[net-next-2.6.git] / net / rds / af_rds.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/module.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/in.h>
#include <linux/poll.h>
#include <net/sock.h>

#include "rds.h"
#include "rdma.h"

/* this is just used for stats gathering :/ */
static DEFINE_SPINLOCK(rds_sock_lock);
static unsigned long rds_sock_count;
static LIST_HEAD(rds_sock_list);
DECLARE_WAIT_QUEUE_HEAD(rds_poll_waitq);

/*
 * This is called as the final descriptor referencing this socket is closed.
 * We have to unbind the socket so that another socket can be bound to the
 * address it was using.
 *
 * We have to be careful about racing with the incoming path.  sock_orphan()
 * sets SOCK_DEAD and we use that as an indicator to the rx path that new
 * messages shouldn't be queued.
 */
static int rds_release(struct socket *sock)
{
	struct sock *sk = sock->sk;
	struct rds_sock *rs;
	unsigned long flags;

	if (sk == NULL)
		goto out;

	rs = rds_sk_to_rs(sk);

	sock_orphan(sk);
	/* Note - rds_clear_recv_queue grabs rs_recv_lock, so
	 * that ensures the recv path has completed messing
	 * with the socket. */
	rds_clear_recv_queue(rs);
	rds_cong_remove_socket(rs);
	rds_remove_bound(rs);
	rds_send_drop_to(rs, NULL);
	rds_rdma_drop_keys(rs);
	rds_notify_queue_get(rs, NULL);

	spin_lock_irqsave(&rds_sock_lock, flags);
	list_del_init(&rs->rs_item);
	rds_sock_count--;
	spin_unlock_irqrestore(&rds_sock_lock, flags);

	sock->sk = NULL;
	sock_put(sk);
out:
	return 0;
}

/*
 * Careful not to race with rds_release -> sock_orphan which clears sk_sleep.
 * _bh() isn't OK here, we're called from interrupt handlers.  It's probably OK
 * to wake the waitqueue after sk_sleep is clear as we hold a sock ref, but
 * this seems more conservative.
 * NB - normally, one would use sk_callback_lock for this, but we can
 * get here from interrupts, whereas the network code grabs sk_callback_lock
 * with _lock_bh only - so relying on sk_callback_lock introduces livelocks.
 */
void rds_wake_sk_sleep(struct rds_sock *rs)
{
	unsigned long flags;

	read_lock_irqsave(&rs->rs_recv_lock, flags);
	__rds_wake_sk_sleep(rds_rs_to_sk(rs));
	read_unlock_irqrestore(&rs->rs_recv_lock, flags);
}

static int rds_getname(struct socket *sock, struct sockaddr *uaddr,
		       int *uaddr_len, int peer)
{
	struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
	struct rds_sock *rs = rds_sk_to_rs(sock->sk);

	memset(sin->sin_zero, 0, sizeof(sin->sin_zero));

	/* racey, don't care */
	if (peer) {
		if (!rs->rs_conn_addr)
			return -ENOTCONN;

		sin->sin_port = rs->rs_conn_port;
		sin->sin_addr.s_addr = rs->rs_conn_addr;
	} else {
		sin->sin_port = rs->rs_bound_port;
		sin->sin_addr.s_addr = rs->rs_bound_addr;
	}

	sin->sin_family = AF_INET;

	*uaddr_len = sizeof(*sin);
	return 0;
}

/*
 * RDS' poll is without a doubt the least intuitive part of the interface,
 * as POLLIN and POLLOUT do not behave entirely as you would expect from
 * a network protocol.
 *
 * POLLIN is asserted if
 *  -	there is data on the receive queue.
 *  -	to signal that a previously congested destination may have become
 *	uncongested
 *  -	A notification has been queued to the socket (this can be a congestion
 *	update, or a RDMA completion).
 *
 * POLLOUT is asserted if there is room on the send queue. This does not mean
 * however, that the next sendmsg() call will succeed. If the application tries
 * to send to a congested destination, the system call may still fail (and
 * return ENOBUFS).
 */
static unsigned int rds_poll(struct file *file, struct socket *sock,
			     poll_table *wait)
{
	struct sock *sk = sock->sk;
	struct rds_sock *rs = rds_sk_to_rs(sk);
	unsigned int mask = 0;
	unsigned long flags;

	poll_wait(file, sk->sk_sleep, wait);

	poll_wait(file, &rds_poll_waitq, wait);

