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>
52 #include <rte_per_lcore.h>
53 #include <rte_launch.h>
54 #include <rte_atomic.h>
55 #include <rte_cycles.h>
56 #include <rte_prefetch.h>
57 #include <rte_lcore.h>
58 #include <rte_per_lcore.h>
59 #include <rte_branch_prediction.h>
60 #include <rte_interrupts.h>
62 #include <rte_random.h>
63 #include <rte_debug.h>
64 #include <rte_ether.h>
65 #include <rte_ethdev.h>
67 #include <rte_mempool.h>
72 #include <rte_string_fns.h>
75 #define DO_RFC_1812_CHECKS
77 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
79 #define MAX_JUMBO_PKT_LEN 9600
81 #define MEMPOOL_CACHE_SIZE 256
83 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
86 * This expression is used to calculate the number of mbufs needed
87 * depending on user input, taking into account memory for rx and tx hardware
88 * rings, cache per lcore and mtable per port per lcore.
89 * RTE_MAX is used to ensure that NB_MBUF never goes below a
90 * minimum value of 8192
93 #define NB_MBUF RTE_MAX(\
94 (nb_ports * nb_rx_queue*RTE_TEST_RX_DESC_DEFAULT + \
95 nb_ports * nb_lcores * MAX_PKT_BURST + \
96 nb_ports * n_tx_queue * RTE_TEST_TX_DESC_DEFAULT + \
97 nb_lcores * MEMPOOL_CACHE_SIZE), \
100 #define MAX_PKT_BURST 32
101 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
105 /* Configure how many packets ahead to prefetch, when reading packets */
106 #define PREFETCH_OFFSET 3
109 * Configurable number of RX/TX ring descriptors
111 #define RTE_TEST_RX_DESC_DEFAULT 128
112 #define RTE_TEST_TX_DESC_DEFAULT 512
113 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
114 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
116 /* ethernet addresses of ports */
117 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
119 /* mask of enabled ports */
120 static uint32_t enabled_port_mask;
121 static int promiscuous_on; /**< Ports set in promiscuous mode off by default. */
122 static int numa_on = 1; /**< NUMA is enabled by default. */
126 struct rte_mbuf *m_table[MAX_PKT_BURST];
129 struct lcore_rx_queue {
132 } __rte_cache_aligned;
134 #define MAX_RX_QUEUE_PER_LCORE 16
135 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
136 #define MAX_RX_QUEUE_PER_PORT 128
138 #define MAX_LCORE_PARAMS 1024
139 struct lcore_params {
143 } __rte_cache_aligned;
145 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
146 static struct lcore_params lcore_params_array_default[] = {
158 static struct lcore_params *lcore_params = lcore_params_array_default;
159 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
160 sizeof(lcore_params_array_default[0]);
162 static struct rte_eth_conf port_conf = {
164 .mq_mode = ETH_MQ_RX_RSS,
165 .max_rx_pkt_len = ETHER_MAX_LEN,
167 .header_split = 0, /**< Header Split disabled */
168 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
169 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
170 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
171 .hw_strip_crc = 0, /**< CRC stripped by hardware */
176 .rss_hf = ETH_RSS_IP | ETH_RSS_UDP |
177 ETH_RSS_TCP | ETH_RSS_SCTP,
181 .mq_mode = ETH_MQ_TX_NONE,
185 static struct rte_mempool *pktmbuf_pool[NB_SOCKETS];
187 /***********************start of ACL part******************************/
188 #ifdef DO_RFC_1812_CHECKS
190 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len);
193 send_single_packet(struct rte_mbuf *m, uint8_t port);
195 #define MAX_ACL_RULE_NUM 100000
196 #define DEFAULT_MAX_CATEGORIES 1
197 #define L3FWD_ACL_IPV4_NAME "l3fwd-acl-ipv4"
198 #define L3FWD_ACL_IPV6_NAME "l3fwd-acl-ipv6"
199 #define ACL_LEAD_CHAR ('@')
200 #define ROUTE_LEAD_CHAR ('R')
201 #define COMMENT_LEAD_CHAR ('#')
202 #define OPTION_CONFIG "config"
203 #define OPTION_NONUMA "no-numa"
204 #define OPTION_ENBJMO "enable-jumbo"
205 #define OPTION_RULE_IPV4 "rule_ipv4"
206 #define OPTION_RULE_IPV6 "rule_ipv6"
207 #define OPTION_SCALAR "scalar"
208 #define ACL_DENY_SIGNATURE 0xf0000000
209 #define RTE_LOGTYPE_L3FWDACL RTE_LOGTYPE_USER3
210 #define acl_log(format, ...) RTE_LOG(ERR, L3FWDACL, format, ##__VA_ARGS__)
211 #define uint32_t_to_char(ip, a, b, c, d) do {\
212 *a = (unsigned char)(ip >> 24 & 0xff);\
213 *b = (unsigned char)(ip >> 16 & 0xff);\
214 *c = (unsigned char)(ip >> 8 & 0xff);\
215 *d = (unsigned char)(ip & 0xff);\
217 #define OFF_ETHHEAD (sizeof(struct ether_hdr))
218 #define OFF_IPV42PROTO (offsetof(struct ipv4_hdr, next_proto_id))
219 #define OFF_IPV62PROTO (offsetof(struct ipv6_hdr, proto))
220 #define MBUF_IPV4_2PROTO(m) \
221 (rte_pktmbuf_mtod((m), uint8_t *) + OFF_ETHHEAD + OFF_IPV42PROTO)
222 #define MBUF_IPV6_2PROTO(m) \
223 (rte_pktmbuf_mtod((m), uint8_t *) + OFF_ETHHEAD + OFF_IPV62PROTO)
225 #define GET_CB_FIELD(in, fd, base, lim, dlm) do { \
229 val = strtoul((in), &end, (base)); \
230 if (errno != 0 || end[0] != (dlm) || val > (lim)) \
232 (fd) = (typeof(fd))val; \
237 * ACL rules should have higher priorities than route ones to ensure ACL rule
238 * always be found when input packets have multi-matches in the database.
239 * A exception case is performance measure, which can define route rules with
240 * higher priority and route rules will always be returned in each lookup.
241 * Reserve range from ACL_RULE_PRIORITY_MAX + 1 to
242 * RTE_ACL_MAX_PRIORITY for route entries in performance measure
244 #define ACL_RULE_PRIORITY_MAX 0x10000000
247 * Forward port info save in ACL lib starts from 1
248 * since ACL assume 0 is invalid.
249 * So, need add 1 when saving and minus 1 when forwarding packets.
251 #define FWD_PORT_SHIFT 1
254 * Rule and trace formats definitions.
