net: add rte prefix to ether defines
[dpdk.git] / drivers / net / sfc / sfc_flow.c
index 05be5a5..e4a9ba0 100644 (file)
@@ -1,36 +1,16 @@
-/*-
- * Copyright (c) 2017 Solarflare Communications Inc.
+/* SPDX-License-Identifier: BSD-3-Clause
+ *
+ * Copyright (c) 2017-2018 Solarflare Communications Inc.
  * All rights reserved.
  *
  * This software was jointly developed between OKTET Labs (under contract
  * for Solarflare) and Solarflare Communications, Inc.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions are met:
- *
- * 1. Redistributions of source code must retain the above copyright notice,
- *    this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright notice,
- *    this list of conditions and the following disclaimer in the documentation
- *    and/or other materials provided with the distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
- * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
- * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
- * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
- * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
- * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
- * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
- * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
- * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
- * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
- * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <rte_byteorder.h>
 #include <rte_tailq.h>
 #include <rte_common.h>
-#include <rte_ethdev.h>
-#include <rte_eth_ctrl.h>
+#include <rte_ethdev_driver.h>
 #include <rte_ether.h>
 #include <rte_flow.h>
 #include <rte_flow_driver.h>
 #include "sfc_filter.h"
 #include "sfc_flow.h"
 #include "sfc_log.h"
+#include "sfc_dp_rx.h"
 
 /*
  * At now flow API is implemented in such a manner that each
- * flow rule is converted to a hardware filter.
+ * flow rule is converted to one or more hardware filters.
  * All elements of flow rule (attributes, pattern items, actions)
  * correspond to one or more fields in the efx_filter_spec_s structure
  * that is responsible for the hardware filter.
+ * If some required field is unset in the flow rule, then a handful
+ * of filter copies will be created to cover all possible values
+ * of such a field.
  */
 
 enum sfc_flow_item_layers {
@@ -56,6 +40,7 @@ enum sfc_flow_item_layers {
        SFC_FLOW_ITEM_START_LAYER,
        SFC_FLOW_ITEM_L2,
        SFC_FLOW_ITEM_L3,
+       SFC_FLOW_ITEM_L4,
 };
 
 typedef int (sfc_flow_item_parse)(const struct rte_flow_item *item,
@@ -73,6 +58,42 @@ static sfc_flow_item_parse sfc_flow_parse_void;
 static sfc_flow_item_parse sfc_flow_parse_eth;
 static sfc_flow_item_parse sfc_flow_parse_vlan;
 static sfc_flow_item_parse sfc_flow_parse_ipv4;
+static sfc_flow_item_parse sfc_flow_parse_ipv6;
+static sfc_flow_item_parse sfc_flow_parse_tcp;
+static sfc_flow_item_parse sfc_flow_parse_udp;
+static sfc_flow_item_parse sfc_flow_parse_vxlan;
+static sfc_flow_item_parse sfc_flow_parse_geneve;
+static sfc_flow_item_parse sfc_flow_parse_nvgre;
+
+typedef int (sfc_flow_spec_set_vals)(struct sfc_flow_spec *spec,
+                                    unsigned int filters_count_for_one_val,
+                                    struct rte_flow_error *error);
+
+typedef boolean_t (sfc_flow_spec_check)(efx_filter_match_flags_t match,
+                                       efx_filter_spec_t *spec,
+                                       struct sfc_filter *filter);
+
+struct sfc_flow_copy_flag {
+       /* EFX filter specification match flag */
+       efx_filter_match_flags_t flag;
+       /* Number of values of corresponding field */
+       unsigned int vals_count;
+       /* Function to set values in specifications */
+       sfc_flow_spec_set_vals *set_vals;
+       /*
+        * Function to check that the specification is suitable
+        * for adding this match flag
+        */
+       sfc_flow_spec_check *spec_check;
+};
+
+static sfc_flow_spec_set_vals sfc_flow_set_unknown_dst_flags;
+static sfc_flow_spec_check sfc_flow_check_unknown_dst_flags;
+static sfc_flow_spec_set_vals sfc_flow_set_ethertypes;
+static sfc_flow_spec_set_vals sfc_flow_set_ifrm_unknown_dst_flags;
+static sfc_flow_spec_check sfc_flow_check_ifrm_unknown_dst_flags;
+static sfc_flow_spec_set_vals sfc_flow_set_outer_vid_flag;
+static sfc_flow_spec_check sfc_flow_check_outer_vid_flag;
 
 static boolean_t
 sfc_flow_is_zero(const uint8_t *buf, unsigned int size)
@@ -101,7 +122,6 @@ sfc_flow_parse_init(const struct rte_flow_item *item,
        const uint8_t *spec;
        const uint8_t *mask;
        const uint8_t *last;
-       uint8_t match;
        uint8_t supp;
        unsigned int i;
 
@@ -131,13 +151,13 @@ sfc_flow_parse_init(const struct rte_flow_item *item,
                        return -rte_errno;
                }
 
-               mask = (const uint8_t *)def_mask;
+               mask = def_mask;
        } else {
-               mask = (const uint8_t *)item->mask;
+               mask = item->mask;
        }
 
-       spec = (const uint8_t *)item->spec;
-       last = (const uint8_t *)item->last;
+       spec = item->spec;
+       last = item->last;
 
        if (spec == NULL)
                goto exit;
@@ -162,12 +182,11 @@ sfc_flow_parse_init(const struct rte_flow_item *item,
                return -rte_errno;
        }
 
-       /* Check that mask and spec not asks for more match than supp_mask */
+       /* Check that mask does not ask for more match than supp_mask */
        for (i = 0; i < size; i++) {
-               match = spec[i] | mask[i];
                supp = ((const uint8_t *)supp_mask)[i];
 
-               if ((match | supp) != supp) {
+               if (~supp & mask[i]) {
                        rte_flow_error_set(error, ENOTSUP,
                                           RTE_FLOW_ERROR_TYPE_ITEM, item,
                                           "Item's field is not supported");
@@ -200,9 +219,11 @@ sfc_flow_parse_void(__rte_unused const struct rte_flow_item *item,
  * Convert Ethernet item to EFX filter specification.
  *
  * @param item[in]
- *   Item specification. Only source and destination addresses and
- *   Ethernet type fields are supported. If the mask is NULL, default
- *   mask will be used. Ranging is not supported.
+ *   Item specification. Outer frame specification may only comprise
+ *   source/destination addresses and Ethertype field.
+ *   Inner frame specification may contain destination address only.
+ *   There is support for individual/group mask as well as for empty and full.
+ *   If the mask is NULL, default mask will be used. Ranging is not supported.
  * @param efx_spec[in, out]
  *   EFX filter specification to update.
  * @param[out] error
@@ -221,12 +242,32 @@ sfc_flow_parse_eth(const struct rte_flow_item *item,
                .src.addr_bytes = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff },
                .type = 0xffff,
        };
+       const struct rte_flow_item_eth ifrm_supp_mask = {
+               .dst.addr_bytes = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff },
+       };
+       const uint8_t ig_mask[EFX_MAC_ADDR_LEN] = {
+               0x01, 0x00, 0x00, 0x00, 0x00, 0x00
+       };
+       const struct rte_flow_item_eth *supp_mask_p;
+       const struct rte_flow_item_eth *def_mask_p;
+       uint8_t *loc_mac = NULL;
+       boolean_t is_ifrm = (efx_spec->efs_encap_type !=
+               EFX_TUNNEL_PROTOCOL_NONE);
+
+       if (is_ifrm) {
+               supp_mask_p = &ifrm_supp_mask;
+               def_mask_p = &ifrm_supp_mask;
+               loc_mac = efx_spec->efs_ifrm_loc_mac;
+       } else {
+               supp_mask_p = &supp_mask;
+               def_mask_p = &rte_flow_item_eth_mask;
+               loc_mac = efx_spec->efs_loc_mac;
+       }
 
        rc = sfc_flow_parse_init(item,
                                 (const void **)&spec,
                                 (const void **)&mask,
-                                &supp_mask,
-                                &rte_flow_item_eth_mask,
+                                supp_mask_p, def_mask_p,
                                 sizeof(struct rte_flow_item_eth),
                                 error);
        if (rc != 0)
@@ -236,19 +277,36 @@ sfc_flow_parse_eth(const struct rte_flow_item *item,
        if (spec == NULL)
                return 0;
 
-       if (is_same_ether_addr(&mask->dst, &supp_mask.dst)) {
-               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_LOC_MAC;
-               rte_memcpy(efx_spec->efs_loc_mac, spec->dst.addr_bytes,
+       if (rte_is_same_ether_addr(&mask->dst, &supp_mask.dst)) {
+               efx_spec->efs_match_flags |= is_ifrm ?
+                       EFX_FILTER_MATCH_IFRM_LOC_MAC :
+                       EFX_FILTER_MATCH_LOC_MAC;
+               rte_memcpy(loc_mac, spec->dst.addr_bytes,
                           EFX_MAC_ADDR_LEN);
-       } else if (!is_zero_ether_addr(&mask->dst)) {
+       } else if (memcmp(mask->dst.addr_bytes, ig_mask,
+                         EFX_MAC_ADDR_LEN) == 0) {
+               if (rte_is_unicast_ether_addr(&spec->dst))
+                       efx_spec->efs_match_flags |= is_ifrm ?
+                               EFX_FILTER_MATCH_IFRM_UNKNOWN_UCAST_DST :
+                               EFX_FILTER_MATCH_UNKNOWN_UCAST_DST;
+               else
+                       efx_spec->efs_match_flags |= is_ifrm ?
+                               EFX_FILTER_MATCH_IFRM_UNKNOWN_MCAST_DST :
+                               EFX_FILTER_MATCH_UNKNOWN_MCAST_DST;
+       } else if (!rte_is_zero_ether_addr(&mask->dst)) {
                goto fail_bad_mask;
        }
 
-       if (is_same_ether_addr(&mask->src, &supp_mask.src)) {
+       /*
+        * ifrm_supp_mask ensures that the source address and
+        * ethertype masks are equal to zero in inner frame,
+        * so these fields are filled in only for the outer frame
+        */
+       if (rte_is_same_ether_addr(&mask->src, &supp_mask.src)) {
                efx_spec->efs_match_flags |= EFX_FILTER_MATCH_REM_MAC;
                rte_memcpy(efx_spec->efs_rem_mac, spec->src.addr_bytes,
                           EFX_MAC_ADDR_LEN);
-       } else if (!is_zero_ether_addr(&mask->src)) {
+       } else if (!rte_is_zero_ether_addr(&mask->src)) {
                goto fail_bad_mask;
        }
 
