net/smc: replace sock_put worker by socket refcounting

Proper socket refcounting makes the sock_put worker obsolete.

Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

6 files changed, 88 insertions, 68 deletions

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 90c22a854f28..732a37ddbc21 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -115,7 +115,6 @@ static int smc_release(struct socket *sock)
 		goto out;
 
 	smc = smc_sk(sk);
-	sock_hold(sk);
 	if (sk->sk_state == SMC_LISTEN)
 		/* smc_close_non_accepted() is called and acquires
 		 * sock lock for child sockets again
@@ -124,10 +123,7 @@ static int smc_release(struct socket *sock)
 	else
 		lock_sock(sk);
 
-	if (smc->use_fallback) {
-		sk->sk_state = SMC_CLOSED;
-		sk->sk_state_change(sk);
-	} else {
+	if (!smc->use_fallback) {
 		rc = smc_close_active(smc);
 		sock_set_flag(sk, SOCK_DEAD);
 		sk->sk_shutdown |= SHUTDOWN_MASK;
@@ -136,20 +132,21 @@ static int smc_release(struct socket *sock)
 		sock_release(smc->clcsock);
 		smc->clcsock = NULL;
 	}
+	if (smc->use_fallback) {
+		sock_put(sk); /* passive closing */
+		sk->sk_state = SMC_CLOSED;
+		sk->sk_state_change(sk);
+	}
 
 	/* detach socket */
 	sock_orphan(sk);
 	sock->sk = NULL;
-	if (smc->use_fallback) {
-		schedule_delayed_work(&smc->sock_put_work, TCP_TIMEWAIT_LEN);
-	} else if (sk->sk_state == SMC_CLOSED) {
+	if (!smc->use_fallback && sk->sk_state == SMC_CLOSED)
 		smc_conn_free(&smc->conn);
-		schedule_delayed_work(&smc->sock_put_work,
-				      SMC_CLOSE_SOCK_PUT_DELAY);
-	}
 	release_sock(sk);
 
-	sock_put(sk);
+	sk->sk_prot->unhash(sk);
+	sock_put(sk); /* final sock_put */
 out:
 	return rc;
 }
@@ -181,7 +178,6 @@ static struct sock *smc_sock_alloc(struct net *net, struct socket *sock)
 	INIT_WORK(&smc->tcp_listen_work, smc_tcp_listen_work);
 	INIT_LIST_HEAD(&smc->accept_q);
 	spin_lock_init(&smc->accept_q_lock);
-	INIT_DELAYED_WORK(&smc->sock_put_work, smc_close_sock_put_work);
 	sk->sk_prot->hash(sk);
 	sk_refcnt_debug_inc(sk);
 
@@ -399,6 +395,8 @@ static int smc_connect_rdma(struct smc_sock *smc)
 	int rc = 0;
 	u8 ibport;
 
+	sock_hold(&smc->sk); /* sock put in passive closing */
+
 	if (!tcp_sk(smc->clcsock->sk)->syn_smc) {
 		/* peer has not signalled SMC-capability */
 		smc->use_fallback = true;
@@ -542,6 +540,8 @@ out_err_unlock:
 	mutex_unlock(&smc_create_lgr_pending);
 	smc_conn_free(&smc->conn);
 out_err:
+	if (smc->sk.sk_state == SMC_INIT)
+		sock_put(&smc->sk); /* passive closing */
 	return rc;
 }
 
@@ -620,7 +620,7 @@ static int smc_clcsock_accept(struct smc_sock *lsmc, struct smc_sock **new_smc)
 		new_sk->sk_state = SMC_CLOSED;
 		sock_set_flag(new_sk, SOCK_DEAD);
 		new_sk->sk_prot->unhash(new_sk);
-		sock_put(new_sk);
+		sock_put(new_sk); /* final */
 		*new_smc = NULL;
 		goto out;
 	}
@@ -637,7 +637,7 @@ static void smc_accept_enqueue(struct sock *parent, struct sock *sk)
 {
 	struct smc_sock *par = smc_sk(parent);
 
