xref: /linux/net/tipc/socket.c (revision cea0f76a483d1270ac6f6513964e3e75193dda48)
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 NAGLE_START_INIT	4
52 #define NAGLE_START_MAX		1024
53 #define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
54 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
55 #define TIPC_FWD_MSG		1
56 #define TIPC_MAX_PORT		0xffffffff
57 #define TIPC_MIN_PORT		1
58 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
59 
60 enum {
61 	TIPC_LISTEN = TCP_LISTEN,
62 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
63 	TIPC_OPEN = TCP_CLOSE,
64 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
65 	TIPC_CONNECTING = TCP_SYN_SENT,
66 };
67 
68 struct sockaddr_pair {
69 	struct sockaddr_tipc sock;
70 	struct sockaddr_tipc member;
71 };
72 
73 /**
74  * struct tipc_sock - TIPC socket structure
75  * @sk: socket - interacts with 'port' and with user via the socket API
76  * @conn_type: TIPC type used when connection was established
77  * @conn_instance: TIPC instance used when connection was established
78  * @published: non-zero if port has one or more associated names
79  * @max_pkt: maximum packet size "hint" used when building messages sent by port
80  * @maxnagle: maximum size of msg which can be subject to nagle
81  * @portid: unique port identity in TIPC socket hash table
82  * @phdr: preformatted message header used when sending messages
83  * #cong_links: list of congested links
84  * @publications: list of publications for port
85  * @blocking_link: address of the congested link we are currently sleeping on
86  * @pub_count: total # of publications port has made during its lifetime
87  * @conn_timeout: the time we can wait for an unresponded setup request
88  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
89  * @cong_link_cnt: number of congested links
90  * @snt_unacked: # messages sent by socket, and not yet acked by peer
91  * @rcv_unacked: # messages read by user, but not yet acked back to peer
92  * @peer: 'connected' peer for dgram/rdm
93  * @node: hash table node
94  * @mc_method: cookie for use between socket and broadcast layer
95  * @rcu: rcu struct for tipc_sock
96  */
97 struct tipc_sock {
98 	struct sock sk;
99 	u32 conn_type;
100 	u32 conn_instance;
101 	int published;
102 	u32 max_pkt;
103 	u32 maxnagle;
104 	u32 portid;
105 	struct tipc_msg phdr;
106 	struct list_head cong_links;
107 	struct list_head publications;
108 	u32 pub_count;
109 	atomic_t dupl_rcvcnt;
110 	u16 conn_timeout;
111 	bool probe_unacked;
112 	u16 cong_link_cnt;
113 	u16 snt_unacked;
114 	u16 snd_win;
115 	u16 peer_caps;
116 	u16 rcv_unacked;
117 	u16 rcv_win;
118 	struct sockaddr_tipc peer;
119 	struct rhash_head node;
120 	struct tipc_mc_method mc_method;
121 	struct rcu_head rcu;
122 	struct tipc_group *group;
123 	u32 oneway;
124 	u32 nagle_start;
125 	u16 snd_backlog;
126 	u16 msg_acc;
127 	u16 pkt_cnt;
128 	bool expect_ack;
129 	bool nodelay;
130 	bool group_is_open;
131 };
132 
133 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
134 static void tipc_data_ready(struct sock *sk);
135 static void tipc_write_space(struct sock *sk);
136 static void tipc_sock_destruct(struct sock *sk);
137 static int tipc_release(struct socket *sock);
138 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
139 		       bool kern);
140 static void tipc_sk_timeout(struct timer_list *t);
141 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
142 			   struct tipc_name_seq const *seq);
143 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
144 			    struct tipc_name_seq const *seq);
145 static int tipc_sk_leave(struct tipc_sock *tsk);
146 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
147 static int tipc_sk_insert(struct tipc_sock *tsk);
148 static void tipc_sk_remove(struct tipc_sock *tsk);
149 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
150 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
151 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
152 
153 static const struct proto_ops packet_ops;
154 static const struct proto_ops stream_ops;
155 static const struct proto_ops msg_ops;
156 static struct proto tipc_proto;
157 static const struct rhashtable_params tsk_rht_params;
158 
159 static u32 tsk_own_node(struct tipc_sock *tsk)
160 {
161 	return msg_prevnode(&tsk->phdr);
162 }
163 
164 static u32 tsk_peer_node(struct tipc_sock *tsk)
165 {
166 	return msg_destnode(&tsk->phdr);
167 }
168 
169 static u32 tsk_peer_port(struct tipc_sock *tsk)
170 {
171 	return msg_destport(&tsk->phdr);
172 }
173 
174 static  bool tsk_unreliable(struct tipc_sock *tsk)
175 {
176 	return msg_src_droppable(&tsk->phdr) != 0;
177 }
178 
179 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
180 {
181 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
182 }
183 
184 static bool tsk_unreturnable(struct tipc_sock *tsk)
185 {
186 	return msg_dest_droppable(&tsk->phdr) != 0;
187 }
188 
189 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
190 {
191 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
192 }
193 
194 static int tsk_importance(struct tipc_sock *tsk)
195 {
196 	return msg_importance(&tsk->phdr);
197 }
198 
199 static struct tipc_sock *tipc_sk(const struct sock *sk)
200 {
201 	return container_of(sk, struct tipc_sock, sk);
202 }
203 
204 int tsk_set_importance(struct sock *sk, int imp)
205 {
206 	if (imp > TIPC_CRITICAL_IMPORTANCE)
207 		return -EINVAL;
208 	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
209 	return 0;
210 }
211 
212 static bool tsk_conn_cong(struct tipc_sock *tsk)
213 {
214 	return tsk->snt_unacked > tsk->snd_win;
215 }
216 
217 static u16 tsk_blocks(int len)
218 {
219 	return ((len / FLOWCTL_BLK_SZ) + 1);
220 }
221 
222 /* tsk_blocks(): translate a buffer size in bytes to number of
223  * advertisable blocks, taking into account the ratio truesize(len)/len
224  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
225  */
226 static u16 tsk_adv_blocks(int len)
227 {
228 	return len / FLOWCTL_BLK_SZ / 4;
229 }
230 
231 /* tsk_inc(): increment counter for sent or received data
232  * - If block based flow control is not supported by peer we
233  *   fall back to message based ditto, incrementing the counter
234  */
235 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
236 {
237 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
238 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
239 	return 1;
240 }
241 
242 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
243  */
244 static void tsk_set_nagle(struct tipc_sock *tsk)
245 {
246 	struct sock *sk = &tsk->sk;
247 
248 	tsk->maxnagle = 0;
249 	if (sk->sk_type != SOCK_STREAM)
250 		return;
251 	if (tsk->nodelay)
252 		return;
253 	if (!(tsk->peer_caps & TIPC_NAGLE))
254 		return;
255 	/* Limit node local buffer size to avoid receive queue overflow */
256 	if (tsk->max_pkt == MAX_MSG_SIZE)
257 		tsk->maxnagle = 1500;
258 	else
259 		tsk->maxnagle = tsk->max_pkt;
260 }
261 
262 /**
263  * tsk_advance_rx_queue - discard first buffer in socket receive queue
264  *
265  * Caller must hold socket lock
266  */
267 static void tsk_advance_rx_queue(struct sock *sk)
268 {
269 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
270 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
271 }
272 
273 /* tipc_sk_respond() : send response message back to sender
274  */
275 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
276 {
277 	u32 selector;
278 	u32 dnode;
279 	u32 onode = tipc_own_addr(sock_net(sk));
280 
281 	if (!tipc_msg_reverse(onode, &skb, err))
282 		return;
283 
284 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
285 	dnode = msg_destnode(buf_msg(skb));
286 	selector = msg_origport(buf_msg(skb));
287 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
288 }
289 
290 /**
291  * tsk_rej_rx_queue - reject all buffers in socket receive queue
292  *
293  * Caller must hold socket lock
294  */
295 static void tsk_rej_rx_queue(struct sock *sk, int error)
296 {
297 	struct sk_buff *skb;
298 
299 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
300 		tipc_sk_respond(sk, skb, error);
301 }
302 
303 static bool tipc_sk_connected(struct sock *sk)
304 {
305 	return sk->sk_state == TIPC_ESTABLISHED;
306 }
307 
308 /* tipc_sk_type_connectionless - check if the socket is datagram socket
309  * @sk: socket
310  *
311  * Returns true if connection less, false otherwise
312  */
313 static bool tipc_sk_type_connectionless(struct sock *sk)
314 {
315 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
316 }
317 
318 /* tsk_peer_msg - verify if message was sent by connected port's peer
319  *
320  * Handles cases where the node's network address has changed from
321  * the default of <0.0.0> to its configured setting.
322  */
323 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
324 {
325 	struct sock *sk = &tsk->sk;
326 	u32 self = tipc_own_addr(sock_net(sk));
327 	u32 peer_port = tsk_peer_port(tsk);
328 	u32 orig_node, peer_node;
329 
330 	if (unlikely(!tipc_sk_connected(sk)))
331 		return false;
332 
333 	if (unlikely(msg_origport(msg) != peer_port))
334 		return false;
335 
336 	orig_node = msg_orignode(msg);
337 	peer_node = tsk_peer_node(tsk);
338 
339 	if (likely(orig_node == peer_node))
340 		return true;
341 
342 	if (!orig_node && peer_node == self)
343 		return true;
344 
345 	if (!peer_node && orig_node == self)
346 		return true;
347 
348 	return false;
349 }
350 
351 /* tipc_set_sk_state - set the sk_state of the socket
352  * @sk: socket
353  *
354  * Caller must hold socket lock
355  *
356  * Returns 0 on success, errno otherwise
357  */
358 static int tipc_set_sk_state(struct sock *sk, int state)
359 {
360 	int oldsk_state = sk->sk_state;
361 	int res = -EINVAL;
362 
363 	switch (state) {
364 	case TIPC_OPEN:
365 		res = 0;
366 		break;
367 	case TIPC_LISTEN:
368 	case TIPC_CONNECTING:
369 		if (oldsk_state == TIPC_OPEN)
370 			res = 0;
371 		break;
372 	case TIPC_ESTABLISHED:
373 		if (oldsk_state == TIPC_CONNECTING ||
374 		    oldsk_state == TIPC_OPEN)
375 			res = 0;
376 		break;
377 	case TIPC_DISCONNECTING:
378 		if (oldsk_state == TIPC_CONNECTING ||
379 		    oldsk_state == TIPC_ESTABLISHED)
380 			res = 0;
381 		break;
382 	}
383 
384 	if (!res)
385 		sk->sk_state = state;
386 
387 	return res;
388 }
389 
390 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
391 {
392 	struct sock *sk = sock->sk;
393 	int err = sock_error(sk);
394 	int typ = sock->type;
395 
396 	if (err)
397 		return err;
398 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
399 		if (sk->sk_state == TIPC_DISCONNECTING)
400 			return -EPIPE;
401 		else if (!tipc_sk_connected(sk))
402 			return -ENOTCONN;
403 	}
404 	if (!*timeout)
405 		return -EAGAIN;
406 	if (signal_pending(current))
407 		return sock_intr_errno(*timeout);
408 
409 	return 0;
410 }
411 
412 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
413 ({                                                                             \
414 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
415 	struct sock *sk_;						       \
416 	int rc_;							       \
417 									       \
418 	while ((rc_ = !(condition_))) {					       \
419 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
420 		smp_rmb();                                                     \
421 		sk_ = (sock_)->sk;					       \
422 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
423 		if (rc_)						       \
424 			break;						       \
425 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
426 		release_sock(sk_);					       \
427 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
428 		sched_annotate_sleep();				               \
429 		lock_sock(sk_);						       \
430 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
431 	}								       \
432 	rc_;								       \
433 })
434 
435 /**
436  * tipc_sk_create - create a TIPC socket
437  * @net: network namespace (must be default network)
438  * @sock: pre-allocated socket structure
439  * @protocol: protocol indicator (must be 0)
440  * @kern: caused by kernel or by userspace?
441  *
442  * This routine creates additional data structures used by the TIPC socket,
443  * initializes them, and links them together.
444  *
445  * Returns 0 on success, errno otherwise
446  */
447 static int tipc_sk_create(struct net *net, struct socket *sock,
448 			  int protocol, int kern)
449 {
450 	const struct proto_ops *ops;
451 	struct sock *sk;
452 	struct tipc_sock *tsk;
453 	struct tipc_msg *msg;
454 
455 	/* Validate arguments */
456 	if (unlikely(protocol != 0))
457 		return -EPROTONOSUPPORT;
458 
459 	switch (sock->type) {
460 	case SOCK_STREAM:
461 		ops = &stream_ops;
462 		break;
463 	case SOCK_SEQPACKET:
464 		ops = &packet_ops;
465 		break;
466 	case SOCK_DGRAM:
467 	case SOCK_RDM:
468 		ops = &msg_ops;
469 		break;
470 	default:
471 		return -EPROTOTYPE;
472 	}
473 
474 	/* Allocate socket's protocol area */
475 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
476 	if (sk == NULL)
477 		return -ENOMEM;
478 
479 	tsk = tipc_sk(sk);
480 	tsk->max_pkt = MAX_PKT_DEFAULT;
481 	tsk->maxnagle = 0;
482 	tsk->nagle_start = NAGLE_START_INIT;
483 	INIT_LIST_HEAD(&tsk->publications);
484 	INIT_LIST_HEAD(&tsk->cong_links);
485 	msg = &tsk->phdr;
486 
487 	/* Finish initializing socket data structures */
488 	sock->ops = ops;
489 	sock_init_data(sock, sk);
490 	tipc_set_sk_state(sk, TIPC_OPEN);
491 	if (tipc_sk_insert(tsk)) {
492 		pr_warn("Socket create failed; port number exhausted\n");
493 		return -EINVAL;
494 	}
495 
496 	/* Ensure tsk is visible before we read own_addr. */
497 	smp_mb();
498 
499 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
500 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
501 
502 	msg_set_origport(msg, tsk->portid);
503 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
504 	sk->sk_shutdown = 0;
505 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
506 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
507 	sk->sk_data_ready = tipc_data_ready;
508 	sk->sk_write_space = tipc_write_space;
509 	sk->sk_destruct = tipc_sock_destruct;
510 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
511 	tsk->group_is_open = true;
512 	atomic_set(&tsk->dupl_rcvcnt, 0);
513 
514 	/* Start out with safe limits until we receive an advertised window */
515 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
516 	tsk->rcv_win = tsk->snd_win;
517 
518 	if (tipc_sk_type_connectionless(sk)) {
519 		tsk_set_unreturnable(tsk, true);
520 		if (sock->type == SOCK_DGRAM)
521 			tsk_set_unreliable(tsk, true);
522 	}
523 	__skb_queue_head_init(&tsk->mc_method.deferredq);
524 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
525 	return 0;
526 }
527 
528 static void tipc_sk_callback(struct rcu_head *head)
529 {
530 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
531 
532 	sock_put(&tsk->sk);
533 }
534 
535 /* Caller should hold socket lock for the socket. */
536 static void __tipc_shutdown(struct socket *sock, int error)
537 {
538 	struct sock *sk = sock->sk;
539 	struct tipc_sock *tsk = tipc_sk(sk);
540 	struct net *net = sock_net(sk);
541 	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
542 	u32 dnode = tsk_peer_node(tsk);
543 	struct sk_buff *skb;
544 
545 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
546 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
547 					    !tsk_conn_cong(tsk)));
548 
549 	/* Push out delayed messages if in Nagle mode */
550 	tipc_sk_push_backlog(tsk, false);
551 	/* Remove pending SYN */
552 	__skb_queue_purge(&sk->sk_write_queue);
553 
554 	/* Remove partially received buffer if any */
555 	skb = skb_peek(&sk->sk_receive_queue);
556 	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
557 		__skb_unlink(skb, &sk->sk_receive_queue);
558 		kfree_skb(skb);
559 	}
560 
561 	/* Reject all unreceived messages if connectionless */
562 	if (tipc_sk_type_connectionless(sk)) {
563 		tsk_rej_rx_queue(sk, error);
564 		return;
565 	}
566 
567 	switch (sk->sk_state) {
568 	case TIPC_CONNECTING:
569 	case TIPC_ESTABLISHED:
570 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
571 		tipc_node_remove_conn(net, dnode, tsk->portid);
572 		/* Send a FIN+/- to its peer */
573 		skb = __skb_dequeue(&sk->sk_receive_queue);
574 		if (skb) {
575 			__skb_queue_purge(&sk->sk_receive_queue);
576 			tipc_sk_respond(sk, skb, error);
577 			break;
578 		}
579 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
580 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
581 				      tsk_own_node(tsk), tsk_peer_port(tsk),
582 				      tsk->portid, error);
583 		if (skb)
584 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
585 		break;
586 	case TIPC_LISTEN:
587 		/* Reject all SYN messages */
588 		tsk_rej_rx_queue(sk, error);
589 		break;
590 	default:
591 		__skb_queue_purge(&sk->sk_receive_queue);
592 		break;
593 	}
594 }
595 
596 /**
597  * tipc_release - destroy a TIPC socket
598  * @sock: socket to destroy
599  *
600  * This routine cleans up any messages that are still queued on the socket.
