[deliverable/linux.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);

	rds_trans_put(rs->rs_transport);

	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 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 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
5adb5bc6 ZB	93	rds_trans_put(rs->rs_transport);
5adb5bc6 ZB	94
639b321b AG	95	sock->sk = NULL;
	96	sock_put(sk);
	97	out:
	98	return 0;
	99	}
	100
	101	/*
	102	* Careful not to race with rds_release -> sock_orphan which clears sk_sleep.
	103	* _bh() isn't OK here, we're called from interrupt handlers. It's probably OK
	104	* to wake the waitqueue after sk_sleep is clear as we hold a sock ref, but
	105	* this seems more conservative.
	106	* NB - normally, one would use sk_callback_lock for this, but we can
	107	* get here from interrupts, whereas the network code grabs sk_callback_lock
	108	* with _lock_bh only - so relying on sk_callback_lock introduces livelocks.
	109	*/
	110	void rds_wake_sk_sleep(struct rds_sock *rs)
	111	{
	112	unsigned long flags;
	113
	114	read_lock_irqsave(&rs->rs_recv_lock, flags);
	115	__rds_wake_sk_sleep(rds_rs_to_sk(rs));
	116	read_unlock_irqrestore(&rs->rs_recv_lock, flags);
	117	}
	118
	119	static int rds_getname(struct socket sock, struct sockaddr uaddr,
	120	int *uaddr_len, int peer)
	121	{
	122	struct sockaddr_in sin = (struct sockaddr_in )uaddr;
	123	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	124
	125	memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
	126
	127	/* racey, don't care */
	128	if (peer) {
	129	if (!rs->rs_conn_addr)
	130	return -ENOTCONN;
	131
	132	sin->sin_port = rs->rs_conn_port;
	133	sin->sin_addr.s_addr = rs->rs_conn_addr;
	134	} else {
	135	sin->sin_port = rs->rs_bound_port;
	136	sin->sin_addr.s_addr = rs->rs_bound_addr;
	137	}
	138
	139	sin->sin_family = AF_INET;
	140
	141	uaddr_len = sizeof(sin);
	142	return 0;
	143	}
	144
	145	/*
	146	* RDS' poll is without a doubt the least intuitive part of the interface,
	147	* as POLLIN and POLLOUT do not behave entirely as you would expect from
	148	* a network protocol.
	149	*
	150	* POLLIN is asserted if
	151	* - there is data on the receive queue.
	152	* - to signal that a previously congested destination may have become
	153	* uncongested
	154	* - A notification has been queued to the socket (this can be a congestion
	155	* update, or a RDMA completion).
	156	*
	157	* POLLOUT is asserted if there is room on the send queue. This does not mean
	158	* however, that the next sendmsg() call will succeed. If the application tries
159	* to send to a congested destination, the system call may still fail (and
160	* return ENOBUFS).
161	*/
162	static unsigned int rds_poll(struct file file, struct socket sock,
163	poll_table *wait)
164	{
165	struct sock *sk = sock->sk;
166	struct rds_sock *rs = rds_sk_to_rs(sk);
167	unsigned int mask = 0;
168	unsigned long flags;
169
aa395145	170	poll_wait(file, sk_sleep(sk), wait);
639b321b	171
b98ba52f AG	172	if (rs->rs_seen_congestion)
b98ba52f AG	173	poll_wait(file, &rds_poll_waitq, wait);
639b321b AG	174
	175	read_lock_irqsave(&rs->rs_recv_lock, flags);
	176	if (!rs->rs_cong_monitor) {
	177	/* When a congestion map was updated, we signal POLLIN for
	178	* "historical" reasons. Applications can also poll for
	179	* WRBAND instead. */
	180	if (rds_cong_updated_since(&rs->rs_cong_track))
	181	mask \|= (POLLIN \| POLLRDNORM \| POLLWRBAND);
	182	} else {
	183	spin_lock(&rs->rs_lock);
	184	if (rs->rs_cong_notify)
	185	mask \|= (POLLIN \| POLLRDNORM);
