xref: /linux/net/tipc/socket.c (revision 79b6bb73f888933cbcd20b0ef3976cde67951b72)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include <linux/rhashtable.h>
38 #include <linux/sched/signal.h>
39 
40 #include "core.h"
41 #include "name_table.h"
42 #include "node.h"
43 #include "link.h"
44 #include "name_distr.h"
45 #include "socket.h"
46 #include "bcast.h"
47 #include "netlink.h"
48 #include "group.h"
49 #include "trace.h"
50 
51 #define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
52 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
53 #define TIPC_FWD_MSG		1
54 #define TIPC_MAX_PORT		0xffffffff
55 #define TIPC_MIN_PORT		1
56 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
57 
58 enum {
59 	TIPC_LISTEN = TCP_LISTEN,
60 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
61 	TIPC_OPEN = TCP_CLOSE,
62 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
63 	TIPC_CONNECTING = TCP_SYN_SENT,
64 };
65 
66 struct sockaddr_pair {
67 	struct sockaddr_tipc sock;
68 	struct sockaddr_tipc member;
69 };
70 
71 /**
72  * struct tipc_sock - TIPC socket structure
73  * @sk: socket - interacts with 'port' and with user via the socket API
74  * @conn_type: TIPC type used when connection was established
75  * @conn_instance: TIPC instance used when connection was established
76  * @published: non-zero if port has one or more associated names
77  * @max_pkt: maximum packet size "hint" used when building messages sent by port
78  * @maxnagle: maximum size of msg which can be subject to nagle
79  * @portid: unique port identity in TIPC socket hash table
80  * @phdr: preformatted message header used when sending messages
81  * #cong_links: list of congested links
82  * @publications: list of publications for port
83  * @blocking_link: address of the congested link we are currently sleeping on
84  * @pub_count: total # of publications port has made during its lifetime
85  * @conn_timeout: the time we can wait for an unresponded setup request
86  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87  * @cong_link_cnt: number of congested links
88  * @snt_unacked: # messages sent by socket, and not yet acked by peer
89  * @rcv_unacked: # messages read by user, but not yet acked back to peer
90  * @peer: 'connected' peer for dgram/rdm
91  * @node: hash table node
92  * @mc_method: cookie for use between socket and broadcast layer
93  * @rcu: rcu struct for tipc_sock
94  */
95 struct tipc_sock {
96 	struct sock sk;
97 	u32 conn_type;
98 	u32 conn_instance;
99 	int published;
100 	u32 max_pkt;
101 	u32 maxnagle;
102 	u32 portid;
103 	struct tipc_msg phdr;
104 	struct list_head cong_links;
105 	struct list_head publications;
106 	u32 pub_count;
107 	atomic_t dupl_rcvcnt;
108 	u16 conn_timeout;
109 	bool probe_unacked;
110 	u16 cong_link_cnt;
111 	u16 snt_unacked;
112 	u16 snd_win;
113 	u16 peer_caps;
114 	u16 rcv_unacked;
115 	u16 rcv_win;
116 	struct sockaddr_tipc peer;
117 	struct rhash_head node;
118 	struct tipc_mc_method mc_method;
119 	struct rcu_head rcu;
120 	struct tipc_group *group;
121 	u32 oneway;
122 	u16 snd_backlog;
123 	bool expect_ack;
124 	bool nodelay;
125 	bool group_is_open;
126 };
127 
128 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
129 static void tipc_data_ready(struct sock *sk);
130 static void tipc_write_space(struct sock *sk);
131 static void tipc_sock_destruct(struct sock *sk);
132 static int tipc_release(struct socket *sock);
133 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
134 		       bool kern);
135 static void tipc_sk_timeout(struct timer_list *t);
136 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
137 			   struct tipc_name_seq const *seq);
138 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
139 			    struct tipc_name_seq const *seq);
140 static int tipc_sk_leave(struct tipc_sock *tsk);
141 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
142 static int tipc_sk_insert(struct tipc_sock *tsk);
143 static void tipc_sk_remove(struct tipc_sock *tsk);
144 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
145 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
146 static void tipc_sk_push_backlog(struct tipc_sock *tsk);
147 
148 static const struct proto_ops packet_ops;
149 static const struct proto_ops stream_ops;
150 static const struct proto_ops msg_ops;
151 static struct proto tipc_proto;
152 static const struct rhashtable_params tsk_rht_params;
153 
154 static u32 tsk_own_node(struct tipc_sock *tsk)
155 {
156 	return msg_prevnode(&tsk->phdr);
157 }
158 
159 static u32 tsk_peer_node(struct tipc_sock *tsk)
160 {
161 	return msg_destnode(&tsk->phdr);
162 }
163 
164 static u32 tsk_peer_port(struct tipc_sock *tsk)
165 {
166 	return msg_destport(&tsk->phdr);
167 }
168 
169 static  bool tsk_unreliable(struct tipc_sock *tsk)
170 {
171 	return msg_src_droppable(&tsk->phdr) != 0;
172 }
173 
174 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
175 {
176 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
177 }
178 
179 static bool tsk_unreturnable(struct tipc_sock *tsk)
180 {
181 	return msg_dest_droppable(&tsk->phdr) != 0;
182 }
183 
184 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
185 {
186 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
187 }
188 
189 static int tsk_importance(struct tipc_sock *tsk)
190 {
191 	return msg_importance(&tsk->phdr);
192 }
193 
194 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
195 {
196 	if (imp > TIPC_CRITICAL_IMPORTANCE)
197 		return -EINVAL;
198 	msg_set_importance(&tsk->phdr, (u32)imp);
199 	return 0;
200 }
201 
202 static struct tipc_sock *tipc_sk(const struct sock *sk)
203 {
204 	return container_of(sk, struct tipc_sock, sk);
205 }
206 
207 static bool tsk_conn_cong(struct tipc_sock *tsk)
208 {
209 	return tsk->snt_unacked > tsk->snd_win;
210 }
211 
212 static u16 tsk_blocks(int len)
213 {
214 	return ((len / FLOWCTL_BLK_SZ) + 1);
215 }
216 
217 /* tsk_blocks(): translate a buffer size in bytes to number of
218  * advertisable blocks, taking into account the ratio truesize(len)/len
219  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
220  */
221 static u16 tsk_adv_blocks(int len)
222 {
223 	return len / FLOWCTL_BLK_SZ / 4;
224 }
225 
226 /* tsk_inc(): increment counter for sent or received data
227  * - If block based flow control is not supported by peer we
228  *   fall back to message based ditto, incrementing the counter
229  */
230 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
231 {
232 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
233 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
234 	return 1;
235 }
236 
237 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
238  */
239 static void tsk_set_nagle(struct tipc_sock *tsk)
240 {
241 	struct sock *sk = &tsk->sk;
242 
243 	tsk->maxnagle = 0;
244 	if (sk->sk_type != SOCK_STREAM)
245 		return;
246 	if (tsk->nodelay)
247 		return;
248 	if (!(tsk->peer_caps & TIPC_NAGLE))
249 		return;
250 	/* Limit node local buffer size to avoid receive queue overflow */
251 	if (tsk->max_pkt == MAX_MSG_SIZE)
252 		tsk->maxnagle = 1500;
253 	else
254 		tsk->maxnagle = tsk->max_pkt;
255 }
256 
257 /**
258  * tsk_advance_rx_queue - discard first buffer in socket receive queue
259  *
260  * Caller must hold socket lock
261  */
262 static void tsk_advance_rx_queue(struct sock *sk)
263 {
264 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
265 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
266 }
267 
268 /* tipc_sk_respond() : send response message back to sender
269  */
270 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
271 {
272 	u32 selector;
273 	u32 dnode;
274 	u32 onode = tipc_own_addr(sock_net(sk));
275 
276 	if (!tipc_msg_reverse(onode, &skb, err))
277 		return;
278 
279 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
280 	dnode = msg_destnode(buf_msg(skb));
281 	selector = msg_origport(buf_msg(skb));
282 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
283 }
284 
285 /**
286  * tsk_rej_rx_queue - reject all buffers in socket receive queue
287  *
288  * Caller must hold socket lock
289  */
290 static void tsk_rej_rx_queue(struct sock *sk)
291 {
292 	struct sk_buff *skb;
293 
294 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
295 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
296 }
297 
298 static bool tipc_sk_connected(struct sock *sk)
299 {
300 	return sk->sk_state == TIPC_ESTABLISHED;
301 }
302 
303 /* tipc_sk_type_connectionless - check if the socket is datagram socket
304  * @sk: socket
305  *
306  * Returns true if connection less, false otherwise
307  */
308 static bool tipc_sk_type_connectionless(struct sock *sk)
309 {
310 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
311 }
312 
313 /* tsk_peer_msg - verify if message was sent by connected port's peer
314  *
315  * Handles cases where the node's network address has changed from
316  * the default of <0.0.0> to its configured setting.
317  */
318 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
319 {
320 	struct sock *sk = &tsk->sk;
321 	u32 self = tipc_own_addr(sock_net(sk));
322 	u32 peer_port = tsk_peer_port(tsk);
323 	u32 orig_node, peer_node;
324 
325 	if (unlikely(!tipc_sk_connected(sk)))
326 		return false;
327 
328 	if (unlikely(msg_origport(msg) != peer_port))
329 		return false;
330 
331 	orig_node = msg_orignode(msg);
332 	peer_node = tsk_peer_node(tsk);
333 
334 	if (likely(orig_node == peer_node))
335 		return true;
336 
337 	if (!orig_node && peer_node == self)
338 		return true;
339 
340 	if (!peer_node && orig_node == self)
341 		return true;
342 
343 	return false;
344 }
345 
346 /* tipc_set_sk_state - set the sk_state of the socket
347  * @sk: socket
348  *
349  * Caller must hold socket lock
350  *
351  * Returns 0 on success, errno otherwise
352  */
353 static int tipc_set_sk_state(struct sock *sk, int state)
354 {
355 	int oldsk_state = sk->sk_state;
356 	int res = -EINVAL;
357 
358 	switch (state) {
359 	case TIPC_OPEN:
360 		res = 0;
361 		break;
362 	case TIPC_LISTEN:
363 	case TIPC_CONNECTING:
364 		if (oldsk_state == TIPC_OPEN)
365 			res = 0;
366 		break;
367 	case TIPC_ESTABLISHED:
368 		if (oldsk_state == TIPC_CONNECTING ||
369 		    oldsk_state == TIPC_OPEN)
370 			res = 0;
371 		break;
372 	case TIPC_DISCONNECTING:
373 		if (oldsk_state == TIPC_CONNECTING ||
374 		    oldsk_state == TIPC_ESTABLISHED)
375 			res = 0;
376 		break;
377 	}
378 
379 	if (!res)
380 		sk->sk_state = state;
381 
382 	return res;
383 }
384 
385 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
386 {
387 	struct sock *sk = sock->sk;
388 	int err = sock_error(sk);
389 	int typ = sock->type;
390 
391 	if (err)
392 		return err;
393 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
394 		if (sk->sk_state == TIPC_DISCONNECTING)
395 			return -EPIPE;
396 		else if (!tipc_sk_connected(sk))
397 			return -ENOTCONN;
398 	}
399 	if (!*timeout)
400 		return -EAGAIN;
401 	if (signal_pending(current))
402 		return sock_intr_errno(*timeout);
403 
404 	return 0;
405 }
406 
407 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
408 ({                                                                             \
409 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
410 	struct sock *sk_;						       \
411 	int rc_;							       \
412 									       \
413 	while ((rc_ = !(condition_))) {					       \
414 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
415 		smp_rmb();                                                     \
416 		sk_ = (sock_)->sk;					       \
417 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
418 		if (rc_)						       \
419 			break;						       \
420 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
421 		release_sock(sk_);					       \
422 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
423 		sched_annotate_sleep();				               \
424 		lock_sock(sk_);						       \
425 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
426 	}								       \
427 	rc_;								       \
428 })
429 
430 /**
431  * tipc_sk_create - create a TIPC socket
432  * @net: network namespace (must be default network)
433  * @sock: pre-allocated socket structure
434  * @protocol: protocol indicator (must be 0)
435  * @kern: caused by kernel or by userspace?
436  *
437  * This routine creates additional data structures used by the TIPC socket,
438  * initializes them, and links them together.
439  *
440  * Returns 0 on success, errno otherwise
441  */
442 static int tipc_sk_create(struct net *net, struct socket *sock,
443 			  int protocol, int kern)
444 {
445 	const struct proto_ops *ops;
446 	struct sock *sk;
447 	struct tipc_sock *tsk;
448 	struct tipc_msg *msg;
449 
450 	/* Validate arguments */
451 	if (unlikely(protocol != 0))
452 		return -EPROTONOSUPPORT;
453 
454 	switch (sock->type) {
455 	case SOCK_STREAM:
456 		ops = &stream_ops;
457 		break;
458 	case SOCK_SEQPACKET:
459 		ops = &packet_ops;
460 		break;
461 	case SOCK_DGRAM:
462 	case SOCK_RDM:
463 		ops = &msg_ops;
464 		break;
465 	default:
466 		return -EPROTOTYPE;
467 	}
468 
469 	/* Allocate socket's protocol area */
470 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
471 	if (sk == NULL)
472 		return -ENOMEM;
473 
474 	tsk = tipc_sk(sk);
475 	tsk->max_pkt = MAX_PKT_DEFAULT;
476 	tsk->maxnagle = 0;
477 	INIT_LIST_HEAD(&tsk->publications);
478 	INIT_LIST_HEAD(&tsk->cong_links);
479 	msg = &tsk->phdr;
480 
481 	/* Finish initializing socket data structures */
482 	sock->ops = ops;
483 	sock_init_data(sock, sk);
484 	tipc_set_sk_state(sk, TIPC_OPEN);
485 	if (tipc_sk_insert(tsk)) {
486 		pr_warn("Socket create failed; port number exhausted\n");
487 		return -EINVAL;
488 	}
489 
490 	/* Ensure tsk is visible before we read own_addr. */
491 	smp_mb();
492 
493 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
494 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
495 
496 	msg_set_origport(msg, tsk->portid);
497 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
498 	sk->sk_shutdown = 0;
499 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
500 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
501 	sk->sk_data_ready = tipc_data_ready;
502 	sk->sk_write_space = tipc_write_space;
503 	sk->sk_destruct = tipc_sock_destruct;
504 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
505 	tsk->group_is_open = true;
506 	atomic_set(&tsk->dupl_rcvcnt, 0);
507 
508 	/* Start out with safe limits until we receive an advertised window */
509 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
510 	tsk->rcv_win = tsk->snd_win;
511 
512 	if (tipc_sk_type_connectionless(sk)) {
513 		tsk_set_unreturnable(tsk, true);
514 		if (sock->type == SOCK_DGRAM)
515 			tsk_set_unreliable(tsk, true);
516 	}
517 	__skb_queue_head_init(&tsk->mc_method.deferredq);
518 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
519 	return 0;
520 }
521 
522 static void tipc_sk_callback(struct rcu_head *head)
523 {
524 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
525 
526 	sock_put(&tsk->sk);
527 }
528 
529 /* Caller should hold socket lock for the socket. */
530 static void __tipc_shutdown(struct socket *sock, int error)
531 {
532 	struct sock *sk = sock->sk;
533 	struct tipc_sock *tsk = tipc_sk(sk);
534 	struct net *net = sock_net(sk);
535 	long timeout = CONN_TIMEOUT_DEFAULT;
536 	u32 dnode = tsk_peer_node(tsk);
537 	struct sk_buff *skb;
538 
539 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
540 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
541 					    !tsk_conn_cong(tsk)));
542 
543 	/* Push out unsent messages or remove if pending SYN */
544 	skb = skb_peek(&sk->sk_write_queue);
545 	if (skb && !msg_is_syn(buf_msg(skb)))
546 		tipc_sk_push_backlog(tsk);
547 	else
548 		__skb_queue_purge(&sk->sk_write_queue);
549 
550 	/* Reject all unreceived messages, except on an active connection
551 	 * (which disconnects locally & sends a 'FIN+' to peer).
