examples/ipsec-secgw: support flow director
[dpdk.git] / examples / ipsec-secgw / sa.c
index 8043ed5..e3a1a5a 100644 (file)
@@ -1,5 +1,5 @@
 /* SPDX-License-Identifier: BSD-3-Clause
- * Copyright(c) 2016-2017 Intel Corporation
+ * Copyright(c) 2016-2020 Intel Corporation
  */
 
 /*
 #include "ipsec.h"
 #include "esp.h"
 #include "parser.h"
+#include "sad.h"
 
 #define IPDEFTTL 64
 
+#define IP4_FULL_MASK (sizeof(((struct ip_addr *)NULL)->ip.ip4) * CHAR_BIT)
+
+#define IP6_FULL_MASK (sizeof(((struct ip_addr *)NULL)->ip.ip6.ip6) * CHAR_BIT)
+
+#define MBUF_NO_SEC_OFFLOAD(m) ((m->ol_flags & PKT_RX_SEC_OFFLOAD) == 0)
+
 struct supported_cipher_algo {
        const char *keyword;
        enum rte_crypto_cipher_algorithm algo;
@@ -69,6 +76,13 @@ const struct supported_cipher_algo cipher_algos[] = {
                .block_size = 16,
                .key_len = 16
        },
+       {
+               .keyword = "aes-192-cbc",
+               .algo = RTE_CRYPTO_CIPHER_AES_CBC,
+               .iv_len = 16,
+               .block_size = 16,
+               .key_len = 24
+       },
        {
                .keyword = "aes-256-cbc",
                .algo = RTE_CRYPTO_CIPHER_AES_CBC,
@@ -109,7 +123,7 @@ const struct supported_auth_algo auth_algos[] = {
        {
                .keyword = "sha256-hmac",
                .algo = RTE_CRYPTO_AUTH_SHA256_HMAC,
-               .digest_len = 12,
+               .digest_len = 16,
                .key_len = 32
        }
 };
@@ -123,14 +137,38 @@ const struct supported_aead_algo aead_algos[] = {
                .key_len = 20,
                .digest_len = 16,
                .aad_len = 8,
+       },
+       {
+               .keyword = "aes-192-gcm",
+               .algo = RTE_CRYPTO_AEAD_AES_GCM,
+               .iv_len = 8,
+               .block_size = 4,
+               .key_len = 28,
+               .digest_len = 16,
+               .aad_len = 8,
+       },
+       {
+               .keyword = "aes-256-gcm",
+               .algo = RTE_CRYPTO_AEAD_AES_GCM,
+               .iv_len = 8,
+               .block_size = 4,
+               .key_len = 36,
+               .digest_len = 16,
+               .aad_len = 8,
        }
 };
 
-static struct ipsec_sa sa_out[IPSEC_SA_MAX_ENTRIES];
-static uint32_t nb_sa_out;
+#define SA_INIT_NB     128
+
+struct ipsec_sa *sa_out;
+uint32_t nb_sa_out;
+static uint32_t sa_out_sz;
+static struct ipsec_sa_cnt sa_out_cnt;
 
-static struct ipsec_sa sa_in[IPSEC_SA_MAX_ENTRIES];
-static uint32_t nb_sa_in;
+struct ipsec_sa *sa_in;
+uint32_t nb_sa_in;
+static uint32_t sa_in_sz;
+static struct ipsec_sa_cnt sa_in_cnt;
 
 static const struct supported_cipher_algo *
 find_match_cipher_algo(const char *cipher_keyword)
@@ -215,13 +253,40 @@ parse_key_string(const char *key_str, uint8_t *key)
        return nb_bytes;
 }
 
+static int
+extend_sa_arr(struct ipsec_sa **sa_tbl, uint32_t cur_cnt, uint32_t *cur_sz)
+{
+       if (*sa_tbl == NULL) {
+               *sa_tbl = calloc(SA_INIT_NB, sizeof(struct ipsec_sa));
+               if (*sa_tbl == NULL)
+                       return -1;
+               *cur_sz = SA_INIT_NB;
+               return 0;
+       }
+
+       if (cur_cnt >= *cur_sz) {
+               *sa_tbl = realloc(*sa_tbl,
+                       *cur_sz * sizeof(struct ipsec_sa) * 2);
+               if (*sa_tbl == NULL)
+                       return -1;
+               /* clean reallocated extra space */
+               memset(&(*sa_tbl)[*cur_sz], 0,
+                       *cur_sz * sizeof(struct ipsec_sa));
+               *cur_sz *= 2;
+       }
+
+       return 0;
+}
+
 void
 parse_sa_tokens(char **tokens, uint32_t n_tokens,
        struct parse_status *status)
 {
        struct ipsec_sa *rule = NULL;
+       struct rte_ipsec_session *ips;
        uint32_t ti; /*token index*/
        uint32_t *ri /*rule index*/;
+       struct ipsec_sa_cnt *sa_cnt;
        uint32_t cipher_algo_p = 0;
        uint32_t auth_algo_p = 0;
        uint32_t aead_algo_p = 0;
@@ -230,25 +295,23 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
        uint32_t mode_p = 0;
        uint32_t type_p = 0;
        uint32_t portid_p = 0;
+       uint32_t fallback_p = 0;
+       int16_t status_p = 0;
 
        if (strcmp(tokens[0], "in") == 0) {
                ri = &nb_sa_in;
-
-               APP_CHECK(*ri <= IPSEC_SA_MAX_ENTRIES - 1, status,
-                       "too many sa rules, abort insertion\n");
-               if (status->status < 0)
+               sa_cnt = &sa_in_cnt;
+               if (extend_sa_arr(&sa_in, nb_sa_in, &sa_in_sz) < 0)
                        return;
-
                rule = &sa_in[*ri];
+               rule->direction = RTE_SECURITY_IPSEC_SA_DIR_INGRESS;
        } else {
                ri = &nb_sa_out;
-
-               APP_CHECK(*ri <= IPSEC_SA_MAX_ENTRIES - 1, status,
-                       "too many sa rules, abort insertion\n");
-               if (status->status < 0)
+               sa_cnt = &sa_out_cnt;
+               if (extend_sa_arr(&sa_out, nb_sa_out, &sa_out_sz) < 0)
                        return;
-
                rule = &sa_out[*ri];
+               rule->direction = RTE_SECURITY_IPSEC_SA_DIR_EGRESS;
        }
 
