Linux Audio

Check our new training course

Loading...
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}
v5.4
   1/*
   2 * net/tipc/socket.c: TIPC socket API
   3 *
   4 * Copyright (c) 2001-2007, 2012-2017, 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 <linux/sched/signal.h>
  39
  40#include "core.h"
  41#include "name_table.h"
  42#include "node.h"
  43#include "link.h"
  44#include "name_distr.h"
  45#include "socket.h"
  46#include "bcast.h"
  47#include "netlink.h"
  48#include "group.h"
  49#include "trace.h"
  50
  51#define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
  52#define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
  53#define TIPC_FWD_MSG		1
  54#define TIPC_MAX_PORT		0xffffffff
  55#define TIPC_MIN_PORT		1
  56#define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
  57
  58enum {
  59	TIPC_LISTEN = TCP_LISTEN,
  60	TIPC_ESTABLISHED = TCP_ESTABLISHED,
  61	TIPC_OPEN = TCP_CLOSE,
  62	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
  63	TIPC_CONNECTING = TCP_SYN_SENT,
  64};
  65
  66struct sockaddr_pair {
  67	struct sockaddr_tipc sock;
  68	struct sockaddr_tipc member;
  69};
  70
  71/**
  72 * struct tipc_sock - TIPC socket structure
  73 * @sk: socket - interacts with 'port' and with user via the socket API
  74 * @conn_type: TIPC type used when connection was established
  75 * @conn_instance: TIPC instance used when connection was established
  76 * @published: non-zero if port has one or more associated names
  77 * @max_pkt: maximum packet size "hint" used when building messages sent by port
  78 * @portid: unique port identity in TIPC socket hash table
  79 * @phdr: preformatted message header used when sending messages
  80 * #cong_links: list of congested links
  81 * @publications: list of publications for port
  82 * @blocking_link: address of the congested link we are currently sleeping on
  83 * @pub_count: total # of publications port has made during its lifetime
 
  84 * @conn_timeout: the time we can wait for an unresponded setup request
  85 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
  86 * @cong_link_cnt: number of congested links
  87 * @snt_unacked: # messages sent by socket, and not yet acked by peer
  88 * @rcv_unacked: # messages read by user, but not yet acked back to peer
  89 * @peer: 'connected' peer for dgram/rdm
  90 * @node: hash table node
  91 * @mc_method: cookie for use between socket and broadcast layer
  92 * @rcu: rcu struct for tipc_sock
  93 */
  94struct tipc_sock {
  95	struct sock sk;
  96	u32 conn_type;
  97	u32 conn_instance;
  98	int published;
  99	u32 max_pkt;
 100	u32 portid;
 101	struct tipc_msg phdr;
 102	struct list_head cong_links;
 103	struct list_head publications;
 104	u32 pub_count;
 
 105	atomic_t dupl_rcvcnt;
 106	u16 conn_timeout;
 107	bool probe_unacked;
 108	u16 cong_link_cnt;
 109	u16 snt_unacked;
 110	u16 snd_win;
 111	u16 peer_caps;
 112	u16 rcv_unacked;
 113	u16 rcv_win;
 114	struct sockaddr_tipc peer;
 115	struct rhash_head node;
 116	struct tipc_mc_method mc_method;
 117	struct rcu_head rcu;
 118	struct tipc_group *group;
 119	bool group_is_open;
 120};
 121
 122static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
 123static void tipc_data_ready(struct sock *sk);
 124static void tipc_write_space(struct sock *sk);
 125static void tipc_sock_destruct(struct sock *sk);
 126static int tipc_release(struct socket *sock);
 127static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
 128		       bool kern);
 129static void tipc_sk_timeout(struct timer_list *t);
 130static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
 131			   struct tipc_name_seq const *seq);
 132static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
 133			    struct tipc_name_seq const *seq);
 134static int tipc_sk_leave(struct tipc_sock *tsk);
 135static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
 136static int tipc_sk_insert(struct tipc_sock *tsk);
 137static void tipc_sk_remove(struct tipc_sock *tsk);
 138static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
 
 139static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
 140
 141static const struct proto_ops packet_ops;
 142static const struct proto_ops stream_ops;
 143static const struct proto_ops msg_ops;
 144static struct proto tipc_proto;
 145static const struct rhashtable_params tsk_rht_params;
 146
 147static u32 tsk_own_node(struct tipc_sock *tsk)
 148{
 149	return msg_prevnode(&tsk->phdr);
 150}
 151
 152static u32 tsk_peer_node(struct tipc_sock *tsk)
 153{
 154	return msg_destnode(&tsk->phdr);
 155}
 156
 157static u32 tsk_peer_port(struct tipc_sock *tsk)
 158{
 159	return msg_destport(&tsk->phdr);
 160}
 161
 162static  bool tsk_unreliable(struct tipc_sock *tsk)
 163{
 164	return msg_src_droppable(&tsk->phdr) != 0;
 165}
 166
 167static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
 168{
 169	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
 170}
 171
 172static bool tsk_unreturnable(struct tipc_sock *tsk)
 173{
 174	return msg_dest_droppable(&tsk->phdr) != 0;
 175}
 176
 177static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
 178{
 179	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
 180}
 181
 182static int tsk_importance(struct tipc_sock *tsk)
 183{
 184	return msg_importance(&tsk->phdr);
 185}
 186
 187static int tsk_set_importance(struct tipc_sock *tsk, int imp)
 188{
 189	if (imp > TIPC_CRITICAL_IMPORTANCE)
 190		return -EINVAL;
 191	msg_set_importance(&tsk->phdr, (u32)imp);
 192	return 0;
 193}
 194
 195static struct tipc_sock *tipc_sk(const struct sock *sk)
 196{
 197	return container_of(sk, struct tipc_sock, sk);
 198}
 199
 200static bool tsk_conn_cong(struct tipc_sock *tsk)
 201{
 202	return tsk->snt_unacked > tsk->snd_win;
 203}
 204
 205static u16 tsk_blocks(int len)
 206{
 207	return ((len / FLOWCTL_BLK_SZ) + 1);
 208}
 209
 210/* tsk_blocks(): translate a buffer size in bytes to number of
 211 * advertisable blocks, taking into account the ratio truesize(len)/len
 212 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
 213 */
 214static u16 tsk_adv_blocks(int len)
 215{
 216	return len / FLOWCTL_BLK_SZ / 4;
 217}
 218
 219/* tsk_inc(): increment counter for sent or received data
 220 * - If block based flow control is not supported by peer we
 221 *   fall back to message based ditto, incrementing the counter
 222 */
 223static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
 224{
 225	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
 226		return ((msglen / FLOWCTL_BLK_SZ) + 1);
 227	return 1;
 228}
 229
 230/**
 231 * tsk_advance_rx_queue - discard first buffer in socket receive queue
 232 *
 233 * Caller must hold socket lock
 234 */
 235static void tsk_advance_rx_queue(struct sock *sk)
 236{
 237	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
 238	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
 239}
 240
 241/* tipc_sk_respond() : send response message back to sender
 242 */
 243static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
 244{
 245	u32 selector;
 246	u32 dnode;
 247	u32 onode = tipc_own_addr(sock_net(sk));
 248
 249	if (!tipc_msg_reverse(onode, &skb, err))
 250		return;
 251
 252	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
 253	dnode = msg_destnode(buf_msg(skb));
 254	selector = msg_origport(buf_msg(skb));
 255	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
 256}
 257
 258/**
 259 * tsk_rej_rx_queue - reject all buffers in socket receive queue
 260 *
 261 * Caller must hold socket lock
 262 */
 263static void tsk_rej_rx_queue(struct sock *sk)
 264{
 265	struct sk_buff *skb;
 266
 267	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
 268		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
 269}
 270
 271static bool tipc_sk_connected(struct sock *sk)
 272{
 273	return sk->sk_state == TIPC_ESTABLISHED;
 274}
 275
 276/* tipc_sk_type_connectionless - check if the socket is datagram socket
 277 * @sk: socket
 278 *
 279 * Returns true if connection less, false otherwise
 280 */
 281static bool tipc_sk_type_connectionless(struct sock *sk)
 282{
 283	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
 284}
 285
 286/* tsk_peer_msg - verify if message was sent by connected port's peer
 287 *
 288 * Handles cases where the node's network address has changed from
 289 * the default of <0.0.0> to its configured setting.
 290 */
 291static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
 292{
 293	struct sock *sk = &tsk->sk;
 294	u32 self = tipc_own_addr(sock_net(sk));
 295	u32 peer_port = tsk_peer_port(tsk);
 296	u32 orig_node, peer_node;
 
 297
 298	if (unlikely(!tipc_sk_connected(sk)))
 299		return false;
 300
 301	if (unlikely(msg_origport(msg) != peer_port))
 302		return false;
 303
 304	orig_node = msg_orignode(msg);
 305	peer_node = tsk_peer_node(tsk);
 306
 307	if (likely(orig_node == peer_node))
 308		return true;
 309
 310	if (!orig_node && peer_node == self)
 311		return true;
 312
 313	if (!peer_node && orig_node == self)
 314		return true;
 315
 316	return false;
 317}
 318
 319/* tipc_set_sk_state - set the sk_state of the socket
 320 * @sk: socket
 321 *
 322 * Caller must hold socket lock
 323 *
 324 * Returns 0 on success, errno otherwise
 325 */
 326static int tipc_set_sk_state(struct sock *sk, int state)
 327{
 328	int oldsk_state = sk->sk_state;
 329	int res = -EINVAL;
 330
 331	switch (state) {
 332	case TIPC_OPEN:
 333		res = 0;
 334		break;
 335	case TIPC_LISTEN:
 336	case TIPC_CONNECTING:
 337		if (oldsk_state == TIPC_OPEN)
 338			res = 0;
 339		break;
 340	case TIPC_ESTABLISHED:
 341		if (oldsk_state == TIPC_CONNECTING ||
 342		    oldsk_state == TIPC_OPEN)
 343			res = 0;
 344		break;
 345	case TIPC_DISCONNECTING:
 346		if (oldsk_state == TIPC_CONNECTING ||
 347		    oldsk_state == TIPC_ESTABLISHED)
 348			res = 0;
 349		break;
 350	}
 351
 352	if (!res)
 353		sk->sk_state = state;
 354
 355	return res;
 356}
 357
 358static int tipc_sk_sock_err(struct socket *sock, long *timeout)
 359{
 360	struct sock *sk = sock->sk;
 361	int err = sock_error(sk);
 362	int typ = sock->type;
 363
 364	if (err)
 365		return err;
 366	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
 367		if (sk->sk_state == TIPC_DISCONNECTING)
 368			return -EPIPE;
 369		else if (!tipc_sk_connected(sk))
 370			return -ENOTCONN;
 371	}
 372	if (!*timeout)
 373		return -EAGAIN;
 374	if (signal_pending(current))
 375		return sock_intr_errno(*timeout);
 376
 377	return 0;
 378}
 379
 380#define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
 381({                                                                             \
 382	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
 383	struct sock *sk_;						       \
 384	int rc_;							       \
 385									       \
 386	while ((rc_ = !(condition_))) {					       \
 387		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
 388		smp_rmb();                                                     \
 389		sk_ = (sock_)->sk;					       \
 390		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
 391		if (rc_)						       \
 392			break;						       \
 393		add_wait_queue(sk_sleep(sk_), &wait_);                         \
 394		release_sock(sk_);					       \
 395		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
 396		sched_annotate_sleep();				               \
 397		lock_sock(sk_);						       \
 398		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
 399	}								       \
 400	rc_;								       \
 401})
 402
 403/**
 404 * tipc_sk_create - create a TIPC socket
 405 * @net: network namespace (must be default network)
 406 * @sock: pre-allocated socket structure
 407 * @protocol: protocol indicator (must be 0)
 408 * @kern: caused by kernel or by userspace?
 409 *
 410 * This routine creates additional data structures used by the TIPC socket,
 411 * initializes them, and links them together.
 412 *
 413 * Returns 0 on success, errno otherwise
 414 */
 415static int tipc_sk_create(struct net *net, struct socket *sock,
 416			  int protocol, int kern)
 417{
 
 418	const struct proto_ops *ops;
 419	struct sock *sk;
 420	struct tipc_sock *tsk;
 421	struct tipc_msg *msg;
 422
 423	/* Validate arguments */
 424	if (unlikely(protocol != 0))
 425		return -EPROTONOSUPPORT;
 426
 427	switch (sock->type) {
 428	case SOCK_STREAM:
 429		ops = &stream_ops;
 430		break;
 431	case SOCK_SEQPACKET:
 432		ops = &packet_ops;
 433		break;
 434	case SOCK_DGRAM:
 435	case SOCK_RDM:
 436		ops = &msg_ops;
 437		break;
 438	default:
 439		return -EPROTOTYPE;
 440	}
 441
 442	/* Allocate socket's protocol area */
 443	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
 444	if (sk == NULL)
 445		return -ENOMEM;
 446
 447	tsk = tipc_sk(sk);
 448	tsk->max_pkt = MAX_PKT_DEFAULT;
 449	INIT_LIST_HEAD(&tsk->publications);
 450	INIT_LIST_HEAD(&tsk->cong_links);
 451	msg = &tsk->phdr;
 
 
 
