Quelle  bridge.c   Sprache: C

// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
/* Copyright (c) 2021 Mellanox Technologies. */

#include <linux/build_bug.h>
#include <linux/list.h>
#include <linux/notifier.h>
#include <net/netevent.h>
#include <net/switchdev.h>
#include "lib/devcom.h"
#include "bridge.h"
#include "eswitch.h"
#include "bridge_priv.h"
#define CREATE_TRACE_POINTS
#include "diag/bridge_tracepoint.h"

static const struct rhashtable_params fdb_ht_params = {
 .key_offset = offsetof(struct mlx5_esw_bridge_fdb_entry, key),
 .key_len = sizeof(struct mlx5_esw_bridge_fdb_key),
 .head_offset = offsetof(struct mlx5_esw_bridge_fdb_entry, ht_node),
 .automatic_shrinking = true,
};

static void
mlx5_esw_bridge_fdb_offload_notify(struct net_device *dev, const unsigned char *addr, u16 vid,
       unsigned long val)
{
 struct switchdev_notifier_fdb_info send_info = {};

 send_info.addr = addr;
 send_info.vid = vid;
 send_info.offloaded = true;
 call_switchdev_notifiers(val, dev, &send_info.info, NULL);
}

static void
mlx5_esw_bridge_fdb_del_notify(struct mlx5_esw_bridge_fdb_entry *entry)
{
 if (!(entry->flags & (MLX5_ESW_BRIDGE_FLAG_ADDED_BY_USER | MLX5_ESW_BRIDGE_FLAG_PEER)))
  mlx5_esw_bridge_fdb_offload_notify(entry->dev, entry->key.addr,
         entry->key.vid,
         SWITCHDEV_FDB_DEL_TO_BRIDGE);
}

static bool mlx5_esw_bridge_pkt_reformat_vlan_pop_supported(struct mlx5_eswitch *esw)
{
 return BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, reformat_remove)) &&
  MLX5_CAP_GEN_2(esw->dev, max_reformat_remove_size) >= sizeof(struct vlan_hdr) &&
  MLX5_CAP_GEN_2(esw->dev, max_reformat_remove_offset) >=
  offsetof(struct vlan_ethhdr, h_vlan_proto);
}

static struct mlx5_pkt_reformat *
mlx5_esw_bridge_pkt_reformat_vlan_pop_create(struct mlx5_eswitch *esw)
{
 struct mlx5_pkt_reformat_params reformat_params = {};

 reformat_params.type = MLX5_REFORMAT_TYPE_REMOVE_HDR;
 reformat_params.param_0 = MLX5_REFORMAT_CONTEXT_ANCHOR_MAC_START;
 reformat_params.param_1 = offsetof(struct vlan_ethhdr, h_vlan_proto);
 reformat_params.size = sizeof(struct vlan_hdr);
 return mlx5_packet_reformat_alloc(esw->dev, &reformat_params, MLX5_FLOW_NAMESPACE_FDB);
}

struct mlx5_flow_table *
mlx5_esw_bridge_table_create(int max_fte, u32 level, struct mlx5_eswitch *esw)
{
 struct mlx5_flow_table_attr ft_attr = {};
 struct mlx5_core_dev *dev = esw->dev;
 struct mlx5_flow_namespace *ns;
 struct mlx5_flow_table *fdb;

 ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_FDB);
 if (!ns) {
  esw_warn(dev, "Failed to get FDB namespace\n");
  return ERR_PTR(-ENOENT);
 }

 ft_attr.flags = MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
 ft_attr.max_fte = max_fte;
 ft_attr.level = level;
 ft_attr.prio = FDB_BR_OFFLOAD;
 fdb = mlx5_create_flow_table(ns, &ft_attr);
 if (IS_ERR(fdb))
  esw_warn(dev, "Failed to create bridge FDB Table (err=%ld)\n", PTR_ERR(fdb));

 return fdb;
}

static struct mlx5_flow_group *
mlx5_esw_bridge_ingress_vlan_proto_fg_create(unsigned int from, unsigned int to, u16 vlan_proto,
          struct mlx5_eswitch *esw,
          struct mlx5_flow_table *ingress_ft)
{
 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 struct mlx5_flow_group *fg;
 u32 *in, *match;

 in = kvzalloc(inlen, GFP_KERNEL);
 if (!in)
  return ERR_PTR(-ENOMEM);

 MLX5_SET(create_flow_group_in, in, match_criteria_enable,
   MLX5_MATCH_OUTER_HEADERS | MLX5_MATCH_MISC_PARAMETERS_2);
 match = MLX5_ADDR_OF(create_flow_group_in, in, match_criteria);

 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.smac_47_16);
 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.smac_15_0);
 if (vlan_proto == ETH_P_8021Q)
  MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.cvlan_tag);
 else if (vlan_proto == ETH_P_8021AD)
  MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.svlan_tag);
 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.first_vid);

 MLX5_SET(fte_match_param, match, misc_parameters_2.metadata_reg_c_0,
   mlx5_eswitch_get_vport_metadata_mask());

 MLX5_SET(create_flow_group_in, in, start_flow_index, from);
 MLX5_SET(create_flow_group_in, in, end_flow_index, to);

 fg = mlx5_create_flow_group(ingress_ft, in);
 kvfree(in);
 if (IS_ERR(fg))
  esw_warn(esw->dev,
    "Failed to create VLAN(proto=%x) flow group for bridge ingress table (err=%ld)\n",
    vlan_proto, PTR_ERR(fg));

 return fg;
}

static struct mlx5_flow_group *
mlx5_esw_bridge_ingress_vlan_fg_create(struct mlx5_eswitch *esw,
           struct mlx5_flow_table *ingress_ft)
{
 unsigned int from = MLX5_ESW_BRIDGE_INGRESS_TABLE_VLAN_GRP_IDX_FROM;
 unsigned int to = MLX5_ESW_BRIDGE_INGRESS_TABLE_VLAN_GRP_IDX_TO;

 return mlx5_esw_bridge_ingress_vlan_proto_fg_create(from, to, ETH_P_8021Q, esw, ingress_ft);
}

static struct mlx5_flow_group *
mlx5_esw_bridge_ingress_qinq_fg_create(struct mlx5_eswitch *esw,
           struct mlx5_flow_table *ingress_ft)
{
 unsigned int from = MLX5_ESW_BRIDGE_INGRESS_TABLE_QINQ_GRP_IDX_FROM;
 unsigned int to = MLX5_ESW_BRIDGE_INGRESS_TABLE_QINQ_GRP_IDX_TO;

 return mlx5_esw_bridge_ingress_vlan_proto_fg_create(from, to, ETH_P_8021AD, esw,
           ingress_ft);
}

static struct mlx5_flow_group *
mlx5_esw_bridge_ingress_vlan_proto_filter_fg_create(unsigned int from, unsigned int to,
          u16 vlan_proto, struct mlx5_eswitch *esw,
          struct mlx5_flow_table *ingress_ft)
{
 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 struct mlx5_flow_group *fg;
 u32 *in, *match;

 in = kvzalloc(inlen, GFP_KERNEL);
 if (!in)
  return ERR_PTR(-ENOMEM);

 MLX5_SET(create_flow_group_in, in, match_criteria_enable,
   MLX5_MATCH_OUTER_HEADERS | MLX5_MATCH_MISC_PARAMETERS_2);
 match = MLX5_ADDR_OF(create_flow_group_in, in, match_criteria);

 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.smac_47_16);
 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.smac_15_0);
 if (vlan_proto == ETH_P_8021Q)
  MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.cvlan_tag);
 else if (vlan_proto == ETH_P_8021AD)
  MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.svlan_tag);
 MLX5_SET(fte_match_param, match, misc_parameters_2.metadata_reg_c_0,
   mlx5_eswitch_get_vport_metadata_mask());

 MLX5_SET(create_flow_group_in, in, start_flow_index, from);
 MLX5_SET(create_flow_group_in, in, end_flow_index, to);

 fg = mlx5_create_flow_group(ingress_ft, in);
 if (IS_ERR(fg))
  esw_warn(esw->dev,
    "Failed to create bridge ingress table VLAN filter flow group (err=%ld)\n",
    PTR_ERR(fg));
 kvfree(in);
 return fg;
}

static struct mlx5_flow_group *
mlx5_esw_bridge_ingress_vlan_filter_fg_create(struct mlx5_eswitch *esw,
           struct mlx5_flow_table *ingress_ft)
{
 unsigned int from = MLX5_ESW_BRIDGE_INGRESS_TABLE_VLAN_FILTER_GRP_IDX_FROM;
 unsigned int to = MLX5_ESW_BRIDGE_INGRESS_TABLE_VLAN_FILTER_GRP_IDX_TO;

 return mlx5_esw_bridge_ingress_vlan_proto_filter_fg_create(from, to, ETH_P_8021Q, esw,
           ingress_ft);
}

static struct mlx5_flow_group *
mlx5_esw_bridge_ingress_qinq_filter_fg_create(struct mlx5_eswitch *esw,
           struct mlx5_flow_table *ingress_ft)
{
 unsigned int from = MLX5_ESW_BRIDGE_INGRESS_TABLE_QINQ_FILTER_GRP_IDX_FROM;
 unsigned int to = MLX5_ESW_BRIDGE_INGRESS_TABLE_QINQ_FILTER_GRP_IDX_TO;

 return mlx5_esw_bridge_ingress_vlan_proto_filter_fg_create(from, to, ETH_P_8021AD, esw,
           ingress_ft);
}

static struct mlx5_flow_group *
mlx5_esw_bridge_ingress_mac_fg_create(struct mlx5_eswitch *esw, struct mlx5_flow_table *ingress_ft)
{
 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 struct mlx5_flow_group *fg;
 u32 *in, *match;

 in = kvzalloc(inlen, GFP_KERNEL);
 if (!in)
  return ERR_PTR(-ENOMEM);

 MLX5_SET(create_flow_group_in, in, match_criteria_enable,
   MLX5_MATCH_OUTER_HEADERS | MLX5_MATCH_MISC_PARAMETERS_2);
 match = MLX5_ADDR_OF(create_flow_group_in, in, match_criteria);

