4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
38 #include <sys/types.h>
40 #include <sys/queue.h>
45 #include <rte_common.h>
46 #include <rte_byteorder.h>
48 #include <rte_memory.h>
49 #include <rte_memcpy.h>
50 #include <rte_memzone.h>
52 #include <rte_per_lcore.h>
53 #include <rte_launch.h>
54 #include <rte_atomic.h>
55 #include <rte_cycles.h>
56 #include <rte_prefetch.h>
57 #include <rte_lcore.h>
58 #include <rte_per_lcore.h>
59 #include <rte_branch_prediction.h>
60 #include <rte_interrupts.h>
62 #include <rte_random.h>
63 #include <rte_debug.h>
64 #include <rte_ether.h>
65 #include <rte_ethdev.h>
67 #include <rte_mempool.h>
72 #include <rte_string_fns.h>
75 #define DO_RFC_1812_CHECKS
77 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
79 #define MAX_JUMBO_PKT_LEN 9600
81 #define MEMPOOL_CACHE_SIZE 256
84 * This expression is used to calculate the number of mbufs needed
85 * depending on user input, taking into account memory for rx and tx hardware
86 * rings, cache per lcore and mtable per port per lcore.
87 * RTE_MAX is used to ensure that NB_MBUF never goes below a
88 * minimum value of 8192
91 #define NB_MBUF RTE_MAX(\
92 (nb_ports * nb_rx_queue*RTE_TEST_RX_DESC_DEFAULT + \
93 nb_ports * nb_lcores * MAX_PKT_BURST + \
94 nb_ports * n_tx_queue * RTE_TEST_TX_DESC_DEFAULT + \
95 nb_lcores * MEMPOOL_CACHE_SIZE), \
98 #define MAX_PKT_BURST 32
99 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
103 /* Configure how many packets ahead to prefetch, when reading packets */
104 #define PREFETCH_OFFSET 3
107 * Configurable number of RX/TX ring descriptors
109 #define RTE_TEST_RX_DESC_DEFAULT 128
110 #define RTE_TEST_TX_DESC_DEFAULT 512
111 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
112 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
114 /* ethernet addresses of ports */
115 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
117 /* mask of enabled ports */
118 static uint32_t enabled_port_mask;
119 static int promiscuous_on; /**< Ports set in promiscuous mode off by default. */
120 static int numa_on = 1; /**< NUMA is enabled by default. */
124 struct rte_mbuf *m_table[MAX_PKT_BURST];
127 struct lcore_rx_queue {
130 } __rte_cache_aligned;
132 #define MAX_RX_QUEUE_PER_LCORE 16
133 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
134 #define MAX_RX_QUEUE_PER_PORT 128
136 #define MAX_LCORE_PARAMS 1024
137 struct lcore_params {
141 } __rte_cache_aligned;
143 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
144 static struct lcore_params lcore_params_array_default[] = {
156 static struct lcore_params *lcore_params = lcore_params_array_default;
157 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
158 sizeof(lcore_params_array_default[0]);
160 static struct rte_eth_conf port_conf = {
162 .mq_mode = ETH_MQ_RX_RSS,
163 .max_rx_pkt_len = ETHER_MAX_LEN,
165 .header_split = 0, /**< Header Split disabled */
166 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
167 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
168 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
169 .hw_strip_crc = 0, /**< CRC stripped by hardware */
174 .rss_hf = ETH_RSS_IP | ETH_RSS_UDP |
175 ETH_RSS_TCP | ETH_RSS_SCTP,
179 .mq_mode = ETH_MQ_TX_NONE,
183 static struct rte_mempool *pktmbuf_pool[NB_SOCKETS];
185 /***********************start of ACL part******************************/
186 #ifdef DO_RFC_1812_CHECKS
188 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len);
191 send_single_packet(struct rte_mbuf *m, uint8_t port);
193 #define MAX_ACL_RULE_NUM 100000
194 #define DEFAULT_MAX_CATEGORIES 1
195 #define L3FWD_ACL_IPV4_NAME "l3fwd-acl-ipv4"
196 #define L3FWD_ACL_IPV6_NAME "l3fwd-acl-ipv6"
197 #define ACL_LEAD_CHAR ('@')
198 #define ROUTE_LEAD_CHAR ('R')
199 #define COMMENT_LEAD_CHAR ('#')
200 #define OPTION_CONFIG "config"
201 #define OPTION_NONUMA "no-numa"
202 #define OPTION_ENBJMO "enable-jumbo"
203 #define OPTION_RULE_IPV4 "rule_ipv4"
204 #define OPTION_RULE_IPV6 "rule_ipv6"
205 #define OPTION_SCALAR "scalar"
206 #define ACL_DENY_SIGNATURE 0xf0000000
207 #define RTE_LOGTYPE_L3FWDACL RTE_LOGTYPE_USER3
208 #define acl_log(format, ...) RTE_LOG(ERR, L3FWDACL, format, ##__VA_ARGS__)
209 #define uint32_t_to_char(ip, a, b, c, d) do {\
210 *a = (unsigned char)(ip >> 24 & 0xff);\
211 *b = (unsigned char)(ip >> 16 & 0xff);\
212 *c = (unsigned char)(ip >> 8 & 0xff);\
213 *d = (unsigned char)(ip & 0xff);\
215 #define OFF_ETHHEAD (sizeof(struct ether_hdr))
216 #define OFF_IPV42PROTO (offsetof(struct ipv4_hdr, next_proto_id))
217 #define OFF_IPV62PROTO (offsetof(struct ipv6_hdr, proto))
218 #define MBUF_IPV4_2PROTO(m) \
219 rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV42PROTO)
220 #define MBUF_IPV6_2PROTO(m) \
221 rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV62PROTO)
223 #define GET_CB_FIELD(in, fd, base, lim, dlm) do { \
227 val = strtoul((in), &end, (base)); \
228 if (errno != 0 || end[0] != (dlm) || val > (lim)) \
230 (fd) = (typeof(fd))val; \
235 * ACL rules should have higher priorities than route ones to ensure ACL rule
236 * always be found when input packets have multi-matches in the database.
237 * A exception case is performance measure, which can define route rules with
238 * higher priority and route rules will always be returned in each lookup.
239 * Reserve range from ACL_RULE_PRIORITY_MAX + 1 to
240 * RTE_ACL_MAX_PRIORITY for route entries in performance measure
242 #define ACL_RULE_PRIORITY_MAX 0x10000000
245 * Forward port info save in ACL lib starts from 1
246 * since ACL assume 0 is invalid.
247 * So, need add 1 when saving and minus 1 when forwarding packets.
249 #define FWD_PORT_SHIFT 1
252 * Rule and trace formats definitions.
