app/crypto-perf: support AES-CCM
[dpdk.git] / drivers / net / i40e / i40e_flow.c
index 08c0f8e..0d9c972 100644 (file)
@@ -52,8 +52,7 @@
 #include "base/i40e_prototype.h"
 #include "i40e_ethdev.h"
 
-#define I40E_IPV4_TC_SHIFT     4
-#define I40E_IPV6_TC_MASK      (0x00FF << I40E_IPV4_TC_SHIFT)
+#define I40E_IPV6_TC_MASK      (0xFF << I40E_FDIR_IPv6_TC_OFFSET)
 #define I40E_IPV6_FRAG_HEADER  44
 #define I40E_TENANT_ARRAY_NUM  3
 #define I40E_TCI_MASK          0xFFFF
@@ -85,11 +84,11 @@ static int i40e_flow_parse_ethertype_action(struct rte_eth_dev *dev,
 static int i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                        const struct rte_flow_item *pattern,
                                        struct rte_flow_error *error,
-                                       struct rte_eth_fdir_filter *filter);
+                                       struct i40e_fdir_filter_conf *filter);
 static int i40e_flow_parse_fdir_action(struct rte_eth_dev *dev,
                                       const struct rte_flow_action *actions,
                                       struct rte_flow_error *error,
-                                      struct rte_eth_fdir_filter *filter);
+                                      struct i40e_fdir_filter_conf *filter);
 static int i40e_flow_parse_tunnel_action(struct rte_eth_dev *dev,
                                 const struct rte_flow_action *actions,
                                 struct rte_flow_error *error,
@@ -126,6 +125,12 @@ static int i40e_flow_parse_mpls_filter(struct rte_eth_dev *dev,
                                       const struct rte_flow_action actions[],
                                       struct rte_flow_error *error,
                                       union i40e_filter_t *filter);
+static int i40e_flow_parse_gtp_filter(struct rte_eth_dev *dev,
+                                     const struct rte_flow_attr *attr,
+                                     const struct rte_flow_item pattern[],
+                                     const struct rte_flow_action actions[],
+                                     struct rte_flow_error *error,
+                                     union i40e_filter_t *filter);
 static int i40e_flow_destroy_ethertype_filter(struct i40e_pf *pf,
                                      struct i40e_ethertype_filter *filter);
 static int i40e_flow_destroy_tunnel_filter(struct i40e_pf *pf,
@@ -190,6 +195,40 @@ static enum rte_flow_item_type pattern_fdir_ipv4_sctp[] = {
        RTE_FLOW_ITEM_TYPE_END,
 };
 
+static enum rte_flow_item_type pattern_fdir_ipv4_gtpc[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV4,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPC,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_fdir_ipv4_gtpu[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV4,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPU,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_fdir_ipv4_gtpu_ipv4[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV4,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPU,
+       RTE_FLOW_ITEM_TYPE_IPV4,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_fdir_ipv4_gtpu_ipv6[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV4,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPU,
+       RTE_FLOW_ITEM_TYPE_IPV6,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
 static enum rte_flow_item_type pattern_fdir_ipv6[] = {
        RTE_FLOW_ITEM_TYPE_ETH,
        RTE_FLOW_ITEM_TYPE_IPV6,
@@ -217,6 +256,40 @@ static enum rte_flow_item_type pattern_fdir_ipv6_sctp[] = {
        RTE_FLOW_ITEM_TYPE_END,
 };
 
