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

/*
 * Copyright (c) 2006, 2018 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/gfp.h>
#include <linux/in.h>
#include <net/tcp.h>

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

void 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 */

	sock_set_keepalive(sock->sk);
	tcp_sock_set_keepcnt(sock->sk, keepcnt);
	tcp_sock_set_keepidle(sock->sk, keepidle);
	/* KEEPINTVL is the interval between successive probes. We follow
	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
	 */
	tcp_sock_set_keepintvl(sock->sk, keepidle);
}

/* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
 * client's ipaddr < server's ipaddr. Otherwise, close the accepted
 * socket and force a reconneect from smaller -> larger ip addr. The reason
 * we special case cp_index 0 is to allow the rds probe ping itself to itself
 * get through efficiently.
 * Since reconnects are only initiated from the node with the numerically
 * smaller ip address, we recycle conns in RDS_CONN_ERROR on the passive side
 * by moving them to CONNECTING in this function.
 */
static
struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
{
	int i;
	int npaths = max_t(int, 1, conn->c_npaths);

	/* for mprds, all paths MUST be initiated by the peer
	 * with the smaller address.
	 */
	if (rds_addr_cmp(&conn->c_faddr, &conn->c_laddr) >= 0) {
		/* Make sure we initiate at least one path if this
		 * has not already been done; rds_start_mprds() will
		 * take care of additional paths, if necessary.
		 */
		if (npaths == 1)
			rds_conn_path_connect_if_down(&conn->c_path[0]);
		return NULL;
	}

	for (i = 0; i < npaths; i++) {
		struct rds_conn_path *cp = &conn->c_path[i];

		if (rds_conn_path_transition(cp, RDS_CONN_DOWN,
					     RDS_CONN_CONNECTING) ||
		    rds_conn_path_transition(cp, RDS_CONN_ERROR,
					     RDS_CONN_CONNECTING)) {
			return cp->cp_transport_data;
		}
	}
	return NULL;
}

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;
	struct in6_addr *my_addr, *peer_addr;
#if !IS_ENABLED(CONFIG_IPV6)
	struct in6_addr saddr, daddr;
#endif
	int dev_if = 0;

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

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

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

	/* sock_create_lite() does not get a hold on the owner module so we
	 * need to do it here.  Note that sock_release() uses sock->ops to
	 * determine if it needs to decrement the reference count.  So set
	 * sock->ops after calling accept() in case that fails.  And there's
	 * no need to do try_module_get() as the listener should have a hold
	 * already.
	 */
	new_sock->ops = sock->ops;
	__module_get(new_sock->ops->owner);

	rds_tcp_keepalive(new_sock);
	rds_tcp_tune(new_sock);

	inet = inet_sk(new_sock->sk);

#if IS_ENABLED(CONFIG_IPV6)
	my_addr = &new_sock->sk->sk_v6_rcv_saddr;
	peer_addr = &new_sock->sk->sk_v6_daddr;
#else
	ipv6_addr_set_v4mapped(inet->inet_saddr, &saddr);
	ipv6_addr_set_v4mapped(inet->inet_daddr, &daddr);
	my_addr = &saddr;
	peer_addr = &daddr;
#endif
	rdsdebug("accepted family %d tcp %pI6c:%u -> %pI6c:%u\n",
		 sock->sk->sk_family,
		 my_addr, ntohs(inet->inet_sport),
		 peer_addr, ntohs(inet->inet_dport));

#if IS_ENABLED(CONFIG_IPV6)
	/* sk_bound_dev_if is not set if the peer address is not link local
	 * address.  In this case, it happens that mcast_oif is set.  So
	 * just use it.
	 */
	if ((ipv6_addr_type(my_addr) & IPV6_ADDR_LINKLOCAL) &&
	    !(ipv6_addr_type(peer_addr) & IPV6_ADDR_LINKLOCAL)) {
		struct ipv6_pinfo *inet6;

		inet6 = inet6_sk(new_sock->sk);
		dev_if = inet6->mcast_oif;
	} else {
		dev_if = new_sock->sk->sk_bound_dev_if;
	}
#endif

	conn = rds_conn_create(sock_net(sock->sk),
			       my_addr, peer_addr,
			       &rds_tcp_transport, 0, GFP_KERNEL, dev_if);