264 struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
266 .type = RTE_ACL_FIELD_TYPE_BITMASK,
267 .size = sizeof(uint8_t),
268 .field_index = PROTO_FIELD_IPV4,
269 .input_index = RTE_ACL_IPV4VLAN_PROTO,
273 .type = RTE_ACL_FIELD_TYPE_MASK,
274 .size = sizeof(uint32_t),
275 .field_index = SRC_FIELD_IPV4,
276 .input_index = RTE_ACL_IPV4VLAN_SRC,
277 .offset = offsetof(struct ipv4_hdr, src_addr) -
278 offsetof(struct ipv4_hdr, next_proto_id),
281 .type = RTE_ACL_FIELD_TYPE_MASK,
282 .size = sizeof(uint32_t),
283 .field_index = DST_FIELD_IPV4,
284 .input_index = RTE_ACL_IPV4VLAN_DST,
285 .offset = offsetof(struct ipv4_hdr, dst_addr) -
286 offsetof(struct ipv4_hdr, next_proto_id),
289 .type = RTE_ACL_FIELD_TYPE_RANGE,
290 .size = sizeof(uint16_t),
291 .field_index = SRCP_FIELD_IPV4,
292 .input_index = RTE_ACL_IPV4VLAN_PORTS,
293 .offset = sizeof(struct ipv4_hdr) -
294 offsetof(struct ipv4_hdr, next_proto_id),
297 .type = RTE_ACL_FIELD_TYPE_RANGE,
298 .size = sizeof(uint16_t),
299 .field_index = DSTP_FIELD_IPV4,
300 .input_index = RTE_ACL_IPV4VLAN_PORTS,
301 .offset = sizeof(struct ipv4_hdr) -
302 offsetof(struct ipv4_hdr, next_proto_id) +
307 #define IPV6_ADDR_LEN 16
308 #define IPV6_ADDR_U16 (IPV6_ADDR_LEN / sizeof(uint16_t))
309 #define IPV6_ADDR_U32 (IPV6_ADDR_LEN / sizeof(uint32_t))
326 struct rte_acl_field_def ipv6_defs[NUM_FIELDS_IPV6] = {
328 .type = RTE_ACL_FIELD_TYPE_BITMASK,
329 .size = sizeof(uint8_t),
330 .field_index = PROTO_FIELD_IPV6,
331 .input_index = PROTO_FIELD_IPV6,
335 .type = RTE_ACL_FIELD_TYPE_MASK,
336 .size = sizeof(uint32_t),
337 .field_index = SRC1_FIELD_IPV6,
338 .input_index = SRC1_FIELD_IPV6,
339 .offset = offsetof(struct ipv6_hdr, src_addr) -
340 offsetof(struct ipv6_hdr, proto),
343 .type = RTE_ACL_FIELD_TYPE_MASK,
344 .size = sizeof(uint32_t),
345 .field_index = SRC2_FIELD_IPV6,
346 .input_index = SRC2_FIELD_IPV6,
347 .offset = offsetof(struct ipv6_hdr, src_addr) -
348 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
351 .type = RTE_ACL_FIELD_TYPE_MASK,
352 .size = sizeof(uint32_t),
353 .field_index = SRC3_FIELD_IPV6,
354 .input_index = SRC3_FIELD_IPV6,
355 .offset = offsetof(struct ipv6_hdr, src_addr) -
356 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
359 .type = RTE_ACL_FIELD_TYPE_MASK,
360 .size = sizeof(uint32_t),
361 .field_index = SRC4_FIELD_IPV6,
362 .input_index = SRC4_FIELD_IPV6,
363 .offset = offsetof(struct ipv6_hdr, src_addr) -
364 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
367 .type = RTE_ACL_FIELD_TYPE_MASK,
368 .size = sizeof(uint32_t),
369 .field_index = DST1_FIELD_IPV6,
370 .input_index = DST1_FIELD_IPV6,
371 .offset = offsetof(struct ipv6_hdr, dst_addr)
372 - offsetof(struct ipv6_hdr, proto),
375 .type = RTE_ACL_FIELD_TYPE_MASK,
376 .size = sizeof(uint32_t),
377 .field_index = DST2_FIELD_IPV6,
378 .input_index = DST2_FIELD_IPV6,
379 .offset = offsetof(struct ipv6_hdr, dst_addr) -
380 offsetof(struct ipv6_hdr, proto) + sizeof(uint32_t),
383 .type = RTE_ACL_FIELD_TYPE_MASK,
384 .size = sizeof(uint32_t),
385 .field_index = DST3_FIELD_IPV6,
386 .input_index = DST3_FIELD_IPV6,
387 .offset = offsetof(struct ipv6_hdr, dst_addr) -
388 offsetof(struct ipv6_hdr, proto) + 2 * sizeof(uint32_t),
391 .type = RTE_ACL_FIELD_TYPE_MASK,
392 .size = sizeof(uint32_t),
393 .field_index = DST4_FIELD_IPV6,
394 .input_index = DST4_FIELD_IPV6,
395 .offset = offsetof(struct ipv6_hdr, dst_addr) -
396 offsetof(struct ipv6_hdr, proto) + 3 * sizeof(uint32_t),
399 .type = RTE_ACL_FIELD_TYPE_RANGE,
400 .size = sizeof(uint16_t),
401 .field_index = SRCP_FIELD_IPV6,
402 .input_index = SRCP_FIELD_IPV6,
403 .offset = sizeof(struct ipv6_hdr) -
404 offsetof(struct ipv6_hdr, proto),
407 .type = RTE_ACL_FIELD_TYPE_RANGE,
408 .size = sizeof(uint16_t),
409 .field_index = DSTP_FIELD_IPV6,
410 .input_index = SRCP_FIELD_IPV6,
411 .offset = sizeof(struct ipv6_hdr) -
412 offsetof(struct ipv6_hdr, proto) + sizeof(uint16_t),
421 CB_FLD_SRC_PORT_HIGH,
424 CB_FLD_DST_PORT_HIGH,
430 RTE_ACL_RULE_DEF(acl4_rule, RTE_DIM(ipv4_defs));
431 RTE_ACL_RULE_DEF(acl6_rule, RTE_DIM(ipv6_defs));
433 struct acl_search_t {
434 const uint8_t *data_ipv4[MAX_PKT_BURST];
435 struct rte_mbuf *m_ipv4[MAX_PKT_BURST];
436 uint32_t res_ipv4[MAX_PKT_BURST];
439 const uint8_t *data_ipv6[MAX_PKT_BURST];
440 struct rte_mbuf *m_ipv6[MAX_PKT_BURST];
441 uint32_t res_ipv6[MAX_PKT_BURST];
446 char mapped[NB_SOCKETS];
447 struct rte_acl_ctx *acx_ipv4[NB_SOCKETS];
448 struct rte_acl_ctx *acx_ipv6[NB_SOCKETS];
449 #ifdef L3FWDACL_DEBUG
450 struct acl4_rule *rule_ipv4;
451 struct acl6_rule *rule_ipv6;
456 const char *rule_ipv4_name;
457 const char *rule_ipv6_name;
461 const char cb_port_delim[] = ":";
464 print_one_ipv4_rule(struct acl4_rule *rule, int extra)
466 unsigned char a, b, c, d;
468 uint32_t_to_char(rule->field[SRC_FIELD_IPV4].value.u32,
470 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
471 rule->field[SRC_FIELD_IPV4].mask_range.u32);
472 uint32_t_to_char(rule->field[DST_FIELD_IPV4].value.u32,
474 printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
475 rule->field[DST_FIELD_IPV4].mask_range.u32);
476 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
477 rule->field[SRCP_FIELD_IPV4].value.u16,
478 rule->field[SRCP_FIELD_IPV4].mask_range.u16,
479 rule->field[DSTP_FIELD_IPV4].value.u16,
480 rule->field[DSTP_FIELD_IPV4].mask_range.u16,
481 rule->field[PROTO_FIELD_IPV4].value.u8,
482 rule->field[PROTO_FIELD_IPV4].mask_range.u8);
484 printf("0x%x-0x%x-0x%x ",
485 rule->data.category_mask,
487 rule->data.userdata);
491 print_one_ipv6_rule(struct acl6_rule *rule, int extra)
493 unsigned char a, b, c, d;
495 uint32_t_to_char(rule->field[SRC1_FIELD_IPV6].value.u32,
497 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
498 uint32_t_to_char(rule->field[SRC2_FIELD_IPV6].value.u32,
500 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
501 uint32_t_to_char(rule->field[SRC3_FIELD_IPV6].value.u32,
503 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
504 uint32_t_to_char(rule->field[SRC4_FIELD_IPV6].value.u32,
506 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
507 rule->field[SRC1_FIELD_IPV6].mask_range.u32
508 + rule->field[SRC2_FIELD_IPV6].mask_range.u32
509 + rule->field[SRC3_FIELD_IPV6].mask_range.u32
510 + rule->field[SRC4_FIELD_IPV6].mask_range.u32);
