PageRenderTime 93ms CodeModel.GetById 3ms app.highlight 67ms RepoModel.GetById 1ms app.codeStats 2ms

/kernel/linux-source-2.6.32/net/tipc/socket.c

https://bitbucket.org/ChuloChumo/sctp_thesis
C | 1901 lines | 1149 code | 279 blank | 473 comment | 314 complexity | 461f3d189dcb750e77f528ec8ab462ad MD5 | raw file
   1/*
   2 * net/tipc/socket.c: TIPC socket API
   3 *
   4 * Copyright (c) 2001-2007, Ericsson AB
   5 * Copyright (c) 2004-2008, 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 <linux/module.h>
  38#include <linux/types.h>
  39#include <linux/net.h>
  40#include <linux/socket.h>
  41#include <linux/errno.h>
  42#include <linux/mm.h>
  43#include <linux/slab.h>
  44#include <linux/poll.h>
  45#include <linux/fcntl.h>
  46#include <asm/string.h>
  47#include <asm/atomic.h>
  48#include <net/sock.h>
  49
  50#include <linux/tipc.h>
  51#include <linux/tipc_config.h>
  52#include <net/tipc/tipc_msg.h>
  53#include <net/tipc/tipc_port.h>
  54
  55#include "core.h"
  56
  57#define SS_LISTENING	-1	/* socket is listening */
  58#define SS_READY	-2	/* socket is connectionless */
  59
  60#define OVERLOAD_LIMIT_BASE	5000
  61#define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
  62
  63struct tipc_sock {
  64	struct sock sk;
  65	struct tipc_port *p;
  66	struct tipc_portid peer_name;
  67};
  68
  69#define tipc_sk(sk) ((struct tipc_sock *)(sk))
  70#define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
  71
  72static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
  73static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
  74static void wakeupdispatch(struct tipc_port *tport);
  75
  76static const struct proto_ops packet_ops;
  77static const struct proto_ops stream_ops;
  78static const struct proto_ops msg_ops;
  79
  80static struct proto tipc_proto;
  81
  82static int sockets_enabled = 0;
  83
  84static atomic_t tipc_queue_size = ATOMIC_INIT(0);
  85
  86/*
  87 * Revised TIPC socket locking policy:
  88 *
  89 * Most socket operations take the standard socket lock when they start
  90 * and hold it until they finish (or until they need to sleep).  Acquiring
  91 * this lock grants the owner exclusive access to the fields of the socket
  92 * data structures, with the exception of the backlog queue.  A few socket
  93 * operations can be done without taking the socket lock because they only
  94 * read socket information that never changes during the life of the socket.
  95 *
  96 * Socket operations may acquire the lock for the associated TIPC port if they
  97 * need to perform an operation on the port.  If any routine needs to acquire
  98 * both the socket lock and the port lock it must take the socket lock first
  99 * to avoid the risk of deadlock.
 100 *
 101 * The dispatcher handling incoming messages cannot grab the socket lock in
 102 * the standard fashion, since invoked it runs at the BH level and cannot block.
 103 * Instead, it checks to see if the socket lock is currently owned by someone,
 104 * and either handles the message itself or adds it to the socket's backlog
 105 * queue; in the latter case the queued message is processed once the process
 106 * owning the socket lock releases it.
 107 *
 108 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
 109 * the problem of a blocked socket operation preventing any other operations
 110 * from occurring.  However, applications must be careful if they have
 111 * multiple threads trying to send (or receive) on the same socket, as these
 112 * operations might interfere with each other.  For example, doing a connect
 113 * and a receive at the same time might allow the receive to consume the
 114 * ACK message meant for the connect.  While additional work could be done
 115 * to try and overcome this, it doesn't seem to be worthwhile at the present.
 116 *
 117 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
 118 * that another operation that must be performed in a non-blocking manner is
 119 * not delayed for very long because the lock has already been taken.
 120 *
 121 * NOTE: This code assumes that certain fields of a port/socket pair are
 122 * constant over its lifetime; such fields can be examined without taking
 123 * the socket lock and/or port lock, and do not need to be re-read even
 124 * after resuming processing after waiting.  These fields include:
 125 *   - socket type
 126 *   - pointer to socket sk structure (aka tipc_sock structure)
 127 *   - pointer to port structure
 128 *   - port reference
 129 */
 130
 131/**
 132 * advance_rx_queue - discard first buffer in socket receive queue
 133 *
 134 * Caller must hold socket lock
 135 */
 136
 137static void advance_rx_queue(struct sock *sk)
 138{
 139	buf_discard(__skb_dequeue(&sk->sk_receive_queue));
 140	atomic_dec(&tipc_queue_size);
 141}
 142
 143/**
 144 * discard_rx_queue - discard all buffers in socket receive queue
 145 *
 146 * Caller must hold socket lock
 147 */
 148
 149static void discard_rx_queue(struct sock *sk)
 150{
 151	struct sk_buff *buf;
 152
 153	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
 154		atomic_dec(&tipc_queue_size);
 155		buf_discard(buf);
 156	}
 157}
 158
 159/**
 160 * reject_rx_queue - reject all buffers in socket receive queue
 161 *
 162 * Caller must hold socket lock
 163 */
 164
 165static void reject_rx_queue(struct sock *sk)
 166{
 167	struct sk_buff *buf;
 168
 169	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
 170		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
 171		atomic_dec(&tipc_queue_size);
 172	}
 173}
 174
 175/**
 176 * tipc_create - create a TIPC socket
 177 * @net: network namespace (must be default network)
 178 * @sock: pre-allocated socket structure
 179 * @protocol: protocol indicator (must be 0)
 180 *
 181 * This routine creates additional data structures used by the TIPC socket,
 182 * initializes them, and links them together.
 183 *
 184 * Returns 0 on success, errno otherwise
 185 */
 186
 187static int tipc_create(struct net *net, struct socket *sock, int protocol)
 188{
 189	const struct proto_ops *ops;
 190	socket_state state;
 191	struct sock *sk;
 192	struct tipc_port *tp_ptr;
 193
 194	/* Validate arguments */
 195
 196	if (net != &init_net)
 197		return -EAFNOSUPPORT;
 198
 199	if (unlikely(protocol != 0))
 200		return -EPROTONOSUPPORT;
 201
 202	switch (sock->type) {
 203	case SOCK_STREAM:
 204		ops = &stream_ops;
 205		state = SS_UNCONNECTED;
 206		break;
 207	case SOCK_SEQPACKET:
 208		ops = &packet_ops;
 209		state = SS_UNCONNECTED;
 210		break;
 211	case SOCK_DGRAM:
 212	case SOCK_RDM:
 213		ops = &msg_ops;
 214		state = SS_READY;
 215		break;
 216	default:
 217		return -EPROTOTYPE;
 218	}
 219
 220	/* Allocate socket's protocol area */
 221
 222	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
 223	if (sk == NULL)
 224		return -ENOMEM;
 225
 226	/* Allocate TIPC port for socket to use */
 227
 228	tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
 229				     TIPC_LOW_IMPORTANCE);
 230	if (unlikely(!tp_ptr)) {
 231		sk_free(sk);
 232		return -ENOMEM;
 233	}
 234
 235	/* Finish initializing socket data structures */
 236
 237	sock->ops = ops;
 238	sock->state = state;
 239
 240	sock_init_data(sock, sk);
 241	sk->sk_rcvtimeo = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
 242	sk->sk_backlog_rcv = backlog_rcv;
 243	tipc_sk(sk)->p = tp_ptr;
 244
 245	spin_unlock_bh(tp_ptr->lock);
 246
 247	if (sock->state == SS_READY) {
 248		tipc_set_portunreturnable(tp_ptr->ref, 1);
 249		if (sock->type == SOCK_DGRAM)
 250			tipc_set_portunreliable(tp_ptr->ref, 1);
 251	}
 252
 253	atomic_inc(&tipc_user_count);
 254	return 0;
 255}
 256
 257/**
 258 * release - destroy a TIPC socket
 259 * @sock: socket to destroy
 260 *
 261 * This routine cleans up any messages that are still queued on the socket.
