net: add rte prefix to ether defines
[dpdk.git] / drivers / net / e1000 / igb_flow.c
index 43eebba..3429861 100644 (file)
@@ -1,34 +1,5 @@
-/*-
- *   BSD LICENSE
- *
- *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
- *   All rights reserved.
- *
- *   Redistribution and use in source and binary forms, with or without
- *   modification, are permitted provided that the following conditions
- *   are met:
- *
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * 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.
- *     * Neither the name of Intel Corporation nor the names of its
- *       contributors may be used to endorse or promote products derived
- *       from this software without specific prior written permission.
- *
- *   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.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2016 Intel Corporation
  */
 
 #include <sys/queue.h>
 #include <rte_debug.h>
 #include <rte_pci.h>
 #include <rte_ether.h>
-#include <rte_ethdev.h>
+#include <rte_ethdev_driver.h>
 #include <rte_ethdev_pci.h>
 #include <rte_memory.h>
-#include <rte_memzone.h>
 #include <rte_eal.h>
 #include <rte_atomic.h>
 #include <rte_malloc.h>
@@ -77,6 +47,8 @@
                }                                               \
        } while (0)
 
+#define        IGB_FLEX_RAW_NUM        12
+
 /**
  * Please aware there's an asumption for all the parsers.
  * rte_flow_item is using big endian, rte_flow_attr and
@@ -203,7 +175,7 @@ cons_parse_ntuple_filter(const struct rte_flow_attr *attr,
                return -rte_errno;
        }
 
-       ipv4_mask = (const struct rte_flow_item_ipv4 *)item->mask;
+       ipv4_mask = item->mask;
        /**
         * Only support src & dst addresses, protocol,
         * others should be masked.
@@ -226,7 +198,7 @@ cons_parse_ntuple_filter(const struct rte_flow_attr *attr,
        filter->src_ip_mask = ipv4_mask->hdr.src_addr;
        filter->proto_mask  = ipv4_mask->hdr.next_proto_id;
 
-       ipv4_spec = (const struct rte_flow_item_ipv4 *)item->spec;
+       ipv4_spec = item->spec;
        filter->dst_ip = ipv4_spec->hdr.dst_addr;
        filter->src_ip = ipv4_spec->hdr.src_addr;
        filter->proto  = ipv4_spec->hdr.next_proto_id;
@@ -256,7 +228,7 @@ cons_parse_ntuple_filter(const struct rte_flow_attr *attr,
        /* get the TCP/UDP/SCTP info */
        if (item->type == RTE_FLOW_ITEM_TYPE_TCP) {
                if (item->spec && item->mask) {
-                       tcp_mask = (const struct rte_flow_item_tcp *)item->mask;
+                       tcp_mask = item->mask;
 
                        /**
                         * Only support src & dst ports, tcp flags,
@@ -291,14 +263,14 @@ cons_parse_ntuple_filter(const struct rte_flow_attr *attr,
                                return -rte_errno;
                        }
 
-                       tcp_spec = (const struct rte_flow_item_tcp *)item->spec;
+                       tcp_spec = item->spec;
                        filter->dst_port  = tcp_spec->hdr.dst_port;
                        filter->src_port  = tcp_spec->hdr.src_port;
                        filter->tcp_flags = tcp_spec->hdr.tcp_flags;
                }
        } else if (item->type == RTE_FLOW_ITEM_TYPE_UDP) {
                if (item->spec && item->mask) {
-                       udp_mask = (const struct rte_flow_item_udp *)item->mask;
+                       udp_mask = item->mask;
 
                        /**
                         * Only support src & dst ports,
@@ -317,14 +289,13 @@ cons_parse_ntuple_filter(const struct rte_flow_attr *attr,
                        filter->dst_port_mask = udp_mask->hdr.dst_port;
                        filter->src_port_mask = udp_mask->hdr.src_port;
 
-                       udp_spec = (const struct rte_flow_item_udp *)item->spec;
+                       udp_spec = item->spec;
                        filter->dst_port = udp_spec->hdr.dst_port;
                        filter->src_port = udp_spec->hdr.src_port;
                }
        } else {
                if (item->spec && item->mask) {
-                       sctp_mask = (const struct rte_flow_item_sctp *)
-                                       item->mask;
+                       sctp_mask = item->mask;
 
                        /**
                         * Only support src & dst ports,
@@ -408,6 +379,15 @@ cons_parse_ntuple_filter(const struct rte_flow_attr *attr,
                return -rte_errno;
        }
 
+       /* not supported */
+       if (attr->transfer) {
+               memset(filter, 0, sizeof(struct rte_eth_ntuple_filter));
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER,
+                                  attr, "No support for transfer.");
+               return -rte_errno;
+       }
+
        if (attr->priority > 0xFFFF) {
                memset(filter, 0, sizeof(struct rte_eth_ntuple_filter));
                rte_flow_error_set(error, EINVAL,
@@ -561,16 +541,16 @@ cons_parse_ethertype_filter(const struct rte_flow_attr *attr,
                return -rte_errno;
        }
 
-       eth_spec = (const struct rte_flow_item_eth *)item->spec;
-       eth_mask = (const struct rte_flow_item_eth *)item->mask;
+       eth_spec = item->spec;
+       eth_mask = item->mask;
 
        /* Mask bits of source MAC address must be full of 0.
         * Mask bits of destination MAC address must be full
         * of 1 or full of 0.
         */
-       if (!is_zero_ether_addr(&eth_mask->src) ||
-           (!is_zero_ether_addr(&eth_mask->dst) &&
-            !is_broadcast_ether_addr(&eth_mask->dst))) {
+       if (!rte_is_zero_ether_addr(&eth_mask->src) ||
+           (!rte_is_zero_ether_addr(&eth_mask->dst) &&
+            !rte_is_broadcast_ether_addr(&eth_mask->dst))) {
                rte_flow_error_set(error, EINVAL,
                                RTE_FLOW_ERROR_TYPE_ITEM,
                                item, "Invalid ether address mask");
@@ -587,7 +567,7 @@ cons_parse_ethertype_filter(const struct rte_flow_attr *attr,
        /* If mask bits of destination MAC address
         * are full of 1, set RTE_ETHTYPE_FLAGS_MAC.
         */
-       if (is_broadcast_ether_addr(&eth_mask->dst)) {
+       if (rte_is_broadcast_ether_addr(&eth_mask->dst)) {
                filter->mac_addr = eth_spec->dst;
                filter->flags |= RTE_ETHTYPE_FLAGS_MAC;
        } else {
@@ -652,6 +632,14 @@ cons_parse_ethertype_filter(const struct rte_flow_attr *attr,
                return -rte_errno;
        }
 
