These changes are the raw update to linux-4.4.6-rt14. Kernel sources
[kvmfornfv.git] / kernel / drivers / net / ethernet / mellanox / mlx5 / core / en_flow_table.c
diff --git a/kernel/drivers/net/ethernet/mellanox/mlx5/core/en_flow_table.c b/kernel/drivers/net/ethernet/mellanox/mlx5/core/en_flow_table.c
new file mode 100644 (file)
index 0000000..22d603f
--- /dev/null
@@ -0,0 +1,907 @@
+/*
+ * Copyright (c) 2015, Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     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.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/list.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/tcp.h>
+#include <linux/mlx5/flow_table.h>
+#include "en.h"
+
+enum {
+       MLX5E_FULLMATCH = 0,
+       MLX5E_ALLMULTI  = 1,
+       MLX5E_PROMISC   = 2,
+};
+
+enum {
+       MLX5E_UC        = 0,
+       MLX5E_MC_IPV4   = 1,
+       MLX5E_MC_IPV6   = 2,
+       MLX5E_MC_OTHER  = 3,
+};
+
+enum {
+       MLX5E_ACTION_NONE = 0,
+       MLX5E_ACTION_ADD  = 1,
+       MLX5E_ACTION_DEL  = 2,
+};
+
+struct mlx5e_eth_addr_hash_node {
+       struct hlist_node          hlist;
+       u8                         action;
+       struct mlx5e_eth_addr_info ai;
+};
+
+static inline int mlx5e_hash_eth_addr(u8 *addr)
+{
+       return addr[5];
+}
+
+static void mlx5e_add_eth_addr_to_hash(struct hlist_head *hash, u8 *addr)
+{
+       struct mlx5e_eth_addr_hash_node *hn;
+       int ix = mlx5e_hash_eth_addr(addr);
+       int found = 0;
+
+       hlist_for_each_entry(hn, &hash[ix], hlist)
+               if (ether_addr_equal_64bits(hn->ai.addr, addr)) {
+                       found = 1;
+                       break;
+               }
+
+       if (found) {
+               hn->action = MLX5E_ACTION_NONE;
+               return;
+       }
+
+       hn = kzalloc(sizeof(*hn), GFP_ATOMIC);
+       if (!hn)
+               return;
+
+       ether_addr_copy(hn->ai.addr, addr);
+       hn->action = MLX5E_ACTION_ADD;
+
+       hlist_add_head(&hn->hlist, &hash[ix]);
+}
+
+static void mlx5e_del_eth_addr_from_hash(struct mlx5e_eth_addr_hash_node *hn)
+{
+       hlist_del(&hn->hlist);
+       kfree(hn);
+}
+
+static void mlx5e_del_eth_addr_from_flow_table(struct mlx5e_priv *priv,
+                                              struct mlx5e_eth_addr_info *ai)
+{
+       void *ft = priv->ft.main;
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV6_IPSEC_ESP))
+               mlx5_del_flow_table_entry(ft,
+                                         ai->ft_ix[MLX5E_TT_IPV6_IPSEC_ESP]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV4_IPSEC_ESP))
+               mlx5_del_flow_table_entry(ft,
+                                         ai->ft_ix[MLX5E_TT_IPV4_IPSEC_ESP]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV6_IPSEC_AH))
+               mlx5_del_flow_table_entry(ft,
+                                         ai->ft_ix[MLX5E_TT_IPV6_IPSEC_AH]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV4_IPSEC_AH))
+               mlx5_del_flow_table_entry(ft,
+                                         ai->ft_ix[MLX5E_TT_IPV4_IPSEC_AH]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV6_TCP))
+               mlx5_del_flow_table_entry(ft, ai->ft_ix[MLX5E_TT_IPV6_TCP]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV4_TCP))
+               mlx5_del_flow_table_entry(ft, ai->ft_ix[MLX5E_TT_IPV4_TCP]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV6_UDP))
+               mlx5_del_flow_table_entry(ft, ai->ft_ix[MLX5E_TT_IPV6_UDP]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV4_UDP))
+               mlx5_del_flow_table_entry(ft, ai->ft_ix[MLX5E_TT_IPV4_UDP]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV6))
