Merge tag 'mlx5e-updates-2018-12-14' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5e-updates-2018-12-14 (VF Lag)

From Aviv Heller,

Subsequent patches introduce VF LAG, which provdies load-balancing and
high-availability capabilities for VFs associated with different
physical ports of the same Connect-X card.

This series consists of the following:
 - mlx5 devcom, driver infrastructure that facilitates operations that involve
   both core devices (physical functions) of the same card, to synchronize and
   communicate between two driver instances of the same card.
 - Infrastructure for TC rule duplication.
 - Changes to LAG logic to enable its use when SR-IOV is enabled
 - PFs in switchdev mode is the only mode currently supported.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

13 files changed, 1000 insertions, 137 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/Makefile b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
index 9678051b8ff1..9de9abacf7f6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/Makefile
+++ b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
@@ -15,7 +15,7 @@ mlx5_core-y :=	main.o cmd.o debugfs.o fw.o eq.o uar.o pagealloc.o \
 		health.o mcg.o cq.o alloc.o qp.o port.o mr.o pd.o \
 		mad.o transobj.o vport.o sriov.o fs_cmd.o fs_core.o \
 		fs_counters.o rl.o lag.o dev.o events.o wq.o lib/gid.o \
-		diag/fs_tracepoint.o diag/fw_tracer.o
+		lib/devcom.o diag/fs_tracepoint.o diag/fw_tracer.o
 
 #
 # Netdev basic
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
index c1515f013501..b92f8b3ff6b2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
@@ -14,7 +14,8 @@ static int mlx5e_route_lookup_ipv4(struct mlx5e_priv *priv,
 				   u8 *out_ttl)
 {
 	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
-	struct mlx5e_rep_priv *uplink_rpriv;
+	struct net_device *uplink_dev, *uplink_upper;
+	bool dst_is_lag_dev;
 	struct rtable *rt;
 	struct neighbour *n = NULL;
 
@@ -28,10 +29,20 @@ static int mlx5e_route_lookup_ipv4(struct mlx5e_priv *priv,
 #else
 	return -EOPNOTSUPP;
 #endif
-	uplink_rpriv = mlx5_eswitch_get_uplink_priv(esw, REP_ETH);
-	/* if the egress device isn't on the same HW e-switch, we use the uplink */
-	if (!switchdev_port_same_parent_id(priv->netdev, rt->dst.dev))
-		*out_dev = uplink_rpriv->netdev;
+
+	uplink_dev = mlx5_eswitch_uplink_get_proto_dev(esw, REP_ETH);
+	uplink_upper = netdev_master_upper_dev_get(uplink_dev);
+	dst_is_lag_dev = (uplink_upper &&
+			  netif_is_lag_master(uplink_upper) &&
+			  rt->dst.dev == uplink_upper &&
+			  mlx5_lag_is_sriov(priv->mdev));
+
+	/* if the egress device isn't on the same HW e-switch or
+	 * it's a LAG device, use the uplink
+	 */
+	if (!switchdev_port_same_parent_id(priv->netdev, rt->dst.dev) ||
+	    dst_is_lag_dev)
+		*out_dev = uplink_dev;
 	else
 		*out_dev = rt->dst.dev;
 
@@ -65,8 +76,9 @@ static int mlx5e_route_lookup_ipv6(struct mlx5e_priv *priv,
 	struct dst_entry *dst;
 
 #if IS_ENABLED(CONFIG_INET) && IS_ENABLED(CONFIG_IPV6)
-	struct mlx5e_rep_priv *uplink_rpriv;
 	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
+	struct net_device *uplink_dev, *uplink_upper;
+	bool dst_is_lag_dev;
 	int ret;
 
 	ret = ipv6_stub->ipv6_dst_lookup(dev_net(mirred_dev), NULL, &dst,
@@ -77,10 +89,19 @@ static int mlx5e_route_lookup_ipv6(struct mlx5e_priv *priv,
 	if (!(*out_ttl))
 		*out_ttl = ip6_dst_hoplimit(dst);
 
-	uplink_rpriv = mlx5_eswitch_get_uplink_priv(esw, REP_ETH);
-	/* if the egress device isn't on the same HW e-switch, we use the uplink */
-	if (!switchdev_port_same_parent_id(priv->netdev, dst->dev))
-		*out_dev = uplink_rpriv->netdev;
+	uplink_dev = mlx5_eswitch_uplink_get_proto_dev(esw, REP_ETH);
+	uplink_upper = netdev_master_upper_dev_get(uplink_dev);
+	dst_is_lag_dev = (uplink_upper &&
+			  netif_is_lag_master(uplink_upper) &&
+			  dst->dev == uplink_upper &&
+			  mlx5_lag_is_sriov(priv->mdev));
+
+	/* if the egress device isn't on the same HW e-switch or
+	 * it's a LAG device, use the uplink
+	 */
+	if (!switchdev_port_same_parent_id(priv->netdev, dst->dev) ||
+	    dst_is_lag_dev)
+		*out_dev = uplink_dev;
 	else
 		*out_dev = dst->dev;
 #else
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
index e5230049612f..e4a34c9ef700 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -297,17 +297,31 @@ static const struct ethtool_ops mlx5e_rep_ethtool_ops = {
 int mlx5e_attr_get(struct net_device *dev, struct switchdev_attr *attr)
 {
 	struct mlx5e_priv *priv = netdev_priv(dev);
-	struct mlx5e_rep_priv *rpriv = priv->ppriv;
-	struct mlx5_eswitch_rep *rep = rpriv->rep;
 	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
+	struct net_device *uplink_upper = NULL;
+	struct mlx5e_priv *uplink_priv = NULL;
+	struct net_device *uplink_dev;
 
 	if (esw->mode == SRIOV_NONE)
 		return -EOPNOTSUPP;
 
+	uplink_dev = mlx5_eswitch_uplink_get_proto_dev(esw, REP_ETH);
+	if (uplink_dev) {
+		uplink_upper = netdev_master_upper_dev_get(uplink_dev);
+		uplink_priv = netdev_priv(uplink_dev);
+	}
+
 	switch (attr->id) {
 	case SWITCHDEV_ATTR_ID_PORT_PARENT_ID:
 		attr->u.ppid.id_len = ETH_ALEN;
-		ether_addr_copy(attr->u.ppid.id, rep->hw_id);
+		if (uplink_upper && mlx5_lag_is_sriov(uplink_priv->mdev)) {
+			ether_addr_copy(attr->u.ppid.id, uplink_upper->dev_addr);
+		} else {
+			struct mlx5e_rep_priv *rpriv = priv->ppriv;
+			struct mlx5_eswitch_rep *rep = rpriv->rep;
+
+			ether_addr_copy(attr->u.ppid.id, rep->hw_id);
+		}
 		break;
 	default:
 		return -EOPNOTSUPP;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
index 0921213561cb..53ebb5a48018 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
@@ -52,6 +52,7 @@
 #include "fs_core.h"
 #include "en/port.h"
 #include "en/tc_tun.h"
+#include "lib/devcom.h"
 
 struct mlx5_nic_flow_attr {
 	u32 action;
@@ -74,6 +75,7 @@ enum {
 	MLX5E_TC_FLOW_HAIRPIN	= BIT(MLX5E_TC_FLOW_BASE + 3),
 	MLX5E_TC_FLOW_HAIRPIN_RSS = BIT(MLX5E_TC_FLOW_BASE + 4),
 	MLX5E_TC_FLOW_SLOW	  = BIT(MLX5E_TC_FLOW_BASE + 5),
+	MLX5E_TC_FLOW_DUP         = BIT(MLX5E_TC_FLOW_BASE + 6),
 };
 
 #define MLX5E_TC_MAX_SPLITS 1
@@ -111,8 +113,10 @@ struct mlx5e_tc_flow {
 	 * destinations.
 	 */
 	struct encap_flow_item encaps[MLX5_MAX_FLOW_FWD_VPORTS];
+	struct mlx5e_tc_flow    *peer_flow;
 	struct list_head	mod_hdr; /* flows sharing the same mod hdr ID */
 	struct list_head	hairpin; /* flows sharing the same hairpin */
+	struct list_head	peer;    /* flows with peer flow */
 	union {
 		struct mlx5_esw_flow_attr esw_attr[0];
 		struct mlx5_nic_flow_attr nic_attr[0];
@@ -990,7 +994,7 @@ mlx5e_tc_add_fdb_flow(struct mlx5e_priv *priv,
 	}
 