 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.smac_47_16);
 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.smac_15_0);

 MLX5_SET(fte_match_param, match, misc_parameters_2.metadata_reg_c_0,
   mlx5_eswitch_get_vport_metadata_mask());

 MLX5_SET(create_flow_group_in, in, start_flow_index,
   MLX5_ESW_BRIDGE_INGRESS_TABLE_MAC_GRP_IDX_FROM);
 MLX5_SET(create_flow_group_in, in, end_flow_index,
   MLX5_ESW_BRIDGE_INGRESS_TABLE_MAC_GRP_IDX_TO);

 fg = mlx5_create_flow_group(ingress_ft, in);
 if (IS_ERR(fg))
  esw_warn(esw->dev,
    "Failed to create MAC flow group for bridge ingress table (err=%ld)\n",
    PTR_ERR(fg));

 kvfree(in);
 return fg;
}

static struct mlx5_flow_group *
mlx5_esw_bridge_egress_vlan_proto_fg_create(unsigned int from, unsigned int to, u16 vlan_proto,
         struct mlx5_eswitch *esw,
         struct mlx5_flow_table *egress_ft)
{
 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 struct mlx5_flow_group *fg;
 u32 *in, *match;

 in = kvzalloc(inlen, GFP_KERNEL);
 if (!in)
  return ERR_PTR(-ENOMEM);

 MLX5_SET(create_flow_group_in, in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
 match = MLX5_ADDR_OF(create_flow_group_in, in, match_criteria);

 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.dmac_47_16);
 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.dmac_15_0);
 if (vlan_proto == ETH_P_8021Q)
  MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.cvlan_tag);
 else if (vlan_proto == ETH_P_8021AD)
  MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.svlan_tag);
 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.first_vid);

 MLX5_SET(create_flow_group_in, in, start_flow_index, from);
 MLX5_SET(create_flow_group_in, in, end_flow_index, to);

 fg = mlx5_create_flow_group(egress_ft, in);
 if (IS_ERR(fg))
  esw_warn(esw->dev,
    "Failed to create VLAN flow group for bridge egress table (err=%ld)\n",
    PTR_ERR(fg));
 kvfree(in);
 return fg;
}

static struct mlx5_flow_group *
mlx5_esw_bridge_egress_vlan_fg_create(struct mlx5_eswitch *esw, struct mlx5_flow_table *egress_ft)
{
 unsigned int from = MLX5_ESW_BRIDGE_EGRESS_TABLE_VLAN_GRP_IDX_FROM;
 unsigned int to = MLX5_ESW_BRIDGE_EGRESS_TABLE_VLAN_GRP_IDX_TO;

 return mlx5_esw_bridge_egress_vlan_proto_fg_create(from, to, ETH_P_8021Q, esw, egress_ft);
}

static struct mlx5_flow_group *
mlx5_esw_bridge_egress_qinq_fg_create(struct mlx5_eswitch *esw,
          struct mlx5_flow_table *egress_ft)
{
 unsigned int from = MLX5_ESW_BRIDGE_EGRESS_TABLE_QINQ_GRP_IDX_FROM;
 unsigned int to = MLX5_ESW_BRIDGE_EGRESS_TABLE_QINQ_GRP_IDX_TO;

 return mlx5_esw_bridge_egress_vlan_proto_fg_create(from, to, ETH_P_8021AD, esw, egress_ft);
}

static struct mlx5_flow_group *
mlx5_esw_bridge_egress_mac_fg_create(struct mlx5_eswitch *esw, struct mlx5_flow_table *egress_ft)
{
 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 struct mlx5_flow_group *fg;
 u32 *in, *match;

 in = kvzalloc(inlen, GFP_KERNEL);
 if (!in)
  return ERR_PTR(-ENOMEM);

 MLX5_SET(create_flow_group_in, in, match_criteria_enable, MLX5_MATCH_OUTER_HEADERS);
 match = MLX5_ADDR_OF(create_flow_group_in, in, match_criteria);

 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.dmac_47_16);
 MLX5_SET_TO_ONES(fte_match_param, match, outer_headers.dmac_15_0);

 MLX5_SET(create_flow_group_in, in, start_flow_index,
   MLX5_ESW_BRIDGE_EGRESS_TABLE_MAC_GRP_IDX_FROM);
 MLX5_SET(create_flow_group_in, in, end_flow_index,
   MLX5_ESW_BRIDGE_EGRESS_TABLE_MAC_GRP_IDX_TO);

 fg = mlx5_create_flow_group(egress_ft, in);
 if (IS_ERR(fg))
  esw_warn(esw->dev,
    "Failed to create bridge egress table MAC flow group (err=%ld)\n",
    PTR_ERR(fg));
 kvfree(in);
 return fg;
}

static struct mlx5_flow_group *
mlx5_esw_bridge_egress_miss_fg_create(struct mlx5_eswitch *esw, struct mlx5_flow_table *egress_ft)
{
 int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
 struct mlx5_flow_group *fg;
 u32 *in, *match;

 in = kvzalloc(inlen, GFP_KERNEL);
 if (!in)
  return ERR_PTR(-ENOMEM);

 MLX5_SET(create_flow_group_in, in, match_criteria_enable, MLX5_MATCH_MISC_PARAMETERS_2);
 match = MLX5_ADDR_OF(create_flow_group_in, in, match_criteria);

 MLX5_SET(fte_match_param, match, misc_parameters_2.metadata_reg_c_1, ESW_TUN_MASK);

 MLX5_SET(create_flow_group_in, in, start_flow_index,
   MLX5_ESW_BRIDGE_EGRESS_TABLE_MISS_GRP_IDX_FROM);
 MLX5_SET(create_flow_group_in, in, end_flow_index,
   MLX5_ESW_BRIDGE_EGRESS_TABLE_MISS_GRP_IDX_TO);

 fg = mlx5_create_flow_group(egress_ft, in);
 if (IS_ERR(fg))
  esw_warn(esw->dev,
    "Failed to create bridge egress table miss flow group (err=%ld)\n",
    PTR_ERR(fg));
 kvfree(in);
 return fg;
}

static int
mlx5_esw_bridge_ingress_table_init(struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_flow_group *mac_fg, *qinq_filter_fg, *qinq_fg, *vlan_filter_fg, *vlan_fg;
 struct mlx5_flow_table *ingress_ft, *skip_ft;
 struct mlx5_eswitch *esw = br_offloads->esw;
 int err;

 if (!mlx5_eswitch_vport_match_metadata_enabled(esw))
  return -EOPNOTSUPP;

 ingress_ft = mlx5_esw_bridge_table_create(MLX5_ESW_BRIDGE_INGRESS_TABLE_SIZE,
        MLX5_ESW_BRIDGE_LEVEL_INGRESS_TABLE,
        esw);
 if (IS_ERR(ingress_ft))
  return PTR_ERR(ingress_ft);

 skip_ft = mlx5_esw_bridge_table_create(MLX5_ESW_BRIDGE_SKIP_TABLE_SIZE,
            MLX5_ESW_BRIDGE_LEVEL_SKIP_TABLE,
            esw);
 if (IS_ERR(skip_ft)) {
  err = PTR_ERR(skip_ft);
  goto err_skip_tbl;
 }

 vlan_fg = mlx5_esw_bridge_ingress_vlan_fg_create(esw, ingress_ft);
 if (IS_ERR(vlan_fg)) {
  err = PTR_ERR(vlan_fg);
  goto err_vlan_fg;
 }

 vlan_filter_fg = mlx5_esw_bridge_ingress_vlan_filter_fg_create(esw, ingress_ft);
 if (IS_ERR(vlan_filter_fg)) {
  err = PTR_ERR(vlan_filter_fg);
  goto err_vlan_filter_fg;
 }

 qinq_fg = mlx5_esw_bridge_ingress_qinq_fg_create(esw, ingress_ft);
 if (IS_ERR(qinq_fg)) {
  err = PTR_ERR(qinq_fg);
  goto err_qinq_fg;
 }

 qinq_filter_fg = mlx5_esw_bridge_ingress_qinq_filter_fg_create(esw, ingress_ft);
 if (IS_ERR(qinq_filter_fg)) {
  err = PTR_ERR(qinq_filter_fg);
  goto err_qinq_filter_fg;
 }

 mac_fg = mlx5_esw_bridge_ingress_mac_fg_create(esw, ingress_ft);
 if (IS_ERR(mac_fg)) {
  err = PTR_ERR(mac_fg);
  goto err_mac_fg;
 }

 br_offloads->ingress_ft = ingress_ft;
 br_offloads->skip_ft = skip_ft;
 br_offloads->ingress_vlan_fg = vlan_fg;
 br_offloads->ingress_vlan_filter_fg = vlan_filter_fg;
 br_offloads->ingress_qinq_fg = qinq_fg;
 br_offloads->ingress_qinq_filter_fg = qinq_filter_fg;
 br_offloads->ingress_mac_fg = mac_fg;
 return 0;

err_mac_fg:
 mlx5_destroy_flow_group(qinq_filter_fg);
err_qinq_filter_fg:
 mlx5_destroy_flow_group(qinq_fg);
err_qinq_fg:
 mlx5_destroy_flow_group(vlan_filter_fg);
err_vlan_filter_fg:
 mlx5_destroy_flow_group(vlan_fg);
err_vlan_fg:
 mlx5_destroy_flow_table(skip_ft);
err_skip_tbl:
 mlx5_destroy_flow_table(ingress_ft);
 return err;
}

static void
mlx5_esw_bridge_ingress_table_cleanup(struct mlx5_esw_bridge_offloads *br_offloads)
{
 mlx5_destroy_flow_group(br_offloads->ingress_mac_fg);
 br_offloads->ingress_mac_fg = NULL;
 mlx5_destroy_flow_group(br_offloads->ingress_qinq_filter_fg);
 br_offloads->ingress_qinq_filter_fg = NULL;
 mlx5_destroy_flow_group(br_offloads->ingress_qinq_fg);
 br_offloads->ingress_qinq_fg = NULL;
 mlx5_destroy_flow_group(br_offloads->ingress_vlan_filter_fg);
 br_offloads->ingress_vlan_filter_fg = NULL;
 mlx5_destroy_flow_group(br_offloads->ingress_vlan_fg);
 br_offloads->ingress_vlan_fg = NULL;
 mlx5_destroy_flow_table(br_offloads->skip_ft);
 br_offloads->skip_ft = NULL;
 mlx5_destroy_flow_table(br_offloads->ingress_ft);
 br_offloads->ingress_ft = NULL;
}

