xref: /linux/net/tipc/socket.c (revision e0a2e73e501c77037c8756137e87b12c7c3c9793)
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->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1424 		return rc;
1425 	}
1426 
1427 	do {
1428 		rc = tipc_wait_for_cond(sock, &timeout,
1429 					(!tsk->cong_link_cnt &&
1430 					 !tsk_conn_cong(tsk) &&
1431 					 tipc_sk_connected(sk)));
1432 		if (unlikely(rc))
1433 			break;
1434 
1435 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1436 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1437 		if (unlikely(rc != send))
1438 			break;
1439 
1440 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1441 		if (unlikely(rc == -ELINKCONG)) {
1442 			tsk->cong_link_cnt = 1;
1443 			rc = 0;
1444 		}
1445 		if (likely(!rc)) {
1446 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1447 			sent += send;
1448 		}
1449 	} while (sent < dlen && !rc);
1450 
1451 	return sent ? sent : rc;
1452 }
1453 
1454 /**
1455  * tipc_send_packet - send a connection-oriented message
1456  * @sock: socket structure
1457  * @m: message to send
1458  * @dsz: length of data to be transmitted
1459  *
1460  * Used for SOCK_SEQPACKET messages.
1461  *
1462  * Returns the number of bytes sent on success, or errno otherwise
1463  */
1464 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1465 {
1466 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1467 		return -EMSGSIZE;
1468 
1469 	return tipc_sendstream(sock, m, dsz);
1470 }
1471 
1472 /* tipc_sk_finish_conn - complete the setup of a connection
1473  */
1474 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1475 				u32 peer_node)
1476 {
1477 	struct sock *sk = &tsk->sk;
1478 	struct net *net = sock_net(sk);
1479 	struct tipc_msg *msg = &tsk->phdr;
1480 
1481 	msg_set_destnode(msg, peer_node);
1482 	msg_set_destport(msg, peer_port);
1483 	msg_set_type(msg, TIPC_CONN_MSG);
1484 	msg_set_lookup_scope(msg, 0);
1485 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1486 
1487 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1488 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1489 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1490 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1491 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1492 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1493 		return;
1494 
1495 	/* Fall back to message based flow control */
1496 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1497 	tsk->snd_win = FLOWCTL_MSG_WIN;
1498 }
1499 
1500 /**
1501  * tipc_sk_set_orig_addr - capture sender's address for received message
1502  * @m: descriptor for message info
1503  * @hdr: received message header
1504  *
1505  * Note: Address is not captured if not requested by receiver.
1506  */
1507 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1508 {
1509 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1510 	struct tipc_msg *hdr = buf_msg(skb);
1511 
1512 	if (!srcaddr)
1513 		return;
1514 
1515 	srcaddr->sock.family = AF_TIPC;
1516 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1517 	srcaddr->sock.scope = 0;
1518 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1519 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1520 	srcaddr->sock.addr.name.domain = 0;
1521 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1522 
1523 	if (!msg_in_group(hdr))
1524 		return;
1525 
1526 	/* Group message users may also want to know sending member's id */
1527 	srcaddr->member.family = AF_TIPC;
1528 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1529 	srcaddr->member.scope = 0;
1530 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1531 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1532 	srcaddr->member.addr.name.domain = 0;
1533 	m->msg_namelen = sizeof(*srcaddr);
1534 }
1535 
1536 /**
1537  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1538  * @m: descriptor for message info
1539  * @msg: received message header
1540  * @tsk: TIPC port associated with message
1541  *
1542  * Note: Ancillary data is not captured if not requested by receiver.
1543  *
1544  * Returns 0 if successful, otherwise errno
1545  */
1546 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1547 				 struct tipc_sock *tsk)
1548 {
1549 	u32 anc_data[3];
1550 	u32 err;
1551 	u32 dest_type;
1552 	int has_name;
1553 	int res;
1554 
1555 	if (likely(m->msg_controllen == 0))
1556 		return 0;
1557 
1558 	/* Optionally capture errored message object(s) */
1559 	err = msg ? msg_errcode(msg) : 0;
1560 	if (unlikely(err)) {
1561 		anc_data[0] = err;
1562 		anc_data[1] = msg_data_sz(msg);
1563 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1564 		if (res)
1565 			return res;
1566 		if (anc_data[1]) {
1567 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1568 				       msg_data(msg));
1569 			if (res)
1570 				return res;
1571 		}
1572 	}
1573 
1574 	/* Optionally capture message destination object */
1575 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1576 	switch (dest_type) {
1577 	case TIPC_NAMED_MSG:
1578 		has_name = 1;
1579 		anc_data[0] = msg_nametype(msg);
1580 		anc_data[1] = msg_namelower(msg);
1581 		anc_data[2] = msg_namelower(msg);
1582 		break;
1583 	case TIPC_MCAST_MSG:
1584 		has_name = 1;
1585 		anc_data[0] = msg_nametype(msg);
1586 		anc_data[1] = msg_namelower(msg);
1587 		anc_data[2] = msg_nameupper(msg);
1588 		break;
1589 	case TIPC_CONN_MSG:
1590 		has_name = (tsk->conn_type != 0);
1591 		anc_data[0] = tsk->conn_type;
1592 		anc_data[1] = tsk->conn_instance;
1593 		anc_data[2] = tsk->conn_instance;
1594 		break;
1595 	default:
1596 		has_name = 0;
1597 	}
1598 	if (has_name) {
1599 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1600 		if (res)
1601 			return res;
1602 	}
1603 
1604 	return 0;
1605 }
1606 
1607 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1608 {
1609 	struct sock *sk = &tsk->sk;
1610 	struct net *net = sock_net(sk);
1611 	struct sk_buff *skb = NULL;
1612 	struct tipc_msg *msg;
1613 	u32 peer_port = tsk_peer_port(tsk);
1614 	u32 dnode = tsk_peer_node(tsk);
1615 
1616 	if (!tipc_sk_connected(sk))
1617 		return;
1618 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1619 			      dnode, tsk_own_node(tsk), peer_port,
1620 			      tsk->portid, TIPC_OK);
1621 	if (!skb)
1622 		return;
1623 	msg = buf_msg(skb);
1624 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1625 	tsk->rcv_unacked = 0;
1626 
1627 	/* Adjust to and advertize the correct window limit */
1628 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1629 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1630 		msg_set_adv_win(msg, tsk->rcv_win);
1631 	}
1632 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1633 }
1634 
1635 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1636 {
1637 	struct sock *sk = sock->sk;
1638 	DEFINE_WAIT(wait);
1639 	long timeo = *timeop;
1640 	int err = sock_error(sk);
1641 
1642 	if (err)
1643 		return err;
1644 
1645 	for (;;) {
1646 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1647 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1648 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1649 				err = -ENOTCONN;
1650 				break;
1651 			}
1652 			release_sock(sk);
1653 			timeo = schedule_timeout(timeo);
1654 			lock_sock(sk);
1655 		}
1656 		err = 0;
1657 		if (!skb_queue_empty(&sk->sk_receive_queue))
1658 			break;
1659 		err = -EAGAIN;
1660 		if (!timeo)
1661 			break;
1662 		err = sock_intr_errno(timeo);
1663 		if (signal_pending(current))
1664 			break;
1665 
1666 		err = sock_error(sk);
1667 		if (err)
1668 			break;
1669 	}
1670 	finish_wait(sk_sleep(sk), &wait);
1671 	*timeop = timeo;
1672 	return err;
1673 }
1674 
1675 /**
1676  * tipc_recvmsg - receive packet-oriented message
1677  * @m: descriptor for message info
1678  * @buflen: length of user buffer area
1679  * @flags: receive flags
1680  *
1681  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1682  * If the complete message doesn't fit in user area, truncate it.
