1 files changed, 346 insertions, 210 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index 1a909d47aee2..0b4b43ebae9a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -260,7 +260,8 @@ mlx5_eswitch_set_rule_source_port(struct mlx5_eswitch *esw,
 								   attr->in_rep->vport));
 
 		misc2 = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, misc_parameters_2);
-		MLX5_SET_TO_ONES(fte_match_set_misc2, misc2, metadata_reg_c_0);
+		MLX5_SET(fte_match_set_misc2, misc2, metadata_reg_c_0,
+			 mlx5_eswitch_get_vport_metadata_mask());
 
 		spec->match_criteria_enable |= MLX5_MATCH_MISC_PARAMETERS_2;
 		misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, misc_parameters);
@@ -299,6 +300,7 @@ mlx5_eswitch_add_offloaded_rule(struct mlx5_eswitch *esw,
 	bool split = !!(attr->split_count);
 	struct mlx5_flow_handle *rule;
 	struct mlx5_flow_table *fdb;
+	bool hairpin = false;
 	int j, i = 0;
 
 	if (esw->mode != MLX5_ESWITCH_OFFLOADS)
@@ -323,7 +325,12 @@ mlx5_eswitch_add_offloaded_rule(struct mlx5_eswitch *esw,
 	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_FWD_DEST) {
 		struct mlx5_flow_table *ft;
 
-		if (attr->flags & MLX5_ESW_ATTR_FLAG_SLOW_PATH) {
+		if (attr->dest_ft) {
+			flow_act.flags |= FLOW_ACT_IGNORE_FLOW_LEVEL;
+			dest[i].type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
+			dest[i].ft = attr->dest_ft;
+			i++;
+		} else if (attr->flags & MLX5_ESW_ATTR_FLAG_SLOW_PATH) {
 			flow_act.flags |= FLOW_ACT_IGNORE_FLOW_LEVEL;
 			dest[i].type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
 			dest[i].ft = mlx5_esw_chains_get_tc_end_ft(esw);
@@ -377,31 +384,41 @@ mlx5_eswitch_add_offloaded_rule(struct mlx5_eswitch *esw,
 	if (split) {
 		fdb = esw_vport_tbl_get(esw, attr);
 	} else {
-		fdb = mlx5_esw_chains_get_table(esw, attr->chain, attr->prio,
-						0);
-		mlx5_eswitch_set_rule_source_port(esw, spec, attr);
+		if (attr->chain || attr->prio)
+			fdb = mlx5_esw_chains_get_table(esw, attr->chain,
+							attr->prio, 0);
+		else
+			fdb = attr->fdb;
+
+		if (!(attr->flags & MLX5_ESW_ATTR_FLAG_NO_IN_PORT))
+			mlx5_eswitch_set_rule_source_port(esw, spec, attr);
 	}
 	if (IS_ERR(fdb)) {
 		rule = ERR_CAST(fdb);
 		goto err_esw_get;
 	}
 
-	if (mlx5_eswitch_termtbl_required(esw, &flow_act, spec))
+	if (mlx5_eswitch_termtbl_required(esw, attr, &flow_act, spec)) {
 		rule = mlx5_eswitch_add_termtbl_rule(esw, fdb, spec, attr,
 						     &flow_act, dest, i);
-	else
+		hairpin = true;
+	} else {
 		rule = mlx5_add_flow_rules(fdb, spec, &flow_act, dest, i);
+	}
 	if (IS_ERR(rule))
 		goto err_add_rule;
 	else
 		atomic64_inc(&esw->offloads.num_flows);
 
+	if (hairpin)
+		attr->flags |= MLX5_ESW_ATTR_FLAG_HAIRPIN;
+
 	return rule;
 
 err_add_rule:
 	if (split)
 		esw_vport_tbl_put(esw, attr);
-	else
+	else if (attr->chain || attr->prio)
 		mlx5_esw_chains_put_table(esw, attr->chain, attr->prio, 0);
 err_esw_get:
 	if (!(attr->flags & MLX5_ESW_ATTR_FLAG_SLOW_PATH) && attr->dest_chain)
@@ -484,10 +501,12 @@ __mlx5_eswitch_del_rule(struct mlx5_eswitch *esw,
 
 	mlx5_del_flow_rules(rule);
 
-	/* unref the term table */
-	for (i = 0; i < MLX5_MAX_FLOW_FWD_VPORTS; i++) {
-		if (attr->dests[i].termtbl)
-			mlx5_eswitch_termtbl_put(esw, attr->dests[i].termtbl);
+	if (attr->flags & MLX5_ESW_ATTR_FLAG_HAIRPIN) {
+		/* unref the term table */
+		for (i = 0; i < MLX5_MAX_FLOW_FWD_VPORTS; i++) {
+			if (attr->dests[i].termtbl)
+				mlx5_eswitch_termtbl_put(esw, attr->dests[i].termtbl);
+		}
 	}
 
 	atomic64_dec(&esw->offloads.num_flows);
@@ -498,7 +517,7 @@ __mlx5_eswitch_del_rule(struct mlx5_eswitch *esw,
 	} else {
 		if (split)
 			esw_vport_tbl_put(esw, attr);
-		else
+		else if (attr->chain || attr->prio)
 			mlx5_esw_chains_put_table(esw, attr->chain, attr->prio,
 						  0);
 		if (attr->dest_chain)
@@ -762,14 +781,21 @@ void mlx5_eswitch_del_send_to_vport_rule(struct mlx5_flow_handle *rule)
 	mlx5_del_flow_rules(rule);
 }
 
