Merge branch 'net-smc-add-failover-processing'

Karsten Graul says:

====================
net/smc: add failover processing

This patch series adds the actual SMC-R link failover processing and
improved link group termination. There will be one more (very small)
series after this which will complete the SMC-R link failover support.
====================

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

11 files changed, 514 insertions, 75 deletions

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index c67272007f41..4e4421c95ca1 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -390,6 +390,7 @@ static int smcr_clnt_conf_first_link(struct smc_sock *smc)
 				      SMC_CLC_DECLINE, CLC_WAIT_TIME_SHORT);
 		return rc == -EAGAIN ? SMC_CLC_DECL_TIMEOUT_CL : rc;
 	}
+	smc_llc_save_peer_uid(qentry);
 	rc = smc_llc_eval_conf_link(qentry, SMC_LLC_REQ);
 	smc_llc_flow_qentry_del(&link->lgr->llc_flow_lcl);
 	if (rc)
@@ -1056,6 +1057,7 @@ static int smcr_serv_conf_first_link(struct smc_sock *smc)
 				      SMC_CLC_DECLINE, CLC_WAIT_TIME_SHORT);
 		return rc == -EAGAIN ? SMC_CLC_DECL_TIMEOUT_CL : rc;
 	}
+	smc_llc_save_peer_uid(qentry);
 	rc = smc_llc_eval_conf_link(qentry, SMC_LLC_RESP);
 	smc_llc_flow_qentry_del(&link->lgr->llc_flow_lcl);
 	if (rc)
diff --git a/net/smc/smc.h b/net/smc/smc.h
index 1a084afa7372..6f1c42da7a4c 100644
--- a/net/smc/smc.h
+++ b/net/smc/smc.h
@@ -143,6 +143,9 @@ struct smc_connection {
 						 * .prod cf. TCP snd_nxt
 						 * .cons cf. TCP sends ack
 						 */
+	union smc_host_cursor	local_tx_ctrl_fin;
+						/* prod crsr - confirmed by peer
+						 */
 	union smc_host_cursor	tx_curs_prep;	/* tx - prepared data
 						 * snd_max..wmem_alloc
 						 */
@@ -154,6 +157,7 @@ struct smc_connection {
 						 */
 	atomic_t		sndbuf_space;	/* remaining space in sndbuf */
 	u16			tx_cdc_seq;	/* sequence # for CDC send */
+	u16			tx_cdc_seq_fin;	/* sequence # - tx completed */
 	spinlock_t		send_lock;	/* protect wr_sends */
 	struct delayed_work	tx_work;	/* retry of smc_cdc_msg_send */
 	u32			tx_off;		/* base offset in peer rmb */
@@ -184,12 +188,14 @@ struct smc_connection {
 	spinlock_t		acurs_lock;	/* protect cursors */
 #endif
 	struct work_struct	close_work;	/* peer sent some closing */
+	struct work_struct	abort_work;	/* abort the connection */
 	struct tasklet_struct	rx_tsklet;	/* Receiver tasklet for SMC-D */
 	u8			rx_off;		/* receive offset:
 						 * 0 for SMC-R, 32 for SMC-D
 						 */
 	u64			peer_token;	/* SMC-D token of peer */
 	u8			killed : 1;	/* abnormal termination */
+	u8			out_of_sync : 1; /* out of sync with peer */
 };
 
 struct smc_sock {				/* smc sock container */
diff --git a/net/smc/smc_cdc.c b/net/smc/smc_cdc.c
index f64589d823aa..b2b85e1be72c 100644
--- a/net/smc/smc_cdc.c
+++ b/net/smc/smc_cdc.c
@@ -47,17 +47,20 @@ static void smc_cdc_tx_handler(struct smc_wr_tx_pend_priv *pnd_snd,
 		/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
 		smp_mb__after_atomic();
 		smc_curs_copy(&conn->tx_curs_fin, &cdcpend->cursor, conn);
+		smc_curs_copy(&conn->local_tx_ctrl_fin, &cdcpend->p_cursor,
+			      conn);
+		conn->tx_cdc_seq_fin = cdcpend->ctrl_seq;
 	}
 	smc_tx_sndbuf_nonfull(smc);
 	bh_unlock_sock(&smc->sk);
 }
 
 int smc_cdc_get_free_slot(struct smc_connection *conn,
+			  struct smc_link *link,
 			  struct smc_wr_buf **wr_buf,
 			  struct smc_rdma_wr **wr_rdma_buf,
 			  struct smc_cdc_tx_pend **pend)
 {
-	struct smc_link *link = conn->lnk;
 	int rc;
 
 	rc = smc_wr_tx_get_free_slot(link, smc_cdc_tx_handler, wr_buf,
@@ -104,22 +107,64 @@ int smc_cdc_msg_send(struct smc_connection *conn,
 	if (!rc) {
 		smc_curs_copy(&conn->rx_curs_confirmed, &cfed, conn);
 		conn->local_rx_ctrl.prod_flags.cons_curs_upd_req = 0;
+	} else {
+		conn->tx_cdc_seq--;
+		conn->local_tx_ctrl.seqno = conn->tx_cdc_seq;
 	}
 
 	return rc;
 }
 
+/* send a validation msg indicating the move of a conn to an other QP link */
+int smcr_cdc_msg_send_validation(struct smc_connection *conn)
+{
+	struct smc_host_cdc_msg *local = &conn->local_tx_ctrl;
+	struct smc_link *link = conn->lnk;
+	struct smc_cdc_tx_pend *pend;
+	struct smc_wr_buf *wr_buf;
+	struct smc_cdc_msg *peer;
+	int rc;
+
+	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, NULL, &pend);
+	if (rc)
+		return rc;
+
+	peer = (struct smc_cdc_msg *)wr_buf;
+	peer->common.type = local->common.type;
+	peer->len = local->len;
+	peer->seqno = htons(conn->tx_cdc_seq_fin); /* seqno last compl. tx */
+	peer->token = htonl(local->token);
+	peer->prod_flags.failover_validation = 1;
+
+	rc = smc_wr_tx_send(link, (struct smc_wr_tx_pend_priv *)pend);
+	return rc;
+}
+
 static int smcr_cdc_get_slot_and_msg_send(struct smc_connection *conn)
 {
 	struct smc_cdc_tx_pend *pend;
 	struct smc_wr_buf *wr_buf;
+	struct smc_link *link;
+	bool again = false;
 	int rc;
 