266 struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
268 .type = RTE_ACL_FIELD_TYPE_BITMASK,
269 .size = sizeof(uint8_t),
270 .field_index = PROTO_FIELD_IPV4,
271 .input_index = RTE_ACL_IPV4VLAN_PROTO,
275 .type = RTE_ACL_FIELD_TYPE_MASK,
276 .size = sizeof(uint32_t),
277 .field_index = SRC_FIELD_IPV4,
278 .input_index = RTE_ACL_IPV4VLAN_SRC,
279 .offset = offsetof(struct ipv4_hdr, src_addr) -
280 offsetof(struct ipv4_hdr, next_proto_id),
283 .type = RTE_ACL_FIELD_TYPE_MASK,
284 .size = sizeof(uint32_t),
285 .field_index = DST_FIELD_IPV4,
286 .input_index = RTE_ACL_IPV4VLAN_DST,
287 .offset = offsetof(struct ipv4_hdr, dst_addr) -
288 offsetof(struct ipv4_hdr, next_proto_id),
291 .type = RTE_ACL_FIELD_TYPE_RANGE,
292 .size = sizeof(uint16_t),
293 .field_index = SRCP_FIELD_IPV4,
294 .input_index = RTE_ACL_IPV4VLAN_PORTS,
295 .offset = sizeof(struct ipv4_hdr) -
296 offsetof(struct ipv4_hdr, next_proto_id),
299 .type = RTE_ACL_FIELD_TYPE_RANGE,
300 .size = sizeof(uint16_t),
301 .field_index = DSTP_FIELD_IPV4,
302 .input_index = RTE_ACL_IPV4VLAN_PORTS,
303 .offset = sizeof(struct ipv4_hdr) -
304 offsetof(struct ipv4_hdr, next_proto_id) +
309 #define IPV6_ADDR_LEN 16
310 #define IPV6_ADDR_U16 (IPV6_ADDR_LEN / sizeof(uint16_t))
311 #define IPV6_ADDR_U32 (IPV6_ADDR_LEN / sizeof(uint32_t))
328 struct rte_acl_field_def ipv6_defs[NUM_FIELDS_IPV6] = {
330 .type = RTE_ACL_FIELD_TYPE_BITMASK,
331 .size = sizeof(uint8_t),
332 .field_index = PROTO_FIELD_IPV6,
333 .input_index = PROTO_FIELD_IPV6,
337 .type = RTE_ACL_FIELD_TYPE_MASK,
338 .size = sizeof(uint32_t),
339 .field_index = SRC1_FIELD_IPV6,
340 .input_index = SRC1_FIELD_IPV6,
341 .offset = offsetof(struct ipv6_hdr, src_addr) -
342 offsetof(struct ipv6_hdr, proto),
345 .type = RTE_ACL_FIELD_TYPE_MASK,
346 .size = sizeof(uint32_t),
347 .field_index = SRC2_FIELD_IPV6,
348 .input_index = SRC2_FIELD_IPV6,
349 .offset = offsetof(struct ipv6_hdr, src_addr) -
350 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
353 .type = RTE_ACL_FIELD_TYPE_MASK,
354 .size = sizeof(uint32_t),
355 .field_index = SRC3_FIELD_IPV6,
356 .input_index = SRC3_FIELD_IPV6,
357 .offset = offsetof(struct ipv6_hdr, src_addr) -
358 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
361 .type = RTE_ACL_FIELD_TYPE_MASK,
362 .size = sizeof(uint32_t),
363 .field_index = SRC4_FIELD_IPV6,
364 .input_index = SRC4_FIELD_IPV6,
365 .offset = offsetof(struct ipv6_hdr, src_addr) -
366 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
369 .type = RTE_ACL_FIELD_TYPE_MASK,
370 .size = sizeof(uint32_t),
371 .field_index = DST1_FIELD_IPV6,
372 .input_index = DST1_FIELD_IPV6,
373 .offset = offsetof(struct ipv6_hdr, dst_addr)
374 - offsetof(struct ipv6_hdr, proto),
377 .type = RTE_ACL_FIELD_TYPE_MASK,
378 .size = sizeof(uint32_t),
379 .field_index = DST2_FIELD_IPV6,
380 .input_index = DST2_FIELD_IPV6,
381 .offset = offsetof(struct ipv6_hdr, dst_addr) -
382 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
385 .type = RTE_ACL_FIELD_TYPE_MASK,
386 .size = sizeof(uint32_t),
387 .field_index = DST3_FIELD_IPV6,
388 .input_index = DST3_FIELD_IPV6,
389 .offset = offsetof(struct ipv6_hdr, dst_addr) -
390 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
393 .type = RTE_ACL_FIELD_TYPE_MASK,
394 .size = sizeof(uint32_t),
395 .field_index = DST4_FIELD_IPV6,
396 .input_index = DST4_FIELD_IPV6,
397 .offset = offsetof(struct ipv6_hdr, dst_addr) -
398 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
401 .type = RTE_ACL_FIELD_TYPE_RANGE,
402 .size = sizeof(uint16_t),
403 .field_index = SRCP_FIELD_IPV6,
404 .input_index = SRCP_FIELD_IPV6,
405 .offset = sizeof(struct ipv6_hdr) -
406 offsetof(struct ipv6_hdr, proto),
409 .type = RTE_ACL_FIELD_TYPE_RANGE,
410 .size = sizeof(uint16_t),
411 .field_index = DSTP_FIELD_IPV6,
412 .input_index = SRCP_FIELD_IPV6,
413 .offset = sizeof(struct ipv6_hdr) -
414 offsetof(struct ipv6_hdr, proto) + sizeof(uint16_t),
423 CB_FLD_SRC_PORT_HIGH,
426 CB_FLD_DST_PORT_HIGH,
432 RTE_ACL_RULE_DEF(acl4_rule, RTE_DIM(ipv4_defs));
433 RTE_ACL_RULE_DEF(acl6_rule, RTE_DIM(ipv6_defs));
435 struct acl_search_t {
436 const uint8_t *data_ipv4[MAX_PKT_BURST];
437 struct rte_mbuf *m_ipv4[MAX_PKT_BURST];
438 uint32_t res_ipv4[MAX_PKT_BURST];
441 const uint8_t *data_ipv6[MAX_PKT_BURST];
442 struct rte_mbuf *m_ipv6[MAX_PKT_BURST];
443 uint32_t res_ipv6[MAX_PKT_BURST];
448 char mapped[NB_SOCKETS];
449 struct rte_acl_ctx *acx_ipv4[NB_SOCKETS];
450 struct rte_acl_ctx *acx_ipv6[NB_SOCKETS];
451 #ifdef L3FWDACL_DEBUG
452 struct acl4_rule *rule_ipv4;
453 struct acl6_rule *rule_ipv6;
458 const char *rule_ipv4_name;
459 const char *rule_ipv6_name;
463 const char cb_port_delim[] = ":";
466 print_one_ipv4_rule(struct acl4_rule *rule, int extra)
468 unsigned char a, b, c, d;
470 uint32_t_to_char(rule->field[SRC_FIELD_IPV4].value.u32,
472 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
473 rule->field[SRC_FIELD_IPV4].mask_range.u32);
474 uint32_t_to_char(rule->field[DST_FIELD_IPV4].value.u32,
476 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
477 rule->field[DST_FIELD_IPV4].mask_range.u32);
478 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
479 rule->field[SRCP_FIELD_IPV4].value.u16,
480 rule->field[SRCP_FIELD_IPV4].mask_range.u16,
481 rule->field[DSTP_FIELD_IPV4].value.u16,
482 rule->field[DSTP_FIELD_IPV4].mask_range.u16,
483 rule->field[PROTO_FIELD_IPV4].value.u8,
484 rule->field[PROTO_FIELD_IPV4].mask_range.u8);
486 printf("0x%x-0x%x-0x%x ",
487 rule->data.category_mask,
489 rule->data.userdata);
493 print_one_ipv6_rule(struct acl6_rule *rule, int extra)
495 unsigned char a, b, c, d;
497 uint32_t_to_char(rule->field[SRC1_FIELD_IPV6].value.u32,
499 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
500 uint32_t_to_char(rule->field[SRC2_FIELD_IPV6].value.u32,
502 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
503 uint32_t_to_char(rule->field[SRC3_FIELD_IPV6].value.u32,
505 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
506 uint32_t_to_char(rule->field[SRC4_FIELD_IPV6].value.u32,
508 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
509 rule->field[SRC1_FIELD_IPV6].mask_range.u32
510 + rule->field[SRC2_FIELD_IPV6].mask_range.u32
511 + rule->field[SRC3_FIELD_IPV6].mask_range.u32
512 + rule->field[SRC4_FIELD_IPV6].mask_range.u32);
514 uint32_t_to_char(rule->field[DST1_FIELD_IPV6].value.u32,