+static bool mlx5_eswitch_reg_c1_loopback_supported(struct mlx5_eswitch *esw)
+{
+	return MLX5_CAP_ESW_FLOWTABLE(esw->dev, fdb_to_vport_reg_c_id) &
+	       MLX5_FDB_TO_VPORT_REG_C_1;
+}
+
 static int esw_set_passing_vport_metadata(struct mlx5_eswitch *esw, bool enable)
 {
 	u32 out[MLX5_ST_SZ_DW(query_esw_vport_context_out)] = {};
 	u32 in[MLX5_ST_SZ_DW(modify_esw_vport_context_in)] = {};
-	u8 fdb_to_vport_reg_c_id;
+	u8 curr, wanted;
 	int err;
 
-	if (!mlx5_eswitch_vport_match_metadata_enabled(esw))
+	if (!mlx5_eswitch_reg_c1_loopback_supported(esw) &&
+	    !mlx5_eswitch_vport_match_metadata_enabled(esw))
 		return 0;
 
 	err = mlx5_eswitch_query_esw_vport_context(esw->dev, 0, false,
@@ -777,22 +803,33 @@ static int esw_set_passing_vport_metadata(struct mlx5_eswitch *esw, bool enable)
 	if (err)
 		return err;
 
-	fdb_to_vport_reg_c_id = MLX5_GET(query_esw_vport_context_out, out,
-					 esw_vport_context.fdb_to_vport_reg_c_id);
+	curr = MLX5_GET(query_esw_vport_context_out, out,
+			esw_vport_context.fdb_to_vport_reg_c_id);
+	wanted = MLX5_FDB_TO_VPORT_REG_C_0;
+	if (mlx5_eswitch_reg_c1_loopback_supported(esw))
+		wanted |= MLX5_FDB_TO_VPORT_REG_C_1;
 
 	if (enable)
-		fdb_to_vport_reg_c_id |= MLX5_FDB_TO_VPORT_REG_C_0;
+		curr |= wanted;
 	else
-		fdb_to_vport_reg_c_id &= ~MLX5_FDB_TO_VPORT_REG_C_0;
+		curr &= ~wanted;
 
 	MLX5_SET(modify_esw_vport_context_in, in,
-		 esw_vport_context.fdb_to_vport_reg_c_id, fdb_to_vport_reg_c_id);
+		 esw_vport_context.fdb_to_vport_reg_c_id, curr);
 
 	MLX5_SET(modify_esw_vport_context_in, in,
 		 field_select.fdb_to_vport_reg_c_id, 1);
 
-	return mlx5_eswitch_modify_esw_vport_context(esw->dev, 0, false,
-						     in, sizeof(in));
+	err = mlx5_eswitch_modify_esw_vport_context(esw->dev, 0, false, in,
+						    sizeof(in));
+	if (!err) {
+		if (enable && (curr & MLX5_FDB_TO_VPORT_REG_C_1))
+			esw->flags |= MLX5_ESWITCH_REG_C1_LOOPBACK_ENABLED;
+		else
+			esw->flags &= ~MLX5_ESWITCH_REG_C1_LOOPBACK_ENABLED;
+	}
+
+	return err;
 }
 
 static void peer_miss_rules_setup(struct mlx5_eswitch *esw,
@@ -805,7 +842,8 @@ static void peer_miss_rules_setup(struct mlx5_eswitch *esw,
 	if (mlx5_eswitch_vport_match_metadata_enabled(esw)) {
 		misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
 				    misc_parameters_2);
-		MLX5_SET_TO_ONES(fte_match_set_misc2, misc, metadata_reg_c_0);
+		MLX5_SET(fte_match_set_misc2, misc, metadata_reg_c_0,
+			 mlx5_eswitch_get_vport_metadata_mask());
 
 		spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS_2;
 	} else {
@@ -1020,6 +1058,56 @@ out:
 	return err;
 }
 