-	rc = smc_cdc_get_free_slot(conn, &wr_buf, NULL, &pend);
+again:
+	link = conn->lnk;
+	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, NULL, &pend);
 	if (rc)
 		return rc;
 
 	spin_lock_bh(&conn->send_lock);
+	if (link != conn->lnk) {
+		/* link of connection changed, try again one time*/
+		spin_unlock_bh(&conn->send_lock);
+		smc_wr_tx_put_slot(link,
+				   (struct smc_wr_tx_pend_priv *)pend);
+		if (again)
+			return -ENOLINK;
+		again = true;
+		goto again;
+	}
 	rc = smc_cdc_msg_send(conn, wr_buf, pend);
 	spin_unlock_bh(&conn->send_lock);
 	return rc;
@@ -237,6 +282,28 @@ static void smc_cdc_handle_urg_data_arrival(struct smc_sock *smc,
 	sk_send_sigurg(&smc->sk);
 }
 
+static void smc_cdc_msg_validate(struct smc_sock *smc, struct smc_cdc_msg *cdc,
+				 struct smc_link *link)
+{
+	struct smc_connection *conn = &smc->conn;
+	u16 recv_seq = ntohs(cdc->seqno);
+	s16 diff;
+
+	/* check that seqnum was seen before */
+	diff = conn->local_rx_ctrl.seqno - recv_seq;
+	if (diff < 0) { /* diff larger than 0x7fff */
+		/* drop connection */
+		conn->out_of_sync = 1;	/* prevent any further receives */
+		spin_lock_bh(&conn->send_lock);
+		conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
+		conn->lnk = link;
+		spin_unlock_bh(&conn->send_lock);
+		sock_hold(&smc->sk); /* sock_put in abort_work */
+		if (!schedule_work(&conn->abort_work))
+			sock_put(&smc->sk);
+	}
+}
+
 static void smc_cdc_msg_recv_action(struct smc_sock *smc,
 				    struct smc_cdc_msg *cdc)
 {
@@ -367,16 +434,19 @@ static void smc_cdc_rx_handler(struct ib_wc *wc, void *buf)
 	read_lock_bh(&lgr->conns_lock);
 	conn = smc_lgr_find_conn(ntohl(cdc->token), lgr);
 	read_unlock_bh(&lgr->conns_lock);
-	if (!conn)
+	if (!conn || conn->out_of_sync)
 		return;
 	smc = container_of(conn, struct smc_sock, conn);
 
-	if (!cdc->prod_flags.failover_validation) {
-		if (smc_cdc_before(ntohs(cdc->seqno),
-				   conn->local_rx_ctrl.seqno))
-			/* received seqno is old */
-			return;
+	if (cdc->prod_flags.failover_validation) {
+		smc_cdc_msg_validate(smc, cdc, link);
+		return;
 	}
+	if (smc_cdc_before(ntohs(cdc->seqno),
+			   conn->local_rx_ctrl.seqno))
+		/* received seqno is old */
+		return;
+
 	smc_cdc_msg_recv(smc, cdc);
 }
 
diff --git a/net/smc/smc_cdc.h b/net/smc/smc_cdc.h
index 861dc24c588c..9cfabc9af120 100644
--- a/net/smc/smc_cdc.h
+++ b/net/smc/smc_cdc.h
@@ -304,6 +304,7 @@ struct smc_cdc_tx_pend {
 };
 
 int smc_cdc_get_free_slot(struct smc_connection *conn,
+			  struct smc_link *link,
 			  struct smc_wr_buf **wr_buf,
 			  struct smc_rdma_wr **wr_rdma_buf,
 			  struct smc_cdc_tx_pend **pend);
@@ -312,6 +313,7 @@ int smc_cdc_msg_send(struct smc_connection *conn, struct smc_wr_buf *wr_buf,
 		     struct smc_cdc_tx_pend *pend);
 int smc_cdc_get_slot_and_msg_send(struct smc_connection *conn);
 int smcd_cdc_msg_send(struct smc_connection *conn);
+int smcr_cdc_msg_send_validation(struct smc_connection *conn);
 int smc_cdc_init(void) __init;
 void smcd_cdc_rx_init(struct smc_connection *conn);
 
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index 32a6cadc5c1f..fb5f685ff494 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -121,16 +121,60 @@ static void smc_lgr_add_alert_token(struct smc_connection *conn)
 	rb_insert_color(&conn->alert_node, &conn->lgr->conns_all);
 }
 
+/* assign an SMC-R link to the connection */
+static int smcr_lgr_conn_assign_link(struct smc_connection *conn, bool first)
+{
+	enum smc_link_state expected = first ? SMC_LNK_ACTIVATING :
+				       SMC_LNK_ACTIVE;
+	int i, j;
+
+	/* do link balancing */
+	for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
+		struct smc_link *lnk = &conn->lgr->lnk[i];
+
+		if (lnk->state != expected || lnk->link_is_asym)
+			continue;
+		if (conn->lgr->role == SMC_CLNT) {
+			conn->lnk = lnk; /* temporary, SMC server assigns link*/
+			break;
+		}
+		if (conn->lgr->conns_num % 2) {
+			for (j = i + 1; j < SMC_LINKS_PER_LGR_MAX; j++) {
+				struct smc_link *lnk2;
+
+				lnk2 = &conn->lgr->lnk[j];
+				if (lnk2->state == expected &&
+				    !lnk2->link_is_asym) {
+					conn->lnk = lnk2;
+					break;
+				}
+			}
+		}
+		if (!conn->lnk)
+			conn->lnk = lnk;
+		break;
+	}
+	if (!conn->lnk)
+		return SMC_CLC_DECL_NOACTLINK;
+	return 0;
+}
+
 /* Register connection in link group by assigning an alert token
  * registered in a search tree.
  * Requires @conns_lock
  * Note that '0' is a reserved value and not assigned.
  */
-static int smc_lgr_register_conn(struct smc_connection *conn)
+static int smc_lgr_register_conn(struct smc_connection *conn, bool first)
 {
 	struct smc_sock *smc = container_of(conn, struct smc_sock, conn);
 	static atomic_t nexttoken = ATOMIC_INIT(0);
+	int rc;
 
+	if (!conn->lgr->is_smcd) {
+		rc = smcr_lgr_conn_assign_link(conn, first);
+		if (rc)
+			return rc;
+	}
 	/* find a new alert_token_local value not yet used by some connection
 	 * in this link group
 	 */
@@ -141,22 +185,6 @@ static int smc_lgr_register_conn(struct smc_connection *conn)
 			conn->alert_token_local = 0;
 	}
 	smc_lgr_add_alert_token(conn);
-
-	/* assign the new connection to a link */
-	if (!conn->lgr->is_smcd) {
-		struct smc_link *lnk;
-		int i;
-
-		/* tbd - link balancing */
-		for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
-			lnk = &conn->lgr->lnk[i];
-			if (lnk->state == SMC_LNK_ACTIVATING ||
-			    lnk->state == SMC_LNK_ACTIVE)
-				conn->lnk = lnk;
-		}
-		if (!conn->lnk)
-			return SMC_CLC_DECL_NOACTLINK;
-	}
 	conn->lgr->conns_num++;
 	return 0;
 }
@@ -209,6 +237,19 @@ void smc_lgr_cleanup_early(struct smc_connection *conn)
 	smc_lgr_schedule_free_work_fast(lgr);
 }
 
