Merge tag 'mlx5-updates-2021-08-16' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2021-08-16

The following patchset provides two separate mlx5 updates
1) Ethtool RSS context and MQPRIO channel mode support:
  1.1) enable mlx5e netdev driver to allow creating Transport Interface RX
       (TIRs) objects on the fly to be used for ethtool RSS contexts and
       TX MQPRIO channel mode
  1.2) Introduce mlx5e_rss object to manage such TIRs.
  1.3) Ethtool support for RSS context
  1.4) Support MQPRIO channel mode

2) Bridge offloads Lag support:
   to allow adding bond net devices to mlx5 bridge
  2.1) Address bridge port by (vport_num, esw_owner_vhca_id) pair
       since vport_num is only unique per eswitch and in lag mode we
       need to manage ports from both eswitches.
  2.2) Allow connectivity between representors of different eswitch
       instances that are attached to same bridge
  2.3) Bridge LAG, Require representors to be in shared FDB mode and
       introduce local and peer ports representors,
       match on paired eswitch metadata in peer FDB entries,
       And finally support addition/deletion and aging of peer flows.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/mellanox/mlx5/core/Makefile

@@ -22,13 +22,13 @@ mlx5_core-y :=	main.o cmd.o debugfs.o fw.o eq.o uar.o pagealloc.o \
 #
 # Netdev basic
 #
-mlx5_core-$(CONFIG_MLX5_CORE_EN) += en_main.o en_common.o en_fs.o en_ethtool.o \
+mlx5_core-$(CONFIG_MLX5_CORE_EN) += en/rqt.o en/tir.o en/rss.o en/rx_res.o \
+		en/channels.o en_main.o en_common.o en_fs.o en_ethtool.o \
 		en_tx.o en_rx.o en_dim.o en_txrx.o en/xdp.o en_stats.o \
 		en_selftest.o en/port.o en/monitor_stats.o en/health.o \
 		en/reporter_tx.o en/reporter_rx.o en/params.o en/xsk/pool.o \
 		en/xsk/setup.o en/xsk/rx.o en/xsk/tx.o en/devlink.o en/ptp.o \
-		en/qos.o en/trap.o en/fs_tt_redirect.o en/rqt.o en/tir.o \
-		en/rx_res.o en/channels.o
+		en/qos.o en/trap.o en/fs_tt_redirect.o
 
 #
 # Netdev extra

drivers/net/ethernet/mellanox/mlx5/core/en.h

@@ -72,6 +72,7 @@ struct page_pool;
 #define MLX5E_SW2HW_MTU(params, swmtu) ((swmtu) + ((params)->hard_mtu))
 
 #define MLX5E_MAX_NUM_TC	8
+#define MLX5E_MAX_NUM_MQPRIO_CH_TC TC_QOPT_MAX_QUEUE
 
 #define MLX5_RX_HEADROOM NET_SKB_PAD
 #define MLX5_SKB_FRAG_SZ(len)	(SKB_DATA_ALIGN(len) +	\
@@ -248,7 +249,10 @@ struct mlx5e_params {
 	u8  rq_wq_type;
 	u8  log_rq_mtu_frames;
 	u16 num_channels;
-	u8  num_tc;
+	struct {
+		u16 mode;
+		u8 num_tc;
+	} mqprio;
 	bool rx_cqe_compress_def;
 	bool tunneled_offload_en;
 	struct dim_cq_moder rx_cq_moderation;
@@ -268,6 +272,12 @@ struct mlx5e_params {
 	bool ptp_rx;
 };
 
+static inline u8 mlx5e_get_dcb_num_tc(struct mlx5e_params *params)
+{
+	return params->mqprio.mode == TC_MQPRIO_MODE_DCB ?
+		params->mqprio.num_tc : 1;
+}
+
 enum {
 	MLX5E_RQ_STATE_ENABLED,
 	MLX5E_RQ_STATE_RECOVERING,

drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c

@@ -326,13 +326,14 @@ static int mlx5e_ptp_open_txqsqs(struct mlx5e_ptp *c,
 				 struct mlx5e_ptp_params *cparams)
 {
 	struct mlx5e_params *params = &cparams->params;
+	u8 num_tc = mlx5e_get_dcb_num_tc(params);
 	int ix_base;
 	int err;
 	int tc;
 
-	ix_base = params->num_tc * params->num_channels;
+	ix_base = num_tc * params->num_channels;
 
-	for (tc = 0; tc < params->num_tc; tc++) {
+	for (tc = 0; tc < num_tc; tc++) {
 		int txq_ix = ix_base + tc;
 
 		err = mlx5e_ptp_open_txqsq(c, c->priv->tisn[c->lag_port][tc], txq_ix,
@@ -365,9 +366,12 @@ static int mlx5e_ptp_open_tx_cqs(struct mlx5e_ptp *c,
 	struct mlx5e_create_cq_param ccp = {};
 	struct dim_cq_moder ptp_moder = {};
 	struct mlx5e_cq_param *cq_param;
+	u8 num_tc;
 	int err;
 	int tc;
 
+	num_tc = mlx5e_get_dcb_num_tc(params);
+
 	ccp.node     = dev_to_node(mlx5_core_dma_dev(c->mdev));
 	ccp.ch_stats = c->stats;
 	ccp.napi     = &c->napi;
@@ -375,7 +379,7 @@ static int mlx5e_ptp_open_tx_cqs(struct mlx5e_ptp *c,
 
 	cq_param = &cparams->txq_sq_param.cqp;
 
-	for (tc = 0; tc < params->num_tc; tc++) {
+	for (tc = 0; tc < num_tc; tc++) {
 		struct mlx5e_cq *cq = &c->ptpsq[tc].txqsq.cq;
 
 		err = mlx5e_open_cq(c->priv, ptp_moder, cq_param, &ccp, cq);
@@ -383,7 +387,7 @@ static int mlx5e_ptp_open_tx_cqs(struct mlx5e_ptp *c,
 			goto out_err_txqsq_cq;
 	}
 
-	for (tc = 0; tc < params->num_tc; tc++) {
+	for (tc = 0; tc < num_tc; tc++) {
 		struct mlx5e_cq *cq = &c->ptpsq[tc].ts_cq;
 		struct mlx5e_ptpsq *ptpsq = &c->ptpsq[tc];
 
@@ -399,7 +403,7 @@ static int mlx5e_ptp_open_tx_cqs(struct mlx5e_ptp *c,
 out_err_ts_cq:
 	for (--tc; tc >= 0; tc--)
 		mlx5e_close_cq(&c->ptpsq[tc].ts_cq);
-	tc = params->num_tc;
+	tc = num_tc;
 out_err_txqsq_cq:
 	for (--tc; tc >= 0; tc--)
 		mlx5e_close_cq(&c->ptpsq[tc].txqsq.cq);
@@ -475,7 +479,7 @@ static void mlx5e_ptp_build_params(struct mlx5e_ptp *c,
 	params->num_channels = orig->num_channels;
 	params->hard_mtu = orig->hard_mtu;
 	params->sw_mtu = orig->sw_mtu;
-	params->num_tc = orig->num_tc;
+	params->mqprio = orig->mqprio;
 
 	/* SQ */
 	if (test_bit(MLX5E_PTP_STATE_TX, c->state)) {
@@ -680,7 +684,7 @@ int mlx5e_ptp_open(struct mlx5e_priv *priv, struct mlx5e_params *params,
 	c->pdev     = mlx5_core_dma_dev(priv->mdev);
 	c->netdev   = priv->netdev;
 	c->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey.key);
-	c->num_tc   = params->num_tc;
+	c->num_tc   = mlx5e_get_dcb_num_tc(params);
 	c->stats    = &priv->ptp_stats.ch;
 	c->lag_port = lag_port;
 

drivers/net/ethernet/mellanox/mlx5/core/en/qos.c

@@ -132,7 +132,7 @@ static u16 mlx5e_qid_from_qos(struct mlx5e_channels *chs, u16 qid)
 	 */
 	bool is_ptp = MLX5E_GET_PFLAG(&chs->params, MLX5E_PFLAG_TX_PORT_TS);
 
-	return (chs->params.num_channels + is_ptp) * chs->params.num_tc + qid;
+	return (chs->params.num_channels + is_ptp) * mlx5e_get_dcb_num_tc(&chs->params) + qid;
 }
 
 int mlx5e_get_txq_by_classid(struct mlx5e_priv *priv, u16 classid)

drivers/net/ethernet/mellanox/mlx5/core/en/rep/bridge.c

@@ -15,9 +15,116 @@ struct mlx5_bridge_switchdev_fdb_work {
 	struct work_struct work;
 	struct switchdev_notifier_fdb_info fdb_info;
 	struct net_device *dev;
+	struct mlx5_esw_bridge_offloads *br_offloads;
 	bool add;
 };
 
+static bool mlx5_esw_bridge_dev_same_esw(struct net_device *dev, struct mlx5_eswitch *esw)
+{
+	struct mlx5e_priv *priv = netdev_priv(dev);
+
+	return esw == priv->mdev->priv.eswitch;
+}
+
+static bool mlx5_esw_bridge_dev_same_hw(struct net_device *dev, struct mlx5_eswitch *esw)
+{
+	struct mlx5e_priv *priv = netdev_priv(dev);
+	struct mlx5_core_dev *mdev, *esw_mdev;
+	u64 system_guid, esw_system_guid;
+
+	mdev = priv->mdev;
+	esw_mdev = esw->dev;
+
+	system_guid = mlx5_query_nic_system_image_guid(mdev);
+	esw_system_guid = mlx5_query_nic_system_image_guid(esw_mdev);
+
+	return system_guid == esw_system_guid;
+}
+
+static struct net_device *
+mlx5_esw_bridge_lag_rep_get(struct net_device *dev, struct mlx5_eswitch *esw)
+{
+	struct net_device *lower;
+	struct list_head *iter;
+
+	netdev_for_each_lower_dev(dev, lower, iter) {
+		struct mlx5_core_dev *mdev;
+		struct mlx5e_priv *priv;
+
+		if (!mlx5e_eswitch_rep(lower))
+			continue;
+
+		priv = netdev_priv(lower);
+		mdev = priv->mdev;
+		if (mlx5_lag_is_shared_fdb(mdev) && mlx5_esw_bridge_dev_same_esw(lower, esw))
+			return lower;
+	}
+
+	return NULL;
+}
+
+static struct net_device *
+mlx5_esw_bridge_rep_vport_num_vhca_id_get(struct net_device *dev, struct mlx5_eswitch *esw,
+					  u16 *vport_num, u16 *esw_owner_vhca_id)
+{
+	struct mlx5e_rep_priv *rpriv;
+	struct mlx5e_priv *priv;
+
+	if (netif_is_lag_master(dev))
+		dev = mlx5_esw_bridge_lag_rep_get(dev, esw);
+
+	if (!dev || !mlx5e_eswitch_rep(dev) || !mlx5_esw_bridge_dev_same_hw(dev, esw))
+		return NULL;
+
+	priv = netdev_priv(dev);
+	rpriv = priv->ppriv;
+	*vport_num = rpriv->rep->vport;
+	*esw_owner_vhca_id = MLX5_CAP_GEN(priv->mdev, vhca_id);
+	return dev;
+}
+
+static struct net_device *
+mlx5_esw_bridge_lower_rep_vport_num_vhca_id_get(struct net_device *dev, struct mlx5_eswitch *esw,
+						u16 *vport_num, u16 *esw_owner_vhca_id)
+{
+	struct net_device *lower_dev;
+	struct list_head *iter;
+
+	if (netif_is_lag_master(dev) || mlx5e_eswitch_rep(dev))
+		return mlx5_esw_bridge_rep_vport_num_vhca_id_get(dev, esw, vport_num,
+								 esw_owner_vhca_id);
+
+	netdev_for_each_lower_dev(dev, lower_dev, iter) {
+		struct net_device *rep;
+
+		if (netif_is_bridge_master(lower_dev))
+			continue;
+
+		rep = mlx5_esw_bridge_lower_rep_vport_num_vhca_id_get(lower_dev, esw, vport_num,
+								      esw_owner_vhca_id);
+		if (rep)
+			return rep;
+	}
+
+	return NULL;
+}
+
+static bool mlx5_esw_bridge_is_local(struct net_device *dev, struct net_device *rep,
+				     struct mlx5_eswitch *esw)
+{
+	struct mlx5_core_dev *mdev;
+	struct mlx5e_priv *priv;
+
+	if (!mlx5_esw_bridge_dev_same_esw(rep, esw))
+		return false;
+
+	priv = netdev_priv(rep);
+	mdev = priv->mdev;
+	if (netif_is_lag_master(dev))
+		return mlx5_lag_is_shared_fdb(mdev) && mlx5_lag_is_master(mdev);
+	return true;
+}
+
 static int mlx5_esw_bridge_port_changeupper(struct notifier_block *nb, void *ptr)
 {
 	struct mlx5_esw_bridge_offloads *br_offloads = container_of(nb,
@@ -25,37 +132,36 @@ static int mlx5_esw_bridge_port_changeupper(struct notifier_block *nb, void *ptr
 								    netdev_nb);
 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 	struct netdev_notifier_changeupper_info *info = ptr;
+	struct net_device *upper = info->upper_dev, *rep;
+	struct mlx5_eswitch *esw = br_offloads->esw;
+	u16 vport_num, esw_owner_vhca_id;
 	struct netlink_ext_ack *extack;
-	struct mlx5e_rep_priv *rpriv;
-	struct mlx5_eswitch *esw;
-	struct mlx5_vport *vport;
-	struct net_device *upper;
-	struct mlx5e_priv *priv;
-	u16 vport_num;
-
-	if (!mlx5e_eswitch_rep(dev))
-		return 0;
+	int ifindex = upper->ifindex;
+	int err;
 
-	upper = info->upper_dev;
 	if (!netif_is_bridge_master(upper))
 		return 0;
 
-	esw = br_offloads->esw;
-	priv = netdev_priv(dev);
-	if (esw != priv->mdev->priv.eswitch)
+	rep = mlx5_esw_bridge_rep_vport_num_vhca_id_get(dev, esw, &vport_num, &esw_owner_vhca_id);
+	if (!rep)
 		return 0;
 
-	rpriv = priv->ppriv;
-	vport_num = rpriv->rep->vport;
-	vport = mlx5_eswitch_get_vport(esw, vport_num);
-	if (IS_ERR(vport))
-		return PTR_ERR(vport);
-
 	extack = netdev_notifier_info_to_extack(&info->info);
 
-	return info->linking ?
-		mlx5_esw_bridge_vport_link(upper->ifindex, br_offloads, vport, extack) :
-		mlx5_esw_bridge_vport_unlink(upper->ifindex, br_offloads, vport, extack);
+	if (mlx5_esw_bridge_is_local(dev, rep, esw))
+		err = info->linking ?
+			mlx5_esw_bridge_vport_link(ifindex, vport_num, esw_owner_vhca_id,
+						   br_offloads, extack) :
+			mlx5_esw_bridge_vport_unlink(ifindex, vport_num, esw_owner_vhca_id,
+						     br_offloads, extack);
+	else if (mlx5_esw_bridge_dev_same_hw(rep, esw))
+		err = info->linking ?
+			mlx5_esw_bridge_vport_peer_link(ifindex, vport_num, esw_owner_vhca_id,
+							br_offloads, extack) :
+			mlx5_esw_bridge_vport_peer_unlink(ifindex, vport_num, esw_owner_vhca_id,
+							  br_offloads, extack);
+
+	return err;
 }
 
 static int mlx5_esw_bridge_switchdev_port_event(struct notifier_block *nb,
@@ -75,31 +181,28 @@ static int mlx5_esw_bridge_switchdev_port_event(struct notifier_block *nb,
 	return notifier_from_errno(err);
 }
 
-static int mlx5_esw_bridge_port_obj_add(struct net_device *dev,
-					const void *ctx,
-					const struct switchdev_obj *obj,
-					struct netlink_ext_ack *extack)
+static int
+mlx5_esw_bridge_port_obj_add(struct net_device *dev,
+			     struct switchdev_notifier_port_obj_info *port_obj_info,
+			     struct mlx5_esw_bridge_offloads *br_offloads)
 {
+	struct netlink_ext_ack *extack = switchdev_notifier_info_to_extack(&port_obj_info->info);
+	const struct switchdev_obj *obj = port_obj_info->obj;
 	const struct switchdev_obj_port_vlan *vlan;
-	struct mlx5e_rep_priv *rpriv;
-	struct mlx5_eswitch *esw;
-	struct mlx5_vport *vport;
-	struct mlx5e_priv *priv;
-	u16 vport_num;
-	int err = 0;
+	u16 vport_num, esw_owner_vhca_id;
+	int err;
 
-	priv = netdev_priv(dev);
-	rpriv = priv->ppriv;
-	vport_num = rpriv->rep->vport;
-	esw = priv->mdev->priv.eswitch;
-	vport = mlx5_eswitch_get_vport(esw, vport_num);
-	if (IS_ERR(vport))
-		return PTR_ERR(vport);
+	if (!mlx5_esw_bridge_rep_vport_num_vhca_id_get(dev, br_offloads->esw, &vport_num,
+						       &esw_owner_vhca_id))
+		return 0;
+
+	port_obj_info->handled = true;
 
 	switch (obj->id) {
 	case SWITCHDEV_OBJ_ID_PORT_VLAN:
 		vlan = SWITCHDEV_OBJ_PORT_VLAN(obj);
-		err = mlx5_esw_bridge_port_vlan_add(vlan->vid, vlan->flags, esw, vport, extack);
+		err = mlx5_esw_bridge_port_vlan_add(vport_num, esw_owner_vhca_id, vlan->vid,
+						    vlan->flags, br_offloads, extack);
 		break;
 	default:
 		return -EOPNOTSUPP;
@@ -107,29 +210,25 @@ static int mlx5_esw_bridge_port_obj_add(struct net_device *dev,
 	return err;
 }
 
