1/*
   2 * net/tipc/socket.c: TIPC socket API
   3 *
   4 * Copyright (c) 2001-2007, Ericsson AB
   5 * Copyright (c) 2004-2008, 2010-2011, Wind River Systems
   6 * All rights reserved.
   7 *
   8 * Redistribution and use in source and binary forms, with or without
   9 * modification, are permitted provided that the following conditions are met:
  10 *
  11 * 1. Redistributions of source code must retain the above copyright
  12 *    notice, this list of conditions and the following disclaimer.
  13 * 2. Redistributions in binary form must reproduce the above copyright
  14 *    notice, this list of conditions and the following disclaimer in the
  15 *    documentation and/or other materials provided with the distribution.
  16 * 3. Neither the names of the copyright holders nor the names of its
  17 *    contributors may be used to endorse or promote products derived from
  18 *    this software without specific prior written permission.
  19 *
  20 * Alternatively, this software may be distributed under the terms of the
  21 * GNU General Public License ("GPL") version 2 as published by the Free
  22 * Software Foundation.
  23 *
  24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  34 * POSSIBILITY OF SUCH DAMAGE.
  35 */
  36
 
  37#include <net/sock.h>
  38
  39#include "core.h"
  40#include "port.h"
  41
  42#define SS_LISTENING	-1	/* socket is listening */
  43#define SS_READY	-2	/* socket is connectionless */
  44
  45#define OVERLOAD_LIMIT_BASE	5000
  46#define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
  47
  48struct tipc_sock {
  49	struct sock sk;
  50	struct tipc_port *p;
  51	struct tipc_portid peer_name;
  52	long conn_timeout;
  53};
  54
  55#define tipc_sk(sk) ((struct tipc_sock *)(sk))
  56#define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
  57
  58#define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
  59			(sock->state == SS_DISCONNECTING))
  60
  61static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
  62static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
  63static void wakeupdispatch(struct tipc_port *tport);
  64
  65static const struct proto_ops packet_ops;
  66static const struct proto_ops stream_ops;
  67static const struct proto_ops msg_ops;
  68
  69static struct proto tipc_proto;
  70
  71static int sockets_enabled;
  72
  73static atomic_t tipc_queue_size = ATOMIC_INIT(0);
  74
  75/*
  76 * Revised TIPC socket locking policy:
  77 *
  78 * Most socket operations take the standard socket lock when they start
  79 * and hold it until they finish (or until they need to sleep).  Acquiring
  80 * this lock grants the owner exclusive access to the fields of the socket
  81 * data structures, with the exception of the backlog queue.  A few socket
  82 * operations can be done without taking the socket lock because they only
  83 * read socket information that never changes during the life of the socket.
  84 *
  85 * Socket operations may acquire the lock for the associated TIPC port if they
  86 * need to perform an operation on the port.  If any routine needs to acquire
  87 * both the socket lock and the port lock it must take the socket lock first
  88 * to avoid the risk of deadlock.
  89 *
  90 * The dispatcher handling incoming messages cannot grab the socket lock in
  91 * the standard fashion, since invoked it runs at the BH level and cannot block.
  92 * Instead, it checks to see if the socket lock is currently owned by someone,
  93 * and either handles the message itself or adds it to the socket's backlog
  94 * queue; in the latter case the queued message is processed once the process
  95 * owning the socket lock releases it.
  96 *
  97 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
  98 * the problem of a blocked socket operation preventing any other operations
  99 * from occurring.  However, applications must be careful if they have
 100 * multiple threads trying to send (or receive) on the same socket, as these
 101 * operations might interfere with each other.  For example, doing a connect
 102 * and a receive at the same time might allow the receive to consume the
 103 * ACK message meant for the connect.  While additional work could be done
 104 * to try and overcome this, it doesn't seem to be worthwhile at the present.
 105 *
 106 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
 107 * that another operation that must be performed in a non-blocking manner is
 108 * not delayed for very long because the lock has already been taken.
 109 *
 110 * NOTE: This code assumes that certain fields of a port/socket pair are
 111 * constant over its lifetime; such fields can be examined without taking
 112 * the socket lock and/or port lock, and do not need to be re-read even
 113 * after resuming processing after waiting.  These fields include:
 114 *   - socket type
 115 *   - pointer to socket sk structure (aka tipc_sock structure)
 116 *   - pointer to port structure
 117 *   - port reference
 118 */
 119
 120/**
 121 * advance_rx_queue - discard first buffer in socket receive queue
 122 *
 123 * Caller must hold socket lock
 124 */
 125
 126static void advance_rx_queue(struct sock *sk)
 127{
 128	buf_discard(__skb_dequeue(&sk->sk_receive_queue));
 129	atomic_dec(&tipc_queue_size);
 130}
 131
 132/**
 133 * discard_rx_queue - discard all buffers in socket receive queue
 134 *
 135 * Caller must hold socket lock
 136 */
 137
 138static void discard_rx_queue(struct sock *sk)
 139{
 140	struct sk_buff *buf;
 141
 142	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
 143		atomic_dec(&tipc_queue_size);
 144		buf_discard(buf);
 145	}
 146}
 147
 148/**
 149 * reject_rx_queue - reject all buffers in socket receive queue
 150 *
 151 * Caller must hold socket lock
 152 */
 153
 154static void reject_rx_queue(struct sock *sk)
 155{
 156	struct sk_buff *buf;
 157
 158	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
 159		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
 160		atomic_dec(&tipc_queue_size);
 161	}
 162}
 163
 164/**
 165 * tipc_create - create a TIPC socket
 166 * @net: network namespace (must be default network)
 167 * @sock: pre-allocated socket structure
 168 * @protocol: protocol indicator (must be 0)
 169 * @kern: caused by kernel or by userspace?
 170 *
 171 * This routine creates additional data structures used by the TIPC socket,
 172 * initializes them, and links them together.
 173 *
 174 * Returns 0 on success, errno otherwise
 175 */
 176
 177static int tipc_create(struct net *net, struct socket *sock, int protocol,
 178		       int kern)
 179{
 180	const struct proto_ops *ops;
 181	socket_state state;
 182	struct sock *sk;
 183	struct tipc_port *tp_ptr;
 184
 185	/* Validate arguments */
 186
 187	if (!net_eq(net, &init_net))
 188		return -EAFNOSUPPORT;
 189
 190	if (unlikely(protocol != 0))
 191		return -EPROTONOSUPPORT;
 192
 193	switch (sock->type) {
 194	case SOCK_STREAM:
 195		ops = &stream_ops;
 196		state = SS_UNCONNECTED;
 197		break;
 198	case SOCK_SEQPACKET:
 199		ops = &packet_ops;
 200		state = SS_UNCONNECTED;
 201		break;
 202	case SOCK_DGRAM:
 203	case SOCK_RDM:
 204		ops = &msg_ops;
 205		state = SS_READY;
 206		break;
 207	default:
 208		return -EPROTOTYPE;
 209	}
 210
 211	/* Allocate socket's protocol area */
 212
 213	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
 214	if (sk == NULL)
 215		return -ENOMEM;
 216
 217	/* Allocate TIPC port for socket to use */
 218
 219	tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
 220				     TIPC_LOW_IMPORTANCE);
 221	if (unlikely(!tp_ptr)) {
 222		sk_free(sk);
 223		return -ENOMEM;
 224	}
 225
 226	/* Finish initializing socket data structures */
 227
 228	sock->ops = ops;
 229	sock->state = state;
 230
 231	sock_init_data(sock, sk);
 232	sk->sk_backlog_rcv = backlog_rcv;
 233	tipc_sk(sk)->p = tp_ptr;
 234	tipc_sk(sk)->conn_timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
 235
 236	spin_unlock_bh(tp_ptr->lock);
 237
 238	if (sock->state == SS_READY) {
 239		tipc_set_portunreturnable(tp_ptr->ref, 1);
 240		if (sock->type == SOCK_DGRAM)
 241			tipc_set_portunreliable(tp_ptr->ref, 1);
 242	}
 243
 244	return 0;
 245}
 246
 247/**
 248 * release - destroy a TIPC socket
 249 * @sock: socket to destroy
 250 *
 251 * This routine cleans up any messages that are still queued on the socket.