601  * For DGRAM and RDM socket types, all queued messages are rejected.
602  * For SEQPACKET and STREAM socket types, the first message is rejected
603  * and any others are discarded.  (If the first message on a STREAM socket
604  * is partially-read, it is discarded and the next one is rejected instead.)
605  *
606  * NOTE: Rejected messages are not necessarily returned to the sender!  They
607  * are returned or discarded according to the "destination droppable" setting
608  * specified for the message by the sender.
609  *
610  * Returns 0 on success, errno otherwise
611  */
612 static int tipc_release(struct socket *sock)
613 {
614 	struct sock *sk = sock->sk;
615 	struct tipc_sock *tsk;
616 
617 	/*
618 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
619 	 * releases a pre-allocated child socket that was never used)
620 	 */
621 	if (sk == NULL)
622 		return 0;
623 
624 	tsk = tipc_sk(sk);
625 	lock_sock(sk);
626 
627 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
628 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
629 	sk->sk_shutdown = SHUTDOWN_MASK;
630 	tipc_sk_leave(tsk);
631 	tipc_sk_withdraw(tsk, 0, NULL);
632 	__skb_queue_purge(&tsk->mc_method.deferredq);
633 	sk_stop_timer(sk, &sk->sk_timer);
634 	tipc_sk_remove(tsk);
635 
636 	sock_orphan(sk);
637 	/* Reject any messages that accumulated in backlog queue */
638 	release_sock(sk);
639 	tipc_dest_list_purge(&tsk->cong_links);
640 	tsk->cong_link_cnt = 0;
641 	call_rcu(&tsk->rcu, tipc_sk_callback);
642 	sock->sk = NULL;
643 
644 	return 0;
645 }
646 
647 /**
648  * tipc_bind - associate or disassocate TIPC name(s) with a socket
649  * @sock: socket structure
650  * @uaddr: socket address describing name(s) and desired operation
651  * @uaddr_len: size of socket address data structure
652  *
653  * Name and name sequence binding is indicated using a positive scope value;
654  * a negative scope value unbinds the specified name.  Specifying no name
655  * (i.e. a socket address length of 0) unbinds all names from the socket.
656  *
657  * Returns 0 on success, errno otherwise
658  *
659  * NOTE: This routine doesn't need to take the socket lock since it doesn't
660  *       access any non-constant socket information.
661  */
662 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
663 		     int uaddr_len)
664 {
665 	struct sock *sk = sock->sk;
666 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
667 	struct tipc_sock *tsk = tipc_sk(sk);
668 	int res = -EINVAL;
669 
670 	lock_sock(sk);
671 	if (unlikely(!uaddr_len)) {
672 		res = tipc_sk_withdraw(tsk, 0, NULL);
673 		goto exit;
674 	}
675 	if (tsk->group) {
676 		res = -EACCES;
677 		goto exit;
678 	}
679 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
680 		res = -EINVAL;
681 		goto exit;
682 	}
683 	if (addr->family != AF_TIPC) {
684 		res = -EAFNOSUPPORT;
685 		goto exit;
686 	}
687 
688 	if (addr->addrtype == TIPC_ADDR_NAME)
689 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
690 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
691 		res = -EAFNOSUPPORT;
692 		goto exit;
693 	}
694 
695 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
696 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
697 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
698 		res = -EACCES;
699 		goto exit;
700 	}
701 
702 	res = (addr->scope >= 0) ?
703 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
704 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
705 exit:
706 	release_sock(sk);
707 	return res;
708 }
709 
710 /**
711  * tipc_getname - get port ID of socket or peer socket
712  * @sock: socket structure
713  * @uaddr: area for returned socket address
714  * @uaddr_len: area for returned length of socket address
715  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
716  *
717  * Returns 0 on success, errno otherwise
718  *
719  * NOTE: This routine doesn't need to take the socket lock since it only
720  *       accesses socket information that is unchanging (or which changes in
721  *       a completely predictable manner).
722  */
723 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
724 			int peer)
725 {
726 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
727 	struct sock *sk = sock->sk;
728 	struct tipc_sock *tsk = tipc_sk(sk);
729 
730 	memset(addr, 0, sizeof(*addr));
731 	if (peer) {
732 		if ((!tipc_sk_connected(sk)) &&
733 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
734 			return -ENOTCONN;
735 		addr->addr.id.ref = tsk_peer_port(tsk);
736 		addr->addr.id.node = tsk_peer_node(tsk);
737 	} else {
738 		addr->addr.id.ref = tsk->portid;
739 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
740 	}
741 
742 	addr->addrtype = TIPC_ADDR_ID;
743 	addr->family = AF_TIPC;
744 	addr->scope = 0;
745 	addr->addr.name.domain = 0;
746 
747 	return sizeof(*addr);
748 }
749 
750 /**
751  * tipc_poll - read and possibly block on pollmask
752  * @file: file structure associated with the socket
753  * @sock: socket for which to calculate the poll bits
754  * @wait: ???
755  *
756  * Returns pollmask value
757  *
758  * COMMENTARY:
759  * It appears that the usual socket locking mechanisms are not useful here
760  * since the pollmask info is potentially out-of-date the moment this routine
761  * exits.  TCP and other protocols seem to rely on higher level poll routines
762  * to handle any preventable race conditions, so TIPC will do the same ...
763  *
764  * IMPORTANT: The fact that a read or write operation is indicated does NOT
765  * imply that the operation will succeed, merely that it should be performed
766  * and will not block.
767  */
768 static __poll_t tipc_poll(struct file *file, struct socket *sock,
769 			      poll_table *wait)
770 {
771 	struct sock *sk = sock->sk;
772 	struct tipc_sock *tsk = tipc_sk(sk);
773 	__poll_t revents = 0;
774 
775 	sock_poll_wait(file, sock, wait);
776 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
777 
778 	if (sk->sk_shutdown & RCV_SHUTDOWN)
779 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
780 	if (sk->sk_shutdown == SHUTDOWN_MASK)
781 		revents |= EPOLLHUP;
782 
783 	switch (sk->sk_state) {
784 	case TIPC_ESTABLISHED:
785 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
786 			revents |= EPOLLOUT;
787 		/* fall through */
788 	case TIPC_LISTEN:
789 	case TIPC_CONNECTING:
790 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
791 			revents |= EPOLLIN | EPOLLRDNORM;
792 		break;
793 	case TIPC_OPEN:
794 		if (tsk->group_is_open && !tsk->cong_link_cnt)
795 			revents |= EPOLLOUT;
796 		if (!tipc_sk_type_connectionless(sk))
797 			break;
798 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
799 			break;
800 		revents |= EPOLLIN | EPOLLRDNORM;
801 		break;
802 	case TIPC_DISCONNECTING:
803 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
804 		break;
805 	}
806 	return revents;
807 }
808 
809 /**
810  * tipc_sendmcast - send multicast message
811  * @sock: socket structure
812  * @seq: destination address
813  * @msg: message to send
814  * @dlen: length of data to send
815  * @timeout: timeout to wait for wakeup
816  *
817  * Called from function tipc_sendmsg(), which has done all sanity checks
818  * Returns the number of bytes sent on success, or errno
819  */
820 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
821 			  struct msghdr *msg, size_t dlen, long timeout)
822 {
823 	struct sock *sk = sock->sk;
824 	struct tipc_sock *tsk = tipc_sk(sk);
825 	struct tipc_msg *hdr = &tsk->phdr;
826 	struct net *net = sock_net(sk);
827 	int mtu = tipc_bcast_get_mtu(net);
828 	struct tipc_mc_method *method = &tsk->mc_method;
829 	struct sk_buff_head pkts;
830 	struct tipc_nlist dsts;
831 	int rc;
832 
833 	if (tsk->group)
834 		return -EACCES;
835 
836 	/* Block or return if any destination link is congested */
837 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
838 	if (unlikely(rc))
839 		return rc;
840 
841 	/* Lookup destination nodes */
842 	tipc_nlist_init(&dsts, tipc_own_addr(net));
843 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
844 				      seq->upper, &dsts);
845 	if (!dsts.local && !dsts.remote)
846 		return -EHOSTUNREACH;
847 
848 	/* Build message header */
849 	msg_set_type(hdr, TIPC_MCAST_MSG);
850 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
851 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
852 	msg_set_destport(hdr, 0);
853 	msg_set_destnode(hdr, 0);
854 	msg_set_nametype(hdr, seq->type);
855 	msg_set_namelower(hdr, seq->lower);
856 	msg_set_nameupper(hdr, seq->upper);
857 
858 	/* Build message as chain of buffers */
859 	__skb_queue_head_init(&pkts);
860 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
861 
862 	/* Send message if build was successful */
863 	if (unlikely(rc == dlen)) {
864 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
865 					TIPC_DUMP_SK_SNDQ, " ");
866 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
867 				     &tsk->cong_link_cnt);
868 	}
869 
870 	tipc_nlist_purge(&dsts);
871 
872 	return rc ? rc : dlen;
873 }
874 
875 /**
876  * tipc_send_group_msg - send a message to a member in the group
877  * @net: network namespace
878  * @m: message to send
879  * @mb: group member
880  * @dnode: destination node
881  * @dport: destination port
882  * @dlen: total length of message data
883  */
884 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
885 			       struct msghdr *m, struct tipc_member *mb,
886 			       u32 dnode, u32 dport, int dlen)
887 {
888 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
889 	struct tipc_mc_method *method = &tsk->mc_method;
890 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
891 	struct tipc_msg *hdr = &tsk->phdr;
892 	struct sk_buff_head pkts;
893 	int mtu, rc;
894 
895 	/* Complete message header */
896 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
897 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
898 	msg_set_destport(hdr, dport);
899 	msg_set_destnode(hdr, dnode);
900 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
901 
902 	/* Build message as chain of buffers */
903 	__skb_queue_head_init(&pkts);
904 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
905 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
906 	if (unlikely(rc != dlen))
907 		return rc;
908 
909 	/* Send message */
910 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
911 	if (unlikely(rc == -ELINKCONG)) {
912 		tipc_dest_push(&tsk->cong_links, dnode, 0);
913 		tsk->cong_link_cnt++;
914 	}
915 
916 	/* Update send window */
917 	tipc_group_update_member(mb, blks);
918 
919 	/* A broadcast sent within next EXPIRE period must follow same path */
920 	method->rcast = true;
921 	method->mandatory = true;
922 	return dlen;
923 }
924 
925 /**
926  * tipc_send_group_unicast - send message to a member in the group
927  * @sock: socket structure
928  * @m: message to send
929  * @dlen: total length of message data
930  * @timeout: timeout to wait for wakeup
931  *
932  * Called from function tipc_sendmsg(), which has done all sanity checks
933  * Returns the number of bytes sent on success, or errno
934  */
935 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
936 				   int dlen, long timeout)
937 {
938 	struct sock *sk = sock->sk;
939 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
940 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
941 	struct tipc_sock *tsk = tipc_sk(sk);
942 	struct net *net = sock_net(sk);
943 	struct tipc_member *mb = NULL;
944 	u32 node, port;
945 	int rc;
946 
947 	node = dest->addr.id.node;
948 	port = dest->addr.id.ref;
949 	if (!port && !node)
950 		return -EHOSTUNREACH;
951 
952 	/* Block or return if destination link or member is congested */
953 	rc = tipc_wait_for_cond(sock, &timeout,
954 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
955 				tsk->group &&
956 				!tipc_group_cong(tsk->group, node, port, blks,
957 						 &mb));
958 	if (unlikely(rc))
959 		return rc;
960 
961 	if (unlikely(!mb))
962 		return -EHOSTUNREACH;
963 
964 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
965 
966 	return rc ? rc : dlen;
967 }
968 
969 /**
970  * tipc_send_group_anycast - send message to any member with given identity
971  * @sock: socket structure
972  * @m: message to send
973  * @dlen: total length of message data
974  * @timeout: timeout to wait for wakeup
975  *
976  * Called from function tipc_sendmsg(), which has done all sanity checks
977  * Returns the number of bytes sent on success, or errno
978  */
979 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
980 				   int dlen, long timeout)
981 {
982 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
983 	struct sock *sk = sock->sk;
984 	struct tipc_sock *tsk = tipc_sk(sk);
985 	struct list_head *cong_links = &tsk->cong_links;
986 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
987 	struct tipc_msg *hdr = &tsk->phdr;
988 	struct tipc_member *first = NULL;
989 	struct tipc_member *mbr = NULL;
990 	struct net *net = sock_net(sk);
991 	u32 node, port, exclude;
992 	struct list_head dsts;
993 	u32 type, inst, scope;
994 	int lookups = 0;
995 	int dstcnt, rc;
996 	bool cong;
997 
998 	INIT_LIST_HEAD(&dsts);
999 
1000 	type = msg_nametype(hdr);
1001 	inst = dest->addr.name.name.instance;
1002 	scope = msg_lookup_scope(hdr);
1003 
1004 	while (++lookups < 4) {
1005 		exclude = tipc_group_exclude(tsk->group);
1006 
1007 		first = NULL;
1008 
1009 		/* Look for a non-congested destination member, if any */
1010 		while (1) {
1011 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1012 						 &dstcnt, exclude, false))
1013 				return -EHOSTUNREACH;
1014 			tipc_dest_pop(&dsts, &node, &port);
1015 			cong = tipc_group_cong(tsk->group, node, port, blks,
1016 					       &mbr);
1017 			if (!cong)
1018 				break;
1019 			if (mbr == first)
1020 				break;
1021 			if (!first)
1022 				first = mbr;
1023 		}
1024 
1025 		/* Start over if destination was not in member list */
1026 		if (unlikely(!mbr))
1027 			continue;
1028 
1029 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1030 			break;
1031 
1032 		/* Block or return if destination link or member is congested */
1033 		rc = tipc_wait_for_cond(sock, &timeout,
1034 					!tipc_dest_find(cong_links, node, 0) &&
1035 					tsk->group &&
1036 					!tipc_group_cong(tsk->group, node, port,
1037 							 blks, &mbr));
1038 		if (unlikely(rc))
1039 			return rc;
1040 
1041 		/* Send, unless destination disappeared while waiting */
1042 		if (likely(mbr))
1043 			break;
1044 	}
1045 
1046 	if (unlikely(lookups >= 4))
1047 		return -EHOSTUNREACH;
1048 
1049 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1050 
1051 	return rc ? rc : dlen;
1052 }
1053 
1054 /**
1055  * tipc_send_group_bcast - send message to all members in communication group
1056  * @sk: socket structure
1057  * @m: message to send
1058  * @dlen: total length of message data
1059  * @timeout: timeout to wait for wakeup
1060  *
1061  * Called from function tipc_sendmsg(), which has done all sanity checks
1062  * Returns the number of bytes sent on success, or errno