+       /* Not supported */
+       if (attr->transfer) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER,
+                               attr, "No support for transfer.");
+               return -rte_errno;
+       }
+
        /* Not supported */
        if (attr->priority) {
                rte_flow_error_set(error, EINVAL,
@@ -692,7 +680,8 @@ igb_parse_ethertype_filter(struct rte_eth_dev *dev,
 
        if (hw->mac.type == e1000_82576) {
                if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576) {
-                       memset(filter, 0, sizeof(struct rte_eth_ntuple_filter));
+                       memset(filter, 0, sizeof(
+                                       struct rte_eth_ethertype_filter));
                        rte_flow_error_set(error, EINVAL,
                                RTE_FLOW_ERROR_TYPE_ITEM,
                                NULL, "queue number not supported "
@@ -701,7 +690,8 @@ igb_parse_ethertype_filter(struct rte_eth_dev *dev,
                }
        } else {
                if (filter->queue >= IGB_MAX_RX_QUEUE_NUM) {
-                       memset(filter, 0, sizeof(struct rte_eth_ntuple_filter));
+                       memset(filter, 0, sizeof(
+                                       struct rte_eth_ethertype_filter));
                        rte_flow_error_set(error, EINVAL,
                                RTE_FLOW_ERROR_TYPE_ITEM,
                                NULL, "queue number not supported "
@@ -710,8 +700,8 @@ igb_parse_ethertype_filter(struct rte_eth_dev *dev,
                }
        }
 
-       if (filter->ether_type == ETHER_TYPE_IPv4 ||
-               filter->ether_type == ETHER_TYPE_IPv6) {
+       if (filter->ether_type == RTE_ETHER_TYPE_IPv4 ||
+               filter->ether_type == RTE_ETHER_TYPE_IPv6) {
                memset(filter, 0, sizeof(struct rte_eth_ethertype_filter));
                rte_flow_error_set(error, EINVAL,
                        RTE_FLOW_ERROR_TYPE_ITEM,
@@ -739,40 +729,1187 @@ igb_parse_ethertype_filter(struct rte_eth_dev *dev,
 }
 
 /**
- * Check if the flow rule is supported by igb.
- * It only checkes the format. Don't guarantee the rule can be programmed into
- * the HW. Because there can be no enough room for the rule.
+ * Parse the rule to see if it is a TCP SYN rule.
+ * And get the TCP SYN filter info BTW.
+ * pattern:
+ * The first not void item must be ETH.
+ * The second not void item must be IPV4 or IPV6.
+ * The third not void item must be TCP.
+ * The next not void item must be END.
+ * action:
+ * The first not void action should be QUEUE.
+ * The next not void action should be END.
+ * pattern example:
+ * ITEM                Spec                    Mask
+ * ETH         NULL                    NULL
+ * IPV4/IPV6   NULL                    NULL
+ * TCP         tcp_flags       0x02    0xFF
+ * END
+ * other members in mask and spec should set to 0x00.
+ * item->last should be NULL.
  */
 static int
-igb_flow_validate(__rte_unused 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)
+cons_parse_syn_filter(const struct rte_flow_attr *attr,
+                               const struct rte_flow_item pattern[],
+                               const struct rte_flow_action actions[],
+                               struct rte_eth_syn_filter *filter,
+                               struct rte_flow_error *error)
 {
-       struct rte_eth_ntuple_filter ntuple_filter;
-       struct rte_eth_ethertype_filter ethertype_filter;
+       const struct rte_flow_item *item;
+       const struct rte_flow_action *act;
+       const struct rte_flow_item_tcp *tcp_spec;
+       const struct rte_flow_item_tcp *tcp_mask;
+       const struct rte_flow_action_queue *act_q;
+       uint32_t index;
+
+       if (!pattern) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM_NUM,
+                               NULL, "NULL pattern.");
+               return -rte_errno;
+       }
+
+       if (!actions) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION_NUM,
+                               NULL, "NULL action.");
+               return -rte_errno;
+       }
+
+       if (!attr) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR,
+                                  NULL, "NULL attribute.");
+               return -rte_errno;
+       }
+
+       /* parse pattern */
+       index = 0;
+
+       /* the first not void item should be MAC or IPv4 or IPv6 or TCP */
+       NEXT_ITEM_OF_PATTERN(item, pattern, index);
+       if (item->type != RTE_FLOW_ITEM_TYPE_ETH &&
+           item->type != RTE_FLOW_ITEM_TYPE_IPV4 &&
+           item->type != RTE_FLOW_ITEM_TYPE_IPV6 &&
+           item->type != RTE_FLOW_ITEM_TYPE_TCP) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by syn filter");
+               return -rte_errno;
+       }
+               /*Not supported last point for range*/
+       if (item->last) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                       item, "Not supported last point for range");
+               return -rte_errno;
