Linux Audio

Check our new training course

Yocto / OpenEmbedded training

Feb 10-13, 2025
Register
Loading...
v3.15
   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}
v4.10.11
   1/*
   2 * net/tipc/socket.c: TIPC socket API
   3 *
   4 * Copyright (c) 2001-2007, 2012-2016, 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 <linux/rhashtable.h>
  38#include "core.h"
  39#include "name_table.h"
  40#include "node.h"
  41#include "link.h"
  42#include "name_distr.h"
  43#include "socket.h"
  44#include "bcast.h"
  45#include "netlink.h"
  46
  47#define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
  48#define CONN_PROBING_INTERVAL	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
  49#define TIPC_FWD_MSG		1
  50#define TIPC_MAX_PORT		0xffffffff
  51#define TIPC_MIN_PORT		1
  52
  53enum {
  54	TIPC_LISTEN = TCP_LISTEN,
  55	TIPC_ESTABLISHED = TCP_ESTABLISHED,
  56	TIPC_OPEN = TCP_CLOSE,
  57	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
  58	TIPC_CONNECTING = TCP_SYN_SENT,
  59};
  60
  61/**
  62 * struct tipc_sock - TIPC socket structure
  63 * @sk: socket - interacts with 'port' and with user via the socket API
  64 * @conn_type: TIPC type used when connection was established
  65 * @conn_instance: TIPC instance used when connection was established
  66 * @published: non-zero if port has one or more associated names
  67 * @max_pkt: maximum packet size "hint" used when building messages sent by port
  68 * @portid: unique port identity in TIPC socket hash table
  69 * @phdr: preformatted message header used when sending messages
  70 * @publications: list of publications for port
  71 * @pub_count: total # of publications port has made during its lifetime
  72 * @probing_state:
  73 * @conn_timeout: the time we can wait for an unresponded setup request
  74 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
  75 * @link_cong: non-zero if owner must sleep because of link congestion
  76 * @sent_unacked: # messages sent by socket, and not yet acked by peer
  77 * @rcv_unacked: # messages read by user, but not yet acked back to peer
  78 * @peer: 'connected' peer for dgram/rdm
  79 * @node: hash table node
  80 * @rcu: rcu struct for tipc_sock
  81 */
  82struct tipc_sock {
  83	struct sock sk;
  84	u32 conn_type;
  85	u32 conn_instance;
  86	int published;
  87	u32 max_pkt;
  88	u32 portid;
  89	struct tipc_msg phdr;
  90	struct list_head sock_list;
  91	struct list_head publications;
  92	u32 pub_count;
  93	uint conn_timeout;
  94	atomic_t dupl_rcvcnt;
  95	bool probe_unacked;
  96	bool link_cong;
  97	u16 snt_unacked;
  98	u16 snd_win;
  99	u16 peer_caps;
 100	u16 rcv_unacked;
 101	u16 rcv_win;
 102	struct sockaddr_tipc peer;
 103	struct rhash_head node;
 104	struct rcu_head rcu;
 105};
 106
 107static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
 108static void tipc_data_ready(struct sock *sk);
 109static void tipc_write_space(struct sock *sk);
 110static void tipc_sock_destruct(struct sock *sk);
 111static int tipc_release(struct socket *sock);
 112static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
 113static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
 114static void tipc_sk_timeout(unsigned long data);
 115static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
 116			   struct tipc_name_seq const *seq);
 117static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
 118			    struct tipc_name_seq const *seq);
 119static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
 120static int tipc_sk_insert(struct tipc_sock *tsk);
 121static void tipc_sk_remove(struct tipc_sock *tsk);
 122static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
 123			      size_t dsz);
 124static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
 125
 126static const struct proto_ops packet_ops;
 127static const struct proto_ops stream_ops;
 128static const struct proto_ops msg_ops;
 
 129static struct proto tipc_proto;
 130static const struct rhashtable_params tsk_rht_params;
 131
 132static u32 tsk_own_node(struct tipc_sock *tsk)
 133{
 134	return msg_prevnode(&tsk->phdr);
 135}
 136
 137static u32 tsk_peer_node(struct tipc_sock *tsk)
 138{
 139	return msg_destnode(&tsk->phdr);
 140}
 141
 142static u32 tsk_peer_port(struct tipc_sock *tsk)
 143{
 144	return msg_destport(&tsk->phdr);
 145}
 146
 147static  bool tsk_unreliable(struct tipc_sock *tsk)
 148{
 149	return msg_src_droppable(&tsk->phdr) != 0;
 150}
 151
 152static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
 153{
 154	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
 155}
 156
 157static bool tsk_unreturnable(struct tipc_sock *tsk)
 158{
 159	return msg_dest_droppable(&tsk->phdr) != 0;
 160}
 161
 162static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
 163{
 164	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
 165}
 166
 167static int tsk_importance(struct tipc_sock *tsk)
 168{
 169	return msg_importance(&tsk->phdr);
 170}
 171
 172static int tsk_set_importance(struct tipc_sock *tsk, int imp)
 173{
 174	if (imp > TIPC_CRITICAL_IMPORTANCE)
 175		return -EINVAL;
 176	msg_set_importance(&tsk->phdr, (u32)imp);
 177	return 0;
 178}
 179
 180static struct tipc_sock *tipc_sk(const struct sock *sk)
 181{
 182	return container_of(sk, struct tipc_sock, sk);
 183}
 184
 185static bool tsk_conn_cong(struct tipc_sock *tsk)
 186{
 187	return tsk->snt_unacked > tsk->snd_win;
 188}
 189
 190/* tsk_blocks(): translate a buffer size in bytes to number of
 191 * advertisable blocks, taking into account the ratio truesize(len)/len
 192 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
 193 */
 194static u16 tsk_adv_blocks(int len)
 195{
 196	return len / FLOWCTL_BLK_SZ / 4;
 197}
 198
 199/* tsk_inc(): increment counter for sent or received data
 200 * - If block based flow control is not supported by peer we
 201 *   fall back to message based ditto, incrementing the counter
 202 */
 203static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
 204{
 205	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
 206		return ((msglen / FLOWCTL_BLK_SZ) + 1);
 207	return 1;
 208}
 209
 210/**
 211 * tsk_advance_rx_queue - discard first buffer in socket receive queue
 212 *
 213 * Caller must hold socket lock
 214 */
 215static void tsk_advance_rx_queue(struct sock *sk)
 216{
 217	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
 218}
 219
 220/* tipc_sk_respond() : send response message back to sender
 221 */
 222static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
 223{
 224	u32 selector;
 225	u32 dnode;
 226	u32 onode = tipc_own_addr(sock_net(sk));
 227
 228	if (!tipc_msg_reverse(onode, &skb, err))
 229		return;
 230
 231	dnode = msg_destnode(buf_msg(skb));
 232	selector = msg_origport(buf_msg(skb));
 233	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
 234}
 235
 236/**
 237 * tsk_rej_rx_queue - reject all buffers in socket receive queue
 238 *
 239 * Caller must hold socket lock
 240 */
 241static void tsk_rej_rx_queue(struct sock *sk)
 242{
 243	struct sk_buff *skb;
 244
 245	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
 246		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
 247}
 248
 249static bool tipc_sk_connected(struct sock *sk)
 250{
 251	return sk->sk_state == TIPC_ESTABLISHED;
 252}
 253
 254/* tipc_sk_type_connectionless - check if the socket is datagram socket
 255 * @sk: socket
 256 *
 257 * Returns true if connection less, false otherwise
 258 */
 259static bool tipc_sk_type_connectionless(struct sock *sk)
 260{
 261	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
 262}
 263
 264/* tsk_peer_msg - verify if message was sent by connected port's peer
 265 *
 266 * Handles cases where the node's network address has changed from
 267 * the default of <0.0.0> to its configured setting.
 268 */
 269static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
 270{
 271	struct sock *sk = &tsk->sk;
 272	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
 273	u32 peer_port = tsk_peer_port(tsk);
 274	u32 orig_node;
 275	u32 peer_node;
 276
 277	if (unlikely(!tipc_sk_connected(sk)))
 278		return false;
 279
 280	if (unlikely(msg_origport(msg) != peer_port))
 281		return false;
 282
 283	orig_node = msg_orignode(msg);
 284	peer_node = tsk_peer_node(tsk);
 285
 286	if (likely(orig_node == peer_node))
 287		return true;
 288
 289	if (!orig_node && (peer_node == tn->own_addr))
 290		return true;
 291
 292	if (!peer_node && (orig_node == tn->own_addr))
 293		return true;
 294
 295	return false;
 296}
 297
 298/* tipc_set_sk_state - set the sk_state of the socket
 299 * @sk: socket
 300 *
 301 * Caller must hold socket lock
 302 *
 303 * Returns 0 on success, errno otherwise
 304 */
 305static int tipc_set_sk_state(struct sock *sk, int state)
 306{
 307	int oldsk_state = sk->sk_state;
 308	int res = -EINVAL;
 309
 310	switch (state) {
 311	case TIPC_OPEN:
 312		res = 0;
 313		break;
 314	case TIPC_LISTEN:
 315	case TIPC_CONNECTING:
 316		if (oldsk_state == TIPC_OPEN)
 317			res = 0;
 318		break;
 319	case TIPC_ESTABLISHED:
 320		if (oldsk_state == TIPC_CONNECTING ||
 321		    oldsk_state == TIPC_OPEN)
 322			res = 0;
 323		break;
 324	case TIPC_DISCONNECTING:
 325		if (oldsk_state == TIPC_CONNECTING ||
 326		    oldsk_state == TIPC_ESTABLISHED)
 327			res = 0;
 328		break;
 329	}
 330
 331	if (!res)
 332		sk->sk_state = state;
 333
 334	return res;
 335}
 336
 337/**
 338 * tipc_sk_create - create a TIPC socket
 339 * @net: network namespace (must be default network)
 340 * @sock: pre-allocated socket structure
 341 * @protocol: protocol indicator (must be 0)
 342 * @kern: caused by kernel or by userspace?
 343 *
 344 * This routine creates additional data structures used by the TIPC socket,
 345 * initializes them, and links them together.
 346 *
 347 * Returns 0 on success, errno otherwise
 348 */
 349static int tipc_sk_create(struct net *net, struct socket *sock,
 350			  int protocol, int kern)
 351{
 352	struct tipc_net *tn;
 353	const struct proto_ops *ops;
 
 354	struct sock *sk;
 355	struct tipc_sock *tsk;
 356	struct tipc_msg *msg;
 
 357
 358	/* Validate arguments */
 359	if (unlikely(protocol != 0))
 360		return -EPROTONOSUPPORT;
 361
 362	switch (sock->type) {
 363	case SOCK_STREAM:
 364		ops = &stream_ops;
 
 365		break;
 366	case SOCK_SEQPACKET:
 367		ops = &packet_ops;
 
 368		break;
 369	case SOCK_DGRAM:
 370	case SOCK_RDM:
 371		ops = &msg_ops;
 
 372		break;
 373	default:
 374		return -EPROTOTYPE;
 375	}
 376
 377	/* Allocate socket's protocol area */
 378	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
 
 
 
 
 379	if (sk == NULL)
 380		return -ENOMEM;
 381
 382	tsk = tipc_sk(sk);
 383	tsk->max_pkt = MAX_PKT_DEFAULT;
 384	INIT_LIST_HEAD(&tsk->publications);
 385	msg = &tsk->phdr;
 386	tn = net_generic(sock_net(sk), tipc_net_id);
 387	tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
 388		      NAMED_H_SIZE, 0);
 