1683  *
1684  * Returns size of returned message data, errno otherwise
1685  */
1686 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1687 			size_t buflen,	int flags)
1688 {
1689 	struct sock *sk = sock->sk;
1690 	bool connected = !tipc_sk_type_connectionless(sk);
1691 	struct tipc_sock *tsk = tipc_sk(sk);
1692 	int rc, err, hlen, dlen, copy;
1693 	struct sk_buff_head xmitq;
1694 	struct tipc_msg *hdr;
1695 	struct sk_buff *skb;
1696 	bool grp_evt;
1697 	long timeout;
1698 
1699 	/* Catch invalid receive requests */
1700 	if (unlikely(!buflen))
1701 		return -EINVAL;
1702 
1703 	lock_sock(sk);
1704 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1705 		rc = -ENOTCONN;
1706 		goto exit;
1707 	}
1708 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1709 
1710 	/* Step rcv queue to first msg with data or error; wait if necessary */
1711 	do {
1712 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1713 		if (unlikely(rc))
1714 			goto exit;
1715 		skb = skb_peek(&sk->sk_receive_queue);
1716 		hdr = buf_msg(skb);
1717 		dlen = msg_data_sz(hdr);
1718 		hlen = msg_hdr_sz(hdr);
1719 		err = msg_errcode(hdr);
1720 		grp_evt = msg_is_grp_evt(hdr);
1721 		if (likely(dlen || err))
1722 			break;
1723 		tsk_advance_rx_queue(sk);
1724 	} while (1);
1725 
1726 	/* Collect msg meta data, including error code and rejected data */
1727 	tipc_sk_set_orig_addr(m, skb);
1728 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1729 	if (unlikely(rc))
1730 		goto exit;
1731 
1732 	/* Capture data if non-error msg, otherwise just set return value */
1733 	if (likely(!err)) {
1734 		copy = min_t(int, dlen, buflen);
1735 		if (unlikely(copy != dlen))
1736 			m->msg_flags |= MSG_TRUNC;
1737 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1738 	} else {
1739 		copy = 0;
1740 		rc = 0;
1741 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1742 			rc = -ECONNRESET;
1743 	}
1744 	if (unlikely(rc))
1745 		goto exit;
1746 
1747 	/* Mark message as group event if applicable */
1748 	if (unlikely(grp_evt)) {
1749 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1750 			m->msg_flags |= MSG_EOR;
1751 		m->msg_flags |= MSG_OOB;
1752 		copy = 0;
1753 	}
1754 
1755 	/* Caption of data or error code/rejected data was successful */
1756 	if (unlikely(flags & MSG_PEEK))
1757 		goto exit;
1758 
1759 	/* Send group flow control advertisement when applicable */
1760 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1761 		skb_queue_head_init(&xmitq);
1762 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1763 					  msg_orignode(hdr), msg_origport(hdr),
1764 					  &xmitq);
1765 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1766 	}
1767 
1768 	tsk_advance_rx_queue(sk);
1769 
1770 	if (likely(!connected))
1771 		goto exit;
1772 
1773 	/* Send connection flow control advertisement when applicable */
1774 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1775 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1776 		tipc_sk_send_ack(tsk);
1777 exit:
1778 	release_sock(sk);
1779 	return rc ? rc : copy;
1780 }
1781 
1782 /**
1783  * tipc_recvstream - receive stream-oriented data
1784  * @m: descriptor for message info
1785  * @buflen: total size of user buffer area
1786  * @flags: receive flags
1787  *
1788  * Used for SOCK_STREAM messages only.  If not enough data is available
1789  * will optionally wait for more; never truncates data.
1790  *
1791  * Returns size of returned message data, errno otherwise
1792  */
1793 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1794 			   size_t buflen, int flags)
1795 {
1796 	struct sock *sk = sock->sk;
1797 	struct tipc_sock *tsk = tipc_sk(sk);
1798 	struct sk_buff *skb;
1799 	struct tipc_msg *hdr;
1800 	struct tipc_skb_cb *skb_cb;
1801 	bool peek = flags & MSG_PEEK;
1802 	int offset, required, copy, copied = 0;
1803 	int hlen, dlen, err, rc;
1804 	long timeout;
1805 
1806 	/* Catch invalid receive attempts */
1807 	if (unlikely(!buflen))
1808 		return -EINVAL;
1809 
1810 	lock_sock(sk);
1811 
1812 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1813 		rc = -ENOTCONN;
1814 		goto exit;
1815 	}
1816 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1817 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1818 
1819 	do {
1820 		/* Look at first msg in receive queue; wait if necessary */
1821 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1822 		if (unlikely(rc))
1823 			break;
1824 		skb = skb_peek(&sk->sk_receive_queue);
1825 		skb_cb = TIPC_SKB_CB(skb);
1826 		hdr = buf_msg(skb);
1827 		dlen = msg_data_sz(hdr);
1828 		hlen = msg_hdr_sz(hdr);
1829 		err = msg_errcode(hdr);
1830 
1831 		/* Discard any empty non-errored (SYN-) message */
1832 		if (unlikely(!dlen && !err)) {
1833 			tsk_advance_rx_queue(sk);
1834 			continue;
1835 		}
1836 
1837 		/* Collect msg meta data, incl. error code and rejected data */
1838 		if (!copied) {
1839 			tipc_sk_set_orig_addr(m, skb);
1840 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1841 			if (rc)
1842 				break;
1843 		}
1844 
1845 		/* Copy data if msg ok, otherwise return error/partial data */
1846 		if (likely(!err)) {
1847 			offset = skb_cb->bytes_read;
1848 			copy = min_t(int, dlen - offset, buflen - copied);
1849 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1850 			if (unlikely(rc))
1851 				break;
1852 			copied += copy;
1853 			offset += copy;
1854 			if (unlikely(offset < dlen)) {
1855 				if (!peek)
1856 					skb_cb->bytes_read = offset;
1857 				break;
1858 			}
1859 		} else {
1860 			rc = 0;
1861 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1862 				rc = -ECONNRESET;
1863 			if (copied || rc)
1864 				break;
1865 		}
1866 
1867 		if (unlikely(peek))
1868 			break;
1869 
1870 		tsk_advance_rx_queue(sk);
1871 
1872 		/* Send connection flow control advertisement when applicable */
1873 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1874 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1875 			tipc_sk_send_ack(tsk);
1876 
1877 		/* Exit if all requested data or FIN/error received */
1878 		if (copied == buflen || err)
1879 			break;
1880 
1881 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1882 exit:
1883 	release_sock(sk);
1884 	return copied ? copied : rc;
1885 }
1886 
1887 /**
1888  * tipc_write_space - wake up thread if port congestion is released
1889  * @sk: socket
1890  */
1891 static void tipc_write_space(struct sock *sk)
1892 {
1893 	struct socket_wq *wq;
1894 
1895 	rcu_read_lock();
1896 	wq = rcu_dereference(sk->sk_wq);
1897 	if (skwq_has_sleeper(wq))
1898 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1899 						EPOLLWRNORM | EPOLLWRBAND);
1900 	rcu_read_unlock();
1901 }
1902 
1903 /**
1904  * tipc_data_ready - wake up threads to indicate messages have been received
1905  * @sk: socket
1906  * @len: the length of messages
1907  */
1908 static void tipc_data_ready(struct sock *sk)
1909 {
1910 	struct socket_wq *wq;
1911 
1912 	rcu_read_lock();
1913 	wq = rcu_dereference(sk->sk_wq);
1914 	if (skwq_has_sleeper(wq))
1915 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1916 						EPOLLRDNORM | EPOLLRDBAND);
1917 	rcu_read_unlock();
1918 }
1919 
1920 static void tipc_sock_destruct(struct sock *sk)
1921 {
1922 	__skb_queue_purge(&sk->sk_receive_queue);
1923 }
1924 
1925 static void tipc_sk_proto_rcv(struct sock *sk,
1926 			      struct sk_buff_head *inputq,
1927 			      struct sk_buff_head *xmitq)
1928 {
1929 	struct sk_buff *skb = __skb_dequeue(inputq);
1930 	struct tipc_sock *tsk = tipc_sk(sk);
1931 	struct tipc_msg *hdr = buf_msg(skb);
1932 	struct tipc_group *grp = tsk->group;
1933 	bool wakeup = false;
1934 
1935 	switch (msg_user(hdr)) {
1936 	case CONN_MANAGER:
1937 		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1938 		return;
1939 	case SOCK_WAKEUP:
1940 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1941 		tsk->cong_link_cnt--;
1942 		wakeup = true;
1943 		break;
1944 	case GROUP_PROTOCOL:
1945 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1946 		break;
1947 	case TOP_SRV:
1948 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1949 				      hdr, inputq, xmitq);