+       }
+
+       /* Skip Ethernet */
+       if (item->type == RTE_FLOW_ITEM_TYPE_ETH) {
+               /* if the item is MAC, the content should be NULL */
+               if (item->spec || item->mask) {
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Invalid SYN address mask");
+                       return -rte_errno;
+               }
+
+               /* check if the next not void item is IPv4 or IPv6 */
+               index++;
+               NEXT_ITEM_OF_PATTERN(item, pattern, index);
+               if (item->type != RTE_FLOW_ITEM_TYPE_IPV4 &&
+                   item->type != RTE_FLOW_ITEM_TYPE_IPV6) {
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by syn filter");
+                       return -rte_errno;
+               }
+       }
+
+       /* Skip IP */
+       if (item->type == RTE_FLOW_ITEM_TYPE_IPV4 ||
+           item->type == RTE_FLOW_ITEM_TYPE_IPV6) {
+               /* if the item is IP, the content should be NULL */
+               if (item->spec || item->mask) {
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Invalid SYN mask");
+                       return -rte_errno;
+               }
+
+               /* check if the next not void item is TCP */
+               index++;
+               NEXT_ITEM_OF_PATTERN(item, pattern, index);
+               if (item->type != RTE_FLOW_ITEM_TYPE_TCP) {
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by syn filter");
+                       return -rte_errno;
+               }
+       }
+
+       /* Get the TCP info. Only support SYN. */
+       if (!item->spec || !item->mask) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Invalid SYN mask");
+               return -rte_errno;
+       }
+       /*Not supported last point for range*/
+       if (item->last) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                       item, "Not supported last point for range");
+               return -rte_errno;
+       }
+
+       tcp_spec = item->spec;
+       tcp_mask = item->mask;
+       if (!(tcp_spec->hdr.tcp_flags & TCP_SYN_FLAG) ||
+           tcp_mask->hdr.src_port ||
+           tcp_mask->hdr.dst_port ||
+           tcp_mask->hdr.sent_seq ||
+           tcp_mask->hdr.recv_ack ||
+           tcp_mask->hdr.data_off ||
+           tcp_mask->hdr.tcp_flags != TCP_SYN_FLAG ||
+           tcp_mask->hdr.rx_win ||
+           tcp_mask->hdr.cksum ||
+           tcp_mask->hdr.tcp_urp) {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by syn filter");
+               return -rte_errno;
+       }
+
+       /* check if the next not void item is END */
+       index++;
+       NEXT_ITEM_OF_PATTERN(item, pattern, index);
+       if (item->type != RTE_FLOW_ITEM_TYPE_END) {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by syn filter");
+               return -rte_errno;
+       }
+
+       /* parse action */
+       index = 0;
+
+       /* check if the first not void action is QUEUE. */
+       NEXT_ITEM_OF_ACTION(act, actions, index);
+       if (act->type != RTE_FLOW_ACTION_TYPE_QUEUE) {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION,
+                               act, "Not supported action.");
+               return -rte_errno;
+       }
+
+       act_q = (const struct rte_flow_action_queue *)act->conf;
+       filter->queue = act_q->index;
+
+       /* check if the next not void item is END */
+       index++;
+       NEXT_ITEM_OF_ACTION(act, actions, index);
+       if (act->type != RTE_FLOW_ACTION_TYPE_END) {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION,
+                               act, "Not supported action.");
+               return -rte_errno;
+       }
+
+       /* parse attr */
+       /* must be input direction */
+       if (!attr->ingress) {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ATTR_INGRESS,
+                       attr, "Only support ingress.");
+               return -rte_errno;
+       }
+
+       /* not supported */
+       if (attr->egress) {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ATTR_EGRESS,
+                       attr, "Not support egress.");
+               return -rte_errno;
+       }
+
+       /* not supported */
+       if (attr->transfer) {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER,
+                       attr, "No support for transfer.");
+               return -rte_errno;
+       }
+
+       /* Support 2 priorities, the lowest or highest. */
+       if (!attr->priority) {
+               filter->hig_pri = 0;
+       } else if (attr->priority == (uint32_t)~0U) {
+               filter->hig_pri = 1;
+       } else {
+               memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY,
+                       attr, "Not support priority.");
+               return -rte_errno;
+       }
+
+       return 0;
+}
+
+static int
+igb_parse_syn_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_eth_syn_filter *filter,
+                            struct rte_flow_error *error)
+{
+       struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
        int ret;
 
