examples/l3fwd-acl/main.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2010-2016 Intel Corporation
   3  */
   4
   5 #include <stdio.h>
   6 #include <stdlib.h>
   7 #include <stdint.h>
   8 #include <inttypes.h>
   9 #include <sys/types.h>
  10 #include <string.h>
  11 #include <sys/queue.h>
  12 #include <stdarg.h>
  13 #include <errno.h>
  14 #include <getopt.h>
  15
  16 #include <rte_common.h>
  17 #include <rte_byteorder.h>
  18 #include <rte_log.h>
  19 #include <rte_memory.h>
  20 #include <rte_memcpy.h>
  21 #include <rte_eal.h>
  22 #include <rte_launch.h>
  23 #include <rte_atomic.h>
  24 #include <rte_cycles.h>
  25 #include <rte_prefetch.h>
  26 #include <rte_lcore.h>
  27 #include <rte_per_lcore.h>
  28 #include <rte_branch_prediction.h>
  29 #include <rte_interrupts.h>
  30 #include <rte_random.h>
  31 #include <rte_debug.h>
  32 #include <rte_ether.h>
  33 #include <rte_ethdev.h>
  34 #include <rte_mempool.h>
  35 #include <rte_mbuf.h>
  36 #include <rte_ip.h>
  37 #include <rte_tcp.h>
  38 #include <rte_udp.h>
  39 #include <rte_string_fns.h>
  40 #include <rte_acl.h>
  41
  42 #if RTE_LOG_DP_LEVEL >= RTE_LOG_DEBUG
  43 #define L3FWDACL_DEBUG
  44 #endif
  45 #define DO_RFC_1812_CHECKS
  46
  47 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
  48
  49 #define MAX_JUMBO_PKT_LEN  9600
  50
  51 #define MEMPOOL_CACHE_SIZE 256
  52
  53 /*
  54  * This expression is used to calculate the number of mbufs needed
  55  * depending on user input, taking into account memory for rx and tx hardware
  56  * rings, cache per lcore and mtable per port per lcore.
  57  * RTE_MAX is used to ensure that NB_MBUF never goes below a
  58  * minimum value of 8192
  59  */
  60
  61 #define NB_MBUF RTE_MAX(\
  62         (nb_ports * nb_rx_queue * nb_rxd +      \
  63         nb_ports * nb_lcores * MAX_PKT_BURST +  \
  64         nb_ports * n_tx_queue * nb_txd +        \
  65         nb_lcores * MEMPOOL_CACHE_SIZE),        \
  66         (unsigned)8192)
  67
  68 #define MAX_PKT_BURST 32
  69 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
  70
  71 #define NB_SOCKETS 8
  72
  73 /* Configure how many packets ahead to prefetch, when reading packets */
  74 #define PREFETCH_OFFSET 3
  75
  76 /*
  77  * Configurable number of RX/TX ring descriptors
  78  */
  79 #define RTE_TEST_RX_DESC_DEFAULT 1024
  80 #define RTE_TEST_TX_DESC_DEFAULT 1024
  81 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
  82 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
  83
  84 /* ethernet addresses of ports */
  85 static struct rte_ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
  86
  87 /* mask of enabled ports */
  88 static uint32_t enabled_port_mask;
  89 static int promiscuous_on; /**< Ports set in promiscuous mode off by default. */
  90 static int numa_on = 1; /**< NUMA is enabled by default. */
  91
  92 struct lcore_rx_queue {
  93         uint16_t port_id;
  94         uint8_t queue_id;
  95 } __rte_cache_aligned;
  96
  97 #define MAX_RX_QUEUE_PER_LCORE 16
  98 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
  99 #define MAX_RX_QUEUE_PER_PORT 128
 100
 101 #define MAX_LCORE_PARAMS 1024
 102 struct lcore_params {
 103         uint16_t port_id;
 104         uint8_t queue_id;
 105         uint8_t lcore_id;
 106 } __rte_cache_aligned;
 107
 108 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
 109 static struct lcore_params lcore_params_array_default[] = {
 110         {0, 0, 2},
 111         {0, 1, 2},
 112         {0, 2, 2},
 113         {1, 0, 2},
 114         {1, 1, 2},
 115         {1, 2, 2},
 116         {2, 0, 2},
 117         {3, 0, 3},
 118         {3, 1, 3},
 119 };
 120
 121 static struct lcore_params *lcore_params = lcore_params_array_default;
 122 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
 123                                 sizeof(lcore_params_array_default[0]);
 124
 125 static struct rte_eth_conf port_conf = {
 126         .rxmode = {
 127                 .mq_mode        = ETH_MQ_RX_RSS,
 128                 .max_rx_pkt_len = RTE_ETHER_MAX_LEN,
 129                 .split_hdr_size = 0,
 130                 .offloads = DEV_RX_OFFLOAD_CHECKSUM,
 131         },
 132         .rx_adv_conf = {
 133                 .rss_conf = {
 134                         .rss_key = NULL,
 135                         .rss_hf = ETH_RSS_IP | ETH_RSS_UDP |
 136                                 ETH_RSS_TCP | ETH_RSS_SCTP,
 137                 },
 138         },
 139         .txmode = {
 140                 .mq_mode = ETH_MQ_TX_NONE,
 141         },
 142 };
 143
 144 static struct rte_mempool *pktmbuf_pool[NB_SOCKETS];
 145
 146 /***********************start of ACL part******************************/
 147 #ifdef DO_RFC_1812_CHECKS
 148 static inline int
 149 is_valid_ipv4_pkt(struct rte_ipv4_hdr *pkt, uint32_t link_len);
 150 #endif
 151 static inline void
 152 send_single_packet(struct rte_mbuf *m, uint16_t port);
 153
 154 #define MAX_ACL_RULE_NUM        100000
 155 #define DEFAULT_MAX_CATEGORIES  1
 156 #define L3FWD_ACL_IPV4_NAME     "l3fwd-acl-ipv4"
 157 #define L3FWD_ACL_IPV6_NAME     "l3fwd-acl-ipv6"
 158 #define ACL_LEAD_CHAR           ('@')
 159 #define ROUTE_LEAD_CHAR         ('R')
 160 #define COMMENT_LEAD_CHAR       ('#')
 161 #define OPTION_CONFIG           "config"
 162 #define OPTION_NONUMA           "no-numa"
 163 #define OPTION_ENBJMO           "enable-jumbo"
 164 #define OPTION_RULE_IPV4        "rule_ipv4"
 165 #define OPTION_RULE_IPV6        "rule_ipv6"
 166 #define OPTION_SCALAR           "scalar"
 167 #define ACL_DENY_SIGNATURE      0xf0000000
 168 #define RTE_LOGTYPE_L3FWDACL    RTE_LOGTYPE_USER3
 169 #define acl_log(format, ...)    RTE_LOG(ERR, L3FWDACL, format, ##__VA_ARGS__)
 170 #define uint32_t_to_char(ip, a, b, c, d) do {\
 171                 *a = (unsigned char)(ip >> 24 & 0xff);\
 172                 *b = (unsigned char)(ip >> 16 & 0xff);\
 173                 *c = (unsigned char)(ip >> 8 & 0xff);\
 174                 *d = (unsigned char)(ip & 0xff);\
 175         } while (0)
 176 #define OFF_ETHHEAD     (sizeof(struct rte_ether_hdr))
 177 #define OFF_IPV42PROTO (offsetof(struct rte_ipv4_hdr, next_proto_id))
 178 #define OFF_IPV62PROTO (offsetof(struct rte_ipv6_hdr, proto))
 179 #define MBUF_IPV4_2PROTO(m)     \
 180         rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV42PROTO)
 181 #define MBUF_IPV6_2PROTO(m)     \
 182         rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV62PROTO)
 183
 184 #define GET_CB_FIELD(in, fd, base, lim, dlm)    do {            \
 185         unsigned long val;                                      \
 186         char *end;                                              \
 187         errno = 0;                                              \
 188         val = strtoul((in), &end, (base));                      \
 189         if (errno != 0 || end[0] != (dlm) || val > (lim))       \
 190                 return -EINVAL;                               \
 191         (fd) = (typeof(fd))val;                                 \
 192         (in) = end + 1;                                         \
 193 } while (0)
 194
 195 /*
 196   * ACL rules should have higher priorities than route ones to ensure ACL rule
 197   * always be found when input packets have multi-matches in the database.
 198   * A exception case is performance measure, which can define route rules with
 199   * higher priority and route rules will always be returned in each lookup.
 200   * Reserve range from ACL_RULE_PRIORITY_MAX + 1 to
 201   * RTE_ACL_MAX_PRIORITY for route entries in performance measure
 202   */
 203 #define ACL_RULE_PRIORITY_MAX 0x10000000
 204
 205 /*
 206   * Forward port info save in ACL lib starts from 1
 207   * since ACL assume 0 is invalid.
 208   * So, need add 1 when saving and minus 1 when forwarding packets.
 209   */
 210 #define FWD_PORT_SHIFT 1
 211
 212 /*
 213  * Rule and trace formats definitions.
