xref: /linux-6.15/net/tipc/socket.c (revision 34c201ae)
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 
50 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
51 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
52 #define TIPC_FWD_MSG		1
53 #define TIPC_MAX_PORT		0xffffffff
54 #define TIPC_MIN_PORT		1
55 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
56 
57 enum {
58 	TIPC_LISTEN = TCP_LISTEN,
59 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
60 	TIPC_OPEN = TCP_CLOSE,
61 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
62 	TIPC_CONNECTING = TCP_SYN_SENT,
63 };
64 
65 struct sockaddr_pair {
66 	struct sockaddr_tipc sock;
67 	struct sockaddr_tipc member;
68 };
69 
70 /**
71  * struct tipc_sock - TIPC socket structure
72  * @sk: socket - interacts with 'port' and with user via the socket API
73  * @conn_type: TIPC type used when connection was established
74  * @conn_instance: TIPC instance used when connection was established
75  * @published: non-zero if port has one or more associated names
76  * @max_pkt: maximum packet size "hint" used when building messages sent by port
77  * @portid: unique port identity in TIPC socket hash table
78  * @phdr: preformatted message header used when sending messages
79  * #cong_links: list of congested links
80  * @publications: list of publications for port
81  * @blocking_link: address of the congested link we are currently sleeping on
82  * @pub_count: total # of publications port has made during its lifetime
83  * @probing_state:
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  */
94 struct 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 	uint conn_timeout;
106 	atomic_t dupl_rcvcnt;
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 
122 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
123 static void tipc_data_ready(struct sock *sk);
124 static void tipc_write_space(struct sock *sk);
125 static void tipc_sock_destruct(struct sock *sk);
126 static int tipc_release(struct socket *sock);
127 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
128 		       bool kern);
129 static void tipc_sk_timeout(struct timer_list *t);
130 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
131 			   struct tipc_name_seq const *seq);
132 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
133 			    struct tipc_name_seq const *seq);
134 static int tipc_sk_leave(struct tipc_sock *tsk);
135 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
136 static int tipc_sk_insert(struct tipc_sock *tsk);
137 static void tipc_sk_remove(struct tipc_sock *tsk);
138 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
139 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
140 
141 static const struct proto_ops packet_ops;
142 static const struct proto_ops stream_ops;
143 static const struct proto_ops msg_ops;
144 static struct proto tipc_proto;
145 static const struct rhashtable_params tsk_rht_params;
146 
147 static u32 tsk_own_node(struct tipc_sock *tsk)
148 {
149 	return msg_prevnode(&tsk->phdr);
150 }
151 
152 static u32 tsk_peer_node(struct tipc_sock *tsk)
153 {
154 	return msg_destnode(&tsk->phdr);
155 }
156 
157 static u32 tsk_peer_port(struct tipc_sock *tsk)
158 {
159 	return msg_destport(&tsk->phdr);
160 }
161 
162 static  bool tsk_unreliable(struct tipc_sock *tsk)
163 {
164 	return msg_src_droppable(&tsk->phdr) != 0;
165 }
166 
167 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
168 {
169 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
170 }
171 
172 static bool tsk_unreturnable(struct tipc_sock *tsk)
173 {
174 	return msg_dest_droppable(&tsk->phdr) != 0;
175 }
176 
177 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
178 {
179 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
180 }
181 
182 static int tsk_importance(struct tipc_sock *tsk)
183 {
184 	return msg_importance(&tsk->phdr);
185 }
186 
187 static 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 
195 static struct tipc_sock *tipc_sk(const struct sock *sk)
196 {
197 	return container_of(sk, struct tipc_sock, sk);
198 }
199 
200 static bool tsk_conn_cong(struct tipc_sock *tsk)
201 {
202 	return tsk->snt_unacked > tsk->snd_win;
203 }
204 
205 static 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  */
214 static 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  */
223 static 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  */
235 static void tsk_advance_rx_queue(struct sock *sk)
236 {
237 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
238 }
239 
240 /* tipc_sk_respond() : send response message back to sender
241  */
242 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
243 {
244 	u32 selector;
245 	u32 dnode;
246 	u32 onode = tipc_own_addr(sock_net(sk));
247 
248 	if (!tipc_msg_reverse(onode, &skb, err))
249 		return;
250 
251 	dnode = msg_destnode(buf_msg(skb));
252 	selector = msg_origport(buf_msg(skb));
253 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
254 }
255 
256 /**
257  * tsk_rej_rx_queue - reject all buffers in socket receive queue
258  *
259  * Caller must hold socket lock
260  */
261 static void tsk_rej_rx_queue(struct sock *sk)
262 {
263 	struct sk_buff *skb;
264 
265 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
266 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
267 }
268 
269 static bool tipc_sk_connected(struct sock *sk)
270 {
271 	return sk->sk_state == TIPC_ESTABLISHED;
272 }
273 
274 /* tipc_sk_type_connectionless - check if the socket is datagram socket
275  * @sk: socket
276  *
277  * Returns true if connection less, false otherwise
278  */
279 static bool tipc_sk_type_connectionless(struct sock *sk)
280 {
281 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
282 }
283 
284 /* tsk_peer_msg - verify if message was sent by connected port's peer
285  *
286  * Handles cases where the node's network address has changed from
287  * the default of <0.0.0> to its configured setting.
288  */
289 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
290 {
291 	struct sock *sk = &tsk->sk;
292 	u32 self = tipc_own_addr(sock_net(sk));
293 	u32 peer_port = tsk_peer_port(tsk);
294 	u32 orig_node, peer_node;
295 
296 	if (unlikely(!tipc_sk_connected(sk)))
297 		return false;
298 
299 	if (unlikely(msg_origport(msg) != peer_port))
300 		return false;
301 
302 	orig_node = msg_orignode(msg);
303 	peer_node = tsk_peer_node(tsk);
304 
305 	if (likely(orig_node == peer_node))
306 		return true;
307 
308 	if (!orig_node && peer_node == self)
309 		return true;
310 
311 	if (!peer_node && orig_node == self)
312 		return true;
313 
314 	return false;
315 }
316 
317 /* tipc_set_sk_state - set the sk_state of the socket
318  * @sk: socket
319  *
320  * Caller must hold socket lock
321  *
322  * Returns 0 on success, errno otherwise
323  */
324 static int tipc_set_sk_state(struct sock *sk, int state)
325 {
326 	int oldsk_state = sk->sk_state;
327 	int res = -EINVAL;
328 
329 	switch (state) {
330 	case TIPC_OPEN:
331 		res = 0;
332 		break;
333 	case TIPC_LISTEN:
334 	case TIPC_CONNECTING:
335 		if (oldsk_state == TIPC_OPEN)
336 			res = 0;
337 		break;
338 	case TIPC_ESTABLISHED:
339 		if (oldsk_state == TIPC_CONNECTING ||
340 		    oldsk_state == TIPC_OPEN)
341 			res = 0;
342 		break;
343 	case TIPC_DISCONNECTING:
344 		if (oldsk_state == TIPC_CONNECTING ||
345 		    oldsk_state == TIPC_ESTABLISHED)
346 			res = 0;
347 		break;
348 	}
349 
350 	if (!res)
351 		sk->sk_state = state;
352 
353 	return res;
354 }
355 
356 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
357 {
358 	struct sock *sk = sock->sk;
359 	int err = sock_error(sk);
360 	int typ = sock->type;
361 
362 	if (err)
363 		return err;
364 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
365 		if (sk->sk_state == TIPC_DISCONNECTING)
366 			return -EPIPE;
367 		else if (!tipc_sk_connected(sk))
368 			return -ENOTCONN;
369 	}
370 	if (!*timeout)
371 		return -EAGAIN;
372 	if (signal_pending(current))
373 		return sock_intr_errno(*timeout);
374 
375 	return 0;
376 }
377 
378 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
379 ({                                                                             \
380 	struct sock *sk_;						       \
381 	int rc_;							       \
382 									       \
383 	while ((rc_ = !(condition_))) {					       \
384 		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
385 		sk_ = (sock_)->sk;					       \
386 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
387 		if (rc_)						       \
388 			break;						       \
389 		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
390 		release_sock(sk_);					       \
391 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
392 		sched_annotate_sleep();				               \
393 		lock_sock(sk_);						       \
394 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
395 	}								       \
396 	rc_;								       \
397 })
398 
399 /**
400  * tipc_sk_create - create a TIPC socket
401  * @net: network namespace (must be default network)
402  * @sock: pre-allocated socket structure
403  * @protocol: protocol indicator (must be 0)
404  * @kern: caused by kernel or by userspace?
405  *
406  * This routine creates additional data structures used by the TIPC socket,
407  * initializes them, and links them together.
408  *
409  * Returns 0 on success, errno otherwise
410  */
411 static int tipc_sk_create(struct net *net, struct socket *sock,
412 			  int protocol, int kern)
413 {
414 	const struct proto_ops *ops;
415 	struct sock *sk;
416 	struct tipc_sock *tsk;
417 	struct tipc_msg *msg;
418 
419 	/* Validate arguments */
420 	if (unlikely(protocol != 0))
421 		return -EPROTONOSUPPORT;
422 
423 	switch (sock->type) {
424 	case SOCK_STREAM:
425 		ops = &stream_ops;
426 		break;
427 	case SOCK_SEQPACKET:
428 		ops = &packet_ops;
429 		break;
430 	case SOCK_DGRAM:
431 	case SOCK_RDM:
432 		ops = &msg_ops;
433 		break;
434 	default:
435 		return -EPROTOTYPE;
436 	}
437 
438 	/* Allocate socket's protocol area */
439 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
440 	if (sk == NULL)
441 		return -ENOMEM;
442 
443 	tsk = tipc_sk(sk);
444 	tsk->max_pkt = MAX_PKT_DEFAULT;
445 	INIT_LIST_HEAD(&tsk->publications);
446 	INIT_LIST_HEAD(&tsk->cong_links);
447 	msg = &tsk->phdr;
448 
449 	/* Finish initializing socket data structures */
450 	sock->ops = ops;
451 	sock_init_data(sock, sk);
452 	tipc_set_sk_state(sk, TIPC_OPEN);
453 	if (tipc_sk_insert(tsk)) {
454 		pr_warn("Socket create failed; port number exhausted\n");
455 		return -EINVAL;
456 	}
457 
458 	/* Ensure tsk is visible before we read own_addr. */
459 	smp_mb();
460 
461 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
462 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
463 
464 	msg_set_origport(msg, tsk->portid);
465 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
466 	sk->sk_shutdown = 0;
467 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
468 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
469 	sk->sk_data_ready = tipc_data_ready;
470 	sk->sk_write_space = tipc_write_space;
471 	sk->sk_destruct = tipc_sock_destruct;
472 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
473 	tsk->group_is_open = true;
474 	atomic_set(&tsk->dupl_rcvcnt, 0);
475 
476 	/* Start out with safe limits until we receive an advertised window */
477 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
478 	tsk->rcv_win = tsk->snd_win;
479 
480 	if (tipc_sk_type_connectionless(sk)) {
481 		tsk_set_unreturnable(tsk, true);
482 		if (sock->type == SOCK_DGRAM)
483 			tsk_set_unreliable(tsk, true);
484 	}
485 
486 	return 0;
487 }
488 
489 static void tipc_sk_callback(struct rcu_head *head)
490 {
491 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
492 
493 	sock_put(&tsk->sk);
494 }
495 
496 /* Caller should hold socket lock for the socket. */
497 static void __tipc_shutdown(struct socket *sock, int error)
498 {
499 	struct sock *sk = sock->sk;
500 	struct tipc_sock *tsk = tipc_sk(sk);
501 	struct net *net = sock_net(sk);
502 	long timeout = CONN_TIMEOUT_DEFAULT;
503 	u32 dnode = tsk_peer_node(tsk);
504 	struct sk_buff *skb;
505 
506 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
507 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
508 					    !tsk_conn_cong(tsk)));
509 
510 	/* Reject all unreceived messages, except on an active connection
511 	 * (which disconnects locally & sends a 'FIN+' to peer).
512 	 */
513 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
514 		if (TIPC_SKB_CB(skb)->bytes_read) {
515 			kfree_skb(skb);
516 			continue;
517 		}
518 		if (!tipc_sk_type_connectionless(sk) &&
519 		    sk->sk_state != TIPC_DISCONNECTING) {
520 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
521 			tipc_node_remove_conn(net, dnode, tsk->portid);
522 		}
523 		tipc_sk_respond(sk, skb, error);
524 	}
525 
526 	if (tipc_sk_type_connectionless(sk))
527 		return;
528 
529 	if (sk->sk_state != TIPC_DISCONNECTING) {
530 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
531 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
532 				      tsk_own_node(tsk), tsk_peer_port(tsk),
533 				      tsk->portid, error);
534 		if (skb)
535 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
536 		tipc_node_remove_conn(net, dnode, tsk->portid);
537 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
538 	}
539 }
540 
541 /**
542  * tipc_release - destroy a TIPC socket
543  * @sock: socket to destroy
544  *
545  * This routine cleans up any messages that are still queued on the socket.
546  * For DGRAM and RDM socket types, all queued messages are rejected.
547  * For SEQPACKET and STREAM socket types, the first message is rejected
548  * and any others are discarded.  (If the first message on a STREAM socket
549  * is partially-read, it is discarded and the next one is rejected instead.)