-       memset(&ntuple_filter, 0, sizeof(struct rte_eth_ntuple_filter));
-       ret = igb_parse_ntuple_filter(dev, attr, pattern,
-                               actions, &ntuple_filter, error);
-       if (!ret)
-               return 0;
+       MAC_TYPE_FILTER_SUP(hw->mac.type);
 
-       memset(&ethertype_filter, 0, sizeof(struct rte_eth_ethertype_filter));
-       ret = igb_parse_ethertype_filter(dev, attr, pattern,
-                               actions, &ethertype_filter, error);
-       if (!ret)
-               return 0;
+       ret = cons_parse_syn_filter(attr, pattern,
+                                       actions, filter, error);
 
-       return ret;
+       if (hw->mac.type == e1000_82576) {
+               if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576) {
+                       memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               NULL, "queue number not "
+                                       "supported by syn filter");
+                       return -rte_errno;
+               }
+       } else {
+               if (filter->queue >= IGB_MAX_RX_QUEUE_NUM) {
+                       memset(filter, 0, sizeof(struct rte_eth_syn_filter));
+                       rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               NULL, "queue number not "
+                                       "supported by syn filter");
+                       return -rte_errno;
+               }
+       }
+
+       if (ret)
+               return ret;
+
+       return 0;
+}
+
+/**
+ * Parse the rule to see if it is a flex byte rule.
+ * And get the flex byte filter info BTW.
+ * pattern:
+ * The first not void item must be RAW.
+ * The second not void item can be RAW or END.
+ * The third not void item can be RAW or END.
+ * The last not void item must be END.
+ * action:
+ * The first not void action should be QUEUE.
+ * The next not void action should be END.
+ * pattern example:
+ * ITEM                Spec                    Mask
+ * RAW         relative        0               0x1
+ *                     offset  0               0xFFFFFFFF
+ *                     pattern {0x08, 0x06}            {0xFF, 0xFF}
+ * RAW         relative        1               0x1
+ *                     offset  100             0xFFFFFFFF
+ *                     pattern {0x11, 0x22, 0x33}      {0xFF, 0xFF, 0xFF}
+ * END
+ * other members in mask and spec should set to 0x00.
+ * item->last should be NULL.
+ */
+static int
+cons_parse_flex_filter(const struct rte_flow_attr *attr,
+                               const struct rte_flow_item pattern[],
+                               const struct rte_flow_action actions[],
+                               struct rte_eth_flex_filter *filter,
+                               struct rte_flow_error *error)
+{
+       const struct rte_flow_item *item;
+       const struct rte_flow_action *act;
+       const struct rte_flow_item_raw *raw_spec;
+       const struct rte_flow_item_raw *raw_mask;
+       const struct rte_flow_action_queue *act_q;
+       uint32_t index, i, offset, total_offset;
+       uint32_t max_offset = 0;
+       int32_t shift, j, raw_index = 0;
+       int32_t relative[IGB_FLEX_RAW_NUM] = {0};
+       int32_t raw_offset[IGB_FLEX_RAW_NUM] = {0};
+
+       if (!pattern) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM_NUM,
+                               NULL, "NULL pattern.");
+               return -rte_errno;
+       }
+
+       if (!actions) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION_NUM,
+                               NULL, "NULL action.");
+               return -rte_errno;
+       }
+
+       if (!attr) {
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR,
+                                  NULL, "NULL attribute.");
+               return -rte_errno;
+       }
+
+       /* parse pattern */
+       index = 0;
+
+item_loop:
+
+       /* the first not void item should be RAW */
+       NEXT_ITEM_OF_PATTERN(item, pattern, index);
+       if (item->type != RTE_FLOW_ITEM_TYPE_RAW) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by flex filter");
+               return -rte_errno;
+       }
+               /*Not supported last point for range*/
+       if (item->last) {
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                       item, "Not supported last point for range");
+               return -rte_errno;
+       }
+
+       raw_spec = item->spec;
+       raw_mask = item->mask;
+
+       if (!raw_mask->length ||
+           !raw_mask->relative) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by flex filter");
+               return -rte_errno;
+       }
+
+       if (raw_mask->offset)
+               offset = raw_spec->offset;
+       else
+               offset = 0;
+
+       for (j = 0; j < raw_spec->length; j++) {
+               if (raw_mask->pattern[j] != 0xFF) {
+                       memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+                       rte_flow_error_set(error, EINVAL,
+                                       RTE_FLOW_ERROR_TYPE_ITEM,
+                                       item, "Not supported by flex filter");
+                       return -rte_errno;
+               }
+       }
+
+       total_offset = 0;
+
+       if (raw_spec->relative) {
+               for (j = raw_index; j > 0; j--) {
+                       total_offset += raw_offset[j - 1];
+                       if (!relative[j - 1])
+                               break;
+               }
+               if (total_offset + raw_spec->length + offset > max_offset)
+                       max_offset = total_offset + raw_spec->length + offset;
+       } else {
+               if (raw_spec->length + offset > max_offset)
+                       max_offset = raw_spec->length + offset;
+       }
+
+       if ((raw_spec->length + offset + total_offset) >
+                       RTE_FLEX_FILTER_MAXLEN) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by flex filter");
+               return -rte_errno;
+       }
+
+       if (raw_spec->relative == 0) {
+               for (j = 0; j < raw_spec->length; j++)
+                       filter->bytes[offset + j] =
+                       raw_spec->pattern[j];
+               j = offset / CHAR_BIT;
+               shift = offset % CHAR_BIT;
+       } else {
+               for (j = 0; j < raw_spec->length; j++)
+                       filter->bytes[total_offset + offset + j] =
+                               raw_spec->pattern[j];
+               j = (total_offset + offset) / CHAR_BIT;
+               shift = (total_offset + offset) % CHAR_BIT;
+       }
+
+       i = 0;
+
+       for ( ; shift < CHAR_BIT; shift++) {
+               filter->mask[j] |= (0x80 >> shift);
+               i++;
+               if (i == raw_spec->length)
+                       break;
+               if (shift == (CHAR_BIT - 1)) {
+                       j++;
+                       shift = -1;
+               }
+       }
+
+       relative[raw_index] = raw_spec->relative;
+       raw_offset[raw_index] = offset + raw_spec->length;
+       raw_index++;
+
+       /* check if the next not void item is RAW */
+       index++;
+       NEXT_ITEM_OF_PATTERN(item, pattern, index);
