xref: /openbmc/linux/net/tipc/socket.c (revision 3a448205)
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  * @conn_timeout: the time we can wait for an unresponded setup request
84  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
85  * @cong_link_cnt: number of congested links
86  * @snt_unacked: # messages sent by socket, and not yet acked by peer
87  * @rcv_unacked: # messages read by user, but not yet acked back to peer
88  * @peer: 'connected' peer for dgram/rdm
89  * @node: hash table node
90  * @mc_method: cookie for use between socket and broadcast layer
91  * @rcu: rcu struct for tipc_sock
92  */
93 struct tipc_sock {
94 	struct sock sk;
95 	u32 conn_type;
96 	u32 conn_instance;
97 	int published;
98 	u32 max_pkt;
99 	u32 portid;
100 	struct tipc_msg phdr;
101 	struct list_head cong_links;
102 	struct list_head publications;
103 	u32 pub_count;
104 	atomic_t dupl_rcvcnt;
105 	u16 conn_timeout;
106 	bool probe_unacked;
107 	u16 cong_link_cnt;
108 	u16 snt_unacked;
109 	u16 snd_win;
110 	u16 peer_caps;
111 	u16 rcv_unacked;
112 	u16 rcv_win;
113 	struct sockaddr_tipc peer;
114 	struct rhash_head node;
115 	struct tipc_mc_method mc_method;
116 	struct rcu_head rcu;
117 	struct tipc_group *group;
118 	bool group_is_open;
119 };
120 
121 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
122 static void tipc_data_ready(struct sock *sk);
123 static void tipc_write_space(struct sock *sk);
124 static void tipc_sock_destruct(struct sock *sk);
125 static int tipc_release(struct socket *sock);
126 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
127 		       bool kern);
128 static void tipc_sk_timeout(struct timer_list *t);
129 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
130 			   struct tipc_name_seq const *seq);
131 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
132 			    struct tipc_name_seq const *seq);
133 static int tipc_sk_leave(struct tipc_sock *tsk);
134 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
135 static int tipc_sk_insert(struct tipc_sock *tsk);
136 static void tipc_sk_remove(struct tipc_sock *tsk);
137 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
138 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
139 
140 static const struct proto_ops packet_ops;
141 static const struct proto_ops stream_ops;
142 static const struct proto_ops msg_ops;
143 static struct proto tipc_proto;
144 static const struct rhashtable_params tsk_rht_params;
145 
146 static u32 tsk_own_node(struct tipc_sock *tsk)
147 {
148 	return msg_prevnode(&tsk->phdr);
149 }
150 
151 static u32 tsk_peer_node(struct tipc_sock *tsk)
152 {
153 	return msg_destnode(&tsk->phdr);
154 }
155 
156 static u32 tsk_peer_port(struct tipc_sock *tsk)
157 {
158 	return msg_destport(&tsk->phdr);
159 }
160 
161 static  bool tsk_unreliable(struct tipc_sock *tsk)
162 {
163 	return msg_src_droppable(&tsk->phdr) != 0;
164 }
165 
166 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
167 {
168 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
169 }
170 
171 static bool tsk_unreturnable(struct tipc_sock *tsk)
172 {
173 	return msg_dest_droppable(&tsk->phdr) != 0;
174 }
175 
176 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
177 {
178 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
179 }
180 
181 static int tsk_importance(struct tipc_sock *tsk)
182 {
183 	return msg_importance(&tsk->phdr);
184 }
185 
186 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
187 {
188 	if (imp > TIPC_CRITICAL_IMPORTANCE)
189 		return -EINVAL;
190 	msg_set_importance(&tsk->phdr, (u32)imp);
191 	return 0;
192 }
193 
194 static struct tipc_sock *tipc_sk(const struct sock *sk)
195 {
196 	return container_of(sk, struct tipc_sock, sk);
197 }
198 
199 static bool tsk_conn_cong(struct tipc_sock *tsk)
200 {
201 	return tsk->snt_unacked > tsk->snd_win;
202 }
203 
204 static u16 tsk_blocks(int len)
205 {
206 	return ((len / FLOWCTL_BLK_SZ) + 1);
207 }
208 
209 /* tsk_blocks(): translate a buffer size in bytes to number of
210  * advertisable blocks, taking into account the ratio truesize(len)/len
211  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
212  */
213 static u16 tsk_adv_blocks(int len)
214 {
215 	return len / FLOWCTL_BLK_SZ / 4;
216 }
217 
218 /* tsk_inc(): increment counter for sent or received data
219  * - If block based flow control is not supported by peer we
220  *   fall back to message based ditto, incrementing the counter
221  */
222 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
223 {
224 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
225 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
226 	return 1;
227 }
228 
229 /**
230  * tsk_advance_rx_queue - discard first buffer in socket receive queue
231  *
232  * Caller must hold socket lock
233  */
234 static void tsk_advance_rx_queue(struct sock *sk)
235 {
236 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
237 }
238 
239 /* tipc_sk_respond() : send response message back to sender
240  */
241 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
242 {
243 	u32 selector;
244 	u32 dnode;
245 	u32 onode = tipc_own_addr(sock_net(sk));
246 
247 	if (!tipc_msg_reverse(onode, &skb, err))
248 		return;
249 
250 	dnode = msg_destnode(buf_msg(skb));
251 	selector = msg_origport(buf_msg(skb));
252 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
253 }
254 
255 /**
256  * tsk_rej_rx_queue - reject all buffers in socket receive queue
257  *
258  * Caller must hold socket lock
259  */
260 static void tsk_rej_rx_queue(struct sock *sk)
261 {
262 	struct sk_buff *skb;
263 
264 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
265 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
266 }
267 
268 static bool tipc_sk_connected(struct sock *sk)
269 {
270 	return sk->sk_state == TIPC_ESTABLISHED;
271 }
272 
273 /* tipc_sk_type_connectionless - check if the socket is datagram socket
274  * @sk: socket
275  *
276  * Returns true if connection less, false otherwise
277  */
278 static bool tipc_sk_type_connectionless(struct sock *sk)
279 {
280 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
281 }
282 
283 /* tsk_peer_msg - verify if message was sent by connected port's peer
284  *
285  * Handles cases where the node's network address has changed from
286  * the default of <0.0.0> to its configured setting.
287  */
288 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
289 {
290 	struct sock *sk = &tsk->sk;
291 	u32 self = tipc_own_addr(sock_net(sk));
292 	u32 peer_port = tsk_peer_port(tsk);
293 	u32 orig_node, peer_node;
294 
295 	if (unlikely(!tipc_sk_connected(sk)))
296 		return false;
297 
298 	if (unlikely(msg_origport(msg) != peer_port))
299 		return false;
300 
301 	orig_node = msg_orignode(msg);
302 	peer_node = tsk_peer_node(tsk);
303 
304 	if (likely(orig_node == peer_node))
305 		return true;
306 
307 	if (!orig_node && peer_node == self)
308 		return true;
309 
310 	if (!peer_node && orig_node == self)
311 		return true;
312 
313 	return false;
314 }
315 
316 /* tipc_set_sk_state - set the sk_state of the socket
317  * @sk: socket
318  *
319  * Caller must hold socket lock
320  *
321  * Returns 0 on success, errno otherwise
322  */
323 static int tipc_set_sk_state(struct sock *sk, int state)
324 {
325 	int oldsk_state = sk->sk_state;
326 	int res = -EINVAL;
327 
328 	switch (state) {
329 	case TIPC_OPEN:
330 		res = 0;
331 		break;
332 	case TIPC_LISTEN:
333 	case TIPC_CONNECTING:
334 		if (oldsk_state == TIPC_OPEN)
335 			res = 0;
336 		break;
337 	case TIPC_ESTABLISHED:
338 		if (oldsk_state == TIPC_CONNECTING ||
339 		    oldsk_state == TIPC_OPEN)
340 			res = 0;
341 		break;
342 	case TIPC_DISCONNECTING:
343 		if (oldsk_state == TIPC_CONNECTING ||
344 		    oldsk_state == TIPC_ESTABLISHED)
345 			res = 0;
346 		break;
347 	}
348 
349 	if (!res)
350 		sk->sk_state = state;
351 
352 	return res;
353 }
354 
355 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
356 {
357 	struct sock *sk = sock->sk;
358 	int err = sock_error(sk);
359 	int typ = sock->type;
360 
361 	if (err)
362 		return err;
363 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
364 		if (sk->sk_state == TIPC_DISCONNECTING)
365 			return -EPIPE;
366 		else if (!tipc_sk_connected(sk))
367 			return -ENOTCONN;
368 	}
369 	if (!*timeout)
370 		return -EAGAIN;
371 	if (signal_pending(current))
372 		return sock_intr_errno(*timeout);
373 
374 	return 0;
375 }
376 
377 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
378 ({                                                                             \
379 	struct sock *sk_;						       \
380 	int rc_;							       \
381 									       \
382 	while ((rc_ = !(condition_))) {					       \
383 		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
384 		sk_ = (sock_)->sk;					       \
385 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
386 		if (rc_)						       \
387 			break;						       \
388 		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
389 		release_sock(sk_);					       \
390 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
391 		sched_annotate_sleep();				               \
392 		lock_sock(sk_);						       \
393 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
394 	}								       \
395 	rc_;								       \
396 })
397 
398 /**
399  * tipc_sk_create - create a TIPC socket
400  * @net: network namespace (must be default network)
401  * @sock: pre-allocated socket structure
402  * @protocol: protocol indicator (must be 0)
403  * @kern: caused by kernel or by userspace?
404  *
405  * This routine creates additional data structures used by the TIPC socket,
406  * initializes them, and links them together.
407  *
408  * Returns 0 on success, errno otherwise
409  */
410 static int tipc_sk_create(struct net *net, struct socket *sock,
411 			  int protocol, int kern)
412 {
413 	const struct proto_ops *ops;
414 	struct sock *sk;
415 	struct tipc_sock *tsk;
416 	struct tipc_msg *msg;
417 
418 	/* Validate arguments */
419 	if (unlikely(protocol != 0))
420 		return -EPROTONOSUPPORT;
421 
422 	switch (sock->type) {
423 	case SOCK_STREAM:
424 		ops = &stream_ops;
425 		break;
426 	case SOCK_SEQPACKET:
427 		ops = &packet_ops;
428 		break;
429 	case SOCK_DGRAM:
430 	case SOCK_RDM:
431 		ops = &msg_ops;
432 		break;
433 	default:
434 		return -EPROTOTYPE;
435 	}
436 
437 	/* Allocate socket's protocol area */
438 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
439 	if (sk == NULL)
440 		return -ENOMEM;
441 
442 	tsk = tipc_sk(sk);
443 	tsk->max_pkt = MAX_PKT_DEFAULT;
444 	INIT_LIST_HEAD(&tsk->publications);
445 	INIT_LIST_HEAD(&tsk->cong_links);
446 	msg = &tsk->phdr;
447 
448 	/* Finish initializing socket data structures */
449 	sock->ops = ops;
450 	sock_init_data(sock, sk);
451 	tipc_set_sk_state(sk, TIPC_OPEN);
452 	if (tipc_sk_insert(tsk)) {
453 		pr_warn("Socket create failed; port number exhausted\n");
454 		return -EINVAL;
455 	}
456 
457 	/* Ensure tsk is visible before we read own_addr. */
458 	smp_mb();
459 
460 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
461 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
462 
463 	msg_set_origport(msg, tsk->portid);
464 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
465 	sk->sk_shutdown = 0;
466 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
467 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
468 	sk->sk_data_ready = tipc_data_ready;
469 	sk->sk_write_space = tipc_write_space;
470 	sk->sk_destruct = tipc_sock_destruct;
471 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
472 	tsk->group_is_open = true;
473 	atomic_set(&tsk->dupl_rcvcnt, 0);
474 
475 	/* Start out with safe limits until we receive an advertised window */
476 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
477 	tsk->rcv_win = tsk->snd_win;
478 
479 	if (tipc_sk_type_connectionless(sk)) {
480 		tsk_set_unreturnable(tsk, true);
481 		if (sock->type == SOCK_DGRAM)
482 			tsk_set_unreliable(tsk, true);
483 	}
484 
485 	return 0;
486 }
487 
488 static void tipc_sk_callback(struct rcu_head *head)
489 {
490 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
491 
492 	sock_put(&tsk->sk);
493 }
494 
495 /* Caller should hold socket lock for the socket. */
496 static void __tipc_shutdown(struct socket *sock, int error)
497 {
498 	struct sock *sk = sock->sk;
499 	struct tipc_sock *tsk = tipc_sk(sk);
500 	struct net *net = sock_net(sk);
501 	long timeout = CONN_TIMEOUT_DEFAULT;
502 	u32 dnode = tsk_peer_node(tsk);
503 	struct sk_buff *skb;
504 
505 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
506 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
507 					    !tsk_conn_cong(tsk)));
508 
509 	/* Remove any pending SYN message */
510 	__skb_queue_purge(&sk->sk_write_queue);
511 
512 	/* Reject all unreceived messages, except on an active connection
513 	 * (which disconnects locally & sends a 'FIN+' to peer).
514 	 */
515 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
516 		if (TIPC_SKB_CB(skb)->bytes_read) {
517 			kfree_skb(skb);
518 			continue;
519 		}
520 		if (!tipc_sk_type_connectionless(sk) &&
521 		    sk->sk_state != TIPC_DISCONNECTING) {
522 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
523 			tipc_node_remove_conn(net, dnode, tsk->portid);
524 		}
525 		tipc_sk_respond(sk, skb, error);
526 	}
527 
528 	if (tipc_sk_type_connectionless(sk))
529 		return;
530 
531 	if (sk->sk_state != TIPC_DISCONNECTING) {
532 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
533 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
534 				      tsk_own_node(tsk), tsk_peer_port(tsk),
535 				      tsk->portid, error);
536 		if (skb)
537 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
538 		tipc_node_remove_conn(net, dnode, tsk->portid);
539 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
540 	}
541 }
542 
543 /**
544  * tipc_release - destroy a TIPC socket
545  * @sock: socket to destroy
546  *
547  * This routine cleans up any messages that are still queued on the socket.
