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

/*
 * Copyright (c) 2006 Oracle.  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 (!IS_CANONICAL(cp->cp_conn->c_laddr, cp->cp_conn->c_faddr) &&
		    rds_conn_path_transition(cp, RDS_CONN_CONNECTING,
					     RDS_CONN_ERROR)) {
			rds_conn_path_drop(cp);
		} else {
			rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
		}
		break;
	case TCP_CLOSE_WAIT:
	case TCP_CLOSE:
		rds_conn_path_drop(cp);
	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_in src, dest;
	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;
	}
	ret = sock_create_kern(rds_conn_net(conn), PF_INET,
			       SOCK_STREAM, IPPROTO_TCP, &sock);
	if (ret < 0)
		goto out;

	rds_tcp_tune(sock);

	src.sin_family = AF_INET;
	src.sin_addr.s_addr = (__force u32)conn->c_laddr;
	src.sin_port = (__force u16)htons(0);

	ret = sock->ops->bind(sock, (struct sockaddr *)&src, sizeof(src));
	if (ret) {
		rdsdebug("bind failed with %d at address %pI4\n",
			 ret, &conn->c_laddr);
		goto out;
	}

	dest.sin_family = AF_INET;
	dest.sin_addr.s_addr = (__force u32)conn->c_faddr;
	dest.sin_port = (__force u16)htons(RDS_TCP_PORT);

	/*
	 * once we call connect() we can start getting callbacks and they
	 * own the socket
	 */
	rds_tcp_set_callbacks(sock, cp);
	ret = sock->ops->connect(sock, (struct sockaddr *)&dest, sizeof(dest),
				 O_NONBLOCK);

	rdsdebug("connect to address %pI4 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 (cp->cp_conn->c_destroy_in_prog)
			rds_tcp_set_linger(sock);
		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 AG	1	/*
	2	* Copyright (c) 2006 Oracle. All rights reserved.
	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	*/
00354de5	69	if (!IS_CANONICAL(cp->cp_conn->c_laddr, cp->cp_conn->c_faddr) &&
1a0e100f SV	70	rds_conn_path_transition(cp, RDS_CONN_CONNECTING,
	71	RDS_CONN_ERROR)) {
	72	rds_conn_path_drop(cp);
	73	} else {
	74	rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
	75	}
5c3da57d JH	76	break;
	77	case TCP_CLOSE_WAIT:
	78	case TCP_CLOSE:
ea3b1ea5	79	rds_conn_path_drop(cp);
5c3da57d JH	80	default:
5c3da57d JH	81	break;
70041088 AG	82	}
70041088 AG	83	out:
38036629	84	read_unlock_bh(&sk->sk_callback_lock);
70041088 AG	85	state_change(sk);
	86	}
	87
b04e8554	88	int rds_tcp_conn_path_connect(struct rds_conn_path *cp)
70041088 AG	89	{
	90	struct socket *sock = NULL;
	91	struct sockaddr_in src, dest;
	92	int ret;
b04e8554 SV	93	struct rds_connection *conn = cp->cp_conn;
b04e8554 SV	94	struct rds_tcp_connection *tc = cp->cp_transport_data;
bd7c5f98	95
5916e2c1 SV	96	/* for multipath rds,we only trigger the connection after
	97	* the handshake probe has determined the number of paths.
	98	*/
	99	if (cp->cp_index > 0 && cp->cp_conn->c_npaths < 2)
	100	return -EAGAIN;
	101
02105b2c	102	mutex_lock(&tc->t_conn_path_lock);
70041088	103
b04e8554	104	if (rds_conn_path_up(cp)) {
02105b2c	105	mutex_unlock(&tc->t_conn_path_lock);
bd7c5f98 SV	106	return 0;
bd7c5f98 SV	107	}
d5a8ac28 SV	108	ret = sock_create_kern(rds_conn_net(conn), PF_INET,
d5a8ac28 SV	109	SOCK_STREAM, IPPROTO_TCP, &sock);
70041088 AG	110	if (ret < 0)
	111	goto out;
	112
	113	rds_tcp_tune(sock);
	114
	115	src.sin_family = AF_INET;
	116	src.sin_addr.s_addr = (__force u32)conn->c_laddr;
	117	src.sin_port = (__force u16)htons(0);
	118
	119	ret = sock->ops->bind(sock, (struct sockaddr *)&src, sizeof(src));
	120	if (ret) {
6884b348 JP	121	rdsdebug("bind failed with %d at address %pI4\n",
6884b348 JP	122	ret, &conn->c_laddr);
70041088 AG	123	goto out;
	124	}
	125
	126	dest.sin_family = AF_INET;
	127	dest.sin_addr.s_addr = (__force u32)conn->c_faddr;
	128	dest.sin_port = (__force u16)htons(RDS_TCP_PORT);
	129
	130	/*
	131	* once we call connect() we can start getting callbacks and they
	132	* own the socket
	133	*/
ea3b1ea5	134	rds_tcp_set_callbacks(sock, cp);
70041088 AG	135	ret = sock->ops->connect(sock, (struct sockaddr *)&dest, sizeof(dest),
70041088 AG	136	O_NONBLOCK);
70041088	137
6884b348	138	rdsdebug("connect to address %pI4 returned %d\n", &conn->c_faddr, ret);
70041088 AG	139	if (ret == -EINPROGRESS)
70041088 AG	140	ret = 0;
467fa153 SV	141	if (ret == 0) {
467fa153 SV	142	rds_tcp_keepalive(sock);
eb74cc97	143	sock = NULL;
467fa153	144	} else {
b04e8554	145	rds_tcp_restore_callbacks(sock, cp->cp_transport_data);
467fa153	146	}
70041088 AG	147
70041088 AG	148	out:
02105b2c	149	mutex_unlock(&tc->t_conn_path_lock);
70041088 AG	150	if (sock)
	151	sock_release(sock);
	152	return ret;
	153	}
	154
	155	/*
	156	* Before killing the tcp socket this needs to serialize with callbacks. The
	157	* caller has already grabbed the sending sem so we're serialized with other
	158	* senders.
	159	*
	160	* TCP calls the callbacks with the sock lock so we hold it while we reset the
	161	* callbacks to those set by TCP. Our callbacks won't execute again once we
	162	* hold the sock lock.
	163	*/
226f7a7d	164	void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
70041088	165	{
226f7a7d	166	struct rds_tcp_connection *tc = cp->cp_transport_data;
70041088 AG	167	struct socket *sock = tc->t_sock;
70041088 AG	168
226f7a7d SV	169	rdsdebug("shutting down conn %p tc %p sock %p\n",
226f7a7d SV	170	cp->cp_conn, tc, sock);
70041088 AG	171
70041088 AG	172	if (sock) {
c14b0366 SV	173	if (cp->cp_conn->c_destroy_in_prog)
c14b0366 SV	174	rds_tcp_set_linger(sock);
70041088 AG	175	sock->ops->shutdown(sock, RCV_SHUTDOWN \| SEND_SHUTDOWN);
	176	lock_sock(sock->sk);
	177	rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
	178
	179	release_sock(sock->sk);
	180	sock_release(sock);
ccbd6a5a	181	}
70041088 AG	182
	183	if (tc->t_tinc) {
	184	rds_inc_put(&tc->t_tinc->ti_inc);
	185	tc->t_tinc = NULL;
	186	}
	187	tc->t_tinc_hdr_rem = sizeof(struct rds_header);
	188	tc->t_tinc_data_rem = 0;
	189	}