Linux Audio

Check our new training course

Loading...
v5.9
   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 NAGLE_START_INIT	4
  52#define NAGLE_START_MAX		1024
  53#define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
  54#define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
  55#define TIPC_FWD_MSG		1
  56#define TIPC_MAX_PORT		0xffffffff
  57#define TIPC_MIN_PORT		1
  58#define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
  59
  60enum {
  61	TIPC_LISTEN = TCP_LISTEN,
  62	TIPC_ESTABLISHED = TCP_ESTABLISHED,
  63	TIPC_OPEN = TCP_CLOSE,
  64	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
  65	TIPC_CONNECTING = TCP_SYN_SENT,
  66};
  67
  68struct sockaddr_pair {
  69	struct sockaddr_tipc sock;
  70	struct sockaddr_tipc member;
  71};
  72
  73/**
  74 * struct tipc_sock - TIPC socket structure
  75 * @sk: socket - interacts with 'port' and with user via the socket API
  76 * @conn_type: TIPC type used when connection was established
  77 * @conn_instance: TIPC instance used when connection was established
  78 * @published: non-zero if port has one or more associated names
  79 * @max_pkt: maximum packet size "hint" used when building messages sent by port
  80 * @maxnagle: maximum size of msg which can be subject to nagle
  81 * @portid: unique port identity in TIPC socket hash table
  82 * @phdr: preformatted message header used when sending messages
  83 * #cong_links: list of congested links
  84 * @publications: list of publications for port
  85 * @blocking_link: address of the congested link we are currently sleeping on
  86 * @pub_count: total # of publications port has made during its lifetime
  87 * @conn_timeout: the time we can wait for an unresponded setup request
  88 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
  89 * @cong_link_cnt: number of congested links
  90 * @snt_unacked: # messages sent by socket, and not yet acked by peer
  91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
  92 * @peer: 'connected' peer for dgram/rdm
  93 * @node: hash table node
  94 * @mc_method: cookie for use between socket and broadcast layer
  95 * @rcu: rcu struct for tipc_sock
  96 */
  97struct tipc_sock {
  98	struct sock sk;
  99	u32 conn_type;
 100	u32 conn_instance;
 101	int published;
 102	u32 max_pkt;
 103	u32 maxnagle;
 104	u32 portid;
 105	struct tipc_msg phdr;
 106	struct list_head cong_links;
 107	struct list_head publications;
 108	u32 pub_count;
 109	atomic_t dupl_rcvcnt;
 110	u16 conn_timeout;
 111	bool probe_unacked;
 112	u16 cong_link_cnt;
 113	u16 snt_unacked;
 114	u16 snd_win;
 115	u16 peer_caps;
 116	u16 rcv_unacked;
 117	u16 rcv_win;
 118	struct sockaddr_tipc peer;
 119	struct rhash_head node;
 120	struct tipc_mc_method mc_method;
 121	struct rcu_head rcu;
 122	struct tipc_group *group;
 123	u32 oneway;
 124	u32 nagle_start;
 125	u16 snd_backlog;
 126	u16 msg_acc;
 127	u16 pkt_cnt;
 128	bool expect_ack;
 129	bool nodelay;
 130	bool group_is_open;
 131};
 132
 133static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
 134static void tipc_data_ready(struct sock *sk);
 135static void tipc_write_space(struct sock *sk);
 136static void tipc_sock_destruct(struct sock *sk);
 137static int tipc_release(struct socket *sock);
 138static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
 139		       bool kern);
 140static void tipc_sk_timeout(struct timer_list *t);
 141static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
 142			   struct tipc_name_seq const *seq);
 143static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
 144			    struct tipc_name_seq const *seq);
 145static int tipc_sk_leave(struct tipc_sock *tsk);
 146static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
 147static int tipc_sk_insert(struct tipc_sock *tsk);
 148static void tipc_sk_remove(struct tipc_sock *tsk);
 149static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
 150static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
 151static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
 152
 153static const struct proto_ops packet_ops;
 154static const struct proto_ops stream_ops;
 155static const struct proto_ops msg_ops;
 156static struct proto tipc_proto;
 157static const struct rhashtable_params tsk_rht_params;
 158
 159static u32 tsk_own_node(struct tipc_sock *tsk)
 160{
 161	return msg_prevnode(&tsk->phdr);
 162}
 163
 164static u32 tsk_peer_node(struct tipc_sock *tsk)
 165{
 166	return msg_destnode(&tsk->phdr);
 167}
 168
 169static u32 tsk_peer_port(struct tipc_sock *tsk)
 170{
 171	return msg_destport(&tsk->phdr);
 172}
 173
 174static  bool tsk_unreliable(struct tipc_sock *tsk)
 175{
 176	return msg_src_droppable(&tsk->phdr) != 0;
 177}
 178
 179static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
 180{
 181	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
 182}
 183
 184static bool tsk_unreturnable(struct tipc_sock *tsk)
 185{
 186	return msg_dest_droppable(&tsk->phdr) != 0;
 187}
 188
 189static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
 190{
 191	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
 192}
 193
 194static int tsk_importance(struct tipc_sock *tsk)
 195{
 196	return msg_importance(&tsk->phdr);
 197}
 198
 199static struct tipc_sock *tipc_sk(const struct sock *sk)
 200{
 201	return container_of(sk, struct tipc_sock, sk);
 202}
 203
 204int tsk_set_importance(struct sock *sk, int imp)
 205{
 206	if (imp > TIPC_CRITICAL_IMPORTANCE)
 207		return -EINVAL;
 208	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
 209	return 0;
 210}
 211
 
 
 
 
 
 212static bool tsk_conn_cong(struct tipc_sock *tsk)
 213{
 214	return tsk->snt_unacked > tsk->snd_win;
 215}
 216
 217static u16 tsk_blocks(int len)
 218{
 219	return ((len / FLOWCTL_BLK_SZ) + 1);
 220}
 221
 222/* tsk_blocks(): translate a buffer size in bytes to number of
 223 * advertisable blocks, taking into account the ratio truesize(len)/len
 224 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
 225 */
 226static u16 tsk_adv_blocks(int len)
 227{
 228	return len / FLOWCTL_BLK_SZ / 4;
 229}
 230
 231/* tsk_inc(): increment counter for sent or received data
 232 * - If block based flow control is not supported by peer we
 233 *   fall back to message based ditto, incrementing the counter
 234 */
 235static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
 236{
 237	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
 238		return ((msglen / FLOWCTL_BLK_SZ) + 1);
 239	return 1;
 240}
 241
 242/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
 243 */
 244static void tsk_set_nagle(struct tipc_sock *tsk)
 245{
 246	struct sock *sk = &tsk->sk;
 247
 248	tsk->maxnagle = 0;
 249	if (sk->sk_type != SOCK_STREAM)
 250		return;
 251	if (tsk->nodelay)
 252		return;
 253	if (!(tsk->peer_caps & TIPC_NAGLE))
 254		return;
 255	/* Limit node local buffer size to avoid receive queue overflow */
 256	if (tsk->max_pkt == MAX_MSG_SIZE)
 257		tsk->maxnagle = 1500;
 258	else
 259		tsk->maxnagle = tsk->max_pkt;
 260}
 261
 262/**
 263 * tsk_advance_rx_queue - discard first buffer in socket receive queue
 264 *
 265 * Caller must hold socket lock
 266 */
 267static void tsk_advance_rx_queue(struct sock *sk)
 268{
 269	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
 270	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
 271}
 272
 273/* tipc_sk_respond() : send response message back to sender
 274 */
 275static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
 276{
 277	u32 selector;
 278	u32 dnode;
 279	u32 onode = tipc_own_addr(sock_net(sk));
 280
 281	if (!tipc_msg_reverse(onode, &skb, err))
 282		return;
 283
 284	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
 285	dnode = msg_destnode(buf_msg(skb));
 286	selector = msg_origport(buf_msg(skb));
 287	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
 288}
 289
 290/**
 291 * tsk_rej_rx_queue - reject all buffers in socket receive queue
 292 *
 293 * Caller must hold socket lock
 294 */
 295static void tsk_rej_rx_queue(struct sock *sk, int error)
 296{
 297	struct sk_buff *skb;
 298
 299	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
 300		tipc_sk_respond(sk, skb, error);
 301}
 302
 303static bool tipc_sk_connected(struct sock *sk)
 304{
 305	return sk->sk_state == TIPC_ESTABLISHED;
 306}
 307
 308/* tipc_sk_type_connectionless - check if the socket is datagram socket
 309 * @sk: socket
 310 *
 311 * Returns true if connection less, false otherwise
 312 */
 313static bool tipc_sk_type_connectionless(struct sock *sk)
 314{
 315	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
 316}
 317
 318/* tsk_peer_msg - verify if message was sent by connected port's peer
 319 *
 320 * Handles cases where the node's network address has changed from
 321 * the default of <0.0.0> to its configured setting.
 322 */
 323static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
 324{
 325	struct sock *sk = &tsk->sk;
 326	u32 self = tipc_own_addr(sock_net(sk));
 327	u32 peer_port = tsk_peer_port(tsk);
 328	u32 orig_node, peer_node;
 329
 330	if (unlikely(!tipc_sk_connected(sk)))
 331		return false;
 332
 333	if (unlikely(msg_origport(msg) != peer_port))
 334		return false;
 335
 336	orig_node = msg_orignode(msg);
 337	peer_node = tsk_peer_node(tsk);
 338
 339	if (likely(orig_node == peer_node))
 340		return true;
 341
 342	if (!orig_node && peer_node == self)
 343		return true;
 344
 345	if (!peer_node && orig_node == self)
 346		return true;
 347
 348	return false;
 349}
 350
 351/* tipc_set_sk_state - set the sk_state of the socket
 352 * @sk: socket
 353 *
 354 * Caller must hold socket lock
 355 *
 356 * Returns 0 on success, errno otherwise
 357 */
 358static int tipc_set_sk_state(struct sock *sk, int state)
 359{
 360	int oldsk_state = sk->sk_state;
 361	int res = -EINVAL;
 362
 363	switch (state) {
 364	case TIPC_OPEN:
 365		res = 0;
 366		break;
 367	case TIPC_LISTEN:
 368	case TIPC_CONNECTING:
 369		if (oldsk_state == TIPC_OPEN)
 370			res = 0;
 371		break;
 372	case TIPC_ESTABLISHED:
 373		if (oldsk_state == TIPC_CONNECTING ||
 374		    oldsk_state == TIPC_OPEN)
 375			res = 0;
 376		break;
 377	case TIPC_DISCONNECTING:
 378		if (oldsk_state == TIPC_CONNECTING ||
 379		    oldsk_state == TIPC_ESTABLISHED)
 380			res = 0;
 381		break;
 382	}
 383
 384	if (!res)
 385		sk->sk_state = state;
 386
 387	return res;
 388}
 389
 390static int tipc_sk_sock_err(struct socket *sock, long *timeout)
 391{
 392	struct sock *sk = sock->sk;
 393	int err = sock_error(sk);
 394	int typ = sock->type;
 395
 396	if (err)
 397		return err;
 398	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
 399		if (sk->sk_state == TIPC_DISCONNECTING)
 400			return -EPIPE;
 401		else if (!tipc_sk_connected(sk))
 402			return -ENOTCONN;
 403	}
 404	if (!*timeout)
 405		return -EAGAIN;
 406	if (signal_pending(current))
 407		return sock_intr_errno(*timeout);
 408
 409	return 0;
 410}
 411
 412#define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
 413({                                                                             \
 414	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
 415	struct sock *sk_;						       \
 416	int rc_;							       \
 417									       \
 418	while ((rc_ = !(condition_))) {					       \
 419		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
 420		smp_rmb();                                                     \
 421		sk_ = (sock_)->sk;					       \
 422		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
 423		if (rc_)						       \
 424			break;						       \
 425		add_wait_queue(sk_sleep(sk_), &wait_);                         \
 426		release_sock(sk_);					       \
 427		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
 428		sched_annotate_sleep();				               \
 429		lock_sock(sk_);						       \
 430		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
 431	}								       \
 432	rc_;								       \
 433})
 434
 435/**
 436 * tipc_sk_create - create a TIPC socket
 437 * @net: network namespace (must be default network)
 438 * @sock: pre-allocated socket structure
 439 * @protocol: protocol indicator (must be 0)
 440 * @kern: caused by kernel or by userspace?
 441 *
 442 * This routine creates additional data structures used by the TIPC socket,
 443 * initializes them, and links them together.
 444 *
 445 * Returns 0 on success, errno otherwise
 446 */
 447static int tipc_sk_create(struct net *net, struct socket *sock,
 448			  int protocol, int kern)
 449{
 450	const struct proto_ops *ops;
 451	struct sock *sk;
 452	struct tipc_sock *tsk;
 453	struct tipc_msg *msg;
 454
 455	/* Validate arguments */
 456	if (unlikely(protocol != 0))
 457		return -EPROTONOSUPPORT;
 458
 459	switch (sock->type) {
 460	case SOCK_STREAM:
 461		ops = &stream_ops;
 462		break;
 463	case SOCK_SEQPACKET:
 464		ops = &packet_ops;
 465		break;
 466	case SOCK_DGRAM:
 467	case SOCK_RDM:
 468		ops = &msg_ops;
 469		break;
 470	default:
 471		return -EPROTOTYPE;
 472	}
 473
 474	/* Allocate socket's protocol area */
 475	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
 476	if (sk == NULL)
 477		return -ENOMEM;
 478
 479	tsk = tipc_sk(sk);
 480	tsk->max_pkt = MAX_PKT_DEFAULT;
 481	tsk->maxnagle = 0;
 482	tsk->nagle_start = NAGLE_START_INIT;
 483	INIT_LIST_HEAD(&tsk->publications);
 484	INIT_LIST_HEAD(&tsk->cong_links);
 485	msg = &tsk->phdr;
 486
 487	/* Finish initializing socket data structures */
 488	sock->ops = ops;
 489	sock_init_data(sock, sk);
 490	tipc_set_sk_state(sk, TIPC_OPEN);
 491	if (tipc_sk_insert(tsk)) {
 492		pr_warn("Socket create failed; port number exhausted\n");
 493		return -EINVAL;
 494	}
 495
 496	/* Ensure tsk is visible before we read own_addr. */
 497	smp_mb();
 498
 499	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
 500		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
 501
 502	msg_set_origport(msg, tsk->portid);
 503	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
 504	sk->sk_shutdown = 0;
 505	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
 506	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
 507	sk->sk_data_ready = tipc_data_ready;
 508	sk->sk_write_space = tipc_write_space;
 509	sk->sk_destruct = tipc_sock_destruct;
 510	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
 511	tsk->group_is_open = true;
 512	atomic_set(&tsk->dupl_rcvcnt, 0);
 513
 514	/* Start out with safe limits until we receive an advertised window */
 515	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
 516	tsk->rcv_win = tsk->snd_win;
 517
 518	if (tipc_sk_type_connectionless(sk)) {
 519		tsk_set_unreturnable(tsk, true);
 520		if (sock->type == SOCK_DGRAM)
 521			tsk_set_unreliable(tsk, true);
 522	}
 523	__skb_queue_head_init(&tsk->mc_method.deferredq);
 524	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
 525	return 0;
 526}
 527
 528static void tipc_sk_callback(struct rcu_head *head)
 529{
 530	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
 531
 532	sock_put(&tsk->sk);
 533}
 534
 535/* Caller should hold socket lock for the socket. */
 536static void __tipc_shutdown(struct socket *sock, int error)
 537{
 538	struct sock *sk = sock->sk;
 539	struct tipc_sock *tsk = tipc_sk(sk);
 540	struct net *net = sock_net(sk);
 541	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
 542	u32 dnode = tsk_peer_node(tsk);
 543	struct sk_buff *skb;
 544
 545	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
 546	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
 547					    !tsk_conn_cong(tsk)));
 548
 549	/* Push out delayed messages if in Nagle mode */
 550	tipc_sk_push_backlog(tsk, false);
 551	/* Remove pending SYN */
 552	__skb_queue_purge(&sk->sk_write_queue);
 553
 554	/* Remove partially received buffer if any */
 555	skb = skb_peek(&sk->sk_receive_queue);
 556	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
 557		__skb_unlink(skb, &sk->sk_receive_queue);
 558		kfree_skb(skb);
 