+static void smcr_lgr_link_deactivate_all(struct smc_link_group *lgr)
+{
+	int i;
+
+	for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
+		struct smc_link *lnk = &lgr->lnk[i];
+
+		if (smc_link_usable(lnk))
+			lnk->state = SMC_LNK_INACTIVE;
+	}
+	wake_up_interruptible_all(&lgr->llc_waiter);
+}
+
 static void smc_lgr_free(struct smc_link_group *lgr);
 
 static void smc_lgr_free_work(struct work_struct *work)
@@ -218,7 +259,6 @@ static void smc_lgr_free_work(struct work_struct *work)
 						  free_work);
 	spinlock_t *lgr_lock;
 	bool conns;
-	int i;
 
 	smc_lgr_list_head(lgr, &lgr_lock);
 	spin_lock_bh(lgr_lock);
@@ -238,17 +278,13 @@ static void smc_lgr_free_work(struct work_struct *work)
 	spin_unlock_bh(lgr_lock);
 	cancel_delayed_work(&lgr->free_work);
 
+	if (!lgr->is_smcd && !lgr->terminating)
+		smc_llc_send_link_delete_all(lgr, true,
+					     SMC_LLC_DEL_PROG_INIT_TERM);
 	if (lgr->is_smcd && !lgr->terminating)
 		smc_ism_signal_shutdown(lgr);
-	if (!lgr->is_smcd) {
-		for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
-			struct smc_link *lnk = &lgr->lnk[i];
-
-			if (smc_link_usable(lnk))
-				lnk->state = SMC_LNK_INACTIVE;
-		}
-		wake_up_interruptible_all(&lgr->llc_waiter);
-	}
+	if (!lgr->is_smcd)
+		smcr_lgr_link_deactivate_all(lgr);
 	smc_lgr_free(lgr);
 }
 
@@ -295,6 +331,7 @@ int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk,
 	lnk->smcibdev = ini->ib_dev;
 	lnk->ibport = ini->ib_port;
 	lnk->path_mtu = ini->ib_dev->pattr[ini->ib_port - 1].active_mtu;
+	smc_llc_link_set_uid(lnk);
 	INIT_WORK(&lnk->link_down_wrk, smc_link_down_work);
 	if (!ini->ib_dev->initialized) {
 		rc = (int)smc_ib_setup_per_ibdev(ini->ib_dev);
@@ -432,6 +469,135 @@ out:
 	return rc;
 }
 
+static int smc_write_space(struct smc_connection *conn)
+{
+	int buffer_len = conn->peer_rmbe_size;
+	union smc_host_cursor prod;
+	union smc_host_cursor cons;
+	int space;
+
+	smc_curs_copy(&prod, &conn->local_tx_ctrl.prod, conn);
+	smc_curs_copy(&cons, &conn->local_rx_ctrl.cons, conn);
+	/* determine rx_buf space */
+	space = buffer_len - smc_curs_diff(buffer_len, &cons, &prod);
+	return space;
+}
+
+static int smc_switch_cursor(struct smc_sock *smc)
+{
+	struct smc_connection *conn = &smc->conn;
+	union smc_host_cursor cons, fin;
+	int rc = 0;
+	int diff;
+
+	smc_curs_copy(&conn->tx_curs_sent, &conn->tx_curs_fin, conn);
+	smc_curs_copy(&fin, &conn->local_tx_ctrl_fin, conn);
+	/* set prod cursor to old state, enforce tx_rdma_writes() */
+	smc_curs_copy(&conn->local_tx_ctrl.prod, &fin, conn);
+	smc_curs_copy(&cons, &conn->local_rx_ctrl.cons, conn);
+
+	if (smc_curs_comp(conn->peer_rmbe_size, &cons, &fin) < 0) {
+		/* cons cursor advanced more than fin, and prod was set
+		 * fin above, so now prod is smaller than cons. Fix that.
+		 */
+		diff = smc_curs_diff(conn->peer_rmbe_size, &fin, &cons);
+		smc_curs_add(conn->sndbuf_desc->len,
+			     &conn->tx_curs_sent, diff);
+		smc_curs_add(conn->sndbuf_desc->len,
+			     &conn->tx_curs_fin, diff);
+
+		smp_mb__before_atomic();
+		atomic_add(diff, &conn->sndbuf_space);
+		smp_mb__after_atomic();
+
+		smc_curs_add(conn->peer_rmbe_size,
+			     &conn->local_tx_ctrl.prod, diff);
+		smc_curs_add(conn->peer_rmbe_size,
+			     &conn->local_tx_ctrl_fin, diff);
+	}
+	/* recalculate, value is used by tx_rdma_writes() */
+	atomic_set(&smc->conn.peer_rmbe_space, smc_write_space(conn));
+
+	if (smc->sk.sk_state != SMC_INIT &&
+	    smc->sk.sk_state != SMC_CLOSED) {
+		rc = smcr_cdc_msg_send_validation(conn);
+		if (!rc) {
+			schedule_delayed_work(&conn->tx_work, 0);
+			smc->sk.sk_data_ready(&smc->sk);
+		}
+	}
+	return rc;
+}
+
+struct smc_link *smc_switch_conns(struct smc_link_group *lgr,
+				  struct smc_link *from_lnk, bool is_dev_err)
+{
+	struct smc_link *to_lnk = NULL;
+	struct smc_connection *conn;
+	struct smc_sock *smc;
+	struct rb_node *node;
+	int i, rc = 0;
+
+	/* link is inactive, wake up tx waiters */
+	smc_wr_wakeup_tx_wait(from_lnk);
+
+	for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
+		if (lgr->lnk[i].state != SMC_LNK_ACTIVE ||
+		    i == from_lnk->link_idx)
+			continue;
+		if (is_dev_err && from_lnk->smcibdev == lgr->lnk[i].smcibdev &&
+		    from_lnk->ibport == lgr->lnk[i].ibport) {
+			continue;
+		}
+		to_lnk = &lgr->lnk[i];
+		break;
+	}
+	if (!to_lnk) {
+		smc_lgr_terminate_sched(lgr);
+		return NULL;
+	}
+again:
+	read_lock_bh(&lgr->conns_lock);
+	for (node = rb_first(&lgr->conns_all); node; node = rb_next(node)) {
+		conn = rb_entry(node, struct smc_connection, alert_node);
+		if (conn->lnk != from_lnk)
+			continue;
+		smc = container_of(conn, struct smc_sock, conn);
+		/* conn->lnk not yet set in SMC_INIT state */
+		if (smc->sk.sk_state == SMC_INIT)
+			continue;
+		if (smc->sk.sk_state == SMC_CLOSED ||
+		    smc->sk.sk_state == SMC_PEERCLOSEWAIT1 ||
+		    smc->sk.sk_state == SMC_PEERCLOSEWAIT2 ||
+		    smc->sk.sk_state == SMC_APPFINCLOSEWAIT ||
+		    smc->sk.sk_state == SMC_APPCLOSEWAIT1 ||
+		    smc->sk.sk_state == SMC_APPCLOSEWAIT2 ||
+		    smc->sk.sk_state == SMC_PEERFINCLOSEWAIT ||
+		    smc->sk.sk_state == SMC_PEERABORTWAIT ||
+		    smc->sk.sk_state == SMC_PROCESSABORT) {
+			spin_lock_bh(&conn->send_lock);
+			conn->lnk = to_lnk;
+			spin_unlock_bh(&conn->send_lock);
+			continue;
+		}
+		sock_hold(&smc->sk);
+		read_unlock_bh(&lgr->conns_lock);
+		/* avoid race with smcr_tx_sndbuf_nonempty() */
+		spin_lock_bh(&conn->send_lock);
+		conn->lnk = to_lnk;
+		rc = smc_switch_cursor(smc);
+		spin_unlock_bh(&conn->send_lock);
+		sock_put(&smc->sk);
+		if (rc) {
+			smcr_link_down_cond_sched(to_lnk);
+			return NULL;
+		}
+		goto again;
+	}
+	read_unlock_bh(&lgr->conns_lock);
+	return to_lnk;
+}
+
 static void smcr_buf_unuse(struct smc_buf_desc *rmb_desc,
 			   struct smc_link_group *lgr)
 {
@@ -486,6 +652,8 @@ void smc_conn_free(struct smc_connection *conn)
 		tasklet_kill(&conn->rx_tsklet);
 	} else {
 		smc_cdc_tx_dismiss_slots(conn);
+		if (current_work() != &conn->abort_work)
+			cancel_work_sync(&conn->abort_work);
 	}
 	if (!list_empty(&lgr->list)) {
 		smc_lgr_unregister_conn(conn);
@@ -640,6 +808,16 @@ static void smc_lgr_free(struct smc_link_group *lgr)
 {
 	int i;
 