+               mlx5_del_flow_table_entry(ft, ai->ft_ix[MLX5E_TT_IPV6]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_IPV4))
+               mlx5_del_flow_table_entry(ft, ai->ft_ix[MLX5E_TT_IPV4]);
+
+       if (ai->tt_vec & BIT(MLX5E_TT_ANY))
+               mlx5_del_flow_table_entry(ft, ai->ft_ix[MLX5E_TT_ANY]);
+}
+
+static int mlx5e_get_eth_addr_type(u8 *addr)
+{
+       if (is_unicast_ether_addr(addr))
+               return MLX5E_UC;
+
+       if ((addr[0] == 0x01) &&
+           (addr[1] == 0x00) &&
+           (addr[2] == 0x5e) &&
+          !(addr[3] &  0x80))
+               return MLX5E_MC_IPV4;
+
+       if ((addr[0] == 0x33) &&
+           (addr[1] == 0x33))
+               return MLX5E_MC_IPV6;
+
+       return MLX5E_MC_OTHER;
+}
+
+static u32 mlx5e_get_tt_vec(struct mlx5e_eth_addr_info *ai, int type)
+{
+       int eth_addr_type;
+       u32 ret;
+
+       switch (type) {
+       case MLX5E_FULLMATCH:
+               eth_addr_type = mlx5e_get_eth_addr_type(ai->addr);
+               switch (eth_addr_type) {
+               case MLX5E_UC:
+                       ret =
+                               BIT(MLX5E_TT_IPV4_TCP)       |
+                               BIT(MLX5E_TT_IPV6_TCP)       |
+                               BIT(MLX5E_TT_IPV4_UDP)       |
+                               BIT(MLX5E_TT_IPV6_UDP)       |
+                               BIT(MLX5E_TT_IPV4_IPSEC_AH)  |
+                               BIT(MLX5E_TT_IPV6_IPSEC_AH)  |
+                               BIT(MLX5E_TT_IPV4_IPSEC_ESP) |
+                               BIT(MLX5E_TT_IPV6_IPSEC_ESP) |
+                               BIT(MLX5E_TT_IPV4)           |
+                               BIT(MLX5E_TT_IPV6)           |
+                               BIT(MLX5E_TT_ANY)            |
+                               0;
+                       break;
+
+               case MLX5E_MC_IPV4:
+                       ret =
+                               BIT(MLX5E_TT_IPV4_UDP)       |
+                               BIT(MLX5E_TT_IPV4)           |
+                               0;
+                       break;
+
+               case MLX5E_MC_IPV6:
+                       ret =
+                               BIT(MLX5E_TT_IPV6_UDP)       |
+                               BIT(MLX5E_TT_IPV6)           |
+                               0;
+                       break;
+
+               case MLX5E_MC_OTHER:
+                       ret =
+                               BIT(MLX5E_TT_ANY)            |
+                               0;
+                       break;
+               }
+
+               break;
+
+       case MLX5E_ALLMULTI:
+               ret =
+                       BIT(MLX5E_TT_IPV4_UDP) |
+                       BIT(MLX5E_TT_IPV6_UDP) |
+                       BIT(MLX5E_TT_IPV4)     |
+                       BIT(MLX5E_TT_IPV6)     |
+                       BIT(MLX5E_TT_ANY)      |
+                       0;
+               break;
+
+       default: /* MLX5E_PROMISC */
+               ret =
+                       BIT(MLX5E_TT_IPV4_TCP)       |
+                       BIT(MLX5E_TT_IPV6_TCP)       |
+                       BIT(MLX5E_TT_IPV4_UDP)       |
+                       BIT(MLX5E_TT_IPV6_UDP)       |
+                       BIT(MLX5E_TT_IPV4_IPSEC_AH)  |
+                       BIT(MLX5E_TT_IPV6_IPSEC_AH)  |
+                       BIT(MLX5E_TT_IPV4_IPSEC_ESP) |
+                       BIT(MLX5E_TT_IPV6_IPSEC_ESP) |
+                       BIT(MLX5E_TT_IPV4)           |
+                       BIT(MLX5E_TT_IPV6)           |
+                       BIT(MLX5E_TT_ANY)            |
+                       0;
+               break;