 
 
 
 
 
 
 
 
 559	}
 560
 561	/* Reject all unreceived messages if connectionless */
 562	if (tipc_sk_type_connectionless(sk)) {
 563		tsk_rej_rx_queue(sk, error);
 564		return;
 565	}
 566
 567	switch (sk->sk_state) {
 568	case TIPC_CONNECTING:
 569	case TIPC_ESTABLISHED:
 570		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
 571		tipc_node_remove_conn(net, dnode, tsk->portid);
 572		/* Send a FIN+/- to its peer */
 573		skb = __skb_dequeue(&sk->sk_receive_queue);
 574		if (skb) {
 575			__skb_queue_purge(&sk->sk_receive_queue);
 576			tipc_sk_respond(sk, skb, error);
 577			break;
 578		}
 579		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
 580				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
 581				      tsk_own_node(tsk), tsk_peer_port(tsk),
 582				      tsk->portid, error);
 583		if (skb)
 584			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
 585		break;
 586	case TIPC_LISTEN:
 587		/* Reject all SYN messages */
 588		tsk_rej_rx_queue(sk, error);
 589		break;
 590	default:
 591		__skb_queue_purge(&sk->sk_receive_queue);
 592		break;
 593	}
 594}
 595
 596/**
 597 * tipc_release - destroy a TIPC socket
 598 * @sock: socket to destroy
 599 *
 600 * This routine cleans up any messages that are still queued on the socket.
 601 * For DGRAM and RDM socket types, all queued messages are rejected.
 602 * For SEQPACKET and STREAM socket types, the first message is rejected
 603 * and any others are discarded.  (If the first message on a STREAM socket
 604 * is partially-read, it is discarded and the next one is rejected instead.)
 605 *
 606 * NOTE: Rejected messages are not necessarily returned to the sender!  They
 607 * are returned or discarded according to the "destination droppable" setting
 608 * specified for the message by the sender.
 609 *
 610 * Returns 0 on success, errno otherwise
 611 */
 612static int tipc_release(struct socket *sock)
 613{
 614	struct sock *sk = sock->sk;
 615	struct tipc_sock *tsk;
 616
 617	/*
 618	 * Exit if socket isn't fully initialized (occurs when a failed accept()
 619	 * releases a pre-allocated child socket that was never used)
 620	 */
 621	if (sk == NULL)
 622		return 0;
 623
 624	tsk = tipc_sk(sk);
 625	lock_sock(sk);
 626
 627	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
 628	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
 629	sk->sk_shutdown = SHUTDOWN_MASK;
 630	tipc_sk_leave(tsk);
 631	tipc_sk_withdraw(tsk, 0, NULL);
 632	__skb_queue_purge(&tsk->mc_method.deferredq);
 633	sk_stop_timer(sk, &sk->sk_timer);
 634	tipc_sk_remove(tsk);
 635
 636	sock_orphan(sk);
 637	/* Reject any messages that accumulated in backlog queue */
 638	release_sock(sk);
 639	tipc_dest_list_purge(&tsk->cong_links);
 640	tsk->cong_link_cnt = 0;
 641	call_rcu(&tsk->rcu, tipc_sk_callback);
 642	sock->sk = NULL;
 643
 644	return 0;
 645}
 646
 647/**
 648 * tipc_bind - associate or disassocate TIPC name(s) with a socket
 649 * @sock: socket structure
 650 * @uaddr: socket address describing name(s) and desired operation
 651 * @uaddr_len: size of socket address data structure
 652 *
 653 * Name and name sequence binding is indicated using a positive scope value;
 654 * a negative scope value unbinds the specified name.  Specifying no name
 655 * (i.e. a socket address length of 0) unbinds all names from the socket.
 656 *
 657 * Returns 0 on success, errno otherwise
 658 *
 659 * NOTE: This routine doesn't need to take the socket lock since it doesn't
 660 *       access any non-constant socket information.
 661 */
 662static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
 663		     int uaddr_len)
 664{
 665	struct sock *sk = sock->sk;
 666	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 667	struct tipc_sock *tsk = tipc_sk(sk);
 668	int res = -EINVAL;
 669
 670	lock_sock(sk);
 671	if (unlikely(!uaddr_len)) {
 672		res = tipc_sk_withdraw(tsk, 0, NULL);
 673		goto exit;
 674	}
 675	if (tsk->group) {
 676		res = -EACCES;
 677		goto exit;
 678	}
 679	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
 680		res = -EINVAL;
 681		goto exit;
 682	}
 683	if (addr->family != AF_TIPC) {
 684		res = -EAFNOSUPPORT;
 685		goto exit;
 686	}
 687
 688	if (addr->addrtype == TIPC_ADDR_NAME)
 689		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
 690	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
 691		res = -EAFNOSUPPORT;
 692		goto exit;
 693	}
 694
 695	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
 696	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
 697	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
 698		res = -EACCES;
 699		goto exit;
 700	}
 701
 702	res = (addr->scope >= 0) ?
 703		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
 704		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
 705exit:
 706	release_sock(sk);
 707	return res;
 708}
 709
 710/**
 711 * tipc_getname - get port ID of socket or peer socket
 712 * @sock: socket structure
 713 * @uaddr: area for returned socket address
 
