Merge branch 'nfp-flower-ct-offload'

Simon Horman says:

====================
nfp: flower: conntrack offload

Louis Peens says:

This series takes the preparation from previous two series
and finally creates the structures and control messages
to offload the conntrack flows to the card. First we
do a bit of refactoring in the existing functions
to make them re-usable for the conntrack implementation,
after which the control messages are compiled and
transmitted to the card. Lastly we add stats handling
for the conntrack flows.
====================

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

7 files changed, 946 insertions, 201 deletions

diff --git a/drivers/net/ethernet/netronome/nfp/flower/action.c b/drivers/net/ethernet/netronome/nfp/flower/action.c
index 1cbe2c9f3959..2a432de11858 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/action.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/action.c
@@ -262,10 +262,10 @@ nfp_fl_output(struct nfp_app *app, struct nfp_fl_output *output,
 }
 
 static bool
-nfp_flower_tun_is_gre(struct flow_cls_offload *flow, int start_idx)
+nfp_flower_tun_is_gre(struct flow_rule *rule, int start_idx)
 {
-	struct flow_action_entry *act = flow->rule->action.entries;
-	int num_act = flow->rule->action.num_entries;
+	struct flow_action_entry *act = rule->action.entries;
+	int num_act = rule->action.num_entries;
 	int act_idx;
 
 	/* Preparse action list for next mirred or redirect action */
@@ -279,7 +279,7 @@ nfp_flower_tun_is_gre(struct flow_cls_offload *flow, int start_idx)
 
 static enum nfp_flower_tun_type
 nfp_fl_get_tun_from_act(struct nfp_app *app,
-			struct flow_cls_offload *flow,
+			struct flow_rule *rule,
 			const struct flow_action_entry *act, int act_idx)
 {
 	const struct ip_tunnel_info *tun = act->tunnel;
@@ -288,7 +288,7 @@ nfp_fl_get_tun_from_act(struct nfp_app *app,
 	/* Determine the tunnel type based on the egress netdev
 	 * in the mirred action for tunnels without l4.
 	 */
-	if (nfp_flower_tun_is_gre(flow, act_idx))
+	if (nfp_flower_tun_is_gre(rule, act_idx))
 		return NFP_FL_TUNNEL_GRE;
 
 	switch (tun->key.tp_dst) {
@@ -788,11 +788,10 @@ struct nfp_flower_pedit_acts {
 };
 
 static int
-nfp_fl_commit_mangle(struct flow_cls_offload *flow, char *nfp_action,
+nfp_fl_commit_mangle(struct flow_rule *rule, char *nfp_action,
 		     int *a_len, struct nfp_flower_pedit_acts *set_act,
 		     u32 *csum_updated)
 {
-	struct flow_rule *rule = flow_cls_offload_flow_rule(flow);
 	size_t act_size = 0;
 	u8 ip_proto = 0;
 
@@ -890,7 +889,7 @@ nfp_fl_commit_mangle(struct flow_cls_offload *flow, char *nfp_action,
 
 static int
 nfp_fl_pedit(const struct flow_action_entry *act,
-	     struct flow_cls_offload *flow, char *nfp_action, int *a_len,
+	     char *nfp_action, int *a_len,
 	     u32 *csum_updated, struct nfp_flower_pedit_acts *set_act,
 	     struct netlink_ext_ack *extack)
 {
@@ -977,7 +976,7 @@ nfp_flower_output_action(struct nfp_app *app,
 
 static int
 nfp_flower_loop_action(struct nfp_app *app, const struct flow_action_entry *act,
-		       struct flow_cls_offload *flow,
+		       struct flow_rule *rule,
 		       struct nfp_fl_payload *nfp_fl, int *a_len,
 		       struct net_device *netdev,
 		       enum nfp_flower_tun_type *tun_type, int *tun_out_cnt,
@@ -1045,7 +1044,7 @@ nfp_flower_loop_action(struct nfp_app *app, const struct flow_action_entry *act,
 	case FLOW_ACTION_TUNNEL_ENCAP: {
 		const struct ip_tunnel_info *ip_tun = act->tunnel;
 
-		*tun_type = nfp_fl_get_tun_from_act(app, flow, act, act_idx);
+		*tun_type = nfp_fl_get_tun_from_act(app, rule, act, act_idx);
 		if (*tun_type == NFP_FL_TUNNEL_NONE) {
 			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: unsupported tunnel type in action list");
 			return -EOPNOTSUPP;
@@ -1086,7 +1085,7 @@ nfp_flower_loop_action(struct nfp_app *app, const struct flow_action_entry *act,
 		/* Tunnel decap is handled by default so accept action. */
 		return 0;
 	case FLOW_ACTION_MANGLE:
-		if (nfp_fl_pedit(act, flow, &nfp_fl->action_data[*a_len],
+		if (nfp_fl_pedit(act, &nfp_fl->action_data[*a_len],
 				 a_len, csum_updated, set_act, extack))
 			return -EOPNOTSUPP;
 		break;
@@ -1195,7 +1194,7 @@ static bool nfp_fl_check_mangle_end(struct flow_action *flow_act,
 }
 
 int nfp_flower_compile_action(struct nfp_app *app,
-			      struct flow_cls_offload *flow,
+			      struct flow_rule *rule,
 			      struct net_device *netdev,
 			      struct nfp_fl_payload *nfp_flow,
 			      struct netlink_ext_ack *extack)
@@ -1207,7 +1206,7 @@ int nfp_flower_compile_action(struct nfp_app *app,
 	bool pkt_host = false;
 	u32 csum_updated = 0;
 
-	if (!flow_action_hw_stats_check(&flow->rule->action, extack,
+	if (!flow_action_hw_stats_check(&rule->action, extack,
 					FLOW_ACTION_HW_STATS_DELAYED_BIT))
 		return -EOPNOTSUPP;
 
@@ -1219,18 +1218,18 @@ int nfp_flower_compile_action(struct nfp_app *app,
 	tun_out_cnt = 0;
 	out_cnt = 0;
 
-	flow_action_for_each(i, act, &flow->rule->action) {
-		if (nfp_fl_check_mangle_start(&flow->rule->action, i))
+	flow_action_for_each(i, act, &rule->action) {
+		if (nfp_fl_check_mangle_start(&rule->action, i))
 			memset(&set_act, 0, sizeof(set_act));
-		err = nfp_flower_loop_action(app, act, flow, nfp_flow, &act_len,
+		err = nfp_flower_loop_action(app, act, rule, nfp_flow, &act_len,
 					     netdev, &tun_type, &tun_out_cnt,
 					     &out_cnt, &csum_updated,
 					     &set_act, &pkt_host, extack, i);
 		if (err)
 			return err;
 		act_cnt++;
-		if (nfp_fl_check_mangle_end(&flow->rule->action, i))
-			nfp_fl_commit_mangle(flow,
+		if (nfp_fl_check_mangle_end(&rule->action, i))
+			nfp_fl_commit_mangle(rule,
 					     &nfp_flow->action_data[act_len],
 					     &act_len, &set_act, &csum_updated);
 	}
diff --git a/drivers/net/ethernet/netronome/nfp/flower/conntrack.c b/drivers/net/ethernet/netronome/nfp/flower/conntrack.c
index 128020b1573e..2abf02eed7fb 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/conntrack.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/conntrack.c
@@ -2,6 +2,7 @@
 /* Copyright (C) 2021 Corigine, Inc. */
 
 #include "conntrack.h"
+#include "../nfp_port.h"
 
 const struct rhashtable_params nfp_tc_ct_merge_params = {
 	.head_offset		= offsetof(struct nfp_fl_ct_tc_merge,
@@ -407,15 +408,487 @@ static int nfp_ct_check_meta(struct nfp_fl_ct_flow_entry *post_ct_entry,
 	return -EINVAL;
 }
 