+       }
+
+       return ret;
+}
+
+static int __mlx5e_add_eth_addr_rule(struct mlx5e_priv *priv,
+                                    struct mlx5e_eth_addr_info *ai, int type,
+                                    void *flow_context, void *match_criteria)
+{
+       u8 match_criteria_enable = 0;
+       void *match_value;
+       void *dest;
+       u8   *dmac;
+       u8   *match_criteria_dmac;
+       void *ft   = priv->ft.main;
+       u32  *tirn = priv->tirn;
+       u32  *ft_ix;
+       u32  tt_vec;
+       int  err;
+
+       match_value = MLX5_ADDR_OF(flow_context, flow_context, match_value);
+       dmac = MLX5_ADDR_OF(fte_match_param, match_value,
+                           outer_headers.dmac_47_16);
+       match_criteria_dmac = MLX5_ADDR_OF(fte_match_param, match_criteria,
+                                          outer_headers.dmac_47_16);
+       dest = MLX5_ADDR_OF(flow_context, flow_context, destination);
+
+       MLX5_SET(flow_context, flow_context, action,
+                MLX5_FLOW_CONTEXT_ACTION_FWD_DEST);
+       MLX5_SET(flow_context, flow_context, destination_list_size, 1);
+       MLX5_SET(dest_format_struct, dest, destination_type,
+                MLX5_FLOW_CONTEXT_DEST_TYPE_TIR);
+
+       switch (type) {
+       case MLX5E_FULLMATCH:
+               match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+               memset(match_criteria_dmac, 0xff, ETH_ALEN);
+               ether_addr_copy(dmac, ai->addr);
+               break;
+
+       case MLX5E_ALLMULTI:
+               match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+               match_criteria_dmac[0] = 0x01;
+               dmac[0] = 0x01;
+               break;
+
+       case MLX5E_PROMISC:
+               break;
+       }
+
+       tt_vec = mlx5e_get_tt_vec(ai, type);
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_ANY];
+       if (tt_vec & BIT(MLX5E_TT_ANY)) {
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_ANY]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_ANY);
+       }
+
+       match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       MLX5_SET_TO_ONES(fte_match_param, match_criteria,
+                        outer_headers.ethertype);
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV4];
+       if (tt_vec & BIT(MLX5E_TT_IPV4)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IP);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV4]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV4);
+       }
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV6];
+       if (tt_vec & BIT(MLX5E_TT_IPV6)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IPV6);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV6]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV6);
+       }
+
+       MLX5_SET_TO_ONES(fte_match_param, match_criteria,
+                        outer_headers.ip_protocol);
+       MLX5_SET(fte_match_param, match_value, outer_headers.ip_protocol,
+                IPPROTO_UDP);
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV4_UDP];
+       if (tt_vec & BIT(MLX5E_TT_IPV4_UDP)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IP);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV4_UDP]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV4_UDP);
+       }
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV6_UDP];
+       if (tt_vec & BIT(MLX5E_TT_IPV6_UDP)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IPV6);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV6_UDP]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV6_UDP);
+       }
+
+       MLX5_SET(fte_match_param, match_value, outer_headers.ip_protocol,
+                IPPROTO_TCP);
