1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright(c) 2010-2016 Intel Corporation
11 #include <sys/queue.h>
16 #include <rte_common.h>
17 #include <rte_byteorder.h>
19 #include <rte_memory.h>
20 #include <rte_memcpy.h>
22 #include <rte_launch.h>
23 #include <rte_atomic.h>
24 #include <rte_cycles.h>
25 #include <rte_prefetch.h>
26 #include <rte_lcore.h>
27 #include <rte_per_lcore.h>
28 #include <rte_branch_prediction.h>
29 #include <rte_interrupts.h>
30 #include <rte_random.h>
31 #include <rte_debug.h>
32 #include <rte_ether.h>
33 #include <rte_ethdev.h>
34 #include <rte_mempool.h>
39 #include <rte_string_fns.h>
42 #include <cmdline_parse.h>
43 #include <cmdline_parse_etheraddr.h>
45 #if RTE_LOG_DP_LEVEL >= RTE_LOG_DEBUG
46 #define L3FWDACL_DEBUG
48 #define DO_RFC_1812_CHECKS
50 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
52 #define MAX_JUMBO_PKT_LEN 9600
54 #define MEMPOOL_CACHE_SIZE 256
57 * This expression is used to calculate the number of mbufs needed
58 * depending on user input, taking into account memory for rx and tx hardware
59 * rings, cache per lcore and mtable per port per lcore.
60 * RTE_MAX is used to ensure that NB_MBUF never goes below a
61 * minimum value of 8192
64 #define NB_MBUF RTE_MAX(\
65 (nb_ports * nb_rx_queue * nb_rxd + \
66 nb_ports * nb_lcores * MAX_PKT_BURST + \
67 nb_ports * n_tx_queue * nb_txd + \
68 nb_lcores * MEMPOOL_CACHE_SIZE), \
71 #define MAX_PKT_BURST 32
72 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
76 /* Configure how many packets ahead to prefetch, when reading packets */
77 #define PREFETCH_OFFSET 3
80 * Configurable number of RX/TX ring descriptors
82 #define RTE_TEST_RX_DESC_DEFAULT 1024
83 #define RTE_TEST_TX_DESC_DEFAULT 1024
84 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
85 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
87 /* mask of enabled ports */
88 static uint32_t enabled_port_mask;
89 static int promiscuous_on; /**< Ports set in promiscuous mode off by default. */
90 static int numa_on = 1; /**< NUMA is enabled by default. */
92 struct lcore_rx_queue {
95 } __rte_cache_aligned;
97 #define MAX_RX_QUEUE_PER_LCORE 16
98 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
99 #define MAX_RX_QUEUE_PER_PORT 128
101 #define MAX_LCORE_PARAMS 1024
102 struct lcore_params {
106 } __rte_cache_aligned;
108 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
109 static struct lcore_params lcore_params_array_default[] = {
121 static struct lcore_params *lcore_params = lcore_params_array_default;
122 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
123 sizeof(lcore_params_array_default[0]);
125 static struct rte_eth_conf port_conf = {
127 .mq_mode = ETH_MQ_RX_RSS,
128 .max_rx_pkt_len = RTE_ETHER_MAX_LEN,
130 .offloads = DEV_RX_OFFLOAD_CHECKSUM,
135 .rss_hf = ETH_RSS_IP | ETH_RSS_UDP |
136 ETH_RSS_TCP | ETH_RSS_SCTP,
140 .mq_mode = ETH_MQ_TX_NONE,
144 static struct rte_mempool *pktmbuf_pool[NB_SOCKETS];
146 /* ethernet addresses of ports */
147 static struct rte_ether_hdr port_l2hdr[RTE_MAX_ETHPORTS];
149 /***********************start of ACL part******************************/
150 #ifdef DO_RFC_1812_CHECKS
152 is_valid_ipv4_pkt(struct rte_ipv4_hdr *pkt, uint32_t link_len);
155 send_single_packet(struct rte_mbuf *m, uint16_t port);
157 #define MAX_ACL_RULE_NUM 100000
158 #define DEFAULT_MAX_CATEGORIES 1
159 #define L3FWD_ACL_IPV4_NAME "l3fwd-acl-ipv4"
160 #define L3FWD_ACL_IPV6_NAME "l3fwd-acl-ipv6"
161 #define ACL_LEAD_CHAR ('@')
162 #define ROUTE_LEAD_CHAR ('R')
163 #define COMMENT_LEAD_CHAR ('#')
164 #define OPTION_CONFIG "config"
165 #define OPTION_NONUMA "no-numa"
166 #define OPTION_ENBJMO "enable-jumbo"
167 #define OPTION_RULE_IPV4 "rule_ipv4"
168 #define OPTION_RULE_IPV6 "rule_ipv6"
169 #define OPTION_SCALAR "scalar"
170 #define OPTION_ETH_DEST "eth-dest"
171 #define ACL_DENY_SIGNATURE 0xf0000000
172 #define RTE_LOGTYPE_L3FWDACL RTE_LOGTYPE_USER3
173 #define acl_log(format, ...) RTE_LOG(ERR, L3FWDACL, format, ##__VA_ARGS__)
174 #define uint32_t_to_char(ip, a, b, c, d) do {\
175 *a = (unsigned char)(ip >> 24 & 0xff);\
176 *b = (unsigned char)(ip >> 16 & 0xff);\
177 *c = (unsigned char)(ip >> 8 & 0xff);\
178 *d = (unsigned char)(ip & 0xff);\
180 #define OFF_ETHHEAD (sizeof(struct rte_ether_hdr))
181 #define OFF_IPV42PROTO (offsetof(struct rte_ipv4_hdr, next_proto_id))
182 #define OFF_IPV62PROTO (offsetof(struct rte_ipv6_hdr, proto))
183 #define MBUF_IPV4_2PROTO(m) \
184 rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV42PROTO)
185 #define MBUF_IPV6_2PROTO(m) \
186 rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV62PROTO)
188 #define GET_CB_FIELD(in, fd, base, lim, dlm) do { \
192 val = strtoul((in), &end, (base)); \
193 if (errno != 0 || end[0] != (dlm) || val > (lim)) \
195 (fd) = (typeof(fd))val; \
200 * ACL rules should have higher priorities than route ones to ensure ACL rule
201 * always be found when input packets have multi-matches in the database.
202 * A exception case is performance measure, which can define route rules with
203 * higher priority and route rules will always be returned in each lookup.
204 * Reserve range from ACL_RULE_PRIORITY_MAX + 1 to
205 * RTE_ACL_MAX_PRIORITY for route entries in performance measure
207 #define ACL_RULE_PRIORITY_MAX 0x10000000
210 * Forward port info save in ACL lib starts from 1
211 * since ACL assume 0 is invalid.
212 * So, need add 1 when saving and minus 1 when forwarding packets.
214 #define FWD_PORT_SHIFT 1
217 * Rule and trace formats definitions.