552 	 */
553 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
554 		if (TIPC_SKB_CB(skb)->bytes_read) {
555 			kfree_skb(skb);
556 			continue;
557 		}
558 		if (!tipc_sk_type_connectionless(sk) &&
559 		    sk->sk_state != TIPC_DISCONNECTING) {
560 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
561 			tipc_node_remove_conn(net, dnode, tsk->portid);
562 		}
563 		tipc_sk_respond(sk, skb, error);
564 	}
565 
566 	if (tipc_sk_type_connectionless(sk))
567 		return;
568 
569 	if (sk->sk_state != TIPC_DISCONNECTING) {
570 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
571 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
572 				      tsk_own_node(tsk), tsk_peer_port(tsk),
573 				      tsk->portid, error);
574 		if (skb)
575 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
576 		tipc_node_remove_conn(net, dnode, tsk->portid);
577 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
578 	}
579 }
580 
581 /**
582  * tipc_release - destroy a TIPC socket
583  * @sock: socket to destroy
584  *
585  * This routine cleans up any messages that are still queued on the socket.
586  * For DGRAM and RDM socket types, all queued messages are rejected.
587  * For SEQPACKET and STREAM socket types, the first message is rejected
588  * and any others are discarded.  (If the first message on a STREAM socket
589  * is partially-read, it is discarded and the next one is rejected instead.)
590  *
591  * NOTE: Rejected messages are not necessarily returned to the sender!  They
592  * are returned or discarded according to the "destination droppable" setting
593  * specified for the message by the sender.
594  *
595  * Returns 0 on success, errno otherwise
596  */
597 static int tipc_release(struct socket *sock)
598 {
599 	struct sock *sk = sock->sk;
600 	struct tipc_sock *tsk;
601 
602 	/*
603 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
604 	 * releases a pre-allocated child socket that was never used)
605 	 */
606 	if (sk == NULL)
607 		return 0;
608 
609 	tsk = tipc_sk(sk);
610 	lock_sock(sk);
611 
612 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
613 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
614 	sk->sk_shutdown = SHUTDOWN_MASK;
615 	tipc_sk_leave(tsk);
616 	tipc_sk_withdraw(tsk, 0, NULL);
617 	__skb_queue_purge(&tsk->mc_method.deferredq);
618 	sk_stop_timer(sk, &sk->sk_timer);
619 	tipc_sk_remove(tsk);
620 
621 	sock_orphan(sk);
622 	/* Reject any messages that accumulated in backlog queue */
623 	release_sock(sk);
624 	tipc_dest_list_purge(&tsk->cong_links);
625 	tsk->cong_link_cnt = 0;
626 	call_rcu(&tsk->rcu, tipc_sk_callback);
627 	sock->sk = NULL;
628 
629 	return 0;
630 }
631 
632 /**
633  * tipc_bind - associate or disassocate TIPC name(s) with a socket
634  * @sock: socket structure
635  * @uaddr: socket address describing name(s) and desired operation
636  * @uaddr_len: size of socket address data structure
637  *
638  * Name and name sequence binding is indicated using a positive scope value;
639  * a negative scope value unbinds the specified name.  Specifying no name
640  * (i.e. a socket address length of 0) unbinds all names from the socket.
641  *
642  * Returns 0 on success, errno otherwise
643  *
644  * NOTE: This routine doesn't need to take the socket lock since it doesn't
645  *       access any non-constant socket information.
646  */
647 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
648 		     int uaddr_len)
649 {
650 	struct sock *sk = sock->sk;
651 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
652 	struct tipc_sock *tsk = tipc_sk(sk);
653 	int res = -EINVAL;
654 
655 	lock_sock(sk);
656 	if (unlikely(!uaddr_len)) {
657 		res = tipc_sk_withdraw(tsk, 0, NULL);
658 		goto exit;
659 	}
660 	if (tsk->group) {
661 		res = -EACCES;
662 		goto exit;
663 	}
664 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
665 		res = -EINVAL;
666 		goto exit;
667 	}
668 	if (addr->family != AF_TIPC) {
669 		res = -EAFNOSUPPORT;
670 		goto exit;
671 	}
672 
673 	if (addr->addrtype == TIPC_ADDR_NAME)
674 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
675 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
676 		res = -EAFNOSUPPORT;
677 		goto exit;
678 	}
679 
680 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
681 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
682 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
683 		res = -EACCES;
684 		goto exit;
685 	}
686 
687 	res = (addr->scope >= 0) ?
688 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
689 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
690 exit:
691 	release_sock(sk);
692 	return res;
693 }
694 
695 /**
696  * tipc_getname - get port ID of socket or peer socket
697  * @sock: socket structure
698  * @uaddr: area for returned socket address
699  * @uaddr_len: area for returned length of socket address
700  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
701  *
702  * Returns 0 on success, errno otherwise
703  *
704  * NOTE: This routine doesn't need to take the socket lock since it only
705  *       accesses socket information that is unchanging (or which changes in
706  *       a completely predictable manner).
707  */
708 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
709 			int peer)
710 {
711 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
712 	struct sock *sk = sock->sk;
713 	struct tipc_sock *tsk = tipc_sk(sk);
714 
715 	memset(addr, 0, sizeof(*addr));
716 	if (peer) {
717 		if ((!tipc_sk_connected(sk)) &&
718 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
719 			return -ENOTCONN;
720 		addr->addr.id.ref = tsk_peer_port(tsk);
721 		addr->addr.id.node = tsk_peer_node(tsk);
722 	} else {
723 		addr->addr.id.ref = tsk->portid;
724 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
725 	}
726 
727 	addr->addrtype = TIPC_ADDR_ID;
728 	addr->family = AF_TIPC;
729 	addr->scope = 0;
730 	addr->addr.name.domain = 0;
731 
732 	return sizeof(*addr);
733 }
734 
735 /**
736  * tipc_poll - read and possibly block on pollmask
737  * @file: file structure associated with the socket
738  * @sock: socket for which to calculate the poll bits
739  * @wait: ???
740  *
741  * Returns pollmask value
742  *
743  * COMMENTARY:
744  * It appears that the usual socket locking mechanisms are not useful here
745  * since the pollmask info is potentially out-of-date the moment this routine
746  * exits.  TCP and other protocols seem to rely on higher level poll routines
747  * to handle any preventable race conditions, so TIPC will do the same ...
748  *
749  * IMPORTANT: The fact that a read or write operation is indicated does NOT
750  * imply that the operation will succeed, merely that it should be performed
751  * and will not block.
752  */
753 static __poll_t tipc_poll(struct file *file, struct socket *sock,
754 			      poll_table *wait)
755 {
756 	struct sock *sk = sock->sk;
757 	struct tipc_sock *tsk = tipc_sk(sk);
758 	__poll_t revents = 0;
759 
760 	sock_poll_wait(file, sock, wait);
761 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
762 
763 	if (sk->sk_shutdown & RCV_SHUTDOWN)
764 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
765 	if (sk->sk_shutdown == SHUTDOWN_MASK)
766 		revents |= EPOLLHUP;
767 
768 	switch (sk->sk_state) {
769 	case TIPC_ESTABLISHED:
770 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
771 			revents |= EPOLLOUT;
772 		/* fall through */
773 	case TIPC_LISTEN:
774 	case TIPC_CONNECTING:
775 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
776 			revents |= EPOLLIN | EPOLLRDNORM;
777 		break;
778 	case TIPC_OPEN:
779 		if (tsk->group_is_open && !tsk->cong_link_cnt)
780 			revents |= EPOLLOUT;
781 		if (!tipc_sk_type_connectionless(sk))
782 			break;
783 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
784 			break;
785 		revents |= EPOLLIN | EPOLLRDNORM;
786 		break;
787 	case TIPC_DISCONNECTING:
788 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
789 		break;
790 	}
791 	return revents;
792 }
793 
794 /**
795  * tipc_sendmcast - send multicast message
796  * @sock: socket structure
797  * @seq: destination address
798  * @msg: message to send
799  * @dlen: length of data to send
800  * @timeout: timeout to wait for wakeup
801  *
802  * Called from function tipc_sendmsg(), which has done all sanity checks
803  * Returns the number of bytes sent on success, or errno
804  */
805 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
806 			  struct msghdr *msg, size_t dlen, long timeout)
807 {
808 	struct sock *sk = sock->sk;
809 	struct tipc_sock *tsk = tipc_sk(sk);
810 	struct tipc_msg *hdr = &tsk->phdr;
811 	struct net *net = sock_net(sk);
812 	int mtu = tipc_bcast_get_mtu(net);
813 	struct tipc_mc_method *method = &tsk->mc_method;
814 	struct sk_buff_head pkts;
815 	struct tipc_nlist dsts;
816 	int rc;
817 
818 	if (tsk->group)
819 		return -EACCES;
820 
821 	/* Block or return if any destination link is congested */
822 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
823 	if (unlikely(rc))
824 		return rc;
825 
826 	/* Lookup destination nodes */
827 	tipc_nlist_init(&dsts, tipc_own_addr(net));
828 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
829 				      seq->upper, &dsts);
830 	if (!dsts.local && !dsts.remote)
831 		return -EHOSTUNREACH;
832 
833 	/* Build message header */
834 	msg_set_type(hdr, TIPC_MCAST_MSG);
835 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
836 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
837 	msg_set_destport(hdr, 0);
838 	msg_set_destnode(hdr, 0);
839 	msg_set_nametype(hdr, seq->type);
840 	msg_set_namelower(hdr, seq->lower);
841 	msg_set_nameupper(hdr, seq->upper);
842 
843 	/* Build message as chain of buffers */
844 	__skb_queue_head_init(&pkts);
845 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
846 
847 	/* Send message if build was successful */
848 	if (unlikely(rc == dlen)) {
849 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
850 					TIPC_DUMP_SK_SNDQ, " ");
851 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
852 				     &tsk->cong_link_cnt);
853 	}
854 
855 	tipc_nlist_purge(&dsts);
856 
857 	return rc ? rc : dlen;
858 }
859 
860 /**
861  * tipc_send_group_msg - send a message to a member in the group
862  * @net: network namespace
863  * @m: message to send
864  * @mb: group member
865  * @dnode: destination node
866  * @dport: destination port
867  * @dlen: total length of message data
868  */
869 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
870 			       struct msghdr *m, struct tipc_member *mb,
871 			       u32 dnode, u32 dport, int dlen)
872 {
873 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
874 	struct tipc_mc_method *method = &tsk->mc_method;
875 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
876 	struct tipc_msg *hdr = &tsk->phdr;
877 	struct sk_buff_head pkts;
878 	int mtu, rc;
879 
880 	/* Complete message header */
881 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
882 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
883 	msg_set_destport(hdr, dport);
884 	msg_set_destnode(hdr, dnode);
885 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
886 
887 	/* Build message as chain of buffers */
888 	__skb_queue_head_init(&pkts);
889 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
890 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
891 	if (unlikely(rc != dlen))
892 		return rc;
893 
894 	/* Send message */
895 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
896 	if (unlikely(rc == -ELINKCONG)) {
897 		tipc_dest_push(&tsk->cong_links, dnode, 0);
898 		tsk->cong_link_cnt++;
899 	}
900 
901 	/* Update send window */
902 	tipc_group_update_member(mb, blks);
903 
904 	/* A broadcast sent within next EXPIRE period must follow same path */
905 	method->rcast = true;
906 	method->mandatory = true;
907 	return dlen;
908 }
909 
910 /**
911  * tipc_send_group_unicast - send message to a member in the group
912  * @sock: socket structure
913  * @m: message to send
914  * @dlen: total length of message data
915  * @timeout: timeout to wait for wakeup
916  *
917  * Called from function tipc_sendmsg(), which has done all sanity checks
918  * Returns the number of bytes sent on success, or errno
919  */
920 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
921 				   int dlen, long timeout)
922 {
923 	struct sock *sk = sock->sk;
924 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
925 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
926 	struct tipc_sock *tsk = tipc_sk(sk);
927 	struct net *net = sock_net(sk);
928 	struct tipc_member *mb = NULL;
929 	u32 node, port;
930 	int rc;
931 
932 	node = dest->addr.id.node;
933 	port = dest->addr.id.ref;
934 	if (!port && !node)
935 		return -EHOSTUNREACH;
936 
937 	/* Block or return if destination link or member is congested */
938 	rc = tipc_wait_for_cond(sock, &timeout,
939 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
940 				tsk->group &&
941 				!tipc_group_cong(tsk->group, node, port, blks,
942 						 &mb));
943 	if (unlikely(rc))
944 		return rc;
945 
946 	if (unlikely(!mb))
947 		return -EHOSTUNREACH;
948 
949 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
950 
951 	return rc ? rc : dlen;
952 }
953 
954 /**
955  * tipc_send_group_anycast - send message to any member with given identity
956  * @sock: socket structure
957  * @m: message to send
958  * @dlen: total length of message data
959  * @timeout: timeout to wait for wakeup
960  *
961  * Called from function tipc_sendmsg(), which has done all sanity checks
962  * Returns the number of bytes sent on success, or errno
963  */
964 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
965 				   int dlen, long timeout)
966 {
967 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
968 	struct sock *sk = sock->sk;
969 	struct tipc_sock *tsk = tipc_sk(sk);
970 	struct list_head *cong_links = &tsk->cong_links;
971 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
972 	struct tipc_msg *hdr = &tsk->phdr;
973 	struct tipc_member *first = NULL;
974 	struct tipc_member *mbr = NULL;
975 	struct net *net = sock_net(sk);
976 	u32 node, port, exclude;
977 	struct list_head dsts;
978 	u32 type, inst, scope;
979 	int lookups = 0;
980 	int dstcnt, rc;
981 	bool cong;
982 
983 	INIT_LIST_HEAD(&dsts);
984 
985 	type = msg_nametype(hdr);
986 	inst = dest->addr.name.name.instance;
987 	scope = msg_lookup_scope(hdr);
988 
989 	while (++lookups < 4) {
990 		exclude = tipc_group_exclude(tsk->group);
991 
992 		first = NULL;
993 
994 		/* Look for a non-congested destination member, if any */
995 		while (1) {
996 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
997 						 &dstcnt, exclude, false))
998 				return -EHOSTUNREACH;
999 			tipc_dest_pop(&dsts, &node, &port);
1000 			cong = tipc_group_cong(tsk->group, node, port, blks,
1001 					       &mbr);
1002 			if (!cong)
1003 				break;
1004 			if (mbr == first)
1005 				break;
1006 			if (!first)
1007 				first = mbr;
1008 		}
1009 
1010 		/* Start over if destination was not in member list */
1011 		if (unlikely(!mbr))
1012 			continue;
1013 