+static enum rte_flow_item_type pattern_fdir_ipv6_gtpc[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV6,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPC,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_fdir_ipv6_gtpu[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV6,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPU,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_fdir_ipv6_gtpu_ipv4[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV6,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPU,
+       RTE_FLOW_ITEM_TYPE_IPV4,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_fdir_ipv6_gtpu_ipv6[] = {
+       RTE_FLOW_ITEM_TYPE_ETH,
+       RTE_FLOW_ITEM_TYPE_IPV6,
+       RTE_FLOW_ITEM_TYPE_UDP,
+       RTE_FLOW_ITEM_TYPE_GTPU,
+       RTE_FLOW_ITEM_TYPE_IPV6,
+       RTE_FLOW_ITEM_TYPE_END,
+};
+
 static enum rte_flow_item_type pattern_fdir_ethertype_raw_1[] = {
        RTE_FLOW_ITEM_TYPE_ETH,
        RTE_FLOW_ITEM_TYPE_RAW,
@@ -1577,10 +1650,18 @@ static struct i40e_valid_pattern i40e_supported_patterns[] = {
        { pattern_fdir_ipv4_udp, i40e_flow_parse_fdir_filter },
        { pattern_fdir_ipv4_tcp, i40e_flow_parse_fdir_filter },
        { pattern_fdir_ipv4_sctp, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv4_gtpc, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv4_gtpu, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv4_gtpu_ipv4, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv4_gtpu_ipv6, i40e_flow_parse_fdir_filter },
        { pattern_fdir_ipv6, i40e_flow_parse_fdir_filter },
        { pattern_fdir_ipv6_udp, i40e_flow_parse_fdir_filter },
        { pattern_fdir_ipv6_tcp, i40e_flow_parse_fdir_filter },
        { pattern_fdir_ipv6_sctp, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv6_gtpc, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv6_gtpu, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv6_gtpu_ipv4, i40e_flow_parse_fdir_filter },
+       { pattern_fdir_ipv6_gtpu_ipv6, i40e_flow_parse_fdir_filter },
        /* FDIR - support default flow type with flexible payload */
        { pattern_fdir_ethertype_raw_1, i40e_flow_parse_fdir_filter },
        { pattern_fdir_ethertype_raw_2, i40e_flow_parse_fdir_filter },
@@ -1733,6 +1814,11 @@ static struct i40e_valid_pattern i40e_supported_patterns[] = {
        { pattern_mpls_2, i40e_flow_parse_mpls_filter },
        { pattern_mpls_3, i40e_flow_parse_mpls_filter },
        { pattern_mpls_4, i40e_flow_parse_mpls_filter },
+       /* GTP-C & GTP-U */
+       { pattern_fdir_ipv4_gtpc, i40e_flow_parse_gtp_filter },
+       { pattern_fdir_ipv4_gtpu, i40e_flow_parse_gtp_filter },
+       { pattern_fdir_ipv6_gtpc, i40e_flow_parse_gtp_filter },
+       { pattern_fdir_ipv6_gtpu, i40e_flow_parse_gtp_filter },
        /* QINQ */
        { pattern_qinq_1, i40e_flow_parse_qinq_filter },
 };
@@ -2303,20 +2389,58 @@ i40e_flow_set_fdir_inset(struct i40e_pf *pf,
        return 0;
 }
 
+static uint8_t
+i40e_flow_fdir_get_pctype_value(struct i40e_pf *pf,
+                               enum rte_flow_item_type item_type,
+                               struct i40e_fdir_filter_conf *filter)
+{
+       struct i40e_customized_pctype *cus_pctype = NULL;
+
+       switch (item_type) {
+       case RTE_FLOW_ITEM_TYPE_GTPC:
+               cus_pctype = i40e_find_customized_pctype(pf,
+                                                        I40E_CUSTOMIZED_GTPC);
+               break;
+       case RTE_FLOW_ITEM_TYPE_GTPU:
+               if (!filter->input.flow_ext.inner_ip)
+                       cus_pctype = i40e_find_customized_pctype(pf,
+                                                        I40E_CUSTOMIZED_GTPU);
+               else if (filter->input.flow_ext.iip_type ==
+                        I40E_FDIR_IPTYPE_IPV4)
+                       cus_pctype = i40e_find_customized_pctype(pf,
+                                                I40E_CUSTOMIZED_GTPU_IPV4);
+               else if (filter->input.flow_ext.iip_type ==
+                        I40E_FDIR_IPTYPE_IPV6)
+                       cus_pctype = i40e_find_customized_pctype(pf,
+                                                I40E_CUSTOMIZED_GTPU_IPV6);
+               break;
+       default:
+               PMD_DRV_LOG(ERR, "Unsupported item type");
+               break;
+       }
+
+       if (cus_pctype)
+               return cus_pctype->pctype;
+
+       return I40E_FILTER_PCTYPE_INVALID;
+}
+
 /* 1. Last in item should be NULL as range is not supported.
  * 2. Supported patterns: refer to array i40e_supported_patterns.
- * 3. Supported flow type and input set: refer to array
+ * 3. Default supported flow type and input set: refer to array
  *    valid_fdir_inset_table in i40e_ethdev.c.
  * 4. Mask of fields which need to be matched should be
  *    filled with 1.
  * 5. Mask of fields which needn't to be matched should be
  *    filled with 0.
+ * 6. GTP profile supports GTPv1 only.
+ * 7. GTP-C response message ('source_port' = 2123) is not supported.
  */
 static int
 i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                             const struct rte_flow_item *pattern,
                             struct rte_flow_error *error,
-                            struct rte_eth_fdir_filter *filter)
+                            struct i40e_fdir_filter_conf *filter)
 {
        struct i40e_pf *pf = I40E_DEV_PRIVATE_TO_PF(dev->data->dev_private);
        const struct rte_flow_item *item = pattern;
@@ -2327,15 +2451,16 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
        const struct rte_flow_item_tcp *tcp_spec, *tcp_mask;
        const struct rte_flow_item_udp *udp_spec, *udp_mask;
        const struct rte_flow_item_sctp *sctp_spec, *sctp_mask;
+       const struct rte_flow_item_gtp *gtp_spec, *gtp_mask;
        const struct rte_flow_item_raw *raw_spec, *raw_mask;
        const struct rte_flow_item_vf *vf_spec;
 