	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 = rds_tcp_accept_one_path(conn);
	if (!rs_tcp)
		goto rst_nsk;
	mutex_lock(&rs_tcp->t_conn_path_lock);
	cp = rs_tcp->t_cpath;
	conn_state = rds_conn_path_state(cp);
	WARN_ON(conn_state == RDS_CONN_UP);
	if (conn_state != RDS_CONN_CONNECTING && conn_state != RDS_CONN_ERROR)
		goto rst_nsk;
	if (rs_tcp->t_sock) {
		/* Duelling SYN has been handled in rds_tcp_accept_one() */
		rds_tcp_reset_callbacks(new_sock, cp);
		/* 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;
	if (conn->c_npaths == 0)
		rds_send_ping(cp->cp_conn, cp->cp_index);
	goto out;
rst_nsk:
	/* reset the newly returned accept sock and bail.
	 * It is safe to set linger on new_sock because the RDS connection
	 * has not been brought up on new_sock, so no RDS-level data could
	 * be pending on it. By setting linger, we achieve the side-effect
	 * of avoiding TIME_WAIT state on new_sock.
	 */
	sock_no_linger(new_sock->sk);
	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
	 *
	 * (*ready)() may be null if we are racing with netns delete, and
	 * the listen socket is being torn down.
	 */
	if (sk->sk_state == TCP_LISTEN)
		rds_tcp_accept_work(sk);
	else
		ready = rds_tcp_listen_sock_def_readable(sock_net(sk));

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

struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
{
	struct socket *sock = NULL;
	struct sockaddr_storage ss;
	struct sockaddr_in6 *sin6;
	struct sockaddr_in *sin;
	int addr_len;
	int ret;

	ret = sock_create_kern(net, isv6 ? PF_INET6 : PF_INET, SOCK_STREAM,
			       IPPROTO_TCP, &sock);
	if (ret < 0) {
		rdsdebug("could not create %s listener socket: %d\n",
			 isv6 ? "IPv6" : "IPv4", ret);
		goto out;
	}

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

	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);

	if (isv6) {
		sin6 = (struct sockaddr_in6 *)&ss;
		sin6->sin6_family = PF_INET6;
		sin6->sin6_addr = in6addr_any;
		sin6->sin6_port = (__force u16)htons(RDS_TCP_PORT);
		sin6->sin6_scope_id = 0;
		sin6->sin6_flowinfo = 0;
		addr_len = sizeof(*sin6);
	} else {
		sin = (struct sockaddr_in *)&ss;
		sin->sin_family = PF_INET;
		sin->sin_addr.s_addr = INADDR_ANY;
		sin->sin_port = (__force u16)htons(RDS_TCP_PORT);
		addr_len = sizeof(*sin);
	}