550  *
551  * NOTE: Rejected messages are not necessarily returned to the sender!  They
552  * are returned or discarded according to the "destination droppable" setting
553  * specified for the message by the sender.
554  *
555  * Returns 0 on success, errno otherwise
556  */
557 static int tipc_release(struct socket *sock)
558 {
559 	struct sock *sk = sock->sk;
560 	struct tipc_sock *tsk;
561 
562 	/*
563 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
564 	 * releases a pre-allocated child socket that was never used)
565 	 */
566 	if (sk == NULL)
567 		return 0;
568 
569 	tsk = tipc_sk(sk);
570 	lock_sock(sk);
571 
572 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
573 	sk->sk_shutdown = SHUTDOWN_MASK;
574 	tipc_sk_leave(tsk);
575 	tipc_sk_withdraw(tsk, 0, NULL);
576 	sk_stop_timer(sk, &sk->sk_timer);
577 	tipc_sk_remove(tsk);
578 
579 	sock_orphan(sk);
580 	/* Reject any messages that accumulated in backlog queue */
581 	release_sock(sk);
582 	tipc_dest_list_purge(&tsk->cong_links);
583 	tsk->cong_link_cnt = 0;
584 	call_rcu(&tsk->rcu, tipc_sk_callback);
585 	sock->sk = NULL;
586 
587 	return 0;
588 }
589 
590 /**
591  * tipc_bind - associate or disassocate TIPC name(s) with a socket
592  * @sock: socket structure
593  * @uaddr: socket address describing name(s) and desired operation
594  * @uaddr_len: size of socket address data structure
595  *
596  * Name and name sequence binding is indicated using a positive scope value;
597  * a negative scope value unbinds the specified name.  Specifying no name
598  * (i.e. a socket address length of 0) unbinds all names from the socket.
599  *
600  * Returns 0 on success, errno otherwise
601  *
602  * NOTE: This routine doesn't need to take the socket lock since it doesn't
603  *       access any non-constant socket information.
604  */
605 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
606 		     int uaddr_len)
607 {
608 	struct sock *sk = sock->sk;
609 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
610 	struct tipc_sock *tsk = tipc_sk(sk);
611 	int res = -EINVAL;
612 
613 	lock_sock(sk);
614 	if (unlikely(!uaddr_len)) {
615 		res = tipc_sk_withdraw(tsk, 0, NULL);
616 		goto exit;
617 	}
618 	if (tsk->group) {
619 		res = -EACCES;
620 		goto exit;
621 	}
622 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
623 		res = -EINVAL;
624 		goto exit;
625 	}
626 	if (addr->family != AF_TIPC) {
627 		res = -EAFNOSUPPORT;
628 		goto exit;
629 	}
630 
631 	if (addr->addrtype == TIPC_ADDR_NAME)
632 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
633 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
634 		res = -EAFNOSUPPORT;
635 		goto exit;
636 	}
637 
638 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
639 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
640 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
641 		res = -EACCES;
642 		goto exit;
643 	}
644 
645 	res = (addr->scope >= 0) ?
646 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
647 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
648 exit:
649 	release_sock(sk);
650 	return res;
651 }
652 
653 /**
654  * tipc_getname - get port ID of socket or peer socket
655  * @sock: socket structure
656  * @uaddr: area for returned socket address
657  * @uaddr_len: area for returned length of socket address
658  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
659  *
660  * Returns 0 on success, errno otherwise
661  *
662  * NOTE: This routine doesn't need to take the socket lock since it only
663  *       accesses socket information that is unchanging (or which changes in
664  *       a completely predictable manner).
665  */
666 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
667 			int peer)
668 {
669 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
670 	struct sock *sk = sock->sk;
671 	struct tipc_sock *tsk = tipc_sk(sk);
672 
673 	memset(addr, 0, sizeof(*addr));
674 	if (peer) {
675 		if ((!tipc_sk_connected(sk)) &&
676 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
677 			return -ENOTCONN;
678 		addr->addr.id.ref = tsk_peer_port(tsk);
679 		addr->addr.id.node = tsk_peer_node(tsk);
680 	} else {
681 		addr->addr.id.ref = tsk->portid;
682 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
683 	}
684 
685 	addr->addrtype = TIPC_ADDR_ID;
686 	addr->family = AF_TIPC;
687 	addr->scope = 0;
688 	addr->addr.name.domain = 0;
689 
690 	return sizeof(*addr);
691 }
692 
693 /**
694  * tipc_poll - read and possibly block on pollmask
695  * @file: file structure associated with the socket
696  * @sock: socket for which to calculate the poll bits
697  * @wait: ???
698  *
699  * Returns pollmask value
700  *
701  * COMMENTARY:
702  * It appears that the usual socket locking mechanisms are not useful here
703  * since the pollmask info is potentially out-of-date the moment this routine
704  * exits.  TCP and other protocols seem to rely on higher level poll routines
705  * to handle any preventable race conditions, so TIPC will do the same ...
706  *
707  * IMPORTANT: The fact that a read or write operation is indicated does NOT
708  * imply that the operation will succeed, merely that it should be performed
709  * and will not block.
710  */
711 static __poll_t tipc_poll(struct file *file, struct socket *sock,
712 			      poll_table *wait)
713 {
714 	struct sock *sk = sock->sk;
715 	struct tipc_sock *tsk = tipc_sk(sk);
716 	__poll_t revents = 0;
717 
718 	sock_poll_wait(file, wait);
719 
720 	if (sk->sk_shutdown & RCV_SHUTDOWN)
721 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
722 	if (sk->sk_shutdown == SHUTDOWN_MASK)
723 		revents |= EPOLLHUP;
724 
725 	switch (sk->sk_state) {
726 	case TIPC_ESTABLISHED:
727 	case TIPC_CONNECTING:
728 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
729 			revents |= EPOLLOUT;
730 		/* fall thru' */
731 	case TIPC_LISTEN:
732 		if (!skb_queue_empty(&sk->sk_receive_queue))
733 			revents |= EPOLLIN | EPOLLRDNORM;
734 		break;
735 	case TIPC_OPEN:
736 		if (tsk->group_is_open && !tsk->cong_link_cnt)
737 			revents |= EPOLLOUT;
738 		if (!tipc_sk_type_connectionless(sk))
739 			break;
740 		if (skb_queue_empty(&sk->sk_receive_queue))
741 			break;
742 		revents |= EPOLLIN | EPOLLRDNORM;
743 		break;
744 	case TIPC_DISCONNECTING:
745 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
746 		break;
747 	}
748 	return revents;
749 }
750 
751 /**
752  * tipc_sendmcast - send multicast message
753  * @sock: socket structure
754  * @seq: destination address
755  * @msg: message to send
756  * @dlen: length of data to send
757  * @timeout: timeout to wait for wakeup
758  *
759  * Called from function tipc_sendmsg(), which has done all sanity checks
760  * Returns the number of bytes sent on success, or errno
761  */
762 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
763 			  struct msghdr *msg, size_t dlen, long timeout)
764 {
765 	struct sock *sk = sock->sk;
766 	struct tipc_sock *tsk = tipc_sk(sk);
767 	struct tipc_msg *hdr = &tsk->phdr;
768 	struct net *net = sock_net(sk);
769 	int mtu = tipc_bcast_get_mtu(net);
770 	struct tipc_mc_method *method = &tsk->mc_method;
771 	struct sk_buff_head pkts;
772 	struct tipc_nlist dsts;
773 	int rc;
774 
775 	if (tsk->group)
776 		return -EACCES;
777 
778 	/* Block or return if any destination link is congested */
779 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
780 	if (unlikely(rc))
781 		return rc;
782 
783 	/* Lookup destination nodes */
784 	tipc_nlist_init(&dsts, tipc_own_addr(net));
785 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
786 				      seq->upper, &dsts);
787 	if (!dsts.local && !dsts.remote)
788 		return -EHOSTUNREACH;
789 
790 	/* Build message header */
791 	msg_set_type(hdr, TIPC_MCAST_MSG);
792 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
793 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
794 	msg_set_destport(hdr, 0);
795 	msg_set_destnode(hdr, 0);
796 	msg_set_nametype(hdr, seq->type);
797 	msg_set_namelower(hdr, seq->lower);
798 	msg_set_nameupper(hdr, seq->upper);
799 
800 	/* Build message as chain of buffers */
801 	skb_queue_head_init(&pkts);
802 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
803 
804 	/* Send message if build was successful */
805 	if (unlikely(rc == dlen))
806 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
807 				     &tsk->cong_link_cnt);
808 
809 	tipc_nlist_purge(&dsts);
810 
811 	return rc ? rc : dlen;
812 }
813 
814 /**
815  * tipc_send_group_msg - send a message to a member in the group
816  * @net: network namespace
817  * @m: message to send
818  * @mb: group member
819  * @dnode: destination node
820  * @dport: destination port
821  * @dlen: total length of message data
822  */
823 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
824 			       struct msghdr *m, struct tipc_member *mb,
825 			       u32 dnode, u32 dport, int dlen)
826 {
827 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
828 	struct tipc_mc_method *method = &tsk->mc_method;
829 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
830 	struct tipc_msg *hdr = &tsk->phdr;
831 	struct sk_buff_head pkts;
832 	int mtu, rc;
833 
834 	/* Complete message header */
835 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
836 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
837 	msg_set_destport(hdr, dport);
838 	msg_set_destnode(hdr, dnode);
839 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
840 
841 	/* Build message as chain of buffers */
842 	skb_queue_head_init(&pkts);
843 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
844 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
845 	if (unlikely(rc != dlen))
846 		return rc;
847 
848 	/* Send message */
849 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
850 	if (unlikely(rc == -ELINKCONG)) {
851 		tipc_dest_push(&tsk->cong_links, dnode, 0);
852 		tsk->cong_link_cnt++;
853 	}
854 
855 	/* Update send window */
856 	tipc_group_update_member(mb, blks);
857 
858 	/* A broadcast sent within next EXPIRE period must follow same path */
859 	method->rcast = true;
860 	method->mandatory = true;
861 	return dlen;
862 }
863 
864 /**
865  * tipc_send_group_unicast - send message to a member in the group
866  * @sock: socket structure
867  * @m: message to send
868  * @dlen: total length of message data
869  * @timeout: timeout to wait for wakeup
870  *
871  * Called from function tipc_sendmsg(), which has done all sanity checks
872  * Returns the number of bytes sent on success, or errno
873  */
874 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
875 				   int dlen, long timeout)
876 {
877 	struct sock *sk = sock->sk;
878 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
879 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
880 	struct tipc_sock *tsk = tipc_sk(sk);
881 	struct tipc_group *grp = tsk->group;
882 	struct net *net = sock_net(sk);
883 	struct tipc_member *mb = NULL;
884 	u32 node, port;
885 	int rc;
886 
887 	node = dest->addr.id.node;
888 	port = dest->addr.id.ref;
889 	if (!port && !node)
890 		return -EHOSTUNREACH;
891 
892 	/* Block or return if destination link or member is congested */
893 	rc = tipc_wait_for_cond(sock, &timeout,
894 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
895 				!tipc_group_cong(grp, node, port, blks, &mb));
896 	if (unlikely(rc))
897 		return rc;
898 
899 	if (unlikely(!mb))
900 		return -EHOSTUNREACH;
901 
902 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
903 
904 	return rc ? rc : dlen;
905 }
906 
907 /**
908  * tipc_send_group_anycast - send message to any member with given identity
909  * @sock: socket structure
910  * @m: message to send
911  * @dlen: total length of message data
912  * @timeout: timeout to wait for wakeup
913  *
914  * Called from function tipc_sendmsg(), which has done all sanity checks
915  * Returns the number of bytes sent on success, or errno
916  */
917 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
918 				   int dlen, long timeout)
919 {
920 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
921 	struct sock *sk = sock->sk;
922 	struct tipc_sock *tsk = tipc_sk(sk);
923 	struct list_head *cong_links = &tsk->cong_links;
924 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
925 	struct tipc_group *grp = tsk->group;
926 	struct tipc_msg *hdr = &tsk->phdr;
927 	struct tipc_member *first = NULL;
928 	struct tipc_member *mbr = NULL;
929 	struct net *net = sock_net(sk);
930 	u32 node, port, exclude;
931 	struct list_head dsts;
932 	u32 type, inst, scope;
933 	int lookups = 0;
934 	int dstcnt, rc;
935 	bool cong;
936 
937 	INIT_LIST_HEAD(&dsts);
938 
939 	type = msg_nametype(hdr);
940 	inst = dest->addr.name.name.instance;
941 	scope = msg_lookup_scope(hdr);
942 	exclude = tipc_group_exclude(grp);
943 
944 	while (++lookups < 4) {
945 		first = NULL;
946 
947 		/* Look for a non-congested destination member, if any */
948 		while (1) {
949 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
950 						 &dstcnt, exclude, false))
951 				return -EHOSTUNREACH;
952 			tipc_dest_pop(&dsts, &node, &port);
953 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
954 			if (!cong)
955 				break;
956 			if (mbr == first)
957 				break;
958 			if (!first)
959 				first = mbr;
960 		}
961 
962 		/* Start over if destination was not in member list */
963 		if (unlikely(!mbr))
964 			continue;
965 
966 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
967 			break;
968 
969 		/* Block or return if destination link or member is congested */