	read_lock_irqsave(&rs->rs_recv_lock, flags);
	if (!rs->rs_cong_monitor) {
		/* When a congestion map was updated, we signal POLLIN for
		 * "historical" reasons. Applications can also poll for
		 * WRBAND instead. */
		if (rds_cong_updated_since(&rs->rs_cong_track))
			mask |= (POLLIN | POLLRDNORM | POLLWRBAND);
	} else {
		spin_lock(&rs->rs_lock);
		if (rs->rs_cong_notify)
			mask |= (POLLIN | POLLRDNORM);
		spin_unlock(&rs->rs_lock);
	}
	if (!list_empty(&rs->rs_recv_queue)
	 || !list_empty(&rs->rs_notify_queue))
		mask |= (POLLIN | POLLRDNORM);
	if (rs->rs_snd_bytes < rds_sk_sndbuf(rs))
		mask |= (POLLOUT | POLLWRNORM);
	read_unlock_irqrestore(&rs->rs_recv_lock, flags);

	return mask;
}

static int rds_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
{
	return -ENOIOCTLCMD;
}

static int rds_cancel_sent_to(struct rds_sock *rs, char __user *optval,
			      int len)
{
	struct sockaddr_in sin;
	int ret = 0;

	/* racing with another thread binding seems ok here */
	if (rs->rs_bound_addr == 0) {
		ret = -ENOTCONN; /* XXX not a great errno */
		goto out;
	}

	if (len < sizeof(struct sockaddr_in)) {
		ret = -EINVAL;
		goto out;
	}

	if (copy_from_user(&sin, optval, sizeof(sin))) {
		ret = -EFAULT;
		goto out;
	}

	rds_send_drop_to(rs, &sin);
out:
	return ret;
}

static int rds_set_bool_option(unsigned char *optvar, char __user *optval,
			       int optlen)
{
	int value;

	if (optlen < sizeof(int))
		return -EINVAL;
	if (get_user(value, (int __user *) optval))
		return -EFAULT;
	*optvar = !!value;
	return 0;
}

static int rds_cong_monitor(struct rds_sock *rs, char __user *optval,
			    int optlen)
{
	int ret;

	ret = rds_set_bool_option(&rs->rs_cong_monitor, optval, optlen);
	if (ret == 0) {
		if (rs->rs_cong_monitor) {
			rds_cong_add_socket(rs);
		} else {
			rds_cong_remove_socket(rs);
			rs->rs_cong_mask = 0;
			rs->rs_cong_notify = 0;
		}
	}
	return ret;
}

static int rds_setsockopt(struct socket *sock, int level, int optname,
			  char __user *optval, int optlen)
{
	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	int ret;

	if (level != SOL_RDS) {
		ret = -ENOPROTOOPT;
		goto out;
	}

	switch (optname) {
	case RDS_CANCEL_SENT_TO:
		ret = rds_cancel_sent_to(rs, optval, optlen);
		break;
	case RDS_GET_MR:
		ret = rds_get_mr(rs, optval, optlen);
		break;
	case RDS_FREE_MR:
		ret = rds_free_mr(rs, optval, optlen);
		break;
	case RDS_RECVERR:
		ret = rds_set_bool_option(&rs->rs_recverr, optval, optlen);
		break;
	case RDS_CONG_MONITOR:
		ret = rds_cong_monitor(rs, optval, optlen);
		break;
	default:
		ret = -ENOPROTOOPT;
	}
out:
	return ret;
}

static int rds_getsockopt(struct socket *sock, int level, int optname,
			  char __user *optval, int __user *optlen)
{
	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	int ret = -ENOPROTOOPT, len;

	if (level != SOL_RDS)
		goto out;

	if (get_user(len, optlen)) {
		ret = -EFAULT;
		goto out;
	}

	switch (optname) {
	case RDS_INFO_FIRST ... RDS_INFO_LAST:
		ret = rds_info_getsockopt(sock, optname, optval,
					  optlen);
		break;

	case RDS_RECVERR:
		if (len < sizeof(int))
			ret = -EINVAL;
		else
		if (put_user(rs->rs_recverr, (int __user *) optval)
		 || put_user(sizeof(int), optlen))
			ret = -EFAULT;
		else
			ret = 0;
		break;
	default:
		break;
	}

out:
	return ret;

