4 * Copyright(c) 2010-2016 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 #if RTE_LOG_LEVEL >= RTE_LOG_DEBUG
76 #define L3FWDACL_DEBUG
78 #define DO_RFC_1812_CHECKS
80 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
82 #define MAX_JUMBO_PKT_LEN 9600
84 #define MEMPOOL_CACHE_SIZE 256
87 * This expression is used to calculate the number of mbufs needed
88 * depending on user input, taking into account memory for rx and tx hardware
89 * rings, cache per lcore and mtable per port per lcore.
90 * RTE_MAX is used to ensure that NB_MBUF never goes below a
91 * minimum value of 8192
94 #define NB_MBUF RTE_MAX(\
95 (nb_ports * nb_rx_queue*RTE_TEST_RX_DESC_DEFAULT + \
96 nb_ports * nb_lcores * MAX_PKT_BURST + \
97 nb_ports * n_tx_queue * RTE_TEST_TX_DESC_DEFAULT + \
98 nb_lcores * MEMPOOL_CACHE_SIZE), \
101 #define MAX_PKT_BURST 32
102 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
106 /* Configure how many packets ahead to prefetch, when reading packets */
107 #define PREFETCH_OFFSET 3
110 * Configurable number of RX/TX ring descriptors
112 #define RTE_TEST_RX_DESC_DEFAULT 128
113 #define RTE_TEST_TX_DESC_DEFAULT 512
114 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
115 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
117 /* ethernet addresses of ports */
118 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
120 /* mask of enabled ports */
121 static uint32_t enabled_port_mask;
122 static int promiscuous_on; /**< Ports set in promiscuous mode off by default. */
123 static int numa_on = 1; /**< NUMA is enabled by default. */
125 struct lcore_rx_queue {
128 } __rte_cache_aligned;
130 #define MAX_RX_QUEUE_PER_LCORE 16
131 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
132 #define MAX_RX_QUEUE_PER_PORT 128
134 #define MAX_LCORE_PARAMS 1024
135 struct lcore_params {
139 } __rte_cache_aligned;
141 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
142 static struct lcore_params lcore_params_array_default[] = {
154 static struct lcore_params *lcore_params = lcore_params_array_default;
155 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
156 sizeof(lcore_params_array_default[0]);
158 static struct rte_eth_conf port_conf = {
160 .mq_mode = ETH_MQ_RX_RSS,
161 .max_rx_pkt_len = ETHER_MAX_LEN,
163 .header_split = 0, /**< Header Split disabled */
164 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
165 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
166 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
167 .hw_strip_crc = 0, /**< CRC stripped by hardware */
172 .rss_hf = ETH_RSS_IP | ETH_RSS_UDP |
173 ETH_RSS_TCP | ETH_RSS_SCTP,
177 .mq_mode = ETH_MQ_TX_NONE,
181 static struct rte_mempool *pktmbuf_pool[NB_SOCKETS];
183 /***********************start of ACL part******************************/
184 #ifdef DO_RFC_1812_CHECKS
186 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len);
189 send_single_packet(struct rte_mbuf *m, uint8_t port);
191 #define MAX_ACL_RULE_NUM 100000
192 #define DEFAULT_MAX_CATEGORIES 1
193 #define L3FWD_ACL_IPV4_NAME "l3fwd-acl-ipv4"
194 #define L3FWD_ACL_IPV6_NAME "l3fwd-acl-ipv6"
195 #define ACL_LEAD_CHAR ('@')
196 #define ROUTE_LEAD_CHAR ('R')
197 #define COMMENT_LEAD_CHAR ('#')
198 #define OPTION_CONFIG "config"
199 #define OPTION_NONUMA "no-numa"
200 #define OPTION_ENBJMO "enable-jumbo"
201 #define OPTION_RULE_IPV4 "rule_ipv4"
202 #define OPTION_RULE_IPV6 "rule_ipv6"
203 #define OPTION_SCALAR "scalar"
204 #define ACL_DENY_SIGNATURE 0xf0000000
205 #define RTE_LOGTYPE_L3FWDACL RTE_LOGTYPE_USER3
206 #define acl_log(format, ...) RTE_LOG(ERR, L3FWDACL, format, ##__VA_ARGS__)
207 #define uint32_t_to_char(ip, a, b, c, d) do {\
208 *a = (unsigned char)(ip >> 24 & 0xff);\
209 *b = (unsigned char)(ip >> 16 & 0xff);\
210 *c = (unsigned char)(ip >> 8 & 0xff);\
211 *d = (unsigned char)(ip & 0xff);\
213 #define OFF_ETHHEAD (sizeof(struct ether_hdr))
214 #define OFF_IPV42PROTO (offsetof(struct ipv4_hdr, next_proto_id))
215 #define OFF_IPV62PROTO (offsetof(struct ipv6_hdr, proto))
216 #define MBUF_IPV4_2PROTO(m) \
217 rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV42PROTO)
218 #define MBUF_IPV6_2PROTO(m) \
219 rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV62PROTO)
221 #define GET_CB_FIELD(in, fd, base, lim, dlm) do { \
225 val = strtoul((in), &end, (base)); \
226 if (errno != 0 || end[0] != (dlm) || val > (lim)) \
228 (fd) = (typeof(fd))val; \
233 * ACL rules should have higher priorities than route ones to ensure ACL rule
234 * always be found when input packets have multi-matches in the database.
235 * A exception case is performance measure, which can define route rules with
236 * higher priority and route rules will always be returned in each lookup.
237 * Reserve range from ACL_RULE_PRIORITY_MAX + 1 to
238 * RTE_ACL_MAX_PRIORITY for route entries in performance measure
240 #define ACL_RULE_PRIORITY_MAX 0x10000000
243 * Forward port info save in ACL lib starts from 1
244 * since ACL assume 0 is invalid.
245 * So, need add 1 when saving and minus 1 when forwarding packets.
247 #define FWD_PORT_SHIFT 1
250 * Rule and trace formats definitions.