 262 * For DGRAM and RDM socket types, all queued messages are rejected.
 263 * For SEQPACKET and STREAM socket types, the first message is rejected
 264 * and any others are discarded.  (If the first message on a STREAM socket
 265 * is partially-read, it is discarded and the next one is rejected instead.)
 266 *
 267 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 268 * are returned or discarded according to the "destination droppable" setting
 269 * specified for the message by the sender.
 270 *
 271 * Returns 0 on success, errno otherwise
 272 */
 273
 274static int release(struct socket *sock)
 275{
 276	struct sock *sk = sock->sk;
 277	struct tipc_port *tport;
 278	struct sk_buff *buf;
 279	int res;
 280
 281	/*
 282	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 283	 * releases a pre-allocated child socket that was never used)
 284	 */
 285
 286	if (sk == NULL)
 287		return 0;
 288
 289	tport = tipc_sk_port(sk);
 290	lock_sock(sk);
 291
 292	/*
 293	 * Reject all unreceived messages, except on an active connection
 294	 * (which disconnects locally & sends a 'FIN+' to peer)
 295	 */
 296
 297	while (sock->state != SS_DISCONNECTING) {
 298		buf = __skb_dequeue(&sk->sk_receive_queue);
 299		if (buf == NULL)
 300			break;
 301		atomic_dec(&tipc_queue_size);
 302		if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf)))
 303			buf_discard(buf);
 304		else {
 305			if ((sock->state == SS_CONNECTING) ||
 306			    (sock->state == SS_CONNECTED)) {
 307				sock->state = SS_DISCONNECTING;
 308				tipc_disconnect(tport->ref);
 309			}
 310			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
 311		}
 312	}
 313
 314	/*
 315	 * Delete TIPC port; this ensures no more messages are queued
 316	 * (also disconnects an active connection & sends a 'FIN-' to peer)
 317	 */
 318
 319	res = tipc_deleteport(tport->ref);
 320
 321	/* Discard any remaining (connection-based) messages in receive queue */
 322
 323	discard_rx_queue(sk);
 324
 325	/* Reject any messages that accumulated in backlog queue */
 326
 327	sock->state = SS_DISCONNECTING;
 328	release_sock(sk);
 329
 330	sock_put(sk);
 331	sock->sk = NULL;
 332
 333	atomic_dec(&tipc_user_count);
 334	return res;
 335}
 336
 337/**
 338 * bind - associate or disassocate TIPC name(s) with a socket
 339 * @sock: socket structure
 340 * @uaddr: socket address describing name(s) and desired operation
 341 * @uaddr_len: size of socket address data structure
 342 *
 343 * Name and name sequence binding is indicated using a positive scope value;
 344 * a negative scope value unbinds the specified name.  Specifying no name
 345 * (i.e. a socket address length of 0) unbinds all names from the socket.
 346 *
 347 * Returns 0 on success, errno otherwise
 348 *
 349 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 350 *       access any non-constant socket information.
 351 */
 352
 353static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
 354{
 355	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 356	u32 portref = tipc_sk_port(sock->sk)->ref;
 357
 358	if (unlikely(!uaddr_len))
 359		return tipc_withdraw(portref, 0, NULL);
 360
 361	if (uaddr_len < sizeof(struct sockaddr_tipc))
 362		return -EINVAL;
 363	if (addr->family != AF_TIPC)
 364		return -EAFNOSUPPORT;
 365
 366	if (addr->addrtype == TIPC_ADDR_NAME)
 367		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
 368	else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
 369		return -EAFNOSUPPORT;
 370
 371	return (addr->scope > 0) ?
 372		tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
 373		tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
 374}
 375
 376/**
 377 * get_name - get port ID of socket or peer socket
 378 * @sock: socket structure
 379 * @uaddr: area for returned socket address
 380 * @uaddr_len: area for returned length of socket address
 381 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 382 *
 383 * Returns 0 on success, errno otherwise
 384 *
 385 * NOTE: This routine doesn't need to take the socket lock since it only
 386 *       accesses socket information that is unchanging (or which changes in
 387 * 	 a completely predictable manner).
 388 */
 389
 390static int get_name(struct socket *sock, struct sockaddr *uaddr,
 391		    int *uaddr_len, int peer)
 392{
 393	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 394	struct tipc_sock *tsock = tipc_sk(sock->sk);
 395
 396	memset(addr, 0, sizeof(*addr));
 397	if (peer) {
 398		if ((sock->state != SS_CONNECTED) &&
 399			((peer != 2) || (sock->state != SS_DISCONNECTING)))
 400			return -ENOTCONN;
 401		addr->addr.id.ref = tsock->peer_name.ref;
 402		addr->addr.id.node = tsock->peer_name.node;
 403	} else {
 404		tipc_ownidentity(tsock->p->ref, &addr->addr.id);
 405	}
 406
 407	*uaddr_len = sizeof(*addr);
 408	addr->addrtype = TIPC_ADDR_ID;
 409	addr->family = AF_TIPC;
 410	addr->scope = 0;
 411	addr->addr.name.domain = 0;
 412
 413	return 0;
 414}
 415
 416/**
 417 * poll - read and possibly block on pollmask
 418 * @file: file structure associated with the socket
 419 * @sock: socket for which to calculate the poll bits
 420 * @wait: ???
 421 *
 422 * Returns pollmask value
 423 *
 424 * COMMENTARY:
 425 * It appears that the usual socket locking mechanisms are not useful here
 426 * since the pollmask info is potentially out-of-date the moment this routine
 427 * exits.  TCP and other protocols seem to rely on higher level poll routines
 428 * to handle any preventable race conditions, so TIPC will do the same ...
 429 *
 430 * TIPC sets the returned events as follows:
 431 * a) POLLRDNORM and POLLIN are set if the socket's receive queue is non-empty
 432 *    or if a connection-oriented socket is does not have an active connection
 433 *    (i.e. a read operation will not block).
