examples/ipsec-secgw: make number of buffers dynamic
[dpdk.git] / examples / ipsec-secgw / ipsec-secgw.c
index 3b5aaf6..af6e3de 100644 (file)
@@ -2,6 +2,7 @@
  * Copyright(c) 2016 Intel Corporation
  */
 
+#include <stdbool.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <stdint.h>
 #include <sys/queue.h>
 #include <stdarg.h>
 #include <errno.h>
+#include <signal.h>
 #include <getopt.h>
 
 #include <rte_common.h>
+#include <rte_bitmap.h>
 #include <rte_byteorder.h>
 #include <rte_log.h>
 #include <rte_eal.h>
 #include <rte_jhash.h>
 #include <rte_cryptodev.h>
 #include <rte_security.h>
+#include <rte_eventdev.h>
 #include <rte_ip.h>
 #include <rte_ip_frag.h>
 
+#include "event_helper.h"
 #include "ipsec.h"
+#include "ipsec_worker.h"
 #include "parser.h"
+#include "sad.h"
 
-#define RTE_LOGTYPE_IPSEC RTE_LOGTYPE_USER1
+volatile bool force_quit;
 
 #define MAX_JUMBO_PKT_LEN  9600
 
 #define MEMPOOL_CACHE_SIZE 256
 
-#define NB_MBUF        (32000)
-
 #define CDEV_QUEUE_DESC 2048
 #define CDEV_MAP_ENTRIES 16384
 #define CDEV_MP_NB_OBJS 1024
@@ -63,8 +68,6 @@
 
 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
 
-#define NB_SOCKETS 4
-
 /* Configure how many packets ahead to prefetch, when reading packets */
 #define PREFETCH_OFFSET        3
 
@@ -72,8 +75,6 @@
 
 #define MAX_LCORE_PARAMS 1024
 
-#define UNPROTECTED_PORT(port) (unprotected_port_mask & (1 << portid))
-
 /*
  * Configurable number of RX/TX ring descriptors
  */
 static uint16_t nb_rxd = IPSEC_SECGW_RX_DESC_DEFAULT;
 static uint16_t nb_txd = IPSEC_SECGW_TX_DESC_DEFAULT;
 
-#if RTE_BYTE_ORDER != RTE_LITTLE_ENDIAN
-#define __BYTES_TO_UINT64(a, b, c, d, e, f, g, h) \
-       (((uint64_t)((a) & 0xff) << 56) | \
-       ((uint64_t)((b) & 0xff) << 48) | \
-       ((uint64_t)((c) & 0xff) << 40) | \
-       ((uint64_t)((d) & 0xff) << 32) | \
-       ((uint64_t)((e) & 0xff) << 24) | \
-       ((uint64_t)((f) & 0xff) << 16) | \
-       ((uint64_t)((g) & 0xff) << 8)  | \
-       ((uint64_t)(h) & 0xff))
-#else
-#define __BYTES_TO_UINT64(a, b, c, d, e, f, g, h) \
-       (((uint64_t)((h) & 0xff) << 56) | \
-       ((uint64_t)((g) & 0xff) << 48) | \
-       ((uint64_t)((f) & 0xff) << 40) | \
-       ((uint64_t)((e) & 0xff) << 32) | \
-       ((uint64_t)((d) & 0xff) << 24) | \
-       ((uint64_t)((c) & 0xff) << 16) | \
-       ((uint64_t)((b) & 0xff) << 8) | \
-       ((uint64_t)(a) & 0xff))
-#endif
-#define ETHADDR(a, b, c, d, e, f) (__BYTES_TO_UINT64(a, b, c, d, e, f, 0, 0))
-
 #define ETHADDR_TO_UINT64(addr) __BYTES_TO_UINT64( \
                (addr)->addr_bytes[0], (addr)->addr_bytes[1], \
                (addr)->addr_bytes[2], (addr)->addr_bytes[3], \
@@ -116,11 +94,6 @@ static uint16_t nb_txd = IPSEC_SECGW_TX_DESC_DEFAULT;
 
 #define MTU_TO_FRAMELEN(x)     ((x) + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN)
 
-/* port/source ethernet addr and destination ethernet addr */
-struct ethaddr_info {
-       uint64_t src, dst;
-};
-
 struct ethaddr_info ethaddr_tbl[RTE_MAX_ETHPORTS] = {
        { 0, ETHADDR(0x00, 0x16, 0x3e, 0x7e, 0x94, 0x9a) },
        { 0, ETHADDR(0x00, 0x16, 0x3e, 0x22, 0xa1, 0xd9) },
@@ -128,15 +101,25 @@ struct ethaddr_info ethaddr_tbl[RTE_MAX_ETHPORTS] = {
        { 0, ETHADDR(0x00, 0x16, 0x3e, 0x49, 0x9e, 0xdd) }
 };
 
