xref: /linux/net/tipc/socket.c (revision a4989fa91110508b64eea7ccde63d062113988ff)
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_MAX_PORT		0xffffffff
56 #define TIPC_MIN_PORT		1
57 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of rcv window size */
58 
59 enum {
60 	TIPC_LISTEN = TCP_LISTEN,
61 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
62 	TIPC_OPEN = TCP_CLOSE,
63 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
64 	TIPC_CONNECTING = TCP_SYN_SENT,
65 };
66 
67 struct sockaddr_pair {
68 	struct sockaddr_tipc sock;
69 	struct sockaddr_tipc member;
70 };
71 
72 /**
73  * struct tipc_sock - TIPC socket structure
74  * @sk: socket - interacts with 'port' and with user via the socket API
75  * @conn_type: TIPC type used when connection was established
76  * @conn_instance: TIPC instance used when connection was established
77  * @published: non-zero if port has one or more associated names
78  * @max_pkt: maximum packet size "hint" used when building messages sent by port
79  * @maxnagle: maximum size of msg which can be subject to nagle
80  * @portid: unique port identity in TIPC socket hash table
81  * @phdr: preformatted message header used when sending messages
82  * #cong_links: list of congested links
83  * @publications: list of publications for port
84  * @blocking_link: address of the congested link we are currently sleeping on
85  * @pub_count: total # of publications port has made during its lifetime
86  * @conn_timeout: the time we can wait for an unresponded setup request
87  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
88  * @cong_link_cnt: number of congested links
89  * @snt_unacked: # messages sent by socket, and not yet acked by peer
90  * @rcv_unacked: # messages read by user, but not yet acked back to peer
91  * @peer: 'connected' peer for dgram/rdm
92  * @node: hash table node
93  * @mc_method: cookie for use between socket and broadcast layer
94  * @rcu: rcu struct for tipc_sock
95  */
96 struct tipc_sock {
97 	struct sock sk;
98 	u32 conn_type;
99 	u32 conn_instance;
100 	int published;
101 	u32 max_pkt;
102 	u32 maxnagle;
103 	u32 portid;
104 	struct tipc_msg phdr;
105 	struct list_head cong_links;
106 	struct list_head publications;
107 	u32 pub_count;
108 	atomic_t dupl_rcvcnt;
109 	u16 conn_timeout;
110 	bool probe_unacked;
111 	u16 cong_link_cnt;
112 	u16 snt_unacked;
113 	u16 snd_win;
114 	u16 peer_caps;
115 	u16 rcv_unacked;
116 	u16 rcv_win;
117 	struct sockaddr_tipc peer;
118 	struct rhash_head node;
119 	struct tipc_mc_method mc_method;
120 	struct rcu_head rcu;
121 	struct tipc_group *group;
122 	u32 oneway;
123 	u32 nagle_start;
124 	u16 snd_backlog;
125 	u16 msg_acc;
126 	u16 pkt_cnt;
127 	bool expect_ack;
128 	bool nodelay;
129 	bool group_is_open;
130 };
131 
132 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
133 static void tipc_data_ready(struct sock *sk);
134 static void tipc_write_space(struct sock *sk);
135 static void tipc_sock_destruct(struct sock *sk);
136 static int tipc_release(struct socket *sock);
137 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
138 		       bool kern);
139 static void tipc_sk_timeout(struct timer_list *t);
140 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
141 			   struct tipc_name_seq const *seq);
142 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
143 			    struct tipc_name_seq const *seq);
144 static int tipc_sk_leave(struct tipc_sock *tsk);
145 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
146 static int tipc_sk_insert(struct tipc_sock *tsk);
147 static void tipc_sk_remove(struct tipc_sock *tsk);
148 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
149 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
150 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
151 
152 static const struct proto_ops packet_ops;
153 static const struct proto_ops stream_ops;
154 static const struct proto_ops msg_ops;
155 static struct proto tipc_proto;
156 static const struct rhashtable_params tsk_rht_params;
157 
158 static u32 tsk_own_node(struct tipc_sock *tsk)
159 {
160 	return msg_prevnode(&tsk->phdr);
161 }
162 
163 static u32 tsk_peer_node(struct tipc_sock *tsk)
164 {
165 	return msg_destnode(&tsk->phdr);
166 }
167 
168 static u32 tsk_peer_port(struct tipc_sock *tsk)
169 {
170 	return msg_destport(&tsk->phdr);
171 }
172 
173 static  bool tsk_unreliable(struct tipc_sock *tsk)
174 {
175 	return msg_src_droppable(&tsk->phdr) != 0;
176 }
177 
178 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
179 {
180 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
181 }
182 
183 static bool tsk_unreturnable(struct tipc_sock *tsk)
184 {
185 	return msg_dest_droppable(&tsk->phdr) != 0;
186 }
187 
188 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
189 {
190 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
191 }
192 
193 static int tsk_importance(struct tipc_sock *tsk)
194 {
195 	return msg_importance(&tsk->phdr);
196 }
197 
198 static struct tipc_sock *tipc_sk(const struct sock *sk)
199 {
200 	return container_of(sk, struct tipc_sock, sk);
201 }
202 
203 int tsk_set_importance(struct sock *sk, int imp)
204 {
205 	if (imp > TIPC_CRITICAL_IMPORTANCE)
206 		return -EINVAL;
207 	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
208 	return 0;
209 }
210 
211 static bool tsk_conn_cong(struct tipc_sock *tsk)
212 {
213 	return tsk->snt_unacked > tsk->snd_win;
214 }
215 
216 static u16 tsk_blocks(int len)
217 {
218 	return ((len / FLOWCTL_BLK_SZ) + 1);
219 }
220 
221 /* tsk_blocks(): translate a buffer size in bytes to number of
222  * advertisable blocks, taking into account the ratio truesize(len)/len
223  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
224  */
225 static u16 tsk_adv_blocks(int len)
226 {
227 	return len / FLOWCTL_BLK_SZ / 4;
228 }
229 
230 /* tsk_inc(): increment counter for sent or received data
231  * - If block based flow control is not supported by peer we
232  *   fall back to message based ditto, incrementing the counter
233  */
234 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
235 {
236 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
237 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
238 	return 1;
239 }
240 
241 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
242  */
243 static void tsk_set_nagle(struct tipc_sock *tsk)
244 {
245 	struct sock *sk = &tsk->sk;
246 
247 	tsk->maxnagle = 0;
248 	if (sk->sk_type != SOCK_STREAM)
249 		return;
250 	if (tsk->nodelay)
251 		return;
252 	if (!(tsk->peer_caps & TIPC_NAGLE))
253 		return;
254 	/* Limit node local buffer size to avoid receive queue overflow */
255 	if (tsk->max_pkt == MAX_MSG_SIZE)
256 		tsk->maxnagle = 1500;
257 	else
258 		tsk->maxnagle = tsk->max_pkt;
259 }
260 
261 /**
262  * tsk_advance_rx_queue - discard first buffer in socket receive queue
263  *
264  * Caller must hold socket lock
265  */
266 static void tsk_advance_rx_queue(struct sock *sk)
267 {
268 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
269 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
270 }
271 
272 /* tipc_sk_respond() : send response message back to sender
273  */
274 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
275 {
276 	u32 selector;
277 	u32 dnode;
278 	u32 onode = tipc_own_addr(sock_net(sk));
279 
280 	if (!tipc_msg_reverse(onode, &skb, err))
281 		return;
282 
283 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
284 	dnode = msg_destnode(buf_msg(skb));
285 	selector = msg_origport(buf_msg(skb));
286 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
287 }
288 
289 /**
290  * tsk_rej_rx_queue - reject all buffers in socket receive queue
291  *
292  * Caller must hold socket lock
293  */
294 static void tsk_rej_rx_queue(struct sock *sk, int error)
295 {
296 	struct sk_buff *skb;
297 
298 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
299 		tipc_sk_respond(sk, skb, error);
300 }
301 
302 static bool tipc_sk_connected(struct sock *sk)
303 {
304 	return sk->sk_state == TIPC_ESTABLISHED;
305 }
306 
307 /* tipc_sk_type_connectionless - check if the socket is datagram socket
308  * @sk: socket
309  *
310  * Returns true if connection less, false otherwise
311  */
312 static bool tipc_sk_type_connectionless(struct sock *sk)
313 {
314 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
315 }
316 
317 /* tsk_peer_msg - verify if message was sent by connected port's peer
318  *
319  * Handles cases where the node's network address has changed from
320  * the default of <0.0.0> to its configured setting.
321  */
322 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
323 {
324 	struct sock *sk = &tsk->sk;
325 	u32 self = tipc_own_addr(sock_net(sk));
326 	u32 peer_port = tsk_peer_port(tsk);
327 	u32 orig_node, peer_node;
328 
329 	if (unlikely(!tipc_sk_connected(sk)))
330 		return false;
331 
332 	if (unlikely(msg_origport(msg) != peer_port))
333 		return false;
334 
335 	orig_node = msg_orignode(msg);
336 	peer_node = tsk_peer_node(tsk);
337 
338 	if (likely(orig_node == peer_node))
339 		return true;
340 
341 	if (!orig_node && peer_node == self)
342 		return true;
343 
344 	if (!peer_node && orig_node == self)
345 		return true;
346 
347 	return false;
348 }
349 
350 /* tipc_set_sk_state - set the sk_state of the socket
351  * @sk: socket
352  *
353  * Caller must hold socket lock
354  *
355  * Returns 0 on success, errno otherwise
356  */
357 static int tipc_set_sk_state(struct sock *sk, int state)
358 {
359 	int oldsk_state = sk->sk_state;
360 	int res = -EINVAL;
361 
362 	switch (state) {
363 	case TIPC_OPEN:
364 		res = 0;
365 		break;
366 	case TIPC_LISTEN:
367 	case TIPC_CONNECTING:
368 		if (oldsk_state == TIPC_OPEN)
369 			res = 0;
370 		break;
371 	case TIPC_ESTABLISHED:
372 		if (oldsk_state == TIPC_CONNECTING ||
373 		    oldsk_state == TIPC_OPEN)
374 			res = 0;
375 		break;
376 	case TIPC_DISCONNECTING:
377 		if (oldsk_state == TIPC_CONNECTING ||
378 		    oldsk_state == TIPC_ESTABLISHED)
379 			res = 0;
380 		break;
381 	}
382 
383 	if (!res)
384 		sk->sk_state = state;
385 
386 	return res;
387 }
388 
389 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
390 {
391 	struct sock *sk = sock->sk;
392 	int err = sock_error(sk);
393 	int typ = sock->type;
394 
395 	if (err)
396 		return err;
397 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
398 		if (sk->sk_state == TIPC_DISCONNECTING)
399 			return -EPIPE;
400 		else if (!tipc_sk_connected(sk))
401 			return -ENOTCONN;
402 	}
403 	if (!*timeout)
404 		return -EAGAIN;
405 	if (signal_pending(current))
406 		return sock_intr_errno(*timeout);
407 
408 	return 0;
409 }
410 
411 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
412 ({                                                                             \
413 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
414 	struct sock *sk_;						       \
415 	int rc_;							       \
416 									       \
417 	while ((rc_ = !(condition_))) {					       \
418 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
419 		smp_rmb();                                                     \
420 		sk_ = (sock_)->sk;					       \
421 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
422 		if (rc_)						       \
423 			break;						       \
424 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
425 		release_sock(sk_);					       \
426 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
427 		sched_annotate_sleep();				               \
428 		lock_sock(sk_);						       \
429 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
430 	}								       \
431 	rc_;								       \
432 })
433 
434 /**
435  * tipc_sk_create - create a TIPC socket
436  * @net: network namespace (must be default network)
437  * @sock: pre-allocated socket structure
438  * @protocol: protocol indicator (must be 0)
439  * @kern: caused by kernel or by userspace?
440  *
441  * This routine creates additional data structures used by the TIPC socket,
442  * initializes them, and links them together.
443  *
444  * Returns 0 on success, errno otherwise
445  */
446 static int tipc_sk_create(struct net *net, struct socket *sock,
447 			  int protocol, int kern)
448 {
449 	const struct proto_ops *ops;
450 	struct sock *sk;
451 	struct tipc_sock *tsk;
452 	struct tipc_msg *msg;
453 
454 	/* Validate arguments */
455 	if (unlikely(protocol != 0))
456 		return -EPROTONOSUPPORT;
457 
458 	switch (sock->type) {
459 	case SOCK_STREAM:
460 		ops = &stream_ops;
461 		break;
462 	case SOCK_SEQPACKET:
463 		ops = &packet_ops;
464 		break;
465 	case SOCK_DGRAM:
466 	case SOCK_RDM:
467 		ops = &msg_ops;
468 		break;
469 	default:
470 		return -EPROTOTYPE;
471 	}
472 
473 	/* Allocate socket's protocol area */
474 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
475 	if (sk == NULL)
476 		return -ENOMEM;
477 
478 	tsk = tipc_sk(sk);
479 	tsk->max_pkt = MAX_PKT_DEFAULT;
480 	tsk->maxnagle = 0;
481 	tsk->nagle_start = NAGLE_START_INIT;
482 	INIT_LIST_HEAD(&tsk->publications);
483 	INIT_LIST_HEAD(&tsk->cong_links);
484 	msg = &tsk->phdr;
485 
486 	/* Finish initializing socket data structures */
487 	sock->ops = ops;
488 	sock_init_data(sock, sk);
489 	tipc_set_sk_state(sk, TIPC_OPEN);
490 	if (tipc_sk_insert(tsk)) {
491 		pr_warn("Socket create failed; port number exhausted\n");
492 		return -EINVAL;
493 	}
494 
495 	/* Ensure tsk is visible before we read own_addr. */
496 	smp_mb();
497 
498 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
499 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
500 
501 	msg_set_origport(msg, tsk->portid);
502 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
503 	sk->sk_shutdown = 0;
504 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
505 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
506 	sk->sk_data_ready = tipc_data_ready;
507 	sk->sk_write_space = tipc_write_space;
508 	sk->sk_destruct = tipc_sock_destruct;
509 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
510 	tsk->group_is_open = true;
511 	atomic_set(&tsk->dupl_rcvcnt, 0);
512 
513 	/* Start out with safe limits until we receive an advertised window */
514 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
515 	tsk->rcv_win = tsk->snd_win;
516 
517 	if (tipc_sk_type_connectionless(sk)) {
518 		tsk_set_unreturnable(tsk, true);
519 		if (sock->type == SOCK_DGRAM)
520 			tsk_set_unreliable(tsk, true);
521 	}
522 	__skb_queue_head_init(&tsk->mc_method.deferredq);
523 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
524 	return 0;
525 }
526 
527 static void tipc_sk_callback(struct rcu_head *head)
528 {
529 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
530 
531 	sock_put(&tsk->sk);
532 }
533 
534 /* Caller should hold socket lock for the socket. */
535 static void __tipc_shutdown(struct socket *sock, int error)
536 {
537 	struct sock *sk = sock->sk;
538 	struct tipc_sock *tsk = tipc_sk(sk);
539 	struct net *net = sock_net(sk);
540 	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
541 	u32 dnode = tsk_peer_node(tsk);
542 	struct sk_buff *skb;
543 
544 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
545 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
546 					    !tsk_conn_cong(tsk)));
547 
548 	/* Push out delayed messages if in Nagle mode */
549 	tipc_sk_push_backlog(tsk, false);
550 	/* Remove pending SYN */
551 	__skb_queue_purge(&sk->sk_write_queue);
552 
553 	/* Remove partially received buffer if any */
554 	skb = skb_peek(&sk->sk_receive_queue);
555 	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
556 		__skb_unlink(skb, &sk->sk_receive_queue);
557 		kfree_skb(skb);
558 	}
559 
560 	/* Reject all unreceived messages if connectionless */
561 	if (tipc_sk_type_connectionless(sk)) {
562 		tsk_rej_rx_queue(sk, error);
563 		return;
564 	}
565 
566 	switch (sk->sk_state) {
567 	case TIPC_CONNECTING:
568 	case TIPC_ESTABLISHED:
569 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
570 		tipc_node_remove_conn(net, dnode, tsk->portid);
571 		/* Send a FIN+/- to its peer */
572 		skb = __skb_dequeue(&sk->sk_receive_queue);
573 		if (skb) {
574 			__skb_queue_purge(&sk->sk_receive_queue);
575 			tipc_sk_respond(sk, skb, error);
576 			break;
577 		}
578 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
579 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
580 				      tsk_own_node(tsk), tsk_peer_port(tsk),
581 				      tsk->portid, error);
582 		if (skb)
583 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
584 		break;
585 	case TIPC_LISTEN:
586 		/* Reject all SYN messages */
587 		tsk_rej_rx_queue(sk, error);
588 		break;
589 	default:
590 		__skb_queue_purge(&sk->sk_receive_queue);
591 		break;
592 	}
593 }
594 
595 /**
596  * tipc_release - destroy a TIPC socket
597  * @sock: socket to destroy
598  *
599  * This routine cleans up any messages that are still queued on the socket.