516 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
517 uint32_t_to_char(rule->field[DST2_FIELD_IPV6].value.u32,
519 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
520 uint32_t_to_char(rule->field[DST3_FIELD_IPV6].value.u32,
522 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
523 uint32_t_to_char(rule->field[DST4_FIELD_IPV6].value.u32,
525 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
526 rule->field[DST1_FIELD_IPV6].mask_range.u32
527 + rule->field[DST2_FIELD_IPV6].mask_range.u32
528 + rule->field[DST3_FIELD_IPV6].mask_range.u32
529 + rule->field[DST4_FIELD_IPV6].mask_range.u32);
531 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
532 rule->field[SRCP_FIELD_IPV6].value.u16,
533 rule->field[SRCP_FIELD_IPV6].mask_range.u16,
534 rule->field[DSTP_FIELD_IPV6].value.u16,
535 rule->field[DSTP_FIELD_IPV6].mask_range.u16,
536 rule->field[PROTO_FIELD_IPV6].value.u8,
537 rule->field[PROTO_FIELD_IPV6].mask_range.u8);
539 printf("0x%x-0x%x-0x%x ",
540 rule->data.category_mask,
542 rule->data.userdata);
545 /* Bypass comment and empty lines */
547 is_bypass_line(char *buff)
552 if (buff[0] == COMMENT_LEAD_CHAR)
555 while (buff[i] != '\0') {
556 if (!isspace(buff[i]))
563 #ifdef L3FWDACL_DEBUG
565 dump_acl4_rule(struct rte_mbuf *m, uint32_t sig)
567 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
568 unsigned char a, b, c, d;
569 struct ipv4_hdr *ipv4_hdr = (struct ipv4_hdr *)
570 (rte_pktmbuf_mtod(m, unsigned char *) +
571 sizeof(struct ether_hdr));
573 uint32_t_to_char(rte_bswap32(ipv4_hdr->src_addr), &a, &b, &c, &d);
574 printf("Packet Src:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
575 uint32_t_to_char(rte_bswap32(ipv4_hdr->dst_addr), &a, &b, &c, &d);
576 printf("Dst:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
578 printf("Src port:%hu,Dst port:%hu ",
579 rte_bswap16(*(uint16_t *)(ipv4_hdr + 1)),
580 rte_bswap16(*((uint16_t *)(ipv4_hdr + 1) + 1)));
581 printf("hit ACL %d - ", offset);
583 print_one_ipv4_rule(acl_config.rule_ipv4 + offset, 1);
589 dump_acl6_rule(struct rte_mbuf *m, uint32_t sig)
592 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
593 struct ipv6_hdr *ipv6_hdr = (struct ipv6_hdr *)
594 (rte_pktmbuf_mtod(m, unsigned char *) +
595 sizeof(struct ether_hdr));
597 printf("Packet Src");
598 for (i = 0; i < RTE_DIM(ipv6_hdr->src_addr); i += sizeof(uint16_t))
600 ipv6_hdr->src_addr[i], ipv6_hdr->src_addr[i + 1]);
603 for (i = 0; i < RTE_DIM(ipv6_hdr->dst_addr); i += sizeof(uint16_t))
605 ipv6_hdr->dst_addr[i], ipv6_hdr->dst_addr[i + 1]);
607 printf("\nSrc port:%hu,Dst port:%hu ",
608 rte_bswap16(*(uint16_t *)(ipv6_hdr + 1)),
609 rte_bswap16(*((uint16_t *)(ipv6_hdr + 1) + 1)));
610 printf("hit ACL %d - ", offset);
612 print_one_ipv6_rule(acl_config.rule_ipv6 + offset, 1);
616 #endif /* L3FWDACL_DEBUG */
619 dump_ipv4_rules(struct acl4_rule *rule, int num, int extra)
623 for (i = 0; i < num; i++, rule++) {
624 printf("\t%d:", i + 1);
625 print_one_ipv4_rule(rule, extra);
631 dump_ipv6_rules(struct acl6_rule *rule, int num, int extra)
635 for (i = 0; i < num; i++, rule++) {
636 printf("\t%d:", i + 1);
637 print_one_ipv6_rule(rule, extra);
642 #ifdef DO_RFC_1812_CHECKS
644 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
647 struct ipv4_hdr *ipv4_hdr;
648 struct rte_mbuf *pkt = pkts_in[index];
650 int type = pkt->ol_flags & (PKT_RX_IPV4_HDR | PKT_RX_IPV6_HDR);
652 if (type == PKT_RX_IPV4_HDR) {
654 ipv4_hdr = (struct ipv4_hdr *)(rte_pktmbuf_mtod(pkt,
655 unsigned char *) + sizeof(struct ether_hdr));
657 /* Check to make sure the packet is valid (RFC1812) */
658 if (is_valid_ipv4_pkt(ipv4_hdr, pkt->pkt_len) >= 0) {
660 /* Update time to live and header checksum */
661 --(ipv4_hdr->time_to_live);
662 ++(ipv4_hdr->hdr_checksum);
664 /* Fill acl structure */
665 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
666 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
669 /* Not a valid IPv4 packet */
670 rte_pktmbuf_free(pkt);
673 } else if (type == PKT_RX_IPV6_HDR) {
675 /* Fill acl structure */
676 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
677 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
680 /* Unknown type, drop the packet */
681 rte_pktmbuf_free(pkt);
687 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
690 struct rte_mbuf *pkt = pkts_in[index];
692 int type = pkt->ol_flags & (PKT_RX_IPV4_HDR | PKT_RX_IPV6_HDR);
694 if (type == PKT_RX_IPV4_HDR) {
696 /* Fill acl structure */
697 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
698 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
701 } else if (type == PKT_RX_IPV6_HDR) {
703 /* Fill acl structure */
704 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
705 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
707 /* Unknown type, drop the packet */
708 rte_pktmbuf_free(pkt);
711 #endif /* DO_RFC_1812_CHECKS */
714 prepare_acl_parameter(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
722 /* Prefetch first packets */
723 for (i = 0; i < PREFETCH_OFFSET && i < nb_rx; i++) {
724 rte_prefetch0(rte_pktmbuf_mtod(
725 pkts_in[i], void *));
728 for (i = 0; i < (nb_rx - PREFETCH_OFFSET); i++) {
729 rte_prefetch0(rte_pktmbuf_mtod(pkts_in[
730 i + PREFETCH_OFFSET], void *));
731 prepare_one_packet(pkts_in, acl, i);
734 /* Process left packets */
735 for (; i < nb_rx; i++)
736 prepare_one_packet(pkts_in, acl, i);
740 send_one_packet(struct rte_mbuf *m, uint32_t res)
742 if (likely((res & ACL_DENY_SIGNATURE) == 0 && res != 0)) {
743 /* forward packets */
744 send_single_packet(m,
745 (uint8_t)(res - FWD_PORT_SHIFT));
747 /* in the ACL list, drop it */
748 #ifdef L3FWDACL_DEBUG
749 if ((res & ACL_DENY_SIGNATURE) != 0) {
750 if (m->ol_flags & PKT_RX_IPV4_HDR)
751 dump_acl4_rule(m, res);
753 dump_acl6_rule(m, res);
763 send_packets(struct rte_mbuf **m, uint32_t *res, int num)
767 /* Prefetch first packets */
768 for (i = 0; i < PREFETCH_OFFSET && i < num; i++) {
769 rte_prefetch0(rte_pktmbuf_mtod(
773 for (i = 0; i < (num - PREFETCH_OFFSET); i++) {
774 rte_prefetch0(rte_pktmbuf_mtod(m[
775 i + PREFETCH_OFFSET], void *));
776 send_one_packet(m[i], res[i]);
779 /* Process left packets */
781 send_one_packet(m[i], res[i]);
785 * Parses IPV6 address, exepcts the following format:
786 * XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX (where X - is a hexedecimal digit).