        /* spi number */
@@ -258,6 +321,8 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
        if (atoi(tokens[1]) == INVALID_SPI)
                return;
        rule->spi = atoi(tokens[1]);
+       rule->portid = UINT16_MAX;
+       ips = ipsec_get_primary_session(rule);
 
        for (ti = 2; ti < n_tokens; ti++) {
                if (strcmp(tokens[ti], "mode") == 0) {
@@ -269,13 +334,17 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                        if (status->status < 0)
                                return;
 
-                       if (strcmp(tokens[ti], "ipv4-tunnel") == 0)
+                       if (strcmp(tokens[ti], "ipv4-tunnel") == 0) {
+                               sa_cnt->nb_v4++;
                                rule->flags = IP4_TUNNEL;
-                       else if (strcmp(tokens[ti], "ipv6-tunnel") == 0)
+                       } else if (strcmp(tokens[ti], "ipv6-tunnel") == 0) {
+                               sa_cnt->nb_v6++;
                                rule->flags = IP6_TUNNEL;
-                       else if (strcmp(tokens[ti], "transport") == 0)
+                       } else if (strcmp(tokens[ti], "transport") == 0) {
+                               sa_cnt->nb_v4++;
+                               sa_cnt->nb_v6++;
                                rule->flags = TRANSPORT;
-                       else {
+                       else {
                                APP_CHECK(0, status, "unrecognized "
                                        "input \"%s\"", tokens[ti]);
                                return;
@@ -303,6 +372,9 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                        APP_CHECK(algo != NULL, status, "unrecognized "
                                "input \"%s\"", tokens[ti]);
 
+                       if (status->status < 0)
+                               return;
+
                        rule->cipher_algo = algo->algo;
                        rule->block_size = algo->block_size;
                        rule->iv_len = algo->iv_len;
@@ -367,6 +439,9 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                        APP_CHECK(algo != NULL, status, "unrecognized "
                                "input \"%s\"", tokens[ti]);
 
+                       if (status->status < 0)
+                               return;
+
                        rule->auth_algo = algo->algo;
                        rule->auth_key_len = algo->key_len;
                        rule->digest_len = algo->digest_len;
@@ -422,6 +497,9 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                        APP_CHECK(algo != NULL, status, "unrecognized "
                                "input \"%s\"", tokens[ti]);
 
+                       if (status->status < 0)
+                               return;
+
                        rule->aead_algo = algo->algo;
                        rule->cipher_key_len = algo->key_len;
                        rule->digest_len = algo->digest_len;
@@ -468,7 +546,7 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                        if (status->status < 0)
                                return;
 
-                       if (rule->flags == IP4_TUNNEL) {
+                       if (IS_IP4_TUNNEL(rule->flags)) {
                                struct in_addr ip;
 
                                APP_CHECK(parse_ipv4_addr(tokens[ti],
@@ -480,7 +558,7 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                                        return;
                                rule->src.ip.ip4 = rte_bswap32(
                                        (uint32_t)ip.s_addr);
-                       } else if (rule->flags == IP6_TUNNEL) {
+                       } else if (IS_IP6_TUNNEL(rule->flags)) {
                                struct in6_addr ip;
 
                                APP_CHECK(parse_ipv6_addr(tokens[ti], &ip,
@@ -492,7 +570,7 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                                        return;
                                memcpy(rule->src.ip.ip6.ip6_b,
                                        ip.s6_addr, 16);
-                       } else if (rule->flags == TRANSPORT) {
+                       } else if (IS_TRANSPORT(rule->flags)) {
                                APP_CHECK(0, status, "unrecognized input "
                                        "\"%s\"", tokens[ti]);
                                return;
@@ -511,7 +589,7 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                        if (status->status < 0)
                                return;
 
-                       if (rule->flags == IP4_TUNNEL) {
+                       if (IS_IP4_TUNNEL(rule->flags)) {
                                struct in_addr ip;
 
                                APP_CHECK(parse_ipv4_addr(tokens[ti],
@@ -523,7 +601,7 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                                        return;
                                rule->dst.ip.ip4 = rte_bswap32(
                                        (uint32_t)ip.s_addr);
-                       } else if (rule->flags == IP6_TUNNEL) {
+                       } else if (IS_IP6_TUNNEL(rule->flags)) {
                                struct in6_addr ip;
 
                                APP_CHECK(parse_ipv6_addr(tokens[ti], &ip,
@@ -534,7 +612,7 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                                if (status->status < 0)
                                        return;
                                memcpy(rule->dst.ip.ip6.ip6_b, ip.s6_addr, 16);
-                       } else if (rule->flags == TRANSPORT) {
+                       } else if (IS_TRANSPORT(rule->flags)) {
                                APP_CHECK(0, status, "unrecognized "
                                        "input \"%s\"", tokens[ti]);
                                return;
@@ -554,18 +632,20 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                                return;
 
                        if (strcmp(tokens[ti], "inline-crypto-offload") == 0)
-                               rule->type =
+                               ips->type =
                                        RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO;
                        else if (strcmp(tokens[ti],
                                        "inline-protocol-offload") == 0)
-                               rule->type =
+                               ips->type =
                                RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL;
                        else if (strcmp(tokens[ti],
                                        "lookaside-protocol-offload") == 0)
-                               rule->type =
+                               ips->type =
                                RTE_SECURITY_ACTION_TYPE_LOOKASIDE_PROTOCOL;
                        else if (strcmp(tokens[ti], "no-offload") == 0)
-                               rule->type = RTE_SECURITY_ACTION_TYPE_NONE;
+                               ips->type = RTE_SECURITY_ACTION_TYPE_NONE;
+                       else if (strcmp(tokens[ti], "cpu-crypto") == 0)
+                               ips->type = RTE_SECURITY_ACTION_TYPE_CPU_CRYPTO;
                        else {
                                APP_CHECK(0, status, "Invalid input \"%s\"",
                                                tokens[ti]);
@@ -583,13 +663,99 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
                        INCREMENT_TOKEN_INDEX(ti, n_tokens, status);
                        if (status->status < 0)
                                return;
-                       rule->portid = atoi(tokens[ti]);
-                       if (status->status < 0)
+                       if (rule->portid == UINT16_MAX)
+                               rule->portid = atoi(tokens[ti]);
+                       else if (rule->portid != atoi(tokens[ti])) {
+                               APP_CHECK(0, status,
+                                       "portid %s not matching with already assigned portid %u",
+                                       tokens[ti], rule->portid);
                                return;
+                       }
                        portid_p = 1;
                        continue;
                }
 