 452
 453	/* Finish initializing socket data structures */
 454	sock->ops = ops;
 455	sock_init_data(sock, sk);
 456	tipc_set_sk_state(sk, TIPC_OPEN);
 457	if (tipc_sk_insert(tsk)) {
 458		pr_warn("Socket create failed; port number exhausted\n");
 459		return -EINVAL;
 460	}
 461
 462	/* Ensure tsk is visible before we read own_addr. */
 463	smp_mb();
 464
 465	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
 466		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
 467
 468	msg_set_origport(msg, tsk->portid);
 469	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
 470	sk->sk_shutdown = 0;
 471	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
 472	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
 473	sk->sk_data_ready = tipc_data_ready;
 474	sk->sk_write_space = tipc_write_space;
 475	sk->sk_destruct = tipc_sock_destruct;
 476	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
 477	tsk->group_is_open = true;
 478	atomic_set(&tsk->dupl_rcvcnt, 0);
 479
 480	/* Start out with safe limits until we receive an advertised window */
 481	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
 482	tsk->rcv_win = tsk->snd_win;
 483
 484	if (tipc_sk_type_connectionless(sk)) {
 485		tsk_set_unreturnable(tsk, true);
 486		if (sock->type == SOCK_DGRAM)
 487			tsk_set_unreliable(tsk, true);
 488	}
 489	__skb_queue_head_init(&tsk->mc_method.deferredq);
 490	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
 491	return 0;
 492}
 493
 494static void tipc_sk_callback(struct rcu_head *head)
 495{
 496	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
 497
 498	sock_put(&tsk->sk);
 499}
 500
 501/* Caller should hold socket lock for the socket. */
 502static void __tipc_shutdown(struct socket *sock, int error)
 503{
 504	struct sock *sk = sock->sk;
 505	struct tipc_sock *tsk = tipc_sk(sk);
 506	struct net *net = sock_net(sk);
 507	long timeout = CONN_TIMEOUT_DEFAULT;
 508	u32 dnode = tsk_peer_node(tsk);
 509	struct sk_buff *skb;
 510
 511	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
 512	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
 513					    !tsk_conn_cong(tsk)));
 514
 515	/* Remove any pending SYN message */
 516	__skb_queue_purge(&sk->sk_write_queue);
 517
 518	/* Reject all unreceived messages, except on an active connection
 519	 * (which disconnects locally & sends a 'FIN+' to peer).
 520	 */
 521	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
 522		if (TIPC_SKB_CB(skb)->bytes_read) {
 523			kfree_skb(skb);
 524			continue;
 525		}
 526		if (!tipc_sk_type_connectionless(sk) &&
 527		    sk->sk_state != TIPC_DISCONNECTING) {
 528			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 529			tipc_node_remove_conn(net, dnode, tsk->portid);
 530		}
 531		tipc_sk_respond(sk, skb, error);
 532	}
 533
 534	if (tipc_sk_type_connectionless(sk))
 535		return;
 536
 537	if (sk->sk_state != TIPC_DISCONNECTING) {
 538		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
 539				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
 540				      tsk_own_node(tsk), tsk_peer_port(tsk),
 541				      tsk->portid, error);
 542		if (skb)
 543			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
 544		tipc_node_remove_conn(net, dnode, tsk->portid);
 545		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 546	}
 547}
 548
 549/**
 550 * tipc_release - destroy a TIPC socket
 551 * @sock: socket to destroy
 552 *
 553 * This routine cleans up any messages that are still queued on the socket.
 554 * For DGRAM and RDM socket types, all queued messages are rejected.
 555 * For SEQPACKET and STREAM socket types, the first message is rejected
 556 * and any others are discarded.  (If the first message on a STREAM socket
 557 * is partially-read, it is discarded and the next one is rejected instead.)
 558 *
 559 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 560 * are returned or discarded according to the "destination droppable" setting
 561 * specified for the message by the sender.
 562 *
 563 * Returns 0 on success, errno otherwise
 564 */
 565static int tipc_release(struct socket *sock)
 566{
 567	struct sock *sk = sock->sk;
 568	struct tipc_sock *tsk;
 569
 570	/*
 571	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 572	 * releases a pre-allocated child socket that was never used)
 573	 */
 574	if (sk == NULL)
 575		return 0;
 576
 577	tsk = tipc_sk(sk);
 578	lock_sock(sk);
 579
 580	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
 581	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
 582	sk->sk_shutdown = SHUTDOWN_MASK;
 583	tipc_sk_leave(tsk);
 584	tipc_sk_withdraw(tsk, 0, NULL);
 585	__skb_queue_purge(&tsk->mc_method.deferredq);
 586	sk_stop_timer(sk, &sk->sk_timer);
 587	tipc_sk_remove(tsk);
 588
 589	sock_orphan(sk);
 590	/* Reject any messages that accumulated in backlog queue */
 591	release_sock(sk);
 592	tipc_dest_list_purge(&tsk->cong_links);
 593	tsk->cong_link_cnt = 0;
 594	call_rcu(&tsk->rcu, tipc_sk_callback);
 595	sock->sk = NULL;
 596
 597	return 0;
 598}
 599
 600/**
 601 * tipc_bind - associate or disassocate TIPC name(s) with a socket
 602 * @sock: socket structure
 603 * @uaddr: socket address describing name(s) and desired operation
 604 * @uaddr_len: size of socket address data structure
 605 *
 606 * Name and name sequence binding is indicated using a positive scope value;
 607 * a negative scope value unbinds the specified name.  Specifying no name
 608 * (i.e. a socket address length of 0) unbinds all names from the socket.
 609 *
 610 * Returns 0 on success, errno otherwise
 611 *
 612 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 613 *       access any non-constant socket information.
 614 */
 615static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
 616		     int uaddr_len)
 617{
 618	struct sock *sk = sock->sk;
 619	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 620	struct tipc_sock *tsk = tipc_sk(sk);
 621	int res = -EINVAL;
 622
 623	lock_sock(sk);
 624	if (unlikely(!uaddr_len)) {
 625		res = tipc_sk_withdraw(tsk, 0, NULL);
 626		goto exit;
 627	}
 628	if (tsk->group) {
 629		res = -EACCES;
 630		goto exit;
 631	}
 632	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
 633		res = -EINVAL;
 634		goto exit;
 635	}
 636	if (addr->family != AF_TIPC) {
 637		res = -EAFNOSUPPORT;
 638		goto exit;
 639	}
 640
 641	if (addr->addrtype == TIPC_ADDR_NAME)
 642		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
 643	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
 644		res = -EAFNOSUPPORT;
 645		goto exit;
 646	}
 647
 648	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
 649	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
 650	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
 651		res = -EACCES;
 652		goto exit;
 653	}
 654
 655	res = (addr->scope >= 0) ?
 656		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
 657		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
 658exit:
 659	release_sock(sk);
 660	return res;
 661}
 662
 663/**
 664 * tipc_getname - get port ID of socket or peer socket
 665 * @sock: socket structure
 666 * @uaddr: area for returned socket address
 667 * @uaddr_len: area for returned length of socket address
 668 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 669 *
 670 * Returns 0 on success, errno otherwise
 671 *
 672 * NOTE: This routine doesn't need to take the socket lock since it only
 673 *       accesses socket information that is unchanging (or which changes in
 674 *       a completely predictable manner).
 675 */
 676static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
 677			int peer)
 678{
 679	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 680	struct sock *sk = sock->sk;
 681	struct tipc_sock *tsk = tipc_sk(sk);
 
 682
 683	memset(addr, 0, sizeof(*addr));
 684	if (peer) {
 685		if ((!tipc_sk_connected(sk)) &&
 686		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
 687			return -ENOTCONN;
 688		addr->addr.id.ref = tsk_peer_port(tsk);
 689		addr->addr.id.node = tsk_peer_node(tsk);
 690	} else {
 691		addr->addr.id.ref = tsk->portid;
 692		addr->addr.id.node = tipc_own_addr(sock_net(sk));
 693	}
 694
 
 695	addr->addrtype = TIPC_ADDR_ID;
 696	addr->family = AF_TIPC;
 697	addr->scope = 0;
 698	addr->addr.name.domain = 0;
 699
 700	return sizeof(*addr);
 701}
 702
 703/**
 704 * tipc_poll - read and possibly block on pollmask
 705 * @file: file structure associated with the socket
 706 * @sock: socket for which to calculate the poll bits
 707 * @wait: ???
 708 *
 709 * Returns pollmask value
 710 *
 711 * COMMENTARY:
 712 * It appears that the usual socket locking mechanisms are not useful here
 713 * since the pollmask info is potentially out-of-date the moment this routine
 714 * exits.  TCP and other protocols seem to rely on higher level poll routines
 715 * to handle any preventable race conditions, so TIPC will do the same ...
 716 *
 717 * IMPORTANT: The fact that a read or write operation is indicated does NOT
 718 * imply that the operation will succeed, merely that it should be performed
 719 * and will not block.
 720 */
 721static __poll_t tipc_poll(struct file *file, struct socket *sock,
 722			      poll_table *wait)
 723{
 724	struct sock *sk = sock->sk;
 725	struct tipc_sock *tsk = tipc_sk(sk);
 726	__poll_t revents = 0;
 727
 728	sock_poll_wait(file, sock, wait);
 729	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
 730
 731	if (sk->sk_shutdown & RCV_SHUTDOWN)
 732		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
 733	if (sk->sk_shutdown == SHUTDOWN_MASK)
 734		revents |= EPOLLHUP;
 735
 736	switch (sk->sk_state) {
 737	case TIPC_ESTABLISHED:
 738		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
 739			revents |= EPOLLOUT;
 740		/* fall through */
 741	case TIPC_LISTEN:
 742	case TIPC_CONNECTING:
 743		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
 744			revents |= EPOLLIN | EPOLLRDNORM;
 745		break;
 746	case TIPC_OPEN:
 747		if (tsk->group_is_open && !tsk->cong_link_cnt)
 748			revents |= EPOLLOUT;
 749		if (!tipc_sk_type_connectionless(sk))
 750			break;
 751		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
 752			break;
 753		revents |= EPOLLIN | EPOLLRDNORM;
 754		break;
 755	case TIPC_DISCONNECTING:
 756		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
 757		break;
 758	}
 759	return revents;
 
 760}
 761
 762/**
 763 * tipc_sendmcast - send multicast message
 764 * @sock: socket structure
 765 * @seq: destination address
 766 * @msg: message to send
 767 * @dlen: length of data to send
 768 * @timeout: timeout to wait for wakeup
 769 *
 770 * Called from function tipc_sendmsg(), which has done all sanity checks
 771 * Returns the number of bytes sent on success, or errno
 772 */
 773static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
 774			  struct msghdr *msg, size_t dlen, long timeout)
 775{
 776	struct sock *sk = sock->sk;
 777	struct tipc_sock *tsk = tipc_sk(sk);
 778	struct tipc_msg *hdr = &tsk->phdr;
 779	struct net *net = sock_net(sk);
 780	int mtu = tipc_bcast_get_mtu(net);
 781	struct tipc_mc_method *method = &tsk->mc_method;
 782	struct sk_buff_head pkts;
 783	struct tipc_nlist dsts;
 784	int rc;
 785
 786	if (tsk->group)
 787		return -EACCES;
 788
 789	/* Block or return if any destination link is congested */
 790	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
 791	if (unlikely(rc))
 
 
 
 
 
 
 
 
 
 
 