 	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
-		counter = mlx5_fc_create(esw->dev, true);
+		counter = mlx5_fc_create(attr->counter_dev, true);
 		if (IS_ERR(counter)) {
 			err = PTR_ERR(counter);
 			goto err_create_counter;
@@ -1019,7 +1023,7 @@ mlx5e_tc_add_fdb_flow(struct mlx5e_priv *priv,
 	return 0;
 
 err_add_rule:
-	mlx5_fc_destroy(esw->dev, counter);
+	mlx5_fc_destroy(attr->counter_dev, counter);
 err_create_counter:
 	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
 		mlx5e_detach_mod_hdr(priv, flow);
@@ -1060,7 +1064,7 @@ static void mlx5e_tc_del_fdb_flow(struct mlx5e_priv *priv,
 		mlx5e_detach_mod_hdr(priv, flow);
 
 	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_COUNT)
-		mlx5_fc_destroy(esw->dev, attr->counter);
+		mlx5_fc_destroy(attr->counter_dev, attr->counter);
 }
 
 void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv,
@@ -1249,13 +1253,48 @@ static void mlx5e_detach_encap(struct mlx5e_priv *priv,
 	}
 }
 
+static void __mlx5e_tc_del_fdb_peer_flow(struct mlx5e_tc_flow *flow)
+{
+	struct mlx5_eswitch *esw = flow->priv->mdev->priv.eswitch;
+
+	if (!(flow->flags & MLX5E_TC_FLOW_ESWITCH) ||
+	    !(flow->flags & MLX5E_TC_FLOW_DUP))
+		return;
+
+	mutex_lock(&esw->offloads.peer_mutex);
+	list_del(&flow->peer);
+	mutex_unlock(&esw->offloads.peer_mutex);
+
+	flow->flags &= ~MLX5E_TC_FLOW_DUP;
+
+	mlx5e_tc_del_fdb_flow(flow->peer_flow->priv, flow->peer_flow);
+	kvfree(flow->peer_flow);
+	flow->peer_flow = NULL;
+}
+
+static void mlx5e_tc_del_fdb_peer_flow(struct mlx5e_tc_flow *flow)
+{
+	struct mlx5_core_dev *dev = flow->priv->mdev;
+	struct mlx5_devcom *devcom = dev->priv.devcom;
+	struct mlx5_eswitch *peer_esw;
+
+	peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+	if (!peer_esw)
+		return;
+
+	__mlx5e_tc_del_fdb_peer_flow(flow);
+	mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+}
+
 static void mlx5e_tc_del_flow(struct mlx5e_priv *priv,
 			      struct mlx5e_tc_flow *flow)
 {
-	if (flow->flags & MLX5E_TC_FLOW_ESWITCH)
+	if (flow->flags & MLX5E_TC_FLOW_ESWITCH) {
+		mlx5e_tc_del_fdb_peer_flow(flow);
 		mlx5e_tc_del_fdb_flow(priv, flow);
-	else
+	} else {
 		mlx5e_tc_del_nic_flow(priv, flow);
+	}
 }
 
 
@@ -2535,6 +2574,15 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 			if (switchdev_port_same_parent_id(priv->netdev,
 							  out_dev) ||
 			    is_merged_eswitch_dev(priv, out_dev)) {
+				struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
+				struct net_device *uplink_dev = mlx5_eswitch_uplink_get_proto_dev(esw, REP_ETH);
+				struct net_device *uplink_upper = netdev_master_upper_dev_get(uplink_dev);
+
+				if (uplink_upper &&
+				    netif_is_lag_master(uplink_upper) &&
+				    uplink_upper == out_dev)
+					out_dev = uplink_dev;
+
 				out_priv = netdev_priv(out_dev);
 				rpriv = out_priv->ppriv;
 				attr->dests[attr->out_count].rep = rpriv->rep;
@@ -2660,6 +2708,20 @@ static struct rhashtable *get_tc_ht(struct mlx5e_priv *priv)
 		return &priv->fs.tc.ht;
 }
 
+static bool is_peer_flow_needed(struct mlx5e_tc_flow *flow)
+{
+	struct mlx5_esw_flow_attr *attr = flow->esw_attr;
+	bool is_rep_ingress = attr->in_rep->vport != FDB_UPLINK_VPORT &&
+			      flow->flags & MLX5E_TC_FLOW_INGRESS;
+	bool act_is_encap = !!(attr->action &
+			       MLX5_FLOW_CONTEXT_ACTION_PACKET_REFORMAT);
+	bool esw_paired = mlx5_devcom_is_paired(attr->in_mdev->priv.devcom,
+						MLX5_DEVCOM_ESW_OFFLOADS);
+
+	return esw_paired && mlx5_lag_is_sriov(attr->in_mdev) &&
+	       (is_rep_ingress || act_is_encap);
+}
+
 static int
 mlx5e_alloc_flow(struct mlx5e_priv *priv, int attr_size,
 		 struct tc_cls_flower_offload *f, u16 flow_flags,
@@ -2693,13 +2755,16 @@ err_free:
 }
 
 static int
-mlx5e_add_fdb_flow(struct mlx5e_priv *priv,
-		   struct tc_cls_flower_offload *f,
-		   u16 flow_flags,
-		   struct net_device *filter_dev,
-		   struct mlx5e_tc_flow **__flow)
+__mlx5e_add_fdb_flow(struct mlx5e_priv *priv,
+		     struct tc_cls_flower_offload *f,
+		     u16 flow_flags,
+		     struct net_device *filter_dev,
+		     struct mlx5_eswitch_rep *in_rep,
+		     struct mlx5_core_dev *in_mdev,
+		     struct mlx5e_tc_flow **__flow)
 {
 	struct netlink_ext_ack *extack = f->common.extack;
+	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
 	struct mlx5e_tc_flow_parse_attr *parse_attr;
 	struct mlx5e_tc_flow *flow;
 	int attr_size, err;
@@ -2723,6 +2788,15 @@ mlx5e_add_fdb_flow(struct mlx5e_priv *priv,
 	if (err)
 		goto err_free;
 
+	flow->esw_attr->in_rep = in_rep;
+	flow->esw_attr->in_mdev = in_mdev;
+
+	if (MLX5_CAP_ESW(esw->dev, counter_eswitch_affinity) ==
+	    MLX5_COUNTER_SOURCE_ESWITCH)
+		flow->esw_attr->counter_dev = in_mdev;
+	else
+		flow->esw_attr->counter_dev = priv->mdev;
+
 	err = mlx5e_tc_add_fdb_flow(priv, parse_attr, flow, extack);
 	if (err)
 		goto err_free;
@@ -2738,6 +2812,87 @@ out:
 	return err;
 }
 