+       if (item->type != RTE_FLOW_ITEM_TYPE_RAW &&
+               item->type != RTE_FLOW_ITEM_TYPE_END) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ITEM,
+                               item, "Not supported by flex filter");
+               return -rte_errno;
+       }
+
+       /* go back to parser */
+       if (item->type == RTE_FLOW_ITEM_TYPE_RAW) {
+               /* if the item is RAW, the content should be parse */
+               goto item_loop;
+       }
+
+       filter->len = RTE_ALIGN(max_offset, 8);
+
+       /* parse action */
+       index = 0;
+
+       /* check if the first not void action is QUEUE. */
+       NEXT_ITEM_OF_ACTION(act, actions, index);
+       if (act->type != RTE_FLOW_ACTION_TYPE_QUEUE) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION,
+                               act, "Not supported action.");
+               return -rte_errno;
+       }
+
+       act_q = (const struct rte_flow_action_queue *)act->conf;
+       filter->queue = act_q->index;
+
+       /* check if the next not void item is END */
+       index++;
+       NEXT_ITEM_OF_ACTION(act, actions, index);
+       if (act->type != RTE_FLOW_ACTION_TYPE_END) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION,
+                               act, "Not supported action.");
+               return -rte_errno;
+       }
+
+       /* parse attr */
+       /* must be input direction */
+       if (!attr->ingress) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ATTR_INGRESS,
+                       attr, "Only support ingress.");
+               return -rte_errno;
+       }
+
+       /* not supported */
+       if (attr->egress) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ATTR_EGRESS,
+                       attr, "Not support egress.");
+               return -rte_errno;
+       }
+
+       /* not supported */
+       if (attr->transfer) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER,
+                       attr, "No support for transfer.");
+               return -rte_errno;
+       }
+
+       if (attr->priority > 0xFFFF) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY,
+                                  attr, "Error priority.");
+               return -rte_errno;
+       }
+
+       filter->priority = (uint16_t)attr->priority;
+
+       return 0;
+}
+
+static int
+igb_parse_flex_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_eth_flex_filter *filter,
+                            struct rte_flow_error *error)
+{
+       struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       int ret;
+
+       MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
+
+       ret = cons_parse_flex_filter(attr, pattern,
+                                       actions, filter, error);
+
+       if (filter->queue >= IGB_MAX_RX_QUEUE_NUM) {
+               memset(filter, 0, sizeof(struct rte_eth_flex_filter));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ITEM,
+                       NULL, "queue number not supported by flex filter");
+               return -rte_errno;
+       }
+
+       if (filter->len == 0 || filter->len > E1000_MAX_FLEX_FILTER_LEN ||
+               filter->len % sizeof(uint64_t) != 0) {
+               PMD_DRV_LOG(ERR, "filter's length is out of range");
+               return -EINVAL;
+       }
+
+       if (filter->priority > E1000_MAX_FLEX_FILTER_PRI) {
+               PMD_DRV_LOG(ERR, "filter's priority is out of range");
+               return -EINVAL;
+       }
+
+       if (ret)
+               return ret;
+
+       return 0;
+}
+
+static int
+igb_parse_rss_filter(struct rte_eth_dev *dev,
+                       const struct rte_flow_attr *attr,
+                       const struct rte_flow_action actions[],
+                       struct igb_rte_flow_rss_conf *rss_conf,
+                       struct rte_flow_error *error)
+{
+       struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       const struct rte_flow_action *act;
+       const struct rte_flow_action_rss *rss;
+       uint16_t n, index;
+
+       /**
+        * rss only supports forwarding,
+        * check if the first not void action is RSS.
+        */
+       index = 0;
+       NEXT_ITEM_OF_ACTION(act, actions, index);
+       if (act->type != RTE_FLOW_ACTION_TYPE_RSS) {
+               memset(rss_conf, 0, sizeof(struct igb_rte_flow_rss_conf));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ACTION,
+                       act, "Not supported action.");
+               return -rte_errno;
+       }
+
+       rss = (const struct rte_flow_action_rss *)act->conf;
+
+       if (!rss || !rss->queue_num) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION,
+                               act,
+                          "no valid queues");
+               return -rte_errno;
+       }
+
+       for (n = 0; n < rss->queue_num; n++) {
+               if (rss->queue[n] >= dev->data->nb_rx_queues) {
+                       rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ACTION,
+                                  act,
+                                  "queue id > max number of queues");
+                       return -rte_errno;
+               }
+       }
+
+       if (rss->func != RTE_ETH_HASH_FUNCTION_DEFAULT)
+               return rte_flow_error_set
+                       (error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION, act,
+                        "non-default RSS hash functions are not supported");
+       if (rss->level)
+               return rte_flow_error_set
+                       (error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION, act,
+                        "a nonzero RSS encapsulation level is not supported");
+       if (rss->key_len && rss->key_len != RTE_DIM(rss_conf->key))
+               return rte_flow_error_set
+                       (error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION, act,
+                        "RSS hash key must be exactly 40 bytes");
+       if (((hw->mac.type == e1000_82576) &&
+            (rss->queue_num > IGB_MAX_RX_QUEUE_NUM_82576)) ||
+           ((hw->mac.type != e1000_82576) &&
+            (rss->queue_num > IGB_MAX_RX_QUEUE_NUM)))
+               return rte_flow_error_set
+                       (error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION, act,
+                        "too many queues for RSS context");
+       if (igb_rss_conf_init(dev, rss_conf, rss))
+               return rte_flow_error_set
+                       (error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION, act,
+                        "RSS context initialization failure");
+
+       /* check if the next not void item is END */
+       index++;
+       NEXT_ITEM_OF_ACTION(act, actions, index);
+       if (act->type != RTE_FLOW_ACTION_TYPE_END) {
+               memset(rss_conf, 0, sizeof(struct rte_eth_rss_conf));
+               rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ACTION,
+                       act, "Not supported action.");
+               return -rte_errno;
+       }
+
+       /* parse attr */
+       /* must be input direction */
+       if (!attr->ingress) {
+               memset(rss_conf, 0, sizeof(struct igb_rte_flow_rss_conf));
