[deliverable/linux.git] / net / rds / tcp_listen.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/gfp.h>
#include <linux/in.h>
#include <net/tcp.h>

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

int rds_tcp_keepalive(struct socket *sock)
{
	/* values below based on xs_udp_default_timeout */
	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
	int keepalive = 1;
	int ret = 0;

	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
				(char *)&keepalive, sizeof(keepalive));
	if (ret < 0)
		goto bail;

	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
				(char *)&keepcnt, sizeof(keepcnt));
	if (ret < 0)
		goto bail;

	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
				(char *)&keepidle, sizeof(keepidle));
	if (ret < 0)
		goto bail;

	/* KEEPINTVL is the interval between successive probes. We follow
	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
	 */
	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
				(char *)&keepidle, sizeof(keepidle));
bail:
	return ret;
}

int rds_tcp_accept_one(struct socket *sock)
{
	struct socket *new_sock = NULL;
	struct rds_connection *conn;
	int ret;
	struct inet_sock *inet;
	struct rds_tcp_connection *rs_tcp = NULL;
	int conn_state;
	struct sock *nsk;

	ret = sock_create_kern(sock_net(sock->sk), sock->sk->sk_family,
			       sock->sk->sk_type, sock->sk->sk_protocol,
			       &new_sock);
	if (ret)
		goto out;

	new_sock->type = sock->type;
	new_sock->ops = sock->ops;
	ret = sock->ops->accept(sock, new_sock, O_NONBLOCK);
	if (ret < 0)
		goto out;

	ret = rds_tcp_keepalive(new_sock);
	if (ret < 0)
		goto out;

	rds_tcp_tune(new_sock);

	inet = inet_sk(new_sock->sk);

	rdsdebug("accepted tcp %pI4:%u -> %pI4:%u\n",
		 &inet->inet_saddr, ntohs(inet->inet_sport),
		 &inet->inet_daddr, ntohs(inet->inet_dport));

	conn = rds_conn_create(sock_net(sock->sk),
			       inet->inet_saddr, inet->inet_daddr,
			       &rds_tcp_transport, GFP_KERNEL);
	if (IS_ERR(conn)) {
		ret = PTR_ERR(conn);
		goto out;
	}
	/* An incoming SYN request came in, and TCP just accepted it.
	 *
	 * If the client reboots, this conn will need to be cleaned up.
	 * rds_tcp_state_change() will do that cleanup
	 */
	rs_tcp = (struct rds_tcp_connection *)conn->c_transport_data;
	rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING);
	mutex_lock(&rs_tcp->t_conn_lock);
	conn_state = rds_conn_state(conn);
	if (conn_state != RDS_CONN_CONNECTING && conn_state != RDS_CONN_UP)
		goto rst_nsk;
	if (rs_tcp->t_sock) {
		/* Need to resolve a duelling SYN between peers.
		 * We have an outstanding SYN to this peer, which may
		 * potentially have transitioned to the RDS_CONN_UP state,
		 * so we must quiesce any send threads before resetting
		 * c_transport_data.
		 */
		wait_event(conn->c_waitq,
			   !test_bit(RDS_IN_XMIT, &conn->c_flags));
		if (ntohl(inet->inet_saddr) < ntohl(inet->inet_daddr)) {
			goto rst_nsk;
		} else if (rs_tcp->t_sock) {
			rds_tcp_restore_callbacks(rs_tcp->t_sock, rs_tcp);
			conn->c_outgoing = 0;
		}
	}
	rds_tcp_set_callbacks(new_sock, conn);
	rds_connect_complete(conn); /* marks RDS_CONN_UP */
	new_sock = NULL;
	ret = 0;
	goto out;
rst_nsk:
	/* reset the newly returned accept sock and bail */
	nsk = new_sock->sk;
	rds_tcp_stats_inc(s_tcp_listen_closed_stale);
	nsk->sk_user_data = NULL;
	nsk->sk_prot->disconnect(nsk, 0);
	tcp_done(nsk);
	new_sock = NULL;
	ret = 0;
out:
	if (rs_tcp)
		mutex_unlock(&rs_tcp->t_conn_lock);
	if (new_sock)
		sock_release(new_sock);
	return ret;
}