-	sock_hold(sk);
+	sock_hold(sk); /* sock_put in smc_accept_unlink () */
 	spin_lock(&par->accept_q_lock);
 	list_add_tail(&smc_sk(sk)->accept_q, &par->accept_q);
 	spin_unlock(&par->accept_q_lock);
@@ -653,7 +653,7 @@ static void smc_accept_unlink(struct sock *sk)
 	list_del_init(&smc_sk(sk)->accept_q);
 	spin_unlock(&par->accept_q_lock);
 	sk_acceptq_removed(&smc_sk(sk)->listen_smc->sk);
-	sock_put(sk);
+	sock_put(sk); /* sock_hold in smc_accept_enqueue */
 }
 
 /* remove a sock from the accept queue to bind it to a new socket created
@@ -671,7 +671,7 @@ struct sock *smc_accept_dequeue(struct sock *parent,
 		smc_accept_unlink(new_sk);
 		if (new_sk->sk_state == SMC_CLOSED) {
 			new_sk->sk_prot->unhash(new_sk);
-			sock_put(new_sk);
+			sock_put(new_sk); /* final */
 			continue;
 		}
 		if (new_sock)
@@ -686,14 +686,11 @@ void smc_close_non_accepted(struct sock *sk)
 {
 	struct smc_sock *smc = smc_sk(sk);
 
-	sock_hold(sk);
 	lock_sock(sk);
 	if (!sk->sk_lingertime)
 		/* wait for peer closing */
 		sk->sk_lingertime = SMC_MAX_STREAM_WAIT_TIMEOUT;
-	if (smc->use_fallback) {
-		sk->sk_state = SMC_CLOSED;
-	} else {
+	if (!smc->use_fallback) {
 		smc_close_active(smc);
 		sock_set_flag(sk, SOCK_DEAD);
 		sk->sk_shutdown |= SHUTDOWN_MASK;
@@ -706,14 +703,15 @@ void smc_close_non_accepted(struct sock *sk)
 		sock_release(tcp);
 	}
 	if (smc->use_fallback) {
-		schedule_delayed_work(&smc->sock_put_work, TCP_TIMEWAIT_LEN);
-	} else if (sk->sk_state == SMC_CLOSED) {
-		smc_conn_free(&smc->conn);
-		schedule_delayed_work(&smc->sock_put_work,
-				      SMC_CLOSE_SOCK_PUT_DELAY);
+		sock_put(sk); /* passive closing */
+		sk->sk_state = SMC_CLOSED;
+	} else {
+		if (sk->sk_state == SMC_CLOSED)
+			smc_conn_free(&smc->conn);
 	}
 	release_sock(sk);
-	sock_put(sk);
+	sk->sk_prot->unhash(sk);
+	sock_put(sk); /* final sock_put */
 }
 
 static int smc_serv_conf_first_link(struct smc_sock *smc)
@@ -937,6 +935,8 @@ out_err_unlock:
 		smc_lgr_forget(new_smc->conn.lgr);
 	mutex_unlock(&smc_create_lgr_pending);
 out_err:
+	if (newsmcsk->sk_state == SMC_INIT)
+		sock_put(&new_smc->sk); /* passive closing */
 	newsmcsk->sk_state = SMC_CLOSED;
 	smc_conn_free(&new_smc->conn);
 	goto enqueue; /* queue new sock with sk_err set */
@@ -963,12 +963,15 @@ static void smc_tcp_listen_work(struct work_struct *work)
 		sock_hold(lsk); /* sock_put in smc_listen_work */
 		INIT_WORK(&new_smc->smc_listen_work, smc_listen_work);
 		smc_copy_sock_settings_to_smc(new_smc);
-		schedule_work(&new_smc->smc_listen_work);
+		sock_hold(&new_smc->sk); /* sock_put in passive closing */
+		if (!schedule_work(&new_smc->smc_listen_work))
+			sock_put(&new_smc->sk);
 	}
 
 out:
 	release_sock(lsk);
 	lsk->sk_data_ready(lsk); /* no more listening, wake accept */
+	sock_put(&lsmc->sk); /* sock_hold in smc_listen */
 }
 