@@ -294,6 +352,7 @@ sfc_flow_parse_vlan(const struct rte_flow_item *item,
        const struct rte_flow_item_vlan *mask = NULL;
        const struct rte_flow_item_vlan supp_mask = {
                .tci = rte_cpu_to_be_16(ETH_VLAN_ID_MAX),
+               .inner_type = RTE_BE16(0xffff),
        };
 
        rc = sfc_flow_parse_init(item,
@@ -313,7 +372,8 @@ sfc_flow_parse_vlan(const struct rte_flow_item *item,
         * the outer tag and the next matches the inner tag.
         */
        if (mask->tci == supp_mask.tci) {
-               vid = rte_bswap16(spec->tci);
+               /* Apply mask to keep VID only */
+               vid = rte_bswap16(spec->tci & mask->tci);
 
                if (!(efx_spec->efs_match_flags &
                      EFX_FILTER_MATCH_OUTER_VID)) {
@@ -336,6 +396,22 @@ sfc_flow_parse_vlan(const struct rte_flow_item *item,
                return -rte_errno;
        }
 
+       if (efx_spec->efs_match_flags & EFX_FILTER_MATCH_ETHER_TYPE) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ITEM, item,
+                                  "VLAN TPID matching is not supported");
+               return -rte_errno;
+       }
+       if (mask->inner_type == supp_mask.inner_type) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_ETHER_TYPE;
+               efx_spec->efs_ether_type = rte_bswap16(spec->inner_type);
+       } else if (mask->inner_type) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ITEM, item,
+                                  "Bad mask for VLAN inner_type");
+               return -rte_errno;
+       }
+
        return 0;
 }
 
@@ -429,186 +505,1060 @@ fail_bad_mask:
        return -rte_errno;
 }
 
-static const struct sfc_flow_item sfc_flow_items[] = {
-       {
-               .type = RTE_FLOW_ITEM_TYPE_VOID,
-               .prev_layer = SFC_FLOW_ITEM_ANY_LAYER,
-               .layer = SFC_FLOW_ITEM_ANY_LAYER,
-               .parse = sfc_flow_parse_void,
-       },
-       {
-               .type = RTE_FLOW_ITEM_TYPE_ETH,
-               .prev_layer = SFC_FLOW_ITEM_START_LAYER,
-               .layer = SFC_FLOW_ITEM_L2,
-               .parse = sfc_flow_parse_eth,
-       },
-       {
-               .type = RTE_FLOW_ITEM_TYPE_VLAN,
-               .prev_layer = SFC_FLOW_ITEM_L2,
-               .layer = SFC_FLOW_ITEM_L2,
-               .parse = sfc_flow_parse_vlan,
-       },
-       {
-               .type = RTE_FLOW_ITEM_TYPE_IPV4,
-               .prev_layer = SFC_FLOW_ITEM_L2,
-               .layer = SFC_FLOW_ITEM_L3,
-               .parse = sfc_flow_parse_ipv4,
-       },
-};
-
-/*
- * Protocol-independent flow API support
+/**
+ * Convert IPv6 item to EFX filter specification.
+ *
+ * @param item[in]
+ *   Item specification. Only source and destination addresses and
+ *   next header fields are supported. If the mask is NULL, default
+ *   mask will be used. Ranging is not supported.
+ * @param efx_spec[in, out]
+ *   EFX filter specification to update.
+ * @param[out] error
+ *   Perform verbose error reporting if not NULL.
  */
 static int
-sfc_flow_parse_attr(const struct rte_flow_attr *attr,
-                   struct rte_flow *flow,
+sfc_flow_parse_ipv6(const struct rte_flow_item *item,
+                   efx_filter_spec_t *efx_spec,
                    struct rte_flow_error *error)
 {
-       if (attr == NULL) {
+       int rc;
+       const struct rte_flow_item_ipv6 *spec = NULL;
+       const struct rte_flow_item_ipv6 *mask = NULL;
+       const uint16_t ether_type_ipv6 = rte_cpu_to_le_16(EFX_ETHER_TYPE_IPV6);
+       const struct rte_flow_item_ipv6 supp_mask = {
+               .hdr = {
+                       .src_addr = { 0xff, 0xff, 0xff, 0xff,
+                                     0xff, 0xff, 0xff, 0xff,
+                                     0xff, 0xff, 0xff, 0xff,
+                                     0xff, 0xff, 0xff, 0xff },
+                       .dst_addr = { 0xff, 0xff, 0xff, 0xff,
+                                     0xff, 0xff, 0xff, 0xff,
+                                     0xff, 0xff, 0xff, 0xff,
+                                     0xff, 0xff, 0xff, 0xff },
+                       .proto = 0xff,
+               }
+       };
+
+       rc = sfc_flow_parse_init(item,
+                                (const void **)&spec,
+                                (const void **)&mask,
+                                &supp_mask,
+                                &rte_flow_item_ipv6_mask,
+                                sizeof(struct rte_flow_item_ipv6),
+                                error);
+       if (rc != 0)
+               return rc;
+
+       /*
+        * Filtering by IPv6 source and destination addresses requires
+        * the appropriate ETHER_TYPE in hardware filters
+        */
+       if (!(efx_spec->efs_match_flags & EFX_FILTER_MATCH_ETHER_TYPE)) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_ETHER_TYPE;
+               efx_spec->efs_ether_type = ether_type_ipv6;
+       } else if (efx_spec->efs_ether_type != ether_type_ipv6) {
                rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ATTR, NULL,
-                                  "NULL attribute");
-               return -rte_errno;
-       }
-       if (attr->group != 0) {
-               rte_flow_error_set(error, ENOTSUP,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_GROUP, attr,
-                                  "Groups are not supported");
-               return -rte_errno;
-       }
-       if (attr->priority != 0) {
-               rte_flow_error_set(error, ENOTSUP,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY, attr,
-                                  "Priorities are not supported");
-               return -rte_errno;
-       }
-       if (attr->egress != 0) {
-               rte_flow_error_set(error, ENOTSUP,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, attr,
-                                  "Egress is not supported");
+                       RTE_FLOW_ERROR_TYPE_ITEM, item,
+                       "Ethertype in pattern with IPV6 item should be appropriate");
                return -rte_errno;
        }
-       if (attr->ingress == 0) {
-               rte_flow_error_set(error, ENOTSUP,
-                                  RTE_FLOW_ERROR_TYPE_ATTR_INGRESS, attr,
-                                  "Only ingress is supported");
-               return -rte_errno;
+
+       if (spec == NULL)
+               return 0;
+
+       /*
+        * IPv6 addresses are in big-endian byte order in item and in
+        * efx_spec
+        */
+       if (memcmp(mask->hdr.src_addr, supp_mask.hdr.src_addr,
+                  sizeof(mask->hdr.src_addr)) == 0) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_REM_HOST;
+
+               RTE_BUILD_BUG_ON(sizeof(efx_spec->efs_rem_host) !=
+                                sizeof(spec->hdr.src_addr));
+               rte_memcpy(&efx_spec->efs_rem_host, spec->hdr.src_addr,
+                          sizeof(efx_spec->efs_rem_host));
+       } else if (!sfc_flow_is_zero(mask->hdr.src_addr,
+                                    sizeof(mask->hdr.src_addr))) {
+               goto fail_bad_mask;
        }
 
-       flow->spec.efs_flags |= EFX_FILTER_FLAG_RX;
-       flow->spec.efs_rss_context = EFX_FILTER_SPEC_RSS_CONTEXT_DEFAULT;
+       if (memcmp(mask->hdr.dst_addr, supp_mask.hdr.dst_addr,
+                  sizeof(mask->hdr.dst_addr)) == 0) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_LOC_HOST;
 
-       return 0;
-}
+               RTE_BUILD_BUG_ON(sizeof(efx_spec->efs_loc_host) !=
+                                sizeof(spec->hdr.dst_addr));
+               rte_memcpy(&efx_spec->efs_loc_host, spec->hdr.dst_addr,
+                          sizeof(efx_spec->efs_loc_host));
+       } else if (!sfc_flow_is_zero(mask->hdr.dst_addr,
+                                    sizeof(mask->hdr.dst_addr))) {
+               goto fail_bad_mask;
+       }
 
-/* Get item from array sfc_flow_items */
-static const struct sfc_flow_item *
-sfc_flow_get_item(enum rte_flow_item_type type)
-{
-       unsigned int i;
+       if (mask->hdr.proto == supp_mask.hdr.proto) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_IP_PROTO;
+               efx_spec->efs_ip_proto = spec->hdr.proto;
+       } else if (mask->hdr.proto != 0) {
+               goto fail_bad_mask;
+       }
 
-       for (i = 0; i < RTE_DIM(sfc_flow_items); i++)
-               if (sfc_flow_items[i].type == type)
-                       return &sfc_flow_items[i];
+       return 0;
 
-       return NULL;
+fail_bad_mask:
+       rte_flow_error_set(error, EINVAL,
+                          RTE_FLOW_ERROR_TYPE_ITEM, item,
+                          "Bad mask in the IPV6 pattern item");
+       return -rte_errno;
 }
 
+/**
+ * Convert TCP item to EFX filter specification.
+ *
+ * @param item[in]
+ *   Item specification. Only source and destination ports fields
+ *   are supported. If the mask is NULL, default mask will be used.
+ *   Ranging is not supported.
+ * @param efx_spec[in, out]
+ *   EFX filter specification to update.
+ * @param[out] error
+ *   Perform verbose error reporting if not NULL.
+ */
 static int
-sfc_flow_parse_pattern(const struct rte_flow_item pattern[],
-                      struct rte_flow *flow,
-                      struct rte_flow_error *error)
+sfc_flow_parse_tcp(const struct rte_flow_item *item,
+                  efx_filter_spec_t *efx_spec,
+                  struct rte_flow_error *error)
 {
        int rc;
-       unsigned int prev_layer = SFC_FLOW_ITEM_ANY_LAYER;
-       const struct sfc_flow_item *item;
+       const struct rte_flow_item_tcp *spec = NULL;
+       const struct rte_flow_item_tcp *mask = NULL;
+       const struct rte_flow_item_tcp supp_mask = {
+               .hdr = {
+                       .src_port = 0xffff,
+                       .dst_port = 0xffff,
+               }
+       };
 
