[linux-block.git] / net / rds / tcp_connect.c

/*
 * Copyright (c) 2006, 2017 Oracle and/or its affiliates. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */
#include <linux/kernel.h>
#include <linux/in.h>
#include <net/tcp.h>

#include "rds.h"
#include "tcp.h"

void rds_tcp_state_change(struct sock *sk)
{
	void (*state_change)(struct sock *sk);
	struct rds_conn_path *cp;
	struct rds_tcp_connection *tc;

	read_lock_bh(&sk->sk_callback_lock);
	cp = sk->sk_user_data;
	if (!cp) {
		state_change = sk->sk_state_change;
		goto out;
	}
	tc = cp->cp_transport_data;
	state_change = tc->t_orig_state_change;

	rdsdebug("sock %p state_change to %d\n", tc->t_sock, sk->sk_state);

	switch (sk->sk_state) {
	/* ignore connecting sockets as they make progress */
	case TCP_SYN_SENT:
	case TCP_SYN_RECV:
		break;
	case TCP_ESTABLISHED:
		/* Force the peer to reconnect so that we have the
		 * TCP ports going from <smaller-ip>.<transient> to
		 * <larger-ip>.<RDS_TCP_PORT>. We avoid marking the
		 * RDS connection as RDS_CONN_UP until the reconnect,
		 * to avoid RDS datagram loss.
		 */
		if (rds_addr_cmp(&cp->cp_conn->c_laddr,
				 &cp->cp_conn->c_faddr) >= 0 &&
		    rds_conn_path_transition(cp, RDS_CONN_CONNECTING,
					     RDS_CONN_ERROR)) {
			rds_conn_path_drop(cp, false);
		} else {
			rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
		}
		break;
	case TCP_CLOSE_WAIT:
	case TCP_CLOSE:
		rds_conn_path_drop(cp, false);
		break;
	default:
		break;
	}
out:
	read_unlock_bh(&sk->sk_callback_lock);
	state_change(sk);
}

int rds_tcp_conn_path_connect(struct rds_conn_path *cp)
{
	struct socket *sock = NULL;
	struct sockaddr_in6 sin6;
	struct sockaddr_in sin;
	struct sockaddr *addr;
	int addrlen;
	bool isv6;
	int ret;
	struct rds_connection *conn = cp->cp_conn;
	struct rds_tcp_connection *tc = cp->cp_transport_data;

	/* for multipath rds,we only trigger the connection after
	 * the handshake probe has determined the number of paths.
	 */
	if (cp->cp_index > 0 && cp->cp_conn->c_npaths < 2)
		return -EAGAIN;

	mutex_lock(&tc->t_conn_path_lock);

	if (rds_conn_path_up(cp)) {
		mutex_unlock(&tc->t_conn_path_lock);
		return 0;
	}
	if (ipv6_addr_v4mapped(&conn->c_laddr)) {
		ret = sock_create_kern(rds_conn_net(conn), PF_INET,
				       SOCK_STREAM, IPPROTO_TCP, &sock);
		isv6 = false;
	} else {
		ret = sock_create_kern(rds_conn_net(conn), PF_INET6,
				       SOCK_STREAM, IPPROTO_TCP, &sock);
		isv6 = true;
	}

	if (ret < 0)
		goto out;

	if (!rds_tcp_tune(sock)) {
		ret = -EINVAL;
		goto out;
	}

	if (isv6) {
		sin6.sin6_family = AF_INET6;
		sin6.sin6_addr = conn->c_laddr;
		sin6.sin6_port = 0;
		sin6.sin6_flowinfo = 0;
		sin6.sin6_scope_id = conn->c_dev_if;
		addr = (struct sockaddr *)&sin6;
		addrlen = sizeof(sin6);
	} else {
		sin.sin_family = AF_INET;
		sin.sin_addr.s_addr = conn->c_laddr.s6_addr32[3];
		sin.sin_port = 0;
		addr = (struct sockaddr *)&sin;
		addrlen = sizeof(sin);
	}

	ret = kernel_bind(sock, addr, addrlen);
	if (ret) {
		rdsdebug("bind failed with %d at address %pI6c\n",
			 ret, &conn->c_laddr);
		goto out;
	}