230 * That effectively defines order of IPV4VLAN classifications:
232 * - VLAN (TAG and DOMAIN)
235 * - PORTS (SRC and DST)
238 RTE_ACL_IPV4VLAN_PROTO,
239 RTE_ACL_IPV4VLAN_VLAN,
240 RTE_ACL_IPV4VLAN_SRC,
241 RTE_ACL_IPV4VLAN_DST,
242 RTE_ACL_IPV4VLAN_PORTS,
246 struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
248 .type = RTE_ACL_FIELD_TYPE_BITMASK,
249 .size = sizeof(uint8_t),
250 .field_index = PROTO_FIELD_IPV4,
251 .input_index = RTE_ACL_IPV4VLAN_PROTO,
255 .type = RTE_ACL_FIELD_TYPE_MASK,
256 .size = sizeof(uint32_t),
257 .field_index = SRC_FIELD_IPV4,
258 .input_index = RTE_ACL_IPV4VLAN_SRC,
259 .offset = offsetof(struct rte_ipv4_hdr, src_addr) -
260 offsetof(struct rte_ipv4_hdr, next_proto_id),
263 .type = RTE_ACL_FIELD_TYPE_MASK,
264 .size = sizeof(uint32_t),
265 .field_index = DST_FIELD_IPV4,
266 .input_index = RTE_ACL_IPV4VLAN_DST,
267 .offset = offsetof(struct rte_ipv4_hdr, dst_addr) -
268 offsetof(struct rte_ipv4_hdr, next_proto_id),
271 .type = RTE_ACL_FIELD_TYPE_RANGE,
272 .size = sizeof(uint16_t),
273 .field_index = SRCP_FIELD_IPV4,
274 .input_index = RTE_ACL_IPV4VLAN_PORTS,
275 .offset = sizeof(struct rte_ipv4_hdr) -
276 offsetof(struct rte_ipv4_hdr, next_proto_id),
279 .type = RTE_ACL_FIELD_TYPE_RANGE,
280 .size = sizeof(uint16_t),
281 .field_index = DSTP_FIELD_IPV4,
282 .input_index = RTE_ACL_IPV4VLAN_PORTS,
283 .offset = sizeof(struct rte_ipv4_hdr) -
284 offsetof(struct rte_ipv4_hdr, next_proto_id) +
289 #define IPV6_ADDR_LEN 16
290 #define IPV6_ADDR_U16 (IPV6_ADDR_LEN / sizeof(uint16_t))
291 #define IPV6_ADDR_U32 (IPV6_ADDR_LEN / sizeof(uint32_t))
308 struct rte_acl_field_def ipv6_defs[NUM_FIELDS_IPV6] = {
310 .type = RTE_ACL_FIELD_TYPE_BITMASK,
311 .size = sizeof(uint8_t),
312 .field_index = PROTO_FIELD_IPV6,
313 .input_index = PROTO_FIELD_IPV6,
317 .type = RTE_ACL_FIELD_TYPE_MASK,
318 .size = sizeof(uint32_t),
319 .field_index = SRC1_FIELD_IPV6,
320 .input_index = SRC1_FIELD_IPV6,
321 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
322 offsetof(struct rte_ipv6_hdr, proto),
325 .type = RTE_ACL_FIELD_TYPE_MASK,
326 .size = sizeof(uint32_t),
327 .field_index = SRC2_FIELD_IPV6,
328 .input_index = SRC2_FIELD_IPV6,
329 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
330 offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint32_t),
333 .type = RTE_ACL_FIELD_TYPE_MASK,
334 .size = sizeof(uint32_t),
335 .field_index = SRC3_FIELD_IPV6,
336 .input_index = SRC3_FIELD_IPV6,
337 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
338 offsetof(struct rte_ipv6_hdr, proto) +
339 2 * sizeof(uint32_t),
342 .type = RTE_ACL_FIELD_TYPE_MASK,
343 .size = sizeof(uint32_t),
344 .field_index = SRC4_FIELD_IPV6,
345 .input_index = SRC4_FIELD_IPV6,
346 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
347 offsetof(struct rte_ipv6_hdr, proto) +
348 3 * sizeof(uint32_t),
351 .type = RTE_ACL_FIELD_TYPE_MASK,
352 .size = sizeof(uint32_t),
353 .field_index = DST1_FIELD_IPV6,
354 .input_index = DST1_FIELD_IPV6,
355 .offset = offsetof(struct rte_ipv6_hdr, dst_addr)
356 - offsetof(struct rte_ipv6_hdr, proto),
359 .type = RTE_ACL_FIELD_TYPE_MASK,
360 .size = sizeof(uint32_t),
361 .field_index = DST2_FIELD_IPV6,
362 .input_index = DST2_FIELD_IPV6,
363 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
364 offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint32_t),
367 .type = RTE_ACL_FIELD_TYPE_MASK,
368 .size = sizeof(uint32_t),
369 .field_index = DST3_FIELD_IPV6,
370 .input_index = DST3_FIELD_IPV6,
371 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
372 offsetof(struct rte_ipv6_hdr, proto) +
373 2 * sizeof(uint32_t),
376 .type = RTE_ACL_FIELD_TYPE_MASK,
377 .size = sizeof(uint32_t),
378 .field_index = DST4_FIELD_IPV6,
379 .input_index = DST4_FIELD_IPV6,
380 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
381 offsetof(struct rte_ipv6_hdr, proto) +
382 3 * sizeof(uint32_t),
385 .type = RTE_ACL_FIELD_TYPE_RANGE,
386 .size = sizeof(uint16_t),
387 .field_index = SRCP_FIELD_IPV6,
388 .input_index = SRCP_FIELD_IPV6,
389 .offset = sizeof(struct rte_ipv6_hdr) -
390 offsetof(struct rte_ipv6_hdr, proto),
393 .type = RTE_ACL_FIELD_TYPE_RANGE,
394 .size = sizeof(uint16_t),
395 .field_index = DSTP_FIELD_IPV6,
396 .input_index = SRCP_FIELD_IPV6,
397 .offset = sizeof(struct rte_ipv6_hdr) -
398 offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint16_t),
407 CB_FLD_SRC_PORT_HIGH,
410 CB_FLD_DST_PORT_HIGH,
416 RTE_ACL_RULE_DEF(acl4_rule, RTE_DIM(ipv4_defs));
417 RTE_ACL_RULE_DEF(acl6_rule, RTE_DIM(ipv6_defs));
419 struct acl_search_t {
420 const uint8_t *data_ipv4[MAX_PKT_BURST];
421 struct rte_mbuf *m_ipv4[MAX_PKT_BURST];
422 uint32_t res_ipv4[MAX_PKT_BURST];
425 const uint8_t *data_ipv6[MAX_PKT_BURST];
426 struct rte_mbuf *m_ipv6[MAX_PKT_BURST];
427 uint32_t res_ipv6[MAX_PKT_BURST];
432 char mapped[NB_SOCKETS];
433 struct rte_acl_ctx *acx_ipv4[NB_SOCKETS];
434 struct rte_acl_ctx *acx_ipv6[NB_SOCKETS];
435 #ifdef L3FWDACL_DEBUG
436 struct acl4_rule *rule_ipv4;
437 struct acl6_rule *rule_ipv6;
442 const char *rule_ipv4_name;
443 const char *rule_ipv6_name;
447 const char cb_port_delim[] = ":";
450 print_one_ipv4_rule(struct acl4_rule *rule, int extra)
452 unsigned char a, b, c, d;
454 uint32_t_to_char(rule->field[SRC_FIELD_IPV4].value.u32,
456 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
457 rule->field[SRC_FIELD_IPV4].mask_range.u32);
458 uint32_t_to_char(rule->field[DST_FIELD_IPV4].value.u32,
460 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
461 rule->field[DST_FIELD_IPV4].mask_range.u32);
462 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
463 rule->field[SRCP_FIELD_IPV4].value.u16,
464 rule->field[SRCP_FIELD_IPV4].mask_range.u16,
465 rule->field[DSTP_FIELD_IPV4].value.u16,
466 rule->field[DSTP_FIELD_IPV4].mask_range.u16,
467 rule->field[PROTO_FIELD_IPV4].value.u8,
468 rule->field[PROTO_FIELD_IPV4].mask_range.u8);
470 printf("0x%x-0x%x-0x%x ",
471 rule->data.category_mask,
473 rule->data.userdata);
477 print_one_ipv6_rule(struct acl6_rule *rule, int extra)
479 unsigned char a, b, c, d;
481 uint32_t_to_char(rule->field[SRC1_FIELD_IPV6].value.u32,
483 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
484 uint32_t_to_char(rule->field[SRC2_FIELD_IPV6].value.u32,
486 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
487 uint32_t_to_char(rule->field[SRC3_FIELD_IPV6].value.u32,
489 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
490 uint32_t_to_char(rule->field[SRC4_FIELD_IPV6].value.u32,
492 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
493 rule->field[SRC1_FIELD_IPV6].mask_range.u32