static struct mlx5_flow_handle *
mlx5_esw_bridge_egress_miss_flow_create(struct mlx5_flow_table *egress_ft,
     struct mlx5_flow_table *skip_ft,
     struct mlx5_pkt_reformat *pkt_reformat);

static int
mlx5_esw_bridge_egress_table_init(struct mlx5_esw_bridge_offloads *br_offloads,
      struct mlx5_esw_bridge *bridge)
{
 struct mlx5_flow_group *miss_fg = NULL, *mac_fg, *vlan_fg, *qinq_fg;
 struct mlx5_pkt_reformat *miss_pkt_reformat = NULL;
 struct mlx5_flow_handle *miss_handle = NULL;
 struct mlx5_eswitch *esw = br_offloads->esw;
 struct mlx5_flow_table *egress_ft;
 int err;

 egress_ft = mlx5_esw_bridge_table_create(MLX5_ESW_BRIDGE_EGRESS_TABLE_SIZE,
       MLX5_ESW_BRIDGE_LEVEL_EGRESS_TABLE,
       esw);
 if (IS_ERR(egress_ft))
  return PTR_ERR(egress_ft);

 vlan_fg = mlx5_esw_bridge_egress_vlan_fg_create(esw, egress_ft);
 if (IS_ERR(vlan_fg)) {
  err = PTR_ERR(vlan_fg);
  goto err_vlan_fg;
 }

 qinq_fg = mlx5_esw_bridge_egress_qinq_fg_create(esw, egress_ft);
 if (IS_ERR(qinq_fg)) {
  err = PTR_ERR(qinq_fg);
  goto err_qinq_fg;
 }

 mac_fg = mlx5_esw_bridge_egress_mac_fg_create(esw, egress_ft);
 if (IS_ERR(mac_fg)) {
  err = PTR_ERR(mac_fg);
  goto err_mac_fg;
 }

 if (mlx5_esw_bridge_pkt_reformat_vlan_pop_supported(esw)) {
  miss_fg = mlx5_esw_bridge_egress_miss_fg_create(esw, egress_ft);
  if (IS_ERR(miss_fg)) {
   esw_warn(esw->dev, "Failed to create miss flow group (err=%ld)\n",
     PTR_ERR(miss_fg));
   miss_fg = NULL;
   goto skip_miss_flow;
  }

  miss_pkt_reformat = mlx5_esw_bridge_pkt_reformat_vlan_pop_create(esw);
  if (IS_ERR(miss_pkt_reformat)) {
   esw_warn(esw->dev,
     "Failed to alloc packet reformat REMOVE_HEADER (err=%ld)\n",
     PTR_ERR(miss_pkt_reformat));
   miss_pkt_reformat = NULL;
   mlx5_destroy_flow_group(miss_fg);
   miss_fg = NULL;
   goto skip_miss_flow;
  }

  miss_handle = mlx5_esw_bridge_egress_miss_flow_create(egress_ft,
              br_offloads->skip_ft,
              miss_pkt_reformat);
  if (IS_ERR(miss_handle)) {
   esw_warn(esw->dev, "Failed to create miss flow (err=%ld)\n",
     PTR_ERR(miss_handle));
   miss_handle = NULL;
   mlx5_packet_reformat_dealloc(esw->dev, miss_pkt_reformat);
   miss_pkt_reformat = NULL;
   mlx5_destroy_flow_group(miss_fg);
   miss_fg = NULL;
   goto skip_miss_flow;
  }
 }
skip_miss_flow:

 bridge->egress_ft = egress_ft;
 bridge->egress_vlan_fg = vlan_fg;
 bridge->egress_qinq_fg = qinq_fg;
 bridge->egress_mac_fg = mac_fg;
 bridge->egress_miss_fg = miss_fg;
 bridge->egress_miss_pkt_reformat = miss_pkt_reformat;
 bridge->egress_miss_handle = miss_handle;
 return 0;

err_mac_fg:
 mlx5_destroy_flow_group(qinq_fg);
err_qinq_fg:
 mlx5_destroy_flow_group(vlan_fg);
err_vlan_fg:
 mlx5_destroy_flow_table(egress_ft);
 return err;
}

static void
mlx5_esw_bridge_egress_table_cleanup(struct mlx5_esw_bridge *bridge)
{
 if (bridge->egress_miss_handle)
  mlx5_del_flow_rules(bridge->egress_miss_handle);
 if (bridge->egress_miss_pkt_reformat)
  mlx5_packet_reformat_dealloc(bridge->br_offloads->esw->dev,
          bridge->egress_miss_pkt_reformat);
 if (bridge->egress_miss_fg)
  mlx5_destroy_flow_group(bridge->egress_miss_fg);
 mlx5_destroy_flow_group(bridge->egress_mac_fg);
 mlx5_destroy_flow_group(bridge->egress_qinq_fg);
 mlx5_destroy_flow_group(bridge->egress_vlan_fg);
 mlx5_destroy_flow_table(bridge->egress_ft);
}

static struct mlx5_flow_handle *
mlx5_esw_bridge_ingress_flow_with_esw_create(u16 vport_num, const unsigned char *addr,
          struct mlx5_esw_bridge_vlan *vlan,
          struct mlx5_fc *counter,
          struct mlx5_esw_bridge *bridge,
          struct mlx5_eswitch *esw)
{
 struct mlx5_esw_bridge_offloads *br_offloads = bridge->br_offloads;
 struct mlx5_flow_act flow_act = {
  .action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST | MLX5_FLOW_CONTEXT_ACTION_COUNT,
  .flags = FLOW_ACT_NO_APPEND,
 };
 struct mlx5_flow_destination dests[2] = {};
 struct mlx5_flow_spec *rule_spec;
 struct mlx5_flow_handle *handle;
 u8 *smac_v, *smac_c;

 rule_spec = kvzalloc(sizeof(*rule_spec), GFP_KERNEL);
 if (!rule_spec)
  return ERR_PTR(-ENOMEM);

 rule_spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS | MLX5_MATCH_MISC_PARAMETERS_2;

 smac_v = MLX5_ADDR_OF(fte_match_param, rule_spec->match_value,
         outer_headers.smac_47_16);
 ether_addr_copy(smac_v, addr);
 smac_c = MLX5_ADDR_OF(fte_match_param, rule_spec->match_criteria,
         outer_headers.smac_47_16);
 eth_broadcast_addr(smac_c);

 MLX5_SET(fte_match_param, rule_spec->match_criteria,
   misc_parameters_2.metadata_reg_c_0, mlx5_eswitch_get_vport_metadata_mask());
 MLX5_SET(fte_match_param, rule_spec->match_value, misc_parameters_2.metadata_reg_c_0,
   mlx5_eswitch_get_vport_metadata_for_match(esw, vport_num));

 if (vlan && vlan->pkt_reformat_push) {
  flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_PACKET_REFORMAT |
   MLX5_FLOW_CONTEXT_ACTION_MOD_HDR;
  flow_act.pkt_reformat = vlan->pkt_reformat_push;
  flow_act.modify_hdr = vlan->pkt_mod_hdr_push_mark;
 } else if (vlan) {
  if (bridge->vlan_proto == ETH_P_8021Q) {
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
      outer_headers.cvlan_tag);
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_value,
      outer_headers.cvlan_tag);
  } else if (bridge->vlan_proto == ETH_P_8021AD) {
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
      outer_headers.svlan_tag);
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_value,
      outer_headers.svlan_tag);
  }
  MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
     outer_headers.first_vid);
  MLX5_SET(fte_match_param, rule_spec->match_value, outer_headers.first_vid,
    vlan->vid);
 }

 dests[0].type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
 dests[0].ft = bridge->egress_ft;
 dests[1].type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
 dests[1].counter = counter;

 handle = mlx5_add_flow_rules(br_offloads->ingress_ft, rule_spec, &flow_act, dests,
         ARRAY_SIZE(dests));

 kvfree(rule_spec);
 return handle;
}

static struct mlx5_flow_handle *
mlx5_esw_bridge_ingress_flow_create(u16 vport_num, const unsigned char *addr,
        struct mlx5_esw_bridge_vlan *vlan,
        struct mlx5_fc *counter,
        struct mlx5_esw_bridge *bridge)
{
 return mlx5_esw_bridge_ingress_flow_with_esw_create(vport_num, addr, vlan, counter,
           bridge, bridge->br_offloads->esw);
}

static struct mlx5_flow_handle *
mlx5_esw_bridge_ingress_flow_peer_create(u16 vport_num, u16 esw_owner_vhca_id,
      const unsigned char *addr,
      struct mlx5_esw_bridge_vlan *vlan,
      struct mlx5_fc *counter,
      struct mlx5_esw_bridge *bridge)
{
 struct mlx5_devcom_comp_dev *devcom = bridge->br_offloads->esw->devcom, *pos;
 struct mlx5_eswitch *tmp, *peer_esw = NULL;
 static struct mlx5_flow_handle *handle;

 if (!mlx5_devcom_for_each_peer_begin(devcom))
  return ERR_PTR(-ENODEV);

 mlx5_devcom_for_each_peer_entry(devcom, tmp, pos) {
  if (mlx5_esw_is_owner(tmp, vport_num, esw_owner_vhca_id)) {
   peer_esw = tmp;
   break;
  }
 }

 if (!peer_esw) {
  handle = ERR_PTR(-ENODEV);
  goto out;
 }

 handle = mlx5_esw_bridge_ingress_flow_with_esw_create(vport_num, addr, vlan, counter,
             bridge, peer_esw);

out:
 mlx5_devcom_for_each_peer_end(devcom);
 return handle;
}

static struct mlx5_flow_handle *
mlx5_esw_bridge_ingress_filter_flow_create(u16 vport_num, const unsigned char *addr,
        struct mlx5_esw_bridge *bridge)
{
 struct mlx5_esw_bridge_offloads *br_offloads = bridge->br_offloads;
 struct mlx5_flow_destination dest = {
  .type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE,
  .ft = br_offloads->skip_ft,
 };
 struct mlx5_flow_act flow_act = {
  .action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
  .flags = FLOW_ACT_NO_APPEND,
 };
 struct mlx5_flow_spec *rule_spec;
 struct mlx5_flow_handle *handle;
 u8 *smac_v, *smac_c;

 rule_spec = kvzalloc(sizeof(*rule_spec), GFP_KERNEL);
 if (!rule_spec)
  return ERR_PTR(-ENOMEM);

 rule_spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS | MLX5_MATCH_MISC_PARAMETERS_2;

 smac_v = MLX5_ADDR_OF(fte_match_param, rule_spec->match_value,
         outer_headers.smac_47_16);
 ether_addr_copy(smac_v, addr);
 smac_c = MLX5_ADDR_OF(fte_match_param, rule_spec->match_criteria,
         outer_headers.smac_47_16);
 eth_broadcast_addr(smac_c);