512 uint32_t_to_char(rule->field[DST1_FIELD_IPV6].value.u32,
514 printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
515 uint32_t_to_char(rule->field[DST2_FIELD_IPV6].value.u32,
517 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
518 uint32_t_to_char(rule->field[DST3_FIELD_IPV6].value.u32,
520 printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
521 uint32_t_to_char(rule->field[DST4_FIELD_IPV6].value.u32,
523 printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
524 rule->field[DST1_FIELD_IPV6].mask_range.u32
525 + rule->field[DST2_FIELD_IPV6].mask_range.u32
526 + rule->field[DST3_FIELD_IPV6].mask_range.u32
527 + rule->field[DST4_FIELD_IPV6].mask_range.u32);
529 printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
530 rule->field[SRCP_FIELD_IPV6].value.u16,
531 rule->field[SRCP_FIELD_IPV6].mask_range.u16,
532 rule->field[DSTP_FIELD_IPV6].value.u16,
533 rule->field[DSTP_FIELD_IPV6].mask_range.u16,
534 rule->field[PROTO_FIELD_IPV6].value.u8,
535 rule->field[PROTO_FIELD_IPV6].mask_range.u8);
537 printf("0x%x-0x%x-0x%x ",
538 rule->data.category_mask,
540 rule->data.userdata);
543 /* Bypass comment and empty lines */
545 is_bypass_line(char *buff)
550 if (buff[0] == COMMENT_LEAD_CHAR)
553 while (buff[i] != '\0') {
554 if (!isspace(buff[i]))
561 #ifdef L3FWDACL_DEBUG
563 dump_acl4_rule(struct rte_mbuf *m, uint32_t sig)
565 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
566 unsigned char a, b, c, d;
567 struct ipv4_hdr *ipv4_hdr = rte_pktmbuf_mtod_offset(m,
569 sizeof(struct ether_hdr));
571 uint32_t_to_char(rte_bswap32(ipv4_hdr->src_addr), &a, &b, &c, &d);
572 printf("Packet Src:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
573 uint32_t_to_char(rte_bswap32(ipv4_hdr->dst_addr), &a, &b, &c, &d);
574 printf("Dst:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
576 printf("Src port:%hu,Dst port:%hu ",
577 rte_bswap16(*(uint16_t *)(ipv4_hdr + 1)),
578 rte_bswap16(*((uint16_t *)(ipv4_hdr + 1) + 1)));
579 printf("hit ACL %d - ", offset);
581 print_one_ipv4_rule(acl_config.rule_ipv4 + offset, 1);
587 dump_acl6_rule(struct rte_mbuf *m, uint32_t sig)
590 uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
591 struct ipv6_hdr *ipv6_hdr = rte_pktmbuf_mtod_offset(m,
593 sizeof(struct ether_hdr));
595 printf("Packet Src");
596 for (i = 0; i < RTE_DIM(ipv6_hdr->src_addr); i += sizeof(uint16_t))
598 ipv6_hdr->src_addr[i], ipv6_hdr->src_addr[i + 1]);
601 for (i = 0; i < RTE_DIM(ipv6_hdr->dst_addr); i += sizeof(uint16_t))
603 ipv6_hdr->dst_addr[i], ipv6_hdr->dst_addr[i + 1]);
605 printf("\nSrc port:%hu,Dst port:%hu ",
606 rte_bswap16(*(uint16_t *)(ipv6_hdr + 1)),
607 rte_bswap16(*((uint16_t *)(ipv6_hdr + 1) + 1)));
608 printf("hit ACL %d - ", offset);
610 print_one_ipv6_rule(acl_config.rule_ipv6 + offset, 1);
614 #endif /* L3FWDACL_DEBUG */
617 dump_ipv4_rules(struct acl4_rule *rule, int num, int extra)
621 for (i = 0; i < num; i++, rule++) {
622 printf("\t%d:", i + 1);
623 print_one_ipv4_rule(rule, extra);
629 dump_ipv6_rules(struct acl6_rule *rule, int num, int extra)
633 for (i = 0; i < num; i++, rule++) {
634 printf("\t%d:", i + 1);
635 print_one_ipv6_rule(rule, extra);
640 #ifdef DO_RFC_1812_CHECKS
642 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
645 struct ipv4_hdr *ipv4_hdr;
646 struct rte_mbuf *pkt = pkts_in[index];
648 if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
649 ipv4_hdr = rte_pktmbuf_mtod_offset(pkt, struct ipv4_hdr *,
650 sizeof(struct ether_hdr));
652 /* Check to make sure the packet is valid (RFC1812) */
653 if (is_valid_ipv4_pkt(ipv4_hdr, pkt->pkt_len) >= 0) {
655 /* Update time to live and header checksum */
656 --(ipv4_hdr->time_to_live);
657 ++(ipv4_hdr->hdr_checksum);
659 /* Fill acl structure */
660 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
661 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
664 /* Not a valid IPv4 packet */
665 rte_pktmbuf_free(pkt);
667 } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
668 /* Fill acl structure */
669 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
670 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
673 /* Unknown type, drop the packet */
674 rte_pktmbuf_free(pkt);
680 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
683 struct rte_mbuf *pkt = pkts_in[index];
685 if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
686 /* Fill acl structure */
687 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
688 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
690 } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
691 /* Fill acl structure */
692 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
693 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
695 /* Unknown type, drop the packet */
696 rte_pktmbuf_free(pkt);
699 #endif /* DO_RFC_1812_CHECKS */
702 prepare_acl_parameter(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
710 /* Prefetch first packets */
711 for (i = 0; i < PREFETCH_OFFSET && i < nb_rx; i++) {
712 rte_prefetch0(rte_pktmbuf_mtod(
713 pkts_in[i], void *));
716 for (i = 0; i < (nb_rx - PREFETCH_OFFSET); i++) {
717 rte_prefetch0(rte_pktmbuf_mtod(pkts_in[
718 i + PREFETCH_OFFSET], void *));
719 prepare_one_packet(pkts_in, acl, i);
722 /* Process left packets */
723 for (; i < nb_rx; i++)
724 prepare_one_packet(pkts_in, acl, i);
728 send_one_packet(struct rte_mbuf *m, uint32_t res)
730 if (likely((res & ACL_DENY_SIGNATURE) == 0 && res != 0)) {
731 /* forward packets */
732 send_single_packet(m,
733 (uint8_t)(res - FWD_PORT_SHIFT));
735 /* in the ACL list, drop it */
736 #ifdef L3FWDACL_DEBUG
737 if ((res & ACL_DENY_SIGNATURE) != 0) {
738 if (RTE_ETH_IS_IPV4_HDR(m->packet_type))
739 dump_acl4_rule(m, res);
740 else if (RTE_ETH_IS_IPV6_HDR(m->packet_type))
741 dump_acl6_rule(m, res);
751 send_packets(struct rte_mbuf **m, uint32_t *res, int num)
755 /* Prefetch first packets */
756 for (i = 0; i < PREFETCH_OFFSET && i < num; i++) {
757 rte_prefetch0(rte_pktmbuf_mtod(
761 for (i = 0; i < (num - PREFETCH_OFFSET); i++) {
762 rte_prefetch0(rte_pktmbuf_mtod(m[
763 i + PREFETCH_OFFSET], void *));
764 send_one_packet(m[i], res[i]);
767 /* Process left packets */
769 send_one_packet(m[i], res[i]);
773 * Parses IPV6 address, exepcts the following format:
774 * XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX (where X - is a hexedecimal digit).