 214  */
 215
 216 enum {
 217         PROTO_FIELD_IPV4,
 218         SRC_FIELD_IPV4,
 219         DST_FIELD_IPV4,
 220         SRCP_FIELD_IPV4,
 221         DSTP_FIELD_IPV4,
 222         NUM_FIELDS_IPV4
 223 };
 224
 225 /*
 226  * That effectively defines order of IPV4VLAN classifications:
 227  *  - PROTO
 228  *  - VLAN (TAG and DOMAIN)
 229  *  - SRC IP ADDRESS
 230  *  - DST IP ADDRESS
 231  *  - PORTS (SRC and DST)
 232  */
 233 enum {
 234         RTE_ACL_IPV4VLAN_PROTO,
 235         RTE_ACL_IPV4VLAN_VLAN,
 236         RTE_ACL_IPV4VLAN_SRC,
 237         RTE_ACL_IPV4VLAN_DST,
 238         RTE_ACL_IPV4VLAN_PORTS,
 239         RTE_ACL_IPV4VLAN_NUM
 240 };
 241
 242 struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
 243         {
 244                 .type = RTE_ACL_FIELD_TYPE_BITMASK,
 245                 .size = sizeof(uint8_t),
 246                 .field_index = PROTO_FIELD_IPV4,
 247                 .input_index = RTE_ACL_IPV4VLAN_PROTO,
 248                 .offset = 0,
 249         },
 250         {
 251                 .type = RTE_ACL_FIELD_TYPE_MASK,
 252                 .size = sizeof(uint32_t),
 253                 .field_index = SRC_FIELD_IPV4,
 254                 .input_index = RTE_ACL_IPV4VLAN_SRC,
 255                 .offset = offsetof(struct rte_ipv4_hdr, src_addr) -
 256                         offsetof(struct rte_ipv4_hdr, next_proto_id),
 257         },
 258         {
 259                 .type = RTE_ACL_FIELD_TYPE_MASK,
 260                 .size = sizeof(uint32_t),
 261                 .field_index = DST_FIELD_IPV4,
 262                 .input_index = RTE_ACL_IPV4VLAN_DST,
 263                 .offset = offsetof(struct rte_ipv4_hdr, dst_addr) -
 264                         offsetof(struct rte_ipv4_hdr, next_proto_id),
 265         },
 266         {
 267                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 268                 .size = sizeof(uint16_t),
 269                 .field_index = SRCP_FIELD_IPV4,
 270                 .input_index = RTE_ACL_IPV4VLAN_PORTS,
 271                 .offset = sizeof(struct rte_ipv4_hdr) -
 272                         offsetof(struct rte_ipv4_hdr, next_proto_id),
 273         },
 274         {
 275                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 276                 .size = sizeof(uint16_t),
 277                 .field_index = DSTP_FIELD_IPV4,
 278                 .input_index = RTE_ACL_IPV4VLAN_PORTS,
 279                 .offset = sizeof(struct rte_ipv4_hdr) -
 280                         offsetof(struct rte_ipv4_hdr, next_proto_id) +
 281                         sizeof(uint16_t),
 282         },
 283 };
 284
 285 #define IPV6_ADDR_LEN   16
 286 #define IPV6_ADDR_U16   (IPV6_ADDR_LEN / sizeof(uint16_t))
 287 #define IPV6_ADDR_U32   (IPV6_ADDR_LEN / sizeof(uint32_t))
 288
 289 enum {
 290         PROTO_FIELD_IPV6,
 291         SRC1_FIELD_IPV6,
 292         SRC2_FIELD_IPV6,
 293         SRC3_FIELD_IPV6,
 294         SRC4_FIELD_IPV6,
 295         DST1_FIELD_IPV6,
 296         DST2_FIELD_IPV6,
 297         DST3_FIELD_IPV6,
 298         DST4_FIELD_IPV6,
 299         SRCP_FIELD_IPV6,
 300         DSTP_FIELD_IPV6,
 301         NUM_FIELDS_IPV6
 302 };
 303
 304 struct rte_acl_field_def ipv6_defs[NUM_FIELDS_IPV6] = {
 305         {
 306                 .type = RTE_ACL_FIELD_TYPE_BITMASK,
 307                 .size = sizeof(uint8_t),
 308                 .field_index = PROTO_FIELD_IPV6,
 309                 .input_index = PROTO_FIELD_IPV6,
 310                 .offset = 0,
 311         },
 312         {
 313                 .type = RTE_ACL_FIELD_TYPE_MASK,
 314                 .size = sizeof(uint32_t),
 315                 .field_index = SRC1_FIELD_IPV6,
 316                 .input_index = SRC1_FIELD_IPV6,
 317                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 318                         offsetof(struct rte_ipv6_hdr, proto),
 319         },
 320         {
 321                 .type = RTE_ACL_FIELD_TYPE_MASK,
 322                 .size = sizeof(uint32_t),
 323                 .field_index = SRC2_FIELD_IPV6,
 324                 .input_index = SRC2_FIELD_IPV6,
 325                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 326                         offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint32_t),
 327         },
 328         {
 329                 .type = RTE_ACL_FIELD_TYPE_MASK,
 330                 .size = sizeof(uint32_t),
 331                 .field_index = SRC3_FIELD_IPV6,
 332                 .input_index = SRC3_FIELD_IPV6,
 333                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 334                         offsetof(struct rte_ipv6_hdr, proto) +
 335                         2 * sizeof(uint32_t),
 336         },
 337         {
 338                 .type = RTE_ACL_FIELD_TYPE_MASK,
 339                 .size = sizeof(uint32_t),
 340                 .field_index = SRC4_FIELD_IPV6,
 341                 .input_index = SRC4_FIELD_IPV6,
 342                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 343                         offsetof(struct rte_ipv6_hdr, proto) +
 344                         3 * sizeof(uint32_t),
 345         },
 346         {
 347                 .type = RTE_ACL_FIELD_TYPE_MASK,
 348                 .size = sizeof(uint32_t),
 349                 .field_index = DST1_FIELD_IPV6,
 350                 .input_index = DST1_FIELD_IPV6,
 351                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr)
 352                                 - offsetof(struct rte_ipv6_hdr, proto),
 353         },
 354         {
 355                 .type = RTE_ACL_FIELD_TYPE_MASK,
 356                 .size = sizeof(uint32_t),
 357                 .field_index = DST2_FIELD_IPV6,
 358                 .input_index = DST2_FIELD_IPV6,
 359                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
 360                         offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint32_t),
 361         },
 362         {
 363                 .type = RTE_ACL_FIELD_TYPE_MASK,
 364                 .size = sizeof(uint32_t),
 365                 .field_index = DST3_FIELD_IPV6,
 366                 .input_index = DST3_FIELD_IPV6,
 367                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
 368                         offsetof(struct rte_ipv6_hdr, proto) +
 369                         2 * sizeof(uint32_t),
 370         },
 371         {
 372                 .type = RTE_ACL_FIELD_TYPE_MASK,
 373                 .size = sizeof(uint32_t),
 374                 .field_index = DST4_FIELD_IPV6,
 375                 .input_index = DST4_FIELD_IPV6,
 376                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
 377                         offsetof(struct rte_ipv6_hdr, proto) +
 378                         3 * sizeof(uint32_t),
 379         },
 380         {
 381                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 382                 .size = sizeof(uint16_t),
 383                 .field_index = SRCP_FIELD_IPV6,
 384                 .input_index = SRCP_FIELD_IPV6,
 385                 .offset = sizeof(struct rte_ipv6_hdr) -
 386                         offsetof(struct rte_ipv6_hdr, proto),
 387         },
 388         {
 389                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 390                 .size = sizeof(uint16_t),
 391                 .field_index = DSTP_FIELD_IPV6,
 392                 .input_index = SRCP_FIELD_IPV6,
 393                 .offset = sizeof(struct rte_ipv6_hdr) -
 394                         offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint16_t),
 395         },
 396 };
 397
 398 enum {
 399         CB_FLD_SRC_ADDR,
 400         CB_FLD_DST_ADDR,
 401         CB_FLD_SRC_PORT_LOW,
 402         CB_FLD_SRC_PORT_DLM,
 403         CB_FLD_SRC_PORT_HIGH,
 404         CB_FLD_DST_PORT_LOW,
 405         CB_FLD_DST_PORT_DLM,
 406         CB_FLD_DST_PORT_HIGH,
 407         CB_FLD_PROTO,
 408         CB_FLD_USERDATA,
 409         CB_FLD_NUM,
 410 };
 411
 412 RTE_ACL_RULE_DEF(acl4_rule, RTE_DIM(ipv4_defs));
 413 RTE_ACL_RULE_DEF(acl6_rule, RTE_DIM(ipv6_defs));
 414
 415 struct acl_search_t {
 416         const uint8_t *data_ipv4[MAX_PKT_BURST];
 417         struct rte_mbuf *m_ipv4[MAX_PKT_BURST];
 418         uint32_t res_ipv4[MAX_PKT_BURST];
 419         int num_ipv4;
 420
 421         const uint8_t *data_ipv6[MAX_PKT_BURST];
 422         struct rte_mbuf *m_ipv6[MAX_PKT_BURST];
 423         uint32_t res_ipv6[MAX_PKT_BURST];
 424         int num_ipv6;
 425 };
 426
 427 static struct {
 428         char mapped[NB_SOCKETS];
 429         struct rte_acl_ctx *acx_ipv4[NB_SOCKETS];
 430         struct rte_acl_ctx *acx_ipv6[NB_SOCKETS];
 431 #ifdef L3FWDACL_DEBUG
 432         struct acl4_rule *rule_ipv4;
 433         struct acl6_rule *rule_ipv6;
 434 #endif
 435 } acl_config;
 436
 437 static struct{
 438         const char *rule_ipv4_name;
 439         const char *rule_ipv6_name;
 440         int scalar;
 441 } parm_config;
 442
 443 const char cb_port_delim[] = ":";
 444
 445 static inline void
 446 print_one_ipv4_rule(struct acl4_rule *rule, int extra)
 447 {
 448         unsigned char a, b, c, d;
 449
 450         uint32_t_to_char(rule->field[SRC_FIELD_IPV4].value.u32,
 451                         &a, &b, &c, &d);
 452         printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
 453                         rule->field[SRC_FIELD_IPV4].mask_range.u32);
 454         uint32_t_to_char(rule->field[DST_FIELD_IPV4].value.u32,
 455                         &a, &b, &c, &d);
 456         printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
 457                         rule->field[DST_FIELD_IPV4].mask_range.u32);
 458         printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
 459                 rule->field[SRCP_FIELD_IPV4].value.u16,
 460                 rule->field[SRCP_FIELD_IPV4].mask_range.u16,
 461                 rule->field[DSTP_FIELD_IPV4].value.u16,
 462                 rule->field[DSTP_FIELD_IPV4].mask_range.u16,
 463                 rule->field[PROTO_FIELD_IPV4].value.u8,
 464                 rule->field[PROTO_FIELD_IPV4].mask_range.u8);
 465         if (extra)
 466                 printf("0x%x-0x%x-0x%x ",
 467                         rule->data.category_mask,
 468                         rule->data.priority,
 469                         rule->data.userdata);
 470 }
 471
 472 static inline void
 473 print_one_ipv6_rule(struct acl6_rule *rule, int extra)
 474 {
 475         unsigned char a, b, c, d;
 476
 477         uint32_t_to_char(rule->field[SRC1_FIELD_IPV6].value.u32,
 478                 &a, &b, &c, &d);
 479         printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
 480         uint32_t_to_char(rule->field[SRC2_FIELD_IPV6].value.u32,
 481                 &a, &b, &c, &d);
 482         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 483         uint32_t_to_char(rule->field[SRC3_FIELD_IPV6].value.u32,