 MLX5_SET(fte_match_param, rule_spec->match_criteria,
   misc_parameters_2.metadata_reg_c_0, mlx5_eswitch_get_vport_metadata_mask());
 MLX5_SET(fte_match_param, rule_spec->match_value, misc_parameters_2.metadata_reg_c_0,
   mlx5_eswitch_get_vport_metadata_for_match(br_offloads->esw, vport_num));

 if (bridge->vlan_proto == ETH_P_8021Q) {
  MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
     outer_headers.cvlan_tag);
  MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_value,
     outer_headers.cvlan_tag);
 } else if (bridge->vlan_proto == ETH_P_8021AD) {
  MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
     outer_headers.svlan_tag);
  MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_value,
     outer_headers.svlan_tag);
 }

 handle = mlx5_add_flow_rules(br_offloads->ingress_ft, rule_spec, &flow_act, &dest, 1);

 kvfree(rule_spec);
 return handle;
}

static struct mlx5_flow_handle *
mlx5_esw_bridge_egress_flow_create(u16 vport_num, u16 esw_owner_vhca_id, const unsigned char *addr,
       struct mlx5_esw_bridge_vlan *vlan,
       struct mlx5_esw_bridge *bridge)
{
 struct mlx5_flow_destination dest = {
  .type = MLX5_FLOW_DESTINATION_TYPE_VPORT,
  .vport.num = vport_num,
 };
 struct mlx5_flow_act flow_act = {
  .action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
  .flags = FLOW_ACT_NO_APPEND,
 };
 struct mlx5_flow_spec *rule_spec;
 struct mlx5_flow_handle *handle;
 u8 *dmac_v, *dmac_c;

 rule_spec = kvzalloc(sizeof(*rule_spec), GFP_KERNEL);
 if (!rule_spec)
  return ERR_PTR(-ENOMEM);

 if (MLX5_CAP_ESW_FLOWTABLE(bridge->br_offloads->esw->dev, flow_source) &&
     vport_num == MLX5_VPORT_UPLINK)
  rule_spec->flow_context.flow_source =
   MLX5_FLOW_CONTEXT_FLOW_SOURCE_LOCAL_VPORT;
 rule_spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;

 dmac_v = MLX5_ADDR_OF(fte_match_param, rule_spec->match_value,
         outer_headers.dmac_47_16);
 ether_addr_copy(dmac_v, addr);
 dmac_c = MLX5_ADDR_OF(fte_match_param, rule_spec->match_criteria,
         outer_headers.dmac_47_16);
 eth_broadcast_addr(dmac_c);

 if (vlan) {
  if (vlan->pkt_reformat_pop) {
   flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_PACKET_REFORMAT;
   flow_act.pkt_reformat = vlan->pkt_reformat_pop;
  }

  if (bridge->vlan_proto == ETH_P_8021Q) {
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
      outer_headers.cvlan_tag);
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_value,
      outer_headers.cvlan_tag);
  } else if (bridge->vlan_proto == ETH_P_8021AD) {
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
      outer_headers.svlan_tag);
   MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_value,
      outer_headers.svlan_tag);
  }
  MLX5_SET_TO_ONES(fte_match_param, rule_spec->match_criteria,
     outer_headers.first_vid);
  MLX5_SET(fte_match_param, rule_spec->match_value, outer_headers.first_vid,
    vlan->vid);
 }

 if (MLX5_CAP_ESW(bridge->br_offloads->esw->dev, merged_eswitch)) {
  dest.vport.flags = MLX5_FLOW_DEST_VPORT_VHCA_ID;
  dest.vport.vhca_id = esw_owner_vhca_id;
 }
 handle = mlx5_add_flow_rules(bridge->egress_ft, rule_spec, &flow_act, &dest, 1);

 kvfree(rule_spec);
 return handle;
}

static struct mlx5_flow_handle *
mlx5_esw_bridge_egress_miss_flow_create(struct mlx5_flow_table *egress_ft,
     struct mlx5_flow_table *skip_ft,
     struct mlx5_pkt_reformat *pkt_reformat)
{
 struct mlx5_flow_destination dest = {
  .type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE,
  .ft = skip_ft,
 };
 struct mlx5_flow_act flow_act = {
  .action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST |
  MLX5_FLOW_CONTEXT_ACTION_PACKET_REFORMAT,
  .flags = FLOW_ACT_NO_APPEND,
  .pkt_reformat = pkt_reformat,
 };
 struct mlx5_flow_spec *rule_spec;
 struct mlx5_flow_handle *handle;

 rule_spec = kvzalloc(sizeof(*rule_spec), GFP_KERNEL);
 if (!rule_spec)
  return ERR_PTR(-ENOMEM);

 rule_spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS_2;

 MLX5_SET(fte_match_param, rule_spec->match_criteria,
   misc_parameters_2.metadata_reg_c_1, ESW_TUN_MASK);
 MLX5_SET(fte_match_param, rule_spec->match_value, misc_parameters_2.metadata_reg_c_1,
   ESW_TUN_BRIDGE_INGRESS_PUSH_VLAN_MARK);

 handle = mlx5_add_flow_rules(egress_ft, rule_spec, &flow_act, &dest, 1);

 kvfree(rule_spec);
 return handle;
}

static struct mlx5_esw_bridge *mlx5_esw_bridge_create(struct net_device *br_netdev,
            struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge *bridge;
 int err;

 bridge = kvzalloc(sizeof(*bridge), GFP_KERNEL);
 if (!bridge)
  return ERR_PTR(-ENOMEM);

 bridge->br_offloads = br_offloads;
 err = mlx5_esw_bridge_egress_table_init(br_offloads, bridge);
 if (err)
  goto err_egress_tbl;

 err = rhashtable_init(&bridge->fdb_ht, &fdb_ht_params);
 if (err)
  goto err_fdb_ht;

 err = mlx5_esw_bridge_mdb_init(bridge);
 if (err)
  goto err_mdb_ht;

 INIT_LIST_HEAD(&bridge->fdb_list);
 bridge->ifindex = br_netdev->ifindex;
 bridge->refcnt = 1;
 bridge->ageing_time = clock_t_to_jiffies(BR_DEFAULT_AGEING_TIME);
 bridge->vlan_proto = ETH_P_8021Q;
 list_add(&bridge->list, &br_offloads->bridges);
 mlx5_esw_bridge_debugfs_init(br_netdev, bridge);

 return bridge;

err_mdb_ht:
 rhashtable_destroy(&bridge->fdb_ht);
err_fdb_ht:
 mlx5_esw_bridge_egress_table_cleanup(bridge);
err_egress_tbl:
 kvfree(bridge);
 return ERR_PTR(err);
}

static void mlx5_esw_bridge_get(struct mlx5_esw_bridge *bridge)
{
 bridge->refcnt++;
}

static void mlx5_esw_bridge_put(struct mlx5_esw_bridge_offloads *br_offloads,
    struct mlx5_esw_bridge *bridge)
{
 if (--bridge->refcnt)
  return;

 mlx5_esw_bridge_debugfs_cleanup(bridge);
 mlx5_esw_bridge_egress_table_cleanup(bridge);
 mlx5_esw_bridge_mcast_disable(bridge);
 list_del(&bridge->list);
 mlx5_esw_bridge_mdb_cleanup(bridge);
 rhashtable_destroy(&bridge->fdb_ht);
 kvfree(bridge);

 if (list_empty(&br_offloads->bridges))
  mlx5_esw_bridge_ingress_table_cleanup(br_offloads);
}

static struct mlx5_esw_bridge *
mlx5_esw_bridge_lookup(struct net_device *br_netdev, struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge *bridge;