 252 * For DGRAM and RDM socket types, all queued messages are rejected.
 253 * For SEQPACKET and STREAM socket types, the first message is rejected
 254 * and any others are discarded.  (If the first message on a STREAM socket
 255 * is partially-read, it is discarded and the next one is rejected instead.)
 256 *
 257 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 258 * are returned or discarded according to the "destination droppable" setting
 259 * specified for the message by the sender.
 260 *
 261 * Returns 0 on success, errno otherwise
 262 */
 263
 264static int release(struct socket *sock)
 265{
 266	struct sock *sk = sock->sk;
 267	struct tipc_port *tport;
 268	struct sk_buff *buf;
 269	int res;
 270
 271	/*
 272	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 273	 * releases a pre-allocated child socket that was never used)
 274	 */
 275
 276	if (sk == NULL)
 277		return 0;
 278
 279	tport = tipc_sk_port(sk);
 280	lock_sock(sk);
 281
 282	/*
 283	 * Reject all unreceived messages, except on an active connection
 284	 * (which disconnects locally & sends a 'FIN+' to peer)
 285	 */
 286
 287	while (sock->state != SS_DISCONNECTING) {
 288		buf = __skb_dequeue(&sk->sk_receive_queue);
 289		if (buf == NULL)
 290			break;
 291		atomic_dec(&tipc_queue_size);
 292		if (TIPC_SKB_CB(buf)->handle != 0)
 293			buf_discard(buf);
 294		else {
 295			if ((sock->state == SS_CONNECTING) ||
 296			    (sock->state == SS_CONNECTED)) {
 297				sock->state = SS_DISCONNECTING;
 298				tipc_disconnect(tport->ref);
 299			}
 300			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
 301		}
 302	}
 303
 304	/*
 305	 * Delete TIPC port; this ensures no more messages are queued
 306	 * (also disconnects an active connection & sends a 'FIN-' to peer)
 307	 */
 308
 309	res = tipc_deleteport(tport->ref);
 310
 311	/* Discard any remaining (connection-based) messages in receive queue */
 312
 313	discard_rx_queue(sk);
 314
 315	/* Reject any messages that accumulated in backlog queue */
 316
 317	sock->state = SS_DISCONNECTING;
 318	release_sock(sk);
 319
 320	sock_put(sk);
 321	sock->sk = NULL;
 322
 323	return res;
 324}
 325
 326/**
 327 * bind - associate or disassocate TIPC name(s) with a socket
 328 * @sock: socket structure
 329 * @uaddr: socket address describing name(s) and desired operation
 330 * @uaddr_len: size of socket address data structure
 331 *
 332 * Name and name sequence binding is indicated using a positive scope value;
 333 * a negative scope value unbinds the specified name.  Specifying no name
 334 * (i.e. a socket address length of 0) unbinds all names from the socket.
 335 *
 336 * Returns 0 on success, errno otherwise
 337 *
 338 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 339 *       access any non-constant socket information.
 340 */
 341
 342static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
 343{
 344	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 345	u32 portref = tipc_sk_port(sock->sk)->ref;
 346
 347	if (unlikely(!uaddr_len))
 348		return tipc_withdraw(portref, 0, NULL);
 349
 350	if (uaddr_len < sizeof(struct sockaddr_tipc))
 351		return -EINVAL;
 352	if (addr->family != AF_TIPC)
 353		return -EAFNOSUPPORT;
 354
 355	if (addr->addrtype == TIPC_ADDR_NAME)
 356		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
 357	else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
 358		return -EAFNOSUPPORT;
 359
 
 
 
 360	return (addr->scope > 0) ?
 361		tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
 362		tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
 363}
 364
 365/**
 366 * get_name - get port ID of socket or peer socket
 367 * @sock: socket structure
 368 * @uaddr: area for returned socket address
 369 * @uaddr_len: area for returned length of socket address
 370 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 371 *
 372 * Returns 0 on success, errno otherwise
 373 *
 374 * NOTE: This routine doesn't need to take the socket lock since it only
 375 *       accesses socket information that is unchanging (or which changes in
 376 *       a completely predictable manner).
 377 */
 378
 379static int get_name(struct socket *sock, struct sockaddr *uaddr,
 380		    int *uaddr_len, int peer)
 381{
 382	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 383	struct tipc_sock *tsock = tipc_sk(sock->sk);
 384
 385	memset(addr, 0, sizeof(*addr));
 386	if (peer) {
 387		if ((sock->state != SS_CONNECTED) &&
 388			((peer != 2) || (sock->state != SS_DISCONNECTING)))
 389			return -ENOTCONN;
 390		addr->addr.id.ref = tsock->peer_name.ref;
 391		addr->addr.id.node = tsock->peer_name.node;
 392	} else {
 393		addr->addr.id.ref = tsock->p->ref;
 394		addr->addr.id.node = tipc_own_addr;
 395	}
 396
 397	*uaddr_len = sizeof(*addr);
 398	addr->addrtype = TIPC_ADDR_ID;
 399	addr->family = AF_TIPC;
 400	addr->scope = 0;
 401	addr->addr.name.domain = 0;
 402
 403	return 0;
 404}
 405
 406/**
 407 * poll - read and possibly block on pollmask
 408 * @file: file structure associated with the socket
 409 * @sock: socket for which to calculate the poll bits
 410 * @wait: ???
 411 *
 412 * Returns pollmask value
 413 *
 414 * COMMENTARY:
 415 * It appears that the usual socket locking mechanisms are not useful here
 416 * since the pollmask info is potentially out-of-date the moment this routine
 417 * exits.  TCP and other protocols seem to rely on higher level poll routines
 418 * to handle any preventable race conditions, so TIPC will do the same ...
 419 *
 420 * TIPC sets the returned events as follows:
 421 *
 422 * socket state		flags set
 423 * ------------		---------
 424 * unconnected		no read flags
 425 *			no write flags
 426 *
 427 * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
 428 *			no write flags
 429 *
 430 * connected		POLLIN/POLLRDNORM if data in rx queue
 431 *			POLLOUT if port is not congested
 432 *
 433 * disconnecting	POLLIN/POLLRDNORM/POLLHUP
 434 *			no write flags
 435 *
 436 * listening		POLLIN if SYN in rx queue
 437 *			no write flags
 438 *
 439 * ready		POLLIN/POLLRDNORM if data in rx queue
 440 * [connectionless]	POLLOUT (since port cannot be congested)
 441 *
 442 * IMPORTANT: The fact that a read or write operation is indicated does NOT
 443 * imply that the operation will succeed, merely that it should be performed
 444 * and will not block.