548  * For DGRAM and RDM socket types, all queued messages are rejected.
549  * For SEQPACKET and STREAM socket types, the first message is rejected
550  * and any others are discarded.  (If the first message on a STREAM socket
551  * is partially-read, it is discarded and the next one is rejected instead.)
552  *
553  * NOTE: Rejected messages are not necessarily returned to the sender!  They
554  * are returned or discarded according to the "destination droppable" setting
555  * specified for the message by the sender.
556  *
557  * Returns 0 on success, errno otherwise
558  */
559 static int tipc_release(struct socket *sock)
560 {
561 	struct sock *sk = sock->sk;
562 	struct tipc_sock *tsk;
563 
564 	/*
565 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
566 	 * releases a pre-allocated child socket that was never used)
567 	 */
568 	if (sk == NULL)
569 		return 0;
570 
571 	tsk = tipc_sk(sk);
572 	lock_sock(sk);
573 
574 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
575 	sk->sk_shutdown = SHUTDOWN_MASK;
576 	tipc_sk_leave(tsk);
577 	tipc_sk_withdraw(tsk, 0, NULL);
578 	sk_stop_timer(sk, &sk->sk_timer);
579 	tipc_sk_remove(tsk);
580 
581 	sock_orphan(sk);
582 	/* Reject any messages that accumulated in backlog queue */
583 	release_sock(sk);
584 	tipc_dest_list_purge(&tsk->cong_links);
585 	tsk->cong_link_cnt = 0;
586 	call_rcu(&tsk->rcu, tipc_sk_callback);
587 	sock->sk = NULL;
588 
589 	return 0;
590 }
591 
592 /**
593  * tipc_bind - associate or disassocate TIPC name(s) with a socket
594  * @sock: socket structure
595  * @uaddr: socket address describing name(s) and desired operation
596  * @uaddr_len: size of socket address data structure
597  *
598  * Name and name sequence binding is indicated using a positive scope value;
599  * a negative scope value unbinds the specified name.  Specifying no name
600  * (i.e. a socket address length of 0) unbinds all names from the socket.
601  *
602  * Returns 0 on success, errno otherwise
603  *
604  * NOTE: This routine doesn't need to take the socket lock since it doesn't
605  *       access any non-constant socket information.
606  */
607 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
608 		     int uaddr_len)
609 {
610 	struct sock *sk = sock->sk;
611 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
612 	struct tipc_sock *tsk = tipc_sk(sk);
613 	int res = -EINVAL;
614 
615 	lock_sock(sk);
616 	if (unlikely(!uaddr_len)) {
617 		res = tipc_sk_withdraw(tsk, 0, NULL);
618 		goto exit;
619 	}
620 	if (tsk->group) {
621 		res = -EACCES;
622 		goto exit;
623 	}
624 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
625 		res = -EINVAL;
626 		goto exit;
627 	}
628 	if (addr->family != AF_TIPC) {
629 		res = -EAFNOSUPPORT;
630 		goto exit;
631 	}
632 
633 	if (addr->addrtype == TIPC_ADDR_NAME)
634 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
635 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
636 		res = -EAFNOSUPPORT;
637 		goto exit;
638 	}
639 
640 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
641 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
642 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
643 		res = -EACCES;
644 		goto exit;
645 	}
646 
647 	res = (addr->scope >= 0) ?
648 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
649 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
650 exit:
651 	release_sock(sk);
652 	return res;
653 }
654 
655 /**
656  * tipc_getname - get port ID of socket or peer socket
657  * @sock: socket structure
658  * @uaddr: area for returned socket address
659  * @uaddr_len: area for returned length of socket address
660  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
661  *
662  * Returns 0 on success, errno otherwise
663  *
664  * NOTE: This routine doesn't need to take the socket lock since it only
665  *       accesses socket information that is unchanging (or which changes in
666  *       a completely predictable manner).
667  */
668 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
669 			int peer)
670 {
671 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
672 	struct sock *sk = sock->sk;
673 	struct tipc_sock *tsk = tipc_sk(sk);
674 
675 	memset(addr, 0, sizeof(*addr));
676 	if (peer) {
677 		if ((!tipc_sk_connected(sk)) &&
678 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
679 			return -ENOTCONN;
680 		addr->addr.id.ref = tsk_peer_port(tsk);
681 		addr->addr.id.node = tsk_peer_node(tsk);
682 	} else {
683 		addr->addr.id.ref = tsk->portid;
684 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
685 	}
686 
687 	addr->addrtype = TIPC_ADDR_ID;
688 	addr->family = AF_TIPC;
689 	addr->scope = 0;
690 	addr->addr.name.domain = 0;
691 
692 	return sizeof(*addr);
693 }
694 
695 /**
696  * tipc_poll - read and possibly block on pollmask
697  * @file: file structure associated with the socket
698  * @sock: socket for which to calculate the poll bits
699  * @wait: ???
700  *
701  * Returns pollmask value
702  *
703  * COMMENTARY:
704  * It appears that the usual socket locking mechanisms are not useful here
705  * since the pollmask info is potentially out-of-date the moment this routine
706  * exits.  TCP and other protocols seem to rely on higher level poll routines
707  * to handle any preventable race conditions, so TIPC will do the same ...
708  *
709  * IMPORTANT: The fact that a read or write operation is indicated does NOT
710  * imply that the operation will succeed, merely that it should be performed
711  * and will not block.
712  */
713 static __poll_t tipc_poll(struct file *file, struct socket *sock,
714 			      poll_table *wait)
715 {
716 	struct sock *sk = sock->sk;
717 	struct tipc_sock *tsk = tipc_sk(sk);
718 	__poll_t revents = 0;
719 
720 	sock_poll_wait(file, sock, wait);
721 
722 	if (sk->sk_shutdown & RCV_SHUTDOWN)
723 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
724 	if (sk->sk_shutdown == SHUTDOWN_MASK)
725 		revents |= EPOLLHUP;
726 
727 	switch (sk->sk_state) {
728 	case TIPC_ESTABLISHED:
729 	case TIPC_CONNECTING:
730 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
731 			revents |= EPOLLOUT;
732 		/* fall thru' */
733 	case TIPC_LISTEN:
734 		if (!skb_queue_empty(&sk->sk_receive_queue))
735 			revents |= EPOLLIN | EPOLLRDNORM;
736 		break;
737 	case TIPC_OPEN:
738 		if (tsk->group_is_open && !tsk->cong_link_cnt)
739 			revents |= EPOLLOUT;
740 		if (!tipc_sk_type_connectionless(sk))
741 			break;
742 		if (skb_queue_empty(&sk->sk_receive_queue))
743 			break;
744 		revents |= EPOLLIN | EPOLLRDNORM;
745 		break;
746 	case TIPC_DISCONNECTING:
747 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
748 		break;
749 	}
750 	return revents;
751 }
752 
753 /**
754  * tipc_sendmcast - send multicast message
755  * @sock: socket structure
756  * @seq: destination address
757  * @msg: message to send
758  * @dlen: length of data to send
759  * @timeout: timeout to wait for wakeup
760  *
761  * Called from function tipc_sendmsg(), which has done all sanity checks
762  * Returns the number of bytes sent on success, or errno
763  */
764 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
765 			  struct msghdr *msg, size_t dlen, long timeout)
766 {
767 	struct sock *sk = sock->sk;
768 	struct tipc_sock *tsk = tipc_sk(sk);
769 	struct tipc_msg *hdr = &tsk->phdr;
770 	struct net *net = sock_net(sk);
771 	int mtu = tipc_bcast_get_mtu(net);
772 	struct tipc_mc_method *method = &tsk->mc_method;
773 	struct sk_buff_head pkts;
774 	struct tipc_nlist dsts;
775 	int rc;
776 
777 	if (tsk->group)
778 		return -EACCES;
779 
780 	/* Block or return if any destination link is congested */
781 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
782 	if (unlikely(rc))
783 		return rc;
784 
785 	/* Lookup destination nodes */
786 	tipc_nlist_init(&dsts, tipc_own_addr(net));
787 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
788 				      seq->upper, &dsts);
789 	if (!dsts.local && !dsts.remote)
790 		return -EHOSTUNREACH;
791 
792 	/* Build message header */
793 	msg_set_type(hdr, TIPC_MCAST_MSG);
794 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
795 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
796 	msg_set_destport(hdr, 0);
797 	msg_set_destnode(hdr, 0);
798 	msg_set_nametype(hdr, seq->type);
799 	msg_set_namelower(hdr, seq->lower);
800 	msg_set_nameupper(hdr, seq->upper);
801 
802 	/* Build message as chain of buffers */
803 	skb_queue_head_init(&pkts);
804 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
805 
806 	/* Send message if build was successful */
807 	if (unlikely(rc == dlen))
808 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
809 				     &tsk->cong_link_cnt);
810 
811 	tipc_nlist_purge(&dsts);
812 
813 	return rc ? rc : dlen;
814 }
815 
816 /**
817  * tipc_send_group_msg - send a message to a member in the group
818  * @net: network namespace
819  * @m: message to send
820  * @mb: group member
821  * @dnode: destination node
822  * @dport: destination port
823  * @dlen: total length of message data
824  */
825 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
826 			       struct msghdr *m, struct tipc_member *mb,
827 			       u32 dnode, u32 dport, int dlen)
828 {
829 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
830 	struct tipc_mc_method *method = &tsk->mc_method;
831 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
832 	struct tipc_msg *hdr = &tsk->phdr;
833 	struct sk_buff_head pkts;
834 	int mtu, rc;
835 
836 	/* Complete message header */
837 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
838 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
839 	msg_set_destport(hdr, dport);
840 	msg_set_destnode(hdr, dnode);
841 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
842 
843 	/* Build message as chain of buffers */
844 	skb_queue_head_init(&pkts);
845 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
846 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
847 	if (unlikely(rc != dlen))
848 		return rc;
849 
850 	/* Send message */
851 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
852 	if (unlikely(rc == -ELINKCONG)) {
853 		tipc_dest_push(&tsk->cong_links, dnode, 0);
854 		tsk->cong_link_cnt++;
855 	}
856 
857 	/* Update send window */
858 	tipc_group_update_member(mb, blks);
859 
860 	/* A broadcast sent within next EXPIRE period must follow same path */
861 	method->rcast = true;
862 	method->mandatory = true;
863 	return dlen;
864 }
865 
866 /**
867  * tipc_send_group_unicast - send message to a member in the group
868  * @sock: socket structure
869  * @m: message to send
870  * @dlen: total length of message data
871  * @timeout: timeout to wait for wakeup
872  *
873  * Called from function tipc_sendmsg(), which has done all sanity checks
874  * Returns the number of bytes sent on success, or errno
875  */
876 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
877 				   int dlen, long timeout)
878 {
879 	struct sock *sk = sock->sk;
880 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
881 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
882 	struct tipc_sock *tsk = tipc_sk(sk);
883 	struct tipc_group *grp = tsk->group;
884 	struct net *net = sock_net(sk);
885 	struct tipc_member *mb = NULL;
886 	u32 node, port;
887 	int rc;
888 
889 	node = dest->addr.id.node;
890 	port = dest->addr.id.ref;
891 	if (!port && !node)
892 		return -EHOSTUNREACH;
893 
894 	/* Block or return if destination link or member is congested */
895 	rc = tipc_wait_for_cond(sock, &timeout,
896 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
897 				!tipc_group_cong(grp, node, port, blks, &mb));
898 	if (unlikely(rc))
899 		return rc;
900 
901 	if (unlikely(!mb))
902 		return -EHOSTUNREACH;
903 
904 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
905 
906 	return rc ? rc : dlen;
907 }
908 
909 /**
910  * tipc_send_group_anycast - send message to any member with given identity
911  * @sock: socket structure
912  * @m: message to send
913  * @dlen: total length of message data
914  * @timeout: timeout to wait for wakeup
915  *
916  * Called from function tipc_sendmsg(), which has done all sanity checks
917  * Returns the number of bytes sent on success, or errno
918  */
919 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
920 				   int dlen, long timeout)
921 {
922 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
923 	struct sock *sk = sock->sk;
924 	struct tipc_sock *tsk = tipc_sk(sk);
925 	struct list_head *cong_links = &tsk->cong_links;
926 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
927 	struct tipc_group *grp = tsk->group;
928 	struct tipc_msg *hdr = &tsk->phdr;
929 	struct tipc_member *first = NULL;
930 	struct tipc_member *mbr = NULL;
931 	struct net *net = sock_net(sk);
932 	u32 node, port, exclude;
933 	struct list_head dsts;
934 	u32 type, inst, scope;
935 	int lookups = 0;
936 	int dstcnt, rc;
937 	bool cong;
938 
939 	INIT_LIST_HEAD(&dsts);
940 
941 	type = msg_nametype(hdr);
942 	inst = dest->addr.name.name.instance;
943 	scope = msg_lookup_scope(hdr);
944 	exclude = tipc_group_exclude(grp);
945 
946 	while (++lookups < 4) {
947 		first = NULL;
948 
949 		/* Look for a non-congested destination member, if any */
950 		while (1) {
951 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
952 						 &dstcnt, exclude, false))
953 				return -EHOSTUNREACH;
954 			tipc_dest_pop(&dsts, &node, &port);
955 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
956 			if (!cong)
957 				break;
958 			if (mbr == first)
959 				break;
960 			if (!first)
961 				first = mbr;
962 		}
963 
964 		/* Start over if destination was not in member list */
965 		if (unlikely(!mbr))
966 			continue;
967 
968 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
969 			break;
970 
971 		/* Block or return if destination link or member is congested */
972 		rc = tipc_wait_for_cond(sock, &timeout,
973 					!tipc_dest_find(cong_links, node, 0) &&