 
 792		return rc;
 793
 794	/* Lookup destination nodes */
 795	tipc_nlist_init(&dsts, tipc_own_addr(net));
 796	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
 797				      seq->upper, &dsts);
 798	if (!dsts.local && !dsts.remote)
 799		return -EHOSTUNREACH;
 800
 801	/* Build message header */
 802	msg_set_type(hdr, TIPC_MCAST_MSG);
 803	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
 804	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
 805	msg_set_destport(hdr, 0);
 806	msg_set_destnode(hdr, 0);
 807	msg_set_nametype(hdr, seq->type);
 808	msg_set_namelower(hdr, seq->lower);
 809	msg_set_nameupper(hdr, seq->upper);
 810
 811	/* Build message as chain of buffers */
 812	__skb_queue_head_init(&pkts);
 813	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
 814
 815	/* Send message if build was successful */
 816	if (unlikely(rc == dlen)) {
 817		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
 818					TIPC_DUMP_SK_SNDQ, " ");
 819		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
 820				     &tsk->cong_link_cnt);
 821	}
 822
 823	tipc_nlist_purge(&dsts);
 824
 825	return rc ? rc : dlen;
 826}
 827
 828/**
 829 * tipc_send_group_msg - send a message to a member in the group
 830 * @net: network namespace
 831 * @m: message to send
 832 * @mb: group member
 833 * @dnode: destination node
 834 * @dport: destination port
 835 * @dlen: total length of message data
 836 */
 837static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
 838			       struct msghdr *m, struct tipc_member *mb,
 839			       u32 dnode, u32 dport, int dlen)
 840{
 841	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
 842	struct tipc_mc_method *method = &tsk->mc_method;
 843	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 844	struct tipc_msg *hdr = &tsk->phdr;
 845	struct sk_buff_head pkts;
 846	int mtu, rc;
 847
 848	/* Complete message header */
 849	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
 850	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
 851	msg_set_destport(hdr, dport);
 852	msg_set_destnode(hdr, dnode);
 853	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
 854
 855	/* Build message as chain of buffers */
 856	__skb_queue_head_init(&pkts);
 857	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
 858	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
 859	if (unlikely(rc != dlen))
 860		return rc;
 861
 862	/* Send message */
 863	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
 864	if (unlikely(rc == -ELINKCONG)) {
 865		tipc_dest_push(&tsk->cong_links, dnode, 0);
 866		tsk->cong_link_cnt++;
 867	}
 868
 869	/* Update send window */
 870	tipc_group_update_member(mb, blks);
 871
 872	/* A broadcast sent within next EXPIRE period must follow same path */
 873	method->rcast = true;
 874	method->mandatory = true;
 875	return dlen;
 876}
 877
 878/**
 879 * tipc_send_group_unicast - send message to a member in the group
 880 * @sock: socket structure
 881 * @m: message to send
 882 * @dlen: total length of message data
 883 * @timeout: timeout to wait for wakeup
 884 *
 885 * Called from function tipc_sendmsg(), which has done all sanity checks
 886 * Returns the number of bytes sent on success, or errno
 887 */
 888static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
 889				   int dlen, long timeout)
 890{
 891	struct sock *sk = sock->sk;
 892	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 893	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 894	struct tipc_sock *tsk = tipc_sk(sk);
 895	struct net *net = sock_net(sk);
 896	struct tipc_member *mb = NULL;
 897	u32 node, port;
 898	int rc;
 899
 900	node = dest->addr.id.node;
 901	port = dest->addr.id.ref;
 902	if (!port && !node)
 903		return -EHOSTUNREACH;
 904
 905	/* Block or return if destination link or member is congested */
 906	rc = tipc_wait_for_cond(sock, &timeout,
 907				!tipc_dest_find(&tsk->cong_links, node, 0) &&
 908				tsk->group &&
 909				!tipc_group_cong(tsk->group, node, port, blks,
 910						 &mb));
 911	if (unlikely(rc))
 912		return rc;
 913
 914	if (unlikely(!mb))
 915		return -EHOSTUNREACH;
 916
 917	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
 918
 919	return rc ? rc : dlen;
 920}
 921
 922/**
 923 * tipc_send_group_anycast - send message to any member with given identity
 924 * @sock: socket structure
 925 * @m: message to send
 926 * @dlen: total length of message data
 927 * @timeout: timeout to wait for wakeup
 928 *
 929 * Called from function tipc_sendmsg(), which has done all sanity checks
 930 * Returns the number of bytes sent on success, or errno
 931 */
 932static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
 933				   int dlen, long timeout)
 934{
 935	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 936	struct sock *sk = sock->sk;
 937	struct tipc_sock *tsk = tipc_sk(sk);
 938	struct list_head *cong_links = &tsk->cong_links;
 939	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 940	struct tipc_msg *hdr = &tsk->phdr;
 941	struct tipc_member *first = NULL;
 942	struct tipc_member *mbr = NULL;
 943	struct net *net = sock_net(sk);
 944	u32 node, port, exclude;
 945	struct list_head dsts;
 946	u32 type, inst, scope;
 947	int lookups = 0;
 948	int dstcnt, rc;
 949	bool cong;
 950
 951	INIT_LIST_HEAD(&dsts);
 952
 953	type = msg_nametype(hdr);
 954	inst = dest->addr.name.name.instance;
 955	scope = msg_lookup_scope(hdr);
 956
 957	while (++lookups < 4) {
 958		exclude = tipc_group_exclude(tsk->group);
 959
 960		first = NULL;
 961
 962		/* Look for a non-congested destination member, if any */
 963		while (1) {
 964			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
 965						 &dstcnt, exclude, false))
 966				return -EHOSTUNREACH;
 967			tipc_dest_pop(&dsts, &node, &port);
 968			cong = tipc_group_cong(tsk->group, node, port, blks,
 969					       &mbr);
 970			if (!cong)
 971				break;
 972			if (mbr == first)
 973				break;
 974			if (!first)
 975				first = mbr;
 976		}
 977
 978		/* Start over if destination was not in member list */
 979		if (unlikely(!mbr))
 980			continue;
 981
 982		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
 983			break;
 984
 985		/* Block or return if destination link or member is congested */
 986		rc = tipc_wait_for_cond(sock, &timeout,
 987					!tipc_dest_find(cong_links, node, 0) &&
 988					tsk->group &&
 989					!tipc_group_cong(tsk->group, node, port,
 990							 blks, &mbr));
 991		if (unlikely(rc))
 992			return rc;
 993
 994		/* Send, unless destination disappeared while waiting */
 995		if (likely(mbr))
 996			break;
 997	}
 998
 999	if (unlikely(lookups >= 4))
1000		return -EHOSTUNREACH;
1001
1002	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1003
1004	return rc ? rc : dlen;
1005}
1006
1007/**
1008 * tipc_send_group_bcast - send message to all members in communication group
1009 * @sk: socket structure
1010 * @m: message to send
1011 * @dlen: total length of message data
1012 * @timeout: timeout to wait for wakeup
1013 *
1014 * Called from function tipc_sendmsg(), which has done all sanity checks
1015 * Returns the number of bytes sent on success, or errno
1016 */
1017static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1018				 int dlen, long timeout)
1019{
1020	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1021	struct sock *sk = sock->sk;
1022	struct net *net = sock_net(sk);
1023	struct tipc_sock *tsk = tipc_sk(sk);
1024	struct tipc_nlist *dsts;
1025	struct tipc_mc_method *method = &tsk->mc_method;
1026	bool ack = method->mandatory && method->rcast;
1027	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1028	struct tipc_msg *hdr = &tsk->phdr;
1029	int mtu = tipc_bcast_get_mtu(net);
1030	struct sk_buff_head pkts;
1031	int rc = -EHOSTUNREACH;
1032
1033	/* Block or return if any destination link or member is congested */
1034	rc = tipc_wait_for_cond(sock, &timeout,
1035				!tsk->cong_link_cnt && tsk->group &&
1036				!tipc_group_bc_cong(tsk->group, blks));
1037	if (unlikely(rc))
1038		return rc;
1039
1040	dsts = tipc_group_dests(tsk->group);
1041	if (!dsts->local && !dsts->remote)
1042		return -EHOSTUNREACH;
1043
1044	/* Complete message header */
1045	if (dest) {
1046		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1047		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1048	} else {
1049		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1050		msg_set_nameinst(hdr, 0);
1051	}
1052	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1053	msg_set_destport(hdr, 0);
1054	msg_set_destnode(hdr, 0);
1055	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1056
1057	/* Avoid getting stuck with repeated forced replicasts */
1058	msg_set_grp_bc_ack_req(hdr, ack);
1059
1060	/* Build message as chain of buffers */
1061	__skb_queue_head_init(&pkts);
1062	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1063	if (unlikely(rc != dlen))
1064		return rc;
1065
1066	/* Send message */
1067	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1068	if (unlikely(rc))
1069		return rc;
1070
1071	/* Update broadcast sequence number and send windows */
1072	tipc_group_update_bc_members(tsk->group, blks, ack);
1073
1074	/* Broadcast link is now free to choose method for next broadcast */
1075	method->mandatory = false;
1076	method->expires = jiffies;
1077
1078	return dlen;
1079}
1080
1081/**
1082 * tipc_send_group_mcast - send message to all members with given identity
1083 * @sock: socket structure
1084 * @m: message to send
1085 * @dlen: total length of message data
1086 * @timeout: timeout to wait for wakeup
1087 *
1088 * Called from function tipc_sendmsg(), which has done all sanity checks
1089 * Returns the number of bytes sent on success, or errno
1090 */
1091static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1092				 int dlen, long timeout)
1093{
1094	struct sock *sk = sock->sk;
1095	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1096	struct tipc_sock *tsk = tipc_sk(sk);
1097	struct tipc_group *grp = tsk->group;
1098	struct tipc_msg *hdr = &tsk->phdr;
1099	struct net *net = sock_net(sk);
1100	u32 type, inst, scope, exclude;
1101	struct list_head dsts;
1102	u32 dstcnt;
1103
1104	INIT_LIST_HEAD(&dsts);
1105
1106	type = msg_nametype(hdr);
1107	inst = dest->addr.name.name.instance;
1108	scope = msg_lookup_scope(hdr);
1109	exclude = tipc_group_exclude(grp);
1110
1111	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1112				 &dstcnt, exclude, true))
1113		return -EHOSTUNREACH;
1114
1115	if (dstcnt == 1) {
1116		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1117		return tipc_send_group_unicast(sock, m, dlen, timeout);
1118	}
1119
1120	tipc_dest_list_purge(&dsts);
1121	return tipc_send_group_bcast(sock, m, dlen, timeout);
1122}
1123
1124/**
1125 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1126 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1127 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1128 *
1129 * Multi-threaded: parallel calls with reference to same queues may occur
1130 */
1131void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1132		       struct sk_buff_head *inputq)
1133{
1134	u32 self = tipc_own_addr(net);
1135	u32 type, lower, upper, scope;
 
 
 
 
1136	struct sk_buff *skb, *_skb;
1137	u32 portid, onode;
1138	struct sk_buff_head tmpq;
1139	struct list_head dports;
1140	struct tipc_msg *hdr;
1141	int user, mtyp, hlen;
1142	bool exact;
1143
1144	__skb_queue_head_init(&tmpq);
1145	INIT_LIST_HEAD(&dports);
1146
1147	skb = tipc_skb_peek(arrvq, &inputq->lock);
1148	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1149		hdr = buf_msg(skb);
1150		user = msg_user(hdr);
1151		mtyp = msg_type(hdr);
1152		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1153		onode = msg_orignode(hdr);
1154		type = msg_nametype(hdr);
1155
1156		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1157			spin_lock_bh(&inputq->lock);
1158			if (skb_peek(arrvq) == skb) {
1159				__skb_dequeue(arrvq);
1160				__skb_queue_tail(inputq, skb);
1161			}
1162			kfree_skb(skb);
1163			spin_unlock_bh(&inputq->lock);
1164			continue;
1165		}
1166
1167		/* Group messages require exact scope match */
1168		if (msg_in_group(hdr)) {
1169			lower = 0;
1170			upper = ~0;
1171			scope = msg_lookup_scope(hdr);
1172			exact = true;
1173		} else {
1174			/* TIPC_NODE_SCOPE means "any scope" in this context */
1175			if (onode == self)
1176				scope = TIPC_NODE_SCOPE;
1177			else
1178				scope = TIPC_CLUSTER_SCOPE;
1179			exact = false;
1180			lower = msg_namelower(hdr);
1181			upper = msg_nameupper(hdr);
1182		}
1183
1184		/* Create destination port list: */
1185		tipc_nametbl_mc_lookup(net, type, lower, upper,
1186				       scope, exact, &dports);
1187
1188		/* Clone message per destination */
1189		while (tipc_dest_pop(&dports, NULL, &portid)) {
1190			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1191			if (_skb) {
1192				msg_set_destport(buf_msg(_skb), portid);
1193				__skb_queue_tail(&tmpq, _skb);
1194				continue;
1195			}
1196			pr_warn("Failed to clone mcast rcv buffer\n");
1197		}
1198		/* Append to inputq if not already done by other thread */
1199		spin_lock_bh(&inputq->lock);
1200		if (skb_peek(arrvq) == skb) {
1201			skb_queue_splice_tail_init(&tmpq, inputq);
1202			kfree_skb(__skb_dequeue(arrvq));
1203		}
1204		spin_unlock_bh(&inputq->lock);
1205		__skb_queue_purge(&tmpq);
1206		kfree_skb(skb);
1207	}
1208	tipc_sk_rcv(net, inputq);
1209}
1210
1211/**
1212 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1213 * @tsk: receiving socket
1214 * @skb: pointer to message buffer.
1215 */
1216static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1217				   struct sk_buff_head *inputq,
1218				   struct sk_buff_head *xmitq)
1219{
 
 
1220	struct tipc_msg *hdr = buf_msg(skb);
1221	u32 onode = tsk_own_node(tsk);
1222	struct sock *sk = &tsk->sk;
1223	int mtyp = msg_type(hdr);
1224	bool conn_cong;
1225
1226	/* Ignore if connection cannot be validated: */
1227	if (!tsk_peer_msg(tsk, hdr)) {
1228		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1229		goto exit;
1230	}
1231
1232	if (unlikely(msg_errcode(hdr))) {
1233		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1234		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1235				      tsk_peer_port(tsk));
1236		sk->sk_state_change(sk);
1237
1238		/* State change is ignored if socket already awake,
1239		 * - convert msg to abort msg and add to inqueue
1240		 */
1241		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1242		msg_set_type(hdr, TIPC_CONN_MSG);
1243		msg_set_size(hdr, BASIC_H_SIZE);
1244		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1245		__skb_queue_tail(inputq, skb);
1246		return;
1247	}
1248
1249	tsk->probe_unacked = false;
1250
1251	if (mtyp == CONN_PROBE) {
1252		msg_set_type(hdr, CONN_PROBE_REPLY);
1253		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1254			__skb_queue_tail(xmitq, skb);
1255		return;
1256	} else if (mtyp == CONN_ACK) {
1257		conn_cong = tsk_conn_cong(tsk);
1258		tsk->snt_unacked -= msg_conn_ack(hdr);
1259		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1260			tsk->snd_win = msg_adv_win(hdr);
1261		if (conn_cong)
1262			sk->sk_write_space(sk);
1263	} else if (mtyp != CONN_PROBE_REPLY) {
1264		pr_warn("Received unknown CONN_PROTO msg\n");
1265	}
1266exit:
1267	kfree_skb(skb);
1268}
1269
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1270/**
1271 * tipc_sendmsg - send message in connectionless manner
1272 * @sock: socket structure
1273 * @m: message to send
1274 * @dsz: amount of user data to be sent
1275 *
1276 * Message must have an destination specified explicitly.
1277 * Used for SOCK_RDM and SOCK_DGRAM messages,
1278 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1279 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1280 *
1281 * Returns the number of bytes sent on success, or errno otherwise
1282 */
1283static int tipc_sendmsg(struct socket *sock,
1284			struct msghdr *m, size_t dsz)
1285{
1286	struct sock *sk = sock->sk;
1287	int ret;
1288
1289	lock_sock(sk);
1290	ret = __tipc_sendmsg(sock, m, dsz);
1291	release_sock(sk);
1292
1293	return ret;
1294}
1295
1296static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1297{
 
1298	struct sock *sk = sock->sk;
 
1299	struct net *net = sock_net(sk);
1300	struct tipc_sock *tsk = tipc_sk(sk);
1301	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1302	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1303	struct list_head *clinks = &tsk->cong_links;
1304	bool syn = !tipc_sk_type_connectionless(sk);
1305	struct tipc_group *grp = tsk->group;
1306	struct tipc_msg *hdr = &tsk->phdr;
1307	struct tipc_name_seq *seq;
1308	struct sk_buff_head pkts;
1309	u32 dport, dnode = 0;
1310	u32 type, inst;
1311	int mtu, rc;
1312
1313	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1314		return -EMSGSIZE;
1315
1316	if (likely(dest)) {
1317		if (unlikely(m->msg_namelen < sizeof(*dest)))
1318			return -EINVAL;
1319		if (unlikely(dest->family != AF_TIPC))
1320			return -EINVAL;
1321	}
1322
1323	if (grp) {
1324		if (!dest)
1325			return tipc_send_group_bcast(sock, m, dlen, timeout);
1326		if (dest->addrtype == TIPC_ADDR_NAME)
1327			return tipc_send_group_anycast(sock, m, dlen, timeout);
1328		if (dest->addrtype == TIPC_ADDR_ID)
1329			return tipc_send_group_unicast(sock, m, dlen, timeout);
1330		if (dest->addrtype == TIPC_ADDR_MCAST)
1331			return tipc_send_group_mcast(sock, m, dlen, timeout);
1332		return -EINVAL;
1333	}
1334
1335	if (unlikely(!dest)) {
1336		dest = &tsk->peer;
1337		if (!syn && dest->family != AF_TIPC)
 
1338			return -EDESTADDRREQ;
 
 
 
1339	}
1340
1341	if (unlikely(syn)) {
1342		if (sk->sk_state == TIPC_LISTEN)
1343			return -EPIPE;
1344		if (sk->sk_state != TIPC_OPEN)
1345			return -EISCONN;
1346		if (tsk->published)
1347			return -EOPNOTSUPP;
1348		if (dest->addrtype == TIPC_ADDR_NAME) {
1349			tsk->conn_type = dest->addr.name.name.type;
1350			tsk->conn_instance = dest->addr.name.name.instance;
1351		}
1352		msg_set_syn(hdr, 1);
1353	}
1354
1355	seq = &dest->addr.nameseq;
1356	if (dest->addrtype == TIPC_ADDR_MCAST)
1357		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1358
1359	if (dest->addrtype == TIPC_ADDR_NAME) {
1360		type = dest->addr.name.name.type;
1361		inst = dest->addr.name.name.instance;
1362		dnode = dest->addr.name.domain;
1363		msg_set_type(hdr, TIPC_NAMED_MSG);
1364		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1365		msg_set_nametype(hdr, type);
1366		msg_set_nameinst(hdr, inst);
1367		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
 
 
 