-static int mlx5_esw_bridge_port_obj_del(struct net_device *dev,
-					const void *ctx,
-					const struct switchdev_obj *obj)
+static int
+mlx5_esw_bridge_port_obj_del(struct net_device *dev,
+			     struct switchdev_notifier_port_obj_info *port_obj_info,
+			     struct mlx5_esw_bridge_offloads *br_offloads)
 {
+	const struct switchdev_obj *obj = port_obj_info->obj;
 	const struct switchdev_obj_port_vlan *vlan;
-	struct mlx5e_rep_priv *rpriv;
-	struct mlx5_eswitch *esw;
-	struct mlx5_vport *vport;
-	struct mlx5e_priv *priv;
-	u16 vport_num;
+	u16 vport_num, esw_owner_vhca_id;
 
-	priv = netdev_priv(dev);
-	rpriv = priv->ppriv;
-	vport_num = rpriv->rep->vport;
-	esw = priv->mdev->priv.eswitch;
-	vport = mlx5_eswitch_get_vport(esw, vport_num);
-	if (IS_ERR(vport))
-		return PTR_ERR(vport);
+	if (!mlx5_esw_bridge_rep_vport_num_vhca_id_get(dev, br_offloads->esw, &vport_num,
+						       &esw_owner_vhca_id))
+		return 0;
+
+	port_obj_info->handled = true;
 
 	switch (obj->id) {
 	case SWITCHDEV_OBJ_ID_PORT_VLAN:
 		vlan = SWITCHDEV_OBJ_PORT_VLAN(obj);
-		mlx5_esw_bridge_port_vlan_del(vlan->vid, esw, vport);
+		mlx5_esw_bridge_port_vlan_del(vport_num, esw_owner_vhca_id, vlan->vid, br_offloads);
 		break;
 	default:
 		return -EOPNOTSUPP;
@@ -137,25 +236,21 @@ static int mlx5_esw_bridge_port_obj_del(struct net_device *dev,
 	return 0;
 }
 
-static int mlx5_esw_bridge_port_obj_attr_set(struct net_device *dev,
-					     const void *ctx,
-					     const struct switchdev_attr *attr,
-					     struct netlink_ext_ack *extack)
+static int
+mlx5_esw_bridge_port_obj_attr_set(struct net_device *dev,
+				  struct switchdev_notifier_port_attr_info *port_attr_info,
+				  struct mlx5_esw_bridge_offloads *br_offloads)
 {
-	struct mlx5e_rep_priv *rpriv;
-	struct mlx5_eswitch *esw;
-	struct mlx5_vport *vport;
-	struct mlx5e_priv *priv;
-	u16 vport_num;
-	int err = 0;
+	struct netlink_ext_ack *extack = switchdev_notifier_info_to_extack(&port_attr_info->info);
+	const struct switchdev_attr *attr = port_attr_info->attr;
+	u16 vport_num, esw_owner_vhca_id;
+	int err;
 
-	priv = netdev_priv(dev);
-	rpriv = priv->ppriv;
-	vport_num = rpriv->rep->vport;
-	esw = priv->mdev->priv.eswitch;
-	vport = mlx5_eswitch_get_vport(esw, vport_num);
-	if (IS_ERR(vport))
-		return PTR_ERR(vport);
+	if (!mlx5_esw_bridge_lower_rep_vport_num_vhca_id_get(dev, br_offloads->esw, &vport_num,
+							     &esw_owner_vhca_id))
+		return 0;
+
+	port_attr_info->handled = true;
 
 	switch (attr->id) {
 	case SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS:
@@ -167,10 +262,12 @@ static int mlx5_esw_bridge_port_obj_attr_set(struct net_device *dev,
 	case SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS:
 		break;
 	case SWITCHDEV_ATTR_ID_BRIDGE_AGEING_TIME:
-		err = mlx5_esw_bridge_ageing_time_set(attr->u.ageing_time, esw, vport);
+		err = mlx5_esw_bridge_ageing_time_set(vport_num, esw_owner_vhca_id,
+						      attr->u.ageing_time, br_offloads);
 		break;
 	case SWITCHDEV_ATTR_ID_BRIDGE_VLAN_FILTERING:
-		err = mlx5_esw_bridge_vlan_filtering_set(attr->u.vlan_filtering, esw, vport);
+		err = mlx5_esw_bridge_vlan_filtering_set(vport_num, esw_owner_vhca_id,
+							 attr->u.vlan_filtering, br_offloads);
 		break;
 	default:
 		err = -EOPNOTSUPP;
@@ -179,27 +276,24 @@ static int mlx5_esw_bridge_port_obj_attr_set(struct net_device *dev,
 	return err;
 }
 
-static int mlx5_esw_bridge_event_blocking(struct notifier_block *unused,
+static int mlx5_esw_bridge_event_blocking(struct notifier_block *nb,
 					  unsigned long event, void *ptr)
 {
+	struct mlx5_esw_bridge_offloads *br_offloads = container_of(nb,
+								    struct mlx5_esw_bridge_offloads,
+								    nb_blk);
 	struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
 	int err;
 
 	switch (event) {
 	case SWITCHDEV_PORT_OBJ_ADD:
-		err = switchdev_handle_port_obj_add(dev, ptr,
-						    mlx5e_eswitch_rep,
-						    mlx5_esw_bridge_port_obj_add);
+		err = mlx5_esw_bridge_port_obj_add(dev, ptr, br_offloads);
 		break;
 	case SWITCHDEV_PORT_OBJ_DEL:
-		err = switchdev_handle_port_obj_del(dev, ptr,
-						    mlx5e_eswitch_rep,
-						    mlx5_esw_bridge_port_obj_del);
+		err = mlx5_esw_bridge_port_obj_del(dev, ptr, br_offloads);
 		break;
 	case SWITCHDEV_PORT_ATTR_SET:
-		err = switchdev_handle_port_attr_set(dev, ptr,
-						     mlx5e_eswitch_rep,
-						     mlx5_esw_bridge_port_obj_attr_set);
+		err = mlx5_esw_bridge_port_obj_attr_set(dev, ptr, br_offloads);
 		break;
 	default:
 		err = 0;
@@ -222,27 +316,23 @@ static void mlx5_esw_bridge_switchdev_fdb_event_work(struct work_struct *work)
 		container_of(work, struct mlx5_bridge_switchdev_fdb_work, work);
 	struct switchdev_notifier_fdb_info *fdb_info =
 		&fdb_work->fdb_info;
+	struct mlx5_esw_bridge_offloads *br_offloads =
+		fdb_work->br_offloads;
 	struct net_device *dev = fdb_work->dev;
-	struct mlx5e_rep_priv *rpriv;
-	struct mlx5_eswitch *esw;
-	struct mlx5_vport *vport;
-	struct mlx5e_priv *priv;
-	u16 vport_num;
+	u16 vport_num, esw_owner_vhca_id;
 
 	rtnl_lock();
 
-	priv = netdev_priv(dev);
-	rpriv = priv->ppriv;
-	vport_num = rpriv->rep->vport;
-	esw = priv->mdev->priv.eswitch;
-	vport = mlx5_eswitch_get_vport(esw, vport_num);
-	if (IS_ERR(vport))
+	if (!mlx5_esw_bridge_rep_vport_num_vhca_id_get(dev, br_offloads->esw, &vport_num,
+						       &esw_owner_vhca_id))
 		goto out;
 
 	if (fdb_work->add)
-		mlx5_esw_bridge_fdb_create(dev, esw, vport, fdb_info);
+		mlx5_esw_bridge_fdb_create(dev, vport_num, esw_owner_vhca_id, br_offloads,
+					   fdb_info);
 	else
-		mlx5_esw_bridge_fdb_remove(dev, esw, vport, fdb_info);
+		mlx5_esw_bridge_fdb_remove(dev, vport_num, esw_owner_vhca_id, br_offloads,
+					   fdb_info);
 
 out:
 	rtnl_unlock();
@@ -251,7 +341,8 @@ static void mlx5_esw_bridge_switchdev_fdb_event_work(struct work_struct *work)
 
 static struct mlx5_bridge_switchdev_fdb_work *
 mlx5_esw_bridge_init_switchdev_fdb_work(struct net_device *dev, bool add,
-					struct switchdev_notifier_fdb_info *fdb_info)
+					struct switchdev_notifier_fdb_info *fdb_info,
+					struct mlx5_esw_bridge_offloads *br_offloads)
 {
 	struct mlx5_bridge_switchdev_fdb_work *work;
 	u8 *addr;
@@ -273,6 +364,7 @@ mlx5_esw_bridge_init_switchdev_fdb_work(struct net_device *dev, bool add,
 
 	dev_hold(dev);
 	work->dev = dev;
+	work->br_offloads = br_offloads;
 	work->add = add;
 	return work;
 }
@@ -286,20 +378,14 @@ static int mlx5_esw_bridge_switchdev_event(struct notifier_block *nb,
 	struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
 	struct switchdev_notifier_fdb_info *fdb_info;
 	struct mlx5_bridge_switchdev_fdb_work *work;
+	struct mlx5_eswitch *esw = br_offloads->esw;
 	struct switchdev_notifier_info *info = ptr;
-	struct net_device *upper;
-	struct mlx5e_priv *priv;
-
-	if (!mlx5e_eswitch_rep(dev))
-		return NOTIFY_DONE;
-	priv = netdev_priv(dev);
-	if (priv->mdev->priv.eswitch != br_offloads->esw)
-		return NOTIFY_DONE;
+	u16 vport_num, esw_owner_vhca_id;
+	struct net_device *upper, *rep;
 
 	if (event == SWITCHDEV_PORT_ATTR_SET) {
-		int err = switchdev_handle_port_attr_set(dev, ptr,
-							 mlx5e_eswitch_rep,
-							 mlx5_esw_bridge_port_obj_attr_set);
+		int err = mlx5_esw_bridge_port_obj_attr_set(dev, ptr, br_offloads);
+
 		return notifier_from_errno(err);
 	}
 
@@ -309,7 +395,27 @@ static int mlx5_esw_bridge_switchdev_event(struct notifier_block *nb,
 	if (!netif_is_bridge_master(upper))
 		return NOTIFY_DONE;
 
+	rep = mlx5_esw_bridge_rep_vport_num_vhca_id_get(dev, esw, &vport_num, &esw_owner_vhca_id);
+	if (!rep)
+		return NOTIFY_DONE;
+
 	switch (event) {
+	case SWITCHDEV_FDB_ADD_TO_BRIDGE:
+		/* only handle the event on native eswtich of representor */
+		if (!mlx5_esw_bridge_is_local(dev, rep, esw))
+			break;
+
+		fdb_info = container_of(info,
+					struct switchdev_notifier_fdb_info,
+					info);
+		mlx5_esw_bridge_fdb_update_used(dev, vport_num, esw_owner_vhca_id, br_offloads,
+						fdb_info);
+		break;
+	case SWITCHDEV_FDB_DEL_TO_BRIDGE:
+		/* only handle the event on peers */
+		if (mlx5_esw_bridge_is_local(dev, rep, esw))
+			break;
+		fallthrough;
 	case SWITCHDEV_FDB_ADD_TO_DEVICE:
 	case SWITCHDEV_FDB_DEL_TO_DEVICE:
 		fdb_info = container_of(info,
@@ -318,7 +424,8 @@ static int mlx5_esw_bridge_switchdev_event(struct notifier_block *nb,
 
 		work = mlx5_esw_bridge_init_switchdev_fdb_work(dev,
 							       event == SWITCHDEV_FDB_ADD_TO_DEVICE,
-							       fdb_info);
+							       fdb_info,
+							       br_offloads);
 		if (IS_ERR(work)) {
 			WARN_ONCE(1, "Failed to init switchdev work, err=%ld",
 				  PTR_ERR(work));

drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c

@@ -372,7 +372,7 @@ static int mlx5e_tx_reporter_diagnose(struct devlink_health_reporter *reporter,
 	for (i = 0; i < priv->channels.num; i++) {
 		struct mlx5e_channel *c = priv->channels.c[i];
 
-		for (tc = 0; tc < priv->channels.params.num_tc; tc++) {
+		for (tc = 0; tc < mlx5e_get_dcb_num_tc(&priv->channels.params); tc++) {
 			struct mlx5e_txqsq *sq = &c->sq[tc];
 
 			err = mlx5e_tx_reporter_build_diagnose_output(fmsg, sq, tc);
@@ -384,7 +384,7 @@ static int mlx5e_tx_reporter_diagnose(struct devlink_health_reporter *reporter,
 	if (!ptp_ch || !test_bit(MLX5E_PTP_STATE_TX, ptp_ch->state))
 		goto close_sqs_nest;
 
-	for (tc = 0; tc < priv->channels.params.num_tc; tc++) {
+	for (tc = 0; tc < mlx5e_get_dcb_num_tc(&priv->channels.params); tc++) {
 		err = mlx5e_tx_reporter_build_diagnose_output_ptpsq(fmsg,
 								    &ptp_ch->ptpsq[tc],
 								    tc);
@@ -494,7 +494,7 @@ static int mlx5e_tx_reporter_dump_all_sqs(struct mlx5e_priv *priv,
 	for (i = 0; i < priv->channels.num; i++) {
 		struct mlx5e_channel *c = priv->channels.c[i];
 
-		for (tc = 0; tc < priv->channels.params.num_tc; tc++) {
+		for (tc = 0; tc < mlx5e_get_dcb_num_tc(&priv->channels.params); tc++) {
 			struct mlx5e_txqsq *sq = &c->sq[tc];
 
 			err = mlx5e_health_queue_dump(priv, fmsg, sq->sqn, "SQ");
@@ -504,7 +504,7 @@ static int mlx5e_tx_reporter_dump_all_sqs(struct mlx5e_priv *priv,
 	}
 
 	if (ptp_ch && test_bit(MLX5E_PTP_STATE_TX, ptp_ch->state)) {
-		for (tc = 0; tc < priv->channels.params.num_tc; tc++) {
+		for (tc = 0; tc < mlx5e_get_dcb_num_tc(&priv->channels.params); tc++) {
 			struct mlx5e_txqsq *sq = &ptp_ch->ptpsq[tc].txqsq;
 
 			err = mlx5e_health_queue_dump(priv, fmsg, sq->sqn, "PTP SQ");