+static int
+nfp_fl_calc_key_layers_sz(struct nfp_fl_key_ls in_key_ls, uint16_t *map)
+{
+	int key_size;
+
+	/* This field must always be present */
+	key_size = sizeof(struct nfp_flower_meta_tci);
+	map[FLOW_PAY_META_TCI] = 0;
+
+	if (in_key_ls.key_layer & NFP_FLOWER_LAYER_EXT_META) {
+		map[FLOW_PAY_EXT_META] = key_size;
+		key_size += sizeof(struct nfp_flower_ext_meta);
+	}
+	if (in_key_ls.key_layer & NFP_FLOWER_LAYER_PORT) {
+		map[FLOW_PAY_INPORT] = key_size;
+		key_size += sizeof(struct nfp_flower_in_port);
+	}
+	if (in_key_ls.key_layer & NFP_FLOWER_LAYER_MAC) {
+		map[FLOW_PAY_MAC_MPLS] = key_size;
+		key_size += sizeof(struct nfp_flower_mac_mpls);
+	}
+	if (in_key_ls.key_layer & NFP_FLOWER_LAYER_TP) {
+		map[FLOW_PAY_L4] = key_size;
+		key_size += sizeof(struct nfp_flower_tp_ports);
+	}
+	if (in_key_ls.key_layer & NFP_FLOWER_LAYER_IPV4) {
+		map[FLOW_PAY_IPV4] = key_size;
+		key_size += sizeof(struct nfp_flower_ipv4);
+	}
+	if (in_key_ls.key_layer & NFP_FLOWER_LAYER_IPV6) {
+		map[FLOW_PAY_IPV6] = key_size;
+		key_size += sizeof(struct nfp_flower_ipv6);
+	}
+
+	if (in_key_ls.key_layer_two & NFP_FLOWER_LAYER2_GRE) {
+		map[FLOW_PAY_GRE] = key_size;
+		if (in_key_ls.key_layer_two & NFP_FLOWER_LAYER2_TUN_IPV6)
+			key_size += sizeof(struct nfp_flower_ipv6_gre_tun);
+		else
+			key_size += sizeof(struct nfp_flower_ipv4_gre_tun);
+	}
+
+	if (in_key_ls.key_layer_two & NFP_FLOWER_LAYER2_QINQ) {
+		map[FLOW_PAY_QINQ] = key_size;
+		key_size += sizeof(struct nfp_flower_vlan);
+	}
+
+	if ((in_key_ls.key_layer & NFP_FLOWER_LAYER_VXLAN) ||
+	    (in_key_ls.key_layer_two & NFP_FLOWER_LAYER2_GENEVE)) {
+		map[FLOW_PAY_UDP_TUN] = key_size;
+		if (in_key_ls.key_layer_two & NFP_FLOWER_LAYER2_TUN_IPV6)
+			key_size += sizeof(struct nfp_flower_ipv6_udp_tun);
+		else
+			key_size += sizeof(struct nfp_flower_ipv4_udp_tun);
+	}
+
+	if (in_key_ls.key_layer_two & NFP_FLOWER_LAYER2_GENEVE_OP) {
+		map[FLOW_PAY_GENEVE_OPT] = key_size;
+		key_size += sizeof(struct nfp_flower_geneve_options);
+	}
+
+	return key_size;
+}
+
+static int nfp_fl_merge_actions_offload(struct flow_rule **rules,
+					struct nfp_flower_priv *priv,
+					struct net_device *netdev,
+					struct nfp_fl_payload *flow_pay)
+{
+	struct flow_action_entry *a_in;
+	int i, j, num_actions, id;
+	struct flow_rule *a_rule;
+	int err = 0, offset = 0;
+
+	num_actions = rules[CT_TYPE_PRE_CT]->action.num_entries +
+		      rules[CT_TYPE_NFT]->action.num_entries +
+		      rules[CT_TYPE_POST_CT]->action.num_entries;
+
+	a_rule = flow_rule_alloc(num_actions);
+	if (!a_rule)
+		return -ENOMEM;
+
+	/* Actions need a BASIC dissector. */
+	a_rule->match = rules[CT_TYPE_PRE_CT]->match;
+
+	/* Copy actions */
+	for (j = 0; j < _CT_TYPE_MAX; j++) {
+		if (flow_rule_match_key(rules[j], FLOW_DISSECTOR_KEY_BASIC)) {
+			struct flow_match_basic match;
+
+			/* ip_proto is the only field that needed in later compile_action,
+			 * needed to set the correct checksum flags. It doesn't really matter
+			 * which input rule's ip_proto field we take as the earlier merge checks
+			 * would have made sure that they don't conflict. We do not know which
+			 * of the subflows would have the ip_proto filled in, so we need to iterate
+			 * through the subflows and assign the proper subflow to a_rule
+			 */
+			flow_rule_match_basic(rules[j], &match);
+			if (match.mask->ip_proto)
+				a_rule->match = rules[j]->match;
+		}
+
+		for (i = 0; i < rules[j]->action.num_entries; i++) {
+			a_in = &rules[j]->action.entries[i];
+			id = a_in->id;
+
+			/* Ignore CT related actions as these would already have
+			 * been taken care of by previous checks, and we do not send
+			 * any CT actions to the firmware.
+			 */
+			switch (id) {
+			case FLOW_ACTION_CT:
+			case FLOW_ACTION_GOTO:
+			case FLOW_ACTION_CT_METADATA:
+				continue;
+			default:
+				memcpy(&a_rule->action.entries[offset++],
+				       a_in, sizeof(struct flow_action_entry));
+				break;
+			}
+		}
+	}
+
+	/* Some actions would have been ignored, so update the num_entries field */
+	a_rule->action.num_entries = offset;
+	err = nfp_flower_compile_action(priv->app, a_rule, netdev, flow_pay, NULL);
+	kfree(a_rule);
+
+	return err;
+}
+
 static int nfp_fl_ct_add_offload(struct nfp_fl_nft_tc_merge *m_entry)
 {
-	return 0;
+	enum nfp_flower_tun_type tun_type = NFP_FL_TUNNEL_NONE;
+	struct nfp_fl_ct_zone_entry *zt = m_entry->zt;
+	struct nfp_fl_key_ls key_layer, tmp_layer;
+	struct nfp_flower_priv *priv = zt->priv;
+	u16 key_map[_FLOW_PAY_LAYERS_MAX];
+	struct nfp_fl_payload *flow_pay;
+
+	struct flow_rule *rules[_CT_TYPE_MAX];
+	u8 *key, *msk, *kdata, *mdata;
+	struct nfp_port *port = NULL;
+	struct net_device *netdev;
+	bool qinq_sup;
+	u32 port_id;
+	u16 offset;
+	int i, err;
+
+	netdev = m_entry->netdev;
+	qinq_sup = !!(priv->flower_ext_feats & NFP_FL_FEATS_VLAN_QINQ);
+
+	rules[CT_TYPE_PRE_CT] = m_entry->tc_m_parent->pre_ct_parent->rule;
+	rules[CT_TYPE_NFT] = m_entry->nft_parent->rule;
+	rules[CT_TYPE_POST_CT] = m_entry->tc_m_parent->post_ct_parent->rule;
+
+	memset(&key_layer, 0, sizeof(struct nfp_fl_key_ls));
+	memset(&key_map, 0, sizeof(key_map));
+
+	/* Calculate the resultant key layer and size for offload */
+	for (i = 0; i < _CT_TYPE_MAX; i++) {
+		err = nfp_flower_calculate_key_layers(priv->app,
+						      m_entry->netdev,
+						      &tmp_layer, rules[i],
+						      &tun_type, NULL);
+		if (err)
+			return err;
+
+		key_layer.key_layer |= tmp_layer.key_layer;
+		key_layer.key_layer_two |= tmp_layer.key_layer_two;
+	}
+	key_layer.key_size = nfp_fl_calc_key_layers_sz(key_layer, key_map);
+
+	flow_pay = nfp_flower_allocate_new(&key_layer);
+	if (!flow_pay)
+		return -ENOMEM;
+
+	memset(flow_pay->unmasked_data, 0, key_layer.key_size);
+	memset(flow_pay->mask_data, 0, key_layer.key_size);
+
+	kdata = flow_pay->unmasked_data;
+	mdata = flow_pay->mask_data;
+
+	offset = key_map[FLOW_PAY_META_TCI];
+	key = kdata + offset;
+	msk = mdata + offset;
+	nfp_flower_compile_meta((struct nfp_flower_meta_tci *)key,
+				(struct nfp_flower_meta_tci *)msk,
+				key_layer.key_layer);
+
+	if (NFP_FLOWER_LAYER_EXT_META & key_layer.key_layer) {
+		offset =  key_map[FLOW_PAY_EXT_META];
+		key = kdata + offset;
+		msk = mdata + offset;
+		nfp_flower_compile_ext_meta((struct nfp_flower_ext_meta *)key,
+					    key_layer.key_layer_two);
+		nfp_flower_compile_ext_meta((struct nfp_flower_ext_meta *)msk,
+					    key_layer.key_layer_two);
+	}
+
+	/* Using in_port from the -trk rule. The tc merge checks should already
+	 * be checking that the ingress netdevs are the same
+	 */
+	port_id = nfp_flower_get_port_id_from_netdev(priv->app, netdev);
+	offset = key_map[FLOW_PAY_INPORT];
+	key = kdata + offset;
+	msk = mdata + offset;
+	err = nfp_flower_compile_port((struct nfp_flower_in_port *)key,
+				      port_id, false, tun_type, NULL);
+	if (err)
+		goto ct_offload_err;
+	err = nfp_flower_compile_port((struct nfp_flower_in_port *)msk,
+				      port_id, true, tun_type, NULL);
+	if (err)
+		goto ct_offload_err;
+
+	/* This following part works on the assumption that previous checks has
+	 * already filtered out flows that has different values for the different
+	 * layers. Here we iterate through all three rules and merge their respective
+	 * masked value(cared bits), basic method is:
+	 * final_key = (r1_key & r1_mask) | (r2_key & r2_mask) | (r3_key & r3_mask)
+	 * final_mask = r1_mask | r2_mask | r3_mask
+	 * If none of the rules contains a match that is also fine, that simply means
+	 * that the layer is not present.
+	 */
+	if (!qinq_sup) {
+		for (i = 0; i < _CT_TYPE_MAX; i++) {
+			offset = key_map[FLOW_PAY_META_TCI];
+			key = kdata + offset;
+			msk = mdata + offset;
+			nfp_flower_compile_tci((struct nfp_flower_meta_tci *)key,
+					       (struct nfp_flower_meta_tci *)msk,
+					       rules[i]);
+		}
+	}
+
+	if (NFP_FLOWER_LAYER_MAC & key_layer.key_layer) {
+		offset = key_map[FLOW_PAY_MAC_MPLS];
+		key = kdata + offset;
+		msk = mdata + offset;
+		for (i = 0; i < _CT_TYPE_MAX; i++) {
+			nfp_flower_compile_mac((struct nfp_flower_mac_mpls *)key,
+					       (struct nfp_flower_mac_mpls *)msk,
+					       rules[i]);
+			err = nfp_flower_compile_mpls((struct nfp_flower_mac_mpls *)key,
+						      (struct nfp_flower_mac_mpls *)msk,
+						      rules[i], NULL);
+			if (err)
+				goto ct_offload_err;
+		}
+	}
+
+	if (NFP_FLOWER_LAYER_IPV4 & key_layer.key_layer) {
+		offset = key_map[FLOW_PAY_IPV4];
+		key = kdata + offset;
+		msk = mdata + offset;
+		for (i = 0; i < _CT_TYPE_MAX; i++) {
+			nfp_flower_compile_ipv4((struct nfp_flower_ipv4 *)key,
+						(struct nfp_flower_ipv4 *)msk,
+						rules[i]);
+		}
+	}
+
+	if (NFP_FLOWER_LAYER_IPV6 & key_layer.key_layer) {
+		offset = key_map[FLOW_PAY_IPV6];
+		key = kdata + offset;
+		msk = mdata + offset;
+		for (i = 0; i < _CT_TYPE_MAX; i++) {
+			nfp_flower_compile_ipv6((struct nfp_flower_ipv6 *)key,
+						(struct nfp_flower_ipv6 *)msk,
+						rules[i]);
+		}
+	}
+
+	if (NFP_FLOWER_LAYER_TP & key_layer.key_layer) {
+		offset = key_map[FLOW_PAY_L4];
+		key = kdata + offset;
+		msk = mdata + offset;
+		for (i = 0; i < _CT_TYPE_MAX; i++) {
+			nfp_flower_compile_tport((struct nfp_flower_tp_ports *)key,
+						 (struct nfp_flower_tp_ports *)msk,
+						 rules[i]);
+		}
+	}
+
+	if (key_layer.key_layer_two & NFP_FLOWER_LAYER2_GRE) {
+		offset = key_map[FLOW_PAY_GRE];
+		key = kdata + offset;
+		msk = mdata + offset;
+		if (key_layer.key_layer_two & NFP_FLOWER_LAYER2_TUN_IPV6) {
+			struct nfp_flower_ipv6_gre_tun *gre_match;
+			struct nfp_ipv6_addr_entry *entry;
+			struct in6_addr *dst;
+
+			for (i = 0; i < _CT_TYPE_MAX; i++) {
+				nfp_flower_compile_ipv6_gre_tun((void *)key,
+								(void *)msk, rules[i]);
+			}
+			gre_match = (struct nfp_flower_ipv6_gre_tun *)key;
+			dst = &gre_match->ipv6.dst;
+
+			entry = nfp_tunnel_add_ipv6_off(priv->app, dst);
+			if (!entry)
+				goto ct_offload_err;
+
+			flow_pay->nfp_tun_ipv6 = entry;
+		} else {
+			__be32 dst;
+
+			for (i = 0; i < _CT_TYPE_MAX; i++) {
+				nfp_flower_compile_ipv4_gre_tun((void *)key,
+								(void *)msk, rules[i]);
+			}
+			dst = ((struct nfp_flower_ipv4_gre_tun *)key)->ipv4.dst;
+
+			/* Store the tunnel destination in the rule data.
+			 * This must be present and be an exact match.
+			 */
+			flow_pay->nfp_tun_ipv4_addr = dst;
+			nfp_tunnel_add_ipv4_off(priv->app, dst);
+		}
+	}
+
+	if (NFP_FLOWER_LAYER2_QINQ & key_layer.key_layer_two) {
+		offset = key_map[FLOW_PAY_QINQ];
+		key = kdata + offset;
+		msk = mdata + offset;
+		for (i = 0; i < _CT_TYPE_MAX; i++) {
+			nfp_flower_compile_vlan((struct nfp_flower_vlan *)key,
+						(struct nfp_flower_vlan *)msk,
+						rules[i]);
+		}
+	}
+
+	if (key_layer.key_layer & NFP_FLOWER_LAYER_VXLAN ||
+	    key_layer.key_layer_two & NFP_FLOWER_LAYER2_GENEVE) {
+		offset = key_map[FLOW_PAY_UDP_TUN];
+		key = kdata + offset;
+		msk = mdata + offset;
+		if (key_layer.key_layer_two & NFP_FLOWER_LAYER2_TUN_IPV6) {
+			struct nfp_flower_ipv6_udp_tun *udp_match;
+			struct nfp_ipv6_addr_entry *entry;
+			struct in6_addr *dst;
+
+			for (i = 0; i < _CT_TYPE_MAX; i++) {
+				nfp_flower_compile_ipv6_udp_tun((void *)key,
+								(void *)msk, rules[i]);
+			}
+			udp_match = (struct nfp_flower_ipv6_udp_tun *)key;
+			dst = &udp_match->ipv6.dst;
+
+			entry = nfp_tunnel_add_ipv6_off(priv->app, dst);
+			if (!entry)
+				goto ct_offload_err;
+
+			flow_pay->nfp_tun_ipv6 = entry;
+		} else {
+			__be32 dst;
+
+			for (i = 0; i < _CT_TYPE_MAX; i++) {
+				nfp_flower_compile_ipv4_udp_tun((void *)key,
+								(void *)msk, rules[i]);
+			}
+			dst = ((struct nfp_flower_ipv4_udp_tun *)key)->ipv4.dst;
+
+			/* Store the tunnel destination in the rule data.
+			 * This must be present and be an exact match.
+			 */
+			flow_pay->nfp_tun_ipv4_addr = dst;
+			nfp_tunnel_add_ipv4_off(priv->app, dst);
+		}
+
+		if (key_layer.key_layer_two & NFP_FLOWER_LAYER2_GENEVE_OP) {
+			offset = key_map[FLOW_PAY_GENEVE_OPT];
+			key = kdata + offset;
+			msk = mdata + offset;
+			for (i = 0; i < _CT_TYPE_MAX; i++)
+				nfp_flower_compile_geneve_opt(key, msk, rules[i]);
+		}
+	}
+
+	/* Merge actions into flow_pay */
+	err = nfp_fl_merge_actions_offload(rules, priv, netdev, flow_pay);
+	if (err)
+		goto ct_offload_err;
+
+	/* Use the pointer address as the cookie, but set the last bit to 1.
+	 * This is to avoid the 'is_merge_flow' check from detecting this as
+	 * an already merged flow. This works since address alignment means
+	 * that the last bit for pointer addresses will be 0.
+	 */
+	flow_pay->tc_flower_cookie = ((unsigned long)flow_pay) | 0x1;
+	err = nfp_compile_flow_metadata(priv->app, flow_pay->tc_flower_cookie,
+					flow_pay, netdev, NULL);
+	if (err)
+		goto ct_offload_err;
+
+	if (nfp_netdev_is_nfp_repr(netdev))
+		port = nfp_port_from_netdev(netdev);
+
+	err = rhashtable_insert_fast(&priv->flow_table, &flow_pay->fl_node,
+				     nfp_flower_table_params);
+	if (err)
+		goto ct_release_offload_meta_err;
+
+	err = nfp_flower_xmit_flow(priv->app, flow_pay,
+				   NFP_FLOWER_CMSG_TYPE_FLOW_ADD);
+	if (err)
+		goto ct_remove_rhash_err;
+
+	m_entry->tc_flower_cookie = flow_pay->tc_flower_cookie;
+	m_entry->flow_pay = flow_pay;
+
+	if (port)
+		port->tc_offload_cnt++;
+
+	return err;
+
+ct_remove_rhash_err:
+	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
+					    &flow_pay->fl_node,
+					    nfp_flower_table_params));
+ct_release_offload_meta_err:
+	nfp_modify_flow_metadata(priv->app, flow_pay);
+ct_offload_err:
+	if (flow_pay->nfp_tun_ipv4_addr)
+		nfp_tunnel_del_ipv4_off(priv->app, flow_pay->nfp_tun_ipv4_addr);
+	if (flow_pay->nfp_tun_ipv6)
+		nfp_tunnel_put_ipv6_off(priv->app, flow_pay->nfp_tun_ipv6);
+	kfree(flow_pay->action_data);
+	kfree(flow_pay->mask_data);
+	kfree(flow_pay->unmasked_data);
+	kfree(flow_pay);
+	return err;
 }
 