}

static int rds_connect(struct socket *sock, struct sockaddr *uaddr,
		       int addr_len, int flags)
{
	struct sock *sk = sock->sk;
	struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
	struct rds_sock *rs = rds_sk_to_rs(sk);
	int ret = 0;

	lock_sock(sk);

	if (addr_len != sizeof(struct sockaddr_in)) {
		ret = -EINVAL;
		goto out;
	}

	if (sin->sin_family != AF_INET) {
		ret = -EAFNOSUPPORT;
		goto out;
	}

	if (sin->sin_addr.s_addr == htonl(INADDR_ANY)) {
		ret = -EDESTADDRREQ;
		goto out;
	}

	rs->rs_conn_addr = sin->sin_addr.s_addr;
	rs->rs_conn_port = sin->sin_port;

out:
	release_sock(sk);
	return ret;
}

static struct proto rds_proto = {
	.name	  = "RDS",
	.owner	  = THIS_MODULE,
	.obj_size = sizeof(struct rds_sock),
};

static struct proto_ops rds_proto_ops = {
	.family =	AF_RDS,
	.owner =	THIS_MODULE,
	.release =	rds_release,
	.bind =		rds_bind,
	.connect =	rds_connect,
	.socketpair =	sock_no_socketpair,
	.accept =	sock_no_accept,
	.getname =	rds_getname,
	.poll =		rds_poll,
	.ioctl =	rds_ioctl,
	.listen =	sock_no_listen,
	.shutdown =	sock_no_shutdown,
	.setsockopt =	rds_setsockopt,
	.getsockopt =	rds_getsockopt,
	.sendmsg =	rds_sendmsg,
	.recvmsg =	rds_recvmsg,
	.mmap =		sock_no_mmap,
	.sendpage =	sock_no_sendpage,
};

static int __rds_create(struct socket *sock, struct sock *sk, int protocol)
{
	unsigned long flags;
	struct rds_sock *rs;

	sock_init_data(sock, sk);
	sock->ops		= &rds_proto_ops;
	sk->sk_protocol		= protocol;

	rs = rds_sk_to_rs(sk);
	spin_lock_init(&rs->rs_lock);
	rwlock_init(&rs->rs_recv_lock);
	INIT_LIST_HEAD(&rs->rs_send_queue);
	INIT_LIST_HEAD(&rs->rs_recv_queue);
	INIT_LIST_HEAD(&rs->rs_notify_queue);
	INIT_LIST_HEAD(&rs->rs_cong_list);
	spin_lock_init(&rs->rs_rdma_lock);
	rs->rs_rdma_keys = RB_ROOT;

	spin_lock_irqsave(&rds_sock_lock, flags);
	list_add_tail(&rs->rs_item, &rds_sock_list);
	rds_sock_count++;
	spin_unlock_irqrestore(&rds_sock_lock, flags);

	return 0;
}

static int rds_create(struct net *net, struct socket *sock, int protocol)
{
	struct sock *sk;

	if (sock->type != SOCK_SEQPACKET || protocol)
		return -ESOCKTNOSUPPORT;

	sk = sk_alloc(net, AF_RDS, GFP_ATOMIC, &rds_proto);
	if (!sk)
		return -ENOMEM;

	return __rds_create(sock, sk, protocol);
}

void rds_sock_addref(struct rds_sock *rs)
{
	sock_hold(rds_rs_to_sk(rs));
}

void rds_sock_put(struct rds_sock *rs)
{
	sock_put(rds_rs_to_sk(rs));
}

static struct net_proto_family rds_family_ops = {
	.family =	AF_RDS,
	.create =	rds_create,
	.owner	=	THIS_MODULE,
};

static void rds_sock_inc_info(struct socket *sock, unsigned int len,
			      struct rds_info_iterator *iter,
			      struct rds_info_lengths *lens)
{
	struct rds_sock *rs;
	struct sock *sk;
	struct rds_incoming *inc;
	unsigned long flags;
	unsigned int total = 0;

	len /= sizeof(struct rds_info_message);

	spin_lock_irqsave(&rds_sock_lock, flags);

	list_for_each_entry(rs, &rds_sock_list, rs_item) {
		sk = rds_rs_to_sk(rs);
		read_lock(&rs->rs_recv_lock);