600  * For DGRAM and RDM socket types, all queued messages are rejected.
601  * For SEQPACKET and STREAM socket types, the first message is rejected
602  * and any others are discarded.  (If the first message on a STREAM socket
603  * is partially-read, it is discarded and the next one is rejected instead.)
604  *
605  * NOTE: Rejected messages are not necessarily returned to the sender!  They
606  * are returned or discarded according to the "destination droppable" setting
607  * specified for the message by the sender.
608  *
609  * Returns 0 on success, errno otherwise
610  */
611 static int tipc_release(struct socket *sock)
612 {
613 	struct sock *sk = sock->sk;
614 	struct tipc_sock *tsk;
615 
616 	/*
617 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
618 	 * releases a pre-allocated child socket that was never used)
619 	 */
620 	if (sk == NULL)
621 		return 0;
622 
623 	tsk = tipc_sk(sk);
624 	lock_sock(sk);
625 
626 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
627 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
628 	sk->sk_shutdown = SHUTDOWN_MASK;
629 	tipc_sk_leave(tsk);
630 	tipc_sk_withdraw(tsk, 0, NULL);
631 	__skb_queue_purge(&tsk->mc_method.deferredq);
632 	sk_stop_timer(sk, &sk->sk_timer);
633 	tipc_sk_remove(tsk);
634 
635 	sock_orphan(sk);
636 	/* Reject any messages that accumulated in backlog queue */
637 	release_sock(sk);
638 	tipc_dest_list_purge(&tsk->cong_links);
639 	tsk->cong_link_cnt = 0;
640 	call_rcu(&tsk->rcu, tipc_sk_callback);
641 	sock->sk = NULL;
642 
643 	return 0;
644 }
645 
646 /**
647  * tipc_bind - associate or disassocate TIPC name(s) with a socket
648  * @sock: socket structure
649  * @uaddr: socket address describing name(s) and desired operation
650  * @uaddr_len: size of socket address data structure
651  *
652  * Name and name sequence binding is indicated using a positive scope value;
653  * a negative scope value unbinds the specified name.  Specifying no name
654  * (i.e. a socket address length of 0) unbinds all names from the socket.
655  *
656  * Returns 0 on success, errno otherwise
657  *
658  * NOTE: This routine doesn't need to take the socket lock since it doesn't
659  *       access any non-constant socket information.
660  */
661 
662 int tipc_sk_bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
663 {
664 	struct sock *sk = sock->sk;
665 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
666 	struct tipc_sock *tsk = tipc_sk(sk);
667 	int res = -EINVAL;
668 
669 	lock_sock(sk);
670 	if (unlikely(!uaddr_len)) {
671 		res = tipc_sk_withdraw(tsk, 0, NULL);
672 		goto exit;
673 	}
674 	if (tsk->group) {
675 		res = -EACCES;
676 		goto exit;
677 	}
678 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
679 		res = -EINVAL;
680 		goto exit;
681 	}
682 	if (addr->family != AF_TIPC) {
683 		res = -EAFNOSUPPORT;
684 		goto exit;
685 	}
686 
687 	if (addr->addrtype == TIPC_ADDR_NAME)
688 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
689 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
690 		res = -EAFNOSUPPORT;
691 		goto exit;
692 	}
693 
694 	res = (addr->scope >= 0) ?
695 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
696 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
697 exit:
698 	release_sock(sk);
699 	return res;
700 }
701 
702 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
703 {
704 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)skaddr;
705 
706 	if (alen) {
707 		if (alen < sizeof(struct sockaddr_tipc))
708 			return -EINVAL;
709 		if (addr->addr.nameseq.type < TIPC_RESERVED_TYPES) {
710 			pr_warn_once("Can't bind to reserved service type %u\n",
711 				     addr->addr.nameseq.type);
712 			return -EACCES;
713 		}
714 	}
715 	return tipc_sk_bind(sock, skaddr, alen);
716 }
717 
718 /**
719  * tipc_getname - get port ID of socket or peer socket
720  * @sock: socket structure
721  * @uaddr: area for returned socket address
722  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
723  *
724  * Returns 0 on success, errno otherwise
725  *
726  * NOTE: This routine doesn't need to take the socket lock since it only
727  *       accesses socket information that is unchanging (or which changes in
728  *       a completely predictable manner).
729  */
730 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
731 			int peer)
732 {
733 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
734 	struct sock *sk = sock->sk;
735 	struct tipc_sock *tsk = tipc_sk(sk);
736 
737 	memset(addr, 0, sizeof(*addr));
738 	if (peer) {
739 		if ((!tipc_sk_connected(sk)) &&
740 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
741 			return -ENOTCONN;
742 		addr->addr.id.ref = tsk_peer_port(tsk);
743 		addr->addr.id.node = tsk_peer_node(tsk);
744 	} else {
745 		addr->addr.id.ref = tsk->portid;
746 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
747 	}
748 
749 	addr->addrtype = TIPC_ADDR_ID;
750 	addr->family = AF_TIPC;
751 	addr->scope = 0;
752 	addr->addr.name.domain = 0;
753 
754 	return sizeof(*addr);
755 }
756 
757 /**
758  * tipc_poll - read and possibly block on pollmask
759  * @file: file structure associated with the socket
760  * @sock: socket for which to calculate the poll bits
761  * @wait: ???
762  *
763  * Returns pollmask value
764  *
765  * COMMENTARY:
766  * It appears that the usual socket locking mechanisms are not useful here
767  * since the pollmask info is potentially out-of-date the moment this routine
768  * exits.  TCP and other protocols seem to rely on higher level poll routines
769  * to handle any preventable race conditions, so TIPC will do the same ...
770  *
771  * IMPORTANT: The fact that a read or write operation is indicated does NOT
772  * imply that the operation will succeed, merely that it should be performed
773  * and will not block.
774  */
775 static __poll_t tipc_poll(struct file *file, struct socket *sock,
776 			      poll_table *wait)
777 {
778 	struct sock *sk = sock->sk;
779 	struct tipc_sock *tsk = tipc_sk(sk);
780 	__poll_t revents = 0;
781 
782 	sock_poll_wait(file, sock, wait);
783 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
784 
785 	if (sk->sk_shutdown & RCV_SHUTDOWN)
786 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
787 	if (sk->sk_shutdown == SHUTDOWN_MASK)
788 		revents |= EPOLLHUP;
789 
790 	switch (sk->sk_state) {
791 	case TIPC_ESTABLISHED:
792 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
793 			revents |= EPOLLOUT;
794 		fallthrough;
795 	case TIPC_LISTEN:
796 	case TIPC_CONNECTING:
797 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
798 			revents |= EPOLLIN | EPOLLRDNORM;
799 		break;
800 	case TIPC_OPEN:
801 		if (tsk->group_is_open && !tsk->cong_link_cnt)
802 			revents |= EPOLLOUT;
803 		if (!tipc_sk_type_connectionless(sk))
804 			break;
805 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
806 			break;
807 		revents |= EPOLLIN | EPOLLRDNORM;
808 		break;
809 	case TIPC_DISCONNECTING:
810 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
811 		break;
812 	}
813 	return revents;
814 }
815 
816 /**
817  * tipc_sendmcast - send multicast message
818  * @sock: socket structure
819  * @seq: destination address
820  * @msg: message to send
821  * @dlen: length of data to send
822  * @timeout: timeout to wait for wakeup
823  *
824  * Called from function tipc_sendmsg(), which has done all sanity checks
825  * Returns the number of bytes sent on success, or errno
826  */
827 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
828 			  struct msghdr *msg, size_t dlen, long timeout)
829 {
830 	struct sock *sk = sock->sk;
831 	struct tipc_sock *tsk = tipc_sk(sk);
832 	struct tipc_msg *hdr = &tsk->phdr;
833 	struct net *net = sock_net(sk);
834 	int mtu = tipc_bcast_get_mtu(net);
835 	struct tipc_mc_method *method = &tsk->mc_method;
836 	struct sk_buff_head pkts;
837 	struct tipc_nlist dsts;
838 	int rc;
839 
840 	if (tsk->group)
841 		return -EACCES;
842 
843 	/* Block or return if any destination link is congested */
844 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
845 	if (unlikely(rc))
846 		return rc;
847 
848 	/* Lookup destination nodes */
849 	tipc_nlist_init(&dsts, tipc_own_addr(net));
850 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
851 				      seq->upper, &dsts);
852 	if (!dsts.local && !dsts.remote)
853 		return -EHOSTUNREACH;
854 
855 	/* Build message header */
856 	msg_set_type(hdr, TIPC_MCAST_MSG);
857 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
858 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
859 	msg_set_destport(hdr, 0);
860 	msg_set_destnode(hdr, 0);
861 	msg_set_nametype(hdr, seq->type);
862 	msg_set_namelower(hdr, seq->lower);
863 	msg_set_nameupper(hdr, seq->upper);
864 
865 	/* Build message as chain of buffers */
866 	__skb_queue_head_init(&pkts);
867 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
868 
869 	/* Send message if build was successful */
870 	if (unlikely(rc == dlen)) {
871 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
872 					TIPC_DUMP_SK_SNDQ, " ");
873 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
874 				     &tsk->cong_link_cnt);
875 	}
876 
877 	tipc_nlist_purge(&dsts);
878 
879 	return rc ? rc : dlen;
880 }
881 
882 /**
883  * tipc_send_group_msg - send a message to a member in the group
884  * @net: network namespace
885  * @m: message to send
886  * @mb: group member
887  * @dnode: destination node
888  * @dport: destination port
889  * @dlen: total length of message data
890  */
891 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
892 			       struct msghdr *m, struct tipc_member *mb,
893 			       u32 dnode, u32 dport, int dlen)
894 {
895 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
896 	struct tipc_mc_method *method = &tsk->mc_method;
897 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
898 	struct tipc_msg *hdr = &tsk->phdr;
899 	struct sk_buff_head pkts;
900 	int mtu, rc;
901 
902 	/* Complete message header */
903 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
904 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
905 	msg_set_destport(hdr, dport);
906 	msg_set_destnode(hdr, dnode);
907 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
908 
909 	/* Build message as chain of buffers */
910 	__skb_queue_head_init(&pkts);
911 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
912 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
913 	if (unlikely(rc != dlen))
914 		return rc;
915 
916 	/* Send message */
917 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
918 	if (unlikely(rc == -ELINKCONG)) {
919 		tipc_dest_push(&tsk->cong_links, dnode, 0);
920 		tsk->cong_link_cnt++;
921 	}
922 
923 	/* Update send window */
924 	tipc_group_update_member(mb, blks);
925 
926 	/* A broadcast sent within next EXPIRE period must follow same path */
927 	method->rcast = true;
928 	method->mandatory = true;
929 	return dlen;
930 }
931 
932 /**
933  * tipc_send_group_unicast - send message to a member in the group
934  * @sock: socket structure
935  * @m: message to send
936  * @dlen: total length of message data
937  * @timeout: timeout to wait for wakeup
938  *
939  * Called from function tipc_sendmsg(), which has done all sanity checks
940  * Returns the number of bytes sent on success, or errno
941  */
942 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
943 				   int dlen, long timeout)
944 {
945 	struct sock *sk = sock->sk;
946 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
947 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
948 	struct tipc_sock *tsk = tipc_sk(sk);
949 	struct net *net = sock_net(sk);
950 	struct tipc_member *mb = NULL;
951 	u32 node, port;
952 	int rc;
953 
954 	node = dest->addr.id.node;
955 	port = dest->addr.id.ref;
956 	if (!port && !node)
957 		return -EHOSTUNREACH;
958 
959 	/* Block or return if destination link or member is congested */
960 	rc = tipc_wait_for_cond(sock, &timeout,
961 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
962 				tsk->group &&
963 				!tipc_group_cong(tsk->group, node, port, blks,
964 						 &mb));
965 	if (unlikely(rc))
966 		return rc;
967 
968 	if (unlikely(!mb))
969 		return -EHOSTUNREACH;
970 
971 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
972 
973 	return rc ? rc : dlen;
974 }
975 
976 /**
977  * tipc_send_group_anycast - send message to any member with given identity
978  * @sock: socket structure
979  * @m: message to send
980  * @dlen: total length of message data
981  * @timeout: timeout to wait for wakeup
982  *
983  * Called from function tipc_sendmsg(), which has done all sanity checks
984  * Returns the number of bytes sent on success, or errno
985  */
986 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
987 				   int dlen, long timeout)
988 {
989 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
990 	struct sock *sk = sock->sk;
991 	struct tipc_sock *tsk = tipc_sk(sk);
992 	struct list_head *cong_links = &tsk->cong_links;
993 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
994 	struct tipc_msg *hdr = &tsk->phdr;
995 	struct tipc_member *first = NULL;
996 	struct tipc_member *mbr = NULL;
997 	struct net *net = sock_net(sk);
998 	u32 node, port, exclude;
999 	struct list_head dsts;
1000 	u32 type, inst, scope;
1001 	int lookups = 0;
1002 	int dstcnt, rc;
1003 	bool cong;
1004 
1005 	INIT_LIST_HEAD(&dsts);
1006 
1007 	type = msg_nametype(hdr);
1008 	inst = dest->addr.name.name.instance;
1009 	scope = msg_lookup_scope(hdr);
1010 
1011 	while (++lookups < 4) {
1012 		exclude = tipc_group_exclude(tsk->group);
1013 
1014 		first = NULL;
1015 
1016 		/* Look for a non-congested destination member, if any */
1017 		while (1) {
1018 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1019 						 &dstcnt, exclude, false))
1020 				return -EHOSTUNREACH;
1021 			tipc_dest_pop(&dsts, &node, &port);
1022 			cong = tipc_group_cong(tsk->group, node, port, blks,
1023 					       &mbr);
1024 			if (!cong)
1025 				break;
1026 			if (mbr == first)
1027 				break;
1028 			if (!first)
1029 				first = mbr;
1030 		}
1031 
1032 		/* Start over if destination was not in member list */
1033 		if (unlikely(!mbr))
1034 			continue;
1035 
1036 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1037 			break;
1038 
1039 		/* Block or return if destination link or member is congested */
1040 		rc = tipc_wait_for_cond(sock, &timeout,
1041 					!tipc_dest_find(cong_links, node, 0) &&
1042 					tsk->group &&
1043 					!tipc_group_cong(tsk->group, node, port,
1044 							 blks, &mbr));
1045 		if (unlikely(rc))
1046 			return rc;
1047 
1048 		/* Send, unless destination disappeared while waiting */
1049 		if (likely(mbr))
1050 			break;
1051 	}
1052 
1053 	if (unlikely(lookups >= 4))
1054 		return -EHOSTUNREACH;
1055 
1056 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1057 
1058 	return rc ? rc : dlen;
1059 }
1060 
1061 /**
1062  * tipc_send_group_bcast - send message to all members in communication group
1063  * @sock: socket structure
1064  * @m: message to send
1065  * @dlen: total length of message data
1066  * @timeout: timeout to wait for wakeup
1067  *