974 					!tipc_group_cong(grp, node, port,
975 							 blks, &mbr));
976 		if (unlikely(rc))
977 			return rc;
978 
979 		/* Send, unless destination disappeared while waiting */
980 		if (likely(mbr))
981 			break;
982 	}
983 
984 	if (unlikely(lookups >= 4))
985 		return -EHOSTUNREACH;
986 
987 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
988 
989 	return rc ? rc : dlen;
990 }
991 
992 /**
993  * tipc_send_group_bcast - send message to all members in communication group
994  * @sk: socket structure
995  * @m: message to send
996  * @dlen: total length of message data
997  * @timeout: timeout to wait for wakeup
998  *
999  * Called from function tipc_sendmsg(), which has done all sanity checks
1000  * Returns the number of bytes sent on success, or errno
1001  */
1002 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1003 				 int dlen, long timeout)
1004 {
1005 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1006 	struct sock *sk = sock->sk;
1007 	struct net *net = sock_net(sk);
1008 	struct tipc_sock *tsk = tipc_sk(sk);
1009 	struct tipc_group *grp = tsk->group;
1010 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1011 	struct tipc_mc_method *method = &tsk->mc_method;
1012 	bool ack = method->mandatory && method->rcast;
1013 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1014 	struct tipc_msg *hdr = &tsk->phdr;
1015 	int mtu = tipc_bcast_get_mtu(net);
1016 	struct sk_buff_head pkts;
1017 	int rc = -EHOSTUNREACH;
1018 
1019 	if (!dsts->local && !dsts->remote)
1020 		return -EHOSTUNREACH;
1021 
1022 	/* Block or return if any destination link or member is congested */
1023 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1024 				!tipc_group_bc_cong(grp, blks));
1025 	if (unlikely(rc))
1026 		return rc;
1027 
1028 	/* Complete message header */
1029 	if (dest) {
1030 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1031 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1032 	} else {
1033 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1034 		msg_set_nameinst(hdr, 0);
1035 	}
1036 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1037 	msg_set_destport(hdr, 0);
1038 	msg_set_destnode(hdr, 0);
1039 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1040 
1041 	/* Avoid getting stuck with repeated forced replicasts */
1042 	msg_set_grp_bc_ack_req(hdr, ack);
1043 
1044 	/* Build message as chain of buffers */
1045 	skb_queue_head_init(&pkts);
1046 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1047 	if (unlikely(rc != dlen))
1048 		return rc;
1049 
1050 	/* Send message */
1051 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1052 	if (unlikely(rc))
1053 		return rc;
1054 
1055 	/* Update broadcast sequence number and send windows */
1056 	tipc_group_update_bc_members(tsk->group, blks, ack);
1057 
1058 	/* Broadcast link is now free to choose method for next broadcast */
1059 	method->mandatory = false;
1060 	method->expires = jiffies;
1061 
1062 	return dlen;
1063 }
1064 
1065 /**
1066  * tipc_send_group_mcast - send message to all members with given identity
1067  * @sock: socket structure
1068  * @m: message to send
1069  * @dlen: total length of message data
1070  * @timeout: timeout to wait for wakeup
1071  *
1072  * Called from function tipc_sendmsg(), which has done all sanity checks
1073  * Returns the number of bytes sent on success, or errno
1074  */
1075 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1076 				 int dlen, long timeout)
1077 {
1078 	struct sock *sk = sock->sk;
1079 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1080 	struct tipc_sock *tsk = tipc_sk(sk);
1081 	struct tipc_group *grp = tsk->group;
1082 	struct tipc_msg *hdr = &tsk->phdr;
1083 	struct net *net = sock_net(sk);
1084 	u32 type, inst, scope, exclude;
1085 	struct list_head dsts;
1086 	u32 dstcnt;
1087 
1088 	INIT_LIST_HEAD(&dsts);
1089 
1090 	type = msg_nametype(hdr);
1091 	inst = dest->addr.name.name.instance;
1092 	scope = msg_lookup_scope(hdr);
1093 	exclude = tipc_group_exclude(grp);
1094 
1095 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1096 				 &dstcnt, exclude, true))
1097 		return -EHOSTUNREACH;
1098 
1099 	if (dstcnt == 1) {
1100 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1101 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1102 	}
1103 
1104 	tipc_dest_list_purge(&dsts);
1105 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1106 }
1107 
1108 /**
1109  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1110  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1111  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1112  *
1113  * Multi-threaded: parallel calls with reference to same queues may occur
1114  */
1115 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1116 		       struct sk_buff_head *inputq)
1117 {
1118 	u32 self = tipc_own_addr(net);
1119 	u32 type, lower, upper, scope;
1120 	struct sk_buff *skb, *_skb;
1121 	u32 portid, onode;
1122 	struct sk_buff_head tmpq;
1123 	struct list_head dports;
1124 	struct tipc_msg *hdr;
1125 	int user, mtyp, hlen;
1126 	bool exact;
1127 
1128 	__skb_queue_head_init(&tmpq);
1129 	INIT_LIST_HEAD(&dports);
1130 
1131 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1132 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1133 		hdr = buf_msg(skb);
1134 		user = msg_user(hdr);
1135 		mtyp = msg_type(hdr);
1136 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1137 		onode = msg_orignode(hdr);
1138 		type = msg_nametype(hdr);
1139 
1140 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1141 			spin_lock_bh(&inputq->lock);
1142 			if (skb_peek(arrvq) == skb) {
1143 				__skb_dequeue(arrvq);
1144 				__skb_queue_tail(inputq, skb);
1145 			}
1146 			kfree_skb(skb);
1147 			spin_unlock_bh(&inputq->lock);
1148 			continue;
1149 		}
1150 
1151 		/* Group messages require exact scope match */
1152 		if (msg_in_group(hdr)) {
1153 			lower = 0;
1154 			upper = ~0;
1155 			scope = msg_lookup_scope(hdr);
1156 			exact = true;
1157 		} else {
1158 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1159 			if (onode == self)
1160 				scope = TIPC_NODE_SCOPE;
1161 			else
1162 				scope = TIPC_CLUSTER_SCOPE;
1163 			exact = false;
1164 			lower = msg_namelower(hdr);
1165 			upper = msg_nameupper(hdr);
1166 		}
1167 
1168 		/* Create destination port list: */
1169 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1170 				       scope, exact, &dports);
1171 
1172 		/* Clone message per destination */
1173 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1174 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1175 			if (_skb) {
1176 				msg_set_destport(buf_msg(_skb), portid);
1177 				__skb_queue_tail(&tmpq, _skb);
1178 				continue;
1179 			}
1180 			pr_warn("Failed to clone mcast rcv buffer\n");
1181 		}
1182 		/* Append to inputq if not already done by other thread */
1183 		spin_lock_bh(&inputq->lock);
1184 		if (skb_peek(arrvq) == skb) {
1185 			skb_queue_splice_tail_init(&tmpq, inputq);
1186 			kfree_skb(__skb_dequeue(arrvq));
1187 		}
1188 		spin_unlock_bh(&inputq->lock);
1189 		__skb_queue_purge(&tmpq);
1190 		kfree_skb(skb);
1191 	}
1192 	tipc_sk_rcv(net, inputq);
1193 }
1194 
1195 /**
1196  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1197  * @tsk: receiving socket
1198  * @skb: pointer to message buffer.
1199  */
1200 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1201 				   struct sk_buff_head *inputq,
1202 				   struct sk_buff_head *xmitq)
1203 {
1204 	struct tipc_msg *hdr = buf_msg(skb);
1205 	u32 onode = tsk_own_node(tsk);
1206 	struct sock *sk = &tsk->sk;
1207 	int mtyp = msg_type(hdr);
1208 	bool conn_cong;
1209 
1210 	/* Ignore if connection cannot be validated: */
1211 	if (!tsk_peer_msg(tsk, hdr))
1212 		goto exit;
1213 
1214 	if (unlikely(msg_errcode(hdr))) {
1215 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1216 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1217 				      tsk_peer_port(tsk));
1218 		sk->sk_state_change(sk);
1219 
1220 		/* State change is ignored if socket already awake,
1221 		 * - convert msg to abort msg and add to inqueue
1222 		 */
1223 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1224 		msg_set_type(hdr, TIPC_CONN_MSG);
1225 		msg_set_size(hdr, BASIC_H_SIZE);
1226 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1227 		__skb_queue_tail(inputq, skb);
1228 		return;
1229 	}
1230 
1231 	tsk->probe_unacked = false;
1232 
1233 	if (mtyp == CONN_PROBE) {
1234 		msg_set_type(hdr, CONN_PROBE_REPLY);
1235 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1236 			__skb_queue_tail(xmitq, skb);
1237 		return;
1238 	} else if (mtyp == CONN_ACK) {
1239 		conn_cong = tsk_conn_cong(tsk);
1240 		tsk->snt_unacked -= msg_conn_ack(hdr);
1241 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1242 			tsk->snd_win = msg_adv_win(hdr);
1243 		if (conn_cong)
1244 			sk->sk_write_space(sk);
1245 	} else if (mtyp != CONN_PROBE_REPLY) {
1246 		pr_warn("Received unknown CONN_PROTO msg\n");
1247 	}
1248 exit:
1249 	kfree_skb(skb);
1250 }
1251 
1252 /**
1253  * tipc_sendmsg - send message in connectionless manner
1254  * @sock: socket structure
1255  * @m: message to send
1256  * @dsz: amount of user data to be sent
1257  *
1258  * Message must have an destination specified explicitly.
1259  * Used for SOCK_RDM and SOCK_DGRAM messages,
1260  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1261  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1262  *
1263  * Returns the number of bytes sent on success, or errno otherwise
1264  */
1265 static int tipc_sendmsg(struct socket *sock,
1266 			struct msghdr *m, size_t dsz)
1267 {
1268 	struct sock *sk = sock->sk;
1269 	int ret;
1270 
1271 	lock_sock(sk);
1272 	ret = __tipc_sendmsg(sock, m, dsz);
1273 	release_sock(sk);
1274 
1275 	return ret;
1276 }
1277 
1278 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1279 {
1280 	struct sock *sk = sock->sk;
1281 	struct net *net = sock_net(sk);
1282 	struct tipc_sock *tsk = tipc_sk(sk);
1283 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1284 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1285 	struct list_head *clinks = &tsk->cong_links;
1286 	bool syn = !tipc_sk_type_connectionless(sk);
1287 	struct tipc_group *grp = tsk->group;
1288 	struct tipc_msg *hdr = &tsk->phdr;
1289 	struct tipc_name_seq *seq;
1290 	struct sk_buff_head pkts;
1291 	u32 dport, dnode = 0;
1292 	u32 type, inst;
1293 	int mtu, rc;
1294 
1295 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1296 		return -EMSGSIZE;
1297 
1298 	if (likely(dest)) {
1299 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1300 			return -EINVAL;
1301 		if (unlikely(dest->family != AF_TIPC))
1302 			return -EINVAL;
1303 	}
1304 
1305 	if (grp) {
1306 		if (!dest)
1307 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1308 		if (dest->addrtype == TIPC_ADDR_NAME)
1309 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1310 		if (dest->addrtype == TIPC_ADDR_ID)
1311 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1312 		if (dest->addrtype == TIPC_ADDR_MCAST)
1313 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1314 		return -EINVAL;
1315 	}
1316 
1317 	if (unlikely(!dest)) {
1318 		dest = &tsk->peer;
1319 		if (!syn || dest->family != AF_TIPC)
1320 			return -EDESTADDRREQ;
1321 	}
1322 
1323 	if (unlikely(syn)) {
1324 		if (sk->sk_state == TIPC_LISTEN)
1325 			return -EPIPE;
1326 		if (sk->sk_state != TIPC_OPEN)
1327 			return -EISCONN;
1328 		if (tsk->published)
1329 			return -EOPNOTSUPP;
1330 		if (dest->addrtype == TIPC_ADDR_NAME) {
1331 			tsk->conn_type = dest->addr.name.name.type;
1332 			tsk->conn_instance = dest->addr.name.name.instance;
1333 		}
1334 		msg_set_syn(hdr, 1);
1335 	}
1336 
1337 	seq = &dest->addr.nameseq;
1338 	if (dest->addrtype == TIPC_ADDR_MCAST)
1339 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1340 
1341 	if (dest->addrtype == TIPC_ADDR_NAME) {
1342 		type = dest->addr.name.name.type;
1343 		inst = dest->addr.name.name.instance;
1344 		dnode = dest->addr.name.domain;
1345 		msg_set_type(hdr, TIPC_NAMED_MSG);
1346 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1347 		msg_set_nametype(hdr, type);
1348 		msg_set_nameinst(hdr, inst);
1349 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1350 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1351 		msg_set_destnode(hdr, dnode);
1352 		msg_set_destport(hdr, dport);
1353 		if (unlikely(!dport && !dnode))
1354 			return -EHOSTUNREACH;
1355 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1356 		dnode = dest->addr.id.node;
1357 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1358 		msg_set_lookup_scope(hdr, 0);
1359 		msg_set_destnode(hdr, dnode);
1360 		msg_set_destport(hdr, dest->addr.id.ref);
1361 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1362 	} else {
1363 		return -EINVAL;
1364 	}
1365 
1366 	/* Block or return if destination link is congested */
1367 	rc = tipc_wait_for_cond(sock, &timeout,
1368 				!tipc_dest_find(clinks, dnode, 0));
1369 	if (unlikely(rc))
1370 		return rc;
1371 
1372 	skb_queue_head_init(&pkts);
1373 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1374 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1375 	if (unlikely(rc != dlen))
1376 		return rc;
1377 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1378 		return -ENOMEM;
1379 
1380 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1381 	if (unlikely(rc == -ELINKCONG)) {
1382 		tipc_dest_push(clinks, dnode, 0);
1383 		tsk->cong_link_cnt++;
1384 		rc = 0;
1385 	}
1386 
1387 	if (unlikely(syn && !rc))
1388 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1389 
1390 	return rc ? rc : dlen;
1391 }
1392 
1393 /**
1394  * tipc_sendstream - send stream-oriented data
1395  * @sock: socket structure
1396  * @m: data to send
1397  * @dsz: total length of data to be transmitted
1398  *
1399  * Used for SOCK_STREAM data.