789 parse_ipv6_addr(const char *in, const char **end, uint32_t v[IPV6_ADDR_U32],
792 uint32_t addr[IPV6_ADDR_U16];
794 GET_CB_FIELD(in, addr[0], 16, UINT16_MAX, ':');
795 GET_CB_FIELD(in, addr[1], 16, UINT16_MAX, ':');
796 GET_CB_FIELD(in, addr[2], 16, UINT16_MAX, ':');
797 GET_CB_FIELD(in, addr[3], 16, UINT16_MAX, ':');
798 GET_CB_FIELD(in, addr[4], 16, UINT16_MAX, ':');
799 GET_CB_FIELD(in, addr[5], 16, UINT16_MAX, ':');
800 GET_CB_FIELD(in, addr[6], 16, UINT16_MAX, ':');
801 GET_CB_FIELD(in, addr[7], 16, UINT16_MAX, dlm);
805 v[0] = (addr[0] << 16) + addr[1];
806 v[1] = (addr[2] << 16) + addr[3];
807 v[2] = (addr[4] << 16) + addr[5];
808 v[3] = (addr[6] << 16) + addr[7];
814 parse_ipv6_net(const char *in, struct rte_acl_field field[4])
819 const uint32_t nbu32 = sizeof(uint32_t) * CHAR_BIT;
822 rc = parse_ipv6_addr(in, &mp, v, '/');
827 GET_CB_FIELD(mp, m, 0, CHAR_BIT * sizeof(v), 0);
829 /* put all together. */
830 for (i = 0; i != RTE_DIM(v); i++) {
831 if (m >= (i + 1) * nbu32)
832 field[i].mask_range.u32 = nbu32;
834 field[i].mask_range.u32 = m > (i * nbu32) ?
837 field[i].value.u32 = v[i];
844 parse_cb_ipv6_rule(char *str, struct rte_acl_rule *v, int has_userdata)
847 char *s, *sp, *in[CB_FLD_NUM];
848 static const char *dlm = " \t\n";
849 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
852 for (i = 0; i != dim; i++, s = NULL) {
853 in[i] = strtok_r(s, dlm, &sp);
858 rc = parse_ipv6_net(in[CB_FLD_SRC_ADDR], v->field + SRC1_FIELD_IPV6);
860 acl_log("failed to read source address/mask: %s\n",
861 in[CB_FLD_SRC_ADDR]);
865 rc = parse_ipv6_net(in[CB_FLD_DST_ADDR], v->field + DST1_FIELD_IPV6);
867 acl_log("failed to read destination address/mask: %s\n",
868 in[CB_FLD_DST_ADDR]);
873 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
874 v->field[SRCP_FIELD_IPV6].value.u16,
876 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
877 v->field[SRCP_FIELD_IPV6].mask_range.u16,
880 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
881 sizeof(cb_port_delim)) != 0)
884 /* destination port. */
885 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
886 v->field[DSTP_FIELD_IPV6].value.u16,
888 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
889 v->field[DSTP_FIELD_IPV6].mask_range.u16,
892 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
893 sizeof(cb_port_delim)) != 0)
896 if (v->field[SRCP_FIELD_IPV6].mask_range.u16
897 < v->field[SRCP_FIELD_IPV6].value.u16
898 || v->field[DSTP_FIELD_IPV6].mask_range.u16
899 < v->field[DSTP_FIELD_IPV6].value.u16)
902 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].value.u8,
904 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].mask_range.u8,
908 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata,
915 * Parse ClassBench rules file.
