1/*
   2 * net/tipc/socket.c: TIPC socket API
   3 *
   4 * Copyright (c) 2001-2007, 2012-2014, Ericsson AB
   5 * Copyright (c) 2004-2008, 2010-2013, 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 "core.h"
  38#include "port.h"
  39
  40#include <linux/export.h>
  41
  42#define SS_LISTENING	-1	/* socket is listening */
  43#define SS_READY	-2	/* socket is connectionless */
  44
 
  45#define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
  46
 
 
 
 
 
 
 
 
 
 
 
 
 
  47static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
  48static void tipc_data_ready(struct sock *sk);
  49static void tipc_write_space(struct sock *sk);
  50static int tipc_release(struct socket *sock);
  51static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
  52
  53static const struct proto_ops packet_ops;
  54static const struct proto_ops stream_ops;
  55static const struct proto_ops msg_ops;
  56
  57static struct proto tipc_proto;
  58static struct proto tipc_proto_kern;
 
 
 
  59
  60/*
  61 * Revised TIPC socket locking policy:
  62 *
  63 * Most socket operations take the standard socket lock when they start
  64 * and hold it until they finish (or until they need to sleep).  Acquiring
  65 * this lock grants the owner exclusive access to the fields of the socket
  66 * data structures, with the exception of the backlog queue.  A few socket
  67 * operations can be done without taking the socket lock because they only
  68 * read socket information that never changes during the life of the socket.
  69 *
  70 * Socket operations may acquire the lock for the associated TIPC port if they
  71 * need to perform an operation on the port.  If any routine needs to acquire
  72 * both the socket lock and the port lock it must take the socket lock first
  73 * to avoid the risk of deadlock.
  74 *
  75 * The dispatcher handling incoming messages cannot grab the socket lock in
  76 * the standard fashion, since invoked it runs at the BH level and cannot block.
  77 * Instead, it checks to see if the socket lock is currently owned by someone,
  78 * and either handles the message itself or adds it to the socket's backlog
  79 * queue; in the latter case the queued message is processed once the process
  80 * owning the socket lock releases it.
  81 *
  82 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
  83 * the problem of a blocked socket operation preventing any other operations
  84 * from occurring.  However, applications must be careful if they have
  85 * multiple threads trying to send (or receive) on the same socket, as these
  86 * operations might interfere with each other.  For example, doing a connect
  87 * and a receive at the same time might allow the receive to consume the
  88 * ACK message meant for the connect.  While additional work could be done
  89 * to try and overcome this, it doesn't seem to be worthwhile at the present.
  90 *
  91 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
  92 * that another operation that must be performed in a non-blocking manner is
  93 * not delayed for very long because the lock has already been taken.
  94 *
  95 * NOTE: This code assumes that certain fields of a port/socket pair are
  96 * constant over its lifetime; such fields can be examined without taking
  97 * the socket lock and/or port lock, and do not need to be re-read even
  98 * after resuming processing after waiting.  These fields include:
  99 *   - socket type
 100 *   - pointer to socket sk structure (aka tipc_sock structure)
 101 *   - pointer to port structure
 102 *   - port reference
 103 */
 104
 105#include "socket.h"
 106
 107/**
 108 * advance_rx_queue - discard first buffer in socket receive queue
 109 *
 110 * Caller must hold socket lock
 111 */
 112static void advance_rx_queue(struct sock *sk)
 113{
 114	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 115}
 116
 117/**
 118 * reject_rx_queue - reject all buffers in socket receive queue
 119 *
 120 * Caller must hold socket lock
 121 */
 122static void reject_rx_queue(struct sock *sk)
 123{
 124	struct sk_buff *buf;
 125
 126	while ((buf = __skb_dequeue(&sk->sk_receive_queue)))
 127		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
 
 
 128}
 129
 130/**
 131 * tipc_sk_create - create a TIPC socket
 132 * @net: network namespace (must be default network)
 133 * @sock: pre-allocated socket structure
 134 * @protocol: protocol indicator (must be 0)
 135 * @kern: caused by kernel or by userspace?
 136 *
 137 * This routine creates additional data structures used by the TIPC socket,
 138 * initializes them, and links them together.
 139 *
 140 * Returns 0 on success, errno otherwise
 141 */
 142static int tipc_sk_create(struct net *net, struct socket *sock,
 143			  int protocol, int kern)
 144{
 145	const struct proto_ops *ops;
 146	socket_state state;
 147	struct sock *sk;
 148	struct tipc_sock *tsk;
 149	struct tipc_port *port;
 150	u32 ref;
 151
 152	/* Validate arguments */
 153	if (unlikely(protocol != 0))
 154		return -EPROTONOSUPPORT;
 155
 156	switch (sock->type) {
 157	case SOCK_STREAM:
 158		ops = &stream_ops;
 159		state = SS_UNCONNECTED;
 160		break;
 161	case SOCK_SEQPACKET:
 162		ops = &packet_ops;
 163		state = SS_UNCONNECTED;
 164		break;
 165	case SOCK_DGRAM:
 166	case SOCK_RDM:
 167		ops = &msg_ops;
 168		state = SS_READY;
 169		break;
 170	default:
 171		return -EPROTOTYPE;
 172	}
 173
 174	/* Allocate socket's protocol area */
 175	if (!kern)
 176		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
 177	else
 178		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
 179
 180	if (sk == NULL)
 181		return -ENOMEM;
 182
 183	tsk = tipc_sk(sk);
 184	port = &tsk->port;
 185
 186	ref = tipc_port_init(port, TIPC_LOW_IMPORTANCE);
 187	if (!ref) {
 188		pr_warn("Socket registration failed, ref. table exhausted\n");
 189		sk_free(sk);
 190		return -ENOMEM;
 191	}
 192
 193	/* Finish initializing socket data structures */
 194	sock->ops = ops;
 195	sock->state = state;
 196
 197	sock_init_data(sock, sk);
 198	sk->sk_backlog_rcv = backlog_rcv;
 199	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
 200	sk->sk_data_ready = tipc_data_ready;
 201	sk->sk_write_space = tipc_write_space;
 202	tipc_sk(sk)->conn_timeout = CONN_TIMEOUT_DEFAULT;
 203	tipc_port_unlock(port);
 
 204
 205	if (sock->state == SS_READY) {
 206		tipc_port_set_unreturnable(port, true);
 207		if (sock->type == SOCK_DGRAM)
 208			tipc_port_set_unreliable(port, true);
 209	}
 210	return 0;
 211}
 212
 213/**
 214 * tipc_sock_create_local - create TIPC socket from inside TIPC module
 215 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
 216 *
 217 * We cannot use sock_creat_kern here because it bumps module user count.
 218 * Since socket owner and creator is the same module we must make sure
 219 * that module count remains zero for module local sockets, otherwise
 220 * we cannot do rmmod.
 221 *
 222 * Returns 0 on success, errno otherwise
 223 */
 224int tipc_sock_create_local(int type, struct socket **res)
 225{
 226	int rc;
 227
 228	rc = sock_create_lite(AF_TIPC, type, 0, res);
 229	if (rc < 0) {
 230		pr_err("Failed to create kernel socket\n");
 231		return rc;
 232	}
 233	tipc_sk_create(&init_net, *res, 0, 1);
 234
 235	return 0;
 236}
 237
 238/**
 239 * tipc_sock_release_local - release socket created by tipc_sock_create_local
 240 * @sock: the socket to be released.
 241 *
 242 * Module reference count is not incremented when such sockets are created,
 243 * so we must keep it from being decremented when they are released.
 244 */
 245void tipc_sock_release_local(struct socket *sock)
 246{
 247	tipc_release(sock);
 248	sock->ops = NULL;
 249	sock_release(sock);
 250}
 251
 252/**
 253 * tipc_sock_accept_local - accept a connection on a socket created
 254 * with tipc_sock_create_local. Use this function to avoid that
 255 * module reference count is inadvertently incremented.
 256 *
 257 * @sock:    the accepting socket
 258 * @newsock: reference to the new socket to be created
 259 * @flags:   socket flags
 260 */
 261
 262int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
 263			   int flags)
 264{
 265	struct sock *sk = sock->sk;
 266	int ret;
 267
 268	ret = sock_create_lite(sk->sk_family, sk->sk_type,
 269			       sk->sk_protocol, newsock);
 270	if (ret < 0)
 271		return ret;
 272
 273	ret = tipc_accept(sock, *newsock, flags);
 274	if (ret < 0) {
 275		sock_release(*newsock);
 276		return ret;
 277	}
 278	(*newsock)->ops = sock->ops;
 279	return ret;
 280}
 281
 282/**
 283 * tipc_release - destroy a TIPC socket
 284 * @sock: socket to destroy
 285 *
 286 * This routine cleans up any messages that are still queued on the socket.
 287 * For DGRAM and RDM socket types, all queued messages are rejected.
 288 * For SEQPACKET and STREAM socket types, the first message is rejected
 289 * and any others are discarded.  (If the first message on a STREAM socket
 290 * is partially-read, it is discarded and the next one is rejected instead.)
 291 *
 292 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 293 * are returned or discarded according to the "destination droppable" setting
 294 * specified for the message by the sender.
 295 *
 296 * Returns 0 on success, errno otherwise
 297 */
 298static int tipc_release(struct socket *sock)
 299{
 300	struct sock *sk = sock->sk;
 301	struct tipc_sock *tsk;
 302	struct tipc_port *port;
 303	struct sk_buff *buf;
 