1014 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1015 			break;
1016 
1017 		/* Block or return if destination link or member is congested */
1018 		rc = tipc_wait_for_cond(sock, &timeout,
1019 					!tipc_dest_find(cong_links, node, 0) &&
1020 					tsk->group &&
1021 					!tipc_group_cong(tsk->group, node, port,
1022 							 blks, &mbr));
1023 		if (unlikely(rc))
1024 			return rc;
1025 
1026 		/* Send, unless destination disappeared while waiting */
1027 		if (likely(mbr))
1028 			break;
1029 	}
1030 
1031 	if (unlikely(lookups >= 4))
1032 		return -EHOSTUNREACH;
1033 
1034 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1035 
1036 	return rc ? rc : dlen;
1037 }
1038 
1039 /**
1040  * tipc_send_group_bcast - send message to all members in communication group
1041  * @sk: socket structure
1042  * @m: message to send
1043  * @dlen: total length of message data
1044  * @timeout: timeout to wait for wakeup
1045  *
1046  * Called from function tipc_sendmsg(), which has done all sanity checks
1047  * Returns the number of bytes sent on success, or errno
1048  */
1049 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1050 				 int dlen, long timeout)
1051 {
1052 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1053 	struct sock *sk = sock->sk;
1054 	struct net *net = sock_net(sk);
1055 	struct tipc_sock *tsk = tipc_sk(sk);
1056 	struct tipc_nlist *dsts;
1057 	struct tipc_mc_method *method = &tsk->mc_method;
1058 	bool ack = method->mandatory && method->rcast;
1059 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1060 	struct tipc_msg *hdr = &tsk->phdr;
1061 	int mtu = tipc_bcast_get_mtu(net);
1062 	struct sk_buff_head pkts;
1063 	int rc = -EHOSTUNREACH;
1064 
1065 	/* Block or return if any destination link or member is congested */
1066 	rc = tipc_wait_for_cond(sock, &timeout,
1067 				!tsk->cong_link_cnt && tsk->group &&
1068 				!tipc_group_bc_cong(tsk->group, blks));
1069 	if (unlikely(rc))
1070 		return rc;
1071 
1072 	dsts = tipc_group_dests(tsk->group);
1073 	if (!dsts->local && !dsts->remote)
1074 		return -EHOSTUNREACH;
1075 
1076 	/* Complete message header */
1077 	if (dest) {
1078 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1079 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1080 	} else {
1081 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1082 		msg_set_nameinst(hdr, 0);
1083 	}
1084 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1085 	msg_set_destport(hdr, 0);
1086 	msg_set_destnode(hdr, 0);
1087 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1088 
1089 	/* Avoid getting stuck with repeated forced replicasts */
1090 	msg_set_grp_bc_ack_req(hdr, ack);
1091 
1092 	/* Build message as chain of buffers */
1093 	__skb_queue_head_init(&pkts);
1094 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1095 	if (unlikely(rc != dlen))
1096 		return rc;
1097 
1098 	/* Send message */
1099 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1100 	if (unlikely(rc))
1101 		return rc;
1102 
1103 	/* Update broadcast sequence number and send windows */
1104 	tipc_group_update_bc_members(tsk->group, blks, ack);
1105 
1106 	/* Broadcast link is now free to choose method for next broadcast */
1107 	method->mandatory = false;
1108 	method->expires = jiffies;
1109 
1110 	return dlen;
1111 }
1112 
1113 /**
1114  * tipc_send_group_mcast - send message to all members with given identity
1115  * @sock: socket structure
1116  * @m: message to send
1117  * @dlen: total length of message data
1118  * @timeout: timeout to wait for wakeup
1119  *
1120  * Called from function tipc_sendmsg(), which has done all sanity checks
1121  * Returns the number of bytes sent on success, or errno
1122  */
1123 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1124 				 int dlen, long timeout)
1125 {
1126 	struct sock *sk = sock->sk;
1127 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1128 	struct tipc_sock *tsk = tipc_sk(sk);
1129 	struct tipc_group *grp = tsk->group;
1130 	struct tipc_msg *hdr = &tsk->phdr;
1131 	struct net *net = sock_net(sk);
1132 	u32 type, inst, scope, exclude;
1133 	struct list_head dsts;
1134 	u32 dstcnt;
1135 
1136 	INIT_LIST_HEAD(&dsts);
1137 
1138 	type = msg_nametype(hdr);
1139 	inst = dest->addr.name.name.instance;
1140 	scope = msg_lookup_scope(hdr);
1141 	exclude = tipc_group_exclude(grp);
1142 
1143 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1144 				 &dstcnt, exclude, true))
1145 		return -EHOSTUNREACH;
1146 
1147 	if (dstcnt == 1) {
1148 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1149 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1150 	}
1151 
1152 	tipc_dest_list_purge(&dsts);
1153 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1154 }
1155 
1156 /**
1157  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1158  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1159  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1160  *
1161  * Multi-threaded: parallel calls with reference to same queues may occur
1162  */
1163 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1164 		       struct sk_buff_head *inputq)
1165 {
1166 	u32 self = tipc_own_addr(net);
1167 	u32 type, lower, upper, scope;
1168 	struct sk_buff *skb, *_skb;
1169 	u32 portid, onode;
1170 	struct sk_buff_head tmpq;
1171 	struct list_head dports;
1172 	struct tipc_msg *hdr;
1173 	int user, mtyp, hlen;
1174 	bool exact;
1175 
1176 	__skb_queue_head_init(&tmpq);
1177 	INIT_LIST_HEAD(&dports);
1178 
1179 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1180 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1181 		hdr = buf_msg(skb);
1182 		user = msg_user(hdr);
1183 		mtyp = msg_type(hdr);
1184 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1185 		onode = msg_orignode(hdr);
1186 		type = msg_nametype(hdr);
1187 
1188 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1189 			spin_lock_bh(&inputq->lock);
1190 			if (skb_peek(arrvq) == skb) {
1191 				__skb_dequeue(arrvq);
1192 				__skb_queue_tail(inputq, skb);
1193 			}
1194 			kfree_skb(skb);
1195 			spin_unlock_bh(&inputq->lock);
1196 			continue;
1197 		}
1198 
1199 		/* Group messages require exact scope match */
1200 		if (msg_in_group(hdr)) {
1201 			lower = 0;
1202 			upper = ~0;
1203 			scope = msg_lookup_scope(hdr);
1204 			exact = true;
1205 		} else {
1206 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1207 			if (onode == self)
1208 				scope = TIPC_NODE_SCOPE;
1209 			else
1210 				scope = TIPC_CLUSTER_SCOPE;
1211 			exact = false;
1212 			lower = msg_namelower(hdr);
1213 			upper = msg_nameupper(hdr);
1214 		}
1215 
1216 		/* Create destination port list: */
1217 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1218 				       scope, exact, &dports);
1219 
1220 		/* Clone message per destination */
1221 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1222 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1223 			if (_skb) {
1224 				msg_set_destport(buf_msg(_skb), portid);
1225 				__skb_queue_tail(&tmpq, _skb);
1226 				continue;
1227 			}
1228 			pr_warn("Failed to clone mcast rcv buffer\n");
1229 		}
1230 		/* Append to inputq if not already done by other thread */
1231 		spin_lock_bh(&inputq->lock);
1232 		if (skb_peek(arrvq) == skb) {
1233 			skb_queue_splice_tail_init(&tmpq, inputq);
1234 			kfree_skb(__skb_dequeue(arrvq));
1235 		}
1236 		spin_unlock_bh(&inputq->lock);
1237 		__skb_queue_purge(&tmpq);
1238 		kfree_skb(skb);
1239 	}
1240 	tipc_sk_rcv(net, inputq);
1241 }
1242 
1243 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1244  *                         when socket is in Nagle mode
1245  */
1246 static void tipc_sk_push_backlog(struct tipc_sock *tsk)
1247 {
1248 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1249 	struct net *net = sock_net(&tsk->sk);
1250 	u32 dnode = tsk_peer_node(tsk);
1251 	int rc;
1252 
1253 	if (skb_queue_empty(txq) || tsk->cong_link_cnt)
1254 		return;
1255 
1256 	tsk->snt_unacked += tsk->snd_backlog;
1257 	tsk->snd_backlog = 0;
1258 	tsk->expect_ack = true;
1259 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1260 	if (rc == -ELINKCONG)
1261 		tsk->cong_link_cnt = 1;
1262 }
1263 
1264 /**
1265  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1266  * @tsk: receiving socket
1267  * @skb: pointer to message buffer.
1268  */
1269 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1270 				   struct sk_buff_head *inputq,
1271 				   struct sk_buff_head *xmitq)
1272 {
1273 	struct tipc_msg *hdr = buf_msg(skb);
1274 	u32 onode = tsk_own_node(tsk);
1275 	struct sock *sk = &tsk->sk;
1276 	int mtyp = msg_type(hdr);
1277 	bool was_cong;
1278 
1279 	/* Ignore if connection cannot be validated: */
1280 	if (!tsk_peer_msg(tsk, hdr)) {
1281 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1282 		goto exit;
1283 	}
1284 
1285 	if (unlikely(msg_errcode(hdr))) {
1286 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1287 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1288 				      tsk_peer_port(tsk));
1289 		sk->sk_state_change(sk);
1290 
1291 		/* State change is ignored if socket already awake,
1292 		 * - convert msg to abort msg and add to inqueue
1293 		 */
1294 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1295 		msg_set_type(hdr, TIPC_CONN_MSG);
1296 		msg_set_size(hdr, BASIC_H_SIZE);
1297 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1298 		__skb_queue_tail(inputq, skb);
1299 		return;
1300 	}
1301 
1302 	tsk->probe_unacked = false;
1303 
1304 	if (mtyp == CONN_PROBE) {
1305 		msg_set_type(hdr, CONN_PROBE_REPLY);
1306 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1307 			__skb_queue_tail(xmitq, skb);
1308 		return;
1309 	} else if (mtyp == CONN_ACK) {
1310 		was_cong = tsk_conn_cong(tsk);
1311 		tsk->expect_ack = false;
1312 		tipc_sk_push_backlog(tsk);
1313 		tsk->snt_unacked -= msg_conn_ack(hdr);
1314 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1315 			tsk->snd_win = msg_adv_win(hdr);
1316 		if (was_cong && !tsk_conn_cong(tsk))
1317 			sk->sk_write_space(sk);
1318 	} else if (mtyp != CONN_PROBE_REPLY) {
1319 		pr_warn("Received unknown CONN_PROTO msg\n");
1320 	}
1321 exit:
1322 	kfree_skb(skb);
1323 }
1324 
1325 /**
1326  * tipc_sendmsg - send message in connectionless manner
1327  * @sock: socket structure
1328  * @m: message to send
1329  * @dsz: amount of user data to be sent
1330  *
1331  * Message must have an destination specified explicitly.
1332  * Used for SOCK_RDM and SOCK_DGRAM messages,
1333  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1334  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1335  *
1336  * Returns the number of bytes sent on success, or errno otherwise
1337  */
1338 static int tipc_sendmsg(struct socket *sock,
1339 			struct msghdr *m, size_t dsz)
1340 {
1341 	struct sock *sk = sock->sk;
1342 	int ret;
1343 
1344 	lock_sock(sk);
1345 	ret = __tipc_sendmsg(sock, m, dsz);
1346 	release_sock(sk);
1347 
1348 	return ret;
1349 }
1350 
1351 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1352 {
1353 	struct sock *sk = sock->sk;
1354 	struct net *net = sock_net(sk);
1355 	struct tipc_sock *tsk = tipc_sk(sk);
1356 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1357 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1358 	struct list_head *clinks = &tsk->cong_links;
1359 	bool syn = !tipc_sk_type_connectionless(sk);
1360 	struct tipc_group *grp = tsk->group;
1361 	struct tipc_msg *hdr = &tsk->phdr;
1362 	struct tipc_name_seq *seq;
1363 	struct sk_buff_head pkts;
1364 	u32 dport, dnode = 0;
1365 	u32 type, inst;
1366 	int mtu, rc;
1367 
1368 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1369 		return -EMSGSIZE;
1370 
1371 	if (likely(dest)) {
1372 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1373 			return -EINVAL;
1374 		if (unlikely(dest->family != AF_TIPC))
1375 			return -EINVAL;
1376 	}
1377 
1378 	if (grp) {
1379 		if (!dest)
1380 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1381 		if (dest->addrtype == TIPC_ADDR_NAME)
1382 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1383 		if (dest->addrtype == TIPC_ADDR_ID)
1384 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1385 		if (dest->addrtype == TIPC_ADDR_MCAST)
1386 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1387 		return -EINVAL;
1388 	}
1389 
1390 	if (unlikely(!dest)) {
1391 		dest = &tsk->peer;
1392 		if (!syn && dest->family != AF_TIPC)
1393 			return -EDESTADDRREQ;
1394 	}
1395 
1396 	if (unlikely(syn)) {
1397 		if (sk->sk_state == TIPC_LISTEN)
1398 			return -EPIPE;
1399 		if (sk->sk_state != TIPC_OPEN)
1400 			return -EISCONN;
1401 		if (tsk->published)
1402 			return -EOPNOTSUPP;
1403 		if (dest->addrtype == TIPC_ADDR_NAME) {
1404 			tsk->conn_type = dest->addr.name.name.type;
1405 			tsk->conn_instance = dest->addr.name.name.instance;
1406 		}
1407 		msg_set_syn(hdr, 1);
1408 	}
1409 
1410 	seq = &dest->addr.nameseq;
1411 	if (dest->addrtype == TIPC_ADDR_MCAST)
1412 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1413 
1414 	if (dest->addrtype == TIPC_ADDR_NAME) {
1415 		type = dest->addr.name.name.type;
1416 		inst = dest->addr.name.name.instance;
1417 		dnode = dest->addr.name.domain;
1418 		msg_set_type(hdr, TIPC_NAMED_MSG);
1419 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1420 		msg_set_nametype(hdr, type);
1421 		msg_set_nameinst(hdr, inst);
1422 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1423 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1424 		msg_set_destnode(hdr, dnode);
1425 		msg_set_destport(hdr, dport);
1426 		if (unlikely(!dport && !dnode))
1427 			return -EHOSTUNREACH;
1428 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1429 		dnode = dest->addr.id.node;
1430 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1431 		msg_set_lookup_scope(hdr, 0);
1432 		msg_set_destnode(hdr, dnode);
1433 		msg_set_destport(hdr, dest->addr.id.ref);
1434 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1435 	} else {
1436 		return -EINVAL;
1437 	}
1438 
1439 	/* Block or return if destination link is congested */
1440 	rc = tipc_wait_for_cond(sock, &timeout,
1441 				!tipc_dest_find(clinks, dnode, 0));
1442 	if (unlikely(rc))
1443 		return rc;
1444 
1445 	__skb_queue_head_init(&pkts);
1446 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
1447 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1448 	if (unlikely(rc != dlen))
1449 		return rc;
1450 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1451 		return -ENOMEM;
1452 
1453 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1454 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1455 	if (unlikely(rc == -ELINKCONG)) {
1456 		tipc_dest_push(clinks, dnode, 0);
1457 		tsk->cong_link_cnt++;
1458 		rc = 0;
1459 	}
1460 
1461 	if (unlikely(syn && !rc))
1462 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1463 
1464 	return rc ? rc : dlen;
1465 }
1466 
1467 /**
1468  * tipc_sendstream - send stream-oriented data
1469  * @sock: socket structure
1470  * @m: data to send
1471  * @dsz: total length of data to be transmitted
1472  *
1473  * Used for SOCK_STREAM data.