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV4_TCP];
+       if (tt_vec & BIT(MLX5E_TT_IPV4_TCP)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IP);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV4_TCP]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV4_TCP);
+       }
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV6_TCP];
+       if (tt_vec & BIT(MLX5E_TT_IPV6_TCP)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IPV6);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV6_TCP]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV6_TCP);
+       }
+
+       MLX5_SET(fte_match_param, match_value, outer_headers.ip_protocol,
+                IPPROTO_AH);
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV4_IPSEC_AH];
+       if (tt_vec & BIT(MLX5E_TT_IPV4_IPSEC_AH)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IP);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV4_IPSEC_AH]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV4_IPSEC_AH);
+       }
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV6_IPSEC_AH];
+       if (tt_vec & BIT(MLX5E_TT_IPV6_IPSEC_AH)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IPV6);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV6_IPSEC_AH]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV6_IPSEC_AH);
+       }
+
+       MLX5_SET(fte_match_param, match_value, outer_headers.ip_protocol,
+                IPPROTO_ESP);
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV4_IPSEC_ESP];
+       if (tt_vec & BIT(MLX5E_TT_IPV4_IPSEC_ESP)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IP);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV4_IPSEC_ESP]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV4_IPSEC_ESP);
+       }
+
+       ft_ix = &ai->ft_ix[MLX5E_TT_IPV6_IPSEC_ESP];
+       if (tt_vec & BIT(MLX5E_TT_IPV6_IPSEC_ESP)) {
+               MLX5_SET(fte_match_param, match_value, outer_headers.ethertype,
+                        ETH_P_IPV6);
+               MLX5_SET(dest_format_struct, dest, destination_id,
+                        tirn[MLX5E_TT_IPV6_IPSEC_ESP]);
+               err = mlx5_add_flow_table_entry(ft, match_criteria_enable,
+                                               match_criteria, flow_context,
+                                               ft_ix);
+               if (err)
+                       goto err_del_ai;
+
+               ai->tt_vec |= BIT(MLX5E_TT_IPV6_IPSEC_ESP);
+       }
+
+       return 0;
+
+err_del_ai:
+       mlx5e_del_eth_addr_from_flow_table(priv, ai);
+
+       return err;
+}
+
+static int mlx5e_add_eth_addr_rule(struct mlx5e_priv *priv,
+                                  struct mlx5e_eth_addr_info *ai, int type)
+{
+       u32 *flow_context;
+       u32 *match_criteria;
+       int err;
+
+       flow_context   = mlx5_vzalloc(MLX5_ST_SZ_BYTES(flow_context) +
+                                     MLX5_ST_SZ_BYTES(dest_format_struct));
+       match_criteria = mlx5_vzalloc(MLX5_ST_SZ_BYTES(fte_match_param));
+       if (!flow_context || !match_criteria) {
+               netdev_err(priv->netdev, "%s: alloc failed\n", __func__);
+               err = -ENOMEM;
+               goto add_eth_addr_rule_out;
+       }
+
+       err = __mlx5e_add_eth_addr_rule(priv, ai, type, flow_context,
+                                       match_criteria);
+       if (err)
+               netdev_err(priv->netdev, "%s: failed\n", __func__);
+
+add_eth_addr_rule_out:
+       kvfree(match_criteria);
+       kvfree(flow_context);
+       return err;
+}
+
+enum mlx5e_vlan_rule_type {
+       MLX5E_VLAN_RULE_TYPE_UNTAGGED,
+       MLX5E_VLAN_RULE_TYPE_ANY_VID,
+       MLX5E_VLAN_RULE_TYPE_MATCH_VID,
+};
+