+struct flow_info flow_info_tbl[RTE_MAX_ETHPORTS];
+
 #define CMD_LINE_OPT_CONFIG            "config"
 #define CMD_LINE_OPT_SINGLE_SA         "single-sa"
 #define CMD_LINE_OPT_CRYPTODEV_MASK    "cryptodev_mask"
+#define CMD_LINE_OPT_TRANSFER_MODE     "transfer-mode"
+#define CMD_LINE_OPT_SCHEDULE_TYPE     "event-schedule-type"
 #define CMD_LINE_OPT_RX_OFFLOAD                "rxoffload"
 #define CMD_LINE_OPT_TX_OFFLOAD                "txoffload"
 #define CMD_LINE_OPT_REASSEMBLE                "reassemble"
 #define CMD_LINE_OPT_MTU               "mtu"
 #define CMD_LINE_OPT_FRAG_TTL          "frag-ttl"
 
+#define CMD_LINE_ARG_EVENT     "event"
+#define CMD_LINE_ARG_POLL      "poll"
+#define CMD_LINE_ARG_ORDERED   "ordered"
+#define CMD_LINE_ARG_ATOMIC    "atomic"
+#define CMD_LINE_ARG_PARALLEL  "parallel"
+
 enum {
        /* long options mapped to a short option */
 
@@ -147,6 +130,8 @@ enum {
        CMD_LINE_OPT_CONFIG_NUM,
        CMD_LINE_OPT_SINGLE_SA_NUM,
        CMD_LINE_OPT_CRYPTODEV_MASK_NUM,
+       CMD_LINE_OPT_TRANSFER_MODE_NUM,
+       CMD_LINE_OPT_SCHEDULE_TYPE_NUM,
        CMD_LINE_OPT_RX_OFFLOAD_NUM,
        CMD_LINE_OPT_TX_OFFLOAD_NUM,
        CMD_LINE_OPT_REASSEMBLE_NUM,
@@ -158,6 +143,8 @@ static const struct option lgopts[] = {
        {CMD_LINE_OPT_CONFIG, 1, 0, CMD_LINE_OPT_CONFIG_NUM},
        {CMD_LINE_OPT_SINGLE_SA, 1, 0, CMD_LINE_OPT_SINGLE_SA_NUM},
        {CMD_LINE_OPT_CRYPTODEV_MASK, 1, 0, CMD_LINE_OPT_CRYPTODEV_MASK_NUM},
+       {CMD_LINE_OPT_TRANSFER_MODE, 1, 0, CMD_LINE_OPT_TRANSFER_MODE_NUM},
+       {CMD_LINE_OPT_SCHEDULE_TYPE, 1, 0, CMD_LINE_OPT_SCHEDULE_TYPE_NUM},
        {CMD_LINE_OPT_RX_OFFLOAD, 1, 0, CMD_LINE_OPT_RX_OFFLOAD_NUM},
        {CMD_LINE_OPT_TX_OFFLOAD, 1, 0, CMD_LINE_OPT_TX_OFFLOAD_NUM},
        {CMD_LINE_OPT_REASSEMBLE, 1, 0, CMD_LINE_OPT_REASSEMBLE_NUM},
@@ -166,15 +153,16 @@ static const struct option lgopts[] = {
        {NULL, 0, 0, 0}
 };
 
+uint32_t unprotected_port_mask;
+uint32_t single_sa_idx;
 /* mask of enabled ports */
 static uint32_t enabled_port_mask;
 static uint64_t enabled_cryptodev_mask = UINT64_MAX;
-static uint32_t unprotected_port_mask;
 static int32_t promiscuous_on = 1;
 static int32_t numa_on = 1; /**< NUMA is enabled by default. */
 static uint32_t nb_lcores;
 static uint32_t single_sa;
-static uint32_t single_sa_idx;
+static uint32_t nb_bufs_in_pool;
 
 /*
  * RX/TX HW offload capabilities to enable/use on ethernet ports.
@@ -192,7 +180,10 @@ static uint32_t mtu_size = RTE_ETHER_MTU;
 static uint64_t frag_ttl_ns = MAX_FRAG_TTL_NS;
 
 /* application wide librte_ipsec/SA parameters */
-struct app_sa_prm app_sa_prm = {.enable = 0};
+struct app_sa_prm app_sa_prm = {
+                       .enable = 0,
+                       .cache_sz = SA_CACHE_SZ
+               };
 static const char *cfgfile;
 
 struct lcore_rx_queue {
@@ -257,7 +248,7 @@ static struct rte_eth_conf port_conf = {
        },
 };
 
-static struct socket_ctx socket_ctx[NB_SOCKETS];
+struct socket_ctx socket_ctx[NB_SOCKETS];
 
