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_cycles.h>
  24 #include <rte_prefetch.h>
  25 #include <rte_lcore.h>
  26 #include <rte_per_lcore.h>
  27 #include <rte_branch_prediction.h>
  28 #include <rte_interrupts.h>
  29 #include <rte_random.h>
  30 #include <rte_debug.h>
  31 #include <rte_ether.h>
  32 #include <rte_ethdev.h>
  33 #include <rte_mempool.h>
  34 #include <rte_mbuf.h>
  35 #include <rte_ip.h>
  36 #include <rte_tcp.h>
  37 #include <rte_udp.h>
  38 #include <rte_string_fns.h>
  39 #include <rte_acl.h>
  40
  41 #include <cmdline_parse.h>
  42 #include <cmdline_parse_etheraddr.h>
  43
  44 #if RTE_LOG_DP_LEVEL >= RTE_LOG_DEBUG
  45 #define L3FWDACL_DEBUG
  46 #endif
  47 #define DO_RFC_1812_CHECKS
  48
  49 #define RTE_LOGTYPE_L3FWD RTE_LOGTYPE_USER1
  50
  51 #define MAX_JUMBO_PKT_LEN  9600
  52
  53 #define MEMPOOL_CACHE_SIZE 256
  54
  55 /*
  56  * This expression is used to calculate the number of mbufs needed
  57  * depending on user input, taking into account memory for rx and tx hardware
  58  * rings, cache per lcore and mtable per port per lcore.
  59  * RTE_MAX is used to ensure that NB_MBUF never goes below a
  60  * minimum value of 8192
  61  */
  62
  63 #define NB_MBUF RTE_MAX(\
  64         (nb_ports * nb_rx_queue * nb_rxd +      \
  65         nb_ports * nb_lcores * MAX_PKT_BURST +  \
  66         nb_ports * n_tx_queue * nb_txd +        \
  67         nb_lcores * MEMPOOL_CACHE_SIZE),        \
  68         (unsigned)8192)
  69
  70 #define MAX_PKT_BURST 32
  71 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
  72
  73 #define NB_SOCKETS 8
  74
  75 /* Configure how many packets ahead to prefetch, when reading packets */
  76 #define PREFETCH_OFFSET 3
  77
  78 /*
  79  * Configurable number of RX/TX ring descriptors
  80  */
  81 #define RTE_TEST_RX_DESC_DEFAULT 1024
  82 #define RTE_TEST_TX_DESC_DEFAULT 1024
  83 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
  84 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
  85
  86 /* mask of enabled ports */
  87 static uint32_t enabled_port_mask;
  88 static int promiscuous_on; /**< Ports set in promiscuous mode off by default. */
  89 static int numa_on = 1; /**< NUMA is enabled by default. */
  90
  91 struct lcore_rx_queue {
  92         uint16_t port_id;
  93         uint8_t queue_id;
  94 } __rte_cache_aligned;
  95
  96 #define MAX_RX_QUEUE_PER_LCORE 16
  97 #define MAX_TX_QUEUE_PER_PORT RTE_MAX_ETHPORTS
  98 #define MAX_RX_QUEUE_PER_PORT 128
  99
 100 #define MAX_LCORE_PARAMS 1024
 101 struct lcore_params {
 102         uint16_t port_id;
 103         uint8_t queue_id;
 104         uint8_t lcore_id;
 105 } __rte_cache_aligned;
 106
 107 static struct lcore_params lcore_params_array[MAX_LCORE_PARAMS];
 108 static struct lcore_params lcore_params_array_default[] = {
 109         {0, 0, 2},
 110         {0, 1, 2},
 111         {0, 2, 2},
 112         {1, 0, 2},
 113         {1, 1, 2},
 114         {1, 2, 2},
 115         {2, 0, 2},
 116         {3, 0, 3},
 117         {3, 1, 3},
 118 };
 119
 120 static struct lcore_params *lcore_params = lcore_params_array_default;
 121 static uint16_t nb_lcore_params = sizeof(lcore_params_array_default) /
 122                                 sizeof(lcore_params_array_default[0]);
 123
 124 static struct rte_eth_conf port_conf = {
 125         .rxmode = {
 126                 .mq_mode        = RTE_ETH_MQ_RX_RSS,
 127                 .split_hdr_size = 0,
 128                 .offloads = RTE_ETH_RX_OFFLOAD_CHECKSUM,
 129         },
 130         .rx_adv_conf = {
 131                 .rss_conf = {
 132                         .rss_key = NULL,
 133                         .rss_hf = RTE_ETH_RSS_IP | RTE_ETH_RSS_UDP |
 134                                 RTE_ETH_RSS_TCP | RTE_ETH_RSS_SCTP,
 135                 },
 136         },
 137         .txmode = {
 138                 .mq_mode = RTE_ETH_MQ_TX_NONE,
 139         },
 140 };
 141
 142 static uint32_t max_pkt_len;
 143
 144 static struct rte_mempool *pktmbuf_pool[NB_SOCKETS];
 145
 146 /* ethernet addresses of ports */
 147 static struct rte_ether_hdr port_l2hdr[RTE_MAX_ETHPORTS];
 148
 149 static const struct {
 150         const char *name;
 151         enum rte_acl_classify_alg alg;
 152 } acl_alg[] = {
 153         {
 154                 .name = "scalar",
 155                 .alg = RTE_ACL_CLASSIFY_SCALAR,
 156         },
 157         {
 158                 .name = "sse",
 159                 .alg = RTE_ACL_CLASSIFY_SSE,
 160         },
 161         {
 162                 .name = "avx2",
 163                 .alg = RTE_ACL_CLASSIFY_AVX2,
 164         },
 165         {
 166                 .name = "neon",
 167                 .alg = RTE_ACL_CLASSIFY_NEON,
 168         },
 169         {
 170                 .name = "altivec",
 171                 .alg = RTE_ACL_CLASSIFY_ALTIVEC,
 172         },
 173         {
 174                 .name = "avx512x16",
 175                 .alg = RTE_ACL_CLASSIFY_AVX512X16,
 176         },
 177         {
 178                 .name = "avx512x32",
 179                 .alg = RTE_ACL_CLASSIFY_AVX512X32,
 180         },
 181 };
 182
 183 /***********************start of ACL part******************************/
 184 #ifdef DO_RFC_1812_CHECKS
 185 static inline int
 186 is_valid_ipv4_pkt(struct rte_ipv4_hdr *pkt, uint32_t link_len);
 187 #endif
 188 static inline void
 189 send_single_packet(struct rte_mbuf *m, uint16_t port);
 190
 191 #define MAX_ACL_RULE_NUM        100000
 192 #define DEFAULT_MAX_CATEGORIES  1
 193 #define L3FWD_ACL_IPV4_NAME     "l3fwd-acl-ipv4"
 194 #define L3FWD_ACL_IPV6_NAME     "l3fwd-acl-ipv6"
 195 #define ACL_LEAD_CHAR           ('@')
 196 #define ROUTE_LEAD_CHAR         ('R')
 197 #define COMMENT_LEAD_CHAR       ('#')
 198
 199 enum {
 200 #define OPT_CONFIG      "config"
 201         OPT_CONFIG_NUM = 256,
 202 #define OPT_NONUMA      "no-numa"
 203         OPT_NONUMA_NUM,
 204 #define OPT_MAX_PKT_LEN "max-pkt-len"
 205         OPT_MAX_PKT_LEN_NUM,
 206 #define OPT_RULE_IPV4   "rule_ipv4"
 207         OPT_RULE_IPV4_NUM,
 208 #define OPT_RULE_IPV6   "rule_ipv6"
 209         OPT_RULE_IPV6_NUM,
 210 #define OPT_ALG         "alg"
 211         OPT_ALG_NUM,
 212 #define OPT_ETH_DEST    "eth-dest"
 213         OPT_ETH_DEST_NUM,
 214 };
 215
 216 #define ACL_DENY_SIGNATURE      0xf0000000
 217 #define RTE_LOGTYPE_L3FWDACL    RTE_LOGTYPE_USER3
 218 #define acl_log(format, ...)    RTE_LOG(ERR, L3FWDACL, format, ##__VA_ARGS__)
 219 #define uint32_t_to_char(ip, a, b, c, d) do {\
 220                 *a = (unsigned char)(ip >> 24 & 0xff);\
 221                 *b = (unsigned char)(ip >> 16 & 0xff);\
 222                 *c = (unsigned char)(ip >> 8 & 0xff);\
 223                 *d = (unsigned char)(ip & 0xff);\
 224         } while (0)
 225 #define OFF_ETHHEAD     (sizeof(struct rte_ether_hdr))
 226 #define OFF_IPV42PROTO (offsetof(struct rte_ipv4_hdr, next_proto_id))
 227 #define OFF_IPV62PROTO (offsetof(struct rte_ipv6_hdr, proto))
 228 #define MBUF_IPV4_2PROTO(m)     \
 229         rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV42PROTO)
 230 #define MBUF_IPV6_2PROTO(m)     \
 231         rte_pktmbuf_mtod_offset((m), uint8_t *, OFF_ETHHEAD + OFF_IPV62PROTO)
 232
 233 #define GET_CB_FIELD(in, fd, base, lim, dlm)    do {            \
 234         unsigned long val;                                      \
 235         char *end;                                              \
 236         errno = 0;                                              \
 237         val = strtoul((in), &end, (base));                      \
 238         if (errno != 0 || end[0] != (dlm) || val > (lim))       \
 239                 return -EINVAL;                               \
 240         (fd) = (typeof(fd))val;                                 \
 241         (in) = end + 1;                                         \
 242 } while (0)
 243
 244 /*
 245   * ACL rules should have higher priorities than route ones to ensure ACL rule
 246   * always be found when input packets have multi-matches in the database.
 247   * A exception case is performance measure, which can define route rules with
 248   * higher priority and route rules will always be returned in each lookup.
 249   * Reserve range from ACL_RULE_PRIORITY_MAX + 1 to
 250   * RTE_ACL_MAX_PRIORITY for route entries in performance measure
 251   */
 252 #define ACL_RULE_PRIORITY_MAX 0x10000000
 253
 254 /*
 255   * Forward port info save in ACL lib starts from 1
 256   * since ACL assume 0 is invalid.
 257   * So, need add 1 when saving and minus 1 when forwarding packets.
 258   */
 259 #define FWD_PORT_SHIFT 1
 260
 261 /*
 262  * Rule and trace formats definitions.