917 * '@'<src_ipv4_addr>'/'<masklen> <space> \
918 * <dst_ipv4_addr>'/'<masklen> <space> \
919 * <src_port_low> <space> ":" <src_port_high> <space> \
920 * <dst_port_low> <space> ":" <dst_port_high> <space> \
924 parse_ipv4_net(const char *in, uint32_t *addr, uint32_t *mask_len)
926 uint8_t a, b, c, d, m;
928 GET_CB_FIELD(in, a, 0, UINT8_MAX, '.');
929 GET_CB_FIELD(in, b, 0, UINT8_MAX, '.');
930 GET_CB_FIELD(in, c, 0, UINT8_MAX, '.');
931 GET_CB_FIELD(in, d, 0, UINT8_MAX, '/');
932 GET_CB_FIELD(in, m, 0, sizeof(uint32_t) * CHAR_BIT, 0);
934 addr[0] = IPv4(a, b, c, d);
941 parse_cb_ipv4vlan_rule(char *str, struct rte_acl_rule *v, int has_userdata)
944 char *s, *sp, *in[CB_FLD_NUM];
945 static const char *dlm = " \t\n";
946 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
949 for (i = 0; i != dim; i++, s = NULL) {
950 in[i] = strtok_r(s, dlm, &sp);
955 rc = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
956 &v->field[SRC_FIELD_IPV4].value.u32,
957 &v->field[SRC_FIELD_IPV4].mask_range.u32);
959 acl_log("failed to read source address/mask: %s\n",
960 in[CB_FLD_SRC_ADDR]);
964 rc = parse_ipv4_net(in[CB_FLD_DST_ADDR],
965 &v->field[DST_FIELD_IPV4].value.u32,
966 &v->field[DST_FIELD_IPV4].mask_range.u32);
968 acl_log("failed to read destination address/mask: %s\n",
969 in[CB_FLD_DST_ADDR]);
973 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
974 v->field[SRCP_FIELD_IPV4].value.u16,
976 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
977 v->field[SRCP_FIELD_IPV4].mask_range.u16,
980 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
981 sizeof(cb_port_delim)) != 0)
984 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
985 v->field[DSTP_FIELD_IPV4].value.u16,
987 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
988 v->field[DSTP_FIELD_IPV4].mask_range.u16,
991 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
992 sizeof(cb_port_delim)) != 0)
995 if (v->field[SRCP_FIELD_IPV4].mask_range.u16
996 < v->field[SRCP_FIELD_IPV4].value.u16
997 || v->field[DSTP_FIELD_IPV4].mask_range.u16
998 < v->field[DSTP_FIELD_IPV4].value.u16)
1001 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].value.u8,
1003 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].mask_range.u8,
1007 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata, 0,
1014 add_rules(const char *rule_path,
1015 struct rte_acl_rule **proute_base,
1016 unsigned int *proute_num,
1017 struct rte_acl_rule **pacl_base,
1018 unsigned int *pacl_num, uint32_t rule_size,
1019 int (*parser)(char *, struct rte_acl_rule*, int))
1021 uint8_t *acl_rules, *route_rules;
1022 struct rte_acl_rule *next;
1023 unsigned int acl_num = 0, route_num = 0, total_num = 0;
1024 unsigned int acl_cnt = 0, route_cnt = 0;
1025 char buff[LINE_MAX];
1026 FILE *fh = fopen(rule_path, "rb");
1030 rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
1033 while ((fgets(buff, LINE_MAX, fh) != NULL)) {
1034 if (buff[0] == ROUTE_LEAD_CHAR)
1036 else if (buff[0] == ACL_LEAD_CHAR)
1041 rte_exit(EXIT_FAILURE, "Not find any route entries in %s!\n",
1044 fseek(fh, 0, SEEK_SET);
1046 acl_rules = calloc(acl_num, rule_size);
1048 if (NULL == acl_rules)
1049 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1052 route_rules = calloc(route_num, rule_size);
1054 if (NULL == route_rules)
1055 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1059 while (fgets(buff, LINE_MAX, fh) != NULL) {
1062 if (is_bypass_line(buff))
1068 if (s == ROUTE_LEAD_CHAR)
1069 next = (struct rte_acl_rule *)(route_rules +
1070 route_cnt * rule_size);
1073 else if (s == ACL_LEAD_CHAR)
1074 next = (struct rte_acl_rule *)(acl_rules +
1075 acl_cnt * rule_size);
1079 rte_exit(EXIT_FAILURE,
1080 "%s Line %u: should start with leading "
1082 rule_path, i, ROUTE_LEAD_CHAR, ACL_LEAD_CHAR);
1084 if (parser(buff + 1, next, s == ROUTE_LEAD_CHAR) != 0)
1085 rte_exit(EXIT_FAILURE,
1086 "%s Line %u: parse rules error\n",
1089 if (s == ROUTE_LEAD_CHAR) {
1090 /* Check the forwarding port number */
1091 if ((enabled_port_mask & (1 << next->data.userdata)) ==
1093 rte_exit(EXIT_FAILURE,
1094 "%s Line %u: fwd number illegal:%u\n",
1095 rule_path, i, next->data.userdata);
1096 next->data.userdata += FWD_PORT_SHIFT;
1099 next->data.userdata = ACL_DENY_SIGNATURE + acl_cnt;
1103 next->data.priority = RTE_ACL_MAX_PRIORITY - total_num;
1104 next->data.category_mask = -1;
1110 *pacl_base = (struct rte_acl_rule *)acl_rules;
1111 *pacl_num = acl_num;
1112 *proute_base = (struct rte_acl_rule *)route_rules;
1113 *proute_num = route_cnt;
1119 dump_acl_config(void)
1121 printf("ACL option are:\n");
1122 printf(OPTION_RULE_IPV4": %s\n", parm_config.rule_ipv4_name);
1123 printf(OPTION_RULE_IPV6": %s\n", parm_config.rule_ipv6_name);
1124 printf(OPTION_SCALAR": %d\n", parm_config.scalar);
1128 check_acl_config(void)
1130 if (parm_config.rule_ipv4_name == NULL) {
1131 acl_log("ACL IPv4 rule file not specified\n");
1133 } else if (parm_config.rule_ipv6_name == NULL) {
1134 acl_log("ACL IPv6 rule file not specified\n");
1141 static struct rte_acl_ctx*
1142 setup_acl(struct rte_acl_rule *route_base,
1143 struct rte_acl_rule *acl_base, unsigned int route_num,
1144 unsigned int acl_num, int ipv6, int socketid)
1146 char name[PATH_MAX];
1147 struct rte_acl_param acl_param;
1148 struct rte_acl_config acl_build_param;
1149 struct rte_acl_ctx *context;
1150 int dim = ipv6 ? RTE_DIM(ipv6_defs) : RTE_DIM(ipv4_defs);
1152 /* Create ACL contexts */
1153 snprintf(name, sizeof(name), "%s%d",
1154 ipv6 ? L3FWD_ACL_IPV6_NAME : L3FWD_ACL_IPV4_NAME,
1157 acl_param.name = name;
1158 acl_param.socket_id = socketid;
1159 acl_param.rule_size = RTE_ACL_RULE_SZ(dim);
1160 acl_param.max_rule_num = MAX_ACL_RULE_NUM;
1162 if ((context = rte_acl_create(&acl_param)) == NULL)
1163 rte_exit(EXIT_FAILURE, "Failed to create ACL context\n");
1165 if (parm_config.scalar && rte_acl_set_ctx_classify(context,
1166 RTE_ACL_CLASSIFY_SCALAR) != 0)
1167 rte_exit(EXIT_FAILURE,
1168 "Failed to setup classify method for ACL context\n");
1170 if (rte_acl_add_rules(context, route_base, route_num) < 0)
1171 rte_exit(EXIT_FAILURE, "add rules failed\n");
1173 if (rte_acl_add_rules(context, acl_base, acl_num) < 0)
1174 rte_exit(EXIT_FAILURE, "add rules failed\n");
1176 /* Perform builds */
1177 memset(&acl_build_param, 0, sizeof(acl_build_param));
1179 acl_build_param.num_categories = DEFAULT_MAX_CATEGORIES;
1180 acl_build_param.num_fields = dim;
1181 memcpy(&acl_build_param.defs, ipv6 ? ipv6_defs : ipv4_defs,