 434 * b) POLLOUT is set except when a socket's connection has been terminated
 435 *    (i.e. a write operation will not block).
 436 * c) POLLHUP is set when a socket's connection has been terminated.
 437 *
 438 * IMPORTANT: The fact that a read or write operation will not block does NOT
 439 * imply that the operation will succeed!
 440 */
 441
 442static unsigned int poll(struct file *file, struct socket *sock,
 443			 poll_table *wait)
 444{
 445	struct sock *sk = sock->sk;
 446	u32 mask;
 447
 448	poll_wait(file, sk->sk_sleep, wait);
 449
 450	if (!skb_queue_empty(&sk->sk_receive_queue) ||
 451	    (sock->state == SS_UNCONNECTED) ||
 452	    (sock->state == SS_DISCONNECTING))
 453		mask = (POLLRDNORM | POLLIN);
 454	else
 455		mask = 0;
 456
 457	if (sock->state == SS_DISCONNECTING)
 458		mask |= POLLHUP;
 459	else
 460		mask |= POLLOUT;
 461
 462	return mask;
 463}
 464
 465/**
 466 * dest_name_check - verify user is permitted to send to specified port name
 467 * @dest: destination address
 468 * @m: descriptor for message to be sent
 469 *
 470 * Prevents restricted configuration commands from being issued by
 471 * unauthorized users.
 472 *
 473 * Returns 0 if permission is granted, otherwise errno
 474 */
 475
 476static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
 477{
 478	struct tipc_cfg_msg_hdr hdr;
 479
 480	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
 481		return 0;
 482	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
 483		return 0;
 484	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
 485		return -EACCES;
 486
 487	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
 488		return -EFAULT;
 489	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
 490		return -EACCES;
 491
 492	return 0;
 493}
 494
 495/**
 496 * send_msg - send message in connectionless manner
 497 * @iocb: if NULL, indicates that socket lock is already held
 498 * @sock: socket structure
 499 * @m: message to send
 500 * @total_len: length of message
 501 *
 502 * Message must have an destination specified explicitly.
 503 * Used for SOCK_RDM and SOCK_DGRAM messages,
 504 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
 505 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
 506 *
 507 * Returns the number of bytes sent on success, or errno otherwise
 508 */
 509
 510static int send_msg(struct kiocb *iocb, struct socket *sock,
 511		    struct msghdr *m, size_t total_len)
 512{
 513	struct sock *sk = sock->sk;
 514	struct tipc_port *tport = tipc_sk_port(sk);
 515	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
 516	int needs_conn;
 517	int res = -EINVAL;
 518
 519	if (unlikely(!dest))
 520		return -EDESTADDRREQ;
 521	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
 522		     (dest->family != AF_TIPC)))
 523		return -EINVAL;
 524
 525	if (iocb)
 526		lock_sock(sk);
 527
 528	needs_conn = (sock->state != SS_READY);
 529	if (unlikely(needs_conn)) {
 530		if (sock->state == SS_LISTENING) {
 531			res = -EPIPE;
 532			goto exit;
 533		}
 534		if (sock->state != SS_UNCONNECTED) {
 535			res = -EISCONN;
 536			goto exit;
 537		}
 538		if ((tport->published) ||
 539		    ((sock->type == SOCK_STREAM) && (total_len != 0))) {
 540			res = -EOPNOTSUPP;
 541			goto exit;
 542		}
 543		if (dest->addrtype == TIPC_ADDR_NAME) {
 544			tport->conn_type = dest->addr.name.name.type;
 545			tport->conn_instance = dest->addr.name.name.instance;
 546		}
 547
 548		/* Abort any pending connection attempts (very unlikely) */
 549
 550		reject_rx_queue(sk);
 551	}
 552
 553	do {
 554		if (dest->addrtype == TIPC_ADDR_NAME) {
 555			if ((res = dest_name_check(dest, m)))
 556				break;
 557			res = tipc_send2name(tport->ref,
 558					     &dest->addr.name.name,
 559					     dest->addr.name.domain,
 560					     m->msg_iovlen,
 561					     m->msg_iov);
 562		}
 563		else if (dest->addrtype == TIPC_ADDR_ID) {
 564			res = tipc_send2port(tport->ref,
 565					     &dest->addr.id,
 566					     m->msg_iovlen,
 567					     m->msg_iov);
 568		}
 569		else if (dest->addrtype == TIPC_ADDR_MCAST) {
 570			if (needs_conn) {
 571				res = -EOPNOTSUPP;
 572				break;
 573			}
 574			if ((res = dest_name_check(dest, m)))
 575				break;
 576			res = tipc_multicast(tport->ref,
 577					     &dest->addr.nameseq,
 578					     0,
 579					     m->msg_iovlen,
 580					     m->msg_iov);
 581		}
 582		if (likely(res != -ELINKCONG)) {
 583			if (needs_conn && (res >= 0)) {
 584				sock->state = SS_CONNECTING;
 585			}
 586			break;
 587		}
 588		if (m->msg_flags & MSG_DONTWAIT) {
 589			res = -EWOULDBLOCK;
 590			break;
 591		}
 592		release_sock(sk);
 593		res = wait_event_interruptible(*sk->sk_sleep,
 594					       !tport->congested);
 595		lock_sock(sk);
 596		if (res)
 597			break;
 598	} while (1);
 599
 600exit:
 601	if (iocb)
 602		release_sock(sk);
 603	return res;
 604}
 605
 606/**
 607 * send_packet - send a connection-oriented message
 608 * @iocb: if NULL, indicates that socket lock is already held
 609 * @sock: socket structure
 610 * @m: message to send
 611 * @total_len: length of message
 612 *
 613 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
 614 *
 615 * Returns the number of bytes sent on success, or errno otherwise
 616 */
 617
 618static int send_packet(struct kiocb *iocb, struct socket *sock,
 619		       struct msghdr *m, size_t total_len)
 620{
 621	struct sock *sk = sock->sk;
 622	struct tipc_port *tport = tipc_sk_port(sk);
 623	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
 624	int res;
 625
 626	/* Handle implied connection establishment */
 627
 628	if (unlikely(dest))
 629		return send_msg(iocb, sock, m, total_len);
 630
 631	if (iocb)
 632		lock_sock(sk);
 633
 634	do {
 635		if (unlikely(sock->state != SS_CONNECTED)) {
 636			if (sock->state == SS_DISCONNECTING)
 637				res = -EPIPE;
 638			else
 639				res = -ENOTCONN;
 640			break;
 641		}
 642
 643		res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov);
 644		if (likely(res != -ELINKCONG)) {
 645			break;
 646		}
 647		if (m->msg_flags & MSG_DONTWAIT) {
 648			res = -EWOULDBLOCK;
 649			break;
 650		}
 651		release_sock(sk);
 652		res = wait_event_interruptible(*sk->sk_sleep,
 653			(!tport->congested || !tport->connected));
 654		lock_sock(sk);
 655		if (res)
 656			break;
 657	} while (1);
 658
 659	if (iocb)
 660		release_sock(sk);
 661	return res;
 662}
 663
 664/**
 665 * send_stream - send stream-oriented data
 666 * @iocb: (unused)
 667 * @sock: socket structure
 668 * @m: data to send
 669 * @total_len: total length of data to be sent
 670 *
 671 * Used for SOCK_STREAM data.