 263  */
 264
 265 enum {
 266         PROTO_FIELD_IPV4,
 267         SRC_FIELD_IPV4,
 268         DST_FIELD_IPV4,
 269         SRCP_FIELD_IPV4,
 270         DSTP_FIELD_IPV4,
 271         NUM_FIELDS_IPV4
 272 };
 273
 274 /*
 275  * That effectively defines order of IPV4VLAN classifications:
 276  *  - PROTO
 277  *  - VLAN (TAG and DOMAIN)
 278  *  - SRC IP ADDRESS
 279  *  - DST IP ADDRESS
 280  *  - PORTS (SRC and DST)
 281  */
 282 enum {
 283         RTE_ACL_IPV4VLAN_PROTO,
 284         RTE_ACL_IPV4VLAN_VLAN,
 285         RTE_ACL_IPV4VLAN_SRC,
 286         RTE_ACL_IPV4VLAN_DST,
 287         RTE_ACL_IPV4VLAN_PORTS,
 288         RTE_ACL_IPV4VLAN_NUM
 289 };
 290
 291 struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
 292         {
 293                 .type = RTE_ACL_FIELD_TYPE_BITMASK,
 294                 .size = sizeof(uint8_t),
 295                 .field_index = PROTO_FIELD_IPV4,
 296                 .input_index = RTE_ACL_IPV4VLAN_PROTO,
 297                 .offset = 0,
 298         },
 299         {
 300                 .type = RTE_ACL_FIELD_TYPE_MASK,
 301                 .size = sizeof(uint32_t),
 302                 .field_index = SRC_FIELD_IPV4,
 303                 .input_index = RTE_ACL_IPV4VLAN_SRC,
 304                 .offset = offsetof(struct rte_ipv4_hdr, src_addr) -
 305                         offsetof(struct rte_ipv4_hdr, next_proto_id),
 306         },
 307         {
 308                 .type = RTE_ACL_FIELD_TYPE_MASK,
 309                 .size = sizeof(uint32_t),
 310                 .field_index = DST_FIELD_IPV4,
 311                 .input_index = RTE_ACL_IPV4VLAN_DST,
 312                 .offset = offsetof(struct rte_ipv4_hdr, dst_addr) -
 313                         offsetof(struct rte_ipv4_hdr, next_proto_id),
 314         },
 315         {
 316                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 317                 .size = sizeof(uint16_t),
 318                 .field_index = SRCP_FIELD_IPV4,
 319                 .input_index = RTE_ACL_IPV4VLAN_PORTS,
 320                 .offset = sizeof(struct rte_ipv4_hdr) -
 321                         offsetof(struct rte_ipv4_hdr, next_proto_id),
 322         },
 323         {
 324                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 325                 .size = sizeof(uint16_t),
 326                 .field_index = DSTP_FIELD_IPV4,
 327                 .input_index = RTE_ACL_IPV4VLAN_PORTS,
 328                 .offset = sizeof(struct rte_ipv4_hdr) -
 329                         offsetof(struct rte_ipv4_hdr, next_proto_id) +
 330                         sizeof(uint16_t),
 331         },
 332 };
 333
 334 #define IPV6_ADDR_LEN   16
 335 #define IPV6_ADDR_U16   (IPV6_ADDR_LEN / sizeof(uint16_t))
 336 #define IPV6_ADDR_U32   (IPV6_ADDR_LEN / sizeof(uint32_t))
 337
 338 enum {
 339         PROTO_FIELD_IPV6,
 340         SRC1_FIELD_IPV6,
 341         SRC2_FIELD_IPV6,
 342         SRC3_FIELD_IPV6,
 343         SRC4_FIELD_IPV6,
 344         DST1_FIELD_IPV6,
 345         DST2_FIELD_IPV6,
 346         DST3_FIELD_IPV6,
 347         DST4_FIELD_IPV6,
 348         SRCP_FIELD_IPV6,
 349         DSTP_FIELD_IPV6,
 350         NUM_FIELDS_IPV6
 351 };
 352
 353 struct rte_acl_field_def ipv6_defs[NUM_FIELDS_IPV6] = {
 354         {
 355                 .type = RTE_ACL_FIELD_TYPE_BITMASK,
 356                 .size = sizeof(uint8_t),
 357                 .field_index = PROTO_FIELD_IPV6,
 358                 .input_index = PROTO_FIELD_IPV6,
 359                 .offset = 0,
 360         },
 361         {
 362                 .type = RTE_ACL_FIELD_TYPE_MASK,
 363                 .size = sizeof(uint32_t),
 364                 .field_index = SRC1_FIELD_IPV6,
 365                 .input_index = SRC1_FIELD_IPV6,
 366                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 367                         offsetof(struct rte_ipv6_hdr, proto),
 368         },
 369         {
 370                 .type = RTE_ACL_FIELD_TYPE_MASK,
 371                 .size = sizeof(uint32_t),
 372                 .field_index = SRC2_FIELD_IPV6,
 373                 .input_index = SRC2_FIELD_IPV6,
 374                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 375                         offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint32_t),
 376         },
 377         {
 378                 .type = RTE_ACL_FIELD_TYPE_MASK,
 379                 .size = sizeof(uint32_t),
 380                 .field_index = SRC3_FIELD_IPV6,
 381                 .input_index = SRC3_FIELD_IPV6,
 382                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 383                         offsetof(struct rte_ipv6_hdr, proto) +
 384                         2 * sizeof(uint32_t),
 385         },
 386         {
 387                 .type = RTE_ACL_FIELD_TYPE_MASK,
 388                 .size = sizeof(uint32_t),
 389                 .field_index = SRC4_FIELD_IPV6,
 390                 .input_index = SRC4_FIELD_IPV6,
 391                 .offset = offsetof(struct rte_ipv6_hdr, src_addr) -
 392                         offsetof(struct rte_ipv6_hdr, proto) +
 393                         3 * sizeof(uint32_t),
 394         },
 395         {
 396                 .type = RTE_ACL_FIELD_TYPE_MASK,
 397                 .size = sizeof(uint32_t),
 398                 .field_index = DST1_FIELD_IPV6,
 399                 .input_index = DST1_FIELD_IPV6,
 400                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr)
 401                                 - offsetof(struct rte_ipv6_hdr, proto),
 402         },
 403         {
 404                 .type = RTE_ACL_FIELD_TYPE_MASK,
 405                 .size = sizeof(uint32_t),
 406                 .field_index = DST2_FIELD_IPV6,
 407                 .input_index = DST2_FIELD_IPV6,
 408                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
 409                         offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint32_t),
 410         },
 411         {
 412                 .type = RTE_ACL_FIELD_TYPE_MASK,
 413                 .size = sizeof(uint32_t),
 414                 .field_index = DST3_FIELD_IPV6,
 415                 .input_index = DST3_FIELD_IPV6,
 416                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
 417                         offsetof(struct rte_ipv6_hdr, proto) +
 418                         2 * sizeof(uint32_t),
 419         },
 420         {
 421                 .type = RTE_ACL_FIELD_TYPE_MASK,
 422                 .size = sizeof(uint32_t),
 423                 .field_index = DST4_FIELD_IPV6,
 424                 .input_index = DST4_FIELD_IPV6,
 425                 .offset = offsetof(struct rte_ipv6_hdr, dst_addr) -
 426                         offsetof(struct rte_ipv6_hdr, proto) +
 427                         3 * sizeof(uint32_t),
 428         },
 429         {
 430                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 431                 .size = sizeof(uint16_t),
 432                 .field_index = SRCP_FIELD_IPV6,
 433                 .input_index = SRCP_FIELD_IPV6,
 434                 .offset = sizeof(struct rte_ipv6_hdr) -
 435                         offsetof(struct rte_ipv6_hdr, proto),
 436         },
 437         {
 438                 .type = RTE_ACL_FIELD_TYPE_RANGE,
 439                 .size = sizeof(uint16_t),
 440                 .field_index = DSTP_FIELD_IPV6,
 441                 .input_index = SRCP_FIELD_IPV6,
 442                 .offset = sizeof(struct rte_ipv6_hdr) -
 443                         offsetof(struct rte_ipv6_hdr, proto) + sizeof(uint16_t),
 444         },
 445 };
 446
 447 enum {
 448         CB_FLD_SRC_ADDR,
 449         CB_FLD_DST_ADDR,
 450         CB_FLD_SRC_PORT_LOW,
 451         CB_FLD_SRC_PORT_DLM,
 452         CB_FLD_SRC_PORT_HIGH,
 453         CB_FLD_DST_PORT_LOW,
 454         CB_FLD_DST_PORT_DLM,
 455         CB_FLD_DST_PORT_HIGH,
 456         CB_FLD_PROTO,
 457         CB_FLD_USERDATA,
 458         CB_FLD_NUM,
 459 };
 460
 461 RTE_ACL_RULE_DEF(acl4_rule, RTE_DIM(ipv4_defs));
 462 RTE_ACL_RULE_DEF(acl6_rule, RTE_DIM(ipv6_defs));
 463
 464 struct acl_search_t {
 465         const uint8_t *data_ipv4[MAX_PKT_BURST];
 466         struct rte_mbuf *m_ipv4[MAX_PKT_BURST];
 467         uint32_t res_ipv4[MAX_PKT_BURST];
 468         int num_ipv4;
 469
 470         const uint8_t *data_ipv6[MAX_PKT_BURST];
 471         struct rte_mbuf *m_ipv6[MAX_PKT_BURST];
 472         uint32_t res_ipv6[MAX_PKT_BURST];
 473         int num_ipv6;
 474 };
 475
 476 static struct {
 477         char mapped[NB_SOCKETS];
 478         struct rte_acl_ctx *acx_ipv4[NB_SOCKETS];
 479         struct rte_acl_ctx *acx_ipv6[NB_SOCKETS];
 480 #ifdef L3FWDACL_DEBUG
 481         struct acl4_rule *rule_ipv4;
 482         struct acl6_rule *rule_ipv6;
 483 #endif
 484 } acl_config;
 485
 486 static struct{
 487         const char *rule_ipv4_name;
 488         const char *rule_ipv6_name;
 489         enum rte_acl_classify_alg alg;
 490 } parm_config;
 491
 492 const char cb_port_delim[] = ":";
 493
 494 static inline void
 495 print_one_ipv4_rule(struct acl4_rule *rule, int extra)
 496 {
 497         unsigned char a, b, c, d;
 498
 499         uint32_t_to_char(rule->field[SRC_FIELD_IPV4].value.u32,
 500                         &a, &b, &c, &d);
 501         printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
 502                         rule->field[SRC_FIELD_IPV4].mask_range.u32);
 503         uint32_t_to_char(rule->field[DST_FIELD_IPV4].value.u32,
 504                         &a, &b, &c, &d);
 505         printf("%hhu.%hhu.%hhu.%hhu/%u ", a, b, c, d,
 506                         rule->field[DST_FIELD_IPV4].mask_range.u32);
 507         printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
 508                 rule->field[SRCP_FIELD_IPV4].value.u16,
 509                 rule->field[SRCP_FIELD_IPV4].mask_range.u16,
 510                 rule->field[DSTP_FIELD_IPV4].value.u16,
 511                 rule->field[DSTP_FIELD_IPV4].mask_range.u16,
 512                 rule->field[PROTO_FIELD_IPV4].value.u8,
 513                 rule->field[PROTO_FIELD_IPV4].mask_range.u8);
 514         if (extra)
 515                 printf("0x%x-0x%x-0x%x ",
 516                         rule->data.category_mask,
 517                         rule->data.priority,
 518                         rule->data.userdata);
 519 }
 520
 521 static inline void
 522 print_one_ipv6_rule(struct acl6_rule *rule, int extra)
 523 {
 524         unsigned char a, b, c, d;
 525
 526         uint32_t_to_char(rule->field[SRC1_FIELD_IPV6].value.u32,
 527                 &a, &b, &c, &d);
 528         printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
 529         uint32_t_to_char(rule->field[SRC2_FIELD_IPV6].value.u32,
 530                 &a, &b, &c, &d);
 531         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 532         uint32_t_to_char(rule->field[SRC3_FIELD_IPV6].value.u32,