1182 ipv6 ? sizeof(ipv6_defs) : sizeof(ipv4_defs));
1184 if (rte_acl_build(context, &acl_build_param) != 0)
1185 rte_exit(EXIT_FAILURE, "Failed to build ACL trie\n");
1187 rte_acl_dump(context);
1198 struct rte_acl_rule *acl_base_ipv4, *route_base_ipv4,
1199 *acl_base_ipv6, *route_base_ipv6;
1200 unsigned int acl_num_ipv4 = 0, route_num_ipv4 = 0,
1201 acl_num_ipv6 = 0, route_num_ipv6 = 0;
1203 if (check_acl_config() != 0)
1204 rte_exit(EXIT_FAILURE, "Failed to get valid ACL options\n");
1208 /* Load rules from the input file */
1209 if (add_rules(parm_config.rule_ipv4_name, &route_base_ipv4,
1210 &route_num_ipv4, &acl_base_ipv4, &acl_num_ipv4,
1211 sizeof(struct acl4_rule), &parse_cb_ipv4vlan_rule) < 0)
1212 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1214 acl_log("IPv4 Route entries %u:\n", route_num_ipv4);
1215 dump_ipv4_rules((struct acl4_rule *)route_base_ipv4, route_num_ipv4, 1);
1217 acl_log("IPv4 ACL entries %u:\n", acl_num_ipv4);
1218 dump_ipv4_rules((struct acl4_rule *)acl_base_ipv4, acl_num_ipv4, 1);
1220 if (add_rules(parm_config.rule_ipv6_name, &route_base_ipv6,
1222 &acl_base_ipv6, &acl_num_ipv6,
1223 sizeof(struct acl6_rule), &parse_cb_ipv6_rule) < 0)
1224 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1226 acl_log("IPv6 Route entries %u:\n", route_num_ipv6);
1227 dump_ipv6_rules((struct acl6_rule *)route_base_ipv6, route_num_ipv6, 1);
1229 acl_log("IPv6 ACL entries %u:\n", acl_num_ipv6);
1230 dump_ipv6_rules((struct acl6_rule *)acl_base_ipv6, acl_num_ipv6, 1);
1232 memset(&acl_config, 0, sizeof(acl_config));
1234 /* Check sockets a context should be created on */
1236 acl_config.mapped[0] = 1;
1238 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1239 if (rte_lcore_is_enabled(lcore_id) == 0)
1242 socketid = rte_lcore_to_socket_id(lcore_id);
1243 if (socketid >= NB_SOCKETS) {
1244 acl_log("Socket %d of lcore %u is out "
1246 socketid, lcore_id, NB_SOCKETS);
1247 free(route_base_ipv4);
1248 free(route_base_ipv6);
1249 free(acl_base_ipv4);
1250 free(acl_base_ipv6);
1254 acl_config.mapped[socketid] = 1;
1258 for (i = 0; i < NB_SOCKETS; i++) {
1259 if (acl_config.mapped[i]) {
1260 acl_config.acx_ipv4[i] = setup_acl(route_base_ipv4,
1261 acl_base_ipv4, route_num_ipv4, acl_num_ipv4,
1264 acl_config.acx_ipv6[i] = setup_acl(route_base_ipv6,
1265 acl_base_ipv6, route_num_ipv6, acl_num_ipv6,
1270 free(route_base_ipv4);
1271 free(route_base_ipv6);
1273 #ifdef L3FWDACL_DEBUG
1274 acl_config.rule_ipv4 = (struct acl4_rule *)acl_base_ipv4;
1275 acl_config.rule_ipv6 = (struct acl6_rule *)acl_base_ipv6;
1277 free(acl_base_ipv4);
1278 free(acl_base_ipv6);
1284 /***********************end of ACL part******************************/
1287 uint16_t n_rx_queue;
1288 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
1289 uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
1290 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
1291 } __rte_cache_aligned;
1293 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
1295 /* Send burst of packets on an output interface */
1297 send_burst(struct lcore_conf *qconf, uint16_t n, uint8_t port)
1299 struct rte_mbuf **m_table;
1303 queueid = qconf->tx_queue_id[port];
1304 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
1306 ret = rte_eth_tx_burst(port, queueid, m_table, n);
1307 if (unlikely(ret < n)) {
1309 rte_pktmbuf_free(m_table[ret]);
1310 } while (++ret < n);
1316 /* Enqueue a single packet, and send burst if queue is filled */
1318 send_single_packet(struct rte_mbuf *m, uint8_t port)
1322 struct lcore_conf *qconf;
1324 lcore_id = rte_lcore_id();
1326 qconf = &lcore_conf[lcore_id];
1327 len = qconf->tx_mbufs[port].len;
1328 qconf->tx_mbufs[port].m_table[len] = m;
1331 /* enough pkts to be sent */
1332 if (unlikely(len == MAX_PKT_BURST)) {
1333 send_burst(qconf, MAX_PKT_BURST, port);
1337 qconf->tx_mbufs[port].len = len;
1341 #ifdef DO_RFC_1812_CHECKS
1343 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
1345 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
1347 * 1. The packet length reported by the Link Layer must be large
1348 * enough to hold the minimum length legal IP datagram (20 bytes).
1350 if (link_len < sizeof(struct ipv4_hdr))
1353 /* 2. The IP checksum must be correct. */
1354 /* this is checked in H/W */
1357 * 3. The IP version number must be 4. If the version number is not 4
1358 * then the packet may be another version of IP, such as IPng or
1361 if (((pkt->version_ihl) >> 4) != 4)
1364 * 4. The IP header length field must be large enough to hold the
1365 * minimum length legal IP datagram (20 bytes = 5 words).
1367 if ((pkt->version_ihl & 0xf) < 5)
1371 * 5. The IP total length field must be large enough to hold the IP
1372 * datagram header, whose length is specified in the IP header length
1375 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
1382 /* main processing loop */
1384 main_loop(__attribute__((unused)) void *dummy)
1386 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1388 uint64_t prev_tsc, diff_tsc, cur_tsc;
1390 uint8_t portid, queueid;
1391 struct lcore_conf *qconf;
1393 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
1394 / US_PER_S * BURST_TX_DRAIN_US;
1397 lcore_id = rte_lcore_id();
1398 qconf = &lcore_conf[lcore_id];
1399 socketid = rte_lcore_to_socket_id(lcore_id);
1401 if (qconf->n_rx_queue == 0) {
1402 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
1406 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
1408 for (i = 0; i < qconf->n_rx_queue; i++) {
1410 portid = qconf->rx_queue_list[i].port_id;
1411 queueid = qconf->rx_queue_list[i].queue_id;
1412 RTE_LOG(INFO, L3FWD,
1413 " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n",
1414 lcore_id, portid, queueid);
1419 cur_tsc = rte_rdtsc();
1422 * TX burst queue drain
1424 diff_tsc = cur_tsc - prev_tsc;
1425 if (unlikely(diff_tsc > drain_tsc)) {
1428 * This could be optimized (use queueid instead of
1429 * portid), but it is not called so often
1431 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
1432 if (qconf->tx_mbufs[portid].len == 0)
1434 send_burst(&lcore_conf[lcore_id],
1435 qconf->tx_mbufs[portid].len,
1437 qconf->tx_mbufs[portid].len = 0;
1444 * Read packet from RX queues
1446 for (i = 0; i < qconf->n_rx_queue; ++i) {
1448 portid = qconf->rx_queue_list[i].port_id;
1449 queueid = qconf->rx_queue_list[i].queue_id;
1450 nb_rx = rte_eth_rx_burst(portid, queueid,
1451 pkts_burst, MAX_PKT_BURST);
1454 struct acl_search_t acl_search;
1456 prepare_acl_parameter(pkts_burst, &acl_search,