 /*
  * Determine is multi-segment support required:
@@ -319,6 +310,7 @@ prepare_one_packet(struct rte_mbuf *pkt, struct ipsec_traffic *t)
                }
                pkt->l2_len = 0;
                pkt->l3_len = sizeof(*iph4);
+               pkt->packet_type |= RTE_PTYPE_L3_IPV4;
        } else if (eth->ether_type == rte_cpu_to_be_16(RTE_ETHER_TYPE_IPV6)) {
                int next_proto;
                size_t l3len, ext_len;
@@ -353,6 +345,7 @@ prepare_one_packet(struct rte_mbuf *pkt, struct ipsec_traffic *t)
                }
                pkt->l2_len = 0;
                pkt->l3_len = l3len;
+               pkt->packet_type |= RTE_PTYPE_L3_IPV6;
        } else {
                /* Unknown/Unsupported type, drop the packet */
                RTE_LOG(ERR, IPSEC, "Unsupported packet type 0x%x\n",
@@ -601,7 +594,7 @@ inbound_sp_sa(struct sp_ctx *sp, struct sa_ctx *sa, struct traffic_type *ip,
                        continue;
                }
 
-               sa_idx = SPI2IDX(res);
+               sa_idx = res - 1;
                if (!inbound_sa_check(sa, m, sa_idx)) {
                        rte_pktmbuf_free(m);
                        continue;
@@ -688,7 +681,7 @@ outbound_sp(struct sp_ctx *sp, struct traffic_type *ip,
        j = 0;
        for (i = 0; i < ip->num; i++) {
                m = ip->pkts[i];
-               sa_idx = SPI2IDX(ip->res[i]);
+               sa_idx = ip->res[i] - 1;
                if (ip->res[i] == DISCARD)
                        rte_pktmbuf_free(m);
                else if (ip->res[i] == BYPASS)
@@ -976,12 +969,12 @@ process_pkts(struct lcore_conf *qconf, struct rte_mbuf **pkts,
        prepare_traffic(pkts, &traffic, nb_pkts);
 
        if (unlikely(single_sa)) {
-               if (UNPROTECTED_PORT(portid))
+               if (is_unprotected_port(portid))
                        process_pkts_inbound_nosp(&qconf->inbound, &traffic);
                else
                        process_pkts_outbound_nosp(&qconf->outbound, &traffic);
        } else {
-               if (UNPROTECTED_PORT(portid))
+               if (is_unprotected_port(portid))
                        process_pkts_inbound(&qconf->inbound, &traffic);
                else
                        process_pkts_outbound(&qconf->outbound, &traffic);
@@ -1092,8 +1085,8 @@ drain_outbound_crypto_queues(const struct lcore_conf *qconf,
 }
 
 /* main processing loop */
-static int32_t
-main_loop(__attribute__((unused)) void *dummy)
+void
+ipsec_poll_mode_worker(void)
 {
        struct rte_mbuf *pkts[MAX_PKT_BURST];
        uint32_t lcore_id;
@@ -1102,7 +1095,7 @@ main_loop(__attribute__((unused)) void *dummy)
        uint16_t portid;
        uint8_t queueid;
        struct lcore_conf *qconf;
-       int32_t socket_id;
+       int32_t rc, socket_id;
        const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
                        / US_PER_S * BURST_TX_DRAIN_US;
        struct lcore_rx_queue *rxql;
@@ -1132,10 +1125,18 @@ main_loop(__attribute__((unused)) void *dummy)
        qconf->frag.pool_dir = socket_ctx[socket_id].mbuf_pool;
        qconf->frag.pool_indir = socket_ctx[socket_id].mbuf_pool_indir;
 
+       rc = ipsec_sad_lcore_cache_init(app_sa_prm.cache_sz);
+       if (rc != 0) {
+               RTE_LOG(ERR, IPSEC,
+                       "SAD cache init on lcore %u, failed with code: %d\n",
+                       lcore_id, rc);
+               return;
+       }
+
        if (qconf->nb_rx_queue == 0) {
                RTE_LOG(DEBUG, IPSEC, "lcore %u has nothing to do\n",
                        lcore_id);
-               return 0;
+               return;
        }
 
        RTE_LOG(INFO, IPSEC, "entering main loop on lcore %u\n", lcore_id);
@@ -1148,7 +1149,7 @@ main_loop(__attribute__((unused)) void *dummy)
                        lcore_id, portid, queueid);
        }
 
-       while (1) {
+       while (!force_quit) {
                cur_tsc = rte_rdtsc();
 
                /* TX queue buffer drain */
@@ -1172,7 +1173,7 @@ main_loop(__attribute__((unused)) void *dummy)
                                process_pkts(qconf, pkts, nb_rx, portid);
 