	if (isv6) {
		sin6.sin6_family = AF_INET6;
		sin6.sin6_addr = conn->c_faddr;
		sin6.sin6_port = htons(RDS_TCP_PORT);
		sin6.sin6_flowinfo = 0;
		sin6.sin6_scope_id = conn->c_dev_if;
		addr = (struct sockaddr *)&sin6;
		addrlen = sizeof(sin6);
	} else {
		sin.sin_family = AF_INET;
		sin.sin_addr.s_addr = conn->c_faddr.s6_addr32[3];
		sin.sin_port = htons(RDS_TCP_PORT);
		addr = (struct sockaddr *)&sin;
		addrlen = sizeof(sin);
	}

	/*
	 * once we call connect() we can start getting callbacks and they
	 * own the socket
	 */
	rds_tcp_set_callbacks(sock, cp);
	ret = kernel_connect(sock, addr, addrlen, O_NONBLOCK);

	rdsdebug("connect to address %pI6c returned %d\n", &conn->c_faddr, ret);
	if (ret == -EINPROGRESS)
		ret = 0;
	if (ret == 0) {
		rds_tcp_keepalive(sock);
		sock = NULL;
	} else {
		rds_tcp_restore_callbacks(sock, cp->cp_transport_data);
	}

out:
	mutex_unlock(&tc->t_conn_path_lock);
	if (sock)
		sock_release(sock);
	return ret;
}

/*
 * Before killing the tcp socket this needs to serialize with callbacks.  The
 * caller has already grabbed the sending sem so we're serialized with other
 * senders.
 *
 * TCP calls the callbacks with the sock lock so we hold it while we reset the
 * callbacks to those set by TCP.  Our callbacks won't execute again once we
 * hold the sock lock.
 */
void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
{
	struct rds_tcp_connection *tc = cp->cp_transport_data;
	struct socket *sock = tc->t_sock;

	rdsdebug("shutting down conn %p tc %p sock %p\n",
		 cp->cp_conn, tc, sock);

	if (sock) {
		if (rds_destroy_pending(cp->cp_conn))
			sock_no_linger(sock->sk);
		sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
		lock_sock(sock->sk);
		rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */

		release_sock(sock->sk);
		sock_release(sock);
	}

	if (tc->t_tinc) {
		rds_inc_put(&tc->t_tinc->ti_inc);
		tc->t_tinc = NULL;
	}
	tc->t_tinc_hdr_rem = sizeof(struct rds_header);
	tc->t_tinc_data_rem = 0;
}
Commit	Line	Data
70041088	1	/*
eee2fa6a	2	* Copyright (c) 2006, 2017 Oracle and/or its affiliates. All rights reserved.
70041088 AG	3	*
	4	* This software is available to you under a choice of one of two
	5	* licenses. You may choose to be licensed under the terms of the GNU
	6	* General Public License (GPL) Version 2, available from the file
	7	* COPYING in the main directory of this source tree, or the
	8	* OpenIB.org BSD license below:
	9	*
	10	* Redistribution and use in source and binary forms, with or
	11	* without modification, are permitted provided that the following
	12	* conditions are met:
	13	*
	14	* - Redistributions of source code must retain the above
	15	* copyright notice, this list of conditions and the following
	16	* disclaimer.
	17	*
	18	* - Redistributions in binary form must reproduce the above
	19	* copyright notice, this list of conditions and the following
	20	* disclaimer in the documentation and/or other materials
	21	* provided with the distribution.
	22	*
	23	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
	24	* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
	25	* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
	26	* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
	27	* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
	28	* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
	29	* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
	30	* SOFTWARE.
	31	*
	32	*/
	33	#include <linux/kernel.h>
	34	#include <linux/in.h>
	35	#include <net/tcp.h>
	36
	37	#include "rds.h"
	38	#include "tcp.h"
	39
	40	void rds_tcp_state_change(struct sock *sk)
	41	{
	42	void (state_change)(struct sock sk);
ea3b1ea5	43	struct rds_conn_path *cp;
70041088 AG	44	struct rds_tcp_connection *tc;
70041088 AG	45
38036629	46	read_lock_bh(&sk->sk_callback_lock);
ea3b1ea5 SV	47	cp = sk->sk_user_data;
ea3b1ea5 SV	48	if (!cp) {
70041088 AG	49	state_change = sk->sk_state_change;
	50	goto out;
	51	}
ea3b1ea5	52	tc = cp->cp_transport_data;
70041088 AG	53	state_change = tc->t_orig_state_change;
	54
	55	rdsdebug("sock %p state_change to %d\n", tc->t_sock, sk->sk_state);
	56
5c3da57d JH	57	switch (sk->sk_state) {
	58	/* ignore connecting sockets as they make progress */
	59	case TCP_SYN_SENT:
	60	case TCP_SYN_RECV:
	61	break;
	62	case TCP_ESTABLISHED:
1a0e100f SV	63	/* Force the peer to reconnect so that we have the
	64	* TCP ports going from <smaller-ip>.<transient> to
	65	* <larger-ip>.<RDS_TCP_PORT>. We avoid marking the
	66	* RDS connection as RDS_CONN_UP until the reconnect,
	67	* to avoid RDS datagram loss.
	68	*/
eee2fa6a KCP	69	if (rds_addr_cmp(&cp->cp_conn->c_laddr,
eee2fa6a KCP	70	&cp->cp_conn->c_faddr) >= 0 &&
1a0e100f SV	71	rds_conn_path_transition(cp, RDS_CONN_CONNECTING,
1a0e100f SV	72	RDS_CONN_ERROR)) {
aed20a53	73	rds_conn_path_drop(cp, false);
1a0e100f SV	74	} else {
	75	rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
	76	}
5c3da57d JH	77	break;
	78	case TCP_CLOSE_WAIT:
	79	case TCP_CLOSE:
aed20a53	80	rds_conn_path_drop(cp, false);
3754fa74	81	break;
5c3da57d JH	82	default:
5c3da57d JH	83	break;
70041088 AG	84	}
70041088 AG	85	out:
38036629	86	read_unlock_bh(&sk->sk_callback_lock);
70041088 AG	87	state_change(sk);
	88	}
	89
b04e8554	90	int rds_tcp_conn_path_connect(struct rds_conn_path *cp)
70041088 AG	91	{
70041088 AG	92	struct socket *sock = NULL;
1e2b44e7	93	struct sockaddr_in6 sin6;
eee2fa6a KCP	94	struct sockaddr_in sin;
	95	struct sockaddr *addr;
	96	int addrlen;
1e2b44e7	97	bool isv6;
70041088	98	int ret;
b04e8554 SV	99	struct rds_connection *conn = cp->cp_conn;
b04e8554 SV	100	struct rds_tcp_connection *tc = cp->cp_transport_data;
bd7c5f98	101
5916e2c1 SV	102	/* for multipath rds,we only trigger the connection after
	103	* the handshake probe has determined the number of paths.