 714 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
 715 *
 716 * Returns 0 on success, errno otherwise
 717 *
 718 * NOTE: This routine doesn't need to take the socket lock since it only
 719 *       accesses socket information that is unchanging (or which changes in
 720 *       a completely predictable manner).
 721 */
 722static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
 723			int peer)
 724{
 725	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
 726	struct sock *sk = sock->sk;
 727	struct tipc_sock *tsk = tipc_sk(sk);
 728
 729	memset(addr, 0, sizeof(*addr));
 730	if (peer) {
 731		if ((!tipc_sk_connected(sk)) &&
 732		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
 733			return -ENOTCONN;
 734		addr->addr.id.ref = tsk_peer_port(tsk);
 735		addr->addr.id.node = tsk_peer_node(tsk);
 736	} else {
 737		addr->addr.id.ref = tsk->portid;
 738		addr->addr.id.node = tipc_own_addr(sock_net(sk));
 739	}
 740
 741	addr->addrtype = TIPC_ADDR_ID;
 742	addr->family = AF_TIPC;
 743	addr->scope = 0;
 744	addr->addr.name.domain = 0;
 745
 746	return sizeof(*addr);
 747}
 748
 749/**
 750 * tipc_poll - read and possibly block on pollmask
 751 * @file: file structure associated with the socket
 752 * @sock: socket for which to calculate the poll bits
 753 * @wait: ???
 754 *
 755 * Returns pollmask value
 756 *
 757 * COMMENTARY:
 758 * It appears that the usual socket locking mechanisms are not useful here
 759 * since the pollmask info is potentially out-of-date the moment this routine
 760 * exits.  TCP and other protocols seem to rely on higher level poll routines
 761 * to handle any preventable race conditions, so TIPC will do the same ...
 762 *
 763 * IMPORTANT: The fact that a read or write operation is indicated does NOT
 764 * imply that the operation will succeed, merely that it should be performed
 765 * and will not block.
 766 */
 767static __poll_t tipc_poll(struct file *file, struct socket *sock,
 768			      poll_table *wait)
 769{
 770	struct sock *sk = sock->sk;
 771	struct tipc_sock *tsk = tipc_sk(sk);
 772	__poll_t revents = 0;
 773
 774	sock_poll_wait(file, sock, wait);
 775	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
 776
 777	if (sk->sk_shutdown & RCV_SHUTDOWN)
 778		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
 779	if (sk->sk_shutdown == SHUTDOWN_MASK)
 780		revents |= EPOLLHUP;
 781
 782	switch (sk->sk_state) {
 783	case TIPC_ESTABLISHED:
 784		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
 785			revents |= EPOLLOUT;
 786		fallthrough;
 787	case TIPC_LISTEN:
 788	case TIPC_CONNECTING:
 789		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
 790			revents |= EPOLLIN | EPOLLRDNORM;
 791		break;
 792	case TIPC_OPEN:
 793		if (tsk->group_is_open && !tsk->cong_link_cnt)
 794			revents |= EPOLLOUT;
 795		if (!tipc_sk_type_connectionless(sk))
 796			break;
 797		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
 798			break;
 799		revents |= EPOLLIN | EPOLLRDNORM;
 800		break;
 801	case TIPC_DISCONNECTING:
 802		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
 803		break;
 804	}
 805	return revents;
 806}
 807
 808/**
 809 * tipc_sendmcast - send multicast message
 810 * @sock: socket structure
 811 * @seq: destination address
 812 * @msg: message to send
 813 * @dlen: length of data to send
 814 * @timeout: timeout to wait for wakeup
 815 *
 816 * Called from function tipc_sendmsg(), which has done all sanity checks
 817 * Returns the number of bytes sent on success, or errno
 818 */
 819static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
 820			  struct msghdr *msg, size_t dlen, long timeout)
 821{
 822	struct sock *sk = sock->sk;
 823	struct tipc_sock *tsk = tipc_sk(sk);
 824	struct tipc_msg *hdr = &tsk->phdr;
 825	struct net *net = sock_net(sk);
 826	int mtu = tipc_bcast_get_mtu(net);
 827	struct tipc_mc_method *method = &tsk->mc_method;
 828	struct sk_buff_head pkts;
 829	struct tipc_nlist dsts;
 830	int rc;
 831
 832	if (tsk->group)
 833		return -EACCES;
 834
 835	/* Block or return if any destination link is congested */
 836	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
 837	if (unlikely(rc))
 838		return rc;
 839
 840	/* Lookup destination nodes */
 841	tipc_nlist_init(&dsts, tipc_own_addr(net));
 842	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
 843				      seq->upper, &dsts);
 844	if (!dsts.local && !dsts.remote)
 845		return -EHOSTUNREACH;
 846
 847	/* Build message header */
 848	msg_set_type(hdr, TIPC_MCAST_MSG);
 849	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
 850	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
 851	msg_set_destport(hdr, 0);
 852	msg_set_destnode(hdr, 0);
 853	msg_set_nametype(hdr, seq->type);
 854	msg_set_namelower(hdr, seq->lower);
 855	msg_set_nameupper(hdr, seq->upper);
 856
 857	/* Build message as chain of buffers */
 858	__skb_queue_head_init(&pkts);
 859	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
 860
 861	/* Send message if build was successful */
 862	if (unlikely(rc == dlen)) {
 863		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
 864					TIPC_DUMP_SK_SNDQ, " ");
 865		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
 866				     &tsk->cong_link_cnt);
 867	}
 868
 869	tipc_nlist_purge(&dsts);
 870
 871	return rc ? rc : dlen;
 872}
 873
 874/**
 875 * tipc_send_group_msg - send a message to a member in the group
 876 * @net: network namespace
 877 * @m: message to send
 878 * @mb: group member
 879 * @dnode: destination node
 880 * @dport: destination port
 881 * @dlen: total length of message data
 882 */
 883static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
 884			       struct msghdr *m, struct tipc_member *mb,
 885			       u32 dnode, u32 dport, int dlen)
 886{
 887	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
 888	struct tipc_mc_method *method = &tsk->mc_method;
 889	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 890	struct tipc_msg *hdr = &tsk->phdr;
 891	struct sk_buff_head pkts;
 892	int mtu, rc;
 893
 894	/* Complete message header */
 895	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
 896	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
 897	msg_set_destport(hdr, dport);
 898	msg_set_destnode(hdr, dnode);
 899	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
 900
 901	/* Build message as chain of buffers */
 902	__skb_queue_head_init(&pkts);
 903	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
 904	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
 905	if (unlikely(rc != dlen))
 906		return rc;
 907
 908	/* Send message */
 909	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
 910	if (unlikely(rc == -ELINKCONG)) {
 911		tipc_dest_push(&tsk->cong_links, dnode, 0);
 912		tsk->cong_link_cnt++;
 913	}
 914
 915	/* Update send window */
 916	tipc_group_update_member(mb, blks);
 917
 918	/* A broadcast sent within next EXPIRE period must follow same path */
 919	method->rcast = true;
 920	method->mandatory = true;
 921	return dlen;
 922}
 923
 924/**
 925 * tipc_send_group_unicast - send message to a member in the group
 926 * @sock: socket structure
 927 * @m: message to send
 928 * @dlen: total length of message data
 929 * @timeout: timeout to wait for wakeup
 930 *
 931 * Called from function tipc_sendmsg(), which has done all sanity checks
 932 * Returns the number of bytes sent on success, or errno
 933 */
 934static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
 935				   int dlen, long timeout)
 936{
 937	struct sock *sk = sock->sk;
 938	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 939	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 940	struct tipc_sock *tsk = tipc_sk(sk);
 941	struct net *net = sock_net(sk);
 942	struct tipc_member *mb = NULL;
 943	u32 node, port;
 944	int rc;
 945
 946	node = dest->addr.id.node;
 947	port = dest->addr.id.ref;
 948	if (!port && !node)
 949		return -EHOSTUNREACH;
 950
 951	/* Block or return if destination link or member is congested */
 952	rc = tipc_wait_for_cond(sock, &timeout,
 953				!tipc_dest_find(&tsk->cong_links, node, 0) &&
 954				tsk->group &&
 955				!tipc_group_cong(tsk->group, node, port, blks,
 956						 &mb));
 957	if (unlikely(rc))
 958		return rc;
 959
 960	if (unlikely(!mb))
 961		return -EHOSTUNREACH;
 962
 963	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
 964
 965	return rc ? rc : dlen;
 966}
 967
 968/**
 969 * tipc_send_group_anycast - send message to any member with given identity
 970 * @sock: socket structure
 971 * @m: message to send
 972 * @dlen: total length of message data
 973 * @timeout: timeout to wait for wakeup
 974 *
 975 * Called from function tipc_sendmsg(), which has done all sanity checks
 976 * Returns the number of bytes sent on success, or errno
 977 */
 978static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
 979				   int dlen, long timeout)
 980{
 981	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
 982	struct sock *sk = sock->sk;
 983	struct tipc_sock *tsk = tipc_sk(sk);
 984	struct list_head *cong_links = &tsk->cong_links;
 985	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
 986	struct tipc_msg *hdr = &tsk->phdr;
 987	struct tipc_member *first = NULL;
 988	struct tipc_member *mbr = NULL;
 989	struct net *net = sock_net(sk);
 990	u32 node, port, exclude;
 991	struct list_head dsts;
 992	u32 type, inst, scope;
 993	int lookups = 0;
 994	int dstcnt, rc;
 995	bool cong;
 996
 997	INIT_LIST_HEAD(&dsts);
 998
 999	type = msg_nametype(hdr);
1000	inst = dest->addr.name.name.instance;
1001	scope = msg_lookup_scope(hdr);
1002
1003	while (++lookups < 4) {
1004		exclude = tipc_group_exclude(tsk->group);
1005
1006		first = NULL;
1007
1008		/* Look for a non-congested destination member, if any */
1009		while (1) {
1010			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1011						 &dstcnt, exclude, false))
1012				return -EHOSTUNREACH;
1013			tipc_dest_pop(&dsts, &node, &port);
1014			cong = tipc_group_cong(tsk->group, node, port, blks,
1015					       &mbr);
1016			if (!cong)
1017				break;
1018			if (mbr == first)
1019				break;
1020			if (!first)
1021				first = mbr;
1022		}
1023
1024		/* Start over if destination was not in member list */
1025		if (unlikely(!mbr))
1026			continue;
1027
1028		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1029			break;
1030
1031		/* Block or return if destination link or member is congested */
1032		rc = tipc_wait_for_cond(sock, &timeout,
1033					!tipc_dest_find(cong_links, node, 0) &&
1034					tsk->group &&
1035					!tipc_group_cong(tsk->group, node, port,
1036							 blks, &mbr));
1037		if (unlikely(rc))
1038			return rc;
1039
1040		/* Send, unless destination disappeared while waiting */
1041		if (likely(mbr))
1042			break;
1043	}
1044
1045	if (unlikely(lookups >= 4))
1046		return -EHOSTUNREACH;
1047
1048	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1049
1050	return rc ? rc : dlen;
1051}
1052
1053/**
1054 * tipc_send_group_bcast - send message to all members in communication group
1055 * @sock: socket structure
1056 * @m: message to send
1057 * @dlen: total length of message data
1058 * @timeout: timeout to wait for wakeup
1059 *
1060 * Called from function tipc_sendmsg(), which has done all sanity checks
1061 * Returns the number of bytes sent on success, or errno
1062 */
1063static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1064				 int dlen, long timeout)
1065{
1066	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1067	struct sock *sk = sock->sk;
1068	struct net *net = sock_net(sk);
1069	struct tipc_sock *tsk = tipc_sk(sk);
1070	struct tipc_nlist *dsts;
1071	struct tipc_mc_method *method = &tsk->mc_method;
1072	bool ack = method->mandatory && method->rcast;
1073	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1074	struct tipc_msg *hdr = &tsk->phdr;
1075	int mtu = tipc_bcast_get_mtu(net);
1076	struct sk_buff_head pkts;
1077	int rc = -EHOSTUNREACH;
1078
1079	/* Block or return if any destination link or member is congested */
1080	rc = tipc_wait_for_cond(sock, &timeout,
1081				!tsk->cong_link_cnt && tsk->group &&
1082				!tipc_group_bc_cong(tsk->group, blks));
1083	if (unlikely(rc))
1084		return rc;
1085
1086	dsts = tipc_group_dests(tsk->group);
1087	if (!dsts->local && !dsts->remote)
1088		return -EHOSTUNREACH;
1089
1090	/* Complete message header */
1091	if (dest) {
1092		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1093		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1094	} else {
1095		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1096		msg_set_nameinst(hdr, 0);
1097	}
1098	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1099	msg_set_destport(hdr, 0);
1100	msg_set_destnode(hdr, 0);
1101	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1102
1103	/* Avoid getting stuck with repeated forced replicasts */
1104	msg_set_grp_bc_ack_req(hdr, ack);
1105
1106	/* Build message as chain of buffers */
1107	__skb_queue_head_init(&pkts);
1108	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1109	if (unlikely(rc != dlen))
1110		return rc;
1111
1112	/* Send message */
1113	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1114	if (unlikely(rc))
1115		return rc;
1116
1117	/* Update broadcast sequence number and send windows */
1118	tipc_group_update_bc_members(tsk->group, blks, ack);
1119
1120	/* Broadcast link is now free to choose method for next broadcast */
1121	method->mandatory = false;
1122	method->expires = jiffies;
1123
1124	return dlen;
1125}
1126
1127/**
1128 * tipc_send_group_mcast - send message to all members with given identity
1129 * @sock: socket structure
1130 * @m: message to send
1131 * @dlen: total length of message data
1132 * @timeout: timeout to wait for wakeup
1133 *
1134 * Called from function tipc_sendmsg(), which has done all sanity checks
1135 * Returns the number of bytes sent on success, or errno
1136 */
1137static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1138				 int dlen, long timeout)
1139{
1140	struct sock *sk = sock->sk;
1141	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1142	struct tipc_sock *tsk = tipc_sk(sk);
1143	struct tipc_group *grp = tsk->group;
1144	struct tipc_msg *hdr = &tsk->phdr;
1145	struct net *net = sock_net(sk);
1146	u32 type, inst, scope, exclude;
1147	struct list_head dsts;
1148	u32 dstcnt;
1149
1150	INIT_LIST_HEAD(&dsts);
1151
1152	type = msg_nametype(hdr);
1153	inst = dest->addr.name.name.instance;
1154	scope = msg_lookup_scope(hdr);
1155	exclude = tipc_group_exclude(grp);
1156
1157	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1158				 &dstcnt, exclude, true))
1159		return -EHOSTUNREACH;
1160
1161	if (dstcnt == 1) {
1162		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1163		return tipc_send_group_unicast(sock, m, dlen, timeout);
1164	}
1165
1166	tipc_dest_list_purge(&dsts);
1167	return tipc_send_group_bcast(sock, m, dlen, timeout);
1168}
1169
1170/**
1171 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1172 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1173 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1174 *
1175 * Multi-threaded: parallel calls with reference to same queues may occur
1176 */
1177void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1178		       struct sk_buff_head *inputq)
1179{
1180	u32 self = tipc_own_addr(net);
1181	u32 type, lower, upper, scope;
1182	struct sk_buff *skb, *_skb;
1183	u32 portid, onode;
1184	struct sk_buff_head tmpq;
1185	struct list_head dports;
1186	struct tipc_msg *hdr;
1187	int user, mtyp, hlen;
1188	bool exact;
1189
1190	__skb_queue_head_init(&tmpq);
1191	INIT_LIST_HEAD(&dports);
1192
1193	skb = tipc_skb_peek(arrvq, &inputq->lock);
1194	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1195		hdr = buf_msg(skb);
1196		user = msg_user(hdr);
1197		mtyp = msg_type(hdr);
1198		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1199		onode = msg_orignode(hdr);
1200		type = msg_nametype(hdr);
1201
1202		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1203			spin_lock_bh(&inputq->lock);
1204			if (skb_peek(arrvq) == skb) {
1205				__skb_dequeue(arrvq);
1206				__skb_queue_tail(inputq, skb);
1207			}
1208			kfree_skb(skb);
1209			spin_unlock_bh(&inputq->lock);
1210			continue;
1211		}
1212
1213		/* Group messages require exact scope match */
1214		if (msg_in_group(hdr)) {
1215			lower = 0;
1216			upper = ~0;
1217			scope = msg_lookup_scope(hdr);
1218			exact = true;
1219		} else {
1220			/* TIPC_NODE_SCOPE means "any scope" in this context */
1221			if (onode == self)
1222				scope = TIPC_NODE_SCOPE;
1223			else
1224				scope = TIPC_CLUSTER_SCOPE;
1225			exact = false;
1226			lower = msg_namelower(hdr);
1227			upper = msg_nameupper(hdr);
1228		}
1229
1230		/* Create destination port list: */
1231		tipc_nametbl_mc_lookup(net, type, lower, upper,
1232				       scope, exact, &dports);
1233
1234		/* Clone message per destination */
1235		while (tipc_dest_pop(&dports, NULL, &portid)) {
1236			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1237			if (_skb) {
1238				msg_set_destport(buf_msg(_skb), portid);
1239				__skb_queue_tail(&tmpq, _skb);
1240				continue;
1241			}
1242			pr_warn("Failed to clone mcast rcv buffer\n");
1243		}
1244		/* Append to inputq if not already done by other thread */
1245		spin_lock_bh(&inputq->lock);
1246		if (skb_peek(arrvq) == skb) {
1247			skb_queue_splice_tail_init(&tmpq, inputq);
1248			kfree_skb(__skb_dequeue(arrvq));
1249		}
1250		spin_unlock_bh(&inputq->lock);
1251		__skb_queue_purge(&tmpq);
1252		kfree_skb(skb);
1253	}
1254	tipc_sk_rcv(net, inputq);
1255}
1256
1257/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1258 *                         when socket is in Nagle mode
1259 */
1260static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1261{
1262	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1263	struct sk_buff *skb = skb_peek_tail(txq);
1264	struct net *net = sock_net(&tsk->sk);
1265	u32 dnode = tsk_peer_node(tsk);
1266	int rc;
1267
1268	if (nagle_ack) {
1269		tsk->pkt_cnt += skb_queue_len(txq);
1270		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1271			tsk->oneway = 0;
1272			if (tsk->nagle_start < NAGLE_START_MAX)
1273				tsk->nagle_start *= 2;
1274			tsk->expect_ack = false;
1275			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1276				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1277				 tsk->nagle_start);
1278		} else {
1279			tsk->nagle_start = NAGLE_START_INIT;
1280			if (skb) {
1281				msg_set_ack_required(buf_msg(skb));
1282				tsk->expect_ack = true;
1283			} else {
1284				tsk->expect_ack = false;
1285			}
1286		}
1287		tsk->msg_acc = 0;
1288		tsk->pkt_cnt = 0;
1289	}
1290
1291	if (!skb || tsk->cong_link_cnt)
1292		return;
1293
1294	/* Do not send SYN again after congestion */
1295	if (msg_is_syn(buf_msg(skb)))
1296		return;
1297
1298	if (tsk->msg_acc)
1299		tsk->pkt_cnt += skb_queue_len(txq);
1300	tsk->snt_unacked += tsk->snd_backlog;
1301	tsk->snd_backlog = 0;
1302	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1303	if (rc == -ELINKCONG)
1304		tsk->cong_link_cnt = 1;
1305}
1306
1307/**
1308 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1309 * @tsk: receiving socket
1310 * @skb: pointer to message buffer.
1311 */
1312static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1313				   struct sk_buff_head *inputq,
1314				   struct sk_buff_head *xmitq)
1315{
1316	struct tipc_msg *hdr = buf_msg(skb);
1317	u32 onode = tsk_own_node(tsk);
1318	struct sock *sk = &tsk->sk;
1319	int mtyp = msg_type(hdr);
1320	bool was_cong;
1321
1322	/* Ignore if connection cannot be validated: */
1323	if (!tsk_peer_msg(tsk, hdr)) {
1324		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1325		goto exit;
1326	}
1327
1328	if (unlikely(msg_errcode(hdr))) {
1329		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1330		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1331				      tsk_peer_port(tsk));
1332		sk->sk_state_change(sk);
1333
1334		/* State change is ignored if socket already awake,
1335		 * - convert msg to abort msg and add to inqueue
1336		 */
1337		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1338		msg_set_type(hdr, TIPC_CONN_MSG);
1339		msg_set_size(hdr, BASIC_H_SIZE);
1340		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1341		__skb_queue_tail(inputq, skb);
1342		return;
1343	}
1344
1345	tsk->probe_unacked = false;
1346
1347	if (mtyp == CONN_PROBE) {
1348		msg_set_type(hdr, CONN_PROBE_REPLY);
1349		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1350			__skb_queue_tail(xmitq, skb);
1351		return;
1352	} else if (mtyp == CONN_ACK) {
1353		was_cong = tsk_conn_cong(tsk);
1354		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1355		tsk->snt_unacked -= msg_conn_ack(hdr);
1356		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1357			tsk->snd_win = msg_adv_win(hdr);
1358		if (was_cong && !tsk_conn_cong(tsk))
1359			sk->sk_write_space(sk);
1360	} else if (mtyp != CONN_PROBE_REPLY) {
1361		pr_warn("Received unknown CONN_PROTO msg\n");
1362	}
1363exit:
1364	kfree_skb(skb);
1365}
1366
1367/**
1368 * tipc_sendmsg - send message in connectionless manner
1369 * @sock: socket structure
1370 * @m: message to send
1371 * @dsz: amount of user data to be sent
1372 *
1373 * Message must have an destination specified explicitly.
1374 * Used for SOCK_RDM and SOCK_DGRAM messages,
1375 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1376 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1377 *
1378 * Returns the number of bytes sent on success, or errno otherwise
1379 */
1380static int tipc_sendmsg(struct socket *sock,
1381			struct msghdr *m, size_t dsz)
1382{
1383	struct sock *sk = sock->sk;
1384	int ret;
1385
1386	lock_sock(sk);
1387	ret = __tipc_sendmsg(sock, m, dsz);
1388	release_sock(sk);
1389
1390	return ret;
1391}
1392
1393static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1394{
1395	struct sock *sk = sock->sk;
1396	struct net *net = sock_net(sk);
1397	struct tipc_sock *tsk = tipc_sk(sk);
1398	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1399	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1400	struct list_head *clinks = &tsk->cong_links;
1401	bool syn = !tipc_sk_type_connectionless(sk);
1402	struct tipc_group *grp = tsk->group;
1403	struct tipc_msg *hdr = &tsk->phdr;
1404	struct tipc_name_seq *seq;
1405	struct sk_buff_head pkts;
1406	u32 dport = 0, dnode = 0;
1407	u32 type = 0, inst = 0;
1408	int mtu, rc;
1409
1410	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1411		return -EMSGSIZE;
1412
1413	if (likely(dest)) {
1414		if (unlikely(m->msg_namelen < sizeof(*dest)))
1415			return -EINVAL;
1416		if (unlikely(dest->family != AF_TIPC))
1417			return -EINVAL;
1418	}
1419
1420	if (grp) {
1421		if (!dest)
1422			return tipc_send_group_bcast(sock, m, dlen, timeout);
1423		if (dest->addrtype == TIPC_ADDR_NAME)
1424			return tipc_send_group_anycast(sock, m, dlen, timeout);
1425		if (dest->addrtype == TIPC_ADDR_ID)
1426			return tipc_send_group_unicast(sock, m, dlen, timeout);
1427		if (dest->addrtype == TIPC_ADDR_MCAST)
1428			return tipc_send_group_mcast(sock, m, dlen, timeout);
1429		return -EINVAL;
1430	}
1431
1432	if (unlikely(!dest)) {
1433		dest = &tsk->peer;
1434		if (!syn && dest->family != AF_TIPC)
1435			return -EDESTADDRREQ;
1436	}
1437
1438	if (unlikely(syn)) {
1439		if (sk->sk_state == TIPC_LISTEN)
1440			return -EPIPE;
1441		if (sk->sk_state != TIPC_OPEN)
1442			return -EISCONN;
1443		if (tsk->published)
1444			return -EOPNOTSUPP;
1445		if (dest->addrtype == TIPC_ADDR_NAME) {
1446			tsk->conn_type = dest->addr.name.name.type;
1447			tsk->conn_instance = dest->addr.name.name.instance;
1448		}
1449		msg_set_syn(hdr, 1);
1450	}
1451
1452	seq = &dest->addr.nameseq;
1453	if (dest->addrtype == TIPC_ADDR_MCAST)
1454		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1455
1456	if (dest->addrtype == TIPC_ADDR_NAME) {
1457		type = dest->addr.name.name.type;
1458		inst = dest->addr.name.name.instance;
1459		dnode = dest->addr.name.domain;
 