	ret = sock->ops->bind(sock, (struct sockaddr *)&ss, addr_len);
	if (ret < 0) {
		rdsdebug("could not bind %s listener socket: %d\n",
			 isv6 ? "IPv6" : "IPv4", ret);
		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 work_struct *acceptor)
{
	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);
	flush_work(acceptor);
	sock_release(sock);
}
Commit	Line	Data
70041088	1	/*
eee2fa6a	2	* Copyright (c) 2006, 2018 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>
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
480aeb96	41	void 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 */
f711a6ae	46
ce3d9544	47	sock_set_keepalive(sock->sk);
480aeb96	48	tcp_sock_set_keepcnt(sock->sk, keepcnt);
71c48eb8	49	tcp_sock_set_keepidle(sock->sk, keepidle);
f711a6ae SV	50	/* KEEPINTVL is the interval between successive probes. We follow
	51	* the model in xs_tcp_finish_connecting() and re-use keepidle.
	52	*/
d41ecaac	53	tcp_sock_set_keepintvl(sock->sk, keepidle);
f711a6ae SV	54	}
f711a6ae SV	55
5916e2c1 SV	56	/* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
	57	* client's ipaddr < server's ipaddr. Otherwise, close the accepted
	58	* socket and force a reconneect from smaller -> larger ip addr. The reason
	59	* we special case cp_index 0 is to allow the rds probe ping itself to itself
	60	* get through efficiently.
	61	* Since reconnects are only initiated from the node with the numerically
	62	* smaller ip address, we recycle conns in RDS_CONN_ERROR on the passive side
	63	* by moving them to CONNECTING in this function.
	64	*/
bb789763	65	static
5916e2c1 SV	66	struct rds_tcp_connection rds_tcp_accept_one_path(struct rds_connection conn)
	67	{
	68	int i;
1a0e100f	69	int npaths = max_t(int, 1, conn->c_npaths);
5916e2c1	70
1a0e100f	71	/* for mprds, all paths MUST be initiated by the peer
5916e2c1 SV	72	* with the smaller address.
5916e2c1 SV	73	*/
eee2fa6a	74	if (rds_addr_cmp(&conn->c_faddr, &conn->c_laddr) >= 0) {
1a0e100f SV	75	/* Make sure we initiate at least one path if this
	76	* has not already been done; rds_start_mprds() will
	77	* take care of additional paths, if necessary.
	78	*/
	79	if (npaths == 1)
	80	rds_conn_path_connect_if_down(&conn->c_path[0]);
5916e2c1	81	return NULL;
1a0e100f	82	}
5916e2c1	83
117d15bb	84	for (i = 0; i < npaths; i++) {
5916e2c1 SV	85	struct rds_conn_path *cp = &conn->c_path[i];
	86
	87	if (rds_conn_path_transition(cp, RDS_CONN_DOWN,
	88	RDS_CONN_CONNECTING) \|\|
	89	rds_conn_path_transition(cp, RDS_CONN_ERROR,
	90	RDS_CONN_CONNECTING)) {
	91	return cp->cp_transport_data;
	92	}
	93	}
	94	return NULL;
	95	}
	96
467fa153	97	int rds_tcp_accept_one(struct socket *sock)
70041088 AG	98	{
	99	struct socket *new_sock = NULL;
	100	struct rds_connection *conn;
	101	int ret;
	102	struct inet_sock *inet;
bd7c5f98 SV	103	struct rds_tcp_connection *rs_tcp = NULL;
bd7c5f98 SV	104	int conn_state;
ea3b1ea5	105	struct rds_conn_path *cp;
1e2b44e7	106	struct in6_addr my_addr, peer_addr;
e65d4d96 KCP	107	#if !IS_ENABLED(CONFIG_IPV6)
	108	struct in6_addr saddr, daddr;
	109	#endif
	110	int dev_if = 0;
70041088	111
37e14f4f SV	112	if (!sock) /* module unload or netns delete in progress */
	113	return -ENETUNREACH;
	114
0933a578	115	ret = sock_create_lite(sock->sk->sk_family,
d5a8ac28 SV	116	sock->sk->sk_type, sock->sk->sk_protocol,
d5a8ac28 SV	117	&new_sock);
70041088 AG	118	if (ret)
	119	goto out;
	120
cdfbabfb	121	ret = sock->ops->accept(sock, new_sock, O_NONBLOCK, true);
70041088 AG	122	if (ret < 0)
	123	goto out;
	124
84eef2b2 KCP	125	/* sock_create_lite() does not get a hold on the owner module so we
	126	* need to do it here. Note that sock_release() uses sock->ops to
	127	* determine if it needs to decrement the reference count. So set