970 		rc = tipc_wait_for_cond(sock, &timeout,
971 					!tipc_dest_find(cong_links, node, 0) &&
972 					!tipc_group_cong(grp, node, port,
973 							 blks, &mbr));
974 		if (unlikely(rc))
975 			return rc;
976 
977 		/* Send, unless destination disappeared while waiting */
978 		if (likely(mbr))
979 			break;
980 	}
981 
982 	if (unlikely(lookups >= 4))
983 		return -EHOSTUNREACH;
984 
985 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
986 
987 	return rc ? rc : dlen;
988 }
989 
990 /**
991  * tipc_send_group_bcast - send message to all members in communication group
992  * @sk: socket structure
993  * @m: message to send
994  * @dlen: total length of message data
995  * @timeout: timeout to wait for wakeup
996  *
997  * Called from function tipc_sendmsg(), which has done all sanity checks
998  * Returns the number of bytes sent on success, or errno
999  */
1000 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1001 				 int dlen, long timeout)
1002 {
1003 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1004 	struct sock *sk = sock->sk;
1005 	struct net *net = sock_net(sk);
1006 	struct tipc_sock *tsk = tipc_sk(sk);
1007 	struct tipc_group *grp = tsk->group;
1008 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1009 	struct tipc_mc_method *method = &tsk->mc_method;
1010 	bool ack = method->mandatory && method->rcast;
1011 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1012 	struct tipc_msg *hdr = &tsk->phdr;
1013 	int mtu = tipc_bcast_get_mtu(net);
1014 	struct sk_buff_head pkts;
1015 	int rc = -EHOSTUNREACH;
1016 
1017 	if (!dsts->local && !dsts->remote)
1018 		return -EHOSTUNREACH;
1019 
1020 	/* Block or return if any destination link or member is congested */
1021 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1022 				!tipc_group_bc_cong(grp, blks));
1023 	if (unlikely(rc))
1024 		return rc;
1025 
1026 	/* Complete message header */
1027 	if (dest) {
1028 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1029 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1030 	} else {
1031 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1032 		msg_set_nameinst(hdr, 0);
1033 	}
1034 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1035 	msg_set_destport(hdr, 0);
1036 	msg_set_destnode(hdr, 0);
1037 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1038 
1039 	/* Avoid getting stuck with repeated forced replicasts */
1040 	msg_set_grp_bc_ack_req(hdr, ack);
1041 
1042 	/* Build message as chain of buffers */
1043 	skb_queue_head_init(&pkts);
1044 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1045 	if (unlikely(rc != dlen))
1046 		return rc;
1047 
1048 	/* Send message */
1049 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1050 	if (unlikely(rc))
1051 		return rc;
1052 
1053 	/* Update broadcast sequence number and send windows */
1054 	tipc_group_update_bc_members(tsk->group, blks, ack);
1055 
1056 	/* Broadcast link is now free to choose method for next broadcast */
1057 	method->mandatory = false;
1058 	method->expires = jiffies;
1059 
1060 	return dlen;
1061 }
1062 
1063 /**
1064  * tipc_send_group_mcast - send message to all members with given identity
1065  * @sock: socket structure
1066  * @m: message to send
1067  * @dlen: total length of message data
1068  * @timeout: timeout to wait for wakeup
1069  *
1070  * Called from function tipc_sendmsg(), which has done all sanity checks
1071  * Returns the number of bytes sent on success, or errno
1072  */
1073 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1074 				 int dlen, long timeout)
1075 {
1076 	struct sock *sk = sock->sk;
1077 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1078 	struct tipc_sock *tsk = tipc_sk(sk);
1079 	struct tipc_group *grp = tsk->group;
1080 	struct tipc_msg *hdr = &tsk->phdr;
1081 	struct net *net = sock_net(sk);
1082 	u32 type, inst, scope, exclude;
1083 	struct list_head dsts;
1084 	u32 dstcnt;
1085 
1086 	INIT_LIST_HEAD(&dsts);
1087 
1088 	type = msg_nametype(hdr);
1089 	inst = dest->addr.name.name.instance;
1090 	scope = msg_lookup_scope(hdr);
1091 	exclude = tipc_group_exclude(grp);
1092 
1093 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1094 				 &dstcnt, exclude, true))
1095 		return -EHOSTUNREACH;
1096 
1097 	if (dstcnt == 1) {
1098 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1099 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1100 	}
1101 
1102 	tipc_dest_list_purge(&dsts);
1103 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1104 }
1105 
1106 /**
1107  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1108  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1109  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1110  *
1111  * Multi-threaded: parallel calls with reference to same queues may occur
1112  */
1113 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1114 		       struct sk_buff_head *inputq)
1115 {
1116 	u32 self = tipc_own_addr(net);
1117 	u32 type, lower, upper, scope;
1118 	struct sk_buff *skb, *_skb;
1119 	u32 portid, onode;
1120 	struct sk_buff_head tmpq;
1121 	struct list_head dports;
1122 	struct tipc_msg *hdr;
1123 	int user, mtyp, hlen;
1124 	bool exact;
1125 
1126 	__skb_queue_head_init(&tmpq);
1127 	INIT_LIST_HEAD(&dports);
1128 
1129 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1130 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1131 		hdr = buf_msg(skb);
1132 		user = msg_user(hdr);
1133 		mtyp = msg_type(hdr);
1134 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1135 		onode = msg_orignode(hdr);
1136 		type = msg_nametype(hdr);
1137 
1138 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1139 			spin_lock_bh(&inputq->lock);
1140 			if (skb_peek(arrvq) == skb) {
1141 				__skb_dequeue(arrvq);
1142 				__skb_queue_tail(inputq, skb);
1143 			}
1144 			kfree_skb(skb);
1145 			spin_unlock_bh(&inputq->lock);
1146 			continue;
1147 		}
1148 
1149 		/* Group messages require exact scope match */
1150 		if (msg_in_group(hdr)) {
1151 			lower = 0;
1152 			upper = ~0;
1153 			scope = msg_lookup_scope(hdr);
1154 			exact = true;
1155 		} else {
1156 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1157 			if (onode == self)
1158 				scope = TIPC_NODE_SCOPE;
1159 			else
1160 				scope = TIPC_CLUSTER_SCOPE;
1161 			exact = false;
1162 			lower = msg_namelower(hdr);
1163 			upper = msg_nameupper(hdr);
1164 		}
1165 
1166 		/* Create destination port list: */
1167 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1168 				       scope, exact, &dports);
1169 
1170 		/* Clone message per destination */
1171 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1172 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1173 			if (_skb) {
1174 				msg_set_destport(buf_msg(_skb), portid);
1175 				__skb_queue_tail(&tmpq, _skb);
1176 				continue;
1177 			}
1178 			pr_warn("Failed to clone mcast rcv buffer\n");
1179 		}
1180 		/* Append to inputq if not already done by other thread */
1181 		spin_lock_bh(&inputq->lock);
1182 		if (skb_peek(arrvq) == skb) {
1183 			skb_queue_splice_tail_init(&tmpq, inputq);
1184 			kfree_skb(__skb_dequeue(arrvq));
1185 		}
1186 		spin_unlock_bh(&inputq->lock);
1187 		__skb_queue_purge(&tmpq);
1188 		kfree_skb(skb);
1189 	}
1190 	tipc_sk_rcv(net, inputq);
1191 }
1192 
1193 /**
1194  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1195  * @tsk: receiving socket
1196  * @skb: pointer to message buffer.
1197  */
1198 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1199 				   struct sk_buff_head *xmitq)
1200 {
1201 	struct tipc_msg *hdr = buf_msg(skb);
1202 	u32 onode = tsk_own_node(tsk);
1203 	struct sock *sk = &tsk->sk;
1204 	int mtyp = msg_type(hdr);
1205 	bool conn_cong;
1206 
1207 	/* Ignore if connection cannot be validated: */
1208 	if (!tsk_peer_msg(tsk, hdr))
1209 		goto exit;
1210 
1211 	if (unlikely(msg_errcode(hdr))) {
1212 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1213 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1214 				      tsk_peer_port(tsk));
1215 		sk->sk_state_change(sk);
1216 		goto exit;
1217 	}
1218 
1219 	tsk->probe_unacked = false;
1220 
1221 	if (mtyp == CONN_PROBE) {
1222 		msg_set_type(hdr, CONN_PROBE_REPLY);
1223 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1224 			__skb_queue_tail(xmitq, skb);
1225 		return;
1226 	} else if (mtyp == CONN_ACK) {
1227 		conn_cong = tsk_conn_cong(tsk);
1228 		tsk->snt_unacked -= msg_conn_ack(hdr);
1229 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1230 			tsk->snd_win = msg_adv_win(hdr);
1231 		if (conn_cong)
1232 			sk->sk_write_space(sk);
1233 	} else if (mtyp != CONN_PROBE_REPLY) {
1234 		pr_warn("Received unknown CONN_PROTO msg\n");
1235 	}
1236 exit:
1237 	kfree_skb(skb);
1238 }
1239 
1240 /**
1241  * tipc_sendmsg - send message in connectionless manner
1242  * @sock: socket structure
1243  * @m: message to send
1244  * @dsz: amount of user data to be sent
1245  *
1246  * Message must have an destination specified explicitly.
1247  * Used for SOCK_RDM and SOCK_DGRAM messages,
1248  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1249  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1250  *
1251  * Returns the number of bytes sent on success, or errno otherwise
1252  */
1253 static int tipc_sendmsg(struct socket *sock,
1254 			struct msghdr *m, size_t dsz)
1255 {
1256 	struct sock *sk = sock->sk;
1257 	int ret;
1258 
1259 	lock_sock(sk);
1260 	ret = __tipc_sendmsg(sock, m, dsz);
1261 	release_sock(sk);
1262 
1263 	return ret;
1264 }
1265 
1266 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1267 {
1268 	struct sock *sk = sock->sk;
1269 	struct net *net = sock_net(sk);
1270 	struct tipc_sock *tsk = tipc_sk(sk);
1271 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1272 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1273 	struct list_head *clinks = &tsk->cong_links;
1274 	bool syn = !tipc_sk_type_connectionless(sk);
1275 	struct tipc_group *grp = tsk->group;
1276 	struct tipc_msg *hdr = &tsk->phdr;
1277 	struct tipc_name_seq *seq;
1278 	struct sk_buff_head pkts;
1279 	u32 dport, dnode = 0;
1280 	u32 type, inst;
1281 	int mtu, rc;
1282 
1283 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1284 		return -EMSGSIZE;
1285 
1286 	if (likely(dest)) {
1287 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1288 			return -EINVAL;
1289 		if (unlikely(dest->family != AF_TIPC))
1290 			return -EINVAL;
1291 	}
1292 
1293 	if (grp) {
1294 		if (!dest)
1295 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1296 		if (dest->addrtype == TIPC_ADDR_NAME)
1297 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1298 		if (dest->addrtype == TIPC_ADDR_ID)
1299 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1300 		if (dest->addrtype == TIPC_ADDR_MCAST)
1301 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1302 		return -EINVAL;
1303 	}
1304 
1305 	if (unlikely(!dest)) {
1306 		dest = &tsk->peer;
1307 		if (!syn || dest->family != AF_TIPC)
1308 			return -EDESTADDRREQ;
1309 	}
1310 
1311 	if (unlikely(syn)) {
1312 		if (sk->sk_state == TIPC_LISTEN)
1313 			return -EPIPE;
1314 		if (sk->sk_state != TIPC_OPEN)
1315 			return -EISCONN;
1316 		if (tsk->published)
1317 			return -EOPNOTSUPP;
1318 		if (dest->addrtype == TIPC_ADDR_NAME) {
1319 			tsk->conn_type = dest->addr.name.name.type;
1320 			tsk->conn_instance = dest->addr.name.name.instance;
1321 		}
1322 	}
1323 
1324 	seq = &dest->addr.nameseq;
1325 	if (dest->addrtype == TIPC_ADDR_MCAST)
1326 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1327 
1328 	if (dest->addrtype == TIPC_ADDR_NAME) {
1329 		type = dest->addr.name.name.type;
1330 		inst = dest->addr.name.name.instance;
1331 		dnode = dest->addr.name.domain;
1332 		msg_set_type(hdr, TIPC_NAMED_MSG);
1333 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1334 		msg_set_nametype(hdr, type);
1335 		msg_set_nameinst(hdr, inst);
1336 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1337 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1338 		msg_set_destnode(hdr, dnode);
1339 		msg_set_destport(hdr, dport);
1340 		if (unlikely(!dport && !dnode))
1341 			return -EHOSTUNREACH;
1342 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1343 		dnode = dest->addr.id.node;
1344 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1345 		msg_set_lookup_scope(hdr, 0);
1346 		msg_set_destnode(hdr, dnode);
1347 		msg_set_destport(hdr, dest->addr.id.ref);
1348 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1349 	} else {
1350 		return -EINVAL;
1351 	}
1352 
1353 	/* Block or return if destination link is congested */
1354 	rc = tipc_wait_for_cond(sock, &timeout,
1355 				!tipc_dest_find(clinks, dnode, 0));
1356 	if (unlikely(rc))
1357 		return rc;
1358 
1359 	skb_queue_head_init(&pkts);
1360 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1361 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1362 	if (unlikely(rc != dlen))
1363 		return rc;
1364 
1365 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1366 	if (unlikely(rc == -ELINKCONG)) {
1367 		tipc_dest_push(clinks, dnode, 0);
1368 		tsk->cong_link_cnt++;
1369 		rc = 0;
1370 	}
1371 
1372 	if (unlikely(syn && !rc))
1373 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1374 
1375 	return rc ? rc : dlen;
1376 }
1377 
1378 /**
1379  * tipc_sendstream - send stream-oriented data
1380  * @sock: socket structure
1381  * @m: data to send
1382  * @dsz: total length of data to be transmitted
1383  *
1384  * Used for SOCK_STREAM data.