drivers/net/ethernet/mellanox/mlx5/core/en/rss.c

+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+// Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES.
+
+#include "rss.h"
+
+#define mlx5e_rss_warn(__dev, format, ...)			\
+	dev_warn((__dev)->device, "%s:%d:(pid %d): " format,	\
+		 __func__, __LINE__, current->pid,		\
+		 ##__VA_ARGS__)
+
+static const struct mlx5e_rss_params_traffic_type rss_default_config[MLX5E_NUM_INDIR_TIRS] = {
+	[MLX5_TT_IPV4_TCP] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
+		.l4_prot_type = MLX5_L4_PROT_TYPE_TCP,
+		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
+	},
+	[MLX5_TT_IPV6_TCP] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
+		.l4_prot_type = MLX5_L4_PROT_TYPE_TCP,
+		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
+	},
+	[MLX5_TT_IPV4_UDP] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
+		.l4_prot_type = MLX5_L4_PROT_TYPE_UDP,
+		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
+	},
+	[MLX5_TT_IPV6_UDP] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
+		.l4_prot_type = MLX5_L4_PROT_TYPE_UDP,
+		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
+	},
+	[MLX5_TT_IPV4_IPSEC_AH] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
+		.l4_prot_type = 0,
+		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
+	},
+	[MLX5_TT_IPV6_IPSEC_AH] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
+		.l4_prot_type = 0,
+		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
+	},
+	[MLX5_TT_IPV4_IPSEC_ESP] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
+		.l4_prot_type = 0,
+		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
+	},
+	[MLX5_TT_IPV6_IPSEC_ESP] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
+		.l4_prot_type = 0,
+		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
+	},
+	[MLX5_TT_IPV4] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
+		.l4_prot_type = 0,
+		.rx_hash_fields = MLX5_HASH_IP,
+	},
+	[MLX5_TT_IPV6] = {
+		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
+		.l4_prot_type = 0,
+		.rx_hash_fields = MLX5_HASH_IP,
+	},
+};
+
+struct mlx5e_rss_params_traffic_type
+mlx5e_rss_get_default_tt_config(enum mlx5_traffic_types tt)
+{
+	return rss_default_config[tt];
+}
+
+struct mlx5e_rss {
+	struct mlx5e_rss_params_hash hash;
+	struct mlx5e_rss_params_indir indir;
+	u32 rx_hash_fields[MLX5E_NUM_INDIR_TIRS];
+	struct mlx5e_tir *tir[MLX5E_NUM_INDIR_TIRS];
+	struct mlx5e_tir *inner_tir[MLX5E_NUM_INDIR_TIRS];
+	struct mlx5e_rqt rqt;
+	struct mlx5_core_dev *mdev;
+	u32 drop_rqn;
+	bool inner_ft_support;
+	bool enabled;
+	refcount_t refcnt;
+};
+
+struct mlx5e_rss *mlx5e_rss_alloc(void)
+{
+	return kvzalloc(sizeof(struct mlx5e_rss), GFP_KERNEL);
+}
+
+void mlx5e_rss_free(struct mlx5e_rss *rss)
+{
+	kvfree(rss);
+}
+
+static void mlx5e_rss_params_init(struct mlx5e_rss *rss)
+{
+	enum mlx5_traffic_types tt;
+
+	rss->hash.hfunc = ETH_RSS_HASH_TOP;
+	netdev_rss_key_fill(rss->hash.toeplitz_hash_key,
+			    sizeof(rss->hash.toeplitz_hash_key));
+	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++)
+		rss->rx_hash_fields[tt] =
+			mlx5e_rss_get_default_tt_config(tt).rx_hash_fields;
+}
+
+static struct mlx5e_tir **rss_get_tirp(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+				       bool inner)
+{
+	return inner ? &rss->inner_tir[tt] : &rss->tir[tt];
+}
+
+static struct mlx5e_tir *rss_get_tir(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+				     bool inner)
+{
+	return *rss_get_tirp(rss, tt, inner);
+}
+
+static struct mlx5e_rss_params_traffic_type
+mlx5e_rss_get_tt_config(struct mlx5e_rss *rss, enum mlx5_traffic_types tt)
+{
+	struct mlx5e_rss_params_traffic_type rss_tt;
+
+	rss_tt = mlx5e_rss_get_default_tt_config(tt);
+	rss_tt.rx_hash_fields = rss->rx_hash_fields[tt];
+	return rss_tt;
+}
+
+static int mlx5e_rss_create_tir(struct mlx5e_rss *rss,
+				enum mlx5_traffic_types tt,
+				const struct mlx5e_lro_param *init_lro_param,
+				bool inner)
+{
+	struct mlx5e_rss_params_traffic_type rss_tt;
+	struct mlx5e_tir_builder *builder;
+	struct mlx5e_tir **tir_p;
+	struct mlx5e_tir *tir;
+	u32 rqtn;
+	int err;
+
+	if (inner && !rss->inner_ft_support) {
+		mlx5e_rss_warn(rss->mdev,
+			       "Cannot create inner indirect TIR[%d], RSS inner FT is not supported.\n",
+			       tt);
+		return -EINVAL;
+	}
+
+	tir_p = rss_get_tirp(rss, tt, inner);
+	if (*tir_p)
+		return -EINVAL;
+
+	tir = kvzalloc(sizeof(*tir), GFP_KERNEL);
+	if (!tir)
+		return -ENOMEM;
+
+	builder = mlx5e_tir_builder_alloc(false);
+	if (!builder) {
+		err = -ENOMEM;
+		goto free_tir;
+	}
+
+	rqtn = mlx5e_rqt_get_rqtn(&rss->rqt);
+	mlx5e_tir_builder_build_rqt(builder, rss->mdev->mlx5e_res.hw_objs.td.tdn,
+				    rqtn, rss->inner_ft_support);
+	mlx5e_tir_builder_build_lro(builder, init_lro_param);
+	rss_tt = mlx5e_rss_get_tt_config(rss, tt);
+	mlx5e_tir_builder_build_rss(builder, &rss->hash, &rss_tt, inner);
+
+	err = mlx5e_tir_init(tir, builder, rss->mdev, true);
+	mlx5e_tir_builder_free(builder);
+	if (err) {
+		mlx5e_rss_warn(rss->mdev, "Failed to create %sindirect TIR: err = %d, tt = %d\n",
+			       inner ? "inner " : "", err, tt);
+		goto free_tir;
+	}
+
+	*tir_p = tir;
+	return 0;
+
+free_tir:
+	kvfree(tir);
+	return err;
+}
+
+static void mlx5e_rss_destroy_tir(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+				  bool inner)
+{
+	struct mlx5e_tir **tir_p;
+	struct mlx5e_tir *tir;
+
+	tir_p = rss_get_tirp(rss, tt, inner);
+	if (!*tir_p)
+		return;
+
+	tir = *tir_p;
+	mlx5e_tir_destroy(tir);
+	kvfree(tir);
+	*tir_p = NULL;
+}
+
+static int mlx5e_rss_create_tirs(struct mlx5e_rss *rss,
+				 const struct mlx5e_lro_param *init_lro_param,
+				 bool inner)
+{
+	enum mlx5_traffic_types tt, max_tt;
+	int err;
+
+	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++) {
+		err = mlx5e_rss_create_tir(rss, tt, init_lro_param, inner);
+		if (err)
+			goto err_destroy_tirs;
+	}
+
+	return 0;
+
+err_destroy_tirs:
+	max_tt = tt;
+	for (tt = 0; tt < max_tt; tt++)
+		mlx5e_rss_destroy_tir(rss, tt, inner);
+	return err;
+}
+
+static void mlx5e_rss_destroy_tirs(struct mlx5e_rss *rss, bool inner)
+{
+	enum mlx5_traffic_types tt;
+
+	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++)
+		mlx5e_rss_destroy_tir(rss, tt, inner);
+}
+
+static int mlx5e_rss_update_tir(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+				bool inner)
+{
+	struct mlx5e_rss_params_traffic_type rss_tt;
+	struct mlx5e_tir_builder *builder;
+	struct mlx5e_tir *tir;
+	int err;
+
+	tir = rss_get_tir(rss, tt, inner);
+	if (!tir)
+		return 0;
+
+	builder = mlx5e_tir_builder_alloc(true);
+	if (!builder)
+		return -ENOMEM;
+
+	rss_tt = mlx5e_rss_get_tt_config(rss, tt);
+
+	mlx5e_tir_builder_build_rss(builder, &rss->hash, &rss_tt, inner);
+	err = mlx5e_tir_modify(tir, builder);
+
+	mlx5e_tir_builder_free(builder);
+	return err;
+}
+
+static int mlx5e_rss_update_tirs(struct mlx5e_rss *rss)
+{
+	enum mlx5_traffic_types tt;
+	int err, retval;
+
+	retval = 0;
+
+	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++) {
+		err = mlx5e_rss_update_tir(rss, tt, false);
+		if (err) {
+			retval = retval ? : err;
+			mlx5e_rss_warn(rss->mdev,
+				       "Failed to update RSS hash of indirect TIR for traffic type %d: err = %d\n",
+				       tt, err);
+		}
+
+		if (!rss->inner_ft_support)
+			continue;
+
+		err = mlx5e_rss_update_tir(rss, tt, true);
+		if (err) {
+			retval = retval ? : err;
+			mlx5e_rss_warn(rss->mdev,
+				       "Failed to update RSS hash of inner indirect TIR for traffic type %d: err = %d\n",
+				       tt, err);
+		}
+	}
+	return retval;
+}
+
+int mlx5e_rss_init_no_tirs(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
+			   bool inner_ft_support, u32 drop_rqn)
+{
+	rss->mdev = mdev;
+	rss->inner_ft_support = inner_ft_support;
+	rss->drop_rqn = drop_rqn;
+
+	mlx5e_rss_params_init(rss);
+	refcount_set(&rss->refcnt, 1);
+
+	return mlx5e_rqt_init_direct(&rss->rqt, mdev, true, drop_rqn);
+}
+
+int mlx5e_rss_init(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
+		   bool inner_ft_support, u32 drop_rqn,
+		   const struct mlx5e_lro_param *init_lro_param)
+{
+	int err;
+
+	err = mlx5e_rss_init_no_tirs(rss, mdev, inner_ft_support, drop_rqn);
+	if (err)
+		goto err_out;
+
+	err = mlx5e_rss_create_tirs(rss, init_lro_param, false);
+	if (err)
+		goto err_destroy_rqt;
+
+	if (inner_ft_support) {
+		err = mlx5e_rss_create_tirs(rss, init_lro_param, true);
+		if (err)
+			goto err_destroy_tirs;
+	}
+
+	return 0;
+
+err_destroy_tirs:
+	mlx5e_rss_destroy_tirs(rss, false);
+err_destroy_rqt:
+	mlx5e_rqt_destroy(&rss->rqt);
+err_out:
+	return err;
+}
+
+int mlx5e_rss_cleanup(struct mlx5e_rss *rss)
+{
+	if (!refcount_dec_if_one(&rss->refcnt))
+		return -EBUSY;
+
+	mlx5e_rss_destroy_tirs(rss, false);
+
+	if (rss->inner_ft_support)
+		mlx5e_rss_destroy_tirs(rss, true);
+
+	mlx5e_rqt_destroy(&rss->rqt);
+
+	return 0;
+}
+
+void mlx5e_rss_refcnt_inc(struct mlx5e_rss *rss)
+{
+	refcount_inc(&rss->refcnt);
+}
+
+void mlx5e_rss_refcnt_dec(struct mlx5e_rss *rss)
+{
+	refcount_dec(&rss->refcnt);
+}
+
+unsigned int mlx5e_rss_refcnt_read(struct mlx5e_rss *rss)
+{
+	return refcount_read(&rss->refcnt);
+}
+
+u32 mlx5e_rss_get_tirn(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+		       bool inner)
+{
+	struct mlx5e_tir *tir;
+
+	WARN_ON(inner && !rss->inner_ft_support);
+	tir = rss_get_tir(rss, tt, inner);
+	WARN_ON(!tir);
+
+	return mlx5e_tir_get_tirn(tir);
+}
+
+/* Fill the "tirn" output parameter.
+ * Create the requested TIR if it's its first usage.
+ */
+int mlx5e_rss_obtain_tirn(struct mlx5e_rss *rss,
+			  enum mlx5_traffic_types tt,
+			  const struct mlx5e_lro_param *init_lro_param,
+			  bool inner, u32 *tirn)
+{
+	struct mlx5e_tir *tir;
+
+	tir = rss_get_tir(rss, tt, inner);
+	if (!tir) { /* TIR doesn't exist, create one */
+		int err;
+
+		err = mlx5e_rss_create_tir(rss, tt, init_lro_param, inner);
+		if (err)
+			return err;
+		tir = rss_get_tir(rss, tt, inner);
+	}
+
+	*tirn = mlx5e_tir_get_tirn(tir);
+	return 0;
+}
+
+static void mlx5e_rss_apply(struct mlx5e_rss *rss, u32 *rqns, unsigned int num_rqns)
+{
+	int err;
+
+	err = mlx5e_rqt_redirect_indir(&rss->rqt, rqns, num_rqns, rss->hash.hfunc, &rss->indir);
+	if (err)
+		mlx5e_rss_warn(rss->mdev, "Failed to redirect RQT %#x to channels: err = %d\n",
+			       mlx5e_rqt_get_rqtn(&rss->rqt), err);
+}
+
+void mlx5e_rss_enable(struct mlx5e_rss *rss, u32 *rqns, unsigned int num_rqns)
+{
+	rss->enabled = true;
+	mlx5e_rss_apply(rss, rqns, num_rqns);
+}
+
+void mlx5e_rss_disable(struct mlx5e_rss *rss)
+{
+	int err;
+
+	rss->enabled = false;
+	err = mlx5e_rqt_redirect_direct(&rss->rqt, rss->drop_rqn);
+	if (err)
+		mlx5e_rss_warn(rss->mdev, "Failed to redirect RQT %#x to drop RQ %#x: err = %d\n",
+			       mlx5e_rqt_get_rqtn(&rss->rqt), rss->drop_rqn, err);
+}
+
+int mlx5e_rss_lro_set_param(struct mlx5e_rss *rss, struct mlx5e_lro_param *lro_param)
+{
+	struct mlx5e_tir_builder *builder;
+	enum mlx5_traffic_types tt;
+	int err, final_err;
+
+	builder = mlx5e_tir_builder_alloc(true);
+	if (!builder)
+		return -ENOMEM;
+
+	mlx5e_tir_builder_build_lro(builder, lro_param);
+
+	final_err = 0;
+
+	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++) {
+		struct mlx5e_tir *tir;
+
+		tir = rss_get_tir(rss, tt, false);
+		if (!tir)
+			goto inner_tir;
+		err = mlx5e_tir_modify(tir, builder);
+		if (err) {
+			mlx5e_rss_warn(rss->mdev, "Failed to update LRO state of indirect TIR %#x for traffic type %d: err = %d\n",
+				       mlx5e_tir_get_tirn(tir), tt, err);
+			if (!final_err)
+				final_err = err;
+		}
+
+inner_tir:
+		if (!rss->inner_ft_support)
+			continue;
+
+		tir = rss_get_tir(rss, tt, true);
+		if (!tir)
+			continue;
+		err = mlx5e_tir_modify(tir, builder);
+		if (err) {
+			mlx5e_rss_warn(rss->mdev, "Failed to update LRO state of inner indirect TIR %#x for traffic type %d: err = %d\n",
+				       mlx5e_tir_get_tirn(tir), tt, err);
+			if (!final_err)
+				final_err = err;
+		}
+	}
+
+	mlx5e_tir_builder_free(builder);
+	return final_err;
+}
+
+int mlx5e_rss_get_rxfh(struct mlx5e_rss *rss, u32 *indir, u8 *key, u8 *hfunc)
+{
+	unsigned int i;
+
+	if (indir)
+		for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++)
+			indir[i] = rss->indir.table[i];
+
+	if (key)
+		memcpy(key, rss->hash.toeplitz_hash_key,
+		       sizeof(rss->hash.toeplitz_hash_key));
+
+	if (hfunc)
+		*hfunc = rss->hash.hfunc;
+
+	return 0;
+}
+
+int mlx5e_rss_set_rxfh(struct mlx5e_rss *rss, const u32 *indir,
+		       const u8 *key, const u8 *hfunc,
+		       u32 *rqns, unsigned int num_rqns)
+{
+	bool changed_indir = false;
+	bool changed_hash = false;
+
+	if (hfunc && *hfunc != rss->hash.hfunc) {
+		switch (*hfunc) {
+		case ETH_RSS_HASH_XOR:
+		case ETH_RSS_HASH_TOP:
+			break;
+		default:
+			return -EINVAL;
+		}
+		changed_hash = true;
+		changed_indir = true;
+		rss->hash.hfunc = *hfunc;
+	}
+
+	if (key) {
+		if (rss->hash.hfunc == ETH_RSS_HASH_TOP)
+			changed_hash = true;
+		memcpy(rss->hash.toeplitz_hash_key, key,
+		       sizeof(rss->hash.toeplitz_hash_key));
+	}
+
+	if (indir) {
+		unsigned int i;
+
+		changed_indir = true;
+
+		for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++)
+			rss->indir.table[i] = indir[i];
+	}
+
+	if (changed_indir && rss->enabled)
+		mlx5e_rss_apply(rss, rqns, num_rqns);
+
+	if (changed_hash)
+		mlx5e_rss_update_tirs(rss);
+
+	return 0;
+}
+
+struct mlx5e_rss_params_hash mlx5e_rss_get_hash(struct mlx5e_rss *rss)
+{
+	return rss->hash;
+}
+
+u8 mlx5e_rss_get_hash_fields(struct mlx5e_rss *rss, enum mlx5_traffic_types tt)
+{
+	return rss->rx_hash_fields[tt];
+}
+
+int mlx5e_rss_set_hash_fields(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+			      u8 rx_hash_fields)
+{
+	u8 old_rx_hash_fields;
+	int err;
+
+	old_rx_hash_fields = rss->rx_hash_fields[tt];
+
+	if (old_rx_hash_fields == rx_hash_fields)
+		return 0;
+
+	rss->rx_hash_fields[tt] = rx_hash_fields;
+
+	err = mlx5e_rss_update_tir(rss, tt, false);
+	if (err) {
+		rss->rx_hash_fields[tt] = old_rx_hash_fields;
+		mlx5e_rss_warn(rss->mdev,
+			       "Failed to update RSS hash fields of indirect TIR for traffic type %d: err = %d\n",
+			       tt, err);
+		return err;
+	}
+
+	if (!(rss->inner_ft_support))
+		return 0;
+
+	err = mlx5e_rss_update_tir(rss, tt, true);
+	if (err) {
+		/* Partial update happened. Try to revert - it may fail too, but
+		 * there is nothing more we can do.
+		 */
+		rss->rx_hash_fields[tt] = old_rx_hash_fields;
+		mlx5e_rss_warn(rss->mdev,
+			       "Failed to update RSS hash fields of inner indirect TIR for traffic type %d: err = %d\n",
+			       tt, err);
+		if (mlx5e_rss_update_tir(rss, tt, false))
+			mlx5e_rss_warn(rss->mdev,
+				       "Partial update of RSS hash fields happened: failed to revert indirect TIR for traffic type %d to the old values\n",
+				       tt);
+	}
+
+	return err;
+}
+
+void mlx5e_rss_set_indir_uniform(struct mlx5e_rss *rss, unsigned int nch)
+{
+	mlx5e_rss_params_indir_init_uniform(&rss->indir, nch);
+}

drivers/net/ethernet/mellanox/mlx5/core/en/rss.h

+/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
+/* Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES. */
+
+#ifndef __MLX5_EN_RSS_H__
+#define __MLX5_EN_RSS_H__
+
+#include "rqt.h"
+#include "tir.h"
+#include "fs.h"
+
+struct mlx5e_rss_params_traffic_type
+mlx5e_rss_get_default_tt_config(enum mlx5_traffic_types tt);
+
+struct mlx5e_rss;
+
+struct mlx5e_rss *mlx5e_rss_alloc(void);
+void mlx5e_rss_free(struct mlx5e_rss *rss);
+int mlx5e_rss_init(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
+		   bool inner_ft_support, u32 drop_rqn,
+		   const struct mlx5e_lro_param *init_lro_param);
+int mlx5e_rss_init_no_tirs(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
+			   bool inner_ft_support, u32 drop_rqn);
+int mlx5e_rss_cleanup(struct mlx5e_rss *rss);
+
+void mlx5e_rss_refcnt_inc(struct mlx5e_rss *rss);
+void mlx5e_rss_refcnt_dec(struct mlx5e_rss *rss);
+unsigned int mlx5e_rss_refcnt_read(struct mlx5e_rss *rss);
+
+u32 mlx5e_rss_get_tirn(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+		       bool inner);
+int mlx5e_rss_obtain_tirn(struct mlx5e_rss *rss,
+			  enum mlx5_traffic_types tt,
+			  const struct mlx5e_lro_param *init_lro_param,
+			  bool inner, u32 *tirn);
+
+void mlx5e_rss_enable(struct mlx5e_rss *rss, u32 *rqns, unsigned int num_rqns);
+void mlx5e_rss_disable(struct mlx5e_rss *rss);
+
+int mlx5e_rss_lro_set_param(struct mlx5e_rss *rss, struct mlx5e_lro_param *lro_param);
+int mlx5e_rss_get_rxfh(struct mlx5e_rss *rss, u32 *indir, u8 *key, u8 *hfunc);
+int mlx5e_rss_set_rxfh(struct mlx5e_rss *rss, const u32 *indir,
+		       const u8 *key, const u8 *hfunc,
+		       u32 *rqns, unsigned int num_rqns);
+struct mlx5e_rss_params_hash mlx5e_rss_get_hash(struct mlx5e_rss *rss);
+u8 mlx5e_rss_get_hash_fields(struct mlx5e_rss *rss, enum mlx5_traffic_types tt);
+int mlx5e_rss_set_hash_fields(struct mlx5e_rss *rss, enum mlx5_traffic_types tt,
+			      u8 rx_hash_fields);
+void mlx5e_rss_set_indir_uniform(struct mlx5e_rss *rss, unsigned int nch);
+#endif /* __MLX5_EN_RSS_H__ */

drivers/net/ethernet/mellanox/mlx5/core/en/rx_res.c

@@ -5,64 +5,7 @@
 #include "channels.h"
 #include "params.h"
 