                        /* dequeue and process completed crypto-ops */
-                       if (UNPROTECTED_PORT(portid))
+                       if (is_unprotected_port(portid))
                                drain_inbound_crypto_queues(qconf,
                                        &qconf->inbound);
                        else
@@ -1183,13 +1184,19 @@ main_loop(__attribute__((unused)) void *dummy)
 }
 
 static int32_t
-check_params(void)
+check_poll_mode_params(struct eh_conf *eh_conf)
 {
        uint8_t lcore;
        uint16_t portid;
        uint16_t i;
        int32_t socket_id;
 
+       if (!eh_conf)
+               return -EINVAL;
+
+       if (eh_conf->mode != EH_PKT_TRANSFER_MODE_POLL)
+               return 0;
+
        if (lcore_params == NULL) {
                printf("Error: No port/queue/core mappings\n");
                return -1;
@@ -1271,10 +1278,14 @@ print_usage(const char *prgname)
                " [-w REPLAY_WINDOW_SIZE]"
                " [-e]"
                " [-a]"
+               " [-c]"
+               " [-s NUMBER_OF_MBUFS_IN_PKT_POOL]"
                " -f CONFIG_FILE"
                " --config (port,queue,lcore)[,(port,queue,lcore)]"
                " [--single-sa SAIDX]"
                " [--cryptodev_mask MASK]"
+               " [--transfer-mode MODE]"
+               " [--event-schedule-type TYPE]"
                " [--" CMD_LINE_OPT_RX_OFFLOAD " RX_OFFLOAD_MASK]"
                " [--" CMD_LINE_OPT_TX_OFFLOAD " TX_OFFLOAD_MASK]"
                " [--" CMD_LINE_OPT_REASSEMBLE " REASSEMBLE_TABLE_SIZE]"
@@ -1290,12 +1301,32 @@ print_usage(const char *prgname)
                "     size for each SA\n"
                "  -e enables ESN\n"
                "  -a enables SA SQN atomic behaviour\n"
+               "  -c specifies inbound SAD cache size,\n"
+               "     zero value disables the cache (default value: 128)\n"
+               "  -s number of mbufs in packet pool, if not specified number\n"
+               "     of mbufs will be calculated based on number of cores,\n"
+               "     ports and crypto queues\n"
                "  -f CONFIG_FILE: Configuration file\n"
-               "  --config (port,queue,lcore): Rx queue configuration\n"
-               "  --single-sa SAIDX: Use single SA index for outbound traffic,\n"
-               "                     bypassing the SP\n"
+               "  --config (port,queue,lcore): Rx queue configuration. In poll\n"
+               "                               mode determines which queues from\n"
+               "                               which ports are mapped to which cores.\n"
+               "                               In event mode this option is not used\n"
+               "                               as packets are dynamically scheduled\n"
+               "                               to cores by HW.\n"
+               "  --single-sa SAIDX: In poll mode use single SA index for\n"
+               "                     outbound traffic, bypassing the SP\n"
+               "                     In event mode selects driver submode,\n"
+               "                     SA index value is ignored\n"
                "  --cryptodev_mask MASK: Hexadecimal bitmask of the crypto\n"
                "                         devices to configure\n"
+               "  --transfer-mode MODE\n"
+               "               \"poll\"  : Packet transfer via polling (default)\n"
+               "               \"event\" : Packet transfer via event device\n"
+               "  --event-schedule-type TYPE queue schedule type, used only when\n"
+               "                             transfer mode is set to event\n"
+               "               \"ordered\"  : Ordered (default)\n"
+               "               \"atomic\"   : Atomic\n"
+               "               \"parallel\" : Parallel\n"
                "  --" CMD_LINE_OPT_RX_OFFLOAD
                ": bitmask of the RX HW offload capabilities to enable/use\n"
                "                         (DEV_RX_OFFLOAD_*)\n"
@@ -1430,8 +1461,45 @@ print_app_sa_prm(const struct app_sa_prm *prm)
        printf("Frag TTL: %" PRIu64 " ns\n", frag_ttl_ns);
 }
 