+static int mlx5e_add_vlan_rule(struct mlx5e_priv *priv,
+                              enum mlx5e_vlan_rule_type rule_type, u16 vid)
+{
+       u8 match_criteria_enable = 0;
+       u32 *flow_context;
+       void *match_value;
+       void *dest;
+       u32 *match_criteria;
+       u32 *ft_ix;
+       int err;
+
+       flow_context   = mlx5_vzalloc(MLX5_ST_SZ_BYTES(flow_context) +
+                                     MLX5_ST_SZ_BYTES(dest_format_struct));
+       match_criteria = mlx5_vzalloc(MLX5_ST_SZ_BYTES(fte_match_param));
+       if (!flow_context || !match_criteria) {
+               netdev_err(priv->netdev, "%s: alloc failed\n", __func__);
+               err = -ENOMEM;
+               goto add_vlan_rule_out;
+       }
+       match_value = MLX5_ADDR_OF(flow_context, flow_context, match_value);
+       dest = MLX5_ADDR_OF(flow_context, flow_context, destination);
+
+       MLX5_SET(flow_context, flow_context, action,
+                MLX5_FLOW_CONTEXT_ACTION_FWD_DEST);
+       MLX5_SET(flow_context, flow_context, destination_list_size, 1);
+       MLX5_SET(dest_format_struct, dest, destination_type,
+                MLX5_FLOW_CONTEXT_DEST_TYPE_FLOW_TABLE);
+       MLX5_SET(dest_format_struct, dest, destination_id,
+                mlx5_get_flow_table_id(priv->ft.main));
+
+       match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       MLX5_SET_TO_ONES(fte_match_param, match_criteria,
+                        outer_headers.vlan_tag);
+
+       switch (rule_type) {
+       case MLX5E_VLAN_RULE_TYPE_UNTAGGED:
+               ft_ix = &priv->vlan.untagged_rule_ft_ix;
+               break;
+       case MLX5E_VLAN_RULE_TYPE_ANY_VID:
+               ft_ix = &priv->vlan.any_vlan_rule_ft_ix;
+               MLX5_SET(fte_match_param, match_value, outer_headers.vlan_tag,
+                        1);
+               break;
+       default: /* MLX5E_VLAN_RULE_TYPE_MATCH_VID */
+               ft_ix = &priv->vlan.active_vlans_ft_ix[vid];
+               MLX5_SET(fte_match_param, match_value, outer_headers.vlan_tag,
+                        1);
+               MLX5_SET_TO_ONES(fte_match_param, match_criteria,
+                                outer_headers.first_vid);
+               MLX5_SET(fte_match_param, match_value, outer_headers.first_vid,
+                        vid);
+               break;
+       }
+
+       err = mlx5_add_flow_table_entry(priv->ft.vlan, match_criteria_enable,
+                                       match_criteria, flow_context, ft_ix);
+       if (err)
+               netdev_err(priv->netdev, "%s: failed\n", __func__);
+
+add_vlan_rule_out:
+       kvfree(match_criteria);
+       kvfree(flow_context);
+       return err;
+}
+
+static void mlx5e_del_vlan_rule(struct mlx5e_priv *priv,
+                               enum mlx5e_vlan_rule_type rule_type, u16 vid)
+{
+       switch (rule_type) {
+       case MLX5E_VLAN_RULE_TYPE_UNTAGGED:
+               mlx5_del_flow_table_entry(priv->ft.vlan,
+                                         priv->vlan.untagged_rule_ft_ix);
+               break;
+       case MLX5E_VLAN_RULE_TYPE_ANY_VID:
+               mlx5_del_flow_table_entry(priv->ft.vlan,
+                                         priv->vlan.any_vlan_rule_ft_ix);
+               break;
+       case MLX5E_VLAN_RULE_TYPE_MATCH_VID:
+               mlx5_del_flow_table_entry(priv->ft.vlan,
+                                         priv->vlan.active_vlans_ft_ix[vid]);
+               break;
+       }
+}
+
+void mlx5e_enable_vlan_filter(struct mlx5e_priv *priv)
+{
+       if (!priv->vlan.filter_disabled)
+               return;
+
+       priv->vlan.filter_disabled = false;
+       if (priv->netdev->flags & IFF_PROMISC)