 445 */
 446
 447static unsigned int poll(struct file *file, struct socket *sock,
 448			 poll_table *wait)
 449{
 450	struct sock *sk = sock->sk;
 451	u32 mask = 0;
 452
 453	poll_wait(file, sk_sleep(sk), wait);
 454
 455	switch ((int)sock->state) {
 456	case SS_READY:
 457	case SS_CONNECTED:
 458		if (!tipc_sk_port(sk)->congested)
 459			mask |= POLLOUT;
 460		/* fall thru' */
 461	case SS_CONNECTING:
 462	case SS_LISTENING:
 463		if (!skb_queue_empty(&sk->sk_receive_queue))
 464			mask |= (POLLIN | POLLRDNORM);
 465		break;
 466	case SS_DISCONNECTING:
 467		mask = (POLLIN | POLLRDNORM | POLLHUP);
 468		break;
 469	}
 470
 471	return mask;
 472}
 473
 474/**
 475 * dest_name_check - verify user is permitted to send to specified port name
 476 * @dest: destination address
 477 * @m: descriptor for message to be sent
 478 *
 479 * Prevents restricted configuration commands from being issued by
 480 * unauthorized users.
 481 *
 482 * Returns 0 if permission is granted, otherwise errno
 483 */
 484
 485static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
 486{
 487	struct tipc_cfg_msg_hdr hdr;
 488
 489	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
 490		return 0;
 491	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
 492		return 0;
 493	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
 494		return -EACCES;
 495
 496	if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
 497		return -EMSGSIZE;
 498	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
 499		return -EFAULT;
 500	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
 501		return -EACCES;
 502
 503	return 0;
 504}
 505
 506/**
 507 * send_msg - send message in connectionless manner
 508 * @iocb: if NULL, indicates that socket lock is already held
 509 * @sock: socket structure
 510 * @m: message to send
 511 * @total_len: length of message
 512 *
 513 * Message must have an destination specified explicitly.
 514 * Used for SOCK_RDM and SOCK_DGRAM messages,
 515 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
 516 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
 517 *
 518 * Returns the number of bytes sent on success, or errno otherwise
 519 */
 520
 521static int send_msg(struct kiocb *iocb, struct socket *sock,
 522		    struct msghdr *m, size_t total_len)
 523{
 524	struct sock *sk = sock->sk;
 525	struct tipc_port *tport = tipc_sk_port(sk);
 526	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
 527	int needs_conn;
 
 528	int res = -EINVAL;
 529
 530	if (unlikely(!dest))
 531		return -EDESTADDRREQ;
 532	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
 533		     (dest->family != AF_TIPC)))
 534		return -EINVAL;
 535	if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
 536	    (m->msg_iovlen > (unsigned)INT_MAX))
 537		return -EMSGSIZE;
 538
 539	if (iocb)
 540		lock_sock(sk);
 541
 542	needs_conn = (sock->state != SS_READY);
 543	if (unlikely(needs_conn)) {
 544		if (sock->state == SS_LISTENING) {
 545			res = -EPIPE;
 546			goto exit;
 547		}
 548		if (sock->state != SS_UNCONNECTED) {
 549			res = -EISCONN;
 550			goto exit;
 551		}
 552		if ((tport->published) ||
 553		    ((sock->type == SOCK_STREAM) && (total_len != 0))) {
 554			res = -EOPNOTSUPP;
 555			goto exit;
 556		}
 557		if (dest->addrtype == TIPC_ADDR_NAME) {
 558			tport->conn_type = dest->addr.name.name.type;
 559			tport->conn_instance = dest->addr.name.name.instance;
 560		}
 561
 562		/* Abort any pending connection attempts (very unlikely) */
 563
 564		reject_rx_queue(sk);
 565	}
 566
 
 
 567	do {
 568		if (dest->addrtype == TIPC_ADDR_NAME) {
 569			res = dest_name_check(dest, m);
 570			if (res)
 571				break;
 572			res = tipc_send2name(tport->ref,
 573					     &dest->addr.name.name,
 574					     dest->addr.name.domain,
 575					     m->msg_iovlen,
 576					     m->msg_iov,
 577					     total_len);
 578		} else if (dest->addrtype == TIPC_ADDR_ID) {
 579			res = tipc_send2port(tport->ref,
 580					     &dest->addr.id,
 581					     m->msg_iovlen,
 582					     m->msg_iov,
 583					     total_len);
 584		} else if (dest->addrtype == TIPC_ADDR_MCAST) {
 585			if (needs_conn) {
 586				res = -EOPNOTSUPP;
 587				break;
 588			}
 589			res = dest_name_check(dest, m);
 590			if (res)
 591				break;
 592			res = tipc_multicast(tport->ref,
 593					     &dest->addr.nameseq,
 594					     m->msg_iovlen,
 595					     m->msg_iov,
 596					     total_len);
 597		}
 598		if (likely(res != -ELINKCONG)) {
 599			if (needs_conn && (res >= 0))
 600				sock->state = SS_CONNECTING;
 601			break;
 602		}
 603		if (m->msg_flags & MSG_DONTWAIT) {
 604			res = -EWOULDBLOCK;
 605			break;
 606		}
 607		release_sock(sk);
 608		res = wait_event_interruptible(*sk_sleep(sk),
 609					       !tport->congested);
 610		lock_sock(sk);
 611		if (res)
 612			break;
 613	} while (1);
 614
 615exit:
 616	if (iocb)
 617		release_sock(sk);
 618	return res;
 619}
 620
 621/**
 622 * send_packet - send a connection-oriented message
 623 * @iocb: if NULL, indicates that socket lock is already held
 624 * @sock: socket structure
 625 * @m: message to send
 626 * @total_len: length of message
 627 *
 628 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
 629 *
 630 * Returns the number of bytes sent on success, or errno otherwise
 631 */
 632
 633static int send_packet(struct kiocb *iocb, struct socket *sock,
 634		       struct msghdr *m, size_t total_len)
 635{
 636	struct sock *sk = sock->sk;
 637	struct tipc_port *tport = tipc_sk_port(sk);
 638	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
 
 639	int res;
 640
 641	/* Handle implied connection establishment */
 642
 643	if (unlikely(dest))
 644		return send_msg(iocb, sock, m, total_len);
 645
 646	if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
 647	    (m->msg_iovlen > (unsigned)INT_MAX))
 648		return -EMSGSIZE;
 649
 650	if (iocb)
 651		lock_sock(sk);
 652
 
 
 653	do {
 654		if (unlikely(sock->state != SS_CONNECTED)) {
 655			if (sock->state == SS_DISCONNECTING)
 656				res = -EPIPE;
 657			else
 658				res = -ENOTCONN;
 659			break;
 660		}
 661
 662		res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
 663				total_len);
 664		if (likely(res != -ELINKCONG))
 665			break;
 666		if (m->msg_flags & MSG_DONTWAIT) {
 667			res = -EWOULDBLOCK;
 668			break;
 669		}
 670		release_sock(sk);
 671		res = wait_event_interruptible(*sk_sleep(sk),
 672			(!tport->congested || !tport->connected));
 673		lock_sock(sk);
 674		if (res)
 675			break;
 676	} while (1);
 677
 678	if (iocb)
 679		release_sock(sk);
 680	return res;
 681}
 682
 683/**
 684 * send_stream - send stream-oriented data
 685 * @iocb: (unused)
 686 * @sock: socket structure
 687 * @m: data to send
 688 * @total_len: total length of data to be sent
 689 *
 690 * Used for SOCK_STREAM data.