777 parse_ipv6_addr(const char *in, const char **end, uint32_t v[IPV6_ADDR_U32],
780 uint32_t addr[IPV6_ADDR_U16];
782 GET_CB_FIELD(in, addr[0], 16, UINT16_MAX, ':');
783 GET_CB_FIELD(in, addr[1], 16, UINT16_MAX, ':');
784 GET_CB_FIELD(in, addr[2], 16, UINT16_MAX, ':');
785 GET_CB_FIELD(in, addr[3], 16, UINT16_MAX, ':');
786 GET_CB_FIELD(in, addr[4], 16, UINT16_MAX, ':');
787 GET_CB_FIELD(in, addr[5], 16, UINT16_MAX, ':');
788 GET_CB_FIELD(in, addr[6], 16, UINT16_MAX, ':');
789 GET_CB_FIELD(in, addr[7], 16, UINT16_MAX, dlm);
793 v[0] = (addr[0] << 16) + addr[1];
794 v[1] = (addr[2] << 16) + addr[3];
795 v[2] = (addr[4] << 16) + addr[5];
796 v[3] = (addr[6] << 16) + addr[7];
802 parse_ipv6_net(const char *in, struct rte_acl_field field[4])
807 const uint32_t nbu32 = sizeof(uint32_t) * CHAR_BIT;
810 rc = parse_ipv6_addr(in, &mp, v, '/');
815 GET_CB_FIELD(mp, m, 0, CHAR_BIT * sizeof(v), 0);
817 /* put all together. */
818 for (i = 0; i != RTE_DIM(v); i++) {
819 if (m >= (i + 1) * nbu32)
820 field[i].mask_range.u32 = nbu32;
822 field[i].mask_range.u32 = m > (i * nbu32) ?
825 field[i].value.u32 = v[i];
832 parse_cb_ipv6_rule(char *str, struct rte_acl_rule *v, int has_userdata)
835 char *s, *sp, *in[CB_FLD_NUM];
836 static const char *dlm = " \t\n";
837 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
840 for (i = 0; i != dim; i++, s = NULL) {
841 in[i] = strtok_r(s, dlm, &sp);
846 rc = parse_ipv6_net(in[CB_FLD_SRC_ADDR], v->field + SRC1_FIELD_IPV6);
848 acl_log("failed to read source address/mask: %s\n",
849 in[CB_FLD_SRC_ADDR]);
853 rc = parse_ipv6_net(in[CB_FLD_DST_ADDR], v->field + DST1_FIELD_IPV6);
855 acl_log("failed to read destination address/mask: %s\n",
856 in[CB_FLD_DST_ADDR]);
861 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
862 v->field[SRCP_FIELD_IPV6].value.u16,
864 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
865 v->field[SRCP_FIELD_IPV6].mask_range.u16,
868 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
869 sizeof(cb_port_delim)) != 0)
872 /* destination port. */
873 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
874 v->field[DSTP_FIELD_IPV6].value.u16,
876 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
877 v->field[DSTP_FIELD_IPV6].mask_range.u16,
880 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
881 sizeof(cb_port_delim)) != 0)
884 if (v->field[SRCP_FIELD_IPV6].mask_range.u16
885 < v->field[SRCP_FIELD_IPV6].value.u16
886 || v->field[DSTP_FIELD_IPV6].mask_range.u16
887 < v->field[DSTP_FIELD_IPV6].value.u16)
890 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].value.u8,
892 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].mask_range.u8,
896 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata,
903 * Parse ClassBench rules file.
905 * '@'<src_ipv4_addr>'/'<masklen> <space> \
906 * <dst_ipv4_addr>'/'<masklen> <space> \
907 * <src_port_low> <space> ":" <src_port_high> <space> \
908 * <dst_port_low> <space> ":" <dst_port_high> <space> \
912 parse_ipv4_net(const char *in, uint32_t *addr, uint32_t *mask_len)
914 uint8_t a, b, c, d, m;
916 GET_CB_FIELD(in, a, 0, UINT8_MAX, '.');
917 GET_CB_FIELD(in, b, 0, UINT8_MAX, '.');
918 GET_CB_FIELD(in, c, 0, UINT8_MAX, '.');
919 GET_CB_FIELD(in, d, 0, UINT8_MAX, '/');
920 GET_CB_FIELD(in, m, 0, sizeof(uint32_t) * CHAR_BIT, 0);
922 addr[0] = IPv4(a, b, c, d);
929 parse_cb_ipv4vlan_rule(char *str, struct rte_acl_rule *v, int has_userdata)
932 char *s, *sp, *in[CB_FLD_NUM];
933 static const char *dlm = " \t\n";
934 int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
937 for (i = 0; i != dim; i++, s = NULL) {
938 in[i] = strtok_r(s, dlm, &sp);
943 rc = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
944 &v->field[SRC_FIELD_IPV4].value.u32,
945 &v->field[SRC_FIELD_IPV4].mask_range.u32);
947 acl_log("failed to read source address/mask: %s\n",
948 in[CB_FLD_SRC_ADDR]);
952 rc = parse_ipv4_net(in[CB_FLD_DST_ADDR],
953 &v->field[DST_FIELD_IPV4].value.u32,
954 &v->field[DST_FIELD_IPV4].mask_range.u32);
956 acl_log("failed to read destination address/mask: %s\n",
957 in[CB_FLD_DST_ADDR]);
961 GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
962 v->field[SRCP_FIELD_IPV4].value.u16,
964 GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
965 v->field[SRCP_FIELD_IPV4].mask_range.u16,
968 if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
969 sizeof(cb_port_delim)) != 0)
972 GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
973 v->field[DSTP_FIELD_IPV4].value.u16,
975 GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
976 v->field[DSTP_FIELD_IPV4].mask_range.u16,
979 if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
980 sizeof(cb_port_delim)) != 0)
983 if (v->field[SRCP_FIELD_IPV4].mask_range.u16
984 < v->field[SRCP_FIELD_IPV4].value.u16
985 || v->field[DSTP_FIELD_IPV4].mask_range.u16
986 < v->field[DSTP_FIELD_IPV4].value.u16)
989 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].value.u8,
991 GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].mask_range.u8,
995 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata, 0,
1002 add_rules(const char *rule_path,
1003 struct rte_acl_rule **proute_base,
1004 unsigned int *proute_num,
1005 struct rte_acl_rule **pacl_base,
1006 unsigned int *pacl_num, uint32_t rule_size,
1007 int (*parser)(char *, struct rte_acl_rule*, int))
1009 uint8_t *acl_rules, *route_rules;
1010 struct rte_acl_rule *next;
1011 unsigned int acl_num = 0, route_num = 0, total_num = 0;
1012 unsigned int acl_cnt = 0, route_cnt = 0;
1013 char buff[LINE_MAX];
1014 FILE *fh = fopen(rule_path, "rb");
1018 rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
1021 while ((fgets(buff, LINE_MAX, fh) != NULL)) {
1022 if (buff[0] == ROUTE_LEAD_CHAR)
1024 else if (buff[0] == ACL_LEAD_CHAR)
