diff --git a/include/net/netfilter/nf_conntrack_act_ct.h b/include/net/netfilter/nf_conntrack_act_ct.h
index 078d3c52c03f..e5f2f0b73a9a 100644
--- a/include/net/netfilter/nf_conntrack_act_ct.h
+++ b/include/net/netfilter/nf_conntrack_act_ct.h
@@ -20,21 +20,6 @@ static inline struct nf_conn_act_ct_ext *nf_conn_act_ct_ext_find(const struct nf
 #endif
 }
 
-static inline struct nf_conn_act_ct_ext *nf_conn_act_ct_ext_add(struct nf_conn *ct)
-{
-#if IS_ENABLED(CONFIG_NET_ACT_CT)
-	struct nf_conn_act_ct_ext *act_ct = nf_ct_ext_find(ct, NF_CT_EXT_ACT_CT);
-
-	if (act_ct)
-		return act_ct;
-
-	act_ct = nf_ct_ext_add(ct, NF_CT_EXT_ACT_CT, GFP_ATOMIC);
-	return act_ct;
-#else
-	return NULL;
-#endif
-}
-
 static inline void nf_conn_act_ct_ext_fill(struct sk_buff *skb, struct nf_conn *ct,
 					   enum ip_conntrack_info ctinfo)
 {
@@ -47,4 +32,23 @@ static inline void nf_conn_act_ct_ext_fill(struct sk_buff *skb, struct nf_conn *
 #endif
 }
 
+static inline struct
+nf_conn_act_ct_ext *nf_conn_act_ct_ext_add(struct sk_buff *skb,
+					   struct nf_conn *ct,
+					   enum ip_conntrack_info ctinfo)
+{
+#if IS_ENABLED(CONFIG_NET_ACT_CT)
+	struct nf_conn_act_ct_ext *act_ct = nf_ct_ext_find(ct, NF_CT_EXT_ACT_CT);
+
+	if (act_ct)
+		return act_ct;
+
+	act_ct = nf_ct_ext_add(ct, NF_CT_EXT_ACT_CT, GFP_ATOMIC);
+	nf_conn_act_ct_ext_fill(skb, ct, ctinfo);
+	return act_ct;
+#else
+	return NULL;
+#endif
+}
+
 #endif /* _NF_CONNTRACK_ACT_CT_H */
diff --git a/net/openvswitch/conntrack.c b/net/openvswitch/conntrack.c
index 0b9a785dea45..3019a4406ca4 100644
--- a/net/openvswitch/conntrack.c
+++ b/net/openvswitch/conntrack.c
@@ -985,7 +985,7 @@ static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
 		if (err)
 			return err;
 
-		nf_conn_act_ct_ext_add(ct);
+		nf_conn_act_ct_ext_add(skb, ct, ctinfo);
 	} else if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
 		   labels_nonzero(&info->labels.mask)) {
 		err = ovs_ct_set_labels(ct, key, &info->labels.value,
diff --git a/net/sched/act_ct.c b/net/sched/act_ct.c
index 9583645e86c2..0db0ecf1d110 100644
--- a/net/sched/act_ct.c
+++ b/net/sched/act_ct.c
@@ -376,6 +376,17 @@ static void tcf_ct_flow_tc_ifidx(struct flow_offload *entry,
 	entry->tuplehash[dir].tuple.tc.iifidx = act_ct_ext->ifindex[dir];
 }
 
+static void tcf_ct_flow_ct_ext_ifidx_update(struct flow_offload *entry)
+{
+	struct nf_conn_act_ct_ext *act_ct_ext;
+
+	act_ct_ext = nf_conn_act_ct_ext_find(entry->ct);
+	if (act_ct_ext) {
+		tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_ORIGINAL);
+		tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_REPLY);
+	}
+}
+
 static void tcf_ct_flow_table_add(struct tcf_ct_flow_table *ct_ft,
 				  struct nf_conn *ct,
 				  bool tcp, bool bidirectional)
@@ -671,6 +682,8 @@ static bool tcf_ct_flow_table_lookup(struct tcf_ct_params *p,
 	else
 		ctinfo = IP_CT_ESTABLISHED_REPLY;
 
+	nf_conn_act_ct_ext_fill(skb, ct, ctinfo);
+	tcf_ct_flow_ct_ext_ifidx_update(flow);
 	flow_offload_refresh(nf_ft, flow, force_refresh);
 	if (!test_bit(IPS_ASSURED_BIT, &ct->status)) {
 		/* Process this flow in SW to allow promoting to ASSURED */
@@ -1034,7 +1047,7 @@ do_nat:
 		tcf_ct_act_set_labels(ct, p->labels, p->labels_mask);
 
 		if (!nf_ct_is_confirmed(ct))
-			nf_conn_act_ct_ext_add(ct);
+			nf_conn_act_ct_ext_add(skb, ct, ctinfo);
 
 		/* This will take care of sending queued events
 		 * even if the connection is already confirmed.