 
 389
 390	/* Finish initializing socket data structures */
 391	sock->ops = ops;
 
 
 392	sock_init_data(sock, sk);
 393	tipc_set_sk_state(sk, TIPC_OPEN);
 394	if (tipc_sk_insert(tsk)) {
 395		pr_warn("Socket create failed; port number exhausted\n");
 396		return -EINVAL;
 397	}
 398	msg_set_origport(msg, tsk->portid);
 399	setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
 400	sk->sk_shutdown = 0;
 401	sk->sk_backlog_rcv = tipc_backlog_rcv;
 402	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
 403	sk->sk_data_ready = tipc_data_ready;
 404	sk->sk_write_space = tipc_write_space;
 405	sk->sk_destruct = tipc_sock_destruct;
 406	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
 407	atomic_set(&tsk->dupl_rcvcnt, 0);
 408
 409	/* Start out with safe limits until we receive an advertised window */
 410	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
 411	tsk->rcv_win = tsk->snd_win;
 412
 413	if (tipc_sk_type_connectionless(sk)) {
 414		tsk_set_unreturnable(tsk, true);
 415		if (sock->type == SOCK_DGRAM)
 416			tsk_set_unreliable(tsk, true);
 417	}
 418
 419	return 0;
 420}
 421
 422static void tipc_sk_callback(struct rcu_head *head)
 
 
 
 
 
 
 
 
 
 
 
 423{
 424	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
 425
 426	sock_put(&tsk->sk);
 
 
 
 
 
 
 
 427}
 428
 429/* Caller should hold socket lock for the socket. */
 430static void __tipc_shutdown(struct socket *sock, int error)
 
 
 
 
 
 
 431{
 432	struct sock *sk = sock->sk;
 433	struct tipc_sock *tsk = tipc_sk(sk);
 434	struct net *net = sock_net(sk);
 435	u32 dnode = tsk_peer_node(tsk);
 436	struct sk_buff *skb;
 437
 438	/* Reject all unreceived messages, except on an active connection
 439	 * (which disconnects locally & sends a 'FIN+' to peer).
 440	 */
 441	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
 442		if (TIPC_SKB_CB(skb)->bytes_read) {
 443			kfree_skb(skb);
 444			continue;
 445		}
 446		if (!tipc_sk_type_connectionless(sk) &&
 447		    sk->sk_state != TIPC_DISCONNECTING) {
 448			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 449			tipc_node_remove_conn(net, dnode, tsk->portid);
 450		}
 451		tipc_sk_respond(sk, skb, error);
 452	}
 453
 454	if (tipc_sk_type_connectionless(sk))
 455		return;
 
 
 
 456
 457	if (sk->sk_state != TIPC_DISCONNECTING) {
 458		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
 459				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
 460				      tsk_own_node(tsk), tsk_peer_port(tsk),
 461				      tsk->portid, error);
 462		if (skb)
 463			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
 464		tipc_node_remove_conn(net, dnode, tsk->portid);
 465		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 466	}
 
 
 467}
 468
 469/**
 470 * tipc_release - destroy a TIPC socket
 471 * @sock: socket to destroy
 472 *
 473 * This routine cleans up any messages that are still queued on the socket.
 474 * For DGRAM and RDM socket types, all queued messages are rejected.
 475 * For SEQPACKET and STREAM socket types, the first message is rejected
 476 * and any others are discarded.  (If the first message on a STREAM socket
 477 * is partially-read, it is discarded and the next one is rejected instead.)
 478 *
 479 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 480 * are returned or discarded according to the "destination droppable" setting
 481 * specified for the message by the sender.
 482 *
 483 * Returns 0 on success, errno otherwise
 484 */
 485static int tipc_release(struct socket *sock)
 486{
 487	struct sock *sk = sock->sk;
 488	struct tipc_sock *tsk;
 
 
 489
 490	/*
 491	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 492	 * releases a pre-allocated child socket that was never used)
 493	 */
 494	if (sk == NULL)
 495		return 0;
 496
 497	tsk = tipc_sk(sk);
 
 498	lock_sock(sk);
 499
 500	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
 501	sk->sk_shutdown = SHUTDOWN_MASK;
 502	tipc_sk_withdraw(tsk, 0, NULL);
 503	sk_stop_timer(sk, &sk->sk_timer);
 504	tipc_sk_remove(tsk);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 505
 506	/* Reject any messages that accumulated in backlog queue */
 
 507	release_sock(sk);
 508
 509	call_rcu(&tsk->rcu, tipc_sk_callback);
 510	sock->sk = NULL;
 511
 512	return 0;
 513}
 514
 515/**
 516 * tipc_bind - associate or disassocate TIPC name(s) with a socket
 517 * @sock: socket structure
 518 * @uaddr: socket address describing name(s) and desired operation
 519 * @uaddr_len: size of socket address data structure
 520 *
 521 * Name and name sequence binding is indicated using a positive scope value;
 522 * a negative scope value unbinds the specified name.  Specifying no name
 523 * (i.e. a socket address length of 0) unbinds all names from the socket.
 524 *
 525 * Returns 0 on success, errno otherwise
 526 *
 527 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 528 *       access any non-constant socket information.
 529 */
 530static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
 531		     int uaddr_len)
 532{
 533	struct sock *sk = sock->sk;
 534	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 535	struct tipc_sock *tsk = tipc_sk(sk);
 536	int res = -EINVAL;
 537
 538	lock_sock(sk);
 539	if (unlikely(!uaddr_len)) {
 540		res = tipc_sk_withdraw(tsk, 0, NULL);
 541		goto exit;
 542	}
 543
 544	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
 545		res = -EINVAL;
 546		goto exit;
 547	}
 548	if (addr->family != AF_TIPC) {
 549		res = -EAFNOSUPPORT;
 550		goto exit;
 551	}
 552
 553	if (addr->addrtype == TIPC_ADDR_NAME)
 554		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
 555	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
 556		res = -EAFNOSUPPORT;
 557		goto exit;
 558	}
 559
 560	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
 561	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
 562	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
 563		res = -EACCES;
 564		goto exit;
 565	}
 566
 567	res = (addr->scope > 0) ?
 568		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
 569		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
 570exit:
 571	release_sock(sk);
 572	return res;
 573}
 574
 575/**
 576 * tipc_getname - get port ID of socket or peer socket
 577 * @sock: socket structure
 578 * @uaddr: area for returned socket address
 579 * @uaddr_len: area for returned length of socket address
 580 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 581 *
 582 * Returns 0 on success, errno otherwise
 583 *
 584 * NOTE: This routine doesn't need to take the socket lock since it only
 585 *       accesses socket information that is unchanging (or which changes in
 586 *       a completely predictable manner).
 587 */
 588static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
 589			int *uaddr_len, int peer)
 590{
 591	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 592	struct sock *sk = sock->sk;
 593	struct tipc_sock *tsk = tipc_sk(sk);
 594	struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
 595
 596	memset(addr, 0, sizeof(*addr));
 597	if (peer) {
 598		if ((!tipc_sk_connected(sk)) &&
 599		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
 600			return -ENOTCONN;
 601		addr->addr.id.ref = tsk_peer_port(tsk);
 602		addr->addr.id.node = tsk_peer_node(tsk);
 603	} else {
 604		addr->addr.id.ref = tsk->portid;
 605		addr->addr.id.node = tn->own_addr;
 606	}
 607
 608	*uaddr_len = sizeof(*addr);
 609	addr->addrtype = TIPC_ADDR_ID;
 610	addr->family = AF_TIPC;
 611	addr->scope = 0;
 612	addr->addr.name.domain = 0;
 613
 614	return 0;
 615}
 616
 617/**
 618 * tipc_poll - read and possibly block on pollmask
 619 * @file: file structure associated with the socket
 620 * @sock: socket for which to calculate the poll bits
 621 * @wait: ???
 622 *
 623 * Returns pollmask value
 624 *
 625 * COMMENTARY:
 626 * It appears that the usual socket locking mechanisms are not useful here
 627 * since the pollmask info is potentially out-of-date the moment this routine
 628 * exits.  TCP and other protocols seem to rely on higher level poll routines
 629 * to handle any preventable race conditions, so TIPC will do the same ...
 630 *
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 631 * IMPORTANT: The fact that a read or write operation is indicated does NOT
 632 * imply that the operation will succeed, merely that it should be performed
 633 * and will not block.
 634 */
 635static unsigned int tipc_poll(struct file *file, struct socket *sock,
 636			      poll_table *wait)
 637{
 638	struct sock *sk = sock->sk;
 639	struct tipc_sock *tsk = tipc_sk(sk);
 640	u32 mask = 0;
 641
 642	sock_poll_wait(file, sk_sleep(sk), wait);
 643
 644	if (sk->sk_shutdown & RCV_SHUTDOWN)
 645		mask |= POLLRDHUP | POLLIN | POLLRDNORM;
 646	if (sk->sk_shutdown == SHUTDOWN_MASK)
 647		mask |= POLLHUP;
 648
 649	switch (sk->sk_state) {
 650	case TIPC_ESTABLISHED:
 651		if (!tsk->link_cong && !tsk_conn_cong(tsk))
 652			mask |= POLLOUT;
 653		/* fall thru' */
 654	case TIPC_LISTEN:
 655	case TIPC_CONNECTING:
 656		if (!skb_queue_empty(&sk->sk_receive_queue))
 657			mask |= (POLLIN | POLLRDNORM);
 658		break;
 659	case TIPC_OPEN:
 660		if (!tsk->link_cong)
 661			mask |= POLLOUT;
 662		if (tipc_sk_type_connectionless(sk) &&
 663		    (!skb_queue_empty(&sk->sk_receive_queue)))
 664			mask |= (POLLIN | POLLRDNORM);
 665		break;
 666	case TIPC_DISCONNECTING:
 667		mask = (POLLIN | POLLRDNORM | POLLHUP);
 668		break;
 669	}
 670
 671	return mask;
 672}
 673
 674/**
 675 * tipc_sendmcast - send multicast message
 676 * @sock: socket structure
 677 * @seq: destination address
 678 * @msg: message to send
 679 * @dsz: total length of message data
 680 * @timeo: timeout to wait for wakeup
 681 *
 682 * Called from function tipc_sendmsg(), which has done all sanity checks
 683 * Returns the number of bytes sent on success, or errno
 684 */
 685static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
 686			  struct msghdr *msg, size_t dsz, long timeo)
 687{
 688	struct sock *sk = sock->sk;
 689	struct tipc_sock *tsk = tipc_sk(sk);
 690	struct net *net = sock_net(sk);
 691	struct tipc_msg *mhdr = &tsk->phdr;
 692	struct sk_buff_head pktchain;
 693	struct iov_iter save = msg->msg_iter;
 694	uint mtu;
 695	int rc;
 696
 697	if (!timeo && tsk->link_cong)
 698		return -ELINKCONG;
 699
 700	msg_set_type(mhdr, TIPC_MCAST_MSG);
 701	msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
 702	msg_set_destport(mhdr, 0);
 703	msg_set_destnode(mhdr, 0);
 704	msg_set_nametype(mhdr, seq->type);
 705	msg_set_namelower(mhdr, seq->lower);
 706	msg_set_nameupper(mhdr, seq->upper);
 707	msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
 708
 709	skb_queue_head_init(&pktchain);
 710
 711new_mtu:
 712	mtu = tipc_bcast_get_mtu(net);
 713	rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
 714	if (unlikely(rc < 0))
 715		return rc;
 716
 717	do {
 718		rc = tipc_bcast_xmit(net, &pktchain);
 719		if (likely(!rc))
 720			return dsz;
 721
 722		if (rc == -ELINKCONG) {
 723			tsk->link_cong = 1;
 724			rc = tipc_wait_for_sndmsg(sock, &timeo);
 725			if (!rc)
 726				continue;
 727		}
 728		__skb_queue_purge(&pktchain);
 729		if (rc == -EMSGSIZE) {
 730			msg->msg_iter = save;
 731			goto new_mtu;
 732		}
 733		break;
 734	} while (1);
 735	return rc;
 736}
 737
 738/**
 739 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
 740 * @arrvq: queue with arriving messages, to be cloned after destination lookup
 741 * @inputq: queue with cloned messages, delivered to socket after dest lookup
 742 *
 743 * Multi-threaded: parallel calls with reference to same queues may occur
 744 */
 745void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
 746		       struct sk_buff_head *inputq)
 747{
 748	struct tipc_msg *msg;
 749	struct tipc_plist dports;
 750	u32 portid;
 751	u32 scope = TIPC_CLUSTER_SCOPE;
 752	struct sk_buff_head tmpq;
 753	uint hsz;
 754	struct sk_buff *skb, *_skb;
 755
 756	__skb_queue_head_init(&tmpq);
 757	tipc_plist_init(&dports);
 758
 759	skb = tipc_skb_peek(arrvq, &inputq->lock);
 760	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
 761		msg = buf_msg(skb);
 762		hsz = skb_headroom(skb) + msg_hdr_sz(msg);
 763
 764		if (in_own_node(net, msg_orignode(msg)))
 765			scope = TIPC_NODE_SCOPE;
 766
 767		/* Create destination port list and message clones: */
 768		tipc_nametbl_mc_translate(net,
 769					  msg_nametype(msg), msg_namelower(msg),
 770					  msg_nameupper(msg), scope, &dports);
 771		portid = tipc_plist_pop(&dports);
 772		for (; portid; portid = tipc_plist_pop(&dports)) {
 773			_skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
 774			if (_skb) {
 775				msg_set_destport(buf_msg(_skb), portid);
 776				__skb_queue_tail(&tmpq, _skb);
 777				continue;
 778			}
 779			pr_warn("Failed to clone mcast rcv buffer\n");
 780		}
 781		/* Append to inputq if not already done by other thread */
 782		spin_lock_bh(&inputq->lock);
 783		if (skb_peek(arrvq) == skb) {
 784			skb_queue_splice_tail_init(&tmpq, inputq);
 785			kfree_skb(__skb_dequeue(arrvq));
 786		}
 787		spin_unlock_bh(&inputq->lock);
 788		__skb_queue_purge(&tmpq);
 789		kfree_skb(skb);
 790	}
 791	tipc_sk_rcv(net, inputq);
 792}
 793
 794/**
 795 * tipc_sk_proto_rcv - receive a connection mng protocol message
 796 * @tsk: receiving socket
 797 * @skb: pointer to message buffer.
 798 */
 799static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
 800			      struct sk_buff_head *xmitq)
 801{
 802	struct sock *sk = &tsk->sk;
 803	u32 onode = tsk_own_node(tsk);
 804	struct tipc_msg *hdr = buf_msg(skb);
 805	int mtyp = msg_type(hdr);
 806	bool conn_cong;
 807
 808	/* Ignore if connection cannot be validated: */
 809	if (!tsk_peer_msg(tsk, hdr))
 810		goto exit;
 