+static int mlx5e_tc_add_fdb_peer_flow(struct tc_cls_flower_offload *f,
+				      struct mlx5e_tc_flow *flow)
+{
+	struct mlx5e_priv *priv = flow->priv, *peer_priv;
+	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch, *peer_esw;
+	struct mlx5_devcom *devcom = priv->mdev->priv.devcom;
+	struct mlx5e_tc_flow_parse_attr *parse_attr;
+	struct mlx5e_rep_priv *peer_urpriv;
+	struct mlx5e_tc_flow *peer_flow;
+	struct mlx5_core_dev *in_mdev;
+	int err = 0;
+
+	peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+	if (!peer_esw)
+		return -ENODEV;
+
+	peer_urpriv = mlx5_eswitch_get_uplink_priv(peer_esw, REP_ETH);
+	peer_priv = netdev_priv(peer_urpriv->netdev);
+
+	/* in_mdev is assigned of which the packet originated from.
+	 * So packets redirected to uplink use the same mdev of the
+	 * original flow and packets redirected from uplink use the
+	 * peer mdev.
+	 */
+	if (flow->esw_attr->in_rep->vport == FDB_UPLINK_VPORT)
+		in_mdev = peer_priv->mdev;
+	else
+		in_mdev = priv->mdev;
+
+	parse_attr = flow->esw_attr->parse_attr;
+	err = __mlx5e_add_fdb_flow(peer_priv, f, flow->flags,
+				   parse_attr->filter_dev,
+				   flow->esw_attr->in_rep, in_mdev, &peer_flow);
+	if (err)
+		goto out;
+
+	flow->peer_flow = peer_flow;
+	flow->flags |= MLX5E_TC_FLOW_DUP;
+	mutex_lock(&esw->offloads.peer_mutex);
+	list_add_tail(&flow->peer, &esw->offloads.peer_flows);
+	mutex_unlock(&esw->offloads.peer_mutex);
+
+out:
+	mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+	return err;
+}
+
+static int
+mlx5e_add_fdb_flow(struct mlx5e_priv *priv,
+		   struct tc_cls_flower_offload *f,
+		   u16 flow_flags,
+		   struct net_device *filter_dev,
+		   struct mlx5e_tc_flow **__flow)
+{
+	struct mlx5e_rep_priv *rpriv = priv->ppriv;
+	struct mlx5_eswitch_rep *in_rep = rpriv->rep;
+	struct mlx5_core_dev *in_mdev = priv->mdev;
+	struct mlx5e_tc_flow *flow;
+	int err;
+
+	err = __mlx5e_add_fdb_flow(priv, f, flow_flags, filter_dev, in_rep,
+				   in_mdev, &flow);
+	if (err)
+		goto out;
+
+	if (is_peer_flow_needed(flow)) {
+		err = mlx5e_tc_add_fdb_peer_flow(f, flow);
+		if (err) {
+			mlx5e_tc_del_fdb_flow(priv, flow);
+			goto out;
+		}
+	}
+
+	*__flow = flow;
+
+	return 0;
+
+out:
+	return err;
+}
+
 static int
 mlx5e_add_nic_flow(struct mlx5e_priv *priv,
 		   struct tc_cls_flower_offload *f,
@@ -2882,7 +3037,9 @@ int mlx5e_delete_flower(struct net_device *dev, struct mlx5e_priv *priv,
 int mlx5e_stats_flower(struct net_device *dev, struct mlx5e_priv *priv,
 		       struct tc_cls_flower_offload *f, int flags)
 {
+	struct mlx5_devcom *devcom = priv->mdev->priv.devcom;
 	struct rhashtable *tc_ht = get_tc_ht(priv);
+	struct mlx5_eswitch *peer_esw;
 	struct mlx5e_tc_flow *flow;
 	struct mlx5_fc *counter;
 	u64 bytes;
@@ -2902,6 +3059,27 @@ int mlx5e_stats_flower(struct net_device *dev, struct mlx5e_priv *priv,
 
 	mlx5_fc_query_cached(counter, &bytes, &packets, &lastuse);
 
+	peer_esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+	if (!peer_esw)
+		goto out;
+
+	if ((flow->flags & MLX5E_TC_FLOW_DUP) &&
+	    (flow->peer_flow->flags & MLX5E_TC_FLOW_OFFLOADED)) {
+		u64 bytes2;
+		u64 packets2;
+		u64 lastuse2;
+
+		counter = mlx5e_tc_get_counter(flow->peer_flow);
+		mlx5_fc_query_cached(counter, &bytes2, &packets2, &lastuse2);
+
+		bytes += bytes2;
+		packets += packets2;
+		lastuse = max_t(u64, lastuse, lastuse2);
+	}
+
+	mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+
+out:
 	tcf_exts_stats_update(f->exts, bytes, packets, lastuse);
 
 	return 0;
@@ -3014,3 +3192,11 @@ int mlx5e_tc_num_filters(struct mlx5e_priv *priv)
 
 	return atomic_read(&tc_ht->nelems);
 }
+
+void mlx5e_tc_clean_fdb_peer_flows(struct mlx5_eswitch *esw)
+{
+	struct mlx5e_tc_flow *flow, *tmp;
+
+	list_for_each_entry_safe(flow, tmp, &esw->offloads.peer_flows, peer)
+		__mlx5e_tc_del_fdb_peer_flow(flow);
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index e6a9b19d8626..ef2bec50423d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -39,7 +39,6 @@
 #include "lib/eq.h"
 #include "eswitch.h"
 #include "fs_core.h"
-#include "lib/eq.h"
 
 #define UPLINK_VPORT 0xFFFF
 
@@ -1633,6 +1632,8 @@ int mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode)
 	esw_info(esw->dev, "E-Switch enable SRIOV: nvfs(%d) mode (%d)\n", nvfs, mode);
 	esw->mode = mode;
 
+	mlx5_lag_update(esw->dev);
+
 	if (mode == SRIOV_LEGACY) {
 		err = esw_create_legacy_fdb_table(esw);
 	} else {
@@ -1709,6 +1710,8 @@ void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw)
 	old_mode = esw->mode;
 	esw->mode = SRIOV_NONE;
 
+	mlx5_lag_update(esw->dev);
+
 	if (old_mode == SRIOV_OFFLOADS)
 		mlx5_reload_interface(esw->dev, MLX5_INTERFACE_PROTOCOL_IB);
 }
@@ -2226,3 +2229,14 @@ u8 mlx5_eswitch_mode(struct mlx5_eswitch *esw)
 	return ESW_ALLOWED(esw) ? esw->mode : SRIOV_NONE;
 }
 EXPORT_SYMBOL_GPL(mlx5_eswitch_mode);
+
+bool mlx5_esw_lag_prereq(struct mlx5_core_dev *dev0, struct mlx5_core_dev *dev1)
+{
+	if ((dev0->priv.eswitch->mode == SRIOV_NONE &&
+	     dev1->priv.eswitch->mode == SRIOV_NONE) ||
+	    (dev0->priv.eswitch->mode == SRIOV_OFFLOADS &&
+	     dev1->priv.eswitch->mode == SRIOV_OFFLOADS))
+		return true;
+
+	return false;
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index 87c9dea9bccf..9c89eea9b2c3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -143,6 +143,8 @@ struct mlx5_eswitch_fdb {
 		struct offloads_fdb {
 			struct mlx5_flow_table *slow_fdb;
 			struct mlx5_flow_group *send_to_vport_grp;
+			struct mlx5_flow_group *peer_miss_grp;
+			struct mlx5_flow_handle **peer_miss_rules;
 			struct mlx5_flow_group *miss_grp;
 			struct mlx5_flow_handle *miss_rule_uni;
 			struct mlx5_flow_handle *miss_rule_multi;
@@ -165,6 +167,8 @@ struct mlx5_esw_offload {
 	struct mlx5_flow_table *ft_offloads;
 	struct mlx5_flow_group *vport_rx_group;
 	struct mlx5_eswitch_rep *vport_reps;
+	struct list_head peer_flows;
+	struct mutex peer_mutex;
 	DECLARE_HASHTABLE(encap_tbl, 8);
 	DECLARE_HASHTABLE(mod_hdr_tbl, 8);
 	u8 inline_mode;
@@ -289,6 +293,7 @@ enum {
 struct mlx5_esw_flow_attr {
 	struct mlx5_eswitch_rep *in_rep;
 	struct mlx5_core_dev	*in_mdev;
+	struct mlx5_core_dev    *counter_dev;
 
 	int split_count;
 	int out_count;
@@ -346,6 +351,9 @@ static inline bool mlx5_eswitch_vlan_actions_supported(struct mlx5_core_dev *dev
 		MLX5_CAP_ESW_FLOWTABLE_FDB(dev, push_vlan_2);
 }
 
+bool mlx5_esw_lag_prereq(struct mlx5_core_dev *dev0,
+			 struct mlx5_core_dev *dev1);
+
 #define MLX5_DEBUG_ESWITCH_MASK BIT(3)
 
 #define esw_info(dev, format, ...)				\
@@ -362,6 +370,7 @@ static inline int  mlx5_eswitch_init(struct mlx5_core_dev *dev) { return 0; }
 static inline void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw) {}
 static inline int  mlx5_eswitch_enable_sriov(struct mlx5_eswitch *esw, int nvfs, int mode) { return 0; }
 static inline void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw) {}
+static inline bool mlx5_esw_lag_prereq(struct mlx5_core_dev *dev0, struct mlx5_core_dev *dev1) { return true; }
 
 #define FDB_MAX_CHAIN 1
 #define FDB_SLOW_PATH_CHAIN (FDB_MAX_CHAIN + 1)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index bde1fb8c284b..76cb57202474 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -39,6 +39,7 @@
 #include "eswitch.h"
 #include "en.h"
 #include "fs_core.h"
+#include "lib/devcom.h"
 
 enum {
 	FDB_FAST_PATH = 0,
@@ -541,6 +542,98 @@ void mlx5_eswitch_del_send_to_vport_rule(struct mlx5_flow_handle *rule)
 	mlx5_del_flow_rules(rule);
 }
 