 484                 &a, &b, &c, &d);
 485         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 486         uint32_t_to_char(rule->field[SRC4_FIELD_IPV6].value.u32,
 487                 &a, &b, &c, &d);
 488         printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
 489                         rule->field[SRC1_FIELD_IPV6].mask_range.u32
 490                         + rule->field[SRC2_FIELD_IPV6].mask_range.u32
 491                         + rule->field[SRC3_FIELD_IPV6].mask_range.u32
 492                         + rule->field[SRC4_FIELD_IPV6].mask_range.u32);
 493
 494         uint32_t_to_char(rule->field[DST1_FIELD_IPV6].value.u32,
 495                 &a, &b, &c, &d);
 496         printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
 497         uint32_t_to_char(rule->field[DST2_FIELD_IPV6].value.u32,
 498                 &a, &b, &c, &d);
 499         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 500         uint32_t_to_char(rule->field[DST3_FIELD_IPV6].value.u32,
 501                 &a, &b, &c, &d);
 502         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 503         uint32_t_to_char(rule->field[DST4_FIELD_IPV6].value.u32,
 504                 &a, &b, &c, &d);
 505         printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
 506                         rule->field[DST1_FIELD_IPV6].mask_range.u32
 507                         + rule->field[DST2_FIELD_IPV6].mask_range.u32
 508                         + rule->field[DST3_FIELD_IPV6].mask_range.u32
 509                         + rule->field[DST4_FIELD_IPV6].mask_range.u32);
 510
 511         printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
 512                 rule->field[SRCP_FIELD_IPV6].value.u16,
 513                 rule->field[SRCP_FIELD_IPV6].mask_range.u16,
 514                 rule->field[DSTP_FIELD_IPV6].value.u16,
 515                 rule->field[DSTP_FIELD_IPV6].mask_range.u16,
 516                 rule->field[PROTO_FIELD_IPV6].value.u8,
 517                 rule->field[PROTO_FIELD_IPV6].mask_range.u8);
 518         if (extra)
 519                 printf("0x%x-0x%x-0x%x ",
 520                         rule->data.category_mask,
 521                         rule->data.priority,
 522                         rule->data.userdata);
 523 }
 524
 525 /* Bypass comment and empty lines */
 526 static inline int
 527 is_bypass_line(char *buff)
 528 {
 529         int i = 0;
 530
 531         /* comment line */
 532         if (buff[0] == COMMENT_LEAD_CHAR)
 533                 return 1;
 534         /* empty line */
 535         while (buff[i] != '\0') {
 536                 if (!isspace(buff[i]))
 537                         return 0;
 538                 i++;
 539         }
 540         return 1;
 541 }
 542
 543 #ifdef L3FWDACL_DEBUG
 544 static inline void
 545 dump_acl4_rule(struct rte_mbuf *m, uint32_t sig)
 546 {
 547         uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
 548         unsigned char a, b, c, d;
 549         struct rte_ipv4_hdr *ipv4_hdr =
 550                 rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
 551                                         sizeof(struct rte_ether_hdr));
 552
 553         uint32_t_to_char(rte_bswap32(ipv4_hdr->src_addr), &a, &b, &c, &d);
 554         printf("Packet Src:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
 555         uint32_t_to_char(rte_bswap32(ipv4_hdr->dst_addr), &a, &b, &c, &d);
 556         printf("Dst:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
 557
 558         printf("Src port:%hu,Dst port:%hu ",
 559                         rte_bswap16(*(uint16_t *)(ipv4_hdr + 1)),
 560                         rte_bswap16(*((uint16_t *)(ipv4_hdr + 1) + 1)));
 561         printf("hit ACL %d - ", offset);
 562
 563         print_one_ipv4_rule(acl_config.rule_ipv4 + offset, 1);
 564
 565         printf("\n\n");
 566 }
 567
 568 static inline void
 569 dump_acl6_rule(struct rte_mbuf *m, uint32_t sig)
 570 {
 571         unsigned i;
 572         uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
 573         struct rte_ipv6_hdr *ipv6_hdr =
 574                 rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
 575                                         sizeof(struct rte_ether_hdr));
 576
 577         printf("Packet Src");
 578         for (i = 0; i < RTE_DIM(ipv6_hdr->src_addr); i += sizeof(uint16_t))
 579                 printf(":%.2x%.2x",
 580                         ipv6_hdr->src_addr[i], ipv6_hdr->src_addr[i + 1]);
 581
 582         printf("\nDst");
 583         for (i = 0; i < RTE_DIM(ipv6_hdr->dst_addr); i += sizeof(uint16_t))
 584                 printf(":%.2x%.2x",
 585                         ipv6_hdr->dst_addr[i], ipv6_hdr->dst_addr[i + 1]);
 586
 587         printf("\nSrc port:%hu,Dst port:%hu ",
 588                         rte_bswap16(*(uint16_t *)(ipv6_hdr + 1)),
 589                         rte_bswap16(*((uint16_t *)(ipv6_hdr + 1) + 1)));
 590         printf("hit ACL %d - ", offset);
 591
 592         print_one_ipv6_rule(acl_config.rule_ipv6 + offset, 1);
 593
 594         printf("\n\n");
 595 }
 596 #endif /* L3FWDACL_DEBUG */
 597
 598 static inline void
 599 dump_ipv4_rules(struct acl4_rule *rule, int num, int extra)
 600 {
 601         int i;
 602
 603         for (i = 0; i < num; i++, rule++) {
 604                 printf("\t%d:", i + 1);
 605                 print_one_ipv4_rule(rule, extra);
 606                 printf("\n");
 607         }
 608 }
 609
 610 static inline void
 611 dump_ipv6_rules(struct acl6_rule *rule, int num, int extra)
 612 {
 613         int i;
 614
 615         for (i = 0; i < num; i++, rule++) {
 616                 printf("\t%d:", i + 1);
 617                 print_one_ipv6_rule(rule, extra);
 618                 printf("\n");
 619         }
 620 }
 621
 622 #ifdef DO_RFC_1812_CHECKS
 623 static inline void
 624 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
 625         int index)
 626 {
 627         struct rte_ipv4_hdr *ipv4_hdr;
 628         struct rte_mbuf *pkt = pkts_in[index];
 629
 630         if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
 631                 ipv4_hdr = rte_pktmbuf_mtod_offset(pkt, struct rte_ipv4_hdr *,
 632                                                 sizeof(struct rte_ether_hdr));
 633
 634                 /* Check to make sure the packet is valid (RFC1812) */
 635                 if (is_valid_ipv4_pkt(ipv4_hdr, pkt->pkt_len) >= 0) {
 636
 637                         /* Update time to live and header checksum */
 638                         --(ipv4_hdr->time_to_live);
 639                         ++(ipv4_hdr->hdr_checksum);
 640
 641                         /* Fill acl structure */
 642                         acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
 643                         acl->m_ipv4[(acl->num_ipv4)++] = pkt;
 644
 645                 } else {
 646                         /* Not a valid IPv4 packet */
 647                         rte_pktmbuf_free(pkt);
 648                 }
 649         } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
 650                 /* Fill acl structure */
 651                 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
 652                 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
 653
 654         } else {
 655                 /* Unknown type, drop the packet */
 656                 rte_pktmbuf_free(pkt);
 657         }
 658 }
 659
 660 #else
 661 static inline void
 662 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
 663         int index)
 664 {
 665         struct rte_mbuf *pkt = pkts_in[index];
 666
 667         if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
 668                 /* Fill acl structure */
 669                 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
 670                 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
 671
 672         } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
 673                 /* Fill acl structure */
 674                 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
 675                 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
 676         } else {
 677                 /* Unknown type, drop the packet */
 678                 rte_pktmbuf_free(pkt);
 679         }
 680 }
 681 #endif /* DO_RFC_1812_CHECKS */
 682
 683 static inline void
 684 prepare_acl_parameter(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
 685         int nb_rx)
 686 {
 687         int i;
 688
 689         acl->num_ipv4 = 0;
 690         acl->num_ipv6 = 0;
 691
 692         /* Prefetch first packets */
 693         for (i = 0; i < PREFETCH_OFFSET && i < nb_rx; i++) {
 694                 rte_prefetch0(rte_pktmbuf_mtod(
 695                                 pkts_in[i], void *));
 696         }
 697
 698         for (i = 0; i < (nb_rx - PREFETCH_OFFSET); i++) {
 699                 rte_prefetch0(rte_pktmbuf_mtod(pkts_in[
 700                                 i + PREFETCH_OFFSET], void *));
 701                 prepare_one_packet(pkts_in, acl, i);
 702         }
 703
 704         /* Process left packets */
 705         for (; i < nb_rx; i++)
 706                 prepare_one_packet(pkts_in, acl, i);
 707 }
 708
 709 static inline void
 710 send_one_packet(struct rte_mbuf *m, uint32_t res)
 711 {
 712         if (likely((res & ACL_DENY_SIGNATURE) == 0 && res != 0)) {
 713                 /* forward packets */
 714                 send_single_packet(m,
 715                         (uint8_t)(res - FWD_PORT_SHIFT));
 716         } else{
 717                 /* in the ACL list, drop it */
 718 #ifdef L3FWDACL_DEBUG
 719                 if ((res & ACL_DENY_SIGNATURE) != 0) {
 720                         if (RTE_ETH_IS_IPV4_HDR(m->packet_type))
 721                                 dump_acl4_rule(m, res);
 722                         else if (RTE_ETH_IS_IPV6_HDR(m->packet_type))
 723                                 dump_acl6_rule(m, res);
 724                 }
 725 #endif
 726                 rte_pktmbuf_free(m);
 727         }
 728 }
 729
 730
 731
 732 static inline void
 733 send_packets(struct rte_mbuf **m, uint32_t *res, int num)
 734 {
 735         int i;
 736
 737         /* Prefetch first packets */
 738         for (i = 0; i < PREFETCH_OFFSET && i < num; i++) {
 739                 rte_prefetch0(rte_pktmbuf_mtod(
 740                                 m[i], void *));
 741         }
 742
 743         for (i = 0; i < (num - PREFETCH_OFFSET); i++) {
 744                 rte_prefetch0(rte_pktmbuf_mtod(m[
 745                                 i + PREFETCH_OFFSET], void *));
 746                 send_one_packet(m[i], res[i]);
 747         }
 748
 749         /* Process left packets */
 750         for (; i < num; i++)
 751                 send_one_packet(m[i], res[i]);
 752 }
 753
 754 /*
 755  * Parses IPV6 address, exepcts the following format:
 756  * XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX (where X - is a hexedecimal digit).
