diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
index 5491a41fd1be..0273f40b85f7 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
@@ -3527,6 +3527,10 @@ DEFINE_SHOW_ATTRIBUTE(meminfo);
 
 static int chcr_stats_show(struct seq_file *seq, void *v)
 {
+#if IS_ENABLED(CONFIG_CHELSIO_TLS_DEVICE)
+	struct ch_ktls_port_stats_debug *ktls_port;
+	int i = 0;
+#endif
 	struct adapter *adap = seq->private;
 
 	seq_puts(seq, "Chelsio Crypto Accelerator Stats \n");
@@ -3557,18 +3561,6 @@ static int chcr_stats_show(struct seq_file *seq, void *v)
 	seq_puts(seq, "\nChelsio KTLS Crypto Accelerator Stats\n");
 	seq_printf(seq, "Tx TLS offload refcount:          %20u\n",
 		   refcount_read(&adap->chcr_ktls.ktls_refcount));
-	seq_printf(seq, "Tx HW offload contexts added:     %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_ctx));
-	seq_printf(seq, "Tx connection created:            %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_connection_open));
-	seq_printf(seq, "Tx connection failed:             %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_connection_fail));
-	seq_printf(seq, "Tx connection closed:             %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_connection_close));
-	seq_printf(seq, "Packets passed for encryption :   %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_encrypted_packets));
-	seq_printf(seq, "Bytes passed for encryption :     %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_encrypted_bytes));
 	seq_printf(seq, "Tx records send:                  %20llu\n",
 		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_send_records));
 	seq_printf(seq, "Tx partial start of records:      %20llu\n",
@@ -3581,14 +3573,17 @@ static int chcr_stats_show(struct seq_file *seq, void *v)
 		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_complete_pkts));
 	seq_printf(seq, "TX trim pkts :                    %20llu\n",
 		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_trimmed_pkts));
-	seq_printf(seq, "Tx out of order packets:          %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_ooo));
-	seq_printf(seq, "Tx drop pkts before HW offload:   %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_skip_no_sync_data));
-	seq_printf(seq, "Tx drop not synced packets:       %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_drop_no_sync_data));
-	seq_printf(seq, "Tx drop bypass req:               %20llu\n",
-		   atomic64_read(&adap->ch_ktls_stats.ktls_tx_drop_bypass_req));
+	while (i < MAX_NPORTS) {
+		ktls_port = &adap->ch_ktls_stats.ktls_port[i];
+		seq_printf(seq, "Port %d\n", i);
+		seq_printf(seq, "Tx connection created:            %20llu\n",
+			   atomic64_read(&ktls_port->ktls_tx_connection_open));
+		seq_printf(seq, "Tx connection failed:             %20llu\n",
+			   atomic64_read(&ktls_port->ktls_tx_connection_fail));
+		seq_printf(seq, "Tx connection closed:             %20llu\n",
+			   atomic64_read(&ktls_port->ktls_tx_connection_close));
+		i++;
+	}
 #endif
 	return 0;
 }
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c
index 8c6c9bedcba1..61ea3ec5c3fc 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c
@@ -117,14 +117,6 @@ static const char stats_strings[][ETH_GSTRING_LEN] = {
 	"vlan_insertions        ",
 	"gro_packets            ",
 	"gro_merged             ",
-};
-
-static char adapter_stats_strings[][ETH_GSTRING_LEN] = {
-	"db_drop                ",
-	"db_full                ",
-	"db_empty               ",
-	"write_coal_success     ",
-	"write_coal_fail        ",
 #if  IS_ENABLED(CONFIG_CHELSIO_TLS_DEVICE)
 	"tx_tls_encrypted_packets",
 	"tx_tls_encrypted_bytes  ",
@@ -136,6 +128,14 @@ static char adapter_stats_strings[][ETH_GSTRING_LEN] = {
 #endif
 };
 
+static char adapter_stats_strings[][ETH_GSTRING_LEN] = {
+	"db_drop                ",
+	"db_full                ",
+	"db_empty               ",
+	"write_coal_success     ",
+	"write_coal_fail        ",
+};
+
 static char loopback_stats_strings[][ETH_GSTRING_LEN] = {
 	"-------Loopback----------- ",
 	"octets_ok              ",
@@ -257,14 +257,6 @@ struct queue_port_stats {
 	u64 vlan_ins;
 	u64 gro_pkts;
 	u64 gro_merged;
-};
-
-struct adapter_stats {
-	u64 db_drop;
-	u64 db_full;
-	u64 db_empty;
-	u64 wc_success;
-	u64 wc_fail;
 #if IS_ENABLED(CONFIG_CHELSIO_TLS_DEVICE)
 	u64 tx_tls_encrypted_packets;
 	u64 tx_tls_encrypted_bytes;
@@ -276,12 +268,23 @@ struct adapter_stats {
 #endif
 };
 