+               if (strcmp(tokens[ti], "fallback") == 0) {
+                       struct rte_ipsec_session *fb;
+
+                       APP_CHECK(app_sa_prm.enable, status, "Fallback session "
+                               "not allowed for legacy mode.");
+                       if (status->status < 0)
+                               return;
+                       APP_CHECK(ips->type ==
+                               RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO, status,
+                               "Fallback session allowed if primary session "
+                               "is of type inline-crypto-offload only.");
+                       if (status->status < 0)
+                               return;
+                       APP_CHECK(rule->direction ==
+                               RTE_SECURITY_IPSEC_SA_DIR_INGRESS, status,
+                               "Fallback session not allowed for egress "
+                               "rule");
+                       if (status->status < 0)
+                               return;
+                       APP_CHECK_PRESENCE(fallback_p, tokens[ti], status);
+                       if (status->status < 0)
+                               return;
+                       INCREMENT_TOKEN_INDEX(ti, n_tokens, status);
+                       if (status->status < 0)
+                               return;
+                       fb = ipsec_get_fallback_session(rule);
+                       if (strcmp(tokens[ti], "lookaside-none") == 0)
+                               fb->type = RTE_SECURITY_ACTION_TYPE_NONE;
+                       else if (strcmp(tokens[ti], "cpu-crypto") == 0)
+                               fb->type = RTE_SECURITY_ACTION_TYPE_CPU_CRYPTO;
+                       else {
+                               APP_CHECK(0, status, "unrecognized fallback "
+                                       "type %s.", tokens[ti]);
+                               return;
+                       }
+
+                       rule->fallback_sessions = 1;
+                       fallback_p = 1;
+                       continue;
+               }
+               if (strcmp(tokens[ti], "flow-direction") == 0) {
+                       switch (ips->type) {
+                       case RTE_SECURITY_ACTION_TYPE_NONE:
+                       case RTE_SECURITY_ACTION_TYPE_CPU_CRYPTO:
+                               rule->fdir_flag = 1;
+                               INCREMENT_TOKEN_INDEX(ti, n_tokens, status);
+                               if (status->status < 0)
+                                       return;
+                               if (rule->portid == UINT16_MAX)
+                                       rule->portid = atoi(tokens[ti]);
+                               else if (rule->portid != atoi(tokens[ti])) {
+                                       APP_CHECK(0, status,
+                                               "portid %s not matching with already assigned portid %u",
+                                               tokens[ti], rule->portid);
+                                       return;
+                               }
+                               INCREMENT_TOKEN_INDEX(ti, n_tokens, status);
+                               if (status->status < 0)
+                                       return;
+                               rule->fdir_qid = atoi(tokens[ti]);
+                               /* validating portid and queueid */
+                               status_p = check_flow_params(rule->portid,
+                                               rule->fdir_qid);
+                               if (status_p < 0) {
+                                       printf("port id %u / queue id %u is "
+                                               "not valid\n", rule->portid,
+                                                rule->fdir_qid);
+                               }
+                               break;
+                       case RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO:
+                       case RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL:
+                       case RTE_SECURITY_ACTION_TYPE_LOOKASIDE_PROTOCOL:
+                       default:
+                               APP_CHECK(0, status,
+                                       "flow director not supported for security session type %d",
+                                       ips->type);
+                               return;
+                       }
+                       continue;
+               }
+
                /* unrecognizeable input */
                APP_CHECK(0, status, "unrecognized input \"%s\"",
                        tokens[ti]);
@@ -620,12 +786,13 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
        if (status->status < 0)
                return;
 
-       if ((rule->type != RTE_SECURITY_ACTION_TYPE_NONE) && (portid_p == 0))
+       if ((ips->type != RTE_SECURITY_ACTION_TYPE_NONE && ips->type !=
+                       RTE_SECURITY_ACTION_TYPE_CPU_CRYPTO) && (portid_p == 0))
                printf("Missing portid option, falling back to non-offload\n");
 
-       if (!type_p || !portid_p) {
-               rule->type = RTE_SECURITY_ACTION_TYPE_NONE;
-               rule->portid = -1;
+       if (!type_p || (!portid_p && ips->type !=
+                       RTE_SECURITY_ACTION_TYPE_CPU_CRYPTO)) {
+               ips->type = RTE_SECURITY_ACTION_TYPE_NONE;
        }
 
        *ri = *ri + 1;
@@ -636,6 +803,8 @@ print_one_sa_rule(const struct ipsec_sa *sa, int inbound)
 {
        uint32_t i;
        uint8_t a, b, c, d;
+       const struct rte_ipsec_session *ips;
+       const struct rte_ipsec_session *fallback_ips;
 
        printf("\tspi_%s(%3u):", inbound?"in":"out", sa->spi);
 
@@ -655,7 +824,8 @@ print_one_sa_rule(const struct ipsec_sa *sa, int inbound)
        }
 
        for (i = 0; i < RTE_DIM(aead_algos); i++) {
-               if (aead_algos[i].algo == sa->aead_algo) {
+               if (aead_algos[i].algo == sa->aead_algo &&
+                               aead_algos[i].key_len-4 == sa->cipher_key_len) {
                        printf("%s ", aead_algos[i].keyword);
                        break;
                }
@@ -663,7 +833,7 @@ print_one_sa_rule(const struct ipsec_sa *sa, int inbound)
 
        printf("mode:");
 