 304
 305	/*
 306	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 307	 * releases a pre-allocated child socket that was never used)
 308	 */
 309	if (sk == NULL)
 310		return 0;
 311
 312	tsk = tipc_sk(sk);
 313	port = &tsk->port;
 314	lock_sock(sk);
 315
 316	/*
 317	 * Reject all unreceived messages, except on an active connection
 318	 * (which disconnects locally & sends a 'FIN+' to peer)
 319	 */
 320	while (sock->state != SS_DISCONNECTING) {
 321		buf = __skb_dequeue(&sk->sk_receive_queue);
 322		if (buf == NULL)
 323			break;
 324		if (TIPC_SKB_CB(buf)->handle != NULL)
 
 325			kfree_skb(buf);
 326		else {
 327			if ((sock->state == SS_CONNECTING) ||
 328			    (sock->state == SS_CONNECTED)) {
 329				sock->state = SS_DISCONNECTING;
 330				tipc_port_disconnect(port->ref);
 331			}
 332			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
 333		}
 334	}
 335
 336	/* Destroy TIPC port; also disconnects an active connection and
 337	 * sends a 'FIN-' to peer.
 
 338	 */
 339	tipc_port_destroy(port);
 340
 341	/* Discard any remaining (connection-based) messages in receive queue */
 342	__skb_queue_purge(&sk->sk_receive_queue);
 343
 344	/* Reject any messages that accumulated in backlog queue */
 345	sock->state = SS_DISCONNECTING;
 346	release_sock(sk);
 347
 348	sock_put(sk);
 349	sock->sk = NULL;
 350
 351	return 0;
 352}
 353
 354/**
 355 * tipc_bind - associate or disassocate TIPC name(s) with a socket
 356 * @sock: socket structure
 357 * @uaddr: socket address describing name(s) and desired operation
 358 * @uaddr_len: size of socket address data structure
 359 *
 360 * Name and name sequence binding is indicated using a positive scope value;
 361 * a negative scope value unbinds the specified name.  Specifying no name
 362 * (i.e. a socket address length of 0) unbinds all names from the socket.
 363 *
 364 * Returns 0 on success, errno otherwise
 365 *
 366 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 367 *       access any non-constant socket information.
 368 */
 369static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
 370		     int uaddr_len)
 371{
 372	struct sock *sk = sock->sk;
 373	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 374	struct tipc_sock *tsk = tipc_sk(sk);
 375	int res = -EINVAL;
 376
 377	lock_sock(sk);
 378	if (unlikely(!uaddr_len)) {
 379		res = tipc_withdraw(&tsk->port, 0, NULL);
 380		goto exit;
 381	}
 382
 383	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
 384		res = -EINVAL;
 385		goto exit;
 386	}
 387	if (addr->family != AF_TIPC) {
 388		res = -EAFNOSUPPORT;
 389		goto exit;
 390	}
 391
 392	if (addr->addrtype == TIPC_ADDR_NAME)
 393		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
 394	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
 395		res = -EAFNOSUPPORT;
 396		goto exit;
 397	}
 398
 399	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
 400	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
 401	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
 402		res = -EACCES;
 403		goto exit;
 404	}
 405
 406	res = (addr->scope > 0) ?
 407		tipc_publish(&tsk->port, addr->scope, &addr->addr.nameseq) :
 408		tipc_withdraw(&tsk->port, -addr->scope, &addr->addr.nameseq);
 409exit:
 410	release_sock(sk);
 411	return res;
 412}
 413
 414/**
 415 * tipc_getname - get port ID of socket or peer socket
 416 * @sock: socket structure
 417 * @uaddr: area for returned socket address
 418 * @uaddr_len: area for returned length of socket address
 419 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 420 *
 421 * Returns 0 on success, errno otherwise
 422 *
 423 * NOTE: This routine doesn't need to take the socket lock since it only
 424 *       accesses socket information that is unchanging (or which changes in
 425 *       a completely predictable manner).
 426 */
 427static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
 428			int *uaddr_len, int peer)
 429{
 430	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 431	struct tipc_sock *tsk = tipc_sk(sock->sk);
 432
 433	memset(addr, 0, sizeof(*addr));
 434	if (peer) {
 435		if ((sock->state != SS_CONNECTED) &&
 436			((peer != 2) || (sock->state != SS_DISCONNECTING)))
 437			return -ENOTCONN;
 438		addr->addr.id.ref = tipc_port_peerport(&tsk->port);
 439		addr->addr.id.node = tipc_port_peernode(&tsk->port);
 440	} else {
 441		addr->addr.id.ref = tsk->port.ref;
 442		addr->addr.id.node = tipc_own_addr;
 443	}
 444
 445	*uaddr_len = sizeof(*addr);
 446	addr->addrtype = TIPC_ADDR_ID;
 447	addr->family = AF_TIPC;
 448	addr->scope = 0;
 449	addr->addr.name.domain = 0;
 450
 451	return 0;
 452}
 453
 454/**
 455 * tipc_poll - read and possibly block on pollmask
 456 * @file: file structure associated with the socket
 457 * @sock: socket for which to calculate the poll bits
 458 * @wait: ???
 459 *
 460 * Returns pollmask value
 461 *
 462 * COMMENTARY:
 463 * It appears that the usual socket locking mechanisms are not useful here
 464 * since the pollmask info is potentially out-of-date the moment this routine
 465 * exits.  TCP and other protocols seem to rely on higher level poll routines
 466 * to handle any preventable race conditions, so TIPC will do the same ...
 467 *
 468 * TIPC sets the returned events as follows:
 469 *
 470 * socket state		flags set
 471 * ------------		---------
 472 * unconnected		no read flags
 473 *			POLLOUT if port is not congested
 474 *
 475 * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
 476 *			no write flags
 477 *
 478 * connected		POLLIN/POLLRDNORM if data in rx queue
 479 *			POLLOUT if port is not congested
 480 *
 481 * disconnecting	POLLIN/POLLRDNORM/POLLHUP
 482 *			no write flags
 483 *
 484 * listening		POLLIN if SYN in rx queue
 485 *			no write flags
 486 *
 487 * ready		POLLIN/POLLRDNORM if data in rx queue
 488 * [connectionless]	POLLOUT (since port cannot be congested)
 489 *
 490 * IMPORTANT: The fact that a read or write operation is indicated does NOT
 491 * imply that the operation will succeed, merely that it should be performed
 492 * and will not block.
 493 */
 494static unsigned int tipc_poll(struct file *file, struct socket *sock,
 495			      poll_table *wait)
 496{
 497	struct sock *sk = sock->sk;
 498	struct tipc_sock *tsk = tipc_sk(sk);
 499	u32 mask = 0;
 500
 501	sock_poll_wait(file, sk_sleep(sk), wait);
 502
 503	switch ((int)sock->state) {
 504	case SS_UNCONNECTED:
 505		if (!tsk->port.congested)
 506			mask |= POLLOUT;
 507		break;
 508	case SS_READY:
 509	case SS_CONNECTED:
 510		if (!tsk->port.congested)
 511			mask |= POLLOUT;
 512		/* fall thru' */
 513	case SS_CONNECTING:
 514	case SS_LISTENING:
 515		if (!skb_queue_empty(&sk->sk_receive_queue))
 516			mask |= (POLLIN | POLLRDNORM);
 517		break;
 518	case SS_DISCONNECTING:
 519		mask = (POLLIN | POLLRDNORM | POLLHUP);
 520		break;
 521	}
 522
 523	return mask;
 524}
 525
 526/**
 527 * dest_name_check - verify user is permitted to send to specified port name
 528 * @dest: destination address
 529 * @m: descriptor for message to be sent
 530 *
 531 * Prevents restricted configuration commands from being issued by
 532 * unauthorized users.
 533 *
 534 * Returns 0 if permission is granted, otherwise errno
 535 */
 536static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
 537{
 538	struct tipc_cfg_msg_hdr hdr;
 539
 540	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
 541		return 0;
 542	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
 543		return 0;
 544	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
 545		return -EACCES;
 546
 547	if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
 548		return -EMSGSIZE;
 549	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
 550		return -EFAULT;
 551	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
 552		return -EACCES;
 553
 554	return 0;
 555}
 556
 557static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
 558{
 559	struct sock *sk = sock->sk;
 560	struct tipc_sock *tsk = tipc_sk(sk);
 561	DEFINE_WAIT(wait);
 562	int done;
 563
 564	do {
 565		int err = sock_error(sk);
 566		if (err)
 567			return err;
 568		if (sock->state == SS_DISCONNECTING)
 569			return -EPIPE;
 570		if (!*timeo_p)
 571			return -EAGAIN;
 572		if (signal_pending(current))
 573			return sock_intr_errno(*timeo_p);
 574
 575		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
 576		done = sk_wait_event(sk, timeo_p, !tsk->port.congested);
 577		finish_wait(sk_sleep(sk), &wait);
 578	} while (!done);
 579	return 0;
 580}
 581
 582
 583/**
 584 * tipc_sendmsg - send message in connectionless manner
 585 * @iocb: if NULL, indicates that socket lock is already held
 586 * @sock: socket structure
 587 * @m: message to send
 588 * @total_len: length of message
 589 *
 590 * Message must have an destination specified explicitly.
 591 * Used for SOCK_RDM and SOCK_DGRAM messages,
 592 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
 593 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
 594 *
 595 * Returns the number of bytes sent on success, or errno otherwise
 596 */
 597static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
 598			struct msghdr *m, size_t total_len)
 599{
 600	struct sock *sk = sock->sk;
 601	struct tipc_sock *tsk = tipc_sk(sk);
 602	struct tipc_port *port = &tsk->port;
 603	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 604	int needs_conn;
 605	long timeo;
 606	int res = -EINVAL;
 607
 608	if (unlikely(!dest))
 609		return -EDESTADDRREQ;
 610	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
 611		     (dest->family != AF_TIPC)))
 612		return -EINVAL;
 613	if (total_len > TIPC_MAX_USER_MSG_SIZE)
 