+               return;
+       mlx5e_del_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_ANY_VID, 0);
+}
+
+void mlx5e_disable_vlan_filter(struct mlx5e_priv *priv)
+{
+       if (priv->vlan.filter_disabled)
+               return;
+
+       priv->vlan.filter_disabled = true;
+       if (priv->netdev->flags & IFF_PROMISC)
+               return;
+       mlx5e_add_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_ANY_VID, 0);
+}
+
+int mlx5e_vlan_rx_add_vid(struct net_device *dev, __always_unused __be16 proto,
+                         u16 vid)
+{
+       struct mlx5e_priv *priv = netdev_priv(dev);
+
+       return mlx5e_add_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_MATCH_VID, vid);
+}
+
+int mlx5e_vlan_rx_kill_vid(struct net_device *dev, __always_unused __be16 proto,
+                          u16 vid)
+{
+       struct mlx5e_priv *priv = netdev_priv(dev);
+
+       mlx5e_del_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_MATCH_VID, vid);
+
+       return 0;
+}
+
+#define mlx5e_for_each_hash_node(hn, tmp, hash, i) \
+       for (i = 0; i < MLX5E_ETH_ADDR_HASH_SIZE; i++) \
+               hlist_for_each_entry_safe(hn, tmp, &hash[i], hlist)
+
+static void mlx5e_execute_action(struct mlx5e_priv *priv,
+                                struct mlx5e_eth_addr_hash_node *hn)
+{
+       switch (hn->action) {
+       case MLX5E_ACTION_ADD:
+               mlx5e_add_eth_addr_rule(priv, &hn->ai, MLX5E_FULLMATCH);
+               hn->action = MLX5E_ACTION_NONE;
+               break;
+
+       case MLX5E_ACTION_DEL:
+               mlx5e_del_eth_addr_from_flow_table(priv, &hn->ai);
+               mlx5e_del_eth_addr_from_hash(hn);
+               break;
+       }
+}
+
+static void mlx5e_sync_netdev_addr(struct mlx5e_priv *priv)
+{
+       struct net_device *netdev = priv->netdev;
+       struct netdev_hw_addr *ha;
+
+       netif_addr_lock_bh(netdev);
+
+       mlx5e_add_eth_addr_to_hash(priv->eth_addr.netdev_uc,
+                                  priv->netdev->dev_addr);
+
+       netdev_for_each_uc_addr(ha, netdev)
+               mlx5e_add_eth_addr_to_hash(priv->eth_addr.netdev_uc, ha->addr);
+
+       netdev_for_each_mc_addr(ha, netdev)
+               mlx5e_add_eth_addr_to_hash(priv->eth_addr.netdev_mc, ha->addr);
+
+       netif_addr_unlock_bh(netdev);
+}
+
+static void mlx5e_apply_netdev_addr(struct mlx5e_priv *priv)
+{
+       struct mlx5e_eth_addr_hash_node *hn;
+       struct hlist_node *tmp;
+       int i;
+
+       mlx5e_for_each_hash_node(hn, tmp, priv->eth_addr.netdev_uc, i)
+               mlx5e_execute_action(priv, hn);
+
+       mlx5e_for_each_hash_node(hn, tmp, priv->eth_addr.netdev_mc, i)
+               mlx5e_execute_action(priv, hn);
+}
+
+static void mlx5e_handle_netdev_addr(struct mlx5e_priv *priv)
+{
+       struct mlx5e_eth_addr_hash_node *hn;
+       struct hlist_node *tmp;
+       int i;
+
+       mlx5e_for_each_hash_node(hn, tmp, priv->eth_addr.netdev_uc, i)
+               hn->action = MLX5E_ACTION_DEL;
+       mlx5e_for_each_hash_node(hn, tmp, priv->eth_addr.netdev_mc, i)
+               hn->action = MLX5E_ACTION_DEL;
+
+       if (!test_bit(MLX5E_STATE_DESTROYING, &priv->state))
+               mlx5e_sync_netdev_addr(priv);
+
+       mlx5e_apply_netdev_addr(priv);
+}
+
+void mlx5e_set_rx_mode_work(struct work_struct *work)
+{
+       struct mlx5e_priv *priv = container_of(work, struct mlx5e_priv,
+                                              set_rx_mode_work);
+
+       struct mlx5e_eth_addr_db *ea = &priv->eth_addr;
+       struct net_device *ndev = priv->netdev;
+
+       bool rx_mode_enable   = !test_bit(MLX5E_STATE_DESTROYING, &priv->state);