1474  *
1475  * Returns the number of bytes sent on success (or partial success),
1476  * or errno if no data sent
1477  */
1478 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1479 {
1480 	struct sock *sk = sock->sk;
1481 	int ret;
1482 
1483 	lock_sock(sk);
1484 	ret = __tipc_sendstream(sock, m, dsz);
1485 	release_sock(sk);
1486 
1487 	return ret;
1488 }
1489 
1490 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1491 {
1492 	struct sock *sk = sock->sk;
1493 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1494 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1495 	struct sk_buff_head *txq = &sk->sk_write_queue;
1496 	struct tipc_sock *tsk = tipc_sk(sk);
1497 	struct tipc_msg *hdr = &tsk->phdr;
1498 	struct net *net = sock_net(sk);
1499 	u32 dnode = tsk_peer_node(tsk);
1500 	int maxnagle = tsk->maxnagle;
1501 	int maxpkt = tsk->max_pkt;
1502 	int send, sent = 0;
1503 	int blocks, rc = 0;
1504 
1505 	if (unlikely(dlen > INT_MAX))
1506 		return -EMSGSIZE;
1507 
1508 	/* Handle implicit connection setup */
1509 	if (unlikely(dest)) {
1510 		rc = __tipc_sendmsg(sock, m, dlen);
1511 		if (dlen && dlen == rc) {
1512 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1513 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1514 		}
1515 		return rc;
1516 	}
1517 
1518 	do {
1519 		rc = tipc_wait_for_cond(sock, &timeout,
1520 					(!tsk->cong_link_cnt &&
1521 					 !tsk_conn_cong(tsk) &&
1522 					 tipc_sk_connected(sk)));
1523 		if (unlikely(rc))
1524 			break;
1525 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1526 		blocks = tsk->snd_backlog;
1527 		if (tsk->oneway++ >= 4 && send <= maxnagle) {
1528 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1529 			if (unlikely(rc < 0))
1530 				break;
1531 			blocks += rc;
1532 			if (blocks <= 64 && tsk->expect_ack) {
1533 				tsk->snd_backlog = blocks;
1534 				sent += send;
1535 				break;
1536 			}
1537 			tsk->expect_ack = true;
1538 		} else {
1539 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1540 			if (unlikely(rc != send))
1541 				break;
1542 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1543 		}
1544 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1545 					 TIPC_DUMP_SK_SNDQ, " ");
1546 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1547 		if (unlikely(rc == -ELINKCONG)) {
1548 			tsk->cong_link_cnt = 1;
1549 			rc = 0;
1550 		}
1551 		if (likely(!rc)) {
1552 			tsk->snt_unacked += blocks;
1553 			tsk->snd_backlog = 0;
1554 			sent += send;
1555 		}
1556 	} while (sent < dlen && !rc);
1557 
1558 	return sent ? sent : rc;
1559 }
1560 
1561 /**
1562  * tipc_send_packet - send a connection-oriented message
1563  * @sock: socket structure
1564  * @m: message to send
1565  * @dsz: length of data to be transmitted
1566  *
1567  * Used for SOCK_SEQPACKET messages.
1568  *
1569  * Returns the number of bytes sent on success, or errno otherwise
1570  */
1571 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1572 {
1573 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1574 		return -EMSGSIZE;
1575 
1576 	return tipc_sendstream(sock, m, dsz);
1577 }
1578 
1579 /* tipc_sk_finish_conn - complete the setup of a connection
1580  */
1581 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1582 				u32 peer_node)
1583 {
1584 	struct sock *sk = &tsk->sk;
1585 	struct net *net = sock_net(sk);
1586 	struct tipc_msg *msg = &tsk->phdr;
1587 
1588 	msg_set_syn(msg, 0);
1589 	msg_set_destnode(msg, peer_node);
1590 	msg_set_destport(msg, peer_port);
1591 	msg_set_type(msg, TIPC_CONN_MSG);
1592 	msg_set_lookup_scope(msg, 0);
1593 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1594 
1595 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1596 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1597 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1598 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1599 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1600 	tsk_set_nagle(tsk);
1601 	__skb_queue_purge(&sk->sk_write_queue);
1602 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1603 		return;
1604 
1605 	/* Fall back to message based flow control */
1606 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1607 	tsk->snd_win = FLOWCTL_MSG_WIN;
1608 }
1609 
1610 /**
1611  * tipc_sk_set_orig_addr - capture sender's address for received message
1612  * @m: descriptor for message info
1613  * @hdr: received message header
1614  *
1615  * Note: Address is not captured if not requested by receiver.
1616  */
1617 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1618 {
1619 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1620 	struct tipc_msg *hdr = buf_msg(skb);
1621 
1622 	if (!srcaddr)
1623 		return;
1624 
1625 	srcaddr->sock.family = AF_TIPC;
1626 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1627 	srcaddr->sock.scope = 0;
1628 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1629 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1630 	srcaddr->sock.addr.name.domain = 0;
1631 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1632 
1633 	if (!msg_in_group(hdr))
1634 		return;
1635 
1636 	/* Group message users may also want to know sending member's id */
1637 	srcaddr->member.family = AF_TIPC;
1638 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1639 	srcaddr->member.scope = 0;
1640 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1641 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1642 	srcaddr->member.addr.name.domain = 0;
1643 	m->msg_namelen = sizeof(*srcaddr);
1644 }
1645 
1646 /**
1647  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1648  * @m: descriptor for message info
1649  * @skb: received message buffer
1650  * @tsk: TIPC port associated with message
1651  *
1652  * Note: Ancillary data is not captured if not requested by receiver.
1653  *
1654  * Returns 0 if successful, otherwise errno
1655  */
1656 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1657 				 struct tipc_sock *tsk)
1658 {
1659 	struct tipc_msg *msg;
1660 	u32 anc_data[3];
1661 	u32 err;
1662 	u32 dest_type;
1663 	int has_name;
1664 	int res;
1665 
1666 	if (likely(m->msg_controllen == 0))
1667 		return 0;
1668 	msg = buf_msg(skb);
1669 
1670 	/* Optionally capture errored message object(s) */
1671 	err = msg ? msg_errcode(msg) : 0;
1672 	if (unlikely(err)) {
1673 		anc_data[0] = err;
1674 		anc_data[1] = msg_data_sz(msg);
1675 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1676 		if (res)
1677 			return res;
1678 		if (anc_data[1]) {
1679 			if (skb_linearize(skb))
1680 				return -ENOMEM;
1681 			msg = buf_msg(skb);
1682 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1683 				       msg_data(msg));
1684 			if (res)
1685 				return res;
1686 		}
1687 	}
1688 
1689 	/* Optionally capture message destination object */
1690 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1691 	switch (dest_type) {
1692 	case TIPC_NAMED_MSG:
1693 		has_name = 1;
1694 		anc_data[0] = msg_nametype(msg);
1695 		anc_data[1] = msg_namelower(msg);
1696 		anc_data[2] = msg_namelower(msg);
1697 		break;
1698 	case TIPC_MCAST_MSG:
1699 		has_name = 1;
1700 		anc_data[0] = msg_nametype(msg);
1701 		anc_data[1] = msg_namelower(msg);
1702 		anc_data[2] = msg_nameupper(msg);
1703 		break;
1704 	case TIPC_CONN_MSG:
1705 		has_name = (tsk->conn_type != 0);
1706 		anc_data[0] = tsk->conn_type;
1707 		anc_data[1] = tsk->conn_instance;
1708 		anc_data[2] = tsk->conn_instance;
1709 		break;
1710 	default:
1711 		has_name = 0;
1712 	}
1713 	if (has_name) {
1714 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1715 		if (res)
1716 			return res;
1717 	}
1718 
1719 	return 0;
1720 }
1721 
1722 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1723 {
1724 	struct sock *sk = &tsk->sk;
1725 	struct net *net = sock_net(sk);
1726 	struct sk_buff *skb = NULL;
1727 	struct tipc_msg *msg;
1728 	u32 peer_port = tsk_peer_port(tsk);
1729 	u32 dnode = tsk_peer_node(tsk);
1730 
1731 	if (!tipc_sk_connected(sk))
1732 		return;
1733 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1734 			      dnode, tsk_own_node(tsk), peer_port,
1735 			      tsk->portid, TIPC_OK);
1736 	if (!skb)
1737 		return;
1738 	msg = buf_msg(skb);
1739 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1740 	tsk->rcv_unacked = 0;
1741 
1742 	/* Adjust to and advertize the correct window limit */
1743 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1744 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1745 		msg_set_adv_win(msg, tsk->rcv_win);
1746 	}
1747 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1748 }
1749 
1750 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1751 {
1752 	struct sock *sk = sock->sk;
1753 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1754 	long timeo = *timeop;
1755 	int err = sock_error(sk);
1756 
1757 	if (err)
1758 		return err;
1759 
1760 	for (;;) {
1761 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1762 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1763 				err = -ENOTCONN;
1764 				break;
1765 			}
1766 			add_wait_queue(sk_sleep(sk), &wait);
1767 			release_sock(sk);
1768 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1769 			sched_annotate_sleep();
1770 			lock_sock(sk);
1771 			remove_wait_queue(sk_sleep(sk), &wait);
1772 		}
1773 		err = 0;
1774 		if (!skb_queue_empty(&sk->sk_receive_queue))
1775 			break;
1776 		err = -EAGAIN;
1777 		if (!timeo)
1778 			break;
1779 		err = sock_intr_errno(timeo);
1780 		if (signal_pending(current))
1781 			break;
1782 
1783 		err = sock_error(sk);
1784 		if (err)
1785 			break;
1786 	}
1787 	*timeop = timeo;
1788 	return err;
1789 }
1790 
1791 /**
1792  * tipc_recvmsg - receive packet-oriented message
1793  * @m: descriptor for message info
1794  * @buflen: length of user buffer area
1795  * @flags: receive flags
1796  *
1797  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1798  * If the complete message doesn't fit in user area, truncate it.
1799  *
1800  * Returns size of returned message data, errno otherwise
1801  */
1802 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1803 			size_t buflen,	int flags)
1804 {
1805 	struct sock *sk = sock->sk;
1806 	bool connected = !tipc_sk_type_connectionless(sk);
1807 	struct tipc_sock *tsk = tipc_sk(sk);
1808 	int rc, err, hlen, dlen, copy;
1809 	struct sk_buff_head xmitq;
1810 	struct tipc_msg *hdr;
1811 	struct sk_buff *skb;
1812 	bool grp_evt;
1813 	long timeout;
1814 
1815 	/* Catch invalid receive requests */
1816 	if (unlikely(!buflen))
1817 		return -EINVAL;
1818 
1819 	lock_sock(sk);
1820 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1821 		rc = -ENOTCONN;
1822 		goto exit;
1823 	}
1824 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1825 
1826 	/* Step rcv queue to first msg with data or error; wait if necessary */
1827 	do {
1828 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1829 		if (unlikely(rc))
1830 			goto exit;
1831 		skb = skb_peek(&sk->sk_receive_queue);
1832 		hdr = buf_msg(skb);
1833 		dlen = msg_data_sz(hdr);
1834 		hlen = msg_hdr_sz(hdr);
1835 		err = msg_errcode(hdr);
1836 		grp_evt = msg_is_grp_evt(hdr);
1837 		if (likely(dlen || err))
1838 			break;
1839 		tsk_advance_rx_queue(sk);
1840 	} while (1);
1841 
1842 	/* Collect msg meta data, including error code and rejected data */
1843 	tipc_sk_set_orig_addr(m, skb);
1844 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1845 	if (unlikely(rc))
1846 		goto exit;
1847 	hdr = buf_msg(skb);
1848 
1849 	/* Capture data if non-error msg, otherwise just set return value */
1850 	if (likely(!err)) {
1851 		copy = min_t(int, dlen, buflen);
1852 		if (unlikely(copy != dlen))
1853 			m->msg_flags |= MSG_TRUNC;
1854 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1855 	} else {
1856 		copy = 0;
1857 		rc = 0;
1858 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1859 			rc = -ECONNRESET;
1860 	}
1861 	if (unlikely(rc))
1862 		goto exit;
1863 
1864 	/* Mark message as group event if applicable */
1865 	if (unlikely(grp_evt)) {
1866 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1867 			m->msg_flags |= MSG_EOR;
1868 		m->msg_flags |= MSG_OOB;
1869 		copy = 0;
1870 	}
1871 
1872 	/* Caption of data or error code/rejected data was successful */
1873 	if (unlikely(flags & MSG_PEEK))
1874 		goto exit;
1875 
1876 	/* Send group flow control advertisement when applicable */
1877 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1878 		__skb_queue_head_init(&xmitq);
1879 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1880 					  msg_orignode(hdr), msg_origport(hdr),
1881 					  &xmitq);
1882 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1883 	}
1884 
1885 	tsk_advance_rx_queue(sk);
1886 
1887 	if (likely(!connected))
1888 		goto exit;
1889 
1890 	/* Send connection flow control advertisement when applicable */
1891 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1892 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1893 		tipc_sk_send_ack(tsk);
1894 exit:
1895 	release_sock(sk);
1896 	return rc ? rc : copy;
1897 }
1898 
1899 /**
1900  * tipc_recvstream - receive stream-oriented data
1901  * @m: descriptor for message info
1902  * @buflen: total size of user buffer area
1903  * @flags: receive flags
1904  *
1905  * Used for SOCK_STREAM messages only.  If not enough data is available
1906  * will optionally wait for more; never truncates data.
