xref: /linux/net/tipc/socket.c (revision b76960c0f6b25d447a1493c4388defb9e8e76c63)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include <linux/rhashtable.h>
38 #include <linux/sched/signal.h>
39 
40 #include "core.h"
41 #include "name_table.h"
42 #include "node.h"
43 #include "link.h"
44 #include "name_distr.h"
45 #include "socket.h"
46 #include "bcast.h"
47 #include "netlink.h"
48 #include "group.h"
49 
50 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
51 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
52 #define TIPC_FWD_MSG		1
53 #define TIPC_MAX_PORT		0xffffffff
54 #define TIPC_MIN_PORT		1
55 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
56 
57 enum {
58 	TIPC_LISTEN = TCP_LISTEN,
59 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
60 	TIPC_OPEN = TCP_CLOSE,
61 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
62 	TIPC_CONNECTING = TCP_SYN_SENT,
63 };
64 
65 struct sockaddr_pair {
66 	struct sockaddr_tipc sock;
67 	struct sockaddr_tipc member;
68 };
69 
70 /**
71  * struct tipc_sock - TIPC socket structure
72  * @sk: socket - interacts with 'port' and with user via the socket API
73  * @conn_type: TIPC type used when connection was established
74  * @conn_instance: TIPC instance used when connection was established
75  * @published: non-zero if port has one or more associated names
76  * @max_pkt: maximum packet size "hint" used when building messages sent by port
77  * @portid: unique port identity in TIPC socket hash table
78  * @phdr: preformatted message header used when sending messages
79  * #cong_links: list of congested links
80  * @publications: list of publications for port
81  * @blocking_link: address of the congested link we are currently sleeping on
82  * @pub_count: total # of publications port has made during its lifetime
83  * @probing_state:
84  * @conn_timeout: the time we can wait for an unresponded setup request
85  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
86  * @cong_link_cnt: number of congested links
87  * @snt_unacked: # messages sent by socket, and not yet acked by peer
88  * @rcv_unacked: # messages read by user, but not yet acked back to peer
89  * @peer: 'connected' peer for dgram/rdm
90  * @node: hash table node
91  * @mc_method: cookie for use between socket and broadcast layer
92  * @rcu: rcu struct for tipc_sock
93  */
94 struct tipc_sock {
95 	struct sock sk;
96 	u32 conn_type;
97 	u32 conn_instance;
98 	int published;
99 	u32 max_pkt;
100 	u32 portid;
101 	struct tipc_msg phdr;
102 	struct list_head cong_links;
103 	struct list_head publications;
104 	u32 pub_count;
105 	uint conn_timeout;
106 	atomic_t dupl_rcvcnt;
107 	bool probe_unacked;
108 	u16 cong_link_cnt;
109 	u16 snt_unacked;
110 	u16 snd_win;
111 	u16 peer_caps;
112 	u16 rcv_unacked;
113 	u16 rcv_win;
114 	struct sockaddr_tipc peer;
115 	struct rhash_head node;
116 	struct tipc_mc_method mc_method;
117 	struct rcu_head rcu;
118 	struct tipc_group *group;
119 	bool group_is_open;
120 };
121 
122 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
123 static void tipc_data_ready(struct sock *sk);
124 static void tipc_write_space(struct sock *sk);
125 static void tipc_sock_destruct(struct sock *sk);
126 static int tipc_release(struct socket *sock);
127 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
128 		       bool kern);
129 static void tipc_sk_timeout(struct timer_list *t);
130 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
131 			   struct tipc_name_seq const *seq);
132 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
133 			    struct tipc_name_seq const *seq);
134 static int tipc_sk_leave(struct tipc_sock *tsk);
135 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
136 static int tipc_sk_insert(struct tipc_sock *tsk);
137 static void tipc_sk_remove(struct tipc_sock *tsk);
138 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
139 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
140 
141 static const struct proto_ops packet_ops;
142 static const struct proto_ops stream_ops;
143 static const struct proto_ops msg_ops;
144 static struct proto tipc_proto;
145 static const struct rhashtable_params tsk_rht_params;
146 
147 static u32 tsk_own_node(struct tipc_sock *tsk)
148 {
149 	return msg_prevnode(&tsk->phdr);
150 }
151 
152 static u32 tsk_peer_node(struct tipc_sock *tsk)
153 {
154 	return msg_destnode(&tsk->phdr);
155 }
156 
157 static u32 tsk_peer_port(struct tipc_sock *tsk)
158 {
159 	return msg_destport(&tsk->phdr);
160 }
161 
162 static  bool tsk_unreliable(struct tipc_sock *tsk)
163 {
164 	return msg_src_droppable(&tsk->phdr) != 0;
165 }
166 
167 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
168 {
169 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
170 }
171 
172 static bool tsk_unreturnable(struct tipc_sock *tsk)
173 {
174 	return msg_dest_droppable(&tsk->phdr) != 0;
175 }
176 
177 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
178 {
179 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
180 }
181 
182 static int tsk_importance(struct tipc_sock *tsk)
183 {
184 	return msg_importance(&tsk->phdr);
185 }
186 
187 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
188 {
189 	if (imp > TIPC_CRITICAL_IMPORTANCE)
190 		return -EINVAL;
191 	msg_set_importance(&tsk->phdr, (u32)imp);
192 	return 0;
193 }
194 
195 static struct tipc_sock *tipc_sk(const struct sock *sk)
196 {
197 	return container_of(sk, struct tipc_sock, sk);
198 }
199 
200 static bool tsk_conn_cong(struct tipc_sock *tsk)
201 {
202 	return tsk->snt_unacked > tsk->snd_win;
203 }
204 
205 static u16 tsk_blocks(int len)
206 {
207 	return ((len / FLOWCTL_BLK_SZ) + 1);
208 }
209 
210 /* tsk_blocks(): translate a buffer size in bytes to number of
211  * advertisable blocks, taking into account the ratio truesize(len)/len
212  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
213  */
214 static u16 tsk_adv_blocks(int len)
215 {
216 	return len / FLOWCTL_BLK_SZ / 4;
217 }
218 
219 /* tsk_inc(): increment counter for sent or received data
220  * - If block based flow control is not supported by peer we
221  *   fall back to message based ditto, incrementing the counter
222  */
223 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
224 {
225 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
226 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
227 	return 1;
228 }
229 
230 /**
231  * tsk_advance_rx_queue - discard first buffer in socket receive queue
232  *
233  * Caller must hold socket lock
234  */
235 static void tsk_advance_rx_queue(struct sock *sk)
236 {
237 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
238 }
239 
240 /* tipc_sk_respond() : send response message back to sender
241  */
242 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
243 {
244 	u32 selector;
245 	u32 dnode;
246 	u32 onode = tipc_own_addr(sock_net(sk));
247 
248 	if (!tipc_msg_reverse(onode, &skb, err))
249 		return;
250 
251 	dnode = msg_destnode(buf_msg(skb));
252 	selector = msg_origport(buf_msg(skb));
253 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
254 }
255 
256 /**
257  * tsk_rej_rx_queue - reject all buffers in socket receive queue
258  *
259  * Caller must hold socket lock
260  */
261 static void tsk_rej_rx_queue(struct sock *sk)
262 {
263 	struct sk_buff *skb;
264 
265 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
266 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
267 }
268 
269 static bool tipc_sk_connected(struct sock *sk)
270 {
271 	return sk->sk_state == TIPC_ESTABLISHED;
272 }
273 
274 /* tipc_sk_type_connectionless - check if the socket is datagram socket
275  * @sk: socket
276  *
277  * Returns true if connection less, false otherwise
278  */
279 static bool tipc_sk_type_connectionless(struct sock *sk)
280 {
281 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
282 }
283 
284 /* tsk_peer_msg - verify if message was sent by connected port's peer
285  *
286  * Handles cases where the node's network address has changed from
287  * the default of <0.0.0> to its configured setting.
288  */
289 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
290 {
291 	struct sock *sk = &tsk->sk;
292 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
293 	u32 peer_port = tsk_peer_port(tsk);
294 	u32 orig_node;
295 	u32 peer_node;
296 
297 	if (unlikely(!tipc_sk_connected(sk)))
298 		return false;
299 
300 	if (unlikely(msg_origport(msg) != peer_port))
301 		return false;
302 
303 	orig_node = msg_orignode(msg);
304 	peer_node = tsk_peer_node(tsk);
305 
306 	if (likely(orig_node == peer_node))
307 		return true;
308 
309 	if (!orig_node && (peer_node == tn->own_addr))
310 		return true;
311 
312 	if (!peer_node && (orig_node == tn->own_addr))
313 		return true;
314 
315 	return false;
316 }
317 
318 /* tipc_set_sk_state - set the sk_state of the socket
319  * @sk: socket
320  *
321  * Caller must hold socket lock
322  *
323  * Returns 0 on success, errno otherwise
324  */
325 static int tipc_set_sk_state(struct sock *sk, int state)
326 {
327 	int oldsk_state = sk->sk_state;
328 	int res = -EINVAL;
329 
330 	switch (state) {
331 	case TIPC_OPEN:
332 		res = 0;
333 		break;
334 	case TIPC_LISTEN:
335 	case TIPC_CONNECTING:
336 		if (oldsk_state == TIPC_OPEN)
337 			res = 0;
338 		break;
339 	case TIPC_ESTABLISHED:
340 		if (oldsk_state == TIPC_CONNECTING ||
341 		    oldsk_state == TIPC_OPEN)
342 			res = 0;
343 		break;
344 	case TIPC_DISCONNECTING:
345 		if (oldsk_state == TIPC_CONNECTING ||
346 		    oldsk_state == TIPC_ESTABLISHED)
347 			res = 0;
348 		break;
349 	}
350 
351 	if (!res)
352 		sk->sk_state = state;
353 
354 	return res;
355 }
356 
357 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
358 {
359 	struct sock *sk = sock->sk;
360 	int err = sock_error(sk);
361 	int typ = sock->type;
362 
363 	if (err)
364 		return err;
365 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
366 		if (sk->sk_state == TIPC_DISCONNECTING)
367 			return -EPIPE;
368 		else if (!tipc_sk_connected(sk))
369 			return -ENOTCONN;
370 	}
371 	if (!*timeout)
372 		return -EAGAIN;
373 	if (signal_pending(current))
374 		return sock_intr_errno(*timeout);
375 
376 	return 0;
377 }
378 
379 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
380 ({                                                                             \
381 	struct sock *sk_;						       \
382 	int rc_;							       \
383 									       \
384 	while ((rc_ = !(condition_))) {					       \
385 		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
386 		sk_ = (sock_)->sk;					       \
387 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
388 		if (rc_)						       \
389 			break;						       \
390 		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
391 		release_sock(sk_);					       \
392 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
393 		sched_annotate_sleep();				               \
394 		lock_sock(sk_);						       \
395 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
396 	}								       \
397 	rc_;								       \
398 })
399 
400 /**
401  * tipc_sk_create - create a TIPC socket
402  * @net: network namespace (must be default network)
403  * @sock: pre-allocated socket structure
404  * @protocol: protocol indicator (must be 0)
405  * @kern: caused by kernel or by userspace?
406  *
407  * This routine creates additional data structures used by the TIPC socket,
408  * initializes them, and links them together.