-       if (pattern == NULL) {
+       rc = sfc_flow_parse_init(item,
+                                (const void **)&spec,
+                                (const void **)&mask,
+                                &supp_mask,
+                                &rte_flow_item_tcp_mask,
+                                sizeof(struct rte_flow_item_tcp),
+                                error);
+       if (rc != 0)
+               return rc;
+
+       /*
+        * Filtering by TCP source and destination ports requires
+        * the appropriate IP_PROTO in hardware filters
+        */
+       if (!(efx_spec->efs_match_flags & EFX_FILTER_MATCH_IP_PROTO)) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_IP_PROTO;
+               efx_spec->efs_ip_proto = EFX_IPPROTO_TCP;
+       } else if (efx_spec->efs_ip_proto != EFX_IPPROTO_TCP) {
                rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ITEM_NUM, NULL,
-                                  "NULL pattern");
+                       RTE_FLOW_ERROR_TYPE_ITEM, item,
+                       "IP proto in pattern with TCP item should be appropriate");
                return -rte_errno;
        }
 
-       for (; pattern != NULL &&
-              pattern->type != RTE_FLOW_ITEM_TYPE_END; pattern++) {
-               item = sfc_flow_get_item(pattern->type);
-               if (item == NULL) {
-                       rte_flow_error_set(error, ENOTSUP,
-                                          RTE_FLOW_ERROR_TYPE_ITEM, pattern,
-                                          "Unsupported pattern item");
-                       return -rte_errno;
-               }
-
-               /*
-                * Omitting one or several protocol layers at the beginning
-                * of pattern is supported
-                */
-               if (item->prev_layer != SFC_FLOW_ITEM_ANY_LAYER &&
-                   prev_layer != SFC_FLOW_ITEM_ANY_LAYER &&
-                   item->prev_layer != prev_layer) {
-                       rte_flow_error_set(error, ENOTSUP,
-                                          RTE_FLOW_ERROR_TYPE_ITEM, pattern,
-                                          "Unexpected sequence of pattern items");
-                       return -rte_errno;
-               }
-
-               rc = item->parse(pattern, &flow->spec, error);
-               if (rc != 0)
-                       return rc;
+       if (spec == NULL)
+               return 0;
 
-               if (item->layer != SFC_FLOW_ITEM_ANY_LAYER)
-                       prev_layer = item->layer;
+       /*
+        * Source and destination ports are in big-endian byte order in item and
+        * in little-endian in efx_spec, so byte swap is used
+        */
+       if (mask->hdr.src_port == supp_mask.hdr.src_port) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_REM_PORT;
+               efx_spec->efs_rem_port = rte_bswap16(spec->hdr.src_port);
+       } else if (mask->hdr.src_port != 0) {
+               goto fail_bad_mask;
        }
 
-       if (pattern == NULL) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ITEM, NULL,
-                                  "NULL item");
-               return -rte_errno;
+       if (mask->hdr.dst_port == supp_mask.hdr.dst_port) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_LOC_PORT;
+               efx_spec->efs_loc_port = rte_bswap16(spec->hdr.dst_port);
+       } else if (mask->hdr.dst_port != 0) {
+               goto fail_bad_mask;
        }
 
        return 0;
-}
-
-static int
-sfc_flow_parse_queue(struct sfc_adapter *sa,
-                    const struct rte_flow_action_queue *queue,
-                    struct rte_flow *flow)
-{
-       struct sfc_rxq *rxq;
 
-       if (queue->index >= sa->rxq_count)
-               return -EINVAL;
-
-       rxq = sa->rxq_info[queue->index].rxq;
-       flow->spec.efs_dmaq_id = (uint16_t)rxq->hw_index;
-
-       return 0;
+fail_bad_mask:
+       rte_flow_error_set(error, EINVAL,
+                          RTE_FLOW_ERROR_TYPE_ITEM, item,
+                          "Bad mask in the TCP pattern item");
+       return -rte_errno;
 }
 