-       switch (sa->flags) {
+       switch (WITHOUT_TRANSPORT_VERSION(sa->flags)) {
        case IP4_TUNNEL:
                printf("IP4Tunnel ");
                uint32_t_to_char(sa->src.ip.ip4, &a, &b, &c, &d);
@@ -691,8 +861,10 @@ print_one_sa_rule(const struct ipsec_sa *sa, int inbound)
                printf("Transport ");
                break;
        }
+
+       ips = &sa->sessions[IPSEC_SESSION_PRIMARY];
        printf(" type:");
-       switch (sa->type) {
+       switch (ips->type) {
        case RTE_SECURITY_ACTION_TYPE_NONE:
                printf("no-offload ");
                break;
@@ -705,23 +877,35 @@ print_one_sa_rule(const struct ipsec_sa *sa, int inbound)
        case RTE_SECURITY_ACTION_TYPE_LOOKASIDE_PROTOCOL:
                printf("lookaside-protocol-offload ");
                break;
+       case RTE_SECURITY_ACTION_TYPE_CPU_CRYPTO:
+               printf("cpu-crypto-accelerated ");
+               break;
        }
+
+       fallback_ips = &sa->sessions[IPSEC_SESSION_FALLBACK];
+       if (fallback_ips != NULL && sa->fallback_sessions > 0) {
+               printf("inline fallback: ");
+               switch (fallback_ips->type) {
+               case RTE_SECURITY_ACTION_TYPE_NONE:
+                       printf("lookaside-none");
+                       break;
+               case RTE_SECURITY_ACTION_TYPE_CPU_CRYPTO:
+                       printf("cpu-crypto-accelerated");
+                       break;
+               default:
+                       printf("invalid");
+                       break;
+               }
+       }
+       if (sa->fdir_flag == 1)
+               printf("flow-direction port %d queue %d", sa->portid,
+                               sa->fdir_qid);
+
        printf("\n");
 }
 
-struct sa_ctx {
-       void *satbl; /* pointer to array of rte_ipsec_sa objects*/
-       struct ipsec_sa sa[IPSEC_SA_MAX_ENTRIES];
-       union {
-               struct {
-                       struct rte_crypto_sym_xform a;
-                       struct rte_crypto_sym_xform b;
-               };
-       } xf[IPSEC_SA_MAX_ENTRIES];
-};
-
 static struct sa_ctx *
-sa_create(const char *name, int32_t socket_id)
+sa_create(const char *name, int32_t socket_id, uint32_t nb_sa)
 {
        char s[PATH_MAX];
        struct sa_ctx *sa_ctx;
@@ -730,20 +914,31 @@ sa_create(const char *name, int32_t socket_id)
 
        snprintf(s, sizeof(s), "%s_%u", name, socket_id);
 
-       /* Create SA array table */
+       /* Create SA context */
        printf("Creating SA context with %u maximum entries on socket %d\n",
-                       IPSEC_SA_MAX_ENTRIES, socket_id);
+                       nb_sa, socket_id);
 
-       mz_size = sizeof(struct sa_ctx);
+       mz_size = sizeof(struct ipsec_xf) * nb_sa;
        mz = rte_memzone_reserve(s, mz_size, socket_id,
                        RTE_MEMZONE_1GB | RTE_MEMZONE_SIZE_HINT_ONLY);
        if (mz == NULL) {
-               printf("Failed to allocate SA DB memory\n");
-               rte_errno = -ENOMEM;
+               printf("Failed to allocate SA XFORM memory\n");
+               rte_errno = ENOMEM;
+               return NULL;
+       }
+
+       sa_ctx = rte_zmalloc(NULL, sizeof(struct sa_ctx) +
+               sizeof(struct ipsec_sa) * nb_sa, RTE_CACHE_LINE_SIZE);
+
+       if (sa_ctx == NULL) {
+               printf("Failed to allocate SA CTX memory\n");
+               rte_errno = ENOMEM;
+               rte_memzone_free(mz);
                return NULL;
        }
 
-       sa_ctx = (struct sa_ctx *)mz->addr;
+       sa_ctx->xf = (struct ipsec_xf *)mz->addr;
+       sa_ctx->nb_sa = nb_sa;
 
        return sa_ctx;
 }
@@ -752,8 +947,16 @@ static int
 check_eth_dev_caps(uint16_t portid, uint32_t inbound)
 {
        struct rte_eth_dev_info dev_info;
+       int retval;
 
-       rte_eth_dev_info_get(portid, &dev_info);
+       retval = rte_eth_dev_info_get(portid, &dev_info);
+       if (retval != 0) {
+               RTE_LOG(ERR, IPSEC,
+                       "Error during getting device (port %u) info: %s\n",
+                       portid, strerror(-retval));
+
+               return retval;
+       }
 
        if (inbound) {
                if ((dev_info.rx_offload_capa &
@@ -774,20 +977,111 @@ check_eth_dev_caps(uint16_t portid, uint32_t inbound)
        return 0;
 }
 
+/*
+ * Helper function, tries to determine next_proto for SPI
+ * by searching though SP rules.
+ */
+static int
+get_spi_proto(uint32_t spi, enum rte_security_ipsec_sa_direction dir,
+               struct ip_addr ip_addr[2], uint32_t mask[2])
+{
+       int32_t rc4, rc6;
+
+       rc4 = sp4_spi_present(spi, dir == RTE_SECURITY_IPSEC_SA_DIR_INGRESS,
+                               ip_addr, mask);
+       rc6 = sp6_spi_present(spi, dir == RTE_SECURITY_IPSEC_SA_DIR_INGRESS,
+                               ip_addr, mask);
+
+       if (rc4 >= 0) {
+               if (rc6 >= 0) {
+                       RTE_LOG(ERR, IPSEC,
+                               "%s: SPI %u used simultaeously by "
+                               "IPv4(%d) and IPv6 (%d) SP rules\n",
+                               __func__, spi, rc4, rc6);
+                       return -EINVAL;
+               } else
+                       return IPPROTO_IPIP;
+       } else if (rc6 < 0) {
+               RTE_LOG(ERR, IPSEC,
+                       "%s: SPI %u is not used by any SP rule\n",
+                       __func__, spi);
+               return -EINVAL;
+       } else
+               return IPPROTO_IPV6;
+}
+
+/*
+ * Helper function for getting source and destination IP addresses
+ * from SP. Needed for inline crypto transport mode, as addresses are not
+ * provided in config file for that mode. It checks if SP for current SA exists,
+ * and based on what type of protocol is returned, it stores appropriate
+ * addresses got from SP into SA.
+ */
+static int
+sa_add_address_inline_crypto(struct ipsec_sa *sa)
+{
+       int protocol;
+       struct ip_addr ip_addr[2];
+       uint32_t mask[2];
+
+       protocol = get_spi_proto(sa->spi, sa->direction, ip_addr, mask);
+       if (protocol < 0)
+               return protocol;
+       else if (protocol == IPPROTO_IPIP) {
+               sa->flags |= IP4_TRANSPORT;
+               if (mask[0] == IP4_FULL_MASK &&
+                               mask[1] == IP4_FULL_MASK &&
+                               ip_addr[0].ip.ip4 != 0 &&
+                               ip_addr[1].ip.ip4 != 0) {
+
+                       sa->src.ip.ip4 = ip_addr[0].ip.ip4;
+                       sa->dst.ip.ip4 = ip_addr[1].ip.ip4;
+               } else {
+                       RTE_LOG(ERR, IPSEC,
+                       "%s: No valid address or mask entry in"
+                       " IPv4 SP rule for SPI %u\n",
+                       __func__, sa->spi);
+                       return -EINVAL;
+               }
+       } else if (protocol == IPPROTO_IPV6) {
+               sa->flags |= IP6_TRANSPORT;
+               if (mask[0] == IP6_FULL_MASK &&
+                               mask[1] == IP6_FULL_MASK &&
+                               (ip_addr[0].ip.ip6.ip6[0] != 0 ||
+                               ip_addr[0].ip.ip6.ip6[1] != 0) &&
+                               (ip_addr[1].ip.ip6.ip6[0] != 0 ||
+                               ip_addr[1].ip.ip6.ip6[1] != 0)) {
+
+                       sa->src.ip.ip6 = ip_addr[0].ip.ip6;
+                       sa->dst.ip.ip6 = ip_addr[1].ip.ip6;
+               } else {
+                       RTE_LOG(ERR, IPSEC,
+                       "%s: No valid address or mask entry in"
+                       " IPv6 SP rule for SPI %u\n",
+                       __func__, sa->spi);
+                       return -EINVAL;
+               }
+       }
+       return 0;
+}
 