void rds_tcp_listen_data_ready(struct sock *sk)
{
	void (*ready)(struct sock *sk);

	rdsdebug("listen data ready sk %p\n", sk);

	read_lock(&sk->sk_callback_lock);
	ready = sk->sk_user_data;
	if (!ready) { /* check for teardown race */
		ready = sk->sk_data_ready;
		goto out;
	}

	/*
	 * ->sk_data_ready is also called for a newly established child socket
	 * before it has been accepted and the accepter has set up their
	 * data_ready.. we only want to queue listen work for our listening
	 * socket
	 */
	if (sk->sk_state == TCP_LISTEN)
		rds_tcp_accept_work(sk);

out:
	read_unlock(&sk->sk_callback_lock);
	ready(sk);
}

struct socket *rds_tcp_listen_init(struct net *net)
{
	struct sockaddr_in sin;
	struct socket *sock = NULL;
	int ret;

	ret = sock_create_kern(net, PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
	if (ret < 0)
		goto out;

	sock->sk->sk_reuse = SK_CAN_REUSE;
	rds_tcp_nonagle(sock);

	write_lock_bh(&sock->sk->sk_callback_lock);
	sock->sk->sk_user_data = sock->sk->sk_data_ready;
	sock->sk->sk_data_ready = rds_tcp_listen_data_ready;
	write_unlock_bh(&sock->sk->sk_callback_lock);

	sin.sin_family = PF_INET;
	sin.sin_addr.s_addr = (__force u32)htonl(INADDR_ANY);
	sin.sin_port = (__force u16)htons(RDS_TCP_PORT);

	ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
	if (ret < 0)
		goto out;

	ret = sock->ops->listen(sock, 64);
	if (ret < 0)
		goto out;

	return sock;
out:
	if (sock)
		sock_release(sock);
	return NULL;
}

void rds_tcp_listen_stop(struct socket *sock)
{
	struct sock *sk;

	if (!sock)
		return;

	sk = sock->sk;

	/* serialize with and prevent further callbacks */
	lock_sock(sk);
	write_lock_bh(&sk->sk_callback_lock);
	if (sk->sk_user_data) {
		sk->sk_data_ready = sk->sk_user_data;
		sk->sk_user_data = NULL;
	}
	write_unlock_bh(&sk->sk_callback_lock);
	release_sock(sk);