-static const struct mlx5e_rss_params_traffic_type rss_default_config[MLX5E_NUM_INDIR_TIRS] = {
-	[MLX5_TT_IPV4_TCP] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
-		.l4_prot_type = MLX5_L4_PROT_TYPE_TCP,
-		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
-	},
-	[MLX5_TT_IPV6_TCP] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
-		.l4_prot_type = MLX5_L4_PROT_TYPE_TCP,
-		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
-	},
-	[MLX5_TT_IPV4_UDP] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
-		.l4_prot_type = MLX5_L4_PROT_TYPE_UDP,
-		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
-	},
-	[MLX5_TT_IPV6_UDP] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
-		.l4_prot_type = MLX5_L4_PROT_TYPE_UDP,
-		.rx_hash_fields = MLX5_HASH_IP_L4PORTS,
-	},
-	[MLX5_TT_IPV4_IPSEC_AH] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
-		.l4_prot_type = 0,
-		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
-	},
-	[MLX5_TT_IPV6_IPSEC_AH] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
-		.l4_prot_type = 0,
-		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
-	},
-	[MLX5_TT_IPV4_IPSEC_ESP] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
-		.l4_prot_type = 0,
-		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
-	},
-	[MLX5_TT_IPV6_IPSEC_ESP] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
-		.l4_prot_type = 0,
-		.rx_hash_fields = MLX5_HASH_IP_IPSEC_SPI,
-	},
-	[MLX5_TT_IPV4] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV4,
-		.l4_prot_type = 0,
-		.rx_hash_fields = MLX5_HASH_IP,
-	},
-	[MLX5_TT_IPV6] = {
-		.l3_prot_type = MLX5_L3_PROT_TYPE_IPV6,
-		.l4_prot_type = 0,
-		.rx_hash_fields = MLX5_HASH_IP,
-	},
-};
-
-struct mlx5e_rss_params_traffic_type
-mlx5e_rss_get_default_tt_config(enum mlx5_traffic_types tt)
-{
-	return rss_default_config[tt];
-}
+#define MLX5E_MAX_NUM_RSS 16
 
 struct mlx5e_rx_res {
 	struct mlx5_core_dev *mdev;
@@ -70,18 +13,7 @@ struct mlx5e_rx_res {
 	unsigned int max_nch;
 	u32 drop_rqn;
 
-	struct {
-		struct mlx5e_rss_params_hash hash;
-		struct mlx5e_rss_params_indir indir;
-		u32 rx_hash_fields[MLX5E_NUM_INDIR_TIRS];
-	} rss_params;
-
-	struct mlx5e_rqt indir_rqt;
-	struct {
-		struct mlx5e_tir indir_tir;
-		struct mlx5e_tir inner_indir_tir;
-	} rss[MLX5E_NUM_INDIR_TIRS];
-
+	struct mlx5e_rss *rss[MLX5E_MAX_NUM_RSS];
 	bool rss_active;
 	u32 rss_rqns[MLX5E_INDIR_RQT_SIZE];
 	unsigned int rss_nch;
@@ -99,103 +31,247 @@ struct mlx5e_rx_res {
 	} ptp;
 };
 
-struct mlx5e_rx_res *mlx5e_rx_res_alloc(void)
-{
-	return kvzalloc(sizeof(struct mlx5e_rx_res), GFP_KERNEL);
-}
+/* API for rx_res_rss_* */
 
-static void mlx5e_rx_res_rss_params_init(struct mlx5e_rx_res *res, unsigned int init_nch)
+static int mlx5e_rx_res_rss_init_def(struct mlx5e_rx_res *res,
+				     const struct mlx5e_lro_param *init_lro_param,
+				     unsigned int init_nch)
 {
-	enum mlx5_traffic_types tt;
+	bool inner_ft_support = res->features & MLX5E_RX_RES_FEATURE_INNER_FT;
+	struct mlx5e_rss *rss;
+	int err;
+
+	if (WARN_ON(res->rss[0]))
+		return -EINVAL;
 
-	res->rss_params.hash.hfunc = ETH_RSS_HASH_TOP;
-	netdev_rss_key_fill(res->rss_params.hash.toeplitz_hash_key,
-			    sizeof(res->rss_params.hash.toeplitz_hash_key));
-	mlx5e_rss_params_indir_init_uniform(&res->rss_params.indir, init_nch);
-	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++)
-		res->rss_params.rx_hash_fields[tt] =
-			mlx5e_rss_get_default_tt_config(tt).rx_hash_fields;
+	rss = mlx5e_rss_alloc();
+	if (!rss)
+		return -ENOMEM;
+
+	err = mlx5e_rss_init(rss, res->mdev, inner_ft_support, res->drop_rqn,
+			     init_lro_param);
+	if (err)
+		goto err_rss_free;
+
+	mlx5e_rss_set_indir_uniform(rss, init_nch);
+
+	res->rss[0] = rss;
+
+	return 0;
+
+err_rss_free:
+	mlx5e_rss_free(rss);
+	return err;
 }
 
-static int mlx5e_rx_res_rss_init(struct mlx5e_rx_res *res,
-				 const struct mlx5e_lro_param *init_lro_param)
+int mlx5e_rx_res_rss_init(struct mlx5e_rx_res *res, u32 *rss_idx, unsigned int init_nch)
 {
 	bool inner_ft_support = res->features & MLX5E_RX_RES_FEATURE_INNER_FT;
-	enum mlx5_traffic_types tt, max_tt;
-	struct mlx5e_tir_builder *builder;
-	u32 indir_rqtn;
-	int err;
+	struct mlx5e_rss *rss;
+	int err, i;
 
-	builder = mlx5e_tir_builder_alloc(false);
-	if (!builder)
+	for (i = 1; i < MLX5E_MAX_NUM_RSS; i++)
+		if (!res->rss[i])
+			break;
+
+	if (i == MLX5E_MAX_NUM_RSS)
+		return -ENOSPC;
+
+	rss = mlx5e_rss_alloc();
+	if (!rss)
 		return -ENOMEM;
 
-	err = mlx5e_rqt_init_direct(&res->indir_rqt, res->mdev, true, res->drop_rqn);
+	err = mlx5e_rss_init_no_tirs(rss, res->mdev, inner_ft_support, res->drop_rqn);
 	if (err)
-		goto out;
+		goto err_rss_free;
 
-	indir_rqtn = mlx5e_rqt_get_rqtn(&res->indir_rqt);
+	mlx5e_rss_set_indir_uniform(rss, init_nch);
+	if (res->rss_active)
+		mlx5e_rss_enable(rss, res->rss_rqns, res->rss_nch);
 
-	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++) {
-		struct mlx5e_rss_params_traffic_type rss_tt;
+	res->rss[i] = rss;
+	*rss_idx = i;
 
-		mlx5e_tir_builder_build_rqt(builder, res->mdev->mlx5e_res.hw_objs.td.tdn,
-					    indir_rqtn, inner_ft_support);
-		mlx5e_tir_builder_build_lro(builder, init_lro_param);
-		rss_tt = mlx5e_rx_res_rss_get_current_tt_config(res, tt);
-		mlx5e_tir_builder_build_rss(builder, &res->rss_params.hash, &rss_tt, false);
+	return 0;
 
-		err = mlx5e_tir_init(&res->rss[tt].indir_tir, builder, res->mdev, true);
-		if (err) {
-			mlx5_core_warn(res->mdev, "Failed to create an indirect TIR: err = %d, tt = %d\n",
-				       err, tt);
-			goto err_destroy_tirs;
-		}
+err_rss_free:
+	mlx5e_rss_free(rss);
+	return err;
+}
 
-		mlx5e_tir_builder_clear(builder);
-	}
+static int __mlx5e_rx_res_rss_destroy(struct mlx5e_rx_res *res, u32 rss_idx)
+{
+	struct mlx5e_rss *rss = res->rss[rss_idx];
+	int err;
 
-	if (!inner_ft_support)
-		goto out;
+	err = mlx5e_rss_cleanup(rss);
+	if (err)
+		return err;
 
-	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++) {
-		struct mlx5e_rss_params_traffic_type rss_tt;
+	mlx5e_rss_free(rss);
+	res->rss[rss_idx] = NULL;
 
-		mlx5e_tir_builder_build_rqt(builder, res->mdev->mlx5e_res.hw_objs.td.tdn,
-					    indir_rqtn, inner_ft_support);
-		mlx5e_tir_builder_build_lro(builder, init_lro_param);
-		rss_tt = mlx5e_rx_res_rss_get_current_tt_config(res, tt);
-		mlx5e_tir_builder_build_rss(builder, &res->rss_params.hash, &rss_tt, true);
+	return 0;
+}
+
+int mlx5e_rx_res_rss_destroy(struct mlx5e_rx_res *res, u32 rss_idx)
+{
+	struct mlx5e_rss *rss;
+
+	if (rss_idx >= MLX5E_MAX_NUM_RSS)
+		return -EINVAL;
+
+	rss = res->rss[rss_idx];
+	if (!rss)
+		return -EINVAL;
+
+	return __mlx5e_rx_res_rss_destroy(res, rss_idx);
+}
+
+static void mlx5e_rx_res_rss_destroy_all(struct mlx5e_rx_res *res)
+{
+	int i;
+
+	for (i = 0; i < MLX5E_MAX_NUM_RSS; i++) {
+		struct mlx5e_rss *rss = res->rss[i];
+		int err;
 
-		err = mlx5e_tir_init(&res->rss[tt].inner_indir_tir, builder, res->mdev, true);
+		if (!rss)
+			continue;
+
+		err = __mlx5e_rx_res_rss_destroy(res, i);
 		if (err) {
-			mlx5_core_warn(res->mdev, "Failed to create an inner indirect TIR: err = %d, tt = %d\n",
-				       err, tt);
-			goto err_destroy_inner_tirs;
+			unsigned int refcount;
+
+			refcount = mlx5e_rss_refcnt_read(rss);
+			mlx5_core_warn(res->mdev,
+				       "Failed to destroy RSS context %d, refcount = %u, err = %d\n",
+				       i, refcount, err);
 		}
+	}
+}
 
-		mlx5e_tir_builder_clear(builder);
+static void mlx5e_rx_res_rss_enable(struct mlx5e_rx_res *res)
+{
+	int i;
+
+	res->rss_active = true;
+
+	for (i = 0; i < MLX5E_MAX_NUM_RSS; i++) {
+		struct mlx5e_rss *rss = res->rss[i];
+
+		if (!rss)
+			continue;
+		mlx5e_rss_enable(rss, res->rss_rqns, res->rss_nch);
 	}
+}
 
-	goto out;
+static void mlx5e_rx_res_rss_disable(struct mlx5e_rx_res *res)
+{
+	int i;
 
-err_destroy_inner_tirs:
-	max_tt = tt;
-	for (tt = 0; tt < max_tt; tt++)
-		mlx5e_tir_destroy(&res->rss[tt].inner_indir_tir);
+	res->rss_active = false;
 
-	tt = MLX5E_NUM_INDIR_TIRS;
-err_destroy_tirs:
-	max_tt = tt;
-	for (tt = 0; tt < max_tt; tt++)
-		mlx5e_tir_destroy(&res->rss[tt].indir_tir);
+	for (i = 0; i < MLX5E_MAX_NUM_RSS; i++) {
+		struct mlx5e_rss *rss = res->rss[i];
 
-	mlx5e_rqt_destroy(&res->indir_rqt);
+		if (!rss)
+			continue;
+		mlx5e_rss_disable(rss);
+	}
+}
 
-out:
-	mlx5e_tir_builder_free(builder);
+/* Updates the indirection table SW shadow, does not update the HW resources yet */
+void mlx5e_rx_res_rss_set_indir_uniform(struct mlx5e_rx_res *res, unsigned int nch)
+{
+	WARN_ON_ONCE(res->rss_active);
+	mlx5e_rss_set_indir_uniform(res->rss[0], nch);
+}
 
-	return err;
+int mlx5e_rx_res_rss_get_rxfh(struct mlx5e_rx_res *res, u32 rss_idx,
+			      u32 *indir, u8 *key, u8 *hfunc)
+{
+	struct mlx5e_rss *rss;
+
+	if (rss_idx >= MLX5E_MAX_NUM_RSS)
+		return -EINVAL;
+
+	rss = res->rss[rss_idx];
+	if (!rss)
+		return -ENOENT;
+
+	return mlx5e_rss_get_rxfh(rss, indir, key, hfunc);
+}
+
+int mlx5e_rx_res_rss_set_rxfh(struct mlx5e_rx_res *res, u32 rss_idx,
+			      const u32 *indir, const u8 *key, const u8 *hfunc)
+{
+	struct mlx5e_rss *rss;
+
+	if (rss_idx >= MLX5E_MAX_NUM_RSS)
+		return -EINVAL;
+
+	rss = res->rss[rss_idx];
+	if (!rss)
+		return -ENOENT;
+
+	return mlx5e_rss_set_rxfh(rss, indir, key, hfunc, res->rss_rqns, res->rss_nch);
+}
+
+u8 mlx5e_rx_res_rss_get_hash_fields(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt)
+{
+	struct mlx5e_rss *rss = res->rss[0];
+
+	return mlx5e_rss_get_hash_fields(rss, tt);
+}
+
+int mlx5e_rx_res_rss_set_hash_fields(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt,
+				     u8 rx_hash_fields)
+{
+	struct mlx5e_rss *rss = res->rss[0];
+
+	return mlx5e_rss_set_hash_fields(rss, tt, rx_hash_fields);
+}
+
+int mlx5e_rx_res_rss_cnt(struct mlx5e_rx_res *res)
+{
+	int i, cnt;
+
+	cnt = 0;
+	for (i = 0; i < MLX5E_MAX_NUM_RSS; i++)
+		if (res->rss[i])
+			cnt++;
+
+	return cnt;
+}
+
+int mlx5e_rx_res_rss_index(struct mlx5e_rx_res *res, struct mlx5e_rss *rss)
+{
+	int i;
+
+	if (!rss)
+		return -EINVAL;
+
+	for (i = 0; i < MLX5E_MAX_NUM_RSS; i++)
+		if (rss == res->rss[i])
+			return i;
+
+	return -ENOENT;
+}
+
+struct mlx5e_rss *mlx5e_rx_res_rss_get(struct mlx5e_rx_res *res, u32 rss_idx)
+{
+	if (rss_idx >= MLX5E_MAX_NUM_RSS)
+		return NULL;
+
+	return res->rss[rss_idx];
+}
+
+/* End of API rx_res_rss_* */
+
+struct mlx5e_rx_res *mlx5e_rx_res_alloc(void)
+{
+	return kvzalloc(sizeof(struct mlx5e_rx_res), GFP_KERNEL);
 }
 
 static int mlx5e_rx_res_channels_init(struct mlx5e_rx_res *res,
@@ -335,20 +411,6 @@ static int mlx5e_rx_res_ptp_init(struct mlx5e_rx_res *res)
 	return err;
 }
 
-static void mlx5e_rx_res_rss_destroy(struct mlx5e_rx_res *res)
-{
-	enum mlx5_traffic_types tt;
-
-	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++)
-		mlx5e_tir_destroy(&res->rss[tt].indir_tir);
-
-	if (res->features & MLX5E_RX_RES_FEATURE_INNER_FT)
-		for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++)
-			mlx5e_tir_destroy(&res->rss[tt].inner_indir_tir);
-
-	mlx5e_rqt_destroy(&res->indir_rqt);
-}
-
 static void mlx5e_rx_res_channels_destroy(struct mlx5e_rx_res *res)
 {
 	unsigned int ix;
@@ -385,11 +447,9 @@ int mlx5e_rx_res_init(struct mlx5e_rx_res *res, struct mlx5_core_dev *mdev,
 	res->max_nch = max_nch;
 	res->drop_rqn = drop_rqn;
 
-	mlx5e_rx_res_rss_params_init(res, init_nch);
-
-	err = mlx5e_rx_res_rss_init(res, init_lro_param);
+	err = mlx5e_rx_res_rss_init_def(res, init_lro_param, init_nch);
 	if (err)
-		return err;
+		goto err_out;
 
 	err = mlx5e_rx_res_channels_init(res, init_lro_param);
 	if (err)
@@ -404,7 +464,8 @@ int mlx5e_rx_res_init(struct mlx5e_rx_res *res, struct mlx5_core_dev *mdev,
 err_channels_destroy:
 	mlx5e_rx_res_channels_destroy(res);
 err_rss_destroy:
-	mlx5e_rx_res_rss_destroy(res);
+	__mlx5e_rx_res_rss_destroy(res, 0);
+err_out:
 	return err;
 }
 