 533                 &a, &b, &c, &d);
 534         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 535         uint32_t_to_char(rule->field[SRC4_FIELD_IPV6].value.u32,
 536                 &a, &b, &c, &d);
 537         printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
 538                         rule->field[SRC1_FIELD_IPV6].mask_range.u32
 539                         + rule->field[SRC2_FIELD_IPV6].mask_range.u32
 540                         + rule->field[SRC3_FIELD_IPV6].mask_range.u32
 541                         + rule->field[SRC4_FIELD_IPV6].mask_range.u32);
 542
 543         uint32_t_to_char(rule->field[DST1_FIELD_IPV6].value.u32,
 544                 &a, &b, &c, &d);
 545         printf("%.2x%.2x:%.2x%.2x", a, b, c, d);
 546         uint32_t_to_char(rule->field[DST2_FIELD_IPV6].value.u32,
 547                 &a, &b, &c, &d);
 548         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 549         uint32_t_to_char(rule->field[DST3_FIELD_IPV6].value.u32,
 550                 &a, &b, &c, &d);
 551         printf(":%.2x%.2x:%.2x%.2x", a, b, c, d);
 552         uint32_t_to_char(rule->field[DST4_FIELD_IPV6].value.u32,
 553                 &a, &b, &c, &d);
 554         printf(":%.2x%.2x:%.2x%.2x/%u ", a, b, c, d,
 555                         rule->field[DST1_FIELD_IPV6].mask_range.u32
 556                         + rule->field[DST2_FIELD_IPV6].mask_range.u32
 557                         + rule->field[DST3_FIELD_IPV6].mask_range.u32
 558                         + rule->field[DST4_FIELD_IPV6].mask_range.u32);
 559
 560         printf("%hu : %hu %hu : %hu 0x%hhx/0x%hhx ",
 561                 rule->field[SRCP_FIELD_IPV6].value.u16,
 562                 rule->field[SRCP_FIELD_IPV6].mask_range.u16,
 563                 rule->field[DSTP_FIELD_IPV6].value.u16,
 564                 rule->field[DSTP_FIELD_IPV6].mask_range.u16,
 565                 rule->field[PROTO_FIELD_IPV6].value.u8,
 566                 rule->field[PROTO_FIELD_IPV6].mask_range.u8);
 567         if (extra)
 568                 printf("0x%x-0x%x-0x%x ",
 569                         rule->data.category_mask,
 570                         rule->data.priority,
 571                         rule->data.userdata);
 572 }
 573
 574 /* Bypass comment and empty lines */
 575 static inline int
 576 is_bypass_line(char *buff)
 577 {
 578         int i = 0;
 579
 580         /* comment line */
 581         if (buff[0] == COMMENT_LEAD_CHAR)
 582                 return 1;
 583         /* empty line */
 584         while (buff[i] != '\0') {
 585                 if (!isspace(buff[i]))
 586                         return 0;
 587                 i++;
 588         }
 589         return 1;
 590 }
 591
 592 #ifdef L3FWDACL_DEBUG
 593 static inline void
 594 dump_acl4_rule(struct rte_mbuf *m, uint32_t sig)
 595 {
 596         uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
 597         unsigned char a, b, c, d;
 598         struct rte_ipv4_hdr *ipv4_hdr =
 599                 rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
 600                                         sizeof(struct rte_ether_hdr));
 601
 602         uint32_t_to_char(rte_bswap32(ipv4_hdr->src_addr), &a, &b, &c, &d);
 603         printf("Packet Src:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
 604         uint32_t_to_char(rte_bswap32(ipv4_hdr->dst_addr), &a, &b, &c, &d);
 605         printf("Dst:%hhu.%hhu.%hhu.%hhu ", a, b, c, d);
 606
 607         printf("Src port:%hu,Dst port:%hu ",
 608                         rte_bswap16(*(uint16_t *)(ipv4_hdr + 1)),
 609                         rte_bswap16(*((uint16_t *)(ipv4_hdr + 1) + 1)));
 610         printf("hit ACL %d - ", offset);
 611
 612         print_one_ipv4_rule(acl_config.rule_ipv4 + offset, 1);
 613
 614         printf("\n\n");
 615 }
 616
 617 static inline void
 618 dump_acl6_rule(struct rte_mbuf *m, uint32_t sig)
 619 {
 620         unsigned i;
 621         uint32_t offset = sig & ~ACL_DENY_SIGNATURE;
 622         struct rte_ipv6_hdr *ipv6_hdr =
 623                 rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
 624                                         sizeof(struct rte_ether_hdr));
 625
 626         printf("Packet Src");
 627         for (i = 0; i < RTE_DIM(ipv6_hdr->src_addr); i += sizeof(uint16_t))
 628                 printf(":%.2x%.2x",
 629                         ipv6_hdr->src_addr[i], ipv6_hdr->src_addr[i + 1]);
 630
 631         printf("\nDst");
 632         for (i = 0; i < RTE_DIM(ipv6_hdr->dst_addr); i += sizeof(uint16_t))
 633                 printf(":%.2x%.2x",
 634                         ipv6_hdr->dst_addr[i], ipv6_hdr->dst_addr[i + 1]);
 635
 636         printf("\nSrc port:%hu,Dst port:%hu ",
 637                         rte_bswap16(*(uint16_t *)(ipv6_hdr + 1)),
 638                         rte_bswap16(*((uint16_t *)(ipv6_hdr + 1) + 1)));
 639         printf("hit ACL %d - ", offset);
 640
 641         print_one_ipv6_rule(acl_config.rule_ipv6 + offset, 1);
 642
 643         printf("\n\n");
 644 }
 645 #endif /* L3FWDACL_DEBUG */
 646
 647 static inline void
 648 dump_ipv4_rules(struct acl4_rule *rule, int num, int extra)
 649 {
 650         int i;
 651
 652         for (i = 0; i < num; i++, rule++) {
 653                 printf("\t%d:", i + 1);
 654                 print_one_ipv4_rule(rule, extra);
 655                 printf("\n");
 656         }
 657 }
 658
 659 static inline void
 660 dump_ipv6_rules(struct acl6_rule *rule, int num, int extra)
 661 {
 662         int i;
 663
 664         for (i = 0; i < num; i++, rule++) {
 665                 printf("\t%d:", i + 1);
 666                 print_one_ipv6_rule(rule, extra);
 667                 printf("\n");
 668         }
 669 }
 670
 671 #ifdef DO_RFC_1812_CHECKS
 672 static inline void
 673 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
 674         int index)
 675 {
 676         struct rte_ipv4_hdr *ipv4_hdr;
 677         struct rte_mbuf *pkt = pkts_in[index];
 678
 679         if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
 680                 ipv4_hdr = rte_pktmbuf_mtod_offset(pkt, struct rte_ipv4_hdr *,
 681                                                 sizeof(struct rte_ether_hdr));
 682
 683                 /* Check to make sure the packet is valid (RFC1812) */
 684                 if (is_valid_ipv4_pkt(ipv4_hdr, pkt->pkt_len) >= 0) {
 685
 686                         /* Update time to live and header checksum */
 687                         --(ipv4_hdr->time_to_live);
 688                         ++(ipv4_hdr->hdr_checksum);
 689
 690                         /* Fill acl structure */
 691                         acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
 692                         acl->m_ipv4[(acl->num_ipv4)++] = pkt;
 693
 694                 } else {
 695                         /* Not a valid IPv4 packet */
 696                         rte_pktmbuf_free(pkt);
 697                 }
 698         } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
 699                 /* Fill acl structure */
 700                 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
 701                 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
 702
 703         } else {
 704                 /* Unknown type, drop the packet */
 705                 rte_pktmbuf_free(pkt);
 706         }
 707 }
 708
 709 #else
 710 static inline void
 711 prepare_one_packet(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
 712         int index)
 713 {
 714         struct rte_mbuf *pkt = pkts_in[index];
 715
 716         if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
 717                 /* Fill acl structure */
 718                 acl->data_ipv4[acl->num_ipv4] = MBUF_IPV4_2PROTO(pkt);
 719                 acl->m_ipv4[(acl->num_ipv4)++] = pkt;
 720
 721         } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
 722                 /* Fill acl structure */
 723                 acl->data_ipv6[acl->num_ipv6] = MBUF_IPV6_2PROTO(pkt);
 724                 acl->m_ipv6[(acl->num_ipv6)++] = pkt;
 725         } else {
 726                 /* Unknown type, drop the packet */
 727                 rte_pktmbuf_free(pkt);
 728         }
 729 }
 730 #endif /* DO_RFC_1812_CHECKS */
 731
 732 static inline void
 733 prepare_acl_parameter(struct rte_mbuf **pkts_in, struct acl_search_t *acl,
 734         int nb_rx)
 735 {
 736         int i;
 737
 738         acl->num_ipv4 = 0;
 739         acl->num_ipv6 = 0;
 740
 741         /* Prefetch first packets */
 742         for (i = 0; i < PREFETCH_OFFSET && i < nb_rx; i++) {
 743                 rte_prefetch0(rte_pktmbuf_mtod(
 744                                 pkts_in[i], void *));
 745         }
 746
 747         for (i = 0; i < (nb_rx - PREFETCH_OFFSET); i++) {
 748                 rte_prefetch0(rte_pktmbuf_mtod(pkts_in[
 749                                 i + PREFETCH_OFFSET], void *));
 750                 prepare_one_packet(pkts_in, acl, i);
 751         }
 752
 753         /* Process left packets */
 754         for (; i < nb_rx; i++)
 755                 prepare_one_packet(pkts_in, acl, i);
 756 }
 757
 758 static inline void
 759 send_one_packet(struct rte_mbuf *m, uint32_t res)
 760 {
 761         if (likely((res & ACL_DENY_SIGNATURE) == 0 && res != 0)) {
 762                 /* forward packets */
 763                 send_single_packet(m,
 764                         (uint8_t)(res - FWD_PORT_SHIFT));
 765         } else{
 766                 /* in the ACL list, drop it */
 767 #ifdef L3FWDACL_DEBUG
 768                 if ((res & ACL_DENY_SIGNATURE) != 0) {
 769                         if (RTE_ETH_IS_IPV4_HDR(m->packet_type))
 770                                 dump_acl4_rule(m, res);
 771                         else if (RTE_ETH_IS_IPV6_HDR(m->packet_type))
 772                                 dump_acl6_rule(m, res);
 773                 }
 774 #endif
 775                 rte_pktmbuf_free(m);
 776         }
 777 }
 778
 779
 780
 781 static inline void
 782 send_packets(struct rte_mbuf **m, uint32_t *res, int num)
 783 {
 784         int i;
 785
 786         /* Prefetch first packets */
 787         for (i = 0; i < PREFETCH_OFFSET && i < num; i++) {
 788                 rte_prefetch0(rte_pktmbuf_mtod(
 789                                 m[i], void *));
 790         }
 791
 792         for (i = 0; i < (num - PREFETCH_OFFSET); i++) {
 793                 rte_prefetch0(rte_pktmbuf_mtod(m[
 794                                 i + PREFETCH_OFFSET], void *));
 795                 send_one_packet(m[i], res[i]);
 796         }
 797
 798         /* Process left packets */
 799         for (; i < num; i++)
 800                 send_one_packet(m[i], res[i]);
 801 }
 802
 803 /*
 804  * Parses IPV6 address, exepcts the following format:
 805  * XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX (where X - is a hexedecimal digit).