 672 *
 673 * Returns the number of bytes sent on success (or partial success),
 674 * or errno if no data sent
 675 */
 676
 677static int send_stream(struct kiocb *iocb, struct socket *sock,
 678		       struct msghdr *m, size_t total_len)
 679{
 680	struct sock *sk = sock->sk;
 681	struct tipc_port *tport = tipc_sk_port(sk);
 682	struct msghdr my_msg;
 683	struct iovec my_iov;
 684	struct iovec *curr_iov;
 685	int curr_iovlen;
 686	char __user *curr_start;
 687	u32 hdr_size;
 688	int curr_left;
 689	int bytes_to_send;
 690	int bytes_sent;
 691	int res;
 692
 693	lock_sock(sk);
 694
 695	/* Handle special cases where there is no connection */
 696
 697	if (unlikely(sock->state != SS_CONNECTED)) {
 698		if (sock->state == SS_UNCONNECTED) {
 699			res = send_packet(NULL, sock, m, total_len);
 700			goto exit;
 701		} else if (sock->state == SS_DISCONNECTING) {
 702			res = -EPIPE;
 703			goto exit;
 704		} else {
 705			res = -ENOTCONN;
 706			goto exit;
 707		}
 708	}
 709
 710	if (unlikely(m->msg_name)) {
 711		res = -EISCONN;
 712		goto exit;
 713	}
 714
 715	/*
 716	 * Send each iovec entry using one or more messages
 717	 *
 718	 * Note: This algorithm is good for the most likely case
 719	 * (i.e. one large iovec entry), but could be improved to pass sets
 720	 * of small iovec entries into send_packet().
 721	 */
 722
 723	curr_iov = m->msg_iov;
 724	curr_iovlen = m->msg_iovlen;
 725	my_msg.msg_iov = &my_iov;
 726	my_msg.msg_iovlen = 1;
 727	my_msg.msg_flags = m->msg_flags;
 728	my_msg.msg_name = NULL;
 729	bytes_sent = 0;
 730
 731	hdr_size = msg_hdr_sz(&tport->phdr);
 732
 733	while (curr_iovlen--) {
 734		curr_start = curr_iov->iov_base;
 735		curr_left = curr_iov->iov_len;
 736
 737		while (curr_left) {
 738			bytes_to_send = tport->max_pkt - hdr_size;
 739			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
 740				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
 741			if (curr_left < bytes_to_send)
 742				bytes_to_send = curr_left;
 743			my_iov.iov_base = curr_start;
 744			my_iov.iov_len = bytes_to_send;
 745			if ((res = send_packet(NULL, sock, &my_msg, 0)) < 0) {
 746				if (bytes_sent)
 747					res = bytes_sent;
 748				goto exit;
 749			}
 750			curr_left -= bytes_to_send;
 751			curr_start += bytes_to_send;
 752			bytes_sent += bytes_to_send;
 753		}
 754
 755		curr_iov++;
 756	}
 757	res = bytes_sent;
 758exit:
 759	release_sock(sk);
 760	return res;
 761}
 762
 763/**
 764 * auto_connect - complete connection setup to a remote port
 765 * @sock: socket structure
 766 * @msg: peer's response message
 767 *
 768 * Returns 0 on success, errno otherwise
 769 */
 770
 771static int auto_connect(struct socket *sock, struct tipc_msg *msg)
 772{
 773	struct tipc_sock *tsock = tipc_sk(sock->sk);
 774
 775	if (msg_errcode(msg)) {
 776		sock->state = SS_DISCONNECTING;
 777		return -ECONNREFUSED;
 778	}
 779
 780	tsock->peer_name.ref = msg_origport(msg);
 781	tsock->peer_name.node = msg_orignode(msg);
 782	tipc_connect2port(tsock->p->ref, &tsock->peer_name);
 783	tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
 784	sock->state = SS_CONNECTED;
 785	return 0;
 786}
 787
 788/**
 789 * set_orig_addr - capture sender's address for received message
 790 * @m: descriptor for message info
 791 * @msg: received message header
 792 *
 793 * Note: Address is not captured if not requested by receiver.
 794 */
 795
 796static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
 797{
 798	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
 799
 800	if (addr) {
 801		addr->family = AF_TIPC;
 802		addr->addrtype = TIPC_ADDR_ID;
 803		addr->addr.id.ref = msg_origport(msg);
 804		addr->addr.id.node = msg_orignode(msg);
 805		addr->addr.name.domain = 0;   	/* could leave uninitialized */
 806		addr->scope = 0;   		/* could leave uninitialized */
 807		m->msg_namelen = sizeof(struct sockaddr_tipc);
 808	}
 809}
 810
 811/**
 812 * anc_data_recv - optionally capture ancillary data for received message
 813 * @m: descriptor for message info
 814 * @msg: received message header
 815 * @tport: TIPC port associated with message
 816 *
 817 * Note: Ancillary data is not captured if not requested by receiver.
 818 *
 819 * Returns 0 if successful, otherwise errno
 820 */
 821
 822static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
 823				struct tipc_port *tport)
 824{
 825	u32 anc_data[3];
 826	u32 err;
 827	u32 dest_type;
 828	int has_name;
 829	int res;
 830
 831	if (likely(m->msg_controllen == 0))
 832		return 0;
 833
 834	/* Optionally capture errored message object(s) */
 835
 836	err = msg ? msg_errcode(msg) : 0;
 837	if (unlikely(err)) {
 838		anc_data[0] = err;
 839		anc_data[1] = msg_data_sz(msg);
 840		if ((res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data)))
 841			return res;
 842		if (anc_data[1] &&
 843		    (res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
 844				    msg_data(msg))))
 845			return res;
 846	}
 847
 848	/* Optionally capture message destination object */
 849
 850	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
 851	switch (dest_type) {
 852	case TIPC_NAMED_MSG:
 853		has_name = 1;
 854		anc_data[0] = msg_nametype(msg);
 855		anc_data[1] = msg_namelower(msg);
 856		anc_data[2] = msg_namelower(msg);
 857		break;
 858	case TIPC_MCAST_MSG:
 859		has_name = 1;
 860		anc_data[0] = msg_nametype(msg);
 861		anc_data[1] = msg_namelower(msg);
 862		anc_data[2] = msg_nameupper(msg);
 863		break;
 864	case TIPC_CONN_MSG:
 865		has_name = (tport->conn_type != 0);
 866		anc_data[0] = tport->conn_type;
 867		anc_data[1] = tport->conn_instance;
 868		anc_data[2] = tport->conn_instance;
 869		break;
 870	default:
 871		has_name = 0;
 872	}
 873	if (has_name &&
 874	    (res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data)))
 875		return res;
 876
 877	return 0;
 878}
 879
 880/**
 881 * recv_msg - receive packet-oriented message
 882 * @iocb: (unused)
 883 * @m: descriptor for message info
 884 * @buf_len: total size of user buffer area
 885 * @flags: receive flags
 886 *
 887 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
 888 * If the complete message doesn't fit in user area, truncate it.