+/**
+ * Convert UDP item to EFX filter specification.
+ *
+ * @param item[in]
+ *   Item specification. Only source and destination ports fields
+ *   are supported. If the mask is NULL, default mask will be used.
+ *   Ranging is not supported.
+ * @param efx_spec[in, out]
+ *   EFX filter specification to update.
+ * @param[out] error
+ *   Perform verbose error reporting if not NULL.
+ */
 static int
-sfc_flow_parse_actions(struct sfc_adapter *sa,
-                      const struct rte_flow_action actions[],
-                      struct rte_flow *flow,
-                      struct rte_flow_error *error)
+sfc_flow_parse_udp(const struct rte_flow_item *item,
+                  efx_filter_spec_t *efx_spec,
+                  struct rte_flow_error *error)
 {
        int rc;
-       boolean_t is_specified = B_FALSE;
-
-       if (actions == NULL) {
-               rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION_NUM, NULL,
-                                  "NULL actions");
-               return -rte_errno;
-       }
+       const struct rte_flow_item_udp *spec = NULL;
+       const struct rte_flow_item_udp *mask = NULL;
+       const struct rte_flow_item_udp supp_mask = {
+               .hdr = {
+                       .src_port = 0xffff,
+                       .dst_port = 0xffff,
+               }
+       };
+
+       rc = sfc_flow_parse_init(item,
+                                (const void **)&spec,
+                                (const void **)&mask,
+                                &supp_mask,
+                                &rte_flow_item_udp_mask,
+                                sizeof(struct rte_flow_item_udp),
+                                error);
+       if (rc != 0)
+               return rc;
+
+       /*
+        * Filtering by UDP source and destination ports requires
+        * the appropriate IP_PROTO in hardware filters
+        */
+       if (!(efx_spec->efs_match_flags & EFX_FILTER_MATCH_IP_PROTO)) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_IP_PROTO;
+               efx_spec->efs_ip_proto = EFX_IPPROTO_UDP;
+       } else if (efx_spec->efs_ip_proto != EFX_IPPROTO_UDP) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ITEM, item,
+                       "IP proto in pattern with UDP item should be appropriate");
+               return -rte_errno;
+       }
+
+       if (spec == NULL)
+               return 0;
+
+       /*
+        * Source and destination ports are in big-endian byte order in item and
+        * in little-endian in efx_spec, so byte swap is used
+        */
+       if (mask->hdr.src_port == supp_mask.hdr.src_port) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_REM_PORT;
+               efx_spec->efs_rem_port = rte_bswap16(spec->hdr.src_port);
+       } else if (mask->hdr.src_port != 0) {
+               goto fail_bad_mask;
+       }
+
+       if (mask->hdr.dst_port == supp_mask.hdr.dst_port) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_LOC_PORT;
+               efx_spec->efs_loc_port = rte_bswap16(spec->hdr.dst_port);
+       } else if (mask->hdr.dst_port != 0) {
+               goto fail_bad_mask;
+       }
+
+       return 0;
+
+fail_bad_mask:
+       rte_flow_error_set(error, EINVAL,
+                          RTE_FLOW_ERROR_TYPE_ITEM, item,
+                          "Bad mask in the UDP pattern item");
+       return -rte_errno;
+}
+
+/*
+ * Filters for encapsulated packets match based on the EtherType and IP
+ * protocol in the outer frame.
+ */
+static int
+sfc_flow_set_match_flags_for_encap_pkts(const struct rte_flow_item *item,
+                                       efx_filter_spec_t *efx_spec,
+                                       uint8_t ip_proto,
+                                       struct rte_flow_error *error)
+{
+       if (!(efx_spec->efs_match_flags & EFX_FILTER_MATCH_IP_PROTO)) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_IP_PROTO;
+               efx_spec->efs_ip_proto = ip_proto;
+       } else if (efx_spec->efs_ip_proto != ip_proto) {
+               switch (ip_proto) {
+               case EFX_IPPROTO_UDP:
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM, item,
+                               "Outer IP header protocol must be UDP "
+                               "in VxLAN/GENEVE pattern");
+                       return -rte_errno;
+
+               case EFX_IPPROTO_GRE:
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM, item,
+                               "Outer IP header protocol must be GRE "
+                               "in NVGRE pattern");
+                       return -rte_errno;
+
+               default:
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM, item,
+                               "Only VxLAN/GENEVE/NVGRE tunneling patterns "
+                               "are supported");
+                       return -rte_errno;
+               }
+       }
+
+       if (efx_spec->efs_match_flags & EFX_FILTER_MATCH_ETHER_TYPE &&
+           efx_spec->efs_ether_type != EFX_ETHER_TYPE_IPV4 &&
+           efx_spec->efs_ether_type != EFX_ETHER_TYPE_IPV6) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ITEM, item,
+                       "Outer frame EtherType in pattern with tunneling "
+                       "must be IPv4 or IPv6");
+               return -rte_errno;
+       }
+
+       return 0;
+}
+
+static int
+sfc_flow_set_efx_spec_vni_or_vsid(efx_filter_spec_t *efx_spec,
+                                 const uint8_t *vni_or_vsid_val,
+                                 const uint8_t *vni_or_vsid_mask,
+                                 const struct rte_flow_item *item,
+                                 struct rte_flow_error *error)
+{
+       const uint8_t vni_or_vsid_full_mask[EFX_VNI_OR_VSID_LEN] = {
+               0xff, 0xff, 0xff
+       };
+
+       if (memcmp(vni_or_vsid_mask, vni_or_vsid_full_mask,
+                  EFX_VNI_OR_VSID_LEN) == 0) {
+               efx_spec->efs_match_flags |= EFX_FILTER_MATCH_VNI_OR_VSID;
+               rte_memcpy(efx_spec->efs_vni_or_vsid, vni_or_vsid_val,
+                          EFX_VNI_OR_VSID_LEN);
+       } else if (!sfc_flow_is_zero(vni_or_vsid_mask, EFX_VNI_OR_VSID_LEN)) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ITEM, item,
+                                  "Unsupported VNI/VSID mask");
+               return -rte_errno;
+       }
+
+       return 0;
+}
+
+/**
+ * Convert VXLAN item to EFX filter specification.
+ *
+ * @param item[in]
+ *   Item specification. Only VXLAN network identifier field is supported.
+ *   If the mask is NULL, default mask will be used.
+ *   Ranging is not supported.
+ * @param efx_spec[in, out]
+ *   EFX filter specification to update.
+ * @param[out] error
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_parse_vxlan(const struct rte_flow_item *item,
+                    efx_filter_spec_t *efx_spec,
+                    struct rte_flow_error *error)
+{
+       int rc;
+       const struct rte_flow_item_vxlan *spec = NULL;
+       const struct rte_flow_item_vxlan *mask = NULL;
+       const struct rte_flow_item_vxlan supp_mask = {
+               .vni = { 0xff, 0xff, 0xff }
+       };
+
+       rc = sfc_flow_parse_init(item,
+                                (const void **)&spec,
+                                (const void **)&mask,
+                                &supp_mask,
+                                &rte_flow_item_vxlan_mask,
+                                sizeof(struct rte_flow_item_vxlan),
+                                error);
+       if (rc != 0)
+               return rc;
+
+       rc = sfc_flow_set_match_flags_for_encap_pkts(item, efx_spec,
+                                                    EFX_IPPROTO_UDP, error);
+       if (rc != 0)
+               return rc;
+
+       efx_spec->efs_encap_type = EFX_TUNNEL_PROTOCOL_VXLAN;
+       efx_spec->efs_match_flags |= EFX_FILTER_MATCH_ENCAP_TYPE;
+
+       if (spec == NULL)
+               return 0;
+
+       rc = sfc_flow_set_efx_spec_vni_or_vsid(efx_spec, spec->vni,
+                                              mask->vni, item, error);
+
+       return rc;
+}
+
+/**
+ * Convert GENEVE item to EFX filter specification.
+ *
+ * @param item[in]
+ *   Item specification. Only Virtual Network Identifier and protocol type
+ *   fields are supported. But protocol type can be only Ethernet (0x6558).
+ *   If the mask is NULL, default mask will be used.
+ *   Ranging is not supported.
+ * @param efx_spec[in, out]
+ *   EFX filter specification to update.
+ * @param[out] error
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_parse_geneve(const struct rte_flow_item *item,
+                     efx_filter_spec_t *efx_spec,
+                     struct rte_flow_error *error)
+{
+       int rc;
+       const struct rte_flow_item_geneve *spec = NULL;
+       const struct rte_flow_item_geneve *mask = NULL;
+       const struct rte_flow_item_geneve supp_mask = {
+               .protocol = RTE_BE16(0xffff),
+               .vni = { 0xff, 0xff, 0xff }
+       };
+
+       rc = sfc_flow_parse_init(item,
+                                (const void **)&spec,
+                                (const void **)&mask,
+                                &supp_mask,
+                                &rte_flow_item_geneve_mask,
+                                sizeof(struct rte_flow_item_geneve),
+                                error);
+       if (rc != 0)
+               return rc;
+
+       rc = sfc_flow_set_match_flags_for_encap_pkts(item, efx_spec,
+                                                    EFX_IPPROTO_UDP, error);
+       if (rc != 0)
+               return rc;
+
+       efx_spec->efs_encap_type = EFX_TUNNEL_PROTOCOL_GENEVE;
+       efx_spec->efs_match_flags |= EFX_FILTER_MATCH_ENCAP_TYPE;
+
+       if (spec == NULL)
+               return 0;
+
+       if (mask->protocol == supp_mask.protocol) {
+               if (spec->protocol != rte_cpu_to_be_16(RTE_ETHER_TYPE_TEB)) {
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM, item,
+                               "GENEVE encap. protocol must be Ethernet "
+                               "(0x6558) in the GENEVE pattern item");
+                       return -rte_errno;
+               }
+       } else if (mask->protocol != 0) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ITEM, item,
+                       "Unsupported mask for GENEVE encap. protocol");
+               return -rte_errno;
+       }
+
+       rc = sfc_flow_set_efx_spec_vni_or_vsid(efx_spec, spec->vni,
+                                              mask->vni, item, error);
+
+       return rc;
+}
+
+/**
+ * Convert NVGRE item to EFX filter specification.
+ *
+ * @param item[in]
+ *   Item specification. Only virtual subnet ID field is supported.
+ *   If the mask is NULL, default mask will be used.
+ *   Ranging is not supported.
+ * @param efx_spec[in, out]
+ *   EFX filter specification to update.
+ * @param[out] error
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_parse_nvgre(const struct rte_flow_item *item,
+                    efx_filter_spec_t *efx_spec,
+                    struct rte_flow_error *error)
+{
+       int rc;
+       const struct rte_flow_item_nvgre *spec = NULL;
+       const struct rte_flow_item_nvgre *mask = NULL;
+       const struct rte_flow_item_nvgre supp_mask = {
+               .tni = { 0xff, 0xff, 0xff }
+       };
+
+       rc = sfc_flow_parse_init(item,
+                                (const void **)&spec,
+                                (const void **)&mask,
+                                &supp_mask,
+                                &rte_flow_item_nvgre_mask,
+                                sizeof(struct rte_flow_item_nvgre),
+                                error);
+       if (rc != 0)
+               return rc;
+
+       rc = sfc_flow_set_match_flags_for_encap_pkts(item, efx_spec,
+                                                    EFX_IPPROTO_GRE, error);
+       if (rc != 0)
+               return rc;
+
+       efx_spec->efs_encap_type = EFX_TUNNEL_PROTOCOL_NVGRE;
+       efx_spec->efs_match_flags |= EFX_FILTER_MATCH_ENCAP_TYPE;
+
+       if (spec == NULL)
+               return 0;
+
+       rc = sfc_flow_set_efx_spec_vni_or_vsid(efx_spec, spec->tni,
+                                              mask->tni, item, error);
+
+       return rc;
+}
+
+static const struct sfc_flow_item sfc_flow_items[] = {
+       {
+               .type = RTE_FLOW_ITEM_TYPE_VOID,
+               .prev_layer = SFC_FLOW_ITEM_ANY_LAYER,
+               .layer = SFC_FLOW_ITEM_ANY_LAYER,
+               .parse = sfc_flow_parse_void,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_ETH,
+               .prev_layer = SFC_FLOW_ITEM_START_LAYER,
+               .layer = SFC_FLOW_ITEM_L2,
+               .parse = sfc_flow_parse_eth,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_VLAN,
+               .prev_layer = SFC_FLOW_ITEM_L2,
+               .layer = SFC_FLOW_ITEM_L2,
+               .parse = sfc_flow_parse_vlan,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_IPV4,
+               .prev_layer = SFC_FLOW_ITEM_L2,
+               .layer = SFC_FLOW_ITEM_L3,
+               .parse = sfc_flow_parse_ipv4,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_IPV6,
+               .prev_layer = SFC_FLOW_ITEM_L2,
+               .layer = SFC_FLOW_ITEM_L3,
+               .parse = sfc_flow_parse_ipv6,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_TCP,
+               .prev_layer = SFC_FLOW_ITEM_L3,
+               .layer = SFC_FLOW_ITEM_L4,
+               .parse = sfc_flow_parse_tcp,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_UDP,
+               .prev_layer = SFC_FLOW_ITEM_L3,
+               .layer = SFC_FLOW_ITEM_L4,
+               .parse = sfc_flow_parse_udp,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_VXLAN,
+               .prev_layer = SFC_FLOW_ITEM_L4,
+               .layer = SFC_FLOW_ITEM_START_LAYER,
+               .parse = sfc_flow_parse_vxlan,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_GENEVE,
+               .prev_layer = SFC_FLOW_ITEM_L4,
+               .layer = SFC_FLOW_ITEM_START_LAYER,
+               .parse = sfc_flow_parse_geneve,
+       },
+       {
+               .type = RTE_FLOW_ITEM_TYPE_NVGRE,
+               .prev_layer = SFC_FLOW_ITEM_L3,
+               .layer = SFC_FLOW_ITEM_START_LAYER,
+               .parse = sfc_flow_parse_nvgre,
+       },
+};
+
+/*
+ * Protocol-independent flow API support
+ */
+static int
+sfc_flow_parse_attr(const struct rte_flow_attr *attr,
+                   struct rte_flow *flow,
+                   struct rte_flow_error *error)
+{
+       if (attr == NULL) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR, NULL,
+                                  "NULL attribute");
+               return -rte_errno;
+       }
+       if (attr->group != 0) {
+               rte_flow_error_set(error, ENOTSUP,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_GROUP, attr,
+                                  "Groups are not supported");
+               return -rte_errno;
+       }
+       if (attr->priority != 0) {
+               rte_flow_error_set(error, ENOTSUP,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY, attr,
+                                  "Priorities are not supported");