 806  */
 807 static int
 808 parse_ipv6_addr(const char *in, const char **end, uint32_t v[IPV6_ADDR_U32],
 809         char dlm)
 810 {
 811         uint32_t addr[IPV6_ADDR_U16];
 812
 813         GET_CB_FIELD(in, addr[0], 16, UINT16_MAX, ':');
 814         GET_CB_FIELD(in, addr[1], 16, UINT16_MAX, ':');
 815         GET_CB_FIELD(in, addr[2], 16, UINT16_MAX, ':');
 816         GET_CB_FIELD(in, addr[3], 16, UINT16_MAX, ':');
 817         GET_CB_FIELD(in, addr[4], 16, UINT16_MAX, ':');
 818         GET_CB_FIELD(in, addr[5], 16, UINT16_MAX, ':');
 819         GET_CB_FIELD(in, addr[6], 16, UINT16_MAX, ':');
 820         GET_CB_FIELD(in, addr[7], 16, UINT16_MAX, dlm);
 821
 822         *end = in;
 823
 824         v[0] = (addr[0] << 16) + addr[1];
 825         v[1] = (addr[2] << 16) + addr[3];
 826         v[2] = (addr[4] << 16) + addr[5];
 827         v[3] = (addr[6] << 16) + addr[7];
 828
 829         return 0;
 830 }
 831
 832 static int
 833 parse_ipv6_net(const char *in, struct rte_acl_field field[4])
 834 {
 835         int32_t rc;
 836         const char *mp;
 837         uint32_t i, m, v[4];
 838         const uint32_t nbu32 = sizeof(uint32_t) * CHAR_BIT;
 839
 840         /* get address. */
 841         rc = parse_ipv6_addr(in, &mp, v, '/');
 842         if (rc != 0)
 843                 return rc;
 844
 845         /* get mask. */
 846         GET_CB_FIELD(mp, m, 0, CHAR_BIT * sizeof(v), 0);
 847
 848         /* put all together. */
 849         for (i = 0; i != RTE_DIM(v); i++) {
 850                 if (m >= (i + 1) * nbu32)
 851                         field[i].mask_range.u32 = nbu32;
 852                 else
 853                         field[i].mask_range.u32 = m > (i * nbu32) ?
 854                                 m - (i * 32) : 0;
 855
 856                 field[i].value.u32 = v[i];
 857         }
 858
 859         return 0;
 860 }
 861
 862 static int
 863 parse_cb_ipv6_rule(char *str, struct rte_acl_rule *v, int has_userdata)
 864 {
 865         int i, rc;
 866         char *s, *sp, *in[CB_FLD_NUM];
 867         static const char *dlm = " \t\n";
 868         int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
 869         s = str;
 870
 871         for (i = 0; i != dim; i++, s = NULL) {
 872                 in[i] = strtok_r(s, dlm, &sp);
 873                 if (in[i] == NULL)
 874                         return -EINVAL;
 875         }
 876
 877         rc = parse_ipv6_net(in[CB_FLD_SRC_ADDR], v->field + SRC1_FIELD_IPV6);
 878         if (rc != 0) {
 879                 acl_log("failed to read source address/mask: %s\n",
 880                         in[CB_FLD_SRC_ADDR]);
 881                 return rc;
 882         }
 883
 884         rc = parse_ipv6_net(in[CB_FLD_DST_ADDR], v->field + DST1_FIELD_IPV6);
 885         if (rc != 0) {
 886                 acl_log("failed to read destination address/mask: %s\n",
 887                         in[CB_FLD_DST_ADDR]);
 888                 return rc;
 889         }
 890
 891         /* source port. */
 892         GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
 893                 v->field[SRCP_FIELD_IPV6].value.u16,
 894                 0, UINT16_MAX, 0);
 895         GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
 896                 v->field[SRCP_FIELD_IPV6].mask_range.u16,
 897                 0, UINT16_MAX, 0);
 898
 899         if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
 900                         sizeof(cb_port_delim)) != 0)
 901                 return -EINVAL;
 902
 903         /* destination port. */
 904         GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
 905                 v->field[DSTP_FIELD_IPV6].value.u16,
 906                 0, UINT16_MAX, 0);
 907         GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
 908                 v->field[DSTP_FIELD_IPV6].mask_range.u16,
 909                 0, UINT16_MAX, 0);
 910
 911         if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
 912                         sizeof(cb_port_delim)) != 0)
 913                 return -EINVAL;
 914
 915         if (v->field[SRCP_FIELD_IPV6].mask_range.u16
 916                         < v->field[SRCP_FIELD_IPV6].value.u16
 917                         || v->field[DSTP_FIELD_IPV6].mask_range.u16
 918                         < v->field[DSTP_FIELD_IPV6].value.u16)
 919                 return -EINVAL;
 920
 921         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].value.u8,
 922                 0, UINT8_MAX, '/');
 923         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV6].mask_range.u8,
 924                 0, UINT8_MAX, 0);
 925
 926         if (has_userdata)
 927                 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata,
 928                         0, UINT32_MAX, 0);
 929
 930         return 0;
 931 }
 932
 933 /*
 934  * Parse ClassBench rules file.
 935  * Expected format:
 936  * '@'<src_ipv4_addr>'/'<masklen> <space> \
 937  * <dst_ipv4_addr>'/'<masklen> <space> \
 938  * <src_port_low> <space> ":" <src_port_high> <space> \
 939  * <dst_port_low> <space> ":" <dst_port_high> <space> \
 940  * <proto>'/'<mask>
 941  */
 942 static int
 943 parse_ipv4_net(const char *in, uint32_t *addr, uint32_t *mask_len)
 944 {
 945         uint8_t a, b, c, d, m;
 946
 947         GET_CB_FIELD(in, a, 0, UINT8_MAX, '.');
 948         GET_CB_FIELD(in, b, 0, UINT8_MAX, '.');
 949         GET_CB_FIELD(in, c, 0, UINT8_MAX, '.');
 950         GET_CB_FIELD(in, d, 0, UINT8_MAX, '/');
 951         GET_CB_FIELD(in, m, 0, sizeof(uint32_t) * CHAR_BIT, 0);
 952
 953         addr[0] = RTE_IPV4(a, b, c, d);
 954         mask_len[0] = m;
 955
 956         return 0;
 957 }
 958
 959 static int
 960 parse_cb_ipv4vlan_rule(char *str, struct rte_acl_rule *v, int has_userdata)
 961 {
 962         int i, rc;
 963         char *s, *sp, *in[CB_FLD_NUM];
 964         static const char *dlm = " \t\n";
 965         int dim = has_userdata ? CB_FLD_NUM : CB_FLD_USERDATA;
 966         s = str;
 967
 968         for (i = 0; i != dim; i++, s = NULL) {
 969                 in[i] = strtok_r(s, dlm, &sp);
 970                 if (in[i] == NULL)
 971                         return -EINVAL;
 972         }
 973
 974         rc = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
 975                         &v->field[SRC_FIELD_IPV4].value.u32,
 976                         &v->field[SRC_FIELD_IPV4].mask_range.u32);
 977         if (rc != 0) {
 978                         acl_log("failed to read source address/mask: %s\n",
 979                         in[CB_FLD_SRC_ADDR]);
 980                 return rc;
 981         }
 982
 983         rc = parse_ipv4_net(in[CB_FLD_DST_ADDR],
 984                         &v->field[DST_FIELD_IPV4].value.u32,
 985                         &v->field[DST_FIELD_IPV4].mask_range.u32);
 986         if (rc != 0) {
 987                 acl_log("failed to read destination address/mask: %s\n",
 988                         in[CB_FLD_DST_ADDR]);
 989                 return rc;
 990         }
 991
 992         GET_CB_FIELD(in[CB_FLD_SRC_PORT_LOW],
 993                 v->field[SRCP_FIELD_IPV4].value.u16,
 994                 0, UINT16_MAX, 0);
 995         GET_CB_FIELD(in[CB_FLD_SRC_PORT_HIGH],
 996                 v->field[SRCP_FIELD_IPV4].mask_range.u16,
 997                 0, UINT16_MAX, 0);
 998
 999         if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
1000                         sizeof(cb_port_delim)) != 0)
1001                 return -EINVAL;
1002
1003         GET_CB_FIELD(in[CB_FLD_DST_PORT_LOW],
1004                 v->field[DSTP_FIELD_IPV4].value.u16,
1005                 0, UINT16_MAX, 0);
1006         GET_CB_FIELD(in[CB_FLD_DST_PORT_HIGH],
1007                 v->field[DSTP_FIELD_IPV4].mask_range.u16,
1008                 0, UINT16_MAX, 0);
1009
1010         if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
1011                         sizeof(cb_port_delim)) != 0)
1012                 return -EINVAL;
1013
1014         if (v->field[SRCP_FIELD_IPV4].mask_range.u16
1015                         < v->field[SRCP_FIELD_IPV4].value.u16
1016                         || v->field[DSTP_FIELD_IPV4].mask_range.u16
1017                         < v->field[DSTP_FIELD_IPV4].value.u16)
1018                 return -EINVAL;
1019
1020         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].value.u8,
1021                 0, UINT8_MAX, '/');
1022         GET_CB_FIELD(in[CB_FLD_PROTO], v->field[PROTO_FIELD_IPV4].mask_range.u8,
1023                 0, UINT8_MAX, 0);
1024
1025         if (has_userdata)
1026                 GET_CB_FIELD(in[CB_FLD_USERDATA], v->data.userdata, 0,
1027                         UINT32_MAX, 0);
1028
1029         return 0;
1030 }
1031
1032 static int
1033 add_rules(const char *rule_path,
1034                 struct rte_acl_rule **proute_base,
1035                 unsigned int *proute_num,
1036                 struct rte_acl_rule **pacl_base,
1037                 unsigned int *pacl_num, uint32_t rule_size,
1038                 int (*parser)(char *, struct rte_acl_rule*, int))
1039 {
1040         uint8_t *acl_rules, *route_rules;
1041         struct rte_acl_rule *next;
1042         unsigned int acl_num = 0, route_num = 0, total_num = 0;
1043         unsigned int acl_cnt = 0, route_cnt = 0;
1044         char buff[LINE_MAX];
1045         FILE *fh = fopen(rule_path, "rb");
1046         unsigned int i = 0;
1047         int val;
1048
1049         if (fh == NULL)
1050                 rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
1051                         rule_path);
1052
1053         while ((fgets(buff, LINE_MAX, fh) != NULL)) {
1054                 if (buff[0] == ROUTE_LEAD_CHAR)
1055                         route_num++;
1056                 else if (buff[0] == ACL_LEAD_CHAR)
1057                         acl_num++;
1058         }
1059
1060         if (0 == route_num)
1061                 rte_exit(EXIT_FAILURE, "Not find any route entries in %s!\n",
1062                                 rule_path);
1063
1064         val = fseek(fh, 0, SEEK_SET);
1065         if (val < 0) {
1066                 rte_exit(EXIT_FAILURE, "%s: File seek operation failed\n",
1067                         __func__);
1068         }
1069
1070         acl_rules = calloc(acl_num, rule_size);
1071
1072         if (NULL == acl_rules)
1073                 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1074                         __func__);
1075
1076         route_rules = calloc(route_num, rule_size);
1077
1078         if (NULL == route_rules)
1079                 rte_exit(EXIT_FAILURE, "%s: failed to malloc memory\n",
1080                         __func__);
1081
1082         i = 0;
1083         while (fgets(buff, LINE_MAX, fh) != NULL) {
1084                 i++;
1085
1086                 if (is_bypass_line(buff))
1087                         continue;
1088
1089                 char s = buff[0];
1090
1091                 /* Route entry */
1092                 if (s == ROUTE_LEAD_CHAR)
1093                         next = (struct rte_acl_rule *)(route_rules +
1094                                 route_cnt * rule_size);
1095
1096                 /* ACL entry */
1097                 else if (s == ACL_LEAD_CHAR)
1098                         next = (struct rte_acl_rule *)(acl_rules +
1099                                 acl_cnt * rule_size);
1100
1101                 /* Illegal line */
1102                 else
1103                         rte_exit(EXIT_FAILURE,