 691 *
 692 * Returns the number of bytes sent on success (or partial success),
 693 * or errno if no data sent
 694 */
 695
 696static int send_stream(struct kiocb *iocb, struct socket *sock,
 697		       struct msghdr *m, size_t total_len)
 698{
 699	struct sock *sk = sock->sk;
 700	struct tipc_port *tport = tipc_sk_port(sk);
 701	struct msghdr my_msg;
 702	struct iovec my_iov;
 703	struct iovec *curr_iov;
 704	int curr_iovlen;
 705	char __user *curr_start;
 706	u32 hdr_size;
 707	int curr_left;
 708	int bytes_to_send;
 709	int bytes_sent;
 710	int res;
 711
 712	lock_sock(sk);
 713
 714	/* Handle special cases where there is no connection */
 715
 716	if (unlikely(sock->state != SS_CONNECTED)) {
 717		if (sock->state == SS_UNCONNECTED) {
 718			res = send_packet(NULL, sock, m, total_len);
 719			goto exit;
 720		} else if (sock->state == SS_DISCONNECTING) {
 721			res = -EPIPE;
 722			goto exit;
 723		} else {
 724			res = -ENOTCONN;
 725			goto exit;
 726		}
 727	}
 728
 729	if (unlikely(m->msg_name)) {
 730		res = -EISCONN;
 731		goto exit;
 732	}
 733
 734	if ((total_len > (unsigned)INT_MAX) ||
 735	    (m->msg_iovlen > (unsigned)INT_MAX)) {
 736		res = -EMSGSIZE;
 737		goto exit;
 738	}
 739
 740	/*
 741	 * Send each iovec entry using one or more messages
 742	 *
 743	 * Note: This algorithm is good for the most likely case
 744	 * (i.e. one large iovec entry), but could be improved to pass sets
 745	 * of small iovec entries into send_packet().
 746	 */
 747
 748	curr_iov = m->msg_iov;
 749	curr_iovlen = m->msg_iovlen;
 750	my_msg.msg_iov = &my_iov;
 751	my_msg.msg_iovlen = 1;
 752	my_msg.msg_flags = m->msg_flags;
 753	my_msg.msg_name = NULL;
 754	bytes_sent = 0;
 755
 756	hdr_size = msg_hdr_sz(&tport->phdr);
 757
 758	while (curr_iovlen--) {
 759		curr_start = curr_iov->iov_base;
 760		curr_left = curr_iov->iov_len;
 761
 762		while (curr_left) {
 763			bytes_to_send = tport->max_pkt - hdr_size;
 764			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
 765				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
 766			if (curr_left < bytes_to_send)
 767				bytes_to_send = curr_left;
 768			my_iov.iov_base = curr_start;
 769			my_iov.iov_len = bytes_to_send;
 770			res = send_packet(NULL, sock, &my_msg, bytes_to_send);
 771			if (res < 0) {
 772				if (bytes_sent)
 773					res = bytes_sent;
 774				goto exit;
 775			}
 776			curr_left -= bytes_to_send;
 777			curr_start += bytes_to_send;
 778			bytes_sent += bytes_to_send;
 779		}
 780
 781		curr_iov++;
 782	}
 783	res = bytes_sent;
 784exit:
 785	release_sock(sk);
 786	return res;
 787}
 788
 789/**
 790 * auto_connect - complete connection setup to a remote port
 791 * @sock: socket structure
 792 * @msg: peer's response message
 793 *
 794 * Returns 0 on success, errno otherwise
 795 */
 796
 797static int auto_connect(struct socket *sock, struct tipc_msg *msg)
 798{
 799	struct tipc_sock *tsock = tipc_sk(sock->sk);
 800
 801	if (msg_errcode(msg)) {
 802		sock->state = SS_DISCONNECTING;
 803		return -ECONNREFUSED;
 804	}
 805
 806	tsock->peer_name.ref = msg_origport(msg);
 807	tsock->peer_name.node = msg_orignode(msg);
 808	tipc_connect2port(tsock->p->ref, &tsock->peer_name);
 809	tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
 810	sock->state = SS_CONNECTED;
 811	return 0;
 812}
 813
 814/**
 815 * set_orig_addr - capture sender's address for received message
 816 * @m: descriptor for message info
 817 * @msg: received message header
 818 *
 819 * Note: Address is not captured if not requested by receiver.
 820 */
 821
 822static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
 823{
 824	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
 825
 826	if (addr) {
 827		addr->family = AF_TIPC;
 828		addr->addrtype = TIPC_ADDR_ID;
 829		addr->addr.id.ref = msg_origport(msg);
 830		addr->addr.id.node = msg_orignode(msg);
 831		addr->addr.name.domain = 0;	/* could leave uninitialized */
 832		addr->scope = 0;		/* could leave uninitialized */
 833		m->msg_namelen = sizeof(struct sockaddr_tipc);
 834	}
 835}
 836
 837/**
 838 * anc_data_recv - optionally capture ancillary data for received message
 839 * @m: descriptor for message info
 840 * @msg: received message header
 841 * @tport: TIPC port associated with message
 842 *
 843 * Note: Ancillary data is not captured if not requested by receiver.
 844 *
 845 * Returns 0 if successful, otherwise errno
 846 */
 847
 848static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
 849				struct tipc_port *tport)
 850{
 851	u32 anc_data[3];
 852	u32 err;
 853	u32 dest_type;
 854	int has_name;
 855	int res;
 856
 857	if (likely(m->msg_controllen == 0))
 858		return 0;
 859
 860	/* Optionally capture errored message object(s) */
 861
 862	err = msg ? msg_errcode(msg) : 0;
 863	if (unlikely(err)) {
 864		anc_data[0] = err;
 865		anc_data[1] = msg_data_sz(msg);
 866		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
 867		if (res)
 868			return res;
 869		if (anc_data[1]) {
 870			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
 871				       msg_data(msg));
 872			if (res)
 873				return res;
 874		}
 875	}
 876
 877	/* Optionally capture message destination object */
 878
 879	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
 880	switch (dest_type) {
 881	case TIPC_NAMED_MSG:
 882		has_name = 1;
 883		anc_data[0] = msg_nametype(msg);
 884		anc_data[1] = msg_namelower(msg);
 885		anc_data[2] = msg_namelower(msg);
 886		break;
 887	case TIPC_MCAST_MSG:
 888		has_name = 1;
 889		anc_data[0] = msg_nametype(msg);
 890		anc_data[1] = msg_namelower(msg);
 891		anc_data[2] = msg_nameupper(msg);
 892		break;
 893	case TIPC_CONN_MSG:
 894		has_name = (tport->conn_type != 0);
 895		anc_data[0] = tport->conn_type;
 896		anc_data[1] = tport->conn_instance;
 897		anc_data[2] = tport->conn_instance;
 898		break;
 899	default:
 900		has_name = 0;
 901	}
 902	if (has_name) {
 903		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
 904		if (res)
 905			return res;
 906	}
 907
 908	return 0;
 909}
 910
 911/**
 912 * recv_msg - receive packet-oriented message
 913 * @iocb: (unused)
 914 * @m: descriptor for message info
 915 * @buf_len: total size of user buffer area
 916 * @flags: receive flags
 917 *
 918 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
 919 * If the complete message doesn't fit in user area, truncate it.
