[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_single_path.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 rds_conn_path *cp;

	if (!sock) /* module unload or netns delete in progress */
		return -ENETUNREACH;

	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;
	cp = &conn->c_path[0];
	rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING);
	mutex_lock(&rs_tcp->t_conn_path_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.
		 */
		if (ntohl(inet->inet_saddr) < ntohl(inet->inet_daddr) ||
		    !conn->c_path[0].cp_outgoing) {
			goto rst_nsk;
		} else {
			rds_tcp_reset_callbacks(new_sock, cp);
			conn->c_path[0].cp_outgoing = 0;
			/* rds_connect_path_complete() marks RDS_CONN_UP */
			rds_connect_path_complete(cp, RDS_CONN_RESETTING);
		}
	} else {
		rds_tcp_set_callbacks(new_sock, cp);
		rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
	}
	new_sock = NULL;
	ret = 0;
	goto out;
rst_nsk:
	/* reset the newly returned accept sock and bail */
	kernel_sock_shutdown(new_sock, SHUT_RDWR);
	ret = 0;
out:
	if (rs_tcp)
		mutex_unlock(&rs_tcp->t_conn_path_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_bh(&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_bh(&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
0cb43965	38	#include "rds_single_path.h"
70041088 AG	39	#include "rds.h"
	40	#include "tcp.h"
	41
467fa153	42	int rds_tcp_keepalive(struct socket *sock)
f711a6ae SV	43	{
	44	/* values below based on xs_udp_default_timeout */
	45	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
	46	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
	47	int keepalive = 1;
	48	int ret = 0;
	49
	50	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
	51	(char *)&keepalive, sizeof(keepalive));
	52	if (ret < 0)
	53	goto bail;
	54
	55	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
	56	(char *)&keepcnt, sizeof(keepcnt));
	57	if (ret < 0)
	58	goto bail;
	59
	60	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
	61	(char *)&keepidle, sizeof(keepidle));
	62	if (ret < 0)
	63	goto bail;
	64
	65	/* KEEPINTVL is the interval between successive probes. We follow
	66	* the model in xs_tcp_finish_connecting() and re-use keepidle.
	67	*/
	68	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
	69	(char *)&keepidle, sizeof(keepidle));
	70	bail:
	71	return ret;
	72	}
	73
467fa153	74	int rds_tcp_accept_one(struct socket *sock)
70041088 AG	75	{
	76	struct socket *new_sock = NULL;
	77	struct rds_connection *conn;
	78	int ret;
	79	struct inet_sock *inet;
bd7c5f98 SV	80	struct rds_tcp_connection *rs_tcp = NULL;
bd7c5f98 SV	81	int conn_state;
ea3b1ea5	82	struct rds_conn_path *cp;
70041088	83
37e14f4f SV	84	if (!sock) /* module unload or netns delete in progress */
	85	return -ENETUNREACH;
	86
d5a8ac28 SV	87	ret = sock_create_kern(sock_net(sock->sk), sock->sk->sk_family,
	88	sock->sk->sk_type, sock->sk->sk_protocol,
	89	&new_sock);
70041088 AG	90	if (ret)
	91	goto out;
	92
	93	new_sock->type = sock->type;
	94	new_sock->ops = sock->ops;
	95	ret = sock->ops->accept(sock, new_sock, O_NONBLOCK);
	96	if (ret < 0)
	97	goto out;
	98
f711a6ae SV	99	ret = rds_tcp_keepalive(new_sock);
	100	if (ret < 0)
	101	goto out;
	102
70041088 AG	103	rds_tcp_tune(new_sock);
	104
	105	inet = inet_sk(new_sock->sk);
	106
6884b348 JP	107	rdsdebug("accepted tcp %pI4:%u -> %pI4:%u\n",
	108	&inet->inet_saddr, ntohs(inet->inet_sport),
	109	&inet->inet_daddr, ntohs(inet->inet_dport));
70041088	110
d5a8ac28 SV	111	conn = rds_conn_create(sock_net(sock->sk),
d5a8ac28 SV	112	inet->inet_saddr, inet->inet_daddr,
c720c7e8	113	&rds_tcp_transport, GFP_KERNEL);
70041088 AG	114	if (IS_ERR(conn)) {
	115	ret = PTR_ERR(conn);
	116	goto out;
	117	}
f711a6ae	118	/* An incoming SYN request came in, and TCP just accepted it.
f711a6ae SV	119	*
	120	* If the client reboots, this conn will need to be cleaned up.
	121	* rds_tcp_state_change() will do that cleanup
	122	*/
	123	rs_tcp = (struct rds_tcp_connection *)conn->c_transport_data;
ea3b1ea5	124	cp = &conn->c_path[0];
3b20fc38	125	rds_conn_transition(conn, RDS_CONN_DOWN, RDS_CONN_CONNECTING);
02105b2c	126	mutex_lock(&rs_tcp->t_conn_path_lock);
bd7c5f98 SV	127	conn_state = rds_conn_state(conn);
	128	if (conn_state != RDS_CONN_CONNECTING && conn_state != RDS_CONN_UP)
	129	goto rst_nsk;
eb192840 SV	130	if (rs_tcp->t_sock) {
	131	/* Need to resolve a duelling SYN between peers.