 ASSERT_RTNL();

 list_for_each_entry(bridge, &br_offloads->bridges, list) {
  if (bridge->ifindex == br_netdev->ifindex) {
   mlx5_esw_bridge_get(bridge);
   return bridge;
  }
 }

 if (!br_offloads->ingress_ft) {
  int err = mlx5_esw_bridge_ingress_table_init(br_offloads);

  if (err)
   return ERR_PTR(err);
 }

 bridge = mlx5_esw_bridge_create(br_netdev, br_offloads);
 if (IS_ERR(bridge) && list_empty(&br_offloads->bridges))
  mlx5_esw_bridge_ingress_table_cleanup(br_offloads);
 return bridge;
}

static unsigned long mlx5_esw_bridge_port_key_from_data(u16 vport_num, u16 esw_owner_vhca_id)
{
 return vport_num | (unsigned long)esw_owner_vhca_id << sizeof(vport_num) * BITS_PER_BYTE;
}

unsigned long mlx5_esw_bridge_port_key(struct mlx5_esw_bridge_port *port)
{
 return mlx5_esw_bridge_port_key_from_data(port->vport_num, port->esw_owner_vhca_id);
}

static int mlx5_esw_bridge_port_insert(struct mlx5_esw_bridge_port *port,
           struct mlx5_esw_bridge_offloads *br_offloads)
{
 return xa_insert(&br_offloads->ports, mlx5_esw_bridge_port_key(port), port, GFP_KERNEL);
}

static struct mlx5_esw_bridge_port *
mlx5_esw_bridge_port_lookup(u16 vport_num, u16 esw_owner_vhca_id,
       struct mlx5_esw_bridge_offloads *br_offloads)
{
 return xa_load(&br_offloads->ports, mlx5_esw_bridge_port_key_from_data(vport_num,
                esw_owner_vhca_id));
}

static void mlx5_esw_bridge_port_erase(struct mlx5_esw_bridge_port *port,
           struct mlx5_esw_bridge_offloads *br_offloads)
{
 xa_erase(&br_offloads->ports, mlx5_esw_bridge_port_key(port));
}

static struct mlx5_esw_bridge *
mlx5_esw_bridge_from_port_lookup(u16 vport_num, u16 esw_owner_vhca_id,
     struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge_port *port;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!port)
  return NULL;

 return port->bridge;
}

static void mlx5_esw_bridge_fdb_entry_refresh(struct mlx5_esw_bridge_fdb_entry *entry)
{
 trace_mlx5_esw_bridge_fdb_entry_refresh(entry);

 mlx5_esw_bridge_fdb_offload_notify(entry->dev, entry->key.addr,
        entry->key.vid,
        SWITCHDEV_FDB_ADD_TO_BRIDGE);
}

static void
mlx5_esw_bridge_fdb_entry_cleanup(struct mlx5_esw_bridge_fdb_entry *entry,
      struct mlx5_esw_bridge *bridge)
{
 trace_mlx5_esw_bridge_fdb_entry_cleanup(entry);

 rhashtable_remove_fast(&bridge->fdb_ht, &entry->ht_node, fdb_ht_params);
 mlx5_del_flow_rules(entry->egress_handle);
 if (entry->filter_handle)
  mlx5_del_flow_rules(entry->filter_handle);
 mlx5_del_flow_rules(entry->ingress_handle);
 mlx5_fc_destroy(bridge->br_offloads->esw->dev, entry->ingress_counter);
 list_del(&entry->vlan_list);
 list_del(&entry->list);
 kvfree(entry);
}

static void
mlx5_esw_bridge_fdb_entry_notify_and_cleanup(struct mlx5_esw_bridge_fdb_entry *entry,
          struct mlx5_esw_bridge *bridge)
{
 mlx5_esw_bridge_fdb_del_notify(entry);
 mlx5_esw_bridge_fdb_entry_cleanup(entry, bridge);
}

static void mlx5_esw_bridge_fdb_flush(struct mlx5_esw_bridge *bridge)
{
 struct mlx5_esw_bridge_fdb_entry *entry, *tmp;

 list_for_each_entry_safe(entry, tmp, &bridge->fdb_list, list)
  mlx5_esw_bridge_fdb_entry_notify_and_cleanup(entry, bridge);
}

static struct mlx5_esw_bridge_vlan *
mlx5_esw_bridge_vlan_lookup(u16 vid, struct mlx5_esw_bridge_port *port)
{
 return xa_load(&port->vlans, vid);
}

static int
mlx5_esw_bridge_vlan_push_create(u16 vlan_proto, struct mlx5_esw_bridge_vlan *vlan,
     struct mlx5_eswitch *esw)
{
 struct {
  __be16 h_vlan_proto;
  __be16 h_vlan_TCI;
 } vlan_hdr = { htons(vlan_proto), htons(vlan->vid) };
 struct mlx5_pkt_reformat_params reformat_params = {};
 struct mlx5_pkt_reformat *pkt_reformat;

 if (!BIT(MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, reformat_insert)) ||
     MLX5_CAP_GEN_2(esw->dev, max_reformat_insert_size) < sizeof(vlan_hdr) ||
     MLX5_CAP_GEN_2(esw->dev, max_reformat_insert_offset) <
     offsetof(struct vlan_ethhdr, h_vlan_proto)) {
  esw_warn(esw->dev, "Packet reformat INSERT_HEADER is not supported\n");
  return -EOPNOTSUPP;
 }

 reformat_params.type = MLX5_REFORMAT_TYPE_INSERT_HDR;
 reformat_params.param_0 = MLX5_REFORMAT_CONTEXT_ANCHOR_MAC_START;
 reformat_params.param_1 = offsetof(struct vlan_ethhdr, h_vlan_proto);
 reformat_params.size = sizeof(vlan_hdr);
 reformat_params.data = &vlan_hdr;
 pkt_reformat = mlx5_packet_reformat_alloc(esw->dev,
        &reformat_params,
        MLX5_FLOW_NAMESPACE_FDB);
 if (IS_ERR(pkt_reformat)) {
  esw_warn(esw->dev, "Failed to alloc packet reformat INSERT_HEADER (err=%ld)\n",
    PTR_ERR(pkt_reformat));
  return PTR_ERR(pkt_reformat);
 }

 vlan->pkt_reformat_push = pkt_reformat;
 return 0;
}

static void
mlx5_esw_bridge_vlan_push_cleanup(struct mlx5_esw_bridge_vlan *vlan, struct mlx5_eswitch *esw)
{
 mlx5_packet_reformat_dealloc(esw->dev, vlan->pkt_reformat_push);
 vlan->pkt_reformat_push = NULL;
}

static int
mlx5_esw_bridge_vlan_pop_create(struct mlx5_esw_bridge_vlan *vlan, struct mlx5_eswitch *esw)
{
 struct mlx5_pkt_reformat *pkt_reformat;

 if (!mlx5_esw_bridge_pkt_reformat_vlan_pop_supported(esw)) {
  esw_warn(esw->dev, "Packet reformat REMOVE_HEADER is not supported\n");
  return -EOPNOTSUPP;
 }

 pkt_reformat = mlx5_esw_bridge_pkt_reformat_vlan_pop_create(esw);
 if (IS_ERR(pkt_reformat)) {
  esw_warn(esw->dev, "Failed to alloc packet reformat REMOVE_HEADER (err=%ld)\n",
    PTR_ERR(pkt_reformat));
  return PTR_ERR(pkt_reformat);
 }

 vlan->pkt_reformat_pop = pkt_reformat;
 return 0;
}

static void
mlx5_esw_bridge_vlan_pop_cleanup(struct mlx5_esw_bridge_vlan *vlan, struct mlx5_eswitch *esw)
{
 mlx5_packet_reformat_dealloc(esw->dev, vlan->pkt_reformat_pop);
 vlan->pkt_reformat_pop = NULL;
}

static int
mlx5_esw_bridge_vlan_push_mark_create(struct mlx5_esw_bridge_vlan *vlan, struct mlx5_eswitch *esw)
{
 u8 action[MLX5_UN_SZ_BYTES(set_add_copy_action_in_auto)] = {};
 struct mlx5_modify_hdr *pkt_mod_hdr;

 MLX5_SET(set_action_in, action, action_type, MLX5_ACTION_TYPE_SET);
 MLX5_SET(set_action_in, action, field, MLX5_ACTION_IN_FIELD_METADATA_REG_C_1);
 MLX5_SET(set_action_in, action, offset, 8);
 MLX5_SET(set_action_in, action, length, ESW_TUN_OPTS_BITS + ESW_TUN_ID_BITS);
 MLX5_SET(set_action_in, action, data, ESW_TUN_BRIDGE_INGRESS_PUSH_VLAN);

 pkt_mod_hdr = mlx5_modify_header_alloc(esw->dev, MLX5_FLOW_NAMESPACE_FDB, 1, action);
 if (IS_ERR(pkt_mod_hdr))
  return PTR_ERR(pkt_mod_hdr);

 vlan->pkt_mod_hdr_push_mark = pkt_mod_hdr;
 return 0;
}

static void
mlx5_esw_bridge_vlan_push_mark_cleanup(struct mlx5_esw_bridge_vlan *vlan, struct mlx5_eswitch *esw)
{
 mlx5_modify_header_dealloc(esw->dev, vlan->pkt_mod_hdr_push_mark);
 vlan->pkt_mod_hdr_push_mark = NULL;
}

static int
mlx5_esw_bridge_vlan_push_pop_fhs_create(u16 vlan_proto, struct mlx5_esw_bridge_port *port,
      struct mlx5_esw_bridge_vlan *vlan)
{
 return mlx5_esw_bridge_vlan_mcast_init(vlan_proto, port, vlan);
}

static void
mlx5_esw_bridge_vlan_push_pop_fhs_cleanup(struct mlx5_esw_bridge_vlan *vlan)
{
 mlx5_esw_bridge_vlan_mcast_cleanup(vlan);
}

static int
mlx5_esw_bridge_vlan_push_pop_create(u16 vlan_proto, u16 flags, struct mlx5_esw_bridge_port *port,
         struct mlx5_esw_bridge_vlan *vlan, struct mlx5_eswitch *esw)
{
 int err;

 if (flags & BRIDGE_VLAN_INFO_PVID) {
  err = mlx5_esw_bridge_vlan_push_create(vlan_proto, vlan, esw);
  if (err)
   return err;

  err = mlx5_esw_bridge_vlan_push_mark_create(vlan, esw);
  if (err)
   goto err_vlan_push_mark;
 }

 if (flags & BRIDGE_VLAN_INFO_UNTAGGED) {
  err = mlx5_esw_bridge_vlan_pop_create(vlan, esw);
  if (err)
   goto err_vlan_pop;

  err = mlx5_esw_bridge_vlan_push_pop_fhs_create(vlan_proto, port, vlan);
  if (err)
   goto err_vlan_pop_fhs;
 }

 return 0;

err_vlan_pop_fhs:
 mlx5_esw_bridge_vlan_pop_cleanup(vlan, esw);
err_vlan_pop:
 if (vlan->pkt_mod_hdr_push_mark)
  mlx5_esw_bridge_vlan_push_mark_cleanup(vlan, esw);
err_vlan_push_mark:
 if (vlan->pkt_reformat_push)
  mlx5_esw_bridge_vlan_push_cleanup(vlan, esw);
 return err;
}