1950 		break;
1951 	default:
1952 		break;
1953 	}
1954 
1955 	if (wakeup)
1956 		sk->sk_write_space(sk);
1957 
1958 	kfree_skb(skb);
1959 }
1960 
1961 /**
1962  * tipc_filter_connect - Handle incoming message for a connection-based socket
1963  * @tsk: TIPC socket
1964  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1965  *
1966  * Returns true if everything ok, false otherwise
1967  */
1968 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1969 {
1970 	struct sock *sk = &tsk->sk;
1971 	struct net *net = sock_net(sk);
1972 	struct tipc_msg *hdr = buf_msg(skb);
1973 	u32 pport = msg_origport(hdr);
1974 	u32 pnode = msg_orignode(hdr);
1975 
1976 	if (unlikely(msg_mcast(hdr)))
1977 		return false;
1978 
1979 	switch (sk->sk_state) {
1980 	case TIPC_CONNECTING:
1981 		/* Accept only ACK or NACK message */
1982 		if (unlikely(!msg_connected(hdr))) {
1983 			if (pport != tsk_peer_port(tsk) ||
1984 			    pnode != tsk_peer_node(tsk))
1985 				return false;
1986 
1987 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1988 			sk->sk_err = ECONNREFUSED;
1989 			sk->sk_state_change(sk);
1990 			return true;
1991 		}
1992 
1993 		if (unlikely(msg_errcode(hdr))) {
1994 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1995 			sk->sk_err = ECONNREFUSED;
1996 			sk->sk_state_change(sk);
1997 			return true;
1998 		}
1999 
2000 		if (unlikely(!msg_isdata(hdr))) {
2001 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2002 			sk->sk_err = EINVAL;
2003 			sk->sk_state_change(sk);
2004 			return true;
2005 		}
2006 
2007 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2008 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
2009 
2010 		/* If 'ACK+' message, add to socket receive queue */
2011 		if (msg_data_sz(hdr))
2012 			return true;
2013 
2014 		/* If empty 'ACK-' message, wake up sleeping connect() */
2015 		sk->sk_data_ready(sk);
2016 
2017 		/* 'ACK-' message is neither accepted nor rejected: */
2018 		msg_set_dest_droppable(hdr, 1);
2019 		return false;
2020 
2021 	case TIPC_OPEN:
2022 	case TIPC_DISCONNECTING:
2023 		break;
2024 	case TIPC_LISTEN:
2025 		/* Accept only SYN message */
2026 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2027 			return true;
2028 		break;
2029 	case TIPC_ESTABLISHED:
2030 		/* Accept only connection-based messages sent by peer */
2031 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
2032 			return false;
2033 
2034 		if (unlikely(msg_errcode(hdr))) {
2035 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2036 			/* Let timer expire on it's own */
2037 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2038 					      tsk->portid);
2039 			sk->sk_state_change(sk);
2040 		}
2041 		return true;
2042 	default:
2043 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2044 	}
2045 
2046 	return false;
2047 }
2048 
2049 /**
2050  * rcvbuf_limit - get proper overload limit of socket receive queue
2051  * @sk: socket
2052  * @skb: message
2053  *
2054  * For connection oriented messages, irrespective of importance,
2055  * default queue limit is 2 MB.
2056  *
2057  * For connectionless messages, queue limits are based on message
2058  * importance as follows:
2059  *
2060  * TIPC_LOW_IMPORTANCE       (2 MB)
2061  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2062  * TIPC_HIGH_IMPORTANCE      (8 MB)
2063  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2064  *
2065  * Returns overload limit according to corresponding message importance
2066  */
2067 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2068 {
2069 	struct tipc_sock *tsk = tipc_sk(sk);
2070 	struct tipc_msg *hdr = buf_msg(skb);
2071 
2072 	if (unlikely(msg_in_group(hdr)))
2073 		return sk->sk_rcvbuf;
2074 
2075 	if (unlikely(!msg_connected(hdr)))
2076 		return sk->sk_rcvbuf << msg_importance(hdr);
2077 
2078 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2079 		return sk->sk_rcvbuf;
2080 
2081 	return FLOWCTL_MSG_LIM;
2082 }
2083 
2084 /**
2085  * tipc_sk_filter_rcv - validate incoming message
2086  * @sk: socket
2087  * @skb: pointer to message.
2088  *
2089  * Enqueues message on receive queue if acceptable; optionally handles
2090  * disconnect indication for a connected socket.
2091  *
2092  * Called with socket lock already taken
2093  *
2094  */
2095 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2096 			       struct sk_buff_head *xmitq)
2097 {
2098 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2099 	struct tipc_sock *tsk = tipc_sk(sk);
2100 	struct tipc_group *grp = tsk->group;
2101 	struct tipc_msg *hdr = buf_msg(skb);
2102 	struct net *net = sock_net(sk);
2103 	struct sk_buff_head inputq;
2104 	int limit, err = TIPC_OK;
2105 
2106 	TIPC_SKB_CB(skb)->bytes_read = 0;
2107 	__skb_queue_head_init(&inputq);
2108 	__skb_queue_tail(&inputq, skb);
2109 
2110 	if (unlikely(!msg_isdata(hdr)))
2111 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2112 
2113 	if (unlikely(grp))
2114 		tipc_group_filter_msg(grp, &inputq, xmitq);
2115 
2116 	/* Validate and add to receive buffer if there is space */
2117 	while ((skb = __skb_dequeue(&inputq))) {
2118 		hdr = buf_msg(skb);
2119 		limit = rcvbuf_limit(sk, skb);
2120 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2121 		    (!sk_conn && msg_connected(hdr)) ||
2122 		    (!grp && msg_in_group(hdr)))
2123 			err = TIPC_ERR_NO_PORT;
2124 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2125 			atomic_inc(&sk->sk_drops);
2126 			err = TIPC_ERR_OVERLOAD;
2127 		}
2128 
2129 		if (unlikely(err)) {
2130 			tipc_skb_reject(net, err, skb, xmitq);
2131 			err = TIPC_OK;
2132 			continue;
2133 		}
2134 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2135 		skb_set_owner_r(skb, sk);
2136 		sk->sk_data_ready(sk);
2137 	}
2138 }
2139 
2140 /**
2141  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2142  * @sk: socket
2143  * @skb: message
2144  *
2145  * Caller must hold socket lock
2146  */
2147 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2148 {
2149 	unsigned int before = sk_rmem_alloc_get(sk);
2150 	struct sk_buff_head xmitq;
2151 	unsigned int added;
2152 
2153 	__skb_queue_head_init(&xmitq);
2154 
2155 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2156 	added = sk_rmem_alloc_get(sk) - before;
2157 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2158 
2159 	/* Send pending response/rejected messages, if any */
2160 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2161 	return 0;
2162 }
2163 
2164 /**
2165  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2166  *                   inputq and try adding them to socket or backlog queue
2167  * @inputq: list of incoming buffers with potentially different destinations
2168  * @sk: socket where the buffers should be enqueued
2169  * @dport: port number for the socket
2170  *
2171  * Caller must hold socket lock
2172  */
2173 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2174 			    u32 dport, struct sk_buff_head *xmitq)
2175 {
2176 	unsigned long time_limit = jiffies + 2;
2177 	struct sk_buff *skb;
2178 	unsigned int lim;
2179 	atomic_t *dcnt;
2180 	u32 onode;
2181 
2182 	while (skb_queue_len(inputq)) {
2183 		if (unlikely(time_after_eq(jiffies, time_limit)))
2184 			return;
2185 
2186 		skb = tipc_skb_dequeue(inputq, dport);
2187 		if (unlikely(!skb))
2188 			return;
2189 
2190 		/* Add message directly to receive queue if possible */
2191 		if (!sock_owned_by_user(sk)) {
2192 			tipc_sk_filter_rcv(sk, skb, xmitq);
2193 			continue;
2194 		}
2195 
2196 		/* Try backlog, compensating for double-counted bytes */
2197 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2198 		if (!sk->sk_backlog.len)
2199 			atomic_set(dcnt, 0);
2200 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2201 		if (likely(!sk_add_backlog(sk, skb, lim)))
2202 			continue;
2203 
2204 		/* Overload => reject message back to sender */
2205 		onode = tipc_own_addr(sock_net(sk));
2206 		atomic_inc(&sk->sk_drops);
2207 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2208 			__skb_queue_tail(xmitq, skb);
2209 		break;
2210 	}
2211 }
2212 
2213 /**
2214  * tipc_sk_rcv - handle a chain of incoming buffers