 static int nfp_fl_ct_del_offload(struct nfp_app *app, unsigned long cookie,
 				 struct net_device *netdev)
 {
-	return 0;
+	struct nfp_flower_priv *priv = app->priv;
+	struct nfp_fl_payload *flow_pay;
+	struct nfp_port *port = NULL;
+	int err = 0;
+
+	if (nfp_netdev_is_nfp_repr(netdev))
+		port = nfp_port_from_netdev(netdev);
+
+	flow_pay = nfp_flower_search_fl_table(app, cookie, netdev);
+	if (!flow_pay)
+		return -ENOENT;
+
+	err = nfp_modify_flow_metadata(app, flow_pay);
+	if (err)
+		goto err_free_merge_flow;
+
+	if (flow_pay->nfp_tun_ipv4_addr)
+		nfp_tunnel_del_ipv4_off(app, flow_pay->nfp_tun_ipv4_addr);
+
+	if (flow_pay->nfp_tun_ipv6)
+		nfp_tunnel_put_ipv6_off(app, flow_pay->nfp_tun_ipv6);
+
+	if (!flow_pay->in_hw) {
+		err = 0;
+		goto err_free_merge_flow;
+	}
+
+	err = nfp_flower_xmit_flow(app, flow_pay,
+				   NFP_FLOWER_CMSG_TYPE_FLOW_DEL);
+
+err_free_merge_flow:
+	nfp_flower_del_linked_merge_flows(app, flow_pay);
+	if (port)
+		port->tc_offload_cnt--;
+	kfree(flow_pay->action_data);
+	kfree(flow_pay->mask_data);
+	kfree(flow_pay->unmasked_data);
+	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
+					    &flow_pay->fl_node,
+					    nfp_flower_table_params));
+	kfree_rcu(flow_pay, rcu);
+	return err;
 }
 
 static int nfp_ct_do_nft_merge(struct nfp_fl_ct_zone_entry *zt,
@@ -1048,6 +1521,139 @@ int nfp_fl_ct_handle_post_ct(struct nfp_flower_priv *priv,
 	return 0;
 }
 