-       uint32_t flow_type = RTE_ETH_FLOW_UNKNOWN;
-       enum i40e_filter_pctype pctype;
+       uint8_t pctype = 0;
        uint64_t input_set = I40E_INSET_NONE;
        uint16_t frag_off;
        enum rte_flow_item_type item_type;
        enum rte_flow_item_type l3 = RTE_FLOW_ITEM_TYPE_END;
+       enum rte_flow_item_type cus_proto = RTE_FLOW_ITEM_TYPE_END;
        uint32_t i, j;
        uint8_t  ipv6_addr_mask[16] = {
                0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
@@ -2352,12 +2477,15 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
        bool cfg_flex_msk = true;
        uint16_t outer_tpid;
        uint16_t ether_type;
+       uint32_t vtc_flow_cpu;
+       bool outer_ip = true;
        int ret;
 
        memset(off_arr, 0, sizeof(off_arr));
        memset(len_arr, 0, sizeof(len_arr));
        memset(flex_mask, 0, I40E_FDIR_MAX_FLEX_LEN);
        outer_tpid = i40e_get_outer_vlan(dev);
+       filter->input.flow_ext.customized_pctype = false;
        for (; item->type != RTE_FLOW_ITEM_TYPE_END; item++) {
                if (item->last) {
                        rte_flow_error_set(error, EINVAL,
@@ -2402,7 +2530,7 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                }
                        }
 
-                       flow_type = RTE_ETH_FLOW_L2_PAYLOAD;
+                       pctype = I40E_FILTER_PCTYPE_L2_PAYLOAD;
                        layer_idx = I40E_FLXPLD_L2_IDX;
 
                        break;
@@ -2420,7 +2548,7 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                }
                        }
 
-                       flow_type = RTE_ETH_FLOW_L2_PAYLOAD;
+                       pctype = I40E_FILTER_PCTYPE_L2_PAYLOAD;
                        layer_idx = I40E_FLXPLD_L2_IDX;
 
                        break;
@@ -2431,7 +2559,7 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                        ipv4_mask =
                                (const struct rte_flow_item_ipv4 *)item->mask;
 
-                       if (ipv4_spec && ipv4_mask) {
+                       if (ipv4_spec && ipv4_mask && outer_ip) {
                                /* Check IPv4 mask and update input set */
                                if (ipv4_mask->hdr.version_ihl ||
                                    ipv4_mask->hdr.total_length ||
@@ -2457,13 +2585,13 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                        input_set |= I40E_INSET_IPV4_PROTO;
 
                                /* Get filter info */
-                               flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_OTHER;
+                               pctype = I40E_FILTER_PCTYPE_NONF_IPV4_OTHER;
                                /* Check if it is fragment. */
                                frag_off = ipv4_spec->hdr.fragment_offset;
                                frag_off = rte_be_to_cpu_16(frag_off);
                                if (frag_off & IPV4_HDR_OFFSET_MASK ||
                                    frag_off & IPV4_HDR_MF_FLAG)
-                                       flow_type = RTE_ETH_FLOW_FRAG_IPV4;
+                                       pctype = I40E_FILTER_PCTYPE_FRAG_IPV4;
 
                                /* Get the filter info */
                                filter->input.flow.ip4_flow.proto =
@@ -2476,9 +2604,22 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                        ipv4_spec->hdr.src_addr;
                                filter->input.flow.ip4_flow.dst_ip =
                                        ipv4_spec->hdr.dst_addr;
+
+                               layer_idx = I40E_FLXPLD_L3_IDX;
+                       } else if (!ipv4_spec && !ipv4_mask && !outer_ip) {
+                               filter->input.flow_ext.inner_ip = true;
+                               filter->input.flow_ext.iip_type =
+                                       I40E_FDIR_IPTYPE_IPV4;
+                       } else if ((ipv4_spec || ipv4_mask) && !outer_ip) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid inner IPv4 mask.");
+                               return -rte_errno;
                        }
 