		/* XXX too lazy to maintain counts.. */
		list_for_each_entry(inc, &rs->rs_recv_queue, i_item) {
			total++;
			if (total <= len)
				rds_inc_info_copy(inc, iter, inc->i_saddr,
						  rs->rs_bound_addr, 1);
		}

		read_unlock(&rs->rs_recv_lock);
	}

	spin_unlock_irqrestore(&rds_sock_lock, flags);

	lens->nr = total;
	lens->each = sizeof(struct rds_info_message);
}

static void rds_sock_info(struct socket *sock, unsigned int len,
			  struct rds_info_iterator *iter,
			  struct rds_info_lengths *lens)
{
	struct rds_info_socket sinfo;
	struct rds_sock *rs;
	unsigned long flags;

	len /= sizeof(struct rds_info_socket);

	spin_lock_irqsave(&rds_sock_lock, flags);

	if (len < rds_sock_count)
		goto out;

	list_for_each_entry(rs, &rds_sock_list, rs_item) {
		sinfo.sndbuf = rds_sk_sndbuf(rs);
		sinfo.rcvbuf = rds_sk_rcvbuf(rs);
		sinfo.bound_addr = rs->rs_bound_addr;
		sinfo.connected_addr = rs->rs_conn_addr;
		sinfo.bound_port = rs->rs_bound_port;
		sinfo.connected_port = rs->rs_conn_port;
		sinfo.inum = sock_i_ino(rds_rs_to_sk(rs));

		rds_info_copy(iter, &sinfo, sizeof(sinfo));
	}

out:
	lens->nr = rds_sock_count;
	lens->each = sizeof(struct rds_info_socket);

	spin_unlock_irqrestore(&rds_sock_lock, flags);
}

static void __exit rds_exit(void)
{
	sock_unregister(rds_family_ops.family);
	proto_unregister(&rds_proto);
	rds_conn_exit();
	rds_cong_exit();
	rds_sysctl_exit();
	rds_threads_exit();
	rds_stats_exit();
	rds_page_exit();
	rds_info_deregister_func(RDS_INFO_SOCKETS, rds_sock_info);
	rds_info_deregister_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);
}
module_exit(rds_exit);

static int __init rds_init(void)
{
	int ret;

	ret = rds_conn_init();
	if (ret)
		goto out;
	ret = rds_threads_init();
	if (ret)
		goto out_conn;
	ret = rds_sysctl_init();
	if (ret)
		goto out_threads;
	ret = rds_stats_init();
	if (ret)
		goto out_sysctl;
	ret = proto_register(&rds_proto, 1);
	if (ret)
		goto out_stats;
	ret = sock_register(&rds_family_ops);
	if (ret)
		goto out_proto;

	rds_info_register_func(RDS_INFO_SOCKETS, rds_sock_info);
	rds_info_register_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);

	goto out;

out_proto:
	proto_unregister(&rds_proto);
out_stats:
	rds_stats_exit();
out_sysctl:
	rds_sysctl_exit();
out_threads:
	rds_threads_exit();
out_conn:
	rds_conn_exit();
	rds_cong_exit();
	rds_page_exit();
out:
	return ret;
}
module_init(rds_init);

#define DRV_VERSION     "4.0"
#define DRV_RELDATE     "Feb 12, 2009"

MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
MODULE_DESCRIPTION("RDS: Reliable Datagram Sockets"
		   " v" DRV_VERSION " (" DRV_RELDATE ")");
MODULE_VERSION(DRV_VERSION);
MODULE_LICENSE("Dual BSD/GPL");
MODULE_ALIAS_NETPROTO(PF_RDS);
Commit	Line	Data
639b321b 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/module.h>
	34	#include <linux/errno.h>
	35	#include <linux/kernel.h>
	36	#include <linux/in.h>
	37	#include <linux/poll.h>
639b321b AG	38	#include <net/sock.h>
	39
	40	#include "rds.h"
	41	#include "rdma.h"
639b321b AG	42
	43	/* this is just used for stats gathering :/ */
	44	static DEFINE_SPINLOCK(rds_sock_lock);
	45	static unsigned long rds_sock_count;
	46	static LIST_HEAD(rds_sock_list);
	47	DECLARE_WAIT_QUEUE_HEAD(rds_poll_waitq);
	48
	49	/*
	50	* This is called as the final descriptor referencing this socket is closed.
	51	* We have to unbind the socket so that another socket can be bound to the
	52	* address it was using.
	53	*
	54	* We have to be careful about racing with the incoming path. sock_orphan()
	55	* sets SOCK_DEAD and we use that as an indicator to the rx path that new
	56	* messages shouldn't be queued.
	57	*/
	58	static int rds_release(struct socket *sock)
	59	{
	60	struct sock *sk = sock->sk;
	61	struct rds_sock *rs;
	62	unsigned long flags;
	63
	64	if (sk == NULL)
	65	goto out;
	66
	67	rs = rds_sk_to_rs(sk);
	68
	69	sock_orphan(sk);
	70	/* Note - rds_clear_recv_queue grabs rs_recv_lock, so
	71	* that ensures the recv path has completed messing
	72	* with the socket. */
	73	rds_clear_recv_queue(rs);
	74	rds_cong_remove_socket(rs);
	75	rds_remove_bound(rs);
	76	rds_send_drop_to(rs, NULL);
	77	rds_rdma_drop_keys(rs);
	78	rds_notify_queue_get(rs, NULL);
	79
	80	spin_lock_irqsave(&rds_sock_lock, flags);
	81	list_del_init(&rs->rs_item);
	82	rds_sock_count--;
	83	spin_unlock_irqrestore(&rds_sock_lock, flags);
	84
	85	sock->sk = NULL;
	86	sock_put(sk);
	87	out:
	88	return 0;
	89	}
	90
	91	/*
	92	* Careful not to race with rds_release -> sock_orphan which clears sk_sleep.
	93	* _bh() isn't OK here, we're called from interrupt handlers. It's probably OK
	94	* to wake the waitqueue after sk_sleep is clear as we hold a sock ref, but
	95	* this seems more conservative.
	96	* NB - normally, one would use sk_callback_lock for this, but we can
	97	* get here from interrupts, whereas the network code grabs sk_callback_lock
	98	* with _lock_bh only - so relying on sk_callback_lock introduces livelocks.
	99	*/
	100	void rds_wake_sk_sleep(struct rds_sock *rs)
	101	{
	102	unsigned long flags;
	103
	104	read_lock_irqsave(&rs->rs_recv_lock, flags);
	105	__rds_wake_sk_sleep(rds_rs_to_sk(rs));
106	read_unlock_irqrestore(&rs->rs_recv_lock, flags);
107	}
108
109	static int rds_getname(struct socket sock, struct sockaddr uaddr,
110	int *uaddr_len, int peer)
111	{
112	struct sockaddr_in sin = (struct sockaddr_in )uaddr;
113	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
114
115	memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
116
117	/* racey, don't care */
118	if (peer) {
119	if (!rs->rs_conn_addr)
120	return -ENOTCONN;
121
122	sin->sin_port = rs->rs_conn_port;
123	sin->sin_addr.s_addr = rs->rs_conn_addr;
124	} else {
125	sin->sin_port = rs->rs_bound_port;
126	sin->sin_addr.s_addr = rs->rs_bound_addr;
127	}
128
129	sin->sin_family = AF_INET;
130
131	uaddr_len = sizeof(sin);
132	return 0;
133	}
134
135	/*
136	* RDS' poll is without a doubt the least intuitive part of the interface,
137	* as POLLIN and POLLOUT do not behave entirely as you would expect from
138	* a network protocol.
139	*
140	* POLLIN is asserted if
141	* - there is data on the receive queue.