1029 rte_exit(EXIT_FAILURE, "Not find any route entries in %s!\n",
1032 fseek(fh, 0, SEEK_SET);
1034 acl_rules = calloc(acl_num, rule_size);
1036 if (NULL == acl_rules)
1037 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1040 route_rules = calloc(route_num, rule_size);
1042 if (NULL == route_rules)
1043 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1047 while (fgets(buff, LINE_MAX, fh) != NULL) {
1050 if (is_bypass_line(buff))
1056 if (s == ROUTE_LEAD_CHAR)
1057 next = (struct rte_acl_rule *)(route_rules +
1058 route_cnt * rule_size);
1061 else if (s == ACL_LEAD_CHAR)
1062 next = (struct rte_acl_rule *)(acl_rules +
1063 acl_cnt * rule_size);
1067 rte_exit(EXIT_FAILURE,
1068 "%s Line %u: should start with leading "
1070 rule_path, i, ROUTE_LEAD_CHAR, ACL_LEAD_CHAR);
1072 if (parser(buff + 1, next, s == ROUTE_LEAD_CHAR) != 0)
1073 rte_exit(EXIT_FAILURE,
1074 "%s Line %u: parse rules error\n",
1077 if (s == ROUTE_LEAD_CHAR) {
1078 /* Check the forwarding port number */
1079 if ((enabled_port_mask & (1 << next->data.userdata)) ==
1081 rte_exit(EXIT_FAILURE,
1082 "%s Line %u: fwd number illegal:%u\n",
1083 rule_path, i, next->data.userdata);
1084 next->data.userdata += FWD_PORT_SHIFT;
1087 next->data.userdata = ACL_DENY_SIGNATURE + acl_cnt;
1091 next->data.priority = RTE_ACL_MAX_PRIORITY - total_num;
1092 next->data.category_mask = -1;
1098 *pacl_base = (struct rte_acl_rule *)acl_rules;
1099 *pacl_num = acl_num;
1100 *proute_base = (struct rte_acl_rule *)route_rules;
1101 *proute_num = route_cnt;
1107 dump_acl_config(void)
1109 printf("ACL option are:\n");
1110 printf(OPTION_RULE_IPV4": %s\n", parm_config.rule_ipv4_name);
1111 printf(OPTION_RULE_IPV6": %s\n", parm_config.rule_ipv6_name);
1112 printf(OPTION_SCALAR": %d\n", parm_config.scalar);
1116 check_acl_config(void)
1118 if (parm_config.rule_ipv4_name == NULL) {
1119 acl_log("ACL IPv4 rule file not specified\n");
1121 } else if (parm_config.rule_ipv6_name == NULL) {
1122 acl_log("ACL IPv6 rule file not specified\n");
1129 static struct rte_acl_ctx*
1130 setup_acl(struct rte_acl_rule *route_base,
1131 struct rte_acl_rule *acl_base, unsigned int route_num,
1132 unsigned int acl_num, int ipv6, int socketid)
1134 char name[PATH_MAX];
1135 struct rte_acl_param acl_param;
1136 struct rte_acl_config acl_build_param;
1137 struct rte_acl_ctx *context;
1138 int dim = ipv6 ? RTE_DIM(ipv6_defs) : RTE_DIM(ipv4_defs);
1140 /* Create ACL contexts */
1141 snprintf(name, sizeof(name), "%s%d",
1142 ipv6 ? L3FWD_ACL_IPV6_NAME : L3FWD_ACL_IPV4_NAME,
1145 acl_param.name = name;
1146 acl_param.socket_id = socketid;
1147 acl_param.rule_size = RTE_ACL_RULE_SZ(dim);
1148 acl_param.max_rule_num = MAX_ACL_RULE_NUM;
1150 if ((context = rte_acl_create(&acl_param)) == NULL)
1151 rte_exit(EXIT_FAILURE, "Failed to create ACL context\n");
1153 if (parm_config.scalar && rte_acl_set_ctx_classify(context,
1154 RTE_ACL_CLASSIFY_SCALAR) != 0)
1155 rte_exit(EXIT_FAILURE,
1156 "Failed to setup classify method for ACL context\n");
1158 if (rte_acl_add_rules(context, route_base, route_num) < 0)
1159 rte_exit(EXIT_FAILURE, "add rules failed\n");
1161 if (rte_acl_add_rules(context, acl_base, acl_num) < 0)
1162 rte_exit(EXIT_FAILURE, "add rules failed\n");
1164 /* Perform builds */
1165 memset(&acl_build_param, 0, sizeof(acl_build_param));
1167 acl_build_param.num_categories = DEFAULT_MAX_CATEGORIES;
1168 acl_build_param.num_fields = dim;
1169 memcpy(&acl_build_param.defs, ipv6 ? ipv6_defs : ipv4_defs,
1170 ipv6 ? sizeof(ipv6_defs) : sizeof(ipv4_defs));
1172 if (rte_acl_build(context, &acl_build_param) != 0)
1173 rte_exit(EXIT_FAILURE, "Failed to build ACL trie\n");
1175 rte_acl_dump(context);
1186 struct rte_acl_rule *acl_base_ipv4, *route_base_ipv4,
1187 *acl_base_ipv6, *route_base_ipv6;
1188 unsigned int acl_num_ipv4 = 0, route_num_ipv4 = 0,
1189 acl_num_ipv6 = 0, route_num_ipv6 = 0;
1191 if (check_acl_config() != 0)
1192 rte_exit(EXIT_FAILURE, "Failed to get valid ACL options\n");
1196 /* Load rules from the input file */
1197 if (add_rules(parm_config.rule_ipv4_name, &route_base_ipv4,
1198 &route_num_ipv4, &acl_base_ipv4, &acl_num_ipv4,
1199 sizeof(struct acl4_rule), &parse_cb_ipv4vlan_rule) < 0)
1200 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1202 acl_log("IPv4 Route entries %u:\n", route_num_ipv4);
1203 dump_ipv4_rules((struct acl4_rule *)route_base_ipv4, route_num_ipv4, 1);
1205 acl_log("IPv4 ACL entries %u:\n", acl_num_ipv4);
1206 dump_ipv4_rules((struct acl4_rule *)acl_base_ipv4, acl_num_ipv4, 1);
1208 if (add_rules(parm_config.rule_ipv6_name, &route_base_ipv6,
1210 &acl_base_ipv6, &acl_num_ipv6,
1211 sizeof(struct acl6_rule), &parse_cb_ipv6_rule) < 0)
1212 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1214 acl_log("IPv6 Route entries %u:\n", route_num_ipv6);
1215 dump_ipv6_rules((struct acl6_rule *)route_base_ipv6, route_num_ipv6, 1);
1217 acl_log("IPv6 ACL entries %u:\n", acl_num_ipv6);
1218 dump_ipv6_rules((struct acl6_rule *)acl_base_ipv6, acl_num_ipv6, 1);
1220 memset(&acl_config, 0, sizeof(acl_config));
1222 /* Check sockets a context should be created on */
1224 acl_config.mapped[0] = 1;
1226 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1227 if (rte_lcore_is_enabled(lcore_id) == 0)
1230 socketid = rte_lcore_to_socket_id(lcore_id);
1231 if (socketid >= NB_SOCKETS) {
1232 acl_log("Socket %d of lcore %u is out "
1234 socketid, lcore_id, NB_SOCKETS);
1235 free(route_base_ipv4);
1236 free(route_base_ipv6);
1237 free(acl_base_ipv4);
1238 free(acl_base_ipv6);
1242 acl_config.mapped[socketid] = 1;
1246 for (i = 0; i < NB_SOCKETS; i++) {
1247 if (acl_config.mapped[i]) {
1248 acl_config.acx_ipv4[i] = setup_acl(route_base_ipv4,
1249 acl_base_ipv4, route_num_ipv4, acl_num_ipv4,