 920 *
 921 * Returns size of returned message data, errno otherwise
 922 */
 923
 924static int recv_msg(struct kiocb *iocb, struct socket *sock,
 925		    struct msghdr *m, size_t buf_len, int flags)
 926{
 927	struct sock *sk = sock->sk;
 928	struct tipc_port *tport = tipc_sk_port(sk);
 929	struct sk_buff *buf;
 930	struct tipc_msg *msg;
 931	long timeout;
 932	unsigned int sz;
 933	u32 err;
 934	int res;
 935
 936	/* Catch invalid receive requests */
 937
 938	if (unlikely(!buf_len))
 939		return -EINVAL;
 940
 941	lock_sock(sk);
 942
 943	if (unlikely(sock->state == SS_UNCONNECTED)) {
 944		res = -ENOTCONN;
 945		goto exit;
 946	}
 947
 948	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
 949restart:
 950
 951	/* Look for a message in receive queue; wait if necessary */
 952
 953	while (skb_queue_empty(&sk->sk_receive_queue)) {
 954		if (sock->state == SS_DISCONNECTING) {
 955			res = -ENOTCONN;
 956			goto exit;
 957		}
 958		if (timeout <= 0L) {
 959			res = timeout ? timeout : -EWOULDBLOCK;
 960			goto exit;
 961		}
 962		release_sock(sk);
 963		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
 964							   tipc_rx_ready(sock),
 965							   timeout);
 966		lock_sock(sk);
 967	}
 968
 969	/* Look at first message in receive queue */
 970
 971	buf = skb_peek(&sk->sk_receive_queue);
 972	msg = buf_msg(buf);
 973	sz = msg_data_sz(msg);
 974	err = msg_errcode(msg);
 975
 976	/* Complete connection setup for an implied connect */
 977
 978	if (unlikely(sock->state == SS_CONNECTING)) {
 979		res = auto_connect(sock, msg);
 980		if (res)
 981			goto exit;
 982	}
 983
 984	/* Discard an empty non-errored message & try again */
 985
 986	if ((!sz) && (!err)) {
 987		advance_rx_queue(sk);
 988		goto restart;
 989	}
 990
 991	/* Capture sender's address (optional) */
 992
 993	set_orig_addr(m, msg);
 994
 995	/* Capture ancillary data (optional) */
 996
 997	res = anc_data_recv(m, msg, tport);
 998	if (res)
 999		goto exit;
1000
1001	/* Capture message data (if valid) & compute return value (always) */
1002
1003	if (!err) {
1004		if (unlikely(buf_len < sz)) {
1005			sz = buf_len;
1006			m->msg_flags |= MSG_TRUNC;
1007		}
1008		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1009					      m->msg_iov, sz);
1010		if (res)
1011			goto exit;
1012		res = sz;
1013	} else {
1014		if ((sock->state == SS_READY) ||
1015		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1016			res = 0;
1017		else
1018			res = -ECONNRESET;
1019	}
1020
1021	/* Consume received message (optional) */
1022
1023	if (likely(!(flags & MSG_PEEK))) {
1024		if ((sock->state != SS_READY) &&
1025		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1026			tipc_acknowledge(tport->ref, tport->conn_unacked);
1027		advance_rx_queue(sk);
1028	}
1029exit:
1030	release_sock(sk);
1031	return res;
1032}
1033
1034/**
1035 * recv_stream - receive stream-oriented data
1036 * @iocb: (unused)
1037 * @m: descriptor for message info
1038 * @buf_len: total size of user buffer area
1039 * @flags: receive flags
1040 *
1041 * Used for SOCK_STREAM messages only.  If not enough data is available
1042 * will optionally wait for more; never truncates data.
1043 *
1044 * Returns size of returned message data, errno otherwise
1045 */
1046
1047static int recv_stream(struct kiocb *iocb, struct socket *sock,
1048		       struct msghdr *m, size_t buf_len, int flags)
1049{
1050	struct sock *sk = sock->sk;
1051	struct tipc_port *tport = tipc_sk_port(sk);
1052	struct sk_buff *buf;
1053	struct tipc_msg *msg;
1054	long timeout;
1055	unsigned int sz;
1056	int sz_to_copy, target, needed;
1057	int sz_copied = 0;
1058	u32 err;
1059	int res = 0;
1060
1061	/* Catch invalid receive attempts */
1062
1063	if (unlikely(!buf_len))
1064		return -EINVAL;
1065
1066	lock_sock(sk);
1067
1068	if (unlikely((sock->state == SS_UNCONNECTED) ||
1069		     (sock->state == SS_CONNECTING))) {
1070		res = -ENOTCONN;
1071		goto exit;
1072	}
1073
1074	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1075	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1076restart:
1077
 
1078	/* Look for a message in receive queue; wait if necessary */
1079
1080	while (skb_queue_empty(&sk->sk_receive_queue)) {
1081		if (sock->state == SS_DISCONNECTING) {
1082			res = -ENOTCONN;
1083			goto exit;
1084		}
1085		if (timeout <= 0L) {
1086			res = timeout ? timeout : -EWOULDBLOCK;
1087			goto exit;
1088		}
1089		release_sock(sk);
1090		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1091							   tipc_rx_ready(sock),
1092							   timeout);
1093		lock_sock(sk);
1094	}
1095
1096	/* Look at first message in receive queue */
1097
1098	buf = skb_peek(&sk->sk_receive_queue);
1099	msg = buf_msg(buf);
1100	sz = msg_data_sz(msg);
1101	err = msg_errcode(msg);
1102
1103	/* Discard an empty non-errored message & try again */
1104
1105	if ((!sz) && (!err)) {
1106		advance_rx_queue(sk);
1107		goto restart;
1108	}
1109
1110	/* Optionally capture sender's address & ancillary data of first msg */
1111
1112	if (sz_copied == 0) {
1113		set_orig_addr(m, msg);
1114		res = anc_data_recv(m, msg, tport);
1115		if (res)
1116			goto exit;
1117	}
1118
1119	/* Capture message data (if valid) & compute return value (always) */
1120
1121	if (!err) {
1122		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1123
1124		sz -= offset;
1125		needed = (buf_len - sz_copied);
1126		sz_to_copy = (sz <= needed) ? sz : needed;
1127
1128		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1129					      m->msg_iov, sz_to_copy);
1130		if (res)
1131			goto exit;
1132
1133		sz_copied += sz_to_copy;
1134
1135		if (sz_to_copy < sz) {
1136			if (!(flags & MSG_PEEK))
1137				TIPC_SKB_CB(buf)->handle =
1138				(void *)(unsigned long)(offset + sz_to_copy);
1139			goto exit;
1140		}
1141	} else {
1142		if (sz_copied != 0)
1143			goto exit; /* can't add error msg to valid data */
1144
1145		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1146			res = 0;
1147		else
1148			res = -ECONNRESET;
1149	}
1150
1151	/* Consume received message (optional) */
1152
1153	if (likely(!(flags & MSG_PEEK))) {
1154		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1155			tipc_acknowledge(tport->ref, tport->conn_unacked);
1156		advance_rx_queue(sk);
1157	}
1158
1159	/* Loop around if more data is required */
1160
1161	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1162	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1163	    (sz_copied < target)) &&	/* and more is ready or required */
1164	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1165	    (!err))			/* and haven't reached a FIN */
1166		goto restart;
1167
1168exit:
1169	release_sock(sk);
1170	return sz_copied ? sz_copied : res;
1171}
1172
1173/**
1174 * rx_queue_full - determine if receive queue can accept another message
1175 * @msg: message to be added to queue
1176 * @queue_size: current size of queue
1177 * @base: nominal maximum size of queue
1178 *
1179 * Returns 1 if queue is unable to accept message, 0 otherwise
1180 */
1181
1182static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1183{
1184	u32 threshold;
1185	u32 imp = msg_importance(msg);
1186
1187	if (imp == TIPC_LOW_IMPORTANCE)
1188		threshold = base;
1189	else if (imp == TIPC_MEDIUM_IMPORTANCE)
1190		threshold = base * 2;
1191	else if (imp == TIPC_HIGH_IMPORTANCE)
1192		threshold = base * 100;
1193	else
1194		return 0;
1195
1196	if (msg_connected(msg))
1197		threshold *= 4;
1198
1199	return queue_size >= threshold;
1200}
1201
1202/**
1203 * filter_rcv - validate incoming message
1204 * @sk: socket
1205 * @buf: message
1206 *
1207 * Enqueues message on receive queue if acceptable; optionally handles
1208 * disconnect indication for a connected socket.