	186	spin_unlock(&rs->rs_lock);
	187	}
f64f9e71 JP	188	if (!list_empty(&rs->rs_recv_queue) \|\|
f64f9e71 JP	189	!list_empty(&rs->rs_notify_queue))
639b321b AG	190	mask \|= (POLLIN \| POLLRDNORM);
	191	if (rs->rs_snd_bytes < rds_sk_sndbuf(rs))
	192	mask \|= (POLLOUT \| POLLWRNORM);
	193	read_unlock_irqrestore(&rs->rs_recv_lock, flags);
	194
b98ba52f AG	195	/* clear state any time we wake a seen-congested socket */
	196	if (mask)
	197	rs->rs_seen_congestion = 0;
	198
639b321b AG	199	return mask;
	200	}
	201
	202	static int rds_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
	203	{
	204	return -ENOIOCTLCMD;
	205	}
	206
	207	static int rds_cancel_sent_to(struct rds_sock rs, char __user optval,
	208	int len)
	209	{
	210	struct sockaddr_in sin;
	211	int ret = 0;
	212
	213	/* racing with another thread binding seems ok here */
	214	if (rs->rs_bound_addr == 0) {
	215	ret = -ENOTCONN; /* XXX not a great errno */
	216	goto out;
	217	}
	218
	219	if (len < sizeof(struct sockaddr_in)) {
	220	ret = -EINVAL;
	221	goto out;
	222	}
	223
	224	if (copy_from_user(&sin, optval, sizeof(sin))) {
	225	ret = -EFAULT;
	226	goto out;
	227	}
	228
	229	rds_send_drop_to(rs, &sin);
	230	out:
	231	return ret;
	232	}
	233
	234	static int rds_set_bool_option(unsigned char optvar, char __user optval,
	235	int optlen)
	236	{
	237	int value;
	238
	239	if (optlen < sizeof(int))
	240	return -EINVAL;
	241	if (get_user(value, (int __user *) optval))
	242	return -EFAULT;
	243	*optvar = !!value;
	244	return 0;
	245	}
	246
	247	static int rds_cong_monitor(struct rds_sock rs, char __user optval,
	248	int optlen)
	249	{
	250	int ret;
	251
	252	ret = rds_set_bool_option(&rs->rs_cong_monitor, optval, optlen);
	253	if (ret == 0) {
	254	if (rs->rs_cong_monitor) {
	255	rds_cong_add_socket(rs);
	256	} else {
	257	rds_cong_remove_socket(rs);
	258	rs->rs_cong_mask = 0;
	259	rs->rs_cong_notify = 0;
	260	}
	261	}
	262	return ret;
263	}
264
265	static int rds_setsockopt(struct socket *sock, int level, int optname,
b7058842	266	char __user *optval, unsigned int optlen)
639b321b AG	267	{
	268	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	269	int ret;
	270
	271	if (level != SOL_RDS) {
	272	ret = -ENOPROTOOPT;
	273	goto out;
	274	}
	275
	276	switch (optname) {
	277	case RDS_CANCEL_SENT_TO:
	278	ret = rds_cancel_sent_to(rs, optval, optlen);
	279	break;
	280	case RDS_GET_MR:
	281	ret = rds_get_mr(rs, optval, optlen);
	282	break;
244546f0 AG	283	case RDS_GET_MR_FOR_DEST:
	284	ret = rds_get_mr_for_dest(rs, optval, optlen);
	285	break;
639b321b AG	286	case RDS_FREE_MR:
	287	ret = rds_free_mr(rs, optval, optlen);
	288	break;
	289	case RDS_RECVERR:
	290	ret = rds_set_bool_option(&rs->rs_recverr, optval, optlen);
	291	break;
	292	case RDS_CONG_MONITOR:
	293	ret = rds_cong_monitor(rs, optval, optlen);
	294	break;
	295	default:
	296	ret = -ENOPROTOOPT;
	297	}
	298	out:
	299	return ret;
	300	}
	301
	302	static int rds_getsockopt(struct socket *sock, int level, int optname,
	303	char __user optval, int __user optlen)
	304	{
	305	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
	306	int ret = -ENOPROTOOPT, len;
	307
	308	if (level != SOL_RDS)
	309	goto out;
	310
	311	if (get_user(len, optlen)) {
	312	ret = -EFAULT;
	313	goto out;
	314	}
	315
	316	switch (optname) {
	317	case RDS_INFO_FIRST ... RDS_INFO_LAST:
	318	ret = rds_info_getsockopt(sock, optname, optval,
	319	optlen);
	320	break;
	321
	322	case RDS_RECVERR:
	323	if (len < sizeof(int))
	324	ret = -EINVAL;
	325	else