 
 
 811
 812	tsk->probe_unacked = false;
 813
 814	if (mtyp == CONN_PROBE) {
 815		msg_set_type(hdr, CONN_PROBE_REPLY);
 816		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
 817			__skb_queue_tail(xmitq, skb);
 818		return;
 819	} else if (mtyp == CONN_ACK) {
 820		conn_cong = tsk_conn_cong(tsk);
 821		tsk->snt_unacked -= msg_conn_ack(hdr);
 822		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
 823			tsk->snd_win = msg_adv_win(hdr);
 824		if (conn_cong)
 825			sk->sk_write_space(sk);
 826	} else if (mtyp != CONN_PROBE_REPLY) {
 827		pr_warn("Received unknown CONN_PROTO msg\n");
 828	}
 829exit:
 830	kfree_skb(skb);
 831}
 832
 833static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
 834{
 835	DEFINE_WAIT_FUNC(wait, woken_wake_function);
 836	struct sock *sk = sock->sk;
 837	struct tipc_sock *tsk = tipc_sk(sk);
 
 838	int done;
 839
 840	do {
 841		int err = sock_error(sk);
 842		if (err)
 843			return err;
 844		if (sk->sk_shutdown & SEND_SHUTDOWN)
 845			return -EPIPE;
 846		if (!*timeo_p)
 847			return -EAGAIN;
 848		if (signal_pending(current))
 849			return sock_intr_errno(*timeo_p);
 850
 851		add_wait_queue(sk_sleep(sk), &wait);
 852		done = sk_wait_event(sk, timeo_p, !tsk->link_cong, &wait);
 853		remove_wait_queue(sk_sleep(sk), &wait);
 854	} while (!done);
 855	return 0;
 856}
 857
 
 858/**
 859 * tipc_sendmsg - send message in connectionless manner
 
 860 * @sock: socket structure
 861 * @m: message to send
 862 * @dsz: amount of user data to be sent
 863 *
 864 * Message must have an destination specified explicitly.
 865 * Used for SOCK_RDM and SOCK_DGRAM messages,
 866 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
 867 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
 868 *
 869 * Returns the number of bytes sent on success, or errno otherwise
 870 */
 871static int tipc_sendmsg(struct socket *sock,
 872			struct msghdr *m, size_t dsz)
 873{
 874	struct sock *sk = sock->sk;
 875	int ret;
 876
 877	lock_sock(sk);
 878	ret = __tipc_sendmsg(sock, m, dsz);
 879	release_sock(sk);
 880
 881	return ret;
 882}
 883
 884static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
 885{
 886	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 887	struct sock *sk = sock->sk;
 888	struct tipc_sock *tsk = tipc_sk(sk);
 889	struct net *net = sock_net(sk);
 890	struct tipc_msg *mhdr = &tsk->phdr;
 891	u32 dnode, dport;
 892	struct sk_buff_head pktchain;
 893	bool is_connectionless = tipc_sk_type_connectionless(sk);
 894	struct sk_buff *skb;
 895	struct tipc_name_seq *seq;
 896	struct iov_iter save;
 897	u32 mtu;
 898	long timeo;
 899	int rc;
 900
 901	if (dsz > TIPC_MAX_USER_MSG_SIZE)
 
 
 
 
 
 902		return -EMSGSIZE;
 903	if (unlikely(!dest)) {
 904		if (is_connectionless && tsk->peer.family == AF_TIPC)
 905			dest = &tsk->peer;
 906		else
 907			return -EDESTADDRREQ;
 908	} else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
 909		   dest->family != AF_TIPC) {
 910		return -EINVAL;
 911	}
 912	if (!is_connectionless) {
 913		if (sk->sk_state == TIPC_LISTEN)
 914			return -EPIPE;
 915		if (sk->sk_state != TIPC_OPEN)
 916			return -EISCONN;
 917		if (tsk->published)
 918			return -EOPNOTSUPP;
 
 
 919		if (dest->addrtype == TIPC_ADDR_NAME) {
 920			tsk->conn_type = dest->addr.name.name.type;
 921			tsk->conn_instance = dest->addr.name.name.instance;
 922		}
 
 
 
 923	}
 924	seq = &dest->addr.nameseq;
 925	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
 926
 927	if (dest->addrtype == TIPC_ADDR_MCAST) {
 928		return tipc_sendmcast(sock, seq, m, dsz, timeo);
 929	} else if (dest->addrtype == TIPC_ADDR_NAME) {
 930		u32 type = dest->addr.name.name.type;
 931		u32 inst = dest->addr.name.name.instance;
 932		u32 domain = dest->addr.name.domain;
 933
 934		dnode = domain;
 935		msg_set_type(mhdr, TIPC_NAMED_MSG);
 936		msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
 937		msg_set_nametype(mhdr, type);
 938		msg_set_nameinst(mhdr, inst);
 939		msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
 940		dport = tipc_nametbl_translate(net, type, inst, &dnode);
 941		msg_set_destnode(mhdr, dnode);
 942		msg_set_destport(mhdr, dport);
 943		if (unlikely(!dport && !dnode))
 944			return -EHOSTUNREACH;
 945	} else if (dest->addrtype == TIPC_ADDR_ID) {
 946		dnode = dest->addr.id.node;
 947		msg_set_type(mhdr, TIPC_DIRECT_MSG);
 948		msg_set_lookup_scope(mhdr, 0);
 949		msg_set_destnode(mhdr, dnode);
 950		msg_set_destport(mhdr, dest->addr.id.ref);
 951		msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
 952	}
 953
 954	skb_queue_head_init(&pktchain);
 955	save = m->msg_iter;
 956new_mtu:
 957	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
 958	rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
 959	if (rc < 0)
 960		return rc;
 961
 962	do {
 963		skb = skb_peek(&pktchain);
 964		TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
 965		rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
 966		if (likely(!rc)) {
 967			if (!is_connectionless)
 968				tipc_set_sk_state(sk, TIPC_CONNECTING);
 969			return dsz;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 970		}
 971		if (rc == -ELINKCONG) {
 972			tsk->link_cong = 1;
 973			rc = tipc_wait_for_sndmsg(sock, &timeo);
 974			if (!rc)
 975				continue;
 976		}
 977		__skb_queue_purge(&pktchain);
 978		if (rc == -EMSGSIZE) {
 979			m->msg_iter = save;
 980			goto new_mtu;
 981		}
 982		break;
 983	} while (1);
 984
 985	return rc;
 
 
 
 986}
 987
 988static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
 989{
 990	DEFINE_WAIT_FUNC(wait, woken_wake_function);
 991	struct sock *sk = sock->sk;
 992	struct tipc_sock *tsk = tipc_sk(sk);
 
 
 993	int done;
 994
 995	do {
 996		int err = sock_error(sk);
 997		if (err)
 998			return err;
 999		if (sk->sk_state == TIPC_DISCONNECTING)
1000			return -EPIPE;
1001		else if (!tipc_sk_connected(sk))
1002			return -ENOTCONN;
1003		if (!*timeo_p)
1004			return -EAGAIN;
1005		if (signal_pending(current))
1006			return sock_intr_errno(*timeo_p);
1007
1008		add_wait_queue(sk_sleep(sk), &wait);
1009		done = sk_wait_event(sk, timeo_p,
1010				     (!tsk->link_cong &&
1011				      !tsk_conn_cong(tsk)) ||
1012				      !tipc_sk_connected(sk), &wait);
1013		remove_wait_queue(sk_sleep(sk), &wait);
1014	} while (!done);
1015	return 0;
1016}
1017
1018/**
1019 * tipc_send_stream - send stream-oriented data
 
1020 * @sock: socket structure
1021 * @m: data to send
1022 * @dsz: total length of data to be transmitted
1023 *
1024 * Used for SOCK_STREAM data.
1025 *
1026 * Returns the number of bytes sent on success (or partial success),
1027 * or errno if no data sent
1028 */
1029static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
 