+static void peer_miss_rules_setup(struct mlx5_core_dev *peer_dev,
+				  struct mlx5_flow_spec *spec,
+				  struct mlx5_flow_destination *dest)
+{
+	void *misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
+				  misc_parameters);
+
+	MLX5_SET(fte_match_set_misc, misc, source_eswitch_owner_vhca_id,
+		 MLX5_CAP_GEN(peer_dev, vhca_id));
+
+	spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS;
+
+	misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
+			    misc_parameters);
+	MLX5_SET_TO_ONES(fte_match_set_misc, misc, source_port);
+	MLX5_SET_TO_ONES(fte_match_set_misc, misc,
+			 source_eswitch_owner_vhca_id);
+
+	dest->type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
+	dest->vport.num = 0;
+	dest->vport.vhca_id = MLX5_CAP_GEN(peer_dev, vhca_id);
+	dest->vport.flags |= MLX5_FLOW_DEST_VPORT_VHCA_ID;
+}
+
+static int esw_add_fdb_peer_miss_rules(struct mlx5_eswitch *esw,
+				       struct mlx5_core_dev *peer_dev)
+{
+	struct mlx5_flow_destination dest = {};
+	struct mlx5_flow_act flow_act = {0};
+	struct mlx5_flow_handle **flows;
+	struct mlx5_flow_handle *flow;
+	struct mlx5_flow_spec *spec;
+	/* total vports is the same for both e-switches */
+	int nvports = esw->total_vports;
+	void *misc;
+	int err, i;
+
+	spec = kvzalloc(sizeof(*spec), GFP_KERNEL);
+	if (!spec)
+		return -ENOMEM;
+
+	peer_miss_rules_setup(peer_dev, spec, &dest);
+
+	flows = kvzalloc(nvports * sizeof(*flows), GFP_KERNEL);
+	if (!flows) {
+		err = -ENOMEM;
+		goto alloc_flows_err;
+	}
+
+	flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
+	misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
+			    misc_parameters);
+
+	for (i = 1; i < nvports; i++) {
+		MLX5_SET(fte_match_set_misc, misc, source_port, i);
+		flow = mlx5_add_flow_rules(esw->fdb_table.offloads.slow_fdb,
+					   spec, &flow_act, &dest, 1);
+		if (IS_ERR(flow)) {
+			err = PTR_ERR(flow);
+			esw_warn(esw->dev, "FDB: Failed to add peer miss flow rule err %d\n", err);
+			goto add_flow_err;
+		}
+		flows[i] = flow;
+	}
+
+	esw->fdb_table.offloads.peer_miss_rules = flows;
+
+	kvfree(spec);
+	return 0;
+
+add_flow_err:
+	for (i--; i > 0; i--)
+		mlx5_del_flow_rules(flows[i]);
+	kvfree(flows);
+alloc_flows_err:
+	kvfree(spec);
+	return err;
+}
+
+static void esw_del_fdb_peer_miss_rules(struct mlx5_eswitch *esw)
+{
+	struct mlx5_flow_handle **flows;
+	int i;
+
+	flows = esw->fdb_table.offloads.peer_miss_rules;
+
+	for (i = 1; i < esw->total_vports; i++)
+		mlx5_del_flow_rules(flows[i]);
+
+	kvfree(flows);
+}
+
 static int esw_add_fdb_miss_rule(struct mlx5_eswitch *esw)
 {
 	struct mlx5_flow_act flow_act = {0};
@@ -811,7 +904,8 @@ static int esw_create_offloads_fdb_tables(struct mlx5_eswitch *esw, int nvports)
 		esw->fdb_table.offloads.fdb_left[i] =
 			ESW_POOLS[i] <= fdb_max ? ESW_SIZE / ESW_POOLS[i] : 0;
 
-	table_size = nvports * MAX_SQ_NVPORTS + MAX_PF_SQ + 2;
+	table_size = nvports * MAX_SQ_NVPORTS + MAX_PF_SQ + 2 +
+		esw->total_vports;
 
 	/* create the slow path fdb with encap set, so further table instances
 	 * can be created at run time while VFs are probed if the FW allows that.
@@ -866,6 +960,34 @@ static int esw_create_offloads_fdb_tables(struct mlx5_eswitch *esw, int nvports)
 	}
 	esw->fdb_table.offloads.send_to_vport_grp = g;
 
+	/* create peer esw miss group */
+	memset(flow_group_in, 0, inlen);
+	MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
+		 MLX5_MATCH_MISC_PARAMETERS);
+
+	match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in,
+				      match_criteria);
+
+	MLX5_SET_TO_ONES(fte_match_param, match_criteria,
+			 misc_parameters.source_port);
+	MLX5_SET_TO_ONES(fte_match_param, match_criteria,
+			 misc_parameters.source_eswitch_owner_vhca_id);
+
+	MLX5_SET(create_flow_group_in, flow_group_in,
+		 source_eswitch_owner_vhca_id_valid, 1);
+	MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, ix);
+	MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index,
+		 ix + esw->total_vports - 1);
+	ix += esw->total_vports;
+
+	g = mlx5_create_flow_group(fdb, flow_group_in);
+	if (IS_ERR(g)) {
+		err = PTR_ERR(g);
+		esw_warn(dev, "Failed to create peer miss flow group err(%d)\n", err);
+		goto peer_miss_err;
+	}
+	esw->fdb_table.offloads.peer_miss_grp = g;
+
 	/* create miss group */
 	memset(flow_group_in, 0, inlen);
 	MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
@@ -898,6 +1020,8 @@ static int esw_create_offloads_fdb_tables(struct mlx5_eswitch *esw, int nvports)
 miss_rule_err:
 	mlx5_destroy_flow_group(esw->fdb_table.offloads.miss_grp);
 miss_err:
+	mlx5_destroy_flow_group(esw->fdb_table.offloads.peer_miss_grp);
+peer_miss_err:
 	mlx5_destroy_flow_group(esw->fdb_table.offloads.send_to_vport_grp);
 send_vport_err:
 	esw_destroy_offloads_fast_fdb_tables(esw);
@@ -917,6 +1041,7 @@ static void esw_destroy_offloads_fdb_tables(struct mlx5_eswitch *esw)
 	mlx5_del_flow_rules(esw->fdb_table.offloads.miss_rule_multi);
 	mlx5_del_flow_rules(esw->fdb_table.offloads.miss_rule_uni);
 	mlx5_destroy_flow_group(esw->fdb_table.offloads.send_to_vport_grp);
+	mlx5_destroy_flow_group(esw->fdb_table.offloads.peer_miss_grp);
 	mlx5_destroy_flow_group(esw->fdb_table.offloads.miss_grp);
 
 	mlx5_destroy_flow_table(esw->fdb_table.offloads.slow_fdb);
@@ -1173,6 +1298,105 @@ err_reps:
 	return err;
 }
 