1385  *
1386  * Returns the number of bytes sent on success (or partial success),
1387  * or errno if no data sent
1388  */
1389 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1390 {
1391 	struct sock *sk = sock->sk;
1392 	int ret;
1393 
1394 	lock_sock(sk);
1395 	ret = __tipc_sendstream(sock, m, dsz);
1396 	release_sock(sk);
1397 
1398 	return ret;
1399 }
1400 
1401 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1402 {
1403 	struct sock *sk = sock->sk;
1404 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1405 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1406 	struct tipc_sock *tsk = tipc_sk(sk);
1407 	struct tipc_msg *hdr = &tsk->phdr;
1408 	struct net *net = sock_net(sk);
1409 	struct sk_buff_head pkts;
1410 	u32 dnode = tsk_peer_node(tsk);
1411 	int send, sent = 0;
1412 	int rc = 0;
1413 
1414 	skb_queue_head_init(&pkts);
1415 
1416 	if (unlikely(dlen > INT_MAX))
1417 		return -EMSGSIZE;
1418 
1419 	/* Handle implicit connection setup */
1420 	if (unlikely(dest)) {
1421 		rc = __tipc_sendmsg(sock, m, dlen);
1422 		if (dlen && dlen == rc) {
1423 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1424 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1425 		}
1426 		return rc;
1427 	}
1428 
1429 	do {
1430 		rc = tipc_wait_for_cond(sock, &timeout,
1431 					(!tsk->cong_link_cnt &&
1432 					 !tsk_conn_cong(tsk) &&
1433 					 tipc_sk_connected(sk)));
1434 		if (unlikely(rc))
1435 			break;
1436 
1437 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1438 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1439 		if (unlikely(rc != send))
1440 			break;
1441 
1442 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1443 		if (unlikely(rc == -ELINKCONG)) {
1444 			tsk->cong_link_cnt = 1;
1445 			rc = 0;
1446 		}
1447 		if (likely(!rc)) {
1448 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1449 			sent += send;
1450 		}
1451 	} while (sent < dlen && !rc);
1452 
1453 	return sent ? sent : rc;
1454 }
1455 
1456 /**
1457  * tipc_send_packet - send a connection-oriented message
1458  * @sock: socket structure
1459  * @m: message to send
1460  * @dsz: length of data to be transmitted
1461  *
1462  * Used for SOCK_SEQPACKET messages.
1463  *
1464  * Returns the number of bytes sent on success, or errno otherwise
1465  */
1466 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1467 {
1468 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1469 		return -EMSGSIZE;
1470 
1471 	return tipc_sendstream(sock, m, dsz);
1472 }
1473 
1474 /* tipc_sk_finish_conn - complete the setup of a connection
1475  */
1476 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1477 				u32 peer_node)
1478 {
1479 	struct sock *sk = &tsk->sk;
1480 	struct net *net = sock_net(sk);
1481 	struct tipc_msg *msg = &tsk->phdr;
1482 
1483 	msg_set_destnode(msg, peer_node);
1484 	msg_set_destport(msg, peer_port);
1485 	msg_set_type(msg, TIPC_CONN_MSG);
1486 	msg_set_lookup_scope(msg, 0);
1487 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1488 
1489 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1490 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1491 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1492 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1493 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1494 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1495 		return;
1496 
1497 	/* Fall back to message based flow control */
1498 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1499 	tsk->snd_win = FLOWCTL_MSG_WIN;
1500 }
1501 
1502 /**
1503  * tipc_sk_set_orig_addr - capture sender's address for received message
1504  * @m: descriptor for message info
1505  * @hdr: received message header
1506  *
1507  * Note: Address is not captured if not requested by receiver.
1508  */
1509 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1510 {
1511 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1512 	struct tipc_msg *hdr = buf_msg(skb);
1513 
1514 	if (!srcaddr)
1515 		return;
1516 
1517 	srcaddr->sock.family = AF_TIPC;
1518 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1519 	srcaddr->sock.scope = 0;
1520 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1521 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1522 	srcaddr->sock.addr.name.domain = 0;
1523 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1524 
1525 	if (!msg_in_group(hdr))
1526 		return;
1527 
1528 	/* Group message users may also want to know sending member's id */
1529 	srcaddr->member.family = AF_TIPC;
1530 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1531 	srcaddr->member.scope = 0;
1532 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1533 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1534 	srcaddr->member.addr.name.domain = 0;
1535 	m->msg_namelen = sizeof(*srcaddr);
1536 }
1537 
1538 /**
1539  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1540  * @m: descriptor for message info
1541  * @msg: received message header
1542  * @tsk: TIPC port associated with message
1543  *
1544  * Note: Ancillary data is not captured if not requested by receiver.
1545  *
1546  * Returns 0 if successful, otherwise errno
1547  */
1548 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1549 				 struct tipc_sock *tsk)
1550 {
1551 	u32 anc_data[3];
1552 	u32 err;
1553 	u32 dest_type;
1554 	int has_name;
1555 	int res;
1556 
1557 	if (likely(m->msg_controllen == 0))
1558 		return 0;
1559 
1560 	/* Optionally capture errored message object(s) */
1561 	err = msg ? msg_errcode(msg) : 0;
1562 	if (unlikely(err)) {
1563 		anc_data[0] = err;
1564 		anc_data[1] = msg_data_sz(msg);
1565 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1566 		if (res)
1567 			return res;
1568 		if (anc_data[1]) {
1569 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1570 				       msg_data(msg));
1571 			if (res)
1572 				return res;
1573 		}
1574 	}
1575 
1576 	/* Optionally capture message destination object */
1577 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1578 	switch (dest_type) {
1579 	case TIPC_NAMED_MSG:
1580 		has_name = 1;
1581 		anc_data[0] = msg_nametype(msg);
1582 		anc_data[1] = msg_namelower(msg);
1583 		anc_data[2] = msg_namelower(msg);
1584 		break;
1585 	case TIPC_MCAST_MSG:
1586 		has_name = 1;
1587 		anc_data[0] = msg_nametype(msg);
1588 		anc_data[1] = msg_namelower(msg);
1589 		anc_data[2] = msg_nameupper(msg);
1590 		break;
1591 	case TIPC_CONN_MSG:
1592 		has_name = (tsk->conn_type != 0);
1593 		anc_data[0] = tsk->conn_type;
1594 		anc_data[1] = tsk->conn_instance;
1595 		anc_data[2] = tsk->conn_instance;
1596 		break;
1597 	default:
1598 		has_name = 0;
1599 	}
1600 	if (has_name) {
1601 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1602 		if (res)
1603 			return res;
1604 	}
1605 
1606 	return 0;
1607 }
1608 
1609 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1610 {
1611 	struct sock *sk = &tsk->sk;
1612 	struct net *net = sock_net(sk);
1613 	struct sk_buff *skb = NULL;
1614 	struct tipc_msg *msg;
1615 	u32 peer_port = tsk_peer_port(tsk);
1616 	u32 dnode = tsk_peer_node(tsk);
1617 
1618 	if (!tipc_sk_connected(sk))
1619 		return;
1620 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1621 			      dnode, tsk_own_node(tsk), peer_port,
1622 			      tsk->portid, TIPC_OK);
1623 	if (!skb)
1624 		return;
1625 	msg = buf_msg(skb);
1626 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1627 	tsk->rcv_unacked = 0;
1628 
1629 	/* Adjust to and advertize the correct window limit */
1630 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1631 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1632 		msg_set_adv_win(msg, tsk->rcv_win);
1633 	}
1634 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1635 }
1636 
1637 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1638 {
1639 	struct sock *sk = sock->sk;
1640 	DEFINE_WAIT(wait);
1641 	long timeo = *timeop;
1642 	int err = sock_error(sk);
1643 
1644 	if (err)
1645 		return err;
1646 
1647 	for (;;) {
1648 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1649 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1650 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1651 				err = -ENOTCONN;
1652 				break;
1653 			}
1654 			release_sock(sk);
1655 			timeo = schedule_timeout(timeo);
1656 			lock_sock(sk);
1657 		}
1658 		err = 0;
1659 		if (!skb_queue_empty(&sk->sk_receive_queue))
1660 			break;
1661 		err = -EAGAIN;
1662 		if (!timeo)
1663 			break;
1664 		err = sock_intr_errno(timeo);
1665 		if (signal_pending(current))
1666 			break;
1667 
1668 		err = sock_error(sk);
1669 		if (err)
1670 			break;
1671 	}
1672 	finish_wait(sk_sleep(sk), &wait);
1673 	*timeop = timeo;
1674 	return err;
1675 }
1676 
1677 /**
1678  * tipc_recvmsg - receive packet-oriented message
1679  * @m: descriptor for message info
1680  * @buflen: length of user buffer area
1681  * @flags: receive flags
1682  *
1683  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1684  * If the complete message doesn't fit in user area, truncate it.
1685  *
1686  * Returns size of returned message data, errno otherwise
1687  */
1688 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1689 			size_t buflen,	int flags)
1690 {
1691 	struct sock *sk = sock->sk;
1692 	bool connected = !tipc_sk_type_connectionless(sk);
1693 	struct tipc_sock *tsk = tipc_sk(sk);
1694 	int rc, err, hlen, dlen, copy;
1695 	struct sk_buff_head xmitq;
1696 	struct tipc_msg *hdr;
1697 	struct sk_buff *skb;
1698 	bool grp_evt;
1699 	long timeout;
1700 
1701 	/* Catch invalid receive requests */
1702 	if (unlikely(!buflen))
1703 		return -EINVAL;
1704 
1705 	lock_sock(sk);
1706 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1707 		rc = -ENOTCONN;
1708 		goto exit;
1709 	}
1710 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1711 
1712 	/* Step rcv queue to first msg with data or error; wait if necessary */
1713 	do {
1714 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1715 		if (unlikely(rc))
1716 			goto exit;
1717 		skb = skb_peek(&sk->sk_receive_queue);
1718 		hdr = buf_msg(skb);
1719 		dlen = msg_data_sz(hdr);
1720 		hlen = msg_hdr_sz(hdr);
1721 		err = msg_errcode(hdr);
1722 		grp_evt = msg_is_grp_evt(hdr);
1723 		if (likely(dlen || err))
1724 			break;
1725 		tsk_advance_rx_queue(sk);
1726 	} while (1);
1727 
1728 	/* Collect msg meta data, including error code and rejected data */
1729 	tipc_sk_set_orig_addr(m, skb);
1730 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1731 	if (unlikely(rc))
1732 		goto exit;
1733 
1734 	/* Capture data if non-error msg, otherwise just set return value */
1735 	if (likely(!err)) {
1736 		copy = min_t(int, dlen, buflen);
1737 		if (unlikely(copy != dlen))
1738 			m->msg_flags |= MSG_TRUNC;
1739 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1740 	} else {
1741 		copy = 0;
1742 		rc = 0;
1743 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1744 			rc = -ECONNRESET;
1745 	}
1746 	if (unlikely(rc))
1747 		goto exit;
1748 
1749 	/* Mark message as group event if applicable */
1750 	if (unlikely(grp_evt)) {
1751 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1752 			m->msg_flags |= MSG_EOR;
1753 		m->msg_flags |= MSG_OOB;
1754 		copy = 0;
1755 	}
1756 
1757 	/* Caption of data or error code/rejected data was successful */
1758 	if (unlikely(flags & MSG_PEEK))
1759 		goto exit;
1760 
1761 	/* Send group flow control advertisement when applicable */
1762 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1763 		skb_queue_head_init(&xmitq);
1764 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1765 					  msg_orignode(hdr), msg_origport(hdr),
1766 					  &xmitq);
1767 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1768 	}
1769 
1770 	tsk_advance_rx_queue(sk);
1771 
1772 	if (likely(!connected))
1773 		goto exit;
1774 
1775 	/* Send connection flow control advertisement when applicable */
1776 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1777 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1778 		tipc_sk_send_ack(tsk);
1779 exit:
1780 	release_sock(sk);
1781 	return rc ? rc : copy;
1782 }
1783 
1784 /**
1785  * tipc_recvstream - receive stream-oriented data
1786  * @m: descriptor for message info
1787  * @buflen: total size of user buffer area
1788  * @flags: receive flags
1789  *
1790  * Used for SOCK_STREAM messages only.  If not enough data is available
1791  * will optionally wait for more; never truncates data.