+struct mlx5_flow_handle *
+esw_add_restore_rule(struct mlx5_eswitch *esw, u32 tag)
+{
+	struct mlx5_flow_act flow_act = { .flags = FLOW_ACT_NO_APPEND, };
+	struct mlx5_flow_table *ft = esw->offloads.ft_offloads_restore;
+	struct mlx5_flow_context *flow_context;
+	struct mlx5_flow_handle *flow_rule;
+	struct mlx5_flow_destination dest;
+	struct mlx5_flow_spec *spec;
+	void *misc;
+
+	spec = kzalloc(sizeof(*spec), GFP_KERNEL);
+	if (!spec)
+		return ERR_PTR(-ENOMEM);
+
+	misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria,
+			    misc_parameters_2);
+	MLX5_SET(fte_match_set_misc2, misc, metadata_reg_c_0,
+		 ESW_CHAIN_TAG_METADATA_MASK);
+	misc = MLX5_ADDR_OF(fte_match_param, spec->match_value,
+			    misc_parameters_2);
+	MLX5_SET(fte_match_set_misc2, misc, metadata_reg_c_0, tag);
+	spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS_2;
+	flow_act.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST |
+			  MLX5_FLOW_CONTEXT_ACTION_MOD_HDR;
+	flow_act.modify_hdr = esw->offloads.restore_copy_hdr_id;
+
+	flow_context = &spec->flow_context;
+	flow_context->flags |= FLOW_CONTEXT_HAS_TAG;
+	flow_context->flow_tag = tag;
+	dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
+	dest.ft = esw->offloads.ft_offloads;
+
+	flow_rule = mlx5_add_flow_rules(ft, spec, &flow_act, &dest, 1);
+	kfree(spec);
+
+	if (IS_ERR(flow_rule))
+		esw_warn(esw->dev,
+			 "Failed to create restore rule for tag: %d, err(%d)\n",
+			 tag, (int)PTR_ERR(flow_rule));
+
+	return flow_rule;
+}
+
+u32
+esw_get_max_restore_tag(struct mlx5_eswitch *esw)
+{
+	return ESW_CHAIN_TAG_METADATA_MASK;
+}
+
 #define MAX_PF_SQ 256
 #define MAX_SQ_NVPORTS 32
 
@@ -1035,8 +1123,9 @@ static void esw_set_flow_group_source_port(struct mlx5_eswitch *esw,
 			 match_criteria_enable,
 			 MLX5_MATCH_MISC_PARAMETERS_2);
 
-		MLX5_SET_TO_ONES(fte_match_param, match_criteria,
-				 misc_parameters_2.metadata_reg_c_0);
+		MLX5_SET(fte_match_param, match_criteria,
+			 misc_parameters_2.metadata_reg_c_0,
+			 mlx5_eswitch_get_vport_metadata_mask());
 	} else {
 		MLX5_SET(create_flow_group_in, flow_group_in,
 			 match_criteria_enable,
@@ -1241,6 +1330,7 @@ static int esw_create_offloads_table(struct mlx5_eswitch *esw, int nvports)
 	}
 
 	ft_attr.max_fte = nvports + MLX5_ESW_MISS_FLOWS;
+	ft_attr.prio = 1;
 
 	ft_offloads = mlx5_create_flow_table(ns, &ft_attr);
 	if (IS_ERR(ft_offloads)) {
@@ -1318,7 +1408,8 @@ mlx5_eswitch_create_vport_rx_rule(struct mlx5_eswitch *esw, u16 vport,
 			 mlx5_eswitch_get_vport_metadata_for_match(esw, vport));
 
 		misc = MLX5_ADDR_OF(fte_match_param, spec->match_criteria, misc_parameters_2);
-		MLX5_SET_TO_ONES(fte_match_set_misc2, misc, metadata_reg_c_0);
+		MLX5_SET(fte_match_set_misc2, misc, metadata_reg_c_0,
+			 mlx5_eswitch_get_vport_metadata_mask());
 
 		spec->match_criteria_enable = MLX5_MATCH_MISC_PARAMETERS_2;
 	} else {
@@ -1344,6 +1435,7 @@ out:
 	return flow_rule;
 }
 