1063  */
1064 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1065 				 int dlen, long timeout)
1066 {
1067 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1068 	struct sock *sk = sock->sk;
1069 	struct net *net = sock_net(sk);
1070 	struct tipc_sock *tsk = tipc_sk(sk);
1071 	struct tipc_nlist *dsts;
1072 	struct tipc_mc_method *method = &tsk->mc_method;
1073 	bool ack = method->mandatory && method->rcast;
1074 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1075 	struct tipc_msg *hdr = &tsk->phdr;
1076 	int mtu = tipc_bcast_get_mtu(net);
1077 	struct sk_buff_head pkts;
1078 	int rc = -EHOSTUNREACH;
1079 
1080 	/* Block or return if any destination link or member is congested */
1081 	rc = tipc_wait_for_cond(sock, &timeout,
1082 				!tsk->cong_link_cnt && tsk->group &&
1083 				!tipc_group_bc_cong(tsk->group, blks));
1084 	if (unlikely(rc))
1085 		return rc;
1086 
1087 	dsts = tipc_group_dests(tsk->group);
1088 	if (!dsts->local && !dsts->remote)
1089 		return -EHOSTUNREACH;
1090 
1091 	/* Complete message header */
1092 	if (dest) {
1093 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1094 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1095 	} else {
1096 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1097 		msg_set_nameinst(hdr, 0);
1098 	}
1099 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1100 	msg_set_destport(hdr, 0);
1101 	msg_set_destnode(hdr, 0);
1102 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1103 
1104 	/* Avoid getting stuck with repeated forced replicasts */
1105 	msg_set_grp_bc_ack_req(hdr, ack);
1106 
1107 	/* Build message as chain of buffers */
1108 	__skb_queue_head_init(&pkts);
1109 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1110 	if (unlikely(rc != dlen))
1111 		return rc;
1112 
1113 	/* Send message */
1114 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1115 	if (unlikely(rc))
1116 		return rc;
1117 
1118 	/* Update broadcast sequence number and send windows */
1119 	tipc_group_update_bc_members(tsk->group, blks, ack);
1120 
1121 	/* Broadcast link is now free to choose method for next broadcast */
1122 	method->mandatory = false;
1123 	method->expires = jiffies;
1124 
1125 	return dlen;
1126 }
1127 
1128 /**
1129  * tipc_send_group_mcast - send message to all members with given identity
1130  * @sock: socket structure
1131  * @m: message to send
1132  * @dlen: total length of message data
1133  * @timeout: timeout to wait for wakeup
1134  *
1135  * Called from function tipc_sendmsg(), which has done all sanity checks
1136  * Returns the number of bytes sent on success, or errno
1137  */
1138 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1139 				 int dlen, long timeout)
1140 {
1141 	struct sock *sk = sock->sk;
1142 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1143 	struct tipc_sock *tsk = tipc_sk(sk);
1144 	struct tipc_group *grp = tsk->group;
1145 	struct tipc_msg *hdr = &tsk->phdr;
1146 	struct net *net = sock_net(sk);
1147 	u32 type, inst, scope, exclude;
1148 	struct list_head dsts;
1149 	u32 dstcnt;
1150 
1151 	INIT_LIST_HEAD(&dsts);
1152 
1153 	type = msg_nametype(hdr);
1154 	inst = dest->addr.name.name.instance;
1155 	scope = msg_lookup_scope(hdr);
1156 	exclude = tipc_group_exclude(grp);
1157 
1158 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1159 				 &dstcnt, exclude, true))
1160 		return -EHOSTUNREACH;
1161 
1162 	if (dstcnt == 1) {
1163 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1164 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1165 	}
1166 
1167 	tipc_dest_list_purge(&dsts);
1168 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1169 }
1170 
1171 /**
1172  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1173  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1174  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1175  *
1176  * Multi-threaded: parallel calls with reference to same queues may occur
1177  */
1178 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1179 		       struct sk_buff_head *inputq)
1180 {
1181 	u32 self = tipc_own_addr(net);
1182 	u32 type, lower, upper, scope;
1183 	struct sk_buff *skb, *_skb;
1184 	u32 portid, onode;
1185 	struct sk_buff_head tmpq;
1186 	struct list_head dports;
1187 	struct tipc_msg *hdr;
1188 	int user, mtyp, hlen;
1189 	bool exact;
1190 
1191 	__skb_queue_head_init(&tmpq);
1192 	INIT_LIST_HEAD(&dports);
1193 
1194 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1195 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1196 		hdr = buf_msg(skb);
1197 		user = msg_user(hdr);
1198 		mtyp = msg_type(hdr);
1199 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1200 		onode = msg_orignode(hdr);
1201 		type = msg_nametype(hdr);
1202 
1203 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1204 			spin_lock_bh(&inputq->lock);
1205 			if (skb_peek(arrvq) == skb) {
1206 				__skb_dequeue(arrvq);
1207 				__skb_queue_tail(inputq, skb);
1208 			}
1209 			kfree_skb(skb);
1210 			spin_unlock_bh(&inputq->lock);
1211 			continue;
1212 		}
1213 
1214 		/* Group messages require exact scope match */
1215 		if (msg_in_group(hdr)) {
1216 			lower = 0;
1217 			upper = ~0;
1218 			scope = msg_lookup_scope(hdr);
1219 			exact = true;
1220 		} else {
1221 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1222 			if (onode == self)
1223 				scope = TIPC_NODE_SCOPE;
1224 			else
1225 				scope = TIPC_CLUSTER_SCOPE;
1226 			exact = false;
1227 			lower = msg_namelower(hdr);
1228 			upper = msg_nameupper(hdr);
1229 		}
1230 
1231 		/* Create destination port list: */
1232 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1233 				       scope, exact, &dports);
1234 
1235 		/* Clone message per destination */
1236 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1237 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1238 			if (_skb) {
1239 				msg_set_destport(buf_msg(_skb), portid);
1240 				__skb_queue_tail(&tmpq, _skb);
1241 				continue;
1242 			}
1243 			pr_warn("Failed to clone mcast rcv buffer\n");
1244 		}
1245 		/* Append to inputq if not already done by other thread */
1246 		spin_lock_bh(&inputq->lock);
1247 		if (skb_peek(arrvq) == skb) {
1248 			skb_queue_splice_tail_init(&tmpq, inputq);
1249 			kfree_skb(__skb_dequeue(arrvq));
1250 		}
1251 		spin_unlock_bh(&inputq->lock);
1252 		__skb_queue_purge(&tmpq);
1253 		kfree_skb(skb);
1254 	}
1255 	tipc_sk_rcv(net, inputq);
1256 }
1257 
1258 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1259  *                         when socket is in Nagle mode
1260  */
1261 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1262 {
1263 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1264 	struct sk_buff *skb = skb_peek_tail(txq);
1265 	struct net *net = sock_net(&tsk->sk);
1266 	u32 dnode = tsk_peer_node(tsk);
1267 	int rc;
1268 
1269 	if (nagle_ack) {
1270 		tsk->pkt_cnt += skb_queue_len(txq);
1271 		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1272 			tsk->oneway = 0;
1273 			if (tsk->nagle_start < NAGLE_START_MAX)
1274 				tsk->nagle_start *= 2;
1275 			tsk->expect_ack = false;
1276 			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1277 				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1278 				 tsk->nagle_start);
1279 		} else {
1280 			tsk->nagle_start = NAGLE_START_INIT;
1281 			if (skb) {
1282 				msg_set_ack_required(buf_msg(skb));
1283 				tsk->expect_ack = true;
1284 			} else {
1285 				tsk->expect_ack = false;
1286 			}
1287 		}
1288 		tsk->msg_acc = 0;
1289 		tsk->pkt_cnt = 0;
1290 	}
1291 
1292 	if (!skb || tsk->cong_link_cnt)
1293 		return;
1294 
1295 	/* Do not send SYN again after congestion */
1296 	if (msg_is_syn(buf_msg(skb)))
1297 		return;
1298 
1299 	if (tsk->msg_acc)
1300 		tsk->pkt_cnt += skb_queue_len(txq);
1301 	tsk->snt_unacked += tsk->snd_backlog;
1302 	tsk->snd_backlog = 0;
1303 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1304 	if (rc == -ELINKCONG)
1305 		tsk->cong_link_cnt = 1;
1306 }
1307 
1308 /**
1309  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1310  * @tsk: receiving socket
1311  * @skb: pointer to message buffer.
1312  */
1313 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1314 				   struct sk_buff_head *inputq,
1315 				   struct sk_buff_head *xmitq)
1316 {
1317 	struct tipc_msg *hdr = buf_msg(skb);
1318 	u32 onode = tsk_own_node(tsk);
1319 	struct sock *sk = &tsk->sk;
1320 	int mtyp = msg_type(hdr);
1321 	bool was_cong;
1322 
1323 	/* Ignore if connection cannot be validated: */
1324 	if (!tsk_peer_msg(tsk, hdr)) {
1325 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1326 		goto exit;
1327 	}
1328 
1329 	if (unlikely(msg_errcode(hdr))) {
1330 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1331 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1332 				      tsk_peer_port(tsk));
1333 		sk->sk_state_change(sk);
1334 
1335 		/* State change is ignored if socket already awake,
1336 		 * - convert msg to abort msg and add to inqueue
1337 		 */
1338 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1339 		msg_set_type(hdr, TIPC_CONN_MSG);
1340 		msg_set_size(hdr, BASIC_H_SIZE);
1341 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1342 		__skb_queue_tail(inputq, skb);
1343 		return;
1344 	}
1345 
1346 	tsk->probe_unacked = false;
1347 
1348 	if (mtyp == CONN_PROBE) {
1349 		msg_set_type(hdr, CONN_PROBE_REPLY);
1350 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1351 			__skb_queue_tail(xmitq, skb);
1352 		return;
1353 	} else if (mtyp == CONN_ACK) {
1354 		was_cong = tsk_conn_cong(tsk);
1355 		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1356 		tsk->snt_unacked -= msg_conn_ack(hdr);
1357 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1358 			tsk->snd_win = msg_adv_win(hdr);
1359 		if (was_cong && !tsk_conn_cong(tsk))
1360 			sk->sk_write_space(sk);
1361 	} else if (mtyp != CONN_PROBE_REPLY) {
1362 		pr_warn("Received unknown CONN_PROTO msg\n");
1363 	}
1364 exit:
1365 	kfree_skb(skb);
1366 }
1367 
1368 /**
1369  * tipc_sendmsg - send message in connectionless manner
1370  * @sock: socket structure
1371  * @m: message to send
1372  * @dsz: amount of user data to be sent
1373  *
1374  * Message must have an destination specified explicitly.
1375  * Used for SOCK_RDM and SOCK_DGRAM messages,
1376  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1377  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1378  *
1379  * Returns the number of bytes sent on success, or errno otherwise
1380  */
1381 static int tipc_sendmsg(struct socket *sock,
1382 			struct msghdr *m, size_t dsz)
1383 {
1384 	struct sock *sk = sock->sk;
1385 	int ret;
1386 
1387 	lock_sock(sk);
1388 	ret = __tipc_sendmsg(sock, m, dsz);
1389 	release_sock(sk);
1390 
1391 	return ret;
1392 }
1393 
1394 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1395 {
1396 	struct sock *sk = sock->sk;
1397 	struct net *net = sock_net(sk);
1398 	struct tipc_sock *tsk = tipc_sk(sk);
1399 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1400 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1401 	struct list_head *clinks = &tsk->cong_links;
1402 	bool syn = !tipc_sk_type_connectionless(sk);
1403 	struct tipc_group *grp = tsk->group;
1404 	struct tipc_msg *hdr = &tsk->phdr;
1405 	struct tipc_name_seq *seq;
1406 	struct sk_buff_head pkts;
1407 	u32 dport = 0, dnode = 0;
1408 	u32 type = 0, inst = 0;
1409 	int mtu, rc;
1410 
1411 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1412 		return -EMSGSIZE;
1413 
1414 	if (likely(dest)) {
1415 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1416 			return -EINVAL;
1417 		if (unlikely(dest->family != AF_TIPC))
1418 			return -EINVAL;
1419 	}
1420 
1421 	if (grp) {
1422 		if (!dest)
1423 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1424 		if (dest->addrtype == TIPC_ADDR_NAME)
1425 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1426 		if (dest->addrtype == TIPC_ADDR_ID)
1427 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1428 		if (dest->addrtype == TIPC_ADDR_MCAST)
1429 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1430 		return -EINVAL;
1431 	}
1432 
1433 	if (unlikely(!dest)) {
1434 		dest = &tsk->peer;
1435 		if (!syn && dest->family != AF_TIPC)
1436 			return -EDESTADDRREQ;
1437 	}
1438 
1439 	if (unlikely(syn)) {
1440 		if (sk->sk_state == TIPC_LISTEN)
1441 			return -EPIPE;
1442 		if (sk->sk_state != TIPC_OPEN)
1443 			return -EISCONN;
1444 		if (tsk->published)
1445 			return -EOPNOTSUPP;
1446 		if (dest->addrtype == TIPC_ADDR_NAME) {
1447 			tsk->conn_type = dest->addr.name.name.type;
1448 			tsk->conn_instance = dest->addr.name.name.instance;
1449 		}
1450 		msg_set_syn(hdr, 1);
1451 	}
1452 
1453 	seq = &dest->addr.nameseq;
1454 	if (dest->addrtype == TIPC_ADDR_MCAST)
1455 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1456 
1457 	if (dest->addrtype == TIPC_ADDR_NAME) {
1458 		type = dest->addr.name.name.type;
1459 		inst = dest->addr.name.name.instance;
1460 		dnode = dest->addr.name.domain;
1461 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1462 		if (unlikely(!dport && !dnode))
1463 			return -EHOSTUNREACH;
1464 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1465 		dnode = dest->addr.id.node;
1466 	} else {
1467 		return -EINVAL;
1468 	}
1469 
1470 	/* Block or return if destination link is congested */
1471 	rc = tipc_wait_for_cond(sock, &timeout,
1472 				!tipc_dest_find(clinks, dnode, 0));
1473 	if (unlikely(rc))
1474 		return rc;
1475 
1476 	if (dest->addrtype == TIPC_ADDR_NAME) {
1477 		msg_set_type(hdr, TIPC_NAMED_MSG);
1478 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1479 		msg_set_nametype(hdr, type);
1480 		msg_set_nameinst(hdr, inst);
1481 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1482 		msg_set_destnode(hdr, dnode);
1483 		msg_set_destport(hdr, dport);
1484 	} else { /* TIPC_ADDR_ID */
1485 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1486 		msg_set_lookup_scope(hdr, 0);
1487 		msg_set_destnode(hdr, dnode);
1488 		msg_set_destport(hdr, dest->addr.id.ref);
1489 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1490 	}
1491 
1492 	__skb_queue_head_init(&pkts);
1493 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1494 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1495 	if (unlikely(rc != dlen))
1496 		return rc;
1497 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1498 		__skb_queue_purge(&pkts);
1499 		return -ENOMEM;
1500 	}
1501 
1502 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1503 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1504 	if (unlikely(rc == -ELINKCONG)) {
1505 		tipc_dest_push(clinks, dnode, 0);
1506 		tsk->cong_link_cnt++;
1507 		rc = 0;
1508 	}
1509 
1510 	if (unlikely(syn && !rc))
1511 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1512 
1513 	return rc ? rc : dlen;
1514 }
1515 
1516 /**
1517  * tipc_sendstream - send stream-oriented data
1518  * @sock: socket structure
1519  * @m: data to send
1520  * @dsz: total length of data to be transmitted
1521  *
1522  * Used for SOCK_STREAM data.