-                       layer_idx = I40E_FLXPLD_L3_IDX;
+                       if (outer_ip)
+                               outer_ip = false;
 
                        break;
                case RTE_FLOW_ITEM_TYPE_IPV6:
@@ -2488,7 +2629,7 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                        ipv6_mask =
                                (const struct rte_flow_item_ipv6 *)item->mask;
 
-                       if (ipv6_spec && ipv6_mask) {
+                       if (ipv6_spec && ipv6_mask && outer_ip) {
                                /* Check IPv6 mask and update input set */
                                if (ipv6_mask->hdr.payload_len) {
                                        rte_flow_error_set(error, EINVAL,
@@ -2508,8 +2649,8 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                        input_set |= I40E_INSET_IPV6_DST;
 
                                if ((ipv6_mask->hdr.vtc_flow &
-                                    rte_cpu_to_be_16(I40E_IPV6_TC_MASK))
-                                   == rte_cpu_to_be_16(I40E_IPV6_TC_MASK))
+                                    rte_cpu_to_be_32(I40E_IPV6_TC_MASK))
+                                   == rte_cpu_to_be_32(I40E_IPV6_TC_MASK))
                                        input_set |= I40E_INSET_IPV6_TC;
                                if (ipv6_mask->hdr.proto == UINT8_MAX)
                                        input_set |= I40E_INSET_IPV6_NEXT_HDR;
@@ -2517,9 +2658,11 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                        input_set |= I40E_INSET_IPV6_HOP_LIMIT;
 
                                /* Get filter info */
+                               vtc_flow_cpu =
+                                     rte_be_to_cpu_32(ipv6_spec->hdr.vtc_flow);
                                filter->input.flow.ipv6_flow.tc =
-                                       (uint8_t)(ipv6_spec->hdr.vtc_flow <<
-                                                 I40E_IPV4_TC_SHIFT);
+                                       (uint8_t)(vtc_flow_cpu >>
+                                                 I40E_FDIR_IPv6_TC_OFFSET);
                                filter->input.flow.ipv6_flow.proto =
                                        ipv6_spec->hdr.proto;
                                filter->input.flow.ipv6_flow.hop_limits =
@@ -2533,15 +2676,26 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                                /* Check if it is fragment. */
                                if (ipv6_spec->hdr.proto ==
                                    I40E_IPV6_FRAG_HEADER)
-                                       flow_type =
-                                               RTE_ETH_FLOW_FRAG_IPV6;
+                                       pctype = I40E_FILTER_PCTYPE_FRAG_IPV6;
                                else
-                                       flow_type =
-                                               RTE_ETH_FLOW_NONFRAG_IPV6_OTHER;
+                                       pctype =
+                                            I40E_FILTER_PCTYPE_NONF_IPV6_OTHER;
+
+                               layer_idx = I40E_FLXPLD_L3_IDX;
+                       } else if (!ipv6_spec && !ipv6_mask && !outer_ip) {
+                               filter->input.flow_ext.inner_ip = true;
+                               filter->input.flow_ext.iip_type =
+                                       I40E_FDIR_IPTYPE_IPV6;
+                       } else if ((ipv6_spec || ipv6_mask) && !outer_ip) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid inner IPv6 mask");
+                               return -rte_errno;
                        }
 
-                       layer_idx = I40E_FLXPLD_L3_IDX;
-
+                       if (outer_ip)
+                               outer_ip = false;
                        break;
                case RTE_FLOW_ITEM_TYPE_TCP:
                        tcp_spec = (const struct rte_flow_item_tcp *)item->spec;
@@ -2570,11 +2724,11 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
 
                                /* Get filter info */
                                if (l3 == RTE_FLOW_ITEM_TYPE_IPV4)
-                                       flow_type =
-                                               RTE_ETH_FLOW_NONFRAG_IPV4_TCP;
+                                       pctype =
+                                              I40E_FILTER_PCTYPE_NONF_IPV4_TCP;
                                else if (l3 == RTE_FLOW_ITEM_TYPE_IPV6)
-                                       flow_type =
-                                               RTE_ETH_FLOW_NONFRAG_IPV6_TCP;
+                                       pctype =
+                                              I40E_FILTER_PCTYPE_NONF_IPV6_TCP;
 