+               return -rte_errno;
+       }
+       if (attr->egress != 0) {
+               rte_flow_error_set(error, ENOTSUP,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_EGRESS, attr,
+                                  "Egress is not supported");
+               return -rte_errno;
+       }
+       if (attr->transfer != 0) {
+               rte_flow_error_set(error, ENOTSUP,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER, attr,
+                                  "Transfer is not supported");
+               return -rte_errno;
+       }
+       if (attr->ingress == 0) {
+               rte_flow_error_set(error, ENOTSUP,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_INGRESS, attr,
+                                  "Only ingress is supported");
+               return -rte_errno;
+       }
+
+       flow->spec.template.efs_flags |= EFX_FILTER_FLAG_RX;
+       flow->spec.template.efs_rss_context = EFX_RSS_CONTEXT_DEFAULT;
+
+       return 0;
+}
+
+/* Get item from array sfc_flow_items */
+static const struct sfc_flow_item *
+sfc_flow_get_item(enum rte_flow_item_type type)
+{
+       unsigned int i;
+
+       for (i = 0; i < RTE_DIM(sfc_flow_items); i++)
+               if (sfc_flow_items[i].type == type)
+                       return &sfc_flow_items[i];
+
+       return NULL;
+}
+
+static int
+sfc_flow_parse_pattern(const struct rte_flow_item pattern[],
+                      struct rte_flow *flow,
+                      struct rte_flow_error *error)
+{
+       int rc;
+       unsigned int prev_layer = SFC_FLOW_ITEM_ANY_LAYER;
+       boolean_t is_ifrm = B_FALSE;
+       const struct sfc_flow_item *item;
+
+       if (pattern == NULL) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ITEM_NUM, NULL,
+                                  "NULL pattern");
+               return -rte_errno;
+       }
+
+       for (; pattern->type != RTE_FLOW_ITEM_TYPE_END; pattern++) {
+               item = sfc_flow_get_item(pattern->type);
+               if (item == NULL) {
+                       rte_flow_error_set(error, ENOTSUP,
+                                          RTE_FLOW_ERROR_TYPE_ITEM, pattern,
+                                          "Unsupported pattern item");
+                       return -rte_errno;
+               }
+
+               /*
+                * Omitting one or several protocol layers at the beginning
+                * of pattern is supported
+                */
+               if (item->prev_layer != SFC_FLOW_ITEM_ANY_LAYER &&
+                   prev_layer != SFC_FLOW_ITEM_ANY_LAYER &&
+                   item->prev_layer != prev_layer) {
+                       rte_flow_error_set(error, ENOTSUP,
+                                          RTE_FLOW_ERROR_TYPE_ITEM, pattern,
+                                          "Unexpected sequence of pattern items");
+                       return -rte_errno;
+               }
+
+               /*
+                * Allow only VOID and ETH pattern items in the inner frame.
+                * Also check that there is only one tunneling protocol.
+                */
+               switch (item->type) {
+               case RTE_FLOW_ITEM_TYPE_VOID:
+               case RTE_FLOW_ITEM_TYPE_ETH:
+                       break;
+
+               case RTE_FLOW_ITEM_TYPE_VXLAN:
+               case RTE_FLOW_ITEM_TYPE_GENEVE:
+               case RTE_FLOW_ITEM_TYPE_NVGRE:
+                       if (is_ifrm) {
+                               rte_flow_error_set(error, EINVAL,
+                                       RTE_FLOW_ERROR_TYPE_ITEM,
+                                       pattern,
+                                       "More than one tunneling protocol");
+                               return -rte_errno;
+                       }
+                       is_ifrm = B_TRUE;
+                       break;
+
+               default:
+                       if (is_ifrm) {
+                               rte_flow_error_set(error, EINVAL,
+                                       RTE_FLOW_ERROR_TYPE_ITEM,
+                                       pattern,
+                                       "There is an unsupported pattern item "
+                                       "in the inner frame");
+                               return -rte_errno;
+                       }
+                       break;
+               }
+
+               rc = item->parse(pattern, &flow->spec.template, error);
+               if (rc != 0)
+                       return rc;
+
+               if (item->layer != SFC_FLOW_ITEM_ANY_LAYER)
+                       prev_layer = item->layer;
+       }
+
+       return 0;
+}
+
+static int
+sfc_flow_parse_queue(struct sfc_adapter *sa,
+                    const struct rte_flow_action_queue *queue,
+                    struct rte_flow *flow)
+{
+       struct sfc_rxq *rxq;
+
+       if (queue->index >= sfc_sa2shared(sa)->rxq_count)
+               return -EINVAL;
+
+       rxq = &sa->rxq_ctrl[queue->index];
+       flow->spec.template.efs_dmaq_id = (uint16_t)rxq->hw_index;
+
+       return 0;
+}
+
+static int
+sfc_flow_parse_rss(struct sfc_adapter *sa,
+                  const struct rte_flow_action_rss *action_rss,
+                  struct rte_flow *flow)
+{
+       struct sfc_adapter_shared * const sas = sfc_sa2shared(sa);
+       struct sfc_rss *rss = &sas->rss;
+       unsigned int rxq_sw_index;
+       struct sfc_rxq *rxq;
+       unsigned int rxq_hw_index_min;
+       unsigned int rxq_hw_index_max;
+       efx_rx_hash_type_t efx_hash_types;
+       const uint8_t *rss_key;
+       struct sfc_flow_rss *sfc_rss_conf = &flow->rss_conf;
+       unsigned int i;
+
+       if (action_rss->queue_num == 0)
+               return -EINVAL;
+
+       rxq_sw_index = sfc_sa2shared(sa)->rxq_count - 1;
+       rxq = &sa->rxq_ctrl[rxq_sw_index];
+       rxq_hw_index_min = rxq->hw_index;
+       rxq_hw_index_max = 0;
+
+       for (i = 0; i < action_rss->queue_num; ++i) {
+               rxq_sw_index = action_rss->queue[i];
+
+               if (rxq_sw_index >= sfc_sa2shared(sa)->rxq_count)
+                       return -EINVAL;
+
+               rxq = &sa->rxq_ctrl[rxq_sw_index];
+
+               if (rxq->hw_index < rxq_hw_index_min)
+                       rxq_hw_index_min = rxq->hw_index;
+
+               if (rxq->hw_index > rxq_hw_index_max)
+                       rxq_hw_index_max = rxq->hw_index;
+       }
+
+       switch (action_rss->func) {
+       case RTE_ETH_HASH_FUNCTION_DEFAULT:
+       case RTE_ETH_HASH_FUNCTION_TOEPLITZ:
+               break;
+       default:
+               return -EINVAL;
+       }
+
+       if (action_rss->level)
+               return -EINVAL;
+
+       /*
+        * Dummy RSS action with only one queue and no specific settings
+        * for hash types and key does not require dedicated RSS context
+        * and may be simplified to single queue action.
+        */
+       if (action_rss->queue_num == 1 && action_rss->types == 0 &&
+           action_rss->key_len == 0) {
+               flow->spec.template.efs_dmaq_id = rxq_hw_index_min;
+               return 0;
+       }
+
+       if (action_rss->types) {
+               int rc;
+
+               rc = sfc_rx_hf_rte_to_efx(sa, action_rss->types,
+                                         &efx_hash_types);
+               if (rc != 0)
+                       return -rc;
+       } else {
+               unsigned int i;
+
+               efx_hash_types = 0;
+               for (i = 0; i < rss->hf_map_nb_entries; ++i)
+                       efx_hash_types |= rss->hf_map[i].efx;
+       }
+
+       if (action_rss->key_len) {
+               if (action_rss->key_len != sizeof(rss->key))
+                       return -EINVAL;
+
+               rss_key = action_rss->key;
+       } else {
+               rss_key = rss->key;
+       }
+
+       flow->rss = B_TRUE;
+
+       sfc_rss_conf->rxq_hw_index_min = rxq_hw_index_min;
+       sfc_rss_conf->rxq_hw_index_max = rxq_hw_index_max;
+       sfc_rss_conf->rss_hash_types = efx_hash_types;
+       rte_memcpy(sfc_rss_conf->rss_key, rss_key, sizeof(rss->key));
+
+       for (i = 0; i < RTE_DIM(sfc_rss_conf->rss_tbl); ++i) {
+               unsigned int nb_queues = action_rss->queue_num;
+               unsigned int rxq_sw_index = action_rss->queue[i % nb_queues];
+               struct sfc_rxq *rxq = &sa->rxq_ctrl[rxq_sw_index];
+
+               sfc_rss_conf->rss_tbl[i] = rxq->hw_index - rxq_hw_index_min;
+       }
+
+       return 0;
+}
+
+static int
+sfc_flow_spec_flush(struct sfc_adapter *sa, struct sfc_flow_spec *spec,
+                   unsigned int filters_count)
+{
+       unsigned int i;
+       int ret = 0;
+
+       for (i = 0; i < filters_count; i++) {
+               int rc;
+
+               rc = efx_filter_remove(sa->nic, &spec->filters[i]);
+               if (ret == 0 && rc != 0) {
+                       sfc_err(sa, "failed to remove filter specification "
+                               "(rc = %d)", rc);
+                       ret = rc;
+               }
+       }
+
+       return ret;
+}
+
+static int
+sfc_flow_spec_insert(struct sfc_adapter *sa, struct sfc_flow_spec *spec)
+{
+       unsigned int i;
+       int rc = 0;
+
+       for (i = 0; i < spec->count; i++) {
+               rc = efx_filter_insert(sa->nic, &spec->filters[i]);
+               if (rc != 0) {
+                       sfc_flow_spec_flush(sa, spec, i);
+                       break;
+               }
+       }
+
+       return rc;
+}
+
+static int
+sfc_flow_spec_remove(struct sfc_adapter *sa, struct sfc_flow_spec *spec)
+{
+       return sfc_flow_spec_flush(sa, spec, spec->count);
+}
+
+static int
+sfc_flow_filter_insert(struct sfc_adapter *sa,
+                      struct rte_flow *flow)
+{
+       struct sfc_adapter_shared * const sas = sfc_sa2shared(sa);
+       struct sfc_rss *rss = &sas->rss;
+       struct sfc_flow_rss *flow_rss = &flow->rss_conf;
+       uint32_t efs_rss_context = EFX_RSS_CONTEXT_DEFAULT;
+       unsigned int i;
+       int rc = 0;
+
+       if (flow->rss) {
+               unsigned int rss_spread = MIN(flow_rss->rxq_hw_index_max -
+                                             flow_rss->rxq_hw_index_min + 1,
+                                             EFX_MAXRSS);
+
+               rc = efx_rx_scale_context_alloc(sa->nic,
+                                               EFX_RX_SCALE_EXCLUSIVE,
+                                               rss_spread,
+                                               &efs_rss_context);
+               if (rc != 0)
+                       goto fail_scale_context_alloc;
+
+               rc = efx_rx_scale_mode_set(sa->nic, efs_rss_context,
+                                          rss->hash_alg,
+                                          flow_rss->rss_hash_types, B_TRUE);
+               if (rc != 0)
+                       goto fail_scale_mode_set;
+
+               rc = efx_rx_scale_key_set(sa->nic, efs_rss_context,
+                                         flow_rss->rss_key,
+                                         sizeof(rss->key));
+               if (rc != 0)
+                       goto fail_scale_key_set;
+
+               /*
+                * At this point, fully elaborated filter specifications
+                * have been produced from the template. To make sure that
+                * RSS behaviour is consistent between them, set the same
+                * RSS context value everywhere.
+                */
+               for (i = 0; i < flow->spec.count; i++) {
+                       efx_filter_spec_t *spec = &flow->spec.filters[i];
+
+                       spec->efs_rss_context = efs_rss_context;
+                       spec->efs_dmaq_id = flow_rss->rxq_hw_index_min;
+                       spec->efs_flags |= EFX_FILTER_FLAG_RX_RSS;
+               }
+       }
+
+       rc = sfc_flow_spec_insert(sa, &flow->spec);
+       if (rc != 0)
+               goto fail_filter_insert;
+
+       if (flow->rss) {
+               /*
+                * Scale table is set after filter insertion because
+                * the table entries are relative to the base RxQ ID
+                * and the latter is submitted to the HW by means of
+                * inserting a filter, so by the time of the request
+                * the HW knows all the information needed to verify
+                * the table entries, and the operation will succeed
+                */
+               rc = efx_rx_scale_tbl_set(sa->nic, efs_rss_context,
+                                         flow_rss->rss_tbl,
+                                         RTE_DIM(flow_rss->rss_tbl));
+               if (rc != 0)
+                       goto fail_scale_tbl_set;
+       }
+
+       return 0;
+
+fail_scale_tbl_set:
+       sfc_flow_spec_remove(sa, &flow->spec);
+
+fail_filter_insert:
+fail_scale_key_set:
+fail_scale_mode_set:
+       if (efs_rss_context != EFX_RSS_CONTEXT_DEFAULT)
+               efx_rx_scale_context_free(sa->nic, efs_rss_context);
+
+fail_scale_context_alloc:
+       return rc;
+}
+
+static int
+sfc_flow_filter_remove(struct sfc_adapter *sa,
+                      struct rte_flow *flow)
+{
+       int rc = 0;
+
+       rc = sfc_flow_spec_remove(sa, &flow->spec);
+       if (rc != 0)
+               return rc;
+
+       if (flow->rss) {
+               /*
+                * All specifications for a given flow rule have the same RSS
+                * context, so that RSS context value is taken from the first
+                * filter specification
+                */
+               efx_filter_spec_t *spec = &flow->spec.filters[0];
+
+               rc = efx_rx_scale_context_free(sa->nic, spec->efs_rss_context);
+       }
+
+       return rc;
+}
+
+static int
+sfc_flow_parse_mark(struct sfc_adapter *sa,
+                   const struct rte_flow_action_mark *mark,
+                   struct rte_flow *flow)
+{
+       const efx_nic_cfg_t *encp = efx_nic_cfg_get(sa->nic);
+
+       if (mark == NULL || mark->id > encp->enc_filter_action_mark_max)
+               return EINVAL;
+
+       flow->spec.template.efs_flags |= EFX_FILTER_FLAG_ACTION_MARK;
+       flow->spec.template.efs_mark = mark->id;
+
+       return 0;
+}
+
+static int
+sfc_flow_parse_actions(struct sfc_adapter *sa,
+                      const struct rte_flow_action actions[],
+                      struct rte_flow *flow,
+                      struct rte_flow_error *error)
+{
+       int rc;
+       const unsigned int dp_rx_features = sa->priv.dp_rx->features;
+       uint32_t actions_set = 0;
+       const uint32_t fate_actions_mask = (1UL << RTE_FLOW_ACTION_TYPE_QUEUE) |
+                                          (1UL << RTE_FLOW_ACTION_TYPE_RSS) |
+                                          (1UL << RTE_FLOW_ACTION_TYPE_DROP);
+       const uint32_t mark_actions_mask = (1UL << RTE_FLOW_ACTION_TYPE_MARK) |
+                                          (1UL << RTE_FLOW_ACTION_TYPE_FLAG);
+
+       if (actions == NULL) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ACTION_NUM, NULL,
+                                  "NULL actions");
+               return -rte_errno;
+       }
+
+#define SFC_BUILD_SET_OVERFLOW(_action, _set) \
+       RTE_BUILD_BUG_ON(_action >= sizeof(_set) * CHAR_BIT)
 