+struct adapter_stats {
+	u64 db_drop;
+	u64 db_full;
+	u64 db_empty;
+	u64 wc_success;
+	u64 wc_fail;
+};
+
 static void collect_sge_port_stats(const struct adapter *adap,
 				   const struct port_info *p,
 				   struct queue_port_stats *s)
 {
 	const struct sge_eth_txq *tx = &adap->sge.ethtxq[p->first_qset];
 	const struct sge_eth_rxq *rx = &adap->sge.ethrxq[p->first_qset];
+#if IS_ENABLED(CONFIG_CHELSIO_TLS_DEVICE)
+	const struct ch_ktls_port_stats_debug *ktls_stats;
+#endif
 	struct sge_eohw_txq *eohw_tx;
 	unsigned int i;
 
@@ -306,6 +309,21 @@ static void collect_sge_port_stats(const struct adapter *adap,
 			s->vlan_ins += eohw_tx->vlan_ins;
 		}
 	}
+#if IS_ENABLED(CONFIG_CHELSIO_TLS_DEVICE)
+	ktls_stats = &adap->ch_ktls_stats.ktls_port[p->port_id];
+	s->tx_tls_encrypted_packets =
+		atomic64_read(&ktls_stats->ktls_tx_encrypted_packets);
+	s->tx_tls_encrypted_bytes =
+		atomic64_read(&ktls_stats->ktls_tx_encrypted_bytes);
+	s->tx_tls_ctx = atomic64_read(&ktls_stats->ktls_tx_ctx);
+	s->tx_tls_ooo = atomic64_read(&ktls_stats->ktls_tx_ooo);
+	s->tx_tls_skip_no_sync_data =
+		atomic64_read(&ktls_stats->ktls_tx_skip_no_sync_data);
+	s->tx_tls_drop_no_sync_data =
+		atomic64_read(&ktls_stats->ktls_tx_drop_no_sync_data);
+	s->tx_tls_drop_bypass_req =
+		atomic64_read(&ktls_stats->ktls_tx_drop_bypass_req);
+#endif
 }
 
 static void collect_adapter_stats(struct adapter *adap, struct adapter_stats *s)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.c
index b154190e1ee2..743af9e654aa 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.c
@@ -690,8 +690,8 @@ int cxgb4_set_ktls_feature(struct adapter *adap, bool enable)
 			 * ULD is/are already active, return failure.
 			 */
 			if (cxgb4_uld_in_use(adap)) {
-				dev_warn(adap->pdev_dev,
-					 "ULD connections (tid/stid) active. Can't enable kTLS\n");
+				dev_dbg(adap->pdev_dev,
+					"ULD connections (tid/stid) active. Can't enable kTLS\n");
 				return -EINVAL;
 			}
 			ret = t4_set_params(adap, adap->mbox, adap->pf,
@@ -699,7 +699,7 @@ int cxgb4_set_ktls_feature(struct adapter *adap, bool enable)
 			if (ret)
 				return ret;
 			refcount_set(&adap->chcr_ktls.ktls_refcount, 1);
-			pr_info("kTLS has been enabled. Restrictions placed on ULD support\n");
+			pr_debug("kTLS has been enabled. Restrictions placed on ULD support\n");
 		} else {
 			/* ktls settings already up, just increment refcount. */
 			refcount_inc(&adap->chcr_ktls.ktls_refcount);
@@ -716,7 +716,7 @@ int cxgb4_set_ktls_feature(struct adapter *adap, bool enable)
 					    0, 1, &params, &params);
 			if (ret)
 				return ret;
-			pr_info("kTLS is disabled. Restrictions on ULD support removed\n");
+			pr_debug("kTLS is disabled. Restrictions on ULD support removed\n");
 		}
 	}
 
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
index ea2fabbdd934..b169776ab484 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
@@ -44,6 +44,7 @@
 #include "cxgb4.h"
 