 889 *
 890 * Returns size of returned message data, errno otherwise
 891 */
 892
 893static int recv_msg(struct kiocb *iocb, struct socket *sock,
 894		    struct msghdr *m, size_t buf_len, int flags)
 895{
 896	struct sock *sk = sock->sk;
 897	struct tipc_port *tport = tipc_sk_port(sk);
 898	struct sk_buff *buf;
 899	struct tipc_msg *msg;
 900	unsigned int sz;
 901	u32 err;
 902	int res;
 903
 904	/* Catch invalid receive requests */
 905
 906	if (m->msg_iovlen != 1)
 907		return -EOPNOTSUPP;   /* Don't do multiple iovec entries yet */
 908
 909	if (unlikely(!buf_len))
 910		return -EINVAL;
 911
 912	lock_sock(sk);
 913
 914	if (unlikely(sock->state == SS_UNCONNECTED)) {
 915		res = -ENOTCONN;
 916		goto exit;
 917	}
 918
 919restart:
 920
 921	/* Look for a message in receive queue; wait if necessary */
 922
 923	while (skb_queue_empty(&sk->sk_receive_queue)) {
 924		if (sock->state == SS_DISCONNECTING) {
 925			res = -ENOTCONN;
 926			goto exit;
 927		}
 928		if (flags & MSG_DONTWAIT) {
 929			res = -EWOULDBLOCK;
 930			goto exit;
 931		}
 932		release_sock(sk);
 933		res = wait_event_interruptible(*sk->sk_sleep,
 934			(!skb_queue_empty(&sk->sk_receive_queue) ||
 935			 (sock->state == SS_DISCONNECTING)));
 936		lock_sock(sk);
 937		if (res)
 938			goto exit;
 939	}
 940
 941	/* Look at first message in receive queue */
 942
 943	buf = skb_peek(&sk->sk_receive_queue);
 944	msg = buf_msg(buf);
 945	sz = msg_data_sz(msg);
 946	err = msg_errcode(msg);
 947
 948	/* Complete connection setup for an implied connect */
 949
 950	if (unlikely(sock->state == SS_CONNECTING)) {
 951		res = auto_connect(sock, msg);
 952		if (res)
 953			goto exit;
 954	}
 955
 956	/* Discard an empty non-errored message & try again */
 957
 958	if ((!sz) && (!err)) {
 959		advance_rx_queue(sk);
 960		goto restart;
 961	}
 962
 963	/* Capture sender's address (optional) */
 964
 965	set_orig_addr(m, msg);
 966
 967	/* Capture ancillary data (optional) */
 968
 969	res = anc_data_recv(m, msg, tport);
 970	if (res)
 971		goto exit;
 972
 973	/* Capture message data (if valid) & compute return value (always) */
 974
 975	if (!err) {
 976		if (unlikely(buf_len < sz)) {
 977			sz = buf_len;
 978			m->msg_flags |= MSG_TRUNC;
 979		}
 980		if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg),
 981					  sz))) {
 982			res = -EFAULT;
 983			goto exit;
 984		}
 985		res = sz;
 986	} else {
 987		if ((sock->state == SS_READY) ||
 988		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
 989			res = 0;
 990		else
 991			res = -ECONNRESET;
 992	}
 993
 994	/* Consume received message (optional) */
 995
 996	if (likely(!(flags & MSG_PEEK))) {
 997		if ((sock->state != SS_READY) &&
 998		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
 999			tipc_acknowledge(tport->ref, tport->conn_unacked);
1000		advance_rx_queue(sk);
1001	}
1002exit:
1003	release_sock(sk);
1004	return res;
1005}
1006
1007/**
1008 * recv_stream - receive stream-oriented data
1009 * @iocb: (unused)
1010 * @m: descriptor for message info
1011 * @buf_len: total size of user buffer area
1012 * @flags: receive flags
1013 *
1014 * Used for SOCK_STREAM messages only.  If not enough data is available
1015 * will optionally wait for more; never truncates data.
1016 *
1017 * Returns size of returned message data, errno otherwise
1018 */
1019
1020static int recv_stream(struct kiocb *iocb, struct socket *sock,
1021		       struct msghdr *m, size_t buf_len, int flags)
1022{
1023	struct sock *sk = sock->sk;
1024	struct tipc_port *tport = tipc_sk_port(sk);
1025	struct sk_buff *buf;
1026	struct tipc_msg *msg;
1027	unsigned int sz;
1028	int sz_to_copy;
1029	int sz_copied = 0;
1030	int needed;
1031	char __user *crs = m->msg_iov->iov_base;
1032	unsigned char *buf_crs;
1033	u32 err;
1034	int res = 0;
1035
1036	/* Catch invalid receive attempts */
1037
1038	if (m->msg_iovlen != 1)
1039		return -EOPNOTSUPP;   /* Don't do multiple iovec entries yet */
1040
1041	if (unlikely(!buf_len))
1042		return -EINVAL;
1043
1044	lock_sock(sk);
1045
1046	if (unlikely((sock->state == SS_UNCONNECTED) ||
1047		     (sock->state == SS_CONNECTING))) {
1048		res = -ENOTCONN;
1049		goto exit;
1050	}
1051
1052restart:
1053
1054	/* Look for a message in receive queue; wait if necessary */
1055
1056	while (skb_queue_empty(&sk->sk_receive_queue)) {
1057		if (sock->state == SS_DISCONNECTING) {
1058			res = -ENOTCONN;
1059			goto exit;
1060		}
1061		if (flags & MSG_DONTWAIT) {
1062			res = -EWOULDBLOCK;
1063			goto exit;
1064		}
1065		release_sock(sk);
1066		res = wait_event_interruptible(*sk->sk_sleep,
1067			(!skb_queue_empty(&sk->sk_receive_queue) ||
1068			 (sock->state == SS_DISCONNECTING)));
1069		lock_sock(sk);
1070		if (res)
1071			goto exit;
1072	}
1073
1074	/* Look at first message in receive queue */
1075
1076	buf = skb_peek(&sk->sk_receive_queue);
1077	msg = buf_msg(buf);
1078	sz = msg_data_sz(msg);
1079	err = msg_errcode(msg);
1080
1081	/* Discard an empty non-errored message & try again */
1082
1083	if ((!sz) && (!err)) {
1084		advance_rx_queue(sk);
1085		goto restart;
1086	}
1087
1088	/* Optionally capture sender's address & ancillary data of first msg */
1089
1090	if (sz_copied == 0) {
1091		set_orig_addr(m, msg);
1092		res = anc_data_recv(m, msg, tport);
1093		if (res)
1094			goto exit;
1095	}
1096
1097	/* Capture message data (if valid) & compute return value (always) */
1098
1099	if (!err) {
1100		buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle);
1101		sz = (unsigned char *)msg + msg_size(msg) - buf_crs;
1102
1103		needed = (buf_len - sz_copied);
1104		sz_to_copy = (sz <= needed) ? sz : needed;
1105		if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) {