static struct mlx5_esw_bridge_vlan *
mlx5_esw_bridge_vlan_create(u16 vlan_proto, u16 vid, u16 flags, struct mlx5_esw_bridge_port *port,
       struct mlx5_eswitch *esw)
{
 struct mlx5_esw_bridge_vlan *vlan;
 int err;

 vlan = kvzalloc(sizeof(*vlan), GFP_KERNEL);
 if (!vlan)
  return ERR_PTR(-ENOMEM);

 vlan->vid = vid;
 vlan->flags = flags;
 INIT_LIST_HEAD(&vlan->fdb_list);

 err = mlx5_esw_bridge_vlan_push_pop_create(vlan_proto, flags, port, vlan, esw);
 if (err)
  goto err_vlan_push_pop;

 err = xa_insert(&port->vlans, vid, vlan, GFP_KERNEL);
 if (err)
  goto err_xa_insert;

 trace_mlx5_esw_bridge_vlan_create(vlan);
 return vlan;

err_xa_insert:
 if (vlan->mcast_handle)
  mlx5_esw_bridge_vlan_push_pop_fhs_cleanup(vlan);
 if (vlan->pkt_reformat_pop)
  mlx5_esw_bridge_vlan_pop_cleanup(vlan, esw);
 if (vlan->pkt_mod_hdr_push_mark)
  mlx5_esw_bridge_vlan_push_mark_cleanup(vlan, esw);
 if (vlan->pkt_reformat_push)
  mlx5_esw_bridge_vlan_push_cleanup(vlan, esw);
err_vlan_push_pop:
 kvfree(vlan);
 return ERR_PTR(err);
}

static void mlx5_esw_bridge_vlan_erase(struct mlx5_esw_bridge_port *port,
           struct mlx5_esw_bridge_vlan *vlan)
{
 xa_erase(&port->vlans, vlan->vid);
}

static void mlx5_esw_bridge_vlan_flush(struct mlx5_esw_bridge_port *port,
           struct mlx5_esw_bridge_vlan *vlan,
           struct mlx5_esw_bridge *bridge)
{
 struct mlx5_eswitch *esw = bridge->br_offloads->esw;
 struct mlx5_esw_bridge_fdb_entry *entry, *tmp;

 list_for_each_entry_safe(entry, tmp, &vlan->fdb_list, vlan_list)
  mlx5_esw_bridge_fdb_entry_notify_and_cleanup(entry, bridge);
 mlx5_esw_bridge_port_mdb_vlan_flush(port, vlan);

 if (vlan->mcast_handle)
  mlx5_esw_bridge_vlan_push_pop_fhs_cleanup(vlan);
 if (vlan->pkt_reformat_pop)
  mlx5_esw_bridge_vlan_pop_cleanup(vlan, esw);
 if (vlan->pkt_mod_hdr_push_mark)
  mlx5_esw_bridge_vlan_push_mark_cleanup(vlan, esw);
 if (vlan->pkt_reformat_push)
  mlx5_esw_bridge_vlan_push_cleanup(vlan, esw);
}

static void mlx5_esw_bridge_vlan_cleanup(struct mlx5_esw_bridge_port *port,
      struct mlx5_esw_bridge_vlan *vlan,
      struct mlx5_esw_bridge *bridge)
{
 trace_mlx5_esw_bridge_vlan_cleanup(vlan);
 mlx5_esw_bridge_vlan_flush(port, vlan, bridge);
 mlx5_esw_bridge_vlan_erase(port, vlan);
 kvfree(vlan);
}

static void mlx5_esw_bridge_port_vlans_flush(struct mlx5_esw_bridge_port *port,
          struct mlx5_esw_bridge *bridge)
{
 struct mlx5_esw_bridge_vlan *vlan;
 unsigned long index;

 xa_for_each(&port->vlans, index, vlan)
  mlx5_esw_bridge_vlan_cleanup(port, vlan, bridge);
}

static int mlx5_esw_bridge_port_vlans_recreate(struct mlx5_esw_bridge_port *port,
            struct mlx5_esw_bridge *bridge)
{
 struct mlx5_esw_bridge_offloads *br_offloads = bridge->br_offloads;
 struct mlx5_esw_bridge_vlan *vlan;
 unsigned long i;
 int err;

 xa_for_each(&port->vlans, i, vlan) {
  mlx5_esw_bridge_vlan_flush(port, vlan, bridge);
  err = mlx5_esw_bridge_vlan_push_pop_create(bridge->vlan_proto, vlan->flags, port,
          vlan, br_offloads->esw);
  if (err) {
   esw_warn(br_offloads->esw->dev,
     "Failed to create VLAN=%u(proto=%x) push/pop actions (vport=%u,err=%d)\n",
     vlan->vid, bridge->vlan_proto, port->vport_num,
     err);
   return err;
  }
 }

 return 0;
}

static int
mlx5_esw_bridge_vlans_recreate(struct mlx5_esw_bridge *bridge)
{
 struct mlx5_esw_bridge_offloads *br_offloads = bridge->br_offloads;
 struct mlx5_esw_bridge_port *port;
 unsigned long i;
 int err;

 xa_for_each(&br_offloads->ports, i, port) {
  if (port->bridge != bridge)
   continue;

  err = mlx5_esw_bridge_port_vlans_recreate(port, bridge);
  if (err)
   return err;
 }

 return 0;
}

static struct mlx5_esw_bridge_vlan *
mlx5_esw_bridge_port_vlan_lookup(u16 vid, u16 vport_num, u16 esw_owner_vhca_id,
     struct mlx5_esw_bridge *bridge, struct mlx5_eswitch *esw)
{
 struct mlx5_esw_bridge_port *port;
 struct mlx5_esw_bridge_vlan *vlan;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, bridge->br_offloads);
 if (!port) {
  /* FDB is added asynchronously on wq while port might have been deleted
 * concurrently. Report on 'info' logging level and skip the FDB offload.
 */
  esw_info(esw->dev, "Failed to lookup bridge port (vport=%u)\n", vport_num);
  return ERR_PTR(-EINVAL);
 }

 vlan = mlx5_esw_bridge_vlan_lookup(vid, port);
 if (!vlan) {
  /* FDB is added asynchronously on wq while vlan might have been deleted
 * concurrently. Report on 'info' logging level and skip the FDB offload.
 */
  esw_info(esw->dev, "Failed to lookup bridge port vlan metadata (vport=%u)\n",
    vport_num);
  return ERR_PTR(-EINVAL);
 }

 return vlan;
}

static struct mlx5_esw_bridge_fdb_entry *
mlx5_esw_bridge_fdb_lookup(struct mlx5_esw_bridge *bridge,
      const unsigned char *addr, u16 vid)
{
 struct mlx5_esw_bridge_fdb_key key = {};

 ether_addr_copy(key.addr, addr);
 key.vid = vid;
 return rhashtable_lookup_fast(&bridge->fdb_ht, &key, fdb_ht_params);
}

static struct mlx5_esw_bridge_fdb_entry *
mlx5_esw_bridge_fdb_entry_init(struct net_device *dev, u16 vport_num, u16 esw_owner_vhca_id,
          const unsigned char *addr, u16 vid, bool added_by_user, bool peer,
          struct mlx5_eswitch *esw, struct mlx5_esw_bridge *bridge)
{
 struct mlx5_esw_bridge_vlan *vlan = NULL;
 struct mlx5_esw_bridge_fdb_entry *entry;
 struct mlx5_flow_handle *handle;
 struct mlx5_fc *counter;
 int err;

 if (bridge->flags & MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG && vid) {
  vlan = mlx5_esw_bridge_port_vlan_lookup(vid, vport_num, esw_owner_vhca_id, bridge,
       esw);
  if (IS_ERR(vlan))
   return ERR_CAST(vlan);
 }

 entry = mlx5_esw_bridge_fdb_lookup(bridge, addr, vid);
 if (entry)
  mlx5_esw_bridge_fdb_entry_notify_and_cleanup(entry, bridge);

 entry = kvzalloc(sizeof(*entry), GFP_KERNEL);
 if (!entry)
  return ERR_PTR(-ENOMEM);

 ether_addr_copy(entry->key.addr, addr);
 entry->key.vid = vid;
 entry->dev = dev;
 entry->vport_num = vport_num;
 entry->esw_owner_vhca_id = esw_owner_vhca_id;
 entry->lastuse = jiffies;
 if (added_by_user)
  entry->flags |= MLX5_ESW_BRIDGE_FLAG_ADDED_BY_USER;
 if (peer)
  entry->flags |= MLX5_ESW_BRIDGE_FLAG_PEER;

 counter = mlx5_fc_create(esw->dev, true);
 if (IS_ERR(counter)) {
  err = PTR_ERR(counter);
  goto err_ingress_fc_create;
 }
 entry->ingress_counter = counter;

 handle = peer ?
  mlx5_esw_bridge_ingress_flow_peer_create(vport_num, esw_owner_vhca_id,
        addr, vlan, counter, bridge) :
  mlx5_esw_bridge_ingress_flow_create(vport_num, addr, vlan,
          counter, bridge);
 if (IS_ERR(handle)) {
  err = PTR_ERR(handle);
  esw_warn(esw->dev, "Failed to create ingress flow(vport=%u,err=%d,peer=%d)\n",
    vport_num, err, peer);
  goto err_ingress_flow_create;
 }
 entry->ingress_handle = handle;

 if (bridge->flags & MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG) {
  handle = mlx5_esw_bridge_ingress_filter_flow_create(vport_num, addr, bridge);
  if (IS_ERR(handle)) {
   err = PTR_ERR(handle);
   esw_warn(esw->dev, "Failed to create ingress filter(vport=%u,err=%d)\n",
     vport_num, err);
   goto err_ingress_filter_flow_create;
  }
  entry->filter_handle = handle;
 }

 handle = mlx5_esw_bridge_egress_flow_create(vport_num, esw_owner_vhca_id, addr, vlan,
          bridge);
 if (IS_ERR(handle)) {
  err = PTR_ERR(handle);
  esw_warn(esw->dev, "Failed to create egress flow(vport=%u,err=%d)\n",
    vport_num, err);
  goto err_egress_flow_create;
 }
 entry->egress_handle = handle;