        for (; actions->type != RTE_FLOW_ACTION_TYPE_END; actions++) {
                switch (actions->type) {
                case RTE_FLOW_ACTION_TYPE_VOID:
+                       SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_VOID,
+                                              actions_set);
                        break;
 
                case RTE_FLOW_ACTION_TYPE_QUEUE:
+                       SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_QUEUE,
+                                              actions_set);
+                       if ((actions_set & fate_actions_mask) != 0)
+                               goto fail_fate_actions;
+
                        rc = sfc_flow_parse_queue(sa, actions->conf, flow);
                        if (rc != 0) {
                                rte_flow_error_set(error, EINVAL,
@@ -616,8 +1566,70 @@ sfc_flow_parse_actions(struct sfc_adapter *sa,
                                        "Bad QUEUE action");
                                return -rte_errno;
                        }
+                       break;
+
+               case RTE_FLOW_ACTION_TYPE_RSS:
+                       SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_RSS,
+                                              actions_set);
+                       if ((actions_set & fate_actions_mask) != 0)
+                               goto fail_fate_actions;
+
+                       rc = sfc_flow_parse_rss(sa, actions->conf, flow);
+                       if (rc != 0) {
+                               rte_flow_error_set(error, -rc,
+                                       RTE_FLOW_ERROR_TYPE_ACTION, actions,
+                                       "Bad RSS action");
+                               return -rte_errno;
+                       }
+                       break;
+
+               case RTE_FLOW_ACTION_TYPE_DROP:
+                       SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_DROP,
+                                              actions_set);
+                       if ((actions_set & fate_actions_mask) != 0)
+                               goto fail_fate_actions;
+
+                       flow->spec.template.efs_dmaq_id =
+                               EFX_FILTER_SPEC_RX_DMAQ_ID_DROP;
+                       break;
+
+               case RTE_FLOW_ACTION_TYPE_FLAG:
+                       SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_FLAG,
+                                              actions_set);
+                       if ((actions_set & mark_actions_mask) != 0)
+                               goto fail_actions_overlap;
+
+                       if ((dp_rx_features & SFC_DP_RX_FEAT_FLOW_FLAG) == 0) {
+                               rte_flow_error_set(error, ENOTSUP,
+                                       RTE_FLOW_ERROR_TYPE_ACTION, NULL,
+                                       "FLAG action is not supported on the current Rx datapath");
+                               return -rte_errno;
+                       }
+
+                       flow->spec.template.efs_flags |=
+                               EFX_FILTER_FLAG_ACTION_FLAG;
+                       break;
+
+               case RTE_FLOW_ACTION_TYPE_MARK:
+                       SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_MARK,
+                                              actions_set);
+                       if ((actions_set & mark_actions_mask) != 0)
+                               goto fail_actions_overlap;
+
+                       if ((dp_rx_features & SFC_DP_RX_FEAT_FLOW_MARK) == 0) {
+                               rte_flow_error_set(error, ENOTSUP,
+                                       RTE_FLOW_ERROR_TYPE_ACTION, NULL,
+                                       "MARK action is not supported on the current Rx datapath");
+                               return -rte_errno;
+                       }
 
-                       is_specified = B_TRUE;
+                       rc = sfc_flow_parse_mark(sa, actions->conf, flow);
+                       if (rc != 0) {
+                               rte_flow_error_set(error, rc,
+                                       RTE_FLOW_ERROR_TYPE_ACTION, actions,
+                                       "Bad MARK action");
+                               return -rte_errno;
+                       }
                        break;
 