 
1368		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1369		msg_set_destnode(hdr, dnode);
1370		msg_set_destport(hdr, dport);
1371		if (unlikely(!dport && !dnode))
1372			return -EHOSTUNREACH;
1373	} else if (dest->addrtype == TIPC_ADDR_ID) {
1374		dnode = dest->addr.id.node;
1375		msg_set_type(hdr, TIPC_DIRECT_MSG);
1376		msg_set_lookup_scope(hdr, 0);
1377		msg_set_destnode(hdr, dnode);
1378		msg_set_destport(hdr, dest->addr.id.ref);
1379		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1380	} else {
1381		return -EINVAL;
1382	}
1383
1384	/* Block or return if destination link is congested */
1385	rc = tipc_wait_for_cond(sock, &timeout,
1386				!tipc_dest_find(clinks, dnode, 0));
1387	if (unlikely(rc))
 
 
1388		return rc;
1389
1390	__skb_queue_head_init(&pkts);
1391	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1392	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1393	if (unlikely(rc != dlen))
1394		return rc;
1395	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1396		return -ENOMEM;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1397
1398	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1399	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1400	if (unlikely(rc == -ELINKCONG)) {
1401		tipc_dest_push(clinks, dnode, 0);
1402		tsk->cong_link_cnt++;
1403		rc = 0;
1404	}
1405
1406	if (unlikely(syn && !rc))
1407		tipc_set_sk_state(sk, TIPC_CONNECTING);
 
 
 
 
 
 
 
 
 
 
1408
1409	return rc ? rc : dlen;
 
 
 
 
 
 
 
1410}
1411
1412/**
1413 * tipc_sendstream - send stream-oriented data
1414 * @sock: socket structure
1415 * @m: data to send
1416 * @dsz: total length of data to be transmitted
1417 *
1418 * Used for SOCK_STREAM data.
1419 *
1420 * Returns the number of bytes sent on success (or partial success),
1421 * or errno if no data sent
1422 */
1423static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1424{
1425	struct sock *sk = sock->sk;
1426	int ret;
1427
1428	lock_sock(sk);
1429	ret = __tipc_sendstream(sock, m, dsz);
1430	release_sock(sk);
1431
1432	return ret;
1433}
1434
1435static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1436{
1437	struct sock *sk = sock->sk;
 
 
 
 
1438	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1439	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1440	struct tipc_sock *tsk = tipc_sk(sk);
1441	struct tipc_msg *hdr = &tsk->phdr;
1442	struct net *net = sock_net(sk);
1443	struct sk_buff_head pkts;
1444	u32 dnode = tsk_peer_node(tsk);
1445	int send, sent = 0;
1446	int rc = 0;
1447
1448	__skb_queue_head_init(&pkts);
 
 
 
 
 
 
 
 
 
1449
1450	if (unlikely(dlen > INT_MAX))
1451		return -EMSGSIZE;
 
 
 
 
1452
1453	/* Handle implicit connection setup */
1454	if (unlikely(dest)) {
1455		rc = __tipc_sendmsg(sock, m, dlen);
1456		if (dlen && dlen == rc) {
1457			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1458			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1459		}
 
 
 
 
 
 
1460		return rc;
1461	}
1462
1463	do {
1464		rc = tipc_wait_for_cond(sock, &timeout,
1465					(!tsk->cong_link_cnt &&
1466					 !tsk_conn_cong(tsk) &&
1467					 tipc_sk_connected(sk)));
1468		if (unlikely(rc))
1469			break;
 
 
 
 
 
 
 
 
 
 
 
 
1470
1471		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1472		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1473		if (unlikely(rc != send))
1474			break;
1475
1476		trace_tipc_sk_sendstream(sk, skb_peek(&pkts),
1477					 TIPC_DUMP_SK_SNDQ, " ");
1478		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1479		if (unlikely(rc == -ELINKCONG)) {
1480			tsk->cong_link_cnt = 1;
1481			rc = 0;
1482		}
1483		if (likely(!rc)) {
1484			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1485			sent += send;
1486		}
1487	} while (sent < dlen && !rc);
1488
 
1489	return sent ? sent : rc;
1490}
1491
1492/**
1493 * tipc_send_packet - send a connection-oriented message
1494 * @sock: socket structure
1495 * @m: message to send
1496 * @dsz: length of data to be transmitted
1497 *
1498 * Used for SOCK_SEQPACKET messages.
1499 *
1500 * Returns the number of bytes sent on success, or errno otherwise
1501 */
1502static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1503{
1504	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1505		return -EMSGSIZE;
1506
1507	return tipc_sendstream(sock, m, dsz);
1508}
1509
1510/* tipc_sk_finish_conn - complete the setup of a connection
1511 */
1512static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1513				u32 peer_node)
1514{
1515	struct sock *sk = &tsk->sk;
1516	struct net *net = sock_net(sk);
1517	struct tipc_msg *msg = &tsk->phdr;
1518
1519	msg_set_syn(msg, 0);
1520	msg_set_destnode(msg, peer_node);
1521	msg_set_destport(msg, peer_port);
1522	msg_set_type(msg, TIPC_CONN_MSG);
1523	msg_set_lookup_scope(msg, 0);
1524	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1525
1526	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1527	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1528	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1529	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1530	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1531	__skb_queue_purge(&sk->sk_write_queue);
1532	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1533		return;
1534
1535	/* Fall back to message based flow control */
1536	tsk->rcv_win = FLOWCTL_MSG_WIN;
1537	tsk->snd_win = FLOWCTL_MSG_WIN;
1538}
1539
1540/**
1541 * tipc_sk_set_orig_addr - capture sender's address for received message
1542 * @m: descriptor for message info
1543 * @hdr: received message header
1544 *
1545 * Note: Address is not captured if not requested by receiver.
1546 */
1547static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1548{
1549	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1550	struct tipc_msg *hdr = buf_msg(skb);
1551
1552	if (!srcaddr)
1553		return;
1554
1555	srcaddr->sock.family = AF_TIPC;
1556	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1557	srcaddr->sock.scope = 0;
1558	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1559	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1560	srcaddr->sock.addr.name.domain = 0;
1561	m->msg_namelen = sizeof(struct sockaddr_tipc);
1562
1563	if (!msg_in_group(hdr))
1564		return;
1565
1566	/* Group message users may also want to know sending member's id */
1567	srcaddr->member.family = AF_TIPC;
1568	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1569	srcaddr->member.scope = 0;
1570	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1571	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1572	srcaddr->member.addr.name.domain = 0;
1573	m->msg_namelen = sizeof(*srcaddr);
1574}
1575
1576/**
1577 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1578 * @m: descriptor for message info
1579 * @skb: received message buffer
1580 * @tsk: TIPC port associated with message
1581 *
1582 * Note: Ancillary data is not captured if not requested by receiver.
1583 *
1584 * Returns 0 if successful, otherwise errno
1585 */
1586static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1587				 struct tipc_sock *tsk)
1588{
1589	struct tipc_msg *msg;
1590	u32 anc_data[3];
1591	u32 err;
1592	u32 dest_type;
1593	int has_name;
1594	int res;
1595
1596	if (likely(m->msg_controllen == 0))
1597		return 0;
1598	msg = buf_msg(skb);
1599
1600	/* Optionally capture errored message object(s) */
1601	err = msg ? msg_errcode(msg) : 0;
1602	if (unlikely(err)) {
1603		anc_data[0] = err;
1604		anc_data[1] = msg_data_sz(msg);
1605		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1606		if (res)
1607			return res;
1608		if (anc_data[1]) {
1609			if (skb_linearize(skb))
1610				return -ENOMEM;
1611			msg = buf_msg(skb);
1612			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1613				       msg_data(msg));
1614			if (res)
1615				return res;
1616		}
1617	}
1618
1619	/* Optionally capture message destination object */
1620	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1621	switch (dest_type) {
1622	case TIPC_NAMED_MSG:
1623		has_name = 1;
1624		anc_data[0] = msg_nametype(msg);
1625		anc_data[1] = msg_namelower(msg);
1626		anc_data[2] = msg_namelower(msg);
1627		break;
1628	case TIPC_MCAST_MSG:
1629		has_name = 1;
1630		anc_data[0] = msg_nametype(msg);
1631		anc_data[1] = msg_namelower(msg);
1632		anc_data[2] = msg_nameupper(msg);
1633		break;
1634	case TIPC_CONN_MSG:
1635		has_name = (tsk->conn_type != 0);
1636		anc_data[0] = tsk->conn_type;
1637		anc_data[1] = tsk->conn_instance;
1638		anc_data[2] = tsk->conn_instance;
1639		break;
1640	default:
1641		has_name = 0;
1642	}
1643	if (has_name) {
1644		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1645		if (res)
1646			return res;
1647	}
1648
1649	return 0;
1650}
1651
1652static void tipc_sk_send_ack(struct tipc_sock *tsk)
1653{
1654	struct sock *sk = &tsk->sk;
1655	struct net *net = sock_net(sk);
1656	struct sk_buff *skb = NULL;
1657	struct tipc_msg *msg;
1658	u32 peer_port = tsk_peer_port(tsk);
1659	u32 dnode = tsk_peer_node(tsk);
1660
1661	if (!tipc_sk_connected(sk))
1662		return;
1663	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1664			      dnode, tsk_own_node(tsk), peer_port,
1665			      tsk->portid, TIPC_OK);
1666	if (!skb)
1667		return;
1668	msg = buf_msg(skb);
1669	msg_set_conn_ack(msg, tsk->rcv_unacked);
1670	tsk->rcv_unacked = 0;
1671
1672	/* Adjust to and advertize the correct window limit */
1673	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1674		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1675		msg_set_adv_win(msg, tsk->rcv_win);
1676	}
1677	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1678}
1679
1680static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1681{
1682	struct sock *sk = sock->sk;
1683	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1684	long timeo = *timeop;
1685	int err = sock_error(sk);
1686
1687	if (err)
1688		return err;
1689
1690	for (;;) {
 
1691		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1692			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1693				err = -ENOTCONN;
1694				break;
1695			}
1696			add_wait_queue(sk_sleep(sk), &wait);
1697			release_sock(sk);
1698			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1699			sched_annotate_sleep();
1700			lock_sock(sk);
1701			remove_wait_queue(sk_sleep(sk), &wait);
1702		}
1703		err = 0;
1704		if (!skb_queue_empty(&sk->sk_receive_queue))
1705			break;
1706		err = -EAGAIN;
1707		if (!timeo)
1708			break;
1709		err = sock_intr_errno(timeo);
1710		if (signal_pending(current))
1711			break;
1712
1713		err = sock_error(sk);
1714		if (err)
1715			break;
1716	}
 
1717	*timeop = timeo;
1718	return err;
1719}
1720
1721/**
1722 * tipc_recvmsg - receive packet-oriented message
1723 * @m: descriptor for message info
1724 * @buflen: length of user buffer area
1725 * @flags: receive flags
1726 *
1727 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1728 * If the complete message doesn't fit in user area, truncate it.
1729 *
1730 * Returns size of returned message data, errno otherwise
1731 */
1732static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1733			size_t buflen,	int flags)
1734{
1735	struct sock *sk = sock->sk;
1736	bool connected = !tipc_sk_type_connectionless(sk);
1737	struct tipc_sock *tsk = tipc_sk(sk);
1738	int rc, err, hlen, dlen, copy;
1739	struct sk_buff_head xmitq;
1740	struct tipc_msg *hdr;
1741	struct sk_buff *skb;
1742	bool grp_evt;
1743	long timeout;
 
1744
1745	/* Catch invalid receive requests */
1746	if (unlikely(!buflen))
1747		return -EINVAL;
1748
1749	lock_sock(sk);
1750	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1751		rc = -ENOTCONN;
 
1752		goto exit;
1753	}
1754	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1755
1756	/* Step rcv queue to first msg with data or error; wait if necessary */
1757	do {
1758		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1759		if (unlikely(rc))
1760			goto exit;
1761		skb = skb_peek(&sk->sk_receive_queue);
1762		hdr = buf_msg(skb);
1763		dlen = msg_data_sz(hdr);
1764		hlen = msg_hdr_sz(hdr);
1765		err = msg_errcode(hdr);
1766		grp_evt = msg_is_grp_evt(hdr);
1767		if (likely(dlen || err))
1768			break;
 
 
 
 
1769		tsk_advance_rx_queue(sk);
1770	} while (1);
 
 
 
 
1771
1772	/* Collect msg meta data, including error code and rejected data */
1773	tipc_sk_set_orig_addr(m, skb);
1774	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1775	if (unlikely(rc))
1776		goto exit;
1777	hdr = buf_msg(skb);
1778
1779	/* Capture data if non-error msg, otherwise just set return value */
1780	if (likely(!err)) {
1781		copy = min_t(int, dlen, buflen);
1782		if (unlikely(copy != dlen))
1783			m->msg_flags |= MSG_TRUNC;
1784		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
 
 
 
 
1785	} else {
1786		copy = 0;
1787		rc = 0;
1788		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1789			rc = -ECONNRESET;
 
1790	}
1791	if (unlikely(rc))
1792		goto exit;
1793
1794	/* Mark message as group event if applicable */
1795	if (unlikely(grp_evt)) {
1796		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1797			m->msg_flags |= MSG_EOR;
1798		m->msg_flags |= MSG_OOB;
1799		copy = 0;
1800	}
1801
1802	/* Caption of data or error code/rejected data was successful */
1803	if (unlikely(flags & MSG_PEEK))
1804		goto exit;
1805
1806	/* Send group flow control advertisement when applicable */
1807	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1808		__skb_queue_head_init(&xmitq);
1809		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1810					  msg_orignode(hdr), msg_origport(hdr),
1811					  &xmitq);
1812		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1813	}
1814
1815	tsk_advance_rx_queue(sk);
1816
1817	if (likely(!connected))
1818		goto exit;
1819
1820	/* Send connection flow control advertisement when applicable */
1821	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1822	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1823		tipc_sk_send_ack(tsk);
1824exit:
1825	release_sock(sk);
1826	return rc ? rc : copy;
1827}
1828
1829/**
1830 * tipc_recvstream - receive stream-oriented data
1831 * @m: descriptor for message info
1832 * @buflen: total size of user buffer area
1833 * @flags: receive flags
1834 *
1835 * Used for SOCK_STREAM messages only.  If not enough data is available
1836 * will optionally wait for more; never truncates data.
1837 *
1838 * Returns size of returned message data, errno otherwise
1839 */
1840static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1841			   size_t buflen, int flags)
1842{
1843	struct sock *sk = sock->sk;
1844	struct tipc_sock *tsk = tipc_sk(sk);
1845	struct sk_buff *skb;
1846	struct tipc_msg *hdr;
1847	struct tipc_skb_cb *skb_cb;
1848	bool peek = flags & MSG_PEEK;
1849	int offset, required, copy, copied = 0;
1850	int hlen, dlen, err, rc;
1851	long timeout;
 