@@ -412,7 +473,7 @@ void mlx5e_rx_res_destroy(struct mlx5e_rx_res *res)
 {
 	mlx5e_rx_res_ptp_destroy(res);
 	mlx5e_rx_res_channels_destroy(res);
-	mlx5e_rx_res_rss_destroy(res);
+	mlx5e_rx_res_rss_destroy_all(res);
 }
 
 void mlx5e_rx_res_free(struct mlx5e_rx_res *res)
@@ -434,13 +495,16 @@ u32 mlx5e_rx_res_get_tirn_xsk(struct mlx5e_rx_res *res, unsigned int ix)
 
 u32 mlx5e_rx_res_get_tirn_rss(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt)
 {
-	return mlx5e_tir_get_tirn(&res->rss[tt].indir_tir);
+	struct mlx5e_rss *rss = res->rss[0];
+
+	return mlx5e_rss_get_tirn(rss, tt, false);
 }
 
 u32 mlx5e_rx_res_get_tirn_rss_inner(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt)
 {
-	WARN_ON(!(res->features & MLX5E_RX_RES_FEATURE_INNER_FT));
-	return mlx5e_tir_get_tirn(&res->rss[tt].inner_indir_tir);
+	struct mlx5e_rss *rss = res->rss[0];
+
+	return mlx5e_rss_get_tirn(rss, tt, true);
 }
 
 u32 mlx5e_rx_res_get_tirn_ptp(struct mlx5e_rx_res *res)
@@ -454,32 +518,6 @@ u32 mlx5e_rx_res_get_rqtn_direct(struct mlx5e_rx_res *res, unsigned int ix)
 	return mlx5e_rqt_get_rqtn(&res->channels[ix].direct_rqt);
 }
 
-static void mlx5e_rx_res_rss_enable(struct mlx5e_rx_res *res)
-{
-	int err;
-
-	res->rss_active = true;
-
-	err = mlx5e_rqt_redirect_indir(&res->indir_rqt, res->rss_rqns, res->rss_nch,
-				       res->rss_params.hash.hfunc,
-				       &res->rss_params.indir);
-	if (err)
-		mlx5_core_warn(res->mdev, "Failed to redirect indirect RQT %#x to channels: err = %d\n",
-			       mlx5e_rqt_get_rqtn(&res->indir_rqt), err);
-}
-
-static void mlx5e_rx_res_rss_disable(struct mlx5e_rx_res *res)
-{
-	int err;
-
-	res->rss_active = false;
-
-	err = mlx5e_rqt_redirect_direct(&res->indir_rqt, res->drop_rqn);
-	if (err)
-		mlx5_core_warn(res->mdev, "Failed to redirect indirect RQT %#x to drop RQ %#x: err = %d\n",
-			       mlx5e_rqt_get_rqtn(&res->indir_rqt), res->drop_rqn, err);
-}
-
 void mlx5e_rx_res_channels_activate(struct mlx5e_rx_res *res, struct mlx5e_channels *chs)
 {
 	unsigned int nch, ix;
@@ -607,178 +645,9 @@ int mlx5e_rx_res_xsk_deactivate(struct mlx5e_rx_res *res, unsigned int ix)
 	return err;
 }
 
-struct mlx5e_rss_params_traffic_type
-mlx5e_rx_res_rss_get_current_tt_config(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt)
-{
-	struct mlx5e_rss_params_traffic_type rss_tt;
-
-	rss_tt = mlx5e_rss_get_default_tt_config(tt);
-	rss_tt.rx_hash_fields = res->rss_params.rx_hash_fields[tt];
-	return rss_tt;
-}
-
-void mlx5e_rx_res_rss_set_indir_uniform(struct mlx5e_rx_res *res, unsigned int nch)
-{
-	mlx5e_rss_params_indir_init_uniform(&res->rss_params.indir, nch);
-
-	if (!res->rss_active)
-		return;
-
-	mlx5e_rx_res_rss_enable(res);
-}
-
-void mlx5e_rx_res_rss_get_rxfh(struct mlx5e_rx_res *res, u32 *indir, u8 *key, u8 *hfunc)
-{
-	unsigned int i;
-
-	if (indir)
-		for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++)
-			indir[i] = res->rss_params.indir.table[i];
-
-	if (key)
-		memcpy(key, res->rss_params.hash.toeplitz_hash_key,
-		       sizeof(res->rss_params.hash.toeplitz_hash_key));
-
-	if (hfunc)
-		*hfunc = res->rss_params.hash.hfunc;
-}
-
-static int mlx5e_rx_res_rss_update_tir(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt,
-				       bool inner)
-{
-	struct mlx5e_rss_params_traffic_type rss_tt;
-	struct mlx5e_tir_builder *builder;
-	struct mlx5e_tir *tir;
-	int err;
-
-	builder = mlx5e_tir_builder_alloc(true);
-	if (!builder)
-		return -ENOMEM;
-
-	rss_tt = mlx5e_rx_res_rss_get_current_tt_config(res, tt);
-
-	mlx5e_tir_builder_build_rss(builder, &res->rss_params.hash, &rss_tt, inner);
-	tir = inner ? &res->rss[tt].inner_indir_tir : &res->rss[tt].indir_tir;
-	err = mlx5e_tir_modify(tir, builder);
-
-	mlx5e_tir_builder_free(builder);
-	return err;
-}
-
-int mlx5e_rx_res_rss_set_rxfh(struct mlx5e_rx_res *res, const u32 *indir,
-			      const u8 *key, const u8 *hfunc)
-{
-	enum mlx5_traffic_types tt;
-	bool changed_indir = false;
-	bool changed_hash = false;
-	int err;
-
-	if (hfunc && *hfunc != res->rss_params.hash.hfunc) {
-		switch (*hfunc) {
-		case ETH_RSS_HASH_XOR:
-		case ETH_RSS_HASH_TOP:
-			break;
-		default:
-			return -EINVAL;
-		}
-		changed_hash = true;
-		changed_indir = true;
-		res->rss_params.hash.hfunc = *hfunc;
-	}
-
-	if (key) {
-		if (res->rss_params.hash.hfunc == ETH_RSS_HASH_TOP)
-			changed_hash = true;
-		memcpy(res->rss_params.hash.toeplitz_hash_key, key,
-		       sizeof(res->rss_params.hash.toeplitz_hash_key));
-	}
-
-	if (indir) {
-		unsigned int i;
-
-		changed_indir = true;
-
-		for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++)
-			res->rss_params.indir.table[i] = indir[i];
-	}
-
-	if (changed_indir && res->rss_active) {
-		err = mlx5e_rqt_redirect_indir(&res->indir_rqt, res->rss_rqns, res->rss_nch,
-					       res->rss_params.hash.hfunc,
-					       &res->rss_params.indir);
-		if (err)
-			mlx5_core_warn(res->mdev, "Failed to redirect indirect RQT %#x to channels: err = %d\n",
-				       mlx5e_rqt_get_rqtn(&res->indir_rqt), err);
-	}
-
-	if (changed_hash)
-		for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++) {
-			err = mlx5e_rx_res_rss_update_tir(res, tt, false);
-			if (err)
-				mlx5_core_warn(res->mdev, "Failed to update RSS hash of indirect TIR for traffic type %d: err = %d\n",
-					       tt, err);
-
-			if (!(res->features & MLX5E_RX_RES_FEATURE_INNER_FT))
-				continue;
-
-			err = mlx5e_rx_res_rss_update_tir(res, tt, true);
-			if (err)
-				mlx5_core_warn(res->mdev, "Failed to update RSS hash of inner indirect TIR for traffic type %d: err = %d\n",
-					       tt, err);
-		}
-
-	return 0;
-}
-
-u8 mlx5e_rx_res_rss_get_hash_fields(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt)
-{
-	return res->rss_params.rx_hash_fields[tt];
-}
-
-int mlx5e_rx_res_rss_set_hash_fields(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt,
-				     u8 rx_hash_fields)
-{
-	u8 old_rx_hash_fields;
-	int err;
-
-	old_rx_hash_fields = res->rss_params.rx_hash_fields[tt];
-
-	if (old_rx_hash_fields == rx_hash_fields)
-		return 0;
-
-	res->rss_params.rx_hash_fields[tt] = rx_hash_fields;
-
-	err = mlx5e_rx_res_rss_update_tir(res, tt, false);
-	if (err) {
-		res->rss_params.rx_hash_fields[tt] = old_rx_hash_fields;
-		mlx5_core_warn(res->mdev, "Failed to update RSS hash fields of indirect TIR for traffic type %d: err = %d\n",
-			       tt, err);
-		return err;
-	}
-
-	if (!(res->features & MLX5E_RX_RES_FEATURE_INNER_FT))
-		return 0;
-
-	err = mlx5e_rx_res_rss_update_tir(res, tt, true);
-	if (err) {
-		/* Partial update happened. Try to revert - it may fail too, but
-		 * there is nothing more we can do.
-		 */
-		res->rss_params.rx_hash_fields[tt] = old_rx_hash_fields;
-		mlx5_core_warn(res->mdev, "Failed to update RSS hash fields of inner indirect TIR for traffic type %d: err = %d\n",
-			       tt, err);
-		if (mlx5e_rx_res_rss_update_tir(res, tt, false))
-			mlx5_core_warn(res->mdev, "Partial update of RSS hash fields happened: failed to revert indirect TIR for traffic type %d to the old values\n",
-				       tt);
-	}
-
-	return err;
-}
-
 int mlx5e_rx_res_lro_set_param(struct mlx5e_rx_res *res, struct mlx5e_lro_param *lro_param)
 {
 	struct mlx5e_tir_builder *builder;
-	enum mlx5_traffic_types tt;
 	int err, final_err;
 	unsigned int ix;
 
@@ -790,25 +659,15 @@ int mlx5e_rx_res_lro_set_param(struct mlx5e_rx_res *res, struct mlx5e_lro_param
 
 	final_err = 0;
 
-	for (tt = 0; tt < MLX5E_NUM_INDIR_TIRS; tt++) {
-		err = mlx5e_tir_modify(&res->rss[tt].indir_tir, builder);
-		if (err) {
-			mlx5_core_warn(res->mdev, "Failed to update LRO state of indirect TIR %#x for traffic type %d: err = %d\n",
-				       mlx5e_tir_get_tirn(&res->rss[tt].indir_tir), tt, err);
-			if (!final_err)
-				final_err = err;
-		}
+	for (ix = 0; ix < MLX5E_MAX_NUM_RSS; ix++) {
+		struct mlx5e_rss *rss = res->rss[ix];
 
-		if (!(res->features & MLX5E_RX_RES_FEATURE_INNER_FT))
+		if (!rss)
 			continue;
 
-		err = mlx5e_tir_modify(&res->rss[tt].inner_indir_tir, builder);
-		if (err) {
-			mlx5_core_warn(res->mdev, "Failed to update LRO state of inner indirect TIR %#x for traffic type %d: err = %d\n",
-				       mlx5e_tir_get_tirn(&res->rss[tt].inner_indir_tir), tt, err);
-			if (!final_err)
-				final_err = err;
-		}
+		err = mlx5e_rss_lro_set_param(rss, lro_param);
+		if (err)
+			final_err = final_err ? : err;
 	}
 
 	for (ix = 0; ix < res->max_nch; ix++) {
@@ -827,5 +686,5 @@ int mlx5e_rx_res_lro_set_param(struct mlx5e_rx_res *res, struct mlx5e_lro_param
 
 struct mlx5e_rss_params_hash mlx5e_rx_res_get_current_hash(struct mlx5e_rx_res *res)
 {
-	return res->rss_params.hash;
+	return mlx5e_rss_get_hash(res->rss[0]);
 }

drivers/net/ethernet/mellanox/mlx5/core/en/rx_res.h

@@ -8,6 +8,7 @@
 #include "rqt.h"
 #include "tir.h"
 #include "fs.h"
+#include "rss.h"
 
 struct mlx5e_rx_res;
 
@@ -20,9 +21,6 @@ enum mlx5e_rx_res_features {
 	MLX5E_RX_RES_FEATURE_PTP = BIT(2),
 };
 
-struct mlx5e_rss_params_traffic_type
-mlx5e_rss_get_default_tt_config(enum mlx5_traffic_types tt);
-
 /* Setup */
 struct mlx5e_rx_res *mlx5e_rx_res_alloc(void);
 int mlx5e_rx_res_init(struct mlx5e_rx_res *res, struct mlx5_core_dev *mdev,
@@ -50,17 +48,23 @@ int mlx5e_rx_res_xsk_activate(struct mlx5e_rx_res *res, struct mlx5e_channels *c
 int mlx5e_rx_res_xsk_deactivate(struct mlx5e_rx_res *res, unsigned int ix);
 
 /* Configuration API */
-struct mlx5e_rss_params_traffic_type
-mlx5e_rx_res_rss_get_current_tt_config(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt);
 void mlx5e_rx_res_rss_set_indir_uniform(struct mlx5e_rx_res *res, unsigned int nch);
-void mlx5e_rx_res_rss_get_rxfh(struct mlx5e_rx_res *res, u32 *indir, u8 *key, u8 *hfunc);
-int mlx5e_rx_res_rss_set_rxfh(struct mlx5e_rx_res *res, const u32 *indir,
-			      const u8 *key, const u8 *hfunc);
+int mlx5e_rx_res_rss_get_rxfh(struct mlx5e_rx_res *res, u32 rss_idx,
+			      u32 *indir, u8 *key, u8 *hfunc);
+int mlx5e_rx_res_rss_set_rxfh(struct mlx5e_rx_res *res, u32 rss_idx,
+			      const u32 *indir, const u8 *key, const u8 *hfunc);
+
 u8 mlx5e_rx_res_rss_get_hash_fields(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt);
 int mlx5e_rx_res_rss_set_hash_fields(struct mlx5e_rx_res *res, enum mlx5_traffic_types tt,
 				     u8 rx_hash_fields);
 int mlx5e_rx_res_lro_set_param(struct mlx5e_rx_res *res, struct mlx5e_lro_param *lro_param);
 
+int mlx5e_rx_res_rss_init(struct mlx5e_rx_res *res, u32 *rss_idx, unsigned int init_nch);
+int mlx5e_rx_res_rss_destroy(struct mlx5e_rx_res *res, u32 rss_idx);
+int mlx5e_rx_res_rss_cnt(struct mlx5e_rx_res *res);
+int mlx5e_rx_res_rss_index(struct mlx5e_rx_res *res, struct mlx5e_rss *rss);
+struct mlx5e_rss *mlx5e_rx_res_rss_get(struct mlx5e_rx_res *res, u32 rss_idx);
+
 /* Workaround for hairpin */
 struct mlx5e_rss_params_hash mlx5e_rx_res_get_current_hash(struct mlx5e_rx_res *res);
 

drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c

@@ -420,6 +420,7 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 	unsigned int count = ch->combined_count;
 	struct mlx5e_params new_params;
 	bool arfs_enabled;
+	int rss_cnt;
 	bool opened;
 	int err = 0;
 
@@ -455,6 +456,27 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 		goto out;
 	}
 
+	/* Don't allow changing the number of channels if non-default RSS contexts exist,
+	 * the kernel doesn't protect against set_channels operations that break them.
+	 */
+	rss_cnt = mlx5e_rx_res_rss_cnt(priv->rx_res) - 1;
+	if (rss_cnt) {
+		err = -EINVAL;
+		netdev_err(priv->netdev, "%s: Non-default RSS contexts exist (%d), cannot change the number of channels\n",
+			   __func__, rss_cnt);
+		goto out;
+	}
+
+	/* Don't allow changing the number of channels if MQPRIO mode channel offload is active,
+	 * because it defines a partition over the channels queues.
+	 */
+	if (cur_params->mqprio.mode == TC_MQPRIO_MODE_CHANNEL) {
+		err = -EINVAL;
+		netdev_err(priv->netdev, "%s: MQPRIO mode channel offload is active, cannot change the number of channels\n",
+			   __func__);
+		goto out;
+	}
+
 	new_params = *cur_params;
 	new_params.num_channels = count;
 
@@ -1194,16 +1216,51 @@ static u32 mlx5e_get_rxfh_indir_size(struct net_device *netdev)
 	return mlx5e_ethtool_get_rxfh_indir_size(priv);
 }
 
-int mlx5e_get_rxfh(struct net_device *netdev, u32 *indir, u8 *key,
-		   u8 *hfunc)
+static int mlx5e_get_rxfh_context(struct net_device *dev, u32 *indir,
+				  u8 *key, u8 *hfunc, u32 rss_context)
 {
-	struct mlx5e_priv *priv = netdev_priv(netdev);
+	struct mlx5e_priv *priv = netdev_priv(dev);
+	int err;
 
 	mutex_lock(&priv->state_lock);
-	mlx5e_rx_res_rss_get_rxfh(priv->rx_res, indir, key, hfunc);
+	err = mlx5e_rx_res_rss_get_rxfh(priv->rx_res, rss_context, indir, key, hfunc);
 	mutex_unlock(&priv->state_lock);
+	return err;
+}
 