+static int
+parse_transfer_mode(struct eh_conf *conf, const char *optarg)
+{
+       if (!strcmp(CMD_LINE_ARG_POLL, optarg))
+               conf->mode = EH_PKT_TRANSFER_MODE_POLL;
+       else if (!strcmp(CMD_LINE_ARG_EVENT, optarg))
+               conf->mode = EH_PKT_TRANSFER_MODE_EVENT;
+       else {
+               printf("Unsupported packet transfer mode\n");
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static int
+parse_schedule_type(struct eh_conf *conf, const char *optarg)
+{
+       struct eventmode_conf *em_conf = NULL;
+
+       /* Get eventmode conf */
+       em_conf = conf->mode_params;
+
+       if (!strcmp(CMD_LINE_ARG_ORDERED, optarg))
+               em_conf->ext_params.sched_type = RTE_SCHED_TYPE_ORDERED;
+       else if (!strcmp(CMD_LINE_ARG_ATOMIC, optarg))
+               em_conf->ext_params.sched_type = RTE_SCHED_TYPE_ATOMIC;
+       else if (!strcmp(CMD_LINE_ARG_PARALLEL, optarg))
+               em_conf->ext_params.sched_type = RTE_SCHED_TYPE_PARALLEL;
+       else {
+               printf("Unsupported queue schedule type\n");
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
 static int32_t
-parse_args(int32_t argc, char **argv)
+parse_args(int32_t argc, char **argv, struct eh_conf *eh_conf)
 {
        int opt;
        int64_t ret;
@@ -1442,7 +1510,7 @@ parse_args(int32_t argc, char **argv)
 
        argvopt = argv;
 
-       while ((opt = getopt_long(argc, argvopt, "aelp:Pu:f:j:w:",
+       while ((opt = getopt_long(argc, argvopt, "aelp:Pu:f:j:w:c:s:",
                                lgopts, &option_index)) != EOF) {
 
                switch (opt) {
@@ -1476,6 +1544,19 @@ parse_args(int32_t argc, char **argv)
                        cfgfile = optarg;
                        f_present = 1;
                        break;
+
+               case 's':
+                       ret = parse_decimal(optarg);
+                       if (ret < 0) {
+                               printf("Invalid number of buffers in a pool: "
+                                       "%s\n", optarg);
+                               print_usage(prgname);
+                               return -1;
+                       }
+
+                       nb_bufs_in_pool = ret;
+                       break;
+
                case 'j':
                        ret = parse_decimal(optarg);
                        if (ret < RTE_MBUF_DEFAULT_BUF_SIZE ||
@@ -1501,6 +1582,15 @@ parse_args(int32_t argc, char **argv)
                        app_sa_prm.enable = 1;
                        app_sa_prm.flags |= RTE_IPSEC_SAFLAG_SQN_ATOM;
                        break;
+               case 'c':
+                       ret = parse_decimal(optarg);
+                       if (ret < 0) {
+                               printf("Invalid SA cache size: %s\n", optarg);
+                               print_usage(prgname);
+                               return -1;
+                       }
+                       app_sa_prm.cache_sz = ret;
+                       break;
                case CMD_LINE_OPT_CONFIG_NUM:
                        ret = parse_config(optarg);
                        if (ret) {
@@ -1520,6 +1610,7 @@ parse_args(int32_t argc, char **argv)
                        /* else */
                        single_sa = 1;
                        single_sa_idx = ret;
+                       eh_conf->ipsec_mode = EH_IPSEC_MODE_TYPE_DRIVER;
                        printf("Configured with single SA index %u\n",
                                        single_sa_idx);
                        break;
@@ -1534,6 +1625,25 @@ parse_args(int32_t argc, char **argv)
                        /* else */
                        enabled_cryptodev_mask = ret;
                        break;
+
+               case CMD_LINE_OPT_TRANSFER_MODE_NUM:
+                       ret = parse_transfer_mode(eh_conf, optarg);
+                       if (ret < 0) {
+                               printf("Invalid packet transfer mode\n");
+                               print_usage(prgname);
+                               return -1;
+                       }
+                       break;
+
+               case CMD_LINE_OPT_SCHEDULE_TYPE_NUM:
+                       ret = parse_schedule_type(eh_conf, optarg);
+                       if (ret < 0) {
+                               printf("Invalid queue schedule type\n");
+                               print_usage(prgname);
+                               return -1;
+                       }
+                       break;
+
                case CMD_LINE_OPT_RX_OFFLOAD_NUM:
                        ret = parse_mask(optarg, &dev_rx_offload);
                        if (ret != 0) {
@@ -1819,12 +1929,12 @@ check_cryptodev_mask(uint8_t cdev_id)
        return -1;
 }
 
-static int32_t
+static uint16_t
 cryptodevs_init(void)
 {
        struct rte_cryptodev_config dev_conf;
        struct rte_cryptodev_qp_conf qp_conf;
-       uint16_t idx, max_nb_qps, qp, i;
+       uint16_t idx, max_nb_qps, qp, total_nb_qps, i;
        int16_t cdev_id;
        struct rte_hash_parameters params = { 0 };
 
@@ -1852,6 +1962,7 @@ cryptodevs_init(void)
        printf("lcore/cryptodev/qp mappings:\n");
 
        idx = 0;
+       total_nb_qps = 0;
        for (cdev_id = 0; cdev_id < rte_cryptodev_count(); cdev_id++) {
                struct rte_cryptodev_info cdev_info;
 