+	if (!lgr->is_smcd) {
+		mutex_lock(&lgr->llc_conf_mutex);
+		for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
+			if (lgr->lnk[i].state != SMC_LNK_UNUSED)
+				smcr_link_clear(&lgr->lnk[i]);
+		}
+		mutex_unlock(&lgr->llc_conf_mutex);
+		smc_llc_lgr_clear(lgr);
+	}
+
 	smc_lgr_free_bufs(lgr);
 	if (lgr->is_smcd) {
 		if (!lgr->terminating) {
@@ -649,11 +827,6 @@ static void smc_lgr_free(struct smc_link_group *lgr)
 		if (!atomic_dec_return(&lgr->smcd->lgr_cnt))
 			wake_up(&lgr->smcd->lgrs_deleted);
 	} else {
-		for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
-			if (lgr->lnk[i].state != SMC_LNK_UNUSED)
-				smcr_link_clear(&lgr->lnk[i]);
-		}
-		smc_llc_lgr_clear(lgr);
 		if (!atomic_dec_return(&lgr_cnt))
 			wake_up(&lgrs_deleted);
 	}
@@ -708,21 +881,18 @@ static void smc_conn_kill(struct smc_connection *conn, bool soft)
 
 static void smc_lgr_cleanup(struct smc_link_group *lgr)
 {
-	int i;
-
 	if (lgr->is_smcd) {
 		smc_ism_signal_shutdown(lgr);
 		smcd_unregister_all_dmbs(lgr);
 		smc_ism_put_vlan(lgr->smcd, lgr->vlan_id);
 		put_device(&lgr->smcd->dev);
 	} else {
-		for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
-			struct smc_link *lnk = &lgr->lnk[i];
+		u32 rsn = lgr->llc_termination_rsn;
 
-			if (smc_link_usable(lnk))
-				lnk->state = SMC_LNK_INACTIVE;
-		}
-		wake_up_interruptible_all(&lgr->llc_waiter);
+		if (!rsn)
+			rsn = SMC_LLC_DEL_PROG_INIT_TERM;
+		smc_llc_send_link_delete_all(lgr, false, rsn);
+		smcr_lgr_link_deactivate_all(lgr);
 	}
 }
 
@@ -738,8 +908,8 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr, bool soft)
 
 	if (lgr->terminating)
 		return;	/* lgr already terminating */
-	if (!soft)
-		cancel_delayed_work_sync(&lgr->free_work);
+	/* cancel free_work sync, will terminate when lgr->freeing is set */
+	cancel_delayed_work_sync(&lgr->free_work);
 	lgr->terminating = 1;
 
 	/* kill remaining link group connections */
@@ -759,10 +929,7 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr, bool soft)
 	}
 	read_unlock_bh(&lgr->conns_lock);
 	smc_lgr_cleanup(lgr);
-	if (soft)
-		smc_lgr_schedule_free_work_fast(lgr);
-	else
-		smc_lgr_free(lgr);
+	smc_lgr_free(lgr);
 }
 
 /* unlink link group and schedule termination */
@@ -777,6 +944,7 @@ void smc_lgr_terminate_sched(struct smc_link_group *lgr)
 		return;	/* lgr already terminating */
 	}
 	list_del_init(&lgr->list);
+	lgr->freeing = 1;
 	spin_unlock_bh(lgr_lock);
 	schedule_work(&lgr->terminate_work);
 }
@@ -795,6 +963,7 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan)
 			if (peer_gid) /* peer triggered termination */
 				lgr->peer_shutdown = 1;
 			list_move(&lgr->list, &lgr_free_list);
+			lgr->freeing = 1;
 		}
 	}
 	spin_unlock_bh(&dev->lgr_lock);
@@ -854,6 +1023,7 @@ void smc_smcr_terminate_all(struct smc_ib_device *smcibdev)
 
 	list_for_each_entry_safe(lgr, lg, &lgr_free_list, list) {
 		list_del_init(&lgr->list);
+		smc_llc_set_termination_rsn(lgr, SMC_LLC_DEL_OP_INIT_TERM);
 		__smc_lgr_terminate(lgr, false);
 	}
 
@@ -867,6 +1037,37 @@ void smc_smcr_terminate_all(struct smc_ib_device *smcibdev)
 	}
 }
 