494 + rule->field[SRC2_FIELD_IPV6].mask_range.u32
495 + rule->field[SRC3_FIELD_IPV6].mask_range.u32
496 + rule->field[SRC4_FIELD_IPV6].mask_range.u32);
498 uint32_t_to_char(rule->field[DST1_FIELD_IPV6].value.u32,
500 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
501 uint32_t_to_char(rule->field[DST2_FIELD_IPV6].value.u32,
503 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
504 uint32_t_to_char(rule->field[DST3_FIELD_IPV6].value.u32,
506 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
507 uint32_t_to_char(rule->field[DST4_FIELD_IPV6].value.u32,
509 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
510 rule->field[DST1_FIELD_IPV6].mask_range.u32
511 + rule->field[DST2_FIELD_IPV6].mask_range.u32
512 + rule->field[DST3_FIELD_IPV6].mask_range.u32
513 + rule->field[DST4_FIELD_IPV6].mask_range.u32);
515 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
516 rule->field[SRCP_FIELD_IPV6].value.u16,
517 rule->field[SRCP_FIELD_IPV6].mask_range.u16,
518 rule->field[DSTP_FIELD_IPV6].value.u16,
519 rule->field[DSTP_FIELD_IPV6].mask_range.u16,
520 rule->field[PROTO_FIELD_IPV6].value.u8,
521 rule->field[PROTO_FIELD_IPV6].mask_range.u8);
523 printf("0x%x-0x%x-0x%x ",
524 rule->data.category_mask,
526 rule->data.userdata);
529 /* Bypass comment and empty lines */
531 is_bypass_line(char *buff)
536 if (buff[0] == COMMENT_LEAD_CHAR)
539 while (buff[i] != '\0') {
540 if (!isspace(buff[i]))
547 #ifdef L3FWDACL_DEBUG
549 dump_acl4_rule(struct rte_mbuf *m, uint32_t sig)
551 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
552 unsigned char a, b, c, d;
553 struct rte_ipv4_hdr *ipv4_hdr =
554 rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
555 sizeof(struct rte_ether_hdr));
557 uint32_t_to_char(rte_bswap32(ipv4_hdr->src_addr), &a, &b, &c, &d);
558 printf("Packet Src:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
559 uint32_t_to_char(rte_bswap32(ipv4_hdr->dst_addr), &a, &b, &c, &d);
560 printf("Dst:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
562 printf("Src port:%hu,Dst port:%hu ",
563 rte_bswap16(*(uint16_t *)(ipv4_hdr + 1)),
564 rte_bswap16(*((uint16_t *)(ipv4_hdr + 1) + 1)));
565 printf("hit ACL %d - ", offset);
567 print_one_ipv4_rule(acl_config.rule_ipv4 + offset, 1);
573 dump_acl6_rule(struct rte_mbuf *m, uint32_t sig)
576 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
577 struct rte_ipv6_hdr *ipv6_hdr =
578 rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
579 sizeof(struct rte_ether_hdr));
581 printf("Packet Src");
582 for (i = 0; i < RTE_DIM(ipv6_hdr->src_addr); i += sizeof(uint16_t))
584 ipv6_hdr->src_addr[i], ipv6_hdr->src_addr[i + 1]);
587 for (i = 0; i < RTE_DIM(ipv6_hdr->dst_addr); i += sizeof(uint16_t))
589 ipv6_hdr->dst_addr[i], ipv6_hdr->dst_addr[i + 1]);
591 printf("\nSrc port:%hu,Dst port:%hu ",
592 rte_bswap16(*(uint16_t *)(ipv6_hdr + 1)),
593 rte_bswap16(*((uint16_t *)(ipv6_hdr + 1) + 1)));
594 printf("hit ACL %d - ", offset);
596 print_one_ipv6_rule(acl_config.rule_ipv6 + offset, 1);
600 #endif /* L3FWDACL_DEBUG */
603 dump_ipv4_rules(struct acl4_rule *rule, int num, int extra)
607 for (i = 0; i < num; i++, rule++) {
608 printf("\t%d:", i + 1);
609 print_one_ipv4_rule(rule, extra);
615 dump_ipv6_rules(struct acl6_rule *rule, int num, int extra)
619 for (i = 0; i < num; i++, rule++) {
620 printf("\t%d:", i + 1);
621 print_one_ipv6_rule(rule, extra);
626 #ifdef DO_RFC_1812_CHECKS
628 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
631 struct rte_ipv4_hdr *ipv4_hdr;
632 struct rte_mbuf *pkt = pkts_in[index];
634 if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
635 ipv4_hdr = rte_pktmbuf_mtod_offset(pkt, struct rte_ipv4_hdr *,
636 sizeof(struct rte_ether_hdr));
638 /* Check to make sure the packet is valid (RFC1812) */
639 if (is_valid_ipv4_pkt(ipv4_hdr, pkt->pkt_len) >= 0) {
641 /* Update time to live and header checksum */
642 --(ipv4_hdr->time_to_live);
643 ++(ipv4_hdr->hdr_checksum);
645 /* Fill acl structure */
646 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
647 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
650 /* Not a valid IPv4 packet */
651 rte_pktmbuf_free(pkt);
653 } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
654 /* Fill acl structure */
655 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
656 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
659 /* Unknown type, drop the packet */
660 rte_pktmbuf_free(pkt);
666 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
669 struct rte_mbuf *pkt = pkts_in[index];
671 if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
672 /* Fill acl structure */
673 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
674 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
676 } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
677 /* Fill acl structure */
678 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
679 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
681 /* Unknown type, drop the packet */
682 rte_pktmbuf_free(pkt);
685 #endif /* DO_RFC_1812_CHECKS */
688 prepare_acl_parameter(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
696 /* Prefetch first packets */
697 for (i = 0; i < PREFETCH_OFFSET && i < nb_rx; i++) {
698 rte_prefetch0(rte_pktmbuf_mtod(
699 pkts_in[i], void *));
702 for (i = 0; i < (nb_rx - PREFETCH_OFFSET); i++) {
703 rte_prefetch0(rte_pktmbuf_mtod(pkts_in[
704 i + PREFETCH_OFFSET], void *));
705 prepare_one_packet(pkts_in, acl, i);
708 /* Process left packets */
709 for (; i < nb_rx; i++)
710 prepare_one_packet(pkts_in, acl, i);
714 send_one_packet(struct rte_mbuf *m, uint32_t res)
716 if (likely((res & ACL_DENY_SIGNATURE) == 0 && res != 0)) {
717 /* forward packets */
718 send_single_packet(m,
719 (uint8_t)(res - FWD_PORT_SHIFT));
721 /* in the ACL list, drop it */
722 #ifdef L3FWDACL_DEBUG
723 if ((res & ACL_DENY_SIGNATURE) != 0) {
724 if (RTE_ETH_IS_IPV4_HDR(m->packet_type))
725 dump_acl4_rule(m, res);
726 else if (RTE_ETH_IS_IPV6_HDR(m->packet_type))
727 dump_acl6_rule(m, res);
737 send_packets(struct rte_mbuf **m, uint32_t *res, int num)
741 /* Prefetch first packets */
742 for (i = 0; i < PREFETCH_OFFSET && i < num; i++) {
743 rte_prefetch0(rte_pktmbuf_mtod(
747 for (i = 0; i < (num - PREFETCH_OFFSET); i++) {
748 rte_prefetch0(rte_pktmbuf_mtod(m[
749 i + PREFETCH_OFFSET], void *));
750 send_one_packet(m[i], res[i]);
753 /* Process left packets */
755 send_one_packet(m[i], res[i]);
759 * Parses IPV6 address, exepcts the following format:
760 * XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX (where X - is a hexedecimal digit).