2215  * @inputq: buffer list containing the buffers
2216  * Consumes all buffers in list until inputq is empty
2217  * Note: may be called in multiple threads referring to the same queue
2218  */
2219 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2220 {
2221 	struct sk_buff_head xmitq;
2222 	u32 dnode, dport = 0;
2223 	int err;
2224 	struct tipc_sock *tsk;
2225 	struct sock *sk;
2226 	struct sk_buff *skb;
2227 
2228 	__skb_queue_head_init(&xmitq);
2229 	while (skb_queue_len(inputq)) {
2230 		dport = tipc_skb_peek_port(inputq, dport);
2231 		tsk = tipc_sk_lookup(net, dport);
2232 
2233 		if (likely(tsk)) {
2234 			sk = &tsk->sk;
2235 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2236 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2237 				spin_unlock_bh(&sk->sk_lock.slock);
2238 			}
2239 			/* Send pending response/rejected messages, if any */
2240 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2241 			sock_put(sk);
2242 			continue;
2243 		}
2244 		/* No destination socket => dequeue skb if still there */
2245 		skb = tipc_skb_dequeue(inputq, dport);
2246 		if (!skb)
2247 			return;
2248 
2249 		/* Try secondary lookup if unresolved named message */
2250 		err = TIPC_ERR_NO_PORT;
2251 		if (tipc_msg_lookup_dest(net, skb, &err))
2252 			goto xmit;
2253 
2254 		/* Prepare for message rejection */
2255 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2256 			continue;
2257 xmit:
2258 		dnode = msg_destnode(buf_msg(skb));
2259 		tipc_node_xmit_skb(net, skb, dnode, dport);
2260 	}
2261 }
2262 
2263 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2264 {
2265 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2266 	struct sock *sk = sock->sk;
2267 	int done;
2268 
2269 	do {
2270 		int err = sock_error(sk);
2271 		if (err)
2272 			return err;
2273 		if (!*timeo_p)
2274 			return -ETIMEDOUT;
2275 		if (signal_pending(current))
2276 			return sock_intr_errno(*timeo_p);
2277 
2278 		add_wait_queue(sk_sleep(sk), &wait);
2279 		done = sk_wait_event(sk, timeo_p,
2280 				     sk->sk_state != TIPC_CONNECTING, &wait);
2281 		remove_wait_queue(sk_sleep(sk), &wait);
2282 	} while (!done);
2283 	return 0;
2284 }
2285 
2286 /**
2287  * tipc_connect - establish a connection to another TIPC port
2288  * @sock: socket structure
2289  * @dest: socket address for destination port
2290  * @destlen: size of socket address data structure
2291  * @flags: file-related flags associated with socket
2292  *
2293  * Returns 0 on success, errno otherwise
2294  */
2295 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2296 			int destlen, int flags)
2297 {
2298 	struct sock *sk = sock->sk;
2299 	struct tipc_sock *tsk = tipc_sk(sk);
2300 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2301 	struct msghdr m = {NULL,};
2302 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2303 	int previous;
2304 	int res = 0;
2305 
2306 	if (destlen != sizeof(struct sockaddr_tipc))
2307 		return -EINVAL;
2308 
2309 	lock_sock(sk);
2310 
2311 	if (tsk->group) {
2312 		res = -EINVAL;
2313 		goto exit;
2314 	}
2315 
2316 	if (dst->family == AF_UNSPEC) {
2317 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2318 		if (!tipc_sk_type_connectionless(sk))
2319 			res = -EINVAL;
2320 		goto exit;
2321 	} else if (dst->family != AF_TIPC) {
2322 		res = -EINVAL;
2323 	}
2324 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2325 		res = -EINVAL;
2326 	if (res)
2327 		goto exit;
2328 
2329 	/* DGRAM/RDM connect(), just save the destaddr */
2330 	if (tipc_sk_type_connectionless(sk)) {
2331 		memcpy(&tsk->peer, dest, destlen);
2332 		goto exit;
2333 	}
2334 
2335 	previous = sk->sk_state;
2336 
2337 	switch (sk->sk_state) {
2338 	case TIPC_OPEN:
2339 		/* Send a 'SYN-' to destination */
2340 		m.msg_name = dest;
2341 		m.msg_namelen = destlen;
2342 
2343 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2344 		 * indicate send_msg() is never blocked.
2345 		 */
2346 		if (!timeout)
2347 			m.msg_flags = MSG_DONTWAIT;
2348 
2349 		res = __tipc_sendmsg(sock, &m, 0);
2350 		if ((res < 0) && (res != -EWOULDBLOCK))
2351 			goto exit;
2352 
2353 		/* Just entered TIPC_CONNECTING state; the only
2354 		 * difference is that return value in non-blocking
2355 		 * case is EINPROGRESS, rather than EALREADY.
2356 		 */
2357 		res = -EINPROGRESS;
2358 		/* fall thru' */
2359 	case TIPC_CONNECTING:
2360 		if (!timeout) {
2361 			if (previous == TIPC_CONNECTING)
2362 				res = -EALREADY;
2363 			goto exit;
2364 		}
2365 		timeout = msecs_to_jiffies(timeout);
2366 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2367 		res = tipc_wait_for_connect(sock, &timeout);
2368 		break;
2369 	case TIPC_ESTABLISHED:
2370 		res = -EISCONN;
2371 		break;
2372 	default:
2373 		res = -EINVAL;
2374 	}
2375 
2376 exit:
2377 	release_sock(sk);
2378 	return res;
2379 }
2380 
2381 /**
2382  * tipc_listen - allow socket to listen for incoming connections
2383  * @sock: socket structure
2384  * @len: (unused)
2385  *
2386  * Returns 0 on success, errno otherwise
2387  */
2388 static int tipc_listen(struct socket *sock, int len)
2389 {
2390 	struct sock *sk = sock->sk;
2391 	int res;
2392 
2393 	lock_sock(sk);
2394 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2395 	release_sock(sk);
2396 
2397 	return res;
2398 }
2399 
2400 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2401 {
2402 	struct sock *sk = sock->sk;
2403 	DEFINE_WAIT(wait);
2404 	int err;
2405 
2406 	/* True wake-one mechanism for incoming connections: only
2407 	 * one process gets woken up, not the 'whole herd'.
2408 	 * Since we do not 'race & poll' for established sockets
2409 	 * anymore, the common case will execute the loop only once.
2410 	*/
2411 	for (;;) {
2412 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2413 					  TASK_INTERRUPTIBLE);
2414 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2415 			release_sock(sk);
2416 			timeo = schedule_timeout(timeo);
2417 			lock_sock(sk);
2418 		}
2419 		err = 0;
2420 		if (!skb_queue_empty(&sk->sk_receive_queue))
2421 			break;
2422 		err = -EAGAIN;
2423 		if (!timeo)
2424 			break;
2425 		err = sock_intr_errno(timeo);
2426 		if (signal_pending(current))
2427 			break;
2428 	}
2429 	finish_wait(sk_sleep(sk), &wait);
2430 	return err;
2431 }
2432 
2433 /**
2434  * tipc_accept - wait for connection request
2435  * @sock: listening socket
2436  * @newsock: new socket that is to be connected
2437  * @flags: file-related flags associated with socket
2438  *
2439  * Returns 0 on success, errno otherwise
2440  */
2441 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2442 		       bool kern)
2443 {
2444 	struct sock *new_sk, *sk = sock->sk;
2445 	struct sk_buff *buf;
2446 	struct tipc_sock *new_tsock;
2447 	struct tipc_msg *msg;
2448 	long timeo;
2449 	int res;
2450 
2451 	lock_sock(sk);
2452 
2453 	if (sk->sk_state != TIPC_LISTEN) {
2454 		res = -EINVAL;
2455 		goto exit;
2456 	}
2457 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2458 	res = tipc_wait_for_accept(sock, timeo);
2459 	if (res)
2460 		goto exit;
2461 
2462 	buf = skb_peek(&sk->sk_receive_queue);
2463 
2464 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2465 	if (res)
2466 		goto exit;
2467 	security_sk_clone(sock->sk, new_sock->sk);
2468 
2469 	new_sk = new_sock->sk;
2470 	new_tsock = tipc_sk(new_sk);
2471 	msg = buf_msg(buf);
2472 
2473 	/* we lock on new_sk; but lockdep sees the lock on sk */
2474 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2475 
2476 	/*
2477 	 * Reject any stray messages received by new socket
2478 	 * before the socket lock was taken (very, very unlikely)
2479 	 */
2480 	tsk_rej_rx_queue(new_sk);
2481 
2482 	/* Connect new socket to it's peer */
2483 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2484 
2485 	tsk_set_importance(new_tsock, msg_importance(msg));
2486 	if (msg_named(msg)) {
2487 		new_tsock->conn_type = msg_nametype(msg);
2488 		new_tsock->conn_instance = msg_nameinst(msg);
2489 	}
2490 
2491 	/*
2492 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2493 	 * Respond to 'SYN+' by queuing it on new socket.