 err = rhashtable_insert_fast(&bridge->fdb_ht, &entry->ht_node, fdb_ht_params);
 if (err) {
  esw_warn(esw->dev, "Failed to insert FDB flow(vport=%u,err=%d)\n", vport_num, err);
  goto err_ht_init;
 }

 if (vlan)
  list_add(&entry->vlan_list, &vlan->fdb_list);
 else
  INIT_LIST_HEAD(&entry->vlan_list);
 list_add(&entry->list, &bridge->fdb_list);

 trace_mlx5_esw_bridge_fdb_entry_init(entry);
 return entry;

err_ht_init:
 mlx5_del_flow_rules(entry->egress_handle);
err_egress_flow_create:
 if (entry->filter_handle)
  mlx5_del_flow_rules(entry->filter_handle);
err_ingress_filter_flow_create:
 mlx5_del_flow_rules(entry->ingress_handle);
err_ingress_flow_create:
 mlx5_fc_destroy(esw->dev, entry->ingress_counter);
err_ingress_fc_create:
 kvfree(entry);
 return ERR_PTR(err);
}

int mlx5_esw_bridge_ageing_time_set(u16 vport_num, u16 esw_owner_vhca_id, unsigned long ageing_time,
        struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge *bridge;

 bridge = mlx5_esw_bridge_from_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!bridge)
  return -EINVAL;

 bridge->ageing_time = clock_t_to_jiffies(ageing_time);
 return 0;
}

int mlx5_esw_bridge_vlan_filtering_set(u16 vport_num, u16 esw_owner_vhca_id, bool enable,
           struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge *bridge;
 bool filtering;

 bridge = mlx5_esw_bridge_from_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!bridge)
  return -EINVAL;

 filtering = bridge->flags & MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG;
 if (filtering == enable)
  return 0;

 mlx5_esw_bridge_fdb_flush(bridge);
 mlx5_esw_bridge_mdb_flush(bridge);
 if (enable)
  bridge->flags |= MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG;
 else
  bridge->flags &= ~MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG;

 return 0;
}

int mlx5_esw_bridge_vlan_proto_set(u16 vport_num, u16 esw_owner_vhca_id, u16 proto,
       struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge *bridge;

 bridge = mlx5_esw_bridge_from_port_lookup(vport_num, esw_owner_vhca_id,
        br_offloads);
 if (!bridge)
  return -EINVAL;

 if (bridge->vlan_proto == proto)
  return 0;
 if (proto != ETH_P_8021Q && proto != ETH_P_8021AD) {
  esw_warn(br_offloads->esw->dev, "Can't set unsupported VLAN protocol %x", proto);
  return -EOPNOTSUPP;
 }

 mlx5_esw_bridge_fdb_flush(bridge);
 mlx5_esw_bridge_mdb_flush(bridge);
 bridge->vlan_proto = proto;
 mlx5_esw_bridge_vlans_recreate(bridge);

 return 0;
}

int mlx5_esw_bridge_mcast_set(u16 vport_num, u16 esw_owner_vhca_id, bool enable,
         struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_eswitch *esw = br_offloads->esw;
 struct mlx5_esw_bridge *bridge;
 int err = 0;
 bool mcast;

 if (!(MLX5_CAP_ESW_FLOWTABLE((esw)->dev, fdb_multi_path_any_table) ||
       MLX5_CAP_ESW_FLOWTABLE((esw)->dev, fdb_multi_path_any_table_limit_regc)) ||
     !MLX5_CAP_ESW_FLOWTABLE((esw)->dev, fdb_uplink_hairpin) ||
     !MLX5_CAP_ESW_FLOWTABLE_FDB((esw)->dev, ignore_flow_level))
  return -EOPNOTSUPP;

 bridge = mlx5_esw_bridge_from_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!bridge)
  return -EINVAL;

 mcast = bridge->flags & MLX5_ESW_BRIDGE_MCAST_FLAG;
 if (mcast == enable)
  return 0;

 if (enable)
  err = mlx5_esw_bridge_mcast_enable(bridge);
 else
  mlx5_esw_bridge_mcast_disable(bridge);

 return err;
}

static int mlx5_esw_bridge_vport_init(u16 vport_num, u16 esw_owner_vhca_id, u16 flags,
          struct mlx5_esw_bridge_offloads *br_offloads,
          struct mlx5_esw_bridge *bridge)
{
 struct mlx5_eswitch *esw = br_offloads->esw;
 struct mlx5_esw_bridge_port *port;
 int err;

 port = kvzalloc(sizeof(*port), GFP_KERNEL);
 if (!port)
  return -ENOMEM;

 port->vport_num = vport_num;
 port->esw_owner_vhca_id = esw_owner_vhca_id;
 port->bridge = bridge;
 port->flags |= flags;
 xa_init(&port->vlans);

 err = mlx5_esw_bridge_port_mcast_init(port);
 if (err) {
  esw_warn(esw->dev,
    "Failed to initialize port multicast (vport=%u,esw_owner_vhca_id=%u,err=%d)\n",
    port->vport_num, port->esw_owner_vhca_id, err);
  goto err_port_mcast;
 }

 err = mlx5_esw_bridge_port_insert(port, br_offloads);
 if (err) {
  esw_warn(esw->dev,
    "Failed to insert port metadata (vport=%u,esw_owner_vhca_id=%u,err=%d)\n",
    port->vport_num, port->esw_owner_vhca_id, err);
  goto err_port_insert;
 }
 trace_mlx5_esw_bridge_vport_init(port);

 return 0;

err_port_insert:
 mlx5_esw_bridge_port_mcast_cleanup(port);
err_port_mcast:
 kvfree(port);
 return err;
}

static int mlx5_esw_bridge_vport_cleanup(struct mlx5_esw_bridge_offloads *br_offloads,
      struct mlx5_esw_bridge_port *port)
{
 u16 vport_num = port->vport_num, esw_owner_vhca_id = port->esw_owner_vhca_id;
 struct mlx5_esw_bridge *bridge = port->bridge;
 struct mlx5_esw_bridge_fdb_entry *entry, *tmp;

 list_for_each_entry_safe(entry, tmp, &bridge->fdb_list, list)
  if (entry->vport_num == vport_num && entry->esw_owner_vhca_id == esw_owner_vhca_id)
   mlx5_esw_bridge_fdb_entry_cleanup(entry, bridge);

 trace_mlx5_esw_bridge_vport_cleanup(port);
 mlx5_esw_bridge_port_vlans_flush(port, bridge);
 mlx5_esw_bridge_port_mcast_cleanup(port);
 mlx5_esw_bridge_port_erase(port, br_offloads);
 kvfree(port);
 mlx5_esw_bridge_put(br_offloads, bridge);
 return 0;
}

static int mlx5_esw_bridge_vport_link_with_flags(struct net_device *br_netdev, u16 vport_num,
       u16 esw_owner_vhca_id, u16 flags,
       struct mlx5_esw_bridge_offloads *br_offloads,
       struct netlink_ext_ack *extack)
{
 struct mlx5_esw_bridge *bridge;
 int err;

 bridge = mlx5_esw_bridge_lookup(br_netdev, br_offloads);
 if (IS_ERR(bridge)) {
  NL_SET_ERR_MSG_MOD(extack, "Error checking for existing bridge with same ifindex");
  return PTR_ERR(bridge);
 }

 err = mlx5_esw_bridge_vport_init(vport_num, esw_owner_vhca_id, flags, br_offloads, bridge);
 if (err) {
  NL_SET_ERR_MSG_MOD(extack, "Error initializing port");
  goto err_vport;
 }
 return 0;

err_vport:
 mlx5_esw_bridge_put(br_offloads, bridge);
 return err;
}

int mlx5_esw_bridge_vport_link(struct net_device *br_netdev, u16 vport_num, u16 esw_owner_vhca_id,
          struct mlx5_esw_bridge_offloads *br_offloads,
          struct netlink_ext_ack *extack)
{
 return mlx5_esw_bridge_vport_link_with_flags(br_netdev, vport_num, esw_owner_vhca_id, 0,
           br_offloads, extack);
}

int mlx5_esw_bridge_vport_unlink(struct net_device *br_netdev, u16 vport_num,
     u16 esw_owner_vhca_id,
     struct mlx5_esw_bridge_offloads *br_offloads,
     struct netlink_ext_ack *extack)
{
 struct mlx5_esw_bridge_port *port;
 int err;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!port) {
  NL_SET_ERR_MSG_MOD(extack, "Port is not attached to any bridge");
  return -EINVAL;
 }
 if (port->bridge->ifindex != br_netdev->ifindex) {
  NL_SET_ERR_MSG_MOD(extack, "Port is attached to another bridge");
  return -EINVAL;
 }

 err = mlx5_esw_bridge_vport_cleanup(br_offloads, port);
 if (err)
  NL_SET_ERR_MSG_MOD(extack, "Port cleanup failed");
 return err;
}

int mlx5_esw_bridge_vport_peer_link(struct net_device *br_netdev, u16 vport_num,
        u16 esw_owner_vhca_id,
        struct mlx5_esw_bridge_offloads *br_offloads,
        struct netlink_ext_ack *extack)
{
 if (!MLX5_CAP_ESW(br_offloads->esw->dev, merged_eswitch))
  return 0;

 return mlx5_esw_bridge_vport_link_with_flags(br_netdev, vport_num, esw_owner_vhca_id,
           MLX5_ESW_BRIDGE_PORT_FLAG_PEER,
           br_offloads, extack);
}

int mlx5_esw_bridge_vport_peer_unlink(struct net_device *br_netdev, u16 vport_num,
          u16 esw_owner_vhca_id,
          struct mlx5_esw_bridge_offloads *br_offloads,
          struct netlink_ext_ack *extack)
{
 return mlx5_esw_bridge_vport_unlink(br_netdev, vport_num, esw_owner_vhca_id, br_offloads,
         extack);
}

int mlx5_esw_bridge_port_vlan_add(u16 vport_num, u16 esw_owner_vhca_id, u16 vid, u16 flags,
      struct mlx5_esw_bridge_offloads *br_offloads,
      struct netlink_ext_ack *extack)
{
 struct mlx5_esw_bridge_port *port;
 struct mlx5_esw_bridge_vlan *vlan;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!port)
  return -EINVAL;