1030{
1031	struct sock *sk = sock->sk;
1032	int ret;
 
 
 
 
 
 
 
 
 
 
 
 
 
1033
1034	lock_sock(sk);
1035	ret = __tipc_send_stream(sock, m, dsz);
1036	release_sock(sk);
 
 
 
 
1037
1038	return ret;
 
 
 
 
 
 
 
 
 
 
 
 
1039}
1040
1041static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
 
 
 
 
 
 
 
 
 
 
 
 
 
1042{
1043	struct sock *sk = sock->sk;
1044	struct net *net = sock_net(sk);
1045	struct tipc_sock *tsk = tipc_sk(sk);
1046	struct tipc_msg *mhdr = &tsk->phdr;
1047	struct sk_buff_head pktchain;
1048	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1049	u32 portid = tsk->portid;
1050	int rc = -EINVAL;
1051	long timeo;
1052	u32 dnode;
1053	uint mtu, send, sent = 0;
1054	struct iov_iter save;
1055	int hlen = MIN_H_SIZE;
1056
1057	/* Handle implied connection establishment */
1058	if (unlikely(dest)) {
1059		rc = __tipc_sendmsg(sock, m, dsz);
1060		hlen = msg_hdr_sz(mhdr);
1061		if (dsz && (dsz == rc))
1062			tsk->snt_unacked = tsk_inc(tsk, dsz + hlen);
1063		return rc;
1064	}
1065	if (dsz > (uint)INT_MAX)
1066		return -EMSGSIZE;
1067
1068	if (unlikely(!tipc_sk_connected(sk))) {
1069		if (sk->sk_state == TIPC_DISCONNECTING)
1070			return -EPIPE;
 
1071		else
1072			return -ENOTCONN;
 
1073	}
1074
1075	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1076	if (!timeo && tsk->link_cong)
1077		return -ELINKCONG;
 
1078
1079	dnode = tsk_peer_node(tsk);
1080	skb_queue_head_init(&pktchain);
 
 
1081
1082next:
1083	save = m->msg_iter;
1084	mtu = tsk->max_pkt;
1085	send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1086	rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1087	if (unlikely(rc < 0))
1088		return rc;
1089
1090	do {
1091		if (likely(!tsk_conn_cong(tsk))) {
1092			rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1093			if (likely(!rc)) {
1094				tsk->snt_unacked += tsk_inc(tsk, send + hlen);
1095				sent += send;
1096				if (sent == dsz)
1097					return dsz;
1098				goto next;
1099			}
1100			if (rc == -EMSGSIZE) {
1101				__skb_queue_purge(&pktchain);
1102				tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1103								 portid);
1104				m->msg_iter = save;
1105				goto next;
 
 
 
 
 
 
 
 
 
 
 
1106			}
1107			if (rc != -ELINKCONG)
1108				break;
1109
1110			tsk->link_cong = 1;
1111		}
1112		rc = tipc_wait_for_sndpkt(sock, &timeo);
1113	} while (!rc);
1114
1115	__skb_queue_purge(&pktchain);
1116	return sent ? sent : rc;
 
 
 
 
1117}
1118
1119/**
1120 * tipc_send_packet - send a connection-oriented message
1121 * @sock: socket structure
1122 * @m: message to send
1123 * @dsz: length of data to be transmitted
1124 *
1125 * Used for SOCK_SEQPACKET messages.
1126 *
1127 * Returns the number of bytes sent on success, or errno otherwise
1128 */
1129static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1130{
1131	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1132		return -EMSGSIZE;
 
1133
1134	return tipc_send_stream(sock, m, dsz);
1135}
1136
1137/* tipc_sk_finish_conn - complete the setup of a connection
1138 */
1139static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1140				u32 peer_node)
1141{
1142	struct sock *sk = &tsk->sk;
1143	struct net *net = sock_net(sk);
1144	struct tipc_msg *msg = &tsk->phdr;
1145
1146	msg_set_destnode(msg, peer_node);
1147	msg_set_destport(msg, peer_port);
1148	msg_set_type(msg, TIPC_CONN_MSG);
1149	msg_set_lookup_scope(msg, 0);
1150	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1151
1152	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
1153	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1154	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1155	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1156	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1157	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1158		return;
1159
1160	/* Fall back to message based flow control */
1161	tsk->rcv_win = FLOWCTL_MSG_WIN;
1162	tsk->snd_win = FLOWCTL_MSG_WIN;
1163}
1164
1165/**
1166 * set_orig_addr - capture sender's address for received message
1167 * @m: descriptor for message info
1168 * @msg: received message header
1169 *
1170 * Note: Address is not captured if not requested by receiver.
1171 */
1172static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1173{
1174	DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1175
1176	if (addr) {
1177		addr->family = AF_TIPC;
1178		addr->addrtype = TIPC_ADDR_ID;
1179		memset(&addr->addr, 0, sizeof(addr->addr));
1180		addr->addr.id.ref = msg_origport(msg);
1181		addr->addr.id.node = msg_orignode(msg);
1182		addr->addr.name.domain = 0;	/* could leave uninitialized */
1183		addr->scope = 0;		/* could leave uninitialized */
1184		m->msg_namelen = sizeof(struct sockaddr_tipc);
1185	}
1186}
1187
1188/**
1189 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1190 * @m: descriptor for message info
1191 * @msg: received message header
1192 * @tsk: TIPC port associated with message
1193 *
1194 * Note: Ancillary data is not captured if not requested by receiver.
1195 *
1196 * Returns 0 if successful, otherwise errno
1197 */
1198static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1199				 struct tipc_sock *tsk)
1200{
1201	u32 anc_data[3];
1202	u32 err;
1203	u32 dest_type;
1204	int has_name;
1205	int res;
1206
1207	if (likely(m->msg_controllen == 0))
1208		return 0;
1209
1210	/* Optionally capture errored message object(s) */
1211	err = msg ? msg_errcode(msg) : 0;
1212	if (unlikely(err)) {
1213		anc_data[0] = err;
1214		anc_data[1] = msg_data_sz(msg);
1215		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1216		if (res)
1217			return res;
1218		if (anc_data[1]) {
1219			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1220				       msg_data(msg));
1221			if (res)
1222				return res;
1223		}
1224	}
1225
1226	/* Optionally capture message destination object */
1227	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1228	switch (dest_type) {
1229	case TIPC_NAMED_MSG:
1230		has_name = 1;
1231		anc_data[0] = msg_nametype(msg);
1232		anc_data[1] = msg_namelower(msg);
1233		anc_data[2] = msg_namelower(msg);
1234		break;
1235	case TIPC_MCAST_MSG:
1236		has_name = 1;
1237		anc_data[0] = msg_nametype(msg);
1238		anc_data[1] = msg_namelower(msg);
1239		anc_data[2] = msg_nameupper(msg);
1240		break;
1241	case TIPC_CONN_MSG:
1242		has_name = (tsk->conn_type != 0);
1243		anc_data[0] = tsk->conn_type;
1244		anc_data[1] = tsk->conn_instance;
1245		anc_data[2] = tsk->conn_instance;
1246		break;
1247	default:
1248		has_name = 0;
1249	}
1250	if (has_name) {
1251		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1252		if (res)
1253			return res;
1254	}
1255
1256	return 0;
1257}
1258
1259static void tipc_sk_send_ack(struct tipc_sock *tsk)
1260{
1261	struct sock *sk = &tsk->sk;
1262	struct net *net = sock_net(sk);
1263	struct sk_buff *skb = NULL;
1264	struct tipc_msg *msg;
1265	u32 peer_port = tsk_peer_port(tsk);
1266	u32 dnode = tsk_peer_node(tsk);
1267
1268	if (!tipc_sk_connected(sk))
1269		return;
1270	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1271			      dnode, tsk_own_node(tsk), peer_port,
1272			      tsk->portid, TIPC_OK);
1273	if (!skb)
1274		return;
1275	msg = buf_msg(skb);
1276	msg_set_conn_ack(msg, tsk->rcv_unacked);
1277	tsk->rcv_unacked = 0;
1278
1279	/* Adjust to and advertize the correct window limit */
1280	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1281		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1282		msg_set_adv_win(msg, tsk->rcv_win);
1283	}
1284	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1285}
1286
1287static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1288{
1289	struct sock *sk = sock->sk;
1290	DEFINE_WAIT(wait);
1291	long timeo = *timeop;
1292	int err;
1293
1294	for (;;) {
1295		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1296		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1297			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1298				err = -ENOTCONN;
1299				break;
1300			}
1301			release_sock(sk);
1302			timeo = schedule_timeout(timeo);
1303			lock_sock(sk);
1304		}
1305		err = 0;
1306		if (!skb_queue_empty(&sk->sk_receive_queue))
1307			break;
 
 
 
1308		err = -EAGAIN;
1309		if (!timeo)
1310			break;
1311		err = sock_intr_errno(timeo);
1312		if (signal_pending(current))
1313			break;
1314	}
1315	finish_wait(sk_sleep(sk), &wait);
1316	*timeop = timeo;
1317	return err;
1318}
1319
1320/**
1321 * tipc_recvmsg - receive packet-oriented message
 
1322 * @m: descriptor for message info
1323 * @buf_len: total size of user buffer area
1324 * @flags: receive flags
1325 *
1326 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1327 * If the complete message doesn't fit in user area, truncate it.
1328 *
1329 * Returns size of returned message data, errno otherwise
1330 */
1331static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1332			int flags)
1333{
1334	struct sock *sk = sock->sk;
1335	struct tipc_sock *tsk = tipc_sk(sk);
 
1336	struct sk_buff *buf;
1337	struct tipc_msg *msg;
1338	bool is_connectionless = tipc_sk_type_connectionless(sk);
1339	long timeo;
1340	unsigned int sz;
1341	u32 err;
1342	int res, hlen;
1343
1344	/* Catch invalid receive requests */
1345	if (unlikely(!buf_len))
1346		return -EINVAL;
1347
1348	lock_sock(sk);
1349
1350	if (!is_connectionless && unlikely(sk->sk_state == TIPC_OPEN)) {
1351		res = -ENOTCONN;
1352		goto exit;
1353	}
1354
1355	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1356restart:
1357
1358	/* Look for a message in receive queue; wait if necessary */
1359	res = tipc_wait_for_rcvmsg(sock, &timeo);
1360	if (res)
1361		goto exit;
1362
1363	/* Look at first message in receive queue */
1364	buf = skb_peek(&sk->sk_receive_queue);
1365	msg = buf_msg(buf);
1366	sz = msg_data_sz(msg);
1367	hlen = msg_hdr_sz(msg);
1368	err = msg_errcode(msg);
1369
1370	/* Discard an empty non-errored message & try again */
1371	if ((!sz) && (!err)) {
1372		tsk_advance_rx_queue(sk);
1373		goto restart;
1374	}
1375
1376	/* Capture sender's address (optional) */
1377	set_orig_addr(m, msg);
1378
1379	/* Capture ancillary data (optional) */
1380	res = tipc_sk_anc_data_recv(m, msg, tsk);
1381	if (res)
1382		goto exit;
1383
1384	/* Capture message data (if valid) & compute return value (always) */
1385	if (!err) {
1386		if (unlikely(buf_len < sz)) {
1387			sz = buf_len;
1388			m->msg_flags |= MSG_TRUNC;
1389		}
1390		res = skb_copy_datagram_msg(buf, hlen, m, sz);
 