142	* - to signal that a previously congested destination may have become
143	* uncongested
144	* - A notification has been queued to the socket (this can be a congestion
145	* update, or a RDMA completion).
146	*
147	* POLLOUT is asserted if there is room on the send queue. This does not mean
148	* however, that the next sendmsg() call will succeed. If the application tries
149	* to send to a congested destination, the system call may still fail (and
150	* return ENOBUFS).
151	*/
152	static unsigned int rds_poll(struct file file, struct socket sock,
153	poll_table *wait)
154	{
155	struct sock *sk = sock->sk;
156	struct rds_sock *rs = rds_sk_to_rs(sk);
157	unsigned int mask = 0;
158	unsigned long flags;
159
160	poll_wait(file, sk->sk_sleep, wait);
161
162	poll_wait(file, &rds_poll_waitq, wait);
163
164	read_lock_irqsave(&rs->rs_recv_lock, flags);
165	if (!rs->rs_cong_monitor) {
166	/* When a congestion map was updated, we signal POLLIN for
167	* "historical" reasons. Applications can also poll for
168	* WRBAND instead. */
169	if (rds_cong_updated_since(&rs->rs_cong_track))
170	mask \|= (POLLIN \| POLLRDNORM \| POLLWRBAND);
171	} else {
172	spin_lock(&rs->rs_lock);
173	if (rs->rs_cong_notify)
174	mask \|= (POLLIN \| POLLRDNORM);
175	spin_unlock(&rs->rs_lock);
176	}
177	if (!list_empty(&rs->rs_recv_queue)
178	\|\| !list_empty(&rs->rs_notify_queue))
179	mask \|= (POLLIN \| POLLRDNORM);
180	if (rs->rs_snd_bytes < rds_sk_sndbuf(rs))
181	mask \|= (POLLOUT \| POLLWRNORM);
182	read_unlock_irqrestore(&rs->rs_recv_lock, flags);
183
184	return mask;
185	}
186
187	static int rds_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
188	{
189	return -ENOIOCTLCMD;
190	}
191
192	static int rds_cancel_sent_to(struct rds_sock rs, char __user optval,
193	int len)
194	{
195	struct sockaddr_in sin;
196	int ret = 0;
197
198	/* racing with another thread binding seems ok here */
199	if (rs->rs_bound_addr == 0) {
200	ret = -ENOTCONN; /* XXX not a great errno */
201	goto out;
202	}
203
204	if (len < sizeof(struct sockaddr_in)) {
205	ret = -EINVAL;
206	goto out;
207	}
208
209	if (copy_from_user(&sin, optval, sizeof(sin))) {
210	ret = -EFAULT;
211	goto out;
212	}
213
214	rds_send_drop_to(rs, &sin);
215	out:
216	return ret;
217	}
218
219	static int rds_set_bool_option(unsigned char optvar, char __user optval,
220	int optlen)
221	{
222	int value;
223
224	if (optlen < sizeof(int))
225	return -EINVAL;
226	if (get_user(value, (int __user *) optval))
227	return -EFAULT;
228	*optvar = !!value;
229	return 0;
230	}
231
232	static int rds_cong_monitor(struct rds_sock rs, char __user optval,
233	int optlen)
234	{
235	int ret;
236
237	ret = rds_set_bool_option(&rs->rs_cong_monitor, optval, optlen);
238	if (ret == 0) {
239	if (rs->rs_cong_monitor) {
240	rds_cong_add_socket(rs);
241	} else {
242	rds_cong_remove_socket(rs);
243	rs->rs_cong_mask = 0;
244	rs->rs_cong_notify = 0;
245	}
246	}
247	return ret;
248	}
249
250	static int rds_setsockopt(struct socket *sock, int level, int optname,
251	char __user *optval, int optlen)
252	{
253	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
254	int ret;
255
256	if (level != SOL_RDS) {
257	ret = -ENOPROTOOPT;
258	goto out;
259	}
260
261	switch (optname) {
262	case RDS_CANCEL_SENT_TO:
263	ret = rds_cancel_sent_to(rs, optval, optlen);
264	break;
265	case RDS_GET_MR:
266	ret = rds_get_mr(rs, optval, optlen);
267	break;
268	case RDS_FREE_MR:
269	ret = rds_free_mr(rs, optval, optlen);
270	break;
271	case RDS_RECVERR:
272	ret = rds_set_bool_option(&rs->rs_recverr, optval, optlen);
273	break;
274	case RDS_CONG_MONITOR:
275	ret = rds_cong_monitor(rs, optval, optlen);
276	break;
277	default:
278	ret = -ENOPROTOOPT;