1209 *
1210 * Called with socket lock already taken; port lock may also be taken.
1211 *
1212 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1213 */
1214
1215static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1216{
1217	struct socket *sock = sk->sk_socket;
1218	struct tipc_msg *msg = buf_msg(buf);
1219	u32 recv_q_len;
1220
1221	/* Reject message if it is wrong sort of message for socket */
1222
1223	/*
1224	 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1225	 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1226	 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1227	 */
1228
1229	if (sock->state == SS_READY) {
1230		if (msg_connected(msg))
1231			return TIPC_ERR_NO_PORT;
1232	} else {
1233		if (msg_mcast(msg))
1234			return TIPC_ERR_NO_PORT;
1235		if (sock->state == SS_CONNECTED) {
1236			if (!msg_connected(msg))
 
1237				return TIPC_ERR_NO_PORT;
1238		} else if (sock->state == SS_CONNECTING) {
1239			if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1240				return TIPC_ERR_NO_PORT;
1241		} else if (sock->state == SS_LISTENING) {
1242			if (msg_connected(msg) || msg_errcode(msg))
1243				return TIPC_ERR_NO_PORT;
1244		} else if (sock->state == SS_DISCONNECTING) {
1245			return TIPC_ERR_NO_PORT;
1246		} else /* (sock->state == SS_UNCONNECTED) */ {
1247			if (msg_connected(msg) || msg_errcode(msg))
1248				return TIPC_ERR_NO_PORT;
1249		}
1250	}
1251
1252	/* Reject message if there isn't room to queue it */
1253
1254	recv_q_len = (u32)atomic_read(&tipc_queue_size);
1255	if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1256		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1257			return TIPC_ERR_OVERLOAD;
1258	}
1259	recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1260	if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1261		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1262			return TIPC_ERR_OVERLOAD;
1263	}
1264
1265	/* Enqueue message (finally!) */
1266
1267	TIPC_SKB_CB(buf)->handle = 0;
1268	atomic_inc(&tipc_queue_size);
1269	__skb_queue_tail(&sk->sk_receive_queue, buf);
1270
1271	/* Initiate connection termination for an incoming 'FIN' */
1272
1273	if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1274		sock->state = SS_DISCONNECTING;
1275		tipc_disconnect_port(tipc_sk_port(sk));
1276	}
1277
1278	if (waitqueue_active(sk_sleep(sk)))
1279		wake_up_interruptible(sk_sleep(sk));
1280	return TIPC_OK;
1281}
1282
1283/**
1284 * backlog_rcv - handle incoming message from backlog queue
1285 * @sk: socket
1286 * @buf: message
1287 *
1288 * Caller must hold socket lock, but not port lock.
1289 *
1290 * Returns 0
1291 */
1292
1293static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1294{
1295	u32 res;
1296
1297	res = filter_rcv(sk, buf);
1298	if (res)
1299		tipc_reject_msg(buf, res);
1300	return 0;
1301}
1302
1303/**
1304 * dispatch - handle incoming message
1305 * @tport: TIPC port that received message
1306 * @buf: message
1307 *
1308 * Called with port lock already taken.
1309 *
1310 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1311 */
1312
1313static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1314{
1315	struct sock *sk = (struct sock *)tport->usr_handle;
1316	u32 res;
1317
1318	/*
1319	 * Process message if socket is unlocked; otherwise add to backlog queue
1320	 *
1321	 * This code is based on sk_receive_skb(), but must be distinct from it
1322	 * since a TIPC-specific filter/reject mechanism is utilized
1323	 */
1324
1325	bh_lock_sock(sk);
1326	if (!sock_owned_by_user(sk)) {
1327		res = filter_rcv(sk, buf);
1328	} else {
1329		if (sk_add_backlog(sk, buf))
1330			res = TIPC_ERR_OVERLOAD;
1331		else
1332			res = TIPC_OK;
1333	}
1334	bh_unlock_sock(sk);
1335
1336	return res;
1337}
1338
1339/**
1340 * wakeupdispatch - wake up port after congestion
1341 * @tport: port to wakeup
1342 *
1343 * Called with port lock already taken.
1344 */
1345
1346static void wakeupdispatch(struct tipc_port *tport)
1347{
1348	struct sock *sk = (struct sock *)tport->usr_handle;
1349
1350	if (waitqueue_active(sk_sleep(sk)))
1351		wake_up_interruptible(sk_sleep(sk));
1352}
1353
1354/**
1355 * connect - establish a connection to another TIPC port
1356 * @sock: socket structure
1357 * @dest: socket address for destination port
1358 * @destlen: size of socket address data structure
1359 * @flags: file-related flags associated with socket
1360 *
1361 * Returns 0 on success, errno otherwise
1362 */
1363
1364static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1365		   int flags)
1366{
1367	struct sock *sk = sock->sk;
1368	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1369	struct msghdr m = {NULL,};
1370	struct sk_buff *buf;
1371	struct tipc_msg *msg;
1372	long timeout;
1373	int res;
1374
1375	lock_sock(sk);
1376
1377	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1378
1379	if (sock->state == SS_READY) {
1380		res = -EOPNOTSUPP;
1381		goto exit;
1382	}
1383
1384	/* For now, TIPC does not support the non-blocking form of connect() */
1385
1386	if (flags & O_NONBLOCK) {
1387		res = -EOPNOTSUPP;
1388		goto exit;
1389	}
1390
1391	/* Issue Posix-compliant error code if socket is in the wrong state */
1392
1393	if (sock->state == SS_LISTENING) {
1394		res = -EOPNOTSUPP;
1395		goto exit;
1396	}
1397	if (sock->state == SS_CONNECTING) {
1398		res = -EALREADY;
1399		goto exit;
1400	}
1401	if (sock->state != SS_UNCONNECTED) {
1402		res = -EISCONN;
1403		goto exit;
1404	}
1405
1406	/*
1407	 * Reject connection attempt using multicast address
1408	 *
1409	 * Note: send_msg() validates the rest of the address fields,
1410	 *       so there's no need to do it here
1411	 */
1412
1413	if (dst->addrtype == TIPC_ADDR_MCAST) {
1414		res = -EINVAL;
1415		goto exit;
1416	}
1417
1418	/* Reject any messages already in receive queue (very unlikely) */
1419
1420	reject_rx_queue(sk);
1421
1422	/* Send a 'SYN-' to destination */
1423
1424	m.msg_name = dest;