 614		return -EMSGSIZE;
 615
 616	if (iocb)
 617		lock_sock(sk);
 618
 619	needs_conn = (sock->state != SS_READY);
 620	if (unlikely(needs_conn)) {
 621		if (sock->state == SS_LISTENING) {
 622			res = -EPIPE;
 623			goto exit;
 624		}
 625		if (sock->state != SS_UNCONNECTED) {
 626			res = -EISCONN;
 627			goto exit;
 628		}
 629		if (tsk->port.published) {
 
 630			res = -EOPNOTSUPP;
 631			goto exit;
 632		}
 633		if (dest->addrtype == TIPC_ADDR_NAME) {
 634			tsk->port.conn_type = dest->addr.name.name.type;
 635			tsk->port.conn_instance = dest->addr.name.name.instance;
 636		}
 637
 638		/* Abort any pending connection attempts (very unlikely) */
 639		reject_rx_queue(sk);
 640	}
 641
 642	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
 
 643	do {
 644		if (dest->addrtype == TIPC_ADDR_NAME) {
 645			res = dest_name_check(dest, m);
 646			if (res)
 647				break;
 648			res = tipc_send2name(port,
 649					     &dest->addr.name.name,
 650					     dest->addr.name.domain,
 
 651					     m->msg_iov,
 652					     total_len);
 653		} else if (dest->addrtype == TIPC_ADDR_ID) {
 654			res = tipc_send2port(port,
 655					     &dest->addr.id,
 
 656					     m->msg_iov,
 657					     total_len);
 658		} else if (dest->addrtype == TIPC_ADDR_MCAST) {
 659			if (needs_conn) {
 660				res = -EOPNOTSUPP;
 661				break;
 662			}
 663			res = dest_name_check(dest, m);
 664			if (res)
 665				break;
 666			res = tipc_port_mcast_xmit(port,
 667						   &dest->addr.nameseq,
 668						   m->msg_iov,
 669						   total_len);
 
 670		}
 671		if (likely(res != -ELINKCONG)) {
 672			if (needs_conn && (res >= 0))
 673				sock->state = SS_CONNECTING;
 674			break;
 675		}
 676		res = tipc_wait_for_sndmsg(sock, &timeo);
 677		if (res)
 678			break;
 
 
 
 
 
 679	} while (1);
 680
 681exit:
 682	if (iocb)
 683		release_sock(sk);
 684	return res;
 685}
 686
 687static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
 688{
 689	struct sock *sk = sock->sk;
 690	struct tipc_sock *tsk = tipc_sk(sk);
 691	struct tipc_port *port = &tsk->port;
 692	DEFINE_WAIT(wait);
 693	int done;
 694
 695	do {
 696		int err = sock_error(sk);
 697		if (err)
 698			return err;
 699		if (sock->state == SS_DISCONNECTING)
 700			return -EPIPE;
 701		else if (sock->state != SS_CONNECTED)
 702			return -ENOTCONN;
 703		if (!*timeo_p)
 704			return -EAGAIN;
 705		if (signal_pending(current))
 706			return sock_intr_errno(*timeo_p);
 707
 708		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
 709		done = sk_wait_event(sk, timeo_p,
 710				     (!port->congested || !port->connected));
 711		finish_wait(sk_sleep(sk), &wait);
 712	} while (!done);
 713	return 0;
 714}
 715
 716/**
 717 * tipc_send_packet - send a connection-oriented message
 718 * @iocb: if NULL, indicates that socket lock is already held
 719 * @sock: socket structure
 720 * @m: message to send
 721 * @total_len: length of message
 722 *
 723 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
 724 *
 725 * Returns the number of bytes sent on success, or errno otherwise
 726 */
 727static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
 728			    struct msghdr *m, size_t total_len)
 729{
 730	struct sock *sk = sock->sk;
 731	struct tipc_sock *tsk = tipc_sk(sk);
 732	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 733	int res = -EINVAL;
 734	long timeo;
 735
 736	/* Handle implied connection establishment */
 737	if (unlikely(dest))
 738		return tipc_sendmsg(iocb, sock, m, total_len);
 739
 740	if (total_len > TIPC_MAX_USER_MSG_SIZE)
 
 741		return -EMSGSIZE;
 742
 743	if (iocb)
 744		lock_sock(sk);
 745
 746	if (unlikely(sock->state != SS_CONNECTED)) {
 747		if (sock->state == SS_DISCONNECTING)
 748			res = -EPIPE;
 749		else
 750			res = -ENOTCONN;
 751		goto exit;
 752	}
 753
 754	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
 755	do {
 756		res = tipc_send(&tsk->port, m->msg_iov, total_len);
 
 
 
 
 
 
 
 
 
 757		if (likely(res != -ELINKCONG))
 758			break;
 759		res = tipc_wait_for_sndpkt(sock, &timeo);
 760		if (res)
 761			break;
 