1400  *
1401  * Returns the number of bytes sent on success (or partial success),
1402  * or errno if no data sent
1403  */
1404 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1405 {
1406 	struct sock *sk = sock->sk;
1407 	int ret;
1408 
1409 	lock_sock(sk);
1410 	ret = __tipc_sendstream(sock, m, dsz);
1411 	release_sock(sk);
1412 
1413 	return ret;
1414 }
1415 
1416 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1417 {
1418 	struct sock *sk = sock->sk;
1419 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1420 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1421 	struct tipc_sock *tsk = tipc_sk(sk);
1422 	struct tipc_msg *hdr = &tsk->phdr;
1423 	struct net *net = sock_net(sk);
1424 	struct sk_buff_head pkts;
1425 	u32 dnode = tsk_peer_node(tsk);
1426 	int send, sent = 0;
1427 	int rc = 0;
1428 
1429 	skb_queue_head_init(&pkts);
1430 
1431 	if (unlikely(dlen > INT_MAX))
1432 		return -EMSGSIZE;
1433 
1434 	/* Handle implicit connection setup */
1435 	if (unlikely(dest)) {
1436 		rc = __tipc_sendmsg(sock, m, dlen);
1437 		if (dlen && dlen == rc) {
1438 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1439 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1440 		}
1441 		return rc;
1442 	}
1443 
1444 	do {
1445 		rc = tipc_wait_for_cond(sock, &timeout,
1446 					(!tsk->cong_link_cnt &&
1447 					 !tsk_conn_cong(tsk) &&
1448 					 tipc_sk_connected(sk)));
1449 		if (unlikely(rc))
1450 			break;
1451 
1452 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1453 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1454 		if (unlikely(rc != send))
1455 			break;
1456 
1457 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1458 		if (unlikely(rc == -ELINKCONG)) {
1459 			tsk->cong_link_cnt = 1;
1460 			rc = 0;
1461 		}
1462 		if (likely(!rc)) {
1463 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1464 			sent += send;
1465 		}
1466 	} while (sent < dlen && !rc);
1467 
1468 	return sent ? sent : rc;
1469 }
1470 
1471 /**
1472  * tipc_send_packet - send a connection-oriented message
1473  * @sock: socket structure
1474  * @m: message to send
1475  * @dsz: length of data to be transmitted
1476  *
1477  * Used for SOCK_SEQPACKET messages.
1478  *
1479  * Returns the number of bytes sent on success, or errno otherwise
1480  */
1481 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1482 {
1483 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1484 		return -EMSGSIZE;
1485 
1486 	return tipc_sendstream(sock, m, dsz);
1487 }
1488 
1489 /* tipc_sk_finish_conn - complete the setup of a connection
1490  */
1491 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1492 				u32 peer_node)
1493 {
1494 	struct sock *sk = &tsk->sk;
1495 	struct net *net = sock_net(sk);
1496 	struct tipc_msg *msg = &tsk->phdr;
1497 
1498 	msg_set_syn(msg, 0);
1499 	msg_set_destnode(msg, peer_node);
1500 	msg_set_destport(msg, peer_port);
1501 	msg_set_type(msg, TIPC_CONN_MSG);
1502 	msg_set_lookup_scope(msg, 0);
1503 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1504 
1505 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1506 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1507 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1508 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1509 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1510 	__skb_queue_purge(&sk->sk_write_queue);
1511 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1512 		return;
1513 
1514 	/* Fall back to message based flow control */
1515 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1516 	tsk->snd_win = FLOWCTL_MSG_WIN;
1517 }
1518 
1519 /**
1520  * tipc_sk_set_orig_addr - capture sender's address for received message
1521  * @m: descriptor for message info
1522  * @hdr: received message header
1523  *
1524  * Note: Address is not captured if not requested by receiver.
1525  */
1526 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1527 {
1528 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1529 	struct tipc_msg *hdr = buf_msg(skb);
1530 
1531 	if (!srcaddr)
1532 		return;
1533 
1534 	srcaddr->sock.family = AF_TIPC;
1535 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1536 	srcaddr->sock.scope = 0;
1537 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1538 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1539 	srcaddr->sock.addr.name.domain = 0;
1540 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1541 
1542 	if (!msg_in_group(hdr))
1543 		return;
1544 
1545 	/* Group message users may also want to know sending member's id */
1546 	srcaddr->member.family = AF_TIPC;
1547 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1548 	srcaddr->member.scope = 0;
1549 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1550 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1551 	srcaddr->member.addr.name.domain = 0;
1552 	m->msg_namelen = sizeof(*srcaddr);
1553 }
1554 
1555 /**
1556  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1557  * @m: descriptor for message info
1558  * @skb: received message buffer
1559  * @tsk: TIPC port associated with message
1560  *
1561  * Note: Ancillary data is not captured if not requested by receiver.
1562  *
1563  * Returns 0 if successful, otherwise errno
1564  */
1565 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1566 				 struct tipc_sock *tsk)
1567 {
1568 	struct tipc_msg *msg;
1569 	u32 anc_data[3];
1570 	u32 err;
1571 	u32 dest_type;
1572 	int has_name;
1573 	int res;
1574 
1575 	if (likely(m->msg_controllen == 0))
1576 		return 0;
1577 	msg = buf_msg(skb);
1578 
1579 	/* Optionally capture errored message object(s) */
1580 	err = msg ? msg_errcode(msg) : 0;
1581 	if (unlikely(err)) {
1582 		anc_data[0] = err;
1583 		anc_data[1] = msg_data_sz(msg);
1584 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1585 		if (res)
1586 			return res;
1587 		if (anc_data[1]) {
1588 			if (skb_linearize(skb))
1589 				return -ENOMEM;
1590 			msg = buf_msg(skb);
1591 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1592 				       msg_data(msg));
1593 			if (res)
1594 				return res;
1595 		}
1596 	}
1597 
1598 	/* Optionally capture message destination object */
1599 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1600 	switch (dest_type) {
1601 	case TIPC_NAMED_MSG:
1602 		has_name = 1;
1603 		anc_data[0] = msg_nametype(msg);
1604 		anc_data[1] = msg_namelower(msg);
1605 		anc_data[2] = msg_namelower(msg);
1606 		break;
1607 	case TIPC_MCAST_MSG:
1608 		has_name = 1;
1609 		anc_data[0] = msg_nametype(msg);
1610 		anc_data[1] = msg_namelower(msg);
1611 		anc_data[2] = msg_nameupper(msg);
1612 		break;
1613 	case TIPC_CONN_MSG:
1614 		has_name = (tsk->conn_type != 0);
1615 		anc_data[0] = tsk->conn_type;
1616 		anc_data[1] = tsk->conn_instance;
1617 		anc_data[2] = tsk->conn_instance;
1618 		break;
1619 	default:
1620 		has_name = 0;
1621 	}
1622 	if (has_name) {
1623 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1624 		if (res)
1625 			return res;
1626 	}
1627 
1628 	return 0;
1629 }
1630 
1631 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1632 {
1633 	struct sock *sk = &tsk->sk;
1634 	struct net *net = sock_net(sk);
1635 	struct sk_buff *skb = NULL;
1636 	struct tipc_msg *msg;
1637 	u32 peer_port = tsk_peer_port(tsk);
1638 	u32 dnode = tsk_peer_node(tsk);
1639 
1640 	if (!tipc_sk_connected(sk))
1641 		return;
1642 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1643 			      dnode, tsk_own_node(tsk), peer_port,
1644 			      tsk->portid, TIPC_OK);
1645 	if (!skb)
1646 		return;
1647 	msg = buf_msg(skb);
1648 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1649 	tsk->rcv_unacked = 0;
1650 
1651 	/* Adjust to and advertize the correct window limit */
1652 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1653 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1654 		msg_set_adv_win(msg, tsk->rcv_win);
1655 	}
1656 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1657 }
1658 
1659 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1660 {
1661 	struct sock *sk = sock->sk;
1662 	DEFINE_WAIT(wait);
1663 	long timeo = *timeop;
1664 	int err = sock_error(sk);
1665 
1666 	if (err)
1667 		return err;
1668 
1669 	for (;;) {
1670 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1671 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1672 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1673 				err = -ENOTCONN;
1674 				break;
1675 			}
1676 			release_sock(sk);
1677 			timeo = schedule_timeout(timeo);
1678 			lock_sock(sk);
1679 		}
1680 		err = 0;
1681 		if (!skb_queue_empty(&sk->sk_receive_queue))
1682 			break;
1683 		err = -EAGAIN;
1684 		if (!timeo)
1685 			break;
1686 		err = sock_intr_errno(timeo);
1687 		if (signal_pending(current))
1688 			break;
1689 
1690 		err = sock_error(sk);
1691 		if (err)
1692 			break;
1693 	}
1694 	finish_wait(sk_sleep(sk), &wait);
1695 	*timeop = timeo;
1696 	return err;
1697 }
1698 
1699 /**
1700  * tipc_recvmsg - receive packet-oriented message
1701  * @m: descriptor for message info
1702  * @buflen: length of user buffer area
1703  * @flags: receive flags
1704  *
1705  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1706  * If the complete message doesn't fit in user area, truncate it.
1707  *
1708  * Returns size of returned message data, errno otherwise
1709  */
1710 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1711 			size_t buflen,	int flags)
1712 {
1713 	struct sock *sk = sock->sk;
1714 	bool connected = !tipc_sk_type_connectionless(sk);
1715 	struct tipc_sock *tsk = tipc_sk(sk);
1716 	int rc, err, hlen, dlen, copy;
1717 	struct sk_buff_head xmitq;
1718 	struct tipc_msg *hdr;
1719 	struct sk_buff *skb;
1720 	bool grp_evt;
1721 	long timeout;
1722 
1723 	/* Catch invalid receive requests */
1724 	if (unlikely(!buflen))
1725 		return -EINVAL;
1726 
1727 	lock_sock(sk);
1728 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1729 		rc = -ENOTCONN;
1730 		goto exit;
1731 	}
1732 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1733 
1734 	/* Step rcv queue to first msg with data or error; wait if necessary */
1735 	do {
1736 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1737 		if (unlikely(rc))
1738 			goto exit;
1739 		skb = skb_peek(&sk->sk_receive_queue);
1740 		hdr = buf_msg(skb);
1741 		dlen = msg_data_sz(hdr);
1742 		hlen = msg_hdr_sz(hdr);
1743 		err = msg_errcode(hdr);
1744 		grp_evt = msg_is_grp_evt(hdr);
1745 		if (likely(dlen || err))
1746 			break;
1747 		tsk_advance_rx_queue(sk);
1748 	} while (1);
1749 
1750 	/* Collect msg meta data, including error code and rejected data */
1751 	tipc_sk_set_orig_addr(m, skb);
1752 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1753 	if (unlikely(rc))
1754 		goto exit;
1755 	hdr = buf_msg(skb);
1756 
1757 	/* Capture data if non-error msg, otherwise just set return value */
1758 	if (likely(!err)) {
1759 		copy = min_t(int, dlen, buflen);
1760 		if (unlikely(copy != dlen))
1761 			m->msg_flags |= MSG_TRUNC;
1762 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1763 	} else {
1764 		copy = 0;
1765 		rc = 0;
1766 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1767 			rc = -ECONNRESET;
1768 	}
1769 	if (unlikely(rc))
1770 		goto exit;
1771 
1772 	/* Mark message as group event if applicable */
1773 	if (unlikely(grp_evt)) {
1774 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1775 			m->msg_flags |= MSG_EOR;
1776 		m->msg_flags |= MSG_OOB;
1777 		copy = 0;
1778 	}
1779 
1780 	/* Caption of data or error code/rejected data was successful */
1781 	if (unlikely(flags & MSG_PEEK))
1782 		goto exit;
1783 
1784 	/* Send group flow control advertisement when applicable */
1785 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1786 		skb_queue_head_init(&xmitq);
1787 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1788 					  msg_orignode(hdr), msg_origport(hdr),
1789 					  &xmitq);
1790 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1791 	}
1792 
1793 	tsk_advance_rx_queue(sk);
1794 
1795 	if (likely(!connected))
1796 		goto exit;
1797 
1798 	/* Send connection flow control advertisement when applicable */
1799 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1800 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1801 		tipc_sk_send_ack(tsk);
1802 exit:
1803 	release_sock(sk);
1804 	return rc ? rc : copy;
1805 }
1806 
1807 /**
1808  * tipc_recvstream - receive stream-oriented data
1809  * @m: descriptor for message info
1810  * @buflen: total size of user buffer area
1811  * @flags: receive flags
1812  *
1813  * Used for SOCK_STREAM messages only.  If not enough data is available
1814  * will optionally wait for more; never truncates data.