1907  *
1908  * Returns size of returned message data, errno otherwise
1909  */
1910 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1911 			   size_t buflen, int flags)
1912 {
1913 	struct sock *sk = sock->sk;
1914 	struct tipc_sock *tsk = tipc_sk(sk);
1915 	struct sk_buff *skb;
1916 	struct tipc_msg *hdr;
1917 	struct tipc_skb_cb *skb_cb;
1918 	bool peek = flags & MSG_PEEK;
1919 	int offset, required, copy, copied = 0;
1920 	int hlen, dlen, err, rc;
1921 	bool ack = false;
1922 	long timeout;
1923 
1924 	/* Catch invalid receive attempts */
1925 	if (unlikely(!buflen))
1926 		return -EINVAL;
1927 
1928 	lock_sock(sk);
1929 
1930 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1931 		rc = -ENOTCONN;
1932 		goto exit;
1933 	}
1934 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1935 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1936 
1937 	do {
1938 		/* Look at first msg in receive queue; wait if necessary */
1939 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1940 		if (unlikely(rc))
1941 			break;
1942 		skb = skb_peek(&sk->sk_receive_queue);
1943 		skb_cb = TIPC_SKB_CB(skb);
1944 		hdr = buf_msg(skb);
1945 		dlen = msg_data_sz(hdr);
1946 		hlen = msg_hdr_sz(hdr);
1947 		err = msg_errcode(hdr);
1948 
1949 		/* Discard any empty non-errored (SYN-) message */
1950 		if (unlikely(!dlen && !err)) {
1951 			tsk_advance_rx_queue(sk);
1952 			continue;
1953 		}
1954 
1955 		/* Collect msg meta data, incl. error code and rejected data */
1956 		if (!copied) {
1957 			tipc_sk_set_orig_addr(m, skb);
1958 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
1959 			if (rc)
1960 				break;
1961 			hdr = buf_msg(skb);
1962 		}
1963 
1964 		/* Copy data if msg ok, otherwise return error/partial data */
1965 		if (likely(!err)) {
1966 			ack = msg_ack_required(hdr);
1967 			offset = skb_cb->bytes_read;
1968 			copy = min_t(int, dlen - offset, buflen - copied);
1969 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1970 			if (unlikely(rc))
1971 				break;
1972 			copied += copy;
1973 			offset += copy;
1974 			if (unlikely(offset < dlen)) {
1975 				if (!peek)
1976 					skb_cb->bytes_read = offset;
1977 				break;
1978 			}
1979 		} else {
1980 			rc = 0;
1981 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1982 				rc = -ECONNRESET;
1983 			if (copied || rc)
1984 				break;
1985 		}
1986 
1987 		if (unlikely(peek))
1988 			break;
1989 
1990 		tsk_advance_rx_queue(sk);
1991 
1992 		/* Send connection flow control advertisement when applicable */
1993 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1994 		if (ack || tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1995 			tipc_sk_send_ack(tsk);
1996 
1997 		/* Exit if all requested data or FIN/error received */
1998 		if (copied == buflen || err)
1999 			break;
2000 
2001 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2002 exit:
2003 	release_sock(sk);
2004 	return copied ? copied : rc;
2005 }
2006 
2007 /**
2008  * tipc_write_space - wake up thread if port congestion is released
2009  * @sk: socket
2010  */
2011 static void tipc_write_space(struct sock *sk)
2012 {
2013 	struct socket_wq *wq;
2014 
2015 	rcu_read_lock();
2016 	wq = rcu_dereference(sk->sk_wq);
2017 	if (skwq_has_sleeper(wq))
2018 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2019 						EPOLLWRNORM | EPOLLWRBAND);
2020 	rcu_read_unlock();
2021 }
2022 
2023 /**
2024  * tipc_data_ready - wake up threads to indicate messages have been received
2025  * @sk: socket
2026  * @len: the length of messages
2027  */
2028 static void tipc_data_ready(struct sock *sk)
2029 {
2030 	struct socket_wq *wq;
2031 
2032 	rcu_read_lock();
2033 	wq = rcu_dereference(sk->sk_wq);
2034 	if (skwq_has_sleeper(wq))
2035 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2036 						EPOLLRDNORM | EPOLLRDBAND);
2037 	rcu_read_unlock();
2038 }
2039 
2040 static void tipc_sock_destruct(struct sock *sk)
2041 {
2042 	__skb_queue_purge(&sk->sk_receive_queue);
2043 }
2044 
2045 static void tipc_sk_proto_rcv(struct sock *sk,
2046 			      struct sk_buff_head *inputq,
2047 			      struct sk_buff_head *xmitq)
2048 {
2049 	struct sk_buff *skb = __skb_dequeue(inputq);
2050 	struct tipc_sock *tsk = tipc_sk(sk);
2051 	struct tipc_msg *hdr = buf_msg(skb);
2052 	struct tipc_group *grp = tsk->group;
2053 	bool wakeup = false;
2054 
2055 	switch (msg_user(hdr)) {
2056 	case CONN_MANAGER:
2057 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2058 		return;
2059 	case SOCK_WAKEUP:
2060 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2061 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2062 		smp_wmb();
2063 		tsk->cong_link_cnt--;
2064 		wakeup = true;
2065 		tipc_sk_push_backlog(tsk);
2066 		break;
2067 	case GROUP_PROTOCOL:
2068 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2069 		break;
2070 	case TOP_SRV:
2071 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2072 				      hdr, inputq, xmitq);
2073 		break;
2074 	default:
2075 		break;
2076 	}
2077 
2078 	if (wakeup)
2079 		sk->sk_write_space(sk);
2080 
2081 	kfree_skb(skb);
2082 }
2083 
2084 /**
2085  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2086  * @tsk: TIPC socket
2087  * @skb: pointer to message buffer.
2088  * Returns true if message should be added to receive queue, false otherwise
2089  */
2090 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2091 {
2092 	struct sock *sk = &tsk->sk;
2093 	struct net *net = sock_net(sk);
2094 	struct tipc_msg *hdr = buf_msg(skb);
2095 	bool con_msg = msg_connected(hdr);
2096 	u32 pport = tsk_peer_port(tsk);
2097 	u32 pnode = tsk_peer_node(tsk);
2098 	u32 oport = msg_origport(hdr);
2099 	u32 onode = msg_orignode(hdr);
2100 	int err = msg_errcode(hdr);
2101 	unsigned long delay;
2102 
2103 	if (unlikely(msg_mcast(hdr)))
2104 		return false;
2105 	tsk->oneway = 0;
2106 
2107 	switch (sk->sk_state) {
2108 	case TIPC_CONNECTING:
2109 		/* Setup ACK */
2110 		if (likely(con_msg)) {
2111 			if (err)
2112 				break;
2113 			tipc_sk_finish_conn(tsk, oport, onode);
2114 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2115 			/* ACK+ message with data is added to receive queue */
2116 			if (msg_data_sz(hdr))
2117 				return true;
2118 			/* Empty ACK-, - wake up sleeping connect() and drop */
2119 			sk->sk_state_change(sk);
2120 			msg_set_dest_droppable(hdr, 1);
2121 			return false;
2122 		}
2123 		/* Ignore connectionless message if not from listening socket */
2124 		if (oport != pport || onode != pnode)
2125 			return false;
2126 
2127 		/* Rejected SYN */
2128 		if (err != TIPC_ERR_OVERLOAD)
2129 			break;
2130 
2131 		/* Prepare for new setup attempt if we have a SYN clone */
2132 		if (skb_queue_empty(&sk->sk_write_queue))
2133 			break;
2134 		get_random_bytes(&delay, 2);
2135 		delay %= (tsk->conn_timeout / 4);
2136 		delay = msecs_to_jiffies(delay + 100);
2137 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2138 		return false;
2139 	case TIPC_OPEN:
2140 	case TIPC_DISCONNECTING:
2141 		return false;
2142 	case TIPC_LISTEN:
2143 		/* Accept only SYN message */
2144 		if (!msg_is_syn(hdr) &&
2145 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2146 			return false;
2147 		if (!con_msg && !err)
2148 			return true;
2149 		return false;
2150 	case TIPC_ESTABLISHED:
2151 		if (!skb_queue_empty(&sk->sk_write_queue))
2152 			tipc_sk_push_backlog(tsk);
2153 		/* Accept only connection-based messages sent by peer */
2154 		if (likely(con_msg && !err && pport == oport && pnode == onode))
2155 			return true;
2156 		if (!tsk_peer_msg(tsk, hdr))
2157 			return false;
2158 		if (!err)
2159 			return true;
2160 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2161 		tipc_node_remove_conn(net, pnode, tsk->portid);
2162 		sk->sk_state_change(sk);
2163 		return true;
2164 	default:
2165 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2166 	}
2167 	/* Abort connection setup attempt */
2168 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2169 	sk->sk_err = ECONNREFUSED;
2170 	sk->sk_state_change(sk);
2171 	return true;
2172 }
2173 
2174 /**
2175  * rcvbuf_limit - get proper overload limit of socket receive queue
2176  * @sk: socket
2177  * @skb: message
2178  *
2179  * For connection oriented messages, irrespective of importance,
2180  * default queue limit is 2 MB.
2181  *
2182  * For connectionless messages, queue limits are based on message
2183  * importance as follows:
2184  *
2185  * TIPC_LOW_IMPORTANCE       (2 MB)
2186  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2187  * TIPC_HIGH_IMPORTANCE      (8 MB)
2188  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2189  *
2190  * Returns overload limit according to corresponding message importance
2191  */
2192 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2193 {
2194 	struct tipc_sock *tsk = tipc_sk(sk);
2195 	struct tipc_msg *hdr = buf_msg(skb);
2196 
2197 	if (unlikely(msg_in_group(hdr)))
2198 		return READ_ONCE(sk->sk_rcvbuf);
2199 
2200 	if (unlikely(!msg_connected(hdr)))
2201 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2202 
2203 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2204 		return READ_ONCE(sk->sk_rcvbuf);
2205 
2206 	return FLOWCTL_MSG_LIM;
2207 }
2208 
2209 /**
2210  * tipc_sk_filter_rcv - validate incoming message
2211  * @sk: socket
2212  * @skb: pointer to message.
2213  *
2214  * Enqueues message on receive queue if acceptable; optionally handles
2215  * disconnect indication for a connected socket.
2216  *
2217  * Called with socket lock already taken
2218  *
2219  */
2220 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2221 			       struct sk_buff_head *xmitq)
2222 {
2223 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2224 	struct tipc_sock *tsk = tipc_sk(sk);
2225 	struct tipc_group *grp = tsk->group;
2226 	struct tipc_msg *hdr = buf_msg(skb);
2227 	struct net *net = sock_net(sk);
2228 	struct sk_buff_head inputq;
2229 	int mtyp = msg_type(hdr);
2230 	int limit, err = TIPC_OK;
2231 
2232 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2233 	TIPC_SKB_CB(skb)->bytes_read = 0;
2234 	__skb_queue_head_init(&inputq);
2235 	__skb_queue_tail(&inputq, skb);
2236 
2237 	if (unlikely(!msg_isdata(hdr)))
2238 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2239 
2240 	if (unlikely(grp))
2241 		tipc_group_filter_msg(grp, &inputq, xmitq);
2242 
2243 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2244 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2245 
2246 	/* Validate and add to receive buffer if there is space */
2247 	while ((skb = __skb_dequeue(&inputq))) {
2248 		hdr = buf_msg(skb);
2249 		limit = rcvbuf_limit(sk, skb);
2250 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2251 		    (!sk_conn && msg_connected(hdr)) ||
2252 		    (!grp && msg_in_group(hdr)))
2253 			err = TIPC_ERR_NO_PORT;
2254 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2255 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2256 					   "err_overload2!");
2257 			atomic_inc(&sk->sk_drops);
2258 			err = TIPC_ERR_OVERLOAD;
2259 		}
2260 
2261 		if (unlikely(err)) {
2262 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2263 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2264 						      "@filter_rcv!");
2265 				__skb_queue_tail(xmitq, skb);
2266 			}
2267 			err = TIPC_OK;
2268 			continue;
2269 		}
2270 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2271 		skb_set_owner_r(skb, sk);
2272 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2273 					 "rcvq >90% allocated!");
2274 		sk->sk_data_ready(sk);
2275 	}
2276 }
2277 
2278 /**
2279  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2280  * @sk: socket
2281  * @skb: message
2282  *
2283  * Caller must hold socket lock
2284  */
2285 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2286 {
2287 	unsigned int before = sk_rmem_alloc_get(sk);
2288 	struct sk_buff_head xmitq;
2289 	unsigned int added;
2290 
2291 	__skb_queue_head_init(&xmitq);
2292 
2293 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2294 	added = sk_rmem_alloc_get(sk) - before;
2295 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2296 
2297 	/* Send pending response/rejected messages, if any */
2298 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2299 	return 0;
2300 }
2301 
2302 /**
2303  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2304  *                   inputq and try adding them to socket or backlog queue
2305  * @inputq: list of incoming buffers with potentially different destinations
2306  * @sk: socket where the buffers should be enqueued
2307  * @dport: port number for the socket
2308  *
2309  * Caller must hold socket lock
2310  */
2311 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2312 			    u32 dport, struct sk_buff_head *xmitq)
2313 {
2314 	unsigned long time_limit = jiffies + 2;
2315 	struct sk_buff *skb;
2316 	unsigned int lim;
2317 	atomic_t *dcnt;
2318 	u32 onode;
2319 
2320 	while (skb_queue_len(inputq)) {
2321 		if (unlikely(time_after_eq(jiffies, time_limit)))
2322 			return;
2323 
2324 		skb = tipc_skb_dequeue(inputq, dport);
2325 		if (unlikely(!skb))
2326 			return;
2327 
2328 		/* Add message directly to receive queue if possible */
2329 		if (!sock_owned_by_user(sk)) {
2330 			tipc_sk_filter_rcv(sk, skb, xmitq);
2331 			continue;
2332 		}
2333 
2334 		/* Try backlog, compensating for double-counted bytes */
2335 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2336 		if (!sk->sk_backlog.len)
2337 			atomic_set(dcnt, 0);
2338 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2339 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2340 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2341 						 "bklg & rcvq >90% allocated!");
2342 			continue;
2343 		}
2344 
2345 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2346 		/* Overload => reject message back to sender */
2347 		onode = tipc_own_addr(sock_net(sk));
2348 		atomic_inc(&sk->sk_drops);
2349 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2350 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2351 					      "@sk_enqueue!");
2352 			__skb_queue_tail(xmitq, skb);
2353 		}
2354 		break;
2355 	}
2356 }
2357 
2358 /**
2359  * tipc_sk_rcv - handle a chain of incoming buffers
2360  * @inputq: buffer list containing the buffers
2361  * Consumes all buffers in list until inputq is empty
2362  * Note: may be called in multiple threads referring to the same queue
2363  */
2364 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2365 {
2366 	struct sk_buff_head xmitq;
2367 	u32 dnode, dport = 0;
2368 	int err;
2369 	struct tipc_sock *tsk;
2370 	struct sock *sk;
2371 	struct sk_buff *skb;
2372 
2373 	__skb_queue_head_init(&xmitq);
2374 	while (skb_queue_len(inputq)) {
2375 		dport = tipc_skb_peek_port(inputq, dport);
2376 		tsk = tipc_sk_lookup(net, dport);
2377 
2378 		if (likely(tsk)) {
2379 			sk = &tsk->sk;
2380 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2381 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2382 				spin_unlock_bh(&sk->sk_lock.slock);
2383 			}
2384 			/* Send pending response/rejected messages, if any */
2385 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2386 			sock_put(sk);
2387 			continue;
2388 		}
2389 		/* No destination socket => dequeue skb if still there */
2390 		skb = tipc_skb_dequeue(inputq, dport);
2391 		if (!skb)
2392 			return;
2393 
2394 		/* Try secondary lookup if unresolved named message */
2395 		err = TIPC_ERR_NO_PORT;
2396 		if (tipc_msg_lookup_dest(net, skb, &err))
2397 			goto xmit;
2398 
2399 		/* Prepare for message rejection */
2400 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2401 			continue;
2402 
2403 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2404 xmit:
2405 		dnode = msg_destnode(buf_msg(skb));
2406 		tipc_node_xmit_skb(net, skb, dnode, dport);
2407 	}
2408 }
2409 
2410 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2411 {
2412 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2413 	struct sock *sk = sock->sk;
2414 	int done;
2415 
2416 	do {
2417 		int err = sock_error(sk);
2418 		if (err)
2419 			return err;
2420 		if (!*timeo_p)
2421 			return -ETIMEDOUT;
2422 		if (signal_pending(current))
2423 			return sock_intr_errno(*timeo_p);
2424 
2425 		add_wait_queue(sk_sleep(sk), &wait);
2426 		done = sk_wait_event(sk, timeo_p,
2427 				     sk->sk_state != TIPC_CONNECTING, &wait);
2428 		remove_wait_queue(sk_sleep(sk), &wait);
2429 	} while (!done);
2430 	return 0;
2431 }
2432 
2433 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2434 {
2435 	if (addr->family != AF_TIPC)
2436 		return false;
2437 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2438 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2439 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2440 		addr->addrtype == TIPC_SOCKET_ADDR);
2441 }
2442 
2443 /**
2444  * tipc_connect - establish a connection to another TIPC port
2445  * @sock: socket structure
2446  * @dest: socket address for destination port
2447  * @destlen: size of socket address data structure
2448  * @flags: file-related flags associated with socket
2449  *
2450  * Returns 0 on success, errno otherwise
2451  */
2452 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2453 			int destlen, int flags)
2454 {
2455 	struct sock *sk = sock->sk;
2456 	struct tipc_sock *tsk = tipc_sk(sk);
2457 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2458 	struct msghdr m = {NULL,};
2459 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2460 	int previous;
2461 	int res = 0;
2462 
2463 	if (destlen != sizeof(struct sockaddr_tipc))
2464 		return -EINVAL;
2465 
2466 	lock_sock(sk);
2467 
2468 	if (tsk->group) {
2469 		res = -EINVAL;
2470 		goto exit;
2471 	}
2472 
2473 	if (dst->family == AF_UNSPEC) {
2474 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2475 		if (!tipc_sk_type_connectionless(sk))
2476 			res = -EINVAL;
2477 		goto exit;
2478 	}
2479 	if (!tipc_sockaddr_is_sane(dst)) {
2480 		res = -EINVAL;
2481 		goto exit;
2482 	}
2483 	/* DGRAM/RDM connect(), just save the destaddr */
2484 	if (tipc_sk_type_connectionless(sk)) {
2485 		memcpy(&tsk->peer, dest, destlen);
2486 		goto exit;
2487 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2488 		res = -EINVAL;
2489 		goto exit;
2490 	}
2491 
2492 	previous = sk->sk_state;
2493 
2494 	switch (sk->sk_state) {
2495 	case TIPC_OPEN:
2496 		/* Send a 'SYN-' to destination */
2497 		m.msg_name = dest;
2498 		m.msg_namelen = destlen;
2499 
2500 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2501 		 * indicate send_msg() is never blocked.
