[mirror_ubuntu-zesty-kernel.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/gfp.h>
#include <linux/in.h>
#include <linux/poll.h>
#include <net/sock.h>

#include "rds.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)
		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);

	/*
	 * the binding lookup hash uses rcu, we need to
	 * make sure we sychronize_rcu before we free our
	 * entry
	 */
	rds_remove_bound(rs);
	synchronize_rcu();

	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_sleep(sk), wait);

	if (rs->rs_seen_congestion)
		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);

	/* clear state any time we wake a seen-congested socket */
	if (mask)
		rs->rs_seen_congestion = 0;

	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, unsigned 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_GET_MR_FOR_DEST:
		ret = rds_get_mr_for_dest(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 const 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,
		      int kern)
{
	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 const 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 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) {
		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>
5a0e3ad6	36	#include <linux/gfp.h>
639b321b AG	37	#include <linux/in.h>
639b321b AG	38	#include <linux/poll.h>
639b321b AG	39	#include <net/sock.h>
	40
	41	#include "rds.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
8690bfa1	64	if (!sk)
639b321b AG	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);
38a4e5e6 CM	75
	76	/*
	77	* the binding lookup hash uses rcu, we need to
	78	* make sure we sychronize_rcu before we free our
	79	* entry
	80	*/
639b321b	81	rds_remove_bound(rs);
38a4e5e6 CM	82	synchronize_rcu();
38a4e5e6 CM	83
639b321b AG	84	rds_send_drop_to(rs, NULL);
	85	rds_rdma_drop_keys(rs);
	86	rds_notify_queue_get(rs, NULL);
	87
	88	spin_lock_irqsave(&rds_sock_lock, flags);
	89	list_del_init(&rs->rs_item);
	90	rds_sock_count--;
	91	spin_unlock_irqrestore(&rds_sock_lock, flags);
	92
	93	sock->sk = NULL;
	94	sock_put(sk);
	95	out:
	96	return 0;
	97	}
	98
	99	/*
	100	* Careful not to race with rds_release -> sock_orphan which clears sk_sleep.
	101	* _bh() isn't OK here, we're called from interrupt handlers. It's probably OK
	102	* to wake the waitqueue after sk_sleep is clear as we hold a sock ref, but
	103	* this seems more conservative.
	104	* NB - normally, one would use sk_callback_lock for this, but we can
	105	* get here from interrupts, whereas the network code grabs sk_callback_lock
	106	* with _lock_bh only - so relying on sk_callback_lock introduces livelocks.
	107	*/
	108	void rds_wake_sk_sleep(struct rds_sock *rs)
	109	{
	110	unsigned long flags;
	111
	112	read_lock_irqsave(&rs->rs_recv_lock, flags);
	113	__rds_wake_sk_sleep(rds_rs_to_sk(rs));
	114	read_unlock_irqrestore(&rs->rs_recv_lock, flags);
	115	}
	116
	117	static int rds_getname(struct socket sock, struct sockaddr uaddr,
	118	int *uaddr_len, int peer)
	119	{
	120	struct sockaddr_in sin = (struct sockaddr_in )uaddr;
	121	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	122
	123	memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
	124
	125	/* racey, don't care */
	126	if (peer) {
	127	if (!rs->rs_conn_addr)
	128	return -ENOTCONN;
	129
	130	sin->sin_port = rs->rs_conn_port;
	131	sin->sin_addr.s_addr = rs->rs_conn_addr;
	132	} else {
	133	sin->sin_port = rs->rs_bound_port;
	134	sin->sin_addr.s_addr = rs->rs_bound_addr;
	135	}
	136
	137	sin->sin_family = AF_INET;
	138
	139	uaddr_len = sizeof(sin);
	140	return 0;
	141	}
	142
	143	/*
	144	* RDS' poll is without a doubt the least intuitive part of the interface,
	145	* as POLLIN and POLLOUT do not behave entirely as you would expect from
	146	* a network protocol.
	147	*
148	* POLLIN is asserted if
149	* - there is data on the receive queue.