 757  */
 758 static int
 759 parse_ipv6_addr(const char *in, const char **end, uint32_t v[IPV6_ADDR_U32],
 760         char dlm)
 761 {
 762         uint32_t addr[IPV6_ADDR_U16];
 763
 764         GET_CB_FIELD(in, addr[0], 16, UINT16_MAX, ':');
 765         GET_CB_FIELD(in, addr[1], 16, UINT16_MAX, ':');
 766         GET_CB_FIELD(in, addr[2], 16, UINT16_MAX, ':');
 767         GET_CB_FIELD(in, addr[3], 16, UINT16_MAX, ':');
 768         GET_CB_FIELD(in, addr[4], 16, UINT16_MAX, ':');
 769         GET_CB_FIELD(in, addr[5], 16, UINT16_MAX, ':');
 770         GET_CB_FIELD(in, addr[6], 16, UINT16_MAX, ':');
 771         GET_CB_FIELD(in, addr[7], 16, UINT16_MAX, dlm);
 772
 773         *end = in;
 774
 775         v[0] = (addr[0] << 16) + addr[1];
 776         v[1] = (addr[2] << 16) + addr[3];
 777         v[2] = (addr[4] << 16) + addr[5];
 778         v[3] = (addr[6] << 16) + addr[7];
 779
 780         return 0;
 781 }
 782
 783 static int
 784 parse_ipv6_net(const char *in, struct rte_acl_field field[4])
 785 {
 786         int32_t rc;
 787         const char *mp;
 788         uint32_t i, m, v[4];
 789         const uint32_t nbu32 = sizeof(uint32_t) * CHAR_BIT;
 790
 791         /* get address. */
 792         rc = parse_ipv6_addr(in, &mp, v, '/');
 793         if (rc != 0)
 794                 return rc;
 795
 796         /* get mask. */
 797         GET_CB_FIELD(mp, m, 0, CHAR_BIT * sizeof(v), 0);
 798
 799         /* put all together. */
 800         for (i = 0; i != RTE_DIM(v); i++) {
 801                 if (m >= (i + 1) * nbu32)
 802                         field[i].mask_range.u32 = nbu32;
 803                 else
 804                         field[i].mask_range.u32 = m > (i * nbu32) ?
 805                                 m - (i * 32) : 0;
 806
 807                 field[i].value.u32 = v[i];
 808         }
 809
 810         return 0;
 811 }
 812
 813 static int
 814 parse_cb_ipv6_rule(char *str, struct rte_acl_rule *v, int has_userdata)
 815 {
 816         int i, rc;
 817         char *s, *sp, *in[CB_FLD_NUM];
 818         static const char *dlm = " \t\n";
 819         int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
 820         s = str;
 821
 822         for (i = 0; i != dim; i++, s = NULL) {
 823                 in[i] = strtok_r(s, dlm, &sp);
 824                 if (in[i] == NULL)
 825                         return -EINVAL;
 826         }
 827
 828         rc = parse_ipv6_net(in[CB_FLD_SRC_ADDR], v->field + SRC1_FIELD_IPV6);
 829         if (rc != 0) {
 830                 acl_log("failed to read source address/mask: %s\n",
 831                         in[CB_FLD_SRC_ADDR]);
 832                 return rc;
 833         }
 834
 835         rc = parse_ipv6_net(in[CB_FLD_DST_ADDR], v->field + DST1_FIELD_IPV6);
 836         if (rc != 0) {
 837                 acl_log("failed to read destination address/mask: %s\n",
 838                         in[CB_FLD_DST_ADDR]);
 839                 return rc;
 840         }
 841
 842         /* source port. */
 843         GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
 844                 v->field[SRCP_FIELD_IPV6].value.u16,
 845                 0, UINT16_MAX, 0);
 846         GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
 847                 v->field[SRCP_FIELD_IPV6].mask_range.u16,
 848                 0, UINT16_MAX, 0);
 849
 850         if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
 851                         sizeof(cb_port_delim)) != 0)
 852                 return -EINVAL;
 853
 854         /* destination port. */
 855         GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
 856                 v->field[DSTP_FIELD_IPV6].value.u16,
 857                 0, UINT16_MAX, 0);
 858         GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
 859                 v->field[DSTP_FIELD_IPV6].mask_range.u16,
 860                 0, UINT16_MAX, 0);
 861
 862         if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
 863                         sizeof(cb_port_delim)) != 0)
 864                 return -EINVAL;
 865
 866         if (v->field[SRCP_FIELD_IPV6].mask_range.u16
 867                         < v->field[SRCP_FIELD_IPV6].value.u16
 868                         || v->field[DSTP_FIELD_IPV6].mask_range.u16
 869                         < v->field[DSTP_FIELD_IPV6].value.u16)
 870                 return -EINVAL;
 871
 872         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].value.u8,
 873                 0, UINT8_MAX, '/');
 874         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].mask_range.u8,
 875                 0, UINT8_MAX, 0);
 876
 877         if (has_userdata)
 878                 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata,
 879                         0, UINT32_MAX, 0);
 880
 881         return 0;
 882 }
 883
 884 /*
 885  * Parse ClassBench rules file.
 886  * Expected format:
 887  * '@'<src_ipv4_addr>'/'<masklen> <space> \
 888  * <dst_ipv4_addr>'/'<masklen> <space> \
 889  * <src_port_low> <space> ":" <src_port_high> <space> \
 890  * <dst_port_low> <space> ":" <dst_port_high> <space> \
 891  * <proto>'/'<mask>
 892  */
 893 static int
 894 parse_ipv4_net(const char *in, uint32_t *addr, uint32_t *mask_len)
 895 {
 896         uint8_t a, b, c, d, m;
 897
 898         GET_CB_FIELD(in, a, 0, UINT8_MAX, '.');
 899         GET_CB_FIELD(in, b, 0, UINT8_MAX, '.');
 900         GET_CB_FIELD(in, c, 0, UINT8_MAX, '.');
 901         GET_CB_FIELD(in, d, 0, UINT8_MAX, '/');
 902         GET_CB_FIELD(in, m, 0, sizeof(uint32_t) * CHAR_BIT, 0);
 903
 904         addr[0] = RTE_IPV4(a, b, c, d);
 905         mask_len[0] = m;
 906
 907         return 0;
 908 }
 909
 910 static int
 911 parse_cb_ipv4vlan_rule(char *str, struct rte_acl_rule *v, int has_userdata)
 912 {
 913         int i, rc;
 914         char *s, *sp, *in[CB_FLD_NUM];
 915         static const char *dlm = " \t\n";
 916         int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
 917         s = str;
 918
 919         for (i = 0; i != dim; i++, s = NULL) {
 920                 in[i] = strtok_r(s, dlm, &sp);
 921                 if (in[i] == NULL)
 922                         return -EINVAL;
 923         }
 924
 925         rc = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
 926                         &v->field[SRC_FIELD_IPV4].value.u32,
 927                         &v->field[SRC_FIELD_IPV4].mask_range.u32);
 928         if (rc != 0) {
 929                         acl_log("failed to read source address/mask: %s\n",
 930                         in[CB_FLD_SRC_ADDR]);
 931                 return rc;
 932         }
 933
 934         rc = parse_ipv4_net(in[CB_FLD_DST_ADDR],
 935                         &v->field[DST_FIELD_IPV4].value.u32,
 936                         &v->field[DST_FIELD_IPV4].mask_range.u32);
 937         if (rc != 0) {
 938                 acl_log("failed to read destination address/mask: %s\n",
 939                         in[CB_FLD_DST_ADDR]);
 940                 return rc;
 941         }
 942
 943         GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
 944                 v->field[SRCP_FIELD_IPV4].value.u16,
 945                 0, UINT16_MAX, 0);
 946         GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
 947                 v->field[SRCP_FIELD_IPV4].mask_range.u16,
 948                 0, UINT16_MAX, 0);
 949
 950         if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
 951                         sizeof(cb_port_delim)) != 0)
 952                 return -EINVAL;
 953
 954         GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
 955                 v->field[DSTP_FIELD_IPV4].value.u16,
 956                 0, UINT16_MAX, 0);
 957         GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
 958                 v->field[DSTP_FIELD_IPV4].mask_range.u16,
 959                 0, UINT16_MAX, 0);
 960
 961         if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
 962                         sizeof(cb_port_delim)) != 0)
 963                 return -EINVAL;
 964
 965         if (v->field[SRCP_FIELD_IPV4].mask_range.u16
 966                         < v->field[SRCP_FIELD_IPV4].value.u16
 967                         || v->field[DSTP_FIELD_IPV4].mask_range.u16
 968                         < v->field[DSTP_FIELD_IPV4].value.u16)
 969                 return -EINVAL;
 970
 971         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].value.u8,
 972                 0, UINT8_MAX, '/');
 973         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].mask_range.u8,
 974                 0, UINT8_MAX, 0);
 975
 976         if (has_userdata)
 977                 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata, 0,
 978                         UINT32_MAX, 0);
 979
 980         return 0;
 981 }
 982
 983 static int
 984 add_rules(const char *rule_path,
 985                 struct rte_acl_rule **proute_base,
 986                 unsigned int *proute_num,
 987                 struct rte_acl_rule **pacl_base,
 988                 unsigned int *pacl_num, uint32_t rule_size,
 989                 int (*parser)(char *, struct rte_acl_rule*, int))
 990 {
 991         uint8_t *acl_rules, *route_rules;
 992         struct rte_acl_rule *next;
 993         unsigned int acl_num = 0, route_num = 0, total_num = 0;
 994         unsigned int acl_cnt = 0, route_cnt = 0;
 995         char buff[LINE_MAX];
 996         FILE *fh = fopen(rule_path, "rb");
 997         unsigned int i = 0;
 998         int val;
 999
1000         if (fh == NULL)
1001                 rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
1002                         rule_path);
1003
1004         while ((fgets(buff, LINE_MAX, fh) != NULL)) {
1005                 if (buff[0] == ROUTE_LEAD_CHAR)
1006                         route_num++;
1007                 else if (buff[0] == ACL_LEAD_CHAR)
1008                         acl_num++;
1009         }
1010
1011         if (0 == route_num)
1012                 rte_exit(EXIT_FAILURE, "Not find any route entries in %s!\n",
1013                                 rule_path);
1014
1015         val = fseek(fh, 0, SEEK_SET);
1016         if (val < 0) {
1017                 rte_exit(EXIT_FAILURE, "%s: File seek operation failed\n",
1018                         __func__);
1019         }
1020
1021         acl_rules = calloc(acl_num, rule_size);
1022
1023         if (NULL == acl_rules)
1024                 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1025                         __func__);
1026
1027         route_rules = calloc(route_num, rule_size);
1028
1029         if (NULL == route_rules)
1030                 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1031                         __func__);
1032
1033         i = 0;