279	}
280	out:
281	return ret;
282	}
283
284	static int rds_getsockopt(struct socket *sock, int level, int optname,
285	char __user optval, int __user optlen)
286	{
287	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
288	int ret = -ENOPROTOOPT, len;
289
290	if (level != SOL_RDS)
291	goto out;
292
293	if (get_user(len, optlen)) {
294	ret = -EFAULT;
295	goto out;
296	}
297
298	switch (optname) {
299	case RDS_INFO_FIRST ... RDS_INFO_LAST:
300	ret = rds_info_getsockopt(sock, optname, optval,
301	optlen);
302	break;
303
304	case RDS_RECVERR:
305	if (len < sizeof(int))
306	ret = -EINVAL;
307	else
308	if (put_user(rs->rs_recverr, (int __user *) optval)
309	\|\| put_user(sizeof(int), optlen))
310	ret = -EFAULT;
311	else
312	ret = 0;
313	break;
314	default:
315	break;
316	}
317
318	out:
319	return ret;
320
321	}
322
323	static int rds_connect(struct socket sock, struct sockaddr uaddr,
324	int addr_len, int flags)
325	{
326	struct sock *sk = sock->sk;
327	struct sockaddr_in sin = (struct sockaddr_in )uaddr;
328	struct rds_sock *rs = rds_sk_to_rs(sk);
329	int ret = 0;
330
331	lock_sock(sk);
332
333	if (addr_len != sizeof(struct sockaddr_in)) {
334	ret = -EINVAL;
335	goto out;
336	}
337
338	if (sin->sin_family != AF_INET) {
339	ret = -EAFNOSUPPORT;
340	goto out;
341	}
342
343	if (sin->sin_addr.s_addr == htonl(INADDR_ANY)) {
344	ret = -EDESTADDRREQ;
345	goto out;
346	}
347
348	rs->rs_conn_addr = sin->sin_addr.s_addr;
349	rs->rs_conn_port = sin->sin_port;
350
351	out:
352	release_sock(sk);
353	return ret;
354	}
355
356	static struct proto rds_proto = {
357	.name = "RDS",
358	.owner = THIS_MODULE,
359	.obj_size = sizeof(struct rds_sock),
360	};
361
362	static struct proto_ops rds_proto_ops = {
363	.family = AF_RDS,
364	.owner = THIS_MODULE,
365	.release = rds_release,
366	.bind = rds_bind,
367	.connect = rds_connect,
368	.socketpair = sock_no_socketpair,
369	.accept = sock_no_accept,
370	.getname = rds_getname,
371	.poll = rds_poll,
372	.ioctl = rds_ioctl,
373	.listen = sock_no_listen,
374	.shutdown = sock_no_shutdown,
375	.setsockopt = rds_setsockopt,
376	.getsockopt = rds_getsockopt,
377	.sendmsg = rds_sendmsg,
378	.recvmsg = rds_recvmsg,
379	.mmap = sock_no_mmap,
380	.sendpage = sock_no_sendpage,
381	};
382
383	static int __rds_create(struct socket sock, struct sock sk, int protocol)
384	{
385	unsigned long flags;
386	struct rds_sock *rs;
387
388	sock_init_data(sock, sk);
389	sock->ops = &rds_proto_ops;
390	sk->sk_protocol = protocol;
391
392	rs = rds_sk_to_rs(sk);
393	spin_lock_init(&rs->rs_lock);
394	rwlock_init(&rs->rs_recv_lock);
395	INIT_LIST_HEAD(&rs->rs_send_queue);
396	INIT_LIST_HEAD(&rs->rs_recv_queue);
397	INIT_LIST_HEAD(&rs->rs_notify_queue);
398	INIT_LIST_HEAD(&rs->rs_cong_list);
399	spin_lock_init(&rs->rs_rdma_lock);
400	rs->rs_rdma_keys = RB_ROOT;
401
402	spin_lock_irqsave(&rds_sock_lock, flags);
403	list_add_tail(&rs->rs_item, &rds_sock_list);
404	rds_sock_count++;
405	spin_unlock_irqrestore(&rds_sock_lock, flags);
406
407	return 0;
408	}
409
410	static int rds_create(struct net net, struct socket sock, int protocol)
411	{
412	struct sock *sk;
413
414	if (sock->type != SOCK_SEQPACKET \|\| protocol)
415	return -ESOCKTNOSUPPORT;
416
417	sk = sk_alloc(net, AF_RDS, GFP_ATOMIC, &rds_proto);
418	if (!sk)
419	return -ENOMEM;
420
421	return __rds_create(sock, sk, protocol);
422	}
423
424	void rds_sock_addref(struct rds_sock *rs)
425	{
426	sock_hold(rds_rs_to_sk(rs));
427	}
428
429	void rds_sock_put(struct rds_sock *rs)
430	{
431	sock_put(rds_rs_to_sk(rs));
432	}
433
434	static struct net_proto_family rds_family_ops = {
435	.family = AF_RDS,
436	.create = rds_create,
437	.owner = THIS_MODULE,