 #define MAX_ULD_QSETS 16
+#define MAX_ULD_NPORTS 4
 
 /* CPL message priority levels */
 enum {
@@ -365,17 +366,10 @@ struct cxgb4_virt_res {                      /* virtualized HW resources */
 };
 
 #if IS_ENABLED(CONFIG_CHELSIO_TLS_DEVICE)
-struct ch_ktls_stats_debug {
+struct ch_ktls_port_stats_debug {
 	atomic64_t ktls_tx_connection_open;
 	atomic64_t ktls_tx_connection_fail;
 	atomic64_t ktls_tx_connection_close;
-	atomic64_t ktls_tx_send_records;
-	atomic64_t ktls_tx_end_pkts;
-	atomic64_t ktls_tx_start_pkts;
-	atomic64_t ktls_tx_middle_pkts;
-	atomic64_t ktls_tx_retransmit_pkts;
-	atomic64_t ktls_tx_complete_pkts;
-	atomic64_t ktls_tx_trimmed_pkts;
 	atomic64_t ktls_tx_encrypted_packets;
 	atomic64_t ktls_tx_encrypted_bytes;
 	atomic64_t ktls_tx_ctx;
@@ -384,6 +378,17 @@ struct ch_ktls_stats_debug {
 	atomic64_t ktls_tx_drop_no_sync_data;
 	atomic64_t ktls_tx_drop_bypass_req;
 };
+
+struct ch_ktls_stats_debug {
+	struct ch_ktls_port_stats_debug ktls_port[MAX_ULD_NPORTS];
+	atomic64_t ktls_tx_send_records;
+	atomic64_t ktls_tx_end_pkts;
+	atomic64_t ktls_tx_start_pkts;
+	atomic64_t ktls_tx_middle_pkts;
+	atomic64_t ktls_tx_retransmit_pkts;
+	atomic64_t ktls_tx_complete_pkts;
+	atomic64_t ktls_tx_trimmed_pkts;
+};
 #endif
 
 struct chcr_stats_debug {
diff --git a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c
index 4609f1f78426..5195f692f14d 100644
--- a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c
+++ b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c
@@ -125,60 +125,6 @@ out:
 	return ret;
 }
 
-static int chcr_ktls_update_connection_state(struct chcr_ktls_info *tx_info,
-					     int new_state)
-{
-	/* This function can be called from both rx (interrupt context) and tx
-	 * queue contexts.
-	 */
-	spin_lock_bh(&tx_info->lock);
-	switch (tx_info->connection_state) {
-	case KTLS_CONN_CLOSED:
-		tx_info->connection_state = new_state;
-		break;
-
-	case KTLS_CONN_ACT_OPEN_REQ:
-		/* only go forward if state is greater than current state. */
-		if (new_state <= tx_info->connection_state)
-			break;
-		/* update to the next state and also initialize TCB */
-		tx_info->connection_state = new_state;
-		fallthrough;
-	case KTLS_CONN_ACT_OPEN_RPL:
-		/* if we are stuck in this state, means tcb init might not
-		 * received by HW, try sending it again.
-		 */
-		if (!chcr_init_tcb_fields(tx_info))
-			tx_info->connection_state = KTLS_CONN_SET_TCB_REQ;
-		break;
-
-	case KTLS_CONN_SET_TCB_REQ:
-		/* only go forward if state is greater than current state. */
-		if (new_state <= tx_info->connection_state)
-			break;
-		/* update to the next state and check if l2t_state is valid  */
-		tx_info->connection_state = new_state;
-		fallthrough;
-	case KTLS_CONN_SET_TCB_RPL:
-		/* Check if l2t state is valid, then move to ready state. */
-		if (cxgb4_check_l2t_valid(tx_info->l2te)) {
-			tx_info->connection_state = KTLS_CONN_TX_READY;
-			atomic64_inc(&tx_info->adap->ch_ktls_stats.ktls_tx_ctx);
-		}
-		break;
-
-	case KTLS_CONN_TX_READY:
-		/* nothing to be done here */
-		break;
-
-	default:
-		pr_err("unknown KTLS connection state\n");
-		break;
-	}
-	spin_unlock_bh(&tx_info->lock);
-
-	return tx_info->connection_state;
-}
 /*
  * chcr_ktls_act_open_req: creates TCB entry for ipv4 connection.
  * @sk - tcp socket.
@@ -298,27 +244,17 @@ static int chcr_setup_connection(struct sock *sk,
 		return -EINVAL;
 
 	tx_info->atid = atid;
-	tx_info->ip_family = sk->sk_family;
 