1104                                 "%s Line %u: should start with leading "
1105                                 "char %c or %c\n",
1106                                 rule_path, i, ROUTE_LEAD_CHAR, ACL_LEAD_CHAR);
1107
1108                 if (parser(buff + 1, next, s == ROUTE_LEAD_CHAR) != 0)
1109                         rte_exit(EXIT_FAILURE,
1110                                 "%s Line %u: parse rules error\n",
1111                                 rule_path, i);
1112
1113                 if (s == ROUTE_LEAD_CHAR) {
1114                         /* Check the forwarding port number */
1115                         if ((enabled_port_mask & (1 << next->data.userdata)) ==
1116                                         0)
1117                                 rte_exit(EXIT_FAILURE,
1118                                         "%s Line %u: fwd number illegal:%u\n",
1119                                         rule_path, i, next->data.userdata);
1120                         next->data.userdata += FWD_PORT_SHIFT;
1121                         route_cnt++;
1122                 } else {
1123                         next->data.userdata = ACL_DENY_SIGNATURE + acl_cnt;
1124                         acl_cnt++;
1125                 }
1126
1127                 next->data.priority = RTE_ACL_MAX_PRIORITY - total_num;
1128                 next->data.category_mask = -1;
1129                 total_num++;
1130         }
1131
1132         fclose(fh);
1133
1134         *pacl_base = (struct rte_acl_rule *)acl_rules;
1135         *pacl_num = acl_num;
1136         *proute_base = (struct rte_acl_rule *)route_rules;
1137         *proute_num = route_cnt;
1138
1139         return 0;
1140 }
1141
1142 static int
1143 usage_acl_alg(char *buf, size_t sz)
1144 {
1145         uint32_t i, n, rc, tn;
1146
1147         n = 0;
1148         tn = 0;
1149         for (i = 0; i < RTE_DIM(acl_alg); i++) {
1150                 rc = snprintf(buf + n, sz - n,
1151                         i == RTE_DIM(acl_alg) - 1 ? "%s" : "%s|",
1152                         acl_alg[i].name);
1153                 tn += rc;
1154                 if (rc < sz - n)
1155                         n += rc;
1156         }
1157
1158         return tn;
1159 }
1160
1161 static const char *
1162 str_acl_alg(enum rte_acl_classify_alg alg)
1163 {
1164         uint32_t i;
1165
1166         for (i = 0; i != RTE_DIM(acl_alg); i++) {
1167                 if (alg == acl_alg[i].alg)
1168                         return acl_alg[i].name;
1169         }
1170
1171         return "default";
1172 }
1173
1174 static enum rte_acl_classify_alg
1175 parse_acl_alg(const char *alg)
1176 {
1177         uint32_t i;
1178
1179         for (i = 0; i != RTE_DIM(acl_alg); i++) {
1180                 if (strcmp(alg, acl_alg[i].name) == 0)
1181                         return acl_alg[i].alg;
1182         }
1183
1184         return RTE_ACL_CLASSIFY_DEFAULT;
1185 }
1186
1187 static void
1188 dump_acl_config(void)
1189 {
1190         printf("ACL option are:\n");
1191         printf(OPT_RULE_IPV4": %s\n", parm_config.rule_ipv4_name);
1192         printf(OPT_RULE_IPV6": %s\n", parm_config.rule_ipv6_name);
1193         printf(OPT_ALG": %s\n", str_acl_alg(parm_config.alg));
1194 }
1195
1196 static int
1197 check_acl_config(void)
1198 {
1199         if (parm_config.rule_ipv4_name == NULL) {
1200                 acl_log("ACL IPv4 rule file not specified\n");
1201                 return -1;
1202         } else if (parm_config.rule_ipv6_name == NULL) {
1203                 acl_log("ACL IPv6 rule file not specified\n");
1204                 return -1;
1205         }
1206
1207         return 0;
1208 }
1209
1210 static struct rte_acl_ctx*
1211 setup_acl(struct rte_acl_rule *route_base,
1212                 struct rte_acl_rule *acl_base, unsigned int route_num,
1213                 unsigned int acl_num, int ipv6, int socketid)
1214 {
1215         char name[PATH_MAX];
1216         struct rte_acl_param acl_param;
1217         struct rte_acl_config acl_build_param;
1218         struct rte_acl_ctx *context;
1219         int dim = ipv6 ? RTE_DIM(ipv6_defs) : RTE_DIM(ipv4_defs);
1220
1221         /* Create ACL contexts */
1222         snprintf(name, sizeof(name), "%s%d",
1223                         ipv6 ? L3FWD_ACL_IPV6_NAME : L3FWD_ACL_IPV4_NAME,
1224                         socketid);
1225
1226         acl_param.name = name;
1227         acl_param.socket_id = socketid;
1228         acl_param.rule_size = RTE_ACL_RULE_SZ(dim);
1229         acl_param.max_rule_num = MAX_ACL_RULE_NUM;
1230
1231         if ((context = rte_acl_create(&acl_param)) == NULL)
1232                 rte_exit(EXIT_FAILURE, "Failed to create ACL context\n");
1233
1234         if (parm_config.alg != RTE_ACL_CLASSIFY_DEFAULT &&
1235                         rte_acl_set_ctx_classify(context, parm_config.alg) != 0)
1236                 rte_exit(EXIT_FAILURE,
1237                         "Failed to setup classify method for  ACL context\n");
1238
1239         if (rte_acl_add_rules(context, route_base, route_num) < 0)
1240                         rte_exit(EXIT_FAILURE, "add rules failed\n");
1241
1242         if (rte_acl_add_rules(context, acl_base, acl_num) < 0)
1243                         rte_exit(EXIT_FAILURE, "add rules failed\n");
1244
1245         /* Perform builds */
1246         memset(&acl_build_param, 0, sizeof(acl_build_param));
1247
1248         acl_build_param.num_categories = DEFAULT_MAX_CATEGORIES;
1249         acl_build_param.num_fields = dim;
1250         memcpy(&acl_build_param.defs, ipv6 ? ipv6_defs : ipv4_defs,
1251                 ipv6 ? sizeof(ipv6_defs) : sizeof(ipv4_defs));
1252
1253         if (rte_acl_build(context, &acl_build_param) != 0)
1254                 rte_exit(EXIT_FAILURE, "Failed to build ACL trie\n");
1255
1256         rte_acl_dump(context);
1257
1258         return context;
1259 }
1260
1261 static int
1262 app_acl_init(void)
1263 {
1264         unsigned lcore_id;
1265         unsigned int i;
1266         int socketid;
1267         struct rte_acl_rule *acl_base_ipv4, *route_base_ipv4,
1268                 *acl_base_ipv6, *route_base_ipv6;
1269         unsigned int acl_num_ipv4 = 0, route_num_ipv4 = 0,
1270                 acl_num_ipv6 = 0, route_num_ipv6 = 0;
1271
1272         if (check_acl_config() != 0)
1273                 rte_exit(EXIT_FAILURE, "Failed to get valid ACL options\n");
1274
1275         dump_acl_config();
1276
1277         /* Load  rules from the input file */
1278         if (add_rules(parm_config.rule_ipv4_name, &route_base_ipv4,
1279                         &route_num_ipv4, &acl_base_ipv4, &acl_num_ipv4,
1280                         sizeof(struct acl4_rule), &parse_cb_ipv4vlan_rule) < 0)
1281                 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1282
1283         acl_log("IPv4 Route entries %u:\n", route_num_ipv4);
1284         dump_ipv4_rules((struct acl4_rule *)route_base_ipv4, route_num_ipv4, 1);
1285
1286         acl_log("IPv4 ACL entries %u:\n", acl_num_ipv4);
1287         dump_ipv4_rules((struct acl4_rule *)acl_base_ipv4, acl_num_ipv4, 1);
1288
1289         if (add_rules(parm_config.rule_ipv6_name, &route_base_ipv6,
1290                         &route_num_ipv6,
1291                         &acl_base_ipv6, &acl_num_ipv6,
1292                         sizeof(struct acl6_rule), &parse_cb_ipv6_rule) < 0)
1293                 rte_exit(EXIT_FAILURE, "Failed to add rules\n");
1294
1295         acl_log("IPv6 Route entries %u:\n", route_num_ipv6);
1296         dump_ipv6_rules((struct acl6_rule *)route_base_ipv6, route_num_ipv6, 1);
1297
1298         acl_log("IPv6 ACL entries %u:\n", acl_num_ipv6);
1299         dump_ipv6_rules((struct acl6_rule *)acl_base_ipv6, acl_num_ipv6, 1);
1300
1301         memset(&acl_config, 0, sizeof(acl_config));
1302
1303         /* Check sockets a context should be created on */
1304         if (!numa_on)
1305                 acl_config.mapped[0] = 1;
1306         else {
1307                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1308                         if (rte_lcore_is_enabled(lcore_id) == 0)
1309                                 continue;
1310
1311                         socketid = rte_lcore_to_socket_id(lcore_id);
1312                         if (socketid >= NB_SOCKETS) {
1313                                 acl_log("Socket %d of lcore %u is out "
1314                                         "of range %d\n",
1315                                         socketid, lcore_id, NB_SOCKETS);
1316                                 free(route_base_ipv4);
1317                                 free(route_base_ipv6);
1318                                 free(acl_base_ipv4);
1319                                 free(acl_base_ipv6);
1320                                 return -1;
1321                         }
1322
1323                         acl_config.mapped[socketid] = 1;
1324                 }
1325         }
1326
1327         for (i = 0; i < NB_SOCKETS; i++) {
1328                 if (acl_config.mapped[i]) {
1329                         acl_config.acx_ipv4[i] = setup_acl(route_base_ipv4,
1330                                 acl_base_ipv4, route_num_ipv4, acl_num_ipv4,
1331                                 0, i);
1332
1333                         acl_config.acx_ipv6[i] = setup_acl(route_base_ipv6,
1334                                 acl_base_ipv6, route_num_ipv6, acl_num_ipv6,
1335                                 1, i);
1336                 }
1337         }
1338
1339         free(route_base_ipv4);
1340         free(route_base_ipv6);
1341
1342 #ifdef L3FWDACL_DEBUG
1343         acl_config.rule_ipv4 = (struct acl4_rule *)acl_base_ipv4;
1344         acl_config.rule_ipv6 = (struct acl6_rule *)acl_base_ipv6;
1345 #else
1346         free(acl_base_ipv4);
1347         free(acl_base_ipv6);
1348 #endif
1349
1350         return 0;
1351 }
1352
1353 /***********************end of ACL part******************************/
1354
1355 struct lcore_conf {
1356         uint16_t n_rx_queue;
1357         struct lcore_rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
1358         uint16_t n_tx_port;
1359         uint16_t tx_port_id[RTE_MAX_ETHPORTS];
1360         uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
1361         struct rte_eth_dev_tx_buffer *tx_buffer[RTE_MAX_ETHPORTS];
1362 } __rte_cache_aligned;
1363
1364 static struct lcore_conf lcore_conf[RTE_MAX_LCORE];
1365
1366 /* Enqueue a single packet, and send burst if queue is filled */
1367 static inline void
1368 send_single_packet(struct rte_mbuf *m, uint16_t port)
1369 {
1370         uint32_t lcore_id;
1371         struct lcore_conf *qconf;
1372         struct rte_ether_hdr *eh;
1373
1374         lcore_id = rte_lcore_id();
1375
1376         /* update src and dst mac*/
1377         eh = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
1378         memcpy(eh, &port_l2hdr[port],
1379                         sizeof(eh->dst_addr) + sizeof(eh->src_addr));
1380
1381         qconf = &lcore_conf[lcore_id];
1382         rte_eth_tx_buffer(port, qconf->tx_queue_id[port],
1383                         qconf->tx_buffer[port], m);
1384 }
1385
1386 #ifdef DO_RFC_1812_CHECKS
1387 static inline int
1388 is_valid_ipv4_pkt(struct rte_ipv4_hdr *pkt, uint32_t link_len)
1389 {
1390         /* From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2 */
1391         /*
1392          * 1. The packet length reported by the Link Layer must be large
1393          * enough to hold the minimum length legal IP datagram (20 bytes).