+static void
+nfp_fl_ct_sub_stats(struct nfp_fl_nft_tc_merge *nft_merge,
+		    enum ct_entry_type type, u64 *m_pkts,
+		    u64 *m_bytes, u64 *m_used)
+{
+	struct nfp_flower_priv *priv = nft_merge->zt->priv;
+	struct nfp_fl_payload *nfp_flow;
+	u32 ctx_id;
+
+	nfp_flow = nft_merge->flow_pay;
+	if (!nfp_flow)
+		return;
+
+	ctx_id = be32_to_cpu(nfp_flow->meta.host_ctx_id);
+	*m_pkts += priv->stats[ctx_id].pkts;
+	*m_bytes += priv->stats[ctx_id].bytes;
+	*m_used = max_t(u64, *m_used, priv->stats[ctx_id].used);
+
+	/* If request is for a sub_flow which is part of a tunnel merged
+	 * flow then update stats from tunnel merged flows first.
+	 */
+	if (!list_empty(&nfp_flow->linked_flows))
+		nfp_flower_update_merge_stats(priv->app, nfp_flow);
+
+	if (type != CT_TYPE_NFT) {
+		/* Update nft cached stats */
+		flow_stats_update(&nft_merge->nft_parent->stats,
+				  priv->stats[ctx_id].bytes,
+				  priv->stats[ctx_id].pkts,
+				  0, priv->stats[ctx_id].used,
+				  FLOW_ACTION_HW_STATS_DELAYED);
+	} else {
+		/* Update pre_ct cached stats */
+		flow_stats_update(&nft_merge->tc_m_parent->pre_ct_parent->stats,
+				  priv->stats[ctx_id].bytes,
+				  priv->stats[ctx_id].pkts,
+				  0, priv->stats[ctx_id].used,
+				  FLOW_ACTION_HW_STATS_DELAYED);
+		/* Update post_ct cached stats */
+		flow_stats_update(&nft_merge->tc_m_parent->post_ct_parent->stats,
+				  priv->stats[ctx_id].bytes,
+				  priv->stats[ctx_id].pkts,
+				  0, priv->stats[ctx_id].used,
+				  FLOW_ACTION_HW_STATS_DELAYED);
+	}
+	/* Reset stats from the nfp */
+	priv->stats[ctx_id].pkts = 0;
+	priv->stats[ctx_id].bytes = 0;
+}
+
+int nfp_fl_ct_stats(struct flow_cls_offload *flow,
+		    struct nfp_fl_ct_map_entry *ct_map_ent)
+{
+	struct nfp_fl_ct_flow_entry *ct_entry = ct_map_ent->ct_entry;
+	struct nfp_fl_nft_tc_merge *nft_merge, *nft_m_tmp;
+	struct nfp_fl_ct_tc_merge *tc_merge, *tc_m_tmp;
+
+	u64 pkts = 0, bytes = 0, used = 0;
+	u64 m_pkts, m_bytes, m_used;
+
+	spin_lock_bh(&ct_entry->zt->priv->stats_lock);
+
+	if (ct_entry->type == CT_TYPE_PRE_CT) {
+		/* Iterate tc_merge entries associated with this flow */
+		list_for_each_entry_safe(tc_merge, tc_m_tmp, &ct_entry->children,
+					 pre_ct_list) {
+			m_pkts = 0;
+			m_bytes = 0;
+			m_used = 0;
+			/* Iterate nft_merge entries associated with this tc_merge flow */
+			list_for_each_entry_safe(nft_merge, nft_m_tmp, &tc_merge->children,
+						 tc_merge_list) {
+				nfp_fl_ct_sub_stats(nft_merge, CT_TYPE_PRE_CT,
+						    &m_pkts, &m_bytes, &m_used);
+			}
+			pkts += m_pkts;
+			bytes += m_bytes;
+			used = max_t(u64, used, m_used);
+			/* Update post_ct partner */
+			flow_stats_update(&tc_merge->post_ct_parent->stats,
+					  m_bytes, m_pkts, 0, m_used,
+					  FLOW_ACTION_HW_STATS_DELAYED);
+		}
+	} else if (ct_entry->type == CT_TYPE_POST_CT) {
+		/* Iterate tc_merge entries associated with this flow */
+		list_for_each_entry_safe(tc_merge, tc_m_tmp, &ct_entry->children,
+					 post_ct_list) {
+			m_pkts = 0;
+			m_bytes = 0;
+			m_used = 0;
+			/* Iterate nft_merge entries associated with this tc_merge flow */
+			list_for_each_entry_safe(nft_merge, nft_m_tmp, &tc_merge->children,
+						 tc_merge_list) {
+				nfp_fl_ct_sub_stats(nft_merge, CT_TYPE_POST_CT,
+						    &m_pkts, &m_bytes, &m_used);
+			}
+			pkts += m_pkts;
+			bytes += m_bytes;
+			used = max_t(u64, used, m_used);
+			/* Update pre_ct partner */
+			flow_stats_update(&tc_merge->pre_ct_parent->stats,
+					  m_bytes, m_pkts, 0, m_used,
+					  FLOW_ACTION_HW_STATS_DELAYED);
+		}
+	} else  {
+		/* Iterate nft_merge entries associated with this nft flow */
+		list_for_each_entry_safe(nft_merge, nft_m_tmp, &ct_entry->children,
+					 nft_flow_list) {
+			nfp_fl_ct_sub_stats(nft_merge, CT_TYPE_NFT,
+					    &pkts, &bytes, &used);
+		}
+	}
+
+	/* Add stats from this request to stats potentially cached by
+	 * previous requests.
+	 */
+	flow_stats_update(&ct_entry->stats, bytes, pkts, 0, used,
+			  FLOW_ACTION_HW_STATS_DELAYED);
+	/* Finally update the flow stats from the original stats request */
+	flow_stats_update(&flow->stats, ct_entry->stats.bytes,
+			  ct_entry->stats.pkts, 0,
+			  ct_entry->stats.lastused,
+			  FLOW_ACTION_HW_STATS_DELAYED);
+	/* Stats has been synced to original flow, can now clear
+	 * the cache.
+	 */
+	ct_entry->stats.pkts = 0;
+	ct_entry->stats.bytes = 0;
+	spin_unlock_bh(&ct_entry->zt->priv->stats_lock);
+
+	return 0;
+}
+
 static int
 nfp_fl_ct_offload_nft_flow(struct nfp_fl_ct_zone_entry *zt, struct flow_cls_offload *flow)
 {
@@ -1080,7 +1686,11 @@ nfp_fl_ct_offload_nft_flow(struct nfp_fl_ct_zone_entry *zt, struct flow_cls_offl
 						    nfp_ct_map_params);
 		return nfp_fl_ct_del_flow(ct_map_ent);
 	case FLOW_CLS_STATS:
-		return 0;
+		ct_map_ent = rhashtable_lookup_fast(&zt->priv->ct_map_table, &flow->cookie,
+						    nfp_ct_map_params);
+		if (ct_map_ent)
+			return nfp_fl_ct_stats(flow, ct_map_ent);
+		break;
 	default:
 		break;
 	}
diff --git a/drivers/net/ethernet/netronome/nfp/flower/conntrack.h b/drivers/net/ethernet/netronome/nfp/flower/conntrack.h
index 170b6cdb8cd0..beb6cceff9d8 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/conntrack.h
+++ b/drivers/net/ethernet/netronome/nfp/flower/conntrack.h
@@ -83,6 +83,24 @@ enum ct_entry_type {
 	CT_TYPE_PRE_CT,
 	CT_TYPE_NFT,
 	CT_TYPE_POST_CT,
+	_CT_TYPE_MAX,
+};
+
+enum nfp_nfp_layer_name {
+	FLOW_PAY_META_TCI =    0,
+	FLOW_PAY_INPORT,
+	FLOW_PAY_EXT_META,
+	FLOW_PAY_MAC_MPLS,
+	FLOW_PAY_L4,
+	FLOW_PAY_IPV4,
+	FLOW_PAY_IPV6,
+	FLOW_PAY_CT,
+	FLOW_PAY_GRE,
+	FLOW_PAY_QINQ,
+	FLOW_PAY_UDP_TUN,
+	FLOW_PAY_GENEVE_OPT,
+
+	_FLOW_PAY_LAYERS_MAX
 };
 
 /**
@@ -228,4 +246,12 @@ int nfp_fl_ct_del_flow(struct nfp_fl_ct_map_entry *ct_map_ent);
  */
 int nfp_fl_ct_handle_nft_flow(enum tc_setup_type type, void *type_data,
 			      void *cb_priv);
+
+/**
+ * nfp_fl_ct_stats() - Handle flower stats callbacks for ct flows
+ * @flow:	TC flower classifier offload structure.
+ * @ct_map_ent:	ct map entry for the flow that needs deleting
+ */
+int nfp_fl_ct_stats(struct flow_cls_offload *flow,
+		    struct nfp_fl_ct_map_entry *ct_map_ent);
 #endif
diff --git a/drivers/net/ethernet/netronome/nfp/flower/main.h b/drivers/net/ethernet/netronome/nfp/flower/main.h
index 0fbd682ccf72..917c450a7aad 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/main.h
+++ b/drivers/net/ethernet/netronome/nfp/flower/main.h
@@ -413,20 +413,73 @@ int nfp_flower_setup_tc(struct nfp_app *app, struct net_device *netdev,
 int nfp_flower_merge_offloaded_flows(struct nfp_app *app,
 				     struct nfp_fl_payload *sub_flow1,
 				     struct nfp_fl_payload *sub_flow2);
+void
+nfp_flower_compile_meta(struct nfp_flower_meta_tci *ext,
+			struct nfp_flower_meta_tci *msk, u8 key_type);
+void
+nfp_flower_compile_tci(struct nfp_flower_meta_tci *ext,
+		       struct nfp_flower_meta_tci *msk,
+		       struct flow_rule *rule);
+void
+nfp_flower_compile_ext_meta(struct nfp_flower_ext_meta *frame, u32 key_ext);
+int
+nfp_flower_compile_port(struct nfp_flower_in_port *frame, u32 cmsg_port,
+			bool mask_version, enum nfp_flower_tun_type tun_type,
+			struct netlink_ext_ack *extack);
+void
+nfp_flower_compile_mac(struct nfp_flower_mac_mpls *ext,
+		       struct nfp_flower_mac_mpls *msk,
+		       struct flow_rule *rule);
+int
+nfp_flower_compile_mpls(struct nfp_flower_mac_mpls *ext,
+			struct nfp_flower_mac_mpls *msk,
+			struct flow_rule *rule,
+			struct netlink_ext_ack *extack);
+void
+nfp_flower_compile_tport(struct nfp_flower_tp_ports *ext,
+			 struct nfp_flower_tp_ports *msk,
+			 struct flow_rule *rule);
+void
+nfp_flower_compile_vlan(struct nfp_flower_vlan *ext,
+			struct nfp_flower_vlan *msk,
+			struct flow_rule *rule);
+void
+nfp_flower_compile_ipv4(struct nfp_flower_ipv4 *ext,
+			struct nfp_flower_ipv4 *msk, struct flow_rule *rule);
+void
+nfp_flower_compile_ipv6(struct nfp_flower_ipv6 *ext,
+			struct nfp_flower_ipv6 *msk, struct flow_rule *rule);
+void
+nfp_flower_compile_geneve_opt(u8 *ext, u8 *msk, struct flow_rule *rule);
+void
+nfp_flower_compile_ipv4_gre_tun(struct nfp_flower_ipv4_gre_tun *ext,
+				struct nfp_flower_ipv4_gre_tun *msk,
+				struct flow_rule *rule);
+void
+nfp_flower_compile_ipv4_udp_tun(struct nfp_flower_ipv4_udp_tun *ext,
+				struct nfp_flower_ipv4_udp_tun *msk,
+				struct flow_rule *rule);
+void
+nfp_flower_compile_ipv6_udp_tun(struct nfp_flower_ipv6_udp_tun *ext,
+				struct nfp_flower_ipv6_udp_tun *msk,
+				struct flow_rule *rule);
+void
+nfp_flower_compile_ipv6_gre_tun(struct nfp_flower_ipv6_gre_tun *ext,
+				struct nfp_flower_ipv6_gre_tun *msk,
+				struct flow_rule *rule);
 int nfp_flower_compile_flow_match(struct nfp_app *app,
-				  struct flow_cls_offload *flow,
+				  struct flow_rule *rule,
 				  struct nfp_fl_key_ls *key_ls,
 				  struct net_device *netdev,
 				  struct nfp_fl_payload *nfp_flow,
 				  enum nfp_flower_tun_type tun_type,
 				  struct netlink_ext_ack *extack);
 int nfp_flower_compile_action(struct nfp_app *app,
-			      struct flow_cls_offload *flow,
+			      struct flow_rule *rule,
 			      struct net_device *netdev,
 			      struct nfp_fl_payload *nfp_flow,
 			      struct netlink_ext_ack *extack);
-int nfp_compile_flow_metadata(struct nfp_app *app,
-			      struct flow_cls_offload *flow,
+int nfp_compile_flow_metadata(struct nfp_app *app, u32 cookie,
 			      struct nfp_fl_payload *nfp_flow,
 			      struct net_device *netdev,
 			      struct netlink_ext_ack *extack);
@@ -498,4 +551,22 @@ int nfp_flower_xmit_pre_tun_flow(struct nfp_app *app,
 				 struct nfp_fl_payload *flow);
 int nfp_flower_xmit_pre_tun_del_flow(struct nfp_app *app,
 				     struct nfp_fl_payload *flow);
+
+struct nfp_fl_payload *
+nfp_flower_allocate_new(struct nfp_fl_key_ls *key_layer);
+int nfp_flower_calculate_key_layers(struct nfp_app *app,
+				    struct net_device *netdev,
+				    struct nfp_fl_key_ls *ret_key_ls,
+				    struct flow_rule *flow,
+				    enum nfp_flower_tun_type *tun_type,
+				    struct netlink_ext_ack *extack);
+void
+nfp_flower_del_linked_merge_flows(struct nfp_app *app,
+				  struct nfp_fl_payload *sub_flow);
+int
+nfp_flower_xmit_flow(struct nfp_app *app, struct nfp_fl_payload *nfp_flow,
+		     u8 mtype);
+void
+nfp_flower_update_merge_stats(struct nfp_app *app,
+			      struct nfp_fl_payload *sub_flow);
 #endif
diff --git a/drivers/net/ethernet/netronome/nfp/flower/match.c b/drivers/net/ethernet/netronome/nfp/flower/match.c
index 255a4dff6288..9d86eea4dc16 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/match.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/match.c
@@ -7,51 +7,68 @@
 #include "cmsg.h"
 #include "main.h"
 