763 parse_ipv6_addr(const char *in, const char **end, uint32_t v[IPV6_ADDR_U32],
766 uint32_t addr[IPV6_ADDR_U16];
768 GET_CB_FIELD(in, addr[0], 16, UINT16_MAX, ':');
769 GET_CB_FIELD(in, addr[1], 16, UINT16_MAX, ':');
770 GET_CB_FIELD(in, addr[2], 16, UINT16_MAX, ':');
771 GET_CB_FIELD(in, addr[3], 16, UINT16_MAX, ':');
772 GET_CB_FIELD(in, addr[4], 16, UINT16_MAX, ':');
773 GET_CB_FIELD(in, addr[5], 16, UINT16_MAX, ':');
774 GET_CB_FIELD(in, addr[6], 16, UINT16_MAX, ':');
775 GET_CB_FIELD(in, addr[7], 16, UINT16_MAX, dlm);
779 v[0] = (addr[0] << 16) + addr[1];
780 v[1] = (addr[2] << 16) + addr[3];
781 v[2] = (addr[4] << 16) + addr[5];
782 v[3] = (addr[6] << 16) + addr[7];
788 parse_ipv6_net(const char *in, struct rte_acl_field field[4])
793 const uint32_t nbu32 = sizeof(uint32_t) * CHAR_BIT;
796 rc = parse_ipv6_addr(in, &mp, v, '/');
801 GET_CB_FIELD(mp, m, 0, CHAR_BIT * sizeof(v), 0);
803 /* put all together. */
804 for (i = 0; i != RTE_DIM(v); i++) {
805 if (m >= (i + 1) * nbu32)
806 field[i].mask_range.u32 = nbu32;
808 field[i].mask_range.u32 = m > (i * nbu32) ?
811 field[i].value.u32 = v[i];
818 parse_cb_ipv6_rule(char *str, struct rte_acl_rule *v, int has_userdata)
821 char *s, *sp, *in[CB_FLD_NUM];
822 static const char *dlm = " \t\n";
823 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
826 for (i = 0; i != dim; i++, s = NULL) {
827 in[i] = strtok_r(s, dlm, &sp);
832 rc = parse_ipv6_net(in[CB_FLD_SRC_ADDR], v->field + SRC1_FIELD_IPV6);
834 acl_log("failed to read source address/mask: %s\n",
835 in[CB_FLD_SRC_ADDR]);
839 rc = parse_ipv6_net(in[CB_FLD_DST_ADDR], v->field + DST1_FIELD_IPV6);
841 acl_log("failed to read destination address/mask: %s\n",
842 in[CB_FLD_DST_ADDR]);
847 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
848 v->field[SRCP_FIELD_IPV6].value.u16,
850 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
851 v->field[SRCP_FIELD_IPV6].mask_range.u16,
854 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
855 sizeof(cb_port_delim)) != 0)
858 /* destination port. */
859 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
860 v->field[DSTP_FIELD_IPV6].value.u16,
862 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
863 v->field[DSTP_FIELD_IPV6].mask_range.u16,
866 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
867 sizeof(cb_port_delim)) != 0)
870 if (v->field[SRCP_FIELD_IPV6].mask_range.u16
871 < v->field[SRCP_FIELD_IPV6].value.u16
872 || v->field[DSTP_FIELD_IPV6].mask_range.u16
873 < v->field[DSTP_FIELD_IPV6].value.u16)
876 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].value.u8,
878 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].mask_range.u8,
882 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata,
889 * Parse ClassBench rules file.
891 * '@'<src_ipv4_addr>'/'<masklen> <space> \
892 * <dst_ipv4_addr>'/'<masklen> <space> \
893 * <src_port_low> <space> ":" <src_port_high> <space> \
894 * <dst_port_low> <space> ":" <dst_port_high> <space> \
898 parse_ipv4_net(const char *in, uint32_t *addr, uint32_t *mask_len)
900 uint8_t a, b, c, d, m;
902 GET_CB_FIELD(in, a, 0, UINT8_MAX, '.');
903 GET_CB_FIELD(in, b, 0, UINT8_MAX, '.');
904 GET_CB_FIELD(in, c, 0, UINT8_MAX, '.');
905 GET_CB_FIELD(in, d, 0, UINT8_MAX, '/');
906 GET_CB_FIELD(in, m, 0, sizeof(uint32_t) * CHAR_BIT, 0);
908 addr[0] = RTE_IPV4(a, b, c, d);
915 parse_cb_ipv4vlan_rule(char *str, struct rte_acl_rule *v, int has_userdata)
918 char *s, *sp, *in[CB_FLD_NUM];
919 static const char *dlm = " \t\n";
920 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
923 for (i = 0; i != dim; i++, s = NULL) {
924 in[i] = strtok_r(s, dlm, &sp);
929 rc = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
930 &v->field[SRC_FIELD_IPV4].value.u32,
931 &v->field[SRC_FIELD_IPV4].mask_range.u32);
933 acl_log("failed to read source address/mask: %s\n",
934 in[CB_FLD_SRC_ADDR]);
938 rc = parse_ipv4_net(in[CB_FLD_DST_ADDR],
939 &v->field[DST_FIELD_IPV4].value.u32,
940 &v->field[DST_FIELD_IPV4].mask_range.u32);
942 acl_log("failed to read destination address/mask: %s\n",
943 in[CB_FLD_DST_ADDR]);
947 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
948 v->field[SRCP_FIELD_IPV4].value.u16,
950 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
951 v->field[SRCP_FIELD_IPV4].mask_range.u16,
954 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
955 sizeof(cb_port_delim)) != 0)
958 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
959 v->field[DSTP_FIELD_IPV4].value.u16,
961 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
962 v->field[DSTP_FIELD_IPV4].mask_range.u16,
965 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
966 sizeof(cb_port_delim)) != 0)
969 if (v->field[SRCP_FIELD_IPV4].mask_range.u16
970 < v->field[SRCP_FIELD_IPV4].value.u16
971 || v->field[DSTP_FIELD_IPV4].mask_range.u16
972 < v->field[DSTP_FIELD_IPV4].value.u16)
975 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].value.u8,
977 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].mask_range.u8,
981 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata, 0,
988 add_rules(const char *rule_path,
989 struct rte_acl_rule **proute_base,
990 unsigned int *proute_num,
991 struct rte_acl_rule **pacl_base,
992 unsigned int *pacl_num, uint32_t rule_size,
993 int (*parser)(char *, struct rte_acl_rule*, int))
995 uint8_t *acl_rules, *route_rules;
996 struct rte_acl_rule *next;
997 unsigned int acl_num = 0, route_num = 0, total_num = 0;
998 unsigned int acl_cnt = 0, route_cnt = 0;
1000 FILE *fh = fopen(rule_path, "rb");
1005 rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
1008 while ((fgets(buff, LINE_MAX, fh) != NULL)) {
1009 if (buff[0] == ROUTE_LEAD_CHAR)
1011 else if (buff[0] == ACL_LEAD_CHAR)
1016 rte_exit(EXIT_FAILURE, "Not find any route entries in %s!\n",
1019 val = fseek(fh, 0, SEEK_SET);
1021 rte_exit(EXIT_FAILURE, "%s: File seek operation failed\n",
1025 acl_rules = calloc(acl_num, rule_size);
1027 if (NULL == acl_rules)
1028 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1031 route_rules = calloc(route_num, rule_size);
1033 if (NULL == route_rules)
1034 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1038 while (fgets(buff, LINE_MAX, fh) != NULL) {
1041 if (is_bypass_line(buff))
1047 if (s == ROUTE_LEAD_CHAR)
1048 next = (struct rte_acl_rule *)(route_rules +
1049 route_cnt * rule_size);
1052 else if (s == ACL_LEAD_CHAR)
1053 next = (struct rte_acl_rule *)(acl_rules +
1054 acl_cnt * rule_size);
1058 rte_exit(EXIT_FAILURE,
1059 "%s Line %u: should start with leading "
1061 rule_path, i, ROUTE_LEAD_CHAR, ACL_LEAD_CHAR);