263 * That effectively defines order of IPV4VLAN classifications:
265 * - VLAN (TAG and DOMAIN)
268 * - PORTS (SRC and DST)
271 RTE_ACL_IPV4VLAN_PROTO,
272 RTE_ACL_IPV4VLAN_VLAN,
273 RTE_ACL_IPV4VLAN_SRC,
274 RTE_ACL_IPV4VLAN_DST,
275 RTE_ACL_IPV4VLAN_PORTS,
279 struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
281 .type = RTE_ACL_FIELD_TYPE_BITMASK,
282 .size = sizeof(uint8_t),
283 .field_index = PROTO_FIELD_IPV4,
284 .input_index = RTE_ACL_IPV4VLAN_PROTO,
288 .type = RTE_ACL_FIELD_TYPE_MASK,
289 .size = sizeof(uint32_t),
290 .field_index = SRC_FIELD_IPV4,
291 .input_index = RTE_ACL_IPV4VLAN_SRC,
292 .offset = offsetof(struct ipv4_hdr, src_addr) -
293 offsetof(struct ipv4_hdr, next_proto_id),
296 .type = RTE_ACL_FIELD_TYPE_MASK,
297 .size = sizeof(uint32_t),
298 .field_index = DST_FIELD_IPV4,
299 .input_index = RTE_ACL_IPV4VLAN_DST,
300 .offset = offsetof(struct ipv4_hdr, dst_addr) -
301 offsetof(struct ipv4_hdr, next_proto_id),
304 .type = RTE_ACL_FIELD_TYPE_RANGE,
305 .size = sizeof(uint16_t),
306 .field_index = SRCP_FIELD_IPV4,
307 .input_index = RTE_ACL_IPV4VLAN_PORTS,
308 .offset = sizeof(struct ipv4_hdr) -
309 offsetof(struct ipv4_hdr, next_proto_id),
312 .type = RTE_ACL_FIELD_TYPE_RANGE,
313 .size = sizeof(uint16_t),
314 .field_index = DSTP_FIELD_IPV4,
315 .input_index = RTE_ACL_IPV4VLAN_PORTS,
316 .offset = sizeof(struct ipv4_hdr) -
317 offsetof(struct ipv4_hdr, next_proto_id) +
322 #define IPV6_ADDR_LEN 16
323 #define IPV6_ADDR_U16 (IPV6_ADDR_LEN / sizeof(uint16_t))
324 #define IPV6_ADDR_U32 (IPV6_ADDR_LEN / sizeof(uint32_t))
341 struct rte_acl_field_def ipv6_defs[NUM_FIELDS_IPV6] = {
343 .type = RTE_ACL_FIELD_TYPE_BITMASK,
344 .size = sizeof(uint8_t),
345 .field_index = PROTO_FIELD_IPV6,
346 .input_index = PROTO_FIELD_IPV6,
350 .type = RTE_ACL_FIELD_TYPE_MASK,
351 .size = sizeof(uint32_t),
352 .field_index = SRC1_FIELD_IPV6,
353 .input_index = SRC1_FIELD_IPV6,
354 .offset = offsetof(struct ipv6_hdr, src_addr) -
355 offsetof(struct ipv6_hdr, proto),
358 .type = RTE_ACL_FIELD_TYPE_MASK,
359 .size = sizeof(uint32_t),
360 .field_index = SRC2_FIELD_IPV6,
361 .input_index = SRC2_FIELD_IPV6,
362 .offset = offsetof(struct ipv6_hdr, src_addr) -
363 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
366 .type = RTE_ACL_FIELD_TYPE_MASK,
367 .size = sizeof(uint32_t),
368 .field_index = SRC3_FIELD_IPV6,
369 .input_index = SRC3_FIELD_IPV6,
370 .offset = offsetof(struct ipv6_hdr, src_addr) -
371 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
374 .type = RTE_ACL_FIELD_TYPE_MASK,
375 .size = sizeof(uint32_t),
376 .field_index = SRC4_FIELD_IPV6,
377 .input_index = SRC4_FIELD_IPV6,
378 .offset = offsetof(struct ipv6_hdr, src_addr) -
379 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
382 .type = RTE_ACL_FIELD_TYPE_MASK,
383 .size = sizeof(uint32_t),
384 .field_index = DST1_FIELD_IPV6,
385 .input_index = DST1_FIELD_IPV6,
386 .offset = offsetof(struct ipv6_hdr, dst_addr)
387 - offsetof(struct ipv6_hdr, proto),
390 .type = RTE_ACL_FIELD_TYPE_MASK,
391 .size = sizeof(uint32_t),
392 .field_index = DST2_FIELD_IPV6,
393 .input_index = DST2_FIELD_IPV6,
394 .offset = offsetof(struct ipv6_hdr, dst_addr) -
395 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
398 .type = RTE_ACL_FIELD_TYPE_MASK,
399 .size = sizeof(uint32_t),
400 .field_index = DST3_FIELD_IPV6,
401 .input_index = DST3_FIELD_IPV6,
402 .offset = offsetof(struct ipv6_hdr, dst_addr) -
403 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
406 .type = RTE_ACL_FIELD_TYPE_MASK,
407 .size = sizeof(uint32_t),
408 .field_index = DST4_FIELD_IPV6,
409 .input_index = DST4_FIELD_IPV6,
410 .offset = offsetof(struct ipv6_hdr, dst_addr) -
411 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
414 .type = RTE_ACL_FIELD_TYPE_RANGE,
415 .size = sizeof(uint16_t),
416 .field_index = SRCP_FIELD_IPV6,
417 .input_index = SRCP_FIELD_IPV6,
418 .offset = sizeof(struct ipv6_hdr) -
419 offsetof(struct ipv6_hdr, proto),
422 .type = RTE_ACL_FIELD_TYPE_RANGE,
423 .size = sizeof(uint16_t),
424 .field_index = DSTP_FIELD_IPV6,
425 .input_index = SRCP_FIELD_IPV6,
426 .offset = sizeof(struct ipv6_hdr) -
427 offsetof(struct ipv6_hdr, proto) + sizeof(uint16_t),
436 CB_FLD_SRC_PORT_HIGH,
439 CB_FLD_DST_PORT_HIGH,
445 RTE_ACL_RULE_DEF(acl4_rule, RTE_DIM(ipv4_defs));
446 RTE_ACL_RULE_DEF(acl6_rule, RTE_DIM(ipv6_defs));
448 struct acl_search_t {
449 const uint8_t *data_ipv4[MAX_PKT_BURST];
450 struct rte_mbuf *m_ipv4[MAX_PKT_BURST];
451 uint32_t res_ipv4[MAX_PKT_BURST];
454 const uint8_t *data_ipv6[MAX_PKT_BURST];
455 struct rte_mbuf *m_ipv6[MAX_PKT_BURST];
456 uint32_t res_ipv6[MAX_PKT_BURST];
461 char mapped[NB_SOCKETS];
462 struct rte_acl_ctx *acx_ipv4[NB_SOCKETS];
463 struct rte_acl_ctx *acx_ipv6[NB_SOCKETS];
464 #ifdef L3FWDACL_DEBUG
465 struct acl4_rule *rule_ipv4;
466 struct acl6_rule *rule_ipv6;
471 const char *rule_ipv4_name;
472 const char *rule_ipv6_name;
476 const char cb_port_delim[] = ":";
479 print_one_ipv4_rule(struct acl4_rule *rule, int extra)
481 unsigned char a, b, c, d;
483 uint32_t_to_char(rule->field[SRC_FIELD_IPV4].value.u32,
485 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
486 rule->field[SRC_FIELD_IPV4].mask_range.u32);
487 uint32_t_to_char(rule->field[DST_FIELD_IPV4].value.u32,
489 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
490 rule->field[DST_FIELD_IPV4].mask_range.u32);
491 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
492 rule->field[SRCP_FIELD_IPV4].value.u16,
493 rule->field[SRCP_FIELD_IPV4].mask_range.u16,
494 rule->field[DSTP_FIELD_IPV4].value.u16,
495 rule->field[DSTP_FIELD_IPV4].mask_range.u16,
496 rule->field[PROTO_FIELD_IPV4].value.u8,
497 rule->field[PROTO_FIELD_IPV4].mask_range.u8);
499 printf("0x%x-0x%x-0x%x ",
500 rule->data.category_mask,
502 rule->data.userdata);
506 print_one_ipv6_rule(struct acl6_rule *rule, int extra)
508 unsigned char a, b, c, d;
510 uint32_t_to_char(rule->field[SRC1_FIELD_IPV6].value.u32,
512 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
513 uint32_t_to_char(rule->field[SRC2_FIELD_IPV6].value.u32,
515 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
516 uint32_t_to_char(rule->field[SRC3_FIELD_IPV6].value.u32,
518 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
519 uint32_t_to_char(rule->field[SRC4_FIELD_IPV6].value.u32,
521 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
522 rule->field[SRC1_FIELD_IPV6].mask_range.u32
523 + rule->field[SRC2_FIELD_IPV6].mask_range.u32