2502 		 */
2503 		if (!timeout)
2504 			m.msg_flags = MSG_DONTWAIT;
2505 
2506 		res = __tipc_sendmsg(sock, &m, 0);
2507 		if ((res < 0) && (res != -EWOULDBLOCK))
2508 			goto exit;
2509 
2510 		/* Just entered TIPC_CONNECTING state; the only
2511 		 * difference is that return value in non-blocking
2512 		 * case is EINPROGRESS, rather than EALREADY.
2513 		 */
2514 		res = -EINPROGRESS;
2515 		/* fall through */
2516 	case TIPC_CONNECTING:
2517 		if (!timeout) {
2518 			if (previous == TIPC_CONNECTING)
2519 				res = -EALREADY;
2520 			goto exit;
2521 		}
2522 		timeout = msecs_to_jiffies(timeout);
2523 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2524 		res = tipc_wait_for_connect(sock, &timeout);
2525 		break;
2526 	case TIPC_ESTABLISHED:
2527 		res = -EISCONN;
2528 		break;
2529 	default:
2530 		res = -EINVAL;
2531 	}
2532 
2533 exit:
2534 	release_sock(sk);
2535 	return res;
2536 }
2537 
2538 /**
2539  * tipc_listen - allow socket to listen for incoming connections
2540  * @sock: socket structure
2541  * @len: (unused)
2542  *
2543  * Returns 0 on success, errno otherwise
2544  */
2545 static int tipc_listen(struct socket *sock, int len)
2546 {
2547 	struct sock *sk = sock->sk;
2548 	int res;
2549 
2550 	lock_sock(sk);
2551 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2552 	release_sock(sk);
2553 
2554 	return res;
2555 }
2556 
2557 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2558 {
2559 	struct sock *sk = sock->sk;
2560 	DEFINE_WAIT(wait);
2561 	int err;
2562 
2563 	/* True wake-one mechanism for incoming connections: only
2564 	 * one process gets woken up, not the 'whole herd'.
2565 	 * Since we do not 'race & poll' for established sockets
2566 	 * anymore, the common case will execute the loop only once.
2567 	*/
2568 	for (;;) {
2569 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2570 					  TASK_INTERRUPTIBLE);
2571 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2572 			release_sock(sk);
2573 			timeo = schedule_timeout(timeo);
2574 			lock_sock(sk);
2575 		}
2576 		err = 0;
2577 		if (!skb_queue_empty(&sk->sk_receive_queue))
2578 			break;
2579 		err = -EAGAIN;
2580 		if (!timeo)
2581 			break;
2582 		err = sock_intr_errno(timeo);
2583 		if (signal_pending(current))
2584 			break;
2585 	}
2586 	finish_wait(sk_sleep(sk), &wait);
2587 	return err;
2588 }
2589 
2590 /**
2591  * tipc_accept - wait for connection request
2592  * @sock: listening socket
2593  * @newsock: new socket that is to be connected
2594  * @flags: file-related flags associated with socket
2595  *
2596  * Returns 0 on success, errno otherwise
2597  */
2598 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2599 		       bool kern)
2600 {
2601 	struct sock *new_sk, *sk = sock->sk;
2602 	struct sk_buff *buf;
2603 	struct tipc_sock *new_tsock;
2604 	struct tipc_msg *msg;
2605 	long timeo;
2606 	int res;
2607 
2608 	lock_sock(sk);
2609 
2610 	if (sk->sk_state != TIPC_LISTEN) {
2611 		res = -EINVAL;
2612 		goto exit;
2613 	}
2614 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2615 	res = tipc_wait_for_accept(sock, timeo);
2616 	if (res)
2617 		goto exit;
2618 
2619 	buf = skb_peek(&sk->sk_receive_queue);
2620 
2621 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2622 	if (res)
2623 		goto exit;
2624 	security_sk_clone(sock->sk, new_sock->sk);
2625 
2626 	new_sk = new_sock->sk;
2627 	new_tsock = tipc_sk(new_sk);
2628 	msg = buf_msg(buf);
2629 
2630 	/* we lock on new_sk; but lockdep sees the lock on sk */
2631 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2632 
2633 	/*
2634 	 * Reject any stray messages received by new socket
2635 	 * before the socket lock was taken (very, very unlikely)
2636 	 */
2637 	tsk_rej_rx_queue(new_sk);
2638 
2639 	/* Connect new socket to it's peer */
2640 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2641 
2642 	tsk_set_importance(new_tsock, msg_importance(msg));
2643 	if (msg_named(msg)) {
2644 		new_tsock->conn_type = msg_nametype(msg);
2645 		new_tsock->conn_instance = msg_nameinst(msg);
2646 	}
2647 
2648 	/*
2649 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2650 	 * Respond to 'SYN+' by queuing it on new socket.
2651 	 */
2652 	if (!msg_data_sz(msg)) {
2653 		struct msghdr m = {NULL,};
2654 
2655 		tsk_advance_rx_queue(sk);
2656 		__tipc_sendstream(new_sock, &m, 0);
2657 	} else {
2658 		__skb_dequeue(&sk->sk_receive_queue);
2659 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2660 		skb_set_owner_r(buf, new_sk);
2661 	}
2662 	release_sock(new_sk);
2663 exit:
2664 	release_sock(sk);
2665 	return res;
2666 }
2667 
2668 /**
2669  * tipc_shutdown - shutdown socket connection
2670  * @sock: socket structure
2671  * @how: direction to close (must be SHUT_RDWR)
2672  *
2673  * Terminates connection (if necessary), then purges socket's receive queue.
2674  *
2675  * Returns 0 on success, errno otherwise
2676  */
2677 static int tipc_shutdown(struct socket *sock, int how)
2678 {
2679 	struct sock *sk = sock->sk;
2680 	int res;
2681 
2682 	if (how != SHUT_RDWR)
2683 		return -EINVAL;
2684 
2685 	lock_sock(sk);
2686 
2687 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2688 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2689 	sk->sk_shutdown = SEND_SHUTDOWN;
2690 
2691 	if (sk->sk_state == TIPC_DISCONNECTING) {
2692 		/* Discard any unreceived messages */
2693 		__skb_queue_purge(&sk->sk_receive_queue);
2694 
2695 		/* Wake up anyone sleeping in poll */
2696 		sk->sk_state_change(sk);
2697 		res = 0;
2698 	} else {
2699 		res = -ENOTCONN;
2700 	}
2701 
2702 	release_sock(sk);
2703 	return res;
2704 }
2705 
2706 static void tipc_sk_check_probing_state(struct sock *sk,
2707 					struct sk_buff_head *list)
2708 {
2709 	struct tipc_sock *tsk = tipc_sk(sk);
2710 	u32 pnode = tsk_peer_node(tsk);
2711 	u32 pport = tsk_peer_port(tsk);
2712 	u32 self = tsk_own_node(tsk);
2713 	u32 oport = tsk->portid;
2714 	struct sk_buff *skb;
2715 
2716 	if (tsk->probe_unacked) {
2717 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2718 		sk->sk_err = ECONNABORTED;
2719 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2720 		sk->sk_state_change(sk);
2721 		return;
2722 	}
2723 	/* Prepare new probe */
2724 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2725 			      pnode, self, pport, oport, TIPC_OK);
2726 	if (skb)
2727 		__skb_queue_tail(list, skb);
2728 	tsk->probe_unacked = true;
2729 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2730 }
2731 
2732 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2733 {
2734 	struct tipc_sock *tsk = tipc_sk(sk);
2735 
2736 	/* Try again later if dest link is congested */
2737 	if (tsk->cong_link_cnt) {
2738 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2739 		return;
2740 	}
2741 	/* Prepare SYN for retransmit */
2742 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2743 }
2744 
2745 static void tipc_sk_timeout(struct timer_list *t)
2746 {
2747 	struct sock *sk = from_timer(sk, t, sk_timer);
2748 	struct tipc_sock *tsk = tipc_sk(sk);
2749 	u32 pnode = tsk_peer_node(tsk);
2750 	struct sk_buff_head list;
2751 	int rc = 0;
2752 
2753 	__skb_queue_head_init(&list);
2754 	bh_lock_sock(sk);
2755 
2756 	/* Try again later if socket is busy */
2757 	if (sock_owned_by_user(sk)) {
2758 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2759 		bh_unlock_sock(sk);
2760 		return;
2761 	}
2762 
2763 	if (sk->sk_state == TIPC_ESTABLISHED)
2764 		tipc_sk_check_probing_state(sk, &list);
2765 	else if (sk->sk_state == TIPC_CONNECTING)
2766 		tipc_sk_retry_connect(sk, &list);
2767 
2768 	bh_unlock_sock(sk);
2769 
2770 	if (!skb_queue_empty(&list))
2771 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2772 
2773 	/* SYN messages may cause link congestion */
2774 	if (rc == -ELINKCONG) {
2775 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2776 		tsk->cong_link_cnt = 1;
2777 	}
2778 	sock_put(sk);
2779 }
2780 
2781 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2782 			   struct tipc_name_seq const *seq)
2783 {
2784 	struct sock *sk = &tsk->sk;
2785 	struct net *net = sock_net(sk);
2786 	struct publication *publ;
2787 	u32 key;
2788 
2789 	if (scope != TIPC_NODE_SCOPE)
2790 		scope = TIPC_CLUSTER_SCOPE;
2791 
2792 	if (tipc_sk_connected(sk))
2793 		return -EINVAL;
2794 	key = tsk->portid + tsk->pub_count + 1;
2795 	if (key == tsk->portid)
2796 		return -EADDRINUSE;
2797 
2798 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2799 				    scope, tsk->portid, key);
2800 	if (unlikely(!publ))
2801 		return -EINVAL;
2802 
2803 	list_add(&publ->binding_sock, &tsk->publications);
2804 	tsk->pub_count++;
2805 	tsk->published = 1;
2806 	return 0;
2807 }
2808 
2809 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2810 			    struct tipc_name_seq const *seq)
2811 {
2812 	struct net *net = sock_net(&tsk->sk);
2813 	struct publication *publ;
2814 	struct publication *safe;
2815 	int rc = -EINVAL;
2816 
2817 	if (scope != TIPC_NODE_SCOPE)
2818 		scope = TIPC_CLUSTER_SCOPE;
2819 
2820 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2821 		if (seq) {
2822 			if (publ->scope != scope)
2823 				continue;
2824 			if (publ->type != seq->type)
2825 				continue;
2826 			if (publ->lower != seq->lower)
2827 				continue;
2828 			if (publ->upper != seq->upper)
2829 				break;
2830 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2831 					      publ->upper, publ->key);
2832 			rc = 0;
2833 			break;
2834 		}
2835 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2836 				      publ->upper, publ->key);
2837 		rc = 0;
2838 	}
2839 	if (list_empty(&tsk->publications))
2840 		tsk->published = 0;
2841 	return rc;
2842 }
2843 
2844 /* tipc_sk_reinit: set non-zero address in all existing sockets
2845  *                 when we go from standalone to network mode.