	128	* sock->ops after calling accept() in case that fails. And there's
	129	* no need to do try_module_get() as the listener should have a hold
	130	* already.
	131	*/
	132	new_sock->ops = sock->ops;
	133	__module_get(new_sock->ops->owner);
	134
480aeb96	135	rds_tcp_keepalive(new_sock);
70041088 AG	136	rds_tcp_tune(new_sock);
	137
	138	inet = inet_sk(new_sock->sk);
	139
e65d4d96	140	#if IS_ENABLED(CONFIG_IPV6)
1e2b44e7 KCP	141	my_addr = &new_sock->sk->sk_v6_rcv_saddr;
1e2b44e7 KCP	142	peer_addr = &new_sock->sk->sk_v6_daddr;
e65d4d96 KCP	143	#else
	144	ipv6_addr_set_v4mapped(inet->inet_saddr, &saddr);
	145	ipv6_addr_set_v4mapped(inet->inet_daddr, &daddr);
	146	my_addr = &saddr;
	147	peer_addr = &daddr;
	148	#endif
	149	rdsdebug("accepted family %d tcp %pI6c:%u -> %pI6c:%u\n",
	150	sock->sk->sk_family,
1e2b44e7 KCP	151	my_addr, ntohs(inet->inet_sport),
1e2b44e7 KCP	152	peer_addr, ntohs(inet->inet_dport));
70041088	153
e65d4d96	154	#if IS_ENABLED(CONFIG_IPV6)
1e2b44e7 KCP	155	/* sk_bound_dev_if is not set if the peer address is not link local
	156	* address. In this case, it happens that mcast_oif is set. So
	157	* just use it.
	158	*/
	159	if ((ipv6_addr_type(my_addr) & IPV6_ADDR_LINKLOCAL) &&
	160	!(ipv6_addr_type(peer_addr) & IPV6_ADDR_LINKLOCAL)) {
	161	struct ipv6_pinfo *inet6;
	162
	163	inet6 = inet6_sk(new_sock->sk);
	164	dev_if = inet6->mcast_oif;
	165	} else {
	166	dev_if = new_sock->sk->sk_bound_dev_if;
	167	}
e65d4d96 KCP	168	#endif
e65d4d96 KCP	169
d5a8ac28	170	conn = rds_conn_create(sock_net(sock->sk),
e65d4d96	171	my_addr, peer_addr,
3eb45036	172	&rds_tcp_transport, 0, GFP_KERNEL, dev_if);
eee2fa6a	173
70041088 AG	174	if (IS_ERR(conn)) {
	175	ret = PTR_ERR(conn);
	176	goto out;
	177	}
f711a6ae	178	/* An incoming SYN request came in, and TCP just accepted it.
f711a6ae SV	179	*
	180	* If the client reboots, this conn will need to be cleaned up.
	181	* rds_tcp_state_change() will do that cleanup
	182	*/
5916e2c1 SV	183	rs_tcp = rds_tcp_accept_one_path(conn);
	184	if (!rs_tcp)
	185	goto rst_nsk;
02105b2c	186	mutex_lock(&rs_tcp->t_conn_path_lock);
5916e2c1 SV	187	cp = rs_tcp->t_cpath;
5916e2c1 SV	188	conn_state = rds_conn_path_state(cp);
1a0e100f SV	189	WARN_ON(conn_state == RDS_CONN_UP);
1a0e100f SV	190	if (conn_state != RDS_CONN_CONNECTING && conn_state != RDS_CONN_ERROR)
bd7c5f98	191	goto rst_nsk;
eb192840	192	if (rs_tcp->t_sock) {
41500c3e SV	193	/* Duelling SYN has been handled in rds_tcp_accept_one() */
	194	rds_tcp_reset_callbacks(new_sock, cp);
	195	/* rds_connect_path_complete() marks RDS_CONN_UP */
	196	rds_connect_path_complete(cp, RDS_CONN_RESETTING);
335b48d9	197	} else {
ea3b1ea5 SV	198	rds_tcp_set_callbacks(new_sock, cp);
ea3b1ea5 SV	199	rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
eb192840	200	}
70041088 AG	201	new_sock = NULL;
70041088 AG	202	ret = 0;
69b92b5b SV	203	if (conn->c_npaths == 0)
69b92b5b SV	204	rds_send_ping(cp->cp_conn, cp->cp_index);
bd7c5f98 SV	205	goto out;
bd7c5f98 SV	206	rst_nsk:
10beea7d SV	207	/* reset the newly returned accept sock and bail.
	208	* It is safe to set linger on new_sock because the RDS connection
	209	* has not been brought up on new_sock, so no RDS-level data could
	210	* be pending on it. By setting linger, we achieve the side-effect
	211	* of avoiding TIME_WAIT state on new_sock.
	212	*/
c433594c	213	sock_no_linger(new_sock->sk);
335b48d9	214	kernel_sock_shutdown(new_sock, SHUT_RDWR);
bd7c5f98	215	ret = 0;
70041088	216	out:
bd7c5f98	217	if (rs_tcp)
02105b2c	218	mutex_unlock(&rs_tcp->t_conn_path_lock);
70041088 AG	219	if (new_sock)
	220	sock_release(new_sock);
	221	return ret;
	222	}
	223
676d2369	224	void rds_tcp_listen_data_ready(struct sock *sk)