+#define ESW_OFFLOADS_DEVCOM_PAIR	(0)
+#define ESW_OFFLOADS_DEVCOM_UNPAIR	(1)
+
+static int mlx5_esw_offloads_pair(struct mlx5_eswitch *esw,
+				  struct mlx5_eswitch *peer_esw)
+{
+	int err;
+
+	err = esw_add_fdb_peer_miss_rules(esw, peer_esw->dev);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+void mlx5e_tc_clean_fdb_peer_flows(struct mlx5_eswitch *esw);
+
+static void mlx5_esw_offloads_unpair(struct mlx5_eswitch *esw)
+{
+	mlx5e_tc_clean_fdb_peer_flows(esw);
+	esw_del_fdb_peer_miss_rules(esw);
+}
+
+static int mlx5_esw_offloads_devcom_event(int event,
+					  void *my_data,
+					  void *event_data)
+{
+	struct mlx5_eswitch *esw = my_data;
+	struct mlx5_eswitch *peer_esw = event_data;
+	struct mlx5_devcom *devcom = esw->dev->priv.devcom;
+	int err;
+
+	switch (event) {
+	case ESW_OFFLOADS_DEVCOM_PAIR:
+		err = mlx5_esw_offloads_pair(esw, peer_esw);
+		if (err)
+			goto err_out;
+
+		err = mlx5_esw_offloads_pair(peer_esw, esw);
+		if (err)
+			goto err_pair;
+
+		mlx5_devcom_set_paired(devcom, MLX5_DEVCOM_ESW_OFFLOADS, true);
+		break;
+
+	case ESW_OFFLOADS_DEVCOM_UNPAIR:
+		if (!mlx5_devcom_is_paired(devcom, MLX5_DEVCOM_ESW_OFFLOADS))
+			break;
+
+		mlx5_devcom_set_paired(devcom, MLX5_DEVCOM_ESW_OFFLOADS, false);
+		mlx5_esw_offloads_unpair(peer_esw);
+		mlx5_esw_offloads_unpair(esw);
+		break;
+	}
+
+	return 0;
+
+err_pair:
+	mlx5_esw_offloads_unpair(esw);
+
+err_out:
+	mlx5_core_err(esw->dev, "esw offloads devcom event failure, event %u err %d",
+		      event, err);
+	return err;
+}
+
+static void esw_offloads_devcom_init(struct mlx5_eswitch *esw)
+{
+	struct mlx5_devcom *devcom = esw->dev->priv.devcom;
+
+	INIT_LIST_HEAD(&esw->offloads.peer_flows);
+	mutex_init(&esw->offloads.peer_mutex);
+
+	if (!MLX5_CAP_ESW(esw->dev, merged_eswitch))
+		return;
+
+	mlx5_devcom_register_component(devcom,
+				       MLX5_DEVCOM_ESW_OFFLOADS,
+				       mlx5_esw_offloads_devcom_event,
+				       esw);
+
+	mlx5_devcom_send_event(devcom,
+			       MLX5_DEVCOM_ESW_OFFLOADS,
+			       ESW_OFFLOADS_DEVCOM_PAIR, esw);
+}
+
+static void esw_offloads_devcom_cleanup(struct mlx5_eswitch *esw)
+{
+	struct mlx5_devcom *devcom = esw->dev->priv.devcom;
+
+	if (!MLX5_CAP_ESW(esw->dev, merged_eswitch))
+		return;
+
+	mlx5_devcom_send_event(devcom, MLX5_DEVCOM_ESW_OFFLOADS,
+			       ESW_OFFLOADS_DEVCOM_UNPAIR, esw);
+
+	mlx5_devcom_unregister_component(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+}
+
 int esw_offloads_init(struct mlx5_eswitch *esw, int nvports)
 {
 	int err;
@@ -1195,6 +1419,7 @@ int esw_offloads_init(struct mlx5_eswitch *esw, int nvports)
 	if (err)
 		goto err_reps;
 
+	esw_offloads_devcom_init(esw);
 	return 0;
 
 err_reps:
@@ -1233,6 +1458,7 @@ static int esw_offloads_stop(struct mlx5_eswitch *esw,
 
 void esw_offloads_cleanup(struct mlx5_eswitch *esw, int nvports)
 {
+	esw_offloads_devcom_cleanup(esw);
 	esw_offloads_unload_reps(esw, nvports);
 	esw_destroy_vport_rx_group(esw);
 	esw_destroy_offloads_table(esw);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag.c b/drivers/net/ethernet/mellanox/mlx5/core/lag.c
index 582b2f18010a..5187dc7a72a3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag.c
@@ -34,11 +34,15 @@
 #include <linux/mlx5/driver.h>
 #include <linux/mlx5/vport.h>
 #include "mlx5_core.h"
+#include "eswitch.h"
 
 enum {
-	MLX5_LAG_FLAG_BONDED = 1 << 0,
+	MLX5_LAG_FLAG_ROCE   = 1 << 0,
+	MLX5_LAG_FLAG_SRIOV  = 1 << 1,
 };
 
+#define MLX5_LAG_MODE_FLAGS (MLX5_LAG_FLAG_ROCE | MLX5_LAG_FLAG_SRIOV)
+
 struct lag_func {
 	struct mlx5_core_dev *dev;
 	struct net_device    *netdev;
@@ -61,11 +65,6 @@ struct mlx5_lag {
 	struct lag_tracker        tracker;
 	struct delayed_work       bond_work;
 	struct notifier_block     nb;
-
-	/* Admin state. Allow lag only if allowed is true
-	 * even if network conditions for lag were met
-	 */
-	bool                      allowed;
 };
 
 /* General purpose, use for short periods of time.
@@ -165,9 +164,19 @@ static int mlx5_lag_dev_get_netdev_idx(struct mlx5_lag *ldev,
 	return -1;
 }
 
-static bool mlx5_lag_is_bonded(struct mlx5_lag *ldev)
+static bool __mlx5_lag_is_roce(struct mlx5_lag *ldev)
+{
+	return !!(ldev->flags & MLX5_LAG_FLAG_ROCE);
+}
+
+static bool __mlx5_lag_is_sriov(struct mlx5_lag *ldev)
 {
-	return !!(ldev->flags & MLX5_LAG_FLAG_BONDED);
+	return !!(ldev->flags & MLX5_LAG_FLAG_SRIOV);
+}
+
+static bool __mlx5_lag_is_active(struct mlx5_lag *ldev)
+{
+	return !!(ldev->flags & MLX5_LAG_MODE_FLAGS);
 }
 
 static void mlx5_infer_tx_affinity_mapping(struct lag_tracker *tracker,
@@ -186,36 +195,128 @@ static void mlx5_infer_tx_affinity_mapping(struct lag_tracker *tracker,
 		*port2 = 1;
 }
 
-static void mlx5_activate_lag(struct mlx5_lag *ldev,
-			      struct lag_tracker *tracker)
+static void mlx5_modify_lag(struct mlx5_lag *ldev,
+			    struct lag_tracker *tracker)
 {
 	struct mlx5_core_dev *dev0 = ldev->pf[0].dev;
+	u8 v2p_port1, v2p_port2;
 	int err;
 
-	ldev->flags |= MLX5_LAG_FLAG_BONDED;
+	mlx5_infer_tx_affinity_mapping(tracker, &v2p_port1,
+				       &v2p_port2);
+
+	if (v2p_port1 != ldev->v2p_map[0] ||
+	    v2p_port2 != ldev->v2p_map[1]) {
+		ldev->v2p_map[0] = v2p_port1;
+		ldev->v2p_map[1] = v2p_port2;
+
+		mlx5_core_info(dev0, "modify lag map port 1:%d port 2:%d",
+			       ldev->v2p_map[0], ldev->v2p_map[1]);
+
+		err = mlx5_cmd_modify_lag(dev0, v2p_port1, v2p_port2);
+		if (err)
+			mlx5_core_err(dev0,
+				      "Failed to modify LAG (%d)\n",
+				      err);
+	}
+}
+
+static int mlx5_create_lag(struct mlx5_lag *ldev,
+			   struct lag_tracker *tracker)
+{
+	struct mlx5_core_dev *dev0 = ldev->pf[0].dev;
+	int err;
 
 	mlx5_infer_tx_affinity_mapping(tracker, &ldev->v2p_map[0],
 				       &ldev->v2p_map[1]);
 
+	mlx5_core_info(dev0, "lag map port 1:%d port 2:%d",
+		       ldev->v2p_map[0], ldev->v2p_map[1]);
+
 	err = mlx5_cmd_create_lag(dev0, ldev->v2p_map[0], ldev->v2p_map[1]);
 	if (err)
 		mlx5_core_err(dev0,
 			      "Failed to create LAG (%d)\n",
 			      err);
+	return err;
 }
 
-static void mlx5_deactivate_lag(struct mlx5_lag *ldev)
+static int mlx5_activate_lag(struct mlx5_lag *ldev,
+			     struct lag_tracker *tracker,
+			     u8 flags)
 {
+	bool roce_lag = !!(flags & MLX5_LAG_FLAG_ROCE);
 	struct mlx5_core_dev *dev0 = ldev->pf[0].dev;
 	int err;
 