                                if (l3 == RTE_FLOW_ITEM_TYPE_IPV4) {
                                        filter->input.flow.tcp4_flow.src_port =
@@ -2614,11 +2768,11 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
 
                                /* Get filter info */
                                if (l3 == RTE_FLOW_ITEM_TYPE_IPV4)
-                                       flow_type =
-                                               RTE_ETH_FLOW_NONFRAG_IPV4_UDP;
+                                       pctype =
+                                              I40E_FILTER_PCTYPE_NONF_IPV4_UDP;
                                else if (l3 == RTE_FLOW_ITEM_TYPE_IPV6)
-                                       flow_type =
-                                               RTE_ETH_FLOW_NONFRAG_IPV6_UDP;
+                                       pctype =
+                                              I40E_FILTER_PCTYPE_NONF_IPV6_UDP;
 
                                if (l3 == RTE_FLOW_ITEM_TYPE_IPV4) {
                                        filter->input.flow.udp4_flow.src_port =
@@ -2635,6 +2789,37 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
 
                        layer_idx = I40E_FLXPLD_L4_IDX;
 
+                       break;
+               case RTE_FLOW_ITEM_TYPE_GTPC:
+               case RTE_FLOW_ITEM_TYPE_GTPU:
+                       if (!pf->gtp_support) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Unsupported protocol");
+                               return -rte_errno;
+                       }
+
+                       gtp_spec = (const struct rte_flow_item_gtp *)item->spec;
+                       gtp_mask = (const struct rte_flow_item_gtp *)item->mask;
+
+                       if (gtp_spec && gtp_mask) {
+                               if (gtp_mask->v_pt_rsv_flags ||
+                                   gtp_mask->msg_type ||
+                                   gtp_mask->msg_len ||
+                                   gtp_mask->teid != UINT32_MAX) {
+                                       rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid GTP mask");
+                                       return -rte_errno;
+                               }
+
+                               filter->input.flow.gtp_flow.teid =
+                                       gtp_spec->teid;
+                               filter->input.flow_ext.customized_pctype = true;
+                               cus_proto = item_type;
+                       }
                        break;
                case RTE_FLOW_ITEM_TYPE_SCTP:
                        sctp_spec =
@@ -2661,11 +2846,11 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
 
                                /* Get filter info */
                                if (l3 == RTE_FLOW_ITEM_TYPE_IPV4)
-                                       flow_type =
-                                               RTE_ETH_FLOW_NONFRAG_IPV4_SCTP;
+                                       pctype =
+                                             I40E_FILTER_PCTYPE_NONF_IPV4_SCTP;
                                else if (l3 == RTE_FLOW_ITEM_TYPE_IPV6)
-                                       flow_type =
-                                               RTE_ETH_FLOW_NONFRAG_IPV6_SCTP;
+                                       pctype =
+                                             I40E_FILTER_PCTYPE_NONF_IPV6_SCTP;
 
                                if (l3 == RTE_FLOW_ITEM_TYPE_IPV4) {
                                        filter->input.flow.sctp4_flow.src_port =
@@ -2774,51 +2959,58 @@ i40e_flow_parse_fdir_pattern(struct rte_eth_dev *dev,
                }
        }
 
-       pctype = i40e_flowtype_to_pctype(flow_type);
-       if (pctype == 0 || pctype > I40E_FILTER_PCTYPE_L2_PAYLOAD) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ITEM, item,
-                                  "Unsupported flow type");
-               return -rte_errno;
+       /* Get customized pctype value */
+       if (filter->input.flow_ext.customized_pctype) {
+               pctype = i40e_flow_fdir_get_pctype_value(pf, cus_proto, filter);
+               if (pctype == I40E_FILTER_PCTYPE_INVALID) {
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM,
+                                          item,
+                                          "Unsupported pctype");
+                       return -rte_errno;
+               }
        }
 
-       ret = i40e_flow_set_fdir_inset(pf, pctype, input_set);
-       if (ret == -1) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ITEM, item,
-                                  "Conflict with the first rule's input set.");
-               return -rte_errno;
-       } else if (ret == -EINVAL) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ITEM, item,
-                                  "Invalid pattern mask.");
-               return -rte_errno;
-       }
+       /* If customized pctype is not used, set fdir configuration.*/
+       if (!filter->input.flow_ext.customized_pctype) {
+               ret = i40e_flow_set_fdir_inset(pf, pctype, input_set);
+               if (ret == -1) {
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM, item,
+                                          "Conflict with the first rule's input set.");
+                       return -rte_errno;
+               } else if (ret == -EINVAL) {
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM, item,
+                                          "Invalid pattern mask.");
+                       return -rte_errno;
+               }
 