+/* set new lgr type and clear all asymmetric link tagging */
+void smcr_lgr_set_type(struct smc_link_group *lgr, enum smc_lgr_type new_type)
+{
+	int i;
+
+	for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++)
+		if (smc_link_usable(&lgr->lnk[i]))
+			lgr->lnk[i].link_is_asym = false;
+	lgr->type = new_type;
+}
+
+/* set new lgr type and tag a link as asymmetric */
+void smcr_lgr_set_type_asym(struct smc_link_group *lgr,
+			    enum smc_lgr_type new_type, int asym_lnk_idx)
+{
+	smcr_lgr_set_type(lgr, new_type);
+	lgr->lnk[asym_lnk_idx].link_is_asym = true;
+}
+
+/* abort connection, abort_work scheduled from tasklet context */
+static void smc_conn_abort_work(struct work_struct *work)
+{
+	struct smc_connection *conn = container_of(work,
+						   struct smc_connection,
+						   abort_work);
+	struct smc_sock *smc = container_of(conn, struct smc_sock, conn);
+
+	smc_conn_kill(conn, true);
+	sock_put(&smc->sk); /* sock_hold done by schedulers of abort_work */
+}
+
 /* link is up - establish alternate link if applicable */
 static void smcr_link_up(struct smc_link_group *lgr,
 			 struct smc_ib_device *smcibdev, u8 ibport)
@@ -943,13 +1144,12 @@ static void smcr_link_down(struct smc_link *lnk)
 		return;
 
 	smc_ib_modify_qp_reset(lnk);
-	to_lnk = NULL;
-	/* tbd: call to_lnk = smc_switch_conns(lgr, lnk, true); */
+	to_lnk = smc_switch_conns(lgr, lnk, true);
 	if (!to_lnk) { /* no backup link available */
 		smcr_link_clear(lnk);
 		return;
 	}
-	lgr->type = SMC_LGR_SINGLE;
+	smcr_lgr_set_type(lgr, SMC_LGR_SINGLE);
 	del_link_id = lnk->link_id;
 
 	if (lgr->role == SMC_SERV) {
@@ -1138,7 +1338,7 @@ int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini)
 			/* link group found */
 			ini->cln_first_contact = SMC_REUSE_CONTACT;
 			conn->lgr = lgr;
-			rc = smc_lgr_register_conn(conn); /* add conn to lgr */
+			rc = smc_lgr_register_conn(conn, false);
 			write_unlock_bh(&lgr->conns_lock);
 			if (!rc && delayed_work_pending(&lgr->free_work))
 				cancel_delayed_work(&lgr->free_work);
@@ -1166,7 +1366,7 @@ create:
 			goto out;
 		lgr = conn->lgr;
 		write_lock_bh(&lgr->conns_lock);
-		rc = smc_lgr_register_conn(conn); /* add smc conn to lgr */
+		rc = smc_lgr_register_conn(conn, true);
 		write_unlock_bh(&lgr->conns_lock);
 		if (rc)
 			goto out;
@@ -1174,6 +1374,7 @@ create:
 	conn->local_tx_ctrl.common.type = SMC_CDC_MSG_TYPE;
 	conn->local_tx_ctrl.len = SMC_WR_TX_SIZE;
 	conn->urg_state = SMC_URG_READ;
+	INIT_WORK(&smc->conn.abort_work, smc_conn_abort_work);
 	if (ini->is_smcd) {
 		conn->rx_off = sizeof(struct smcd_cdc_msg);
 		smcd_cdc_rx_init(conn); /* init tasklet for this conn */
diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h
index 7fe53feb9dc4..4ae76802214f 100644
--- a/net/smc/smc_core.h
+++ b/net/smc/smc_core.h
@@ -70,6 +70,8 @@ struct smc_rdma_wr {				/* work requests per message
 	struct ib_rdma_wr	wr_tx_rdma[SMC_MAX_RDMA_WRITES];
 };
 
+#define SMC_LGR_ID_SIZE		4
+
 struct smc_link {
 	struct smc_ib_device	*smcibdev;	/* ib-device */
 	u8			ibport;		/* port - values 1 | 2 */
@@ -85,6 +87,7 @@ struct smc_link {
 	struct smc_rdma_sges	*wr_tx_rdma_sges;/*RDMA WRITE gather meta data*/
 	struct smc_rdma_wr	*wr_tx_rdmas;	/* WR RDMA WRITE */
 	struct smc_wr_tx_pend	*wr_tx_pends;	/* WR send waiting for CQE */
+	struct completion	*wr_tx_compl;	/* WR send CQE completion */
 	/* above four vectors have wr_tx_cnt elements and use the same index */
 	dma_addr_t		wr_tx_dma_addr;	/* DMA address of wr_tx_bufs */
 	atomic_long_t		wr_tx_id;	/* seq # of last sent WR */
@@ -115,7 +118,10 @@ struct smc_link {
 	u8			peer_mac[ETH_ALEN];	/* = gid[8:10||13:15] */
 	u8			peer_gid[SMC_GID_SIZE];	/* gid of peer*/
 	u8			link_id;	/* unique # within link group */
+	u8			link_uid[SMC_LGR_ID_SIZE]; /* unique lnk id */
+	u8			peer_link_uid[SMC_LGR_ID_SIZE]; /* peer uid */
 	u8			link_idx;	/* index in lgr link array */
+	u8			link_is_asym;	/* is link asymmetric? */
 	struct smc_link_group	*lgr;		/* parent link group */
 	struct work_struct	link_down_wrk;	/* wrk to bring link down */
 
@@ -176,7 +182,6 @@ struct smc_rtoken {				/* address/key of remote RMB */
 	u32			rkey;
 };
 
-#define SMC_LGR_ID_SIZE		4
 #define SMC_BUF_MIN_SIZE	16384	/* minimum size of an RMB */
 #define SMC_RMBE_SIZES		16	/* number of distinct RMBE sizes */
 /* theoretically, the RFC states that largest size would be 512K,
@@ -269,6 +274,8 @@ struct smc_link_group {
 						/* protects llc flow */
 			int			llc_testlink_time;
 						/* link keep alive time */
+			u32			llc_termination_rsn;
+						/* rsn code for termination */
 		};
 		struct { /* SMC-D */
 			u64			peer_gid;
@@ -379,7 +386,12 @@ int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk,
 void smcr_link_clear(struct smc_link *lnk);
 int smcr_buf_map_lgr(struct smc_link *lnk);
 int smcr_buf_reg_lgr(struct smc_link *lnk);
+void smcr_lgr_set_type(struct smc_link_group *lgr, enum smc_lgr_type new_type);
+void smcr_lgr_set_type_asym(struct smc_link_group *lgr,
+			    enum smc_lgr_type new_type, int asym_lnk_idx);
 int smcr_link_reg_rmb(struct smc_link *link, struct smc_buf_desc *rmb_desc);
+struct smc_link *smc_switch_conns(struct smc_link_group *lgr,
+				  struct smc_link *from_lnk, bool is_dev_err);
 void smcr_link_down_cond(struct smc_link *lnk);
 void smcr_link_down_cond_sched(struct smc_link *lnk);
 