 
 
 
 
 762	} while (1);
 763exit:
 764	if (iocb)
 765		release_sock(sk);
 766	return res;
 767}
 768
 769/**
 770 * tipc_send_stream - send stream-oriented data
 771 * @iocb: (unused)
 772 * @sock: socket structure
 773 * @m: data to send
 774 * @total_len: total length of data to be sent
 775 *
 776 * Used for SOCK_STREAM data.
 777 *
 778 * Returns the number of bytes sent on success (or partial success),
 779 * or errno if no data sent
 780 */
 781static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
 782			    struct msghdr *m, size_t total_len)
 783{
 784	struct sock *sk = sock->sk;
 785	struct tipc_sock *tsk = tipc_sk(sk);
 786	struct msghdr my_msg;
 787	struct iovec my_iov;
 788	struct iovec *curr_iov;
 789	int curr_iovlen;
 790	char __user *curr_start;
 791	u32 hdr_size;
 792	int curr_left;
 793	int bytes_to_send;
 794	int bytes_sent;
 795	int res;
 796
 797	lock_sock(sk);
 798
 799	/* Handle special cases where there is no connection */
 800	if (unlikely(sock->state != SS_CONNECTED)) {
 801		if (sock->state == SS_UNCONNECTED)
 802			res = tipc_send_packet(NULL, sock, m, total_len);
 803		else
 804			res = sock->state == SS_DISCONNECTING ? -EPIPE : -ENOTCONN;
 805		goto exit;
 
 
 
 
 
 806	}
 807
 808	if (unlikely(m->msg_name)) {
 809		res = -EISCONN;
 810		goto exit;
 811	}
 812
 813	if (total_len > (unsigned int)INT_MAX) {
 
 814		res = -EMSGSIZE;
 815		goto exit;
 816	}
 817
 818	/*
 819	 * Send each iovec entry using one or more messages
 820	 *
 821	 * Note: This algorithm is good for the most likely case
 822	 * (i.e. one large iovec entry), but could be improved to pass sets
 823	 * of small iovec entries into send_packet().
 824	 */
 825	curr_iov = m->msg_iov;
 826	curr_iovlen = m->msg_iovlen;
 827	my_msg.msg_iov = &my_iov;
 828	my_msg.msg_iovlen = 1;
 829	my_msg.msg_flags = m->msg_flags;
 830	my_msg.msg_name = NULL;
 831	bytes_sent = 0;
 832
 833	hdr_size = msg_hdr_sz(&tsk->port.phdr);
 834
 835	while (curr_iovlen--) {
 836		curr_start = curr_iov->iov_base;
 837		curr_left = curr_iov->iov_len;
 838
 839		while (curr_left) {
 840			bytes_to_send = tsk->port.max_pkt - hdr_size;
 841			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
 842				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
 843			if (curr_left < bytes_to_send)
 844				bytes_to_send = curr_left;
 845			my_iov.iov_base = curr_start;
 846			my_iov.iov_len = bytes_to_send;
 847			res = tipc_send_packet(NULL, sock, &my_msg,
 848					       bytes_to_send);
 849			if (res < 0) {
 850				if (bytes_sent)
 851					res = bytes_sent;
 852				goto exit;
 853			}
 854			curr_left -= bytes_to_send;
 855			curr_start += bytes_to_send;
 856			bytes_sent += bytes_to_send;
 857		}
 858
 859		curr_iov++;
 860	}
 861	res = bytes_sent;
 862exit:
 863	release_sock(sk);
 864	return res;
 865}
 866
 867/**
 868 * auto_connect - complete connection setup to a remote port
 869 * @tsk: tipc socket structure
 870 * @msg: peer's response message
 871 *
 872 * Returns 0 on success, errno otherwise
 873 */
 874static int auto_connect(struct tipc_sock *tsk, struct tipc_msg *msg)
 875{
 876	struct tipc_port *port = &tsk->port;
 877	struct socket *sock = tsk->sk.sk_socket;
 878	struct tipc_portid peer;
 879
 880	peer.ref = msg_origport(msg);
 881	peer.node = msg_orignode(msg);
 882
 883	__tipc_port_connect(port->ref, port, &peer);
 
 
 
 884
 885	if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)
 886		return -EINVAL;
 887	msg_set_importance(&port->phdr, (u32)msg_importance(msg));
 
 888	sock->state = SS_CONNECTED;
 889	return 0;
 890}
 891
 892/**
 893 * set_orig_addr - capture sender's address for received message
 894 * @m: descriptor for message info
 895 * @msg: received message header
 896 *
 897 * Note: Address is not captured if not requested by receiver.
 898 */
 899static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
 900{
 901	DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
 902
 903	if (addr) {
 904		addr->family = AF_TIPC;
 905		addr->addrtype = TIPC_ADDR_ID;
 906		memset(&addr->addr, 0, sizeof(addr->addr));
 907		addr->addr.id.ref = msg_origport(msg);
 908		addr->addr.id.node = msg_orignode(msg);
 909		addr->addr.name.domain = 0;	/* could leave uninitialized */
 910		addr->scope = 0;		/* could leave uninitialized */
 911		m->msg_namelen = sizeof(struct sockaddr_tipc);
 912	}
 913}
 914
 915/**
 916 * anc_data_recv - optionally capture ancillary data for received message
 917 * @m: descriptor for message info
 918 * @msg: received message header
 919 * @tport: TIPC port associated with message
 920 *
 921 * Note: Ancillary data is not captured if not requested by receiver.
 922 *
 923 * Returns 0 if successful, otherwise errno
 924 */
 925static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
 926			 struct tipc_port *tport)
 927{
 928	u32 anc_data[3];
 929	u32 err;
 930	u32 dest_type;
 931	int has_name;
 932	int res;
 933
 934	if (likely(m->msg_controllen == 0))
 935		return 0;
 936
 937	/* Optionally capture errored message object(s) */
 938	err = msg ? msg_errcode(msg) : 0;
 939	if (unlikely(err)) {
 940		anc_data[0] = err;
 941		anc_data[1] = msg_data_sz(msg);
 942		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
 943		if (res)
 944			return res;
 945		if (anc_data[1]) {
 946			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
 947				       msg_data(msg));
 948			if (res)
 949				return res;
 950		}
 951	}
 952
 953	/* Optionally capture message destination object */
 954	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
 955	switch (dest_type) {
 956	case TIPC_NAMED_MSG:
 957		has_name = 1;
 958		anc_data[0] = msg_nametype(msg);
 959		anc_data[1] = msg_namelower(msg);
 960		anc_data[2] = msg_namelower(msg);
 961		break;
 962	case TIPC_MCAST_MSG:
 963		has_name = 1;
 964		anc_data[0] = msg_nametype(msg);
 965		anc_data[1] = msg_namelower(msg);
 966		anc_data[2] = msg_nameupper(msg);
 967		break;
 968	case TIPC_CONN_MSG:
 969		has_name = (tport->conn_type != 0);
 970		anc_data[0] = tport->conn_type;
 971		anc_data[1] = tport->conn_instance;
 972		anc_data[2] = tport->conn_instance;
 973		break;
 974	default:
 975		has_name = 0;
 976	}
 977	if (has_name) {
 978		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
 979		if (res)
 980			return res;
 981	}
 982
 983	return 0;
 984}
 985
 986static int tipc_wait_for_rcvmsg(struct socket *sock, long timeo)
 987{
 988	struct sock *sk = sock->sk;
 989	DEFINE_WAIT(wait);
 990	int err;
 991
 992	for (;;) {
 993		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
 994		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
 995			if (sock->state == SS_DISCONNECTING) {
 996				err = -ENOTCONN;
 997				break;
 998			}
 999			release_sock(sk);
1000			timeo = schedule_timeout(timeo);
1001			lock_sock(sk);
1002		}
1003		err = 0;
1004		if (!skb_queue_empty(&sk->sk_receive_queue))
1005			break;
1006		err = sock_intr_errno(timeo);
1007		if (signal_pending(current))
1008			break;
1009		err = -EAGAIN;
1010		if (!timeo)
1011			break;
1012	}
1013	finish_wait(sk_sleep(sk), &wait);
1014	return err;
1015}
1016
1017/**
1018 * tipc_recvmsg - receive packet-oriented message
1019 * @iocb: (unused)
1020 * @m: descriptor for message info
1021 * @buf_len: total size of user buffer area
1022 * @flags: receive flags
1023 *
1024 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1025 * If the complete message doesn't fit in user area, truncate it.
1026 *
1027 * Returns size of returned message data, errno otherwise
1028 */
1029static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1030			struct msghdr *m, size_t buf_len, int flags)
1031{
1032	struct sock *sk = sock->sk;
1033	struct tipc_sock *tsk = tipc_sk(sk);
1034	struct tipc_port *port = &tsk->port;
1035	struct sk_buff *buf;
1036	struct tipc_msg *msg;
1037	long timeo;
1038	unsigned int sz;
1039	u32 err;
1040	int res;
1041
1042	/* Catch invalid receive requests */
1043	if (unlikely(!buf_len))
1044		return -EINVAL;
1045
1046	lock_sock(sk);
1047
1048	if (unlikely(sock->state == SS_UNCONNECTED)) {
1049		res = -ENOTCONN;
1050		goto exit;
1051	}
1052
1053	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1054restart:
1055
1056	/* Look for a message in receive queue; wait if necessary */
1057	res = tipc_wait_for_rcvmsg(sock, timeo);
1058	if (res)
1059		goto exit;
 