	132	* We have an outstanding SYN to this peer, which may
	133	* potentially have transitioned to the RDS_CONN_UP state,
	134	* so we must quiesce any send threads before resetting
	135	* c_transport_data.
	136	*/
c948bb5c	137	if (ntohl(inet->inet_saddr) < ntohl(inet->inet_daddr) \|\|
0cb43965	138	!conn->c_path[0].cp_outgoing) {
bd7c5f98	139	goto rst_nsk;
c948bb5c	140	} else {
ea3b1ea5	141	rds_tcp_reset_callbacks(new_sock, cp);
0cb43965	142	conn->c_path[0].cp_outgoing = 0;
9c79440e	143	/* rds_connect_path_complete() marks RDS_CONN_UP */
ea3b1ea5	144	rds_connect_path_complete(cp, RDS_CONN_RESETTING);
eb192840	145	}
335b48d9	146	} else {
ea3b1ea5 SV	147	rds_tcp_set_callbacks(new_sock, cp);
ea3b1ea5 SV	148	rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
eb192840	149	}
70041088 AG	150	new_sock = NULL;
70041088 AG	151	ret = 0;
bd7c5f98 SV	152	goto out;
	153	rst_nsk:
	154	/* reset the newly returned accept sock and bail */
335b48d9	155	kernel_sock_shutdown(new_sock, SHUT_RDWR);
bd7c5f98	156	ret = 0;
70041088	157	out:
bd7c5f98	158	if (rs_tcp)
02105b2c	159	mutex_unlock(&rs_tcp->t_conn_path_lock);
70041088 AG	160	if (new_sock)
	161	sock_release(new_sock);
	162	return ret;
	163	}
	164
676d2369	165	void rds_tcp_listen_data_ready(struct sock *sk)
70041088	166	{
676d2369	167	void (ready)(struct sock sk);
70041088 AG	168
	169	rdsdebug("listen data ready sk %p\n", sk);
	170
38036629	171	read_lock_bh(&sk->sk_callback_lock);
70041088	172	ready = sk->sk_user_data;
8690bfa1	173	if (!ready) { /* check for teardown race */
70041088 AG	174	ready = sk->sk_data_ready;
	175	goto out;
	176	}
	177
	178	/*
	179	* ->sk_data_ready is also called for a newly established child socket
	180	* before it has been accepted and the accepter has set up their
	181	* data_ready.. we only want to queue listen work for our listening
	182	* socket
	183	*/
	184	if (sk->sk_state == TCP_LISTEN)
467fa153	185	rds_tcp_accept_work(sk);
70041088 AG	186
70041088 AG	187	out:
38036629	188	read_unlock_bh(&sk->sk_callback_lock);
676d2369	189	ready(sk);
70041088 AG	190	}
70041088 AG	191
467fa153	192	struct socket rds_tcp_listen_init(struct net net)
70041088 AG	193	{
	194	struct sockaddr_in sin;
	195	struct socket *sock = NULL;
	196	int ret;
	197
467fa153	198	ret = sock_create_kern(net, PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
70041088 AG	199	if (ret < 0)
	200	goto out;
	201
4a17fd52	202	sock->sk->sk_reuse = SK_CAN_REUSE;
70041088 AG	203	rds_tcp_nonagle(sock);
	204
	205	write_lock_bh(&sock->sk->sk_callback_lock);
	206	sock->sk->sk_user_data = sock->sk->sk_data_ready;
	207	sock->sk->sk_data_ready = rds_tcp_listen_data_ready;
	208	write_unlock_bh(&sock->sk->sk_callback_lock);
	209
01728371	210	sin.sin_family = PF_INET;
70041088 AG	211	sin.sin_addr.s_addr = (__force u32)htonl(INADDR_ANY);
	212	sin.sin_port = (__force u16)htons(RDS_TCP_PORT);
	213
	214	ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
	215	if (ret < 0)
	216	goto out;
	217
	218	ret = sock->ops->listen(sock, 64);
	219	if (ret < 0)
	220	goto out;
	221
467fa153	222	return sock;
70041088 AG	223	out:
	224	if (sock)
	225	sock_release(sock);
467fa153	226	return NULL;
70041088 AG	227	}
70041088 AG	228
467fa153	229	void rds_tcp_listen_stop(struct socket *sock)
70041088	230	{
70041088 AG	231	struct sock *sk;
70041088 AG	232
8690bfa1	233	if (!sock)
70041088 AG	234	return;
	235
	236	sk = sock->sk;
	237
	238	/* serialize with and prevent further callbacks */
	239	lock_sock(sk);
	240	write_lock_bh(&sk->sk_callback_lock);
	241	if (sk->sk_user_data) {
	242	sk->sk_data_ready = sk->sk_user_data;
	243	sk->sk_user_data = NULL;
	244	}
	245	write_unlock_bh(&sk->sk_callback_lock);
	246	release_sock(sk);
	247
	248	/* wait for accepts to stop and close the socket */
	249	flush_workqueue(rds_wq);
	250	sock_release(sock);
70041088	251	}