524 + rule->field[SRC3_FIELD_IPV6].mask_range.u32
525 + rule->field[SRC4_FIELD_IPV6].mask_range.u32);
527 uint32_t_to_char(rule->field[DST1_FIELD_IPV6].value.u32,
529 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
530 uint32_t_to_char(rule->field[DST2_FIELD_IPV6].value.u32,
532 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
533 uint32_t_to_char(rule->field[DST3_FIELD_IPV6].value.u32,
535 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
536 uint32_t_to_char(rule->field[DST4_FIELD_IPV6].value.u32,
538 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
539 rule->field[DST1_FIELD_IPV6].mask_range.u32
540 + rule->field[DST2_FIELD_IPV6].mask_range.u32
541 + rule->field[DST3_FIELD_IPV6].mask_range.u32
542 + rule->field[DST4_FIELD_IPV6].mask_range.u32);
544 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
545 rule->field[SRCP_FIELD_IPV6].value.u16,
546 rule->field[SRCP_FIELD_IPV6].mask_range.u16,
547 rule->field[DSTP_FIELD_IPV6].value.u16,
548 rule->field[DSTP_FIELD_IPV6].mask_range.u16,
549 rule->field[PROTO_FIELD_IPV6].value.u8,
550 rule->field[PROTO_FIELD_IPV6].mask_range.u8);
552 printf("0x%x-0x%x-0x%x ",
553 rule->data.category_mask,
555 rule->data.userdata);
558 /* Bypass comment and empty lines */
560 is_bypass_line(char *buff)
565 if (buff[0] == COMMENT_LEAD_CHAR)
568 while (buff[i] != '\0') {
569 if (!isspace(buff[i]))
576 #ifdef L3FWDACL_DEBUG
578 dump_acl4_rule(struct rte_mbuf *m, uint32_t sig)
580 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
581 unsigned char a, b, c, d;
582 struct ipv4_hdr *ipv4_hdr = rte_pktmbuf_mtod_offset(m,
584 sizeof(struct ether_hdr));
586 uint32_t_to_char(rte_bswap32(ipv4_hdr->src_addr), &a, &b, &c, &d);
587 printf("Packet Src:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
588 uint32_t_to_char(rte_bswap32(ipv4_hdr->dst_addr), &a, &b, &c, &d);
589 printf("Dst:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
591 printf("Src port:%hu,Dst port:%hu ",
592 rte_bswap16(*(uint16_t *)(ipv4_hdr + 1)),
593 rte_bswap16(*((uint16_t *)(ipv4_hdr + 1) + 1)));
594 printf("hit ACL %d - ", offset);
596 print_one_ipv4_rule(acl_config.rule_ipv4 + offset, 1);
602 dump_acl6_rule(struct rte_mbuf *m, uint32_t sig)
605 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
606 struct ipv6_hdr *ipv6_hdr = rte_pktmbuf_mtod_offset(m,
608 sizeof(struct ether_hdr));
610 printf("Packet Src");
611 for (i = 0; i < RTE_DIM(ipv6_hdr->src_addr); i += sizeof(uint16_t))
613 ipv6_hdr->src_addr[i], ipv6_hdr->src_addr[i + 1]);
616 for (i = 0; i < RTE_DIM(ipv6_hdr->dst_addr); i += sizeof(uint16_t))
618 ipv6_hdr->dst_addr[i], ipv6_hdr->dst_addr[i + 1]);
620 printf("\nSrc port:%hu,Dst port:%hu ",
621 rte_bswap16(*(uint16_t *)(ipv6_hdr + 1)),
622 rte_bswap16(*((uint16_t *)(ipv6_hdr + 1) + 1)));
623 printf("hit ACL %d - ", offset);
625 print_one_ipv6_rule(acl_config.rule_ipv6 + offset, 1);
629 #endif /* L3FWDACL_DEBUG */
632 dump_ipv4_rules(struct acl4_rule *rule, int num, int extra)
636 for (i = 0; i < num; i++, rule++) {
637 printf("\t%d:", i + 1);
638 print_one_ipv4_rule(rule, extra);
644 dump_ipv6_rules(struct acl6_rule *rule, int num, int extra)
648 for (i = 0; i < num; i++, rule++) {
649 printf("\t%d:", i + 1);
650 print_one_ipv6_rule(rule, extra);
655 #ifdef DO_RFC_1812_CHECKS
657 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
660 struct ipv4_hdr *ipv4_hdr;
661 struct rte_mbuf *pkt = pkts_in[index];
663 if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
664 ipv4_hdr = rte_pktmbuf_mtod_offset(pkt, struct ipv4_hdr *,
665 sizeof(struct ether_hdr));
667 /* Check to make sure the packet is valid (RFC1812) */
668 if (is_valid_ipv4_pkt(ipv4_hdr, pkt->pkt_len) >= 0) {
670 /* Update time to live and header checksum */
671 --(ipv4_hdr->time_to_live);
672 ++(ipv4_hdr->hdr_checksum);
674 /* Fill acl structure */
675 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
676 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
679 /* Not a valid IPv4 packet */
680 rte_pktmbuf_free(pkt);
682 } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
683 /* Fill acl structure */
684 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
685 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
688 /* Unknown type, drop the packet */
689 rte_pktmbuf_free(pkt);
695 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
698 struct rte_mbuf *pkt = pkts_in[index];
700 if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
701 /* Fill acl structure */
702 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
703 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
705 } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
706 /* Fill acl structure */
707 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
708 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
710 /* Unknown type, drop the packet */
711 rte_pktmbuf_free(pkt);
714 #endif /* DO_RFC_1812_CHECKS */
717 prepare_acl_parameter(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
725 /* Prefetch first packets */
726 for (i = 0; i < PREFETCH_OFFSET && i < nb_rx; i++) {
727 rte_prefetch0(rte_pktmbuf_mtod(
728 pkts_in[i], void *));
731 for (i = 0; i < (nb_rx - PREFETCH_OFFSET); i++) {
732 rte_prefetch0(rte_pktmbuf_mtod(pkts_in[
733 i + PREFETCH_OFFSET], void *));
734 prepare_one_packet(pkts_in, acl, i);
737 /* Process left packets */
738 for (; i < nb_rx; i++)
739 prepare_one_packet(pkts_in, acl, i);
743 send_one_packet(struct rte_mbuf *m, uint32_t res)
745 if (likely((res & ACL_DENY_SIGNATURE) == 0 && res != 0)) {
746 /* forward packets */
747 send_single_packet(m,
748 (uint8_t)(res - FWD_PORT_SHIFT));
750 /* in the ACL list, drop it */
751 #ifdef L3FWDACL_DEBUG
752 if ((res & ACL_DENY_SIGNATURE) != 0) {
753 if (RTE_ETH_IS_IPV4_HDR(m->packet_type))
754 dump_acl4_rule(m, res);
755 else if (RTE_ETH_IS_IPV6_HDR(m->packet_type))
756 dump_acl6_rule(m, res);
766 send_packets(struct rte_mbuf **m, uint32_t *res, int num)
770 /* Prefetch first packets */
771 for (i = 0; i < PREFETCH_OFFSET && i < num; i++) {
772 rte_prefetch0(rte_pktmbuf_mtod(
776 for (i = 0; i < (num - PREFETCH_OFFSET); i++) {
777 rte_prefetch0(rte_pktmbuf_mtod(m[
778 i + PREFETCH_OFFSET], void *));
779 send_one_packet(m[i], res[i]);
782 /* Process left packets */
784 send_one_packet(m[i], res[i]);
788 * Parses IPV6 address, exepcts the following format:
789 * XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX (where X - is a hexedecimal digit).