-	if (sk->sk_family == AF_INET) {
-		tx_info->ip_family = AF_INET;
+	if (tx_info->ip_family == AF_INET) {
 		ret = chcr_ktls_act_open_req(sk, tx_info, atid);
 #if IS_ENABLED(CONFIG_IPV6)
 	} else {
-		if (!sk->sk_ipv6only &&
-		    ipv6_addr_type(&sk->sk_v6_daddr) == IPV6_ADDR_MAPPED) {
-			tx_info->ip_family = AF_INET;
-			ret = chcr_ktls_act_open_req(sk, tx_info, atid);
-		} else {
-			tx_info->ip_family = AF_INET6;
-			ret = cxgb4_clip_get(tx_info->netdev,
-					     (const u32 *)
-					     &sk->sk_v6_rcv_saddr.s6_addr,
-					     1);
-			if (ret)
-				goto out;
-			ret = chcr_ktls_act_open_req6(sk, tx_info, atid);
-		}
+		ret = cxgb4_clip_get(tx_info->netdev, (const u32 *)
+				     &sk->sk_v6_rcv_saddr,
+				     1);
+		if (ret)
+			return ret;
+		ret = chcr_ktls_act_open_req6(sk, tx_info, atid);
 #endif
 	}
 
@@ -326,16 +262,21 @@ static int chcr_setup_connection(struct sock *sk,
 	 * success, if any other return type clear atid and return that failure.
 	 */
 	if (ret) {
-		if (ret == NET_XMIT_CN)
+		if (ret == NET_XMIT_CN) {
 			ret = 0;
-		else
+		} else {
+#if IS_ENABLED(CONFIG_IPV6)
+			/* clear clip entry */
+			if (tx_info->ip_family == AF_INET6)
+				cxgb4_clip_release(tx_info->netdev,
+						   (const u32 *)
+						   &sk->sk_v6_rcv_saddr,
+						   1);
+#endif
 			cxgb4_free_atid(t, atid);
-		goto out;
+		}
 	}
 
-	/* update the connection state */
-	chcr_ktls_update_connection_state(tx_info, KTLS_CONN_ACT_OPEN_REQ);
-out:
 	return ret;
 }
 
@@ -396,15 +337,10 @@ static void chcr_ktls_dev_del(struct net_device *netdev,
 	struct chcr_ktls_ofld_ctx_tx *tx_ctx =
 				chcr_get_ktls_tx_context(tls_ctx);
 	struct chcr_ktls_info *tx_info = tx_ctx->chcr_info;
-	struct sock *sk;
+	struct ch_ktls_port_stats_debug *port_stats;
 
 	if (!tx_info)
 		return;
-	sk = tx_info->sk;
-
-	spin_lock(&tx_info->lock);
-	tx_info->connection_state = KTLS_CONN_CLOSED;
-	spin_unlock(&tx_info->lock);
 
 	/* clear l2t entry */
 	if (tx_info->l2te)
@@ -413,8 +349,8 @@ static void chcr_ktls_dev_del(struct net_device *netdev,
 #if IS_ENABLED(CONFIG_IPV6)
 	/* clear clip entry */
 	if (tx_info->ip_family == AF_INET6)
-		cxgb4_clip_release(netdev,
-				   (const u32 *)&sk->sk_v6_daddr.in6_u.u6_addr8,
+		cxgb4_clip_release(netdev, (const u32 *)
+				   &tx_info->sk->sk_v6_rcv_saddr,
 				   1);
 #endif
 
@@ -426,7 +362,8 @@ static void chcr_ktls_dev_del(struct net_device *netdev,
 				 tx_info->tid, tx_info->ip_family);
 	}
 