2846  */
2847 void tipc_sk_reinit(struct net *net)
2848 {
2849 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2850 	struct rhashtable_iter iter;
2851 	struct tipc_sock *tsk;
2852 	struct tipc_msg *msg;
2853 
2854 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2855 
2856 	do {
2857 		rhashtable_walk_start(&iter);
2858 
2859 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2860 			sock_hold(&tsk->sk);
2861 			rhashtable_walk_stop(&iter);
2862 			lock_sock(&tsk->sk);
2863 			msg = &tsk->phdr;
2864 			msg_set_prevnode(msg, tipc_own_addr(net));
2865 			msg_set_orignode(msg, tipc_own_addr(net));
2866 			release_sock(&tsk->sk);
2867 			rhashtable_walk_start(&iter);
2868 			sock_put(&tsk->sk);
2869 		}
2870 
2871 		rhashtable_walk_stop(&iter);
2872 	} while (tsk == ERR_PTR(-EAGAIN));
2873 
2874 	rhashtable_walk_exit(&iter);
2875 }
2876 
2877 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2878 {
2879 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2880 	struct tipc_sock *tsk;
2881 
2882 	rcu_read_lock();
2883 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2884 	if (tsk)
2885 		sock_hold(&tsk->sk);
2886 	rcu_read_unlock();
2887 
2888 	return tsk;
2889 }
2890 
2891 static int tipc_sk_insert(struct tipc_sock *tsk)
2892 {
2893 	struct sock *sk = &tsk->sk;
2894 	struct net *net = sock_net(sk);
2895 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2896 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2897 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2898 
2899 	while (remaining--) {
2900 		portid++;
2901 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2902 			portid = TIPC_MIN_PORT;
2903 		tsk->portid = portid;
2904 		sock_hold(&tsk->sk);
2905 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2906 						   tsk_rht_params))
2907 			return 0;
2908 		sock_put(&tsk->sk);
2909 	}
2910 
2911 	return -1;
2912 }
2913 
2914 static void tipc_sk_remove(struct tipc_sock *tsk)
2915 {
2916 	struct sock *sk = &tsk->sk;
2917 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2918 
2919 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2920 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2921 		__sock_put(sk);
2922 	}
2923 }
2924 
2925 static const struct rhashtable_params tsk_rht_params = {
2926 	.nelem_hint = 192,
2927 	.head_offset = offsetof(struct tipc_sock, node),
2928 	.key_offset = offsetof(struct tipc_sock, portid),
2929 	.key_len = sizeof(u32), /* portid */
2930 	.max_size = 1048576,
2931 	.min_size = 256,
2932 	.automatic_shrinking = true,
2933 };
2934 
2935 int tipc_sk_rht_init(struct net *net)
2936 {
2937 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2938 
2939 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2940 }
2941 
2942 void tipc_sk_rht_destroy(struct net *net)
2943 {
2944 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2945 
2946 	/* Wait for socket readers to complete */
2947 	synchronize_net();
2948 
2949 	rhashtable_destroy(&tn->sk_rht);
2950 }
2951 
2952 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2953 {
2954 	struct net *net = sock_net(&tsk->sk);
2955 	struct tipc_group *grp = tsk->group;
2956 	struct tipc_msg *hdr = &tsk->phdr;
2957 	struct tipc_name_seq seq;
2958 	int rc;
2959 
2960 	if (mreq->type < TIPC_RESERVED_TYPES)
2961 		return -EACCES;
2962 	if (mreq->scope > TIPC_NODE_SCOPE)
2963 		return -EINVAL;
2964 	if (grp)
2965 		return -EACCES;
2966 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2967 	if (!grp)
2968 		return -ENOMEM;
2969 	tsk->group = grp;
2970 	msg_set_lookup_scope(hdr, mreq->scope);
2971 	msg_set_nametype(hdr, mreq->type);
2972 	msg_set_dest_droppable(hdr, true);
2973 	seq.type = mreq->type;
2974 	seq.lower = mreq->instance;
2975 	seq.upper = seq.lower;
2976 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2977 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2978 	if (rc) {
2979 		tipc_group_delete(net, grp);
2980 		tsk->group = NULL;
2981 		return rc;
2982 	}
2983 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2984 	tsk->mc_method.rcast = true;
2985 	tsk->mc_method.mandatory = true;
2986 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2987 	return rc;
2988 }
2989 
2990 static int tipc_sk_leave(struct tipc_sock *tsk)
2991 {
2992 	struct net *net = sock_net(&tsk->sk);
2993 	struct tipc_group *grp = tsk->group;
2994 	struct tipc_name_seq seq;
2995 	int scope;
2996 
2997 	if (!grp)
2998 		return -EINVAL;
2999 	tipc_group_self(grp, &seq, &scope);
3000 	tipc_group_delete(net, grp);
3001 	tsk->group = NULL;
3002 	tipc_sk_withdraw(tsk, scope, &seq);
3003 	return 0;
3004 }
3005 
3006 /**
3007  * tipc_setsockopt - set socket option
3008  * @sock: socket structure
3009  * @lvl: option level
3010  * @opt: option identifier
3011  * @ov: pointer to new option value
3012  * @ol: length of option value
3013  *
3014  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3015  * (to ease compatibility).
3016  *
3017  * Returns 0 on success, errno otherwise
3018  */
3019 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3020 			   char __user *ov, unsigned int ol)
3021 {
3022 	struct sock *sk = sock->sk;
3023 	struct tipc_sock *tsk = tipc_sk(sk);
3024 	struct tipc_group_req mreq;
3025 	u32 value = 0;
3026 	int res = 0;
3027 
3028 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3029 		return 0;
3030 	if (lvl != SOL_TIPC)
3031 		return -ENOPROTOOPT;
3032 
3033 	switch (opt) {
3034 	case TIPC_IMPORTANCE:
3035 	case TIPC_SRC_DROPPABLE:
3036 	case TIPC_DEST_DROPPABLE:
3037 	case TIPC_CONN_TIMEOUT:
3038 	case TIPC_NODELAY:
3039 		if (ol < sizeof(value))
3040 			return -EINVAL;
3041 		if (get_user(value, (u32 __user *)ov))
3042 			return -EFAULT;
3043 		break;
3044 	case TIPC_GROUP_JOIN:
3045 		if (ol < sizeof(mreq))
3046 			return -EINVAL;
3047 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
3048 			return -EFAULT;
3049 		break;
3050 	default:
3051 		if (ov || ol)
3052 			return -EINVAL;
3053 	}
3054 
3055 	lock_sock(sk);
3056 
3057 	switch (opt) {
3058 	case TIPC_IMPORTANCE:
3059 		res = tsk_set_importance(tsk, value);
3060 		break;
3061 	case TIPC_SRC_DROPPABLE:
3062 		if (sock->type != SOCK_STREAM)
3063 			tsk_set_unreliable(tsk, value);
3064 		else
3065 			res = -ENOPROTOOPT;
3066 		break;
3067 	case TIPC_DEST_DROPPABLE:
3068 		tsk_set_unreturnable(tsk, value);
3069 		break;
3070 	case TIPC_CONN_TIMEOUT:
3071 		tipc_sk(sk)->conn_timeout = value;
3072 		break;
3073 	case TIPC_MCAST_BROADCAST:
3074 		tsk->mc_method.rcast = false;
3075 		tsk->mc_method.mandatory = true;
3076 		break;
3077 	case TIPC_MCAST_REPLICAST:
3078 		tsk->mc_method.rcast = true;
3079 		tsk->mc_method.mandatory = true;
3080 		break;
3081 	case TIPC_GROUP_JOIN:
3082 		res = tipc_sk_join(tsk, &mreq);
3083 		break;
3084 	case TIPC_GROUP_LEAVE:
3085 		res = tipc_sk_leave(tsk);
3086 		break;
3087 	case TIPC_NODELAY:
3088 		tsk->nodelay = !!value;
3089 		tsk_set_nagle(tsk);
3090 		break;
3091 	default:
3092 		res = -EINVAL;
3093 	}
3094 
3095 	release_sock(sk);
3096 
3097 	return res;
3098 }
3099 
3100 /**
3101  * tipc_getsockopt - get socket option
3102  * @sock: socket structure
3103  * @lvl: option level
3104  * @opt: option identifier
3105  * @ov: receptacle for option value
3106  * @ol: receptacle for length of option value
3107  *
3108  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3109  * (to ease compatibility).
3110  *
3111  * Returns 0 on success, errno otherwise
3112  */
3113 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3114 			   char __user *ov, int __user *ol)
3115 {
3116 	struct sock *sk = sock->sk;
3117 	struct tipc_sock *tsk = tipc_sk(sk);
3118 	struct tipc_name_seq seq;
3119 	int len, scope;
3120 	u32 value;
3121 	int res;
3122 
3123 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3124 		return put_user(0, ol);
3125 	if (lvl != SOL_TIPC)
3126 		return -ENOPROTOOPT;
3127 	res = get_user(len, ol);
3128 	if (res)
3129 		return res;
3130 
3131 	lock_sock(sk);
3132 
3133 	switch (opt) {
3134 	case TIPC_IMPORTANCE:
3135 		value = tsk_importance(tsk);
3136 		break;
3137 	case TIPC_SRC_DROPPABLE:
3138 		value = tsk_unreliable(tsk);
3139 		break;
3140 	case TIPC_DEST_DROPPABLE:
3141 		value = tsk_unreturnable(tsk);
3142 		break;
3143 	case TIPC_CONN_TIMEOUT:
3144 		value = tsk->conn_timeout;
3145 		/* no need to set "res", since already 0 at this point */
3146 		break;
3147 	case TIPC_NODE_RECVQ_DEPTH:
3148 		value = 0; /* was tipc_queue_size, now obsolete */
3149 		break;
3150 	case TIPC_SOCK_RECVQ_DEPTH:
3151 		value = skb_queue_len(&sk->sk_receive_queue);
3152 		break;
3153 	case TIPC_SOCK_RECVQ_USED:
3154 		value = sk_rmem_alloc_get(sk);
3155 		break;
3156 	case TIPC_GROUP_JOIN:
3157 		seq.type = 0;
3158 		if (tsk->group)
3159 			tipc_group_self(tsk->group, &seq, &scope);
3160 		value = seq.type;
3161 		break;
3162 	default:
3163 		res = -EINVAL;
3164 	}
3165 
3166 	release_sock(sk);
3167 
3168 	if (res)
3169 		return res;	/* "get" failed */
3170 
3171 	if (len < sizeof(value))
3172 		return -EINVAL;
3173 
3174 	if (copy_to_user(ov, &value, sizeof(value)))
3175 		return -EFAULT;
3176 
3177 	return put_user(sizeof(value), ol);
3178 }
3179 
3180 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3181 {
3182 	struct net *net = sock_net(sock->sk);
3183 	struct tipc_sioc_nodeid_req nr = {0};
3184 	struct tipc_sioc_ln_req lnr;
3185 	void __user *argp = (void __user *)arg;
3186 
3187 	switch (cmd) {
3188 	case SIOCGETLINKNAME:
3189 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3190 			return -EFAULT;
3191 		if (!tipc_node_get_linkname(net,
3192 					    lnr.bearer_id & 0xffff, lnr.peer,
3193 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3194 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3195 				return -EFAULT;
3196 			return 0;
3197 		}
3198 		return -EADDRNOTAVAIL;
3199 	case SIOCGETNODEID:
3200 		if (copy_from_user(&nr, argp, sizeof(nr)))
3201 			return -EFAULT;
3202 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3203 			return -EADDRNOTAVAIL;
3204 		if (copy_to_user(argp, &nr, sizeof(nr)))
3205 			return -EFAULT;
3206 		return 0;
3207 	default:
3208 		return -ENOIOCTLCMD;
3209 	}
3210 }
3211 
3212 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3213 {
3214 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3215 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3216 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3217 
3218 	tsk1->peer.family = AF_TIPC;
3219 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3220 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3221 	tsk1->peer.addr.id.ref = tsk2->portid;
3222 	tsk1->peer.addr.id.node = onode;
3223 	tsk2->peer.family = AF_TIPC;
3224 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3225 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3226 	tsk2->peer.addr.id.ref = tsk1->portid;
3227 	tsk2->peer.addr.id.node = onode;
3228 
3229 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3230 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3231 	return 0;
3232 }
3233 
3234 /* Protocol switches for the various types of TIPC sockets */
3235 
3236 static const struct proto_ops msg_ops = {
3237 	.owner		= THIS_MODULE,
3238 	.family		= AF_TIPC,
3239 	.release	= tipc_release,
3240 	.bind		= tipc_bind,
3241 	.connect	= tipc_connect,
3242 	.socketpair	= tipc_socketpair,
3243 	.accept		= sock_no_accept,
3244 	.getname	= tipc_getname,
3245 	.poll		= tipc_poll,
3246 	.ioctl		= tipc_ioctl,
3247 	.listen		= sock_no_listen,
3248 	.shutdown	= tipc_shutdown,
3249 	.setsockopt	= tipc_setsockopt,
3250 	.getsockopt	= tipc_getsockopt,
3251 	.sendmsg	= tipc_sendmsg,
3252 	.recvmsg	= tipc_recvmsg,
3253 	.mmap		= sock_no_mmap,
3254 	.sendpage	= sock_no_sendpage
3255 };
3256 
3257 static const struct proto_ops packet_ops = {
3258 	.owner		= THIS_MODULE,
3259 	.family		= AF_TIPC,
3260 	.release	= tipc_release,
3261 	.bind		= tipc_bind,
3262 	.connect	= tipc_connect,
3263 	.socketpair	= tipc_socketpair,
3264 	.accept		= tipc_accept,
3265 	.getname	= tipc_getname,
3266 	.poll		= tipc_poll,
3267 	.ioctl		= tipc_ioctl,
3268 	.listen		= tipc_listen,
3269 	.shutdown	= tipc_shutdown,
3270 	.setsockopt	= tipc_setsockopt,
3271 	.getsockopt	= tipc_getsockopt,
3272 	.sendmsg	= tipc_send_packet,
3273 	.recvmsg	= tipc_recvmsg,
3274 	.mmap		= sock_no_mmap,
3275 	.sendpage	= sock_no_sendpage
3276 };
3277 
3278 static const struct proto_ops stream_ops = {
3279 	.owner		= THIS_MODULE,
3280 	.family		= AF_TIPC,
3281 	.release	= tipc_release,
3282 	.bind		= tipc_bind,
3283 	.connect	= tipc_connect,
3284 	.socketpair	= tipc_socketpair,
3285 	.accept		= tipc_accept,
3286 	.getname	= tipc_getname,
3287 	.poll		= tipc_poll,
3288 	.ioctl		= tipc_ioctl,
3289 	.listen		= tipc_listen,
3290 	.shutdown	= tipc_shutdown,
3291 	.setsockopt	= tipc_setsockopt,
3292 	.getsockopt	= tipc_getsockopt,
3293 	.sendmsg	= tipc_sendstream,
3294 	.recvmsg	= tipc_recvstream,
3295 	.mmap		= sock_no_mmap,
3296 	.sendpage	= sock_no_sendpage
3297 };
3298 
3299 static const struct net_proto_family tipc_family_ops = {
3300 	.owner		= THIS_MODULE,
3301 	.family		= AF_TIPC,
3302 	.create		= tipc_sk_create
3303 };
3304 
3305 static struct proto tipc_proto = {
3306 	.name		= "TIPC",
3307 	.owner		= THIS_MODULE,
3308 	.obj_size	= sizeof(struct tipc_sock),
3309 	.sysctl_rmem	= sysctl_tipc_rmem
3310 };
3311 
3312 /**
3313  * tipc_socket_init - initialize TIPC socket interface
3314  *
3315  * Returns 0 on success, errno otherwise
3316  */
3317 int tipc_socket_init(void)
3318 {
3319 	int res;
3320 
3321 	res = proto_register(&tipc_proto, 1);
3322 	if (res) {
3323 		pr_err("Failed to register TIPC protocol type\n");
3324 		goto out;
3325 	}
3326 
3327 	res = sock_register(&tipc_family_ops);
3328 	if (res) {
3329 		pr_err("Failed to register TIPC socket type\n");
3330 		proto_unregister(&tipc_proto);
3331 		goto out;
3332 	}
3333  out:
3334 	return res;
3335 }
3336 
3337 /**
3338  * tipc_socket_stop - stop TIPC socket interface
3339  */
3340 void tipc_socket_stop(void)
3341 {
3342 	sock_unregister(tipc_family_ops.family);
3343 	proto_unregister(&tipc_proto);
3344 }
3345 
3346 /* Caller should hold socket lock for the passed tipc socket. */
3347 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3348 {
3349 	u32 peer_node;
3350 	u32 peer_port;
3351 	struct nlattr *nest;
3352 
3353 	peer_node = tsk_peer_node(tsk);
3354 	peer_port = tsk_peer_port(tsk);
3355 
3356 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3357 	if (!nest)
3358 		return -EMSGSIZE;
3359 
3360 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3361 		goto msg_full;
3362 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3363 		goto msg_full;
3364 
3365 	if (tsk->conn_type != 0) {
3366 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3367 			goto msg_full;
3368 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3369 			goto msg_full;
3370 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3371 			goto msg_full;
3372 	}
3373 	nla_nest_end(skb, nest);
3374 
3375 	return 0;
3376 
3377 msg_full:
3378 	nla_nest_cancel(skb, nest);
3379 
3380 	return -EMSGSIZE;
3381 }
3382 
3383 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3384 			  *tsk)
3385 {
3386 	struct net *net = sock_net(skb->sk);
3387 	struct sock *sk = &tsk->sk;
3388 
3389 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3390 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3391 		return -EMSGSIZE;