1523  *
1524  * Returns the number of bytes sent on success (or partial success),
1525  * or errno if no data sent
1526  */
1527 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1528 {
1529 	struct sock *sk = sock->sk;
1530 	int ret;
1531 
1532 	lock_sock(sk);
1533 	ret = __tipc_sendstream(sock, m, dsz);
1534 	release_sock(sk);
1535 
1536 	return ret;
1537 }
1538 
1539 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1540 {
1541 	struct sock *sk = sock->sk;
1542 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1543 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1544 	struct sk_buff_head *txq = &sk->sk_write_queue;
1545 	struct tipc_sock *tsk = tipc_sk(sk);
1546 	struct tipc_msg *hdr = &tsk->phdr;
1547 	struct net *net = sock_net(sk);
1548 	struct sk_buff *skb;
1549 	u32 dnode = tsk_peer_node(tsk);
1550 	int maxnagle = tsk->maxnagle;
1551 	int maxpkt = tsk->max_pkt;
1552 	int send, sent = 0;
1553 	int blocks, rc = 0;
1554 
1555 	if (unlikely(dlen > INT_MAX))
1556 		return -EMSGSIZE;
1557 
1558 	/* Handle implicit connection setup */
1559 	if (unlikely(dest)) {
1560 		rc = __tipc_sendmsg(sock, m, dlen);
1561 		if (dlen && dlen == rc) {
1562 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1563 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1564 		}
1565 		return rc;
1566 	}
1567 
1568 	do {
1569 		rc = tipc_wait_for_cond(sock, &timeout,
1570 					(!tsk->cong_link_cnt &&
1571 					 !tsk_conn_cong(tsk) &&
1572 					 tipc_sk_connected(sk)));
1573 		if (unlikely(rc))
1574 			break;
1575 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1576 		blocks = tsk->snd_backlog;
1577 		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1578 		    send <= maxnagle) {
1579 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1580 			if (unlikely(rc < 0))
1581 				break;
1582 			blocks += rc;
1583 			tsk->msg_acc++;
1584 			if (blocks <= 64 && tsk->expect_ack) {
1585 				tsk->snd_backlog = blocks;
1586 				sent += send;
1587 				break;
1588 			} else if (blocks > 64) {
1589 				tsk->pkt_cnt += skb_queue_len(txq);
1590 			} else {
1591 				skb = skb_peek_tail(txq);
1592 				if (skb) {
1593 					msg_set_ack_required(buf_msg(skb));
1594 					tsk->expect_ack = true;
1595 				} else {
1596 					tsk->expect_ack = false;
1597 				}
1598 				tsk->msg_acc = 0;
1599 				tsk->pkt_cnt = 0;
1600 			}
1601 		} else {
1602 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1603 			if (unlikely(rc != send))
1604 				break;
1605 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1606 		}
1607 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1608 					 TIPC_DUMP_SK_SNDQ, " ");
1609 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1610 		if (unlikely(rc == -ELINKCONG)) {
1611 			tsk->cong_link_cnt = 1;
1612 			rc = 0;
1613 		}
1614 		if (likely(!rc)) {
1615 			tsk->snt_unacked += blocks;
1616 			tsk->snd_backlog = 0;
1617 			sent += send;
1618 		}
1619 	} while (sent < dlen && !rc);
1620 
1621 	return sent ? sent : rc;
1622 }
1623 
1624 /**
1625  * tipc_send_packet - send a connection-oriented message
1626  * @sock: socket structure
1627  * @m: message to send
1628  * @dsz: length of data to be transmitted
1629  *
1630  * Used for SOCK_SEQPACKET messages.
1631  *
1632  * Returns the number of bytes sent on success, or errno otherwise
1633  */
1634 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1635 {
1636 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1637 		return -EMSGSIZE;
1638 
1639 	return tipc_sendstream(sock, m, dsz);
1640 }
1641 
1642 /* tipc_sk_finish_conn - complete the setup of a connection
1643  */
1644 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1645 				u32 peer_node)
1646 {
1647 	struct sock *sk = &tsk->sk;
1648 	struct net *net = sock_net(sk);
1649 	struct tipc_msg *msg = &tsk->phdr;
1650 
1651 	msg_set_syn(msg, 0);
1652 	msg_set_destnode(msg, peer_node);
1653 	msg_set_destport(msg, peer_port);
1654 	msg_set_type(msg, TIPC_CONN_MSG);
1655 	msg_set_lookup_scope(msg, 0);
1656 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1657 
1658 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1659 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1660 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1661 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1662 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1663 	tsk_set_nagle(tsk);
1664 	__skb_queue_purge(&sk->sk_write_queue);
1665 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1666 		return;
1667 
1668 	/* Fall back to message based flow control */
1669 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1670 	tsk->snd_win = FLOWCTL_MSG_WIN;
1671 }
1672 
1673 /**
1674  * tipc_sk_set_orig_addr - capture sender's address for received message
1675  * @m: descriptor for message info
1676  * @hdr: received message header
1677  *
1678  * Note: Address is not captured if not requested by receiver.
1679  */
1680 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1681 {
1682 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1683 	struct tipc_msg *hdr = buf_msg(skb);
1684 
1685 	if (!srcaddr)
1686 		return;
1687 
1688 	srcaddr->sock.family = AF_TIPC;
1689 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1690 	srcaddr->sock.scope = 0;
1691 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1692 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1693 	srcaddr->sock.addr.name.domain = 0;
1694 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1695 
1696 	if (!msg_in_group(hdr))
1697 		return;
1698 
1699 	/* Group message users may also want to know sending member's id */
1700 	srcaddr->member.family = AF_TIPC;
1701 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1702 	srcaddr->member.scope = 0;
1703 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1704 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1705 	srcaddr->member.addr.name.domain = 0;
1706 	m->msg_namelen = sizeof(*srcaddr);
1707 }
1708 
1709 /**
1710  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1711  * @m: descriptor for message info
1712  * @skb: received message buffer
1713  * @tsk: TIPC port associated with message
1714  *
1715  * Note: Ancillary data is not captured if not requested by receiver.
1716  *
1717  * Returns 0 if successful, otherwise errno
1718  */
1719 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1720 				 struct tipc_sock *tsk)
1721 {
1722 	struct tipc_msg *msg;
1723 	u32 anc_data[3];
1724 	u32 err;
1725 	u32 dest_type;
1726 	int has_name;
1727 	int res;
1728 
1729 	if (likely(m->msg_controllen == 0))
1730 		return 0;
1731 	msg = buf_msg(skb);
1732 
1733 	/* Optionally capture errored message object(s) */
1734 	err = msg ? msg_errcode(msg) : 0;
1735 	if (unlikely(err)) {
1736 		anc_data[0] = err;
1737 		anc_data[1] = msg_data_sz(msg);
1738 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1739 		if (res)
1740 			return res;
1741 		if (anc_data[1]) {
1742 			if (skb_linearize(skb))
1743 				return -ENOMEM;
1744 			msg = buf_msg(skb);
1745 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1746 				       msg_data(msg));
1747 			if (res)
1748 				return res;
1749 		}
1750 	}
1751 
1752 	/* Optionally capture message destination object */
1753 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1754 	switch (dest_type) {
1755 	case TIPC_NAMED_MSG:
1756 		has_name = 1;
1757 		anc_data[0] = msg_nametype(msg);
1758 		anc_data[1] = msg_namelower(msg);
1759 		anc_data[2] = msg_namelower(msg);
1760 		break;
1761 	case TIPC_MCAST_MSG:
1762 		has_name = 1;
1763 		anc_data[0] = msg_nametype(msg);
1764 		anc_data[1] = msg_namelower(msg);
1765 		anc_data[2] = msg_nameupper(msg);
1766 		break;
1767 	case TIPC_CONN_MSG:
1768 		has_name = (tsk->conn_type != 0);
1769 		anc_data[0] = tsk->conn_type;
1770 		anc_data[1] = tsk->conn_instance;
1771 		anc_data[2] = tsk->conn_instance;
1772 		break;
1773 	default:
1774 		has_name = 0;
1775 	}
1776 	if (has_name) {
1777 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1778 		if (res)
1779 			return res;
1780 	}
1781 
1782 	return 0;
1783 }
1784 
1785 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1786 {
1787 	struct sock *sk = &tsk->sk;
1788 	struct sk_buff *skb = NULL;
1789 	struct tipc_msg *msg;
1790 	u32 peer_port = tsk_peer_port(tsk);
1791 	u32 dnode = tsk_peer_node(tsk);
1792 
1793 	if (!tipc_sk_connected(sk))
1794 		return NULL;
1795 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1796 			      dnode, tsk_own_node(tsk), peer_port,
1797 			      tsk->portid, TIPC_OK);
1798 	if (!skb)
1799 		return NULL;
1800 	msg = buf_msg(skb);
1801 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1802 	tsk->rcv_unacked = 0;
1803 
1804 	/* Adjust to and advertize the correct window limit */
1805 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1806 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1807 		msg_set_adv_win(msg, tsk->rcv_win);
1808 	}
1809 	return skb;
1810 }
1811 
1812 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1813 {
1814 	struct sk_buff *skb;
1815 
1816 	skb = tipc_sk_build_ack(tsk);
1817 	if (!skb)
1818 		return;
1819 
1820 	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1821 			   msg_link_selector(buf_msg(skb)));
1822 }
1823 
1824 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1825 {
1826 	struct sock *sk = sock->sk;
1827 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1828 	long timeo = *timeop;
1829 	int err = sock_error(sk);
1830 
1831 	if (err)
1832 		return err;
1833 
1834 	for (;;) {
1835 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1836 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1837 				err = -ENOTCONN;
1838 				break;
1839 			}
1840 			add_wait_queue(sk_sleep(sk), &wait);
1841 			release_sock(sk);
1842 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1843 			sched_annotate_sleep();
1844 			lock_sock(sk);
1845 			remove_wait_queue(sk_sleep(sk), &wait);
1846 		}
1847 		err = 0;
1848 		if (!skb_queue_empty(&sk->sk_receive_queue))
1849 			break;
1850 		err = -EAGAIN;
1851 		if (!timeo)
1852 			break;
1853 		err = sock_intr_errno(timeo);
1854 		if (signal_pending(current))
1855 			break;
1856 
1857 		err = sock_error(sk);
1858 		if (err)
1859 			break;
1860 	}
1861 	*timeop = timeo;
1862 	return err;
1863 }
1864 
1865 /**
1866  * tipc_recvmsg - receive packet-oriented message
1867  * @m: descriptor for message info
1868  * @buflen: length of user buffer area
1869  * @flags: receive flags
1870  *
1871  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1872  * If the complete message doesn't fit in user area, truncate it.
1873  *
1874  * Returns size of returned message data, errno otherwise
1875  */
1876 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1877 			size_t buflen,	int flags)
1878 {
1879 	struct sock *sk = sock->sk;
1880 	bool connected = !tipc_sk_type_connectionless(sk);
1881 	struct tipc_sock *tsk = tipc_sk(sk);
1882 	int rc, err, hlen, dlen, copy;
1883 	struct sk_buff_head xmitq;
1884 	struct tipc_msg *hdr;
1885 	struct sk_buff *skb;
1886 	bool grp_evt;
1887 	long timeout;
1888 
1889 	/* Catch invalid receive requests */
1890 	if (unlikely(!buflen))
1891 		return -EINVAL;
1892 
1893 	lock_sock(sk);
1894 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1895 		rc = -ENOTCONN;
1896 		goto exit;
1897 	}
1898 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1899 
1900 	/* Step rcv queue to first msg with data or error; wait if necessary */
1901 	do {
1902 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1903 		if (unlikely(rc))
1904 			goto exit;
1905 		skb = skb_peek(&sk->sk_receive_queue);
1906 		hdr = buf_msg(skb);
1907 		dlen = msg_data_sz(hdr);
1908 		hlen = msg_hdr_sz(hdr);
1909 		err = msg_errcode(hdr);
1910 		grp_evt = msg_is_grp_evt(hdr);
1911 		if (likely(dlen || err))
1912 			break;
1913 		tsk_advance_rx_queue(sk);
1914 	} while (1);
1915 
1916 	/* Collect msg meta data, including error code and rejected data */
1917 	tipc_sk_set_orig_addr(m, skb);
1918 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1919 	if (unlikely(rc))
1920 		goto exit;
1921 	hdr = buf_msg(skb);
1922 
1923 	/* Capture data if non-error msg, otherwise just set return value */
1924 	if (likely(!err)) {
1925 		copy = min_t(int, dlen, buflen);
1926 		if (unlikely(copy != dlen))
1927 			m->msg_flags |= MSG_TRUNC;
1928 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1929 	} else {
1930 		copy = 0;
1931 		rc = 0;
1932 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1933 			rc = -ECONNRESET;
1934 	}
1935 	if (unlikely(rc))
1936 		goto exit;
1937 
1938 	/* Mark message as group event if applicable */
1939 	if (unlikely(grp_evt)) {
1940 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1941 			m->msg_flags |= MSG_EOR;
1942 		m->msg_flags |= MSG_OOB;
1943 		copy = 0;
1944 	}
1945 
1946 	/* Caption of data or error code/rejected data was successful */
1947 	if (unlikely(flags & MSG_PEEK))
1948 		goto exit;
1949 
1950 	/* Send group flow control advertisement when applicable */
1951 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1952 		__skb_queue_head_init(&xmitq);
1953 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1954 					  msg_orignode(hdr), msg_origport(hdr),
1955 					  &xmitq);
1956 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1957 	}
1958 
1959 	tsk_advance_rx_queue(sk);
1960 
1961 	if (likely(!connected))
1962 		goto exit;
1963 
1964 	/* Send connection flow control advertisement when applicable */
1965 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1966 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1967 		tipc_sk_send_ack(tsk);
1968 exit:
1969 	release_sock(sk);
1970 	return rc ? rc : copy;
1971 }
1972 
1973 /**
1974  * tipc_recvstream - receive stream-oriented data
1975  * @m: descriptor for message info
1976  * @buflen: total size of user buffer area
1977  * @flags: receive flags
1978  *
1979  * Used for SOCK_STREAM messages only.  If not enough data is available
1980  * will optionally wait for more; never truncates data.