1252 acl_config.acx_ipv6[i] = setup_acl(route_base_ipv6,
1253 acl_base_ipv6, route_num_ipv6, acl_num_ipv6,
1258 free(route_base_ipv4);
1259 free(route_base_ipv6);
1261 #ifdef L3FWDACL_DEBUG
1262 acl_config.rule_ipv4 = (struct acl4_rule *)acl_base_ipv4;
1263 acl_config.rule_ipv6 = (struct acl6_rule *)acl_base_ipv6;
1265 free(acl_base_ipv4);
1266 free(acl_base_ipv6);
1272 /***********************end of ACL part******************************/
1275 uint16_t n_rx_queue;
1276 struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
1277 uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
1278 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
1279 } __rte_cache_aligned;
1281 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
1283 /* Send burst of packets on an output interface */
1285 send_burst(struct lcore_conf *qconf, uint16_t n, uint8_t port)
1287 struct rte_mbuf **m_table;
1291 queueid = qconf->tx_queue_id[port];
1292 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
1294 ret = rte_eth_tx_burst(port, queueid, m_table, n);
1295 if (unlikely(ret < n)) {
1297 rte_pktmbuf_free(m_table[ret]);
1298 } while (++ret < n);
1304 /* Enqueue a single packet, and send burst if queue is filled */
1306 send_single_packet(struct rte_mbuf *m, uint8_t port)
1310 struct lcore_conf *qconf;
1312 lcore_id = rte_lcore_id();
1314 qconf = &lcore_conf[lcore_id];
1315 len = qconf->tx_mbufs[port].len;
1316 qconf->tx_mbufs[port].m_table[len] = m;
1319 /* enough pkts to be sent */
1320 if (unlikely(len == MAX_PKT_BURST)) {
1321 send_burst(qconf, MAX_PKT_BURST, port);
1325 qconf->tx_mbufs[port].len = len;
1329 #ifdef DO_RFC_1812_CHECKS
1331 is_valid_ipv4_pkt(struct ipv4_hdr *pkt, uint32_t link_len)
1333 /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
1335 * 1. The packet length reported by the Link Layer must be large
1336 * enough to hold the minimum length legal IP datagram (20 bytes).
1338 if (link_len < sizeof(struct ipv4_hdr))
1341 /* 2. The IP checksum must be correct. */
1342 /* this is checked in H/W */
1345 * 3. The IP version number must be 4. If the version number is not 4
1346 * then the packet may be another version of IP, such as IPng or
1349 if (((pkt->version_ihl) >> 4) != 4)
1352 * 4. The IP header length field must be large enough to hold the
1353 * minimum length legal IP datagram (20 bytes = 5 words).
1355 if ((pkt->version_ihl & 0xf) < 5)
1359 * 5. The IP total length field must be large enough to hold the IP
1360 * datagram header, whose length is specified in the IP header length
1363 if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct ipv4_hdr))
1370 /* main processing loop */
1372 main_loop(__attribute__((unused)) void *dummy)
1374 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1376 uint64_t prev_tsc, diff_tsc, cur_tsc;
1378 uint8_t portid, queueid;
1379 struct lcore_conf *qconf;
1381 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
1382 / US_PER_S * BURST_TX_DRAIN_US;
1385 lcore_id = rte_lcore_id();
1386 qconf = &lcore_conf[lcore_id];
1387 socketid = rte_lcore_to_socket_id(lcore_id);
1389 if (qconf->n_rx_queue == 0) {
1390 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
1394 RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
1396 for (i = 0; i < qconf->n_rx_queue; i++) {
1398 portid = qconf->rx_queue_list[i].port_id;
1399 queueid = qconf->rx_queue_list[i].queue_id;
1400 RTE_LOG(INFO, L3FWD,
1401 " -- lcoreid=%u portid=%hhu rxqueueid=%hhu\n",
1402 lcore_id, portid, queueid);
1407 cur_tsc = rte_rdtsc();
1410 * TX burst queue drain
1412 diff_tsc = cur_tsc - prev_tsc;
1413 if (unlikely(diff_tsc > drain_tsc)) {
1416 * This could be optimized (use queueid instead of
1417 * portid), but it is not called so often
1419 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
1420 if (qconf->tx_mbufs[portid].len == 0)
1422 send_burst(&lcore_conf[lcore_id],
1423 qconf->tx_mbufs[portid].len,
1425 qconf->tx_mbufs[portid].len = 0;
1432 * Read packet from RX queues
1434 for (i = 0; i < qconf->n_rx_queue; ++i) {
1436 portid = qconf->rx_queue_list[i].port_id;
1437 queueid = qconf->rx_queue_list[i].queue_id;
1438 nb_rx = rte_eth_rx_burst(portid, queueid,
1439 pkts_burst, MAX_PKT_BURST);
1442 struct acl_search_t acl_search;
1444 prepare_acl_parameter(pkts_burst, &acl_search,
1447 if (acl_search.num_ipv4) {
1449 acl_config.acx_ipv4[socketid],
1450 acl_search.data_ipv4,
1451 acl_search.res_ipv4,
1452 acl_search.num_ipv4,
1453 DEFAULT_MAX_CATEGORIES);
1455 send_packets(acl_search.m_ipv4,
1456 acl_search.res_ipv4,
1457 acl_search.num_ipv4);
1460 if (acl_search.num_ipv6) {
1462 acl_config.acx_ipv6[socketid],
1463 acl_search.data_ipv6,
1464 acl_search.res_ipv6,
1465 acl_search.num_ipv6,
1466 DEFAULT_MAX_CATEGORIES);
1468 send_packets(acl_search.m_ipv6,
1469 acl_search.res_ipv6,
1470 acl_search.num_ipv6);
1478 check_lcore_params(void)
1480 uint8_t queue, lcore;
1484 for (i = 0; i < nb_lcore_params; ++i) {
1485 queue = lcore_params[i].queue_id;
1486 if (queue >= MAX_RX_QUEUE_PER_PORT) {
1487 printf("invalid queue number: %hhu\n", queue);
1490 lcore = lcore_params[i].lcore_id;
1491 if (!rte_lcore_is_enabled(lcore)) {
1492 printf("error: lcore %hhu is not enabled in "
1493 "lcore mask\n", lcore);
1496 socketid = rte_lcore_to_socket_id(lcore);
1497 if (socketid != 0 && numa_on == 0) {
1498 printf("warning: lcore %hhu is on socket %d "
1507 check_port_config(const unsigned nb_ports)
1512 for (i = 0; i < nb_lcore_params; ++i) {
1513 portid = lcore_params[i].port_id;
1515 if ((enabled_port_mask & (1 << portid)) == 0) {
1516 printf("port %u is not enabled in port mask\n", portid);
1519 if (portid >= nb_ports) {
1520 printf("port %u is not present on the board\n", portid);
1528 get_port_n_rx_queues(const uint8_t port)