-	return 0;
+static int mlx5e_set_rxfh_context(struct net_device *dev, const u32 *indir,
+				  const u8 *key, const u8 hfunc,
+				  u32 *rss_context, bool delete)
+{
+	struct mlx5e_priv *priv = netdev_priv(dev);
+	int err;
+
+	mutex_lock(&priv->state_lock);
+	if (delete) {
+		err = mlx5e_rx_res_rss_destroy(priv->rx_res, *rss_context);
+		goto unlock;
+	}
+
+	if (*rss_context == ETH_RXFH_CONTEXT_ALLOC) {
+		unsigned int count = priv->channels.params.num_channels;
+
+		err = mlx5e_rx_res_rss_init(priv->rx_res, rss_context, count);
+		if (err)
+			goto unlock;
+	}
+
+	err = mlx5e_rx_res_rss_set_rxfh(priv->rx_res, *rss_context, indir, key,
+					hfunc == ETH_RSS_HASH_NO_CHANGE ? NULL : &hfunc);
+
+unlock:
+	mutex_unlock(&priv->state_lock);
+	return err;
+}
+
+int mlx5e_get_rxfh(struct net_device *netdev, u32 *indir, u8 *key,
+		   u8 *hfunc)
+{
+	return mlx5e_get_rxfh_context(netdev, indir, key, hfunc, 0);
 }
 
 int mlx5e_set_rxfh(struct net_device *dev, const u32 *indir,
@@ -1213,7 +1270,7 @@ int mlx5e_set_rxfh(struct net_device *dev, const u32 *indir,
 	int err;
 
 	mutex_lock(&priv->state_lock);
-	err = mlx5e_rx_res_rss_set_rxfh(priv->rx_res, indir, key,
+	err = mlx5e_rx_res_rss_set_rxfh(priv->rx_res, 0, indir, key,
 					hfunc == ETH_RSS_HASH_NO_CHANGE ? NULL : &hfunc);
 	mutex_unlock(&priv->state_lock);
 	return err;
@@ -2299,6 +2356,8 @@ const struct ethtool_ops mlx5e_ethtool_ops = {
 	.get_rxfh_indir_size = mlx5e_get_rxfh_indir_size,
 	.get_rxfh          = mlx5e_get_rxfh,
 	.set_rxfh          = mlx5e_set_rxfh,
+	.get_rxfh_context  = mlx5e_get_rxfh_context,
+	.set_rxfh_context  = mlx5e_set_rxfh_context,
 	.get_rxnfc         = mlx5e_get_rxnfc,
 	.set_rxnfc         = mlx5e_set_rxnfc,
 	.get_tunable       = mlx5e_get_tunable,

drivers/net/ethernet/mellanox/mlx5/core/en_fs_ethtool.c

@@ -35,11 +35,19 @@
 #include "en/params.h"
 #include "en/xsk/pool.h"
 
+static int flow_type_to_traffic_type(u32 flow_type);
+
+static u32 flow_type_mask(u32 flow_type)
+{
+	return flow_type & ~(FLOW_EXT | FLOW_MAC_EXT | FLOW_RSS);
+}
+
 struct mlx5e_ethtool_rule {
 	struct list_head             list;
 	struct ethtool_rx_flow_spec  flow_spec;
 	struct mlx5_flow_handle	     *rule;
 	struct mlx5e_ethtool_table   *eth_ft;
+	struct mlx5e_rss             *rss;
 };
 
 static void put_flow_table(struct mlx5e_ethtool_table *eth_ft)
@@ -66,7 +74,7 @@ static struct mlx5e_ethtool_table *get_flow_table(struct mlx5e_priv *priv,
 	int table_size;
 	int prio;
 
-	switch (fs->flow_type & ~(FLOW_EXT | FLOW_MAC_EXT)) {
+	switch (flow_type_mask(fs->flow_type)) {
 	case TCP_V4_FLOW:
 	case UDP_V4_FLOW:
 	case TCP_V6_FLOW:
@@ -329,7 +337,7 @@ static int set_flow_attrs(u32 *match_c, u32 *match_v,
 					     outer_headers);
 	void *outer_headers_v = MLX5_ADDR_OF(fte_match_param, match_v,
 					     outer_headers);
-	u32 flow_type = fs->flow_type & ~(FLOW_EXT | FLOW_MAC_EXT);
+	u32 flow_type = flow_type_mask(fs->flow_type);
 
 	switch (flow_type) {
 	case TCP_V4_FLOW:
@@ -397,10 +405,53 @@ static bool outer_header_zero(u32 *match_criteria)
 						  size - 1);
 }
 
+static int flow_get_tirn(struct mlx5e_priv *priv,
+			 struct mlx5e_ethtool_rule *eth_rule,
+			 struct ethtool_rx_flow_spec *fs,
+			 u32 rss_context, u32 *tirn)
+{
+	if (fs->flow_type & FLOW_RSS) {
+		struct mlx5e_lro_param lro_param;
+		struct mlx5e_rss *rss;
+		u32 flow_type;
+		int err;
+		int tt;
+
+		rss = mlx5e_rx_res_rss_get(priv->rx_res, rss_context);
+		if (!rss)
+			return -ENOENT;
+
+		flow_type = flow_type_mask(fs->flow_type);
+		tt = flow_type_to_traffic_type(flow_type);
+		if (tt < 0)
+			return -EINVAL;
+
+		lro_param = mlx5e_get_lro_param(&priv->channels.params);
+		err = mlx5e_rss_obtain_tirn(rss, tt, &lro_param, false, tirn);
+		if (err)
+			return err;
+		eth_rule->rss = rss;
+		mlx5e_rss_refcnt_inc(eth_rule->rss);
+	} else {
+		struct mlx5e_params *params = &priv->channels.params;
+		enum mlx5e_rq_group group;
+		u16 ix;
+
+		mlx5e_qid_get_ch_and_group(params, fs->ring_cookie, &ix, &group);
+
+		*tirn = group == MLX5E_RQ_GROUP_XSK ?
+			mlx5e_rx_res_get_tirn_xsk(priv->rx_res, ix) :
+			mlx5e_rx_res_get_tirn_direct(priv->rx_res, ix);
+	}
+
+	return 0;
+}
+
 static struct mlx5_flow_handle *
 add_ethtool_flow_rule(struct mlx5e_priv *priv,
+		      struct mlx5e_ethtool_rule *eth_rule,
 		      struct mlx5_flow_table *ft,
-		      struct ethtool_rx_flow_spec *fs)
+		      struct ethtool_rx_flow_spec *fs, u32 rss_context)
 {
 	struct mlx5_flow_act flow_act = { .flags = FLOW_ACT_NO_APPEND };
 	struct mlx5_flow_destination *dst = NULL;
@@ -419,23 +470,17 @@ add_ethtool_flow_rule(struct mlx5e_priv *priv,
 	if (fs->ring_cookie == RX_CLS_FLOW_DISC) {
 		flow_act.action = MLX5_FLOW_CONTEXT_ACTION_DROP;
 	} else {
-		struct mlx5e_params *params = &priv->channels.params;
-		enum mlx5e_rq_group group;
-		u16 ix;
-
-		mlx5e_qid_get_ch_and_group(params, fs->ring_cookie, &ix, &group);
-
 		dst = kzalloc(sizeof(*dst), GFP_KERNEL);
 		if (!dst) {
 			err = -ENOMEM;
 			goto free;
 		}
 
+		err = flow_get_tirn(priv, eth_rule, fs, rss_context, &dst->tir_num);
+		if (err)
+			goto free;
+
 		dst->type = MLX5_FLOW_DESTINATION_TYPE_TIR;
-		if (group == MLX5E_RQ_GROUP_XSK)
-			dst->tir_num = mlx5e_rx_res_get_tirn_xsk(priv->rx_res, ix);
-		else
-			dst->tir_num = mlx5e_rx_res_get_tirn_direct(priv->rx_res, ix);
 		flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
 	}
 
@@ -459,6 +504,8 @@ static void del_ethtool_rule(struct mlx5e_priv *priv,
 {
 	if (eth_rule->rule)
 		mlx5_del_flow_rules(eth_rule->rule);
+	if (eth_rule->rss)
+		mlx5e_rss_refcnt_dec(eth_rule->rss);
 	list_del(&eth_rule->list);
 	priv->fs.ethtool.tot_num_rules--;
 	put_flow_table(eth_rule->eth_ft);
@@ -619,7 +666,7 @@ static int validate_flow(struct mlx5e_priv *priv,
 					fs->ring_cookie))
 			return -EINVAL;
 
-	switch (fs->flow_type & ~(FLOW_EXT | FLOW_MAC_EXT)) {
+	switch (flow_type_mask(fs->flow_type)) {
 	case ETHER_FLOW:
 		num_tuples += validate_ethter(fs);
 		break;
@@ -668,7 +715,7 @@ static int validate_flow(struct mlx5e_priv *priv,
 
 static int
 mlx5e_ethtool_flow_replace(struct mlx5e_priv *priv,
-			   struct ethtool_rx_flow_spec *fs)
+			   struct ethtool_rx_flow_spec *fs, u32 rss_context)
 {
 	struct mlx5e_ethtool_table *eth_ft;
 	struct mlx5e_ethtool_rule *eth_rule;
@@ -699,7 +746,7 @@ mlx5e_ethtool_flow_replace(struct mlx5e_priv *priv,
 		err = -EINVAL;
 		goto del_ethtool_rule;
 	}
-	rule = add_ethtool_flow_rule(priv, eth_ft->ft, fs);
+	rule = add_ethtool_flow_rule(priv, eth_rule, eth_ft->ft, fs, rss_context);
 	if (IS_ERR(rule)) {
 		err = PTR_ERR(rule);
 		goto del_ethtool_rule;
@@ -745,10 +792,20 @@ mlx5e_ethtool_get_flow(struct mlx5e_priv *priv,
 		return -EINVAL;
 
 	list_for_each_entry(eth_rule, &priv->fs.ethtool.rules, list) {
-		if (eth_rule->flow_spec.location == location) {
-			info->fs = eth_rule->flow_spec;
+		int index;
+
+		if (eth_rule->flow_spec.location != location)
+			continue;
+		if (!info)
 			return 0;
-		}
+		info->fs = eth_rule->flow_spec;
+		if (!eth_rule->rss)
+			return 0;
+		index = mlx5e_rx_res_rss_index(priv->rx_res, eth_rule->rss);
+		if (index < 0)
+			return index;
+		info->rss_context = index;
+		return 0;
 	}
 
 	return -ENOENT;
@@ -764,7 +821,7 @@ mlx5e_ethtool_get_all_flows(struct mlx5e_priv *priv,
 
 	info->data = MAX_NUM_OF_ETHTOOL_RULES;
 	while ((!err || err == -ENOENT) && idx < info->rule_cnt) {
-		err = mlx5e_ethtool_get_flow(priv, info, location);
+		err = mlx5e_ethtool_get_flow(priv, NULL, location);
 		if (!err)
 			rule_locs[idx++] = location;
 		location++;
@@ -887,7 +944,7 @@ int mlx5e_ethtool_set_rxnfc(struct net_device *dev, struct ethtool_rxnfc *cmd)
 
 	switch (cmd->cmd) {
 	case ETHTOOL_SRXCLSRLINS:
-		err = mlx5e_ethtool_flow_replace(priv, &cmd->fs);
+		err = mlx5e_ethtool_flow_replace(priv, &cmd->fs, cmd->rss_context);
 		break;
 	case ETHTOOL_SRXCLSRLDEL:
 		err = mlx5e_ethtool_flow_remove(priv, cmd->fs.location);

drivers/net/ethernet/mellanox/mlx5/core/en_main.c

@@ -1711,7 +1711,7 @@ static int mlx5e_open_sqs(struct mlx5e_channel *c,
 {
 	int err, tc;
 
-	for (tc = 0; tc < params->num_tc; tc++) {
+	for (tc = 0; tc < mlx5e_get_dcb_num_tc(params); tc++) {
 		int txq_ix = c->ix + tc * params->num_channels;
 
 		err = mlx5e_open_txqsq(c, c->priv->tisn[c->lag_port][tc], txq_ix,
@@ -1992,7 +1992,7 @@ static int mlx5e_open_channel(struct mlx5e_priv *priv, int ix,
 	c->pdev     = mlx5_core_dma_dev(priv->mdev);
 	c->netdev   = priv->netdev;
 	c->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey.key);
-	c->num_tc   = params->num_tc;
+	c->num_tc   = mlx5e_get_dcb_num_tc(params);
 	c->xdp      = !!params->xdp_prog;
 	c->stats    = &priv->channel_stats[ix].ch;
 	c->aff_mask = irq_get_effective_affinity_mask(irq);
@@ -2263,22 +2263,34 @@ void mlx5e_set_netdev_mtu_boundaries(struct mlx5e_priv *priv)
 				ETH_MAX_MTU);
 }
 
-static void mlx5e_netdev_set_tcs(struct net_device *netdev, u16 nch, u8 ntc)
+static int mlx5e_netdev_set_tcs(struct net_device *netdev, u16 nch, u8 ntc,
+				struct tc_mqprio_qopt_offload *mqprio)
 {
-	int tc;
+	int tc, err;
 
 	netdev_reset_tc(netdev);
 
 	if (ntc == 1)
-		return;
+		return 0;
 
-	netdev_set_num_tc(netdev, ntc);
+	err = netdev_set_num_tc(netdev, ntc);
+	if (err) {
+		netdev_WARN(netdev, "netdev_set_num_tc failed (%d), ntc = %d\n", err, ntc);
+		return err;
+	}
 
-	/* Map netdev TCs to offset 0
-	 * We have our own UP to TXQ mapping for QoS
-	 */
-	for (tc = 0; tc < ntc; tc++)
-		netdev_set_tc_queue(netdev, tc, nch, 0);
+	for (tc = 0; tc < ntc; tc++) {
+		u16 count, offset;
+
+		/* For DCB mode, map netdev TCs to offset 0
+		 * We have our own UP to TXQ mapping for QoS
+		 */
+		count = mqprio ? mqprio->qopt.count[tc] : nch;
+		offset = mqprio ? mqprio->qopt.offset[tc] : 0;
+		netdev_set_tc_queue(netdev, tc, count, offset);
+	}
+
+	return 0;
 }
 
 int mlx5e_update_tx_netdev_queues(struct mlx5e_priv *priv)
@@ -2288,7 +2300,7 @@ int mlx5e_update_tx_netdev_queues(struct mlx5e_priv *priv)
 	qos_queues = mlx5e_qos_cur_leaf_nodes(priv);
 
 	nch = priv->channels.params.num_channels;
-	ntc = priv->channels.params.num_tc;
+	ntc = mlx5e_get_dcb_num_tc(&priv->channels.params);
 	num_txqs = nch * ntc + qos_queues;
 	if (MLX5E_GET_PFLAG(&priv->channels.params, MLX5E_PFLAG_TX_PORT_TS))
 		num_txqs += ntc;
@@ -2312,11 +2324,12 @@ static int mlx5e_update_netdev_queues(struct mlx5e_priv *priv)
 	old_ntc = netdev->num_tc ? : 1;
 
 	nch = priv->channels.params.num_channels;
-	ntc = priv->channels.params.num_tc;
+	ntc = mlx5e_get_dcb_num_tc(&priv->channels.params);
 	num_rxqs = nch * priv->profile->rq_groups;
 
-	mlx5e_netdev_set_tcs(netdev, nch, ntc);
-
+	err = mlx5e_netdev_set_tcs(netdev, nch, ntc, NULL);
+	if (err)
+		goto err_out;
 	err = mlx5e_update_tx_netdev_queues(priv);
 	if (err)
 		goto err_tcs;
@@ -2337,7 +2350,8 @@ static int mlx5e_update_netdev_queues(struct mlx5e_priv *priv)
 	WARN_ON_ONCE(netif_set_real_num_tx_queues(netdev, old_num_txqs));
 
 err_tcs:
-	mlx5e_netdev_set_tcs(netdev, old_num_txqs / old_ntc, old_ntc);
+	mlx5e_netdev_set_tcs(netdev, old_num_txqs / old_ntc, old_ntc, NULL);
+err_out:
 	return err;
 }
 
@@ -2387,7 +2401,7 @@ static void mlx5e_build_txq_maps(struct mlx5e_priv *priv)
 	int i, ch, tc, num_tc;
 
 	ch = priv->channels.num;
-	num_tc = priv->channels.params.num_tc;
+	num_tc = mlx5e_get_dcb_num_tc(&priv->channels.params);
 
 	for (i = 0; i < ch; i++) {
 		for (tc = 0; tc < num_tc; tc++) {
@@ -2418,7 +2432,7 @@ static void mlx5e_update_num_tc_x_num_ch(struct mlx5e_priv *priv)
 {
 	/* Sync with mlx5e_select_queue. */
 	WRITE_ONCE(priv->num_tc_x_num_ch,
-		   priv->channels.params.num_tc * priv->channels.num);
+		   mlx5e_get_dcb_num_tc(&priv->channels.params) * priv->channels.num);
 }
 
 void mlx5e_activate_priv_channels(struct mlx5e_priv *priv)
@@ -2847,41 +2861,124 @@ static int mlx5e_modify_channels_vsd(struct mlx5e_channels *chs, bool vsd)
 	return 0;
 }
 
-static int mlx5e_setup_tc_mqprio(struct mlx5e_priv *priv,
-				 struct tc_mqprio_qopt *mqprio)
+static int mlx5e_setup_tc_mqprio_dcb(struct mlx5e_priv *priv,
+				     struct tc_mqprio_qopt *mqprio)
 {
 	struct mlx5e_params new_params;
 	u8 tc = mqprio->num_tc;
-	int err = 0;
+	int err;
 
 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
 
 	if (tc && tc != MLX5E_MAX_NUM_TC)
 		return -EINVAL;
 