2494 	 */
2495 	if (!msg_data_sz(msg)) {
2496 		struct msghdr m = {NULL,};
2497 
2498 		tsk_advance_rx_queue(sk);
2499 		__tipc_sendstream(new_sock, &m, 0);
2500 	} else {
2501 		__skb_dequeue(&sk->sk_receive_queue);
2502 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2503 		skb_set_owner_r(buf, new_sk);
2504 	}
2505 	release_sock(new_sk);
2506 exit:
2507 	release_sock(sk);
2508 	return res;
2509 }
2510 
2511 /**
2512  * tipc_shutdown - shutdown socket connection
2513  * @sock: socket structure
2514  * @how: direction to close (must be SHUT_RDWR)
2515  *
2516  * Terminates connection (if necessary), then purges socket's receive queue.
2517  *
2518  * Returns 0 on success, errno otherwise
2519  */
2520 static int tipc_shutdown(struct socket *sock, int how)
2521 {
2522 	struct sock *sk = sock->sk;
2523 	int res;
2524 
2525 	if (how != SHUT_RDWR)
2526 		return -EINVAL;
2527 
2528 	lock_sock(sk);
2529 
2530 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2531 	sk->sk_shutdown = SEND_SHUTDOWN;
2532 
2533 	if (sk->sk_state == TIPC_DISCONNECTING) {
2534 		/* Discard any unreceived messages */
2535 		__skb_queue_purge(&sk->sk_receive_queue);
2536 
2537 		/* Wake up anyone sleeping in poll */
2538 		sk->sk_state_change(sk);
2539 		res = 0;
2540 	} else {
2541 		res = -ENOTCONN;
2542 	}
2543 
2544 	release_sock(sk);
2545 	return res;
2546 }
2547 
2548 static void tipc_sk_timeout(struct timer_list *t)
2549 {
2550 	struct sock *sk = from_timer(sk, t, sk_timer);
2551 	struct tipc_sock *tsk = tipc_sk(sk);
2552 	u32 peer_port = tsk_peer_port(tsk);
2553 	u32 peer_node = tsk_peer_node(tsk);
2554 	u32 own_node = tsk_own_node(tsk);
2555 	u32 own_port = tsk->portid;
2556 	struct net *net = sock_net(sk);
2557 	struct sk_buff *skb = NULL;
2558 
2559 	bh_lock_sock(sk);
2560 	if (!tipc_sk_connected(sk))
2561 		goto exit;
2562 
2563 	/* Try again later if socket is busy */
2564 	if (sock_owned_by_user(sk)) {
2565 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2566 		goto exit;
2567 	}
2568 
2569 	if (tsk->probe_unacked) {
2570 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2571 		tipc_node_remove_conn(net, peer_node, peer_port);
2572 		sk->sk_state_change(sk);
2573 		goto exit;
2574 	}
2575 	/* Send new probe */
2576 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2577 			      peer_node, own_node, peer_port, own_port,
2578 			      TIPC_OK);
2579 	tsk->probe_unacked = true;
2580 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2581 exit:
2582 	bh_unlock_sock(sk);
2583 	if (skb)
2584 		tipc_node_xmit_skb(net, skb, peer_node, own_port);
2585 	sock_put(sk);
2586 }
2587 
2588 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2589 			   struct tipc_name_seq const *seq)
2590 {
2591 	struct sock *sk = &tsk->sk;
2592 	struct net *net = sock_net(sk);
2593 	struct publication *publ;
2594 	u32 key;
2595 
2596 	if (scope != TIPC_NODE_SCOPE)
2597 		scope = TIPC_CLUSTER_SCOPE;
2598 
2599 	if (tipc_sk_connected(sk))
2600 		return -EINVAL;
2601 	key = tsk->portid + tsk->pub_count + 1;
2602 	if (key == tsk->portid)
2603 		return -EADDRINUSE;
2604 
2605 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2606 				    scope, tsk->portid, key);
2607 	if (unlikely(!publ))
2608 		return -EINVAL;
2609 
2610 	list_add(&publ->binding_sock, &tsk->publications);
2611 	tsk->pub_count++;
2612 	tsk->published = 1;
2613 	return 0;
2614 }
2615 
2616 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2617 			    struct tipc_name_seq const *seq)
2618 {
2619 	struct net *net = sock_net(&tsk->sk);
2620 	struct publication *publ;
2621 	struct publication *safe;
2622 	int rc = -EINVAL;
2623 
2624 	if (scope != TIPC_NODE_SCOPE)
2625 		scope = TIPC_CLUSTER_SCOPE;
2626 
2627 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2628 		if (seq) {
2629 			if (publ->scope != scope)
2630 				continue;
2631 			if (publ->type != seq->type)
2632 				continue;
2633 			if (publ->lower != seq->lower)
2634 				continue;
2635 			if (publ->upper != seq->upper)
2636 				break;
2637 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2638 					      publ->upper, publ->key);
2639 			rc = 0;
2640 			break;
2641 		}
2642 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2643 				      publ->upper, publ->key);
2644 		rc = 0;
2645 	}
2646 	if (list_empty(&tsk->publications))
2647 		tsk->published = 0;
2648 	return rc;
2649 }
2650 
2651 /* tipc_sk_reinit: set non-zero address in all existing sockets
2652  *                 when we go from standalone to network mode.