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_INGRESS,
+                                  attr, "Only support ingress.");
+               return -rte_errno;
+       }
+
+       /* not supported */
+       if (attr->egress) {
+               memset(rss_conf, 0, sizeof(struct igb_rte_flow_rss_conf));
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_EGRESS,
+                                  attr, "Not support egress.");
+               return -rte_errno;
+       }
+
+       /* not supported */
+       if (attr->transfer) {
+               memset(rss_conf, 0, sizeof(struct igb_rte_flow_rss_conf));
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_TRANSFER,
+                                  attr, "No support for transfer.");
+               return -rte_errno;
+       }
+
+       if (attr->priority > 0xFFFF) {
+               memset(rss_conf, 0, sizeof(struct igb_rte_flow_rss_conf));
+               rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY,
+                                  attr, "Error priority.");
+               return -rte_errno;
+       }
+
+       return 0;
+}
+
+/**
+ * Create a flow rule.
+ * Theorically one rule can match more than one filters.
+ * We will let it use the filter which it hitt first.
+ * So, the sequence matters.
+ */
+static struct rte_flow *
+igb_flow_create(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)
+{
+       int ret;
+       struct rte_eth_ntuple_filter ntuple_filter;
+       struct rte_eth_ethertype_filter ethertype_filter;
+       struct rte_eth_syn_filter syn_filter;
+       struct rte_eth_flex_filter flex_filter;
+       struct igb_rte_flow_rss_conf rss_conf;
+       struct rte_flow *flow = NULL;
+       struct igb_ntuple_filter_ele *ntuple_filter_ptr;
+       struct igb_ethertype_filter_ele *ethertype_filter_ptr;
+       struct igb_eth_syn_filter_ele *syn_filter_ptr;
+       struct igb_flex_filter_ele *flex_filter_ptr;
+       struct igb_rss_conf_ele *rss_filter_ptr;
+       struct igb_flow_mem *igb_flow_mem_ptr;
+
+       flow = rte_zmalloc("igb_rte_flow", sizeof(struct rte_flow), 0);
+       if (!flow) {
+               PMD_DRV_LOG(ERR, "failed to allocate memory");
+               return (struct rte_flow *)flow;
+       }
+       igb_flow_mem_ptr = rte_zmalloc("igb_flow_mem",
+                       sizeof(struct igb_flow_mem), 0);
+       if (!igb_flow_mem_ptr) {
+               PMD_DRV_LOG(ERR, "failed to allocate memory");
+               rte_free(flow);
+               return NULL;
+       }
+       igb_flow_mem_ptr->flow = flow;
+       igb_flow_mem_ptr->dev = dev;
+       TAILQ_INSERT_TAIL(&igb_flow_list,
+                               igb_flow_mem_ptr, entries);
+
+       memset(&ntuple_filter, 0, sizeof(struct rte_eth_ntuple_filter));
+       ret = igb_parse_ntuple_filter(dev, attr, pattern,
+                       actions, &ntuple_filter, error);
+       if (!ret) {
+               ret = igb_add_del_ntuple_filter(dev, &ntuple_filter, TRUE);
+               if (!ret) {
+                       ntuple_filter_ptr = rte_zmalloc("igb_ntuple_filter",
+                               sizeof(struct igb_ntuple_filter_ele), 0);
+                       if (!ntuple_filter_ptr) {
+                               PMD_DRV_LOG(ERR, "failed to allocate memory");
+                               goto out;
+                       }
+
+                       rte_memcpy(&ntuple_filter_ptr->filter_info,
+                               &ntuple_filter,
+                               sizeof(struct rte_eth_ntuple_filter));
+                       TAILQ_INSERT_TAIL(&igb_filter_ntuple_list,
+                               ntuple_filter_ptr, entries);
+                       flow->rule = ntuple_filter_ptr;
+                       flow->filter_type = RTE_ETH_FILTER_NTUPLE;
+                       return flow;
+               }
+               goto out;
+       }
+
+       memset(&ethertype_filter, 0, sizeof(struct rte_eth_ethertype_filter));
+       ret = igb_parse_ethertype_filter(dev, attr, pattern,
+                               actions, &ethertype_filter, error);
+       if (!ret) {
+               ret = igb_add_del_ethertype_filter(dev,
+                               &ethertype_filter, TRUE);
+               if (!ret) {
+                       ethertype_filter_ptr = rte_zmalloc(
+                               "igb_ethertype_filter",
+                               sizeof(struct igb_ethertype_filter_ele), 0);
+                       if (!ethertype_filter_ptr) {
+                               PMD_DRV_LOG(ERR, "failed to allocate memory");
+                               goto out;
+                       }
+
+                       rte_memcpy(&ethertype_filter_ptr->filter_info,
+                               &ethertype_filter,
+                               sizeof(struct rte_eth_ethertype_filter));
+                       TAILQ_INSERT_TAIL(&igb_filter_ethertype_list,
+                               ethertype_filter_ptr, entries);
+                       flow->rule = ethertype_filter_ptr;
+                       flow->filter_type = RTE_ETH_FILTER_ETHERTYPE;
+                       return flow;
+               }
+               goto out;
+       }
+
+       memset(&syn_filter, 0, sizeof(struct rte_eth_syn_filter));
+       ret = igb_parse_syn_filter(dev, attr, pattern,
+                               actions, &syn_filter, error);
+       if (!ret) {
+               ret = eth_igb_syn_filter_set(dev, &syn_filter, TRUE);
+               if (!ret) {
+                       syn_filter_ptr = rte_zmalloc("igb_syn_filter",
+                               sizeof(struct igb_eth_syn_filter_ele), 0);
+                       if (!syn_filter_ptr) {
+                               PMD_DRV_LOG(ERR, "failed to allocate memory");
+                               goto out;
+                       }
+
+                       rte_memcpy(&syn_filter_ptr->filter_info,
+                               &syn_filter,
+                               sizeof(struct rte_eth_syn_filter));
+                       TAILQ_INSERT_TAIL(&igb_filter_syn_list,
+                               syn_filter_ptr,
+                               entries);
+                       flow->rule = syn_filter_ptr;
+                       flow->filter_type = RTE_ETH_FILTER_SYN;
+                       return flow;
+               }
+               goto out;
+       }
+
+       memset(&flex_filter, 0, sizeof(struct rte_eth_flex_filter));
+       ret = igb_parse_flex_filter(dev, attr, pattern,
+                                       actions, &flex_filter, error);
+       if (!ret) {
+               ret = eth_igb_add_del_flex_filter(dev, &flex_filter, TRUE);
+               if (!ret) {
+                       flex_filter_ptr = rte_zmalloc("igb_flex_filter",
+                               sizeof(struct igb_flex_filter_ele), 0);
+                       if (!flex_filter_ptr) {
+                               PMD_DRV_LOG(ERR, "failed to allocate memory");
+                               goto out;
+                       }
+
+                       rte_memcpy(&flex_filter_ptr->filter_info,
+                               &flex_filter,
+                               sizeof(struct rte_eth_flex_filter));