+
 static int mlx5_eswitch_inline_mode_get(const struct mlx5_eswitch *esw, u8 *mode)
 {
 	u8 prev_mlx5_mode, mlx5_mode = MLX5_INLINE_MODE_L2;
@@ -1379,6 +1471,102 @@ query_vports:
 out:
 	*mode = mlx5_mode;
 	return 0;
+}       
+
+static void esw_destroy_restore_table(struct mlx5_eswitch *esw)
+{
+	struct mlx5_esw_offload *offloads = &esw->offloads;
+
+	mlx5_modify_header_dealloc(esw->dev, offloads->restore_copy_hdr_id);
+	mlx5_destroy_flow_group(offloads->restore_group);
+	mlx5_destroy_flow_table(offloads->ft_offloads_restore);
+}
+
+static int esw_create_restore_table(struct mlx5_eswitch *esw)
+{
+	u8 modact[MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto)] = {};
+	int inlen = MLX5_ST_SZ_BYTES(create_flow_group_in);
+	struct mlx5_flow_table_attr ft_attr = {};
+	struct mlx5_core_dev *dev = esw->dev;
+	struct mlx5_flow_namespace *ns;
+	struct mlx5_modify_hdr *mod_hdr;
+	void *match_criteria, *misc;
+	struct mlx5_flow_table *ft;
+	struct mlx5_flow_group *g;
+	u32 *flow_group_in;
+	int err = 0;
+
+	ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_OFFLOADS);
+	if (!ns) {
+		esw_warn(esw->dev, "Failed to get offloads flow namespace\n");
+		return -EOPNOTSUPP;
+	}
+
+	flow_group_in = kvzalloc(inlen, GFP_KERNEL);
+	if (!flow_group_in) {
+		err = -ENOMEM;
+		goto out_free;
+	}
+
+	ft_attr.max_fte = 1 << ESW_CHAIN_TAG_METADATA_BITS;
+	ft = mlx5_create_flow_table(ns, &ft_attr);
+	if (IS_ERR(ft)) {
+		err = PTR_ERR(ft);
+		esw_warn(esw->dev, "Failed to create restore table, err %d\n",
+			 err);
+		goto out_free;
+	}
+
+	memset(flow_group_in, 0, inlen);
+	match_criteria = MLX5_ADDR_OF(create_flow_group_in, flow_group_in,
+				      match_criteria);
+	misc = MLX5_ADDR_OF(fte_match_param, match_criteria,
+			    misc_parameters_2);
+
+	MLX5_SET(fte_match_set_misc2, misc, metadata_reg_c_0,
+		 ESW_CHAIN_TAG_METADATA_MASK);
+	MLX5_SET(create_flow_group_in, flow_group_in, start_flow_index, 0);
+	MLX5_SET(create_flow_group_in, flow_group_in, end_flow_index,
+		 ft_attr.max_fte - 1);
+	MLX5_SET(create_flow_group_in, flow_group_in, match_criteria_enable,
+		 MLX5_MATCH_MISC_PARAMETERS_2);
+	g = mlx5_create_flow_group(ft, flow_group_in);
+	if (IS_ERR(g)) {
+		err = PTR_ERR(g);
+		esw_warn(dev, "Failed to create restore flow group, err: %d\n",
+			 err);
+		goto err_group;
+	}
+
+	MLX5_SET(copy_action_in, modact, action_type, MLX5_ACTION_TYPE_COPY);
+	MLX5_SET(copy_action_in, modact, src_field,
+		 MLX5_ACTION_IN_FIELD_METADATA_REG_C_1);
+	MLX5_SET(copy_action_in, modact, dst_field,
+		 MLX5_ACTION_IN_FIELD_METADATA_REG_B);
+	mod_hdr = mlx5_modify_header_alloc(esw->dev,
+					   MLX5_FLOW_NAMESPACE_KERNEL, 1,
+					   modact);
+	if (IS_ERR(mod_hdr)) {
+		esw_warn(dev, "Failed to create restore mod header, err: %d\n",
+			 err);
+		err = PTR_ERR(mod_hdr);
+		goto err_mod_hdr;
+	}
+
+	esw->offloads.ft_offloads_restore = ft;
+	esw->offloads.restore_group = g;
+	esw->offloads.restore_copy_hdr_id = mod_hdr;
+
+	return 0;
+
+err_mod_hdr:
+	mlx5_destroy_flow_group(g);
+err_group:
+	mlx5_destroy_flow_table(ft);
+out_free:
+	kvfree(flow_group_in);
+
+	return err;
 }
 
 static int esw_offloads_start(struct mlx5_eswitch *esw,
@@ -1454,187 +1642,66 @@ static void __esw_offloads_unload_rep(struct mlx5_eswitch *esw,
 		esw->offloads.rep_ops[rep_type]->unload(rep);
 }
 
-static void __unload_reps_special_vport(struct mlx5_eswitch *esw, u8 rep_type)
-{
-	struct mlx5_eswitch_rep *rep;
-
-	if (mlx5_ecpf_vport_exists(esw->dev)) {
-		rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_ECPF);
-		__esw_offloads_unload_rep(esw, rep, rep_type);
-	}
-
-	if (mlx5_core_is_ecpf_esw_manager(esw->dev)) {
-		rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_PF);
-		__esw_offloads_unload_rep(esw, rep, rep_type);
-	}
-
-	rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_UPLINK);
-	__esw_offloads_unload_rep(esw, rep, rep_type);
-}
-
-static void __unload_reps_vf_vport(struct mlx5_eswitch *esw, int nvports,
-				   u8 rep_type)
+static void __unload_reps_all_vport(struct mlx5_eswitch *esw, u8 rep_type)
 {
 	struct mlx5_eswitch_rep *rep;
 	int i;
 
-	mlx5_esw_for_each_vf_rep_reverse(esw, i, rep, nvports)
+	mlx5_esw_for_each_vf_rep_reverse(esw, i, rep, esw->esw_funcs.num_vfs)
 		__esw_offloads_unload_rep(esw, rep, rep_type);
-}
-
-static void esw_offloads_unload_vf_reps(struct mlx5_eswitch *esw, int nvports)
-{
-	u8 rep_type = NUM_REP_TYPES;
-
-	while (rep_type-- > 0)
-		__unload_reps_vf_vport(esw, nvports, rep_type);
-}
-
-static void __unload_reps_all_vport(struct mlx5_eswitch *esw, u8 rep_type)
-{
-	__unload_reps_vf_vport(esw, esw->esw_funcs.num_vfs, rep_type);
-
-	/* Special vports must be the last to unload. */
-	__unload_reps_special_vport(esw, rep_type);
-}
-
-static void esw_offloads_unload_all_reps(struct mlx5_eswitch *esw)
-{
-	u8 rep_type = NUM_REP_TYPES;
-
-	while (rep_type-- > 0)
-		__unload_reps_all_vport(esw, rep_type);
-}
-
-static int __esw_offloads_load_rep(struct mlx5_eswitch *esw,
-				   struct mlx5_eswitch_rep *rep, u8 rep_type)
-{
-	int err = 0;
-
-	if (atomic_cmpxchg(&rep->rep_data[rep_type].state,
-			   REP_REGISTERED, REP_LOADED) == REP_REGISTERED) {
-		err = esw->offloads.rep_ops[rep_type]->load(esw->dev, rep);
-		if (err)
-			atomic_set(&rep->rep_data[rep_type].state,
-				   REP_REGISTERED);
-	}
-
-	return err;
-}
-
-static int __load_reps_special_vport(struct mlx5_eswitch *esw, u8 rep_type)
-{
-	struct mlx5_eswitch_rep *rep;
-	int err;
-
-	rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_UPLINK);
-	err = __esw_offloads_load_rep(esw, rep, rep_type);
-	if (err)
-		return err;
-
-	if (mlx5_core_is_ecpf_esw_manager(esw->dev)) {
-		rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_PF);
-		err = __esw_offloads_load_rep(esw, rep, rep_type);
-		if (err)
-			goto err_pf;
-	}
 