1106			res = -EFAULT;
1107			goto exit;
1108		}
1109		sz_copied += sz_to_copy;
1110
1111		if (sz_to_copy < sz) {
1112			if (!(flags & MSG_PEEK))
1113				TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy;
1114			goto exit;
1115		}
1116
1117		crs += sz_to_copy;
1118	} else {
1119		if (sz_copied != 0)
1120			goto exit; /* can't add error msg to valid data */
1121
1122		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1123			res = 0;
1124		else
1125			res = -ECONNRESET;
1126	}
1127
1128	/* Consume received message (optional) */
1129
1130	if (likely(!(flags & MSG_PEEK))) {
1131		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1132			tipc_acknowledge(tport->ref, tport->conn_unacked);
1133		advance_rx_queue(sk);
1134	}
1135
1136	/* Loop around if more data is required */
1137
1138	if ((sz_copied < buf_len)    /* didn't get all requested data */
1139	    && (!skb_queue_empty(&sk->sk_receive_queue) ||
1140		(flags & MSG_WAITALL))
1141				     /* ... and more is ready or required */
1142	    && (!(flags & MSG_PEEK)) /* ... and aren't just peeking at data */
1143	    && (!err)                /* ... and haven't reached a FIN */
1144	    )
1145		goto restart;
1146
1147exit:
1148	release_sock(sk);
1149	return sz_copied ? sz_copied : res;
1150}
1151
1152/**
1153 * rx_queue_full - determine if receive queue can accept another message
1154 * @msg: message to be added to queue
1155 * @queue_size: current size of queue
1156 * @base: nominal maximum size of queue
1157 *
1158 * Returns 1 if queue is unable to accept message, 0 otherwise
1159 */
1160
1161static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1162{
1163	u32 threshold;
1164	u32 imp = msg_importance(msg);
1165
1166	if (imp == TIPC_LOW_IMPORTANCE)
1167		threshold = base;
1168	else if (imp == TIPC_MEDIUM_IMPORTANCE)
1169		threshold = base * 2;
1170	else if (imp == TIPC_HIGH_IMPORTANCE)
1171		threshold = base * 100;
1172	else
1173		return 0;
1174
1175	if (msg_connected(msg))
1176		threshold *= 4;
1177
1178	return (queue_size >= threshold);
1179}
1180
1181/**
1182 * filter_rcv - validate incoming message
1183 * @sk: socket
1184 * @buf: message
1185 *
1186 * Enqueues message on receive queue if acceptable; optionally handles
1187 * disconnect indication for a connected socket.
1188 *
1189 * Called with socket lock already taken; port lock may also be taken.
1190 *
1191 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1192 */
1193
1194static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1195{
1196	struct socket *sock = sk->sk_socket;
1197	struct tipc_msg *msg = buf_msg(buf);
1198	u32 recv_q_len;
1199
1200	/* Reject message if it is wrong sort of message for socket */
1201
1202	/*
1203	 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1204	 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1205	 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1206	 */
1207
1208	if (sock->state == SS_READY) {
1209		if (msg_connected(msg)) {
1210			msg_dbg(msg, "dispatch filter 1\n");
1211			return TIPC_ERR_NO_PORT;
1212		}
1213	} else {
1214		if (msg_mcast(msg)) {
1215			msg_dbg(msg, "dispatch filter 2\n");
1216			return TIPC_ERR_NO_PORT;
1217		}
1218		if (sock->state == SS_CONNECTED) {
1219			if (!msg_connected(msg)) {
1220				msg_dbg(msg, "dispatch filter 3\n");
1221				return TIPC_ERR_NO_PORT;
1222			}
1223		}
1224		else if (sock->state == SS_CONNECTING) {
1225			if (!msg_connected(msg) && (msg_errcode(msg) == 0)) {
1226				msg_dbg(msg, "dispatch filter 4\n");
1227				return TIPC_ERR_NO_PORT;
1228			}
1229		}
1230		else if (sock->state == SS_LISTENING) {
1231			if (msg_connected(msg) || msg_errcode(msg)) {
1232				msg_dbg(msg, "dispatch filter 5\n");
1233				return TIPC_ERR_NO_PORT;
1234			}
1235		}
1236		else if (sock->state == SS_DISCONNECTING) {
1237			msg_dbg(msg, "dispatch filter 6\n");
1238			return TIPC_ERR_NO_PORT;
1239		}
1240		else /* (sock->state == SS_UNCONNECTED) */ {
1241			if (msg_connected(msg) || msg_errcode(msg)) {
1242				msg_dbg(msg, "dispatch filter 7\n");
1243				return TIPC_ERR_NO_PORT;
1244			}
1245		}
1246	}
1247
1248	/* Reject message if there isn't room to queue it */
1249
1250	recv_q_len = (u32)atomic_read(&tipc_queue_size);
1251	if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1252		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1253			return TIPC_ERR_OVERLOAD;
1254	}
1255	recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1256	if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1257		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1258			return TIPC_ERR_OVERLOAD;
1259	}
1260
1261	/* Enqueue message (finally!) */
1262
1263	msg_dbg(msg, "<DISP<: ");
1264	TIPC_SKB_CB(buf)->handle = msg_data(msg);
1265	atomic_inc(&tipc_queue_size);
1266	__skb_queue_tail(&sk->sk_receive_queue, buf);
1267
1268	/* Initiate connection termination for an incoming 'FIN' */
1269
1270	if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1271		sock->state = SS_DISCONNECTING;
1272		tipc_disconnect_port(tipc_sk_port(sk));
1273	}
1274
1275	if (waitqueue_active(sk->sk_sleep))
1276		wake_up_interruptible(sk->sk_sleep);
1277	return TIPC_OK;
1278}
1279
1280/**
1281 * backlog_rcv - handle incoming message from backlog queue
1282 * @sk: socket
1283 * @buf: message
1284 *
1285 * Caller must hold socket lock, but not port lock.
1286 *
1287 * Returns 0
1288 */
1289
1290static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1291{
1292	u32 res;
1293
1294	res = filter_rcv(sk, buf);
1295	if (res)
1296		tipc_reject_msg(buf, res);
1297	return 0;
1298}
1299
1300/**
1301 * dispatch - handle incoming message
1302 * @tport: TIPC port that received message
1303 * @buf: message
1304 *
1305 * Called with port lock already taken.