1981  *
1982  * Returns size of returned message data, errno otherwise
1983  */
1984 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1985 			   size_t buflen, int flags)
1986 {
1987 	struct sock *sk = sock->sk;
1988 	struct tipc_sock *tsk = tipc_sk(sk);
1989 	struct sk_buff *skb;
1990 	struct tipc_msg *hdr;
1991 	struct tipc_skb_cb *skb_cb;
1992 	bool peek = flags & MSG_PEEK;
1993 	int offset, required, copy, copied = 0;
1994 	int hlen, dlen, err, rc;
1995 	long timeout;
1996 
1997 	/* Catch invalid receive attempts */
1998 	if (unlikely(!buflen))
1999 		return -EINVAL;
2000 
2001 	lock_sock(sk);
2002 
2003 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2004 		rc = -ENOTCONN;
2005 		goto exit;
2006 	}
2007 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2008 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2009 
2010 	do {
2011 		/* Look at first msg in receive queue; wait if necessary */
2012 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2013 		if (unlikely(rc))
2014 			break;
2015 		skb = skb_peek(&sk->sk_receive_queue);
2016 		skb_cb = TIPC_SKB_CB(skb);
2017 		hdr = buf_msg(skb);
2018 		dlen = msg_data_sz(hdr);
2019 		hlen = msg_hdr_sz(hdr);
2020 		err = msg_errcode(hdr);
2021 
2022 		/* Discard any empty non-errored (SYN-) message */
2023 		if (unlikely(!dlen && !err)) {
2024 			tsk_advance_rx_queue(sk);
2025 			continue;
2026 		}
2027 
2028 		/* Collect msg meta data, incl. error code and rejected data */
2029 		if (!copied) {
2030 			tipc_sk_set_orig_addr(m, skb);
2031 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2032 			if (rc)
2033 				break;
2034 			hdr = buf_msg(skb);
2035 		}
2036 
2037 		/* Copy data if msg ok, otherwise return error/partial data */
2038 		if (likely(!err)) {
2039 			offset = skb_cb->bytes_read;
2040 			copy = min_t(int, dlen - offset, buflen - copied);
2041 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2042 			if (unlikely(rc))
2043 				break;
2044 			copied += copy;
2045 			offset += copy;
2046 			if (unlikely(offset < dlen)) {
2047 				if (!peek)
2048 					skb_cb->bytes_read = offset;
2049 				break;
2050 			}
2051 		} else {
2052 			rc = 0;
2053 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2054 				rc = -ECONNRESET;
2055 			if (copied || rc)
2056 				break;
2057 		}
2058 
2059 		if (unlikely(peek))
2060 			break;
2061 
2062 		tsk_advance_rx_queue(sk);
2063 
2064 		/* Send connection flow control advertisement when applicable */
2065 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2066 		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2067 			tipc_sk_send_ack(tsk);
2068 
2069 		/* Exit if all requested data or FIN/error received */
2070 		if (copied == buflen || err)
2071 			break;
2072 
2073 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2074 exit:
2075 	release_sock(sk);
2076 	return copied ? copied : rc;
2077 }
2078 
2079 /**
2080  * tipc_write_space - wake up thread if port congestion is released
2081  * @sk: socket
2082  */
2083 static void tipc_write_space(struct sock *sk)
2084 {
2085 	struct socket_wq *wq;
2086 
2087 	rcu_read_lock();
2088 	wq = rcu_dereference(sk->sk_wq);
2089 	if (skwq_has_sleeper(wq))
2090 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2091 						EPOLLWRNORM | EPOLLWRBAND);
2092 	rcu_read_unlock();
2093 }
2094 
2095 /**
2096  * tipc_data_ready - wake up threads to indicate messages have been received
2097  * @sk: socket
2098  * @len: the length of messages
2099  */
2100 static void tipc_data_ready(struct sock *sk)
2101 {
2102 	struct socket_wq *wq;
2103 
2104 	rcu_read_lock();
2105 	wq = rcu_dereference(sk->sk_wq);
2106 	if (skwq_has_sleeper(wq))
2107 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2108 						EPOLLRDNORM | EPOLLRDBAND);
2109 	rcu_read_unlock();
2110 }
2111 
2112 static void tipc_sock_destruct(struct sock *sk)
2113 {
2114 	__skb_queue_purge(&sk->sk_receive_queue);
2115 }
2116 
2117 static void tipc_sk_proto_rcv(struct sock *sk,
2118 			      struct sk_buff_head *inputq,
2119 			      struct sk_buff_head *xmitq)
2120 {
2121 	struct sk_buff *skb = __skb_dequeue(inputq);
2122 	struct tipc_sock *tsk = tipc_sk(sk);
2123 	struct tipc_msg *hdr = buf_msg(skb);
2124 	struct tipc_group *grp = tsk->group;
2125 	bool wakeup = false;
2126 
2127 	switch (msg_user(hdr)) {
2128 	case CONN_MANAGER:
2129 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2130 		return;
2131 	case SOCK_WAKEUP:
2132 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2133 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2134 		smp_wmb();
2135 		tsk->cong_link_cnt--;
2136 		wakeup = true;
2137 		tipc_sk_push_backlog(tsk, false);
2138 		break;
2139 	case GROUP_PROTOCOL:
2140 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2141 		break;
2142 	case TOP_SRV:
2143 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2144 				      hdr, inputq, xmitq);
2145 		break;
2146 	default:
2147 		break;
2148 	}
2149 
2150 	if (wakeup)
2151 		sk->sk_write_space(sk);
2152 
2153 	kfree_skb(skb);
2154 }
2155 
2156 /**
2157  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2158  * @tsk: TIPC socket
2159  * @skb: pointer to message buffer.
2160  * @xmitq: for Nagle ACK if any
2161  * Returns true if message should be added to receive queue, false otherwise
2162  */
2163 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2164 				   struct sk_buff_head *xmitq)
2165 {
2166 	struct sock *sk = &tsk->sk;
2167 	struct net *net = sock_net(sk);
2168 	struct tipc_msg *hdr = buf_msg(skb);
2169 	bool con_msg = msg_connected(hdr);
2170 	u32 pport = tsk_peer_port(tsk);
2171 	u32 pnode = tsk_peer_node(tsk);
2172 	u32 oport = msg_origport(hdr);
2173 	u32 onode = msg_orignode(hdr);
2174 	int err = msg_errcode(hdr);
2175 	unsigned long delay;
2176 
2177 	if (unlikely(msg_mcast(hdr)))
2178 		return false;
2179 	tsk->oneway = 0;
2180 
2181 	switch (sk->sk_state) {
2182 	case TIPC_CONNECTING:
2183 		/* Setup ACK */
2184 		if (likely(con_msg)) {
2185 			if (err)
2186 				break;
2187 			tipc_sk_finish_conn(tsk, oport, onode);
2188 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2189 			/* ACK+ message with data is added to receive queue */
2190 			if (msg_data_sz(hdr))
2191 				return true;
2192 			/* Empty ACK-, - wake up sleeping connect() and drop */
2193 			sk->sk_state_change(sk);
2194 			msg_set_dest_droppable(hdr, 1);
2195 			return false;
2196 		}
2197 		/* Ignore connectionless message if not from listening socket */
2198 		if (oport != pport || onode != pnode)
2199 			return false;
2200 
2201 		/* Rejected SYN */
2202 		if (err != TIPC_ERR_OVERLOAD)
2203 			break;
2204 
2205 		/* Prepare for new setup attempt if we have a SYN clone */
2206 		if (skb_queue_empty(&sk->sk_write_queue))
2207 			break;
2208 		get_random_bytes(&delay, 2);
2209 		delay %= (tsk->conn_timeout / 4);
2210 		delay = msecs_to_jiffies(delay + 100);
2211 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2212 		return false;
2213 	case TIPC_OPEN:
2214 	case TIPC_DISCONNECTING:
2215 		return false;
2216 	case TIPC_LISTEN:
2217 		/* Accept only SYN message */
2218 		if (!msg_is_syn(hdr) &&
2219 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2220 			return false;
2221 		if (!con_msg && !err)
2222 			return true;
2223 		return false;
2224 	case TIPC_ESTABLISHED:
2225 		if (!skb_queue_empty(&sk->sk_write_queue))
2226 			tipc_sk_push_backlog(tsk, false);
2227 		/* Accept only connection-based messages sent by peer */
2228 		if (likely(con_msg && !err && pport == oport &&
2229 			   pnode == onode)) {
2230 			if (msg_ack_required(hdr)) {
2231 				struct sk_buff *skb;
2232 
2233 				skb = tipc_sk_build_ack(tsk);
2234 				if (skb) {
2235 					msg_set_nagle_ack(buf_msg(skb));
2236 					__skb_queue_tail(xmitq, skb);
2237 				}
2238 			}
2239 			return true;
2240 		}
2241 		if (!tsk_peer_msg(tsk, hdr))
2242 			return false;
2243 		if (!err)
2244 			return true;
2245 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2246 		tipc_node_remove_conn(net, pnode, tsk->portid);
2247 		sk->sk_state_change(sk);
2248 		return true;
2249 	default:
2250 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2251 	}
2252 	/* Abort connection setup attempt */
2253 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2254 	sk->sk_err = ECONNREFUSED;
2255 	sk->sk_state_change(sk);
2256 	return true;
2257 }
2258 
2259 /**
2260  * rcvbuf_limit - get proper overload limit of socket receive queue
2261  * @sk: socket
2262  * @skb: message
2263  *
2264  * For connection oriented messages, irrespective of importance,
2265  * default queue limit is 2 MB.
2266  *
2267  * For connectionless messages, queue limits are based on message
2268  * importance as follows:
2269  *
2270  * TIPC_LOW_IMPORTANCE       (2 MB)
2271  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2272  * TIPC_HIGH_IMPORTANCE      (8 MB)
2273  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2274  *
2275  * Returns overload limit according to corresponding message importance
2276  */
2277 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2278 {
2279 	struct tipc_sock *tsk = tipc_sk(sk);
2280 	struct tipc_msg *hdr = buf_msg(skb);
2281 
2282 	if (unlikely(msg_in_group(hdr)))
2283 		return READ_ONCE(sk->sk_rcvbuf);
2284 
2285 	if (unlikely(!msg_connected(hdr)))
2286 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2287 
2288 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2289 		return READ_ONCE(sk->sk_rcvbuf);
2290 
2291 	return FLOWCTL_MSG_LIM;
2292 }
2293 
2294 /**
2295  * tipc_sk_filter_rcv - validate incoming message
2296  * @sk: socket
2297  * @skb: pointer to message.
2298  *
2299  * Enqueues message on receive queue if acceptable; optionally handles
2300  * disconnect indication for a connected socket.
2301  *
2302  * Called with socket lock already taken
2303  *
2304  */
2305 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2306 			       struct sk_buff_head *xmitq)
2307 {
2308 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2309 	struct tipc_sock *tsk = tipc_sk(sk);
2310 	struct tipc_group *grp = tsk->group;
2311 	struct tipc_msg *hdr = buf_msg(skb);
2312 	struct net *net = sock_net(sk);
2313 	struct sk_buff_head inputq;
2314 	int mtyp = msg_type(hdr);
2315 	int limit, err = TIPC_OK;
2316 
2317 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2318 	TIPC_SKB_CB(skb)->bytes_read = 0;
2319 	__skb_queue_head_init(&inputq);
2320 	__skb_queue_tail(&inputq, skb);
2321 
2322 	if (unlikely(!msg_isdata(hdr)))
2323 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2324 
2325 	if (unlikely(grp))
2326 		tipc_group_filter_msg(grp, &inputq, xmitq);
2327 
2328 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2329 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2330 
2331 	/* Validate and add to receive buffer if there is space */
2332 	while ((skb = __skb_dequeue(&inputq))) {
2333 		hdr = buf_msg(skb);
2334 		limit = rcvbuf_limit(sk, skb);
2335 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2336 		    (!sk_conn && msg_connected(hdr)) ||
2337 		    (!grp && msg_in_group(hdr)))
2338 			err = TIPC_ERR_NO_PORT;
2339 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2340 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2341 					   "err_overload2!");
2342 			atomic_inc(&sk->sk_drops);
2343 			err = TIPC_ERR_OVERLOAD;
2344 		}
2345 
2346 		if (unlikely(err)) {
2347 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2348 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2349 						      "@filter_rcv!");
2350 				__skb_queue_tail(xmitq, skb);
2351 			}
2352 			err = TIPC_OK;
2353 			continue;
2354 		}
2355 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2356 		skb_set_owner_r(skb, sk);
2357 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2358 					 "rcvq >90% allocated!");
2359 		sk->sk_data_ready(sk);
2360 	}
2361 }
2362 
2363 /**
2364  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2365  * @sk: socket
2366  * @skb: message
2367  *
2368  * Caller must hold socket lock
2369  */
2370 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2371 {
2372 	unsigned int before = sk_rmem_alloc_get(sk);
2373 	struct sk_buff_head xmitq;
2374 	unsigned int added;
2375 
2376 	__skb_queue_head_init(&xmitq);
2377 
2378 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2379 	added = sk_rmem_alloc_get(sk) - before;
2380 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2381 
2382 	/* Send pending response/rejected messages, if any */
2383 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2384 	return 0;
2385 }
2386 
2387 /**
2388  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2389  *                   inputq and try adding them to socket or backlog queue
2390  * @inputq: list of incoming buffers with potentially different destinations
2391  * @sk: socket where the buffers should be enqueued
2392  * @dport: port number for the socket
2393  *
2394  * Caller must hold socket lock
2395  */
2396 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2397 			    u32 dport, struct sk_buff_head *xmitq)
2398 {
2399 	unsigned long time_limit = jiffies + 2;
2400 	struct sk_buff *skb;
2401 	unsigned int lim;
2402 	atomic_t *dcnt;
2403 	u32 onode;
2404 
2405 	while (skb_queue_len(inputq)) {
2406 		if (unlikely(time_after_eq(jiffies, time_limit)))
2407 			return;
2408 
2409 		skb = tipc_skb_dequeue(inputq, dport);
2410 		if (unlikely(!skb))
2411 			return;
2412 
2413 		/* Add message directly to receive queue if possible */
2414 		if (!sock_owned_by_user(sk)) {
2415 			tipc_sk_filter_rcv(sk, skb, xmitq);
2416 			continue;
2417 		}
2418 
2419 		/* Try backlog, compensating for double-counted bytes */
2420 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2421 		if (!sk->sk_backlog.len)
2422 			atomic_set(dcnt, 0);
2423 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2424 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2425 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2426 						 "bklg & rcvq >90% allocated!");
2427 			continue;
2428 		}
2429 
2430 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2431 		/* Overload => reject message back to sender */
2432 		onode = tipc_own_addr(sock_net(sk));
2433 		atomic_inc(&sk->sk_drops);
2434 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2435 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2436 					      "@sk_enqueue!");
2437 			__skb_queue_tail(xmitq, skb);
2438 		}
2439 		break;
2440 	}
2441 }
2442 
2443 /**
2444  * tipc_sk_rcv - handle a chain of incoming buffers
2445  * @inputq: buffer list containing the buffers
2446  * Consumes all buffers in list until inputq is empty
2447  * Note: may be called in multiple threads referring to the same queue
2448  */
2449 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2450 {
2451 	struct sk_buff_head xmitq;
2452 	u32 dnode, dport = 0;
2453 	int err;
2454 	struct tipc_sock *tsk;
2455 	struct sock *sk;
2456 	struct sk_buff *skb;
2457 
2458 	__skb_queue_head_init(&xmitq);
2459 	while (skb_queue_len(inputq)) {
2460 		dport = tipc_skb_peek_port(inputq, dport);
2461 		tsk = tipc_sk_lookup(net, dport);
2462 
2463 		if (likely(tsk)) {
2464 			sk = &tsk->sk;
2465 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2466 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2467 				spin_unlock_bh(&sk->sk_lock.slock);
2468 			}
2469 			/* Send pending response/rejected messages, if any */
2470 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2471 			sock_put(sk);
2472 			continue;
2473 		}
2474 		/* No destination socket => dequeue skb if still there */
2475 		skb = tipc_skb_dequeue(inputq, dport);
2476 		if (!skb)
2477 			return;
2478 
2479 		/* Try secondary lookup if unresolved named message */
2480 		err = TIPC_ERR_NO_PORT;
2481 		if (tipc_msg_lookup_dest(net, skb, &err))
2482 			goto xmit;
2483 
2484 		/* Prepare for message rejection */
2485 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2486 			continue;
2487 
2488 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2489 xmit:
2490 		dnode = msg_destnode(buf_msg(skb));
2491 		tipc_node_xmit_skb(net, skb, dnode, dport);
2492 	}
2493 }
2494 
2495 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2496 {
2497 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2498 	struct sock *sk = sock->sk;
2499 	int done;
2500 
2501 	do {
2502 		int err = sock_error(sk);
2503 		if (err)
2504 			return err;
2505 		if (!*timeo_p)
2506 			return -ETIMEDOUT;
2507 		if (signal_pending(current))
2508 			return sock_intr_errno(*timeo_p);
2509 		if (sk->sk_state == TIPC_DISCONNECTING)
2510 			break;
2511 
2512 		add_wait_queue(sk_sleep(sk), &wait);
2513 		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2514 				     &wait);
2515 		remove_wait_queue(sk_sleep(sk), &wait);
2516 	} while (!done);
2517 	return 0;
2518 }
2519 
2520 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2521 {
2522 	if (addr->family != AF_TIPC)
2523 		return false;
2524 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2525 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2526 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2527 		addr->addrtype == TIPC_SOCKET_ADDR);
2528 }
2529 
2530 /**
2531  * tipc_connect - establish a connection to another TIPC port
2532  * @sock: socket structure
2533  * @dest: socket address for destination port
2534  * @destlen: size of socket address data structure
2535  * @flags: file-related flags associated with socket
2536  *
2537  * Returns 0 on success, errno otherwise
2538  */
2539 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2540 			int destlen, int flags)
2541 {
2542 	struct sock *sk = sock->sk;
2543 	struct tipc_sock *tsk = tipc_sk(sk);
2544 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2545 	struct msghdr m = {NULL,};
2546 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2547 	int previous;
2548 	int res = 0;
2549 
2550 	if (destlen != sizeof(struct sockaddr_tipc))
2551 		return -EINVAL;
2552 
2553 	lock_sock(sk);
2554 
2555 	if (tsk->group) {
2556 		res = -EINVAL;
2557 		goto exit;
2558 	}
2559 
2560 	if (dst->family == AF_UNSPEC) {
2561 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2562 		if (!tipc_sk_type_connectionless(sk))
2563 			res = -EINVAL;
2564 		goto exit;
2565 	}
2566 	if (!tipc_sockaddr_is_sane(dst)) {
2567 		res = -EINVAL;
2568 		goto exit;
2569 	}
2570 	/* DGRAM/RDM connect(), just save the destaddr */
2571 	if (tipc_sk_type_connectionless(sk)) {
2572 		memcpy(&tsk->peer, dest, destlen);
2573 		goto exit;
2574 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2575 		res = -EINVAL;
2576 		goto exit;
2577 	}
2578 
2579 	previous = sk->sk_state;
2580 
2581 	switch (sk->sk_state) {
2582 	case TIPC_OPEN:
2583 		/* Send a 'SYN-' to destination */
2584 		m.msg_name = dest;
2585 		m.msg_namelen = destlen;
2586 
2587 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2588 		 * indicate send_msg() is never blocked.