 	if (mlx5_ecpf_vport_exists(esw->dev)) {
 		rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_ECPF);
-		err = __esw_offloads_load_rep(esw, rep, rep_type);
-		if (err)
-			goto err_ecpf;
+		__esw_offloads_unload_rep(esw, rep, rep_type);
 	}
 
-	return 0;
-
-err_ecpf:
 	if (mlx5_core_is_ecpf_esw_manager(esw->dev)) {
 		rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_PF);
 		__esw_offloads_unload_rep(esw, rep, rep_type);
 	}
 
-err_pf:
 	rep = mlx5_eswitch_get_rep(esw, MLX5_VPORT_UPLINK);
 	__esw_offloads_unload_rep(esw, rep, rep_type);
-	return err;
 }
 
-static int __load_reps_vf_vport(struct mlx5_eswitch *esw, int nvports,
-				u8 rep_type)
+int esw_offloads_load_rep(struct mlx5_eswitch *esw, u16 vport_num)
 {
 	struct mlx5_eswitch_rep *rep;
-	int err, i;
-
-	mlx5_esw_for_each_vf_rep(esw, i, rep, nvports) {
-		err = __esw_offloads_load_rep(esw, rep, rep_type);
-		if (err)
-			goto err_vf;
-	}
-
-	return 0;
-
-err_vf:
-	__unload_reps_vf_vport(esw, --i, rep_type);
-	return err;
-}
-
-static int __load_reps_all_vport(struct mlx5_eswitch *esw, u8 rep_type)
-{
+	int rep_type;
 	int err;
 
-	/* Special vports must be loaded first, uplink rep creates mdev resource. */
-	err = __load_reps_special_vport(esw, rep_type);
-	if (err)
-		return err;
+	if (esw->mode != MLX5_ESWITCH_OFFLOADS)
+		return 0;
 
-	err = __load_reps_vf_vport(esw, esw->esw_funcs.num_vfs, rep_type);
-	if (err)
-		goto err_vfs;
+	rep = mlx5_eswitch_get_rep(esw, vport_num);
+	for (rep_type = 0; rep_type < NUM_REP_TYPES; rep_type++)
+		if (atomic_cmpxchg(&rep->rep_data[rep_type].state,
+				   REP_REGISTERED, REP_LOADED) == REP_REGISTERED) {
+			err = esw->offloads.rep_ops[rep_type]->load(esw->dev, rep);
+			if (err)
+				goto err_reps;
+		}
 
 	return 0;
 
-err_vfs:
-	__unload_reps_special_vport(esw, rep_type);
-	return err;
-}
-
-static int esw_offloads_load_vf_reps(struct mlx5_eswitch *esw, int nvports)
-{
-	u8 rep_type = 0;
-	int err;
-
-	for (rep_type = 0; rep_type < NUM_REP_TYPES; rep_type++) {
-		err = __load_reps_vf_vport(esw, nvports, rep_type);
-		if (err)
-			goto err_reps;
-	}
-
-	return err;
-
 err_reps:
-	while (rep_type-- > 0)
-		__unload_reps_vf_vport(esw, nvports, rep_type);
+	atomic_set(&rep->rep_data[rep_type].state, REP_REGISTERED);
+	for (--rep_type; rep_type >= 0; rep_type--)
+		__esw_offloads_unload_rep(esw, rep, rep_type);
 	return err;
 }
 
-static int esw_offloads_load_all_reps(struct mlx5_eswitch *esw)
+void esw_offloads_unload_rep(struct mlx5_eswitch *esw, u16 vport_num)
 {
-	u8 rep_type = 0;
-	int err;
-
-	for (rep_type = 0; rep_type < NUM_REP_TYPES; rep_type++) {
-		err = __load_reps_all_vport(esw, rep_type);
-		if (err)
-			goto err_reps;
-	}
+	struct mlx5_eswitch_rep *rep;
+	int rep_type;
 
-	return err;
+	if (esw->mode != MLX5_ESWITCH_OFFLOADS)
+		return;
 