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 7675ccd6f3c3..66ddc9cf5e2f 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -361,7 +361,6 @@ static int smc_llc_add_pending_send(struct smc_link *link,
 int smc_llc_send_confirm_link(struct smc_link *link,
 			      enum smc_llc_reqresp reqresp)
 {
-	struct smc_link_group *lgr = smc_get_lgr(link);
 	struct smc_llc_msg_confirm_link *confllc;
 	struct smc_wr_tx_pend_priv *pend;
 	struct smc_wr_buf *wr_buf;
@@ -382,7 +381,7 @@ int smc_llc_send_confirm_link(struct smc_link *link,
 	memcpy(confllc->sender_gid, link->gid, SMC_GID_SIZE);
 	hton24(confllc->sender_qp_num, link->roce_qp->qp_num);
 	confllc->link_num = link->link_id;
-	memcpy(confllc->link_uid, lgr->id, SMC_LGR_ID_SIZE);
+	memcpy(confllc->link_uid, link->link_uid, SMC_LGR_ID_SIZE);
 	confllc->max_links = SMC_LLC_ADD_LNK_MAX_LINKS;
 	/* send llc message */
 	rc = smc_wr_tx_send(link, pend);
@@ -560,6 +559,25 @@ static int smc_llc_send_message(struct smc_link *link, void *llcbuf)
 	return smc_wr_tx_send(link, pend);
 }
 
+/* schedule an llc send on link, may wait for buffers,
+ * and wait for send completion notification.
+ * @return 0 on success
+ */
+static int smc_llc_send_message_wait(struct smc_link *link, void *llcbuf)
+{
+	struct smc_wr_tx_pend_priv *pend;
+	struct smc_wr_buf *wr_buf;
+	int rc;
+
+	if (!smc_link_usable(link))
+		return -ENOLINK;
+	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
+	if (rc)
+		return rc;
+	memcpy(wr_buf, llcbuf, sizeof(union smc_llc_msg));
+	return smc_wr_tx_send_wait(link, pend, SMC_LLC_WAIT_TIME);
+}
+
 /********************************* receive ***********************************/
 
 static int smc_llc_alloc_alt_link(struct smc_link_group *lgr,
@@ -752,6 +770,7 @@ static int smc_llc_cli_conf_link(struct smc_link *link,
 		smc_llc_flow_qentry_del(&lgr->llc_flow_lcl);
 		return -ENOLINK;
 	}
+	smc_llc_save_peer_uid(qentry);
 	smc_llc_flow_qentry_del(&lgr->llc_flow_lcl);
 
 	rc = smc_ib_modify_qp_rts(link_new);
@@ -777,7 +796,11 @@ static int smc_llc_cli_conf_link(struct smc_link *link,
 		return -ENOLINK;
 	}
 	smc_llc_link_active(link_new);
-	lgr->type = lgr_new_t;
+	if (lgr_new_t == SMC_LGR_ASYMMETRIC_LOCAL ||
+	    lgr_new_t == SMC_LGR_ASYMMETRIC_PEER)
+		smcr_lgr_set_type_asym(lgr, lgr_new_t, link_new->link_idx);
+	else
+		smcr_lgr_set_type(lgr, lgr_new_t);
 	return 0;
 }
 
@@ -822,7 +845,8 @@ int smc_llc_cli_add_link(struct smc_link *link, struct smc_llc_qentry *qentry)
 	if (rc)
 		goto out_reject;
 	smc_llc_save_add_link_info(lnk_new, llc);
-	lnk_new->link_id = llc->link_num;
+	lnk_new->link_id = llc->link_num;	/* SMC server assigns link id */
+	smc_llc_link_set_uid(lnk_new);
 
 	rc = smc_ib_ready_link(lnk_new);
 	if (rc)
@@ -933,7 +957,7 @@ static void smc_llc_delete_asym_link(struct smc_link_group *lgr)
 		return; /* no asymmetric link */
 	if (!smc_link_downing(&lnk_asym->state))
 		return;
-	/* tbd: lnk_new = smc_switch_conns(lgr, lnk_asym, false); */
+	lnk_new = smc_switch_conns(lgr, lnk_asym, false);
 	smc_wr_tx_wait_no_pending_sends(lnk_asym);
 	if (!lnk_new)
 		goto out_free;
@@ -1018,8 +1042,13 @@ static int smc_llc_srv_conf_link(struct smc_link *link,
 					 false, SMC_LLC_DEL_LOST_PATH);
 		return -ENOLINK;
 	}
+	smc_llc_save_peer_uid(qentry);
 	smc_llc_link_active(link_new);
-	lgr->type = lgr_new_t;
+	if (lgr_new_t == SMC_LGR_ASYMMETRIC_LOCAL ||
+	    lgr_new_t == SMC_LGR_ASYMMETRIC_PEER)
+		smcr_lgr_set_type_asym(lgr, lgr_new_t, link_new->link_idx);
+	else
+		smcr_lgr_set_type(lgr, lgr_new_t);
 	smc_llc_flow_qentry_del(&lgr->llc_flow_lcl);
 	return 0;
 }
@@ -1195,7 +1224,7 @@ static void smc_llc_process_cli_delete_link(struct smc_link_group *lgr)
 	smc_llc_send_message(lnk, &qentry->msg); /* response */
 
 	if (smc_link_downing(&lnk_del->state)) {
-		/* tbd: call smc_switch_conns(lgr, lnk_del, false); */
+		smc_switch_conns(lgr, lnk_del, false);
 		smc_wr_tx_wait_no_pending_sends(lnk_del);
 	}
 	smcr_link_clear(lnk_del);
@@ -1204,9 +1233,9 @@ static void smc_llc_process_cli_delete_link(struct smc_link_group *lgr)
 	if (lnk_del == lnk_asym) {
 		/* expected deletion of asym link, don't change lgr state */
 	} else if (active_links == 1) {
-		lgr->type = SMC_LGR_SINGLE;
+		smcr_lgr_set_type(lgr, SMC_LGR_SINGLE);
 	} else if (!active_links) {
-		lgr->type = SMC_LGR_NONE;
+		smcr_lgr_set_type(lgr, SMC_LGR_NONE);
 		smc_lgr_terminate_sched(lgr);
 	}
 out_unlock:
@@ -1215,6 +1244,29 @@ out:
 	kfree(qentry);
 }
 