409  *
410  * Returns 0 on success, errno otherwise
411  */
412 static int tipc_sk_create(struct net *net, struct socket *sock,
413 			  int protocol, int kern)
414 {
415 	struct tipc_net *tn;
416 	const struct proto_ops *ops;
417 	struct sock *sk;
418 	struct tipc_sock *tsk;
419 	struct tipc_msg *msg;
420 
421 	/* Validate arguments */
422 	if (unlikely(protocol != 0))
423 		return -EPROTONOSUPPORT;
424 
425 	switch (sock->type) {
426 	case SOCK_STREAM:
427 		ops = &stream_ops;
428 		break;
429 	case SOCK_SEQPACKET:
430 		ops = &packet_ops;
431 		break;
432 	case SOCK_DGRAM:
433 	case SOCK_RDM:
434 		ops = &msg_ops;
435 		break;
436 	default:
437 		return -EPROTOTYPE;
438 	}
439 
440 	/* Allocate socket's protocol area */
441 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
442 	if (sk == NULL)
443 		return -ENOMEM;
444 
445 	tsk = tipc_sk(sk);
446 	tsk->max_pkt = MAX_PKT_DEFAULT;
447 	INIT_LIST_HEAD(&tsk->publications);
448 	INIT_LIST_HEAD(&tsk->cong_links);
449 	msg = &tsk->phdr;
450 	tn = net_generic(sock_net(sk), tipc_net_id);
451 
452 	/* Finish initializing socket data structures */
453 	sock->ops = ops;
454 	sock_init_data(sock, sk);
455 	tipc_set_sk_state(sk, TIPC_OPEN);
456 	if (tipc_sk_insert(tsk)) {
457 		pr_warn("Socket create failed; port number exhausted\n");
458 		return -EINVAL;
459 	}
460 
461 	/* Ensure tsk is visible before we read own_addr. */
462 	smp_mb();
463 
464 	tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
465 		      NAMED_H_SIZE, 0);
466 
467 	msg_set_origport(msg, tsk->portid);
468 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
469 	sk->sk_shutdown = 0;
470 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
471 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
472 	sk->sk_data_ready = tipc_data_ready;
473 	sk->sk_write_space = tipc_write_space;
474 	sk->sk_destruct = tipc_sock_destruct;
475 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
476 	tsk->group_is_open = true;
477 	atomic_set(&tsk->dupl_rcvcnt, 0);
478 
479 	/* Start out with safe limits until we receive an advertised window */
480 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
481 	tsk->rcv_win = tsk->snd_win;
482 
483 	if (tipc_sk_type_connectionless(sk)) {
484 		tsk_set_unreturnable(tsk, true);
485 		if (sock->type == SOCK_DGRAM)
486 			tsk_set_unreliable(tsk, true);
487 	}
488 
489 	return 0;
490 }
491 
492 static void tipc_sk_callback(struct rcu_head *head)
493 {
494 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
495 
496 	sock_put(&tsk->sk);
497 }
498 
499 /* Caller should hold socket lock for the socket. */
500 static void __tipc_shutdown(struct socket *sock, int error)
501 {
502 	struct sock *sk = sock->sk;
503 	struct tipc_sock *tsk = tipc_sk(sk);
504 	struct net *net = sock_net(sk);
505 	long timeout = CONN_TIMEOUT_DEFAULT;
506 	u32 dnode = tsk_peer_node(tsk);
507 	struct sk_buff *skb;
508 
509 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
510 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
511 					    !tsk_conn_cong(tsk)));
512 
513 	/* Reject all unreceived messages, except on an active connection
514 	 * (which disconnects locally & sends a 'FIN+' to peer).
515 	 */
516 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
517 		if (TIPC_SKB_CB(skb)->bytes_read) {
518 			kfree_skb(skb);
519 			continue;
520 		}
521 		if (!tipc_sk_type_connectionless(sk) &&
522 		    sk->sk_state != TIPC_DISCONNECTING) {
523 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
524 			tipc_node_remove_conn(net, dnode, tsk->portid);
525 		}
526 		tipc_sk_respond(sk, skb, error);
527 	}
528 
529 	if (tipc_sk_type_connectionless(sk))
530 		return;
531 
532 	if (sk->sk_state != TIPC_DISCONNECTING) {
533 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
534 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
535 				      tsk_own_node(tsk), tsk_peer_port(tsk),
536 				      tsk->portid, error);
537 		if (skb)
538 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
539 		tipc_node_remove_conn(net, dnode, tsk->portid);
540 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
541 	}
542 }
543 
544 /**
545  * tipc_release - destroy a TIPC socket
546  * @sock: socket to destroy
547  *
548  * This routine cleans up any messages that are still queued on the socket.
549  * For DGRAM and RDM socket types, all queued messages are rejected.
550  * For SEQPACKET and STREAM socket types, the first message is rejected
551  * and any others are discarded.  (If the first message on a STREAM socket
552  * is partially-read, it is discarded and the next one is rejected instead.)
553  *
554  * NOTE: Rejected messages are not necessarily returned to the sender!  They
555  * are returned or discarded according to the "destination droppable" setting
556  * specified for the message by the sender.
557  *
558  * Returns 0 on success, errno otherwise
559  */
560 static int tipc_release(struct socket *sock)
561 {
562 	struct sock *sk = sock->sk;
563 	struct tipc_sock *tsk;
564 
565 	/*
566 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
567 	 * releases a pre-allocated child socket that was never used)
568 	 */
569 	if (sk == NULL)
570 		return 0;
571 
572 	tsk = tipc_sk(sk);
573 	lock_sock(sk);
574 
575 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
576 	sk->sk_shutdown = SHUTDOWN_MASK;
577 	tipc_sk_leave(tsk);
578 	tipc_sk_withdraw(tsk, 0, NULL);
579 	sk_stop_timer(sk, &sk->sk_timer);
580 	tipc_sk_remove(tsk);
581 
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 *uaddr_len, 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 	struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
675 
676 	memset(addr, 0, sizeof(*addr));
677 	if (peer) {
678 		if ((!tipc_sk_connected(sk)) &&
679 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
680 			return -ENOTCONN;
681 		addr->addr.id.ref = tsk_peer_port(tsk);
682 		addr->addr.id.node = tsk_peer_node(tsk);
683 	} else {
684 		addr->addr.id.ref = tsk->portid;
685 		addr->addr.id.node = tn->own_addr;
686 	}
687 
688 	*uaddr_len = sizeof(*addr);
689 	addr->addrtype = TIPC_ADDR_ID;
690 	addr->family = AF_TIPC;
691 	addr->scope = 0;
692 	addr->addr.name.domain = 0;
693 
694 	return 0;
695 }
696 
697 /**
698  * tipc_poll - read and possibly block on pollmask
699  * @file: file structure associated with the socket
700  * @sock: socket for which to calculate the poll bits
701  * @wait: ???
702  *
703  * Returns pollmask value
704  *
705  * COMMENTARY:
706  * It appears that the usual socket locking mechanisms are not useful here
707  * since the pollmask info is potentially out-of-date the moment this routine
708  * exits.  TCP and other protocols seem to rely on higher level poll routines
709  * to handle any preventable race conditions, so TIPC will do the same ...
710  *
711  * IMPORTANT: The fact that a read or write operation is indicated does NOT
712  * imply that the operation will succeed, merely that it should be performed
713  * and will not block.
714  */
715 static __poll_t tipc_poll(struct file *file, struct socket *sock,
716 			      poll_table *wait)
717 {
718 	struct sock *sk = sock->sk;
719 	struct tipc_sock *tsk = tipc_sk(sk);
720 	__poll_t revents = 0;
721 
722 	sock_poll_wait(file, sk_sleep(sk), wait);
723 
724 	if (sk->sk_shutdown & RCV_SHUTDOWN)
725 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
726 	if (sk->sk_shutdown == SHUTDOWN_MASK)
727 		revents |= EPOLLHUP;
728 
729 	switch (sk->sk_state) {
730 	case TIPC_ESTABLISHED:
731 	case TIPC_CONNECTING:
732 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
733 			revents |= EPOLLOUT;
734 		/* fall thru' */
735 	case TIPC_LISTEN:
736 		if (!skb_queue_empty(&sk->sk_receive_queue))
737 			revents |= EPOLLIN | EPOLLRDNORM;
738 		break;
739 	case TIPC_OPEN:
740 		if (tsk->group_is_open && !tsk->cong_link_cnt)
741 			revents |= EPOLLOUT;
742 		if (!tipc_sk_type_connectionless(sk))
743 			break;
744 		if (skb_queue_empty(&sk->sk_receive_queue))
745 			break;
746 		revents |= EPOLLIN | EPOLLRDNORM;
747 		break;
748 	case TIPC_DISCONNECTING:
749 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
750 		break;
751 	}
752 	return revents;
753 }
754 
755 /**
756  * tipc_sendmcast - send multicast message
757  * @sock: socket structure
758  * @seq: destination address
759  * @msg: message to send
760  * @dlen: length of data to send
761  * @timeout: timeout to wait for wakeup
762  *
763  * Called from function tipc_sendmsg(), which has done all sanity checks
764  * Returns the number of bytes sent on success, or errno
765  */
766 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
767 			  struct msghdr *msg, size_t dlen, long timeout)
768 {
769 	struct sock *sk = sock->sk;
770 	struct tipc_sock *tsk = tipc_sk(sk);
771 	struct tipc_msg *hdr = &tsk->phdr;
772 	struct net *net = sock_net(sk);
773 	int mtu = tipc_bcast_get_mtu(net);
774 	struct tipc_mc_method *method = &tsk->mc_method;
775 	struct sk_buff_head pkts;
776 	struct tipc_nlist dsts;
777 	int rc;
778 
779 	if (tsk->group)
780 		return -EACCES;
781 
782 	/* Block or return if any destination link is congested */
783 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
784 	if (unlikely(rc))
785 		return rc;
786 
787 	/* Lookup destination nodes */
788 	tipc_nlist_init(&dsts, tipc_own_addr(net));
789 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
790 				      seq->upper, &dsts);
791 	if (!dsts.local && !dsts.remote)
792 		return -EHOSTUNREACH;
793 
794 	/* Build message header */
795 	msg_set_type(hdr, TIPC_MCAST_MSG);
796 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
797 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
798 	msg_set_destport(hdr, 0);
799 	msg_set_destnode(hdr, 0);
800 	msg_set_nametype(hdr, seq->type);
801 	msg_set_namelower(hdr, seq->lower);
802 	msg_set_nameupper(hdr, seq->upper);
803 
804 	/* Build message as chain of buffers */
805 	skb_queue_head_init(&pkts);
806 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
807 
808 	/* Send message if build was successful */
809 	if (unlikely(rc == dlen))
810 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
811 				     &tsk->cong_link_cnt);
812 
813 	tipc_nlist_purge(&dsts);
814 
815 	return rc ? rc : dlen;
816 }
817 
818 /**
819  * tipc_send_group_msg - send a message to a member in the group
820  * @net: network namespace
821  * @m: message to send
822  * @mb: group member
823  * @dnode: destination node
824  * @dport: destination port
825  * @dlen: total length of message data
826  */
827 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
828 			       struct msghdr *m, struct tipc_member *mb,
829 			       u32 dnode, u32 dport, int dlen)
830 {
831 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
832 	struct tipc_mc_method *method = &tsk->mc_method;
833 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
834 	struct tipc_msg *hdr = &tsk->phdr;
835 	struct sk_buff_head pkts;
836 	int mtu, rc;
837 
838 	/* Complete message header */
839 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
840 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
841 	msg_set_destport(hdr, dport);
842 	msg_set_destnode(hdr, dnode);
843 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
844 
845 	/* Build message as chain of buffers */
846 	skb_queue_head_init(&pkts);
847 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
848 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
849 	if (unlikely(rc != dlen))
850 		return rc;
851 
852 	/* Send message */
853 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
854 	if (unlikely(rc == -ELINKCONG)) {
855 		tipc_dest_push(&tsk->cong_links, dnode, 0);
856 		tsk->cong_link_cnt++;
857 	}
858 
859 	/* Update send window */
860 	tipc_group_update_member(mb, blks);
861 
862 	/* A broadcast sent within next EXPIRE period must follow same path */
863 	method->rcast = true;
864 	method->mandatory = true;
865 	return dlen;
866 }
867 
868 /**
869  * tipc_send_group_unicast - send message to a member in the group
870  * @sock: socket structure
871  * @m: message to send
872  * @dlen: total length of message data
873  * @timeout: timeout to wait for wakeup
874  *
875  * Called from function tipc_sendmsg(), which has done all sanity checks
876  * Returns the number of bytes sent on success, or errno
877  */
878 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
879 				   int dlen, long timeout)
880 {
881 	struct sock *sk = sock->sk;
882 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
883 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
884 	struct tipc_sock *tsk = tipc_sk(sk);
885 	struct tipc_group *grp = tsk->group;
886 	struct net *net = sock_net(sk);
887 	struct tipc_member *mb = NULL;
888 	u32 node, port;
889 	int rc;
890 
891 	node = dest->addr.id.node;
892 	port = dest->addr.id.ref;
893 	if (!port && !node)
894 		return -EHOSTUNREACH;
895 
896 	/* Block or return if destination link or member is congested */
897 	rc = tipc_wait_for_cond(sock, &timeout,
898 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
899 				!tipc_group_cong(grp, node, port, blks, &mb));
900 	if (unlikely(rc))
901 		return rc;
902 
903 	if (unlikely(!mb))
904 		return -EHOSTUNREACH;
905 
906 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
907 
908 	return rc ? rc : dlen;
909 }
910 
911 /**
912  * tipc_send_group_anycast - send message to any member with given identity
913  * @sock: socket structure
914  * @m: message to send
915  * @dlen: total length of message data
916  * @timeout: timeout to wait for wakeup
917  *
918  * Called from function tipc_sendmsg(), which has done all sanity checks
919  * Returns the number of bytes sent on success, or errno
920  */
921 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
922 				   int dlen, long timeout)
923 {
924 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
925 	struct sock *sk = sock->sk;
926 	struct tipc_sock *tsk = tipc_sk(sk);
927 	struct list_head *cong_links = &tsk->cong_links;
928 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
929 	struct tipc_group *grp = tsk->group;
930 	struct tipc_msg *hdr = &tsk->phdr;
931 	struct tipc_member *first = NULL;
932 	struct tipc_member *mbr = NULL;
933 	struct net *net = sock_net(sk);
934 	u32 node, port, exclude;
935 	struct list_head dsts;
936 	u32 type, inst, scope;
937 	int lookups = 0;
938 	int dstcnt, rc;
939 	bool cong;
940 
941 	INIT_LIST_HEAD(&dsts);
942 
