891 if (print_flag == 1) {
1892 if (link.link_status)
1893 printf("Port %d Link Up - speed %u "
1894 "Mbps - %s\n", (uint8_t)portid,
1895 (unsigned)link.link_speed,
1896 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1897 ("full-duplex") : ("half-duplex\n"));
1899 printf("Port %d Link Down\n",
1903 /* clear all_ports_up flag if any link down */
1904 if (link.link_status == 0) {
1909 /* after finally printing all link status, get out */
1910 if (print_flag == 1)
1913 if (all_ports_up == 0) {
1916 rte_delay_ms(CHECK_INTERVAL);
1919 /* set the print_flag if all ports up or timeout */
1920 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1928 main(int argc, char **argv)
1930 struct lcore_conf *qconf;
1931 struct rte_eth_dev_info dev_info;
1932 struct rte_eth_txconf *txconf;
1937 uint32_t n_tx_queue, nb_lcores;
1938 uint8_t portid, nb_rx_queue, queue, socketid;
1941 ret = rte_eal_init(argc, argv);
1943 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1947 /* parse application arguments (after the EAL ones) */
1948 ret = parse_args(argc, argv);
1950 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
1952 if (check_lcore_params() < 0)
1953 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1955 ret = init_lcore_rx_queues();
1957 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1959 nb_ports = rte_eth_dev_count();
1960 if (nb_ports > RTE_MAX_ETHPORTS)
1961 nb_ports = RTE_MAX_ETHPORTS;
1963 if (check_port_config(nb_ports) < 0)
1964 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1966 /* Add ACL rules and route entries, build trie */
1967 if (app_acl_init() < 0)
1968 rte_exit(EXIT_FAILURE, "app_acl_init failed\n");
1970 nb_lcores = rte_lcore_count();
1972 /* initialize all ports */
1973 for (portid = 0; portid < nb_ports; portid++) {
1974 /* skip ports that are not enabled */
1975 if ((enabled_port_mask & (1 << portid)) == 0) {
1976 printf("\nSkipping disabled port %d\n", portid);
1981 printf("Initializing port %d ... ", portid);
1984 nb_rx_queue = get_port_n_rx_queues(portid);
1985 n_tx_queue = nb_lcores;
1986 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1987 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1988 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1989 nb_rx_queue, (unsigned)n_tx_queue);
1990 ret = rte_eth_dev_configure(portid, nb_rx_queue,
1991 (uint16_t)n_tx_queue, &port_conf);
1993 rte_exit(EXIT_FAILURE,
1994 "Cannot configure device: err=%d, port=%d\n",
1997 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1998 print_ethaddr(" Address:", &ports_eth_addr[portid]);
2002 ret = init_mem(NB_MBUF);
2004 rte_exit(EXIT_FAILURE, "init_mem failed\n");
2006 /* init one TX queue per couple (lcore,port) */
2008 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2009 if (rte_lcore_is_enabled(lcore_id) == 0)
2013 socketid = (uint8_t)
2014 rte_lcore_to_socket_id(lcore_id);
2018 printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
2021 rte_eth_dev_info_get(portid, &dev_info);
2022 txconf = &dev_info.default_txconf;
2023 if (port_conf.rxmode.jumbo_frame)
2024 txconf->txq_flags = 0;
2025 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
2028 rte_exit(EXIT_FAILURE,
2029 "rte_eth_tx_queue_setup: err=%d, "
2030 "port=%d\n", ret, portid);
2032 qconf = &lcore_conf[lcore_id];
2033 qconf->tx_queue_id[portid] = queueid;
2039 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2040 if (rte_lcore_is_enabled(lcore_id) == 0)
2042 qconf = &lcore_conf[lcore_id];
2043 printf("\nInitializing rx queues on lcore %u ... ", lcore_id);
2045 /* init RX queues */
2046 for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
2047 portid = qconf->rx_queue_list[queue].port_id;
2048 queueid = qconf->rx_queue_list[queue].queue_id;
2051 socketid = (uint8_t)
2052 rte_lcore_to_socket_id(lcore_id);
2056 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
2059 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
2061 pktmbuf_pool[socketid]);
2063 rte_exit(EXIT_FAILURE,
2064 "rte_eth_rx_queue_setup: err=%d,"
2065 "port=%d\n", ret, portid);
2072 for (portid = 0; portid < nb_ports; portid++) {
2073 if ((enabled_port_mask & (1 << portid)) == 0)
2077 ret = rte_eth_dev_start(portid);
2079 rte_exit(EXIT_FAILURE,
2080 "rte_eth_dev_start: err=%d, port=%d\n",
2084 * If enabled, put device in promiscuous mode.
2085 * This allows IO forwarding mode to forward packets
2086 * to itself through 2 cross-connected ports of the
2090 rte_eth_promiscuous_enable(portid);
2093 check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
2095 /* launch per-lcore init on every lcore */
2096 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
2097 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
2098 if (rte_eal_wait_lcore(lcore_id) < 0)