792 parse_ipv6_addr(const char *in, const char **end, uint32_t v[IPV6_ADDR_U32],
795 uint32_t addr[IPV6_ADDR_U16];
797 GET_CB_FIELD(in, addr[0], 16, UINT16_MAX, ':');
798 GET_CB_FIELD(in, addr[1], 16, UINT16_MAX, ':');
799 GET_CB_FIELD(in, addr[2], 16, UINT16_MAX, ':');
800 GET_CB_FIELD(in, addr[3], 16, UINT16_MAX, ':');
801 GET_CB_FIELD(in, addr[4], 16, UINT16_MAX, ':');
802 GET_CB_FIELD(in, addr[5], 16, UINT16_MAX, ':');
803 GET_CB_FIELD(in, addr[6], 16, UINT16_MAX, ':');
804 GET_CB_FIELD(in, addr[7], 16, UINT16_MAX, dlm);
808 v[0] = (addr[0] << 16) + addr[1];
809 v[1] = (addr[2] << 16) + addr[3];
810 v[2] = (addr[4] << 16) + addr[5];
811 v[3] = (addr[6] << 16) + addr[7];
817 parse_ipv6_net(const char *in, struct rte_acl_field field[4])
822 const uint32_t nbu32 = sizeof(uint32_t) * CHAR_BIT;
825 rc = parse_ipv6_addr(in, &mp, v, '/');
830 GET_CB_FIELD(mp, m, 0, CHAR_BIT * sizeof(v), 0);
832 /* put all together. */
833 for (i = 0; i != RTE_DIM(v); i++) {
834 if (m >= (i + 1) * nbu32)
835 field[i].mask_range.u32 = nbu32;
837 field[i].mask_range.u32 = m > (i * nbu32) ?
840 field[i].value.u32 = v[i];
847 parse_cb_ipv6_rule(char *str, struct rte_acl_rule *v, int has_userdata)
850 char *s, *sp, *in[CB_FLD_NUM];
851 static const char *dlm = " \t\n";
852 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
855 for (i = 0; i != dim; i++, s = NULL) {
856 in[i] = strtok_r(s, dlm, &sp);
861 rc = parse_ipv6_net(in[CB_FLD_SRC_ADDR], v->field + SRC1_FIELD_IPV6);
863 acl_log("failed to read source address/mask: %s\n",
864 in[CB_FLD_SRC_ADDR]);
868 rc = parse_ipv6_net(in[CB_FLD_DST_ADDR], v->field + DST1_FIELD_IPV6);
870 acl_log("failed to read destination address/mask: %s\n",
871 in[CB_FLD_DST_ADDR]);
876 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
877 v->field[SRCP_FIELD_IPV6].value.u16,
879 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
880 v->field[SRCP_FIELD_IPV6].mask_range.u16,
883 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
884 sizeof(cb_port_delim)) != 0)
887 /* destination port. */
888 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
889 v->field[DSTP_FIELD_IPV6].value.u16,
891 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
892 v->field[DSTP_FIELD_IPV6].mask_range.u16,
895 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
896 sizeof(cb_port_delim)) != 0)
899 if (v->field[SRCP_FIELD_IPV6].mask_range.u16
900 < v->field[SRCP_FIELD_IPV6].value.u16
901 || v->field[DSTP_FIELD_IPV6].mask_range.u16
902 < v->field[DSTP_FIELD_IPV6].value.u16)
905 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].value.u8,
907 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].mask_range.u8,
911 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata,
918 * Parse ClassBench rules file.