1034         while (fgets(buff, LINE_MAX, fh) != NULL) {
1035                 i++;
1036
1037                 if (is_bypass_line(buff))
1038                         continue;
1039
1040                 char s = buff[0];
1041
1042                 /* Route entry */
1043                 if (s == ROUTE_LEAD_CHAR)
1044                         next = (struct rte_acl_rule *)(route_rules +
1045                                 route_cnt * rule_size);
1046
1047                 /* ACL entry */
1048                 else if (s == ACL_LEAD_CHAR)
1049                         next = (struct rte_acl_rule *)(acl_rules +
1050                                 acl_cnt * rule_size);
1051
1052                 /* Illegal line */
1053                 else
1054                         rte_exit(EXIT_FAILURE,
1055                                 "%s Line %u: should start with leading "
1056                                 "char %c or %c\n",
1057                                 rule_path, i, ROUTE_LEAD_CHAR, ACL_LEAD_CHAR);
1058
1059                 if (parser(buff + 1, next, s == ROUTE_LEAD_CHAR) != 0)
1060                         rte_exit(EXIT_FAILURE,
1061                                 "%s Line %u: parse rules error\n",
1062                                 rule_path, i);
1063
1064                 if (s == ROUTE_LEAD_CHAR) {
1065                         /* Check the forwarding port number */
1066                         if ((enabled_port_mask & (1 << next->data.userdata)) ==
1067                                         0)
1068                                 rte_exit(EXIT_FAILURE,
1069                                         "%s Line %u: fwd number illegal:%u\n",
1070                                         rule_path, i, next->data.userdata);
1071                         next->data.userdata += FWD_PORT_SHIFT;
1072                         route_cnt++;
1073                 } else {
1074                         next->data.userdata = ACL_DENY_SIGNATURE + acl_cnt;
1075                         acl_cnt++;
1076                 }
1077
1078                 next->data.priority = RTE_ACL_MAX_PRIORITY - total_num;
1079                 next->data.category_mask = -1;
1080                 total_num++;
1081         }
1082
1083         fclose(fh);
1084
1085         *pacl_base = (struct rte_acl_rule *)acl_rules;
1086         *pacl_num = acl_num;
1087         *proute_base = (struct rte_acl_rule *)route_rules;
1088         *proute_num = route_cnt;
1089
1090         return 0;
1091 }
1092
1093 static void
1094 dump_acl_config(void)
1095 {
1096         printf("ACL option are:\n");
1097         printf(OPTION_RULE_IPV4": %s\n", parm_config.rule_ipv4_name);
1098         printf(OPTION_RULE_IPV6": %s\n", parm_config.rule_ipv6_name);
1099         printf(OPTION_SCALAR": %d\n", parm_config.scalar);
1100 }
1101
1102 static int
1103 check_acl_config(void)
1104 {
1105         if (parm_config.rule_ipv4_name == NULL) {
1106                 acl_log("ACL IPv4 rule file not specified\n");
1107                 return -1;
1108         } else if (parm_config.rule_ipv6_name == NULL) {
1109                 acl_log("ACL IPv6 rule file not specified\n");
1110                 return -1;
1111         }
1112
1113         return 0;
1114 }
1115
1116 static struct rte_acl_ctx*
1117 setup_acl(struct rte_acl_rule *route_base,
1118                 struct rte_acl_rule *acl_base, unsigned int route_num,
1119                 unsigned int acl_num, int ipv6, int socketid)
1120 {
1121         char name[PATH_MAX];
1122         struct rte_acl_param acl_param;
1123         struct rte_acl_config acl_build_param;
1124         struct rte_acl_ctx *context;
1125         int dim = ipv6 ? RTE_DIM(ipv6_defs) : RTE_DIM(ipv4_defs);
1126
1127         /* Create ACL contexts */
1128         snprintf(name, sizeof(name), "%s%d",
1129                         ipv6 ? L3FWD_ACL_IPV6_NAME : L3FWD_ACL_IPV4_NAME,
1130                         socketid);
1131
1132         acl_param.name = name;
1133         acl_param.socket_id = socketid;
1134         acl_param.rule_size = RTE_ACL_RULE_SZ(dim);
1135         acl_param.max_rule_num = MAX_ACL_RULE_NUM;
1136
1137         if ((context = rte_acl_create(&acl_param)) == NULL)
1138                 rte_exit(EXIT_FAILURE, "Failed to create ACL context\n");
1139
1140         if (parm_config.scalar && rte_acl_set_ctx_classify(context,
1141                         RTE_ACL_CLASSIFY_SCALAR) != 0)
1142                 rte_exit(EXIT_FAILURE,
1143                         "Failed to setup classify method for  ACL context\n");
1144
1145         if (rte_acl_add_rules(context, route_base, route_num) < 0)
1146                         rte_exit(EXIT_FAILURE, "add rules failed\n");
1147
1148         if (rte_acl_add_rules(context, acl_base, acl_num) < 0)
1149                         rte_exit(EXIT_FAILURE, "add rules failed\n");
1150
1151         /* Perform builds */
1152         memset(&acl_build_param, 0, sizeof(acl_build_param));
1153
1154         acl_build_param.num_categories = DEFAULT_MAX_CATEGORIES;
1155         acl_build_param.num_fields = dim;
1156         memcpy(&acl_build_param.defs, ipv6 ? ipv6_defs : ipv4_defs,
1157                 ipv6 ? sizeof(ipv6_defs) : sizeof(ipv4_defs));
1158
1159         if (rte_acl_build(context, &acl_build_param) != 0)
1160                 rte_exit(EXIT_FAILURE, "Failed to build ACL trie\n");
1161
1162         rte_acl_dump(context);
1163
1164         return context;
1165 }
1166
1167 static int
1168 app_acl_init(void)
1169 {
1170         unsigned lcore_id;
1171         unsigned int i;
1172         int socketid;
1173         struct rte_acl_rule *acl_base_ipv4, *route_base_ipv4,
1174                 *acl_base_ipv6, *route_base_ipv6;
1175         unsigned int acl_num_ipv4 = 0, route_num_ipv4 = 0,
1176                 acl_num_ipv6 = 0, route_num_ipv6 = 0;
1177
1178         if (check_acl_config() != 0)
1179                 rte_exit(EXIT_FAILURE, "Failed to get valid ACL options\n");
1180
1181         dump_acl_config();
1182
1183         /* Load  rules from the input file */
1184         if (add_rules(parm_config.rule_ipv4_name, &route_base_ipv4,
1185                         &route_num_ipv4, &acl_base_ipv4, &acl_num_ipv4,
1186                         sizeof(struct acl4_rule), &parse_cb_ipv4vlan_rule) < 0)
1187                 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1188
1189         acl_log("IPv4 Route entries %u:\n", route_num_ipv4);
1190         dump_ipv4_rules((struct acl4_rule *)route_base_ipv4, route_num_ipv4, 1);
1191
1192         acl_log("IPv4 ACL entries %u:\n", acl_num_ipv4);
1193         dump_ipv4_rules((struct acl4_rule *)acl_base_ipv4, acl_num_ipv4, 1);
1194
1195         if (add_rules(parm_config.rule_ipv6_name, &route_base_ipv6,
1196                         &route_num_ipv6,
1197                         &acl_base_ipv6, &acl_num_ipv6,
1198                         sizeof(struct acl6_rule), &parse_cb_ipv6_rule) < 0)
1199                 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1200
1201         acl_log("IPv6 Route entries %u:\n", route_num_ipv6);
1202         dump_ipv6_rules((struct acl6_rule *)route_base_ipv6, route_num_ipv6, 1);
1203
1204         acl_log("IPv6 ACL entries %u:\n", acl_num_ipv6);
1205         dump_ipv6_rules((struct acl6_rule *)acl_base_ipv6, acl_num_ipv6, 1);
1206
1207         memset(&acl_config, 0, sizeof(acl_config));
1208
1209         /* Check sockets a context should be created on */
1210         if (!numa_on)
1211                 acl_config.mapped[0] = 1;
1212         else {
1213                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1214                         if (rte_lcore_is_enabled(lcore_id) == 0)
1215                                 continue;
1216
1217                         socketid = rte_lcore_to_socket_id(lcore_id);
1218                         if (socketid >= NB_SOCKETS) {
1219                                 acl_log("Socket %d of lcore %u is out "
1220                                         "of range %d\n",
1221                                         socketid, lcore_id, NB_SOCKETS);
1222                                 free(route_base_ipv4);
1223                                 free(route_base_ipv6);
1224                                 free(acl_base_ipv4);
1225                                 free(acl_base_ipv6);
1226                                 return -1;
1227                         }
1228
1229                         acl_config.mapped[socketid] = 1;
1230                 }
1231         }
1232
1233         for (i = 0; i < NB_SOCKETS; i++) {
1234                 if (acl_config.mapped[i]) {
1235                         acl_config.acx_ipv4[i] = setup_acl(route_base_ipv4,
1236                                 acl_base_ipv4, route_num_ipv4, acl_num_ipv4,
1237                                 0, i);
1238
1239                         acl_config.acx_ipv6[i] = setup_acl(route_base_ipv6,
1240                                 acl_base_ipv6, route_num_ipv6, acl_num_ipv6,
1241                                 1, i);
1242                 }
1243         }
1244
1245         free(route_base_ipv4);
1246         free(route_base_ipv6);
1247
1248 #ifdef L3FWDACL_DEBUG
1249         acl_config.rule_ipv4 = (struct acl4_rule *)acl_base_ipv4;
1250         acl_config.rule_ipv6 = (struct acl6_rule *)acl_base_ipv6;
1251 #else
1252         free(acl_base_ipv4);
1253         free(acl_base_ipv6);
1254 #endif
1255
1256         return 0;
1257 }
1258
1259 /***********************end of ACL part******************************/
1260
1261 struct lcore_conf {
1262         uint16_t n_rx_queue;
1263         struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
1264         uint16_t n_tx_port;
1265         uint16_t tx_port_id[RTE_MAX_ETHPORTS];
1266         uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
1267         struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
1268 } __rte_cache_aligned;
1269
1270 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
1271
1272 /* Enqueue a single packet, and send burst if queue is filled */
1273 static inline void
1274 send_single_packet(struct rte_mbuf *m, uint16_t port)
1275 {
1276         uint32_t lcore_id;
1277         struct lcore_conf *qconf;
1278
1279         lcore_id = rte_lcore_id();
1280
1281         qconf = &lcore_conf[lcore_id];
1282         rte_eth_tx_buffer(port, qconf->tx_queue_id[port],
1283                         qconf->tx_buffer[port], m);
1284 }
1285
1286 #ifdef DO_RFC_1812_CHECKS
1287 static inline int
1288 is_valid_ipv4_pkt(struct rte_ipv4_hdr *pkt, uint32_t link_len)
1289 {
1290         /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
1291         /*
1292          * 1. The packet length reported by the Link Layer must be large
1293          * enough to hold the minimum length legal IP datagram (20 bytes).