-	atomic64_inc(&tx_info->adap->ch_ktls_stats.ktls_tx_connection_close);
+	port_stats = &tx_info->adap->ch_ktls_stats.ktls_port[tx_info->port_id];
+	atomic64_inc(&port_stats->ktls_tx_connection_close);
 	kvfree(tx_info);
 	tx_ctx->chcr_info = NULL;
 	/* release module refcount */
@@ -448,6 +385,7 @@ static int chcr_ktls_dev_add(struct net_device *netdev, struct sock *sk,
 			     u32 start_offload_tcp_sn)
 {
 	struct tls_context *tls_ctx = tls_get_ctx(sk);
+	struct ch_ktls_port_stats_debug *port_stats;
 	struct chcr_ktls_ofld_ctx_tx *tx_ctx;
 	struct chcr_ktls_info *tx_info;
 	struct dst_entry *dst;
@@ -461,30 +399,23 @@ static int chcr_ktls_dev_add(struct net_device *netdev, struct sock *sk,
 
 	pi = netdev_priv(netdev);
 	adap = pi->adapter;
+	port_stats = &adap->ch_ktls_stats.ktls_port[pi->port_id];
+	atomic64_inc(&port_stats->ktls_tx_connection_open);
+
 	if (direction == TLS_OFFLOAD_CTX_DIR_RX) {
 		pr_err("not expecting for RX direction\n");
-		ret = -EINVAL;
 		goto out;
 	}
-	if (tx_ctx->chcr_info) {
-		ret = -EINVAL;
+
+	if (tx_ctx->chcr_info)
 		goto out;
-	}
 
 	tx_info = kvzalloc(sizeof(*tx_info), GFP_KERNEL);
-	if (!tx_info) {
-		ret = -ENOMEM;
+	if (!tx_info)
 		goto out;
-	}
-
-	spin_lock_init(&tx_info->lock);
-
-	/* clear connection state */
-	spin_lock(&tx_info->lock);
-	tx_info->connection_state = KTLS_CONN_CLOSED;
-	spin_unlock(&tx_info->lock);
 
 	tx_info->sk = sk;
+	spin_lock_init(&tx_info->lock);
 	/* initialize tid and atid to -1, 0 is a also a valid id. */
 	tx_info->tid = -1;
 	tx_info->atid = -1;
@@ -495,10 +426,12 @@ static int chcr_ktls_dev_add(struct net_device *netdev, struct sock *sk,
 	tx_info->tx_chan = pi->tx_chan;
 	tx_info->smt_idx = pi->smt_idx;
 	tx_info->port_id = pi->port_id;
+	tx_info->prev_ack = 0;
+	tx_info->prev_win = 0;
 
 	tx_info->rx_qid = chcr_get_first_rx_qid(adap);
 	if (unlikely(tx_info->rx_qid < 0))
-		goto out2;
+		goto free_tx_info;
 
 	tx_info->prev_seq = start_offload_tcp_sn;
 	tx_info->tcp_start_seq_number = start_offload_tcp_sn;
@@ -506,18 +439,22 @@ static int chcr_ktls_dev_add(struct net_device *netdev, struct sock *sk,
 	/* save crypto keys */
 	ret = chcr_ktls_save_keys(tx_info, crypto_info, direction);
 	if (ret < 0)
-		goto out2;
+		goto free_tx_info;
 
 	/* get peer ip */
 	if (sk->sk_family == AF_INET) {
 		memcpy(daaddr, &sk->sk_daddr, 4);
+		tx_info->ip_family = AF_INET;
 #if IS_ENABLED(CONFIG_IPV6)
 	} else {
 		if (!sk->sk_ipv6only &&
-		    ipv6_addr_type(&sk->sk_v6_daddr) == IPV6_ADDR_MAPPED)
+		    ipv6_addr_type(&sk->sk_v6_daddr) == IPV6_ADDR_MAPPED) {
 			memcpy(daaddr, &sk->sk_daddr, 4);
-		else
+			tx_info->ip_family = AF_INET;
+		} else {
 			memcpy(daaddr, sk->sk_v6_daddr.in6_u.u6_addr8, 16);
+			tx_info->ip_family = AF_INET6;
+		}
 #endif
 	}
 
@@ -525,13 +462,13 @@ static int chcr_ktls_dev_add(struct net_device *netdev, struct sock *sk,
 	dst = sk_dst_get(sk);
 	if (!dst) {
 		pr_err("DST entry not found\n");
-		goto out2;
+		goto free_tx_info;
 	}
 	n = dst_neigh_lookup(dst, daaddr);
 	if (!n || !n->dev) {
 		pr_err("neighbour not found\n");
 		dst_release(dst);
-		goto out2;
+		goto free_tx_info;
 	}
 	tx_info->l2te  = cxgb4_l2t_get(adap->l2t, n, n->dev, 0);
 