1533 for (i = 0; i < nb_lcore_params; ++i) {
1534 if (lcore_params[i].port_id == port &&
1535 lcore_params[i].queue_id > queue)
1536 queue = lcore_params[i].queue_id;
1538 return (uint8_t)(++queue);
1542 init_lcore_rx_queues(void)
1544 uint16_t i, nb_rx_queue;
1547 for (i = 0; i < nb_lcore_params; ++i) {
1548 lcore = lcore_params[i].lcore_id;
1549 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
1550 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
1551 printf("error: too many queues (%u) for lcore: %u\n",
1552 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
1555 lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
1556 lcore_params[i].port_id;
1557 lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
1558 lcore_params[i].queue_id;
1559 lcore_conf[lcore].n_rx_queue++;
1567 print_usage(const char *prgname)
1569 printf("%s [EAL options] -- -p PORTMASK -P"
1570 "--"OPTION_RULE_IPV4"=FILE"
1571 "--"OPTION_RULE_IPV6"=FILE"
1572 " [--"OPTION_CONFIG" (port,queue,lcore)[,(port,queue,lcore]]"
1573 " [--"OPTION_ENBJMO" [--max-pkt-len PKTLEN]]\n"
1574 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
1575 " -P : enable promiscuous mode\n"
1576 " --"OPTION_CONFIG": (port,queue,lcore): "
1577 "rx queues configuration\n"
1578 " --"OPTION_NONUMA": optional, disable numa awareness\n"
1579 " --"OPTION_ENBJMO": enable jumbo frame"
1580 " which max packet len is PKTLEN in decimal (64-9600)\n"
1581 " --"OPTION_RULE_IPV4"=FILE: specify the ipv4 rules entries "
1583 "Each rule occupy one line. "
1584 "2 kinds of rules are supported. "
1585 "One is ACL entry at while line leads with character '%c', "
1586 "another is route entry at while line leads with "
1588 " --"OPTION_RULE_IPV6"=FILE: specify the ipv6 rules "
1590 " --"OPTION_SCALAR": Use scalar function to do lookup\n",
1591 prgname, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR);
1595 parse_max_pkt_len(const char *pktlen)
1600 /* parse decimal string */
1601 len = strtoul(pktlen, &end, 10);
1602 if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1612 parse_portmask(const char *portmask)
1617 /* parse hexadecimal string */
1618 pm = strtoul(portmask, &end, 16);
1619 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1629 parse_config(const char *q_arg)
1632 const char *p, *p0 = q_arg;
1640 unsigned long int_fld[_NUM_FLD];
1641 char *str_fld[_NUM_FLD];
1645 nb_lcore_params = 0;
1647 while ((p = strchr(p0, '(')) != NULL) {
1649 if ((p0 = strchr(p, ')')) == NULL)
1653 if (size >= sizeof(s))
1656 snprintf(s, sizeof(s), "%.*s", size, p);
1657 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') !=
1660 for (i = 0; i < _NUM_FLD; i++) {
1662 int_fld[i] = strtoul(str_fld[i], &end, 0);
1663 if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1666 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1667 printf("exceeded max number of lcore params: %hu\n",
1671 lcore_params_array[nb_lcore_params].port_id =
1672 (uint8_t)int_fld[FLD_PORT];
1673 lcore_params_array[nb_lcore_params].queue_id =
1674 (uint8_t)int_fld[FLD_QUEUE];
1675 lcore_params_array[nb_lcore_params].lcore_id =
1676 (uint8_t)int_fld[FLD_LCORE];
1679 lcore_params = lcore_params_array;
1683 /* Parse the argument given in the command line of the application */
1685 parse_args(int argc, char **argv)
1690 char *prgname = argv[0];
1691 static struct option lgopts[] = {
1692 {OPTION_CONFIG, 1, 0, 0},
1693 {OPTION_NONUMA, 0, 0, 0},
1694 {OPTION_ENBJMO, 0, 0, 0},
1695 {OPTION_RULE_IPV4, 1, 0, 0},
1696 {OPTION_RULE_IPV6, 1, 0, 0},
1697 {OPTION_SCALAR, 0, 0, 0},
1703 while ((opt = getopt_long(argc, argvopt, "p:P",
1704 lgopts, &option_index)) != EOF) {
1709 enabled_port_mask = parse_portmask(optarg);
1710 if (enabled_port_mask == 0) {
1711 printf("invalid portmask\n");
1712 print_usage(prgname);
1717 printf("Promiscuous mode selected\n");
1723 if (!strncmp(lgopts[option_index].name,
1725 sizeof(OPTION_CONFIG))) {
1726 ret = parse_config(optarg);
1728 printf("invalid config\n");
1729 print_usage(prgname);
1734 if (!strncmp(lgopts[option_index].name,
1736 sizeof(OPTION_NONUMA))) {
1737 printf("numa is disabled\n");
1741 if (!strncmp(lgopts[option_index].name,
1742 OPTION_ENBJMO, sizeof(OPTION_ENBJMO))) {
1743 struct option lenopts = {
1750 printf("jumbo frame is enabled\n");
1751 port_conf.rxmode.jumbo_frame = 1;
1754 * if no max-pkt-len set, then use the
1755 * default value ETHER_MAX_LEN
1757 if (0 == getopt_long(argc, argvopt, "",
1758 &lenopts, &option_index)) {
1759 ret = parse_max_pkt_len(optarg);
1761 (ret > MAX_JUMBO_PKT_LEN)) {
1762 printf("invalid packet "
1764 print_usage(prgname);
1767 port_conf.rxmode.max_rx_pkt_len = ret;
1769 printf("set jumbo frame max packet length "
1772 port_conf.rxmode.max_rx_pkt_len);
1775 if (!strncmp(lgopts[option_index].name,
1777 sizeof(OPTION_RULE_IPV4)))
1778 parm_config.rule_ipv4_name = optarg;
1780 if (!strncmp(lgopts[option_index].name,
1782 sizeof(OPTION_RULE_IPV6))) {
1783 parm_config.rule_ipv6_name = optarg;
1786 if (!strncmp(lgopts[option_index].name,
1787 OPTION_SCALAR, sizeof(OPTION_SCALAR)))
1788 parm_config.scalar = 1;
1794 print_usage(prgname);
1800 argv[optind-1] = prgname;
1803 optind = 0; /* reset getopt lib */
1808 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
1810 char buf[ETHER_ADDR_FMT_SIZE];
1811 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
1812 printf("%s%s", name, buf);
1816 init_mem(unsigned nb_mbuf)
1822 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1823 if (rte_lcore_is_enabled(lcore_id) == 0)
1827 socketid = rte_lcore_to_socket_id(lcore_id);
1831 if (socketid >= NB_SOCKETS) {
1832 rte_exit(EXIT_FAILURE,
1833 "Socket %d of lcore %u is out of range %d\n",
1834 socketid, lcore_id, NB_SOCKETS);
1836 if (pktmbuf_pool[socketid] == NULL) {