1294          */
1295         if (link_len < sizeof(struct rte_ipv4_hdr))
1296                 return -1;
1297
1298         /* 2. The IP checksum must be correct. */
1299         /* this is checked in H/W */
1300
1301         /*
1302          * 3. The IP version number must be 4. If the version number is not 4
1303          * then the packet may be another version of IP, such as IPng or
1304          * ST-II.
1305          */
1306         if (((pkt->version_ihl) >> 4) != 4)
1307                 return -3;
1308         /*
1309          * 4. The IP header length field must be large enough to hold the
1310          * minimum length legal IP datagram (20 bytes = 5 words).
1311          */
1312         if ((pkt->version_ihl & 0xf) < 5)
1313                 return -4;
1314
1315         /*
1316          * 5. The IP total length field must be large enough to hold the IP
1317          * datagram header, whose length is specified in the IP header length
1318          * field.
1319          */
1320         if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct rte_ipv4_hdr))
1321                 return -5;
1322
1323         return 0;
1324 }
1325 #endif
1326
1327 /* main processing loop */
1328 static int
1329 main_loop(__rte_unused void *dummy)
1330 {
1331         struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1332         unsigned lcore_id;
1333         uint64_t prev_tsc, diff_tsc, cur_tsc;
1334         int i, nb_rx;
1335         uint16_t portid;
1336         uint8_t queueid;
1337         struct lcore_conf *qconf;
1338         int socketid;
1339         const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
1340                         / US_PER_S * BURST_TX_DRAIN_US;
1341
1342         prev_tsc = 0;
1343         lcore_id = rte_lcore_id();
1344         qconf = &lcore_conf[lcore_id];
1345         socketid = rte_lcore_to_socket_id(lcore_id);
1346
1347         if (qconf->n_rx_queue == 0) {
1348                 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
1349                 return 0;
1350         }
1351
1352         RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
1353
1354         for (i = 0; i < qconf->n_rx_queue; i++) {
1355
1356                 portid = qconf->rx_queue_list[i].port_id;
1357                 queueid = qconf->rx_queue_list[i].queue_id;
1358                 RTE_LOG(INFO, L3FWD,
1359                         " -- lcoreid=%u portid=%u rxqueueid=%hhu\n",
1360                         lcore_id, portid, queueid);
1361         }
1362
1363         while (1) {
1364
1365                 cur_tsc = rte_rdtsc();
1366
1367                 /*
1368                  * TX burst queue drain
1369                  */
1370                 diff_tsc = cur_tsc - prev_tsc;
1371                 if (unlikely(diff_tsc > drain_tsc)) {
1372                         for (i = 0; i < qconf->n_tx_port; ++i) {
1373                                 portid = qconf->tx_port_id[i];
1374                                 rte_eth_tx_buffer_flush(portid,
1375                                                 qconf->tx_queue_id[portid],
1376                                                 qconf->tx_buffer[portid]);
1377                         }
1378                         prev_tsc = cur_tsc;
1379                 }
1380
1381                 /*
1382                  * Read packet from RX queues
1383                  */
1384                 for (i = 0; i < qconf->n_rx_queue; ++i) {
1385
1386                         portid = qconf->rx_queue_list[i].port_id;
1387                         queueid = qconf->rx_queue_list[i].queue_id;
1388                         nb_rx = rte_eth_rx_burst(portid, queueid,
1389                                 pkts_burst, MAX_PKT_BURST);
1390
1391                         if (nb_rx > 0) {
1392                                 struct acl_search_t acl_search;
1393
1394                                 prepare_acl_parameter(pkts_burst, &acl_search,
1395                                         nb_rx);
1396
1397                                 if (acl_search.num_ipv4) {
1398                                         rte_acl_classify(
1399                                                 acl_config.acx_ipv4[socketid],
1400                                                 acl_search.data_ipv4,
1401                                                 acl_search.res_ipv4,
1402                                                 acl_search.num_ipv4,
1403                                                 DEFAULT_MAX_CATEGORIES);
1404
1405                                         send_packets(acl_search.m_ipv4,
1406                                                 acl_search.res_ipv4,
1407                                                 acl_search.num_ipv4);
1408                                 }
1409
1410                                 if (acl_search.num_ipv6) {
1411                                         rte_acl_classify(
1412                                                 acl_config.acx_ipv6[socketid],
1413                                                 acl_search.data_ipv6,
1414                                                 acl_search.res_ipv6,
1415                                                 acl_search.num_ipv6,
1416                                                 DEFAULT_MAX_CATEGORIES);
1417
1418                                         send_packets(acl_search.m_ipv6,
1419                                                 acl_search.res_ipv6,
1420                                                 acl_search.num_ipv6);
1421                                 }
1422                         }
1423                 }
1424         }
1425 }
1426
1427 static int
1428 check_lcore_params(void)
1429 {
1430         uint8_t queue, lcore;
1431         uint16_t i;
1432         int socketid;
1433
1434         for (i = 0; i < nb_lcore_params; ++i) {
1435                 queue = lcore_params[i].queue_id;
1436                 if (queue >= MAX_RX_QUEUE_PER_PORT) {
1437                         printf("invalid queue number: %hhu\n", queue);
1438                         return -1;
1439                 }
1440                 lcore = lcore_params[i].lcore_id;
1441                 if (!rte_lcore_is_enabled(lcore)) {
1442                         printf("error: lcore %hhu is not enabled in "
1443                                 "lcore mask\n", lcore);
1444                         return -1;
1445                 }
1446                 socketid = rte_lcore_to_socket_id(lcore);
1447                 if (socketid != 0 && numa_on == 0) {
1448                         printf("warning: lcore %hhu is on socket %d "
1449                                 "with numa off\n",
1450                                 lcore, socketid);
1451                 }
1452         }
1453         return 0;
1454 }
1455
1456 static int
1457 check_port_config(void)
1458 {
1459         unsigned portid;
1460         uint16_t i;
1461
1462         for (i = 0; i < nb_lcore_params; ++i) {
1463                 portid = lcore_params[i].port_id;
1464
1465                 if ((enabled_port_mask & (1 << portid)) == 0) {
1466                         printf("port %u is not enabled in port mask\n", portid);
1467                         return -1;
1468                 }
1469                 if (!rte_eth_dev_is_valid_port(portid)) {
1470                         printf("port %u is not present on the board\n", portid);
1471                         return -1;
1472                 }
1473         }
1474         return 0;
1475 }
1476
1477 static uint8_t
1478 get_port_n_rx_queues(const uint16_t port)
1479 {
1480         int queue = -1;
1481         uint16_t i;
1482
1483         for (i = 0; i < nb_lcore_params; ++i) {
1484                 if (lcore_params[i].port_id == port &&
1485                                 lcore_params[i].queue_id > queue)
1486                         queue = lcore_params[i].queue_id;
1487         }
1488         return (uint8_t)(++queue);
1489 }
1490
1491 static int
1492 init_lcore_rx_queues(void)
1493 {
1494         uint16_t i, nb_rx_queue;
1495         uint8_t lcore;
1496
1497         for (i = 0; i < nb_lcore_params; ++i) {
1498                 lcore = lcore_params[i].lcore_id;
1499                 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
1500                 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
1501                         printf("error: too many queues (%u) for lcore: %u\n",
1502                                 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
1503                         return -1;
1504                 } else {
1505                         lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
1506                                 lcore_params[i].port_id;
1507                         lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
1508                                 lcore_params[i].queue_id;
1509                         lcore_conf[lcore].n_rx_queue++;
1510                 }
1511         }
1512         return 0;
1513 }
1514
1515 /* display usage */
1516 static void
1517 print_usage(const char *prgname)
1518 {
1519         printf("%s [EAL options] -- -p PORTMASK -P"
1520                 "--"OPTION_RULE_IPV4"=FILE"
1521                 "--"OPTION_RULE_IPV6"=FILE"
1522                 "  [--"OPTION_CONFIG" (port,queue,lcore)[,(port,queue,lcore]]"
1523                 "  [--"OPTION_ENBJMO" [--max-pkt-len PKTLEN]]\n"
1524                 "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
1525                 "  -P : enable promiscuous mode\n"
1526                 "  --"OPTION_CONFIG": (port,queue,lcore): "
1527                 "rx queues configuration\n"
1528                 "  --"OPTION_NONUMA": optional, disable numa awareness\n"
1529                 "  --"OPTION_ENBJMO": enable jumbo frame"
1530                 " which max packet len is PKTLEN in decimal (64-9600)\n"
1531                 "  --"OPTION_RULE_IPV4"=FILE: specify the ipv4 rules entries "
1532                 "file. "
1533                 "Each rule occupy one line. "
1534                 "2 kinds of rules are supported. "
1535                 "One is ACL entry at while line leads with character '%c', "
1536                 "another is route entry at while line leads with "
1537                 "character '%c'.\n"
1538                 "  --"OPTION_RULE_IPV6"=FILE: specify the ipv6 rules "
1539                 "entries file.\n"
1540                 "  --"OPTION_SCALAR": Use scalar function to do lookup\n",
1541                 prgname, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR);
1542 }
1543
1544 static int
1545 parse_max_pkt_len(const char *pktlen)
1546 {
1547         char *end = NULL;
1548         unsigned long len;
1549
1550         /* parse decimal string */
1551         len = strtoul(pktlen, &end, 10);
1552         if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1553                 return -1;
1554
1555         if (len == 0)
1556                 return -1;
1557
1558         return len;
1559 }
1560
1561 static int
1562 parse_portmask(const char *portmask)
1563 {
1564         char *end = NULL;
1565         unsigned long pm;
1566
1567         /* parse hexadecimal string */
1568         pm = strtoul(portmask, &end, 16);
1569         if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1570                 return -1;
1571
1572         if (pm == 0)
1573                 return -1;
1574
1575         return pm;
1576 }
1577
1578 static int
1579 parse_config(const char *q_arg)
1580 {
1581         char s[256];
1582         const char *p, *p0 = q_arg;
1583         char *end;
1584         enum fieldnames {
1585                 FLD_PORT = 0,
1586                 FLD_QUEUE,
1587                 FLD_LCORE,
1588                 _NUM_FLD
1589         };
1590         unsigned long int_fld[_NUM_FLD];
1591         char *str_fld[_NUM_FLD];
1592         int i;
1593         unsigned size;
1594
1595         nb_lcore_params = 0;
1596
1597         while ((p = strchr(p0, '(')) != NULL) {
1598                 ++p;
1599                 if ((p0 = strchr(p, ')')) == NULL)
1600                         return -1;
1601
1602                 size = p0 - p;