2653  */
2654 void tipc_sk_reinit(struct net *net)
2655 {
2656 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2657 	struct rhashtable_iter iter;
2658 	struct tipc_sock *tsk;
2659 	struct tipc_msg *msg;
2660 
2661 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2662 
2663 	do {
2664 		rhashtable_walk_start(&iter);
2665 
2666 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2667 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2668 			msg = &tsk->phdr;
2669 			msg_set_prevnode(msg, tipc_own_addr(net));
2670 			msg_set_orignode(msg, tipc_own_addr(net));
2671 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2672 		}
2673 
2674 		rhashtable_walk_stop(&iter);
2675 	} while (tsk == ERR_PTR(-EAGAIN));
2676 
2677 	rhashtable_walk_exit(&iter);
2678 }
2679 
2680 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2681 {
2682 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2683 	struct tipc_sock *tsk;
2684 
2685 	rcu_read_lock();
2686 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2687 	if (tsk)
2688 		sock_hold(&tsk->sk);
2689 	rcu_read_unlock();
2690 
2691 	return tsk;
2692 }
2693 
2694 static int tipc_sk_insert(struct tipc_sock *tsk)
2695 {
2696 	struct sock *sk = &tsk->sk;
2697 	struct net *net = sock_net(sk);
2698 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2699 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2700 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2701 
2702 	while (remaining--) {
2703 		portid++;
2704 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2705 			portid = TIPC_MIN_PORT;
2706 		tsk->portid = portid;
2707 		sock_hold(&tsk->sk);
2708 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2709 						   tsk_rht_params))
2710 			return 0;
2711 		sock_put(&tsk->sk);
2712 	}
2713 
2714 	return -1;
2715 }
2716 
2717 static void tipc_sk_remove(struct tipc_sock *tsk)
2718 {
2719 	struct sock *sk = &tsk->sk;
2720 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2721 
2722 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2723 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2724 		__sock_put(sk);
2725 	}
2726 }
2727 
2728 static const struct rhashtable_params tsk_rht_params = {
2729 	.nelem_hint = 192,
2730 	.head_offset = offsetof(struct tipc_sock, node),
2731 	.key_offset = offsetof(struct tipc_sock, portid),
2732 	.key_len = sizeof(u32), /* portid */
2733 	.max_size = 1048576,
2734 	.min_size = 256,
2735 	.automatic_shrinking = true,
2736 };
2737 
2738 int tipc_sk_rht_init(struct net *net)
2739 {
2740 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2741 
2742 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2743 }
2744 
2745 void tipc_sk_rht_destroy(struct net *net)
2746 {
2747 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2748 
2749 	/* Wait for socket readers to complete */
2750 	synchronize_net();
2751 
2752 	rhashtable_destroy(&tn->sk_rht);
2753 }
2754 
2755 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2756 {
2757 	struct net *net = sock_net(&tsk->sk);
2758 	struct tipc_group *grp = tsk->group;
2759 	struct tipc_msg *hdr = &tsk->phdr;
2760 	struct tipc_name_seq seq;
2761 	int rc;
2762 
2763 	if (mreq->type < TIPC_RESERVED_TYPES)
2764 		return -EACCES;
2765 	if (mreq->scope > TIPC_NODE_SCOPE)
2766 		return -EINVAL;
2767 	if (grp)
2768 		return -EACCES;
2769 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2770 	if (!grp)
2771 		return -ENOMEM;
2772 	tsk->group = grp;
2773 	msg_set_lookup_scope(hdr, mreq->scope);
2774 	msg_set_nametype(hdr, mreq->type);
2775 	msg_set_dest_droppable(hdr, true);
2776 	seq.type = mreq->type;
2777 	seq.lower = mreq->instance;
2778 	seq.upper = seq.lower;
2779 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2780 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2781 	if (rc) {
2782 		tipc_group_delete(net, grp);
2783 		tsk->group = NULL;
2784 		return rc;
2785 	}
2786 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2787 	tsk->mc_method.rcast = true;
2788 	tsk->mc_method.mandatory = true;
2789 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2790 	return rc;
2791 }
2792 
2793 static int tipc_sk_leave(struct tipc_sock *tsk)
2794 {
2795 	struct net *net = sock_net(&tsk->sk);
2796 	struct tipc_group *grp = tsk->group;
2797 	struct tipc_name_seq seq;
2798 	int scope;
2799 
2800 	if (!grp)
2801 		return -EINVAL;
2802 	tipc_group_self(grp, &seq, &scope);
2803 	tipc_group_delete(net, grp);
2804 	tsk->group = NULL;
2805 	tipc_sk_withdraw(tsk, scope, &seq);
2806 	return 0;
2807 }
2808 
2809 /**
2810  * tipc_setsockopt - set socket option
2811  * @sock: socket structure
2812  * @lvl: option level
2813  * @opt: option identifier
2814  * @ov: pointer to new option value
2815  * @ol: length of option value
2816  *
2817  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2818  * (to ease compatibility).
2819  *
2820  * Returns 0 on success, errno otherwise
2821  */
2822 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2823 			   char __user *ov, unsigned int ol)
2824 {
2825 	struct sock *sk = sock->sk;
2826 	struct tipc_sock *tsk = tipc_sk(sk);
2827 	struct tipc_group_req mreq;
2828 	u32 value = 0;
2829 	int res = 0;
2830 
2831 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2832 		return 0;
2833 	if (lvl != SOL_TIPC)
2834 		return -ENOPROTOOPT;
2835 
2836 	switch (opt) {
2837 	case TIPC_IMPORTANCE:
2838 	case TIPC_SRC_DROPPABLE:
2839 	case TIPC_DEST_DROPPABLE:
2840 	case TIPC_CONN_TIMEOUT:
2841 		if (ol < sizeof(value))
2842 			return -EINVAL;
2843 		if (get_user(value, (u32 __user *)ov))
2844 			return -EFAULT;
2845 		break;
2846 	case TIPC_GROUP_JOIN:
2847 		if (ol < sizeof(mreq))
2848 			return -EINVAL;
2849 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2850 			return -EFAULT;
2851 		break;
2852 	default:
2853 		if (ov || ol)
2854 			return -EINVAL;
2855 	}
2856 
2857 	lock_sock(sk);
2858 
2859 	switch (opt) {
2860 	case TIPC_IMPORTANCE:
2861 		res = tsk_set_importance(tsk, value);
2862 		break;
2863 	case TIPC_SRC_DROPPABLE:
2864 		if (sock->type != SOCK_STREAM)
2865 			tsk_set_unreliable(tsk, value);
2866 		else
2867 			res = -ENOPROTOOPT;
2868 		break;
2869 	case TIPC_DEST_DROPPABLE:
2870 		tsk_set_unreturnable(tsk, value);
2871 		break;
2872 	case TIPC_CONN_TIMEOUT:
2873 		tipc_sk(sk)->conn_timeout = value;
2874 		break;
2875 	case TIPC_MCAST_BROADCAST:
2876 		tsk->mc_method.rcast = false;
2877 		tsk->mc_method.mandatory = true;
2878 		break;
2879 	case TIPC_MCAST_REPLICAST:
2880 		tsk->mc_method.rcast = true;
2881 		tsk->mc_method.mandatory = true;
2882 		break;
2883 	case TIPC_GROUP_JOIN:
2884 		res = tipc_sk_join(tsk, &mreq);
2885 		break;
2886 	case TIPC_GROUP_LEAVE:
2887 		res = tipc_sk_leave(tsk);
2888 		break;
2889 	default:
2890 		res = -EINVAL;
2891 	}
2892 
2893 	release_sock(sk);
2894 
2895 	return res;
2896 }
2897 
2898 /**
2899  * tipc_getsockopt - get socket option
2900  * @sock: socket structure
2901  * @lvl: option level
2902  * @opt: option identifier
2903  * @ov: receptacle for option value
2904  * @ol: receptacle for length of option value
2905  *
2906  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2907  * (to ease compatibility).