150	* - to signal that a previously congested destination may have become
151	* uncongested
152	* - A notification has been queued to the socket (this can be a congestion
153	* update, or a RDMA completion).
154	*
155	* POLLOUT is asserted if there is room on the send queue. This does not mean
156	* however, that the next sendmsg() call will succeed. If the application tries
157	* to send to a congested destination, the system call may still fail (and
158	* return ENOBUFS).
159	*/
160	static unsigned int rds_poll(struct file file, struct socket sock,
161	poll_table *wait)
162	{
163	struct sock *sk = sock->sk;
164	struct rds_sock *rs = rds_sk_to_rs(sk);
165	unsigned int mask = 0;
166	unsigned long flags;
167
aa395145	168	poll_wait(file, sk_sleep(sk), wait);
639b321b	169
b98ba52f AG	170	if (rs->rs_seen_congestion)
b98ba52f AG	171	poll_wait(file, &rds_poll_waitq, wait);
639b321b AG	172
	173	read_lock_irqsave(&rs->rs_recv_lock, flags);
	174	if (!rs->rs_cong_monitor) {
	175	/* When a congestion map was updated, we signal POLLIN for
	176	* "historical" reasons. Applications can also poll for
	177	* WRBAND instead. */
	178	if (rds_cong_updated_since(&rs->rs_cong_track))
	179	mask \|= (POLLIN \| POLLRDNORM \| POLLWRBAND);
	180	} else {
	181	spin_lock(&rs->rs_lock);
	182	if (rs->rs_cong_notify)
	183	mask \|= (POLLIN \| POLLRDNORM);
	184	spin_unlock(&rs->rs_lock);
	185	}
f64f9e71 JP	186	if (!list_empty(&rs->rs_recv_queue) \|\|
f64f9e71 JP	187	!list_empty(&rs->rs_notify_queue))
639b321b AG	188	mask \|= (POLLIN \| POLLRDNORM);
	189	if (rs->rs_snd_bytes < rds_sk_sndbuf(rs))
	190	mask \|= (POLLOUT \| POLLWRNORM);
	191	read_unlock_irqrestore(&rs->rs_recv_lock, flags);
	192
b98ba52f AG	193	/* clear state any time we wake a seen-congested socket */
	194	if (mask)
	195	rs->rs_seen_congestion = 0;
	196
639b321b AG	197	return mask;
	198	}
	199
	200	static int rds_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
	201	{
	202	return -ENOIOCTLCMD;
	203	}
	204
	205	static int rds_cancel_sent_to(struct rds_sock rs, char __user optval,
	206	int len)
	207	{
	208	struct sockaddr_in sin;
	209	int ret = 0;
	210
	211	/* racing with another thread binding seems ok here */
	212	if (rs->rs_bound_addr == 0) {
	213	ret = -ENOTCONN; /* XXX not a great errno */
	214	goto out;
	215	}
	216
	217	if (len < sizeof(struct sockaddr_in)) {
	218	ret = -EINVAL;
	219	goto out;
	220	}
	221
	222	if (copy_from_user(&sin, optval, sizeof(sin))) {
	223	ret = -EFAULT;
	224	goto out;
	225	}
	226
	227	rds_send_drop_to(rs, &sin);
	228	out:
	229	return ret;
	230	}
	231
	232	static int rds_set_bool_option(unsigned char optvar, char __user optval,
	233	int optlen)
	234	{
	235	int value;
	236
	237	if (optlen < sizeof(int))
	238	return -EINVAL;
	239	if (get_user(value, (int __user *) optval))
	240	return -EFAULT;
	241	*optvar = !!value;
	242	return 0;
	243	}
	244
	245	static int rds_cong_monitor(struct rds_sock rs, char __user optval,
	246	int optlen)
	247	{
	248	int ret;