1792  *
1793  * Returns size of returned message data, errno otherwise
1794  */
1795 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1796 			   size_t buflen, int flags)
1797 {
1798 	struct sock *sk = sock->sk;
1799 	struct tipc_sock *tsk = tipc_sk(sk);
1800 	struct sk_buff *skb;
1801 	struct tipc_msg *hdr;
1802 	struct tipc_skb_cb *skb_cb;
1803 	bool peek = flags & MSG_PEEK;
1804 	int offset, required, copy, copied = 0;
1805 	int hlen, dlen, err, rc;
1806 	long timeout;
1807 
1808 	/* Catch invalid receive attempts */
1809 	if (unlikely(!buflen))
1810 		return -EINVAL;
1811 
1812 	lock_sock(sk);
1813 
1814 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1815 		rc = -ENOTCONN;
1816 		goto exit;
1817 	}
1818 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1819 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1820 
1821 	do {
1822 		/* Look at first msg in receive queue; wait if necessary */
1823 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1824 		if (unlikely(rc))
1825 			break;
1826 		skb = skb_peek(&sk->sk_receive_queue);
1827 		skb_cb = TIPC_SKB_CB(skb);
1828 		hdr = buf_msg(skb);
1829 		dlen = msg_data_sz(hdr);
1830 		hlen = msg_hdr_sz(hdr);
1831 		err = msg_errcode(hdr);
1832 
1833 		/* Discard any empty non-errored (SYN-) message */
1834 		if (unlikely(!dlen && !err)) {
1835 			tsk_advance_rx_queue(sk);
1836 			continue;
1837 		}
1838 
1839 		/* Collect msg meta data, incl. error code and rejected data */
1840 		if (!copied) {
1841 			tipc_sk_set_orig_addr(m, skb);
1842 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1843 			if (rc)
1844 				break;
1845 		}
1846 
1847 		/* Copy data if msg ok, otherwise return error/partial data */
1848 		if (likely(!err)) {
1849 			offset = skb_cb->bytes_read;
1850 			copy = min_t(int, dlen - offset, buflen - copied);
1851 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1852 			if (unlikely(rc))
1853 				break;
1854 			copied += copy;
1855 			offset += copy;
1856 			if (unlikely(offset < dlen)) {
1857 				if (!peek)
1858 					skb_cb->bytes_read = offset;
1859 				break;
1860 			}
1861 		} else {
1862 			rc = 0;
1863 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1864 				rc = -ECONNRESET;
1865 			if (copied || rc)
1866 				break;
1867 		}
1868 
1869 		if (unlikely(peek))
1870 			break;
1871 
1872 		tsk_advance_rx_queue(sk);
1873 
1874 		/* Send connection flow control advertisement when applicable */
1875 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1876 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1877 			tipc_sk_send_ack(tsk);
1878 
1879 		/* Exit if all requested data or FIN/error received */
1880 		if (copied == buflen || err)
1881 			break;
1882 
1883 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1884 exit:
1885 	release_sock(sk);
1886 	return copied ? copied : rc;
1887 }
1888 
1889 /**
1890  * tipc_write_space - wake up thread if port congestion is released
1891  * @sk: socket
1892  */
1893 static void tipc_write_space(struct sock *sk)
1894 {
1895 	struct socket_wq *wq;
1896 
1897 	rcu_read_lock();
1898 	wq = rcu_dereference(sk->sk_wq);
1899 	if (skwq_has_sleeper(wq))
1900 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1901 						EPOLLWRNORM | EPOLLWRBAND);
1902 	rcu_read_unlock();
1903 }
1904 
1905 /**
1906  * tipc_data_ready - wake up threads to indicate messages have been received
1907  * @sk: socket
1908  * @len: the length of messages
1909  */
1910 static void tipc_data_ready(struct sock *sk)
1911 {
1912 	struct socket_wq *wq;
1913 
1914 	rcu_read_lock();
1915 	wq = rcu_dereference(sk->sk_wq);
1916 	if (skwq_has_sleeper(wq))
1917 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1918 						EPOLLRDNORM | EPOLLRDBAND);
1919 	rcu_read_unlock();
1920 }
1921 
1922 static void tipc_sock_destruct(struct sock *sk)
1923 {
1924 	__skb_queue_purge(&sk->sk_receive_queue);
1925 }
1926 
1927 static void tipc_sk_proto_rcv(struct sock *sk,
1928 			      struct sk_buff_head *inputq,
1929 			      struct sk_buff_head *xmitq)
1930 {
1931 	struct sk_buff *skb = __skb_dequeue(inputq);
1932 	struct tipc_sock *tsk = tipc_sk(sk);
1933 	struct tipc_msg *hdr = buf_msg(skb);
1934 	struct tipc_group *grp = tsk->group;
1935 	bool wakeup = false;
1936 
1937 	switch (msg_user(hdr)) {
1938 	case CONN_MANAGER:
1939 		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1940 		return;
1941 	case SOCK_WAKEUP:
1942 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1943 		tsk->cong_link_cnt--;
1944 		wakeup = true;
1945 		break;
1946 	case GROUP_PROTOCOL:
1947 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1948 		break;
1949 	case TOP_SRV:
1950 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1951 				      hdr, inputq, xmitq);
1952 		break;
1953 	default:
1954 		break;
1955 	}
1956 
1957 	if (wakeup)
1958 		sk->sk_write_space(sk);
1959 
1960 	kfree_skb(skb);
1961 }
1962 
1963 /**
1964  * tipc_filter_connect - Handle incoming message for a connection-based socket
1965  * @tsk: TIPC socket
1966  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1967  *
1968  * Returns true if everything ok, false otherwise
1969  */
1970 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1971 {
1972 	struct sock *sk = &tsk->sk;
1973 	struct net *net = sock_net(sk);
1974 	struct tipc_msg *hdr = buf_msg(skb);
1975 	u32 pport = msg_origport(hdr);
1976 	u32 pnode = msg_orignode(hdr);
1977 
1978 	if (unlikely(msg_mcast(hdr)))
1979 		return false;
1980 
1981 	switch (sk->sk_state) {
1982 	case TIPC_CONNECTING:
1983 		/* Accept only ACK or NACK message */
1984 		if (unlikely(!msg_connected(hdr))) {
1985 			if (pport != tsk_peer_port(tsk) ||
1986 			    pnode != tsk_peer_node(tsk))
1987 				return false;
1988 
1989 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1990 			sk->sk_err = ECONNREFUSED;
1991 			sk->sk_state_change(sk);
1992 			return true;
1993 		}
1994 
1995 		if (unlikely(msg_errcode(hdr))) {
1996 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1997 			sk->sk_err = ECONNREFUSED;
1998 			sk->sk_state_change(sk);
1999 			return true;
2000 		}
2001 
2002 		if (unlikely(!msg_isdata(hdr))) {
2003 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2004 			sk->sk_err = EINVAL;
2005 			sk->sk_state_change(sk);
2006 			return true;
2007 		}
2008 
2009 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2010 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
2011 
2012 		/* If 'ACK+' message, add to socket receive queue */
2013 		if (msg_data_sz(hdr))
2014 			return true;
2015 
2016 		/* If empty 'ACK-' message, wake up sleeping connect() */
2017 		sk->sk_data_ready(sk);
2018 
2019 		/* 'ACK-' message is neither accepted nor rejected: */
2020 		msg_set_dest_droppable(hdr, 1);
2021 		return false;
2022 
2023 	case TIPC_OPEN:
2024 	case TIPC_DISCONNECTING:
2025 		break;
2026 	case TIPC_LISTEN:
2027 		/* Accept only SYN message */
2028 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2029 			return true;
2030 		break;
2031 	case TIPC_ESTABLISHED:
2032 		/* Accept only connection-based messages sent by peer */
2033 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
2034 			return false;
2035 
2036 		if (unlikely(msg_errcode(hdr))) {
2037 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2038 			/* Let timer expire on it's own */
2039 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2040 					      tsk->portid);
2041 			sk->sk_state_change(sk);
2042 		}
2043 		return true;
2044 	default:
2045 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2046 	}
2047 
2048 	return false;
2049 }
2050 
2051 /**
2052  * rcvbuf_limit - get proper overload limit of socket receive queue
2053  * @sk: socket
2054  * @skb: message
2055  *
2056  * For connection oriented messages, irrespective of importance,
2057  * default queue limit is 2 MB.
2058  *
2059  * For connectionless messages, queue limits are based on message
2060  * importance as follows:
2061  *
2062  * TIPC_LOW_IMPORTANCE       (2 MB)
2063  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2064  * TIPC_HIGH_IMPORTANCE      (8 MB)
2065  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2066  *
2067  * Returns overload limit according to corresponding message importance
2068  */
2069 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2070 {
2071 	struct tipc_sock *tsk = tipc_sk(sk);
2072 	struct tipc_msg *hdr = buf_msg(skb);
2073 
2074 	if (unlikely(msg_in_group(hdr)))
2075 		return sk->sk_rcvbuf;
2076 
2077 	if (unlikely(!msg_connected(hdr)))
2078 		return sk->sk_rcvbuf << msg_importance(hdr);
2079 
2080 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2081 		return sk->sk_rcvbuf;
2082 
2083 	return FLOWCTL_MSG_LIM;
2084 }
2085 
2086 /**
2087  * tipc_sk_filter_rcv - validate incoming message
2088  * @sk: socket
2089  * @skb: pointer to message.
2090  *
2091  * Enqueues message on receive queue if acceptable; optionally handles
2092  * disconnect indication for a connected socket.
2093  *
2094  * Called with socket lock already taken
2095  *
2096  */
2097 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2098 			       struct sk_buff_head *xmitq)
2099 {
2100 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2101 	struct tipc_sock *tsk = tipc_sk(sk);
2102 	struct tipc_group *grp = tsk->group;
2103 	struct tipc_msg *hdr = buf_msg(skb);
2104 	struct net *net = sock_net(sk);
2105 	struct sk_buff_head inputq;
2106 	int limit, err = TIPC_OK;
2107 
2108 	TIPC_SKB_CB(skb)->bytes_read = 0;
2109 	__skb_queue_head_init(&inputq);
2110 	__skb_queue_tail(&inputq, skb);
2111 
2112 	if (unlikely(!msg_isdata(hdr)))
2113 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2114 
2115 	if (unlikely(grp))
2116 		tipc_group_filter_msg(grp, &inputq, xmitq);
2117 
2118 	/* Validate and add to receive buffer if there is space */
2119 	while ((skb = __skb_dequeue(&inputq))) {
2120 		hdr = buf_msg(skb);
2121 		limit = rcvbuf_limit(sk, skb);
2122 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2123 		    (!sk_conn && msg_connected(hdr)) ||
2124 		    (!grp && msg_in_group(hdr)))
2125 			err = TIPC_ERR_NO_PORT;
2126 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2127 			atomic_inc(&sk->sk_drops);
2128 			err = TIPC_ERR_OVERLOAD;
2129 		}
2130 
2131 		if (unlikely(err)) {
2132 			tipc_skb_reject(net, err, skb, xmitq);
2133 			err = TIPC_OK;
2134 			continue;
2135 		}
2136 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2137 		skb_set_owner_r(skb, sk);
2138 		sk->sk_data_ready(sk);
2139 	}
2140 }
2141 
2142 /**
2143  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2144  * @sk: socket
2145  * @skb: message
2146  *
2147  * Caller must hold socket lock
2148  */
2149 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2150 {
2151 	unsigned int before = sk_rmem_alloc_get(sk);
2152 	struct sk_buff_head xmitq;
2153 	unsigned int added;
2154 
2155 	__skb_queue_head_init(&xmitq);
2156 
2157 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2158 	added = sk_rmem_alloc_get(sk) - before;
2159 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2160 
2161 	/* Send pending response/rejected messages, if any */
2162 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2163 	return 0;
2164 }
2165 
2166 /**
2167  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2168  *                   inputq and try adding them to socket or backlog queue
2169  * @inputq: list of incoming buffers with potentially different destinations
2170  * @sk: socket where the buffers should be enqueued
2171  * @dport: port number for the socket
2172  *
2173  * Caller must hold socket lock
2174  */
2175 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2176 			    u32 dport, struct sk_buff_head *xmitq)
2177 {
2178 	unsigned long time_limit = jiffies + 2;
2179 	struct sk_buff *skb;
2180 	unsigned int lim;
2181 	atomic_t *dcnt;
2182 	u32 onode;
2183 
2184 	while (skb_queue_len(inputq)) {
2185 		if (unlikely(time_after_eq(jiffies, time_limit)))
2186 			return;
2187 
2188 		skb = tipc_skb_dequeue(inputq, dport);
2189 		if (unlikely(!skb))
2190 			return;
2191 
2192 		/* Add message directly to receive queue if possible */
2193 		if (!sock_owned_by_user(sk)) {
2194 			tipc_sk_filter_rcv(sk, skb, xmitq);
2195 			continue;
2196 		}
2197 
2198 		/* Try backlog, compensating for double-counted bytes */
2199 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2200 		if (!sk->sk_backlog.len)
2201 			atomic_set(dcnt, 0);
2202 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2203 		if (likely(!sk_add_backlog(sk, skb, lim)))
2204 			continue;
2205 
2206 		/* Overload => reject message back to sender */
2207 		onode = tipc_own_addr(sock_net(sk));
2208 		atomic_inc(&sk->sk_drops);
2209 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2210 			__skb_queue_tail(xmitq, skb);
2211 		break;
2212 	}
2213 }
2214 
2215 /**
2216  * tipc_sk_rcv - handle a chain of incoming buffers
2217  * @inputq: buffer list containing the buffers
2218  * Consumes all buffers in list until inputq is empty
2219  * Note: may be called in multiple threads referring to the same queue
2220  */
2221 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2222 {
2223 	struct sk_buff_head xmitq;
2224 	u32 dnode, dport = 0;
2225 	int err;
2226 	struct tipc_sock *tsk;
2227 	struct sock *sk;
2228 	struct sk_buff *skb;
2229 
2230 	__skb_queue_head_init(&xmitq);
2231 	while (skb_queue_len(inputq)) {
2232 		dport = tipc_skb_peek_port(inputq, dport);
2233 		tsk = tipc_sk_lookup(net, dport);
2234 
2235 		if (likely(tsk)) {
2236 			sk = &tsk->sk;
2237 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2238 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2239 				spin_unlock_bh(&sk->sk_lock.slock);
2240 			}
2241 			/* Send pending response/rejected messages, if any */
2242 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2243 			sock_put(sk);
2244 			continue;
2245 		}
2246 		/* No destination socket => dequeue skb if still there */
2247 		skb = tipc_skb_dequeue(inputq, dport);
2248 		if (!skb)
2249 			return;
2250 
2251 		/* Try secondary lookup if unresolved named message */
2252 		err = TIPC_ERR_NO_PORT;
2253 		if (tipc_msg_lookup_dest(net, skb, &err))
2254 			goto xmit;
2255 
2256 		/* Prepare for message rejection */
2257 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2258 			continue;
2259 xmit:
2260 		dnode = msg_destnode(buf_msg(skb));
2261 		tipc_node_xmit_skb(net, skb, dnode, dport);
2262 	}
2263 }
2264 
2265 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2266 {
2267 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2268 	struct sock *sk = sock->sk;
2269 	int done;
2270 
2271 	do {
2272 		int err = sock_error(sk);
2273 		if (err)
2274 			return err;
2275 		if (!*timeo_p)
2276 			return -ETIMEDOUT;
2277 		if (signal_pending(current))
2278 			return sock_intr_errno(*timeo_p);
2279 
2280 		add_wait_queue(sk_sleep(sk), &wait);
2281 		done = sk_wait_event(sk, timeo_p,
2282 				     sk->sk_state != TIPC_CONNECTING, &wait);
2283 		remove_wait_queue(sk_sleep(sk), &wait);
2284 	} while (!done);
2285 	return 0;
2286 }
2287 
2288 /**
2289  * tipc_connect - establish a connection to another TIPC port
2290  * @sock: socket structure
2291  * @dest: socket address for destination port
2292  * @destlen: size of socket address data structure
2293  * @flags: file-related flags associated with socket
2294  *
2295  * Returns 0 on success, errno otherwise
2296  */
2297 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2298 			int destlen, int flags)
2299 {
2300 	struct sock *sk = sock->sk;
2301 	struct tipc_sock *tsk = tipc_sk(sk);
2302 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2303 	struct msghdr m = {NULL,};
2304 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2305 	int previous;
2306 	int res = 0;
2307 
2308 	if (destlen != sizeof(struct sockaddr_tipc))
2309 		return -EINVAL;
2310 
2311 	lock_sock(sk);
2312 
2313 	if (tsk->group) {
2314 		res = -EINVAL;
2315 		goto exit;
2316 	}
2317 
2318 	if (dst->family == AF_UNSPEC) {
2319 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2320 		if (!tipc_sk_type_connectionless(sk))
2321 			res = -EINVAL;
2322 		goto exit;
2323 	} else if (dst->family != AF_TIPC) {
2324 		res = -EINVAL;
2325 	}
2326 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2327 		res = -EINVAL;
2328 	if (res)
2329 		goto exit;
2330 
2331 	/* DGRAM/RDM connect(), just save the destaddr */
2332 	if (tipc_sk_type_connectionless(sk)) {
2333 		memcpy(&tsk->peer, dest, destlen);
2334 		goto exit;
2335 	}
2336 
2337 	previous = sk->sk_state;
2338 
2339 	switch (sk->sk_state) {
2340 	case TIPC_OPEN:
2341 		/* Send a 'SYN-' to destination */
2342 		m.msg_name = dest;
2343 		m.msg_namelen = destlen;
2344 
2345 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2346 		 * indicate send_msg() is never blocked.