-err_reps:
-	while (rep_type-- > 0)
-		__unload_reps_all_vport(esw, rep_type);
-	return err;
+	rep = mlx5_eswitch_get_rep(esw, vport_num);
+	for (rep_type = NUM_REP_TYPES - 1; rep_type >= 0; rep_type--)
+		__esw_offloads_unload_rep(esw, rep, rep_type);
 }
 
 #define ESW_OFFLOADS_DEVCOM_PAIR	(0)
@@ -1822,14 +1889,21 @@ static int esw_vport_add_ingress_acl_modify_metadata(struct mlx5_eswitch *esw,
 						     struct mlx5_vport *vport)
 {
 	u8 action[MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto)] = {};
-	static const struct mlx5_flow_spec spec = {};
 	struct mlx5_flow_act flow_act = {};
 	int err = 0;
+	u32 key;
+
+	key = mlx5_eswitch_get_vport_metadata_for_match(esw, vport->vport);
+	key >>= ESW_SOURCE_PORT_METADATA_OFFSET;
 
 	MLX5_SET(set_action_in, action, action_type, MLX5_ACTION_TYPE_SET);
-	MLX5_SET(set_action_in, action, field, MLX5_ACTION_IN_FIELD_METADATA_REG_C_0);
-	MLX5_SET(set_action_in, action, data,
-		 mlx5_eswitch_get_vport_metadata_for_match(esw, vport->vport));
+	MLX5_SET(set_action_in, action, field,
+		 MLX5_ACTION_IN_FIELD_METADATA_REG_C_0);
+	MLX5_SET(set_action_in, action, data, key);
+	MLX5_SET(set_action_in, action, offset,
+		 ESW_SOURCE_PORT_METADATA_OFFSET);
+	MLX5_SET(set_action_in, action, length,
+		 ESW_SOURCE_PORT_METADATA_BITS);
 
 	vport->ingress.offloads.modify_metadata =
 		mlx5_modify_header_alloc(esw->dev, MLX5_FLOW_NAMESPACE_ESW_INGRESS,
@@ -1846,7 +1920,7 @@ static int esw_vport_add_ingress_acl_modify_metadata(struct mlx5_eswitch *esw,
 	flow_act.modify_hdr = vport->ingress.offloads.modify_metadata;
 	vport->ingress.offloads.modify_metadata_rule =
 				mlx5_add_flow_rules(vport->ingress.acl,
-						    &spec, &flow_act, NULL, 0);
+						    NULL, &flow_act, NULL, 0);
 	if (IS_ERR(vport->ingress.offloads.modify_metadata_rule)) {
 		err = PTR_ERR(vport->ingress.offloads.modify_metadata_rule);
 		esw_warn(esw->dev,
@@ -2144,13 +2218,17 @@ static int esw_offloads_steering_init(struct mlx5_eswitch *esw)
 	if (err)
 		return err;
 
-	err = esw_create_offloads_fdb_tables(esw, total_vports);
+	err = esw_create_offloads_table(esw, total_vports);
 	if (err)
-		goto create_fdb_err;
+		goto create_offloads_err;
 
-	err = esw_create_offloads_table(esw, total_vports);
+	err = esw_create_restore_table(esw);
 	if (err)
-		goto create_ft_err;
+		goto create_restore_err;
+
+	err = esw_create_offloads_fdb_tables(esw, total_vports);
+	if (err)
+		goto create_fdb_err;
 
 	err = esw_create_vport_rx_group(esw, total_vports);
 	if (err)
@@ -2162,12 +2240,12 @@ static int esw_offloads_steering_init(struct mlx5_eswitch *esw)
 	return 0;
 
 create_fg_err:
-	esw_destroy_offloads_table(esw);
-
-create_ft_err:
 	esw_destroy_offloads_fdb_tables(esw);
-
 create_fdb_err:
+	esw_destroy_restore_table(esw);
+create_restore_err:
+	esw_destroy_offloads_table(esw);
+create_offloads_err:
 	esw_destroy_uplink_offloads_acl_tables(esw);
 
 	return err;
@@ -2177,8 +2255,9 @@ static void esw_offloads_steering_cleanup(struct mlx5_eswitch *esw)
 {
 	mutex_destroy(&esw->fdb_table.offloads.vports.lock);
 	esw_destroy_vport_rx_group(esw);
-	esw_destroy_offloads_table(esw);
 	esw_destroy_offloads_fdb_tables(esw);
+	esw_destroy_restore_table(esw);
+	esw_destroy_offloads_table(esw);
 	esw_destroy_uplink_offloads_acl_tables(esw);
 }
 
@@ -2198,11 +2277,12 @@ esw_vfs_changed_event_handler(struct mlx5_eswitch *esw, const u32 *out)
 
 	/* Number of VFs can only change from "0 to x" or "x to 0". */
 	if (esw->esw_funcs.num_vfs > 0) {
-		esw_offloads_unload_vf_reps(esw, esw->esw_funcs.num_vfs);
+		mlx5_eswitch_unload_vf_vports(esw, esw->esw_funcs.num_vfs);
 	} else {
 		int err;
 
-		err = esw_offloads_load_vf_reps(esw, new_num_vfs);
+		err = mlx5_eswitch_load_vf_vports(esw, new_num_vfs,
+						  MLX5_VPORT_UC_ADDR_CHANGE);
 		if (err)
 			return;
 	}
@@ -2260,6 +2340,7 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)
 	else
 		esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
 