943 	type = msg_nametype(hdr);
944 	inst = dest->addr.name.name.instance;
945 	scope = msg_lookup_scope(hdr);
946 	exclude = tipc_group_exclude(grp);
947 
948 	while (++lookups < 4) {
949 		first = NULL;
950 
951 		/* Look for a non-congested destination member, if any */
952 		while (1) {
953 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
954 						 &dstcnt, exclude, false))
955 				return -EHOSTUNREACH;
956 			tipc_dest_pop(&dsts, &node, &port);
957 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
958 			if (!cong)
959 				break;
960 			if (mbr == first)
961 				break;
962 			if (!first)
963 				first = mbr;
964 		}
965 
966 		/* Start over if destination was not in member list */
967 		if (unlikely(!mbr))
968 			continue;
969 
970 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
971 			break;
972 
973 		/* Block or return if destination link or member is congested */
974 		rc = tipc_wait_for_cond(sock, &timeout,
975 					!tipc_dest_find(cong_links, node, 0) &&
976 					!tipc_group_cong(grp, node, port,
977 							 blks, &mbr));
978 		if (unlikely(rc))
979 			return rc;
980 
981 		/* Send, unless destination disappeared while waiting */
982 		if (likely(mbr))
983 			break;
984 	}
985 
986 	if (unlikely(lookups >= 4))
987 		return -EHOSTUNREACH;
988 
989 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
990 
991 	return rc ? rc : dlen;
992 }
993 
994 /**
995  * tipc_send_group_bcast - send message to all members in communication group
996  * @sk: socket structure
997  * @m: message to send
998  * @dlen: total length of message data
999  * @timeout: timeout to wait for wakeup
1000  *
1001  * Called from function tipc_sendmsg(), which has done all sanity checks
1002  * Returns the number of bytes sent on success, or errno
1003  */
1004 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1005 				 int dlen, long timeout)
1006 {
1007 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1008 	struct sock *sk = sock->sk;
1009 	struct net *net = sock_net(sk);
1010 	struct tipc_sock *tsk = tipc_sk(sk);
1011 	struct tipc_group *grp = tsk->group;
1012 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1013 	struct tipc_mc_method *method = &tsk->mc_method;
1014 	bool ack = method->mandatory && method->rcast;
1015 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1016 	struct tipc_msg *hdr = &tsk->phdr;
1017 	int mtu = tipc_bcast_get_mtu(net);
1018 	struct sk_buff_head pkts;
1019 	int rc = -EHOSTUNREACH;
1020 
1021 	if (!dsts->local && !dsts->remote)
1022 		return -EHOSTUNREACH;
1023 
1024 	/* Block or return if any destination link or member is congested */
1025 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1026 				!tipc_group_bc_cong(grp, blks));
1027 	if (unlikely(rc))
1028 		return rc;
1029 
1030 	/* Complete message header */
1031 	if (dest) {
1032 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1033 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1034 	} else {
1035 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1036 		msg_set_nameinst(hdr, 0);
1037 	}
1038 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1039 	msg_set_destport(hdr, 0);
1040 	msg_set_destnode(hdr, 0);
1041 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1042 
1043 	/* Avoid getting stuck with repeated forced replicasts */
1044 	msg_set_grp_bc_ack_req(hdr, ack);
1045 
1046 	/* Build message as chain of buffers */
1047 	skb_queue_head_init(&pkts);
1048 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1049 	if (unlikely(rc != dlen))
1050 		return rc;
1051 
1052 	/* Send message */
1053 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1054 	if (unlikely(rc))
1055 		return rc;
1056 
1057 	/* Update broadcast sequence number and send windows */
1058 	tipc_group_update_bc_members(tsk->group, blks, ack);
1059 
1060 	/* Broadcast link is now free to choose method for next broadcast */
1061 	method->mandatory = false;
1062 	method->expires = jiffies;
1063 
1064 	return dlen;
1065 }
1066 
1067 /**
1068  * tipc_send_group_mcast - send message to all members with given identity
1069  * @sock: socket structure
1070  * @m: message to send
1071  * @dlen: total length of message data
1072  * @timeout: timeout to wait for wakeup
1073  *
1074  * Called from function tipc_sendmsg(), which has done all sanity checks
1075  * Returns the number of bytes sent on success, or errno
1076  */
1077 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1078 				 int dlen, long timeout)
1079 {
1080 	struct sock *sk = sock->sk;
1081 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1082 	struct tipc_sock *tsk = tipc_sk(sk);
1083 	struct tipc_group *grp = tsk->group;
1084 	struct tipc_msg *hdr = &tsk->phdr;
1085 	struct net *net = sock_net(sk);
1086 	u32 type, inst, scope, exclude;
1087 	struct list_head dsts;
1088 	u32 dstcnt;
1089 
1090 	INIT_LIST_HEAD(&dsts);
1091 
1092 	type = msg_nametype(hdr);
1093 	inst = dest->addr.name.name.instance;
1094 	scope = msg_lookup_scope(hdr);
1095 	exclude = tipc_group_exclude(grp);
1096 
1097 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1098 				 &dstcnt, exclude, true))
1099 		return -EHOSTUNREACH;
1100 
1101 	if (dstcnt == 1) {
1102 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1103 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1104 	}
1105 
1106 	tipc_dest_list_purge(&dsts);
1107 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1108 }
1109 
1110 /**
1111  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1112  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1113  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1114  *
1115  * Multi-threaded: parallel calls with reference to same queues may occur
1116  */
1117 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1118 		       struct sk_buff_head *inputq)
1119 {
1120 	u32 self = tipc_own_addr(net);
1121 	u32 type, lower, upper, scope;
1122 	struct sk_buff *skb, *_skb;
1123 	u32 portid, oport, onode;
1124 	struct sk_buff_head tmpq;
1125 	struct list_head dports;
1126 	struct tipc_msg *hdr;
1127 	int user, mtyp, hlen;
1128 	bool exact;
1129 
1130 	__skb_queue_head_init(&tmpq);
1131 	INIT_LIST_HEAD(&dports);
1132 
1133 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1134 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1135 		hdr = buf_msg(skb);
1136 		user = msg_user(hdr);
1137 		mtyp = msg_type(hdr);
1138 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1139 		oport = msg_origport(hdr);
1140 		onode = msg_orignode(hdr);
1141 		type = msg_nametype(hdr);
1142 
1143 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1144 			spin_lock_bh(&inputq->lock);
1145 			if (skb_peek(arrvq) == skb) {
1146 				__skb_dequeue(arrvq);
1147 				__skb_queue_tail(inputq, skb);
1148 			}
1149 			kfree_skb(skb);
1150 			spin_unlock_bh(&inputq->lock);
1151 			continue;
1152 		}
1153 
1154 		/* Group messages require exact scope match */
1155 		if (msg_in_group(hdr)) {
1156 			lower = 0;
1157 			upper = ~0;
1158 			scope = msg_lookup_scope(hdr);
1159 			exact = true;
1160 		} else {
1161 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1162 			if (onode == self)
1163 				scope = TIPC_NODE_SCOPE;
1164 			else
1165 				scope = TIPC_CLUSTER_SCOPE;
1166 			exact = false;
1167 			lower = msg_namelower(hdr);
1168 			upper = msg_nameupper(hdr);
1169 		}
1170 
1171 		/* Create destination port list: */
1172 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1173 				       scope, exact, &dports);
1174 
1175 		/* Clone message per destination */
1176 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1177 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1178 			if (_skb) {
1179 				msg_set_destport(buf_msg(_skb), portid);
1180 				__skb_queue_tail(&tmpq, _skb);
1181 				continue;
1182 			}
1183 			pr_warn("Failed to clone mcast rcv buffer\n");
1184 		}
1185 		/* Append to inputq if not already done by other thread */
1186 		spin_lock_bh(&inputq->lock);
1187 		if (skb_peek(arrvq) == skb) {
1188 			skb_queue_splice_tail_init(&tmpq, inputq);
1189 			kfree_skb(__skb_dequeue(arrvq));
1190 		}
1191 		spin_unlock_bh(&inputq->lock);
1192 		__skb_queue_purge(&tmpq);
1193 		kfree_skb(skb);
1194 	}
1195 	tipc_sk_rcv(net, inputq);
1196 }
1197 
1198 /**
1199  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1200  * @tsk: receiving socket
1201  * @skb: pointer to message buffer.
1202  */
1203 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1204 				   struct sk_buff_head *xmitq)
1205 {
1206 	struct tipc_msg *hdr = buf_msg(skb);
1207 	u32 onode = tsk_own_node(tsk);
1208 	struct sock *sk = &tsk->sk;
1209 	int mtyp = msg_type(hdr);
1210 	bool conn_cong;
1211 
1212 	/* Ignore if connection cannot be validated: */
1213 	if (!tsk_peer_msg(tsk, hdr))
1214 		goto exit;
1215 
1216 	if (unlikely(msg_errcode(hdr))) {
1217 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1218 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1219 				      tsk_peer_port(tsk));
1220 		sk->sk_state_change(sk);
1221 		goto exit;
1222 	}
1223 
1224 	tsk->probe_unacked = false;
1225 
1226 	if (mtyp == CONN_PROBE) {
1227 		msg_set_type(hdr, CONN_PROBE_REPLY);
1228 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1229 			__skb_queue_tail(xmitq, skb);
1230 		return;
1231 	} else if (mtyp == CONN_ACK) {
1232 		conn_cong = tsk_conn_cong(tsk);
1233 		tsk->snt_unacked -= msg_conn_ack(hdr);
1234 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1235 			tsk->snd_win = msg_adv_win(hdr);
1236 		if (conn_cong)
1237 			sk->sk_write_space(sk);
1238 	} else if (mtyp != CONN_PROBE_REPLY) {
1239 		pr_warn("Received unknown CONN_PROTO msg\n");
1240 	}
1241 exit:
1242 	kfree_skb(skb);
1243 }
1244 
1245 /**
1246  * tipc_sendmsg - send message in connectionless manner
1247  * @sock: socket structure
1248  * @m: message to send
1249  * @dsz: amount of user data to be sent
1250  *
1251  * Message must have an destination specified explicitly.
1252  * Used for SOCK_RDM and SOCK_DGRAM messages,
1253  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1254  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1255  *
1256  * Returns the number of bytes sent on success, or errno otherwise
1257  */
1258 static int tipc_sendmsg(struct socket *sock,
1259 			struct msghdr *m, size_t dsz)
1260 {
1261 	struct sock *sk = sock->sk;
1262 	int ret;
1263 
1264 	lock_sock(sk);
1265 	ret = __tipc_sendmsg(sock, m, dsz);
1266 	release_sock(sk);
1267 
1268 	return ret;
1269 }
1270 
1271 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1272 {
1273 	struct sock *sk = sock->sk;
1274 	struct net *net = sock_net(sk);
1275 	struct tipc_sock *tsk = tipc_sk(sk);
1276 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1277 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1278 	struct list_head *clinks = &tsk->cong_links;
1279 	bool syn = !tipc_sk_type_connectionless(sk);
1280 	struct tipc_group *grp = tsk->group;
1281 	struct tipc_msg *hdr = &tsk->phdr;
1282 	struct tipc_name_seq *seq;
1283 	struct sk_buff_head pkts;
1284 	u32 type, inst, domain;
1285 	u32 dnode, dport;
1286 	int mtu, rc;
1287 
1288 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1289 		return -EMSGSIZE;
1290 
1291 	if (likely(dest)) {
1292 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1293 			return -EINVAL;
1294 		if (unlikely(dest->family != AF_TIPC))
1295 			return -EINVAL;
1296 	}
1297 
1298 	if (grp) {
1299 		if (!dest)
1300 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1301 		if (dest->addrtype == TIPC_ADDR_NAME)
1302 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1303 		if (dest->addrtype == TIPC_ADDR_ID)
1304 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1305 		if (dest->addrtype == TIPC_ADDR_MCAST)
1306 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1307 		return -EINVAL;
1308 	}
1309 
1310 	if (unlikely(!dest)) {
1311 		dest = &tsk->peer;
1312 		if (!syn || dest->family != AF_TIPC)
1313 			return -EDESTADDRREQ;
1314 	}
1315 
1316 	if (unlikely(syn)) {
1317 		if (sk->sk_state == TIPC_LISTEN)
1318 			return -EPIPE;
1319 		if (sk->sk_state != TIPC_OPEN)
1320 			return -EISCONN;
1321 		if (tsk->published)
1322 			return -EOPNOTSUPP;
1323 		if (dest->addrtype == TIPC_ADDR_NAME) {
1324 			tsk->conn_type = dest->addr.name.name.type;
1325 			tsk->conn_instance = dest->addr.name.name.instance;
1326 		}
1327 	}
1328 
1329 	seq = &dest->addr.nameseq;
1330 	if (dest->addrtype == TIPC_ADDR_MCAST)
1331 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1332 
1333 	if (dest->addrtype == TIPC_ADDR_NAME) {
1334 		type = dest->addr.name.name.type;
1335 		inst = dest->addr.name.name.instance;
1336 		domain = dest->addr.name.domain;
1337 		dnode = domain;
1338 		msg_set_type(hdr, TIPC_NAMED_MSG);
1339 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1340 		msg_set_nametype(hdr, type);
1341 		msg_set_nameinst(hdr, inst);
1342 		msg_set_lookup_scope(hdr, tipc_addr_scope(domain));
1343 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1344 		msg_set_destnode(hdr, dnode);
1345 		msg_set_destport(hdr, dport);
1346 		if (unlikely(!dport && !dnode))
1347 			return -EHOSTUNREACH;
1348 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1349 		dnode = dest->addr.id.node;
1350 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1351 		msg_set_lookup_scope(hdr, 0);
1352 		msg_set_destnode(hdr, dnode);
1353 		msg_set_destport(hdr, dest->addr.id.ref);
1354 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1355 	}
1356 
1357 	/* Block or return if destination link is congested */
1358 	rc = tipc_wait_for_cond(sock, &timeout,
1359 				!tipc_dest_find(clinks, dnode, 0));
1360 	if (unlikely(rc))
1361 		return rc;
1362 
1363 	skb_queue_head_init(&pkts);
1364 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1365 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1366 	if (unlikely(rc != dlen))
1367 		return rc;
1368 
1369 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1370 	if (unlikely(rc == -ELINKCONG)) {
1371 		tipc_dest_push(clinks, dnode, 0);
1372 		tsk->cong_link_cnt++;
1373 		rc = 0;
1374 	}
1375 
1376 	if (unlikely(syn && !rc))
1377 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1378 
1379 	return rc ? rc : dlen;
1380 }
1381 
1382 /**
1383  * tipc_sendstream - send stream-oriented data
1384  * @sock: socket structure
1385  * @m: data to send
1386  * @dsz: total length of data to be transmitted
1387  *
1388  * Used for SOCK_STREAM data.