	249
	250	ret = rds_set_bool_option(&rs->rs_cong_monitor, optval, optlen);
	251	if (ret == 0) {
	252	if (rs->rs_cong_monitor) {
	253	rds_cong_add_socket(rs);
	254	} else {
	255	rds_cong_remove_socket(rs);
	256	rs->rs_cong_mask = 0;
	257	rs->rs_cong_notify = 0;
	258	}
	259	}
	260	return ret;
261	}
262
263	static int rds_setsockopt(struct socket *sock, int level, int optname,
b7058842	264	char __user *optval, unsigned int optlen)
639b321b AG	265	{
	266	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	267	int ret;
	268
	269	if (level != SOL_RDS) {
	270	ret = -ENOPROTOOPT;
	271	goto out;
	272	}
	273
	274	switch (optname) {
	275	case RDS_CANCEL_SENT_TO:
	276	ret = rds_cancel_sent_to(rs, optval, optlen);
	277	break;
	278	case RDS_GET_MR:
	279	ret = rds_get_mr(rs, optval, optlen);
	280	break;
244546f0 AG	281	case RDS_GET_MR_FOR_DEST:
	282	ret = rds_get_mr_for_dest(rs, optval, optlen);
	283	break;
639b321b AG	284	case RDS_FREE_MR:
	285	ret = rds_free_mr(rs, optval, optlen);
	286	break;
	287	case RDS_RECVERR:
	288	ret = rds_set_bool_option(&rs->rs_recverr, optval, optlen);
	289	break;
	290	case RDS_CONG_MONITOR:
	291	ret = rds_cong_monitor(rs, optval, optlen);
	292	break;
	293	default:
	294	ret = -ENOPROTOOPT;
	295	}
	296	out:
	297	return ret;
	298	}
	299
	300	static int rds_getsockopt(struct socket *sock, int level, int optname,
	301	char __user optval, int __user optlen)
	302	{
	303	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	304	int ret = -ENOPROTOOPT, len;
	305
	306	if (level != SOL_RDS)
	307	goto out;
	308
	309	if (get_user(len, optlen)) {
	310	ret = -EFAULT;
	311	goto out;
	312	}
	313
	314	switch (optname) {
	315	case RDS_INFO_FIRST ... RDS_INFO_LAST:
	316	ret = rds_info_getsockopt(sock, optname, optval,
	317	optlen);
	318	break;
	319
	320	case RDS_RECVERR:
	321	if (len < sizeof(int))
	322	ret = -EINVAL;
	323	else
f64f9e71 JP	324	if (put_user(rs->rs_recverr, (int __user *) optval) \|\|
f64f9e71 JP	325	put_user(sizeof(int), optlen))
639b321b AG	326	ret = -EFAULT;
	327	else
	328	ret = 0;
	329	break;
	330	default:
	331	break;
	332	}
	333
	334	out:
	335	return ret;
	336
	337	}
	338
	339	static int rds_connect(struct socket sock, struct sockaddr uaddr,
	340	int addr_len, int flags)
	341	{
	342	struct sock *sk = sock->sk;
	343	struct sockaddr_in sin = (struct sockaddr_in )uaddr;
	344	struct rds_sock *rs = rds_sk_to_rs(sk);
	345	int ret = 0;
	346
	347	lock_sock(sk);
	348
	349	if (addr_len != sizeof(struct sockaddr_in)) {
	350	ret = -EINVAL;
	351	goto out;
	352	}
	353
	354	if (sin->sin_family != AF_INET) {
	355	ret = -EAFNOSUPPORT;
	356	goto out;
	357	}
	358
	359	if (sin->sin_addr.s_addr == htonl(INADDR_ANY)) {
	360	ret = -EDESTADDRREQ;
	361	goto out;
	362	}
	363
	364	rs->rs_conn_addr = sin->sin_addr.s_addr;
	365	rs->rs_conn_port = sin->sin_port;
	366
	367	out:
	368	release_sock(sk);
	369	return ret;
	370	}
	371
	372	static struct proto rds_proto = {
	373	.name = "RDS",
	374	.owner = THIS_MODULE,