+                       TAILQ_INSERT_TAIL(&igb_filter_flex_list,
+                               flex_filter_ptr, entries);
+                       flow->rule = flex_filter_ptr;
+                       flow->filter_type = RTE_ETH_FILTER_FLEXIBLE;
+                       return flow;
+               }
+       }
+
+       memset(&rss_conf, 0, sizeof(struct igb_rte_flow_rss_conf));
+       ret = igb_parse_rss_filter(dev, attr,
+                                       actions, &rss_conf, error);
+       if (!ret) {
+               ret = igb_config_rss_filter(dev, &rss_conf, TRUE);
+               if (!ret) {
+                       rss_filter_ptr = rte_zmalloc("igb_rss_filter",
+                               sizeof(struct igb_rss_conf_ele), 0);
+                       if (!rss_filter_ptr) {
+                               PMD_DRV_LOG(ERR, "failed to allocate memory");
+                               goto out;
+                       }
+                       igb_rss_conf_init(dev, &rss_filter_ptr->filter_info,
+                                         &rss_conf.conf);
+                       TAILQ_INSERT_TAIL(&igb_filter_rss_list,
+                               rss_filter_ptr, entries);
+                       flow->rule = rss_filter_ptr;
+                       flow->filter_type = RTE_ETH_FILTER_HASH;
+                       return flow;
+               }
+       }
+
+out:
+       TAILQ_REMOVE(&igb_flow_list,
+               igb_flow_mem_ptr, entries);
+       rte_flow_error_set(error, -ret,
+                          RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+                          "Failed to create flow.");
+       rte_free(igb_flow_mem_ptr);
+       rte_free(flow);
+       return NULL;
+}
+
+/**
+ * Check if the flow rule is supported by igb.
+ * It only checkes the format. Don't guarantee the rule can be programmed into
+ * the HW. Because there can be no enough room for the rule.
+ */
+static int
+igb_flow_validate(__rte_unused 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)
+{
+       struct rte_eth_ntuple_filter ntuple_filter;
+       struct rte_eth_ethertype_filter ethertype_filter;
+       struct rte_eth_syn_filter syn_filter;
+       struct rte_eth_flex_filter flex_filter;
+       struct igb_rte_flow_rss_conf rss_conf;
+       int ret;
+
+       memset(&ntuple_filter, 0, sizeof(struct rte_eth_ntuple_filter));
+       ret = igb_parse_ntuple_filter(dev, attr, pattern,
+                               actions, &ntuple_filter, error);
+       if (!ret)
+               return 0;
+
+       memset(&ethertype_filter, 0, sizeof(struct rte_eth_ethertype_filter));
+       ret = igb_parse_ethertype_filter(dev, attr, pattern,
+                               actions, &ethertype_filter, error);
+       if (!ret)
+               return 0;
+
+       memset(&syn_filter, 0, sizeof(struct rte_eth_syn_filter));
+       ret = igb_parse_syn_filter(dev, attr, pattern,
+                               actions, &syn_filter, error);
+       if (!ret)
+               return 0;
+
+       memset(&flex_filter, 0, sizeof(struct rte_eth_flex_filter));
+       ret = igb_parse_flex_filter(dev, attr, pattern,
+                               actions, &flex_filter, error);
+       if (!ret)
+               return 0;
+
+       memset(&rss_conf, 0, sizeof(struct igb_rte_flow_rss_conf));
+       ret = igb_parse_rss_filter(dev, attr,
+                                       actions, &rss_conf, error);
+
+       return ret;
+}
+
+/* Destroy a flow rule on igb. */
+static int
+igb_flow_destroy(struct rte_eth_dev *dev,
+               struct rte_flow *flow,
+               struct rte_flow_error *error)
+{
+       int ret;
+       struct rte_flow *pmd_flow = flow;
+       enum rte_filter_type filter_type = pmd_flow->filter_type;
+       struct igb_ntuple_filter_ele *ntuple_filter_ptr;
+       struct igb_ethertype_filter_ele *ethertype_filter_ptr;
+       struct igb_eth_syn_filter_ele *syn_filter_ptr;
+       struct igb_flex_filter_ele *flex_filter_ptr;
+       struct igb_flow_mem *igb_flow_mem_ptr;
+       struct igb_rss_conf_ele *rss_filter_ptr;
+
+       switch (filter_type) {
+       case RTE_ETH_FILTER_NTUPLE:
+               ntuple_filter_ptr = (struct igb_ntuple_filter_ele *)
+                                       pmd_flow->rule;
+               ret = igb_add_del_ntuple_filter(dev,
+                               &ntuple_filter_ptr->filter_info, FALSE);
+               if (!ret) {
+                       TAILQ_REMOVE(&igb_filter_ntuple_list,
+                       ntuple_filter_ptr, entries);
+                       rte_free(ntuple_filter_ptr);
+               }
+               break;
+       case RTE_ETH_FILTER_ETHERTYPE:
+               ethertype_filter_ptr = (struct igb_ethertype_filter_ele *)
+                                       pmd_flow->rule;
+               ret = igb_add_del_ethertype_filter(dev,
+                               &ethertype_filter_ptr->filter_info, FALSE);
+               if (!ret) {
+                       TAILQ_REMOVE(&igb_filter_ethertype_list,
+                               ethertype_filter_ptr, entries);
+                       rte_free(ethertype_filter_ptr);
+               }
+               break;
+       case RTE_ETH_FILTER_SYN:
+               syn_filter_ptr = (struct igb_eth_syn_filter_ele *)
+                               pmd_flow->rule;
+               ret = eth_igb_syn_filter_set(dev,
+                               &syn_filter_ptr->filter_info, FALSE);
+               if (!ret) {
+                       TAILQ_REMOVE(&igb_filter_syn_list,
+                               syn_filter_ptr, entries);
+                       rte_free(syn_filter_ptr);
+               }
+               break;
+       case RTE_ETH_FILTER_FLEXIBLE:
+               flex_filter_ptr = (struct igb_flex_filter_ele *)
+                               pmd_flow->rule;
+               ret = eth_igb_add_del_flex_filter(dev,
+                               &flex_filter_ptr->filter_info, FALSE);
+               if (!ret) {
+                       TAILQ_REMOVE(&igb_filter_flex_list,
+                               flex_filter_ptr, entries);
+                       rte_free(flex_filter_ptr);
+               }
+               break;
+       case RTE_ETH_FILTER_HASH:
+               rss_filter_ptr = (struct igb_rss_conf_ele *)
+                               pmd_flow->rule;
+               ret = igb_config_rss_filter(dev,
+                                       &rss_filter_ptr->filter_info, FALSE);
+               if (!ret) {
+                       TAILQ_REMOVE(&igb_filter_rss_list,
+                               rss_filter_ptr, entries);
+                       rte_free(rss_filter_ptr);
+               }
+               break;
+       default:
+               PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
+                           filter_type);
+               ret = -EINVAL;
+               break;
+       }
+
+       if (ret) {
+               rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_HANDLE,
+                               NULL, "Failed to destroy flow");
+               return ret;
+       }
+
+       TAILQ_FOREACH(igb_flow_mem_ptr, &igb_flow_list, entries) {