 
 
 
 
 
 
 
 
 
 
 
1060
1061	/* Look at first message in receive queue */
1062	buf = skb_peek(&sk->sk_receive_queue);
1063	msg = buf_msg(buf);
1064	sz = msg_data_sz(msg);
1065	err = msg_errcode(msg);
1066
 
 
 
 
 
 
 
1067	/* Discard an empty non-errored message & try again */
1068	if ((!sz) && (!err)) {
1069		advance_rx_queue(sk);
1070		goto restart;
1071	}
1072
1073	/* Capture sender's address (optional) */
1074	set_orig_addr(m, msg);
1075
1076	/* Capture ancillary data (optional) */
1077	res = anc_data_recv(m, msg, port);
1078	if (res)
1079		goto exit;
1080
1081	/* Capture message data (if valid) & compute return value (always) */
1082	if (!err) {
1083		if (unlikely(buf_len < sz)) {
1084			sz = buf_len;
1085			m->msg_flags |= MSG_TRUNC;
1086		}
1087		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1088					      m->msg_iov, sz);
1089		if (res)
1090			goto exit;
1091		res = sz;
1092	} else {
1093		if ((sock->state == SS_READY) ||
1094		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1095			res = 0;
1096		else
1097			res = -ECONNRESET;
1098	}
1099
1100	/* Consume received message (optional) */
1101	if (likely(!(flags & MSG_PEEK))) {
1102		if ((sock->state != SS_READY) &&
1103		    (++port->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1104			tipc_acknowledge(port->ref, port->conn_unacked);
1105		advance_rx_queue(sk);
1106	}
1107exit:
1108	release_sock(sk);
1109	return res;
1110}
1111
1112/**
1113 * tipc_recv_stream - receive stream-oriented data
1114 * @iocb: (unused)
1115 * @m: descriptor for message info
1116 * @buf_len: total size of user buffer area
1117 * @flags: receive flags
1118 *
1119 * Used for SOCK_STREAM messages only.  If not enough data is available
1120 * will optionally wait for more; never truncates data.
1121 *
1122 * Returns size of returned message data, errno otherwise
1123 */
1124static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1125			    struct msghdr *m, size_t buf_len, int flags)
1126{
1127	struct sock *sk = sock->sk;
1128	struct tipc_sock *tsk = tipc_sk(sk);
1129	struct tipc_port *port = &tsk->port;
1130	struct sk_buff *buf;
1131	struct tipc_msg *msg;
1132	long timeo;
1133	unsigned int sz;
1134	int sz_to_copy, target, needed;
1135	int sz_copied = 0;
1136	u32 err;
1137	int res = 0;
1138
1139	/* Catch invalid receive attempts */
1140	if (unlikely(!buf_len))
1141		return -EINVAL;
1142
1143	lock_sock(sk);
1144
1145	if (unlikely(sock->state == SS_UNCONNECTED)) {
 
1146		res = -ENOTCONN;
1147		goto exit;
1148	}
1149
1150	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1151	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1152
1153restart:
1154	/* Look for a message in receive queue; wait if necessary */
1155	res = tipc_wait_for_rcvmsg(sock, timeo);
1156	if (res)
1157		goto exit;
 
 
 
 
 
 
 
 
 
 
 