	/* wait for accepts to stop and close the socket */
	flush_workqueue(rds_wq);
	sock_release(sock);
}
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>
5a0e3ad6	34	#include <linux/gfp.h>
70041088 AG	35	#include <linux/in.h>
	36	#include <net/tcp.h>
	37
	38	#include "rds.h"
	39	#include "tcp.h"
	40
467fa153	41	int rds_tcp_keepalive(struct socket *sock)
f711a6ae SV	42	{
	43	/* values below based on xs_udp_default_timeout */
	44	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
	45	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
	46	int keepalive = 1;
	47	int ret = 0;
	48
	49	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
	50	(char *)&keepalive, sizeof(keepalive));
	51	if (ret < 0)
	52	goto bail;
	53
	54	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
	55	(char *)&keepcnt, sizeof(keepcnt));
	56	if (ret < 0)
	57	goto bail;
	58
	59	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
	60	(char *)&keepidle, sizeof(keepidle));
	61	if (ret < 0)
	62	goto bail;
	63
	64	/* KEEPINTVL is the interval between successive probes. We follow
	65	* the model in xs_tcp_finish_connecting() and re-use keepidle.
	66	*/
	67	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
	68	(char *)&keepidle, sizeof(keepidle));
	69	bail:
	70	return ret;
	71	}
	72
467fa153	73	int rds_tcp_accept_one(struct socket *sock)
70041088 AG	74	{
	75	struct socket *new_sock = NULL;
	76	struct rds_connection *conn;
	77	int ret;
	78	struct inet_sock *inet;
bd7c5f98 SV	79	struct rds_tcp_connection *rs_tcp = NULL;
	80	int conn_state;
	81	struct sock *nsk;
70041088	82
d5a8ac28 SV	83	ret = sock_create_kern(sock_net(sock->sk), sock->sk->sk_family,
	84	sock->sk->sk_type, sock->sk->sk_protocol,
	85	&new_sock);
70041088 AG	86	if (ret)
	87	goto out;
	88
	89	new_sock->type = sock->type;
	90	new_sock->ops = sock->ops;
	91	ret = sock->ops->accept(sock, new_sock, O_NONBLOCK);
	92	if (ret < 0)
	93	goto out;
	94
f711a6ae SV	95	ret = rds_tcp_keepalive(new_sock);
	96	if (ret < 0)
	97	goto out;
	98
70041088 AG	99	rds_tcp_tune(new_sock);
	100
	101	inet = inet_sk(new_sock->sk);
	102
6884b348 JP	103	rdsdebug("accepted tcp %pI4:%u -> %pI4:%u\n",
	104	&inet->inet_saddr, ntohs(inet->inet_sport),
	105	&inet->inet_daddr, ntohs(inet->inet_dport));
70041088	106
d5a8ac28 SV	107	conn = rds_conn_create(sock_net(sock->sk),
d5a8ac28 SV	108	inet->inet_saddr, inet->inet_daddr,
c720c7e8	109	&rds_tcp_transport, GFP_KERNEL);
70041088 AG	110	if (IS_ERR(conn)) {
	111	ret = PTR_ERR(conn);
	112	goto out;
	113	}
f711a6ae	114	/* An incoming SYN request came in, and TCP just accepted it.
f711a6ae SV	115	*
	116	* If the client reboots, this conn will need to be cleaned up.
	117	* rds_tcp_state_change() will do that cleanup
	118	*/
	119	rs_tcp = (struct rds_tcp_connection *)conn->c_transport_data;
3b20fc38	120	rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING);
bd7c5f98 SV	121	mutex_lock(&rs_tcp->t_conn_lock);
	122	conn_state = rds_conn_state(conn);
	123	if (conn_state != RDS_CONN_CONNECTING && conn_state != RDS_CONN_UP)
	124	goto rst_nsk;
eb192840 SV	125	if (rs_tcp->t_sock) {
	126	/* Need to resolve a duelling SYN between peers.
	127	* We have an outstanding SYN to this peer, which may
	128	* potentially have transitioned to the RDS_CONN_UP state,
	129	* so we must quiesce any send threads before resetting
	130	* c_transport_data.
	131	*/
	132	wait_event(conn->c_waitq,
	133	!test_bit(RDS_IN_XMIT, &conn->c_flags));
	134	if (ntohl(inet->inet_saddr) < ntohl(inet->inet_daddr)) {