1306 *
1307 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1308 */
1309
1310static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1311{
1312	struct sock *sk = (struct sock *)tport->usr_handle;
1313	u32 res;
1314
1315	/*
1316	 * Process message if socket is unlocked; otherwise add to backlog queue
1317	 *
1318	 * This code is based on sk_receive_skb(), but must be distinct from it
1319	 * since a TIPC-specific filter/reject mechanism is utilized
1320	 */
1321
1322	bh_lock_sock(sk);
1323	if (!sock_owned_by_user(sk)) {
1324		res = filter_rcv(sk, buf);
1325	} else {
1326		if (sk_add_backlog_limited(sk, buf))
1327			res = TIPC_ERR_OVERLOAD;
1328		else
1329			res = TIPC_OK;
1330	}
1331	bh_unlock_sock(sk);
1332
1333	return res;
1334}
1335
1336/**
1337 * wakeupdispatch - wake up port after congestion
1338 * @tport: port to wakeup
1339 *
1340 * Called with port lock already taken.
1341 */
1342
1343static void wakeupdispatch(struct tipc_port *tport)
1344{
1345	struct sock *sk = (struct sock *)tport->usr_handle;
1346
1347	if (waitqueue_active(sk->sk_sleep))
1348		wake_up_interruptible(sk->sk_sleep);
1349}
1350
1351/**
1352 * connect - establish a connection to another TIPC port
1353 * @sock: socket structure
1354 * @dest: socket address for destination port
1355 * @destlen: size of socket address data structure
1356 * @flags: file-related flags associated with socket
1357 *
1358 * Returns 0 on success, errno otherwise
1359 */
1360
1361static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1362		   int flags)
1363{
1364	struct sock *sk = sock->sk;
1365	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1366	struct msghdr m = {NULL,};
1367	struct sk_buff *buf;
1368	struct tipc_msg *msg;
1369	int res;
1370
1371	lock_sock(sk);
1372
1373	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1374
1375	if (sock->state == SS_READY) {
1376		res = -EOPNOTSUPP;
1377		goto exit;
1378	}
1379
1380	/* For now, TIPC does not support the non-blocking form of connect() */
1381
1382	if (flags & O_NONBLOCK) {
1383		res = -EWOULDBLOCK;
1384		goto exit;
1385	}
1386
1387	/* Issue Posix-compliant error code if socket is in the wrong state */
1388
1389	if (sock->state == SS_LISTENING) {
1390		res = -EOPNOTSUPP;
1391		goto exit;
1392	}
1393	if (sock->state == SS_CONNECTING) {
1394		res = -EALREADY;
1395		goto exit;
1396	}
1397	if (sock->state != SS_UNCONNECTED) {
1398		res = -EISCONN;
1399		goto exit;
1400	}
1401
1402	/*
1403	 * Reject connection attempt using multicast address
1404	 *
1405	 * Note: send_msg() validates the rest of the address fields,
1406	 *       so there's no need to do it here
1407	 */
1408
1409	if (dst->addrtype == TIPC_ADDR_MCAST) {
1410		res = -EINVAL;
1411		goto exit;
1412	}
1413
1414	/* Reject any messages already in receive queue (very unlikely) */
1415
1416	reject_rx_queue(sk);
1417
1418	/* Send a 'SYN-' to destination */
1419
1420	m.msg_name = dest;
1421	m.msg_namelen = destlen;
1422	res = send_msg(NULL, sock, &m, 0);
1423	if (res < 0) {
1424		goto exit;
1425	}
1426
1427	/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1428
1429	release_sock(sk);
1430	res = wait_event_interruptible_timeout(*sk->sk_sleep,
1431			(!skb_queue_empty(&sk->sk_receive_queue) ||
1432			(sock->state != SS_CONNECTING)),
1433			sk->sk_rcvtimeo);
1434	lock_sock(sk);
1435
1436	if (res > 0) {
1437		buf = skb_peek(&sk->sk_receive_queue);
1438		if (buf != NULL) {
1439			msg = buf_msg(buf);
1440			res = auto_connect(sock, msg);
1441			if (!res) {
1442				if (!msg_data_sz(msg))
1443					advance_rx_queue(sk);
1444			}
1445		} else {
1446			if (sock->state == SS_CONNECTED) {
1447				res = -EISCONN;
1448			} else {
1449				res = -ECONNREFUSED;
1450			}
1451		}
1452	} else {
1453		if (res == 0)
1454			res = -ETIMEDOUT;
1455		else
1456			; /* leave "res" unchanged */
1457		sock->state = SS_DISCONNECTING;
1458	}
1459
1460exit:
1461	release_sock(sk);
1462	return res;
1463}
1464
1465/**
1466 * listen - allow socket to listen for incoming connections
1467 * @sock: socket structure
1468 * @len: (unused)
1469 *
1470 * Returns 0 on success, errno otherwise
1471 */
1472
1473static int listen(struct socket *sock, int len)
1474{
1475	struct sock *sk = sock->sk;
1476	int res;
1477
1478	lock_sock(sk);
1479
1480	if (sock->state == SS_READY)
1481		res = -EOPNOTSUPP;
1482	else if (sock->state != SS_UNCONNECTED)
1483		res = -EINVAL;
1484	else {
1485		sock->state = SS_LISTENING;
1486		res = 0;
1487	}
1488
1489	release_sock(sk);
1490	return res;
1491}
1492
1493/**
1494 * accept - wait for connection request
1495 * @sock: listening socket
1496 * @newsock: new socket that is to be connected
1497 * @flags: file-related flags associated with socket
1498 *
1499 * Returns 0 on success, errno otherwise
1500 */
1501
1502static int accept(struct socket *sock, struct socket *new_sock, int flags)
1503{
1504	struct sock *sk = sock->sk;
1505	struct sk_buff *buf;
1506	int res;
1507
1508	lock_sock(sk);
1509
1510	if (sock->state == SS_READY) {
1511		res = -EOPNOTSUPP;
1512		goto exit;
1513	}
1514	if (sock->state != SS_LISTENING) {
1515		res = -EINVAL;
1516		goto exit;
1517	}
1518
1519	while (skb_queue_empty(&sk->sk_receive_queue)) {
1520		if (flags & O_NONBLOCK) {
1521			res = -EWOULDBLOCK;
1522			goto exit;
1523		}
1524		release_sock(sk);
1525		res = wait_event_interruptible(*sk->sk_sleep,
1526				(!skb_queue_empty(&sk->sk_receive_queue)));
1527		lock_sock(sk);
1528		if (res)
1529			goto exit;
1530	}
1531
1532	buf = skb_peek(&sk->sk_receive_queue);
1533
1534	res = tipc_create(sock_net(sock->sk), new_sock, 0);
1535	if (!res) {
1536		struct sock *new_sk = new_sock->sk;
1537		struct tipc_sock *new_tsock = tipc_sk(new_sk);
1538		struct tipc_port *new_tport = new_tsock->p;
1539		u32 new_ref = new_tport->ref;
1540		struct tipc_msg *msg = buf_msg(buf);
1541
1542		lock_sock(new_sk);
1543
1544		/*
1545		 * Reject any stray messages received by new socket
1546		 * before the socket lock was taken (very, very unlikely)
1547		 */
1548
1549		reject_rx_queue(new_sk);
1550
1551		/* Connect new socket to it's peer */
1552
1553		new_tsock->peer_name.ref = msg_origport(msg);
1554		new_tsock->peer_name.node = msg_orignode(msg);
1555		tipc_connect2port(new_ref, &new_tsock->peer_name);
1556		new_sock->state = SS_CONNECTED;
1557
1558		tipc_set_portimportance(new_ref, msg_importance(msg));
1559		if (msg_named(msg)) {
1560			new_tport->conn_type = msg_nametype(msg);
1561			new_tport->conn_instance = msg_nameinst(msg);
1562		}
1563
1564		/*
1565		 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1566		 * Respond to 'SYN+' by queuing it on new socket.