1063 if (parser(buff + 1, next, s == ROUTE_LEAD_CHAR) != 0)
1064 rte_exit(EXIT_FAILURE,
1065 "%s Line %u: parse rules error\n",
1068 if (s == ROUTE_LEAD_CHAR) {
1069 /* Check the forwarding port number */
1070 if ((enabled_port_mask & (1 << next->data.userdata)) ==
1072 rte_exit(EXIT_FAILURE,
1073 "%s Line %u: fwd number illegal:%u\n",
1074 rule_path, i, next->data.userdata);
1075 next->data.userdata += FWD_PORT_SHIFT;
1078 next->data.userdata = ACL_DENY_SIGNATURE + acl_cnt;
1082 next->data.priority = RTE_ACL_MAX_PRIORITY - total_num;
1083 next->data.category_mask = -1;
1089 *pacl_base = (struct rte_acl_rule *)acl_rules;
1090 *pacl_num = acl_num;
1091 *proute_base = (struct rte_acl_rule *)route_rules;
1092 *proute_num = route_cnt;
1098 dump_acl_config(void)
1100 printf("ACL option are:\n");
1101 printf(OPTION_RULE_IPV4": %s\n", parm_config.rule_ipv4_name);
1102 printf(OPTION_RULE_IPV6": %s\n", parm_config.rule_ipv6_name);
1103 printf(OPTION_SCALAR": %d\n", parm_config.scalar);
1107 check_acl_config(void)
1109 if (parm_config.rule_ipv4_name == NULL) {
1110 acl_log("ACL IPv4 rule file not specified\n");
1112 } else if (parm_config.rule_ipv6_name == NULL) {
1113 acl_log("ACL IPv6 rule file not specified\n");
1120 static struct rte_acl_ctx*
1121 setup_acl(struct rte_acl_rule *route_base,
1122 struct rte_acl_rule *acl_base, unsigned int route_num,
1123 unsigned int acl_num, int ipv6, int socketid)
1125 char name[PATH_MAX];
1126 struct rte_acl_param acl_param;
1127 struct rte_acl_config acl_build_param;
1128 struct rte_acl_ctx *context;
1129 int dim = ipv6 ? RTE_DIM(ipv6_defs) : RTE_DIM(ipv4_defs);
1131 /* Create ACL contexts */
1132 snprintf(name, sizeof(name), "%s%d",
1133 ipv6 ? L3FWD_ACL_IPV6_NAME : L3FWD_ACL_IPV4_NAME,
1136 acl_param.name = name;
1137 acl_param.socket_id = socketid;
1138 acl_param.rule_size = RTE_ACL_RULE_SZ(dim);
1139 acl_param.max_rule_num = MAX_ACL_RULE_NUM;
1141 if ((context = rte_acl_create(&acl_param)) == NULL)
1142 rte_exit(EXIT_FAILURE, "Failed to create ACL context\n");
1144 if (parm_config.scalar && rte_acl_set_ctx_classify(context,
1145 RTE_ACL_CLASSIFY_SCALAR) != 0)
1146 rte_exit(EXIT_FAILURE,
1147 "Failed to setup classify method for ACL context\n");
1149 if (rte_acl_add_rules(context, route_base, route_num) < 0)
1150 rte_exit(EXIT_FAILURE, "add rules failed\n");
1152 if (rte_acl_add_rules(context, acl_base, acl_num) < 0)
1153 rte_exit(EXIT_FAILURE, "add rules failed\n");
1155 /* Perform builds */
1156 memset(&acl_build_param, 0, sizeof(acl_build_param));
1158 acl_build_param.num_categories = DEFAULT_MAX_CATEGORIES;
1159 acl_build_param.num_fields = dim;
1160 memcpy(&acl_build_param.defs, ipv6 ? ipv6_defs : ipv4_defs,
1161 ipv6 ? sizeof(ipv6_defs) : sizeof(ipv4_defs));
1163 if (rte_acl_build(context, &acl_build_param) != 0)
1164 rte_exit(EXIT_FAILURE, "Failed to build ACL trie\n");
1166 rte_acl_dump(context);
1177 struct rte_acl_rule *acl_base_ipv4, *route_base_ipv4,
1178 *acl_base_ipv6, *route_base_ipv6;
1179 unsigned int acl_num_ipv4 = 0, route_num_ipv4 = 0,
1180 acl_num_ipv6 = 0, route_num_ipv6 = 0;
1182 if (check_acl_config() != 0)
1183 rte_exit(EXIT_FAILURE, "Failed to get valid ACL options\n");
1187 /* Load rules from the input file */
1188 if (add_rules(parm_config.rule_ipv4_name, &route_base_ipv4,
1189 &route_num_ipv4, &acl_base_ipv4, &acl_num_ipv4,
1190 sizeof(struct acl4_rule), &parse_cb_ipv4vlan_rule) < 0)
1191 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1193 acl_log("IPv4 Route entries %u:\n", route_num_ipv4);
1194 dump_ipv4_rules((struct acl4_rule *)route_base_ipv4, route_num_ipv4, 1);
1196 acl_log("IPv4 ACL entries %u:\n", acl_num_ipv4);
1197 dump_ipv4_rules((struct acl4_rule *)acl_base_ipv4, acl_num_ipv4, 1);
1199 if (add_rules(parm_config.rule_ipv6_name, &route_base_ipv6,
1201 &acl_base_ipv6, &acl_num_ipv6,
1202 sizeof(struct acl6_rule), &parse_cb_ipv6_rule) < 0)
1203 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1205 acl_log("IPv6 Route entries %u:\n", route_num_ipv6);
1206 dump_ipv6_rules((struct acl6_rule *)route_base_ipv6, route_num_ipv6, 1);
1208 acl_log("IPv6 ACL entries %u:\n", acl_num_ipv6);
1209 dump_ipv6_rules((struct acl6_rule *)acl_base_ipv6, acl_num_ipv6, 1);
1211 memset(&acl_config, 0, sizeof(acl_config));
1213 /* Check sockets a context should be created on */
1215 acl_config.mapped[0] = 1;
1217 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1218 if (rte_lcore_is_enabled(lcore_id) == 0)
1221 socketid = rte_lcore_to_socket_id(lcore_id);
1222 if (socketid >= NB_SOCKETS) {
1223 acl_log("Socket %d of lcore %u is out "
1225 socketid, lcore_id, NB_SOCKETS);
1226 free(route_base_ipv4);
1227 free(route_base_ipv6);
1228 free(acl_base_ipv4);
1229 free(acl_base_ipv6);
1233 acl_config.mapped[socketid] = 1;
1237 for (i = 0; i < NB_SOCKETS; i++) {
1238 if (acl_config.mapped[i]) {
1239 acl_config.acx_ipv4[i] = setup_acl(route_base_ipv4,
1240 acl_base_ipv4, route_num_ipv4, acl_num_ipv4,
1243 acl_config.acx_ipv6[i] = setup_acl(route_base_ipv6,
1244 acl_base_ipv6, route_num_ipv6, acl_num_ipv6,
1249 free(route_base_ipv4);
1250 free(route_base_ipv6);
1252 #ifdef L3FWDACL_DEBUG
1253 acl_config.rule_ipv4 = (struct acl4_rule *)acl_base_ipv4;
1254 acl_config.rule_ipv6 = (struct acl6_rule *)acl_base_ipv6;
1256 free(acl_base_ipv4);
1257 free(acl_base_ipv6);
1263 /***********************end of ACL part******************************/
1266 uint16_t n_rx_queue;
1267 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
1269 uint16_t tx_port_id[RTE_MAX_ETHPORTS];
1270 uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
1271 struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
1272 } __rte_cache_aligned;
1274 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
1276 /* Enqueue a single packet, and send burst if queue is filled */
1278 send_single_packet(struct rte_mbuf *m, uint16_t port)
1281 struct lcore_conf *qconf;
1282 struct rte_ether_hdr *eh;
1284 lcore_id = rte_lcore_id();
1286 /* update src and dst mac*/
1287 eh = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
1288 memcpy(eh, &port_l2hdr[port], sizeof(eh->d_addr) + sizeof(eh->s_addr));
1290 qconf = &lcore_conf[lcore_id];
1291 rte_eth_tx_buffer(port, qconf->tx_queue_id[port],
1292 qconf->tx_buffer[port], m);
1295 #ifdef DO_RFC_1812_CHECKS
1297 is_valid_ipv4_pkt(struct rte_ipv4_hdr *pkt, uint32_t link_len)
1299 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
1301 * 1. The packet length reported by the Link Layer must be large
1302 * enough to hold the minimum length legal IP datagram (20 bytes).