+	mutex_init(&esw->offloads.termtbl_mutex);
 	mlx5_rdma_enable_roce(esw->dev);
 	err = esw_offloads_steering_init(esw);
 	if (err)
@@ -2273,27 +2354,28 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)
 	mlx5_esw_for_each_vf_vport(esw, i, vport, esw->esw_funcs.num_vfs)
 		vport->info.link_state = MLX5_VPORT_ADMIN_STATE_DOWN;
 
-	err = mlx5_eswitch_enable_pf_vf_vports(esw, MLX5_VPORT_UC_ADDR_CHANGE);
+	/* Uplink vport rep must load first. */
+	err = esw_offloads_load_rep(esw, MLX5_VPORT_UPLINK);
 	if (err)
-		goto err_vports;
+		goto err_uplink;
 
-	err = esw_offloads_load_all_reps(esw);
+	err = mlx5_eswitch_enable_pf_vf_vports(esw, MLX5_VPORT_UC_ADDR_CHANGE);
 	if (err)
-		goto err_reps;
+		goto err_vports;
 
 	esw_offloads_devcom_init(esw);
-	mutex_init(&esw->offloads.termtbl_mutex);
 
 	return 0;
 
-err_reps:
-	mlx5_eswitch_disable_pf_vf_vports(esw);
 err_vports:
+	esw_offloads_unload_rep(esw, MLX5_VPORT_UPLINK);
+err_uplink:
 	esw_set_passing_vport_metadata(esw, false);
 err_vport_metadata:
 	esw_offloads_steering_cleanup(esw);
 err_steering_init:
 	mlx5_rdma_disable_roce(esw->dev);
+	mutex_destroy(&esw->offloads.termtbl_mutex);
 	return err;
 }
 
@@ -2319,11 +2401,12 @@ static int esw_offloads_stop(struct mlx5_eswitch *esw,
 void esw_offloads_disable(struct mlx5_eswitch *esw)
 {
 	esw_offloads_devcom_cleanup(esw);
-	esw_offloads_unload_all_reps(esw);
 	mlx5_eswitch_disable_pf_vf_vports(esw);
+	esw_offloads_unload_rep(esw, MLX5_VPORT_UPLINK);
 	esw_set_passing_vport_metadata(esw, false);
 	esw_offloads_steering_cleanup(esw);
 	mlx5_rdma_disable_roce(esw->dev);
+	mutex_destroy(&esw->offloads.termtbl_mutex);
 	esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
 }
 
@@ -2403,10 +2486,8 @@ static int esw_inline_mode_to_devlink(u8 mlx5_mode, u8 *mode)
 	return 0;
 }
 
-static int mlx5_devlink_eswitch_check(struct devlink *devlink)
+static int mlx5_eswitch_check(const struct mlx5_core_dev *dev)
 {
-	struct mlx5_core_dev *dev = devlink_priv(devlink);
-
 	if (MLX5_CAP_GEN(dev, port_type) != MLX5_CAP_PORT_TYPE_ETH)
 		return -EOPNOTSUPP;
 
@@ -2427,7 +2508,7 @@ int mlx5_devlink_eswitch_mode_set(struct devlink *devlink, u16 mode,
 	u16 cur_mlx5_mode, mlx5_mode = 0;
 	int err;
 
-	err = mlx5_devlink_eswitch_check(devlink);
+	err = mlx5_eswitch_check(dev);
 	if (err)
 		return err;
 
@@ -2452,7 +2533,7 @@ int mlx5_devlink_eswitch_mode_get(struct devlink *devlink, u16 *mode)
 	struct mlx5_core_dev *dev = devlink_priv(devlink);
 	int err;
 
-	err = mlx5_devlink_eswitch_check(devlink);
+	err = mlx5_eswitch_check(dev);
 	if (err)
 		return err;
 
@@ -2467,7 +2548,7 @@ int mlx5_devlink_eswitch_inline_mode_set(struct devlink *devlink, u8 mode,
 	int err, vport, num_vport;
 	u8 mlx5_mode;
 
-	err = mlx5_devlink_eswitch_check(devlink);
+	err = mlx5_eswitch_check(dev);
 	if (err)
 		return err;
 
@@ -2521,7 +2602,7 @@ int mlx5_devlink_eswitch_inline_mode_get(struct devlink *devlink, u8 *mode)
 	struct mlx5_eswitch *esw = dev->priv.eswitch;
 	int err;
 
-	err = mlx5_devlink_eswitch_check(devlink);
+	err = mlx5_eswitch_check(dev);
 	if (err)
 		return err;
 
@@ -2536,7 +2617,7 @@ int mlx5_devlink_eswitch_encap_mode_set(struct devlink *devlink,
 	struct mlx5_eswitch *esw = dev->priv.eswitch;
 	int err;
 
-	err = mlx5_devlink_eswitch_check(devlink);
+	err = mlx5_eswitch_check(dev);
 	if (err)
 		return err;
 