                default:
@@ -626,12 +1638,607 @@ sfc_flow_parse_actions(struct sfc_adapter *sa,
                                           "Action is not supported");
                        return -rte_errno;
                }
+
+               actions_set |= (1UL << actions->type);
+       }
+#undef SFC_BUILD_SET_OVERFLOW
+
+       /* When fate is unknown, drop traffic. */
+       if ((actions_set & fate_actions_mask) == 0) {
+               flow->spec.template.efs_dmaq_id =
+                       EFX_FILTER_SPEC_RX_DMAQ_ID_DROP;
+       }
+
+       return 0;
+
+fail_fate_actions:
+       rte_flow_error_set(error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION, actions,
+                          "Cannot combine several fate-deciding actions, "
+                          "choose between QUEUE, RSS or DROP");
+       return -rte_errno;
+
+fail_actions_overlap:
+       rte_flow_error_set(error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION, actions,
+                          "Overlapping actions are not supported");
+       return -rte_errno;
+}
+
+/**
+ * Set the EFX_FILTER_MATCH_UNKNOWN_UCAST_DST
+ * and EFX_FILTER_MATCH_UNKNOWN_MCAST_DST match flags in the same
+ * specifications after copying.
+ *
+ * @param spec[in, out]
+ *   SFC flow specification to update.
+ * @param filters_count_for_one_val[in]
+ *   How many specifications should have the same match flag, what is the
+ *   number of specifications before copying.
+ * @param error[out]
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_set_unknown_dst_flags(struct sfc_flow_spec *spec,
+                              unsigned int filters_count_for_one_val,
+                              struct rte_flow_error *error)
+{
+       unsigned int i;
+       static const efx_filter_match_flags_t vals[] = {
+               EFX_FILTER_MATCH_UNKNOWN_UCAST_DST,
+               EFX_FILTER_MATCH_UNKNOWN_MCAST_DST
+       };
+
+       if (filters_count_for_one_val * RTE_DIM(vals) != spec->count) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                       "Number of specifications is incorrect while copying "
+                       "by unknown destination flags");
+               return -rte_errno;
+       }
+
+       for (i = 0; i < spec->count; i++) {
+               /* The check above ensures that divisor can't be zero here */
+               spec->filters[i].efs_match_flags |=
+                       vals[i / filters_count_for_one_val];
+       }
+
+       return 0;
+}
+
+/**
+ * Check that the following conditions are met:
+ * - the list of supported filters has a filter
+ *   with EFX_FILTER_MATCH_UNKNOWN_MCAST_DST flag instead of
+ *   EFX_FILTER_MATCH_UNKNOWN_UCAST_DST, since this filter will also
+ *   be inserted.
+ *
+ * @param match[in]
+ *   The match flags of filter.
+ * @param spec[in]
+ *   Specification to be supplemented.
+ * @param filter[in]
+ *   SFC filter with list of supported filters.
+ */
+static boolean_t
+sfc_flow_check_unknown_dst_flags(efx_filter_match_flags_t match,
+                                __rte_unused efx_filter_spec_t *spec,
+                                struct sfc_filter *filter)
+{
+       unsigned int i;
+       efx_filter_match_flags_t match_mcast_dst;
+
+       match_mcast_dst =
+               (match & ~EFX_FILTER_MATCH_UNKNOWN_UCAST_DST) |
+               EFX_FILTER_MATCH_UNKNOWN_MCAST_DST;
+       for (i = 0; i < filter->supported_match_num; i++) {
+               if (match_mcast_dst == filter->supported_match[i])
+                       return B_TRUE;
+       }
+
+       return B_FALSE;
+}
+
+/**
+ * Set the EFX_FILTER_MATCH_ETHER_TYPE match flag and EFX_ETHER_TYPE_IPV4 and
+ * EFX_ETHER_TYPE_IPV6 values of the corresponding field in the same
+ * specifications after copying.
+ *
+ * @param spec[in, out]
+ *   SFC flow specification to update.
+ * @param filters_count_for_one_val[in]
+ *   How many specifications should have the same EtherType value, what is the
+ *   number of specifications before copying.
+ * @param error[out]
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_set_ethertypes(struct sfc_flow_spec *spec,
+                       unsigned int filters_count_for_one_val,
+                       struct rte_flow_error *error)
+{
+       unsigned int i;
+       static const uint16_t vals[] = {
+               EFX_ETHER_TYPE_IPV4, EFX_ETHER_TYPE_IPV6
+       };
+
+       if (filters_count_for_one_val * RTE_DIM(vals) != spec->count) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                       "Number of specifications is incorrect "
+                       "while copying by Ethertype");
+               return -rte_errno;
+       }
+
+       for (i = 0; i < spec->count; i++) {
+               spec->filters[i].efs_match_flags |=
+                       EFX_FILTER_MATCH_ETHER_TYPE;
+
+               /*
+                * The check above ensures that
+                * filters_count_for_one_val is not 0
+                */
+               spec->filters[i].efs_ether_type =
+                       vals[i / filters_count_for_one_val];
+       }
+
+       return 0;
+}
+
+/**
+ * Set the EFX_FILTER_MATCH_OUTER_VID match flag with value 0
+ * in the same specifications after copying.
+ *
+ * @param spec[in, out]
+ *   SFC flow specification to update.
+ * @param filters_count_for_one_val[in]
+ *   How many specifications should have the same match flag, what is the
+ *   number of specifications before copying.
+ * @param error[out]
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_set_outer_vid_flag(struct sfc_flow_spec *spec,
+                           unsigned int filters_count_for_one_val,
+                           struct rte_flow_error *error)
+{
+       unsigned int i;
+
+       if (filters_count_for_one_val != spec->count) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                       "Number of specifications is incorrect "
+                       "while copying by outer VLAN ID");
+               return -rte_errno;
+       }
+
+       for (i = 0; i < spec->count; i++) {
+               spec->filters[i].efs_match_flags |=
+                       EFX_FILTER_MATCH_OUTER_VID;
+
+               spec->filters[i].efs_outer_vid = 0;
+       }
+
+       return 0;
+}
+
+/**
+ * Set the EFX_FILTER_MATCH_IFRM_UNKNOWN_UCAST_DST and
+ * EFX_FILTER_MATCH_IFRM_UNKNOWN_MCAST_DST match flags in the same
+ * specifications after copying.
+ *
+ * @param spec[in, out]
+ *   SFC flow specification to update.
+ * @param filters_count_for_one_val[in]
+ *   How many specifications should have the same match flag, what is the
+ *   number of specifications before copying.
+ * @param error[out]
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_set_ifrm_unknown_dst_flags(struct sfc_flow_spec *spec,
+                                   unsigned int filters_count_for_one_val,
+                                   struct rte_flow_error *error)
+{
+       unsigned int i;
+       static const efx_filter_match_flags_t vals[] = {
+               EFX_FILTER_MATCH_IFRM_UNKNOWN_UCAST_DST,
+               EFX_FILTER_MATCH_IFRM_UNKNOWN_MCAST_DST
+       };
+
+       if (filters_count_for_one_val * RTE_DIM(vals) != spec->count) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                       "Number of specifications is incorrect while copying "
+                       "by inner frame unknown destination flags");
+               return -rte_errno;
+       }
+
+       for (i = 0; i < spec->count; i++) {
+               /* The check above ensures that divisor can't be zero here */
+               spec->filters[i].efs_match_flags |=
+                       vals[i / filters_count_for_one_val];
+       }
+
+       return 0;
+}
+
+/**
+ * Check that the following conditions are met:
+ * - the specification corresponds to a filter for encapsulated traffic
+ * - the list of supported filters has a filter
+ *   with EFX_FILTER_MATCH_IFRM_UNKNOWN_MCAST_DST flag instead of
+ *   EFX_FILTER_MATCH_IFRM_UNKNOWN_UCAST_DST, since this filter will also
+ *   be inserted.
+ *
+ * @param match[in]
+ *   The match flags of filter.
+ * @param spec[in]
+ *   Specification to be supplemented.
+ * @param filter[in]
+ *   SFC filter with list of supported filters.
+ */
+static boolean_t
+sfc_flow_check_ifrm_unknown_dst_flags(efx_filter_match_flags_t match,
+                                     efx_filter_spec_t *spec,
+                                     struct sfc_filter *filter)
+{
+       unsigned int i;
+       efx_tunnel_protocol_t encap_type = spec->efs_encap_type;
+       efx_filter_match_flags_t match_mcast_dst;
+
+       if (encap_type == EFX_TUNNEL_PROTOCOL_NONE)
+               return B_FALSE;
+
+       match_mcast_dst =
+               (match & ~EFX_FILTER_MATCH_IFRM_UNKNOWN_UCAST_DST) |
+               EFX_FILTER_MATCH_IFRM_UNKNOWN_MCAST_DST;
+       for (i = 0; i < filter->supported_match_num; i++) {
+               if (match_mcast_dst == filter->supported_match[i])
+                       return B_TRUE;
+       }
+
+       return B_FALSE;
+}
+
+/**
+ * Check that the list of supported filters has a filter that differs
+ * from @p match in that it has no flag EFX_FILTER_MATCH_OUTER_VID
+ * in this case that filter will be used and the flag
+ * EFX_FILTER_MATCH_OUTER_VID is not needed.
+ *
+ * @param match[in]
+ *   The match flags of filter.
+ * @param spec[in]
+ *   Specification to be supplemented.
+ * @param filter[in]
+ *   SFC filter with list of supported filters.
+ */
+static boolean_t
+sfc_flow_check_outer_vid_flag(efx_filter_match_flags_t match,
+                             __rte_unused efx_filter_spec_t *spec,
+                             struct sfc_filter *filter)
+{
+       unsigned int i;
+       efx_filter_match_flags_t match_without_vid =
+               match & ~EFX_FILTER_MATCH_OUTER_VID;
+
+       for (i = 0; i < filter->supported_match_num; i++) {
+               if (match_without_vid == filter->supported_match[i])
+                       return B_FALSE;
+       }
+
+       return B_TRUE;
+}
+
+/*
+ * Match flags that can be automatically added to filters.
+ * Selecting the last minimum when searching for the copy flag ensures that the
+ * EFX_FILTER_MATCH_UNKNOWN_UCAST_DST flag has a higher priority than
+ * EFX_FILTER_MATCH_ETHER_TYPE. This is because the filter
+ * EFX_FILTER_MATCH_UNKNOWN_UCAST_DST is at the end of the list of supported
+ * filters.
+ */
+static const struct sfc_flow_copy_flag sfc_flow_copy_flags[] = {
+       {
+               .flag = EFX_FILTER_MATCH_UNKNOWN_UCAST_DST,
+               .vals_count = 2,
+               .set_vals = sfc_flow_set_unknown_dst_flags,
+               .spec_check = sfc_flow_check_unknown_dst_flags,
+       },
+       {
+               .flag = EFX_FILTER_MATCH_ETHER_TYPE,
+               .vals_count = 2,
+               .set_vals = sfc_flow_set_ethertypes,
+               .spec_check = NULL,
+       },
+       {
+               .flag = EFX_FILTER_MATCH_IFRM_UNKNOWN_UCAST_DST,
+               .vals_count = 2,
+               .set_vals = sfc_flow_set_ifrm_unknown_dst_flags,
+               .spec_check = sfc_flow_check_ifrm_unknown_dst_flags,
+       },
+       {
+               .flag = EFX_FILTER_MATCH_OUTER_VID,
+               .vals_count = 1,
+               .set_vals = sfc_flow_set_outer_vid_flag,
+               .spec_check = sfc_flow_check_outer_vid_flag,
+       },
+};
+
+/* Get item from array sfc_flow_copy_flags */
+static const struct sfc_flow_copy_flag *
+sfc_flow_get_copy_flag(efx_filter_match_flags_t flag)
+{
+       unsigned int i;
+
+       for (i = 0; i < RTE_DIM(sfc_flow_copy_flags); i++) {
+               if (sfc_flow_copy_flags[i].flag == flag)
+                       return &sfc_flow_copy_flags[i];
+       }
+
+       return NULL;
+}
+
+/**
+ * Make copies of the specifications, set match flag and values
+ * of the field that corresponds to it.
+ *
+ * @param spec[in, out]
+ *   SFC flow specification to update.
+ * @param flag[in]
+ *   The match flag to add.
+ * @param error[out]
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_spec_add_match_flag(struct sfc_flow_spec *spec,
+                            efx_filter_match_flags_t flag,
+                            struct rte_flow_error *error)
+{
+       unsigned int i;
+       unsigned int new_filters_count;
+       unsigned int filters_count_for_one_val;
+       const struct sfc_flow_copy_flag *copy_flag;
+       int rc;
+
+       copy_flag = sfc_flow_get_copy_flag(flag);
+       if (copy_flag == NULL) {
+               rte_flow_error_set(error, ENOTSUP,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                                  "Unsupported spec field for copying");
+               return -rte_errno;
        }
 