1815  *
1816  * Returns size of returned message data, errno otherwise
1817  */
1818 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1819 			   size_t buflen, int flags)
1820 {
1821 	struct sock *sk = sock->sk;
1822 	struct tipc_sock *tsk = tipc_sk(sk);
1823 	struct sk_buff *skb;
1824 	struct tipc_msg *hdr;
1825 	struct tipc_skb_cb *skb_cb;
1826 	bool peek = flags & MSG_PEEK;
1827 	int offset, required, copy, copied = 0;
1828 	int hlen, dlen, err, rc;
1829 	long timeout;
1830 
1831 	/* Catch invalid receive attempts */
1832 	if (unlikely(!buflen))
1833 		return -EINVAL;
1834 
1835 	lock_sock(sk);
1836 
1837 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1838 		rc = -ENOTCONN;
1839 		goto exit;
1840 	}
1841 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1842 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1843 
1844 	do {
1845 		/* Look at first msg in receive queue; wait if necessary */
1846 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1847 		if (unlikely(rc))
1848 			break;
1849 		skb = skb_peek(&sk->sk_receive_queue);
1850 		skb_cb = TIPC_SKB_CB(skb);
1851 		hdr = buf_msg(skb);
1852 		dlen = msg_data_sz(hdr);
1853 		hlen = msg_hdr_sz(hdr);
1854 		err = msg_errcode(hdr);
1855 
1856 		/* Discard any empty non-errored (SYN-) message */
1857 		if (unlikely(!dlen && !err)) {
1858 			tsk_advance_rx_queue(sk);
1859 			continue;
1860 		}
1861 
1862 		/* Collect msg meta data, incl. error code and rejected data */
1863 		if (!copied) {
1864 			tipc_sk_set_orig_addr(m, skb);
1865 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
1866 			if (rc)
1867 				break;
1868 			hdr = buf_msg(skb);
1869 		}
1870 
1871 		/* Copy data if msg ok, otherwise return error/partial data */
1872 		if (likely(!err)) {
1873 			offset = skb_cb->bytes_read;
1874 			copy = min_t(int, dlen - offset, buflen - copied);
1875 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1876 			if (unlikely(rc))
1877 				break;
1878 			copied += copy;
1879 			offset += copy;
1880 			if (unlikely(offset < dlen)) {
1881 				if (!peek)
1882 					skb_cb->bytes_read = offset;
1883 				break;
1884 			}
1885 		} else {
1886 			rc = 0;
1887 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1888 				rc = -ECONNRESET;
1889 			if (copied || rc)
1890 				break;
1891 		}
1892 
1893 		if (unlikely(peek))
1894 			break;
1895 
1896 		tsk_advance_rx_queue(sk);
1897 
1898 		/* Send connection flow control advertisement when applicable */
1899 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1900 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1901 			tipc_sk_send_ack(tsk);
1902 
1903 		/* Exit if all requested data or FIN/error received */
1904 		if (copied == buflen || err)
1905 			break;
1906 
1907 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1908 exit:
1909 	release_sock(sk);
1910 	return copied ? copied : rc;
1911 }
1912 
1913 /**
1914  * tipc_write_space - wake up thread if port congestion is released
1915  * @sk: socket
1916  */
1917 static void tipc_write_space(struct sock *sk)
1918 {
1919 	struct socket_wq *wq;
1920 
1921 	rcu_read_lock();
1922 	wq = rcu_dereference(sk->sk_wq);
1923 	if (skwq_has_sleeper(wq))
1924 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1925 						EPOLLWRNORM | EPOLLWRBAND);
1926 	rcu_read_unlock();
1927 }
1928 
1929 /**
1930  * tipc_data_ready - wake up threads to indicate messages have been received
1931  * @sk: socket
1932  * @len: the length of messages
1933  */
1934 static void tipc_data_ready(struct sock *sk)
1935 {
1936 	struct socket_wq *wq;
1937 
1938 	rcu_read_lock();
1939 	wq = rcu_dereference(sk->sk_wq);
1940 	if (skwq_has_sleeper(wq))
1941 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1942 						EPOLLRDNORM | EPOLLRDBAND);
1943 	rcu_read_unlock();
1944 }
1945 
1946 static void tipc_sock_destruct(struct sock *sk)
1947 {
1948 	__skb_queue_purge(&sk->sk_receive_queue);
1949 }
1950 
1951 static void tipc_sk_proto_rcv(struct sock *sk,
1952 			      struct sk_buff_head *inputq,
1953 			      struct sk_buff_head *xmitq)
1954 {
1955 	struct sk_buff *skb = __skb_dequeue(inputq);
1956 	struct tipc_sock *tsk = tipc_sk(sk);
1957 	struct tipc_msg *hdr = buf_msg(skb);
1958 	struct tipc_group *grp = tsk->group;
1959 	bool wakeup = false;
1960 
1961 	switch (msg_user(hdr)) {
1962 	case CONN_MANAGER:
1963 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1964 		return;
1965 	case SOCK_WAKEUP:
1966 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1967 		tsk->cong_link_cnt--;
1968 		wakeup = true;
1969 		break;
1970 	case GROUP_PROTOCOL:
1971 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1972 		break;
1973 	case TOP_SRV:
1974 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1975 				      hdr, inputq, xmitq);
1976 		break;
1977 	default:
1978 		break;
1979 	}
1980 
1981 	if (wakeup)
1982 		sk->sk_write_space(sk);
1983 
1984 	kfree_skb(skb);
1985 }
1986 
1987 /**
1988  * tipc_sk_filter_connect - check incoming message for a connection-based socket
1989  * @tsk: TIPC socket
1990  * @skb: pointer to message buffer.
1991  * Returns true if message should be added to receive queue, false otherwise
1992  */
1993 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1994 {
1995 	struct sock *sk = &tsk->sk;
1996 	struct net *net = sock_net(sk);
1997 	struct tipc_msg *hdr = buf_msg(skb);
1998 	bool con_msg = msg_connected(hdr);
1999 	u32 pport = tsk_peer_port(tsk);
2000 	u32 pnode = tsk_peer_node(tsk);
2001 	u32 oport = msg_origport(hdr);
2002 	u32 onode = msg_orignode(hdr);
2003 	int err = msg_errcode(hdr);
2004 	unsigned long delay;
2005 
2006 	if (unlikely(msg_mcast(hdr)))
2007 		return false;
2008 
2009 	switch (sk->sk_state) {
2010 	case TIPC_CONNECTING:
2011 		/* Setup ACK */
2012 		if (likely(con_msg)) {
2013 			if (err)
2014 				break;
2015 			tipc_sk_finish_conn(tsk, oport, onode);
2016 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2017 			/* ACK+ message with data is added to receive queue */
2018 			if (msg_data_sz(hdr))
2019 				return true;
2020 			/* Empty ACK-, - wake up sleeping connect() and drop */
2021 			sk->sk_data_ready(sk);
2022 			msg_set_dest_droppable(hdr, 1);
2023 			return false;
2024 		}
2025 		/* Ignore connectionless message if not from listening socket */
2026 		if (oport != pport || onode != pnode)
2027 			return false;
2028 
2029 		/* Rejected SYN */
2030 		if (err != TIPC_ERR_OVERLOAD)
2031 			break;
2032 
2033 		/* Prepare for new setup attempt if we have a SYN clone */
2034 		if (skb_queue_empty(&sk->sk_write_queue))
2035 			break;
2036 		get_random_bytes(&delay, 2);
2037 		delay %= (tsk->conn_timeout / 4);
2038 		delay = msecs_to_jiffies(delay + 100);
2039 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2040 		return false;
2041 	case TIPC_OPEN:
2042 	case TIPC_DISCONNECTING:
2043 		return false;
2044 	case TIPC_LISTEN:
2045 		/* Accept only SYN message */
2046 		if (!msg_is_syn(hdr) &&
2047 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2048 			return false;
2049 		if (!con_msg && !err)
2050 			return true;
2051 		return false;
2052 	case TIPC_ESTABLISHED:
2053 		/* Accept only connection-based messages sent by peer */
2054 		if (likely(con_msg && !err && pport == oport && pnode == onode))
2055 			return true;
2056 		if (!tsk_peer_msg(tsk, hdr))
2057 			return false;
2058 		if (!err)
2059 			return true;
2060 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2061 		tipc_node_remove_conn(net, pnode, tsk->portid);
2062 		sk->sk_state_change(sk);
2063 		return true;
2064 	default:
2065 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2066 	}
2067 	/* Abort connection setup attempt */
2068 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2069 	sk->sk_err = ECONNREFUSED;
2070 	sk->sk_state_change(sk);
2071 	return true;
2072 }
2073 
2074 /**
2075  * rcvbuf_limit - get proper overload limit of socket receive queue
2076  * @sk: socket
2077  * @skb: message
2078  *
2079  * For connection oriented messages, irrespective of importance,
2080  * default queue limit is 2 MB.
2081  *
2082  * For connectionless messages, queue limits are based on message
2083  * importance as follows:
2084  *
2085  * TIPC_LOW_IMPORTANCE       (2 MB)
2086  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2087  * TIPC_HIGH_IMPORTANCE      (8 MB)
2088  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2089  *
2090  * Returns overload limit according to corresponding message importance
2091  */
2092 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2093 {
2094 	struct tipc_sock *tsk = tipc_sk(sk);
2095 	struct tipc_msg *hdr = buf_msg(skb);
2096 
2097 	if (unlikely(msg_in_group(hdr)))
2098 		return sk->sk_rcvbuf;
2099 
2100 	if (unlikely(!msg_connected(hdr)))
2101 		return sk->sk_rcvbuf << msg_importance(hdr);
2102 
2103 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2104 		return sk->sk_rcvbuf;
2105 
2106 	return FLOWCTL_MSG_LIM;
2107 }
2108 
2109 /**
2110  * tipc_sk_filter_rcv - validate incoming message
2111  * @sk: socket
2112  * @skb: pointer to message.
2113  *
2114  * Enqueues message on receive queue if acceptable; optionally handles
2115  * disconnect indication for a connected socket.
2116  *
2117  * Called with socket lock already taken
2118  *
2119  */
2120 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2121 			       struct sk_buff_head *xmitq)
2122 {
2123 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2124 	struct tipc_sock *tsk = tipc_sk(sk);
2125 	struct tipc_group *grp = tsk->group;
2126 	struct tipc_msg *hdr = buf_msg(skb);
2127 	struct net *net = sock_net(sk);
2128 	struct sk_buff_head inputq;
2129 	int limit, err = TIPC_OK;
2130 
2131 	TIPC_SKB_CB(skb)->bytes_read = 0;
2132 	__skb_queue_head_init(&inputq);
2133 	__skb_queue_tail(&inputq, skb);
2134 
2135 	if (unlikely(!msg_isdata(hdr)))
2136 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2137 
2138 	if (unlikely(grp))
2139 		tipc_group_filter_msg(grp, &inputq, xmitq);
2140 
2141 	/* Validate and add to receive buffer if there is space */
2142 	while ((skb = __skb_dequeue(&inputq))) {
2143 		hdr = buf_msg(skb);
2144 		limit = rcvbuf_limit(sk, skb);
2145 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2146 		    (!sk_conn && msg_connected(hdr)) ||
2147 		    (!grp && msg_in_group(hdr)))
2148 			err = TIPC_ERR_NO_PORT;
2149 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2150 			atomic_inc(&sk->sk_drops);
2151 			err = TIPC_ERR_OVERLOAD;
2152 		}
2153 
2154 		if (unlikely(err)) {
2155 			tipc_skb_reject(net, err, skb, xmitq);
2156 			err = TIPC_OK;
2157 			continue;
2158 		}
2159 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2160 		skb_set_owner_r(skb, sk);
2161 		sk->sk_data_ready(sk);
2162 	}
2163 }
2164 
2165 /**
2166  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2167  * @sk: socket
2168  * @skb: message
2169  *
2170  * Caller must hold socket lock
2171  */
2172 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2173 {
2174 	unsigned int before = sk_rmem_alloc_get(sk);
2175 	struct sk_buff_head xmitq;
2176 	unsigned int added;
2177 
2178 	__skb_queue_head_init(&xmitq);
2179 
2180 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2181 	added = sk_rmem_alloc_get(sk) - before;
2182 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2183 
2184 	/* Send pending response/rejected messages, if any */
2185 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2186 	return 0;
2187 }
2188 
2189 /**
2190  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2191  *                   inputq and try adding them to socket or backlog queue
2192  * @inputq: list of incoming buffers with potentially different destinations
2193  * @sk: socket where the buffers should be enqueued
2194  * @dport: port number for the socket
2195  *
2196  * Caller must hold socket lock
2197  */
2198 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2199 			    u32 dport, struct sk_buff_head *xmitq)
2200 {
2201 	unsigned long time_limit = jiffies + 2;
2202 	struct sk_buff *skb;
2203 	unsigned int lim;
2204 	atomic_t *dcnt;
2205 	u32 onode;
2206 
2207 	while (skb_queue_len(inputq)) {
2208 		if (unlikely(time_after_eq(jiffies, time_limit)))
2209 			return;
2210 
2211 		skb = tipc_skb_dequeue(inputq, dport);
2212 		if (unlikely(!skb))
2213 			return;
2214 
2215 		/* Add message directly to receive queue if possible */
2216 		if (!sock_owned_by_user(sk)) {
2217 			tipc_sk_filter_rcv(sk, skb, xmitq);
2218 			continue;
2219 		}
2220 
2221 		/* Try backlog, compensating for double-counted bytes */
2222 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2223 		if (!sk->sk_backlog.len)
2224 			atomic_set(dcnt, 0);
2225 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2226 		if (likely(!sk_add_backlog(sk, skb, lim)))
2227 			continue;
2228 
2229 		/* Overload => reject message back to sender */
2230 		onode = tipc_own_addr(sock_net(sk));
2231 		atomic_inc(&sk->sk_drops);
2232 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2233 			__skb_queue_tail(xmitq, skb);
2234 		break;
2235 	}
2236 }
2237 
2238 /**
2239  * tipc_sk_rcv - handle a chain of incoming buffers
2240  * @inputq: buffer list containing the buffers
2241  * Consumes all buffers in list until inputq is empty
2242  * Note: may be called in multiple threads referring to the same queue
2243  */
2244 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2245 {
2246 	struct sk_buff_head xmitq;
2247 	u32 dnode, dport = 0;
2248 	int err;
2249 	struct tipc_sock *tsk;
2250 	struct sock *sk;
2251 	struct sk_buff *skb;
2252 
2253 	__skb_queue_head_init(&xmitq);
2254 	while (skb_queue_len(inputq)) {
2255 		dport = tipc_skb_peek_port(inputq, dport);
2256 		tsk = tipc_sk_lookup(net, dport);
2257 
2258 		if (likely(tsk)) {
2259 			sk = &tsk->sk;
2260 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2261 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2262 				spin_unlock_bh(&sk->sk_lock.slock);
2263 			}
2264 			/* Send pending response/rejected messages, if any */
2265 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2266 			sock_put(sk);
2267 			continue;
2268 		}
2269 		/* No destination socket => dequeue skb if still there */
2270 		skb = tipc_skb_dequeue(inputq, dport);
2271 		if (!skb)
2272 			return;
2273 
2274 		/* Try secondary lookup if unresolved named message */
2275 		err = TIPC_ERR_NO_PORT;
2276 		if (tipc_msg_lookup_dest(net, skb, &err))
2277 			goto xmit;
2278 
2279 		/* Prepare for message rejection */
2280 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2281 			continue;
2282 xmit:
2283 		dnode = msg_destnode(buf_msg(skb));
2284 		tipc_node_xmit_skb(net, skb, dnode, dport);
2285 	}
2286 }
2287 
2288 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2289 {
2290 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2291 	struct sock *sk = sock->sk;
2292 	int done;
2293 
2294 	do {
2295 		int err = sock_error(sk);
2296 		if (err)
2297 			return err;
2298 		if (!*timeo_p)
2299 			return -ETIMEDOUT;
2300 		if (signal_pending(current))
2301 			return sock_intr_errno(*timeo_p);
2302 
2303 		add_wait_queue(sk_sleep(sk), &wait);
2304 		done = sk_wait_event(sk, timeo_p,
2305 				     sk->sk_state != TIPC_CONNECTING, &wait);
2306 		remove_wait_queue(sk_sleep(sk), &wait);
2307 	} while (!done);
2308 	return 0;
2309 }
2310 
2311 /**
2312  * tipc_connect - establish a connection to another TIPC port
2313  * @sock: socket structure
2314  * @dest: socket address for destination port
2315  * @destlen: size of socket address data structure
2316  * @flags: file-related flags associated with socket
2317  *
2318  * Returns 0 on success, errno otherwise
2319  */
2320 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2321 			int destlen, int flags)
2322 {
2323 	struct sock *sk = sock->sk;
2324 	struct tipc_sock *tsk = tipc_sk(sk);
2325 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2326 	struct msghdr m = {NULL,};
2327 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2328 	int previous;
2329 	int res = 0;
2330 
2331 	if (destlen != sizeof(struct sockaddr_tipc))
2332 		return -EINVAL;
2333 
2334 	lock_sock(sk);
2335 
2336 	if (tsk->group) {
2337 		res = -EINVAL;
2338 		goto exit;
2339 	}
2340 
2341 	if (dst->family == AF_UNSPEC) {
2342 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2343 		if (!tipc_sk_type_connectionless(sk))
2344 			res = -EINVAL;
2345 		goto exit;
2346 	} else if (dst->family != AF_TIPC) {
2347 		res = -EINVAL;
2348 	}
2349 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2350 		res = -EINVAL;
2351 	if (res)
2352 		goto exit;
2353 
2354 	/* DGRAM/RDM connect(), just save the destaddr */
2355 	if (tipc_sk_type_connectionless(sk)) {
2356 		memcpy(&tsk->peer, dest, destlen);
2357 		goto exit;
2358 	}
2359 
2360 	previous = sk->sk_state;
2361 
2362 	switch (sk->sk_state) {
2363 	case TIPC_OPEN:
2364 		/* Send a 'SYN-' to destination */
2365 		m.msg_name = dest;
2366 		m.msg_namelen = destlen;
2367 
2368 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2369 		 * indicate send_msg() is never blocked.