 
 
 
 
1460		dport = tipc_nametbl_translate(net, type, inst, &dnode);
 
 
1461		if (unlikely(!dport && !dnode))
1462			return -EHOSTUNREACH;
1463	} else if (dest->addrtype == TIPC_ADDR_ID) {
1464		dnode = dest->addr.id.node;
 
 
 
 
 
1465	} else {
1466		return -EINVAL;
1467	}
1468
1469	/* Block or return if destination link is congested */
1470	rc = tipc_wait_for_cond(sock, &timeout,
1471				!tipc_dest_find(clinks, dnode, 0));
1472	if (unlikely(rc))
1473		return rc;
1474
1475	if (dest->addrtype == TIPC_ADDR_NAME) {
1476		msg_set_type(hdr, TIPC_NAMED_MSG);
1477		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1478		msg_set_nametype(hdr, type);
1479		msg_set_nameinst(hdr, inst);
1480		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1481		msg_set_destnode(hdr, dnode);
1482		msg_set_destport(hdr, dport);
1483	} else { /* TIPC_ADDR_ID */
1484		msg_set_type(hdr, TIPC_DIRECT_MSG);
1485		msg_set_lookup_scope(hdr, 0);
1486		msg_set_destnode(hdr, dnode);
1487		msg_set_destport(hdr, dest->addr.id.ref);
1488		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1489	}
1490
1491	__skb_queue_head_init(&pkts);
1492	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1493	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1494	if (unlikely(rc != dlen))
1495		return rc;
1496	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1497		__skb_queue_purge(&pkts);
1498		return -ENOMEM;
1499	}
1500
1501	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1502	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1503	if (unlikely(rc == -ELINKCONG)) {
1504		tipc_dest_push(clinks, dnode, 0);
1505		tsk->cong_link_cnt++;
1506		rc = 0;
1507	}
1508
1509	if (unlikely(syn && !rc))
1510		tipc_set_sk_state(sk, TIPC_CONNECTING);
1511
1512	return rc ? rc : dlen;
1513}
1514
1515/**
1516 * tipc_sendstream - send stream-oriented data
1517 * @sock: socket structure
1518 * @m: data to send
1519 * @dsz: total length of data to be transmitted
1520 *
1521 * Used for SOCK_STREAM data.
1522 *
1523 * Returns the number of bytes sent on success (or partial success),
1524 * or errno if no data sent
1525 */
1526static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1527{
1528	struct sock *sk = sock->sk;
1529	int ret;
1530
1531	lock_sock(sk);
1532	ret = __tipc_sendstream(sock, m, dsz);
1533	release_sock(sk);
1534
1535	return ret;
1536}
1537
1538static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1539{
1540	struct sock *sk = sock->sk;
1541	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1542	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1543	struct sk_buff_head *txq = &sk->sk_write_queue;
1544	struct tipc_sock *tsk = tipc_sk(sk);
1545	struct tipc_msg *hdr = &tsk->phdr;
1546	struct net *net = sock_net(sk);
1547	struct sk_buff *skb;
1548	u32 dnode = tsk_peer_node(tsk);
1549	int maxnagle = tsk->maxnagle;
1550	int maxpkt = tsk->max_pkt;
1551	int send, sent = 0;
1552	int blocks, rc = 0;
 
 
1553
1554	if (unlikely(dlen > INT_MAX))
1555		return -EMSGSIZE;
1556
1557	/* Handle implicit connection setup */
1558	if (unlikely(dest)) {
1559		rc = __tipc_sendmsg(sock, m, dlen);
1560		if (dlen && dlen == rc) {
1561			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1562			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1563		}
1564		return rc;
1565	}
1566
1567	do {
1568		rc = tipc_wait_for_cond(sock, &timeout,
1569					(!tsk->cong_link_cnt &&
1570					 !tsk_conn_cong(tsk) &&
1571					 tipc_sk_connected(sk)));
1572		if (unlikely(rc))
1573			break;
 
1574		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1575		blocks = tsk->snd_backlog;
1576		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1577		    send <= maxnagle) {
1578			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1579			if (unlikely(rc < 0))
1580				break;
1581			blocks += rc;
1582			tsk->msg_acc++;
1583			if (blocks <= 64 && tsk->expect_ack) {
1584				tsk->snd_backlog = blocks;
1585				sent += send;
1586				break;
1587			} else if (blocks > 64) {
1588				tsk->pkt_cnt += skb_queue_len(txq);
1589			} else {
1590				skb = skb_peek_tail(txq);
1591				if (skb) {
1592					msg_set_ack_required(buf_msg(skb));
1593					tsk->expect_ack = true;
1594				} else {
1595					tsk->expect_ack = false;
1596				}
1597				tsk->msg_acc = 0;
1598				tsk->pkt_cnt = 0;
1599			}
1600		} else {
1601			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1602			if (unlikely(rc != send))
1603				break;
1604			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1605		}
1606		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1607					 TIPC_DUMP_SK_SNDQ, " ");
1608		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1609		if (unlikely(rc == -ELINKCONG)) {
1610			tsk->cong_link_cnt = 1;
1611			rc = 0;
1612		}
1613		if (likely(!rc)) {
1614			tsk->snt_unacked += blocks;
1615			tsk->snd_backlog = 0;
1616			sent += send;
1617		}
1618	} while (sent < dlen && !rc);
1619
1620	return sent ? sent : rc;
1621}
1622
1623/**
1624 * tipc_send_packet - send a connection-oriented message
1625 * @sock: socket structure
1626 * @m: message to send
1627 * @dsz: length of data to be transmitted
1628 *
1629 * Used for SOCK_SEQPACKET messages.
1630 *
1631 * Returns the number of bytes sent on success, or errno otherwise
1632 */
1633static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1634{
1635	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1636		return -EMSGSIZE;
1637
1638	return tipc_sendstream(sock, m, dsz);
1639}
1640
1641/* tipc_sk_finish_conn - complete the setup of a connection
1642 */
1643static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1644				u32 peer_node)
1645{
1646	struct sock *sk = &tsk->sk;
1647	struct net *net = sock_net(sk);
1648	struct tipc_msg *msg = &tsk->phdr;
1649
1650	msg_set_syn(msg, 0);
1651	msg_set_destnode(msg, peer_node);
1652	msg_set_destport(msg, peer_port);
1653	msg_set_type(msg, TIPC_CONN_MSG);
1654	msg_set_lookup_scope(msg, 0);
1655	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1656
1657	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1658	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1659	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1660	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1661	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1662	tsk_set_nagle(tsk);
1663	__skb_queue_purge(&sk->sk_write_queue);
1664	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1665		return;
1666
1667	/* Fall back to message based flow control */
1668	tsk->rcv_win = FLOWCTL_MSG_WIN;
1669	tsk->snd_win = FLOWCTL_MSG_WIN;
1670}
1671
1672/**
1673 * tipc_sk_set_orig_addr - capture sender's address for received message
1674 * @m: descriptor for message info
1675 * @skb: received message
1676 *
1677 * Note: Address is not captured if not requested by receiver.
1678 */
1679static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1680{
1681	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1682	struct tipc_msg *hdr = buf_msg(skb);
1683
1684	if (!srcaddr)
1685		return;
1686
1687	srcaddr->sock.family = AF_TIPC;
1688	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1689	srcaddr->sock.scope = 0;
1690	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1691	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1692	srcaddr->sock.addr.name.domain = 0;
1693	m->msg_namelen = sizeof(struct sockaddr_tipc);
1694
1695	if (!msg_in_group(hdr))
1696		return;
1697
1698	/* Group message users may also want to know sending member's id */
1699	srcaddr->member.family = AF_TIPC;
1700	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1701	srcaddr->member.scope = 0;
1702	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1703	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1704	srcaddr->member.addr.name.domain = 0;
1705	m->msg_namelen = sizeof(*srcaddr);
1706}
1707
1708/**
1709 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1710 * @m: descriptor for message info
1711 * @skb: received message buffer
1712 * @tsk: TIPC port associated with message
1713 *
1714 * Note: Ancillary data is not captured if not requested by receiver.
1715 *
1716 * Returns 0 if successful, otherwise errno
1717 */
1718static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1719				 struct tipc_sock *tsk)
1720{
1721	struct tipc_msg *msg;
1722	u32 anc_data[3];
1723	u32 err;
1724	u32 dest_type;
1725	int has_name;
1726	int res;
1727
1728	if (likely(m->msg_controllen == 0))
1729		return 0;
1730	msg = buf_msg(skb);
1731
1732	/* Optionally capture errored message object(s) */
1733	err = msg ? msg_errcode(msg) : 0;
1734	if (unlikely(err)) {
1735		anc_data[0] = err;
1736		anc_data[1] = msg_data_sz(msg);
1737		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1738		if (res)
1739			return res;
1740		if (anc_data[1]) {
1741			if (skb_linearize(skb))
1742				return -ENOMEM;
1743			msg = buf_msg(skb);
1744			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1745				       msg_data(msg));
1746			if (res)
1747				return res;
1748		}
1749	}
1750
1751	/* Optionally capture message destination object */
1752	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1753	switch (dest_type) {
1754	case TIPC_NAMED_MSG:
1755		has_name = 1;
1756		anc_data[0] = msg_nametype(msg);
1757		anc_data[1] = msg_namelower(msg);
1758		anc_data[2] = msg_namelower(msg);
1759		break;
1760	case TIPC_MCAST_MSG:
1761		has_name = 1;
1762		anc_data[0] = msg_nametype(msg);
1763		anc_data[1] = msg_namelower(msg);
1764		anc_data[2] = msg_nameupper(msg);
1765		break;
1766	case TIPC_CONN_MSG:
1767		has_name = (tsk->conn_type != 0);
1768		anc_data[0] = tsk->conn_type;
1769		anc_data[1] = tsk->conn_instance;
1770		anc_data[2] = tsk->conn_instance;
1771		break;
1772	default:
1773		has_name = 0;
1774	}
1775	if (has_name) {
1776		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1777		if (res)
1778			return res;
1779	}
1780
1781	return 0;
1782}
1783
1784static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1785{
1786	struct sock *sk = &tsk->sk;
 