1068  * Called from function tipc_sendmsg(), which has done all sanity checks
1069  * Returns the number of bytes sent on success, or errno
1070  */
1071 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1072 				 int dlen, long timeout)
1073 {
1074 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1075 	struct sock *sk = sock->sk;
1076 	struct net *net = sock_net(sk);
1077 	struct tipc_sock *tsk = tipc_sk(sk);
1078 	struct tipc_nlist *dsts;
1079 	struct tipc_mc_method *method = &tsk->mc_method;
1080 	bool ack = method->mandatory && method->rcast;
1081 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1082 	struct tipc_msg *hdr = &tsk->phdr;
1083 	int mtu = tipc_bcast_get_mtu(net);
1084 	struct sk_buff_head pkts;
1085 	int rc = -EHOSTUNREACH;
1086 
1087 	/* Block or return if any destination link or member is congested */
1088 	rc = tipc_wait_for_cond(sock, &timeout,
1089 				!tsk->cong_link_cnt && tsk->group &&
1090 				!tipc_group_bc_cong(tsk->group, blks));
1091 	if (unlikely(rc))
1092 		return rc;
1093 
1094 	dsts = tipc_group_dests(tsk->group);
1095 	if (!dsts->local && !dsts->remote)
1096 		return -EHOSTUNREACH;
1097 
1098 	/* Complete message header */
1099 	if (dest) {
1100 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1101 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1102 	} else {
1103 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1104 		msg_set_nameinst(hdr, 0);
1105 	}
1106 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1107 	msg_set_destport(hdr, 0);
1108 	msg_set_destnode(hdr, 0);
1109 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1110 
1111 	/* Avoid getting stuck with repeated forced replicasts */
1112 	msg_set_grp_bc_ack_req(hdr, ack);
1113 
1114 	/* Build message as chain of buffers */
1115 	__skb_queue_head_init(&pkts);
1116 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1117 	if (unlikely(rc != dlen))
1118 		return rc;
1119 
1120 	/* Send message */
1121 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1122 	if (unlikely(rc))
1123 		return rc;
1124 
1125 	/* Update broadcast sequence number and send windows */
1126 	tipc_group_update_bc_members(tsk->group, blks, ack);
1127 
1128 	/* Broadcast link is now free to choose method for next broadcast */
1129 	method->mandatory = false;
1130 	method->expires = jiffies;
1131 
1132 	return dlen;
1133 }
1134 
1135 /**
1136  * tipc_send_group_mcast - send message to all members with given identity
1137  * @sock: socket structure
1138  * @m: message to send
1139  * @dlen: total length of message data
1140  * @timeout: timeout to wait for wakeup
1141  *
1142  * Called from function tipc_sendmsg(), which has done all sanity checks
1143  * Returns the number of bytes sent on success, or errno
1144  */
1145 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1146 				 int dlen, long timeout)
1147 {
1148 	struct sock *sk = sock->sk;
1149 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1150 	struct tipc_sock *tsk = tipc_sk(sk);
1151 	struct tipc_group *grp = tsk->group;
1152 	struct tipc_msg *hdr = &tsk->phdr;
1153 	struct net *net = sock_net(sk);
1154 	u32 type, inst, scope, exclude;
1155 	struct list_head dsts;
1156 	u32 dstcnt;
1157 
1158 	INIT_LIST_HEAD(&dsts);
1159 
1160 	type = msg_nametype(hdr);
1161 	inst = dest->addr.name.name.instance;
1162 	scope = msg_lookup_scope(hdr);
1163 	exclude = tipc_group_exclude(grp);
1164 
1165 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1166 				 &dstcnt, exclude, true))
1167 		return -EHOSTUNREACH;
1168 
1169 	if (dstcnt == 1) {
1170 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1171 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1172 	}
1173 
1174 	tipc_dest_list_purge(&dsts);
1175 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1176 }
1177 
1178 /**
1179  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1180  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1181  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1182  *
1183  * Multi-threaded: parallel calls with reference to same queues may occur
1184  */
1185 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1186 		       struct sk_buff_head *inputq)
1187 {
1188 	u32 self = tipc_own_addr(net);
1189 	u32 type, lower, upper, scope;
1190 	struct sk_buff *skb, *_skb;
1191 	u32 portid, onode;
1192 	struct sk_buff_head tmpq;
1193 	struct list_head dports;
1194 	struct tipc_msg *hdr;
1195 	int user, mtyp, hlen;
1196 	bool exact;
1197 
1198 	__skb_queue_head_init(&tmpq);
1199 	INIT_LIST_HEAD(&dports);
1200 
1201 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1202 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1203 		hdr = buf_msg(skb);
1204 		user = msg_user(hdr);
1205 		mtyp = msg_type(hdr);
1206 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1207 		onode = msg_orignode(hdr);
1208 		type = msg_nametype(hdr);
1209 
1210 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1211 			spin_lock_bh(&inputq->lock);
1212 			if (skb_peek(arrvq) == skb) {
1213 				__skb_dequeue(arrvq);
1214 				__skb_queue_tail(inputq, skb);
1215 			}
1216 			kfree_skb(skb);
1217 			spin_unlock_bh(&inputq->lock);
1218 			continue;
1219 		}
1220 
1221 		/* Group messages require exact scope match */
1222 		if (msg_in_group(hdr)) {
1223 			lower = 0;
1224 			upper = ~0;
1225 			scope = msg_lookup_scope(hdr);
1226 			exact = true;
1227 		} else {
1228 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1229 			if (onode == self)
1230 				scope = TIPC_NODE_SCOPE;
1231 			else
1232 				scope = TIPC_CLUSTER_SCOPE;
1233 			exact = false;
1234 			lower = msg_namelower(hdr);
1235 			upper = msg_nameupper(hdr);
1236 		}
1237 
1238 		/* Create destination port list: */
1239 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1240 				       scope, exact, &dports);
1241 
1242 		/* Clone message per destination */
1243 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1244 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1245 			if (_skb) {
1246 				msg_set_destport(buf_msg(_skb), portid);
1247 				__skb_queue_tail(&tmpq, _skb);
1248 				continue;
1249 			}
1250 			pr_warn("Failed to clone mcast rcv buffer\n");
1251 		}
1252 		/* Append to inputq if not already done by other thread */
1253 		spin_lock_bh(&inputq->lock);
1254 		if (skb_peek(arrvq) == skb) {
1255 			skb_queue_splice_tail_init(&tmpq, inputq);
1256 			kfree_skb(__skb_dequeue(arrvq));
1257 		}
1258 		spin_unlock_bh(&inputq->lock);
1259 		__skb_queue_purge(&tmpq);
1260 		kfree_skb(skb);
1261 	}
1262 	tipc_sk_rcv(net, inputq);
1263 }
1264 
1265 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1266  *                         when socket is in Nagle mode
1267  */
1268 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1269 {
1270 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1271 	struct sk_buff *skb = skb_peek_tail(txq);
1272 	struct net *net = sock_net(&tsk->sk);
1273 	u32 dnode = tsk_peer_node(tsk);
1274 	int rc;
1275 
1276 	if (nagle_ack) {
1277 		tsk->pkt_cnt += skb_queue_len(txq);
1278 		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1279 			tsk->oneway = 0;
1280 			if (tsk->nagle_start < NAGLE_START_MAX)
1281 				tsk->nagle_start *= 2;
1282 			tsk->expect_ack = false;
1283 			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1284 				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1285 				 tsk->nagle_start);
1286 		} else {
1287 			tsk->nagle_start = NAGLE_START_INIT;
1288 			if (skb) {
1289 				msg_set_ack_required(buf_msg(skb));
1290 				tsk->expect_ack = true;
1291 			} else {
1292 				tsk->expect_ack = false;
1293 			}
1294 		}
1295 		tsk->msg_acc = 0;
1296 		tsk->pkt_cnt = 0;
1297 	}
1298 
1299 	if (!skb || tsk->cong_link_cnt)
1300 		return;
1301 
1302 	/* Do not send SYN again after congestion */
1303 	if (msg_is_syn(buf_msg(skb)))
1304 		return;
1305 
1306 	if (tsk->msg_acc)
1307 		tsk->pkt_cnt += skb_queue_len(txq);
1308 	tsk->snt_unacked += tsk->snd_backlog;
1309 	tsk->snd_backlog = 0;
1310 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1311 	if (rc == -ELINKCONG)
1312 		tsk->cong_link_cnt = 1;
1313 }
1314 
1315 /**
1316  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1317  * @tsk: receiving socket
1318  * @skb: pointer to message buffer.
1319  */
1320 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1321 				   struct sk_buff_head *inputq,
1322 				   struct sk_buff_head *xmitq)
1323 {
1324 	struct tipc_msg *hdr = buf_msg(skb);
1325 	u32 onode = tsk_own_node(tsk);
1326 	struct sock *sk = &tsk->sk;
1327 	int mtyp = msg_type(hdr);
1328 	bool was_cong;
1329 
1330 	/* Ignore if connection cannot be validated: */
1331 	if (!tsk_peer_msg(tsk, hdr)) {
1332 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1333 		goto exit;
1334 	}
1335 
1336 	if (unlikely(msg_errcode(hdr))) {
1337 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1338 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1339 				      tsk_peer_port(tsk));
1340 		sk->sk_state_change(sk);
1341 
1342 		/* State change is ignored if socket already awake,
1343 		 * - convert msg to abort msg and add to inqueue
1344 		 */
1345 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1346 		msg_set_type(hdr, TIPC_CONN_MSG);
1347 		msg_set_size(hdr, BASIC_H_SIZE);
1348 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1349 		__skb_queue_tail(inputq, skb);
1350 		return;
1351 	}
1352 
1353 	tsk->probe_unacked = false;
1354 
1355 	if (mtyp == CONN_PROBE) {
1356 		msg_set_type(hdr, CONN_PROBE_REPLY);
1357 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1358 			__skb_queue_tail(xmitq, skb);
1359 		return;
1360 	} else if (mtyp == CONN_ACK) {
1361 		was_cong = tsk_conn_cong(tsk);
1362 		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1363 		tsk->snt_unacked -= msg_conn_ack(hdr);
1364 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1365 			tsk->snd_win = msg_adv_win(hdr);
1366 		if (was_cong && !tsk_conn_cong(tsk))
1367 			sk->sk_write_space(sk);
1368 	} else if (mtyp != CONN_PROBE_REPLY) {
1369 		pr_warn("Received unknown CONN_PROTO msg\n");
1370 	}
1371 exit:
1372 	kfree_skb(skb);
1373 }
1374 
1375 /**
1376  * tipc_sendmsg - send message in connectionless manner
1377  * @sock: socket structure
1378  * @m: message to send
1379  * @dsz: amount of user data to be sent
1380  *
1381  * Message must have an destination specified explicitly.
1382  * Used for SOCK_RDM and SOCK_DGRAM messages,
1383  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1384  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1385  *
1386  * Returns the number of bytes sent on success, or errno otherwise
1387  */
1388 static int tipc_sendmsg(struct socket *sock,
1389 			struct msghdr *m, size_t dsz)
1390 {
1391 	struct sock *sk = sock->sk;
1392 	int ret;
1393 
1394 	lock_sock(sk);
1395 	ret = __tipc_sendmsg(sock, m, dsz);
1396 	release_sock(sk);
1397 
1398 	return ret;
1399 }
1400 
1401 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1402 {
1403 	struct sock *sk = sock->sk;
1404 	struct net *net = sock_net(sk);
1405 	struct tipc_sock *tsk = tipc_sk(sk);
1406 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1407 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1408 	struct list_head *clinks = &tsk->cong_links;
1409 	bool syn = !tipc_sk_type_connectionless(sk);
1410 	struct tipc_group *grp = tsk->group;
1411 	struct tipc_msg *hdr = &tsk->phdr;
1412 	struct tipc_name_seq *seq;
1413 	struct sk_buff_head pkts;
1414 	u32 dport = 0, dnode = 0;
1415 	u32 type = 0, inst = 0;
1416 	int mtu, rc;
1417 
1418 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1419 		return -EMSGSIZE;
1420 
1421 	if (likely(dest)) {
1422 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1423 			return -EINVAL;
1424 		if (unlikely(dest->family != AF_TIPC))
1425 			return -EINVAL;
1426 	}
1427 
1428 	if (grp) {
1429 		if (!dest)
1430 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1431 		if (dest->addrtype == TIPC_ADDR_NAME)
1432 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1433 		if (dest->addrtype == TIPC_ADDR_ID)
1434 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1435 		if (dest->addrtype == TIPC_ADDR_MCAST)
1436 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1437 		return -EINVAL;
1438 	}
1439 
1440 	if (unlikely(!dest)) {
1441 		dest = &tsk->peer;
1442 		if (!syn && dest->family != AF_TIPC)
1443 			return -EDESTADDRREQ;
1444 	}
1445 
1446 	if (unlikely(syn)) {
1447 		if (sk->sk_state == TIPC_LISTEN)
1448 			return -EPIPE;
1449 		if (sk->sk_state != TIPC_OPEN)
1450 			return -EISCONN;
1451 		if (tsk->published)
1452 			return -EOPNOTSUPP;
1453 		if (dest->addrtype == TIPC_ADDR_NAME) {
1454 			tsk->conn_type = dest->addr.name.name.type;
1455 			tsk->conn_instance = dest->addr.name.name.instance;
1456 		}
1457 		msg_set_syn(hdr, 1);
1458 	}
1459 
1460 	seq = &dest->addr.nameseq;
1461 	if (dest->addrtype == TIPC_ADDR_MCAST)
1462 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1463 
1464 	if (dest->addrtype == TIPC_ADDR_NAME) {
1465 		type = dest->addr.name.name.type;
1466 		inst = dest->addr.name.name.instance;
1467 		dnode = dest->addr.name.domain;
1468 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1469 		if (unlikely(!dport && !dnode))
1470 			return -EHOSTUNREACH;
1471 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1472 		dnode = dest->addr.id.node;
1473 	} else {
1474 		return -EINVAL;
1475 	}
1476 
1477 	/* Block or return if destination link is congested */
1478 	rc = tipc_wait_for_cond(sock, &timeout,
1479 				!tipc_dest_find(clinks, dnode, 0));
1480 	if (unlikely(rc))
1481 		return rc;
1482 
1483 	if (dest->addrtype == TIPC_ADDR_NAME) {
1484 		msg_set_type(hdr, TIPC_NAMED_MSG);
1485 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1486 		msg_set_nametype(hdr, type);
1487 		msg_set_nameinst(hdr, inst);
1488 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1489 		msg_set_destnode(hdr, dnode);
1490 		msg_set_destport(hdr, dport);
1491 	} else { /* TIPC_ADDR_ID */
1492 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1493 		msg_set_lookup_scope(hdr, 0);
1494 		msg_set_destnode(hdr, dnode);
1495 		msg_set_destport(hdr, dest->addr.id.ref);
1496 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1497 	}
1498 
1499 	__skb_queue_head_init(&pkts);
1500 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1501 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1502 	if (unlikely(rc != dlen))
1503 		return rc;
1504 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1505 		__skb_queue_purge(&pkts);
1506 		return -ENOMEM;
1507 	}
1508 
1509 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1510 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1511 	if (unlikely(rc == -ELINKCONG)) {
1512 		tipc_dest_push(clinks, dnode, 0);
1513 		tsk->cong_link_cnt++;
1514 		rc = 0;
1515 	}
1516 
1517 	if (unlikely(syn && !rc))
1518 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1519 
1520 	return rc ? rc : dlen;
1521 }
1522 
1523 /**
1524  * tipc_sendstream - send stream-oriented data
1525  * @sock: socket structure
1526  * @m: data to send
1527  * @dsz: total length of data to be transmitted
1528  *
1529  * Used for SOCK_STREAM data.