1391		if (res)
1392			goto exit;
1393		res = sz;
1394	} else {
1395		if (is_connectionless || err == TIPC_CONN_SHUTDOWN ||
1396		    m->msg_control)
1397			res = 0;
1398		else
1399			res = -ECONNRESET;
1400	}
1401
1402	if (unlikely(flags & MSG_PEEK))
1403		goto exit;
1404
1405	if (likely(!is_connectionless)) {
1406		tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1407		if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1408			tipc_sk_send_ack(tsk);
1409	}
1410	tsk_advance_rx_queue(sk);
1411exit:
1412	release_sock(sk);
1413	return res;
1414}
1415
1416/**
1417 * tipc_recv_stream - receive stream-oriented data
 
1418 * @m: descriptor for message info
1419 * @buf_len: total size of user buffer area
1420 * @flags: receive flags
1421 *
1422 * Used for SOCK_STREAM messages only.  If not enough data is available
1423 * will optionally wait for more; never truncates data.
1424 *
1425 * Returns size of returned message data, errno otherwise
1426 */
1427static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1428			    size_t buf_len, int flags)
1429{
1430	struct sock *sk = sock->sk;
1431	struct tipc_sock *tsk = tipc_sk(sk);
 
1432	struct sk_buff *buf;
1433	struct tipc_msg *msg;
1434	long timeo;
1435	unsigned int sz;
1436	int target;
1437	int sz_copied = 0;
1438	u32 err;
1439	int res = 0, hlen;
1440
1441	/* Catch invalid receive attempts */
1442	if (unlikely(!buf_len))
1443		return -EINVAL;
1444
1445	lock_sock(sk);
1446
1447	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1448		res = -ENOTCONN;
1449		goto exit;
1450	}
1451
1452	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1453	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1454
1455restart:
1456	/* Look for a message in receive queue; wait if necessary */
1457	res = tipc_wait_for_rcvmsg(sock, &timeo);
1458	if (res)
1459		goto exit;
1460
1461	/* Look at first message in receive queue */
1462	buf = skb_peek(&sk->sk_receive_queue);
1463	msg = buf_msg(buf);
1464	sz = msg_data_sz(msg);
1465	hlen = msg_hdr_sz(msg);
1466	err = msg_errcode(msg);
1467
1468	/* Discard an empty non-errored message & try again */
1469	if ((!sz) && (!err)) {
1470		tsk_advance_rx_queue(sk);
1471		goto restart;
1472	}
1473
1474	/* Optionally capture sender's address & ancillary data of first msg */
1475	if (sz_copied == 0) {
1476		set_orig_addr(m, msg);
1477		res = tipc_sk_anc_data_recv(m, msg, tsk);
1478		if (res)
1479			goto exit;
1480	}
1481
1482	/* Capture message data (if valid) & compute return value (always) */
1483	if (!err) {
1484		u32 offset = TIPC_SKB_CB(buf)->bytes_read;
1485		u32 needed;
1486		int sz_to_copy;
1487
1488		sz -= offset;
1489		needed = (buf_len - sz_copied);
1490		sz_to_copy = min(sz, needed);
1491
1492		res = skb_copy_datagram_msg(buf, hlen + offset, m, sz_to_copy);
 
1493		if (res)
1494			goto exit;
1495
1496		sz_copied += sz_to_copy;
1497
1498		if (sz_to_copy < sz) {
1499			if (!(flags & MSG_PEEK))
1500				TIPC_SKB_CB(buf)->bytes_read =
1501					offset + sz_to_copy;
1502			goto exit;
1503		}
1504	} else {
1505		if (sz_copied != 0)
1506			goto exit; /* can't add error msg to valid data */
1507
1508		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1509			res = 0;
1510		else
1511			res = -ECONNRESET;
1512	}
1513
1514	if (unlikely(flags & MSG_PEEK))
1515		goto exit;
1516
1517	tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1518	if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1519		tipc_sk_send_ack(tsk);
1520	tsk_advance_rx_queue(sk);
1521
1522	/* Loop around if more data is required */
1523	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1524	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1525	    (sz_copied < target)) &&	/* and more is ready or required */
 
1526	    (!err))			/* and haven't reached a FIN */
1527		goto restart;
1528
1529exit:
1530	release_sock(sk);
1531	return sz_copied ? sz_copied : res;
1532}
1533
1534/**
1535 * tipc_write_space - wake up thread if port congestion is released
1536 * @sk: socket
1537 */
1538static void tipc_write_space(struct sock *sk)
1539{
1540	struct socket_wq *wq;
1541
1542	rcu_read_lock();
1543	wq = rcu_dereference(sk->sk_wq);
1544	if (skwq_has_sleeper(wq))
1545		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1546						POLLWRNORM | POLLWRBAND);
1547	rcu_read_unlock();
1548}
1549
1550/**
1551 * tipc_data_ready - wake up threads to indicate messages have been received
1552 * @sk: socket
1553 * @len: the length of messages
1554 */
1555static void tipc_data_ready(struct sock *sk)
1556{
1557	struct socket_wq *wq;
1558
1559	rcu_read_lock();
1560	wq = rcu_dereference(sk->sk_wq);
1561	if (skwq_has_sleeper(wq))
1562		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1563						POLLRDNORM | POLLRDBAND);
1564	rcu_read_unlock();
1565}
1566
1567static void tipc_sock_destruct(struct sock *sk)
1568{
1569	__skb_queue_purge(&sk->sk_receive_queue);
1570}
1571
1572/**
1573 * filter_connect - Handle all incoming messages for a connection-based socket
1574 * @tsk: TIPC socket
1575 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1576 *
1577 * Returns true if everything ok, false otherwise
 
1578 */
1579static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1580{
1581	struct sock *sk = &tsk->sk;
1582	struct net *net = sock_net(sk);
1583	struct tipc_msg *hdr = buf_msg(skb);
 
 
 
 
1584
1585	if (unlikely(msg_mcast(hdr)))
1586		return false;
1587
1588	switch (sk->sk_state) {
1589	case TIPC_CONNECTING:
 
 
 
 
 
 
 
 
 
 
1590		/* Accept only ACK or NACK message */
1591		if (unlikely(!msg_connected(hdr)))
1592			return false;
1593
1594		if (unlikely(msg_errcode(hdr))) {
1595			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1596			sk->sk_err = ECONNREFUSED;
1597			return true;
 
1598		}
1599
1600		if (unlikely(!msg_isdata(hdr))) {
1601			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1602			sk->sk_err = EINVAL;
1603			return true;
 
 
 
 
 
1604		}
1605
1606		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1607		msg_set_importance(&tsk->phdr, msg_importance(hdr));
1608
1609		/* If 'ACK+' message, add to socket receive queue */
1610		if (msg_data_sz(hdr))
1611			return true;
1612
1613		/* If empty 'ACK-' message, wake up sleeping connect() */
1614		if (waitqueue_active(sk_sleep(sk)))
1615			wake_up_interruptible(sk_sleep(sk));
1616
1617		/* 'ACK-' message is neither accepted nor rejected: */
1618		msg_set_dest_droppable(hdr, 1);
1619		return false;
1620
1621	case TIPC_OPEN:
1622	case TIPC_DISCONNECTING:
1623		break;
1624	case TIPC_LISTEN:
 
1625		/* Accept only SYN message */
1626		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1627			return true;
 
 
1628		break;
1629	case TIPC_ESTABLISHED:
1630		/* Accept only connection-based messages sent by peer */
1631		if (unlikely(!tsk_peer_msg(tsk, hdr)))
1632			return false;
1633
1634		if (unlikely(msg_errcode(hdr))) {
1635			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1636			/* Let timer expire on it's own */
1637			tipc_node_remove_conn(net, tsk_peer_node(tsk),
1638					      tsk->portid);
1639			sk->sk_state_change(sk);
1640		}
1641		return true;
1642	default:
1643		pr_err("Unknown sk_state %u\n", sk->sk_state);
1644	}
1645
1646	return false;
1647}
1648
1649/**
1650 * rcvbuf_limit - get proper overload limit of socket receive queue
1651 * @sk: socket
1652 * @skb: message
1653 *
1654 * For connection oriented messages, irrespective of importance,
1655 * default queue limit is 2 MB.
1656 *
1657 * For connectionless messages, queue limits are based on message
1658 * importance as follows:
1659 *
1660 * TIPC_LOW_IMPORTANCE       (2 MB)
1661 * TIPC_MEDIUM_IMPORTANCE    (4 MB)
1662 * TIPC_HIGH_IMPORTANCE      (8 MB)
1663 * TIPC_CRITICAL_IMPORTANCE  (16 MB)
1664 *
1665 * Returns overload limit according to corresponding message importance
1666 */
1667static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
1668{
1669	struct tipc_sock *tsk = tipc_sk(sk);
1670	struct tipc_msg *hdr = buf_msg(skb);
1671
1672	if (unlikely(!msg_connected(hdr)))
1673		return sk->sk_rcvbuf << msg_importance(hdr);
1674
1675	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
1676		return sk->sk_rcvbuf;
1677
1678	return FLOWCTL_MSG_LIM;
 
1679}
1680
1681/**
1682 * filter_rcv - validate incoming message
1683 * @sk: socket
1684 * @skb: pointer to message.
1685 *
1686 * Enqueues message on receive queue if acceptable; optionally handles
1687 * disconnect indication for a connected socket.
1688 *
1689 * Called with socket lock already taken
1690 *
1691 * Returns true if message was added to socket receive queue, otherwise false
1692 */
1693static bool filter_rcv(struct sock *sk, struct sk_buff *skb,
1694		       struct sk_buff_head *xmitq)
1695{
 