-static void
-nfp_flower_compile_meta_tci(struct nfp_flower_meta_tci *ext,
-			    struct nfp_flower_meta_tci *msk,
-			    struct flow_rule *rule, u8 key_type, bool qinq_sup)
+void
+nfp_flower_compile_meta(struct nfp_flower_meta_tci *ext,
+			struct nfp_flower_meta_tci *msk, u8 key_type)
 {
-	u16 tmp_tci;
-
-	memset(ext, 0, sizeof(struct nfp_flower_meta_tci));
-	memset(msk, 0, sizeof(struct nfp_flower_meta_tci));
-
 	/* Populate the metadata frame. */
 	ext->nfp_flow_key_layer = key_type;
 	ext->mask_id = ~0;
 
 	msk->nfp_flow_key_layer = key_type;
 	msk->mask_id = ~0;
+}
 
-	if (!qinq_sup && flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
+void
+nfp_flower_compile_tci(struct nfp_flower_meta_tci *ext,
+		       struct nfp_flower_meta_tci *msk,
+		       struct flow_rule *rule)
+{
+	u16 msk_tci, key_tci;
+
+	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
 		struct flow_match_vlan match;
 
 		flow_rule_match_vlan(rule, &match);
 		/* Populate the tci field. */
-		tmp_tci = NFP_FLOWER_MASK_VLAN_PRESENT;
-		tmp_tci |= FIELD_PREP(NFP_FLOWER_MASK_VLAN_PRIO,
+		key_tci = NFP_FLOWER_MASK_VLAN_PRESENT;
+		key_tci |= FIELD_PREP(NFP_FLOWER_MASK_VLAN_PRIO,
 				      match.key->vlan_priority) |
 			   FIELD_PREP(NFP_FLOWER_MASK_VLAN_VID,
 				      match.key->vlan_id);
-		ext->tci = cpu_to_be16(tmp_tci);
 
-		tmp_tci = NFP_FLOWER_MASK_VLAN_PRESENT;
-		tmp_tci |= FIELD_PREP(NFP_FLOWER_MASK_VLAN_PRIO,
+		msk_tci = NFP_FLOWER_MASK_VLAN_PRESENT;
+		msk_tci |= FIELD_PREP(NFP_FLOWER_MASK_VLAN_PRIO,
 				      match.mask->vlan_priority) |
 			   FIELD_PREP(NFP_FLOWER_MASK_VLAN_VID,
 				      match.mask->vlan_id);
-		msk->tci = cpu_to_be16(tmp_tci);
+
+		ext->tci |= cpu_to_be16((key_tci & msk_tci));
+		msk->tci |= cpu_to_be16(msk_tci);
 	}
 }
 
 static void
+nfp_flower_compile_meta_tci(struct nfp_flower_meta_tci *ext,
+			    struct nfp_flower_meta_tci *msk,
+			    struct flow_rule *rule, u8 key_type, bool qinq_sup)
+{
+	memset(ext, 0, sizeof(struct nfp_flower_meta_tci));
+	memset(msk, 0, sizeof(struct nfp_flower_meta_tci));
+
+	nfp_flower_compile_meta(ext, msk, key_type);
+
+	if (!qinq_sup)
+		nfp_flower_compile_tci(ext, msk, rule);
+}
+
+void
 nfp_flower_compile_ext_meta(struct nfp_flower_ext_meta *frame, u32 key_ext)
 {
 	frame->nfp_flow_key_layer2 = cpu_to_be32(key_ext);
 }
 
-static int
+int
 nfp_flower_compile_port(struct nfp_flower_in_port *frame, u32 cmsg_port,
 			bool mask_version, enum nfp_flower_tun_type tun_type,
 			struct netlink_ext_ack *extack)
@@ -74,28 +91,37 @@ nfp_flower_compile_port(struct nfp_flower_in_port *frame, u32 cmsg_port,
 	return 0;
 }
 
-static int
+void
 nfp_flower_compile_mac(struct nfp_flower_mac_mpls *ext,
-		       struct nfp_flower_mac_mpls *msk, struct flow_rule *rule,
-		       struct netlink_ext_ack *extack)
+		       struct nfp_flower_mac_mpls *msk,
+		       struct flow_rule *rule)
 {
-	memset(ext, 0, sizeof(struct nfp_flower_mac_mpls));
-	memset(msk, 0, sizeof(struct nfp_flower_mac_mpls));
-
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
 		struct flow_match_eth_addrs match;
+		int i;
 
 		flow_rule_match_eth_addrs(rule, &match);
 		/* Populate mac frame. */
-		ether_addr_copy(ext->mac_dst, &match.key->dst[0]);
-		ether_addr_copy(ext->mac_src, &match.key->src[0]);
-		ether_addr_copy(msk->mac_dst, &match.mask->dst[0]);
-		ether_addr_copy(msk->mac_src, &match.mask->src[0]);
+		for (i = 0; i < ETH_ALEN; i++) {
+			ext->mac_dst[i] |= match.key->dst[i] &
+					   match.mask->dst[i];
+			msk->mac_dst[i] |= match.mask->dst[i];
+			ext->mac_src[i] |= match.key->src[i] &
+					   match.mask->src[i];
+			msk->mac_src[i] |= match.mask->src[i];
+		}
 	}
+}
 
+int
+nfp_flower_compile_mpls(struct nfp_flower_mac_mpls *ext,
+			struct nfp_flower_mac_mpls *msk,
+			struct flow_rule *rule,
+			struct netlink_ext_ack *extack)
+{
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_MPLS)) {
 		struct flow_match_mpls match;
-		u32 t_mpls;
+		u32 key_mpls, msk_mpls;
 
 		flow_rule_match_mpls(rule, &match);
 
@@ -106,22 +132,24 @@ nfp_flower_compile_mac(struct nfp_flower_mac_mpls *ext,
 			return -EOPNOTSUPP;
 		}
 