	375	.obj_size = sizeof(struct rds_sock),
	376	};
	377
5708e868	378	static const struct proto_ops rds_proto_ops = {
639b321b AG	379	.family = AF_RDS,
	380	.owner = THIS_MODULE,
	381	.release = rds_release,
	382	.bind = rds_bind,
	383	.connect = rds_connect,
	384	.socketpair = sock_no_socketpair,
	385	.accept = sock_no_accept,
	386	.getname = rds_getname,
	387	.poll = rds_poll,
	388	.ioctl = rds_ioctl,
	389	.listen = sock_no_listen,
	390	.shutdown = sock_no_shutdown,
	391	.setsockopt = rds_setsockopt,
	392	.getsockopt = rds_getsockopt,
	393	.sendmsg = rds_sendmsg,
	394	.recvmsg = rds_recvmsg,
	395	.mmap = sock_no_mmap,
	396	.sendpage = sock_no_sendpage,
	397	};
	398
	399	static int __rds_create(struct socket sock, struct sock sk, int protocol)
	400	{
	401	unsigned long flags;
	402	struct rds_sock *rs;
	403
	404	sock_init_data(sock, sk);
	405	sock->ops = &rds_proto_ops;
	406	sk->sk_protocol = protocol;
	407
	408	rs = rds_sk_to_rs(sk);
	409	spin_lock_init(&rs->rs_lock);
	410	rwlock_init(&rs->rs_recv_lock);
	411	INIT_LIST_HEAD(&rs->rs_send_queue);
	412	INIT_LIST_HEAD(&rs->rs_recv_queue);
	413	INIT_LIST_HEAD(&rs->rs_notify_queue);
	414	INIT_LIST_HEAD(&rs->rs_cong_list);
	415	spin_lock_init(&rs->rs_rdma_lock);
	416	rs->rs_rdma_keys = RB_ROOT;
	417
	418	spin_lock_irqsave(&rds_sock_lock, flags);
	419	list_add_tail(&rs->rs_item, &rds_sock_list);
	420	rds_sock_count++;
	421	spin_unlock_irqrestore(&rds_sock_lock, flags);
	422
	423	return 0;
	424	}
	425
3f378b68 EP	426	static int rds_create(struct net net, struct socket sock, int protocol,
3f378b68 EP	427	int kern)
639b321b AG	428	{
	429	struct sock *sk;
	430
	431	if (sock->type != SOCK_SEQPACKET \|\| protocol)
	432	return -ESOCKTNOSUPPORT;
	433
	434	sk = sk_alloc(net, AF_RDS, GFP_ATOMIC, &rds_proto);
	435	if (!sk)
	436	return -ENOMEM;
	437
	438	return __rds_create(sock, sk, protocol);
	439	}
	440
	441	void rds_sock_addref(struct rds_sock *rs)
	442	{
	443	sock_hold(rds_rs_to_sk(rs));
	444	}
	445
	446	void rds_sock_put(struct rds_sock *rs)
	447	{
	448	sock_put(rds_rs_to_sk(rs));
	449	}
	450
ec1b4cf7	451	static const struct net_proto_family rds_family_ops = {
639b321b AG	452	.family = AF_RDS,
	453	.create = rds_create,
	454	.owner = THIS_MODULE,
	455	};
	456
	457	static void rds_sock_inc_info(struct socket *sock, unsigned int len,
	458	struct rds_info_iterator *iter,
	459	struct rds_info_lengths *lens)
	460	{
	461	struct rds_sock *rs;
639b321b AG	462	struct rds_incoming *inc;
	463	unsigned long flags;
	464	unsigned int total = 0;
	465
	466	len /= sizeof(struct rds_info_message);
	467
	468	spin_lock_irqsave(&rds_sock_lock, flags);
	469
	470	list_for_each_entry(rs, &rds_sock_list, rs_item) {
639b321b AG	471	read_lock(&rs->rs_recv_lock);
	472
	473	/* XXX too lazy to maintain counts.. */
	474	list_for_each_entry(inc, &rs->rs_recv_queue, i_item) {
	475	total++;
	476	if (total <= len)
	477	rds_inc_info_copy(inc, iter, inc->i_saddr,
	478	rs->rs_bound_addr, 1);
	479	}
	480