1837 snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
1838 pktmbuf_pool[socketid] =
1839 rte_pktmbuf_pool_create(s, nb_mbuf,
1840 MEMPOOL_CACHE_SIZE, 0,
1841 RTE_MBUF_DEFAULT_BUF_SIZE,
1843 if (pktmbuf_pool[socketid] == NULL)
1844 rte_exit(EXIT_FAILURE,
1845 "Cannot init mbuf pool on socket %d\n",
1848 printf("Allocated mbuf pool on socket %d\n",
1855 /* Check the link status of all ports in up to 9s, and print them finally */
1857 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
1859 #define CHECK_INTERVAL 100 /* 100ms */
1860 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1861 uint8_t portid, count, all_ports_up, print_flag = 0;
1862 struct rte_eth_link link;
1864 printf("\nChecking link status");
1866 for (count = 0; count <= MAX_CHECK_TIME; count++) {
1868 for (portid = 0; portid < port_num; portid++) {
1869 if ((port_mask & (1 << portid)) == 0)
1871 memset(&link, 0, sizeof(link));
1872 rte_eth_link_get_nowait(portid, &link);
1873 /* print link status if flag set */
1874 if (print_flag == 1) {
1875 if (link.link_status)
1876 printf("Port %d Link Up - speed %u "
1877 "Mbps - %s\n", (uint8_t)portid,
1878 (unsigned)link.link_speed,
1879 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1880 ("full-duplex") : ("half-duplex\n"));
1882 printf("Port %d Link Down\n",
1886 /* clear all_ports_up flag if any link down */
1887 if (link.link_status == 0) {
1892 /* after finally printing all link status, get out */
1893 if (print_flag == 1)
1896 if (all_ports_up == 0) {
1899 rte_delay_ms(CHECK_INTERVAL);
1902 /* set the print_flag if all ports up or timeout */
1903 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1911 main(int argc, char **argv)
1913 struct lcore_conf *qconf;
1914 struct rte_eth_dev_info dev_info;
1915 struct rte_eth_txconf *txconf;
1920 uint32_t n_tx_queue, nb_lcores;
1921 uint8_t portid, nb_rx_queue, queue, socketid;
1924 ret = rte_eal_init(argc, argv);
1926 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1930 /* parse application arguments (after the EAL ones) */
1931 ret = parse_args(argc, argv);
1933 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
1935 if (check_lcore_params() < 0)
1936 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1938 ret = init_lcore_rx_queues();
1940 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1942 nb_ports = rte_eth_dev_count();
1943 if (nb_ports > RTE_MAX_ETHPORTS)
1944 nb_ports = RTE_MAX_ETHPORTS;
1946 if (check_port_config(nb_ports) < 0)
1947 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1949 /* Add ACL rules and route entries, build trie */
1950 if (app_acl_init() < 0)
1951 rte_exit(EXIT_FAILURE, "app_acl_init failed\n");
1953 nb_lcores = rte_lcore_count();
1955 /* initialize all ports */
1956 for (portid = 0; portid < nb_ports; portid++) {
1957 /* skip ports that are not enabled */
1958 if ((enabled_port_mask & (1 << portid)) == 0) {
1959 printf("\nSkipping disabled port %d\n", portid);
1964 printf("Initializing port %d ... ", portid);
1967 nb_rx_queue = get_port_n_rx_queues(portid);
1968 n_tx_queue = nb_lcores;
1969 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1970 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1971 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1972 nb_rx_queue, (unsigned)n_tx_queue);
1973 ret = rte_eth_dev_configure(portid, nb_rx_queue,
1974 (uint16_t)n_tx_queue, &port_conf);
1976 rte_exit(EXIT_FAILURE,
1977 "Cannot configure device: err=%d, port=%d\n",
1980 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1981 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1985 ret = init_mem(NB_MBUF);
1987 rte_exit(EXIT_FAILURE, "init_mem failed\n");
1989 /* init one TX queue per couple (lcore,port) */
1991 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1992 if (rte_lcore_is_enabled(lcore_id) == 0)
1996 socketid = (uint8_t)
1997 rte_lcore_to_socket_id(lcore_id);
2001 printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
2004 rte_eth_dev_info_get(portid, &dev_info);
2005 txconf = &dev_info.default_txconf;
2006 if (port_conf.rxmode.jumbo_frame)
2007 txconf->txq_flags = 0;
2008 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
2011 rte_exit(EXIT_FAILURE,
2012 "rte_eth_tx_queue_setup: err=%d, "
2013 "port=%d\n", ret, portid);
2015 qconf = &lcore_conf[lcore_id];
2016 qconf->tx_queue_id[portid] = queueid;
2022 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2023 if (rte_lcore_is_enabled(lcore_id) == 0)
2025 qconf = &lcore_conf[lcore_id];
2026 printf("\nInitializing rx queues on lcore %u ... ", lcore_id);
2028 /* init RX queues */
2029 for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
2030 portid = qconf->rx_queue_list[queue].port_id;
2031 queueid = qconf->rx_queue_list[queue].queue_id;
2034 socketid = (uint8_t)
2035 rte_lcore_to_socket_id(lcore_id);
2039 printf("rxq=%d,%d,%d ", portid, queueid, socketid);
2042 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
2044 pktmbuf_pool[socketid]);
2046 rte_exit(EXIT_FAILURE,
2047 "rte_eth_rx_queue_setup: err=%d,"
2048 "port=%d\n", ret, portid);
2055 for (portid = 0; portid < nb_ports; portid++) {
2056 if ((enabled_port_mask & (1 << portid)) == 0)
2060 ret = rte_eth_dev_start(portid);
2062 rte_exit(EXIT_FAILURE,
2063 "rte_eth_dev_start: err=%d, port=%d\n",
2067 * If enabled, put device in promiscuous mode.
2068 * This allows IO forwarding mode to forward packets
2069 * to itself through 2 cross-connected ports of the
2073 rte_eth_promiscuous_enable(portid);
2076 check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
2078 /* launch per-lcore init on every lcore */
2079 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
2080 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
2081 if (rte_eal_wait_lcore(lcore_id) < 0)