+/* try to send a DELETE LINK ALL request on any active link,
+ * waiting for send completion
+ */
+void smc_llc_send_link_delete_all(struct smc_link_group *lgr, bool ord, u32 rsn)
+{
+	struct smc_llc_msg_del_link delllc = {0};
+	int i;
+
+	delllc.hd.common.type = SMC_LLC_DELETE_LINK;
+	delllc.hd.length = sizeof(delllc);
+	if (ord)
+		delllc.hd.flags |= SMC_LLC_FLAG_DEL_LINK_ORDERLY;
+	delllc.hd.flags |= SMC_LLC_FLAG_DEL_LINK_ALL;
+	delllc.reason = htonl(rsn);
+
+	for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
+		if (!smc_link_usable(&lgr->lnk[i]))
+			continue;
+		if (!smc_llc_send_message_wait(&lgr->lnk[i], &delllc))
+			break;
+	}
+}
+
 static void smc_llc_process_srv_delete_link(struct smc_link_group *lgr)
 {
 	struct smc_llc_msg_del_link *del_llc;
@@ -1230,6 +1282,8 @@ static void smc_llc_process_srv_delete_link(struct smc_link_group *lgr)
 
 	if (qentry->msg.delete_link.hd.flags & SMC_LLC_FLAG_DEL_LINK_ALL) {
 		/* delete entire lgr */
+		smc_llc_send_link_delete_all(lgr, true, ntohl(
+					      qentry->msg.delete_link.reason));
 		smc_lgr_terminate_sched(lgr);
 		goto out;
 	}
@@ -1245,7 +1299,7 @@ static void smc_llc_process_srv_delete_link(struct smc_link_group *lgr)
 		goto out; /* asymmetric link already deleted */
 
 	if (smc_link_downing(&lnk_del->state)) {
-		/* tbd: call smc_switch_conns(lgr, lnk_del, false); */
+		smc_switch_conns(lgr, lnk_del, false);
 		smc_wr_tx_wait_no_pending_sends(lnk_del);
 	}
 	if (!list_empty(&lgr->list)) {
@@ -1270,9 +1324,9 @@ static void smc_llc_process_srv_delete_link(struct smc_link_group *lgr)
 
 	active_links = smc_llc_active_link_count(lgr);
 	if (active_links == 1) {
-		lgr->type = SMC_LGR_SINGLE;
+		smcr_lgr_set_type(lgr, SMC_LGR_SINGLE);
 	} else if (!active_links) {
-		lgr->type = SMC_LGR_NONE;
+		smcr_lgr_set_type(lgr, SMC_LGR_NONE);
 		smc_lgr_terminate_sched(lgr);
 	}
 
@@ -1368,6 +1422,14 @@ static void smc_llc_rmt_delete_rkey(struct smc_link_group *lgr)
 	smc_llc_flow_qentry_del(&lgr->llc_flow_rmt);
 }
 
+static void smc_llc_protocol_violation(struct smc_link_group *lgr, u8 type)
+{
+	pr_warn_ratelimited("smc: SMC-R lg %*phN LLC protocol violation: "
+			    "llc_type %d\n", SMC_LGR_ID_SIZE, &lgr->id, type);
+	smc_llc_set_termination_rsn(lgr, SMC_LLC_DEL_PROT_VIOL);
+	smc_lgr_terminate_sched(lgr);
+}
+
 /* flush the llc event queue */
 static void smc_llc_event_flush(struct smc_link_group *lgr)
 {
@@ -1468,6 +1530,9 @@ static void smc_llc_event_handler(struct smc_llc_qentry *qentry)
 			smc_llc_flow_stop(lgr, &lgr->llc_flow_rmt);
 		}
 		return;
+	default:
+		smc_llc_protocol_violation(lgr, llc->raw.hdr.common.type);
+		break;
 	}
 out:
 	kfree(qentry);
@@ -1527,6 +1592,9 @@ static void smc_llc_rx_response(struct smc_link *link,
 	case SMC_LLC_CONFIRM_RKEY_CONT:
 		/* not used because max links is 3 */
 		break;
+	default:
+		smc_llc_protocol_violation(link->lgr, llc_type);
+		break;
 	}
 	kfree(qentry);
 }
@@ -1709,12 +1777,29 @@ out:
 	return rc;
 }
 
+void smc_llc_link_set_uid(struct smc_link *link)
+{
+	__be32 link_uid;
+
+	link_uid = htonl(*((u32 *)link->lgr->id) + link->link_id);
+	memcpy(link->link_uid, &link_uid, SMC_LGR_ID_SIZE);
+}
+
+/* save peers link user id, used for debug purposes */
+void smc_llc_save_peer_uid(struct smc_llc_qentry *qentry)
+{
+	memcpy(qentry->link->peer_link_uid, qentry->msg.confirm_link.link_uid,
+	       SMC_LGR_ID_SIZE);
+}
+
 /* evaluate confirm link request or response */
 int smc_llc_eval_conf_link(struct smc_llc_qentry *qentry,
 			   enum smc_llc_reqresp type)
 {
-	if (type == SMC_LLC_REQ)	/* SMC server assigns link_id */
+	if (type == SMC_LLC_REQ) {	/* SMC server assigns link_id */
 		qentry->link->link_id = qentry->msg.confirm_link.link_num;
+		smc_llc_link_set_uid(qentry->link);
+	}
 	if (!(qentry->msg.raw.hdr.flags & SMC_LLC_FLAG_NO_RMBE_EYEC))
 		return -ENOTSUPP;
 	return 0;
diff --git a/net/smc/smc_llc.h b/net/smc/smc_llc.h
index c335fc5f363c..55287376112d 100644
--- a/net/smc/smc_llc.h
+++ b/net/smc/smc_llc.h
@@ -60,6 +60,14 @@ static inline struct smc_link *smc_llc_usable_link(struct smc_link_group *lgr)
 	return NULL;
 }
 