3392 
3393 	if (tipc_sk_connected(sk)) {
3394 		if (__tipc_nl_add_sk_con(skb, tsk))
3395 			return -EMSGSIZE;
3396 	} else if (!list_empty(&tsk->publications)) {
3397 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3398 			return -EMSGSIZE;
3399 	}
3400 	return 0;
3401 }
3402 
3403 /* Caller should hold socket lock for the passed tipc socket. */
3404 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3405 			    struct tipc_sock *tsk)
3406 {
3407 	struct nlattr *attrs;
3408 	void *hdr;
3409 
3410 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3411 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3412 	if (!hdr)
3413 		goto msg_cancel;
3414 
3415 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3416 	if (!attrs)
3417 		goto genlmsg_cancel;
3418 
3419 	if (__tipc_nl_add_sk_info(skb, tsk))
3420 		goto attr_msg_cancel;
3421 
3422 	nla_nest_end(skb, attrs);
3423 	genlmsg_end(skb, hdr);
3424 
3425 	return 0;
3426 
3427 attr_msg_cancel:
3428 	nla_nest_cancel(skb, attrs);
3429 genlmsg_cancel:
3430 	genlmsg_cancel(skb, hdr);
3431 msg_cancel:
3432 	return -EMSGSIZE;
3433 }
3434 
3435 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3436 		    int (*skb_handler)(struct sk_buff *skb,
3437 				       struct netlink_callback *cb,
3438 				       struct tipc_sock *tsk))
3439 {
3440 	struct rhashtable_iter *iter = (void *)cb->args[4];
3441 	struct tipc_sock *tsk;
3442 	int err;
3443 
3444 	rhashtable_walk_start(iter);
3445 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3446 		if (IS_ERR(tsk)) {
3447 			err = PTR_ERR(tsk);
3448 			if (err == -EAGAIN) {
3449 				err = 0;
3450 				continue;
3451 			}
3452 			break;
3453 		}
3454 
3455 		sock_hold(&tsk->sk);
3456 		rhashtable_walk_stop(iter);
3457 		lock_sock(&tsk->sk);
3458 		err = skb_handler(skb, cb, tsk);
3459 		if (err) {
3460 			release_sock(&tsk->sk);
3461 			sock_put(&tsk->sk);
3462 			goto out;
3463 		}
3464 		release_sock(&tsk->sk);
3465 		rhashtable_walk_start(iter);
3466 		sock_put(&tsk->sk);
3467 	}
3468 	rhashtable_walk_stop(iter);
3469 out:
3470 	return skb->len;
3471 }
3472 EXPORT_SYMBOL(tipc_nl_sk_walk);
3473 
3474 int tipc_dump_start(struct netlink_callback *cb)
3475 {
3476 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3477 }
3478 EXPORT_SYMBOL(tipc_dump_start);
3479 
3480 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3481 {
3482 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3483 	struct rhashtable_iter *iter = (void *)cb->args[4];
3484 	struct tipc_net *tn = tipc_net(net);
3485 
3486 	if (!iter) {
3487 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3488 		if (!iter)
3489 			return -ENOMEM;
3490 
3491 		cb->args[4] = (long)iter;
3492 	}
3493 
3494 	rhashtable_walk_enter(&tn->sk_rht, iter);
3495 	return 0;
3496 }
3497 
3498 int tipc_dump_done(struct netlink_callback *cb)
3499 {
3500 	struct rhashtable_iter *hti = (void *)cb->args[4];
3501 
3502 	rhashtable_walk_exit(hti);
3503 	kfree(hti);
3504 	return 0;
3505 }
3506 EXPORT_SYMBOL(tipc_dump_done);
3507 
3508 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3509 			   struct tipc_sock *tsk, u32 sk_filter_state,
3510 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3511 {
3512 	struct sock *sk = &tsk->sk;
3513 	struct nlattr *attrs;
3514 	struct nlattr *stat;
3515 
3516 	/*filter response w.r.t sk_state*/
3517 	if (!(sk_filter_state & (1 << sk->sk_state)))
3518 		return 0;
3519 
3520 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3521 	if (!attrs)
3522 		goto msg_cancel;
3523 
3524 	if (__tipc_nl_add_sk_info(skb, tsk))
3525 		goto attr_msg_cancel;
3526 
3527 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3528 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3529 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3530 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3531 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3532 					 sock_i_uid(sk))) ||
3533 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3534 			      tipc_diag_gen_cookie(sk),
3535 			      TIPC_NLA_SOCK_PAD))
3536 		goto attr_msg_cancel;
3537 
3538 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3539 	if (!stat)
3540 		goto attr_msg_cancel;
3541 
3542 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3543 			skb_queue_len(&sk->sk_receive_queue)) ||
3544 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3545 			skb_queue_len(&sk->sk_write_queue)) ||
3546 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3547 			atomic_read(&sk->sk_drops)))
3548 		goto stat_msg_cancel;
3549 
3550 	if (tsk->cong_link_cnt &&
3551 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3552 		goto stat_msg_cancel;
3553 
3554 	if (tsk_conn_cong(tsk) &&
3555 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3556 		goto stat_msg_cancel;
3557 
3558 	nla_nest_end(skb, stat);
3559 
3560 	if (tsk->group)
3561 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3562 			goto stat_msg_cancel;
3563 
3564 	nla_nest_end(skb, attrs);
3565 
3566 	return 0;
3567 
3568 stat_msg_cancel:
3569 	nla_nest_cancel(skb, stat);
3570 attr_msg_cancel:
3571 	nla_nest_cancel(skb, attrs);
3572 msg_cancel:
3573 	return -EMSGSIZE;
3574 }
3575 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3576 
3577 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3578 {
3579 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3580 }
3581 
3582 /* Caller should hold socket lock for the passed tipc socket. */
3583 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3584 				 struct netlink_callback *cb,
3585 				 struct publication *publ)
3586 {
3587 	void *hdr;
3588 	struct nlattr *attrs;
3589 
3590 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3591 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3592 	if (!hdr)
3593 		goto msg_cancel;
3594 
3595 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3596 	if (!attrs)
3597 		goto genlmsg_cancel;
3598 
3599 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3600 		goto attr_msg_cancel;
3601 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3602 		goto attr_msg_cancel;
3603 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3604 		goto attr_msg_cancel;
3605 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3606 		goto attr_msg_cancel;
3607 
3608 	nla_nest_end(skb, attrs);
3609 	genlmsg_end(skb, hdr);
3610 
3611 	return 0;
3612 
3613 attr_msg_cancel:
3614 	nla_nest_cancel(skb, attrs);
3615 genlmsg_cancel:
3616 	genlmsg_cancel(skb, hdr);
3617 msg_cancel:
3618 	return -EMSGSIZE;
3619 }
3620 
3621 /* Caller should hold socket lock for the passed tipc socket. */
3622 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3623 				  struct netlink_callback *cb,
3624 				  struct tipc_sock *tsk, u32 *last_publ)
3625 {
3626 	int err;
3627 	struct publication *p;
3628 
3629 	if (*last_publ) {
3630 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3631 			if (p->key == *last_publ)
3632 				break;
3633 		}
3634 		if (p->key != *last_publ) {
3635 			/* We never set seq or call nl_dump_check_consistent()
3636 			 * this means that setting prev_seq here will cause the
3637 			 * consistence check to fail in the netlink callback
3638 			 * handler. Resulting in the last NLMSG_DONE message
3639 			 * having the NLM_F_DUMP_INTR flag set.
3640 			 */
3641 			cb->prev_seq = 1;
3642 			*last_publ = 0;
3643 			return -EPIPE;
3644 		}
3645 	} else {
3646 		p = list_first_entry(&tsk->publications, struct publication,
3647 				     binding_sock);
3648 	}
3649 
3650 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3651 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3652 		if (err) {
3653 			*last_publ = p->key;
3654 			return err;
3655 		}
3656 	}
3657 	*last_publ = 0;
3658 
3659 	return 0;
3660 }
3661 
3662 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3663 {
3664 	int err;
3665 	u32 tsk_portid = cb->args[0];
3666 	u32 last_publ = cb->args[1];
3667 	u32 done = cb->args[2];
3668 	struct net *net = sock_net(skb->sk);
3669 	struct tipc_sock *tsk;
3670 
3671 	if (!tsk_portid) {
3672 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3673 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3674 
3675 		if (!attrs[TIPC_NLA_SOCK])
3676 			return -EINVAL;
3677 
3678 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3679 						  attrs[TIPC_NLA_SOCK],
3680 						  tipc_nl_sock_policy, NULL);
3681 		if (err)
3682 			return err;
3683 
3684 		if (!sock[TIPC_NLA_SOCK_REF])
3685 			return -EINVAL;
3686 
3687 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3688 	}
3689 
3690 	if (done)
3691 		return 0;
3692 
3693 	tsk = tipc_sk_lookup(net, tsk_portid);
3694 	if (!tsk)
3695 		return -EINVAL;
3696 
3697 	lock_sock(&tsk->sk);
3698 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3699 	if (!err)
3700 		done = 1;
3701 	release_sock(&tsk->sk);
3702 	sock_put(&tsk->sk);
3703 
3704 	cb->args[0] = tsk_portid;
3705 	cb->args[1] = last_publ;
3706 	cb->args[2] = done;
3707 
3708 	return skb->len;
3709 }
3710 
3711 /**
3712  * tipc_sk_filtering - check if a socket should be traced
3713  * @sk: the socket to be examined
3714  * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3715  *  (portid, sock type, name type, name lower, name upper)
3716  *
3717  * Returns true if the socket meets the socket tuple data
3718  * (value 0 = 'any') or when there is no tuple set (all = 0),
3719  * otherwise false
3720  */
3721 bool tipc_sk_filtering(struct sock *sk)
3722 {
3723 	struct tipc_sock *tsk;
3724 	struct publication *p;
3725 	u32 _port, _sktype, _type, _lower, _upper;
3726 	u32 type = 0, lower = 0, upper = 0;
3727 
3728 	if (!sk)
3729 		return true;
3730 
3731 	tsk = tipc_sk(sk);
3732 
3733 	_port = sysctl_tipc_sk_filter[0];
3734 	_sktype = sysctl_tipc_sk_filter[1];
3735 	_type = sysctl_tipc_sk_filter[2];
3736 	_lower = sysctl_tipc_sk_filter[3];
3737 	_upper = sysctl_tipc_sk_filter[4];
3738 
3739 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3740 		return true;
3741 
3742 	if (_port)
3743 		return (_port == tsk->portid);
3744 
3745 	if (_sktype && _sktype != sk->sk_type)
3746 		return false;
3747 
3748 	if (tsk->published) {
3749 		p = list_first_entry_or_null(&tsk->publications,
3750 					     struct publication, binding_sock);
3751 		if (p) {
3752 			type = p->type;
3753 			lower = p->lower;
3754 			upper = p->upper;
3755 		}
3756 	}
3757 
3758 	if (!tipc_sk_type_connectionless(sk)) {
3759 		type = tsk->conn_type;
3760 		lower = tsk->conn_instance;
3761 		upper = tsk->conn_instance;
3762 	}
3763 
3764 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3765 	    (_upper && _upper != upper))
3766 		return false;
3767 
3768 	return true;
3769 }
3770 
3771 u32 tipc_sock_get_portid(struct sock *sk)
3772 {
3773 	return (sk) ? (tipc_sk(sk))->portid : 0;
3774 }
3775 
3776 /**
3777  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3778  *			both the rcv and backlog queues are considered
3779  * @sk: tipc sk to be checked
3780  * @skb: tipc msg to be checked
3781  *
3782  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3783  */
3784 
3785 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3786 {
3787 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3788 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3789 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3790 
3791 	return (qsize > lim * 90 / 100);
3792 }
3793 
3794 /**
3795  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3796  *			only the rcv queue is considered
3797  * @sk: tipc sk to be checked
3798  * @skb: tipc msg to be checked
3799  *
3800  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3801  */
3802 
3803 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3804 {
3805 	unsigned int lim = rcvbuf_limit(sk, skb);
3806 	unsigned int qsize = sk_rmem_alloc_get(sk);
3807 
3808 	return (qsize > lim * 90 / 100);
3809 }
3810 
3811 /**
3812  * tipc_sk_dump - dump TIPC socket
3813  * @sk: tipc sk to be dumped
3814  * @dqueues: bitmask to decide if any socket queue to be dumped?
3815  *           - TIPC_DUMP_NONE: don't dump socket queues
3816  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3817  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3818  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3819  *           - TIPC_DUMP_ALL: dump all the socket queues above
3820  * @buf: returned buffer of dump data in format
3821  */
3822 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3823 {
3824 	int i = 0;
3825 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3826 	struct tipc_sock *tsk;
3827 	struct publication *p;
3828 	bool tsk_connected;
3829 
3830 	if (!sk) {
3831 		i += scnprintf(buf, sz, "sk data: (null)\n");
3832 		return i;
3833 	}
3834 
3835 	tsk = tipc_sk(sk);
3836 	tsk_connected = !tipc_sk_type_connectionless(sk);
3837 
3838 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3839 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3840 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3841 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3842 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3843 	if (tsk_connected) {
3844 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3845 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3846 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3847 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3848 	}
3849 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3850 	if (tsk->published) {
3851 		p = list_first_entry_or_null(&tsk->publications,
3852 					     struct publication, binding_sock);
3853 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3854 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3855 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3856 	}
3857 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3858 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3859 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3860 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3861 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3862 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3863 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3864 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3865 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3866 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3867 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3868 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3869 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3870 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3871 
3872 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3873 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3874 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3875 	}
3876 
3877 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3878 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3879 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3880 	}
3881 
3882 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3883 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3884 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3885 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3886 			i += scnprintf(buf + i, sz - i, "  tail ");
3887 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3888 					   buf + i);
3889 		}
3890 	}
3891 
3892 	return i;
3893 }
3894