1394          */
1395         if (link_len < sizeof(struct rte_ipv4_hdr))
1396                 return -1;
1397
1398         /* 2. The IP checksum must be correct. */
1399         /* this is checked in H/W */
1400
1401         /*
1402          * 3. The IP version number must be 4. If the version number is not 4
1403          * then the packet may be another version of IP, such as IPng or
1404          * ST-II.
1405          */
1406         if (((pkt->version_ihl) >> 4) != 4)
1407                 return -3;
1408         /*
1409          * 4. The IP header length field must be large enough to hold the
1410          * minimum length legal IP datagram (20 bytes = 5 words).
1411          */
1412         if ((pkt->version_ihl & 0xf) < 5)
1413                 return -4;
1414
1415         /*
1416          * 5. The IP total length field must be large enough to hold the IP
1417          * datagram header, whose length is specified in the IP header length
1418          * field.
1419          */
1420         if (rte_cpu_to_be_16(pkt->total_length) < sizeof(struct rte_ipv4_hdr))
1421                 return -5;
1422
1423         return 0;
1424 }
1425 #endif
1426
1427 /* main processing loop */
1428 static int
1429 main_loop(__rte_unused void *dummy)
1430 {
1431         struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
1432         unsigned lcore_id;
1433         uint64_t prev_tsc, diff_tsc, cur_tsc;
1434         int i, nb_rx;
1435         uint16_t portid;
1436         uint8_t queueid;
1437         struct lcore_conf *qconf;
1438         int socketid;
1439         const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
1440                         / US_PER_S * BURST_TX_DRAIN_US;
1441
1442         prev_tsc = 0;
1443         lcore_id = rte_lcore_id();
1444         qconf = &lcore_conf[lcore_id];
1445         socketid = rte_lcore_to_socket_id(lcore_id);
1446
1447         if (qconf->n_rx_queue == 0) {
1448                 RTE_LOG(INFO, L3FWD, "lcore %u has nothing to do\n", lcore_id);
1449                 return 0;
1450         }
1451
1452         RTE_LOG(INFO, L3FWD, "entering main loop on lcore %u\n", lcore_id);
1453
1454         for (i = 0; i < qconf->n_rx_queue; i++) {
1455
1456                 portid = qconf->rx_queue_list[i].port_id;
1457                 queueid = qconf->rx_queue_list[i].queue_id;
1458                 RTE_LOG(INFO, L3FWD,
1459                         " -- lcoreid=%u portid=%u rxqueueid=%hhu\n",
1460                         lcore_id, portid, queueid);
1461         }
1462
1463         while (1) {
1464
1465                 cur_tsc = rte_rdtsc();
1466
1467                 /*
1468                  * TX burst queue drain
1469                  */
1470                 diff_tsc = cur_tsc - prev_tsc;
1471                 if (unlikely(diff_tsc > drain_tsc)) {
1472                         for (i = 0; i < qconf->n_tx_port; ++i) {
1473                                 portid = qconf->tx_port_id[i];
1474                                 rte_eth_tx_buffer_flush(portid,
1475                                                 qconf->tx_queue_id[portid],
1476                                                 qconf->tx_buffer[portid]);
1477                         }
1478                         prev_tsc = cur_tsc;
1479                 }
1480
1481                 /*
1482                  * Read packet from RX queues
1483                  */
1484                 for (i = 0; i < qconf->n_rx_queue; ++i) {
1485
1486                         portid = qconf->rx_queue_list[i].port_id;
1487                         queueid = qconf->rx_queue_list[i].queue_id;
1488                         nb_rx = rte_eth_rx_burst(portid, queueid,
1489                                 pkts_burst, MAX_PKT_BURST);
1490
1491                         if (nb_rx > 0) {
1492                                 struct acl_search_t acl_search;
1493
1494                                 prepare_acl_parameter(pkts_burst, &acl_search,
1495                                         nb_rx);
1496
1497                                 if (acl_search.num_ipv4) {
1498                                         rte_acl_classify(
1499                                                 acl_config.acx_ipv4[socketid],
1500                                                 acl_search.data_ipv4,
1501                                                 acl_search.res_ipv4,
1502                                                 acl_search.num_ipv4,
1503                                                 DEFAULT_MAX_CATEGORIES);
1504
1505                                         send_packets(acl_search.m_ipv4,
1506                                                 acl_search.res_ipv4,
1507                                                 acl_search.num_ipv4);
1508                                 }
1509
1510                                 if (acl_search.num_ipv6) {
1511                                         rte_acl_classify(
1512                                                 acl_config.acx_ipv6[socketid],
1513                                                 acl_search.data_ipv6,
1514                                                 acl_search.res_ipv6,
1515                                                 acl_search.num_ipv6,
1516                                                 DEFAULT_MAX_CATEGORIES);
1517
1518                                         send_packets(acl_search.m_ipv6,
1519                                                 acl_search.res_ipv6,
1520                                                 acl_search.num_ipv6);
1521                                 }
1522                         }
1523                 }
1524         }
1525 }
1526
1527 static int
1528 check_lcore_params(void)
1529 {
1530         uint8_t queue, lcore;
1531         uint16_t i;
1532         int socketid;
1533
1534         for (i = 0; i < nb_lcore_params; ++i) {
1535                 queue = lcore_params[i].queue_id;
1536                 if (queue >= MAX_RX_QUEUE_PER_PORT) {
1537                         printf("invalid queue number: %hhu\n", queue);
1538                         return -1;
1539                 }
1540                 lcore = lcore_params[i].lcore_id;
1541                 if (!rte_lcore_is_enabled(lcore)) {
1542                         printf("error: lcore %hhu is not enabled in "
1543                                 "lcore mask\n", lcore);
1544                         return -1;
1545                 }
1546                 socketid = rte_lcore_to_socket_id(lcore);
1547                 if (socketid != 0 && numa_on == 0) {
1548                         printf("warning: lcore %hhu is on socket %d "
1549                                 "with numa off\n",
1550                                 lcore, socketid);
1551                 }
1552         }
1553         return 0;
1554 }
1555
1556 static int
1557 check_port_config(void)
1558 {
1559         unsigned portid;
1560         uint16_t i;
1561
1562         for (i = 0; i < nb_lcore_params; ++i) {
1563                 portid = lcore_params[i].port_id;
1564
1565                 if ((enabled_port_mask & (1 << portid)) == 0) {
1566                         printf("port %u is not enabled in port mask\n", portid);
1567                         return -1;
1568                 }
1569                 if (!rte_eth_dev_is_valid_port(portid)) {
1570                         printf("port %u is not present on the board\n", portid);
1571                         return -1;
1572                 }
1573         }
1574         return 0;
1575 }
1576
1577 static uint8_t
1578 get_port_n_rx_queues(const uint16_t port)
1579 {
1580         int queue = -1;
1581         uint16_t i;
1582
1583         for (i = 0; i < nb_lcore_params; ++i) {
1584                 if (lcore_params[i].port_id == port &&
1585                                 lcore_params[i].queue_id > queue)
1586                         queue = lcore_params[i].queue_id;
1587         }
1588         return (uint8_t)(++queue);
1589 }
1590
1591 static int
1592 init_lcore_rx_queues(void)
1593 {
1594         uint16_t i, nb_rx_queue;
1595         uint8_t lcore;
1596
1597         for (i = 0; i < nb_lcore_params; ++i) {
1598                 lcore = lcore_params[i].lcore_id;
1599                 nb_rx_queue = lcore_conf[lcore].n_rx_queue;
1600                 if (nb_rx_queue >= MAX_RX_QUEUE_PER_LCORE) {
1601                         printf("error: too many queues (%u) for lcore: %u\n",
1602                                 (unsigned)nb_rx_queue + 1, (unsigned)lcore);
1603                         return -1;
1604                 } else {
1605                         lcore_conf[lcore].rx_queue_list[nb_rx_queue].port_id =
1606                                 lcore_params[i].port_id;
1607                         lcore_conf[lcore].rx_queue_list[nb_rx_queue].queue_id =
1608                                 lcore_params[i].queue_id;
1609                         lcore_conf[lcore].n_rx_queue++;
1610                 }
1611         }
1612         return 0;
1613 }
1614
1615 /* display usage */
1616 static void
1617 print_usage(const char *prgname)
1618 {
1619         char alg[PATH_MAX];
1620
1621         usage_acl_alg(alg, sizeof(alg));
1622         printf("%s [EAL options] -- -p PORTMASK -P"
1623                 "  --"OPT_RULE_IPV4"=FILE"
1624                 "  --"OPT_RULE_IPV6"=FILE"
1625                 "  [--"OPT_CONFIG" (port,queue,lcore)[,(port,queue,lcore]]"
1626                 "  [--"OPT_MAX_PKT_LEN" PKTLEN]\n"
1627                 "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
1628                 "  -P: enable promiscuous mode\n"
1629                 "  --"OPT_CONFIG" (port,queue,lcore): rx queues configuration\n"
1630                 "  --"OPT_NONUMA": optional, disable numa awareness\n"
1631                 "  --"OPT_MAX_PKT_LEN" PKTLEN: maximum packet length in decimal (64-9600)\n"
1632                 "  --"OPT_RULE_IPV4"=FILE: specify the ipv4 rules entries file. "
1633                 "Each rule occupy one line. "
1634                 "2 kinds of rules are supported. "
1635                 "One is ACL entry at while line leads with character '%c', "
1636                 "another is route entry at while line leads with character '%c'.\n"
1637                 "  --"OPT_RULE_IPV6"=FILE: specify the ipv6 rules entries file.\n"
1638                 "  --"OPT_ALG": ACL classify method to use, one of: %s\n",
1639                 prgname, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR, alg);
1640 }
1641
1642 static int
1643 parse_max_pkt_len(const char *pktlen)
1644 {
1645         char *end = NULL;
1646         unsigned long len;
1647
1648         /* parse decimal string */
1649         len = strtoul(pktlen, &end, 10);
1650         if ((pktlen[0] == '\0') || (end == NULL) || (*end != '\0'))
1651                 return -1;
1652
1653         if (len == 0)
1654                 return -1;
1655
1656         return len;
1657 }
1658
1659 static int
1660 parse_portmask(const char *portmask)
1661 {
1662         char *end = NULL;
1663         unsigned long pm;
1664
1665         /* parse hexadecimal string */
1666         pm = strtoul(portmask, &end, 16);
1667         if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
1668                 return 0;
1669
1670         return pm;
1671 }
1672
1673 static int
1674 parse_config(const char *q_arg)
1675 {
1676         char s[256];
1677         const char *p, *p0 = q_arg;
1678         char *end;
1679         enum fieldnames {
1680                 FLD_PORT = 0,
1681                 FLD_QUEUE,
1682                 FLD_LCORE,
1683                 _NUM_FLD
1684         };
1685         unsigned long int_fld[_NUM_FLD];
1686         char *str_fld[_NUM_FLD];
1687         int i;
1688         unsigned size;