1530  *
1531  * Returns the number of bytes sent on success (or partial success),
1532  * or errno if no data sent
1533  */
1534 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1535 {
1536 	struct sock *sk = sock->sk;
1537 	int ret;
1538 
1539 	lock_sock(sk);
1540 	ret = __tipc_sendstream(sock, m, dsz);
1541 	release_sock(sk);
1542 
1543 	return ret;
1544 }
1545 
1546 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1547 {
1548 	struct sock *sk = sock->sk;
1549 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1550 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1551 	struct sk_buff_head *txq = &sk->sk_write_queue;
1552 	struct tipc_sock *tsk = tipc_sk(sk);
1553 	struct tipc_msg *hdr = &tsk->phdr;
1554 	struct net *net = sock_net(sk);
1555 	struct sk_buff *skb;
1556 	u32 dnode = tsk_peer_node(tsk);
1557 	int maxnagle = tsk->maxnagle;
1558 	int maxpkt = tsk->max_pkt;
1559 	int send, sent = 0;
1560 	int blocks, rc = 0;
1561 
1562 	if (unlikely(dlen > INT_MAX))
1563 		return -EMSGSIZE;
1564 
1565 	/* Handle implicit connection setup */
1566 	if (unlikely(dest)) {
1567 		rc = __tipc_sendmsg(sock, m, dlen);
1568 		if (dlen && dlen == rc) {
1569 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1570 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1571 		}
1572 		return rc;
1573 	}
1574 
1575 	do {
1576 		rc = tipc_wait_for_cond(sock, &timeout,
1577 					(!tsk->cong_link_cnt &&
1578 					 !tsk_conn_cong(tsk) &&
1579 					 tipc_sk_connected(sk)));
1580 		if (unlikely(rc))
1581 			break;
1582 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1583 		blocks = tsk->snd_backlog;
1584 		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1585 		    send <= maxnagle) {
1586 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1587 			if (unlikely(rc < 0))
1588 				break;
1589 			blocks += rc;
1590 			tsk->msg_acc++;
1591 			if (blocks <= 64 && tsk->expect_ack) {
1592 				tsk->snd_backlog = blocks;
1593 				sent += send;
1594 				break;
1595 			} else if (blocks > 64) {
1596 				tsk->pkt_cnt += skb_queue_len(txq);
1597 			} else {
1598 				skb = skb_peek_tail(txq);
1599 				if (skb) {
1600 					msg_set_ack_required(buf_msg(skb));
1601 					tsk->expect_ack = true;
1602 				} else {
1603 					tsk->expect_ack = false;
1604 				}
1605 				tsk->msg_acc = 0;
1606 				tsk->pkt_cnt = 0;
1607 			}
1608 		} else {
1609 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1610 			if (unlikely(rc != send))
1611 				break;
1612 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1613 		}
1614 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1615 					 TIPC_DUMP_SK_SNDQ, " ");
1616 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1617 		if (unlikely(rc == -ELINKCONG)) {
1618 			tsk->cong_link_cnt = 1;
1619 			rc = 0;
1620 		}
1621 		if (likely(!rc)) {
1622 			tsk->snt_unacked += blocks;
1623 			tsk->snd_backlog = 0;
1624 			sent += send;
1625 		}
1626 	} while (sent < dlen && !rc);
1627 
1628 	return sent ? sent : rc;
1629 }
1630 
1631 /**
1632  * tipc_send_packet - send a connection-oriented message
1633  * @sock: socket structure
1634  * @m: message to send
1635  * @dsz: length of data to be transmitted
1636  *
1637  * Used for SOCK_SEQPACKET messages.
1638  *
1639  * Returns the number of bytes sent on success, or errno otherwise
1640  */
1641 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1642 {
1643 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1644 		return -EMSGSIZE;
1645 
1646 	return tipc_sendstream(sock, m, dsz);
1647 }
1648 
1649 /* tipc_sk_finish_conn - complete the setup of a connection
1650  */
1651 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1652 				u32 peer_node)
1653 {
1654 	struct sock *sk = &tsk->sk;
1655 	struct net *net = sock_net(sk);
1656 	struct tipc_msg *msg = &tsk->phdr;
1657 
1658 	msg_set_syn(msg, 0);
1659 	msg_set_destnode(msg, peer_node);
1660 	msg_set_destport(msg, peer_port);
1661 	msg_set_type(msg, TIPC_CONN_MSG);
1662 	msg_set_lookup_scope(msg, 0);
1663 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1664 
1665 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1666 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1667 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1668 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1669 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1670 	tsk_set_nagle(tsk);
1671 	__skb_queue_purge(&sk->sk_write_queue);
1672 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1673 		return;
1674 
1675 	/* Fall back to message based flow control */
1676 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1677 	tsk->snd_win = FLOWCTL_MSG_WIN;
1678 }
1679 
1680 /**
1681  * tipc_sk_set_orig_addr - capture sender's address for received message
1682  * @m: descriptor for message info
1683  * @skb: received message
1684  *
1685  * Note: Address is not captured if not requested by receiver.
1686  */
1687 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1688 {
1689 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1690 	struct tipc_msg *hdr = buf_msg(skb);
1691 
1692 	if (!srcaddr)
1693 		return;
1694 
1695 	srcaddr->sock.family = AF_TIPC;
1696 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1697 	srcaddr->sock.scope = 0;
1698 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1699 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1700 	srcaddr->sock.addr.name.domain = 0;
1701 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1702 
1703 	if (!msg_in_group(hdr))
1704 		return;
1705 
1706 	/* Group message users may also want to know sending member's id */
1707 	srcaddr->member.family = AF_TIPC;
1708 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1709 	srcaddr->member.scope = 0;
1710 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1711 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1712 	srcaddr->member.addr.name.domain = 0;
1713 	m->msg_namelen = sizeof(*srcaddr);
1714 }
1715 
1716 /**
1717  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1718  * @m: descriptor for message info
1719  * @skb: received message buffer
1720  * @tsk: TIPC port associated with message
1721  *
1722  * Note: Ancillary data is not captured if not requested by receiver.
1723  *
1724  * Returns 0 if successful, otherwise errno
1725  */
1726 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1727 				 struct tipc_sock *tsk)
1728 {
1729 	struct tipc_msg *msg;
1730 	u32 anc_data[3];
1731 	u32 err;
1732 	u32 dest_type;
1733 	int has_name;
1734 	int res;
1735 
1736 	if (likely(m->msg_controllen == 0))
1737 		return 0;
1738 	msg = buf_msg(skb);
1739 
1740 	/* Optionally capture errored message object(s) */
1741 	err = msg ? msg_errcode(msg) : 0;
1742 	if (unlikely(err)) {
1743 		anc_data[0] = err;
1744 		anc_data[1] = msg_data_sz(msg);
1745 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1746 		if (res)
1747 			return res;
1748 		if (anc_data[1]) {
1749 			if (skb_linearize(skb))
1750 				return -ENOMEM;
1751 			msg = buf_msg(skb);
1752 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1753 				       msg_data(msg));
1754 			if (res)
1755 				return res;
1756 		}
1757 	}
1758 
1759 	/* Optionally capture message destination object */
1760 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1761 	switch (dest_type) {
1762 	case TIPC_NAMED_MSG:
1763 		has_name = 1;
1764 		anc_data[0] = msg_nametype(msg);
1765 		anc_data[1] = msg_namelower(msg);
1766 		anc_data[2] = msg_namelower(msg);
1767 		break;
1768 	case TIPC_MCAST_MSG:
1769 		has_name = 1;
1770 		anc_data[0] = msg_nametype(msg);
1771 		anc_data[1] = msg_namelower(msg);
1772 		anc_data[2] = msg_nameupper(msg);
1773 		break;
1774 	case TIPC_CONN_MSG:
1775 		has_name = (tsk->conn_type != 0);
1776 		anc_data[0] = tsk->conn_type;
1777 		anc_data[1] = tsk->conn_instance;
1778 		anc_data[2] = tsk->conn_instance;
1779 		break;
1780 	default:
1781 		has_name = 0;
1782 	}
1783 	if (has_name) {
1784 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1785 		if (res)
1786 			return res;
1787 	}
1788 
1789 	return 0;
1790 }
1791 
1792 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1793 {
1794 	struct sock *sk = &tsk->sk;
1795 	struct sk_buff *skb = NULL;
1796 	struct tipc_msg *msg;
1797 	u32 peer_port = tsk_peer_port(tsk);
1798 	u32 dnode = tsk_peer_node(tsk);
1799 
1800 	if (!tipc_sk_connected(sk))
1801 		return NULL;
1802 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1803 			      dnode, tsk_own_node(tsk), peer_port,
1804 			      tsk->portid, TIPC_OK);
1805 	if (!skb)
1806 		return NULL;
1807 	msg = buf_msg(skb);
1808 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1809 	tsk->rcv_unacked = 0;
1810 
1811 	/* Adjust to and advertize the correct window limit */
1812 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1813 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1814 		msg_set_adv_win(msg, tsk->rcv_win);
1815 	}
1816 	return skb;
1817 }
1818 
1819 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1820 {
1821 	struct sk_buff *skb;
1822 
1823 	skb = tipc_sk_build_ack(tsk);
1824 	if (!skb)
1825 		return;
1826 
1827 	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1828 			   msg_link_selector(buf_msg(skb)));
1829 }
1830 
1831 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1832 {
1833 	struct sock *sk = sock->sk;
1834 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1835 	long timeo = *timeop;
1836 	int err = sock_error(sk);
1837 
1838 	if (err)
1839 		return err;
1840 
1841 	for (;;) {
1842 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1843 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1844 				err = -ENOTCONN;
1845 				break;
1846 			}
1847 			add_wait_queue(sk_sleep(sk), &wait);
1848 			release_sock(sk);
1849 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1850 			sched_annotate_sleep();
1851 			lock_sock(sk);
1852 			remove_wait_queue(sk_sleep(sk), &wait);
1853 		}
1854 		err = 0;
1855 		if (!skb_queue_empty(&sk->sk_receive_queue))
1856 			break;
1857 		err = -EAGAIN;
1858 		if (!timeo)
1859 			break;
1860 		err = sock_intr_errno(timeo);
1861 		if (signal_pending(current))
1862 			break;
1863 
1864 		err = sock_error(sk);
1865 		if (err)
1866 			break;
1867 	}
1868 	*timeop = timeo;
1869 	return err;
1870 }
1871 
1872 /**
1873  * tipc_recvmsg - receive packet-oriented message
1874  * @m: descriptor for message info
1875  * @buflen: length of user buffer area
1876  * @flags: receive flags
1877  *
1878  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1879  * If the complete message doesn't fit in user area, truncate it.
1880  *
1881  * Returns size of returned message data, errno otherwise
1882  */
1883 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1884 			size_t buflen,	int flags)
1885 {
1886 	struct sock *sk = sock->sk;
1887 	bool connected = !tipc_sk_type_connectionless(sk);
1888 	struct tipc_sock *tsk = tipc_sk(sk);
1889 	int rc, err, hlen, dlen, copy;
1890 	struct sk_buff_head xmitq;
1891 	struct tipc_msg *hdr;
1892 	struct sk_buff *skb;
1893 	bool grp_evt;
1894 	long timeout;
1895 
1896 	/* Catch invalid receive requests */
1897 	if (unlikely(!buflen))
1898 		return -EINVAL;
1899 
1900 	lock_sock(sk);
1901 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1902 		rc = -ENOTCONN;
1903 		goto exit;
1904 	}
1905 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1906 
1907 	/* Step rcv queue to first msg with data or error; wait if necessary */
1908 	do {
1909 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1910 		if (unlikely(rc))
1911 			goto exit;
1912 		skb = skb_peek(&sk->sk_receive_queue);
1913 		hdr = buf_msg(skb);
1914 		dlen = msg_data_sz(hdr);
1915 		hlen = msg_hdr_sz(hdr);
1916 		err = msg_errcode(hdr);
1917 		grp_evt = msg_is_grp_evt(hdr);
1918 		if (likely(dlen || err))
1919 			break;
1920 		tsk_advance_rx_queue(sk);
1921 	} while (1);
1922 
1923 	/* Collect msg meta data, including error code and rejected data */
1924 	tipc_sk_set_orig_addr(m, skb);
1925 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1926 	if (unlikely(rc))
1927 		goto exit;
1928 	hdr = buf_msg(skb);
1929 
1930 	/* Capture data if non-error msg, otherwise just set return value */
1931 	if (likely(!err)) {
1932 		copy = min_t(int, dlen, buflen);
1933 		if (unlikely(copy != dlen))
1934 			m->msg_flags |= MSG_TRUNC;
1935 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1936 	} else {
1937 		copy = 0;
1938 		rc = 0;
1939 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1940 			rc = -ECONNRESET;
1941 	}
1942 	if (unlikely(rc))
1943 		goto exit;
1944 
1945 	/* Mark message as group event if applicable */
1946 	if (unlikely(grp_evt)) {
1947 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1948 			m->msg_flags |= MSG_EOR;
1949 		m->msg_flags |= MSG_OOB;
1950 		copy = 0;
1951 	}
1952 
1953 	/* Caption of data or error code/rejected data was successful */
1954 	if (unlikely(flags & MSG_PEEK))
1955 		goto exit;
1956 
1957 	/* Send group flow control advertisement when applicable */
1958 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1959 		__skb_queue_head_init(&xmitq);
1960 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1961 					  msg_orignode(hdr), msg_origport(hdr),
1962 					  &xmitq);
1963 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1964 	}
1965 
1966 	tsk_advance_rx_queue(sk);
1967 
1968 	if (likely(!connected))
1969 		goto exit;
1970 
1971 	/* Send connection flow control advertisement when applicable */
1972 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1973 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1974 		tipc_sk_send_ack(tsk);
1975 exit:
1976 	release_sock(sk);
1977 	return rc ? rc : copy;
1978 }
1979 
1980 /**
1981  * tipc_recvstream - receive stream-oriented data
1982  * @m: descriptor for message info
1983  * @buflen: total size of user buffer area
1984  * @flags: receive flags
1985  *
1986  * Used for SOCK_STREAM messages only.  If not enough data is available
1987  * will optionally wait for more; never truncates data.