-		t_mpls = FIELD_PREP(NFP_FLOWER_MASK_MPLS_LB,
-				    match.key->ls[0].mpls_label) |
-			 FIELD_PREP(NFP_FLOWER_MASK_MPLS_TC,
-				    match.key->ls[0].mpls_tc) |
-			 FIELD_PREP(NFP_FLOWER_MASK_MPLS_BOS,
-				    match.key->ls[0].mpls_bos) |
-			 NFP_FLOWER_MASK_MPLS_Q;
-		ext->mpls_lse = cpu_to_be32(t_mpls);
-		t_mpls = FIELD_PREP(NFP_FLOWER_MASK_MPLS_LB,
-				    match.mask->ls[0].mpls_label) |
-			 FIELD_PREP(NFP_FLOWER_MASK_MPLS_TC,
-				    match.mask->ls[0].mpls_tc) |
-			 FIELD_PREP(NFP_FLOWER_MASK_MPLS_BOS,
-				    match.mask->ls[0].mpls_bos) |
-			 NFP_FLOWER_MASK_MPLS_Q;
-		msk->mpls_lse = cpu_to_be32(t_mpls);
+		key_mpls = FIELD_PREP(NFP_FLOWER_MASK_MPLS_LB,
+				      match.key->ls[0].mpls_label) |
+			   FIELD_PREP(NFP_FLOWER_MASK_MPLS_TC,
+				      match.key->ls[0].mpls_tc) |
+			   FIELD_PREP(NFP_FLOWER_MASK_MPLS_BOS,
+				      match.key->ls[0].mpls_bos) |
+			   NFP_FLOWER_MASK_MPLS_Q;
+
+		msk_mpls = FIELD_PREP(NFP_FLOWER_MASK_MPLS_LB,
+				      match.mask->ls[0].mpls_label) |
+			   FIELD_PREP(NFP_FLOWER_MASK_MPLS_TC,
+				      match.mask->ls[0].mpls_tc) |
+			   FIELD_PREP(NFP_FLOWER_MASK_MPLS_BOS,
+				      match.mask->ls[0].mpls_bos) |
+			   NFP_FLOWER_MASK_MPLS_Q;
+
+		ext->mpls_lse |= cpu_to_be32((key_mpls & msk_mpls));
+		msk->mpls_lse |= cpu_to_be32(msk_mpls);
 	} else if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
 		/* Check for mpls ether type and set NFP_FLOWER_MASK_MPLS_Q
 		 * bit, which indicates an mpls ether type but without any
@@ -132,30 +160,41 @@ nfp_flower_compile_mac(struct nfp_flower_mac_mpls *ext,
 		flow_rule_match_basic(rule, &match);
 		if (match.key->n_proto == cpu_to_be16(ETH_P_MPLS_UC) ||
 		    match.key->n_proto == cpu_to_be16(ETH_P_MPLS_MC)) {
-			ext->mpls_lse = cpu_to_be32(NFP_FLOWER_MASK_MPLS_Q);
-			msk->mpls_lse = cpu_to_be32(NFP_FLOWER_MASK_MPLS_Q);
+			ext->mpls_lse |= cpu_to_be32(NFP_FLOWER_MASK_MPLS_Q);
+			msk->mpls_lse |= cpu_to_be32(NFP_FLOWER_MASK_MPLS_Q);
 		}
 	}
 
 	return 0;
 }
 
-static void
+static int
+nfp_flower_compile_mac_mpls(struct nfp_flower_mac_mpls *ext,
+			    struct nfp_flower_mac_mpls *msk,
+			    struct flow_rule *rule,
+			    struct netlink_ext_ack *extack)
+{
+	memset(ext, 0, sizeof(struct nfp_flower_mac_mpls));
+	memset(msk, 0, sizeof(struct nfp_flower_mac_mpls));
+
+	nfp_flower_compile_mac(ext, msk, rule);
+
+	return nfp_flower_compile_mpls(ext, msk, rule, extack);
+}
+
+void
 nfp_flower_compile_tport(struct nfp_flower_tp_ports *ext,
 			 struct nfp_flower_tp_ports *msk,
 			 struct flow_rule *rule)
 {
-	memset(ext, 0, sizeof(struct nfp_flower_tp_ports));
-	memset(msk, 0, sizeof(struct nfp_flower_tp_ports));
-
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS)) {
 		struct flow_match_ports match;
 
 		flow_rule_match_ports(rule, &match);
-		ext->port_src = match.key->src;
-		ext->port_dst = match.key->dst;
-		msk->port_src = match.mask->src;
-		msk->port_dst = match.mask->dst;
+		ext->port_src |= match.key->src & match.mask->src;
+		ext->port_dst |= match.key->dst & match.mask->dst;
+		msk->port_src |= match.mask->src;
+		msk->port_dst |= match.mask->dst;
 	}
 }
 
@@ -167,18 +206,18 @@ nfp_flower_compile_ip_ext(struct nfp_flower_ip_ext *ext,
 		struct flow_match_basic match;
 
 		flow_rule_match_basic(rule, &match);
-		ext->proto = match.key->ip_proto;
-		msk->proto = match.mask->ip_proto;
+		ext->proto |= match.key->ip_proto & match.mask->ip_proto;
+		msk->proto |= match.mask->ip_proto;
 	}
 
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IP)) {
 		struct flow_match_ip match;
 
 		flow_rule_match_ip(rule, &match);
-		ext->tos = match.key->tos;
-		ext->ttl = match.key->ttl;
-		msk->tos = match.mask->tos;
-		msk->ttl = match.mask->ttl;
+		ext->tos |= match.key->tos & match.mask->tos;
+		ext->ttl |= match.key->ttl & match.mask->ttl;
+		msk->tos |= match.mask->tos;
+		msk->ttl |= match.mask->ttl;
 	}
 
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_TCP)) {
@@ -231,99 +270,108 @@ nfp_flower_compile_ip_ext(struct nfp_flower_ip_ext *ext,
 }
 
 static void
-nfp_flower_fill_vlan(struct flow_dissector_key_vlan *key,
-		     struct nfp_flower_vlan *frame,
-		     bool outer_vlan)
+nfp_flower_fill_vlan(struct flow_match_vlan *match,
+		     struct nfp_flower_vlan *ext,
+		     struct nfp_flower_vlan *msk, bool outer_vlan)
 {
-	u16 tci;
-
-	tci = NFP_FLOWER_MASK_VLAN_PRESENT;
-	tci |= FIELD_PREP(NFP_FLOWER_MASK_VLAN_PRIO,
-			  key->vlan_priority) |
-	       FIELD_PREP(NFP_FLOWER_MASK_VLAN_VID,
-			  key->vlan_id);
+	struct flow_dissector_key_vlan *mask = match->mask;
+	struct flow_dissector_key_vlan *key = match->key;
+	u16 msk_tci, key_tci;
+
+	key_tci = NFP_FLOWER_MASK_VLAN_PRESENT;
+	key_tci |= FIELD_PREP(NFP_FLOWER_MASK_VLAN_PRIO,
+			      key->vlan_priority) |
+		   FIELD_PREP(NFP_FLOWER_MASK_VLAN_VID,
+			      key->vlan_id);
+	msk_tci = NFP_FLOWER_MASK_VLAN_PRESENT;
+	msk_tci |= FIELD_PREP(NFP_FLOWER_MASK_VLAN_PRIO,
+			      mask->vlan_priority) |
+		   FIELD_PREP(NFP_FLOWER_MASK_VLAN_VID,
+			      mask->vlan_id);
 
 	if (outer_vlan) {
-		frame->outer_tci = cpu_to_be16(tci);
-		frame->outer_tpid = key->vlan_tpid;
+		ext->outer_tci |= cpu_to_be16((key_tci & msk_tci));
+		ext->outer_tpid |= key->vlan_tpid & mask->vlan_tpid;
+		msk->outer_tci |= cpu_to_be16(msk_tci);
+		msk->outer_tpid |= mask->vlan_tpid;
 	} else {
-		frame->inner_tci = cpu_to_be16(tci);
-		frame->inner_tpid = key->vlan_tpid;
+		ext->inner_tci |= cpu_to_be16((key_tci & msk_tci));
+		ext->inner_tpid |= key->vlan_tpid & mask->vlan_tpid;
+		msk->inner_tci |= cpu_to_be16(msk_tci);
+		msk->inner_tpid |= mask->vlan_tpid;
 	}
 }
 
-static void
+void
 nfp_flower_compile_vlan(struct nfp_flower_vlan *ext,
 			struct nfp_flower_vlan *msk,
 			struct flow_rule *rule)
 {
 	struct flow_match_vlan match;
 
-	memset(ext, 0, sizeof(struct nfp_flower_vlan));
-	memset(msk, 0, sizeof(struct nfp_flower_vlan));
-
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
 		flow_rule_match_vlan(rule, &match);
-		nfp_flower_fill_vlan(match.key, ext, true);
-		nfp_flower_fill_vlan(match.mask, msk, true);
+		nfp_flower_fill_vlan(&match, ext, msk, true);
 	}
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CVLAN)) {
 		flow_rule_match_cvlan(rule, &match);
-		nfp_flower_fill_vlan(match.key, ext, false);
-		nfp_flower_fill_vlan(match.mask, msk, false);
+		nfp_flower_fill_vlan(&match, ext, msk, false);
 	}
 }
 
-static void
+void
 nfp_flower_compile_ipv4(struct nfp_flower_ipv4 *ext,
 			struct nfp_flower_ipv4 *msk, struct flow_rule *rule)
 {
-	struct flow_match_ipv4_addrs match;
-
-	memset(ext, 0, sizeof(struct nfp_flower_ipv4));
-	memset(msk, 0, sizeof(struct nfp_flower_ipv4));
-
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
+		struct flow_match_ipv4_addrs match;
+
 		flow_rule_match_ipv4_addrs(rule, &match);
-		ext->ipv4_src = match.key->src;
-		ext->ipv4_dst = match.key->dst;
-		msk->ipv4_src = match.mask->src;
-		msk->ipv4_dst = match.mask->dst;
+		ext->ipv4_src |= match.key->src & match.mask->src;
+		ext->ipv4_dst |= match.key->dst & match.mask->dst;
+		msk->ipv4_src |= match.mask->src;
+		msk->ipv4_dst |= match.mask->dst;
 	}
 
 	nfp_flower_compile_ip_ext(&ext->ip_ext, &msk->ip_ext, rule);
 }
 
-static void
+void
 nfp_flower_compile_ipv6(struct nfp_flower_ipv6 *ext,
 			struct nfp_flower_ipv6 *msk, struct flow_rule *rule)
 {
-	memset(ext, 0, sizeof(struct nfp_flower_ipv6));
-	memset(msk, 0, sizeof(struct nfp_flower_ipv6));
-
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
 		struct flow_match_ipv6_addrs match;
+		int i;
 
 		flow_rule_match_ipv6_addrs(rule, &match);
-		ext->ipv6_src = match.key->src;
-		ext->ipv6_dst = match.key->dst;
-		msk->ipv6_src = match.mask->src;
-		msk->ipv6_dst = match.mask->dst;
+		for (i = 0; i < sizeof(ext->ipv6_src); i++) {
+			ext->ipv6_src.s6_addr[i] |= match.key->src.s6_addr[i] &
+						    match.mask->src.s6_addr[i];
+			ext->ipv6_dst.s6_addr[i] |= match.key->dst.s6_addr[i] &
+						    match.mask->dst.s6_addr[i];
+			msk->ipv6_src.s6_addr[i] |= match.mask->src.s6_addr[i];
+			msk->ipv6_dst.s6_addr[i] |= match.mask->dst.s6_addr[i];
+		}
 	}
 
 	nfp_flower_compile_ip_ext(&ext->ip_ext, &msk->ip_ext, rule);
 }
 
-static int
-nfp_flower_compile_geneve_opt(void *ext, void *msk, struct flow_rule *rule)
+void
+nfp_flower_compile_geneve_opt(u8 *ext, u8 *msk, struct flow_rule *rule)
 {
 	struct flow_match_enc_opts match;
+	int i;
 