2908  *
2909  * Returns 0 on success, errno otherwise
2910  */
2911 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2912 			   char __user *ov, int __user *ol)
2913 {
2914 	struct sock *sk = sock->sk;
2915 	struct tipc_sock *tsk = tipc_sk(sk);
2916 	struct tipc_name_seq seq;
2917 	int len, scope;
2918 	u32 value;
2919 	int res;
2920 
2921 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2922 		return put_user(0, ol);
2923 	if (lvl != SOL_TIPC)
2924 		return -ENOPROTOOPT;
2925 	res = get_user(len, ol);
2926 	if (res)
2927 		return res;
2928 
2929 	lock_sock(sk);
2930 
2931 	switch (opt) {
2932 	case TIPC_IMPORTANCE:
2933 		value = tsk_importance(tsk);
2934 		break;
2935 	case TIPC_SRC_DROPPABLE:
2936 		value = tsk_unreliable(tsk);
2937 		break;
2938 	case TIPC_DEST_DROPPABLE:
2939 		value = tsk_unreturnable(tsk);
2940 		break;
2941 	case TIPC_CONN_TIMEOUT:
2942 		value = tsk->conn_timeout;
2943 		/* no need to set "res", since already 0 at this point */
2944 		break;
2945 	case TIPC_NODE_RECVQ_DEPTH:
2946 		value = 0; /* was tipc_queue_size, now obsolete */
2947 		break;
2948 	case TIPC_SOCK_RECVQ_DEPTH:
2949 		value = skb_queue_len(&sk->sk_receive_queue);
2950 		break;
2951 	case TIPC_GROUP_JOIN:
2952 		seq.type = 0;
2953 		if (tsk->group)
2954 			tipc_group_self(tsk->group, &seq, &scope);
2955 		value = seq.type;
2956 		break;
2957 	default:
2958 		res = -EINVAL;
2959 	}
2960 
2961 	release_sock(sk);
2962 
2963 	if (res)
2964 		return res;	/* "get" failed */
2965 
2966 	if (len < sizeof(value))
2967 		return -EINVAL;
2968 
2969 	if (copy_to_user(ov, &value, sizeof(value)))
2970 		return -EFAULT;
2971 
2972 	return put_user(sizeof(value), ol);
2973 }
2974 
2975 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2976 {
2977 	struct net *net = sock_net(sock->sk);
2978 	struct tipc_sioc_nodeid_req nr = {0};
2979 	struct tipc_sioc_ln_req lnr;
2980 	void __user *argp = (void __user *)arg;
2981 
2982 	switch (cmd) {
2983 	case SIOCGETLINKNAME:
2984 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2985 			return -EFAULT;
2986 		if (!tipc_node_get_linkname(net,
2987 					    lnr.bearer_id & 0xffff, lnr.peer,
2988 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2989 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2990 				return -EFAULT;
2991 			return 0;
2992 		}
2993 		return -EADDRNOTAVAIL;
2994 	case SIOCGETNODEID:
2995 		if (copy_from_user(&nr, argp, sizeof(nr)))
2996 			return -EFAULT;
2997 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
2998 			return -EADDRNOTAVAIL;
2999 		if (copy_to_user(argp, &nr, sizeof(nr)))
3000 			return -EFAULT;
3001 		return 0;
3002 	default:
3003 		return -ENOIOCTLCMD;
3004 	}
3005 }
3006 
3007 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3008 {
3009 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3010 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3011 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3012 
3013 	tsk1->peer.family = AF_TIPC;
3014 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3015 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3016 	tsk1->peer.addr.id.ref = tsk2->portid;
3017 	tsk1->peer.addr.id.node = onode;
3018 	tsk2->peer.family = AF_TIPC;
3019 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3020 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3021 	tsk2->peer.addr.id.ref = tsk1->portid;
3022 	tsk2->peer.addr.id.node = onode;
3023 
3024 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3025 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3026 	return 0;
3027 }
3028 
3029 /* Protocol switches for the various types of TIPC sockets */
3030 
3031 static const struct proto_ops msg_ops = {
3032 	.owner		= THIS_MODULE,
3033 	.family		= AF_TIPC,
3034 	.release	= tipc_release,
3035 	.bind		= tipc_bind,
3036 	.connect	= tipc_connect,
3037 	.socketpair	= tipc_socketpair,
3038 	.accept		= sock_no_accept,
3039 	.getname	= tipc_getname,
3040 	.poll		= tipc_poll,
3041 	.ioctl		= tipc_ioctl,
3042 	.listen		= sock_no_listen,
3043 	.shutdown	= tipc_shutdown,
3044 	.setsockopt	= tipc_setsockopt,
3045 	.getsockopt	= tipc_getsockopt,
3046 	.sendmsg	= tipc_sendmsg,
3047 	.recvmsg	= tipc_recvmsg,
3048 	.mmap		= sock_no_mmap,
3049 	.sendpage	= sock_no_sendpage
3050 };
3051 
3052 static const struct proto_ops packet_ops = {
3053 	.owner		= THIS_MODULE,
3054 	.family		= AF_TIPC,
3055 	.release	= tipc_release,
3056 	.bind		= tipc_bind,
3057 	.connect	= tipc_connect,
3058 	.socketpair	= tipc_socketpair,
3059 	.accept		= tipc_accept,
3060 	.getname	= tipc_getname,
3061 	.poll		= tipc_poll,
3062 	.ioctl		= tipc_ioctl,
3063 	.listen		= tipc_listen,
3064 	.shutdown	= tipc_shutdown,
3065 	.setsockopt	= tipc_setsockopt,
3066 	.getsockopt	= tipc_getsockopt,
3067 	.sendmsg	= tipc_send_packet,
3068 	.recvmsg	= tipc_recvmsg,
3069 	.mmap		= sock_no_mmap,
3070 	.sendpage	= sock_no_sendpage
3071 };
3072 
3073 static const struct proto_ops stream_ops = {
3074 	.owner		= THIS_MODULE,
3075 	.family		= AF_TIPC,
3076 	.release	= tipc_release,
3077 	.bind		= tipc_bind,
3078 	.connect	= tipc_connect,
3079 	.socketpair	= tipc_socketpair,
3080 	.accept		= tipc_accept,
3081 	.getname	= tipc_getname,
3082 	.poll		= tipc_poll,
3083 	.ioctl		= tipc_ioctl,
3084 	.listen		= tipc_listen,
3085 	.shutdown	= tipc_shutdown,
3086 	.setsockopt	= tipc_setsockopt,
3087 	.getsockopt	= tipc_getsockopt,
3088 	.sendmsg	= tipc_sendstream,
3089 	.recvmsg	= tipc_recvstream,
3090 	.mmap		= sock_no_mmap,
3091 	.sendpage	= sock_no_sendpage
3092 };
3093 
3094 static const struct net_proto_family tipc_family_ops = {
3095 	.owner		= THIS_MODULE,
3096 	.family		= AF_TIPC,
3097 	.create		= tipc_sk_create
3098 };
3099 
3100 static struct proto tipc_proto = {
3101 	.name		= "TIPC",
3102 	.owner		= THIS_MODULE,
3103 	.obj_size	= sizeof(struct tipc_sock),
3104 	.sysctl_rmem	= sysctl_tipc_rmem
3105 };
3106 
3107 /**
3108  * tipc_socket_init - initialize TIPC socket interface
3109  *
3110  * Returns 0 on success, errno otherwise
3111  */
3112 int tipc_socket_init(void)
3113 {
3114 	int res;
3115 
3116 	res = proto_register(&tipc_proto, 1);
3117 	if (res) {
3118 		pr_err("Failed to register TIPC protocol type\n");
3119 		goto out;
3120 	}
3121 
3122 	res = sock_register(&tipc_family_ops);
3123 	if (res) {
3124 		pr_err("Failed to register TIPC socket type\n");
3125 		proto_unregister(&tipc_proto);
3126 		goto out;
3127 	}
3128  out:
3129 	return res;
3130 }
3131 
3132 /**
3133  * tipc_socket_stop - stop TIPC socket interface
3134  */
3135 void tipc_socket_stop(void)
3136 {
3137 	sock_unregister(tipc_family_ops.family);
3138 	proto_unregister(&tipc_proto);
3139 }
3140 
3141 /* Caller should hold socket lock for the passed tipc socket. */
3142 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3143 {
3144 	u32 peer_node;
3145 	u32 peer_port;
3146 	struct nlattr *nest;
3147 
3148 	peer_node = tsk_peer_node(tsk);
3149 	peer_port = tsk_peer_port(tsk);
3150 
3151 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3152 
3153 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3154 		goto msg_full;
3155 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3156 		goto msg_full;
3157 
3158 	if (tsk->conn_type != 0) {
3159 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3160 			goto msg_full;
3161 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3162 			goto msg_full;
3163 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3164 			goto msg_full;
3165 	}
3166 	nla_nest_end(skb, nest);
3167 
3168 	return 0;
3169 
3170 msg_full:
3171 	nla_nest_cancel(skb, nest);
3172 
3173 	return -EMSGSIZE;
3174 }
3175 
3176 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3177 			  *tsk)
3178 {
3179 	struct net *net = sock_net(skb->sk);
3180 	struct sock *sk = &tsk->sk;
3181 
3182 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3183 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3184 		return -EMSGSIZE;
3185 
3186 	if (tipc_sk_connected(sk)) {
3187 		if (__tipc_nl_add_sk_con(skb, tsk))
3188 			return -EMSGSIZE;
3189 	} else if (!list_empty(&tsk->publications)) {
3190 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3191 			return -EMSGSIZE;
3192 	}
3193 	return 0;
3194 }
3195 
3196 /* Caller should hold socket lock for the passed tipc socket. */
3197 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3198 			    struct tipc_sock *tsk)
3199 {
3200 	struct nlattr *attrs;
3201 	void *hdr;
3202 