1787	struct sk_buff *skb = NULL;
1788	struct tipc_msg *msg;
1789	u32 peer_port = tsk_peer_port(tsk);
1790	u32 dnode = tsk_peer_node(tsk);
1791
1792	if (!tipc_sk_connected(sk))
1793		return NULL;
1794	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1795			      dnode, tsk_own_node(tsk), peer_port,
1796			      tsk->portid, TIPC_OK);
1797	if (!skb)
1798		return NULL;
1799	msg = buf_msg(skb);
1800	msg_set_conn_ack(msg, tsk->rcv_unacked);
1801	tsk->rcv_unacked = 0;
1802
1803	/* Adjust to and advertize the correct window limit */
1804	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1805		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1806		msg_set_adv_win(msg, tsk->rcv_win);
1807	}
1808	return skb;
1809}
1810
1811static void tipc_sk_send_ack(struct tipc_sock *tsk)
1812{
1813	struct sk_buff *skb;
1814
1815	skb = tipc_sk_build_ack(tsk);
1816	if (!skb)
1817		return;
1818
1819	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1820			   msg_link_selector(buf_msg(skb)));
1821}
1822
1823static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1824{
1825	struct sock *sk = sock->sk;
1826	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1827	long timeo = *timeop;
1828	int err = sock_error(sk);
1829
1830	if (err)
1831		return err;
1832
1833	for (;;) {
1834		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1835			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1836				err = -ENOTCONN;
1837				break;
1838			}
1839			add_wait_queue(sk_sleep(sk), &wait);
1840			release_sock(sk);
1841			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1842			sched_annotate_sleep();
1843			lock_sock(sk);
1844			remove_wait_queue(sk_sleep(sk), &wait);
1845		}
1846		err = 0;
1847		if (!skb_queue_empty(&sk->sk_receive_queue))
1848			break;
1849		err = -EAGAIN;
1850		if (!timeo)
1851			break;
1852		err = sock_intr_errno(timeo);
1853		if (signal_pending(current))
1854			break;
1855
1856		err = sock_error(sk);
1857		if (err)
1858			break;
1859	}
1860	*timeop = timeo;
1861	return err;
1862}
1863
1864/**
1865 * tipc_recvmsg - receive packet-oriented message
1866 * @m: descriptor for message info
1867 * @buflen: length of user buffer area
1868 * @flags: receive flags
1869 *
1870 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1871 * If the complete message doesn't fit in user area, truncate it.
1872 *
1873 * Returns size of returned message data, errno otherwise
1874 */
1875static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1876			size_t buflen,	int flags)
1877{
1878	struct sock *sk = sock->sk;
1879	bool connected = !tipc_sk_type_connectionless(sk);
1880	struct tipc_sock *tsk = tipc_sk(sk);
1881	int rc, err, hlen, dlen, copy;
1882	struct sk_buff_head xmitq;
1883	struct tipc_msg *hdr;
1884	struct sk_buff *skb;
1885	bool grp_evt;
1886	long timeout;
1887
1888	/* Catch invalid receive requests */
1889	if (unlikely(!buflen))
1890		return -EINVAL;
1891
1892	lock_sock(sk);
1893	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1894		rc = -ENOTCONN;
1895		goto exit;
1896	}
1897	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1898
1899	/* Step rcv queue to first msg with data or error; wait if necessary */
1900	do {
1901		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1902		if (unlikely(rc))
1903			goto exit;
1904		skb = skb_peek(&sk->sk_receive_queue);
1905		hdr = buf_msg(skb);
1906		dlen = msg_data_sz(hdr);
1907		hlen = msg_hdr_sz(hdr);
1908		err = msg_errcode(hdr);
1909		grp_evt = msg_is_grp_evt(hdr);
1910		if (likely(dlen || err))
1911			break;
1912		tsk_advance_rx_queue(sk);
1913	} while (1);
1914
1915	/* Collect msg meta data, including error code and rejected data */
1916	tipc_sk_set_orig_addr(m, skb);
1917	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1918	if (unlikely(rc))
1919		goto exit;
1920	hdr = buf_msg(skb);
1921
1922	/* Capture data if non-error msg, otherwise just set return value */
1923	if (likely(!err)) {
1924		copy = min_t(int, dlen, buflen);
1925		if (unlikely(copy != dlen))
1926			m->msg_flags |= MSG_TRUNC;
1927		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1928	} else {
1929		copy = 0;
1930		rc = 0;
1931		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1932			rc = -ECONNRESET;
1933	}
1934	if (unlikely(rc))
1935		goto exit;
1936
1937	/* Mark message as group event if applicable */
1938	if (unlikely(grp_evt)) {
1939		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1940			m->msg_flags |= MSG_EOR;
1941		m->msg_flags |= MSG_OOB;
1942		copy = 0;
1943	}
1944
1945	/* Caption of data or error code/rejected data was successful */
1946	if (unlikely(flags & MSG_PEEK))
1947		goto exit;
1948
1949	/* Send group flow control advertisement when applicable */
1950	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1951		__skb_queue_head_init(&xmitq);
1952		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1953					  msg_orignode(hdr), msg_origport(hdr),
1954					  &xmitq);
1955		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1956	}
1957
1958	tsk_advance_rx_queue(sk);
1959
1960	if (likely(!connected))
1961		goto exit;
1962
1963	/* Send connection flow control advertisement when applicable */
1964	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1965	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1966		tipc_sk_send_ack(tsk);
1967exit:
1968	release_sock(sk);
1969	return rc ? rc : copy;
1970}
1971
1972/**
1973 * tipc_recvstream - receive stream-oriented data
1974 * @m: descriptor for message info
1975 * @buflen: total size of user buffer area
1976 * @flags: receive flags
1977 *
1978 * Used for SOCK_STREAM messages only.  If not enough data is available
1979 * will optionally wait for more; never truncates data.
1980 *
1981 * Returns size of returned message data, errno otherwise
1982 */
1983static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1984			   size_t buflen, int flags)
1985{
1986	struct sock *sk = sock->sk;
1987	struct tipc_sock *tsk = tipc_sk(sk);
1988	struct sk_buff *skb;
1989	struct tipc_msg *hdr;
1990	struct tipc_skb_cb *skb_cb;
1991	bool peek = flags & MSG_PEEK;
1992	int offset, required, copy, copied = 0;
1993	int hlen, dlen, err, rc;
1994	long timeout;
1995
1996	/* Catch invalid receive attempts */
1997	if (unlikely(!buflen))
1998		return -EINVAL;
1999
2000	lock_sock(sk);
2001
2002	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2003		rc = -ENOTCONN;
2004		goto exit;
2005	}
2006	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2007	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2008
2009	do {
2010		/* Look at first msg in receive queue; wait if necessary */
2011		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2012		if (unlikely(rc))
2013			break;
2014		skb = skb_peek(&sk->sk_receive_queue);
2015		skb_cb = TIPC_SKB_CB(skb);
2016		hdr = buf_msg(skb);
2017		dlen = msg_data_sz(hdr);
2018		hlen = msg_hdr_sz(hdr);
2019		err = msg_errcode(hdr);
2020
2021		/* Discard any empty non-errored (SYN-) message */
2022		if (unlikely(!dlen && !err)) {
2023			tsk_advance_rx_queue(sk);
2024			continue;
2025		}
2026
2027		/* Collect msg meta data, incl. error code and rejected data */
2028		if (!copied) {
2029			tipc_sk_set_orig_addr(m, skb);
2030			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2031			if (rc)
2032				break;
2033			hdr = buf_msg(skb);
2034		}
2035
2036		/* Copy data if msg ok, otherwise return error/partial data */
2037		if (likely(!err)) {
2038			offset = skb_cb->bytes_read;
2039			copy = min_t(int, dlen - offset, buflen - copied);
2040			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2041			if (unlikely(rc))
2042				break;
2043			copied += copy;
2044			offset += copy;
2045			if (unlikely(offset < dlen)) {
2046				if (!peek)
2047					skb_cb->bytes_read = offset;
2048				break;
2049			}
2050		} else {
2051			rc = 0;
2052			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2053				rc = -ECONNRESET;
2054			if (copied || rc)
2055				break;
2056		}
2057
2058		if (unlikely(peek))
2059			break;
2060
2061		tsk_advance_rx_queue(sk);
2062
2063		/* Send connection flow control advertisement when applicable */
2064		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2065		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2066			tipc_sk_send_ack(tsk);
2067
2068		/* Exit if all requested data or FIN/error received */
2069		if (copied == buflen || err)
2070			break;
2071
2072	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2073exit:
2074	release_sock(sk);
2075	return copied ? copied : rc;
2076}
2077
2078/**
2079 * tipc_write_space - wake up thread if port congestion is released
2080 * @sk: socket
2081 */
2082static void tipc_write_space(struct sock *sk)
2083{
2084	struct socket_wq *wq;
2085
2086	rcu_read_lock();
2087	wq = rcu_dereference(sk->sk_wq);
2088	if (skwq_has_sleeper(wq))
2089		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2090						EPOLLWRNORM | EPOLLWRBAND);
2091	rcu_read_unlock();
2092}
2093
2094/**
2095 * tipc_data_ready - wake up threads to indicate messages have been received
2096 * @sk: socket
 