-	flow_rule_match_enc_opts(rule, &match);
-	memcpy(ext, match.key->data, match.key->len);
-	memcpy(msk, match.mask->data, match.mask->len);
+	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
+		flow_rule_match_enc_opts(rule, &match);
 
-	return 0;
+		for (i = 0; i < match.mask->len; i++) {
+			ext[i] |= match.key->data[i] & match.mask->data[i];
+			msk[i] |= match.mask->data[i];
+		}
+	}
 }
 
 static void
@@ -335,10 +383,10 @@ nfp_flower_compile_tun_ipv4_addrs(struct nfp_flower_tun_ipv4 *ext,
 		struct flow_match_ipv4_addrs match;
 
 		flow_rule_match_enc_ipv4_addrs(rule, &match);
-		ext->src = match.key->src;
-		ext->dst = match.key->dst;
-		msk->src = match.mask->src;
-		msk->dst = match.mask->dst;
+		ext->src |= match.key->src & match.mask->src;
+		ext->dst |= match.key->dst & match.mask->dst;
+		msk->src |= match.mask->src;
+		msk->dst |= match.mask->dst;
 	}
 }
 
@@ -349,12 +397,17 @@ nfp_flower_compile_tun_ipv6_addrs(struct nfp_flower_tun_ipv6 *ext,
 {
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
 		struct flow_match_ipv6_addrs match;
+		int i;
 
 		flow_rule_match_enc_ipv6_addrs(rule, &match);
-		ext->src = match.key->src;
-		ext->dst = match.key->dst;
-		msk->src = match.mask->src;
-		msk->dst = match.mask->dst;
+		for (i = 0; i < sizeof(ext->src); i++) {
+			ext->src.s6_addr[i] |= match.key->src.s6_addr[i] &
+					       match.mask->src.s6_addr[i];
+			ext->dst.s6_addr[i] |= match.key->dst.s6_addr[i] &
+					       match.mask->dst.s6_addr[i];
+			msk->src.s6_addr[i] |= match.mask->src.s6_addr[i];
+			msk->dst.s6_addr[i] |= match.mask->dst.s6_addr[i];
+		}
 	}
 }
 
@@ -367,10 +420,10 @@ nfp_flower_compile_tun_ip_ext(struct nfp_flower_tun_ip_ext *ext,
 		struct flow_match_ip match;
 
 		flow_rule_match_enc_ip(rule, &match);
-		ext->tos = match.key->tos;
-		ext->ttl = match.key->ttl;
-		msk->tos = match.mask->tos;
-		msk->ttl = match.mask->ttl;
+		ext->tos |= match.key->tos & match.mask->tos;
+		ext->ttl |= match.key->ttl & match.mask->ttl;
+		msk->tos |= match.mask->tos;
+		msk->ttl |= match.mask->ttl;
 	}
 }
 
@@ -383,10 +436,11 @@ nfp_flower_compile_tun_udp_key(__be32 *key, __be32 *key_msk,
 		u32 vni;
 
 		flow_rule_match_enc_keyid(rule, &match);
-		vni = be32_to_cpu(match.key->keyid) << NFP_FL_TUN_VNI_OFFSET;
-		*key = cpu_to_be32(vni);
+		vni = be32_to_cpu((match.key->keyid & match.mask->keyid)) <<
+		      NFP_FL_TUN_VNI_OFFSET;
+		*key |= cpu_to_be32(vni);
 		vni = be32_to_cpu(match.mask->keyid) << NFP_FL_TUN_VNI_OFFSET;
-		*key_msk = cpu_to_be32(vni);
+		*key_msk |= cpu_to_be32(vni);
 	}
 }
 
@@ -398,22 +452,19 @@ nfp_flower_compile_tun_gre_key(__be32 *key, __be32 *key_msk, __be16 *flags,
 		struct flow_match_enc_keyid match;
 
 		flow_rule_match_enc_keyid(rule, &match);
-		*key = match.key->keyid;
-		*key_msk = match.mask->keyid;
+		*key |= match.key->keyid & match.mask->keyid;
+		*key_msk |= match.mask->keyid;
 
 		*flags = cpu_to_be16(NFP_FL_GRE_FLAG_KEY);
 		*flags_msk = cpu_to_be16(NFP_FL_GRE_FLAG_KEY);
 	}
 }
 
-static void
+void
 nfp_flower_compile_ipv4_gre_tun(struct nfp_flower_ipv4_gre_tun *ext,
 				struct nfp_flower_ipv4_gre_tun *msk,
 				struct flow_rule *rule)
 {
-	memset(ext, 0, sizeof(struct nfp_flower_ipv4_gre_tun));
-	memset(msk, 0, sizeof(struct nfp_flower_ipv4_gre_tun));
-
 	/* NVGRE is the only supported GRE tunnel type */
 	ext->ethertype = cpu_to_be16(ETH_P_TEB);
 	msk->ethertype = cpu_to_be16(~0);
@@ -424,40 +475,31 @@ nfp_flower_compile_ipv4_gre_tun(struct nfp_flower_ipv4_gre_tun *ext,
 				       &ext->tun_flags, &msk->tun_flags, rule);
 }
 
-static void
+void
 nfp_flower_compile_ipv4_udp_tun(struct nfp_flower_ipv4_udp_tun *ext,
 				struct nfp_flower_ipv4_udp_tun *msk,
 				struct flow_rule *rule)
 {
-	memset(ext, 0, sizeof(struct nfp_flower_ipv4_udp_tun));
-	memset(msk, 0, sizeof(struct nfp_flower_ipv4_udp_tun));
-
 	nfp_flower_compile_tun_ipv4_addrs(&ext->ipv4, &msk->ipv4, rule);
 	nfp_flower_compile_tun_ip_ext(&ext->ip_ext, &msk->ip_ext, rule);
 	nfp_flower_compile_tun_udp_key(&ext->tun_id, &msk->tun_id, rule);
 }
 
-static void
+void
 nfp_flower_compile_ipv6_udp_tun(struct nfp_flower_ipv6_udp_tun *ext,
 				struct nfp_flower_ipv6_udp_tun *msk,
 				struct flow_rule *rule)
 {
-	memset(ext, 0, sizeof(struct nfp_flower_ipv6_udp_tun));
-	memset(msk, 0, sizeof(struct nfp_flower_ipv6_udp_tun));
-
 	nfp_flower_compile_tun_ipv6_addrs(&ext->ipv6, &msk->ipv6, rule);
 	nfp_flower_compile_tun_ip_ext(&ext->ip_ext, &msk->ip_ext, rule);
 	nfp_flower_compile_tun_udp_key(&ext->tun_id, &msk->tun_id, rule);
 }
 
-static void
+void
 nfp_flower_compile_ipv6_gre_tun(struct nfp_flower_ipv6_gre_tun *ext,
 				struct nfp_flower_ipv6_gre_tun *msk,
 				struct flow_rule *rule)
 {
-	memset(ext, 0, sizeof(struct nfp_flower_ipv6_gre_tun));
-	memset(msk, 0, sizeof(struct nfp_flower_ipv6_gre_tun));
-
 	/* NVGRE is the only supported GRE tunnel type */
 	ext->ethertype = cpu_to_be16(ETH_P_TEB);
 	msk->ethertype = cpu_to_be16(~0);
@@ -469,14 +511,13 @@ nfp_flower_compile_ipv6_gre_tun(struct nfp_flower_ipv6_gre_tun *ext,
 }
 
 int nfp_flower_compile_flow_match(struct nfp_app *app,
-				  struct flow_cls_offload *flow,
+				  struct flow_rule *rule,
 				  struct nfp_fl_key_ls *key_ls,
 				  struct net_device *netdev,
 				  struct nfp_fl_payload *nfp_flow,
 				  enum nfp_flower_tun_type tun_type,
 				  struct netlink_ext_ack *extack)
 {
-	struct flow_rule *rule = flow_cls_offload_flow_rule(flow);
 	struct nfp_flower_priv *priv = app->priv;
 	bool qinq_sup;
 	u32 port_id;
@@ -527,9 +568,9 @@ int nfp_flower_compile_flow_match(struct nfp_app *app,
 	msk += sizeof(struct nfp_flower_in_port);
 
 	if (NFP_FLOWER_LAYER_MAC & key_ls->key_layer) {
-		err = nfp_flower_compile_mac((struct nfp_flower_mac_mpls *)ext,
-					     (struct nfp_flower_mac_mpls *)msk,
-					     rule, extack);
+		err = nfp_flower_compile_mac_mpls((struct nfp_flower_mac_mpls *)ext,
+						  (struct nfp_flower_mac_mpls *)msk,
+						  rule, extack);
 		if (err)
 			return err;
 
@@ -640,9 +681,7 @@ int nfp_flower_compile_flow_match(struct nfp_app *app,
 		}
 
 		if (key_ls->key_layer_two & NFP_FLOWER_LAYER2_GENEVE_OP) {
-			err = nfp_flower_compile_geneve_opt(ext, msk, rule);
-			if (err)
-				return err;
+			nfp_flower_compile_geneve_opt(ext, msk, rule);
 		}
 	}
 
diff --git a/drivers/net/ethernet/netronome/nfp/flower/metadata.c b/drivers/net/ethernet/netronome/nfp/flower/metadata.c
index 621113650a9b..2af9faee96c5 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/metadata.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/metadata.c
@@ -290,8 +290,7 @@ nfp_check_mask_remove(struct nfp_app *app, char *mask_data, u32 mask_len,
 	return true;
 }
 