70041088	225	{
676d2369	226	void (ready)(struct sock sk);
70041088 AG	227
	228	rdsdebug("listen data ready sk %p\n", sk);
	229
38036629	230	read_lock_bh(&sk->sk_callback_lock);
70041088	231	ready = sk->sk_user_data;
8690bfa1	232	if (!ready) { /* check for teardown race */
70041088 AG	233	ready = sk->sk_data_ready;
	234	goto out;
	235	}
	236
	237	/*
	238	* ->sk_data_ready is also called for a newly established child socket
	239	* before it has been accepted and the accepter has set up their
	240	* data_ready.. we only want to queue listen work for our listening
	241	* socket
b21dd450 SV	242	*
	243	* (*ready)() may be null if we are racing with netns delete, and
	244	* the listen socket is being torn down.
70041088 AG	245	*/
70041088 AG	246	if (sk->sk_state == TCP_LISTEN)
467fa153	247	rds_tcp_accept_work(sk);
a93d01f5 SV	248	else
a93d01f5 SV	249	ready = rds_tcp_listen_sock_def_readable(sock_net(sk));
70041088 AG	250
70041088 AG	251	out:
38036629	252	read_unlock_bh(&sk->sk_callback_lock);
b21dd450 SV	253	if (ready)
b21dd450 SV	254	ready(sk);
70041088 AG	255	}
70041088 AG	256
1e2b44e7	257	struct socket rds_tcp_listen_init(struct net net, bool isv6)
70041088	258	{
70041088	259	struct socket *sock = NULL;
1e2b44e7 KCP	260	struct sockaddr_storage ss;
	261	struct sockaddr_in6 *sin6;
	262	struct sockaddr_in *sin;
	263	int addr_len;
70041088 AG	264	int ret;
70041088 AG	265
1e2b44e7 KCP	266	ret = sock_create_kern(net, isv6 ? PF_INET6 : PF_INET, SOCK_STREAM,
	267	IPPROTO_TCP, &sock);
	268	if (ret < 0) {
	269	rdsdebug("could not create %s listener socket: %d\n",
	270	isv6 ? "IPv6" : "IPv4", ret);
70041088	271	goto out;
1e2b44e7	272	}
70041088	273
4a17fd52	274	sock->sk->sk_reuse = SK_CAN_REUSE;
12abc5ee	275	tcp_sock_set_nodelay(sock->sk);
70041088 AG	276
	277	write_lock_bh(&sock->sk->sk_callback_lock);
	278	sock->sk->sk_user_data = sock->sk->sk_data_ready;
	279	sock->sk->sk_data_ready = rds_tcp_listen_data_ready;
	280	write_unlock_bh(&sock->sk->sk_callback_lock);
	281
1e2b44e7 KCP	282	if (isv6) {
	283	sin6 = (struct sockaddr_in6 *)&ss;
	284	sin6->sin6_family = PF_INET6;
	285	sin6->sin6_addr = in6addr_any;
	286	sin6->sin6_port = (__force u16)htons(RDS_TCP_PORT);
	287	sin6->sin6_scope_id = 0;
	288	sin6->sin6_flowinfo = 0;
	289	addr_len = sizeof(*sin6);
	290	} else {
	291	sin = (struct sockaddr_in *)&ss;
	292	sin->sin_family = PF_INET;
	293	sin->sin_addr.s_addr = INADDR_ANY;
	294	sin->sin_port = (__force u16)htons(RDS_TCP_PORT);
	295	addr_len = sizeof(*sin);
	296	}
70041088	297
1e2b44e7 KCP	298	ret = sock->ops->bind(sock, (struct sockaddr *)&ss, addr_len);
	299	if (ret < 0) {
	300	rdsdebug("could not bind %s listener socket: %d\n",
	301	isv6 ? "IPv6" : "IPv4", ret);
70041088	302	goto out;
1e2b44e7	303	}
70041088 AG	304
	305	ret = sock->ops->listen(sock, 64);
	306	if (ret < 0)
	307	goto out;
	308
467fa153	309	return sock;
70041088 AG	310	out:
	311	if (sock)
	312	sock_release(sock);
467fa153	313	return NULL;
70041088 AG	314	}
70041088 AG	315
b21dd450	316	void rds_tcp_listen_stop(struct socket sock, struct work_struct acceptor)
70041088	317	{
70041088 AG	318	struct sock *sk;
70041088 AG	319
8690bfa1	320	if (!sock)
70041088 AG	321	return;
	322
	323	sk = sock->sk;
	324
	325	/* serialize with and prevent further callbacks */
	326	lock_sock(sk);
	327	write_lock_bh(&sk->sk_callback_lock);
	328	if (sk->sk_user_data) {
	329	sk->sk_data_ready = sk->sk_user_data;
	330	sk->sk_user_data = NULL;
	331	}
	332	write_unlock_bh(&sk->sk_callback_lock);
	333	release_sock(sk);
	334
	335	/* wait for accepts to stop and close the socket */
	336	flush_workqueue(rds_wq);
b21dd450	337	flush_work(acceptor);
70041088	338	sock_release(sock);
70041088	339	}