2097 */
2098static void tipc_data_ready(struct sock *sk)
2099{
2100	struct socket_wq *wq;
2101
2102	rcu_read_lock();
2103	wq = rcu_dereference(sk->sk_wq);
2104	if (skwq_has_sleeper(wq))
2105		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2106						EPOLLRDNORM | EPOLLRDBAND);
2107	rcu_read_unlock();
2108}
2109
2110static void tipc_sock_destruct(struct sock *sk)
2111{
2112	__skb_queue_purge(&sk->sk_receive_queue);
2113}
2114
2115static void tipc_sk_proto_rcv(struct sock *sk,
2116			      struct sk_buff_head *inputq,
2117			      struct sk_buff_head *xmitq)
2118{
2119	struct sk_buff *skb = __skb_dequeue(inputq);
2120	struct tipc_sock *tsk = tipc_sk(sk);
2121	struct tipc_msg *hdr = buf_msg(skb);
2122	struct tipc_group *grp = tsk->group;
2123	bool wakeup = false;
2124
2125	switch (msg_user(hdr)) {
2126	case CONN_MANAGER:
2127		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2128		return;
2129	case SOCK_WAKEUP:
2130		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2131		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2132		smp_wmb();
2133		tsk->cong_link_cnt--;
2134		wakeup = true;
2135		tipc_sk_push_backlog(tsk, false);
2136		break;
2137	case GROUP_PROTOCOL:
2138		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2139		break;
2140	case TOP_SRV:
2141		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2142				      hdr, inputq, xmitq);
2143		break;
2144	default:
2145		break;
2146	}
2147
2148	if (wakeup)
2149		sk->sk_write_space(sk);
2150
2151	kfree_skb(skb);
2152}
2153
2154/**
2155 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2156 * @tsk: TIPC socket
2157 * @skb: pointer to message buffer.
2158 * @xmitq: for Nagle ACK if any
2159 * Returns true if message should be added to receive queue, false otherwise
2160 */
2161static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2162				   struct sk_buff_head *xmitq)
2163{
2164	struct sock *sk = &tsk->sk;
2165	struct net *net = sock_net(sk);
2166	struct tipc_msg *hdr = buf_msg(skb);
2167	bool con_msg = msg_connected(hdr);
2168	u32 pport = tsk_peer_port(tsk);
2169	u32 pnode = tsk_peer_node(tsk);
2170	u32 oport = msg_origport(hdr);
2171	u32 onode = msg_orignode(hdr);
2172	int err = msg_errcode(hdr);
2173	unsigned long delay;
2174
2175	if (unlikely(msg_mcast(hdr)))
2176		return false;
2177	tsk->oneway = 0;
2178
2179	switch (sk->sk_state) {
2180	case TIPC_CONNECTING:
2181		/* Setup ACK */
2182		if (likely(con_msg)) {
2183			if (err)
2184				break;
2185			tipc_sk_finish_conn(tsk, oport, onode);
2186			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2187			/* ACK+ message with data is added to receive queue */
2188			if (msg_data_sz(hdr))
2189				return true;
2190			/* Empty ACK-, - wake up sleeping connect() and drop */
2191			sk->sk_state_change(sk);
2192			msg_set_dest_droppable(hdr, 1);
2193			return false;
2194		}
2195		/* Ignore connectionless message if not from listening socket */
2196		if (oport != pport || onode != pnode)
2197			return false;
2198
2199		/* Rejected SYN */
2200		if (err != TIPC_ERR_OVERLOAD)
2201			break;
2202
2203		/* Prepare for new setup attempt if we have a SYN clone */
2204		if (skb_queue_empty(&sk->sk_write_queue))
2205			break;
2206		get_random_bytes(&delay, 2);
2207		delay %= (tsk->conn_timeout / 4);
2208		delay = msecs_to_jiffies(delay + 100);
2209		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2210		return false;
2211	case TIPC_OPEN:
2212	case TIPC_DISCONNECTING:
2213		return false;
2214	case TIPC_LISTEN:
2215		/* Accept only SYN message */
2216		if (!msg_is_syn(hdr) &&
2217		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2218			return false;
2219		if (!con_msg && !err)
2220			return true;
2221		return false;
2222	case TIPC_ESTABLISHED:
2223		if (!skb_queue_empty(&sk->sk_write_queue))
2224			tipc_sk_push_backlog(tsk, false);
2225		/* Accept only connection-based messages sent by peer */
2226		if (likely(con_msg && !err && pport == oport &&
2227			   pnode == onode)) {
2228			if (msg_ack_required(hdr)) {
2229				struct sk_buff *skb;
2230
2231				skb = tipc_sk_build_ack(tsk);
2232				if (skb) {
2233					msg_set_nagle_ack(buf_msg(skb));
2234					__skb_queue_tail(xmitq, skb);
2235				}
2236			}
2237			return true;
2238		}
2239		if (!tsk_peer_msg(tsk, hdr))
2240			return false;
2241		if (!err)
2242			return true;
2243		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2244		tipc_node_remove_conn(net, pnode, tsk->portid);
2245		sk->sk_state_change(sk);
2246		return true;
2247	default:
2248		pr_err("Unknown sk_state %u\n", sk->sk_state);
2249	}
2250	/* Abort connection setup attempt */
2251	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2252	sk->sk_err = ECONNREFUSED;
2253	sk->sk_state_change(sk);
2254	return true;
2255}
2256
2257/**
2258 * rcvbuf_limit - get proper overload limit of socket receive queue
2259 * @sk: socket
2260 * @skb: message
2261 *
2262 * For connection oriented messages, irrespective of importance,
2263 * default queue limit is 2 MB.
2264 *
2265 * For connectionless messages, queue limits are based on message
2266 * importance as follows:
2267 *
2268 * TIPC_LOW_IMPORTANCE       (2 MB)
2269 * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2270 * TIPC_HIGH_IMPORTANCE      (8 MB)
2271 * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2272 *
2273 * Returns overload limit according to corresponding message importance
2274 */
2275static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2276{
2277	struct tipc_sock *tsk = tipc_sk(sk);
2278	struct tipc_msg *hdr = buf_msg(skb);
2279
2280	if (unlikely(msg_in_group(hdr)))
2281		return READ_ONCE(sk->sk_rcvbuf);
2282
2283	if (unlikely(!msg_connected(hdr)))
2284		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2285
2286	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2287		return READ_ONCE(sk->sk_rcvbuf);
2288
2289	return FLOWCTL_MSG_LIM;
2290}
2291
2292/**
2293 * tipc_sk_filter_rcv - validate incoming message
2294 * @sk: socket
2295 * @skb: pointer to message.
2296 *
2297 * Enqueues message on receive queue if acceptable; optionally handles
2298 * disconnect indication for a connected socket.
2299 *
2300 * Called with socket lock already taken
2301 *
2302 */
2303static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2304			       struct sk_buff_head *xmitq)
2305{
2306	bool sk_conn = !tipc_sk_type_connectionless(sk);
2307	struct tipc_sock *tsk = tipc_sk(sk);
2308	struct tipc_group *grp = tsk->group;
2309	struct tipc_msg *hdr = buf_msg(skb);
2310	struct net *net = sock_net(sk);
2311	struct sk_buff_head inputq;
2312	int mtyp = msg_type(hdr);
2313	int limit, err = TIPC_OK;
2314
2315	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2316	TIPC_SKB_CB(skb)->bytes_read = 0;
2317	__skb_queue_head_init(&inputq);
2318	__skb_queue_tail(&inputq, skb);
2319
2320	if (unlikely(!msg_isdata(hdr)))
2321		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2322
2323	if (unlikely(grp))
2324		tipc_group_filter_msg(grp, &inputq, xmitq);
2325
2326	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2327		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2328
2329	/* Validate and add to receive buffer if there is space */
2330	while ((skb = __skb_dequeue(&inputq))) {
2331		hdr = buf_msg(skb);
2332		limit = rcvbuf_limit(sk, skb);
2333		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2334		    (!sk_conn && msg_connected(hdr)) ||
2335		    (!grp && msg_in_group(hdr)))
2336			err = TIPC_ERR_NO_PORT;
2337		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2338			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2339					   "err_overload2!");
2340			atomic_inc(&sk->sk_drops);
2341			err = TIPC_ERR_OVERLOAD;
2342		}
2343
2344		if (unlikely(err)) {
2345			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2346				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2347						      "@filter_rcv!");
2348				__skb_queue_tail(xmitq, skb);
2349			}
2350			err = TIPC_OK;
2351			continue;
2352		}
2353		__skb_queue_tail(&sk->sk_receive_queue, skb);
2354		skb_set_owner_r(skb, sk);
2355		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2356					 "rcvq >90% allocated!");
2357		sk->sk_data_ready(sk);
2358	}
2359}
2360
2361/**
2362 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2363 * @sk: socket
2364 * @skb: message
2365 *
2366 * Caller must hold socket lock
2367 */
2368static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2369{
2370	unsigned int before = sk_rmem_alloc_get(sk);
2371	struct sk_buff_head xmitq;
2372	unsigned int added;
2373
2374	__skb_queue_head_init(&xmitq);
2375
2376	tipc_sk_filter_rcv(sk, skb, &xmitq);
2377	added = sk_rmem_alloc_get(sk) - before;
2378	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2379
2380	/* Send pending response/rejected messages, if any */
2381	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2382	return 0;
2383}
2384
2385/**
2386 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2387 *                   inputq and try adding them to socket or backlog queue
2388 * @inputq: list of incoming buffers with potentially different destinations
2389 * @sk: socket where the buffers should be enqueued
2390 * @dport: port number for the socket
2391 *
2392 * Caller must hold socket lock
2393 */
2394static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2395			    u32 dport, struct sk_buff_head *xmitq)
2396{
2397	unsigned long time_limit = jiffies + 2;
2398	struct sk_buff *skb;
2399	unsigned int lim;
2400	atomic_t *dcnt;
2401	u32 onode;
2402
2403	while (skb_queue_len(inputq)) {
2404		if (unlikely(time_after_eq(jiffies, time_limit)))
2405			return;
2406
2407		skb = tipc_skb_dequeue(inputq, dport);
2408		if (unlikely(!skb))
2409			return;
2410
2411		/* Add message directly to receive queue if possible */
2412		if (!sock_owned_by_user(sk)) {
2413			tipc_sk_filter_rcv(sk, skb, xmitq);
2414			continue;
2415		}
2416
2417		/* Try backlog, compensating for double-counted bytes */
2418		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2419		if (!sk->sk_backlog.len)
2420			atomic_set(dcnt, 0);
2421		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2422		if (likely(!sk_add_backlog(sk, skb, lim))) {
2423			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2424						 "bklg & rcvq >90% allocated!");
2425			continue;
2426		}
2427
2428		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2429		/* Overload => reject message back to sender */
2430		onode = tipc_own_addr(sock_net(sk));
2431		atomic_inc(&sk->sk_drops);
2432		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2433			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2434					      "@sk_enqueue!");
2435			__skb_queue_tail(xmitq, skb);
2436		}
2437		break;
2438	}
2439}
2440
2441/**
2442 * tipc_sk_rcv - handle a chain of incoming buffers
2443 * @inputq: buffer list containing the buffers
2444 * Consumes all buffers in list until inputq is empty
2445 * Note: may be called in multiple threads referring to the same queue
2446 */
2447void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2448{
2449	struct sk_buff_head xmitq;
2450	u32 dnode, dport = 0;
2451	int err;
2452	struct tipc_sock *tsk;
2453	struct sock *sk;
2454	struct sk_buff *skb;
2455
2456	__skb_queue_head_init(&xmitq);
2457	while (skb_queue_len(inputq)) {
2458		dport = tipc_skb_peek_port(inputq, dport);
2459		tsk = tipc_sk_lookup(net, dport);
2460
2461		if (likely(tsk)) {
2462			sk = &tsk->sk;
2463			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2464				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2465				spin_unlock_bh(&sk->sk_lock.slock);
2466			}
2467			/* Send pending response/rejected messages, if any */
2468			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2469			sock_put(sk);
2470			continue;
2471		}
2472		/* No destination socket => dequeue skb if still there */
2473		skb = tipc_skb_dequeue(inputq, dport);
2474		if (!skb)
2475			return;
2476
2477		/* Try secondary lookup if unresolved named message */
2478		err = TIPC_ERR_NO_PORT;
2479		if (tipc_msg_lookup_dest(net, skb, &err))
2480			goto xmit;
2481
2482		/* Prepare for message rejection */
2483		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2484			continue;
2485
2486		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2487xmit:
2488		dnode = msg_destnode(buf_msg(skb));
2489		tipc_node_xmit_skb(net, skb, dnode, dport);
2490	}
2491}
2492
2493static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2494{
2495	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2496	struct sock *sk = sock->sk;
2497	int done;
2498
2499	do {
2500		int err = sock_error(sk);
2501		if (err)
2502			return err;
2503		if (!*timeo_p)
2504			return -ETIMEDOUT;
2505		if (signal_pending(current))
2506			return sock_intr_errno(*timeo_p);
2507		if (sk->sk_state == TIPC_DISCONNECTING)
2508			break;
2509
2510		add_wait_queue(sk_sleep(sk), &wait);
2511		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2512				     &wait);
2513		remove_wait_queue(sk_sleep(sk), &wait);
2514	} while (!done);
2515	return 0;
2516}
2517
2518static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2519{
2520	if (addr->family != AF_TIPC)
2521		return false;
2522	if (addr->addrtype == TIPC_SERVICE_RANGE)
2523		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2524	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2525		addr->addrtype == TIPC_SOCKET_ADDR);
2526}
2527
2528/**
2529 * tipc_connect - establish a connection to another TIPC port
2530 * @sock: socket structure
2531 * @dest: socket address for destination port
2532 * @destlen: size of socket address data structure
2533 * @flags: file-related flags associated with socket
2534 *
2535 * Returns 0 on success, errno otherwise
2536 */
2537static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2538			int destlen, int flags)
2539{
2540	struct sock *sk = sock->sk;
2541	struct tipc_sock *tsk = tipc_sk(sk);
2542	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2543	struct msghdr m = {NULL,};
2544	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2545	int previous;
2546	int res = 0;
2547
2548	if (destlen != sizeof(struct sockaddr_tipc))
2549		return -EINVAL;
2550
2551	lock_sock(sk);
2552
2553	if (tsk->group) {
2554		res = -EINVAL;
2555		goto exit;
2556	}
2557
2558	if (dst->family == AF_UNSPEC) {
2559		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2560		if (!tipc_sk_type_connectionless(sk))
2561			res = -EINVAL;
2562		goto exit;
2563	}
2564	if (!tipc_sockaddr_is_sane(dst)) {
2565		res = -EINVAL;
2566		goto exit;
2567	}
2568	/* DGRAM/RDM connect(), just save the destaddr */
2569	if (tipc_sk_type_connectionless(sk)) {
2570		memcpy(&tsk->peer, dest, destlen);
2571		goto exit;
2572	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2573		res = -EINVAL;
2574		goto exit;
2575	}
2576
2577	previous = sk->sk_state;
2578
2579	switch (sk->sk_state) {
2580	case TIPC_OPEN:
2581		/* Send a 'SYN-' to destination */
2582		m.msg_name = dest;
2583		m.msg_namelen = destlen;
2584
2585		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2586		 * indicate send_msg() is never blocked.
2587		 */
2588		if (!timeout)
2589			m.msg_flags = MSG_DONTWAIT;
2590
2591		res = __tipc_sendmsg(sock, &m, 0);
2592		if ((res < 0) && (res != -EWOULDBLOCK))
2593			goto exit;
2594
2595		/* Just entered TIPC_CONNECTING state; the only
2596		 * difference is that return value in non-blocking
2597		 * case is EINPROGRESS, rather than EALREADY.
2598		 */
2599		res = -EINPROGRESS;
2600		fallthrough;
2601	case TIPC_CONNECTING:
2602		if (!timeout) {
2603			if (previous == TIPC_CONNECTING)
2604				res = -EALREADY;
2605			goto exit;
2606		}
2607		timeout = msecs_to_jiffies(timeout);
2608		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2609		res = tipc_wait_for_connect(sock, &timeout);
2610		break;
2611	case TIPC_ESTABLISHED:
2612		res = -EISCONN;
2613		break;
2614	default:
2615		res = -EINVAL;
2616	}
2617
2618exit:
2619	release_sock(sk);
2620	return res;
2621}
2622
2623/**
2624 * tipc_listen - allow socket to listen for incoming connections
2625 * @sock: socket structure
2626 * @len: (unused)
2627 *
2628 * Returns 0 on success, errno otherwise
2629 */
2630static int tipc_listen(struct socket *sock, int len)
2631{
2632	struct sock *sk = sock->sk;
2633	int res;
2634
2635	lock_sock(sk);
2636	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2637	release_sock(sk);
2638
2639	return res;
2640}
2641
2642static int tipc_wait_for_accept(struct socket *sock, long timeo)
2643{
2644	struct sock *sk = sock->sk;
2645	DEFINE_WAIT(wait);
2646	int err;
2647
2648	/* True wake-one mechanism for incoming connections: only
2649	 * one process gets woken up, not the 'whole herd'.
2650	 * Since we do not 'race & poll' for established sockets
2651	 * anymore, the common case will execute the loop only once.
2652	*/
2653	for (;;) {
2654		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2655					  TASK_INTERRUPTIBLE);
2656		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2657			release_sock(sk);
2658			timeo = schedule_timeout(timeo);
2659			lock_sock(sk);
2660		}
2661		err = 0;
2662		if (!skb_queue_empty(&sk->sk_receive_queue))
2663			break;
2664		err = -EAGAIN;
2665		if (!timeo)
2666			break;
2667		err = sock_intr_errno(timeo);
2668		if (signal_pending(current))
2669			break;
2670	}
2671	finish_wait(sk_sleep(sk), &wait);
2672	return err;
2673}
2674
2675/**
2676 * tipc_accept - wait for connection request
2677 * @sock: listening socket
2678 * @new_sock: new socket that is to be connected
2679 * @flags: file-related flags associated with socket
2680 *
2681 * Returns 0 on success, errno otherwise
2682 */
2683static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2684		       bool kern)
2685{
2686	struct sock *new_sk, *sk = sock->sk;
2687	struct sk_buff *buf;
2688	struct tipc_sock *new_tsock;
2689	struct tipc_msg *msg;
2690	long timeo;
2691	int res;
2692
2693	lock_sock(sk);
2694
2695	if (sk->sk_state != TIPC_LISTEN) {
2696		res = -EINVAL;
2697		goto exit;
2698	}
2699	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2700	res = tipc_wait_for_accept(sock, timeo);
2701	if (res)
2702		goto exit;
2703
2704	buf = skb_peek(&sk->sk_receive_queue);
2705
2706	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2707	if (res)
2708		goto exit;
2709	security_sk_clone(sock->sk, new_sock->sk);
2710
2711	new_sk = new_sock->sk;
2712	new_tsock = tipc_sk(new_sk);
2713	msg = buf_msg(buf);
2714
2715	/* we lock on new_sk; but lockdep sees the lock on sk */
2716	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2717
2718	/*
2719	 * Reject any stray messages received by new socket
2720	 * before the socket lock was taken (very, very unlikely)
2721	 */
2722	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2723
2724	/* Connect new socket to it's peer */
2725	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2726
2727	tsk_set_importance(new_sk, msg_importance(msg));
2728	if (msg_named(msg)) {
2729		new_tsock->conn_type = msg_nametype(msg);
2730		new_tsock->conn_instance = msg_nameinst(msg);
2731	}
2732
2733	/*
2734	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2735	 * Respond to 'SYN+' by queuing it on new socket.
2736	 */
2737	if (!msg_data_sz(msg)) {
2738		struct msghdr m = {NULL,};
2739
2740		tsk_advance_rx_queue(sk);
2741		__tipc_sendstream(new_sock, &m, 0);
2742	} else {
2743		__skb_dequeue(&sk->sk_receive_queue);
2744		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2745		skb_set_owner_r(buf, new_sk);
2746	}
2747	release_sock(new_sk);
2748exit:
2749	release_sock(sk);
2750	return res;
2751}
2752
2753/**
2754 * tipc_shutdown - shutdown socket connection
2755 * @sock: socket structure
2756 * @how: direction to close (must be SHUT_RDWR)
2757 *
2758 * Terminates connection (if necessary), then purges socket's receive queue.
2759 *
2760 * Returns 0 on success, errno otherwise
2761 */
2762static int tipc_shutdown(struct socket *sock, int how)
2763{
2764	struct sock *sk = sock->sk;
2765	int res;
2766
2767	if (how != SHUT_RDWR)
2768		return -EINVAL;
2769
2770	lock_sock(sk);
2771
2772	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2773	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2774	sk->sk_shutdown = SHUTDOWN_MASK;
2775
2776	if (sk->sk_state == TIPC_DISCONNECTING) {
2777		/* Discard any unreceived messages */
2778		__skb_queue_purge(&sk->sk_receive_queue);
2779
 