-       filter->input.flow_type = flow_type;
+               /* Store flex mask to SW */
+               ret = i40e_flow_store_flex_mask(pf, pctype, flex_mask);
+               if (ret == -1) {
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM,
+                                          item,
+                                          "Exceed maximal number of bitmasks");
+                       return -rte_errno;
+               } else if (ret == -2) {
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM,
+                                          item,
+                                          "Conflict with the first flexible rule");
+                       return -rte_errno;
+               } else if (ret > 0)
+                       cfg_flex_msk = false;
 
-       /* Store flex mask to SW */
-       ret = i40e_flow_store_flex_mask(pf, pctype, flex_mask);
-       if (ret == -1) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                  item,
-                                  "Exceed maximal number of bitmasks");
-               return -rte_errno;
-       } else if (ret == -2) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ITEM,
-                                  item,
-                                  "Conflict with the first flexible rule");
-               return -rte_errno;
-       } else if (ret > 0)
-               cfg_flex_msk = false;
+               if (cfg_flex_pit)
+                       i40e_flow_set_fdir_flex_pit(pf, layer_idx, raw_id);
 
-       if (cfg_flex_pit)
-               i40e_flow_set_fdir_flex_pit(pf, layer_idx, raw_id);
+               if (cfg_flex_msk)
+                       i40e_flow_set_fdir_flex_msk(pf, pctype);
+       }
 
-       if (cfg_flex_msk)
-               i40e_flow_set_fdir_flex_msk(pf, pctype);
+       filter->input.pctype = pctype;
 
        return 0;
 }
@@ -2830,7 +3022,7 @@ static int
 i40e_flow_parse_fdir_action(struct rte_eth_dev *dev,
                            const struct rte_flow_action *actions,
                            struct rte_flow_error *error,
-                           struct rte_eth_fdir_filter *filter)
+                           struct i40e_fdir_filter_conf *filter)
 {
        struct i40e_pf *pf = I40E_DEV_PRIVATE_TO_PF(dev->data->dev_private);
        const struct rte_flow_action *act;
@@ -2844,19 +3036,22 @@ i40e_flow_parse_fdir_action(struct rte_eth_dev *dev,
        case RTE_FLOW_ACTION_TYPE_QUEUE:
                act_q = (const struct rte_flow_action_queue *)act->conf;
                filter->action.rx_queue = act_q->index;
-               if (filter->action.rx_queue >= pf->dev_data->nb_rx_queues) {
+               if ((!filter->input.flow_ext.is_vf &&
+                    filter->action.rx_queue >= pf->dev_data->nb_rx_queues) ||
+                   (filter->input.flow_ext.is_vf &&
+                    filter->action.rx_queue >= pf->vf_nb_qps)) {
                        rte_flow_error_set(error, EINVAL,
                                           RTE_FLOW_ERROR_TYPE_ACTION, act,
                                           "Invalid queue ID for FDIR.");
                        return -rte_errno;
                }
-               filter->action.behavior = RTE_ETH_FDIR_ACCEPT;
+               filter->action.behavior = I40E_FDIR_ACCEPT;
                break;
        case RTE_FLOW_ACTION_TYPE_DROP:
-               filter->action.behavior = RTE_ETH_FDIR_REJECT;
+               filter->action.behavior = I40E_FDIR_REJECT;
                break;
        case RTE_FLOW_ACTION_TYPE_PASSTHRU:
-               filter->action.behavior = RTE_ETH_FDIR_PASSTHRU;
+               filter->action.behavior = I40E_FDIR_PASSTHRU;
                break;
        default:
                rte_flow_error_set(error, EINVAL,
@@ -2871,11 +3066,11 @@ i40e_flow_parse_fdir_action(struct rte_eth_dev *dev,
        switch (act->type) {
        case RTE_FLOW_ACTION_TYPE_MARK:
                mark_spec = (const struct rte_flow_action_mark *)act->conf;
-               filter->action.report_status = RTE_ETH_FDIR_REPORT_ID;
+               filter->action.report_status = I40E_FDIR_REPORT_ID;
                filter->soft_id = mark_spec->id;
                break;
        case RTE_FLOW_ACTION_TYPE_FLAG:
-               filter->action.report_status = RTE_ETH_FDIR_NO_REPORT_STATUS;
+               filter->action.report_status = I40E_FDIR_NO_REPORT_STATUS;
                break;
        case RTE_FLOW_ACTION_TYPE_END:
                return 0;
@@ -2906,7 +3101,7 @@ i40e_flow_parse_fdir_filter(struct rte_eth_dev *dev,
                            struct rte_flow_error *error,
                            union i40e_filter_t *filter)
 {
-       struct rte_eth_fdir_filter *fdir_filter =
+       struct i40e_fdir_filter_conf *fdir_filter =
                &filter->fdir_filter;
        int ret;
 