 vlan = mlx5_esw_bridge_vlan_lookup(vid, port);
 if (vlan) {
  if (vlan->flags == flags)
   return 0;
  mlx5_esw_bridge_vlan_cleanup(port, vlan, port->bridge);
 }

 vlan = mlx5_esw_bridge_vlan_create(port->bridge->vlan_proto, vid, flags, port,
        br_offloads->esw);
 if (IS_ERR(vlan)) {
  NL_SET_ERR_MSG_MOD(extack, "Failed to create VLAN entry");
  return PTR_ERR(vlan);
 }
 return 0;
}

void mlx5_esw_bridge_port_vlan_del(u16 vport_num, u16 esw_owner_vhca_id, u16 vid,
       struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge_port *port;
 struct mlx5_esw_bridge_vlan *vlan;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!port)
  return;

 vlan = mlx5_esw_bridge_vlan_lookup(vid, port);
 if (!vlan)
  return;
 mlx5_esw_bridge_vlan_cleanup(port, vlan, port->bridge);
}

void mlx5_esw_bridge_fdb_update_used(struct net_device *dev, u16 vport_num, u16 esw_owner_vhca_id,
         struct mlx5_esw_bridge_offloads *br_offloads,
         struct switchdev_notifier_fdb_info *fdb_info)
{
 struct mlx5_esw_bridge_fdb_entry *entry;
 struct mlx5_esw_bridge *bridge;

 bridge = mlx5_esw_bridge_from_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!bridge)
  return;

 entry = mlx5_esw_bridge_fdb_lookup(bridge, fdb_info->addr, fdb_info->vid);
 if (!entry) {
  esw_debug(br_offloads->esw->dev,
     "FDB update entry with specified key not found (MAC=%pM,vid=%u,vport=%u)\n",
     fdb_info->addr, fdb_info->vid, vport_num);
  return;
 }

 entry->lastuse = jiffies;
}

void mlx5_esw_bridge_fdb_mark_deleted(struct net_device *dev, u16 vport_num, u16 esw_owner_vhca_id,
          struct mlx5_esw_bridge_offloads *br_offloads,
          struct switchdev_notifier_fdb_info *fdb_info)
{
 struct mlx5_esw_bridge_fdb_entry *entry;
 struct mlx5_esw_bridge *bridge;

 bridge = mlx5_esw_bridge_from_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!bridge)
  return;

 entry = mlx5_esw_bridge_fdb_lookup(bridge, fdb_info->addr, fdb_info->vid);
 if (!entry) {
  esw_debug(br_offloads->esw->dev,
     "FDB mark deleted entry with specified key not found (MAC=%pM,vid=%u,vport=%u)\n",
     fdb_info->addr, fdb_info->vid, vport_num);
  return;
 }

 entry->flags |= MLX5_ESW_BRIDGE_FLAG_DELETED;
}

void mlx5_esw_bridge_fdb_create(struct net_device *dev, u16 vport_num, u16 esw_owner_vhca_id,
    struct mlx5_esw_bridge_offloads *br_offloads,
    struct switchdev_notifier_fdb_info *fdb_info)
{
 struct mlx5_esw_bridge_fdb_entry *entry;
 struct mlx5_esw_bridge_port *port;
 struct mlx5_esw_bridge *bridge;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!port)
  return;

 bridge = port->bridge;
 entry = mlx5_esw_bridge_fdb_entry_init(dev, vport_num, esw_owner_vhca_id, fdb_info->addr,
            fdb_info->vid, fdb_info->added_by_user,
            port->flags & MLX5_ESW_BRIDGE_PORT_FLAG_PEER,
            br_offloads->esw, bridge);
 if (IS_ERR(entry))
  return;

 if (entry->flags & MLX5_ESW_BRIDGE_FLAG_ADDED_BY_USER)
  mlx5_esw_bridge_fdb_offload_notify(dev, entry->key.addr, entry->key.vid,
         SWITCHDEV_FDB_OFFLOADED);
 else if (!(entry->flags & MLX5_ESW_BRIDGE_FLAG_PEER))
  /* Take over dynamic entries to prevent kernel bridge from aging them out. */
  mlx5_esw_bridge_fdb_offload_notify(dev, entry->key.addr, entry->key.vid,
         SWITCHDEV_FDB_ADD_TO_BRIDGE);
}

void mlx5_esw_bridge_fdb_remove(struct net_device *dev, u16 vport_num, u16 esw_owner_vhca_id,
    struct mlx5_esw_bridge_offloads *br_offloads,
    struct switchdev_notifier_fdb_info *fdb_info)
{
 struct mlx5_eswitch *esw = br_offloads->esw;
 struct mlx5_esw_bridge_fdb_entry *entry;
 struct mlx5_esw_bridge *bridge;

 bridge = mlx5_esw_bridge_from_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!bridge)
  return;

 entry = mlx5_esw_bridge_fdb_lookup(bridge, fdb_info->addr, fdb_info->vid);
 if (!entry) {
  esw_debug(esw->dev,
     "FDB remove entry with specified key not found (MAC=%pM,vid=%u,vport=%u)\n",
     fdb_info->addr, fdb_info->vid, vport_num);
  return;
 }

 mlx5_esw_bridge_fdb_entry_notify_and_cleanup(entry, bridge);
}

void mlx5_esw_bridge_update(struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge_fdb_entry *entry, *tmp;
 struct mlx5_esw_bridge *bridge;

 list_for_each_entry(bridge, &br_offloads->bridges, list) {
  list_for_each_entry_safe(entry, tmp, &bridge->fdb_list, list) {
   unsigned long lastuse =
    (unsigned long)mlx5_fc_query_lastuse(entry->ingress_counter);

   if (entry->flags & (MLX5_ESW_BRIDGE_FLAG_ADDED_BY_USER |
         MLX5_ESW_BRIDGE_FLAG_DELETED))
    continue;

   if (time_after(lastuse, entry->lastuse))
    mlx5_esw_bridge_fdb_entry_refresh(entry);
   else if (!(entry->flags & MLX5_ESW_BRIDGE_FLAG_PEER) &&
     time_is_before_jiffies(entry->lastuse + bridge->ageing_time))
    mlx5_esw_bridge_fdb_entry_notify_and_cleanup(entry, bridge);
  }
 }
}

int mlx5_esw_bridge_port_mdb_add(struct net_device *dev, u16 vport_num, u16 esw_owner_vhca_id,
     const unsigned char *addr, u16 vid,
     struct mlx5_esw_bridge_offloads *br_offloads,
     struct netlink_ext_ack *extack)
{
 struct mlx5_esw_bridge_vlan *vlan;
 struct mlx5_esw_bridge_port *port;
 struct mlx5_esw_bridge *bridge;
 int err;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!port) {
  esw_warn(br_offloads->esw->dev,
    "Failed to lookup bridge port to add MDB (MAC=%pM,vport=%u)\n",
    addr, vport_num);
  NL_SET_ERR_MSG_FMT_MOD(extack,
           "Failed to lookup bridge port to add MDB (MAC=%pM,vport=%u)",
           addr, vport_num);
  return -EINVAL;
 }

 bridge = port->bridge;
 if (bridge->flags & MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG && vid) {
  vlan = mlx5_esw_bridge_vlan_lookup(vid, port);
  if (!vlan) {
   esw_warn(br_offloads->esw->dev,
     "Failed to lookup bridge port vlan metadata to create MDB (MAC=%pM,vid=%u,vport=%u)\n",
     addr, vid, vport_num);
   NL_SET_ERR_MSG_FMT_MOD(extack,
            "Failed to lookup vlan metadata for MDB (MAC=%pM,vid=%u,vport=%u)",
            addr, vid, vport_num);
   return -EINVAL;
  }
 }

 err = mlx5_esw_bridge_port_mdb_attach(dev, port, addr, vid);
 if (err) {
  NL_SET_ERR_MSG_FMT_MOD(extack, "Failed to add MDB (MAC=%pM,vid=%u,vport=%u)",
           addr, vid, vport_num);
  return err;
 }

 return 0;
}

void mlx5_esw_bridge_port_mdb_del(struct net_device *dev, u16 vport_num, u16 esw_owner_vhca_id,
      const unsigned char *addr, u16 vid,
      struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge_port *port;

 port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 if (!port)
  return;

 mlx5_esw_bridge_port_mdb_detach(dev, port, addr, vid);
}

static void mlx5_esw_bridge_flush(struct mlx5_esw_bridge_offloads *br_offloads)
{
 struct mlx5_esw_bridge_port *port;
 unsigned long i;

 xa_for_each(&br_offloads->ports, i, port)
  mlx5_esw_bridge_vport_cleanup(br_offloads, port);

 WARN_ONCE(!list_empty(&br_offloads->bridges),
    "Cleaning up bridge offloads while still having bridges attached\n");
}

struct mlx5_esw_bridge_offloads *mlx5_esw_bridge_init(struct mlx5_eswitch *esw)
{
 struct mlx5_esw_bridge_offloads *br_offloads;

 ASSERT_RTNL();

 br_offloads = kvzalloc(sizeof(*br_offloads), GFP_KERNEL);
 if (!br_offloads)
  return ERR_PTR(-ENOMEM);

 INIT_LIST_HEAD(&br_offloads->bridges);
 xa_init(&br_offloads->ports);
 br_offloads->esw = esw;
 esw->br_offloads = br_offloads;
 mlx5_esw_bridge_debugfs_offloads_init(br_offloads);

 return br_offloads;
}

void mlx5_esw_bridge_cleanup(struct mlx5_eswitch *esw)
{
 struct mlx5_esw_bridge_offloads *br_offloads = esw->br_offloads;

 ASSERT_RTNL();

 if (!br_offloads)
  return;

 mlx5_esw_bridge_flush(br_offloads);
 WARN_ON(!xa_empty(&br_offloads->ports));
 mlx5_esw_bridge_debugfs_offloads_cleanup(br_offloads);

 esw->br_offloads = NULL;
 kvfree(br_offloads);
}