2347 		 */
2348 		if (!timeout)
2349 			m.msg_flags = MSG_DONTWAIT;
2350 
2351 		res = __tipc_sendmsg(sock, &m, 0);
2352 		if ((res < 0) && (res != -EWOULDBLOCK))
2353 			goto exit;
2354 
2355 		/* Just entered TIPC_CONNECTING state; the only
2356 		 * difference is that return value in non-blocking
2357 		 * case is EINPROGRESS, rather than EALREADY.
2358 		 */
2359 		res = -EINPROGRESS;
2360 		/* fall thru' */
2361 	case TIPC_CONNECTING:
2362 		if (!timeout) {
2363 			if (previous == TIPC_CONNECTING)
2364 				res = -EALREADY;
2365 			goto exit;
2366 		}
2367 		timeout = msecs_to_jiffies(timeout);
2368 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2369 		res = tipc_wait_for_connect(sock, &timeout);
2370 		break;
2371 	case TIPC_ESTABLISHED:
2372 		res = -EISCONN;
2373 		break;
2374 	default:
2375 		res = -EINVAL;
2376 	}
2377 
2378 exit:
2379 	release_sock(sk);
2380 	return res;
2381 }
2382 
2383 /**
2384  * tipc_listen - allow socket to listen for incoming connections
2385  * @sock: socket structure
2386  * @len: (unused)
2387  *
2388  * Returns 0 on success, errno otherwise
2389  */
2390 static int tipc_listen(struct socket *sock, int len)
2391 {
2392 	struct sock *sk = sock->sk;
2393 	int res;
2394 
2395 	lock_sock(sk);
2396 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2397 	release_sock(sk);
2398 
2399 	return res;
2400 }
2401 
2402 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2403 {
2404 	struct sock *sk = sock->sk;
2405 	DEFINE_WAIT(wait);
2406 	int err;
2407 
2408 	/* True wake-one mechanism for incoming connections: only
2409 	 * one process gets woken up, not the 'whole herd'.
2410 	 * Since we do not 'race & poll' for established sockets
2411 	 * anymore, the common case will execute the loop only once.
2412 	*/
2413 	for (;;) {
2414 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2415 					  TASK_INTERRUPTIBLE);
2416 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2417 			release_sock(sk);
2418 			timeo = schedule_timeout(timeo);
2419 			lock_sock(sk);
2420 		}
2421 		err = 0;
2422 		if (!skb_queue_empty(&sk->sk_receive_queue))
2423 			break;
2424 		err = -EAGAIN;
2425 		if (!timeo)
2426 			break;
2427 		err = sock_intr_errno(timeo);
2428 		if (signal_pending(current))
2429 			break;
2430 	}
2431 	finish_wait(sk_sleep(sk), &wait);
2432 	return err;
2433 }
2434 
2435 /**
2436  * tipc_accept - wait for connection request
2437  * @sock: listening socket
2438  * @newsock: new socket that is to be connected
2439  * @flags: file-related flags associated with socket
2440  *
2441  * Returns 0 on success, errno otherwise
2442  */
2443 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2444 		       bool kern)
2445 {
2446 	struct sock *new_sk, *sk = sock->sk;
2447 	struct sk_buff *buf;
2448 	struct tipc_sock *new_tsock;
2449 	struct tipc_msg *msg;
2450 	long timeo;
2451 	int res;
2452 
2453 	lock_sock(sk);
2454 
2455 	if (sk->sk_state != TIPC_LISTEN) {
2456 		res = -EINVAL;
2457 		goto exit;
2458 	}
2459 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2460 	res = tipc_wait_for_accept(sock, timeo);
2461 	if (res)
2462 		goto exit;
2463 
2464 	buf = skb_peek(&sk->sk_receive_queue);
2465 
2466 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2467 	if (res)
2468 		goto exit;
2469 	security_sk_clone(sock->sk, new_sock->sk);
2470 
2471 	new_sk = new_sock->sk;
2472 	new_tsock = tipc_sk(new_sk);
2473 	msg = buf_msg(buf);
2474 
2475 	/* we lock on new_sk; but lockdep sees the lock on sk */
2476 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2477 
2478 	/*
2479 	 * Reject any stray messages received by new socket
2480 	 * before the socket lock was taken (very, very unlikely)
2481 	 */
2482 	tsk_rej_rx_queue(new_sk);
2483 
2484 	/* Connect new socket to it's peer */
2485 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2486 
2487 	tsk_set_importance(new_tsock, msg_importance(msg));
2488 	if (msg_named(msg)) {
2489 		new_tsock->conn_type = msg_nametype(msg);
2490 		new_tsock->conn_instance = msg_nameinst(msg);
2491 	}
2492 
2493 	/*
2494 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2495 	 * Respond to 'SYN+' by queuing it on new socket.
2496 	 */
2497 	if (!msg_data_sz(msg)) {
2498 		struct msghdr m = {NULL,};
2499 
2500 		tsk_advance_rx_queue(sk);
2501 		__tipc_sendstream(new_sock, &m, 0);
2502 	} else {
2503 		__skb_dequeue(&sk->sk_receive_queue);
2504 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2505 		skb_set_owner_r(buf, new_sk);
2506 	}
2507 	release_sock(new_sk);
2508 exit:
2509 	release_sock(sk);
2510 	return res;
2511 }
2512 
2513 /**
2514  * tipc_shutdown - shutdown socket connection
2515  * @sock: socket structure
2516  * @how: direction to close (must be SHUT_RDWR)
2517  *
2518  * Terminates connection (if necessary), then purges socket's receive queue.
2519  *
2520  * Returns 0 on success, errno otherwise
2521  */
2522 static int tipc_shutdown(struct socket *sock, int how)
2523 {
2524 	struct sock *sk = sock->sk;
2525 	int res;
2526 
2527 	if (how != SHUT_RDWR)
2528 		return -EINVAL;
2529 
2530 	lock_sock(sk);
2531 
2532 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2533 	sk->sk_shutdown = SEND_SHUTDOWN;
2534 
2535 	if (sk->sk_state == TIPC_DISCONNECTING) {
2536 		/* Discard any unreceived messages */
2537 		__skb_queue_purge(&sk->sk_receive_queue);
2538 
2539 		/* Wake up anyone sleeping in poll */
2540 		sk->sk_state_change(sk);
2541 		res = 0;
2542 	} else {
2543 		res = -ENOTCONN;
2544 	}
2545 
2546 	release_sock(sk);
2547 	return res;
2548 }
2549 
2550 static void tipc_sk_timeout(struct timer_list *t)
2551 {
2552 	struct sock *sk = from_timer(sk, t, sk_timer);
2553 	struct tipc_sock *tsk = tipc_sk(sk);
2554 	u32 peer_port = tsk_peer_port(tsk);
2555 	u32 peer_node = tsk_peer_node(tsk);
2556 	u32 own_node = tsk_own_node(tsk);
2557 	u32 own_port = tsk->portid;
2558 	struct net *net = sock_net(sk);
2559 	struct sk_buff *skb = NULL;
2560 
2561 	bh_lock_sock(sk);
2562 	if (!tipc_sk_connected(sk))
2563 		goto exit;
2564 
2565 	/* Try again later if socket is busy */
2566 	if (sock_owned_by_user(sk)) {
2567 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2568 		goto exit;
2569 	}
2570 
2571 	if (tsk->probe_unacked) {
2572 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2573 		tipc_node_remove_conn(net, peer_node, peer_port);
2574 		sk->sk_state_change(sk);
2575 		goto exit;
2576 	}
2577 	/* Send new probe */
2578 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2579 			      peer_node, own_node, peer_port, own_port,
2580 			      TIPC_OK);
2581 	tsk->probe_unacked = true;
2582 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2583 exit:
2584 	bh_unlock_sock(sk);
2585 	if (skb)
2586 		tipc_node_xmit_skb(net, skb, peer_node, own_port);
2587 	sock_put(sk);
2588 }
2589 
2590 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2591 			   struct tipc_name_seq const *seq)
2592 {
2593 	struct sock *sk = &tsk->sk;
2594 	struct net *net = sock_net(sk);
2595 	struct publication *publ;
2596 	u32 key;
2597 
2598 	if (scope != TIPC_NODE_SCOPE)
2599 		scope = TIPC_CLUSTER_SCOPE;
2600 
2601 	if (tipc_sk_connected(sk))
2602 		return -EINVAL;
2603 	key = tsk->portid + tsk->pub_count + 1;
2604 	if (key == tsk->portid)
2605 		return -EADDRINUSE;
2606 
2607 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2608 				    scope, tsk->portid, key);
2609 	if (unlikely(!publ))
2610 		return -EINVAL;
2611 
2612 	list_add(&publ->binding_sock, &tsk->publications);
2613 	tsk->pub_count++;
2614 	tsk->published = 1;
2615 	return 0;
2616 }
2617 
2618 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2619 			    struct tipc_name_seq const *seq)
2620 {
2621 	struct net *net = sock_net(&tsk->sk);
2622 	struct publication *publ;
2623 	struct publication *safe;
2624 	int rc = -EINVAL;
2625 
2626 	if (scope != TIPC_NODE_SCOPE)
2627 		scope = TIPC_CLUSTER_SCOPE;
2628 
2629 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2630 		if (seq) {
2631 			if (publ->scope != scope)
2632 				continue;
2633 			if (publ->type != seq->type)
2634 				continue;
2635 			if (publ->lower != seq->lower)
2636 				continue;
2637 			if (publ->upper != seq->upper)
2638 				break;
2639 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2640 					      publ->upper, publ->key);
2641 			rc = 0;
2642 			break;
2643 		}
2644 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2645 				      publ->upper, publ->key);
2646 		rc = 0;
2647 	}
2648 	if (list_empty(&tsk->publications))
2649 		tsk->published = 0;
2650 	return rc;
2651 }
2652 
2653 /* tipc_sk_reinit: set non-zero address in all existing sockets
2654  *                 when we go from standalone to network mode.