-	ldev->flags &= ~MLX5_LAG_FLAG_BONDED;
+	err = mlx5_create_lag(ldev, tracker);
+	if (err) {
+		if (roce_lag) {
+			mlx5_core_err(dev0,
+				      "Failed to activate RoCE LAG\n");
+		} else {
+			mlx5_core_err(dev0,
+				      "Failed to activate VF LAG\n"
+				      "Make sure all VFs are unbound prior to VF LAG activation or deactivation\n");
+		}
+		return err;
+	}
+
+	ldev->flags |= flags;
+	return 0;
+}
+
+static int mlx5_deactivate_lag(struct mlx5_lag *ldev)
+{
+	struct mlx5_core_dev *dev0 = ldev->pf[0].dev;
+	bool roce_lag = __mlx5_lag_is_roce(ldev);
+	int err;
+
+	ldev->flags &= ~MLX5_LAG_MODE_FLAGS;
 
 	err = mlx5_cmd_destroy_lag(dev0);
-	if (err)
-		mlx5_core_err(dev0,
-			      "Failed to destroy LAG (%d)\n",
-			      err);
+	if (err) {
+		if (roce_lag) {
+			mlx5_core_err(dev0,
+				      "Failed to deactivate RoCE LAG; driver restart required\n");
+		} else {
+			mlx5_core_err(dev0,
+				      "Failed to deactivate VF LAG; driver restart required\n"
+				      "Make sure all VFs are unbound prior to VF LAG activation or deactivation\n");
+		}
+	}
+
+	return err;
+}
+
+static bool mlx5_lag_check_prereq(struct mlx5_lag *ldev)
+{
+	if (ldev->pf[0].dev &&
+	    ldev->pf[1].dev &&
+	    mlx5_esw_lag_prereq(ldev->pf[0].dev, ldev->pf[1].dev))
+		return true;
+	else
+		return false;
+}
+
+static void mlx5_lag_add_ib_devices(struct mlx5_lag *ldev)
+{
+	int i;
+
+	for (i = 0; i < MLX5_MAX_PORTS; i++)
+		if (ldev->pf[i].dev)
+			mlx5_add_dev_by_protocol(ldev->pf[i].dev,
+						 MLX5_INTERFACE_PROTOCOL_IB);
+}
+
+static void mlx5_lag_remove_ib_devices(struct mlx5_lag *ldev)
+{
+	int i;
+
+	for (i = 0; i < MLX5_MAX_PORTS; i++)
+		if (ldev->pf[i].dev)
+			mlx5_remove_dev_by_protocol(ldev->pf[i].dev,
+						    MLX5_INTERFACE_PROTOCOL_IB);
 }
 
 static void mlx5_do_bond(struct mlx5_lag *ldev)
@@ -223,9 +324,8 @@ static void mlx5_do_bond(struct mlx5_lag *ldev)
 	struct mlx5_core_dev *dev0 = ldev->pf[0].dev;
 	struct mlx5_core_dev *dev1 = ldev->pf[1].dev;
 	struct lag_tracker tracker;
-	u8 v2p_port1, v2p_port2;
-	int i, err;
-	bool do_bond;
+	bool do_bond, roce_lag;
+	int err;
 
 	if (!dev0 || !dev1)
 		return;
@@ -234,42 +334,45 @@ static void mlx5_do_bond(struct mlx5_lag *ldev)
 	tracker = ldev->tracker;
 	mutex_unlock(&lag_mutex);
 
-	do_bond = tracker.is_bonded && ldev->allowed;
+	do_bond = tracker.is_bonded && mlx5_lag_check_prereq(ldev);
 
-	if (do_bond && !mlx5_lag_is_bonded(ldev)) {
-		for (i = 0; i < MLX5_MAX_PORTS; i++)
-			mlx5_remove_dev_by_protocol(ldev->pf[i].dev,
-						    MLX5_INTERFACE_PROTOCOL_IB);
+	if (do_bond && !__mlx5_lag_is_active(ldev)) {
+		roce_lag = !mlx5_sriov_is_enabled(dev0) &&
+			   !mlx5_sriov_is_enabled(dev1);
 
-		mlx5_activate_lag(ldev, &tracker);
+		if (roce_lag)
+			mlx5_lag_remove_ib_devices(ldev);
 
-		mlx5_add_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
-		mlx5_nic_vport_enable_roce(dev1);
-	} else if (do_bond && mlx5_lag_is_bonded(ldev)) {
-		mlx5_infer_tx_affinity_mapping(&tracker, &v2p_port1,
-					       &v2p_port2);
+		err = mlx5_activate_lag(ldev, &tracker,
+					roce_lag ? MLX5_LAG_FLAG_ROCE :
+					MLX5_LAG_FLAG_SRIOV);
+		if (err) {
+			if (roce_lag)
+				mlx5_lag_add_ib_devices(ldev);
 
-		if ((v2p_port1 != ldev->v2p_map[0]) ||
-		    (v2p_port2 != ldev->v2p_map[1])) {
-			ldev->v2p_map[0] = v2p_port1;
-			ldev->v2p_map[1] = v2p_port2;
+			return;
+		}
 
-			err = mlx5_cmd_modify_lag(dev0, v2p_port1, v2p_port2);
-			if (err)
-				mlx5_core_err(dev0,
-					      "Failed to modify LAG (%d)\n",
-					      err);
+		if (roce_lag) {
+			mlx5_add_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
+			mlx5_nic_vport_enable_roce(dev1);
+		}
+	} else if (do_bond && __mlx5_lag_is_active(ldev)) {
+		mlx5_modify_lag(ldev, &tracker);
+	} else if (!do_bond && __mlx5_lag_is_active(ldev)) {
+		roce_lag = __mlx5_lag_is_roce(ldev);
+
+		if (roce_lag) {
+			mlx5_remove_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
+			mlx5_nic_vport_disable_roce(dev1);
 		}
-	} else if (!do_bond && mlx5_lag_is_bonded(ldev)) {
-		mlx5_remove_dev_by_protocol(dev0, MLX5_INTERFACE_PROTOCOL_IB);
-		mlx5_nic_vport_disable_roce(dev1);
 
-		mlx5_deactivate_lag(ldev);
+		err = mlx5_deactivate_lag(ldev);
+		if (err)
+			return;
 
-		for (i = 0; i < MLX5_MAX_PORTS; i++)
-			if (ldev->pf[i].dev)
-				mlx5_add_dev_by_protocol(ldev->pf[i].dev,
-							 MLX5_INTERFACE_PROTOCOL_IB);
+		if (roce_lag)
+			mlx5_lag_add_ib_devices(ldev);
 	}
 }
 
@@ -419,15 +522,6 @@ static int mlx5_lag_netdev_event(struct notifier_block *this,
 	return NOTIFY_DONE;
 }
 
-static bool mlx5_lag_check_prereq(struct mlx5_lag *ldev)
-{
-	if ((ldev->pf[0].dev && mlx5_sriov_is_enabled(ldev->pf[0].dev)) ||
-	    (ldev->pf[1].dev && mlx5_sriov_is_enabled(ldev->pf[1].dev)))
-		return false;
-	else
-		return true;
-}
-
 static struct mlx5_lag *mlx5_lag_dev_alloc(void)
 {
 	struct mlx5_lag *ldev;
@@ -437,7 +531,6 @@ static struct mlx5_lag *mlx5_lag_dev_alloc(void)
 		return NULL;
 
 	INIT_DELAYED_WORK(&ldev->bond_work, mlx5_do_bond_work);
-	ldev->allowed = mlx5_lag_check_prereq(ldev);
 
 	return ldev;
 }
@@ -462,7 +555,6 @@ static void mlx5_lag_dev_add_pf(struct mlx5_lag *ldev,
 	ldev->tracker.netdev_state[fn].link_up = 0;
 	ldev->tracker.netdev_state[fn].tx_enabled = 0;
 
-	ldev->allowed = mlx5_lag_check_prereq(ldev);
 	dev->priv.lag = ldev;
 
 	mutex_unlock(&lag_mutex);
@@ -484,7 +576,6 @@ static void mlx5_lag_dev_remove_pf(struct mlx5_lag *ldev,
 	memset(&ldev->pf[i], 0, sizeof(*ldev->pf));
 
 	dev->priv.lag = NULL;
-	ldev->allowed = mlx5_lag_check_prereq(ldev);
 	mutex_unlock(&lag_mutex);
 }
 
@@ -532,7 +623,7 @@ void mlx5_lag_remove(struct mlx5_core_dev *dev)
 	if (!ldev)
 		return;
 
-	if (mlx5_lag_is_bonded(ldev))
+	if (__mlx5_lag_is_active(ldev))
 		mlx5_deactivate_lag(ldev);
 
 	mlx5_lag_dev_remove_pf(ldev, dev);
@@ -549,56 +640,61 @@ void mlx5_lag_remove(struct mlx5_core_dev *dev)
 	}
 }
 
-bool mlx5_lag_is_active(struct mlx5_core_dev *dev)
+bool mlx5_lag_is_roce(struct mlx5_core_dev *dev)
 {
 	struct mlx5_lag *ldev;
 	bool res;
 
 	mutex_lock(&lag_mutex);
 	ldev = mlx5_lag_dev_get(dev);
-	res  = ldev && mlx5_lag_is_bonded(ldev);
+	res  = ldev && __mlx5_lag_is_roce(ldev);
 	mutex_unlock(&lag_mutex);
 
 	return res;
 }
-EXPORT_SYMBOL(mlx5_lag_is_active);
+EXPORT_SYMBOL(mlx5_lag_is_roce);
 