 
1158
1159	/* Look at first message in receive queue */
1160	buf = skb_peek(&sk->sk_receive_queue);
1161	msg = buf_msg(buf);
1162	sz = msg_data_sz(msg);
1163	err = msg_errcode(msg);
1164
1165	/* Discard an empty non-errored message & try again */
1166	if ((!sz) && (!err)) {
1167		advance_rx_queue(sk);
1168		goto restart;
1169	}
1170
1171	/* Optionally capture sender's address & ancillary data of first msg */
1172	if (sz_copied == 0) {
1173		set_orig_addr(m, msg);
1174		res = anc_data_recv(m, msg, port);
1175		if (res)
1176			goto exit;
1177	}
1178
1179	/* Capture message data (if valid) & compute return value (always) */
1180	if (!err) {
1181		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1182
1183		sz -= offset;
1184		needed = (buf_len - sz_copied);
1185		sz_to_copy = (sz <= needed) ? sz : needed;
1186
1187		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1188					      m->msg_iov, sz_to_copy);
1189		if (res)
1190			goto exit;
1191
1192		sz_copied += sz_to_copy;
1193
1194		if (sz_to_copy < sz) {
1195			if (!(flags & MSG_PEEK))
1196				TIPC_SKB_CB(buf)->handle =
1197				(void *)(unsigned long)(offset + sz_to_copy);
1198			goto exit;
1199		}
1200	} else {
1201		if (sz_copied != 0)
1202			goto exit; /* can't add error msg to valid data */
1203
1204		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1205			res = 0;
1206		else
1207			res = -ECONNRESET;
1208	}
1209
1210	/* Consume received message (optional) */
1211	if (likely(!(flags & MSG_PEEK))) {
1212		if (unlikely(++port->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1213			tipc_acknowledge(port->ref, port->conn_unacked);
1214		advance_rx_queue(sk);
1215	}
1216
1217	/* Loop around if more data is required */
1218	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1219	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1220	    (sz_copied < target)) &&	/* and more is ready or required */
1221	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1222	    (!err))			/* and haven't reached a FIN */
1223		goto restart;
1224
1225exit:
1226	release_sock(sk);
1227	return sz_copied ? sz_copied : res;
1228}
1229
1230/**
1231 * tipc_write_space - wake up thread if port congestion is released
1232 * @sk: socket
1233 */
1234static void tipc_write_space(struct sock *sk)
1235{
1236	struct socket_wq *wq;
1237
1238	rcu_read_lock();
1239	wq = rcu_dereference(sk->sk_wq);
1240	if (wq_has_sleeper(wq))
1241		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1242						POLLWRNORM | POLLWRBAND);
1243	rcu_read_unlock();
1244}
1245
1246/**
1247 * tipc_data_ready - wake up threads to indicate messages have been received
1248 * @sk: socket
1249 * @len: the length of messages
1250 */
1251static void tipc_data_ready(struct sock *sk)
1252{
1253	struct socket_wq *wq;
1254
1255	rcu_read_lock();
1256	wq = rcu_dereference(sk->sk_wq);
1257	if (wq_has_sleeper(wq))
1258		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1259						POLLRDNORM | POLLRDBAND);
1260	rcu_read_unlock();
1261}
1262
1263/**
1264 * filter_connect - Handle all incoming messages for a connection-based socket
1265 * @tsk: TIPC socket
1266 * @msg: message
1267 *
1268 * Returns TIPC error status code and socket error status code
1269 * once it encounters some errors
1270 */
1271static u32 filter_connect(struct tipc_sock *tsk, struct sk_buff **buf)
1272{
1273	struct sock *sk = &tsk->sk;
1274	struct tipc_port *port = &tsk->port;
1275	struct socket *sock = sk->sk_socket;
1276	struct tipc_msg *msg = buf_msg(*buf);
1277
1278	u32 retval = TIPC_ERR_NO_PORT;
1279	int res;
1280
1281	if (msg_mcast(msg))
1282		return retval;
1283
1284	switch ((int)sock->state) {
1285	case SS_CONNECTED:
1286		/* Accept only connection-based messages sent by peer */
1287		if (msg_connected(msg) && tipc_port_peer_msg(port, msg)) {
1288			if (unlikely(msg_errcode(msg))) {
1289				sock->state = SS_DISCONNECTING;
1290				__tipc_port_disconnect(port);
1291			}
1292			retval = TIPC_OK;
1293		}
1294		break;
1295	case SS_CONNECTING:
1296		/* Accept only ACK or NACK message */
1297		if (unlikely(msg_errcode(msg))) {
1298			sock->state = SS_DISCONNECTING;
1299			sk->sk_err = ECONNREFUSED;
1300			retval = TIPC_OK;
1301			break;
1302		}
1303
1304		if (unlikely(!msg_connected(msg)))
1305			break;
1306
1307		res = auto_connect(tsk, msg);
1308		if (res) {
1309			sock->state = SS_DISCONNECTING;
1310			sk->sk_err = -res;
1311			retval = TIPC_OK;
1312			break;
1313		}
1314
1315		/* If an incoming message is an 'ACK-', it should be
1316		 * discarded here because it doesn't contain useful
1317		 * data. In addition, we should try to wake up
1318		 * connect() routine if sleeping.
1319		 */
1320		if (msg_data_sz(msg) == 0) {
1321			kfree_skb(*buf);
1322			*buf = NULL;
1323			if (waitqueue_active(sk_sleep(sk)))
1324				wake_up_interruptible(sk_sleep(sk));
1325		}
1326		retval = TIPC_OK;
1327		break;
1328	case SS_LISTENING:
1329	case SS_UNCONNECTED:
1330		/* Accept only SYN message */
1331		if (!msg_connected(msg) && !(msg_errcode(msg)))
1332			retval = TIPC_OK;
1333		break;
1334	case SS_DISCONNECTING:
1335		break;
1336	default:
1337		pr_err("Unknown socket state %u\n", sock->state);
1338	}
1339	return retval;
1340}
1341
1342/**
1343 * rcvbuf_limit - get proper overload limit of socket receive queue
1344 * @sk: socket
1345 * @buf: message
1346 *
1347 * For all connection oriented messages, irrespective of importance,
1348 * the default overload value (i.e. 67MB) is set as limit.
1349 *
1350 * For all connectionless messages, by default new queue limits are
1351 * as belows:
1352 *
1353 * TIPC_LOW_IMPORTANCE       (4 MB)
1354 * TIPC_MEDIUM_IMPORTANCE    (8 MB)
1355 * TIPC_HIGH_IMPORTANCE      (16 MB)
1356 * TIPC_CRITICAL_IMPORTANCE  (32 MB)
1357 *
1358 * Returns overload limit according to corresponding message importance
1359 */
1360static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1361{
1362	struct tipc_msg *msg = buf_msg(buf);
1363
1364	if (msg_connected(msg))
1365		return sysctl_tipc_rmem[2];
1366
1367	return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1368		msg_importance(msg);
1369}
1370
1371/**
1372 * filter_rcv - validate incoming message
1373 * @sk: socket
1374 * @buf: message
1375 *
1376 * Enqueues message on receive queue if acceptable; optionally handles
1377 * disconnect indication for a connected socket.
1378 *
1379 * Called with socket lock already taken; port lock may also be taken.
1380 *
1381 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1382 */
1383static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1384{
1385	struct socket *sock = sk->sk_socket;
1386	struct tipc_sock *tsk = tipc_sk(sk);
1387	struct tipc_msg *msg = buf_msg(buf);
1388	unsigned int limit = rcvbuf_limit(sk, buf);
1389	u32 res = TIPC_OK;
1390
1391	/* Reject message if it is wrong sort of message for socket */
1392	if (msg_type(msg) > TIPC_DIRECT_MSG)
1393		return TIPC_ERR_NO_PORT;
1394
1395	if (sock->state == SS_READY) {
1396		if (msg_connected(msg))
1397			return TIPC_ERR_NO_PORT;
1398	} else {
1399		res = filter_connect(tsk, &buf);
1400		if (res != TIPC_OK || buf == NULL)
1401			return res;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1402	}
1403
1404	/* Reject message if there isn't room to queue it */
1405	if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1406		return TIPC_ERR_OVERLOAD;
 
 
 
 
 
 
 
 
1407
1408	/* Enqueue message */
1409	TIPC_SKB_CB(buf)->handle = NULL;
 
1410	__skb_queue_tail(&sk->sk_receive_queue, buf);
1411	skb_set_owner_r(buf, sk);
1412
1413	sk->sk_data_ready(sk);
 
 
 
 
 
 
 
1414	return TIPC_OK;
1415}
1416
1417/**
1418 * backlog_rcv - handle incoming message from backlog queue
1419 * @sk: socket
1420 * @buf: message
1421 *
1422 * Caller must hold socket lock, but not port lock.
1423 *
1424 * Returns 0
1425 */
1426static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1427{
1428	u32 res;
1429
1430	res = filter_rcv(sk, buf);
1431	if (res)
1432		tipc_reject_msg(buf, res);
1433	return 0;
1434}
1435
1436/**
1437 * tipc_sk_rcv - handle incoming message
1438 * @sk:  socket receiving message
1439 * @buf: message
1440 *
1441 * Called with port lock already taken.
1442 *
1443 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1444 */
1445u32 tipc_sk_rcv(struct sock *sk, struct sk_buff *buf)
1446{
 
1447	u32 res;
1448
1449	/*
1450	 * Process message if socket is unlocked; otherwise add to backlog queue
1451	 *
1452	 * This code is based on sk_receive_skb(), but must be distinct from it
1453	 * since a TIPC-specific filter/reject mechanism is utilized
1454	 */
1455	bh_lock_sock(sk);
1456	if (!sock_owned_by_user(sk)) {
1457		res = filter_rcv(sk, buf);
1458	} else {
1459		if (sk_add_backlog(sk, buf, rcvbuf_limit(sk, buf)))
1460			res = TIPC_ERR_OVERLOAD;
1461		else
1462			res = TIPC_OK;
1463	}
1464	bh_unlock_sock(sk);
1465
1466	return res;
1467}
1468
1469static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
 
 
 
 
 
 
1470{
1471	struct sock *sk = sock->sk;
1472	DEFINE_WAIT(wait);
1473	int done;
1474
1475	do {
1476		int err = sock_error(sk);
1477		if (err)
1478			return err;
1479		if (!*timeo_p)
1480			return -ETIMEDOUT;
1481		if (signal_pending(current))
1482			return sock_intr_errno(*timeo_p);
1483
1484		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1485		done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1486		finish_wait(sk_sleep(sk), &wait);
1487	} while (!done);
1488	return 0;
1489}
1490
1491/**
1492 * tipc_connect - establish a connection to another TIPC port
1493 * @sock: socket structure
1494 * @dest: socket address for destination port
1495 * @destlen: size of socket address data structure
1496 * @flags: file-related flags associated with socket
1497 *
1498 * Returns 0 on success, errno otherwise
1499 */
1500static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1501			int destlen, int flags)
1502{
1503	struct sock *sk = sock->sk;
1504	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1505	struct msghdr m = {NULL,};
1506	long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1507	socket_state previous;
 