1425	m.msg_namelen = destlen;
1426	res = send_msg(NULL, sock, &m, 0);
1427	if (res < 0)
1428		goto exit;
1429
1430	/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1431
1432	timeout = tipc_sk(sk)->conn_timeout;
1433	release_sock(sk);
1434	res = wait_event_interruptible_timeout(*sk_sleep(sk),
1435			(!skb_queue_empty(&sk->sk_receive_queue) ||
1436			(sock->state != SS_CONNECTING)),
1437			timeout ? timeout : MAX_SCHEDULE_TIMEOUT);
 
1438	lock_sock(sk);
1439
1440	if (res > 0) {
1441		buf = skb_peek(&sk->sk_receive_queue);
1442		if (buf != NULL) {
1443			msg = buf_msg(buf);
1444			res = auto_connect(sock, msg);
1445			if (!res) {
1446				if (!msg_data_sz(msg))
1447					advance_rx_queue(sk);
1448			}
1449		} else {
1450			if (sock->state == SS_CONNECTED)
1451				res = -EISCONN;
1452			else
1453				res = -ECONNREFUSED;
1454		}
1455	} else {
1456		if (res == 0)
1457			res = -ETIMEDOUT;
1458		else
1459			; /* leave "res" unchanged */
1460		sock->state = SS_DISCONNECTING;
1461	}
1462
1463exit:
1464	release_sock(sk);
1465	return res;
1466}
1467
1468/**
1469 * listen - allow socket to listen for incoming connections
1470 * @sock: socket structure
1471 * @len: (unused)
1472 *
1473 * Returns 0 on success, errno otherwise
1474 */
1475
1476static int listen(struct socket *sock, int len)
1477{
1478	struct sock *sk = sock->sk;
1479	int res;
1480
1481	lock_sock(sk);
1482
1483	if (sock->state == SS_READY)
1484		res = -EOPNOTSUPP;
1485	else if (sock->state != SS_UNCONNECTED)
1486		res = -EINVAL;
1487	else {
1488		sock->state = SS_LISTENING;
1489		res = 0;
1490	}
1491
1492	release_sock(sk);
1493	return res;
1494}
1495
1496/**
1497 * accept - wait for connection request
1498 * @sock: listening socket
1499 * @newsock: new socket that is to be connected
1500 * @flags: file-related flags associated with socket
1501 *
1502 * Returns 0 on success, errno otherwise
1503 */
1504
1505static int accept(struct socket *sock, struct socket *new_sock, int flags)
1506{
1507	struct sock *sk = sock->sk;
1508	struct sk_buff *buf;
1509	int res;
1510
1511	lock_sock(sk);
1512
1513	if (sock->state == SS_READY) {
1514		res = -EOPNOTSUPP;
1515		goto exit;
1516	}
1517	if (sock->state != SS_LISTENING) {
1518		res = -EINVAL;
1519		goto exit;
1520	}
1521
1522	while (skb_queue_empty(&sk->sk_receive_queue)) {
1523		if (flags & O_NONBLOCK) {
1524			res = -EWOULDBLOCK;
1525			goto exit;
1526		}
1527		release_sock(sk);
1528		res = wait_event_interruptible(*sk_sleep(sk),
1529				(!skb_queue_empty(&sk->sk_receive_queue)));
1530		lock_sock(sk);
1531		if (res)
1532			goto exit;
1533	}
1534
1535	buf = skb_peek(&sk->sk_receive_queue);
1536
1537	res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1538	if (!res) {
1539		struct sock *new_sk = new_sock->sk;
1540		struct tipc_sock *new_tsock = tipc_sk(new_sk);
1541		struct tipc_port *new_tport = new_tsock->p;
1542		u32 new_ref = new_tport->ref;
1543		struct tipc_msg *msg = buf_msg(buf);
1544
1545		lock_sock(new_sk);
1546
1547		/*
1548		 * Reject any stray messages received by new socket
1549		 * before the socket lock was taken (very, very unlikely)
1550		 */
1551
1552		reject_rx_queue(new_sk);
1553
1554		/* Connect new socket to it's peer */
1555
1556		new_tsock->peer_name.ref = msg_origport(msg);
1557		new_tsock->peer_name.node = msg_orignode(msg);
1558		tipc_connect2port(new_ref, &new_tsock->peer_name);
1559		new_sock->state = SS_CONNECTED;
1560
1561		tipc_set_portimportance(new_ref, msg_importance(msg));
1562		if (msg_named(msg)) {
1563			new_tport->conn_type = msg_nametype(msg);
1564			new_tport->conn_instance = msg_nameinst(msg);
1565		}
1566
1567		/*
1568		 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1569		 * Respond to 'SYN+' by queuing it on new socket.
1570		 */
1571
1572		if (!msg_data_sz(msg)) {
1573			struct msghdr m = {NULL,};
1574
1575			advance_rx_queue(sk);
1576			send_packet(NULL, new_sock, &m, 0);
1577		} else {
1578			__skb_dequeue(&sk->sk_receive_queue);
1579			__skb_queue_head(&new_sk->sk_receive_queue, buf);
1580		}
1581		release_sock(new_sk);
1582	}
1583exit:
1584	release_sock(sk);
1585	return res;
1586}
1587
1588/**
1589 * shutdown - shutdown socket connection
1590 * @sock: socket structure
1591 * @how: direction to close (must be SHUT_RDWR)
1592 *
1593 * Terminates connection (if necessary), then purges socket's receive queue.
1594 *
1595 * Returns 0 on success, errno otherwise
1596 */
1597
1598static int shutdown(struct socket *sock, int how)
1599{
1600	struct sock *sk = sock->sk;
1601	struct tipc_port *tport = tipc_sk_port(sk);
1602	struct sk_buff *buf;
1603	int res;
1604
1605	if (how != SHUT_RDWR)
1606		return -EINVAL;
1607
1608	lock_sock(sk);
1609
1610	switch (sock->state) {
1611	case SS_CONNECTING:
1612	case SS_CONNECTED:
1613
1614		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1615restart:
 
1616		buf = __skb_dequeue(&sk->sk_receive_queue);
1617		if (buf) {
1618			atomic_dec(&tipc_queue_size);
1619			if (TIPC_SKB_CB(buf)->handle != 0) {
1620				buf_discard(buf);
1621				goto restart;
1622			}
1623			tipc_disconnect(tport->ref);
1624			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1625		} else {
1626			tipc_shutdown(tport->ref);
1627		}
1628
1629		sock->state = SS_DISCONNECTING;
1630
1631		/* fall through */
1632
1633	case SS_DISCONNECTING:
1634
1635		/* Discard any unreceived messages; wake up sleeping tasks */
1636
1637		discard_rx_queue(sk);
1638		if (waitqueue_active(sk_sleep(sk)))
1639			wake_up_interruptible(sk_sleep(sk));
1640		res = 0;
1641		break;
1642
1643	default:
1644		res = -ENOTCONN;
1645	}
1646
1647	release_sock(sk);
1648	return res;
1649}
1650
1651/**
1652 * setsockopt - set socket option
1653 * @sock: socket structure
1654 * @lvl: option level
1655 * @opt: option identifier
1656 * @ov: pointer to new option value
1657 * @ol: length of option value
1658 *
1659 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1660 * (to ease compatibility).