@@ -3640,6 +3835,148 @@ i40e_flow_parse_mpls_filter(struct rte_eth_dev *dev,
        return ret;
 }
 
+/* 1. Last in item should be NULL as range is not supported.
+ * 2. Supported filter types: GTP TEID.
+ * 3. Mask of fields which need to be matched should be
+ *    filled with 1.
+ * 4. Mask of fields which needn't to be matched should be
+ *    filled with 0.
+ * 5. GTP profile supports GTPv1 only.
+ * 6. GTP-C response message ('source_port' = 2123) is not supported.
+ */
+static int
+i40e_flow_parse_gtp_pattern(struct rte_eth_dev *dev,
+                           const struct rte_flow_item *pattern,
+                           struct rte_flow_error *error,
+                           struct i40e_tunnel_filter_conf *filter)
+{
+       struct i40e_pf *pf = I40E_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+       const struct rte_flow_item *item = pattern;
+       const struct rte_flow_item_gtp *gtp_spec;
+       const struct rte_flow_item_gtp *gtp_mask;
+       enum rte_flow_item_type item_type;
+
+       if (!pf->gtp_support) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                  item,
+                                  "GTP is not supported by default.");
+               return -rte_errno;
+       }
+
+       for (; item->type != RTE_FLOW_ITEM_TYPE_END; item++) {
+               if (item->last) {
+                       rte_flow_error_set(error, EINVAL,
+                                          RTE_FLOW_ERROR_TYPE_ITEM,
+                                          item,
+                                          "Not support range");
+                       return -rte_errno;
+               }
+               item_type = item->type;
+               switch (item_type) {
+               case RTE_FLOW_ITEM_TYPE_ETH:
+                       if (item->spec || item->mask) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid ETH item");
+                               return -rte_errno;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_IPV4:
+                       filter->ip_type = I40E_TUNNEL_IPTYPE_IPV4;
+                       /* IPv4 is used to describe protocol,
+                        * spec and mask should be NULL.
+                        */
+                       if (item->spec || item->mask) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid IPv4 item");
+                               return -rte_errno;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_UDP:
+                       if (item->spec || item->mask) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid UDP item");
+                               return -rte_errno;
+                       }
+                       break;
+               case RTE_FLOW_ITEM_TYPE_GTPC:
+               case RTE_FLOW_ITEM_TYPE_GTPU:
+                       gtp_spec =
+                               (const struct rte_flow_item_gtp *)item->spec;
+                       gtp_mask =
+                               (const struct rte_flow_item_gtp *)item->mask;
+
+                       if (!gtp_spec || !gtp_mask) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid GTP item");
+                               return -rte_errno;
+                       }
+
+                       if (gtp_mask->v_pt_rsv_flags ||
+                           gtp_mask->msg_type ||
+                           gtp_mask->msg_len ||
+                           gtp_mask->teid != UINT32_MAX) {
+                               rte_flow_error_set(error, EINVAL,
+                                                  RTE_FLOW_ERROR_TYPE_ITEM,
+                                                  item,
+                                                  "Invalid GTP mask");
+                               return -rte_errno;
+                       }
+
+                       if (item_type == RTE_FLOW_ITEM_TYPE_GTPC)
+                               filter->tunnel_type = I40E_TUNNEL_TYPE_GTPC;
+                       else if (item_type == RTE_FLOW_ITEM_TYPE_GTPU)
+                               filter->tunnel_type = I40E_TUNNEL_TYPE_GTPU;
+
+                       filter->tenant_id = rte_be_to_cpu_32(gtp_spec->teid);
+
+                       break;
+               default:
+                       break;
+               }
+       }
+
+       return 0;
+}
+
+static int
+i40e_flow_parse_gtp_filter(struct rte_eth_dev *dev,
+                          const struct rte_flow_attr *attr,
+                          const struct rte_flow_item pattern[],
+                          const struct rte_flow_action actions[],
+                          struct rte_flow_error *error,
+                          union i40e_filter_t *filter)
+{
+       struct i40e_tunnel_filter_conf *tunnel_filter =
+               &filter->consistent_tunnel_filter;
+       int ret;
+
+       ret = i40e_flow_parse_gtp_pattern(dev, pattern,
+                                         error, tunnel_filter);
+       if (ret)
+               return ret;
+
+       ret = i40e_flow_parse_tunnel_action(dev, actions, error, tunnel_filter);
+       if (ret)
+               return ret;
+
+       ret = i40e_flow_parse_attr(attr, error);
+       if (ret)
+               return ret;
+
+       cons_filter_type = RTE_ETH_FILTER_TUNNEL;
+
+       return ret;
+}
+
 /* 1. Last in item should be NULL as range is not supported.
  * 2. Supported filter types: QINQ.
  * 3. Mask of fields which need to be matched should be
@@ -3714,8 +4051,10 @@ i40e_flow_parse_qinq_pattern(__rte_unused struct rte_eth_dev *dev,
        }
 