+       bool promisc_enabled   = rx_mode_enable && (ndev->flags & IFF_PROMISC);
+       bool allmulti_enabled  = rx_mode_enable && (ndev->flags & IFF_ALLMULTI);
+       bool broadcast_enabled = rx_mode_enable;
+
+       bool enable_promisc    = !ea->promisc_enabled   &&  promisc_enabled;
+       bool disable_promisc   =  ea->promisc_enabled   && !promisc_enabled;
+       bool enable_allmulti   = !ea->allmulti_enabled  &&  allmulti_enabled;
+       bool disable_allmulti  =  ea->allmulti_enabled  && !allmulti_enabled;
+       bool enable_broadcast  = !ea->broadcast_enabled &&  broadcast_enabled;
+       bool disable_broadcast =  ea->broadcast_enabled && !broadcast_enabled;
+
+       if (enable_promisc) {
+               mlx5e_add_eth_addr_rule(priv, &ea->promisc, MLX5E_PROMISC);
+               if (!priv->vlan.filter_disabled)
+                       mlx5e_add_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_ANY_VID,
+                                           0);
+       }
+       if (enable_allmulti)
+               mlx5e_add_eth_addr_rule(priv, &ea->allmulti, MLX5E_ALLMULTI);
+       if (enable_broadcast)
+               mlx5e_add_eth_addr_rule(priv, &ea->broadcast, MLX5E_FULLMATCH);
+
+       mlx5e_handle_netdev_addr(priv);
+
+       if (disable_broadcast)
+               mlx5e_del_eth_addr_from_flow_table(priv, &ea->broadcast);
+       if (disable_allmulti)
+               mlx5e_del_eth_addr_from_flow_table(priv, &ea->allmulti);
+       if (disable_promisc) {
+               if (!priv->vlan.filter_disabled)
+                       mlx5e_del_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_ANY_VID,
+                                           0);
+               mlx5e_del_eth_addr_from_flow_table(priv, &ea->promisc);
+       }
+
+       ea->promisc_enabled   = promisc_enabled;
+       ea->allmulti_enabled  = allmulti_enabled;
+       ea->broadcast_enabled = broadcast_enabled;
+}
+
+void mlx5e_init_eth_addr(struct mlx5e_priv *priv)
+{
+       ether_addr_copy(priv->eth_addr.broadcast.addr, priv->netdev->broadcast);
+}
+
+static int mlx5e_create_main_flow_table(struct mlx5e_priv *priv)
+{
+       struct mlx5_flow_table_group *g;
+       u8 *dmac;
+
+       g = kcalloc(9, sizeof(*g), GFP_KERNEL);
+       if (!g)
+               return -ENOMEM;
+
+       g[0].log_sz = 3;
+       g[0].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       MLX5_SET_TO_ONES(fte_match_param, g[0].match_criteria,
+                        outer_headers.ethertype);
+       MLX5_SET_TO_ONES(fte_match_param, g[0].match_criteria,
+                        outer_headers.ip_protocol);
+
+       g[1].log_sz = 1;
+       g[1].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       MLX5_SET_TO_ONES(fte_match_param, g[1].match_criteria,
+                        outer_headers.ethertype);
+
+       g[2].log_sz = 0;
+
+       g[3].log_sz = 14;
+       g[3].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       dmac = MLX5_ADDR_OF(fte_match_param, g[3].match_criteria,
+                           outer_headers.dmac_47_16);
+       memset(dmac, 0xff, ETH_ALEN);
+       MLX5_SET_TO_ONES(fte_match_param, g[3].match_criteria,
+                        outer_headers.ethertype);
+       MLX5_SET_TO_ONES(fte_match_param, g[3].match_criteria,
+                        outer_headers.ip_protocol);
+
+       g[4].log_sz = 13;
+       g[4].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       dmac = MLX5_ADDR_OF(fte_match_param, g[4].match_criteria,
+                           outer_headers.dmac_47_16);
+       memset(dmac, 0xff, ETH_ALEN);
+       MLX5_SET_TO_ONES(fte_match_param, g[4].match_criteria,
+                        outer_headers.ethertype);
+
+       g[5].log_sz = 11;