-	mutex_lock(&priv->state_lock);
-
-	/* MQPRIO is another toplevel qdisc that can't be attached
-	 * simultaneously with the offloaded HTB.
-	 */
-	if (WARN_ON(priv->htb.maj_id)) {
-		err = -EINVAL;
-		goto out;
-	}
-
 	new_params = priv->channels.params;
-	new_params.num_tc = tc ? tc : 1;
+	new_params.mqprio.mode = TC_MQPRIO_MODE_DCB;
+	new_params.mqprio.num_tc = tc ? tc : 1;
 
 	err = mlx5e_safe_switch_params(priv, &new_params,
 				       mlx5e_num_channels_changed_ctx, NULL, true);
 
-out:
 	priv->max_opened_tc = max_t(u8, priv->max_opened_tc,
-				    priv->channels.params.num_tc);
-	mutex_unlock(&priv->state_lock);
+				    mlx5e_get_dcb_num_tc(&priv->channels.params));
+	return err;
+}
+
+static int mlx5e_mqprio_channel_validate(struct mlx5e_priv *priv,
+					 struct tc_mqprio_qopt_offload *mqprio)
+{
+	struct net_device *netdev = priv->netdev;
+	int agg_count = 0;
+	int i;
+
+	if (mqprio->qopt.offset[0] != 0 || mqprio->qopt.num_tc < 1 ||
+	    mqprio->qopt.num_tc > MLX5E_MAX_NUM_MQPRIO_CH_TC)
+		return -EINVAL;
+
+	for (i = 0; i < mqprio->qopt.num_tc; i++) {
+		if (!mqprio->qopt.count[i]) {
+			netdev_err(netdev, "Zero size for queue-group (%d) is not supported\n", i);
+			return -EINVAL;
+		}
+		if (mqprio->min_rate[i]) {
+			netdev_err(netdev, "Min tx rate is not supported\n");
+			return -EINVAL;
+		}
+		if (mqprio->max_rate[i]) {
+			netdev_err(netdev, "Max tx rate is not supported\n");
+			return -EINVAL;
+		}
+
+		if (mqprio->qopt.offset[i] != agg_count) {
+			netdev_err(netdev, "Discontinuous queues config is not supported\n");
+			return -EINVAL;
+		}
+		agg_count += mqprio->qopt.count[i];
+	}
+
+	if (priv->channels.params.num_channels < agg_count) {
+		netdev_err(netdev, "Num of queues (%d) exceeds available (%d)\n",
+			   agg_count, priv->channels.params.num_channels);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int mlx5e_mqprio_channel_set_tcs_ctx(struct mlx5e_priv *priv, void *ctx)
+{
+	struct tc_mqprio_qopt_offload *mqprio = (struct tc_mqprio_qopt_offload *)ctx;
+	struct net_device *netdev = priv->netdev;
+	u8 num_tc;
+
+	if (priv->channels.params.mqprio.mode != TC_MQPRIO_MODE_CHANNEL)
+		return -EINVAL;
+
+	num_tc = priv->channels.params.mqprio.num_tc;
+	mlx5e_netdev_set_tcs(netdev, 0, num_tc, mqprio);
+
+	return 0;
+}
+
+static int mlx5e_setup_tc_mqprio_channel(struct mlx5e_priv *priv,
+					 struct tc_mqprio_qopt_offload *mqprio)
+{
+	struct mlx5e_params new_params;
+	int err;
+
+	err = mlx5e_mqprio_channel_validate(priv, mqprio);
+	if (err)
+		return err;
+
+	new_params = priv->channels.params;
+	new_params.mqprio.mode = TC_MQPRIO_MODE_CHANNEL;
+	new_params.mqprio.num_tc = mqprio->qopt.num_tc;
+	err = mlx5e_safe_switch_params(priv, &new_params,
+				       mlx5e_mqprio_channel_set_tcs_ctx, mqprio, true);
+
 	return err;
 }
 
+static int mlx5e_setup_tc_mqprio(struct mlx5e_priv *priv,
+				 struct tc_mqprio_qopt_offload *mqprio)
+{
+	/* MQPRIO is another toplevel qdisc that can't be attached
+	 * simultaneously with the offloaded HTB.
+	 */
+	if (WARN_ON(priv->htb.maj_id))
+		return -EINVAL;
+
+	switch (mqprio->mode) {
+	case TC_MQPRIO_MODE_DCB:
+		return mlx5e_setup_tc_mqprio_dcb(priv, &mqprio->qopt);
+	case TC_MQPRIO_MODE_CHANNEL:
+		return mlx5e_setup_tc_mqprio_channel(priv, mqprio);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static int mlx5e_setup_tc_htb(struct mlx5e_priv *priv, struct tc_htb_qopt_offload *htb)
 {
 	int res;
@@ -2951,7 +3048,10 @@ static int mlx5e_setup_tc(struct net_device *dev, enum tc_setup_type type,
 						  priv, priv, true);
 	}
 	case TC_SETUP_QDISC_MQPRIO:
-		return mlx5e_setup_tc_mqprio(priv, type_data);
+		mutex_lock(&priv->state_lock);
+		err = mlx5e_setup_tc_mqprio(priv, type_data);
+		mutex_unlock(&priv->state_lock);
+		return err;
 	case TC_SETUP_QDISC_HTB:
 		mutex_lock(&priv->state_lock);
 		err = mlx5e_setup_tc_htb(priv, type_data);
@@ -4093,12 +4193,12 @@ void mlx5e_build_nic_params(struct mlx5e_priv *priv, struct mlx5e_xsk *xsk, u16
 	params->hard_mtu = MLX5E_ETH_HARD_MTU;
 	params->num_channels = min_t(unsigned int, MLX5E_MAX_NUM_CHANNELS / 2,
 				     priv->max_nch);
-	params->num_tc       = 1;
+	params->mqprio.num_tc = 1;
 
 	/* Set an initial non-zero value, so that mlx5e_select_queue won't
 	 * divide by zero if called before first activating channels.
 	 */
-	priv->num_tc_x_num_ch = params->num_channels * params->num_tc;
+	priv->num_tc_x_num_ch = params->num_channels * params->mqprio.num_tc;
 
 	/* SQ */
 	params->log_sq_size = is_kdump_kernel() ?

drivers/net/ethernet/mellanox/mlx5/core/en_rep.c

@@ -394,7 +394,8 @@ int mlx5e_add_sqs_fwd_rules(struct mlx5e_priv *priv)
 	int err = -ENOMEM;
 	u32 *sqs;
 
-	sqs = kcalloc(priv->channels.num * priv->channels.params.num_tc, sizeof(*sqs), GFP_KERNEL);
+	sqs = kcalloc(priv->channels.num * mlx5e_get_dcb_num_tc(&priv->channels.params),
+		      sizeof(*sqs), GFP_KERNEL);
 	if (!sqs)
 		goto out;
 
@@ -611,7 +612,7 @@ static void mlx5e_build_rep_params(struct net_device *netdev)
 	params->rx_dim_enabled = MLX5_CAP_GEN(mdev, cq_moderation);
 	mlx5e_set_rx_cq_mode_params(params, cq_period_mode);
 
-	params->num_tc                = 1;
+	params->mqprio.num_tc       = 1;
 	params->tunneled_offload_en = false;
 
 	mlx5_query_min_inline(mdev, &params->tx_min_inline_mode);

drivers/net/ethernet/mellanox/mlx5/core/esw/bridge.c

@@ -5,6 +5,7 @@
 #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"
@@ -56,7 +57,6 @@ struct mlx5_esw_bridge {
 
 	struct list_head fdb_list;
 	struct rhashtable fdb_ht;
-	struct xarray vports;
 
 	struct mlx5_flow_table *egress_ft;
 	struct mlx5_flow_group *egress_vlan_fg;
@@ -77,6 +77,15 @@ mlx5_esw_bridge_fdb_offload_notify(struct net_device *dev, const unsigned char *
 	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 struct mlx5_flow_table *
 mlx5_esw_bridge_table_create(int max_fte, u32 level, struct mlx5_eswitch *esw)
 {
@@ -400,9 +409,10 @@ mlx5_esw_bridge_egress_table_cleanup(struct mlx5_esw_bridge *bridge)
 }
 
 static struct mlx5_flow_handle *
-mlx5_esw_bridge_ingress_flow_create(u16 vport_num, const unsigned char *addr,
-				    struct mlx5_esw_bridge_vlan *vlan, u32 counter_id,
-				    struct mlx5_esw_bridge *bridge)
+mlx5_esw_bridge_ingress_flow_with_esw_create(u16 vport_num, const unsigned char *addr,
+					     struct mlx5_esw_bridge_vlan *vlan, u32 counter_id,
+					     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 = {
@@ -430,7 +440,7 @@ mlx5_esw_bridge_ingress_flow_create(u16 vport_num, const unsigned char *addr,
 	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));
+		 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;
@@ -458,6 +468,35 @@ mlx5_esw_bridge_ingress_flow_create(u16 vport_num, const unsigned char *addr,
 	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, u32 counter_id,
+				    struct mlx5_esw_bridge *bridge)
+{
+	return mlx5_esw_bridge_ingress_flow_with_esw_create(vport_num, addr, vlan, counter_id,
+							    bridge, bridge->br_offloads->esw);
+}
+
+static struct mlx5_flow_handle *
+mlx5_esw_bridge_ingress_flow_peer_create(u16 vport_num, const unsigned char *addr,
+					 struct mlx5_esw_bridge_vlan *vlan, u32 counter_id,
+					 struct mlx5_esw_bridge *bridge)
+{
+	struct mlx5_devcom *devcom = bridge->br_offloads->esw->dev->priv.devcom;
+	static struct mlx5_flow_handle *handle;
+	struct mlx5_eswitch *peer_esw;
+
+	peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+	if (!peer_esw)
+		return ERR_PTR(-ENODEV);
+
+	handle = mlx5_esw_bridge_ingress_flow_with_esw_create(vport_num, addr, vlan, counter_id,
+							      bridge, peer_esw);
+
+	mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+	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)
@@ -505,7 +544,7 @@ mlx5_esw_bridge_ingress_filter_flow_create(u16 vport_num, const unsigned char *a
 }
 
 static struct mlx5_flow_handle *
-mlx5_esw_bridge_egress_flow_create(u16 vport_num, const unsigned char *addr,
+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)
 {
@@ -550,6 +589,10 @@ mlx5_esw_bridge_egress_flow_create(u16 vport_num, const unsigned char *addr,
 			 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);
@@ -576,7 +619,6 @@ static struct mlx5_esw_bridge *mlx5_esw_bridge_create(int ifindex,
 		goto err_fdb_ht;
 
 	INIT_LIST_HEAD(&bridge->fdb_list);
-	xa_init(&bridge->vports);
 	bridge->ifindex = ifindex;
 	bridge->refcnt = 1;
 	bridge->ageing_time = clock_t_to_jiffies(BR_DEFAULT_AGEING_TIME);
@@ -603,7 +645,6 @@ static void mlx5_esw_bridge_put(struct mlx5_esw_bridge_offloads *br_offloads,
 		return;
 
 	mlx5_esw_bridge_egress_table_cleanup(bridge);
-	WARN_ON(!xa_empty(&bridge->vports));
 	list_del(&bridge->list);
 	rhashtable_destroy(&bridge->fdb_ht);
 	kvfree(bridge);
@@ -639,30 +680,40 @@ mlx5_esw_bridge_lookup(int ifindex, struct mlx5_esw_bridge_offloads *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;
+}
+
+static 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 *bridge)
+				       struct mlx5_esw_bridge_offloads *br_offloads)
 {
-	return xa_insert(&bridge->vports, port->vport_num, port, GFP_KERNEL);
+	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, struct mlx5_esw_bridge *bridge)
+mlx5_esw_bridge_port_lookup(u16 vport_num, u16 esw_owner_vhca_id,
+			    struct mlx5_esw_bridge_offloads *br_offloads)
 {
-	return xa_load(&bridge->vports, vport_num);
+	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 *bridge)
+				       struct mlx5_esw_bridge_offloads *br_offloads)
 {
-	xa_erase(&bridge->vports, port->vport_num);
+	xa_erase(&br_offloads->ports, mlx5_esw_bridge_port_key(port));
 }
 
-static void mlx5_esw_bridge_fdb_entry_refresh(unsigned long lastuse,
-					      struct mlx5_esw_bridge_fdb_entry *entry)
+static void mlx5_esw_bridge_fdb_entry_refresh(struct mlx5_esw_bridge_fdb_entry *entry)
 {
 	trace_mlx5_esw_bridge_fdb_entry_refresh(entry);
 
-	entry->lastuse = lastuse;
 	mlx5_esw_bridge_fdb_offload_notify(entry->dev, entry->key.addr,
 					   entry->key.vid,
 					   SWITCHDEV_FDB_ADD_TO_BRIDGE);
@@ -690,10 +741,7 @@ 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) {
-		if (!(entry->flags & MLX5_ESW_BRIDGE_FLAG_ADDED_BY_USER))
-			mlx5_esw_bridge_fdb_offload_notify(entry->dev, entry->key.addr,
-							   entry->key.vid,
-							   SWITCHDEV_FDB_DEL_TO_BRIDGE);
+		mlx5_esw_bridge_fdb_del_notify(entry);
 		mlx5_esw_bridge_fdb_entry_cleanup(entry, bridge);
 	}
 }
@@ -841,10 +889,7 @@ static void mlx5_esw_bridge_vlan_flush(struct mlx5_esw_bridge_vlan *vlan,
 	struct mlx5_esw_bridge_fdb_entry *entry, *tmp;
 
 	list_for_each_entry_safe(entry, tmp, &vlan->fdb_list, vlan_list) {
-		if (!(entry->flags & MLX5_ESW_BRIDGE_FLAG_ADDED_BY_USER))
-			mlx5_esw_bridge_fdb_offload_notify(entry->dev, entry->key.addr,
-							   entry->key.vid,
-							   SWITCHDEV_FDB_DEL_TO_BRIDGE);
+		mlx5_esw_bridge_fdb_del_notify(entry);
 		mlx5_esw_bridge_fdb_entry_cleanup(entry, bridge);
 	}
 
@@ -875,13 +920,13 @@ static void mlx5_esw_bridge_port_vlans_flush(struct mlx5_esw_bridge_port *port,
 }
 
 static struct mlx5_esw_bridge_vlan *
-mlx5_esw_bridge_port_vlan_lookup(u16 vid, u16 vport_num, struct mlx5_esw_bridge *bridge,
-				 struct mlx5_eswitch *esw)
+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, bridge);
+	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.
@@ -904,24 +949,23 @@ mlx5_esw_bridge_port_vlan_lookup(u16 vid, u16 vport_num, struct mlx5_esw_bridge
 }
 
 static struct mlx5_esw_bridge_fdb_entry *
-mlx5_esw_bridge_fdb_entry_init(struct net_device *dev, u16 vport_num, const unsigned char *addr,
-			       u16 vid, bool added_by_user, struct mlx5_eswitch *esw,
-			       struct mlx5_esw_bridge *bridge)
+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;
-	struct mlx5e_priv *priv;
 	int err;
 
 	if (bridge->flags & MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG && vid) {
-		vlan = mlx5_esw_bridge_port_vlan_lookup(vid, vport_num, bridge, esw);
+		vlan = mlx5_esw_bridge_port_vlan_lookup(vid, vport_num, esw_owner_vhca_id, bridge,
+							esw);
 		if (IS_ERR(vlan))
 			return ERR_CAST(vlan);
 	}
 
-	priv = netdev_priv(dev);
 	entry = kvzalloc(sizeof(*entry), GFP_KERNEL);
 	if (!entry)
 		return ERR_PTR(-ENOMEM);
@@ -930,19 +974,25 @@ mlx5_esw_bridge_fdb_entry_init(struct net_device *dev, u16 vport_num, const unsi
 	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(priv->mdev, true);
+	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 = mlx5_esw_bridge_ingress_flow_create(vport_num, addr, vlan, mlx5_fc_id(counter),
-						     bridge);
+	handle = peer ?
+		mlx5_esw_bridge_ingress_flow_peer_create(vport_num, addr, vlan,
+							 mlx5_fc_id(counter), bridge) :
+		mlx5_esw_bridge_ingress_flow_create(vport_num, addr, vlan,
+						    mlx5_fc_id(counter), bridge);
 	if (IS_ERR(handle)) {
 		err = PTR_ERR(handle);
 		esw_warn(esw->dev, "Failed to create ingress flow(vport=%u,err=%d)\n",
@@ -962,7 +1012,8 @@ mlx5_esw_bridge_fdb_entry_init(struct net_device *dev, u16 vport_num, const unsi
 		entry->filter_handle = handle;
 	}
 
-	handle = mlx5_esw_bridge_egress_flow_create(vport_num, addr, vlan, bridge);
+	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",
@@ -994,32 +1045,37 @@ mlx5_esw_bridge_fdb_entry_init(struct net_device *dev, u16 vport_num, const unsi
 err_ingress_filter_flow_create:
 	mlx5_del_flow_rules(entry->ingress_handle);
 err_ingress_flow_create:
-	mlx5_fc_destroy(priv->mdev, entry->ingress_counter);
+	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(unsigned long ageing_time, struct mlx5_eswitch *esw,
-				    struct mlx5_vport *vport)
+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)
 {
-	if (!vport->bridge)
+	struct mlx5_esw_bridge_port *port;
+
+	port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
+	if (!port)
 		return -EINVAL;
 
-	vport->bridge->ageing_time = clock_t_to_jiffies(ageing_time);
+	port->bridge->ageing_time = clock_t_to_jiffies(ageing_time);
 	return 0;
 }
 
-int mlx5_esw_bridge_vlan_filtering_set(bool enable, struct mlx5_eswitch *esw,
-				       struct mlx5_vport *vport)
+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_port *port;
 	struct mlx5_esw_bridge *bridge;
 	bool filtering;
 
-	if (!vport->bridge)
+	port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
+	if (!port)
 		return -EINVAL;
 