1389  *
1390  * Returns the number of bytes sent on success (or partial success),
1391  * or errno if no data sent
1392  */
1393 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1394 {
1395 	struct sock *sk = sock->sk;
1396 	int ret;
1397 
1398 	lock_sock(sk);
1399 	ret = __tipc_sendstream(sock, m, dsz);
1400 	release_sock(sk);
1401 
1402 	return ret;
1403 }
1404 
1405 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1406 {
1407 	struct sock *sk = sock->sk;
1408 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1409 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1410 	struct tipc_sock *tsk = tipc_sk(sk);
1411 	struct tipc_msg *hdr = &tsk->phdr;
1412 	struct net *net = sock_net(sk);
1413 	struct sk_buff_head pkts;
1414 	u32 dnode = tsk_peer_node(tsk);
1415 	int send, sent = 0;
1416 	int rc = 0;
1417 
1418 	skb_queue_head_init(&pkts);
1419 
1420 	if (unlikely(dlen > INT_MAX))
1421 		return -EMSGSIZE;
1422 
1423 	/* Handle implicit connection setup */
1424 	if (unlikely(dest)) {
1425 		rc = __tipc_sendmsg(sock, m, dlen);
1426 		if (dlen && (dlen == rc))
1427 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1428 		return rc;
1429 	}
1430 
1431 	do {
1432 		rc = tipc_wait_for_cond(sock, &timeout,
1433 					(!tsk->cong_link_cnt &&
1434 					 !tsk_conn_cong(tsk) &&
1435 					 tipc_sk_connected(sk)));
1436 		if (unlikely(rc))
1437 			break;
1438 
1439 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1440 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1441 		if (unlikely(rc != send))
1442 			break;
1443 
1444 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1445 		if (unlikely(rc == -ELINKCONG)) {
1446 			tsk->cong_link_cnt = 1;
1447 			rc = 0;
1448 		}
1449 		if (likely(!rc)) {
1450 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1451 			sent += send;
1452 		}
1453 	} while (sent < dlen && !rc);
1454 
1455 	return sent ? sent : rc;
1456 }
1457 
1458 /**
1459  * tipc_send_packet - send a connection-oriented message
1460  * @sock: socket structure
1461  * @m: message to send
1462  * @dsz: length of data to be transmitted
1463  *
1464  * Used for SOCK_SEQPACKET messages.
1465  *
1466  * Returns the number of bytes sent on success, or errno otherwise
1467  */
1468 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1469 {
1470 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1471 		return -EMSGSIZE;
1472 
1473 	return tipc_sendstream(sock, m, dsz);
1474 }
1475 
1476 /* tipc_sk_finish_conn - complete the setup of a connection
1477  */
1478 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1479 				u32 peer_node)
1480 {
1481 	struct sock *sk = &tsk->sk;
1482 	struct net *net = sock_net(sk);
1483 	struct tipc_msg *msg = &tsk->phdr;
1484 
1485 	msg_set_destnode(msg, peer_node);
1486 	msg_set_destport(msg, peer_port);
1487 	msg_set_type(msg, TIPC_CONN_MSG);
1488 	msg_set_lookup_scope(msg, 0);
1489 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1490 
1491 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1492 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1493 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1494 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1495 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1496 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1497 		return;
1498 
1499 	/* Fall back to message based flow control */
1500 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1501 	tsk->snd_win = FLOWCTL_MSG_WIN;
1502 }
1503 
1504 /**
1505  * tipc_sk_set_orig_addr - capture sender's address for received message
1506  * @m: descriptor for message info
1507  * @hdr: received message header
1508  *
1509  * Note: Address is not captured if not requested by receiver.
1510  */
1511 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1512 {
1513 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1514 	struct tipc_msg *hdr = buf_msg(skb);
1515 
1516 	if (!srcaddr)
1517 		return;
1518 
1519 	srcaddr->sock.family = AF_TIPC;
1520 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1521 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1522 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1523 	srcaddr->sock.addr.name.domain = 0;
1524 	srcaddr->sock.scope = 0;
1525 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1526 
1527 	if (!msg_in_group(hdr))
1528 		return;
1529 
1530 	/* Group message users may also want to know sending member's id */
1531 	srcaddr->member.family = AF_TIPC;
1532 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1533 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1534 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1535 	srcaddr->member.addr.name.domain = 0;
1536 	m->msg_namelen = sizeof(*srcaddr);
1537 }
1538 
1539 /**
1540  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1541  * @m: descriptor for message info
1542  * @msg: received message header
1543  * @tsk: TIPC port associated with message
1544  *
1545  * Note: Ancillary data is not captured if not requested by receiver.
1546  *
1547  * Returns 0 if successful, otherwise errno
1548  */
1549 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1550 				 struct tipc_sock *tsk)
1551 {
1552 	u32 anc_data[3];
1553 	u32 err;
1554 	u32 dest_type;
1555 	int has_name;
1556 	int res;
1557 
1558 	if (likely(m->msg_controllen == 0))
1559 		return 0;
1560 
1561 	/* Optionally capture errored message object(s) */
1562 	err = msg ? msg_errcode(msg) : 0;
1563 	if (unlikely(err)) {
1564 		anc_data[0] = err;
1565 		anc_data[1] = msg_data_sz(msg);
1566 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1567 		if (res)
1568 			return res;
1569 		if (anc_data[1]) {
1570 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1571 				       msg_data(msg));
1572 			if (res)
1573 				return res;
1574 		}
1575 	}
1576 
1577 	/* Optionally capture message destination object */
1578 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1579 	switch (dest_type) {
1580 	case TIPC_NAMED_MSG:
1581 		has_name = 1;
1582 		anc_data[0] = msg_nametype(msg);
1583 		anc_data[1] = msg_namelower(msg);
1584 		anc_data[2] = msg_namelower(msg);
1585 		break;
1586 	case TIPC_MCAST_MSG:
1587 		has_name = 1;
1588 		anc_data[0] = msg_nametype(msg);
1589 		anc_data[1] = msg_namelower(msg);
1590 		anc_data[2] = msg_nameupper(msg);
1591 		break;
1592 	case TIPC_CONN_MSG:
1593 		has_name = (tsk->conn_type != 0);
1594 		anc_data[0] = tsk->conn_type;
1595 		anc_data[1] = tsk->conn_instance;
1596 		anc_data[2] = tsk->conn_instance;
1597 		break;
1598 	default:
1599 		has_name = 0;
1600 	}
1601 	if (has_name) {
1602 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1603 		if (res)
1604 			return res;
1605 	}
1606 
1607 	return 0;
1608 }
1609 
1610 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1611 {
1612 	struct sock *sk = &tsk->sk;
1613 	struct net *net = sock_net(sk);
1614 	struct sk_buff *skb = NULL;
1615 	struct tipc_msg *msg;
1616 	u32 peer_port = tsk_peer_port(tsk);
1617 	u32 dnode = tsk_peer_node(tsk);
1618 
1619 	if (!tipc_sk_connected(sk))
1620 		return;
1621 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1622 			      dnode, tsk_own_node(tsk), peer_port,
1623 			      tsk->portid, TIPC_OK);
1624 	if (!skb)
1625 		return;
1626 	msg = buf_msg(skb);
1627 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1628 	tsk->rcv_unacked = 0;
1629 
1630 	/* Adjust to and advertize the correct window limit */
1631 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1632 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1633 		msg_set_adv_win(msg, tsk->rcv_win);
1634 	}
1635 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1636 }
1637 
1638 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1639 {
1640 	struct sock *sk = sock->sk;
1641 	DEFINE_WAIT(wait);
1642 	long timeo = *timeop;
1643 	int err = sock_error(sk);
1644 
1645 	if (err)
1646 		return err;
1647 
1648 	for (;;) {
1649 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1650 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1651 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1652 				err = -ENOTCONN;
1653 				break;
1654 			}
1655 			release_sock(sk);
1656 			timeo = schedule_timeout(timeo);
1657 			lock_sock(sk);
1658 		}
1659 		err = 0;
1660 		if (!skb_queue_empty(&sk->sk_receive_queue))
1661 			break;
1662 		err = -EAGAIN;
1663 		if (!timeo)
1664 			break;
1665 		err = sock_intr_errno(timeo);
1666 		if (signal_pending(current))
1667 			break;
1668 
1669 		err = sock_error(sk);
1670 		if (err)
1671 			break;
1672 	}
1673 	finish_wait(sk_sleep(sk), &wait);
1674 	*timeop = timeo;
1675 	return err;
1676 }
1677 
1678 /**
1679  * tipc_recvmsg - receive packet-oriented message
1680  * @m: descriptor for message info
1681  * @buflen: length of user buffer area
1682  * @flags: receive flags
1683  *
1684  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1685  * If the complete message doesn't fit in user area, truncate it.
1686  *
1687  * Returns size of returned message data, errno otherwise
1688  */
1689 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1690 			size_t buflen,	int flags)
1691 {
1692 	struct sock *sk = sock->sk;
1693 	bool connected = !tipc_sk_type_connectionless(sk);
1694 	struct tipc_sock *tsk = tipc_sk(sk);
1695 	int rc, err, hlen, dlen, copy;
1696 	struct sk_buff_head xmitq;
1697 	struct tipc_msg *hdr;
1698 	struct sk_buff *skb;
1699 	bool grp_evt;
1700 	long timeout;
1701 
1702 	/* Catch invalid receive requests */
1703 	if (unlikely(!buflen))
1704 		return -EINVAL;
1705 
1706 	lock_sock(sk);
1707 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1708 		rc = -ENOTCONN;
1709 		goto exit;
1710 	}
1711 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1712 
1713 	/* Step rcv queue to first msg with data or error; wait if necessary */
1714 	do {
1715 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1716 		if (unlikely(rc))
1717 			goto exit;
1718 		skb = skb_peek(&sk->sk_receive_queue);
1719 		hdr = buf_msg(skb);
1720 		dlen = msg_data_sz(hdr);
1721 		hlen = msg_hdr_sz(hdr);
1722 		err = msg_errcode(hdr);
1723 		grp_evt = msg_is_grp_evt(hdr);
1724 		if (likely(dlen || err))
1725 			break;
1726 		tsk_advance_rx_queue(sk);
1727 	} while (1);
1728 
1729 	/* Collect msg meta data, including error code and rejected data */
1730 	tipc_sk_set_orig_addr(m, skb);
1731 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1732 	if (unlikely(rc))
1733 		goto exit;
1734 
1735 	/* Capture data if non-error msg, otherwise just set return value */
1736 	if (likely(!err)) {
1737 		copy = min_t(int, dlen, buflen);
1738 		if (unlikely(copy != dlen))
1739 			m->msg_flags |= MSG_TRUNC;
1740 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1741 	} else {
1742 		copy = 0;
1743 		rc = 0;
1744 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1745 			rc = -ECONNRESET;
1746 	}
1747 	if (unlikely(rc))
1748 		goto exit;
1749 
1750 	/* Mark message as group event if applicable */
1751 	if (unlikely(grp_evt)) {
1752 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1753 			m->msg_flags |= MSG_EOR;
1754 		m->msg_flags |= MSG_OOB;
1755 		copy = 0;
1756 	}
1757 
1758 	/* Caption of data or error code/rejected data was successful */
1759 	if (unlikely(flags & MSG_PEEK))
1760 		goto exit;
1761 
1762 	/* Send group flow control advertisement when applicable */
1763 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1764 		skb_queue_head_init(&xmitq);
1765 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1766 					  msg_orignode(hdr), msg_origport(hdr),
1767 					  &xmitq);
1768 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1769 	}
1770 
1771 	tsk_advance_rx_queue(sk);
1772 
1773 	if (likely(!connected))
1774 		goto exit;
1775 
1776 	/* Send connection flow control advertisement when applicable */
1777 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1778 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1779 		tipc_sk_send_ack(tsk);
1780 exit:
1781 	release_sock(sk);
1782 	return rc ? rc : copy;
1783 }
1784 
1785 /**
1786  * tipc_recvstream - receive stream-oriented data
1787  * @m: descriptor for message info
1788  * @buflen: total size of user buffer area
1789  * @flags: receive flags
1790  *
1791  * Used for SOCK_STREAM messages only.  If not enough data is available
1792  * will optionally wait for more; never truncates data.