1988  *
1989  * Returns size of returned message data, errno otherwise
1990  */
1991 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1992 			   size_t buflen, int flags)
1993 {
1994 	struct sock *sk = sock->sk;
1995 	struct tipc_sock *tsk = tipc_sk(sk);
1996 	struct sk_buff *skb;
1997 	struct tipc_msg *hdr;
1998 	struct tipc_skb_cb *skb_cb;
1999 	bool peek = flags & MSG_PEEK;
2000 	int offset, required, copy, copied = 0;
2001 	int hlen, dlen, err, rc;
2002 	long timeout;
2003 
2004 	/* Catch invalid receive attempts */
2005 	if (unlikely(!buflen))
2006 		return -EINVAL;
2007 
2008 	lock_sock(sk);
2009 
2010 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2011 		rc = -ENOTCONN;
2012 		goto exit;
2013 	}
2014 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2015 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2016 
2017 	do {
2018 		/* Look at first msg in receive queue; wait if necessary */
2019 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2020 		if (unlikely(rc))
2021 			break;
2022 		skb = skb_peek(&sk->sk_receive_queue);
2023 		skb_cb = TIPC_SKB_CB(skb);
2024 		hdr = buf_msg(skb);
2025 		dlen = msg_data_sz(hdr);
2026 		hlen = msg_hdr_sz(hdr);
2027 		err = msg_errcode(hdr);
2028 
2029 		/* Discard any empty non-errored (SYN-) message */
2030 		if (unlikely(!dlen && !err)) {
2031 			tsk_advance_rx_queue(sk);
2032 			continue;
2033 		}
2034 
2035 		/* Collect msg meta data, incl. error code and rejected data */
2036 		if (!copied) {
2037 			tipc_sk_set_orig_addr(m, skb);
2038 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2039 			if (rc)
2040 				break;
2041 			hdr = buf_msg(skb);
2042 		}
2043 
2044 		/* Copy data if msg ok, otherwise return error/partial data */
2045 		if (likely(!err)) {
2046 			offset = skb_cb->bytes_read;
2047 			copy = min_t(int, dlen - offset, buflen - copied);
2048 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2049 			if (unlikely(rc))
2050 				break;
2051 			copied += copy;
2052 			offset += copy;
2053 			if (unlikely(offset < dlen)) {
2054 				if (!peek)
2055 					skb_cb->bytes_read = offset;
2056 				break;
2057 			}
2058 		} else {
2059 			rc = 0;
2060 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2061 				rc = -ECONNRESET;
2062 			if (copied || rc)
2063 				break;
2064 		}
2065 
2066 		if (unlikely(peek))
2067 			break;
2068 
2069 		tsk_advance_rx_queue(sk);
2070 
2071 		/* Send connection flow control advertisement when applicable */
2072 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2073 		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2074 			tipc_sk_send_ack(tsk);
2075 
2076 		/* Exit if all requested data or FIN/error received */
2077 		if (copied == buflen || err)
2078 			break;
2079 
2080 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2081 exit:
2082 	release_sock(sk);
2083 	return copied ? copied : rc;
2084 }
2085 
2086 /**
2087  * tipc_write_space - wake up thread if port congestion is released
2088  * @sk: socket
2089  */
2090 static void tipc_write_space(struct sock *sk)
2091 {
2092 	struct socket_wq *wq;
2093 
2094 	rcu_read_lock();
2095 	wq = rcu_dereference(sk->sk_wq);
2096 	if (skwq_has_sleeper(wq))
2097 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2098 						EPOLLWRNORM | EPOLLWRBAND);
2099 	rcu_read_unlock();
2100 }
2101 
2102 /**
2103  * tipc_data_ready - wake up threads to indicate messages have been received
2104  * @sk: socket
2105  */
2106 static void tipc_data_ready(struct sock *sk)
2107 {
2108 	struct socket_wq *wq;
2109 
2110 	rcu_read_lock();
2111 	wq = rcu_dereference(sk->sk_wq);
2112 	if (skwq_has_sleeper(wq))
2113 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2114 						EPOLLRDNORM | EPOLLRDBAND);
2115 	rcu_read_unlock();
2116 }
2117 
2118 static void tipc_sock_destruct(struct sock *sk)
2119 {
2120 	__skb_queue_purge(&sk->sk_receive_queue);
2121 }
2122 
2123 static void tipc_sk_proto_rcv(struct sock *sk,
2124 			      struct sk_buff_head *inputq,
2125 			      struct sk_buff_head *xmitq)
2126 {
2127 	struct sk_buff *skb = __skb_dequeue(inputq);
2128 	struct tipc_sock *tsk = tipc_sk(sk);
2129 	struct tipc_msg *hdr = buf_msg(skb);
2130 	struct tipc_group *grp = tsk->group;
2131 	bool wakeup = false;
2132 
2133 	switch (msg_user(hdr)) {
2134 	case CONN_MANAGER:
2135 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2136 		return;
2137 	case SOCK_WAKEUP:
2138 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2139 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2140 		smp_wmb();
2141 		tsk->cong_link_cnt--;
2142 		wakeup = true;
2143 		tipc_sk_push_backlog(tsk, false);
2144 		break;
2145 	case GROUP_PROTOCOL:
2146 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2147 		break;
2148 	case TOP_SRV:
2149 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2150 				      hdr, inputq, xmitq);
2151 		break;
2152 	default:
2153 		break;
2154 	}
2155 
2156 	if (wakeup)
2157 		sk->sk_write_space(sk);
2158 
2159 	kfree_skb(skb);
2160 }
2161 
2162 /**
2163  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2164  * @tsk: TIPC socket
2165  * @skb: pointer to message buffer.
2166  * @xmitq: for Nagle ACK if any
2167  * Returns true if message should be added to receive queue, false otherwise
2168  */
2169 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2170 				   struct sk_buff_head *xmitq)
2171 {
2172 	struct sock *sk = &tsk->sk;
2173 	struct net *net = sock_net(sk);
2174 	struct tipc_msg *hdr = buf_msg(skb);
2175 	bool con_msg = msg_connected(hdr);
2176 	u32 pport = tsk_peer_port(tsk);
2177 	u32 pnode = tsk_peer_node(tsk);
2178 	u32 oport = msg_origport(hdr);
2179 	u32 onode = msg_orignode(hdr);
2180 	int err = msg_errcode(hdr);
2181 	unsigned long delay;
2182 
2183 	if (unlikely(msg_mcast(hdr)))
2184 		return false;
2185 	tsk->oneway = 0;
2186 
2187 	switch (sk->sk_state) {
2188 	case TIPC_CONNECTING:
2189 		/* Setup ACK */
2190 		if (likely(con_msg)) {
2191 			if (err)
2192 				break;
2193 			tipc_sk_finish_conn(tsk, oport, onode);
2194 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2195 			/* ACK+ message with data is added to receive queue */
2196 			if (msg_data_sz(hdr))
2197 				return true;
2198 			/* Empty ACK-, - wake up sleeping connect() and drop */
2199 			sk->sk_state_change(sk);
2200 			msg_set_dest_droppable(hdr, 1);
2201 			return false;
2202 		}
2203 		/* Ignore connectionless message if not from listening socket */
2204 		if (oport != pport || onode != pnode)
2205 			return false;
2206 
2207 		/* Rejected SYN */
2208 		if (err != TIPC_ERR_OVERLOAD)
2209 			break;
2210 
2211 		/* Prepare for new setup attempt if we have a SYN clone */
2212 		if (skb_queue_empty(&sk->sk_write_queue))
2213 			break;
2214 		get_random_bytes(&delay, 2);
2215 		delay %= (tsk->conn_timeout / 4);
2216 		delay = msecs_to_jiffies(delay + 100);
2217 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2218 		return false;
2219 	case TIPC_OPEN:
2220 	case TIPC_DISCONNECTING:
2221 		return false;
2222 	case TIPC_LISTEN:
2223 		/* Accept only SYN message */
2224 		if (!msg_is_syn(hdr) &&
2225 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2226 			return false;
2227 		if (!con_msg && !err)
2228 			return true;
2229 		return false;
2230 	case TIPC_ESTABLISHED:
2231 		if (!skb_queue_empty(&sk->sk_write_queue))
2232 			tipc_sk_push_backlog(tsk, false);
2233 		/* Accept only connection-based messages sent by peer */
2234 		if (likely(con_msg && !err && pport == oport &&
2235 			   pnode == onode)) {
2236 			if (msg_ack_required(hdr)) {
2237 				struct sk_buff *skb;
2238 
2239 				skb = tipc_sk_build_ack(tsk);
2240 				if (skb) {
2241 					msg_set_nagle_ack(buf_msg(skb));
2242 					__skb_queue_tail(xmitq, skb);
2243 				}
2244 			}
2245 			return true;
2246 		}
2247 		if (!tsk_peer_msg(tsk, hdr))
2248 			return false;
2249 		if (!err)
2250 			return true;
2251 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2252 		tipc_node_remove_conn(net, pnode, tsk->portid);
2253 		sk->sk_state_change(sk);
2254 		return true;
2255 	default:
2256 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2257 	}
2258 	/* Abort connection setup attempt */
2259 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2260 	sk->sk_err = ECONNREFUSED;
2261 	sk->sk_state_change(sk);
2262 	return true;
2263 }
2264 
2265 /**
2266  * rcvbuf_limit - get proper overload limit of socket receive queue
2267  * @sk: socket
2268  * @skb: message
2269  *
2270  * For connection oriented messages, irrespective of importance,
2271  * default queue limit is 2 MB.
2272  *
2273  * For connectionless messages, queue limits are based on message
2274  * importance as follows:
2275  *
2276  * TIPC_LOW_IMPORTANCE       (2 MB)
2277  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2278  * TIPC_HIGH_IMPORTANCE      (8 MB)
2279  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2280  *
2281  * Returns overload limit according to corresponding message importance
2282  */
2283 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2284 {
2285 	struct tipc_sock *tsk = tipc_sk(sk);
2286 	struct tipc_msg *hdr = buf_msg(skb);
2287 
2288 	if (unlikely(msg_in_group(hdr)))
2289 		return READ_ONCE(sk->sk_rcvbuf);
2290 
2291 	if (unlikely(!msg_connected(hdr)))
2292 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2293 
2294 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2295 		return READ_ONCE(sk->sk_rcvbuf);
2296 
2297 	return FLOWCTL_MSG_LIM;
2298 }
2299 
2300 /**
2301  * tipc_sk_filter_rcv - validate incoming message
2302  * @sk: socket
2303  * @skb: pointer to message.
2304  *
2305  * Enqueues message on receive queue if acceptable; optionally handles
2306  * disconnect indication for a connected socket.
2307  *
2308  * Called with socket lock already taken
2309  *
2310  */
2311 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2312 			       struct sk_buff_head *xmitq)
2313 {
2314 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2315 	struct tipc_sock *tsk = tipc_sk(sk);
2316 	struct tipc_group *grp = tsk->group;
2317 	struct tipc_msg *hdr = buf_msg(skb);
2318 	struct net *net = sock_net(sk);
2319 	struct sk_buff_head inputq;
2320 	int mtyp = msg_type(hdr);
2321 	int limit, err = TIPC_OK;
2322 
2323 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2324 	TIPC_SKB_CB(skb)->bytes_read = 0;
2325 	__skb_queue_head_init(&inputq);
2326 	__skb_queue_tail(&inputq, skb);
2327 
2328 	if (unlikely(!msg_isdata(hdr)))
2329 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2330 
2331 	if (unlikely(grp))
2332 		tipc_group_filter_msg(grp, &inputq, xmitq);
2333 
2334 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2335 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2336 
2337 	/* Validate and add to receive buffer if there is space */
2338 	while ((skb = __skb_dequeue(&inputq))) {
2339 		hdr = buf_msg(skb);
2340 		limit = rcvbuf_limit(sk, skb);
2341 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2342 		    (!sk_conn && msg_connected(hdr)) ||
2343 		    (!grp && msg_in_group(hdr)))
2344 			err = TIPC_ERR_NO_PORT;
2345 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2346 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2347 					   "err_overload2!");
2348 			atomic_inc(&sk->sk_drops);
2349 			err = TIPC_ERR_OVERLOAD;
2350 		}
2351 
2352 		if (unlikely(err)) {
2353 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2354 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2355 						      "@filter_rcv!");
2356 				__skb_queue_tail(xmitq, skb);
2357 			}
2358 			err = TIPC_OK;
2359 			continue;
2360 		}
2361 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2362 		skb_set_owner_r(skb, sk);
2363 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2364 					 "rcvq >90% allocated!");
2365 		sk->sk_data_ready(sk);
2366 	}
2367 }
2368 
2369 /**
2370  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2371  * @sk: socket
2372  * @skb: message
2373  *
2374  * Caller must hold socket lock
2375  */
2376 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2377 {
2378 	unsigned int before = sk_rmem_alloc_get(sk);
2379 	struct sk_buff_head xmitq;
2380 	unsigned int added;
2381 
2382 	__skb_queue_head_init(&xmitq);
2383 
2384 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2385 	added = sk_rmem_alloc_get(sk) - before;
2386 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2387 
2388 	/* Send pending response/rejected messages, if any */
2389 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2390 	return 0;
2391 }
2392 
2393 /**
2394  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2395  *                   inputq and try adding them to socket or backlog queue
2396  * @inputq: list of incoming buffers with potentially different destinations
2397  * @sk: socket where the buffers should be enqueued
2398  * @dport: port number for the socket
2399  *
2400  * Caller must hold socket lock
2401  */
2402 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2403 			    u32 dport, struct sk_buff_head *xmitq)
2404 {
2405 	unsigned long time_limit = jiffies + 2;
2406 	struct sk_buff *skb;
2407 	unsigned int lim;
2408 	atomic_t *dcnt;
2409 	u32 onode;
2410 
2411 	while (skb_queue_len(inputq)) {
2412 		if (unlikely(time_after_eq(jiffies, time_limit)))
2413 			return;
2414 
2415 		skb = tipc_skb_dequeue(inputq, dport);
2416 		if (unlikely(!skb))
2417 			return;
2418 
2419 		/* Add message directly to receive queue if possible */
2420 		if (!sock_owned_by_user(sk)) {
2421 			tipc_sk_filter_rcv(sk, skb, xmitq);
2422 			continue;
2423 		}
2424 
2425 		/* Try backlog, compensating for double-counted bytes */
2426 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2427 		if (!sk->sk_backlog.len)
2428 			atomic_set(dcnt, 0);
2429 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2430 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2431 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2432 						 "bklg & rcvq >90% allocated!");
2433 			continue;
2434 		}
2435 
2436 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2437 		/* Overload => reject message back to sender */
2438 		onode = tipc_own_addr(sock_net(sk));
2439 		atomic_inc(&sk->sk_drops);
2440 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2441 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2442 					      "@sk_enqueue!");
2443 			__skb_queue_tail(xmitq, skb);
2444 		}
2445 		break;
2446 	}
2447 }
2448 
2449 /**
2450  * tipc_sk_rcv - handle a chain of incoming buffers
2451  * @inputq: buffer list containing the buffers
2452  * Consumes all buffers in list until inputq is empty
2453  * Note: may be called in multiple threads referring to the same queue
2454  */
2455 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2456 {
2457 	struct sk_buff_head xmitq;
2458 	u32 dnode, dport = 0;
2459 	int err;
2460 	struct tipc_sock *tsk;
2461 	struct sock *sk;
2462 	struct sk_buff *skb;
2463 
2464 	__skb_queue_head_init(&xmitq);
2465 	while (skb_queue_len(inputq)) {
2466 		dport = tipc_skb_peek_port(inputq, dport);
2467 		tsk = tipc_sk_lookup(net, dport);
2468 
2469 		if (likely(tsk)) {
2470 			sk = &tsk->sk;
2471 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2472 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2473 				spin_unlock_bh(&sk->sk_lock.slock);
2474 			}
2475 			/* Send pending response/rejected messages, if any */
2476 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2477 			sock_put(sk);
2478 			continue;
2479 		}
2480 		/* No destination socket => dequeue skb if still there */
2481 		skb = tipc_skb_dequeue(inputq, dport);
2482 		if (!skb)
2483 			return;
2484 
2485 		/* Try secondary lookup if unresolved named message */
2486 		err = TIPC_ERR_NO_PORT;
2487 		if (tipc_msg_lookup_dest(net, skb, &err))
2488 			goto xmit;
2489 
2490 		/* Prepare for message rejection */
2491 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2492 			continue;
2493 
2494 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2495 xmit:
2496 		dnode = msg_destnode(buf_msg(skb));
2497 		tipc_node_xmit_skb(net, skb, dnode, dport);
2498 	}
2499 }
2500 
2501 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2502 {
2503 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2504 	struct sock *sk = sock->sk;
2505 	int done;
2506 
2507 	do {
2508 		int err = sock_error(sk);
2509 		if (err)
2510 			return err;
2511 		if (!*timeo_p)
2512 			return -ETIMEDOUT;
2513 		if (signal_pending(current))
2514 			return sock_intr_errno(*timeo_p);
2515 		if (sk->sk_state == TIPC_DISCONNECTING)
2516 			break;
2517 
2518 		add_wait_queue(sk_sleep(sk), &wait);
2519 		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2520 				     &wait);
2521 		remove_wait_queue(sk_sleep(sk), &wait);
2522 	} while (!done);
2523 	return 0;
2524 }
2525 
2526 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2527 {
2528 	if (addr->family != AF_TIPC)
2529 		return false;
2530 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2531 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2532 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2533 		addr->addrtype == TIPC_SOCKET_ADDR);
2534 }
2535 
2536 /**
2537  * tipc_connect - establish a connection to another TIPC port
2538  * @sock: socket structure
2539  * @dest: socket address for destination port
2540  * @destlen: size of socket address data structure
2541  * @flags: file-related flags associated with socket
2542  *
2543  * Returns 0 on success, errno otherwise
2544  */
2545 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2546 			int destlen, int flags)
2547 {
2548 	struct sock *sk = sock->sk;
2549 	struct tipc_sock *tsk = tipc_sk(sk);
2550 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2551 	struct msghdr m = {NULL,};
2552 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2553 	int previous;
2554 	int res = 0;
2555 
2556 	if (destlen != sizeof(struct sockaddr_tipc))
2557 		return -EINVAL;
2558 
2559 	lock_sock(sk);
2560 
2561 	if (tsk->group) {
2562 		res = -EINVAL;
2563 		goto exit;
2564 	}
2565 
2566 	if (dst->family == AF_UNSPEC) {
2567 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2568 		if (!tipc_sk_type_connectionless(sk))
2569 			res = -EINVAL;
2570 		goto exit;
2571 	}
2572 	if (!tipc_sockaddr_is_sane(dst)) {
2573 		res = -EINVAL;
2574 		goto exit;
2575 	}
2576 	/* DGRAM/RDM connect(), just save the destaddr */
2577 	if (tipc_sk_type_connectionless(sk)) {
2578 		memcpy(&tsk->peer, dest, destlen);
2579 		goto exit;
2580 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2581 		res = -EINVAL;
2582 		goto exit;
2583 	}
2584 
2585 	previous = sk->sk_state;
2586 
2587 	switch (sk->sk_state) {
2588 	case TIPC_OPEN:
2589 		/* Send a 'SYN-' to destination */
2590 		m.msg_name = dest;
2591 		m.msg_namelen = destlen;
2592 
2593 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2594 		 * indicate send_msg() is never blocked.