@@ -540,31 +477,86 @@ static int chcr_ktls_dev_add(struct net_device *netdev, struct sock *sk,
 
 	if (!tx_info->l2te) {
 		pr_err("l2t entry not found\n");
-		goto out2;
+		goto free_tx_info;
 	}
 
-	tx_ctx->chcr_info = tx_info;
+	/* Driver shouldn't be removed until any single connection exists */
+	if (!try_module_get(THIS_MODULE))
+		goto free_l2t;
 
+	init_completion(&tx_info->completion);
 	/* create a filter and call cxgb4_l2t_send to send the packet out, which
 	 * will take care of updating l2t entry in hw if not already done.
 	 */
-	ret = chcr_setup_connection(sk, tx_info);
-	if (ret)
-		goto out2;
+	tx_info->open_state = CH_KTLS_OPEN_PENDING;
 
-	/* Driver shouldn't be removed until any single connection exists */
-	if (!try_module_get(THIS_MODULE)) {
-		ret = -EINVAL;
-		goto out2;
+	if (chcr_setup_connection(sk, tx_info))
+		goto put_module;
+
+	/* Wait for reply */
+	wait_for_completion_timeout(&tx_info->completion, 30 * HZ);
+	spin_lock_bh(&tx_info->lock);
+	if (tx_info->open_state) {
+		/* need to wait for hw response, can't free tx_info yet. */
+		if (tx_info->open_state == CH_KTLS_OPEN_PENDING)
+			tx_info->pending_close = true;
+		/* free the lock after the cleanup */
+		goto put_module;
 	}
+	spin_unlock_bh(&tx_info->lock);
+
+	/* initialize tcb */
+	reinit_completion(&tx_info->completion);
+	/* mark it pending for hw response */
+	tx_info->open_state = CH_KTLS_OPEN_PENDING;
+
+	if (chcr_init_tcb_fields(tx_info))
+		goto free_tid;
+
+	/* Wait for reply */
+	wait_for_completion_timeout(&tx_info->completion, 30 * HZ);
+	spin_lock_bh(&tx_info->lock);
+	if (tx_info->open_state) {
+		/* need to wait for hw response, can't free tx_info yet. */
+		tx_info->pending_close = true;
+		/* free the lock after cleanup */
+		goto free_tid;
+	}
+	spin_unlock_bh(&tx_info->lock);
+
+	if (!cxgb4_check_l2t_valid(tx_info->l2te))
+		goto free_tid;
+
+	atomic64_inc(&port_stats->ktls_tx_ctx);
+	tx_ctx->chcr_info = tx_info;
 
-	atomic64_inc(&adap->ch_ktls_stats.ktls_tx_connection_open);
 	return 0;
-out2:
-	kvfree(tx_info);
+
+free_tid:
+	chcr_ktls_mark_tcb_close(tx_info);
+#if IS_ENABLED(CONFIG_IPV6)
+	/* clear clip entry */
+	if (tx_info->ip_family == AF_INET6)
+		cxgb4_clip_release(netdev, (const u32 *)
+				   &sk->sk_v6_rcv_saddr,
+				   1);
+#endif
+	cxgb4_remove_tid(&tx_info->adap->tids, tx_info->tx_chan,
+			 tx_info->tid, tx_info->ip_family);
+
+put_module:
+	/* release module refcount */
+	module_put(THIS_MODULE);
+free_l2t:
+	cxgb4_l2t_release(tx_info->l2te);
+free_tx_info:
+	if (tx_info->pending_close)
+		spin_unlock_bh(&tx_info->lock);
+	else
+		kvfree(tx_info);
 out:
-	atomic64_inc(&adap->ch_ktls_stats.ktls_tx_connection_fail);
-	return ret;
+	atomic64_inc(&port_stats->ktls_tx_connection_fail);
+	return -1;
 }
 
 /*
@@ -627,20 +619,39 @@ static int chcr_ktls_cpl_act_open_rpl(struct adapter *adap,
 	tx_info = lookup_atid(t, atid);
 
 	if (!tx_info || tx_info->atid != atid) {
-		pr_err("tx_info or atid is not correct\n");
+		pr_err("%s: incorrect tx_info or atid\n", __func__);
 		return -1;
 	}
 