1793  *
1794  * Returns size of returned message data, errno otherwise
1795  */
1796 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1797 			   size_t buflen, int flags)
1798 {
1799 	struct sock *sk = sock->sk;
1800 	struct tipc_sock *tsk = tipc_sk(sk);
1801 	struct sk_buff *skb;
1802 	struct tipc_msg *hdr;
1803 	struct tipc_skb_cb *skb_cb;
1804 	bool peek = flags & MSG_PEEK;
1805 	int offset, required, copy, copied = 0;
1806 	int hlen, dlen, err, rc;
1807 	long timeout;
1808 
1809 	/* Catch invalid receive attempts */
1810 	if (unlikely(!buflen))
1811 		return -EINVAL;
1812 
1813 	lock_sock(sk);
1814 
1815 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1816 		rc = -ENOTCONN;
1817 		goto exit;
1818 	}
1819 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1820 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1821 
1822 	do {
1823 		/* Look at first msg in receive queue; wait if necessary */
1824 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1825 		if (unlikely(rc))
1826 			break;
1827 		skb = skb_peek(&sk->sk_receive_queue);
1828 		skb_cb = TIPC_SKB_CB(skb);
1829 		hdr = buf_msg(skb);
1830 		dlen = msg_data_sz(hdr);
1831 		hlen = msg_hdr_sz(hdr);
1832 		err = msg_errcode(hdr);
1833 
1834 		/* Discard any empty non-errored (SYN-) message */
1835 		if (unlikely(!dlen && !err)) {
1836 			tsk_advance_rx_queue(sk);
1837 			continue;
1838 		}
1839 
1840 		/* Collect msg meta data, incl. error code and rejected data */
1841 		if (!copied) {
1842 			tipc_sk_set_orig_addr(m, skb);
1843 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1844 			if (rc)
1845 				break;
1846 		}
1847 
1848 		/* Copy data if msg ok, otherwise return error/partial data */
1849 		if (likely(!err)) {
1850 			offset = skb_cb->bytes_read;
1851 			copy = min_t(int, dlen - offset, buflen - copied);
1852 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1853 			if (unlikely(rc))
1854 				break;
1855 			copied += copy;
1856 			offset += copy;
1857 			if (unlikely(offset < dlen)) {
1858 				if (!peek)
1859 					skb_cb->bytes_read = offset;
1860 				break;
1861 			}
1862 		} else {
1863 			rc = 0;
1864 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1865 				rc = -ECONNRESET;
1866 			if (copied || rc)
1867 				break;
1868 		}
1869 
1870 		if (unlikely(peek))
1871 			break;
1872 
1873 		tsk_advance_rx_queue(sk);
1874 
1875 		/* Send connection flow control advertisement when applicable */
1876 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1877 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1878 			tipc_sk_send_ack(tsk);
1879 
1880 		/* Exit if all requested data or FIN/error received */
1881 		if (copied == buflen || err)
1882 			break;
1883 
1884 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1885 exit:
1886 	release_sock(sk);
1887 	return copied ? copied : rc;
1888 }
1889 
1890 /**
1891  * tipc_write_space - wake up thread if port congestion is released
1892  * @sk: socket
1893  */
1894 static void tipc_write_space(struct sock *sk)
1895 {
1896 	struct socket_wq *wq;
1897 
1898 	rcu_read_lock();
1899 	wq = rcu_dereference(sk->sk_wq);
1900 	if (skwq_has_sleeper(wq))
1901 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1902 						EPOLLWRNORM | EPOLLWRBAND);
1903 	rcu_read_unlock();
1904 }
1905 
1906 /**
1907  * tipc_data_ready - wake up threads to indicate messages have been received
1908  * @sk: socket
1909  * @len: the length of messages
1910  */
1911 static void tipc_data_ready(struct sock *sk)
1912 {
1913 	struct socket_wq *wq;
1914 
1915 	rcu_read_lock();
1916 	wq = rcu_dereference(sk->sk_wq);
1917 	if (skwq_has_sleeper(wq))
1918 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1919 						EPOLLRDNORM | EPOLLRDBAND);
1920 	rcu_read_unlock();
1921 }
1922 
1923 static void tipc_sock_destruct(struct sock *sk)
1924 {
1925 	__skb_queue_purge(&sk->sk_receive_queue);
1926 }
1927 
1928 static void tipc_sk_proto_rcv(struct sock *sk,
1929 			      struct sk_buff_head *inputq,
1930 			      struct sk_buff_head *xmitq)
1931 {
1932 	struct sk_buff *skb = __skb_dequeue(inputq);
1933 	struct tipc_sock *tsk = tipc_sk(sk);
1934 	struct tipc_msg *hdr = buf_msg(skb);
1935 	struct tipc_group *grp = tsk->group;
1936 	bool wakeup = false;
1937 
1938 	switch (msg_user(hdr)) {
1939 	case CONN_MANAGER:
1940 		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1941 		return;
1942 	case SOCK_WAKEUP:
1943 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1944 		tsk->cong_link_cnt--;
1945 		wakeup = true;
1946 		break;
1947 	case GROUP_PROTOCOL:
1948 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1949 		break;
1950 	case TOP_SRV:
1951 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1952 				      hdr, inputq, xmitq);
1953 		break;
1954 	default:
1955 		break;
1956 	}
1957 
1958 	if (wakeup)
1959 		sk->sk_write_space(sk);
1960 
1961 	kfree_skb(skb);
1962 }
1963 
1964 /**
1965  * tipc_filter_connect - Handle incoming message for a connection-based socket
1966  * @tsk: TIPC socket
1967  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1968  *
1969  * Returns true if everything ok, false otherwise
1970  */
1971 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1972 {
1973 	struct sock *sk = &tsk->sk;
1974 	struct net *net = sock_net(sk);
1975 	struct tipc_msg *hdr = buf_msg(skb);
1976 	u32 pport = msg_origport(hdr);
1977 	u32 pnode = msg_orignode(hdr);
1978 
1979 	if (unlikely(msg_mcast(hdr)))
1980 		return false;
1981 
1982 	switch (sk->sk_state) {
1983 	case TIPC_CONNECTING:
1984 		/* Accept only ACK or NACK message */
1985 		if (unlikely(!msg_connected(hdr))) {
1986 			if (pport != tsk_peer_port(tsk) ||
1987 			    pnode != tsk_peer_node(tsk))
1988 				return false;
1989 
1990 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1991 			sk->sk_err = ECONNREFUSED;
1992 			sk->sk_state_change(sk);
1993 			return true;
1994 		}
1995 
1996 		if (unlikely(msg_errcode(hdr))) {
1997 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1998 			sk->sk_err = ECONNREFUSED;
1999 			sk->sk_state_change(sk);
2000 			return true;
2001 		}
2002 
2003 		if (unlikely(!msg_isdata(hdr))) {
2004 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2005 			sk->sk_err = EINVAL;
2006 			sk->sk_state_change(sk);
2007 			return true;
2008 		}
2009 
2010 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2011 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
2012 
2013 		/* If 'ACK+' message, add to socket receive queue */
2014 		if (msg_data_sz(hdr))
2015 			return true;
2016 
2017 		/* If empty 'ACK-' message, wake up sleeping connect() */
2018 		sk->sk_data_ready(sk);
2019 
2020 		/* 'ACK-' message is neither accepted nor rejected: */
2021 		msg_set_dest_droppable(hdr, 1);
2022 		return false;
2023 
2024 	case TIPC_OPEN:
2025 	case TIPC_DISCONNECTING:
2026 		break;
2027 	case TIPC_LISTEN:
2028 		/* Accept only SYN message */
2029 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2030 			return true;
2031 		break;
2032 	case TIPC_ESTABLISHED:
2033 		/* Accept only connection-based messages sent by peer */
2034 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
2035 			return false;
2036 
2037 		if (unlikely(msg_errcode(hdr))) {
2038 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2039 			/* Let timer expire on it's own */
2040 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2041 					      tsk->portid);
2042 			sk->sk_state_change(sk);
2043 		}
2044 		return true;
2045 	default:
2046 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2047 	}
2048 
2049 	return false;
2050 }
2051 
2052 /**
2053  * rcvbuf_limit - get proper overload limit of socket receive queue
2054  * @sk: socket
2055  * @skb: message
2056  *
2057  * For connection oriented messages, irrespective of importance,
2058  * default queue limit is 2 MB.
2059  *
2060  * For connectionless messages, queue limits are based on message
2061  * importance as follows:
2062  *
2063  * TIPC_LOW_IMPORTANCE       (2 MB)
2064  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2065  * TIPC_HIGH_IMPORTANCE      (8 MB)
2066  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2067  *
2068  * Returns overload limit according to corresponding message importance
2069  */
2070 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2071 {
2072 	struct tipc_sock *tsk = tipc_sk(sk);
2073 	struct tipc_msg *hdr = buf_msg(skb);
2074 
2075 	if (unlikely(msg_in_group(hdr)))
2076 		return sk->sk_rcvbuf;
2077 
2078 	if (unlikely(!msg_connected(hdr)))
2079 		return sk->sk_rcvbuf << msg_importance(hdr);
2080 
2081 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2082 		return sk->sk_rcvbuf;
2083 
2084 	return FLOWCTL_MSG_LIM;
2085 }
2086 
2087 /**
2088  * tipc_sk_filter_rcv - validate incoming message
2089  * @sk: socket
2090  * @skb: pointer to message.
2091  *
2092  * Enqueues message on receive queue if acceptable; optionally handles
2093  * disconnect indication for a connected socket.