2370 		 */
2371 		if (!timeout)
2372 			m.msg_flags = MSG_DONTWAIT;
2373 
2374 		res = __tipc_sendmsg(sock, &m, 0);
2375 		if ((res < 0) && (res != -EWOULDBLOCK))
2376 			goto exit;
2377 
2378 		/* Just entered TIPC_CONNECTING state; the only
2379 		 * difference is that return value in non-blocking
2380 		 * case is EINPROGRESS, rather than EALREADY.
2381 		 */
2382 		res = -EINPROGRESS;
2383 		/* fall thru' */
2384 	case TIPC_CONNECTING:
2385 		if (!timeout) {
2386 			if (previous == TIPC_CONNECTING)
2387 				res = -EALREADY;
2388 			goto exit;
2389 		}
2390 		timeout = msecs_to_jiffies(timeout);
2391 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2392 		res = tipc_wait_for_connect(sock, &timeout);
2393 		break;
2394 	case TIPC_ESTABLISHED:
2395 		res = -EISCONN;
2396 		break;
2397 	default:
2398 		res = -EINVAL;
2399 	}
2400 
2401 exit:
2402 	release_sock(sk);
2403 	return res;
2404 }
2405 
2406 /**
2407  * tipc_listen - allow socket to listen for incoming connections
2408  * @sock: socket structure
2409  * @len: (unused)
2410  *
2411  * Returns 0 on success, errno otherwise
2412  */
2413 static int tipc_listen(struct socket *sock, int len)
2414 {
2415 	struct sock *sk = sock->sk;
2416 	int res;
2417 
2418 	lock_sock(sk);
2419 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2420 	release_sock(sk);
2421 
2422 	return res;
2423 }
2424 
2425 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2426 {
2427 	struct sock *sk = sock->sk;
2428 	DEFINE_WAIT(wait);
2429 	int err;
2430 
2431 	/* True wake-one mechanism for incoming connections: only
2432 	 * one process gets woken up, not the 'whole herd'.
2433 	 * Since we do not 'race & poll' for established sockets
2434 	 * anymore, the common case will execute the loop only once.
2435 	*/
2436 	for (;;) {
2437 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2438 					  TASK_INTERRUPTIBLE);
2439 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2440 			release_sock(sk);
2441 			timeo = schedule_timeout(timeo);
2442 			lock_sock(sk);
2443 		}
2444 		err = 0;
2445 		if (!skb_queue_empty(&sk->sk_receive_queue))
2446 			break;
2447 		err = -EAGAIN;
2448 		if (!timeo)
2449 			break;
2450 		err = sock_intr_errno(timeo);
2451 		if (signal_pending(current))
2452 			break;
2453 	}
2454 	finish_wait(sk_sleep(sk), &wait);
2455 	return err;
2456 }
2457 
2458 /**
2459  * tipc_accept - wait for connection request
2460  * @sock: listening socket
2461  * @newsock: new socket that is to be connected
2462  * @flags: file-related flags associated with socket
2463  *
2464  * Returns 0 on success, errno otherwise
2465  */
2466 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2467 		       bool kern)
2468 {
2469 	struct sock *new_sk, *sk = sock->sk;
2470 	struct sk_buff *buf;
2471 	struct tipc_sock *new_tsock;
2472 	struct tipc_msg *msg;
2473 	long timeo;
2474 	int res;
2475 
2476 	lock_sock(sk);
2477 
2478 	if (sk->sk_state != TIPC_LISTEN) {
2479 		res = -EINVAL;
2480 		goto exit;
2481 	}
2482 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2483 	res = tipc_wait_for_accept(sock, timeo);
2484 	if (res)
2485 		goto exit;
2486 
2487 	buf = skb_peek(&sk->sk_receive_queue);
2488 
2489 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2490 	if (res)
2491 		goto exit;
2492 	security_sk_clone(sock->sk, new_sock->sk);
2493 
2494 	new_sk = new_sock->sk;
2495 	new_tsock = tipc_sk(new_sk);
2496 	msg = buf_msg(buf);
2497 
2498 	/* we lock on new_sk; but lockdep sees the lock on sk */
2499 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2500 
2501 	/*
2502 	 * Reject any stray messages received by new socket
2503 	 * before the socket lock was taken (very, very unlikely)
2504 	 */
2505 	tsk_rej_rx_queue(new_sk);
2506 
2507 	/* Connect new socket to it's peer */
2508 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2509 
2510 	tsk_set_importance(new_tsock, msg_importance(msg));
2511 	if (msg_named(msg)) {
2512 		new_tsock->conn_type = msg_nametype(msg);
2513 		new_tsock->conn_instance = msg_nameinst(msg);
2514 	}
2515 
2516 	/*
2517 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2518 	 * Respond to 'SYN+' by queuing it on new socket.
2519 	 */
2520 	if (!msg_data_sz(msg)) {
2521 		struct msghdr m = {NULL,};
2522 
2523 		tsk_advance_rx_queue(sk);
2524 		__tipc_sendstream(new_sock, &m, 0);
2525 	} else {
2526 		__skb_dequeue(&sk->sk_receive_queue);
2527 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2528 		skb_set_owner_r(buf, new_sk);
2529 	}
2530 	release_sock(new_sk);
2531 exit:
2532 	release_sock(sk);
2533 	return res;
2534 }
2535 
2536 /**
2537  * tipc_shutdown - shutdown socket connection
2538  * @sock: socket structure
2539  * @how: direction to close (must be SHUT_RDWR)
2540  *
2541  * Terminates connection (if necessary), then purges socket's receive queue.
2542  *
2543  * Returns 0 on success, errno otherwise
2544  */
2545 static int tipc_shutdown(struct socket *sock, int how)
2546 {
2547 	struct sock *sk = sock->sk;
2548 	int res;
2549 
2550 	if (how != SHUT_RDWR)
2551 		return -EINVAL;
2552 
2553 	lock_sock(sk);
2554 
2555 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2556 	sk->sk_shutdown = SEND_SHUTDOWN;
2557 
2558 	if (sk->sk_state == TIPC_DISCONNECTING) {
2559 		/* Discard any unreceived messages */
2560 		__skb_queue_purge(&sk->sk_receive_queue);
2561 
2562 		/* Wake up anyone sleeping in poll */
2563 		sk->sk_state_change(sk);
2564 		res = 0;
2565 	} else {
2566 		res = -ENOTCONN;
2567 	}
2568 
2569 	release_sock(sk);
2570 	return res;
2571 }
2572 
2573 static void tipc_sk_check_probing_state(struct sock *sk,
2574 					struct sk_buff_head *list)
2575 {
2576 	struct tipc_sock *tsk = tipc_sk(sk);
2577 	u32 pnode = tsk_peer_node(tsk);
2578 	u32 pport = tsk_peer_port(tsk);
2579 	u32 self = tsk_own_node(tsk);
2580 	u32 oport = tsk->portid;
2581 	struct sk_buff *skb;
2582 
2583 	if (tsk->probe_unacked) {
2584 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2585 		sk->sk_err = ECONNABORTED;
2586 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2587 		sk->sk_state_change(sk);
2588 		return;
2589 	}
2590 	/* Prepare new probe */
2591 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2592 			      pnode, self, pport, oport, TIPC_OK);
2593 	if (skb)
2594 		__skb_queue_tail(list, skb);
2595 	tsk->probe_unacked = true;
2596 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2597 }
2598 
2599 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2600 {
2601 	struct tipc_sock *tsk = tipc_sk(sk);
2602 
2603 	/* Try again later if dest link is congested */
2604 	if (tsk->cong_link_cnt) {
2605 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2606 		return;
2607 	}
2608 	/* Prepare SYN for retransmit */
2609 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2610 }
2611 
2612 static void tipc_sk_timeout(struct timer_list *t)
2613 {
2614 	struct sock *sk = from_timer(sk, t, sk_timer);
2615 	struct tipc_sock *tsk = tipc_sk(sk);
2616 	u32 pnode = tsk_peer_node(tsk);
2617 	struct sk_buff_head list;
2618 	int rc = 0;
2619 
2620 	skb_queue_head_init(&list);
2621 	bh_lock_sock(sk);
2622 
2623 	/* Try again later if socket is busy */
2624 	if (sock_owned_by_user(sk)) {
2625 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2626 		bh_unlock_sock(sk);
2627 		return;
2628 	}
2629 
2630 	if (sk->sk_state == TIPC_ESTABLISHED)
2631 		tipc_sk_check_probing_state(sk, &list);
2632 	else if (sk->sk_state == TIPC_CONNECTING)
2633 		tipc_sk_retry_connect(sk, &list);
2634 
2635 	bh_unlock_sock(sk);
2636 
2637 	if (!skb_queue_empty(&list))
2638 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2639 
2640 	/* SYN messages may cause link congestion */
2641 	if (rc == -ELINKCONG) {
2642 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2643 		tsk->cong_link_cnt = 1;
2644 	}
2645 	sock_put(sk);
2646 }
2647 
2648 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2649 			   struct tipc_name_seq const *seq)
2650 {
2651 	struct sock *sk = &tsk->sk;
2652 	struct net *net = sock_net(sk);
2653 	struct publication *publ;
2654 	u32 key;
2655 
2656 	if (scope != TIPC_NODE_SCOPE)
2657 		scope = TIPC_CLUSTER_SCOPE;
2658 
2659 	if (tipc_sk_connected(sk))
2660 		return -EINVAL;
2661 	key = tsk->portid + tsk->pub_count + 1;
2662 	if (key == tsk->portid)
2663 		return -EADDRINUSE;
2664 
2665 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2666 				    scope, tsk->portid, key);
2667 	if (unlikely(!publ))
2668 		return -EINVAL;
2669 
2670 	list_add(&publ->binding_sock, &tsk->publications);
2671 	tsk->pub_count++;
2672 	tsk->published = 1;
2673 	return 0;
2674 }
2675 
2676 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2677 			    struct tipc_name_seq const *seq)
2678 {
2679 	struct net *net = sock_net(&tsk->sk);
2680 	struct publication *publ;
2681 	struct publication *safe;
2682 	int rc = -EINVAL;
2683 
2684 	if (scope != TIPC_NODE_SCOPE)
2685 		scope = TIPC_CLUSTER_SCOPE;
2686 
2687 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2688 		if (seq) {
2689 			if (publ->scope != scope)
2690 				continue;
2691 			if (publ->type != seq->type)
2692 				continue;
2693 			if (publ->lower != seq->lower)
2694 				continue;
2695 			if (publ->upper != seq->upper)
2696 				break;
2697 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2698 					      publ->upper, publ->key);
2699 			rc = 0;
2700 			break;
2701 		}
2702 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2703 				      publ->upper, publ->key);
2704 		rc = 0;
2705 	}
2706 	if (list_empty(&tsk->publications))
2707 		tsk->published = 0;
2708 	return rc;
2709 }
2710 
2711 /* tipc_sk_reinit: set non-zero address in all existing sockets
2712  *                 when we go from standalone to network mode.