f64f9e71 JP	326	if (put_user(rs->rs_recverr, (int __user *) optval) \|\|
f64f9e71 JP	327	put_user(sizeof(int), optlen))
639b321b AG	328	ret = -EFAULT;
	329	else
	330	ret = 0;
	331	break;
	332	default:
	333	break;
	334	}
	335
	336	out:
	337	return ret;
	338
	339	}
	340
	341	static int rds_connect(struct socket sock, struct sockaddr uaddr,
	342	int addr_len, int flags)
	343	{
	344	struct sock *sk = sock->sk;
	345	struct sockaddr_in sin = (struct sockaddr_in )uaddr;
	346	struct rds_sock *rs = rds_sk_to_rs(sk);
	347	int ret = 0;
	348
	349	lock_sock(sk);
	350
	351	if (addr_len != sizeof(struct sockaddr_in)) {
	352	ret = -EINVAL;
	353	goto out;
	354	}
	355
	356	if (sin->sin_family != AF_INET) {
	357	ret = -EAFNOSUPPORT;
	358	goto out;
	359	}
	360
	361	if (sin->sin_addr.s_addr == htonl(INADDR_ANY)) {
	362	ret = -EDESTADDRREQ;
	363	goto out;
	364	}
	365
	366	rs->rs_conn_addr = sin->sin_addr.s_addr;
	367	rs->rs_conn_port = sin->sin_port;
	368
	369	out:
	370	release_sock(sk);
	371	return ret;
	372	}
	373
	374	static struct proto rds_proto = {
	375	.name = "RDS",
	376	.owner = THIS_MODULE,
	377	.obj_size = sizeof(struct rds_sock),
	378	};
	379
5708e868	380	static const struct proto_ops rds_proto_ops = {
639b321b AG	381	.family = AF_RDS,
	382	.owner = THIS_MODULE,
	383	.release = rds_release,
	384	.bind = rds_bind,
	385	.connect = rds_connect,
	386	.socketpair = sock_no_socketpair,
	387	.accept = sock_no_accept,
	388	.getname = rds_getname,
	389	.poll = rds_poll,
	390	.ioctl = rds_ioctl,
	391	.listen = sock_no_listen,
	392	.shutdown = sock_no_shutdown,
	393	.setsockopt = rds_setsockopt,
	394	.getsockopt = rds_getsockopt,
	395	.sendmsg = rds_sendmsg,
	396	.recvmsg = rds_recvmsg,
	397	.mmap = sock_no_mmap,
	398	.sendpage = sock_no_sendpage,
	399	};
	400
	401	static int __rds_create(struct socket sock, struct sock sk, int protocol)
	402	{
	403	unsigned long flags;
	404	struct rds_sock *rs;
	405
	406	sock_init_data(sock, sk);
	407	sock->ops = &rds_proto_ops;
	408	sk->sk_protocol = protocol;
	409
	410	rs = rds_sk_to_rs(sk);
	411	spin_lock_init(&rs->rs_lock);
	412	rwlock_init(&rs->rs_recv_lock);
	413	INIT_LIST_HEAD(&rs->rs_send_queue);
	414	INIT_LIST_HEAD(&rs->rs_recv_queue);
	415	INIT_LIST_HEAD(&rs->rs_notify_queue);
	416	INIT_LIST_HEAD(&rs->rs_cong_list);
	417	spin_lock_init(&rs->rs_rdma_lock);
	418	rs->rs_rdma_keys = RB_ROOT;
	419
	420	spin_lock_irqsave(&rds_sock_lock, flags);
	421	list_add_tail(&rs->rs_item, &rds_sock_list);
	422	rds_sock_count++;
	423	spin_unlock_irqrestore(&rds_sock_lock, flags);
	424
	425	return 0;
	426	}
	427
3f378b68 EP	428	static int rds_create(struct net net, struct socket sock, int protocol,
3f378b68 EP	429	int kern)
639b321b AG	430	{
	431	struct sock *sk;
	432
	433	if (sock->type != SOCK_SEQPACKET \|\| protocol)
	434	return -ESOCKTNOSUPPORT;
	435
	436	sk = sk_alloc(net, AF_RDS, GFP_ATOMIC, &rds_proto);
	437	if (!sk)
	438	return -ENOMEM;
	439
	440	return __rds_create(sock, sk, protocol);
	441	}
	442
	443	void rds_sock_addref(struct rds_sock *rs)
	444	{
	445	sock_hold(rds_rs_to_sk(rs));
	446	}
	447
	448	void rds_sock_put(struct rds_sock *rs)
	449	{
	450	sock_put(rds_rs_to_sk(rs));
	451	}
	452
ec1b4cf7	453	static const struct net_proto_family rds_family_ops = {
639b321b AG	454	.family = AF_RDS,
	455	.create = rds_create,
	456	.owner = THIS_MODULE,
	457	};
	458
	459	static void rds_sock_inc_info(struct socket *sock, unsigned int len,
	460	struct rds_info_iterator *iter,
	461	struct rds_info_lengths *lens)
	462	{
	463	struct rds_sock *rs;
639b321b AG	464	struct rds_incoming *inc;