+	cxgb4_free_atid(t, atid);
+	tx_info->atid = -1;
+
+	spin_lock(&tx_info->lock);
+	/* HW response is very close, finish pending cleanup */
+	if (tx_info->pending_close) {
+		spin_unlock(&tx_info->lock);
+		if (!status) {
+			/* it's a late success, tcb status is establised,
+			 * mark it close.
+			 */
+			chcr_ktls_mark_tcb_close(tx_info);
+			cxgb4_remove_tid(&tx_info->adap->tids, tx_info->tx_chan,
+					 tid, tx_info->ip_family);
+		}
+		kvfree(tx_info);
+		return 0;
+	}
+
 	if (!status) {
 		tx_info->tid = tid;
 		cxgb4_insert_tid(t, tx_info, tx_info->tid, tx_info->ip_family);
-
-		cxgb4_free_atid(t, atid);
-		tx_info->atid = -1;
-		/* update the connection state */
-		chcr_ktls_update_connection_state(tx_info,
-						  KTLS_CONN_ACT_OPEN_RPL);
+		tx_info->open_state = CH_KTLS_OPEN_SUCCESS;
+	} else {
+		tx_info->open_state = CH_KTLS_OPEN_FAILURE;
 	}
+	spin_unlock(&tx_info->lock);
+
+	complete(&tx_info->completion);
 	return 0;
 }
 
@@ -658,12 +669,22 @@ static int chcr_ktls_cpl_set_tcb_rpl(struct adapter *adap, unsigned char *input)
 
 	t = &adap->tids;
 	tx_info = lookup_tid(t, tid);
+
 	if (!tx_info || tx_info->tid != tid) {
-		pr_err("tx_info or atid is not correct\n");
+		pr_err("%s: incorrect tx_info or tid\n", __func__);
 		return -1;
 	}
-	/* update the connection state */
-	chcr_ktls_update_connection_state(tx_info, KTLS_CONN_SET_TCB_RPL);
+
+	spin_lock(&tx_info->lock);
+	if (tx_info->pending_close) {
+		spin_unlock(&tx_info->lock);
+		kvfree(tx_info);
+		return 0;
+	}
+	tx_info->open_state = false;
+	spin_unlock(&tx_info->lock);
+
+	complete(&tx_info->completion);
 	return 0;
 }
 
@@ -765,6 +786,7 @@ static int chcr_ktls_xmit_tcb_cpls(struct chcr_ktls_info *tx_info,
 				   u64 tcp_ack, u64 tcp_win)
 {
 	bool first_wr = ((tx_info->prev_ack == 0) && (tx_info->prev_win == 0));
+	struct ch_ktls_port_stats_debug *port_stats;
 	u32 len, cpl = 0, ndesc, wr_len;
 	struct fw_ulptx_wr *wr;
 	int credits;
@@ -798,12 +820,14 @@ static int chcr_ktls_xmit_tcb_cpls(struct chcr_ktls_info *tx_info,
 	/* reset snd una if it's a re-transmit pkt */
 	if (tcp_seq != tx_info->prev_seq) {
 		/* reset snd_una */
+		port_stats =
+			&tx_info->adap->ch_ktls_stats.ktls_port[tx_info->port_id];
 		pos = chcr_write_cpl_set_tcb_ulp(tx_info, q, tx_info->tid, pos,
 						 TCB_SND_UNA_RAW_W,
 						 TCB_SND_UNA_RAW_V
 						 (TCB_SND_UNA_RAW_M),
 						 TCB_SND_UNA_RAW_V(0), 0);
-		atomic64_inc(&tx_info->adap->ch_ktls_stats.ktls_tx_ooo);
+		atomic64_inc(&port_stats->ktls_tx_ooo);
 		cpl++;
 	}
 	/* update ack */
@@ -1836,6 +1860,7 @@ out:
 /* nic tls TX handler */
 static int chcr_ktls_xmit(struct sk_buff *skb, struct net_device *dev)
 {
+	struct ch_ktls_port_stats_debug *port_stats;
 	struct chcr_ktls_ofld_ctx_tx *tx_ctx;
 	struct ch_ktls_stats_debug *stats;
 	struct tcphdr *th = tcp_hdr(skb);
@@ -1845,7 +1870,6 @@ static int chcr_ktls_xmit(struct sk_buff *skb, struct net_device *dev)
 	u32 tls_end_offset, tcp_seq;
 	struct tls_context *tls_ctx;
 	struct sk_buff *local_skb;
-	int new_connection_state;
 	struct sge_eth_txq *q;
 	struct adapter *adap;
 	unsigned long flags;
@@ -1868,15 +1892,6 @@ static int chcr_ktls_xmit(struct sk_buff *skb, struct net_device *dev)
 	if (unlikely(!tx_info))
 		goto out;
 