2589 		 */
2590 		if (!timeout)
2591 			m.msg_flags = MSG_DONTWAIT;
2592 
2593 		res = __tipc_sendmsg(sock, &m, 0);
2594 		if ((res < 0) && (res != -EWOULDBLOCK))
2595 			goto exit;
2596 
2597 		/* Just entered TIPC_CONNECTING state; the only
2598 		 * difference is that return value in non-blocking
2599 		 * case is EINPROGRESS, rather than EALREADY.
2600 		 */
2601 		res = -EINPROGRESS;
2602 		/* fall through */
2603 	case TIPC_CONNECTING:
2604 		if (!timeout) {
2605 			if (previous == TIPC_CONNECTING)
2606 				res = -EALREADY;
2607 			goto exit;
2608 		}
2609 		timeout = msecs_to_jiffies(timeout);
2610 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2611 		res = tipc_wait_for_connect(sock, &timeout);
2612 		break;
2613 	case TIPC_ESTABLISHED:
2614 		res = -EISCONN;
2615 		break;
2616 	default:
2617 		res = -EINVAL;
2618 	}
2619 
2620 exit:
2621 	release_sock(sk);
2622 	return res;
2623 }
2624 
2625 /**
2626  * tipc_listen - allow socket to listen for incoming connections
2627  * @sock: socket structure
2628  * @len: (unused)
2629  *
2630  * Returns 0 on success, errno otherwise
2631  */
2632 static int tipc_listen(struct socket *sock, int len)
2633 {
2634 	struct sock *sk = sock->sk;
2635 	int res;
2636 
2637 	lock_sock(sk);
2638 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2639 	release_sock(sk);
2640 
2641 	return res;
2642 }
2643 
2644 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2645 {
2646 	struct sock *sk = sock->sk;
2647 	DEFINE_WAIT(wait);
2648 	int err;
2649 
2650 	/* True wake-one mechanism for incoming connections: only
2651 	 * one process gets woken up, not the 'whole herd'.
2652 	 * Since we do not 'race & poll' for established sockets
2653 	 * anymore, the common case will execute the loop only once.
2654 	*/
2655 	for (;;) {
2656 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2657 					  TASK_INTERRUPTIBLE);
2658 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2659 			release_sock(sk);
2660 			timeo = schedule_timeout(timeo);
2661 			lock_sock(sk);
2662 		}
2663 		err = 0;
2664 		if (!skb_queue_empty(&sk->sk_receive_queue))
2665 			break;
2666 		err = -EAGAIN;
2667 		if (!timeo)
2668 			break;
2669 		err = sock_intr_errno(timeo);
2670 		if (signal_pending(current))
2671 			break;
2672 	}
2673 	finish_wait(sk_sleep(sk), &wait);
2674 	return err;
2675 }
2676 
2677 /**
2678  * tipc_accept - wait for connection request
2679  * @sock: listening socket
2680  * @newsock: new socket that is to be connected
2681  * @flags: file-related flags associated with socket
2682  *
2683  * Returns 0 on success, errno otherwise
2684  */
2685 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2686 		       bool kern)
2687 {
2688 	struct sock *new_sk, *sk = sock->sk;
2689 	struct sk_buff *buf;
2690 	struct tipc_sock *new_tsock;
2691 	struct tipc_msg *msg;
2692 	long timeo;
2693 	int res;
2694 
2695 	lock_sock(sk);
2696 
2697 	if (sk->sk_state != TIPC_LISTEN) {
2698 		res = -EINVAL;
2699 		goto exit;
2700 	}
2701 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2702 	res = tipc_wait_for_accept(sock, timeo);
2703 	if (res)
2704 		goto exit;
2705 
2706 	buf = skb_peek(&sk->sk_receive_queue);
2707 
2708 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2709 	if (res)
2710 		goto exit;
2711 	security_sk_clone(sock->sk, new_sock->sk);
2712 
2713 	new_sk = new_sock->sk;
2714 	new_tsock = tipc_sk(new_sk);
2715 	msg = buf_msg(buf);
2716 
2717 	/* we lock on new_sk; but lockdep sees the lock on sk */
2718 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2719 
2720 	/*
2721 	 * Reject any stray messages received by new socket
2722 	 * before the socket lock was taken (very, very unlikely)
2723 	 */
2724 	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2725 
2726 	/* Connect new socket to it's peer */
2727 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2728 
2729 	tsk_set_importance(new_sk, msg_importance(msg));
2730 	if (msg_named(msg)) {
2731 		new_tsock->conn_type = msg_nametype(msg);
2732 		new_tsock->conn_instance = msg_nameinst(msg);
2733 	}
2734 
2735 	/*
2736 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2737 	 * Respond to 'SYN+' by queuing it on new socket.
2738 	 */
2739 	if (!msg_data_sz(msg)) {
2740 		struct msghdr m = {NULL,};
2741 
2742 		tsk_advance_rx_queue(sk);
2743 		__tipc_sendstream(new_sock, &m, 0);
2744 	} else {
2745 		__skb_dequeue(&sk->sk_receive_queue);
2746 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2747 		skb_set_owner_r(buf, new_sk);
2748 	}
2749 	release_sock(new_sk);
2750 exit:
2751 	release_sock(sk);
2752 	return res;
2753 }
2754 
2755 /**
2756  * tipc_shutdown - shutdown socket connection
2757  * @sock: socket structure
2758  * @how: direction to close (must be SHUT_RDWR)
2759  *
2760  * Terminates connection (if necessary), then purges socket's receive queue.
2761  *
2762  * Returns 0 on success, errno otherwise
2763  */
2764 static int tipc_shutdown(struct socket *sock, int how)
2765 {
2766 	struct sock *sk = sock->sk;
2767 	int res;
2768 
2769 	if (how != SHUT_RDWR)
2770 		return -EINVAL;
2771 
2772 	lock_sock(sk);
2773 
2774 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2775 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2776 	sk->sk_shutdown = SEND_SHUTDOWN;
2777 
2778 	if (sk->sk_state == TIPC_DISCONNECTING) {
2779 		/* Discard any unreceived messages */
2780 		__skb_queue_purge(&sk->sk_receive_queue);
2781 
2782 		/* Wake up anyone sleeping in poll */
2783 		sk->sk_state_change(sk);
2784 		res = 0;
2785 	} else {
2786 		res = -ENOTCONN;
2787 	}
2788 
2789 	release_sock(sk);
2790 	return res;
2791 }
2792 
2793 static void tipc_sk_check_probing_state(struct sock *sk,
2794 					struct sk_buff_head *list)
2795 {
2796 	struct tipc_sock *tsk = tipc_sk(sk);
2797 	u32 pnode = tsk_peer_node(tsk);
2798 	u32 pport = tsk_peer_port(tsk);
2799 	u32 self = tsk_own_node(tsk);
2800 	u32 oport = tsk->portid;
2801 	struct sk_buff *skb;
2802 
2803 	if (tsk->probe_unacked) {
2804 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2805 		sk->sk_err = ECONNABORTED;
2806 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2807 		sk->sk_state_change(sk);
2808 		return;
2809 	}
2810 	/* Prepare new probe */
2811 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2812 			      pnode, self, pport, oport, TIPC_OK);
2813 	if (skb)
2814 		__skb_queue_tail(list, skb);
2815 	tsk->probe_unacked = true;
2816 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2817 }
2818 
2819 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2820 {
2821 	struct tipc_sock *tsk = tipc_sk(sk);
2822 
2823 	/* Try again later if dest link is congested */
2824 	if (tsk->cong_link_cnt) {
2825 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2826 		return;
2827 	}
2828 	/* Prepare SYN for retransmit */
2829 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2830 }
2831 
2832 static void tipc_sk_timeout(struct timer_list *t)
2833 {
2834 	struct sock *sk = from_timer(sk, t, sk_timer);
2835 	struct tipc_sock *tsk = tipc_sk(sk);
2836 	u32 pnode = tsk_peer_node(tsk);
2837 	struct sk_buff_head list;
2838 	int rc = 0;
2839 
2840 	__skb_queue_head_init(&list);
2841 	bh_lock_sock(sk);
2842 
2843 	/* Try again later if socket is busy */
2844 	if (sock_owned_by_user(sk)) {
2845 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2846 		bh_unlock_sock(sk);
2847 		sock_put(sk);
2848 		return;
2849 	}
2850 
2851 	if (sk->sk_state == TIPC_ESTABLISHED)
2852 		tipc_sk_check_probing_state(sk, &list);
2853 	else if (sk->sk_state == TIPC_CONNECTING)
2854 		tipc_sk_retry_connect(sk, &list);
2855 
2856 	bh_unlock_sock(sk);
2857 
2858 	if (!skb_queue_empty(&list))
2859 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2860 
2861 	/* SYN messages may cause link congestion */
2862 	if (rc == -ELINKCONG) {
2863 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2864 		tsk->cong_link_cnt = 1;
2865 	}
2866 	sock_put(sk);
2867 }
2868 
2869 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2870 			   struct tipc_name_seq const *seq)
2871 {
2872 	struct sock *sk = &tsk->sk;
2873 	struct net *net = sock_net(sk);
2874 	struct publication *publ;
2875 	u32 key;
2876 
2877 	if (scope != TIPC_NODE_SCOPE)
2878 		scope = TIPC_CLUSTER_SCOPE;
2879 
2880 	if (tipc_sk_connected(sk))
2881 		return -EINVAL;
2882 	key = tsk->portid + tsk->pub_count + 1;
2883 	if (key == tsk->portid)
2884 		return -EADDRINUSE;
2885 
2886 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2887 				    scope, tsk->portid, key);
2888 	if (unlikely(!publ))
2889 		return -EINVAL;
2890 
2891 	list_add(&publ->binding_sock, &tsk->publications);
2892 	tsk->pub_count++;
2893 	tsk->published = 1;
2894 	return 0;
2895 }
2896 
2897 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2898 			    struct tipc_name_seq const *seq)
2899 {
2900 	struct net *net = sock_net(&tsk->sk);
2901 	struct publication *publ;
2902 	struct publication *safe;
2903 	int rc = -EINVAL;
2904 
2905 	if (scope != TIPC_NODE_SCOPE)
2906 		scope = TIPC_CLUSTER_SCOPE;
2907 
2908 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2909 		if (seq) {
2910 			if (publ->scope != scope)
2911 				continue;
2912 			if (publ->type != seq->type)
2913 				continue;
2914 			if (publ->lower != seq->lower)
2915 				continue;
2916 			if (publ->upper != seq->upper)
2917 				break;
2918 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2919 					      publ->upper, publ->key);
2920 			rc = 0;
2921 			break;
2922 		}
2923 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2924 				      publ->upper, publ->key);
2925 		rc = 0;
2926 	}
2927 	if (list_empty(&tsk->publications))
2928 		tsk->published = 0;
2929 	return rc;
2930 }
2931 
2932 /* tipc_sk_reinit: set non-zero address in all existing sockets
2933  *                 when we go from standalone to network mode.