2595 		 */
2596 		if (!timeout)
2597 			m.msg_flags = MSG_DONTWAIT;
2598 
2599 		res = __tipc_sendmsg(sock, &m, 0);
2600 		if ((res < 0) && (res != -EWOULDBLOCK))
2601 			goto exit;
2602 
2603 		/* Just entered TIPC_CONNECTING state; the only
2604 		 * difference is that return value in non-blocking
2605 		 * case is EINPROGRESS, rather than EALREADY.
2606 		 */
2607 		res = -EINPROGRESS;
2608 		fallthrough;
2609 	case TIPC_CONNECTING:
2610 		if (!timeout) {
2611 			if (previous == TIPC_CONNECTING)
2612 				res = -EALREADY;
2613 			goto exit;
2614 		}
2615 		timeout = msecs_to_jiffies(timeout);
2616 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2617 		res = tipc_wait_for_connect(sock, &timeout);
2618 		break;
2619 	case TIPC_ESTABLISHED:
2620 		res = -EISCONN;
2621 		break;
2622 	default:
2623 		res = -EINVAL;
2624 	}
2625 
2626 exit:
2627 	release_sock(sk);
2628 	return res;
2629 }
2630 
2631 /**
2632  * tipc_listen - allow socket to listen for incoming connections
2633  * @sock: socket structure
2634  * @len: (unused)
2635  *
2636  * Returns 0 on success, errno otherwise
2637  */
2638 static int tipc_listen(struct socket *sock, int len)
2639 {
2640 	struct sock *sk = sock->sk;
2641 	int res;
2642 
2643 	lock_sock(sk);
2644 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2645 	release_sock(sk);
2646 
2647 	return res;
2648 }
2649 
2650 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2651 {
2652 	struct sock *sk = sock->sk;
2653 	DEFINE_WAIT(wait);
2654 	int err;
2655 
2656 	/* True wake-one mechanism for incoming connections: only
2657 	 * one process gets woken up, not the 'whole herd'.
2658 	 * Since we do not 'race & poll' for established sockets
2659 	 * anymore, the common case will execute the loop only once.
2660 	*/
2661 	for (;;) {
2662 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2663 					  TASK_INTERRUPTIBLE);
2664 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2665 			release_sock(sk);
2666 			timeo = schedule_timeout(timeo);
2667 			lock_sock(sk);
2668 		}
2669 		err = 0;
2670 		if (!skb_queue_empty(&sk->sk_receive_queue))
2671 			break;
2672 		err = -EAGAIN;
2673 		if (!timeo)
2674 			break;
2675 		err = sock_intr_errno(timeo);
2676 		if (signal_pending(current))
2677 			break;
2678 	}
2679 	finish_wait(sk_sleep(sk), &wait);
2680 	return err;
2681 }
2682 
2683 /**
2684  * tipc_accept - wait for connection request
2685  * @sock: listening socket
2686  * @new_sock: new socket that is to be connected
2687  * @flags: file-related flags associated with socket
2688  *
2689  * Returns 0 on success, errno otherwise
2690  */
2691 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2692 		       bool kern)
2693 {
2694 	struct sock *new_sk, *sk = sock->sk;
2695 	struct sk_buff *buf;
2696 	struct tipc_sock *new_tsock;
2697 	struct tipc_msg *msg;
2698 	long timeo;
2699 	int res;
2700 
2701 	lock_sock(sk);
2702 
2703 	if (sk->sk_state != TIPC_LISTEN) {
2704 		res = -EINVAL;
2705 		goto exit;
2706 	}
2707 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2708 	res = tipc_wait_for_accept(sock, timeo);
2709 	if (res)
2710 		goto exit;
2711 
2712 	buf = skb_peek(&sk->sk_receive_queue);
2713 
2714 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2715 	if (res)
2716 		goto exit;
2717 	security_sk_clone(sock->sk, new_sock->sk);
2718 
2719 	new_sk = new_sock->sk;
2720 	new_tsock = tipc_sk(new_sk);
2721 	msg = buf_msg(buf);
2722 
2723 	/* we lock on new_sk; but lockdep sees the lock on sk */
2724 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2725 
2726 	/*
2727 	 * Reject any stray messages received by new socket
2728 	 * before the socket lock was taken (very, very unlikely)
2729 	 */
2730 	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2731 
2732 	/* Connect new socket to it's peer */
2733 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2734 
2735 	tsk_set_importance(new_sk, msg_importance(msg));
2736 	if (msg_named(msg)) {
2737 		new_tsock->conn_type = msg_nametype(msg);
2738 		new_tsock->conn_instance = msg_nameinst(msg);
2739 	}
2740 
2741 	/*
2742 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2743 	 * Respond to 'SYN+' by queuing it on new socket.
2744 	 */
2745 	if (!msg_data_sz(msg)) {
2746 		struct msghdr m = {NULL,};
2747 
2748 		tsk_advance_rx_queue(sk);
2749 		__tipc_sendstream(new_sock, &m, 0);
2750 	} else {
2751 		__skb_dequeue(&sk->sk_receive_queue);
2752 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2753 		skb_set_owner_r(buf, new_sk);
2754 	}
2755 	release_sock(new_sk);
2756 exit:
2757 	release_sock(sk);
2758 	return res;
2759 }
2760 
2761 /**
2762  * tipc_shutdown - shutdown socket connection
2763  * @sock: socket structure
2764  * @how: direction to close (must be SHUT_RDWR)
2765  *
2766  * Terminates connection (if necessary), then purges socket's receive queue.
2767  *
2768  * Returns 0 on success, errno otherwise
2769  */
2770 static int tipc_shutdown(struct socket *sock, int how)
2771 {
2772 	struct sock *sk = sock->sk;
2773 	int res;
2774 
2775 	if (how != SHUT_RDWR)
2776 		return -EINVAL;
2777 
2778 	lock_sock(sk);
2779 
2780 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2781 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2782 	sk->sk_shutdown = SHUTDOWN_MASK;
2783 
2784 	if (sk->sk_state == TIPC_DISCONNECTING) {
2785 		/* Discard any unreceived messages */
2786 		__skb_queue_purge(&sk->sk_receive_queue);
2787 
2788 		res = 0;
2789 	} else {
2790 		res = -ENOTCONN;
2791 	}
2792 	/* Wake up anyone sleeping in poll. */
2793 	sk->sk_state_change(sk);
2794 
2795 	release_sock(sk);
2796 	return res;
2797 }
2798 
2799 static void tipc_sk_check_probing_state(struct sock *sk,
2800 					struct sk_buff_head *list)
2801 {
2802 	struct tipc_sock *tsk = tipc_sk(sk);
2803 	u32 pnode = tsk_peer_node(tsk);
2804 	u32 pport = tsk_peer_port(tsk);
2805 	u32 self = tsk_own_node(tsk);
2806 	u32 oport = tsk->portid;
2807 	struct sk_buff *skb;
2808 
2809 	if (tsk->probe_unacked) {
2810 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2811 		sk->sk_err = ECONNABORTED;
2812 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2813 		sk->sk_state_change(sk);
2814 		return;
2815 	}
2816 	/* Prepare new probe */
2817 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2818 			      pnode, self, pport, oport, TIPC_OK);
2819 	if (skb)
2820 		__skb_queue_tail(list, skb);
2821 	tsk->probe_unacked = true;
2822 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2823 }
2824 
2825 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2826 {
2827 	struct tipc_sock *tsk = tipc_sk(sk);
2828 
2829 	/* Try again later if dest link is congested */
2830 	if (tsk->cong_link_cnt) {
2831 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2832 		return;
2833 	}
2834 	/* Prepare SYN for retransmit */
2835 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2836 }
2837 
2838 static void tipc_sk_timeout(struct timer_list *t)
2839 {
2840 	struct sock *sk = from_timer(sk, t, sk_timer);
2841 	struct tipc_sock *tsk = tipc_sk(sk);
2842 	u32 pnode = tsk_peer_node(tsk);
2843 	struct sk_buff_head list;
2844 	int rc = 0;
2845 
2846 	__skb_queue_head_init(&list);
2847 	bh_lock_sock(sk);
2848 
2849 	/* Try again later if socket is busy */
2850 	if (sock_owned_by_user(sk)) {
2851 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2852 		bh_unlock_sock(sk);
2853 		sock_put(sk);
2854 		return;
2855 	}
2856 
2857 	if (sk->sk_state == TIPC_ESTABLISHED)
2858 		tipc_sk_check_probing_state(sk, &list);
2859 	else if (sk->sk_state == TIPC_CONNECTING)
2860 		tipc_sk_retry_connect(sk, &list);
2861 
2862 	bh_unlock_sock(sk);
2863 
2864 	if (!skb_queue_empty(&list))
2865 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2866 
2867 	/* SYN messages may cause link congestion */
2868 	if (rc == -ELINKCONG) {
2869 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2870 		tsk->cong_link_cnt = 1;
2871 	}
2872 	sock_put(sk);
2873 }
2874 
2875 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2876 			   struct tipc_name_seq const *seq)
2877 {
2878 	struct sock *sk = &tsk->sk;
2879 	struct net *net = sock_net(sk);
2880 	struct publication *publ;
2881 	u32 key;
2882 
2883 	if (scope != TIPC_NODE_SCOPE)
2884 		scope = TIPC_CLUSTER_SCOPE;
2885 
2886 	if (tipc_sk_connected(sk))
2887 		return -EINVAL;
2888 	key = tsk->portid + tsk->pub_count + 1;
2889 	if (key == tsk->portid)
2890 		return -EADDRINUSE;
2891 
2892 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2893 				    scope, tsk->portid, key);
2894 	if (unlikely(!publ))
2895 		return -EINVAL;
2896 
2897 	list_add(&publ->binding_sock, &tsk->publications);
2898 	tsk->pub_count++;
2899 	tsk->published = 1;
2900 	return 0;
2901 }
2902 
2903 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2904 			    struct tipc_name_seq const *seq)
2905 {
2906 	struct net *net = sock_net(&tsk->sk);
2907 	struct publication *publ;
2908 	struct publication *safe;
2909 	int rc = -EINVAL;
2910 
2911 	if (scope != TIPC_NODE_SCOPE)
2912 		scope = TIPC_CLUSTER_SCOPE;
2913 
2914 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2915 		if (seq) {
2916 			if (publ->scope != scope)
2917 				continue;
2918 			if (publ->type != seq->type)
2919 				continue;
2920 			if (publ->lower != seq->lower)
2921 				continue;
2922 			if (publ->upper != seq->upper)
2923 				break;
2924 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2925 					      publ->upper, publ->key);
2926 			rc = 0;
2927 			break;
2928 		}
2929 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2930 				      publ->upper, publ->key);
2931 		rc = 0;
2932 	}
2933 	if (list_empty(&tsk->publications))
2934 		tsk->published = 0;
2935 	return rc;
2936 }
2937 
2938 /* tipc_sk_reinit: set non-zero address in all existing sockets
2939  *                 when we go from standalone to network mode.