2094  *
2095  * Called with socket lock already taken
2096  *
2097  */
2098 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2099 			       struct sk_buff_head *xmitq)
2100 {
2101 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2102 	struct tipc_sock *tsk = tipc_sk(sk);
2103 	struct tipc_group *grp = tsk->group;
2104 	struct tipc_msg *hdr = buf_msg(skb);
2105 	struct net *net = sock_net(sk);
2106 	struct sk_buff_head inputq;
2107 	int limit, err = TIPC_OK;
2108 
2109 	TIPC_SKB_CB(skb)->bytes_read = 0;
2110 	__skb_queue_head_init(&inputq);
2111 	__skb_queue_tail(&inputq, skb);
2112 
2113 	if (unlikely(!msg_isdata(hdr)))
2114 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2115 
2116 	if (unlikely(grp))
2117 		tipc_group_filter_msg(grp, &inputq, xmitq);
2118 
2119 	/* Validate and add to receive buffer if there is space */
2120 	while ((skb = __skb_dequeue(&inputq))) {
2121 		hdr = buf_msg(skb);
2122 		limit = rcvbuf_limit(sk, skb);
2123 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2124 		    (!sk_conn && msg_connected(hdr)) ||
2125 		    (!grp && msg_in_group(hdr)))
2126 			err = TIPC_ERR_NO_PORT;
2127 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit)
2128 			err = TIPC_ERR_OVERLOAD;
2129 
2130 		if (unlikely(err)) {
2131 			tipc_skb_reject(net, err, skb, xmitq);
2132 			err = TIPC_OK;
2133 			continue;
2134 		}
2135 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2136 		skb_set_owner_r(skb, sk);
2137 		sk->sk_data_ready(sk);
2138 	}
2139 }
2140 
2141 /**
2142  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2143  * @sk: socket
2144  * @skb: message
2145  *
2146  * Caller must hold socket lock
2147  */
2148 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2149 {
2150 	unsigned int before = sk_rmem_alloc_get(sk);
2151 	struct sk_buff_head xmitq;
2152 	unsigned int added;
2153 
2154 	__skb_queue_head_init(&xmitq);
2155 
2156 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2157 	added = sk_rmem_alloc_get(sk) - before;
2158 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2159 
2160 	/* Send pending response/rejected messages, if any */
2161 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2162 	return 0;
2163 }
2164 
2165 /**
2166  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2167  *                   inputq and try adding them to socket or backlog queue
2168  * @inputq: list of incoming buffers with potentially different destinations
2169  * @sk: socket where the buffers should be enqueued
2170  * @dport: port number for the socket
2171  *
2172  * Caller must hold socket lock
2173  */
2174 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2175 			    u32 dport, struct sk_buff_head *xmitq)
2176 {
2177 	unsigned long time_limit = jiffies + 2;
2178 	struct sk_buff *skb;
2179 	unsigned int lim;
2180 	atomic_t *dcnt;
2181 	u32 onode;
2182 
2183 	while (skb_queue_len(inputq)) {
2184 		if (unlikely(time_after_eq(jiffies, time_limit)))
2185 			return;
2186 
2187 		skb = tipc_skb_dequeue(inputq, dport);
2188 		if (unlikely(!skb))
2189 			return;
2190 
2191 		/* Add message directly to receive queue if possible */
2192 		if (!sock_owned_by_user(sk)) {
2193 			tipc_sk_filter_rcv(sk, skb, xmitq);
2194 			continue;
2195 		}
2196 
2197 		/* Try backlog, compensating for double-counted bytes */
2198 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2199 		if (!sk->sk_backlog.len)
2200 			atomic_set(dcnt, 0);
2201 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2202 		if (likely(!sk_add_backlog(sk, skb, lim)))
2203 			continue;
2204 
2205 		/* Overload => reject message back to sender */
2206 		onode = tipc_own_addr(sock_net(sk));
2207 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2208 			__skb_queue_tail(xmitq, skb);
2209 		break;
2210 	}
2211 }
2212 
2213 /**
2214  * tipc_sk_rcv - handle a chain of incoming buffers
2215  * @inputq: buffer list containing the buffers
2216  * Consumes all buffers in list until inputq is empty
2217  * Note: may be called in multiple threads referring to the same queue
2218  */
2219 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2220 {
2221 	struct sk_buff_head xmitq;
2222 	u32 dnode, dport = 0;
2223 	int err;
2224 	struct tipc_sock *tsk;
2225 	struct sock *sk;
2226 	struct sk_buff *skb;
2227 
2228 	__skb_queue_head_init(&xmitq);
2229 	while (skb_queue_len(inputq)) {
2230 		dport = tipc_skb_peek_port(inputq, dport);
2231 		tsk = tipc_sk_lookup(net, dport);
2232 
2233 		if (likely(tsk)) {
2234 			sk = &tsk->sk;
2235 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2236 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2237 				spin_unlock_bh(&sk->sk_lock.slock);
2238 			}
2239 			/* Send pending response/rejected messages, if any */
2240 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2241 			sock_put(sk);
2242 			continue;
2243 		}
2244 		/* No destination socket => dequeue skb if still there */
2245 		skb = tipc_skb_dequeue(inputq, dport);
2246 		if (!skb)
2247 			return;
2248 
2249 		/* Try secondary lookup if unresolved named message */
2250 		err = TIPC_ERR_NO_PORT;
2251 		if (tipc_msg_lookup_dest(net, skb, &err))
2252 			goto xmit;
2253 
2254 		/* Prepare for message rejection */
2255 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2256 			continue;
2257 xmit:
2258 		dnode = msg_destnode(buf_msg(skb));
2259 		tipc_node_xmit_skb(net, skb, dnode, dport);
2260 	}
2261 }
2262 
2263 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2264 {
2265 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2266 	struct sock *sk = sock->sk;
2267 	int done;
2268 
2269 	do {
2270 		int err = sock_error(sk);
2271 		if (err)
2272 			return err;
2273 		if (!*timeo_p)
2274 			return -ETIMEDOUT;
2275 		if (signal_pending(current))
2276 			return sock_intr_errno(*timeo_p);
2277 
2278 		add_wait_queue(sk_sleep(sk), &wait);
2279 		done = sk_wait_event(sk, timeo_p,
2280 				     sk->sk_state != TIPC_CONNECTING, &wait);
2281 		remove_wait_queue(sk_sleep(sk), &wait);
2282 	} while (!done);
2283 	return 0;
2284 }
2285 
2286 /**
2287  * tipc_connect - establish a connection to another TIPC port
2288  * @sock: socket structure
2289  * @dest: socket address for destination port
2290  * @destlen: size of socket address data structure
2291  * @flags: file-related flags associated with socket
2292  *
2293  * Returns 0 on success, errno otherwise
2294  */
2295 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2296 			int destlen, int flags)
2297 {
2298 	struct sock *sk = sock->sk;
2299 	struct tipc_sock *tsk = tipc_sk(sk);
2300 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2301 	struct msghdr m = {NULL,};
2302 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2303 	int previous;
2304 	int res = 0;
2305 
2306 	if (destlen != sizeof(struct sockaddr_tipc))
2307 		return -EINVAL;
2308 
2309 	lock_sock(sk);
2310 
2311 	if (tsk->group) {
2312 		res = -EINVAL;
2313 		goto exit;
2314 	}
2315 
2316 	if (dst->family == AF_UNSPEC) {
2317 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2318 		if (!tipc_sk_type_connectionless(sk))
2319 			res = -EINVAL;
2320 		goto exit;
2321 	} else if (dst->family != AF_TIPC) {
2322 		res = -EINVAL;
2323 	}
2324 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2325 		res = -EINVAL;
2326 	if (res)
2327 		goto exit;
2328 
2329 	/* DGRAM/RDM connect(), just save the destaddr */
2330 	if (tipc_sk_type_connectionless(sk)) {
2331 		memcpy(&tsk->peer, dest, destlen);
2332 		goto exit;
2333 	}
2334 
2335 	previous = sk->sk_state;
2336 
2337 	switch (sk->sk_state) {
2338 	case TIPC_OPEN:
2339 		/* Send a 'SYN-' to destination */
2340 		m.msg_name = dest;
2341 		m.msg_namelen = destlen;
2342 
2343 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2344 		 * indicate send_msg() is never blocked.
2345 		 */
2346 		if (!timeout)
2347 			m.msg_flags = MSG_DONTWAIT;
2348 
2349 		res = __tipc_sendmsg(sock, &m, 0);
2350 		if ((res < 0) && (res != -EWOULDBLOCK))
2351 			goto exit;
2352 
2353 		/* Just entered TIPC_CONNECTING state; the only
2354 		 * difference is that return value in non-blocking
2355 		 * case is EINPROGRESS, rather than EALREADY.
2356 		 */
2357 		res = -EINPROGRESS;
2358 		/* fall thru' */
2359 	case TIPC_CONNECTING:
2360 		if (!timeout) {
2361 			if (previous == TIPC_CONNECTING)
2362 				res = -EALREADY;
2363 			goto exit;
2364 		}
2365 		timeout = msecs_to_jiffies(timeout);
2366 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2367 		res = tipc_wait_for_connect(sock, &timeout);
2368 		break;
2369 	case TIPC_ESTABLISHED:
2370 		res = -EISCONN;
2371 		break;
2372 	default:
2373 		res = -EINVAL;
2374 	}
2375 
2376 exit:
2377 	release_sock(sk);
2378 	return res;
2379 }
2380 
2381 /**
2382  * tipc_listen - allow socket to listen for incoming connections
2383  * @sock: socket structure
2384  * @len: (unused)
2385  *
2386  * Returns 0 on success, errno otherwise
2387  */
2388 static int tipc_listen(struct socket *sock, int len)
2389 {
2390 	struct sock *sk = sock->sk;
2391 	int res;
2392 
2393 	lock_sock(sk);
2394 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2395 	release_sock(sk);
2396 
2397 	return res;
2398 }
2399 
2400 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2401 {
2402 	struct sock *sk = sock->sk;
2403 	DEFINE_WAIT(wait);
2404 	int err;
2405 
2406 	/* True wake-one mechanism for incoming connections: only
2407 	 * one process gets woken up, not the 'whole herd'.
2408 	 * Since we do not 'race & poll' for established sockets
2409 	 * anymore, the common case will execute the loop only once.
2410 	*/
2411 	for (;;) {
2412 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2413 					  TASK_INTERRUPTIBLE);
2414 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2415 			release_sock(sk);
2416 			timeo = schedule_timeout(timeo);
2417 			lock_sock(sk);
2418 		}
2419 		err = 0;
2420 		if (!skb_queue_empty(&sk->sk_receive_queue))
2421 			break;
2422 		err = -EAGAIN;
2423 		if (!timeo)
2424 			break;
2425 		err = sock_intr_errno(timeo);
2426 		if (signal_pending(current))
2427 			break;
2428 	}
2429 	finish_wait(sk_sleep(sk), &wait);
2430 	return err;
2431 }
2432 
2433 /**
2434  * tipc_accept - wait for connection request
2435  * @sock: listening socket
2436  * @newsock: new socket that is to be connected
2437  * @flags: file-related flags associated with socket
2438  *
2439  * Returns 0 on success, errno otherwise
2440  */
2441 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2442 		       bool kern)
2443 {
2444 	struct sock *new_sk, *sk = sock->sk;
2445 	struct sk_buff *buf;
2446 	struct tipc_sock *new_tsock;
2447 	struct tipc_msg *msg;
2448 	long timeo;
2449 	int res;
2450 
2451 	lock_sock(sk);
2452 
2453 	if (sk->sk_state != TIPC_LISTEN) {
2454 		res = -EINVAL;
2455 		goto exit;
2456 	}
2457 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2458 	res = tipc_wait_for_accept(sock, timeo);
2459 	if (res)
2460 		goto exit;
2461 
2462 	buf = skb_peek(&sk->sk_receive_queue);
2463 
2464 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2465 	if (res)
2466 		goto exit;
2467 	security_sk_clone(sock->sk, new_sock->sk);
2468 
2469 	new_sk = new_sock->sk;
2470 	new_tsock = tipc_sk(new_sk);
2471 	msg = buf_msg(buf);
2472 
2473 	/* we lock on new_sk; but lockdep sees the lock on sk */
2474 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2475 
2476 	/*
2477 	 * Reject any stray messages received by new socket
2478 	 * before the socket lock was taken (very, very unlikely)
2479 	 */
2480 	tsk_rej_rx_queue(new_sk);
2481 
2482 	/* Connect new socket to it's peer */
2483 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2484 
2485 	tsk_set_importance(new_tsock, msg_importance(msg));
2486 	if (msg_named(msg)) {
2487 		new_tsock->conn_type = msg_nametype(msg);
2488 		new_tsock->conn_instance = msg_nameinst(msg);
2489 	}
2490 
2491 	/*
2492 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2493 	 * Respond to 'SYN+' by queuing it on new socket.
2494 	 */
2495 	if (!msg_data_sz(msg)) {
2496 		struct msghdr m = {NULL,};
2497 
2498 		tsk_advance_rx_queue(sk);
2499 		__tipc_sendstream(new_sock, &m, 0);
2500 	} else {
2501 		__skb_dequeue(&sk->sk_receive_queue);
2502 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2503 		skb_set_owner_r(buf, new_sk);
2504 	}
2505 	release_sock(new_sk);
2506 exit:
2507 	release_sock(sk);
2508 	return res;
2509 }
2510 
2511 /**
2512  * tipc_shutdown - shutdown socket connection
2513  * @sock: socket structure
2514  * @how: direction to close (must be SHUT_RDWR)
2515  *
2516  * Terminates connection (if necessary), then purges socket's receive queue.