1459 if (acl_search.num_ipv4) {
1461 acl_config.acx_ipv4[socketid],
1462 acl_search.data_ipv4,
1463 acl_search.res_ipv4,
1464 acl_search.num_ipv4,
1465 DEFAULT_MAX_CATEGORIES);
1467 send_packets(acl_search.m_ipv4,
1468 acl_search.res_ipv4,
1469 acl_search.num_ipv4);
1472 if (acl_search.num_ipv6) {
1474 acl_config.acx_ipv6[socketid],
1475 acl_search.data_ipv6,
1476 acl_search.res_ipv6,
1477 acl_search.num_ipv6,
1478 DEFAULT_MAX_CATEGORIES);
1480 send_packets(acl_search.m_ipv6,
1481 acl_search.res_ipv6,
1482 acl_search.num_ipv6);
1490 check_lcore_params(void)
1492 uint8_t queue, lcore;
1496 for (i = 0; i < nb_lcore_params; ++i) {
1497 queue = lcore_params[i].queue_id;
1498 if (queue >= MAX_RX_QUEUE_PER_PORT) {
1499 printf("invalid queue number: %hhu\n", queue);
1502 lcore = lcore_params[i].lcore_id;
1503 if (!rte_lcore_is_enabled(lcore)) {
1504 printf("error: lcore %hhu is not enabled in "
1505 "lcore mask\n", lcore);
1508 socketid = rte_lcore_to_socket_id(lcore);
1509 if (socketid != 0 && numa_on == 0) {
1510 printf("warning: lcore %hhu is on socket %d "
1519 check_port_config(const unsigned nb_ports)
1524 for (i = 0; i < nb_lcore_params; ++i) {
1525 portid = lcore_params[i].port_id;
1527 if ((enabled_port_mask & (1 << portid)) == 0) {
1528 printf("port %u is not enabled in port mask\n", portid);
1531 if (portid >= nb_ports) {
1532 printf("port %u is not present on the board\n", portid);
1540 get_port_n_rx_queues(const uint8_t port)
1545 for (i = 0; i < nb_lcore_params; ++i) {
1546 if (lcore_params[i].port_id == port &&
1547 lcore_params[i].queue_id > queue)
1548 queue = lcore_params[i].queue_id;
1550 return (uint8_t)(++queue);
1554 init_lcore_rx_queues(void)
1556 uint16_t i, nb_rx_queue;
1559 for (i = 0; i < nb_lcore_params; ++i) {
1560 lcore = lcore_params[i].lcore_id;
1561 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
1562 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
1563 printf("error: too many queues (%u) for lcore: %u\n",
1564 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
1567 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
1568 lcore_params[i].port_id;
1569 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
1570 lcore_params[i].queue_id;
1571 lcore_conf[lcore].n_rx_queue++;
1579 print_usage(const char *prgname)
1581 printf("%s [EAL options] -- -p PORTMASK -P"
1582 "--"OPTION_RULE_IPV4"=FILE"
1583 "--"OPTION_RULE_IPV6"=FILE"
1584 " [--"OPTION_CONFIG" (port,queue,lcore)[,(port,queue,lcore]]"
1585 " [--"OPTION_ENBJMO" [--max-pkt-len PKTLEN]]\n"
1586 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
1587 " -P : enable promiscuous mode\n"
1588 " --"OPTION_CONFIG": (port,queue,lcore): "
1589 "rx queues configuration\n"
1590 " --"OPTION_NONUMA": optional, disable numa awareness\n"
1591 " --"OPTION_ENBJMO": enable jumbo frame"
1592 " which max packet len is PKTLEN in decimal (64-9600)\n"
1593 " --"OPTION_RULE_IPV4"=FILE: specify the ipv4 rules entries "
1595 "Each rule occupy one line. "
1596 "2 kinds of rules are supported. "
1597 "One is ACL entry at while line leads with character '%c', "
1598 "another is route entry at while line leads with "
1600 " --"OPTION_RULE_IPV6"=FILE: specify the ipv6 rules "
1602 " --"OPTION_SCALAR": Use scalar function to do lookup\n",
1603 prgname, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR);
1607 parse_max_pkt_len(const char *pktlen)
1612 /* parse decimal string */
1613 len = strtoul(pktlen, &end, 10);
1614 if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1624 parse_portmask(const char *portmask)
1629 /* parse hexadecimal string */
1630 pm = strtoul(portmask, &end, 16);
1631 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1641 parse_config(const char *q_arg)
1644 const char *p, *p0 = q_arg;
1652 unsigned long int_fld[_NUM_FLD];
1653 char *str_fld[_NUM_FLD];
1657 nb_lcore_params = 0;
1659 while ((p = strchr(p0, '(')) != NULL) {
1661 if ((p0 = strchr(p, ')')) == NULL)
1665 if (size >= sizeof(s))
1668 snprintf(s, sizeof(s), "%.*s", size, p);
1669 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') !=
1672 for (i = 0; i < _NUM_FLD; i++) {
1674 int_fld[i] = strtoul(str_fld[i], &end, 0);
1675 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1678 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1679 printf("exceeded max number of lcore params: %hu\n",
1683 lcore_params_array[nb_lcore_params].port_id =
1684 (uint8_t)int_fld[FLD_PORT];
1685 lcore_params_array[nb_lcore_params].queue_id =
1686 (uint8_t)int_fld[FLD_QUEUE];
1687 lcore_params_array[nb_lcore_params].lcore_id =
1688 (uint8_t)int_fld[FLD_LCORE];
1691 lcore_params = lcore_params_array;
1695 /* Parse the argument given in the command line of the application */
1697 parse_args(int argc, char **argv)
1702 char *prgname = argv[0];
1703 static struct option lgopts[] = {
1704 {OPTION_CONFIG, 1, 0, 0},
1705 {OPTION_NONUMA, 0, 0, 0},
1706 {OPTION_ENBJMO, 0, 0, 0},
1707 {OPTION_RULE_IPV4, 1, 0, 0},
1708 {OPTION_RULE_IPV6, 1, 0, 0},
1709 {OPTION_SCALAR, 0, 0, 0},
1715 while ((opt = getopt_long(argc, argvopt, "p:P",
1716 lgopts, &option_index)) != EOF) {
1721 enabled_port_mask = parse_portmask(optarg);
1722 if (enabled_port_mask == 0) {
1723 printf("invalid portmask\n");
1724 print_usage(prgname);
1729 printf("Promiscuous mode selected\n");
1735 if (!strncmp(lgopts[option_index].name,
1737 sizeof(OPTION_CONFIG))) {
1738 ret = parse_config(optarg);
1740 printf("invalid config\n");
1741 print_usage(prgname);
1746 if (!strncmp(lgopts[option_index].name,
1748 sizeof(OPTION_NONUMA))) {
1749 printf("numa is disabled\n");
1753 if (!strncmp(lgopts[option_index].name,
1754 OPTION_ENBJMO, sizeof(OPTION_ENBJMO))) {
1755 struct option lenopts = {
1762 printf("jumbo frame is enabled\n");
1763 port_conf.rxmode.jumbo_frame = 1;
1766 * if no max-pkt-len set, then use the
1767 * default value ETHER_MAX_LEN
1769 if (0 == getopt_long(argc, argvopt, "",
1770 &lenopts, &option_index)) {
1771 ret = parse_max_pkt_len(optarg);
1773 (ret > MAX_JUMBO_PKT_LEN)) {
1774 printf("invalid packet "
1776 print_usage(prgname);
1779 port_conf.rxmode.max_rx_pkt_len = ret;
1781 printf("set jumbo frame max packet length "
1784 port_conf.rxmode.max_rx_pkt_len);
1787 if (!strncmp(lgopts[option_index].name,
1789 sizeof(OPTION_RULE_IPV4)))
1790 parm_config.rule_ipv4_name = optarg;
1792 if (!strncmp(lgopts[option_index].name,
1794 sizeof(OPTION_RULE_IPV6))) {
1795 parm_config.rule_ipv6_name = optarg;
1798 if (!strncmp(lgopts[option_index].name,
1799 OPTION_SCALAR, sizeof(OPTION_SCALAR)))