920 * '@'<src_ipv4_addr>'/'<masklen> <space> \
921 * <dst_ipv4_addr>'/'<masklen> <space> \
922 * <src_port_low> <space> ":" <src_port_high> <space> \
923 * <dst_port_low> <space> ":" <dst_port_high> <space> \
927 parse_ipv4_net(const char *in, uint32_t *addr, uint32_t *mask_len)
929 uint8_t a, b, c, d, m;
931 GET_CB_FIELD(in, a, 0, UINT8_MAX, '.');
932 GET_CB_FIELD(in, b, 0, UINT8_MAX, '.');
933 GET_CB_FIELD(in, c, 0, UINT8_MAX, '.');
934 GET_CB_FIELD(in, d, 0, UINT8_MAX, '/');
935 GET_CB_FIELD(in, m, 0, sizeof(uint32_t) * CHAR_BIT, 0);
937 addr[0] = IPv4(a, b, c, d);
944 parse_cb_ipv4vlan_rule(char *str, struct rte_acl_rule *v, int has_userdata)
947 char *s, *sp, *in[CB_FLD_NUM];
948 static const char *dlm = " \t\n";
949 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
952 for (i = 0; i != dim; i++, s = NULL) {
953 in[i] = strtok_r(s, dlm, &sp);
958 rc = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
959 &v->field[SRC_FIELD_IPV4].value.u32,
960 &v->field[SRC_FIELD_IPV4].mask_range.u32);
962 acl_log("failed to read source address/mask: %s\n",
963 in[CB_FLD_SRC_ADDR]);
967 rc = parse_ipv4_net(in[CB_FLD_DST_ADDR],
968 &v->field[DST_FIELD_IPV4].value.u32,
969 &v->field[DST_FIELD_IPV4].mask_range.u32);
971 acl_log("failed to read destination address/mask: %s\n",
972 in[CB_FLD_DST_ADDR]);
976 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
977 v->field[SRCP_FIELD_IPV4].value.u16,
979 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
980 v->field[SRCP_FIELD_IPV4].mask_range.u16,
983 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
984 sizeof(cb_port_delim)) != 0)
987 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
988 v->field[DSTP_FIELD_IPV4].value.u16,
990 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
991 v->field[DSTP_FIELD_IPV4].mask_range.u16,
994 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
995 sizeof(cb_port_delim)) != 0)
998 if (v->field[SRCP_FIELD_IPV4].mask_range.u16
999 < v->field[SRCP_FIELD_IPV4].value.u16
1000 || v->field[DSTP_FIELD_IPV4].mask_range.u16
1001 < v->field[DSTP_FIELD_IPV4].value.u16)
1004 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].value.u8,
1006 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].mask_range.u8,
1010 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata, 0,
1017 add_rules(const char *rule_path,
1018 struct rte_acl_rule **proute_base,
1019 unsigned int *proute_num,
1020 struct rte_acl_rule **pacl_base,
1021 unsigned int *pacl_num, uint32_t rule_size,
1022 int (*parser)(char *, struct rte_acl_rule*, int))
1024 uint8_t *acl_rules, *route_rules;
1025 struct rte_acl_rule *next;
1026 unsigned int acl_num = 0, route_num = 0, total_num = 0;
1027 unsigned int acl_cnt = 0, route_cnt = 0;
1028 char buff[LINE_MAX];
1029 FILE *fh = fopen(rule_path, "rb");
1033 rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
1036 while ((fgets(buff, LINE_MAX, fh) != NULL)) {
1037 if (buff[0] == ROUTE_LEAD_CHAR)
1039 else if (buff[0] == ACL_LEAD_CHAR)
1044 rte_exit(EXIT_FAILURE, "Not find any route entries in %s!\n",
1047 fseek(fh, 0, SEEK_SET);
1049 acl_rules = calloc(acl_num, rule_size);
1051 if (NULL == acl_rules)
1052 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1055 route_rules = calloc(route_num, rule_size);
1057 if (NULL == route_rules)
1058 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1062 while (fgets(buff, LINE_MAX, fh) != NULL) {
1065 if (is_bypass_line(buff))
1071 if (s == ROUTE_LEAD_CHAR)
1072 next = (struct rte_acl_rule *)(route_rules +
1073 route_cnt * rule_size);
1076 else if (s == ACL_LEAD_CHAR)
1077 next = (struct rte_acl_rule *)(acl_rules +
1078 acl_cnt * rule_size);
1082 rte_exit(EXIT_FAILURE,
1083 "%s Line %u: should start with leading "
1085 rule_path, i, ROUTE_LEAD_CHAR, ACL_LEAD_CHAR);
1087 if (parser(buff + 1, next, s == ROUTE_LEAD_CHAR) != 0)
1088 rte_exit(EXIT_FAILURE,
1089 "%s Line %u: parse rules error\n",
1092 if (s == ROUTE_LEAD_CHAR) {
1093 /* Check the forwarding port number */
1094 if ((enabled_port_mask & (1 << next->data.userdata)) ==
1096 rte_exit(EXIT_FAILURE,
1097 "%s Line %u: fwd number illegal:%u\n",
1098 rule_path, i, next->data.userdata);
1099 next->data.userdata += FWD_PORT_SHIFT;
1102 next->data.userdata = ACL_DENY_SIGNATURE + acl_cnt;
1106 next->data.priority = RTE_ACL_MAX_PRIORITY - total_num;
1107 next->data.category_mask = -1;
1113 *pacl_base = (struct rte_acl_rule *)acl_rules;
1114 *pacl_num = acl_num;
1115 *proute_base = (struct rte_acl_rule *)route_rules;
1116 *proute_num = route_cnt;
1122 dump_acl_config(void)
1124 printf("ACL option are:\n");
1125 printf(OPTION_RULE_IPV4": %s\n", parm_config.rule_ipv4_name);
1126 printf(OPTION_RULE_IPV6": %s\n", parm_config.rule_ipv6_name);
1127 printf(OPTION_SCALAR": %d\n", parm_config.scalar);
1131 check_acl_config(void)
1133 if (parm_config.rule_ipv4_name == NULL) {
1134 acl_log("ACL IPv4 rule file not specified\n");
1136 } else if (parm_config.rule_ipv6_name == NULL) {
1137 acl_log("ACL IPv6 rule file not specified\n");
1144 static struct rte_acl_ctx*
1145 setup_acl(struct rte_acl_rule *route_base,
1146 struct rte_acl_rule *acl_base, unsigned int route_num,
1147 unsigned int acl_num, int ipv6, int socketid)
1149 char name[PATH_MAX];
1150 struct rte_acl_param acl_param;
1151 struct rte_acl_config acl_build_param;
1152 struct rte_acl_ctx *context;
1153 int dim = ipv6 ? RTE_DIM(ipv6_defs) : RTE_DIM(ipv4_defs);
1155 /* Create ACL contexts */
1156 snprintf(name, sizeof(name), "%s%d",
1157 ipv6 ? L3FWD_ACL_IPV6_NAME : L3FWD_ACL_IPV4_NAME,
1160 acl_param.name = name;
1161 acl_param.socket_id = socketid;
1162 acl_param.rule_size = RTE_ACL_RULE_SZ(dim);
1163 acl_param.max_rule_num = MAX_ACL_RULE_NUM;
1165 if ((context = rte_acl_create(&acl_param)) == NULL)
1166 rte_exit(EXIT_FAILURE, "Failed to create ACL context\n");
1168 if (parm_config.scalar && rte_acl_set_ctx_classify(context,
1169 RTE_ACL_CLASSIFY_SCALAR) != 0)
1170 rte_exit(EXIT_FAILURE,
1171 "Failed to setup classify method for ACL context\n");
1173 if (rte_acl_add_rules(context, route_base, route_num) < 0)
1174 rte_exit(EXIT_FAILURE, "add rules failed\n");
1176 if (rte_acl_add_rules(context, acl_base, acl_num) < 0)
1177 rte_exit(EXIT_FAILURE, "add rules failed\n");
1179 /* Perform builds */