 
2780		res = 0;
2781	} else {
2782		res = -ENOTCONN;
2783	}
2784	/* Wake up anyone sleeping in poll. */
2785	sk->sk_state_change(sk);
2786
2787	release_sock(sk);
2788	return res;
2789}
2790
2791static void tipc_sk_check_probing_state(struct sock *sk,
2792					struct sk_buff_head *list)
2793{
2794	struct tipc_sock *tsk = tipc_sk(sk);
2795	u32 pnode = tsk_peer_node(tsk);
2796	u32 pport = tsk_peer_port(tsk);
2797	u32 self = tsk_own_node(tsk);
2798	u32 oport = tsk->portid;
2799	struct sk_buff *skb;
2800
2801	if (tsk->probe_unacked) {
2802		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2803		sk->sk_err = ECONNABORTED;
2804		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2805		sk->sk_state_change(sk);
2806		return;
2807	}
2808	/* Prepare new probe */
2809	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2810			      pnode, self, pport, oport, TIPC_OK);
2811	if (skb)
2812		__skb_queue_tail(list, skb);
2813	tsk->probe_unacked = true;
2814	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2815}
2816
2817static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2818{
2819	struct tipc_sock *tsk = tipc_sk(sk);
2820
2821	/* Try again later if dest link is congested */
2822	if (tsk->cong_link_cnt) {
2823		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2824		return;
2825	}
2826	/* Prepare SYN for retransmit */
2827	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2828}
2829
2830static void tipc_sk_timeout(struct timer_list *t)
2831{
2832	struct sock *sk = from_timer(sk, t, sk_timer);
2833	struct tipc_sock *tsk = tipc_sk(sk);
2834	u32 pnode = tsk_peer_node(tsk);
2835	struct sk_buff_head list;
2836	int rc = 0;
2837
2838	__skb_queue_head_init(&list);
2839	bh_lock_sock(sk);
2840
2841	/* Try again later if socket is busy */
2842	if (sock_owned_by_user(sk)) {
2843		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2844		bh_unlock_sock(sk);
2845		sock_put(sk);
2846		return;
2847	}
2848
2849	if (sk->sk_state == TIPC_ESTABLISHED)
2850		tipc_sk_check_probing_state(sk, &list);
2851	else if (sk->sk_state == TIPC_CONNECTING)
2852		tipc_sk_retry_connect(sk, &list);
2853
2854	bh_unlock_sock(sk);
2855
2856	if (!skb_queue_empty(&list))
2857		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2858
2859	/* SYN messages may cause link congestion */
2860	if (rc == -ELINKCONG) {
2861		tipc_dest_push(&tsk->cong_links, pnode, 0);
2862		tsk->cong_link_cnt = 1;
2863	}
2864	sock_put(sk);
2865}
2866
2867static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2868			   struct tipc_name_seq const *seq)
2869{
2870	struct sock *sk = &tsk->sk;
2871	struct net *net = sock_net(sk);
2872	struct publication *publ;
2873	u32 key;
2874
2875	if (scope != TIPC_NODE_SCOPE)
2876		scope = TIPC_CLUSTER_SCOPE;
2877
2878	if (tipc_sk_connected(sk))
2879		return -EINVAL;
2880	key = tsk->portid + tsk->pub_count + 1;
2881	if (key == tsk->portid)
2882		return -EADDRINUSE;
2883
2884	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2885				    scope, tsk->portid, key);
2886	if (unlikely(!publ))
2887		return -EINVAL;
2888
2889	list_add(&publ->binding_sock, &tsk->publications);
2890	tsk->pub_count++;
2891	tsk->published = 1;
2892	return 0;
2893}
2894
2895static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2896			    struct tipc_name_seq const *seq)
2897{
2898	struct net *net = sock_net(&tsk->sk);
2899	struct publication *publ;
2900	struct publication *safe;
2901	int rc = -EINVAL;
2902
2903	if (scope != TIPC_NODE_SCOPE)
2904		scope = TIPC_CLUSTER_SCOPE;
2905
2906	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2907		if (seq) {
2908			if (publ->scope != scope)
2909				continue;
2910			if (publ->type != seq->type)
2911				continue;
2912			if (publ->lower != seq->lower)
2913				continue;
2914			if (publ->upper != seq->upper)
2915				break;
2916			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2917					      publ->upper, publ->key);
2918			rc = 0;
2919			break;
2920		}
2921		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2922				      publ->upper, publ->key);
2923		rc = 0;
2924	}
2925	if (list_empty(&tsk->publications))
2926		tsk->published = 0;
2927	return rc;
2928}
2929
2930/* tipc_sk_reinit: set non-zero address in all existing sockets
2931 *                 when we go from standalone to network mode.
2932 */
2933void tipc_sk_reinit(struct net *net)
2934{
2935	struct tipc_net *tn = net_generic(net, tipc_net_id);
2936	struct rhashtable_iter iter;
2937	struct tipc_sock *tsk;
2938	struct tipc_msg *msg;
2939
2940	rhashtable_walk_enter(&tn->sk_rht, &iter);
2941
2942	do {
2943		rhashtable_walk_start(&iter);
2944
2945		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2946			sock_hold(&tsk->sk);
2947			rhashtable_walk_stop(&iter);
2948			lock_sock(&tsk->sk);
2949			msg = &tsk->phdr;
2950			msg_set_prevnode(msg, tipc_own_addr(net));
2951			msg_set_orignode(msg, tipc_own_addr(net));
2952			release_sock(&tsk->sk);
2953			rhashtable_walk_start(&iter);
2954			sock_put(&tsk->sk);
2955		}
2956
2957		rhashtable_walk_stop(&iter);
2958	} while (tsk == ERR_PTR(-EAGAIN));
2959
2960	rhashtable_walk_exit(&iter);
2961}
2962
2963static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2964{
2965	struct tipc_net *tn = net_generic(net, tipc_net_id);
2966	struct tipc_sock *tsk;
2967
2968	rcu_read_lock();
2969	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2970	if (tsk)
2971		sock_hold(&tsk->sk);
2972	rcu_read_unlock();
2973
2974	return tsk;
2975}
2976
2977static int tipc_sk_insert(struct tipc_sock *tsk)
2978{
2979	struct sock *sk = &tsk->sk;
2980	struct net *net = sock_net(sk);
2981	struct tipc_net *tn = net_generic(net, tipc_net_id);
2982	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2983	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2984
2985	while (remaining--) {
2986		portid++;
2987		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2988			portid = TIPC_MIN_PORT;
2989		tsk->portid = portid;
2990		sock_hold(&tsk->sk);
2991		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2992						   tsk_rht_params))
2993			return 0;
2994		sock_put(&tsk->sk);
2995	}
2996
2997	return -1;
2998}
2999
3000static void tipc_sk_remove(struct tipc_sock *tsk)
3001{
3002	struct sock *sk = &tsk->sk;
3003	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3004
3005	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3006		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3007		__sock_put(sk);
3008	}
3009}
3010
3011static const struct rhashtable_params tsk_rht_params = {
3012	.nelem_hint = 192,
3013	.head_offset = offsetof(struct tipc_sock, node),
3014	.key_offset = offsetof(struct tipc_sock, portid),
3015	.key_len = sizeof(u32), /* portid */
3016	.max_size = 1048576,
3017	.min_size = 256,
3018	.automatic_shrinking = true,
3019};
3020
3021int tipc_sk_rht_init(struct net *net)
3022{
3023	struct tipc_net *tn = net_generic(net, tipc_net_id);
3024
3025	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3026}
3027
3028void tipc_sk_rht_destroy(struct net *net)
3029{
3030	struct tipc_net *tn = net_generic(net, tipc_net_id);
3031
3032	/* Wait for socket readers to complete */
3033	synchronize_net();
3034
3035	rhashtable_destroy(&tn->sk_rht);
3036}
3037
3038static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3039{
3040	struct net *net = sock_net(&tsk->sk);
3041	struct tipc_group *grp = tsk->group;
3042	struct tipc_msg *hdr = &tsk->phdr;
3043	struct tipc_name_seq seq;
3044	int rc;
3045
3046	if (mreq->type < TIPC_RESERVED_TYPES)
3047		return -EACCES;
3048	if (mreq->scope > TIPC_NODE_SCOPE)
3049		return -EINVAL;
3050	if (grp)
3051		return -EACCES;
3052	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3053	if (!grp)
3054		return -ENOMEM;
3055	tsk->group = grp;
3056	msg_set_lookup_scope(hdr, mreq->scope);
3057	msg_set_nametype(hdr, mreq->type);
3058	msg_set_dest_droppable(hdr, true);
3059	seq.type = mreq->type;
3060	seq.lower = mreq->instance;
3061	seq.upper = seq.lower;
3062	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3063	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3064	if (rc) {
3065		tipc_group_delete(net, grp);
3066		tsk->group = NULL;
3067		return rc;
3068	}
3069	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3070	tsk->mc_method.rcast = true;
3071	tsk->mc_method.mandatory = true;
3072	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3073	return rc;
3074}
3075
3076static int tipc_sk_leave(struct tipc_sock *tsk)
3077{
3078	struct net *net = sock_net(&tsk->sk);
3079	struct tipc_group *grp = tsk->group;
3080	struct tipc_name_seq seq;
3081	int scope;
3082
3083	if (!grp)
3084		return -EINVAL;
3085	tipc_group_self(grp, &seq, &scope);
3086	tipc_group_delete(net, grp);
3087	tsk->group = NULL;
3088	tipc_sk_withdraw(tsk, scope, &seq);
3089	return 0;
3090}
3091
3092/**
3093 * tipc_setsockopt - set socket option
3094 * @sock: socket structure
3095 * @lvl: option level
3096 * @opt: option identifier
3097 * @ov: pointer to new option value
3098 * @ol: length of option value
3099 *
3100 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3101 * (to ease compatibility).
3102 *
3103 * Returns 0 on success, errno otherwise
3104 */
3105static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3106			   sockptr_t ov, unsigned int ol)
3107{
3108	struct sock *sk = sock->sk;
3109	struct tipc_sock *tsk = tipc_sk(sk);
3110	struct tipc_group_req mreq;
3111	u32 value = 0;
3112	int res = 0;
3113
3114	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3115		return 0;
3116	if (lvl != SOL_TIPC)
3117		return -ENOPROTOOPT;
3118
3119	switch (opt) {
3120	case TIPC_IMPORTANCE:
3121	case TIPC_SRC_DROPPABLE:
3122	case TIPC_DEST_DROPPABLE:
3123	case TIPC_CONN_TIMEOUT:
3124	case TIPC_NODELAY:
3125		if (ol < sizeof(value))
3126			return -EINVAL;
3127		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3128			return -EFAULT;
3129		break;
3130	case TIPC_GROUP_JOIN:
3131		if (ol < sizeof(mreq))
3132			return -EINVAL;
3133		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3134			return -EFAULT;
3135		break;
3136	default:
3137		if (!sockptr_is_null(ov) || ol)
3138			return -EINVAL;
3139	}
3140
3141	lock_sock(sk);
3142
3143	switch (opt) {
3144	case TIPC_IMPORTANCE:
3145		res = tsk_set_importance(sk, value);
3146		break;
3147	case TIPC_SRC_DROPPABLE:
3148		if (sock->type != SOCK_STREAM)
3149			tsk_set_unreliable(tsk, value);
3150		else
3151			res = -ENOPROTOOPT;
3152		break;
3153	case TIPC_DEST_DROPPABLE:
3154		tsk_set_unreturnable(tsk, value);
3155		break;
3156	case TIPC_CONN_TIMEOUT:
3157		tipc_sk(sk)->conn_timeout = value;
3158		break;
3159	case TIPC_MCAST_BROADCAST:
3160		tsk->mc_method.rcast = false;
3161		tsk->mc_method.mandatory = true;
3162		break;
3163	case TIPC_MCAST_REPLICAST:
3164		tsk->mc_method.rcast = true;
3165		tsk->mc_method.mandatory = true;
3166		break;
3167	case TIPC_GROUP_JOIN:
3168		res = tipc_sk_join(tsk, &mreq);
3169		break;
3170	case TIPC_GROUP_LEAVE:
3171		res = tipc_sk_leave(tsk);
3172		break;
3173	case TIPC_NODELAY:
3174		tsk->nodelay = !!value;
3175		tsk_set_nagle(tsk);
3176		break;
3177	default:
3178		res = -EINVAL;
3179	}
3180
3181	release_sock(sk);
3182
3183	return res;
3184}
3185
3186/**
3187 * tipc_getsockopt - get socket option
3188 * @sock: socket structure
3189 * @lvl: option level
3190 * @opt: option identifier
3191 * @ov: receptacle for option value
3192 * @ol: receptacle for length of option value
3193 *
3194 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3195 * (to ease compatibility).
3196 *
3197 * Returns 0 on success, errno otherwise
3198 */
3199static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3200			   char __user *ov, int __user *ol)
3201{
3202	struct sock *sk = sock->sk;
3203	struct tipc_sock *tsk = tipc_sk(sk);
3204	struct tipc_name_seq seq;
3205	int len, scope;
3206	u32 value;
3207	int res;
3208
3209	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3210		return put_user(0, ol);
3211	if (lvl != SOL_TIPC)
3212		return -ENOPROTOOPT;
3213	res = get_user(len, ol);
3214	if (res)
3215		return res;
3216
3217	lock_sock(sk);
3218
3219	switch (opt) {
3220	case TIPC_IMPORTANCE:
3221		value = tsk_importance(tsk);
3222		break;
3223	case TIPC_SRC_DROPPABLE:
3224		value = tsk_unreliable(tsk);
3225		break;
3226	case TIPC_DEST_DROPPABLE:
3227		value = tsk_unreturnable(tsk);
3228		break;
3229	case TIPC_CONN_TIMEOUT:
3230		value = tsk->conn_timeout;
3231		/* no need to set "res", since already 0 at this point */
3232		break;
3233	case TIPC_NODE_RECVQ_DEPTH:
3234		value = 0; /* was tipc_queue_size, now obsolete */
3235		break;
3236	case TIPC_SOCK_RECVQ_DEPTH:
3237		value = skb_queue_len(&sk->sk_receive_queue);
3238		break;
3239	case TIPC_SOCK_RECVQ_USED:
3240		value = sk_rmem_alloc_get(sk);
3241		break;
3242	case TIPC_GROUP_JOIN:
3243		seq.type = 0;
3244		if (tsk->group)
3245			tipc_group_self(tsk->group, &seq, &scope);
3246		value = seq.type;
3247		break;
3248	default:
3249		res = -EINVAL;
3250	}
3251
3252	release_sock(sk);
3253
3254	if (res)
3255		return res;	/* "get" failed */
3256
3257	if (len < sizeof(value))
3258		return -EINVAL;
3259
3260	if (copy_to_user(ov, &value, sizeof(value)))
3261		return -EFAULT;
3262
3263	return put_user(sizeof(value), ol);
3264}
3265
3266static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3267{
3268	struct net *net = sock_net(sock->sk);
3269	struct tipc_sioc_nodeid_req nr = {0};
3270	struct tipc_sioc_ln_req lnr;
3271	void __user *argp = (void __user *)arg;
3272
3273	switch (cmd) {
3274	case SIOCGETLINKNAME:
3275		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3276			return -EFAULT;
3277		if (!tipc_node_get_linkname(net,
3278					    lnr.bearer_id & 0xffff, lnr.peer,
3279					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3280			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3281				return -EFAULT;
3282			return 0;
3283		}
3284		return -EADDRNOTAVAIL;
3285	case SIOCGETNODEID:
3286		if (copy_from_user(&nr, argp, sizeof(nr)))
3287			return -EFAULT;
3288		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3289			return -EADDRNOTAVAIL;
3290		if (copy_to_user(argp, &nr, sizeof(nr)))
3291			return -EFAULT;
3292		return 0;
3293	default:
3294		return -ENOIOCTLCMD;
3295	}
3296}
3297
3298static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3299{
3300	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3301	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3302	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3303
3304	tsk1->peer.family = AF_TIPC;
3305	tsk1->peer.addrtype = TIPC_ADDR_ID;
3306	tsk1->peer.scope = TIPC_NODE_SCOPE;
3307	tsk1->peer.addr.id.ref = tsk2->portid;
3308	tsk1->peer.addr.id.node = onode;
3309	tsk2->peer.family = AF_TIPC;
3310	tsk2->peer.addrtype = TIPC_ADDR_ID;
3311	tsk2->peer.scope = TIPC_NODE_SCOPE;
3312	tsk2->peer.addr.id.ref = tsk1->portid;
3313	tsk2->peer.addr.id.node = onode;
3314
3315	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3316	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3317	return 0;
3318}
3319
3320/* Protocol switches for the various types of TIPC sockets */
3321
3322static const struct proto_ops msg_ops = {
3323	.owner		= THIS_MODULE,
3324	.family		= AF_TIPC,
3325	.release	= tipc_release,
3326	.bind		= tipc_bind,
3327	.connect	= tipc_connect,
3328	.socketpair	= tipc_socketpair,
3329	.accept		= sock_no_accept,
3330	.getname	= tipc_getname,
3331	.poll		= tipc_poll,
3332	.ioctl		= tipc_ioctl,
3333	.listen		= sock_no_listen,
3334	.shutdown	= tipc_shutdown,
3335	.setsockopt	= tipc_setsockopt,
3336	.getsockopt	= tipc_getsockopt,
3337	.sendmsg	= tipc_sendmsg,
3338	.recvmsg	= tipc_recvmsg,
3339	.mmap		= sock_no_mmap,
3340	.sendpage	= sock_no_sendpage
3341};
3342
3343static const struct proto_ops packet_ops = {
3344	.owner		= THIS_MODULE,
3345	.family		= AF_TIPC,
3346	.release	= tipc_release,
3347	.bind		= tipc_bind,
3348	.connect	= tipc_connect,
3349	.socketpair	= tipc_socketpair,
3350	.accept		= tipc_accept,
3351	.getname	= tipc_getname,
3352	.poll		= tipc_poll,
3353	.ioctl		= tipc_ioctl,
3354	.listen		= tipc_listen,
3355	.shutdown	= tipc_shutdown,
3356	.setsockopt	= tipc_setsockopt,
3357	.getsockopt	= tipc_getsockopt,
3358	.sendmsg	= tipc_send_packet,
3359	.recvmsg	= tipc_recvmsg,
3360	.mmap		= sock_no_mmap,
3361	.sendpage	= sock_no_sendpage
3362};
3363
3364static const struct proto_ops stream_ops = {
3365	.owner		= THIS_MODULE,
3366	.family		= AF_TIPC,
3367	.release	= tipc_release,
3368	.bind		= tipc_bind,
3369	.connect	= tipc_connect,
3370	.socketpair	= tipc_socketpair,
3371	.accept		= tipc_accept,
3372	.getname	= tipc_getname,
3373	.poll		= tipc_poll,
3374	.ioctl		= tipc_ioctl,
3375	.listen		= tipc_listen,
3376	.shutdown	= tipc_shutdown,
3377	.setsockopt	= tipc_setsockopt,
3378	.getsockopt	= tipc_getsockopt,
3379	.sendmsg	= tipc_sendstream,
3380	.recvmsg	= tipc_recvstream,
3381	.mmap		= sock_no_mmap,
3382	.sendpage	= sock_no_sendpage
3383};
3384
3385static const struct net_proto_family tipc_family_ops = {
3386	.owner		= THIS_MODULE,
3387	.family		= AF_TIPC,
3388	.create		= tipc_sk_create
3389};
3390
3391static struct proto tipc_proto = {
3392	.name		= "TIPC",
3393	.owner		= THIS_MODULE,
3394	.obj_size	= sizeof(struct tipc_sock),
3395	.sysctl_rmem	= sysctl_tipc_rmem
3396};
3397
3398/**
3399 * tipc_socket_init - initialize TIPC socket interface
3400 *
3401 * Returns 0 on success, errno otherwise
3402 */
3403int tipc_socket_init(void)
3404{
3405	int res;
3406
3407	res = proto_register(&tipc_proto, 1);
3408	if (res) {
3409		pr_err("Failed to register TIPC protocol type\n");
3410		goto out;
3411	}
3412
3413	res = sock_register(&tipc_family_ops);
3414	if (res) {
3415		pr_err("Failed to register TIPC socket type\n");
3416		proto_unregister(&tipc_proto);
3417		goto out;
3418	}
3419 out:
3420	return res;
3421}
3422
3423/**
3424 * tipc_socket_stop - stop TIPC socket interface
3425 */
3426void tipc_socket_stop(void)
3427{
3428	sock_unregister(tipc_family_ops.family);
3429	proto_unregister(&tipc_proto);
3430}
3431
3432/* Caller should hold socket lock for the passed tipc socket. */
3433static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3434{
3435	u32 peer_node;
3436	u32 peer_port;
3437	struct nlattr *nest;
3438
3439	peer_node = tsk_peer_node(tsk);
3440	peer_port = tsk_peer_port(tsk);
3441
3442	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3443	if (!nest)
3444		return -EMSGSIZE;
3445
3446	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3447		goto msg_full;
3448	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3449		goto msg_full;
3450
3451	if (tsk->conn_type != 0) {
3452		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3453			goto msg_full;
3454		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3455			goto msg_full;
3456		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3457			goto msg_full;
3458	}
3459	nla_nest_end(skb, nest);
3460
3461	return 0;
3462
3463msg_full:
3464	nla_nest_cancel(skb, nest);
3465
3466	return -EMSGSIZE;
3467}
3468
3469static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3470			  *tsk)
3471{
3472	struct net *net = sock_net(skb->sk);
3473	struct sock *sk = &tsk->sk;
3474
3475	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3476	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3477		return -EMSGSIZE;
3478
3479	if (tipc_sk_connected(sk)) {
3480		if (__tipc_nl_add_sk_con(skb, tsk))
3481			return -EMSGSIZE;
3482	} else if (!list_empty(&tsk->publications)) {
3483		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3484			return -EMSGSIZE;
3485	}
3486	return 0;
3487}
3488
3489/* Caller should hold socket lock for the passed tipc socket. */
3490static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3491			    struct tipc_sock *tsk)
3492{
3493	struct nlattr *attrs;
3494	void *hdr;
3495
3496	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3497			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3498	if (!hdr)
3499		goto msg_cancel;
3500
3501	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3502	if (!attrs)
3503		goto genlmsg_cancel;
3504
3505	if (__tipc_nl_add_sk_info(skb, tsk))
3506		goto attr_msg_cancel;
3507
3508	nla_nest_end(skb, attrs);
3509	genlmsg_end(skb, hdr);
3510
3511	return 0;
3512
3513attr_msg_cancel:
3514	nla_nest_cancel(skb, attrs);
3515genlmsg_cancel:
3516	genlmsg_cancel(skb, hdr);
3517msg_cancel:
3518	return -EMSGSIZE;
3519}
3520
3521int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3522		    int (*skb_handler)(struct sk_buff *skb,
3523				       struct netlink_callback *cb,
3524				       struct tipc_sock *tsk))
3525{
3526	struct rhashtable_iter *iter = (void *)cb->args[4];
3527	struct tipc_sock *tsk;
3528	int err;
3529
3530	rhashtable_walk_start(iter);
3531	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3532		if (IS_ERR(tsk)) {
3533			err = PTR_ERR(tsk);
3534			if (err == -EAGAIN) {
3535				err = 0;
3536				continue;
3537			}
3538			break;
3539		}
3540
3541		sock_hold(&tsk->sk);
3542		rhashtable_walk_stop(iter);
3543		lock_sock(&tsk->sk);
3544		err = skb_handler(skb, cb, tsk);
3545		if (err) {
3546			release_sock(&tsk->sk);
3547			sock_put(&tsk->sk);
3548			goto out;
3549		}
3550		release_sock(&tsk->sk);
3551		rhashtable_walk_start(iter);
3552		sock_put(&tsk->sk);
3553	}
3554	rhashtable_walk_stop(iter);
3555out:
3556	return skb->len;
3557}
3558EXPORT_SYMBOL(tipc_nl_sk_walk);
3559
3560int tipc_dump_start(struct netlink_callback *cb)
3561{
3562	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3563}
3564EXPORT_SYMBOL(tipc_dump_start);
3565
3566int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3567{
3568	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3569	struct rhashtable_iter *iter = (void *)cb->args[4];
3570	struct tipc_net *tn = tipc_net(net);
3571
3572	if (!iter) {
3573		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3574		if (!iter)
3575			return -ENOMEM;
3576
3577		cb->args[4] = (long)iter;
3578	}
3579
3580	rhashtable_walk_enter(&tn->sk_rht, iter);
3581	return 0;
3582}
3583
3584int tipc_dump_done(struct netlink_callback *cb)
3585{
3586	struct rhashtable_iter *hti = (void *)cb->args[4];
3587
3588	rhashtable_walk_exit(hti);
3589	kfree(hti);
3590	return 0;
3591}
3592EXPORT_SYMBOL(tipc_dump_done);
3593
3594int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3595			   struct tipc_sock *tsk, u32 sk_filter_state,
3596			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3597{
3598	struct sock *sk = &tsk->sk;
3599	struct nlattr *attrs;
3600	struct nlattr *stat;
3601
3602	/*filter response w.r.t sk_state*/
3603	if (!(sk_filter_state & (1 << sk->sk_state)))
3604		return 0;
3605
3606	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3607	if (!attrs)
3608		goto msg_cancel;
3609
3610	if (__tipc_nl_add_sk_info(skb, tsk))
3611		goto attr_msg_cancel;
3612
3613	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3614	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3615	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3616	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3617			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3618					 sock_i_uid(sk))) ||
3619	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3620			      tipc_diag_gen_cookie(sk),
3621			      TIPC_NLA_SOCK_PAD))
3622		goto attr_msg_cancel;
3623
3624	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3625	if (!stat)
3626		goto attr_msg_cancel;
3627
3628	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3629			skb_queue_len(&sk->sk_receive_queue)) ||
3630	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3631			skb_queue_len(&sk->sk_write_queue)) ||
3632	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3633			atomic_read(&sk->sk_drops)))
3634		goto stat_msg_cancel;
3635
3636	if (tsk->cong_link_cnt &&
3637	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3638		goto stat_msg_cancel;
3639
3640	if (tsk_conn_cong(tsk) &&
3641	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3642		goto stat_msg_cancel;
3643
3644	nla_nest_end(skb, stat);
3645
3646	if (tsk->group)
3647		if (tipc_group_fill_sock_diag(tsk->group, skb))
3648			goto stat_msg_cancel;
3649
3650	nla_nest_end(skb, attrs);
3651
3652	return 0;
3653
3654stat_msg_cancel:
3655	nla_nest_cancel(skb, stat);
3656attr_msg_cancel:
3657	nla_nest_cancel(skb, attrs);
3658msg_cancel:
3659	return -EMSGSIZE;
3660}
3661EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3662
3663int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3664{
3665	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3666}
3667
3668/* Caller should hold socket lock for the passed tipc socket. */
3669static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3670				 struct netlink_callback *cb,
3671				 struct publication *publ)
3672{
3673	void *hdr;
3674	struct nlattr *attrs;
3675
3676	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3677			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3678	if (!hdr)
3679		goto msg_cancel;
3680
3681	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3682	if (!attrs)
3683		goto genlmsg_cancel;
3684
3685	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3686		goto attr_msg_cancel;
3687	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3688		goto attr_msg_cancel;
3689	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3690		goto attr_msg_cancel;
3691	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3692		goto attr_msg_cancel;
3693
3694	nla_nest_end(skb, attrs);
3695	genlmsg_end(skb, hdr);
3696
3697	return 0;
3698
3699attr_msg_cancel:
3700	nla_nest_cancel(skb, attrs);
3701genlmsg_cancel:
3702	genlmsg_cancel(skb, hdr);
3703msg_cancel:
3704	return -EMSGSIZE;
3705}
3706
3707/* Caller should hold socket lock for the passed tipc socket. */
3708static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3709				  struct netlink_callback *cb,
3710				  struct tipc_sock *tsk, u32 *last_publ)
3711{
3712	int err;
3713	struct publication *p;
3714
3715	if (*last_publ) {
3716		list_for_each_entry(p, &tsk->publications, binding_sock) {
3717			if (p->key == *last_publ)
3718				break;
3719		}
3720		if (p->key != *last_publ) {
3721			/* We never set seq or call nl_dump_check_consistent()
3722			 * this means that setting prev_seq here will cause the
3723			 * consistence check to fail in the netlink callback
3724			 * handler. Resulting in the last NLMSG_DONE message
3725			 * having the NLM_F_DUMP_INTR flag set.
3726			 */
3727			cb->prev_seq = 1;
3728			*last_publ = 0;
3729			return -EPIPE;
3730		}
3731	} else {
3732		p = list_first_entry(&tsk->publications, struct publication,
3733				     binding_sock);
3734	}
3735
3736	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3737		err = __tipc_nl_add_sk_publ(skb, cb, p);
3738		if (err) {
3739			*last_publ = p->key;
3740			return err;
3741		}
3742	}
3743	*last_publ = 0;
3744
3745	return 0;
3746}
3747
3748int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3749{
3750	int err;
3751	u32 tsk_portid = cb->args[0];
3752	u32 last_publ = cb->args[1];
3753	u32 done = cb->args[2];
3754	struct net *net = sock_net(skb->sk);
3755	struct tipc_sock *tsk;
3756
3757	if (!tsk_portid) {
3758		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3759		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
 