1800 parm_config.scalar = 1;
1806 print_usage(prgname);
1812 argv[optind-1] = prgname;
1815 optind = 0; /* reset getopt lib */
1820 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
1822 char buf[ETHER_ADDR_FMT_SIZE];
1823 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
1824 printf("%s%s", name, buf);
1828 init_mem(unsigned nb_mbuf)
1834 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1835 if (rte_lcore_is_enabled(lcore_id) == 0)
1839 socketid = rte_lcore_to_socket_id(lcore_id);
1843 if (socketid >= NB_SOCKETS) {
1844 rte_exit(EXIT_FAILURE,
1845 "Socket %d of lcore %u is out of range %d\n",
1846 socketid, lcore_id, NB_SOCKETS);
1848 if (pktmbuf_pool[socketid] == NULL) {
1849 snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
1850 pktmbuf_pool[socketid] =
1851 rte_mempool_create(s, nb_mbuf, MBUF_SIZE,
1853 sizeof(struct rte_pktmbuf_pool_private),
1854 rte_pktmbuf_pool_init, NULL,
1855 rte_pktmbuf_init, NULL,
1857 if (pktmbuf_pool[socketid] == NULL)
1858 rte_exit(EXIT_FAILURE,
1859 "Cannot init mbuf pool on socket %d\n",
1862 printf("Allocated mbuf pool on socket %d\n",
1869 /* Check the link status of all ports in up to 9s, and print them finally */
1871 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
1873 #define CHECK_INTERVAL 100 /* 100ms */
1874 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1875 uint8_t portid, count, all_ports_up, print_flag = 0;
1876 struct rte_eth_link link;
1878 printf("\nChecking link status");
1880 for (count = 0; count <= MAX_CHECK_TIME; count++) {
1882 for (portid = 0; portid < port_num; portid++) {
1883 if ((port_mask & (1 << portid)) == 0)
1885 memset(&link, 0, sizeof(link));
1886 rte_eth_link_get_nowait(portid, &link);
1887 /* print link status if flag set */
1888 if (print_flag == 1) {
1889 if (link.link_status)
1890 printf("Port %d Link Up - speed %u "
1891 "Mbps - %s\n", (uint8_t)portid,
1892 (unsigned)link.link_speed,
1893 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1894 ("full-duplex") : ("half-duplex\n"));
1896 printf("Port %d Link Down\n",
1900 /* clear all_ports_up flag if any link down */
1901 if (link.link_status == 0) {
1906 /* after finally printing all link status, get out */
1907 if (print_flag == 1)
1910 if (all_ports_up == 0) {
1913 rte_delay_ms(CHECK_INTERVAL);
1916 /* set the print_flag if all ports up or timeout */
1917 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1925 main(int argc, char **argv)
1927 struct lcore_conf *qconf;
1928 struct rte_eth_dev_info dev_info;
1929 struct rte_eth_txconf *txconf;
1934 uint32_t n_tx_queue, nb_lcores;
1935 uint8_t portid, nb_rx_queue, queue, socketid;
1938 ret = rte_eal_init(argc, argv);
1940 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1944 /* parse application arguments (after the EAL ones) */
1945 ret = parse_args(argc, argv);
1947 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
1949 if (check_lcore_params() < 0)
1950 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1952 ret = init_lcore_rx_queues();
1954 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1956 nb_ports = rte_eth_dev_count();
1957 if (nb_ports > RTE_MAX_ETHPORTS)
1958 nb_ports = RTE_MAX_ETHPORTS;
1960 if (check_port_config(nb_ports) < 0)
1961 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1963 /* Add ACL rules and route entries, build trie */
1964 if (app_acl_init() < 0)
1965 rte_exit(EXIT_FAILURE, "app_acl_init failed\n");
1967 nb_lcores = rte_lcore_count();
1969 /* initialize all ports */
1970 for (portid = 0; portid < nb_ports; portid++) {
1971 /* skip ports that are not enabled */
1972 if ((enabled_port_mask & (1 << portid)) == 0) {
1973 printf("\nSkipping disabled port %d\n", portid);
1978 printf("Initializing port %d ... ", portid);
1981 nb_rx_queue = get_port_n_rx_queues(portid);
1982 n_tx_queue = nb_lcores;
1983 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1984 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1985 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1986 nb_rx_queue, (unsigned)n_tx_queue);
1987 ret = rte_eth_dev_configure(portid, nb_rx_queue,
1988 (uint16_t)n_tx_queue, &port_conf);
1990 rte_exit(EXIT_FAILURE,
1991 "Cannot configure device: err=%d, port=%d\n",
1994 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1995 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1999 ret = init_mem(NB_MBUF);
2001 rte_exit(EXIT_FAILURE, "init_mem failed\n");
2003 /* init one TX queue per couple (lcore,port) */
2005 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2006 if (rte_lcore_is_enabled(lcore_id) == 0)
2010 socketid = (uint8_t)
2011 rte_lcore_to_socket_id(lcore_id);
2015 printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
2018 rte_eth_dev_info_get(portid, &dev_info);
2019 txconf = &dev_info.default_txconf;
2020 if (port_conf.rxmode.jumbo_frame)
2021 txconf->txq_flags = 0;
2022 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
2025 rte_exit(EXIT_FAILURE,
2026 "rte_eth_tx_queue_setup: err=%d, "
2027 "port=%d\n", ret, portid);
2029 qconf = &lcore_conf[lcore_id];
2030 qconf->tx_queue_id[portid] = queueid;
2036 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2037 if (rte_lcore_is_enabled(lcore_id) == 0)
2039 qconf = &lcore_conf[lcore_id];
2040 printf("\nInitializing rx queues on lcore %u ... ", lcore_id);
2042 /* init RX queues */
2043 for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
2044 portid = qconf->rx_queue_list[queue].port_id;
2045 queueid = qconf->rx_queue_list[queue].queue_id;
2048 socketid = (uint8_t)
2049 rte_lcore_to_socket_id(lcore_id);
2053 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
2056 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
2058 pktmbuf_pool[socketid]);
2060 rte_exit(EXIT_FAILURE,
2061 "rte_eth_rx_queue_setup: err=%d,"
2062 "port=%d\n", ret, portid);
2069 for (portid = 0; portid < nb_ports; portid++) {
2070 if ((enabled_port_mask & (1 << portid)) == 0)
2074 ret = rte_eth_dev_start(portid);
2076 rte_exit(EXIT_FAILURE,
2077 "rte_eth_dev_start: err=%d, port=%d\n",
2081 * If enabled, put device in promiscuous mode.
2082 * This allows IO forwarding mode to forward packets
2083 * to itself through 2 cross-connected ports of the
2087 rte_eth_promiscuous_enable(portid);
2090 check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
2092 /* launch per-lcore init on every lcore */
2093 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
2094 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
2095 if (rte_eal_wait_lcore(lcore_id) < 0)