1603                 if (size >= sizeof(s))
1604                         return -1;
1605
1606                 snprintf(s, sizeof(s), "%.*s", size, p);
1607                 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') !=
1608                                 _NUM_FLD)
1609                         return -1;
1610                 for (i = 0; i < _NUM_FLD; i++) {
1611                         errno = 0;
1612                         int_fld[i] = strtoul(str_fld[i], &end, 0);
1613                         if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1614                                 return -1;
1615                 }
1616                 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1617                         printf("exceeded max number of lcore params: %hu\n",
1618                                 nb_lcore_params);
1619                         return -1;
1620                 }
1621                 lcore_params_array[nb_lcore_params].port_id =
1622                         (uint8_t)int_fld[FLD_PORT];
1623                 lcore_params_array[nb_lcore_params].queue_id =
1624                         (uint8_t)int_fld[FLD_QUEUE];
1625                 lcore_params_array[nb_lcore_params].lcore_id =
1626                         (uint8_t)int_fld[FLD_LCORE];
1627                 ++nb_lcore_params;
1628         }
1629         lcore_params = lcore_params_array;
1630         return 0;
1631 }
1632
1633 /* Parse the argument given in the command line of the application */
1634 static int
1635 parse_args(int argc, char **argv)
1636 {
1637         int opt, ret;
1638         char **argvopt;
1639         int option_index;
1640         char *prgname = argv[0];
1641         static struct option lgopts[] = {
1642                 {OPTION_CONFIG, 1, 0, 0},
1643                 {OPTION_NONUMA, 0, 0, 0},
1644                 {OPTION_ENBJMO, 0, 0, 0},
1645                 {OPTION_RULE_IPV4, 1, 0, 0},
1646                 {OPTION_RULE_IPV6, 1, 0, 0},
1647                 {OPTION_SCALAR, 0, 0, 0},
1648                 {NULL, 0, 0, 0}
1649         };
1650
1651         argvopt = argv;
1652
1653         while ((opt = getopt_long(argc, argvopt, "p:P",
1654                                 lgopts, &option_index)) != EOF) {
1655
1656                 switch (opt) {
1657                 /* portmask */
1658                 case 'p':
1659                         enabled_port_mask = parse_portmask(optarg);
1660                         if (enabled_port_mask == 0) {
1661                                 printf("invalid portmask\n");
1662                                 print_usage(prgname);
1663                                 return -1;
1664                         }
1665                         break;
1666                 case 'P':
1667                         printf("Promiscuous mode selected\n");
1668                         promiscuous_on = 1;
1669                         break;
1670
1671                 /* long options */
1672                 case 0:
1673                         if (!strncmp(lgopts[option_index].name,
1674                                         OPTION_CONFIG,
1675                                         sizeof(OPTION_CONFIG))) {
1676                                 ret = parse_config(optarg);
1677                                 if (ret) {
1678                                         printf("invalid config\n");
1679                                         print_usage(prgname);
1680                                         return -1;
1681                                 }
1682                         }
1683
1684                         if (!strncmp(lgopts[option_index].name,
1685                                         OPTION_NONUMA,
1686                                         sizeof(OPTION_NONUMA))) {
1687                                 printf("numa is disabled\n");
1688                                 numa_on = 0;
1689                         }
1690
1691                         if (!strncmp(lgopts[option_index].name,
1692                                         OPTION_ENBJMO, sizeof(OPTION_ENBJMO))) {
1693                                 struct option lenopts = {
1694                                         "max-pkt-len",
1695                                         required_argument,
1696                                         0,
1697                                         0
1698                                 };
1699
1700                                 printf("jumbo frame is enabled\n");
1701                                 port_conf.rxmode.offloads |=
1702                                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1703                                 port_conf.txmode.offloads |=
1704                                                 DEV_TX_OFFLOAD_MULTI_SEGS;
1705
1706                                 /*
1707                                  * if no max-pkt-len set, then use the
1708                                  * default value RTE_ETHER_MAX_LEN
1709                                  */
1710                                 if (0 == getopt_long(argc, argvopt, "",
1711                                                 &lenopts, &option_index)) {
1712                                         ret = parse_max_pkt_len(optarg);
1713                                         if ((ret < 64) ||
1714                                                 (ret > MAX_JUMBO_PKT_LEN)) {
1715                                                 printf("invalid packet "
1716                                                         "length\n");
1717                                                 print_usage(prgname);
1718                                                 return -1;
1719                                         }
1720                                         port_conf.rxmode.max_rx_pkt_len = ret;
1721                                 }
1722                                 printf("set jumbo frame max packet length "
1723                                         "to %u\n",
1724                                         (unsigned int)
1725                                         port_conf.rxmode.max_rx_pkt_len);
1726                         }
1727
1728                         if (!strncmp(lgopts[option_index].name,
1729                                         OPTION_RULE_IPV4,
1730                                         sizeof(OPTION_RULE_IPV4)))
1731                                 parm_config.rule_ipv4_name = optarg;
1732
1733                         if (!strncmp(lgopts[option_index].name,
1734                                         OPTION_RULE_IPV6,
1735                                         sizeof(OPTION_RULE_IPV6))) {
1736                                 parm_config.rule_ipv6_name = optarg;
1737                         }
1738
1739                         if (!strncmp(lgopts[option_index].name,
1740                                         OPTION_SCALAR, sizeof(OPTION_SCALAR)))
1741                                 parm_config.scalar = 1;
1742
1743
1744                         break;
1745
1746                 default:
1747                         print_usage(prgname);
1748                         return -1;
1749                 }
1750         }
1751
1752         if (optind >= 0)
1753                 argv[optind-1] = prgname;
1754
1755         ret = optind-1;
1756         optind = 1; /* reset getopt lib */
1757         return ret;
1758 }
1759
1760 static void
1761 print_ethaddr(const char *name, const struct rte_ether_addr *eth_addr)
1762 {
1763         char buf[RTE_ETHER_ADDR_FMT_SIZE];
1764         rte_ether_format_addr(buf, RTE_ETHER_ADDR_FMT_SIZE, eth_addr);
1765         printf("%s%s", name, buf);
1766 }
1767
1768 static int
1769 init_mem(unsigned nb_mbuf)
1770 {
1771         int socketid;
1772         unsigned lcore_id;
1773         char s[64];
1774
1775         for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1776                 if (rte_lcore_is_enabled(lcore_id) == 0)
1777                         continue;
1778
1779                 if (numa_on)
1780                         socketid = rte_lcore_to_socket_id(lcore_id);
1781                 else
1782                         socketid = 0;
1783
1784                 if (socketid >= NB_SOCKETS) {
1785                         rte_exit(EXIT_FAILURE,
1786                                 "Socket %d of lcore %u is out of range %d\n",
1787                                 socketid, lcore_id, NB_SOCKETS);
1788                 }
1789                 if (pktmbuf_pool[socketid] == NULL) {
1790                         snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
1791                         pktmbuf_pool[socketid] =
1792                                 rte_pktmbuf_pool_create(s, nb_mbuf,
1793                                         MEMPOOL_CACHE_SIZE, 0,
1794                                         RTE_MBUF_DEFAULT_BUF_SIZE,
1795                                         socketid);
1796                         if (pktmbuf_pool[socketid] == NULL)
1797                                 rte_exit(EXIT_FAILURE,
1798                                         "Cannot init mbuf pool on socket %d\n",
1799                                         socketid);
1800                         else
1801                                 printf("Allocated mbuf pool on socket %d\n",
1802                                         socketid);
1803                 }
1804         }
1805         return 0;
1806 }
1807
1808 /* Check the link status of all ports in up to 9s, and print them finally */
1809 static void
1810 check_all_ports_link_status(uint32_t port_mask)
1811 {
1812 #define CHECK_INTERVAL 100 /* 100ms */
1813 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1814         uint16_t portid;
1815         uint8_t count, all_ports_up, print_flag = 0;
1816         struct rte_eth_link link;
1817         int ret;
1818
1819         printf("\nChecking link status");
1820         fflush(stdout);
1821         for (count = 0; count <= MAX_CHECK_TIME; count++) {
1822                 all_ports_up = 1;
1823                 RTE_ETH_FOREACH_DEV(portid) {
1824                         if ((port_mask & (1 << portid)) == 0)
1825                                 continue;
1826                         memset(&link, 0, sizeof(link));
1827                         ret = rte_eth_link_get_nowait(portid, &link);
1828                         if (ret < 0) {
1829                                 all_ports_up = 0;
1830                                 if (print_flag == 1)
1831                                         printf("Port %u link get failed: %s\n",
1832                                                 portid, rte_strerror(-ret));
1833                                 continue;
1834                         }
1835                         /* print link status if flag set */
1836                         if (print_flag == 1) {
1837                                 if (link.link_status)
1838                                         printf(
1839                                         "Port%d Link Up. Speed %u Mbps %s\n",
1840                                                 portid, link.link_speed,
1841                                 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
1842                                         ("full-duplex") : ("half-duplex"));
1843                                 else
1844                                         printf("Port %d Link Down\n", portid);
1845                                 continue;
1846                         }
1847                         /* clear all_ports_up flag if any link down */
1848                         if (link.link_status == ETH_LINK_DOWN) {
1849                                 all_ports_up = 0;
1850                                 break;
1851                         }
1852                 }
1853                 /* after finally printing all link status, get out */
1854                 if (print_flag == 1)
1855                         break;
1856
1857                 if (all_ports_up == 0) {
1858                         printf(".");
1859                         fflush(stdout);
1860                         rte_delay_ms(CHECK_INTERVAL);
1861                 }