	104	*/
	105	if (cp->cp_index > 0 && cp->cp_conn->c_npaths < 2)
	106	return -EAGAIN;
	107
02105b2c	108	mutex_lock(&tc->t_conn_path_lock);
70041088	109
b04e8554	110	if (rds_conn_path_up(cp)) {
02105b2c	111	mutex_unlock(&tc->t_conn_path_lock);
bd7c5f98 SV	112	return 0;
bd7c5f98 SV	113	}
1e2b44e7 KCP	114	if (ipv6_addr_v4mapped(&conn->c_laddr)) {
	115	ret = sock_create_kern(rds_conn_net(conn), PF_INET,
	116	SOCK_STREAM, IPPROTO_TCP, &sock);
	117	isv6 = false;
	118	} else {
	119	ret = sock_create_kern(rds_conn_net(conn), PF_INET6,
	120	SOCK_STREAM, IPPROTO_TCP, &sock);
	121	isv6 = true;
	122	}
	123
70041088 AG	124	if (ret < 0)
	125	goto out;
	126
6997fbd7 TH	127	if (!rds_tcp_tune(sock)) {
	128	ret = -EINVAL;
	129	goto out;
	130	}
70041088	131
1e2b44e7 KCP	132	if (isv6) {
	133	sin6.sin6_family = AF_INET6;
	134	sin6.sin6_addr = conn->c_laddr;
	135	sin6.sin6_port = 0;
	136	sin6.sin6_flowinfo = 0;
	137	sin6.sin6_scope_id = conn->c_dev_if;
	138	addr = (struct sockaddr *)&sin6;
	139	addrlen = sizeof(sin6);
	140	} else {
	141	sin.sin_family = AF_INET;
	142	sin.sin_addr.s_addr = conn->c_laddr.s6_addr32[3];
	143	sin.sin_port = 0;
	144	addr = (struct sockaddr *)&sin;
	145	addrlen = sizeof(sin);
	146	}
70041088	147
c889a99a	148	ret = kernel_bind(sock, addr, addrlen);
70041088	149	if (ret) {
eee2fa6a	150	rdsdebug("bind failed with %d at address %pI6c\n",
6884b348	151	ret, &conn->c_laddr);
70041088 AG	152	goto out;
	153	}
	154
1e2b44e7 KCP	155	if (isv6) {
	156	sin6.sin6_family = AF_INET6;
	157	sin6.sin6_addr = conn->c_faddr;
	158	sin6.sin6_port = htons(RDS_TCP_PORT);
	159	sin6.sin6_flowinfo = 0;
	160	sin6.sin6_scope_id = conn->c_dev_if;
	161	addr = (struct sockaddr *)&sin6;
	162	addrlen = sizeof(sin6);
	163	} else {
	164	sin.sin_family = AF_INET;
	165	sin.sin_addr.s_addr = conn->c_faddr.s6_addr32[3];
	166	sin.sin_port = htons(RDS_TCP_PORT);
	167	addr = (struct sockaddr *)&sin;
	168	addrlen = sizeof(sin);
	169	}
70041088 AG	170
	171	/*
	172	* once we call connect() we can start getting callbacks and they
	173	* own the socket
	174	*/
ea3b1ea5	175	rds_tcp_set_callbacks(sock, cp);
26297b4c	176	ret = kernel_connect(sock, addr, addrlen, O_NONBLOCK);
70041088	177
eee2fa6a	178	rdsdebug("connect to address %pI6c returned %d\n", &conn->c_faddr, ret);
70041088 AG	179	if (ret == -EINPROGRESS)
70041088 AG	180	ret = 0;
467fa153 SV	181	if (ret == 0) {
467fa153 SV	182	rds_tcp_keepalive(sock);
eb74cc97	183	sock = NULL;
467fa153	184	} else {
b04e8554	185	rds_tcp_restore_callbacks(sock, cp->cp_transport_data);
467fa153	186	}
70041088 AG	187
70041088 AG	188	out:
02105b2c	189	mutex_unlock(&tc->t_conn_path_lock);
70041088 AG	190	if (sock)
	191	sock_release(sock);
	192	return ret;
	193	}
	194
	195	/*
	196	* Before killing the tcp socket this needs to serialize with callbacks. The
	197	* caller has already grabbed the sending sem so we're serialized with other
	198	* senders.
	199	*
	200	* TCP calls the callbacks with the sock lock so we hold it while we reset the
	201	* callbacks to those set by TCP. Our callbacks won't execute again once we
	202	* hold the sock lock.
	203	*/
226f7a7d	204	void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
70041088	205	{
226f7a7d	206	struct rds_tcp_connection *tc = cp->cp_transport_data;
70041088 AG	207	struct socket *sock = tc->t_sock;
70041088 AG	208
226f7a7d SV	209	rdsdebug("shutting down conn %p tc %p sock %p\n",
226f7a7d SV	210	cp->cp_conn, tc, sock);
70041088 AG	211
70041088 AG	212	if (sock) {
ebeeb1ad	213	if (rds_destroy_pending(cp->cp_conn))
c433594c	214	sock_no_linger(sock->sk);
70041088 AG	215	sock->ops->shutdown(sock, RCV_SHUTDOWN \| SEND_SHUTDOWN);
	216	lock_sock(sock->sk);
	217	rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
	218
	219	release_sock(sock->sk);
	220	sock_release(sock);
ccbd6a5a	221	}
70041088 AG	222
	223	if (tc->t_tinc) {
	224	rds_inc_put(&tc->t_tinc->ti_inc);
	225	tc->t_tinc = NULL;
	226	}
	227	tc->t_tinc_hdr_rem = sizeof(struct rds_header);
	228	tc->t_tinc_data_rem = 0;
	229	}