-	bridge = vport->bridge;
+	bridge = port->bridge;
 	filtering = bridge->flags & MLX5_ESW_BRIDGE_VLAN_FILTERING_FLAG;
 	if (filtering == enable)
 		return 0;
@@ -1033,114 +1089,143 @@ int mlx5_esw_bridge_vlan_filtering_set(bool enable, struct mlx5_eswitch *esw,
 	return 0;
 }
 
-static int mlx5_esw_bridge_vport_init(struct mlx5_esw_bridge_offloads *br_offloads,
-				      struct mlx5_esw_bridge *bridge,
-				      struct mlx5_vport *vport)
+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) {
-		err = -ENOMEM;
-		goto err_port_alloc;
-	}
+	if (!port)
+		return -ENOMEM;
 
-	port->vport_num = vport->vport;
+	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_insert(port, bridge);
+	err = mlx5_esw_bridge_port_insert(port, br_offloads);
 	if (err) {
-		esw_warn(esw->dev, "Failed to insert port metadata (vport=%u,err=%d)\n",
-			 vport->vport, 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);
 
-	vport->bridge = bridge;
 	return 0;
 
 err_port_insert:
 	kvfree(port);
-err_port_alloc:
-	mlx5_esw_bridge_put(br_offloads, bridge);
 	return err;
 }
 
 static int mlx5_esw_bridge_vport_cleanup(struct mlx5_esw_bridge_offloads *br_offloads,
-					 struct mlx5_vport *vport)
+					 struct mlx5_esw_bridge_port *port)
 {
-	struct mlx5_esw_bridge *bridge = vport->bridge;
+	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;
-	struct mlx5_esw_bridge_port *port;
 
 	list_for_each_entry_safe(entry, tmp, &bridge->fdb_list, list)
-		if (entry->vport_num == vport->vport)
+		if (entry->vport_num == vport_num && entry->esw_owner_vhca_id == esw_owner_vhca_id)
 			mlx5_esw_bridge_fdb_entry_cleanup(entry, bridge);
 
-	port = mlx5_esw_bridge_port_lookup(vport->vport, bridge);
-	if (!port) {
-		WARN(1, "Vport %u metadata not found on bridge", vport->vport);
-		return -EINVAL;
-	}
-
 	trace_mlx5_esw_bridge_vport_cleanup(port);
 	mlx5_esw_bridge_port_vlans_flush(port, bridge);
-	mlx5_esw_bridge_port_erase(port, bridge);
+	mlx5_esw_bridge_port_erase(port, br_offloads);
 	kvfree(port);
 	mlx5_esw_bridge_put(br_offloads, bridge);
-	vport->bridge = NULL;
 	return 0;
 }
 
-int mlx5_esw_bridge_vport_link(int ifindex, struct mlx5_esw_bridge_offloads *br_offloads,
-			       struct mlx5_vport *vport, struct netlink_ext_ack *extack)
+static int mlx5_esw_bridge_vport_link_with_flags(int ifindex, 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;
 
-	WARN_ON(vport->bridge);
-
 	bridge = mlx5_esw_bridge_lookup(ifindex, 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(br_offloads, bridge, vport);
-	if (err)
+	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_unlink(int ifindex, struct mlx5_esw_bridge_offloads *br_offloads,
-				 struct mlx5_vport *vport, struct netlink_ext_ack *extack)
+int mlx5_esw_bridge_vport_link(int ifindex, u16 vport_num, u16 esw_owner_vhca_id,
+			       struct mlx5_esw_bridge_offloads *br_offloads,
+			       struct netlink_ext_ack *extack)
 {
-	struct mlx5_esw_bridge *bridge = vport->bridge;
+	return mlx5_esw_bridge_vport_link_with_flags(ifindex, vport_num, esw_owner_vhca_id, 0,
+						     br_offloads, extack);
+}
+
+int mlx5_esw_bridge_vport_unlink(int ifindex, 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;
 
-	if (!bridge) {
+	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 (bridge->ifindex != ifindex) {
+	if (port->bridge->ifindex != ifindex) {
 		NL_SET_ERR_MSG_MOD(extack, "Port is attached to another bridge");
 		return -EINVAL;
 	}
 
-	err = mlx5_esw_bridge_vport_cleanup(br_offloads, vport);
+	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_port_vlan_add(u16 vid, u16 flags, struct mlx5_eswitch *esw,
-				  struct mlx5_vport *vport, struct netlink_ext_ack *extack)
+int mlx5_esw_bridge_vport_peer_link(int ifindex, 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(ifindex, vport_num, esw_owner_vhca_id,
+						     MLX5_ESW_BRIDGE_PORT_FLAG_PEER,
+						     br_offloads, extack);
+}
+
+int mlx5_esw_bridge_vport_peer_unlink(int ifindex, 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(ifindex, 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->vport, vport->bridge);
+	port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
 	if (!port)
 		return -EINVAL;
 
@@ -1148,10 +1233,10 @@ int mlx5_esw_bridge_port_vlan_add(u16 vid, u16 flags, struct mlx5_eswitch *esw,
 	if (vlan) {
 		if (vlan->flags == flags)
 			return 0;
-		mlx5_esw_bridge_vlan_cleanup(port, vlan, vport->bridge);
+		mlx5_esw_bridge_vlan_cleanup(port, vlan, port->bridge);
 	}
 
-	vlan = mlx5_esw_bridge_vlan_create(vid, flags, port, esw);
+	vlan = mlx5_esw_bridge_vlan_create(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);
@@ -1159,62 +1244,93 @@ int mlx5_esw_bridge_port_vlan_add(u16 vid, u16 flags, struct mlx5_eswitch *esw,
 	return 0;
 }
 
-void mlx5_esw_bridge_port_vlan_del(u16 vid, struct mlx5_eswitch *esw, struct mlx5_vport *vport)
+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->vport, vport->bridge);
+	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, vport->bridge);
+	mlx5_esw_bridge_vlan_cleanup(port, vlan, port->bridge);
 }
 
-void mlx5_esw_bridge_fdb_create(struct net_device *dev, struct mlx5_eswitch *esw,
-				struct mlx5_vport *vport,
-				struct switchdev_notifier_fdb_info *fdb_info)
+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 *bridge = vport->bridge;
 	struct mlx5_esw_bridge_fdb_entry *entry;
-	u16 vport_num = vport->vport;
+	struct mlx5_esw_bridge_fdb_key key;
+	struct mlx5_esw_bridge_port *port;
+	struct mlx5_esw_bridge *bridge;
 
-	if (!bridge) {
-		esw_info(esw->dev, "Vport is not assigned to bridge (vport=%u)\n", vport_num);
+	port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
+	if (!port || port->flags & MLX5_ESW_BRIDGE_PORT_FLAG_PEER)
+		return;
+
+	bridge = port->bridge;
+	ether_addr_copy(key.addr, fdb_info->addr);
+	key.vid = fdb_info->vid;
+	entry = rhashtable_lookup_fast(&bridge->fdb_ht, &key, fdb_ht_params);
+	if (!entry) {
+		esw_debug(br_offloads->esw->dev,
+			  "FDB entry with specified key not found (MAC=%pM,vid=%u,vport=%u)\n",
+			  key.addr, key.vid, vport_num);
 		return;
 	}
 
-	entry = mlx5_esw_bridge_fdb_entry_init(dev, vport_num, fdb_info->addr, fdb_info->vid,
-					       fdb_info->added_by_user, esw, bridge);
+	entry->lastuse = jiffies;
+}
+
+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
+	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, struct mlx5_eswitch *esw,
-				struct mlx5_vport *vport,
+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_esw_bridge *bridge = vport->bridge;
+	struct mlx5_eswitch *esw = br_offloads->esw;
 	struct mlx5_esw_bridge_fdb_entry *entry;
 	struct mlx5_esw_bridge_fdb_key key;
-	u16 vport_num = vport->vport;
+	struct mlx5_esw_bridge_port *port;
+	struct mlx5_esw_bridge *bridge;
 
-	if (!bridge) {
-		esw_warn(esw->dev, "Vport is not assigned to bridge (vport=%u)\n", vport_num);
+	port = mlx5_esw_bridge_port_lookup(vport_num, esw_owner_vhca_id, br_offloads);
+	if (!port)
 		return;
-	}
 
+	bridge = port->bridge;
 	ether_addr_copy(key.addr, fdb_info->addr);
 	key.vid = fdb_info->vid;
 	entry = rhashtable_lookup_fast(&bridge->fdb_ht, &key, fdb_ht_params);
@@ -1225,9 +1341,7 @@ void mlx5_esw_bridge_fdb_remove(struct net_device *dev, struct mlx5_eswitch *esw
 		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_DEL_TO_BRIDGE);
+	mlx5_esw_bridge_fdb_del_notify(entry);
 	mlx5_esw_bridge_fdb_entry_cleanup(entry, bridge);
 }
 
@@ -1245,11 +1359,10 @@ void mlx5_esw_bridge_update(struct mlx5_esw_bridge_offloads *br_offloads)
 				continue;
 
 			if (time_after(lastuse, entry->lastuse)) {
-				mlx5_esw_bridge_fdb_entry_refresh(lastuse, entry);
-			} else if (time_is_before_jiffies(entry->lastuse + bridge->ageing_time)) {
-				mlx5_esw_bridge_fdb_offload_notify(entry->dev, entry->key.addr,
-								   entry->key.vid,
-								   SWITCHDEV_FDB_DEL_TO_BRIDGE);
+				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_del_notify(entry);
 				mlx5_esw_bridge_fdb_entry_cleanup(entry, bridge);
 			}
 		}
@@ -1258,13 +1371,11 @@ void mlx5_esw_bridge_update(struct mlx5_esw_bridge_offloads *br_offloads)
 
 static void mlx5_esw_bridge_flush(struct mlx5_esw_bridge_offloads *br_offloads)
 {
-	struct mlx5_eswitch *esw = br_offloads->esw;
-	struct mlx5_vport *vport;
+	struct mlx5_esw_bridge_port *port;
 	unsigned long i;
 
-	mlx5_esw_for_each_vport(esw, i, vport)
-		if (vport->bridge)
-			mlx5_esw_bridge_vport_cleanup(br_offloads, vport);
+	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");
@@ -1279,6 +1390,7 @@ struct mlx5_esw_bridge_offloads *mlx5_esw_bridge_init(struct mlx5_eswitch *esw)
 		return ERR_PTR(-ENOMEM);
 
 	INIT_LIST_HEAD(&br_offloads->bridges);
+	xa_init(&br_offloads->ports);
 	br_offloads->esw = esw;
 	esw->br_offloads = br_offloads;
 
@@ -1293,6 +1405,7 @@ void mlx5_esw_bridge_cleanup(struct mlx5_eswitch *esw)
 		return;
 
 	mlx5_esw_bridge_flush(br_offloads);
+	WARN_ON(!xa_empty(&br_offloads->ports));
 
 	esw->br_offloads = NULL;
 	kvfree(br_offloads);

drivers/net/ethernet/mellanox/mlx5/core/esw/bridge.h

@@ -7,6 +7,7 @@
 #include <linux/notifier.h>
 #include <linux/list.h>
 #include <linux/workqueue.h>
+#include <linux/xarray.h>
 #include "eswitch.h"
 
 struct mlx5_flow_table;
@@ -15,6 +16,8 @@ struct mlx5_flow_group;
 struct mlx5_esw_bridge_offloads {
 	struct mlx5_eswitch *esw;
 	struct list_head bridges;
+	struct xarray ports;
+
 	struct notifier_block netdev_nb;
 	struct notifier_block nb_blk;
 	struct notifier_block nb;
@@ -31,23 +34,36 @@ struct mlx5_esw_bridge_offloads {
 
 struct mlx5_esw_bridge_offloads *mlx5_esw_bridge_init(struct mlx5_eswitch *esw);
 void mlx5_esw_bridge_cleanup(struct mlx5_eswitch *esw);
-int mlx5_esw_bridge_vport_link(int ifindex, struct mlx5_esw_bridge_offloads *br_offloads,
-			       struct mlx5_vport *vport, struct netlink_ext_ack *extack);
-int mlx5_esw_bridge_vport_unlink(int ifindex, struct mlx5_esw_bridge_offloads *br_offloads,
-				 struct mlx5_vport *vport, struct netlink_ext_ack *extack);
-void mlx5_esw_bridge_fdb_create(struct net_device *dev, struct mlx5_eswitch *esw,
-				struct mlx5_vport *vport,
+int mlx5_esw_bridge_vport_link(int ifindex, u16 vport_num, u16 esw_owner_vhca_id,
+			       struct mlx5_esw_bridge_offloads *br_offloads,
+			       struct netlink_ext_ack *extack);
+int mlx5_esw_bridge_vport_unlink(int ifindex, u16 vport_num, u16 esw_owner_vhca_id,
+				 struct mlx5_esw_bridge_offloads *br_offloads,
+				 struct netlink_ext_ack *extack);
+int mlx5_esw_bridge_vport_peer_link(int ifindex, u16 vport_num, u16 esw_owner_vhca_id,
+				    struct mlx5_esw_bridge_offloads *br_offloads,
+				    struct netlink_ext_ack *extack);
+int mlx5_esw_bridge_vport_peer_unlink(int ifindex, u16 vport_num, u16 esw_owner_vhca_id,
+				      struct mlx5_esw_bridge_offloads *br_offloads,
+				      struct netlink_ext_ack *extack);
+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);
+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);
-void mlx5_esw_bridge_fdb_remove(struct net_device *dev, struct mlx5_eswitch *esw,
-				struct mlx5_vport *vport,
+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);
 void mlx5_esw_bridge_update(struct mlx5_esw_bridge_offloads *br_offloads);
-int mlx5_esw_bridge_ageing_time_set(unsigned long ageing_time, struct mlx5_eswitch *esw,
-				    struct mlx5_vport *vport);
-int mlx5_esw_bridge_vlan_filtering_set(bool enable, struct mlx5_eswitch *esw,
-				       struct mlx5_vport *vport);
-int mlx5_esw_bridge_port_vlan_add(u16 vid, u16 flags, struct mlx5_eswitch *esw,
-				  struct mlx5_vport *vport, struct netlink_ext_ack *extack);
-void mlx5_esw_bridge_port_vlan_del(u16 vid, struct mlx5_eswitch *esw, struct mlx5_vport *vport);
+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);
+int mlx5_esw_bridge_vlan_filtering_set(u16 vport_num, u16 esw_owner_vhca_id, bool enable,
+				       struct mlx5_esw_bridge_offloads *br_offloads);
+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);
+void mlx5_esw_bridge_port_vlan_del(u16 vport_num, u16 esw_owner_vhca_id, u16 vid,
+				   struct mlx5_esw_bridge_offloads *br_offloads);
 
 #endif /* __MLX5_ESW_BRIDGE_H__ */

drivers/net/ethernet/mellanox/mlx5/core/esw/bridge_priv.h

@@ -19,6 +19,11 @@ struct mlx5_esw_bridge_fdb_key {
 
 enum {
 	MLX5_ESW_BRIDGE_FLAG_ADDED_BY_USER = BIT(0),
+	MLX5_ESW_BRIDGE_FLAG_PEER = BIT(1),
+};
+
+enum {
+	MLX5_ESW_BRIDGE_PORT_FLAG_PEER = BIT(0),
 };
 
 struct mlx5_esw_bridge_fdb_entry {
@@ -28,6 +33,7 @@ struct mlx5_esw_bridge_fdb_entry {
 	struct list_head list;
 	struct list_head vlan_list;
 	u16 vport_num;
+	u16 esw_owner_vhca_id;
 	u16 flags;
 
 	struct mlx5_flow_handle *ingress_handle;
@@ -47,6 +53,9 @@ struct mlx5_esw_bridge_vlan {
 
 struct mlx5_esw_bridge_port {
 	u16 vport_num;
+	u16 esw_owner_vhca_id;
+	u16 flags;
+	struct mlx5_esw_bridge *bridge;
 	struct xarray vlans;
 };
 

drivers/net/ethernet/mellanox/mlx5/core/esw/diag/bridge_tracepoint.h

@@ -85,11 +85,18 @@ DECLARE_EVENT_CLASS(mlx5_esw_bridge_port_template,
 		    TP_ARGS(port),
 		    TP_STRUCT__entry(
 			    __field(u16, vport_num)
+			    __field(u16, esw_owner_vhca_id)
+			    __field(u16, flags)
 			    ),
 		    TP_fast_assign(
 			    __entry->vport_num = port->vport_num;
+			    __entry->esw_owner_vhca_id = port->esw_owner_vhca_id;
+			    __entry->flags = port->flags;
 			    ),
-		    TP_printk("vport_num=%hu", __entry->vport_num)
+		    TP_printk("vport_num=%hu esw_owner_vhca_id=%hu flags=%hx",
+			      __entry->vport_num,
+			      __entry->esw_owner_vhca_id,
+			      __entry->flags)
 	);
 
 DEFINE_EVENT(mlx5_esw_bridge_port_template,

drivers/net/ethernet/mellanox/mlx5/core/eswitch.h

@@ -160,8 +160,6 @@ enum mlx5_eswitch_vport_event {
 	MLX5_VPORT_PROMISC_CHANGE = BIT(3),
 };
 
-struct mlx5_esw_bridge;
-
 struct mlx5_vport {
 	struct mlx5_core_dev    *dev;
 	struct hlist_head       uc_list[MLX5_L2_ADDR_HASH_SIZE];
@@ -190,7 +188,6 @@ struct mlx5_vport {
 	enum mlx5_eswitch_vport_event enabled_events;
 	int index;
 	struct devlink_port *dl_port;
-	struct mlx5_esw_bridge *bridge;
 };
 
 struct mlx5_esw_indir_table;