2655  */
2656 void tipc_sk_reinit(struct net *net)
2657 {
2658 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2659 	struct rhashtable_iter iter;
2660 	struct tipc_sock *tsk;
2661 	struct tipc_msg *msg;
2662 
2663 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2664 
2665 	do {
2666 		rhashtable_walk_start(&iter);
2667 
2668 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2669 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2670 			msg = &tsk->phdr;
2671 			msg_set_prevnode(msg, tipc_own_addr(net));
2672 			msg_set_orignode(msg, tipc_own_addr(net));
2673 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2674 		}
2675 
2676 		rhashtable_walk_stop(&iter);
2677 	} while (tsk == ERR_PTR(-EAGAIN));
2678 
2679 	rhashtable_walk_exit(&iter);
2680 }
2681 
2682 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2683 {
2684 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2685 	struct tipc_sock *tsk;
2686 
2687 	rcu_read_lock();
2688 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2689 	if (tsk)
2690 		sock_hold(&tsk->sk);
2691 	rcu_read_unlock();
2692 
2693 	return tsk;
2694 }
2695 
2696 static int tipc_sk_insert(struct tipc_sock *tsk)
2697 {
2698 	struct sock *sk = &tsk->sk;
2699 	struct net *net = sock_net(sk);
2700 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2701 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2702 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2703 
2704 	while (remaining--) {
2705 		portid++;
2706 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2707 			portid = TIPC_MIN_PORT;
2708 		tsk->portid = portid;
2709 		sock_hold(&tsk->sk);
2710 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2711 						   tsk_rht_params))
2712 			return 0;
2713 		sock_put(&tsk->sk);
2714 	}
2715 
2716 	return -1;
2717 }
2718 
2719 static void tipc_sk_remove(struct tipc_sock *tsk)
2720 {
2721 	struct sock *sk = &tsk->sk;
2722 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2723 
2724 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2725 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2726 		__sock_put(sk);
2727 	}
2728 }
2729 
2730 static const struct rhashtable_params tsk_rht_params = {
2731 	.nelem_hint = 192,
2732 	.head_offset = offsetof(struct tipc_sock, node),
2733 	.key_offset = offsetof(struct tipc_sock, portid),
2734 	.key_len = sizeof(u32), /* portid */
2735 	.max_size = 1048576,
2736 	.min_size = 256,
2737 	.automatic_shrinking = true,
2738 };
2739 
2740 int tipc_sk_rht_init(struct net *net)
2741 {
2742 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2743 
2744 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2745 }
2746 
2747 void tipc_sk_rht_destroy(struct net *net)
2748 {
2749 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2750 
2751 	/* Wait for socket readers to complete */
2752 	synchronize_net();
2753 
2754 	rhashtable_destroy(&tn->sk_rht);
2755 }
2756 
2757 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2758 {
2759 	struct net *net = sock_net(&tsk->sk);
2760 	struct tipc_group *grp = tsk->group;
2761 	struct tipc_msg *hdr = &tsk->phdr;
2762 	struct tipc_name_seq seq;
2763 	int rc;
2764 
2765 	if (mreq->type < TIPC_RESERVED_TYPES)
2766 		return -EACCES;
2767 	if (mreq->scope > TIPC_NODE_SCOPE)
2768 		return -EINVAL;
2769 	if (grp)
2770 		return -EACCES;
2771 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2772 	if (!grp)
2773 		return -ENOMEM;
2774 	tsk->group = grp;
2775 	msg_set_lookup_scope(hdr, mreq->scope);
2776 	msg_set_nametype(hdr, mreq->type);
2777 	msg_set_dest_droppable(hdr, true);
2778 	seq.type = mreq->type;
2779 	seq.lower = mreq->instance;
2780 	seq.upper = seq.lower;
2781 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2782 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2783 	if (rc) {
2784 		tipc_group_delete(net, grp);
2785 		tsk->group = NULL;
2786 		return rc;
2787 	}
2788 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2789 	tsk->mc_method.rcast = true;
2790 	tsk->mc_method.mandatory = true;
2791 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2792 	return rc;
2793 }
2794 
2795 static int tipc_sk_leave(struct tipc_sock *tsk)
2796 {
2797 	struct net *net = sock_net(&tsk->sk);
2798 	struct tipc_group *grp = tsk->group;
2799 	struct tipc_name_seq seq;
2800 	int scope;
2801 
2802 	if (!grp)
2803 		return -EINVAL;
2804 	tipc_group_self(grp, &seq, &scope);
2805 	tipc_group_delete(net, grp);
2806 	tsk->group = NULL;
2807 	tipc_sk_withdraw(tsk, scope, &seq);
2808 	return 0;
2809 }
2810 
2811 /**
2812  * tipc_setsockopt - set socket option
2813  * @sock: socket structure
2814  * @lvl: option level
2815  * @opt: option identifier
2816  * @ov: pointer to new option value
2817  * @ol: length of option value
2818  *
2819  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2820  * (to ease compatibility).
2821  *
2822  * Returns 0 on success, errno otherwise
2823  */
2824 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2825 			   char __user *ov, unsigned int ol)
2826 {
2827 	struct sock *sk = sock->sk;
2828 	struct tipc_sock *tsk = tipc_sk(sk);
2829 	struct tipc_group_req mreq;
2830 	u32 value = 0;
2831 	int res = 0;
2832 
2833 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2834 		return 0;
2835 	if (lvl != SOL_TIPC)
2836 		return -ENOPROTOOPT;
2837 
2838 	switch (opt) {
2839 	case TIPC_IMPORTANCE:
2840 	case TIPC_SRC_DROPPABLE:
2841 	case TIPC_DEST_DROPPABLE:
2842 	case TIPC_CONN_TIMEOUT:
2843 		if (ol < sizeof(value))
2844 			return -EINVAL;
2845 		if (get_user(value, (u32 __user *)ov))
2846 			return -EFAULT;
2847 		break;
2848 	case TIPC_GROUP_JOIN:
2849 		if (ol < sizeof(mreq))
2850 			return -EINVAL;
2851 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2852 			return -EFAULT;
2853 		break;
2854 	default:
2855 		if (ov || ol)
2856 			return -EINVAL;
2857 	}
2858 
2859 	lock_sock(sk);
2860 
2861 	switch (opt) {
2862 	case TIPC_IMPORTANCE:
2863 		res = tsk_set_importance(tsk, value);
2864 		break;
2865 	case TIPC_SRC_DROPPABLE:
2866 		if (sock->type != SOCK_STREAM)
2867 			tsk_set_unreliable(tsk, value);
2868 		else
2869 			res = -ENOPROTOOPT;
2870 		break;
2871 	case TIPC_DEST_DROPPABLE:
2872 		tsk_set_unreturnable(tsk, value);
2873 		break;
2874 	case TIPC_CONN_TIMEOUT:
2875 		tipc_sk(sk)->conn_timeout = value;
2876 		break;
2877 	case TIPC_MCAST_BROADCAST:
2878 		tsk->mc_method.rcast = false;
2879 		tsk->mc_method.mandatory = true;
2880 		break;
2881 	case TIPC_MCAST_REPLICAST:
2882 		tsk->mc_method.rcast = true;
2883 		tsk->mc_method.mandatory = true;
2884 		break;
2885 	case TIPC_GROUP_JOIN:
2886 		res = tipc_sk_join(tsk, &mreq);
2887 		break;
2888 	case TIPC_GROUP_LEAVE:
2889 		res = tipc_sk_leave(tsk);
2890 		break;
2891 	default:
2892 		res = -EINVAL;
2893 	}
2894 
2895 	release_sock(sk);
2896 
2897 	return res;
2898 }
2899 
2900 /**
2901  * tipc_getsockopt - get socket option
2902  * @sock: socket structure
2903  * @lvl: option level
2904  * @opt: option identifier
2905  * @ov: receptacle for option value
2906  * @ol: receptacle for length of option value
2907  *
2908  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2909  * (to ease compatibility).
2910  *
2911  * Returns 0 on success, errno otherwise
2912  */
2913 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2914 			   char __user *ov, int __user *ol)
2915 {
2916 	struct sock *sk = sock->sk;
2917 	struct tipc_sock *tsk = tipc_sk(sk);
2918 	struct tipc_name_seq seq;
2919 	int len, scope;
2920 	u32 value;
2921 	int res;
2922 
2923 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2924 		return put_user(0, ol);
2925 	if (lvl != SOL_TIPC)
2926 		return -ENOPROTOOPT;
2927 	res = get_user(len, ol);
2928 	if (res)
2929 		return res;
2930 
2931 	lock_sock(sk);
2932 
2933 	switch (opt) {
2934 	case TIPC_IMPORTANCE:
2935 		value = tsk_importance(tsk);
2936 		break;
2937 	case TIPC_SRC_DROPPABLE:
2938 		value = tsk_unreliable(tsk);
2939 		break;
2940 	case TIPC_DEST_DROPPABLE:
2941 		value = tsk_unreturnable(tsk);
2942 		break;
2943 	case TIPC_CONN_TIMEOUT:
2944 		value = tsk->conn_timeout;
2945 		/* no need to set "res", since already 0 at this point */
2946 		break;
2947 	case TIPC_NODE_RECVQ_DEPTH:
2948 		value = 0; /* was tipc_queue_size, now obsolete */
2949 		break;
2950 	case TIPC_SOCK_RECVQ_DEPTH:
2951 		value = skb_queue_len(&sk->sk_receive_queue);
2952 		break;
2953 	case TIPC_GROUP_JOIN:
2954 		seq.type = 0;
2955 		if (tsk->group)
2956 			tipc_group_self(tsk->group, &seq, &scope);
2957 		value = seq.type;
2958 		break;
2959 	default:
2960 		res = -EINVAL;
2961 	}
2962 
2963 	release_sock(sk);
2964 
2965 	if (res)
2966 		return res;	/* "get" failed */
2967 
2968 	if (len < sizeof(value))
2969 		return -EINVAL;
2970 
2971 	if (copy_to_user(ov, &value, sizeof(value)))
2972 		return -EFAULT;
2973 
2974 	return put_user(sizeof(value), ol);
2975 }
2976 
2977 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2978 {
2979 	struct net *net = sock_net(sock->sk);
2980 	struct tipc_sioc_nodeid_req nr = {0};
2981 	struct tipc_sioc_ln_req lnr;
2982 	void __user *argp = (void __user *)arg;
2983 
2984 	switch (cmd) {
2985 	case SIOCGETLINKNAME:
2986 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2987 			return -EFAULT;
2988 		if (!tipc_node_get_linkname(net,
2989 					    lnr.bearer_id & 0xffff, lnr.peer,
2990 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2991 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2992 				return -EFAULT;
2993 			return 0;
2994 		}
2995 		return -EADDRNOTAVAIL;
2996 	case SIOCGETNODEID:
2997 		if (copy_from_user(&nr, argp, sizeof(nr)))
2998 			return -EFAULT;
2999 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3000 			return -EADDRNOTAVAIL;
3001 		if (copy_to_user(argp, &nr, sizeof(nr)))
3002 			return -EFAULT;
3003 		return 0;
3004 	default:
3005 		return -ENOIOCTLCMD;
3006 	}
3007 }
3008 
3009 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3010 {
3011 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3012 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3013 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3014 
3015 	tsk1->peer.family = AF_TIPC;
3016 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3017 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3018 	tsk1->peer.addr.id.ref = tsk2->portid;
3019 	tsk1->peer.addr.id.node = onode;
3020 	tsk2->peer.family = AF_TIPC;
3021 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3022 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3023 	tsk2->peer.addr.id.ref = tsk1->portid;
3024 	tsk2->peer.addr.id.node = onode;
3025 
3026 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3027 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3028 	return 0;
3029 }
3030 
3031 /* Protocol switches for the various types of TIPC sockets */
3032 
3033 static const struct proto_ops msg_ops = {
3034 	.owner		= THIS_MODULE,
3035 	.family		= AF_TIPC,
3036 	.release	= tipc_release,
3037 	.bind		= tipc_bind,
3038 	.connect	= tipc_connect,
3039 	.socketpair	= tipc_socketpair,
3040 	.accept		= sock_no_accept,
3041 	.getname	= tipc_getname,
3042 	.poll		= tipc_poll,
3043 	.ioctl		= tipc_ioctl,
3044 	.listen		= sock_no_listen,
3045 	.shutdown	= tipc_shutdown,
3046 	.setsockopt	= tipc_setsockopt,
3047 	.getsockopt	= tipc_getsockopt,
3048 	.sendmsg	= tipc_sendmsg,
3049 	.recvmsg	= tipc_recvmsg,
3050 	.mmap		= sock_no_mmap,
3051 	.sendpage	= sock_no_sendpage
3052 };
3053 
3054 static const struct proto_ops packet_ops = {
3055 	.owner		= THIS_MODULE,
3056 	.family		= AF_TIPC,
3057 	.release	= tipc_release,
3058 	.bind		= tipc_bind,
3059 	.connect	= tipc_connect,
3060 	.socketpair	= tipc_socketpair,
3061 	.accept		= tipc_accept,
3062 	.getname	= tipc_getname,
3063 	.poll		= tipc_poll,
3064 	.ioctl		= tipc_ioctl,
3065 	.listen		= tipc_listen,
3066 	.shutdown	= tipc_shutdown,
3067 	.setsockopt	= tipc_setsockopt,
3068 	.getsockopt	= tipc_getsockopt,
3069 	.sendmsg	= tipc_send_packet,
3070 	.recvmsg	= tipc_recvmsg,
3071 	.mmap		= sock_no_mmap,
3072 	.sendpage	= sock_no_sendpage
3073 };
3074 