1852
1853	/* Catch invalid receive attempts */
1854	if (unlikely(!buflen))
1855		return -EINVAL;
1856
1857	lock_sock(sk);
1858
1859	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1860		rc = -ENOTCONN;
1861		goto exit;
1862	}
1863	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1864	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1865
1866	do {
1867		/* Look at first msg in receive queue; wait if necessary */
1868		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1869		if (unlikely(rc))
1870			break;
1871		skb = skb_peek(&sk->sk_receive_queue);
1872		skb_cb = TIPC_SKB_CB(skb);
1873		hdr = buf_msg(skb);
1874		dlen = msg_data_sz(hdr);
1875		hlen = msg_hdr_sz(hdr);
1876		err = msg_errcode(hdr);
1877
1878		/* Discard any empty non-errored (SYN-) message */
1879		if (unlikely(!dlen && !err)) {
1880			tsk_advance_rx_queue(sk);
1881			continue;
1882		}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1883
1884		/* Collect msg meta data, incl. error code and rejected data */
1885		if (!copied) {
1886			tipc_sk_set_orig_addr(m, skb);
1887			rc = tipc_sk_anc_data_recv(m, skb, tsk);
1888			if (rc)
1889				break;
1890			hdr = buf_msg(skb);
1891		}
1892
1893		/* Copy data if msg ok, otherwise return error/partial data */
1894		if (likely(!err)) {
1895			offset = skb_cb->bytes_read;
1896			copy = min_t(int, dlen - offset, buflen - copied);
1897			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1898			if (unlikely(rc))
1899				break;
1900			copied += copy;
1901			offset += copy;
1902			if (unlikely(offset < dlen)) {
1903				if (!peek)
1904					skb_cb->bytes_read = offset;
1905				break;
1906			}
1907		} else {
1908			rc = 0;
1909			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1910				rc = -ECONNRESET;
1911			if (copied || rc)
1912				break;
1913		}
 
 
 
1914
1915		if (unlikely(peek))
1916			break;
 
 
 
1917
1918		tsk_advance_rx_queue(sk);
 
1919
1920		/* Send connection flow control advertisement when applicable */
1921		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1922		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1923			tipc_sk_send_ack(tsk);
1924
1925		/* Exit if all requested data or FIN/error received */
1926		if (copied == buflen || err)
1927			break;
 
 
 
1928
1929	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1930exit:
1931	release_sock(sk);
1932	return copied ? copied : rc;
1933}
1934
1935/**
1936 * tipc_write_space - wake up thread if port congestion is released
1937 * @sk: socket
1938 */
1939static void tipc_write_space(struct sock *sk)
1940{
1941	struct socket_wq *wq;
1942
1943	rcu_read_lock();
1944	wq = rcu_dereference(sk->sk_wq);
1945	if (skwq_has_sleeper(wq))
1946		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1947						EPOLLWRNORM | EPOLLWRBAND);
1948	rcu_read_unlock();
1949}
1950
1951/**
1952 * tipc_data_ready - wake up threads to indicate messages have been received
1953 * @sk: socket
1954 * @len: the length of messages
1955 */
1956static void tipc_data_ready(struct sock *sk)
1957{
1958	struct socket_wq *wq;
1959
1960	rcu_read_lock();
1961	wq = rcu_dereference(sk->sk_wq);
1962	if (skwq_has_sleeper(wq))
1963		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1964						EPOLLRDNORM | EPOLLRDBAND);
1965	rcu_read_unlock();
1966}
1967
1968static void tipc_sock_destruct(struct sock *sk)
1969{
1970	__skb_queue_purge(&sk->sk_receive_queue);
1971}
1972
1973static void tipc_sk_proto_rcv(struct sock *sk,
1974			      struct sk_buff_head *inputq,
1975			      struct sk_buff_head *xmitq)
1976{
1977	struct sk_buff *skb = __skb_dequeue(inputq);
1978	struct tipc_sock *tsk = tipc_sk(sk);
1979	struct tipc_msg *hdr = buf_msg(skb);
1980	struct tipc_group *grp = tsk->group;
1981	bool wakeup = false;
1982
1983	switch (msg_user(hdr)) {
1984	case CONN_MANAGER:
1985		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1986		return;
1987	case SOCK_WAKEUP:
1988		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1989		/* coupled with smp_rmb() in tipc_wait_for_cond() */
1990		smp_wmb();
1991		tsk->cong_link_cnt--;
1992		wakeup = true;
1993		break;
1994	case GROUP_PROTOCOL:
1995		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1996		break;
1997	case TOP_SRV:
1998		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1999				      hdr, inputq, xmitq);
2000		break;
2001	default:
2002		break;
2003	}
2004
2005	if (wakeup)
2006		sk->sk_write_space(sk);
2007
2008	kfree_skb(skb);
2009}
2010
2011/**
2012 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2013 * @tsk: TIPC socket
2014 * @skb: pointer to message buffer.
2015 * Returns true if message should be added to receive queue, false otherwise
 
2016 */
2017static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2018{
2019	struct sock *sk = &tsk->sk;
2020	struct net *net = sock_net(sk);
2021	struct tipc_msg *hdr = buf_msg(skb);
2022	bool con_msg = msg_connected(hdr);
2023	u32 pport = tsk_peer_port(tsk);
2024	u32 pnode = tsk_peer_node(tsk);
2025	u32 oport = msg_origport(hdr);
2026	u32 onode = msg_orignode(hdr);
2027	int err = msg_errcode(hdr);
2028	unsigned long delay;
2029
2030	if (unlikely(msg_mcast(hdr)))
2031		return false;
2032
2033	switch (sk->sk_state) {
2034	case TIPC_CONNECTING:
2035		/* Setup ACK */
2036		if (likely(con_msg)) {
2037			if (err)
2038				break;
2039			tipc_sk_finish_conn(tsk, oport, onode);
2040			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2041			/* ACK+ message with data is added to receive queue */
2042			if (msg_data_sz(hdr))
2043				return true;
2044			/* Empty ACK-, - wake up sleeping connect() and drop */
2045			sk->sk_state_change(sk);
2046			msg_set_dest_droppable(hdr, 1);
2047			return false;
 
 
 
 
 
2048		}
2049		/* Ignore connectionless message if not from listening socket */
2050		if (oport != pport || onode != pnode)
2051			return false;
2052
2053		/* Rejected SYN */
2054		if (err != TIPC_ERR_OVERLOAD)
2055			break;
 
 
 
 
 
 
 
 
 
 
 
 
 
2056
2057		/* Prepare for new setup attempt if we have a SYN clone */
2058		if (skb_queue_empty(&sk->sk_write_queue))
2059			break;
2060		get_random_bytes(&delay, 2);
2061		delay %= (tsk->conn_timeout / 4);
2062		delay = msecs_to_jiffies(delay + 100);
2063		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2064		return false;
 
2065	case TIPC_OPEN:
2066	case TIPC_DISCONNECTING:
2067		return false;
2068	case TIPC_LISTEN:
2069		/* Accept only SYN message */
2070		if (!msg_is_syn(hdr) &&
2071		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2072			return false;
2073		if (!con_msg && !err)
2074			return true;
2075		return false;
2076	case TIPC_ESTABLISHED:
2077		/* Accept only connection-based messages sent by peer */
2078		if (likely(con_msg && !err && pport == oport && pnode == onode))
2079			return true;
2080		if (!tsk_peer_msg(tsk, hdr))
2081			return false;
2082		if (!err)
2083			return true;
2084		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2085		tipc_node_remove_conn(net, pnode, tsk->portid);
2086		sk->sk_state_change(sk);
 
 
 
2087		return true;
2088	default:
2089		pr_err("Unknown sk_state %u\n", sk->sk_state);
2090	}
2091	/* Abort connection setup attempt */
2092	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2093	sk->sk_err = ECONNREFUSED;
2094	sk->sk_state_change(sk);
2095	return true;
2096}
2097
2098/**
2099 * rcvbuf_limit - get proper overload limit of socket receive queue
2100 * @sk: socket
2101 * @skb: message
2102 *
2103 * For connection oriented messages, irrespective of importance,
2104 * default queue limit is 2 MB.
2105 *
2106 * For connectionless messages, queue limits are based on message
2107 * importance as follows:
2108 *
2109 * TIPC_LOW_IMPORTANCE       (2 MB)
2110 * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2111 * TIPC_HIGH_IMPORTANCE      (8 MB)
2112 * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2113 *
2114 * Returns overload limit according to corresponding message importance
2115 */
2116static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2117{
2118	struct tipc_sock *tsk = tipc_sk(sk);
2119	struct tipc_msg *hdr = buf_msg(skb);
2120
2121	if (unlikely(msg_in_group(hdr)))
2122		return READ_ONCE(sk->sk_rcvbuf);
2123
2124	if (unlikely(!msg_connected(hdr)))
2125		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2126
2127	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2128		return READ_ONCE(sk->sk_rcvbuf);
2129
2130	return FLOWCTL_MSG_LIM;
2131}
2132
2133/**
2134 * tipc_sk_filter_rcv - validate incoming message
2135 * @sk: socket
2136 * @skb: pointer to message.
2137 *
2138 * Enqueues message on receive queue if acceptable; optionally handles
2139 * disconnect indication for a connected socket.
2140 *
2141 * Called with socket lock already taken
2142 *
 
2143 */
2144static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2145			       struct sk_buff_head *xmitq)
2146{
2147	bool sk_conn = !tipc_sk_type_connectionless(sk);
2148	struct tipc_sock *tsk = tipc_sk(sk);
2149	struct tipc_group *grp = tsk->group;
2150	struct tipc_msg *hdr = buf_msg(skb);
2151	struct net *net = sock_net(sk);
2152	struct sk_buff_head inputq;
2153	int mtyp = msg_type(hdr);
2154	int limit, err = TIPC_OK;
2155
2156	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2157	TIPC_SKB_CB(skb)->bytes_read = 0;
2158	__skb_queue_head_init(&inputq);
2159	__skb_queue_tail(&inputq, skb);
2160
2161	if (unlikely(!msg_isdata(hdr)))
2162		tipc_sk_proto_rcv(sk, &inputq, xmitq);
 
 
 
 
2163
2164	if (unlikely(grp))
2165		tipc_group_filter_msg(grp, &inputq, xmitq);
 
 
 
2166
2167	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2168		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2169
2170	/* Validate and add to receive buffer if there is space */
2171	while ((skb = __skb_dequeue(&inputq))) {
2172		hdr = buf_msg(skb);
2173		limit = rcvbuf_limit(sk, skb);
2174		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2175		    (!sk_conn && msg_connected(hdr)) ||
2176		    (!grp && msg_in_group(hdr)))
2177			err = TIPC_ERR_NO_PORT;
2178		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2179			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2180					   "err_overload2!");
2181			atomic_inc(&sk->sk_drops);
2182			err = TIPC_ERR_OVERLOAD;
2183		}
 
 
 
 
2184
2185		if (unlikely(err)) {
2186			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2187				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2188						      "@filter_rcv!");
2189				__skb_queue_tail(xmitq, skb);
2190			}
2191			err = TIPC_OK;
2192			continue;
2193		}
2194		__skb_queue_tail(&sk->sk_receive_queue, skb);
2195		skb_set_owner_r(skb, sk);
2196		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2197					 "rcvq >90% allocated!");
2198		sk->sk_data_ready(sk);
2199	}
 
 
 
 
 
 
 
 
 
 
 
 
 
2200}
2201
2202/**
2203 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2204 * @sk: socket
2205 * @skb: message
2206 *
2207 * Caller must hold socket lock
 
 
2208 */
2209static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2210{
2211	unsigned int before = sk_rmem_alloc_get(sk);
2212	struct sk_buff_head xmitq;
2213	unsigned int added;
2214
2215	__skb_queue_head_init(&xmitq);
2216
2217	tipc_sk_filter_rcv(sk, skb, &xmitq);
2218	added = sk_rmem_alloc_get(sk) - before;
2219	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
 
2220
2221	/* Send pending response/rejected messages, if any */
2222	tipc_node_distr_xmit(sock_net(sk), &xmitq);
 
 
 
 
 
 
2223	return 0;
2224}
2225
2226/**
2227 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2228 *                   inputq and try adding them to socket or backlog queue
2229 * @inputq: list of incoming buffers with potentially different destinations
2230 * @sk: socket where the buffers should be enqueued
2231 * @dport: port number for the socket
2232 *
2233 * Caller must hold socket lock
2234 */
2235static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2236			    u32 dport, struct sk_buff_head *xmitq)
2237{
2238	unsigned long time_limit = jiffies + 2;
2239	struct sk_buff *skb;
2240	unsigned int lim;
2241	atomic_t *dcnt;
2242	u32 onode;
2243
2244	while (skb_queue_len(inputq)) {
2245		if (unlikely(time_after_eq(jiffies, time_limit)))
2246			return;
2247
2248		skb = tipc_skb_dequeue(inputq, dport);
2249		if (unlikely(!skb))
2250			return;
2251
2252		/* Add message directly to receive queue if possible */
2253		if (!sock_owned_by_user(sk)) {
2254			tipc_sk_filter_rcv(sk, skb, xmitq);
2255			continue;
2256		}
2257
2258		/* Try backlog, compensating for double-counted bytes */
2259		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2260		if (!sk->sk_backlog.len)
2261			atomic_set(dcnt, 0);
2262		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2263		if (likely(!sk_add_backlog(sk, skb, lim))) {
2264			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2265						 "bklg & rcvq >90% allocated!");
2266			continue;
2267		}
2268
2269		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2270		/* Overload => reject message back to sender */
2271		onode = tipc_own_addr(sock_net(sk));
2272		atomic_inc(&sk->sk_drops);
2273		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2274			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2275					      "@sk_enqueue!");
2276			__skb_queue_tail(xmitq, skb);
2277		}
2278		break;
2279	}
2280}
2281
2282/**
2283 * tipc_sk_rcv - handle a chain of incoming buffers
2284 * @inputq: buffer list containing the buffers
2285 * Consumes all buffers in list until inputq is empty
2286 * Note: may be called in multiple threads referring to the same queue
2287 */
2288void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2289{
2290	struct sk_buff_head xmitq;
2291	u32 dnode, dport = 0;
2292	int err;
2293	struct tipc_sock *tsk;
2294	struct sock *sk;
2295	struct sk_buff *skb;
2296
2297	__skb_queue_head_init(&xmitq);
2298	while (skb_queue_len(inputq)) {
2299		dport = tipc_skb_peek_port(inputq, dport);
2300		tsk = tipc_sk_lookup(net, dport);
2301
2302		if (likely(tsk)) {
2303			sk = &tsk->sk;
2304			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2305				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2306				spin_unlock_bh(&sk->sk_lock.slock);
2307			}
2308			/* Send pending response/rejected messages, if any */
2309			tipc_node_distr_xmit(sock_net(sk), &xmitq);
 
 
 