@@ -1885,6 +1996,7 @@ cryptodevs_init(void)
                if (qp == 0)
                        continue;
 
+               total_nb_qps += qp;
                dev_conf.socket_id = rte_cryptodev_socket_id(cdev_id);
                dev_conf.nb_queue_pairs = qp;
                dev_conf.ff_disable = RTE_CRYPTODEV_FF_ASYMMETRIC_CRYPTO;
@@ -1917,7 +2029,7 @@ cryptodevs_init(void)
 
        printf("\n");
 
-       return 0;
+       return total_nb_qps;
 }
 
 static void
@@ -2406,17 +2518,201 @@ reassemble_init(void)
        return rc;
 }
 
+static void
+create_default_ipsec_flow(uint16_t port_id, uint64_t rx_offloads)
+{
+       struct rte_flow_action action[2];
+       struct rte_flow_item pattern[2];
+       struct rte_flow_attr attr = {0};
+       struct rte_flow_error err;
+       struct rte_flow *flow;
+       int ret;
+
+       if (!(rx_offloads & DEV_RX_OFFLOAD_SECURITY))
+               return;
+
+       /* Add the default rte_flow to enable SECURITY for all ESP packets */
+
+       pattern[0].type = RTE_FLOW_ITEM_TYPE_ESP;
+       pattern[0].spec = NULL;
+       pattern[0].mask = NULL;
+       pattern[0].last = NULL;
+       pattern[1].type = RTE_FLOW_ITEM_TYPE_END;
+
+       action[0].type = RTE_FLOW_ACTION_TYPE_SECURITY;
+       action[0].conf = NULL;
+       action[1].type = RTE_FLOW_ACTION_TYPE_END;
+       action[1].conf = NULL;
+
+       attr.ingress = 1;
+
+       ret = rte_flow_validate(port_id, &attr, pattern, action, &err);
+       if (ret)
+               return;
+
+       flow = rte_flow_create(port_id, &attr, pattern, action, &err);
+       if (flow == NULL)
+               return;
+
+       flow_info_tbl[port_id].rx_def_flow = flow;
+       RTE_LOG(INFO, IPSEC,
+               "Created default flow enabling SECURITY for all ESP traffic on port %d\n",
+               port_id);
+}
+
+static void
+signal_handler(int signum)
+{
+       if (signum == SIGINT || signum == SIGTERM) {
+               printf("\n\nSignal %d received, preparing to exit...\n",
+                               signum);
+               force_quit = true;
+       }
+}
+
+static void
+ev_mode_sess_verify(struct ipsec_sa *sa, int nb_sa)
+{
+       struct rte_ipsec_session *ips;
+       int32_t i;
+
+       if (!sa || !nb_sa)
+               return;
+
+       for (i = 0; i < nb_sa; i++) {
+               ips = ipsec_get_primary_session(&sa[i]);
+               if (ips->type != RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL)
+                       rte_exit(EXIT_FAILURE, "Event mode supports only "
+                                "inline protocol sessions\n");
+       }
+
+}
+
+static int32_t
+check_event_mode_params(struct eh_conf *eh_conf)
+{
+       struct eventmode_conf *em_conf = NULL;
+       struct lcore_params *params;
+       uint16_t portid;
+
+       if (!eh_conf || !eh_conf->mode_params)
+               return -EINVAL;
+
+       /* Get eventmode conf */
+       em_conf = eh_conf->mode_params;
+
+       if (eh_conf->mode == EH_PKT_TRANSFER_MODE_POLL &&
+           em_conf->ext_params.sched_type != SCHED_TYPE_NOT_SET) {
+               printf("error: option --event-schedule-type applies only to "
+                      "event mode\n");
+               return -EINVAL;
+       }
+
+       if (eh_conf->mode != EH_PKT_TRANSFER_MODE_EVENT)
+               return 0;
+
+       /* Set schedule type to ORDERED if it wasn't explicitly set by user */
+       if (em_conf->ext_params.sched_type == SCHED_TYPE_NOT_SET)
+               em_conf->ext_params.sched_type = RTE_SCHED_TYPE_ORDERED;
+
+       /*
+        * Event mode currently supports only inline protocol sessions.
+        * If there are other types of sessions configured then exit with
+        * error.
+        */
+       ev_mode_sess_verify(sa_in, nb_sa_in);
+       ev_mode_sess_verify(sa_out, nb_sa_out);
+
+
+       /* Option --config does not apply to event mode */
+       if (nb_lcore_params > 0) {
+               printf("error: option --config applies only to poll mode\n");
+               return -EINVAL;
+       }
+
+       /*
+        * In order to use the same port_init routine for both poll and event
+        * modes initialize lcore_params with one queue for each eth port
+        */
+       lcore_params = lcore_params_array;
+       RTE_ETH_FOREACH_DEV(portid) {
+               if ((enabled_port_mask & (1 << portid)) == 0)
+                       continue;
+
+               params = &lcore_params[nb_lcore_params++];
+               params->port_id = portid;
+               params->queue_id = 0;
+               params->lcore_id = rte_get_next_lcore(0, 0, 1);
+       }
+
+       return 0;
+}
+
+static void
+inline_sessions_free(struct sa_ctx *sa_ctx)
+{
+       struct rte_ipsec_session *ips;
+       struct ipsec_sa *sa;
+       int32_t ret;
+       uint32_t i;
+
+       if (!sa_ctx)
+               return;
+
+       for (i = 0; i < sa_ctx->nb_sa; i++) {
+
+               sa = &sa_ctx->sa[i];
+               if (!sa->spi)
+                       continue;
+
+               ips = ipsec_get_primary_session(sa);
+               if (ips->type != RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL &&
+                   ips->type != RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO)
+                       continue;
+
+               if (!rte_eth_dev_is_valid_port(sa->portid))
+                       continue;
+
+               ret = rte_security_session_destroy(
+                               rte_eth_dev_get_sec_ctx(sa->portid),
+                               ips->security.ses);
+               if (ret)
+                       RTE_LOG(ERR, IPSEC, "Failed to destroy security "
+                                           "session type %d, spi %d\n",
+                                           ips->type, sa->spi);
+       }
+}
+
+static uint32_t
+calculate_nb_mbufs(uint16_t nb_ports, uint16_t nb_crypto_qp, uint32_t nb_rxq,
+               uint32_t nb_txq)
+{
+       return RTE_MAX((nb_rxq * nb_rxd +
+                       nb_ports * nb_lcores * MAX_PKT_BURST +
+                       nb_ports * nb_txq * nb_txd +
+                       nb_lcores * MEMPOOL_CACHE_SIZE +
+                       nb_crypto_qp * CDEV_QUEUE_DESC +
+                       nb_lcores * frag_tbl_sz *
+                       FRAG_TBL_BUCKET_ENTRIES),
+                      8192U);
+}
+
 int32_t
 main(int32_t argc, char **argv)
 {
        int32_t ret;
-       uint32_t lcore_id;
+       uint32_t lcore_id, nb_txq, nb_rxq = 0;
+       uint32_t cdev_id;
        uint32_t i;
        uint8_t socket_id;
-       uint16_t portid;
-       uint64_t req_rx_offloads, req_tx_offloads;
+       uint16_t portid, nb_crypto_qp, nb_ports = 0;
+       uint64_t req_rx_offloads[RTE_MAX_ETHPORTS];
+       uint64_t req_tx_offloads[RTE_MAX_ETHPORTS];
+       struct eh_conf *eh_conf = NULL;
        size_t sess_sz;
 