 static int
 sa_add_rules(struct sa_ctx *sa_ctx, const struct ipsec_sa entries[],
-               uint32_t nb_entries, uint32_t inbound)
+               uint32_t nb_entries, uint32_t inbound,
+               struct socket_ctx *skt_ctx)
 {
        struct ipsec_sa *sa;
        uint32_t i, idx;
        uint16_t iv_length, aad_length;
+       int inline_status;
+       int32_t rc;
+       struct rte_ipsec_session *ips;
 
        /* for ESN upper 32 bits of SQN also need to be part of AAD */
        aad_length = (app_sa_prm.enable_esn != 0) ? sizeof(uint32_t) : 0;
 
        for (i = 0; i < nb_entries; i++) {
-               idx = SPI2IDX(entries[i].spi);
+               idx = i;
                sa = &sa_ctx->sa[idx];
                if (sa->spi != 0) {
                        printf("Index %u already in use by SPI %u\n",
@@ -795,26 +1089,40 @@ sa_add_rules(struct sa_ctx *sa_ctx, const struct ipsec_sa entries[],
                        return -EINVAL;
                }
                *sa = entries[i];
+
+               if (inbound) {
+                       rc = ipsec_sad_add(&sa_ctx->sad, sa);
+                       if (rc != 0)
+                               return rc;
+               }
+
                sa->seq = 0;
+               ips = ipsec_get_primary_session(sa);
 
-               if (sa->type == RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL ||
-                       sa->type == RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO) {
+               if (ips->type == RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL ||
+                       ips->type == RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO) {
                        if (check_eth_dev_caps(sa->portid, inbound))
                                return -EINVAL;
                }
 
-               sa->direction = (inbound == 1) ?
-                               RTE_SECURITY_IPSEC_SA_DIR_INGRESS :
-                               RTE_SECURITY_IPSEC_SA_DIR_EGRESS;
-
-               switch (sa->flags) {
+               switch (WITHOUT_TRANSPORT_VERSION(sa->flags)) {
                case IP4_TUNNEL:
                        sa->src.ip.ip4 = rte_cpu_to_be_32(sa->src.ip.ip4);
                        sa->dst.ip.ip4 = rte_cpu_to_be_32(sa->dst.ip.ip4);
+                       break;
+               case TRANSPORT:
+                       if (ips->type ==
+                               RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO) {
+                               inline_status =
+                                       sa_add_address_inline_crypto(sa);
+                               if (inline_status < 0)
+                                       return inline_status;
+                       }
+                       break;
                }
 
                if (sa->aead_algo == RTE_CRYPTO_AEAD_AES_GCM) {
-                       iv_length = 16;
+                       iv_length = 12;
 
                        sa_ctx->xf[idx].a.type = RTE_CRYPTO_SYM_XFORM_AEAD;
                        sa_ctx->xf[idx].a.aead.algo = sa->aead_algo;
@@ -833,8 +1141,6 @@ sa_add_rules(struct sa_ctx *sa_ctx, const struct ipsec_sa entries[],
                                sa->digest_len;
 
                        sa->xforms = &sa_ctx->xf[idx].a;
-
-                       print_one_sa_rule(sa, inbound);
                } else {
                        switch (sa->cipher_algo) {
                        case RTE_CRYPTO_CIPHER_NULL:
@@ -899,9 +1205,28 @@ sa_add_rules(struct sa_ctx *sa_ctx, const struct ipsec_sa entries[],
                        sa_ctx->xf[idx].a.next = &sa_ctx->xf[idx].b;
                        sa_ctx->xf[idx].b.next = NULL;
                        sa->xforms = &sa_ctx->xf[idx].a;
+               }
+
+               if (ips->type ==
+                       RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL ||
+                       ips->type ==
+                       RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO) {
+                       rc = create_inline_session(skt_ctx, sa, ips);
+                       if (rc != 0) {
+                               RTE_LOG(ERR, IPSEC_ESP,
+                                       "create_inline_session() failed\n");
+                               return -EINVAL;
+                       }
+               }
 
-                       print_one_sa_rule(sa, inbound);
+               if (sa->fdir_flag && inbound) {
+                       rc = create_ipsec_esp_flow(sa);
+                       if (rc != 0)
+                               RTE_LOG(ERR, IPSEC_ESP,
+                                       "create_ipsec_esp_flow() failed %s\n",
+                                       strerror(rc));
                }
+               print_one_sa_rule(sa, inbound);
        }
 
        return 0;
@@ -909,16 +1234,16 @@ sa_add_rules(struct sa_ctx *sa_ctx, const struct ipsec_sa entries[],
 
 static inline int
 sa_out_add_rules(struct sa_ctx *sa_ctx, const struct ipsec_sa entries[],
-               uint32_t nb_entries)
+               uint32_t nb_entries, struct socket_ctx *skt_ctx)
 {
-       return sa_add_rules(sa_ctx, entries, nb_entries, 0);
+       return sa_add_rules(sa_ctx, entries, nb_entries, 0, skt_ctx);
 }
 
 static inline int
 sa_in_add_rules(struct sa_ctx *sa_ctx, const struct ipsec_sa entries[],
-               uint32_t nb_entries)
+               uint32_t nb_entries, struct socket_ctx *skt_ctx)
 {
-       return sa_add_rules(sa_ctx, entries, nb_entries, 1);
+       return sa_add_rules(sa_ctx, entries, nb_entries, 1, skt_ctx);
 }
 