2713  */
2714 void tipc_sk_reinit(struct net *net)
2715 {
2716 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2717 	struct rhashtable_iter iter;
2718 	struct tipc_sock *tsk;
2719 	struct tipc_msg *msg;
2720 
2721 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2722 
2723 	do {
2724 		rhashtable_walk_start(&iter);
2725 
2726 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2727 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2728 			msg = &tsk->phdr;
2729 			msg_set_prevnode(msg, tipc_own_addr(net));
2730 			msg_set_orignode(msg, tipc_own_addr(net));
2731 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2732 		}
2733 
2734 		rhashtable_walk_stop(&iter);
2735 	} while (tsk == ERR_PTR(-EAGAIN));
2736 
2737 	rhashtable_walk_exit(&iter);
2738 }
2739 
2740 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2741 {
2742 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2743 	struct tipc_sock *tsk;
2744 
2745 	rcu_read_lock();
2746 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2747 	if (tsk)
2748 		sock_hold(&tsk->sk);
2749 	rcu_read_unlock();
2750 
2751 	return tsk;
2752 }
2753 
2754 static int tipc_sk_insert(struct tipc_sock *tsk)
2755 {
2756 	struct sock *sk = &tsk->sk;
2757 	struct net *net = sock_net(sk);
2758 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2759 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2760 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2761 
2762 	while (remaining--) {
2763 		portid++;
2764 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2765 			portid = TIPC_MIN_PORT;
2766 		tsk->portid = portid;
2767 		sock_hold(&tsk->sk);
2768 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2769 						   tsk_rht_params))
2770 			return 0;
2771 		sock_put(&tsk->sk);
2772 	}
2773 
2774 	return -1;
2775 }
2776 
2777 static void tipc_sk_remove(struct tipc_sock *tsk)
2778 {
2779 	struct sock *sk = &tsk->sk;
2780 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2781 
2782 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2783 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2784 		__sock_put(sk);
2785 	}
2786 }
2787 
2788 static const struct rhashtable_params tsk_rht_params = {
2789 	.nelem_hint = 192,
2790 	.head_offset = offsetof(struct tipc_sock, node),
2791 	.key_offset = offsetof(struct tipc_sock, portid),
2792 	.key_len = sizeof(u32), /* portid */
2793 	.max_size = 1048576,
2794 	.min_size = 256,
2795 	.automatic_shrinking = true,
2796 };
2797 
2798 int tipc_sk_rht_init(struct net *net)
2799 {
2800 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2801 
2802 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2803 }
2804 
2805 void tipc_sk_rht_destroy(struct net *net)
2806 {
2807 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2808 
2809 	/* Wait for socket readers to complete */
2810 	synchronize_net();
2811 
2812 	rhashtable_destroy(&tn->sk_rht);
2813 }
2814 
2815 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2816 {
2817 	struct net *net = sock_net(&tsk->sk);
2818 	struct tipc_group *grp = tsk->group;
2819 	struct tipc_msg *hdr = &tsk->phdr;
2820 	struct tipc_name_seq seq;
2821 	int rc;
2822 
2823 	if (mreq->type < TIPC_RESERVED_TYPES)
2824 		return -EACCES;
2825 	if (mreq->scope > TIPC_NODE_SCOPE)
2826 		return -EINVAL;
2827 	if (grp)
2828 		return -EACCES;
2829 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2830 	if (!grp)
2831 		return -ENOMEM;
2832 	tsk->group = grp;
2833 	msg_set_lookup_scope(hdr, mreq->scope);
2834 	msg_set_nametype(hdr, mreq->type);
2835 	msg_set_dest_droppable(hdr, true);
2836 	seq.type = mreq->type;
2837 	seq.lower = mreq->instance;
2838 	seq.upper = seq.lower;
2839 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2840 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2841 	if (rc) {
2842 		tipc_group_delete(net, grp);
2843 		tsk->group = NULL;
2844 		return rc;
2845 	}
2846 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2847 	tsk->mc_method.rcast = true;
2848 	tsk->mc_method.mandatory = true;
2849 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2850 	return rc;
2851 }
2852 
2853 static int tipc_sk_leave(struct tipc_sock *tsk)
2854 {
2855 	struct net *net = sock_net(&tsk->sk);
2856 	struct tipc_group *grp = tsk->group;
2857 	struct tipc_name_seq seq;
2858 	int scope;
2859 
2860 	if (!grp)
2861 		return -EINVAL;
2862 	tipc_group_self(grp, &seq, &scope);
2863 	tipc_group_delete(net, grp);
2864 	tsk->group = NULL;
2865 	tipc_sk_withdraw(tsk, scope, &seq);
2866 	return 0;
2867 }
2868 
2869 /**
2870  * tipc_setsockopt - set socket option
2871  * @sock: socket structure
2872  * @lvl: option level
2873  * @opt: option identifier
2874  * @ov: pointer to new option value
2875  * @ol: length of option value
2876  *
2877  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2878  * (to ease compatibility).
2879  *
2880  * Returns 0 on success, errno otherwise
2881  */
2882 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2883 			   char __user *ov, unsigned int ol)
2884 {
2885 	struct sock *sk = sock->sk;
2886 	struct tipc_sock *tsk = tipc_sk(sk);
2887 	struct tipc_group_req mreq;
2888 	u32 value = 0;
2889 	int res = 0;
2890 
2891 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2892 		return 0;
2893 	if (lvl != SOL_TIPC)
2894 		return -ENOPROTOOPT;
2895 
2896 	switch (opt) {
2897 	case TIPC_IMPORTANCE:
2898 	case TIPC_SRC_DROPPABLE:
2899 	case TIPC_DEST_DROPPABLE:
2900 	case TIPC_CONN_TIMEOUT:
2901 		if (ol < sizeof(value))
2902 			return -EINVAL;
2903 		if (get_user(value, (u32 __user *)ov))
2904 			return -EFAULT;
2905 		break;
2906 	case TIPC_GROUP_JOIN:
2907 		if (ol < sizeof(mreq))
2908 			return -EINVAL;
2909 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2910 			return -EFAULT;
2911 		break;
2912 	default:
2913 		if (ov || ol)
2914 			return -EINVAL;
2915 	}
2916 
2917 	lock_sock(sk);
2918 
2919 	switch (opt) {
2920 	case TIPC_IMPORTANCE:
2921 		res = tsk_set_importance(tsk, value);
2922 		break;
2923 	case TIPC_SRC_DROPPABLE:
2924 		if (sock->type != SOCK_STREAM)
2925 			tsk_set_unreliable(tsk, value);
2926 		else
2927 			res = -ENOPROTOOPT;
2928 		break;
2929 	case TIPC_DEST_DROPPABLE:
2930 		tsk_set_unreturnable(tsk, value);
2931 		break;
2932 	case TIPC_CONN_TIMEOUT:
2933 		tipc_sk(sk)->conn_timeout = value;
2934 		break;
2935 	case TIPC_MCAST_BROADCAST:
2936 		tsk->mc_method.rcast = false;
2937 		tsk->mc_method.mandatory = true;
2938 		break;
2939 	case TIPC_MCAST_REPLICAST:
2940 		tsk->mc_method.rcast = true;
2941 		tsk->mc_method.mandatory = true;
2942 		break;
2943 	case TIPC_GROUP_JOIN:
2944 		res = tipc_sk_join(tsk, &mreq);
2945 		break;
2946 	case TIPC_GROUP_LEAVE:
2947 		res = tipc_sk_leave(tsk);
2948 		break;
2949 	default:
2950 		res = -EINVAL;
2951 	}
2952 
2953 	release_sock(sk);
2954 
2955 	return res;
2956 }
2957 
2958 /**
2959  * tipc_getsockopt - get socket option
2960  * @sock: socket structure
2961  * @lvl: option level
2962  * @opt: option identifier
2963  * @ov: receptacle for option value
2964  * @ol: receptacle for length of option value
2965  *
2966  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2967  * (to ease compatibility).
2968  *
2969  * Returns 0 on success, errno otherwise
2970  */
2971 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2972 			   char __user *ov, int __user *ol)
2973 {
2974 	struct sock *sk = sock->sk;
2975 	struct tipc_sock *tsk = tipc_sk(sk);
2976 	struct tipc_name_seq seq;
2977 	int len, scope;
2978 	u32 value;
2979 	int res;
2980 
2981 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2982 		return put_user(0, ol);
2983 	if (lvl != SOL_TIPC)
2984 		return -ENOPROTOOPT;
2985 	res = get_user(len, ol);
2986 	if (res)
2987 		return res;
2988 
2989 	lock_sock(sk);
2990 
2991 	switch (opt) {
2992 	case TIPC_IMPORTANCE:
2993 		value = tsk_importance(tsk);
2994 		break;
2995 	case TIPC_SRC_DROPPABLE:
2996 		value = tsk_unreliable(tsk);
2997 		break;
2998 	case TIPC_DEST_DROPPABLE:
2999 		value = tsk_unreturnable(tsk);
3000 		break;
3001 	case TIPC_CONN_TIMEOUT:
3002 		value = tsk->conn_timeout;
3003 		/* no need to set "res", since already 0 at this point */
3004 		break;
3005 	case TIPC_NODE_RECVQ_DEPTH:
3006 		value = 0; /* was tipc_queue_size, now obsolete */
3007 		break;
3008 	case TIPC_SOCK_RECVQ_DEPTH:
3009 		value = skb_queue_len(&sk->sk_receive_queue);
3010 		break;
3011 	case TIPC_GROUP_JOIN:
3012 		seq.type = 0;
3013 		if (tsk->group)
3014 			tipc_group_self(tsk->group, &seq, &scope);
3015 		value = seq.type;
3016 		break;
3017 	default:
3018 		res = -EINVAL;
3019 	}
3020 
3021 	release_sock(sk);
3022 
3023 	if (res)
3024 		return res;	/* "get" failed */
3025 
3026 	if (len < sizeof(value))
3027 		return -EINVAL;
3028 
3029 	if (copy_to_user(ov, &value, sizeof(value)))
3030 		return -EFAULT;
3031 
3032 	return put_user(sizeof(value), ol);
3033 }
3034 
3035 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3036 {
3037 	struct net *net = sock_net(sock->sk);
3038 	struct tipc_sioc_nodeid_req nr = {0};
3039 	struct tipc_sioc_ln_req lnr;
3040 	void __user *argp = (void __user *)arg;
3041 
3042 	switch (cmd) {
3043 	case SIOCGETLINKNAME:
3044 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3045 			return -EFAULT;
3046 		if (!tipc_node_get_linkname(net,
3047 					    lnr.bearer_id & 0xffff, lnr.peer,
3048 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3049 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3050 				return -EFAULT;
3051 			return 0;
3052 		}
3053 		return -EADDRNOTAVAIL;
3054 	case SIOCGETNODEID:
3055 		if (copy_from_user(&nr, argp, sizeof(nr)))
3056 			return -EFAULT;
3057 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3058 			return -EADDRNOTAVAIL;
3059 		if (copy_to_user(argp, &nr, sizeof(nr)))
3060 			return -EFAULT;
3061 		return 0;
3062 	default:
3063 		return -ENOIOCTLCMD;
3064 	}
3065 }
3066 
3067 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3068 {
3069 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3070 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3071 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3072 
3073 	tsk1->peer.family = AF_TIPC;
3074 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3075 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3076 	tsk1->peer.addr.id.ref = tsk2->portid;
3077 	tsk1->peer.addr.id.node = onode;
3078 	tsk2->peer.family = AF_TIPC;
3079 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3080 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3081 	tsk2->peer.addr.id.ref = tsk1->portid;
3082 	tsk2->peer.addr.id.node = onode;
3083 
3084 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3085 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3086 	return 0;
3087 }
3088 
3089 /* Protocol switches for the various types of TIPC sockets */
3090 
3091 static const struct proto_ops msg_ops = {
3092 	.owner		= THIS_MODULE,
3093 	.family		= AF_TIPC,
3094 	.release	= tipc_release,
3095 	.bind		= tipc_bind,
3096 	.connect	= tipc_connect,
3097 	.socketpair	= tipc_socketpair,
3098 	.accept		= sock_no_accept,
3099 	.getname	= tipc_getname,
3100 	.poll		= tipc_poll,
3101 	.ioctl		= tipc_ioctl,
3102 	.listen		= sock_no_listen,
3103 	.shutdown	= tipc_shutdown,
3104 	.setsockopt	= tipc_setsockopt,
3105 	.getsockopt	= tipc_getsockopt,
3106 	.sendmsg	= tipc_sendmsg,
3107 	.recvmsg	= tipc_recvmsg,
3108 	.mmap		= sock_no_mmap,
3109 	.sendpage	= sock_no_sendpage
3110 };
3111 
3112 static const struct proto_ops packet_ops = {
3113 	.owner		= THIS_MODULE,
3114 	.family		= AF_TIPC,
3115 	.release	= tipc_release,
3116 	.bind		= tipc_bind,
3117 	.connect	= tipc_connect,
3118 	.socketpair	= tipc_socketpair,
3119 	.accept		= tipc_accept,
3120 	.getname	= tipc_getname,
3121 	.poll		= tipc_poll,
3122 	.ioctl		= tipc_ioctl,
3123 	.listen		= tipc_listen,
3124 	.shutdown	= tipc_shutdown,
3125 	.setsockopt	= tipc_setsockopt,
3126 	.getsockopt	= tipc_getsockopt,
3127 	.sendmsg	= tipc_send_packet,
3128 	.recvmsg	= tipc_recvmsg,
3129 	.mmap		= sock_no_mmap,
3130 	.sendpage	= sock_no_sendpage
3131 };
3132 