2310			sock_put(sk);
2311			continue;
2312		}
 
2313		/* No destination socket => dequeue skb if still there */
2314		skb = tipc_skb_dequeue(inputq, dport);
2315		if (!skb)
2316			return;
2317
2318		/* Try secondary lookup if unresolved named message */
2319		err = TIPC_ERR_NO_PORT;
2320		if (tipc_msg_lookup_dest(net, skb, &err))
2321			goto xmit;
2322
2323		/* Prepare for message rejection */
2324		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2325			continue;
2326
2327		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2328xmit:
2329		dnode = msg_destnode(buf_msg(skb));
2330		tipc_node_xmit_skb(net, skb, dnode, dport);
2331	}
2332}
2333
2334static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2335{
2336	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2337	struct sock *sk = sock->sk;
2338	int done;
2339
2340	do {
2341		int err = sock_error(sk);
2342		if (err)
2343			return err;
2344		if (!*timeo_p)
2345			return -ETIMEDOUT;
2346		if (signal_pending(current))
2347			return sock_intr_errno(*timeo_p);
2348
2349		add_wait_queue(sk_sleep(sk), &wait);
2350		done = sk_wait_event(sk, timeo_p,
2351				     sk->sk_state != TIPC_CONNECTING, &wait);
2352		remove_wait_queue(sk_sleep(sk), &wait);
2353	} while (!done);
2354	return 0;
2355}
2356
2357static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2358{
2359	if (addr->family != AF_TIPC)
2360		return false;
2361	if (addr->addrtype == TIPC_SERVICE_RANGE)
2362		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2363	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2364		addr->addrtype == TIPC_SOCKET_ADDR);
2365}
2366
2367/**
2368 * tipc_connect - establish a connection to another TIPC port
2369 * @sock: socket structure
2370 * @dest: socket address for destination port
2371 * @destlen: size of socket address data structure
2372 * @flags: file-related flags associated with socket
2373 *
2374 * Returns 0 on success, errno otherwise
2375 */
2376static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2377			int destlen, int flags)
2378{
2379	struct sock *sk = sock->sk;
2380	struct tipc_sock *tsk = tipc_sk(sk);
2381	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2382	struct msghdr m = {NULL,};
2383	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2384	int previous;
2385	int res = 0;
2386
2387	if (destlen != sizeof(struct sockaddr_tipc))
2388		return -EINVAL;
2389
2390	lock_sock(sk);
2391
2392	if (tsk->group) {
2393		res = -EINVAL;
 
 
 
 
 
 
 
2394		goto exit;
2395	}
2396
2397	if (dst->family == AF_UNSPEC) {
2398		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2399		if (!tipc_sk_type_connectionless(sk))
2400			res = -EINVAL;
2401		goto exit;
2402	}
2403	if (!tipc_sockaddr_is_sane(dst)) {
2404		res = -EINVAL;
2405		goto exit;
2406	}
2407	/* DGRAM/RDM connect(), just save the destaddr */
2408	if (tipc_sk_type_connectionless(sk)) {
2409		memcpy(&tsk->peer, dest, destlen);
2410		goto exit;
2411	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2412		res = -EINVAL;
2413		goto exit;
2414	}
2415
2416	previous = sk->sk_state;
2417
2418	switch (sk->sk_state) {
2419	case TIPC_OPEN:
2420		/* Send a 'SYN-' to destination */
2421		m.msg_name = dest;
2422		m.msg_namelen = destlen;
2423
2424		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2425		 * indicate send_msg() is never blocked.
2426		 */
2427		if (!timeout)
2428			m.msg_flags = MSG_DONTWAIT;
2429
2430		res = __tipc_sendmsg(sock, &m, 0);
2431		if ((res < 0) && (res != -EWOULDBLOCK))
2432			goto exit;
2433
2434		/* Just entered TIPC_CONNECTING state; the only
2435		 * difference is that return value in non-blocking
2436		 * case is EINPROGRESS, rather than EALREADY.
2437		 */
2438		res = -EINPROGRESS;
2439		/* fall through */
2440	case TIPC_CONNECTING:
2441		if (!timeout) {
2442			if (previous == TIPC_CONNECTING)
2443				res = -EALREADY;
2444			goto exit;
2445		}
2446		timeout = msecs_to_jiffies(timeout);
2447		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2448		res = tipc_wait_for_connect(sock, &timeout);
2449		break;
2450	case TIPC_ESTABLISHED:
2451		res = -EISCONN;
2452		break;
2453	default:
2454		res = -EINVAL;
2455	}
2456
2457exit:
2458	release_sock(sk);
2459	return res;
2460}
2461
2462/**
2463 * tipc_listen - allow socket to listen for incoming connections
2464 * @sock: socket structure
2465 * @len: (unused)
2466 *
2467 * Returns 0 on success, errno otherwise
2468 */
2469static int tipc_listen(struct socket *sock, int len)
2470{
2471	struct sock *sk = sock->sk;
2472	int res;
2473
2474	lock_sock(sk);
2475	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2476	release_sock(sk);
2477
2478	return res;
2479}
2480
2481static int tipc_wait_for_accept(struct socket *sock, long timeo)
2482{
2483	struct sock *sk = sock->sk;
2484	DEFINE_WAIT(wait);
2485	int err;
2486
2487	/* True wake-one mechanism for incoming connections: only
2488	 * one process gets woken up, not the 'whole herd'.
2489	 * Since we do not 'race & poll' for established sockets
2490	 * anymore, the common case will execute the loop only once.
2491	*/
2492	for (;;) {
2493		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2494					  TASK_INTERRUPTIBLE);
2495		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2496			release_sock(sk);
2497			timeo = schedule_timeout(timeo);
2498			lock_sock(sk);
2499		}
2500		err = 0;
2501		if (!skb_queue_empty(&sk->sk_receive_queue))
2502			break;
2503		err = -EAGAIN;
2504		if (!timeo)
2505			break;
2506		err = sock_intr_errno(timeo);
2507		if (signal_pending(current))
2508			break;
2509	}
2510	finish_wait(sk_sleep(sk), &wait);
2511	return err;
2512}
2513
2514/**
2515 * tipc_accept - wait for connection request
2516 * @sock: listening socket
2517 * @newsock: new socket that is to be connected
2518 * @flags: file-related flags associated with socket
2519 *
2520 * Returns 0 on success, errno otherwise
2521 */
2522static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2523		       bool kern)
2524{
2525	struct sock *new_sk, *sk = sock->sk;
2526	struct sk_buff *buf;
2527	struct tipc_sock *new_tsock;
2528	struct tipc_msg *msg;
2529	long timeo;
2530	int res;
2531
2532	lock_sock(sk);
2533
2534	if (sk->sk_state != TIPC_LISTEN) {
2535		res = -EINVAL;
2536		goto exit;
2537	}
2538	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2539	res = tipc_wait_for_accept(sock, timeo);
2540	if (res)
2541		goto exit;
2542
2543	buf = skb_peek(&sk->sk_receive_queue);
2544
2545	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2546	if (res)
2547		goto exit;
2548	security_sk_clone(sock->sk, new_sock->sk);
2549
2550	new_sk = new_sock->sk;
2551	new_tsock = tipc_sk(new_sk);
2552	msg = buf_msg(buf);
2553
2554	/* we lock on new_sk; but lockdep sees the lock on sk */
2555	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2556
2557	/*
2558	 * Reject any stray messages received by new socket
2559	 * before the socket lock was taken (very, very unlikely)
2560	 */
2561	tsk_rej_rx_queue(new_sk);
2562
2563	/* Connect new socket to it's peer */
2564	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2565
2566	tsk_set_importance(new_tsock, msg_importance(msg));
2567	if (msg_named(msg)) {
2568		new_tsock->conn_type = msg_nametype(msg);
2569		new_tsock->conn_instance = msg_nameinst(msg);
2570	}
2571
2572	/*
2573	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2574	 * Respond to 'SYN+' by queuing it on new socket.
2575	 */
2576	if (!msg_data_sz(msg)) {
2577		struct msghdr m = {NULL,};
2578
2579		tsk_advance_rx_queue(sk);
2580		__tipc_sendstream(new_sock, &m, 0);
2581	} else {
2582		__skb_dequeue(&sk->sk_receive_queue);
2583		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2584		skb_set_owner_r(buf, new_sk);
2585	}
2586	release_sock(new_sk);
2587exit:
2588	release_sock(sk);
2589	return res;
2590}
2591
2592/**
2593 * tipc_shutdown - shutdown socket connection
2594 * @sock: socket structure
2595 * @how: direction to close (must be SHUT_RDWR)
2596 *
2597 * Terminates connection (if necessary), then purges socket's receive queue.
2598 *
2599 * Returns 0 on success, errno otherwise
2600 */
2601static int tipc_shutdown(struct socket *sock, int how)
2602{
2603	struct sock *sk = sock->sk;
2604	int res;
2605
2606	if (how != SHUT_RDWR)
2607		return -EINVAL;
2608
2609	lock_sock(sk);
2610
2611	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2612	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2613	sk->sk_shutdown = SEND_SHUTDOWN;
2614
2615	if (sk->sk_state == TIPC_DISCONNECTING) {
2616		/* Discard any unreceived messages */
2617		__skb_queue_purge(&sk->sk_receive_queue);
2618
2619		/* Wake up anyone sleeping in poll */
2620		sk->sk_state_change(sk);
2621		res = 0;
2622	} else {
2623		res = -ENOTCONN;
2624	}
2625
2626	release_sock(sk);
2627	return res;
2628}
2629
2630static void tipc_sk_check_probing_state(struct sock *sk,
2631					struct sk_buff_head *list)
2632{
2633	struct tipc_sock *tsk = tipc_sk(sk);
2634	u32 pnode = tsk_peer_node(tsk);
2635	u32 pport = tsk_peer_port(tsk);
2636	u32 self = tsk_own_node(tsk);
2637	u32 oport = tsk->portid;
2638	struct sk_buff *skb;
2639
2640	if (tsk->probe_unacked) {
2641		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2642		sk->sk_err = ECONNABORTED;
2643		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2644		sk->sk_state_change(sk);
2645		return;
2646	}
2647	/* Prepare new probe */
2648	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2649			      pnode, self, pport, oport, TIPC_OK);
2650	if (skb)
2651		__skb_queue_tail(list, skb);
2652	tsk->probe_unacked = true;
2653	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2654}
2655
2656static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2657{
2658	struct tipc_sock *tsk = tipc_sk(sk);
2659
2660	/* Try again later if dest link is congested */
2661	if (tsk->cong_link_cnt) {
2662		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2663		return;
2664	}
2665	/* Prepare SYN for retransmit */
2666	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2667}
2668
2669static void tipc_sk_timeout(struct timer_list *t)
2670{
2671	struct sock *sk = from_timer(sk, t, sk_timer);
2672	struct tipc_sock *tsk = tipc_sk(sk);
2673	u32 pnode = tsk_peer_node(tsk);
2674	struct sk_buff_head list;
2675	int rc = 0;
2676
2677	__skb_queue_head_init(&list);
2678	bh_lock_sock(sk);
2679
2680	/* Try again later if socket is busy */
2681	if (sock_owned_by_user(sk)) {
2682		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2683		bh_unlock_sock(sk);
2684		return;
2685	}
2686
2687	if (sk->sk_state == TIPC_ESTABLISHED)
2688		tipc_sk_check_probing_state(sk, &list);
2689	else if (sk->sk_state == TIPC_CONNECTING)
2690		tipc_sk_retry_connect(sk, &list);
2691
2692	bh_unlock_sock(sk);
2693
2694	if (!skb_queue_empty(&list))
2695		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2696
2697	/* SYN messages may cause link congestion */
2698	if (rc == -ELINKCONG) {
2699		tipc_dest_push(&tsk->cong_links, pnode, 0);
2700		tsk->cong_link_cnt = 1;
2701	}
2702	sock_put(sk);
2703}
2704
2705static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2706			   struct tipc_name_seq const *seq)
2707{
2708	struct sock *sk = &tsk->sk;
2709	struct net *net = sock_net(sk);
2710	struct publication *publ;
2711	u32 key;
2712
2713	if (scope != TIPC_NODE_SCOPE)
2714		scope = TIPC_CLUSTER_SCOPE;
2715
2716	if (tipc_sk_connected(sk))
2717		return -EINVAL;
2718	key = tsk->portid + tsk->pub_count + 1;
2719	if (key == tsk->portid)
2720		return -EADDRINUSE;
2721
2722	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2723				    scope, tsk->portid, key);
2724	if (unlikely(!publ))
2725		return -EINVAL;
2726
2727	list_add(&publ->binding_sock, &tsk->publications);
2728	tsk->pub_count++;
2729	tsk->published = 1;
2730	return 0;
2731}
2732
2733static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2734			    struct tipc_name_seq const *seq)
2735{
2736	struct net *net = sock_net(&tsk->sk);
2737	struct publication *publ;
2738	struct publication *safe;
2739	int rc = -EINVAL;
2740
2741	if (scope != TIPC_NODE_SCOPE)
2742		scope = TIPC_CLUSTER_SCOPE;
2743
2744	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2745		if (seq) {
2746			if (publ->scope != scope)
2747				continue;
2748			if (publ->type != seq->type)
2749				continue;
2750			if (publ->lower != seq->lower)
2751				continue;
2752			if (publ->upper != seq->upper)
2753				break;
2754			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2755					      publ->upper, publ->key);
2756			rc = 0;
2757			break;
2758		}
2759		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2760				      publ->upper, publ->key);
2761		rc = 0;
2762	}
2763	if (list_empty(&tsk->publications))
2764		tsk->published = 0;
2765	return rc;
2766}
2767
2768/* tipc_sk_reinit: set non-zero address in all existing sockets
2769 *                 when we go from standalone to network mode.
2770 */
2771void tipc_sk_reinit(struct net *net)
2772{
2773	struct tipc_net *tn = net_generic(net, tipc_net_id);
2774	struct rhashtable_iter iter;
 
2775	struct tipc_sock *tsk;
2776	struct tipc_msg *msg;
 