 static int smc_listen(struct socket *sock, int backlog)
@@ -1002,7 +1005,9 @@ static int smc_listen(struct socket *sock, int backlog)
 	sk->sk_ack_backlog = 0;
 	sk->sk_state = SMC_LISTEN;
 	INIT_WORK(&smc->tcp_listen_work, smc_tcp_listen_work);
-	schedule_work(&smc->tcp_listen_work);
+	sock_hold(sk); /* sock_hold in tcp_listen_worker */
+	if (!schedule_work(&smc->tcp_listen_work))
+		sock_put(sk);
 
 out:
 	release_sock(sk);
@@ -1019,6 +1024,7 @@ static int smc_accept(struct socket *sock, struct socket *new_sock,
 	int rc = 0;
 
 	lsmc = smc_sk(sk);
+	sock_hold(sk); /* sock_put below */
 	lock_sock(sk);
 
 	if (lsmc->sk.sk_state != SMC_LISTEN) {
@@ -1053,6 +1059,7 @@ static int smc_accept(struct socket *sock, struct socket *new_sock,
 
 out:
 	release_sock(sk);
+	sock_put(sk); /* sock_hold above */
 	return rc;
 }
 
diff --git a/net/smc/smc.h b/net/smc/smc.h
index 0bee9d16cf29..bfbe20234105 100644
--- a/net/smc/smc.h
+++ b/net/smc/smc.h
@@ -178,7 +178,6 @@ struct smc_sock {				/* smc sock container */
 	struct work_struct	smc_listen_work;/* prepare new accept socket */
 	struct list_head	accept_q;	/* sockets to be accepted */
 	spinlock_t		accept_q_lock;	/* protects accept_q */
-	struct delayed_work	sock_put_work;	/* final socket freeing */
 	bool			use_fallback;	/* fallback to tcp */
 	u8			wait_close_tx_prepared : 1;
 						/* shutdown wr or close
diff --git a/net/smc/smc_cdc.c b/net/smc/smc_cdc.c
index 6e8f5fbe0f09..3cd086e5bd28 100644
--- a/net/smc/smc_cdc.c
+++ b/net/smc/smc_cdc.c
@@ -212,6 +212,14 @@ static void smc_cdc_msg_recv_action(struct smc_sock *smc,
 		smc->sk.sk_data_ready(&smc->sk);
 	}
 
+	/* piggy backed tx info */
+	/* trigger sndbuf consumer: RDMA write into peer RMBE and CDC */
+	if (diff_cons && smc_tx_prepared_sends(conn)) {
+		smc_tx_sndbuf_nonempty(conn);
+		/* trigger socket release if connection closed */
+		smc_close_wake_tx_prepared(smc);
+	}
+
 	if (conn->local_rx_ctrl.conn_state_flags.peer_conn_abort) {
 		smc->sk.sk_err = ECONNRESET;
 		conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
@@ -221,15 +229,9 @@ static void smc_cdc_msg_recv_action(struct smc_sock *smc,
 		if (smc->clcsock && smc->clcsock->sk)
 			smc->clcsock->sk->sk_shutdown |= RCV_SHUTDOWN;
 		sock_set_flag(&smc->sk, SOCK_DONE);
-		schedule_work(&conn->close_work);
-	}
-
-	/* piggy backed tx info */
-	/* trigger sndbuf consumer: RDMA write into peer RMBE and CDC */
-	if (diff_cons && smc_tx_prepared_sends(conn)) {
-		smc_tx_sndbuf_nonempty(conn);
-		/* trigger socket release if connection closed */
-		smc_close_wake_tx_prepared(smc);
+		sock_hold(&smc->sk); /* sock_put in close_work */
+		if (!schedule_work(&conn->close_work))
+			sock_put(&smc->sk);
 	}
 }
 
diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c
index babe05d385e7..4339852a8910 100644
--- a/net/smc/smc_close.c
+++ b/net/smc/smc_close.c
@@ -110,6 +110,7 @@ static void smc_close_active_abort(struct smc_sock *smc)
 		release_sock(sk);
 		cancel_delayed_work_sync(&smc->conn.tx_work);
 		lock_sock(sk);
+		sock_put(sk); /* passive closing */
 		break;
 	case SMC_APPCLOSEWAIT1:
 	case SMC_APPCLOSEWAIT2:
@@ -125,11 +126,13 @@ static void smc_close_active_abort(struct smc_sock *smc)
 	case SMC_PEERCLOSEWAIT1:
 	case SMC_PEERCLOSEWAIT2:
 		if (!txflags->peer_conn_closed) {
+			/* just SHUTDOWN_SEND done */
 			sk->sk_state = SMC_PEERABORTWAIT;
 			sock_release(smc->clcsock);
 		} else {
 			sk->sk_state = SMC_CLOSED;
 		}
+		sock_put(sk); /* passive closing */
 		break;
 	case SMC_PROCESSABORT:
 	case SMC_APPFINCLOSEWAIT:
@@ -138,6 +141,8 @@ static void smc_close_active_abort(struct smc_sock *smc)
 		sk->sk_state = SMC_CLOSED;
 		break;
 	case SMC_PEERFINCLOSEWAIT:
+		sock_put(sk); /* passive closing */
+		break;
 	case SMC_PEERABORTWAIT:
 	case SMC_CLOSED:
 		break;
@@ -229,12 +234,14 @@ again:
 		rc = smc_close_final(conn);
 		if (rc)
 			break;
-		if (smc_cdc_rxed_any_close(conn))
+		if (smc_cdc_rxed_any_close(conn)) {
 			/* peer has closed the socket already */
 			sk->sk_state = SMC_CLOSED;
-		else
+			sock_put(sk); /* postponed passive closing */
+		} else {
 			/* peer has just issued a shutdown write */
 			sk->sk_state = SMC_PEERFINCLOSEWAIT;
+		}
 		break;
 	case SMC_PEERCLOSEWAIT1:
 	case SMC_PEERCLOSEWAIT2:
@@ -272,27 +279,33 @@ static void smc_close_passive_abort_received(struct smc_sock *smc)
 	struct sock *sk = &smc->sk;
 