 /*
@@ -932,37 +1257,7 @@ fill_ipsec_app_sa_prm(struct rte_ipsec_sa_prm *prm,
 
        prm->flags = app_prm->flags;
        prm->ipsec_xform.options.esn = app_prm->enable_esn;
-       prm->replay_win_sz = app_prm->window_size;
-}
-
-/*
- * Helper function, tries to determine next_proto for SPI
- * by searching though SP rules.
- */
-static int
-get_spi_proto(uint32_t spi, enum rte_security_ipsec_sa_direction dir)
-{
-       int32_t rc4, rc6;
-
-       rc4 = sp4_spi_present(spi, dir == RTE_SECURITY_IPSEC_SA_DIR_INGRESS);
-       rc6 = sp6_spi_present(spi, dir == RTE_SECURITY_IPSEC_SA_DIR_INGRESS);
-
-       if (rc4 >= 0) {
-               if (rc6 >= 0) {
-                       RTE_LOG(ERR, IPSEC,
-                               "%s: SPI %u used simultaeously by "
-                               "IPv4(%d) and IPv6 (%d) SP rules\n",
-                               __func__, spi, rc4, rc6);
-                       return -EINVAL;
-               } else
-                       return IPPROTO_IPIP;
-       } else if (rc6 < 0) {
-               RTE_LOG(ERR, IPSEC,
-                       "%s: SPI %u is not used by any SP rule\n",
-                       __func__, spi);
-               return -EINVAL;
-       } else
-               return IPPROTO_IPV6;
+       prm->ipsec_xform.replay_win_sz = app_prm->window_size;
 }
 
 static int
@@ -976,7 +1271,7 @@ fill_ipsec_sa_prm(struct rte_ipsec_sa_prm *prm, const struct ipsec_sa *ss,
         * probably not the optimal way, but there seems nothing
         * better right now.
         */
-       rc = get_spi_proto(ss->spi, ss->direction);
+       rc = get_spi_proto(ss->spi, ss->direction, NULL, NULL);
        if (rc < 0)
                return rc;
 
@@ -988,16 +1283,18 @@ fill_ipsec_sa_prm(struct rte_ipsec_sa_prm *prm, const struct ipsec_sa *ss,
        prm->ipsec_xform.salt = ss->salt;
        prm->ipsec_xform.direction = ss->direction;
        prm->ipsec_xform.proto = RTE_SECURITY_IPSEC_SA_PROTO_ESP;
-       prm->ipsec_xform.mode = (ss->flags == TRANSPORT) ?
+       prm->ipsec_xform.mode = (IS_TRANSPORT(ss->flags)) ?
                RTE_SECURITY_IPSEC_SA_MODE_TRANSPORT :
                RTE_SECURITY_IPSEC_SA_MODE_TUNNEL;
+       prm->ipsec_xform.options.ecn = 1;
+       prm->ipsec_xform.options.copy_dscp = 1;
 
-       if (ss->flags == IP4_TUNNEL) {
+       if (IS_IP4_TUNNEL(ss->flags)) {
                prm->ipsec_xform.tunnel.type = RTE_SECURITY_IPSEC_TUNNEL_IPV4;
                prm->tun.hdr_len = sizeof(*v4);
                prm->tun.next_proto = rc;
                prm->tun.hdr = v4;
-       } else if (ss->flags == IP6_TUNNEL) {
+       } else if (IS_IP6_TUNNEL(ss->flags)) {
                prm->ipsec_xform.tunnel.type = RTE_SECURITY_IPSEC_TUNNEL_IPV6;
                prm->tun.hdr_len = sizeof(*v6);
                prm->tun.next_proto = rc;
@@ -1012,22 +1309,23 @@ fill_ipsec_sa_prm(struct rte_ipsec_sa_prm *prm, const struct ipsec_sa *ss,
        return 0;
 }
 
-static void
-fill_ipsec_session(struct rte_ipsec_session *ss, struct rte_ipsec_sa *sa,
-       const struct ipsec_sa *lsa)
+static int
+fill_ipsec_session(struct rte_ipsec_session *ss, struct rte_ipsec_sa *sa)
 {
+       int32_t rc = 0;
+
        ss->sa = sa;
-       ss->type = lsa->type;
 
-       /* setup crypto section */
-       if (ss->type == RTE_SECURITY_ACTION_TYPE_NONE) {
-               ss->crypto.ses = lsa->crypto_session;
-       /* setup session action type */
-       } else {
-               ss->security.ses = lsa->sec_session;
-               ss->security.ctx = lsa->security_ctx;
-               ss->security.ol_flags = lsa->ol_flags;
+       if (ss->type == RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO ||
+               ss->type == RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL) {
+               if (ss->security.ses != NULL) {
+                       rc = rte_ipsec_session_prepare(ss);
+                       if (rc != 0)
+                               memset(ss, 0, sizeof(*ss));
+               }
        }
+
+       return rc;
 }
 
 /*
@@ -1038,9 +1336,10 @@ ipsec_sa_init(struct ipsec_sa *lsa, struct rte_ipsec_sa *sa, uint32_t sa_size)
 {
        int rc;
        struct rte_ipsec_sa_prm prm;
+       struct rte_ipsec_session *ips;
        struct rte_ipv4_hdr v4  = {
                .version_ihl = IPVERSION << 4 |
-                       sizeof(v4) / IPV4_IHL_MULTIPLIER,
+                       sizeof(v4) / RTE_IPV4_IHL_MULTIPLIER,
                .time_to_live = IPDEFTTL,
                .next_proto_id = IPPROTO_ESP,
                .src_addr = lsa->src.ip.ip4,
@@ -1051,7 +1350,7 @@ ipsec_sa_init(struct ipsec_sa *lsa, struct rte_ipsec_sa *sa, uint32_t sa_size)
                .proto = IPPROTO_ESP,
        };
 