1304 if (link_len < sizeof(struct rte_ipv4_hdr))
1307 /* 2. The IP checksum must be correct. */
1308 /* this is checked in H/W */
1311 * 3. The IP version number must be 4. If the version number is not 4
1312 * then the packet may be another version of IP, such as IPng or
1315 if (((pkt->version_ihl) >> 4) != 4)
1318 * 4. The IP header length field must be large enough to hold the
1319 * minimum length legal IP datagram (20 bytes = 5 words).
1321 if ((pkt->version_ihl & 0xf) < 5)
1325 * 5. The IP total length field must be large enough to hold the IP
1326 * datagram header, whose length is specified in the IP header length
1329 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct rte_ipv4_hdr))
1336 /* main processing loop */
1338 main_loop(__rte_unused void *dummy)
1340 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1342 uint64_t prev_tsc, diff_tsc, cur_tsc;
1346 struct lcore_conf *qconf;
1348 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
1349 / US_PER_S * BURST_TX_DRAIN_US;
1352 lcore_id = rte_lcore_id();
1353 qconf = &lcore_conf[lcore_id];
1354 socketid = rte_lcore_to_socket_id(lcore_id);
1356 if (qconf->n_rx_queue == 0) {
1357 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
1361 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
1363 for (i = 0; i < qconf->n_rx_queue; i++) {
1365 portid = qconf->rx_queue_list[i].port_id;
1366 queueid = qconf->rx_queue_list[i].queue_id;
1367 RTE_LOG(INFO, L3FWD,
1368 " -- lcoreid=%u portid=%u rxqueueid=%hhu\n",
1369 lcore_id, portid, queueid);
1374 cur_tsc = rte_rdtsc();
1377 * TX burst queue drain
1379 diff_tsc = cur_tsc - prev_tsc;
1380 if (unlikely(diff_tsc > drain_tsc)) {
1381 for (i = 0; i < qconf->n_tx_port; ++i) {
1382 portid = qconf->tx_port_id[i];
1383 rte_eth_tx_buffer_flush(portid,
1384 qconf->tx_queue_id[portid],
1385 qconf->tx_buffer[portid]);
1391 * Read packet from RX queues
1393 for (i = 0; i < qconf->n_rx_queue; ++i) {
1395 portid = qconf->rx_queue_list[i].port_id;
1396 queueid = qconf->rx_queue_list[i].queue_id;
1397 nb_rx = rte_eth_rx_burst(portid, queueid,
1398 pkts_burst, MAX_PKT_BURST);
1401 struct acl_search_t acl_search;
1403 prepare_acl_parameter(pkts_burst, &acl_search,
1406 if (acl_search.num_ipv4) {
1408 acl_config.acx_ipv4[socketid],
1409 acl_search.data_ipv4,
1410 acl_search.res_ipv4,
1411 acl_search.num_ipv4,
1412 DEFAULT_MAX_CATEGORIES);
1414 send_packets(acl_search.m_ipv4,
1415 acl_search.res_ipv4,
1416 acl_search.num_ipv4);
1419 if (acl_search.num_ipv6) {
1421 acl_config.acx_ipv6[socketid],
1422 acl_search.data_ipv6,
1423 acl_search.res_ipv6,
1424 acl_search.num_ipv6,
1425 DEFAULT_MAX_CATEGORIES);
1427 send_packets(acl_search.m_ipv6,
1428 acl_search.res_ipv6,
1429 acl_search.num_ipv6);
1437 check_lcore_params(void)
1439 uint8_t queue, lcore;
1443 for (i = 0; i < nb_lcore_params; ++i) {
1444 queue = lcore_params[i].queue_id;
1445 if (queue >= MAX_RX_QUEUE_PER_PORT) {
1446 printf("invalid queue number: %hhu\n", queue);
1449 lcore = lcore_params[i].lcore_id;
1450 if (!rte_lcore_is_enabled(lcore)) {
1451 printf("error: lcore %hhu is not enabled in "
1452 "lcore mask\n", lcore);
1455 socketid = rte_lcore_to_socket_id(lcore);
1456 if (socketid != 0 && numa_on == 0) {
1457 printf("warning: lcore %hhu is on socket %d "
1466 check_port_config(void)
1471 for (i = 0; i < nb_lcore_params; ++i) {
1472 portid = lcore_params[i].port_id;
1474 if ((enabled_port_mask & (1 << portid)) == 0) {
1475 printf("port %u is not enabled in port mask\n", portid);
1478 if (!rte_eth_dev_is_valid_port(portid)) {
1479 printf("port %u is not present on the board\n", portid);
1487 get_port_n_rx_queues(const uint16_t port)
1492 for (i = 0; i < nb_lcore_params; ++i) {
1493 if (lcore_params[i].port_id == port &&
1494 lcore_params[i].queue_id > queue)
1495 queue = lcore_params[i].queue_id;
1497 return (uint8_t)(++queue);
1501 init_lcore_rx_queues(void)
1503 uint16_t i, nb_rx_queue;
1506 for (i = 0; i < nb_lcore_params; ++i) {
1507 lcore = lcore_params[i].lcore_id;
1508 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
1509 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
1510 printf("error: too many queues (%u) for lcore: %u\n",
1511 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
1514 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
1515 lcore_params[i].port_id;
1516 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
1517 lcore_params[i].queue_id;
1518 lcore_conf[lcore].n_rx_queue++;
1526 print_usage(const char *prgname)
1528 printf("%s [EAL options] -- -p PORTMASK -P"
1529 "--"OPTION_RULE_IPV4"=FILE"
1530 "--"OPTION_RULE_IPV6"=FILE"
1531 " [--"OPTION_CONFIG" (port,queue,lcore)[,(port,queue,lcore]]"
1532 " [--"OPTION_ENBJMO" [--max-pkt-len PKTLEN]]\n"
1533 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
1534 " -P : enable promiscuous mode\n"
1535 " --"OPTION_CONFIG": (port,queue,lcore): "
1536 "rx queues configuration\n"
1537 " --"OPTION_NONUMA": optional, disable numa awareness\n"
1538 " --"OPTION_ENBJMO": enable jumbo frame"
1539 " which max packet len is PKTLEN in decimal (64-9600)\n"
1540 " --"OPTION_RULE_IPV4"=FILE: specify the ipv4 rules entries "
1542 "Each rule occupy one line. "
1543 "2 kinds of rules are supported. "
1544 "One is ACL entry at while line leads with character '%c', "
1545 "another is route entry at while line leads with "
1547 " --"OPTION_RULE_IPV6"=FILE: specify the ipv6 rules "
1549 " --"OPTION_SCALAR": Use scalar function to do lookup\n",
1550 prgname, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR);
1554 parse_max_pkt_len(const char *pktlen)
1559 /* parse decimal string */
1560 len = strtoul(pktlen, &end, 10);
1561 if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1571 parse_portmask(const char *portmask)
1576 /* parse hexadecimal string */
1577 pm = strtoul(portmask, &end, 16);
1578 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1585 parse_config(const char *q_arg)
1588 const char *p, *p0 = q_arg;
1596 unsigned long int_fld[_NUM_FLD];
1597 char *str_fld[_NUM_FLD];