-int nfp_compile_flow_metadata(struct nfp_app *app,
-			      struct flow_cls_offload *flow,
+int nfp_compile_flow_metadata(struct nfp_app *app, u32 cookie,
 			      struct nfp_fl_payload *nfp_flow,
 			      struct net_device *netdev,
 			      struct netlink_ext_ack *extack)
@@ -310,7 +309,7 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 	}
 
 	nfp_flow->meta.host_ctx_id = cpu_to_be32(stats_cxt);
-	nfp_flow->meta.host_cookie = cpu_to_be64(flow->cookie);
+	nfp_flow->meta.host_cookie = cpu_to_be64(cookie);
 	nfp_flow->ingress_dev = netdev;
 
 	ctx_entry = kzalloc(sizeof(*ctx_entry), GFP_KERNEL);
@@ -357,7 +356,7 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 	priv->stats[stats_cxt].bytes = 0;
 	priv->stats[stats_cxt].used = jiffies;
 
-	check_entry = nfp_flower_search_fl_table(app, flow->cookie, netdev);
+	check_entry = nfp_flower_search_fl_table(app, cookie, netdev);
 	if (check_entry) {
 		NL_SET_ERR_MSG_MOD(extack, "invalid entry: cannot offload duplicate flow entry");
 		if (nfp_release_stats_entry(app, stats_cxt)) {
diff --git a/drivers/net/ethernet/netronome/nfp/flower/offload.c b/drivers/net/ethernet/netronome/nfp/flower/offload.c
index 2406d33356ad..556c3495211d 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/offload.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/offload.c
@@ -41,6 +41,8 @@
 	 BIT(FLOW_DISSECTOR_KEY_ENC_OPTS) | \
 	 BIT(FLOW_DISSECTOR_KEY_ENC_IP) | \
 	 BIT(FLOW_DISSECTOR_KEY_MPLS) | \
+	 BIT(FLOW_DISSECTOR_KEY_CT) | \
+	 BIT(FLOW_DISSECTOR_KEY_META) | \
 	 BIT(FLOW_DISSECTOR_KEY_IP))
 
 #define NFP_FLOWER_WHITELIST_TUN_DISSECTOR \
@@ -89,7 +91,7 @@ struct nfp_flower_merge_check {
 	};
 };
 
-static int
+int
 nfp_flower_xmit_flow(struct nfp_app *app, struct nfp_fl_payload *nfp_flow,
 		     u8 mtype)
 {
@@ -134,20 +136,16 @@ nfp_flower_xmit_flow(struct nfp_app *app, struct nfp_fl_payload *nfp_flow,
 	return 0;
 }
 
-static bool nfp_flower_check_higher_than_mac(struct flow_cls_offload *f)
+static bool nfp_flower_check_higher_than_mac(struct flow_rule *rule)
 {
-	struct flow_rule *rule = flow_cls_offload_flow_rule(f);
-
 	return flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV4_ADDRS) ||
 	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV6_ADDRS) ||
 	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS) ||
 	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ICMP);
 }
 
-static bool nfp_flower_check_higher_than_l3(struct flow_cls_offload *f)
+static bool nfp_flower_check_higher_than_l3(struct flow_rule *rule)
 {
-	struct flow_rule *rule = flow_cls_offload_flow_rule(f);
-
 	return flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS) ||
 	       flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ICMP);
 }
@@ -236,15 +234,14 @@ nfp_flower_calc_udp_tun_layer(struct flow_dissector_key_ports *enc_ports,
 	return 0;
 }
 
-static int
+int
 nfp_flower_calculate_key_layers(struct nfp_app *app,
 				struct net_device *netdev,
 				struct nfp_fl_key_ls *ret_key_ls,
-				struct flow_cls_offload *flow,
+				struct flow_rule *rule,
 				enum nfp_flower_tun_type *tun_type,
 				struct netlink_ext_ack *extack)
 {
-	struct flow_rule *rule = flow_cls_offload_flow_rule(flow);
 	struct flow_dissector *dissector = rule->match.dissector;
 	struct flow_match_basic basic = { NULL, NULL};
 	struct nfp_flower_priv *priv = app->priv;
@@ -452,7 +449,7 @@ nfp_flower_calculate_key_layers(struct nfp_app *app,
 			NL_SET_ERR_MSG_MOD(extack, "unsupported offload: match on given EtherType is not supported");
 			return -EOPNOTSUPP;
 		}
-	} else if (nfp_flower_check_higher_than_mac(flow)) {
+	} else if (nfp_flower_check_higher_than_mac(rule)) {
 		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: cannot match above L2 without specified EtherType");
 		return -EOPNOTSUPP;
 	}
@@ -471,7 +468,7 @@ nfp_flower_calculate_key_layers(struct nfp_app *app,
 	}
 
 	if (!(key_layer & NFP_FLOWER_LAYER_TP) &&
-	    nfp_flower_check_higher_than_l3(flow)) {
+	    nfp_flower_check_higher_than_l3(rule)) {
 		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: cannot match on L4 information without specified IP protocol type");
 		return -EOPNOTSUPP;
 	}
@@ -543,7 +540,7 @@ nfp_flower_calculate_key_layers(struct nfp_app *app,
 	return 0;
 }
 
-static struct nfp_fl_payload *
+struct nfp_fl_payload *
 nfp_flower_allocate_new(struct nfp_fl_key_ls *key_layer)
 {
 	struct nfp_fl_payload *flow_pay;
@@ -1005,9 +1002,7 @@ int nfp_flower_merge_offloaded_flows(struct nfp_app *app,
 				     struct nfp_fl_payload *sub_flow1,
 				     struct nfp_fl_payload *sub_flow2)
 {
-	struct flow_cls_offload merge_tc_off;
 	struct nfp_flower_priv *priv = app->priv;
-	struct netlink_ext_ack *extack = NULL;
 	struct nfp_fl_payload *merge_flow;
 	struct nfp_fl_key_ls merge_key_ls;
 	struct nfp_merge_info *merge_info;
@@ -1016,7 +1011,6 @@ int nfp_flower_merge_offloaded_flows(struct nfp_app *app,
 
 	ASSERT_RTNL();
 
-	extack = merge_tc_off.common.extack;
 	if (sub_flow1 == sub_flow2 ||
 	    nfp_flower_is_merge_flow(sub_flow1) ||
 	    nfp_flower_is_merge_flow(sub_flow2))
@@ -1061,9 +1055,8 @@ int nfp_flower_merge_offloaded_flows(struct nfp_app *app,
 	if (err)
 		goto err_unlink_sub_flow1;
 
-	merge_tc_off.cookie = merge_flow->tc_flower_cookie;
-	err = nfp_compile_flow_metadata(app, &merge_tc_off, merge_flow,
-					merge_flow->ingress_dev, extack);
+	err = nfp_compile_flow_metadata(app, merge_flow->tc_flower_cookie, merge_flow,
+					merge_flow->ingress_dev, NULL);
 	if (err)
 		goto err_unlink_sub_flow2;
 
@@ -1305,6 +1298,7 @@ static int
 nfp_flower_add_offload(struct nfp_app *app, struct net_device *netdev,
 		       struct flow_cls_offload *flow)
 {
+	struct flow_rule *rule = flow_cls_offload_flow_rule(flow);
 	enum nfp_flower_tun_type tun_type = NFP_FL_TUNNEL_NONE;
 	struct nfp_flower_priv *priv = app->priv;
 	struct netlink_ext_ack *extack = NULL;
@@ -1330,7 +1324,7 @@ nfp_flower_add_offload(struct nfp_app *app, struct net_device *netdev,
 	if (!key_layer)
 		return -ENOMEM;
 
-	err = nfp_flower_calculate_key_layers(app, netdev, key_layer, flow,
+	err = nfp_flower_calculate_key_layers(app, netdev, key_layer, rule,
 					      &tun_type, extack);
 	if (err)
 		goto err_free_key_ls;
@@ -1341,12 +1335,12 @@ nfp_flower_add_offload(struct nfp_app *app, struct net_device *netdev,
 		goto err_free_key_ls;
 	}
 
-	err = nfp_flower_compile_flow_match(app, flow, key_layer, netdev,
+	err = nfp_flower_compile_flow_match(app, rule, key_layer, netdev,
 					    flow_pay, tun_type, extack);
 	if (err)
 		goto err_destroy_flow;
 
-	err = nfp_flower_compile_action(app, flow, netdev, flow_pay, extack);
+	err = nfp_flower_compile_action(app, rule, netdev, flow_pay, extack);
 	if (err)
 		goto err_destroy_flow;
 
@@ -1356,7 +1350,7 @@ nfp_flower_add_offload(struct nfp_app *app, struct net_device *netdev,
 			goto err_destroy_flow;
 	}
 
-	err = nfp_compile_flow_metadata(app, flow, flow_pay, netdev, extack);
+	err = nfp_compile_flow_metadata(app, flow->cookie, flow_pay, netdev, extack);
 	if (err)
 		goto err_destroy_flow;
 
@@ -1476,7 +1470,7 @@ err_free_links:
 	kfree_rcu(merge_flow, rcu);
 }
 
-static void
+void
 nfp_flower_del_linked_merge_flows(struct nfp_app *app,
 				  struct nfp_fl_payload *sub_flow)
 {
@@ -1601,7 +1595,7 @@ __nfp_flower_update_merge_stats(struct nfp_app *app,
 	}
 }
 
-static void
+void
 nfp_flower_update_merge_stats(struct nfp_app *app,
 			      struct nfp_fl_payload *sub_flow)
 {
@@ -1628,10 +1622,17 @@ nfp_flower_get_stats(struct nfp_app *app, struct net_device *netdev,
 		     struct flow_cls_offload *flow)
 {
 	struct nfp_flower_priv *priv = app->priv;
+	struct nfp_fl_ct_map_entry *ct_map_ent;
 	struct netlink_ext_ack *extack = NULL;
 	struct nfp_fl_payload *nfp_flow;
 	u32 ctx_id;
 
+	/* Check ct_map table first */
+	ct_map_ent = rhashtable_lookup_fast(&priv->ct_map_table, &flow->cookie,
+					    nfp_ct_map_params);
+	if (ct_map_ent)
+		return nfp_fl_ct_stats(flow, ct_map_ent);
+
 	extack = flow->common.extack;
 	nfp_flow = nfp_flower_search_fl_table(app, flow->cookie, netdev);
 	if (!nfp_flow) {