-	/* check the connection state, we don't need to pass new connection
-	 * state, state machine will check and update the new state if it is
-	 * stuck due to responses not received from HW.
-	 * Start the tx handling only if state is KTLS_CONN_TX_READY.
-	 */
-	new_connection_state = chcr_ktls_update_connection_state(tx_info, 0);
-	if (new_connection_state != KTLS_CONN_TX_READY)
-		goto out;
-
 	/* don't touch the original skb, make a new skb to extract each records
 	 * and send them separately.
 	 */
@@ -1887,6 +1902,7 @@ static int chcr_ktls_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	adap = tx_info->adap;
 	stats = &adap->ch_ktls_stats;
+	port_stats = &stats->ktls_port[tx_info->port_id];
 
 	qidx = skb->queue_mapping;
 	q = &adap->sge.ethtxq[qidx + tx_info->first_qset];
@@ -1932,13 +1948,13 @@ static int chcr_ktls_xmit(struct sk_buff *skb, struct net_device *dev)
 		 */
 		if (unlikely(!record)) {
 			spin_unlock_irqrestore(&tx_ctx->base.lock, flags);
-			atomic64_inc(&stats->ktls_tx_drop_no_sync_data);
+			atomic64_inc(&port_stats->ktls_tx_drop_no_sync_data);
 			goto out;
 		}
 
 		if (unlikely(tls_record_is_start_marker(record))) {
 			spin_unlock_irqrestore(&tx_ctx->base.lock, flags);
-			atomic64_inc(&stats->ktls_tx_skip_no_sync_data);
+			atomic64_inc(&port_stats->ktls_tx_skip_no_sync_data);
 			goto out;
 		}
 
@@ -2009,9 +2025,8 @@ clear_ref:
 	} while (data_len > 0);
 
 	tx_info->prev_seq = ntohl(th->seq) + skb->data_len;
-
-	atomic64_inc(&stats->ktls_tx_encrypted_packets);
-	atomic64_add(skb->data_len, &stats->ktls_tx_encrypted_bytes);
+	atomic64_inc(&port_stats->ktls_tx_encrypted_packets);
+	atomic64_add(skb->data_len, &port_stats->ktls_tx_encrypted_bytes);
 
 	/* tcp finish is set, send a separate tcp msg including all the options
 	 * as well.
diff --git a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.h b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.h
index 4ae6ae38c406..c1651b1431a0 100644
--- a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.h
+++ b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.h
@@ -27,22 +27,20 @@
 #define CHCR_KTLS_WR_SIZE	(CHCR_PLAIN_TX_DATA_LEN +\
 				 sizeof(struct cpl_tx_sec_pdu))
 
-enum chcr_ktls_conn_state {
-	KTLS_CONN_CLOSED,
-	KTLS_CONN_ACT_OPEN_REQ,
-	KTLS_CONN_ACT_OPEN_RPL,
-	KTLS_CONN_SET_TCB_REQ,
-	KTLS_CONN_SET_TCB_RPL,
-	KTLS_CONN_TX_READY,
+enum ch_ktls_open_state {
+	CH_KTLS_OPEN_SUCCESS = 0,
+	CH_KTLS_OPEN_PENDING = 1,
+	CH_KTLS_OPEN_FAILURE = 2,
 };
 
 struct chcr_ktls_info {
 	struct sock *sk;
-	spinlock_t lock; /* state machine lock */
+	spinlock_t lock; /* lock for pending_close */
 	struct ktls_key_ctx key_ctx;
 	struct adapter *adap;
 	struct l2t_entry *l2te;
 	struct net_device *netdev;
+	struct completion completion;
 	u64 iv;
 	u64 record_no;
 	int tid;
@@ -58,13 +56,14 @@ struct chcr_ktls_info {
 	u32 tcp_start_seq_number;
 	u32 scmd0_short_seqno_numivs;
 	u32 scmd0_short_ivgen_hdrlen;
-	enum chcr_ktls_conn_state connection_state;
 	u16 prev_win;
 	u8 tx_chan;
 	u8 smt_idx;
 	u8 port_id;
 	u8 ip_family;
 	u8 first_qset;
+	enum ch_ktls_open_state open_state;
+	bool pending_close;
 };
 
 struct chcr_ktls_ofld_ctx_tx {