1601 nb_lcore_params = 0;
1603 while ((p = strchr(p0, '(')) != NULL) {
1605 if ((p0 = strchr(p, ')')) == NULL)
1609 if (size >= sizeof(s))
1612 snprintf(s, sizeof(s), "%.*s", size, p);
1613 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') !=
1616 for (i = 0; i < _NUM_FLD; i++) {
1618 int_fld[i] = strtoul(str_fld[i], &end, 0);
1619 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1622 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1623 printf("exceeded max number of lcore params: %hu\n",
1627 lcore_params_array[nb_lcore_params].port_id =
1628 (uint8_t)int_fld[FLD_PORT];
1629 lcore_params_array[nb_lcore_params].queue_id =
1630 (uint8_t)int_fld[FLD_QUEUE];
1631 lcore_params_array[nb_lcore_params].lcore_id =
1632 (uint8_t)int_fld[FLD_LCORE];
1635 lcore_params = lcore_params_array;
1640 parse_eth_dest(const char *optarg)
1642 unsigned long portid;
1646 portid = strtoul(optarg, &port_end, 0);
1647 if (errno != 0 || port_end == optarg || *port_end++ != ',')
1648 return "Invalid format";
1649 else if (portid >= RTE_MAX_ETHPORTS)
1650 return "port value exceeds RTE_MAX_ETHPORTS("
1651 RTE_STR(RTE_MAX_ETHPORTS) ")";
1653 if (cmdline_parse_etheraddr(NULL, port_end, &port_l2hdr[portid].d_addr,
1654 sizeof(port_l2hdr[portid].d_addr)) < 0)
1655 return "Invalid ethernet address";
1659 /* Parse the argument given in the command line of the application */
1661 parse_args(int argc, char **argv)
1666 char *prgname = argv[0];
1667 static struct option lgopts[] = {
1668 {OPTION_CONFIG, 1, 0, 0},
1669 {OPTION_NONUMA, 0, 0, 0},
1670 {OPTION_ENBJMO, 0, 0, 0},
1671 {OPTION_RULE_IPV4, 1, 0, 0},
1672 {OPTION_RULE_IPV6, 1, 0, 0},
1673 {OPTION_SCALAR, 0, 0, 0},
1674 {OPTION_ETH_DEST, 1, 0, 0},
1680 while ((opt = getopt_long(argc, argvopt, "p:P",
1681 lgopts, &option_index)) != EOF) {
1686 enabled_port_mask = parse_portmask(optarg);
1687 if (enabled_port_mask == 0) {
1688 printf("invalid portmask\n");
1689 print_usage(prgname);
1694 printf("Promiscuous mode selected\n");
1700 if (!strncmp(lgopts[option_index].name,
1702 sizeof(OPTION_CONFIG))) {
1703 ret = parse_config(optarg);
1705 printf("invalid config\n");
1706 print_usage(prgname);
1711 if (!strncmp(lgopts[option_index].name,
1713 sizeof(OPTION_NONUMA))) {
1714 printf("numa is disabled\n");
1718 if (!strncmp(lgopts[option_index].name,
1719 OPTION_ENBJMO, sizeof(OPTION_ENBJMO))) {
1720 struct option lenopts = {
1727 printf("jumbo frame is enabled\n");
1728 port_conf.rxmode.offloads |=
1729 DEV_RX_OFFLOAD_JUMBO_FRAME;
1730 port_conf.txmode.offloads |=
1731 DEV_TX_OFFLOAD_MULTI_SEGS;
1734 * if no max-pkt-len set, then use the
1735 * default value RTE_ETHER_MAX_LEN
1737 if (0 == getopt_long(argc, argvopt, "",
1738 &lenopts, &option_index)) {
1739 ret = parse_max_pkt_len(optarg);
1741 (ret > MAX_JUMBO_PKT_LEN)) {
1742 printf("invalid packet "
1744 print_usage(prgname);
1747 port_conf.rxmode.max_rx_pkt_len = ret;
1749 printf("set jumbo frame max packet length "
1752 port_conf.rxmode.max_rx_pkt_len);
1755 if (!strncmp(lgopts[option_index].name,
1757 sizeof(OPTION_RULE_IPV4)))
1758 parm_config.rule_ipv4_name = optarg;
1760 if (!strncmp(lgopts[option_index].name,
1762 sizeof(OPTION_RULE_IPV6))) {
1763 parm_config.rule_ipv6_name = optarg;
1766 if (!strncmp(lgopts[option_index].name,
1767 OPTION_SCALAR, sizeof(OPTION_SCALAR)))
1768 parm_config.scalar = 1;
1770 if (!strncmp(lgopts[option_index].name, OPTION_ETH_DEST,
1771 sizeof(OPTION_ETH_DEST))) {
1772 const char *serr = parse_eth_dest(optarg);
1774 printf("invalid %s value:\"%s\": %s\n",
1775 OPTION_ETH_DEST, optarg, serr);
1776 print_usage(prgname);
1784 print_usage(prgname);
1790 argv[optind-1] = prgname;
1793 optind = 1; /* reset getopt lib */
1798 print_ethaddr(const char *name, const struct rte_ether_addr *eth_addr)
1800 char buf[RTE_ETHER_ADDR_FMT_SIZE];
1801 rte_ether_format_addr(buf, RTE_ETHER_ADDR_FMT_SIZE, eth_addr);
1802 printf("%s%s", name, buf);
1806 init_mem(unsigned nb_mbuf)
1812 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1813 if (rte_lcore_is_enabled(lcore_id) == 0)
1817 socketid = rte_lcore_to_socket_id(lcore_id);
1821 if (socketid >= NB_SOCKETS) {
1822 rte_exit(EXIT_FAILURE,
1823 "Socket %d of lcore %u is out of range %d\n",
1824 socketid, lcore_id, NB_SOCKETS);
1826 if (pktmbuf_pool[socketid] == NULL) {
1827 snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
1828 pktmbuf_pool[socketid] =
1829 rte_pktmbuf_pool_create(s, nb_mbuf,
1830 MEMPOOL_CACHE_SIZE, 0,
1831 RTE_MBUF_DEFAULT_BUF_SIZE,
1833 if (pktmbuf_pool[socketid] == NULL)
1834 rte_exit(EXIT_FAILURE,
1835 "Cannot init mbuf pool on socket %d\n",
1838 printf("Allocated mbuf pool on socket %d\n",
1845 /* Check the link status of all ports in up to 9s, and print them finally */
1847 check_all_ports_link_status(uint32_t port_mask)
1849 #define CHECK_INTERVAL 100 /* 100ms */
1850 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1852 uint8_t count, all_ports_up, print_flag = 0;
1853 struct rte_eth_link link;
1855 char link_status_text[RTE_ETH_LINK_MAX_STR_LEN];
1857 printf("\nChecking link status");
1859 for (count = 0; count <= MAX_CHECK_TIME; count++) {
1861 RTE_ETH_FOREACH_DEV(portid) {
1862 if ((port_mask & (1 << portid)) == 0)
1864 memset(&link, 0, sizeof(link));
1865 ret = rte_eth_link_get_nowait(portid, &link);
1868 if (print_flag == 1)
1869 printf("Port %u link get failed: %s\n",
1870 portid, rte_strerror(-ret));
1873 /* print link status if flag set */
1874 if (print_flag == 1) {
1875 rte_eth_link_to_str(link_status_text,
1876 sizeof(link_status_text), &link);
1877 printf("Port %d %s\n", portid,
1881 /* clear all_ports_up flag if any link down */
1882 if (link.link_status == ETH_LINK_DOWN) {
1887 /* after finally printing all link status, get out */
1888 if (print_flag == 1)
1891 if (all_ports_up == 0) {
1894 rte_delay_ms(CHECK_INTERVAL);
1897 /* set the print_flag if all ports up or timeout */
1898 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1906 * build-up default vaues for dest MACs.