1508	int res;
1509
1510	lock_sock(sk);
1511
1512	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1513	if (sock->state == SS_READY) {
1514		res = -EOPNOTSUPP;
1515		goto exit;
1516	}
1517
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1518	/*
1519	 * Reject connection attempt using multicast address
1520	 *
1521	 * Note: send_msg() validates the rest of the address fields,
1522	 *       so there's no need to do it here
1523	 */
1524	if (dst->addrtype == TIPC_ADDR_MCAST) {
1525		res = -EINVAL;
1526		goto exit;
1527	}
1528
1529	previous = sock->state;
1530	switch (sock->state) {
1531	case SS_UNCONNECTED:
1532		/* Send a 'SYN-' to destination */
1533		m.msg_name = dest;
1534		m.msg_namelen = destlen;
1535
1536		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1537		 * indicate send_msg() is never blocked.
1538		 */
1539		if (!timeout)
1540			m.msg_flags = MSG_DONTWAIT;
 
1541
1542		res = tipc_sendmsg(NULL, sock, &m, 0);
1543		if ((res < 0) && (res != -EWOULDBLOCK))
1544			goto exit;
 
 
 
 
 
 
1545
1546		/* Just entered SS_CONNECTING state; the only
1547		 * difference is that return value in non-blocking
1548		 * case is EINPROGRESS, rather than EALREADY.
1549		 */
1550		res = -EINPROGRESS;
1551	case SS_CONNECTING:
1552		if (previous == SS_CONNECTING)
1553			res = -EALREADY;
1554		if (!timeout)
1555			goto exit;
1556		timeout = msecs_to_jiffies(timeout);
1557		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1558		res = tipc_wait_for_connect(sock, &timeout);
1559		break;
1560	case SS_CONNECTED:
1561		res = -EISCONN;
1562		break;
1563	default:
1564		res = -EINVAL;
1565		break;
 
1566	}
 
1567exit:
1568	release_sock(sk);
1569	return res;
1570}
1571
1572/**
1573 * tipc_listen - allow socket to listen for incoming connections
1574 * @sock: socket structure
1575 * @len: (unused)
1576 *
1577 * Returns 0 on success, errno otherwise
1578 */
1579static int tipc_listen(struct socket *sock, int len)
1580{
1581	struct sock *sk = sock->sk;
1582	int res;
1583
1584	lock_sock(sk);
1585
1586	if (sock->state != SS_UNCONNECTED)
1587		res = -EINVAL;
1588	else {
1589		sock->state = SS_LISTENING;
1590		res = 0;
1591	}
1592
1593	release_sock(sk);
1594	return res;
1595}
1596
1597static int tipc_wait_for_accept(struct socket *sock, long timeo)
1598{
1599	struct sock *sk = sock->sk;
1600	DEFINE_WAIT(wait);
1601	int err;
1602
1603	/* True wake-one mechanism for incoming connections: only
1604	 * one process gets woken up, not the 'whole herd'.
1605	 * Since we do not 'race & poll' for established sockets
1606	 * anymore, the common case will execute the loop only once.
1607	*/
1608	for (;;) {
1609		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1610					  TASK_INTERRUPTIBLE);
1611		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1612			release_sock(sk);
1613			timeo = schedule_timeout(timeo);
1614			lock_sock(sk);
1615		}
1616		err = 0;
1617		if (!skb_queue_empty(&sk->sk_receive_queue))
1618			break;
1619		err = -EINVAL;
1620		if (sock->state != SS_LISTENING)
1621			break;
1622		err = sock_intr_errno(timeo);
1623		if (signal_pending(current))
1624			break;
1625		err = -EAGAIN;
1626		if (!timeo)
1627			break;
1628	}
1629	finish_wait(sk_sleep(sk), &wait);
1630	return err;
1631}
1632
1633/**
1634 * tipc_accept - wait for connection request
1635 * @sock: listening socket
1636 * @newsock: new socket that is to be connected
1637 * @flags: file-related flags associated with socket
1638 *
1639 * Returns 0 on success, errno otherwise
1640 */
1641static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1642{
1643	struct sock *new_sk, *sk = sock->sk;
1644	struct sk_buff *buf;
1645	struct tipc_port *new_port;
1646	struct tipc_msg *msg;
1647	struct tipc_portid peer;
1648	u32 new_ref;
1649	long timeo;
1650	int res;
1651
1652	lock_sock(sk);
1653
1654	if (sock->state != SS_LISTENING) {
1655		res = -EINVAL;
1656		goto exit;
1657	}
1658	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1659	res = tipc_wait_for_accept(sock, timeo);
1660	if (res)
1661		goto exit;
1662
1663	buf = skb_peek(&sk->sk_receive_queue);
1664
1665	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
1666	if (res)
1667		goto exit;
1668
1669	new_sk = new_sock->sk;
1670	new_port = &tipc_sk(new_sk)->port;
1671	new_ref = new_port->ref;
1672	msg = buf_msg(buf);
 
 
 
 
1673
1674	/* we lock on new_sk; but lockdep sees the lock on sk */
1675	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1676
1677	/*
1678	 * Reject any stray messages received by new socket
1679	 * before the socket lock was taken (very, very unlikely)
1680	 */
1681	reject_rx_queue(new_sk);
 
 
 
 
 
 
 
 
 
 
1682
1683	/* Connect new socket to it's peer */
1684	peer.ref = msg_origport(msg);
1685	peer.node = msg_orignode(msg);
1686	tipc_port_connect(new_ref, &peer);
1687	new_sock->state = SS_CONNECTED;
1688
1689	tipc_port_set_importance(new_port, msg_importance(msg));
1690	if (msg_named(msg)) {
1691		new_port->conn_type = msg_nametype(msg);
1692		new_port->conn_instance = msg_nameinst(msg);
1693	}
1694
1695	/*
1696	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1697	 * Respond to 'SYN+' by queuing it on new socket.
1698	 */
1699	if (!msg_data_sz(msg)) {
1700		struct msghdr m = {NULL,};
1701
1702		advance_rx_queue(sk);
1703		tipc_send_packet(NULL, new_sock, &m, 0);
1704	} else {
1705		__skb_dequeue(&sk->sk_receive_queue);
1706		__skb_queue_head(&new_sk->sk_receive_queue, buf);
1707		skb_set_owner_r(buf, new_sk);
 
1708	}
1709	release_sock(new_sk);
1710exit:
1711	release_sock(sk);
1712	return res;
1713}
1714
1715/**
1716 * tipc_shutdown - shutdown socket connection
1717 * @sock: socket structure
1718 * @how: direction to close (must be SHUT_RDWR)
1719 *
1720 * Terminates connection (if necessary), then purges socket's receive queue.
1721 *
1722 * Returns 0 on success, errno otherwise
1723 */
1724static int tipc_shutdown(struct socket *sock, int how)
1725{
1726	struct sock *sk = sock->sk;
1727	struct tipc_sock *tsk = tipc_sk(sk);
1728	struct tipc_port *port = &tsk->port;
1729	struct sk_buff *buf;
1730	int res;
1731
1732	if (how != SHUT_RDWR)
1733		return -EINVAL;
1734
1735	lock_sock(sk);
1736
1737	switch (sock->state) {
1738	case SS_CONNECTING:
1739	case SS_CONNECTED:
1740
1741restart:
1742		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1743		buf = __skb_dequeue(&sk->sk_receive_queue);
1744		if (buf) {
1745			if (TIPC_SKB_CB(buf)->handle != NULL) {
 