+/* set the termination reason code for the link group */
+static inline void smc_llc_set_termination_rsn(struct smc_link_group *lgr,
+					       u32 rsn)
+{
+	if (!lgr->llc_termination_rsn)
+		lgr->llc_termination_rsn = rsn;
+}
+
 /* transmit */
 int smc_llc_send_confirm_link(struct smc_link *lnk,
 			      enum smc_llc_reqresp reqresp);
@@ -84,11 +92,15 @@ int smc_llc_flow_initiate(struct smc_link_group *lgr,
 void smc_llc_flow_stop(struct smc_link_group *lgr, struct smc_llc_flow *flow);
 int smc_llc_eval_conf_link(struct smc_llc_qentry *qentry,
 			   enum smc_llc_reqresp type);
+void smc_llc_link_set_uid(struct smc_link *link);
+void smc_llc_save_peer_uid(struct smc_llc_qentry *qentry);
 struct smc_llc_qentry *smc_llc_wait(struct smc_link_group *lgr,
 				    struct smc_link *lnk,
 				    int time_out, u8 exp_msg);
 struct smc_llc_qentry *smc_llc_flow_qentry_clr(struct smc_llc_flow *flow);
 void smc_llc_flow_qentry_del(struct smc_llc_flow *flow);
+void smc_llc_send_link_delete_all(struct smc_link_group *lgr, bool ord,
+				  u32 rsn);
 int smc_llc_cli_add_link(struct smc_link *link, struct smc_llc_qentry *qentry);
 int smc_llc_srv_add_link(struct smc_link *link);
 void smc_llc_srv_add_link_local(struct smc_link *link);
diff --git a/net/smc/smc_tx.c b/net/smc/smc_tx.c
index 417204572a69..54ba0443847e 100644
--- a/net/smc/smc_tx.c
+++ b/net/smc/smc_tx.c
@@ -482,12 +482,13 @@ static int smc_tx_rdma_writes(struct smc_connection *conn,
 static int smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
 {
 	struct smc_cdc_producer_flags *pflags = &conn->local_tx_ctrl.prod_flags;
+	struct smc_link *link = conn->lnk;
 	struct smc_rdma_wr *wr_rdma_buf;
 	struct smc_cdc_tx_pend *pend;
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
-	rc = smc_cdc_get_free_slot(conn, &wr_buf, &wr_rdma_buf, &pend);
+	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, &wr_rdma_buf, &pend);
 	if (rc < 0) {
 		if (rc == -EBUSY) {
 			struct smc_sock *smc =
@@ -505,10 +506,17 @@ static int smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
 	}
 
 	spin_lock_bh(&conn->send_lock);
+	if (link != conn->lnk) {
+		/* link of connection changed, tx_work will restart */
+		smc_wr_tx_put_slot(link,
+				   (struct smc_wr_tx_pend_priv *)pend);
+		rc = -ENOLINK;
+		goto out_unlock;
+	}
 	if (!pflags->urg_data_present) {
 		rc = smc_tx_rdma_writes(conn, wr_rdma_buf);
 		if (rc) {
-			smc_wr_tx_put_slot(conn->lnk,
+			smc_wr_tx_put_slot(link,
 					   (struct smc_wr_tx_pend_priv *)pend);
 			goto out_unlock;
 		}
diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c
index 3fd27bea4f7a..7239ba9b99dc 100644
--- a/net/smc/smc_wr.c
+++ b/net/smc/smc_wr.c
@@ -44,6 +44,7 @@ struct smc_wr_tx_pend {	/* control data for a pending send request */
 	struct smc_link		*link;
 	u32			idx;
 	struct smc_wr_tx_pend_priv priv;
+	u8			compl_requested;
 };
 
 /******************************** send queue *********************************/
@@ -103,6 +104,8 @@ static inline void smc_wr_tx_process_cqe(struct ib_wc *wc)
 	if (pnd_snd_idx == link->wr_tx_cnt)
 		return;
 	link->wr_tx_pends[pnd_snd_idx].wc_status = wc->status;
+	if (link->wr_tx_pends[pnd_snd_idx].compl_requested)
+		complete(&link->wr_tx_compl[pnd_snd_idx]);
 	memcpy(&pnd_snd, &link->wr_tx_pends[pnd_snd_idx], sizeof(pnd_snd));
 	/* clear the full struct smc_wr_tx_pend including .priv */
 	memset(&link->wr_tx_pends[pnd_snd_idx], 0,
@@ -275,6 +278,33 @@ int smc_wr_tx_send(struct smc_link *link, struct smc_wr_tx_pend_priv *priv)
 	return rc;
 }
 
+/* Send prepared WR slot via ib_post_send and wait for send completion
+ * notification.
+ * @priv: pointer to smc_wr_tx_pend_priv identifying prepared message buffer
+ */
+int smc_wr_tx_send_wait(struct smc_link *link, struct smc_wr_tx_pend_priv *priv,
+			unsigned long timeout)
+{
+	struct smc_wr_tx_pend *pend;
+	int rc;
+
+	pend = container_of(priv, struct smc_wr_tx_pend, priv);
+	pend->compl_requested = 1;
+	init_completion(&link->wr_tx_compl[pend->idx]);
+
+	rc = smc_wr_tx_send(link, priv);
+	if (rc)
+		return rc;
+	/* wait for completion by smc_wr_tx_process_cqe() */
+	rc = wait_for_completion_interruptible_timeout(
+					&link->wr_tx_compl[pend->idx], timeout);
+	if (rc <= 0)
+		rc = -ENODATA;
+	if (rc > 0)
+		rc = 0;
+	return rc;
+}
+
 /* Register a memory region and wait for result. */
 int smc_wr_reg_send(struct smc_link *link, struct ib_mr *mr)
 {
@@ -555,6 +585,8 @@ void smc_wr_free_link(struct smc_link *lnk)
 
 void smc_wr_free_link_mem(struct smc_link *lnk)
 {
+	kfree(lnk->wr_tx_compl);
+	lnk->wr_tx_compl = NULL;
 	kfree(lnk->wr_tx_pends);
 	lnk->wr_tx_pends = NULL;
 	kfree(lnk->wr_tx_mask);
@@ -625,8 +657,15 @@ int smc_wr_alloc_link_mem(struct smc_link *link)
 				    GFP_KERNEL);
 	if (!link->wr_tx_pends)
 		goto no_mem_wr_tx_mask;
+	link->wr_tx_compl = kcalloc(SMC_WR_BUF_CNT,
+				    sizeof(link->wr_tx_compl[0]),
+				    GFP_KERNEL);
+	if (!link->wr_tx_compl)
+		goto no_mem_wr_tx_pends;
 	return 0;
 
+no_mem_wr_tx_pends:
+	kfree(link->wr_tx_pends);
 no_mem_wr_tx_mask:
 	kfree(link->wr_tx_mask);
 no_mem_wr_rx_sges:
diff --git a/net/smc/smc_wr.h b/net/smc/smc_wr.h
index f7eaeb3391f3..423b8709f1c9 100644
--- a/net/smc/smc_wr.h
+++ b/net/smc/smc_wr.h
@@ -101,6 +101,8 @@ int smc_wr_tx_put_slot(struct smc_link *link,
 		       struct smc_wr_tx_pend_priv *wr_pend_priv);
 int smc_wr_tx_send(struct smc_link *link,
 		   struct smc_wr_tx_pend_priv *wr_pend_priv);
+int smc_wr_tx_send_wait(struct smc_link *link, struct smc_wr_tx_pend_priv *priv,
+			unsigned long timeout);
 void smc_wr_tx_cq_handler(struct ib_cq *ib_cq, void *cq_context);
 void smc_wr_tx_dismiss_slots(struct smc_link *lnk, u8 wr_rx_hdr_type,
 			     smc_wr_tx_filter filter,