bd7c5f98	135	goto rst_nsk;
eb192840 SV	136	} else if (rs_tcp->t_sock) {
	137	rds_tcp_restore_callbacks(rs_tcp->t_sock, rs_tcp);
	138	conn->c_outgoing = 0;
	139	}
	140	}
70041088	141	rds_tcp_set_callbacks(new_sock, conn);
eb192840	142	rds_connect_complete(conn); /* marks RDS_CONN_UP */
70041088 AG	143	new_sock = NULL;
70041088 AG	144	ret = 0;
bd7c5f98 SV	145	goto out;
	146	rst_nsk:
	147	/* reset the newly returned accept sock and bail */
	148	nsk = new_sock->sk;
	149	rds_tcp_stats_inc(s_tcp_listen_closed_stale);
	150	nsk->sk_user_data = NULL;
	151	nsk->sk_prot->disconnect(nsk, 0);
	152	tcp_done(nsk);
	153	new_sock = NULL;
	154	ret = 0;
70041088	155	out:
bd7c5f98 SV	156	if (rs_tcp)
bd7c5f98 SV	157	mutex_unlock(&rs_tcp->t_conn_lock);
70041088 AG	158	if (new_sock)
	159	sock_release(new_sock);
	160	return ret;
	161	}
	162
676d2369	163	void rds_tcp_listen_data_ready(struct sock *sk)
70041088	164	{
676d2369	165	void (ready)(struct sock sk);
70041088 AG	166
	167	rdsdebug("listen data ready sk %p\n", sk);
	168
bfdc587c	169	read_lock(&sk->sk_callback_lock);
70041088	170	ready = sk->sk_user_data;
8690bfa1	171	if (!ready) { /* check for teardown race */
70041088 AG	172	ready = sk->sk_data_ready;
	173	goto out;
	174	}
	175
	176	/*
	177	* ->sk_data_ready is also called for a newly established child socket
	178	* before it has been accepted and the accepter has set up their
	179	* data_ready.. we only want to queue listen work for our listening
	180	* socket
	181	*/
	182	if (sk->sk_state == TCP_LISTEN)
467fa153	183	rds_tcp_accept_work(sk);
70041088 AG	184
70041088 AG	185	out:
bfdc587c	186	read_unlock(&sk->sk_callback_lock);
676d2369	187	ready(sk);
70041088 AG	188	}
70041088 AG	189
467fa153	190	struct socket rds_tcp_listen_init(struct net net)
70041088 AG	191	{
	192	struct sockaddr_in sin;
	193	struct socket *sock = NULL;
	194	int ret;
	195
467fa153	196	ret = sock_create_kern(net, PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
70041088 AG	197	if (ret < 0)
	198	goto out;
	199
4a17fd52	200	sock->sk->sk_reuse = SK_CAN_REUSE;
70041088 AG	201	rds_tcp_nonagle(sock);
	202
	203	write_lock_bh(&sock->sk->sk_callback_lock);
	204	sock->sk->sk_user_data = sock->sk->sk_data_ready;
	205	sock->sk->sk_data_ready = rds_tcp_listen_data_ready;
	206	write_unlock_bh(&sock->sk->sk_callback_lock);
	207
01728371	208	sin.sin_family = PF_INET;
70041088 AG	209	sin.sin_addr.s_addr = (__force u32)htonl(INADDR_ANY);
	210	sin.sin_port = (__force u16)htons(RDS_TCP_PORT);
	211
	212	ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
	213	if (ret < 0)
	214	goto out;
	215
	216	ret = sock->ops->listen(sock, 64);
	217	if (ret < 0)
	218	goto out;
	219
467fa153	220	return sock;
70041088 AG	221	out:
	222	if (sock)
	223	sock_release(sock);
467fa153	224	return NULL;
70041088 AG	225	}
70041088 AG	226
467fa153	227	void rds_tcp_listen_stop(struct socket *sock)
70041088	228	{
70041088 AG	229	struct sock *sk;
70041088 AG	230
8690bfa1	231	if (!sock)
70041088 AG	232	return;
	233
	234	sk = sock->sk;
	235
	236	/* serialize with and prevent further callbacks */
	237	lock_sock(sk);
	238	write_lock_bh(&sk->sk_callback_lock);
	239	if (sk->sk_user_data) {
	240	sk->sk_data_ready = sk->sk_user_data;
	241	sk->sk_user_data = NULL;
	242	}
	243	write_unlock_bh(&sk->sk_callback_lock);
	244	release_sock(sk);
	245
	246	/* wait for accepts to stop and close the socket */
	247	flush_workqueue(rds_wq);
	248	sock_release(sock);
70041088	249	}