+       g[5].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       dmac = MLX5_ADDR_OF(fte_match_param, g[5].match_criteria,
+                           outer_headers.dmac_47_16);
+       memset(dmac, 0xff, ETH_ALEN);
+
+       g[6].log_sz = 2;
+       g[6].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       dmac = MLX5_ADDR_OF(fte_match_param, g[6].match_criteria,
+                           outer_headers.dmac_47_16);
+       dmac[0] = 0x01;
+       MLX5_SET_TO_ONES(fte_match_param, g[6].match_criteria,
+                        outer_headers.ethertype);
+       MLX5_SET_TO_ONES(fte_match_param, g[6].match_criteria,
+                        outer_headers.ip_protocol);
+
+       g[7].log_sz = 1;
+       g[7].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       dmac = MLX5_ADDR_OF(fte_match_param, g[7].match_criteria,
+                           outer_headers.dmac_47_16);
+       dmac[0] = 0x01;
+       MLX5_SET_TO_ONES(fte_match_param, g[7].match_criteria,
+                        outer_headers.ethertype);
+
+       g[8].log_sz = 0;
+       g[8].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       dmac = MLX5_ADDR_OF(fte_match_param, g[8].match_criteria,
+                           outer_headers.dmac_47_16);
+       dmac[0] = 0x01;
+       priv->ft.main = mlx5_create_flow_table(priv->mdev, 1,
+                                              MLX5_FLOW_TABLE_TYPE_NIC_RCV,
+                                              9, g);
+       kfree(g);
+
+       return priv->ft.main ? 0 : -ENOMEM;
+}
+
+static void mlx5e_destroy_main_flow_table(struct mlx5e_priv *priv)
+{
+       mlx5_destroy_flow_table(priv->ft.main);
+}
+
+static int mlx5e_create_vlan_flow_table(struct mlx5e_priv *priv)
+{
+       struct mlx5_flow_table_group *g;
+
+       g = kcalloc(2, sizeof(*g), GFP_KERNEL);
+       if (!g)
+               return -ENOMEM;
+
+       g[0].log_sz = 12;
+       g[0].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       MLX5_SET_TO_ONES(fte_match_param, g[0].match_criteria,
+                        outer_headers.vlan_tag);
+       MLX5_SET_TO_ONES(fte_match_param, g[0].match_criteria,
+                        outer_headers.first_vid);
+
+       /* untagged + any vlan id */
+       g[1].log_sz = 1;
+       g[1].match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+       MLX5_SET_TO_ONES(fte_match_param, g[1].match_criteria,
+                        outer_headers.vlan_tag);
+
+       priv->ft.vlan = mlx5_create_flow_table(priv->mdev, 0,
+                                              MLX5_FLOW_TABLE_TYPE_NIC_RCV,
+                                              2, g);
+
+       kfree(g);
+       return priv->ft.vlan ? 0 : -ENOMEM;
+}
+
+static void mlx5e_destroy_vlan_flow_table(struct mlx5e_priv *priv)
+{
+       mlx5_destroy_flow_table(priv->ft.vlan);
+}
+
+int mlx5e_create_flow_tables(struct mlx5e_priv *priv)
+{
+       int err;
+
+       err = mlx5e_create_main_flow_table(priv);
+       if (err)
+               return err;
+
+       err = mlx5e_create_vlan_flow_table(priv);
+       if (err)
+               goto err_destroy_main_flow_table;
+
+       err = mlx5e_add_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_UNTAGGED, 0);
+       if (err)
+               goto err_destroy_vlan_flow_table;
+
+       return 0;
+
+err_destroy_vlan_flow_table:
+       mlx5e_destroy_vlan_flow_table(priv);
+
+err_destroy_main_flow_table:
+       mlx5e_destroy_main_flow_table(priv);
+
+       return err;
+}
+
+void mlx5e_destroy_flow_tables(struct mlx5e_priv *priv)
+{
+       mlx5e_del_vlan_rule(priv, MLX5E_VLAN_RULE_TYPE_UNTAGGED, 0);
+       mlx5e_destroy_vlan_flow_table(priv);
+       mlx5e_destroy_main_flow_table(priv);
+}