-       if (lsa->flags == IP6_TUNNEL) {
+       if (IS_IP6_TUNNEL(lsa->flags)) {
                memcpy(v6.src_addr, lsa->src.ip.ip6.ip6_b, sizeof(v6.src_addr));
                memcpy(v6.dst_addr, lsa->dst.ip.ip6.ip6_b, sizeof(v6.dst_addr));
        }
@@ -1062,8 +1361,17 @@ ipsec_sa_init(struct ipsec_sa *lsa, struct rte_ipsec_sa *sa, uint32_t sa_size)
        if (rc < 0)
                return rc;
 
-       fill_ipsec_session(&lsa->ips, sa, lsa);
-       return 0;
+       /* init primary processing session */
+       ips = ipsec_get_primary_session(lsa);
+       rc = fill_ipsec_session(ips, sa);
+       if (rc != 0)
+               return rc;
+
+       /* init inline fallback processing session */
+       if (lsa->fallback_sessions == 1)
+               rc = fill_ipsec_session(ipsec_get_fallback_session(lsa), sa);
+
+       return rc;
 }
 
 /*
@@ -1071,8 +1379,7 @@ ipsec_sa_init(struct ipsec_sa *lsa, struct rte_ipsec_sa *sa, uint32_t sa_size)
  * one per session.
  */
 static int
-ipsec_satbl_init(struct sa_ctx *ctx, const struct ipsec_sa *ent,
-       uint32_t nb_ent, int32_t socket)
+ipsec_satbl_init(struct sa_ctx *ctx, uint32_t nb_ent, int32_t socket)
 {
        int32_t rc, sz;
        uint32_t i, idx;
@@ -1082,7 +1389,7 @@ ipsec_satbl_init(struct sa_ctx *ctx, const struct ipsec_sa *ent,
        struct rte_ipsec_sa_prm prm;
 
        /* determine SA size */
-       idx = SPI2IDX(ent[0].spi);
+       idx = 0;
        fill_ipsec_sa_prm(&prm, ctx->sa + idx, NULL, NULL);
        sz = rte_ipsec_sa_size(&prm);
        if (sz < 0) {
@@ -1105,7 +1412,7 @@ ipsec_satbl_init(struct sa_ctx *ctx, const struct ipsec_sa *ent,
        rc = 0;
        for (i = 0; i != nb_ent && rc == 0; i++) {
 
-               idx = SPI2IDX(ent[i].spi);
+               idx = i;
 
                sa = (struct rte_ipsec_sa *)((uintptr_t)ctx->satbl + sz * i);
                lsa = ctx->sa + idx;
@@ -1116,27 +1423,37 @@ ipsec_satbl_init(struct sa_ctx *ctx, const struct ipsec_sa *ent,
        return rc;
 }
 
+static int
+sa_cmp(const void *p, const void *q)
+{
+       uint32_t spi1 = ((const struct ipsec_sa *)p)->spi;
+       uint32_t spi2 = ((const struct ipsec_sa *)q)->spi;
+
+       return (int)(spi1 - spi2);
+}
+
 /*
  * Walk through all SA rules to find an SA with given SPI
  */
 int
-sa_spi_present(uint32_t spi, int inbound)
+sa_spi_present(struct sa_ctx *sa_ctx, uint32_t spi, int inbound)
 {
-       uint32_t i, num;
+       uint32_t num;
+       struct ipsec_sa *sa;
+       struct ipsec_sa tmpl;
        const struct ipsec_sa *sar;
 
-       if (inbound != 0) {
-               sar = sa_in;
+       sar = sa_ctx->sa;
+       if (inbound != 0)
                num = nb_sa_in;
-       } else {
-               sar = sa_out;
+       else
                num = nb_sa_out;
-       }
 
-       for (i = 0; i != num; i++) {
-               if (sar[i].spi == spi)
-                       return i;
-       }
+       tmpl.spi = spi;
+
+       sa = bsearch(&tmpl, sar, num, sizeof(struct ipsec_sa), sa_cmp);
+       if (sa != NULL)
+               return RTE_PTR_DIFF(sa, sar) / sizeof(struct ipsec_sa);
 
        return -ENOENT;
 }
@@ -1160,16 +1477,21 @@ sa_init(struct socket_ctx *ctx, int32_t socket_id)
 
        if (nb_sa_in > 0) {
                name = "sa_in";
-               ctx->sa_in = sa_create(name, socket_id);
+               ctx->sa_in = sa_create(name, socket_id, nb_sa_in);
                if (ctx->sa_in == NULL)
                        rte_exit(EXIT_FAILURE, "Error [%d] creating SA "
                                "context %s in socket %d\n", rte_errno,
                                name, socket_id);
 
-               sa_in_add_rules(ctx->sa_in, sa_in, nb_sa_in);
+               rc = ipsec_sad_create(name, &ctx->sa_in->sad, socket_id,
+                               &sa_in_cnt);
+               if (rc != 0)
+                       rte_exit(EXIT_FAILURE, "failed to init SAD\n");
+
+               sa_in_add_rules(ctx->sa_in, sa_in, nb_sa_in, ctx);
 
                if (app_sa_prm.enable != 0) {
-                       rc = ipsec_satbl_init(ctx->sa_in, sa_in, nb_sa_in,
+                       rc = ipsec_satbl_init(ctx->sa_in, nb_sa_in,
                                socket_id);
                        if (rc != 0)
                                rte_exit(EXIT_FAILURE,
@@ -1180,16 +1502,16 @@ sa_init(struct socket_ctx *ctx, int32_t socket_id)
 
        if (nb_sa_out > 0) {
                name = "sa_out";
-               ctx->sa_out = sa_create(name, socket_id);
+               ctx->sa_out = sa_create(name, socket_id, nb_sa_out);
                if (ctx->sa_out == NULL)
                        rte_exit(EXIT_FAILURE, "Error [%d] creating SA "
                                "context %s in socket %d\n", rte_errno,
                                name, socket_id);
 