3075 static const struct proto_ops stream_ops = {
3076 	.owner		= THIS_MODULE,
3077 	.family		= AF_TIPC,
3078 	.release	= tipc_release,
3079 	.bind		= tipc_bind,
3080 	.connect	= tipc_connect,
3081 	.socketpair	= tipc_socketpair,
3082 	.accept		= tipc_accept,
3083 	.getname	= tipc_getname,
3084 	.poll		= tipc_poll,
3085 	.ioctl		= tipc_ioctl,
3086 	.listen		= tipc_listen,
3087 	.shutdown	= tipc_shutdown,
3088 	.setsockopt	= tipc_setsockopt,
3089 	.getsockopt	= tipc_getsockopt,
3090 	.sendmsg	= tipc_sendstream,
3091 	.recvmsg	= tipc_recvstream,
3092 	.mmap		= sock_no_mmap,
3093 	.sendpage	= sock_no_sendpage
3094 };
3095 
3096 static const struct net_proto_family tipc_family_ops = {
3097 	.owner		= THIS_MODULE,
3098 	.family		= AF_TIPC,
3099 	.create		= tipc_sk_create
3100 };
3101 
3102 static struct proto tipc_proto = {
3103 	.name		= "TIPC",
3104 	.owner		= THIS_MODULE,
3105 	.obj_size	= sizeof(struct tipc_sock),
3106 	.sysctl_rmem	= sysctl_tipc_rmem
3107 };
3108 
3109 /**
3110  * tipc_socket_init - initialize TIPC socket interface
3111  *
3112  * Returns 0 on success, errno otherwise
3113  */
3114 int tipc_socket_init(void)
3115 {
3116 	int res;
3117 
3118 	res = proto_register(&tipc_proto, 1);
3119 	if (res) {
3120 		pr_err("Failed to register TIPC protocol type\n");
3121 		goto out;
3122 	}
3123 
3124 	res = sock_register(&tipc_family_ops);
3125 	if (res) {
3126 		pr_err("Failed to register TIPC socket type\n");
3127 		proto_unregister(&tipc_proto);
3128 		goto out;
3129 	}
3130  out:
3131 	return res;
3132 }
3133 
3134 /**
3135  * tipc_socket_stop - stop TIPC socket interface
3136  */
3137 void tipc_socket_stop(void)
3138 {
3139 	sock_unregister(tipc_family_ops.family);
3140 	proto_unregister(&tipc_proto);
3141 }
3142 
3143 /* Caller should hold socket lock for the passed tipc socket. */
3144 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3145 {
3146 	u32 peer_node;
3147 	u32 peer_port;
3148 	struct nlattr *nest;
3149 
3150 	peer_node = tsk_peer_node(tsk);
3151 	peer_port = tsk_peer_port(tsk);
3152 
3153 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3154 
3155 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3156 		goto msg_full;
3157 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3158 		goto msg_full;
3159 
3160 	if (tsk->conn_type != 0) {
3161 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3162 			goto msg_full;
3163 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3164 			goto msg_full;
3165 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3166 			goto msg_full;
3167 	}
3168 	nla_nest_end(skb, nest);
3169 
3170 	return 0;
3171 
3172 msg_full:
3173 	nla_nest_cancel(skb, nest);
3174 
3175 	return -EMSGSIZE;
3176 }
3177 
3178 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3179 			  *tsk)
3180 {
3181 	struct net *net = sock_net(skb->sk);
3182 	struct sock *sk = &tsk->sk;
3183 
3184 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3185 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3186 		return -EMSGSIZE;
3187 
3188 	if (tipc_sk_connected(sk)) {
3189 		if (__tipc_nl_add_sk_con(skb, tsk))
3190 			return -EMSGSIZE;
3191 	} else if (!list_empty(&tsk->publications)) {
3192 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3193 			return -EMSGSIZE;
3194 	}
3195 	return 0;
3196 }
3197 
3198 /* Caller should hold socket lock for the passed tipc socket. */
3199 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3200 			    struct tipc_sock *tsk)
3201 {
3202 	struct nlattr *attrs;
3203 	void *hdr;
3204 
3205 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3206 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3207 	if (!hdr)
3208 		goto msg_cancel;
3209 
3210 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3211 	if (!attrs)
3212 		goto genlmsg_cancel;
3213 
3214 	if (__tipc_nl_add_sk_info(skb, tsk))
3215 		goto attr_msg_cancel;
3216 
3217 	nla_nest_end(skb, attrs);
3218 	genlmsg_end(skb, hdr);
3219 
3220 	return 0;
3221 
3222 attr_msg_cancel:
3223 	nla_nest_cancel(skb, attrs);
3224 genlmsg_cancel:
3225 	genlmsg_cancel(skb, hdr);
3226 msg_cancel:
3227 	return -EMSGSIZE;
3228 }
3229 
3230 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3231 		    int (*skb_handler)(struct sk_buff *skb,
3232 				       struct netlink_callback *cb,
3233 				       struct tipc_sock *tsk))
3234 {
3235 	struct rhashtable_iter *iter = (void *)cb->args[4];
3236 	struct tipc_sock *tsk;
3237 	int err;
3238 
3239 	rhashtable_walk_start(iter);
3240 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3241 		if (IS_ERR(tsk)) {
3242 			err = PTR_ERR(tsk);
3243 			if (err == -EAGAIN) {
3244 				err = 0;
3245 				continue;
3246 			}
3247 			break;
3248 		}
3249 
3250 		sock_hold(&tsk->sk);
3251 		rhashtable_walk_stop(iter);
3252 		lock_sock(&tsk->sk);
3253 		err = skb_handler(skb, cb, tsk);
3254 		if (err) {
3255 			release_sock(&tsk->sk);
3256 			sock_put(&tsk->sk);
3257 			goto out;
3258 		}
3259 		release_sock(&tsk->sk);
3260 		rhashtable_walk_start(iter);
3261 		sock_put(&tsk->sk);
3262 	}
3263 	rhashtable_walk_stop(iter);
3264 out:
3265 	return skb->len;
3266 }
3267 EXPORT_SYMBOL(tipc_nl_sk_walk);
3268 
3269 int tipc_dump_start(struct netlink_callback *cb)
3270 {
3271 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3272 }
3273 EXPORT_SYMBOL(tipc_dump_start);
3274 
3275 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3276 {
3277 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3278 	struct rhashtable_iter *iter = (void *)cb->args[4];
3279 	struct tipc_net *tn = tipc_net(net);
3280 
3281 	if (!iter) {
3282 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3283 		if (!iter)
3284 			return -ENOMEM;
3285 
3286 		cb->args[4] = (long)iter;
3287 	}
3288 
3289 	rhashtable_walk_enter(&tn->sk_rht, iter);
3290 	return 0;
3291 }
3292 
3293 int tipc_dump_done(struct netlink_callback *cb)
3294 {
3295 	struct rhashtable_iter *hti = (void *)cb->args[4];
3296 
3297 	rhashtable_walk_exit(hti);
3298 	kfree(hti);
3299 	return 0;
3300 }
3301 EXPORT_SYMBOL(tipc_dump_done);
3302 
3303 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3304 			   struct tipc_sock *tsk, u32 sk_filter_state,
3305 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3306 {
3307 	struct sock *sk = &tsk->sk;
3308 	struct nlattr *attrs;
3309 	struct nlattr *stat;
3310 
3311 	/*filter response w.r.t sk_state*/
3312 	if (!(sk_filter_state & (1 << sk->sk_state)))
3313 		return 0;
3314 
3315 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3316 	if (!attrs)
3317 		goto msg_cancel;
3318 
3319 	if (__tipc_nl_add_sk_info(skb, tsk))
3320 		goto attr_msg_cancel;
3321 
3322 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3323 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3324 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3325 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3326 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3327 					 sock_i_uid(sk))) ||
3328 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3329 			      tipc_diag_gen_cookie(sk),
3330 			      TIPC_NLA_SOCK_PAD))
3331 		goto attr_msg_cancel;
3332 
3333 	stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3334 	if (!stat)
3335 		goto attr_msg_cancel;
3336 
3337 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3338 			skb_queue_len(&sk->sk_receive_queue)) ||
3339 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3340 			skb_queue_len(&sk->sk_write_queue)) ||
3341 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3342 			atomic_read(&sk->sk_drops)))
3343 		goto stat_msg_cancel;
3344 
3345 	if (tsk->cong_link_cnt &&
3346 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3347 		goto stat_msg_cancel;
3348 
3349 	if (tsk_conn_cong(tsk) &&
3350 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3351 		goto stat_msg_cancel;
3352 
3353 	nla_nest_end(skb, stat);
3354 
3355 	if (tsk->group)
3356 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3357 			goto stat_msg_cancel;
3358 
3359 	nla_nest_end(skb, attrs);
3360 
3361 	return 0;
3362 
3363 stat_msg_cancel:
3364 	nla_nest_cancel(skb, stat);
3365 attr_msg_cancel:
3366 	nla_nest_cancel(skb, attrs);
3367 msg_cancel:
3368 	return -EMSGSIZE;
3369 }
3370 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3371 
3372 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3373 {
3374 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3375 }
3376 
3377 /* Caller should hold socket lock for the passed tipc socket. */
3378 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3379 				 struct netlink_callback *cb,
3380 				 struct publication *publ)
3381 {
3382 	void *hdr;
3383 	struct nlattr *attrs;
3384 
3385 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3386 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3387 	if (!hdr)
3388 		goto msg_cancel;
3389 
3390 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3391 	if (!attrs)
3392 		goto genlmsg_cancel;
3393 
3394 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3395 		goto attr_msg_cancel;
3396 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3397 		goto attr_msg_cancel;
3398 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3399 		goto attr_msg_cancel;
3400 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3401 		goto attr_msg_cancel;
3402 
3403 	nla_nest_end(skb, attrs);
3404 	genlmsg_end(skb, hdr);
3405 
3406 	return 0;
3407 
3408 attr_msg_cancel:
3409 	nla_nest_cancel(skb, attrs);
3410 genlmsg_cancel:
3411 	genlmsg_cancel(skb, hdr);
3412 msg_cancel:
3413 	return -EMSGSIZE;
3414 }
3415 
3416 /* Caller should hold socket lock for the passed tipc socket. */
3417 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3418 				  struct netlink_callback *cb,
3419 				  struct tipc_sock *tsk, u32 *last_publ)
3420 {
3421 	int err;
3422 	struct publication *p;
3423 
3424 	if (*last_publ) {
3425 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3426 			if (p->key == *last_publ)
3427 				break;
3428 		}
3429 		if (p->key != *last_publ) {
3430 			/* We never set seq or call nl_dump_check_consistent()
3431 			 * this means that setting prev_seq here will cause the
3432 			 * consistence check to fail in the netlink callback
3433 			 * handler. Resulting in the last NLMSG_DONE message
3434 			 * having the NLM_F_DUMP_INTR flag set.
3435 			 */
3436 			cb->prev_seq = 1;
3437 			*last_publ = 0;
3438 			return -EPIPE;
3439 		}
3440 	} else {
3441 		p = list_first_entry(&tsk->publications, struct publication,
3442 				     binding_sock);
3443 	}
3444 
3445 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3446 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3447 		if (err) {
3448 			*last_publ = p->key;
3449 			return err;
3450 		}
3451 	}
3452 	*last_publ = 0;
3453 
3454 	return 0;
3455 }
3456 
3457 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3458 {
3459 	int err;
3460 	u32 tsk_portid = cb->args[0];
3461 	u32 last_publ = cb->args[1];
3462 	u32 done = cb->args[2];
3463 	struct net *net = sock_net(skb->sk);
3464 	struct tipc_sock *tsk;
3465 
3466 	if (!tsk_portid) {
3467 		struct nlattr **attrs;
3468 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3469 
3470 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3471 		if (err)
3472 			return err;
3473 
3474 		if (!attrs[TIPC_NLA_SOCK])
3475 			return -EINVAL;
3476 
3477 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3478 				       attrs[TIPC_NLA_SOCK],
3479 				       tipc_nl_sock_policy, NULL);
3480 		if (err)
3481 			return err;
3482 
3483 		if (!sock[TIPC_NLA_SOCK_REF])
3484 			return -EINVAL;
3485 
3486 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3487 	}
3488 
3489 	if (done)
3490 		return 0;
3491 
3492 	tsk = tipc_sk_lookup(net, tsk_portid);
3493 	if (!tsk)
3494 		return -EINVAL;
3495 
3496 	lock_sock(&tsk->sk);
3497 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3498 	if (!err)
3499 		done = 1;
3500 	release_sock(&tsk->sk);
3501 	sock_put(&tsk->sk);
3502 
3503 	cb->args[0] = tsk_portid;
3504 	cb->args[1] = last_publ;
3505 	cb->args[2] = done;
3506 
3507 	return skb->len;
3508 }
3509