-       if (!is_specified) {
+       new_filters_count = spec->count * copy_flag->vals_count;
+       if (new_filters_count > SF_FLOW_SPEC_NB_FILTERS_MAX) {
                rte_flow_error_set(error, EINVAL,
-                                  RTE_FLOW_ERROR_TYPE_ACTION_NUM, actions,
-                                  "Action is unspecified");
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                       "Too much EFX specifications in the flow rule");
+               return -rte_errno;
+       }
+
+       /* Copy filters specifications */
+       for (i = spec->count; i < new_filters_count; i++)
+               spec->filters[i] = spec->filters[i - spec->count];
+
+       filters_count_for_one_val = spec->count;
+       spec->count = new_filters_count;
+
+       rc = copy_flag->set_vals(spec, filters_count_for_one_val, error);
+       if (rc != 0)
+               return rc;
+
+       return 0;
+}
+
+/**
+ * Check that the given set of match flags missing in the original filter spec
+ * could be covered by adding spec copies which specify the corresponding
+ * flags and packet field values to match.
+ *
+ * @param miss_flags[in]
+ *   Flags that are missing until the supported filter.
+ * @param spec[in]
+ *   Specification to be supplemented.
+ * @param filter[in]
+ *   SFC filter.
+ *
+ * @return
+ *   Number of specifications after copy or 0, if the flags can not be added.
+ */
+static unsigned int
+sfc_flow_check_missing_flags(efx_filter_match_flags_t miss_flags,
+                            efx_filter_spec_t *spec,
+                            struct sfc_filter *filter)
+{
+       unsigned int i;
+       efx_filter_match_flags_t copy_flags = 0;
+       efx_filter_match_flags_t flag;
+       efx_filter_match_flags_t match = spec->efs_match_flags | miss_flags;
+       sfc_flow_spec_check *check;
+       unsigned int multiplier = 1;
+
+       for (i = 0; i < RTE_DIM(sfc_flow_copy_flags); i++) {
+               flag = sfc_flow_copy_flags[i].flag;
+               check = sfc_flow_copy_flags[i].spec_check;
+               if ((flag & miss_flags) == flag) {
+                       if (check != NULL && (!check(match, spec, filter)))
+                               continue;
+
+                       copy_flags |= flag;
+                       multiplier *= sfc_flow_copy_flags[i].vals_count;
+               }
+       }
+
+       if (copy_flags == miss_flags)
+               return multiplier;
+
+       return 0;
+}
+
+/**
+ * Attempt to supplement the specification template to the minimally
+ * supported set of match flags. To do this, it is necessary to copy
+ * the specifications, filling them with the values of fields that
+ * correspond to the missing flags.
+ * The necessary and sufficient filter is built from the fewest number
+ * of copies which could be made to cover the minimally required set
+ * of flags.
+ *
+ * @param sa[in]
+ *   SFC adapter.
+ * @param spec[in, out]
+ *   SFC flow specification to update.
+ * @param error[out]
+ *   Perform verbose error reporting if not NULL.
+ */
+static int
+sfc_flow_spec_filters_complete(struct sfc_adapter *sa,
+                              struct sfc_flow_spec *spec,
+                              struct rte_flow_error *error)
+{
+       struct sfc_filter *filter = &sa->filter;
+       efx_filter_match_flags_t miss_flags;
+       efx_filter_match_flags_t min_miss_flags = 0;
+       efx_filter_match_flags_t match;
+       unsigned int min_multiplier = UINT_MAX;
+       unsigned int multiplier;
+       unsigned int i;
+       int rc;
+
+       match = spec->template.efs_match_flags;
+       for (i = 0; i < filter->supported_match_num; i++) {
+               if ((match & filter->supported_match[i]) == match) {
+                       miss_flags = filter->supported_match[i] & (~match);
+                       multiplier = sfc_flow_check_missing_flags(miss_flags,
+                               &spec->template, filter);
+                       if (multiplier > 0) {
+                               if (multiplier <= min_multiplier) {
+                                       min_multiplier = multiplier;
+                                       min_miss_flags = miss_flags;
+                               }
+                       }
+               }
+       }
+
+       if (min_multiplier == UINT_MAX) {
+               rte_flow_error_set(error, ENOTSUP,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                                  "The flow rule pattern is unsupported");
+               return -rte_errno;
+       }
+
+       for (i = 0; i < RTE_DIM(sfc_flow_copy_flags); i++) {
+               efx_filter_match_flags_t flag = sfc_flow_copy_flags[i].flag;
+
+               if ((flag & min_miss_flags) == flag) {
+                       rc = sfc_flow_spec_add_match_flag(spec, flag, error);
+                       if (rc != 0)
+                               return rc;
+               }
+       }
+
+       return 0;
+}
+
+/**
+ * Check that set of match flags is referred to by a filter. Filter is
+ * described by match flags with the ability to add OUTER_VID and INNER_VID
+ * flags.
+ *
+ * @param match_flags[in]
+ *   Set of match flags.
+ * @param flags_pattern[in]
+ *   Pattern of filter match flags.
+ */
+static boolean_t
+sfc_flow_is_match_with_vids(efx_filter_match_flags_t match_flags,
+                           efx_filter_match_flags_t flags_pattern)
+{
+       if ((match_flags & flags_pattern) != flags_pattern)
+               return B_FALSE;
+
+       switch (match_flags & ~flags_pattern) {
+       case 0:
+       case EFX_FILTER_MATCH_OUTER_VID:
+       case EFX_FILTER_MATCH_OUTER_VID | EFX_FILTER_MATCH_INNER_VID:
+               return B_TRUE;
+       default:
+               return B_FALSE;
+       }
+}
+
+/**
+ * Check whether the spec maps to a hardware filter which is known to be
+ * ineffective despite being valid.
+ *
+ * @param filter[in]
+ *   SFC filter with list of supported filters.
+ * @param spec[in]
+ *   SFC flow specification.
+ */
+static boolean_t
+sfc_flow_is_match_flags_exception(struct sfc_filter *filter,
+                                 struct sfc_flow_spec *spec)
+{
+       unsigned int i;
+       uint16_t ether_type;
+       uint8_t ip_proto;
+       efx_filter_match_flags_t match_flags;
+
+       for (i = 0; i < spec->count; i++) {
+               match_flags = spec->filters[i].efs_match_flags;
+
+               if (sfc_flow_is_match_with_vids(match_flags,
+                                               EFX_FILTER_MATCH_ETHER_TYPE) ||
+                   sfc_flow_is_match_with_vids(match_flags,
+                                               EFX_FILTER_MATCH_ETHER_TYPE |
+                                               EFX_FILTER_MATCH_LOC_MAC)) {
+                       ether_type = spec->filters[i].efs_ether_type;
+                       if (filter->supports_ip_proto_or_addr_filter &&
+                           (ether_type == EFX_ETHER_TYPE_IPV4 ||
+                            ether_type == EFX_ETHER_TYPE_IPV6))
+                               return B_TRUE;
+               } else if (sfc_flow_is_match_with_vids(match_flags,
+                               EFX_FILTER_MATCH_ETHER_TYPE |
+                               EFX_FILTER_MATCH_IP_PROTO) ||
+                          sfc_flow_is_match_with_vids(match_flags,
+                               EFX_FILTER_MATCH_ETHER_TYPE |
+                               EFX_FILTER_MATCH_IP_PROTO |
+                               EFX_FILTER_MATCH_LOC_MAC)) {
+                       ip_proto = spec->filters[i].efs_ip_proto;
+                       if (filter->supports_rem_or_local_port_filter &&
+                           (ip_proto == EFX_IPPROTO_TCP ||
+                            ip_proto == EFX_IPPROTO_UDP))
+                               return B_TRUE;
+               }
+       }
+
+       return B_FALSE;
+}
+
+static int
+sfc_flow_validate_match_flags(struct sfc_adapter *sa,
+                             struct rte_flow *flow,
+                             struct rte_flow_error *error)
+{
+       efx_filter_spec_t *spec_tmpl = &flow->spec.template;
+       efx_filter_match_flags_t match_flags = spec_tmpl->efs_match_flags;
+       int rc;
+
+       /* Initialize the first filter spec with template */
+       flow->spec.filters[0] = *spec_tmpl;
+       flow->spec.count = 1;
+
+       if (!sfc_filter_is_match_supported(sa, match_flags)) {
+               rc = sfc_flow_spec_filters_complete(sa, &flow->spec, error);
+               if (rc != 0)
+                       return rc;
+       }
+
+       if (sfc_flow_is_match_flags_exception(&sa->filter, &flow->spec)) {
+               rte_flow_error_set(error, ENOTSUP,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                       "The flow rule pattern is unsupported");
                return -rte_errno;
        }
 
@@ -646,11 +2253,9 @@ sfc_flow_parse(struct rte_eth_dev *dev,
               struct rte_flow *flow,
               struct rte_flow_error *error)
 {
-       struct sfc_adapter *sa = dev->data->dev_private;
+       struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
        int rc;
 
-       memset(&flow->spec, 0, sizeof(flow->spec));
-
        rc = sfc_flow_parse_attr(attr, flow, error);
        if (rc != 0)
                goto fail_bad_value;
@@ -663,12 +2268,11 @@ sfc_flow_parse(struct rte_eth_dev *dev,
        if (rc != 0)
                goto fail_bad_value;
 
-       if (!sfc_filter_is_match_supported(sa, flow->spec.efs_match_flags)) {
-               rte_flow_error_set(error, ENOTSUP,
-                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
-                                  "Flow rule pattern is not supported");
-               return -rte_errno;
-       }
+       rc = sfc_flow_validate_match_flags(sa, flow, error);
+       if (rc != 0)
+               goto fail_bad_value;
+
+       return 0;
 
 fail_bad_value:
        return rc;
@@ -683,6 +2287,8 @@ sfc_flow_validate(struct rte_eth_dev *dev,
 {
        struct rte_flow flow;
 
+       memset(&flow, 0, sizeof(flow));
+
        return sfc_flow_parse(dev, attr, pattern, actions, &flow, error);
 }
 
@@ -693,7 +2299,7 @@ sfc_flow_create(struct rte_eth_dev *dev,
                const struct rte_flow_action actions[],
                struct rte_flow_error *error)
 {
-       struct sfc_adapter *sa = dev->data->dev_private;
+       struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
        struct rte_flow *flow = NULL;
        int rc;
 
@@ -714,7 +2320,7 @@ sfc_flow_create(struct rte_eth_dev *dev,
        sfc_adapter_lock(sa);
 
        if (sa->state == SFC_ADAPTER_STARTED) {
-               rc = efx_filter_insert(sa->nic, &flow->spec);
+               rc = sfc_flow_filter_insert(sa, flow);
                if (rc != 0) {
                        rte_flow_error_set(error, rc,
                                RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
@@ -748,7 +2354,7 @@ sfc_flow_remove(struct sfc_adapter *sa,
        SFC_ASSERT(sfc_adapter_is_locked(sa));
 
        if (sa->state == SFC_ADAPTER_STARTED) {
-               rc = efx_filter_remove(sa->nic, &flow->spec);
+               rc = sfc_flow_filter_remove(sa, flow);
                if (rc != 0)
                        rte_flow_error_set(error, rc,
                                RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
@@ -766,7 +2372,7 @@ sfc_flow_destroy(struct rte_eth_dev *dev,
                 struct rte_flow *flow,
                 struct rte_flow_error *error)
 {
-       struct sfc_adapter *sa = dev->data->dev_private;
+       struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
        struct rte_flow *flow_ptr;
        int rc = EINVAL;
 
@@ -795,7 +2401,7 @@ static int
 sfc_flow_flush(struct rte_eth_dev *dev,
               struct rte_flow_error *error)
 {
-       struct sfc_adapter *sa = dev->data->dev_private;
+       struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
        struct rte_flow *flow;
        int rc = 0;
        int ret = 0;
@@ -813,12 +2419,34 @@ sfc_flow_flush(struct rte_eth_dev *dev,
        return -ret;
 }
 
+static int
+sfc_flow_isolate(struct rte_eth_dev *dev, int enable,
+                struct rte_flow_error *error)
+{
+       struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
+       int ret = 0;
+
+       sfc_adapter_lock(sa);
+       if (sa->state != SFC_ADAPTER_INITIALIZED) {
+               rte_flow_error_set(error, EBUSY,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                                  NULL, "please close the port first");
+               ret = -rte_errno;
+       } else {
+               sfc_sa2shared(sa)->isolated = (enable) ? B_TRUE : B_FALSE;
+       }
+       sfc_adapter_unlock(sa);
+
+       return ret;
+}
+
 const struct rte_flow_ops sfc_flow_ops = {
        .validate = sfc_flow_validate,
        .create = sfc_flow_create,
        .destroy = sfc_flow_destroy,
        .flush = sfc_flow_flush,
        .query = NULL,
+       .isolate = sfc_flow_isolate,
 };
 
 void
@@ -850,7 +2478,7 @@ sfc_flow_stop(struct sfc_adapter *sa)
        SFC_ASSERT(sfc_adapter_is_locked(sa));
 
        TAILQ_FOREACH(flow, &sa->filter.flow_list, entries)
-               efx_filter_remove(sa->nic, &flow->spec);
+               sfc_flow_filter_remove(sa, flow);
 }
 
 int
@@ -864,7 +2492,7 @@ sfc_flow_start(struct sfc_adapter *sa)
        SFC_ASSERT(sfc_adapter_is_locked(sa));
 
        TAILQ_FOREACH(flow, &sa->filter.flow_list, entries) {
-               rc = efx_filter_insert(sa->nic, &flow->spec);
+               rc = sfc_flow_filter_insert(sa, flow);
                if (rc != 0)
                        goto fail_bad_flow;
        }