1689
1690         nb_lcore_params = 0;
1691
1692         while ((p = strchr(p0, '(')) != NULL) {
1693                 ++p;
1694                 if ((p0 = strchr(p, ')')) == NULL)
1695                         return -1;
1696
1697                 size = p0 - p;
1698                 if (size >= sizeof(s))
1699                         return -1;
1700
1701                 snprintf(s, sizeof(s), "%.*s", size, p);
1702                 if (rte_strsplit(s, sizeof(s), str_fld, _NUM_FLD, ',') !=
1703                                 _NUM_FLD)
1704                         return -1;
1705                 for (i = 0; i < _NUM_FLD; i++) {
1706                         errno = 0;
1707                         int_fld[i] = strtoul(str_fld[i], &end, 0);
1708                         if (errno != 0 || end == str_fld[i] || int_fld[i] > 255)
1709                                 return -1;
1710                 }
1711                 if (nb_lcore_params >= MAX_LCORE_PARAMS) {
1712                         printf("exceeded max number of lcore params: %hu\n",
1713                                 nb_lcore_params);
1714                         return -1;
1715                 }
1716                 lcore_params_array[nb_lcore_params].port_id =
1717                         (uint8_t)int_fld[FLD_PORT];
1718                 lcore_params_array[nb_lcore_params].queue_id =
1719                         (uint8_t)int_fld[FLD_QUEUE];
1720                 lcore_params_array[nb_lcore_params].lcore_id =
1721                         (uint8_t)int_fld[FLD_LCORE];
1722                 ++nb_lcore_params;
1723         }
1724         lcore_params = lcore_params_array;
1725         return 0;
1726 }
1727
1728 static const char *
1729 parse_eth_dest(const char *optarg)
1730 {
1731         unsigned long portid;
1732         char *port_end;
1733
1734         errno = 0;
1735         portid = strtoul(optarg, &port_end, 0);
1736         if (errno != 0 || port_end == optarg || *port_end++ != ',')
1737                 return "Invalid format";
1738         else if (portid >= RTE_MAX_ETHPORTS)
1739                 return "port value exceeds RTE_MAX_ETHPORTS("
1740                         RTE_STR(RTE_MAX_ETHPORTS) ")";
1741
1742         if (cmdline_parse_etheraddr(NULL, port_end,
1743                         &port_l2hdr[portid].dst_addr,
1744                         sizeof(port_l2hdr[portid].dst_addr)) < 0)
1745                 return "Invalid ethernet address";
1746         return NULL;
1747 }
1748
1749 /* Parse the argument given in the command line of the application */
1750 static int
1751 parse_args(int argc, char **argv)
1752 {
1753         int opt, ret;
1754         char **argvopt;
1755         int option_index;
1756         char *prgname = argv[0];
1757         static struct option lgopts[] = {
1758                 {OPT_CONFIG,      1, NULL, OPT_CONFIG_NUM      },
1759                 {OPT_NONUMA,      0, NULL, OPT_NONUMA_NUM      },
1760                 {OPT_MAX_PKT_LEN, 1, NULL, OPT_MAX_PKT_LEN_NUM },
1761                 {OPT_RULE_IPV4,   1, NULL, OPT_RULE_IPV4_NUM   },
1762                 {OPT_RULE_IPV6,   1, NULL, OPT_RULE_IPV6_NUM   },
1763                 {OPT_ALG,         1, NULL, OPT_ALG_NUM         },
1764                 {OPT_ETH_DEST,    1, NULL, OPT_ETH_DEST_NUM    },
1765                 {NULL,            0, 0,    0                   }
1766         };
1767
1768         argvopt = argv;
1769
1770         while ((opt = getopt_long(argc, argvopt, "p:P",
1771                                 lgopts, &option_index)) != EOF) {
1772
1773                 switch (opt) {
1774                 /* portmask */
1775                 case 'p':
1776                         enabled_port_mask = parse_portmask(optarg);
1777                         if (enabled_port_mask == 0) {
1778                                 printf("invalid portmask\n");
1779                                 print_usage(prgname);
1780                                 return -1;
1781                         }
1782                         break;
1783
1784                 case 'P':
1785                         printf("Promiscuous mode selected\n");
1786                         promiscuous_on = 1;
1787                         break;
1788
1789                 /* long options */
1790                 case OPT_CONFIG_NUM:
1791                         ret = parse_config(optarg);
1792                         if (ret) {
1793                                 printf("invalid config\n");
1794                                 print_usage(prgname);
1795                                 return -1;
1796                         }
1797                         break;
1798
1799                 case OPT_NONUMA_NUM:
1800                         printf("numa is disabled\n");
1801                         numa_on = 0;
1802                         break;
1803
1804                 case OPT_MAX_PKT_LEN_NUM:
1805                         printf("Custom frame size is configured\n");
1806                         max_pkt_len = parse_max_pkt_len(optarg);
1807                         break;
1808
1809                 case OPT_RULE_IPV4_NUM:
1810                         parm_config.rule_ipv4_name = optarg;
1811                         break;
1812
1813                 case OPT_RULE_IPV6_NUM:
1814                         parm_config.rule_ipv6_name = optarg;
1815                         break;
1816
1817                 case OPT_ALG_NUM:
1818                         parm_config.alg = parse_acl_alg(optarg);
1819                         if (parm_config.alg ==
1820                                         RTE_ACL_CLASSIFY_DEFAULT) {
1821                                 printf("unknown %s value:\"%s\"\n",
1822                                         OPT_ALG, optarg);
1823                                 print_usage(prgname);
1824                                 return -1;
1825                         }
1826                         break;
1827
1828                 case OPT_ETH_DEST_NUM:
1829                 {
1830                         const char *serr = parse_eth_dest(optarg);
1831                         if (serr != NULL) {
1832                                 printf("invalid %s value:\"%s\": %s\n",
1833                                         OPT_ETH_DEST, optarg, serr);
1834                                 print_usage(prgname);
1835                                 return -1;
1836                         }
1837                         break;
1838                 }
1839                 default:
1840                         print_usage(prgname);
1841                         return -1;
1842                 }
1843         }
1844
1845         if (optind >= 0)
1846                 argv[optind-1] = prgname;
1847
1848         ret = optind-1;
1849         optind = 1; /* reset getopt lib */
1850         return ret;
1851 }
1852
1853 static void
1854 print_ethaddr(const char *name, const struct rte_ether_addr *eth_addr)
1855 {
1856         char buf[RTE_ETHER_ADDR_FMT_SIZE];
1857         rte_ether_format_addr(buf, RTE_ETHER_ADDR_FMT_SIZE, eth_addr);
1858         printf("%s%s", name, buf);
1859 }
1860
1861 static int
1862 init_mem(unsigned nb_mbuf)
1863 {
1864         int socketid;
1865         unsigned lcore_id;
1866         char s[64];
1867
1868         for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1869                 if (rte_lcore_is_enabled(lcore_id) == 0)
1870                         continue;
1871
1872                 if (numa_on)
1873                         socketid = rte_lcore_to_socket_id(lcore_id);
1874                 else
1875                         socketid = 0;
1876
1877                 if (socketid >= NB_SOCKETS) {
1878                         rte_exit(EXIT_FAILURE,
1879                                 "Socket %d of lcore %u is out of range %d\n",
1880                                 socketid, lcore_id, NB_SOCKETS);
1881                 }
1882                 if (pktmbuf_pool[socketid] == NULL) {
1883                         snprintf(s, sizeof(s), "mbuf_pool_%d", socketid);
1884                         pktmbuf_pool[socketid] =
1885                                 rte_pktmbuf_pool_create(s, nb_mbuf,
1886                                         MEMPOOL_CACHE_SIZE, 0,
1887                                         RTE_MBUF_DEFAULT_BUF_SIZE,
1888                                         socketid);
1889                         if (pktmbuf_pool[socketid] == NULL)
1890                                 rte_exit(EXIT_FAILURE,
1891                                         "Cannot init mbuf pool on socket %d\n",
1892                                         socketid);
1893                         else
1894                                 printf("Allocated mbuf pool on socket %d\n",
1895                                         socketid);
1896                 }
1897         }
1898         return 0;
1899 }
1900
1901 /* Check the link status of all ports in up to 9s, and print them finally */
1902 static void
1903 check_all_ports_link_status(uint32_t port_mask)
1904 {
1905 #define CHECK_INTERVAL 100 /* 100ms */
1906 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
1907         uint16_t portid;
1908         uint8_t count, all_ports_up, print_flag = 0;
1909         struct rte_eth_link link;
1910         int ret;
1911         char link_status_text[RTE_ETH_LINK_MAX_STR_LEN];
1912
1913         printf("\nChecking link status");
1914         fflush(stdout);
1915         for (count = 0; count <= MAX_CHECK_TIME; count++) {
1916                 all_ports_up = 1;
1917                 RTE_ETH_FOREACH_DEV(portid) {
1918                         if ((port_mask & (1 << portid)) == 0)
1919                                 continue;
1920                         memset(&link, 0, sizeof(link));
1921                         ret = rte_eth_link_get_nowait(portid, &link);
1922                         if (ret < 0) {
1923                                 all_ports_up = 0;
1924                                 if (print_flag == 1)
1925                                         printf("Port %u link get failed: %s\n",
1926                                                 portid, rte_strerror(-ret));
1927                                 continue;
1928                         }
1929                         /* print link status if flag set */
1930                         if (print_flag == 1) {
1931                                 rte_eth_link_to_str(link_status_text,
1932                                         sizeof(link_status_text), &link);
1933                                 printf("Port %d %s\n", portid,
1934                                        link_status_text);
1935                                 continue;
1936                         }
1937                         /* clear all_ports_up flag if any link down */
1938                         if (link.link_status == RTE_ETH_LINK_DOWN) {
1939                                 all_ports_up = 0;
1940                                 break;
1941                         }
1942                 }
1943                 /* after finally printing all link status, get out */
1944                 if (print_flag == 1)
1945                         break;
1946
1947                 if (all_ports_up == 0) {
1948                         printf(".");
1949                         fflush(stdout);
1950                         rte_delay_ms(CHECK_INTERVAL);
1951                 }
1952
1953                 /* set the print_flag if all ports up or timeout */
1954                 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
1955                         print_flag = 1;
1956                         printf("done\n");
1957                 }
1958         }
1959 }
1960
1961 /*
1962  * build-up default vaues for dest MACs.