2777
2778	rhashtable_walk_enter(&tn->sk_rht, &iter);
2779
2780	do {
2781		rhashtable_walk_start(&iter);
2782
2783		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2784			sock_hold(&tsk->sk);
2785			rhashtable_walk_stop(&iter);
2786			lock_sock(&tsk->sk);
2787			msg = &tsk->phdr;
2788			msg_set_prevnode(msg, tipc_own_addr(net));
2789			msg_set_orignode(msg, tipc_own_addr(net));
2790			release_sock(&tsk->sk);
2791			rhashtable_walk_start(&iter);
2792			sock_put(&tsk->sk);
2793		}
2794
2795		rhashtable_walk_stop(&iter);
2796	} while (tsk == ERR_PTR(-EAGAIN));
2797
2798	rhashtable_walk_exit(&iter);
2799}
2800
2801static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2802{
2803	struct tipc_net *tn = net_generic(net, tipc_net_id);
2804	struct tipc_sock *tsk;
2805
2806	rcu_read_lock();
2807	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2808	if (tsk)
2809		sock_hold(&tsk->sk);
2810	rcu_read_unlock();
2811
2812	return tsk;
2813}
2814
2815static int tipc_sk_insert(struct tipc_sock *tsk)
2816{
2817	struct sock *sk = &tsk->sk;
2818	struct net *net = sock_net(sk);
2819	struct tipc_net *tn = net_generic(net, tipc_net_id);
2820	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2821	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2822
2823	while (remaining--) {
2824		portid++;
2825		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2826			portid = TIPC_MIN_PORT;
2827		tsk->portid = portid;
2828		sock_hold(&tsk->sk);
2829		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2830						   tsk_rht_params))
2831			return 0;
2832		sock_put(&tsk->sk);
2833	}
2834
2835	return -1;
2836}
2837
2838static void tipc_sk_remove(struct tipc_sock *tsk)
2839{
2840	struct sock *sk = &tsk->sk;
2841	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2842
2843	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2844		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2845		__sock_put(sk);
2846	}
2847}
2848
2849static const struct rhashtable_params tsk_rht_params = {
2850	.nelem_hint = 192,
2851	.head_offset = offsetof(struct tipc_sock, node),
2852	.key_offset = offsetof(struct tipc_sock, portid),
2853	.key_len = sizeof(u32), /* portid */
2854	.max_size = 1048576,
2855	.min_size = 256,
2856	.automatic_shrinking = true,
2857};
2858
2859int tipc_sk_rht_init(struct net *net)
2860{
2861	struct tipc_net *tn = net_generic(net, tipc_net_id);
2862
2863	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2864}
2865
2866void tipc_sk_rht_destroy(struct net *net)
2867{
2868	struct tipc_net *tn = net_generic(net, tipc_net_id);
2869
2870	/* Wait for socket readers to complete */
2871	synchronize_net();
2872
2873	rhashtable_destroy(&tn->sk_rht);
2874}
2875
2876static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2877{
2878	struct net *net = sock_net(&tsk->sk);
2879	struct tipc_group *grp = tsk->group;
2880	struct tipc_msg *hdr = &tsk->phdr;
2881	struct tipc_name_seq seq;
2882	int rc;
2883
2884	if (mreq->type < TIPC_RESERVED_TYPES)
2885		return -EACCES;
2886	if (mreq->scope > TIPC_NODE_SCOPE)
2887		return -EINVAL;
2888	if (grp)
2889		return -EACCES;
2890	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2891	if (!grp)
2892		return -ENOMEM;
2893	tsk->group = grp;
2894	msg_set_lookup_scope(hdr, mreq->scope);
2895	msg_set_nametype(hdr, mreq->type);
2896	msg_set_dest_droppable(hdr, true);
2897	seq.type = mreq->type;
2898	seq.lower = mreq->instance;
2899	seq.upper = seq.lower;
2900	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2901	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2902	if (rc) {
2903		tipc_group_delete(net, grp);
2904		tsk->group = NULL;
2905		return rc;
2906	}
2907	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2908	tsk->mc_method.rcast = true;
2909	tsk->mc_method.mandatory = true;
2910	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2911	return rc;
2912}
2913
2914static int tipc_sk_leave(struct tipc_sock *tsk)
2915{
2916	struct net *net = sock_net(&tsk->sk);
2917	struct tipc_group *grp = tsk->group;
2918	struct tipc_name_seq seq;
2919	int scope;
2920
2921	if (!grp)
2922		return -EINVAL;
2923	tipc_group_self(grp, &seq, &scope);
2924	tipc_group_delete(net, grp);
2925	tsk->group = NULL;
2926	tipc_sk_withdraw(tsk, scope, &seq);
2927	return 0;
2928}
2929
2930/**
2931 * tipc_setsockopt - set socket option
2932 * @sock: socket structure
2933 * @lvl: option level
2934 * @opt: option identifier
2935 * @ov: pointer to new option value
2936 * @ol: length of option value
2937 *
2938 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2939 * (to ease compatibility).
2940 *
2941 * Returns 0 on success, errno otherwise
2942 */
2943static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2944			   char __user *ov, unsigned int ol)
2945{
2946	struct sock *sk = sock->sk;
2947	struct tipc_sock *tsk = tipc_sk(sk);
2948	struct tipc_group_req mreq;
2949	u32 value = 0;
2950	int res = 0;
2951
2952	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2953		return 0;
2954	if (lvl != SOL_TIPC)
2955		return -ENOPROTOOPT;
2956
2957	switch (opt) {
2958	case TIPC_IMPORTANCE:
2959	case TIPC_SRC_DROPPABLE:
2960	case TIPC_DEST_DROPPABLE:
2961	case TIPC_CONN_TIMEOUT:
2962		if (ol < sizeof(value))
2963			return -EINVAL;
2964		if (get_user(value, (u32 __user *)ov))
2965			return -EFAULT;
2966		break;
2967	case TIPC_GROUP_JOIN:
2968		if (ol < sizeof(mreq))
2969			return -EINVAL;
2970		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2971			return -EFAULT;
2972		break;
2973	default:
2974		if (ov || ol)
2975			return -EINVAL;
2976	}
2977
2978	lock_sock(sk);
2979
2980	switch (opt) {
2981	case TIPC_IMPORTANCE:
2982		res = tsk_set_importance(tsk, value);
2983		break;
2984	case TIPC_SRC_DROPPABLE:
2985		if (sock->type != SOCK_STREAM)
2986			tsk_set_unreliable(tsk, value);
2987		else
2988			res = -ENOPROTOOPT;
2989		break;
2990	case TIPC_DEST_DROPPABLE:
2991		tsk_set_unreturnable(tsk, value);
2992		break;
2993	case TIPC_CONN_TIMEOUT:
2994		tipc_sk(sk)->conn_timeout = value;
2995		break;
2996	case TIPC_MCAST_BROADCAST:
2997		tsk->mc_method.rcast = false;
2998		tsk->mc_method.mandatory = true;
2999		break;
3000	case TIPC_MCAST_REPLICAST:
3001		tsk->mc_method.rcast = true;
3002		tsk->mc_method.mandatory = true;
3003		break;
3004	case TIPC_GROUP_JOIN:
3005		res = tipc_sk_join(tsk, &mreq);
3006		break;
3007	case TIPC_GROUP_LEAVE:
3008		res = tipc_sk_leave(tsk);
3009		break;
3010	default:
3011		res = -EINVAL;
3012	}
3013
3014	release_sock(sk);
3015
3016	return res;
3017}
3018
3019/**
3020 * tipc_getsockopt - get socket option
3021 * @sock: socket structure
3022 * @lvl: option level
3023 * @opt: option identifier
3024 * @ov: receptacle for option value
3025 * @ol: receptacle for length of option value
3026 *
3027 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3028 * (to ease compatibility).
3029 *
3030 * Returns 0 on success, errno otherwise
3031 */
3032static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3033			   char __user *ov, int __user *ol)
3034{
3035	struct sock *sk = sock->sk;
3036	struct tipc_sock *tsk = tipc_sk(sk);
3037	struct tipc_name_seq seq;
3038	int len, scope;
3039	u32 value;
3040	int res;
3041
3042	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3043		return put_user(0, ol);
3044	if (lvl != SOL_TIPC)
3045		return -ENOPROTOOPT;
3046	res = get_user(len, ol);
3047	if (res)
3048		return res;
3049
3050	lock_sock(sk);
3051
3052	switch (opt) {
3053	case TIPC_IMPORTANCE:
3054		value = tsk_importance(tsk);
3055		break;
3056	case TIPC_SRC_DROPPABLE:
3057		value = tsk_unreliable(tsk);
3058		break;
3059	case TIPC_DEST_DROPPABLE:
3060		value = tsk_unreturnable(tsk);
3061		break;
3062	case TIPC_CONN_TIMEOUT:
3063		value = tsk->conn_timeout;
3064		/* no need to set "res", since already 0 at this point */
3065		break;
3066	case TIPC_NODE_RECVQ_DEPTH:
3067		value = 0; /* was tipc_queue_size, now obsolete */
3068		break;
3069	case TIPC_SOCK_RECVQ_DEPTH:
3070		value = skb_queue_len(&sk->sk_receive_queue);
3071		break;
3072	case TIPC_SOCK_RECVQ_USED:
3073		value = sk_rmem_alloc_get(sk);
3074		break;
3075	case TIPC_GROUP_JOIN:
3076		seq.type = 0;
3077		if (tsk->group)
3078			tipc_group_self(tsk->group, &seq, &scope);
3079		value = seq.type;
3080		break;
3081	default:
3082		res = -EINVAL;
3083	}
3084
3085	release_sock(sk);
3086
3087	if (res)
3088		return res;	/* "get" failed */
3089
3090	if (len < sizeof(value))
3091		return -EINVAL;
3092
3093	if (copy_to_user(ov, &value, sizeof(value)))
3094		return -EFAULT;
3095
3096	return put_user(sizeof(value), ol);
3097}
3098
3099static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3100{
3101	struct net *net = sock_net(sock->sk);
3102	struct tipc_sioc_nodeid_req nr = {0};
3103	struct tipc_sioc_ln_req lnr;
3104	void __user *argp = (void __user *)arg;
3105
3106	switch (cmd) {
3107	case SIOCGETLINKNAME:
3108		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3109			return -EFAULT;
3110		if (!tipc_node_get_linkname(net,
3111					    lnr.bearer_id & 0xffff, lnr.peer,
3112					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3113			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3114				return -EFAULT;
3115			return 0;
3116		}
3117		return -EADDRNOTAVAIL;
3118	case SIOCGETNODEID:
3119		if (copy_from_user(&nr, argp, sizeof(nr)))
3120			return -EFAULT;
3121		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3122			return -EADDRNOTAVAIL;
3123		if (copy_to_user(argp, &nr, sizeof(nr)))
3124			return -EFAULT;
3125		return 0;
3126	default:
3127		return -ENOIOCTLCMD;
3128	}
3129}
3130
3131static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3132{
3133	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3134	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3135	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3136
3137	tsk1->peer.family = AF_TIPC;
3138	tsk1->peer.addrtype = TIPC_ADDR_ID;
3139	tsk1->peer.scope = TIPC_NODE_SCOPE;
3140	tsk1->peer.addr.id.ref = tsk2->portid;
3141	tsk1->peer.addr.id.node = onode;
3142	tsk2->peer.family = AF_TIPC;
3143	tsk2->peer.addrtype = TIPC_ADDR_ID;
3144	tsk2->peer.scope = TIPC_NODE_SCOPE;
3145	tsk2->peer.addr.id.ref = tsk1->portid;
3146	tsk2->peer.addr.id.node = onode;
3147
3148	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3149	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3150	return 0;
3151}
3152
3153/* Protocol switches for the various types of TIPC sockets */
3154
3155static const struct proto_ops msg_ops = {
3156	.owner		= THIS_MODULE,
3157	.family		= AF_TIPC,
3158	.release	= tipc_release,
3159	.bind		= tipc_bind,
3160	.connect	= tipc_connect,
3161	.socketpair	= tipc_socketpair,
3162	.accept		= sock_no_accept,
3163	.getname	= tipc_getname,
3164	.poll		= tipc_poll,
3165	.ioctl		= tipc_ioctl,
3166	.listen		= sock_no_listen,
3167	.shutdown	= tipc_shutdown,
3168	.setsockopt	= tipc_setsockopt,
3169	.getsockopt	= tipc_getsockopt,
3170	.sendmsg	= tipc_sendmsg,
3171	.recvmsg	= tipc_recvmsg,
3172	.mmap		= sock_no_mmap,
3173	.sendpage	= sock_no_sendpage
3174};
3175
3176static const struct proto_ops packet_ops = {
3177	.owner		= THIS_MODULE,
3178	.family		= AF_TIPC,
3179	.release	= tipc_release,
3180	.bind		= tipc_bind,
3181	.connect	= tipc_connect,
3182	.socketpair	= tipc_socketpair,
3183	.accept		= tipc_accept,
3184	.getname	= tipc_getname,
3185	.poll		= tipc_poll,
3186	.ioctl		= tipc_ioctl,
3187	.listen		= tipc_listen,
3188	.shutdown	= tipc_shutdown,
3189	.setsockopt	= tipc_setsockopt,
3190	.getsockopt	= tipc_getsockopt,
3191	.sendmsg	= tipc_send_packet,
3192	.recvmsg	= tipc_recvmsg,
3193	.mmap		= sock_no_mmap,
3194	.sendpage	= sock_no_sendpage
3195};
3196
3197static const struct proto_ops stream_ops = {
3198	.owner		= THIS_MODULE,
3199	.family		= AF_TIPC,
3200	.release	= tipc_release,
3201	.bind		= tipc_bind,
3202	.connect	= tipc_connect,
3203	.socketpair	= tipc_socketpair,
3204	.accept		= tipc_accept,
3205	.getname	= tipc_getname,
3206	.poll		= tipc_poll,
3207	.ioctl		= tipc_ioctl,
3208	.listen		= tipc_listen,
3209	.shutdown	= tipc_shutdown,
3210	.setsockopt	= tipc_setsockopt,
3211	.getsockopt	= tipc_getsockopt,
3212	.sendmsg	= tipc_sendstream,
3213	.recvmsg	= tipc_recvstream,
3214	.mmap		= sock_no_mmap,
3215	.sendpage	= sock_no_sendpage
3216};
3217
3218static const struct net_proto_family tipc_family_ops = {
3219	.owner		= THIS_MODULE,
3220	.family		= AF_TIPC,
3221	.create		= tipc_sk_create
3222};
3223
3224static struct proto tipc_proto = {
3225	.name		= "TIPC",
3226	.owner		= THIS_MODULE,
3227	.obj_size	= sizeof(struct tipc_sock),
3228	.sysctl_rmem	= sysctl_tipc_rmem
3229};
3230
3231/**
3232 * tipc_socket_init - initialize TIPC socket interface
3233 *
3234 * Returns 0 on success, errno otherwise
3235 */
3236int tipc_socket_init(void)
3237{
3238	int res;
3239
3240	res = proto_register(&tipc_proto, 1);
3241	if (res) {
3242		pr_err("Failed to register TIPC protocol type\n");
3243		goto out;
3244	}
3245
3246	res = sock_register(&tipc_family_ops);
3247	if (res) {
3248		pr_err("Failed to register TIPC socket type\n");
3249		proto_unregister(&tipc_proto);
3250		goto out;
3251	}
3252 out:
3253	return res;
3254}
3255
3256/**
3257 * tipc_socket_stop - stop TIPC socket interface
3258 */
3259void tipc_socket_stop(void)
3260{
3261	sock_unregister(tipc_family_ops.family);
3262	proto_unregister(&tipc_proto);
3263}
3264
3265/* Caller should hold socket lock for the passed tipc socket. */
3266static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3267{
3268	u32 peer_node;
3269	u32 peer_port;
3270	struct nlattr *nest;
3271
3272	peer_node = tsk_peer_node(tsk);
3273	peer_port = tsk_peer_port(tsk);
3274
3275	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3276	if (!nest)
3277		return -EMSGSIZE;
3278
3279	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3280		goto msg_full;
3281	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3282		goto msg_full;
3283
3284	if (tsk->conn_type != 0) {
3285		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3286			goto msg_full;
3287		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3288			goto msg_full;
3289		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3290			goto msg_full;
3291	}
3292	nla_nest_end(skb, nest);
3293
3294	return 0;
3295
3296msg_full:
3297	nla_nest_cancel(skb, nest);
3298
3299	return -EMSGSIZE;
3300}
3301
3302static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3303			  *tsk)
3304{
3305	struct net *net = sock_net(skb->sk);
3306	struct sock *sk = &tsk->sk;
3307
3308	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3309	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3310		return -EMSGSIZE;
3311
3312	if (tipc_sk_connected(sk)) {
3313		if (__tipc_nl_add_sk_con(skb, tsk))
3314			return -EMSGSIZE;
3315	} else if (!list_empty(&tsk->publications)) {
3316		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3317			return -EMSGSIZE;
3318	}
3319	return 0;
3320}
3321
3322/* Caller should hold socket lock for the passed tipc socket. */
3323static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3324			    struct tipc_sock *tsk)
3325{
 