438	};
439
440	static void rds_sock_inc_info(struct socket *sock, unsigned int len,
441	struct rds_info_iterator *iter,
442	struct rds_info_lengths *lens)
443	{
444	struct rds_sock *rs;
445	struct sock *sk;
446	struct rds_incoming *inc;
447	unsigned long flags;
448	unsigned int total = 0;
449
450	len /= sizeof(struct rds_info_message);
451
452	spin_lock_irqsave(&rds_sock_lock, flags);
453
454	list_for_each_entry(rs, &rds_sock_list, rs_item) {
455	sk = rds_rs_to_sk(rs);
456	read_lock(&rs->rs_recv_lock);
457
458	/* XXX too lazy to maintain counts.. */
459	list_for_each_entry(inc, &rs->rs_recv_queue, i_item) {
460	total++;
461	if (total <= len)
462	rds_inc_info_copy(inc, iter, inc->i_saddr,
463	rs->rs_bound_addr, 1);
464	}
465
466	read_unlock(&rs->rs_recv_lock);
467	}
468
469	spin_unlock_irqrestore(&rds_sock_lock, flags);
470
471	lens->nr = total;
472	lens->each = sizeof(struct rds_info_message);
473	}
474
475	static void rds_sock_info(struct socket *sock, unsigned int len,
476	struct rds_info_iterator *iter,
477	struct rds_info_lengths *lens)
478	{
479	struct rds_info_socket sinfo;
480	struct rds_sock *rs;
481	unsigned long flags;
482
483	len /= sizeof(struct rds_info_socket);
484
485	spin_lock_irqsave(&rds_sock_lock, flags);
486
487	if (len < rds_sock_count)
488	goto out;
489
490	list_for_each_entry(rs, &rds_sock_list, rs_item) {
491	sinfo.sndbuf = rds_sk_sndbuf(rs);
492	sinfo.rcvbuf = rds_sk_rcvbuf(rs);
493	sinfo.bound_addr = rs->rs_bound_addr;
494	sinfo.connected_addr = rs->rs_conn_addr;
495	sinfo.bound_port = rs->rs_bound_port;
496	sinfo.connected_port = rs->rs_conn_port;
497	sinfo.inum = sock_i_ino(rds_rs_to_sk(rs));
498
499	rds_info_copy(iter, &sinfo, sizeof(sinfo));
500	}
501
502	out:
503	lens->nr = rds_sock_count;
504	lens->each = sizeof(struct rds_info_socket);
505
506	spin_unlock_irqrestore(&rds_sock_lock, flags);
507	}
508
509	static void __exit rds_exit(void)
510	{
639b321b AG	511	sock_unregister(rds_family_ops.family);
	512	proto_unregister(&rds_proto);
	513	rds_conn_exit();
	514	rds_cong_exit();
	515	rds_sysctl_exit();
	516	rds_threads_exit();
	517	rds_stats_exit();
	518	rds_page_exit();
	519	rds_info_deregister_func(RDS_INFO_SOCKETS, rds_sock_info);
	520	rds_info_deregister_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);
	521	}
	522	module_exit(rds_exit);
	523
	524	static int __init rds_init(void)
	525	{
	526	int ret;
	527
	528	ret = rds_conn_init();
	529	if (ret)
	530	goto out;
	531	ret = rds_threads_init();
	532	if (ret)
	533	goto out_conn;
	534	ret = rds_sysctl_init();
	535	if (ret)
	536	goto out_threads;
	537	ret = rds_stats_init();
	538	if (ret)
	539	goto out_sysctl;
	540	ret = proto_register(&rds_proto, 1);
	541	if (ret)
	542	goto out_stats;
	543	ret = sock_register(&rds_family_ops);
	544	if (ret)
	545	goto out_proto;
	546
	547	rds_info_register_func(RDS_INFO_SOCKETS, rds_sock_info);
	548	rds_info_register_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);
	549
639b321b AG	550	goto out;
639b321b AG	551
639b321b AG	552	out_proto:
	553	proto_unregister(&rds_proto);
	554	out_stats:
	555	rds_stats_exit();
	556	out_sysctl:
	557	rds_sysctl_exit();
	558	out_threads:
	559	rds_threads_exit();
	560	out_conn:
	561	rds_conn_exit();
	562	rds_cong_exit();
	563	rds_page_exit();
	564	out:
	565	return ret;
	566	}
	567	module_init(rds_init);
	568
	569	#define DRV_VERSION "4.0"
	570	#define DRV_RELDATE "Feb 12, 2009"
	571
	572	MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
	573	MODULE_DESCRIPTION("RDS: Reliable Datagram Sockets"
	574	" v" DRV_VERSION " (" DRV_RELDATE ")");
	575	MODULE_VERSION(DRV_VERSION);
	576	MODULE_LICENSE("Dual BSD/GPL");
	577	MODULE_ALIAS_NETPROTO(PF_RDS);