-static int mlx5_lag_set_state(struct mlx5_core_dev *dev, bool allow)
+bool mlx5_lag_is_active(struct mlx5_core_dev *dev)
 {
 	struct mlx5_lag *ldev;
-	int ret = 0;
-	bool lag_active;
-
-	mlx5_dev_list_lock();
+	bool res;
 
+	mutex_lock(&lag_mutex);
 	ldev = mlx5_lag_dev_get(dev);
-	if (!ldev) {
-		ret = -ENODEV;
-		goto unlock;
-	}
-	lag_active = mlx5_lag_is_bonded(ldev);
-	if (!mlx5_lag_check_prereq(ldev) && allow) {
-		ret = -EINVAL;
-		goto unlock;
-	}
-	if (ldev->allowed == allow)
-		goto unlock;
-	ldev->allowed = allow;
-	if ((lag_active && !allow) || allow)
-		mlx5_do_bond(ldev);
-unlock:
-	mlx5_dev_list_unlock();
-	return ret;
+	res  = ldev && __mlx5_lag_is_active(ldev);
+	mutex_unlock(&lag_mutex);
+
+	return res;
 }
+EXPORT_SYMBOL(mlx5_lag_is_active);
 
-int mlx5_lag_forbid(struct mlx5_core_dev *dev)
+bool mlx5_lag_is_sriov(struct mlx5_core_dev *dev)
 {
-	return mlx5_lag_set_state(dev, false);
+	struct mlx5_lag *ldev;
+	bool res;
+
+	mutex_lock(&lag_mutex);
+	ldev = mlx5_lag_dev_get(dev);
+	res  = ldev && __mlx5_lag_is_sriov(ldev);
+	mutex_unlock(&lag_mutex);
+
+	return res;
 }
+EXPORT_SYMBOL(mlx5_lag_is_sriov);
 
-int mlx5_lag_allow(struct mlx5_core_dev *dev)
+void mlx5_lag_update(struct mlx5_core_dev *dev)
 {
-	return mlx5_lag_set_state(dev, true);
+	struct mlx5_lag *ldev;
+
+	mlx5_dev_list_lock();
+	ldev = mlx5_lag_dev_get(dev);
+	if (!ldev)
+		goto unlock;
+
+	mlx5_do_bond(ldev);
+
+unlock:
+	mlx5_dev_list_unlock();
 }
 
 struct net_device *mlx5_lag_get_roce_netdev(struct mlx5_core_dev *dev)
@@ -609,7 +705,7 @@ struct net_device *mlx5_lag_get_roce_netdev(struct mlx5_core_dev *dev)
 	mutex_lock(&lag_mutex);
 	ldev = mlx5_lag_dev_get(dev);
 
-	if (!(ldev && mlx5_lag_is_bonded(ldev)))
+	if (!(ldev && __mlx5_lag_is_roce(ldev)))
 		goto unlock;
 
 	if (ldev->tracker.tx_type == NETDEV_LAG_TX_TYPE_ACTIVEBACKUP) {
@@ -638,7 +734,7 @@ bool mlx5_lag_intf_add(struct mlx5_interface *intf, struct mlx5_priv *priv)
 		return true;
 
 	ldev = mlx5_lag_dev_get(dev);
-	if (!ldev || !mlx5_lag_is_bonded(ldev) || ldev->pf[0].dev == dev)
+	if (!ldev || !__mlx5_lag_is_roce(ldev) || ldev->pf[0].dev == dev)
 		return true;
 
 	/* If bonded, we do not add an IB device for PF1. */
@@ -665,7 +761,7 @@ int mlx5_lag_query_cong_counters(struct mlx5_core_dev *dev,
 