+       nb_bufs_in_pool = 0;
+
        /* init EAL */
        ret = rte_eal_init(argc, argv);
        if (ret < 0)
@@ -2424,8 +2720,17 @@ main(int32_t argc, char **argv)
        argc -= ret;
        argv += ret;
 
+       force_quit = false;
+       signal(SIGINT, signal_handler);
+       signal(SIGTERM, signal_handler);
+
+       /* initialize event helper configuration */
+       eh_conf = eh_conf_init();
+       if (eh_conf == NULL)
+               rte_exit(EXIT_FAILURE, "Failed to init event helper config");
+
        /* parse application arguments (after the EAL ones) */
-       ret = parse_args(argc, argv);
+       ret = parse_args(argc, argv, eh_conf);
        if (ret < 0)
                rte_exit(EXIT_FAILURE, "Invalid parameters\n");
 
@@ -2442,8 +2747,11 @@ main(int32_t argc, char **argv)
                rte_exit(EXIT_FAILURE, "Invalid unprotected portmask 0x%x\n",
                                unprotected_port_mask);
 
-       if (check_params() < 0)
-               rte_exit(EXIT_FAILURE, "check_params failed\n");
+       if (check_poll_mode_params(eh_conf) < 0)
+               rte_exit(EXIT_FAILURE, "check_poll_mode_params failed\n");
+
+       if (check_event_mode_params(eh_conf) < 0)
+               rte_exit(EXIT_FAILURE, "check_event_mode_params failed\n");
 
        ret = init_lcore_rx_queues();
        if (ret < 0)
@@ -2453,6 +2761,22 @@ main(int32_t argc, char **argv)
 
        sess_sz = max_session_size();
 