1862
1863                 /* set the print_flag if all ports up or timeout */
1864                 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1865                         print_flag = 1;
1866                         printf("done\n");
1867                 }
1868         }
1869 }
1870
1871 int
1872 main(int argc, char **argv)
1873 {
1874         struct lcore_conf *qconf;
1875         struct rte_eth_dev_info dev_info;
1876         struct rte_eth_txconf *txconf;
1877         int ret;
1878         unsigned nb_ports;
1879         uint16_t queueid;
1880         unsigned lcore_id;
1881         uint32_t n_tx_queue, nb_lcores;
1882         uint16_t portid;
1883         uint8_t nb_rx_queue, queue, socketid;
1884
1885         /* init EAL */
1886         ret = rte_eal_init(argc, argv);
1887         if (ret < 0)
1888                 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1889         argc -= ret;
1890         argv += ret;
1891
1892         /* parse application arguments (after the EAL ones) */
1893         ret = parse_args(argc, argv);
1894         if (ret < 0)
1895                 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
1896
1897         if (check_lcore_params() < 0)
1898                 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
1899
1900         ret = init_lcore_rx_queues();
1901         if (ret < 0)
1902                 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
1903
1904         nb_ports = rte_eth_dev_count_avail();
1905
1906         if (check_port_config() < 0)
1907                 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
1908
1909         /* Add ACL rules and route entries, build trie */
1910         if (app_acl_init() < 0)
1911                 rte_exit(EXIT_FAILURE, "app_acl_init failed\n");
1912
1913         nb_lcores = rte_lcore_count();
1914
1915         /* initialize all ports */
1916         RTE_ETH_FOREACH_DEV(portid) {
1917                 struct rte_eth_conf local_port_conf = port_conf;
1918
1919                 /* skip ports that are not enabled */
1920                 if ((enabled_port_mask & (1 << portid)) == 0) {
1921                         printf("\nSkipping disabled port %d\n", portid);
1922                         continue;
1923                 }
1924
1925                 /* init port */
1926                 printf("Initializing port %d ... ", portid);
1927                 fflush(stdout);
1928
1929                 nb_rx_queue = get_port_n_rx_queues(portid);
1930                 n_tx_queue = nb_lcores;
1931                 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1932                         n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1933                 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
1934                         nb_rx_queue, (unsigned)n_tx_queue);
1935
1936                 ret = rte_eth_dev_info_get(portid, &dev_info);
1937                 if (ret != 0)
1938                         rte_exit(EXIT_FAILURE,
1939                                 "Error during getting device (port %u) info: %s\n",
1940                                 portid, strerror(-ret));
1941
1942                 if (dev_info.tx_offload_capa & DEV_TX_OFFLOAD_MBUF_FAST_FREE)
1943                         local_port_conf.txmode.offloads |=
1944                                 DEV_TX_OFFLOAD_MBUF_FAST_FREE;
1945
1946                 local_port_conf.rx_adv_conf.rss_conf.rss_hf &=
1947                         dev_info.flow_type_rss_offloads;
1948                 if (local_port_conf.rx_adv_conf.rss_conf.rss_hf !=
1949                                 port_conf.rx_adv_conf.rss_conf.rss_hf) {
1950                         printf("Port %u modified RSS hash function based on hardware support,"
1951                                 "requested:%#"PRIx64" configured:%#"PRIx64"\n",
1952                                 portid,
1953                                 port_conf.rx_adv_conf.rss_conf.rss_hf,
1954                                 local_port_conf.rx_adv_conf.rss_conf.rss_hf);
1955                 }
1956
1957                 ret = rte_eth_dev_configure(portid, nb_rx_queue,
1958                                         (uint16_t)n_tx_queue, &local_port_conf);
1959                 if (ret < 0)
1960                         rte_exit(EXIT_FAILURE,
1961                                 "Cannot configure device: err=%d, port=%d\n",
1962                                 ret, portid);
1963
1964                 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
1965                                                        &nb_txd);
1966                 if (ret < 0)
1967                         rte_exit(EXIT_FAILURE,
1968                                 "rte_eth_dev_adjust_nb_rx_tx_desc: err=%d, port=%d\n",
1969                                 ret, portid);
1970
1971                 ret = rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1972                 if (ret < 0)
1973                         rte_exit(EXIT_FAILURE,
1974                                 "rte_eth_macaddr_get: err=%d, port=%d\n",
1975                                 ret, portid);
1976
1977                 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1978                 printf(", ");
1979
1980                 /* init memory */
1981                 ret = init_mem(NB_MBUF);
1982                 if (ret < 0)
1983                         rte_exit(EXIT_FAILURE, "init_mem failed\n");
1984
1985                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1986                         if (rte_lcore_is_enabled(lcore_id) == 0)
1987                                 continue;
1988
1989                         /* Initialize TX buffers */
1990                         qconf = &lcore_conf[lcore_id];
1991                         qconf->tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
1992                                         RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
1993                                         rte_eth_dev_socket_id(portid));
1994                         if (qconf->tx_buffer[portid] == NULL)
1995                                 rte_exit(EXIT_FAILURE, "Can't allocate tx buffer for port %u\n",
1996                                                 (unsigned) portid);
1997
1998                         rte_eth_tx_buffer_init(qconf->tx_buffer[portid], MAX_PKT_BURST);
1999                 }
2000
2001                 /* init one TX queue per couple (lcore,port) */
2002                 queueid = 0;
2003                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2004                         if (rte_lcore_is_enabled(lcore_id) == 0)
2005                                 continue;
2006
2007                         if (numa_on)
2008                                 socketid = (uint8_t)
2009                                         rte_lcore_to_socket_id(lcore_id);
2010                         else
2011                                 socketid = 0;
2012
2013                         printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
2014                         fflush(stdout);
2015
2016                         ret = rte_eth_dev_info_get(portid, &dev_info);
2017                         if (ret != 0)
2018                                 rte_exit(EXIT_FAILURE,
2019                                         "Error during getting device (port %u) info: %s\n",
2020                                         portid, strerror(-ret));
2021
2022                         txconf = &dev_info.default_txconf;
2023                         txconf->offloads = local_port_conf.txmode.offloads;
2024                         ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
2025                                                      socketid, txconf);
2026                         if (ret < 0)
2027                                 rte_exit(EXIT_FAILURE,
2028                                         "rte_eth_tx_queue_setup: err=%d, "
2029                                         "port=%d\n", ret, portid);
2030
2031                         qconf = &lcore_conf[lcore_id];
2032                         qconf->tx_queue_id[portid] = queueid;
2033                         queueid++;
2034
2035                         qconf->tx_port_id[qconf->n_tx_port] = portid;
2036                         qconf->n_tx_port++;
2037                 }
2038                 printf("\n");
2039         }
2040
2041         for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2042                 if (rte_lcore_is_enabled(lcore_id) == 0)
2043                         continue;
2044                 qconf = &lcore_conf[lcore_id];
2045                 printf("\nInitializing rx queues on lcore %u ... ", lcore_id);
2046                 fflush(stdout);
2047                 /* init RX queues */
2048                 for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
2049                         struct rte_eth_rxconf rxq_conf;
2050
2051                         portid = qconf->rx_queue_list[queue].port_id;
2052                         queueid = qconf->rx_queue_list[queue].queue_id;
2053
2054                         if (numa_on)
2055                                 socketid = (uint8_t)
2056                                         rte_lcore_to_socket_id(lcore_id);
2057                         else
2058                                 socketid = 0;
2059
2060                         printf("rxq=%d,%d,%d ", portid, queueid, socketid);
2061                         fflush(stdout);
2062
2063                         ret = rte_eth_dev_info_get(portid, &dev_info);
2064                         if (ret != 0)
2065                                 rte_exit(EXIT_FAILURE,
2066                                         "Error during getting device (port %u) info: %s\n",
2067                                         portid, strerror(-ret));
2068
2069                         rxq_conf = dev_info.default_rxconf;
2070                         rxq_conf.offloads = port_conf.rxmode.offloads;
2071                         ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
2072                                         socketid, &rxq_conf,
2073                                         pktmbuf_pool[socketid]);
2074                         if (ret < 0)
2075                                 rte_exit(EXIT_FAILURE,
2076                                         "rte_eth_rx_queue_setup: err=%d,"
2077                                         "port=%d\n", ret, portid);
2078                 }
2079         }
2080
2081         printf("\n");
2082
2083         /* start ports */
2084         RTE_ETH_FOREACH_DEV(portid) {
2085                 if ((enabled_port_mask & (1 << portid)) == 0)
2086                         continue;
2087
2088                 /* Start device */
2089                 ret = rte_eth_dev_start(portid);
2090                 if (ret < 0)
2091                         rte_exit(EXIT_FAILURE,
2092                                 "rte_eth_dev_start: err=%d, port=%d\n",
2093                                 ret, portid);
2094
2095                 /*
2096                  * If enabled, put device in promiscuous mode.
2097                  * This allows IO forwarding mode to forward packets
2098                  * to itself through 2 cross-connected  ports of the
2099                  * target machine.
2100                  */
2101                 if (promiscuous_on) {
2102                         ret = rte_eth_promiscuous_enable(portid);
2103                         if (ret != 0)
2104                                 rte_exit(EXIT_FAILURE,
2105                                         "rte_eth_promiscuous_enable: err=%s, port=%u\n",
2106                                         rte_strerror(-ret), portid);
2107                 }
2108         }
2109
2110         check_all_ports_link_status(enabled_port_mask);
2111
2112         /* launch per-lcore init on every lcore */
2113         rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
2114         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
2115                 if (rte_eal_wait_lcore(lcore_id) < 0)
2116                         return -1;
2117         }
2118
2119         return 0;
2120 }