+               if (igb_flow_mem_ptr->flow == pmd_flow) {
+                       TAILQ_REMOVE(&igb_flow_list,
+                               igb_flow_mem_ptr, entries);
+                       rte_free(igb_flow_mem_ptr);
+               }
+       }
+       rte_free(flow);
+
+       return ret;
+}
+
+/* remove all the n-tuple filters */
+static void
+igb_clear_all_ntuple_filter(struct rte_eth_dev *dev)
+{
+       struct e1000_filter_info *filter_info =
+               E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
+       struct e1000_5tuple_filter *p_5tuple;
+       struct e1000_2tuple_filter *p_2tuple;
+
+       while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
+               igb_delete_5tuple_filter_82576(dev, p_5tuple);
+
+       while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list)))
+               igb_delete_2tuple_filter(dev, p_2tuple);
+}
+
+/* remove all the ether type filters */
+static void
+igb_clear_all_ethertype_filter(struct rte_eth_dev *dev)
+{
+       struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       struct e1000_filter_info *filter_info =
+               E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
+       int i;
+
+       for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
+               if (filter_info->ethertype_mask & (1 << i)) {
+                       (void)igb_ethertype_filter_remove(filter_info,
+                                                           (uint8_t)i);
+                       E1000_WRITE_REG(hw, E1000_ETQF(i), 0);
+                       E1000_WRITE_FLUSH(hw);
+               }
+       }
+}
+
+/* remove the SYN filter */
+static void
+igb_clear_syn_filter(struct rte_eth_dev *dev)
+{
+       struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       struct e1000_filter_info *filter_info =
+               E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
+
+       if (filter_info->syn_info & E1000_SYN_FILTER_ENABLE) {
+               filter_info->syn_info = 0;
+               E1000_WRITE_REG(hw, E1000_SYNQF(0), 0);
+               E1000_WRITE_FLUSH(hw);
+       }
+}
+
+/* remove all the flex filters */
+static void
+igb_clear_all_flex_filter(struct rte_eth_dev *dev)
+{
+       struct e1000_filter_info *filter_info =
+               E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
+       struct e1000_flex_filter *flex_filter;
+
+       while ((flex_filter = TAILQ_FIRST(&filter_info->flex_list)))
+               igb_remove_flex_filter(dev, flex_filter);
+}
+
+/* remove the rss filter */
+static void
+igb_clear_rss_filter(struct rte_eth_dev *dev)
+{
+       struct e1000_filter_info *filter =
+               E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
+
+       if (filter->rss_info.conf.queue_num)
+               igb_config_rss_filter(dev, &filter->rss_info, FALSE);
+}
+
+void
+igb_filterlist_flush(struct rte_eth_dev *dev)
+{
+       struct igb_ntuple_filter_ele *ntuple_filter_ptr;
+       struct igb_ethertype_filter_ele *ethertype_filter_ptr;
+       struct igb_eth_syn_filter_ele *syn_filter_ptr;
+       struct igb_flex_filter_ele *flex_filter_ptr;
+       struct igb_rss_conf_ele  *rss_filter_ptr;
+       struct igb_flow_mem *igb_flow_mem_ptr;
+       enum rte_filter_type filter_type;
+       struct rte_flow *pmd_flow;
+
+       TAILQ_FOREACH(igb_flow_mem_ptr, &igb_flow_list, entries) {
+               if (igb_flow_mem_ptr->dev == dev) {
+                       pmd_flow = igb_flow_mem_ptr->flow;
+                       filter_type = pmd_flow->filter_type;
+
+                       switch (filter_type) {
+                       case RTE_ETH_FILTER_NTUPLE:
+                               ntuple_filter_ptr =
+                               (struct igb_ntuple_filter_ele *)
+                                       pmd_flow->rule;
+                               TAILQ_REMOVE(&igb_filter_ntuple_list,
+                                               ntuple_filter_ptr, entries);
+                               rte_free(ntuple_filter_ptr);
+                               break;
+                       case RTE_ETH_FILTER_ETHERTYPE:
+                               ethertype_filter_ptr =
+                               (struct igb_ethertype_filter_ele *)
+                                       pmd_flow->rule;
+                               TAILQ_REMOVE(&igb_filter_ethertype_list,
+                                               ethertype_filter_ptr, entries);
+                               rte_free(ethertype_filter_ptr);
+                               break;
+                       case RTE_ETH_FILTER_SYN:
+                               syn_filter_ptr =
+                                       (struct igb_eth_syn_filter_ele *)
+                                               pmd_flow->rule;
+                               TAILQ_REMOVE(&igb_filter_syn_list,
+                                               syn_filter_ptr, entries);
+                               rte_free(syn_filter_ptr);
+                               break;
+                       case RTE_ETH_FILTER_FLEXIBLE:
+                               flex_filter_ptr =
+                                       (struct igb_flex_filter_ele *)
+                                               pmd_flow->rule;
+                               TAILQ_REMOVE(&igb_filter_flex_list,
+                                               flex_filter_ptr, entries);
+                               rte_free(flex_filter_ptr);
+                               break;
+                       case RTE_ETH_FILTER_HASH:
+                               rss_filter_ptr =
+                                       (struct igb_rss_conf_ele *)
+                                               pmd_flow->rule;
+                               TAILQ_REMOVE(&igb_filter_rss_list,
+                                               rss_filter_ptr, entries);
+                               rte_free(rss_filter_ptr);
+                               break;
+                       default:
+                               PMD_DRV_LOG(WARNING, "Filter type"
+                                       "(%d) not supported", filter_type);
+                               break;
+                       }
+                       TAILQ_REMOVE(&igb_flow_list,
+                                igb_flow_mem_ptr,
+                                entries);
+                       rte_free(igb_flow_mem_ptr->flow);
+                       rte_free(igb_flow_mem_ptr);
+               }
+       }
+}
+
+/*  Destroy all flow rules associated with a port on igb. */
+static int
+igb_flow_flush(struct rte_eth_dev *dev,
+               __rte_unused struct rte_flow_error *error)
+{
+       igb_clear_all_ntuple_filter(dev);
+       igb_clear_all_ethertype_filter(dev);
+       igb_clear_syn_filter(dev);
+       igb_clear_all_flex_filter(dev);
+       igb_clear_rss_filter(dev);
+       igb_filterlist_flush(dev);
+
+       return 0;
 }
 
 const struct rte_flow_ops igb_flow_ops = {
-       igb_flow_validate,
-       NULL,
-       NULL,
-       NULL,
-       NULL,
+       .validate = igb_flow_validate,
+       .create = igb_flow_create,
+       .destroy = igb_flow_destroy,
+       .flush = igb_flow_flush,
 };