3133 static const struct proto_ops stream_ops = {
3134 	.owner		= THIS_MODULE,
3135 	.family		= AF_TIPC,
3136 	.release	= tipc_release,
3137 	.bind		= tipc_bind,
3138 	.connect	= tipc_connect,
3139 	.socketpair	= tipc_socketpair,
3140 	.accept		= tipc_accept,
3141 	.getname	= tipc_getname,
3142 	.poll		= tipc_poll,
3143 	.ioctl		= tipc_ioctl,
3144 	.listen		= tipc_listen,
3145 	.shutdown	= tipc_shutdown,
3146 	.setsockopt	= tipc_setsockopt,
3147 	.getsockopt	= tipc_getsockopt,
3148 	.sendmsg	= tipc_sendstream,
3149 	.recvmsg	= tipc_recvstream,
3150 	.mmap		= sock_no_mmap,
3151 	.sendpage	= sock_no_sendpage
3152 };
3153 
3154 static const struct net_proto_family tipc_family_ops = {
3155 	.owner		= THIS_MODULE,
3156 	.family		= AF_TIPC,
3157 	.create		= tipc_sk_create
3158 };
3159 
3160 static struct proto tipc_proto = {
3161 	.name		= "TIPC",
3162 	.owner		= THIS_MODULE,
3163 	.obj_size	= sizeof(struct tipc_sock),
3164 	.sysctl_rmem	= sysctl_tipc_rmem
3165 };
3166 
3167 /**
3168  * tipc_socket_init - initialize TIPC socket interface
3169  *
3170  * Returns 0 on success, errno otherwise
3171  */
3172 int tipc_socket_init(void)
3173 {
3174 	int res;
3175 
3176 	res = proto_register(&tipc_proto, 1);
3177 	if (res) {
3178 		pr_err("Failed to register TIPC protocol type\n");
3179 		goto out;
3180 	}
3181 
3182 	res = sock_register(&tipc_family_ops);
3183 	if (res) {
3184 		pr_err("Failed to register TIPC socket type\n");
3185 		proto_unregister(&tipc_proto);
3186 		goto out;
3187 	}
3188  out:
3189 	return res;
3190 }
3191 
3192 /**
3193  * tipc_socket_stop - stop TIPC socket interface
3194  */
3195 void tipc_socket_stop(void)
3196 {
3197 	sock_unregister(tipc_family_ops.family);
3198 	proto_unregister(&tipc_proto);
3199 }
3200 
3201 /* Caller should hold socket lock for the passed tipc socket. */
3202 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3203 {
3204 	u32 peer_node;
3205 	u32 peer_port;
3206 	struct nlattr *nest;
3207 
3208 	peer_node = tsk_peer_node(tsk);
3209 	peer_port = tsk_peer_port(tsk);
3210 
3211 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3212 
3213 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3214 		goto msg_full;
3215 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3216 		goto msg_full;
3217 
3218 	if (tsk->conn_type != 0) {
3219 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3220 			goto msg_full;
3221 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3222 			goto msg_full;
3223 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3224 			goto msg_full;
3225 	}
3226 	nla_nest_end(skb, nest);
3227 
3228 	return 0;
3229 
3230 msg_full:
3231 	nla_nest_cancel(skb, nest);
3232 
3233 	return -EMSGSIZE;
3234 }
3235 
3236 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3237 			  *tsk)
3238 {
3239 	struct net *net = sock_net(skb->sk);
3240 	struct sock *sk = &tsk->sk;
3241 
3242 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3243 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3244 		return -EMSGSIZE;
3245 
3246 	if (tipc_sk_connected(sk)) {
3247 		if (__tipc_nl_add_sk_con(skb, tsk))
3248 			return -EMSGSIZE;
3249 	} else if (!list_empty(&tsk->publications)) {
3250 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3251 			return -EMSGSIZE;
3252 	}
3253 	return 0;
3254 }
3255 
3256 /* Caller should hold socket lock for the passed tipc socket. */
3257 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3258 			    struct tipc_sock *tsk)
3259 {
3260 	struct nlattr *attrs;
3261 	void *hdr;
3262 
3263 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3264 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3265 	if (!hdr)
3266 		goto msg_cancel;
3267 
3268 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3269 	if (!attrs)
3270 		goto genlmsg_cancel;
3271 
3272 	if (__tipc_nl_add_sk_info(skb, tsk))
3273 		goto attr_msg_cancel;
3274 
3275 	nla_nest_end(skb, attrs);
3276 	genlmsg_end(skb, hdr);
3277 
3278 	return 0;
3279 
3280 attr_msg_cancel:
3281 	nla_nest_cancel(skb, attrs);
3282 genlmsg_cancel:
3283 	genlmsg_cancel(skb, hdr);
3284 msg_cancel:
3285 	return -EMSGSIZE;
3286 }
3287 
3288 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3289 		    int (*skb_handler)(struct sk_buff *skb,
3290 				       struct netlink_callback *cb,
3291 				       struct tipc_sock *tsk))
3292 {
3293 	struct rhashtable_iter *iter = (void *)cb->args[4];
3294 	struct tipc_sock *tsk;
3295 	int err;
3296 
3297 	rhashtable_walk_start(iter);
3298 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3299 		if (IS_ERR(tsk)) {
3300 			err = PTR_ERR(tsk);
3301 			if (err == -EAGAIN) {
3302 				err = 0;
3303 				continue;
3304 			}
3305 			break;
3306 		}
3307 
3308 		sock_hold(&tsk->sk);
3309 		rhashtable_walk_stop(iter);
3310 		lock_sock(&tsk->sk);
3311 		err = skb_handler(skb, cb, tsk);
3312 		if (err) {
3313 			release_sock(&tsk->sk);
3314 			sock_put(&tsk->sk);
3315 			goto out;
3316 		}
3317 		release_sock(&tsk->sk);
3318 		rhashtable_walk_start(iter);
3319 		sock_put(&tsk->sk);
3320 	}
3321 	rhashtable_walk_stop(iter);
3322 out:
3323 	return skb->len;
3324 }
3325 EXPORT_SYMBOL(tipc_nl_sk_walk);
3326 
3327 int tipc_dump_start(struct netlink_callback *cb)
3328 {
3329 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3330 }
3331 EXPORT_SYMBOL(tipc_dump_start);
3332 
3333 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3334 {
3335 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3336 	struct rhashtable_iter *iter = (void *)cb->args[4];
3337 	struct tipc_net *tn = tipc_net(net);
3338 
3339 	if (!iter) {
3340 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3341 		if (!iter)
3342 			return -ENOMEM;
3343 
3344 		cb->args[4] = (long)iter;
3345 	}
3346 
3347 	rhashtable_walk_enter(&tn->sk_rht, iter);
3348 	return 0;
3349 }
3350 
3351 int tipc_dump_done(struct netlink_callback *cb)
3352 {
3353 	struct rhashtable_iter *hti = (void *)cb->args[4];
3354 
3355 	rhashtable_walk_exit(hti);
3356 	kfree(hti);
3357 	return 0;
3358 }
3359 EXPORT_SYMBOL(tipc_dump_done);
3360 
3361 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3362 			   struct tipc_sock *tsk, u32 sk_filter_state,
3363 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3364 {
3365 	struct sock *sk = &tsk->sk;
3366 	struct nlattr *attrs;
3367 	struct nlattr *stat;
3368 
3369 	/*filter response w.r.t sk_state*/
3370 	if (!(sk_filter_state & (1 << sk->sk_state)))
3371 		return 0;
3372 
3373 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3374 	if (!attrs)
3375 		goto msg_cancel;
3376 
3377 	if (__tipc_nl_add_sk_info(skb, tsk))
3378 		goto attr_msg_cancel;
3379 
3380 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3381 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3382 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3383 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3384 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3385 					 sock_i_uid(sk))) ||
3386 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3387 			      tipc_diag_gen_cookie(sk),
3388 			      TIPC_NLA_SOCK_PAD))
3389 		goto attr_msg_cancel;
3390 
3391 	stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3392 	if (!stat)
3393 		goto attr_msg_cancel;
3394 
3395 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3396 			skb_queue_len(&sk->sk_receive_queue)) ||
3397 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3398 			skb_queue_len(&sk->sk_write_queue)) ||
3399 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3400 			atomic_read(&sk->sk_drops)))
3401 		goto stat_msg_cancel;
3402 
3403 	if (tsk->cong_link_cnt &&
3404 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3405 		goto stat_msg_cancel;
3406 
3407 	if (tsk_conn_cong(tsk) &&
3408 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3409 		goto stat_msg_cancel;
3410 
3411 	nla_nest_end(skb, stat);
3412 
3413 	if (tsk->group)
3414 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3415 			goto stat_msg_cancel;
3416 
3417 	nla_nest_end(skb, attrs);
3418 
3419 	return 0;
3420 
3421 stat_msg_cancel:
3422 	nla_nest_cancel(skb, stat);
3423 attr_msg_cancel:
3424 	nla_nest_cancel(skb, attrs);
3425 msg_cancel:
3426 	return -EMSGSIZE;
3427 }
3428 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3429 
3430 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3431 {
3432 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3433 }
3434 
3435 /* Caller should hold socket lock for the passed tipc socket. */
3436 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3437 				 struct netlink_callback *cb,
3438 				 struct publication *publ)
3439 {
3440 	void *hdr;
3441 	struct nlattr *attrs;
3442 
3443 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3444 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3445 	if (!hdr)
3446 		goto msg_cancel;
3447 
3448 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3449 	if (!attrs)
3450 		goto genlmsg_cancel;
3451 
3452 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3453 		goto attr_msg_cancel;
3454 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3455 		goto attr_msg_cancel;
3456 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3457 		goto attr_msg_cancel;
3458 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3459 		goto attr_msg_cancel;
3460 
3461 	nla_nest_end(skb, attrs);
3462 	genlmsg_end(skb, hdr);
3463 
3464 	return 0;
3465 
3466 attr_msg_cancel:
3467 	nla_nest_cancel(skb, attrs);
3468 genlmsg_cancel:
3469 	genlmsg_cancel(skb, hdr);
3470 msg_cancel:
3471 	return -EMSGSIZE;
3472 }
3473 
3474 /* Caller should hold socket lock for the passed tipc socket. */
3475 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3476 				  struct netlink_callback *cb,
3477 				  struct tipc_sock *tsk, u32 *last_publ)
3478 {
3479 	int err;
3480 	struct publication *p;
3481 
3482 	if (*last_publ) {
3483 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3484 			if (p->key == *last_publ)
3485 				break;
3486 		}
3487 		if (p->key != *last_publ) {
3488 			/* We never set seq or call nl_dump_check_consistent()
3489 			 * this means that setting prev_seq here will cause the
3490 			 * consistence check to fail in the netlink callback
3491 			 * handler. Resulting in the last NLMSG_DONE message
3492 			 * having the NLM_F_DUMP_INTR flag set.
3493 			 */
3494 			cb->prev_seq = 1;
3495 			*last_publ = 0;
3496 			return -EPIPE;
3497 		}
3498 	} else {
3499 		p = list_first_entry(&tsk->publications, struct publication,
3500 				     binding_sock);
3501 	}
3502 
3503 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3504 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3505 		if (err) {
3506 			*last_publ = p->key;
3507 			return err;
3508 		}
3509 	}
3510 	*last_publ = 0;
3511 
3512 	return 0;
3513 }
3514 
3515 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3516 {
3517 	int err;
3518 	u32 tsk_portid = cb->args[0];
3519 	u32 last_publ = cb->args[1];
3520 	u32 done = cb->args[2];
3521 	struct net *net = sock_net(skb->sk);
3522 	struct tipc_sock *tsk;
3523 
3524 	if (!tsk_portid) {
3525 		struct nlattr **attrs;
3526 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3527 
3528 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3529 		if (err)
3530 			return err;
3531 
3532 		if (!attrs[TIPC_NLA_SOCK])
3533 			return -EINVAL;
3534 
3535 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3536 				       attrs[TIPC_NLA_SOCK],
3537 				       tipc_nl_sock_policy, NULL);
3538 		if (err)
3539 			return err;
3540 
3541 		if (!sock[TIPC_NLA_SOCK_REF])
3542 			return -EINVAL;
3543 
3544 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3545 	}
3546 
3547 	if (done)
3548 		return 0;
3549 
3550 	tsk = tipc_sk_lookup(net, tsk_portid);
3551 	if (!tsk)
3552 		return -EINVAL;
3553 
3554 	lock_sock(&tsk->sk);
3555 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3556 	if (!err)
3557 		done = 1;
3558 	release_sock(&tsk->sk);
3559 	sock_put(&tsk->sk);
3560 
3561 	cb->args[0] = tsk_portid;
3562 	cb->args[1] = last_publ;
3563 	cb->args[2] = done;
3564 
3565 	return skb->len;
3566 }
3567