 
 
 
3760
3761		if (!attrs[TIPC_NLA_SOCK])
3762			return -EINVAL;
3763
3764		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3765						  attrs[TIPC_NLA_SOCK],
3766						  tipc_nl_sock_policy, NULL);
3767		if (err)
3768			return err;
3769
3770		if (!sock[TIPC_NLA_SOCK_REF])
3771			return -EINVAL;
3772
3773		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3774	}
3775
3776	if (done)
3777		return 0;
3778
3779	tsk = tipc_sk_lookup(net, tsk_portid);
3780	if (!tsk)
3781		return -EINVAL;
3782
3783	lock_sock(&tsk->sk);
3784	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3785	if (!err)
3786		done = 1;
3787	release_sock(&tsk->sk);
3788	sock_put(&tsk->sk);
3789
3790	cb->args[0] = tsk_portid;
3791	cb->args[1] = last_publ;
3792	cb->args[2] = done;
3793
3794	return skb->len;
3795}
3796
3797/**
3798 * tipc_sk_filtering - check if a socket should be traced
3799 * @sk: the socket to be examined
3800 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3801 *  (portid, sock type, name type, name lower, name upper)
3802 *
3803 * Returns true if the socket meets the socket tuple data
3804 * (value 0 = 'any') or when there is no tuple set (all = 0),
3805 * otherwise false
3806 */
3807bool tipc_sk_filtering(struct sock *sk)
3808{
3809	struct tipc_sock *tsk;
3810	struct publication *p;
3811	u32 _port, _sktype, _type, _lower, _upper;
3812	u32 type = 0, lower = 0, upper = 0;
3813
3814	if (!sk)
3815		return true;
3816
3817	tsk = tipc_sk(sk);
3818
3819	_port = sysctl_tipc_sk_filter[0];
3820	_sktype = sysctl_tipc_sk_filter[1];
3821	_type = sysctl_tipc_sk_filter[2];
3822	_lower = sysctl_tipc_sk_filter[3];
3823	_upper = sysctl_tipc_sk_filter[4];
3824
3825	if (!_port && !_sktype && !_type && !_lower && !_upper)
3826		return true;
3827
3828	if (_port)
3829		return (_port == tsk->portid);
3830
3831	if (_sktype && _sktype != sk->sk_type)
3832		return false;
3833
3834	if (tsk->published) {
3835		p = list_first_entry_or_null(&tsk->publications,
3836					     struct publication, binding_sock);
3837		if (p) {
3838			type = p->type;
3839			lower = p->lower;
3840			upper = p->upper;
3841		}
3842	}
3843
3844	if (!tipc_sk_type_connectionless(sk)) {
3845		type = tsk->conn_type;
3846		lower = tsk->conn_instance;
3847		upper = tsk->conn_instance;
3848	}
3849
3850	if ((_type && _type != type) || (_lower && _lower != lower) ||
3851	    (_upper && _upper != upper))
3852		return false;
3853
3854	return true;
3855}
3856
3857u32 tipc_sock_get_portid(struct sock *sk)
3858{
3859	return (sk) ? (tipc_sk(sk))->portid : 0;
3860}
3861
3862/**
3863 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3864 *			both the rcv and backlog queues are considered
3865 * @sk: tipc sk to be checked
3866 * @skb: tipc msg to be checked
3867 *
3868 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3869 */
3870
3871bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3872{
3873	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3874	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3875	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3876
3877	return (qsize > lim * 90 / 100);
3878}
3879
3880/**
3881 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3882 *			only the rcv queue is considered
3883 * @sk: tipc sk to be checked
3884 * @skb: tipc msg to be checked
3885 *
3886 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3887 */
3888
3889bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3890{
3891	unsigned int lim = rcvbuf_limit(sk, skb);
3892	unsigned int qsize = sk_rmem_alloc_get(sk);
3893
3894	return (qsize > lim * 90 / 100);
3895}
3896
3897/**
3898 * tipc_sk_dump - dump TIPC socket
3899 * @sk: tipc sk to be dumped
3900 * @dqueues: bitmask to decide if any socket queue to be dumped?
3901 *           - TIPC_DUMP_NONE: don't dump socket queues
3902 *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3903 *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3904 *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3905 *           - TIPC_DUMP_ALL: dump all the socket queues above
3906 * @buf: returned buffer of dump data in format
3907 */
3908int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3909{
3910	int i = 0;
3911	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3912	struct tipc_sock *tsk;
3913	struct publication *p;
3914	bool tsk_connected;
3915
3916	if (!sk) {
3917		i += scnprintf(buf, sz, "sk data: (null)\n");
3918		return i;
3919	}
3920
3921	tsk = tipc_sk(sk);
3922	tsk_connected = !tipc_sk_type_connectionless(sk);
3923
3924	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3925	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3926	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3927	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3928	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3929	if (tsk_connected) {
3930		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3931		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3932		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3933		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3934	}
3935	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3936	if (tsk->published) {
3937		p = list_first_entry_or_null(&tsk->publications,
3938					     struct publication, binding_sock);
3939		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3940		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3941		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3942	}
3943	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3944	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3945	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3946	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3947	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3948	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3949	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3950	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3951	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3952	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3953	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3954	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3955	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3956	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3957
3958	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3959		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3960		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3961	}
3962
3963	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3964		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3965		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3966	}
3967
3968	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3969		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3970		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3971		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3972			i += scnprintf(buf + i, sz - i, "  tail ");
3973			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3974					   buf + i);
3975		}
3976	}
3977
3978	return i;
3979}
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}