1567		 */
1568
1569		msg_dbg(msg,"<ACC<: ");
1570		if (!msg_data_sz(msg)) {
1571			struct msghdr m = {NULL,};
1572
1573			advance_rx_queue(sk);
1574			send_packet(NULL, new_sock, &m, 0);
1575		} else {
1576			__skb_dequeue(&sk->sk_receive_queue);
1577			__skb_queue_head(&new_sk->sk_receive_queue, buf);
1578		}
1579		release_sock(new_sk);
1580	}
1581exit:
1582	release_sock(sk);
1583	return res;
1584}
1585
1586/**
1587 * shutdown - shutdown socket connection
1588 * @sock: socket structure
1589 * @how: direction to close (must be SHUT_RDWR)
1590 *
1591 * Terminates connection (if necessary), then purges socket's receive queue.
1592 *
1593 * Returns 0 on success, errno otherwise
1594 */
1595
1596static int shutdown(struct socket *sock, int how)
1597{
1598	struct sock *sk = sock->sk;
1599	struct tipc_port *tport = tipc_sk_port(sk);
1600	struct sk_buff *buf;
1601	int res;
1602
1603	if (how != SHUT_RDWR)
1604		return -EINVAL;
1605
1606	lock_sock(sk);
1607
1608	switch (sock->state) {
1609	case SS_CONNECTING:
1610	case SS_CONNECTED:
1611
1612		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1613restart:
1614		buf = __skb_dequeue(&sk->sk_receive_queue);
1615		if (buf) {
1616			atomic_dec(&tipc_queue_size);
1617			if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) {
1618				buf_discard(buf);
1619				goto restart;
1620			}
1621			tipc_disconnect(tport->ref);
1622			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1623		} else {
1624			tipc_shutdown(tport->ref);
1625		}
1626
1627		sock->state = SS_DISCONNECTING;
1628
1629		/* fall through */
1630
1631	case SS_DISCONNECTING:
1632
1633		/* Discard any unreceived messages; wake up sleeping tasks */
1634
1635		discard_rx_queue(sk);
1636		if (waitqueue_active(sk->sk_sleep))
1637			wake_up_interruptible(sk->sk_sleep);
1638		res = 0;
1639		break;
1640
1641	default:
1642		res = -ENOTCONN;
1643	}
1644
1645	release_sock(sk);
1646	return res;
1647}
1648
1649/**
1650 * setsockopt - set socket option
1651 * @sock: socket structure
1652 * @lvl: option level
1653 * @opt: option identifier
1654 * @ov: pointer to new option value
1655 * @ol: length of option value
1656 *
1657 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1658 * (to ease compatibility).
1659 *
1660 * Returns 0 on success, errno otherwise
1661 */
1662
1663static int setsockopt(struct socket *sock,
1664		      int lvl, int opt, char __user *ov, unsigned int ol)
1665{
1666	struct sock *sk = sock->sk;
1667	struct tipc_port *tport = tipc_sk_port(sk);
1668	u32 value;
1669	int res;
1670
1671	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1672		return 0;
1673	if (lvl != SOL_TIPC)
1674		return -ENOPROTOOPT;
1675	if (ol < sizeof(value))
1676		return -EINVAL;
1677	if ((res = get_user(value, (u32 __user *)ov)))
1678		return res;
1679
1680	lock_sock(sk);
1681
1682	switch (opt) {
1683	case TIPC_IMPORTANCE:
1684		res = tipc_set_portimportance(tport->ref, value);
1685		break;
1686	case TIPC_SRC_DROPPABLE:
1687		if (sock->type != SOCK_STREAM)
1688			res = tipc_set_portunreliable(tport->ref, value);
1689		else
1690			res = -ENOPROTOOPT;
1691		break;
1692	case TIPC_DEST_DROPPABLE:
1693		res = tipc_set_portunreturnable(tport->ref, value);
1694		break;
1695	case TIPC_CONN_TIMEOUT:
1696		sk->sk_rcvtimeo = msecs_to_jiffies(value);
1697		/* no need to set "res", since already 0 at this point */
1698		break;
1699	default:
1700		res = -EINVAL;
1701	}
1702
1703	release_sock(sk);
1704
1705	return res;
1706}
1707
1708/**
1709 * getsockopt - get socket option
1710 * @sock: socket structure
1711 * @lvl: option level
1712 * @opt: option identifier
1713 * @ov: receptacle for option value
1714 * @ol: receptacle for length of option value
1715 *
1716 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1717 * (to ease compatibility).
1718 *
1719 * Returns 0 on success, errno otherwise
1720 */
1721
1722static int getsockopt(struct socket *sock,
1723		      int lvl, int opt, char __user *ov, int __user *ol)
1724{
1725	struct sock *sk = sock->sk;
1726	struct tipc_port *tport = tipc_sk_port(sk);
1727	int len;
1728	u32 value;
1729	int res;
1730
1731	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1732		return put_user(0, ol);
1733	if (lvl != SOL_TIPC)
1734		return -ENOPROTOOPT;
1735	if ((res = get_user(len, ol)))
1736		return res;
1737
1738	lock_sock(sk);
1739
1740	switch (opt) {
1741	case TIPC_IMPORTANCE:
1742		res = tipc_portimportance(tport->ref, &value);
1743		break;
1744	case TIPC_SRC_DROPPABLE:
1745		res = tipc_portunreliable(tport->ref, &value);
1746		break;
1747	case TIPC_DEST_DROPPABLE:
1748		res = tipc_portunreturnable(tport->ref, &value);
1749		break;
1750	case TIPC_CONN_TIMEOUT:
1751		value = jiffies_to_msecs(sk->sk_rcvtimeo);
1752		/* no need to set "res", since already 0 at this point */
1753		break;
1754	 case TIPC_NODE_RECVQ_DEPTH:
1755		value = (u32)atomic_read(&tipc_queue_size);
1756		break;
1757	 case TIPC_SOCK_RECVQ_DEPTH:
1758		value = skb_queue_len(&sk->sk_receive_queue);
1759		break;
1760	default:
1761		res = -EINVAL;
1762	}
1763
1764	release_sock(sk);
1765
1766	if (res) {
1767		/* "get" failed */
1768	}
1769	else if (len < sizeof(value)) {
1770		res = -EINVAL;
1771	}
1772	else if (copy_to_user(ov, &value, sizeof(value))) {
1773		res = -EFAULT;
1774	}
1775	else {
1776		res = put_user(sizeof(value), ol);
1777	}
1778
1779	return res;
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