1696	struct tipc_sock *tsk = tipc_sk(sk);
1697	struct tipc_msg *hdr = buf_msg(skb);
1698	unsigned int limit = rcvbuf_limit(sk, skb);
1699	int err = TIPC_OK;
1700	int usr = msg_user(hdr);
1701
1702	if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1703		tipc_sk_proto_rcv(tsk, skb, xmitq);
1704		return false;
1705	}
1706
1707	if (unlikely(usr == SOCK_WAKEUP)) {
1708		kfree_skb(skb);
1709		tsk->link_cong = 0;
1710		sk->sk_write_space(sk);
1711		return false;
1712	}
1713
1714	/* Drop if illegal message type */
1715	if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1716		kfree_skb(skb);
1717		return false;
1718	}
1719
1720	/* Reject if wrong message type for current socket state */
1721	if (tipc_sk_type_connectionless(sk)) {
1722		if (msg_connected(hdr)) {
1723			err = TIPC_ERR_NO_PORT;
1724			goto reject;
1725		}
1726	} else if (unlikely(!filter_connect(tsk, skb))) {
1727		err = TIPC_ERR_NO_PORT;
1728		goto reject;
1729	}
1730
1731	/* Reject message if there isn't room to queue it */
1732	if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1733		err = TIPC_ERR_OVERLOAD;
1734		goto reject;
1735	}
1736
1737	/* Enqueue message */
1738	TIPC_SKB_CB(skb)->bytes_read = 0;
1739	__skb_queue_tail(&sk->sk_receive_queue, skb);
1740	skb_set_owner_r(skb, sk);
1741
1742	sk->sk_data_ready(sk);
1743	return true;
1744
1745reject:
1746	if (tipc_msg_reverse(tsk_own_node(tsk), &skb, err))
1747		__skb_queue_tail(xmitq, skb);
1748	return false;
1749}
1750
1751/**
1752 * tipc_backlog_rcv - handle incoming message from backlog queue
1753 * @sk: socket
1754 * @skb: message
1755 *
1756 * Caller must hold socket lock
1757 *
1758 * Returns 0
1759 */
1760static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1761{
1762	unsigned int truesize = skb->truesize;
1763	struct sk_buff_head xmitq;
1764	u32 dnode, selector;
1765
1766	__skb_queue_head_init(&xmitq);
1767
1768	if (likely(filter_rcv(sk, skb, &xmitq))) {
1769		atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1770		return 0;
1771	}
1772
1773	if (skb_queue_empty(&xmitq))
1774		return 0;
1775
1776	/* Send response/rejected message */
1777	skb = __skb_dequeue(&xmitq);
1778	dnode = msg_destnode(buf_msg(skb));
1779	selector = msg_origport(buf_msg(skb));
1780	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
1781	return 0;
1782}
1783
1784/**
1785 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1786 *                   inputq and try adding them to socket or backlog queue
1787 * @inputq: list of incoming buffers with potentially different destinations
1788 * @sk: socket where the buffers should be enqueued
1789 * @dport: port number for the socket
1790 *
1791 * Caller must hold socket lock
1792 */
1793static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1794			    u32 dport, struct sk_buff_head *xmitq)
1795{
1796	unsigned long time_limit = jiffies + 2;
1797	struct sk_buff *skb;
1798	unsigned int lim;
1799	atomic_t *dcnt;
1800	u32 onode;
1801
1802	while (skb_queue_len(inputq)) {
1803		if (unlikely(time_after_eq(jiffies, time_limit)))
1804			return;
1805
1806		skb = tipc_skb_dequeue(inputq, dport);
1807		if (unlikely(!skb))
1808			return;
1809
1810		/* Add message directly to receive queue if possible */
1811		if (!sock_owned_by_user(sk)) {
1812			filter_rcv(sk, skb, xmitq);
1813			continue;
1814		}
1815
1816		/* Try backlog, compensating for double-counted bytes */
1817		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1818		if (!sk->sk_backlog.len)
1819			atomic_set(dcnt, 0);
1820		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1821		if (likely(!sk_add_backlog(sk, skb, lim)))
1822			continue;
1823
1824		/* Overload => reject message back to sender */
1825		onode = tipc_own_addr(sock_net(sk));
1826		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
1827			__skb_queue_tail(xmitq, skb);
1828		break;
 
1829	}
1830}
1831
1832/**
1833 * tipc_sk_rcv - handle a chain of incoming buffers
1834 * @inputq: buffer list containing the buffers
1835 * Consumes all buffers in list until inputq is empty
1836 * Note: may be called in multiple threads referring to the same queue
1837 */
1838void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1839{
1840	struct sk_buff_head xmitq;
1841	u32 dnode, dport = 0;
1842	int err;
1843	struct tipc_sock *tsk;
1844	struct sock *sk;
1845	struct sk_buff *skb;
1846
1847	__skb_queue_head_init(&xmitq);
1848	while (skb_queue_len(inputq)) {
1849		dport = tipc_skb_peek_port(inputq, dport);
1850		tsk = tipc_sk_lookup(net, dport);
1851
1852		if (likely(tsk)) {
1853			sk = &tsk->sk;
1854			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1855				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
1856				spin_unlock_bh(&sk->sk_lock.slock);
1857			}
1858			/* Send pending response/rejected messages, if any */
1859			while ((skb = __skb_dequeue(&xmitq))) {
1860				dnode = msg_destnode(buf_msg(skb));
1861				tipc_node_xmit_skb(net, skb, dnode, dport);
1862			}
1863			sock_put(sk);
1864			continue;
1865		}
1866
1867		/* No destination socket => dequeue skb if still there */
1868		skb = tipc_skb_dequeue(inputq, dport);
1869		if (!skb)
1870			return;
1871
1872		/* Try secondary lookup if unresolved named message */
1873		err = TIPC_ERR_NO_PORT;
1874		if (tipc_msg_lookup_dest(net, skb, &err))
1875			goto xmit;
1876
1877		/* Prepare for message rejection */
1878		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1879			continue;
1880xmit:
1881		dnode = msg_destnode(buf_msg(skb));
1882		tipc_node_xmit_skb(net, skb, dnode, dport);
1883	}
1884}
1885
1886static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1887{
1888	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1889	struct sock *sk = sock->sk;
 
1890	int done;
1891
1892	do {
1893		int err = sock_error(sk);
1894		if (err)
1895			return err;
1896		if (!*timeo_p)
1897			return -ETIMEDOUT;
1898		if (signal_pending(current))
1899			return sock_intr_errno(*timeo_p);
1900
1901		add_wait_queue(sk_sleep(sk), &wait);
1902		done = sk_wait_event(sk, timeo_p,
1903				     sk->sk_state != TIPC_CONNECTING, &wait);
1904		remove_wait_queue(sk_sleep(sk), &wait);
1905	} while (!done);
1906	return 0;
1907}
1908
1909/**
1910 * tipc_connect - establish a connection to another TIPC port
1911 * @sock: socket structure
1912 * @dest: socket address for destination port
1913 * @destlen: size of socket address data structure
1914 * @flags: file-related flags associated with socket
1915 *
1916 * Returns 0 on success, errno otherwise
1917 */
1918static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1919			int destlen, int flags)
1920{
1921	struct sock *sk = sock->sk;
1922	struct tipc_sock *tsk = tipc_sk(sk);
1923	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1924	struct msghdr m = {NULL,};
1925	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1926	int previous;
1927	int res = 0;
1928
1929	lock_sock(sk);
1930
1931	/* DGRAM/RDM connect(), just save the destaddr */
1932	if (tipc_sk_type_connectionless(sk)) {
1933		if (dst->family == AF_UNSPEC) {
1934			memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
1935		} else if (destlen != sizeof(struct sockaddr_tipc)) {
1936			res = -EINVAL;
1937		} else {
1938			memcpy(&tsk->peer, dest, destlen);
1939		}
1940		goto exit;
1941	}
1942
1943	/*
1944	 * Reject connection attempt using multicast address
1945	 *
1946	 * Note: send_msg() validates the rest of the address fields,
1947	 *       so there's no need to do it here
1948	 */
1949	if (dst->addrtype == TIPC_ADDR_MCAST) {
1950		res = -EINVAL;
1951		goto exit;
1952	}
1953
1954	previous = sk->sk_state;
1955
1956	switch (sk->sk_state) {
1957	case TIPC_OPEN:
1958		/* Send a 'SYN-' to destination */
1959		m.msg_name = dest;
1960		m.msg_namelen = destlen;
1961
1962		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1963		 * indicate send_msg() is never blocked.
1964		 */
1965		if (!timeout)
1966			m.msg_flags = MSG_DONTWAIT;
1967
1968		res = __tipc_sendmsg(sock, &m, 0);
1969		if ((res < 0) && (res != -EWOULDBLOCK))
1970			goto exit;
1971
1972		/* Just entered TIPC_CONNECTING state; the only
1973		 * difference is that return value in non-blocking
1974		 * case is EINPROGRESS, rather than EALREADY.
1975		 */
1976		res = -EINPROGRESS;
1977		/* fall thru' */
1978	case TIPC_CONNECTING:
1979		if (!timeout) {
1980			if (previous == TIPC_CONNECTING)
1981				res = -EALREADY;
1982			goto exit;
1983		}
1984		timeout = msecs_to_jiffies(timeout);
1985		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1986		res = tipc_wait_for_connect(sock, &timeout);
1987		break;
1988	case TIPC_ESTABLISHED:
1989		res = -EISCONN;
1990		break;
1991	default:
1992		res = -EINVAL;
 
1993	}
1994
1995exit:
1996	release_sock(sk);
1997	return res;
1998}
1999
2000/**
2001 * tipc_listen - allow socket to listen for incoming connections
2002 * @sock: socket structure
2003 * @len: (unused)
2004 *
2005 * Returns 0 on success, errno otherwise
2006 */
2007static int tipc_listen(struct socket *sock, int len)
2008{
2009	struct sock *sk = sock->sk;
2010	int res;
2011
2012	lock_sock(sk);
2013	res = tipc_set_sk_state(sk, TIPC_LISTEN);
 
 
 
 
 
 
 
2014	release_sock(sk);
2015
2016	return res;
2017}
2018
2019static int tipc_wait_for_accept(struct socket *sock, long timeo)
2020{
2021	struct sock *sk = sock->sk;
2022	DEFINE_WAIT(wait);
2023	int err;
2024
2025	/* True wake-one mechanism for incoming connections: only
2026	 * one process gets woken up, not the 'whole herd'.
2027	 * Since we do not 'race & poll' for established sockets
2028	 * anymore, the common case will execute the loop only once.
2029	*/
2030	for (;;) {
2031		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2032					  TASK_INTERRUPTIBLE);
2033		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2034			release_sock(sk);
2035			timeo = schedule_timeout(timeo);
2036			lock_sock(sk);
2037		}
2038		err = 0;
2039		if (!skb_queue_empty(&sk->sk_receive_queue))
2040			break;
2041		err = -EAGAIN;
2042		if (!timeo)
2043			break;
2044		err = sock_intr_errno(timeo);
2045		if (signal_pending(current))
2046			break;
 
 
 
2047	}
2048	finish_wait(sk_sleep(sk), &wait);
2049	return err;
2050}
2051
2052/**
2053 * tipc_accept - wait for connection request
2054 * @sock: listening socket
2055 * @newsock: new socket that is to be connected
2056 * @flags: file-related flags associated with socket
2057 *
2058 * Returns 0 on success, errno otherwise
2059 */
2060static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2061{
2062	struct sock *new_sk, *sk = sock->sk;
2063	struct sk_buff *buf;
2064	struct tipc_sock *new_tsock;
2065	struct tipc_msg *msg;
 
 
2066	long timeo;
2067	int res;
2068
2069	lock_sock(sk);
2070
2071	if (sk->sk_state != TIPC_LISTEN) {
2072		res = -EINVAL;
2073		goto exit;
2074	}
2075	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2076	res = tipc_wait_for_accept(sock, timeo);
2077	if (res)
2078		goto exit;
2079
2080	buf = skb_peek(&sk->sk_receive_queue);
2081
2082	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 0);
2083	if (res)
2084		goto exit;
2085	security_sk_clone(sock->sk, new_sock->sk);
2086
2087	new_sk = new_sock->sk;
2088	new_tsock = tipc_sk(new_sk);
 
2089	msg = buf_msg(buf);
2090
2091	/* we lock on new_sk; but lockdep sees the lock on sk */
2092	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2093
2094	/*
2095	 * Reject any stray messages received by new socket
2096	 * before the socket lock was taken (very, very unlikely)
2097	 */
2098	tsk_rej_rx_queue(new_sk);
2099
2100	/* Connect new socket to it's peer */
2101	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
 
 
 