	465	unsigned long flags;
	466	unsigned int total = 0;
	467
	468	len /= sizeof(struct rds_info_message);
	469
	470	spin_lock_irqsave(&rds_sock_lock, flags);
	471
	472	list_for_each_entry(rs, &rds_sock_list, rs_item) {
639b321b AG	473	read_lock(&rs->rs_recv_lock);
	474
	475	/* XXX too lazy to maintain counts.. */
	476	list_for_each_entry(inc, &rs->rs_recv_queue, i_item) {
	477	total++;
	478	if (total <= len)
	479	rds_inc_info_copy(inc, iter, inc->i_saddr,
	480	rs->rs_bound_addr, 1);
	481	}
	482
	483	read_unlock(&rs->rs_recv_lock);
	484	}
	485
	486	spin_unlock_irqrestore(&rds_sock_lock, flags);
	487
	488	lens->nr = total;
	489	lens->each = sizeof(struct rds_info_message);
	490	}
	491
	492	static void rds_sock_info(struct socket *sock, unsigned int len,
	493	struct rds_info_iterator *iter,
	494	struct rds_info_lengths *lens)
	495	{
	496	struct rds_info_socket sinfo;
	497	struct rds_sock *rs;
	498	unsigned long flags;
	499
	500	len /= sizeof(struct rds_info_socket);
	501
	502	spin_lock_irqsave(&rds_sock_lock, flags);
	503
	504	if (len < rds_sock_count)
	505	goto out;
	506
	507	list_for_each_entry(rs, &rds_sock_list, rs_item) {
	508	sinfo.sndbuf = rds_sk_sndbuf(rs);
	509	sinfo.rcvbuf = rds_sk_rcvbuf(rs);
	510	sinfo.bound_addr = rs->rs_bound_addr;
	511	sinfo.connected_addr = rs->rs_conn_addr;
	512	sinfo.bound_port = rs->rs_bound_port;
	513	sinfo.connected_port = rs->rs_conn_port;
	514	sinfo.inum = sock_i_ino(rds_rs_to_sk(rs));
	515
	516	rds_info_copy(iter, &sinfo, sizeof(sinfo));
	517	}
	518
	519	out:
	520	lens->nr = rds_sock_count;
	521	lens->each = sizeof(struct rds_info_socket);
	522
	523	spin_unlock_irqrestore(&rds_sock_lock, flags);
	524	}
	525
ef87b7ea	526	static void rds_exit(void)
639b321b	527	{
639b321b AG	528	sock_unregister(rds_family_ops.family);
	529	proto_unregister(&rds_proto);
	530	rds_conn_exit();
	531	rds_cong_exit();
	532	rds_sysctl_exit();
	533	rds_threads_exit();
	534	rds_stats_exit();
	535	rds_page_exit();
	536	rds_info_deregister_func(RDS_INFO_SOCKETS, rds_sock_info);
	537	rds_info_deregister_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);
	538	}
	539	module_exit(rds_exit);
	540
ef87b7ea	541	static int rds_init(void)
639b321b AG	542	{
	543	int ret;
	544
	545	ret = rds_conn_init();
	546	if (ret)
	547	goto out;
	548	ret = rds_threads_init();
	549	if (ret)
	550	goto out_conn;
	551	ret = rds_sysctl_init();
	552	if (ret)
	553	goto out_threads;
	554	ret = rds_stats_init();
	555	if (ret)
	556	goto out_sysctl;
	557	ret = proto_register(&rds_proto, 1);
	558	if (ret)
	559	goto out_stats;
	560	ret = sock_register(&rds_family_ops);
	561	if (ret)
	562	goto out_proto;
	563
	564	rds_info_register_func(RDS_INFO_SOCKETS, rds_sock_info);
	565	rds_info_register_func(RDS_INFO_RECV_MESSAGES, rds_sock_inc_info);
	566
639b321b AG	567	goto out;
639b321b AG	568
639b321b AG	569	out_proto:
	570	proto_unregister(&rds_proto);
	571	out_stats:
	572	rds_stats_exit();
	573	out_sysctl:
	574	rds_sysctl_exit();
	575	out_threads:
	576	rds_threads_exit();
	577	out_conn:
	578	rds_conn_exit();
	579	rds_cong_exit();
	580	rds_page_exit();
	581	out:
	582	return ret;
	583	}
	584	module_init(rds_init);
	585
	586	#define DRV_VERSION "4.0"
	587	#define DRV_RELDATE "Feb 12, 2009"
	588
	589	MODULE_AUTHOR("Oracle Corporation <rds-devel@oss.oracle.com>");
	590	MODULE_DESCRIPTION("RDS: Reliable Datagram Sockets"
	591	" v" DRV_VERSION " (" DRV_RELDATE ")");
	592	MODULE_VERSION(DRV_VERSION);
	593	MODULE_LICENSE("Dual BSD/GPL");
	594	MODULE_ALIAS_NETPROTO(PF_RDS);