3203 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3204 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3205 	if (!hdr)
3206 		goto msg_cancel;
3207 
3208 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3209 	if (!attrs)
3210 		goto genlmsg_cancel;
3211 
3212 	if (__tipc_nl_add_sk_info(skb, tsk))
3213 		goto attr_msg_cancel;
3214 
3215 	nla_nest_end(skb, attrs);
3216 	genlmsg_end(skb, hdr);
3217 
3218 	return 0;
3219 
3220 attr_msg_cancel:
3221 	nla_nest_cancel(skb, attrs);
3222 genlmsg_cancel:
3223 	genlmsg_cancel(skb, hdr);
3224 msg_cancel:
3225 	return -EMSGSIZE;
3226 }
3227 
3228 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3229 		    int (*skb_handler)(struct sk_buff *skb,
3230 				       struct netlink_callback *cb,
3231 				       struct tipc_sock *tsk))
3232 {
3233 	struct rhashtable_iter *iter = (void *)cb->args[4];
3234 	struct tipc_sock *tsk;
3235 	int err;
3236 
3237 	rhashtable_walk_start(iter);
3238 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3239 		if (IS_ERR(tsk)) {
3240 			err = PTR_ERR(tsk);
3241 			if (err == -EAGAIN) {
3242 				err = 0;
3243 				continue;
3244 			}
3245 			break;
3246 		}
3247 
3248 		sock_hold(&tsk->sk);
3249 		rhashtable_walk_stop(iter);
3250 		lock_sock(&tsk->sk);
3251 		err = skb_handler(skb, cb, tsk);
3252 		if (err) {
3253 			release_sock(&tsk->sk);
3254 			sock_put(&tsk->sk);
3255 			goto out;
3256 		}
3257 		release_sock(&tsk->sk);
3258 		rhashtable_walk_start(iter);
3259 		sock_put(&tsk->sk);
3260 	}
3261 	rhashtable_walk_stop(iter);
3262 out:
3263 	return skb->len;
3264 }
3265 EXPORT_SYMBOL(tipc_nl_sk_walk);
3266 
3267 int tipc_dump_start(struct netlink_callback *cb)
3268 {
3269 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3270 }
3271 EXPORT_SYMBOL(tipc_dump_start);
3272 
3273 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3274 {
3275 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3276 	struct rhashtable_iter *iter = (void *)cb->args[4];
3277 	struct tipc_net *tn = tipc_net(net);
3278 
3279 	if (!iter) {
3280 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3281 		if (!iter)
3282 			return -ENOMEM;
3283 
3284 		cb->args[4] = (long)iter;
3285 	}
3286 
3287 	rhashtable_walk_enter(&tn->sk_rht, iter);
3288 	return 0;
3289 }
3290 
3291 int tipc_dump_done(struct netlink_callback *cb)
3292 {
3293 	struct rhashtable_iter *hti = (void *)cb->args[4];
3294 
3295 	rhashtable_walk_exit(hti);
3296 	kfree(hti);
3297 	return 0;
3298 }
3299 EXPORT_SYMBOL(tipc_dump_done);
3300 
3301 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3302 			   struct tipc_sock *tsk, u32 sk_filter_state,
3303 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3304 {
3305 	struct sock *sk = &tsk->sk;
3306 	struct nlattr *attrs;
3307 	struct nlattr *stat;
3308 
3309 	/*filter response w.r.t sk_state*/
3310 	if (!(sk_filter_state & (1 << sk->sk_state)))
3311 		return 0;
3312 
3313 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3314 	if (!attrs)
3315 		goto msg_cancel;
3316 
3317 	if (__tipc_nl_add_sk_info(skb, tsk))
3318 		goto attr_msg_cancel;
3319 
3320 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3321 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3322 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3323 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3324 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3325 					 sock_i_uid(sk))) ||
3326 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3327 			      tipc_diag_gen_cookie(sk),
3328 			      TIPC_NLA_SOCK_PAD))
3329 		goto attr_msg_cancel;
3330 
3331 	stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3332 	if (!stat)
3333 		goto attr_msg_cancel;
3334 
3335 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3336 			skb_queue_len(&sk->sk_receive_queue)) ||
3337 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3338 			skb_queue_len(&sk->sk_write_queue)) ||
3339 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3340 			atomic_read(&sk->sk_drops)))
3341 		goto stat_msg_cancel;
3342 
3343 	if (tsk->cong_link_cnt &&
3344 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3345 		goto stat_msg_cancel;
3346 
3347 	if (tsk_conn_cong(tsk) &&
3348 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3349 		goto stat_msg_cancel;
3350 
3351 	nla_nest_end(skb, stat);
3352 
3353 	if (tsk->group)
3354 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3355 			goto stat_msg_cancel;
3356 
3357 	nla_nest_end(skb, attrs);
3358 
3359 	return 0;
3360 
3361 stat_msg_cancel:
3362 	nla_nest_cancel(skb, stat);
3363 attr_msg_cancel:
3364 	nla_nest_cancel(skb, attrs);
3365 msg_cancel:
3366 	return -EMSGSIZE;
3367 }
3368 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3369 
3370 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3371 {
3372 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3373 }
3374 
3375 /* Caller should hold socket lock for the passed tipc socket. */
3376 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3377 				 struct netlink_callback *cb,
3378 				 struct publication *publ)
3379 {
3380 	void *hdr;
3381 	struct nlattr *attrs;
3382 
3383 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3384 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3385 	if (!hdr)
3386 		goto msg_cancel;
3387 
3388 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3389 	if (!attrs)
3390 		goto genlmsg_cancel;
3391 
3392 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3393 		goto attr_msg_cancel;
3394 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3395 		goto attr_msg_cancel;
3396 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3397 		goto attr_msg_cancel;
3398 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3399 		goto attr_msg_cancel;
3400 
3401 	nla_nest_end(skb, attrs);
3402 	genlmsg_end(skb, hdr);
3403 
3404 	return 0;
3405 
3406 attr_msg_cancel:
3407 	nla_nest_cancel(skb, attrs);
3408 genlmsg_cancel:
3409 	genlmsg_cancel(skb, hdr);
3410 msg_cancel:
3411 	return -EMSGSIZE;
3412 }
3413 
3414 /* Caller should hold socket lock for the passed tipc socket. */
3415 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3416 				  struct netlink_callback *cb,
3417 				  struct tipc_sock *tsk, u32 *last_publ)
3418 {
3419 	int err;
3420 	struct publication *p;
3421 
3422 	if (*last_publ) {
3423 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3424 			if (p->key == *last_publ)
3425 				break;
3426 		}
3427 		if (p->key != *last_publ) {
3428 			/* We never set seq or call nl_dump_check_consistent()
3429 			 * this means that setting prev_seq here will cause the
3430 			 * consistence check to fail in the netlink callback
3431 			 * handler. Resulting in the last NLMSG_DONE message
3432 			 * having the NLM_F_DUMP_INTR flag set.
3433 			 */
3434 			cb->prev_seq = 1;
3435 			*last_publ = 0;
3436 			return -EPIPE;
3437 		}
3438 	} else {
3439 		p = list_first_entry(&tsk->publications, struct publication,
3440 				     binding_sock);
3441 	}
3442 
3443 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3444 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3445 		if (err) {
3446 			*last_publ = p->key;
3447 			return err;
3448 		}
3449 	}
3450 	*last_publ = 0;
3451 
3452 	return 0;
3453 }
3454 
3455 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3456 {
3457 	int err;
3458 	u32 tsk_portid = cb->args[0];
3459 	u32 last_publ = cb->args[1];
3460 	u32 done = cb->args[2];
3461 	struct net *net = sock_net(skb->sk);
3462 	struct tipc_sock *tsk;
3463 
3464 	if (!tsk_portid) {
3465 		struct nlattr **attrs;
3466 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3467 
3468 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3469 		if (err)
3470 			return err;
3471 
3472 		if (!attrs[TIPC_NLA_SOCK])
3473 			return -EINVAL;
3474 
3475 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3476 				       attrs[TIPC_NLA_SOCK],
3477 				       tipc_nl_sock_policy, NULL);
3478 		if (err)
3479 			return err;
3480 
3481 		if (!sock[TIPC_NLA_SOCK_REF])
3482 			return -EINVAL;
3483 
3484 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3485 	}
3486 
3487 	if (done)
3488 		return 0;
3489 
3490 	tsk = tipc_sk_lookup(net, tsk_portid);
3491 	if (!tsk)
3492 		return -EINVAL;
3493 
3494 	lock_sock(&tsk->sk);
3495 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3496 	if (!err)
3497 		done = 1;
3498 	release_sock(&tsk->sk);
3499 	sock_put(&tsk->sk);
3500 
3501 	cb->args[0] = tsk_portid;
3502 	cb->args[1] = last_publ;
3503 	cb->args[2] = done;
3504 
3505 	return skb->len;
3506 }
3507