2102
2103	tsk_set_importance(new_tsock, msg_importance(msg));
2104	if (msg_named(msg)) {
2105		new_tsock->conn_type = msg_nametype(msg);
2106		new_tsock->conn_instance = msg_nameinst(msg);
2107	}
2108
2109	/*
2110	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2111	 * Respond to 'SYN+' by queuing it on new socket.
2112	 */
2113	if (!msg_data_sz(msg)) {
2114		struct msghdr m = {NULL,};
2115
2116		tsk_advance_rx_queue(sk);
2117		__tipc_send_stream(new_sock, &m, 0);
2118	} else {
2119		__skb_dequeue(&sk->sk_receive_queue);
2120		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2121		skb_set_owner_r(buf, new_sk);
2122	}
2123	release_sock(new_sk);
2124exit:
2125	release_sock(sk);
2126	return res;
2127}
2128
2129/**
2130 * tipc_shutdown - shutdown socket connection
2131 * @sock: socket structure
2132 * @how: direction to close (must be SHUT_RDWR)
2133 *
2134 * Terminates connection (if necessary), then purges socket's receive queue.
2135 *
2136 * Returns 0 on success, errno otherwise
2137 */
2138static int tipc_shutdown(struct socket *sock, int how)
2139{
2140	struct sock *sk = sock->sk;
 
 
 
2141	int res;
2142
2143	if (how != SHUT_RDWR)
2144		return -EINVAL;
2145
2146	lock_sock(sk);
2147
2148	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2149	sk->sk_shutdown = SEND_SHUTDOWN;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2150
2151	if (sk->sk_state == TIPC_DISCONNECTING) {
2152		/* Discard any unreceived messages */
2153		__skb_queue_purge(&sk->sk_receive_queue);
2154
2155		/* Wake up anyone sleeping in poll */
2156		sk->sk_state_change(sk);
2157		res = 0;
2158	} else {
 
 
2159		res = -ENOTCONN;
2160	}
2161
2162	release_sock(sk);
2163	return res;
2164}
2165
2166static void tipc_sk_timeout(unsigned long data)
2167{
2168	struct tipc_sock *tsk = (struct tipc_sock *)data;
2169	struct sock *sk = &tsk->sk;
2170	struct sk_buff *skb = NULL;
2171	u32 peer_port, peer_node;
2172	u32 own_node = tsk_own_node(tsk);
2173
2174	bh_lock_sock(sk);
2175	if (!tipc_sk_connected(sk)) {
2176		bh_unlock_sock(sk);
2177		goto exit;
2178	}
2179	peer_port = tsk_peer_port(tsk);
2180	peer_node = tsk_peer_node(tsk);
2181
2182	if (tsk->probe_unacked) {
2183		if (!sock_owned_by_user(sk)) {
2184			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2185			tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2186					      tsk_peer_port(tsk));
2187			sk->sk_state_change(sk);
2188		} else {
2189			/* Try again later */
2190			sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2191		}
2192
2193		bh_unlock_sock(sk);
2194		goto exit;
2195	}
2196
2197	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2198			      INT_H_SIZE, 0, peer_node, own_node,
2199			      peer_port, tsk->portid, TIPC_OK);
2200	tsk->probe_unacked = true;
2201	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
2202	bh_unlock_sock(sk);
2203	if (skb)
2204		tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2205exit:
2206	sock_put(sk);
2207}
2208
2209static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2210			   struct tipc_name_seq const *seq)
2211{
2212	struct sock *sk = &tsk->sk;
2213	struct net *net = sock_net(sk);
2214	struct publication *publ;
2215	u32 key;
2216
2217	if (tipc_sk_connected(sk))
2218		return -EINVAL;
2219	key = tsk->portid + tsk->pub_count + 1;
2220	if (key == tsk->portid)
2221		return -EADDRINUSE;
2222
2223	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2224				    scope, tsk->portid, key);
2225	if (unlikely(!publ))
2226		return -EINVAL;
2227
2228	list_add(&publ->pport_list, &tsk->publications);
2229	tsk->pub_count++;
2230	tsk->published = 1;
2231	return 0;
2232}
2233
2234static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2235			    struct tipc_name_seq const *seq)
2236{
2237	struct net *net = sock_net(&tsk->sk);
2238	struct publication *publ;
2239	struct publication *safe;
2240	int rc = -EINVAL;
2241
2242	list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2243		if (seq) {
2244			if (publ->scope != scope)
2245				continue;
2246			if (publ->type != seq->type)
2247				continue;
2248			if (publ->lower != seq->lower)
2249				continue;
2250			if (publ->upper != seq->upper)
2251				break;
2252			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2253					      publ->ref, publ->key);
2254			rc = 0;
2255			break;
2256		}
2257		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2258				      publ->ref, publ->key);
2259		rc = 0;
2260	}
2261	if (list_empty(&tsk->publications))
2262		tsk->published = 0;
2263	return rc;
2264}
2265
2266/* tipc_sk_reinit: set non-zero address in all existing sockets
2267 *                 when we go from standalone to network mode.
2268 */
2269void tipc_sk_reinit(struct net *net)
2270{
2271	struct tipc_net *tn = net_generic(net, tipc_net_id);
2272	const struct bucket_table *tbl;
2273	struct rhash_head *pos;
2274	struct tipc_sock *tsk;
2275	struct tipc_msg *msg;
2276	int i;
2277
2278	rcu_read_lock();
2279	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2280	for (i = 0; i < tbl->size; i++) {
2281		rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2282			spin_lock_bh(&tsk->sk.sk_lock.slock);
2283			msg = &tsk->phdr;
2284			msg_set_prevnode(msg, tn->own_addr);
2285			msg_set_orignode(msg, tn->own_addr);
2286			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2287		}
2288	}
2289	rcu_read_unlock();
2290}
2291
2292static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2293{
2294	struct tipc_net *tn = net_generic(net, tipc_net_id);
2295	struct tipc_sock *tsk;
2296
2297	rcu_read_lock();
2298	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2299	if (tsk)
2300		sock_hold(&tsk->sk);
2301	rcu_read_unlock();
2302
2303	return tsk;
2304}
2305
2306static int tipc_sk_insert(struct tipc_sock *tsk)
2307{
2308	struct sock *sk = &tsk->sk;
2309	struct net *net = sock_net(sk);
2310	struct tipc_net *tn = net_generic(net, tipc_net_id);
2311	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2312	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2313
2314	while (remaining--) {
2315		portid++;
2316		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2317			portid = TIPC_MIN_PORT;
2318		tsk->portid = portid;
2319		sock_hold(&tsk->sk);
2320		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2321						   tsk_rht_params))
2322			return 0;
2323		sock_put(&tsk->sk);
2324	}
2325
2326	return -1;
2327}
2328
2329static void tipc_sk_remove(struct tipc_sock *tsk)
2330{
2331	struct sock *sk = &tsk->sk;
2332	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2333
2334	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2335		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2336		__sock_put(sk);
2337	}
2338}
2339
2340static const struct rhashtable_params tsk_rht_params = {
2341	.nelem_hint = 192,
2342	.head_offset = offsetof(struct tipc_sock, node),
2343	.key_offset = offsetof(struct tipc_sock, portid),
2344	.key_len = sizeof(u32), /* portid */
2345	.max_size = 1048576,
2346	.min_size = 256,
2347	.automatic_shrinking = true,
2348};
2349
2350int tipc_sk_rht_init(struct net *net)
2351{
2352	struct tipc_net *tn = net_generic(net, tipc_net_id);
2353
2354	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2355}
2356
2357void tipc_sk_rht_destroy(struct net *net)
2358{
2359	struct tipc_net *tn = net_generic(net, tipc_net_id);
2360
2361	/* Wait for socket readers to complete */
2362	synchronize_net();
2363
2364	rhashtable_destroy(&tn->sk_rht);
2365}
2366
2367/**
2368 * tipc_setsockopt - set socket option
2369 * @sock: socket structure
2370 * @lvl: option level
2371 * @opt: option identifier
2372 * @ov: pointer to new option value
2373 * @ol: length of option value
2374 *
2375 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2376 * (to ease compatibility).
2377 *
2378 * Returns 0 on success, errno otherwise
2379 */
2380static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2381			   char __user *ov, unsigned int ol)
2382{
2383	struct sock *sk = sock->sk;
2384	struct tipc_sock *tsk = tipc_sk(sk);
 
2385	u32 value;
2386	int res;
2387
2388	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2389		return 0;
2390	if (lvl != SOL_TIPC)
2391		return -ENOPROTOOPT;
2392	if (ol < sizeof(value))
2393		return -EINVAL;
2394	res = get_user(value, (u32 __user *)ov);
2395	if (res)
2396		return res;
2397
2398	lock_sock(sk);
2399
2400	switch (opt) {
2401	case TIPC_IMPORTANCE:
2402		res = tsk_set_importance(tsk, value);
2403		break;
2404	case TIPC_SRC_DROPPABLE:
2405		if (sock->type != SOCK_STREAM)
2406			tsk_set_unreliable(tsk, value);
2407		else
2408			res = -ENOPROTOOPT;
2409		break;
2410	case TIPC_DEST_DROPPABLE:
2411		tsk_set_unreturnable(tsk, value);
2412		break;
2413	case TIPC_CONN_TIMEOUT:
2414		tipc_sk(sk)->conn_timeout = value;
2415		/* no need to set "res", since already 0 at this point */
2416		break;
2417	default:
2418		res = -EINVAL;
2419	}
2420
2421	release_sock(sk);
2422
2423	return res;
2424}
2425
2426/**
2427 * tipc_getsockopt - get socket option
2428 * @sock: socket structure
2429 * @lvl: option level
2430 * @opt: option identifier
2431 * @ov: receptacle for option value
2432 * @ol: receptacle for length of option value
2433 *
2434 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2435 * (to ease compatibility).
2436 *
2437 * Returns 0 on success, errno otherwise
2438 */
2439static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2440			   char __user *ov, int __user *ol)
2441{
2442	struct sock *sk = sock->sk;
2443	struct tipc_sock *tsk = tipc_sk(sk);
 