 	mutex_lock(&lag_mutex);
 	ldev = mlx5_lag_dev_get(dev);
-	if (ldev && mlx5_lag_is_bonded(ldev)) {
+	if (ldev && __mlx5_lag_is_roce(ldev)) {
 		num_ports = MLX5_MAX_PORTS;
 		mdev[0] = ldev->pf[0].dev;
 		mdev[1] = ldev->pf[1].dev;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.c
new file mode 100644
index 000000000000..bced2efe9bef
--- /dev/null
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.c
@@ -0,0 +1,255 @@
+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/* Copyright (c) 2018 Mellanox Technologies */
+
+#include <linux/mlx5/vport.h>
+#include "lib/devcom.h"
+
+static LIST_HEAD(devcom_list);
+
+#define devcom_for_each_component(priv, comp, iter) \
+	for (iter = 0; \
+	     comp = &(priv)->components[iter], iter < MLX5_DEVCOM_NUM_COMPONENTS; \
+	     iter++)
+
+struct mlx5_devcom_component {
+	struct {
+		void *data;
+	} device[MLX5_MAX_PORTS];
+
+	mlx5_devcom_event_handler_t handler;
+	struct rw_semaphore sem;
+	bool paired;
+};
+
+struct mlx5_devcom_list {
+	struct list_head list;
+
+	struct mlx5_devcom_component components[MLX5_DEVCOM_NUM_COMPONENTS];
+	struct mlx5_core_dev *devs[MLX5_MAX_PORTS];
+};
+
+struct mlx5_devcom {
+	struct mlx5_devcom_list *priv;
+	int idx;
+};
+
+static struct mlx5_devcom_list *mlx5_devcom_list_alloc(void)
+{
+	struct mlx5_devcom_component *comp;
+	struct mlx5_devcom_list *priv;
+	int i;
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return NULL;
+
+	devcom_for_each_component(priv, comp, i)
+		init_rwsem(&comp->sem);
+
+	return priv;
+}
+
+static struct mlx5_devcom *mlx5_devcom_alloc(struct mlx5_devcom_list *priv,
+					     u8 idx)
+{
+	struct mlx5_devcom *devcom;
+
+	devcom = kzalloc(sizeof(*devcom), GFP_KERNEL);
+	if (!devcom)
+		return NULL;
+
+	devcom->priv = priv;
+	devcom->idx = idx;
+	return devcom;
+}
+
+/* Must be called with intf_mutex held */
+struct mlx5_devcom *mlx5_devcom_register_device(struct mlx5_core_dev *dev)
+{
+	struct mlx5_devcom_list *priv = NULL, *iter;
+	struct mlx5_devcom *devcom = NULL;
+	bool new_priv = false;
+	u64 sguid0, sguid1;
+	int idx, i;
+
+	if (!mlx5_core_is_pf(dev))
+		return NULL;
+
+	sguid0 = mlx5_query_nic_system_image_guid(dev);
+	list_for_each_entry(iter, &devcom_list, list) {
+		struct mlx5_core_dev *tmp_dev = NULL;
+
+		idx = -1;
+		for (i = 0; i < MLX5_MAX_PORTS; i++) {
+			if (iter->devs[i])
+				tmp_dev = iter->devs[i];
+			else
+				idx = i;
+		}
+
+		if (idx == -1)
+			continue;
+
+		sguid1 = mlx5_query_nic_system_image_guid(tmp_dev);
+		if (sguid0 != sguid1)
+			continue;
+
+		priv = iter;
+		break;
+	}
+
+	if (!priv) {
+		priv = mlx5_devcom_list_alloc();
+		if (!priv)
+			return ERR_PTR(-ENOMEM);
+
+		idx = 0;
+		new_priv = true;
+	}
+
+	priv->devs[idx] = dev;
+	devcom = mlx5_devcom_alloc(priv, idx);
+	if (!devcom) {
+		kfree(priv);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	if (new_priv)
+		list_add(&priv->list, &devcom_list);
+
+	return devcom;
+}
+
+/* Must be called with intf_mutex held */
+void mlx5_devcom_unregister_device(struct mlx5_devcom *devcom)
+{
+	struct mlx5_devcom_list *priv;
+	int i;
+
+	if (IS_ERR_OR_NULL(devcom))
+		return;
+
+	priv = devcom->priv;
+	priv->devs[devcom->idx] = NULL;
+
+	kfree(devcom);
+
+	for (i = 0; i < MLX5_MAX_PORTS; i++)
+		if (priv->devs[i])
+			break;
+
+	if (i != MLX5_MAX_PORTS)
+		return;
+
+	list_del(&priv->list);
+	kfree(priv);
+}
+
+void mlx5_devcom_register_component(struct mlx5_devcom *devcom,
+				    enum mlx5_devcom_components id,
+				    mlx5_devcom_event_handler_t handler,
+				    void *data)
+{
+	struct mlx5_devcom_component *comp;
+
+	if (IS_ERR_OR_NULL(devcom))
+		return;
+
+	WARN_ON(!data);
+
+	comp = &devcom->priv->components[id];
+	down_write(&comp->sem);
+	comp->handler = handler;
+	comp->device[devcom->idx].data = data;
+	up_write(&comp->sem);
+}
+
+void mlx5_devcom_unregister_component(struct mlx5_devcom *devcom,
+				      enum mlx5_devcom_components id)
+{
+	struct mlx5_devcom_component *comp;
+
+	if (IS_ERR_OR_NULL(devcom))
+		return;
+
+	comp = &devcom->priv->components[id];
+	down_write(&comp->sem);
+	comp->device[devcom->idx].data = NULL;
+	up_write(&comp->sem);
+}
+
+int mlx5_devcom_send_event(struct mlx5_devcom *devcom,
+			   enum mlx5_devcom_components id,
+			   int event,
+			   void *event_data)
+{
+	struct mlx5_devcom_component *comp;
+	int err = -ENODEV, i;
+
+	if (IS_ERR_OR_NULL(devcom))
+		return err;
+
+	comp = &devcom->priv->components[id];
+	down_write(&comp->sem);
+	for (i = 0; i < MLX5_MAX_PORTS; i++)
+		if (i != devcom->idx && comp->device[i].data) {
+			err = comp->handler(event, comp->device[i].data,
+					    event_data);
+			break;
+		}
+
+	up_write(&comp->sem);
+	return err;
+}
+
+void mlx5_devcom_set_paired(struct mlx5_devcom *devcom,
+			    enum mlx5_devcom_components id,
+			    bool paired)
+{
+	struct mlx5_devcom_component *comp;
+
+	comp = &devcom->priv->components[id];
+	WARN_ON(!rwsem_is_locked(&comp->sem));
+
+	comp->paired = paired;
+}
+
+bool mlx5_devcom_is_paired(struct mlx5_devcom *devcom,
+			   enum mlx5_devcom_components id)
+{
+	if (IS_ERR_OR_NULL(devcom))
+		return false;
+
+	return devcom->priv->components[id].paired;
+}
+
+void *mlx5_devcom_get_peer_data(struct mlx5_devcom *devcom,
+				enum mlx5_devcom_components id)
+{
+	struct mlx5_devcom_component *comp;
+	int i;
+
+	if (IS_ERR_OR_NULL(devcom))
+		return NULL;
+
+	comp = &devcom->priv->components[id];
+	down_read(&comp->sem);
+	if (!comp->paired) {
+		up_read(&comp->sem);
+		return NULL;
+	}
+
+	for (i = 0; i < MLX5_MAX_PORTS; i++)
+		if (i != devcom->idx)
+			break;
+
+	return comp->device[i].data;
+}
+
+void mlx5_devcom_release_peer_data(struct mlx5_devcom *devcom,
+				   enum mlx5_devcom_components id)
+{
+	struct mlx5_devcom_component *comp = &devcom->priv->components[id];
+
+	up_read(&comp->sem);
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.h b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.h
new file mode 100644
index 000000000000..939d5bf1581b
--- /dev/null
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/devcom.h
@@ -0,0 +1,46 @@
+/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
+/* Copyright (c) 2018 Mellanox Technologies */
+
+#ifndef __LIB_MLX5_DEVCOM_H__
+#define __LIB_MLX5_DEVCOM_H__
+
+#include <linux/mlx5/driver.h>
+
+enum mlx5_devcom_components {
+	MLX5_DEVCOM_ESW_OFFLOADS,
+
+	MLX5_DEVCOM_NUM_COMPONENTS,
+};
+
+typedef int (*mlx5_devcom_event_handler_t)(int event,
+					   void *my_data,
+					   void *event_data);
+
+struct mlx5_devcom *mlx5_devcom_register_device(struct mlx5_core_dev *dev);
+void mlx5_devcom_unregister_device(struct mlx5_devcom *devcom);
+
+void mlx5_devcom_register_component(struct mlx5_devcom *devcom,
+				    enum mlx5_devcom_components id,
+				    mlx5_devcom_event_handler_t handler,
+				    void *data);
+void mlx5_devcom_unregister_component(struct mlx5_devcom *devcom,
+				      enum mlx5_devcom_components id);
+
+int mlx5_devcom_send_event(struct mlx5_devcom *devcom,
+			   enum mlx5_devcom_components id,
+			   int event,
+			   void *event_data);
+
+void mlx5_devcom_set_paired(struct mlx5_devcom *devcom,
+			    enum mlx5_devcom_components id,
+			    bool paired);
+bool mlx5_devcom_is_paired(struct mlx5_devcom *devcom,
+			   enum mlx5_devcom_components id);
+
+void *mlx5_devcom_get_peer_data(struct mlx5_devcom *devcom,
+				enum mlx5_devcom_components id);
+void mlx5_devcom_release_peer_data(struct mlx5_devcom *devcom,
+				   enum mlx5_devcom_components id);
+
+#endif
+
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index 778995573812..c23553164e0d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -63,6 +63,7 @@
 #include "accel/tls.h"
 #include "lib/clock.h"
 #include "lib/vxlan.h"
+#include "lib/devcom.h"
 #include "diag/fw_tracer.h"
 
 MODULE_AUTHOR("Eli Cohen <eli@mellanox.com>");
@@ -722,16 +723,21 @@ static int mlx5_init_once(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	struct pci_dev *pdev = dev->pdev;
 	int err;
 
+	priv->devcom = mlx5_devcom_register_device(dev);
+	if (IS_ERR(priv->devcom))
+		dev_err(&pdev->dev, "failed to register with devcom (0x%p)\n",
+			priv->devcom);
+
 	err = mlx5_query_board_id(dev);
 	if (err) {
 		dev_err(&pdev->dev, "query board id failed\n");
-		goto out;
+		goto err_devcom;
 	}
 
 	err = mlx5_eq_table_init(dev);
 	if (err) {
 		dev_err(&pdev->dev, "failed to initialize eq\n");
-		goto out;
+		goto err_devcom;
 	}
 
 	err = mlx5_events_init(dev);
@@ -807,8 +813,9 @@ err_events_cleanup:
 	mlx5_events_cleanup(dev);
 err_eq_cleanup:
 	mlx5_eq_table_cleanup(dev);
+err_devcom:
+	mlx5_devcom_unregister_device(dev->priv.devcom);
 
-out:
 	return err;
 }
 
@@ -828,6 +835,7 @@ static void mlx5_cleanup_once(struct mlx5_core_dev *dev)
 	mlx5_cq_debugfs_cleanup(dev);
 	mlx5_events_cleanup(dev);
 	mlx5_eq_table_cleanup(dev);
+	mlx5_devcom_unregister_device(dev->priv.devcom);
 }
 
 static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
index fd3141a4b3f1..73bf46599ec6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
@@ -185,10 +185,8 @@ static inline int mlx5_lag_is_lacp_owner(struct mlx5_core_dev *dev)
 		    MLX5_CAP_GEN(dev, lag_master);
 }
 
-int mlx5_lag_allow(struct mlx5_core_dev *dev);
-int mlx5_lag_forbid(struct mlx5_core_dev *dev);
-
 void mlx5_reload_interface(struct mlx5_core_dev *mdev, int protocol);
+void mlx5_lag_update(struct mlx5_core_dev *dev);
 
 enum {
 	MLX5_NIC_IFC_FULL		= 0,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
index a0674962f02c..6e178030d8fb 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
@@ -216,20 +216,10 @@ int mlx5_core_sriov_configure(struct pci_dev *pdev, int num_vfs)
 	if (!mlx5_core_is_pf(dev))
 		return -EPERM;
 
-	if (num_vfs) {
-		int ret;
-
-		ret = mlx5_lag_forbid(dev);
-		if (ret && (ret != -ENODEV))
-			return ret;
-	}
-
-	if (num_vfs) {
+	if (num_vfs)
 		err = mlx5_sriov_enable(pdev, num_vfs);
-	} else {
+	else
 		mlx5_sriov_disable(pdev);
-		mlx5_lag_allow(dev);
-	}
 
 	return err ? err : num_vfs;
 }