1661 *
1662 * Returns 0 on success, errno otherwise
1663 */
1664
1665static int setsockopt(struct socket *sock,
1666		      int lvl, int opt, char __user *ov, unsigned int ol)
1667{
1668	struct sock *sk = sock->sk;
1669	struct tipc_port *tport = tipc_sk_port(sk);
1670	u32 value;
1671	int res;
1672
1673	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1674		return 0;
1675	if (lvl != SOL_TIPC)
1676		return -ENOPROTOOPT;
1677	if (ol < sizeof(value))
1678		return -EINVAL;
1679	res = get_user(value, (u32 __user *)ov);
1680	if (res)
1681		return res;
1682
1683	lock_sock(sk);
1684
1685	switch (opt) {
1686	case TIPC_IMPORTANCE:
1687		res = tipc_set_portimportance(tport->ref, value);
1688		break;
1689	case TIPC_SRC_DROPPABLE:
1690		if (sock->type != SOCK_STREAM)
1691			res = tipc_set_portunreliable(tport->ref, value);
1692		else
1693			res = -ENOPROTOOPT;
1694		break;
1695	case TIPC_DEST_DROPPABLE:
1696		res = tipc_set_portunreturnable(tport->ref, value);
1697		break;
1698	case TIPC_CONN_TIMEOUT:
1699		tipc_sk(sk)->conn_timeout = msecs_to_jiffies(value);
1700		/* no need to set "res", since already 0 at this point */
1701		break;
1702	default:
1703		res = -EINVAL;
1704	}
1705
1706	release_sock(sk);
1707
1708	return res;
1709}
1710
1711/**
1712 * getsockopt - get socket option
1713 * @sock: socket structure
1714 * @lvl: option level
1715 * @opt: option identifier
1716 * @ov: receptacle for option value
1717 * @ol: receptacle for length of option value
1718 *
1719 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1720 * (to ease compatibility).
1721 *
1722 * Returns 0 on success, errno otherwise
1723 */
1724
1725static int getsockopt(struct socket *sock,
1726		      int lvl, int opt, char __user *ov, int __user *ol)
1727{
1728	struct sock *sk = sock->sk;
1729	struct tipc_port *tport = tipc_sk_port(sk);
1730	int len;
1731	u32 value;
1732	int res;
1733
1734	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1735		return put_user(0, ol);
1736	if (lvl != SOL_TIPC)
1737		return -ENOPROTOOPT;
1738	res = get_user(len, ol);
1739	if (res)
1740		return res;
1741
1742	lock_sock(sk);
1743
1744	switch (opt) {
1745	case TIPC_IMPORTANCE:
1746		res = tipc_portimportance(tport->ref, &value);
1747		break;
1748	case TIPC_SRC_DROPPABLE:
1749		res = tipc_portunreliable(tport->ref, &value);
1750		break;
1751	case TIPC_DEST_DROPPABLE:
1752		res = tipc_portunreturnable(tport->ref, &value);
1753		break;
1754	case TIPC_CONN_TIMEOUT:
1755		value = jiffies_to_msecs(tipc_sk(sk)->conn_timeout);
1756		/* no need to set "res", since already 0 at this point */
1757		break;
1758	case TIPC_NODE_RECVQ_DEPTH:
1759		value = (u32)atomic_read(&tipc_queue_size);
1760		break;
1761	case TIPC_SOCK_RECVQ_DEPTH:
1762		value = skb_queue_len(&sk->sk_receive_queue);
1763		break;
1764	default:
1765		res = -EINVAL;
1766	}
1767
1768	release_sock(sk);
1769
1770	if (res)
1771		return res;	/* "get" failed */
1772
1773	if (len < sizeof(value))
1774		return -EINVAL;
1775
1776	if (copy_to_user(ov, &value, sizeof(value)))
1777		return -EFAULT;
1778
1779	return put_user(sizeof(value), ol);
1780}
1781
1782/**
1783 * Protocol switches for the various types of TIPC sockets
1784 */
1785
1786static const struct proto_ops msg_ops = {
1787	.owner		= THIS_MODULE,
1788	.family		= AF_TIPC,
1789	.release	= release,
1790	.bind		= bind,
1791	.connect	= connect,
1792	.socketpair	= sock_no_socketpair,
1793	.accept		= accept,
1794	.getname	= get_name,
1795	.poll		= poll,
1796	.ioctl		= sock_no_ioctl,
1797	.listen		= listen,
1798	.shutdown	= shutdown,
1799	.setsockopt	= setsockopt,
1800	.getsockopt	= getsockopt,
1801	.sendmsg	= send_msg,
1802	.recvmsg	= recv_msg,
1803	.mmap		= sock_no_mmap,
1804	.sendpage	= sock_no_sendpage
1805};
1806
1807static const struct proto_ops packet_ops = {
1808	.owner		= THIS_MODULE,
1809	.family		= AF_TIPC,
1810	.release	= release,
1811	.bind		= bind,
1812	.connect	= connect,
1813	.socketpair	= sock_no_socketpair,
1814	.accept		= accept,
1815	.getname	= get_name,
1816	.poll		= poll,
1817	.ioctl		= sock_no_ioctl,
1818	.listen		= listen,
1819	.shutdown	= shutdown,
1820	.setsockopt	= setsockopt,
1821	.getsockopt	= getsockopt,
1822	.sendmsg	= send_packet,
1823	.recvmsg	= recv_msg,
1824	.mmap		= sock_no_mmap,
1825	.sendpage	= sock_no_sendpage
1826};
1827
1828static const struct proto_ops stream_ops = {
1829	.owner		= THIS_MODULE,
1830	.family		= AF_TIPC,
1831	.release	= release,
1832	.bind		= bind,
1833	.connect	= connect,
1834	.socketpair	= sock_no_socketpair,
1835	.accept		= accept,
1836	.getname	= get_name,
1837	.poll		= poll,
1838	.ioctl		= sock_no_ioctl,
1839	.listen		= listen,
1840	.shutdown	= shutdown,
1841	.setsockopt	= setsockopt,
1842	.getsockopt	= getsockopt,
1843	.sendmsg	= send_stream,
1844	.recvmsg	= recv_stream,
1845	.mmap		= sock_no_mmap,
1846	.sendpage	= sock_no_sendpage
1847};
1848
1849static const struct net_proto_family tipc_family_ops = {
1850	.owner		= THIS_MODULE,
1851	.family		= AF_TIPC,
1852	.create		= tipc_create
1853};
1854
1855static struct proto tipc_proto = {
1856	.name		= "TIPC",
1857	.owner		= THIS_MODULE,
1858	.obj_size	= sizeof(struct tipc_sock)
1859};
1860
1861/**
1862 * tipc_socket_init - initialize TIPC socket interface
1863 *
1864 * Returns 0 on success, errno otherwise
1865 */
1866int tipc_socket_init(void)
1867{
1868	int res;
1869
1870	res = proto_register(&tipc_proto, 1);
1871	if (res) {
1872		err("Failed to register TIPC protocol type\n");
1873		goto out;
1874	}
1875
1876	res = sock_register(&tipc_family_ops);
1877	if (res) {
1878		err("Failed to register TIPC socket type\n");
1879		proto_unregister(&tipc_proto);
1880		goto out;
1881	}
1882
1883	sockets_enabled = 1;
1884 out:
1885	return res;
1886}
1887
1888/**
1889 * tipc_socket_stop - stop TIPC socket interface
1890 */
1891
1892void tipc_socket_stop(void)
1893{
1894	if (!sockets_enabled)
1895		return;
1896
1897	sockets_enabled = 0;
1898	sock_unregister(tipc_family_ops.family);
1899	proto_unregister(&tipc_proto);
1900}
1901