2444	int len;
2445	u32 value;
2446	int res;
2447
2448	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2449		return put_user(0, ol);
2450	if (lvl != SOL_TIPC)
2451		return -ENOPROTOOPT;
2452	res = get_user(len, ol);
2453	if (res)
2454		return res;
2455
2456	lock_sock(sk);
2457
2458	switch (opt) {
2459	case TIPC_IMPORTANCE:
2460		value = tsk_importance(tsk);
2461		break;
2462	case TIPC_SRC_DROPPABLE:
2463		value = tsk_unreliable(tsk);
2464		break;
2465	case TIPC_DEST_DROPPABLE:
2466		value = tsk_unreturnable(tsk);
2467		break;
2468	case TIPC_CONN_TIMEOUT:
2469		value = tsk->conn_timeout;
2470		/* no need to set "res", since already 0 at this point */
2471		break;
2472	case TIPC_NODE_RECVQ_DEPTH:
2473		value = 0; /* was tipc_queue_size, now obsolete */
2474		break;
2475	case TIPC_SOCK_RECVQ_DEPTH:
2476		value = skb_queue_len(&sk->sk_receive_queue);
2477		break;
2478	default:
2479		res = -EINVAL;
2480	}
2481
2482	release_sock(sk);
2483
2484	if (res)
2485		return res;	/* "get" failed */
2486
2487	if (len < sizeof(value))
2488		return -EINVAL;
2489
2490	if (copy_to_user(ov, &value, sizeof(value)))
2491		return -EFAULT;
2492
2493	return put_user(sizeof(value), ol);
2494}
2495
2496static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2497{
2498	struct sock *sk = sock->sk;
2499	struct tipc_sioc_ln_req lnr;
2500	void __user *argp = (void __user *)arg;
2501
2502	switch (cmd) {
2503	case SIOCGETLINKNAME:
2504		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2505			return -EFAULT;
2506		if (!tipc_node_get_linkname(sock_net(sk),
2507					    lnr.bearer_id & 0xffff, lnr.peer,
2508					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2509			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2510				return -EFAULT;
2511			return 0;
2512		}
2513		return -EADDRNOTAVAIL;
2514	default:
2515		return -ENOIOCTLCMD;
2516	}
2517}
2518
2519/* Protocol switches for the various types of TIPC sockets */
2520
2521static const struct proto_ops msg_ops = {
2522	.owner		= THIS_MODULE,
2523	.family		= AF_TIPC,
2524	.release	= tipc_release,
2525	.bind		= tipc_bind,
2526	.connect	= tipc_connect,
2527	.socketpair	= sock_no_socketpair,
2528	.accept		= sock_no_accept,
2529	.getname	= tipc_getname,
2530	.poll		= tipc_poll,
2531	.ioctl		= tipc_ioctl,
2532	.listen		= sock_no_listen,
2533	.shutdown	= tipc_shutdown,
2534	.setsockopt	= tipc_setsockopt,
2535	.getsockopt	= tipc_getsockopt,
2536	.sendmsg	= tipc_sendmsg,
2537	.recvmsg	= tipc_recvmsg,
2538	.mmap		= sock_no_mmap,
2539	.sendpage	= sock_no_sendpage
2540};
2541
2542static const struct proto_ops packet_ops = {
2543	.owner		= THIS_MODULE,
2544	.family		= AF_TIPC,
2545	.release	= tipc_release,
2546	.bind		= tipc_bind,
2547	.connect	= tipc_connect,
2548	.socketpair	= sock_no_socketpair,
2549	.accept		= tipc_accept,
2550	.getname	= tipc_getname,
2551	.poll		= tipc_poll,
2552	.ioctl		= tipc_ioctl,
2553	.listen		= tipc_listen,
2554	.shutdown	= tipc_shutdown,
2555	.setsockopt	= tipc_setsockopt,
2556	.getsockopt	= tipc_getsockopt,
2557	.sendmsg	= tipc_send_packet,
2558	.recvmsg	= tipc_recvmsg,
2559	.mmap		= sock_no_mmap,
2560	.sendpage	= sock_no_sendpage
2561};
2562
2563static const struct proto_ops stream_ops = {
2564	.owner		= THIS_MODULE,
2565	.family		= AF_TIPC,
2566	.release	= tipc_release,
2567	.bind		= tipc_bind,
2568	.connect	= tipc_connect,
2569	.socketpair	= sock_no_socketpair,
2570	.accept		= tipc_accept,
2571	.getname	= tipc_getname,
2572	.poll		= tipc_poll,
2573	.ioctl		= tipc_ioctl,
2574	.listen		= tipc_listen,
2575	.shutdown	= tipc_shutdown,
2576	.setsockopt	= tipc_setsockopt,
2577	.getsockopt	= tipc_getsockopt,
2578	.sendmsg	= tipc_send_stream,
2579	.recvmsg	= tipc_recv_stream,
2580	.mmap		= sock_no_mmap,
2581	.sendpage	= sock_no_sendpage
2582};
2583
2584static const struct net_proto_family tipc_family_ops = {
2585	.owner		= THIS_MODULE,
2586	.family		= AF_TIPC,
2587	.create		= tipc_sk_create
2588};
2589
2590static struct proto tipc_proto = {
2591	.name		= "TIPC",
2592	.owner		= THIS_MODULE,
2593	.obj_size	= sizeof(struct tipc_sock),
2594	.sysctl_rmem	= sysctl_tipc_rmem
2595};
2596
 
 
 
 
 
 
2597/**
2598 * tipc_socket_init - initialize TIPC socket interface
2599 *
2600 * Returns 0 on success, errno otherwise
2601 */
2602int tipc_socket_init(void)
2603{
2604	int res;
2605
2606	res = proto_register(&tipc_proto, 1);
2607	if (res) {
2608		pr_err("Failed to register TIPC protocol type\n");
2609		goto out;
2610	}
2611
2612	res = sock_register(&tipc_family_ops);
2613	if (res) {
2614		pr_err("Failed to register TIPC socket type\n");
2615		proto_unregister(&tipc_proto);
2616		goto out;
2617	}
2618 out:
2619	return res;
2620}
2621
2622/**
2623 * tipc_socket_stop - stop TIPC socket interface
2624 */
2625void tipc_socket_stop(void)
2626{
2627	sock_unregister(tipc_family_ops.family);
2628	proto_unregister(&tipc_proto);
2629}
2630
2631/* Caller should hold socket lock for the passed tipc socket. */
2632static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2633{
2634	u32 peer_node;
2635	u32 peer_port;
2636	struct nlattr *nest;
2637
2638	peer_node = tsk_peer_node(tsk);
2639	peer_port = tsk_peer_port(tsk);
2640
2641	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2642
2643	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2644		goto msg_full;
2645	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2646		goto msg_full;
2647
2648	if (tsk->conn_type != 0) {
2649		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2650			goto msg_full;
2651		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2652			goto msg_full;
2653		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2654			goto msg_full;
2655	}
2656	nla_nest_end(skb, nest);
2657
2658	return 0;
2659
2660msg_full:
2661	nla_nest_cancel(skb, nest);
2662
2663	return -EMSGSIZE;
2664}
2665
2666/* Caller should hold socket lock for the passed tipc socket. */
2667static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2668			    struct tipc_sock *tsk)
2669{
2670	int err;
2671	void *hdr;
2672	struct nlattr *attrs;
2673	struct net *net = sock_net(skb->sk);
2674	struct tipc_net *tn = net_generic(net, tipc_net_id);
2675	struct sock *sk = &tsk->sk;
2676
2677	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2678			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2679	if (!hdr)
2680		goto msg_cancel;
2681
2682	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2683	if (!attrs)
2684		goto genlmsg_cancel;
2685	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2686		goto attr_msg_cancel;
2687	if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2688		goto attr_msg_cancel;
2689
2690	if (tipc_sk_connected(sk)) {
2691		err = __tipc_nl_add_sk_con(skb, tsk);
2692		if (err)
2693			goto attr_msg_cancel;
2694	} else if (!list_empty(&tsk->publications)) {
2695		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2696			goto attr_msg_cancel;
2697	}
2698	nla_nest_end(skb, attrs);
2699	genlmsg_end(skb, hdr);
2700
2701	return 0;
2702
2703attr_msg_cancel:
2704	nla_nest_cancel(skb, attrs);
2705genlmsg_cancel:
2706	genlmsg_cancel(skb, hdr);
2707msg_cancel:
2708	return -EMSGSIZE;
2709}
2710
2711int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2712{
2713	int err;
2714	struct tipc_sock *tsk;
2715	const struct bucket_table *tbl;
2716	struct rhash_head *pos;
2717	struct net *net = sock_net(skb->sk);
2718	struct tipc_net *tn = net_generic(net, tipc_net_id);
2719	u32 tbl_id = cb->args[0];
2720	u32 prev_portid = cb->args[1];
2721
2722	rcu_read_lock();
2723	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2724	for (; tbl_id < tbl->size; tbl_id++) {
2725		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2726			spin_lock_bh(&tsk->sk.sk_lock.slock);
2727			if (prev_portid && prev_portid != tsk->portid) {
2728				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2729				continue;
2730			}
2731
2732			err = __tipc_nl_add_sk(skb, cb, tsk);
2733			if (err) {
2734				prev_portid = tsk->portid;
2735				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2736				goto out;
2737			}
2738			prev_portid = 0;
2739			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2740		}
2741	}
2742out:
2743	rcu_read_unlock();
2744	cb->args[0] = tbl_id;
2745	cb->args[1] = prev_portid;
2746
2747	return skb->len;
2748}
2749
2750/* Caller should hold socket lock for the passed tipc socket. */
2751static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2752				 struct netlink_callback *cb,
2753				 struct publication *publ)
2754{
2755	void *hdr;
2756	struct nlattr *attrs;
2757
2758	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2759			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2760	if (!hdr)
2761		goto msg_cancel;
2762
2763	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2764	if (!attrs)
2765		goto genlmsg_cancel;
2766
2767	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2768		goto attr_msg_cancel;
2769	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2770		goto attr_msg_cancel;
2771	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2772		goto attr_msg_cancel;
2773	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2774		goto attr_msg_cancel;
2775
2776	nla_nest_end(skb, attrs);
2777	genlmsg_end(skb, hdr);
2778
2779	return 0;
2780
2781attr_msg_cancel:
2782	nla_nest_cancel(skb, attrs);
2783genlmsg_cancel:
2784	genlmsg_cancel(skb, hdr);
2785msg_cancel:
2786	return -EMSGSIZE;
2787}
2788
2789/* Caller should hold socket lock for the passed tipc socket. */
2790static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2791				  struct netlink_callback *cb,
2792				  struct tipc_sock *tsk, u32 *last_publ)
2793{
2794	int err;
2795	struct publication *p;
2796
2797	if (*last_publ) {
2798		list_for_each_entry(p, &tsk->publications, pport_list) {
2799			if (p->key == *last_publ)
2800				break;
2801		}
2802		if (p->key != *last_publ) {
2803			/* We never set seq or call nl_dump_check_consistent()
2804			 * this means that setting prev_seq here will cause the
2805			 * consistence check to fail in the netlink callback
2806			 * handler. Resulting in the last NLMSG_DONE message
2807			 * having the NLM_F_DUMP_INTR flag set.
2808			 */
2809			cb->prev_seq = 1;
2810			*last_publ = 0;
2811			return -EPIPE;
2812		}
2813	} else {
2814		p = list_first_entry(&tsk->publications, struct publication,
2815				     pport_list);
2816	}
2817
2818	list_for_each_entry_from(p, &tsk->publications, pport_list) {
2819		err = __tipc_nl_add_sk_publ(skb, cb, p);
2820		if (err) {
2821			*last_publ = p->key;
2822			return err;
2823		}
2824	}
2825	*last_publ = 0;
2826
2827	return 0;
2828}
2829
2830int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2831{
2832	int err;
2833	u32 tsk_portid = cb->args[0];
2834	u32 last_publ = cb->args[1];
2835	u32 done = cb->args[2];
2836	struct net *net = sock_net(skb->sk);
2837	struct tipc_sock *tsk;
2838
2839	if (!tsk_portid) {
2840		struct nlattr **attrs;
2841		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2842
2843		err = tipc_nlmsg_parse(cb->nlh, &attrs);
2844		if (err)
2845			return err;
2846
2847		if (!attrs[TIPC_NLA_SOCK])
2848			return -EINVAL;
2849
2850		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2851				       attrs[TIPC_NLA_SOCK],
2852				       tipc_nl_sock_policy);
2853		if (err)
2854			return err;
2855
2856		if (!sock[TIPC_NLA_SOCK_REF])
2857			return -EINVAL;
2858
2859		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2860	}
2861
2862	if (done)
2863		return 0;
2864
2865	tsk = tipc_sk_lookup(net, tsk_portid);
2866	if (!tsk)
2867		return -EINVAL;
2868
2869	lock_sock(&tsk->sk);
2870	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2871	if (!err)
2872		done = 1;
2873	release_sock(&tsk->sk);
2874	sock_put(&tsk->sk);
2875
2876	cb->args[0] = tsk_portid;
2877	cb->args[1] = last_publ;
2878	cb->args[2] = done;
2879
2880	return skb->len;
2881}