2934  */
2935 void tipc_sk_reinit(struct net *net)
2936 {
2937 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2938 	struct rhashtable_iter iter;
2939 	struct tipc_sock *tsk;
2940 	struct tipc_msg *msg;
2941 
2942 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2943 
2944 	do {
2945 		rhashtable_walk_start(&iter);
2946 
2947 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2948 			sock_hold(&tsk->sk);
2949 			rhashtable_walk_stop(&iter);
2950 			lock_sock(&tsk->sk);
2951 			msg = &tsk->phdr;
2952 			msg_set_prevnode(msg, tipc_own_addr(net));
2953 			msg_set_orignode(msg, tipc_own_addr(net));
2954 			release_sock(&tsk->sk);
2955 			rhashtable_walk_start(&iter);
2956 			sock_put(&tsk->sk);
2957 		}
2958 
2959 		rhashtable_walk_stop(&iter);
2960 	} while (tsk == ERR_PTR(-EAGAIN));
2961 
2962 	rhashtable_walk_exit(&iter);
2963 }
2964 
2965 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2966 {
2967 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2968 	struct tipc_sock *tsk;
2969 
2970 	rcu_read_lock();
2971 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2972 	if (tsk)
2973 		sock_hold(&tsk->sk);
2974 	rcu_read_unlock();
2975 
2976 	return tsk;
2977 }
2978 
2979 static int tipc_sk_insert(struct tipc_sock *tsk)
2980 {
2981 	struct sock *sk = &tsk->sk;
2982 	struct net *net = sock_net(sk);
2983 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2984 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2985 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2986 
2987 	while (remaining--) {
2988 		portid++;
2989 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2990 			portid = TIPC_MIN_PORT;
2991 		tsk->portid = portid;
2992 		sock_hold(&tsk->sk);
2993 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2994 						   tsk_rht_params))
2995 			return 0;
2996 		sock_put(&tsk->sk);
2997 	}
2998 
2999 	return -1;
3000 }
3001 
3002 static void tipc_sk_remove(struct tipc_sock *tsk)
3003 {
3004 	struct sock *sk = &tsk->sk;
3005 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3006 
3007 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3008 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3009 		__sock_put(sk);
3010 	}
3011 }
3012 
3013 static const struct rhashtable_params tsk_rht_params = {
3014 	.nelem_hint = 192,
3015 	.head_offset = offsetof(struct tipc_sock, node),
3016 	.key_offset = offsetof(struct tipc_sock, portid),
3017 	.key_len = sizeof(u32), /* portid */
3018 	.max_size = 1048576,
3019 	.min_size = 256,
3020 	.automatic_shrinking = true,
3021 };
3022 
3023 int tipc_sk_rht_init(struct net *net)
3024 {
3025 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3026 
3027 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3028 }
3029 
3030 void tipc_sk_rht_destroy(struct net *net)
3031 {
3032 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3033 
3034 	/* Wait for socket readers to complete */
3035 	synchronize_net();
3036 
3037 	rhashtable_destroy(&tn->sk_rht);
3038 }
3039 
3040 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3041 {
3042 	struct net *net = sock_net(&tsk->sk);
3043 	struct tipc_group *grp = tsk->group;
3044 	struct tipc_msg *hdr = &tsk->phdr;
3045 	struct tipc_name_seq seq;
3046 	int rc;
3047 
3048 	if (mreq->type < TIPC_RESERVED_TYPES)
3049 		return -EACCES;
3050 	if (mreq->scope > TIPC_NODE_SCOPE)
3051 		return -EINVAL;
3052 	if (grp)
3053 		return -EACCES;
3054 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3055 	if (!grp)
3056 		return -ENOMEM;
3057 	tsk->group = grp;
3058 	msg_set_lookup_scope(hdr, mreq->scope);
3059 	msg_set_nametype(hdr, mreq->type);
3060 	msg_set_dest_droppable(hdr, true);
3061 	seq.type = mreq->type;
3062 	seq.lower = mreq->instance;
3063 	seq.upper = seq.lower;
3064 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3065 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3066 	if (rc) {
3067 		tipc_group_delete(net, grp);
3068 		tsk->group = NULL;
3069 		return rc;
3070 	}
3071 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3072 	tsk->mc_method.rcast = true;
3073 	tsk->mc_method.mandatory = true;
3074 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3075 	return rc;
3076 }
3077 
3078 static int tipc_sk_leave(struct tipc_sock *tsk)
3079 {
3080 	struct net *net = sock_net(&tsk->sk);
3081 	struct tipc_group *grp = tsk->group;
3082 	struct tipc_name_seq seq;
3083 	int scope;
3084 
3085 	if (!grp)
3086 		return -EINVAL;
3087 	tipc_group_self(grp, &seq, &scope);
3088 	tipc_group_delete(net, grp);
3089 	tsk->group = NULL;
3090 	tipc_sk_withdraw(tsk, scope, &seq);
3091 	return 0;
3092 }
3093 
3094 /**
3095  * tipc_setsockopt - set socket option
3096  * @sock: socket structure
3097  * @lvl: option level
3098  * @opt: option identifier
3099  * @ov: pointer to new option value
3100  * @ol: length of option value
3101  *
3102  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3103  * (to ease compatibility).
3104  *
3105  * Returns 0 on success, errno otherwise
3106  */
3107 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3108 			   char __user *ov, unsigned int ol)
3109 {
3110 	struct sock *sk = sock->sk;
3111 	struct tipc_sock *tsk = tipc_sk(sk);
3112 	struct tipc_group_req mreq;
3113 	u32 value = 0;
3114 	int res = 0;
3115 
3116 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3117 		return 0;
3118 	if (lvl != SOL_TIPC)
3119 		return -ENOPROTOOPT;
3120 
3121 	switch (opt) {
3122 	case TIPC_IMPORTANCE:
3123 	case TIPC_SRC_DROPPABLE:
3124 	case TIPC_DEST_DROPPABLE:
3125 	case TIPC_CONN_TIMEOUT:
3126 	case TIPC_NODELAY:
3127 		if (ol < sizeof(value))
3128 			return -EINVAL;
3129 		if (get_user(value, (u32 __user *)ov))
3130 			return -EFAULT;
3131 		break;
3132 	case TIPC_GROUP_JOIN:
3133 		if (ol < sizeof(mreq))
3134 			return -EINVAL;
3135 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
3136 			return -EFAULT;
3137 		break;
3138 	default:
3139 		if (ov || ol)
3140 			return -EINVAL;
3141 	}
3142 
3143 	lock_sock(sk);
3144 
3145 	switch (opt) {
3146 	case TIPC_IMPORTANCE:
3147 		res = tsk_set_importance(sk, value);
3148 		break;
3149 	case TIPC_SRC_DROPPABLE:
3150 		if (sock->type != SOCK_STREAM)
3151 			tsk_set_unreliable(tsk, value);
3152 		else
3153 			res = -ENOPROTOOPT;
3154 		break;
3155 	case TIPC_DEST_DROPPABLE:
3156 		tsk_set_unreturnable(tsk, value);
3157 		break;
3158 	case TIPC_CONN_TIMEOUT:
3159 		tipc_sk(sk)->conn_timeout = value;
3160 		break;
3161 	case TIPC_MCAST_BROADCAST:
3162 		tsk->mc_method.rcast = false;
3163 		tsk->mc_method.mandatory = true;
3164 		break;
3165 	case TIPC_MCAST_REPLICAST:
3166 		tsk->mc_method.rcast = true;
3167 		tsk->mc_method.mandatory = true;
3168 		break;
3169 	case TIPC_GROUP_JOIN:
3170 		res = tipc_sk_join(tsk, &mreq);
3171 		break;
3172 	case TIPC_GROUP_LEAVE:
3173 		res = tipc_sk_leave(tsk);
3174 		break;
3175 	case TIPC_NODELAY:
3176 		tsk->nodelay = !!value;
3177 		tsk_set_nagle(tsk);
3178 		break;
3179 	default:
3180 		res = -EINVAL;
3181 	}
3182 
3183 	release_sock(sk);
3184 
3185 	return res;
3186 }
3187 
3188 /**
3189  * tipc_getsockopt - get socket option
3190  * @sock: socket structure
3191  * @lvl: option level
3192  * @opt: option identifier
3193  * @ov: receptacle for option value
3194  * @ol: receptacle for length of option value
3195  *
3196  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3197  * (to ease compatibility).
3198  *
3199  * Returns 0 on success, errno otherwise
3200  */
3201 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3202 			   char __user *ov, int __user *ol)
3203 {
3204 	struct sock *sk = sock->sk;
3205 	struct tipc_sock *tsk = tipc_sk(sk);
3206 	struct tipc_name_seq seq;
3207 	int len, scope;
3208 	u32 value;
3209 	int res;
3210 
3211 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3212 		return put_user(0, ol);
3213 	if (lvl != SOL_TIPC)
3214 		return -ENOPROTOOPT;
3215 	res = get_user(len, ol);
3216 	if (res)
3217 		return res;
3218 
3219 	lock_sock(sk);
3220 
3221 	switch (opt) {
3222 	case TIPC_IMPORTANCE:
3223 		value = tsk_importance(tsk);
3224 		break;
3225 	case TIPC_SRC_DROPPABLE:
3226 		value = tsk_unreliable(tsk);
3227 		break;
3228 	case TIPC_DEST_DROPPABLE:
3229 		value = tsk_unreturnable(tsk);
3230 		break;
3231 	case TIPC_CONN_TIMEOUT:
3232 		value = tsk->conn_timeout;
3233 		/* no need to set "res", since already 0 at this point */
3234 		break;
3235 	case TIPC_NODE_RECVQ_DEPTH:
3236 		value = 0; /* was tipc_queue_size, now obsolete */
3237 		break;
3238 	case TIPC_SOCK_RECVQ_DEPTH:
3239 		value = skb_queue_len(&sk->sk_receive_queue);
3240 		break;
3241 	case TIPC_SOCK_RECVQ_USED:
3242 		value = sk_rmem_alloc_get(sk);
3243 		break;
3244 	case TIPC_GROUP_JOIN:
3245 		seq.type = 0;
3246 		if (tsk->group)
3247 			tipc_group_self(tsk->group, &seq, &scope);
3248 		value = seq.type;
3249 		break;
3250 	default:
3251 		res = -EINVAL;
3252 	}
3253 
3254 	release_sock(sk);
3255 
3256 	if (res)
3257 		return res;	/* "get" failed */
3258 
3259 	if (len < sizeof(value))
3260 		return -EINVAL;
3261 
3262 	if (copy_to_user(ov, &value, sizeof(value)))
3263 		return -EFAULT;
3264 
3265 	return put_user(sizeof(value), ol);
3266 }
3267 
3268 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3269 {
3270 	struct net *net = sock_net(sock->sk);
3271 	struct tipc_sioc_nodeid_req nr = {0};
3272 	struct tipc_sioc_ln_req lnr;
3273 	void __user *argp = (void __user *)arg;
3274 
3275 	switch (cmd) {
3276 	case SIOCGETLINKNAME:
3277 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3278 			return -EFAULT;
3279 		if (!tipc_node_get_linkname(net,
3280 					    lnr.bearer_id & 0xffff, lnr.peer,
3281 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3282 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3283 				return -EFAULT;
3284 			return 0;
3285 		}
3286 		return -EADDRNOTAVAIL;
3287 	case SIOCGETNODEID:
3288 		if (copy_from_user(&nr, argp, sizeof(nr)))
3289 			return -EFAULT;
3290 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3291 			return -EADDRNOTAVAIL;
3292 		if (copy_to_user(argp, &nr, sizeof(nr)))
3293 			return -EFAULT;
3294 		return 0;
3295 	default:
3296 		return -ENOIOCTLCMD;
3297 	}
3298 }
3299 
3300 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3301 {
3302 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3303 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3304 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3305 
3306 	tsk1->peer.family = AF_TIPC;
3307 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3308 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3309 	tsk1->peer.addr.id.ref = tsk2->portid;
3310 	tsk1->peer.addr.id.node = onode;
3311 	tsk2->peer.family = AF_TIPC;
3312 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3313 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3314 	tsk2->peer.addr.id.ref = tsk1->portid;
3315 	tsk2->peer.addr.id.node = onode;
3316 
3317 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3318 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3319 	return 0;
3320 }
3321 
3322 /* Protocol switches for the various types of TIPC sockets */
3323 
3324 static const struct proto_ops msg_ops = {
3325 	.owner		= THIS_MODULE,
3326 	.family		= AF_TIPC,
3327 	.release	= tipc_release,
3328 	.bind		= tipc_bind,
3329 	.connect	= tipc_connect,
3330 	.socketpair	= tipc_socketpair,
3331 	.accept		= sock_no_accept,
3332 	.getname	= tipc_getname,
3333 	.poll		= tipc_poll,
3334 	.ioctl		= tipc_ioctl,
3335 	.listen		= sock_no_listen,
3336 	.shutdown	= tipc_shutdown,
3337 	.setsockopt	= tipc_setsockopt,
3338 	.getsockopt	= tipc_getsockopt,
3339 	.sendmsg	= tipc_sendmsg,
3340 	.recvmsg	= tipc_recvmsg,
3341 	.mmap		= sock_no_mmap,
3342 	.sendpage	= sock_no_sendpage
3343 };
3344 
3345 static const struct proto_ops packet_ops = {
3346 	.owner		= THIS_MODULE,
3347 	.family		= AF_TIPC,
3348 	.release	= tipc_release,
3349 	.bind		= tipc_bind,
3350 	.connect	= tipc_connect,
3351 	.socketpair	= tipc_socketpair,
3352 	.accept		= tipc_accept,
3353 	.getname	= tipc_getname,
3354 	.poll		= tipc_poll,
3355 	.ioctl		= tipc_ioctl,
3356 	.listen		= tipc_listen,
3357 	.shutdown	= tipc_shutdown,
3358 	.setsockopt	= tipc_setsockopt,
3359 	.getsockopt	= tipc_getsockopt,
3360 	.sendmsg	= tipc_send_packet,
3361 	.recvmsg	= tipc_recvmsg,
3362 	.mmap		= sock_no_mmap,
3363 	.sendpage	= sock_no_sendpage
3364 };
3365 
3366 static const struct proto_ops stream_ops = {
3367 	.owner		= THIS_MODULE,
3368 	.family		= AF_TIPC,
3369 	.release	= tipc_release,
3370 	.bind		= tipc_bind,
3371 	.connect	= tipc_connect,
3372 	.socketpair	= tipc_socketpair,
3373 	.accept		= tipc_accept,
3374 	.getname	= tipc_getname,
3375 	.poll		= tipc_poll,
3376 	.ioctl		= tipc_ioctl,
3377 	.listen		= tipc_listen,
3378 	.shutdown	= tipc_shutdown,
3379 	.setsockopt	= tipc_setsockopt,
3380 	.getsockopt	= tipc_getsockopt,
3381 	.sendmsg	= tipc_sendstream,
3382 	.recvmsg	= tipc_recvstream,
3383 	.mmap		= sock_no_mmap,
3384 	.sendpage	= sock_no_sendpage
3385 };
3386 
3387 static const struct net_proto_family tipc_family_ops = {
3388 	.owner		= THIS_MODULE,
3389 	.family		= AF_TIPC,
3390 	.create		= tipc_sk_create
3391 };
3392 
3393 static struct proto tipc_proto = {
3394 	.name		= "TIPC",
3395 	.owner		= THIS_MODULE,
3396 	.obj_size	= sizeof(struct tipc_sock),
3397 	.sysctl_rmem	= sysctl_tipc_rmem
3398 };
3399 
3400 /**
3401  * tipc_socket_init - initialize TIPC socket interface
3402  *
3403  * Returns 0 on success, errno otherwise
3404  */
3405 int tipc_socket_init(void)
3406 {
3407 	int res;
3408 
3409 	res = proto_register(&tipc_proto, 1);
3410 	if (res) {
3411 		pr_err("Failed to register TIPC protocol type\n");
3412 		goto out;
3413 	}
3414 
3415 	res = sock_register(&tipc_family_ops);
3416 	if (res) {
3417 		pr_err("Failed to register TIPC socket type\n");
3418 		proto_unregister(&tipc_proto);
3419 		goto out;
3420 	}
3421  out:
3422 	return res;
3423 }
3424 
3425 /**
3426  * tipc_socket_stop - stop TIPC socket interface
3427  */
3428 void tipc_socket_stop(void)
3429 {
3430 	sock_unregister(tipc_family_ops.family);
3431 	proto_unregister(&tipc_proto);
3432 }
3433 
3434 /* Caller should hold socket lock for the passed tipc socket. */
3435 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3436 {
3437 	u32 peer_node;
3438 	u32 peer_port;
3439 	struct nlattr *nest;
3440 
3441 	peer_node = tsk_peer_node(tsk);
3442 	peer_port = tsk_peer_port(tsk);
3443 
3444 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3445 	if (!nest)
3446 		return -EMSGSIZE;
3447 
3448 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3449 		goto msg_full;
3450 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3451 		goto msg_full;
3452 
3453 	if (tsk->conn_type != 0) {
3454 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3455 			goto msg_full;
3456 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3457 			goto msg_full;
3458 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3459 			goto msg_full;
3460 	}
3461 	nla_nest_end(skb, nest);
3462 
3463 	return 0;
3464 
3465 msg_full:
3466 	nla_nest_cancel(skb, nest);
3467 
3468 	return -EMSGSIZE;
3469 }
3470 
3471 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3472 			  *tsk)
3473 {
3474 	struct net *net = sock_net(skb->sk);
3475 	struct sock *sk = &tsk->sk;
3476 
3477 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3478 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3479 		return -EMSGSIZE;