2940  */
2941 void tipc_sk_reinit(struct net *net)
2942 {
2943 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2944 	struct rhashtable_iter iter;
2945 	struct tipc_sock *tsk;
2946 	struct tipc_msg *msg;
2947 
2948 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2949 
2950 	do {
2951 		rhashtable_walk_start(&iter);
2952 
2953 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2954 			sock_hold(&tsk->sk);
2955 			rhashtable_walk_stop(&iter);
2956 			lock_sock(&tsk->sk);
2957 			msg = &tsk->phdr;
2958 			msg_set_prevnode(msg, tipc_own_addr(net));
2959 			msg_set_orignode(msg, tipc_own_addr(net));
2960 			release_sock(&tsk->sk);
2961 			rhashtable_walk_start(&iter);
2962 			sock_put(&tsk->sk);
2963 		}
2964 
2965 		rhashtable_walk_stop(&iter);
2966 	} while (tsk == ERR_PTR(-EAGAIN));
2967 
2968 	rhashtable_walk_exit(&iter);
2969 }
2970 
2971 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2972 {
2973 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2974 	struct tipc_sock *tsk;
2975 
2976 	rcu_read_lock();
2977 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2978 	if (tsk)
2979 		sock_hold(&tsk->sk);
2980 	rcu_read_unlock();
2981 
2982 	return tsk;
2983 }
2984 
2985 static int tipc_sk_insert(struct tipc_sock *tsk)
2986 {
2987 	struct sock *sk = &tsk->sk;
2988 	struct net *net = sock_net(sk);
2989 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2990 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2991 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2992 
2993 	while (remaining--) {
2994 		portid++;
2995 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2996 			portid = TIPC_MIN_PORT;
2997 		tsk->portid = portid;
2998 		sock_hold(&tsk->sk);
2999 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3000 						   tsk_rht_params))
3001 			return 0;
3002 		sock_put(&tsk->sk);
3003 	}
3004 
3005 	return -1;
3006 }
3007 
3008 static void tipc_sk_remove(struct tipc_sock *tsk)
3009 {
3010 	struct sock *sk = &tsk->sk;
3011 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3012 
3013 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3014 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3015 		__sock_put(sk);
3016 	}
3017 }
3018 
3019 static const struct rhashtable_params tsk_rht_params = {
3020 	.nelem_hint = 192,
3021 	.head_offset = offsetof(struct tipc_sock, node),
3022 	.key_offset = offsetof(struct tipc_sock, portid),
3023 	.key_len = sizeof(u32), /* portid */
3024 	.max_size = 1048576,
3025 	.min_size = 256,
3026 	.automatic_shrinking = true,
3027 };
3028 
3029 int tipc_sk_rht_init(struct net *net)
3030 {
3031 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3032 
3033 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3034 }
3035 
3036 void tipc_sk_rht_destroy(struct net *net)
3037 {
3038 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3039 
3040 	/* Wait for socket readers to complete */
3041 	synchronize_net();
3042 
3043 	rhashtable_destroy(&tn->sk_rht);
3044 }
3045 
3046 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3047 {
3048 	struct net *net = sock_net(&tsk->sk);
3049 	struct tipc_group *grp = tsk->group;
3050 	struct tipc_msg *hdr = &tsk->phdr;
3051 	struct tipc_name_seq seq;
3052 	int rc;
3053 
3054 	if (mreq->type < TIPC_RESERVED_TYPES)
3055 		return -EACCES;
3056 	if (mreq->scope > TIPC_NODE_SCOPE)
3057 		return -EINVAL;
3058 	if (grp)
3059 		return -EACCES;
3060 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3061 	if (!grp)
3062 		return -ENOMEM;
3063 	tsk->group = grp;
3064 	msg_set_lookup_scope(hdr, mreq->scope);
3065 	msg_set_nametype(hdr, mreq->type);
3066 	msg_set_dest_droppable(hdr, true);
3067 	seq.type = mreq->type;
3068 	seq.lower = mreq->instance;
3069 	seq.upper = seq.lower;
3070 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3071 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3072 	if (rc) {
3073 		tipc_group_delete(net, grp);
3074 		tsk->group = NULL;
3075 		return rc;
3076 	}
3077 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3078 	tsk->mc_method.rcast = true;
3079 	tsk->mc_method.mandatory = true;
3080 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3081 	return rc;
3082 }
3083 
3084 static int tipc_sk_leave(struct tipc_sock *tsk)
3085 {
3086 	struct net *net = sock_net(&tsk->sk);
3087 	struct tipc_group *grp = tsk->group;
3088 	struct tipc_name_seq seq;
3089 	int scope;
3090 
3091 	if (!grp)
3092 		return -EINVAL;
3093 	tipc_group_self(grp, &seq, &scope);
3094 	tipc_group_delete(net, grp);
3095 	tsk->group = NULL;
3096 	tipc_sk_withdraw(tsk, scope, &seq);
3097 	return 0;
3098 }
3099 
3100 /**
3101  * tipc_setsockopt - set socket option
3102  * @sock: socket structure
3103  * @lvl: option level
3104  * @opt: option identifier
3105  * @ov: pointer to new option value
3106  * @ol: length of option value
3107  *
3108  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3109  * (to ease compatibility).
3110  *
3111  * Returns 0 on success, errno otherwise
3112  */
3113 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3114 			   sockptr_t ov, unsigned int ol)
3115 {
3116 	struct sock *sk = sock->sk;
3117 	struct tipc_sock *tsk = tipc_sk(sk);
3118 	struct tipc_group_req mreq;
3119 	u32 value = 0;
3120 	int res = 0;
3121 
3122 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3123 		return 0;
3124 	if (lvl != SOL_TIPC)
3125 		return -ENOPROTOOPT;
3126 
3127 	switch (opt) {
3128 	case TIPC_IMPORTANCE:
3129 	case TIPC_SRC_DROPPABLE:
3130 	case TIPC_DEST_DROPPABLE:
3131 	case TIPC_CONN_TIMEOUT:
3132 	case TIPC_NODELAY:
3133 		if (ol < sizeof(value))
3134 			return -EINVAL;
3135 		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3136 			return -EFAULT;
3137 		break;
3138 	case TIPC_GROUP_JOIN:
3139 		if (ol < sizeof(mreq))
3140 			return -EINVAL;
3141 		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3142 			return -EFAULT;
3143 		break;
3144 	default:
3145 		if (!sockptr_is_null(ov) || ol)
3146 			return -EINVAL;
3147 	}
3148 
3149 	lock_sock(sk);
3150 
3151 	switch (opt) {
3152 	case TIPC_IMPORTANCE:
3153 		res = tsk_set_importance(sk, value);
3154 		break;
3155 	case TIPC_SRC_DROPPABLE:
3156 		if (sock->type != SOCK_STREAM)
3157 			tsk_set_unreliable(tsk, value);
3158 		else
3159 			res = -ENOPROTOOPT;
3160 		break;
3161 	case TIPC_DEST_DROPPABLE:
3162 		tsk_set_unreturnable(tsk, value);
3163 		break;
3164 	case TIPC_CONN_TIMEOUT:
3165 		tipc_sk(sk)->conn_timeout = value;
3166 		break;
3167 	case TIPC_MCAST_BROADCAST:
3168 		tsk->mc_method.rcast = false;
3169 		tsk->mc_method.mandatory = true;
3170 		break;
3171 	case TIPC_MCAST_REPLICAST:
3172 		tsk->mc_method.rcast = true;
3173 		tsk->mc_method.mandatory = true;
3174 		break;
3175 	case TIPC_GROUP_JOIN:
3176 		res = tipc_sk_join(tsk, &mreq);
3177 		break;
3178 	case TIPC_GROUP_LEAVE:
3179 		res = tipc_sk_leave(tsk);
3180 		break;
3181 	case TIPC_NODELAY:
3182 		tsk->nodelay = !!value;
3183 		tsk_set_nagle(tsk);
3184 		break;
3185 	default:
3186 		res = -EINVAL;
3187 	}
3188 
3189 	release_sock(sk);
3190 
3191 	return res;
3192 }
3193 
3194 /**
3195  * tipc_getsockopt - get socket option
3196  * @sock: socket structure
3197  * @lvl: option level
3198  * @opt: option identifier
3199  * @ov: receptacle for option value
3200  * @ol: receptacle for length of option value
3201  *
3202  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3203  * (to ease compatibility).
3204  *
3205  * Returns 0 on success, errno otherwise
3206  */
3207 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3208 			   char __user *ov, int __user *ol)
3209 {
3210 	struct sock *sk = sock->sk;
3211 	struct tipc_sock *tsk = tipc_sk(sk);
3212 	struct tipc_name_seq seq;
3213 	int len, scope;
3214 	u32 value;
3215 	int res;
3216 
3217 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3218 		return put_user(0, ol);
3219 	if (lvl != SOL_TIPC)
3220 		return -ENOPROTOOPT;
3221 	res = get_user(len, ol);
3222 	if (res)
3223 		return res;
3224 
3225 	lock_sock(sk);
3226 
3227 	switch (opt) {
3228 	case TIPC_IMPORTANCE:
3229 		value = tsk_importance(tsk);
3230 		break;
3231 	case TIPC_SRC_DROPPABLE:
3232 		value = tsk_unreliable(tsk);
3233 		break;
3234 	case TIPC_DEST_DROPPABLE:
3235 		value = tsk_unreturnable(tsk);
3236 		break;
3237 	case TIPC_CONN_TIMEOUT:
3238 		value = tsk->conn_timeout;
3239 		/* no need to set "res", since already 0 at this point */
3240 		break;
3241 	case TIPC_NODE_RECVQ_DEPTH:
3242 		value = 0; /* was tipc_queue_size, now obsolete */
3243 		break;
3244 	case TIPC_SOCK_RECVQ_DEPTH:
3245 		value = skb_queue_len(&sk->sk_receive_queue);
3246 		break;
3247 	case TIPC_SOCK_RECVQ_USED:
3248 		value = sk_rmem_alloc_get(sk);
3249 		break;
3250 	case TIPC_GROUP_JOIN:
3251 		seq.type = 0;
3252 		if (tsk->group)
3253 			tipc_group_self(tsk->group, &seq, &scope);
3254 		value = seq.type;
3255 		break;
3256 	default:
3257 		res = -EINVAL;
3258 	}
3259 
3260 	release_sock(sk);
3261 
3262 	if (res)
3263 		return res;	/* "get" failed */
3264 
3265 	if (len < sizeof(value))
3266 		return -EINVAL;
3267 
3268 	if (copy_to_user(ov, &value, sizeof(value)))
3269 		return -EFAULT;
3270 
3271 	return put_user(sizeof(value), ol);
3272 }
3273 
3274 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3275 {
3276 	struct net *net = sock_net(sock->sk);
3277 	struct tipc_sioc_nodeid_req nr = {0};
3278 	struct tipc_sioc_ln_req lnr;
3279 	void __user *argp = (void __user *)arg;
3280 
3281 	switch (cmd) {
3282 	case SIOCGETLINKNAME:
3283 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3284 			return -EFAULT;
3285 		if (!tipc_node_get_linkname(net,
3286 					    lnr.bearer_id & 0xffff, lnr.peer,
3287 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3288 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3289 				return -EFAULT;
3290 			return 0;
3291 		}
3292 		return -EADDRNOTAVAIL;
3293 	case SIOCGETNODEID:
3294 		if (copy_from_user(&nr, argp, sizeof(nr)))
3295 			return -EFAULT;
3296 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3297 			return -EADDRNOTAVAIL;
3298 		if (copy_to_user(argp, &nr, sizeof(nr)))
3299 			return -EFAULT;
3300 		return 0;
3301 	default:
3302 		return -ENOIOCTLCMD;
3303 	}
3304 }
3305 
3306 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3307 {
3308 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3309 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3310 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3311 
3312 	tsk1->peer.family = AF_TIPC;
3313 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3314 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3315 	tsk1->peer.addr.id.ref = tsk2->portid;
3316 	tsk1->peer.addr.id.node = onode;
3317 	tsk2->peer.family = AF_TIPC;
3318 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3319 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3320 	tsk2->peer.addr.id.ref = tsk1->portid;
3321 	tsk2->peer.addr.id.node = onode;
3322 
3323 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3324 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3325 	return 0;
3326 }
3327 
3328 /* Protocol switches for the various types of TIPC sockets */
3329 
3330 static const struct proto_ops msg_ops = {
3331 	.owner		= THIS_MODULE,
3332 	.family		= AF_TIPC,
3333 	.release	= tipc_release,
3334 	.bind		= tipc_bind,
3335 	.connect	= tipc_connect,
3336 	.socketpair	= tipc_socketpair,
3337 	.accept		= sock_no_accept,
3338 	.getname	= tipc_getname,
3339 	.poll		= tipc_poll,
3340 	.ioctl		= tipc_ioctl,
3341 	.listen		= sock_no_listen,
3342 	.shutdown	= tipc_shutdown,
3343 	.setsockopt	= tipc_setsockopt,
3344 	.getsockopt	= tipc_getsockopt,
3345 	.sendmsg	= tipc_sendmsg,
3346 	.recvmsg	= tipc_recvmsg,
3347 	.mmap		= sock_no_mmap,
3348 	.sendpage	= sock_no_sendpage
3349 };
3350 
3351 static const struct proto_ops packet_ops = {
3352 	.owner		= THIS_MODULE,
3353 	.family		= AF_TIPC,
3354 	.release	= tipc_release,
3355 	.bind		= tipc_bind,
3356 	.connect	= tipc_connect,
3357 	.socketpair	= tipc_socketpair,
3358 	.accept		= tipc_accept,
3359 	.getname	= tipc_getname,
3360 	.poll		= tipc_poll,
3361 	.ioctl		= tipc_ioctl,
3362 	.listen		= tipc_listen,
3363 	.shutdown	= tipc_shutdown,
3364 	.setsockopt	= tipc_setsockopt,
3365 	.getsockopt	= tipc_getsockopt,
3366 	.sendmsg	= tipc_send_packet,
3367 	.recvmsg	= tipc_recvmsg,
3368 	.mmap		= sock_no_mmap,
3369 	.sendpage	= sock_no_sendpage
3370 };
3371 
3372 static const struct proto_ops stream_ops = {
3373 	.owner		= THIS_MODULE,
3374 	.family		= AF_TIPC,
3375 	.release	= tipc_release,
3376 	.bind		= tipc_bind,
3377 	.connect	= tipc_connect,
3378 	.socketpair	= tipc_socketpair,
3379 	.accept		= tipc_accept,
3380 	.getname	= tipc_getname,
3381 	.poll		= tipc_poll,
3382 	.ioctl		= tipc_ioctl,
3383 	.listen		= tipc_listen,
3384 	.shutdown	= tipc_shutdown,
3385 	.setsockopt	= tipc_setsockopt,
3386 	.getsockopt	= tipc_getsockopt,
3387 	.sendmsg	= tipc_sendstream,
3388 	.recvmsg	= tipc_recvstream,
3389 	.mmap		= sock_no_mmap,
3390 	.sendpage	= sock_no_sendpage
3391 };
3392 
3393 static const struct net_proto_family tipc_family_ops = {
3394 	.owner		= THIS_MODULE,
3395 	.family		= AF_TIPC,
3396 	.create		= tipc_sk_create
3397 };
3398 
3399 static struct proto tipc_proto = {
3400 	.name		= "TIPC",
3401 	.owner		= THIS_MODULE,
3402 	.obj_size	= sizeof(struct tipc_sock),
3403 	.sysctl_rmem	= sysctl_tipc_rmem
3404 };
3405 
3406 /**
3407  * tipc_socket_init - initialize TIPC socket interface
3408  *
3409  * Returns 0 on success, errno otherwise
3410  */
3411 int tipc_socket_init(void)
3412 {
3413 	int res;
3414 
3415 	res = proto_register(&tipc_proto, 1);
3416 	if (res) {
3417 		pr_err("Failed to register TIPC protocol type\n");
3418 		goto out;
3419 	}
3420 
3421 	res = sock_register(&tipc_family_ops);
3422 	if (res) {
3423 		pr_err("Failed to register TIPC socket type\n");
3424 		proto_unregister(&tipc_proto);
3425 		goto out;
3426 	}
3427  out:
3428 	return res;
3429 }
3430 
3431 /**
3432  * tipc_socket_stop - stop TIPC socket interface
3433  */
3434 void tipc_socket_stop(void)
3435 {
3436 	sock_unregister(tipc_family_ops.family);
3437 	proto_unregister(&tipc_proto);
3438 }
3439 
3440 /* Caller should hold socket lock for the passed tipc socket. */
3441 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3442 {
3443 	u32 peer_node;
3444 	u32 peer_port;
3445 	struct nlattr *nest;
3446 
3447 	peer_node = tsk_peer_node(tsk);
3448 	peer_port = tsk_peer_port(tsk);
3449 
3450 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3451 	if (!nest)
3452 		return -EMSGSIZE;
3453 
3454 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3455 		goto msg_full;
3456 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3457 		goto msg_full;
3458 
3459 	if (tsk->conn_type != 0) {
3460 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3461 			goto msg_full;
3462 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3463 			goto msg_full;
3464 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3465 			goto msg_full;
3466 	}
3467 	nla_nest_end(skb, nest);
3468 
3469 	return 0;
3470 
3471 msg_full:
3472 	nla_nest_cancel(skb, nest);
3473 
3474 	return -EMSGSIZE;
3475 }
3476 
3477 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3478 			  *tsk)
3479 {
3480 	struct net *net = sock_net(skb->sk);
3481 	struct sock *sk = &tsk->sk;
3482 
3483 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3484 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3485 		return -EMSGSIZE;