1180 memset(&acl_build_param, 0, sizeof(acl_build_param));
1182 acl_build_param.num_categories = DEFAULT_MAX_CATEGORIES;
1183 acl_build_param.num_fields = dim;
1184 memcpy(&acl_build_param.defs, ipv6 ? ipv6_defs : ipv4_defs,
1185 ipv6 ? sizeof(ipv6_defs) : sizeof(ipv4_defs));
1187 if (rte_acl_build(context, &acl_build_param) != 0)
1188 rte_exit(EXIT_FAILURE, "Failed to build ACL trie\n");
1190 rte_acl_dump(context);
1201 struct rte_acl_rule *acl_base_ipv4, *route_base_ipv4,
1202 *acl_base_ipv6, *route_base_ipv6;
1203 unsigned int acl_num_ipv4 = 0, route_num_ipv4 = 0,
1204 acl_num_ipv6 = 0, route_num_ipv6 = 0;
1206 if (check_acl_config() != 0)
1207 rte_exit(EXIT_FAILURE, "Failed to get valid ACL options\n");
1211 /* Load rules from the input file */
1212 if (add_rules(parm_config.rule_ipv4_name, &route_base_ipv4,
1213 &route_num_ipv4, &acl_base_ipv4, &acl_num_ipv4,
1214 sizeof(struct acl4_rule), &parse_cb_ipv4vlan_rule) < 0)
1215 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1217 acl_log("IPv4 Route entries %u:\n", route_num_ipv4);
1218 dump_ipv4_rules((struct acl4_rule *)route_base_ipv4, route_num_ipv4, 1);
1220 acl_log("IPv4 ACL entries %u:\n", acl_num_ipv4);
1221 dump_ipv4_rules((struct acl4_rule *)acl_base_ipv4, acl_num_ipv4, 1);
1223 if (add_rules(parm_config.rule_ipv6_name, &route_base_ipv6,
1225 &acl_base_ipv6, &acl_num_ipv6,
1226 sizeof(struct acl6_rule), &parse_cb_ipv6_rule) < 0)
1227 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1229 acl_log("IPv6 Route entries %u:\n", route_num_ipv6);
1230 dump_ipv6_rules((struct acl6_rule *)route_base_ipv6, route_num_ipv6, 1);
1232 acl_log("IPv6 ACL entries %u:\n", acl_num_ipv6);
1233 dump_ipv6_rules((struct acl6_rule *)acl_base_ipv6, acl_num_ipv6, 1);
1235 memset(&acl_config, 0, sizeof(acl_config));
1237 /* Check sockets a context should be created on */
1239 acl_config.mapped[0] = 1;
1241 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1242 if (rte_lcore_is_enabled(lcore_id) == 0)
1245 socketid = rte_lcore_to_socket_id(lcore_id);
1246 if (socketid >= NB_SOCKETS) {
1247 acl_log("Socket %d of lcore %u is out "
1249 socketid, lcore_id, NB_SOCKETS);
1250 free(route_base_ipv4);
1251 free(route_base_ipv6);
1252 free(acl_base_ipv4);
1253 free(acl_base_ipv6);
1257 acl_config.mapped[socketid] = 1;
1261 for (i = 0; i < NB_SOCKETS; i++) {
1262 if (acl_config.mapped[i]) {
1263 acl_config.acx_ipv4[i] = setup_acl(route_base_ipv4,
1264 acl_base_ipv4, route_num_ipv4, acl_num_ipv4,
1267 acl_config.acx_ipv6[i] = setup_acl(route_base_ipv6,
1268 acl_base_ipv6, route_num_ipv6, acl_num_ipv6,
1273 free(route_base_ipv4);
1274 free(route_base_ipv6);
1276 #ifdef L3FWDACL_DEBUG
1277 acl_config.rule_ipv4 = (struct acl4_rule *)acl_base_ipv4;
1278 acl_config.rule_ipv6 = (struct acl6_rule *)acl_base_ipv6;
1280 free(acl_base_ipv4);
1281 free(acl_base_ipv6);
1287 /***********************end of ACL part******************************/
1290 uint16_t n_rx_queue;
1291 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
1293 uint16_t tx_port_id[RTE_MAX_ETHPORTS];
1294 uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
1295 struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
1296 } __rte_cache_aligned;
1298 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
1300 /* Enqueue a single packet, and send burst if queue is filled */
1302 send_single_packet(struct rte_mbuf *m, uint8_t port)
1305 struct lcore_conf *qconf;
1307 lcore_id = rte_lcore_id();
1309 qconf = &lcore_conf[lcore_id];
1310 rte_eth_tx_buffer(port, qconf->tx_queue_id[port],
1311 qconf->tx_buffer[port], m);
1314 #ifdef DO_RFC_1812_CHECKS
1316 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
1318 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
1320 * 1. The packet length reported by the Link Layer must be large
1321 * enough to hold the minimum length legal IP datagram (20 bytes).
1323 if (link_len < sizeof(struct ipv4_hdr))
1326 /* 2. The IP checksum must be correct. */
1327 /* this is checked in H/W */
1330 * 3. The IP version number must be 4. If the version number is not 4
1331 * then the packet may be another version of IP, such as IPng or
1334 if (((pkt->version_ihl) >> 4) != 4)
1337 * 4. The IP header length field must be large enough to hold the
1338 * minimum length legal IP datagram (20 bytes = 5 words).
1340 if ((pkt->version_ihl & 0xf) < 5)
1344 * 5. The IP total length field must be large enough to hold the IP
1345 * datagram header, whose length is specified in the IP header length
1348 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
1355 /* main processing loop */
1357 main_loop(__attribute__((unused)) void *dummy)
1359 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1361 uint64_t prev_tsc, diff_tsc, cur_tsc;
1363 uint8_t portid, queueid;
1364 struct lcore_conf *qconf;
1366 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
1367 / US_PER_S * BURST_TX_DRAIN_US;
1370 lcore_id = rte_lcore_id();
1371 qconf = &lcore_conf[lcore_id];
1372 socketid = rte_lcore_to_socket_id(lcore_id);
1374 if (qconf->n_rx_queue == 0) {
1375 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
1379 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
1381 for (i = 0; i < qconf->n_rx_queue; i++) {
1383 portid = qconf->rx_queue_list[i].port_id;
1384 queueid = qconf->rx_queue_list[i].queue_id;
1385 RTE_LOG(INFO, L3FWD,
1386 " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n",
1387 lcore_id, portid, queueid);
1392 cur_tsc = rte_rdtsc();
1395 * TX burst queue drain
1397 diff_tsc = cur_tsc - prev_tsc;
1398 if (unlikely(diff_tsc > drain_tsc)) {
1399 for (i = 0; i < qconf->n_tx_port; ++i) {
1400 portid = qconf->tx_port_id[i];
1401 rte_eth_tx_buffer_flush(portid,
1402 qconf->tx_queue_id[portid],
1403 qconf->tx_buffer[portid]);
1409 * Read packet from RX queues
1411 for (i = 0; i < qconf->n_rx_queue; ++i) {
1413 portid = qconf->rx_queue_list[i].port_id;
1414 queueid = qconf->rx_queue_list[i].queue_id;
1415 nb_rx = rte_eth_rx_burst(portid, queueid,
1416 pkts_burst, MAX_PKT_BURST);
1419 struct acl_search_t acl_search;
1421 prepare_acl_parameter(pkts_burst, &acl_search,
1424 if (acl_search.num_ipv4) {
1426 acl_config.acx_ipv4[socketid],
1427 acl_search.data_ipv4,
1428 acl_search.res_ipv4,
1429 acl_search.num_ipv4,
1430 DEFAULT_MAX_CATEGORIES);
1432 send_packets(acl_search.m_ipv4,