3480 
3481 	if (tipc_sk_connected(sk)) {
3482 		if (__tipc_nl_add_sk_con(skb, tsk))
3483 			return -EMSGSIZE;
3484 	} else if (!list_empty(&tsk->publications)) {
3485 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3486 			return -EMSGSIZE;
3487 	}
3488 	return 0;
3489 }
3490 
3491 /* Caller should hold socket lock for the passed tipc socket. */
3492 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3493 			    struct tipc_sock *tsk)
3494 {
3495 	struct nlattr *attrs;
3496 	void *hdr;
3497 
3498 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3499 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3500 	if (!hdr)
3501 		goto msg_cancel;
3502 
3503 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3504 	if (!attrs)
3505 		goto genlmsg_cancel;
3506 
3507 	if (__tipc_nl_add_sk_info(skb, tsk))
3508 		goto attr_msg_cancel;
3509 
3510 	nla_nest_end(skb, attrs);
3511 	genlmsg_end(skb, hdr);
3512 
3513 	return 0;
3514 
3515 attr_msg_cancel:
3516 	nla_nest_cancel(skb, attrs);
3517 genlmsg_cancel:
3518 	genlmsg_cancel(skb, hdr);
3519 msg_cancel:
3520 	return -EMSGSIZE;
3521 }
3522 
3523 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3524 		    int (*skb_handler)(struct sk_buff *skb,
3525 				       struct netlink_callback *cb,
3526 				       struct tipc_sock *tsk))
3527 {
3528 	struct rhashtable_iter *iter = (void *)cb->args[4];
3529 	struct tipc_sock *tsk;
3530 	int err;
3531 
3532 	rhashtable_walk_start(iter);
3533 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3534 		if (IS_ERR(tsk)) {
3535 			err = PTR_ERR(tsk);
3536 			if (err == -EAGAIN) {
3537 				err = 0;
3538 				continue;
3539 			}
3540 			break;
3541 		}
3542 
3543 		sock_hold(&tsk->sk);
3544 		rhashtable_walk_stop(iter);
3545 		lock_sock(&tsk->sk);
3546 		err = skb_handler(skb, cb, tsk);
3547 		if (err) {
3548 			release_sock(&tsk->sk);
3549 			sock_put(&tsk->sk);
3550 			goto out;
3551 		}
3552 		release_sock(&tsk->sk);
3553 		rhashtable_walk_start(iter);
3554 		sock_put(&tsk->sk);
3555 	}
3556 	rhashtable_walk_stop(iter);
3557 out:
3558 	return skb->len;
3559 }
3560 EXPORT_SYMBOL(tipc_nl_sk_walk);
3561 
3562 int tipc_dump_start(struct netlink_callback *cb)
3563 {
3564 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3565 }
3566 EXPORT_SYMBOL(tipc_dump_start);
3567 
3568 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3569 {
3570 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3571 	struct rhashtable_iter *iter = (void *)cb->args[4];
3572 	struct tipc_net *tn = tipc_net(net);
3573 
3574 	if (!iter) {
3575 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3576 		if (!iter)
3577 			return -ENOMEM;
3578 
3579 		cb->args[4] = (long)iter;
3580 	}
3581 
3582 	rhashtable_walk_enter(&tn->sk_rht, iter);
3583 	return 0;
3584 }
3585 
3586 int tipc_dump_done(struct netlink_callback *cb)
3587 {
3588 	struct rhashtable_iter *hti = (void *)cb->args[4];
3589 
3590 	rhashtable_walk_exit(hti);
3591 	kfree(hti);
3592 	return 0;
3593 }
3594 EXPORT_SYMBOL(tipc_dump_done);
3595 
3596 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3597 			   struct tipc_sock *tsk, u32 sk_filter_state,
3598 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3599 {
3600 	struct sock *sk = &tsk->sk;
3601 	struct nlattr *attrs;
3602 	struct nlattr *stat;
3603 
3604 	/*filter response w.r.t sk_state*/
3605 	if (!(sk_filter_state & (1 << sk->sk_state)))
3606 		return 0;
3607 
3608 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3609 	if (!attrs)
3610 		goto msg_cancel;
3611 
3612 	if (__tipc_nl_add_sk_info(skb, tsk))
3613 		goto attr_msg_cancel;
3614 
3615 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3616 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3617 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3618 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3619 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3620 					 sock_i_uid(sk))) ||
3621 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3622 			      tipc_diag_gen_cookie(sk),
3623 			      TIPC_NLA_SOCK_PAD))
3624 		goto attr_msg_cancel;
3625 
3626 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3627 	if (!stat)
3628 		goto attr_msg_cancel;
3629 
3630 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3631 			skb_queue_len(&sk->sk_receive_queue)) ||
3632 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3633 			skb_queue_len(&sk->sk_write_queue)) ||
3634 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3635 			atomic_read(&sk->sk_drops)))
3636 		goto stat_msg_cancel;
3637 
3638 	if (tsk->cong_link_cnt &&
3639 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3640 		goto stat_msg_cancel;
3641 
3642 	if (tsk_conn_cong(tsk) &&
3643 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3644 		goto stat_msg_cancel;
3645 
3646 	nla_nest_end(skb, stat);
3647 
3648 	if (tsk->group)
3649 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3650 			goto stat_msg_cancel;
3651 
3652 	nla_nest_end(skb, attrs);
3653 
3654 	return 0;
3655 
3656 stat_msg_cancel:
3657 	nla_nest_cancel(skb, stat);
3658 attr_msg_cancel:
3659 	nla_nest_cancel(skb, attrs);
3660 msg_cancel:
3661 	return -EMSGSIZE;
3662 }
3663 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3664 
3665 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3666 {
3667 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3668 }
3669 
3670 /* Caller should hold socket lock for the passed tipc socket. */
3671 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3672 				 struct netlink_callback *cb,
3673 				 struct publication *publ)
3674 {
3675 	void *hdr;
3676 	struct nlattr *attrs;
3677 
3678 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3679 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3680 	if (!hdr)
3681 		goto msg_cancel;
3682 
3683 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3684 	if (!attrs)
3685 		goto genlmsg_cancel;
3686 
3687 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3688 		goto attr_msg_cancel;
3689 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3690 		goto attr_msg_cancel;
3691 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3692 		goto attr_msg_cancel;
3693 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3694 		goto attr_msg_cancel;
3695 
3696 	nla_nest_end(skb, attrs);
3697 	genlmsg_end(skb, hdr);
3698 
3699 	return 0;
3700 
3701 attr_msg_cancel:
3702 	nla_nest_cancel(skb, attrs);
3703 genlmsg_cancel:
3704 	genlmsg_cancel(skb, hdr);
3705 msg_cancel:
3706 	return -EMSGSIZE;
3707 }
3708 
3709 /* Caller should hold socket lock for the passed tipc socket. */
3710 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3711 				  struct netlink_callback *cb,
3712 				  struct tipc_sock *tsk, u32 *last_publ)
3713 {
3714 	int err;
3715 	struct publication *p;
3716 
3717 	if (*last_publ) {
3718 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3719 			if (p->key == *last_publ)
3720 				break;
3721 		}
3722 		if (p->key != *last_publ) {
3723 			/* We never set seq or call nl_dump_check_consistent()
3724 			 * this means that setting prev_seq here will cause the
3725 			 * consistence check to fail in the netlink callback
3726 			 * handler. Resulting in the last NLMSG_DONE message
3727 			 * having the NLM_F_DUMP_INTR flag set.
3728 			 */
3729 			cb->prev_seq = 1;
3730 			*last_publ = 0;
3731 			return -EPIPE;
3732 		}
3733 	} else {
3734 		p = list_first_entry(&tsk->publications, struct publication,
3735 				     binding_sock);
3736 	}
3737 
3738 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3739 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3740 		if (err) {
3741 			*last_publ = p->key;
3742 			return err;
3743 		}
3744 	}
3745 	*last_publ = 0;
3746 
3747 	return 0;
3748 }
3749 
3750 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3751 {
3752 	int err;
3753 	u32 tsk_portid = cb->args[0];
3754 	u32 last_publ = cb->args[1];
3755 	u32 done = cb->args[2];
3756 	struct net *net = sock_net(skb->sk);
3757 	struct tipc_sock *tsk;
3758 
3759 	if (!tsk_portid) {
3760 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3761 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3762 
3763 		if (!attrs[TIPC_NLA_SOCK])
3764 			return -EINVAL;
3765 
3766 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3767 						  attrs[TIPC_NLA_SOCK],
3768 						  tipc_nl_sock_policy, NULL);
3769 		if (err)
3770 			return err;
3771 
3772 		if (!sock[TIPC_NLA_SOCK_REF])
3773 			return -EINVAL;
3774 
3775 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3776 	}
3777 
3778 	if (done)
3779 		return 0;
3780 
3781 	tsk = tipc_sk_lookup(net, tsk_portid);
3782 	if (!tsk)
3783 		return -EINVAL;
3784 
3785 	lock_sock(&tsk->sk);
3786 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3787 	if (!err)
3788 		done = 1;
3789 	release_sock(&tsk->sk);
3790 	sock_put(&tsk->sk);
3791 
3792 	cb->args[0] = tsk_portid;
3793 	cb->args[1] = last_publ;
3794 	cb->args[2] = done;
3795 
3796 	return skb->len;
3797 }
3798 
3799 /**
3800  * tipc_sk_filtering - check if a socket should be traced
3801  * @sk: the socket to be examined
3802  * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3803  *  (portid, sock type, name type, name lower, name upper)
3804  *
3805  * Returns true if the socket meets the socket tuple data
3806  * (value 0 = 'any') or when there is no tuple set (all = 0),
3807  * otherwise false
3808  */
3809 bool tipc_sk_filtering(struct sock *sk)
3810 {
3811 	struct tipc_sock *tsk;
3812 	struct publication *p;
3813 	u32 _port, _sktype, _type, _lower, _upper;
3814 	u32 type = 0, lower = 0, upper = 0;
3815 
3816 	if (!sk)
3817 		return true;
3818 
3819 	tsk = tipc_sk(sk);
3820 
3821 	_port = sysctl_tipc_sk_filter[0];
3822 	_sktype = sysctl_tipc_sk_filter[1];
3823 	_type = sysctl_tipc_sk_filter[2];
3824 	_lower = sysctl_tipc_sk_filter[3];
3825 	_upper = sysctl_tipc_sk_filter[4];
3826 
3827 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3828 		return true;
3829 
3830 	if (_port)
3831 		return (_port == tsk->portid);
3832 
3833 	if (_sktype && _sktype != sk->sk_type)
3834 		return false;
3835 
3836 	if (tsk->published) {
3837 		p = list_first_entry_or_null(&tsk->publications,
3838 					     struct publication, binding_sock);
3839 		if (p) {
3840 			type = p->type;
3841 			lower = p->lower;
3842 			upper = p->upper;
3843 		}
3844 	}
3845 
3846 	if (!tipc_sk_type_connectionless(sk)) {
3847 		type = tsk->conn_type;
3848 		lower = tsk->conn_instance;
3849 		upper = tsk->conn_instance;
3850 	}
3851 
3852 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3853 	    (_upper && _upper != upper))
3854 		return false;
3855 
3856 	return true;
3857 }
3858 
3859 u32 tipc_sock_get_portid(struct sock *sk)
3860 {
3861 	return (sk) ? (tipc_sk(sk))->portid : 0;
3862 }
3863 
3864 /**
3865  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3866  *			both the rcv and backlog queues are considered
3867  * @sk: tipc sk to be checked
3868  * @skb: tipc msg to be checked
3869  *
3870  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3871  */
3872 
3873 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3874 {
3875 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3876 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3877 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3878 
3879 	return (qsize > lim * 90 / 100);
3880 }
3881 
3882 /**
3883  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3884  *			only the rcv queue is considered
3885  * @sk: tipc sk to be checked
3886  * @skb: tipc msg to be checked
3887  *
3888  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3889  */
3890 
3891 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3892 {
3893 	unsigned int lim = rcvbuf_limit(sk, skb);
3894 	unsigned int qsize = sk_rmem_alloc_get(sk);
3895 
3896 	return (qsize > lim * 90 / 100);
3897 }
3898 
3899 /**
3900  * tipc_sk_dump - dump TIPC socket
3901  * @sk: tipc sk to be dumped
3902  * @dqueues: bitmask to decide if any socket queue to be dumped?
3903  *           - TIPC_DUMP_NONE: don't dump socket queues
3904  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3905  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3906  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3907  *           - TIPC_DUMP_ALL: dump all the socket queues above
3908  * @buf: returned buffer of dump data in format
3909  */
3910 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3911 {
3912 	int i = 0;
3913 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3914 	struct tipc_sock *tsk;
3915 	struct publication *p;
3916 	bool tsk_connected;
3917 
3918 	if (!sk) {
3919 		i += scnprintf(buf, sz, "sk data: (null)\n");
3920 		return i;
3921 	}
3922 
3923 	tsk = tipc_sk(sk);
3924 	tsk_connected = !tipc_sk_type_connectionless(sk);
3925 
3926 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3927 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3928 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3929 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3930 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3931 	if (tsk_connected) {
3932 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3933 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3934 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3935 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3936 	}
3937 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3938 	if (tsk->published) {
3939 		p = list_first_entry_or_null(&tsk->publications,
3940 					     struct publication, binding_sock);
3941 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3942 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3943 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3944 	}
3945 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3946 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3947 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3948 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3949 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3950 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3951 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3952 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3953 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3954 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3955 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3956 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3957 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3958 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3959 
3960 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3961 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3962 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3963 	}
3964 
3965 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3966 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3967 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3968 	}
3969 
3970 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3971 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3972 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3973 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3974 			i += scnprintf(buf + i, sz - i, "  tail ");
3975 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3976 					   buf + i);
3977 		}
3978 	}
3979 
3980 	return i;
3981 }
3982