1963  */
1964 static void
1965 set_default_dest_mac(void)
1966 {
1967         uint32_t i;
1968
1969         for (i = 0; i != RTE_DIM(port_l2hdr); i++) {
1970                 port_l2hdr[i].dst_addr.addr_bytes[0] =
1971                                 RTE_ETHER_LOCAL_ADMIN_ADDR;
1972                 port_l2hdr[i].dst_addr.addr_bytes[5] = i;
1973         }
1974 }
1975
1976 static uint32_t
1977 eth_dev_get_overhead_len(uint32_t max_rx_pktlen, uint16_t max_mtu)
1978 {
1979         uint32_t overhead_len;
1980
1981         if (max_mtu != UINT16_MAX && max_rx_pktlen > max_mtu)
1982                 overhead_len = max_rx_pktlen - max_mtu;
1983         else
1984                 overhead_len = RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN;
1985
1986         return overhead_len;
1987 }
1988
1989 static int
1990 config_port_max_pkt_len(struct rte_eth_conf *conf,
1991                 struct rte_eth_dev_info *dev_info)
1992 {
1993         uint32_t overhead_len;
1994
1995         if (max_pkt_len == 0)
1996                 return 0;
1997
1998         if (max_pkt_len < RTE_ETHER_MIN_LEN || max_pkt_len > MAX_JUMBO_PKT_LEN)
1999                 return -1;
2000
2001         overhead_len = eth_dev_get_overhead_len(dev_info->max_rx_pktlen,
2002                         dev_info->max_mtu);
2003         conf->rxmode.mtu = max_pkt_len - overhead_len;
2004
2005         if (conf->rxmode.mtu > RTE_ETHER_MTU)
2006                 conf->txmode.offloads |= RTE_ETH_TX_OFFLOAD_MULTI_SEGS;
2007
2008         return 0;
2009 }
2010
2011 int
2012 main(int argc, char **argv)
2013 {
2014         struct lcore_conf *qconf;
2015         struct rte_eth_dev_info dev_info;
2016         struct rte_eth_txconf *txconf;
2017         int ret;
2018         unsigned nb_ports;
2019         uint16_t queueid;
2020         unsigned lcore_id;
2021         uint32_t n_tx_queue, nb_lcores;
2022         uint16_t portid;
2023         uint8_t nb_rx_queue, queue, socketid;
2024
2025         /* init EAL */
2026         ret = rte_eal_init(argc, argv);
2027         if (ret < 0)
2028                 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
2029         argc -= ret;
2030         argv += ret;
2031
2032         set_default_dest_mac();
2033
2034         /* parse application arguments (after the EAL ones) */
2035         ret = parse_args(argc, argv);
2036         if (ret < 0)
2037                 rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
2038
2039         if (check_lcore_params() < 0)
2040                 rte_exit(EXIT_FAILURE, "check_lcore_params failed\n");
2041
2042         ret = init_lcore_rx_queues();
2043         if (ret < 0)
2044                 rte_exit(EXIT_FAILURE, "init_lcore_rx_queues failed\n");
2045
2046         nb_ports = rte_eth_dev_count_avail();
2047
2048         if (check_port_config() < 0)
2049                 rte_exit(EXIT_FAILURE, "check_port_config failed\n");
2050
2051         /* Add ACL rules and route entries, build trie */
2052         if (app_acl_init() < 0)
2053                 rte_exit(EXIT_FAILURE, "app_acl_init failed\n");
2054
2055         nb_lcores = rte_lcore_count();
2056
2057         /* initialize all ports */
2058         RTE_ETH_FOREACH_DEV(portid) {
2059                 struct rte_eth_conf local_port_conf = port_conf;
2060
2061                 /* skip ports that are not enabled */
2062                 if ((enabled_port_mask & (1 << portid)) == 0) {
2063                         printf("\nSkipping disabled port %d\n", portid);
2064                         continue;
2065                 }
2066
2067                 /* init port */
2068                 printf("Initializing port %d ... ", portid);
2069                 fflush(stdout);
2070
2071                 nb_rx_queue = get_port_n_rx_queues(portid);
2072                 n_tx_queue = nb_lcores;
2073                 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
2074                         n_tx_queue = MAX_TX_QUEUE_PER_PORT;
2075                 printf("Creating queues: nb_rxq=%d nb_txq=%u... ",
2076                         nb_rx_queue, (unsigned)n_tx_queue);
2077
2078                 ret = rte_eth_dev_info_get(portid, &dev_info);
2079                 if (ret != 0)
2080                         rte_exit(EXIT_FAILURE,
2081                                 "Error during getting device (port %u) info: %s\n",
2082                                 portid, strerror(-ret));
2083
2084                 ret = config_port_max_pkt_len(&local_port_conf, &dev_info);
2085                 if (ret != 0)
2086                         rte_exit(EXIT_FAILURE,
2087                                 "Invalid max packet length: %u (port %u)\n",
2088                                 max_pkt_len, portid);
2089
2090                 if (dev_info.tx_offload_capa & RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE)
2091                         local_port_conf.txmode.offloads |=
2092                                 RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
2093
2094                 local_port_conf.rx_adv_conf.rss_conf.rss_hf &=
2095                         dev_info.flow_type_rss_offloads;
2096                 if (local_port_conf.rx_adv_conf.rss_conf.rss_hf !=
2097                                 port_conf.rx_adv_conf.rss_conf.rss_hf) {
2098                         printf("Port %u modified RSS hash function based on hardware support,"
2099                                 "requested:%#"PRIx64" configured:%#"PRIx64"\n",
2100                                 portid,
2101                                 port_conf.rx_adv_conf.rss_conf.rss_hf,
2102                                 local_port_conf.rx_adv_conf.rss_conf.rss_hf);
2103                 }
2104
2105                 ret = rte_eth_dev_configure(portid, nb_rx_queue,
2106                                         (uint16_t)n_tx_queue, &local_port_conf);
2107                 if (ret < 0)
2108                         rte_exit(EXIT_FAILURE,
2109                                 "Cannot configure device: err=%d, port=%d\n",
2110                                 ret, portid);
2111
2112                 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd,
2113                                                        &nb_txd);
2114                 if (ret < 0)
2115                         rte_exit(EXIT_FAILURE,
2116                                 "rte_eth_dev_adjust_nb_rx_tx_desc: err=%d, port=%d\n",
2117                                 ret, portid);
2118
2119                 ret = rte_eth_macaddr_get(portid, &port_l2hdr[portid].src_addr);
2120                 if (ret < 0)
2121                         rte_exit(EXIT_FAILURE,
2122                                 "rte_eth_macaddr_get: err=%d, port=%d\n",
2123                                 ret, portid);
2124
2125                 print_ethaddr("Dst MAC:", &port_l2hdr[portid].dst_addr);
2126                 print_ethaddr(", Src MAC:", &port_l2hdr[portid].src_addr);
2127                 printf(", ");
2128
2129                 /* init memory */
2130                 ret = init_mem(NB_MBUF);
2131                 if (ret < 0)
2132                         rte_exit(EXIT_FAILURE, "init_mem failed\n");
2133
2134                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2135                         if (rte_lcore_is_enabled(lcore_id) == 0)
2136                                 continue;
2137
2138                         /* Initialize TX buffers */
2139                         qconf = &lcore_conf[lcore_id];
2140                         qconf->tx_buffer[portid] = rte_zmalloc_socket("tx_buffer",
2141                                         RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0,
2142                                         rte_eth_dev_socket_id(portid));
2143                         if (qconf->tx_buffer[portid] == NULL)
2144                                 rte_exit(EXIT_FAILURE, "Can't allocate tx buffer for port %u\n",
2145                                                 (unsigned) portid);
2146
2147                         rte_eth_tx_buffer_init(qconf->tx_buffer[portid], MAX_PKT_BURST);
2148                 }
2149
2150                 /* init one TX queue per couple (lcore,port) */
2151                 queueid = 0;
2152                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2153                         if (rte_lcore_is_enabled(lcore_id) == 0)
2154                                 continue;
2155
2156                         if (numa_on)
2157                                 socketid = (uint8_t)
2158                                         rte_lcore_to_socket_id(lcore_id);
2159                         else
2160                                 socketid = 0;
2161
2162                         printf("txq=%u,%d,%d ", lcore_id, queueid, socketid);
2163                         fflush(stdout);
2164
2165                         ret = rte_eth_dev_info_get(portid, &dev_info);
2166                         if (ret != 0)
2167                                 rte_exit(EXIT_FAILURE,
2168                                         "Error during getting device (port %u) info: %s\n",
2169                                         portid, strerror(-ret));
2170
2171                         txconf = &dev_info.default_txconf;
2172                         txconf->offloads = local_port_conf.txmode.offloads;
2173                         ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
2174                                                      socketid, txconf);
2175                         if (ret < 0)
2176                                 rte_exit(EXIT_FAILURE,
2177                                         "rte_eth_tx_queue_setup: err=%d, "
2178                                         "port=%d\n", ret, portid);
2179
2180                         qconf = &lcore_conf[lcore_id];
2181                         qconf->tx_queue_id[portid] = queueid;
2182                         queueid++;
2183
2184                         qconf->tx_port_id[qconf->n_tx_port] = portid;
2185                         qconf->n_tx_port++;
2186                 }
2187                 printf("\n");
2188         }
2189
2190         for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
2191                 if (rte_lcore_is_enabled(lcore_id) == 0)
2192                         continue;
2193                 qconf = &lcore_conf[lcore_id];
2194                 printf("\nInitializing rx queues on lcore %u ... ", lcore_id);
2195                 fflush(stdout);
2196                 /* init RX queues */
2197                 for (queue = 0; queue < qconf->n_rx_queue; ++queue) {
2198                         struct rte_eth_rxconf rxq_conf;
2199
2200                         portid = qconf->rx_queue_list[queue].port_id;
2201                         queueid = qconf->rx_queue_list[queue].queue_id;
2202
2203                         if (numa_on)
2204                                 socketid = (uint8_t)
2205                                         rte_lcore_to_socket_id(lcore_id);
2206                         else
2207                                 socketid = 0;
2208
2209                         printf("rxq=%d,%d,%d ", portid, queueid, socketid);
2210                         fflush(stdout);
2211
2212                         ret = rte_eth_dev_info_get(portid, &dev_info);
2213                         if (ret != 0)
2214                                 rte_exit(EXIT_FAILURE,
2215                                         "Error during getting device (port %u) info: %s\n",
2216                                         portid, strerror(-ret));
2217
2218                         rxq_conf = dev_info.default_rxconf;
2219                         rxq_conf.offloads = port_conf.rxmode.offloads;
2220                         ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd,
2221                                         socketid, &rxq_conf,
2222                                         pktmbuf_pool[socketid]);
2223                         if (ret < 0)
2224                                 rte_exit(EXIT_FAILURE,
2225                                         "rte_eth_rx_queue_setup: err=%d,"
2226                                         "port=%d\n", ret, portid);
2227                 }
2228         }
2229
2230         printf("\n");
2231
2232         /* start ports */
2233         RTE_ETH_FOREACH_DEV(portid) {
2234                 if ((enabled_port_mask & (1 << portid)) == 0)
2235                         continue;
2236
2237                 /* Start device */
2238                 ret = rte_eth_dev_start(portid);
2239                 if (ret < 0)
2240                         rte_exit(EXIT_FAILURE,
2241                                 "rte_eth_dev_start: err=%d, port=%d\n",
2242                                 ret, portid);
2243
2244                 /*
2245                  * If enabled, put device in promiscuous mode.
2246                  * This allows IO forwarding mode to forward packets
2247                  * to itself through 2 cross-connected  ports of the
2248                  * target machine.
2249                  */
2250                 if (promiscuous_on) {
2251                         ret = rte_eth_promiscuous_enable(portid);
2252                         if (ret != 0)
2253                                 rte_exit(EXIT_FAILURE,
2254                                         "rte_eth_promiscuous_enable: err=%s, port=%u\n",
2255                                         rte_strerror(-ret), portid);
2256                 }
2257         }
2258
2259         check_all_ports_link_status(enabled_port_mask);
2260
2261         /* launch per-lcore init on every lcore */
2262         rte_eal_mp_remote_launch(main_loop, NULL, CALL_MAIN);
2263         RTE_LCORE_FOREACH_WORKER(lcore_id) {
2264                 if (rte_eal_wait_lcore(lcore_id) < 0)
2265                         return -1;
2266         }
2267
2268         /* clean up the EAL */
2269         rte_eal_cleanup();
2270
2271         return 0;
2272 }