	481	read_unlock(&rs->rs_recv_lock);
	482	}
	483
	484	spin_unlock_irqrestore(&rds_sock_lock, flags);
	485
	486	lens->nr = total;
	487	lens->each = sizeof(struct rds_info_message);
	488	}
	489
	490	static void rds_sock_info(struct socket *sock, unsigned int len,
	491	struct rds_info_iterator *iter,
	492	struct rds_info_lengths *lens)
	493	{
	494	struct rds_info_socket sinfo;
	495	struct rds_sock *rs;
	496	unsigned long flags;
	497
	498	len /= sizeof(struct rds_info_socket);
	499
	500	spin_lock_irqsave(&rds_sock_lock, flags);
	501
	502	if (len < rds_sock_count)
	503	goto out;
	504
	505	list_for_each_entry(rs, &rds_sock_list, rs_item) {
	506	sinfo.sndbuf = rds_sk_sndbuf(rs);
	507	sinfo.rcvbuf = rds_sk_rcvbuf(rs);
	508	sinfo.bound_addr = rs->rs_bound_addr;
	509	sinfo.connected_addr = rs->rs_conn_addr;
	510	sinfo.bound_port = rs->rs_bound_port;
	511	sinfo.connected_port = rs->rs_conn_port;
	512	sinfo.inum = sock_i_ino(rds_rs_to_sk(rs));
	513
	514	rds_info_copy(iter, &sinfo, sizeof(sinfo));
	515	}
	516
	517	out:
	518	lens->nr = rds_sock_count;
	519	lens->each = sizeof(struct rds_info_socket);
	520
	521	spin_unlock_irqrestore(&rds_sock_lock, flags);
	522	}
	523
	524	static void __exit rds_exit(void)
	525	{
639b321b AG	526	sock_unregister(rds_family_ops.family);
	527	proto_unregister(&rds_proto);
	528	rds_conn_exit();
	529	rds_cong_exit();
	530	rds_sysctl_exit();
	531	rds_threads_exit();
	532	rds_stats_exit();
	533	rds_page_exit();
	534	rds_info_deregister_func(RDS_INFO_SOCKETS, rds_sock_info);
	535	rds_info_deregister_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);
	536	}
	537	module_exit(rds_exit);
	538
	539	static int __init rds_init(void)
	540	{
	541	int ret;
	542
	543	ret = rds_conn_init();
	544	if (ret)
	545	goto out;
	546	ret = rds_threads_init();
	547	if (ret)
	548	goto out_conn;
	549	ret = rds_sysctl_init();
	550	if (ret)
	551	goto out_threads;
	552	ret = rds_stats_init();
	553	if (ret)
	554	goto out_sysctl;
	555	ret = proto_register(&rds_proto, 1);
	556	if (ret)
	557	goto out_stats;
	558	ret = sock_register(&rds_family_ops);
	559	if (ret)
	560	goto out_proto;
	561
	562	rds_info_register_func(RDS_INFO_SOCKETS, rds_sock_info);
	563	rds_info_register_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);
	564
639b321b AG	565	goto out;
639b321b AG	566
639b321b AG	567	out_proto:
	568	proto_unregister(&rds_proto);
	569	out_stats:
	570	rds_stats_exit();
	571	out_sysctl:
	572	rds_sysctl_exit();
	573	out_threads:
	574	rds_threads_exit();
	575	out_conn:
	576	rds_conn_exit();
	577	rds_cong_exit();
	578	rds_page_exit();
	579	out:
	580	return ret;
	581	}
	582	module_init(rds_init);
	583
	584	#define DRV_VERSION "4.0"
	585	#define DRV_RELDATE "Feb 12, 2009"
	586
	587	MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
	588	MODULE_DESCRIPTION("RDS: Reliable Datagram Sockets"
	589	" v" DRV_VERSION " (" DRV_RELDATE ")");
	590	MODULE_VERSION(DRV_VERSION);
	591	MODULE_LICENSE("Dual BSD/GPL");
	592	MODULE_ALIAS_NETPROTO(PF_RDS);