1909 set_default_dest_mac(void)
1913 for (i = 0; i != RTE_DIM(port_l2hdr); i++) {
1914 port_l2hdr[i].d_addr.addr_bytes[0] = RTE_ETHER_LOCAL_ADMIN_ADDR;
1915 port_l2hdr[i].d_addr.addr_bytes[5] = i;
1920 main(int argc, char **argv)
1922 struct lcore_conf *qconf;
1923 struct rte_eth_dev_info dev_info;
1924 struct rte_eth_txconf *txconf;
1929 uint32_t n_tx_queue, nb_lcores;
1931 uint8_t nb_rx_queue, queue, socketid;
1934 ret = rte_eal_init(argc, argv);
1936 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1940 set_default_dest_mac();
1942 /* parse application arguments (after the EAL ones) */
1943 ret = parse_args(argc, argv);
1945 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
1947 if (check_lcore_params() < 0)
1948 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1950 ret = init_lcore_rx_queues();
1952 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1954 nb_ports = rte_eth_dev_count_avail();
1956 if (check_port_config() < 0)
1957 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1959 /* Add ACL rules and route entries, build trie */
1960 if (app_acl_init() < 0)
1961 rte_exit(EXIT_FAILURE, "app_acl_init failed\n");
1963 nb_lcores = rte_lcore_count();
1965 /* initialize all ports */
1966 RTE_ETH_FOREACH_DEV(portid) {
1967 struct rte_eth_conf local_port_conf = port_conf;
1969 /* skip ports that are not enabled */
1970 if ((enabled_port_mask & (1 << portid)) == 0) {
1971 printf("\nSkipping disabled port %d\n", portid);
1976 printf("Initializing port %d ... ", portid);
1979 nb_rx_queue = get_port_n_rx_queues(portid);
1980 n_tx_queue = nb_lcores;
1981 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1982 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1983 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1984 nb_rx_queue, (unsigned)n_tx_queue);
1986 ret = rte_eth_dev_info_get(portid, &dev_info);
1988 rte_exit(EXIT_FAILURE,
1989 "Error during getting device (port %u) info: %s\n",
1990 portid, strerror(-ret));
1992 if (dev_info.tx_offload_capa & DEV_TX_OFFLOAD_MBUF_FAST_FREE)
1993 local_port_conf.txmode.offloads |=
1994 DEV_TX_OFFLOAD_MBUF_FAST_FREE;
1996 local_port_conf.rx_adv_conf.rss_conf.rss_hf &=
1997 dev_info.flow_type_rss_offloads;
1998 if (local_port_conf.rx_adv_conf.rss_conf.rss_hf !=
1999 port_conf.rx_adv_conf.rss_conf.rss_hf) {
2000 printf("Port %u modified RSS hash function based on hardware support,"
2001 "requested:%#"PRIx64" configured:%#"PRIx64"\n",
2003 port_conf.rx_adv_conf.rss_conf.rss_hf,
2004 local_port_conf.rx_adv_conf.rss_conf.rss_hf);
2007 ret = rte_eth_dev_configure(portid, nb_rx_queue,
2008 (uint16_t)n_tx_queue, &local_port_conf);
2010 rte_exit(EXIT_FAILURE,
2011 "Cannot configure device: err=%d, port=%d\n",
2014 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
2017 rte_exit(EXIT_FAILURE,
2018 "rte_eth_dev_adjust_nb_rx_tx_desc: err=%d, port=%d\n",
2021 ret = rte_eth_macaddr_get(portid, &port_l2hdr[portid].s_addr);
2023 rte_exit(EXIT_FAILURE,
2024 "rte_eth_macaddr_get: err=%d, port=%d\n",
2027 print_ethaddr("Dst MAC:", &port_l2hdr[portid].d_addr);
2028 print_ethaddr(", Src MAC:", &port_l2hdr[portid].s_addr);
2032 ret = init_mem(NB_MBUF);
2034 rte_exit(EXIT_FAILURE, "init_mem failed\n");
2036 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2037 if (rte_lcore_is_enabled(lcore_id) == 0)
2040 /* Initialize TX buffers */
2041 qconf = &lcore_conf[lcore_id];
2042 qconf->tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
2043 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
2044 rte_eth_dev_socket_id(portid));
2045 if (qconf->tx_buffer[portid] == NULL)
2046 rte_exit(EXIT_FAILURE, "Can't allocate tx buffer for port %u\n",
2049 rte_eth_tx_buffer_init(qconf->tx_buffer[portid], MAX_PKT_BURST);
2052 /* init one TX queue per couple (lcore,port) */
2054 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2055 if (rte_lcore_is_enabled(lcore_id) == 0)
2059 socketid = (uint8_t)
2060 rte_lcore_to_socket_id(lcore_id);
2064 printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
2067 ret = rte_eth_dev_info_get(portid, &dev_info);
2069 rte_exit(EXIT_FAILURE,
2070 "Error during getting device (port %u) info: %s\n",
2071 portid, strerror(-ret));
2073 txconf = &dev_info.default_txconf;
2074 txconf->offloads = local_port_conf.txmode.offloads;
2075 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
2078 rte_exit(EXIT_FAILURE,
2079 "rte_eth_tx_queue_setup: err=%d, "
2080 "port=%d\n", ret, portid);
2082 qconf = &lcore_conf[lcore_id];
2083 qconf->tx_queue_id[portid] = queueid;
2086 qconf->tx_port_id[qconf->n_tx_port] = portid;
2092 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2093 if (rte_lcore_is_enabled(lcore_id) == 0)
2095 qconf = &lcore_conf[lcore_id];
2096 printf("\nInitializing rx queues on lcore %u ... ", lcore_id);
2098 /* init RX queues */
2099 for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
2100 struct rte_eth_rxconf rxq_conf;
2102 portid = qconf->rx_queue_list[queue].port_id;
2103 queueid = qconf->rx_queue_list[queue].queue_id;
2106 socketid = (uint8_t)
2107 rte_lcore_to_socket_id(lcore_id);
2111 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
2114 ret = rte_eth_dev_info_get(portid, &dev_info);
2116 rte_exit(EXIT_FAILURE,
2117 "Error during getting device (port %u) info: %s\n",
2118 portid, strerror(-ret));
2120 rxq_conf = dev_info.default_rxconf;
2121 rxq_conf.offloads = port_conf.rxmode.offloads;
2122 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
2123 socketid, &rxq_conf,
2124 pktmbuf_pool[socketid]);
2126 rte_exit(EXIT_FAILURE,
2127 "rte_eth_rx_queue_setup: err=%d,"
2128 "port=%d\n", ret, portid);
2135 RTE_ETH_FOREACH_DEV(portid) {
2136 if ((enabled_port_mask & (1 << portid)) == 0)
2140 ret = rte_eth_dev_start(portid);
2142 rte_exit(EXIT_FAILURE,
2143 "rte_eth_dev_start: err=%d, port=%d\n",
2147 * If enabled, put device in promiscuous mode.
2148 * This allows IO forwarding mode to forward packets
2149 * to itself through 2 cross-connected ports of the
2152 if (promiscuous_on) {
2153 ret = rte_eth_promiscuous_enable(portid);
2155 rte_exit(EXIT_FAILURE,
2156 "rte_eth_promiscuous_enable: err=%s, port=%u\n",
2157 rte_strerror(-ret), portid);
2161 check_all_ports_link_status(enabled_port_mask);
2163 /* launch per-lcore init on every lcore */
2164 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
2165 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
2166 if (rte_eal_wait_lcore(lcore_id) < 0)