2517  *
2518  * Returns 0 on success, errno otherwise
2519  */
2520 static int tipc_shutdown(struct socket *sock, int how)
2521 {
2522 	struct sock *sk = sock->sk;
2523 	int res;
2524 
2525 	if (how != SHUT_RDWR)
2526 		return -EINVAL;
2527 
2528 	lock_sock(sk);
2529 
2530 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2531 	sk->sk_shutdown = SEND_SHUTDOWN;
2532 
2533 	if (sk->sk_state == TIPC_DISCONNECTING) {
2534 		/* Discard any unreceived messages */
2535 		__skb_queue_purge(&sk->sk_receive_queue);
2536 
2537 		/* Wake up anyone sleeping in poll */
2538 		sk->sk_state_change(sk);
2539 		res = 0;
2540 	} else {
2541 		res = -ENOTCONN;
2542 	}
2543 
2544 	release_sock(sk);
2545 	return res;
2546 }
2547 
2548 static void tipc_sk_timeout(struct timer_list *t)
2549 {
2550 	struct sock *sk = from_timer(sk, t, sk_timer);
2551 	struct tipc_sock *tsk = tipc_sk(sk);
2552 	u32 peer_port = tsk_peer_port(tsk);
2553 	u32 peer_node = tsk_peer_node(tsk);
2554 	u32 own_node = tsk_own_node(tsk);
2555 	u32 own_port = tsk->portid;
2556 	struct net *net = sock_net(sk);
2557 	struct sk_buff *skb = NULL;
2558 
2559 	bh_lock_sock(sk);
2560 	if (!tipc_sk_connected(sk))
2561 		goto exit;
2562 
2563 	/* Try again later if socket is busy */
2564 	if (sock_owned_by_user(sk)) {
2565 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2566 		goto exit;
2567 	}
2568 
2569 	if (tsk->probe_unacked) {
2570 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2571 		tipc_node_remove_conn(net, peer_node, peer_port);
2572 		sk->sk_state_change(sk);
2573 		goto exit;
2574 	}
2575 	/* Send new probe */
2576 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2577 			      peer_node, own_node, peer_port, own_port,
2578 			      TIPC_OK);
2579 	tsk->probe_unacked = true;
2580 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2581 exit:
2582 	bh_unlock_sock(sk);
2583 	if (skb)
2584 		tipc_node_xmit_skb(net, skb, peer_node, own_port);
2585 	sock_put(sk);
2586 }
2587 
2588 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2589 			   struct tipc_name_seq const *seq)
2590 {
2591 	struct sock *sk = &tsk->sk;
2592 	struct net *net = sock_net(sk);
2593 	struct publication *publ;
2594 	u32 key;
2595 
2596 	if (tipc_sk_connected(sk))
2597 		return -EINVAL;
2598 	key = tsk->portid + tsk->pub_count + 1;
2599 	if (key == tsk->portid)
2600 		return -EADDRINUSE;
2601 
2602 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2603 				    scope, tsk->portid, key);
2604 	if (unlikely(!publ))
2605 		return -EINVAL;
2606 
2607 	list_add(&publ->pport_list, &tsk->publications);
2608 	tsk->pub_count++;
2609 	tsk->published = 1;
2610 	return 0;
2611 }
2612 
2613 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2614 			    struct tipc_name_seq const *seq)
2615 {
2616 	struct net *net = sock_net(&tsk->sk);
2617 	struct publication *publ;
2618 	struct publication *safe;
2619 	int rc = -EINVAL;
2620 
2621 	list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2622 		if (seq) {
2623 			if (publ->scope != scope)
2624 				continue;
2625 			if (publ->type != seq->type)
2626 				continue;
2627 			if (publ->lower != seq->lower)
2628 				continue;
2629 			if (publ->upper != seq->upper)
2630 				break;
2631 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2632 					      publ->ref, publ->key);
2633 			rc = 0;
2634 			break;
2635 		}
2636 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2637 				      publ->ref, publ->key);
2638 		rc = 0;
2639 	}
2640 	if (list_empty(&tsk->publications))
2641 		tsk->published = 0;
2642 	return rc;
2643 }
2644 
2645 /* tipc_sk_reinit: set non-zero address in all existing sockets
2646  *                 when we go from standalone to network mode.
2647  */
2648 void tipc_sk_reinit(struct net *net)
2649 {
2650 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2651 	struct rhashtable_iter iter;
2652 	struct tipc_sock *tsk;
2653 	struct tipc_msg *msg;
2654 
2655 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2656 
2657 	do {
2658 		rhashtable_walk_start(&iter);
2659 
2660 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2661 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2662 			msg = &tsk->phdr;
2663 			msg_set_prevnode(msg, tn->own_addr);
2664 			msg_set_orignode(msg, tn->own_addr);
2665 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2666 		}
2667 
2668 		rhashtable_walk_stop(&iter);
2669 	} while (tsk == ERR_PTR(-EAGAIN));
2670 }
2671 
2672 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2673 {
2674 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2675 	struct tipc_sock *tsk;
2676 
2677 	rcu_read_lock();
2678 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2679 	if (tsk)
2680 		sock_hold(&tsk->sk);
2681 	rcu_read_unlock();
2682 
2683 	return tsk;
2684 }
2685 
2686 static int tipc_sk_insert(struct tipc_sock *tsk)
2687 {
2688 	struct sock *sk = &tsk->sk;
2689 	struct net *net = sock_net(sk);
2690 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2691 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2692 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2693 
2694 	while (remaining--) {
2695 		portid++;
2696 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2697 			portid = TIPC_MIN_PORT;
2698 		tsk->portid = portid;
2699 		sock_hold(&tsk->sk);
2700 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2701 						   tsk_rht_params))
2702 			return 0;
2703 		sock_put(&tsk->sk);
2704 	}
2705 
2706 	return -1;
2707 }
2708 
2709 static void tipc_sk_remove(struct tipc_sock *tsk)
2710 {
2711 	struct sock *sk = &tsk->sk;
2712 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2713 
2714 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2715 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2716 		__sock_put(sk);
2717 	}
2718 }
2719 
2720 static const struct rhashtable_params tsk_rht_params = {
2721 	.nelem_hint = 192,
2722 	.head_offset = offsetof(struct tipc_sock, node),
2723 	.key_offset = offsetof(struct tipc_sock, portid),
2724 	.key_len = sizeof(u32), /* portid */
2725 	.max_size = 1048576,
2726 	.min_size = 256,
2727 	.automatic_shrinking = true,
2728 };
2729 
2730 int tipc_sk_rht_init(struct net *net)
2731 {
2732 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2733 
2734 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2735 }
2736 
2737 void tipc_sk_rht_destroy(struct net *net)
2738 {
2739 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2740 
2741 	/* Wait for socket readers to complete */
2742 	synchronize_net();
2743 
2744 	rhashtable_destroy(&tn->sk_rht);
2745 }
2746 
2747 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2748 {
2749 	struct net *net = sock_net(&tsk->sk);
2750 	struct tipc_group *grp = tsk->group;
2751 	struct tipc_msg *hdr = &tsk->phdr;
2752 	struct tipc_name_seq seq;
2753 	int rc;
2754 
2755 	if (mreq->type < TIPC_RESERVED_TYPES)
2756 		return -EACCES;
2757 	if (mreq->scope > TIPC_NODE_SCOPE)
2758 		return -EINVAL;
2759 	if (grp)
2760 		return -EACCES;
2761 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2762 	if (!grp)
2763 		return -ENOMEM;
2764 	tsk->group = grp;
2765 	msg_set_lookup_scope(hdr, mreq->scope);
2766 	msg_set_nametype(hdr, mreq->type);
2767 	msg_set_dest_droppable(hdr, true);
2768 	seq.type = mreq->type;
2769 	seq.lower = mreq->instance;
2770 	seq.upper = seq.lower;
2771 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2772 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2773 	if (rc) {
2774 		tipc_group_delete(net, grp);
2775 		tsk->group = NULL;
2776 		return rc;
2777 	}
2778 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2779 	tsk->mc_method.rcast = true;
2780 	tsk->mc_method.mandatory = true;
2781 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2782 	return rc;
2783 }
2784 
2785 static int tipc_sk_leave(struct tipc_sock *tsk)
2786 {
2787 	struct net *net = sock_net(&tsk->sk);
2788 	struct tipc_group *grp = tsk->group;
2789 	struct tipc_name_seq seq;
2790 	int scope;
2791 
2792 	if (!grp)
2793 		return -EINVAL;
2794 	tipc_group_self(grp, &seq, &scope);
2795 	tipc_group_delete(net, grp);
2796 	tsk->group = NULL;
2797 	tipc_sk_withdraw(tsk, scope, &seq);
2798 	return 0;
2799 }
2800 
2801 /**
2802  * tipc_setsockopt - set socket option
2803  * @sock: socket structure
2804  * @lvl: option level
2805  * @opt: option identifier
2806  * @ov: pointer to new option value
2807  * @ol: length of option value
2808  *
2809  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2810  * (to ease compatibility).
2811  *
2812  * Returns 0 on success, errno otherwise
2813  */
2814 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2815 			   char __user *ov, unsigned int ol)
2816 {
2817 	struct sock *sk = sock->sk;
2818 	struct tipc_sock *tsk = tipc_sk(sk);
2819 	struct tipc_group_req mreq;
2820 	u32 value = 0;
2821 	int res = 0;
2822 
2823 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2824 		return 0;
2825 	if (lvl != SOL_TIPC)
2826 		return -ENOPROTOOPT;
2827 
2828 	switch (opt) {
2829 	case TIPC_IMPORTANCE:
2830 	case TIPC_SRC_DROPPABLE:
2831 	case TIPC_DEST_DROPPABLE:
2832 	case TIPC_CONN_TIMEOUT:
2833 		if (ol < sizeof(value))
2834 			return -EINVAL;
2835 		if (get_user(value, (u32 __user *)ov))
2836 			return -EFAULT;
2837 		break;
2838 	case TIPC_GROUP_JOIN:
2839 		if (ol < sizeof(mreq))
2840 			return -EINVAL;
2841 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2842 			return -EFAULT;
2843 		break;
2844 	default:
2845 		if (ov || ol)
2846 			return -EINVAL;
2847 	}
2848 
2849 	lock_sock(sk);
2850 
2851 	switch (opt) {
2852 	case TIPC_IMPORTANCE:
2853 		res = tsk_set_importance(tsk, value);
2854 		break;
2855 	case TIPC_SRC_DROPPABLE:
2856 		if (sock->type != SOCK_STREAM)
2857 			tsk_set_unreliable(tsk, value);
2858 		else
2859 			res = -ENOPROTOOPT;
2860 		break;
2861 	case TIPC_DEST_DROPPABLE:
2862 		tsk_set_unreturnable(tsk, value);
2863 		break;
2864 	case TIPC_CONN_TIMEOUT:
2865 		tipc_sk(sk)->conn_timeout = value;
2866 		break;
2867 	case TIPC_MCAST_BROADCAST:
2868 		tsk->mc_method.rcast = false;
2869 		tsk->mc_method.mandatory = true;
2870 		break;
2871 	case TIPC_MCAST_REPLICAST:
2872 		tsk->mc_method.rcast = true;
2873 		tsk->mc_method.mandatory = true;
2874 		break;
2875 	case TIPC_GROUP_JOIN:
2876 		res = tipc_sk_join(tsk, &mreq);
2877 		break;
2878 	case TIPC_GROUP_LEAVE:
2879 		res = tipc_sk_leave(tsk);
2880 		break;
2881 	default:
2882 		res = -EINVAL;
2883 	}
2884 
2885 	release_sock(sk);
2886 
2887 	return res;
2888 }
2889 
2890 /**
2891  * tipc_getsockopt - get socket option
2892  * @sock: socket structure
2893  * @lvl: option level
2894  * @opt: option identifier
2895  * @ov: receptacle for option value
2896  * @ol: receptacle for length of option value
2897  *
2898  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2899  * (to ease compatibility).