 
3326	struct nlattr *attrs;
3327	void *hdr;
 
 
3328
3329	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3330			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3331	if (!hdr)
3332		goto msg_cancel;
3333
3334	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3335	if (!attrs)
3336		goto genlmsg_cancel;
3337
3338	if (__tipc_nl_add_sk_info(skb, tsk))
 
3339		goto attr_msg_cancel;
3340
 
 
 
 
 
 
 
 
3341	nla_nest_end(skb, attrs);
3342	genlmsg_end(skb, hdr);
3343
3344	return 0;
3345
3346attr_msg_cancel:
3347	nla_nest_cancel(skb, attrs);
3348genlmsg_cancel:
3349	genlmsg_cancel(skb, hdr);
3350msg_cancel:
3351	return -EMSGSIZE;
3352}
3353
3354int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3355		    int (*skb_handler)(struct sk_buff *skb,
3356				       struct netlink_callback *cb,
3357				       struct tipc_sock *tsk))
3358{
3359	struct rhashtable_iter *iter = (void *)cb->args[4];
3360	struct tipc_sock *tsk;
3361	int err;
 
 
 
 
 
3362
3363	rhashtable_walk_start(iter);
3364	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3365		if (IS_ERR(tsk)) {
3366			err = PTR_ERR(tsk);
3367			if (err == -EAGAIN) {
3368				err = 0;
 
3369				continue;
3370			}
3371			break;
3372		}
3373
3374		sock_hold(&tsk->sk);
3375		rhashtable_walk_stop(iter);
3376		lock_sock(&tsk->sk);
3377		err = skb_handler(skb, cb, tsk);
3378		if (err) {
3379			release_sock(&tsk->sk);
3380			sock_put(&tsk->sk);
3381			goto out;
3382		}
3383		release_sock(&tsk->sk);
3384		rhashtable_walk_start(iter);
3385		sock_put(&tsk->sk);
3386	}
3387	rhashtable_walk_stop(iter);
3388out:
 
 
 
 
3389	return skb->len;
3390}
3391EXPORT_SYMBOL(tipc_nl_sk_walk);
3392
3393int tipc_dump_start(struct netlink_callback *cb)
3394{
3395	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3396}
3397EXPORT_SYMBOL(tipc_dump_start);
3398
3399int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3400{
3401	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3402	struct rhashtable_iter *iter = (void *)cb->args[4];
3403	struct tipc_net *tn = tipc_net(net);
3404
3405	if (!iter) {
3406		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3407		if (!iter)
3408			return -ENOMEM;
3409
3410		cb->args[4] = (long)iter;
3411	}
3412
3413	rhashtable_walk_enter(&tn->sk_rht, iter);
3414	return 0;
3415}
3416
3417int tipc_dump_done(struct netlink_callback *cb)
3418{
3419	struct rhashtable_iter *hti = (void *)cb->args[4];
3420
3421	rhashtable_walk_exit(hti);
3422	kfree(hti);
3423	return 0;
3424}
3425EXPORT_SYMBOL(tipc_dump_done);
3426
3427int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3428			   struct tipc_sock *tsk, u32 sk_filter_state,
3429			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3430{
3431	struct sock *sk = &tsk->sk;
3432	struct nlattr *attrs;
3433	struct nlattr *stat;
3434
3435	/*filter response w.r.t sk_state*/
3436	if (!(sk_filter_state & (1 << sk->sk_state)))
3437		return 0;
3438
3439	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3440	if (!attrs)
3441		goto msg_cancel;
3442
3443	if (__tipc_nl_add_sk_info(skb, tsk))
3444		goto attr_msg_cancel;
3445
3446	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3447	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3448	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3449	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3450			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3451					 sock_i_uid(sk))) ||
3452	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3453			      tipc_diag_gen_cookie(sk),
3454			      TIPC_NLA_SOCK_PAD))
3455		goto attr_msg_cancel;
3456
3457	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3458	if (!stat)
3459		goto attr_msg_cancel;
3460
3461	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3462			skb_queue_len(&sk->sk_receive_queue)) ||
3463	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3464			skb_queue_len(&sk->sk_write_queue)) ||
3465	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3466			atomic_read(&sk->sk_drops)))
3467		goto stat_msg_cancel;
3468
3469	if (tsk->cong_link_cnt &&
3470	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3471		goto stat_msg_cancel;
3472
3473	if (tsk_conn_cong(tsk) &&
3474	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3475		goto stat_msg_cancel;
3476
3477	nla_nest_end(skb, stat);
3478
3479	if (tsk->group)
3480		if (tipc_group_fill_sock_diag(tsk->group, skb))
3481			goto stat_msg_cancel;
3482
3483	nla_nest_end(skb, attrs);
3484
3485	return 0;
3486
3487stat_msg_cancel:
3488	nla_nest_cancel(skb, stat);
3489attr_msg_cancel:
3490	nla_nest_cancel(skb, attrs);
3491msg_cancel:
3492	return -EMSGSIZE;
3493}
3494EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3495
3496int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3497{
3498	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3499}
3500
3501/* Caller should hold socket lock for the passed tipc socket. */
3502static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3503				 struct netlink_callback *cb,
3504				 struct publication *publ)
3505{
3506	void *hdr;
3507	struct nlattr *attrs;
3508
3509	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3510			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3511	if (!hdr)
3512		goto msg_cancel;
3513
3514	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3515	if (!attrs)
3516		goto genlmsg_cancel;
3517
3518	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3519		goto attr_msg_cancel;
3520	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3521		goto attr_msg_cancel;
3522	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3523		goto attr_msg_cancel;
3524	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3525		goto attr_msg_cancel;
3526
3527	nla_nest_end(skb, attrs);
3528	genlmsg_end(skb, hdr);
3529
3530	return 0;
3531
3532attr_msg_cancel:
3533	nla_nest_cancel(skb, attrs);
3534genlmsg_cancel:
3535	genlmsg_cancel(skb, hdr);
3536msg_cancel:
3537	return -EMSGSIZE;
3538}
3539
3540/* Caller should hold socket lock for the passed tipc socket. */
3541static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3542				  struct netlink_callback *cb,
3543				  struct tipc_sock *tsk, u32 *last_publ)
3544{
3545	int err;
3546	struct publication *p;
3547
3548	if (*last_publ) {
3549		list_for_each_entry(p, &tsk->publications, binding_sock) {
3550			if (p->key == *last_publ)
3551				break;
3552		}
3553		if (p->key != *last_publ) {
3554			/* We never set seq or call nl_dump_check_consistent()
3555			 * this means that setting prev_seq here will cause the
3556			 * consistence check to fail in the netlink callback
3557			 * handler. Resulting in the last NLMSG_DONE message
3558			 * having the NLM_F_DUMP_INTR flag set.
3559			 */
3560			cb->prev_seq = 1;
3561			*last_publ = 0;
3562			return -EPIPE;
3563		}
3564	} else {
3565		p = list_first_entry(&tsk->publications, struct publication,
3566				     binding_sock);
3567	}
3568
3569	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3570		err = __tipc_nl_add_sk_publ(skb, cb, p);
3571		if (err) {
3572			*last_publ = p->key;
3573			return err;
3574		}
3575	}
3576	*last_publ = 0;
3577
3578	return 0;
3579}
3580
3581int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3582{
3583	int err;
3584	u32 tsk_portid = cb->args[0];
3585	u32 last_publ = cb->args[1];
3586	u32 done = cb->args[2];
3587	struct net *net = sock_net(skb->sk);
3588	struct tipc_sock *tsk;
3589
3590	if (!tsk_portid) {
3591		struct nlattr **attrs;
3592		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3593
3594		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3595		if (err)
3596			return err;
3597
3598		if (!attrs[TIPC_NLA_SOCK])
3599			return -EINVAL;
3600
3601		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3602						  attrs[TIPC_NLA_SOCK],
3603						  tipc_nl_sock_policy, NULL);
3604		if (err)
3605			return err;
3606
3607		if (!sock[TIPC_NLA_SOCK_REF])
3608			return -EINVAL;
3609
3610		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3611	}
3612
3613	if (done)
3614		return 0;
3615
3616	tsk = tipc_sk_lookup(net, tsk_portid);
3617	if (!tsk)
3618		return -EINVAL;
3619
3620	lock_sock(&tsk->sk);
3621	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3622	if (!err)
3623		done = 1;
3624	release_sock(&tsk->sk);
3625	sock_put(&tsk->sk);
3626
3627	cb->args[0] = tsk_portid;
3628	cb->args[1] = last_publ;
3629	cb->args[2] = done;
3630
3631	return skb->len;
3632}
3633
3634/**
3635 * tipc_sk_filtering - check if a socket should be traced
3636 * @sk: the socket to be examined
3637 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3638 *  (portid, sock type, name type, name lower, name upper)
3639 *
3640 * Returns true if the socket meets the socket tuple data
3641 * (value 0 = 'any') or when there is no tuple set (all = 0),
3642 * otherwise false
3643 */
3644bool tipc_sk_filtering(struct sock *sk)
3645{
3646	struct tipc_sock *tsk;
3647	struct publication *p;
3648	u32 _port, _sktype, _type, _lower, _upper;
3649	u32 type = 0, lower = 0, upper = 0;
3650
3651	if (!sk)
3652		return true;
3653
3654	tsk = tipc_sk(sk);
3655
3656	_port = sysctl_tipc_sk_filter[0];
3657	_sktype = sysctl_tipc_sk_filter[1];
3658	_type = sysctl_tipc_sk_filter[2];
3659	_lower = sysctl_tipc_sk_filter[3];
3660	_upper = sysctl_tipc_sk_filter[4];
3661
3662	if (!_port && !_sktype && !_type && !_lower && !_upper)
3663		return true;
3664
3665	if (_port)
3666		return (_port == tsk->portid);
3667
3668	if (_sktype && _sktype != sk->sk_type)
3669		return false;
3670
3671	if (tsk->published) {
3672		p = list_first_entry_or_null(&tsk->publications,
3673					     struct publication, binding_sock);
3674		if (p) {
3675			type = p->type;
3676			lower = p->lower;
3677			upper = p->upper;
3678		}
3679	}
3680
3681	if (!tipc_sk_type_connectionless(sk)) {
3682		type = tsk->conn_type;
3683		lower = tsk->conn_instance;
3684		upper = tsk->conn_instance;
3685	}
3686
3687	if ((_type && _type != type) || (_lower && _lower != lower) ||
3688	    (_upper && _upper != upper))
3689		return false;
3690
3691	return true;
3692}
3693
3694u32 tipc_sock_get_portid(struct sock *sk)
3695{
3696	return (sk) ? (tipc_sk(sk))->portid : 0;
3697}
3698
3699/**
3700 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3701 *			both the rcv and backlog queues are considered
3702 * @sk: tipc sk to be checked
3703 * @skb: tipc msg to be checked
3704 *
3705 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3706 */
3707
3708bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3709{
3710	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3711	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3712	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3713
3714	return (qsize > lim * 90 / 100);
3715}
3716
3717/**
3718 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3719 *			only the rcv queue is considered
3720 * @sk: tipc sk to be checked
3721 * @skb: tipc msg to be checked
3722 *
3723 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3724 */
3725
3726bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3727{
3728	unsigned int lim = rcvbuf_limit(sk, skb);
3729	unsigned int qsize = sk_rmem_alloc_get(sk);
3730
3731	return (qsize > lim * 90 / 100);
3732}
3733
3734/**
3735 * tipc_sk_dump - dump TIPC socket
3736 * @sk: tipc sk to be dumped
3737 * @dqueues: bitmask to decide if any socket queue to be dumped?
3738 *           - TIPC_DUMP_NONE: don't dump socket queues
3739 *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3740 *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3741 *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3742 *           - TIPC_DUMP_ALL: dump all the socket queues above
3743 * @buf: returned buffer of dump data in format
3744 */
3745int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3746{
3747	int i = 0;
3748	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3749	struct tipc_sock *tsk;
3750	struct publication *p;
3751	bool tsk_connected;
3752
3753	if (!sk) {
3754		i += scnprintf(buf, sz, "sk data: (null)\n");
3755		return i;
3756	}
3757
3758	tsk = tipc_sk(sk);
3759	tsk_connected = !tipc_sk_type_connectionless(sk);
3760
3761	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3762	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3763	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3764	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3765	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3766	if (tsk_connected) {
3767		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3768		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3769		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3770		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3771	}
3772	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3773	if (tsk->published) {
3774		p = list_first_entry_or_null(&tsk->publications,
3775					     struct publication, binding_sock);
3776		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3777		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3778		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3779	}
3780	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3781	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3782	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3783	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3784	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3785	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3786	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3787	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3788	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3789	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3790	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3791	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3792	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3793	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3794
3795	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3796		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3797		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3798	}
3799
3800	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3801		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3802		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3803	}
3804
3805	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3806		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3807		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3808		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3809			i += scnprintf(buf + i, sz - i, "  tail ");
3810			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3811					   buf + i);
3812		}
3813	}
3814
3815	return i;
3816}