-               sa_out_add_rules(ctx->sa_out, sa_out, nb_sa_out);
+               sa_out_add_rules(ctx->sa_out, sa_out, nb_sa_out, ctx);
 
                if (app_sa_prm.enable != 0) {
-                       rc = ipsec_satbl_init(ctx->sa_out, sa_out, nb_sa_out,
+                       rc = ipsec_satbl_init(ctx->sa_out, nb_sa_out,
                                socket_id);
                        if (rc != 0)
                                rte_exit(EXIT_FAILURE,
@@ -1215,64 +1537,43 @@ inbound_sa_check(struct sa_ctx *sa_ctx, struct rte_mbuf *m, uint32_t sa_idx)
        return 0;
 }
 
-static inline void
-single_inbound_lookup(struct ipsec_sa *sadb, struct rte_mbuf *pkt,
-               struct ipsec_sa **sa_ret)
+void
+inbound_sa_lookup(struct sa_ctx *sa_ctx, struct rte_mbuf *pkts[],
+               void *sa_arr[], uint16_t nb_pkts)
 {
-       struct rte_esp_hdr *esp;
-       struct ip *ip;
-       uint32_t *src4_addr;
-       uint8_t *src6_addr;
+       uint32_t i;
+       void *result_sa;
        struct ipsec_sa *sa;
 
-       *sa_ret = NULL;
-
-       ip = rte_pktmbuf_mtod(pkt, struct ip *);
-       if (ip->ip_v == IPVERSION)
-               esp = (struct rte_esp_hdr *)(ip + 1);
-       else
-               esp = (struct rte_esp_hdr *)(((struct ip6_hdr *)ip) + 1);
-
-       if (esp->spi == INVALID_SPI)
-               return;
+       sad_lookup(&sa_ctx->sad, pkts, sa_arr, nb_pkts);
 
-       sa = &sadb[SPI2IDX(rte_be_to_cpu_32(esp->spi))];
-       if (rte_be_to_cpu_32(esp->spi) != sa->spi)
-               return;
+       /*
+        * Mark need for inline offload fallback on the LSB of SA pointer.
+        * Thanks to packet grouping mechanism which ipsec_process is using
+        * packets marked for fallback processing will form separate group.
+        *
+        * Because it is not safe to use SA pointer it is casted to generic
+        * pointer to prevent from unintentional use. Use ipsec_mask_saptr
+        * to get valid struct pointer.
+        */
+       for (i = 0; i < nb_pkts; i++) {
+               if (sa_arr[i] == NULL)
+                       continue;
 
-       switch (sa->flags) {
-       case IP4_TUNNEL:
-               src4_addr = RTE_PTR_ADD(ip, offsetof(struct ip, ip_src));
-               if ((ip->ip_v == IPVERSION) &&
-                               (sa->src.ip.ip4 == *src4_addr) &&
-                               (sa->dst.ip.ip4 == *(src4_addr + 1)))
-                       *sa_ret = sa;
-               break;
-       case IP6_TUNNEL:
-               src6_addr = RTE_PTR_ADD(ip, offsetof(struct ip6_hdr, ip6_src));
-               if ((ip->ip_v == IP6_VERSION) &&
-                               !memcmp(&sa->src.ip.ip6.ip6, src6_addr, 16) &&
-                               !memcmp(&sa->dst.ip.ip6.ip6, src6_addr + 16, 16))
-                       *sa_ret = sa;
-               break;
-       case TRANSPORT:
-               *sa_ret = sa;
+               result_sa = sa = sa_arr[i];
+               if (MBUF_NO_SEC_OFFLOAD(pkts[i]) &&
+                       sa->fallback_sessions > 0) {
+                       uintptr_t intsa = (uintptr_t)sa;
+                       intsa |= IPSEC_SA_OFFLOAD_FALLBACK_FLAG;
+                       result_sa = (void *)intsa;
+               }
+               sa_arr[i] = result_sa;
        }
 }
 
-void
-inbound_sa_lookup(struct sa_ctx *sa_ctx, struct rte_mbuf *pkts[],
-               struct ipsec_sa *sa[], uint16_t nb_pkts)
-{
-       uint32_t i;
-
-       for (i = 0; i < nb_pkts; i++)
-               single_inbound_lookup(sa_ctx->sa, pkts[i], &sa[i]);
-}
-
 void
 outbound_sa_lookup(struct sa_ctx *sa_ctx, uint32_t sa_idx[],
-               struct ipsec_sa *sa[], uint16_t nb_pkts)
+               void *sa[], uint16_t nb_pkts)
 {
        uint32_t i;
 
@@ -1289,6 +1590,7 @@ sa_check_offloads(uint16_t port_id, uint64_t *rx_offloads,
 {
        struct ipsec_sa *rule;
        uint32_t idx_sa;
+       enum rte_security_session_action_type rule_type;
 
        *rx_offloads = 0;
        *tx_offloads = 0;
@@ -1296,8 +1598,9 @@ sa_check_offloads(uint16_t port_id, uint64_t *rx_offloads,
        /* Check for inbound rules that use offloads and use this port */
        for (idx_sa = 0; idx_sa < nb_sa_in; idx_sa++) {
                rule = &sa_in[idx_sa];
-               if ((rule->type == RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO ||
-                               rule->type ==
+               rule_type = ipsec_get_action_type(rule);
+               if ((rule_type == RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO ||
+                               rule_type ==
                                RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL)
                                && rule->portid == port_id)
                        *rx_offloads |= DEV_RX_OFFLOAD_SECURITY;
@@ -1306,11 +1609,19 @@ sa_check_offloads(uint16_t port_id, uint64_t *rx_offloads,
        /* Check for outbound rules that use offloads and use this port */
        for (idx_sa = 0; idx_sa < nb_sa_out; idx_sa++) {
                rule = &sa_out[idx_sa];
-               if ((rule->type == RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO ||
-                               rule->type ==
+               rule_type = ipsec_get_action_type(rule);
+               if ((rule_type == RTE_SECURITY_ACTION_TYPE_INLINE_CRYPTO ||
+                               rule_type ==
                                RTE_SECURITY_ACTION_TYPE_INLINE_PROTOCOL)
                                && rule->portid == port_id)
                        *tx_offloads |= DEV_TX_OFFLOAD_SECURITY;
        }
        return 0;
 }
+
+void
+sa_sort_arr(void)
+{
+       qsort(sa_in, nb_sa_in, sizeof(struct ipsec_sa), sa_cmp);
+       qsort(sa_out, nb_sa_out, sizeof(struct ipsec_sa), sa_cmp);
+}