3486 
3487 	if (tipc_sk_connected(sk)) {
3488 		if (__tipc_nl_add_sk_con(skb, tsk))
3489 			return -EMSGSIZE;
3490 	} else if (!list_empty(&tsk->publications)) {
3491 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3492 			return -EMSGSIZE;
3493 	}
3494 	return 0;
3495 }
3496 
3497 /* Caller should hold socket lock for the passed tipc socket. */
3498 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3499 			    struct tipc_sock *tsk)
3500 {
3501 	struct nlattr *attrs;
3502 	void *hdr;
3503 
3504 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3505 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3506 	if (!hdr)
3507 		goto msg_cancel;
3508 
3509 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3510 	if (!attrs)
3511 		goto genlmsg_cancel;
3512 
3513 	if (__tipc_nl_add_sk_info(skb, tsk))
3514 		goto attr_msg_cancel;
3515 
3516 	nla_nest_end(skb, attrs);
3517 	genlmsg_end(skb, hdr);
3518 
3519 	return 0;
3520 
3521 attr_msg_cancel:
3522 	nla_nest_cancel(skb, attrs);
3523 genlmsg_cancel:
3524 	genlmsg_cancel(skb, hdr);
3525 msg_cancel:
3526 	return -EMSGSIZE;
3527 }
3528 
3529 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3530 		    int (*skb_handler)(struct sk_buff *skb,
3531 				       struct netlink_callback *cb,
3532 				       struct tipc_sock *tsk))
3533 {
3534 	struct rhashtable_iter *iter = (void *)cb->args[4];
3535 	struct tipc_sock *tsk;
3536 	int err;
3537 
3538 	rhashtable_walk_start(iter);
3539 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3540 		if (IS_ERR(tsk)) {
3541 			err = PTR_ERR(tsk);
3542 			if (err == -EAGAIN) {
3543 				err = 0;
3544 				continue;
3545 			}
3546 			break;
3547 		}
3548 
3549 		sock_hold(&tsk->sk);
3550 		rhashtable_walk_stop(iter);
3551 		lock_sock(&tsk->sk);
3552 		err = skb_handler(skb, cb, tsk);
3553 		if (err) {
3554 			release_sock(&tsk->sk);
3555 			sock_put(&tsk->sk);
3556 			goto out;
3557 		}
3558 		release_sock(&tsk->sk);
3559 		rhashtable_walk_start(iter);
3560 		sock_put(&tsk->sk);
3561 	}
3562 	rhashtable_walk_stop(iter);
3563 out:
3564 	return skb->len;
3565 }
3566 EXPORT_SYMBOL(tipc_nl_sk_walk);
3567 
3568 int tipc_dump_start(struct netlink_callback *cb)
3569 {
3570 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3571 }
3572 EXPORT_SYMBOL(tipc_dump_start);
3573 
3574 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3575 {
3576 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3577 	struct rhashtable_iter *iter = (void *)cb->args[4];
3578 	struct tipc_net *tn = tipc_net(net);
3579 
3580 	if (!iter) {
3581 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3582 		if (!iter)
3583 			return -ENOMEM;
3584 
3585 		cb->args[4] = (long)iter;
3586 	}
3587 
3588 	rhashtable_walk_enter(&tn->sk_rht, iter);
3589 	return 0;
3590 }
3591 
3592 int tipc_dump_done(struct netlink_callback *cb)
3593 {
3594 	struct rhashtable_iter *hti = (void *)cb->args[4];
3595 
3596 	rhashtable_walk_exit(hti);
3597 	kfree(hti);
3598 	return 0;
3599 }
3600 EXPORT_SYMBOL(tipc_dump_done);
3601 
3602 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3603 			   struct tipc_sock *tsk, u32 sk_filter_state,
3604 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3605 {
3606 	struct sock *sk = &tsk->sk;
3607 	struct nlattr *attrs;
3608 	struct nlattr *stat;
3609 
3610 	/*filter response w.r.t sk_state*/
3611 	if (!(sk_filter_state & (1 << sk->sk_state)))
3612 		return 0;
3613 
3614 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3615 	if (!attrs)
3616 		goto msg_cancel;
3617 
3618 	if (__tipc_nl_add_sk_info(skb, tsk))
3619 		goto attr_msg_cancel;
3620 
3621 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3622 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3623 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3624 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3625 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3626 					 sock_i_uid(sk))) ||
3627 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3628 			      tipc_diag_gen_cookie(sk),
3629 			      TIPC_NLA_SOCK_PAD))
3630 		goto attr_msg_cancel;
3631 
3632 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3633 	if (!stat)
3634 		goto attr_msg_cancel;
3635 
3636 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3637 			skb_queue_len(&sk->sk_receive_queue)) ||
3638 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3639 			skb_queue_len(&sk->sk_write_queue)) ||
3640 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3641 			atomic_read(&sk->sk_drops)))
3642 		goto stat_msg_cancel;
3643 
3644 	if (tsk->cong_link_cnt &&
3645 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3646 		goto stat_msg_cancel;
3647 
3648 	if (tsk_conn_cong(tsk) &&
3649 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3650 		goto stat_msg_cancel;
3651 
3652 	nla_nest_end(skb, stat);
3653 
3654 	if (tsk->group)
3655 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3656 			goto stat_msg_cancel;
3657 
3658 	nla_nest_end(skb, attrs);
3659 
3660 	return 0;
3661 
3662 stat_msg_cancel:
3663 	nla_nest_cancel(skb, stat);
3664 attr_msg_cancel:
3665 	nla_nest_cancel(skb, attrs);
3666 msg_cancel:
3667 	return -EMSGSIZE;
3668 }
3669 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3670 
3671 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3672 {
3673 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3674 }
3675 
3676 /* Caller should hold socket lock for the passed tipc socket. */
3677 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3678 				 struct netlink_callback *cb,
3679 				 struct publication *publ)
3680 {
3681 	void *hdr;
3682 	struct nlattr *attrs;
3683 
3684 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3685 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3686 	if (!hdr)
3687 		goto msg_cancel;
3688 
3689 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3690 	if (!attrs)
3691 		goto genlmsg_cancel;
3692 
3693 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3694 		goto attr_msg_cancel;
3695 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3696 		goto attr_msg_cancel;
3697 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3698 		goto attr_msg_cancel;
3699 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3700 		goto attr_msg_cancel;
3701 
3702 	nla_nest_end(skb, attrs);
3703 	genlmsg_end(skb, hdr);
3704 
3705 	return 0;
3706 
3707 attr_msg_cancel:
3708 	nla_nest_cancel(skb, attrs);
3709 genlmsg_cancel:
3710 	genlmsg_cancel(skb, hdr);
3711 msg_cancel:
3712 	return -EMSGSIZE;
3713 }
3714 
3715 /* Caller should hold socket lock for the passed tipc socket. */
3716 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3717 				  struct netlink_callback *cb,
3718 				  struct tipc_sock *tsk, u32 *last_publ)
3719 {
3720 	int err;
3721 	struct publication *p;
3722 
3723 	if (*last_publ) {
3724 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3725 			if (p->key == *last_publ)
3726 				break;
3727 		}
3728 		if (p->key != *last_publ) {
3729 			/* We never set seq or call nl_dump_check_consistent()
3730 			 * this means that setting prev_seq here will cause the
3731 			 * consistence check to fail in the netlink callback
3732 			 * handler. Resulting in the last NLMSG_DONE message
3733 			 * having the NLM_F_DUMP_INTR flag set.
3734 			 */
3735 			cb->prev_seq = 1;
3736 			*last_publ = 0;
3737 			return -EPIPE;
3738 		}
3739 	} else {
3740 		p = list_first_entry(&tsk->publications, struct publication,
3741 				     binding_sock);
3742 	}
3743 
3744 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3745 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3746 		if (err) {
3747 			*last_publ = p->key;
3748 			return err;
3749 		}
3750 	}
3751 	*last_publ = 0;
3752 
3753 	return 0;
3754 }
3755 
3756 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3757 {
3758 	int err;
3759 	u32 tsk_portid = cb->args[0];
3760 	u32 last_publ = cb->args[1];
3761 	u32 done = cb->args[2];
3762 	struct net *net = sock_net(skb->sk);
3763 	struct tipc_sock *tsk;
3764 
3765 	if (!tsk_portid) {
3766 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3767 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3768 
3769 		if (!attrs[TIPC_NLA_SOCK])
3770 			return -EINVAL;
3771 
3772 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3773 						  attrs[TIPC_NLA_SOCK],
3774 						  tipc_nl_sock_policy, NULL);
3775 		if (err)
3776 			return err;
3777 
3778 		if (!sock[TIPC_NLA_SOCK_REF])
3779 			return -EINVAL;
3780 
3781 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3782 	}
3783 
3784 	if (done)
3785 		return 0;
3786 
3787 	tsk = tipc_sk_lookup(net, tsk_portid);
3788 	if (!tsk)
3789 		return -EINVAL;
3790 
3791 	lock_sock(&tsk->sk);
3792 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3793 	if (!err)
3794 		done = 1;
3795 	release_sock(&tsk->sk);
3796 	sock_put(&tsk->sk);
3797 
3798 	cb->args[0] = tsk_portid;
3799 	cb->args[1] = last_publ;
3800 	cb->args[2] = done;
3801 
3802 	return skb->len;
3803 }
3804 
3805 /**
3806  * tipc_sk_filtering - check if a socket should be traced
3807  * @sk: the socket to be examined
3808  * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3809  *  (portid, sock type, name type, name lower, name upper)
3810  *
3811  * Returns true if the socket meets the socket tuple data
3812  * (value 0 = 'any') or when there is no tuple set (all = 0),
3813  * otherwise false
3814  */
3815 bool tipc_sk_filtering(struct sock *sk)
3816 {
3817 	struct tipc_sock *tsk;
3818 	struct publication *p;
3819 	u32 _port, _sktype, _type, _lower, _upper;
3820 	u32 type = 0, lower = 0, upper = 0;
3821 
3822 	if (!sk)
3823 		return true;
3824 
3825 	tsk = tipc_sk(sk);
3826 
3827 	_port = sysctl_tipc_sk_filter[0];
3828 	_sktype = sysctl_tipc_sk_filter[1];
3829 	_type = sysctl_tipc_sk_filter[2];
3830 	_lower = sysctl_tipc_sk_filter[3];
3831 	_upper = sysctl_tipc_sk_filter[4];
3832 
3833 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3834 		return true;
3835 
3836 	if (_port)
3837 		return (_port == tsk->portid);
3838 
3839 	if (_sktype && _sktype != sk->sk_type)
3840 		return false;
3841 
3842 	if (tsk->published) {
3843 		p = list_first_entry_or_null(&tsk->publications,
3844 					     struct publication, binding_sock);
3845 		if (p) {
3846 			type = p->type;
3847 			lower = p->lower;
3848 			upper = p->upper;
3849 		}
3850 	}
3851 
3852 	if (!tipc_sk_type_connectionless(sk)) {
3853 		type = tsk->conn_type;
3854 		lower = tsk->conn_instance;
3855 		upper = tsk->conn_instance;
3856 	}
3857 
3858 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3859 	    (_upper && _upper != upper))
3860 		return false;
3861 
3862 	return true;
3863 }
3864 
3865 u32 tipc_sock_get_portid(struct sock *sk)
3866 {
3867 	return (sk) ? (tipc_sk(sk))->portid : 0;
3868 }
3869 
3870 /**
3871  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3872  *			both the rcv and backlog queues are considered
3873  * @sk: tipc sk to be checked
3874  * @skb: tipc msg to be checked
3875  *
3876  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3877  */
3878 
3879 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3880 {
3881 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3882 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3883 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3884 
3885 	return (qsize > lim * 90 / 100);
3886 }
3887 
3888 /**
3889  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3890  *			only the rcv queue is considered
3891  * @sk: tipc sk to be checked
3892  * @skb: tipc msg to be checked
3893  *
3894  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3895  */
3896 
3897 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3898 {
3899 	unsigned int lim = rcvbuf_limit(sk, skb);
3900 	unsigned int qsize = sk_rmem_alloc_get(sk);
3901 
3902 	return (qsize > lim * 90 / 100);
3903 }
3904 
3905 /**
3906  * tipc_sk_dump - dump TIPC socket
3907  * @sk: tipc sk to be dumped
3908  * @dqueues: bitmask to decide if any socket queue to be dumped?
3909  *           - TIPC_DUMP_NONE: don't dump socket queues
3910  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3911  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3912  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3913  *           - TIPC_DUMP_ALL: dump all the socket queues above
3914  * @buf: returned buffer of dump data in format
3915  */
3916 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3917 {
3918 	int i = 0;
3919 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3920 	struct tipc_sock *tsk;
3921 	struct publication *p;
3922 	bool tsk_connected;
3923 
3924 	if (!sk) {
3925 		i += scnprintf(buf, sz, "sk data: (null)\n");
3926 		return i;
3927 	}
3928 
3929 	tsk = tipc_sk(sk);
3930 	tsk_connected = !tipc_sk_type_connectionless(sk);
3931 
3932 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3933 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3934 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3935 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3936 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3937 	if (tsk_connected) {
3938 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3939 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3940 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3941 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3942 	}
3943 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3944 	if (tsk->published) {
3945 		p = list_first_entry_or_null(&tsk->publications,
3946 					     struct publication, binding_sock);
3947 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3948 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3949 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3950 	}
3951 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3952 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3953 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3954 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3955 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3956 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3957 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3958 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3959 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3960 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3961 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3962 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3963 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3964 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3965 
3966 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3967 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3968 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3969 	}
3970 
3971 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3972 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3973 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3974 	}
3975 
3976 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3977 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3978 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3979 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3980 			i += scnprintf(buf + i, sz - i, "  tail ");
3981 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3982 					   buf + i);
3983 		}
3984 	}
3985 
3986 	return i;
3987 }
3988