+       nb_crypto_qp = cryptodevs_init();
+
+       if (nb_bufs_in_pool == 0) {
+               RTE_ETH_FOREACH_DEV(portid) {
+                       if ((enabled_port_mask & (1 << portid)) == 0)
+                               continue;
+                       nb_ports++;
+                       nb_rxq += get_port_nb_rx_queues(portid);
+               }
+
+               nb_txq = nb_lcores;
+
+               nb_bufs_in_pool = calculate_nb_mbufs(nb_ports, nb_crypto_qp,
+                                               nb_rxq, nb_txq);
+       }
+
        for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
                if (rte_lcore_is_enabled(lcore_id) == 0)
                        continue;
@@ -2466,32 +2790,43 @@ main(int32_t argc, char **argv)
                if (socket_ctx[socket_id].mbuf_pool)
                        continue;
 
-               pool_init(&socket_ctx[socket_id], socket_id, NB_MBUF);
+               pool_init(&socket_ctx[socket_id], socket_id, nb_bufs_in_pool);
                session_pool_init(&socket_ctx[socket_id], socket_id, sess_sz);
                session_priv_pool_init(&socket_ctx[socket_id], socket_id,
                        sess_sz);
        }
+       printf("Number of mbufs in packet pool %d\n", nb_bufs_in_pool);
 
        RTE_ETH_FOREACH_DEV(portid) {
                if ((enabled_port_mask & (1 << portid)) == 0)
                        continue;
 
-               sa_check_offloads(portid, &req_rx_offloads, &req_tx_offloads);
-               port_init(portid, req_rx_offloads, req_tx_offloads);
+               sa_check_offloads(portid, &req_rx_offloads[portid],
+                               &req_tx_offloads[portid]);
+               port_init(portid, req_rx_offloads[portid],
+                               req_tx_offloads[portid]);
        }
 
-       cryptodevs_init();
+       /*
+        * Set the enabled port mask in helper config for use by helper
+        * sub-system. This will be used while initializing devices using
+        * helper sub-system.
+        */
+       eh_conf->eth_portmask = enabled_port_mask;
+
+       /* Initialize eventmode components */
+       ret = eh_devs_init(eh_conf);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "eh_devs_init failed, err=%d\n", ret);
 
        /* start ports */
        RTE_ETH_FOREACH_DEV(portid) {
                if ((enabled_port_mask & (1 << portid)) == 0)
                        continue;
 
-               /*
-                * Start device
-                * note: device must be started before a flow rule
-                * can be installed.
-                */
+               /* Create flow before starting the device */
+               create_default_ipsec_flow(portid, req_rx_offloads[portid]);
+
                ret = rte_eth_dev_start(portid);
                if (ret < 0)
                        rte_exit(EXIT_FAILURE, "rte_eth_dev_start: "
@@ -2537,11 +2872,54 @@ main(int32_t argc, char **argv)
        check_all_ports_link_status(enabled_port_mask);
 
        /* launch per-lcore init on every lcore */
-       rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
+       rte_eal_mp_remote_launch(ipsec_launch_one_lcore, eh_conf, CALL_MASTER);
        RTE_LCORE_FOREACH_SLAVE(lcore_id) {
                if (rte_eal_wait_lcore(lcore_id) < 0)
                        return -1;
        }
 
+       /* Uninitialize eventmode components */
+       ret = eh_devs_uninit(eh_conf);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "eh_devs_uninit failed, err=%d\n", ret);
+
+       /* Free eventmode configuration memory */
+       eh_conf_uninit(eh_conf);
+
+       /* Destroy inline inbound and outbound sessions */
+       for (i = 0; i < NB_SOCKETS && i < rte_socket_count(); i++) {
+               socket_id = rte_socket_id_by_idx(i);
+               inline_sessions_free(socket_ctx[socket_id].sa_in);
+               inline_sessions_free(socket_ctx[socket_id].sa_out);
+       }
+
+       for (cdev_id = 0; cdev_id < rte_cryptodev_count(); cdev_id++) {
+               printf("Closing cryptodev %d...", cdev_id);
+               rte_cryptodev_stop(cdev_id);
+               rte_cryptodev_close(cdev_id);
+               printf(" Done\n");
+       }
+
+       RTE_ETH_FOREACH_DEV(portid) {
+               if ((enabled_port_mask & (1 << portid)) == 0)
+                       continue;
+
+               printf("Closing port %d...", portid);
+               if (flow_info_tbl[portid].rx_def_flow) {
+                       struct rte_flow_error err;
+
+                       ret = rte_flow_destroy(portid,
+                               flow_info_tbl[portid].rx_def_flow, &err);
+                       if (ret)
+                               RTE_LOG(ERR, IPSEC, "Failed to destroy flow "
+                                       " for port %u, err msg: %s\n", portid,
+                                       err.message);
+               }
+               rte_eth_dev_stop(portid);
+               rte_eth_dev_close(portid);
+               printf(" Done\n");
+       }
+       printf("Bye...\n");
+
        return 0;
 }