@@ -2585,7 +2666,7 @@ int mlx5_devlink_eswitch_encap_mode_get(struct devlink *devlink,
 	struct mlx5_eswitch *esw = dev->priv.eswitch;
 	int err;
 
-	err = mlx5_devlink_eswitch_check(devlink);
+	err = mlx5_eswitch_check(dev);
 	if (err)
 		return err;
 
@@ -2593,6 +2674,21 @@ int mlx5_devlink_eswitch_encap_mode_get(struct devlink *devlink,
 	return 0;
 }
 
+static bool
+mlx5_eswitch_vport_has_rep(const struct mlx5_eswitch *esw, u16 vport_num)
+{
+	/* Currently, only ECPF based device has representor for host PF. */
+	if (vport_num == MLX5_VPORT_PF &&
+	    !mlx5_core_is_ecpf_esw_manager(esw->dev))
+		return false;
+
+	if (vport_num == MLX5_VPORT_ECPF &&
+	    !mlx5_ecpf_vport_exists(esw->dev))
+		return false;
+
+	return true;
+}
+
 void mlx5_eswitch_register_vport_reps(struct mlx5_eswitch *esw,
 				      const struct mlx5_eswitch_rep_ops *ops,
 				      u8 rep_type)
@@ -2603,8 +2699,10 @@ void mlx5_eswitch_register_vport_reps(struct mlx5_eswitch *esw,
 
 	esw->offloads.rep_ops[rep_type] = ops;
 	mlx5_esw_for_all_reps(esw, i, rep) {
-		rep_data = &rep->rep_data[rep_type];
-		atomic_set(&rep_data->state, REP_REGISTERED);
+		if (likely(mlx5_eswitch_vport_has_rep(esw, i))) {
+			rep_data = &rep->rep_data[rep_type];
+			atomic_set(&rep_data->state, REP_REGISTERED);
+		}
 	}
 }
 EXPORT_SYMBOL(mlx5_eswitch_register_vport_reps);
@@ -2664,15 +2762,53 @@ bool mlx5_eswitch_is_vf_vport(const struct mlx5_eswitch *esw, u16 vport_num)
 	       vport_num <= esw->dev->priv.sriov.max_vfs;
 }
 
+bool mlx5_eswitch_reg_c1_loopback_enabled(const struct mlx5_eswitch *esw)
+{
+	return !!(esw->flags & MLX5_ESWITCH_REG_C1_LOOPBACK_ENABLED);
+}
+EXPORT_SYMBOL(mlx5_eswitch_reg_c1_loopback_enabled);
+
 bool mlx5_eswitch_vport_match_metadata_enabled(const struct mlx5_eswitch *esw)
 {
 	return !!(esw->flags & MLX5_ESWITCH_VPORT_MATCH_METADATA);
 }
 EXPORT_SYMBOL(mlx5_eswitch_vport_match_metadata_enabled);
 
-u32 mlx5_eswitch_get_vport_metadata_for_match(const struct mlx5_eswitch *esw,
+u32 mlx5_eswitch_get_vport_metadata_for_match(struct mlx5_eswitch *esw,
 					      u16 vport_num)
 {
-	return ((MLX5_CAP_GEN(esw->dev, vhca_id) & 0xffff) << 16) | vport_num;
+	u32 vport_num_mask = GENMASK(ESW_VPORT_BITS - 1, 0);
+	u32 vhca_id_mask = GENMASK(ESW_VHCA_ID_BITS - 1, 0);
+	u32 vhca_id = MLX5_CAP_GEN(esw->dev, vhca_id);
+	u32 val;
+
+	/* Make sure the vhca_id fits the ESW_VHCA_ID_BITS */
+	WARN_ON_ONCE(vhca_id >= BIT(ESW_VHCA_ID_BITS));
+
+	/* Trim vhca_id to ESW_VHCA_ID_BITS */
+	vhca_id &= vhca_id_mask;
+
+	/* Make sure pf and ecpf map to end of ESW_VPORT_BITS range so they
+	 * don't overlap with VF numbers, and themselves, after trimming.
+	 */
+	WARN_ON_ONCE((MLX5_VPORT_UPLINK & vport_num_mask) <
+		     vport_num_mask - 1);
+	WARN_ON_ONCE((MLX5_VPORT_ECPF & vport_num_mask) <
+		     vport_num_mask - 1);
+	WARN_ON_ONCE((MLX5_VPORT_UPLINK & vport_num_mask) ==
+		     (MLX5_VPORT_ECPF & vport_num_mask));
+
+	/* Make sure that the VF vport_num fits ESW_VPORT_BITS and don't
+	 * overlap with pf and ecpf.
+	 */
+	if (vport_num != MLX5_VPORT_UPLINK &&
+	    vport_num != MLX5_VPORT_ECPF)
+		WARN_ON_ONCE(vport_num >= vport_num_mask - 1);
+
+	/* We can now trim vport_num to ESW_VPORT_BITS */
+	vport_num &= vport_num_mask;
+
+	val = (vhca_id << ESW_VPORT_BITS) | vport_num;
+	return val << (32 - ESW_SOURCE_PORT_METADATA_BITS);
 }
 EXPORT_SYMBOL(mlx5_eswitch_get_vport_metadata_for_match);