1433 acl_search.res_ipv4,
1434 acl_search.num_ipv4);
1437 if (acl_search.num_ipv6) {
1439 acl_config.acx_ipv6[socketid],
1440 acl_search.data_ipv6,
1441 acl_search.res_ipv6,
1442 acl_search.num_ipv6,
1443 DEFAULT_MAX_CATEGORIES);
1445 send_packets(acl_search.m_ipv6,
1446 acl_search.res_ipv6,
1447 acl_search.num_ipv6);
1455 check_lcore_params(void)
1457 uint8_t queue, lcore;
1461 for (i = 0; i < nb_lcore_params; ++i) {
1462 queue = lcore_params[i].queue_id;
1463 if (queue >= MAX_RX_QUEUE_PER_PORT) {
1464 printf("invalid queue number: %hhu\n", queue);
1467 lcore = lcore_params[i].lcore_id;
1468 if (!rte_lcore_is_enabled(lcore)) {
1469 printf("error: lcore %hhu is not enabled in "
1470 "lcore mask\n", lcore);
1473 socketid = rte_lcore_to_socket_id(lcore);
1474 if (socketid != 0 && numa_on == 0) {
1475 printf("warning: lcore %hhu is on socket %d "
1484 check_port_config(const unsigned nb_ports)
1489 for (i = 0; i < nb_lcore_params; ++i) {
1490 portid = lcore_params[i].port_id;
1492 if ((enabled_port_mask & (1 << portid)) == 0) {
1493 printf("port %u is not enabled in port mask\n", portid);
1496 if (portid >= nb_ports) {
1497 printf("port %u is not present on the board\n", portid);
1505 get_port_n_rx_queues(const uint8_t port)
1510 for (i = 0; i < nb_lcore_params; ++i) {
1511 if (lcore_params[i].port_id == port &&
1512 lcore_params[i].queue_id > queue)
1513 queue = lcore_params[i].queue_id;
1515 return (uint8_t)(++queue);
1519 init_lcore_rx_queues(void)
1521 uint16_t i, nb_rx_queue;
1524 for (i = 0; i < nb_lcore_params; ++i) {
1525 lcore = lcore_params[i].lcore_id;
1526 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
1527 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
1528 printf("error: too many queues (%u) for lcore: %u\n",
1529 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
1532 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
1533 lcore_params[i].port_id;
1534 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
1535 lcore_params[i].queue_id;
1536 lcore_conf[lcore].n_rx_queue++;
1544 print_usage(const char *prgname)
1546 printf("%s [EAL options] -- -p PORTMASK -P"
1547 "--"OPTION_RULE_IPV4"=FILE"
1548 "--"OPTION_RULE_IPV6"=FILE"
1549 " [--"OPTION_CONFIG" (port,queue,lcore)[,(port,queue,lcore]]"
1550 " [--"OPTION_ENBJMO" [--max-pkt-len PKTLEN]]\n"
1551 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
1552 " -P : enable promiscuous mode\n"
1553 " --"OPTION_CONFIG": (port,queue,lcore): "
1554 "rx queues configuration\n"
1555 " --"OPTION_NONUMA": optional, disable numa awareness\n"
1556 " --"OPTION_ENBJMO": enable jumbo frame"
1557 " which max packet len is PKTLEN in decimal (64-9600)\n"
1558 " --"OPTION_RULE_IPV4"=FILE: specify the ipv4 rules entries "
1560 "Each rule occupy one line. "
1561 "2 kinds of rules are supported. "
1562 "One is ACL entry at while line leads with character '%c', "
1563 "another is route entry at while line leads with "
1565 " --"OPTION_RULE_IPV6"=FILE: specify the ipv6 rules "
1567 " --"OPTION_SCALAR": Use scalar function to do lookup\n",
1568 prgname, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR);
1572 parse_max_pkt_len(const char *pktlen)
1577 /* parse decimal string */
1578 len = strtoul(pktlen, &end, 10);
1579 if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1589 parse_portmask(const char *portmask)
1594 /* parse hexadecimal string */
1595 pm = strtoul(portmask, &end, 16);
1596 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1606 parse_config(const char *q_arg)
1609 const char *p, *p0 = q_arg;
1617 unsigned long int_fld[_NUM_FLD];
1618 char *str_fld[_NUM_FLD];
1622 nb_lcore_params = 0;
1624 while ((p = strchr(p0, '(')) != NULL) {
1626 if ((p0 = strchr(p, ')')) == NULL)
1630 if (size >= sizeof(s))
1633 snprintf(s, sizeof(s), "%.*s", size, p);
1634 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') !=
1637 for (i = 0; i < _NUM_FLD; i++) {
1639 int_fld[i] = strtoul(str_fld[i], &end, 0);
1640 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1643 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1644 printf("exceeded max number of lcore params: %hu\n",
1648 lcore_params_array[nb_lcore_params].port_id =
1649 (uint8_t)int_fld[FLD_PORT];
1650 lcore_params_array[nb_lcore_params].queue_id =
1651 (uint8_t)int_fld[FLD_QUEUE];
1652 lcore_params_array[nb_lcore_params].lcore_id =
1653 (uint8_t)int_fld[FLD_LCORE];
1656 lcore_params = lcore_params_array;
1660 /* Parse the argument given in the command line of the application */
1662 parse_args(int argc, char **argv)
1667 char *prgname = argv[0];
1668 static struct option lgopts[] = {
1669 {OPTION_CONFIG, 1, 0, 0},
1670 {OPTION_NONUMA, 0, 0, 0},
1671 {OPTION_ENBJMO, 0, 0, 0},
1672 {OPTION_RULE_IPV4, 1, 0, 0},
1673 {OPTION_RULE_IPV6, 1, 0, 0},
1674 {OPTION_SCALAR, 0, 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.jumbo_frame = 1;
1731 * if no max-pkt-len set, then use the
1732 * default value ETHER_MAX_LEN
1734 if (0 == getopt_long(argc, argvopt, "",
1735 &lenopts, &option_index)) {
1736 ret = parse_max_pkt_len(optarg);
1738 (ret > MAX_JUMBO_PKT_LEN)) {
1739 printf("invalid packet "
1741 print_usage(prgname);
1744 port_conf.rxmode.max_rx_pkt_len = ret;
1746 printf("set jumbo frame max packet length "
1749 port_conf.rxmode.max_rx_pkt_len);
1752 if (!strncmp(lgopts[option_index].name,
1754 sizeof(OPTION_RULE_IPV4)))
1755 parm_config.rule_ipv4_name = optarg;
1757 if (!strncmp(lgopts[option_index].name,
1759 sizeof(OPTION_RULE_IPV6))) {
1760 parm_config.rule_ipv6_name = optarg;
1763 if (!strncmp(lgopts[option_index].name,
1764 OPTION_SCALAR, sizeof(OPTION_SCALAR)))
1765 parm_config.scalar = 1;
1771 print_usage(prgname);
1777 argv[optind-1] = prgname;
1780 optind = 0; /* reset getopt lib */
1785 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
1787 char buf[ETHER_ADDR_FMT_SIZE];
1788 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
1789 printf("%s%s", name, buf);
1793 init_mem(unsigned nb_mbuf)
1799 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1800 if (rte_lcore_is_enabled(lcore_id) == 0)
1804 socketid = rte_lcore_to_socket_id(lcore_id);
1808 if (socketid >= NB_SOCKETS) {
1809 rte_exit(EXIT_FAILURE,
1810 "Socket %d of lcore %u is out of range %d\n",
1811 socketid, lcore_id, NB_SOCKETS);