 	switch (sk->sk_state) {
+	case SMC_INIT:
 	case SMC_ACTIVE:
-	case SMC_APPFINCLOSEWAIT:
 	case SMC_APPCLOSEWAIT1:
-	case SMC_APPCLOSEWAIT2:
+		sk->sk_state = SMC_PROCESSABORT;
+		sock_put(sk); /* passive closing */
+		break;
+	case SMC_APPFINCLOSEWAIT:
 		sk->sk_state = SMC_PROCESSABORT;
 		break;
 	case SMC_PEERCLOSEWAIT1:
 	case SMC_PEERCLOSEWAIT2:
 		if (txflags->peer_done_writing &&
-		    !smc_close_sent_any_close(&smc->conn)) {
+		    !smc_close_sent_any_close(&smc->conn))
 			/* just shutdown, but not yet closed locally */
 			sk->sk_state = SMC_PROCESSABORT;
-		} else {
+		else
 			sk->sk_state = SMC_CLOSED;
-		}
+		sock_put(sk); /* passive closing */
 		break;
+	case SMC_APPCLOSEWAIT2:
 	case SMC_PEERFINCLOSEWAIT:
+		sk->sk_state = SMC_CLOSED;
+		sock_put(sk); /* passive closing */
+		break;
 	case SMC_PEERABORTWAIT:
 		sk->sk_state = SMC_CLOSED;
 		break;
-	case SMC_INIT:
 	case SMC_PROCESSABORT:
 	/* nothing to do, add tracing in future patch */
 		break;
@@ -336,13 +349,18 @@ static void smc_close_passive_work(struct work_struct *work)
 	case SMC_INIT:
 		if (atomic_read(&conn->bytes_to_rcv) ||
 		    (rxflags->peer_done_writing &&
-		     !smc_cdc_rxed_any_close(conn)))
+		     !smc_cdc_rxed_any_close(conn))) {
 			sk->sk_state = SMC_APPCLOSEWAIT1;
-		else
+		} else {
 			sk->sk_state = SMC_CLOSED;
+			sock_put(sk); /* passive closing */
+		}
 		break;
 	case SMC_ACTIVE:
 		sk->sk_state = SMC_APPCLOSEWAIT1;
+		/* postpone sock_put() for passive closing to cover
+		 * received SEND_SHUTDOWN as well
+		 */
 		break;
 	case SMC_PEERCLOSEWAIT1:
 		if (rxflags->peer_done_writing)
@@ -360,13 +378,20 @@ static void smc_close_passive_work(struct work_struct *work)
 			/* just shutdown, but not yet closed locally */
 			sk->sk_state = SMC_APPFINCLOSEWAIT;
 		}
+		sock_put(sk); /* passive closing */
 		break;
 	case SMC_PEERFINCLOSEWAIT:
-		if (smc_cdc_rxed_any_close(conn))
+		if (smc_cdc_rxed_any_close(conn)) {
 			sk->sk_state = SMC_CLOSED;
+			sock_put(sk); /* passive closing */
+		}
 		break;
 	case SMC_APPCLOSEWAIT1:
 	case SMC_APPCLOSEWAIT2:
+		/* postpone sock_put() for passive closing to cover
+		 * received SEND_SHUTDOWN as well
+		 */
+		break;
 	case SMC_APPFINCLOSEWAIT:
 	case SMC_PEERABORTWAIT:
 	case SMC_PROCESSABORT:
@@ -382,23 +407,11 @@ wakeup:
 	if (old_state != sk->sk_state) {
 		sk->sk_state_change(sk);
 		if ((sk->sk_state == SMC_CLOSED) &&
-		    (sock_flag(sk, SOCK_DEAD) || !sk->sk_socket)) {
+		    (sock_flag(sk, SOCK_DEAD) || !sk->sk_socket))
 			smc_conn_free(conn);
-			schedule_delayed_work(&smc->sock_put_work,
-					      SMC_CLOSE_SOCK_PUT_DELAY);
-		}
 	}
 	release_sock(sk);
-}
-
-void smc_close_sock_put_work(struct work_struct *work)
-{
-	struct smc_sock *smc = container_of(to_delayed_work(work),
-					    struct smc_sock,
-					    sock_put_work);
-
-	smc->sk.sk_prot->unhash(&smc->sk);
-	sock_put(&smc->sk);
+	sock_put(sk); /* sock_hold done by schedulers of close_work */
 }
 
 int smc_close_shutdown_write(struct smc_sock *smc)
diff --git a/net/smc/smc_close.h b/net/smc/smc_close.h
index 8c498885d758..19eb6a211c23 100644
--- a/net/smc/smc_close.h
+++ b/net/smc/smc_close.h
@@ -21,7 +21,6 @@
 
 void smc_close_wake_tx_prepared(struct smc_sock *smc);
 int smc_close_active(struct smc_sock *smc);
-void smc_close_sock_put_work(struct work_struct *work);
 int smc_close_shutdown_write(struct smc_sock *smc);
 void smc_close_init(struct smc_sock *smc);
 
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index ed5b46d1fe41..2424c7100aaf 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -328,13 +328,13 @@ void smc_lgr_terminate(struct smc_link_group *lgr)
 	while (node) {
 		conn = rb_entry(node, struct smc_connection, alert_node);
 		smc = container_of(conn, struct smc_sock, conn);
-		sock_hold(&smc->sk);
+		sock_hold(&smc->sk); /* sock_put in close work */
 		conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
 		__smc_lgr_unregister_conn(conn);
 		write_unlock_bh(&lgr->conns_lock);
-		schedule_work(&conn->close_work);
+		if (!schedule_work(&conn->close_work))
+			sock_put(&smc->sk);
 		write_lock_bh(&lgr->conns_lock);
-		sock_put(&smc->sk);
 		node = rb_first(&lgr->conns_all);
 	}
 	write_unlock_bh(&lgr->conns_lock);