2900  *
2901  * Returns 0 on success, errno otherwise
2902  */
2903 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2904 			   char __user *ov, int __user *ol)
2905 {
2906 	struct sock *sk = sock->sk;
2907 	struct tipc_sock *tsk = tipc_sk(sk);
2908 	struct tipc_name_seq seq;
2909 	int len, scope;
2910 	u32 value;
2911 	int res;
2912 
2913 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2914 		return put_user(0, ol);
2915 	if (lvl != SOL_TIPC)
2916 		return -ENOPROTOOPT;
2917 	res = get_user(len, ol);
2918 	if (res)
2919 		return res;
2920 
2921 	lock_sock(sk);
2922 
2923 	switch (opt) {
2924 	case TIPC_IMPORTANCE:
2925 		value = tsk_importance(tsk);
2926 		break;
2927 	case TIPC_SRC_DROPPABLE:
2928 		value = tsk_unreliable(tsk);
2929 		break;
2930 	case TIPC_DEST_DROPPABLE:
2931 		value = tsk_unreturnable(tsk);
2932 		break;
2933 	case TIPC_CONN_TIMEOUT:
2934 		value = tsk->conn_timeout;
2935 		/* no need to set "res", since already 0 at this point */
2936 		break;
2937 	case TIPC_NODE_RECVQ_DEPTH:
2938 		value = 0; /* was tipc_queue_size, now obsolete */
2939 		break;
2940 	case TIPC_SOCK_RECVQ_DEPTH:
2941 		value = skb_queue_len(&sk->sk_receive_queue);
2942 		break;
2943 	case TIPC_GROUP_JOIN:
2944 		seq.type = 0;
2945 		if (tsk->group)
2946 			tipc_group_self(tsk->group, &seq, &scope);
2947 		value = seq.type;
2948 		break;
2949 	default:
2950 		res = -EINVAL;
2951 	}
2952 
2953 	release_sock(sk);
2954 
2955 	if (res)
2956 		return res;	/* "get" failed */
2957 
2958 	if (len < sizeof(value))
2959 		return -EINVAL;
2960 
2961 	if (copy_to_user(ov, &value, sizeof(value)))
2962 		return -EFAULT;
2963 
2964 	return put_user(sizeof(value), ol);
2965 }
2966 
2967 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2968 {
2969 	struct sock *sk = sock->sk;
2970 	struct tipc_sioc_ln_req lnr;
2971 	void __user *argp = (void __user *)arg;
2972 
2973 	switch (cmd) {
2974 	case SIOCGETLINKNAME:
2975 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2976 			return -EFAULT;
2977 		if (!tipc_node_get_linkname(sock_net(sk),
2978 					    lnr.bearer_id & 0xffff, lnr.peer,
2979 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2980 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2981 				return -EFAULT;
2982 			return 0;
2983 		}
2984 		return -EADDRNOTAVAIL;
2985 	default:
2986 		return -ENOIOCTLCMD;
2987 	}
2988 }
2989 
2990 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
2991 {
2992 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
2993 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
2994 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
2995 
2996 	tsk1->peer.family = AF_TIPC;
2997 	tsk1->peer.addrtype = TIPC_ADDR_ID;
2998 	tsk1->peer.scope = TIPC_NODE_SCOPE;
2999 	tsk1->peer.addr.id.ref = tsk2->portid;
3000 	tsk1->peer.addr.id.node = onode;
3001 	tsk2->peer.family = AF_TIPC;
3002 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3003 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3004 	tsk2->peer.addr.id.ref = tsk1->portid;
3005 	tsk2->peer.addr.id.node = onode;
3006 
3007 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3008 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3009 	return 0;
3010 }
3011 
3012 /* Protocol switches for the various types of TIPC sockets */
3013 
3014 static const struct proto_ops msg_ops = {
3015 	.owner		= THIS_MODULE,
3016 	.family		= AF_TIPC,
3017 	.release	= tipc_release,
3018 	.bind		= tipc_bind,
3019 	.connect	= tipc_connect,
3020 	.socketpair	= tipc_socketpair,
3021 	.accept		= sock_no_accept,
3022 	.getname	= tipc_getname,
3023 	.poll		= tipc_poll,
3024 	.ioctl		= tipc_ioctl,
3025 	.listen		= sock_no_listen,
3026 	.shutdown	= tipc_shutdown,
3027 	.setsockopt	= tipc_setsockopt,
3028 	.getsockopt	= tipc_getsockopt,
3029 	.sendmsg	= tipc_sendmsg,
3030 	.recvmsg	= tipc_recvmsg,
3031 	.mmap		= sock_no_mmap,
3032 	.sendpage	= sock_no_sendpage
3033 };
3034 
3035 static const struct proto_ops packet_ops = {
3036 	.owner		= THIS_MODULE,
3037 	.family		= AF_TIPC,
3038 	.release	= tipc_release,
3039 	.bind		= tipc_bind,
3040 	.connect	= tipc_connect,
3041 	.socketpair	= tipc_socketpair,
3042 	.accept		= tipc_accept,
3043 	.getname	= tipc_getname,
3044 	.poll		= tipc_poll,
3045 	.ioctl		= tipc_ioctl,
3046 	.listen		= tipc_listen,
3047 	.shutdown	= tipc_shutdown,
3048 	.setsockopt	= tipc_setsockopt,
3049 	.getsockopt	= tipc_getsockopt,
3050 	.sendmsg	= tipc_send_packet,
3051 	.recvmsg	= tipc_recvmsg,
3052 	.mmap		= sock_no_mmap,
3053 	.sendpage	= sock_no_sendpage
3054 };
3055 
3056 static const struct proto_ops stream_ops = {
3057 	.owner		= THIS_MODULE,
3058 	.family		= AF_TIPC,
3059 	.release	= tipc_release,
3060 	.bind		= tipc_bind,
3061 	.connect	= tipc_connect,
3062 	.socketpair	= tipc_socketpair,
3063 	.accept		= tipc_accept,
3064 	.getname	= tipc_getname,
3065 	.poll		= tipc_poll,
3066 	.ioctl		= tipc_ioctl,
3067 	.listen		= tipc_listen,
3068 	.shutdown	= tipc_shutdown,
3069 	.setsockopt	= tipc_setsockopt,
3070 	.getsockopt	= tipc_getsockopt,
3071 	.sendmsg	= tipc_sendstream,
3072 	.recvmsg	= tipc_recvstream,
3073 	.mmap		= sock_no_mmap,
3074 	.sendpage	= sock_no_sendpage
3075 };
3076 
3077 static const struct net_proto_family tipc_family_ops = {
3078 	.owner		= THIS_MODULE,
3079 	.family		= AF_TIPC,
3080 	.create		= tipc_sk_create
3081 };
3082 
3083 static struct proto tipc_proto = {
3084 	.name		= "TIPC",
3085 	.owner		= THIS_MODULE,
3086 	.obj_size	= sizeof(struct tipc_sock),
3087 	.sysctl_rmem	= sysctl_tipc_rmem
3088 };
3089 
3090 /**
3091  * tipc_socket_init - initialize TIPC socket interface
3092  *
3093  * Returns 0 on success, errno otherwise
3094  */
3095 int tipc_socket_init(void)
3096 {
3097 	int res;
3098 
3099 	res = proto_register(&tipc_proto, 1);
3100 	if (res) {
3101 		pr_err("Failed to register TIPC protocol type\n");
3102 		goto out;
3103 	}
3104 
3105 	res = sock_register(&tipc_family_ops);
3106 	if (res) {
3107 		pr_err("Failed to register TIPC socket type\n");
3108 		proto_unregister(&tipc_proto);
3109 		goto out;
3110 	}
3111  out:
3112 	return res;
3113 }
3114 
3115 /**
3116  * tipc_socket_stop - stop TIPC socket interface
3117  */
3118 void tipc_socket_stop(void)
3119 {
3120 	sock_unregister(tipc_family_ops.family);
3121 	proto_unregister(&tipc_proto);
3122 }
3123 
3124 /* Caller should hold socket lock for the passed tipc socket. */
3125 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3126 {
3127 	u32 peer_node;
3128 	u32 peer_port;
3129 	struct nlattr *nest;
3130 
3131 	peer_node = tsk_peer_node(tsk);
3132 	peer_port = tsk_peer_port(tsk);
3133 
3134 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3135 
3136 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3137 		goto msg_full;
3138 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3139 		goto msg_full;
3140 
3141 	if (tsk->conn_type != 0) {
3142 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3143 			goto msg_full;
3144 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3145 			goto msg_full;
3146 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3147 			goto msg_full;
3148 	}
3149 	nla_nest_end(skb, nest);
3150 
3151 	return 0;
3152 
3153 msg_full:
3154 	nla_nest_cancel(skb, nest);
3155 
3156 	return -EMSGSIZE;
3157 }
3158 
3159 /* Caller should hold socket lock for the passed tipc socket. */
3160 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3161 			    struct tipc_sock *tsk)
3162 {
3163 	int err;
3164 	void *hdr;
3165 	struct nlattr *attrs;
3166 	struct net *net = sock_net(skb->sk);
3167 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3168 	struct sock *sk = &tsk->sk;
3169 
3170 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3171 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3172 	if (!hdr)
3173 		goto msg_cancel;
3174 
3175 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3176 	if (!attrs)
3177 		goto genlmsg_cancel;
3178 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
3179 		goto attr_msg_cancel;
3180 	if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
3181 		goto attr_msg_cancel;
3182 
3183 	if (tipc_sk_connected(sk)) {
3184 		err = __tipc_nl_add_sk_con(skb, tsk);
3185 		if (err)
3186 			goto attr_msg_cancel;
3187 	} else if (!list_empty(&tsk->publications)) {
3188 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3189 			goto attr_msg_cancel;
3190 	}
3191 	nla_nest_end(skb, attrs);
3192 	genlmsg_end(skb, hdr);
3193 
3194 	return 0;
3195 
3196 attr_msg_cancel:
3197 	nla_nest_cancel(skb, attrs);
3198 genlmsg_cancel:
3199 	genlmsg_cancel(skb, hdr);
3200 msg_cancel:
3201 	return -EMSGSIZE;
3202 }
3203 
3204 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3205 {
3206 	int err;
3207 	struct tipc_sock *tsk;
3208 	const struct bucket_table *tbl;
3209 	struct rhash_head *pos;
3210 	struct net *net = sock_net(skb->sk);
3211 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3212 	u32 tbl_id = cb->args[0];
3213 	u32 prev_portid = cb->args[1];
3214 
3215 	rcu_read_lock();
3216 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
3217 	for (; tbl_id < tbl->size; tbl_id++) {
3218 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
3219 			spin_lock_bh(&tsk->sk.sk_lock.slock);
3220 			if (prev_portid && prev_portid != tsk->portid) {
3221 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3222 				continue;
3223 			}
3224 
3225 			err = __tipc_nl_add_sk(skb, cb, tsk);
3226 			if (err) {
3227 				prev_portid = tsk->portid;
3228 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3229 				goto out;
3230 			}
3231 			prev_portid = 0;
3232 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
3233 		}
3234 	}
3235 out:
3236 	rcu_read_unlock();
3237 	cb->args[0] = tbl_id;
3238 	cb->args[1] = prev_portid;
3239 
3240 	return skb->len;
3241 }
3242 
3243 /* Caller should hold socket lock for the passed tipc socket. */
3244 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3245 				 struct netlink_callback *cb,
3246 				 struct publication *publ)
3247 {
3248 	void *hdr;
3249 	struct nlattr *attrs;
3250 
3251 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3252 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3253 	if (!hdr)
3254 		goto msg_cancel;
3255 
3256 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3257 	if (!attrs)
3258 		goto genlmsg_cancel;
3259 
3260 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3261 		goto attr_msg_cancel;
3262 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3263 		goto attr_msg_cancel;
3264 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3265 		goto attr_msg_cancel;
3266 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3267 		goto attr_msg_cancel;
3268 
3269 	nla_nest_end(skb, attrs);
3270 	genlmsg_end(skb, hdr);
3271 
3272 	return 0;
3273 
3274 attr_msg_cancel:
3275 	nla_nest_cancel(skb, attrs);
3276 genlmsg_cancel:
3277 	genlmsg_cancel(skb, hdr);
3278 msg_cancel:
3279 	return -EMSGSIZE;
3280 }
3281 
3282 /* Caller should hold socket lock for the passed tipc socket. */
3283 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3284 				  struct netlink_callback *cb,
3285 				  struct tipc_sock *tsk, u32 *last_publ)
3286 {
3287 	int err;
3288 	struct publication *p;
3289 
3290 	if (*last_publ) {
3291 		list_for_each_entry(p, &tsk->publications, pport_list) {
3292 			if (p->key == *last_publ)
3293 				break;
3294 		}
3295 		if (p->key != *last_publ) {
3296 			/* We never set seq or call nl_dump_check_consistent()
3297 			 * this means that setting prev_seq here will cause the
3298 			 * consistence check to fail in the netlink callback
3299 			 * handler. Resulting in the last NLMSG_DONE message
3300 			 * having the NLM_F_DUMP_INTR flag set.
3301 			 */
3302 			cb->prev_seq = 1;
3303 			*last_publ = 0;
3304 			return -EPIPE;
3305 		}
3306 	} else {
3307 		p = list_first_entry(&tsk->publications, struct publication,
3308 				     pport_list);
3309 	}
3310 
3311 	list_for_each_entry_from(p, &tsk->publications, pport_list) {
3312 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3313 		if (err) {
3314 			*last_publ = p->key;
3315 			return err;
3316 		}
3317 	}
3318 	*last_publ = 0;
3319 
3320 	return 0;
3321 }
3322 
3323 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3324 {
3325 	int err;
3326 	u32 tsk_portid = cb->args[0];
3327 	u32 last_publ = cb->args[1];
3328 	u32 done = cb->args[2];
3329 	struct net *net = sock_net(skb->sk);
3330 	struct tipc_sock *tsk;
3331 
3332 	if (!tsk_portid) {
3333 		struct nlattr **attrs;
3334 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3335 
3336 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3337 		if (err)
3338 			return err;
3339 
3340 		if (!attrs[TIPC_NLA_SOCK])
3341 			return -EINVAL;
3342 
3343 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3344 				       attrs[TIPC_NLA_SOCK],
3345 				       tipc_nl_sock_policy, NULL);
3346 		if (err)
3347 			return err;
3348 
3349 		if (!sock[TIPC_NLA_SOCK_REF])
3350 			return -EINVAL;
3351 
3352 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3353 	}
3354 
3355 	if (done)
3356 		return 0;
3357 
3358 	tsk = tipc_sk_lookup(net, tsk_portid);
3359 	if (!tsk)
3360 		return -EINVAL;
3361 
3362 	lock_sock(&tsk->sk);
3363 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3364 	if (!err)
3365 		done = 1;
3366 	release_sock(&tsk->sk);
3367 	sock_put(&tsk->sk);
3368 
3369 	cb->args[0] = tsk_portid;
3370 	cb->args[1] = last_publ;
3371 	cb->args[2] = done;
3372 
3373 	return skb->len;
3374 }
3375