1746				kfree_skb(buf);
1747				goto restart;
1748			}
1749			tipc_port_disconnect(port->ref);
1750			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1751		} else {
1752			tipc_port_shutdown(port->ref);
1753		}
1754
1755		sock->state = SS_DISCONNECTING;
1756
1757		/* fall through */
1758
1759	case SS_DISCONNECTING:
1760
1761		/* Discard any unreceived messages */
1762		__skb_queue_purge(&sk->sk_receive_queue);
1763
1764		/* Wake up anyone sleeping in poll */
1765		sk->sk_state_change(sk);
1766		res = 0;
1767		break;
1768
1769	default:
1770		res = -ENOTCONN;
1771	}
1772
1773	release_sock(sk);
1774	return res;
1775}
1776
1777/**
1778 * tipc_setsockopt - set socket option
1779 * @sock: socket structure
1780 * @lvl: option level
1781 * @opt: option identifier
1782 * @ov: pointer to new option value
1783 * @ol: length of option value
1784 *
1785 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1786 * (to ease compatibility).
1787 *
1788 * Returns 0 on success, errno otherwise
1789 */
1790static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
1791			   char __user *ov, unsigned int ol)
1792{
1793	struct sock *sk = sock->sk;
1794	struct tipc_sock *tsk = tipc_sk(sk);
1795	struct tipc_port *port = &tsk->port;
1796	u32 value;
1797	int res;
1798
1799	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1800		return 0;
1801	if (lvl != SOL_TIPC)
1802		return -ENOPROTOOPT;
1803	if (ol < sizeof(value))
1804		return -EINVAL;
1805	res = get_user(value, (u32 __user *)ov);
1806	if (res)
1807		return res;
1808
1809	lock_sock(sk);
1810
1811	switch (opt) {
1812	case TIPC_IMPORTANCE:
1813		tipc_port_set_importance(port, value);
1814		break;
1815	case TIPC_SRC_DROPPABLE:
1816		if (sock->type != SOCK_STREAM)
1817			tipc_port_set_unreliable(port, value);
1818		else
1819			res = -ENOPROTOOPT;
1820		break;
1821	case TIPC_DEST_DROPPABLE:
1822		tipc_port_set_unreturnable(port, value);
1823		break;
1824	case TIPC_CONN_TIMEOUT:
1825		tipc_sk(sk)->conn_timeout = value;
1826		/* no need to set "res", since already 0 at this point */
1827		break;
1828	default:
1829		res = -EINVAL;
1830	}
1831
1832	release_sock(sk);
1833
1834	return res;
1835}
1836
1837/**
1838 * tipc_getsockopt - get socket option
1839 * @sock: socket structure
1840 * @lvl: option level
1841 * @opt: option identifier
1842 * @ov: receptacle for option value
1843 * @ol: receptacle for length of option value
1844 *
1845 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1846 * (to ease compatibility).
1847 *
1848 * Returns 0 on success, errno otherwise
1849 */
1850static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
1851			   char __user *ov, int __user *ol)
1852{
1853	struct sock *sk = sock->sk;
1854	struct tipc_sock *tsk = tipc_sk(sk);
1855	struct tipc_port *port = &tsk->port;
1856	int len;
1857	u32 value;
1858	int res;
1859
1860	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1861		return put_user(0, ol);
1862	if (lvl != SOL_TIPC)
1863		return -ENOPROTOOPT;
1864	res = get_user(len, ol);
1865	if (res)
1866		return res;
1867
1868	lock_sock(sk);
1869
1870	switch (opt) {
1871	case TIPC_IMPORTANCE:
1872		value = tipc_port_importance(port);
1873		break;
1874	case TIPC_SRC_DROPPABLE:
1875		value = tipc_port_unreliable(port);
1876		break;
1877	case TIPC_DEST_DROPPABLE:
1878		value = tipc_port_unreturnable(port);
1879		break;
1880	case TIPC_CONN_TIMEOUT:
1881		value = tipc_sk(sk)->conn_timeout;
1882		/* no need to set "res", since already 0 at this point */
1883		break;
1884	case TIPC_NODE_RECVQ_DEPTH:
1885		value = 0; /* was tipc_queue_size, now obsolete */
1886		break;
1887	case TIPC_SOCK_RECVQ_DEPTH:
1888		value = skb_queue_len(&sk->sk_receive_queue);
1889		break;
1890	default:
1891		res = -EINVAL;
1892	}
1893
1894	release_sock(sk);
1895
1896	if (res)
1897		return res;	/* "get" failed */
1898
1899	if (len < sizeof(value))
1900		return -EINVAL;
1901
1902	if (copy_to_user(ov, &value, sizeof(value)))
1903		return -EFAULT;
1904
1905	return put_user(sizeof(value), ol);
1906}
1907
1908/* Protocol switches for the various types of TIPC sockets */
1909
 
1910static const struct proto_ops msg_ops = {
1911	.owner		= THIS_MODULE,
1912	.family		= AF_TIPC,
1913	.release	= tipc_release,
1914	.bind		= tipc_bind,
1915	.connect	= tipc_connect,
1916	.socketpair	= sock_no_socketpair,
1917	.accept		= sock_no_accept,
1918	.getname	= tipc_getname,
1919	.poll		= tipc_poll,
1920	.ioctl		= sock_no_ioctl,
1921	.listen		= sock_no_listen,
1922	.shutdown	= tipc_shutdown,
1923	.setsockopt	= tipc_setsockopt,
1924	.getsockopt	= tipc_getsockopt,
1925	.sendmsg	= tipc_sendmsg,
1926	.recvmsg	= tipc_recvmsg,
1927	.mmap		= sock_no_mmap,
1928	.sendpage	= sock_no_sendpage
1929};
1930
1931static const struct proto_ops packet_ops = {
1932	.owner		= THIS_MODULE,
1933	.family		= AF_TIPC,
1934	.release	= tipc_release,
1935	.bind		= tipc_bind,
1936	.connect	= tipc_connect,
1937	.socketpair	= sock_no_socketpair,
1938	.accept		= tipc_accept,
1939	.getname	= tipc_getname,
1940	.poll		= tipc_poll,
1941	.ioctl		= sock_no_ioctl,
1942	.listen		= tipc_listen,
1943	.shutdown	= tipc_shutdown,
1944	.setsockopt	= tipc_setsockopt,
1945	.getsockopt	= tipc_getsockopt,
1946	.sendmsg	= tipc_send_packet,
1947	.recvmsg	= tipc_recvmsg,
1948	.mmap		= sock_no_mmap,
1949	.sendpage	= sock_no_sendpage
1950};
1951
1952static const struct proto_ops stream_ops = {
1953	.owner		= THIS_MODULE,
1954	.family		= AF_TIPC,
1955	.release	= tipc_release,
1956	.bind		= tipc_bind,
1957	.connect	= tipc_connect,
1958	.socketpair	= sock_no_socketpair,
1959	.accept		= tipc_accept,
1960	.getname	= tipc_getname,
1961	.poll		= tipc_poll,
1962	.ioctl		= sock_no_ioctl,
1963	.listen		= tipc_listen,
1964	.shutdown	= tipc_shutdown,
1965	.setsockopt	= tipc_setsockopt,
1966	.getsockopt	= tipc_getsockopt,
1967	.sendmsg	= tipc_send_stream,
1968	.recvmsg	= tipc_recv_stream,
1969	.mmap		= sock_no_mmap,
1970	.sendpage	= sock_no_sendpage
1971};
1972
1973static const struct net_proto_family tipc_family_ops = {
1974	.owner		= THIS_MODULE,
1975	.family		= AF_TIPC,
1976	.create		= tipc_sk_create
1977};
1978
1979static struct proto tipc_proto = {
1980	.name		= "TIPC",
1981	.owner		= THIS_MODULE,
1982	.obj_size	= sizeof(struct tipc_sock),
1983	.sysctl_rmem	= sysctl_tipc_rmem
1984};
1985
1986static struct proto tipc_proto_kern = {
1987	.name		= "TIPC",
1988	.obj_size	= sizeof(struct tipc_sock),
1989	.sysctl_rmem	= sysctl_tipc_rmem
1990};
1991
1992/**
1993 * tipc_socket_init - initialize TIPC socket interface
1994 *
1995 * Returns 0 on success, errno otherwise
1996 */
1997int tipc_socket_init(void)
1998{
1999	int res;
2000
2001	res = proto_register(&tipc_proto, 1);
2002	if (res) {
2003		pr_err("Failed to register TIPC protocol type\n");
2004		goto out;
2005	}
2006
2007	res = sock_register(&tipc_family_ops);
2008	if (res) {
2009		pr_err("Failed to register TIPC socket type\n");
2010		proto_unregister(&tipc_proto);
2011		goto out;
2012	}
 
 
2013 out:
2014	return res;
2015}
2016
2017/**
2018 * tipc_socket_stop - stop TIPC socket interface
2019 */
2020void tipc_socket_stop(void)
2021{
 
 
 
 
2022	sock_unregister(tipc_family_ops.family);
2023	proto_unregister(&tipc_proto);
2024}