1813 if (pktmbuf_pool[socketid] == NULL) {
1814 snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
1815 pktmbuf_pool[socketid] =
1816 rte_pktmbuf_pool_create(s, nb_mbuf,
1817 MEMPOOL_CACHE_SIZE, 0,
1818 RTE_MBUF_DEFAULT_BUF_SIZE,
1820 if (pktmbuf_pool[socketid] == NULL)
1821 rte_exit(EXIT_FAILURE,
1822 "Cannot init mbuf pool on socket %d\n",
1825 printf("Allocated mbuf pool on socket %d\n",
1832 /* Check the link status of all ports in up to 9s, and print them finally */
1834 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
1836 #define CHECK_INTERVAL 100 /* 100ms */
1837 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1838 uint8_t portid, count, all_ports_up, print_flag = 0;
1839 struct rte_eth_link link;
1841 printf("\nChecking link status");
1843 for (count = 0; count <= MAX_CHECK_TIME; count++) {
1845 for (portid = 0; portid < port_num; portid++) {
1846 if ((port_mask & (1 << portid)) == 0)
1848 memset(&link, 0, sizeof(link));
1849 rte_eth_link_get_nowait(portid, &link);
1850 /* print link status if flag set */
1851 if (print_flag == 1) {
1852 if (link.link_status)
1853 printf("Port %d Link Up - speed %u "
1854 "Mbps - %s\n", (uint8_t)portid,
1855 (unsigned)link.link_speed,
1856 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1857 ("full-duplex") : ("half-duplex\n"));
1859 printf("Port %d Link Down\n",
1863 /* clear all_ports_up flag if any link down */
1864 if (link.link_status == ETH_LINK_DOWN) {
1869 /* after finally printing all link status, get out */
1870 if (print_flag == 1)
1873 if (all_ports_up == 0) {
1876 rte_delay_ms(CHECK_INTERVAL);
1879 /* set the print_flag if all ports up or timeout */
1880 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1888 main(int argc, char **argv)
1890 struct lcore_conf *qconf;
1891 struct rte_eth_dev_info dev_info;
1892 struct rte_eth_txconf *txconf;
1897 uint32_t n_tx_queue, nb_lcores;
1898 uint8_t portid, nb_rx_queue, queue, socketid;
1901 ret = rte_eal_init(argc, argv);
1903 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1907 /* parse application arguments (after the EAL ones) */
1908 ret = parse_args(argc, argv);
1910 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
1912 if (check_lcore_params() < 0)
1913 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1915 ret = init_lcore_rx_queues();
1917 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1919 nb_ports = rte_eth_dev_count();
1921 if (check_port_config(nb_ports) < 0)
1922 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1924 /* Add ACL rules and route entries, build trie */
1925 if (app_acl_init() < 0)
1926 rte_exit(EXIT_FAILURE, "app_acl_init failed\n");
1928 nb_lcores = rte_lcore_count();
1930 /* initialize all ports */
1931 for (portid = 0; portid < nb_ports; portid++) {
1932 /* skip ports that are not enabled */
1933 if ((enabled_port_mask & (1 << portid)) == 0) {
1934 printf("\nSkipping disabled port %d\n", portid);
1939 printf("Initializing port %d ... ", portid);
1942 nb_rx_queue = get_port_n_rx_queues(portid);
1943 n_tx_queue = nb_lcores;
1944 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1945 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1946 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1947 nb_rx_queue, (unsigned)n_tx_queue);
1948 ret = rte_eth_dev_configure(portid, nb_rx_queue,
1949 (uint16_t)n_tx_queue, &port_conf);
1951 rte_exit(EXIT_FAILURE,
1952 "Cannot configure device: err=%d, port=%d\n",
1955 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1956 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1960 ret = init_mem(NB_MBUF);
1962 rte_exit(EXIT_FAILURE, "init_mem failed\n");
1964 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1965 if (rte_lcore_is_enabled(lcore_id) == 0)
1968 /* Initialize TX buffers */
1969 qconf = &lcore_conf[lcore_id];
1970 qconf->tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
1971 RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
1972 rte_eth_dev_socket_id(portid));
1973 if (qconf->tx_buffer[portid] == NULL)
1974 rte_exit(EXIT_FAILURE, "Can't allocate tx buffer for port %u\n",
1977 rte_eth_tx_buffer_init(qconf->tx_buffer[portid], MAX_PKT_BURST);
1980 /* init one TX queue per couple (lcore,port) */
1982 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1983 if (rte_lcore_is_enabled(lcore_id) == 0)
1987 socketid = (uint8_t)
1988 rte_lcore_to_socket_id(lcore_id);
1992 printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
1995 rte_eth_dev_info_get(portid, &dev_info);
1996 txconf = &dev_info.default_txconf;
1997 if (port_conf.rxmode.jumbo_frame)
1998 txconf->txq_flags = 0;
1999 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
2002 rte_exit(EXIT_FAILURE,
2003 "rte_eth_tx_queue_setup: err=%d, "
2004 "port=%d\n", ret, portid);
2006 qconf = &lcore_conf[lcore_id];
2007 qconf->tx_queue_id[portid] = queueid;
2010 qconf->tx_port_id[qconf->n_tx_port] = portid;
2016 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2017 if (rte_lcore_is_enabled(lcore_id) == 0)
2019 qconf = &lcore_conf[lcore_id];
2020 printf("\nInitializing rx queues on lcore %u ... ", lcore_id);
2022 /* init RX queues */
2023 for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
2024 portid = qconf->rx_queue_list[queue].port_id;
2025 queueid = qconf->rx_queue_list[queue].queue_id;
2028 socketid = (uint8_t)
2029 rte_lcore_to_socket_id(lcore_id);
2033 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
2036 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
2038 pktmbuf_pool[socketid]);
2040 rte_exit(EXIT_FAILURE,
2041 "rte_eth_rx_queue_setup: err=%d,"
2042 "port=%d\n", ret, portid);
2049 for (portid = 0; portid < nb_ports; portid++) {
2050 if ((enabled_port_mask & (1 << portid)) == 0)
2054 ret = rte_eth_dev_start(portid);
2056 rte_exit(EXIT_FAILURE,
2057 "rte_eth_dev_start: err=%d, port=%d\n",
2061 * If enabled, put device in promiscuous mode.
2062 * This allows IO forwarding mode to forward packets
2063 * to itself through 2 cross-connected ports of the
2067 rte_eth_promiscuous_enable(portid);
2070 check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
2072 /* launch per-lcore init on every lcore */
2073 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
2074 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
2075 if (rte_eal_wait_lcore(lcore_id) < 0)