        /* Get filter specification */
-       if ((o_vlan_mask->tci == rte_cpu_to_be_16(I40E_TCI_MASK)) &&
-           (i_vlan_mask->tci == rte_cpu_to_be_16(I40E_TCI_MASK))) {
+       if ((o_vlan_mask != NULL) && (o_vlan_mask->tci ==
+                       rte_cpu_to_be_16(I40E_TCI_MASK)) &&
+                       (i_vlan_mask != NULL) &&
+                       (i_vlan_mask->tci == rte_cpu_to_be_16(I40E_TCI_MASK))) {
                filter->outer_vlan = rte_be_to_cpu_16(o_vlan_spec->tci)
                        & I40E_TCI_MASK;
                filter->inner_vlan = rte_be_to_cpu_16(i_vlan_spec->tci)
@@ -3870,7 +4209,7 @@ i40e_flow_create(struct rte_eth_dev *dev,
                                        i40e_ethertype_filter_list);
                break;
        case RTE_ETH_FILTER_FDIR:
-               ret = i40e_add_del_fdir_filter(dev,
+               ret = i40e_flow_add_del_fdir_filter(dev,
                                       &cons_filter.fdir_filter, 1);
                if (ret)
                        goto free_flow;
@@ -3920,7 +4259,7 @@ i40e_flow_destroy(struct rte_eth_dev *dev,
                              (struct i40e_tunnel_filter *)flow->rule);
                break;
        case RTE_ETH_FILTER_FDIR:
-               ret = i40e_add_del_fdir_filter(dev,
+               ret = i40e_flow_add_del_fdir_filter(dev,
                       &((struct i40e_fdir_filter *)flow->rule)->fdir, 0);
                break;
        default:
@@ -4009,12 +4348,12 @@ i40e_flow_destroy_tunnel_filter(struct i40e_pf *pf,
                vsi = vf->vsi;
        }
 
-       if (((filter->input.flags & I40E_AQC_ADD_CLOUD_FILTER_TEID_MPLSoUDP) ==
-           I40E_AQC_ADD_CLOUD_FILTER_TEID_MPLSoUDP) ||
-           ((filter->input.flags & I40E_AQC_ADD_CLOUD_FILTER_TEID_MPLSoGRE) ==
-           I40E_AQC_ADD_CLOUD_FILTER_TEID_MPLSoGRE) ||
-           ((filter->input.flags & I40E_AQC_ADD_CLOUD_FILTER_CUSTOM_QINQ) ==
-           I40E_AQC_ADD_CLOUD_FILTER_CUSTOM_QINQ))
+       if (((filter->input.flags & I40E_AQC_ADD_CLOUD_FILTER_0X11) ==
+           I40E_AQC_ADD_CLOUD_FILTER_0X11) ||
+           ((filter->input.flags & I40E_AQC_ADD_CLOUD_FILTER_0X12) ==
+           I40E_AQC_ADD_CLOUD_FILTER_0X12) ||
+           ((filter->input.flags & I40E_AQC_ADD_CLOUD_FILTER_0X10) ==
+           I40E_AQC_ADD_CLOUD_FILTER_0X10))
                big_buffer = 1;
 
        if (big_buffer)